SEED to support OPENSSL_SMALL_FOOTPRINT: ~2x size decrease on x86.

[openssl.git] / crypto / x86cpuid.pl

diff --git a/crypto/x86cpuid.pl b/crypto/x86cpuid.pl
index 7d924a60b77d78c9352c50bbe89795332a78ee56..c3291112032b3ff7b365830ecf123f3a4efc4df8 100644 (file)
--- a/crypto/x86cpuid.pl
+++ b/crypto/x86cpuid.pl
@@ -1,6 +1,7 @@
 #!/usr/bin/env perl
 
-push(@INC,"perlasm");
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+push(@INC, "${dir}perlasm", "perlasm");
 require "x86asm.pl";
 
 &asm_init($ARGV[0],"x86cpuid");
@@ -19,7 +20,7 @@ for (@ARGV) { $sse2=1 if (/-DOPENSSL_IA32_SSE2/); }
        &pop    ("eax");
        &xor    ("ecx","eax");
        &bt     ("ecx",21);
-       &jnc    (&label("nocpuid"));
+       &jnc    (&label("done"));
        &xor    ("eax","eax");
        &cpuid  ();
        &xor    ("eax","eax");
@@ -34,17 +35,20 @@ for (@ARGV) { $sse2=1 if (/-DOPENSSL_IA32_SSE2/); }
        &or     ("ebp","eax");
        &mov    ("eax",1);
        &cpuid  ();
-       &bt     ("edx",28);             # test hyper-threading bit
-       &jnc    (&label("nocpuid"));
        &cmp    ("ebp",0);
-       &jne    (&label("notintel"));
+       &jne    (&label("notP4"));
+       &and    (&HB("eax"),15);        # familiy ID
+       &cmp    (&HB("eax"),15);        # P4?
+       &jne    (&label("notP4"));
        &or     ("edx",1<<20);          # use reserved bit to engage RC4_CHAR
-&set_label("notintel");
+&set_label("notP4");
+       &bt     ("edx",28);             # test hyper-threading bit
+       &jnc    (&label("done"));
        &shr    ("ebx",16);
        &cmp    (&LB("ebx"),1);         # see if cache is shared(*)
-       &ja     (&label("nocpuid"));
-       &and    ("edx",~(1<<28));       # clear hyper-threading bit if not
-&set_label("nocpuid");
+       &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");
@@ -216,6 +220,37 @@ for (@ARGV) { $sse2=1 if (/-DOPENSSL_IA32_SSE2/); }
        }
 &function_end_B("OPENSSL_indirect_call");
 
+&function_begin_B("OPENSSL_cleanse");
+       &mov    ("edx",&wparam(0));
+       &mov    ("ecx",&wparam(1));
+       &xor    ("eax","eax");
+       &cmp    ("ecx",7);
+       &jae    (&label("lot"));
+&set_label("little");
+       &mov    (&BP(0,"edx"),"al");
+       &sub    ("ecx",1);
+       &lea    ("edx",&DWP(1,"edx"));
+       &jnz    (&label("little"));
+       &ret    ();
+
+&set_label("lot",16);
+       &test   ("edx",3);
+       &jz     (&label("aligned"));
+       &mov    (&BP(0,"edx"),"al");
+       &lea    ("ecx",&DWP(-1,"ecx"));
+       &lea    ("edx",&DWP(1,"edx"));
+       &jmp    (&label("lot"));
+&set_label("aligned");
+       &mov    (&DWP(0,"edx"),"eax");
+       &lea    ("ecx",&DWP(-4,"ecx"));
+       &test   ("ecx",-4);
+       &lea    ("edx",&DWP(4,"edx"));
+       &jnz    (&label("aligned"));
+       &cmp    ("ecx",0);
+       &jne    (&label("little"));
+       &ret    ();
+&function_end_B("OPENSSL_cleanse");
+
 &initseg("OPENSSL_cpuid_setup");
 
 &asm_finish();