aboutsummaryrefslogtreecommitdiff
path: root/openssl/crypto/x86cpuid.pl
diff options
context:
space:
mode:
authormarha <marha@users.sourceforge.net>2009-06-28 22:07:26 +0000
committermarha <marha@users.sourceforge.net>2009-06-28 22:07:26 +0000
commit3562e78743202e43aec8727005182a2558117eca (patch)
tree8f9113a77d12470c5c851a2a8e4cb02e89df7d43 /openssl/crypto/x86cpuid.pl
downloadvcxsrv-3562e78743202e43aec8727005182a2558117eca.tar.gz
vcxsrv-3562e78743202e43aec8727005182a2558117eca.tar.bz2
vcxsrv-3562e78743202e43aec8727005182a2558117eca.zip
Checked in the following released items:
xkeyboard-config-1.4.tar.gz ttf-bitstream-vera-1.10.tar.gz font-alias-1.0.1.tar.gz font-sun-misc-1.0.0.tar.gz font-sun-misc-1.0.0.tar.gz font-sony-misc-1.0.0.tar.gz font-schumacher-misc-1.0.0.tar.gz font-mutt-misc-1.0.0.tar.gz font-misc-misc-1.0.0.tar.gz font-misc-meltho-1.0.0.tar.gz font-micro-misc-1.0.0.tar.gz font-jis-misc-1.0.0.tar.gz font-isas-misc-1.0.0.tar.gz font-dec-misc-1.0.0.tar.gz font-daewoo-misc-1.0.0.tar.gz font-cursor-misc-1.0.0.tar.gz font-arabic-misc-1.0.0.tar.gz font-winitzki-cyrillic-1.0.0.tar.gz font-misc-cyrillic-1.0.0.tar.gz font-cronyx-cyrillic-1.0.0.tar.gz font-screen-cyrillic-1.0.1.tar.gz font-xfree86-type1-1.0.1.tar.gz font-adobe-utopia-type1-1.0.1.tar.gz font-ibm-type1-1.0.0.tar.gz font-bitstream-type1-1.0.0.tar.gz font-bitstream-speedo-1.0.0.tar.gz font-bh-ttf-1.0.0.tar.gz font-bh-type1-1.0.0.tar.gz font-bitstream-100dpi-1.0.0.tar.gz font-bh-lucidatypewriter-100dpi-1.0.0.tar.gz font-bh-100dpi-1.0.0.tar.gz font-adobe-utopia-100dpi-1.0.1.tar.gz font-adobe-100dpi-1.0.0.tar.gz font-util-1.0.1.tar.gz font-bitstream-75dpi-1.0.0.tar.gz font-bh-lucidatypewriter-75dpi-1.0.0.tar.gz font-adobe-utopia-75dpi-1.0.1.tar.gz font-bh-75dpi-1.0.0.tar.gz bdftopcf-1.0.1.tar.gz font-adobe-75dpi-1.0.0.tar.gz mkfontscale-1.0.6.tar.gz openssl-0.9.8k.tar.gz bigreqsproto-1.0.2.tar.gz xtrans-1.2.2.tar.gz resourceproto-1.0.2.tar.gz inputproto-1.4.4.tar.gz compositeproto-0.4.tar.gz damageproto-1.1.0.tar.gz zlib-1.2.3.tar.gz xkbcomp-1.0.5.tar.gz freetype-2.3.9.tar.gz pthreads-w32-2-8-0-release.tar.gz pixman-0.12.0.tar.gz kbproto-1.0.3.tar.gz evieext-1.0.2.tar.gz fixesproto-4.0.tar.gz recordproto-1.13.2.tar.gz randrproto-1.2.2.tar.gz scrnsaverproto-1.1.0.tar.gz renderproto-0.9.3.tar.gz xcmiscproto-1.1.2.tar.gz fontsproto-2.0.2.tar.gz xextproto-7.0.3.tar.gz xproto-7.0.14.tar.gz libXdmcp-1.0.2.tar.gz libxkbfile-1.0.5.tar.gz libfontenc-1.0.4.tar.gz libXfont-1.3.4.tar.gz libX11-1.1.5.tar.gz libXau-1.0.4.tar.gz libxcb-1.1.tar.gz xorg-server-1.5.3.tar.gz
Diffstat (limited to 'openssl/crypto/x86cpuid.pl')
-rw-r--r--openssl/crypto/x86cpuid.pl225
1 files changed, 225 insertions, 0 deletions
diff --git a/openssl/crypto/x86cpuid.pl b/openssl/crypto/x86cpuid.pl
new file mode 100644
index 000000000..4408ef293
--- /dev/null
+++ b/openssl/crypto/x86cpuid.pl
@@ -0,0 +1,225 @@
+#!/usr/bin/env perl
+
+push(@INC,"perlasm");
+require "x86asm.pl";
+
+&asm_init($ARGV[0],"x86cpuid");
+
+for (@ARGV) { $sse2=1 if (/-DOPENSSL_IA32_SSE2/); }
+
+&function_begin("OPENSSL_ia32_cpuid");
+ &xor ("edx","edx");
+ &pushf ();
+ &pop ("eax");
+ &mov ("ecx","eax");
+ &xor ("eax",1<<21);
+ &push ("eax");
+ &popf ();
+ &pushf ();
+ &pop ("eax");
+ &xor ("ecx","eax");
+ &bt ("ecx",21);
+ &jnc (&label("done"));
+ &xor ("eax","eax");
+ &cpuid ();
+ &xor ("eax","eax");
+ &cmp ("ebx",0x756e6547); # "Genu"
+ &data_byte(0x0f,0x95,0xc0); #&setne (&LB("eax"));
+ &mov ("ebp","eax");
+ &cmp ("edx",0x49656e69); # "ineI"
+ &data_byte(0x0f,0x95,0xc0); #&setne (&LB("eax"));
+ &or ("ebp","eax");
+ &cmp ("ecx",0x6c65746e); # "ntel"
+ &data_byte(0x0f,0x95,0xc0); #&setne (&LB("eax"));
+ &or ("ebp","eax");
+ &mov ("eax",1);
+ &cpuid ();
+ &cmp ("ebp",0);
+ &jne (&label("notP4"));
+ &and ("eax",15<<8); # familiy ID
+ &cmp ("eax",15<<8); # P4?
+ &jne (&label("notP4"));
+ &or ("edx",1<<20); # use reserved bit to engage RC4_CHAR
+&set_label("notP4");
+ &bt ("edx",28); # test hyper-threading bit
+ &jnc (&label("done"));
+ &shr ("ebx",16);
+ &and ("ebx",0xff);
+ &cmp ("ebx",1); # see if cache is shared(*)
+ &ja (&label("done"));
+ &and ("edx",0xefffffff); # clear hyper-threading bit if not
+&set_label("done");
+ &mov ("eax","edx");
+ &mov ("edx","ecx");
+&function_end("OPENSSL_ia32_cpuid");
+# (*) on Core2 this value is set to 2 denoting the fact that L2
+# cache is shared between cores.
+
+&external_label("OPENSSL_ia32cap_P");
+
+&function_begin_B("OPENSSL_rdtsc","EXTRN\t_OPENSSL_ia32cap_P:DWORD");
+ &xor ("eax","eax");
+ &xor ("edx","edx");
+ &picmeup("ecx","OPENSSL_ia32cap_P");
+ &bt (&DWP(0,"ecx"),4);
+ &jnc (&label("notsc"));
+ &rdtsc ();
+&set_label("notsc");
+ &ret ();
+&function_end_B("OPENSSL_rdtsc");
+
+# This works in Ring 0 only [read DJGPP+MS-DOS+privileged DPMI host],
+# but it's safe to call it on any [supported] 32-bit platform...
+# Just check for [non-]zero return value...
+&function_begin_B("OPENSSL_instrument_halt","EXTRN\t_OPENSSL_ia32cap_P:DWORD");
+ &picmeup("ecx","OPENSSL_ia32cap_P");
+ &bt (&DWP(0,"ecx"),4);
+ &jnc (&label("nohalt")); # no TSC
+
+ &data_word(0x9058900e); # push %cs; pop %eax
+ &and ("eax",3);
+ &jnz (&label("nohalt")); # not enough privileges
+
+ &pushf ();
+ &pop ("eax")
+ &bt ("eax",9);
+ &jnc (&label("nohalt")); # interrupts are disabled
+
+ &rdtsc ();
+ &push ("edx");
+ &push ("eax");
+ &halt ();
+ &rdtsc ();
+
+ &sub ("eax",&DWP(0,"esp"));
+ &sbb ("edx",&DWP(4,"esp"));
+ &add ("esp",8);
+ &ret ();
+
+&set_label("nohalt");
+ &xor ("eax","eax");
+ &xor ("edx","edx");
+ &ret ();
+&function_end_B("OPENSSL_instrument_halt");
+
+# Essentially there is only one use for this function. Under DJGPP:
+#
+# #include <go32.h>
+# ...
+# i=OPENSSL_far_spin(_dos_ds,0x46c);
+# ...
+# to obtain the number of spins till closest timer interrupt.
+
+&function_begin_B("OPENSSL_far_spin");
+ &pushf ();
+ &pop ("eax")
+ &bt ("eax",9);
+ &jnc (&label("nospin")); # interrupts are disabled
+
+ &mov ("eax",&DWP(4,"esp"));
+ &mov ("ecx",&DWP(8,"esp"));
+ &data_word (0x90d88e1e); # push %ds, mov %eax,%ds
+ &xor ("eax","eax");
+ &mov ("edx",&DWP(0,"ecx"));
+ &jmp (&label("spin"));
+
+ &align (16);
+&set_label("spin");
+ &inc ("eax");
+ &cmp ("edx",&DWP(0,"ecx"));
+ &je (&label("spin"));
+
+ &data_word (0x1f909090); # pop %ds
+ &ret ();
+
+&set_label("nospin");
+ &xor ("eax","eax");
+ &xor ("edx","edx");
+ &ret ();
+&function_end_B("OPENSSL_far_spin");
+
+&function_begin_B("OPENSSL_wipe_cpu","EXTRN\t_OPENSSL_ia32cap_P:DWORD");
+ &xor ("eax","eax");
+ &xor ("edx","edx");
+ &picmeup("ecx","OPENSSL_ia32cap_P");
+ &mov ("ecx",&DWP(0,"ecx"));
+ &bt (&DWP(0,"ecx"),1);
+ &jnc (&label("no_x87"));
+ if ($sse2) {
+ &bt (&DWP(0,"ecx"),26);
+ &jnc (&label("no_sse2"));
+ &pxor ("xmm0","xmm0");
+ &pxor ("xmm1","xmm1");
+ &pxor ("xmm2","xmm2");
+ &pxor ("xmm3","xmm3");
+ &pxor ("xmm4","xmm4");
+ &pxor ("xmm5","xmm5");
+ &pxor ("xmm6","xmm6");
+ &pxor ("xmm7","xmm7");
+ &set_label("no_sse2");
+ }
+ # just a bunch of fldz to zap the fp/mm bank followed by finit...
+ &data_word(0xeed9eed9,0xeed9eed9,0xeed9eed9,0xeed9eed9,0x90e3db9b);
+&set_label("no_x87");
+ &lea ("eax",&DWP(4,"esp"));
+ &ret ();
+&function_end_B("OPENSSL_wipe_cpu");
+
+&function_begin_B("OPENSSL_atomic_add");
+ &mov ("edx",&DWP(4,"esp")); # fetch the pointer, 1st arg
+ &mov ("ecx",&DWP(8,"esp")); # fetch the increment, 2nd arg
+ &push ("ebx");
+ &nop ();
+ &mov ("eax",&DWP(0,"edx"));
+&set_label("spin");
+ &lea ("ebx",&DWP(0,"eax","ecx"));
+ &nop ();
+ &data_word(0x1ab10ff0); # lock; cmpxchg %ebx,(%edx) # %eax is envolved and is always reloaded
+ &jne (&label("spin"));
+ &mov ("eax","ebx"); # OpenSSL expects the new value
+ &pop ("ebx");
+ &ret ();
+&function_end_B("OPENSSL_atomic_add");
+
+# This function can become handy under Win32 in situations when
+# we don't know which calling convention, __stdcall or __cdecl(*),
+# indirect callee is using. In C it can be deployed as
+#
+#ifdef OPENSSL_CPUID_OBJ
+# type OPENSSL_indirect_call(void *f,...);
+# ...
+# OPENSSL_indirect_call(func,[up to $max arguments]);
+#endif
+#
+# (*) it's designed to work even for __fastcall if number of
+# arguments is 1 or 2!
+&function_begin_B("OPENSSL_indirect_call");
+ {
+ my $i,$max=7; # $max has to be chosen as 4*n-1
+ # in order to preserve eventual
+ # stack alignment
+ &push ("ebp");
+ &mov ("ebp","esp");
+ &sub ("esp",$max*4);
+ &mov ("ecx",&DWP(12,"ebp"));
+ &mov (&DWP(0,"esp"),"ecx");
+ &mov ("edx",&DWP(16,"ebp"));
+ &mov (&DWP(4,"esp"),"edx");
+ for($i=2;$i<$max;$i++)
+ {
+ # Some copies will be redundant/bogus...
+ &mov ("eax",&DWP(12+$i*4,"ebp"));
+ &mov (&DWP(0+$i*4,"esp"),"eax");
+ }
+ &call_ptr (&DWP(8,"ebp"));# make the call...
+ &mov ("esp","ebp"); # ... and just restore the stack pointer
+ # without paying attention to what we called,
+ # (__cdecl *func) or (__stdcall *one).
+ &pop ("ebp");
+ &ret ();
+ }
+&function_end_B("OPENSSL_indirect_call");
+
+&initseg("OPENSSL_cpuid_setup");
+
+&asm_finish();