diff options
| author | marha <marha@users.sourceforge.net> | 2010-10-11 07:06:15 +0000 |
|---|---|---|
| committer | marha <marha@users.sourceforge.net> | 2010-10-11 07:06:15 +0000 |
| commit | 752f86c17342e07a858aea75106eb8316a226d89 (patch) | |
| tree | 03429163ab9e7a871b346d7a71754b877d99a120 | |
| parent | ca01727646bcc035a36b90f2a2a38b3dd9ce8c93 (diff) | |
| download | vcxsrv-752f86c17342e07a858aea75106eb8316a226d89.tar.gz vcxsrv-752f86c17342e07a858aea75106eb8316a226d89.tar.bz2 vcxsrv-752f86c17342e07a858aea75106eb8316a226d89.zip | |
pixman update 11/10/2010
| -rw-r--r-- | pixman/pixman/pixman-arm-neon.c | 8 | ||||
| -rw-r--r-- | pixman/pixman/pixman-fast-path.c | 51 |
2 files changed, 54 insertions, 5 deletions
diff --git a/pixman/pixman/pixman-arm-neon.c b/pixman/pixman/pixman-arm-neon.c index 40f7b6a60..d22922a4a 100644 --- a/pixman/pixman/pixman-arm-neon.c +++ b/pixman/pixman/pixman-arm-neon.c @@ -355,9 +355,13 @@ BIND_COMBINE_U (out_reverse) pixman_implementation_t *
_pixman_implementation_create_arm_neon (void)
{
- pixman_implementation_t *general = _pixman_implementation_create_fast_path ();
+#ifdef USE_ARM_SIMD
+ pixman_implementation_t *fallback = _pixman_implementation_create_arm_simd ();
+#else
+ pixman_implementation_t *fallback = _pixman_implementation_create_fast_path ();
+#endif
pixman_implementation_t *imp =
- _pixman_implementation_create (general, arm_neon_fast_paths);
+ _pixman_implementation_create (fallback, arm_neon_fast_paths);
imp->combine_32[PIXMAN_OP_OVER] = neon_combine_over_u;
imp->combine_32[PIXMAN_OP_ADD] = neon_combine_add_u;
diff --git a/pixman/pixman/pixman-fast-path.c b/pixman/pixman/pixman-fast-path.c index 25ef9243b..dc8fad06d 100644 --- a/pixman/pixman/pixman-fast-path.c +++ b/pixman/pixman/pixman-fast-path.c @@ -1399,15 +1399,60 @@ FAST_NEAREST (8888_565_cover, 8888, 0565, uint32_t, uint16_t, SRC, COVER); FAST_NEAREST (8888_565_none, 8888, 0565, uint32_t, uint16_t, SRC, NONE);
FAST_NEAREST (8888_565_pad, 8888, 0565, uint32_t, uint16_t, SRC, PAD);
FAST_NEAREST (8888_565_normal, 8888, 0565, uint32_t, uint16_t, SRC, NORMAL);
-FAST_NEAREST (565_565_cover, 0565, 0565, uint16_t, uint16_t, SRC, COVER);
-FAST_NEAREST (565_565_none, 0565, 0565, uint16_t, uint16_t, SRC, NONE);
-FAST_NEAREST (565_565_pad, 0565, 0565, uint16_t, uint16_t, SRC, PAD);
FAST_NEAREST (565_565_normal, 0565, 0565, uint16_t, uint16_t, SRC, NORMAL);
FAST_NEAREST (8888_565_cover, 8888, 0565, uint32_t, uint16_t, OVER, COVER);
FAST_NEAREST (8888_565_none, 8888, 0565, uint32_t, uint16_t, OVER, NONE);
FAST_NEAREST (8888_565_pad, 8888, 0565, uint32_t, uint16_t, OVER, PAD);
FAST_NEAREST (8888_565_normal, 8888, 0565, uint32_t, uint16_t, OVER, NORMAL);
+/* Use more unrolling for src_0565_0565 because it is typically CPU bound */
+static force_inline void
+scaled_nearest_scanline_565_565_SRC (uint16_t * dst,
+ uint16_t * src,
+ int32_t w,
+ pixman_fixed_t vx,
+ pixman_fixed_t unit_x,
+ pixman_fixed_t max_vx)
+{
+ uint16_t tmp1, tmp2, tmp3, tmp4;
+ while ((w -= 4) >= 0)
+ {
+ tmp1 = src[pixman_fixed_to_int (vx)];
+ vx += unit_x;
+ tmp2 = src[pixman_fixed_to_int (vx)];
+ vx += unit_x;
+ tmp3 = src[pixman_fixed_to_int (vx)];
+ vx += unit_x;
+ tmp4 = src[pixman_fixed_to_int (vx)];
+ vx += unit_x;
+ *dst++ = tmp1;
+ *dst++ = tmp2;
+ *dst++ = tmp3;
+ *dst++ = tmp4;
+ }
+ if (w & 2)
+ {
+ tmp1 = src[pixman_fixed_to_int (vx)];
+ vx += unit_x;
+ tmp2 = src[pixman_fixed_to_int (vx)];
+ vx += unit_x;
+ *dst++ = tmp1;
+ *dst++ = tmp2;
+ }
+ if (w & 1)
+ *dst++ = src[pixman_fixed_to_int (vx)];
+}
+
+FAST_NEAREST_MAINLOOP (565_565_cover_SRC,
+ scaled_nearest_scanline_565_565_SRC,
+ uint16_t, uint16_t, COVER);
+FAST_NEAREST_MAINLOOP (565_565_none_SRC,
+ scaled_nearest_scanline_565_565_SRC,
+ uint16_t, uint16_t, NONE);
+FAST_NEAREST_MAINLOOP (565_565_pad_SRC,
+ scaled_nearest_scanline_565_565_SRC,
+ uint16_t, uint16_t, PAD);
+
static force_inline uint32_t
fetch_nearest (pixman_repeat_t src_repeat,
pixman_format_code_t format,
|