Update .cvsignore

[openssl.git] / crypto / whrlpool / asm / wp-mmx.pl

diff --git a/crypto/whrlpool/asm/wp-mmx.pl b/crypto/whrlpool/asm/wp-mmx.pl
index 7f49c778e80a10badfe8f4aff04f0bf4de35d643..d52439bca4b805f668134731e82411b4e19a2bb0 100644 (file)
--- a/crypto/whrlpool/asm/wp-mmx.pl
+++ b/crypto/whrlpool/asm/wp-mmx.pl
@@ -31,7 +31,7 @@
 #
 # (*)  I've sketched even non-MMX assembler, but for the record
 #      I've failed to beat the Intel compiler on P4, without using
 #
 # (*)  I've sketched even non-MMX assembler, but for the record
 #      I've failed to beat the Intel compiler on P4, without using

-*      MMX that is...
+#      MMX that is...

 # (**) ... on AMD on the other hand non-MMX assembler was observed
 #      to perform significantly better, but I figured this MMX
 #      implementation is even faster anyway, so why bother? As for
 # (**) ... on AMD on the other hand non-MMX assembler was observed
 #      to perform significantly better, but I figured this MMX
 #      implementation is even faster anyway, so why bother? As for


@@ -45,7 +45,7 @@
 #      non-MMX implementation would actually pay off, but till
 #      opposite is proved "unlikely" is assumed.
 
 #      non-MMX implementation would actually pay off, but till
 #      opposite is proved "unlikely" is assumed.
 

-push(@INC,"../CVS/HEAD/openssl/crypto/perlasm","../../perlasm");
+push(@INC,"../../perlasm");

 require "x86asm.pl";
 
 &asm_init($ARGV[0],"wp-mmx.pl");
 require "x86asm.pl";
 
 &asm_init($ARGV[0],"wp-mmx.pl");


@@ -99,22 +99,22 @@ $tbl="ebp";
        &call   (&label("pic_point"));
 &set_label("pic_point");
        &blindpop($tbl);
        &call   (&label("pic_point"));
 &set_label("pic_point");
        &blindpop($tbl);

-        &lea    ($tbl,&DWP(&label("table")."-".&label("pic_point"),$tbl));
+       &lea    ($tbl,&DWP(&label("table")."-".&label("pic_point"),$tbl));

 
        &xor    ("ecx","ecx");
        &xor    ("edx","edx");
 
        for($i=0;$i<8;$i++) { &movq(@mm[$i],&QWP($i*8,"esi")); }    # L=H
 &set_label("outerloop");
 
        &xor    ("ecx","ecx");
        &xor    ("edx","edx");
 
        for($i=0;$i<8;$i++) { &movq(@mm[$i],&QWP($i*8,"esi")); }    # L=H
 &set_label("outerloop");

-       for($i=0;$i<8;$i++) { &movq(&QWP($i*8,"esp"),@mm[$i]); }    # K=H
+       for($i=0;$i<8;$i++) { &movq(&QWP($i*8,"esp"),@mm[$i]); }    # K=L

        for($i=0;$i<8;$i++) { &pxor(@mm[$i],&QWP($i*8,"edi")); }    # L^=inp
        for($i=0;$i<8;$i++) { &pxor(@mm[$i],&QWP($i*8,"edi")); }    # L^=inp

-       for($i=0;$i<8;$i++) { &movq(&QWP(64+$i*8,"esp"),@mm[$i]); } #S=L
+       for($i=0;$i<8;$i++) { &movq(&QWP(64+$i*8,"esp"),@mm[$i]); } # S=L

 
        &xor    ("esi","esi");
        &mov    (&DWP(12,"ebx"),"esi");         # zero round counter
 
 &set_label("round",16);
 
        &xor    ("esi","esi");
        &mov    (&DWP(12,"ebx"),"esi");         # zero round counter
 
 &set_label("round",16);

-       &movq   (@mm[0],&DWP(2048*$SCALE,$tbl,"esi",8));        # rc[r]
+       &movq   (@mm[0],&QWP(2048*$SCALE,$tbl,"esi",8));        # rc[r]

        &mov    ("eax",&DWP(0,"esp"));
        &mov    ("ebx",&DWP(4,"esp"));
 for($i=0;$i<8;$i++) {
        &mov    ("eax",&DWP(0,"esp"));
        &mov    ("ebx",&DWP(4,"esp"));
 for($i=0;$i<8;$i++) {


@@ -488,5 +488,5 @@ for($i=0;$i<8;$i++) {
        &L(0xfb,0xee,0x7c,0x66,0xdd,0x17,0x47,0x9e);
        &L(0xca,0x2d,0xbf,0x07,0xad,0x5a,0x83,0x33);
 
        &L(0xfb,0xee,0x7c,0x66,0xdd,0x17,0x47,0x9e);
        &L(0xca,0x2d,0xbf,0x07,0xad,0x5a,0x83,0x33);
 

-&function_end_B("whrilpool_block_mmx");
+&function_end_B("whirlpool_block_mmx");

 &asm_finish(); 
 &asm_finish();