diff options
44 files changed, 11894 insertions, 12116 deletions
diff --git a/libX11/modules/im/ximcp/imLcFlt.c b/libX11/modules/im/ximcp/imLcFlt.c index 014ad6add..7ab56f8a1 100644 --- a/libX11/modules/im/ximcp/imLcFlt.c +++ b/libX11/modules/im/ximcp/imLcFlt.c @@ -45,6 +45,7 @@ _XimLocalFilter(Display *d, Window w, XEvent *ev, XPointer client_data) static char buf[256];
DefTree *b = ic->private.local.base.tree;
DTIndex t;
+ Bool braille = False;
if(ev->xkey.keycode == 0)
return (False);
@@ -58,6 +59,7 @@ _XimLocalFilter(Display *d, Window w, XEvent *ev, XPointer client_data) if(ev->type == KeyPress) {
ic->private.local.brl_pressed |=
1<<(keysym-XK_braille_dot_1);
+ return(True);
} else {
if(!ic->private.local.brl_committing
|| ev->xkey.time - ic->private.local.brl_release_start > 300) {
@@ -65,23 +67,20 @@ _XimLocalFilter(Display *d, Window w, XEvent *ev, XPointer client_data) ic->private.local.brl_release_start = ev->xkey.time;
}
ic->private.local.brl_pressed &= ~(1<<(keysym-XK_braille_dot_1));
- if(!ic->private.local.brl_pressed) {
- if(ic->private.local.brl_committing) {
- ic->private.local.brl_committed =
- ic->private.local.brl_committing;
- ic->private.local.composed = 0;
- ev->type = KeyPress;
- ev->xkey.keycode = 0;
- _XPutBackEvent(d, ev);
- }
+ if(!ic->private.local.brl_pressed && ic->private.local.brl_committing) {
+ /* Commited a braille pattern, let it go through compose tree */
+ keysym = XK_braille_blank | ic->private.local.brl_committing;
+ ev->type = KeyPress;
+ braille = True;
+ } else {
+ return(True);
}
}
- return(True);
}
if( (ev->type != KeyPress)
|| (((Xim)ic->core.im)->private.local.top == 0 ) )
- return(False);
+ goto emit_braille;
for(t = ic->private.local.context; t; t = b[t].next) {
if(((ev->xkey.state & b[t].modifier_mask) == b[t].modifier) &&
@@ -105,11 +104,22 @@ _XimLocalFilter(Display *d, Window w, XEvent *ev, XPointer client_data) }
} else { /* Unmatched */
if(ic->private.local.context == ((Xim)ic->core.im)->private.local.top) {
- return(False);
+ goto emit_braille;
}
/* Error (Sequence Unmatch occured) */
/* initialize internal state for next key sequence */
ic->private.local.context = ((Xim)ic->core.im)->private.local.top;
return(True);
}
+
+emit_braille:
+ if(braille) {
+ /* Braille pattern is not in compose tree, emit alone */
+ ic->private.local.brl_committed = ic->private.local.brl_committing;
+ ic->private.local.composed = 0;
+ ev->xkey.keycode = 0;
+ _XPutBackEvent(d, ev);
+ return(True);
+ }
+ return(False);
}
diff --git a/libX11/modules/im/ximcp/imLcIc.c b/libX11/modules/im/ximcp/imLcIc.c index 53d53efec..04de3c1bd 100644 --- a/libX11/modules/im/ximcp/imLcIc.c +++ b/libX11/modules/im/ximcp/imLcIc.c @@ -1,200 +1,200 @@ -/****************************************************************** - - Copyright 1992,1993, 1994 by FUJITSU LIMITED - -Permission to use, copy, modify, distribute, and sell this software -and its documentation for any purpose is hereby granted without fee, -provided that the above copyright notice appear in all copies and -that both that copyright notice and this permission notice appear -in supporting documentation, and that the name of FUJITSU LIMITED -not be used in advertising or publicity pertaining to distribution -of the software without specific, written prior permission. -FUJITSU LIMITED makes no representations about the suitability of -this software for any purpose. -It is provided "as is" without express or implied warranty. - -FUJITSU LIMITED DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, -INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO -EVENT SHALL FUJITSU LIMITED BE LIABLE FOR ANY SPECIAL, INDIRECT OR -CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF -USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR -OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR -PERFORMANCE OF THIS SOFTWARE. - - Author: Takashi Fujiwara FUJITSU LIMITED - fujiwara@a80.tech.yk.fujitsu.co.jp - -******************************************************************/ - -#ifdef HAVE_CONFIG_H -#include <config.h> -#endif -#include <stdio.h> -#include <X11/Xlib.h> -#include <X11/Xmd.h> -#include "Xlibint.h" -#include "Xlcint.h" -#include "Ximint.h" - -Private void -_XimLocalUnSetFocus( - XIC xic) -{ - Xic ic = (Xic)xic; - ((Xim)ic->core.im)->private.local.current_ic = (XIC)NULL; - - if (ic->core.focus_window) - _XUnregisterFilter(ic->core.im->core.display, - ic->core.focus_window, _XimLocalFilter, (XPointer)ic); - return; -} - -Private void -_XimLocalDestroyIC( - XIC xic) -{ - Xic ic = (Xic)xic; - - if(((Xim)ic->core.im)->private.local.current_ic == (XIC)ic) { - ((Xim)ic->core.im)->private.local.current_ic = (XIC)NULL; - } - if (ic->core.focus_window) - _XUnregisterFilter(ic->core.im->core.display, - ic->core.focus_window, _XimLocalFilter, (XPointer)ic); - if(ic->private.local.ic_resources) { - Xfree(ic->private.local.ic_resources); - ic->private.local.ic_resources = NULL; - } - return; -} - -Private void -_XimLocalSetFocus( - XIC xic) -{ - Xic ic = (Xic)xic; - XIC current_ic = ((Xim)ic->core.im)->private.local.current_ic; - - if (current_ic == (XIC)ic) - return; - - if (current_ic != (XIC)NULL) { - _XimLocalUnSetFocus(current_ic); - } - ((Xim)ic->core.im)->private.local.current_ic = (XIC)ic; - - if (ic->core.focus_window) - _XRegisterFilterByType(ic->core.im->core.display, - ic->core.focus_window, KeyPress, KeyRelease, - _XimLocalFilter, (XPointer)ic); - return; -} - -Private void -_XimLocalReset( - XIC xic) -{ - Xic ic = (Xic)xic; - ic->private.local.composed = 0; - ic->private.local.context = ((Xim)ic->core.im)->private.local.top; - ic->private.local.brl_pressed = 0; - ic->private.local.brl_committing = 0; - ic->private.local.brl_committed = 0; -} - -Private char * -_XimLocalMbReset( - XIC xic) -{ - _XimLocalReset(xic); - return (char *)NULL; -} - -Private wchar_t * -_XimLocalWcReset( - XIC xic) -{ - _XimLocalReset(xic); - return (wchar_t *)NULL; -} - -Private XICMethodsRec Local_ic_methods = { - _XimLocalDestroyIC, /* destroy */ - _XimLocalSetFocus, /* set_focus */ - _XimLocalUnSetFocus, /* unset_focus */ - _XimLocalSetICValues, /* set_values */ - _XimLocalGetICValues, /* get_values */ - _XimLocalMbReset, /* mb_reset */ - _XimLocalWcReset, /* wc_reset */ - _XimLocalMbReset, /* utf8_reset */ - _XimLocalMbLookupString, /* mb_lookup_string */ - _XimLocalWcLookupString, /* wc_lookup_string */ - _XimLocalUtf8LookupString /* utf8_lookup_string */ -}; - -Public XIC -_XimLocalCreateIC( - XIM im, - XIMArg *values) -{ - Xic ic; - XimDefICValues ic_values; - XIMResourceList res; - unsigned int num; - int len; - - if((ic = (Xic)Xmalloc(sizeof(XicRec))) == (Xic)NULL) { - return ((XIC)NULL); - } - bzero((char *)ic, sizeof(XicRec)); - - ic->methods = &Local_ic_methods; - ic->core.im = im; - ic->private.local.base = ((Xim)im)->private.local.base; - ic->private.local.context = ((Xim)im)->private.local.top; - ic->private.local.composed = 0; - ic->private.local.brl_pressed = 0; - ic->private.local.brl_committing = 0; - ic->private.local.brl_committed = 0; - - num = im->core.ic_num_resources; - len = sizeof(XIMResource) * num; - if((res = (XIMResourceList)Xmalloc(len)) == (XIMResourceList)NULL) { - goto Set_Error; - } - (void)memcpy((char *)res, (char *)im->core.ic_resources, len); - ic->private.local.ic_resources = res; - ic->private.local.ic_num_resources = num; - - bzero((char *)&ic_values, sizeof(XimDefICValues)); - if(_XimCheckLocalInputStyle(ic, (XPointer)&ic_values, values, - im->core.styles, res, num) == False) { - goto Set_Error; - } - - _XimSetICMode(res, num, ic_values.input_style); - - if(_XimSetICValueData(ic, (XPointer)&ic_values, - ic->private.local.ic_resources, - ic->private.local.ic_num_resources, - values, XIM_CREATEIC, True)) { - goto Set_Error; - } - ic_values.filter_events = KeyPressMask; - _XimSetCurrentICValues(ic, &ic_values); - if(_XimSetICDefaults(ic, (XPointer)&ic_values, - XIM_SETICDEFAULTS, res, num) == False) { - goto Set_Error; - } - _XimSetCurrentICValues(ic, &ic_values); - - return((XIC)ic); - -Set_Error : - if (ic->private.local.ic_resources) { - Xfree(ic->private.local.ic_resources); - ic->private.local.ic_resources = NULL; - } - Xfree(ic); - return((XIC)NULL); -} +/******************************************************************
+
+ Copyright 1992,1993, 1994 by FUJITSU LIMITED
+
+Permission to use, copy, modify, distribute, and sell this software
+and its documentation for any purpose is hereby granted without fee,
+provided that the above copyright notice appear in all copies and
+that both that copyright notice and this permission notice appear
+in supporting documentation, and that the name of FUJITSU LIMITED
+not be used in advertising or publicity pertaining to distribution
+of the software without specific, written prior permission.
+FUJITSU LIMITED makes no representations about the suitability of
+this software for any purpose.
+It is provided "as is" without express or implied warranty.
+
+FUJITSU LIMITED DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE,
+INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO
+EVENT SHALL FUJITSU LIMITED BE LIABLE FOR ANY SPECIAL, INDIRECT OR
+CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF
+USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR
+OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR
+PERFORMANCE OF THIS SOFTWARE.
+
+ Author: Takashi Fujiwara FUJITSU LIMITED
+ fujiwara@a80.tech.yk.fujitsu.co.jp
+
+******************************************************************/
+
+#ifdef HAVE_CONFIG_H
+#include <config.h>
+#endif
+#include <stdio.h>
+#include <X11/Xlib.h>
+#include <X11/Xmd.h>
+#include "Xlibint.h"
+#include "Xlcint.h"
+#include "Ximint.h"
+
+Private void
+_XimLocalUnSetFocus(
+ XIC xic)
+{
+ Xic ic = (Xic)xic;
+ ((Xim)ic->core.im)->private.local.current_ic = (XIC)NULL;
+
+ if (ic->core.focus_window)
+ _XUnregisterFilter(ic->core.im->core.display,
+ ic->core.focus_window, _XimLocalFilter, (XPointer)ic);
+ return;
+}
+
+Private void
+_XimLocalDestroyIC(
+ XIC xic)
+{
+ Xic ic = (Xic)xic;
+
+ if(((Xim)ic->core.im)->private.local.current_ic == (XIC)ic) {
+ ((Xim)ic->core.im)->private.local.current_ic = (XIC)NULL;
+ }
+ if (ic->core.focus_window)
+ _XUnregisterFilter(ic->core.im->core.display,
+ ic->core.focus_window, _XimLocalFilter, (XPointer)ic);
+ if(ic->private.local.ic_resources) {
+ Xfree(ic->private.local.ic_resources);
+ ic->private.local.ic_resources = NULL;
+ }
+ return;
+}
+
+Private void
+_XimLocalSetFocus(
+ XIC xic)
+{
+ Xic ic = (Xic)xic;
+ XIC current_ic = ((Xim)ic->core.im)->private.local.current_ic;
+
+ if (current_ic == (XIC)ic)
+ return;
+
+ if (current_ic != (XIC)NULL) {
+ _XimLocalUnSetFocus(current_ic);
+ }
+ ((Xim)ic->core.im)->private.local.current_ic = (XIC)ic;
+
+ if (ic->core.focus_window)
+ _XRegisterFilterByType(ic->core.im->core.display,
+ ic->core.focus_window, KeyPress, KeyRelease,
+ _XimLocalFilter, (XPointer)ic);
+ return;
+}
+
+Private void
+_XimLocalReset(
+ XIC xic)
+{
+ Xic ic = (Xic)xic;
+ ic->private.local.composed = 0;
+ ic->private.local.context = ((Xim)ic->core.im)->private.local.top;
+ ic->private.local.brl_pressed = 0;
+ ic->private.local.brl_committing = 0;
+ ic->private.local.brl_committed = 0;
+}
+
+Private char *
+_XimLocalMbReset(
+ XIC xic)
+{
+ _XimLocalReset(xic);
+ return (char *)NULL;
+}
+
+Private wchar_t *
+_XimLocalWcReset(
+ XIC xic)
+{
+ _XimLocalReset(xic);
+ return (wchar_t *)NULL;
+}
+
+Private XICMethodsRec Local_ic_methods = {
+ _XimLocalDestroyIC, /* destroy */
+ _XimLocalSetFocus, /* set_focus */
+ _XimLocalUnSetFocus, /* unset_focus */
+ _XimLocalSetICValues, /* set_values */
+ _XimLocalGetICValues, /* get_values */
+ _XimLocalMbReset, /* mb_reset */
+ _XimLocalWcReset, /* wc_reset */
+ _XimLocalMbReset, /* utf8_reset */
+ _XimLocalMbLookupString, /* mb_lookup_string */
+ _XimLocalWcLookupString, /* wc_lookup_string */
+ _XimLocalUtf8LookupString /* utf8_lookup_string */
+};
+
+Public XIC
+_XimLocalCreateIC(
+ XIM im,
+ XIMArg *values)
+{
+ Xic ic;
+ XimDefICValues ic_values;
+ XIMResourceList res;
+ unsigned int num;
+ int len;
+
+ if((ic = (Xic)Xmalloc(sizeof(XicRec))) == (Xic)NULL) {
+ return ((XIC)NULL);
+ }
+ bzero((char *)ic, sizeof(XicRec));
+
+ ic->methods = &Local_ic_methods;
+ ic->core.im = im;
+ ic->private.local.base = ((Xim)im)->private.local.base;
+ ic->private.local.context = ((Xim)im)->private.local.top;
+ ic->private.local.composed = 0;
+ ic->private.local.brl_pressed = 0;
+ ic->private.local.brl_committing = 0;
+ ic->private.local.brl_committed = 0;
+
+ num = im->core.ic_num_resources;
+ len = sizeof(XIMResource) * num;
+ if((res = (XIMResourceList)Xmalloc(len)) == (XIMResourceList)NULL) {
+ goto Set_Error;
+ }
+ (void)memcpy((char *)res, (char *)im->core.ic_resources, len);
+ ic->private.local.ic_resources = res;
+ ic->private.local.ic_num_resources = num;
+
+ bzero((char *)&ic_values, sizeof(XimDefICValues));
+ if(_XimCheckLocalInputStyle(ic, (XPointer)&ic_values, values,
+ im->core.styles, res, num) == False) {
+ goto Set_Error;
+ }
+
+ _XimSetICMode(res, num, ic_values.input_style);
+
+ if(_XimSetICValueData(ic, (XPointer)&ic_values,
+ ic->private.local.ic_resources,
+ ic->private.local.ic_num_resources,
+ values, XIM_CREATEIC, True)) {
+ goto Set_Error;
+ }
+ ic_values.filter_events = KeyPressMask | KeyReleaseMask;
+ _XimSetCurrentICValues(ic, &ic_values);
+ if(_XimSetICDefaults(ic, (XPointer)&ic_values,
+ XIM_SETICDEFAULTS, res, num) == False) {
+ goto Set_Error;
+ }
+ _XimSetCurrentICValues(ic, &ic_values);
+
+ return((XIC)ic);
+
+Set_Error :
+ if (ic->private.local.ic_resources) {
+ Xfree(ic->private.local.ic_resources);
+ ic->private.local.ic_resources = NULL;
+ }
+ Xfree(ic);
+ return((XIC)NULL);
+}
diff --git a/libX11/modules/im/ximcp/imLcLkup.c b/libX11/modules/im/ximcp/imLcLkup.c index 813efcd25..c62a9106c 100644 --- a/libX11/modules/im/ximcp/imLcLkup.c +++ b/libX11/modules/im/ximcp/imLcLkup.c @@ -1,407 +1,417 @@ -/****************************************************************** - - Copyright 1992 by Fuji Xerox Co., Ltd. - Copyright 1992, 1994 by FUJITSU LIMITED - -Permission to use, copy, modify, distribute, and sell this software -and its documentation for any purpose is hereby granted without fee, -provided that the above copyright notice appear in all copies and -that both that copyright notice and this permission notice appear -in supporting documentation, and that the name of Fuji Xerox, -FUJITSU LIMITED not be used in advertising or publicity pertaining -to distribution of the software without specific, written prior -permission. Fuji Xerox, FUJITSU LIMITED make no representations -about the suitability of this software for any purpose. -It is provided "as is" without express or implied warranty. - -FUJI XEROX, FUJITSU LIMITED DISCLAIM ALL WARRANTIES WITH -REGARD TO THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF -MERCHANTABILITY AND FITNESS, IN NO EVENT SHALL FUJI XEROX, -FUJITSU LIMITED BE LIABLE FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL -DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA -OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER -TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR -PERFORMANCE OF THIS SOFTWARE. - - Author: Kazunori Nishihara Fuji Xerox - Takashi Fujiwara FUJITSU LIMITED - fujiwara@a80.tech.yk.fujitsu.co.jp - -******************************************************************/ - -#ifdef HAVE_CONFIG_H -#include <config.h> -#endif -#include <stdio.h> -#include <X11/Xatom.h> -#include <X11/Xos.h> -#include <X11/Xlib.h> -#include <X11/keysym.h> -#include <X11/Xutil.h> -#include "Xlibint.h" -#include "Xlcint.h" -#include "XlcPubI.h" -#include "Ximint.h" - -Public int -_XimLocalMbLookupString(XIC xic, XKeyEvent *ev, char *buffer, int bytes, - KeySym *keysym, Status *status) -{ - Xic ic = (Xic)xic; - int ret; - DefTree *b = ic->private.local.base.tree; - char *mb = ic->private.local.base.mb; - - if(ev->type != KeyPress) { - if(status) *status = XLookupNone; - return(0); - } - if(ev->keycode == 0 && - ( (ic->private.local.composed != 0) - ||(ic->private.local.brl_committed != 0))) { - if (ic->private.local.brl_committed != 0) { /* Braille Event */ - unsigned char pattern = ic->private.local.brl_committed; - char *mb=alloca(XLC_PUBLIC(ic->core.im->core.lcd, mb_cur_max)); - ret = _Xlcwctomb(ic->core.im->core.lcd, mb, BRL_UC_ROW | pattern); - if(ret < 0) { - if(status) *status = XLookupNone; - return(0); - } - if(ret > bytes) { - if(status) *status = XBufferOverflow; - return(ret); - } - if(keysym) { - *keysym = XK_braille_blank | pattern; - if(status) *status = XLookupBoth; - } else - if(status) *status = XLookupChars; - memcpy(buffer, mb, ret); - } else { /* Composed Event */ - ret = strlen(&mb[b[ic->private.local.composed].mb]); - if(ret > bytes) { - if(status) *status = XBufferOverflow; - return(ret); - } - memcpy(buffer, &mb[b[ic->private.local.composed].mb], ret); - if(keysym) *keysym = b[ic->private.local.composed].ks; - if (ret > 0) { - if (keysym && *keysym != NoSymbol) { - if(status) *status = XLookupBoth; - } else { - if(status) *status = XLookupChars; - } - } else { - if(keysym && *keysym != NoSymbol) { - if(status) *status = XLookupKeySym; - } else { - if(status) *status = XLookupNone; - } - } - } - return (ret); - } else { /* Throughed Event */ - ret = _XimLookupMBText(ic, ev, buffer, bytes, keysym, NULL); - if(ret > 0) { - if (ret > bytes) { - if (status) *status = XBufferOverflow; - } else if (keysym && *keysym != NoSymbol) { - if(status) *status = XLookupBoth; - } else { - if(status) *status = XLookupChars; - } - } else { - if(keysym && *keysym != NoSymbol) { - if(status) *status = XLookupKeySym; - } else { - if(status) *status = XLookupNone; - } - } - } - return (ret); -} - -Public int -_XimLocalWcLookupString(XIC xic, XKeyEvent *ev, wchar_t *buffer, int wlen, - KeySym *keysym, Status *status) -{ - Xic ic = (Xic)xic; - int ret; - DefTree *b = ic->private.local.base.tree; - wchar_t *wc = ic->private.local.base.wc; - - if(ev->type != KeyPress) { - if(status) *status = XLookupNone; - return(0); - } - if(ev->keycode == 0) { - if (ic->private.local.brl_committed != 0) { /* Braille Event */ - unsigned char pattern = ic->private.local.brl_committed; - ret = 1; - if (ret > wlen) { - if(status) *status = XBufferOverflow; - return (ret); - } - *buffer = BRL_UC_ROW | pattern; - if(keysym) { - *keysym = XK_braille_blank | pattern; - if(status) *status = XLookupBoth; - } else - if(status) *status = XLookupChars; - } else { /* Composed Event */ - ret = _Xwcslen(&wc[b[ic->private.local.composed].wc]); - if(ret > wlen) { - if(status) *status = XBufferOverflow; - return (ret); - } - memcpy((char *)buffer, (char *)&wc[b[ic->private.local.composed].wc], - ret * sizeof(wchar_t)); - if(keysym) *keysym = b[ic->private.local.composed].ks; - if (ret > 0) { - if (keysym && *keysym != NoSymbol) { - if(status) *status = XLookupBoth; - } else { - if(status) *status = XLookupChars; - } - } else { - if(keysym && *keysym != NoSymbol) { - if(status) *status = XLookupKeySym; - } else { - if(status) *status = XLookupNone; - } - } - } - return (ret); - } else { /* Throughed Event */ - ret = _XimLookupWCText(ic, ev, buffer, wlen, keysym, NULL); - if(ret > 0) { - if (ret > wlen) { - if (status) *status = XBufferOverflow; - } else if (keysym && *keysym != NoSymbol) { - if(status) *status = XLookupBoth; - } else { - if(status) *status = XLookupChars; - } - } else { - if(keysym && *keysym != NoSymbol) { - if(status) *status = XLookupKeySym; - } else { - if(status) *status = XLookupNone; - } - } - } - return (ret); -} - -Public int -_XimLocalUtf8LookupString(XIC xic, XKeyEvent *ev, char *buffer, int bytes, - KeySym *keysym, Status *status) -{ - Xic ic = (Xic)xic; - int ret; - DefTree *b = ic->private.local.base.tree; - char *utf8 = ic->private.local.base.utf8; - - if(ev->type != KeyPress) { - if(status) *status = XLookupNone; - return(0); - } - if(ev->keycode == 0) { - if (ic->private.local.brl_committed != 0) { /* Braille Event */ - unsigned char pattern = ic->private.local.brl_committed; - ret = 3; - if (ret > bytes) { - if(status) *status = XBufferOverflow; - return (ret); - } - buffer[0] = 0xe0 | ((BRL_UC_ROW >> 12) & 0x0f); - buffer[1] = 0x80 | ((BRL_UC_ROW >> 8) & 0x30) | (pattern >> 6); - buffer[2] = 0x80 | (pattern & 0x3f); - } else { /* Composed Event */ - ret = strlen(&utf8[b[ic->private.local.composed].utf8]); - if(ret > bytes) { - if(status) *status = XBufferOverflow; - return (ret); - } - memcpy(buffer, &utf8[b[ic->private.local.composed].utf8], ret); - if(keysym) *keysym = b[ic->private.local.composed].ks; - if (ret > 0) { - if (keysym && *keysym != NoSymbol) { - if(status) *status = XLookupBoth; - } else { - if(status) *status = XLookupChars; - } - } else { - if(keysym && *keysym != NoSymbol) { - if(status) *status = XLookupKeySym; - } else { - if(status) *status = XLookupNone; - } - } - } - return (ret); - } else { /* Throughed Event */ - ret = _XimLookupUTF8Text(ic, ev, buffer, bytes, keysym, NULL); - if(ret > 0) { - if (ret > bytes) { - if (status) *status = XBufferOverflow; - } else if (keysym && *keysym != NoSymbol) { - if(status) *status = XLookupBoth; - } else { - if(status) *status = XLookupChars; - } - } else { - if(keysym && *keysym != NoSymbol) { - if(status) *status = XLookupKeySym; - } else { - if(status) *status = XLookupNone; - } - } - } - return (ret); -} - -Private int -_XimLcctsconvert( - XlcConv conv, - char *from, - int from_len, - char *to, - int to_len, - Status *state) -{ - int from_left; - int to_left; - int from_savelen; - int to_savelen; - int from_cnvlen; - int to_cnvlen; - char *from_buf; - char *to_buf; - char scratchbuf[BUFSIZ]; - Status tmp_state; - - if (!state) - state = &tmp_state; - - if (!conv || !from || !from_len) { - *state = XLookupNone; - return 0; - } - - /* Reset the converter. The CompoundText at 'from' starts in - initial state. */ - _XlcResetConverter(conv); - - from_left = from_len; - to_left = BUFSIZ; - from_cnvlen = 0; - to_cnvlen = 0; - for (;;) { - from_buf = &from[from_cnvlen]; - from_savelen = from_left; - to_buf = &scratchbuf[to_cnvlen]; - to_savelen = to_left; - if (_XlcConvert(conv, (XPointer *)&from_buf, &from_left, - (XPointer *)&to_buf, &to_left, NULL, 0) < 0) { - *state = XLookupNone; - return 0; - } - from_cnvlen += (from_savelen - from_left); - to_cnvlen += (to_savelen - to_left); - if (from_left == 0) { - if (!to_cnvlen) { - *state = XLookupNone; - return 0; - } - break; - } - } - - if (!to || !to_len || (to_len < to_cnvlen)) { - *state = XBufferOverflow; - } else { - memcpy(to, scratchbuf, to_cnvlen); - *state = XLookupChars; - } - return to_cnvlen; -} - -Public int -_XimLcctstombs(XIM xim, char *from, int from_len, - char *to, int to_len, Status *state) -{ - return _XimLcctsconvert(((Xim)xim)->private.local.ctom_conv, - from, from_len, to, to_len, state); -} - -Public int -_XimLcctstowcs(XIM xim, char *from, int from_len, - wchar_t *to, int to_len, Status *state) -{ - Xim im = (Xim)xim; - XlcConv conv = im->private.local.ctow_conv; - int from_left; - int to_left; - int from_savelen; - int to_savelen; - int from_cnvlen; - int to_cnvlen; - char *from_buf; - wchar_t *to_buf; - wchar_t scratchbuf[BUFSIZ]; - Status tmp_state; - - if (!state) - state = &tmp_state; - - if (!conv || !from || !from_len) { - *state = XLookupNone; - return 0; - } - - /* Reset the converter. The CompoundText at 'from' starts in - initial state. */ - _XlcResetConverter(conv); - - from_left = from_len; - to_left = BUFSIZ; - from_cnvlen = 0; - to_cnvlen = 0; - for (;;) { - from_buf = &from[from_cnvlen]; - from_savelen = from_left; - to_buf = &scratchbuf[to_cnvlen]; - to_savelen = to_left; - if (_XlcConvert(conv, (XPointer *)&from_buf, &from_left, - (XPointer *)&to_buf, &to_left, NULL, 0) < 0) { - *state = XLookupNone; - return 0; - } - from_cnvlen += (from_savelen - from_left); - to_cnvlen += (to_savelen - to_left); - if (from_left == 0) { - if (!to_cnvlen){ - *state = XLookupNone; - return 0; - } - break; - } - } - - if (!to || !to_len || (to_len < to_cnvlen)) { - *state = XBufferOverflow; - } else { - memcpy(to, scratchbuf, to_cnvlen * sizeof(wchar_t)); - *state = XLookupChars; - } - return to_cnvlen; -} - -Public int -_XimLcctstoutf8(XIM xim, char *from, int from_len, - char *to, int to_len, Status *state) -{ - return _XimLcctsconvert(((Xim)xim)->private.local.ctoutf8_conv, - from, from_len, to, to_len, state); -} +/******************************************************************
+
+ Copyright 1992 by Fuji Xerox Co., Ltd.
+ Copyright 1992, 1994 by FUJITSU LIMITED
+
+Permission to use, copy, modify, distribute, and sell this software
+and its documentation for any purpose is hereby granted without fee,
+provided that the above copyright notice appear in all copies and
+that both that copyright notice and this permission notice appear
+in supporting documentation, and that the name of Fuji Xerox,
+FUJITSU LIMITED not be used in advertising or publicity pertaining
+to distribution of the software without specific, written prior
+permission. Fuji Xerox, FUJITSU LIMITED make no representations
+about the suitability of this software for any purpose.
+It is provided "as is" without express or implied warranty.
+
+FUJI XEROX, FUJITSU LIMITED DISCLAIM ALL WARRANTIES WITH
+REGARD TO THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF
+MERCHANTABILITY AND FITNESS, IN NO EVENT SHALL FUJI XEROX,
+FUJITSU LIMITED BE LIABLE FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL
+DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA
+OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER
+TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR
+PERFORMANCE OF THIS SOFTWARE.
+
+ Author: Kazunori Nishihara Fuji Xerox
+ Takashi Fujiwara FUJITSU LIMITED
+ fujiwara@a80.tech.yk.fujitsu.co.jp
+
+******************************************************************/
+
+#ifdef HAVE_CONFIG_H
+#include <config.h>
+#endif
+#include <stdio.h>
+#include <X11/Xatom.h>
+#include <X11/Xos.h>
+#include <X11/Xlib.h>
+#include <X11/keysym.h>
+#include <X11/Xutil.h>
+#include "Xlibint.h"
+#include "Xlcint.h"
+#include "XlcPubI.h"
+#include "Ximint.h"
+
+Public int
+_XimLocalMbLookupString(XIC xic, XKeyEvent *ev, char *buffer, int bytes,
+ KeySym *keysym, Status *status)
+{
+ Xic ic = (Xic)xic;
+ int ret;
+ DefTree *b = ic->private.local.base.tree;
+ char *mb = ic->private.local.base.mb;
+
+ if(ev->type != KeyPress) {
+ if(status) *status = XLookupNone;
+ return(0);
+ }
+ if(ev->keycode == 0 &&
+ ( (ic->private.local.composed != 0)
+ ||(ic->private.local.brl_committed != 0))) {
+ if (ic->private.local.brl_committed != 0) { /* Braille Event */
+ unsigned char pattern = ic->private.local.brl_committed;
+ char *mb=alloca(XLC_PUBLIC(ic->core.im->core.lcd, mb_cur_max));
+ ret = _Xlcwctomb(ic->core.im->core.lcd, mb, BRL_UC_ROW | pattern);
+ if(ret > bytes) {
+ if(status) *status = XBufferOverflow;
+ return(ret);
+ }
+ if(keysym) *keysym = XK_braille_blank | pattern;
+ if(ret > 0) {
+ if (keysym) {
+ if(status) *status = XLookupBoth;
+ } else {
+ if(status) *status = XLookupChars;
+ }
+ memcpy(buffer, mb, ret);
+ } else {
+ if(keysym) {
+ if(status) *status = XLookupKeySym;
+ } else {
+ if(status) *status = XLookupNone;
+ }
+ }
+ } else { /* Composed Event */
+ ret = strlen(&mb[b[ic->private.local.composed].mb]);
+ if(ret > bytes) {
+ if(status) *status = XBufferOverflow;
+ return(ret);
+ }
+ memcpy(buffer, &mb[b[ic->private.local.composed].mb], ret);
+ if(keysym) *keysym = b[ic->private.local.composed].ks;
+ if (ret > 0) {
+ if (keysym && *keysym != NoSymbol) {
+ if(status) *status = XLookupBoth;
+ } else {
+ if(status) *status = XLookupChars;
+ }
+ } else {
+ if(keysym && *keysym != NoSymbol) {
+ if(status) *status = XLookupKeySym;
+ } else {
+ if(status) *status = XLookupNone;
+ }
+ }
+ }
+ return (ret);
+ } else { /* Throughed Event */
+ ret = _XimLookupMBText(ic, ev, buffer, bytes, keysym, NULL);
+ if(ret > 0) {
+ if (ret > bytes) {
+ if (status) *status = XBufferOverflow;
+ } else if (keysym && *keysym != NoSymbol) {
+ if(status) *status = XLookupBoth;
+ } else {
+ if(status) *status = XLookupChars;
+ }
+ } else {
+ if(keysym && *keysym != NoSymbol) {
+ if(status) *status = XLookupKeySym;
+ } else {
+ if(status) *status = XLookupNone;
+ }
+ }
+ }
+ return (ret);
+}
+
+Public int
+_XimLocalWcLookupString(XIC xic, XKeyEvent *ev, wchar_t *buffer, int wlen,
+ KeySym *keysym, Status *status)
+{
+ Xic ic = (Xic)xic;
+ int ret;
+ DefTree *b = ic->private.local.base.tree;
+ wchar_t *wc = ic->private.local.base.wc;
+
+ if(ev->type != KeyPress) {
+ if(status) *status = XLookupNone;
+ return(0);
+ }
+ if(ev->keycode == 0) {
+ if (ic->private.local.brl_committed != 0) { /* Braille Event */
+ unsigned char pattern = ic->private.local.brl_committed;
+ ret = 1;
+ if (ret > wlen) {
+ if(status) *status = XBufferOverflow;
+ return (ret);
+ }
+ *buffer = BRL_UC_ROW | pattern;
+ if(keysym) {
+ *keysym = XK_braille_blank | pattern;
+ if(status) *status = XLookupBoth;
+ } else
+ if(status) *status = XLookupChars;
+ } else { /* Composed Event */
+ ret = _Xwcslen(&wc[b[ic->private.local.composed].wc]);
+ if(ret > wlen) {
+ if(status) *status = XBufferOverflow;
+ return (ret);
+ }
+ memcpy((char *)buffer, (char *)&wc[b[ic->private.local.composed].wc],
+ ret * sizeof(wchar_t));
+ if(keysym) *keysym = b[ic->private.local.composed].ks;
+ if (ret > 0) {
+ if (keysym && *keysym != NoSymbol) {
+ if(status) *status = XLookupBoth;
+ } else {
+ if(status) *status = XLookupChars;
+ }
+ } else {
+ if(keysym && *keysym != NoSymbol) {
+ if(status) *status = XLookupKeySym;
+ } else {
+ if(status) *status = XLookupNone;
+ }
+ }
+ }
+ return (ret);
+ } else { /* Throughed Event */
+ ret = _XimLookupWCText(ic, ev, buffer, wlen, keysym, NULL);
+ if(ret > 0) {
+ if (ret > wlen) {
+ if (status) *status = XBufferOverflow;
+ } else if (keysym && *keysym != NoSymbol) {
+ if(status) *status = XLookupBoth;
+ } else {
+ if(status) *status = XLookupChars;
+ }
+ } else {
+ if(keysym && *keysym != NoSymbol) {
+ if(status) *status = XLookupKeySym;
+ } else {
+ if(status) *status = XLookupNone;
+ }
+ }
+ }
+ return (ret);
+}
+
+Public int
+_XimLocalUtf8LookupString(XIC xic, XKeyEvent *ev, char *buffer, int bytes,
+ KeySym *keysym, Status *status)
+{
+ Xic ic = (Xic)xic;
+ int ret;
+ DefTree *b = ic->private.local.base.tree;
+ char *utf8 = ic->private.local.base.utf8;
+
+ if(ev->type != KeyPress) {
+ if(status) *status = XLookupNone;
+ return(0);
+ }
+ if(ev->keycode == 0) {
+ if (ic->private.local.brl_committed != 0) { /* Braille Event */
+ unsigned char pattern = ic->private.local.brl_committed;
+ ret = 3;
+ if (ret > bytes) {
+ if(status) *status = XBufferOverflow;
+ return (ret);
+ }
+ buffer[0] = 0xe0 | ((BRL_UC_ROW >> 12) & 0x0f);
+ buffer[1] = 0x80 | ((BRL_UC_ROW >> 8) & 0x30) | (pattern >> 6);
+ buffer[2] = 0x80 | (pattern & 0x3f);
+ if(keysym) {
+ *keysym = XK_braille_blank | pattern;
+ if(status) *status = XLookupBoth;
+ } else
+ if(status) *status = XLookupChars;
+ } else { /* Composed Event */
+ ret = strlen(&utf8[b[ic->private.local.composed].utf8]);
+ if(ret > bytes) {
+ if(status) *status = XBufferOverflow;
+ return (ret);
+ }
+ memcpy(buffer, &utf8[b[ic->private.local.composed].utf8], ret);
+ if(keysym) *keysym = b[ic->private.local.composed].ks;
+ if (ret > 0) {
+ if (keysym && *keysym != NoSymbol) {
+ if(status) *status = XLookupBoth;
+ } else {
+ if(status) *status = XLookupChars;
+ }
+ } else {
+ if(keysym && *keysym != NoSymbol) {
+ if(status) *status = XLookupKeySym;
+ } else {
+ if(status) *status = XLookupNone;
+ }
+ }
+ }
+ return (ret);
+ } else { /* Throughed Event */
+ ret = _XimLookupUTF8Text(ic, ev, buffer, bytes, keysym, NULL);
+ if(ret > 0) {
+ if (ret > bytes) {
+ if (status) *status = XBufferOverflow;
+ } else if (keysym && *keysym != NoSymbol) {
+ if(status) *status = XLookupBoth;
+ } else {
+ if(status) *status = XLookupChars;
+ }
+ } else {
+ if(keysym && *keysym != NoSymbol) {
+ if(status) *status = XLookupKeySym;
+ } else {
+ if(status) *status = XLookupNone;
+ }
+ }
+ }
+ return (ret);
+}
+
+Private int
+_XimLcctsconvert(
+ XlcConv conv,
+ char *from,
+ int from_len,
+ char *to,
+ int to_len,
+ Status *state)
+{
+ int from_left;
+ int to_left;
+ int from_savelen;
+ int to_savelen;
+ int from_cnvlen;
+ int to_cnvlen;
+ char *from_buf;
+ char *to_buf;
+ char scratchbuf[BUFSIZ];
+ Status tmp_state;
+
+ if (!state)
+ state = &tmp_state;
+
+ if (!conv || !from || !from_len) {
+ *state = XLookupNone;
+ return 0;
+ }
+
+ /* Reset the converter. The CompoundText at 'from' starts in
+ initial state. */
+ _XlcResetConverter(conv);
+
+ from_left = from_len;
+ to_left = BUFSIZ;
+ from_cnvlen = 0;
+ to_cnvlen = 0;
+ for (;;) {
+ from_buf = &from[from_cnvlen];
+ from_savelen = from_left;
+ to_buf = &scratchbuf[to_cnvlen];
+ to_savelen = to_left;
+ if (_XlcConvert(conv, (XPointer *)&from_buf, &from_left,
+ (XPointer *)&to_buf, &to_left, NULL, 0) < 0) {
+ *state = XLookupNone;
+ return 0;
+ }
+ from_cnvlen += (from_savelen - from_left);
+ to_cnvlen += (to_savelen - to_left);
+ if (from_left == 0) {
+ if (!to_cnvlen) {
+ *state = XLookupNone;
+ return 0;
+ }
+ break;
+ }
+ }
+
+ if (!to || !to_len || (to_len < to_cnvlen)) {
+ *state = XBufferOverflow;
+ } else {
+ memcpy(to, scratchbuf, to_cnvlen);
+ *state = XLookupChars;
+ }
+ return to_cnvlen;
+}
+
+Public int
+_XimLcctstombs(XIM xim, char *from, int from_len,
+ char *to, int to_len, Status *state)
+{
+ return _XimLcctsconvert(((Xim)xim)->private.local.ctom_conv,
+ from, from_len, to, to_len, state);
+}
+
+Public int
+_XimLcctstowcs(XIM xim, char *from, int from_len,
+ wchar_t *to, int to_len, Status *state)
+{
+ Xim im = (Xim)xim;
+ XlcConv conv = im->private.local.ctow_conv;
+ int from_left;
+ int to_left;
+ int from_savelen;
+ int to_savelen;
+ int from_cnvlen;
+ int to_cnvlen;
+ char *from_buf;
+ wchar_t *to_buf;
+ wchar_t scratchbuf[BUFSIZ];
+ Status tmp_state;
+
+ if (!state)
+ state = &tmp_state;
+
+ if (!conv || !from || !from_len) {
+ *state = XLookupNone;
+ return 0;
+ }
+
+ /* Reset the converter. The CompoundText at 'from' starts in
+ initial state. */
+ _XlcResetConverter(conv);
+
+ from_left = from_len;
+ to_left = BUFSIZ;
+ from_cnvlen = 0;
+ to_cnvlen = 0;
+ for (;;) {
+ from_buf = &from[from_cnvlen];
+ from_savelen = from_left;
+ to_buf = &scratchbuf[to_cnvlen];
+ to_savelen = to_left;
+ if (_XlcConvert(conv, (XPointer *)&from_buf, &from_left,
+ (XPointer *)&to_buf, &to_left, NULL, 0) < 0) {
+ *state = XLookupNone;
+ return 0;
+ }
+ from_cnvlen += (from_savelen - from_left);
+ to_cnvlen += (to_savelen - to_left);
+ if (from_left == 0) {
+ if (!to_cnvlen){
+ *state = XLookupNone;
+ return 0;
+ }
+ break;
+ }
+ }
+
+ if (!to || !to_len || (to_len < to_cnvlen)) {
+ *state = XBufferOverflow;
+ } else {
+ memcpy(to, scratchbuf, to_cnvlen * sizeof(wchar_t));
+ *state = XLookupChars;
+ }
+ return to_cnvlen;
+}
+
+Public int
+_XimLcctstoutf8(XIM xim, char *from, int from_len,
+ char *to, int to_len, Status *state)
+{
+ return _XimLcctsconvert(((Xim)xim)->private.local.ctoutf8_conv,
+ from, from_len, to, to_len, state);
+}
diff --git a/mesalib/src/glsl/ast_expr.cpp b/mesalib/src/glsl/ast_expr.cpp index 974beb9f6..e8812e7a1 100644 --- a/mesalib/src/glsl/ast_expr.cpp +++ b/mesalib/src/glsl/ast_expr.cpp @@ -1,98 +1,95 @@ -/* - * Copyright © 2010 Intel Corporation - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING - * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER - * DEALINGS IN THE SOFTWARE. - */ -#include <cstdio> -#include <cassert> -#include "ast.h" - -using std::printf; - -const char * -ast_expression::operator_string(enum ast_operators op) -{ - static const char *const operators[] = { - "=", - "+", - "-", - "+", - "-", - "*", - "/", - "%", - "<<", - ">>", - "<", - ">", - "<=", - ">=", - "==", - "!=", - "&", - "^", - "|", - "~", - "&&", - "^^", - "||", - "!", - - "*=", - "/=", - "%=", - "+=", - "-=", - "<<=", - ">>=", - "&=", - "^=", - "|=", - - "?:", - - "++", - "--", - "++", - "--", - ".", - }; - - assert((unsigned int)op < sizeof(operators) / sizeof(operators[0])); - - return operators[op]; -} - - -ast_expression_bin::ast_expression_bin(int oper, ast_expression *ex0, - ast_expression *ex1) : - ast_expression(oper, ex0, ex1, NULL) -{ - assert((oper >= ast_plus) && (oper <= ast_logic_not)); -} - - -void -ast_expression_bin::print(void) const -{ - subexpressions[0]->print(); - printf("%s ", operator_string(oper)); - subexpressions[1]->print(); -} +/*
+ * Copyright © 2010 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+#include <assert.h>
+#include "ast.h"
+
+const char *
+ast_expression::operator_string(enum ast_operators op)
+{
+ static const char *const operators[] = {
+ "=",
+ "+",
+ "-",
+ "+",
+ "-",
+ "*",
+ "/",
+ "%",
+ "<<",
+ ">>",
+ "<",
+ ">",
+ "<=",
+ ">=",
+ "==",
+ "!=",
+ "&",
+ "^",
+ "|",
+ "~",
+ "&&",
+ "^^",
+ "||",
+ "!",
+
+ "*=",
+ "/=",
+ "%=",
+ "+=",
+ "-=",
+ "<<=",
+ ">>=",
+ "&=",
+ "^=",
+ "|=",
+
+ "?:",
+
+ "++",
+ "--",
+ "++",
+ "--",
+ ".",
+ };
+
+ assert((unsigned int)op < sizeof(operators) / sizeof(operators[0]));
+
+ return operators[op];
+}
+
+
+ast_expression_bin::ast_expression_bin(int oper, ast_expression *ex0,
+ ast_expression *ex1) :
+ ast_expression(oper, ex0, ex1, NULL)
+{
+ assert((oper >= ast_plus) && (oper <= ast_logic_not));
+}
+
+
+void
+ast_expression_bin::print(void) const
+{
+ subexpressions[0]->print();
+ printf("%s ", operator_string(oper));
+ subexpressions[1]->print();
+}
diff --git a/mesalib/src/glsl/ast_type.cpp b/mesalib/src/glsl/ast_type.cpp index 5ddfeec87..8e49bef8a 100644 --- a/mesalib/src/glsl/ast_type.cpp +++ b/mesalib/src/glsl/ast_type.cpp @@ -1,141 +1,138 @@ -/* - * Copyright © 2010 Intel Corporation - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING - * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER - * DEALINGS IN THE SOFTWARE. - */ - -#include <cstdio> -#include "ast.h" -extern "C" { -#include "program/symbol_table.h" -} - -using std::printf; - -void -ast_type_specifier::print(void) const -{ - if (type_specifier == ast_struct) { - structure->print(); - } else { - printf("%s ", type_name); - } - - if (is_array) { - printf("[ "); - - if (array_size) { - array_size->print(); - } - - printf("] "); - } -} - -ast_type_specifier::ast_type_specifier(int specifier) - : type_specifier(ast_types(specifier)), type_name(NULL), structure(NULL), - is_array(false), array_size(NULL), precision(ast_precision_none), - is_precision_statement(false) -{ - static const char *const names[] = { - "void", - "float", - "int", - "uint", - "bool", - "vec2", - "vec3", - "vec4", - "bvec2", - "bvec3", - "bvec4", - "ivec2", - "ivec3", - "ivec4", - "uvec2", - "uvec3", - "uvec4", - "mat2", - "mat2x3", - "mat2x4", - "mat3x2", - "mat3", - "mat3x4", - "mat4x2", - "mat4x3", - "mat4", - "sampler1D", - "sampler2D", - "sampler2DRect", - "sampler3D", - "samplerCube", - "sampler1DShadow", - "sampler2DShadow", - "sampler2DRectShadow", - "samplerCubeShadow", - "sampler1DArray", - "sampler2DArray", - "sampler1DArrayShadow", - "sampler2DArrayShadow", - "isampler1D", - "isampler2D", - "isampler3D", - "isamplerCube", - "isampler1DArray", - "isampler2DArray", - "usampler1D", - "usampler2D", - "usampler3D", - "usamplerCube", - "usampler1DArray", - "usampler2DArray", - - NULL, /* ast_struct */ - NULL /* ast_type_name */ - }; - - type_name = names[specifier]; -} - -bool -ast_fully_specified_type::has_qualifiers() const -{ - return this->qualifier.flags.i != 0; -} - -bool ast_type_qualifier::has_interpolation() const -{ - return this->flags.q.smooth - || this->flags.q.flat - || this->flags.q.noperspective; -} - -const char* -ast_type_qualifier::interpolation_string() const -{ - if (this->flags.q.smooth) - return "smooth"; - else if (this->flags.q.flat) - return "flat"; - else if (this->flags.q.noperspective) - return "noperspective"; - else - return NULL; -} +/*
+ * Copyright © 2010 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include "ast.h"
+extern "C" {
+#include "program/symbol_table.h"
+}
+
+void
+ast_type_specifier::print(void) const
+{
+ if (type_specifier == ast_struct) {
+ structure->print();
+ } else {
+ printf("%s ", type_name);
+ }
+
+ if (is_array) {
+ printf("[ ");
+
+ if (array_size) {
+ array_size->print();
+ }
+
+ printf("] ");
+ }
+}
+
+ast_type_specifier::ast_type_specifier(int specifier)
+ : type_specifier(ast_types(specifier)), type_name(NULL), structure(NULL),
+ is_array(false), array_size(NULL), precision(ast_precision_none),
+ is_precision_statement(false)
+{
+ static const char *const names[] = {
+ "void",
+ "float",
+ "int",
+ "uint",
+ "bool",
+ "vec2",
+ "vec3",
+ "vec4",
+ "bvec2",
+ "bvec3",
+ "bvec4",
+ "ivec2",
+ "ivec3",
+ "ivec4",
+ "uvec2",
+ "uvec3",
+ "uvec4",
+ "mat2",
+ "mat2x3",
+ "mat2x4",
+ "mat3x2",
+ "mat3",
+ "mat3x4",
+ "mat4x2",
+ "mat4x3",
+ "mat4",
+ "sampler1D",
+ "sampler2D",
+ "sampler2DRect",
+ "sampler3D",
+ "samplerCube",
+ "sampler1DShadow",
+ "sampler2DShadow",
+ "sampler2DRectShadow",
+ "samplerCubeShadow",
+ "sampler1DArray",
+ "sampler2DArray",
+ "sampler1DArrayShadow",
+ "sampler2DArrayShadow",
+ "isampler1D",
+ "isampler2D",
+ "isampler3D",
+ "isamplerCube",
+ "isampler1DArray",
+ "isampler2DArray",
+ "usampler1D",
+ "usampler2D",
+ "usampler3D",
+ "usamplerCube",
+ "usampler1DArray",
+ "usampler2DArray",
+
+ NULL, /* ast_struct */
+ NULL /* ast_type_name */
+ };
+
+ type_name = names[specifier];
+}
+
+bool
+ast_fully_specified_type::has_qualifiers() const
+{
+ return this->qualifier.flags.i != 0;
+}
+
+bool ast_type_qualifier::has_interpolation() const
+{
+ return this->flags.q.smooth
+ || this->flags.q.flat
+ || this->flags.q.noperspective;
+}
+
+const char*
+ast_type_qualifier::interpolation_string() const
+{
+ if (this->flags.q.smooth)
+ return "smooth";
+ else if (this->flags.q.flat)
+ return "flat";
+ else if (this->flags.q.noperspective)
+ return "noperspective";
+ else
+ return NULL;
+}
diff --git a/mesalib/src/glsl/glsl_parser_extras.h b/mesalib/src/glsl/glsl_parser_extras.h index c628db94e..0d9c78719 100644 --- a/mesalib/src/glsl/glsl_parser_extras.h +++ b/mesalib/src/glsl/glsl_parser_extras.h @@ -1,281 +1,281 @@ -/* - * Copyright © 2010 Intel Corporation - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING - * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER - * DEALINGS IN THE SOFTWARE. - */ - -#pragma once -#ifndef GLSL_PARSER_EXTRAS_H -#define GLSL_PARSER_EXTRAS_H - -/* - * Most of the definitions here only apply to C++ - */ -#ifdef __cplusplus - - -#include <cstdlib> -#include "glsl_symbol_table.h" - -enum _mesa_glsl_parser_targets { - vertex_shader, - geometry_shader, - fragment_shader -}; - -struct gl_context; - -struct _mesa_glsl_parse_state { - _mesa_glsl_parse_state(struct gl_context *ctx, GLenum target, - void *mem_ctx); - - /* Callers of this ralloc-based new need not call delete. It's - * easier to just ralloc_free 'ctx' (or any of its ancestors). */ - static void* operator new(size_t size, void *ctx) - { - void *mem = rzalloc_size(ctx, size); - assert(mem != NULL); - - return mem; - } - - /* If the user *does* call delete, that's OK, we will just - * ralloc_free in that case. */ - static void operator delete(void *mem, void *ctx) - { - ralloc_free(mem); - } - static void operator delete(void *mem) - { - ralloc_free(mem); - } - - void *scanner; - exec_list translation_unit; - glsl_symbol_table *symbols; - - bool es_shader; - unsigned language_version; - const char *version_string; - enum _mesa_glsl_parser_targets target; - - /** - * Printable list of GLSL versions supported by the current context - * - * \note - * This string should probably be generated per-context instead of per - * invokation of the compiler. This should be changed when the method of - * tracking supported GLSL versions changes. - */ - const char *supported_version_string; - - /** - * Implementation defined limits that affect built-in variables, etc. - * - * \sa struct gl_constants (in mtypes.h) - */ - struct { - /* 1.10 */ - unsigned MaxLights; - unsigned MaxClipPlanes; - unsigned MaxTextureUnits; - unsigned MaxTextureCoords; - unsigned MaxVertexAttribs; - unsigned MaxVertexUniformComponents; - unsigned MaxVaryingFloats; - unsigned MaxVertexTextureImageUnits; - unsigned MaxCombinedTextureImageUnits; - unsigned MaxTextureImageUnits; - unsigned MaxFragmentUniformComponents; - - /* ARB_draw_buffers */ - unsigned MaxDrawBuffers; - - /** - * Set of GLSL versions supported by the current context - * - * Knowing that version X is supported doesn't mean that versions before - * X are also supported. Version 1.00 is only supported in an ES2 - * context or when GL_ARB_ES2_compatibility is supported. In an OpenGL - * 3.0 "forward compatible" context, GLSL 1.10 and 1.20 are \b not - * supported. - */ - /*@{*/ - unsigned GLSL_100ES:1; - unsigned GLSL_110:1; - unsigned GLSL_120:1; - unsigned GLSL_130:1; - /*@}*/ - } Const; - - /** - * During AST to IR conversion, pointer to current IR function - * - * Will be \c NULL whenever the AST to IR conversion is not inside a - * function definition. - */ - class ir_function_signature *current_function; - - /** Have we found a return statement in this function? */ - bool found_return; - - /** Was there an error during compilation? */ - bool error; - - /** - * Are all shader inputs / outputs invariant? - * - * This is set when the 'STDGL invariant(all)' pragma is used. - */ - bool all_invariant; - - /** Loop or switch statement containing the current instructions. */ - class ir_instruction *loop_or_switch_nesting; - class ast_iteration_statement *loop_or_switch_nesting_ast; - - /** List of structures defined in user code. */ - const glsl_type **user_structures; - unsigned num_user_structures; - - char *info_log; - - /** - * \name Enable bits for GLSL extensions - */ - /*@{*/ - unsigned ARB_draw_buffers_enable:1; - unsigned ARB_draw_buffers_warn:1; - unsigned ARB_draw_instanced_enable:1; - unsigned ARB_draw_instanced_warn:1; - unsigned ARB_explicit_attrib_location_enable:1; - unsigned ARB_explicit_attrib_location_warn:1; - unsigned ARB_fragment_coord_conventions_enable:1; - unsigned ARB_fragment_coord_conventions_warn:1; - unsigned ARB_texture_rectangle_enable:1; - unsigned ARB_texture_rectangle_warn:1; - unsigned EXT_texture_array_enable:1; - unsigned EXT_texture_array_warn:1; - unsigned ARB_shader_stencil_export_enable:1; - unsigned ARB_shader_stencil_export_warn:1; - unsigned AMD_conservative_depth_enable:1; - unsigned AMD_conservative_depth_warn:1; - /*@}*/ - - /** Extensions supported by the OpenGL implementation. */ - const struct gl_extensions *extensions; - - /** Shaders containing built-in functions that are used for linking. */ - struct gl_shader *builtins_to_link[16]; - unsigned num_builtins_to_link; -}; - -typedef struct YYLTYPE { - int first_line; - int first_column; - int last_line; - int last_column; - unsigned source; -} YYLTYPE; -# define YYLTYPE_IS_DECLARED 1 -# define YYLTYPE_IS_TRIVIAL 1 - -# define YYLLOC_DEFAULT(Current, Rhs, N) \ -do { \ - if (N) \ - { \ - (Current).first_line = YYRHSLOC(Rhs, 1).first_line; \ - (Current).first_column = YYRHSLOC(Rhs, 1).first_column; \ - (Current).last_line = YYRHSLOC(Rhs, N).last_line; \ - (Current).last_column = YYRHSLOC(Rhs, N).last_column; \ - } \ - else \ - { \ - (Current).first_line = (Current).last_line = \ - YYRHSLOC(Rhs, 0).last_line; \ - (Current).first_column = (Current).last_column = \ - YYRHSLOC(Rhs, 0).last_column; \ - } \ - (Current).source = 0; \ -} while (0) - -extern void _mesa_glsl_error(YYLTYPE *locp, _mesa_glsl_parse_state *state, - const char *fmt, ...); - -/** - * Emit a warning to the shader log - * - * \sa _mesa_glsl_error - */ -extern void _mesa_glsl_warning(const YYLTYPE *locp, - _mesa_glsl_parse_state *state, - const char *fmt, ...); - -extern void _mesa_glsl_lexer_ctor(struct _mesa_glsl_parse_state *state, - const char *string); - -extern void _mesa_glsl_lexer_dtor(struct _mesa_glsl_parse_state *state); - -union YYSTYPE; -extern int _mesa_glsl_lex(union YYSTYPE *yylval, YYLTYPE *yylloc, - void *scanner); - -extern int _mesa_glsl_parse(struct _mesa_glsl_parse_state *); - -/** - * Process elements of the #extension directive - * - * \return - * If \c name and \c behavior are valid, \c true is returned. Otherwise - * \c false is returned. - */ -extern bool _mesa_glsl_process_extension(const char *name, YYLTYPE *name_locp, - const char *behavior, - YYLTYPE *behavior_locp, - _mesa_glsl_parse_state *state); - -/** - * Get the textual name of the specified shader target - */ -extern const char * -_mesa_glsl_shader_target_name(enum _mesa_glsl_parser_targets target); - - -#endif /* __cplusplus */ - - -/* - * These definitions apply to C and C++ - */ -#ifdef __cplusplus -extern "C" { -#endif - -extern int preprocess(void *ctx, const char **shader, char **info_log, - const struct gl_extensions *extensions, int api); - -extern void _mesa_destroy_shader_compiler(); -extern void _mesa_destroy_shader_compiler_caches(); - -#ifdef __cplusplus -} -#endif - - -#endif /* GLSL_PARSER_EXTRAS_H */ +/*
+ * Copyright © 2010 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#pragma once
+#ifndef GLSL_PARSER_EXTRAS_H
+#define GLSL_PARSER_EXTRAS_H
+
+/*
+ * Most of the definitions here only apply to C++
+ */
+#ifdef __cplusplus
+
+
+#include <stdlib.h>
+#include "glsl_symbol_table.h"
+
+enum _mesa_glsl_parser_targets {
+ vertex_shader,
+ geometry_shader,
+ fragment_shader
+};
+
+struct gl_context;
+
+struct _mesa_glsl_parse_state {
+ _mesa_glsl_parse_state(struct gl_context *ctx, GLenum target,
+ void *mem_ctx);
+
+ /* Callers of this ralloc-based new need not call delete. It's
+ * easier to just ralloc_free 'ctx' (or any of its ancestors). */
+ static void* operator new(size_t size, void *ctx)
+ {
+ void *mem = rzalloc_size(ctx, size);
+ assert(mem != NULL);
+
+ return mem;
+ }
+
+ /* If the user *does* call delete, that's OK, we will just
+ * ralloc_free in that case. */
+ static void operator delete(void *mem, void *ctx)
+ {
+ ralloc_free(mem);
+ }
+ static void operator delete(void *mem)
+ {
+ ralloc_free(mem);
+ }
+
+ void *scanner;
+ exec_list translation_unit;
+ glsl_symbol_table *symbols;
+
+ bool es_shader;
+ unsigned language_version;
+ const char *version_string;
+ enum _mesa_glsl_parser_targets target;
+
+ /**
+ * Printable list of GLSL versions supported by the current context
+ *
+ * \note
+ * This string should probably be generated per-context instead of per
+ * invokation of the compiler. This should be changed when the method of
+ * tracking supported GLSL versions changes.
+ */
+ const char *supported_version_string;
+
+ /**
+ * Implementation defined limits that affect built-in variables, etc.
+ *
+ * \sa struct gl_constants (in mtypes.h)
+ */
+ struct {
+ /* 1.10 */
+ unsigned MaxLights;
+ unsigned MaxClipPlanes;
+ unsigned MaxTextureUnits;
+ unsigned MaxTextureCoords;
+ unsigned MaxVertexAttribs;
+ unsigned MaxVertexUniformComponents;
+ unsigned MaxVaryingFloats;
+ unsigned MaxVertexTextureImageUnits;
+ unsigned MaxCombinedTextureImageUnits;
+ unsigned MaxTextureImageUnits;
+ unsigned MaxFragmentUniformComponents;
+
+ /* ARB_draw_buffers */
+ unsigned MaxDrawBuffers;
+
+ /**
+ * Set of GLSL versions supported by the current context
+ *
+ * Knowing that version X is supported doesn't mean that versions before
+ * X are also supported. Version 1.00 is only supported in an ES2
+ * context or when GL_ARB_ES2_compatibility is supported. In an OpenGL
+ * 3.0 "forward compatible" context, GLSL 1.10 and 1.20 are \b not
+ * supported.
+ */
+ /*@{*/
+ unsigned GLSL_100ES:1;
+ unsigned GLSL_110:1;
+ unsigned GLSL_120:1;
+ unsigned GLSL_130:1;
+ /*@}*/
+ } Const;
+
+ /**
+ * During AST to IR conversion, pointer to current IR function
+ *
+ * Will be \c NULL whenever the AST to IR conversion is not inside a
+ * function definition.
+ */
+ class ir_function_signature *current_function;
+
+ /** Have we found a return statement in this function? */
+ bool found_return;
+
+ /** Was there an error during compilation? */
+ bool error;
+
+ /**
+ * Are all shader inputs / outputs invariant?
+ *
+ * This is set when the 'STDGL invariant(all)' pragma is used.
+ */
+ bool all_invariant;
+
+ /** Loop or switch statement containing the current instructions. */
+ class ir_instruction *loop_or_switch_nesting;
+ class ast_iteration_statement *loop_or_switch_nesting_ast;
+
+ /** List of structures defined in user code. */
+ const glsl_type **user_structures;
+ unsigned num_user_structures;
+
+ char *info_log;
+
+ /**
+ * \name Enable bits for GLSL extensions
+ */
+ /*@{*/
+ unsigned ARB_draw_buffers_enable:1;
+ unsigned ARB_draw_buffers_warn:1;
+ unsigned ARB_draw_instanced_enable:1;
+ unsigned ARB_draw_instanced_warn:1;
+ unsigned ARB_explicit_attrib_location_enable:1;
+ unsigned ARB_explicit_attrib_location_warn:1;
+ unsigned ARB_fragment_coord_conventions_enable:1;
+ unsigned ARB_fragment_coord_conventions_warn:1;
+ unsigned ARB_texture_rectangle_enable:1;
+ unsigned ARB_texture_rectangle_warn:1;
+ unsigned EXT_texture_array_enable:1;
+ unsigned EXT_texture_array_warn:1;
+ unsigned ARB_shader_stencil_export_enable:1;
+ unsigned ARB_shader_stencil_export_warn:1;
+ unsigned AMD_conservative_depth_enable:1;
+ unsigned AMD_conservative_depth_warn:1;
+ /*@}*/
+
+ /** Extensions supported by the OpenGL implementation. */
+ const struct gl_extensions *extensions;
+
+ /** Shaders containing built-in functions that are used for linking. */
+ struct gl_shader *builtins_to_link[16];
+ unsigned num_builtins_to_link;
+};
+
+typedef struct YYLTYPE {
+ int first_line;
+ int first_column;
+ int last_line;
+ int last_column;
+ unsigned source;
+} YYLTYPE;
+# define YYLTYPE_IS_DECLARED 1
+# define YYLTYPE_IS_TRIVIAL 1
+
+# define YYLLOC_DEFAULT(Current, Rhs, N) \
+do { \
+ if (N) \
+ { \
+ (Current).first_line = YYRHSLOC(Rhs, 1).first_line; \
+ (Current).first_column = YYRHSLOC(Rhs, 1).first_column; \
+ (Current).last_line = YYRHSLOC(Rhs, N).last_line; \
+ (Current).last_column = YYRHSLOC(Rhs, N).last_column; \
+ } \
+ else \
+ { \
+ (Current).first_line = (Current).last_line = \
+ YYRHSLOC(Rhs, 0).last_line; \
+ (Current).first_column = (Current).last_column = \
+ YYRHSLOC(Rhs, 0).last_column; \
+ } \
+ (Current).source = 0; \
+} while (0)
+
+extern void _mesa_glsl_error(YYLTYPE *locp, _mesa_glsl_parse_state *state,
+ const char *fmt, ...);
+
+/**
+ * Emit a warning to the shader log
+ *
+ * \sa _mesa_glsl_error
+ */
+extern void _mesa_glsl_warning(const YYLTYPE *locp,
+ _mesa_glsl_parse_state *state,
+ const char *fmt, ...);
+
+extern void _mesa_glsl_lexer_ctor(struct _mesa_glsl_parse_state *state,
+ const char *string);
+
+extern void _mesa_glsl_lexer_dtor(struct _mesa_glsl_parse_state *state);
+
+union YYSTYPE;
+extern int _mesa_glsl_lex(union YYSTYPE *yylval, YYLTYPE *yylloc,
+ void *scanner);
+
+extern int _mesa_glsl_parse(struct _mesa_glsl_parse_state *);
+
+/**
+ * Process elements of the #extension directive
+ *
+ * \return
+ * If \c name and \c behavior are valid, \c true is returned. Otherwise
+ * \c false is returned.
+ */
+extern bool _mesa_glsl_process_extension(const char *name, YYLTYPE *name_locp,
+ const char *behavior,
+ YYLTYPE *behavior_locp,
+ _mesa_glsl_parse_state *state);
+
+/**
+ * Get the textual name of the specified shader target
+ */
+extern const char *
+_mesa_glsl_shader_target_name(enum _mesa_glsl_parser_targets target);
+
+
+#endif /* __cplusplus */
+
+
+/*
+ * These definitions apply to C and C++
+ */
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+extern int preprocess(void *ctx, const char **shader, char **info_log,
+ const struct gl_extensions *extensions, int api);
+
+extern void _mesa_destroy_shader_compiler();
+extern void _mesa_destroy_shader_compiler_caches();
+
+#ifdef __cplusplus
+}
+#endif
+
+
+#endif /* GLSL_PARSER_EXTRAS_H */
diff --git a/mesalib/src/glsl/glsl_types.cpp b/mesalib/src/glsl/glsl_types.cpp index f4d9242b2..1802bf0ac 100644 --- a/mesalib/src/glsl/glsl_types.cpp +++ b/mesalib/src/glsl/glsl_types.cpp @@ -1,497 +1,497 @@ -/* - * Copyright © 2009 Intel Corporation - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING - * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER - * DEALINGS IN THE SOFTWARE. - */ - -#include <cstdio> -#include <stdlib.h> -#include "main/core.h" /* for Elements */ -#include "glsl_symbol_table.h" -#include "glsl_parser_extras.h" -#include "glsl_types.h" -#include "builtin_types.h" -extern "C" { -#include "program/hash_table.h" -} - -hash_table *glsl_type::array_types = NULL; -hash_table *glsl_type::record_types = NULL; -void *glsl_type::mem_ctx = NULL; - -void -glsl_type::init_ralloc_type_ctx(void) -{ - if (glsl_type::mem_ctx == NULL) { - glsl_type::mem_ctx = ralloc_autofree_context(); - assert(glsl_type::mem_ctx != NULL); - } -} - -glsl_type::glsl_type(GLenum gl_type, - glsl_base_type base_type, unsigned vector_elements, - unsigned matrix_columns, const char *name) : - gl_type(gl_type), - base_type(base_type), - sampler_dimensionality(0), sampler_shadow(0), sampler_array(0), - sampler_type(0), - vector_elements(vector_elements), matrix_columns(matrix_columns), - length(0) -{ - init_ralloc_type_ctx(); - this->name = ralloc_strdup(this->mem_ctx, name); - /* Neither dimension is zero or both dimensions are zero. - */ - assert((vector_elements == 0) == (matrix_columns == 0)); - memset(& fields, 0, sizeof(fields)); -} - -glsl_type::glsl_type(GLenum gl_type, - enum glsl_sampler_dim dim, bool shadow, bool array, - unsigned type, const char *name) : - gl_type(gl_type), - base_type(GLSL_TYPE_SAMPLER), - sampler_dimensionality(dim), sampler_shadow(shadow), - sampler_array(array), sampler_type(type), - vector_elements(0), matrix_columns(0), - length(0) -{ - init_ralloc_type_ctx(); - this->name = ralloc_strdup(this->mem_ctx, name); - memset(& fields, 0, sizeof(fields)); -} - -glsl_type::glsl_type(const glsl_struct_field *fields, unsigned num_fields, - const char *name) : - base_type(GLSL_TYPE_STRUCT), - sampler_dimensionality(0), sampler_shadow(0), sampler_array(0), - sampler_type(0), - vector_elements(0), matrix_columns(0), - length(num_fields) -{ - unsigned int i; - - init_ralloc_type_ctx(); - this->name = ralloc_strdup(this->mem_ctx, name); - this->fields.structure = ralloc_array(this->mem_ctx, - glsl_struct_field, length); - for (i = 0; i < length; i++) { - this->fields.structure[i].type = fields[i].type; - this->fields.structure[i].name = ralloc_strdup(this->fields.structure, - fields[i].name); - } -} - -static void -add_types_to_symbol_table(glsl_symbol_table *symtab, - const struct glsl_type *types, - unsigned num_types, bool warn) -{ - (void) warn; - - for (unsigned i = 0; i < num_types; i++) { - symtab->add_type(types[i].name, & types[i]); - } -} - -void -glsl_type::generate_100ES_types(glsl_symbol_table *symtab) -{ - add_types_to_symbol_table(symtab, builtin_core_types, - Elements(builtin_core_types), - false); - add_types_to_symbol_table(symtab, builtin_structure_types, - Elements(builtin_structure_types), - false); - add_types_to_symbol_table(symtab, void_type, 1, false); -} - -void -glsl_type::generate_110_types(glsl_symbol_table *symtab) -{ - generate_100ES_types(symtab); - - add_types_to_symbol_table(symtab, builtin_110_types, - Elements(builtin_110_types), - false); - add_types_to_symbol_table(symtab, builtin_110_deprecated_structure_types, - Elements(builtin_110_deprecated_structure_types), - false); -} - - -void -glsl_type::generate_120_types(glsl_symbol_table *symtab) -{ - generate_110_types(symtab); - - add_types_to_symbol_table(symtab, builtin_120_types, - Elements(builtin_120_types), false); -} - - -void -glsl_type::generate_130_types(glsl_symbol_table *symtab) -{ - generate_120_types(symtab); - - add_types_to_symbol_table(symtab, builtin_130_types, - Elements(builtin_130_types), false); - generate_EXT_texture_array_types(symtab, false); -} - - -void -glsl_type::generate_ARB_texture_rectangle_types(glsl_symbol_table *symtab, - bool warn) -{ - add_types_to_symbol_table(symtab, builtin_ARB_texture_rectangle_types, - Elements(builtin_ARB_texture_rectangle_types), - warn); -} - - -void -glsl_type::generate_EXT_texture_array_types(glsl_symbol_table *symtab, - bool warn) -{ - add_types_to_symbol_table(symtab, builtin_EXT_texture_array_types, - Elements(builtin_EXT_texture_array_types), - warn); -} - - -void -_mesa_glsl_initialize_types(struct _mesa_glsl_parse_state *state) -{ - switch (state->language_version) { - case 100: - assert(state->es_shader); - glsl_type::generate_100ES_types(state->symbols); - break; - case 110: - glsl_type::generate_110_types(state->symbols); - break; - case 120: - glsl_type::generate_120_types(state->symbols); - break; - case 130: - glsl_type::generate_130_types(state->symbols); - break; - default: - /* error */ - break; - } - - if (state->ARB_texture_rectangle_enable) { - glsl_type::generate_ARB_texture_rectangle_types(state->symbols, - state->ARB_texture_rectangle_warn); - } - - if (state->EXT_texture_array_enable && state->language_version < 130) { - // These are already included in 130; don't create twice. - glsl_type::generate_EXT_texture_array_types(state->symbols, - state->EXT_texture_array_warn); - } -} - - -const glsl_type *glsl_type::get_base_type() const -{ - switch (base_type) { - case GLSL_TYPE_UINT: - return uint_type; - case GLSL_TYPE_INT: - return int_type; - case GLSL_TYPE_FLOAT: - return float_type; - case GLSL_TYPE_BOOL: - return bool_type; - default: - return error_type; - } -} - - -void -_mesa_glsl_release_types(void) -{ - if (glsl_type::array_types != NULL) { - hash_table_dtor(glsl_type::array_types); - glsl_type::array_types = NULL; - } - - if (glsl_type::record_types != NULL) { - hash_table_dtor(glsl_type::record_types); - glsl_type::record_types = NULL; - } -} - - -glsl_type::glsl_type(const glsl_type *array, unsigned length) : - base_type(GLSL_TYPE_ARRAY), - sampler_dimensionality(0), sampler_shadow(0), sampler_array(0), - sampler_type(0), - vector_elements(0), matrix_columns(0), - name(NULL), length(length) -{ - this->fields.array = array; - /* Inherit the gl type of the base. The GL type is used for - * uniform/statevar handling in Mesa and the arrayness of the type - * is represented by the size rather than the type. - */ - this->gl_type = array->gl_type; - - /* Allow a maximum of 10 characters for the array size. This is enough - * for 32-bits of ~0. The extra 3 are for the '[', ']', and terminating - * NUL. - */ - const unsigned name_length = strlen(array->name) + 10 + 3; - char *const n = (char *) ralloc_size(this->mem_ctx, name_length); - - if (length == 0) - snprintf(n, name_length, "%s[]", array->name); - else - snprintf(n, name_length, "%s[%u]", array->name, length); - - this->name = n; -} - - -const glsl_type * -glsl_type::get_instance(unsigned base_type, unsigned rows, unsigned columns) -{ - if (base_type == GLSL_TYPE_VOID) - return void_type; - - if ((rows < 1) || (rows > 4) || (columns < 1) || (columns > 4)) - return error_type; - - /* Treat GLSL vectors as Nx1 matrices. - */ - if (columns == 1) { - switch (base_type) { - case GLSL_TYPE_UINT: - return uint_type + (rows - 1); - case GLSL_TYPE_INT: - return int_type + (rows - 1); - case GLSL_TYPE_FLOAT: - return float_type + (rows - 1); - case GLSL_TYPE_BOOL: - return bool_type + (rows - 1); - default: - return error_type; - } - } else { - if ((base_type != GLSL_TYPE_FLOAT) || (rows == 1)) - return error_type; - - /* GLSL matrix types are named mat{COLUMNS}x{ROWS}. Only the following - * combinations are valid: - * - * 1 2 3 4 - * 1 - * 2 x x x - * 3 x x x - * 4 x x x - */ -#define IDX(c,r) (((c-1)*3) + (r-1)) - - switch (IDX(columns, rows)) { - case IDX(2,2): return mat2_type; - case IDX(2,3): return mat2x3_type; - case IDX(2,4): return mat2x4_type; - case IDX(3,2): return mat3x2_type; - case IDX(3,3): return mat3_type; - case IDX(3,4): return mat3x4_type; - case IDX(4,2): return mat4x2_type; - case IDX(4,3): return mat4x3_type; - case IDX(4,4): return mat4_type; - default: return error_type; - } - } - - assert(!"Should not get here."); - return error_type; -} - - -const glsl_type * -glsl_type::get_array_instance(const glsl_type *base, unsigned array_size) -{ - - if (array_types == NULL) { - array_types = hash_table_ctor(64, hash_table_string_hash, - hash_table_string_compare); - } - - /* Generate a name using the base type pointer in the key. This is - * done because the name of the base type may not be unique across - * shaders. For example, two shaders may have different record types - * named 'foo'. - */ - char key[128]; - snprintf(key, sizeof(key), "%p[%u]", (void *) base, array_size); - - const glsl_type *t = (glsl_type *) hash_table_find(array_types, key); - if (t == NULL) { - t = new glsl_type(base, array_size); - - hash_table_insert(array_types, (void *) t, ralloc_strdup(mem_ctx, key)); - } - - assert(t->base_type == GLSL_TYPE_ARRAY); - assert(t->length == array_size); - assert(t->fields.array == base); - - return t; -} - - -int -glsl_type::record_key_compare(const void *a, const void *b) -{ - const glsl_type *const key1 = (glsl_type *) a; - const glsl_type *const key2 = (glsl_type *) b; - - /* Return zero is the types match (there is zero difference) or non-zero - * otherwise. - */ - if (strcmp(key1->name, key2->name) != 0) - return 1; - - if (key1->length != key2->length) - return 1; - - for (unsigned i = 0; i < key1->length; i++) { - if (key1->fields.structure[i].type != key2->fields.structure[i].type) - return 1; - if (strcmp(key1->fields.structure[i].name, - key2->fields.structure[i].name) != 0) - return 1; - } - - return 0; -} - - -unsigned -glsl_type::record_key_hash(const void *a) -{ - const glsl_type *const key = (glsl_type *) a; - char hash_key[128]; - unsigned size = 0; - - size = snprintf(hash_key, sizeof(hash_key), "%08x", key->length); - - for (unsigned i = 0; i < key->length; i++) { - if (size >= sizeof(hash_key)) - break; - - size += snprintf(& hash_key[size], sizeof(hash_key) - size, - "%p", (void *) key->fields.structure[i].type); - } - - return hash_table_string_hash(& hash_key); -} - - -const glsl_type * -glsl_type::get_record_instance(const glsl_struct_field *fields, - unsigned num_fields, - const char *name) -{ - const glsl_type key(fields, num_fields, name); - - if (record_types == NULL) { - record_types = hash_table_ctor(64, record_key_hash, record_key_compare); - } - - const glsl_type *t = (glsl_type *) hash_table_find(record_types, & key); - if (t == NULL) { - t = new glsl_type(fields, num_fields, name); - - hash_table_insert(record_types, (void *) t, t); - } - - assert(t->base_type == GLSL_TYPE_STRUCT); - assert(t->length == num_fields); - assert(strcmp(t->name, name) == 0); - - return t; -} - - -const glsl_type * -glsl_type::field_type(const char *name) const -{ - if (this->base_type != GLSL_TYPE_STRUCT) - return error_type; - - for (unsigned i = 0; i < this->length; i++) { - if (strcmp(name, this->fields.structure[i].name) == 0) - return this->fields.structure[i].type; - } - - return error_type; -} - - -int -glsl_type::field_index(const char *name) const -{ - if (this->base_type != GLSL_TYPE_STRUCT) - return -1; - - for (unsigned i = 0; i < this->length; i++) { - if (strcmp(name, this->fields.structure[i].name) == 0) - return i; - } - - return -1; -} - - -unsigned -glsl_type::component_slots() const -{ - switch (this->base_type) { - case GLSL_TYPE_UINT: - case GLSL_TYPE_INT: - case GLSL_TYPE_FLOAT: - case GLSL_TYPE_BOOL: - return this->components(); - - case GLSL_TYPE_STRUCT: { - unsigned size = 0; - - for (unsigned i = 0; i < this->length; i++) - size += this->fields.structure[i].type->component_slots(); - - return size; - } - - case GLSL_TYPE_ARRAY: - return this->length * this->fields.array->component_slots(); - - default: - return 0; - } -} +/*
+ * Copyright © 2009 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include "main/core.h" /* for Elements */
+#include "glsl_symbol_table.h"
+#include "glsl_parser_extras.h"
+#include "glsl_types.h"
+#include "builtin_types.h"
+extern "C" {
+#include "program/hash_table.h"
+}
+
+hash_table *glsl_type::array_types = NULL;
+hash_table *glsl_type::record_types = NULL;
+void *glsl_type::mem_ctx = NULL;
+
+void
+glsl_type::init_ralloc_type_ctx(void)
+{
+ if (glsl_type::mem_ctx == NULL) {
+ glsl_type::mem_ctx = ralloc_autofree_context();
+ assert(glsl_type::mem_ctx != NULL);
+ }
+}
+
+glsl_type::glsl_type(GLenum gl_type,
+ glsl_base_type base_type, unsigned vector_elements,
+ unsigned matrix_columns, const char *name) :
+ gl_type(gl_type),
+ base_type(base_type),
+ sampler_dimensionality(0), sampler_shadow(0), sampler_array(0),
+ sampler_type(0),
+ vector_elements(vector_elements), matrix_columns(matrix_columns),
+ length(0)
+{
+ init_ralloc_type_ctx();
+ this->name = ralloc_strdup(this->mem_ctx, name);
+ /* Neither dimension is zero or both dimensions are zero.
+ */
+ assert((vector_elements == 0) == (matrix_columns == 0));
+ memset(& fields, 0, sizeof(fields));
+}
+
+glsl_type::glsl_type(GLenum gl_type,
+ enum glsl_sampler_dim dim, bool shadow, bool array,
+ unsigned type, const char *name) :
+ gl_type(gl_type),
+ base_type(GLSL_TYPE_SAMPLER),
+ sampler_dimensionality(dim), sampler_shadow(shadow),
+ sampler_array(array), sampler_type(type),
+ vector_elements(0), matrix_columns(0),
+ length(0)
+{
+ init_ralloc_type_ctx();
+ this->name = ralloc_strdup(this->mem_ctx, name);
+ memset(& fields, 0, sizeof(fields));
+}
+
+glsl_type::glsl_type(const glsl_struct_field *fields, unsigned num_fields,
+ const char *name) :
+ base_type(GLSL_TYPE_STRUCT),
+ sampler_dimensionality(0), sampler_shadow(0), sampler_array(0),
+ sampler_type(0),
+ vector_elements(0), matrix_columns(0),
+ length(num_fields)
+{
+ unsigned int i;
+
+ init_ralloc_type_ctx();
+ this->name = ralloc_strdup(this->mem_ctx, name);
+ this->fields.structure = ralloc_array(this->mem_ctx,
+ glsl_struct_field, length);
+ for (i = 0; i < length; i++) {
+ this->fields.structure[i].type = fields[i].type;
+ this->fields.structure[i].name = ralloc_strdup(this->fields.structure,
+ fields[i].name);
+ }
+}
+
+static void
+add_types_to_symbol_table(glsl_symbol_table *symtab,
+ const struct glsl_type *types,
+ unsigned num_types, bool warn)
+{
+ (void) warn;
+
+ for (unsigned i = 0; i < num_types; i++) {
+ symtab->add_type(types[i].name, & types[i]);
+ }
+}
+
+void
+glsl_type::generate_100ES_types(glsl_symbol_table *symtab)
+{
+ add_types_to_symbol_table(symtab, builtin_core_types,
+ Elements(builtin_core_types),
+ false);
+ add_types_to_symbol_table(symtab, builtin_structure_types,
+ Elements(builtin_structure_types),
+ false);
+ add_types_to_symbol_table(symtab, void_type, 1, false);
+}
+
+void
+glsl_type::generate_110_types(glsl_symbol_table *symtab)
+{
+ generate_100ES_types(symtab);
+
+ add_types_to_symbol_table(symtab, builtin_110_types,
+ Elements(builtin_110_types),
+ false);
+ add_types_to_symbol_table(symtab, builtin_110_deprecated_structure_types,
+ Elements(builtin_110_deprecated_structure_types),
+ false);
+}
+
+
+void
+glsl_type::generate_120_types(glsl_symbol_table *symtab)
+{
+ generate_110_types(symtab);
+
+ add_types_to_symbol_table(symtab, builtin_120_types,
+ Elements(builtin_120_types), false);
+}
+
+
+void
+glsl_type::generate_130_types(glsl_symbol_table *symtab)
+{
+ generate_120_types(symtab);
+
+ add_types_to_symbol_table(symtab, builtin_130_types,
+ Elements(builtin_130_types), false);
+ generate_EXT_texture_array_types(symtab, false);
+}
+
+
+void
+glsl_type::generate_ARB_texture_rectangle_types(glsl_symbol_table *symtab,
+ bool warn)
+{
+ add_types_to_symbol_table(symtab, builtin_ARB_texture_rectangle_types,
+ Elements(builtin_ARB_texture_rectangle_types),
+ warn);
+}
+
+
+void
+glsl_type::generate_EXT_texture_array_types(glsl_symbol_table *symtab,
+ bool warn)
+{
+ add_types_to_symbol_table(symtab, builtin_EXT_texture_array_types,
+ Elements(builtin_EXT_texture_array_types),
+ warn);
+}
+
+
+void
+_mesa_glsl_initialize_types(struct _mesa_glsl_parse_state *state)
+{
+ switch (state->language_version) {
+ case 100:
+ assert(state->es_shader);
+ glsl_type::generate_100ES_types(state->symbols);
+ break;
+ case 110:
+ glsl_type::generate_110_types(state->symbols);
+ break;
+ case 120:
+ glsl_type::generate_120_types(state->symbols);
+ break;
+ case 130:
+ glsl_type::generate_130_types(state->symbols);
+ break;
+ default:
+ /* error */
+ break;
+ }
+
+ if (state->ARB_texture_rectangle_enable) {
+ glsl_type::generate_ARB_texture_rectangle_types(state->symbols,
+ state->ARB_texture_rectangle_warn);
+ }
+
+ if (state->EXT_texture_array_enable && state->language_version < 130) {
+ // These are already included in 130; don't create twice.
+ glsl_type::generate_EXT_texture_array_types(state->symbols,
+ state->EXT_texture_array_warn);
+ }
+}
+
+
+const glsl_type *glsl_type::get_base_type() const
+{
+ switch (base_type) {
+ case GLSL_TYPE_UINT:
+ return uint_type;
+ case GLSL_TYPE_INT:
+ return int_type;
+ case GLSL_TYPE_FLOAT:
+ return float_type;
+ case GLSL_TYPE_BOOL:
+ return bool_type;
+ default:
+ return error_type;
+ }
+}
+
+
+void
+_mesa_glsl_release_types(void)
+{
+ if (glsl_type::array_types != NULL) {
+ hash_table_dtor(glsl_type::array_types);
+ glsl_type::array_types = NULL;
+ }
+
+ if (glsl_type::record_types != NULL) {
+ hash_table_dtor(glsl_type::record_types);
+ glsl_type::record_types = NULL;
+ }
+}
+
+
+glsl_type::glsl_type(const glsl_type *array, unsigned length) :
+ base_type(GLSL_TYPE_ARRAY),
+ sampler_dimensionality(0), sampler_shadow(0), sampler_array(0),
+ sampler_type(0),
+ vector_elements(0), matrix_columns(0),
+ name(NULL), length(length)
+{
+ this->fields.array = array;
+ /* Inherit the gl type of the base. The GL type is used for
+ * uniform/statevar handling in Mesa and the arrayness of the type
+ * is represented by the size rather than the type.
+ */
+ this->gl_type = array->gl_type;
+
+ /* Allow a maximum of 10 characters for the array size. This is enough
+ * for 32-bits of ~0. The extra 3 are for the '[', ']', and terminating
+ * NUL.
+ */
+ const unsigned name_length = strlen(array->name) + 10 + 3;
+ char *const n = (char *) ralloc_size(this->mem_ctx, name_length);
+
+ if (length == 0)
+ snprintf(n, name_length, "%s[]", array->name);
+ else
+ snprintf(n, name_length, "%s[%u]", array->name, length);
+
+ this->name = n;
+}
+
+
+const glsl_type *
+glsl_type::get_instance(unsigned base_type, unsigned rows, unsigned columns)
+{
+ if (base_type == GLSL_TYPE_VOID)
+ return void_type;
+
+ if ((rows < 1) || (rows > 4) || (columns < 1) || (columns > 4))
+ return error_type;
+
+ /* Treat GLSL vectors as Nx1 matrices.
+ */
+ if (columns == 1) {
+ switch (base_type) {
+ case GLSL_TYPE_UINT:
+ return uint_type + (rows - 1);
+ case GLSL_TYPE_INT:
+ return int_type + (rows - 1);
+ case GLSL_TYPE_FLOAT:
+ return float_type + (rows - 1);
+ case GLSL_TYPE_BOOL:
+ return bool_type + (rows - 1);
+ default:
+ return error_type;
+ }
+ } else {
+ if ((base_type != GLSL_TYPE_FLOAT) || (rows == 1))
+ return error_type;
+
+ /* GLSL matrix types are named mat{COLUMNS}x{ROWS}. Only the following
+ * combinations are valid:
+ *
+ * 1 2 3 4
+ * 1
+ * 2 x x x
+ * 3 x x x
+ * 4 x x x
+ */
+#define IDX(c,r) (((c-1)*3) + (r-1))
+
+ switch (IDX(columns, rows)) {
+ case IDX(2,2): return mat2_type;
+ case IDX(2,3): return mat2x3_type;
+ case IDX(2,4): return mat2x4_type;
+ case IDX(3,2): return mat3x2_type;
+ case IDX(3,3): return mat3_type;
+ case IDX(3,4): return mat3x4_type;
+ case IDX(4,2): return mat4x2_type;
+ case IDX(4,3): return mat4x3_type;
+ case IDX(4,4): return mat4_type;
+ default: return error_type;
+ }
+ }
+
+ assert(!"Should not get here.");
+ return error_type;
+}
+
+
+const glsl_type *
+glsl_type::get_array_instance(const glsl_type *base, unsigned array_size)
+{
+
+ if (array_types == NULL) {
+ array_types = hash_table_ctor(64, hash_table_string_hash,
+ hash_table_string_compare);
+ }
+
+ /* Generate a name using the base type pointer in the key. This is
+ * done because the name of the base type may not be unique across
+ * shaders. For example, two shaders may have different record types
+ * named 'foo'.
+ */
+ char key[128];
+ snprintf(key, sizeof(key), "%p[%u]", (void *) base, array_size);
+
+ const glsl_type *t = (glsl_type *) hash_table_find(array_types, key);
+ if (t == NULL) {
+ t = new glsl_type(base, array_size);
+
+ hash_table_insert(array_types, (void *) t, ralloc_strdup(mem_ctx, key));
+ }
+
+ assert(t->base_type == GLSL_TYPE_ARRAY);
+ assert(t->length == array_size);
+ assert(t->fields.array == base);
+
+ return t;
+}
+
+
+int
+glsl_type::record_key_compare(const void *a, const void *b)
+{
+ const glsl_type *const key1 = (glsl_type *) a;
+ const glsl_type *const key2 = (glsl_type *) b;
+
+ /* Return zero is the types match (there is zero difference) or non-zero
+ * otherwise.
+ */
+ if (strcmp(key1->name, key2->name) != 0)
+ return 1;
+
+ if (key1->length != key2->length)
+ return 1;
+
+ for (unsigned i = 0; i < key1->length; i++) {
+ if (key1->fields.structure[i].type != key2->fields.structure[i].type)
+ return 1;
+ if (strcmp(key1->fields.structure[i].name,
+ key2->fields.structure[i].name) != 0)
+ return 1;
+ }
+
+ return 0;
+}
+
+
+unsigned
+glsl_type::record_key_hash(const void *a)
+{
+ const glsl_type *const key = (glsl_type *) a;
+ char hash_key[128];
+ unsigned size = 0;
+
+ size = snprintf(hash_key, sizeof(hash_key), "%08x", key->length);
+
+ for (unsigned i = 0; i < key->length; i++) {
+ if (size >= sizeof(hash_key))
+ break;
+
+ size += snprintf(& hash_key[size], sizeof(hash_key) - size,
+ "%p", (void *) key->fields.structure[i].type);
+ }
+
+ return hash_table_string_hash(& hash_key);
+}
+
+
+const glsl_type *
+glsl_type::get_record_instance(const glsl_struct_field *fields,
+ unsigned num_fields,
+ const char *name)
+{
+ const glsl_type key(fields, num_fields, name);
+
+ if (record_types == NULL) {
+ record_types = hash_table_ctor(64, record_key_hash, record_key_compare);
+ }
+
+ const glsl_type *t = (glsl_type *) hash_table_find(record_types, & key);
+ if (t == NULL) {
+ t = new glsl_type(fields, num_fields, name);
+
+ hash_table_insert(record_types, (void *) t, t);
+ }
+
+ assert(t->base_type == GLSL_TYPE_STRUCT);
+ assert(t->length == num_fields);
+ assert(strcmp(t->name, name) == 0);
+
+ return t;
+}
+
+
+const glsl_type *
+glsl_type::field_type(const char *name) const
+{
+ if (this->base_type != GLSL_TYPE_STRUCT)
+ return error_type;
+
+ for (unsigned i = 0; i < this->length; i++) {
+ if (strcmp(name, this->fields.structure[i].name) == 0)
+ return this->fields.structure[i].type;
+ }
+
+ return error_type;
+}
+
+
+int
+glsl_type::field_index(const char *name) const
+{
+ if (this->base_type != GLSL_TYPE_STRUCT)
+ return -1;
+
+ for (unsigned i = 0; i < this->length; i++) {
+ if (strcmp(name, this->fields.structure[i].name) == 0)
+ return i;
+ }
+
+ return -1;
+}
+
+
+unsigned
+glsl_type::component_slots() const
+{
+ switch (this->base_type) {
+ case GLSL_TYPE_UINT:
+ case GLSL_TYPE_INT:
+ case GLSL_TYPE_FLOAT:
+ case GLSL_TYPE_BOOL:
+ return this->components();
+
+ case GLSL_TYPE_STRUCT: {
+ unsigned size = 0;
+
+ for (unsigned i = 0; i < this->length; i++)
+ size += this->fields.structure[i].type->component_slots();
+
+ return size;
+ }
+
+ case GLSL_TYPE_ARRAY:
+ return this->length * this->fields.array->component_slots();
+
+ default:
+ return 0;
+ }
+}
diff --git a/mesalib/src/glsl/glsl_types.h b/mesalib/src/glsl/glsl_types.h index ab6858faa..c3b7255a3 100644 --- a/mesalib/src/glsl/glsl_types.h +++ b/mesalib/src/glsl/glsl_types.h @@ -1,476 +1,476 @@ -/* -*- c++ -*- */ -/* - * Copyright © 2009 Intel Corporation - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING - * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER - * DEALINGS IN THE SOFTWARE. - */ - -#pragma once -#ifndef GLSL_TYPES_H -#define GLSL_TYPES_H - -#include <cstring> -#include <cassert> - -extern "C" { -#include "GL/gl.h" -} - -#include "ralloc.h" - -struct _mesa_glsl_parse_state; -struct glsl_symbol_table; - -extern "C" void -_mesa_glsl_initialize_types(struct _mesa_glsl_parse_state *state); - -extern "C" void -_mesa_glsl_release_types(void); - -enum glsl_base_type { - GLSL_TYPE_UINT = 0, - GLSL_TYPE_INT, - GLSL_TYPE_FLOAT, - GLSL_TYPE_BOOL, - GLSL_TYPE_SAMPLER, - GLSL_TYPE_STRUCT, - GLSL_TYPE_ARRAY, - GLSL_TYPE_VOID, - GLSL_TYPE_ERROR -}; - -enum glsl_sampler_dim { - GLSL_SAMPLER_DIM_1D = 0, - GLSL_SAMPLER_DIM_2D, - GLSL_SAMPLER_DIM_3D, - GLSL_SAMPLER_DIM_CUBE, - GLSL_SAMPLER_DIM_RECT, - GLSL_SAMPLER_DIM_BUF -}; - - -struct glsl_type { - GLenum gl_type; - glsl_base_type base_type; - - unsigned sampler_dimensionality:3; - unsigned sampler_shadow:1; - unsigned sampler_array:1; - unsigned sampler_type:2; /**< Type of data returned using this sampler. - * only \c GLSL_TYPE_FLOAT, \c GLSL_TYPE_INT, - * and \c GLSL_TYPE_UINT are valid. - */ - - /* Callers of this ralloc-based new need not call delete. It's - * easier to just ralloc_free 'mem_ctx' (or any of its ancestors). */ - static void* operator new(size_t size) - { - if (glsl_type::mem_ctx == NULL) { - glsl_type::mem_ctx = ralloc_context(NULL); - assert(glsl_type::mem_ctx != NULL); - } - - void *type; - - type = ralloc_size(glsl_type::mem_ctx, size); - assert(type != NULL); - - return type; - } - - /* If the user *does* call delete, that's OK, we will just - * ralloc_free in that case. */ - static void operator delete(void *type) - { - ralloc_free(type); - } - - /** - * \name Vector and matrix element counts - * - * For scalars, each of these values will be 1. For non-numeric types - * these will be 0. - */ - /*@{*/ - unsigned vector_elements:3; /**< 1, 2, 3, or 4 vector elements. */ - unsigned matrix_columns:3; /**< 1, 2, 3, or 4 matrix columns. */ - /*@}*/ - - /** - * Name of the data type - * - * This may be \c NULL for anonymous structures, for arrays, or for - * function types. - */ - const char *name; - - /** - * For \c GLSL_TYPE_ARRAY, this is the length of the array. For - * \c GLSL_TYPE_STRUCT, it is the number of elements in the structure and - * the number of values pointed to by \c fields.structure (below). - */ - unsigned length; - - /** - * Subtype of composite data types. - */ - union { - const struct glsl_type *array; /**< Type of array elements. */ - const struct glsl_type *parameters; /**< Parameters to function. */ - struct glsl_struct_field *structure; /**< List of struct fields. */ - } fields; - - - /** - * \name Pointers to various public type singletons - */ - /*@{*/ - static const glsl_type *const error_type; - static const glsl_type *const void_type; - static const glsl_type *const int_type; - static const glsl_type *const ivec4_type; - static const glsl_type *const uint_type; - static const glsl_type *const uvec2_type; - static const glsl_type *const uvec3_type; - static const glsl_type *const uvec4_type; - static const glsl_type *const float_type; - static const glsl_type *const vec2_type; - static const glsl_type *const vec3_type; - static const glsl_type *const vec4_type; - static const glsl_type *const bool_type; - static const glsl_type *const mat2_type; - static const glsl_type *const mat2x3_type; - static const glsl_type *const mat2x4_type; - static const glsl_type *const mat3x2_type; - static const glsl_type *const mat3_type; - static const glsl_type *const mat3x4_type; - static const glsl_type *const mat4x2_type; - static const glsl_type *const mat4x3_type; - static const glsl_type *const mat4_type; - /*@}*/ - - - /** - * For numeric and boolean derrived types returns the basic scalar type - * - * If the type is a numeric or boolean scalar, vector, or matrix type, - * this function gets the scalar type of the individual components. For - * all other types, including arrays of numeric or boolean types, the - * error type is returned. - */ - const glsl_type *get_base_type() const; - - /** - * Query the type of elements in an array - * - * \return - * Pointer to the type of elements in the array for array types, or \c NULL - * for non-array types. - */ - const glsl_type *element_type() const - { - return is_array() ? fields.array : NULL; - } - - /** - * Get the instance of a built-in scalar, vector, or matrix type - */ - static const glsl_type *get_instance(unsigned base_type, unsigned rows, - unsigned columns); - - /** - * Get the instance of an array type - */ - static const glsl_type *get_array_instance(const glsl_type *base, - unsigned elements); - - /** - * Get the instance of a record type - */ - static const glsl_type *get_record_instance(const glsl_struct_field *fields, - unsigned num_fields, - const char *name); - - /** - * Query the total number of scalars that make up a scalar, vector or matrix - */ - unsigned components() const - { - return vector_elements * matrix_columns; - } - - /** - * Calculate the number of components slots required to hold this type - * - * This is used to determine how many uniform or varying locations a type - * might occupy. - */ - unsigned component_slots() const; - - - /** - * Query whether or not a type is a scalar (non-vector and non-matrix). - */ - bool is_scalar() const - { - return (vector_elements == 1) - && (base_type >= GLSL_TYPE_UINT) - && (base_type <= GLSL_TYPE_BOOL); - } - - /** - * Query whether or not a type is a vector - */ - bool is_vector() const - { - return (vector_elements > 1) - && (matrix_columns == 1) - && (base_type >= GLSL_TYPE_UINT) - && (base_type <= GLSL_TYPE_BOOL); - } - - /** - * Query whether or not a type is a matrix - */ - bool is_matrix() const - { - /* GLSL only has float matrices. */ - return (matrix_columns > 1) && (base_type == GLSL_TYPE_FLOAT); - } - - /** - * Query whether or not a type is a non-array numeric type - */ - bool is_numeric() const - { - return (base_type >= GLSL_TYPE_UINT) && (base_type <= GLSL_TYPE_FLOAT); - } - - /** - * Query whether or not a type is an integral type - */ - bool is_integer() const - { - return (base_type == GLSL_TYPE_UINT) || (base_type == GLSL_TYPE_INT); - } - - /** - * Query whether or not a type is a float type - */ - bool is_float() const - { - return base_type == GLSL_TYPE_FLOAT; - } - - /** - * Query whether or not a type is a non-array boolean type - */ - bool is_boolean() const - { - return base_type == GLSL_TYPE_BOOL; - } - - /** - * Query whether or not a type is a sampler - */ - bool is_sampler() const - { - return base_type == GLSL_TYPE_SAMPLER; - } - - /** - * Query whether or not a type is an array - */ - bool is_array() const - { - return base_type == GLSL_TYPE_ARRAY; - } - - /** - * Query whether or not a type is a record - */ - bool is_record() const - { - return base_type == GLSL_TYPE_STRUCT; - } - - /** - * Query whether or not a type is the void type singleton. - */ - bool is_void() const - { - return base_type == GLSL_TYPE_VOID; - } - - /** - * Query whether or not a type is the error type singleton. - */ - bool is_error() const - { - return base_type == GLSL_TYPE_ERROR; - } - - /** - * Query the full type of a matrix row - * - * \return - * If the type is not a matrix, \c glsl_type::error_type is returned. - * Otherwise a type matching the rows of the matrix is returned. - */ - const glsl_type *row_type() const - { - return is_matrix() - ? get_instance(base_type, matrix_columns, 1) - : error_type; - } - - /** - * Query the full type of a matrix column - * - * \return - * If the type is not a matrix, \c glsl_type::error_type is returned. - * Otherwise a type matching the columns of the matrix is returned. - */ - const glsl_type *column_type() const - { - return is_matrix() - ? get_instance(base_type, vector_elements, 1) - : error_type; - } - - - /** - * Get the type of a structure field - * - * \return - * Pointer to the type of the named field. If the type is not a structure - * or the named field does not exist, \c glsl_type::error_type is returned. - */ - const glsl_type *field_type(const char *name) const; - - - /** - * Get the location of a filed within a record type - */ - int field_index(const char *name) const; - - - /** - * Query the number of elements in an array type - * - * \return - * The number of elements in the array for array types or -1 for non-array - * types. If the number of elements in the array has not yet been declared, - * zero is returned. - */ - int array_size() const - { - return is_array() ? length : -1; - } - -private: - /** - * ralloc context for all glsl_type allocations - * - * Set on the first call to \c glsl_type::new. - */ - static void *mem_ctx; - - void init_ralloc_type_ctx(void); - - /** Constructor for vector and matrix types */ - glsl_type(GLenum gl_type, - glsl_base_type base_type, unsigned vector_elements, - unsigned matrix_columns, const char *name); - - /** Constructor for sampler types */ - glsl_type(GLenum gl_type, - enum glsl_sampler_dim dim, bool shadow, bool array, - unsigned type, const char *name); - - /** Constructor for record types */ - glsl_type(const glsl_struct_field *fields, unsigned num_fields, - const char *name); - - /** Constructor for array types */ - glsl_type(const glsl_type *array, unsigned length); - - /** Hash table containing the known array types. */ - static struct hash_table *array_types; - - /** Hash table containing the known record types. */ - static struct hash_table *record_types; - - static int record_key_compare(const void *a, const void *b); - static unsigned record_key_hash(const void *key); - - /** - * \name Pointers to various type singletons - */ - /*@{*/ - static const glsl_type _error_type; - static const glsl_type _void_type; - static const glsl_type builtin_core_types[]; - static const glsl_type builtin_structure_types[]; - static const glsl_type builtin_110_deprecated_structure_types[]; - static const glsl_type builtin_110_types[]; - static const glsl_type builtin_120_types[]; - static const glsl_type builtin_130_types[]; - static const glsl_type builtin_ARB_texture_rectangle_types[]; - static const glsl_type builtin_EXT_texture_array_types[]; - static const glsl_type builtin_EXT_texture_buffer_object_types[]; - /*@}*/ - - /** - * \name Methods to populate a symbol table with built-in types. - * - * \internal - * This is one of the truely annoying things about C++. Methods that are - * completely internal and private to a type still have to be advertised to - * the world in a public header file. - */ - /*@{*/ - static void generate_100ES_types(glsl_symbol_table *); - static void generate_110_types(glsl_symbol_table *); - static void generate_120_types(glsl_symbol_table *); - static void generate_130_types(glsl_symbol_table *); - static void generate_ARB_texture_rectangle_types(glsl_symbol_table *, bool); - static void generate_EXT_texture_array_types(glsl_symbol_table *, bool); - /*@}*/ - - /** - * \name Friend functions. - * - * These functions are friends because they must have C linkage and the - * need to call various private methods or access various private static - * data. - */ - /*@{*/ - friend void _mesa_glsl_initialize_types(struct _mesa_glsl_parse_state *); - friend void _mesa_glsl_release_types(void); - /*@}*/ -}; - -struct glsl_struct_field { - const struct glsl_type *type; - const char *name; -}; - -#endif /* GLSL_TYPES_H */ +/* -*- c++ -*- */
+/*
+ * Copyright © 2009 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#pragma once
+#ifndef GLSL_TYPES_H
+#define GLSL_TYPES_H
+
+#include <string.h>
+#include <assert.h>
+
+extern "C" {
+#include "GL/gl.h"
+}
+
+#include "ralloc.h"
+
+struct _mesa_glsl_parse_state;
+struct glsl_symbol_table;
+
+extern "C" void
+_mesa_glsl_initialize_types(struct _mesa_glsl_parse_state *state);
+
+extern "C" void
+_mesa_glsl_release_types(void);
+
+enum glsl_base_type {
+ GLSL_TYPE_UINT = 0,
+ GLSL_TYPE_INT,
+ GLSL_TYPE_FLOAT,
+ GLSL_TYPE_BOOL,
+ GLSL_TYPE_SAMPLER,
+ GLSL_TYPE_STRUCT,
+ GLSL_TYPE_ARRAY,
+ GLSL_TYPE_VOID,
+ GLSL_TYPE_ERROR
+};
+
+enum glsl_sampler_dim {
+ GLSL_SAMPLER_DIM_1D = 0,
+ GLSL_SAMPLER_DIM_2D,
+ GLSL_SAMPLER_DIM_3D,
+ GLSL_SAMPLER_DIM_CUBE,
+ GLSL_SAMPLER_DIM_RECT,
+ GLSL_SAMPLER_DIM_BUF
+};
+
+
+struct glsl_type {
+ GLenum gl_type;
+ glsl_base_type base_type;
+
+ unsigned sampler_dimensionality:3;
+ unsigned sampler_shadow:1;
+ unsigned sampler_array:1;
+ unsigned sampler_type:2; /**< Type of data returned using this sampler.
+ * only \c GLSL_TYPE_FLOAT, \c GLSL_TYPE_INT,
+ * and \c GLSL_TYPE_UINT are valid.
+ */
+
+ /* Callers of this ralloc-based new need not call delete. It's
+ * easier to just ralloc_free 'mem_ctx' (or any of its ancestors). */
+ static void* operator new(size_t size)
+ {
+ if (glsl_type::mem_ctx == NULL) {
+ glsl_type::mem_ctx = ralloc_context(NULL);
+ assert(glsl_type::mem_ctx != NULL);
+ }
+
+ void *type;
+
+ type = ralloc_size(glsl_type::mem_ctx, size);
+ assert(type != NULL);
+
+ return type;
+ }
+
+ /* If the user *does* call delete, that's OK, we will just
+ * ralloc_free in that case. */
+ static void operator delete(void *type)
+ {
+ ralloc_free(type);
+ }
+
+ /**
+ * \name Vector and matrix element counts
+ *
+ * For scalars, each of these values will be 1. For non-numeric types
+ * these will be 0.
+ */
+ /*@{*/
+ unsigned vector_elements:3; /**< 1, 2, 3, or 4 vector elements. */
+ unsigned matrix_columns:3; /**< 1, 2, 3, or 4 matrix columns. */
+ /*@}*/
+
+ /**
+ * Name of the data type
+ *
+ * This may be \c NULL for anonymous structures, for arrays, or for
+ * function types.
+ */
+ const char *name;
+
+ /**
+ * For \c GLSL_TYPE_ARRAY, this is the length of the array. For
+ * \c GLSL_TYPE_STRUCT, it is the number of elements in the structure and
+ * the number of values pointed to by \c fields.structure (below).
+ */
+ unsigned length;
+
+ /**
+ * Subtype of composite data types.
+ */
+ union {
+ const struct glsl_type *array; /**< Type of array elements. */
+ const struct glsl_type *parameters; /**< Parameters to function. */
+ struct glsl_struct_field *structure; /**< List of struct fields. */
+ } fields;
+
+
+ /**
+ * \name Pointers to various public type singletons
+ */
+ /*@{*/
+ static const glsl_type *const error_type;
+ static const glsl_type *const void_type;
+ static const glsl_type *const int_type;
+ static const glsl_type *const ivec4_type;
+ static const glsl_type *const uint_type;
+ static const glsl_type *const uvec2_type;
+ static const glsl_type *const uvec3_type;
+ static const glsl_type *const uvec4_type;
+ static const glsl_type *const float_type;
+ static const glsl_type *const vec2_type;
+ static const glsl_type *const vec3_type;
+ static const glsl_type *const vec4_type;
+ static const glsl_type *const bool_type;
+ static const glsl_type *const mat2_type;
+ static const glsl_type *const mat2x3_type;
+ static const glsl_type *const mat2x4_type;
+ static const glsl_type *const mat3x2_type;
+ static const glsl_type *const mat3_type;
+ static const glsl_type *const mat3x4_type;
+ static const glsl_type *const mat4x2_type;
+ static const glsl_type *const mat4x3_type;
+ static const glsl_type *const mat4_type;
+ /*@}*/
+
+
+ /**
+ * For numeric and boolean derrived types returns the basic scalar type
+ *
+ * If the type is a numeric or boolean scalar, vector, or matrix type,
+ * this function gets the scalar type of the individual components. For
+ * all other types, including arrays of numeric or boolean types, the
+ * error type is returned.
+ */
+ const glsl_type *get_base_type() const;
+
+ /**
+ * Query the type of elements in an array
+ *
+ * \return
+ * Pointer to the type of elements in the array for array types, or \c NULL
+ * for non-array types.
+ */
+ const glsl_type *element_type() const
+ {
+ return is_array() ? fields.array : NULL;
+ }
+
+ /**
+ * Get the instance of a built-in scalar, vector, or matrix type
+ */
+ static const glsl_type *get_instance(unsigned base_type, unsigned rows,
+ unsigned columns);
+
+ /**
+ * Get the instance of an array type
+ */
+ static const glsl_type *get_array_instance(const glsl_type *base,
+ unsigned elements);
+
+ /**
+ * Get the instance of a record type
+ */
+ static const glsl_type *get_record_instance(const glsl_struct_field *fields,
+ unsigned num_fields,
+ const char *name);
+
+ /**
+ * Query the total number of scalars that make up a scalar, vector or matrix
+ */
+ unsigned components() const
+ {
+ return vector_elements * matrix_columns;
+ }
+
+ /**
+ * Calculate the number of components slots required to hold this type
+ *
+ * This is used to determine how many uniform or varying locations a type
+ * might occupy.
+ */
+ unsigned component_slots() const;
+
+
+ /**
+ * Query whether or not a type is a scalar (non-vector and non-matrix).
+ */
+ bool is_scalar() const
+ {
+ return (vector_elements == 1)
+ && (base_type >= GLSL_TYPE_UINT)
+ && (base_type <= GLSL_TYPE_BOOL);
+ }
+
+ /**
+ * Query whether or not a type is a vector
+ */
+ bool is_vector() const
+ {
+ return (vector_elements > 1)
+ && (matrix_columns == 1)
+ && (base_type >= GLSL_TYPE_UINT)
+ && (base_type <= GLSL_TYPE_BOOL);
+ }
+
+ /**
+ * Query whether or not a type is a matrix
+ */
+ bool is_matrix() const
+ {
+ /* GLSL only has float matrices. */
+ return (matrix_columns > 1) && (base_type == GLSL_TYPE_FLOAT);
+ }
+
+ /**
+ * Query whether or not a type is a non-array numeric type
+ */
+ bool is_numeric() const
+ {
+ return (base_type >= GLSL_TYPE_UINT) && (base_type <= GLSL_TYPE_FLOAT);
+ }
+
+ /**
+ * Query whether or not a type is an integral type
+ */
+ bool is_integer() const
+ {
+ return (base_type == GLSL_TYPE_UINT) || (base_type == GLSL_TYPE_INT);
+ }
+
+ /**
+ * Query whether or not a type is a float type
+ */
+ bool is_float() const
+ {
+ return base_type == GLSL_TYPE_FLOAT;
+ }
+
+ /**
+ * Query whether or not a type is a non-array boolean type
+ */
+ bool is_boolean() const
+ {
+ return base_type == GLSL_TYPE_BOOL;
+ }
+
+ /**
+ * Query whether or not a type is a sampler
+ */
+ bool is_sampler() const
+ {
+ return base_type == GLSL_TYPE_SAMPLER;
+ }
+
+ /**
+ * Query whether or not a type is an array
+ */
+ bool is_array() const
+ {
+ return base_type == GLSL_TYPE_ARRAY;
+ }
+
+ /**
+ * Query whether or not a type is a record
+ */
+ bool is_record() const
+ {
+ return base_type == GLSL_TYPE_STRUCT;
+ }
+
+ /**
+ * Query whether or not a type is the void type singleton.
+ */
+ bool is_void() const
+ {
+ return base_type == GLSL_TYPE_VOID;
+ }
+
+ /**
+ * Query whether or not a type is the error type singleton.
+ */
+ bool is_error() const
+ {
+ return base_type == GLSL_TYPE_ERROR;
+ }
+
+ /**
+ * Query the full type of a matrix row
+ *
+ * \return
+ * If the type is not a matrix, \c glsl_type::error_type is returned.
+ * Otherwise a type matching the rows of the matrix is returned.
+ */
+ const glsl_type *row_type() const
+ {
+ return is_matrix()
+ ? get_instance(base_type, matrix_columns, 1)
+ : error_type;
+ }
+
+ /**
+ * Query the full type of a matrix column
+ *
+ * \return
+ * If the type is not a matrix, \c glsl_type::error_type is returned.
+ * Otherwise a type matching the columns of the matrix is returned.
+ */
+ const glsl_type *column_type() const
+ {
+ return is_matrix()
+ ? get_instance(base_type, vector_elements, 1)
+ : error_type;
+ }
+
+
+ /**
+ * Get the type of a structure field
+ *
+ * \return
+ * Pointer to the type of the named field. If the type is not a structure
+ * or the named field does not exist, \c glsl_type::error_type is returned.
+ */
+ const glsl_type *field_type(const char *name) const;
+
+
+ /**
+ * Get the location of a filed within a record type
+ */
+ int field_index(const char *name) const;
+
+
+ /**
+ * Query the number of elements in an array type
+ *
+ * \return
+ * The number of elements in the array for array types or -1 for non-array
+ * types. If the number of elements in the array has not yet been declared,
+ * zero is returned.
+ */
+ int array_size() const
+ {
+ return is_array() ? length : -1;
+ }
+
+private:
+ /**
+ * ralloc context for all glsl_type allocations
+ *
+ * Set on the first call to \c glsl_type::new.
+ */
+ static void *mem_ctx;
+
+ void init_ralloc_type_ctx(void);
+
+ /** Constructor for vector and matrix types */
+ glsl_type(GLenum gl_type,
+ glsl_base_type base_type, unsigned vector_elements,
+ unsigned matrix_columns, const char *name);
+
+ /** Constructor for sampler types */
+ glsl_type(GLenum gl_type,
+ enum glsl_sampler_dim dim, bool shadow, bool array,
+ unsigned type, const char *name);
+
+ /** Constructor for record types */
+ glsl_type(const glsl_struct_field *fields, unsigned num_fields,
+ const char *name);
+
+ /** Constructor for array types */
+ glsl_type(const glsl_type *array, unsigned length);
+
+ /** Hash table containing the known array types. */
+ static struct hash_table *array_types;
+
+ /** Hash table containing the known record types. */
+ static struct hash_table *record_types;
+
+ static int record_key_compare(const void *a, const void *b);
+ static unsigned record_key_hash(const void *key);
+
+ /**
+ * \name Pointers to various type singletons
+ */
+ /*@{*/
+ static const glsl_type _error_type;
+ static const glsl_type _void_type;
+ static const glsl_type builtin_core_types[];
+ static const glsl_type builtin_structure_types[];
+ static const glsl_type builtin_110_deprecated_structure_types[];
+ static const glsl_type builtin_110_types[];
+ static const glsl_type builtin_120_types[];
+ static const glsl_type builtin_130_types[];
+ static const glsl_type builtin_ARB_texture_rectangle_types[];
+ static const glsl_type builtin_EXT_texture_array_types[];
+ static const glsl_type builtin_EXT_texture_buffer_object_types[];
+ /*@}*/
+
+ /**
+ * \name Methods to populate a symbol table with built-in types.
+ *
+ * \internal
+ * This is one of the truely annoying things about C++. Methods that are
+ * completely internal and private to a type still have to be advertised to
+ * the world in a public header file.
+ */
+ /*@{*/
+ static void generate_100ES_types(glsl_symbol_table *);
+ static void generate_110_types(glsl_symbol_table *);
+ static void generate_120_types(glsl_symbol_table *);
+ static void generate_130_types(glsl_symbol_table *);
+ static void generate_ARB_texture_rectangle_types(glsl_symbol_table *, bool);
+ static void generate_EXT_texture_array_types(glsl_symbol_table *, bool);
+ /*@}*/
+
+ /**
+ * \name Friend functions.
+ *
+ * These functions are friends because they must have C linkage and the
+ * need to call various private methods or access various private static
+ * data.
+ */
+ /*@{*/
+ friend void _mesa_glsl_initialize_types(struct _mesa_glsl_parse_state *);
+ friend void _mesa_glsl_release_types(void);
+ /*@}*/
+};
+
+struct glsl_struct_field {
+ const struct glsl_type *type;
+ const char *name;
+};
+
+#endif /* GLSL_TYPES_H */
diff --git a/mesalib/src/glsl/hir_field_selection.cpp b/mesalib/src/glsl/hir_field_selection.cpp index 995f284fa..21f107405 100644 --- a/mesalib/src/glsl/hir_field_selection.cpp +++ b/mesalib/src/glsl/hir_field_selection.cpp @@ -1,104 +1,102 @@ -/* - * Copyright © 2010 Intel Corporation - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING - * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER - * DEALINGS IN THE SOFTWARE. - */ - -#include "ir.h" -#include "program/symbol_table.h" -#include "glsl_parser_extras.h" -#include "ast.h" -#include "glsl_types.h" - -using std::strcmp; - -ir_rvalue * -_mesa_ast_field_selection_to_hir(const ast_expression *expr, - exec_list *instructions, - struct _mesa_glsl_parse_state *state) -{ - void *ctx = state; - ir_rvalue *result = NULL; - ir_rvalue *op; - - op = expr->subexpressions[0]->hir(instructions, state); - - /* There are two kinds of field selection. There is the selection of a - * specific field from a structure, and there is the selection of a - * swizzle / mask from a vector. Which is which is determined entirely - * by the base type of the thing to which the field selection operator is - * being applied. - */ - YYLTYPE loc = expr->get_location(); - if (op->type->is_error()) { - /* silently propagate the error */ - } else if (op->type->is_vector()) { - ir_swizzle *swiz = ir_swizzle::create(op, - expr->primary_expression.identifier, - op->type->vector_elements); - if (swiz != NULL) { - result = swiz; - } else { - /* FINISHME: Logging of error messages should be moved into - * FINISHME: ir_swizzle::create. This allows the generation of more - * FINISHME: specific error messages. - */ - _mesa_glsl_error(& loc, state, "Invalid swizzle / mask `%s'", - expr->primary_expression.identifier); - } - } else if (op->type->base_type == GLSL_TYPE_STRUCT) { - result = new(ctx) ir_dereference_record(op, - expr->primary_expression.identifier); - - if (result->type->is_error()) { - _mesa_glsl_error(& loc, state, "Cannot access field `%s' of " - "structure", - expr->primary_expression.identifier); - } - } else if (expr->subexpressions[1] != NULL) { - /* Handle "method calls" in GLSL 1.20 - namely, array.length() */ - if (state->language_version < 120) - _mesa_glsl_error(&loc, state, "Methods not supported in GLSL 1.10."); - - ast_expression *call = expr->subexpressions[1]; - assert(call->oper == ast_function_call); - - const char *method; - method = call->subexpressions[0]->primary_expression.identifier; - - if (op->type->is_array() && strcmp(method, "length") == 0) { - if (!call->expressions.is_empty()) - _mesa_glsl_error(&loc, state, "length method takes no arguments."); - - if (op->type->array_size() == 0) - _mesa_glsl_error(&loc, state, "length called on unsized array."); - - result = new(ctx) ir_constant(op->type->array_size()); - } else { - _mesa_glsl_error(&loc, state, "Unknown method: `%s'.", method); - } - } else { - _mesa_glsl_error(& loc, state, "Cannot access field `%s' of " - "non-structure / non-vector.", - expr->primary_expression.identifier); - } - - return result ? result : ir_call::get_error_instruction(ctx); -} +/*
+ * Copyright © 2010 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include "ir.h"
+#include "program/symbol_table.h"
+#include "glsl_parser_extras.h"
+#include "ast.h"
+#include "glsl_types.h"
+
+ir_rvalue *
+_mesa_ast_field_selection_to_hir(const ast_expression *expr,
+ exec_list *instructions,
+ struct _mesa_glsl_parse_state *state)
+{
+ void *ctx = state;
+ ir_rvalue *result = NULL;
+ ir_rvalue *op;
+
+ op = expr->subexpressions[0]->hir(instructions, state);
+
+ /* There are two kinds of field selection. There is the selection of a
+ * specific field from a structure, and there is the selection of a
+ * swizzle / mask from a vector. Which is which is determined entirely
+ * by the base type of the thing to which the field selection operator is
+ * being applied.
+ */
+ YYLTYPE loc = expr->get_location();
+ if (op->type->is_error()) {
+ /* silently propagate the error */
+ } else if (op->type->is_vector()) {
+ ir_swizzle *swiz = ir_swizzle::create(op,
+ expr->primary_expression.identifier,
+ op->type->vector_elements);
+ if (swiz != NULL) {
+ result = swiz;
+ } else {
+ /* FINISHME: Logging of error messages should be moved into
+ * FINISHME: ir_swizzle::create. This allows the generation of more
+ * FINISHME: specific error messages.
+ */
+ _mesa_glsl_error(& loc, state, "Invalid swizzle / mask `%s'",
+ expr->primary_expression.identifier);
+ }
+ } else if (op->type->base_type == GLSL_TYPE_STRUCT) {
+ result = new(ctx) ir_dereference_record(op,
+ expr->primary_expression.identifier);
+
+ if (result->type->is_error()) {
+ _mesa_glsl_error(& loc, state, "Cannot access field `%s' of "
+ "structure",
+ expr->primary_expression.identifier);
+ }
+ } else if (expr->subexpressions[1] != NULL) {
+ /* Handle "method calls" in GLSL 1.20 - namely, array.length() */
+ if (state->language_version < 120)
+ _mesa_glsl_error(&loc, state, "Methods not supported in GLSL 1.10.");
+
+ ast_expression *call = expr->subexpressions[1];
+ assert(call->oper == ast_function_call);
+
+ const char *method;
+ method = call->subexpressions[0]->primary_expression.identifier;
+
+ if (op->type->is_array() && strcmp(method, "length") == 0) {
+ if (!call->expressions.is_empty())
+ _mesa_glsl_error(&loc, state, "length method takes no arguments.");
+
+ if (op->type->array_size() == 0)
+ _mesa_glsl_error(&loc, state, "length called on unsized array.");
+
+ result = new(ctx) ir_constant(op->type->array_size());
+ } else {
+ _mesa_glsl_error(&loc, state, "Unknown method: `%s'.", method);
+ }
+ } else {
+ _mesa_glsl_error(& loc, state, "Cannot access field `%s' of "
+ "non-structure / non-vector.",
+ expr->primary_expression.identifier);
+ }
+
+ return result ? result : ir_call::get_error_instruction(ctx);
+}
diff --git a/mesalib/src/glsl/ir.h b/mesalib/src/glsl/ir.h index 74a8b06b1..fcf1931ff 100644 --- a/mesalib/src/glsl/ir.h +++ b/mesalib/src/glsl/ir.h @@ -1,1640 +1,1640 @@ -/* -*- c++ -*- */ -/* - * Copyright © 2010 Intel Corporation - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING - * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER - * DEALINGS IN THE SOFTWARE. - */ - -#pragma once -#ifndef IR_H -#define IR_H - -#include <cstdio> -#include <cstdlib> - -#include "ralloc.h" -#include "glsl_types.h" -#include "list.h" -#include "ir_visitor.h" -#include "ir_hierarchical_visitor.h" - -/** - * \defgroup IR Intermediate representation nodes - * - * @{ - */ - -/** - * Class tags - * - * Each concrete class derived from \c ir_instruction has a value in this - * enumerant. The value for the type is stored in \c ir_instruction::ir_type - * by the constructor. While using type tags is not very C++, it is extremely - * convenient. For example, during debugging you can simply inspect - * \c ir_instruction::ir_type to find out the actual type of the object. - * - * In addition, it is possible to use a switch-statement based on \c - * \c ir_instruction::ir_type to select different behavior for different object - * types. For functions that have only slight differences for several object - * types, this allows writing very straightforward, readable code. - */ -enum ir_node_type { - /** - * Zero is unused so that the IR validator can detect cases where - * \c ir_instruction::ir_type has not been initialized. - */ - ir_type_unset, - ir_type_variable, - ir_type_assignment, - ir_type_call, - ir_type_constant, - ir_type_dereference_array, - ir_type_dereference_record, - ir_type_dereference_variable, - ir_type_discard, - ir_type_expression, - ir_type_function, - ir_type_function_signature, - ir_type_if, - ir_type_loop, - ir_type_loop_jump, - ir_type_return, - ir_type_swizzle, - ir_type_texture, - ir_type_max /**< maximum ir_type enum number, for validation */ -}; - -/** - * Base class of all IR instructions - */ -class ir_instruction : public exec_node { -public: - enum ir_node_type ir_type; - const struct glsl_type *type; - - /** ir_print_visitor helper for debugging. */ - void print(void) const; - - virtual void accept(ir_visitor *) = 0; - virtual ir_visitor_status accept(ir_hierarchical_visitor *) = 0; - virtual ir_instruction *clone(void *mem_ctx, - struct hash_table *ht) const = 0; - - /** - * \name IR instruction downcast functions - * - * These functions either cast the object to a derived class or return - * \c NULL if the object's type does not match the specified derived class. - * Additional downcast functions will be added as needed. - */ - /*@{*/ - virtual class ir_variable * as_variable() { return NULL; } - virtual class ir_function * as_function() { return NULL; } - virtual class ir_dereference * as_dereference() { return NULL; } - virtual class ir_dereference_array * as_dereference_array() { return NULL; } - virtual class ir_dereference_variable *as_dereference_variable() { return NULL; } - virtual class ir_expression * as_expression() { return NULL; } - virtual class ir_rvalue * as_rvalue() { return NULL; } - virtual class ir_loop * as_loop() { return NULL; } - virtual class ir_assignment * as_assignment() { return NULL; } - virtual class ir_call * as_call() { return NULL; } - virtual class ir_return * as_return() { return NULL; } - virtual class ir_if * as_if() { return NULL; } - virtual class ir_swizzle * as_swizzle() { return NULL; } - virtual class ir_constant * as_constant() { return NULL; } - virtual class ir_discard * as_discard() { return NULL; } - /*@}*/ - -protected: - ir_instruction() - { - ir_type = ir_type_unset; - type = NULL; - } -}; - - -class ir_rvalue : public ir_instruction { -public: - virtual ir_rvalue *clone(void *mem_ctx, struct hash_table *) const = 0; - - virtual ir_constant *constant_expression_value() = 0; - - virtual ir_rvalue * as_rvalue() - { - return this; - } - - ir_rvalue *as_rvalue_to_saturate(); - - virtual bool is_lvalue() - { - return false; - } - - /** - * Get the variable that is ultimately referenced by an r-value - */ - virtual ir_variable *variable_referenced() - { - return NULL; - } - - - /** - * If an r-value is a reference to a whole variable, get that variable - * - * \return - * Pointer to a variable that is completely dereferenced by the r-value. If - * the r-value is not a dereference or the dereference does not access the - * entire variable (i.e., it's just one array element, struct field), \c NULL - * is returned. - */ - virtual ir_variable *whole_variable_referenced() - { - return NULL; - } - - /** - * Determine if an r-value has the value zero - * - * The base implementation of this function always returns \c false. The - * \c ir_constant class over-rides this function to return \c true \b only - * for vector and scalar types that have all elements set to the value - * zero (or \c false for booleans). - * - * \sa ir_constant::has_value, ir_rvalue::is_one, ir_rvalue::is_negative_one - */ - virtual bool is_zero() const; - - /** - * Determine if an r-value has the value one - * - * The base implementation of this function always returns \c false. The - * \c ir_constant class over-rides this function to return \c true \b only - * for vector and scalar types that have all elements set to the value - * one (or \c true for booleans). - * - * \sa ir_constant::has_value, ir_rvalue::is_zero, ir_rvalue::is_negative_one - */ - virtual bool is_one() const; - - /** - * Determine if an r-value has the value negative one - * - * The base implementation of this function always returns \c false. The - * \c ir_constant class over-rides this function to return \c true \b only - * for vector and scalar types that have all elements set to the value - * negative one. For boolean times, the result is always \c false. - * - * \sa ir_constant::has_value, ir_rvalue::is_zero, ir_rvalue::is_one - */ - virtual bool is_negative_one() const; - -protected: - ir_rvalue(); -}; - - -/** - * Variable storage classes - */ -enum ir_variable_mode { - ir_var_auto = 0, /**< Function local variables and globals. */ - ir_var_uniform, /**< Variable declared as a uniform. */ - ir_var_in, - ir_var_out, - ir_var_inout, - ir_var_const_in, /**< "in" param that must be a constant expression */ - ir_var_system_value, /**< Ex: front-face, instance-id, etc. */ - ir_var_temporary /**< Temporary variable generated during compilation. */ -}; - -enum ir_variable_interpolation { - ir_var_smooth = 0, - ir_var_flat, - ir_var_noperspective -}; - -/** - * \brief Layout qualifiers for gl_FragDepth. - * - * The AMD_conservative_depth extension allows gl_FragDepth to be redeclared - * with a layout qualifier. - */ -enum ir_depth_layout { - ir_depth_layout_none, /**< No depth layout is specified. */ - ir_depth_layout_any, - ir_depth_layout_greater, - ir_depth_layout_less, - ir_depth_layout_unchanged -}; - -/** - * \brief Convert depth layout qualifier to string. - */ -const char* -depth_layout_string(ir_depth_layout layout); - -class ir_variable : public ir_instruction { -public: - ir_variable(const struct glsl_type *, const char *, ir_variable_mode); - - virtual ir_variable *clone(void *mem_ctx, struct hash_table *ht) const; - - virtual ir_variable *as_variable() - { - return this; - } - - virtual void accept(ir_visitor *v) - { - v->visit(this); - } - - virtual ir_visitor_status accept(ir_hierarchical_visitor *); - - - /** - * Get the string value for the interpolation qualifier - * - * \return The string that would be used in a shader to specify \c - * mode will be returned. - * - * This function should only be used on a shader input or output variable. - */ - const char *interpolation_string() const; - - /** - * Calculate the number of slots required to hold this variable - * - * This is used to determine how many uniform or varying locations a variable - * occupies. The count is in units of floating point components. - */ - unsigned component_slots() const; - - /** - * Delcared name of the variable - */ - const char *name; - - /** - * Highest element accessed with a constant expression array index - * - * Not used for non-array variables. - */ - unsigned max_array_access; - - /** - * Is the variable read-only? - * - * This is set for variables declared as \c const, shader inputs, - * and uniforms. - */ - unsigned read_only:1; - unsigned centroid:1; - unsigned invariant:1; - - /** - * Has this variable been used for reading or writing? - * - * Several GLSL semantic checks require knowledge of whether or not a - * variable has been used. For example, it is an error to redeclare a - * variable as invariant after it has been used. - */ - unsigned used:1; - - /** - * Storage class of the variable. - * - * \sa ir_variable_mode - */ - unsigned mode:3; - - /** - * Interpolation mode for shader inputs / outputs - * - * \sa ir_variable_interpolation - */ - unsigned interpolation:2; - - /** - * Flag that the whole array is assignable - * - * In GLSL 1.20 and later whole arrays are assignable (and comparable for - * equality). This flag enables this behavior. - */ - unsigned array_lvalue:1; - - /** - * \name ARB_fragment_coord_conventions - * @{ - */ - unsigned origin_upper_left:1; - unsigned pixel_center_integer:1; - /*@}*/ - - /** - * \brief Layout qualifier for gl_FragDepth. - * - * This is not equal to \c ir_depth_layout_none if and only if this - * variable is \c gl_FragDepth and a layout qualifier is specified. - */ - ir_depth_layout depth_layout; - - /** - * Was the location explicitly set in the shader? - * - * If the location is explicitly set in the shader, it \b cannot be changed - * by the linker or by the API (e.g., calls to \c glBindAttribLocation have - * no effect). - */ - unsigned explicit_location:1; - - /** - * Storage location of the base of this variable - * - * The precise meaning of this field depends on the nature of the variable. - * - * - Vertex shader input: one of the values from \c gl_vert_attrib. - * - Vertex shader output: one of the values from \c gl_vert_result. - * - Fragment shader input: one of the values from \c gl_frag_attrib. - * - Fragment shader output: one of the values from \c gl_frag_result. - * - Uniforms: Per-stage uniform slot number. - * - Other: This field is not currently used. - * - * If the variable is a uniform, shader input, or shader output, and the - * slot has not been assigned, the value will be -1. - */ - int location; - - /** - * Emit a warning if this variable is accessed. - */ - const char *warn_extension; - - /** - * Value assigned in the initializer of a variable declared "const" - */ - ir_constant *constant_value; -}; - - -/*@{*/ -/** - * The representation of a function instance; may be the full definition or - * simply a prototype. - */ -class ir_function_signature : public ir_instruction { - /* An ir_function_signature will be part of the list of signatures in - * an ir_function. - */ -public: - ir_function_signature(const glsl_type *return_type); - - virtual ir_function_signature *clone(void *mem_ctx, - struct hash_table *ht) const; - ir_function_signature *clone_prototype(void *mem_ctx, - struct hash_table *ht) const; - - virtual void accept(ir_visitor *v) - { - v->visit(this); - } - - virtual ir_visitor_status accept(ir_hierarchical_visitor *); - - /** - * Get the name of the function for which this is a signature - */ - const char *function_name() const; - - /** - * Get a handle to the function for which this is a signature - * - * There is no setter function, this function returns a \c const pointer, - * and \c ir_function_signature::_function is private for a reason. The - * only way to make a connection between a function and function signature - * is via \c ir_function::add_signature. This helps ensure that certain - * invariants (i.e., a function signature is in the list of signatures for - * its \c _function) are met. - * - * \sa ir_function::add_signature - */ - inline const class ir_function *function() const - { - return this->_function; - } - - /** - * Check whether the qualifiers match between this signature's parameters - * and the supplied parameter list. If not, returns the name of the first - * parameter with mismatched qualifiers (for use in error messages). - */ - const char *qualifiers_match(exec_list *params); - - /** - * Replace the current parameter list with the given one. This is useful - * if the current information came from a prototype, and either has invalid - * or missing parameter names. - */ - void replace_parameters(exec_list *new_params); - - /** - * Function return type. - * - * \note This discards the optional precision qualifier. - */ - const struct glsl_type *return_type; - - /** - * List of ir_variable of function parameters. - * - * This represents the storage. The paramaters passed in a particular - * call will be in ir_call::actual_paramaters. - */ - struct exec_list parameters; - - /** Whether or not this function has a body (which may be empty). */ - unsigned is_defined:1; - - /** Whether or not this function signature is a built-in. */ - unsigned is_builtin:1; - - /** Body of instructions in the function. */ - struct exec_list body; - -private: - /** Function of which this signature is one overload. */ - class ir_function *_function; - - friend class ir_function; -}; - - -/** - * Header for tracking multiple overloaded functions with the same name. - * Contains a list of ir_function_signatures representing each of the - * actual functions. - */ -class ir_function : public ir_instruction { -public: - ir_function(const char *name); - - virtual ir_function *clone(void *mem_ctx, struct hash_table *ht) const; - - virtual ir_function *as_function() - { - return this; - } - - virtual void accept(ir_visitor *v) - { - v->visit(this); - } - - virtual ir_visitor_status accept(ir_hierarchical_visitor *); - - void add_signature(ir_function_signature *sig) - { - sig->_function = this; - this->signatures.push_tail(sig); - } - - /** - * Get an iterator for the set of function signatures - */ - exec_list_iterator iterator() - { - return signatures.iterator(); - } - - /** - * Find a signature that matches a set of actual parameters, taking implicit - * conversions into account. - */ - ir_function_signature *matching_signature(const exec_list *actual_param); - - /** - * Find a signature that exactly matches a set of actual parameters without - * any implicit type conversions. - */ - ir_function_signature *exact_matching_signature(const exec_list *actual_ps); - - /** - * Name of the function. - */ - const char *name; - - /** Whether or not this function has a signature that isn't a built-in. */ - bool has_user_signature(); - - /** - * List of ir_function_signature for each overloaded function with this name. - */ - struct exec_list signatures; -}; - -inline const char *ir_function_signature::function_name() const -{ - return this->_function->name; -} -/*@}*/ - - -/** - * IR instruction representing high-level if-statements - */ -class ir_if : public ir_instruction { -public: - ir_if(ir_rvalue *condition) - : condition(condition) - { - ir_type = ir_type_if; - } - - virtual ir_if *clone(void *mem_ctx, struct hash_table *ht) const; - - virtual ir_if *as_if() - { - return this; - } - - virtual void accept(ir_visitor *v) - { - v->visit(this); - } - - virtual ir_visitor_status accept(ir_hierarchical_visitor *); - - ir_rvalue *condition; - /** List of ir_instruction for the body of the then branch */ - exec_list then_instructions; - /** List of ir_instruction for the body of the else branch */ - exec_list else_instructions; -}; - - -/** - * IR instruction representing a high-level loop structure. - */ -class ir_loop : public ir_instruction { -public: - ir_loop(); - - virtual ir_loop *clone(void *mem_ctx, struct hash_table *ht) const; - - virtual void accept(ir_visitor *v) - { - v->visit(this); - } - - virtual ir_visitor_status accept(ir_hierarchical_visitor *); - - virtual ir_loop *as_loop() - { - return this; - } - - /** - * Get an iterator for the instructions of the loop body - */ - exec_list_iterator iterator() - { - return body_instructions.iterator(); - } - - /** List of ir_instruction that make up the body of the loop. */ - exec_list body_instructions; - - /** - * \name Loop counter and controls - * - * Represents a loop like a FORTRAN \c do-loop. - * - * \note - * If \c from and \c to are the same value, the loop will execute once. - */ - /*@{*/ - ir_rvalue *from; /** Value of the loop counter on the first - * iteration of the loop. - */ - ir_rvalue *to; /** Value of the loop counter on the last - * iteration of the loop. - */ - ir_rvalue *increment; - ir_variable *counter; - - /** - * Comparison operation in the loop terminator. - * - * If any of the loop control fields are non-\c NULL, this field must be - * one of \c ir_binop_less, \c ir_binop_greater, \c ir_binop_lequal, - * \c ir_binop_gequal, \c ir_binop_equal, or \c ir_binop_nequal. - */ - int cmp; - /*@}*/ -}; - - -class ir_assignment : public ir_instruction { -public: - ir_assignment(ir_rvalue *lhs, ir_rvalue *rhs, ir_rvalue *condition); - - /** - * Construct an assignment with an explicit write mask - * - * \note - * Since a write mask is supplied, the LHS must already be a bare - * \c ir_dereference. The cannot be any swizzles in the LHS. - */ - ir_assignment(ir_dereference *lhs, ir_rvalue *rhs, ir_rvalue *condition, - unsigned write_mask); - - virtual ir_assignment *clone(void *mem_ctx, struct hash_table *ht) const; - - virtual ir_constant *constant_expression_value(); - - virtual void accept(ir_visitor *v) - { - v->visit(this); - } - - virtual ir_visitor_status accept(ir_hierarchical_visitor *); - - virtual ir_assignment * as_assignment() - { - return this; - } - - /** - * Get a whole variable written by an assignment - * - * If the LHS of the assignment writes a whole variable, the variable is - * returned. Otherwise \c NULL is returned. Examples of whole-variable - * assignment are: - * - * - Assigning to a scalar - * - Assigning to all components of a vector - * - Whole array (or matrix) assignment - * - Whole structure assignment - */ - ir_variable *whole_variable_written(); - - /** - * Set the LHS of an assignment - */ - void set_lhs(ir_rvalue *lhs); - - /** - * Left-hand side of the assignment. - * - * This should be treated as read only. If you need to set the LHS of an - * assignment, use \c ir_assignment::set_lhs. - */ - ir_dereference *lhs; - - /** - * Value being assigned - */ - ir_rvalue *rhs; - - /** - * Optional condition for the assignment. - */ - ir_rvalue *condition; - - - /** - * Component mask written - * - * For non-vector types in the LHS, this field will be zero. For vector - * types, a bit will be set for each component that is written. Note that - * for \c vec2 and \c vec3 types only the lower bits will ever be set. - * - * A partially-set write mask means that each enabled channel gets - * the value from a consecutive channel of the rhs. For example, - * to write just .xyw of gl_FrontColor with color: - * - * (assign (constant bool (1)) (xyw) - * (var_ref gl_FragColor) - * (swiz xyw (var_ref color))) - */ - unsigned write_mask:4; -}; - -/* Update ir_expression::num_operands() and operator_strs when - * updating this list. - */ -enum ir_expression_operation { - ir_unop_bit_not, - ir_unop_logic_not, - ir_unop_neg, - ir_unop_abs, - ir_unop_sign, - ir_unop_rcp, - ir_unop_rsq, - ir_unop_sqrt, - ir_unop_exp, /**< Log base e on gentype */ - ir_unop_log, /**< Natural log on gentype */ - ir_unop_exp2, - ir_unop_log2, - ir_unop_f2i, /**< Float-to-integer conversion. */ - ir_unop_i2f, /**< Integer-to-float conversion. */ - ir_unop_f2b, /**< Float-to-boolean conversion */ - ir_unop_b2f, /**< Boolean-to-float conversion */ - ir_unop_i2b, /**< int-to-boolean conversion */ - ir_unop_b2i, /**< Boolean-to-int conversion */ - ir_unop_u2f, /**< Unsigned-to-float conversion. */ - ir_unop_any, - - /** - * \name Unary floating-point rounding operations. - */ - /*@{*/ - ir_unop_trunc, - ir_unop_ceil, - ir_unop_floor, - ir_unop_fract, - ir_unop_round_even, - /*@}*/ - - /** - * \name Trigonometric operations. - */ - /*@{*/ - ir_unop_sin, - ir_unop_cos, - ir_unop_sin_reduced, /**< Reduced range sin. [-pi, pi] */ - ir_unop_cos_reduced, /**< Reduced range cos. [-pi, pi] */ - /*@}*/ - - /** - * \name Partial derivatives. - */ - /*@{*/ - ir_unop_dFdx, - ir_unop_dFdy, - /*@}*/ - - ir_unop_noise, - - /** - * A sentinel marking the last of the unary operations. - */ - ir_last_unop = ir_unop_noise, - - ir_binop_add, - ir_binop_sub, - ir_binop_mul, - ir_binop_div, - - /** - * Takes one of two combinations of arguments: - * - * - mod(vecN, vecN) - * - mod(vecN, float) - * - * Does not take integer types. - */ - ir_binop_mod, - - /** - * \name Binary comparison operators which return a boolean vector. - * The type of both operands must be equal. - */ - /*@{*/ - ir_binop_less, - ir_binop_greater, - ir_binop_lequal, - ir_binop_gequal, - ir_binop_equal, - ir_binop_nequal, - /** - * Returns single boolean for whether all components of operands[0] - * equal the components of operands[1]. - */ - ir_binop_all_equal, - /** - * Returns single boolean for whether any component of operands[0] - * is not equal to the corresponding component of operands[1]. - */ - ir_binop_any_nequal, - /*@}*/ - - /** - * \name Bit-wise binary operations. - */ - /*@{*/ - ir_binop_lshift, - ir_binop_rshift, - ir_binop_bit_and, - ir_binop_bit_xor, - ir_binop_bit_or, - /*@}*/ - - ir_binop_logic_and, - ir_binop_logic_xor, - ir_binop_logic_or, - - ir_binop_dot, - ir_binop_min, - ir_binop_max, - - ir_binop_pow, - - /** - * A sentinel marking the last of the binary operations. - */ - ir_last_binop = ir_binop_pow, - - ir_quadop_vector, - - /** - * A sentinel marking the last of all operations. - */ - ir_last_opcode = ir_last_binop -}; - -class ir_expression : public ir_rvalue { -public: - /** - * Constructor for unary operation expressions - */ - ir_expression(int op, const struct glsl_type *type, ir_rvalue *); - ir_expression(int op, ir_rvalue *); - - /** - * Constructor for binary operation expressions - */ - ir_expression(int op, const struct glsl_type *type, - ir_rvalue *, ir_rvalue *); - ir_expression(int op, ir_rvalue *op0, ir_rvalue *op1); - - /** - * Constructor for quad operator expressions - */ - ir_expression(int op, const struct glsl_type *type, - ir_rvalue *, ir_rvalue *, ir_rvalue *, ir_rvalue *); - - virtual ir_expression *as_expression() - { - return this; - } - - virtual ir_expression *clone(void *mem_ctx, struct hash_table *ht) const; - - /** - * Attempt to constant-fold the expression - * - * If the expression cannot be constant folded, this method will return - * \c NULL. - */ - virtual ir_constant *constant_expression_value(); - - /** - * Determine the number of operands used by an expression - */ - static unsigned int get_num_operands(ir_expression_operation); - - /** - * Determine the number of operands used by an expression - */ - unsigned int get_num_operands() const - { - return (this->operation == ir_quadop_vector) - ? this->type->vector_elements : get_num_operands(operation); - } - - /** - * Return a string representing this expression's operator. - */ - const char *operator_string(); - - /** - * Return a string representing this expression's operator. - */ - static const char *operator_string(ir_expression_operation); - - - /** - * Do a reverse-lookup to translate the given string into an operator. - */ - static ir_expression_operation get_operator(const char *); - - virtual void accept(ir_visitor *v) - { - v->visit(this); - } - - virtual ir_visitor_status accept(ir_hierarchical_visitor *); - - ir_expression_operation operation; - ir_rvalue *operands[4]; -}; - - -/** - * IR instruction representing a function call - */ -class ir_call : public ir_rvalue { -public: - ir_call(ir_function_signature *callee, exec_list *actual_parameters) - : callee(callee) - { - ir_type = ir_type_call; - assert(callee->return_type != NULL); - type = callee->return_type; - actual_parameters->move_nodes_to(& this->actual_parameters); - } - - virtual ir_call *clone(void *mem_ctx, struct hash_table *ht) const; - - virtual ir_constant *constant_expression_value(); - - virtual ir_call *as_call() - { - return this; - } - - virtual void accept(ir_visitor *v) - { - v->visit(this); - } - - virtual ir_visitor_status accept(ir_hierarchical_visitor *); - - /** - * Get a generic ir_call object when an error occurs - * - * Any allocation will be performed with 'ctx' as ralloc owner. - */ - static ir_call *get_error_instruction(void *ctx); - - /** - * Get an iterator for the set of acutal parameters - */ - exec_list_iterator iterator() - { - return actual_parameters.iterator(); - } - - /** - * Get the name of the function being called. - */ - const char *callee_name() const - { - return callee->function_name(); - } - - /** - * Get the function signature bound to this function call - */ - ir_function_signature *get_callee() - { - return callee; - } - - /** - * Set the function call target - */ - void set_callee(ir_function_signature *sig); - - /** - * Generates an inline version of the function before @ir, - * returning the return value of the function. - */ - ir_rvalue *generate_inline(ir_instruction *ir); - - /* List of ir_rvalue of paramaters passed in this call. */ - exec_list actual_parameters; - -private: - ir_call() - : callee(NULL) - { - this->ir_type = ir_type_call; - } - - ir_function_signature *callee; -}; - - -/** - * \name Jump-like IR instructions. - * - * These include \c break, \c continue, \c return, and \c discard. - */ -/*@{*/ -class ir_jump : public ir_instruction { -protected: - ir_jump() - { - ir_type = ir_type_unset; - } -}; - -class ir_return : public ir_jump { -public: - ir_return() - : value(NULL) - { - this->ir_type = ir_type_return; - } - - ir_return(ir_rvalue *value) - : value(value) - { - this->ir_type = ir_type_return; - } - - virtual ir_return *clone(void *mem_ctx, struct hash_table *) const; - - virtual ir_return *as_return() - { - return this; - } - - ir_rvalue *get_value() const - { - return value; - } - - virtual void accept(ir_visitor *v) - { - v->visit(this); - } - - virtual ir_visitor_status accept(ir_hierarchical_visitor *); - - ir_rvalue *value; -}; - - -/** - * Jump instructions used inside loops - * - * These include \c break and \c continue. The \c break within a loop is - * different from the \c break within a switch-statement. - * - * \sa ir_switch_jump - */ -class ir_loop_jump : public ir_jump { -public: - enum jump_mode { - jump_break, - jump_continue - }; - - ir_loop_jump(jump_mode mode) - { - this->ir_type = ir_type_loop_jump; - this->mode = mode; - this->loop = loop; - } - - virtual ir_loop_jump *clone(void *mem_ctx, struct hash_table *) const; - - virtual void accept(ir_visitor *v) - { - v->visit(this); - } - - virtual ir_visitor_status accept(ir_hierarchical_visitor *); - - bool is_break() const - { - return mode == jump_break; - } - - bool is_continue() const - { - return mode == jump_continue; - } - - /** Mode selector for the jump instruction. */ - enum jump_mode mode; -private: - /** Loop containing this break instruction. */ - ir_loop *loop; -}; - -/** - * IR instruction representing discard statements. - */ -class ir_discard : public ir_jump { -public: - ir_discard() - { - this->ir_type = ir_type_discard; - this->condition = NULL; - } - - ir_discard(ir_rvalue *cond) - { - this->ir_type = ir_type_discard; - this->condition = cond; - } - - virtual ir_discard *clone(void *mem_ctx, struct hash_table *ht) const; - - virtual void accept(ir_visitor *v) - { - v->visit(this); - } - - virtual ir_visitor_status accept(ir_hierarchical_visitor *); - - virtual ir_discard *as_discard() - { - return this; - } - - ir_rvalue *condition; -}; -/*@}*/ - - -/** - * Texture sampling opcodes used in ir_texture - */ -enum ir_texture_opcode { - ir_tex, /**< Regular texture look-up */ - ir_txb, /**< Texture look-up with LOD bias */ - ir_txl, /**< Texture look-up with explicit LOD */ - ir_txd, /**< Texture look-up with partial derivatvies */ - ir_txf /**< Texel fetch with explicit LOD */ -}; - - -/** - * IR instruction to sample a texture - * - * The specific form of the IR instruction depends on the \c mode value - * selected from \c ir_texture_opcodes. In the printed IR, these will - * appear as: - * - * Texel offset (0 or an expression) - * | Projection divisor - * | | Shadow comparitor - * | | | - * v v v - * (tex <sampler> <coordinate> 0 1 ( )) - * (txb <sampler> <coordinate> 0 1 ( ) <bias>) - * (txl <sampler> <coordinate> 0 1 ( ) <lod>) - * (txd <sampler> <coordinate> 0 1 ( ) (dPdx dPdy)) - * (txf <sampler> <coordinate> 0 <lod>) - */ -class ir_texture : public ir_rvalue { -public: - ir_texture(enum ir_texture_opcode op) - : op(op), projector(NULL), shadow_comparitor(NULL), offset(NULL) - { - this->ir_type = ir_type_texture; - } - - virtual ir_texture *clone(void *mem_ctx, struct hash_table *) const; - - virtual ir_constant *constant_expression_value(); - - virtual void accept(ir_visitor *v) - { - v->visit(this); - } - - virtual ir_visitor_status accept(ir_hierarchical_visitor *); - - /** - * Return a string representing the ir_texture_opcode. - */ - const char *opcode_string(); - - /** Set the sampler and infer the type. */ - void set_sampler(ir_dereference *sampler); - - /** - * Do a reverse-lookup to translate a string into an ir_texture_opcode. - */ - static ir_texture_opcode get_opcode(const char *); - - enum ir_texture_opcode op; - - /** Sampler to use for the texture access. */ - ir_dereference *sampler; - - /** Texture coordinate to sample */ - ir_rvalue *coordinate; - - /** - * Value used for projective divide. - * - * If there is no projective divide (the common case), this will be - * \c NULL. Optimization passes should check for this to point to a constant - * of 1.0 and replace that with \c NULL. - */ - ir_rvalue *projector; - - /** - * Coordinate used for comparison on shadow look-ups. - * - * If there is no shadow comparison, this will be \c NULL. For the - * \c ir_txf opcode, this *must* be \c NULL. - */ - ir_rvalue *shadow_comparitor; - - /** Texel offset. */ - ir_rvalue *offset; - - union { - ir_rvalue *lod; /**< Floating point LOD */ - ir_rvalue *bias; /**< Floating point LOD bias */ - struct { - ir_rvalue *dPdx; /**< Partial derivative of coordinate wrt X */ - ir_rvalue *dPdy; /**< Partial derivative of coordinate wrt Y */ - } grad; - } lod_info; -}; - - -struct ir_swizzle_mask { - unsigned x:2; - unsigned y:2; - unsigned z:2; - unsigned w:2; - - /** - * Number of components in the swizzle. - */ - unsigned num_components:3; - - /** - * Does the swizzle contain duplicate components? - * - * L-value swizzles cannot contain duplicate components. - */ - unsigned has_duplicates:1; -}; - - -class ir_swizzle : public ir_rvalue { -public: - ir_swizzle(ir_rvalue *, unsigned x, unsigned y, unsigned z, unsigned w, - unsigned count); - - ir_swizzle(ir_rvalue *val, const unsigned *components, unsigned count); - - ir_swizzle(ir_rvalue *val, ir_swizzle_mask mask); - - virtual ir_swizzle *clone(void *mem_ctx, struct hash_table *) const; - - virtual ir_constant *constant_expression_value(); - - virtual ir_swizzle *as_swizzle() - { - return this; - } - - /** - * Construct an ir_swizzle from the textual representation. Can fail. - */ - static ir_swizzle *create(ir_rvalue *, const char *, unsigned vector_length); - - virtual void accept(ir_visitor *v) - { - v->visit(this); - } - - virtual ir_visitor_status accept(ir_hierarchical_visitor *); - - bool is_lvalue() - { - return val->is_lvalue() && !mask.has_duplicates; - } - - /** - * Get the variable that is ultimately referenced by an r-value - */ - virtual ir_variable *variable_referenced(); - - ir_rvalue *val; - ir_swizzle_mask mask; - -private: - /** - * Initialize the mask component of a swizzle - * - * This is used by the \c ir_swizzle constructors. - */ - void init_mask(const unsigned *components, unsigned count); -}; - - -class ir_dereference : public ir_rvalue { -public: - virtual ir_dereference *clone(void *mem_ctx, struct hash_table *) const = 0; - - virtual ir_dereference *as_dereference() - { - return this; - } - - bool is_lvalue(); - - /** - * Get the variable that is ultimately referenced by an r-value - */ - virtual ir_variable *variable_referenced() = 0; -}; - - -class ir_dereference_variable : public ir_dereference { -public: - ir_dereference_variable(ir_variable *var); - - virtual ir_dereference_variable *clone(void *mem_ctx, - struct hash_table *) const; - - virtual ir_constant *constant_expression_value(); - - virtual ir_dereference_variable *as_dereference_variable() - { - return this; - } - - /** - * Get the variable that is ultimately referenced by an r-value - */ - virtual ir_variable *variable_referenced() - { - return this->var; - } - - virtual ir_variable *whole_variable_referenced() - { - /* ir_dereference_variable objects always dereference the entire - * variable. However, if this dereference is dereferenced by anything - * else, the complete deferefernce chain is not a whole-variable - * dereference. This method should only be called on the top most - * ir_rvalue in a dereference chain. - */ - return this->var; - } - - virtual void accept(ir_visitor *v) - { - v->visit(this); - } - - virtual ir_visitor_status accept(ir_hierarchical_visitor *); - - /** - * Object being dereferenced. - */ - ir_variable *var; -}; - - -class ir_dereference_array : public ir_dereference { -public: - ir_dereference_array(ir_rvalue *value, ir_rvalue *array_index); - - ir_dereference_array(ir_variable *var, ir_rvalue *array_index); - - virtual ir_dereference_array *clone(void *mem_ctx, - struct hash_table *) const; - - virtual ir_constant *constant_expression_value(); - - virtual ir_dereference_array *as_dereference_array() - { - return this; - } - - /** - * Get the variable that is ultimately referenced by an r-value - */ - virtual ir_variable *variable_referenced() - { - return this->array->variable_referenced(); - } - - virtual void accept(ir_visitor *v) - { - v->visit(this); - } - - virtual ir_visitor_status accept(ir_hierarchical_visitor *); - - ir_rvalue *array; - ir_rvalue *array_index; - -private: - void set_array(ir_rvalue *value); -}; - - -class ir_dereference_record : public ir_dereference { -public: - ir_dereference_record(ir_rvalue *value, const char *field); - - ir_dereference_record(ir_variable *var, const char *field); - - virtual ir_dereference_record *clone(void *mem_ctx, - struct hash_table *) const; - - virtual ir_constant *constant_expression_value(); - - /** - * Get the variable that is ultimately referenced by an r-value - */ - virtual ir_variable *variable_referenced() - { - return this->record->variable_referenced(); - } - - virtual void accept(ir_visitor *v) - { - v->visit(this); - } - - virtual ir_visitor_status accept(ir_hierarchical_visitor *); - - ir_rvalue *record; - const char *field; -}; - - -/** - * Data stored in an ir_constant - */ -union ir_constant_data { - unsigned u[16]; - int i[16]; - float f[16]; - bool b[16]; -}; - - -class ir_constant : public ir_rvalue { -public: - ir_constant(const struct glsl_type *type, const ir_constant_data *data); - ir_constant(bool b); - ir_constant(unsigned int u); - ir_constant(int i); - ir_constant(float f); - - /** - * Construct an ir_constant from a list of ir_constant values - */ - ir_constant(const struct glsl_type *type, exec_list *values); - - /** - * Construct an ir_constant from a scalar component of another ir_constant - * - * The new \c ir_constant inherits the type of the component from the - * source constant. - * - * \note - * In the case of a matrix constant, the new constant is a scalar, \b not - * a vector. - */ - ir_constant(const ir_constant *c, unsigned i); - - /** - * Return a new ir_constant of the specified type containing all zeros. - */ - static ir_constant *zero(void *mem_ctx, const glsl_type *type); - - virtual ir_constant *clone(void *mem_ctx, struct hash_table *) const; - - virtual ir_constant *constant_expression_value(); - - virtual ir_constant *as_constant() - { - return this; - } - - virtual void accept(ir_visitor *v) - { - v->visit(this); - } - - virtual ir_visitor_status accept(ir_hierarchical_visitor *); - - /** - * Get a particular component of a constant as a specific type - * - * This is useful, for example, to get a value from an integer constant - * as a float or bool. This appears frequently when constructors are - * called with all constant parameters. - */ - /*@{*/ - bool get_bool_component(unsigned i) const; - float get_float_component(unsigned i) const; - int get_int_component(unsigned i) const; - unsigned get_uint_component(unsigned i) const; - /*@}*/ - - ir_constant *get_array_element(unsigned i) const; - - ir_constant *get_record_field(const char *name); - - /** - * Determine whether a constant has the same value as another constant - * - * \sa ir_constant::is_zero, ir_constant::is_one, - * ir_constant::is_negative_one - */ - bool has_value(const ir_constant *) const; - - virtual bool is_zero() const; - virtual bool is_one() const; - virtual bool is_negative_one() const; - - /** - * Value of the constant. - * - * The field used to back the values supplied by the constant is determined - * by the type associated with the \c ir_instruction. Constants may be - * scalars, vectors, or matrices. - */ - union ir_constant_data value; - - /* Array elements */ - ir_constant **array_elements; - - /* Structure fields */ - exec_list components; - -private: - /** - * Parameterless constructor only used by the clone method - */ - ir_constant(void); -}; - -/*@}*/ - -/** - * Apply a visitor to each IR node in a list - */ -void -visit_exec_list(exec_list *list, ir_visitor *visitor); - -/** - * Validate invariants on each IR node in a list - */ -void validate_ir_tree(exec_list *instructions); - -/** - * Make a clone of each IR instruction in a list - * - * \param in List of IR instructions that are to be cloned - * \param out List to hold the cloned instructions - */ -void -clone_ir_list(void *mem_ctx, exec_list *out, const exec_list *in); - -extern void -_mesa_glsl_initialize_variables(exec_list *instructions, - struct _mesa_glsl_parse_state *state); - -extern void -_mesa_glsl_initialize_functions(_mesa_glsl_parse_state *state); - -extern void -_mesa_glsl_release_functions(void); - -extern void -reparent_ir(exec_list *list, void *mem_ctx); - -struct glsl_symbol_table; - -extern void -import_prototypes(const exec_list *source, exec_list *dest, - struct glsl_symbol_table *symbols, void *mem_ctx); - -extern bool -ir_has_call(ir_instruction *ir); - -extern void -do_set_program_inouts(exec_list *instructions, struct gl_program *prog); - -#endif /* IR_H */ +/* -*- c++ -*- */
+/*
+ * Copyright © 2010 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#pragma once
+#ifndef IR_H
+#define IR_H
+
+#include <stdio.h>
+#include <stdlib.h>
+
+#include "ralloc.h"
+#include "glsl_types.h"
+#include "list.h"
+#include "ir_visitor.h"
+#include "ir_hierarchical_visitor.h"
+
+/**
+ * \defgroup IR Intermediate representation nodes
+ *
+ * @{
+ */
+
+/**
+ * Class tags
+ *
+ * Each concrete class derived from \c ir_instruction has a value in this
+ * enumerant. The value for the type is stored in \c ir_instruction::ir_type
+ * by the constructor. While using type tags is not very C++, it is extremely
+ * convenient. For example, during debugging you can simply inspect
+ * \c ir_instruction::ir_type to find out the actual type of the object.
+ *
+ * In addition, it is possible to use a switch-statement based on \c
+ * \c ir_instruction::ir_type to select different behavior for different object
+ * types. For functions that have only slight differences for several object
+ * types, this allows writing very straightforward, readable code.
+ */
+enum ir_node_type {
+ /**
+ * Zero is unused so that the IR validator can detect cases where
+ * \c ir_instruction::ir_type has not been initialized.
+ */
+ ir_type_unset,
+ ir_type_variable,
+ ir_type_assignment,
+ ir_type_call,
+ ir_type_constant,
+ ir_type_dereference_array,
+ ir_type_dereference_record,
+ ir_type_dereference_variable,
+ ir_type_discard,
+ ir_type_expression,
+ ir_type_function,
+ ir_type_function_signature,
+ ir_type_if,
+ ir_type_loop,
+ ir_type_loop_jump,
+ ir_type_return,
+ ir_type_swizzle,
+ ir_type_texture,
+ ir_type_max /**< maximum ir_type enum number, for validation */
+};
+
+/**
+ * Base class of all IR instructions
+ */
+class ir_instruction : public exec_node {
+public:
+ enum ir_node_type ir_type;
+ const struct glsl_type *type;
+
+ /** ir_print_visitor helper for debugging. */
+ void print(void) const;
+
+ virtual void accept(ir_visitor *) = 0;
+ virtual ir_visitor_status accept(ir_hierarchical_visitor *) = 0;
+ virtual ir_instruction *clone(void *mem_ctx,
+ struct hash_table *ht) const = 0;
+
+ /**
+ * \name IR instruction downcast functions
+ *
+ * These functions either cast the object to a derived class or return
+ * \c NULL if the object's type does not match the specified derived class.
+ * Additional downcast functions will be added as needed.
+ */
+ /*@{*/
+ virtual class ir_variable * as_variable() { return NULL; }
+ virtual class ir_function * as_function() { return NULL; }
+ virtual class ir_dereference * as_dereference() { return NULL; }
+ virtual class ir_dereference_array * as_dereference_array() { return NULL; }
+ virtual class ir_dereference_variable *as_dereference_variable() { return NULL; }
+ virtual class ir_expression * as_expression() { return NULL; }
+ virtual class ir_rvalue * as_rvalue() { return NULL; }
+ virtual class ir_loop * as_loop() { return NULL; }
+ virtual class ir_assignment * as_assignment() { return NULL; }
+ virtual class ir_call * as_call() { return NULL; }
+ virtual class ir_return * as_return() { return NULL; }
+ virtual class ir_if * as_if() { return NULL; }
+ virtual class ir_swizzle * as_swizzle() { return NULL; }
+ virtual class ir_constant * as_constant() { return NULL; }
+ virtual class ir_discard * as_discard() { return NULL; }
+ /*@}*/
+
+protected:
+ ir_instruction()
+ {
+ ir_type = ir_type_unset;
+ type = NULL;
+ }
+};
+
+
+class ir_rvalue : public ir_instruction {
+public:
+ virtual ir_rvalue *clone(void *mem_ctx, struct hash_table *) const = 0;
+
+ virtual ir_constant *constant_expression_value() = 0;
+
+ virtual ir_rvalue * as_rvalue()
+ {
+ return this;
+ }
+
+ ir_rvalue *as_rvalue_to_saturate();
+
+ virtual bool is_lvalue()
+ {
+ return false;
+ }
+
+ /**
+ * Get the variable that is ultimately referenced by an r-value
+ */
+ virtual ir_variable *variable_referenced()
+ {
+ return NULL;
+ }
+
+
+ /**
+ * If an r-value is a reference to a whole variable, get that variable
+ *
+ * \return
+ * Pointer to a variable that is completely dereferenced by the r-value. If
+ * the r-value is not a dereference or the dereference does not access the
+ * entire variable (i.e., it's just one array element, struct field), \c NULL
+ * is returned.
+ */
+ virtual ir_variable *whole_variable_referenced()
+ {
+ return NULL;
+ }
+
+ /**
+ * Determine if an r-value has the value zero
+ *
+ * The base implementation of this function always returns \c false. The
+ * \c ir_constant class over-rides this function to return \c true \b only
+ * for vector and scalar types that have all elements set to the value
+ * zero (or \c false for booleans).
+ *
+ * \sa ir_constant::has_value, ir_rvalue::is_one, ir_rvalue::is_negative_one
+ */
+ virtual bool is_zero() const;
+
+ /**
+ * Determine if an r-value has the value one
+ *
+ * The base implementation of this function always returns \c false. The
+ * \c ir_constant class over-rides this function to return \c true \b only
+ * for vector and scalar types that have all elements set to the value
+ * one (or \c true for booleans).
+ *
+ * \sa ir_constant::has_value, ir_rvalue::is_zero, ir_rvalue::is_negative_one
+ */
+ virtual bool is_one() const;
+
+ /**
+ * Determine if an r-value has the value negative one
+ *
+ * The base implementation of this function always returns \c false. The
+ * \c ir_constant class over-rides this function to return \c true \b only
+ * for vector and scalar types that have all elements set to the value
+ * negative one. For boolean times, the result is always \c false.
+ *
+ * \sa ir_constant::has_value, ir_rvalue::is_zero, ir_rvalue::is_one
+ */
+ virtual bool is_negative_one() const;
+
+protected:
+ ir_rvalue();
+};
+
+
+/**
+ * Variable storage classes
+ */
+enum ir_variable_mode {
+ ir_var_auto = 0, /**< Function local variables and globals. */
+ ir_var_uniform, /**< Variable declared as a uniform. */
+ ir_var_in,
+ ir_var_out,
+ ir_var_inout,
+ ir_var_const_in, /**< "in" param that must be a constant expression */
+ ir_var_system_value, /**< Ex: front-face, instance-id, etc. */
+ ir_var_temporary /**< Temporary variable generated during compilation. */
+};
+
+enum ir_variable_interpolation {
+ ir_var_smooth = 0,
+ ir_var_flat,
+ ir_var_noperspective
+};
+
+/**
+ * \brief Layout qualifiers for gl_FragDepth.
+ *
+ * The AMD_conservative_depth extension allows gl_FragDepth to be redeclared
+ * with a layout qualifier.
+ */
+enum ir_depth_layout {
+ ir_depth_layout_none, /**< No depth layout is specified. */
+ ir_depth_layout_any,
+ ir_depth_layout_greater,
+ ir_depth_layout_less,
+ ir_depth_layout_unchanged
+};
+
+/**
+ * \brief Convert depth layout qualifier to string.
+ */
+const char*
+depth_layout_string(ir_depth_layout layout);
+
+class ir_variable : public ir_instruction {
+public:
+ ir_variable(const struct glsl_type *, const char *, ir_variable_mode);
+
+ virtual ir_variable *clone(void *mem_ctx, struct hash_table *ht) const;
+
+ virtual ir_variable *as_variable()
+ {
+ return this;
+ }
+
+ virtual void accept(ir_visitor *v)
+ {
+ v->visit(this);
+ }
+
+ virtual ir_visitor_status accept(ir_hierarchical_visitor *);
+
+
+ /**
+ * Get the string value for the interpolation qualifier
+ *
+ * \return The string that would be used in a shader to specify \c
+ * mode will be returned.
+ *
+ * This function should only be used on a shader input or output variable.
+ */
+ const char *interpolation_string() const;
+
+ /**
+ * Calculate the number of slots required to hold this variable
+ *
+ * This is used to determine how many uniform or varying locations a variable
+ * occupies. The count is in units of floating point components.
+ */
+ unsigned component_slots() const;
+
+ /**
+ * Delcared name of the variable
+ */
+ const char *name;
+
+ /**
+ * Highest element accessed with a constant expression array index
+ *
+ * Not used for non-array variables.
+ */
+ unsigned max_array_access;
+
+ /**
+ * Is the variable read-only?
+ *
+ * This is set for variables declared as \c const, shader inputs,
+ * and uniforms.
+ */
+ unsigned read_only:1;
+ unsigned centroid:1;
+ unsigned invariant:1;
+
+ /**
+ * Has this variable been used for reading or writing?
+ *
+ * Several GLSL semantic checks require knowledge of whether or not a
+ * variable has been used. For example, it is an error to redeclare a
+ * variable as invariant after it has been used.
+ */
+ unsigned used:1;
+
+ /**
+ * Storage class of the variable.
+ *
+ * \sa ir_variable_mode
+ */
+ unsigned mode:3;
+
+ /**
+ * Interpolation mode for shader inputs / outputs
+ *
+ * \sa ir_variable_interpolation
+ */
+ unsigned interpolation:2;
+
+ /**
+ * Flag that the whole array is assignable
+ *
+ * In GLSL 1.20 and later whole arrays are assignable (and comparable for
+ * equality). This flag enables this behavior.
+ */
+ unsigned array_lvalue:1;
+
+ /**
+ * \name ARB_fragment_coord_conventions
+ * @{
+ */
+ unsigned origin_upper_left:1;
+ unsigned pixel_center_integer:1;
+ /*@}*/
+
+ /**
+ * \brief Layout qualifier for gl_FragDepth.
+ *
+ * This is not equal to \c ir_depth_layout_none if and only if this
+ * variable is \c gl_FragDepth and a layout qualifier is specified.
+ */
+ ir_depth_layout depth_layout;
+
+ /**
+ * Was the location explicitly set in the shader?
+ *
+ * If the location is explicitly set in the shader, it \b cannot be changed
+ * by the linker or by the API (e.g., calls to \c glBindAttribLocation have
+ * no effect).
+ */
+ unsigned explicit_location:1;
+
+ /**
+ * Storage location of the base of this variable
+ *
+ * The precise meaning of this field depends on the nature of the variable.
+ *
+ * - Vertex shader input: one of the values from \c gl_vert_attrib.
+ * - Vertex shader output: one of the values from \c gl_vert_result.
+ * - Fragment shader input: one of the values from \c gl_frag_attrib.
+ * - Fragment shader output: one of the values from \c gl_frag_result.
+ * - Uniforms: Per-stage uniform slot number.
+ * - Other: This field is not currently used.
+ *
+ * If the variable is a uniform, shader input, or shader output, and the
+ * slot has not been assigned, the value will be -1.
+ */
+ int location;
+
+ /**
+ * Emit a warning if this variable is accessed.
+ */
+ const char *warn_extension;
+
+ /**
+ * Value assigned in the initializer of a variable declared "const"
+ */
+ ir_constant *constant_value;
+};
+
+
+/*@{*/
+/**
+ * The representation of a function instance; may be the full definition or
+ * simply a prototype.
+ */
+class ir_function_signature : public ir_instruction {
+ /* An ir_function_signature will be part of the list of signatures in
+ * an ir_function.
+ */
+public:
+ ir_function_signature(const glsl_type *return_type);
+
+ virtual ir_function_signature *clone(void *mem_ctx,
+ struct hash_table *ht) const;
+ ir_function_signature *clone_prototype(void *mem_ctx,
+ struct hash_table *ht) const;
+
+ virtual void accept(ir_visitor *v)
+ {
+ v->visit(this);
+ }
+
+ virtual ir_visitor_status accept(ir_hierarchical_visitor *);
+
+ /**
+ * Get the name of the function for which this is a signature
+ */
+ const char *function_name() const;
+
+ /**
+ * Get a handle to the function for which this is a signature
+ *
+ * There is no setter function, this function returns a \c const pointer,
+ * and \c ir_function_signature::_function is private for a reason. The
+ * only way to make a connection between a function and function signature
+ * is via \c ir_function::add_signature. This helps ensure that certain
+ * invariants (i.e., a function signature is in the list of signatures for
+ * its \c _function) are met.
+ *
+ * \sa ir_function::add_signature
+ */
+ inline const class ir_function *function() const
+ {
+ return this->_function;
+ }
+
+ /**
+ * Check whether the qualifiers match between this signature's parameters
+ * and the supplied parameter list. If not, returns the name of the first
+ * parameter with mismatched qualifiers (for use in error messages).
+ */
+ const char *qualifiers_match(exec_list *params);
+
+ /**
+ * Replace the current parameter list with the given one. This is useful
+ * if the current information came from a prototype, and either has invalid
+ * or missing parameter names.
+ */
+ void replace_parameters(exec_list *new_params);
+
+ /**
+ * Function return type.
+ *
+ * \note This discards the optional precision qualifier.
+ */
+ const struct glsl_type *return_type;
+
+ /**
+ * List of ir_variable of function parameters.
+ *
+ * This represents the storage. The paramaters passed in a particular
+ * call will be in ir_call::actual_paramaters.
+ */
+ struct exec_list parameters;
+
+ /** Whether or not this function has a body (which may be empty). */
+ unsigned is_defined:1;
+
+ /** Whether or not this function signature is a built-in. */
+ unsigned is_builtin:1;
+
+ /** Body of instructions in the function. */
+ struct exec_list body;
+
+private:
+ /** Function of which this signature is one overload. */
+ class ir_function *_function;
+
+ friend class ir_function;
+};
+
+
+/**
+ * Header for tracking multiple overloaded functions with the same name.
+ * Contains a list of ir_function_signatures representing each of the
+ * actual functions.
+ */
+class ir_function : public ir_instruction {
+public:
+ ir_function(const char *name);
+
+ virtual ir_function *clone(void *mem_ctx, struct hash_table *ht) const;
+
+ virtual ir_function *as_function()
+ {
+ return this;
+ }
+
+ virtual void accept(ir_visitor *v)
+ {
+ v->visit(this);
+ }
+
+ virtual ir_visitor_status accept(ir_hierarchical_visitor *);
+
+ void add_signature(ir_function_signature *sig)
+ {
+ sig->_function = this;
+ this->signatures.push_tail(sig);
+ }
+
+ /**
+ * Get an iterator for the set of function signatures
+ */
+ exec_list_iterator iterator()
+ {
+ return signatures.iterator();
+ }
+
+ /**
+ * Find a signature that matches a set of actual parameters, taking implicit
+ * conversions into account.
+ */
+ ir_function_signature *matching_signature(const exec_list *actual_param);
+
+ /**
+ * Find a signature that exactly matches a set of actual parameters without
+ * any implicit type conversions.
+ */
+ ir_function_signature *exact_matching_signature(const exec_list *actual_ps);
+
+ /**
+ * Name of the function.
+ */
+ const char *name;
+
+ /** Whether or not this function has a signature that isn't a built-in. */
+ bool has_user_signature();
+
+ /**
+ * List of ir_function_signature for each overloaded function with this name.
+ */
+ struct exec_list signatures;
+};
+
+inline const char *ir_function_signature::function_name() const
+{
+ return this->_function->name;
+}
+/*@}*/
+
+
+/**
+ * IR instruction representing high-level if-statements
+ */
+class ir_if : public ir_instruction {
+public:
+ ir_if(ir_rvalue *condition)
+ : condition(condition)
+ {
+ ir_type = ir_type_if;
+ }
+
+ virtual ir_if *clone(void *mem_ctx, struct hash_table *ht) const;
+
+ virtual ir_if *as_if()
+ {
+ return this;
+ }
+
+ virtual void accept(ir_visitor *v)
+ {
+ v->visit(this);
+ }
+
+ virtual ir_visitor_status accept(ir_hierarchical_visitor *);
+
+ ir_rvalue *condition;
+ /** List of ir_instruction for the body of the then branch */
+ exec_list then_instructions;
+ /** List of ir_instruction for the body of the else branch */
+ exec_list else_instructions;
+};
+
+
+/**
+ * IR instruction representing a high-level loop structure.
+ */
+class ir_loop : public ir_instruction {
+public:
+ ir_loop();
+
+ virtual ir_loop *clone(void *mem_ctx, struct hash_table *ht) const;
+
+ virtual void accept(ir_visitor *v)
+ {
+ v->visit(this);
+ }
+
+ virtual ir_visitor_status accept(ir_hierarchical_visitor *);
+
+ virtual ir_loop *as_loop()
+ {
+ return this;
+ }
+
+ /**
+ * Get an iterator for the instructions of the loop body
+ */
+ exec_list_iterator iterator()
+ {
+ return body_instructions.iterator();
+ }
+
+ /** List of ir_instruction that make up the body of the loop. */
+ exec_list body_instructions;
+
+ /**
+ * \name Loop counter and controls
+ *
+ * Represents a loop like a FORTRAN \c do-loop.
+ *
+ * \note
+ * If \c from and \c to are the same value, the loop will execute once.
+ */
+ /*@{*/
+ ir_rvalue *from; /** Value of the loop counter on the first
+ * iteration of the loop.
+ */
+ ir_rvalue *to; /** Value of the loop counter on the last
+ * iteration of the loop.
+ */
+ ir_rvalue *increment;
+ ir_variable *counter;
+
+ /**
+ * Comparison operation in the loop terminator.
+ *
+ * If any of the loop control fields are non-\c NULL, this field must be
+ * one of \c ir_binop_less, \c ir_binop_greater, \c ir_binop_lequal,
+ * \c ir_binop_gequal, \c ir_binop_equal, or \c ir_binop_nequal.
+ */
+ int cmp;
+ /*@}*/
+};
+
+
+class ir_assignment : public ir_instruction {
+public:
+ ir_assignment(ir_rvalue *lhs, ir_rvalue *rhs, ir_rvalue *condition);
+
+ /**
+ * Construct an assignment with an explicit write mask
+ *
+ * \note
+ * Since a write mask is supplied, the LHS must already be a bare
+ * \c ir_dereference. The cannot be any swizzles in the LHS.
+ */
+ ir_assignment(ir_dereference *lhs, ir_rvalue *rhs, ir_rvalue *condition,
+ unsigned write_mask);
+
+ virtual ir_assignment *clone(void *mem_ctx, struct hash_table *ht) const;
+
+ virtual ir_constant *constant_expression_value();
+
+ virtual void accept(ir_visitor *v)
+ {
+ v->visit(this);
+ }
+
+ virtual ir_visitor_status accept(ir_hierarchical_visitor *);
+
+ virtual ir_assignment * as_assignment()
+ {
+ return this;
+ }
+
+ /**
+ * Get a whole variable written by an assignment
+ *
+ * If the LHS of the assignment writes a whole variable, the variable is
+ * returned. Otherwise \c NULL is returned. Examples of whole-variable
+ * assignment are:
+ *
+ * - Assigning to a scalar
+ * - Assigning to all components of a vector
+ * - Whole array (or matrix) assignment
+ * - Whole structure assignment
+ */
+ ir_variable *whole_variable_written();
+
+ /**
+ * Set the LHS of an assignment
+ */
+ void set_lhs(ir_rvalue *lhs);
+
+ /**
+ * Left-hand side of the assignment.
+ *
+ * This should be treated as read only. If you need to set the LHS of an
+ * assignment, use \c ir_assignment::set_lhs.
+ */
+ ir_dereference *lhs;
+
+ /**
+ * Value being assigned
+ */
+ ir_rvalue *rhs;
+
+ /**
+ * Optional condition for the assignment.
+ */
+ ir_rvalue *condition;
+
+
+ /**
+ * Component mask written
+ *
+ * For non-vector types in the LHS, this field will be zero. For vector
+ * types, a bit will be set for each component that is written. Note that
+ * for \c vec2 and \c vec3 types only the lower bits will ever be set.
+ *
+ * A partially-set write mask means that each enabled channel gets
+ * the value from a consecutive channel of the rhs. For example,
+ * to write just .xyw of gl_FrontColor with color:
+ *
+ * (assign (constant bool (1)) (xyw)
+ * (var_ref gl_FragColor)
+ * (swiz xyw (var_ref color)))
+ */
+ unsigned write_mask:4;
+};
+
+/* Update ir_expression::num_operands() and operator_strs when
+ * updating this list.
+ */
+enum ir_expression_operation {
+ ir_unop_bit_not,
+ ir_unop_logic_not,
+ ir_unop_neg,
+ ir_unop_abs,
+ ir_unop_sign,
+ ir_unop_rcp,
+ ir_unop_rsq,
+ ir_unop_sqrt,
+ ir_unop_exp, /**< Log base e on gentype */
+ ir_unop_log, /**< Natural log on gentype */
+ ir_unop_exp2,
+ ir_unop_log2,
+ ir_unop_f2i, /**< Float-to-integer conversion. */
+ ir_unop_i2f, /**< Integer-to-float conversion. */
+ ir_unop_f2b, /**< Float-to-boolean conversion */
+ ir_unop_b2f, /**< Boolean-to-float conversion */
+ ir_unop_i2b, /**< int-to-boolean conversion */
+ ir_unop_b2i, /**< Boolean-to-int conversion */
+ ir_unop_u2f, /**< Unsigned-to-float conversion. */
+ ir_unop_any,
+
+ /**
+ * \name Unary floating-point rounding operations.
+ */
+ /*@{*/
+ ir_unop_trunc,
+ ir_unop_ceil,
+ ir_unop_floor,
+ ir_unop_fract,
+ ir_unop_round_even,
+ /*@}*/
+
+ /**
+ * \name Trigonometric operations.
+ */
+ /*@{*/
+ ir_unop_sin,
+ ir_unop_cos,
+ ir_unop_sin_reduced, /**< Reduced range sin. [-pi, pi] */
+ ir_unop_cos_reduced, /**< Reduced range cos. [-pi, pi] */
+ /*@}*/
+
+ /**
+ * \name Partial derivatives.
+ */
+ /*@{*/
+ ir_unop_dFdx,
+ ir_unop_dFdy,
+ /*@}*/
+
+ ir_unop_noise,
+
+ /**
+ * A sentinel marking the last of the unary operations.
+ */
+ ir_last_unop = ir_unop_noise,
+
+ ir_binop_add,
+ ir_binop_sub,
+ ir_binop_mul,
+ ir_binop_div,
+
+ /**
+ * Takes one of two combinations of arguments:
+ *
+ * - mod(vecN, vecN)
+ * - mod(vecN, float)
+ *
+ * Does not take integer types.
+ */
+ ir_binop_mod,
+
+ /**
+ * \name Binary comparison operators which return a boolean vector.
+ * The type of both operands must be equal.
+ */
+ /*@{*/
+ ir_binop_less,
+ ir_binop_greater,
+ ir_binop_lequal,
+ ir_binop_gequal,
+ ir_binop_equal,
+ ir_binop_nequal,
+ /**
+ * Returns single boolean for whether all components of operands[0]
+ * equal the components of operands[1].
+ */
+ ir_binop_all_equal,
+ /**
+ * Returns single boolean for whether any component of operands[0]
+ * is not equal to the corresponding component of operands[1].
+ */
+ ir_binop_any_nequal,
+ /*@}*/
+
+ /**
+ * \name Bit-wise binary operations.
+ */
+ /*@{*/
+ ir_binop_lshift,
+ ir_binop_rshift,
+ ir_binop_bit_and,
+ ir_binop_bit_xor,
+ ir_binop_bit_or,
+ /*@}*/
+
+ ir_binop_logic_and,
+ ir_binop_logic_xor,
+ ir_binop_logic_or,
+
+ ir_binop_dot,
+ ir_binop_min,
+ ir_binop_max,
+
+ ir_binop_pow,
+
+ /**
+ * A sentinel marking the last of the binary operations.
+ */
+ ir_last_binop = ir_binop_pow,
+
+ ir_quadop_vector,
+
+ /**
+ * A sentinel marking the last of all operations.
+ */
+ ir_last_opcode = ir_last_binop
+};
+
+class ir_expression : public ir_rvalue {
+public:
+ /**
+ * Constructor for unary operation expressions
+ */
+ ir_expression(int op, const struct glsl_type *type, ir_rvalue *);
+ ir_expression(int op, ir_rvalue *);
+
+ /**
+ * Constructor for binary operation expressions
+ */
+ ir_expression(int op, const struct glsl_type *type,
+ ir_rvalue *, ir_rvalue *);
+ ir_expression(int op, ir_rvalue *op0, ir_rvalue *op1);
+
+ /**
+ * Constructor for quad operator expressions
+ */
+ ir_expression(int op, const struct glsl_type *type,
+ ir_rvalue *, ir_rvalue *, ir_rvalue *, ir_rvalue *);
+
+ virtual ir_expression *as_expression()
+ {
+ return this;
+ }
+
+ virtual ir_expression *clone(void *mem_ctx, struct hash_table *ht) const;
+
+ /**
+ * Attempt to constant-fold the expression
+ *
+ * If the expression cannot be constant folded, this method will return
+ * \c NULL.
+ */
+ virtual ir_constant *constant_expression_value();
+
+ /**
+ * Determine the number of operands used by an expression
+ */
+ static unsigned int get_num_operands(ir_expression_operation);
+
+ /**
+ * Determine the number of operands used by an expression
+ */
+ unsigned int get_num_operands() const
+ {
+ return (this->operation == ir_quadop_vector)
+ ? this->type->vector_elements : get_num_operands(operation);
+ }
+
+ /**
+ * Return a string representing this expression's operator.
+ */
+ const char *operator_string();
+
+ /**
+ * Return a string representing this expression's operator.
+ */
+ static const char *operator_string(ir_expression_operation);
+
+
+ /**
+ * Do a reverse-lookup to translate the given string into an operator.
+ */
+ static ir_expression_operation get_operator(const char *);
+
+ virtual void accept(ir_visitor *v)
+ {
+ v->visit(this);
+ }
+
+ virtual ir_visitor_status accept(ir_hierarchical_visitor *);
+
+ ir_expression_operation operation;
+ ir_rvalue *operands[4];
+};
+
+
+/**
+ * IR instruction representing a function call
+ */
+class ir_call : public ir_rvalue {
+public:
+ ir_call(ir_function_signature *callee, exec_list *actual_parameters)
+ : callee(callee)
+ {
+ ir_type = ir_type_call;
+ assert(callee->return_type != NULL);
+ type = callee->return_type;
+ actual_parameters->move_nodes_to(& this->actual_parameters);
+ }
+
+ virtual ir_call *clone(void *mem_ctx, struct hash_table *ht) const;
+
+ virtual ir_constant *constant_expression_value();
+
+ virtual ir_call *as_call()
+ {
+ return this;
+ }
+
+ virtual void accept(ir_visitor *v)
+ {
+ v->visit(this);
+ }
+
+ virtual ir_visitor_status accept(ir_hierarchical_visitor *);
+
+ /**
+ * Get a generic ir_call object when an error occurs
+ *
+ * Any allocation will be performed with 'ctx' as ralloc owner.
+ */
+ static ir_call *get_error_instruction(void *ctx);
+
+ /**
+ * Get an iterator for the set of acutal parameters
+ */
+ exec_list_iterator iterator()
+ {
+ return actual_parameters.iterator();
+ }
+
+ /**
+ * Get the name of the function being called.
+ */
+ const char *callee_name() const
+ {
+ return callee->function_name();
+ }
+
+ /**
+ * Get the function signature bound to this function call
+ */
+ ir_function_signature *get_callee()
+ {
+ return callee;
+ }
+
+ /**
+ * Set the function call target
+ */
+ void set_callee(ir_function_signature *sig);
+
+ /**
+ * Generates an inline version of the function before @ir,
+ * returning the return value of the function.
+ */
+ ir_rvalue *generate_inline(ir_instruction *ir);
+
+ /* List of ir_rvalue of paramaters passed in this call. */
+ exec_list actual_parameters;
+
+private:
+ ir_call()
+ : callee(NULL)
+ {
+ this->ir_type = ir_type_call;
+ }
+
+ ir_function_signature *callee;
+};
+
+
+/**
+ * \name Jump-like IR instructions.
+ *
+ * These include \c break, \c continue, \c return, and \c discard.
+ */
+/*@{*/
+class ir_jump : public ir_instruction {
+protected:
+ ir_jump()
+ {
+ ir_type = ir_type_unset;
+ }
+};
+
+class ir_return : public ir_jump {
+public:
+ ir_return()
+ : value(NULL)
+ {
+ this->ir_type = ir_type_return;
+ }
+
+ ir_return(ir_rvalue *value)
+ : value(value)
+ {
+ this->ir_type = ir_type_return;
+ }
+
+ virtual ir_return *clone(void *mem_ctx, struct hash_table *) const;
+
+ virtual ir_return *as_return()
+ {
+ return this;
+ }
+
+ ir_rvalue *get_value() const
+ {
+ return value;
+ }
+
+ virtual void accept(ir_visitor *v)
+ {
+ v->visit(this);
+ }
+
+ virtual ir_visitor_status accept(ir_hierarchical_visitor *);
+
+ ir_rvalue *value;
+};
+
+
+/**
+ * Jump instructions used inside loops
+ *
+ * These include \c break and \c continue. The \c break within a loop is
+ * different from the \c break within a switch-statement.
+ *
+ * \sa ir_switch_jump
+ */
+class ir_loop_jump : public ir_jump {
+public:
+ enum jump_mode {
+ jump_break,
+ jump_continue
+ };
+
+ ir_loop_jump(jump_mode mode)
+ {
+ this->ir_type = ir_type_loop_jump;
+ this->mode = mode;
+ this->loop = loop;
+ }
+
+ virtual ir_loop_jump *clone(void *mem_ctx, struct hash_table *) const;
+
+ virtual void accept(ir_visitor *v)
+ {
+ v->visit(this);
+ }
+
+ virtual ir_visitor_status accept(ir_hierarchical_visitor *);
+
+ bool is_break() const
+ {
+ return mode == jump_break;
+ }
+
+ bool is_continue() const
+ {
+ return mode == jump_continue;
+ }
+
+ /** Mode selector for the jump instruction. */
+ enum jump_mode mode;
+private:
+ /** Loop containing this break instruction. */
+ ir_loop *loop;
+};
+
+/**
+ * IR instruction representing discard statements.
+ */
+class ir_discard : public ir_jump {
+public:
+ ir_discard()
+ {
+ this->ir_type = ir_type_discard;
+ this->condition = NULL;
+ }
+
+ ir_discard(ir_rvalue *cond)
+ {
+ this->ir_type = ir_type_discard;
+ this->condition = cond;
+ }
+
+ virtual ir_discard *clone(void *mem_ctx, struct hash_table *ht) const;
+
+ virtual void accept(ir_visitor *v)
+ {
+ v->visit(this);
+ }
+
+ virtual ir_visitor_status accept(ir_hierarchical_visitor *);
+
+ virtual ir_discard *as_discard()
+ {
+ return this;
+ }
+
+ ir_rvalue *condition;
+};
+/*@}*/
+
+
+/**
+ * Texture sampling opcodes used in ir_texture
+ */
+enum ir_texture_opcode {
+ ir_tex, /**< Regular texture look-up */
+ ir_txb, /**< Texture look-up with LOD bias */
+ ir_txl, /**< Texture look-up with explicit LOD */
+ ir_txd, /**< Texture look-up with partial derivatvies */
+ ir_txf /**< Texel fetch with explicit LOD */
+};
+
+
+/**
+ * IR instruction to sample a texture
+ *
+ * The specific form of the IR instruction depends on the \c mode value
+ * selected from \c ir_texture_opcodes. In the printed IR, these will
+ * appear as:
+ *
+ * Texel offset (0 or an expression)
+ * | Projection divisor
+ * | | Shadow comparitor
+ * | | |
+ * v v v
+ * (tex <sampler> <coordinate> 0 1 ( ))
+ * (txb <sampler> <coordinate> 0 1 ( ) <bias>)
+ * (txl <sampler> <coordinate> 0 1 ( ) <lod>)
+ * (txd <sampler> <coordinate> 0 1 ( ) (dPdx dPdy))
+ * (txf <sampler> <coordinate> 0 <lod>)
+ */
+class ir_texture : public ir_rvalue {
+public:
+ ir_texture(enum ir_texture_opcode op)
+ : op(op), projector(NULL), shadow_comparitor(NULL), offset(NULL)
+ {
+ this->ir_type = ir_type_texture;
+ }
+
+ virtual ir_texture *clone(void *mem_ctx, struct hash_table *) const;
+
+ virtual ir_constant *constant_expression_value();
+
+ virtual void accept(ir_visitor *v)
+ {
+ v->visit(this);
+ }
+
+ virtual ir_visitor_status accept(ir_hierarchical_visitor *);
+
+ /**
+ * Return a string representing the ir_texture_opcode.
+ */
+ const char *opcode_string();
+
+ /** Set the sampler and infer the type. */
+ void set_sampler(ir_dereference *sampler);
+
+ /**
+ * Do a reverse-lookup to translate a string into an ir_texture_opcode.
+ */
+ static ir_texture_opcode get_opcode(const char *);
+
+ enum ir_texture_opcode op;
+
+ /** Sampler to use for the texture access. */
+ ir_dereference *sampler;
+
+ /** Texture coordinate to sample */
+ ir_rvalue *coordinate;
+
+ /**
+ * Value used for projective divide.
+ *
+ * If there is no projective divide (the common case), this will be
+ * \c NULL. Optimization passes should check for this to point to a constant
+ * of 1.0 and replace that with \c NULL.
+ */
+ ir_rvalue *projector;
+
+ /**
+ * Coordinate used for comparison on shadow look-ups.
+ *
+ * If there is no shadow comparison, this will be \c NULL. For the
+ * \c ir_txf opcode, this *must* be \c NULL.
+ */
+ ir_rvalue *shadow_comparitor;
+
+ /** Texel offset. */
+ ir_rvalue *offset;
+
+ union {
+ ir_rvalue *lod; /**< Floating point LOD */
+ ir_rvalue *bias; /**< Floating point LOD bias */
+ struct {
+ ir_rvalue *dPdx; /**< Partial derivative of coordinate wrt X */
+ ir_rvalue *dPdy; /**< Partial derivative of coordinate wrt Y */
+ } grad;
+ } lod_info;
+};
+
+
+struct ir_swizzle_mask {
+ unsigned x:2;
+ unsigned y:2;
+ unsigned z:2;
+ unsigned w:2;
+
+ /**
+ * Number of components in the swizzle.
+ */
+ unsigned num_components:3;
+
+ /**
+ * Does the swizzle contain duplicate components?
+ *
+ * L-value swizzles cannot contain duplicate components.
+ */
+ unsigned has_duplicates:1;
+};
+
+
+class ir_swizzle : public ir_rvalue {
+public:
+ ir_swizzle(ir_rvalue *, unsigned x, unsigned y, unsigned z, unsigned w,
+ unsigned count);
+
+ ir_swizzle(ir_rvalue *val, const unsigned *components, unsigned count);
+
+ ir_swizzle(ir_rvalue *val, ir_swizzle_mask mask);
+
+ virtual ir_swizzle *clone(void *mem_ctx, struct hash_table *) const;
+
+ virtual ir_constant *constant_expression_value();
+
+ virtual ir_swizzle *as_swizzle()
+ {
+ return this;
+ }
+
+ /**
+ * Construct an ir_swizzle from the textual representation. Can fail.
+ */
+ static ir_swizzle *create(ir_rvalue *, const char *, unsigned vector_length);
+
+ virtual void accept(ir_visitor *v)
+ {
+ v->visit(this);
+ }
+
+ virtual ir_visitor_status accept(ir_hierarchical_visitor *);
+
+ bool is_lvalue()
+ {
+ return val->is_lvalue() && !mask.has_duplicates;
+ }
+
+ /**
+ * Get the variable that is ultimately referenced by an r-value
+ */
+ virtual ir_variable *variable_referenced();
+
+ ir_rvalue *val;
+ ir_swizzle_mask mask;
+
+private:
+ /**
+ * Initialize the mask component of a swizzle
+ *
+ * This is used by the \c ir_swizzle constructors.
+ */
+ void init_mask(const unsigned *components, unsigned count);
+};
+
+
+class ir_dereference : public ir_rvalue {
+public:
+ virtual ir_dereference *clone(void *mem_ctx, struct hash_table *) const = 0;
+
+ virtual ir_dereference *as_dereference()
+ {
+ return this;
+ }
+
+ bool is_lvalue();
+
+ /**
+ * Get the variable that is ultimately referenced by an r-value
+ */
+ virtual ir_variable *variable_referenced() = 0;
+};
+
+
+class ir_dereference_variable : public ir_dereference {
+public:
+ ir_dereference_variable(ir_variable *var);
+
+ virtual ir_dereference_variable *clone(void *mem_ctx,
+ struct hash_table *) const;
+
+ virtual ir_constant *constant_expression_value();
+
+ virtual ir_dereference_variable *as_dereference_variable()
+ {
+ return this;
+ }
+
+ /**
+ * Get the variable that is ultimately referenced by an r-value
+ */
+ virtual ir_variable *variable_referenced()
+ {
+ return this->var;
+ }
+
+ virtual ir_variable *whole_variable_referenced()
+ {
+ /* ir_dereference_variable objects always dereference the entire
+ * variable. However, if this dereference is dereferenced by anything
+ * else, the complete deferefernce chain is not a whole-variable
+ * dereference. This method should only be called on the top most
+ * ir_rvalue in a dereference chain.
+ */
+ return this->var;
+ }
+
+ virtual void accept(ir_visitor *v)
+ {
+ v->visit(this);
+ }
+
+ virtual ir_visitor_status accept(ir_hierarchical_visitor *);
+
+ /**
+ * Object being dereferenced.
+ */
+ ir_variable *var;
+};
+
+
+class ir_dereference_array : public ir_dereference {
+public:
+ ir_dereference_array(ir_rvalue *value, ir_rvalue *array_index);
+
+ ir_dereference_array(ir_variable *var, ir_rvalue *array_index);
+
+ virtual ir_dereference_array *clone(void *mem_ctx,
+ struct hash_table *) const;
+
+ virtual ir_constant *constant_expression_value();
+
+ virtual ir_dereference_array *as_dereference_array()
+ {
+ return this;
+ }
+
+ /**
+ * Get the variable that is ultimately referenced by an r-value
+ */
+ virtual ir_variable *variable_referenced()
+ {
+ return this->array->variable_referenced();
+ }
+
+ virtual void accept(ir_visitor *v)
+ {
+ v->visit(this);
+ }
+
+ virtual ir_visitor_status accept(ir_hierarchical_visitor *);
+
+ ir_rvalue *array;
+ ir_rvalue *array_index;
+
+private:
+ void set_array(ir_rvalue *value);
+};
+
+
+class ir_dereference_record : public ir_dereference {
+public:
+ ir_dereference_record(ir_rvalue *value, const char *field);
+
+ ir_dereference_record(ir_variable *var, const char *field);
+
+ virtual ir_dereference_record *clone(void *mem_ctx,
+ struct hash_table *) const;
+
+ virtual ir_constant *constant_expression_value();
+
+ /**
+ * Get the variable that is ultimately referenced by an r-value
+ */
+ virtual ir_variable *variable_referenced()
+ {
+ return this->record->variable_referenced();
+ }
+
+ virtual void accept(ir_visitor *v)
+ {
+ v->visit(this);
+ }
+
+ virtual ir_visitor_status accept(ir_hierarchical_visitor *);
+
+ ir_rvalue *record;
+ const char *field;
+};
+
+
+/**
+ * Data stored in an ir_constant
+ */
+union ir_constant_data {
+ unsigned u[16];
+ int i[16];
+ float f[16];
+ bool b[16];
+};
+
+
+class ir_constant : public ir_rvalue {
+public:
+ ir_constant(const struct glsl_type *type, const ir_constant_data *data);
+ ir_constant(bool b);
+ ir_constant(unsigned int u);
+ ir_constant(int i);
+ ir_constant(float f);
+
+ /**
+ * Construct an ir_constant from a list of ir_constant values
+ */
+ ir_constant(const struct glsl_type *type, exec_list *values);
+
+ /**
+ * Construct an ir_constant from a scalar component of another ir_constant
+ *
+ * The new \c ir_constant inherits the type of the component from the
+ * source constant.
+ *
+ * \note
+ * In the case of a matrix constant, the new constant is a scalar, \b not
+ * a vector.
+ */
+ ir_constant(const ir_constant *c, unsigned i);
+
+ /**
+ * Return a new ir_constant of the specified type containing all zeros.
+ */
+ static ir_constant *zero(void *mem_ctx, const glsl_type *type);
+
+ virtual ir_constant *clone(void *mem_ctx, struct hash_table *) const;
+
+ virtual ir_constant *constant_expression_value();
+
+ virtual ir_constant *as_constant()
+ {
+ return this;
+ }
+
+ virtual void accept(ir_visitor *v)
+ {
+ v->visit(this);
+ }
+
+ virtual ir_visitor_status accept(ir_hierarchical_visitor *);
+
+ /**
+ * Get a particular component of a constant as a specific type
+ *
+ * This is useful, for example, to get a value from an integer constant
+ * as a float or bool. This appears frequently when constructors are
+ * called with all constant parameters.
+ */
+ /*@{*/
+ bool get_bool_component(unsigned i) const;
+ float get_float_component(unsigned i) const;
+ int get_int_component(unsigned i) const;
+ unsigned get_uint_component(unsigned i) const;
+ /*@}*/
+
+ ir_constant *get_array_element(unsigned i) const;
+
+ ir_constant *get_record_field(const char *name);
+
+ /**
+ * Determine whether a constant has the same value as another constant
+ *
+ * \sa ir_constant::is_zero, ir_constant::is_one,
+ * ir_constant::is_negative_one
+ */
+ bool has_value(const ir_constant *) const;
+
+ virtual bool is_zero() const;
+ virtual bool is_one() const;
+ virtual bool is_negative_one() const;
+
+ /**
+ * Value of the constant.
+ *
+ * The field used to back the values supplied by the constant is determined
+ * by the type associated with the \c ir_instruction. Constants may be
+ * scalars, vectors, or matrices.
+ */
+ union ir_constant_data value;
+
+ /* Array elements */
+ ir_constant **array_elements;
+
+ /* Structure fields */
+ exec_list components;
+
+private:
+ /**
+ * Parameterless constructor only used by the clone method
+ */
+ ir_constant(void);
+};
+
+/*@}*/
+
+/**
+ * Apply a visitor to each IR node in a list
+ */
+void
+visit_exec_list(exec_list *list, ir_visitor *visitor);
+
+/**
+ * Validate invariants on each IR node in a list
+ */
+void validate_ir_tree(exec_list *instructions);
+
+/**
+ * Make a clone of each IR instruction in a list
+ *
+ * \param in List of IR instructions that are to be cloned
+ * \param out List to hold the cloned instructions
+ */
+void
+clone_ir_list(void *mem_ctx, exec_list *out, const exec_list *in);
+
+extern void
+_mesa_glsl_initialize_variables(exec_list *instructions,
+ struct _mesa_glsl_parse_state *state);
+
+extern void
+_mesa_glsl_initialize_functions(_mesa_glsl_parse_state *state);
+
+extern void
+_mesa_glsl_release_functions(void);
+
+extern void
+reparent_ir(exec_list *list, void *mem_ctx);
+
+struct glsl_symbol_table;
+
+extern void
+import_prototypes(const exec_list *source, exec_list *dest,
+ struct glsl_symbol_table *symbols, void *mem_ctx);
+
+extern bool
+ir_has_call(ir_instruction *ir);
+
+extern void
+do_set_program_inouts(exec_list *instructions, struct gl_program *prog);
+
+#endif /* IR_H */
diff --git a/mesalib/src/glsl/ir_import_prototypes.cpp b/mesalib/src/glsl/ir_import_prototypes.cpp index be5e0c1d3..03d40bb68 100644 --- a/mesalib/src/glsl/ir_import_prototypes.cpp +++ b/mesalib/src/glsl/ir_import_prototypes.cpp @@ -1,123 +1,122 @@ -/* - * Copyright © 2010 Intel Corporation - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING - * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER - * DEALINGS IN THE SOFTWARE. - */ - -/** - * \file ir_import_prototypes.cpp - * Import function prototypes from one IR tree into another. - * - * \author Ian Romanick - */ -#include <cstdio> -#include "ir.h" -#include "glsl_symbol_table.h" - -/** - * Visitor used to import function prototypes - * - * Normally the \c clone method of either \c ir_function or - * \c ir_function_signature could be used. However, we don't want a complete - * clone of the \c ir_function_signature. We want everything \b except the - * body of the function. - */ -class import_prototype_visitor : public ir_hierarchical_visitor { -public: - /** - */ - import_prototype_visitor(exec_list *list, glsl_symbol_table *symbols, - void *mem_ctx) - { - this->mem_ctx = mem_ctx; - this->list = list; - this->symbols = symbols; - this->function = NULL; - } - - virtual ir_visitor_status visit_enter(ir_function *ir) - { - assert(this->function == NULL); - - this->function = this->symbols->get_function(ir->name); - if (!this->function) { - this->function = new(this->mem_ctx) ir_function(ir->name); - - list->push_tail(this->function); - - /* Add the new function to the symbol table. - */ - this->symbols->add_function(this->function); - } - return visit_continue; - } - - virtual ir_visitor_status visit_leave(ir_function *ir) - { - (void) ir; - assert(this->function != NULL); - - this->function = NULL; - return visit_continue; - } - - ir_visitor_status visit_enter(ir_function_signature *ir) - { - assert(this->function != NULL); - - ir_function_signature *copy = ir->clone_prototype(mem_ctx, NULL); - - this->function->add_signature(copy); - - /* Do not process child nodes of the ir_function_signature. There can - * never be any nodes inside the ir_function_signature that we care - * about. Instead continue with the next sibling. - */ - return visit_continue_with_parent; - } - -private: - exec_list *list; - ir_function *function; - glsl_symbol_table *symbols; - void *mem_ctx; -}; - - -/** - * Import function prototypes from one IR tree into another - * - * \param source Source instruction stream containing functions whose - * prototypes are to be imported - * \param dest Destination instruction stream where new \c ir_function and - * \c ir_function_signature nodes will be stored - * \param symbols Symbol table where new functions will be stored - * \param mem_ctx ralloc memory context used for new allocations - */ -void -import_prototypes(const exec_list *source, exec_list *dest, - glsl_symbol_table *symbols, void *mem_ctx) -{ - import_prototype_visitor v(dest, symbols, mem_ctx); - - /* Making source be const is just extra documentation. - */ - v.run(const_cast<exec_list *>(source)); -} +/*
+ * Copyright © 2010 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file ir_import_prototypes.cpp
+ * Import function prototypes from one IR tree into another.
+ *
+ * \author Ian Romanick
+ */
+#include "ir.h"
+#include "glsl_symbol_table.h"
+
+/**
+ * Visitor used to import function prototypes
+ *
+ * Normally the \c clone method of either \c ir_function or
+ * \c ir_function_signature could be used. However, we don't want a complete
+ * clone of the \c ir_function_signature. We want everything \b except the
+ * body of the function.
+ */
+class import_prototype_visitor : public ir_hierarchical_visitor {
+public:
+ /**
+ */
+ import_prototype_visitor(exec_list *list, glsl_symbol_table *symbols,
+ void *mem_ctx)
+ {
+ this->mem_ctx = mem_ctx;
+ this->list = list;
+ this->symbols = symbols;
+ this->function = NULL;
+ }
+
+ virtual ir_visitor_status visit_enter(ir_function *ir)
+ {
+ assert(this->function == NULL);
+
+ this->function = this->symbols->get_function(ir->name);
+ if (!this->function) {
+ this->function = new(this->mem_ctx) ir_function(ir->name);
+
+ list->push_tail(this->function);
+
+ /* Add the new function to the symbol table.
+ */
+ this->symbols->add_function(this->function);
+ }
+ return visit_continue;
+ }
+
+ virtual ir_visitor_status visit_leave(ir_function *ir)
+ {
+ (void) ir;
+ assert(this->function != NULL);
+
+ this->function = NULL;
+ return visit_continue;
+ }
+
+ ir_visitor_status visit_enter(ir_function_signature *ir)
+ {
+ assert(this->function != NULL);
+
+ ir_function_signature *copy = ir->clone_prototype(mem_ctx, NULL);
+
+ this->function->add_signature(copy);
+
+ /* Do not process child nodes of the ir_function_signature. There can
+ * never be any nodes inside the ir_function_signature that we care
+ * about. Instead continue with the next sibling.
+ */
+ return visit_continue_with_parent;
+ }
+
+private:
+ exec_list *list;
+ ir_function *function;
+ glsl_symbol_table *symbols;
+ void *mem_ctx;
+};
+
+
+/**
+ * Import function prototypes from one IR tree into another
+ *
+ * \param source Source instruction stream containing functions whose
+ * prototypes are to be imported
+ * \param dest Destination instruction stream where new \c ir_function and
+ * \c ir_function_signature nodes will be stored
+ * \param symbols Symbol table where new functions will be stored
+ * \param mem_ctx ralloc memory context used for new allocations
+ */
+void
+import_prototypes(const exec_list *source, exec_list *dest,
+ glsl_symbol_table *symbols, void *mem_ctx)
+{
+ import_prototype_visitor v(dest, symbols, mem_ctx);
+
+ /* Making source be const is just extra documentation.
+ */
+ v.run(const_cast<exec_list *>(source));
+}
diff --git a/mesalib/src/glsl/ir_print_visitor.cpp b/mesalib/src/glsl/ir_print_visitor.cpp index be76945a2..f7dc29689 100644 --- a/mesalib/src/glsl/ir_print_visitor.cpp +++ b/mesalib/src/glsl/ir_print_visitor.cpp @@ -1,474 +1,471 @@ -/* - * Copyright © 2010 Intel Corporation - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING - * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER - * DEALINGS IN THE SOFTWARE. - */ - -#include "ir_print_visitor.h" -#include "glsl_types.h" -#include "glsl_parser_extras.h" - -using std::printf; -using std::strncmp; - -static void print_type(const glsl_type *t); - -void -ir_instruction::print(void) const -{ - ir_instruction *deconsted = const_cast<ir_instruction *>(this); - - ir_print_visitor v; - deconsted->accept(&v); -} - -void -_mesa_print_ir(exec_list *instructions, - struct _mesa_glsl_parse_state *state) -{ - if (state) { - for (unsigned i = 0; i < state->num_user_structures; i++) { - const glsl_type *const s = state->user_structures[i]; - - printf("(structure (%s) (%s@%p) (%u) (\n", - s->name, s->name, (void *) s, s->length); - - for (unsigned j = 0; j < s->length; j++) { - printf("\t(("); - print_type(s->fields.structure[j].type); - printf(")(%s))\n", s->fields.structure[j].name); - } - - printf(")\n"); - } - } - - printf("(\n"); - foreach_iter(exec_list_iterator, iter, *instructions) { - ir_instruction *ir = (ir_instruction *)iter.get(); - ir->print(); - if (ir->ir_type != ir_type_function) - printf("\n"); - } - printf("\n)"); -} - - -void ir_print_visitor::indent(void) -{ - for (int i = 0; i < indentation; i++) - printf(" "); -} - -static void -print_type(const glsl_type *t) -{ - if (t->base_type == GLSL_TYPE_ARRAY) { - printf("(array "); - print_type(t->fields.array); - printf(" %u)", t->length); - } else if ((t->base_type == GLSL_TYPE_STRUCT) - && (strncmp("gl_", t->name, 3) != 0)) { - printf("%s@%p", t->name, (void *) t); - } else { - printf("%s", t->name); - } -} - - -void ir_print_visitor::visit(ir_variable *ir) -{ - printf("(declare "); - - const char *const cent = (ir->centroid) ? "centroid " : ""; - const char *const inv = (ir->invariant) ? "invariant " : ""; - const char *const mode[] = { "", "uniform ", "in ", "out ", "inout ", - "const_in ", "sys ", "temporary " }; - const char *const interp[] = { "", "flat", "noperspective" }; - - printf("(%s%s%s%s) ", - cent, inv, mode[ir->mode], interp[ir->interpolation]); - - print_type(ir->type); - printf(" %s@%p)", ir->name, (void *) ir); -} - - -void ir_print_visitor::visit(ir_function_signature *ir) -{ - printf("(signature "); - indentation++; - - print_type(ir->return_type); - printf("\n"); - indent(); - - printf("(parameters\n"); - indentation++; - - foreach_iter(exec_list_iterator, iter, ir->parameters) { - ir_variable *const inst = (ir_variable *) iter.get(); - - indent(); - inst->accept(this); - printf("\n"); - } - indentation--; - - indent(); - printf(")\n"); - - indent(); - - printf("(\n"); - indentation++; - - foreach_iter(exec_list_iterator, iter, ir->body) { - ir_instruction *const inst = (ir_instruction *) iter.get(); - - indent(); - inst->accept(this); - printf("\n"); - } - indentation--; - indent(); - printf("))\n"); - indentation--; -} - - -void ir_print_visitor::visit(ir_function *ir) -{ - printf("(function %s\n", ir->name); - indentation++; - foreach_iter(exec_list_iterator, iter, *ir) { - ir_function_signature *const sig = (ir_function_signature *) iter.get(); - indent(); - sig->accept(this); - printf("\n"); - } - indentation--; - indent(); - printf(")\n\n"); -} - - -void ir_print_visitor::visit(ir_expression *ir) -{ - printf("(expression "); - - print_type(ir->type); - - printf(" %s ", ir->operator_string()); - - for (unsigned i = 0; i < ir->get_num_operands(); i++) { - ir->operands[i]->accept(this); - } - - printf(") "); -} - - -void ir_print_visitor::visit(ir_texture *ir) -{ - printf("(%s ", ir->opcode_string()); - - ir->sampler->accept(this); - printf(" "); - - ir->coordinate->accept(this); - - printf(" "); - - if (ir->offset != NULL) { - ir->offset->accept(this); - } else { - printf("0"); - } - - printf(" "); - - if (ir->op != ir_txf) { - if (ir->projector) - ir->projector->accept(this); - else - printf("1"); - - if (ir->shadow_comparitor) { - printf(" "); - ir->shadow_comparitor->accept(this); - } else { - printf(" ()"); - } - } - - printf(" "); - switch (ir->op) - { - case ir_tex: - break; - case ir_txb: - ir->lod_info.bias->accept(this); - break; - case ir_txl: - case ir_txf: - ir->lod_info.lod->accept(this); - break; - case ir_txd: - printf("("); - ir->lod_info.grad.dPdx->accept(this); - printf(" "); - ir->lod_info.grad.dPdy->accept(this); - printf(")"); - break; - }; - printf(")"); -} - - -void ir_print_visitor::visit(ir_swizzle *ir) -{ - const unsigned swiz[4] = { - ir->mask.x, - ir->mask.y, - ir->mask.z, - ir->mask.w, - }; - - printf("(swiz "); - for (unsigned i = 0; i < ir->mask.num_components; i++) { - printf("%c", "xyzw"[swiz[i]]); - } - printf(" "); - ir->val->accept(this); - printf(")"); -} - - -void ir_print_visitor::visit(ir_dereference_variable *ir) -{ - ir_variable *var = ir->variable_referenced(); - printf("(var_ref %s@%p) ", var->name, (void *) var); -} - - -void ir_print_visitor::visit(ir_dereference_array *ir) -{ - printf("(array_ref "); - ir->array->accept(this); - ir->array_index->accept(this); - printf(") "); -} - - -void ir_print_visitor::visit(ir_dereference_record *ir) -{ - printf("(record_ref "); - ir->record->accept(this); - printf(" %s) ", ir->field); -} - - -void ir_print_visitor::visit(ir_assignment *ir) -{ - printf("(assign "); - - if (ir->condition) - ir->condition->accept(this); - - char mask[5]; - unsigned j = 0; - - for (unsigned i = 0; i < 4; i++) { - if ((ir->write_mask & (1 << i)) != 0) { - mask[j] = "xyzw"[i]; - j++; - } - } - mask[j] = '\0'; - - printf(" (%s) ", mask); - - ir->lhs->accept(this); - - printf(" "); - - ir->rhs->accept(this); - printf(") "); -} - - -void ir_print_visitor::visit(ir_constant *ir) -{ - const glsl_type *const base_type = ir->type->get_base_type(); - - printf("(constant "); - print_type(ir->type); - printf(" ("); - - if (ir->type->is_array()) { - for (unsigned i = 0; i < ir->type->length; i++) - ir->get_array_element(i)->accept(this); - } else if (ir->type->is_record()) { - ir_constant *value = (ir_constant *) ir->components.get_head(); - for (unsigned i = 0; i < ir->type->length; i++) { - printf("(%s ", ir->type->fields.structure->name); - value->accept(this); - printf(")"); - - value = (ir_constant *) value->next; - } - } else { - for (unsigned i = 0; i < ir->type->components(); i++) { - if (i != 0) - printf(" "); - switch (base_type->base_type) { - case GLSL_TYPE_UINT: printf("%u", ir->value.u[i]); break; - case GLSL_TYPE_INT: printf("%d", ir->value.i[i]); break; - case GLSL_TYPE_FLOAT: printf("%f", ir->value.f[i]); break; - case GLSL_TYPE_BOOL: printf("%d", ir->value.b[i]); break; - default: assert(0); - } - } - } - printf(")) "); -} - - -void -ir_print_visitor::visit(ir_call *ir) -{ - printf("(call %s (", ir->callee_name()); - foreach_iter(exec_list_iterator, iter, *ir) { - ir_instruction *const inst = (ir_instruction *) iter.get(); - - inst->accept(this); - } - printf("))\n"); -} - - -void -ir_print_visitor::visit(ir_return *ir) -{ - printf("(return"); - - ir_rvalue *const value = ir->get_value(); - if (value) { - printf(" "); - value->accept(this); - } - - printf(")"); -} - - -void -ir_print_visitor::visit(ir_discard *ir) -{ - printf("(discard "); - - if (ir->condition != NULL) { - printf(" "); - ir->condition->accept(this); - } - - printf(")"); -} - - -void -ir_print_visitor::visit(ir_if *ir) -{ - printf("(if "); - ir->condition->accept(this); - - printf("(\n"); - indentation++; - - foreach_iter(exec_list_iterator, iter, ir->then_instructions) { - ir_instruction *const inst = (ir_instruction *) iter.get(); - - indent(); - inst->accept(this); - printf("\n"); - } - - indentation--; - indent(); - printf(")\n"); - - indent(); - if (!ir->else_instructions.is_empty()) { - printf("(\n"); - indentation++; - - foreach_iter(exec_list_iterator, iter, ir->else_instructions) { - ir_instruction *const inst = (ir_instruction *) iter.get(); - - indent(); - inst->accept(this); - printf("\n"); - } - indentation--; - indent(); - printf("))\n"); - } else { - printf("())\n"); - } -} - - -void -ir_print_visitor::visit(ir_loop *ir) -{ - printf("(loop ("); - if (ir->counter != NULL) - ir->counter->accept(this); - printf(") ("); - if (ir->from != NULL) - ir->from->accept(this); - printf(") ("); - if (ir->to != NULL) - ir->to->accept(this); - printf(") ("); - if (ir->increment != NULL) - ir->increment->accept(this); - printf(") (\n"); - indentation++; - - foreach_iter(exec_list_iterator, iter, ir->body_instructions) { - ir_instruction *const inst = (ir_instruction *) iter.get(); - - indent(); - inst->accept(this); - printf("\n"); - } - indentation--; - indent(); - printf("))\n"); -} - - -void -ir_print_visitor::visit(ir_loop_jump *ir) -{ - printf("%s", ir->is_break() ? "break" : "continue"); -} +/*
+ * Copyright © 2010 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include "ir_print_visitor.h"
+#include "glsl_types.h"
+#include "glsl_parser_extras.h"
+
+static void print_type(const glsl_type *t);
+
+void
+ir_instruction::print(void) const
+{
+ ir_instruction *deconsted = const_cast<ir_instruction *>(this);
+
+ ir_print_visitor v;
+ deconsted->accept(&v);
+}
+
+void
+_mesa_print_ir(exec_list *instructions,
+ struct _mesa_glsl_parse_state *state)
+{
+ if (state) {
+ for (unsigned i = 0; i < state->num_user_structures; i++) {
+ const glsl_type *const s = state->user_structures[i];
+
+ printf("(structure (%s) (%s@%p) (%u) (\n",
+ s->name, s->name, (void *) s, s->length);
+
+ for (unsigned j = 0; j < s->length; j++) {
+ printf("\t((");
+ print_type(s->fields.structure[j].type);
+ printf(")(%s))\n", s->fields.structure[j].name);
+ }
+
+ printf(")\n");
+ }
+ }
+
+ printf("(\n");
+ foreach_iter(exec_list_iterator, iter, *instructions) {
+ ir_instruction *ir = (ir_instruction *)iter.get();
+ ir->print();
+ if (ir->ir_type != ir_type_function)
+ printf("\n");
+ }
+ printf("\n)");
+}
+
+
+void ir_print_visitor::indent(void)
+{
+ for (int i = 0; i < indentation; i++)
+ printf(" ");
+}
+
+static void
+print_type(const glsl_type *t)
+{
+ if (t->base_type == GLSL_TYPE_ARRAY) {
+ printf("(array ");
+ print_type(t->fields.array);
+ printf(" %u)", t->length);
+ } else if ((t->base_type == GLSL_TYPE_STRUCT)
+ && (strncmp("gl_", t->name, 3) != 0)) {
+ printf("%s@%p", t->name, (void *) t);
+ } else {
+ printf("%s", t->name);
+ }
+}
+
+
+void ir_print_visitor::visit(ir_variable *ir)
+{
+ printf("(declare ");
+
+ const char *const cent = (ir->centroid) ? "centroid " : "";
+ const char *const inv = (ir->invariant) ? "invariant " : "";
+ const char *const mode[] = { "", "uniform ", "in ", "out ", "inout ",
+ "const_in ", "sys ", "temporary " };
+ const char *const interp[] = { "", "flat", "noperspective" };
+
+ printf("(%s%s%s%s) ",
+ cent, inv, mode[ir->mode], interp[ir->interpolation]);
+
+ print_type(ir->type);
+ printf(" %s@%p)", ir->name, (void *) ir);
+}
+
+
+void ir_print_visitor::visit(ir_function_signature *ir)
+{
+ printf("(signature ");
+ indentation++;
+
+ print_type(ir->return_type);
+ printf("\n");
+ indent();
+
+ printf("(parameters\n");
+ indentation++;
+
+ foreach_iter(exec_list_iterator, iter, ir->parameters) {
+ ir_variable *const inst = (ir_variable *) iter.get();
+
+ indent();
+ inst->accept(this);
+ printf("\n");
+ }
+ indentation--;
+
+ indent();
+ printf(")\n");
+
+ indent();
+
+ printf("(\n");
+ indentation++;
+
+ foreach_iter(exec_list_iterator, iter, ir->body) {
+ ir_instruction *const inst = (ir_instruction *) iter.get();
+
+ indent();
+ inst->accept(this);
+ printf("\n");
+ }
+ indentation--;
+ indent();
+ printf("))\n");
+ indentation--;
+}
+
+
+void ir_print_visitor::visit(ir_function *ir)
+{
+ printf("(function %s\n", ir->name);
+ indentation++;
+ foreach_iter(exec_list_iterator, iter, *ir) {
+ ir_function_signature *const sig = (ir_function_signature *) iter.get();
+ indent();
+ sig->accept(this);
+ printf("\n");
+ }
+ indentation--;
+ indent();
+ printf(")\n\n");
+}
+
+
+void ir_print_visitor::visit(ir_expression *ir)
+{
+ printf("(expression ");
+
+ print_type(ir->type);
+
+ printf(" %s ", ir->operator_string());
+
+ for (unsigned i = 0; i < ir->get_num_operands(); i++) {
+ ir->operands[i]->accept(this);
+ }
+
+ printf(") ");
+}
+
+
+void ir_print_visitor::visit(ir_texture *ir)
+{
+ printf("(%s ", ir->opcode_string());
+
+ ir->sampler->accept(this);
+ printf(" ");
+
+ ir->coordinate->accept(this);
+
+ printf(" ");
+
+ if (ir->offset != NULL) {
+ ir->offset->accept(this);
+ } else {
+ printf("0");
+ }
+
+ printf(" ");
+
+ if (ir->op != ir_txf) {
+ if (ir->projector)
+ ir->projector->accept(this);
+ else
+ printf("1");
+
+ if (ir->shadow_comparitor) {
+ printf(" ");
+ ir->shadow_comparitor->accept(this);
+ } else {
+ printf(" ()");
+ }
+ }
+
+ printf(" ");
+ switch (ir->op)
+ {
+ case ir_tex:
+ break;
+ case ir_txb:
+ ir->lod_info.bias->accept(this);
+ break;
+ case ir_txl:
+ case ir_txf:
+ ir->lod_info.lod->accept(this);
+ break;
+ case ir_txd:
+ printf("(");
+ ir->lod_info.grad.dPdx->accept(this);
+ printf(" ");
+ ir->lod_info.grad.dPdy->accept(this);
+ printf(")");
+ break;
+ };
+ printf(")");
+}
+
+
+void ir_print_visitor::visit(ir_swizzle *ir)
+{
+ const unsigned swiz[4] = {
+ ir->mask.x,
+ ir->mask.y,
+ ir->mask.z,
+ ir->mask.w,
+ };
+
+ printf("(swiz ");
+ for (unsigned i = 0; i < ir->mask.num_components; i++) {
+ printf("%c", "xyzw"[swiz[i]]);
+ }
+ printf(" ");
+ ir->val->accept(this);
+ printf(")");
+}
+
+
+void ir_print_visitor::visit(ir_dereference_variable *ir)
+{
+ ir_variable *var = ir->variable_referenced();
+ printf("(var_ref %s@%p) ", var->name, (void *) var);
+}
+
+
+void ir_print_visitor::visit(ir_dereference_array *ir)
+{
+ printf("(array_ref ");
+ ir->array->accept(this);
+ ir->array_index->accept(this);
+ printf(") ");
+}
+
+
+void ir_print_visitor::visit(ir_dereference_record *ir)
+{
+ printf("(record_ref ");
+ ir->record->accept(this);
+ printf(" %s) ", ir->field);
+}
+
+
+void ir_print_visitor::visit(ir_assignment *ir)
+{
+ printf("(assign ");
+
+ if (ir->condition)
+ ir->condition->accept(this);
+
+ char mask[5];
+ unsigned j = 0;
+
+ for (unsigned i = 0; i < 4; i++) {
+ if ((ir->write_mask & (1 << i)) != 0) {
+ mask[j] = "xyzw"[i];
+ j++;
+ }
+ }
+ mask[j] = '\0';
+
+ printf(" (%s) ", mask);
+
+ ir->lhs->accept(this);
+
+ printf(" ");
+
+ ir->rhs->accept(this);
+ printf(") ");
+}
+
+
+void ir_print_visitor::visit(ir_constant *ir)
+{
+ const glsl_type *const base_type = ir->type->get_base_type();
+
+ printf("(constant ");
+ print_type(ir->type);
+ printf(" (");
+
+ if (ir->type->is_array()) {
+ for (unsigned i = 0; i < ir->type->length; i++)
+ ir->get_array_element(i)->accept(this);
+ } else if (ir->type->is_record()) {
+ ir_constant *value = (ir_constant *) ir->components.get_head();
+ for (unsigned i = 0; i < ir->type->length; i++) {
+ printf("(%s ", ir->type->fields.structure->name);
+ value->accept(this);
+ printf(")");
+
+ value = (ir_constant *) value->next;
+ }
+ } else {
+ for (unsigned i = 0; i < ir->type->components(); i++) {
+ if (i != 0)
+ printf(" ");
+ switch (base_type->base_type) {
+ case GLSL_TYPE_UINT: printf("%u", ir->value.u[i]); break;
+ case GLSL_TYPE_INT: printf("%d", ir->value.i[i]); break;
+ case GLSL_TYPE_FLOAT: printf("%f", ir->value.f[i]); break;
+ case GLSL_TYPE_BOOL: printf("%d", ir->value.b[i]); break;
+ default: assert(0);
+ }
+ }
+ }
+ printf(")) ");
+}
+
+
+void
+ir_print_visitor::visit(ir_call *ir)
+{
+ printf("(call %s (", ir->callee_name());
+ foreach_iter(exec_list_iterator, iter, *ir) {
+ ir_instruction *const inst = (ir_instruction *) iter.get();
+
+ inst->accept(this);
+ }
+ printf("))\n");
+}
+
+
+void
+ir_print_visitor::visit(ir_return *ir)
+{
+ printf("(return");
+
+ ir_rvalue *const value = ir->get_value();
+ if (value) {
+ printf(" ");
+ value->accept(this);
+ }
+
+ printf(")");
+}
+
+
+void
+ir_print_visitor::visit(ir_discard *ir)
+{
+ printf("(discard ");
+
+ if (ir->condition != NULL) {
+ printf(" ");
+ ir->condition->accept(this);
+ }
+
+ printf(")");
+}
+
+
+void
+ir_print_visitor::visit(ir_if *ir)
+{
+ printf("(if ");
+ ir->condition->accept(this);
+
+ printf("(\n");
+ indentation++;
+
+ foreach_iter(exec_list_iterator, iter, ir->then_instructions) {
+ ir_instruction *const inst = (ir_instruction *) iter.get();
+
+ indent();
+ inst->accept(this);
+ printf("\n");
+ }
+
+ indentation--;
+ indent();
+ printf(")\n");
+
+ indent();
+ if (!ir->else_instructions.is_empty()) {
+ printf("(\n");
+ indentation++;
+
+ foreach_iter(exec_list_iterator, iter, ir->else_instructions) {
+ ir_instruction *const inst = (ir_instruction *) iter.get();
+
+ indent();
+ inst->accept(this);
+ printf("\n");
+ }
+ indentation--;
+ indent();
+ printf("))\n");
+ } else {
+ printf("())\n");
+ }
+}
+
+
+void
+ir_print_visitor::visit(ir_loop *ir)
+{
+ printf("(loop (");
+ if (ir->counter != NULL)
+ ir->counter->accept(this);
+ printf(") (");
+ if (ir->from != NULL)
+ ir->from->accept(this);
+ printf(") (");
+ if (ir->to != NULL)
+ ir->to->accept(this);
+ printf(") (");
+ if (ir->increment != NULL)
+ ir->increment->accept(this);
+ printf(") (\n");
+ indentation++;
+
+ foreach_iter(exec_list_iterator, iter, ir->body_instructions) {
+ ir_instruction *const inst = (ir_instruction *) iter.get();
+
+ indent();
+ inst->accept(this);
+ printf("\n");
+ }
+ indentation--;
+ indent();
+ printf("))\n");
+}
+
+
+void
+ir_print_visitor::visit(ir_loop_jump *ir)
+{
+ printf("%s", ir->is_break() ? "break" : "continue");
+}
diff --git a/mesalib/src/glsl/ir_validate.cpp b/mesalib/src/glsl/ir_validate.cpp index b0dd6c21f..4d9d4562c 100644 --- a/mesalib/src/glsl/ir_validate.cpp +++ b/mesalib/src/glsl/ir_validate.cpp @@ -1,532 +1,529 @@ -/* - * Copyright © 2010 Intel Corporation - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING - * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER - * DEALINGS IN THE SOFTWARE. - */ - -/** - * \file ir_validate.cpp - * - * Attempts to verify that various invariants of the IR tree are true. - * - * In particular, at the moment it makes sure that no single - * ir_instruction node except for ir_variable appears multiple times - * in the ir tree. ir_variable does appear multiple times: Once as a - * declaration in an exec_list, and multiple times as the endpoint of - * a dereference chain. - */ - -#include <inttypes.h> -#include "ir.h" -#include "ir_hierarchical_visitor.h" -#include "program/hash_table.h" -#include "glsl_types.h" - -using std::abort; -using std::printf; - -class ir_validate : public ir_hierarchical_visitor { -public: - ir_validate() - { - this->ht = hash_table_ctor(0, hash_table_pointer_hash, - hash_table_pointer_compare); - - this->current_function = NULL; - - this->callback = ir_validate::validate_ir; - this->data = ht; - } - - ~ir_validate() - { - hash_table_dtor(this->ht); - } - - virtual ir_visitor_status visit(ir_variable *v); - virtual ir_visitor_status visit(ir_dereference_variable *ir); - virtual ir_visitor_status visit(ir_if *ir); - - virtual ir_visitor_status visit_leave(ir_loop *ir); - virtual ir_visitor_status visit_enter(ir_function *ir); - virtual ir_visitor_status visit_leave(ir_function *ir); - virtual ir_visitor_status visit_enter(ir_function_signature *ir); - - virtual ir_visitor_status visit_leave(ir_expression *ir); - virtual ir_visitor_status visit_leave(ir_swizzle *ir); - - virtual ir_visitor_status visit_enter(ir_assignment *ir); - - static void validate_ir(ir_instruction *ir, void *data); - - ir_function *current_function; - - struct hash_table *ht; -}; - - -ir_visitor_status -ir_validate::visit(ir_dereference_variable *ir) -{ - if ((ir->var == NULL) || (ir->var->as_variable() == NULL)) { - printf("ir_dereference_variable @ %p does not specify a variable %p\n", - (void *) ir, (void *) ir->var); - abort(); - } - - if (hash_table_find(ht, ir->var) == NULL) { - printf("ir_dereference_variable @ %p specifies undeclared variable " - "`%s' @ %p\n", - (void *) ir, ir->var->name, (void *) ir->var); - abort(); - } - - this->validate_ir(ir, this->data); - - return visit_continue; -} - -ir_visitor_status -ir_validate::visit(ir_if *ir) -{ - if (ir->condition->type != glsl_type::bool_type) { - printf("ir_if condition %s type instead of bool.\n", - ir->condition->type->name); - ir->print(); - printf("\n"); - abort(); - } - - return visit_continue; -} - - -ir_visitor_status -ir_validate::visit_leave(ir_loop *ir) -{ - if (ir->counter != NULL) { - if ((ir->from == NULL) || (ir->from == NULL) || (ir->increment == NULL)) { - printf("ir_loop has invalid loop controls:\n" - " counter: %p\n" - " from: %p\n" - " to: %p\n" - " increment: %p\n", - (void *) ir->counter, (void *) ir->from, (void *) ir->to, - (void *) ir->increment); - abort(); - } - - if ((ir->cmp < ir_binop_less) || (ir->cmp > ir_binop_nequal)) { - printf("ir_loop has invalid comparitor %d\n", ir->cmp); - abort(); - } - } else { - if ((ir->from != NULL) || (ir->from != NULL) || (ir->increment != NULL)) { - printf("ir_loop has invalid loop controls:\n" - " counter: %p\n" - " from: %p\n" - " to: %p\n" - " increment: %p\n", - (void *) ir->counter, (void *) ir->from, (void *) ir->to, - (void *) ir->increment); - abort(); - } - } - - return visit_continue; -} - - -ir_visitor_status -ir_validate::visit_enter(ir_function *ir) -{ - /* Function definitions cannot be nested. - */ - if (this->current_function != NULL) { - printf("Function definition nested inside another function " - "definition:\n"); - printf("%s %p inside %s %p\n", - ir->name, (void *) ir, - this->current_function->name, (void *) this->current_function); - abort(); - } - - /* Store the current function hierarchy being traversed. This is used - * by the function signature visitor to ensure that the signatures are - * linked with the correct functions. - */ - this->current_function = ir; - - this->validate_ir(ir, this->data); - - return visit_continue; -} - -ir_visitor_status -ir_validate::visit_leave(ir_function *ir) -{ - assert(ralloc_parent(ir->name) == ir); - - this->current_function = NULL; - return visit_continue; -} - -ir_visitor_status -ir_validate::visit_enter(ir_function_signature *ir) -{ - if (this->current_function != ir->function()) { - printf("Function signature nested inside wrong function " - "definition:\n"); - printf("%p inside %s %p instead of %s %p\n", - (void *) ir, - this->current_function->name, (void *) this->current_function, - ir->function_name(), (void *) ir->function()); - abort(); - } - - this->validate_ir(ir, this->data); - - return visit_continue; -} - -ir_visitor_status -ir_validate::visit_leave(ir_expression *ir) -{ - switch (ir->operation) { - case ir_unop_bit_not: - assert(ir->operands[0]->type == ir->type); - break; - case ir_unop_logic_not: - assert(ir->type->base_type == GLSL_TYPE_BOOL); - assert(ir->operands[0]->type->base_type == GLSL_TYPE_BOOL); - break; - - case ir_unop_neg: - case ir_unop_abs: - case ir_unop_sign: - case ir_unop_rcp: - case ir_unop_rsq: - case ir_unop_sqrt: - assert(ir->type == ir->operands[0]->type); - break; - - case ir_unop_exp: - case ir_unop_log: - case ir_unop_exp2: - case ir_unop_log2: - assert(ir->operands[0]->type->base_type == GLSL_TYPE_FLOAT); - assert(ir->type == ir->operands[0]->type); - break; - - case ir_unop_f2i: - assert(ir->operands[0]->type->base_type == GLSL_TYPE_FLOAT); - assert(ir->type->base_type == GLSL_TYPE_INT); - break; - case ir_unop_i2f: - assert(ir->operands[0]->type->base_type == GLSL_TYPE_INT); - assert(ir->type->base_type == GLSL_TYPE_FLOAT); - break; - case ir_unop_f2b: - assert(ir->operands[0]->type->base_type == GLSL_TYPE_FLOAT); - assert(ir->type->base_type == GLSL_TYPE_BOOL); - break; - case ir_unop_b2f: - assert(ir->operands[0]->type->base_type == GLSL_TYPE_BOOL); - assert(ir->type->base_type == GLSL_TYPE_FLOAT); - break; - case ir_unop_i2b: - assert(ir->operands[0]->type->base_type == GLSL_TYPE_INT); - assert(ir->type->base_type == GLSL_TYPE_BOOL); - break; - case ir_unop_b2i: - assert(ir->operands[0]->type->base_type == GLSL_TYPE_BOOL); - assert(ir->type->base_type == GLSL_TYPE_INT); - break; - case ir_unop_u2f: - assert(ir->operands[0]->type->base_type == GLSL_TYPE_UINT); - assert(ir->type->base_type == GLSL_TYPE_FLOAT); - break; - - case ir_unop_any: - assert(ir->operands[0]->type->base_type == GLSL_TYPE_BOOL); - assert(ir->type == glsl_type::bool_type); - break; - - case ir_unop_trunc: - case ir_unop_round_even: - case ir_unop_ceil: - case ir_unop_floor: - case ir_unop_fract: - case ir_unop_sin: - case ir_unop_cos: - case ir_unop_sin_reduced: - case ir_unop_cos_reduced: - case ir_unop_dFdx: - case ir_unop_dFdy: - assert(ir->operands[0]->type->base_type == GLSL_TYPE_FLOAT); - assert(ir->operands[0]->type == ir->type); - break; - - case ir_unop_noise: - /* XXX what can we assert here? */ - break; - - case ir_binop_add: - case ir_binop_sub: - case ir_binop_mul: - case ir_binop_div: - case ir_binop_mod: - case ir_binop_min: - case ir_binop_max: - case ir_binop_pow: - if (ir->operands[0]->type->is_scalar()) - assert(ir->operands[1]->type == ir->type); - else if (ir->operands[1]->type->is_scalar()) - assert(ir->operands[0]->type == ir->type); - else if (ir->operands[0]->type->is_vector() && - ir->operands[1]->type->is_vector()) { - assert(ir->operands[0]->type == ir->operands[1]->type); - assert(ir->operands[0]->type == ir->type); - } - break; - - case ir_binop_less: - case ir_binop_greater: - case ir_binop_lequal: - case ir_binop_gequal: - case ir_binop_equal: - case ir_binop_nequal: - /* The semantics of the IR operators differ from the GLSL <, >, <=, >=, - * ==, and != operators. The IR operators perform a component-wise - * comparison on scalar or vector types and return a boolean scalar or - * vector type of the same size. - */ - assert(ir->type->base_type == GLSL_TYPE_BOOL); - assert(ir->operands[0]->type == ir->operands[1]->type); - assert(ir->operands[0]->type->is_vector() - || ir->operands[0]->type->is_scalar()); - assert(ir->operands[0]->type->vector_elements - == ir->type->vector_elements); - break; - - case ir_binop_all_equal: - case ir_binop_any_nequal: - /* GLSL == and != operate on scalars, vectors, matrices and arrays, and - * return a scalar boolean. The IR matches that. - */ - assert(ir->type == glsl_type::bool_type); - assert(ir->operands[0]->type == ir->operands[1]->type); - break; - - case ir_binop_lshift: - case ir_binop_rshift: - assert(ir->operands[0]->type->is_integer() && - ir->operands[1]->type->is_integer()); - if (ir->operands[0]->type->is_scalar()) { - assert(ir->operands[1]->type->is_scalar()); - } - if (ir->operands[0]->type->is_vector() && - ir->operands[1]->type->is_vector()) { - assert(ir->operands[0]->type->components() == - ir->operands[1]->type->components()); - } - assert(ir->type == ir->operands[0]->type); - break; - - case ir_binop_bit_and: - case ir_binop_bit_xor: - case ir_binop_bit_or: - assert(ir->operands[0]->type->base_type == - ir->operands[1]->type->base_type); - assert(ir->type->is_integer()); - if (ir->operands[0]->type->is_vector() && - ir->operands[1]->type->is_vector()) { - assert(ir->operands[0]->type->vector_elements == - ir->operands[1]->type->vector_elements); - } - break; - - case ir_binop_logic_and: - case ir_binop_logic_xor: - case ir_binop_logic_or: - assert(ir->type == glsl_type::bool_type); - assert(ir->operands[0]->type == glsl_type::bool_type); - assert(ir->operands[1]->type == glsl_type::bool_type); - break; - - case ir_binop_dot: - assert(ir->type == glsl_type::float_type); - assert(ir->operands[0]->type->base_type == GLSL_TYPE_FLOAT); - assert(ir->operands[0]->type->is_vector()); - assert(ir->operands[0]->type == ir->operands[1]->type); - break; - - case ir_quadop_vector: - /* The vector operator collects some number of scalars and generates a - * vector from them. - * - * - All of the operands must be scalar. - * - Number of operands must matche the size of the resulting vector. - * - Base type of the operands must match the base type of the result. - */ - assert(ir->type->is_vector()); - switch (ir->type->vector_elements) { - case 2: - assert(ir->operands[0]->type->is_scalar()); - assert(ir->operands[0]->type->base_type == ir->type->base_type); - assert(ir->operands[1]->type->is_scalar()); - assert(ir->operands[1]->type->base_type == ir->type->base_type); - assert(ir->operands[2] == NULL); - assert(ir->operands[3] == NULL); - break; - case 3: - assert(ir->operands[0]->type->is_scalar()); - assert(ir->operands[0]->type->base_type == ir->type->base_type); - assert(ir->operands[1]->type->is_scalar()); - assert(ir->operands[1]->type->base_type == ir->type->base_type); - assert(ir->operands[2]->type->is_scalar()); - assert(ir->operands[2]->type->base_type == ir->type->base_type); - assert(ir->operands[3] == NULL); - break; - case 4: - assert(ir->operands[0]->type->is_scalar()); - assert(ir->operands[0]->type->base_type == ir->type->base_type); - assert(ir->operands[1]->type->is_scalar()); - assert(ir->operands[1]->type->base_type == ir->type->base_type); - assert(ir->operands[2]->type->is_scalar()); - assert(ir->operands[2]->type->base_type == ir->type->base_type); - assert(ir->operands[3]->type->is_scalar()); - assert(ir->operands[3]->type->base_type == ir->type->base_type); - break; - default: - /* The is_vector assertion above should prevent execution from ever - * getting here. - */ - assert(!"Should not get here."); - break; - } - } - - return visit_continue; -} - -ir_visitor_status -ir_validate::visit_leave(ir_swizzle *ir) -{ - int chans[4] = {ir->mask.x, ir->mask.y, ir->mask.z, ir->mask.w}; - - for (unsigned int i = 0; i < ir->type->vector_elements; i++) { - if (chans[i] >= ir->val->type->vector_elements) { - printf("ir_swizzle @ %p specifies a channel not present " - "in the value.\n", (void *) ir); - ir->print(); - abort(); - } - } - - return visit_continue; -} - -ir_visitor_status -ir_validate::visit(ir_variable *ir) -{ - /* An ir_variable is the one thing that can (and will) appear multiple times - * in an IR tree. It is added to the hashtable so that it can be used - * in the ir_dereference_variable handler to ensure that a variable is - * declared before it is dereferenced. - */ - if (ir->name) - assert(ralloc_parent(ir->name) == ir); - - hash_table_insert(ht, ir, ir); - return visit_continue; -} - -ir_visitor_status -ir_validate::visit_enter(ir_assignment *ir) -{ - const ir_dereference *const lhs = ir->lhs; - if (lhs->type->is_scalar() || lhs->type->is_vector()) { - if (ir->write_mask == 0) { - printf("Assignment LHS is %s, but write mask is 0:\n", - lhs->type->is_scalar() ? "scalar" : "vector"); - ir->print(); - abort(); - } - - int lhs_components = 0; - for (int i = 0; i < 4; i++) { - if (ir->write_mask & (1 << i)) - lhs_components++; - } - - if (lhs_components != ir->rhs->type->vector_elements) { - printf("Assignment count of LHS write mask channels enabled not\n" - "matching RHS vector size (%d LHS, %d RHS).\n", - lhs_components, ir->rhs->type->vector_elements); - ir->print(); - abort(); - } - } - - this->validate_ir(ir, this->data); - - return visit_continue; -} - -void -ir_validate::validate_ir(ir_instruction *ir, void *data) -{ - struct hash_table *ht = (struct hash_table *) data; - - if (hash_table_find(ht, ir)) { - printf("Instruction node present twice in ir tree:\n"); - ir->print(); - printf("\n"); - abort(); - } - hash_table_insert(ht, ir, ir); -} - -void -check_node_type(ir_instruction *ir, void *data) -{ - (void) data; - - if (ir->ir_type <= ir_type_unset || ir->ir_type >= ir_type_max) { - printf("Instruction node with unset type\n"); - ir->print(); printf("\n"); - } - assert(ir->type != glsl_type::error_type); -} - -void -validate_ir_tree(exec_list *instructions) -{ - ir_validate v; - - v.run(instructions); - - foreach_iter(exec_list_iterator, iter, *instructions) { - ir_instruction *ir = (ir_instruction *)iter.get(); - - visit_tree(ir, check_node_type, NULL); - } -} +/*
+ * Copyright © 2010 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file ir_validate.cpp
+ *
+ * Attempts to verify that various invariants of the IR tree are true.
+ *
+ * In particular, at the moment it makes sure that no single
+ * ir_instruction node except for ir_variable appears multiple times
+ * in the ir tree. ir_variable does appear multiple times: Once as a
+ * declaration in an exec_list, and multiple times as the endpoint of
+ * a dereference chain.
+ */
+
+#include <inttypes.h>
+#include "ir.h"
+#include "ir_hierarchical_visitor.h"
+#include "program/hash_table.h"
+#include "glsl_types.h"
+
+class ir_validate : public ir_hierarchical_visitor {
+public:
+ ir_validate()
+ {
+ this->ht = hash_table_ctor(0, hash_table_pointer_hash,
+ hash_table_pointer_compare);
+
+ this->current_function = NULL;
+
+ this->callback = ir_validate::validate_ir;
+ this->data = ht;
+ }
+
+ ~ir_validate()
+ {
+ hash_table_dtor(this->ht);
+ }
+
+ virtual ir_visitor_status visit(ir_variable *v);
+ virtual ir_visitor_status visit(ir_dereference_variable *ir);
+ virtual ir_visitor_status visit(ir_if *ir);
+
+ virtual ir_visitor_status visit_leave(ir_loop *ir);
+ virtual ir_visitor_status visit_enter(ir_function *ir);
+ virtual ir_visitor_status visit_leave(ir_function *ir);
+ virtual ir_visitor_status visit_enter(ir_function_signature *ir);
+
+ virtual ir_visitor_status visit_leave(ir_expression *ir);
+ virtual ir_visitor_status visit_leave(ir_swizzle *ir);
+
+ virtual ir_visitor_status visit_enter(ir_assignment *ir);
+
+ static void validate_ir(ir_instruction *ir, void *data);
+
+ ir_function *current_function;
+
+ struct hash_table *ht;
+};
+
+
+ir_visitor_status
+ir_validate::visit(ir_dereference_variable *ir)
+{
+ if ((ir->var == NULL) || (ir->var->as_variable() == NULL)) {
+ printf("ir_dereference_variable @ %p does not specify a variable %p\n",
+ (void *) ir, (void *) ir->var);
+ abort();
+ }
+
+ if (hash_table_find(ht, ir->var) == NULL) {
+ printf("ir_dereference_variable @ %p specifies undeclared variable "
+ "`%s' @ %p\n",
+ (void *) ir, ir->var->name, (void *) ir->var);
+ abort();
+ }
+
+ this->validate_ir(ir, this->data);
+
+ return visit_continue;
+}
+
+ir_visitor_status
+ir_validate::visit(ir_if *ir)
+{
+ if (ir->condition->type != glsl_type::bool_type) {
+ printf("ir_if condition %s type instead of bool.\n",
+ ir->condition->type->name);
+ ir->print();
+ printf("\n");
+ abort();
+ }
+
+ return visit_continue;
+}
+
+
+ir_visitor_status
+ir_validate::visit_leave(ir_loop *ir)
+{
+ if (ir->counter != NULL) {
+ if ((ir->from == NULL) || (ir->from == NULL) || (ir->increment == NULL)) {
+ printf("ir_loop has invalid loop controls:\n"
+ " counter: %p\n"
+ " from: %p\n"
+ " to: %p\n"
+ " increment: %p\n",
+ (void *) ir->counter, (void *) ir->from, (void *) ir->to,
+ (void *) ir->increment);
+ abort();
+ }
+
+ if ((ir->cmp < ir_binop_less) || (ir->cmp > ir_binop_nequal)) {
+ printf("ir_loop has invalid comparitor %d\n", ir->cmp);
+ abort();
+ }
+ } else {
+ if ((ir->from != NULL) || (ir->from != NULL) || (ir->increment != NULL)) {
+ printf("ir_loop has invalid loop controls:\n"
+ " counter: %p\n"
+ " from: %p\n"
+ " to: %p\n"
+ " increment: %p\n",
+ (void *) ir->counter, (void *) ir->from, (void *) ir->to,
+ (void *) ir->increment);
+ abort();
+ }
+ }
+
+ return visit_continue;
+}
+
+
+ir_visitor_status
+ir_validate::visit_enter(ir_function *ir)
+{
+ /* Function definitions cannot be nested.
+ */
+ if (this->current_function != NULL) {
+ printf("Function definition nested inside another function "
+ "definition:\n");
+ printf("%s %p inside %s %p\n",
+ ir->name, (void *) ir,
+ this->current_function->name, (void *) this->current_function);
+ abort();
+ }
+
+ /* Store the current function hierarchy being traversed. This is used
+ * by the function signature visitor to ensure that the signatures are
+ * linked with the correct functions.
+ */
+ this->current_function = ir;
+
+ this->validate_ir(ir, this->data);
+
+ return visit_continue;
+}
+
+ir_visitor_status
+ir_validate::visit_leave(ir_function *ir)
+{
+ assert(ralloc_parent(ir->name) == ir);
+
+ this->current_function = NULL;
+ return visit_continue;
+}
+
+ir_visitor_status
+ir_validate::visit_enter(ir_function_signature *ir)
+{
+ if (this->current_function != ir->function()) {
+ printf("Function signature nested inside wrong function "
+ "definition:\n");
+ printf("%p inside %s %p instead of %s %p\n",
+ (void *) ir,
+ this->current_function->name, (void *) this->current_function,
+ ir->function_name(), (void *) ir->function());
+ abort();
+ }
+
+ this->validate_ir(ir, this->data);
+
+ return visit_continue;
+}
+
+ir_visitor_status
+ir_validate::visit_leave(ir_expression *ir)
+{
+ switch (ir->operation) {
+ case ir_unop_bit_not:
+ assert(ir->operands[0]->type == ir->type);
+ break;
+ case ir_unop_logic_not:
+ assert(ir->type->base_type == GLSL_TYPE_BOOL);
+ assert(ir->operands[0]->type->base_type == GLSL_TYPE_BOOL);
+ break;
+
+ case ir_unop_neg:
+ case ir_unop_abs:
+ case ir_unop_sign:
+ case ir_unop_rcp:
+ case ir_unop_rsq:
+ case ir_unop_sqrt:
+ assert(ir->type == ir->operands[0]->type);
+ break;
+
+ case ir_unop_exp:
+ case ir_unop_log:
+ case ir_unop_exp2:
+ case ir_unop_log2:
+ assert(ir->operands[0]->type->base_type == GLSL_TYPE_FLOAT);
+ assert(ir->type == ir->operands[0]->type);
+ break;
+
+ case ir_unop_f2i:
+ assert(ir->operands[0]->type->base_type == GLSL_TYPE_FLOAT);
+ assert(ir->type->base_type == GLSL_TYPE_INT);
+ break;
+ case ir_unop_i2f:
+ assert(ir->operands[0]->type->base_type == GLSL_TYPE_INT);
+ assert(ir->type->base_type == GLSL_TYPE_FLOAT);
+ break;
+ case ir_unop_f2b:
+ assert(ir->operands[0]->type->base_type == GLSL_TYPE_FLOAT);
+ assert(ir->type->base_type == GLSL_TYPE_BOOL);
+ break;
+ case ir_unop_b2f:
+ assert(ir->operands[0]->type->base_type == GLSL_TYPE_BOOL);
+ assert(ir->type->base_type == GLSL_TYPE_FLOAT);
+ break;
+ case ir_unop_i2b:
+ assert(ir->operands[0]->type->base_type == GLSL_TYPE_INT);
+ assert(ir->type->base_type == GLSL_TYPE_BOOL);
+ break;
+ case ir_unop_b2i:
+ assert(ir->operands[0]->type->base_type == GLSL_TYPE_BOOL);
+ assert(ir->type->base_type == GLSL_TYPE_INT);
+ break;
+ case ir_unop_u2f:
+ assert(ir->operands[0]->type->base_type == GLSL_TYPE_UINT);
+ assert(ir->type->base_type == GLSL_TYPE_FLOAT);
+ break;
+
+ case ir_unop_any:
+ assert(ir->operands[0]->type->base_type == GLSL_TYPE_BOOL);
+ assert(ir->type == glsl_type::bool_type);
+ break;
+
+ case ir_unop_trunc:
+ case ir_unop_round_even:
+ case ir_unop_ceil:
+ case ir_unop_floor:
+ case ir_unop_fract:
+ case ir_unop_sin:
+ case ir_unop_cos:
+ case ir_unop_sin_reduced:
+ case ir_unop_cos_reduced:
+ case ir_unop_dFdx:
+ case ir_unop_dFdy:
+ assert(ir->operands[0]->type->base_type == GLSL_TYPE_FLOAT);
+ assert(ir->operands[0]->type == ir->type);
+ break;
+
+ case ir_unop_noise:
+ /* XXX what can we assert here? */
+ break;
+
+ case ir_binop_add:
+ case ir_binop_sub:
+ case ir_binop_mul:
+ case ir_binop_div:
+ case ir_binop_mod:
+ case ir_binop_min:
+ case ir_binop_max:
+ case ir_binop_pow:
+ if (ir->operands[0]->type->is_scalar())
+ assert(ir->operands[1]->type == ir->type);
+ else if (ir->operands[1]->type->is_scalar())
+ assert(ir->operands[0]->type == ir->type);
+ else if (ir->operands[0]->type->is_vector() &&
+ ir->operands[1]->type->is_vector()) {
+ assert(ir->operands[0]->type == ir->operands[1]->type);
+ assert(ir->operands[0]->type == ir->type);
+ }
+ break;
+
+ case ir_binop_less:
+ case ir_binop_greater:
+ case ir_binop_lequal:
+ case ir_binop_gequal:
+ case ir_binop_equal:
+ case ir_binop_nequal:
+ /* The semantics of the IR operators differ from the GLSL <, >, <=, >=,
+ * ==, and != operators. The IR operators perform a component-wise
+ * comparison on scalar or vector types and return a boolean scalar or
+ * vector type of the same size.
+ */
+ assert(ir->type->base_type == GLSL_TYPE_BOOL);
+ assert(ir->operands[0]->type == ir->operands[1]->type);
+ assert(ir->operands[0]->type->is_vector()
+ || ir->operands[0]->type->is_scalar());
+ assert(ir->operands[0]->type->vector_elements
+ == ir->type->vector_elements);
+ break;
+
+ case ir_binop_all_equal:
+ case ir_binop_any_nequal:
+ /* GLSL == and != operate on scalars, vectors, matrices and arrays, and
+ * return a scalar boolean. The IR matches that.
+ */
+ assert(ir->type == glsl_type::bool_type);
+ assert(ir->operands[0]->type == ir->operands[1]->type);
+ break;
+
+ case ir_binop_lshift:
+ case ir_binop_rshift:
+ assert(ir->operands[0]->type->is_integer() &&
+ ir->operands[1]->type->is_integer());
+ if (ir->operands[0]->type->is_scalar()) {
+ assert(ir->operands[1]->type->is_scalar());
+ }
+ if (ir->operands[0]->type->is_vector() &&
+ ir->operands[1]->type->is_vector()) {
+ assert(ir->operands[0]->type->components() ==
+ ir->operands[1]->type->components());
+ }
+ assert(ir->type == ir->operands[0]->type);
+ break;
+
+ case ir_binop_bit_and:
+ case ir_binop_bit_xor:
+ case ir_binop_bit_or:
+ assert(ir->operands[0]->type->base_type ==
+ ir->operands[1]->type->base_type);
+ assert(ir->type->is_integer());
+ if (ir->operands[0]->type->is_vector() &&
+ ir->operands[1]->type->is_vector()) {
+ assert(ir->operands[0]->type->vector_elements ==
+ ir->operands[1]->type->vector_elements);
+ }
+ break;
+
+ case ir_binop_logic_and:
+ case ir_binop_logic_xor:
+ case ir_binop_logic_or:
+ assert(ir->type == glsl_type::bool_type);
+ assert(ir->operands[0]->type == glsl_type::bool_type);
+ assert(ir->operands[1]->type == glsl_type::bool_type);
+ break;
+
+ case ir_binop_dot:
+ assert(ir->type == glsl_type::float_type);
+ assert(ir->operands[0]->type->base_type == GLSL_TYPE_FLOAT);
+ assert(ir->operands[0]->type->is_vector());
+ assert(ir->operands[0]->type == ir->operands[1]->type);
+ break;
+
+ case ir_quadop_vector:
+ /* The vector operator collects some number of scalars and generates a
+ * vector from them.
+ *
+ * - All of the operands must be scalar.
+ * - Number of operands must matche the size of the resulting vector.
+ * - Base type of the operands must match the base type of the result.
+ */
+ assert(ir->type->is_vector());
+ switch (ir->type->vector_elements) {
+ case 2:
+ assert(ir->operands[0]->type->is_scalar());
+ assert(ir->operands[0]->type->base_type == ir->type->base_type);
+ assert(ir->operands[1]->type->is_scalar());
+ assert(ir->operands[1]->type->base_type == ir->type->base_type);
+ assert(ir->operands[2] == NULL);
+ assert(ir->operands[3] == NULL);
+ break;
+ case 3:
+ assert(ir->operands[0]->type->is_scalar());
+ assert(ir->operands[0]->type->base_type == ir->type->base_type);
+ assert(ir->operands[1]->type->is_scalar());
+ assert(ir->operands[1]->type->base_type == ir->type->base_type);
+ assert(ir->operands[2]->type->is_scalar());
+ assert(ir->operands[2]->type->base_type == ir->type->base_type);
+ assert(ir->operands[3] == NULL);
+ break;
+ case 4:
+ assert(ir->operands[0]->type->is_scalar());
+ assert(ir->operands[0]->type->base_type == ir->type->base_type);
+ assert(ir->operands[1]->type->is_scalar());
+ assert(ir->operands[1]->type->base_type == ir->type->base_type);
+ assert(ir->operands[2]->type->is_scalar());
+ assert(ir->operands[2]->type->base_type == ir->type->base_type);
+ assert(ir->operands[3]->type->is_scalar());
+ assert(ir->operands[3]->type->base_type == ir->type->base_type);
+ break;
+ default:
+ /* The is_vector assertion above should prevent execution from ever
+ * getting here.
+ */
+ assert(!"Should not get here.");
+ break;
+ }
+ }
+
+ return visit_continue;
+}
+
+ir_visitor_status
+ir_validate::visit_leave(ir_swizzle *ir)
+{
+ int chans[4] = {ir->mask.x, ir->mask.y, ir->mask.z, ir->mask.w};
+
+ for (unsigned int i = 0; i < ir->type->vector_elements; i++) {
+ if (chans[i] >= ir->val->type->vector_elements) {
+ printf("ir_swizzle @ %p specifies a channel not present "
+ "in the value.\n", (void *) ir);
+ ir->print();
+ abort();
+ }
+ }
+
+ return visit_continue;
+}
+
+ir_visitor_status
+ir_validate::visit(ir_variable *ir)
+{
+ /* An ir_variable is the one thing that can (and will) appear multiple times
+ * in an IR tree. It is added to the hashtable so that it can be used
+ * in the ir_dereference_variable handler to ensure that a variable is
+ * declared before it is dereferenced.
+ */
+ if (ir->name)
+ assert(ralloc_parent(ir->name) == ir);
+
+ hash_table_insert(ht, ir, ir);
+ return visit_continue;
+}
+
+ir_visitor_status
+ir_validate::visit_enter(ir_assignment *ir)
+{
+ const ir_dereference *const lhs = ir->lhs;
+ if (lhs->type->is_scalar() || lhs->type->is_vector()) {
+ if (ir->write_mask == 0) {
+ printf("Assignment LHS is %s, but write mask is 0:\n",
+ lhs->type->is_scalar() ? "scalar" : "vector");
+ ir->print();
+ abort();
+ }
+
+ int lhs_components = 0;
+ for (int i = 0; i < 4; i++) {
+ if (ir->write_mask & (1 << i))
+ lhs_components++;
+ }
+
+ if (lhs_components != ir->rhs->type->vector_elements) {
+ printf("Assignment count of LHS write mask channels enabled not\n"
+ "matching RHS vector size (%d LHS, %d RHS).\n",
+ lhs_components, ir->rhs->type->vector_elements);
+ ir->print();
+ abort();
+ }
+ }
+
+ this->validate_ir(ir, this->data);
+
+ return visit_continue;
+}
+
+void
+ir_validate::validate_ir(ir_instruction *ir, void *data)
+{
+ struct hash_table *ht = (struct hash_table *) data;
+
+ if (hash_table_find(ht, ir)) {
+ printf("Instruction node present twice in ir tree:\n");
+ ir->print();
+ printf("\n");
+ abort();
+ }
+ hash_table_insert(ht, ir, ir);
+}
+
+void
+check_node_type(ir_instruction *ir, void *data)
+{
+ (void) data;
+
+ if (ir->ir_type <= ir_type_unset || ir->ir_type >= ir_type_max) {
+ printf("Instruction node with unset type\n");
+ ir->print(); printf("\n");
+ }
+ assert(ir->type != glsl_type::error_type);
+}
+
+void
+validate_ir_tree(exec_list *instructions)
+{
+ ir_validate v;
+
+ v.run(instructions);
+
+ foreach_iter(exec_list_iterator, iter, *instructions) {
+ ir_instruction *ir = (ir_instruction *)iter.get();
+
+ visit_tree(ir, check_node_type, NULL);
+ }
+}
diff --git a/mesalib/src/glsl/link_functions.cpp b/mesalib/src/glsl/link_functions.cpp index 861fa39b5..bc8758438 100644 --- a/mesalib/src/glsl/link_functions.cpp +++ b/mesalib/src/glsl/link_functions.cpp @@ -1,270 +1,266 @@ -/* - * Copyright © 2010 Intel Corporation - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING - * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER - * DEALINGS IN THE SOFTWARE. - */ - -#include <cstdlib> -#include <cstdio> -#include <cstdarg> - -#include "main/core.h" -#include "glsl_symbol_table.h" -#include "glsl_parser_extras.h" -#include "ir.h" -#include "program.h" -#include "program/hash_table.h" -#include "linker.h" - -static ir_function_signature * -find_matching_signature(const char *name, const exec_list *actual_parameters, - gl_shader **shader_list, unsigned num_shaders); - -class call_link_visitor : public ir_hierarchical_visitor { -public: - call_link_visitor(gl_shader_program *prog, gl_shader *linked, - gl_shader **shader_list, unsigned num_shaders) - { - this->prog = prog; - this->shader_list = shader_list; - this->num_shaders = num_shaders; - this->success = true; - this->linked = linked; - - this->locals = hash_table_ctor(0, hash_table_pointer_hash, - hash_table_pointer_compare); - } - - ~call_link_visitor() - { - hash_table_dtor(this->locals); - } - - virtual ir_visitor_status visit(ir_variable *ir) - { - hash_table_insert(locals, ir, ir); - return visit_continue; - } - - virtual ir_visitor_status visit_enter(ir_call *ir) - { - /* If ir is an ir_call from a function that was imported from another - * shader callee will point to an ir_function_signature in the original - * shader. In this case the function signature MUST NOT BE MODIFIED. - * Doing so will modify the original shader. This may prevent that - * shader from being linkable in other programs. - */ - const ir_function_signature *const callee = ir->get_callee(); - assert(callee != NULL); - const char *const name = callee->function_name(); - - /* Determine if the requested function signature already exists in the - * final linked shader. If it does, use it as the target of the call. - */ - ir_function_signature *sig = - find_matching_signature(name, &callee->parameters, &linked, 1); - if (sig != NULL) { - ir->set_callee(sig); - return visit_continue; - } - - /* Try to find the signature in one of the other shaders that is being - * linked. If it's not found there, return an error. - */ - sig = find_matching_signature(name, &ir->actual_parameters, shader_list, - num_shaders); - if (sig == NULL) { - /* FINISHME: Log the full signature of unresolved function. - */ - linker_error_printf(this->prog, "unresolved reference to function " - "`%s'\n", name); - this->success = false; - return visit_stop; - } - - /* Find the prototype information in the linked shader. Generate any - * details that may be missing. - */ - ir_function *f = linked->symbols->get_function(name); - if (f == NULL) - f = new(linked) ir_function(name); - - ir_function_signature *linked_sig = - f->exact_matching_signature(&callee->parameters); - if (linked_sig == NULL) { - linked_sig = new(linked) ir_function_signature(callee->return_type); - f->add_signature(linked_sig); - } - - /* At this point linked_sig and called may be the same. If ir is an - * ir_call from linked then linked_sig and callee will be - * ir_function_signatures that have no definitions (is_defined is false). - */ - assert(!linked_sig->is_defined); - assert(linked_sig->body.is_empty()); - - /* Create an in-place clone of the function definition. This multistep - * process introduces some complexity here, but it has some advantages. - * The parameter list and the and function body are cloned separately. - * The clone of the parameter list is used to prime the hashtable used - * to replace variable references in the cloned body. - * - * The big advantage is that the ir_function_signature does not change. - * This means that we don't have to process the rest of the IR tree to - * patch ir_call nodes. In addition, there is no way to remove or - * replace signature stored in a function. One could easily be added, - * but this avoids the need. - */ - struct hash_table *ht = hash_table_ctor(0, hash_table_pointer_hash, - hash_table_pointer_compare); - exec_list formal_parameters; - foreach_list_const(node, &sig->parameters) { - const ir_instruction *const original = (ir_instruction *) node; - assert(const_cast<ir_instruction *>(original)->as_variable()); - - ir_instruction *copy = original->clone(linked, ht); - formal_parameters.push_tail(copy); - } - - linked_sig->replace_parameters(&formal_parameters); - - foreach_list_const(node, &sig->body) { - const ir_instruction *const original = (ir_instruction *) node; - - ir_instruction *copy = original->clone(linked, ht); - linked_sig->body.push_tail(copy); - } - - linked_sig->is_defined = true; - hash_table_dtor(ht); - - /* Patch references inside the function to things outside the function - * (i.e., function calls and global variables). - */ - linked_sig->accept(this); - - ir->set_callee(linked_sig); - - return visit_continue; - } - - virtual ir_visitor_status visit(ir_dereference_variable *ir) - { - if (hash_table_find(locals, ir->var) == NULL) { - /* The non-function variable must be a global, so try to find the - * variable in the shader's symbol table. If the variable is not - * found, then it's a global that *MUST* be defined in the original - * shader. - */ - ir_variable *var = linked->symbols->get_variable(ir->var->name); - if (var == NULL) { - /* Clone the ir_variable that the dereference already has and add - * it to the linked shader. - */ - var = ir->var->clone(linked, NULL); - linked->symbols->add_variable(var); - linked->ir->push_head(var); - } else if (var->type->is_array()) { - /* It is possible to have a global array declared in multiple - * shaders without a size. The array is implicitly sized by the - * maximal access to it in *any* shader. Because of this, we - * need to track the maximal access to the array as linking pulls - * more functions in that access the array. - */ - var->max_array_access = - MAX2(var->max_array_access, ir->var->max_array_access); - - if (var->type->length == 0 && ir->var->type->length != 0) - var->type = ir->var->type; - } - - ir->var = var; - } - - return visit_continue; - } - - /** Was function linking successful? */ - bool success; - -private: - /** - * Shader program being linked - * - * This is only used for logging error messages. - */ - gl_shader_program *prog; - - /** List of shaders available for linking. */ - gl_shader **shader_list; - - /** Number of shaders available for linking. */ - unsigned num_shaders; - - /** - * Final linked shader - * - * This is used two ways. It is used to find global variables in the - * linked shader that are accessed by the function. It is also used to add - * global variables from the shader where the function originated. - */ - gl_shader *linked; - - /** - * Table of variables local to the function. - */ - hash_table *locals; -}; - - -/** - * Searches a list of shaders for a particular function definition - */ -ir_function_signature * -find_matching_signature(const char *name, const exec_list *actual_parameters, - gl_shader **shader_list, unsigned num_shaders) -{ - for (unsigned i = 0; i < num_shaders; i++) { - ir_function *const f = shader_list[i]->symbols->get_function(name); - - if (f == NULL) - continue; - - ir_function_signature *sig = f->matching_signature(actual_parameters); - - if ((sig == NULL) || !sig->is_defined) - continue; - - return sig; - } - - return NULL; -} - - -bool -link_function_calls(gl_shader_program *prog, gl_shader *main, - gl_shader **shader_list, unsigned num_shaders) -{ - call_link_visitor v(prog, main, shader_list, num_shaders); - - v.run(main->ir); - return v.success; -} +/*
+ * Copyright © 2010 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include "main/core.h"
+#include "glsl_symbol_table.h"
+#include "glsl_parser_extras.h"
+#include "ir.h"
+#include "program.h"
+#include "program/hash_table.h"
+#include "linker.h"
+
+static ir_function_signature *
+find_matching_signature(const char *name, const exec_list *actual_parameters,
+ gl_shader **shader_list, unsigned num_shaders);
+
+class call_link_visitor : public ir_hierarchical_visitor {
+public:
+ call_link_visitor(gl_shader_program *prog, gl_shader *linked,
+ gl_shader **shader_list, unsigned num_shaders)
+ {
+ this->prog = prog;
+ this->shader_list = shader_list;
+ this->num_shaders = num_shaders;
+ this->success = true;
+ this->linked = linked;
+
+ this->locals = hash_table_ctor(0, hash_table_pointer_hash,
+ hash_table_pointer_compare);
+ }
+
+ ~call_link_visitor()
+ {
+ hash_table_dtor(this->locals);
+ }
+
+ virtual ir_visitor_status visit(ir_variable *ir)
+ {
+ hash_table_insert(locals, ir, ir);
+ return visit_continue;
+ }
+
+ virtual ir_visitor_status visit_enter(ir_call *ir)
+ {
+ /* If ir is an ir_call from a function that was imported from another
+ * shader callee will point to an ir_function_signature in the original
+ * shader. In this case the function signature MUST NOT BE MODIFIED.
+ * Doing so will modify the original shader. This may prevent that
+ * shader from being linkable in other programs.
+ */
+ const ir_function_signature *const callee = ir->get_callee();
+ assert(callee != NULL);
+ const char *const name = callee->function_name();
+
+ /* Determine if the requested function signature already exists in the
+ * final linked shader. If it does, use it as the target of the call.
+ */
+ ir_function_signature *sig =
+ find_matching_signature(name, &callee->parameters, &linked, 1);
+ if (sig != NULL) {
+ ir->set_callee(sig);
+ return visit_continue;
+ }
+
+ /* Try to find the signature in one of the other shaders that is being
+ * linked. If it's not found there, return an error.
+ */
+ sig = find_matching_signature(name, &ir->actual_parameters, shader_list,
+ num_shaders);
+ if (sig == NULL) {
+ /* FINISHME: Log the full signature of unresolved function.
+ */
+ linker_error_printf(this->prog, "unresolved reference to function "
+ "`%s'\n", name);
+ this->success = false;
+ return visit_stop;
+ }
+
+ /* Find the prototype information in the linked shader. Generate any
+ * details that may be missing.
+ */
+ ir_function *f = linked->symbols->get_function(name);
+ if (f == NULL)
+ f = new(linked) ir_function(name);
+
+ ir_function_signature *linked_sig =
+ f->exact_matching_signature(&callee->parameters);
+ if (linked_sig == NULL) {
+ linked_sig = new(linked) ir_function_signature(callee->return_type);
+ f->add_signature(linked_sig);
+ }
+
+ /* At this point linked_sig and called may be the same. If ir is an
+ * ir_call from linked then linked_sig and callee will be
+ * ir_function_signatures that have no definitions (is_defined is false).
+ */
+ assert(!linked_sig->is_defined);
+ assert(linked_sig->body.is_empty());
+
+ /* Create an in-place clone of the function definition. This multistep
+ * process introduces some complexity here, but it has some advantages.
+ * The parameter list and the and function body are cloned separately.
+ * The clone of the parameter list is used to prime the hashtable used
+ * to replace variable references in the cloned body.
+ *
+ * The big advantage is that the ir_function_signature does not change.
+ * This means that we don't have to process the rest of the IR tree to
+ * patch ir_call nodes. In addition, there is no way to remove or
+ * replace signature stored in a function. One could easily be added,
+ * but this avoids the need.
+ */
+ struct hash_table *ht = hash_table_ctor(0, hash_table_pointer_hash,
+ hash_table_pointer_compare);
+ exec_list formal_parameters;
+ foreach_list_const(node, &sig->parameters) {
+ const ir_instruction *const original = (ir_instruction *) node;
+ assert(const_cast<ir_instruction *>(original)->as_variable());
+
+ ir_instruction *copy = original->clone(linked, ht);
+ formal_parameters.push_tail(copy);
+ }
+
+ linked_sig->replace_parameters(&formal_parameters);
+
+ foreach_list_const(node, &sig->body) {
+ const ir_instruction *const original = (ir_instruction *) node;
+
+ ir_instruction *copy = original->clone(linked, ht);
+ linked_sig->body.push_tail(copy);
+ }
+
+ linked_sig->is_defined = true;
+ hash_table_dtor(ht);
+
+ /* Patch references inside the function to things outside the function
+ * (i.e., function calls and global variables).
+ */
+ linked_sig->accept(this);
+
+ ir->set_callee(linked_sig);
+
+ return visit_continue;
+ }
+
+ virtual ir_visitor_status visit(ir_dereference_variable *ir)
+ {
+ if (hash_table_find(locals, ir->var) == NULL) {
+ /* The non-function variable must be a global, so try to find the
+ * variable in the shader's symbol table. If the variable is not
+ * found, then it's a global that *MUST* be defined in the original
+ * shader.
+ */
+ ir_variable *var = linked->symbols->get_variable(ir->var->name);
+ if (var == NULL) {
+ /* Clone the ir_variable that the dereference already has and add
+ * it to the linked shader.
+ */
+ var = ir->var->clone(linked, NULL);
+ linked->symbols->add_variable(var);
+ linked->ir->push_head(var);
+ } else if (var->type->is_array()) {
+ /* It is possible to have a global array declared in multiple
+ * shaders without a size. The array is implicitly sized by the
+ * maximal access to it in *any* shader. Because of this, we
+ * need to track the maximal access to the array as linking pulls
+ * more functions in that access the array.
+ */
+ var->max_array_access =
+ MAX2(var->max_array_access, ir->var->max_array_access);
+
+ if (var->type->length == 0 && ir->var->type->length != 0)
+ var->type = ir->var->type;
+ }
+
+ ir->var = var;
+ }
+
+ return visit_continue;
+ }
+
+ /** Was function linking successful? */
+ bool success;
+
+private:
+ /**
+ * Shader program being linked
+ *
+ * This is only used for logging error messages.
+ */
+ gl_shader_program *prog;
+
+ /** List of shaders available for linking. */
+ gl_shader **shader_list;
+
+ /** Number of shaders available for linking. */
+ unsigned num_shaders;
+
+ /**
+ * Final linked shader
+ *
+ * This is used two ways. It is used to find global variables in the
+ * linked shader that are accessed by the function. It is also used to add
+ * global variables from the shader where the function originated.
+ */
+ gl_shader *linked;
+
+ /**
+ * Table of variables local to the function.
+ */
+ hash_table *locals;
+};
+
+
+/**
+ * Searches a list of shaders for a particular function definition
+ */
+ir_function_signature *
+find_matching_signature(const char *name, const exec_list *actual_parameters,
+ gl_shader **shader_list, unsigned num_shaders)
+{
+ for (unsigned i = 0; i < num_shaders; i++) {
+ ir_function *const f = shader_list[i]->symbols->get_function(name);
+
+ if (f == NULL)
+ continue;
+
+ ir_function_signature *sig = f->matching_signature(actual_parameters);
+
+ if ((sig == NULL) || !sig->is_defined)
+ continue;
+
+ return sig;
+ }
+
+ return NULL;
+}
+
+
+bool
+link_function_calls(gl_shader_program *prog, gl_shader *main,
+ gl_shader **shader_list, unsigned num_shaders)
+{
+ call_link_visitor v(prog, main, shader_list, num_shaders);
+
+ v.run(main->ir);
+ return v.success;
+}
diff --git a/mesalib/src/glsl/linker.cpp b/mesalib/src/glsl/linker.cpp index a9a2a5e20..7760b4f2c 100644 --- a/mesalib/src/glsl/linker.cpp +++ b/mesalib/src/glsl/linker.cpp @@ -63,10 +63,6 @@ *
* \author Ian Romanick <ian.d.romanick@intel.com>
*/
-#include <cstdlib>
-#include <cstdio>
-#include <cstdarg>
-#include <climits>
#include "main/core.h"
#include "glsl_symbol_table.h"
diff --git a/mesalib/src/glsl/loop_controls.cpp b/mesalib/src/glsl/loop_controls.cpp index 9eaa50f22..9b77ef460 100644 --- a/mesalib/src/glsl/loop_controls.cpp +++ b/mesalib/src/glsl/loop_controls.cpp @@ -1,304 +1,304 @@ -/* - * Copyright © 2010 Intel Corporation - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING - * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER - * DEALINGS IN THE SOFTWARE. - */ - -#include <climits> -#include "main/compiler.h" -#include "glsl_types.h" -#include "loop_analysis.h" -#include "ir_hierarchical_visitor.h" - -/** - * Find an initializer of a variable outside a loop - * - * Works backwards from the loop to find the pre-loop value of the variable. - * This is used, for example, to find the initial value of loop induction - * variables. - * - * \param loop Loop where \c var is an induction variable - * \param var Variable whose initializer is to be found - * - * \return - * The \c ir_rvalue assigned to the variable outside the loop. May return - * \c NULL if no initializer can be found. - */ -ir_rvalue * -find_initial_value(ir_loop *loop, ir_variable *var) -{ - for (exec_node *node = loop->prev; - !node->is_head_sentinel(); - node = node->prev) { - ir_instruction *ir = (ir_instruction *) node; - - switch (ir->ir_type) { - case ir_type_call: - case ir_type_loop: - case ir_type_loop_jump: - case ir_type_return: - case ir_type_if: - return NULL; - - case ir_type_function: - case ir_type_function_signature: - assert(!"Should not get here."); - return NULL; - - case ir_type_assignment: { - ir_assignment *assign = ir->as_assignment(); - ir_variable *assignee = assign->lhs->whole_variable_referenced(); - - if (assignee == var) - return (assign->condition != NULL) ? NULL : assign->rhs; - - break; - } - - default: - break; - } - } - - return NULL; -} - - -int -calculate_iterations(ir_rvalue *from, ir_rvalue *to, ir_rvalue *increment, - enum ir_expression_operation op) -{ - if (from == NULL || to == NULL || increment == NULL) - return -1; - - void *mem_ctx = ralloc_context(NULL); - - ir_expression *const sub = - new(mem_ctx) ir_expression(ir_binop_sub, from->type, to, from); - - ir_expression *const div = - new(mem_ctx) ir_expression(ir_binop_div, sub->type, sub, increment); - - ir_constant *iter = div->constant_expression_value(); - - if (iter == NULL) - return -1; - - if (!iter->type->is_integer()) { - ir_rvalue *cast = - new(mem_ctx) ir_expression(ir_unop_f2i, glsl_type::int_type, iter, - NULL); - - iter = cast->constant_expression_value(); - } - - int iter_value = iter->get_int_component(0); - - /* Make sure that the calculated number of iterations satisfies the exit - * condition. This is needed to catch off-by-one errors and some types of - * ill-formed loops. For example, we need to detect that the following - * loop does not have a maximum iteration count. - * - * for (float x = 0.0; x != 0.9; x += 0.2) - * ; - */ - const int bias[] = { -1, 0, 1 }; - bool valid_loop = false; - - for (unsigned i = 0; i < Elements(bias); i++) { - iter = (increment->type->is_integer()) - ? new(mem_ctx) ir_constant(iter_value + bias[i]) - : new(mem_ctx) ir_constant(float(iter_value + bias[i])); - - ir_expression *const mul = - new(mem_ctx) ir_expression(ir_binop_mul, increment->type, iter, - increment); - - ir_expression *const add = - new(mem_ctx) ir_expression(ir_binop_add, mul->type, mul, from); - - ir_expression *const cmp = - new(mem_ctx) ir_expression(op, glsl_type::bool_type, add, to); - - ir_constant *const cmp_result = cmp->constant_expression_value(); - - assert(cmp_result != NULL); - if (cmp_result->get_bool_component(0)) { - iter_value += bias[i]; - valid_loop = true; - break; - } - } - - ralloc_free(mem_ctx); - return (valid_loop) ? iter_value : -1; -} - - -class loop_control_visitor : public ir_hierarchical_visitor { -public: - loop_control_visitor(loop_state *state) - { - this->state = state; - this->progress = false; - } - - virtual ir_visitor_status visit_leave(ir_loop *ir); - - loop_state *state; - - bool progress; -}; - - -ir_visitor_status -loop_control_visitor::visit_leave(ir_loop *ir) -{ - loop_variable_state *const ls = this->state->get(ir); - - /* If we've entered a loop that hasn't been analyzed, something really, - * really bad has happened. - */ - if (ls == NULL) { - assert(ls != NULL); - return visit_continue; - } - - /* Search the loop terminating conditions for one of the form 'i < c' where - * i is a loop induction variable, c is a constant, and < is any relative - * operator. - */ - int max_iterations = ls->max_iterations; - - if(ir->from && ir->to && ir->increment) - max_iterations = calculate_iterations(ir->from, ir->to, ir->increment, (ir_expression_operation)ir->cmp); - - if(max_iterations < 0) - max_iterations = INT_MAX; - - foreach_list(node, &ls->terminators) { - loop_terminator *t = (loop_terminator *) node; - ir_if *if_stmt = t->ir; - - /* If-statements can be either 'if (expr)' or 'if (deref)'. We only care - * about the former here. - */ - ir_expression *cond = if_stmt->condition->as_expression(); - if (cond == NULL) - continue; - - switch (cond->operation) { - case ir_binop_less: - case ir_binop_greater: - case ir_binop_lequal: - case ir_binop_gequal: { - /* The expressions that we care about will either be of the form - * 'counter < limit' or 'limit < counter'. Figure out which is - * which. - */ - ir_rvalue *counter = cond->operands[0]->as_dereference_variable(); - ir_constant *limit = cond->operands[1]->as_constant(); - enum ir_expression_operation cmp = cond->operation; - - if (limit == NULL) { - counter = cond->operands[1]->as_dereference_variable(); - limit = cond->operands[0]->as_constant(); - - switch (cmp) { - case ir_binop_less: cmp = ir_binop_gequal; break; - case ir_binop_greater: cmp = ir_binop_lequal; break; - case ir_binop_lequal: cmp = ir_binop_greater; break; - case ir_binop_gequal: cmp = ir_binop_less; break; - default: assert(!"Should not get here."); - } - } - - if ((counter == NULL) || (limit == NULL)) - break; - - ir_variable *var = counter->variable_referenced(); - - ir_rvalue *init = find_initial_value(ir, var); - - foreach_list(iv_node, &ls->induction_variables) { - loop_variable *lv = (loop_variable *) iv_node; - - if (lv->var == var) { - const int iterations = calculate_iterations(init, limit, - lv->increment, - cmp); - if (iterations >= 0) { - /* If the new iteration count is lower than the previously - * believed iteration count, update the loop control values. - */ - if (iterations < max_iterations) { - ir->from = init->clone(ir, NULL); - ir->to = limit->clone(ir, NULL); - ir->increment = lv->increment->clone(ir, NULL); - ir->counter = lv->var; - ir->cmp = cmp; - - max_iterations = iterations; - } - - /* Remove the conditional break statement. The loop - * controls are now set such that the exit condition will be - * satisfied. - */ - if_stmt->remove(); - - assert(ls->num_loop_jumps > 0); - ls->num_loop_jumps--; - - this->progress = true; - } - - break; - } - } - break; - } - - default: - break; - } - } - - /* If we have proven the one of the loop exit conditions is satisifed before - * running the loop once, remove the loop. - */ - if (max_iterations == 0) - ir->remove(); - else - ls->max_iterations = max_iterations; - - return visit_continue; -} - - -bool -set_loop_controls(exec_list *instructions, loop_state *ls) -{ - loop_control_visitor v(ls); - - v.run(instructions); - - return v.progress; -} +/*
+ * Copyright © 2010 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include <limits.h>
+#include "main/compiler.h"
+#include "glsl_types.h"
+#include "loop_analysis.h"
+#include "ir_hierarchical_visitor.h"
+
+/**
+ * Find an initializer of a variable outside a loop
+ *
+ * Works backwards from the loop to find the pre-loop value of the variable.
+ * This is used, for example, to find the initial value of loop induction
+ * variables.
+ *
+ * \param loop Loop where \c var is an induction variable
+ * \param var Variable whose initializer is to be found
+ *
+ * \return
+ * The \c ir_rvalue assigned to the variable outside the loop. May return
+ * \c NULL if no initializer can be found.
+ */
+ir_rvalue *
+find_initial_value(ir_loop *loop, ir_variable *var)
+{
+ for (exec_node *node = loop->prev;
+ !node->is_head_sentinel();
+ node = node->prev) {
+ ir_instruction *ir = (ir_instruction *) node;
+
+ switch (ir->ir_type) {
+ case ir_type_call:
+ case ir_type_loop:
+ case ir_type_loop_jump:
+ case ir_type_return:
+ case ir_type_if:
+ return NULL;
+
+ case ir_type_function:
+ case ir_type_function_signature:
+ assert(!"Should not get here.");
+ return NULL;
+
+ case ir_type_assignment: {
+ ir_assignment *assign = ir->as_assignment();
+ ir_variable *assignee = assign->lhs->whole_variable_referenced();
+
+ if (assignee == var)
+ return (assign->condition != NULL) ? NULL : assign->rhs;
+
+ break;
+ }
+
+ default:
+ break;
+ }
+ }
+
+ return NULL;
+}
+
+
+int
+calculate_iterations(ir_rvalue *from, ir_rvalue *to, ir_rvalue *increment,
+ enum ir_expression_operation op)
+{
+ if (from == NULL || to == NULL || increment == NULL)
+ return -1;
+
+ void *mem_ctx = ralloc_context(NULL);
+
+ ir_expression *const sub =
+ new(mem_ctx) ir_expression(ir_binop_sub, from->type, to, from);
+
+ ir_expression *const div =
+ new(mem_ctx) ir_expression(ir_binop_div, sub->type, sub, increment);
+
+ ir_constant *iter = div->constant_expression_value();
+
+ if (iter == NULL)
+ return -1;
+
+ if (!iter->type->is_integer()) {
+ ir_rvalue *cast =
+ new(mem_ctx) ir_expression(ir_unop_f2i, glsl_type::int_type, iter,
+ NULL);
+
+ iter = cast->constant_expression_value();
+ }
+
+ int iter_value = iter->get_int_component(0);
+
+ /* Make sure that the calculated number of iterations satisfies the exit
+ * condition. This is needed to catch off-by-one errors and some types of
+ * ill-formed loops. For example, we need to detect that the following
+ * loop does not have a maximum iteration count.
+ *
+ * for (float x = 0.0; x != 0.9; x += 0.2)
+ * ;
+ */
+ const int bias[] = { -1, 0, 1 };
+ bool valid_loop = false;
+
+ for (unsigned i = 0; i < Elements(bias); i++) {
+ iter = (increment->type->is_integer())
+ ? new(mem_ctx) ir_constant(iter_value + bias[i])
+ : new(mem_ctx) ir_constant(float(iter_value + bias[i]));
+
+ ir_expression *const mul =
+ new(mem_ctx) ir_expression(ir_binop_mul, increment->type, iter,
+ increment);
+
+ ir_expression *const add =
+ new(mem_ctx) ir_expression(ir_binop_add, mul->type, mul, from);
+
+ ir_expression *const cmp =
+ new(mem_ctx) ir_expression(op, glsl_type::bool_type, add, to);
+
+ ir_constant *const cmp_result = cmp->constant_expression_value();
+
+ assert(cmp_result != NULL);
+ if (cmp_result->get_bool_component(0)) {
+ iter_value += bias[i];
+ valid_loop = true;
+ break;
+ }
+ }
+
+ ralloc_free(mem_ctx);
+ return (valid_loop) ? iter_value : -1;
+}
+
+
+class loop_control_visitor : public ir_hierarchical_visitor {
+public:
+ loop_control_visitor(loop_state *state)
+ {
+ this->state = state;
+ this->progress = false;
+ }
+
+ virtual ir_visitor_status visit_leave(ir_loop *ir);
+
+ loop_state *state;
+
+ bool progress;
+};
+
+
+ir_visitor_status
+loop_control_visitor::visit_leave(ir_loop *ir)
+{
+ loop_variable_state *const ls = this->state->get(ir);
+
+ /* If we've entered a loop that hasn't been analyzed, something really,
+ * really bad has happened.
+ */
+ if (ls == NULL) {
+ assert(ls != NULL);
+ return visit_continue;
+ }
+
+ /* Search the loop terminating conditions for one of the form 'i < c' where
+ * i is a loop induction variable, c is a constant, and < is any relative
+ * operator.
+ */
+ int max_iterations = ls->max_iterations;
+
+ if(ir->from && ir->to && ir->increment)
+ max_iterations = calculate_iterations(ir->from, ir->to, ir->increment, (ir_expression_operation)ir->cmp);
+
+ if(max_iterations < 0)
+ max_iterations = INT_MAX;
+
+ foreach_list(node, &ls->terminators) {
+ loop_terminator *t = (loop_terminator *) node;
+ ir_if *if_stmt = t->ir;
+
+ /* If-statements can be either 'if (expr)' or 'if (deref)'. We only care
+ * about the former here.
+ */
+ ir_expression *cond = if_stmt->condition->as_expression();
+ if (cond == NULL)
+ continue;
+
+ switch (cond->operation) {
+ case ir_binop_less:
+ case ir_binop_greater:
+ case ir_binop_lequal:
+ case ir_binop_gequal: {
+ /* The expressions that we care about will either be of the form
+ * 'counter < limit' or 'limit < counter'. Figure out which is
+ * which.
+ */
+ ir_rvalue *counter = cond->operands[0]->as_dereference_variable();
+ ir_constant *limit = cond->operands[1]->as_constant();
+ enum ir_expression_operation cmp = cond->operation;
+
+ if (limit == NULL) {
+ counter = cond->operands[1]->as_dereference_variable();
+ limit = cond->operands[0]->as_constant();
+
+ switch (cmp) {
+ case ir_binop_less: cmp = ir_binop_gequal; break;
+ case ir_binop_greater: cmp = ir_binop_lequal; break;
+ case ir_binop_lequal: cmp = ir_binop_greater; break;
+ case ir_binop_gequal: cmp = ir_binop_less; break;
+ default: assert(!"Should not get here.");
+ }
+ }
+
+ if ((counter == NULL) || (limit == NULL))
+ break;
+
+ ir_variable *var = counter->variable_referenced();
+
+ ir_rvalue *init = find_initial_value(ir, var);
+
+ foreach_list(iv_node, &ls->induction_variables) {
+ loop_variable *lv = (loop_variable *) iv_node;
+
+ if (lv->var == var) {
+ const int iterations = calculate_iterations(init, limit,
+ lv->increment,
+ cmp);
+ if (iterations >= 0) {
+ /* If the new iteration count is lower than the previously
+ * believed iteration count, update the loop control values.
+ */
+ if (iterations < max_iterations) {
+ ir->from = init->clone(ir, NULL);
+ ir->to = limit->clone(ir, NULL);
+ ir->increment = lv->increment->clone(ir, NULL);
+ ir->counter = lv->var;
+ ir->cmp = cmp;
+
+ max_iterations = iterations;
+ }
+
+ /* Remove the conditional break statement. The loop
+ * controls are now set such that the exit condition will be
+ * satisfied.
+ */
+ if_stmt->remove();
+
+ assert(ls->num_loop_jumps > 0);
+ ls->num_loop_jumps--;
+
+ this->progress = true;
+ }
+
+ break;
+ }
+ }
+ break;
+ }
+
+ default:
+ break;
+ }
+ }
+
+ /* If we have proven the one of the loop exit conditions is satisifed before
+ * running the loop once, remove the loop.
+ */
+ if (max_iterations == 0)
+ ir->remove();
+ else
+ ls->max_iterations = max_iterations;
+
+ return visit_continue;
+}
+
+
+bool
+set_loop_controls(exec_list *instructions, loop_state *ls)
+{
+ loop_control_visitor v(ls);
+
+ v.run(instructions);
+
+ return v.progress;
+}
diff --git a/mesalib/src/glsl/lower_mat_op_to_vec.cpp b/mesalib/src/glsl/lower_mat_op_to_vec.cpp index bdc53a1f8..dc46ed417 100644 --- a/mesalib/src/glsl/lower_mat_op_to_vec.cpp +++ b/mesalib/src/glsl/lower_mat_op_to_vec.cpp @@ -1,493 +1,490 @@ -/* - * Copyright © 2010 Intel Corporation - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING - * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER - * DEALINGS IN THE SOFTWARE. - */ - -/** - * \file lower_mat_op_to_vec.cpp - * - * Breaks matrix operation expressions down to a series of vector operations. - * - * Generally this is how we have to codegen matrix operations for a - * GPU, so this gives us the chance to constant fold operations on a - * column or row. - */ - -#include "ir.h" -#include "ir_expression_flattening.h" -#include "glsl_types.h" - -using std::abort; -using std::printf; - -class ir_mat_op_to_vec_visitor : public ir_hierarchical_visitor { -public: - ir_mat_op_to_vec_visitor() - { - this->made_progress = false; - this->mem_ctx = NULL; - } - - ir_visitor_status visit_leave(ir_assignment *); - - ir_dereference *get_column(ir_variable *var, int col); - ir_rvalue *get_element(ir_variable *var, int col, int row); - - void do_mul_mat_mat(ir_variable *result_var, - ir_variable *a_var, ir_variable *b_var); - void do_mul_mat_vec(ir_variable *result_var, - ir_variable *a_var, ir_variable *b_var); - void do_mul_vec_mat(ir_variable *result_var, - ir_variable *a_var, ir_variable *b_var); - void do_mul_mat_scalar(ir_variable *result_var, - ir_variable *a_var, ir_variable *b_var); - void do_equal_mat_mat(ir_variable *result_var, ir_variable *a_var, - ir_variable *b_var, bool test_equal); - - void *mem_ctx; - bool made_progress; -}; - -static bool -mat_op_to_vec_predicate(ir_instruction *ir) -{ - ir_expression *expr = ir->as_expression(); - unsigned int i; - - if (!expr) - return false; - - for (i = 0; i < expr->get_num_operands(); i++) { - if (expr->operands[i]->type->is_matrix()) - return true; - } - - return false; -} - -bool -do_mat_op_to_vec(exec_list *instructions) -{ - ir_mat_op_to_vec_visitor v; - - /* Pull out any matrix expression to a separate assignment to a - * temp. This will make our handling of the breakdown to - * operations on the matrix's vector components much easier. - */ - do_expression_flattening(instructions, mat_op_to_vec_predicate); - - visit_list_elements(&v, instructions); - - return v.made_progress; -} - -ir_rvalue * -ir_mat_op_to_vec_visitor::get_element(ir_variable *var, int col, int row) -{ - ir_dereference *deref; - - deref = new(mem_ctx) ir_dereference_variable(var); - - if (var->type->is_matrix()) { - deref = new(mem_ctx) ir_dereference_array(var, - new(mem_ctx) ir_constant(col)); - } else { - assert(col == 0); - } - - return new(mem_ctx) ir_swizzle(deref, row, 0, 0, 0, 1); -} - -ir_dereference * -ir_mat_op_to_vec_visitor::get_column(ir_variable *var, int row) -{ - ir_dereference *deref; - - if (!var->type->is_matrix()) { - deref = new(mem_ctx) ir_dereference_variable(var); - } else { - deref = new(mem_ctx) ir_dereference_variable(var); - deref = new(mem_ctx) ir_dereference_array(deref, - new(mem_ctx) ir_constant(row)); - } - - return deref; -} - -void -ir_mat_op_to_vec_visitor::do_mul_mat_mat(ir_variable *result_var, - ir_variable *a_var, - ir_variable *b_var) -{ - int b_col, i; - ir_assignment *assign; - ir_expression *expr; - - for (b_col = 0; b_col < b_var->type->matrix_columns; b_col++) { - ir_rvalue *a = get_column(a_var, 0); - ir_rvalue *b = get_element(b_var, b_col, 0); - - /* first column */ - expr = new(mem_ctx) ir_expression(ir_binop_mul, - a->type, - a, - b); - - /* following columns */ - for (i = 1; i < a_var->type->matrix_columns; i++) { - ir_expression *mul_expr; - - a = get_column(a_var, i); - b = get_element(b_var, b_col, i); - - mul_expr = new(mem_ctx) ir_expression(ir_binop_mul, - a->type, - a, - b); - expr = new(mem_ctx) ir_expression(ir_binop_add, - a->type, - expr, - mul_expr); - } - - ir_rvalue *result = get_column(result_var, b_col); - assign = new(mem_ctx) ir_assignment(result, - expr, - NULL); - base_ir->insert_before(assign); - } -} - -void -ir_mat_op_to_vec_visitor::do_mul_mat_vec(ir_variable *result_var, - ir_variable *a_var, - ir_variable *b_var) -{ - int i; - ir_rvalue *a = get_column(a_var, 0); - ir_rvalue *b = get_element(b_var, 0, 0); - ir_assignment *assign; - ir_expression *expr; - - /* first column */ - expr = new(mem_ctx) ir_expression(ir_binop_mul, - result_var->type, - a, - b); - - /* following columns */ - for (i = 1; i < a_var->type->matrix_columns; i++) { - ir_expression *mul_expr; - - a = get_column(a_var, i); - b = get_element(b_var, 0, i); - - mul_expr = new(mem_ctx) ir_expression(ir_binop_mul, - result_var->type, - a, - b); - expr = new(mem_ctx) ir_expression(ir_binop_add, - result_var->type, - expr, - mul_expr); - } - - ir_rvalue *result = new(mem_ctx) ir_dereference_variable(result_var); - assign = new(mem_ctx) ir_assignment(result, - expr, - NULL); - base_ir->insert_before(assign); -} - -void -ir_mat_op_to_vec_visitor::do_mul_vec_mat(ir_variable *result_var, - ir_variable *a_var, - ir_variable *b_var) -{ - int i; - - for (i = 0; i < b_var->type->matrix_columns; i++) { - ir_rvalue *a = new(mem_ctx) ir_dereference_variable(a_var); - ir_rvalue *b = get_column(b_var, i); - ir_rvalue *result; - ir_expression *column_expr; - ir_assignment *column_assign; - - result = new(mem_ctx) ir_dereference_variable(result_var); - result = new(mem_ctx) ir_swizzle(result, i, 0, 0, 0, 1); - - column_expr = new(mem_ctx) ir_expression(ir_binop_dot, - result->type, - a, - b); - - column_assign = new(mem_ctx) ir_assignment(result, - column_expr, - NULL); - base_ir->insert_before(column_assign); - } -} - -void -ir_mat_op_to_vec_visitor::do_mul_mat_scalar(ir_variable *result_var, - ir_variable *a_var, - ir_variable *b_var) -{ - int i; - - for (i = 0; i < a_var->type->matrix_columns; i++) { - ir_rvalue *a = get_column(a_var, i); - ir_rvalue *b = new(mem_ctx) ir_dereference_variable(b_var); - ir_rvalue *result = get_column(result_var, i); - ir_expression *column_expr; - ir_assignment *column_assign; - - column_expr = new(mem_ctx) ir_expression(ir_binop_mul, - result->type, - a, - b); - - column_assign = new(mem_ctx) ir_assignment(result, - column_expr, - NULL); - base_ir->insert_before(column_assign); - } -} - -void -ir_mat_op_to_vec_visitor::do_equal_mat_mat(ir_variable *result_var, - ir_variable *a_var, - ir_variable *b_var, - bool test_equal) -{ - /* This essentially implements the following GLSL: - * - * bool equal(mat4 a, mat4 b) - * { - * return !any(bvec4(a[0] != b[0], - * a[1] != b[1], - * a[2] != b[2], - * a[3] != b[3]); - * } - * - * bool nequal(mat4 a, mat4 b) - * { - * return any(bvec4(a[0] != b[0], - * a[1] != b[1], - * a[2] != b[2], - * a[3] != b[3]); - * } - */ - const unsigned columns = a_var->type->matrix_columns; - const glsl_type *const bvec_type = - glsl_type::get_instance(GLSL_TYPE_BOOL, columns, 1); - - ir_variable *const tmp_bvec = - new(this->mem_ctx) ir_variable(bvec_type, "mat_cmp_bvec", - ir_var_temporary); - this->base_ir->insert_before(tmp_bvec); - - for (unsigned i = 0; i < columns; i++) { - ir_dereference *const op0 = get_column(a_var, i); - ir_dereference *const op1 = get_column(b_var, i); - - ir_expression *const cmp = - new(this->mem_ctx) ir_expression(ir_binop_any_nequal, - glsl_type::bool_type, op0, op1); - - ir_dereference *const lhs = - new(this->mem_ctx) ir_dereference_variable(tmp_bvec); - - ir_assignment *const assign = - new(this->mem_ctx) ir_assignment(lhs, cmp, NULL, (1U << i)); - - this->base_ir->insert_before(assign); - } - - ir_rvalue *const val = - new(this->mem_ctx) ir_dereference_variable(tmp_bvec); - - ir_expression *any = - new(this->mem_ctx) ir_expression(ir_unop_any, glsl_type::bool_type, - val, NULL); - - if (test_equal) - any = new(this->mem_ctx) ir_expression(ir_unop_logic_not, - glsl_type::bool_type, - any, NULL); - - ir_rvalue *const result = - new(this->mem_ctx) ir_dereference_variable(result_var); - - ir_assignment *const assign = - new(mem_ctx) ir_assignment(result, any, NULL); - base_ir->insert_before(assign); -} - -static bool -has_matrix_operand(const ir_expression *expr, unsigned &columns) -{ - for (unsigned i = 0; i < expr->get_num_operands(); i++) { - if (expr->operands[i]->type->is_matrix()) { - columns = expr->operands[i]->type->matrix_columns; - return true; - } - } - - return false; -} - - -ir_visitor_status -ir_mat_op_to_vec_visitor::visit_leave(ir_assignment *orig_assign) -{ - ir_expression *orig_expr = orig_assign->rhs->as_expression(); - unsigned int i, matrix_columns = 1; - ir_variable *op_var[2]; - - if (!orig_expr) - return visit_continue; - - if (!has_matrix_operand(orig_expr, matrix_columns)) - return visit_continue; - - assert(orig_expr->get_num_operands() <= 2); - - mem_ctx = ralloc_parent(orig_assign); - - ir_dereference_variable *lhs_deref = - orig_assign->lhs->as_dereference_variable(); - assert(lhs_deref); - - ir_variable *result_var = lhs_deref->var; - - /* Store the expression operands in temps so we can use them - * multiple times. - */ - for (i = 0; i < orig_expr->get_num_operands(); i++) { - ir_assignment *assign; - - op_var[i] = new(mem_ctx) ir_variable(orig_expr->operands[i]->type, - "mat_op_to_vec", - ir_var_temporary); - base_ir->insert_before(op_var[i]); - - lhs_deref = new(mem_ctx) ir_dereference_variable(op_var[i]); - assign = new(mem_ctx) ir_assignment(lhs_deref, - orig_expr->operands[i], - NULL); - base_ir->insert_before(assign); - } - - /* OK, time to break down this matrix operation. */ - switch (orig_expr->operation) { - case ir_unop_neg: { - const unsigned mask = (1U << result_var->type->vector_elements) - 1; - - /* Apply the operation to each column.*/ - for (i = 0; i < matrix_columns; i++) { - ir_rvalue *op0 = get_column(op_var[0], i); - ir_dereference *result = get_column(result_var, i); - ir_expression *column_expr; - ir_assignment *column_assign; - - column_expr = new(mem_ctx) ir_expression(orig_expr->operation, - result->type, - op0, - NULL); - - column_assign = new(mem_ctx) ir_assignment(result, - column_expr, - NULL, - mask); - assert(column_assign->write_mask != 0); - base_ir->insert_before(column_assign); - } - break; - } - case ir_binop_add: - case ir_binop_sub: - case ir_binop_div: - case ir_binop_mod: { - const unsigned mask = (1U << result_var->type->vector_elements) - 1; - - /* For most operations, the matrix version is just going - * column-wise through and applying the operation to each column - * if available. - */ - for (i = 0; i < matrix_columns; i++) { - ir_rvalue *op0 = get_column(op_var[0], i); - ir_rvalue *op1 = get_column(op_var[1], i); - ir_dereference *result = get_column(result_var, i); - ir_expression *column_expr; - ir_assignment *column_assign; - - column_expr = new(mem_ctx) ir_expression(orig_expr->operation, - result->type, - op0, - op1); - - column_assign = new(mem_ctx) ir_assignment(result, - column_expr, - NULL, - mask); - assert(column_assign->write_mask != 0); - base_ir->insert_before(column_assign); - } - break; - } - case ir_binop_mul: - if (op_var[0]->type->is_matrix()) { - if (op_var[1]->type->is_matrix()) { - do_mul_mat_mat(result_var, op_var[0], op_var[1]); - } else if (op_var[1]->type->is_vector()) { - do_mul_mat_vec(result_var, op_var[0], op_var[1]); - } else { - assert(op_var[1]->type->is_scalar()); - do_mul_mat_scalar(result_var, op_var[0], op_var[1]); - } - } else { - assert(op_var[1]->type->is_matrix()); - if (op_var[0]->type->is_vector()) { - do_mul_vec_mat(result_var, op_var[0], op_var[1]); - } else { - assert(op_var[0]->type->is_scalar()); - do_mul_mat_scalar(result_var, op_var[1], op_var[0]); - } - } - break; - - case ir_binop_all_equal: - case ir_binop_any_nequal: - do_equal_mat_mat(result_var, op_var[1], op_var[0], - (orig_expr->operation == ir_binop_all_equal)); - break; - - default: - printf("FINISHME: Handle matrix operation for %s\n", - orig_expr->operator_string()); - abort(); - } - orig_assign->remove(); - this->made_progress = true; - - return visit_continue; -} +/*
+ * Copyright © 2010 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file lower_mat_op_to_vec.cpp
+ *
+ * Breaks matrix operation expressions down to a series of vector operations.
+ *
+ * Generally this is how we have to codegen matrix operations for a
+ * GPU, so this gives us the chance to constant fold operations on a
+ * column or row.
+ */
+
+#include "ir.h"
+#include "ir_expression_flattening.h"
+#include "glsl_types.h"
+
+class ir_mat_op_to_vec_visitor : public ir_hierarchical_visitor {
+public:
+ ir_mat_op_to_vec_visitor()
+ {
+ this->made_progress = false;
+ this->mem_ctx = NULL;
+ }
+
+ ir_visitor_status visit_leave(ir_assignment *);
+
+ ir_dereference *get_column(ir_variable *var, int col);
+ ir_rvalue *get_element(ir_variable *var, int col, int row);
+
+ void do_mul_mat_mat(ir_variable *result_var,
+ ir_variable *a_var, ir_variable *b_var);
+ void do_mul_mat_vec(ir_variable *result_var,
+ ir_variable *a_var, ir_variable *b_var);
+ void do_mul_vec_mat(ir_variable *result_var,
+ ir_variable *a_var, ir_variable *b_var);
+ void do_mul_mat_scalar(ir_variable *result_var,
+ ir_variable *a_var, ir_variable *b_var);
+ void do_equal_mat_mat(ir_variable *result_var, ir_variable *a_var,
+ ir_variable *b_var, bool test_equal);
+
+ void *mem_ctx;
+ bool made_progress;
+};
+
+static bool
+mat_op_to_vec_predicate(ir_instruction *ir)
+{
+ ir_expression *expr = ir->as_expression();
+ unsigned int i;
+
+ if (!expr)
+ return false;
+
+ for (i = 0; i < expr->get_num_operands(); i++) {
+ if (expr->operands[i]->type->is_matrix())
+ return true;
+ }
+
+ return false;
+}
+
+bool
+do_mat_op_to_vec(exec_list *instructions)
+{
+ ir_mat_op_to_vec_visitor v;
+
+ /* Pull out any matrix expression to a separate assignment to a
+ * temp. This will make our handling of the breakdown to
+ * operations on the matrix's vector components much easier.
+ */
+ do_expression_flattening(instructions, mat_op_to_vec_predicate);
+
+ visit_list_elements(&v, instructions);
+
+ return v.made_progress;
+}
+
+ir_rvalue *
+ir_mat_op_to_vec_visitor::get_element(ir_variable *var, int col, int row)
+{
+ ir_dereference *deref;
+
+ deref = new(mem_ctx) ir_dereference_variable(var);
+
+ if (var->type->is_matrix()) {
+ deref = new(mem_ctx) ir_dereference_array(var,
+ new(mem_ctx) ir_constant(col));
+ } else {
+ assert(col == 0);
+ }
+
+ return new(mem_ctx) ir_swizzle(deref, row, 0, 0, 0, 1);
+}
+
+ir_dereference *
+ir_mat_op_to_vec_visitor::get_column(ir_variable *var, int row)
+{
+ ir_dereference *deref;
+
+ if (!var->type->is_matrix()) {
+ deref = new(mem_ctx) ir_dereference_variable(var);
+ } else {
+ deref = new(mem_ctx) ir_dereference_variable(var);
+ deref = new(mem_ctx) ir_dereference_array(deref,
+ new(mem_ctx) ir_constant(row));
+ }
+
+ return deref;
+}
+
+void
+ir_mat_op_to_vec_visitor::do_mul_mat_mat(ir_variable *result_var,
+ ir_variable *a_var,
+ ir_variable *b_var)
+{
+ int b_col, i;
+ ir_assignment *assign;
+ ir_expression *expr;
+
+ for (b_col = 0; b_col < b_var->type->matrix_columns; b_col++) {
+ ir_rvalue *a = get_column(a_var, 0);
+ ir_rvalue *b = get_element(b_var, b_col, 0);
+
+ /* first column */
+ expr = new(mem_ctx) ir_expression(ir_binop_mul,
+ a->type,
+ a,
+ b);
+
+ /* following columns */
+ for (i = 1; i < a_var->type->matrix_columns; i++) {
+ ir_expression *mul_expr;
+
+ a = get_column(a_var, i);
+ b = get_element(b_var, b_col, i);
+
+ mul_expr = new(mem_ctx) ir_expression(ir_binop_mul,
+ a->type,
+ a,
+ b);
+ expr = new(mem_ctx) ir_expression(ir_binop_add,
+ a->type,
+ expr,
+ mul_expr);
+ }
+
+ ir_rvalue *result = get_column(result_var, b_col);
+ assign = new(mem_ctx) ir_assignment(result,
+ expr,
+ NULL);
+ base_ir->insert_before(assign);
+ }
+}
+
+void
+ir_mat_op_to_vec_visitor::do_mul_mat_vec(ir_variable *result_var,
+ ir_variable *a_var,
+ ir_variable *b_var)
+{
+ int i;
+ ir_rvalue *a = get_column(a_var, 0);
+ ir_rvalue *b = get_element(b_var, 0, 0);
+ ir_assignment *assign;
+ ir_expression *expr;
+
+ /* first column */
+ expr = new(mem_ctx) ir_expression(ir_binop_mul,
+ result_var->type,
+ a,
+ b);
+
+ /* following columns */
+ for (i = 1; i < a_var->type->matrix_columns; i++) {
+ ir_expression *mul_expr;
+
+ a = get_column(a_var, i);
+ b = get_element(b_var, 0, i);
+
+ mul_expr = new(mem_ctx) ir_expression(ir_binop_mul,
+ result_var->type,
+ a,
+ b);
+ expr = new(mem_ctx) ir_expression(ir_binop_add,
+ result_var->type,
+ expr,
+ mul_expr);
+ }
+
+ ir_rvalue *result = new(mem_ctx) ir_dereference_variable(result_var);
+ assign = new(mem_ctx) ir_assignment(result,
+ expr,
+ NULL);
+ base_ir->insert_before(assign);
+}
+
+void
+ir_mat_op_to_vec_visitor::do_mul_vec_mat(ir_variable *result_var,
+ ir_variable *a_var,
+ ir_variable *b_var)
+{
+ int i;
+
+ for (i = 0; i < b_var->type->matrix_columns; i++) {
+ ir_rvalue *a = new(mem_ctx) ir_dereference_variable(a_var);
+ ir_rvalue *b = get_column(b_var, i);
+ ir_rvalue *result;
+ ir_expression *column_expr;
+ ir_assignment *column_assign;
+
+ result = new(mem_ctx) ir_dereference_variable(result_var);
+ result = new(mem_ctx) ir_swizzle(result, i, 0, 0, 0, 1);
+
+ column_expr = new(mem_ctx) ir_expression(ir_binop_dot,
+ result->type,
+ a,
+ b);
+
+ column_assign = new(mem_ctx) ir_assignment(result,
+ column_expr,
+ NULL);
+ base_ir->insert_before(column_assign);
+ }
+}
+
+void
+ir_mat_op_to_vec_visitor::do_mul_mat_scalar(ir_variable *result_var,
+ ir_variable *a_var,
+ ir_variable *b_var)
+{
+ int i;
+
+ for (i = 0; i < a_var->type->matrix_columns; i++) {
+ ir_rvalue *a = get_column(a_var, i);
+ ir_rvalue *b = new(mem_ctx) ir_dereference_variable(b_var);
+ ir_rvalue *result = get_column(result_var, i);
+ ir_expression *column_expr;
+ ir_assignment *column_assign;
+
+ column_expr = new(mem_ctx) ir_expression(ir_binop_mul,
+ result->type,
+ a,
+ b);
+
+ column_assign = new(mem_ctx) ir_assignment(result,
+ column_expr,
+ NULL);
+ base_ir->insert_before(column_assign);
+ }
+}
+
+void
+ir_mat_op_to_vec_visitor::do_equal_mat_mat(ir_variable *result_var,
+ ir_variable *a_var,
+ ir_variable *b_var,
+ bool test_equal)
+{
+ /* This essentially implements the following GLSL:
+ *
+ * bool equal(mat4 a, mat4 b)
+ * {
+ * return !any(bvec4(a[0] != b[0],
+ * a[1] != b[1],
+ * a[2] != b[2],
+ * a[3] != b[3]);
+ * }
+ *
+ * bool nequal(mat4 a, mat4 b)
+ * {
+ * return any(bvec4(a[0] != b[0],
+ * a[1] != b[1],
+ * a[2] != b[2],
+ * a[3] != b[3]);
+ * }
+ */
+ const unsigned columns = a_var->type->matrix_columns;
+ const glsl_type *const bvec_type =
+ glsl_type::get_instance(GLSL_TYPE_BOOL, columns, 1);
+
+ ir_variable *const tmp_bvec =
+ new(this->mem_ctx) ir_variable(bvec_type, "mat_cmp_bvec",
+ ir_var_temporary);
+ this->base_ir->insert_before(tmp_bvec);
+
+ for (unsigned i = 0; i < columns; i++) {
+ ir_dereference *const op0 = get_column(a_var, i);
+ ir_dereference *const op1 = get_column(b_var, i);
+
+ ir_expression *const cmp =
+ new(this->mem_ctx) ir_expression(ir_binop_any_nequal,
+ glsl_type::bool_type, op0, op1);
+
+ ir_dereference *const lhs =
+ new(this->mem_ctx) ir_dereference_variable(tmp_bvec);
+
+ ir_assignment *const assign =
+ new(this->mem_ctx) ir_assignment(lhs, cmp, NULL, (1U << i));
+
+ this->base_ir->insert_before(assign);
+ }
+
+ ir_rvalue *const val =
+ new(this->mem_ctx) ir_dereference_variable(tmp_bvec);
+
+ ir_expression *any =
+ new(this->mem_ctx) ir_expression(ir_unop_any, glsl_type::bool_type,
+ val, NULL);
+
+ if (test_equal)
+ any = new(this->mem_ctx) ir_expression(ir_unop_logic_not,
+ glsl_type::bool_type,
+ any, NULL);
+
+ ir_rvalue *const result =
+ new(this->mem_ctx) ir_dereference_variable(result_var);
+
+ ir_assignment *const assign =
+ new(mem_ctx) ir_assignment(result, any, NULL);
+ base_ir->insert_before(assign);
+}
+
+static bool
+has_matrix_operand(const ir_expression *expr, unsigned &columns)
+{
+ for (unsigned i = 0; i < expr->get_num_operands(); i++) {
+ if (expr->operands[i]->type->is_matrix()) {
+ columns = expr->operands[i]->type->matrix_columns;
+ return true;
+ }
+ }
+
+ return false;
+}
+
+
+ir_visitor_status
+ir_mat_op_to_vec_visitor::visit_leave(ir_assignment *orig_assign)
+{
+ ir_expression *orig_expr = orig_assign->rhs->as_expression();
+ unsigned int i, matrix_columns = 1;
+ ir_variable *op_var[2];
+
+ if (!orig_expr)
+ return visit_continue;
+
+ if (!has_matrix_operand(orig_expr, matrix_columns))
+ return visit_continue;
+
+ assert(orig_expr->get_num_operands() <= 2);
+
+ mem_ctx = ralloc_parent(orig_assign);
+
+ ir_dereference_variable *lhs_deref =
+ orig_assign->lhs->as_dereference_variable();
+ assert(lhs_deref);
+
+ ir_variable *result_var = lhs_deref->var;
+
+ /* Store the expression operands in temps so we can use them
+ * multiple times.
+ */
+ for (i = 0; i < orig_expr->get_num_operands(); i++) {
+ ir_assignment *assign;
+
+ op_var[i] = new(mem_ctx) ir_variable(orig_expr->operands[i]->type,
+ "mat_op_to_vec",
+ ir_var_temporary);
+ base_ir->insert_before(op_var[i]);
+
+ lhs_deref = new(mem_ctx) ir_dereference_variable(op_var[i]);
+ assign = new(mem_ctx) ir_assignment(lhs_deref,
+ orig_expr->operands[i],
+ NULL);
+ base_ir->insert_before(assign);
+ }
+
+ /* OK, time to break down this matrix operation. */
+ switch (orig_expr->operation) {
+ case ir_unop_neg: {
+ const unsigned mask = (1U << result_var->type->vector_elements) - 1;
+
+ /* Apply the operation to each column.*/
+ for (i = 0; i < matrix_columns; i++) {
+ ir_rvalue *op0 = get_column(op_var[0], i);
+ ir_dereference *result = get_column(result_var, i);
+ ir_expression *column_expr;
+ ir_assignment *column_assign;
+
+ column_expr = new(mem_ctx) ir_expression(orig_expr->operation,
+ result->type,
+ op0,
+ NULL);
+
+ column_assign = new(mem_ctx) ir_assignment(result,
+ column_expr,
+ NULL,
+ mask);
+ assert(column_assign->write_mask != 0);
+ base_ir->insert_before(column_assign);
+ }
+ break;
+ }
+ case ir_binop_add:
+ case ir_binop_sub:
+ case ir_binop_div:
+ case ir_binop_mod: {
+ const unsigned mask = (1U << result_var->type->vector_elements) - 1;
+
+ /* For most operations, the matrix version is just going
+ * column-wise through and applying the operation to each column
+ * if available.
+ */
+ for (i = 0; i < matrix_columns; i++) {
+ ir_rvalue *op0 = get_column(op_var[0], i);
+ ir_rvalue *op1 = get_column(op_var[1], i);
+ ir_dereference *result = get_column(result_var, i);
+ ir_expression *column_expr;
+ ir_assignment *column_assign;
+
+ column_expr = new(mem_ctx) ir_expression(orig_expr->operation,
+ result->type,
+ op0,
+ op1);
+
+ column_assign = new(mem_ctx) ir_assignment(result,
+ column_expr,
+ NULL,
+ mask);
+ assert(column_assign->write_mask != 0);
+ base_ir->insert_before(column_assign);
+ }
+ break;
+ }
+ case ir_binop_mul:
+ if (op_var[0]->type->is_matrix()) {
+ if (op_var[1]->type->is_matrix()) {
+ do_mul_mat_mat(result_var, op_var[0], op_var[1]);
+ } else if (op_var[1]->type->is_vector()) {
+ do_mul_mat_vec(result_var, op_var[0], op_var[1]);
+ } else {
+ assert(op_var[1]->type->is_scalar());
+ do_mul_mat_scalar(result_var, op_var[0], op_var[1]);
+ }
+ } else {
+ assert(op_var[1]->type->is_matrix());
+ if (op_var[0]->type->is_vector()) {
+ do_mul_vec_mat(result_var, op_var[0], op_var[1]);
+ } else {
+ assert(op_var[0]->type->is_scalar());
+ do_mul_mat_scalar(result_var, op_var[1], op_var[0]);
+ }
+ }
+ break;
+
+ case ir_binop_all_equal:
+ case ir_binop_any_nequal:
+ do_equal_mat_mat(result_var, op_var[1], op_var[0],
+ (orig_expr->operation == ir_binop_all_equal));
+ break;
+
+ default:
+ printf("FINISHME: Handle matrix operation for %s\n",
+ orig_expr->operator_string());
+ abort();
+ }
+ orig_assign->remove();
+ this->made_progress = true;
+
+ return visit_continue;
+}
diff --git a/mesalib/src/glsl/main.cpp b/mesalib/src/glsl/main.cpp index 6321a4a17..873a4fa16 100644 --- a/mesalib/src/glsl/main.cpp +++ b/mesalib/src/glsl/main.cpp @@ -1,347 +1,345 @@ -/* - * Copyright © 2008, 2009 Intel Corporation - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING - * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER - * DEALINGS IN THE SOFTWARE. - */ -#include <cstdlib> -#include <cstdio> -#include <io.h> - -#ifdef _MSC_VER -#define __STDC__ 1 -#include <getopt.h> -#define open _open -#define read _read -#define fstat _fstat -#define stat _stat -#define close _close -#define O_RDONLY _O_RDONLY -#endif - -#include "ast.h" -#include "glsl_parser_extras.h" -#include "glsl_parser.h" -#include "ir_optimization.h" -#include "ir_print_visitor.h" -#include "program.h" -#include "loop_analysis.h" - -extern "C" struct gl_shader * -_mesa_new_shader(struct gl_context *ctx, GLuint name, GLenum type); - -extern "C" void -_mesa_reference_shader(struct gl_context *ctx, struct gl_shader **ptr, - struct gl_shader *sh); - -/* Copied from shader_api.c for the stand-alone compiler. - */ -void -_mesa_reference_shader(struct gl_context *ctx, struct gl_shader **ptr, - struct gl_shader *sh) -{ - *ptr = sh; -} - -struct gl_shader * -_mesa_new_shader(struct gl_context *ctx, GLuint name, GLenum type) -{ - struct gl_shader *shader; - - (void) ctx; - - assert(type == GL_FRAGMENT_SHADER || type == GL_VERTEX_SHADER); - shader = rzalloc(NULL, struct gl_shader); - if (shader) { - shader->Type = type; - shader->Name = name; - shader->RefCount = 1; - } - return shader; -} - -static void -initialize_context(struct gl_context *ctx, gl_api api) -{ - memset(ctx, 0, sizeof(*ctx)); - - ctx->API = api; - - ctx->Extensions.ARB_ES2_compatibility = GL_TRUE; - ctx->Extensions.ARB_draw_buffers = GL_TRUE; - ctx->Extensions.ARB_draw_instanced = GL_TRUE; - ctx->Extensions.ARB_fragment_coord_conventions = GL_TRUE; - ctx->Extensions.EXT_texture_array = GL_TRUE; - ctx->Extensions.NV_texture_rectangle = GL_TRUE; - - /* GLSL 1.30 isn't fully supported, but we need to advertise 1.30 so that - * the built-in functions for 1.30 can be built. - */ - ctx->Const.GLSLVersion = 130; - - /* 1.10 minimums. */ - ctx->Const.MaxLights = 8; - ctx->Const.MaxClipPlanes = 8; - ctx->Const.MaxTextureUnits = 2; - - /* More than the 1.10 minimum to appease parser tests taken from - * apps that (hopefully) already checked the number of coords. - */ - ctx->Const.MaxTextureCoordUnits = 4; - - ctx->Const.VertexProgram.MaxAttribs = 16; - ctx->Const.VertexProgram.MaxUniformComponents = 512; - ctx->Const.MaxVarying = 8; - ctx->Const.MaxVertexTextureImageUnits = 0; - ctx->Const.MaxCombinedTextureImageUnits = 2; - ctx->Const.MaxTextureImageUnits = 2; - ctx->Const.FragmentProgram.MaxUniformComponents = 64; - - ctx->Const.MaxDrawBuffers = 2; - - ctx->Driver.NewShader = _mesa_new_shader; -} - -/* Returned string will have 'ctx' as its ralloc owner. */ -static char * -load_text_file(void *ctx, const char *file_name) -{ - char *text = NULL; - size_t size; - size_t total_read = 0; - FILE *fp = fopen(file_name, "rb"); - - if (!fp) { - return NULL; - } - - fseek(fp, 0L, SEEK_END); - size = ftell(fp); - fseek(fp, 0L, SEEK_SET); - - text = (char *) ralloc_size(ctx, size + 1); - if (text != NULL) { - do { - size_t bytes = fread(text + total_read, - 1, size - total_read, fp); - if (bytes < size - total_read) { - free(text); - text = NULL; - break; - } - - if (bytes == 0) { - break; - } - - total_read += bytes; - } while (total_read < size); - - text[total_read] = '\0'; - } - - fclose(fp); - - return text; -} - -int glsl_es = 0; -int dump_ast = 0; -int dump_hir = 0; -int dump_lir = 0; -int do_link = 0; - -const struct option compiler_opts[] = { - { "glsl-es", 0, &glsl_es, 1 }, - { "dump-ast", 0, &dump_ast, 1 }, - { "dump-hir", 0, &dump_hir, 1 }, - { "dump-lir", 0, &dump_lir, 1 }, - { "link", 0, &do_link, 1 }, - { NULL, 0, NULL, 0 } -}; - -/** - * \brief Print proper usage and exit with failure. - */ -void -usage_fail(const char *name) -{ - - const char *header = - "usage: %s [options] <file.vert | file.geom | file.frag>\n" - "\n" - "Possible options are:\n"; - printf(header, name, name); - for (const struct option *o = compiler_opts; o->name != 0; ++o) { - printf(" --%s\n", o->name); - } - exit(EXIT_FAILURE); -} - - -void -compile_shader(struct gl_context *ctx, struct gl_shader *shader) -{ - struct _mesa_glsl_parse_state *state = - new(shader) _mesa_glsl_parse_state(ctx, shader->Type, shader); - - const char *source = shader->Source; - state->error = preprocess(state, &source, &state->info_log, - state->extensions, ctx->API) != 0; - - if (!state->error) { - _mesa_glsl_lexer_ctor(state, source); - _mesa_glsl_parse(state); - _mesa_glsl_lexer_dtor(state); - } - - if (dump_ast) { - foreach_list_const(n, &state->translation_unit) { - ast_node *ast = exec_node_data(ast_node, n, link); - ast->print(); - } - printf("\n\n"); - } - - shader->ir = new(shader) exec_list; - if (!state->error && !state->translation_unit.is_empty()) - _mesa_ast_to_hir(shader->ir, state); - - /* Print out the unoptimized IR. */ - if (!state->error && dump_hir) { - validate_ir_tree(shader->ir); - _mesa_print_ir(shader->ir, state); - } - - /* Optimization passes */ - if (!state->error && !shader->ir->is_empty()) { - bool progress; - do { - progress = do_common_optimization(shader->ir, false, 32); - } while (progress); - - validate_ir_tree(shader->ir); - } - - - /* Print out the resulting IR */ - if (!state->error && dump_lir) { - _mesa_print_ir(shader->ir, state); - } - - shader->symbols = state->symbols; - shader->CompileStatus = !state->error; - shader->Version = state->language_version; - memcpy(shader->builtins_to_link, state->builtins_to_link, - sizeof(shader->builtins_to_link[0]) * state->num_builtins_to_link); - shader->num_builtins_to_link = state->num_builtins_to_link; - - if (shader->InfoLog) - ralloc_free(shader->InfoLog); - - shader->InfoLog = state->info_log; - - /* Retain any live IR, but trash the rest. */ - reparent_ir(shader->ir, shader); - - ralloc_free(state); - - return; -} - -int -main(int argc, char **argv) -{ - int status = EXIT_SUCCESS; - struct gl_context local_ctx; - struct gl_context *ctx = &local_ctx; - - int c; - int idx = 0; - while ((c = getopt_long(argc, argv, "", compiler_opts, &idx)) != -1) - /* empty */ ; - - - if (argc <= optind) - usage_fail(argv[0]); - - initialize_context(ctx, (glsl_es) ? API_OPENGLES2 : API_OPENGL); - - struct gl_shader_program *whole_program; - - whole_program = rzalloc (NULL, struct gl_shader_program); - assert(whole_program != NULL); - - for (/* empty */; argc > optind; optind++) { - whole_program->Shaders = - reralloc(whole_program, whole_program->Shaders, - struct gl_shader *, whole_program->NumShaders + 1); - assert(whole_program->Shaders != NULL); - - struct gl_shader *shader = rzalloc(whole_program, gl_shader); - - whole_program->Shaders[whole_program->NumShaders] = shader; - whole_program->NumShaders++; - - const unsigned len = strlen(argv[optind]); - if (len < 6) - usage_fail(argv[0]); - - const char *const ext = & argv[optind][len - 5]; - if (strncmp(".vert", ext, 5) == 0) - shader->Type = GL_VERTEX_SHADER; - else if (strncmp(".geom", ext, 5) == 0) - shader->Type = GL_GEOMETRY_SHADER; - else if (strncmp(".frag", ext, 5) == 0) - shader->Type = GL_FRAGMENT_SHADER; - else - usage_fail(argv[0]); - - shader->Source = load_text_file(whole_program, argv[optind]); - if (shader->Source == NULL) { - printf("File \"%s\" does not exist.\n", argv[optind]); - exit(EXIT_FAILURE); - } - - compile_shader(ctx, shader); - - if (!shader->CompileStatus) { - printf("Info log for %s:\n%s\n", argv[optind], shader->InfoLog); - status = EXIT_FAILURE; - break; - } - } - - if ((status == EXIT_SUCCESS) && do_link) { - link_shaders(ctx, whole_program); - status = (whole_program->LinkStatus) ? EXIT_SUCCESS : EXIT_FAILURE; - - if (strlen(whole_program->InfoLog) > 0) - printf("Info log for linking:\n%s\n", whole_program->InfoLog); - } - - for (unsigned i = 0; i < MESA_SHADER_TYPES; i++) - ralloc_free(whole_program->_LinkedShaders[i]); - - ralloc_free(whole_program); - _mesa_glsl_release_types(); - _mesa_glsl_release_functions(); - - return status; -} +/*
+ * Copyright © 2008, 2009 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+#include <io.h>
+
+#ifdef _MSC_VER
+#define __STDC__ 1
+#include <getopt.h>
+#define open _open
+#define read _read
+#define fstat _fstat
+#define stat _stat
+#define close _close
+#define O_RDONLY _O_RDONLY
+#endif
+
+#include "ast.h"
+#include "glsl_parser_extras.h"
+#include "glsl_parser.h"
+#include "ir_optimization.h"
+#include "ir_print_visitor.h"
+#include "program.h"
+#include "loop_analysis.h"
+
+extern "C" struct gl_shader *
+_mesa_new_shader(struct gl_context *ctx, GLuint name, GLenum type);
+
+extern "C" void
+_mesa_reference_shader(struct gl_context *ctx, struct gl_shader **ptr,
+ struct gl_shader *sh);
+
+/* Copied from shader_api.c for the stand-alone compiler.
+ */
+void
+_mesa_reference_shader(struct gl_context *ctx, struct gl_shader **ptr,
+ struct gl_shader *sh)
+{
+ *ptr = sh;
+}
+
+struct gl_shader *
+_mesa_new_shader(struct gl_context *ctx, GLuint name, GLenum type)
+{
+ struct gl_shader *shader;
+
+ (void) ctx;
+
+ assert(type == GL_FRAGMENT_SHADER || type == GL_VERTEX_SHADER);
+ shader = rzalloc(NULL, struct gl_shader);
+ if (shader) {
+ shader->Type = type;
+ shader->Name = name;
+ shader->RefCount = 1;
+ }
+ return shader;
+}
+
+static void
+initialize_context(struct gl_context *ctx, gl_api api)
+{
+ memset(ctx, 0, sizeof(*ctx));
+
+ ctx->API = api;
+
+ ctx->Extensions.ARB_ES2_compatibility = GL_TRUE;
+ ctx->Extensions.ARB_draw_buffers = GL_TRUE;
+ ctx->Extensions.ARB_draw_instanced = GL_TRUE;
+ ctx->Extensions.ARB_fragment_coord_conventions = GL_TRUE;
+ ctx->Extensions.EXT_texture_array = GL_TRUE;
+ ctx->Extensions.NV_texture_rectangle = GL_TRUE;
+
+ /* GLSL 1.30 isn't fully supported, but we need to advertise 1.30 so that
+ * the built-in functions for 1.30 can be built.
+ */
+ ctx->Const.GLSLVersion = 130;
+
+ /* 1.10 minimums. */
+ ctx->Const.MaxLights = 8;
+ ctx->Const.MaxClipPlanes = 8;
+ ctx->Const.MaxTextureUnits = 2;
+
+ /* More than the 1.10 minimum to appease parser tests taken from
+ * apps that (hopefully) already checked the number of coords.
+ */
+ ctx->Const.MaxTextureCoordUnits = 4;
+
+ ctx->Const.VertexProgram.MaxAttribs = 16;
+ ctx->Const.VertexProgram.MaxUniformComponents = 512;
+ ctx->Const.MaxVarying = 8;
+ ctx->Const.MaxVertexTextureImageUnits = 0;
+ ctx->Const.MaxCombinedTextureImageUnits = 2;
+ ctx->Const.MaxTextureImageUnits = 2;
+ ctx->Const.FragmentProgram.MaxUniformComponents = 64;
+
+ ctx->Const.MaxDrawBuffers = 2;
+
+ ctx->Driver.NewShader = _mesa_new_shader;
+}
+
+/* Returned string will have 'ctx' as its ralloc owner. */
+static char *
+load_text_file(void *ctx, const char *file_name)
+{
+ char *text = NULL;
+ size_t size;
+ size_t total_read = 0;
+ FILE *fp = fopen(file_name, "rb");
+
+ if (!fp) {
+ return NULL;
+ }
+
+ fseek(fp, 0L, SEEK_END);
+ size = ftell(fp);
+ fseek(fp, 0L, SEEK_SET);
+
+ text = (char *) ralloc_size(ctx, size + 1);
+ if (text != NULL) {
+ do {
+ size_t bytes = fread(text + total_read,
+ 1, size - total_read, fp);
+ if (bytes < size - total_read) {
+ free(text);
+ text = NULL;
+ break;
+ }
+
+ if (bytes == 0) {
+ break;
+ }
+
+ total_read += bytes;
+ } while (total_read < size);
+
+ text[total_read] = '\0';
+ }
+
+ fclose(fp);
+
+ return text;
+}
+
+int glsl_es = 0;
+int dump_ast = 0;
+int dump_hir = 0;
+int dump_lir = 0;
+int do_link = 0;
+
+const struct option compiler_opts[] = {
+ { "glsl-es", 0, &glsl_es, 1 },
+ { "dump-ast", 0, &dump_ast, 1 },
+ { "dump-hir", 0, &dump_hir, 1 },
+ { "dump-lir", 0, &dump_lir, 1 },
+ { "link", 0, &do_link, 1 },
+ { NULL, 0, NULL, 0 }
+};
+
+/**
+ * \brief Print proper usage and exit with failure.
+ */
+void
+usage_fail(const char *name)
+{
+
+ const char *header =
+ "usage: %s [options] <file.vert | file.geom | file.frag>\n"
+ "\n"
+ "Possible options are:\n";
+ printf(header, name, name);
+ for (const struct option *o = compiler_opts; o->name != 0; ++o) {
+ printf(" --%s\n", o->name);
+ }
+ exit(EXIT_FAILURE);
+}
+
+
+void
+compile_shader(struct gl_context *ctx, struct gl_shader *shader)
+{
+ struct _mesa_glsl_parse_state *state =
+ new(shader) _mesa_glsl_parse_state(ctx, shader->Type, shader);
+
+ const char *source = shader->Source;
+ state->error = preprocess(state, &source, &state->info_log,
+ state->extensions, ctx->API) != 0;
+
+ if (!state->error) {
+ _mesa_glsl_lexer_ctor(state, source);
+ _mesa_glsl_parse(state);
+ _mesa_glsl_lexer_dtor(state);
+ }
+
+ if (dump_ast) {
+ foreach_list_const(n, &state->translation_unit) {
+ ast_node *ast = exec_node_data(ast_node, n, link);
+ ast->print();
+ }
+ printf("\n\n");
+ }
+
+ shader->ir = new(shader) exec_list;
+ if (!state->error && !state->translation_unit.is_empty())
+ _mesa_ast_to_hir(shader->ir, state);
+
+ /* Print out the unoptimized IR. */
+ if (!state->error && dump_hir) {
+ validate_ir_tree(shader->ir);
+ _mesa_print_ir(shader->ir, state);
+ }
+
+ /* Optimization passes */
+ if (!state->error && !shader->ir->is_empty()) {
+ bool progress;
+ do {
+ progress = do_common_optimization(shader->ir, false, 32);
+ } while (progress);
+
+ validate_ir_tree(shader->ir);
+ }
+
+
+ /* Print out the resulting IR */
+ if (!state->error && dump_lir) {
+ _mesa_print_ir(shader->ir, state);
+ }
+
+ shader->symbols = state->symbols;
+ shader->CompileStatus = !state->error;
+ shader->Version = state->language_version;
+ memcpy(shader->builtins_to_link, state->builtins_to_link,
+ sizeof(shader->builtins_to_link[0]) * state->num_builtins_to_link);
+ shader->num_builtins_to_link = state->num_builtins_to_link;
+
+ if (shader->InfoLog)
+ ralloc_free(shader->InfoLog);
+
+ shader->InfoLog = state->info_log;
+
+ /* Retain any live IR, but trash the rest. */
+ reparent_ir(shader->ir, shader);
+
+ ralloc_free(state);
+
+ return;
+}
+
+int
+main(int argc, char **argv)
+{
+ int status = EXIT_SUCCESS;
+ struct gl_context local_ctx;
+ struct gl_context *ctx = &local_ctx;
+
+ int c;
+ int idx = 0;
+ while ((c = getopt_long(argc, argv, "", compiler_opts, &idx)) != -1)
+ /* empty */ ;
+
+
+ if (argc <= optind)
+ usage_fail(argv[0]);
+
+ initialize_context(ctx, (glsl_es) ? API_OPENGLES2 : API_OPENGL);
+
+ struct gl_shader_program *whole_program;
+
+ whole_program = rzalloc (NULL, struct gl_shader_program);
+ assert(whole_program != NULL);
+
+ for (/* empty */; argc > optind; optind++) {
+ whole_program->Shaders =
+ reralloc(whole_program, whole_program->Shaders,
+ struct gl_shader *, whole_program->NumShaders + 1);
+ assert(whole_program->Shaders != NULL);
+
+ struct gl_shader *shader = rzalloc(whole_program, gl_shader);
+
+ whole_program->Shaders[whole_program->NumShaders] = shader;
+ whole_program->NumShaders++;
+
+ const unsigned len = strlen(argv[optind]);
+ if (len < 6)
+ usage_fail(argv[0]);
+
+ const char *const ext = & argv[optind][len - 5];
+ if (strncmp(".vert", ext, 5) == 0)
+ shader->Type = GL_VERTEX_SHADER;
+ else if (strncmp(".geom", ext, 5) == 0)
+ shader->Type = GL_GEOMETRY_SHADER;
+ else if (strncmp(".frag", ext, 5) == 0)
+ shader->Type = GL_FRAGMENT_SHADER;
+ else
+ usage_fail(argv[0]);
+
+ shader->Source = load_text_file(whole_program, argv[optind]);
+ if (shader->Source == NULL) {
+ printf("File \"%s\" does not exist.\n", argv[optind]);
+ exit(EXIT_FAILURE);
+ }
+
+ compile_shader(ctx, shader);
+
+ if (!shader->CompileStatus) {
+ printf("Info log for %s:\n%s\n", argv[optind], shader->InfoLog);
+ status = EXIT_FAILURE;
+ break;
+ }
+ }
+
+ if ((status == EXIT_SUCCESS) && do_link) {
+ link_shaders(ctx, whole_program);
+ status = (whole_program->LinkStatus) ? EXIT_SUCCESS : EXIT_FAILURE;
+
+ if (strlen(whole_program->InfoLog) > 0)
+ printf("Info log for linking:\n%s\n", whole_program->InfoLog);
+ }
+
+ for (unsigned i = 0; i < MESA_SHADER_TYPES; i++)
+ ralloc_free(whole_program->_LinkedShaders[i]);
+
+ ralloc_free(whole_program);
+ _mesa_glsl_release_types();
+ _mesa_glsl_release_functions();
+
+ return status;
+}
diff --git a/mesalib/src/glsl/opt_constant_propagation.cpp b/mesalib/src/glsl/opt_constant_propagation.cpp index 814f6e61f..55c271200 100644 --- a/mesalib/src/glsl/opt_constant_propagation.cpp +++ b/mesalib/src/glsl/opt_constant_propagation.cpp @@ -1,436 +1,434 @@ -/* - * Copyright © 2010 Intel Corporation - * - * Permission is hereby granted, free of charge, to any person obtaining a - * constant of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, constant, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above constantright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE AUTHORS OR CONSTANTRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING - * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER - * DEALINGS IN THE SOFTWARE. - */ - -/** - * \file opt_constant_propagation.cpp - * - * Tracks assignments of constants to channels of variables, and - * usage of those constant channels with direct usage of the constants. - * - * This can lead to constant folding and algebraic optimizations in - * those later expressions, while causing no increase in instruction - * count (due to constants being generally free to load from a - * constant push buffer or as instruction immediate values) and - * possibly reducing register pressure. - */ - -#include "ir.h" -#include "ir_visitor.h" -#include "ir_rvalue_visitor.h" -#include "ir_basic_block.h" -#include "ir_optimization.h" -#include "glsl_types.h" - -using std::memset; - -class acp_entry : public exec_node -{ -public: - acp_entry(ir_variable *var, unsigned write_mask, ir_constant *constant) - { - assert(var); - assert(constant); - this->var = var; - this->write_mask = write_mask; - this->constant = constant; - } - - ir_variable *var; - ir_constant *constant; - unsigned write_mask; -}; - - -class kill_entry : public exec_node -{ -public: - kill_entry(ir_variable *var, unsigned write_mask) - { - assert(var); - this->var = var; - this->write_mask = write_mask; - } - - ir_variable *var; - unsigned write_mask; -}; - -class ir_constant_propagation_visitor : public ir_rvalue_visitor { -public: - ir_constant_propagation_visitor() - { - progress = false; - mem_ctx = ralloc_context(0); - this->acp = new(mem_ctx) exec_list; - this->kills = new(mem_ctx) exec_list; - } - ~ir_constant_propagation_visitor() - { - ralloc_free(mem_ctx); - } - - virtual ir_visitor_status visit_enter(class ir_loop *); - virtual ir_visitor_status visit_enter(class ir_function_signature *); - virtual ir_visitor_status visit_enter(class ir_function *); - virtual ir_visitor_status visit_leave(class ir_assignment *); - virtual ir_visitor_status visit_enter(class ir_call *); - virtual ir_visitor_status visit_enter(class ir_if *); - - void add_constant(ir_assignment *ir); - void kill(ir_variable *ir, unsigned write_mask); - void handle_if_block(exec_list *instructions); - void handle_rvalue(ir_rvalue **rvalue); - - /** List of acp_entry: The available constants to propagate */ - exec_list *acp; - - /** - * List of kill_entry: The masks of variables whose values were - * killed in this block. - */ - exec_list *kills; - - bool progress; - - bool killed_all; - - void *mem_ctx; -}; - - -void -ir_constant_propagation_visitor::handle_rvalue(ir_rvalue **rvalue) -{ - if (this->in_assignee || !*rvalue) - return; - - const glsl_type *type = (*rvalue)->type; - if (!type->is_scalar() && !type->is_vector()) - return; - - ir_swizzle *swiz = NULL; - ir_dereference_variable *deref = (*rvalue)->as_dereference_variable(); - if (!deref) { - swiz = (*rvalue)->as_swizzle(); - if (!swiz) - return; - - deref = swiz->val->as_dereference_variable(); - if (!deref) - return; - } - - ir_constant_data data; - memset(&data, 0, sizeof(data)); - - for (unsigned int i = 0; i < type->components(); i++) { - int channel; - acp_entry *found = NULL; - - if (swiz) { - switch (i) { - case 0: channel = swiz->mask.x; break; - case 1: channel = swiz->mask.y; break; - case 2: channel = swiz->mask.z; break; - case 3: channel = swiz->mask.w; break; - default: assert(!"shouldn't be reached"); channel = 0; break; - } - } else { - channel = i; - } - - foreach_iter(exec_list_iterator, iter, *this->acp) { - acp_entry *entry = (acp_entry *)iter.get(); - if (entry->var == deref->var && entry->write_mask & (1 << channel)) { - found = entry; - break; - } - } - - if (!found) - return; - - int rhs_channel = 0; - for (int j = 0; j < 4; j++) { - if (j == channel) - break; - if (found->write_mask & (1 << j)) - rhs_channel++; - } - - switch (type->base_type) { - case GLSL_TYPE_FLOAT: - data.f[i] = found->constant->value.f[rhs_channel]; - break; - case GLSL_TYPE_INT: - data.i[i] = found->constant->value.i[rhs_channel]; - break; - case GLSL_TYPE_UINT: - data.u[i] = found->constant->value.u[rhs_channel]; - break; - case GLSL_TYPE_BOOL: - data.b[i] = found->constant->value.b[rhs_channel]; - break; - default: - assert(!"not reached"); - break; - } - } - - *rvalue = new(ralloc_parent(deref)) ir_constant(type, &data); - this->progress = true; -} - -ir_visitor_status -ir_constant_propagation_visitor::visit_enter(ir_function_signature *ir) -{ - /* Treat entry into a function signature as a completely separate - * block. Any instructions at global scope will be shuffled into - * main() at link time, so they're irrelevant to us. - */ - exec_list *orig_acp = this->acp; - exec_list *orig_kills = this->kills; - bool orig_killed_all = this->killed_all; - - this->acp = new(mem_ctx) exec_list; - this->kills = new(mem_ctx) exec_list; - this->killed_all = false; - - visit_list_elements(this, &ir->body); - - this->kills = orig_kills; - this->acp = orig_acp; - this->killed_all = orig_killed_all; - - return visit_continue_with_parent; -} - -ir_visitor_status -ir_constant_propagation_visitor::visit_leave(ir_assignment *ir) -{ - if (this->in_assignee) - return visit_continue; - - kill(ir->lhs->variable_referenced(), ir->write_mask); - - add_constant(ir); - - return visit_continue; -} - -ir_visitor_status -ir_constant_propagation_visitor::visit_enter(ir_function *ir) -{ - (void) ir; - return visit_continue; -} - -ir_visitor_status -ir_constant_propagation_visitor::visit_enter(ir_call *ir) -{ - /* Do constant propagation on call parameters, but skip any out params */ - exec_list_iterator sig_param_iter = ir->get_callee()->parameters.iterator(); - foreach_iter(exec_list_iterator, iter, ir->actual_parameters) { - ir_variable *sig_param = (ir_variable *)sig_param_iter.get(); - ir_rvalue *param = (ir_rvalue *)iter.get(); - if (sig_param->mode != ir_var_out && sig_param->mode != ir_var_inout) { - ir_rvalue *new_param = param; - handle_rvalue(&new_param); - if (new_param != param) - param->replace_with(new_param); - else - param->accept(this); - } - sig_param_iter.next(); - } - - /* Since we're unlinked, we don't (necssarily) know the side effects of - * this call. So kill all copies. - */ - acp->make_empty(); - this->killed_all = true; - - return visit_continue_with_parent; -} - -void -ir_constant_propagation_visitor::handle_if_block(exec_list *instructions) -{ - exec_list *orig_acp = this->acp; - exec_list *orig_kills = this->kills; - bool orig_killed_all = this->killed_all; - - this->acp = new(mem_ctx) exec_list; - this->kills = new(mem_ctx) exec_list; - this->killed_all = false; - - /* Populate the initial acp with a constant of the original */ - foreach_iter(exec_list_iterator, iter, *orig_acp) { - acp_entry *a = (acp_entry *)iter.get(); - this->acp->push_tail(new(this->mem_ctx) acp_entry(a->var, a->write_mask, - a->constant)); - } - - visit_list_elements(this, instructions); - - if (this->killed_all) { - orig_acp->make_empty(); - } - - exec_list *new_kills = this->kills; - this->kills = orig_kills; - this->acp = orig_acp; - this->killed_all = this->killed_all || orig_killed_all; - - foreach_iter(exec_list_iterator, iter, *new_kills) { - kill_entry *k = (kill_entry *)iter.get(); - kill(k->var, k->write_mask); - } -} - -ir_visitor_status -ir_constant_propagation_visitor::visit_enter(ir_if *ir) -{ - ir->condition->accept(this); - handle_rvalue(&ir->condition); - - handle_if_block(&ir->then_instructions); - handle_if_block(&ir->else_instructions); - - /* handle_if_block() already descended into the children. */ - return visit_continue_with_parent; -} - -ir_visitor_status -ir_constant_propagation_visitor::visit_enter(ir_loop *ir) -{ - exec_list *orig_acp = this->acp; - exec_list *orig_kills = this->kills; - bool orig_killed_all = this->killed_all; - - /* FINISHME: For now, the initial acp for loops is totally empty. - * We could go through once, then go through again with the acp - * cloned minus the killed entries after the first run through. - */ - this->acp = new(mem_ctx) exec_list; - this->kills = new(mem_ctx) exec_list; - this->killed_all = false; - - visit_list_elements(this, &ir->body_instructions); - - if (this->killed_all) { - orig_acp->make_empty(); - } - - exec_list *new_kills = this->kills; - this->kills = orig_kills; - this->acp = orig_acp; - this->killed_all = this->killed_all || orig_killed_all; - - foreach_iter(exec_list_iterator, iter, *new_kills) { - kill_entry *k = (kill_entry *)iter.get(); - kill(k->var, k->write_mask); - } - - /* already descended into the children. */ - return visit_continue_with_parent; -} - -void -ir_constant_propagation_visitor::kill(ir_variable *var, unsigned write_mask) -{ - assert(var != NULL); - - /* We don't track non-vectors. */ - if (!var->type->is_vector() && !var->type->is_scalar()) - return; - - /* Remove any entries currently in the ACP for this kill. */ - foreach_iter(exec_list_iterator, iter, *this->acp) { - acp_entry *entry = (acp_entry *)iter.get(); - - if (entry->var == var) { - entry->write_mask &= ~write_mask; - if (entry->write_mask == 0) - entry->remove(); - } - } - - /* Add this writemask of the variable to the list of killed - * variables in this block. - */ - foreach_iter(exec_list_iterator, iter, *this->kills) { - kill_entry *entry = (kill_entry *)iter.get(); - - if (entry->var == var) { - entry->write_mask |= write_mask; - return; - } - } - /* Not already in the list. Make new entry. */ - this->kills->push_tail(new(this->mem_ctx) kill_entry(var, write_mask)); -} - -/** - * Adds an entry to the available constant list if it's a plain assignment - * of a variable to a variable. - */ -void -ir_constant_propagation_visitor::add_constant(ir_assignment *ir) -{ - acp_entry *entry; - - if (ir->condition) - return; - - if (!ir->write_mask) - return; - - ir_dereference_variable *deref = ir->lhs->as_dereference_variable(); - ir_constant *constant = ir->rhs->as_constant(); - - if (!deref || !constant) - return; - - /* Only do constant propagation on vectors. Constant matrices, - * arrays, or structures would require more work elsewhere. - */ - if (!deref->var->type->is_vector() && !deref->var->type->is_scalar()) - return; - - entry = new(this->mem_ctx) acp_entry(deref->var, ir->write_mask, constant); - this->acp->push_tail(entry); -} - -/** - * Does a constant propagation pass on the code present in the instruction stream. - */ -bool -do_constant_propagation(exec_list *instructions) -{ - ir_constant_propagation_visitor v; - - visit_list_elements(&v, instructions); - - return v.progress; -} +/*
+ * Copyright © 2010 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * constant of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, constant, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above constantright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR CONSTANTRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file opt_constant_propagation.cpp
+ *
+ * Tracks assignments of constants to channels of variables, and
+ * usage of those constant channels with direct usage of the constants.
+ *
+ * This can lead to constant folding and algebraic optimizations in
+ * those later expressions, while causing no increase in instruction
+ * count (due to constants being generally free to load from a
+ * constant push buffer or as instruction immediate values) and
+ * possibly reducing register pressure.
+ */
+
+#include "ir.h"
+#include "ir_visitor.h"
+#include "ir_rvalue_visitor.h"
+#include "ir_basic_block.h"
+#include "ir_optimization.h"
+#include "glsl_types.h"
+
+class acp_entry : public exec_node
+{
+public:
+ acp_entry(ir_variable *var, unsigned write_mask, ir_constant *constant)
+ {
+ assert(var);
+ assert(constant);
+ this->var = var;
+ this->write_mask = write_mask;
+ this->constant = constant;
+ }
+
+ ir_variable *var;
+ ir_constant *constant;
+ unsigned write_mask;
+};
+
+
+class kill_entry : public exec_node
+{
+public:
+ kill_entry(ir_variable *var, unsigned write_mask)
+ {
+ assert(var);
+ this->var = var;
+ this->write_mask = write_mask;
+ }
+
+ ir_variable *var;
+ unsigned write_mask;
+};
+
+class ir_constant_propagation_visitor : public ir_rvalue_visitor {
+public:
+ ir_constant_propagation_visitor()
+ {
+ progress = false;
+ mem_ctx = ralloc_context(0);
+ this->acp = new(mem_ctx) exec_list;
+ this->kills = new(mem_ctx) exec_list;
+ }
+ ~ir_constant_propagation_visitor()
+ {
+ ralloc_free(mem_ctx);
+ }
+
+ virtual ir_visitor_status visit_enter(class ir_loop *);
+ virtual ir_visitor_status visit_enter(class ir_function_signature *);
+ virtual ir_visitor_status visit_enter(class ir_function *);
+ virtual ir_visitor_status visit_leave(class ir_assignment *);
+ virtual ir_visitor_status visit_enter(class ir_call *);
+ virtual ir_visitor_status visit_enter(class ir_if *);
+
+ void add_constant(ir_assignment *ir);
+ void kill(ir_variable *ir, unsigned write_mask);
+ void handle_if_block(exec_list *instructions);
+ void handle_rvalue(ir_rvalue **rvalue);
+
+ /** List of acp_entry: The available constants to propagate */
+ exec_list *acp;
+
+ /**
+ * List of kill_entry: The masks of variables whose values were
+ * killed in this block.
+ */
+ exec_list *kills;
+
+ bool progress;
+
+ bool killed_all;
+
+ void *mem_ctx;
+};
+
+
+void
+ir_constant_propagation_visitor::handle_rvalue(ir_rvalue **rvalue)
+{
+ if (this->in_assignee || !*rvalue)
+ return;
+
+ const glsl_type *type = (*rvalue)->type;
+ if (!type->is_scalar() && !type->is_vector())
+ return;
+
+ ir_swizzle *swiz = NULL;
+ ir_dereference_variable *deref = (*rvalue)->as_dereference_variable();
+ if (!deref) {
+ swiz = (*rvalue)->as_swizzle();
+ if (!swiz)
+ return;
+
+ deref = swiz->val->as_dereference_variable();
+ if (!deref)
+ return;
+ }
+
+ ir_constant_data data;
+ memset(&data, 0, sizeof(data));
+
+ for (unsigned int i = 0; i < type->components(); i++) {
+ int channel;
+ acp_entry *found = NULL;
+
+ if (swiz) {
+ switch (i) {
+ case 0: channel = swiz->mask.x; break;
+ case 1: channel = swiz->mask.y; break;
+ case 2: channel = swiz->mask.z; break;
+ case 3: channel = swiz->mask.w; break;
+ default: assert(!"shouldn't be reached"); channel = 0; break;
+ }
+ } else {
+ channel = i;
+ }
+
+ foreach_iter(exec_list_iterator, iter, *this->acp) {
+ acp_entry *entry = (acp_entry *)iter.get();
+ if (entry->var == deref->var && entry->write_mask & (1 << channel)) {
+ found = entry;
+ break;
+ }
+ }
+
+ if (!found)
+ return;
+
+ int rhs_channel = 0;
+ for (int j = 0; j < 4; j++) {
+ if (j == channel)
+ break;
+ if (found->write_mask & (1 << j))
+ rhs_channel++;
+ }
+
+ switch (type->base_type) {
+ case GLSL_TYPE_FLOAT:
+ data.f[i] = found->constant->value.f[rhs_channel];
+ break;
+ case GLSL_TYPE_INT:
+ data.i[i] = found->constant->value.i[rhs_channel];
+ break;
+ case GLSL_TYPE_UINT:
+ data.u[i] = found->constant->value.u[rhs_channel];
+ break;
+ case GLSL_TYPE_BOOL:
+ data.b[i] = found->constant->value.b[rhs_channel];
+ break;
+ default:
+ assert(!"not reached");
+ break;
+ }
+ }
+
+ *rvalue = new(ralloc_parent(deref)) ir_constant(type, &data);
+ this->progress = true;
+}
+
+ir_visitor_status
+ir_constant_propagation_visitor::visit_enter(ir_function_signature *ir)
+{
+ /* Treat entry into a function signature as a completely separate
+ * block. Any instructions at global scope will be shuffled into
+ * main() at link time, so they're irrelevant to us.
+ */
+ exec_list *orig_acp = this->acp;
+ exec_list *orig_kills = this->kills;
+ bool orig_killed_all = this->killed_all;
+
+ this->acp = new(mem_ctx) exec_list;
+ this->kills = new(mem_ctx) exec_list;
+ this->killed_all = false;
+
+ visit_list_elements(this, &ir->body);
+
+ this->kills = orig_kills;
+ this->acp = orig_acp;
+ this->killed_all = orig_killed_all;
+
+ return visit_continue_with_parent;
+}
+
+ir_visitor_status
+ir_constant_propagation_visitor::visit_leave(ir_assignment *ir)
+{
+ if (this->in_assignee)
+ return visit_continue;
+
+ kill(ir->lhs->variable_referenced(), ir->write_mask);
+
+ add_constant(ir);
+
+ return visit_continue;
+}
+
+ir_visitor_status
+ir_constant_propagation_visitor::visit_enter(ir_function *ir)
+{
+ (void) ir;
+ return visit_continue;
+}
+
+ir_visitor_status
+ir_constant_propagation_visitor::visit_enter(ir_call *ir)
+{
+ /* Do constant propagation on call parameters, but skip any out params */
+ exec_list_iterator sig_param_iter = ir->get_callee()->parameters.iterator();
+ foreach_iter(exec_list_iterator, iter, ir->actual_parameters) {
+ ir_variable *sig_param = (ir_variable *)sig_param_iter.get();
+ ir_rvalue *param = (ir_rvalue *)iter.get();
+ if (sig_param->mode != ir_var_out && sig_param->mode != ir_var_inout) {
+ ir_rvalue *new_param = param;
+ handle_rvalue(&new_param);
+ if (new_param != param)
+ param->replace_with(new_param);
+ else
+ param->accept(this);
+ }
+ sig_param_iter.next();
+ }
+
+ /* Since we're unlinked, we don't (necssarily) know the side effects of
+ * this call. So kill all copies.
+ */
+ acp->make_empty();
+ this->killed_all = true;
+
+ return visit_continue_with_parent;
+}
+
+void
+ir_constant_propagation_visitor::handle_if_block(exec_list *instructions)
+{
+ exec_list *orig_acp = this->acp;
+ exec_list *orig_kills = this->kills;
+ bool orig_killed_all = this->killed_all;
+
+ this->acp = new(mem_ctx) exec_list;
+ this->kills = new(mem_ctx) exec_list;
+ this->killed_all = false;
+
+ /* Populate the initial acp with a constant of the original */
+ foreach_iter(exec_list_iterator, iter, *orig_acp) {
+ acp_entry *a = (acp_entry *)iter.get();
+ this->acp->push_tail(new(this->mem_ctx) acp_entry(a->var, a->write_mask,
+ a->constant));
+ }
+
+ visit_list_elements(this, instructions);
+
+ if (this->killed_all) {
+ orig_acp->make_empty();
+ }
+
+ exec_list *new_kills = this->kills;
+ this->kills = orig_kills;
+ this->acp = orig_acp;
+ this->killed_all = this->killed_all || orig_killed_all;
+
+ foreach_iter(exec_list_iterator, iter, *new_kills) {
+ kill_entry *k = (kill_entry *)iter.get();
+ kill(k->var, k->write_mask);
+ }
+}
+
+ir_visitor_status
+ir_constant_propagation_visitor::visit_enter(ir_if *ir)
+{
+ ir->condition->accept(this);
+ handle_rvalue(&ir->condition);
+
+ handle_if_block(&ir->then_instructions);
+ handle_if_block(&ir->else_instructions);
+
+ /* handle_if_block() already descended into the children. */
+ return visit_continue_with_parent;
+}
+
+ir_visitor_status
+ir_constant_propagation_visitor::visit_enter(ir_loop *ir)
+{
+ exec_list *orig_acp = this->acp;
+ exec_list *orig_kills = this->kills;
+ bool orig_killed_all = this->killed_all;
+
+ /* FINISHME: For now, the initial acp for loops is totally empty.
+ * We could go through once, then go through again with the acp
+ * cloned minus the killed entries after the first run through.
+ */
+ this->acp = new(mem_ctx) exec_list;
+ this->kills = new(mem_ctx) exec_list;
+ this->killed_all = false;
+
+ visit_list_elements(this, &ir->body_instructions);
+
+ if (this->killed_all) {
+ orig_acp->make_empty();
+ }
+
+ exec_list *new_kills = this->kills;
+ this->kills = orig_kills;
+ this->acp = orig_acp;
+ this->killed_all = this->killed_all || orig_killed_all;
+
+ foreach_iter(exec_list_iterator, iter, *new_kills) {
+ kill_entry *k = (kill_entry *)iter.get();
+ kill(k->var, k->write_mask);
+ }
+
+ /* already descended into the children. */
+ return visit_continue_with_parent;
+}
+
+void
+ir_constant_propagation_visitor::kill(ir_variable *var, unsigned write_mask)
+{
+ assert(var != NULL);
+
+ /* We don't track non-vectors. */
+ if (!var->type->is_vector() && !var->type->is_scalar())
+ return;
+
+ /* Remove any entries currently in the ACP for this kill. */
+ foreach_iter(exec_list_iterator, iter, *this->acp) {
+ acp_entry *entry = (acp_entry *)iter.get();
+
+ if (entry->var == var) {
+ entry->write_mask &= ~write_mask;
+ if (entry->write_mask == 0)
+ entry->remove();
+ }
+ }
+
+ /* Add this writemask of the variable to the list of killed
+ * variables in this block.
+ */
+ foreach_iter(exec_list_iterator, iter, *this->kills) {
+ kill_entry *entry = (kill_entry *)iter.get();
+
+ if (entry->var == var) {
+ entry->write_mask |= write_mask;
+ return;
+ }
+ }
+ /* Not already in the list. Make new entry. */
+ this->kills->push_tail(new(this->mem_ctx) kill_entry(var, write_mask));
+}
+
+/**
+ * Adds an entry to the available constant list if it's a plain assignment
+ * of a variable to a variable.
+ */
+void
+ir_constant_propagation_visitor::add_constant(ir_assignment *ir)
+{
+ acp_entry *entry;
+
+ if (ir->condition)
+ return;
+
+ if (!ir->write_mask)
+ return;
+
+ ir_dereference_variable *deref = ir->lhs->as_dereference_variable();
+ ir_constant *constant = ir->rhs->as_constant();
+
+ if (!deref || !constant)
+ return;
+
+ /* Only do constant propagation on vectors. Constant matrices,
+ * arrays, or structures would require more work elsewhere.
+ */
+ if (!deref->var->type->is_vector() && !deref->var->type->is_scalar())
+ return;
+
+ entry = new(this->mem_ctx) acp_entry(deref->var, ir->write_mask, constant);
+ this->acp->push_tail(entry);
+}
+
+/**
+ * Does a constant propagation pass on the code present in the instruction stream.
+ */
+bool
+do_constant_propagation(exec_list *instructions)
+{
+ ir_constant_propagation_visitor v;
+
+ visit_list_elements(&v, instructions);
+
+ return v.progress;
+}
diff --git a/mesalib/src/glsl/opt_constant_variable.cpp b/mesalib/src/glsl/opt_constant_variable.cpp index 9dca0ff8e..f1aeaa99a 100644 --- a/mesalib/src/glsl/opt_constant_variable.cpp +++ b/mesalib/src/glsl/opt_constant_variable.cpp @@ -1,198 +1,195 @@ -/* - * Copyright © 2010 Intel Corporation - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING - * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER - * DEALINGS IN THE SOFTWARE. - */ - -/** - * \file opt_constant_variable.cpp - * - * Marks variables assigned a single constant value over the course - * of the program as constant. - * - * The goal here is to trigger further constant folding and then dead - * code elimination. This is common with vector/matrix constructors - * and calls to builtin functions. - */ - -#include "ir.h" -#include "ir_visitor.h" -#include "ir_optimization.h" -#include "glsl_types.h" - -using std::calloc; -using std::free; - -struct assignment_entry { - exec_node link; - int assignment_count; - ir_variable *var; - ir_constant *constval; - bool our_scope; -}; - -class ir_constant_variable_visitor : public ir_hierarchical_visitor { -public: - virtual ir_visitor_status visit_enter(ir_dereference_variable *); - virtual ir_visitor_status visit(ir_variable *); - virtual ir_visitor_status visit_enter(ir_assignment *); - virtual ir_visitor_status visit_enter(ir_call *); - - exec_list list; -}; - -static struct assignment_entry * -get_assignment_entry(ir_variable *var, exec_list *list) -{ - struct assignment_entry *entry; - - foreach_list_typed(struct assignment_entry, entry, link, list) { - if (entry->var == var) - return entry; - } - - entry = (struct assignment_entry *)calloc(1, sizeof(*entry)); - entry->var = var; - list->push_head(&entry->link); - return entry; -} - -ir_visitor_status -ir_constant_variable_visitor::visit(ir_variable *ir) -{ - struct assignment_entry *entry = get_assignment_entry(ir, &this->list); - entry->our_scope = true; - return visit_continue; -} - -/* Skip derefs of variables so that we can detect declarations. */ -ir_visitor_status -ir_constant_variable_visitor::visit_enter(ir_dereference_variable *ir) -{ - (void)ir; - return visit_continue_with_parent; -} - -ir_visitor_status -ir_constant_variable_visitor::visit_enter(ir_assignment *ir) -{ - ir_constant *constval; - struct assignment_entry *entry; - - entry = get_assignment_entry(ir->lhs->variable_referenced(), &this->list); - assert(entry); - entry->assignment_count++; - - /* If it's already constant, don't do the work. */ - if (entry->var->constant_value) - return visit_continue; - - /* OK, now find if we actually have all the right conditions for - * this to be a constant value assigned to the var. - */ - if (ir->condition) - return visit_continue; - - ir_variable *var = ir->whole_variable_written(); - if (!var) - return visit_continue; - - constval = ir->rhs->constant_expression_value(); - if (!constval) - return visit_continue; - - /* Mark this entry as having a constant assignment (if the - * assignment count doesn't go >1). do_constant_variable will fix - * up the variable with the constant value later. - */ - entry->constval = constval; - - return visit_continue; -} - -ir_visitor_status -ir_constant_variable_visitor::visit_enter(ir_call *ir) -{ - exec_list_iterator sig_iter = ir->get_callee()->parameters.iterator(); - foreach_iter(exec_list_iterator, iter, *ir) { - ir_rvalue *param_rval = (ir_rvalue *)iter.get(); - ir_variable *param = (ir_variable *)sig_iter.get(); - - if (param->mode == ir_var_out || - param->mode == ir_var_inout) { - ir_variable *var = param_rval->variable_referenced(); - struct assignment_entry *entry; - - assert(var); - entry = get_assignment_entry(var, &this->list); - entry->assignment_count++; - } - sig_iter.next(); - } - return visit_continue; -} - -/** - * Does a copy propagation pass on the code present in the instruction stream. - */ -bool -do_constant_variable(exec_list *instructions) -{ - bool progress = false; - ir_constant_variable_visitor v; - - v.run(instructions); - - while (!v.list.is_empty()) { - - struct assignment_entry *entry; - entry = exec_node_data(struct assignment_entry, v.list.head, link); - - if (entry->assignment_count == 1 && entry->constval && entry->our_scope) { - entry->var->constant_value = entry->constval; - progress = true; - } - entry->link.remove(); - free(entry); - } - - return progress; -} - -bool -do_constant_variable_unlinked(exec_list *instructions) -{ - bool progress = false; - - foreach_iter(exec_list_iterator, iter, *instructions) { - ir_instruction *ir = (ir_instruction *)iter.get(); - ir_function *f = ir->as_function(); - if (f) { - foreach_iter(exec_list_iterator, sigiter, *f) { - ir_function_signature *sig = - (ir_function_signature *) sigiter.get(); - if (do_constant_variable(&sig->body)) - progress = true; - } - } - } - - return progress; -} +/*
+ * Copyright © 2010 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file opt_constant_variable.cpp
+ *
+ * Marks variables assigned a single constant value over the course
+ * of the program as constant.
+ *
+ * The goal here is to trigger further constant folding and then dead
+ * code elimination. This is common with vector/matrix constructors
+ * and calls to builtin functions.
+ */
+
+#include "ir.h"
+#include "ir_visitor.h"
+#include "ir_optimization.h"
+#include "glsl_types.h"
+
+struct assignment_entry {
+ exec_node link;
+ int assignment_count;
+ ir_variable *var;
+ ir_constant *constval;
+ bool our_scope;
+};
+
+class ir_constant_variable_visitor : public ir_hierarchical_visitor {
+public:
+ virtual ir_visitor_status visit_enter(ir_dereference_variable *);
+ virtual ir_visitor_status visit(ir_variable *);
+ virtual ir_visitor_status visit_enter(ir_assignment *);
+ virtual ir_visitor_status visit_enter(ir_call *);
+
+ exec_list list;
+};
+
+static struct assignment_entry *
+get_assignment_entry(ir_variable *var, exec_list *list)
+{
+ struct assignment_entry *entry;
+
+ foreach_list_typed(struct assignment_entry, entry, link, list) {
+ if (entry->var == var)
+ return entry;
+ }
+
+ entry = (struct assignment_entry *)calloc(1, sizeof(*entry));
+ entry->var = var;
+ list->push_head(&entry->link);
+ return entry;
+}
+
+ir_visitor_status
+ir_constant_variable_visitor::visit(ir_variable *ir)
+{
+ struct assignment_entry *entry = get_assignment_entry(ir, &this->list);
+ entry->our_scope = true;
+ return visit_continue;
+}
+
+/* Skip derefs of variables so that we can detect declarations. */
+ir_visitor_status
+ir_constant_variable_visitor::visit_enter(ir_dereference_variable *ir)
+{
+ (void)ir;
+ return visit_continue_with_parent;
+}
+
+ir_visitor_status
+ir_constant_variable_visitor::visit_enter(ir_assignment *ir)
+{
+ ir_constant *constval;
+ struct assignment_entry *entry;
+
+ entry = get_assignment_entry(ir->lhs->variable_referenced(), &this->list);
+ assert(entry);
+ entry->assignment_count++;
+
+ /* If it's already constant, don't do the work. */
+ if (entry->var->constant_value)
+ return visit_continue;
+
+ /* OK, now find if we actually have all the right conditions for
+ * this to be a constant value assigned to the var.
+ */
+ if (ir->condition)
+ return visit_continue;
+
+ ir_variable *var = ir->whole_variable_written();
+ if (!var)
+ return visit_continue;
+
+ constval = ir->rhs->constant_expression_value();
+ if (!constval)
+ return visit_continue;
+
+ /* Mark this entry as having a constant assignment (if the
+ * assignment count doesn't go >1). do_constant_variable will fix
+ * up the variable with the constant value later.
+ */
+ entry->constval = constval;
+
+ return visit_continue;
+}
+
+ir_visitor_status
+ir_constant_variable_visitor::visit_enter(ir_call *ir)
+{
+ exec_list_iterator sig_iter = ir->get_callee()->parameters.iterator();
+ foreach_iter(exec_list_iterator, iter, *ir) {
+ ir_rvalue *param_rval = (ir_rvalue *)iter.get();
+ ir_variable *param = (ir_variable *)sig_iter.get();
+
+ if (param->mode == ir_var_out ||
+ param->mode == ir_var_inout) {
+ ir_variable *var = param_rval->variable_referenced();
+ struct assignment_entry *entry;
+
+ assert(var);
+ entry = get_assignment_entry(var, &this->list);
+ entry->assignment_count++;
+ }
+ sig_iter.next();
+ }
+ return visit_continue;
+}
+
+/**
+ * Does a copy propagation pass on the code present in the instruction stream.
+ */
+bool
+do_constant_variable(exec_list *instructions)
+{
+ bool progress = false;
+ ir_constant_variable_visitor v;
+
+ v.run(instructions);
+
+ while (!v.list.is_empty()) {
+
+ struct assignment_entry *entry;
+ entry = exec_node_data(struct assignment_entry, v.list.head, link);
+
+ if (entry->assignment_count == 1 && entry->constval && entry->our_scope) {
+ entry->var->constant_value = entry->constval;
+ progress = true;
+ }
+ entry->link.remove();
+ free(entry);
+ }
+
+ return progress;
+}
+
+bool
+do_constant_variable_unlinked(exec_list *instructions)
+{
+ bool progress = false;
+
+ foreach_iter(exec_list_iterator, iter, *instructions) {
+ ir_instruction *ir = (ir_instruction *)iter.get();
+ ir_function *f = ir->as_function();
+ if (f) {
+ foreach_iter(exec_list_iterator, sigiter, *f) {
+ ir_function_signature *sig =
+ (ir_function_signature *) sigiter.get();
+ if (do_constant_variable(&sig->body))
+ progress = true;
+ }
+ }
+ }
+
+ return progress;
+}
diff --git a/mesalib/src/glsl/opt_dead_code.cpp b/mesalib/src/glsl/opt_dead_code.cpp index f47b9613e..492ba73a1 100644 --- a/mesalib/src/glsl/opt_dead_code.cpp +++ b/mesalib/src/glsl/opt_dead_code.cpp @@ -1,144 +1,142 @@ -/* - * Copyright © 2010 Intel Corporation - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING - * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER - * DEALINGS IN THE SOFTWARE. - */ - -/** - * \file opt_dead_code.cpp - * - * Eliminates dead assignments and variable declarations from the code. - */ - -#include "ir.h" -#include "ir_visitor.h" -#include "ir_variable_refcount.h" -#include "glsl_types.h" - -using std::printf; - -static bool debug = false; - -/** - * Do a dead code pass over instructions and everything that instructions - * references. - * - * Note that this will remove assignments to globals, so it is not suitable - * for usage on an unlinked instruction stream. - */ -bool -do_dead_code(exec_list *instructions) -{ - ir_variable_refcount_visitor v; - bool progress = false; - - v.run(instructions); - - foreach_iter(exec_list_iterator, iter, v.variable_list) { - variable_entry *entry = (variable_entry *)iter.get(); - - /* Since each assignment is a reference, the refereneced count must be - * greater than or equal to the assignment count. If they are equal, - * then all of the references are assignments, and the variable is - * dead. - * - * Note that if the variable is neither assigned nor referenced, both - * counts will be zero and will be caught by the equality test. - */ - assert(entry->referenced_count >= entry->assigned_count); - - if (debug) { - printf("%s@%p: %d refs, %d assigns, %sdeclared in our scope\n", - entry->var->name, (void *) entry->var, - entry->referenced_count, entry->assigned_count, - entry->declaration ? "" : "not "); - } - - if ((entry->referenced_count > entry->assigned_count) - || !entry->declaration) - continue; - - if (entry->assign) { - /* Remove a single dead assignment to the variable we found. - * Don't do so if it's a shader output, though. - */ - if (entry->var->mode != ir_var_out && - entry->var->mode != ir_var_inout && - !ir_has_call(entry->assign)) { - entry->assign->remove(); - progress = true; - - if (debug) { - printf("Removed assignment to %s@%p\n", - entry->var->name, (void *) entry->var); - } - } - } else { - /* If there are no assignments or references to the variable left, - * then we can remove its declaration. - */ - - /* uniform initializers are precious, and could get used by another - * stage. - */ - if (entry->var->mode == ir_var_uniform && - entry->var->constant_value) - continue; - - entry->var->remove(); - progress = true; - - if (debug) { - printf("Removed declaration of %s@%p\n", - entry->var->name, (void *) entry->var); - } - } - } - - return progress; -} - -/** - * Does a dead code pass on the functions present in the instruction stream. - * - * This is suitable for use while the program is not linked, as it will - * ignore variable declarations (and the assignments to them) for variables - * with global scope. - */ -bool -do_dead_code_unlinked(exec_list *instructions) -{ - bool progress = false; - - foreach_iter(exec_list_iterator, iter, *instructions) { - ir_instruction *ir = (ir_instruction *)iter.get(); - ir_function *f = ir->as_function(); - if (f) { - foreach_iter(exec_list_iterator, sigiter, *f) { - ir_function_signature *sig = - (ir_function_signature *) sigiter.get(); - if (do_dead_code(&sig->body)) - progress = true; - } - } - } - - return progress; -} +/*
+ * Copyright © 2010 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file opt_dead_code.cpp
+ *
+ * Eliminates dead assignments and variable declarations from the code.
+ */
+
+#include "ir.h"
+#include "ir_visitor.h"
+#include "ir_variable_refcount.h"
+#include "glsl_types.h"
+
+static bool debug = false;
+
+/**
+ * Do a dead code pass over instructions and everything that instructions
+ * references.
+ *
+ * Note that this will remove assignments to globals, so it is not suitable
+ * for usage on an unlinked instruction stream.
+ */
+bool
+do_dead_code(exec_list *instructions)
+{
+ ir_variable_refcount_visitor v;
+ bool progress = false;
+
+ v.run(instructions);
+
+ foreach_iter(exec_list_iterator, iter, v.variable_list) {
+ variable_entry *entry = (variable_entry *)iter.get();
+
+ /* Since each assignment is a reference, the refereneced count must be
+ * greater than or equal to the assignment count. If they are equal,
+ * then all of the references are assignments, and the variable is
+ * dead.
+ *
+ * Note that if the variable is neither assigned nor referenced, both
+ * counts will be zero and will be caught by the equality test.
+ */
+ assert(entry->referenced_count >= entry->assigned_count);
+
+ if (debug) {
+ printf("%s@%p: %d refs, %d assigns, %sdeclared in our scope\n",
+ entry->var->name, (void *) entry->var,
+ entry->referenced_count, entry->assigned_count,
+ entry->declaration ? "" : "not ");
+ }
+
+ if ((entry->referenced_count > entry->assigned_count)
+ || !entry->declaration)
+ continue;
+
+ if (entry->assign) {
+ /* Remove a single dead assignment to the variable we found.
+ * Don't do so if it's a shader output, though.
+ */
+ if (entry->var->mode != ir_var_out &&
+ entry->var->mode != ir_var_inout &&
+ !ir_has_call(entry->assign)) {
+ entry->assign->remove();
+ progress = true;
+
+ if (debug) {
+ printf("Removed assignment to %s@%p\n",
+ entry->var->name, (void *) entry->var);
+ }
+ }
+ } else {
+ /* If there are no assignments or references to the variable left,
+ * then we can remove its declaration.
+ */
+
+ /* uniform initializers are precious, and could get used by another
+ * stage.
+ */
+ if (entry->var->mode == ir_var_uniform &&
+ entry->var->constant_value)
+ continue;
+
+ entry->var->remove();
+ progress = true;
+
+ if (debug) {
+ printf("Removed declaration of %s@%p\n",
+ entry->var->name, (void *) entry->var);
+ }
+ }
+ }
+
+ return progress;
+}
+
+/**
+ * Does a dead code pass on the functions present in the instruction stream.
+ *
+ * This is suitable for use while the program is not linked, as it will
+ * ignore variable declarations (and the assignments to them) for variables
+ * with global scope.
+ */
+bool
+do_dead_code_unlinked(exec_list *instructions)
+{
+ bool progress = false;
+
+ foreach_iter(exec_list_iterator, iter, *instructions) {
+ ir_instruction *ir = (ir_instruction *)iter.get();
+ ir_function *f = ir->as_function();
+ if (f) {
+ foreach_iter(exec_list_iterator, sigiter, *f) {
+ ir_function_signature *sig =
+ (ir_function_signature *) sigiter.get();
+ if (do_dead_code(&sig->body))
+ progress = true;
+ }
+ }
+ }
+
+ return progress;
+}
diff --git a/mesalib/src/glsl/opt_dead_code_local.cpp b/mesalib/src/glsl/opt_dead_code_local.cpp index f67b46337..42e5e94c8 100644 --- a/mesalib/src/glsl/opt_dead_code_local.cpp +++ b/mesalib/src/glsl/opt_dead_code_local.cpp @@ -1,224 +1,222 @@ -/* - * Copyright © 2010 Intel Corporation - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING - * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER - * DEALINGS IN THE SOFTWARE. - */ - -/** - * \file opt_dead_code_local.cpp - * - * Eliminates local dead assignments from the code. - * - * This operates on basic blocks, tracking assignments and finding if - * they're used before the variable is completely reassigned. - * - * Compare this to ir_dead_code.cpp, which operates globally looking - * for assignments to variables that are never read. - */ - -#include "ir.h" -#include "ir_basic_block.h" -#include "ir_optimization.h" -#include "glsl_types.h" - -using std::printf; - -static bool debug = false; - -class assignment_entry : public exec_node -{ -public: - assignment_entry(ir_variable *lhs, ir_instruction *ir) - { - assert(lhs); - assert(ir); - this->lhs = lhs; - this->ir = ir; - } - - ir_variable *lhs; - ir_instruction *ir; -}; - -class kill_for_derefs_visitor : public ir_hierarchical_visitor { -public: - kill_for_derefs_visitor(exec_list *assignments) - { - this->assignments = assignments; - } - - virtual ir_visitor_status visit(ir_dereference_variable *ir) - { - ir_variable *const var = ir->variable_referenced(); - - foreach_iter(exec_list_iterator, iter, *this->assignments) { - assignment_entry *entry = (assignment_entry *)iter.get(); - - if (entry->lhs == var) { - if (debug) - printf("kill %s\n", entry->lhs->name); - entry->remove(); - } - } - - return visit_continue; - } - -private: - exec_list *assignments; -}; - -class array_index_visit : public ir_hierarchical_visitor { -public: - array_index_visit(ir_hierarchical_visitor *v) - { - this->visitor = v; - } - - virtual ir_visitor_status visit_enter(class ir_dereference_array *ir) - { - ir->array_index->accept(visitor); - return visit_continue; - } - - static void run(ir_instruction *ir, ir_hierarchical_visitor *v) - { - array_index_visit top_visit(v); - ir->accept(& top_visit); - } - - ir_hierarchical_visitor *visitor; -}; - - -/** - * Adds an entry to the available copy list if it's a plain assignment - * of a variable to a variable. - */ -static bool -process_assignment(void *ctx, ir_assignment *ir, exec_list *assignments) -{ - ir_variable *var = NULL; - bool progress = false; - kill_for_derefs_visitor v(assignments); - - /* Kill assignment entries for things used to produce this assignment. */ - ir->rhs->accept(&v); - if (ir->condition) { - ir->condition->accept(&v); - } - - /* Kill assignment enties used as array indices. - */ - array_index_visit::run(ir->lhs, &v); - var = ir->lhs->variable_referenced(); - assert(var); - - /* Now, check if we did a whole-variable assignment. */ - if (!ir->condition && (ir->whole_variable_written() != NULL)) { - /* We did a whole-variable assignment. So, any instruction in - * the assignment list with the same LHS is dead. - */ - if (debug) - printf("looking for %s to remove\n", var->name); - foreach_iter(exec_list_iterator, iter, *assignments) { - assignment_entry *entry = (assignment_entry *)iter.get(); - - if (entry->lhs == var) { - if (debug) - printf("removing %s\n", var->name); - entry->ir->remove(); - entry->remove(); - progress = true; - } - } - } - - /* Add this instruction to the assignment list available to be removed. - * But not if the assignment has other side effects. - */ - if (ir_has_call(ir)) - return progress; - - assignment_entry *entry = new(ctx) assignment_entry(var, ir); - assignments->push_tail(entry); - - if (debug) { - printf("add %s\n", var->name); - - printf("current entries\n"); - foreach_iter(exec_list_iterator, iter, *assignments) { - assignment_entry *entry = (assignment_entry *)iter.get(); - - printf(" %s\n", entry->lhs->name); - } - } - - return progress; -} - -static void -dead_code_local_basic_block(ir_instruction *first, - ir_instruction *last, - void *data) -{ - ir_instruction *ir, *ir_next; - /* List of avaialble_copy */ - exec_list assignments; - bool *out_progress = (bool *)data; - bool progress = false; - - void *ctx = ralloc_context(NULL); - /* Safe looping, since process_assignment */ - for (ir = first, ir_next = (ir_instruction *)first->next;; - ir = ir_next, ir_next = (ir_instruction *)ir->next) { - ir_assignment *ir_assign = ir->as_assignment(); - - if (debug) { - ir->print(); - printf("\n"); - } - - if (ir_assign) { - progress = process_assignment(ctx, ir_assign, &assignments) || progress; - } else { - kill_for_derefs_visitor kill(&assignments); - ir->accept(&kill); - } - - if (ir == last) - break; - } - *out_progress = progress; - ralloc_free(ctx); -} - -/** - * Does a copy propagation pass on the code present in the instruction stream. - */ -bool -do_dead_code_local(exec_list *instructions) -{ - bool progress = false; - - call_for_basic_blocks(instructions, dead_code_local_basic_block, &progress); - - return progress; -} +/*
+ * Copyright © 2010 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file opt_dead_code_local.cpp
+ *
+ * Eliminates local dead assignments from the code.
+ *
+ * This operates on basic blocks, tracking assignments and finding if
+ * they're used before the variable is completely reassigned.
+ *
+ * Compare this to ir_dead_code.cpp, which operates globally looking
+ * for assignments to variables that are never read.
+ */
+
+#include "ir.h"
+#include "ir_basic_block.h"
+#include "ir_optimization.h"
+#include "glsl_types.h"
+
+static bool debug = false;
+
+class assignment_entry : public exec_node
+{
+public:
+ assignment_entry(ir_variable *lhs, ir_instruction *ir)
+ {
+ assert(lhs);
+ assert(ir);
+ this->lhs = lhs;
+ this->ir = ir;
+ }
+
+ ir_variable *lhs;
+ ir_instruction *ir;
+};
+
+class kill_for_derefs_visitor : public ir_hierarchical_visitor {
+public:
+ kill_for_derefs_visitor(exec_list *assignments)
+ {
+ this->assignments = assignments;
+ }
+
+ virtual ir_visitor_status visit(ir_dereference_variable *ir)
+ {
+ ir_variable *const var = ir->variable_referenced();
+
+ foreach_iter(exec_list_iterator, iter, *this->assignments) {
+ assignment_entry *entry = (assignment_entry *)iter.get();
+
+ if (entry->lhs == var) {
+ if (debug)
+ printf("kill %s\n", entry->lhs->name);
+ entry->remove();
+ }
+ }
+
+ return visit_continue;
+ }
+
+private:
+ exec_list *assignments;
+};
+
+class array_index_visit : public ir_hierarchical_visitor {
+public:
+ array_index_visit(ir_hierarchical_visitor *v)
+ {
+ this->visitor = v;
+ }
+
+ virtual ir_visitor_status visit_enter(class ir_dereference_array *ir)
+ {
+ ir->array_index->accept(visitor);
+ return visit_continue;
+ }
+
+ static void run(ir_instruction *ir, ir_hierarchical_visitor *v)
+ {
+ array_index_visit top_visit(v);
+ ir->accept(& top_visit);
+ }
+
+ ir_hierarchical_visitor *visitor;
+};
+
+
+/**
+ * Adds an entry to the available copy list if it's a plain assignment
+ * of a variable to a variable.
+ */
+static bool
+process_assignment(void *ctx, ir_assignment *ir, exec_list *assignments)
+{
+ ir_variable *var = NULL;
+ bool progress = false;
+ kill_for_derefs_visitor v(assignments);
+
+ /* Kill assignment entries for things used to produce this assignment. */
+ ir->rhs->accept(&v);
+ if (ir->condition) {
+ ir->condition->accept(&v);
+ }
+
+ /* Kill assignment enties used as array indices.
+ */
+ array_index_visit::run(ir->lhs, &v);
+ var = ir->lhs->variable_referenced();
+ assert(var);
+
+ /* Now, check if we did a whole-variable assignment. */
+ if (!ir->condition && (ir->whole_variable_written() != NULL)) {
+ /* We did a whole-variable assignment. So, any instruction in
+ * the assignment list with the same LHS is dead.
+ */
+ if (debug)
+ printf("looking for %s to remove\n", var->name);
+ foreach_iter(exec_list_iterator, iter, *assignments) {
+ assignment_entry *entry = (assignment_entry *)iter.get();
+
+ if (entry->lhs == var) {
+ if (debug)
+ printf("removing %s\n", var->name);
+ entry->ir->remove();
+ entry->remove();
+ progress = true;
+ }
+ }
+ }
+
+ /* Add this instruction to the assignment list available to be removed.
+ * But not if the assignment has other side effects.
+ */
+ if (ir_has_call(ir))
+ return progress;
+
+ assignment_entry *entry = new(ctx) assignment_entry(var, ir);
+ assignments->push_tail(entry);
+
+ if (debug) {
+ printf("add %s\n", var->name);
+
+ printf("current entries\n");
+ foreach_iter(exec_list_iterator, iter, *assignments) {
+ assignment_entry *entry = (assignment_entry *)iter.get();
+
+ printf(" %s\n", entry->lhs->name);
+ }
+ }
+
+ return progress;
+}
+
+static void
+dead_code_local_basic_block(ir_instruction *first,
+ ir_instruction *last,
+ void *data)
+{
+ ir_instruction *ir, *ir_next;
+ /* List of avaialble_copy */
+ exec_list assignments;
+ bool *out_progress = (bool *)data;
+ bool progress = false;
+
+ void *ctx = ralloc_context(NULL);
+ /* Safe looping, since process_assignment */
+ for (ir = first, ir_next = (ir_instruction *)first->next;;
+ ir = ir_next, ir_next = (ir_instruction *)ir->next) {
+ ir_assignment *ir_assign = ir->as_assignment();
+
+ if (debug) {
+ ir->print();
+ printf("\n");
+ }
+
+ if (ir_assign) {
+ progress = process_assignment(ctx, ir_assign, &assignments) || progress;
+ } else {
+ kill_for_derefs_visitor kill(&assignments);
+ ir->accept(&kill);
+ }
+
+ if (ir == last)
+ break;
+ }
+ *out_progress = progress;
+ ralloc_free(ctx);
+}
+
+/**
+ * Does a copy propagation pass on the code present in the instruction stream.
+ */
+bool
+do_dead_code_local(exec_list *instructions)
+{
+ bool progress = false;
+
+ call_for_basic_blocks(instructions, dead_code_local_basic_block, &progress);
+
+ return progress;
+}
diff --git a/mesalib/src/glsl/opt_dead_functions.cpp b/mesalib/src/glsl/opt_dead_functions.cpp index d72eb6115..6d41af5c5 100644 --- a/mesalib/src/glsl/opt_dead_functions.cpp +++ b/mesalib/src/glsl/opt_dead_functions.cpp @@ -1,155 +1,153 @@ - /* - * Copyright © 2010 Intel Corporation - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING - * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER - * DEALINGS IN THE SOFTWARE. - */ - - /** - * \file opt_dead_functions.cpp - * - * Eliminates unused functions from the linked program. - */ - - #include "ir.h" - #include "ir_visitor.h" - #include "ir_expression_flattening.h" - #include "glsl_types.h" - - using std::strcmp; - - class signature_entry : public exec_node - { - public: - signature_entry(ir_function_signature *sig) - { - this->signature = sig; - this->used = false; - } - - ir_function_signature *signature; - bool used; - }; - - class ir_dead_functions_visitor : public ir_hierarchical_visitor { - public: - ir_dead_functions_visitor() - { - this->mem_ctx = ralloc_context(NULL); - } - - ~ir_dead_functions_visitor() - { - ralloc_free(this->mem_ctx); - } - - virtual ir_visitor_status visit_enter(ir_function_signature *); - virtual ir_visitor_status visit_enter(ir_call *); - - signature_entry *get_signature_entry(ir_function_signature *var); - - bool (*predicate)(ir_instruction *ir); - - /* List of signature_entry */ - exec_list signature_list; - void *mem_ctx; - }; - - - signature_entry * - ir_dead_functions_visitor::get_signature_entry(ir_function_signature *sig) - { - foreach_iter(exec_list_iterator, iter, this->signature_list) { - signature_entry *entry = (signature_entry *)iter.get(); - if (entry->signature == sig) - return entry; - } - - signature_entry *entry = new(mem_ctx) signature_entry(sig); - this->signature_list.push_tail(entry); - return entry; - } - - - ir_visitor_status - ir_dead_functions_visitor::visit_enter(ir_function_signature *ir) - { - signature_entry *entry = this->get_signature_entry(ir); - - if (strcmp(ir->function_name(), "main") == 0) { - entry->used = true; - } - - return visit_continue; - } - - - ir_visitor_status - ir_dead_functions_visitor::visit_enter(ir_call *ir) - { - signature_entry *entry = this->get_signature_entry(ir->get_callee()); - - entry->used = true; - - return visit_continue; -} - -bool -do_dead_functions(exec_list *instructions) -{ - ir_dead_functions_visitor v; - bool progress = false; - - visit_list_elements(&v, instructions); - - /* Now that we've figured out which function signatures are used, remove - * the unused ones, and remove function definitions that have no more - * signatures. - */ - foreach_iter(exec_list_iterator, iter, v.signature_list) { - signature_entry *entry = (signature_entry *)iter.get(); - - if (!entry->used) { - entry->signature->remove(); - delete entry->signature; - progress = true; - } - delete(entry); - } - - /* We don't just do this above when we nuked a signature because of - * const pointers. - */ - foreach_iter(exec_list_iterator, iter, *instructions) { - ir_instruction *ir = (ir_instruction *)iter.get(); - ir_function *func = ir->as_function(); - - if (func && func->signatures.is_empty()) { - /* At this point (post-linking), the symbol table is no - * longer in use, so not removing the function from the - * symbol table should be OK. - */ - func->remove(); - delete func; - progress = true; - } - } - - return progress; -} + /*
+ * Copyright © 2010 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+ /**
+ * \file opt_dead_functions.cpp
+ *
+ * Eliminates unused functions from the linked program.
+ */
+
+ #include "ir.h"
+ #include "ir_visitor.h"
+ #include "ir_expression_flattening.h"
+ #include "glsl_types.h"
+
+ class signature_entry : public exec_node
+ {
+ public:
+ signature_entry(ir_function_signature *sig)
+ {
+ this->signature = sig;
+ this->used = false;
+ }
+
+ ir_function_signature *signature;
+ bool used;
+ };
+
+ class ir_dead_functions_visitor : public ir_hierarchical_visitor {
+ public:
+ ir_dead_functions_visitor()
+ {
+ this->mem_ctx = ralloc_context(NULL);
+ }
+
+ ~ir_dead_functions_visitor()
+ {
+ ralloc_free(this->mem_ctx);
+ }
+
+ virtual ir_visitor_status visit_enter(ir_function_signature *);
+ virtual ir_visitor_status visit_enter(ir_call *);
+
+ signature_entry *get_signature_entry(ir_function_signature *var);
+
+ bool (*predicate)(ir_instruction *ir);
+
+ /* List of signature_entry */
+ exec_list signature_list;
+ void *mem_ctx;
+ };
+
+
+ signature_entry *
+ ir_dead_functions_visitor::get_signature_entry(ir_function_signature *sig)
+ {
+ foreach_iter(exec_list_iterator, iter, this->signature_list) {
+ signature_entry *entry = (signature_entry *)iter.get();
+ if (entry->signature == sig)
+ return entry;
+ }
+
+ signature_entry *entry = new(mem_ctx) signature_entry(sig);
+ this->signature_list.push_tail(entry);
+ return entry;
+ }
+
+
+ ir_visitor_status
+ ir_dead_functions_visitor::visit_enter(ir_function_signature *ir)
+ {
+ signature_entry *entry = this->get_signature_entry(ir);
+
+ if (strcmp(ir->function_name(), "main") == 0) {
+ entry->used = true;
+ }
+
+ return visit_continue;
+ }
+
+
+ ir_visitor_status
+ ir_dead_functions_visitor::visit_enter(ir_call *ir)
+ {
+ signature_entry *entry = this->get_signature_entry(ir->get_callee());
+
+ entry->used = true;
+
+ return visit_continue;
+}
+
+bool
+do_dead_functions(exec_list *instructions)
+{
+ ir_dead_functions_visitor v;
+ bool progress = false;
+
+ visit_list_elements(&v, instructions);
+
+ /* Now that we've figured out which function signatures are used, remove
+ * the unused ones, and remove function definitions that have no more
+ * signatures.
+ */
+ foreach_iter(exec_list_iterator, iter, v.signature_list) {
+ signature_entry *entry = (signature_entry *)iter.get();
+
+ if (!entry->used) {
+ entry->signature->remove();
+ delete entry->signature;
+ progress = true;
+ }
+ delete(entry);
+ }
+
+ /* We don't just do this above when we nuked a signature because of
+ * const pointers.
+ */
+ foreach_iter(exec_list_iterator, iter, *instructions) {
+ ir_instruction *ir = (ir_instruction *)iter.get();
+ ir_function *func = ir->as_function();
+
+ if (func && func->signatures.is_empty()) {
+ /* At this point (post-linking), the symbol table is no
+ * longer in use, so not removing the function from the
+ * symbol table should be OK.
+ */
+ func->remove();
+ delete func;
+ progress = true;
+ }
+ }
+
+ return progress;
+}
diff --git a/mesalib/src/glsl/opt_structure_splitting.cpp b/mesalib/src/glsl/opt_structure_splitting.cpp index 8686da06a..ec2315499 100644 --- a/mesalib/src/glsl/opt_structure_splitting.cpp +++ b/mesalib/src/glsl/opt_structure_splitting.cpp @@ -1,364 +1,361 @@ -/* - * Copyright © 2010 Intel Corporation - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING - * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER - * DEALINGS IN THE SOFTWARE. - */ - -/** - * \file opt_structure_splitting.cpp - * - * If a structure is only ever referenced by its components, then - * split those components out to individual variables so they can be - * handled normally by other optimization passes. - * - * This skips structures like uniforms, which need to be accessible as - * structures for their access by the GL. - */ - -#include "ir.h" -#include "ir_visitor.h" -#include "ir_print_visitor.h" -#include "ir_rvalue_visitor.h" -#include "glsl_types.h" - -using std::printf; -using std::strcmp; - -static bool debug = false; - -// XXX using variable_entry2 here to avoid collision (MSVC multiply-defined -// function) with the variable_entry class seen in ir_variable_refcount.h -// Perhaps we can use the one in ir_variable_refcount.h and make this class -// here go away? -class variable_entry2 : public exec_node -{ -public: - variable_entry2(ir_variable *var) - { - this->var = var; - this->whole_structure_access = 0; - this->declaration = false; - this->components = NULL; - this->mem_ctx = NULL; - } - - ir_variable *var; /* The key: the variable's pointer. */ - - /** Number of times the variable is referenced, including assignments. */ - unsigned whole_structure_access; - - bool declaration; /* If the variable had a decl in the instruction stream */ - - ir_variable **components; - - /** ralloc_parent(this->var) -- the shader's ralloc context. */ - void *mem_ctx; -}; - - -class ir_structure_reference_visitor : public ir_hierarchical_visitor { -public: - ir_structure_reference_visitor(void) - { - this->mem_ctx = ralloc_context(NULL); - this->variable_list.make_empty(); - } - - ~ir_structure_reference_visitor(void) - { - ralloc_free(mem_ctx); - } - - virtual ir_visitor_status visit(ir_variable *); - virtual ir_visitor_status visit(ir_dereference_variable *); - virtual ir_visitor_status visit_enter(ir_dereference_record *); - virtual ir_visitor_status visit_enter(ir_assignment *); - virtual ir_visitor_status visit_enter(ir_function_signature *); - - variable_entry2 *get_variable_entry2(ir_variable *var); - - /* List of variable_entry */ - exec_list variable_list; - - void *mem_ctx; -}; - -variable_entry2 * -ir_structure_reference_visitor::get_variable_entry2(ir_variable *var) -{ - assert(var); - - if (!var->type->is_record() || var->mode == ir_var_uniform) - return NULL; - - foreach_iter(exec_list_iterator, iter, this->variable_list) { - variable_entry2 *entry = (variable_entry2 *)iter.get(); - if (entry->var == var) - return entry; - } - - variable_entry2 *entry = new(mem_ctx) variable_entry2(var); - this->variable_list.push_tail(entry); - return entry; -} - - -ir_visitor_status -ir_structure_reference_visitor::visit(ir_variable *ir) -{ - variable_entry2 *entry = this->get_variable_entry2(ir); - - if (entry) - entry->declaration = true; - - return visit_continue; -} - -ir_visitor_status -ir_structure_reference_visitor::visit(ir_dereference_variable *ir) -{ - ir_variable *const var = ir->variable_referenced(); - variable_entry2 *entry = this->get_variable_entry2(var); - - if (entry) - entry->whole_structure_access++; - - return visit_continue; -} - -ir_visitor_status -ir_structure_reference_visitor::visit_enter(ir_dereference_record *ir) -{ - (void) ir; - /* Don't descend into the ir_dereference_variable below. */ - return visit_continue_with_parent; -} - -ir_visitor_status -ir_structure_reference_visitor::visit_enter(ir_assignment *ir) -{ - if (ir->lhs->as_dereference_variable() && - ir->rhs->as_dereference_variable() && - !ir->condition) { - /* We'll split copies of a structure to copies of components, so don't - * descend to the ir_dereference_variables. - */ - return visit_continue_with_parent; - } - return visit_continue; -} - -ir_visitor_status -ir_structure_reference_visitor::visit_enter(ir_function_signature *ir) -{ - /* We don't want to descend into the function parameters and - * dead-code eliminate them, so just accept the body here. - */ - visit_list_elements(this, &ir->body); - return visit_continue_with_parent; -} - -class ir_structure_splitting_visitor : public ir_rvalue_visitor { -public: - ir_structure_splitting_visitor(exec_list *vars) - { - this->variable_list = vars; - } - - virtual ~ir_structure_splitting_visitor() - { - } - - virtual ir_visitor_status visit_leave(ir_assignment *); - - void split_deref(ir_dereference **deref); - void handle_rvalue(ir_rvalue **rvalue); - variable_entry2 *get_splitting_entry(ir_variable *var); - - exec_list *variable_list; - void *mem_ctx; -}; - -variable_entry2 * -ir_structure_splitting_visitor::get_splitting_entry(ir_variable *var) -{ - assert(var); - - if (!var->type->is_record()) - return NULL; - - foreach_iter(exec_list_iterator, iter, *this->variable_list) { - variable_entry2 *entry = (variable_entry2 *)iter.get(); - if (entry->var == var) { - return entry; - } - } - - return NULL; -} - -void -ir_structure_splitting_visitor::split_deref(ir_dereference **deref) -{ - if ((*deref)->ir_type != ir_type_dereference_record) - return; - - ir_dereference_record *deref_record = (ir_dereference_record *)*deref; - ir_dereference_variable *deref_var = deref_record->record->as_dereference_variable(); - if (!deref_var) - return; - - variable_entry2 *entry = get_splitting_entry(deref_var->var); - if (!entry) - return; - - unsigned int i; - for (i = 0; i < entry->var->type->length; i++) { - if (strcmp(deref_record->field, - entry->var->type->fields.structure[i].name) == 0) - break; - } - assert(i != entry->var->type->length); - - *deref = new(entry->mem_ctx) ir_dereference_variable(entry->components[i]); -} - -void -ir_structure_splitting_visitor::handle_rvalue(ir_rvalue **rvalue) -{ - if (!*rvalue) - return; - - ir_dereference *deref = (*rvalue)->as_dereference(); - - if (!deref) - return; - - split_deref(&deref); - *rvalue = deref; -} - -ir_visitor_status -ir_structure_splitting_visitor::visit_leave(ir_assignment *ir) -{ - ir_dereference_variable *lhs_deref = ir->lhs->as_dereference_variable(); - ir_dereference_variable *rhs_deref = ir->rhs->as_dereference_variable(); - variable_entry2 *lhs_entry = lhs_deref ? get_splitting_entry(lhs_deref->var) : NULL; - variable_entry2 *rhs_entry = rhs_deref ? get_splitting_entry(rhs_deref->var) : NULL; - const glsl_type *type = ir->rhs->type; - - if ((lhs_entry || rhs_entry) && !ir->condition) { - for (unsigned int i = 0; i < type->length; i++) { - ir_dereference *new_lhs, *new_rhs; - void *mem_ctx = lhs_entry ? lhs_entry->mem_ctx : rhs_entry->mem_ctx; - - if (lhs_entry) { - new_lhs = new(mem_ctx) ir_dereference_variable(lhs_entry->components[i]); - } else { - new_lhs = new(mem_ctx) - ir_dereference_record(ir->lhs->clone(mem_ctx, NULL), - type->fields.structure[i].name); - } - - if (rhs_entry) { - new_rhs = new(mem_ctx) ir_dereference_variable(rhs_entry->components[i]); - } else { - new_rhs = new(mem_ctx) - ir_dereference_record(ir->rhs->clone(mem_ctx, NULL), - type->fields.structure[i].name); - } - - ir->insert_before(new(mem_ctx) ir_assignment(new_lhs, - new_rhs, - NULL)); - } - ir->remove(); - } else { - handle_rvalue(&ir->rhs); - split_deref(&ir->lhs); - } - - handle_rvalue(&ir->condition); - - return visit_continue; -} - -bool -do_structure_splitting(exec_list *instructions) -{ - ir_structure_reference_visitor refs; - - visit_list_elements(&refs, instructions); - - /* Trim out variables we can't split. */ - foreach_iter(exec_list_iterator, iter, refs.variable_list) { - variable_entry2 *entry = (variable_entry2 *)iter.get(); - - if (debug) { - printf("structure %s@%p: decl %d, whole_access %d\n", - entry->var->name, (void *) entry->var, entry->declaration, - entry->whole_structure_access); - } - - if (!entry->declaration || entry->whole_structure_access) { - entry->remove(); - } - } - - if (refs.variable_list.is_empty()) - return false; - - void *mem_ctx = ralloc_context(NULL); - - /* Replace the decls of the structures to be split with their split - * components. - */ - foreach_iter(exec_list_iterator, iter, refs.variable_list) { - variable_entry2 *entry = (variable_entry2 *)iter.get(); - const struct glsl_type *type = entry->var->type; - - entry->mem_ctx = ralloc_parent(entry->var); - - entry->components = ralloc_array(mem_ctx, - ir_variable *, - type->length); - - for (unsigned int i = 0; i < entry->var->type->length; i++) { - const char *name = ralloc_asprintf(mem_ctx, "%s_%s", - entry->var->name, - type->fields.structure[i].name); - - entry->components[i] = - new(entry->mem_ctx) ir_variable(type->fields.structure[i].type, - name, - ir_var_temporary); - entry->var->insert_before(entry->components[i]); - } - - entry->var->remove(); - } - - ir_structure_splitting_visitor split(&refs.variable_list); - visit_list_elements(&split, instructions); - - ralloc_free(mem_ctx); - - return true; -} +/*
+ * Copyright © 2010 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file opt_structure_splitting.cpp
+ *
+ * If a structure is only ever referenced by its components, then
+ * split those components out to individual variables so they can be
+ * handled normally by other optimization passes.
+ *
+ * This skips structures like uniforms, which need to be accessible as
+ * structures for their access by the GL.
+ */
+
+#include "ir.h"
+#include "ir_visitor.h"
+#include "ir_print_visitor.h"
+#include "ir_rvalue_visitor.h"
+#include "glsl_types.h"
+
+static bool debug = false;
+
+// XXX using variable_entry2 here to avoid collision (MSVC multiply-defined
+// function) with the variable_entry class seen in ir_variable_refcount.h
+// Perhaps we can use the one in ir_variable_refcount.h and make this class
+// here go away?
+class variable_entry2 : public exec_node
+{
+public:
+ variable_entry2(ir_variable *var)
+ {
+ this->var = var;
+ this->whole_structure_access = 0;
+ this->declaration = false;
+ this->components = NULL;
+ this->mem_ctx = NULL;
+ }
+
+ ir_variable *var; /* The key: the variable's pointer. */
+
+ /** Number of times the variable is referenced, including assignments. */
+ unsigned whole_structure_access;
+
+ bool declaration; /* If the variable had a decl in the instruction stream */
+
+ ir_variable **components;
+
+ /** ralloc_parent(this->var) -- the shader's ralloc context. */
+ void *mem_ctx;
+};
+
+
+class ir_structure_reference_visitor : public ir_hierarchical_visitor {
+public:
+ ir_structure_reference_visitor(void)
+ {
+ this->mem_ctx = ralloc_context(NULL);
+ this->variable_list.make_empty();
+ }
+
+ ~ir_structure_reference_visitor(void)
+ {
+ ralloc_free(mem_ctx);
+ }
+
+ virtual ir_visitor_status visit(ir_variable *);
+ virtual ir_visitor_status visit(ir_dereference_variable *);
+ virtual ir_visitor_status visit_enter(ir_dereference_record *);
+ virtual ir_visitor_status visit_enter(ir_assignment *);
+ virtual ir_visitor_status visit_enter(ir_function_signature *);
+
+ variable_entry2 *get_variable_entry2(ir_variable *var);
+
+ /* List of variable_entry */
+ exec_list variable_list;
+
+ void *mem_ctx;
+};
+
+variable_entry2 *
+ir_structure_reference_visitor::get_variable_entry2(ir_variable *var)
+{
+ assert(var);
+
+ if (!var->type->is_record() || var->mode == ir_var_uniform)
+ return NULL;
+
+ foreach_iter(exec_list_iterator, iter, this->variable_list) {
+ variable_entry2 *entry = (variable_entry2 *)iter.get();
+ if (entry->var == var)
+ return entry;
+ }
+
+ variable_entry2 *entry = new(mem_ctx) variable_entry2(var);
+ this->variable_list.push_tail(entry);
+ return entry;
+}
+
+
+ir_visitor_status
+ir_structure_reference_visitor::visit(ir_variable *ir)
+{
+ variable_entry2 *entry = this->get_variable_entry2(ir);
+
+ if (entry)
+ entry->declaration = true;
+
+ return visit_continue;
+}
+
+ir_visitor_status
+ir_structure_reference_visitor::visit(ir_dereference_variable *ir)
+{
+ ir_variable *const var = ir->variable_referenced();
+ variable_entry2 *entry = this->get_variable_entry2(var);
+
+ if (entry)
+ entry->whole_structure_access++;
+
+ return visit_continue;
+}
+
+ir_visitor_status
+ir_structure_reference_visitor::visit_enter(ir_dereference_record *ir)
+{
+ (void) ir;
+ /* Don't descend into the ir_dereference_variable below. */
+ return visit_continue_with_parent;
+}
+
+ir_visitor_status
+ir_structure_reference_visitor::visit_enter(ir_assignment *ir)
+{
+ if (ir->lhs->as_dereference_variable() &&
+ ir->rhs->as_dereference_variable() &&
+ !ir->condition) {
+ /* We'll split copies of a structure to copies of components, so don't
+ * descend to the ir_dereference_variables.
+ */
+ return visit_continue_with_parent;
+ }
+ return visit_continue;
+}
+
+ir_visitor_status
+ir_structure_reference_visitor::visit_enter(ir_function_signature *ir)
+{
+ /* We don't want to descend into the function parameters and
+ * dead-code eliminate them, so just accept the body here.
+ */
+ visit_list_elements(this, &ir->body);
+ return visit_continue_with_parent;
+}
+
+class ir_structure_splitting_visitor : public ir_rvalue_visitor {
+public:
+ ir_structure_splitting_visitor(exec_list *vars)
+ {
+ this->variable_list = vars;
+ }
+
+ virtual ~ir_structure_splitting_visitor()
+ {
+ }
+
+ virtual ir_visitor_status visit_leave(ir_assignment *);
+
+ void split_deref(ir_dereference **deref);
+ void handle_rvalue(ir_rvalue **rvalue);
+ variable_entry2 *get_splitting_entry(ir_variable *var);
+
+ exec_list *variable_list;
+ void *mem_ctx;
+};
+
+variable_entry2 *
+ir_structure_splitting_visitor::get_splitting_entry(ir_variable *var)
+{
+ assert(var);
+
+ if (!var->type->is_record())
+ return NULL;
+
+ foreach_iter(exec_list_iterator, iter, *this->variable_list) {
+ variable_entry2 *entry = (variable_entry2 *)iter.get();
+ if (entry->var == var) {
+ return entry;
+ }
+ }
+
+ return NULL;
+}
+
+void
+ir_structure_splitting_visitor::split_deref(ir_dereference **deref)
+{
+ if ((*deref)->ir_type != ir_type_dereference_record)
+ return;
+
+ ir_dereference_record *deref_record = (ir_dereference_record *)*deref;
+ ir_dereference_variable *deref_var = deref_record->record->as_dereference_variable();
+ if (!deref_var)
+ return;
+
+ variable_entry2 *entry = get_splitting_entry(deref_var->var);
+ if (!entry)
+ return;
+
+ unsigned int i;
+ for (i = 0; i < entry->var->type->length; i++) {
+ if (strcmp(deref_record->field,
+ entry->var->type->fields.structure[i].name) == 0)
+ break;
+ }
+ assert(i != entry->var->type->length);
+
+ *deref = new(entry->mem_ctx) ir_dereference_variable(entry->components[i]);
+}
+
+void
+ir_structure_splitting_visitor::handle_rvalue(ir_rvalue **rvalue)
+{
+ if (!*rvalue)
+ return;
+
+ ir_dereference *deref = (*rvalue)->as_dereference();
+
+ if (!deref)
+ return;
+
+ split_deref(&deref);
+ *rvalue = deref;
+}
+
+ir_visitor_status
+ir_structure_splitting_visitor::visit_leave(ir_assignment *ir)
+{
+ ir_dereference_variable *lhs_deref = ir->lhs->as_dereference_variable();
+ ir_dereference_variable *rhs_deref = ir->rhs->as_dereference_variable();
+ variable_entry2 *lhs_entry = lhs_deref ? get_splitting_entry(lhs_deref->var) : NULL;
+ variable_entry2 *rhs_entry = rhs_deref ? get_splitting_entry(rhs_deref->var) : NULL;
+ const glsl_type *type = ir->rhs->type;
+
+ if ((lhs_entry || rhs_entry) && !ir->condition) {
+ for (unsigned int i = 0; i < type->length; i++) {
+ ir_dereference *new_lhs, *new_rhs;
+ void *mem_ctx = lhs_entry ? lhs_entry->mem_ctx : rhs_entry->mem_ctx;
+
+ if (lhs_entry) {
+ new_lhs = new(mem_ctx) ir_dereference_variable(lhs_entry->components[i]);
+ } else {
+ new_lhs = new(mem_ctx)
+ ir_dereference_record(ir->lhs->clone(mem_ctx, NULL),
+ type->fields.structure[i].name);
+ }
+
+ if (rhs_entry) {
+ new_rhs = new(mem_ctx) ir_dereference_variable(rhs_entry->components[i]);
+ } else {
+ new_rhs = new(mem_ctx)
+ ir_dereference_record(ir->rhs->clone(mem_ctx, NULL),
+ type->fields.structure[i].name);
+ }
+
+ ir->insert_before(new(mem_ctx) ir_assignment(new_lhs,
+ new_rhs,
+ NULL));
+ }
+ ir->remove();
+ } else {
+ handle_rvalue(&ir->rhs);
+ split_deref(&ir->lhs);
+ }
+
+ handle_rvalue(&ir->condition);
+
+ return visit_continue;
+}
+
+bool
+do_structure_splitting(exec_list *instructions)
+{
+ ir_structure_reference_visitor refs;
+
+ visit_list_elements(&refs, instructions);
+
+ /* Trim out variables we can't split. */
+ foreach_iter(exec_list_iterator, iter, refs.variable_list) {
+ variable_entry2 *entry = (variable_entry2 *)iter.get();
+
+ if (debug) {
+ printf("structure %s@%p: decl %d, whole_access %d\n",
+ entry->var->name, (void *) entry->var, entry->declaration,
+ entry->whole_structure_access);
+ }
+
+ if (!entry->declaration || entry->whole_structure_access) {
+ entry->remove();
+ }
+ }
+
+ if (refs.variable_list.is_empty())
+ return false;
+
+ void *mem_ctx = ralloc_context(NULL);
+
+ /* Replace the decls of the structures to be split with their split
+ * components.
+ */
+ foreach_iter(exec_list_iterator, iter, refs.variable_list) {
+ variable_entry2 *entry = (variable_entry2 *)iter.get();
+ const struct glsl_type *type = entry->var->type;
+
+ entry->mem_ctx = ralloc_parent(entry->var);
+
+ entry->components = ralloc_array(mem_ctx,
+ ir_variable *,
+ type->length);
+
+ for (unsigned int i = 0; i < entry->var->type->length; i++) {
+ const char *name = ralloc_asprintf(mem_ctx, "%s_%s",
+ entry->var->name,
+ type->fields.structure[i].name);
+
+ entry->components[i] =
+ new(entry->mem_ctx) ir_variable(type->fields.structure[i].type,
+ name,
+ ir_var_temporary);
+ entry->var->insert_before(entry->components[i]);
+ }
+
+ entry->var->remove();
+ }
+
+ ir_structure_splitting_visitor split(&refs.variable_list);
+ visit_list_elements(&split, instructions);
+
+ ralloc_free(mem_ctx);
+
+ return true;
+}
diff --git a/mesalib/src/glsl/opt_swizzle_swizzle.cpp b/mesalib/src/glsl/opt_swizzle_swizzle.cpp index 8d0e1051d..e23655240 100644 --- a/mesalib/src/glsl/opt_swizzle_swizzle.cpp +++ b/mesalib/src/glsl/opt_swizzle_swizzle.cpp @@ -1,95 +1,93 @@ -/* - * Copyright © 2010 Intel Corporation - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING - * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER - * DEALINGS IN THE SOFTWARE. - */ - -/** - * \file opt_swizzle_swizzle.cpp - * - * Eliminates the second swizzle in a swizzle chain. - */ - -#include "ir.h" -#include "ir_visitor.h" -#include "ir_optimization.h" -#include "glsl_types.h" - -using std::memset; - -class ir_swizzle_swizzle_visitor : public ir_hierarchical_visitor { -public: - ir_swizzle_swizzle_visitor() - { - progress = false; - } - - virtual ir_visitor_status visit_enter(ir_swizzle *); - - bool progress; -}; - -ir_visitor_status -ir_swizzle_swizzle_visitor::visit_enter(ir_swizzle *ir) -{ - int mask2[4]; - - ir_swizzle *swiz2 = ir->val->as_swizzle(); - if (!swiz2) - return visit_continue; - - memset(&mask2, 0, sizeof(mask2)); - if (swiz2->mask.num_components >= 1) - mask2[0] = swiz2->mask.x; - if (swiz2->mask.num_components >= 2) - mask2[1] = swiz2->mask.y; - if (swiz2->mask.num_components >= 3) - mask2[2] = swiz2->mask.z; - if (swiz2->mask.num_components >= 4) - mask2[3] = swiz2->mask.w; - - if (ir->mask.num_components >= 1) - ir->mask.x = mask2[ir->mask.x]; - if (ir->mask.num_components >= 2) - ir->mask.y = mask2[ir->mask.y]; - if (ir->mask.num_components >= 3) - ir->mask.z = mask2[ir->mask.z]; - if (ir->mask.num_components >= 4) - ir->mask.w = mask2[ir->mask.w]; - - ir->val = swiz2->val; - - this->progress = true; - - return visit_continue; -} - -/** - * Does a copy propagation pass on the code present in the instruction stream. - */ -bool -do_swizzle_swizzle(exec_list *instructions) -{ - ir_swizzle_swizzle_visitor v; - - v.run(instructions); - - return v.progress; -} +/*
+ * Copyright © 2010 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file opt_swizzle_swizzle.cpp
+ *
+ * Eliminates the second swizzle in a swizzle chain.
+ */
+
+#include "ir.h"
+#include "ir_visitor.h"
+#include "ir_optimization.h"
+#include "glsl_types.h"
+
+class ir_swizzle_swizzle_visitor : public ir_hierarchical_visitor {
+public:
+ ir_swizzle_swizzle_visitor()
+ {
+ progress = false;
+ }
+
+ virtual ir_visitor_status visit_enter(ir_swizzle *);
+
+ bool progress;
+};
+
+ir_visitor_status
+ir_swizzle_swizzle_visitor::visit_enter(ir_swizzle *ir)
+{
+ int mask2[4];
+
+ ir_swizzle *swiz2 = ir->val->as_swizzle();
+ if (!swiz2)
+ return visit_continue;
+
+ memset(&mask2, 0, sizeof(mask2));
+ if (swiz2->mask.num_components >= 1)
+ mask2[0] = swiz2->mask.x;
+ if (swiz2->mask.num_components >= 2)
+ mask2[1] = swiz2->mask.y;
+ if (swiz2->mask.num_components >= 3)
+ mask2[2] = swiz2->mask.z;
+ if (swiz2->mask.num_components >= 4)
+ mask2[3] = swiz2->mask.w;
+
+ if (ir->mask.num_components >= 1)
+ ir->mask.x = mask2[ir->mask.x];
+ if (ir->mask.num_components >= 2)
+ ir->mask.y = mask2[ir->mask.y];
+ if (ir->mask.num_components >= 3)
+ ir->mask.z = mask2[ir->mask.z];
+ if (ir->mask.num_components >= 4)
+ ir->mask.w = mask2[ir->mask.w];
+
+ ir->val = swiz2->val;
+
+ this->progress = true;
+
+ return visit_continue;
+}
+
+/**
+ * Does a copy propagation pass on the code present in the instruction stream.
+ */
+bool
+do_swizzle_swizzle(exec_list *instructions)
+{
+ ir_swizzle_swizzle_visitor v;
+
+ v.run(instructions);
+
+ return v.progress;
+}
diff --git a/mesalib/src/glsl/opt_tree_grafting.cpp b/mesalib/src/glsl/opt_tree_grafting.cpp index a85ba8234..75bd11c09 100644 --- a/mesalib/src/glsl/opt_tree_grafting.cpp +++ b/mesalib/src/glsl/opt_tree_grafting.cpp @@ -1,370 +1,368 @@ -/* - * Copyright © 2010 Intel Corporation - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING - * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER - * DEALINGS IN THE SOFTWARE. - */ - -/** - * \file opt_tree_grafting.cpp - * - * Takes assignments to variables that are dereferenced only once and - * pastes the RHS expression into where the variable is dereferenced. - * - * In the process of various operations like function inlining and - * tertiary op handling, we'll end up with our expression trees having - * been chopped up into a series of assignments of short expressions - * to temps. Other passes like ir_algebraic.cpp would prefer to see - * the deepest expression trees they can to try to optimize them. - * - * This is a lot like copy propagaton. In comparison, copy - * propagation only acts on plain copies, not arbitrary expressions on - * the RHS. Generally, we wouldn't want to go pasting some - * complicated expression everywhere it got used, though, so we don't - * handle expressions in that pass. - * - * The hard part is making sure we don't move an expression across - * some other assignments that would change the value of the - * expression. So we split this into two passes: First, find the - * variables in our scope which are written to once and read once, and - * then go through basic blocks seeing if we find an opportunity to - * move those expressions safely. - */ - -#include "ir.h" -#include "ir_visitor.h" -#include "ir_variable_refcount.h" -#include "ir_basic_block.h" -#include "ir_optimization.h" -#include "glsl_types.h" - -using std::printf; - -static bool debug = false; - -class ir_tree_grafting_visitor : public ir_hierarchical_visitor { -public: - ir_tree_grafting_visitor(ir_assignment *graft_assign, - ir_variable *graft_var) - { - this->progress = false; - this->graft_assign = graft_assign; - this->graft_var = graft_var; - } - - virtual ir_visitor_status visit_leave(class ir_assignment *); - virtual ir_visitor_status visit_enter(class ir_call *); - virtual ir_visitor_status visit_enter(class ir_expression *); - virtual ir_visitor_status visit_enter(class ir_function *); - virtual ir_visitor_status visit_enter(class ir_function_signature *); - virtual ir_visitor_status visit_enter(class ir_if *); - virtual ir_visitor_status visit_enter(class ir_loop *); - virtual ir_visitor_status visit_enter(class ir_swizzle *); - virtual ir_visitor_status visit_enter(class ir_texture *); - - bool do_graft(ir_rvalue **rvalue); - - bool progress; - ir_variable *graft_var; - ir_assignment *graft_assign; -}; - -struct find_deref_info { - ir_variable *var; - bool found; -}; - -void -dereferences_variable_callback(ir_instruction *ir, void *data) -{ - struct find_deref_info *info = (struct find_deref_info *)data; - ir_dereference_variable *deref = ir->as_dereference_variable(); - - if (deref && deref->var == info->var) - info->found = true; -} - -static bool -dereferences_variable(ir_instruction *ir, ir_variable *var) -{ - struct find_deref_info info; - - info.var = var; - info.found = false; - - visit_tree(ir, dereferences_variable_callback, &info); - - return info.found; -} - -bool -ir_tree_grafting_visitor::do_graft(ir_rvalue **rvalue) -{ - if (!*rvalue) - return false; - - ir_dereference_variable *deref = (*rvalue)->as_dereference_variable(); - - if (!deref || deref->var != this->graft_var) - return false; - - if (debug) { - printf("GRAFTING:\n"); - this->graft_assign->print(); - printf("\n"); - printf("TO:\n"); - (*rvalue)->print(); - printf("\n"); - } - - this->graft_assign->remove(); - *rvalue = this->graft_assign->rhs; - - this->progress = true; - return true; -} - -ir_visitor_status -ir_tree_grafting_visitor::visit_enter(ir_loop *ir) -{ - (void)ir; - /* Do not traverse into the body of the loop since that is a - * different basic block. - */ - return visit_stop; -} - -ir_visitor_status -ir_tree_grafting_visitor::visit_leave(ir_assignment *ir) -{ - if (do_graft(&ir->rhs) || - do_graft(&ir->condition)) - return visit_stop; - - /* If this assignment updates a variable used in the assignment - * we're trying to graft, then we're done. - */ - if (dereferences_variable(this->graft_assign->rhs, - ir->lhs->variable_referenced())) { - if (debug) { - printf("graft killed by: "); - ir->print(); - printf("\n"); - } - return visit_stop; - } - - return visit_continue; -} - -ir_visitor_status -ir_tree_grafting_visitor::visit_enter(ir_function *ir) -{ - (void) ir; - return visit_continue_with_parent; -} - -ir_visitor_status -ir_tree_grafting_visitor::visit_enter(ir_function_signature *ir) -{ - (void)ir; - return visit_continue_with_parent; -} - -ir_visitor_status -ir_tree_grafting_visitor::visit_enter(ir_call *ir) -{ - exec_list_iterator sig_iter = ir->get_callee()->parameters.iterator(); - /* Reminder: iterating ir_call iterates its parameters. */ - foreach_iter(exec_list_iterator, iter, *ir) { - ir_variable *sig_param = (ir_variable *)sig_iter.get(); - ir_rvalue *ir = (ir_rvalue *)iter.get(); - ir_rvalue *new_ir = ir; - - if (sig_param->mode != ir_var_in && sig_param->mode != ir_var_const_in) - continue; - - if (do_graft(&new_ir)) { - ir->replace_with(new_ir); - return visit_stop; - } - sig_iter.next(); - } - - return visit_continue; -} - -ir_visitor_status -ir_tree_grafting_visitor::visit_enter(ir_expression *ir) -{ - for (unsigned int i = 0; i < ir->get_num_operands(); i++) { - if (do_graft(&ir->operands[i])) - return visit_stop; - } - - return visit_continue; -} - -ir_visitor_status -ir_tree_grafting_visitor::visit_enter(ir_if *ir) -{ - if (do_graft(&ir->condition)) - return visit_stop; - - /* Do not traverse into the body of the if-statement since that is a - * different basic block. - */ - return visit_continue_with_parent; -} - -ir_visitor_status -ir_tree_grafting_visitor::visit_enter(ir_swizzle *ir) -{ - if (do_graft(&ir->val)) - return visit_stop; - - return visit_continue; -} - -ir_visitor_status -ir_tree_grafting_visitor::visit_enter(ir_texture *ir) -{ - if (do_graft(&ir->coordinate) || - do_graft(&ir->projector) || - do_graft(&ir->offset) || - do_graft(&ir->shadow_comparitor)) - return visit_stop; - - switch (ir->op) { - case ir_tex: - break; - case ir_txb: - if (do_graft(&ir->lod_info.bias)) - return visit_stop; - break; - case ir_txf: - case ir_txl: - if (do_graft(&ir->lod_info.lod)) - return visit_stop; - break; - case ir_txd: - if (do_graft(&ir->lod_info.grad.dPdx) || - do_graft(&ir->lod_info.grad.dPdy)) - return visit_stop; - break; - } - - return visit_continue; -} - -struct tree_grafting_info { - ir_variable_refcount_visitor *refs; - bool progress; -}; - -static bool -try_tree_grafting(ir_assignment *start, - ir_variable *lhs_var, - ir_instruction *bb_last) -{ - ir_tree_grafting_visitor v(start, lhs_var); - - if (debug) { - printf("trying to graft: "); - lhs_var->print(); - printf("\n"); - } - - for (ir_instruction *ir = (ir_instruction *)start->next; - ir != bb_last->next; - ir = (ir_instruction *)ir->next) { - - if (debug) { - printf("- "); - ir->print(); - printf("\n"); - } - - ir_visitor_status s = ir->accept(&v); - if (s == visit_stop) - return v.progress; - } - - return false; -} - -static void -tree_grafting_basic_block(ir_instruction *bb_first, - ir_instruction *bb_last, - void *data) -{ - struct tree_grafting_info *info = (struct tree_grafting_info *)data; - ir_instruction *ir, *next; - - for (ir = bb_first, next = (ir_instruction *)ir->next; - ir != bb_last->next; - ir = next, next = (ir_instruction *)ir->next) { - ir_assignment *assign = ir->as_assignment(); - - if (!assign) - continue; - - ir_variable *lhs_var = assign->whole_variable_written(); - if (!lhs_var) - continue; - - if (lhs_var->mode == ir_var_out || - lhs_var->mode == ir_var_inout) - continue; - - variable_entry *entry = info->refs->get_variable_entry(lhs_var); - - if (!entry->declaration || - entry->assigned_count != 1 || - entry->referenced_count != 2) - continue; - - assert(assign == entry->assign); - - /* Found a possibly graftable assignment. Now, walk through the - * rest of the BB seeing if the deref is here, and if nothing interfered with - * pasting its expression's values in between. - */ - info->progress |= try_tree_grafting(assign, lhs_var, bb_last); - } -} - -/** - * Does a copy propagation pass on the code present in the instruction stream. - */ -bool -do_tree_grafting(exec_list *instructions) -{ - ir_variable_refcount_visitor refs; - struct tree_grafting_info info; - - info.progress = false; - info.refs = &refs; - - visit_list_elements(info.refs, instructions); - - call_for_basic_blocks(instructions, tree_grafting_basic_block, &info); - - return info.progress; -} +/*
+ * Copyright © 2010 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file opt_tree_grafting.cpp
+ *
+ * Takes assignments to variables that are dereferenced only once and
+ * pastes the RHS expression into where the variable is dereferenced.
+ *
+ * In the process of various operations like function inlining and
+ * tertiary op handling, we'll end up with our expression trees having
+ * been chopped up into a series of assignments of short expressions
+ * to temps. Other passes like ir_algebraic.cpp would prefer to see
+ * the deepest expression trees they can to try to optimize them.
+ *
+ * This is a lot like copy propagaton. In comparison, copy
+ * propagation only acts on plain copies, not arbitrary expressions on
+ * the RHS. Generally, we wouldn't want to go pasting some
+ * complicated expression everywhere it got used, though, so we don't
+ * handle expressions in that pass.
+ *
+ * The hard part is making sure we don't move an expression across
+ * some other assignments that would change the value of the
+ * expression. So we split this into two passes: First, find the
+ * variables in our scope which are written to once and read once, and
+ * then go through basic blocks seeing if we find an opportunity to
+ * move those expressions safely.
+ */
+
+#include "ir.h"
+#include "ir_visitor.h"
+#include "ir_variable_refcount.h"
+#include "ir_basic_block.h"
+#include "ir_optimization.h"
+#include "glsl_types.h"
+
+static bool debug = false;
+
+class ir_tree_grafting_visitor : public ir_hierarchical_visitor {
+public:
+ ir_tree_grafting_visitor(ir_assignment *graft_assign,
+ ir_variable *graft_var)
+ {
+ this->progress = false;
+ this->graft_assign = graft_assign;
+ this->graft_var = graft_var;
+ }
+
+ virtual ir_visitor_status visit_leave(class ir_assignment *);
+ virtual ir_visitor_status visit_enter(class ir_call *);
+ virtual ir_visitor_status visit_enter(class ir_expression *);
+ virtual ir_visitor_status visit_enter(class ir_function *);
+ virtual ir_visitor_status visit_enter(class ir_function_signature *);
+ virtual ir_visitor_status visit_enter(class ir_if *);
+ virtual ir_visitor_status visit_enter(class ir_loop *);
+ virtual ir_visitor_status visit_enter(class ir_swizzle *);
+ virtual ir_visitor_status visit_enter(class ir_texture *);
+
+ bool do_graft(ir_rvalue **rvalue);
+
+ bool progress;
+ ir_variable *graft_var;
+ ir_assignment *graft_assign;
+};
+
+struct find_deref_info {
+ ir_variable *var;
+ bool found;
+};
+
+void
+dereferences_variable_callback(ir_instruction *ir, void *data)
+{
+ struct find_deref_info *info = (struct find_deref_info *)data;
+ ir_dereference_variable *deref = ir->as_dereference_variable();
+
+ if (deref && deref->var == info->var)
+ info->found = true;
+}
+
+static bool
+dereferences_variable(ir_instruction *ir, ir_variable *var)
+{
+ struct find_deref_info info;
+
+ info.var = var;
+ info.found = false;
+
+ visit_tree(ir, dereferences_variable_callback, &info);
+
+ return info.found;
+}
+
+bool
+ir_tree_grafting_visitor::do_graft(ir_rvalue **rvalue)
+{
+ if (!*rvalue)
+ return false;
+
+ ir_dereference_variable *deref = (*rvalue)->as_dereference_variable();
+
+ if (!deref || deref->var != this->graft_var)
+ return false;
+
+ if (debug) {
+ printf("GRAFTING:\n");
+ this->graft_assign->print();
+ printf("\n");
+ printf("TO:\n");
+ (*rvalue)->print();
+ printf("\n");
+ }
+
+ this->graft_assign->remove();
+ *rvalue = this->graft_assign->rhs;
+
+ this->progress = true;
+ return true;
+}
+
+ir_visitor_status
+ir_tree_grafting_visitor::visit_enter(ir_loop *ir)
+{
+ (void)ir;
+ /* Do not traverse into the body of the loop since that is a
+ * different basic block.
+ */
+ return visit_stop;
+}
+
+ir_visitor_status
+ir_tree_grafting_visitor::visit_leave(ir_assignment *ir)
+{
+ if (do_graft(&ir->rhs) ||
+ do_graft(&ir->condition))
+ return visit_stop;
+
+ /* If this assignment updates a variable used in the assignment
+ * we're trying to graft, then we're done.
+ */
+ if (dereferences_variable(this->graft_assign->rhs,
+ ir->lhs->variable_referenced())) {
+ if (debug) {
+ printf("graft killed by: ");
+ ir->print();
+ printf("\n");
+ }
+ return visit_stop;
+ }
+
+ return visit_continue;
+}
+
+ir_visitor_status
+ir_tree_grafting_visitor::visit_enter(ir_function *ir)
+{
+ (void) ir;
+ return visit_continue_with_parent;
+}
+
+ir_visitor_status
+ir_tree_grafting_visitor::visit_enter(ir_function_signature *ir)
+{
+ (void)ir;
+ return visit_continue_with_parent;
+}
+
+ir_visitor_status
+ir_tree_grafting_visitor::visit_enter(ir_call *ir)
+{
+ exec_list_iterator sig_iter = ir->get_callee()->parameters.iterator();
+ /* Reminder: iterating ir_call iterates its parameters. */
+ foreach_iter(exec_list_iterator, iter, *ir) {
+ ir_variable *sig_param = (ir_variable *)sig_iter.get();
+ ir_rvalue *ir = (ir_rvalue *)iter.get();
+ ir_rvalue *new_ir = ir;
+
+ if (sig_param->mode != ir_var_in && sig_param->mode != ir_var_const_in)
+ continue;
+
+ if (do_graft(&new_ir)) {
+ ir->replace_with(new_ir);
+ return visit_stop;
+ }
+ sig_iter.next();
+ }
+
+ return visit_continue;
+}
+
+ir_visitor_status
+ir_tree_grafting_visitor::visit_enter(ir_expression *ir)
+{
+ for (unsigned int i = 0; i < ir->get_num_operands(); i++) {
+ if (do_graft(&ir->operands[i]))
+ return visit_stop;
+ }
+
+ return visit_continue;
+}
+
+ir_visitor_status
+ir_tree_grafting_visitor::visit_enter(ir_if *ir)
+{
+ if (do_graft(&ir->condition))
+ return visit_stop;
+
+ /* Do not traverse into the body of the if-statement since that is a
+ * different basic block.
+ */
+ return visit_continue_with_parent;
+}
+
+ir_visitor_status
+ir_tree_grafting_visitor::visit_enter(ir_swizzle *ir)
+{
+ if (do_graft(&ir->val))
+ return visit_stop;
+
+ return visit_continue;
+}
+
+ir_visitor_status
+ir_tree_grafting_visitor::visit_enter(ir_texture *ir)
+{
+ if (do_graft(&ir->coordinate) ||
+ do_graft(&ir->projector) ||
+ do_graft(&ir->offset) ||
+ do_graft(&ir->shadow_comparitor))
+ return visit_stop;
+
+ switch (ir->op) {
+ case ir_tex:
+ break;
+ case ir_txb:
+ if (do_graft(&ir->lod_info.bias))
+ return visit_stop;
+ break;
+ case ir_txf:
+ case ir_txl:
+ if (do_graft(&ir->lod_info.lod))
+ return visit_stop;
+ break;
+ case ir_txd:
+ if (do_graft(&ir->lod_info.grad.dPdx) ||
+ do_graft(&ir->lod_info.grad.dPdy))
+ return visit_stop;
+ break;
+ }
+
+ return visit_continue;
+}
+
+struct tree_grafting_info {
+ ir_variable_refcount_visitor *refs;
+ bool progress;
+};
+
+static bool
+try_tree_grafting(ir_assignment *start,
+ ir_variable *lhs_var,
+ ir_instruction *bb_last)
+{
+ ir_tree_grafting_visitor v(start, lhs_var);
+
+ if (debug) {
+ printf("trying to graft: ");
+ lhs_var->print();
+ printf("\n");
+ }
+
+ for (ir_instruction *ir = (ir_instruction *)start->next;
+ ir != bb_last->next;
+ ir = (ir_instruction *)ir->next) {
+
+ if (debug) {
+ printf("- ");
+ ir->print();
+ printf("\n");
+ }
+
+ ir_visitor_status s = ir->accept(&v);
+ if (s == visit_stop)
+ return v.progress;
+ }
+
+ return false;
+}
+
+static void
+tree_grafting_basic_block(ir_instruction *bb_first,
+ ir_instruction *bb_last,
+ void *data)
+{
+ struct tree_grafting_info *info = (struct tree_grafting_info *)data;
+ ir_instruction *ir, *next;
+
+ for (ir = bb_first, next = (ir_instruction *)ir->next;
+ ir != bb_last->next;
+ ir = next, next = (ir_instruction *)ir->next) {
+ ir_assignment *assign = ir->as_assignment();
+
+ if (!assign)
+ continue;
+
+ ir_variable *lhs_var = assign->whole_variable_written();
+ if (!lhs_var)
+ continue;
+
+ if (lhs_var->mode == ir_var_out ||
+ lhs_var->mode == ir_var_inout)
+ continue;
+
+ variable_entry *entry = info->refs->get_variable_entry(lhs_var);
+
+ if (!entry->declaration ||
+ entry->assigned_count != 1 ||
+ entry->referenced_count != 2)
+ continue;
+
+ assert(assign == entry->assign);
+
+ /* Found a possibly graftable assignment. Now, walk through the
+ * rest of the BB seeing if the deref is here, and if nothing interfered with
+ * pasting its expression's values in between.
+ */
+ info->progress |= try_tree_grafting(assign, lhs_var, bb_last);
+ }
+}
+
+/**
+ * Does a copy propagation pass on the code present in the instruction stream.
+ */
+bool
+do_tree_grafting(exec_list *instructions)
+{
+ ir_variable_refcount_visitor refs;
+ struct tree_grafting_info info;
+
+ info.progress = false;
+ info.refs = &refs;
+
+ visit_list_elements(info.refs, instructions);
+
+ call_for_basic_blocks(instructions, tree_grafting_basic_block, &info);
+
+ return info.progress;
+}
diff --git a/mesalib/src/glsl/s_expression.cpp b/mesalib/src/glsl/s_expression.cpp index 77ac08ac7..b47b014c6 100644 --- a/mesalib/src/glsl/s_expression.cpp +++ b/mesalib/src/glsl/s_expression.cpp @@ -1,186 +1,183 @@ -/* -*- c++ -*- */ -/* - * Copyright © 2010 Intel Corporation - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING - * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER - * DEALINGS IN THE SOFTWARE. - */ - -#include <cstdio> -#include <cstdlib> -#include <cstring> -#include <assert.h> -#include "s_expression.h" - -s_symbol::s_symbol(const char *tmp, size_t n) -{ - this->str = ralloc_strndup (this, tmp, n); - assert(this->str != NULL); -} - -s_list::s_list() -{ -} - -static void -skip_whitespace(const char *& src) -{ - src += strspn(src, " \v\t\r\n"); - /* Also skip Scheme-style comments: semi-colon 'til end of line */ - if (src[0] == ';') { - src += strcspn(src, "\n"); - skip_whitespace(src); - } -} - -static s_expression * -read_atom(void *ctx, const char *& src) -{ - s_expression *expr = NULL; - - skip_whitespace(src); - - size_t n = strcspn(src, "( \v\t\r\n);"); - if (n == 0) - return NULL; // no atom - - // Check if the atom is a number. - char *float_end = NULL; - double f = glsl_strtod(src, &float_end); - if (float_end != src) { - char *int_end = NULL; - int i = strtol(src, &int_end, 10); - // If strtod matched more characters, it must have a decimal part - if (float_end > int_end) - expr = new(ctx) s_float(f); - else - expr = new(ctx) s_int(i); - } else { - // Not a number; return a symbol. - expr = new(ctx) s_symbol(src, n); - } - - src += n; - - return expr; -} - -s_expression * -s_expression::read_expression(void *ctx, const char *&src) -{ - assert(src != NULL); - - s_expression *atom = read_atom(ctx, src); - if (atom != NULL) - return atom; - - skip_whitespace(src); - if (src[0] == '(') { - ++src; - - s_list *list = new(ctx) s_list; - s_expression *expr; - - while ((expr = read_expression(ctx, src)) != NULL) { - list->subexpressions.push_tail(expr); - } - skip_whitespace(src); - if (src[0] != ')') { - printf("Unclosed expression (check your parenthesis).\n"); - return NULL; - } - ++src; - return list; - } - return NULL; -} - -void s_int::print() -{ - printf("%d", this->val); -} - -void s_float::print() -{ - printf("%f", this->val); -} - -void s_symbol::print() -{ - printf("%s", this->str); -} - -void s_list::print() -{ - printf("("); - foreach_iter(exec_list_iterator, it, this->subexpressions) { - s_expression *expr = (s_expression*) it.get(); - expr->print(); - if (!expr->next->is_tail_sentinel()) - printf(" "); - } - printf(")"); -} - -// -------------------------------------------------- - -bool -s_pattern::match(s_expression *expr) -{ - switch (type) - { - case EXPR: *p_expr = expr; break; - case LIST: if (expr->is_list()) *p_list = (s_list *) expr; break; - case SYMBOL: if (expr->is_symbol()) *p_symbol = (s_symbol *) expr; break; - case NUMBER: if (expr->is_number()) *p_number = (s_number *) expr; break; - case INT: if (expr->is_int()) *p_int = (s_int *) expr; break; - case STRING: - s_symbol *sym = SX_AS_SYMBOL(expr); - if (sym != NULL && strcmp(sym->value(), literal) == 0) - return true; - return false; - }; - - return *p_expr == expr; -} - -bool -s_match(s_expression *top, unsigned n, s_pattern *pattern, bool partial) -{ - s_list *list = SX_AS_LIST(top); - if (list == NULL) - return false; - - unsigned i = 0; - foreach_iter(exec_list_iterator, it, list->subexpressions) { - if (i >= n) - return partial; /* More actual items than the pattern expected */ - - s_expression *expr = (s_expression *) it.get(); - if (expr == NULL || !pattern[i].match(expr)) - return false; - - i++; - } - - if (i < n) - return false; /* Less actual items than the pattern expected */ - - return true; -} +/* -*- c++ -*- */
+/*
+ * Copyright © 2010 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include <assert.h>
+#include "s_expression.h"
+
+s_symbol::s_symbol(const char *tmp, size_t n)
+{
+ this->str = ralloc_strndup (this, tmp, n);
+ assert(this->str != NULL);
+}
+
+s_list::s_list()
+{
+}
+
+static void
+skip_whitespace(const char *& src)
+{
+ src += strspn(src, " \v\t\r\n");
+ /* Also skip Scheme-style comments: semi-colon 'til end of line */
+ if (src[0] == ';') {
+ src += strcspn(src, "\n");
+ skip_whitespace(src);
+ }
+}
+
+static s_expression *
+read_atom(void *ctx, const char *& src)
+{
+ s_expression *expr = NULL;
+
+ skip_whitespace(src);
+
+ size_t n = strcspn(src, "( \v\t\r\n);");
+ if (n == 0)
+ return NULL; // no atom
+
+ // Check if the atom is a number.
+ char *float_end = NULL;
+ double f = glsl_strtod(src, &float_end);
+ if (float_end != src) {
+ char *int_end = NULL;
+ int i = strtol(src, &int_end, 10);
+ // If strtod matched more characters, it must have a decimal part
+ if (float_end > int_end)
+ expr = new(ctx) s_float(f);
+ else
+ expr = new(ctx) s_int(i);
+ } else {
+ // Not a number; return a symbol.
+ expr = new(ctx) s_symbol(src, n);
+ }
+
+ src += n;
+
+ return expr;
+}
+
+s_expression *
+s_expression::read_expression(void *ctx, const char *&src)
+{
+ assert(src != NULL);
+
+ s_expression *atom = read_atom(ctx, src);
+ if (atom != NULL)
+ return atom;
+
+ skip_whitespace(src);
+ if (src[0] == '(') {
+ ++src;
+
+ s_list *list = new(ctx) s_list;
+ s_expression *expr;
+
+ while ((expr = read_expression(ctx, src)) != NULL) {
+ list->subexpressions.push_tail(expr);
+ }
+ skip_whitespace(src);
+ if (src[0] != ')') {
+ printf("Unclosed expression (check your parenthesis).\n");
+ return NULL;
+ }
+ ++src;
+ return list;
+ }
+ return NULL;
+}
+
+void s_int::print()
+{
+ printf("%d", this->val);
+}
+
+void s_float::print()
+{
+ printf("%f", this->val);
+}
+
+void s_symbol::print()
+{
+ printf("%s", this->str);
+}
+
+void s_list::print()
+{
+ printf("(");
+ foreach_iter(exec_list_iterator, it, this->subexpressions) {
+ s_expression *expr = (s_expression*) it.get();
+ expr->print();
+ if (!expr->next->is_tail_sentinel())
+ printf(" ");
+ }
+ printf(")");
+}
+
+// --------------------------------------------------
+
+bool
+s_pattern::match(s_expression *expr)
+{
+ switch (type)
+ {
+ case EXPR: *p_expr = expr; break;
+ case LIST: if (expr->is_list()) *p_list = (s_list *) expr; break;
+ case SYMBOL: if (expr->is_symbol()) *p_symbol = (s_symbol *) expr; break;
+ case NUMBER: if (expr->is_number()) *p_number = (s_number *) expr; break;
+ case INT: if (expr->is_int()) *p_int = (s_int *) expr; break;
+ case STRING:
+ s_symbol *sym = SX_AS_SYMBOL(expr);
+ if (sym != NULL && strcmp(sym->value(), literal) == 0)
+ return true;
+ return false;
+ };
+
+ return *p_expr == expr;
+}
+
+bool
+s_match(s_expression *top, unsigned n, s_pattern *pattern, bool partial)
+{
+ s_list *list = SX_AS_LIST(top);
+ if (list == NULL)
+ return false;
+
+ unsigned i = 0;
+ foreach_iter(exec_list_iterator, it, list->subexpressions) {
+ if (i >= n)
+ return partial; /* More actual items than the pattern expected */
+
+ s_expression *expr = (s_expression *) it.get();
+ if (expr == NULL || !pattern[i].match(expr))
+ return false;
+
+ i++;
+ }
+
+ if (i < n)
+ return false; /* Less actual items than the pattern expected */
+
+ return true;
+}
diff --git a/mesalib/src/mesa/main/imports.c b/mesalib/src/mesa/main/imports.c index f98098230..3f856402f 100644 --- a/mesalib/src/mesa/main/imports.c +++ b/mesalib/src/mesa/main/imports.c @@ -453,6 +453,7 @@ _mesa_inv_sqrtf(float n) #endif
}
+#ifndef __GNUC__
/**
* Find the first bit set in a word.
*/
@@ -496,9 +497,6 @@ _mesa_ffs(int32_t i) int
_mesa_ffsll(int64_t val)
{
-#ifdef ffsll
- return ffsll(val);
-#else
int bit;
assert(sizeof(val) == 8);
@@ -512,27 +510,24 @@ _mesa_ffsll(int64_t val) return 32 + bit;
return 0;
-#endif
}
+#if ((_GNUC__ == 3 && __GNUC_MINOR__ < 4) || __GNUC__ < 4)
/**
* Return number of bits set in given GLuint.
*/
unsigned int
_mesa_bitcount(unsigned int n)
{
-#if defined(__GNUC__) && \
- ((_GNUC__ == 3 && __GNUC_MINOR__ >= 4) || __GNUC__ >= 4)
- return __builtin_popcount(n);
-#else
unsigned int bits;
for (bits = 0; n > 0; n = n >> 1) {
bits += (n & 1);
}
return bits;
-#endif
}
+#endif
+#endif
/**
diff --git a/mesalib/src/mesa/main/imports.h b/mesalib/src/mesa/main/imports.h index af7a8cc00..9922f43e3 100644 --- a/mesalib/src/mesa/main/imports.h +++ b/mesalib/src/mesa/main/imports.h @@ -539,6 +539,18 @@ _mesa_inv_sqrtf(float x); extern void
_mesa_init_sqrt_table(void);
+#ifdef __GNUC__
+#define _mesa_ffs(i) ffs(i)
+#define _mesa_ffsll(i) ffsll(i)
+
+#if ((_GNUC__ == 3 && __GNUC_MINOR__ >= 4) || __GNUC__ >= 4)
+#define _mesa_bitcount(i) __builtin_popcount(i)
+#else
+extern unsigned int
+_mesa_bitcount(unsigned int n);
+#endif
+
+#else
extern int
_mesa_ffs(int32_t i);
@@ -547,6 +559,7 @@ _mesa_ffsll(int64_t i); extern unsigned int
_mesa_bitcount(unsigned int n);
+#endif
extern GLhalfARB
_mesa_float_to_half(float f);
diff --git a/mesalib/src/mesa/program/sampler.cpp b/mesalib/src/mesa/program/sampler.cpp index 12c4a40a2..b219d7016 100644 --- a/mesalib/src/mesa/program/sampler.cpp +++ b/mesalib/src/mesa/program/sampler.cpp @@ -1,138 +1,137 @@ -/* - * Copyright (C) 2005-2007 Brian Paul All Rights Reserved. - * Copyright (C) 2008 VMware, Inc. All Rights Reserved. - * Copyright © 2010 Intel Corporation - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING - * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER - * DEALINGS IN THE SOFTWARE. - */ - -#include <cstdio> -#include "ir.h" -#include "glsl_types.h" -#include "ir_visitor.h" - -extern "C" { -#include "main/compiler.h" -#include "main/mtypes.h" -#include "program/prog_parameter.h" -} - -static void fail_link(struct gl_shader_program *prog, const char *fmt, ...) PRINTFLIKE(2, 3); - -static void fail_link(struct gl_shader_program *prog, const char *fmt, ...) -{ - va_list args; - va_start(args, fmt); - ralloc_vasprintf_append(&prog->InfoLog, fmt, args); - va_end(args); - - prog->LinkStatus = GL_FALSE; -} - -class get_sampler_name : public ir_hierarchical_visitor -{ -public: - get_sampler_name(ir_dereference *last, - struct gl_shader_program *shader_program) - { - this->mem_ctx = ralloc_context(NULL); - this->shader_program = shader_program; - this->name = NULL; - this->offset = 0; - this->last = last; - } - - ~get_sampler_name() - { - ralloc_free(this->mem_ctx); - } - - virtual ir_visitor_status visit(ir_dereference_variable *ir) - { - this->name = ir->var->name; - return visit_continue; - } - - virtual ir_visitor_status visit_leave(ir_dereference_record *ir) - { - this->name = ralloc_asprintf(mem_ctx, "%s.%s", name, ir->field); - return visit_continue; - } - - virtual ir_visitor_status visit_leave(ir_dereference_array *ir) - { - ir_constant *index = ir->array_index->as_constant(); - int i; - - if (index) { - i = index->value.i[0]; - } else { - /* GLSL 1.10 and 1.20 allowed variable sampler array indices, - * while GLSL 1.30 requires that the array indices be - * constant integer expressions. We don't expect any driver - * to actually work with a really variable array index, so - * all that would work would be an unrolled loop counter that ends - * up being constant above. - */ - ralloc_strcat(&shader_program->InfoLog, - "warning: Variable sampler array index unsupported.\n" - "This feature of the language was removed in GLSL 1.20 " - "and is unlikely to be supported for 1.10 in Mesa.\n"); - i = 0; - } - if (ir != last) { - this->name = ralloc_asprintf(mem_ctx, "%s[%d]", name, i); - } else { - offset = i; - } - return visit_continue; - } - - struct gl_shader_program *shader_program; - const char *name; - void *mem_ctx; - int offset; - ir_dereference *last; -}; - -extern "C" { -int -_mesa_get_sampler_uniform_value(class ir_dereference *sampler, - struct gl_shader_program *shader_program, - const struct gl_program *prog) -{ - get_sampler_name getname(sampler, shader_program); - - sampler->accept(&getname); - - GLint index = _mesa_lookup_parameter_index(prog->Parameters, -1, - getname.name); - - if (index < 0) { - fail_link(shader_program, - "failed to find sampler named %s.\n", getname.name); - return 0; - } - - index += getname.offset; - - return prog->Parameters->ParameterValues[index][0]; -} -} +/*
+ * Copyright (C) 2005-2007 Brian Paul All Rights Reserved.
+ * Copyright (C) 2008 VMware, Inc. All Rights Reserved.
+ * Copyright © 2010 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include "ir.h"
+#include "glsl_types.h"
+#include "ir_visitor.h"
+
+extern "C" {
+#include "main/compiler.h"
+#include "main/mtypes.h"
+#include "program/prog_parameter.h"
+}
+
+static void fail_link(struct gl_shader_program *prog, const char *fmt, ...) PRINTFLIKE(2, 3);
+
+static void fail_link(struct gl_shader_program *prog, const char *fmt, ...)
+{
+ va_list args;
+ va_start(args, fmt);
+ ralloc_vasprintf_append(&prog->InfoLog, fmt, args);
+ va_end(args);
+
+ prog->LinkStatus = GL_FALSE;
+}
+
+class get_sampler_name : public ir_hierarchical_visitor
+{
+public:
+ get_sampler_name(ir_dereference *last,
+ struct gl_shader_program *shader_program)
+ {
+ this->mem_ctx = ralloc_context(NULL);
+ this->shader_program = shader_program;
+ this->name = NULL;
+ this->offset = 0;
+ this->last = last;
+ }
+
+ ~get_sampler_name()
+ {
+ ralloc_free(this->mem_ctx);
+ }
+
+ virtual ir_visitor_status visit(ir_dereference_variable *ir)
+ {
+ this->name = ir->var->name;
+ return visit_continue;
+ }
+
+ virtual ir_visitor_status visit_leave(ir_dereference_record *ir)
+ {
+ this->name = ralloc_asprintf(mem_ctx, "%s.%s", name, ir->field);
+ return visit_continue;
+ }
+
+ virtual ir_visitor_status visit_leave(ir_dereference_array *ir)
+ {
+ ir_constant *index = ir->array_index->as_constant();
+ int i;
+
+ if (index) {
+ i = index->value.i[0];
+ } else {
+ /* GLSL 1.10 and 1.20 allowed variable sampler array indices,
+ * while GLSL 1.30 requires that the array indices be
+ * constant integer expressions. We don't expect any driver
+ * to actually work with a really variable array index, so
+ * all that would work would be an unrolled loop counter that ends
+ * up being constant above.
+ */
+ ralloc_strcat(&shader_program->InfoLog,
+ "warning: Variable sampler array index unsupported.\n"
+ "This feature of the language was removed in GLSL 1.20 "
+ "and is unlikely to be supported for 1.10 in Mesa.\n");
+ i = 0;
+ }
+ if (ir != last) {
+ this->name = ralloc_asprintf(mem_ctx, "%s[%d]", name, i);
+ } else {
+ offset = i;
+ }
+ return visit_continue;
+ }
+
+ struct gl_shader_program *shader_program;
+ const char *name;
+ void *mem_ctx;
+ int offset;
+ ir_dereference *last;
+};
+
+extern "C" {
+int
+_mesa_get_sampler_uniform_value(class ir_dereference *sampler,
+ struct gl_shader_program *shader_program,
+ const struct gl_program *prog)
+{
+ get_sampler_name getname(sampler, shader_program);
+
+ sampler->accept(&getname);
+
+ GLint index = _mesa_lookup_parameter_index(prog->Parameters, -1,
+ getname.name);
+
+ if (index < 0) {
+ fail_link(shader_program,
+ "failed to find sampler named %s.\n", getname.name);
+ return 0;
+ }
+
+ index += getname.offset;
+
+ return prog->Parameters->ParameterValues[index][0];
+}
+}
diff --git a/mesalib/src/mesa/state_tracker/st_atom_sampler.c b/mesalib/src/mesa/state_tracker/st_atom_sampler.c index c9dfe248f..17596d4c8 100644 --- a/mesalib/src/mesa/state_tracker/st_atom_sampler.c +++ b/mesalib/src/mesa/state_tracker/st_atom_sampler.c @@ -36,6 +36,7 @@ #include "st_context.h"
#include "st_cb_texture.h"
+#include "st_format.h"
#include "st_atom.h"
#include "pipe/p_context.h"
#include "pipe/p_defines.h"
@@ -117,49 +118,6 @@ gl_filter_to_img_filter(GLenum filter) }
-static void
-xlate_border_color(const GLfloat *colorIn, GLenum baseFormat, GLfloat *colorOut)
-{
- switch (baseFormat) {
- case GL_RED:
- colorOut[0] = colorIn[0];
- colorOut[1] = 0.0F;
- colorOut[2] = 0.0F;
- colorOut[3] = 1.0F;
- break;
- case GL_RG:
- colorOut[0] = colorIn[0];
- colorOut[1] = colorIn[1];
- colorOut[2] = 0.0F;
- colorOut[3] = 1.0F;
- break;
- case GL_RGB:
- colorOut[0] = colorIn[0];
- colorOut[1] = colorIn[1];
- colorOut[2] = colorIn[2];
- colorOut[3] = 1.0F;
- break;
- case GL_ALPHA:
- colorOut[0] = colorOut[1] = colorOut[2] = 0.0;
- colorOut[3] = colorIn[3];
- break;
- case GL_LUMINANCE:
- colorOut[0] = colorOut[1] = colorOut[2] = colorIn[0];
- colorOut[3] = 1.0;
- break;
- case GL_LUMINANCE_ALPHA:
- colorOut[0] = colorOut[1] = colorOut[2] = colorIn[0];
- colorOut[3] = colorIn[3];
- break;
- case GL_INTENSITY:
- colorOut[0] = colorOut[1] = colorOut[2] = colorOut[3] = colorIn[0];
- break;
- default:
- COPY_4V(colorOut, colorIn);
- }
-}
-
-
static void
update_samplers(struct st_context *st)
{
@@ -223,7 +181,7 @@ update_samplers(struct st_context *st) assert(sampler->min_lod <= sampler->max_lod);
}
- xlate_border_color(texobj->BorderColor.f,
+ st_translate_color(texobj->BorderColor.f,
teximg ? teximg->_BaseFormat : GL_RGBA,
sampler->border_color);
diff --git a/mesalib/src/mesa/state_tracker/st_cb_clear.c b/mesalib/src/mesa/state_tracker/st_cb_clear.c index 461f4e315..79e58182f 100644 --- a/mesalib/src/mesa/state_tracker/st_cb_clear.c +++ b/mesalib/src/mesa/state_tracker/st_cb_clear.c @@ -42,6 +42,7 @@ #include "st_cb_accum.h"
#include "st_cb_clear.h"
#include "st_cb_fbo.h"
+#include "st_format.h"
#include "st_program.h"
#include "pipe/p_context.h"
@@ -204,6 +205,7 @@ clear_with_quad(struct gl_context *ctx, const GLfloat x1 = (GLfloat) ctx->DrawBuffer->_Xmax / fb_width * 2.0f - 1.0f;
const GLfloat y0 = (GLfloat) ctx->DrawBuffer->_Ymin / fb_height * 2.0f - 1.0f;
const GLfloat y1 = (GLfloat) ctx->DrawBuffer->_Ymax / fb_height * 2.0f - 1.0f;
+ float clearColor[4];
/*
printf("%s %s%s%s %f,%f %f,%f\n", __FUNCTION__,
@@ -298,9 +300,14 @@ clear_with_quad(struct gl_context *ctx, cso_set_fragment_shader_handle(st->cso_context, st->clear.fs);
cso_set_vertex_shader_handle(st->cso_context, st->clear.vs);
- /* draw quad matching scissor rect (XXX verify coord round-off) */
- draw_quad(st, x0, y0, x1, y1,
- (GLfloat) ctx->Depth.Clear, ctx->Color.ClearColor);
+ if (ctx->DrawBuffer->_ColorDrawBuffers[0]) {
+ st_translate_color(ctx->Color.ClearColor,
+ ctx->DrawBuffer->_ColorDrawBuffers[0]->_BaseFormat,
+ clearColor);
+ }
+
+ /* draw quad matching scissor rect */
+ draw_quad(st, x0, y0, x1, y1, (GLfloat) ctx->Depth.Clear, clearColor);
/* Restore pipe state */
cso_restore_blend(st->cso_context);
@@ -541,12 +548,21 @@ st_Clear(struct gl_context *ctx, GLbitfield mask) * required from the visual. Hence fix this up to avoid potential
* read-modify-write in the driver.
*/
+ float clearColor[4];
+
if ((clear_buffers & PIPE_CLEAR_DEPTHSTENCIL) &&
((clear_buffers & PIPE_CLEAR_DEPTHSTENCIL) != PIPE_CLEAR_DEPTHSTENCIL) &&
(depthRb == stencilRb) &&
(ctx->DrawBuffer->Visual.depthBits == 0 ||
ctx->DrawBuffer->Visual.stencilBits == 0))
clear_buffers |= PIPE_CLEAR_DEPTHSTENCIL;
+
+ if (ctx->DrawBuffer->_ColorDrawBuffers[0]) {
+ st_translate_color(ctx->Color.ClearColor,
+ ctx->DrawBuffer->_ColorDrawBuffers[0]->_BaseFormat,
+ clearColor);
+ }
+
st->pipe->clear(st->pipe, clear_buffers, ctx->Color.ClearColor,
ctx->Depth.Clear, ctx->Stencil.Clear);
}
diff --git a/mesalib/src/mesa/state_tracker/st_cb_texture.c b/mesalib/src/mesa/state_tracker/st_cb_texture.c index c0d1bd94a..708aa1290 100644 --- a/mesalib/src/mesa/state_tracker/st_cb_texture.c +++ b/mesalib/src/mesa/state_tracker/st_cb_texture.c @@ -260,6 +260,55 @@ get_texture_dims(GLenum target) /**
+ * Given the size of a mipmap image, try to compute the size of the level=0
+ * mipmap image.
+ *
+ * Note that this isn't always accurate for odd-sized, non-POW textures.
+ * For example, if level=1 and width=40 then the level=0 width may be 80 or 81.
+ *
+ * \return GL_TRUE for success, GL_FALSE for failure
+ */
+static GLboolean
+guess_base_level_size(GLenum target,
+ GLuint width, GLuint height, GLuint depth, GLuint level,
+ GLuint *width0, GLuint *height0, GLuint *depth0)
+{
+ const GLuint dims = get_texture_dims(target);
+
+ assert(width >= 1);
+ assert(height >= 1);
+ assert(depth >= 1);
+
+ if (level > 0) {
+ /* Depending on the image's size, we can't always make a guess here */
+ if ((dims >= 1 && width == 1) ||
+ (dims >= 2 && height == 1) ||
+ (dims >= 3 && depth == 1)) {
+ /* we can't determine the image size at level=0 */
+ return GL_FALSE;
+ }
+
+ /* grow the image size until we hit level = 0 */
+ while (level > 0) {
+ if (width > 1)
+ width <<= 1;
+ if (height > 1)
+ height <<= 1;
+ if (depth > 1)
+ depth <<= 1;
+ level--;
+ }
+ }
+
+ *width0 = width;
+ *height0 = height;
+ *depth0 = depth;
+
+ return GL_TRUE;
+}
+
+
+/**
* Try to allocate a pipe_resource object for the given st_texture_object.
*
* We use the given st_texture_image as a clue to determine the size of the
@@ -272,8 +321,7 @@ guess_and_alloc_texture(struct st_context *st, struct st_texture_object *stObj,
const struct st_texture_image *stImage)
{
- const GLuint dims = get_texture_dims(stObj->base.Target);
- GLuint level, lastLevel, width, height, depth;
+ GLuint lastLevel, width, height, depth;
GLuint bindings;
GLuint ptWidth, ptHeight, ptDepth, ptLayers;
enum pipe_format fmt;
@@ -282,41 +330,18 @@ guess_and_alloc_texture(struct st_context *st, assert(!stObj->pt);
- level = stImage->level;
- width = stImage->base.Width2; /* size w/out border */
- height = stImage->base.Height2;
- depth = stImage->base.Depth2;
-
- assert(width > 0);
- assert(height > 0);
- assert(depth > 0);
-
- /* Depending on the image's size, we can't always make a guess here.
- */
- if (level > 0) {
- if ( (dims >= 1 && width == 1) ||
- (dims >= 2 && height == 1) ||
- (dims >= 3 && depth == 1) ) {
- /* we can't determine the image size at level=0 */
- stObj->width0 = stObj->height0 = stObj->depth0 = 0;
- /* this is not an out of memory error */
- return GL_TRUE;
- }
+ if (!guess_base_level_size(stObj->base.Target,
+ stImage->base.Width2,
+ stImage->base.Height2,
+ stImage->base.Depth2,
+ stImage->level,
+ &width, &height, &depth)) {
+ /* we can't determine the image size at level=0 */
+ stObj->width0 = stObj->height0 = stObj->depth0 = 0;
+ /* this is not an out of memory error */
+ return GL_TRUE;
}
- /* grow the image size until we hit level = 0 */
- while (level > 0) {
- if (width != 1)
- width <<= 1;
- if (height != 1)
- height <<= 1;
- if (depth != 1)
- depth <<= 1;
- level--;
- }
-
- assert(level == 0);
-
/* At this point, (width x height x depth) is the expected size of
* the level=0 mipmap image.
*/
@@ -1721,9 +1746,24 @@ st_finalize_texture(struct gl_context *ctx, /* Find gallium format for the Mesa texture */
firstImageFormat = st_mesa_format_to_pipe_format(firstImage->base.TexFormat);
- st_gl_texture_dims_to_pipe_dims(stObj->base.Target, stObj->width0,
- stObj->height0, stObj->depth0,
- &ptWidth, &ptHeight, &ptDepth, &ptLayers);
+
+ /* Find size of level=0 Gallium mipmap image, plus number of texture layers */
+ {
+ GLuint width, height, depth;
+ if (!guess_base_level_size(stObj->base.Target,
+ firstImage->base.Width2,
+ firstImage->base.Height2,
+ firstImage->base.Depth2,
+ stObj->base.BaseLevel,
+ &width, &height, &depth)) {
+ width = stObj->width0;
+ height = stObj->height0;
+ depth = stObj->depth0;
+ }
+ /* convert GL dims to Gallium dims */
+ st_gl_texture_dims_to_pipe_dims(stObj->base.Target, width, height, depth,
+ &ptWidth, &ptHeight, &ptDepth, &ptLayers);
+ }
/* If we already have a gallium texture, check that it matches the texture
* object's format, target, size, num_levels, etc.
diff --git a/mesalib/src/mesa/state_tracker/st_draw.c b/mesalib/src/mesa/state_tracker/st_draw.c index 564d63483..7f661d529 100644 --- a/mesalib/src/mesa/state_tracker/st_draw.c +++ b/mesalib/src/mesa/state_tracker/st_draw.c @@ -249,6 +249,7 @@ is_interleaved_arrays(const struct st_vertex_program *vp, const struct gl_buffer_object *firstBufObj = NULL;
GLint firstStride = -1;
const GLubyte *client_addr = NULL;
+ GLboolean user_memory;
for (attr = 0; attr < vpv->num_inputs; attr++) {
const GLuint mesaAttr = vp->index_to_input[attr];
@@ -257,6 +258,7 @@ is_interleaved_arrays(const struct st_vertex_program *vp, if (firstStride < 0) {
firstStride = stride;
+ user_memory = !bufObj || !bufObj->Name;
}
else if (firstStride != stride) {
return GL_FALSE;
@@ -266,6 +268,9 @@ is_interleaved_arrays(const struct st_vertex_program *vp, /* Try to detect if the client-space arrays are
* "close" to each other.
*/
+ if (!user_memory) {
+ return GL_FALSE;
+ }
if (!client_addr) {
client_addr = arrays[mesaAttr]->Ptr;
}
@@ -275,6 +280,9 @@ is_interleaved_arrays(const struct st_vertex_program *vp, }
}
else if (!firstBufObj) {
+ if (user_memory) {
+ return GL_FALSE;
+ }
firstBufObj = bufObj;
}
else if (bufObj != firstBufObj) {
diff --git a/mesalib/src/mesa/state_tracker/st_format.c b/mesalib/src/mesa/state_tracker/st_format.c index a532e089c..3620d0cb6 100644 --- a/mesalib/src/mesa/state_tracker/st_format.c +++ b/mesalib/src/mesa/state_tracker/st_format.c @@ -1,1100 +1,1149 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * Copyright (c) 2008-2010 VMware, Inc. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -/** - * Mesa / Gallium format conversion and format selection code. - * \author Brian Paul - */ - -#include "main/imports.h" -#include "main/context.h" -#include "main/texstore.h" -#include "main/image.h" -#include "main/macros.h" -#include "main/mfeatures.h" - -#include "pipe/p_context.h" -#include "pipe/p_defines.h" -#include "pipe/p_screen.h" -#include "util/u_format.h" -#include "st_context.h" -#include "st_format.h" - - -static GLuint -format_max_bits(enum pipe_format format) -{ - GLuint size = util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 0); - - size = MAX2(size, util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 1)); - size = MAX2(size, util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 2)); - size = MAX2(size, util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 3)); - size = MAX2(size, util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_ZS, 0)); - size = MAX2(size, util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_ZS, 1)); - return size; -} - - -/** - * Return basic GL datatype for the given gallium format. - */ -GLenum -st_format_datatype(enum pipe_format format) -{ - const struct util_format_description *desc; - - desc = util_format_description(format); - assert(desc); - - if (desc->layout == UTIL_FORMAT_LAYOUT_PLAIN) { - if (format == PIPE_FORMAT_B5G5R5A1_UNORM || - format == PIPE_FORMAT_B5G6R5_UNORM) { - return GL_UNSIGNED_SHORT; - } - else if (format == PIPE_FORMAT_Z24_UNORM_S8_USCALED || - format == PIPE_FORMAT_S8_USCALED_Z24_UNORM || - format == PIPE_FORMAT_Z24X8_UNORM || - format == PIPE_FORMAT_X8Z24_UNORM) { - return GL_UNSIGNED_INT_24_8; - } - else { - const GLuint size = format_max_bits(format); - if (size == 8) { - if (desc->channel[0].type == UTIL_FORMAT_TYPE_UNSIGNED) - return GL_UNSIGNED_BYTE; - else - return GL_BYTE; - } - else if (size == 16) { - if (desc->channel[0].type == UTIL_FORMAT_TYPE_UNSIGNED) - return GL_UNSIGNED_SHORT; - else - return GL_SHORT; - } - else { - assert( size <= 32 ); - if (desc->channel[0].type == UTIL_FORMAT_TYPE_UNSIGNED) - return GL_UNSIGNED_INT; - else - return GL_INT; - } - } - } - else if (format == PIPE_FORMAT_UYVY) { - return GL_UNSIGNED_SHORT; - } - else if (format == PIPE_FORMAT_YUYV) { - return GL_UNSIGNED_SHORT; - } - else { - /* compressed format? */ - assert(0); - } - - assert(0); - return GL_NONE; -} - - -/** - * Translate Mesa format to Gallium format. - */ -enum pipe_format -st_mesa_format_to_pipe_format(gl_format mesaFormat) -{ - switch (mesaFormat) { - case MESA_FORMAT_RGBA8888: - return PIPE_FORMAT_A8B8G8R8_UNORM; - case MESA_FORMAT_RGBA8888_REV: - return PIPE_FORMAT_R8G8B8A8_UNORM; - case MESA_FORMAT_ARGB8888: - return PIPE_FORMAT_B8G8R8A8_UNORM; - case MESA_FORMAT_ARGB8888_REV: - return PIPE_FORMAT_A8R8G8B8_UNORM; - case MESA_FORMAT_XRGB8888: - return PIPE_FORMAT_B8G8R8X8_UNORM; - case MESA_FORMAT_XRGB8888_REV: - return PIPE_FORMAT_X8R8G8B8_UNORM; - case MESA_FORMAT_ARGB1555: - return PIPE_FORMAT_B5G5R5A1_UNORM; - case MESA_FORMAT_ARGB4444: - return PIPE_FORMAT_B4G4R4A4_UNORM; - case MESA_FORMAT_RGB565: - return PIPE_FORMAT_B5G6R5_UNORM; - case MESA_FORMAT_RGB332: - return PIPE_FORMAT_B2G3R3_UNORM; - case MESA_FORMAT_ARGB2101010: - return PIPE_FORMAT_B10G10R10A2_UNORM; - case MESA_FORMAT_AL44: - return PIPE_FORMAT_L4A4_UNORM; - case MESA_FORMAT_AL88: - return PIPE_FORMAT_L8A8_UNORM; - case MESA_FORMAT_AL1616: - return PIPE_FORMAT_L16A16_UNORM; - case MESA_FORMAT_A8: - return PIPE_FORMAT_A8_UNORM; - case MESA_FORMAT_A16: - return PIPE_FORMAT_A16_UNORM; - case MESA_FORMAT_L8: - return PIPE_FORMAT_L8_UNORM; - case MESA_FORMAT_L16: - return PIPE_FORMAT_L16_UNORM; - case MESA_FORMAT_I8: - return PIPE_FORMAT_I8_UNORM; - case MESA_FORMAT_I16: - return PIPE_FORMAT_I16_UNORM; - case MESA_FORMAT_Z16: - return PIPE_FORMAT_Z16_UNORM; - case MESA_FORMAT_Z32: - return PIPE_FORMAT_Z32_UNORM; - case MESA_FORMAT_Z24_S8: - return PIPE_FORMAT_S8_USCALED_Z24_UNORM; - case MESA_FORMAT_S8_Z24: - return PIPE_FORMAT_Z24_UNORM_S8_USCALED; - case MESA_FORMAT_Z24_X8: - return PIPE_FORMAT_X8Z24_UNORM; - case MESA_FORMAT_X8_Z24: - return PIPE_FORMAT_Z24X8_UNORM; - case MESA_FORMAT_S8: - return PIPE_FORMAT_S8_USCALED; - case MESA_FORMAT_YCBCR: - return PIPE_FORMAT_UYVY; -#if FEATURE_texture_s3tc - case MESA_FORMAT_RGB_DXT1: - return PIPE_FORMAT_DXT1_RGB; - case MESA_FORMAT_RGBA_DXT1: - return PIPE_FORMAT_DXT1_RGBA; - case MESA_FORMAT_RGBA_DXT3: - return PIPE_FORMAT_DXT3_RGBA; - case MESA_FORMAT_RGBA_DXT5: - return PIPE_FORMAT_DXT5_RGBA; -#if FEATURE_EXT_texture_sRGB - case MESA_FORMAT_SRGB_DXT1: - return PIPE_FORMAT_DXT1_SRGB; - case MESA_FORMAT_SRGBA_DXT1: - return PIPE_FORMAT_DXT1_SRGBA; - case MESA_FORMAT_SRGBA_DXT3: - return PIPE_FORMAT_DXT3_SRGBA; - case MESA_FORMAT_SRGBA_DXT5: - return PIPE_FORMAT_DXT5_SRGBA; -#endif -#endif -#if FEATURE_EXT_texture_sRGB - case MESA_FORMAT_SLA8: - return PIPE_FORMAT_L8A8_SRGB; - case MESA_FORMAT_SL8: - return PIPE_FORMAT_L8_SRGB; - case MESA_FORMAT_SRGB8: - return PIPE_FORMAT_R8G8B8_SRGB; - case MESA_FORMAT_SRGBA8: - return PIPE_FORMAT_A8B8G8R8_SRGB; - case MESA_FORMAT_SARGB8: - return PIPE_FORMAT_B8G8R8A8_SRGB; -#endif - case MESA_FORMAT_R8: - return PIPE_FORMAT_R8_UNORM; - case MESA_FORMAT_R16: - return PIPE_FORMAT_R16_UNORM; - case MESA_FORMAT_RG88: - return PIPE_FORMAT_R8G8_UNORM; - case MESA_FORMAT_RG1616: - return PIPE_FORMAT_R16G16_UNORM; - case MESA_FORMAT_RGBA_16: - return PIPE_FORMAT_R16G16B16A16_UNORM; - - /* signed int formats */ - case MESA_FORMAT_RGBA_INT8: - return PIPE_FORMAT_R8G8B8A8_SSCALED; - case MESA_FORMAT_RGBA_INT16: - return PIPE_FORMAT_R16G16B16A16_SSCALED; - case MESA_FORMAT_RGBA_INT32: - return PIPE_FORMAT_R32G32B32A32_SSCALED; - - /* unsigned int formats */ - case MESA_FORMAT_RGBA_UINT8: - return PIPE_FORMAT_R8G8B8A8_USCALED; - case MESA_FORMAT_RGBA_UINT16: - return PIPE_FORMAT_R16G16B16A16_USCALED; - case MESA_FORMAT_RGBA_UINT32: - return PIPE_FORMAT_R32G32B32A32_USCALED; - - default: - assert(0); - return PIPE_FORMAT_NONE; - } -} - - -/** - * Translate Gallium format to Mesa format. - */ -gl_format -st_pipe_format_to_mesa_format(enum pipe_format format) -{ - switch (format) { - case PIPE_FORMAT_A8B8G8R8_UNORM: - return MESA_FORMAT_RGBA8888; - case PIPE_FORMAT_R8G8B8A8_UNORM: - return MESA_FORMAT_RGBA8888_REV; - case PIPE_FORMAT_B8G8R8A8_UNORM: - return MESA_FORMAT_ARGB8888; - case PIPE_FORMAT_A8R8G8B8_UNORM: - return MESA_FORMAT_ARGB8888_REV; - case PIPE_FORMAT_B8G8R8X8_UNORM: - return MESA_FORMAT_XRGB8888; - case PIPE_FORMAT_X8R8G8B8_UNORM: - return MESA_FORMAT_XRGB8888_REV; - case PIPE_FORMAT_B5G5R5A1_UNORM: - return MESA_FORMAT_ARGB1555; - case PIPE_FORMAT_B4G4R4A4_UNORM: - return MESA_FORMAT_ARGB4444; - case PIPE_FORMAT_B5G6R5_UNORM: - return MESA_FORMAT_RGB565; - case PIPE_FORMAT_B2G3R3_UNORM: - return MESA_FORMAT_RGB332; - case PIPE_FORMAT_B10G10R10A2_UNORM: - return MESA_FORMAT_ARGB2101010; - case PIPE_FORMAT_L4A4_UNORM: - return MESA_FORMAT_AL44; - case PIPE_FORMAT_L8A8_UNORM: - return MESA_FORMAT_AL88; - case PIPE_FORMAT_L16A16_UNORM: - return MESA_FORMAT_AL1616; - case PIPE_FORMAT_A8_UNORM: - return MESA_FORMAT_A8; - case PIPE_FORMAT_A16_UNORM: - return MESA_FORMAT_A16; - case PIPE_FORMAT_L8_UNORM: - return MESA_FORMAT_L8; - case PIPE_FORMAT_L16_UNORM: - return MESA_FORMAT_L16; - case PIPE_FORMAT_I8_UNORM: - return MESA_FORMAT_I8; - case PIPE_FORMAT_I16_UNORM: - return MESA_FORMAT_I16; - case PIPE_FORMAT_S8_USCALED: - return MESA_FORMAT_S8; - - case PIPE_FORMAT_R16G16B16A16_UNORM: - return MESA_FORMAT_RGBA_16; - case PIPE_FORMAT_R16G16B16A16_SNORM: - return MESA_FORMAT_SIGNED_RGBA_16; - - case PIPE_FORMAT_Z16_UNORM: - return MESA_FORMAT_Z16; - case PIPE_FORMAT_Z32_UNORM: - return MESA_FORMAT_Z32; - case PIPE_FORMAT_S8_USCALED_Z24_UNORM: - return MESA_FORMAT_Z24_S8; - case PIPE_FORMAT_X8Z24_UNORM: - return MESA_FORMAT_Z24_X8; - case PIPE_FORMAT_Z24X8_UNORM: - return MESA_FORMAT_X8_Z24; - case PIPE_FORMAT_Z24_UNORM_S8_USCALED: - return MESA_FORMAT_S8_Z24; - - case PIPE_FORMAT_UYVY: - return MESA_FORMAT_YCBCR; - case PIPE_FORMAT_YUYV: - return MESA_FORMAT_YCBCR_REV; - -#if FEATURE_texture_s3tc - case PIPE_FORMAT_DXT1_RGB: - return MESA_FORMAT_RGB_DXT1; - case PIPE_FORMAT_DXT1_RGBA: - return MESA_FORMAT_RGBA_DXT1; - case PIPE_FORMAT_DXT3_RGBA: - return MESA_FORMAT_RGBA_DXT3; - case PIPE_FORMAT_DXT5_RGBA: - return MESA_FORMAT_RGBA_DXT5; -#if FEATURE_EXT_texture_sRGB - case PIPE_FORMAT_DXT1_SRGB: - return MESA_FORMAT_SRGB_DXT1; - case PIPE_FORMAT_DXT1_SRGBA: - return MESA_FORMAT_SRGBA_DXT1; - case PIPE_FORMAT_DXT3_SRGBA: - return MESA_FORMAT_SRGBA_DXT3; - case PIPE_FORMAT_DXT5_SRGBA: - return MESA_FORMAT_SRGBA_DXT5; -#endif -#endif - -#if FEATURE_EXT_texture_sRGB - case PIPE_FORMAT_L8A8_SRGB: - return MESA_FORMAT_SLA8; - case PIPE_FORMAT_L8_SRGB: - return MESA_FORMAT_SL8; - case PIPE_FORMAT_R8G8B8_SRGB: - return MESA_FORMAT_SRGB8; - case PIPE_FORMAT_A8B8G8R8_SRGB: - return MESA_FORMAT_SRGBA8; - case PIPE_FORMAT_B8G8R8A8_SRGB: - return MESA_FORMAT_SARGB8; -#endif - - case PIPE_FORMAT_R8_UNORM: - return MESA_FORMAT_R8; - case PIPE_FORMAT_R16_UNORM: - return MESA_FORMAT_R16; - case PIPE_FORMAT_R8G8_UNORM: - return MESA_FORMAT_RG88; - case PIPE_FORMAT_R16G16_UNORM: - return MESA_FORMAT_RG1616; - - /* signed int formats */ - case PIPE_FORMAT_R8G8B8A8_SSCALED: - return MESA_FORMAT_RGBA_INT8; - case PIPE_FORMAT_R16G16B16A16_SSCALED: - return MESA_FORMAT_RGBA_INT16; - case PIPE_FORMAT_R32G32B32A32_SSCALED: - return MESA_FORMAT_RGBA_INT32; - - /* unsigned int formats */ - case PIPE_FORMAT_R8G8B8A8_USCALED: - return MESA_FORMAT_RGBA_UINT8; - case PIPE_FORMAT_R16G16B16A16_USCALED: - return MESA_FORMAT_RGBA_UINT16; - case PIPE_FORMAT_R32G32B32A32_USCALED: - return MESA_FORMAT_RGBA_UINT32; - - default: - assert(0); - return MESA_FORMAT_NONE; - } -} - - -/** - * Return first supported format from the given list. - */ -static enum pipe_format -find_supported_format(struct pipe_screen *screen, - const enum pipe_format formats[], - uint num_formats, - enum pipe_texture_target target, - unsigned sample_count, - unsigned tex_usage, - unsigned geom_flags) -{ - uint i; - for (i = 0; i < num_formats; i++) { - if (screen->is_format_supported(screen, formats[i], target, - sample_count, tex_usage, geom_flags)) { - return formats[i]; - } - } - return PIPE_FORMAT_NONE; -} - - -/** - * Find an RGBA format supported by the context/winsys. - */ -static enum pipe_format -default_rgba_format(struct pipe_screen *screen, - enum pipe_texture_target target, - unsigned sample_count, - unsigned tex_usage, - unsigned geom_flags) -{ - static const enum pipe_format colorFormats[] = { - PIPE_FORMAT_B8G8R8A8_UNORM, - PIPE_FORMAT_A8R8G8B8_UNORM, - PIPE_FORMAT_A8B8G8R8_UNORM, - PIPE_FORMAT_B5G6R5_UNORM - }; - return find_supported_format(screen, colorFormats, Elements(colorFormats), - target, sample_count, tex_usage, geom_flags); -} - - -/** - * Find an RGB format supported by the context/winsys. - */ -static enum pipe_format -default_rgb_format(struct pipe_screen *screen, - enum pipe_texture_target target, - unsigned sample_count, - unsigned tex_usage, - unsigned geom_flags) -{ - static const enum pipe_format colorFormats[] = { - PIPE_FORMAT_B8G8R8X8_UNORM, - PIPE_FORMAT_X8R8G8B8_UNORM, - PIPE_FORMAT_X8B8G8R8_UNORM, - PIPE_FORMAT_B8G8R8A8_UNORM, - PIPE_FORMAT_A8R8G8B8_UNORM, - PIPE_FORMAT_A8B8G8R8_UNORM, - PIPE_FORMAT_B5G6R5_UNORM - }; - return find_supported_format(screen, colorFormats, Elements(colorFormats), - target, sample_count, tex_usage, geom_flags); -} - -/** - * Find an sRGBA format supported by the context/winsys. - */ -static enum pipe_format -default_srgba_format(struct pipe_screen *screen, - enum pipe_texture_target target, - unsigned sample_count, - unsigned tex_usage, - unsigned geom_flags) -{ - static const enum pipe_format colorFormats[] = { - PIPE_FORMAT_B8G8R8A8_SRGB, - PIPE_FORMAT_A8R8G8B8_SRGB, - PIPE_FORMAT_A8B8G8R8_SRGB, - }; - return find_supported_format(screen, colorFormats, Elements(colorFormats), - target, sample_count, tex_usage, geom_flags); -} - - -/** - * Given an OpenGL internalFormat value for a texture or surface, return - * the best matching PIPE_FORMAT_x, or PIPE_FORMAT_NONE if there's no match. - * This is called during glTexImage2D, for example. - * - * The bindings parameter typically has PIPE_BIND_SAMPLER_VIEW set, plus - * either PIPE_BINDING_RENDER_TARGET or PIPE_BINDING_DEPTH_STENCIL if - * we want render-to-texture ability. - * - * \param internalFormat the user value passed to glTexImage2D - * \param target one of PIPE_TEXTURE_x - * \param bindings bitmask of PIPE_BIND_x flags. - */ -enum pipe_format -st_choose_format(struct pipe_screen *screen, GLenum internalFormat, - enum pipe_texture_target target, unsigned sample_count, - unsigned bindings) -{ - unsigned geom_flags = 0; /* we don't care about POT vs. NPOT here, yet */ - - switch (internalFormat) { - case GL_RGB10: - case GL_RGB10_A2: - if (screen->is_format_supported( screen, PIPE_FORMAT_B10G10R10A2_UNORM, - target, sample_count, bindings, - geom_flags )) - return PIPE_FORMAT_B10G10R10A2_UNORM; - /* Pass through. */ - case 4: - case GL_RGBA: - case GL_RGBA8: - return default_rgba_format( screen, target, sample_count, bindings, - geom_flags ); - - case GL_BGRA: - if (screen->is_format_supported( screen, PIPE_FORMAT_B8G8R8A8_UNORM, - target, sample_count, bindings, - geom_flags )) - return PIPE_FORMAT_B8G8R8A8_UNORM; - return default_rgba_format( screen, target, sample_count, bindings, - geom_flags ); - - case 3: - case GL_RGB: - case GL_RGB8: - return default_rgb_format( screen, target, sample_count, bindings, - geom_flags ); - - case GL_RGB12: - case GL_RGB16: - case GL_RGBA12: - case GL_RGBA16: - if (screen->is_format_supported( screen, PIPE_FORMAT_R16G16B16A16_UNORM, - target, sample_count, bindings, - geom_flags )) - return PIPE_FORMAT_R16G16B16A16_UNORM; - return default_rgba_format( screen, target, sample_count, bindings, - geom_flags ); - - case GL_RGBA4: - case GL_RGBA2: - if (screen->is_format_supported( screen, PIPE_FORMAT_B4G4R4A4_UNORM, - target, sample_count, bindings, - geom_flags )) - return PIPE_FORMAT_B4G4R4A4_UNORM; - return default_rgba_format( screen, target, sample_count, bindings, - geom_flags ); - - case GL_RGB5_A1: - if (screen->is_format_supported( screen, PIPE_FORMAT_B5G5R5A1_UNORM, - target, sample_count, bindings, - geom_flags )) - return PIPE_FORMAT_B5G5R5A1_UNORM; - return default_rgba_format( screen, target, sample_count, bindings, - geom_flags ); - - case GL_R3_G3_B2: - if (screen->is_format_supported( screen, PIPE_FORMAT_B2G3R3_UNORM, - target, sample_count, bindings, - geom_flags )) - return PIPE_FORMAT_B2G3R3_UNORM; - /* Pass through. */ - case GL_RGB5: - case GL_RGB4: - if (screen->is_format_supported( screen, PIPE_FORMAT_B5G6R5_UNORM, - target, sample_count, bindings, - geom_flags )) - return PIPE_FORMAT_B5G6R5_UNORM; - if (screen->is_format_supported( screen, PIPE_FORMAT_B5G5R5A1_UNORM, - target, sample_count, bindings, - geom_flags )) - return PIPE_FORMAT_B5G5R5A1_UNORM; - return default_rgba_format( screen, target, sample_count, bindings, - geom_flags ); - - case GL_ALPHA12: - case GL_ALPHA16: - if (screen->is_format_supported( screen, PIPE_FORMAT_A16_UNORM, target, - sample_count, bindings, geom_flags )) - return PIPE_FORMAT_A16_UNORM; - /* Pass through. */ - case GL_ALPHA: - case GL_ALPHA4: - case GL_ALPHA8: - case GL_COMPRESSED_ALPHA: - if (screen->is_format_supported( screen, PIPE_FORMAT_A8_UNORM, target, - sample_count, bindings, geom_flags )) - return PIPE_FORMAT_A8_UNORM; - return default_rgba_format( screen, target, sample_count, bindings, - geom_flags ); - - case GL_LUMINANCE12: - case GL_LUMINANCE16: - if (screen->is_format_supported( screen, PIPE_FORMAT_L16_UNORM, target, - sample_count, bindings, geom_flags )) - return PIPE_FORMAT_L16_UNORM; - /* Pass through. */ - case 1: - case GL_LUMINANCE: - case GL_LUMINANCE4: - case GL_LUMINANCE8: - case GL_COMPRESSED_LUMINANCE: - if (screen->is_format_supported( screen, PIPE_FORMAT_L8_UNORM, target, - sample_count, bindings, geom_flags )) - return PIPE_FORMAT_L8_UNORM; - return default_rgba_format( screen, target, sample_count, bindings, - geom_flags ); - - case GL_LUMINANCE12_ALPHA4: - case GL_LUMINANCE12_ALPHA12: - case GL_LUMINANCE16_ALPHA16: - if (screen->is_format_supported( screen, PIPE_FORMAT_L16A16_UNORM, target, - sample_count, bindings, geom_flags )) - return PIPE_FORMAT_L16A16_UNORM; - /* Pass through. */ - case 2: - case GL_LUMINANCE_ALPHA: - case GL_LUMINANCE6_ALPHA2: - case GL_LUMINANCE8_ALPHA8: - case GL_COMPRESSED_LUMINANCE_ALPHA: - if (screen->is_format_supported( screen, PIPE_FORMAT_L8A8_UNORM, target, - sample_count, bindings, geom_flags )) - return PIPE_FORMAT_L8A8_UNORM; - return default_rgba_format( screen, target, sample_count, bindings, - geom_flags ); - - case GL_LUMINANCE4_ALPHA4: - if (screen->is_format_supported( screen, PIPE_FORMAT_L4A4_UNORM, target, - sample_count, bindings, geom_flags )) - return PIPE_FORMAT_L4A4_UNORM; - if (screen->is_format_supported( screen, PIPE_FORMAT_L8A8_UNORM, target, - sample_count, bindings, geom_flags )) - return PIPE_FORMAT_L8A8_UNORM; - return default_rgba_format( screen, target, sample_count, bindings, - geom_flags ); - - case GL_INTENSITY12: - case GL_INTENSITY16: - if (screen->is_format_supported( screen, PIPE_FORMAT_I16_UNORM, target, - sample_count, bindings, geom_flags )) - return PIPE_FORMAT_I16_UNORM; - /* Pass through. */ - case GL_INTENSITY: - case GL_INTENSITY4: - case GL_INTENSITY8: - case GL_COMPRESSED_INTENSITY: - if (screen->is_format_supported( screen, PIPE_FORMAT_I8_UNORM, target, - sample_count, bindings, geom_flags )) - return PIPE_FORMAT_I8_UNORM; - return default_rgba_format( screen, target, sample_count, bindings, - geom_flags ); - - case GL_YCBCR_MESA: - if (screen->is_format_supported(screen, PIPE_FORMAT_UYVY, target, - sample_count, bindings, geom_flags)) { - return PIPE_FORMAT_UYVY; - } - if (screen->is_format_supported(screen, PIPE_FORMAT_YUYV, target, - sample_count, bindings, geom_flags)) { - return PIPE_FORMAT_YUYV; - } - return PIPE_FORMAT_NONE; - - case GL_COMPRESSED_RED: - case GL_COMPRESSED_RG: - case GL_COMPRESSED_RGB: - /* can only sample from compressed formats */ - if (bindings & ~PIPE_BIND_SAMPLER_VIEW) - return PIPE_FORMAT_NONE; - else if (screen->is_format_supported(screen, PIPE_FORMAT_DXT1_RGB, - target, sample_count, bindings, - geom_flags)) - return PIPE_FORMAT_DXT1_RGB; - else - return default_rgb_format(screen, target, sample_count, bindings, - geom_flags); - - case GL_COMPRESSED_RGBA: - /* can only sample from compressed formats */ - if (bindings & ~PIPE_BIND_SAMPLER_VIEW) - return PIPE_FORMAT_NONE; - else if (screen->is_format_supported(screen, PIPE_FORMAT_DXT3_RGBA, - target, sample_count, bindings, - geom_flags)) - return PIPE_FORMAT_DXT3_RGBA; - else - return default_rgba_format(screen, target, sample_count, bindings, - geom_flags); - - case GL_RGB_S3TC: - case GL_RGB4_S3TC: - case GL_COMPRESSED_RGB_S3TC_DXT1_EXT: - if (screen->is_format_supported(screen, PIPE_FORMAT_DXT1_RGB, - target, sample_count, bindings, - geom_flags)) - return PIPE_FORMAT_DXT1_RGB; - else - return PIPE_FORMAT_NONE; - - case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT: - if (screen->is_format_supported(screen, PIPE_FORMAT_DXT1_RGBA, - target, sample_count, bindings, - geom_flags)) - return PIPE_FORMAT_DXT1_RGBA; - else - return PIPE_FORMAT_NONE; - - case GL_RGBA_S3TC: - case GL_RGBA4_S3TC: - case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT: - if (screen->is_format_supported(screen, PIPE_FORMAT_DXT3_RGBA, - target, sample_count, bindings, - geom_flags)) - return PIPE_FORMAT_DXT3_RGBA; - else - return PIPE_FORMAT_NONE; - - case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT: - if (screen->is_format_supported(screen, PIPE_FORMAT_DXT5_RGBA, - target, sample_count, bindings, - geom_flags)) - return PIPE_FORMAT_DXT5_RGBA; - else - return PIPE_FORMAT_NONE; - -#if 0 - case GL_COMPRESSED_RGB_FXT1_3DFX: - return PIPE_FORMAT_RGB_FXT1; - case GL_COMPRESSED_RGBA_FXT1_3DFX: - return PIPE_FORMAT_RGB_FXT1; -#endif - - case GL_DEPTH_COMPONENT16: - if (screen->is_format_supported(screen, PIPE_FORMAT_Z16_UNORM, target, - sample_count, bindings, geom_flags)) - return PIPE_FORMAT_Z16_UNORM; - /* fall-through */ - case GL_DEPTH_COMPONENT24: - if (screen->is_format_supported(screen, PIPE_FORMAT_Z24_UNORM_S8_USCALED, - target, sample_count, bindings, geom_flags)) - return PIPE_FORMAT_Z24_UNORM_S8_USCALED; - if (screen->is_format_supported(screen, PIPE_FORMAT_S8_USCALED_Z24_UNORM, - target, sample_count, bindings, geom_flags)) - return PIPE_FORMAT_S8_USCALED_Z24_UNORM; - /* fall-through */ - case GL_DEPTH_COMPONENT32: - if (screen->is_format_supported(screen, PIPE_FORMAT_Z32_UNORM, target, - sample_count, bindings, geom_flags)) - return PIPE_FORMAT_Z32_UNORM; - /* fall-through */ - case GL_DEPTH_COMPONENT: - { - static const enum pipe_format formats[] = { - PIPE_FORMAT_Z32_UNORM, - PIPE_FORMAT_Z24_UNORM_S8_USCALED, - PIPE_FORMAT_S8_USCALED_Z24_UNORM, - PIPE_FORMAT_Z16_UNORM - }; - return find_supported_format(screen, formats, Elements(formats), - target, sample_count, bindings, geom_flags); - } - - case GL_STENCIL_INDEX: - case GL_STENCIL_INDEX1_EXT: - case GL_STENCIL_INDEX4_EXT: - case GL_STENCIL_INDEX8_EXT: - case GL_STENCIL_INDEX16_EXT: - { - static const enum pipe_format formats[] = { - PIPE_FORMAT_S8_USCALED, - PIPE_FORMAT_Z24_UNORM_S8_USCALED, - PIPE_FORMAT_S8_USCALED_Z24_UNORM - }; - return find_supported_format(screen, formats, Elements(formats), - target, sample_count, bindings, geom_flags); - } - - case GL_DEPTH_STENCIL_EXT: - case GL_DEPTH24_STENCIL8_EXT: - { - static const enum pipe_format formats[] = { - PIPE_FORMAT_Z24_UNORM_S8_USCALED, - PIPE_FORMAT_S8_USCALED_Z24_UNORM - }; - return find_supported_format(screen, formats, Elements(formats), - target, sample_count, bindings, geom_flags); - } - - case GL_SRGB_EXT: - case GL_SRGB8_EXT: - case GL_SRGB_ALPHA_EXT: - case GL_SRGB8_ALPHA8_EXT: - return default_srgba_format( screen, target, sample_count, bindings, - geom_flags ); - - case GL_COMPRESSED_SRGB_EXT: - case GL_COMPRESSED_SRGB_S3TC_DXT1_EXT: - if (screen->is_format_supported(screen, PIPE_FORMAT_DXT1_SRGB, target, - sample_count, bindings, geom_flags)) - return PIPE_FORMAT_DXT1_SRGB; - return default_srgba_format( screen, target, sample_count, bindings, - geom_flags ); - - case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT1_EXT: - return PIPE_FORMAT_DXT1_SRGBA; - - case GL_COMPRESSED_SRGB_ALPHA_EXT: - case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT3_EXT: - if (screen->is_format_supported(screen, PIPE_FORMAT_DXT3_SRGBA, target, - sample_count, bindings, geom_flags)) - return PIPE_FORMAT_DXT3_SRGBA; - return default_srgba_format( screen, target, sample_count, bindings, - geom_flags ); - - case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT: - return PIPE_FORMAT_DXT5_SRGBA; - - case GL_SLUMINANCE_ALPHA_EXT: - case GL_SLUMINANCE8_ALPHA8_EXT: - case GL_COMPRESSED_SLUMINANCE_EXT: - case GL_COMPRESSED_SLUMINANCE_ALPHA_EXT: - if (screen->is_format_supported(screen, PIPE_FORMAT_L8A8_SRGB, target, - sample_count, bindings, geom_flags)) - return PIPE_FORMAT_L8A8_SRGB; - return default_srgba_format( screen, target, sample_count, bindings, - geom_flags ); - - case GL_SLUMINANCE_EXT: - case GL_SLUMINANCE8_EXT: - if (screen->is_format_supported(screen, PIPE_FORMAT_L8_SRGB, target, - sample_count, bindings, geom_flags)) - return PIPE_FORMAT_L8_SRGB; - return default_srgba_format( screen, target, sample_count, bindings, - geom_flags ); - - case GL_RED: - case GL_R8: - if (screen->is_format_supported(screen, PIPE_FORMAT_R8_UNORM, target, - sample_count, bindings, geom_flags)) - return PIPE_FORMAT_R8_UNORM; - return PIPE_FORMAT_NONE; - case GL_RG: - case GL_RG8: - if (screen->is_format_supported(screen, PIPE_FORMAT_R8G8_UNORM, target, - sample_count, bindings, geom_flags)) - return PIPE_FORMAT_R8G8_UNORM; - return PIPE_FORMAT_NONE; - - case GL_R16: - if (screen->is_format_supported(screen, PIPE_FORMAT_R16_UNORM, target, - sample_count, bindings, geom_flags)) - return PIPE_FORMAT_R16_UNORM; - return PIPE_FORMAT_NONE; - - case GL_RG16: - if (screen->is_format_supported(screen, PIPE_FORMAT_R16G16_UNORM, target, - sample_count, bindings, geom_flags)) - return PIPE_FORMAT_R16G16_UNORM; - return PIPE_FORMAT_NONE; - - case GL_COMPRESSED_RED_RGTC1: - if (screen->is_format_supported(screen, PIPE_FORMAT_RGTC1_UNORM, target, - sample_count, bindings, geom_flags)) - return PIPE_FORMAT_RGTC1_UNORM; - return PIPE_FORMAT_NONE; - - case GL_COMPRESSED_SIGNED_RED_RGTC1: - if (screen->is_format_supported(screen, PIPE_FORMAT_RGTC1_SNORM, target, - sample_count, bindings, geom_flags)) - return PIPE_FORMAT_RGTC1_SNORM; - return PIPE_FORMAT_NONE; - - case GL_COMPRESSED_RG_RGTC2: - if (screen->is_format_supported(screen, PIPE_FORMAT_RGTC2_UNORM, target, - sample_count, bindings, geom_flags)) - return PIPE_FORMAT_RGTC2_UNORM; - return PIPE_FORMAT_NONE; - - case GL_COMPRESSED_SIGNED_RG_RGTC2: - if (screen->is_format_supported(screen, PIPE_FORMAT_RGTC2_SNORM, target, - sample_count, bindings, geom_flags)) - return PIPE_FORMAT_RGTC2_SNORM; - return PIPE_FORMAT_NONE; - - /* signed/unsigned integer formats. - * XXX Mesa only has formats for RGBA signed/unsigned integer formats. - * If/when new formats are added this code should be updated. - */ - case GL_RED_INTEGER_EXT: - case GL_GREEN_INTEGER_EXT: - case GL_BLUE_INTEGER_EXT: - case GL_ALPHA_INTEGER_EXT: - case GL_RGB_INTEGER_EXT: - case GL_RGBA_INTEGER_EXT: - case GL_BGR_INTEGER_EXT: - case GL_BGRA_INTEGER_EXT: - case GL_LUMINANCE_INTEGER_EXT: - case GL_LUMINANCE_ALPHA_INTEGER_EXT: - /* fall-through */ - case GL_RGBA8I_EXT: - case GL_RGB8I_EXT: - case GL_ALPHA8I_EXT: - case GL_INTENSITY8I_EXT: - case GL_LUMINANCE8I_EXT: - case GL_LUMINANCE_ALPHA8I_EXT: - if (screen->is_format_supported(screen, PIPE_FORMAT_R8G8B8A8_SSCALED, - target, - sample_count, bindings, geom_flags)) - return PIPE_FORMAT_R8G8B8A8_SSCALED; - return PIPE_FORMAT_NONE; - case GL_RGBA16I_EXT: - case GL_RGB16I_EXT: - case GL_ALPHA16I_EXT: - case GL_INTENSITY16I_EXT: - case GL_LUMINANCE16I_EXT: - case GL_LUMINANCE_ALPHA16I_EXT: - if (screen->is_format_supported(screen, PIPE_FORMAT_R16G16B16A16_SSCALED, - target, - sample_count, bindings, geom_flags)) - return PIPE_FORMAT_R16G16B16A16_SSCALED; - return PIPE_FORMAT_NONE; - case GL_RGBA32I_EXT: - case GL_RGB32I_EXT: - case GL_ALPHA32I_EXT: - case GL_INTENSITY32I_EXT: - case GL_LUMINANCE32I_EXT: - case GL_LUMINANCE_ALPHA32I_EXT: - /* xxx */ - if (screen->is_format_supported(screen, PIPE_FORMAT_R32G32B32A32_SSCALED, - target, - sample_count, bindings, geom_flags)) - return PIPE_FORMAT_R32G32B32A32_SSCALED; - return PIPE_FORMAT_NONE; - - case GL_RGBA8UI_EXT: - case GL_RGB8UI_EXT: - case GL_ALPHA8UI_EXT: - case GL_INTENSITY8UI_EXT: - case GL_LUMINANCE8UI_EXT: - case GL_LUMINANCE_ALPHA8UI_EXT: - if (screen->is_format_supported(screen, PIPE_FORMAT_R8G8B8A8_USCALED, - target, - sample_count, bindings, geom_flags)) - return PIPE_FORMAT_R8G8B8A8_USCALED; - return PIPE_FORMAT_NONE; - - case GL_RGBA16UI_EXT: - case GL_RGB16UI_EXT: - case GL_ALPHA16UI_EXT: - case GL_INTENSITY16UI_EXT: - case GL_LUMINANCE16UI_EXT: - case GL_LUMINANCE_ALPHA16UI_EXT: - if (screen->is_format_supported(screen, PIPE_FORMAT_R16G16B16A16_USCALED, - target, - sample_count, bindings, geom_flags)) - return PIPE_FORMAT_R16G16B16A16_USCALED; - return PIPE_FORMAT_NONE; - - case GL_RGBA32UI_EXT: - case GL_RGB32UI_EXT: - case GL_ALPHA32UI_EXT: - case GL_INTENSITY32UI_EXT: - case GL_LUMINANCE32UI_EXT: - case GL_LUMINANCE_ALPHA32UI_EXT: - if (screen->is_format_supported(screen, PIPE_FORMAT_R32G32B32A32_USCALED, - target, - sample_count, bindings, geom_flags)) - return PIPE_FORMAT_R32G32B32A32_USCALED; - return PIPE_FORMAT_NONE; - - default: - return PIPE_FORMAT_NONE; - } -} - - -/** - * Called by FBO code to choose a PIPE_FORMAT_ for drawing surfaces. - */ -enum pipe_format -st_choose_renderbuffer_format(struct pipe_screen *screen, - GLenum internalFormat, unsigned sample_count) -{ - uint usage; - if (_mesa_is_depth_or_stencil_format(internalFormat)) - usage = PIPE_BIND_DEPTH_STENCIL; - else - usage = PIPE_BIND_RENDER_TARGET; - return st_choose_format(screen, internalFormat, PIPE_TEXTURE_2D, - sample_count, usage); -} - - -/** - * Called via ctx->Driver.chooseTextureFormat(). - */ -gl_format -st_ChooseTextureFormat_renderable(struct gl_context *ctx, GLint internalFormat, - GLenum format, GLenum type, GLboolean renderable) -{ - struct pipe_screen *screen = st_context(ctx)->pipe->screen; - enum pipe_format pFormat; - uint bindings; - - (void) format; - (void) type; - - /* GL textures may wind up being render targets, but we don't know - * that in advance. Specify potential render target flags now. - */ - bindings = PIPE_BIND_SAMPLER_VIEW; - if (renderable == GL_TRUE) { - if (_mesa_is_depth_format(internalFormat) || - _mesa_is_depth_or_stencil_format(internalFormat)) - bindings |= PIPE_BIND_DEPTH_STENCIL; - else - bindings |= PIPE_BIND_RENDER_TARGET; - } - - pFormat = st_choose_format(screen, internalFormat, - PIPE_TEXTURE_2D, 0, bindings); - - if (pFormat == PIPE_FORMAT_NONE) { - /* try choosing format again, this time without render target bindings */ - pFormat = st_choose_format(screen, internalFormat, - PIPE_TEXTURE_2D, 0, PIPE_BIND_SAMPLER_VIEW); - } - - if (pFormat == PIPE_FORMAT_NONE) { - /* no luck at all */ - return MESA_FORMAT_NONE; - } - - return st_pipe_format_to_mesa_format(pFormat); -} - -gl_format -st_ChooseTextureFormat(struct gl_context *ctx, GLint internalFormat, - GLenum format, GLenum type) -{ - boolean want_renderable = - internalFormat == 3 || internalFormat == 4 || - internalFormat == GL_RGB || internalFormat == GL_RGBA || - internalFormat == GL_RGB8 || internalFormat == GL_RGBA8 || - internalFormat == GL_BGRA; - - return st_ChooseTextureFormat_renderable(ctx, internalFormat, - format, type, want_renderable); -} - -/** - * Test if a gallium format is equivalent to a GL format/type. - */ -GLboolean -st_equal_formats(enum pipe_format pFormat, GLenum format, GLenum type) -{ - switch (pFormat) { - case PIPE_FORMAT_A8B8G8R8_UNORM: - return format == GL_RGBA && type == GL_UNSIGNED_BYTE; - case PIPE_FORMAT_A8R8G8B8_UNORM: - return format == GL_BGRA && type == GL_UNSIGNED_BYTE; - case PIPE_FORMAT_B5G6R5_UNORM: - return format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5; - /* XXX more combos... */ - default: - return GL_FALSE; - } -} - -GLboolean -st_sampler_compat_formats(enum pipe_format format1, enum pipe_format format2) -{ - if (format1 == format2) - return GL_TRUE; - - if (format1 == PIPE_FORMAT_B8G8R8A8_UNORM && - format2 == PIPE_FORMAT_B8G8R8X8_UNORM) - return GL_TRUE; - - if (format1 == PIPE_FORMAT_B8G8R8X8_UNORM && - format2 == PIPE_FORMAT_B8G8R8A8_UNORM) - return GL_TRUE; - - if (format1 == PIPE_FORMAT_A8B8G8R8_UNORM && - format2 == PIPE_FORMAT_X8B8G8R8_UNORM) - return GL_TRUE; - - if (format1 == PIPE_FORMAT_X8B8G8R8_UNORM && - format2 == PIPE_FORMAT_A8B8G8R8_UNORM) - return GL_TRUE; - - if (format1 == PIPE_FORMAT_A8R8G8B8_UNORM && - format2 == PIPE_FORMAT_X8R8G8B8_UNORM) - return GL_TRUE; - - if (format1 == PIPE_FORMAT_X8R8G8B8_UNORM && - format2 == PIPE_FORMAT_A8R8G8B8_UNORM) - return GL_TRUE; - - return GL_FALSE; -} +/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * Copyright (c) 2008-2010 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+/**
+ * Mesa / Gallium format conversion and format selection code.
+ * \author Brian Paul
+ */
+
+#include "main/imports.h"
+#include "main/context.h"
+#include "main/texstore.h"
+#include "main/image.h"
+#include "main/macros.h"
+#include "main/mfeatures.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_screen.h"
+#include "util/u_format.h"
+#include "st_context.h"
+#include "st_format.h"
+
+
+static GLuint
+format_max_bits(enum pipe_format format)
+{
+ GLuint size = util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 0);
+
+ size = MAX2(size, util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 1));
+ size = MAX2(size, util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 2));
+ size = MAX2(size, util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 3));
+ size = MAX2(size, util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_ZS, 0));
+ size = MAX2(size, util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_ZS, 1));
+ return size;
+}
+
+
+/**
+ * Return basic GL datatype for the given gallium format.
+ */
+GLenum
+st_format_datatype(enum pipe_format format)
+{
+ const struct util_format_description *desc;
+
+ desc = util_format_description(format);
+ assert(desc);
+
+ if (desc->layout == UTIL_FORMAT_LAYOUT_PLAIN) {
+ if (format == PIPE_FORMAT_B5G5R5A1_UNORM ||
+ format == PIPE_FORMAT_B5G6R5_UNORM) {
+ return GL_UNSIGNED_SHORT;
+ }
+ else if (format == PIPE_FORMAT_Z24_UNORM_S8_USCALED ||
+ format == PIPE_FORMAT_S8_USCALED_Z24_UNORM ||
+ format == PIPE_FORMAT_Z24X8_UNORM ||
+ format == PIPE_FORMAT_X8Z24_UNORM) {
+ return GL_UNSIGNED_INT_24_8;
+ }
+ else {
+ const GLuint size = format_max_bits(format);
+ if (size == 8) {
+ if (desc->channel[0].type == UTIL_FORMAT_TYPE_UNSIGNED)
+ return GL_UNSIGNED_BYTE;
+ else
+ return GL_BYTE;
+ }
+ else if (size == 16) {
+ if (desc->channel[0].type == UTIL_FORMAT_TYPE_UNSIGNED)
+ return GL_UNSIGNED_SHORT;
+ else
+ return GL_SHORT;
+ }
+ else {
+ assert( size <= 32 );
+ if (desc->channel[0].type == UTIL_FORMAT_TYPE_UNSIGNED)
+ return GL_UNSIGNED_INT;
+ else
+ return GL_INT;
+ }
+ }
+ }
+ else if (format == PIPE_FORMAT_UYVY) {
+ return GL_UNSIGNED_SHORT;
+ }
+ else if (format == PIPE_FORMAT_YUYV) {
+ return GL_UNSIGNED_SHORT;
+ }
+ else {
+ /* probably a compressed format, unsupported anyway */
+ return GL_NONE;
+ }
+}
+
+
+/**
+ * Translate Mesa format to Gallium format.
+ */
+enum pipe_format
+st_mesa_format_to_pipe_format(gl_format mesaFormat)
+{
+ switch (mesaFormat) {
+ case MESA_FORMAT_RGBA8888:
+ return PIPE_FORMAT_A8B8G8R8_UNORM;
+ case MESA_FORMAT_RGBA8888_REV:
+ return PIPE_FORMAT_R8G8B8A8_UNORM;
+ case MESA_FORMAT_ARGB8888:
+ return PIPE_FORMAT_B8G8R8A8_UNORM;
+ case MESA_FORMAT_ARGB8888_REV:
+ return PIPE_FORMAT_A8R8G8B8_UNORM;
+ case MESA_FORMAT_XRGB8888:
+ return PIPE_FORMAT_B8G8R8X8_UNORM;
+ case MESA_FORMAT_XRGB8888_REV:
+ return PIPE_FORMAT_X8R8G8B8_UNORM;
+ case MESA_FORMAT_ARGB1555:
+ return PIPE_FORMAT_B5G5R5A1_UNORM;
+ case MESA_FORMAT_ARGB4444:
+ return PIPE_FORMAT_B4G4R4A4_UNORM;
+ case MESA_FORMAT_RGB565:
+ return PIPE_FORMAT_B5G6R5_UNORM;
+ case MESA_FORMAT_RGB332:
+ return PIPE_FORMAT_B2G3R3_UNORM;
+ case MESA_FORMAT_ARGB2101010:
+ return PIPE_FORMAT_B10G10R10A2_UNORM;
+ case MESA_FORMAT_AL44:
+ return PIPE_FORMAT_L4A4_UNORM;
+ case MESA_FORMAT_AL88:
+ return PIPE_FORMAT_L8A8_UNORM;
+ case MESA_FORMAT_AL1616:
+ return PIPE_FORMAT_L16A16_UNORM;
+ case MESA_FORMAT_A8:
+ return PIPE_FORMAT_A8_UNORM;
+ case MESA_FORMAT_A16:
+ return PIPE_FORMAT_A16_UNORM;
+ case MESA_FORMAT_L8:
+ return PIPE_FORMAT_L8_UNORM;
+ case MESA_FORMAT_L16:
+ return PIPE_FORMAT_L16_UNORM;
+ case MESA_FORMAT_I8:
+ return PIPE_FORMAT_I8_UNORM;
+ case MESA_FORMAT_I16:
+ return PIPE_FORMAT_I16_UNORM;
+ case MESA_FORMAT_Z16:
+ return PIPE_FORMAT_Z16_UNORM;
+ case MESA_FORMAT_Z32:
+ return PIPE_FORMAT_Z32_UNORM;
+ case MESA_FORMAT_Z24_S8:
+ return PIPE_FORMAT_S8_USCALED_Z24_UNORM;
+ case MESA_FORMAT_S8_Z24:
+ return PIPE_FORMAT_Z24_UNORM_S8_USCALED;
+ case MESA_FORMAT_Z24_X8:
+ return PIPE_FORMAT_X8Z24_UNORM;
+ case MESA_FORMAT_X8_Z24:
+ return PIPE_FORMAT_Z24X8_UNORM;
+ case MESA_FORMAT_S8:
+ return PIPE_FORMAT_S8_USCALED;
+ case MESA_FORMAT_YCBCR:
+ return PIPE_FORMAT_UYVY;
+#if FEATURE_texture_s3tc
+ case MESA_FORMAT_RGB_DXT1:
+ return PIPE_FORMAT_DXT1_RGB;
+ case MESA_FORMAT_RGBA_DXT1:
+ return PIPE_FORMAT_DXT1_RGBA;
+ case MESA_FORMAT_RGBA_DXT3:
+ return PIPE_FORMAT_DXT3_RGBA;
+ case MESA_FORMAT_RGBA_DXT5:
+ return PIPE_FORMAT_DXT5_RGBA;
+#if FEATURE_EXT_texture_sRGB
+ case MESA_FORMAT_SRGB_DXT1:
+ return PIPE_FORMAT_DXT1_SRGB;
+ case MESA_FORMAT_SRGBA_DXT1:
+ return PIPE_FORMAT_DXT1_SRGBA;
+ case MESA_FORMAT_SRGBA_DXT3:
+ return PIPE_FORMAT_DXT3_SRGBA;
+ case MESA_FORMAT_SRGBA_DXT5:
+ return PIPE_FORMAT_DXT5_SRGBA;
+#endif
+#endif
+#if FEATURE_EXT_texture_sRGB
+ case MESA_FORMAT_SLA8:
+ return PIPE_FORMAT_L8A8_SRGB;
+ case MESA_FORMAT_SL8:
+ return PIPE_FORMAT_L8_SRGB;
+ case MESA_FORMAT_SRGB8:
+ return PIPE_FORMAT_R8G8B8_SRGB;
+ case MESA_FORMAT_SRGBA8:
+ return PIPE_FORMAT_A8B8G8R8_SRGB;
+ case MESA_FORMAT_SARGB8:
+ return PIPE_FORMAT_B8G8R8A8_SRGB;
+#endif
+ case MESA_FORMAT_R8:
+ return PIPE_FORMAT_R8_UNORM;
+ case MESA_FORMAT_R16:
+ return PIPE_FORMAT_R16_UNORM;
+ case MESA_FORMAT_RG88:
+ return PIPE_FORMAT_R8G8_UNORM;
+ case MESA_FORMAT_RG1616:
+ return PIPE_FORMAT_R16G16_UNORM;
+ case MESA_FORMAT_RGBA_16:
+ return PIPE_FORMAT_R16G16B16A16_UNORM;
+
+ /* signed int formats */
+ case MESA_FORMAT_RGBA_INT8:
+ return PIPE_FORMAT_R8G8B8A8_SSCALED;
+ case MESA_FORMAT_RGBA_INT16:
+ return PIPE_FORMAT_R16G16B16A16_SSCALED;
+ case MESA_FORMAT_RGBA_INT32:
+ return PIPE_FORMAT_R32G32B32A32_SSCALED;
+
+ /* unsigned int formats */
+ case MESA_FORMAT_RGBA_UINT8:
+ return PIPE_FORMAT_R8G8B8A8_USCALED;
+ case MESA_FORMAT_RGBA_UINT16:
+ return PIPE_FORMAT_R16G16B16A16_USCALED;
+ case MESA_FORMAT_RGBA_UINT32:
+ return PIPE_FORMAT_R32G32B32A32_USCALED;
+
+ default:
+ assert(0);
+ return PIPE_FORMAT_NONE;
+ }
+}
+
+
+/**
+ * Translate Gallium format to Mesa format.
+ */
+gl_format
+st_pipe_format_to_mesa_format(enum pipe_format format)
+{
+ switch (format) {
+ case PIPE_FORMAT_A8B8G8R8_UNORM:
+ return MESA_FORMAT_RGBA8888;
+ case PIPE_FORMAT_R8G8B8A8_UNORM:
+ return MESA_FORMAT_RGBA8888_REV;
+ case PIPE_FORMAT_B8G8R8A8_UNORM:
+ return MESA_FORMAT_ARGB8888;
+ case PIPE_FORMAT_A8R8G8B8_UNORM:
+ return MESA_FORMAT_ARGB8888_REV;
+ case PIPE_FORMAT_B8G8R8X8_UNORM:
+ return MESA_FORMAT_XRGB8888;
+ case PIPE_FORMAT_X8R8G8B8_UNORM:
+ return MESA_FORMAT_XRGB8888_REV;
+ case PIPE_FORMAT_B5G5R5A1_UNORM:
+ return MESA_FORMAT_ARGB1555;
+ case PIPE_FORMAT_B4G4R4A4_UNORM:
+ return MESA_FORMAT_ARGB4444;
+ case PIPE_FORMAT_B5G6R5_UNORM:
+ return MESA_FORMAT_RGB565;
+ case PIPE_FORMAT_B2G3R3_UNORM:
+ return MESA_FORMAT_RGB332;
+ case PIPE_FORMAT_B10G10R10A2_UNORM:
+ return MESA_FORMAT_ARGB2101010;
+ case PIPE_FORMAT_L4A4_UNORM:
+ return MESA_FORMAT_AL44;
+ case PIPE_FORMAT_L8A8_UNORM:
+ return MESA_FORMAT_AL88;
+ case PIPE_FORMAT_L16A16_UNORM:
+ return MESA_FORMAT_AL1616;
+ case PIPE_FORMAT_A8_UNORM:
+ return MESA_FORMAT_A8;
+ case PIPE_FORMAT_A16_UNORM:
+ return MESA_FORMAT_A16;
+ case PIPE_FORMAT_L8_UNORM:
+ return MESA_FORMAT_L8;
+ case PIPE_FORMAT_L16_UNORM:
+ return MESA_FORMAT_L16;
+ case PIPE_FORMAT_I8_UNORM:
+ return MESA_FORMAT_I8;
+ case PIPE_FORMAT_I16_UNORM:
+ return MESA_FORMAT_I16;
+ case PIPE_FORMAT_S8_USCALED:
+ return MESA_FORMAT_S8;
+
+ case PIPE_FORMAT_R16G16B16A16_UNORM:
+ return MESA_FORMAT_RGBA_16;
+ case PIPE_FORMAT_R16G16B16A16_SNORM:
+ return MESA_FORMAT_SIGNED_RGBA_16;
+
+ case PIPE_FORMAT_Z16_UNORM:
+ return MESA_FORMAT_Z16;
+ case PIPE_FORMAT_Z32_UNORM:
+ return MESA_FORMAT_Z32;
+ case PIPE_FORMAT_S8_USCALED_Z24_UNORM:
+ return MESA_FORMAT_Z24_S8;
+ case PIPE_FORMAT_X8Z24_UNORM:
+ return MESA_FORMAT_Z24_X8;
+ case PIPE_FORMAT_Z24X8_UNORM:
+ return MESA_FORMAT_X8_Z24;
+ case PIPE_FORMAT_Z24_UNORM_S8_USCALED:
+ return MESA_FORMAT_S8_Z24;
+
+ case PIPE_FORMAT_UYVY:
+ return MESA_FORMAT_YCBCR;
+ case PIPE_FORMAT_YUYV:
+ return MESA_FORMAT_YCBCR_REV;
+
+#if FEATURE_texture_s3tc
+ case PIPE_FORMAT_DXT1_RGB:
+ return MESA_FORMAT_RGB_DXT1;
+ case PIPE_FORMAT_DXT1_RGBA:
+ return MESA_FORMAT_RGBA_DXT1;
+ case PIPE_FORMAT_DXT3_RGBA:
+ return MESA_FORMAT_RGBA_DXT3;
+ case PIPE_FORMAT_DXT5_RGBA:
+ return MESA_FORMAT_RGBA_DXT5;
+#if FEATURE_EXT_texture_sRGB
+ case PIPE_FORMAT_DXT1_SRGB:
+ return MESA_FORMAT_SRGB_DXT1;
+ case PIPE_FORMAT_DXT1_SRGBA:
+ return MESA_FORMAT_SRGBA_DXT1;
+ case PIPE_FORMAT_DXT3_SRGBA:
+ return MESA_FORMAT_SRGBA_DXT3;
+ case PIPE_FORMAT_DXT5_SRGBA:
+ return MESA_FORMAT_SRGBA_DXT5;
+#endif
+#endif
+
+#if FEATURE_EXT_texture_sRGB
+ case PIPE_FORMAT_L8A8_SRGB:
+ return MESA_FORMAT_SLA8;
+ case PIPE_FORMAT_L8_SRGB:
+ return MESA_FORMAT_SL8;
+ case PIPE_FORMAT_R8G8B8_SRGB:
+ return MESA_FORMAT_SRGB8;
+ case PIPE_FORMAT_A8B8G8R8_SRGB:
+ return MESA_FORMAT_SRGBA8;
+ case PIPE_FORMAT_B8G8R8A8_SRGB:
+ return MESA_FORMAT_SARGB8;
+#endif
+
+ case PIPE_FORMAT_R8_UNORM:
+ return MESA_FORMAT_R8;
+ case PIPE_FORMAT_R16_UNORM:
+ return MESA_FORMAT_R16;
+ case PIPE_FORMAT_R8G8_UNORM:
+ return MESA_FORMAT_RG88;
+ case PIPE_FORMAT_R16G16_UNORM:
+ return MESA_FORMAT_RG1616;
+
+ /* signed int formats */
+ case PIPE_FORMAT_R8G8B8A8_SSCALED:
+ return MESA_FORMAT_RGBA_INT8;
+ case PIPE_FORMAT_R16G16B16A16_SSCALED:
+ return MESA_FORMAT_RGBA_INT16;
+ case PIPE_FORMAT_R32G32B32A32_SSCALED:
+ return MESA_FORMAT_RGBA_INT32;
+
+ /* unsigned int formats */
+ case PIPE_FORMAT_R8G8B8A8_USCALED:
+ return MESA_FORMAT_RGBA_UINT8;
+ case PIPE_FORMAT_R16G16B16A16_USCALED:
+ return MESA_FORMAT_RGBA_UINT16;
+ case PIPE_FORMAT_R32G32B32A32_USCALED:
+ return MESA_FORMAT_RGBA_UINT32;
+
+ default:
+ assert(0);
+ return MESA_FORMAT_NONE;
+ }
+}
+
+
+/**
+ * Return first supported format from the given list.
+ */
+static enum pipe_format
+find_supported_format(struct pipe_screen *screen,
+ const enum pipe_format formats[],
+ uint num_formats,
+ enum pipe_texture_target target,
+ unsigned sample_count,
+ unsigned tex_usage,
+ unsigned geom_flags)
+{
+ uint i;
+ for (i = 0; i < num_formats; i++) {
+ if (screen->is_format_supported(screen, formats[i], target,
+ sample_count, tex_usage, geom_flags)) {
+ return formats[i];
+ }
+ }
+ return PIPE_FORMAT_NONE;
+}
+
+
+/**
+ * Find an RGBA format supported by the context/winsys.
+ */
+static enum pipe_format
+default_rgba_format(struct pipe_screen *screen,
+ enum pipe_texture_target target,
+ unsigned sample_count,
+ unsigned tex_usage,
+ unsigned geom_flags)
+{
+ static const enum pipe_format colorFormats[] = {
+ PIPE_FORMAT_B8G8R8A8_UNORM,
+ PIPE_FORMAT_A8R8G8B8_UNORM,
+ PIPE_FORMAT_A8B8G8R8_UNORM,
+ PIPE_FORMAT_B5G6R5_UNORM
+ };
+ return find_supported_format(screen, colorFormats, Elements(colorFormats),
+ target, sample_count, tex_usage, geom_flags);
+}
+
+
+/**
+ * Find an RGB format supported by the context/winsys.
+ */
+static enum pipe_format
+default_rgb_format(struct pipe_screen *screen,
+ enum pipe_texture_target target,
+ unsigned sample_count,
+ unsigned tex_usage,
+ unsigned geom_flags)
+{
+ static const enum pipe_format colorFormats[] = {
+ PIPE_FORMAT_B8G8R8X8_UNORM,
+ PIPE_FORMAT_X8R8G8B8_UNORM,
+ PIPE_FORMAT_X8B8G8R8_UNORM,
+ PIPE_FORMAT_B8G8R8A8_UNORM,
+ PIPE_FORMAT_A8R8G8B8_UNORM,
+ PIPE_FORMAT_A8B8G8R8_UNORM,
+ PIPE_FORMAT_B5G6R5_UNORM
+ };
+ return find_supported_format(screen, colorFormats, Elements(colorFormats),
+ target, sample_count, tex_usage, geom_flags);
+}
+
+/**
+ * Find an sRGBA format supported by the context/winsys.
+ */
+static enum pipe_format
+default_srgba_format(struct pipe_screen *screen,
+ enum pipe_texture_target target,
+ unsigned sample_count,
+ unsigned tex_usage,
+ unsigned geom_flags)
+{
+ static const enum pipe_format colorFormats[] = {
+ PIPE_FORMAT_B8G8R8A8_SRGB,
+ PIPE_FORMAT_A8R8G8B8_SRGB,
+ PIPE_FORMAT_A8B8G8R8_SRGB,
+ };
+ return find_supported_format(screen, colorFormats, Elements(colorFormats),
+ target, sample_count, tex_usage, geom_flags);
+}
+
+
+/**
+ * Given an OpenGL internalFormat value for a texture or surface, return
+ * the best matching PIPE_FORMAT_x, or PIPE_FORMAT_NONE if there's no match.
+ * This is called during glTexImage2D, for example.
+ *
+ * The bindings parameter typically has PIPE_BIND_SAMPLER_VIEW set, plus
+ * either PIPE_BINDING_RENDER_TARGET or PIPE_BINDING_DEPTH_STENCIL if
+ * we want render-to-texture ability.
+ *
+ * \param internalFormat the user value passed to glTexImage2D
+ * \param target one of PIPE_TEXTURE_x
+ * \param bindings bitmask of PIPE_BIND_x flags.
+ */
+enum pipe_format
+st_choose_format(struct pipe_screen *screen, GLenum internalFormat,
+ enum pipe_texture_target target, unsigned sample_count,
+ unsigned bindings)
+{
+ unsigned geom_flags = 0; /* we don't care about POT vs. NPOT here, yet */
+
+ switch (internalFormat) {
+ case GL_RGB10:
+ case GL_RGB10_A2:
+ if (screen->is_format_supported( screen, PIPE_FORMAT_B10G10R10A2_UNORM,
+ target, sample_count, bindings,
+ geom_flags ))
+ return PIPE_FORMAT_B10G10R10A2_UNORM;
+ /* Pass through. */
+ case 4:
+ case GL_RGBA:
+ case GL_RGBA8:
+ return default_rgba_format( screen, target, sample_count, bindings,
+ geom_flags );
+
+ case GL_BGRA:
+ if (screen->is_format_supported( screen, PIPE_FORMAT_B8G8R8A8_UNORM,
+ target, sample_count, bindings,
+ geom_flags ))
+ return PIPE_FORMAT_B8G8R8A8_UNORM;
+ return default_rgba_format( screen, target, sample_count, bindings,
+ geom_flags );
+
+ case 3:
+ case GL_RGB:
+ case GL_RGB8:
+ return default_rgb_format( screen, target, sample_count, bindings,
+ geom_flags );
+
+ case GL_RGB12:
+ case GL_RGB16:
+ case GL_RGBA12:
+ case GL_RGBA16:
+ if (screen->is_format_supported( screen, PIPE_FORMAT_R16G16B16A16_UNORM,
+ target, sample_count, bindings,
+ geom_flags ))
+ return PIPE_FORMAT_R16G16B16A16_UNORM;
+ return default_rgba_format( screen, target, sample_count, bindings,
+ geom_flags );
+
+ case GL_RGBA4:
+ case GL_RGBA2:
+ if (screen->is_format_supported( screen, PIPE_FORMAT_B4G4R4A4_UNORM,
+ target, sample_count, bindings,
+ geom_flags ))
+ return PIPE_FORMAT_B4G4R4A4_UNORM;
+ return default_rgba_format( screen, target, sample_count, bindings,
+ geom_flags );
+
+ case GL_RGB5_A1:
+ if (screen->is_format_supported( screen, PIPE_FORMAT_B5G5R5A1_UNORM,
+ target, sample_count, bindings,
+ geom_flags ))
+ return PIPE_FORMAT_B5G5R5A1_UNORM;
+ return default_rgba_format( screen, target, sample_count, bindings,
+ geom_flags );
+
+ case GL_R3_G3_B2:
+ if (screen->is_format_supported( screen, PIPE_FORMAT_B2G3R3_UNORM,
+ target, sample_count, bindings,
+ geom_flags ))
+ return PIPE_FORMAT_B2G3R3_UNORM;
+ /* Pass through. */
+ case GL_RGB5:
+ case GL_RGB4:
+ if (screen->is_format_supported( screen, PIPE_FORMAT_B5G6R5_UNORM,
+ target, sample_count, bindings,
+ geom_flags ))
+ return PIPE_FORMAT_B5G6R5_UNORM;
+ if (screen->is_format_supported( screen, PIPE_FORMAT_B5G5R5A1_UNORM,
+ target, sample_count, bindings,
+ geom_flags ))
+ return PIPE_FORMAT_B5G5R5A1_UNORM;
+ return default_rgba_format( screen, target, sample_count, bindings,
+ geom_flags );
+
+ case GL_ALPHA12:
+ case GL_ALPHA16:
+ if (screen->is_format_supported( screen, PIPE_FORMAT_A16_UNORM, target,
+ sample_count, bindings, geom_flags ))
+ return PIPE_FORMAT_A16_UNORM;
+ /* Pass through. */
+ case GL_ALPHA:
+ case GL_ALPHA4:
+ case GL_ALPHA8:
+ case GL_COMPRESSED_ALPHA:
+ if (screen->is_format_supported( screen, PIPE_FORMAT_A8_UNORM, target,
+ sample_count, bindings, geom_flags ))
+ return PIPE_FORMAT_A8_UNORM;
+ return default_rgba_format( screen, target, sample_count, bindings,
+ geom_flags );
+
+ case GL_LUMINANCE12:
+ case GL_LUMINANCE16:
+ if (screen->is_format_supported( screen, PIPE_FORMAT_L16_UNORM, target,
+ sample_count, bindings, geom_flags ))
+ return PIPE_FORMAT_L16_UNORM;
+ /* Pass through. */
+ case 1:
+ case GL_LUMINANCE:
+ case GL_LUMINANCE4:
+ case GL_LUMINANCE8:
+ case GL_COMPRESSED_LUMINANCE:
+ if (screen->is_format_supported( screen, PIPE_FORMAT_L8_UNORM, target,
+ sample_count, bindings, geom_flags ))
+ return PIPE_FORMAT_L8_UNORM;
+ return default_rgba_format( screen, target, sample_count, bindings,
+ geom_flags );
+
+ case GL_LUMINANCE12_ALPHA4:
+ case GL_LUMINANCE12_ALPHA12:
+ case GL_LUMINANCE16_ALPHA16:
+ if (screen->is_format_supported( screen, PIPE_FORMAT_L16A16_UNORM, target,
+ sample_count, bindings, geom_flags ))
+ return PIPE_FORMAT_L16A16_UNORM;
+ /* Pass through. */
+ case 2:
+ case GL_LUMINANCE_ALPHA:
+ case GL_LUMINANCE6_ALPHA2:
+ case GL_LUMINANCE8_ALPHA8:
+ case GL_COMPRESSED_LUMINANCE_ALPHA:
+ if (screen->is_format_supported( screen, PIPE_FORMAT_L8A8_UNORM, target,
+ sample_count, bindings, geom_flags ))
+ return PIPE_FORMAT_L8A8_UNORM;
+ return default_rgba_format( screen, target, sample_count, bindings,
+ geom_flags );
+
+ case GL_LUMINANCE4_ALPHA4:
+ if (screen->is_format_supported( screen, PIPE_FORMAT_L4A4_UNORM, target,
+ sample_count, bindings, geom_flags ))
+ return PIPE_FORMAT_L4A4_UNORM;
+ if (screen->is_format_supported( screen, PIPE_FORMAT_L8A8_UNORM, target,
+ sample_count, bindings, geom_flags ))
+ return PIPE_FORMAT_L8A8_UNORM;
+ return default_rgba_format( screen, target, sample_count, bindings,
+ geom_flags );
+
+ case GL_INTENSITY12:
+ case GL_INTENSITY16:
+ if (screen->is_format_supported( screen, PIPE_FORMAT_I16_UNORM, target,
+ sample_count, bindings, geom_flags ))
+ return PIPE_FORMAT_I16_UNORM;
+ /* Pass through. */
+ case GL_INTENSITY:
+ case GL_INTENSITY4:
+ case GL_INTENSITY8:
+ case GL_COMPRESSED_INTENSITY:
+ if (screen->is_format_supported( screen, PIPE_FORMAT_I8_UNORM, target,
+ sample_count, bindings, geom_flags ))
+ return PIPE_FORMAT_I8_UNORM;
+ return default_rgba_format( screen, target, sample_count, bindings,
+ geom_flags );
+
+ case GL_YCBCR_MESA:
+ if (screen->is_format_supported(screen, PIPE_FORMAT_UYVY, target,
+ sample_count, bindings, geom_flags)) {
+ return PIPE_FORMAT_UYVY;
+ }
+ if (screen->is_format_supported(screen, PIPE_FORMAT_YUYV, target,
+ sample_count, bindings, geom_flags)) {
+ return PIPE_FORMAT_YUYV;
+ }
+ return PIPE_FORMAT_NONE;
+
+ case GL_COMPRESSED_RED:
+ case GL_COMPRESSED_RG:
+ case GL_COMPRESSED_RGB:
+ /* can only sample from compressed formats */
+ if (bindings & ~PIPE_BIND_SAMPLER_VIEW)
+ return PIPE_FORMAT_NONE;
+ else if (screen->is_format_supported(screen, PIPE_FORMAT_DXT1_RGB,
+ target, sample_count, bindings,
+ geom_flags))
+ return PIPE_FORMAT_DXT1_RGB;
+ else
+ return default_rgb_format(screen, target, sample_count, bindings,
+ geom_flags);
+
+ case GL_COMPRESSED_RGBA:
+ /* can only sample from compressed formats */
+ if (bindings & ~PIPE_BIND_SAMPLER_VIEW)
+ return PIPE_FORMAT_NONE;
+ else if (screen->is_format_supported(screen, PIPE_FORMAT_DXT3_RGBA,
+ target, sample_count, bindings,
+ geom_flags))
+ return PIPE_FORMAT_DXT3_RGBA;
+ else
+ return default_rgba_format(screen, target, sample_count, bindings,
+ geom_flags);
+
+ case GL_RGB_S3TC:
+ case GL_RGB4_S3TC:
+ case GL_COMPRESSED_RGB_S3TC_DXT1_EXT:
+ if (screen->is_format_supported(screen, PIPE_FORMAT_DXT1_RGB,
+ target, sample_count, bindings,
+ geom_flags))
+ return PIPE_FORMAT_DXT1_RGB;
+ else
+ return PIPE_FORMAT_NONE;
+
+ case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT:
+ if (screen->is_format_supported(screen, PIPE_FORMAT_DXT1_RGBA,
+ target, sample_count, bindings,
+ geom_flags))
+ return PIPE_FORMAT_DXT1_RGBA;
+ else
+ return PIPE_FORMAT_NONE;
+
+ case GL_RGBA_S3TC:
+ case GL_RGBA4_S3TC:
+ case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT:
+ if (screen->is_format_supported(screen, PIPE_FORMAT_DXT3_RGBA,
+ target, sample_count, bindings,
+ geom_flags))
+ return PIPE_FORMAT_DXT3_RGBA;
+ else
+ return PIPE_FORMAT_NONE;
+
+ case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT:
+ if (screen->is_format_supported(screen, PIPE_FORMAT_DXT5_RGBA,
+ target, sample_count, bindings,
+ geom_flags))
+ return PIPE_FORMAT_DXT5_RGBA;
+ else
+ return PIPE_FORMAT_NONE;
+
+#if 0
+ case GL_COMPRESSED_RGB_FXT1_3DFX:
+ return PIPE_FORMAT_RGB_FXT1;
+ case GL_COMPRESSED_RGBA_FXT1_3DFX:
+ return PIPE_FORMAT_RGB_FXT1;
+#endif
+
+ case GL_DEPTH_COMPONENT16:
+ if (screen->is_format_supported(screen, PIPE_FORMAT_Z16_UNORM, target,
+ sample_count, bindings, geom_flags))
+ return PIPE_FORMAT_Z16_UNORM;
+ /* fall-through */
+ case GL_DEPTH_COMPONENT24:
+ if (screen->is_format_supported(screen, PIPE_FORMAT_Z24_UNORM_S8_USCALED,
+ target, sample_count, bindings, geom_flags))
+ return PIPE_FORMAT_Z24_UNORM_S8_USCALED;
+ if (screen->is_format_supported(screen, PIPE_FORMAT_S8_USCALED_Z24_UNORM,
+ target, sample_count, bindings, geom_flags))
+ return PIPE_FORMAT_S8_USCALED_Z24_UNORM;
+ /* fall-through */
+ case GL_DEPTH_COMPONENT32:
+ if (screen->is_format_supported(screen, PIPE_FORMAT_Z32_UNORM, target,
+ sample_count, bindings, geom_flags))
+ return PIPE_FORMAT_Z32_UNORM;
+ /* fall-through */
+ case GL_DEPTH_COMPONENT:
+ {
+ static const enum pipe_format formats[] = {
+ PIPE_FORMAT_Z32_UNORM,
+ PIPE_FORMAT_Z24_UNORM_S8_USCALED,
+ PIPE_FORMAT_S8_USCALED_Z24_UNORM,
+ PIPE_FORMAT_Z16_UNORM
+ };
+ return find_supported_format(screen, formats, Elements(formats),
+ target, sample_count, bindings, geom_flags);
+ }
+
+ case GL_STENCIL_INDEX:
+ case GL_STENCIL_INDEX1_EXT:
+ case GL_STENCIL_INDEX4_EXT:
+ case GL_STENCIL_INDEX8_EXT:
+ case GL_STENCIL_INDEX16_EXT:
+ {
+ static const enum pipe_format formats[] = {
+ PIPE_FORMAT_S8_USCALED,
+ PIPE_FORMAT_Z24_UNORM_S8_USCALED,
+ PIPE_FORMAT_S8_USCALED_Z24_UNORM
+ };
+ return find_supported_format(screen, formats, Elements(formats),
+ target, sample_count, bindings, geom_flags);
+ }
+
+ case GL_DEPTH_STENCIL_EXT:
+ case GL_DEPTH24_STENCIL8_EXT:
+ {
+ static const enum pipe_format formats[] = {
+ PIPE_FORMAT_Z24_UNORM_S8_USCALED,
+ PIPE_FORMAT_S8_USCALED_Z24_UNORM
+ };
+ return find_supported_format(screen, formats, Elements(formats),
+ target, sample_count, bindings, geom_flags);
+ }
+
+ case GL_SRGB_EXT:
+ case GL_SRGB8_EXT:
+ case GL_SRGB_ALPHA_EXT:
+ case GL_SRGB8_ALPHA8_EXT:
+ return default_srgba_format( screen, target, sample_count, bindings,
+ geom_flags );
+
+ case GL_COMPRESSED_SRGB_EXT:
+ case GL_COMPRESSED_SRGB_S3TC_DXT1_EXT:
+ if (screen->is_format_supported(screen, PIPE_FORMAT_DXT1_SRGB, target,
+ sample_count, bindings, geom_flags))
+ return PIPE_FORMAT_DXT1_SRGB;
+ return default_srgba_format( screen, target, sample_count, bindings,
+ geom_flags );
+
+ case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT1_EXT:
+ return PIPE_FORMAT_DXT1_SRGBA;
+
+ case GL_COMPRESSED_SRGB_ALPHA_EXT:
+ case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT3_EXT:
+ if (screen->is_format_supported(screen, PIPE_FORMAT_DXT3_SRGBA, target,
+ sample_count, bindings, geom_flags))
+ return PIPE_FORMAT_DXT3_SRGBA;
+ return default_srgba_format( screen, target, sample_count, bindings,
+ geom_flags );
+
+ case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT:
+ return PIPE_FORMAT_DXT5_SRGBA;
+
+ case GL_SLUMINANCE_ALPHA_EXT:
+ case GL_SLUMINANCE8_ALPHA8_EXT:
+ case GL_COMPRESSED_SLUMINANCE_EXT:
+ case GL_COMPRESSED_SLUMINANCE_ALPHA_EXT:
+ if (screen->is_format_supported(screen, PIPE_FORMAT_L8A8_SRGB, target,
+ sample_count, bindings, geom_flags))
+ return PIPE_FORMAT_L8A8_SRGB;
+ return default_srgba_format( screen, target, sample_count, bindings,
+ geom_flags );
+
+ case GL_SLUMINANCE_EXT:
+ case GL_SLUMINANCE8_EXT:
+ if (screen->is_format_supported(screen, PIPE_FORMAT_L8_SRGB, target,
+ sample_count, bindings, geom_flags))
+ return PIPE_FORMAT_L8_SRGB;
+ return default_srgba_format( screen, target, sample_count, bindings,
+ geom_flags );
+
+ case GL_RED:
+ case GL_R8:
+ if (screen->is_format_supported(screen, PIPE_FORMAT_R8_UNORM, target,
+ sample_count, bindings, geom_flags))
+ return PIPE_FORMAT_R8_UNORM;
+ return PIPE_FORMAT_NONE;
+ case GL_RG:
+ case GL_RG8:
+ if (screen->is_format_supported(screen, PIPE_FORMAT_R8G8_UNORM, target,
+ sample_count, bindings, geom_flags))
+ return PIPE_FORMAT_R8G8_UNORM;
+ return PIPE_FORMAT_NONE;
+
+ case GL_R16:
+ if (screen->is_format_supported(screen, PIPE_FORMAT_R16_UNORM, target,
+ sample_count, bindings, geom_flags))
+ return PIPE_FORMAT_R16_UNORM;
+ return PIPE_FORMAT_NONE;
+
+ case GL_RG16:
+ if (screen->is_format_supported(screen, PIPE_FORMAT_R16G16_UNORM, target,
+ sample_count, bindings, geom_flags))
+ return PIPE_FORMAT_R16G16_UNORM;
+ return PIPE_FORMAT_NONE;
+
+ case GL_COMPRESSED_RED_RGTC1:
+ if (screen->is_format_supported(screen, PIPE_FORMAT_RGTC1_UNORM, target,
+ sample_count, bindings, geom_flags))
+ return PIPE_FORMAT_RGTC1_UNORM;
+ return PIPE_FORMAT_NONE;
+
+ case GL_COMPRESSED_SIGNED_RED_RGTC1:
+ if (screen->is_format_supported(screen, PIPE_FORMAT_RGTC1_SNORM, target,
+ sample_count, bindings, geom_flags))
+ return PIPE_FORMAT_RGTC1_SNORM;
+ return PIPE_FORMAT_NONE;
+
+ case GL_COMPRESSED_RG_RGTC2:
+ if (screen->is_format_supported(screen, PIPE_FORMAT_RGTC2_UNORM, target,
+ sample_count, bindings, geom_flags))
+ return PIPE_FORMAT_RGTC2_UNORM;
+ return PIPE_FORMAT_NONE;
+
+ case GL_COMPRESSED_SIGNED_RG_RGTC2:
+ if (screen->is_format_supported(screen, PIPE_FORMAT_RGTC2_SNORM, target,
+ sample_count, bindings, geom_flags))
+ return PIPE_FORMAT_RGTC2_SNORM;
+ return PIPE_FORMAT_NONE;
+
+ /* signed/unsigned integer formats.
+ * XXX Mesa only has formats for RGBA signed/unsigned integer formats.
+ * If/when new formats are added this code should be updated.
+ */
+ case GL_RED_INTEGER_EXT:
+ case GL_GREEN_INTEGER_EXT:
+ case GL_BLUE_INTEGER_EXT:
+ case GL_ALPHA_INTEGER_EXT:
+ case GL_RGB_INTEGER_EXT:
+ case GL_RGBA_INTEGER_EXT:
+ case GL_BGR_INTEGER_EXT:
+ case GL_BGRA_INTEGER_EXT:
+ case GL_LUMINANCE_INTEGER_EXT:
+ case GL_LUMINANCE_ALPHA_INTEGER_EXT:
+ /* fall-through */
+ case GL_RGBA8I_EXT:
+ case GL_RGB8I_EXT:
+ case GL_ALPHA8I_EXT:
+ case GL_INTENSITY8I_EXT:
+ case GL_LUMINANCE8I_EXT:
+ case GL_LUMINANCE_ALPHA8I_EXT:
+ if (screen->is_format_supported(screen, PIPE_FORMAT_R8G8B8A8_SSCALED,
+ target,
+ sample_count, bindings, geom_flags))
+ return PIPE_FORMAT_R8G8B8A8_SSCALED;
+ return PIPE_FORMAT_NONE;
+ case GL_RGBA16I_EXT:
+ case GL_RGB16I_EXT:
+ case GL_ALPHA16I_EXT:
+ case GL_INTENSITY16I_EXT:
+ case GL_LUMINANCE16I_EXT:
+ case GL_LUMINANCE_ALPHA16I_EXT:
+ if (screen->is_format_supported(screen, PIPE_FORMAT_R16G16B16A16_SSCALED,
+ target,
+ sample_count, bindings, geom_flags))
+ return PIPE_FORMAT_R16G16B16A16_SSCALED;
+ return PIPE_FORMAT_NONE;
+ case GL_RGBA32I_EXT:
+ case GL_RGB32I_EXT:
+ case GL_ALPHA32I_EXT:
+ case GL_INTENSITY32I_EXT:
+ case GL_LUMINANCE32I_EXT:
+ case GL_LUMINANCE_ALPHA32I_EXT:
+ /* xxx */
+ if (screen->is_format_supported(screen, PIPE_FORMAT_R32G32B32A32_SSCALED,
+ target,
+ sample_count, bindings, geom_flags))
+ return PIPE_FORMAT_R32G32B32A32_SSCALED;
+ return PIPE_FORMAT_NONE;
+
+ case GL_RGBA8UI_EXT:
+ case GL_RGB8UI_EXT:
+ case GL_ALPHA8UI_EXT:
+ case GL_INTENSITY8UI_EXT:
+ case GL_LUMINANCE8UI_EXT:
+ case GL_LUMINANCE_ALPHA8UI_EXT:
+ if (screen->is_format_supported(screen, PIPE_FORMAT_R8G8B8A8_USCALED,
+ target,
+ sample_count, bindings, geom_flags))
+ return PIPE_FORMAT_R8G8B8A8_USCALED;
+ return PIPE_FORMAT_NONE;
+
+ case GL_RGBA16UI_EXT:
+ case GL_RGB16UI_EXT:
+ case GL_ALPHA16UI_EXT:
+ case GL_INTENSITY16UI_EXT:
+ case GL_LUMINANCE16UI_EXT:
+ case GL_LUMINANCE_ALPHA16UI_EXT:
+ if (screen->is_format_supported(screen, PIPE_FORMAT_R16G16B16A16_USCALED,
+ target,
+ sample_count, bindings, geom_flags))
+ return PIPE_FORMAT_R16G16B16A16_USCALED;
+ return PIPE_FORMAT_NONE;
+
+ case GL_RGBA32UI_EXT:
+ case GL_RGB32UI_EXT:
+ case GL_ALPHA32UI_EXT:
+ case GL_INTENSITY32UI_EXT:
+ case GL_LUMINANCE32UI_EXT:
+ case GL_LUMINANCE_ALPHA32UI_EXT:
+ if (screen->is_format_supported(screen, PIPE_FORMAT_R32G32B32A32_USCALED,
+ target,
+ sample_count, bindings, geom_flags))
+ return PIPE_FORMAT_R32G32B32A32_USCALED;
+ return PIPE_FORMAT_NONE;
+
+ default:
+ return PIPE_FORMAT_NONE;
+ }
+}
+
+
+/**
+ * Called by FBO code to choose a PIPE_FORMAT_ for drawing surfaces.
+ */
+enum pipe_format
+st_choose_renderbuffer_format(struct pipe_screen *screen,
+ GLenum internalFormat, unsigned sample_count)
+{
+ uint usage;
+ if (_mesa_is_depth_or_stencil_format(internalFormat))
+ usage = PIPE_BIND_DEPTH_STENCIL;
+ else
+ usage = PIPE_BIND_RENDER_TARGET;
+ return st_choose_format(screen, internalFormat, PIPE_TEXTURE_2D,
+ sample_count, usage);
+}
+
+
+/**
+ * Called via ctx->Driver.chooseTextureFormat().
+ */
+gl_format
+st_ChooseTextureFormat_renderable(struct gl_context *ctx, GLint internalFormat,
+ GLenum format, GLenum type, GLboolean renderable)
+{
+ struct pipe_screen *screen = st_context(ctx)->pipe->screen;
+ enum pipe_format pFormat;
+ uint bindings;
+
+ (void) format;
+ (void) type;
+
+ /* GL textures may wind up being render targets, but we don't know
+ * that in advance. Specify potential render target flags now.
+ */
+ bindings = PIPE_BIND_SAMPLER_VIEW;
+ if (renderable == GL_TRUE) {
+ if (_mesa_is_depth_format(internalFormat) ||
+ _mesa_is_depth_or_stencil_format(internalFormat))
+ bindings |= PIPE_BIND_DEPTH_STENCIL;
+ else
+ bindings |= PIPE_BIND_RENDER_TARGET;
+ }
+
+ pFormat = st_choose_format(screen, internalFormat,
+ PIPE_TEXTURE_2D, 0, bindings);
+
+ if (pFormat == PIPE_FORMAT_NONE) {
+ /* try choosing format again, this time without render target bindings */
+ pFormat = st_choose_format(screen, internalFormat,
+ PIPE_TEXTURE_2D, 0, PIPE_BIND_SAMPLER_VIEW);
+ }
+
+ if (pFormat == PIPE_FORMAT_NONE) {
+ /* no luck at all */
+ return MESA_FORMAT_NONE;
+ }
+
+ return st_pipe_format_to_mesa_format(pFormat);
+}
+
+gl_format
+st_ChooseTextureFormat(struct gl_context *ctx, GLint internalFormat,
+ GLenum format, GLenum type)
+{
+ boolean want_renderable =
+ internalFormat == 3 || internalFormat == 4 ||
+ internalFormat == GL_RGB || internalFormat == GL_RGBA ||
+ internalFormat == GL_RGB8 || internalFormat == GL_RGBA8 ||
+ internalFormat == GL_BGRA;
+
+ return st_ChooseTextureFormat_renderable(ctx, internalFormat,
+ format, type, want_renderable);
+}
+
+/**
+ * Test if a gallium format is equivalent to a GL format/type.
+ */
+GLboolean
+st_equal_formats(enum pipe_format pFormat, GLenum format, GLenum type)
+{
+ switch (pFormat) {
+ case PIPE_FORMAT_A8B8G8R8_UNORM:
+ return format == GL_RGBA && type == GL_UNSIGNED_BYTE;
+ case PIPE_FORMAT_A8R8G8B8_UNORM:
+ return format == GL_BGRA && type == GL_UNSIGNED_BYTE;
+ case PIPE_FORMAT_B5G6R5_UNORM:
+ return format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5;
+ /* XXX more combos... */
+ default:
+ return GL_FALSE;
+ }
+}
+
+GLboolean
+st_sampler_compat_formats(enum pipe_format format1, enum pipe_format format2)
+{
+ if (format1 == format2)
+ return GL_TRUE;
+
+ if (format1 == PIPE_FORMAT_B8G8R8A8_UNORM &&
+ format2 == PIPE_FORMAT_B8G8R8X8_UNORM)
+ return GL_TRUE;
+
+ if (format1 == PIPE_FORMAT_B8G8R8X8_UNORM &&
+ format2 == PIPE_FORMAT_B8G8R8A8_UNORM)
+ return GL_TRUE;
+
+ if (format1 == PIPE_FORMAT_A8B8G8R8_UNORM &&
+ format2 == PIPE_FORMAT_X8B8G8R8_UNORM)
+ return GL_TRUE;
+
+ if (format1 == PIPE_FORMAT_X8B8G8R8_UNORM &&
+ format2 == PIPE_FORMAT_A8B8G8R8_UNORM)
+ return GL_TRUE;
+
+ if (format1 == PIPE_FORMAT_A8R8G8B8_UNORM &&
+ format2 == PIPE_FORMAT_X8R8G8B8_UNORM)
+ return GL_TRUE;
+
+ if (format1 == PIPE_FORMAT_X8R8G8B8_UNORM &&
+ format2 == PIPE_FORMAT_A8R8G8B8_UNORM)
+ return GL_TRUE;
+
+ return GL_FALSE;
+}
+
+
+
+/**
+ * This is used for translating texture border color and the clear
+ * color. For example, the clear color is interpreted according to
+ * the renderbuffer's base format. For example, if clearing a
+ * GL_LUMINANCE buffer, ClearColor[0] = luminance and ClearColor[1] =
+ * alpha. Similarly for texture border colors.
+ */
+void
+st_translate_color(const GLfloat colorIn[4], GLenum baseFormat,
+ GLfloat colorOut[4])
+{
+ switch (baseFormat) {
+ case GL_RED:
+ colorOut[0] = colorIn[0];
+ colorOut[1] = 0.0F;
+ colorOut[2] = 0.0F;
+ colorOut[3] = 1.0F;
+ break;
+ case GL_RG:
+ colorOut[0] = colorIn[0];
+ colorOut[1] = colorIn[1];
+ colorOut[2] = 0.0F;
+ colorOut[3] = 1.0F;
+ break;
+ case GL_RGB:
+ colorOut[0] = colorIn[0];
+ colorOut[1] = colorIn[1];
+ colorOut[2] = colorIn[2];
+ colorOut[3] = 1.0F;
+ break;
+ case GL_ALPHA:
+ colorOut[0] = colorOut[1] = colorOut[2] = 0.0;
+ colorOut[3] = colorIn[3];
+ break;
+ case GL_LUMINANCE:
+ colorOut[0] = colorOut[1] = colorOut[2] = colorIn[0];
+ colorOut[3] = 1.0;
+ break;
+ case GL_LUMINANCE_ALPHA:
+ colorOut[0] = colorOut[1] = colorOut[2] = colorIn[0];
+ colorOut[3] = colorIn[3];
+ break;
+ case GL_INTENSITY:
+ colorOut[0] = colorOut[1] = colorOut[2] = colorOut[3] = colorIn[0];
+ break;
+ default:
+ COPY_4V(colorOut, colorIn);
+ }
+}
diff --git a/mesalib/src/mesa/state_tracker/st_format.h b/mesalib/src/mesa/state_tracker/st_format.h index ffcaf402a..d31e60711 100644 --- a/mesalib/src/mesa/state_tracker/st_format.h +++ b/mesalib/src/mesa/state_tracker/st_format.h @@ -77,4 +77,10 @@ st_equal_formats(enum pipe_format pFormat, GLenum format, GLenum type); extern GLboolean
st_sampler_compat_formats(enum pipe_format format1, enum pipe_format format2);
+
+extern void
+st_translate_color(const GLfloat colorIn[4], GLenum baseFormat,
+ GLfloat colorOut[4]);
+
+
#endif /* ST_FORMAT_H */
diff --git a/mesalib/src/mesa/state_tracker/st_gen_mipmap.c b/mesalib/src/mesa/state_tracker/st_gen_mipmap.c index 5fcd806e0..d04ec696f 100644 --- a/mesalib/src/mesa/state_tracker/st_gen_mipmap.c +++ b/mesalib/src/mesa/state_tracker/st_gen_mipmap.c @@ -336,6 +336,11 @@ st_generate_mipmap(struct gl_context *ctx, GLenum target, if (lastLevel == 0)
return;
+ /* The texture isn't in a "complete" state yet so set the expected
+ * lastLevel here, since it won't get done in st_finalize_texture().
+ */
+ stObj->lastLevel = lastLevel;
+
if (pt->last_level < lastLevel) {
/* The current gallium texture doesn't have space for all the
* mipmap levels we need to generate. So allocate a new texture.
@@ -353,11 +358,6 @@ st_generate_mipmap(struct gl_context *ctx, GLenum target, oldTex->array_size,
oldTex->bind);
- /* The texture isn't in a "complete" state yet so set the expected
- * lastLevel here, since it won't get done in st_finalize_texture().
- */
- stObj->lastLevel = lastLevel;
-
/* This will copy the old texture's base image into the new texture
* which we just allocated.
*/
@@ -366,8 +366,6 @@ st_generate_mipmap(struct gl_context *ctx, GLenum target, /* release the old tex (will likely be freed too) */
pipe_resource_reference(&oldTex, NULL);
pipe_sampler_view_reference(&stObj->sampler_view, NULL);
-
- pt = stObj->pt;
}
else {
/* Make sure that the base texture image data is present in the
@@ -376,6 +374,8 @@ st_generate_mipmap(struct gl_context *ctx, GLenum target, st_finalize_texture(ctx, st->pipe, texObj);
}
+ pt = stObj->pt;
+
assert(pt->last_level >= lastLevel);
/* Try to generate the mipmap by rendering/texturing. If that fails,
diff --git a/mesalib/src/mesa/vbo/vbo_exec.h b/mesalib/src/mesa/vbo/vbo_exec.h index 2508016c0..0dd2286dc 100644 --- a/mesalib/src/mesa/vbo/vbo_exec.h +++ b/mesalib/src/mesa/vbo/vbo_exec.h @@ -152,7 +152,6 @@ struct vbo_exec_context void vbo_exec_init( struct gl_context *ctx );
void vbo_exec_destroy( struct gl_context *ctx );
void vbo_exec_invalidate_state( struct gl_context *ctx, GLuint new_state );
-void vbo_exec_FlushVertices_internal( struct gl_context *ctx, GLboolean unmap );
void vbo_exec_BeginVertices( struct gl_context *ctx );
void vbo_exec_FlushVertices( struct gl_context *ctx, GLuint flags );
diff --git a/mesalib/src/mesa/vbo/vbo_exec_api.c b/mesalib/src/mesa/vbo/vbo_exec_api.c index 0c59c41ba..859db5e2c 100644 --- a/mesalib/src/mesa/vbo/vbo_exec_api.c +++ b/mesalib/src/mesa/vbo/vbo_exec_api.c @@ -159,8 +159,7 @@ static void vbo_exec_copy_to_current( struct vbo_exec_context *exec ) exec->vtx.attrsz[i],
exec->vtx.attrptr[i]);
- if (memcmp(current, tmp, sizeof(tmp)) != 0)
- {
+ if (memcmp(current, tmp, sizeof(tmp)) != 0) {
memcpy(current, tmp, sizeof(tmp));
/* Given that we explicitly state size here, there is no need
@@ -192,14 +191,18 @@ static void vbo_exec_copy_to_current( struct vbo_exec_context *exec ) }
-static void vbo_exec_copy_from_current( struct vbo_exec_context *exec )
+/**
+ * Copy current vertex attribute values into the current vertex.
+ */
+static void
+vbo_exec_copy_from_current(struct vbo_exec_context *exec)
{
struct gl_context *ctx = exec->ctx;
struct vbo_context *vbo = vbo_context(ctx);
GLint i;
- for (i = VBO_ATTRIB_POS+1 ; i < VBO_ATTRIB_MAX ; i++) {
- const GLfloat *current = (GLfloat *)vbo->currval[i].Ptr;
+ for (i = VBO_ATTRIB_POS + 1; i < VBO_ATTRIB_MAX; i++) {
+ const GLfloat *current = (GLfloat *) vbo->currval[i].Ptr;
switch (exec->vtx.attrsz[i]) {
case 4: exec->vtx.attrptr[i][3] = current[3];
case 3: exec->vtx.attrptr[i][2] = current[2];
@@ -213,17 +216,21 @@ static void vbo_exec_copy_from_current( struct vbo_exec_context *exec ) /**
* Flush existing data, set new attrib size, replay copied vertices.
+ * This is called when we transition from a small vertex attribute size
+ * to a larger one. Ex: glTexCoord2f -> glTexCoord4f.
+ * We need to go back over the previous 2-component texcoords and insert
+ * zero and one values.
*/
-static void vbo_exec_wrap_upgrade_vertex( struct vbo_exec_context *exec,
- GLuint attr,
- GLuint newsz )
+static void
+vbo_exec_wrap_upgrade_vertex(struct vbo_exec_context *exec,
+ GLuint attr, GLuint newSize )
{
struct gl_context *ctx = exec->ctx;
struct vbo_context *vbo = vbo_context(ctx);
- GLint lastcount = exec->vtx.vert_count;
+ const GLint lastcount = exec->vtx.vert_count;
GLfloat *old_attrptr[VBO_ATTRIB_MAX];
- GLuint old_vtx_size = exec->vtx.vertex_size;
- GLuint oldsz = exec->vtx.attrsz[attr];
+ const GLuint old_vtx_size = exec->vtx.vertex_size; /* floats per vertex */
+ const GLuint oldSize = exec->vtx.attrsz[attr];
GLuint i;
/* Run pipeline on current vertices, copy wrapped vertices
@@ -239,7 +246,7 @@ static void vbo_exec_wrap_upgrade_vertex( struct vbo_exec_context *exec, memcpy(old_attrptr, exec->vtx.attrptr, sizeof(old_attrptr));
}
- if (unlikely(oldsz)) {
+ if (unlikely(oldSize)) {
/* Do a COPY_TO_CURRENT to ensure back-copying works for the
* case when the attribute already exists in the vertex and is
* having its size increased.
@@ -251,21 +258,21 @@ static void vbo_exec_wrap_upgrade_vertex( struct vbo_exec_context *exec, * begin/end so that they don't bloat the vertices.
*/
if (ctx->Driver.CurrentExecPrimitive == PRIM_OUTSIDE_BEGIN_END &&
- !oldsz && lastcount > 8 && exec->vtx.vertex_size) {
+ !oldSize && lastcount > 8 && exec->vtx.vertex_size) {
vbo_exec_copy_to_current( exec );
reset_attrfv( exec );
}
/* Fix up sizes:
*/
- exec->vtx.attrsz[attr] = newsz;
- exec->vtx.vertex_size += newsz - oldsz;
+ exec->vtx.attrsz[attr] = newSize;
+ exec->vtx.vertex_size += newSize - oldSize;
exec->vtx.max_vert = ((VBO_VERT_BUFFER_SIZE - exec->vtx.buffer_used) /
(exec->vtx.vertex_size * sizeof(GLfloat)));
exec->vtx.vert_count = 0;
exec->vtx.buffer_ptr = exec->vtx.buffer_map;
- if (unlikely(oldsz)) {
+ if (unlikely(oldSize)) {
/* Size changed, recalculate all the attrptr[] values
*/
GLfloat *tmp = exec->vtx.vertex;
@@ -283,11 +290,11 @@ static void vbo_exec_wrap_upgrade_vertex( struct vbo_exec_context *exec, * values.
*/
vbo_exec_copy_from_current( exec );
-
- } else {
+ }
+ else {
/* Just have to append the new attribute at the end */
exec->vtx.attrptr[attr] = exec->vtx.vertex +
- exec->vtx.vertex_size - newsz;
+ exec->vtx.vertex_size - newSize;
}
/* Replay stored vertices to translate them
@@ -311,10 +318,10 @@ static void vbo_exec_wrap_upgrade_vertex( struct vbo_exec_context *exec, GLint new_offset = exec->vtx.attrptr[j] - exec->vtx.vertex;
if (j == attr) {
- if (oldsz) {
+ if (oldSize) {
GLfloat tmp[4];
- COPY_CLEAN_4V(tmp, oldsz, data + old_offset);
- COPY_SZ_4V(dest + new_offset, newsz, tmp);
+ COPY_CLEAN_4V(tmp, oldSize, data + old_offset);
+ COPY_SZ_4V(dest + new_offset, newSize, tmp);
} else {
GLfloat *current = (GLfloat *)vbo->currval[j].Ptr;
COPY_SZ_4V(dest + new_offset, sz, current);
@@ -337,29 +344,34 @@ static void vbo_exec_wrap_upgrade_vertex( struct vbo_exec_context *exec, }
-static void vbo_exec_fixup_vertex( struct gl_context *ctx,
- GLuint attr, GLuint sz )
+/**
+ * This is when a vertex attribute transitions to a different size.
+ * For example, we saw a bunch of glTexCoord2f() calls and now we got a
+ * glTexCoord4f() call. We promote the array from size=2 to size=4.
+ */
+static void
+vbo_exec_fixup_vertex(struct gl_context *ctx, GLuint attr, GLuint newSize)
{
struct vbo_exec_context *exec = &vbo_context(ctx)->exec;
- int i;
- if (sz > exec->vtx.attrsz[attr]) {
+ if (newSize > exec->vtx.attrsz[attr]) {
/* New size is larger. Need to flush existing vertices and get
* an enlarged vertex format.
*/
- vbo_exec_wrap_upgrade_vertex( exec, attr, sz );
+ vbo_exec_wrap_upgrade_vertex( exec, attr, newSize );
}
- else if (sz < exec->vtx.active_sz[attr]) {
+ else if (newSize < exec->vtx.active_sz[attr]) {
static const GLfloat id[4] = { 0, 0, 0, 1 };
+ GLuint i;
/* New size is smaller - just need to fill in some
* zeros. Don't need to flush or wrap.
*/
- for (i = sz ; i <= exec->vtx.attrsz[attr] ; i++)
+ for (i = newSize; i <= exec->vtx.attrsz[attr]; i++)
exec->vtx.attrptr[attr][i-1] = id[i-1];
}
- exec->vtx.active_sz[attr] = sz;
+ exec->vtx.active_sz[attr] = newSize;
/* Does setting NeedFlush belong here? Necessitates resetting
* vtxfmt on each flush (otherwise flags won't get reset
@@ -370,37 +382,44 @@ static void vbo_exec_fixup_vertex( struct gl_context *ctx, }
-/*
+/**
+ * This macro is used to implement all the glVertex, glColor, glTexCoord,
+ * glVertexAttrib, etc functions.
*/
-#define ATTR( A, N, V0, V1, V2, V3 ) \
-do { \
- struct vbo_exec_context *exec = &vbo_context(ctx)->exec; \
+#define ATTR( A, N, V0, V1, V2, V3 ) \
+do { \
+ struct vbo_exec_context *exec = &vbo_context(ctx)->exec; \
\
- if (unlikely(!(exec->ctx->Driver.NeedFlush & FLUSH_UPDATE_CURRENT))) \
- ctx->Driver.BeginVertices( ctx ); \
+ if (unlikely(!(ctx->Driver.NeedFlush & FLUSH_UPDATE_CURRENT))) \
+ ctx->Driver.BeginVertices( ctx ); \
+ \
if (unlikely(exec->vtx.active_sz[A] != N)) \
vbo_exec_fixup_vertex(ctx, A, N); \
\
- { \
- GLfloat *dest = exec->vtx.attrptr[A]; \
- if (N>0) dest[0] = V0; \
- if (N>1) dest[1] = V1; \
- if (N>2) dest[2] = V2; \
- if (N>3) dest[3] = V3; \
- } \
- \
- if ((A) == 0) { \
- GLuint i; \
- \
- for (i = 0; i < exec->vtx.vertex_size; i++) \
- exec->vtx.buffer_ptr[i] = exec->vtx.vertex[i]; \
- \
+ { \
+ GLfloat *dest = exec->vtx.attrptr[A]; \
+ if (N>0) dest[0] = V0; \
+ if (N>1) dest[1] = V1; \
+ if (N>2) dest[2] = V2; \
+ if (N>3) dest[3] = V3; \
+ } \
+ \
+ if ((A) == 0) { \
+ /* This is a glVertex call */ \
+ GLuint i; \
+ \
+ for (i = 0; i < exec->vtx.vertex_size; i++) \
+ exec->vtx.buffer_ptr[i] = exec->vtx.vertex[i]; \
+ \
exec->vtx.buffer_ptr += exec->vtx.vertex_size; \
- exec->ctx->Driver.NeedFlush |= FLUSH_STORED_VERTICES; \
- \
- if (++exec->vtx.vert_count >= exec->vtx.max_vert) \
- vbo_exec_vtx_wrap( exec ); \
- } \
+ \
+ /* Set FLUSH_STORED_VERTICES to indicate that there's now */ \
+ /* something to draw (not just updating a color or texcoord).*/ \
+ ctx->Driver.NeedFlush |= FLUSH_STORED_VERTICES; \
+ \
+ if (++exec->vtx.vert_count >= exec->vtx.max_vert) \
+ vbo_exec_vtx_wrap( exec ); \
+ } \
} while (0)
@@ -514,6 +533,24 @@ static void GLAPIENTRY vbo_exec_EvalPoint2( GLint i, GLint j ) /**
+ * Flush (draw) vertices.
+ * \param unmap - leave VBO unmapped after flushing?
+ */
+static void
+vbo_exec_FlushVertices_internal(struct vbo_exec_context *exec, GLboolean unmap)
+{
+ if (exec->vtx.vert_count || unmap) {
+ vbo_exec_vtx_flush( exec, unmap );
+ }
+
+ if (exec->vtx.vertex_size) {
+ vbo_exec_copy_to_current( exec );
+ reset_attrfv( exec );
+ }
+}
+
+
+/**
* Called via glBegin.
*/
static void GLAPIENTRY vbo_exec_Begin( GLenum mode )
@@ -539,7 +576,7 @@ static void GLAPIENTRY vbo_exec_Begin( GLenum mode ) * begin/end pairs.
*/
if (exec->vtx.vertex_size && !exec->vtx.attrsz[0])
- vbo_exec_FlushVertices_internal( ctx, GL_FALSE );
+ vbo_exec_FlushVertices_internal(exec, GL_FALSE);
i = exec->vtx.prim_count++;
exec->vtx.prim[i].mode = mode;
@@ -894,32 +931,23 @@ void vbo_exec_vtx_destroy( struct vbo_exec_context *exec ) _mesa_reference_buffer_object(ctx, &exec->vtx.bufferobj, NULL);
}
+
+/**
+ * Called upon first glVertex, glColor, glTexCoord, etc.
+ */
void vbo_exec_BeginVertices( struct gl_context *ctx )
{
struct vbo_exec_context *exec = &vbo_context(ctx)->exec;
- if (0) printf("%s\n", __FUNCTION__);
+
vbo_exec_vtx_map( exec );
assert((exec->ctx->Driver.NeedFlush & FLUSH_UPDATE_CURRENT) == 0);
exec->ctx->Driver.NeedFlush |= FLUSH_UPDATE_CURRENT;
}
-void vbo_exec_FlushVertices_internal( struct gl_context *ctx, GLboolean unmap )
-{
- struct vbo_exec_context *exec = &vbo_context(ctx)->exec;
-
- if (exec->vtx.vert_count || unmap) {
- vbo_exec_vtx_flush( exec, unmap );
- }
-
- if (exec->vtx.vertex_size) {
- vbo_exec_copy_to_current( exec );
- reset_attrfv( exec );
- }
-}
-
/**
+ * Called via ctx->Driver.FlushVertices()
* \param flags bitmask of FLUSH_STORED_VERTICES, FLUSH_UPDATE_CURRENT
*/
void vbo_exec_FlushVertices( struct gl_context *ctx, GLuint flags )
@@ -932,10 +960,8 @@ void vbo_exec_FlushVertices( struct gl_context *ctx, GLuint flags ) assert(exec->flush_call_depth == 1);
#endif
- if (0) printf("%s\n", __FUNCTION__);
-
if (exec->ctx->Driver.CurrentExecPrimitive != PRIM_OUTSIDE_BEGIN_END) {
- if (0) printf("%s - inside begin/end\n", __FUNCTION__);
+ /* We've had glBegin but not glEnd! */
#ifdef DEBUG
exec->flush_call_depth--;
assert(exec->flush_call_depth == 0);
@@ -943,7 +969,8 @@ void vbo_exec_FlushVertices( struct gl_context *ctx, GLuint flags ) return;
}
- vbo_exec_FlushVertices_internal( ctx, GL_TRUE );
+ /* Flush (draw), and make sure VBO is left unmapped when done */
+ vbo_exec_FlushVertices_internal(exec, GL_TRUE);
/* Need to do this to ensure BeginVertices gets called again:
*/
diff --git a/mesalib/src/mesa/vbo/vbo_exec_draw.c b/mesalib/src/mesa/vbo/vbo_exec_draw.c index a2a098857..7cc843779 100644 --- a/mesalib/src/mesa/vbo/vbo_exec_draw.c +++ b/mesalib/src/mesa/vbo/vbo_exec_draw.c @@ -253,6 +253,9 @@ vbo_exec_bind_arrays( struct gl_context *ctx ) }
+/**
+ * Unmap the VBO. This is called before drawing.
+ */
static void
vbo_exec_vtx_unmap( struct vbo_exec_context *exec )
{
@@ -285,6 +288,9 @@ vbo_exec_vtx_unmap( struct vbo_exec_context *exec ) }
+/**
+ * Map the vertex buffer to begin storing glVertex, glColor, etc data.
+ */
void
vbo_exec_vtx_map( struct vbo_exec_context *exec )
{
@@ -301,14 +307,12 @@ vbo_exec_vtx_map( struct vbo_exec_context *exec ) if (!_mesa_is_bufferobj(exec->vtx.bufferobj))
return;
- if (exec->vtx.buffer_map != NULL) {
- assert(0);
- exec->vtx.buffer_map = NULL;
- exec->vtx.buffer_ptr = NULL;
- }
+ assert(!exec->vtx.buffer_map);
+ assert(!exec->vtx.buffer_ptr);
if (VBO_VERT_BUFFER_SIZE > exec->vtx.buffer_used + 1024 &&
ctx->Driver.MapBufferRange) {
+ /* The VBO exists and there's room for more */
exec->vtx.buffer_map =
(GLfloat *)ctx->Driver.MapBufferRange(ctx,
target,
@@ -321,6 +325,7 @@ vbo_exec_vtx_map( struct vbo_exec_context *exec ) }
if (!exec->vtx.buffer_map) {
+ /* Need to allocate a new VBO */
exec->vtx.buffer_used = 0;
ctx->Driver.BufferData(ctx, target,
@@ -349,9 +354,10 @@ vbo_exec_vtx_map( struct vbo_exec_context *exec ) /**
* Execute the buffer and save copied verts.
+ * \param keep_unmapped if true, leave the VBO unmapped when we're done.
*/
void
-vbo_exec_vtx_flush( struct vbo_exec_context *exec, GLboolean unmap )
+vbo_exec_vtx_flush(struct vbo_exec_context *exec, GLboolean keepUnmapped)
{
if (0)
vbo_exec_debug_verts( exec );
@@ -391,7 +397,7 @@ vbo_exec_vtx_flush( struct vbo_exec_context *exec, GLboolean unmap ) /* If using a real VBO, get new storage -- unless asked not to.
*/
- if (_mesa_is_bufferobj(exec->vtx.bufferobj) && !unmap) {
+ if (_mesa_is_bufferobj(exec->vtx.bufferobj) && !keepUnmapped) {
vbo_exec_vtx_map( exec );
}
}
@@ -399,14 +405,13 @@ vbo_exec_vtx_flush( struct vbo_exec_context *exec, GLboolean unmap ) /* May have to unmap explicitly if we didn't draw:
*/
- if (unmap &&
+ if (keepUnmapped &&
_mesa_is_bufferobj(exec->vtx.bufferobj) &&
exec->vtx.buffer_map) {
vbo_exec_vtx_unmap( exec );
}
-
- if (unmap || exec->vtx.vertex_size == 0)
+ if (keepUnmapped || exec->vtx.vertex_size == 0)
exec->vtx.max_vert = 0;
else
exec->vtx.max_vert = ((VBO_VERT_BUFFER_SIZE - exec->vtx.buffer_used) /
diff --git a/pixman/configure.ac b/pixman/configure.ac index 189cc0667..4a4759e91 100644 --- a/pixman/configure.ac +++ b/pixman/configure.ac @@ -326,7 +326,7 @@ if test "x$SSE2_CFLAGS" = "x" ; then SSE2_CFLAGS="-xarch=sse2"
fi
else
- SSE2_CFLAGS="-mmmx -msse2 -Winline"
+ SSE2_CFLAGS="-msse2 -Winline"
fi
fi
diff --git a/pixman/pixman/pixman-sse2.c b/pixman/pixman/pixman-sse2.c index b62a54c3f..e7ecc8733 100644 --- a/pixman/pixman/pixman-sse2.c +++ b/pixman/pixman/pixman-sse2.c @@ -30,36 +30,12 @@ #include <config.h>
#endif
-#include <mmintrin.h>
#include <xmmintrin.h> /* for _mm_shuffle_pi16 and _MM_SHUFFLE */
#include <emmintrin.h> /* for SSE2 intrinsics */
#include "pixman-private.h"
#include "pixman-combine32.h"
#include "pixman-fast-path.h"
-#if defined(_MSC_VER) && defined(_M_AMD64)
-/* Windows 64 doesn't allow MMX to be used, so
- * the pixman-x64-mmx-emulation.h file contains
- * implementations of those MMX intrinsics that
- * are used in the SSE2 implementation.
- */
-# include "pixman-x64-mmx-emulation.h"
-#endif
-
-#ifdef USE_SSE2
-
-/* --------------------------------------------------------------------
- * Locals
- */
-
-static __m64 mask_x0080;
-static __m64 mask_x00ff;
-static __m64 mask_x0101;
-static __m64 mask_x_alpha;
-
-static __m64 mask_x565_rgb;
-static __m64 mask_x565_unpack;
-
static __m128i mask_0080;
static __m128i mask_00ff;
static __m128i mask_0101;
@@ -77,9 +53,6 @@ static __m128i mask_blue; static __m128i mask_565_fix_rb;
static __m128i mask_565_fix_g;
-/* ----------------------------------------------------------------------
- * SSE2 Inlines
- */
static force_inline __m128i
unpack_32_1x128 (uint32_t data)
{
@@ -397,146 +370,104 @@ save_128_unaligned (__m128i* dst, _mm_storeu_si128 (dst, data);
}
-/* ------------------------------------------------------------------
- * MMX inlines
- */
-
-static force_inline __m64
-load_32_1x64 (uint32_t data)
-{
- return _mm_cvtsi32_si64 (data);
-}
-
-static force_inline __m64
-unpack_32_1x64 (uint32_t data)
-{
- return _mm_unpacklo_pi8 (load_32_1x64 (data), _mm_setzero_si64 ());
-}
-
-static force_inline __m64
-expand_alpha_1x64 (__m64 data)
+static force_inline __m128i
+load_32_1x128 (uint32_t data)
{
- return _mm_shuffle_pi16 (data, _MM_SHUFFLE (3, 3, 3, 3));
+ return _mm_cvtsi32_si128 (data);
}
-static force_inline __m64
-expand_alpha_rev_1x64 (__m64 data)
+static force_inline __m128i
+expand_alpha_rev_1x128 (__m128i data)
{
- return _mm_shuffle_pi16 (data, _MM_SHUFFLE (0, 0, 0, 0));
+ return _mm_shufflelo_epi16 (data, _MM_SHUFFLE (0, 0, 0, 0));
}
-static force_inline __m64
-expand_pixel_8_1x64 (uint8_t data)
+static force_inline __m128i
+expand_pixel_8_1x128 (uint8_t data)
{
- return _mm_shuffle_pi16 (
- unpack_32_1x64 ((uint32_t)data), _MM_SHUFFLE (0, 0, 0, 0));
+ return _mm_shufflelo_epi16 (
+ unpack_32_1x128 ((uint32_t)data), _MM_SHUFFLE (0, 0, 0, 0));
}
-static force_inline __m64
-pix_multiply_1x64 (__m64 data,
- __m64 alpha)
+static force_inline __m128i
+pix_multiply_1x128 (__m128i data,
+ __m128i alpha)
{
- return _mm_mulhi_pu16 (_mm_adds_pu16 (_mm_mullo_pi16 (data, alpha),
- mask_x0080),
- mask_x0101);
+ return _mm_mulhi_epu16 (_mm_adds_epu16 (_mm_mullo_epi16 (data, alpha),
+ mask_0080),
+ mask_0101);
}
-static force_inline __m64
-pix_add_multiply_1x64 (__m64* src,
- __m64* alpha_dst,
- __m64* dst,
- __m64* alpha_src)
+static force_inline __m128i
+pix_add_multiply_1x128 (__m128i* src,
+ __m128i* alpha_dst,
+ __m128i* dst,
+ __m128i* alpha_src)
{
- __m64 t1 = pix_multiply_1x64 (*src, *alpha_dst);
- __m64 t2 = pix_multiply_1x64 (*dst, *alpha_src);
+ __m128i t1 = pix_multiply_1x128 (*src, *alpha_dst);
+ __m128i t2 = pix_multiply_1x128 (*dst, *alpha_src);
- return _mm_adds_pu8 (t1, t2);
+ return _mm_adds_epu8 (t1, t2);
}
-static force_inline __m64
-negate_1x64 (__m64 data)
+static force_inline __m128i
+negate_1x128 (__m128i data)
{
- return _mm_xor_si64 (data, mask_x00ff);
+ return _mm_xor_si128 (data, mask_00ff);
}
-static force_inline __m64
-invert_colors_1x64 (__m64 data)
+static force_inline __m128i
+invert_colors_1x128 (__m128i data)
{
- return _mm_shuffle_pi16 (data, _MM_SHUFFLE (3, 0, 1, 2));
+ return _mm_shufflelo_epi16 (data, _MM_SHUFFLE (3, 0, 1, 2));
}
-static force_inline __m64
-over_1x64 (__m64 src, __m64 alpha, __m64 dst)
+static force_inline __m128i
+over_1x128 (__m128i src, __m128i alpha, __m128i dst)
{
- return _mm_adds_pu8 (src, pix_multiply_1x64 (dst, negate_1x64 (alpha)));
+ return _mm_adds_epu8 (src, pix_multiply_1x128 (dst, negate_1x128 (alpha)));
}
-static force_inline __m64
-in_over_1x64 (__m64* src, __m64* alpha, __m64* mask, __m64* dst)
+static force_inline __m128i
+in_over_1x128 (__m128i* src, __m128i* alpha, __m128i* mask, __m128i* dst)
{
- return over_1x64 (pix_multiply_1x64 (*src, *mask),
- pix_multiply_1x64 (*alpha, *mask),
- *dst);
+ return over_1x128 (pix_multiply_1x128 (*src, *mask),
+ pix_multiply_1x128 (*alpha, *mask),
+ *dst);
}
-static force_inline __m64
-over_rev_non_pre_1x64 (__m64 src, __m64 dst)
+static force_inline __m128i
+over_rev_non_pre_1x128 (__m128i src, __m128i dst)
{
- __m64 alpha = expand_alpha_1x64 (src);
+ __m128i alpha = expand_alpha_1x128 (src);
- return over_1x64 (pix_multiply_1x64 (invert_colors_1x64 (src),
- _mm_or_si64 (alpha, mask_x_alpha)),
- alpha,
- dst);
+ return over_1x128 (pix_multiply_1x128 (invert_colors_1x128 (src),
+ _mm_or_si128 (alpha, mask_alpha)),
+ alpha,
+ dst);
}
static force_inline uint32_t
-pack_1x64_32 (__m64 data)
+pack_1x128_32 (__m128i data)
{
- return _mm_cvtsi64_si32 (_mm_packs_pu16 (data, _mm_setzero_si64 ()));
+ return _mm_cvtsi128_si32 (_mm_packus_epi16 (data, _mm_setzero_si128 ()));
}
-/* Expand 16 bits positioned at @pos (0-3) of a mmx register into
- *
- * 00RR00GG00BB
- *
- * --- Expanding 565 in the low word ---
- *
- * m = (m << (32 - 3)) | (m << (16 - 5)) | m;
- * m = m & (01f0003f001f);
- * m = m * (008404100840);
- * m = m >> 8;
- *
- * Note the trick here - the top word is shifted by another nibble to
- * avoid it bumping into the middle word
- */
-static force_inline __m64
-expand565_16_1x64 (uint16_t pixel)
+static force_inline __m128i
+expand565_16_1x128 (uint16_t pixel)
{
- __m64 p;
- __m64 t1, t2;
+ __m128i m = _mm_cvtsi32_si128 (pixel);
- p = _mm_cvtsi32_si64 ((uint32_t) pixel);
+ m = unpack_565_to_8888 (m);
- t1 = _mm_slli_si64 (p, 36 - 11);
- t2 = _mm_slli_si64 (p, 16 - 5);
-
- p = _mm_or_si64 (t1, p);
- p = _mm_or_si64 (t2, p);
- p = _mm_and_si64 (p, mask_x565_rgb);
- p = _mm_mullo_pi16 (p, mask_x565_unpack);
-
- return _mm_srli_pi16 (p, 8);
+ return _mm_unpacklo_epi8 (m, _mm_setzero_si128 ());
}
-/* ----------------------------------------------------------------------------
- * Compose Core transformations
- */
static force_inline uint32_t
core_combine_over_u_pixel_sse2 (uint32_t src, uint32_t dst)
{
uint8_t a;
- __m64 ms;
+ __m128i xmms;
a = src >> 24;
@@ -546,9 +477,10 @@ core_combine_over_u_pixel_sse2 (uint32_t src, uint32_t dst) }
else if (src)
{
- ms = unpack_32_1x64 (src);
- return pack_1x64_32 (
- over_1x64 (ms, expand_alpha_1x64 (ms), unpack_32_1x64 (dst)));
+ xmms = unpack_32_1x128 (src);
+ return pack_1x128_32 (
+ over_1x128 (xmms, expand_alpha_1x128 (xmms),
+ unpack_32_1x128 (dst)));
}
return dst;
@@ -561,15 +493,15 @@ combine1 (const uint32_t *ps, const uint32_t *pm) if (pm)
{
- __m64 ms, mm;
+ __m128i ms, mm;
- mm = unpack_32_1x64 (*pm);
- mm = expand_alpha_1x64 (mm);
+ mm = unpack_32_1x128 (*pm);
+ mm = expand_alpha_1x128 (mm);
- ms = unpack_32_1x64 (s);
- ms = pix_multiply_1x64 (ms, mm);
+ ms = unpack_32_1x128 (s);
+ ms = pix_multiply_1x128 (ms, mm);
- s = pack_1x64_32 (ms);
+ s = pack_1x128_32 (ms);
}
return s;
@@ -766,10 +698,12 @@ core_combine_over_u_sse2_no_mask (uint32_t * pd, }
static force_inline void
-core_combine_over_u_sse2 (uint32_t* pd,
- const uint32_t* ps,
- const uint32_t* pm,
- int w)
+sse2_combine_over_u (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * pd,
+ const uint32_t * ps,
+ const uint32_t * pm,
+ int w)
{
if (pm)
core_combine_over_u_sse2_mask (pd, ps, pm, w);
@@ -777,11 +711,13 @@ core_combine_over_u_sse2 (uint32_t* pd, core_combine_over_u_sse2_no_mask (pd, ps, w);
}
-static force_inline void
-core_combine_over_reverse_u_sse2 (uint32_t* pd,
- const uint32_t* ps,
- const uint32_t* pm,
- int w)
+static void
+sse2_combine_over_reverse_u (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * pd,
+ const uint32_t * ps,
+ const uint32_t * pm,
+ int w)
{
uint32_t s, d;
@@ -847,7 +783,7 @@ core_combine_over_reverse_u_sse2 (uint32_t* pd, }
static force_inline uint32_t
-core_combine_in_u_pixelsse2 (uint32_t src, uint32_t dst)
+core_combine_in_u_pixel_sse2 (uint32_t src, uint32_t dst)
{
uint32_t maska = src >> 24;
@@ -857,19 +793,21 @@ core_combine_in_u_pixelsse2 (uint32_t src, uint32_t dst) }
else if (maska != 0xff)
{
- return pack_1x64_32 (
- pix_multiply_1x64 (unpack_32_1x64 (dst),
- expand_alpha_1x64 (unpack_32_1x64 (src))));
+ return pack_1x128_32 (
+ pix_multiply_1x128 (unpack_32_1x128 (dst),
+ expand_alpha_1x128 (unpack_32_1x128 (src))));
}
return dst;
}
-static force_inline void
-core_combine_in_u_sse2 (uint32_t* pd,
- const uint32_t* ps,
- const uint32_t* pm,
- int w)
+static void
+sse2_combine_in_u (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * pd,
+ const uint32_t * ps,
+ const uint32_t * pm,
+ int w)
{
uint32_t s, d;
@@ -881,7 +819,7 @@ core_combine_in_u_sse2 (uint32_t* pd, s = combine1 (ps, pm);
d = *pd;
- *pd++ = core_combine_in_u_pixelsse2 (d, s);
+ *pd++ = core_combine_in_u_pixel_sse2 (d, s);
w--;
ps++;
if (pm)
@@ -916,7 +854,7 @@ core_combine_in_u_sse2 (uint32_t* pd, s = combine1 (ps, pm);
d = *pd;
- *pd++ = core_combine_in_u_pixelsse2 (d, s);
+ *pd++ = core_combine_in_u_pixel_sse2 (d, s);
w--;
ps++;
if (pm)
@@ -924,11 +862,13 @@ core_combine_in_u_sse2 (uint32_t* pd, }
}
-static force_inline void
-core_combine_reverse_in_u_sse2 (uint32_t* pd,
- const uint32_t* ps,
- const uint32_t *pm,
- int w)
+static void
+sse2_combine_in_reverse_u (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * pd,
+ const uint32_t * ps,
+ const uint32_t * pm,
+ int w)
{
uint32_t s, d;
@@ -940,7 +880,7 @@ core_combine_reverse_in_u_sse2 (uint32_t* pd, s = combine1 (ps, pm);
d = *pd;
- *pd++ = core_combine_in_u_pixelsse2 (s, d);
+ *pd++ = core_combine_in_u_pixel_sse2 (s, d);
ps++;
w--;
if (pm)
@@ -975,7 +915,7 @@ core_combine_reverse_in_u_sse2 (uint32_t* pd, s = combine1 (ps, pm);
d = *pd;
- *pd++ = core_combine_in_u_pixelsse2 (s, d);
+ *pd++ = core_combine_in_u_pixel_sse2 (s, d);
w--;
ps++;
if (pm)
@@ -983,21 +923,23 @@ core_combine_reverse_in_u_sse2 (uint32_t* pd, }
}
-static force_inline void
-core_combine_reverse_out_u_sse2 (uint32_t* pd,
- const uint32_t* ps,
- const uint32_t* pm,
- int w)
+static void
+sse2_combine_out_reverse_u (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * pd,
+ const uint32_t * ps,
+ const uint32_t * pm,
+ int w)
{
while (w && ((unsigned long) pd & 15))
{
uint32_t s = combine1 (ps, pm);
uint32_t d = *pd;
- *pd++ = pack_1x64_32 (
- pix_multiply_1x64 (
- unpack_32_1x64 (d), negate_1x64 (
- expand_alpha_1x64 (unpack_32_1x64 (s)))));
+ *pd++ = pack_1x128_32 (
+ pix_multiply_1x128 (
+ unpack_32_1x128 (d), negate_1x128 (
+ expand_alpha_1x128 (unpack_32_1x128 (s)))));
if (pm)
pm++;
@@ -1039,10 +981,10 @@ core_combine_reverse_out_u_sse2 (uint32_t* pd, uint32_t s = combine1 (ps, pm);
uint32_t d = *pd;
- *pd++ = pack_1x64_32 (
- pix_multiply_1x64 (
- unpack_32_1x64 (d), negate_1x64 (
- expand_alpha_1x64 (unpack_32_1x64 (s)))));
+ *pd++ = pack_1x128_32 (
+ pix_multiply_1x128 (
+ unpack_32_1x128 (d), negate_1x128 (
+ expand_alpha_1x128 (unpack_32_1x128 (s)))));
ps++;
if (pm)
pm++;
@@ -1050,21 +992,23 @@ core_combine_reverse_out_u_sse2 (uint32_t* pd, }
}
-static force_inline void
-core_combine_out_u_sse2 (uint32_t* pd,
- const uint32_t* ps,
- const uint32_t* pm,
- int w)
+static void
+sse2_combine_out_u (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * pd,
+ const uint32_t * ps,
+ const uint32_t * pm,
+ int w)
{
while (w && ((unsigned long) pd & 15))
{
uint32_t s = combine1 (ps, pm);
uint32_t d = *pd;
- *pd++ = pack_1x64_32 (
- pix_multiply_1x64 (
- unpack_32_1x64 (s), negate_1x64 (
- expand_alpha_1x64 (unpack_32_1x64 (d)))));
+ *pd++ = pack_1x128_32 (
+ pix_multiply_1x128 (
+ unpack_32_1x128 (s), negate_1x128 (
+ expand_alpha_1x128 (unpack_32_1x128 (d)))));
w--;
ps++;
if (pm)
@@ -1104,10 +1048,10 @@ core_combine_out_u_sse2 (uint32_t* pd, uint32_t s = combine1 (ps, pm);
uint32_t d = *pd;
- *pd++ = pack_1x64_32 (
- pix_multiply_1x64 (
- unpack_32_1x64 (s), negate_1x64 (
- expand_alpha_1x64 (unpack_32_1x64 (d)))));
+ *pd++ = pack_1x128_32 (
+ pix_multiply_1x128 (
+ unpack_32_1x128 (s), negate_1x128 (
+ expand_alpha_1x128 (unpack_32_1x128 (d)))));
w--;
ps++;
if (pm)
@@ -1119,20 +1063,22 @@ static force_inline uint32_t core_combine_atop_u_pixel_sse2 (uint32_t src,
uint32_t dst)
{
- __m64 s = unpack_32_1x64 (src);
- __m64 d = unpack_32_1x64 (dst);
+ __m128i s = unpack_32_1x128 (src);
+ __m128i d = unpack_32_1x128 (dst);
- __m64 sa = negate_1x64 (expand_alpha_1x64 (s));
- __m64 da = expand_alpha_1x64 (d);
+ __m128i sa = negate_1x128 (expand_alpha_1x128 (s));
+ __m128i da = expand_alpha_1x128 (d);
- return pack_1x64_32 (pix_add_multiply_1x64 (&s, &da, &d, &sa));
+ return pack_1x128_32 (pix_add_multiply_1x128 (&s, &da, &d, &sa));
}
-static force_inline void
-core_combine_atop_u_sse2 (uint32_t* pd,
- const uint32_t* ps,
- const uint32_t* pm,
- int w)
+static void
+sse2_combine_atop_u (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * pd,
+ const uint32_t * ps,
+ const uint32_t * pm,
+ int w)
{
uint32_t s, d;
@@ -1201,20 +1147,22 @@ static force_inline uint32_t core_combine_reverse_atop_u_pixel_sse2 (uint32_t src,
uint32_t dst)
{
- __m64 s = unpack_32_1x64 (src);
- __m64 d = unpack_32_1x64 (dst);
+ __m128i s = unpack_32_1x128 (src);
+ __m128i d = unpack_32_1x128 (dst);
- __m64 sa = expand_alpha_1x64 (s);
- __m64 da = negate_1x64 (expand_alpha_1x64 (d));
+ __m128i sa = expand_alpha_1x128 (s);
+ __m128i da = negate_1x128 (expand_alpha_1x128 (d));
- return pack_1x64_32 (pix_add_multiply_1x64 (&s, &da, &d, &sa));
+ return pack_1x128_32 (pix_add_multiply_1x128 (&s, &da, &d, &sa));
}
-static force_inline void
-core_combine_reverse_atop_u_sse2 (uint32_t* pd,
- const uint32_t* ps,
- const uint32_t* pm,
- int w)
+static void
+sse2_combine_atop_reverse_u (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * pd,
+ const uint32_t * ps,
+ const uint32_t * pm,
+ int w)
{
uint32_t s, d;
@@ -1283,20 +1231,22 @@ static force_inline uint32_t core_combine_xor_u_pixel_sse2 (uint32_t src,
uint32_t dst)
{
- __m64 s = unpack_32_1x64 (src);
- __m64 d = unpack_32_1x64 (dst);
+ __m128i s = unpack_32_1x128 (src);
+ __m128i d = unpack_32_1x128 (dst);
- __m64 neg_d = negate_1x64 (expand_alpha_1x64 (d));
- __m64 neg_s = negate_1x64 (expand_alpha_1x64 (s));
+ __m128i neg_d = negate_1x128 (expand_alpha_1x128 (d));
+ __m128i neg_s = negate_1x128 (expand_alpha_1x128 (s));
- return pack_1x64_32 (pix_add_multiply_1x64 (&s, &neg_d, &d, &neg_s));
+ return pack_1x128_32 (pix_add_multiply_1x128 (&s, &neg_d, &d, &neg_s));
}
-static force_inline void
-core_combine_xor_u_sse2 (uint32_t* dst,
- const uint32_t* src,
- const uint32_t *mask,
- int width)
+static void
+sse2_combine_xor_u (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * dst,
+ const uint32_t * src,
+ const uint32_t * mask,
+ int width)
{
int w = width;
uint32_t s, d;
@@ -1368,10 +1318,12 @@ core_combine_xor_u_sse2 (uint32_t* dst, }
static force_inline void
-core_combine_add_u_sse2 (uint32_t* dst,
- const uint32_t* src,
- const uint32_t* mask,
- int width)
+sse2_combine_add_u (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * dst,
+ const uint32_t * src,
+ const uint32_t * mask,
+ int width)
{
int w = width;
uint32_t s, d;
@@ -1387,8 +1339,8 @@ core_combine_add_u_sse2 (uint32_t* dst, ps++;
if (pm)
pm++;
- *pd++ = _mm_cvtsi64_si32 (
- _mm_adds_pu8 (_mm_cvtsi32_si64 (s), _mm_cvtsi32_si64 (d)));
+ *pd++ = _mm_cvtsi128_si32 (
+ _mm_adds_epu8 (_mm_cvtsi32_si128 (s), _mm_cvtsi32_si128 (d)));
w--;
}
@@ -1414,8 +1366,8 @@ core_combine_add_u_sse2 (uint32_t* dst, d = *pd;
ps++;
- *pd++ = _mm_cvtsi64_si32 (
- _mm_adds_pu8 (_mm_cvtsi32_si64 (s), _mm_cvtsi32_si64 (d)));
+ *pd++ = _mm_cvtsi128_si32 (
+ _mm_adds_epu8 (_mm_cvtsi32_si128 (s), _mm_cvtsi32_si128 (d)));
if (pm)
pm++;
}
@@ -1425,25 +1377,27 @@ static force_inline uint32_t core_combine_saturate_u_pixel_sse2 (uint32_t src,
uint32_t dst)
{
- __m64 ms = unpack_32_1x64 (src);
- __m64 md = unpack_32_1x64 (dst);
+ __m128i ms = unpack_32_1x128 (src);
+ __m128i md = unpack_32_1x128 (dst);
uint32_t sa = src >> 24;
uint32_t da = ~dst >> 24;
if (sa > da)
{
- ms = pix_multiply_1x64 (
- ms, expand_alpha_1x64 (unpack_32_1x64 (DIV_UN8 (da, sa) << 24)));
+ ms = pix_multiply_1x128 (
+ ms, expand_alpha_1x128 (unpack_32_1x128 (DIV_UN8 (da, sa) << 24)));
}
- return pack_1x64_32 (_mm_adds_pu16 (md, ms));
+ return pack_1x128_32 (_mm_adds_epu16 (md, ms));
}
-static force_inline void
-core_combine_saturate_u_sse2 (uint32_t * pd,
- const uint32_t *ps,
- const uint32_t *pm,
- int w)
+static void
+sse2_combine_saturate_u (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * pd,
+ const uint32_t * ps,
+ const uint32_t * pm,
+ int w)
{
uint32_t s, d;
@@ -1524,11 +1478,13 @@ core_combine_saturate_u_sse2 (uint32_t * pd, }
}
-static force_inline void
-core_combine_src_ca_sse2 (uint32_t* pd,
- const uint32_t* ps,
- const uint32_t *pm,
- int w)
+static void
+sse2_combine_src_ca (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * pd,
+ const uint32_t * ps,
+ const uint32_t * pm,
+ int w)
{
uint32_t s, m;
@@ -1540,8 +1496,8 @@ core_combine_src_ca_sse2 (uint32_t* pd, {
s = *ps++;
m = *pm++;
- *pd++ = pack_1x64_32 (
- pix_multiply_1x64 (unpack_32_1x64 (s), unpack_32_1x64 (m)));
+ *pd++ = pack_1x128_32 (
+ pix_multiply_1x128 (unpack_32_1x128 (s), unpack_32_1x128 (m)));
w--;
}
@@ -1570,8 +1526,8 @@ core_combine_src_ca_sse2 (uint32_t* pd, {
s = *ps++;
m = *pm++;
- *pd++ = pack_1x64_32 (
- pix_multiply_1x64 (unpack_32_1x64 (s), unpack_32_1x64 (m)));
+ *pd++ = pack_1x128_32 (
+ pix_multiply_1x128 (unpack_32_1x128 (s), unpack_32_1x128 (m)));
w--;
}
}
@@ -1581,19 +1537,21 @@ core_combine_over_ca_pixel_sse2 (uint32_t src, uint32_t mask,
uint32_t dst)
{
- __m64 s = unpack_32_1x64 (src);
- __m64 expAlpha = expand_alpha_1x64 (s);
- __m64 unpk_mask = unpack_32_1x64 (mask);
- __m64 unpk_dst = unpack_32_1x64 (dst);
+ __m128i s = unpack_32_1x128 (src);
+ __m128i expAlpha = expand_alpha_1x128 (s);
+ __m128i unpk_mask = unpack_32_1x128 (mask);
+ __m128i unpk_dst = unpack_32_1x128 (dst);
- return pack_1x64_32 (in_over_1x64 (&s, &expAlpha, &unpk_mask, &unpk_dst));
+ return pack_1x128_32 (in_over_1x128 (&s, &expAlpha, &unpk_mask, &unpk_dst));
}
-static force_inline void
-core_combine_over_ca_sse2 (uint32_t* pd,
- const uint32_t* ps,
- const uint32_t *pm,
- int w)
+static void
+sse2_combine_over_ca (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * pd,
+ const uint32_t * ps,
+ const uint32_t * pm,
+ int w)
{
uint32_t s, m, d;
@@ -1655,19 +1613,21 @@ core_combine_over_reverse_ca_pixel_sse2 (uint32_t src, uint32_t mask,
uint32_t dst)
{
- __m64 d = unpack_32_1x64 (dst);
+ __m128i d = unpack_32_1x128 (dst);
- return pack_1x64_32 (
- over_1x64 (d, expand_alpha_1x64 (d),
- pix_multiply_1x64 (unpack_32_1x64 (src),
- unpack_32_1x64 (mask))));
+ return pack_1x128_32 (
+ over_1x128 (d, expand_alpha_1x128 (d),
+ pix_multiply_1x128 (unpack_32_1x128 (src),
+ unpack_32_1x128 (mask))));
}
-static force_inline void
-core_combine_over_reverse_ca_sse2 (uint32_t* pd,
- const uint32_t* ps,
- const uint32_t *pm,
- int w)
+static void
+sse2_combine_over_reverse_ca (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * pd,
+ const uint32_t * ps,
+ const uint32_t * pm,
+ int w)
{
uint32_t s, m, d;
@@ -1726,11 +1686,13 @@ core_combine_over_reverse_ca_sse2 (uint32_t* pd, }
}
-static force_inline void
-core_combine_in_ca_sse2 (uint32_t * pd,
- const uint32_t *ps,
- const uint32_t *pm,
- int w)
+static void
+sse2_combine_in_ca (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * pd,
+ const uint32_t * ps,
+ const uint32_t * pm,
+ int w)
{
uint32_t s, m, d;
@@ -1745,10 +1707,10 @@ core_combine_in_ca_sse2 (uint32_t * pd, m = *pm++;
d = *pd;
- *pd++ = pack_1x64_32 (
- pix_multiply_1x64 (
- pix_multiply_1x64 (unpack_32_1x64 (s), unpack_32_1x64 (m)),
- expand_alpha_1x64 (unpack_32_1x64 (d))));
+ *pd++ = pack_1x128_32 (
+ pix_multiply_1x128 (
+ pix_multiply_1x128 (unpack_32_1x128 (s), unpack_32_1x128 (m)),
+ expand_alpha_1x128 (unpack_32_1x128 (d))));
w--;
}
@@ -1789,21 +1751,23 @@ core_combine_in_ca_sse2 (uint32_t * pd, m = *pm++;
d = *pd;
- *pd++ = pack_1x64_32 (
- pix_multiply_1x64 (
- pix_multiply_1x64 (
- unpack_32_1x64 (s), unpack_32_1x64 (m)),
- expand_alpha_1x64 (unpack_32_1x64 (d))));
+ *pd++ = pack_1x128_32 (
+ pix_multiply_1x128 (
+ pix_multiply_1x128 (
+ unpack_32_1x128 (s), unpack_32_1x128 (m)),
+ expand_alpha_1x128 (unpack_32_1x128 (d))));
w--;
}
}
-static force_inline void
-core_combine_in_reverse_ca_sse2 (uint32_t * pd,
- const uint32_t *ps,
- const uint32_t *pm,
- int w)
+static void
+sse2_combine_in_reverse_ca (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * pd,
+ const uint32_t * ps,
+ const uint32_t * pm,
+ int w)
{
uint32_t s, m, d;
@@ -1818,11 +1782,11 @@ core_combine_in_reverse_ca_sse2 (uint32_t * pd, m = *pm++;
d = *pd;
- *pd++ = pack_1x64_32 (
- pix_multiply_1x64 (
- unpack_32_1x64 (d),
- pix_multiply_1x64 (unpack_32_1x64 (m),
- expand_alpha_1x64 (unpack_32_1x64 (s)))));
+ *pd++ = pack_1x128_32 (
+ pix_multiply_1x128 (
+ unpack_32_1x128 (d),
+ pix_multiply_1x128 (unpack_32_1x128 (m),
+ expand_alpha_1x128 (unpack_32_1x128 (s)))));
w--;
}
@@ -1861,20 +1825,22 @@ core_combine_in_reverse_ca_sse2 (uint32_t * pd, m = *pm++;
d = *pd;
- *pd++ = pack_1x64_32 (
- pix_multiply_1x64 (
- unpack_32_1x64 (d),
- pix_multiply_1x64 (unpack_32_1x64 (m),
- expand_alpha_1x64 (unpack_32_1x64 (s)))));
+ *pd++ = pack_1x128_32 (
+ pix_multiply_1x128 (
+ unpack_32_1x128 (d),
+ pix_multiply_1x128 (unpack_32_1x128 (m),
+ expand_alpha_1x128 (unpack_32_1x128 (s)))));
w--;
}
}
-static force_inline void
-core_combine_out_ca_sse2 (uint32_t * pd,
- const uint32_t *ps,
- const uint32_t *pm,
- int w)
+static void
+sse2_combine_out_ca (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * pd,
+ const uint32_t * ps,
+ const uint32_t * pm,
+ int w)
{
uint32_t s, m, d;
@@ -1889,11 +1855,11 @@ core_combine_out_ca_sse2 (uint32_t * pd, m = *pm++;
d = *pd;
- *pd++ = pack_1x64_32 (
- pix_multiply_1x64 (
- pix_multiply_1x64 (
- unpack_32_1x64 (s), unpack_32_1x64 (m)),
- negate_1x64 (expand_alpha_1x64 (unpack_32_1x64 (d)))));
+ *pd++ = pack_1x128_32 (
+ pix_multiply_1x128 (
+ pix_multiply_1x128 (
+ unpack_32_1x128 (s), unpack_32_1x128 (m)),
+ negate_1x128 (expand_alpha_1x128 (unpack_32_1x128 (d)))));
w--;
}
@@ -1934,21 +1900,23 @@ core_combine_out_ca_sse2 (uint32_t * pd, m = *pm++;
d = *pd;
- *pd++ = pack_1x64_32 (
- pix_multiply_1x64 (
- pix_multiply_1x64 (
- unpack_32_1x64 (s), unpack_32_1x64 (m)),
- negate_1x64 (expand_alpha_1x64 (unpack_32_1x64 (d)))));
+ *pd++ = pack_1x128_32 (
+ pix_multiply_1x128 (
+ pix_multiply_1x128 (
+ unpack_32_1x128 (s), unpack_32_1x128 (m)),
+ negate_1x128 (expand_alpha_1x128 (unpack_32_1x128 (d)))));
w--;
}
}
-static force_inline void
-core_combine_out_reverse_ca_sse2 (uint32_t * pd,
- const uint32_t *ps,
- const uint32_t *pm,
- int w)
+static void
+sse2_combine_out_reverse_ca (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * pd,
+ const uint32_t * ps,
+ const uint32_t * pm,
+ int w)
{
uint32_t s, m, d;
@@ -1963,12 +1931,12 @@ core_combine_out_reverse_ca_sse2 (uint32_t * pd, m = *pm++;
d = *pd;
- *pd++ = pack_1x64_32 (
- pix_multiply_1x64 (
- unpack_32_1x64 (d),
- negate_1x64 (pix_multiply_1x64 (
- unpack_32_1x64 (m),
- expand_alpha_1x64 (unpack_32_1x64 (s))))));
+ *pd++ = pack_1x128_32 (
+ pix_multiply_1x128 (
+ unpack_32_1x128 (d),
+ negate_1x128 (pix_multiply_1x128 (
+ unpack_32_1x128 (m),
+ expand_alpha_1x128 (unpack_32_1x128 (s))))));
w--;
}
@@ -2011,12 +1979,12 @@ core_combine_out_reverse_ca_sse2 (uint32_t * pd, m = *pm++;
d = *pd;
- *pd++ = pack_1x64_32 (
- pix_multiply_1x64 (
- unpack_32_1x64 (d),
- negate_1x64 (pix_multiply_1x64 (
- unpack_32_1x64 (m),
- expand_alpha_1x64 (unpack_32_1x64 (s))))));
+ *pd++ = pack_1x128_32 (
+ pix_multiply_1x128 (
+ unpack_32_1x128 (d),
+ negate_1x128 (pix_multiply_1x128 (
+ unpack_32_1x128 (m),
+ expand_alpha_1x128 (unpack_32_1x128 (s))))));
w--;
}
}
@@ -2026,23 +1994,25 @@ core_combine_atop_ca_pixel_sse2 (uint32_t src, uint32_t mask,
uint32_t dst)
{
- __m64 m = unpack_32_1x64 (mask);
- __m64 s = unpack_32_1x64 (src);
- __m64 d = unpack_32_1x64 (dst);
- __m64 sa = expand_alpha_1x64 (s);
- __m64 da = expand_alpha_1x64 (d);
+ __m128i m = unpack_32_1x128 (mask);
+ __m128i s = unpack_32_1x128 (src);
+ __m128i d = unpack_32_1x128 (dst);
+ __m128i sa = expand_alpha_1x128 (s);
+ __m128i da = expand_alpha_1x128 (d);
- s = pix_multiply_1x64 (s, m);
- m = negate_1x64 (pix_multiply_1x64 (m, sa));
+ s = pix_multiply_1x128 (s, m);
+ m = negate_1x128 (pix_multiply_1x128 (m, sa));
- return pack_1x64_32 (pix_add_multiply_1x64 (&d, &m, &s, &da));
+ return pack_1x128_32 (pix_add_multiply_1x128 (&d, &m, &s, &da));
}
-static force_inline void
-core_combine_atop_ca_sse2 (uint32_t * pd,
- const uint32_t *ps,
- const uint32_t *pm,
- int w)
+static void
+sse2_combine_atop_ca (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * pd,
+ const uint32_t * ps,
+ const uint32_t * pm,
+ int w)
{
uint32_t s, m, d;
@@ -2116,24 +2086,26 @@ core_combine_reverse_atop_ca_pixel_sse2 (uint32_t src, uint32_t mask,
uint32_t dst)
{
- __m64 m = unpack_32_1x64 (mask);
- __m64 s = unpack_32_1x64 (src);
- __m64 d = unpack_32_1x64 (dst);
+ __m128i m = unpack_32_1x128 (mask);
+ __m128i s = unpack_32_1x128 (src);
+ __m128i d = unpack_32_1x128 (dst);
- __m64 da = negate_1x64 (expand_alpha_1x64 (d));
- __m64 sa = expand_alpha_1x64 (s);
+ __m128i da = negate_1x128 (expand_alpha_1x128 (d));
+ __m128i sa = expand_alpha_1x128 (s);
- s = pix_multiply_1x64 (s, m);
- m = pix_multiply_1x64 (m, sa);
+ s = pix_multiply_1x128 (s, m);
+ m = pix_multiply_1x128 (m, sa);
- return pack_1x64_32 (pix_add_multiply_1x64 (&d, &m, &s, &da));
+ return pack_1x128_32 (pix_add_multiply_1x128 (&d, &m, &s, &da));
}
-static force_inline void
-core_combine_reverse_atop_ca_sse2 (uint32_t * pd,
- const uint32_t *ps,
- const uint32_t *pm,
- int w)
+static void
+sse2_combine_atop_reverse_ca (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * pd,
+ const uint32_t * ps,
+ const uint32_t * pm,
+ int w)
{
uint32_t s, m, d;
@@ -2208,26 +2180,28 @@ core_combine_xor_ca_pixel_sse2 (uint32_t src, uint32_t mask,
uint32_t dst)
{
- __m64 a = unpack_32_1x64 (mask);
- __m64 s = unpack_32_1x64 (src);
- __m64 d = unpack_32_1x64 (dst);
+ __m128i a = unpack_32_1x128 (mask);
+ __m128i s = unpack_32_1x128 (src);
+ __m128i d = unpack_32_1x128 (dst);
- __m64 alpha_dst = negate_1x64 (pix_multiply_1x64 (
- a, expand_alpha_1x64 (s)));
- __m64 dest = pix_multiply_1x64 (s, a);
- __m64 alpha_src = negate_1x64 (expand_alpha_1x64 (d));
+ __m128i alpha_dst = negate_1x128 (pix_multiply_1x128 (
+ a, expand_alpha_1x128 (s)));
+ __m128i dest = pix_multiply_1x128 (s, a);
+ __m128i alpha_src = negate_1x128 (expand_alpha_1x128 (d));
- return pack_1x64_32 (pix_add_multiply_1x64 (&d,
+ return pack_1x128_32 (pix_add_multiply_1x128 (&d,
&alpha_dst,
&dest,
&alpha_src));
}
-static force_inline void
-core_combine_xor_ca_sse2 (uint32_t * pd,
- const uint32_t *ps,
- const uint32_t *pm,
- int w)
+static void
+sse2_combine_xor_ca (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * pd,
+ const uint32_t * ps,
+ const uint32_t * pm,
+ int w)
{
uint32_t s, m, d;
@@ -2299,11 +2273,13 @@ core_combine_xor_ca_sse2 (uint32_t * pd, }
}
-static force_inline void
-core_combine_add_ca_sse2 (uint32_t * pd,
- const uint32_t *ps,
- const uint32_t *pm,
- int w)
+static void
+sse2_combine_add_ca (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * pd,
+ const uint32_t * ps,
+ const uint32_t * pm,
+ int w)
{
uint32_t s, m, d;
@@ -2317,10 +2293,10 @@ core_combine_add_ca_sse2 (uint32_t * pd, m = *pm++;
d = *pd;
- *pd++ = pack_1x64_32 (
- _mm_adds_pu8 (pix_multiply_1x64 (unpack_32_1x64 (s),
- unpack_32_1x64 (m)),
- unpack_32_1x64 (d)));
+ *pd++ = pack_1x128_32 (
+ _mm_adds_epu8 (pix_multiply_1x128 (unpack_32_1x128 (s),
+ unpack_32_1x128 (m)),
+ unpack_32_1x128 (d)));
w--;
}
@@ -2355,36 +2331,20 @@ core_combine_add_ca_sse2 (uint32_t * pd, m = *pm++;
d = *pd;
- *pd++ = pack_1x64_32 (
- _mm_adds_pu8 (pix_multiply_1x64 (unpack_32_1x64 (s),
- unpack_32_1x64 (m)),
- unpack_32_1x64 (d)));
+ *pd++ = pack_1x128_32 (
+ _mm_adds_epu8 (pix_multiply_1x128 (unpack_32_1x128 (s),
+ unpack_32_1x128 (m)),
+ unpack_32_1x128 (d)));
w--;
}
}
-/* ---------------------------------------------------
- * fb_compose_setup_sSE2
- */
-static force_inline __m64
-create_mask_16_64 (uint16_t mask)
-{
- return _mm_set1_pi16 (mask);
-}
-
static force_inline __m128i
create_mask_16_128 (uint16_t mask)
{
return _mm_set1_epi16 (mask);
}
-static force_inline __m64
-create_mask_2x32_64 (uint32_t mask0,
- uint32_t mask1)
-{
- return _mm_set_pi32 (mask0, mask1);
-}
-
/* Work around a code generation bug in Sun Studio 12. */
#if defined(__SUNPRO_C) && (__SUNPRO_C >= 0x590)
# define create_mask_2x32_128(mask0, mask1) \
@@ -2398,276 +2358,6 @@ create_mask_2x32_128 (uint32_t mask0, }
#endif
-/* SSE2 code patch for fbcompose.c */
-
-static void
-sse2_combine_over_u (pixman_implementation_t *imp,
- pixman_op_t op,
- uint32_t * dst,
- const uint32_t * src,
- const uint32_t * mask,
- int width)
-{
- core_combine_over_u_sse2 (dst, src, mask, width);
- _mm_empty ();
-}
-
-static void
-sse2_combine_over_reverse_u (pixman_implementation_t *imp,
- pixman_op_t op,
- uint32_t * dst,
- const uint32_t * src,
- const uint32_t * mask,
- int width)
-{
- core_combine_over_reverse_u_sse2 (dst, src, mask, width);
- _mm_empty ();
-}
-
-static void
-sse2_combine_in_u (pixman_implementation_t *imp,
- pixman_op_t op,
- uint32_t * dst,
- const uint32_t * src,
- const uint32_t * mask,
- int width)
-{
- core_combine_in_u_sse2 (dst, src, mask, width);
- _mm_empty ();
-}
-
-static void
-sse2_combine_in_reverse_u (pixman_implementation_t *imp,
- pixman_op_t op,
- uint32_t * dst,
- const uint32_t * src,
- const uint32_t * mask,
- int width)
-{
- core_combine_reverse_in_u_sse2 (dst, src, mask, width);
- _mm_empty ();
-}
-
-static void
-sse2_combine_out_u (pixman_implementation_t *imp,
- pixman_op_t op,
- uint32_t * dst,
- const uint32_t * src,
- const uint32_t * mask,
- int width)
-{
- core_combine_out_u_sse2 (dst, src, mask, width);
- _mm_empty ();
-}
-
-static void
-sse2_combine_out_reverse_u (pixman_implementation_t *imp,
- pixman_op_t op,
- uint32_t * dst,
- const uint32_t * src,
- const uint32_t * mask,
- int width)
-{
- core_combine_reverse_out_u_sse2 (dst, src, mask, width);
- _mm_empty ();
-}
-
-static void
-sse2_combine_atop_u (pixman_implementation_t *imp,
- pixman_op_t op,
- uint32_t * dst,
- const uint32_t * src,
- const uint32_t * mask,
- int width)
-{
- core_combine_atop_u_sse2 (dst, src, mask, width);
- _mm_empty ();
-}
-
-static void
-sse2_combine_atop_reverse_u (pixman_implementation_t *imp,
- pixman_op_t op,
- uint32_t * dst,
- const uint32_t * src,
- const uint32_t * mask,
- int width)
-{
- core_combine_reverse_atop_u_sse2 (dst, src, mask, width);
- _mm_empty ();
-}
-
-static void
-sse2_combine_xor_u (pixman_implementation_t *imp,
- pixman_op_t op,
- uint32_t * dst,
- const uint32_t * src,
- const uint32_t * mask,
- int width)
-{
- core_combine_xor_u_sse2 (dst, src, mask, width);
- _mm_empty ();
-}
-
-static void
-sse2_combine_add_u (pixman_implementation_t *imp,
- pixman_op_t op,
- uint32_t * dst,
- const uint32_t * src,
- const uint32_t * mask,
- int width)
-{
- core_combine_add_u_sse2 (dst, src, mask, width);
- _mm_empty ();
-}
-
-static void
-sse2_combine_saturate_u (pixman_implementation_t *imp,
- pixman_op_t op,
- uint32_t * dst,
- const uint32_t * src,
- const uint32_t * mask,
- int width)
-{
- core_combine_saturate_u_sse2 (dst, src, mask, width);
- _mm_empty ();
-}
-
-static void
-sse2_combine_src_ca (pixman_implementation_t *imp,
- pixman_op_t op,
- uint32_t * dst,
- const uint32_t * src,
- const uint32_t * mask,
- int width)
-{
- core_combine_src_ca_sse2 (dst, src, mask, width);
- _mm_empty ();
-}
-
-static void
-sse2_combine_over_ca (pixman_implementation_t *imp,
- pixman_op_t op,
- uint32_t * dst,
- const uint32_t * src,
- const uint32_t * mask,
- int width)
-{
- core_combine_over_ca_sse2 (dst, src, mask, width);
- _mm_empty ();
-}
-
-static void
-sse2_combine_over_reverse_ca (pixman_implementation_t *imp,
- pixman_op_t op,
- uint32_t * dst,
- const uint32_t * src,
- const uint32_t * mask,
- int width)
-{
- core_combine_over_reverse_ca_sse2 (dst, src, mask, width);
- _mm_empty ();
-}
-
-static void
-sse2_combine_in_ca (pixman_implementation_t *imp,
- pixman_op_t op,
- uint32_t * dst,
- const uint32_t * src,
- const uint32_t * mask,
- int width)
-{
- core_combine_in_ca_sse2 (dst, src, mask, width);
- _mm_empty ();
-}
-
-static void
-sse2_combine_in_reverse_ca (pixman_implementation_t *imp,
- pixman_op_t op,
- uint32_t * dst,
- const uint32_t * src,
- const uint32_t * mask,
- int width)
-{
- core_combine_in_reverse_ca_sse2 (dst, src, mask, width);
- _mm_empty ();
-}
-
-static void
-sse2_combine_out_ca (pixman_implementation_t *imp,
- pixman_op_t op,
- uint32_t * dst,
- const uint32_t * src,
- const uint32_t * mask,
- int width)
-{
- core_combine_out_ca_sse2 (dst, src, mask, width);
- _mm_empty ();
-}
-
-static void
-sse2_combine_out_reverse_ca (pixman_implementation_t *imp,
- pixman_op_t op,
- uint32_t * dst,
- const uint32_t * src,
- const uint32_t * mask,
- int width)
-{
- core_combine_out_reverse_ca_sse2 (dst, src, mask, width);
- _mm_empty ();
-}
-
-static void
-sse2_combine_atop_ca (pixman_implementation_t *imp,
- pixman_op_t op,
- uint32_t * dst,
- const uint32_t * src,
- const uint32_t * mask,
- int width)
-{
- core_combine_atop_ca_sse2 (dst, src, mask, width);
- _mm_empty ();
-}
-
-static void
-sse2_combine_atop_reverse_ca (pixman_implementation_t *imp,
- pixman_op_t op,
- uint32_t * dst,
- const uint32_t * src,
- const uint32_t * mask,
- int width)
-{
- core_combine_reverse_atop_ca_sse2 (dst, src, mask, width);
- _mm_empty ();
-}
-
-static void
-sse2_combine_xor_ca (pixman_implementation_t *imp,
- pixman_op_t op,
- uint32_t * dst,
- const uint32_t * src,
- const uint32_t * mask,
- int width)
-{
- core_combine_xor_ca_sse2 (dst, src, mask, width);
- _mm_empty ();
-}
-
-static void
-sse2_combine_add_ca (pixman_implementation_t *imp,
- pixman_op_t op,
- uint32_t * dst,
- const uint32_t * src,
- const uint32_t * mask,
- int width)
-{
- core_combine_add_ca_sse2 (dst, src, mask, width);
- _mm_empty ();
-}
-
-/* -------------------------------------------------------------------
- * composite_over_n_8888
- */
-
static void
sse2_composite_over_n_8888 (pixman_implementation_t *imp,
pixman_op_t op,
@@ -2711,9 +2401,9 @@ sse2_composite_over_n_8888 (pixman_implementation_t *imp, while (w && (unsigned long)dst & 15)
{
d = *dst;
- *dst++ = pack_1x64_32 (over_1x64 (_mm_movepi64_pi64 (xmm_src),
- _mm_movepi64_pi64 (xmm_alpha),
- unpack_32_1x64 (d)));
+ *dst++ = pack_1x128_32 (over_1x128 (xmm_src,
+ xmm_alpha,
+ unpack_32_1x128 (d)));
w--;
}
@@ -2738,19 +2428,15 @@ sse2_composite_over_n_8888 (pixman_implementation_t *imp, while (w)
{
d = *dst;
- *dst++ = pack_1x64_32 (over_1x64 (_mm_movepi64_pi64 (xmm_src),
- _mm_movepi64_pi64 (xmm_alpha),
- unpack_32_1x64 (d)));
+ *dst++ = pack_1x128_32 (over_1x128 (xmm_src,
+ xmm_alpha,
+ unpack_32_1x128 (d)));
w--;
}
}
- _mm_empty ();
}
-/* ---------------------------------------------------------------------
- * composite_over_n_0565
- */
static void
sse2_composite_over_n_0565 (pixman_implementation_t *imp,
pixman_op_t op,
@@ -2796,9 +2482,9 @@ sse2_composite_over_n_0565 (pixman_implementation_t *imp, d = *dst;
*dst++ = pack_565_32_16 (
- pack_1x64_32 (over_1x64 (_mm_movepi64_pi64 (xmm_src),
- _mm_movepi64_pi64 (xmm_alpha),
- expand565_16_1x64 (d))));
+ pack_1x128_32 (over_1x128 (xmm_src,
+ xmm_alpha,
+ expand565_16_1x128 (d))));
w--;
}
@@ -2829,18 +2515,13 @@ sse2_composite_over_n_0565 (pixman_implementation_t *imp, {
d = *dst;
*dst++ = pack_565_32_16 (
- pack_1x64_32 (over_1x64 (_mm_movepi64_pi64 (xmm_src),
- _mm_movepi64_pi64 (xmm_alpha),
- expand565_16_1x64 (d))));
+ pack_1x128_32 (over_1x128 (xmm_src, xmm_alpha,
+ expand565_16_1x128 (d))));
}
}
- _mm_empty ();
}
-/* ------------------------------
- * composite_add_n_8888_8888_ca
- */
static void
sse2_composite_add_n_8888_8888_ca (pixman_implementation_t *imp,
pixman_op_t op,
@@ -2866,7 +2547,7 @@ sse2_composite_add_n_8888_8888_ca (pixman_implementation_t *imp, __m128i xmm_dst;
__m128i xmm_mask, xmm_mask_lo, xmm_mask_hi;
- __m64 mmx_src, mmx_alpha, mmx_mask, mmx_dest;
+ __m128i mmx_src, mmx_alpha, mmx_mask, mmx_dest;
src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format);
srca = src >> 24;
@@ -2882,8 +2563,8 @@ sse2_composite_add_n_8888_8888_ca (pixman_implementation_t *imp, xmm_src = _mm_unpacklo_epi8 (
create_mask_2x32_128 (src, src), _mm_setzero_si128 ());
xmm_alpha = expand_alpha_1x128 (xmm_src);
- mmx_src = _mm_movepi64_pi64 (xmm_src);
- mmx_alpha = _mm_movepi64_pi64 (xmm_alpha);
+ mmx_src = xmm_src;
+ mmx_alpha = xmm_alpha;
while (height--)
{
@@ -2902,11 +2583,12 @@ sse2_composite_add_n_8888_8888_ca (pixman_implementation_t *imp, {
d = *pd;
- mmx_mask = unpack_32_1x64 (m);
- mmx_dest = unpack_32_1x64 (d);
+ mmx_mask = unpack_32_1x128 (m);
+ mmx_dest = unpack_32_1x128 (d);
- *pd = pack_1x64_32 (
- _mm_adds_pu8 (pix_multiply_1x64 (mmx_mask, mmx_src), mmx_dest));
+ *pd = pack_1x128_32 (
+ _mm_adds_epu8 (pix_multiply_1x128 (mmx_mask, mmx_src),
+ mmx_dest));
}
pd++;
@@ -2950,11 +2632,12 @@ sse2_composite_add_n_8888_8888_ca (pixman_implementation_t *imp, {
d = *pd;
- mmx_mask = unpack_32_1x64 (m);
- mmx_dest = unpack_32_1x64 (d);
+ mmx_mask = unpack_32_1x128 (m);
+ mmx_dest = unpack_32_1x128 (d);
- *pd = pack_1x64_32 (
- _mm_adds_pu8 (pix_multiply_1x64 (mmx_mask, mmx_src), mmx_dest));
+ *pd = pack_1x128_32 (
+ _mm_adds_epu8 (pix_multiply_1x128 (mmx_mask, mmx_src),
+ mmx_dest));
}
pd++;
@@ -2962,13 +2645,8 @@ sse2_composite_add_n_8888_8888_ca (pixman_implementation_t *imp, }
}
- _mm_empty ();
}
-/* ---------------------------------------------------------------------------
- * composite_over_n_8888_8888_ca
- */
-
static void
sse2_composite_over_n_8888_8888_ca (pixman_implementation_t *imp,
pixman_op_t op,
@@ -2994,7 +2672,7 @@ sse2_composite_over_n_8888_8888_ca (pixman_implementation_t *imp, __m128i xmm_dst, xmm_dst_lo, xmm_dst_hi;
__m128i xmm_mask, xmm_mask_lo, xmm_mask_hi;
- __m64 mmx_src, mmx_alpha, mmx_mask, mmx_dest;
+ __m128i mmx_src, mmx_alpha, mmx_mask, mmx_dest;
src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format);
@@ -3009,8 +2687,8 @@ sse2_composite_over_n_8888_8888_ca (pixman_implementation_t *imp, xmm_src = _mm_unpacklo_epi8 (
create_mask_2x32_128 (src, src), _mm_setzero_si128 ());
xmm_alpha = expand_alpha_1x128 (xmm_src);
- mmx_src = _mm_movepi64_pi64 (xmm_src);
- mmx_alpha = _mm_movepi64_pi64 (xmm_alpha);
+ mmx_src = xmm_src;
+ mmx_alpha = xmm_alpha;
while (height--)
{
@@ -3028,10 +2706,10 @@ sse2_composite_over_n_8888_8888_ca (pixman_implementation_t *imp, if (m)
{
d = *pd;
- mmx_mask = unpack_32_1x64 (m);
- mmx_dest = unpack_32_1x64 (d);
+ mmx_mask = unpack_32_1x128 (m);
+ mmx_dest = unpack_32_1x128 (d);
- *pd = pack_1x64_32 (in_over_1x64 (&mmx_src,
+ *pd = pack_1x128_32 (in_over_1x128 (&mmx_src,
&mmx_alpha,
&mmx_mask,
&mmx_dest));
@@ -3078,11 +2756,11 @@ sse2_composite_over_n_8888_8888_ca (pixman_implementation_t *imp, if (m)
{
d = *pd;
- mmx_mask = unpack_32_1x64 (m);
- mmx_dest = unpack_32_1x64 (d);
+ mmx_mask = unpack_32_1x128 (m);
+ mmx_dest = unpack_32_1x128 (d);
- *pd = pack_1x64_32 (
- in_over_1x64 (&mmx_src, &mmx_alpha, &mmx_mask, &mmx_dest));
+ *pd = pack_1x128_32 (
+ in_over_1x128 (&mmx_src, &mmx_alpha, &mmx_mask, &mmx_dest));
}
pd++;
@@ -3090,13 +2768,8 @@ sse2_composite_over_n_8888_8888_ca (pixman_implementation_t *imp, }
}
- _mm_empty ();
}
-/*---------------------------------------------------------------------
- * composite_over_8888_n_8888
- */
-
static void
sse2_composite_over_8888_n_8888 (pixman_implementation_t *imp,
pixman_op_t op,
@@ -3148,13 +2821,13 @@ sse2_composite_over_8888_n_8888 (pixman_implementation_t *imp, {
uint32_t d = *dst;
- __m64 ms = unpack_32_1x64 (s);
- __m64 alpha = expand_alpha_1x64 (ms);
- __m64 dest = _mm_movepi64_pi64 (xmm_mask);
- __m64 alpha_dst = unpack_32_1x64 (d);
+ __m128i ms = unpack_32_1x128 (s);
+ __m128i alpha = expand_alpha_1x128 (ms);
+ __m128i dest = xmm_mask;
+ __m128i alpha_dst = unpack_32_1x128 (d);
- *dst = pack_1x64_32 (
- in_over_1x64 (&ms, &alpha, &dest, &alpha_dst));
+ *dst = pack_1x128_32 (
+ in_over_1x128 (&ms, &alpha, &dest, &alpha_dst));
}
dst++;
w--;
@@ -3195,13 +2868,13 @@ sse2_composite_over_8888_n_8888 (pixman_implementation_t *imp, {
uint32_t d = *dst;
- __m64 ms = unpack_32_1x64 (s);
- __m64 alpha = expand_alpha_1x64 (ms);
- __m64 mask = _mm_movepi64_pi64 (xmm_mask);
- __m64 dest = unpack_32_1x64 (d);
+ __m128i ms = unpack_32_1x128 (s);
+ __m128i alpha = expand_alpha_1x128 (ms);
+ __m128i mask = xmm_mask;
+ __m128i dest = unpack_32_1x128 (d);
- *dst = pack_1x64_32 (
- in_over_1x64 (&ms, &alpha, &mask, &dest));
+ *dst = pack_1x128_32 (
+ in_over_1x128 (&ms, &alpha, &mask, &dest));
}
dst++;
@@ -3209,13 +2882,8 @@ sse2_composite_over_8888_n_8888 (pixman_implementation_t *imp, }
}
- _mm_empty ();
}
-/*---------------------------------------------------------------------
- * composite_over_8888_n_8888
- */
-
static void
sse2_composite_src_x888_8888 (pixman_implementation_t *imp,
pixman_op_t op,
@@ -3282,12 +2950,8 @@ sse2_composite_src_x888_8888 (pixman_implementation_t *imp, }
}
- _mm_empty ();
}
-/* ---------------------------------------------------------------------
- * composite_over_x888_n_8888
- */
static void
sse2_composite_over_x888_n_8888 (pixman_implementation_t *imp,
pixman_op_t op,
@@ -3336,13 +3000,13 @@ sse2_composite_over_x888_n_8888 (pixman_implementation_t *imp, uint32_t s = (*src++) | 0xff000000;
uint32_t d = *dst;
- __m64 src = unpack_32_1x64 (s);
- __m64 alpha = _mm_movepi64_pi64 (xmm_alpha);
- __m64 mask = _mm_movepi64_pi64 (xmm_mask);
- __m64 dest = unpack_32_1x64 (d);
+ __m128i src = unpack_32_1x128 (s);
+ __m128i alpha = xmm_alpha;
+ __m128i mask = xmm_mask;
+ __m128i dest = unpack_32_1x128 (d);
- *dst++ = pack_1x64_32 (
- in_over_1x64 (&src, &alpha, &mask, &dest));
+ *dst++ = pack_1x128_32 (
+ in_over_1x128 (&src, &alpha, &mask, &dest));
w--;
}
@@ -3375,24 +3039,20 @@ sse2_composite_over_x888_n_8888 (pixman_implementation_t *imp, uint32_t s = (*src++) | 0xff000000;
uint32_t d = *dst;
- __m64 src = unpack_32_1x64 (s);
- __m64 alpha = _mm_movepi64_pi64 (xmm_alpha);
- __m64 mask = _mm_movepi64_pi64 (xmm_mask);
- __m64 dest = unpack_32_1x64 (d);
+ __m128i src = unpack_32_1x128 (s);
+ __m128i alpha = xmm_alpha;
+ __m128i mask = xmm_mask;
+ __m128i dest = unpack_32_1x128 (d);
- *dst++ = pack_1x64_32 (
- in_over_1x64 (&src, &alpha, &mask, &dest));
+ *dst++ = pack_1x128_32 (
+ in_over_1x128 (&src, &alpha, &mask, &dest));
w--;
}
}
- _mm_empty ();
}
-/* --------------------------------------------------------------------
- * composite_over_8888_8888
- */
static void
sse2_composite_over_8888_8888 (pixman_implementation_t *imp,
pixman_op_t op,
@@ -3422,27 +3082,23 @@ sse2_composite_over_8888_8888 (pixman_implementation_t *imp, while (height--)
{
- core_combine_over_u_sse2 (dst, src, NULL, width);
+ sse2_combine_over_u (imp, op, dst, src, NULL, width);
dst += dst_stride;
src += src_stride;
}
- _mm_empty ();
}
-/* ------------------------------------------------------------------
- * composite_over_8888_0565
- */
static force_inline uint16_t
composite_over_8888_0565pixel (uint32_t src, uint16_t dst)
{
- __m64 ms;
+ __m128i ms;
- ms = unpack_32_1x64 (src);
+ ms = unpack_32_1x128 (src);
return pack_565_32_16 (
- pack_1x64_32 (
- over_1x64 (
- ms, expand_alpha_1x64 (ms), expand565_16_1x64 (dst))));
+ pack_1x128_32 (
+ over_1x128 (
+ ms, expand_alpha_1x128 (ms), expand565_16_1x128 (dst))));
}
static void
@@ -3474,15 +3130,6 @@ sse2_composite_over_8888_0565 (pixman_implementation_t *imp, PIXMAN_IMAGE_GET_LINE (
src_image, src_x, src_y, uint32_t, src_stride, src_line, 1);
-#if 0
- /* FIXME
- *
- * I copy the code from MMX one and keep the fixme.
- * If it's a problem there, probably is a problem here.
- */
- assert (src_image->drawable == mask_image->drawable);
-#endif
-
while (height--)
{
dst = dst_line;
@@ -3555,13 +3202,8 @@ sse2_composite_over_8888_0565 (pixman_implementation_t *imp, }
}
- _mm_empty ();
}
-/* -----------------------------------------------------------------
- * composite_over_n_8_8888
- */
-
static void
sse2_composite_over_n_8_8888 (pixman_implementation_t *imp,
pixman_op_t op,
@@ -3588,7 +3230,7 @@ sse2_composite_over_n_8_8888 (pixman_implementation_t *imp, __m128i xmm_dst, xmm_dst_lo, xmm_dst_hi;
__m128i xmm_mask, xmm_mask_lo, xmm_mask_hi;
- __m64 mmx_src, mmx_alpha, mmx_mask, mmx_dest;
+ __m128i mmx_src, mmx_alpha, mmx_mask, mmx_dest;
src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format);
@@ -3604,8 +3246,8 @@ sse2_composite_over_n_8_8888 (pixman_implementation_t *imp, xmm_def = create_mask_2x32_128 (src, src);
xmm_src = expand_pixel_32_1x128 (src);
xmm_alpha = expand_alpha_1x128 (xmm_src);
- mmx_src = _mm_movepi64_pi64 (xmm_src);
- mmx_alpha = _mm_movepi64_pi64 (xmm_alpha);
+ mmx_src = xmm_src;
+ mmx_alpha = xmm_alpha;
while (height--)
{
@@ -3622,10 +3264,10 @@ sse2_composite_over_n_8_8888 (pixman_implementation_t *imp, if (m)
{
d = *dst;
- mmx_mask = expand_pixel_8_1x64 (m);
- mmx_dest = unpack_32_1x64 (d);
+ mmx_mask = expand_pixel_8_1x128 (m);
+ mmx_dest = unpack_32_1x128 (d);
- *dst = pack_1x64_32 (in_over_1x64 (&mmx_src,
+ *dst = pack_1x128_32 (in_over_1x128 (&mmx_src,
&mmx_alpha,
&mmx_mask,
&mmx_dest));
@@ -3677,10 +3319,10 @@ sse2_composite_over_n_8_8888 (pixman_implementation_t *imp, if (m)
{
d = *dst;
- mmx_mask = expand_pixel_8_1x64 (m);
- mmx_dest = unpack_32_1x64 (d);
+ mmx_mask = expand_pixel_8_1x128 (m);
+ mmx_dest = unpack_32_1x128 (d);
- *dst = pack_1x64_32 (in_over_1x64 (&mmx_src,
+ *dst = pack_1x128_32 (in_over_1x128 (&mmx_src,
&mmx_alpha,
&mmx_mask,
&mmx_dest));
@@ -3691,14 +3333,9 @@ sse2_composite_over_n_8_8888 (pixman_implementation_t *imp, }
}
- _mm_empty ();
}
-/* ----------------------------------------------------------------
- * composite_over_n_8_8888
- */
-
-pixman_bool_t
+static pixman_bool_t
pixman_fill_sse2 (uint32_t *bits,
int stride,
int bpp,
@@ -3845,7 +3482,6 @@ pixman_fill_sse2 (uint32_t *bits, }
}
- _mm_empty ();
return TRUE;
}
@@ -3907,9 +3543,8 @@ sse2_composite_src_n_8_8888 (pixman_implementation_t *imp, if (m)
{
- *dst = pack_1x64_32 (
- pix_multiply_1x64 (
- _mm_movepi64_pi64 (xmm_src), expand_pixel_8_1x64 (m)));
+ *dst = pack_1x128_32 (
+ pix_multiply_1x128 (xmm_src, expand_pixel_8_1x128 (m)));
}
else
{
@@ -3962,9 +3597,9 @@ sse2_composite_src_n_8_8888 (pixman_implementation_t *imp, if (m)
{
- *dst = pack_1x64_32 (
- pix_multiply_1x64 (
- _mm_movepi64_pi64 (xmm_src), expand_pixel_8_1x64 (m)));
+ *dst = pack_1x128_32 (
+ pix_multiply_1x128 (
+ xmm_src, expand_pixel_8_1x128 (m)));
}
else
{
@@ -3976,13 +3611,8 @@ sse2_composite_src_n_8_8888 (pixman_implementation_t *imp, }
}
- _mm_empty ();
}
-/*-----------------------------------------------------------------------
- * composite_over_n_8_0565
- */
-
static void
sse2_composite_over_n_8_0565 (pixman_implementation_t *imp,
pixman_op_t op,
@@ -4004,7 +3634,7 @@ sse2_composite_over_n_8_0565 (pixman_implementation_t *imp, int dst_stride, mask_stride;
int32_t w;
uint32_t m;
- __m64 mmx_src, mmx_alpha, mmx_mask, mmx_dest;
+ __m128i mmx_src, mmx_alpha, mmx_mask, mmx_dest;
__m128i xmm_src, xmm_alpha;
__m128i xmm_mask, xmm_mask_lo, xmm_mask_hi;
@@ -4023,8 +3653,8 @@ sse2_composite_over_n_8_0565 (pixman_implementation_t *imp, xmm_src = expand_pixel_32_1x128 (src);
xmm_alpha = expand_alpha_1x128 (xmm_src);
- mmx_src = _mm_movepi64_pi64 (xmm_src);
- mmx_alpha = _mm_movepi64_pi64 (xmm_alpha);
+ mmx_src = xmm_src;
+ mmx_alpha = xmm_alpha;
while (height--)
{
@@ -4041,12 +3671,12 @@ sse2_composite_over_n_8_0565 (pixman_implementation_t *imp, if (m)
{
d = *dst;
- mmx_mask = expand_alpha_rev_1x64 (unpack_32_1x64 (m));
- mmx_dest = expand565_16_1x64 (d);
+ mmx_mask = expand_alpha_rev_1x128 (unpack_32_1x128 (m));
+ mmx_dest = expand565_16_1x128 (d);
*dst = pack_565_32_16 (
- pack_1x64_32 (
- in_over_1x64 (
+ pack_1x128_32 (
+ in_over_1x128 (
&mmx_src, &mmx_alpha, &mmx_mask, &mmx_dest)));
}
@@ -4114,12 +3744,12 @@ sse2_composite_over_n_8_0565 (pixman_implementation_t *imp, if (m)
{
d = *dst;
- mmx_mask = expand_alpha_rev_1x64 (unpack_32_1x64 (m));
- mmx_dest = expand565_16_1x64 (d);
+ mmx_mask = expand_alpha_rev_1x128 (unpack_32_1x128 (m));
+ mmx_dest = expand565_16_1x128 (d);
*dst = pack_565_32_16 (
- pack_1x64_32 (
- in_over_1x64 (
+ pack_1x128_32 (
+ in_over_1x128 (
&mmx_src, &mmx_alpha, &mmx_mask, &mmx_dest)));
}
@@ -4128,13 +3758,8 @@ sse2_composite_over_n_8_0565 (pixman_implementation_t *imp, }
}
- _mm_empty ();
}
-/* -----------------------------------------------------------------------
- * composite_over_pixbuf_0565
- */
-
static void
sse2_composite_over_pixbuf_0565 (pixman_implementation_t *imp,
pixman_op_t op,
@@ -4156,7 +3781,7 @@ sse2_composite_over_pixbuf_0565 (pixman_implementation_t *imp, int32_t w;
uint32_t opaque, zero;
- __m64 ms;
+ __m128i ms;
__m128i xmm_src, xmm_src_lo, xmm_src_hi;
__m128i xmm_dst, xmm_dst0, xmm_dst1, xmm_dst2, xmm_dst3;
@@ -4165,15 +3790,6 @@ sse2_composite_over_pixbuf_0565 (pixman_implementation_t *imp, PIXMAN_IMAGE_GET_LINE (
src_image, src_x, src_y, uint32_t, src_stride, src_line, 1);
-#if 0
- /* FIXME
- *
- * I copy the code from MMX one and keep the fixme.
- * If it's a problem there, probably is a problem here.
- */
- assert (src_image->drawable == mask_image->drawable);
-#endif
-
while (height--)
{
dst = dst_line;
@@ -4187,11 +3803,11 @@ sse2_composite_over_pixbuf_0565 (pixman_implementation_t *imp, s = *src++;
d = *dst;
- ms = unpack_32_1x64 (s);
+ ms = unpack_32_1x128 (s);
*dst++ = pack_565_32_16 (
- pack_1x64_32 (
- over_rev_non_pre_1x64 (ms, expand565_16_1x64 (d))));
+ pack_1x128_32 (
+ over_rev_non_pre_1x128 (ms, expand565_16_1x128 (d))));
w--;
}
@@ -4253,22 +3869,17 @@ sse2_composite_over_pixbuf_0565 (pixman_implementation_t *imp, s = *src++;
d = *dst;
- ms = unpack_32_1x64 (s);
+ ms = unpack_32_1x128 (s);
*dst++ = pack_565_32_16 (
- pack_1x64_32 (
- over_rev_non_pre_1x64 (ms, expand565_16_1x64 (d))));
+ pack_1x128_32 (
+ over_rev_non_pre_1x128 (ms, expand565_16_1x128 (d))));
w--;
}
}
- _mm_empty ();
}
-/* -------------------------------------------------------------------------
- * composite_over_pixbuf_8888
- */
-
static void
sse2_composite_over_pixbuf_8888 (pixman_implementation_t *imp,
pixman_op_t op,
@@ -4298,15 +3909,6 @@ sse2_composite_over_pixbuf_8888 (pixman_implementation_t *imp, PIXMAN_IMAGE_GET_LINE (
src_image, src_x, src_y, uint32_t, src_stride, src_line, 1);
-#if 0
- /* FIXME
- *
- * I copy the code from MMX one and keep the fixme.
- * If it's a problem there, probably is a problem here.
- */
- assert (src_image->drawable == mask_image->drawable);
-#endif
-
while (height--)
{
dst = dst_line;
@@ -4320,9 +3922,9 @@ sse2_composite_over_pixbuf_8888 (pixman_implementation_t *imp, s = *src++;
d = *dst;
- *dst++ = pack_1x64_32 (
- over_rev_non_pre_1x64 (
- unpack_32_1x64 (s), unpack_32_1x64 (d)));
+ *dst++ = pack_1x128_32 (
+ over_rev_non_pre_1x128 (
+ unpack_32_1x128 (s), unpack_32_1x128 (d)));
w--;
}
@@ -4367,21 +3969,16 @@ sse2_composite_over_pixbuf_8888 (pixman_implementation_t *imp, s = *src++;
d = *dst;
- *dst++ = pack_1x64_32 (
- over_rev_non_pre_1x64 (
- unpack_32_1x64 (s), unpack_32_1x64 (d)));
+ *dst++ = pack_1x128_32 (
+ over_rev_non_pre_1x128 (
+ unpack_32_1x128 (s), unpack_32_1x128 (d)));
w--;
}
}
- _mm_empty ();
}
-/* -------------------------------------------------------------------------------------------------
- * composite_over_n_8888_0565_ca
- */
-
static void
sse2_composite_over_n_8888_0565_ca (pixman_implementation_t *imp,
pixman_op_t op,
@@ -4408,7 +4005,7 @@ sse2_composite_over_n_8888_0565_ca (pixman_implementation_t *imp, __m128i xmm_mask, xmm_mask_lo, xmm_mask_hi;
__m128i xmm_dst, xmm_dst0, xmm_dst1, xmm_dst2, xmm_dst3;
- __m64 mmx_src, mmx_alpha, mmx_mask, mmx_dest;
+ __m128i mmx_src, mmx_alpha, mmx_mask, mmx_dest;
src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format);
@@ -4422,8 +4019,8 @@ sse2_composite_over_n_8888_0565_ca (pixman_implementation_t *imp, xmm_src = expand_pixel_32_1x128 (src);
xmm_alpha = expand_alpha_1x128 (xmm_src);
- mmx_src = _mm_movepi64_pi64 (xmm_src);
- mmx_alpha = _mm_movepi64_pi64 (xmm_alpha);
+ mmx_src = xmm_src;
+ mmx_alpha = xmm_alpha;
while (height--)
{
@@ -4440,12 +4037,12 @@ sse2_composite_over_n_8888_0565_ca (pixman_implementation_t *imp, if (m)
{
d = *dst;
- mmx_mask = unpack_32_1x64 (m);
- mmx_dest = expand565_16_1x64 (d);
+ mmx_mask = unpack_32_1x128 (m);
+ mmx_dest = expand565_16_1x128 (d);
*dst = pack_565_32_16 (
- pack_1x64_32 (
- in_over_1x64 (
+ pack_1x128_32 (
+ in_over_1x128 (
&mmx_src, &mmx_alpha, &mmx_mask, &mmx_dest)));
}
@@ -4509,12 +4106,12 @@ sse2_composite_over_n_8888_0565_ca (pixman_implementation_t *imp, if (m)
{
d = *dst;
- mmx_mask = unpack_32_1x64 (m);
- mmx_dest = expand565_16_1x64 (d);
+ mmx_mask = unpack_32_1x128 (m);
+ mmx_dest = expand565_16_1x128 (d);
*dst = pack_565_32_16 (
- pack_1x64_32 (
- in_over_1x64 (
+ pack_1x128_32 (
+ in_over_1x128 (
&mmx_src, &mmx_alpha, &mmx_mask, &mmx_dest)));
}
@@ -4524,13 +4121,8 @@ sse2_composite_over_n_8888_0565_ca (pixman_implementation_t *imp, }
}
- _mm_empty ();
}
-/* -----------------------------------------------------------------------
- * composite_in_n_8_8
- */
-
static void
sse2_composite_in_n_8_8 (pixman_implementation_t *imp,
pixman_op_t op,
@@ -4582,11 +4174,11 @@ sse2_composite_in_n_8_8 (pixman_implementation_t *imp, m = (uint32_t) *mask++;
d = (uint32_t) *dst;
- *dst++ = (uint8_t) pack_1x64_32 (
- pix_multiply_1x64 (
- pix_multiply_1x64 (_mm_movepi64_pi64 (xmm_alpha),
- unpack_32_1x64 (m)),
- unpack_32_1x64 (d)));
+ *dst++ = (uint8_t) pack_1x128_32 (
+ pix_multiply_1x128 (
+ pix_multiply_1x128 (xmm_alpha,
+ unpack_32_1x128 (m)),
+ unpack_32_1x128 (d)));
w--;
}
@@ -4619,22 +4211,17 @@ sse2_composite_in_n_8_8 (pixman_implementation_t *imp, m = (uint32_t) *mask++;
d = (uint32_t) *dst;
- *dst++ = (uint8_t) pack_1x64_32 (
- pix_multiply_1x64 (
- pix_multiply_1x64 (
- _mm_movepi64_pi64 (xmm_alpha), unpack_32_1x64 (m)),
- unpack_32_1x64 (d)));
+ *dst++ = (uint8_t) pack_1x128_32 (
+ pix_multiply_1x128 (
+ pix_multiply_1x128 (
+ xmm_alpha, unpack_32_1x128 (m)),
+ unpack_32_1x128 (d)));
w--;
}
}
- _mm_empty ();
}
-/* -----------------------------------------------------------------------
- * composite_in_n_8
- */
-
static void
sse2_composite_in_n_8 (pixman_implementation_t *imp,
pixman_op_t op,
@@ -4689,10 +4276,10 @@ sse2_composite_in_n_8 (pixman_implementation_t *imp, {
d = (uint32_t) *dst;
- *dst++ = (uint8_t) pack_1x64_32 (
- pix_multiply_1x64 (
- _mm_movepi64_pi64 (xmm_alpha),
- unpack_32_1x64 (d)));
+ *dst++ = (uint8_t) pack_1x128_32 (
+ pix_multiply_1x128 (
+ xmm_alpha,
+ unpack_32_1x128 (d)));
w--;
}
@@ -4717,21 +4304,16 @@ sse2_composite_in_n_8 (pixman_implementation_t *imp, {
d = (uint32_t) *dst;
- *dst++ = (uint8_t) pack_1x64_32 (
- pix_multiply_1x64 (
- _mm_movepi64_pi64 (xmm_alpha),
- unpack_32_1x64 (d)));
+ *dst++ = (uint8_t) pack_1x128_32 (
+ pix_multiply_1x128 (
+ xmm_alpha,
+ unpack_32_1x128 (d)));
w--;
}
}
- _mm_empty ();
}
-/* ---------------------------------------------------------------------------
- * composite_in_8_8
- */
-
static void
sse2_composite_in_8_8 (pixman_implementation_t *imp,
pixman_op_t op,
@@ -4774,9 +4356,9 @@ sse2_composite_in_8_8 (pixman_implementation_t *imp, s = (uint32_t) *src++;
d = (uint32_t) *dst;
- *dst++ = (uint8_t) pack_1x64_32 (
- pix_multiply_1x64 (
- unpack_32_1x64 (s), unpack_32_1x64 (d)));
+ *dst++ = (uint8_t) pack_1x128_32 (
+ pix_multiply_1x128 (
+ unpack_32_1x128 (s), unpack_32_1x128 (d)));
w--;
}
@@ -4805,19 +4387,14 @@ sse2_composite_in_8_8 (pixman_implementation_t *imp, s = (uint32_t) *src++;
d = (uint32_t) *dst;
- *dst++ = (uint8_t) pack_1x64_32 (
- pix_multiply_1x64 (unpack_32_1x64 (s), unpack_32_1x64 (d)));
+ *dst++ = (uint8_t) pack_1x128_32 (
+ pix_multiply_1x128 (unpack_32_1x128 (s), unpack_32_1x128 (d)));
w--;
}
}
- _mm_empty ();
}
-/* -------------------------------------------------------------------------
- * composite_add_n_8_8
- */
-
static void
sse2_composite_add_n_8_8 (pixman_implementation_t *imp,
pixman_op_t op,
@@ -4869,11 +4446,11 @@ sse2_composite_add_n_8_8 (pixman_implementation_t *imp, m = (uint32_t) *mask++;
d = (uint32_t) *dst;
- *dst++ = (uint8_t) pack_1x64_32 (
- _mm_adds_pu16 (
- pix_multiply_1x64 (
- _mm_movepi64_pi64 (xmm_alpha), unpack_32_1x64 (m)),
- unpack_32_1x64 (d)));
+ *dst++ = (uint8_t) pack_1x128_32 (
+ _mm_adds_epu16 (
+ pix_multiply_1x128 (
+ xmm_alpha, unpack_32_1x128 (m)),
+ unpack_32_1x128 (d)));
w--;
}
@@ -4905,23 +4482,18 @@ sse2_composite_add_n_8_8 (pixman_implementation_t *imp, m = (uint32_t) *mask++;
d = (uint32_t) *dst;
- *dst++ = (uint8_t) pack_1x64_32 (
- _mm_adds_pu16 (
- pix_multiply_1x64 (
- _mm_movepi64_pi64 (xmm_alpha), unpack_32_1x64 (m)),
- unpack_32_1x64 (d)));
+ *dst++ = (uint8_t) pack_1x128_32 (
+ _mm_adds_epu16 (
+ pix_multiply_1x128 (
+ xmm_alpha, unpack_32_1x128 (m)),
+ unpack_32_1x128 (d)));
w--;
}
}
- _mm_empty ();
}
-/* -------------------------------------------------------------------------
- * composite_add_n_8_8
- */
-
static void
sse2_composite_add_n_8 (pixman_implementation_t *imp,
pixman_op_t op,
@@ -4973,10 +4545,10 @@ sse2_composite_add_n_8 (pixman_implementation_t *imp, while (w && ((unsigned long)dst & 15))
{
- *dst = (uint8_t)_mm_cvtsi64_si32 (
- _mm_adds_pu8 (
- _mm_movepi64_pi64 (xmm_src),
- _mm_cvtsi32_si64 (*dst)));
+ *dst = (uint8_t)_mm_cvtsi128_si32 (
+ _mm_adds_epu8 (
+ xmm_src,
+ _mm_cvtsi32_si128 (*dst)));
w--;
dst++;
@@ -4993,23 +4565,18 @@ sse2_composite_add_n_8 (pixman_implementation_t *imp, while (w)
{
- *dst = (uint8_t)_mm_cvtsi64_si32 (
- _mm_adds_pu8 (
- _mm_movepi64_pi64 (xmm_src),
- _mm_cvtsi32_si64 (*dst)));
+ *dst = (uint8_t)_mm_cvtsi128_si32 (
+ _mm_adds_epu8 (
+ xmm_src,
+ _mm_cvtsi32_si128 (*dst)));
w--;
dst++;
}
}
- _mm_empty ();
}
-/* ----------------------------------------------------------------------
- * composite_add_8_8
- */
-
static void
sse2_composite_add_8_8 (pixman_implementation_t *imp,
pixman_op_t op,
@@ -5053,7 +4620,8 @@ sse2_composite_add_8_8 (pixman_implementation_t *imp, w--;
}
- core_combine_add_u_sse2 ((uint32_t*)dst, (uint32_t*)src, NULL, w >> 2);
+ sse2_combine_add_u (imp, op,
+ (uint32_t*)dst, (uint32_t*)src, NULL, w >> 2);
/* Small tail */
dst += w & 0xfffc;
@@ -5069,12 +4637,8 @@ sse2_composite_add_8_8 (pixman_implementation_t *imp, }
}
- _mm_empty ();
}
-/* ---------------------------------------------------------------------
- * composite_add_8888_8888
- */
static void
sse2_composite_add_8888_8888 (pixman_implementation_t *imp,
pixman_op_t op,
@@ -5106,16 +4670,11 @@ sse2_composite_add_8888_8888 (pixman_implementation_t *imp, src = src_line;
src_line += src_stride;
- core_combine_add_u_sse2 (dst, src, NULL, width);
+ sse2_combine_add_u (imp, op, dst, src, NULL, width);
}
- _mm_empty ();
}
-/* -------------------------------------------------------------------------------------------------
- * sse2_composite_copy_area
- */
-
static pixman_bool_t
pixman_blt_sse2 (uint32_t *src_bits,
uint32_t *dst_bits,
@@ -5234,7 +4793,6 @@ pixman_blt_sse2 (uint32_t *src_bits, }
}
- _mm_empty ();
return TRUE;
}
@@ -5284,7 +4842,7 @@ sse2_composite_over_x888_8_8888 (pixman_implementation_t *imp, uint32_t m;
int src_stride, mask_stride, dst_stride;
int32_t w;
- __m64 ms;
+ __m128i ms;
__m128i xmm_src, xmm_src_lo, xmm_src_hi;
__m128i xmm_dst, xmm_dst_lo, xmm_dst_hi;
@@ -5313,24 +4871,25 @@ sse2_composite_over_x888_8_8888 (pixman_implementation_t *imp, s = 0xff000000 | *src++;
m = (uint32_t) *mask++;
d = *dst;
- ms = unpack_32_1x64 (s);
+ ms = unpack_32_1x128 (s);
if (m != 0xff)
{
- __m64 ma = expand_alpha_rev_1x64 (unpack_32_1x64 (m));
- __m64 md = unpack_32_1x64 (d);
+ __m128i ma = expand_alpha_rev_1x128 (unpack_32_1x128 (m));
+ __m128i md = unpack_32_1x128 (d);
- ms = in_over_1x64 (&ms, &mask_x00ff, &ma, &md);
+ ms = in_over_1x128 (&ms, &mask_00ff, &ma, &md);
}
- *dst++ = pack_1x64_32 (ms);
+ *dst++ = pack_1x128_32 (ms);
w--;
}
while (w >= 4)
{
m = *(uint32_t*) mask;
- xmm_src = _mm_or_si128 (load_128_unaligned ((__m128i*)src), mask_ff000000);
+ xmm_src = _mm_or_si128 (
+ load_128_unaligned ((__m128i*)src), mask_ff000000);
if (m == 0xffffffff)
{
@@ -5346,9 +4905,12 @@ sse2_composite_over_x888_8_8888 (pixman_implementation_t *imp, unpack_128_2x128 (xmm_mask, &xmm_mask_lo, &xmm_mask_hi);
unpack_128_2x128 (xmm_dst, &xmm_dst_lo, &xmm_dst_hi);
- expand_alpha_rev_2x128 (xmm_mask_lo, xmm_mask_hi, &xmm_mask_lo, &xmm_mask_hi);
+ expand_alpha_rev_2x128 (
+ xmm_mask_lo, xmm_mask_hi, &xmm_mask_lo, &xmm_mask_hi);
- in_over_2x128 (&xmm_src_lo, &xmm_src_hi, &mask_00ff, &mask_00ff, &xmm_mask_lo, &xmm_mask_hi, &xmm_dst_lo, &xmm_dst_hi);
+ in_over_2x128 (&xmm_src_lo, &xmm_src_hi,
+ &mask_00ff, &mask_00ff, &xmm_mask_lo, &xmm_mask_hi,
+ &xmm_dst_lo, &xmm_dst_hi);
save_128_aligned ((__m128i*)dst, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi));
}
@@ -5373,15 +4935,15 @@ sse2_composite_over_x888_8_8888 (pixman_implementation_t *imp, }
else
{
- __m64 ma, md, ms;
+ __m128i ma, md, ms;
d = *dst;
- ma = expand_alpha_rev_1x64 (unpack_32_1x64 (m));
- md = unpack_32_1x64 (d);
- ms = unpack_32_1x64 (s);
+ ma = expand_alpha_rev_1x128 (unpack_32_1x128 (m));
+ md = unpack_32_1x128 (d);
+ ms = unpack_32_1x128 (s);
- *dst = pack_1x64_32 (in_over_1x64 (&ms, &mask_x00ff, &ma, &md));
+ *dst = pack_1x128_32 (in_over_1x128 (&ms, &mask_00ff, &ma, &md));
}
}
@@ -5392,7 +4954,6 @@ sse2_composite_over_x888_8_8888 (pixman_implementation_t *imp, }
}
- _mm_empty ();
}
static void
@@ -5457,15 +5018,15 @@ sse2_composite_over_8888_8_8888 (pixman_implementation_t *imp, }
else
{
- __m64 ms, md, ma, msa;
+ __m128i ms, md, ma, msa;
- ma = expand_alpha_rev_1x64 (load_32_1x64 (m));
- ms = unpack_32_1x64 (s);
- md = unpack_32_1x64 (d);
+ ma = expand_alpha_rev_1x128 (load_32_1x128 (m));
+ ms = unpack_32_1x128 (s);
+ md = unpack_32_1x128 (d);
- msa = expand_alpha_rev_1x64 (load_32_1x64 (sa));
+ msa = expand_alpha_rev_1x128 (load_32_1x128 (sa));
- *dst = pack_1x64_32 (in_over_1x64 (&ms, &msa, &ma, &md));
+ *dst = pack_1x128_32 (in_over_1x128 (&ms, &msa, &ma, &md));
}
}
@@ -5529,15 +5090,15 @@ sse2_composite_over_8888_8_8888 (pixman_implementation_t *imp, }
else
{
- __m64 ms, md, ma, msa;
+ __m128i ms, md, ma, msa;
- ma = expand_alpha_rev_1x64 (load_32_1x64 (m));
- ms = unpack_32_1x64 (s);
- md = unpack_32_1x64 (d);
+ ma = expand_alpha_rev_1x128 (load_32_1x128 (m));
+ ms = unpack_32_1x128 (s);
+ md = unpack_32_1x128 (d);
- msa = expand_alpha_rev_1x64 (load_32_1x64 (sa));
+ msa = expand_alpha_rev_1x128 (load_32_1x128 (sa));
- *dst = pack_1x64_32 (in_over_1x64 (&ms, &msa, &ma, &md));
+ *dst = pack_1x128_32 (in_over_1x128 (&ms, &msa, &ma, &md));
}
}
@@ -5546,7 +5107,6 @@ sse2_composite_over_8888_8_8888 (pixman_implementation_t *imp, }
}
- _mm_empty ();
}
static void
@@ -5591,12 +5151,12 @@ sse2_composite_over_reverse_n_8888 (pixman_implementation_t *imp, while (w && (unsigned long)dst & 15)
{
- __m64 vd;
+ __m128i vd;
- vd = unpack_32_1x64 (*dst);
+ vd = unpack_32_1x128 (*dst);
- *dst = pack_1x64_32 (over_1x64 (vd, expand_alpha_1x64 (vd),
- _mm_movepi64_pi64 (xmm_src)));
+ *dst = pack_1x128_32 (over_1x128 (vd, expand_alpha_1x128 (vd),
+ xmm_src));
w--;
dst++;
}
@@ -5626,19 +5186,18 @@ sse2_composite_over_reverse_n_8888 (pixman_implementation_t *imp, while (w)
{
- __m64 vd;
+ __m128i vd;
- vd = unpack_32_1x64 (*dst);
+ vd = unpack_32_1x128 (*dst);
- *dst = pack_1x64_32 (over_1x64 (vd, expand_alpha_1x64 (vd),
- _mm_movepi64_pi64 (xmm_src)));
+ *dst = pack_1x128_32 (over_1x128 (vd, expand_alpha_1x128 (vd),
+ xmm_src));
w--;
dst++;
}
}
- _mm_empty ();
}
static void
@@ -5703,15 +5262,15 @@ sse2_composite_over_8888_8888_8888 (pixman_implementation_t *imp, }
else
{
- __m64 ms, md, ma, msa;
+ __m128i ms, md, ma, msa;
- ma = expand_alpha_rev_1x64 (load_32_1x64 (m));
- ms = unpack_32_1x64 (s);
- md = unpack_32_1x64 (d);
+ ma = expand_alpha_rev_1x128 (load_32_1x128 (m));
+ ms = unpack_32_1x128 (s);
+ md = unpack_32_1x128 (d);
- msa = expand_alpha_rev_1x64 (load_32_1x64 (sa));
+ msa = expand_alpha_rev_1x128 (load_32_1x128 (sa));
- *dst = pack_1x64_32 (in_over_1x64 (&ms, &msa, &ma, &md));
+ *dst = pack_1x128_32 (in_over_1x128 (&ms, &msa, &ma, &md));
}
}
@@ -5773,15 +5332,15 @@ sse2_composite_over_8888_8888_8888 (pixman_implementation_t *imp, }
else
{
- __m64 ms, md, ma, msa;
+ __m128i ms, md, ma, msa;
- ma = expand_alpha_rev_1x64 (load_32_1x64 (m));
- ms = unpack_32_1x64 (s);
- md = unpack_32_1x64 (d);
+ ma = expand_alpha_rev_1x128 (load_32_1x128 (m));
+ ms = unpack_32_1x128 (s);
+ md = unpack_32_1x128 (d);
- msa = expand_alpha_rev_1x64 (load_32_1x64 (sa));
+ msa = expand_alpha_rev_1x128 (load_32_1x128 (sa));
- *dst = pack_1x64_32 (in_over_1x64 (&ms, &msa, &ma, &md));
+ *dst = pack_1x128_32 (in_over_1x128 (&ms, &msa, &ma, &md));
}
}
@@ -5790,10 +5349,9 @@ sse2_composite_over_8888_8888_8888 (pixman_implementation_t *imp, }
}
- _mm_empty ();
}
-/* A variant of 'core_combine_over_u_sse2' with minor tweaks */
+/* A variant of 'sse2_combine_over_u' with minor tweaks */
static force_inline void
scaled_nearest_scanline_sse2_8888_8888_OVER (uint32_t* pd,
const uint32_t* ps,
@@ -5885,7 +5443,6 @@ scaled_nearest_scanline_sse2_8888_8888_OVER (uint32_t* pd, w--;
}
- _mm_empty ();
}
FAST_NEAREST_MAINLOOP (sse2_8888_8888_cover_OVER,
@@ -5927,13 +5484,13 @@ scaled_nearest_scanline_sse2_8888_n_8888_OVER (const uint32_t * mask, {
uint32_t d = *dst;
- __m64 ms = unpack_32_1x64 (s);
- __m64 alpha = expand_alpha_1x64 (ms);
- __m64 dest = _mm_movepi64_pi64 (xmm_mask);
- __m64 alpha_dst = unpack_32_1x64 (d);
+ __m128i ms = unpack_32_1x128 (s);
+ __m128i alpha = expand_alpha_1x128 (ms);
+ __m128i dest = xmm_mask;
+ __m128i alpha_dst = unpack_32_1x128 (d);
- *dst = pack_1x64_32 (
- in_over_1x64 (&ms, &alpha, &dest, &alpha_dst));
+ *dst = pack_1x128_32 (
+ in_over_1x128 (&ms, &alpha, &dest, &alpha_dst));
}
dst++;
w--;
@@ -5985,20 +5542,19 @@ scaled_nearest_scanline_sse2_8888_n_8888_OVER (const uint32_t * mask, {
uint32_t d = *dst;
- __m64 ms = unpack_32_1x64 (s);
- __m64 alpha = expand_alpha_1x64 (ms);
- __m64 mask = _mm_movepi64_pi64 (xmm_mask);
- __m64 dest = unpack_32_1x64 (d);
+ __m128i ms = unpack_32_1x128 (s);
+ __m128i alpha = expand_alpha_1x128 (ms);
+ __m128i mask = xmm_mask;
+ __m128i dest = unpack_32_1x128 (d);
- *dst = pack_1x64_32 (
- in_over_1x64 (&ms, &alpha, &mask, &dest));
+ *dst = pack_1x128_32 (
+ in_over_1x128 (&ms, &alpha, &mask, &dest));
}
dst++;
w--;
}
- _mm_empty ();
}
FAST_NEAREST_MAINLOOP_COMMON (sse2_8888_n_8888_cover_OVER,
@@ -6374,20 +5930,7 @@ _pixman_implementation_create_sse2 (pixman_implementation_t *fallback) mask_ff000000 = create_mask_2x32_128 (0xff000000, 0xff000000);
mask_alpha = create_mask_2x32_128 (0x00ff0000, 0x00000000);
- /* MMX constants */
- mask_x565_rgb = create_mask_2x32_64 (0x000001f0, 0x003f001f);
- mask_x565_unpack = create_mask_2x32_64 (0x00000084, 0x04100840);
-
- mask_x0080 = create_mask_16_64 (0x0080);
- mask_x00ff = create_mask_16_64 (0x00ff);
- mask_x0101 = create_mask_16_64 (0x0101);
- mask_x_alpha = create_mask_2x32_64 (0x00ff0000, 0x00000000);
-
- _mm_empty ();
-
/* Set up function pointers */
-
- /* SSE code patch for fbcompose.c */
imp->combine_32[PIXMAN_OP_OVER] = sse2_combine_over_u;
imp->combine_32[PIXMAN_OP_OVER_REVERSE] = sse2_combine_over_reverse_u;
imp->combine_32[PIXMAN_OP_IN] = sse2_combine_in_u;
@@ -6420,5 +5963,3 @@ _pixman_implementation_create_sse2 (pixman_implementation_t *fallback) return imp;
}
-
-#endif /* USE_SSE2 */
diff --git a/xorg-server/dix/eventconvert.c b/xorg-server/dix/eventconvert.c index dff1a5631..1009550c3 100644 --- a/xorg-server/dix/eventconvert.c +++ b/xorg-server/dix/eventconvert.c @@ -263,8 +263,20 @@ eventToKeyButtonPointer(DeviceEvent *ev, xEvent **xi, int *count) num_events = (countValuators(ev, &first) + 5)/6; /* valuator ev */
if (num_events <= 0)
{
- *count = 0;
- return BadMatch;
+ switch (ev->type)
+ {
+ case ET_KeyPress:
+ case ET_KeyRelease:
+ case ET_ButtonPress:
+ case ET_ButtonRelease:
+ /* no axes is ok */
+ break;
+ case ET_Motion:
+ case ET_ProximityIn:
+ case ET_ProximityOut:
+ *count = 0;
+ return BadMatch;
+ }
}
num_events++; /* the actual event event */
diff --git a/xorg-server/test/input.c b/xorg-server/test/input.c index 879e14f2f..3e6619181 100644 --- a/xorg-server/test/input.c +++ b/xorg-server/test/input.c @@ -1,1083 +1,1222 @@ -/** - * Copyright © 2009 Red Hat, Inc. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING - * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER - * DEALINGS IN THE SOFTWARE. - */ - -#ifdef HAVE_DIX_CONFIG_H -#include <dix-config.h> -#endif - -#include <stdint.h> -#include <X11/X.h> -#include "misc.h" -#include "resource.h" -#include <X11/Xproto.h> -#include <X11/extensions/XI2proto.h> -#include <X11/Xatom.h> -#include "windowstr.h" -#include "inputstr.h" -#include "eventconvert.h" -#include "exevents.h" -#include "dixgrabs.h" -#include "eventstr.h" -#include "inpututils.h" -#include <glib.h> - -/** - * Init a device with axes. - * Verify values set on the device. - * - * Result: All axes set to default values (usually 0). - */ -static void dix_init_valuators(void) -{ - DeviceIntRec dev; - ValuatorClassPtr val; - const int num_axes = 2; - int i; - Atom atoms[MAX_VALUATORS] = { 0 }; - - - memset(&dev, 0, sizeof(DeviceIntRec)); - dev.type = MASTER_POINTER; /* claim it's a master to stop ptracccel */ - - g_assert(InitValuatorClassDeviceStruct(NULL, 0, atoms, 0, 0) == FALSE); - g_assert(InitValuatorClassDeviceStruct(&dev, num_axes, atoms, 0, Absolute)); - - val = dev.valuator; - g_assert(val); - g_assert(val->numAxes == num_axes); - g_assert(val->numMotionEvents == 0); - g_assert(val->axisVal); - - for (i = 0; i < num_axes; i++) - { - g_assert(val->axisVal[i] == 0); - g_assert(val->axes->min_value == NO_AXIS_LIMITS); - g_assert(val->axes->max_value == NO_AXIS_LIMITS); - g_assert(val->axes->mode == Absolute); - } - - g_assert(dev.last.numValuators == num_axes); -} - -/* just check the known success cases, and that error cases set the client's - * error value correctly. */ -static void dix_check_grab_values(void) -{ - ClientRec client; - GrabParameters param; - int rc; - - memset(&client, 0, sizeof(client)); - - param.grabtype = GRABTYPE_CORE; - param.this_device_mode = GrabModeSync; - param.other_devices_mode = GrabModeSync; - param.modifiers = AnyModifier; - param.ownerEvents = FALSE; - - rc = CheckGrabValues(&client, ¶m); - g_assert(rc == Success); - - param.this_device_mode = GrabModeAsync; - rc = CheckGrabValues(&client, ¶m); - g_assert(rc == Success); - - param.this_device_mode = GrabModeAsync + 1; - rc = CheckGrabValues(&client, ¶m); - g_assert(rc == BadValue); - g_assert(client.errorValue == param.this_device_mode); - g_assert(client.errorValue == GrabModeAsync + 1); - - param.this_device_mode = GrabModeSync; - param.other_devices_mode = GrabModeAsync; - rc = CheckGrabValues(&client, ¶m); - g_assert(rc == Success); - - param.other_devices_mode = GrabModeAsync + 1; - rc = CheckGrabValues(&client, ¶m); - g_assert(rc == BadValue); - g_assert(client.errorValue == param.other_devices_mode); - g_assert(client.errorValue == GrabModeAsync + 1); - - param.other_devices_mode = GrabModeSync; - - param.modifiers = 1 << 13; - rc = CheckGrabValues(&client, ¶m); - g_assert(rc == BadValue); - g_assert(client.errorValue == param.modifiers); - g_assert(client.errorValue == (1 << 13)); - - - param.modifiers = AnyModifier; - param.ownerEvents = TRUE; - rc = CheckGrabValues(&client, ¶m); - g_assert(rc == Success); - - param.ownerEvents = 3; - rc = CheckGrabValues(&client, ¶m); - g_assert(rc == BadValue); - g_assert(client.errorValue == param.ownerEvents); - g_assert(client.errorValue == 3); -} - - -/** - * Convert various internal events to the matching core event and verify the - * parameters. - */ -static void dix_event_to_core(int type) -{ - DeviceEvent ev; - xEvent core; - int time; - int x, y; - int rc; - int state; - int detail; - const int ROOT_WINDOW_ID = 0x100; - - /* EventToCore memsets the event to 0 */ -#define test_event() \ - g_assert(rc == Success); \ - g_assert(core.u.u.type == type); \ - g_assert(core.u.u.detail == detail); \ - g_assert(core.u.keyButtonPointer.time == time); \ - g_assert(core.u.keyButtonPointer.rootX == x); \ - g_assert(core.u.keyButtonPointer.rootY == y); \ - g_assert(core.u.keyButtonPointer.state == state); \ - g_assert(core.u.keyButtonPointer.eventX == 0); \ - g_assert(core.u.keyButtonPointer.eventY == 0); \ - g_assert(core.u.keyButtonPointer.root == ROOT_WINDOW_ID); \ - g_assert(core.u.keyButtonPointer.event == 0); \ - g_assert(core.u.keyButtonPointer.child == 0); \ - g_assert(core.u.keyButtonPointer.sameScreen == FALSE); - - x = 0; - y = 0; - time = 12345; - state = 0; - detail = 0; - - ev.header = 0xFF; - ev.length = sizeof(DeviceEvent); - ev.time = time; - ev.root_y = x; - ev.root_x = y; - SetBit(ev.valuators.mask, 0); - SetBit(ev.valuators.mask, 1); - ev.root = ROOT_WINDOW_ID; - ev.corestate = state; - ev.detail.key = detail; - - ev.type = type; - ev.detail.key = 0; - rc = EventToCore((InternalEvent*)&ev, &core); - test_event(); - - x = 1; - y = 2; - ev.root_x = x; - ev.root_y = y; - rc = EventToCore((InternalEvent*)&ev, &core); - test_event(); - - x = 0x7FFF; - y = 0x7FFF; - ev.root_x = x; - ev.root_y = y; - rc = EventToCore((InternalEvent*)&ev, &core); - test_event(); - - x = 0x8000; /* too high */ - y = 0x8000; /* too high */ - ev.root_x = x; - ev.root_y = y; - rc = EventToCore((InternalEvent*)&ev, &core); - g_assert(core.u.keyButtonPointer.rootX != x); - g_assert(core.u.keyButtonPointer.rootY != y); - - x = 0x7FFF; - y = 0x7FFF; - ev.root_x = x; - ev.root_y = y; - time = 0; - ev.time = time; - rc = EventToCore((InternalEvent*)&ev, &core); - test_event(); - - detail = 1; - ev.detail.key = detail; - rc = EventToCore((InternalEvent*)&ev, &core); - test_event(); - - detail = 0xFF; /* highest value */ - ev.detail.key = detail; - rc = EventToCore((InternalEvent*)&ev, &core); - test_event(); - - detail = 0xFFF; /* too big */ - ev.detail.key = detail; - rc = EventToCore((InternalEvent*)&ev, &core); - g_assert(rc == BadMatch); - - detail = 0xFF; /* too big */ - ev.detail.key = detail; - state = 0xFFFF; /* highest value */ - ev.corestate = state; - rc = EventToCore((InternalEvent*)&ev, &core); - test_event(); - - state = 0x10000; /* too big */ - ev.corestate = state; - rc = EventToCore((InternalEvent*)&ev, &core); - g_assert(core.u.keyButtonPointer.state != state); - g_assert(core.u.keyButtonPointer.state == (state & 0xFFFF)); - -#undef test_event -} - -static void dix_event_to_core_fail(int evtype, int expected_rc) -{ - DeviceEvent ev; - xEvent core; - int rc; - - ev.header = 0xFF; - ev.length = sizeof(DeviceEvent); - - ev.type = evtype; - rc = EventToCore((InternalEvent*)&ev, &core); - g_assert(rc == expected_rc); -} - -static void dix_event_to_core_conversion(void) -{ - dix_event_to_core_fail(0, BadImplementation); - dix_event_to_core_fail(1, BadImplementation); - dix_event_to_core_fail(ET_ProximityOut + 1, BadImplementation); - dix_event_to_core_fail(ET_ProximityIn, BadMatch); - dix_event_to_core_fail(ET_ProximityOut, BadMatch); - - dix_event_to_core(ET_KeyPress); - dix_event_to_core(ET_KeyRelease); - dix_event_to_core(ET_ButtonPress); - dix_event_to_core(ET_ButtonRelease); - dix_event_to_core(ET_Motion); -} - -static void xi2_struct_sizes(void) -{ -#define compare(req) \ - g_assert(sizeof(req) == sz_##req); - - compare(xXIQueryVersionReq); - compare(xXIWarpPointerReq); - compare(xXIChangeCursorReq); - compare(xXIChangeHierarchyReq); - compare(xXISetClientPointerReq); - compare(xXIGetClientPointerReq); - compare(xXISelectEventsReq); - compare(xXIQueryVersionReq); - compare(xXIQueryDeviceReq); - compare(xXISetFocusReq); - compare(xXIGetFocusReq); - compare(xXIGrabDeviceReq); - compare(xXIUngrabDeviceReq); - compare(xXIAllowEventsReq); - compare(xXIPassiveGrabDeviceReq); - compare(xXIPassiveUngrabDeviceReq); - compare(xXIListPropertiesReq); - compare(xXIChangePropertyReq); - compare(xXIDeletePropertyReq); - compare(xXIGetPropertyReq); - compare(xXIGetSelectedEventsReq); -#undef compare -} - - -static void dix_grab_matching(void) -{ - DeviceIntRec xi_all_devices, xi_all_master_devices, dev1, dev2; - GrabRec a, b; - BOOL rc; - - memset(&a, 0, sizeof(a)); - memset(&b, 0, sizeof(b)); - - /* different grabtypes must fail */ - a.grabtype = GRABTYPE_CORE; - b.grabtype = GRABTYPE_XI2; - rc = GrabMatchesSecond(&a, &b, FALSE); - g_assert(rc == FALSE); - rc = GrabMatchesSecond(&b, &a, FALSE); - g_assert(rc == FALSE); - - a.grabtype = GRABTYPE_XI; - b.grabtype = GRABTYPE_XI2; - rc = GrabMatchesSecond(&a, &b, FALSE); - g_assert(rc == FALSE); - rc = GrabMatchesSecond(&b, &a, FALSE); - g_assert(rc == FALSE); - - a.grabtype = GRABTYPE_XI; - b.grabtype = GRABTYPE_CORE; - rc = GrabMatchesSecond(&a, &b, FALSE); - g_assert(rc == FALSE); - rc = GrabMatchesSecond(&b, &a, FALSE); - g_assert(rc == FALSE); - - /* XI2 grabs for different devices must fail, regardless of ignoreDevice - * XI2 grabs for master devices must fail against a slave */ - memset(&xi_all_devices, 0, sizeof(DeviceIntRec)); - memset(&xi_all_master_devices, 0, sizeof(DeviceIntRec)); - memset(&dev1, 0, sizeof(DeviceIntRec)); - memset(&dev2, 0, sizeof(DeviceIntRec)); - - xi_all_devices.id = XIAllDevices; - xi_all_master_devices.id = XIAllMasterDevices; - dev1.id = 10; - dev1.type = SLAVE; - dev2.id = 11; - dev2.type = SLAVE; - - inputInfo.all_devices = &xi_all_devices; - inputInfo.all_master_devices = &xi_all_master_devices; - a.grabtype = GRABTYPE_XI2; - b.grabtype = GRABTYPE_XI2; - a.device = &dev1; - b.device = &dev2; - - rc = GrabMatchesSecond(&a, &b, FALSE); - g_assert(rc == FALSE); - - a.device = &dev2; - b.device = &dev1; - rc = GrabMatchesSecond(&a, &b, FALSE); - g_assert(rc == FALSE); - rc = GrabMatchesSecond(&a, &b, TRUE); - g_assert(rc == FALSE); - - a.device = inputInfo.all_master_devices; - b.device = &dev1; - rc = GrabMatchesSecond(&a, &b, FALSE); - g_assert(rc == FALSE); - rc = GrabMatchesSecond(&a, &b, TRUE); - g_assert(rc == FALSE); - - a.device = &dev1; - b.device = inputInfo.all_master_devices; - rc = GrabMatchesSecond(&a, &b, FALSE); - g_assert(rc == FALSE); - rc = GrabMatchesSecond(&a, &b, TRUE); - g_assert(rc == FALSE); - - /* ignoreDevice FALSE must fail for different devices for CORE and XI */ - a.grabtype = GRABTYPE_XI; - b.grabtype = GRABTYPE_XI; - a.device = &dev1; - b.device = &dev2; - a.modifierDevice = &dev1; - b.modifierDevice = &dev1; - rc = GrabMatchesSecond(&a, &b, FALSE); - g_assert(rc == FALSE); - - a.grabtype = GRABTYPE_CORE; - b.grabtype = GRABTYPE_CORE; - a.device = &dev1; - b.device = &dev2; - a.modifierDevice = &dev1; - b.modifierDevice = &dev1; - rc = GrabMatchesSecond(&a, &b, FALSE); - g_assert(rc == FALSE); - - /* ignoreDevice FALSE must fail for different modifier devices for CORE - * and XI */ - a.grabtype = GRABTYPE_XI; - b.grabtype = GRABTYPE_XI; - a.device = &dev1; - b.device = &dev1; - a.modifierDevice = &dev1; - b.modifierDevice = &dev2; - rc = GrabMatchesSecond(&a, &b, FALSE); - g_assert(rc == FALSE); - - a.grabtype = GRABTYPE_CORE; - b.grabtype = GRABTYPE_CORE; - a.device = &dev1; - b.device = &dev1; - a.modifierDevice = &dev1; - b.modifierDevice = &dev2; - rc = GrabMatchesSecond(&a, &b, FALSE); - g_assert(rc == FALSE); - - /* different event type must fail */ - a.grabtype = GRABTYPE_XI2; - b.grabtype = GRABTYPE_XI2; - a.device = &dev1; - b.device = &dev1; - a.modifierDevice = &dev1; - b.modifierDevice = &dev1; - a.type = XI_KeyPress; - b.type = XI_KeyRelease; - rc = GrabMatchesSecond(&a, &b, FALSE); - g_assert(rc == FALSE); - rc = GrabMatchesSecond(&a, &b, TRUE); - g_assert(rc == FALSE); - - a.grabtype = GRABTYPE_CORE; - b.grabtype = GRABTYPE_CORE; - a.device = &dev1; - b.device = &dev1; - a.modifierDevice = &dev1; - b.modifierDevice = &dev1; - a.type = XI_KeyPress; - b.type = XI_KeyRelease; - rc = GrabMatchesSecond(&a, &b, FALSE); - g_assert(rc == FALSE); - rc = GrabMatchesSecond(&a, &b, TRUE); - g_assert(rc == FALSE); - - a.grabtype = GRABTYPE_XI; - b.grabtype = GRABTYPE_XI; - a.device = &dev1; - b.device = &dev1; - a.modifierDevice = &dev1; - b.modifierDevice = &dev1; - a.type = XI_KeyPress; - b.type = XI_KeyRelease; - rc = GrabMatchesSecond(&a, &b, FALSE); - g_assert(rc == FALSE); - rc = GrabMatchesSecond(&a, &b, TRUE); - g_assert(rc == FALSE); - - /* different modifiers must fail */ - a.grabtype = GRABTYPE_XI2; - b.grabtype = GRABTYPE_XI2; - a.device = &dev1; - b.device = &dev1; - a.modifierDevice = &dev1; - b.modifierDevice = &dev1; - a.type = XI_KeyPress; - b.type = XI_KeyPress; - a.modifiersDetail.exact = 1; - b.modifiersDetail.exact = 2; - rc = GrabMatchesSecond(&a, &b, FALSE); - g_assert(rc == FALSE); - rc = GrabMatchesSecond(&b, &a, FALSE); - g_assert(rc == FALSE); - - a.grabtype = GRABTYPE_CORE; - b.grabtype = GRABTYPE_CORE; - rc = GrabMatchesSecond(&a, &b, FALSE); - g_assert(rc == FALSE); - rc = GrabMatchesSecond(&b, &a, FALSE); - g_assert(rc == FALSE); - - a.grabtype = GRABTYPE_XI; - b.grabtype = GRABTYPE_XI; - rc = GrabMatchesSecond(&a, &b, FALSE); - g_assert(rc == FALSE); - rc = GrabMatchesSecond(&b, &a, FALSE); - g_assert(rc == FALSE); - - /* AnyModifier must fail for XI2 */ - a.grabtype = GRABTYPE_XI2; - b.grabtype = GRABTYPE_XI2; - a.modifiersDetail.exact = AnyModifier; - b.modifiersDetail.exact = 1; - rc = GrabMatchesSecond(&a, &b, FALSE); - g_assert(rc == FALSE); - rc = GrabMatchesSecond(&b, &a, FALSE); - g_assert(rc == FALSE); - - /* XIAnyModifier must fail for CORE and XI */ - a.grabtype = GRABTYPE_XI; - b.grabtype = GRABTYPE_XI; - a.modifiersDetail.exact = XIAnyModifier; - b.modifiersDetail.exact = 1; - rc = GrabMatchesSecond(&a, &b, FALSE); - g_assert(rc == FALSE); - rc = GrabMatchesSecond(&b, &a, FALSE); - g_assert(rc == FALSE); - - a.grabtype = GRABTYPE_CORE; - b.grabtype = GRABTYPE_CORE; - a.modifiersDetail.exact = XIAnyModifier; - b.modifiersDetail.exact = 1; - rc = GrabMatchesSecond(&a, &b, FALSE); - g_assert(rc == FALSE); - rc = GrabMatchesSecond(&b, &a, FALSE); - g_assert(rc == FALSE); - - /* different detail must fail */ - a.grabtype = GRABTYPE_XI2; - b.grabtype = GRABTYPE_XI2; - a.detail.exact = 1; - b.detail.exact = 2; - a.modifiersDetail.exact = 1; - b.modifiersDetail.exact = 1; - rc = GrabMatchesSecond(&a, &b, FALSE); - g_assert(rc == FALSE); - rc = GrabMatchesSecond(&b, &a, FALSE); - g_assert(rc == FALSE); - - a.grabtype = GRABTYPE_XI; - b.grabtype = GRABTYPE_XI; - rc = GrabMatchesSecond(&a, &b, FALSE); - g_assert(rc == FALSE); - rc = GrabMatchesSecond(&b, &a, FALSE); - g_assert(rc == FALSE); - - a.grabtype = GRABTYPE_CORE; - b.grabtype = GRABTYPE_CORE; - rc = GrabMatchesSecond(&a, &b, FALSE); - g_assert(rc == FALSE); - rc = GrabMatchesSecond(&b, &a, FALSE); - g_assert(rc == FALSE); - - /* detail of AnyModifier must fail */ - a.grabtype = GRABTYPE_XI2; - b.grabtype = GRABTYPE_XI2; - a.detail.exact = AnyModifier; - b.detail.exact = 1; - a.modifiersDetail.exact = 1; - b.modifiersDetail.exact = 1; - rc = GrabMatchesSecond(&a, &b, FALSE); - g_assert(rc == FALSE); - rc = GrabMatchesSecond(&b, &a, FALSE); - g_assert(rc == FALSE); - - a.grabtype = GRABTYPE_CORE; - b.grabtype = GRABTYPE_CORE; - rc = GrabMatchesSecond(&a, &b, FALSE); - g_assert(rc == FALSE); - rc = GrabMatchesSecond(&b, &a, FALSE); - g_assert(rc == FALSE); - - a.grabtype = GRABTYPE_XI; - b.grabtype = GRABTYPE_XI; - rc = GrabMatchesSecond(&a, &b, FALSE); - g_assert(rc == FALSE); - rc = GrabMatchesSecond(&b, &a, FALSE); - g_assert(rc == FALSE); - - /* detail of XIAnyModifier must fail */ - a.grabtype = GRABTYPE_XI2; - b.grabtype = GRABTYPE_XI2; - a.detail.exact = XIAnyModifier; - b.detail.exact = 1; - a.modifiersDetail.exact = 1; - b.modifiersDetail.exact = 1; - rc = GrabMatchesSecond(&a, &b, FALSE); - g_assert(rc == FALSE); - rc = GrabMatchesSecond(&b, &a, FALSE); - g_assert(rc == FALSE); - - a.grabtype = GRABTYPE_CORE; - b.grabtype = GRABTYPE_CORE; - rc = GrabMatchesSecond(&a, &b, FALSE); - g_assert(rc == FALSE); - rc = GrabMatchesSecond(&b, &a, FALSE); - g_assert(rc == FALSE); - - a.grabtype = GRABTYPE_XI; - b.grabtype = GRABTYPE_XI; - rc = GrabMatchesSecond(&a, &b, FALSE); - g_assert(rc == FALSE); - rc = GrabMatchesSecond(&b, &a, FALSE); - g_assert(rc == FALSE); - - /* XIAnyModifier or AnyModifer must succeed */ - a.grabtype = GRABTYPE_XI2; - b.grabtype = GRABTYPE_XI2; - a.detail.exact = 1; - b.detail.exact = 1; - a.modifiersDetail.exact = XIAnyModifier; - b.modifiersDetail.exact = 1; - rc = GrabMatchesSecond(&a, &b, FALSE); - g_assert(rc == TRUE); - rc = GrabMatchesSecond(&b, &a, FALSE); - g_assert(rc == TRUE); - - a.grabtype = GRABTYPE_CORE; - b.grabtype = GRABTYPE_CORE; - a.detail.exact = 1; - b.detail.exact = 1; - a.modifiersDetail.exact = AnyModifier; - b.modifiersDetail.exact = 1; - rc = GrabMatchesSecond(&a, &b, FALSE); - g_assert(rc == TRUE); - rc = GrabMatchesSecond(&b, &a, FALSE); - g_assert(rc == TRUE); - - a.grabtype = GRABTYPE_XI; - b.grabtype = GRABTYPE_XI; - a.detail.exact = 1; - b.detail.exact = 1; - a.modifiersDetail.exact = AnyModifier; - b.modifiersDetail.exact = 1; - rc = GrabMatchesSecond(&a, &b, FALSE); - g_assert(rc == TRUE); - rc = GrabMatchesSecond(&b, &a, FALSE); - g_assert(rc == TRUE); - - /* AnyKey or XIAnyKeycode must succeed */ - a.grabtype = GRABTYPE_XI2; - b.grabtype = GRABTYPE_XI2; - a.detail.exact = XIAnyKeycode; - b.detail.exact = 1; - a.modifiersDetail.exact = 1; - b.modifiersDetail.exact = 1; - rc = GrabMatchesSecond(&a, &b, FALSE); - g_assert(rc == TRUE); - rc = GrabMatchesSecond(&b, &a, FALSE); - g_assert(rc == TRUE); - - a.grabtype = GRABTYPE_CORE; - b.grabtype = GRABTYPE_CORE; - a.detail.exact = AnyKey; - b.detail.exact = 1; - a.modifiersDetail.exact = 1; - b.modifiersDetail.exact = 1; - rc = GrabMatchesSecond(&a, &b, FALSE); - g_assert(rc == TRUE); - rc = GrabMatchesSecond(&b, &a, FALSE); - g_assert(rc == TRUE); - - a.grabtype = GRABTYPE_XI; - b.grabtype = GRABTYPE_XI; - a.detail.exact = AnyKey; - b.detail.exact = 1; - a.modifiersDetail.exact = 1; - b.modifiersDetail.exact = 1; - rc = GrabMatchesSecond(&a, &b, FALSE); - g_assert(rc == TRUE); - rc = GrabMatchesSecond(&b, &a, FALSE); - g_assert(rc == TRUE); -} - -static void test_bits_to_byte(int i) -{ - int expected_bytes; - expected_bytes = (i + 7)/8; - - g_assert(bits_to_bytes(i) >= i/8); - g_assert((bits_to_bytes(i) * 8) - i <= 7); - g_assert(expected_bytes == bits_to_bytes(i)); -} - -static void test_bytes_to_int32(int i) -{ - int expected_4byte; - expected_4byte = (i + 3)/4; - - g_assert(bytes_to_int32(i) <= i); - g_assert((bytes_to_int32(i) * 4) - i <= 3); - g_assert(expected_4byte == bytes_to_int32(i)); -} - -static void test_pad_to_int32(int i) -{ - int expected_bytes; - expected_bytes = ((i + 3)/4) * 4; - - g_assert(pad_to_int32(i) >= i); - g_assert(pad_to_int32(i) - i <= 3); - g_assert(expected_bytes == pad_to_int32(i)); -} -static void include_byte_padding_macros(void) -{ - g_test_message("Testing bits_to_bytes()"); - - /* the macros don't provide overflow protection */ - test_bits_to_byte(0); - test_bits_to_byte(1); - test_bits_to_byte(2); - test_bits_to_byte(7); - test_bits_to_byte(8); - test_bits_to_byte(0xFF); - test_bits_to_byte(0x100); - test_bits_to_byte(INT_MAX - 9); - test_bits_to_byte(INT_MAX - 8); - - g_test_message("Testing bytes_to_int32()"); - - test_bytes_to_int32(0); - test_bytes_to_int32(1); - test_bytes_to_int32(2); - test_bytes_to_int32(7); - test_bytes_to_int32(8); - test_bytes_to_int32(0xFF); - test_bytes_to_int32(0x100); - test_bytes_to_int32(0xFFFF); - test_bytes_to_int32(0x10000); - test_bytes_to_int32(0xFFFFFF); - test_bytes_to_int32(0x1000000); - test_bytes_to_int32(INT_MAX - 4); - test_bytes_to_int32(INT_MAX - 3); - - g_test_message("Testing pad_to_int32"); - - test_pad_to_int32(0); - test_pad_to_int32(0); - test_pad_to_int32(1); - test_pad_to_int32(2); - test_pad_to_int32(7); - test_pad_to_int32(8); - test_pad_to_int32(0xFF); - test_pad_to_int32(0x100); - test_pad_to_int32(0xFFFF); - test_pad_to_int32(0x10000); - test_pad_to_int32(0xFFFFFF); - test_pad_to_int32(0x1000000); - test_pad_to_int32(INT_MAX - 4); - test_pad_to_int32(INT_MAX - 3); -} - -static void xi_unregister_handlers(void) -{ - DeviceIntRec dev; - int handler; - - memset(&dev, 0, sizeof(dev)); - - handler = XIRegisterPropertyHandler(&dev, NULL, NULL, NULL); - g_assert(handler == 1); - handler = XIRegisterPropertyHandler(&dev, NULL, NULL, NULL); - g_assert(handler == 2); - handler = XIRegisterPropertyHandler(&dev, NULL, NULL, NULL); - g_assert(handler == 3); - - g_test_message("Unlinking from front."); - - XIUnregisterPropertyHandler(&dev, 4); /* NOOP */ - g_assert(dev.properties.handlers->id == 3); - XIUnregisterPropertyHandler(&dev, 3); - g_assert(dev.properties.handlers->id == 2); - XIUnregisterPropertyHandler(&dev, 2); - g_assert(dev.properties.handlers->id == 1); - XIUnregisterPropertyHandler(&dev, 1); - g_assert(dev.properties.handlers == NULL); - - handler = XIRegisterPropertyHandler(&dev, NULL, NULL, NULL); - g_assert(handler == 4); - handler = XIRegisterPropertyHandler(&dev, NULL, NULL, NULL); - g_assert(handler == 5); - handler = XIRegisterPropertyHandler(&dev, NULL, NULL, NULL); - g_assert(handler == 6); - XIUnregisterPropertyHandler(&dev, 3); /* NOOP */ - g_assert(dev.properties.handlers->next->next->next == NULL); - XIUnregisterPropertyHandler(&dev, 4); - g_assert(dev.properties.handlers->next->next == NULL); - XIUnregisterPropertyHandler(&dev, 5); - g_assert(dev.properties.handlers->next == NULL); - XIUnregisterPropertyHandler(&dev, 6); - g_assert(dev.properties.handlers == NULL); - - handler = XIRegisterPropertyHandler(&dev, NULL, NULL, NULL); - g_assert(handler == 7); - handler = XIRegisterPropertyHandler(&dev, NULL, NULL, NULL); - g_assert(handler == 8); - handler = XIRegisterPropertyHandler(&dev, NULL, NULL, NULL); - g_assert(handler == 9); - - XIDeleteAllDeviceProperties(&dev); - g_assert(dev.properties.handlers == NULL); - XIUnregisterPropertyHandler(&dev, 7); /* NOOP */ - -} - -static void cmp_attr_fields(InputAttributes *attr1, - InputAttributes *attr2) -{ - char **tags1, **tags2; - - g_assert(attr1 && attr2); - g_assert(attr1 != attr2); - g_assert(attr1->flags == attr2->flags); - - if (attr1->product != NULL) - { - g_assert(attr1->product != attr2->product); - g_assert(strcmp(attr1->product, attr2->product) == 0); - } else - g_assert(attr2->product == NULL); - - if (attr1->vendor != NULL) - { - g_assert(attr1->vendor != attr2->vendor); - g_assert(strcmp(attr1->vendor, attr2->vendor) == 0); - } else - g_assert(attr2->vendor == NULL); - - if (attr1->device != NULL) - { - g_assert(attr1->device != attr2->device); - g_assert(strcmp(attr1->device, attr2->device) == 0); - } else - g_assert(attr2->device == NULL); - - if (attr1->pnp_id != NULL) - { - g_assert(attr1->pnp_id != attr2->pnp_id); - g_assert(strcmp(attr1->pnp_id, attr2->pnp_id) == 0); - } else - g_assert(attr2->pnp_id == NULL); - - if (attr1->usb_id != NULL) - { - g_assert(attr1->usb_id != attr2->usb_id); - g_assert(strcmp(attr1->usb_id, attr2->usb_id) == 0); - } else - g_assert(attr2->usb_id == NULL); - - tags1 = attr1->tags; - tags2 = attr2->tags; - - /* if we don't have any tags, skip the tag checking bits */ - if (!tags1) - { - g_assert(!tags2); - return; - } - - /* Don't lug around empty arrays */ - g_assert(*tags1); - g_assert(*tags2); - - /* check for identical content, but duplicated */ - while (*tags1) - { - g_assert(*tags1 != *tags2); - g_assert(strcmp(*tags1, *tags2) == 0); - tags1++; - tags2++; - } - - /* ensure tags1 and tags2 have the same no of elements */ - g_assert(!*tags2); - - /* check for not sharing memory */ - tags1 = attr1->tags; - while (*tags1) - { - tags2 = attr2->tags; - while (*tags2) - g_assert(*tags1 != *tags2++); - - tags1++; - } -} - -static void dix_input_attributes(void) -{ - InputAttributes orig = {0}; - InputAttributes *new; - char *tags[4] = {"tag1", "tag2", "tag2", NULL}; - - new = DuplicateInputAttributes(NULL); - g_assert(!new); - - new = DuplicateInputAttributes(&orig); - g_assert(memcmp(&orig, new, sizeof(InputAttributes)) == 0); - - orig.product = "product name"; - new = DuplicateInputAttributes(&orig); - cmp_attr_fields(&orig, new); - FreeInputAttributes(new); - - orig.vendor = "vendor name"; - new = DuplicateInputAttributes(&orig); - cmp_attr_fields(&orig, new); - FreeInputAttributes(new); - - orig.device = "device path"; - new = DuplicateInputAttributes(&orig); - cmp_attr_fields(&orig, new); - FreeInputAttributes(new); - - orig.pnp_id = "PnPID"; - new = DuplicateInputAttributes(&orig); - cmp_attr_fields(&orig, new); - FreeInputAttributes(new); - - orig.usb_id = "USBID"; - new = DuplicateInputAttributes(&orig); - cmp_attr_fields(&orig, new); - FreeInputAttributes(new); - - orig.flags = 0xF0; - new = DuplicateInputAttributes(&orig); - cmp_attr_fields(&orig, new); - FreeInputAttributes(new); - - orig.tags = tags; - new = DuplicateInputAttributes(&orig); - cmp_attr_fields(&orig, new); - FreeInputAttributes(new); -} - -static void dix_input_valuator_masks(void) -{ - ValuatorMask *mask = NULL, *copy; - int nvaluators = MAX_VALUATORS; - int valuators[nvaluators]; - int i; - int first_val, num_vals; - - for (i = 0; i < nvaluators; i++) - valuators[i] = i; - - mask = valuator_mask_new(nvaluators); - g_assert(mask != NULL); - g_assert(valuator_mask_size(mask) == 0); - g_assert(valuator_mask_num_valuators(mask) == 0); - - for (i = 0; i < nvaluators; i++) - { - g_assert(!valuator_mask_isset(mask, i)); - valuator_mask_set(mask, i, valuators[i]); - g_assert(valuator_mask_isset(mask, i)); - g_assert(valuator_mask_get(mask, i) == valuators[i]); - g_assert(valuator_mask_size(mask) == i + 1); - g_assert(valuator_mask_num_valuators(mask) == i + 1); - } - - for (i = 0; i < nvaluators; i++) - { - g_assert(valuator_mask_isset(mask, i)); - valuator_mask_unset(mask, i); - /* we're removing valuators from the front, so size should stay the - * same until the last bit is removed */ - if (i < nvaluators - 1) - g_assert(valuator_mask_size(mask) == nvaluators); - g_assert(!valuator_mask_isset(mask, i)); - } - - g_assert(valuator_mask_size(mask) == 0); - valuator_mask_zero(mask); - g_assert(valuator_mask_size(mask) == 0); - g_assert(valuator_mask_num_valuators(mask) == 0); - for (i = 0; i < nvaluators; i++) - g_assert(!valuator_mask_isset(mask, i)); - - first_val = 5; - num_vals = 6; - - valuator_mask_set_range(mask, first_val, num_vals, valuators); - g_assert(valuator_mask_size(mask) == first_val + num_vals); - g_assert(valuator_mask_num_valuators(mask) == num_vals); - for (i = 0; i < nvaluators; i++) - { - if (i < first_val || i >= first_val + num_vals) - g_assert(!valuator_mask_isset(mask, i)); - else - { - g_assert(valuator_mask_isset(mask, i)); - g_assert(valuator_mask_get(mask, i) == valuators[i - first_val]); - } - } - - copy = valuator_mask_new(nvaluators); - valuator_mask_copy(copy, mask); - g_assert(mask != copy); - g_assert(valuator_mask_size(mask) == valuator_mask_size(copy)); - g_assert(valuator_mask_num_valuators(mask) == valuator_mask_num_valuators(copy)); - - for (i = 0; i < nvaluators; i++) - { - g_assert(valuator_mask_isset(mask, i) == valuator_mask_isset(copy, i)); - g_assert(valuator_mask_get(mask, i) == valuator_mask_get(copy, i)); - } - - valuator_mask_free(&mask); - g_assert(mask == NULL); -} - -static void dix_valuator_mode(void) -{ - DeviceIntRec dev; - const int num_axes = MAX_VALUATORS; - int i; - Atom atoms[MAX_VALUATORS] = { 0 }; - - memset(&dev, 0, sizeof(DeviceIntRec)); - dev.type = MASTER_POINTER; /* claim it's a master to stop ptracccel */ - - g_assert(InitValuatorClassDeviceStruct(NULL, 0, atoms, 0, 0) == FALSE); - g_assert(InitValuatorClassDeviceStruct(&dev, num_axes, atoms, 0, Absolute)); - - for (i = 0; i < num_axes; i++) - { - g_assert(valuator_get_mode(&dev, i) == Absolute); - valuator_set_mode(&dev, i, Relative); - g_assert(dev.valuator->axes[i].mode == Relative); - g_assert(valuator_get_mode(&dev, i) == Relative); - } - - valuator_set_mode(&dev, VALUATOR_MODE_ALL_AXES, Absolute); - for (i = 0; i < num_axes; i++) - g_assert(valuator_get_mode(&dev, i) == Absolute); - - valuator_set_mode(&dev, VALUATOR_MODE_ALL_AXES, Relative); - for (i = 0; i < num_axes; i++) - g_assert(valuator_get_mode(&dev, i) == Relative); -} - -static void include_bit_test_macros(void) -{ - uint8_t mask[9] = { 0 }; - int i; - - for (i = 0; i < sizeof(mask)/sizeof(mask[0]); i++) - { - g_assert(BitIsOn(mask, i) == 0); - SetBit(mask, i); - g_assert(BitIsOn(mask, i) == 1); - g_assert(!!(mask[i/8] & (1 << (i % 8)))); - g_assert(CountBits(mask, sizeof(mask)) == 1); - ClearBit(mask, i); - g_assert(BitIsOn(mask, i) == 0); - } -} - -int main(int argc, char** argv) -{ - g_test_init(&argc, &argv,NULL); - g_test_bug_base("https://bugzilla.freedesktop.org/show_bug.cgi?id="); - - g_test_add_func("/dix/input/valuator-masks", dix_input_valuator_masks); - g_test_add_func("/dix/input/attributes", dix_input_attributes); - g_test_add_func("/dix/input/init-valuators", dix_init_valuators); - g_test_add_func("/dix/input/event-core-conversion", dix_event_to_core_conversion); - g_test_add_func("/dix/input/check-grab-values", dix_check_grab_values); - g_test_add_func("/dix/input/xi2-struct-sizes", xi2_struct_sizes); - g_test_add_func("/dix/input/grab_matching", dix_grab_matching); - g_test_add_func("/dix/input/valuator_mode", dix_valuator_mode); - g_test_add_func("/include/byte_padding_macros", include_byte_padding_macros); - g_test_add_func("/include/bit_test_macros", include_bit_test_macros); - g_test_add_func("/Xi/xiproperty/register-unregister", xi_unregister_handlers); - - - return g_test_run(); -} +/**
+ * Copyright © 2009 Red Hat, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#ifdef HAVE_DIX_CONFIG_H
+#include <dix-config.h>
+#endif
+
+#include <stdint.h>
+#include <X11/X.h>
+#include "misc.h"
+#include "resource.h"
+#include <X11/Xproto.h>
+#include <X11/extensions/XI2proto.h>
+#include <X11/Xatom.h>
+#include "windowstr.h"
+#include "inputstr.h"
+#include "eventconvert.h"
+#include "exevents.h"
+#include "exglobals.h"
+#include "dixgrabs.h"
+#include "eventstr.h"
+#include "inpututils.h"
+#include <glib.h>
+
+/**
+ * Init a device with axes.
+ * Verify values set on the device.
+ *
+ * Result: All axes set to default values (usually 0).
+ */
+static void dix_init_valuators(void)
+{
+ DeviceIntRec dev;
+ ValuatorClassPtr val;
+ const int num_axes = 2;
+ int i;
+ Atom atoms[MAX_VALUATORS] = { 0 };
+
+
+ memset(&dev, 0, sizeof(DeviceIntRec));
+ dev.type = MASTER_POINTER; /* claim it's a master to stop ptracccel */
+
+ g_assert(InitValuatorClassDeviceStruct(NULL, 0, atoms, 0, 0) == FALSE);
+ g_assert(InitValuatorClassDeviceStruct(&dev, num_axes, atoms, 0, Absolute));
+
+ val = dev.valuator;
+ g_assert(val);
+ g_assert(val->numAxes == num_axes);
+ g_assert(val->numMotionEvents == 0);
+ g_assert(val->axisVal);
+
+ for (i = 0; i < num_axes; i++)
+ {
+ g_assert(val->axisVal[i] == 0);
+ g_assert(val->axes->min_value == NO_AXIS_LIMITS);
+ g_assert(val->axes->max_value == NO_AXIS_LIMITS);
+ g_assert(val->axes->mode == Absolute);
+ }
+
+ g_assert(dev.last.numValuators == num_axes);
+}
+
+/* just check the known success cases, and that error cases set the client's
+ * error value correctly. */
+static void dix_check_grab_values(void)
+{
+ ClientRec client;
+ GrabParameters param;
+ int rc;
+
+ memset(&client, 0, sizeof(client));
+
+ param.grabtype = GRABTYPE_CORE;
+ param.this_device_mode = GrabModeSync;
+ param.other_devices_mode = GrabModeSync;
+ param.modifiers = AnyModifier;
+ param.ownerEvents = FALSE;
+
+ rc = CheckGrabValues(&client, ¶m);
+ g_assert(rc == Success);
+
+ param.this_device_mode = GrabModeAsync;
+ rc = CheckGrabValues(&client, ¶m);
+ g_assert(rc == Success);
+
+ param.this_device_mode = GrabModeAsync + 1;
+ rc = CheckGrabValues(&client, ¶m);
+ g_assert(rc == BadValue);
+ g_assert(client.errorValue == param.this_device_mode);
+ g_assert(client.errorValue == GrabModeAsync + 1);
+
+ param.this_device_mode = GrabModeSync;
+ param.other_devices_mode = GrabModeAsync;
+ rc = CheckGrabValues(&client, ¶m);
+ g_assert(rc == Success);
+
+ param.other_devices_mode = GrabModeAsync + 1;
+ rc = CheckGrabValues(&client, ¶m);
+ g_assert(rc == BadValue);
+ g_assert(client.errorValue == param.other_devices_mode);
+ g_assert(client.errorValue == GrabModeAsync + 1);
+
+ param.other_devices_mode = GrabModeSync;
+
+ param.modifiers = 1 << 13;
+ rc = CheckGrabValues(&client, ¶m);
+ g_assert(rc == BadValue);
+ g_assert(client.errorValue == param.modifiers);
+ g_assert(client.errorValue == (1 << 13));
+
+
+ param.modifiers = AnyModifier;
+ param.ownerEvents = TRUE;
+ rc = CheckGrabValues(&client, ¶m);
+ g_assert(rc == Success);
+
+ param.ownerEvents = 3;
+ rc = CheckGrabValues(&client, ¶m);
+ g_assert(rc == BadValue);
+ g_assert(client.errorValue == param.ownerEvents);
+ g_assert(client.errorValue == 3);
+}
+
+
+/**
+ * Convert various internal events to the matching core event and verify the
+ * parameters.
+ */
+static void dix_event_to_core(int type)
+{
+ DeviceEvent ev;
+ xEvent core;
+ int time;
+ int x, y;
+ int rc;
+ int state;
+ int detail;
+ const int ROOT_WINDOW_ID = 0x100;
+
+ /* EventToCore memsets the event to 0 */
+#define test_event() \
+ g_assert(rc == Success); \
+ g_assert(core.u.u.type == type); \
+ g_assert(core.u.u.detail == detail); \
+ g_assert(core.u.keyButtonPointer.time == time); \
+ g_assert(core.u.keyButtonPointer.rootX == x); \
+ g_assert(core.u.keyButtonPointer.rootY == y); \
+ g_assert(core.u.keyButtonPointer.state == state); \
+ g_assert(core.u.keyButtonPointer.eventX == 0); \
+ g_assert(core.u.keyButtonPointer.eventY == 0); \
+ g_assert(core.u.keyButtonPointer.root == ROOT_WINDOW_ID); \
+ g_assert(core.u.keyButtonPointer.event == 0); \
+ g_assert(core.u.keyButtonPointer.child == 0); \
+ g_assert(core.u.keyButtonPointer.sameScreen == FALSE);
+
+ x = 0;
+ y = 0;
+ time = 12345;
+ state = 0;
+ detail = 0;
+
+ ev.header = 0xFF;
+ ev.length = sizeof(DeviceEvent);
+ ev.time = time;
+ ev.root_y = x;
+ ev.root_x = y;
+ SetBit(ev.valuators.mask, 0);
+ SetBit(ev.valuators.mask, 1);
+ ev.root = ROOT_WINDOW_ID;
+ ev.corestate = state;
+ ev.detail.key = detail;
+
+ ev.type = type;
+ ev.detail.key = 0;
+ rc = EventToCore((InternalEvent*)&ev, &core);
+ test_event();
+
+ x = 1;
+ y = 2;
+ ev.root_x = x;
+ ev.root_y = y;
+ rc = EventToCore((InternalEvent*)&ev, &core);
+ test_event();
+
+ x = 0x7FFF;
+ y = 0x7FFF;
+ ev.root_x = x;
+ ev.root_y = y;
+ rc = EventToCore((InternalEvent*)&ev, &core);
+ test_event();
+
+ x = 0x8000; /* too high */
+ y = 0x8000; /* too high */
+ ev.root_x = x;
+ ev.root_y = y;
+ rc = EventToCore((InternalEvent*)&ev, &core);
+ g_assert(core.u.keyButtonPointer.rootX != x);
+ g_assert(core.u.keyButtonPointer.rootY != y);
+
+ x = 0x7FFF;
+ y = 0x7FFF;
+ ev.root_x = x;
+ ev.root_y = y;
+ time = 0;
+ ev.time = time;
+ rc = EventToCore((InternalEvent*)&ev, &core);
+ test_event();
+
+ detail = 1;
+ ev.detail.key = detail;
+ rc = EventToCore((InternalEvent*)&ev, &core);
+ test_event();
+
+ detail = 0xFF; /* highest value */
+ ev.detail.key = detail;
+ rc = EventToCore((InternalEvent*)&ev, &core);
+ test_event();
+
+ detail = 0xFFF; /* too big */
+ ev.detail.key = detail;
+ rc = EventToCore((InternalEvent*)&ev, &core);
+ g_assert(rc == BadMatch);
+
+ detail = 0xFF; /* too big */
+ ev.detail.key = detail;
+ state = 0xFFFF; /* highest value */
+ ev.corestate = state;
+ rc = EventToCore((InternalEvent*)&ev, &core);
+ test_event();
+
+ state = 0x10000; /* too big */
+ ev.corestate = state;
+ rc = EventToCore((InternalEvent*)&ev, &core);
+ g_assert(core.u.keyButtonPointer.state != state);
+ g_assert(core.u.keyButtonPointer.state == (state & 0xFFFF));
+
+#undef test_event
+}
+
+static void dix_event_to_core_fail(int evtype, int expected_rc)
+{
+ DeviceEvent ev;
+ xEvent core;
+ int rc;
+
+ ev.header = 0xFF;
+ ev.length = sizeof(DeviceEvent);
+
+ ev.type = evtype;
+ rc = EventToCore((InternalEvent*)&ev, &core);
+ g_assert(rc == expected_rc);
+}
+
+static void dix_event_to_core_conversion(void)
+{
+ dix_event_to_core_fail(0, BadImplementation);
+ dix_event_to_core_fail(1, BadImplementation);
+ dix_event_to_core_fail(ET_ProximityOut + 1, BadImplementation);
+ dix_event_to_core_fail(ET_ProximityIn, BadMatch);
+ dix_event_to_core_fail(ET_ProximityOut, BadMatch);
+
+ dix_event_to_core(ET_KeyPress);
+ dix_event_to_core(ET_KeyRelease);
+ dix_event_to_core(ET_ButtonPress);
+ dix_event_to_core(ET_ButtonRelease);
+ dix_event_to_core(ET_Motion);
+}
+
+static void
+_dix_test_xi_convert(DeviceEvent *ev, int expected_rc, int expected_count)
+{
+ xEvent *xi;
+ int count = 0;
+ int rc;
+
+ rc = EventToXI((InternalEvent*)ev, &xi, &count);
+ g_assert(rc == expected_rc);
+ g_assert(count >= expected_count);
+ if (count > 0){
+ deviceKeyButtonPointer *kbp = (deviceKeyButtonPointer*)xi;
+ g_assert(kbp->type == IEventBase + ev->type);
+ g_assert(kbp->detail == ev->detail.key);
+ g_assert(kbp->time == ev->time);
+ g_assert((kbp->deviceid & ~MORE_EVENTS) == ev->deviceid);
+ g_assert(kbp->root_x == ev->root_x);
+ g_assert(kbp->root_y == ev->root_y);
+ g_assert(kbp->state == ev->corestate);
+ g_assert(kbp->event_x == 0);
+ g_assert(kbp->event_y == 0);
+ g_assert(kbp->root == ev->root);
+ g_assert(kbp->event == 0);
+ g_assert(kbp->child == 0);
+ g_assert(kbp->same_screen == FALSE);
+
+ while (--count > 0) {
+ deviceValuator *v = (deviceValuator*)&xi[count];
+ g_assert(v->type == DeviceValuator);
+ g_assert(v->num_valuators <= 6);
+ }
+
+
+ free(xi);
+ }
+}
+
+/**
+ * This tests for internal event → XI1 event conversion
+ * - all conversions should generate the right XI event type
+ * - right number of events generated
+ * - extra events are valuators
+ */
+static void dix_event_to_xi1_conversion(void)
+{
+ DeviceEvent ev = {0};
+ int time;
+ int x, y;
+ int state;
+ int detail;
+ const int ROOT_WINDOW_ID = 0x100;
+ int deviceid;
+
+ IEventBase = 80;
+ DeviceValuator = IEventBase - 1;
+ DeviceKeyPress = IEventBase + ET_KeyPress;
+ DeviceKeyRelease = IEventBase + ET_KeyRelease;
+ DeviceButtonPress = IEventBase + ET_ButtonPress;
+ DeviceButtonRelease = IEventBase + ET_ButtonRelease;
+ DeviceMotionNotify = IEventBase + ET_Motion;
+ DeviceFocusIn = IEventBase + ET_FocusIn;
+ DeviceFocusOut = IEventBase + ET_FocusOut;
+ ProximityIn = IEventBase + ET_ProximityIn;
+ ProximityOut = IEventBase + ET_ProximityOut;
+
+ /* EventToXI callocs */
+ x = 0;
+ y = 0;
+ time = 12345;
+ state = 0;
+ detail = 0;
+ deviceid = 4;
+
+ ev.header = 0xFF;
+
+ ev.header = 0xFF;
+ ev.length = sizeof(DeviceEvent);
+ ev.time = time;
+ ev.root_y = x;
+ ev.root_x = y;
+ SetBit(ev.valuators.mask, 0);
+ SetBit(ev.valuators.mask, 1);
+ ev.root = ROOT_WINDOW_ID;
+ ev.corestate = state;
+ ev.detail.key = detail;
+ ev.deviceid = deviceid;
+
+ /* test all types for bad match */
+ ev.type = ET_KeyPress; _dix_test_xi_convert(&ev, Success, 1);
+ ev.type = ET_KeyRelease; _dix_test_xi_convert(&ev, Success, 1);
+ ev.type = ET_ButtonPress; _dix_test_xi_convert(&ev, Success, 1);
+ ev.type = ET_ButtonRelease; _dix_test_xi_convert(&ev, Success, 1);
+ ev.type = ET_Motion; _dix_test_xi_convert(&ev, Success, 1);
+ ev.type = ET_ProximityIn; _dix_test_xi_convert(&ev, Success, 1);
+ ev.type = ET_ProximityOut; _dix_test_xi_convert(&ev, Success, 1);
+
+ /* No axes */
+ ClearBit(ev.valuators.mask, 0);
+ ClearBit(ev.valuators.mask, 1);
+ ev.type = ET_KeyPress; _dix_test_xi_convert(&ev, Success, 1);
+ ev.type = ET_KeyRelease; _dix_test_xi_convert(&ev, Success, 1);
+ ev.type = ET_ButtonPress; _dix_test_xi_convert(&ev, Success, 1);
+ ev.type = ET_ButtonRelease; _dix_test_xi_convert(&ev, Success, 1);
+ ev.type = ET_Motion; _dix_test_xi_convert(&ev, BadMatch, 0);
+ ev.type = ET_ProximityIn; _dix_test_xi_convert(&ev, BadMatch, 0);
+ ev.type = ET_ProximityOut; _dix_test_xi_convert(&ev, BadMatch, 0);
+
+ /* more than 6 axes → 2 valuator events */
+ SetBit(ev.valuators.mask, 0);
+ SetBit(ev.valuators.mask, 1);
+ SetBit(ev.valuators.mask, 2);
+ SetBit(ev.valuators.mask, 3);
+ SetBit(ev.valuators.mask, 4);
+ SetBit(ev.valuators.mask, 5);
+ SetBit(ev.valuators.mask, 6);
+ ev.type = ET_KeyPress; _dix_test_xi_convert(&ev, Success, 2);
+ ev.type = ET_KeyRelease; _dix_test_xi_convert(&ev, Success, 2);
+ ev.type = ET_ButtonPress; _dix_test_xi_convert(&ev, Success, 2);
+ ev.type = ET_ButtonRelease; _dix_test_xi_convert(&ev, Success, 2);
+ ev.type = ET_Motion; _dix_test_xi_convert(&ev, Success, 2);
+ ev.type = ET_ProximityIn; _dix_test_xi_convert(&ev, Success, 2);
+ ev.type = ET_ProximityOut; _dix_test_xi_convert(&ev, Success, 2);
+
+
+ /* keycode too high */
+ ev.type = ET_KeyPress;
+ ev.detail.key = 256;
+ _dix_test_xi_convert(&ev, Success, 0);
+
+ /* deviceid too high */
+ ev.type = ET_KeyPress;
+ ev.detail.key = 18;
+ ev.deviceid = 128;
+ _dix_test_xi_convert(&ev, Success, 0);
+}
+
+
+static void xi2_struct_sizes(void)
+{
+#define compare(req) \
+ g_assert(sizeof(req) == sz_##req);
+
+ compare(xXIQueryVersionReq);
+ compare(xXIWarpPointerReq);
+ compare(xXIChangeCursorReq);
+ compare(xXIChangeHierarchyReq);
+ compare(xXISetClientPointerReq);
+ compare(xXIGetClientPointerReq);
+ compare(xXISelectEventsReq);
+ compare(xXIQueryVersionReq);
+ compare(xXIQueryDeviceReq);
+ compare(xXISetFocusReq);
+ compare(xXIGetFocusReq);
+ compare(xXIGrabDeviceReq);
+ compare(xXIUngrabDeviceReq);
+ compare(xXIAllowEventsReq);
+ compare(xXIPassiveGrabDeviceReq);
+ compare(xXIPassiveUngrabDeviceReq);
+ compare(xXIListPropertiesReq);
+ compare(xXIChangePropertyReq);
+ compare(xXIDeletePropertyReq);
+ compare(xXIGetPropertyReq);
+ compare(xXIGetSelectedEventsReq);
+#undef compare
+}
+
+
+static void dix_grab_matching(void)
+{
+ DeviceIntRec xi_all_devices, xi_all_master_devices, dev1, dev2;
+ GrabRec a, b;
+ BOOL rc;
+
+ memset(&a, 0, sizeof(a));
+ memset(&b, 0, sizeof(b));
+
+ /* different grabtypes must fail */
+ a.grabtype = GRABTYPE_CORE;
+ b.grabtype = GRABTYPE_XI2;
+ rc = GrabMatchesSecond(&a, &b, FALSE);
+ g_assert(rc == FALSE);
+ rc = GrabMatchesSecond(&b, &a, FALSE);
+ g_assert(rc == FALSE);
+
+ a.grabtype = GRABTYPE_XI;
+ b.grabtype = GRABTYPE_XI2;
+ rc = GrabMatchesSecond(&a, &b, FALSE);
+ g_assert(rc == FALSE);
+ rc = GrabMatchesSecond(&b, &a, FALSE);
+ g_assert(rc == FALSE);
+
+ a.grabtype = GRABTYPE_XI;
+ b.grabtype = GRABTYPE_CORE;
+ rc = GrabMatchesSecond(&a, &b, FALSE);
+ g_assert(rc == FALSE);
+ rc = GrabMatchesSecond(&b, &a, FALSE);
+ g_assert(rc == FALSE);
+
+ /* XI2 grabs for different devices must fail, regardless of ignoreDevice
+ * XI2 grabs for master devices must fail against a slave */
+ memset(&xi_all_devices, 0, sizeof(DeviceIntRec));
+ memset(&xi_all_master_devices, 0, sizeof(DeviceIntRec));
+ memset(&dev1, 0, sizeof(DeviceIntRec));
+ memset(&dev2, 0, sizeof(DeviceIntRec));
+
+ xi_all_devices.id = XIAllDevices;
+ xi_all_master_devices.id = XIAllMasterDevices;
+ dev1.id = 10;
+ dev1.type = SLAVE;
+ dev2.id = 11;
+ dev2.type = SLAVE;
+
+ inputInfo.all_devices = &xi_all_devices;
+ inputInfo.all_master_devices = &xi_all_master_devices;
+ a.grabtype = GRABTYPE_XI2;
+ b.grabtype = GRABTYPE_XI2;
+ a.device = &dev1;
+ b.device = &dev2;
+
+ rc = GrabMatchesSecond(&a, &b, FALSE);
+ g_assert(rc == FALSE);
+
+ a.device = &dev2;
+ b.device = &dev1;
+ rc = GrabMatchesSecond(&a, &b, FALSE);
+ g_assert(rc == FALSE);
+ rc = GrabMatchesSecond(&a, &b, TRUE);
+ g_assert(rc == FALSE);
+
+ a.device = inputInfo.all_master_devices;
+ b.device = &dev1;
+ rc = GrabMatchesSecond(&a, &b, FALSE);
+ g_assert(rc == FALSE);
+ rc = GrabMatchesSecond(&a, &b, TRUE);
+ g_assert(rc == FALSE);
+
+ a.device = &dev1;
+ b.device = inputInfo.all_master_devices;
+ rc = GrabMatchesSecond(&a, &b, FALSE);
+ g_assert(rc == FALSE);
+ rc = GrabMatchesSecond(&a, &b, TRUE);
+ g_assert(rc == FALSE);
+
+ /* ignoreDevice FALSE must fail for different devices for CORE and XI */
+ a.grabtype = GRABTYPE_XI;
+ b.grabtype = GRABTYPE_XI;
+ a.device = &dev1;
+ b.device = &dev2;
+ a.modifierDevice = &dev1;
+ b.modifierDevice = &dev1;
+ rc = GrabMatchesSecond(&a, &b, FALSE);
+ g_assert(rc == FALSE);
+
+ a.grabtype = GRABTYPE_CORE;
+ b.grabtype = GRABTYPE_CORE;
+ a.device = &dev1;
+ b.device = &dev2;
+ a.modifierDevice = &dev1;
+ b.modifierDevice = &dev1;
+ rc = GrabMatchesSecond(&a, &b, FALSE);
+ g_assert(rc == FALSE);
+
+ /* ignoreDevice FALSE must fail for different modifier devices for CORE
+ * and XI */
+ a.grabtype = GRABTYPE_XI;
+ b.grabtype = GRABTYPE_XI;
+ a.device = &dev1;
+ b.device = &dev1;
+ a.modifierDevice = &dev1;
+ b.modifierDevice = &dev2;
+ rc = GrabMatchesSecond(&a, &b, FALSE);
+ g_assert(rc == FALSE);
+
+ a.grabtype = GRABTYPE_CORE;
+ b.grabtype = GRABTYPE_CORE;
+ a.device = &dev1;
+ b.device = &dev1;
+ a.modifierDevice = &dev1;
+ b.modifierDevice = &dev2;
+ rc = GrabMatchesSecond(&a, &b, FALSE);
+ g_assert(rc == FALSE);
+
+ /* different event type must fail */
+ a.grabtype = GRABTYPE_XI2;
+ b.grabtype = GRABTYPE_XI2;
+ a.device = &dev1;
+ b.device = &dev1;
+ a.modifierDevice = &dev1;
+ b.modifierDevice = &dev1;
+ a.type = XI_KeyPress;
+ b.type = XI_KeyRelease;
+ rc = GrabMatchesSecond(&a, &b, FALSE);
+ g_assert(rc == FALSE);
+ rc = GrabMatchesSecond(&a, &b, TRUE);
+ g_assert(rc == FALSE);
+
+ a.grabtype = GRABTYPE_CORE;
+ b.grabtype = GRABTYPE_CORE;
+ a.device = &dev1;
+ b.device = &dev1;
+ a.modifierDevice = &dev1;
+ b.modifierDevice = &dev1;
+ a.type = XI_KeyPress;
+ b.type = XI_KeyRelease;
+ rc = GrabMatchesSecond(&a, &b, FALSE);
+ g_assert(rc == FALSE);
+ rc = GrabMatchesSecond(&a, &b, TRUE);
+ g_assert(rc == FALSE);
+
+ a.grabtype = GRABTYPE_XI;
+ b.grabtype = GRABTYPE_XI;
+ a.device = &dev1;
+ b.device = &dev1;
+ a.modifierDevice = &dev1;
+ b.modifierDevice = &dev1;
+ a.type = XI_KeyPress;
+ b.type = XI_KeyRelease;
+ rc = GrabMatchesSecond(&a, &b, FALSE);
+ g_assert(rc == FALSE);
+ rc = GrabMatchesSecond(&a, &b, TRUE);
+ g_assert(rc == FALSE);
+
+ /* different modifiers must fail */
+ a.grabtype = GRABTYPE_XI2;
+ b.grabtype = GRABTYPE_XI2;
+ a.device = &dev1;
+ b.device = &dev1;
+ a.modifierDevice = &dev1;
+ b.modifierDevice = &dev1;
+ a.type = XI_KeyPress;
+ b.type = XI_KeyPress;
+ a.modifiersDetail.exact = 1;
+ b.modifiersDetail.exact = 2;
+ rc = GrabMatchesSecond(&a, &b, FALSE);
+ g_assert(rc == FALSE);
+ rc = GrabMatchesSecond(&b, &a, FALSE);
+ g_assert(rc == FALSE);
+
+ a.grabtype = GRABTYPE_CORE;
+ b.grabtype = GRABTYPE_CORE;
+ rc = GrabMatchesSecond(&a, &b, FALSE);
+ g_assert(rc == FALSE);
+ rc = GrabMatchesSecond(&b, &a, FALSE);
+ g_assert(rc == FALSE);
+
+ a.grabtype = GRABTYPE_XI;
+ b.grabtype = GRABTYPE_XI;
+ rc = GrabMatchesSecond(&a, &b, FALSE);
+ g_assert(rc == FALSE);
+ rc = GrabMatchesSecond(&b, &a, FALSE);
+ g_assert(rc == FALSE);
+
+ /* AnyModifier must fail for XI2 */
+ a.grabtype = GRABTYPE_XI2;
+ b.grabtype = GRABTYPE_XI2;
+ a.modifiersDetail.exact = AnyModifier;
+ b.modifiersDetail.exact = 1;
+ rc = GrabMatchesSecond(&a, &b, FALSE);
+ g_assert(rc == FALSE);
+ rc = GrabMatchesSecond(&b, &a, FALSE);
+ g_assert(rc == FALSE);
+
+ /* XIAnyModifier must fail for CORE and XI */
+ a.grabtype = GRABTYPE_XI;
+ b.grabtype = GRABTYPE_XI;
+ a.modifiersDetail.exact = XIAnyModifier;
+ b.modifiersDetail.exact = 1;
+ rc = GrabMatchesSecond(&a, &b, FALSE);
+ g_assert(rc == FALSE);
+ rc = GrabMatchesSecond(&b, &a, FALSE);
+ g_assert(rc == FALSE);
+
+ a.grabtype = GRABTYPE_CORE;
+ b.grabtype = GRABTYPE_CORE;
+ a.modifiersDetail.exact = XIAnyModifier;
+ b.modifiersDetail.exact = 1;
+ rc = GrabMatchesSecond(&a, &b, FALSE);
+ g_assert(rc == FALSE);
+ rc = GrabMatchesSecond(&b, &a, FALSE);
+ g_assert(rc == FALSE);
+
+ /* different detail must fail */
+ a.grabtype = GRABTYPE_XI2;
+ b.grabtype = GRABTYPE_XI2;
+ a.detail.exact = 1;
+ b.detail.exact = 2;
+ a.modifiersDetail.exact = 1;
+ b.modifiersDetail.exact = 1;
+ rc = GrabMatchesSecond(&a, &b, FALSE);
+ g_assert(rc == FALSE);
+ rc = GrabMatchesSecond(&b, &a, FALSE);
+ g_assert(rc == FALSE);
+
+ a.grabtype = GRABTYPE_XI;
+ b.grabtype = GRABTYPE_XI;
+ rc = GrabMatchesSecond(&a, &b, FALSE);
+ g_assert(rc == FALSE);
+ rc = GrabMatchesSecond(&b, &a, FALSE);
+ g_assert(rc == FALSE);
+
+ a.grabtype = GRABTYPE_CORE;
+ b.grabtype = GRABTYPE_CORE;
+ rc = GrabMatchesSecond(&a, &b, FALSE);
+ g_assert(rc == FALSE);
+ rc = GrabMatchesSecond(&b, &a, FALSE);
+ g_assert(rc == FALSE);
+
+ /* detail of AnyModifier must fail */
+ a.grabtype = GRABTYPE_XI2;
+ b.grabtype = GRABTYPE_XI2;
+ a.detail.exact = AnyModifier;
+ b.detail.exact = 1;
+ a.modifiersDetail.exact = 1;
+ b.modifiersDetail.exact = 1;
+ rc = GrabMatchesSecond(&a, &b, FALSE);
+ g_assert(rc == FALSE);
+ rc = GrabMatchesSecond(&b, &a, FALSE);
+ g_assert(rc == FALSE);
+
+ a.grabtype = GRABTYPE_CORE;
+ b.grabtype = GRABTYPE_CORE;
+ rc = GrabMatchesSecond(&a, &b, FALSE);
+ g_assert(rc == FALSE);
+ rc = GrabMatchesSecond(&b, &a, FALSE);
+ g_assert(rc == FALSE);
+
+ a.grabtype = GRABTYPE_XI;
+ b.grabtype = GRABTYPE_XI;
+ rc = GrabMatchesSecond(&a, &b, FALSE);
+ g_assert(rc == FALSE);
+ rc = GrabMatchesSecond(&b, &a, FALSE);
+ g_assert(rc == FALSE);
+
+ /* detail of XIAnyModifier must fail */
+ a.grabtype = GRABTYPE_XI2;
+ b.grabtype = GRABTYPE_XI2;
+ a.detail.exact = XIAnyModifier;
+ b.detail.exact = 1;
+ a.modifiersDetail.exact = 1;
+ b.modifiersDetail.exact = 1;
+ rc = GrabMatchesSecond(&a, &b, FALSE);
+ g_assert(rc == FALSE);
+ rc = GrabMatchesSecond(&b, &a, FALSE);
+ g_assert(rc == FALSE);
+
+ a.grabtype = GRABTYPE_CORE;
+ b.grabtype = GRABTYPE_CORE;
+ rc = GrabMatchesSecond(&a, &b, FALSE);
+ g_assert(rc == FALSE);
+ rc = GrabMatchesSecond(&b, &a, FALSE);
+ g_assert(rc == FALSE);
+
+ a.grabtype = GRABTYPE_XI;
+ b.grabtype = GRABTYPE_XI;
+ rc = GrabMatchesSecond(&a, &b, FALSE);
+ g_assert(rc == FALSE);
+ rc = GrabMatchesSecond(&b, &a, FALSE);
+ g_assert(rc == FALSE);
+
+ /* XIAnyModifier or AnyModifer must succeed */
+ a.grabtype = GRABTYPE_XI2;
+ b.grabtype = GRABTYPE_XI2;
+ a.detail.exact = 1;
+ b.detail.exact = 1;
+ a.modifiersDetail.exact = XIAnyModifier;
+ b.modifiersDetail.exact = 1;
+ rc = GrabMatchesSecond(&a, &b, FALSE);
+ g_assert(rc == TRUE);
+ rc = GrabMatchesSecond(&b, &a, FALSE);
+ g_assert(rc == TRUE);
+
+ a.grabtype = GRABTYPE_CORE;
+ b.grabtype = GRABTYPE_CORE;
+ a.detail.exact = 1;
+ b.detail.exact = 1;
+ a.modifiersDetail.exact = AnyModifier;
+ b.modifiersDetail.exact = 1;
+ rc = GrabMatchesSecond(&a, &b, FALSE);
+ g_assert(rc == TRUE);
+ rc = GrabMatchesSecond(&b, &a, FALSE);
+ g_assert(rc == TRUE);
+
+ a.grabtype = GRABTYPE_XI;
+ b.grabtype = GRABTYPE_XI;
+ a.detail.exact = 1;
+ b.detail.exact = 1;
+ a.modifiersDetail.exact = AnyModifier;
+ b.modifiersDetail.exact = 1;
+ rc = GrabMatchesSecond(&a, &b, FALSE);
+ g_assert(rc == TRUE);
+ rc = GrabMatchesSecond(&b, &a, FALSE);
+ g_assert(rc == TRUE);
+
+ /* AnyKey or XIAnyKeycode must succeed */
+ a.grabtype = GRABTYPE_XI2;
+ b.grabtype = GRABTYPE_XI2;
+ a.detail.exact = XIAnyKeycode;
+ b.detail.exact = 1;
+ a.modifiersDetail.exact = 1;
+ b.modifiersDetail.exact = 1;
+ rc = GrabMatchesSecond(&a, &b, FALSE);
+ g_assert(rc == TRUE);
+ rc = GrabMatchesSecond(&b, &a, FALSE);
+ g_assert(rc == TRUE);
+
+ a.grabtype = GRABTYPE_CORE;
+ b.grabtype = GRABTYPE_CORE;
+ a.detail.exact = AnyKey;
+ b.detail.exact = 1;
+ a.modifiersDetail.exact = 1;
+ b.modifiersDetail.exact = 1;
+ rc = GrabMatchesSecond(&a, &b, FALSE);
+ g_assert(rc == TRUE);
+ rc = GrabMatchesSecond(&b, &a, FALSE);
+ g_assert(rc == TRUE);
+
+ a.grabtype = GRABTYPE_XI;
+ b.grabtype = GRABTYPE_XI;
+ a.detail.exact = AnyKey;
+ b.detail.exact = 1;
+ a.modifiersDetail.exact = 1;
+ b.modifiersDetail.exact = 1;
+ rc = GrabMatchesSecond(&a, &b, FALSE);
+ g_assert(rc == TRUE);
+ rc = GrabMatchesSecond(&b, &a, FALSE);
+ g_assert(rc == TRUE);
+}
+
+static void test_bits_to_byte(int i)
+{
+ int expected_bytes;
+ expected_bytes = (i + 7)/8;
+
+ g_assert(bits_to_bytes(i) >= i/8);
+ g_assert((bits_to_bytes(i) * 8) - i <= 7);
+ g_assert(expected_bytes == bits_to_bytes(i));
+}
+
+static void test_bytes_to_int32(int i)
+{
+ int expected_4byte;
+ expected_4byte = (i + 3)/4;
+
+ g_assert(bytes_to_int32(i) <= i);
+ g_assert((bytes_to_int32(i) * 4) - i <= 3);
+ g_assert(expected_4byte == bytes_to_int32(i));
+}
+
+static void test_pad_to_int32(int i)
+{
+ int expected_bytes;
+ expected_bytes = ((i + 3)/4) * 4;
+
+ g_assert(pad_to_int32(i) >= i);
+ g_assert(pad_to_int32(i) - i <= 3);
+ g_assert(expected_bytes == pad_to_int32(i));
+}
+static void include_byte_padding_macros(void)
+{
+ g_test_message("Testing bits_to_bytes()");
+
+ /* the macros don't provide overflow protection */
+ test_bits_to_byte(0);
+ test_bits_to_byte(1);
+ test_bits_to_byte(2);
+ test_bits_to_byte(7);
+ test_bits_to_byte(8);
+ test_bits_to_byte(0xFF);
+ test_bits_to_byte(0x100);
+ test_bits_to_byte(INT_MAX - 9);
+ test_bits_to_byte(INT_MAX - 8);
+
+ g_test_message("Testing bytes_to_int32()");
+
+ test_bytes_to_int32(0);
+ test_bytes_to_int32(1);
+ test_bytes_to_int32(2);
+ test_bytes_to_int32(7);
+ test_bytes_to_int32(8);
+ test_bytes_to_int32(0xFF);
+ test_bytes_to_int32(0x100);
+ test_bytes_to_int32(0xFFFF);
+ test_bytes_to_int32(0x10000);
+ test_bytes_to_int32(0xFFFFFF);
+ test_bytes_to_int32(0x1000000);
+ test_bytes_to_int32(INT_MAX - 4);
+ test_bytes_to_int32(INT_MAX - 3);
+
+ g_test_message("Testing pad_to_int32");
+
+ test_pad_to_int32(0);
+ test_pad_to_int32(0);
+ test_pad_to_int32(1);
+ test_pad_to_int32(2);
+ test_pad_to_int32(7);
+ test_pad_to_int32(8);
+ test_pad_to_int32(0xFF);
+ test_pad_to_int32(0x100);
+ test_pad_to_int32(0xFFFF);
+ test_pad_to_int32(0x10000);
+ test_pad_to_int32(0xFFFFFF);
+ test_pad_to_int32(0x1000000);
+ test_pad_to_int32(INT_MAX - 4);
+ test_pad_to_int32(INT_MAX - 3);
+}
+
+static void xi_unregister_handlers(void)
+{
+ DeviceIntRec dev;
+ int handler;
+
+ memset(&dev, 0, sizeof(dev));
+
+ handler = XIRegisterPropertyHandler(&dev, NULL, NULL, NULL);
+ g_assert(handler == 1);
+ handler = XIRegisterPropertyHandler(&dev, NULL, NULL, NULL);
+ g_assert(handler == 2);
+ handler = XIRegisterPropertyHandler(&dev, NULL, NULL, NULL);
+ g_assert(handler == 3);
+
+ g_test_message("Unlinking from front.");
+
+ XIUnregisterPropertyHandler(&dev, 4); /* NOOP */
+ g_assert(dev.properties.handlers->id == 3);
+ XIUnregisterPropertyHandler(&dev, 3);
+ g_assert(dev.properties.handlers->id == 2);
+ XIUnregisterPropertyHandler(&dev, 2);
+ g_assert(dev.properties.handlers->id == 1);
+ XIUnregisterPropertyHandler(&dev, 1);
+ g_assert(dev.properties.handlers == NULL);
+
+ handler = XIRegisterPropertyHandler(&dev, NULL, NULL, NULL);
+ g_assert(handler == 4);
+ handler = XIRegisterPropertyHandler(&dev, NULL, NULL, NULL);
+ g_assert(handler == 5);
+ handler = XIRegisterPropertyHandler(&dev, NULL, NULL, NULL);
+ g_assert(handler == 6);
+ XIUnregisterPropertyHandler(&dev, 3); /* NOOP */
+ g_assert(dev.properties.handlers->next->next->next == NULL);
+ XIUnregisterPropertyHandler(&dev, 4);
+ g_assert(dev.properties.handlers->next->next == NULL);
+ XIUnregisterPropertyHandler(&dev, 5);
+ g_assert(dev.properties.handlers->next == NULL);
+ XIUnregisterPropertyHandler(&dev, 6);
+ g_assert(dev.properties.handlers == NULL);
+
+ handler = XIRegisterPropertyHandler(&dev, NULL, NULL, NULL);
+ g_assert(handler == 7);
+ handler = XIRegisterPropertyHandler(&dev, NULL, NULL, NULL);
+ g_assert(handler == 8);
+ handler = XIRegisterPropertyHandler(&dev, NULL, NULL, NULL);
+ g_assert(handler == 9);
+
+ XIDeleteAllDeviceProperties(&dev);
+ g_assert(dev.properties.handlers == NULL);
+ XIUnregisterPropertyHandler(&dev, 7); /* NOOP */
+
+}
+
+static void cmp_attr_fields(InputAttributes *attr1,
+ InputAttributes *attr2)
+{
+ char **tags1, **tags2;
+
+ g_assert(attr1 && attr2);
+ g_assert(attr1 != attr2);
+ g_assert(attr1->flags == attr2->flags);
+
+ if (attr1->product != NULL)
+ {
+ g_assert(attr1->product != attr2->product);
+ g_assert(strcmp(attr1->product, attr2->product) == 0);
+ } else
+ g_assert(attr2->product == NULL);
+
+ if (attr1->vendor != NULL)
+ {
+ g_assert(attr1->vendor != attr2->vendor);
+ g_assert(strcmp(attr1->vendor, attr2->vendor) == 0);
+ } else
+ g_assert(attr2->vendor == NULL);
+
+ if (attr1->device != NULL)
+ {
+ g_assert(attr1->device != attr2->device);
+ g_assert(strcmp(attr1->device, attr2->device) == 0);
+ } else
+ g_assert(attr2->device == NULL);
+
+ if (attr1->pnp_id != NULL)
+ {
+ g_assert(attr1->pnp_id != attr2->pnp_id);
+ g_assert(strcmp(attr1->pnp_id, attr2->pnp_id) == 0);
+ } else
+ g_assert(attr2->pnp_id == NULL);
+
+ if (attr1->usb_id != NULL)
+ {
+ g_assert(attr1->usb_id != attr2->usb_id);
+ g_assert(strcmp(attr1->usb_id, attr2->usb_id) == 0);
+ } else
+ g_assert(attr2->usb_id == NULL);
+
+ tags1 = attr1->tags;
+ tags2 = attr2->tags;
+
+ /* if we don't have any tags, skip the tag checking bits */
+ if (!tags1)
+ {
+ g_assert(!tags2);
+ return;
+ }
+
+ /* Don't lug around empty arrays */
+ g_assert(*tags1);
+ g_assert(*tags2);
+
+ /* check for identical content, but duplicated */
+ while (*tags1)
+ {
+ g_assert(*tags1 != *tags2);
+ g_assert(strcmp(*tags1, *tags2) == 0);
+ tags1++;
+ tags2++;
+ }
+
+ /* ensure tags1 and tags2 have the same no of elements */
+ g_assert(!*tags2);
+
+ /* check for not sharing memory */
+ tags1 = attr1->tags;
+ while (*tags1)
+ {
+ tags2 = attr2->tags;
+ while (*tags2)
+ g_assert(*tags1 != *tags2++);
+
+ tags1++;
+ }
+}
+
+static void dix_input_attributes(void)
+{
+ InputAttributes orig = {0};
+ InputAttributes *new;
+ char *tags[4] = {"tag1", "tag2", "tag2", NULL};
+
+ new = DuplicateInputAttributes(NULL);
+ g_assert(!new);
+
+ new = DuplicateInputAttributes(&orig);
+ g_assert(memcmp(&orig, new, sizeof(InputAttributes)) == 0);
+
+ orig.product = "product name";
+ new = DuplicateInputAttributes(&orig);
+ cmp_attr_fields(&orig, new);
+ FreeInputAttributes(new);
+
+ orig.vendor = "vendor name";
+ new = DuplicateInputAttributes(&orig);
+ cmp_attr_fields(&orig, new);
+ FreeInputAttributes(new);
+
+ orig.device = "device path";
+ new = DuplicateInputAttributes(&orig);
+ cmp_attr_fields(&orig, new);
+ FreeInputAttributes(new);
+
+ orig.pnp_id = "PnPID";
+ new = DuplicateInputAttributes(&orig);
+ cmp_attr_fields(&orig, new);
+ FreeInputAttributes(new);
+
+ orig.usb_id = "USBID";
+ new = DuplicateInputAttributes(&orig);
+ cmp_attr_fields(&orig, new);
+ FreeInputAttributes(new);
+
+ orig.flags = 0xF0;
+ new = DuplicateInputAttributes(&orig);
+ cmp_attr_fields(&orig, new);
+ FreeInputAttributes(new);
+
+ orig.tags = tags;
+ new = DuplicateInputAttributes(&orig);
+ cmp_attr_fields(&orig, new);
+ FreeInputAttributes(new);
+}
+
+static void dix_input_valuator_masks(void)
+{
+ ValuatorMask *mask = NULL, *copy;
+ int nvaluators = MAX_VALUATORS;
+ int valuators[nvaluators];
+ int i;
+ int first_val, num_vals;
+
+ for (i = 0; i < nvaluators; i++)
+ valuators[i] = i;
+
+ mask = valuator_mask_new(nvaluators);
+ g_assert(mask != NULL);
+ g_assert(valuator_mask_size(mask) == 0);
+ g_assert(valuator_mask_num_valuators(mask) == 0);
+
+ for (i = 0; i < nvaluators; i++)
+ {
+ g_assert(!valuator_mask_isset(mask, i));
+ valuator_mask_set(mask, i, valuators[i]);
+ g_assert(valuator_mask_isset(mask, i));
+ g_assert(valuator_mask_get(mask, i) == valuators[i]);
+ g_assert(valuator_mask_size(mask) == i + 1);
+ g_assert(valuator_mask_num_valuators(mask) == i + 1);
+ }
+
+ for (i = 0; i < nvaluators; i++)
+ {
+ g_assert(valuator_mask_isset(mask, i));
+ valuator_mask_unset(mask, i);
+ /* we're removing valuators from the front, so size should stay the
+ * same until the last bit is removed */
+ if (i < nvaluators - 1)
+ g_assert(valuator_mask_size(mask) == nvaluators);
+ g_assert(!valuator_mask_isset(mask, i));
+ }
+
+ g_assert(valuator_mask_size(mask) == 0);
+ valuator_mask_zero(mask);
+ g_assert(valuator_mask_size(mask) == 0);
+ g_assert(valuator_mask_num_valuators(mask) == 0);
+ for (i = 0; i < nvaluators; i++)
+ g_assert(!valuator_mask_isset(mask, i));
+
+ first_val = 5;
+ num_vals = 6;
+
+ valuator_mask_set_range(mask, first_val, num_vals, valuators);
+ g_assert(valuator_mask_size(mask) == first_val + num_vals);
+ g_assert(valuator_mask_num_valuators(mask) == num_vals);
+ for (i = 0; i < nvaluators; i++)
+ {
+ if (i < first_val || i >= first_val + num_vals)
+ g_assert(!valuator_mask_isset(mask, i));
+ else
+ {
+ g_assert(valuator_mask_isset(mask, i));
+ g_assert(valuator_mask_get(mask, i) == valuators[i - first_val]);
+ }
+ }
+
+ copy = valuator_mask_new(nvaluators);
+ valuator_mask_copy(copy, mask);
+ g_assert(mask != copy);
+ g_assert(valuator_mask_size(mask) == valuator_mask_size(copy));
+ g_assert(valuator_mask_num_valuators(mask) == valuator_mask_num_valuators(copy));
+
+ for (i = 0; i < nvaluators; i++)
+ {
+ g_assert(valuator_mask_isset(mask, i) == valuator_mask_isset(copy, i));
+ g_assert(valuator_mask_get(mask, i) == valuator_mask_get(copy, i));
+ }
+
+ valuator_mask_free(&mask);
+ g_assert(mask == NULL);
+}
+
+static void dix_valuator_mode(void)
+{
+ DeviceIntRec dev;
+ const int num_axes = MAX_VALUATORS;
+ int i;
+ Atom atoms[MAX_VALUATORS] = { 0 };
+
+ memset(&dev, 0, sizeof(DeviceIntRec));
+ dev.type = MASTER_POINTER; /* claim it's a master to stop ptracccel */
+
+ g_assert(InitValuatorClassDeviceStruct(NULL, 0, atoms, 0, 0) == FALSE);
+ g_assert(InitValuatorClassDeviceStruct(&dev, num_axes, atoms, 0, Absolute));
+
+ for (i = 0; i < num_axes; i++)
+ {
+ g_assert(valuator_get_mode(&dev, i) == Absolute);
+ valuator_set_mode(&dev, i, Relative);
+ g_assert(dev.valuator->axes[i].mode == Relative);
+ g_assert(valuator_get_mode(&dev, i) == Relative);
+ }
+
+ valuator_set_mode(&dev, VALUATOR_MODE_ALL_AXES, Absolute);
+ for (i = 0; i < num_axes; i++)
+ g_assert(valuator_get_mode(&dev, i) == Absolute);
+
+ valuator_set_mode(&dev, VALUATOR_MODE_ALL_AXES, Relative);
+ for (i = 0; i < num_axes; i++)
+ g_assert(valuator_get_mode(&dev, i) == Relative);
+}
+
+static void include_bit_test_macros(void)
+{
+ uint8_t mask[9] = { 0 };
+ int i;
+
+ for (i = 0; i < sizeof(mask)/sizeof(mask[0]); i++)
+ {
+ g_assert(BitIsOn(mask, i) == 0);
+ SetBit(mask, i);
+ g_assert(BitIsOn(mask, i) == 1);
+ g_assert(!!(mask[i/8] & (1 << (i % 8))));
+ g_assert(CountBits(mask, sizeof(mask)) == 1);
+ ClearBit(mask, i);
+ g_assert(BitIsOn(mask, i) == 0);
+ }
+}
+
+int main(int argc, char** argv)
+{
+ g_test_init(&argc, &argv,NULL);
+ g_test_bug_base("https://bugzilla.freedesktop.org/show_bug.cgi?id=");
+
+ g_test_add_func("/dix/input/valuator-masks", dix_input_valuator_masks);
+ g_test_add_func("/dix/input/attributes", dix_input_attributes);
+ g_test_add_func("/dix/input/init-valuators", dix_init_valuators);
+ g_test_add_func("/dix/input/event-core-conversion", dix_event_to_core_conversion);
+ g_test_add_func("/dix/input/event-xi1-conversion", dix_event_to_xi1_conversion);
+ g_test_add_func("/dix/input/check-grab-values", dix_check_grab_values);
+ g_test_add_func("/dix/input/xi2-struct-sizes", xi2_struct_sizes);
+ g_test_add_func("/dix/input/grab_matching", dix_grab_matching);
+ g_test_add_func("/dix/input/valuator_mode", dix_valuator_mode);
+ g_test_add_func("/include/byte_padding_macros", include_byte_padding_macros);
+ g_test_add_func("/include/bit_test_macros", include_bit_test_macros);
+ g_test_add_func("/Xi/xiproperty/register-unregister", xi_unregister_handlers);
+
+
+ return g_test_run();
+}
|