sha[256|512]-586.pl: add more SIMD code paths.
authorAndy Polyakov <appro@openssl.org>
Mon, 10 Jun 2013 20:26:53 +0000 (22:26 +0200)
committerAndy Polyakov <appro@openssl.org>
Mon, 10 Jun 2013 20:26:53 +0000 (22:26 +0200)
crypto/sha/asm/sha256-586.pl
crypto/sha/asm/sha512-586.pl

index 447c50b..bd48b63 100644 (file)
 # May version, >60% over original. Add AVX+shrd code path, >25%
 # improvement on Sandy Bridge over May version, 60% over original.
 #
+# May 2013.
+#
+# Replace AMD XOP code path with SSSE3 to cover more processors.
+# (Biggest improvement coefficient is on upcoming Atom Silvermont,
+# not shown.) Add AVX+BMI code path.
+#
 # Performance in clock cycles per processed byte (less is better):
 #
-#              PIII    P4      AMD K8  Core2   SB      Atom    Bldzr
-# gcc          36      41      27      26      25      50      36
-# icc          33      38      25      23      -       -       -
-# x86 asm(*)   27/24   28      19/15.5 18/15.6 12.3    30/25   16.6
-# x86_64 asm(**)       17.5    15.1    13.9    11.6    22      13.7
+#              gcc     icc     x86 asm(*)      SIMD    x86_64 asm(**)  
+# Pentium      46      57      40/38           -       -
+# PIII         36      33      27/24           -       -
+# P4           41      38      28              -       17.3
+# AMD K8       27      25      19/15.5         -       14.9
+# Core2                26      23      18/15.6         14.3    13.8
+# Westmere     27      -       19/15.7         13.4    12.3
+# Sandy Bridge 25      -       15.9            12.4    11.6
+# Ivy Bridge   24      -       15.0            11.4    10.3
+# Haswell      22      -       13.9            9.46    7.80
+# Bulldozer    36      -       27/22           17.0    13.6
+# VIA Nano     36      -       25/22           16.8    16.5
+# Atom         50      -       30/25           21.9    18.9
 #
-# (*)  numbers after slash are for unrolled loop, where available,
-#      otherwise best applicable such as AVX/XOP;
+# (*)  numbers after slash are for unrolled loop, where applicable;
 # (**) x86_64 assembly performance is presented for reference
-#      purposes.
+#      purposes, results are best-available;
 
 $0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
 push(@INC,"${dir}","${dir}../../perlasm");
@@ -47,21 +60,23 @@ require "x86asm.pl";
 
 &asm_init($ARGV[0],"sha512-586.pl",$ARGV[$#ARGV] eq "386");
 
-$xmm=$ymm=0;
+$xmm=$avx=0;
 for (@ARGV) { $xmm=1 if (/-DOPENSSL_IA32_SSE2/); }
 
-$ymm=1 if ($xmm &&
-               `$ENV{CC} -Wa,-v -c -o /dev/null -x assembler /dev/null 2>&1`
-                       =~ /GNU assembler version ([2-9]\.[0-9]+)/ &&
-               $1>=2.19);      # first version supporting AVX
+if ($xmm &&    `$ENV{CC} -Wa,-v -c -o /dev/null -x assembler /dev/null 2>&1`
+                       =~ /GNU assembler version ([2-9]\.[0-9]+)/) {
+       $avx = ($1>=2.19) + ($1>=2.22);
+}
 
-$ymm=1 if ($xmm && !$ymm && $ARGV[0] eq "win32n" &&
-               `nasm -v 2>&1` =~ /NASM version ([2-9]\.[0-9]+)/ &&
-               $1>=2.03);      # first version supporting AVX
+if ($xmm && !$avx && $ARGV[0] eq "win32n" &&
+               `nasm -v 2>&1` =~ /NASM version ([2-9]\.[0-9]+)/) {
+       $avx = ($1>=2.03) + ($1>=2.10);
+}
 
-$ymm=1 if ($xmm && !$ymm && $ARGV[0] eq "win32" &&
-               `ml 2>&1` =~ /Version ([0-9]+)\./ &&
-               $1>=10);        # first version supporting AVX
+if ($xmm && !$avx && $ARGV[0] eq "win32" &&
+               `ml 2>&1` =~ /Version ([0-9]+)\./) {
+       $avx = ($1>=10) + ($1>=11);
+}
 
 $unroll_after = 64*4;  # If pre-evicted from L1P cache first spin of
                        # fully unrolled loop was measured to run about
@@ -178,17 +193,21 @@ sub BODY_00_15() {
                                                if (!$i386) {
        &picmeup("edx","OPENSSL_ia32cap_P",$K256,&label("K256"));
        &mov    ("ecx",&DWP(0,"edx"));
-       &mov    ("edx",&DWP(4,"edx"));
+       &mov    ("ebx",&DWP(4,"edx"));
        &test   ("ecx",1<<20);          # check for P4
        &jnz    (&label("loop"));
-       &test   ("edx",1<<11);          # check for XOP
-       &jnz    (&label("XOP"))         if ($ymm);
        &and    ("ecx",1<<30);          # mask "Intel CPU" bit
-       &and    ("edx",1<<28);          # mask AVX bit
-       &or     ("ecx","edx");
+       &and    ("ebx",1<<28|1<<9);     # mask AVX and SSSE3 bits
+       &or     ("ecx","ebx");
+       &and    ("ecx",1<<28|1<<30);
        &cmp    ("ecx",1<<28|1<<30);
-       &je     (&label("AVX"))         if ($ymm);
-       &je     (&label("loop_shrd"))   if (!$ymm);
+                                       if ($xmm) {
+       &je     (&label("AVX"))         if ($avx);
+       &test   ("ebx",1<<9);           # check for SSSE3
+       &jnz    (&label("SSSE3"));
+                                       } else {
+       &je     (&label("loop_shrd"));
+                                       }
                                                if ($unroll_after) {
        &sub    ("eax","edi");
        &cmp    ("eax",$unroll_after);
@@ -199,7 +218,7 @@ sub BODY_00_15() {
 sub COMPACT_LOOP() {
 my $suffix=shift;
 
-&set_label("loop$suffix",16);
+&set_label("loop$suffix",$suffix?32:16);
     # copy input block to stack reversing byte and dword order
     for($i=0;$i<4;$i++) {
        &mov    ("eax",&DWP($i*16+0,"edi"));
@@ -292,7 +311,7 @@ my $suffix=shift;
        &COMPACT_LOOP();
        &mov    ("esp",&DWP(12,"esp"));         # restore sp
 &function_end_A();
-                                               if (!$i386 && !$ymm) {
+                                               if (!$i386 && !$xmm) {
        # ~20% improvement on Sandy Bridge
        local *ror = sub { &shrd(@_[0],@_) };
        &COMPACT_LOOP("_shrd");
@@ -318,7 +337,11 @@ my $suffix=shift;
        0x748f82ee,0x78a5636f,0x84c87814,0x8cc70208,
        0x90befffa,0xa4506ceb,0xbef9a3f7,0xc67178f2     );
 &data_word(@K256);
-&data_word(0x00010203,0x04050607,0x08090a0b,0x0c0d0e0f);
+&data_word(0x00010203,0x04050607,0x08090a0b,0x0c0d0e0f);       # byte swap mask
+&asciz("SHA256 block transform for x86, CRYPTOGAMS by <appro\@openssl.org>");
+
+($a,$b,$c,$d,$e,$f,$g,$h)=(0..7);      # offsets
+sub off { &DWP(4*(((shift)-$i)&7),"esp"); }
 
 if (!$i386 && $unroll_after) {
 my @AH=($A,$K256);
@@ -365,8 +388,6 @@ my @AH=($A,$K256);
        &mov    (&DWP(32+12*$i,"esp"),"ebx");
 
     my ($t1,$t2) = ("ecx","esi");
-    my ($a,$b,$c,$d,$e,$f,$g,$h)=(0..7);       # offsets
-    sub off { &DWP(4*(((shift)-$i)&7),"esp"); }
 
     for ($i=0;$i<64;$i++) {
 
@@ -472,15 +493,14 @@ my @AH=($A,$K256);
 
        &mov    ("esp",&DWP(96+12,"esp"));      # restore sp
 &function_end_A();
-
-                                               if ($ymm) {{{
+}
+                                               if (!$i386 && $xmm) {{{
 my @X = map("xmm$_",(0..3));
 my ($t0,$t1,$t2,$t3) = map("xmm$_",(4..7));
 my @AH = ($A,$T);
 
-&set_label("XOP",16);
+&set_label("SSSE3",32);
        &lea    ("esp",&DWP(-96,"esp"));
-       &vzeroall       ();
        # copy ctx->h[0-7] to A,B,C,D,E,F,G,H on stack
        &mov    ($AH[0],&DWP(0,"esi"));
        &mov    ($AH[1],&DWP(4,"esi"));
@@ -500,135 +520,208 @@ my @AH = ($A,$T);
        &mov    ("edi",&DWP(96+4,"esp"));       # inp
        &mov    (&DWP(24,"esp"),"ecx");
        &mov    (&DWP(28,"esp"),"esi");
-       &vmovdqa        ($t3,&QWP(256,$K256));
-       &jmp    (&label("grand_xop"));
+       &movdqa ($t3,&QWP(256,$K256));
+       &jmp    (&label("grand_ssse3"));
 
-&set_label("grand_xop",16);
+&set_label("grand_ssse3",16);
        # load input, reverse byte order, add K256[0..15], save to stack
-       &vmovdqu        (@X[0],&QWP(0,"edi"));
-       &vmovdqu        (@X[1],&QWP(16,"edi"));
-       &vmovdqu        (@X[2],&QWP(32,"edi"));
-       &vmovdqu        (@X[3],&QWP(48,"edi"));
-       &add            ("edi",64);
-       &vpshufb        (@X[0],@X[0],$t3);
-       &mov            (&DWP(96+4,"esp"),"edi");
-       &vpshufb        (@X[1],@X[1],$t3);
-       &vpshufb        (@X[2],@X[2],$t3);
-       &vpaddd         ($t0,@X[0],&QWP(0,$K256));
-       &vpshufb        (@X[3],@X[3],$t3);
-       &vpaddd         ($t1,@X[1],&QWP(16,$K256));
-       &vpaddd         ($t2,@X[2],&QWP(32,$K256));
-       &vpaddd         ($t3,@X[3],&QWP(48,$K256));
-       &vmovdqa        (&QWP(32+0,"esp"),$t0);
-       &vmovdqa        (&QWP(32+16,"esp"),$t1);
-       &vmovdqa        (&QWP(32+32,"esp"),$t2);
-       &vmovdqa        (&QWP(32+48,"esp"),$t3);
-       &jmp            (&label("xop_00_47"));
-
-&set_label("xop_00_47",16);
+       &movdqu (@X[0],&QWP(0,"edi"));
+       &movdqu (@X[1],&QWP(16,"edi"));
+       &movdqu (@X[2],&QWP(32,"edi"));
+       &movdqu (@X[3],&QWP(48,"edi"));
+       &add    ("edi",64);
+       &pshufb (@X[0],$t3);
+       &mov    (&DWP(96+4,"esp"),"edi");
+       &pshufb (@X[1],$t3);
+       &movdqa ($t0,&QWP(0,$K256));
+       &pshufb (@X[2],$t3);
+       &movdqa ($t1,&QWP(16,$K256));
+       &paddd  ($t0,@X[0]);
+       &pshufb (@X[3],$t3);
+       &movdqa ($t2,&QWP(32,$K256));
+       &paddd  ($t1,@X[1]);
+       &movdqa ($t3,&QWP(48,$K256));
+       &movdqa (&QWP(32+0,"esp"),$t0);
+       &paddd  ($t2,@X[2]);
+       &movdqa (&QWP(32+16,"esp"),$t1);
+       &paddd  ($t3,@X[3]);
+       &movdqa (&QWP(32+32,"esp"),$t2);
+       &movdqa (&QWP(32+48,"esp"),$t3);
+       &jmp    (&label("ssse3_00_47"));
+
+&set_label("ssse3_00_47",16);
        &add            ($K256,64);
 
-sub XOP_00_47 () {
+sub SSSE3_00_47 () {
 my $j = shift;
 my $body = shift;
 my @X = @_;
 my @insns = (&$body,&$body,&$body,&$body);     # 120 instructions
 
-       &vpalignr       ($t0,@X[1],@X[0],4);    # X[1..4]
          eval(shift(@insns));
+       &movdqa         ($t0,@X[1]);
+         eval(shift(@insns));                  # @
          eval(shift(@insns));
-        &vpalignr      ($t3,@X[3],@X[2],4);    # X[9..12]
+       &movdqa         ($t3,@X[3]);
          eval(shift(@insns));
          eval(shift(@insns));
-       &vprotd         ($t1,$t0,14);
+       &palignr        ($t0,@X[0],4);          # X[1..4]
          eval(shift(@insns));
+         eval(shift(@insns));                  # @
          eval(shift(@insns));
-       &vpsrld         ($t0,$t0,3);
-        &vpaddd        (@X[0],@X[0],$t3);      # X[0..3] += X[9..12]
+        &palignr       ($t3,@X[2],4);          # X[9..12]
          eval(shift(@insns));
          eval(shift(@insns));
          eval(shift(@insns));
+       &movdqa         ($t1,$t0);
+         eval(shift(@insns));                  # @
          eval(shift(@insns));
-       &vprotd         ($t2,$t1,25-14);
-       &vpxor          ($t0,$t0,$t1);
+       &movdqa         ($t2,$t0);
          eval(shift(@insns));
          eval(shift(@insns));
+       &psrld          ($t0,3);
          eval(shift(@insns));
+         eval(shift(@insns));                  # @
+        &paddd         (@X[0],$t3);            # X[0..3] += X[9..12]
          eval(shift(@insns));
-        &vprotd        ($t3,@X[3],13);
-       &vpxor          ($t0,$t0,$t2);          # sigma0(X[1..4])
          eval(shift(@insns));
+       &psrld          ($t2,7);
          eval(shift(@insns));
-        &vpsrld        ($t2,@X[3],10);
          eval(shift(@insns));
+         eval(shift(@insns));                  # @
          eval(shift(@insns));
-       &vpaddd         (@X[0],@X[0],$t0);      # X[0..3] += sigma0(X[1..4])
+        &pshufd        ($t3,@X[3],0b11111010); # X[14..15]
          eval(shift(@insns));
          eval(shift(@insns));
-        &vprotd        ($t1,$t3,15-13);
-        &vpxor         ($t3,$t3,$t2);
+       &pslld          ($t1,32-18);
          eval(shift(@insns));
+         eval(shift(@insns));                  # @
+       &pxor           ($t0,$t2);
          eval(shift(@insns));
          eval(shift(@insns));
+       &psrld          ($t2,18-7);
          eval(shift(@insns));
-        &vpxor         ($t3,$t3,$t1);          # sigma1(X[14..15])
          eval(shift(@insns));
+         eval(shift(@insns));                  # @
+       &pxor           ($t0,$t1);
          eval(shift(@insns));
          eval(shift(@insns));
+       &pslld          ($t1,18-7);
          eval(shift(@insns));
-       &vpsrldq        ($t3,$t3,8);
          eval(shift(@insns));
+         eval(shift(@insns));                  # @
+       &pxor           ($t0,$t2);
          eval(shift(@insns));
          eval(shift(@insns));
+        &movdqa        ($t2,$t3);
          eval(shift(@insns));
-       &vpaddd         (@X[0],@X[0],$t3);      # X[0..1] += sigma1(X[14..15])
          eval(shift(@insns));
+         eval(shift(@insns));                  # @
+       &pxor           ($t0,$t1);              # sigma0(X[1..4])
          eval(shift(@insns));
          eval(shift(@insns));
+        &psrld         ($t3,10);
          eval(shift(@insns));
-        &vprotd        ($t3,@X[0],13);
          eval(shift(@insns));
+         eval(shift(@insns));                  # @
+       &paddd          (@X[0],$t0);            # X[0..3] += sigma0(X[1..4])
          eval(shift(@insns));
-        &vpsrld        ($t2,@X[0],10);
          eval(shift(@insns));
+        &psrlq         ($t2,17);
          eval(shift(@insns));
-        &vprotd        ($t1,$t3,15-13);
          eval(shift(@insns));
+         eval(shift(@insns));                  # @
+        &pxor          ($t3,$t2);
          eval(shift(@insns));
-        &vpxor         ($t3,$t3,$t2);
          eval(shift(@insns));
+        &psrlq         ($t2,19-17);
          eval(shift(@insns));
          eval(shift(@insns));
+         eval(shift(@insns));                  # @
+        &pxor          ($t3,$t2);
          eval(shift(@insns));
-        &vpxor         ($t3,$t3,$t1);          # sigma1(X[16..17])
          eval(shift(@insns));
+        &pshufd        ($t3,$t3,0b10000000);
          eval(shift(@insns));
          eval(shift(@insns));
+         eval(shift(@insns));                  # @
          eval(shift(@insns));
-       &vpslldq        ($t3,$t3,8);            # 22 instructions
          eval(shift(@insns));
          eval(shift(@insns));
          eval(shift(@insns));
+         eval(shift(@insns));                  # @
          eval(shift(@insns));
-       &vpaddd         (@X[0],@X[0],$t3);      # X[2..3] += sigma1(X[16..17])
+        &psrldq        ($t3,8);
          eval(shift(@insns));
          eval(shift(@insns));
          eval(shift(@insns));
+       &paddd          (@X[0],$t3);            # X[0..1] += sigma1(X[14..15])
+         eval(shift(@insns));                  # @
          eval(shift(@insns));
-       &vpaddd         ($t2,@X[0],&QWP(16*$j,$K256));
+         eval(shift(@insns));
+         eval(shift(@insns));
+         eval(shift(@insns));
+         eval(shift(@insns));                  # @
+         eval(shift(@insns));
+        &pshufd        ($t3,@X[0],0b01010000); # X[16..17]
+         eval(shift(@insns));
+         eval(shift(@insns));
+         eval(shift(@insns));
+        &movdqa        ($t2,$t3);
+         eval(shift(@insns));                  # @
+        &psrld         ($t3,10);
+         eval(shift(@insns));
+        &psrlq         ($t2,17);
+         eval(shift(@insns));
+         eval(shift(@insns));
+         eval(shift(@insns));
+         eval(shift(@insns));                  # @
+        &pxor          ($t3,$t2);
+         eval(shift(@insns));
+         eval(shift(@insns));
+        &psrlq         ($t2,19-17);
+         eval(shift(@insns));
+         eval(shift(@insns));
+         eval(shift(@insns));                  # @
+        &pxor          ($t3,$t2);
+         eval(shift(@insns));
+         eval(shift(@insns));
+         eval(shift(@insns));
+        &pshufd        ($t3,$t3,0b00001000);
+         eval(shift(@insns));
+         eval(shift(@insns));                  # @
+       &movdqa         ($t2,&QWP(16*$j,$K256));
+         eval(shift(@insns));
+         eval(shift(@insns));
+        &pslldq        ($t3,8);
+         eval(shift(@insns));
+         eval(shift(@insns));
+         eval(shift(@insns));                  # @
+         eval(shift(@insns));
+         eval(shift(@insns));
+         eval(shift(@insns));
+         eval(shift(@insns));
+         eval(shift(@insns));                  # @
+       &paddd          (@X[0],$t3);            # X[2..3] += sigma1(X[16..17])
+         eval(shift(@insns));
+         eval(shift(@insns));
+         eval(shift(@insns));
+         eval(shift(@insns));
+       &paddd          ($t2,@X[0]);
+         eval(shift(@insns));                  # @
 
        foreach (@insns) { eval; }              # remaining instructions
 
-       &vmovdqa        (&QWP(32+16*$j,"esp"),$t2);
+       &movdqa         (&QWP(32+16*$j,"esp"),$t2);
 }
 
 sub body_00_15 () {
        (
        '&mov   ("ecx",$E);',
-        '&mov  ("esi",&off($f));',
        '&ror   ($E,25-11);',
-        '&mov  ("edi",&off($g));',
+        '&mov  ("esi",&off($f));',
        '&xor   ($E,"ecx");',
+        '&mov  ("edi",&off($g));',
         '&xor  ("esi","edi");',
        '&ror   ($E,11-6);',
         '&and  ("esi","ecx");',
@@ -637,19 +730,19 @@ sub body_00_15 () {
         '&xor  ("edi","esi");',        # Ch(e,f,g)
        '&ror   ($E,6);',               # T = Sigma1(e)
         '&mov  ("ecx",$AH[0]);',
+        '&add  ($E,"edi");',           # T += Ch(e,f,g)
+        '&mov  ("edi",&off($b));',
        '&mov   ("esi",$AH[0]);',
-        '&add  ($E,&off($h));',        # T += h
 
        '&ror   ("ecx",22-13);',
-        '&add  ($E,"edi");',           # T += Ch(e,f,g)
-        '&mov  ("edi",&off($b));',
-       '&xor   ("ecx",$AH[0]);',
         '&mov  (&off($a),$AH[0]);',    # save $A, modulo-scheduled
+       '&xor   ("ecx",$AH[0]);',
         '&xor  ($AH[0],"edi");',       # a ^= b, (b^c) in next round
+        '&add  ($E,&off($h));',        # T += h
        '&ror   ("ecx",13-2);',
         '&and  ($AH[1],$AH[0]);',      # (b^c) &= (a^b)
-        '&add  ($E,&DWP(32+4*($i&15),"esp"));',        # T += K[i]+X[i]
        '&xor   ("ecx","esi");',
+        '&add  ($E,&DWP(32+4*($i&15),"esp"));',        # T += K[i]+X[i]
         '&xor  ($AH[1],"edi");',       # h = Maj(a,b,c) = Ch(a^b,c,b)
        '&ror   ("ecx",2);',            # Sigma0(a)
 
@@ -662,11 +755,11 @@ sub body_00_15 () {
 }
 
     for ($i=0,$j=0; $j<4; $j++) {
-       &XOP_00_47($j,\&body_00_15,@X);
+       &SSSE3_00_47($j,\&body_00_15,@X);
        push(@X,shift(@X));             # rotate(@X)
     }
        &cmp    (&DWP(16*$j,$K256),0x00010203);
-       &jne    (&label("xop_00_47"));
+       &jne    (&label("ssse3_00_47"));
 
     for ($i=0; $i<16; ) {
        foreach(body_00_15()) { eval; }
@@ -708,16 +801,21 @@ sub body_00_15 () {
         &mov   (&DWP(28,"esp"),"edi");
        &mov    ("edi",&DWP(96+4,"esp"));       # inp
 
-       &vmovdqa        ($t3,&QWP(64,$K256));
+       &movdqa ($t3,&QWP(64,$K256));
        &sub    ($K256,3*64);                   # rewind K
        &cmp    ("edi",&DWP(96+8,"esp"));       # are we done yet?
-       &jb     (&label("grand_xop"));
+       &jb     (&label("grand_ssse3"));
 
        &mov    ("esp",&DWP(96+12,"esp"));      # restore sp
-       &vzeroall       ();
 &function_end_A();
-
-&set_label("AVX",16);
+                                               if ($avx) {
+&set_label("AVX",32);
+                                               if ($avx>1) {
+       &mov    ("edx",&DWP(8,"edx"));
+       &and    ("edx",1<<8|1<<3);              # check for BMI2+BMI1
+       &cmp    ("edx",1<<8|1<<3);
+       &je     (&label("AVX_BMI"));
+                                               }
        &lea    ("esp",&DWP(-96,"esp"));
        &vzeroall       ();
        # copy ctx->h[0-7] to A,B,C,D,E,F,G,H on stack
@@ -742,7 +840,7 @@ sub body_00_15 () {
        &vmovdqa        ($t3,&QWP(256,$K256));
        &jmp    (&label("grand_avx"));
 
-&set_label("grand_avx",16);
+&set_label("grand_avx",32);
        # load input, reverse byte order, add K256[0..15], save to stack
        &vmovdqu        (@X[0],&QWP(0,"edi"));
        &vmovdqu        (@X[1],&QWP(16,"edi"));
@@ -809,12 +907,14 @@ my $j = shift;
 my $body = shift;
 my @X = @_;
 my @insns = (&$body,&$body,&$body,&$body);     # 120 instructions
+my $insn;
 
        foreach (Xupdate_AVX()) {               # 31 instructions
            eval;
            eval(shift(@insns));
            eval(shift(@insns));
-           eval(shift(@insns));
+           eval($insn = shift(@insns));
+           eval(shift(@insns)) if ($insn =~ /rorx/ && @insns[0] =~ /rorx/);
        }
        &vpaddd         ($t2,@X[0],&QWP(16*$j,$K256));
        foreach (@insns) { eval; }              # remaining instructions
@@ -876,9 +976,150 @@ my @insns = (&$body,&$body,&$body,&$body);        # 120 instructions
        &mov    ("esp",&DWP(96+12,"esp"));      # restore sp
        &vzeroall       ();
 &function_end_A();
-                                               }}}
+                                               if ($avx>1) {
+sub bodyx_00_15 () {                   # +10%
+       (
+       '&rorx  ("ecx",$E,6)',
+       '&rorx  ("esi",$E,11)',
+        '&mov  (&off($e),$E)',         # save $E, modulo-scheduled
+       '&rorx  ("edi",$E,25)',
+       '&xor   ("ecx","esi")',
+        '&andn ("esi",$E,&off($g))',
+       '&xor   ("ecx","edi")',         # Sigma1(e)
+        '&and  ($E,&off($f))',
+        '&mov  (&off($a),$AH[0]);',    # save $A, modulo-scheduled
+        '&or   ($E,"esi")',            # T = Ch(e,f,g)
+
+       '&rorx  ("edi",$AH[0],2)',
+       '&rorx  ("esi",$AH[0],13)',
+        '&lea  ($E,&DWP(0,$E,"ecx"))', # T += Sigma1(e)
+       '&rorx  ("ecx",$AH[0],22)',
+       '&xor   ("esi","edi")',
+        '&mov  ("edi",&off($b))',
+       '&xor   ("ecx","esi")',         # Sigma0(a)
+
+        '&xor  ($AH[0],"edi")',        # a ^= b, (b^c) in next round
+        '&add  ($E,&off($h))',         # T += h
+        '&and  ($AH[1],$AH[0])',       # (b^c) &= (a^b)
+        '&add  ($E,&DWP(32+4*($i&15),"esp"))', # T += K[i]+X[i]
+        '&xor  ($AH[1],"edi")',        # h = Maj(a,b,c) = Ch(a^b,c,b)
+
+        '&add  ("ecx",$E)',            # h += T
+        '&add  ($E,&off($d))',         # d += T
+       '&lea   ($AH[1],&DWP(0,$AH[1],"ecx"));'.        # h += Sigma0(a)
+
+       '@AH = reverse(@AH); $i++;'     # rotate(a,h)
+       );
 }
+
+&set_label("AVX_BMI",32);
+       &lea    ("esp",&DWP(-96,"esp"));
+       &vzeroall       ();
+       # copy ctx->h[0-7] to A,B,C,D,E,F,G,H on stack
+       &mov    ($AH[0],&DWP(0,"esi"));
+       &mov    ($AH[1],&DWP(4,"esi"));
+       &mov    ("ecx",&DWP(8,"esi"));
+       &mov    ("edi",&DWP(12,"esi"));
+       #&mov   (&DWP(0,"esp"),$AH[0]);
+       &mov    (&DWP(4,"esp"),$AH[1]);
+       &xor    ($AH[1],"ecx");                 # magic
+       &mov    (&DWP(8,"esp"),"ecx");
+       &mov    (&DWP(12,"esp"),"edi");
+       &mov    ($E,&DWP(16,"esi"));
+       &mov    ("edi",&DWP(20,"esi"));
+       &mov    ("ecx",&DWP(24,"esi"));
+       &mov    ("esi",&DWP(28,"esi"));
+       #&mov   (&DWP(16,"esp"),$E);
+       &mov    (&DWP(20,"esp"),"edi");
+       &mov    ("edi",&DWP(96+4,"esp"));       # inp
+       &mov    (&DWP(24,"esp"),"ecx");
+       &mov    (&DWP(28,"esp"),"esi");
+       &vmovdqa        ($t3,&QWP(256,$K256));
+       &jmp    (&label("grand_avx_bmi"));
+
+&set_label("grand_avx_bmi",32);
+       # load input, reverse byte order, add K256[0..15], save to stack
+       &vmovdqu        (@X[0],&QWP(0,"edi"));
+       &vmovdqu        (@X[1],&QWP(16,"edi"));
+       &vmovdqu        (@X[2],&QWP(32,"edi"));
+       &vmovdqu        (@X[3],&QWP(48,"edi"));
+       &add            ("edi",64);
+       &vpshufb        (@X[0],@X[0],$t3);
+       &mov            (&DWP(96+4,"esp"),"edi");
+       &vpshufb        (@X[1],@X[1],$t3);
+       &vpshufb        (@X[2],@X[2],$t3);
+       &vpaddd         ($t0,@X[0],&QWP(0,$K256));
+       &vpshufb        (@X[3],@X[3],$t3);
+       &vpaddd         ($t1,@X[1],&QWP(16,$K256));
+       &vpaddd         ($t2,@X[2],&QWP(32,$K256));
+       &vpaddd         ($t3,@X[3],&QWP(48,$K256));
+       &vmovdqa        (&QWP(32+0,"esp"),$t0);
+       &vmovdqa        (&QWP(32+16,"esp"),$t1);
+       &vmovdqa        (&QWP(32+32,"esp"),$t2);
+       &vmovdqa        (&QWP(32+48,"esp"),$t3);
+       &jmp            (&label("avx_bmi_00_47"));
+
+&set_label("avx_bmi_00_47",16);
+       &add            ($K256,64);
+
+    for ($i=0,$j=0; $j<4; $j++) {
+       &AVX_00_47($j,\&bodyx_00_15,@X);
+       push(@X,shift(@X));             # rotate(@X)
+    }
+       &cmp    (&DWP(16*$j,$K256),0x00010203);
+       &jne    (&label("avx_bmi_00_47"));
+
+    for ($i=0; $i<16; ) {
+       foreach(bodyx_00_15()) { eval; }
+    }
+
+       &mov    ("esi",&DWP(96,"esp")); #ctx
+                                       #&mov   ($AH[0],&DWP(0,"esp"));
+       &xor    ($AH[1],"edi");         #&mov   ($AH[1],&DWP(4,"esp"));
+                                       #&mov   ("edi", &DWP(8,"esp"));
+       &mov    ("ecx",&DWP(12,"esp"));
+       &add    ($AH[0],&DWP(0,"esi"));
+       &add    ($AH[1],&DWP(4,"esi"));
+       &add    ("edi",&DWP(8,"esi"));
+       &add    ("ecx",&DWP(12,"esi"));
+       &mov    (&DWP(0,"esi"),$AH[0]);
+       &mov    (&DWP(4,"esi"),$AH[1]);
+       &mov    (&DWP(8,"esi"),"edi");
+       &mov    (&DWP(12,"esi"),"ecx");
+        #&mov  (&DWP(0,"esp"),$AH[0]);
+        &mov   (&DWP(4,"esp"),$AH[1]);
+        &xor   ($AH[1],"edi");                 # magic
+        &mov   (&DWP(8,"esp"),"edi");
+        &mov   (&DWP(12,"esp"),"ecx");
+       #&mov   ($E,&DWP(16,"esp"));
+       &mov    ("edi",&DWP(20,"esp"));
+       &mov    ("ecx",&DWP(24,"esp"));
+       &add    ($E,&DWP(16,"esi"));
+       &add    ("edi",&DWP(20,"esi"));
+       &add    ("ecx",&DWP(24,"esi"));
+       &mov    (&DWP(16,"esi"),$E);
+       &mov    (&DWP(20,"esi"),"edi");
+        &mov   (&DWP(20,"esp"),"edi");
+       &mov    ("edi",&DWP(28,"esp"));
+       &mov    (&DWP(24,"esi"),"ecx");
+        #&mov  (&DWP(16,"esp"),$E);
+       &add    ("edi",&DWP(28,"esi"));
+        &mov   (&DWP(24,"esp"),"ecx");
+       &mov    (&DWP(28,"esi"),"edi");
+        &mov   (&DWP(28,"esp"),"edi");
+       &mov    ("edi",&DWP(96+4,"esp"));       # inp
+
+       &vmovdqa        ($t3,&QWP(64,$K256));
+       &sub    ($K256,3*64);                   # rewind K
+       &cmp    ("edi",&DWP(96+8,"esp"));       # are we done yet?
+       &jb     (&label("grand_avx_bmi"));
+
+       &mov    ("esp",&DWP(96+12,"esp"));      # restore sp
+       &vzeroall       ();
+&function_end_A();
+                                               }
+                                               }
+                                               }}}
 &function_end_B("sha256_block_data_order");
-&asciz("SHA256 block transform for x86, CRYPTOGAMS by <appro\@openssl.org>");
 
 &asm_finish();
index 0500b56..77a5c31 100644 (file)
@@ -9,18 +9,30 @@
 #
 # SHA512 block transform for x86. September 2007.
 #
+# May 2013.
+#
+# Add SSSE3 code path, 20-25% improvement [over original SSE2 code].
+#
 # Performance in clock cycles per processed byte (less is better):
 #
-#              PIII    P4      AMD K8  Core2   SB      Atom    Bldzr
-# gcc          75      116     54      66      58      126     121
-# icc          77      95      55      57      -       -       -
-# x86 asm      56      82      36      40      35      68      50
-# SSE2 asm     -       36.2    20.8    19.2    14.9    60(**)  17.1
-# x86_64 asm(*)        -       33      9.6     10.3    11.3    14.7    13.5
+#              gcc     icc     x86 asm SIMD(*) x86_64(**)
+# Pentium      100     97      61      -       -
+# PIII         75      77      56      -       -
+# P4           116     95      82      34.6    30.8
+# AMD K8       54      55      36      20.7    9.57
+# Core2                66      57      40      15.9    9.97
+# Westmere     70      -       38      12.2    9.58
+# Sandy Bridge 58      -       35      11.9    11.2
+# Ivy Bridge   50      -       33      11.5    8.17
+# Haswell      46      -       29      11.3    7.66
+# Bulldozer    121     -       50      14.0    13.5
+# VIA Nano     91      -       52      33      14.7
+# Atom         126     -       68      48(***) 14.7
 #
-# (*)  x86_64 assembler performance is presented for reference
-#      purposes.
-# (**) paddq is increadibly slow on Atom.
+# (*)  whichever best applicable.
+# (**) x86_64 assembler performance is presented for reference
+#      purposes, the results are for integer-only code.
+# (***)        paddq is increadibly slow on Atom.
 #
 # IALU code-path is optimized for elder Pentiums. On vanilla Pentium
 # performance improvement over compiler generated code reaches ~60%,
@@ -82,7 +94,7 @@ sub BODY_00_15_sse2 {
         &pand  ("mm5",$E);                     # f&=e
        &psllq  ($E,23);                        # $E is sliding left
         &movq  ($A,"mm3")                      if ($phase<2);
-        &movq  (&QWP(8*9,"esp"),"mm7")         if ($phase>1);  # save X[i]
+        &movq  (&QWP(8*9,"esp"),"mm7")         # save X[i]
        &movq   ("mm3","mm1");                  # %mm3 is T1
         &psrlq ("mm1",4);
         &pxor  ("mm5","mm6");                  # Ch(e,f,g)
@@ -100,11 +112,11 @@ sub BODY_00_15_sse2 {
        &pxor   ("mm3",$E);                     # T1=Sigma1_512(e)
 
         &movq  ($E,$Dsse2);                    # e = load d, e in next round
-        &movq  ("mm5",$A);                     # %mm5 is sliding right
        &paddq  ("mm3","mm7");                  # T1+=X[i]
+        &movq  ("mm5",$A);                     # %mm5 is sliding right
         &psrlq ("mm5",28);
-        &movq  ("mm6",$A);                     # %mm6 is sliding left
        &paddq  ($E,"mm3");                     # d += T1
+        &movq  ("mm6",$A);                     # %mm6 is sliding left
         &movq  ("mm7","mm5");
         &psllq ("mm6",25);
        &movq   ("mm1",$Bsse2);                 # load b
@@ -290,19 +302,28 @@ sub BODY_00_15_x86 {
 
 if ($sse2) {
        &picmeup("edx","OPENSSL_ia32cap_P",$K512,&label("K512"));
-       &bt     (&DWP(0,"edx"),26);
-       &jnc    (&label("loop_x86"));
+       &mov    ("ecx",&DWP(0,"edx"));
+       &test   ("ecx",1<<26);
+       &jz     (&label("loop_x86"));
+
+       &mov    ("edx",&DWP(4,"edx"));
 
        # load ctx->h[0-7]
        &movq   ($A,&QWP(0,"esi"));
+        &and   ("ecx",1<<24);          # XMM registers availability
        &movq   ("mm1",&QWP(8,"esi"));
+        &and   ("edx",1<<9);           # SSSE3 bit
        &movq   ($BxC,&QWP(16,"esi"));
+        &or    ("ecx","edx");
        &movq   ("mm3",&QWP(24,"esi"));
        &movq   ($E,&QWP(32,"esi"));
        &movq   ("mm5",&QWP(40,"esi"));
        &movq   ("mm6",&QWP(48,"esi"));
        &movq   ("mm7",&QWP(56,"esi"));
+       &cmp    ("ecx",1<<24|1<<9);
+       &je     (&label("SSSE3"));
        &sub    ("esp",8*10);
+       &jmp    (&label("loop_sse2"));
 
 &set_label("loop_sse2",16);
        #&movq  ($Asse2,$A);
@@ -316,32 +337,32 @@ if ($sse2) {
        &movq   ($Hsse2,"mm7");
        &movq   ("mm3",$A);                     # magic
 
-       &mov    ("ecx",&DWP(0,"edi"));
-       &mov    ("edx",&DWP(4,"edi"));
+       &mov    ("eax",&DWP(0,"edi"));
+       &mov    ("ebx",&DWP(4,"edi"));
        &add    ("edi",8);
-       &bswap  ("ecx");
-       &bswap  ("edx");
-       &mov    (&DWP(8*9+4,"esp"),"ecx");
-       &mov    (&DWP(8*9+0,"esp"),"edx");
        &mov    ("edx",15);                     # counter
+       &bswap  ("eax");
+       &bswap  ("ebx");
        &jmp    (&label("00_14_sse2"));
 
 &set_label("00_14_sse2",16);
+       &movd   ("mm1","eax");
        &mov    ("eax",&DWP(0,"edi"));
+       &movd   ("mm7","ebx");
        &mov    ("ebx",&DWP(4,"edi"));
        &add    ("edi",8);
        &bswap  ("eax");
-       &movq   ("mm7",&QWP(8*9,"esp"));        # X[i]
        &bswap  ("ebx");
-       &mov    (&DWP(8*8+4,"esp"),"eax");
-       &mov    (&DWP(8*8+0,"esp"),"ebx");
+       &punpckldq("mm7","mm1");
 
        &BODY_00_15_sse2();
 
        &dec    ("edx");
        &jnz    (&label("00_14_sse2"));
 
-       &movq   ("mm7",&QWP(8*9,"esp"));        # X[i]
+       &movd   ("mm1","eax");
+       &movd   ("mm7","ebx");
+       &punpckldq("mm7","mm1");
 
        &BODY_00_15_sse2(1);
 
@@ -429,6 +450,231 @@ if ($sse2) {
        &mov    ("esp",&DWP(8*10+12,"esp"));    # restore sp
        &emms   ();
 &function_end_A();
+
+&set_label("SSSE3",32);
+{ my ($cnt,$frame)=("ecx","edx");
+  my @X=map("xmm$_",(0..7));
+  my $j;
+  my $i=0;
+
+       &lea    ($frame,&DWP(-64,"esp"));
+       &sub    ("esp",256);
+
+       # fixed stack frame layout
+       #
+       # +0    A B C D E F G H         # backing store
+       # +64   X[0]+K[i] .. X[15]+K[i] # XMM->MM xfer area
+       # +192                          # XMM off-load ring buffer
+       # +256                          # saved parameters
+
+       &movdqa         (@X[1],&QWP(80*8,$K512));               # byte swap mask
+       &movdqu         (@X[0],&QWP(0,"edi"));
+       &pshufb         (@X[0],@X[1]);
+    for ($j=0;$j<8;$j++) {
+       &movdqa         (&QWP(16*(($j-1)%4),$frame),@X[3])      if ($j>4); # off-load
+       &movdqa         (@X[3],&QWP(16*($j%8),$K512));
+       &movdqa         (@X[2],@X[1])                           if ($j<7); # perpetuate byte swap mask
+       &movdqu         (@X[1],&QWP(16*($j+1),"edi"))           if ($j<7); # next input
+       &movdqa         (@X[1],&QWP(16*(($j+1)%4),$frame))      if ($j==7);# restore @X[0]
+       &paddq          (@X[3],@X[0]);
+       &pshufb         (@X[1],@X[2])                           if ($j<7);
+       &movdqa         (&QWP(16*($j%8)-128,$frame),@X[3]);     # xfer X[i]+K[i]
+
+       push(@X,shift(@X));                                     # rotate(@X)
+    }
+       #&jmp           (&label("loop_ssse3"));
+       &nop            ();
+
+&set_label("loop_ssse3",32);
+       &movdqa         (@X[2],&QWP(16*(($j+1)%4),$frame));     # pre-restore @X[1]
+       &movdqa         (&QWP(16*(($j-1)%4),$frame),@X[3]);     # off-load @X[3]
+       &lea            ($K512,&DWP(16*8,$K512));
+
+       #&movq  ($Asse2,$A);                    # off-load A-H
+       &movq   ($Bsse2,"mm1");
+        &mov   ("ebx","edi");
+       &movq   ($Csse2,$BxC);
+        &lea   ("edi",&DWP(128,"edi"));        # advance input
+       &movq   ($Dsse2,"mm3");
+        &cmp   ("edi","eax");
+       #&movq  ($Esse2,$E);
+       &movq   ($Fsse2,"mm5");
+        &cmovb ("ebx","edi");
+       &movq   ($Gsse2,"mm6");
+        &mov   ("ecx",4);                      # loop counter
+       &pxor   ($BxC,"mm1");                   # magic
+       &movq   ($Hsse2,"mm7");
+       &pxor   ("mm3","mm3");                  # magic
+
+       &jmp            (&label("00_47_ssse3"));
+
+sub BODY_00_15_ssse3 {         # "phase-less" copy of BODY_00_15_sse2
+       (
+       '&movq  ("mm1",$E)',                            # %mm1 is sliding right
+       '&movq  ("mm7",&QWP(((-8*$i)%128)-128,$frame))',# X[i]+K[i]
+        '&pxor ("mm5","mm6")',                         # f^=g
+       '&psrlq ("mm1",14)',
+        '&movq (&QWP(8*($i+4)%64,"esp"),$E)',          # modulo-scheduled save e
+        '&pand ("mm5",$E)',                            # f&=e
+       '&psllq ($E,23)',                               # $E is sliding left
+       '&paddq ($A,"mm3")',                            # [h+=Maj(a,b,c)]
+       '&movq  ("mm3","mm1")',                         # %mm3 is T1
+        '&psrlq("mm1",4)',
+        '&pxor ("mm5","mm6")',                         # Ch(e,f,g)
+       '&pxor  ("mm3",$E)',
+        '&psllq($E,23)',
+       '&pxor  ("mm3","mm1")',
+        '&movq (&QWP(8*$i%64,"esp"),$A)',              # modulo-scheduled save a
+        '&paddq("mm7","mm5")',                         # X[i]+=Ch(e,f,g)
+       '&pxor  ("mm3",$E)',
+        '&psrlq("mm1",23)',
+        '&paddq("mm7",&QWP(8*($i+7)%64,"esp"))',       # X[i]+=h
+       '&pxor  ("mm3","mm1")',
+        '&psllq($E,4)',
+       '&pxor  ("mm3",$E)',                            # T1=Sigma1_512(e)
+
+        '&movq ($E,&DWP(8*($i+3)%64,"esp"))',          # e = load d, e in next round
+       '&paddq ("mm3","mm7")',                         # T1+=X[i]
+        '&movq ("mm5",$A)',                            # %mm5 is sliding right
+        '&psrlq("mm5",28)',
+       '&paddq ($E,"mm3")',                            # d += T1
+        '&movq ("mm6",$A)',                            # %mm6 is sliding left
+        '&movq ("mm7","mm5")',
+        '&psllq("mm6",25)',
+       '&movq  ("mm1",&QWP(8*($i+1)%64,"esp"))',       # load b
+        '&psrlq("mm5",6)',
+        '&pxor ("mm7","mm6")',
+        '&psllq("mm6",5)',
+        '&pxor ("mm7","mm5")',
+       '&pxor  ($A,"mm1")',                            # a^b, b^c in next round
+        '&psrlq("mm5",5)',
+        '&pxor ("mm7","mm6")',
+       '&pand  ($BxC,$A)',                             # (b^c)&(a^b)
+        '&psllq("mm6",6)',
+        '&pxor ("mm7","mm5")',
+       '&pxor  ($BxC,"mm1")',                          # [h=]Maj(a,b,c)
+        '&pxor ("mm6","mm7")',                         # Sigma0_512(a)
+        '&movq ("mm5",&QWP(8*($i+5-1)%64,"esp"))',     # pre-load f
+       '&paddq ($BxC,"mm6")',                          # h+=Sigma0(a)
+        '&movq ("mm6",&QWP(8*($i+6-1)%64,"esp"))',     # pre-load g
+
+       '($A,$BxC) = ($BxC,$A); $i--;'
+       );
+}
+
+&set_label("00_47_ssse3",32);
+
+    for(;$j<16;$j++) {
+       my ($t0,$t2,$t1)=@X[2..4];
+       my @insns = (&BODY_00_15_ssse3(),&BODY_00_15_ssse3());
+
+       &movdqa         ($t2,@X[5]);
+       &movdqa         (@X[1],$t0);                    # restore @X[1]
+       &palignr        ($t0,@X[0],8);                  # X[1..2]
+       &movdqa         (&QWP(16*($j%4),$frame),@X[4]); # off-load @X[4]
+        &palignr       ($t2,@X[4],8);                  # X[9..10]
+
+       &movdqa         ($t1,$t0);
+       &psrlq          ($t0,7);
+        &paddq         (@X[0],$t2);                    # X[0..1] += X[9..10]
+       &movdqa         ($t2,$t1);
+       &psrlq          ($t1,1);
+       &psllq          ($t2,64-8);
+       &pxor           ($t0,$t1);
+       &psrlq          ($t1,8-1);
+       &pxor           ($t0,$t2);
+       &psllq          ($t2,8-1);
+       &pxor           ($t0,$t1);
+        &movdqa        ($t1,@X[7]);
+       &pxor           ($t0,$t2);                      # sigma0(X[1..2])
+        &movdqa        ($t2,@X[7]);
+        &psrlq         ($t1,6);
+       &paddq          (@X[0],$t0);                    # X[0..1] += sigma0(X[1..2])
+
+       &movdqa         ($t0,@X[7]);
+       &psrlq          ($t2,19);
+       &psllq          ($t0,64-61);
+       &pxor           ($t1,$t2);
+       &psrlq          ($t2,61-19);
+       &pxor           ($t1,$t0);
+       &psllq          ($t0,61-19);
+       &pxor           ($t1,$t2);
+       &movdqa         ($t2,&QWP(16*(($j+2)%4),$frame));# pre-restore @X[1]
+       &pxor           ($t1,$t0);                      # sigma0(X[1..2])
+       &movdqa         ($t0,&QWP(16*($j%8),$K512));
+        eval(shift(@insns));
+       &paddq          (@X[0],$t1);                    # X[0..1] += sigma0(X[14..15])
+        eval(shift(@insns));
+        eval(shift(@insns));
+        eval(shift(@insns));
+        eval(shift(@insns));
+       &paddq          ($t0,@X[0]);
+        foreach(@insns) { eval; }
+       &movdqa         (&QWP(16*($j%8)-128,$frame),$t0);# xfer X[i]+K[i]
+
+       push(@X,shift(@X));                             # rotate(@X)
+    }
+       &lea            ($K512,&DWP(16*8,$K512));
+       &dec            ("ecx");
+       &jnz            (&label("00_47_ssse3"));
+
+       &movdqa         (@X[1],&QWP(0,$K512));          # byte swap mask
+       &lea            ($K512,&DWP(-80*8,$K512));      # rewind
+       &movdqu         (@X[0],&QWP(0,"ebx"));
+       &pshufb         (@X[0],@X[1]);
+
+    for ($j=0;$j<8;$j++) {     # load next or same block
+       my @insns = (&BODY_00_15_ssse3(),&BODY_00_15_ssse3());
+
+       &movdqa         (&QWP(16*(($j-1)%4),$frame),@X[3])      if ($j>4); # off-load
+       &movdqa         (@X[3],&QWP(16*($j%8),$K512));
+       &movdqa         (@X[2],@X[1])                           if ($j<7); # perpetuate byte swap mask
+       &movdqu         (@X[1],&QWP(16*($j+1),"ebx"))           if ($j<7); # next input
+       &movdqa         (@X[1],&QWP(16*(($j+1)%4),$frame))      if ($j==7);# restore @X[0]
+       &paddq          (@X[3],@X[0]);
+       &pshufb         (@X[1],@X[2])                           if ($j<7);
+        foreach(@insns) { eval; }
+       &movdqa         (&QWP(16*($j%8)-128,$frame),@X[3]);# xfer X[i]+K[i]
+
+       push(@X,shift(@X));                             # rotate(@X)
+    }
+
+       #&movq  ($A,$Asse2);                    # load A-H
+       &movq   ("mm1",$Bsse2);
+       &paddq  ($A,"mm3");                     # from BODY_00_15
+       #&movq  ($BxC,$Csse2);
+       &movq   ("mm3",$Dsse2);
+       #&movq  ($E,$Esse2);
+       #&movq  ("mm5",$Fsse2);
+       #&movq  ("mm6",$Gsse2);
+       &movq   ("mm7",$Hsse2);
+
+       &pxor   ($BxC,"mm1");                   # de-magic
+       &paddq  ($A,&QWP(0,"esi"));
+       &paddq  ("mm1",&QWP(8,"esi"));
+       &paddq  ($BxC,&QWP(16,"esi"));
+       &paddq  ("mm3",&QWP(24,"esi"));
+       &paddq  ($E,&QWP(32,"esi"));
+       &paddq  ("mm5",&QWP(40,"esi"));
+       &paddq  ("mm6",&QWP(48,"esi"));
+       &paddq  ("mm7",&QWP(56,"esi"));
+
+       &movq   (&QWP(0,"esi"),$A);
+       &movq   (&QWP(8,"esi"),"mm1");
+       &movq   (&QWP(16,"esi"),$BxC);
+       &movq   (&QWP(24,"esi"),"mm3");
+       &movq   (&QWP(32,"esi"),$E);
+       &movq   (&QWP(40,"esi"),"mm5");
+       &movq   (&QWP(48,"esi"),"mm6");
+       &movq   (&QWP(56,"esi"),"mm7");
+
+       &cmp    ("edi","eax")                   # are we done yet?
+       &jb     (&label("loop_ssse3"));
+
+       &mov    ("esp",&DWP(64+12,$frame));     # restore sp
+       &emms   ();
+}
+&function_end_A();
 }
 &set_label("loop_x86",16);
     # copy input block to stack reversing byte and qword order
@@ -655,6 +901,9 @@ if ($sse2) {
        &data_word(0xfc657e2a,0x597f299c);      # u64
        &data_word(0x3ad6faec,0x5fcb6fab);      # u64
        &data_word(0x4a475817,0x6c44198c);      # u64
+
+       &data_word(0x04050607,0x00010203);      # byte swap
+       &data_word(0x0c0d0e0f,0x08090a0b);      # mask
 &function_end_B("sha512_block_data_order");
 &asciz("SHA512 block transform for x86, CRYPTOGAMS by <appro\@openssl.org>");