Synchronize SHA1 assembler with md32_common.h update.
authorAndy Polyakov <appro@openssl.org>
Tue, 17 Oct 2006 07:00:23 +0000 (07:00 +0000)
committerAndy Polyakov <appro@openssl.org>
Tue, 17 Oct 2006 07:00:23 +0000 (07:00 +0000)
crypto/sha/asm/sha1-586.pl
crypto/sha/asm/sha1-ia64.pl
crypto/sha/asm/sha1-ppc.pl
crypto/sha/asm/sha1-x86_64.pl

index 8a8228f..10c7d43 100644 (file)
@@ -2,8 +2,9 @@
 
 # ====================================================================
 # [Re]written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
-# project. Rights for redistribution and usage in source and binary
-# forms are granted according to the OpenSSL license.
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
 # ====================================================================
 
 # "[Re]written" was achieved in two major overhauls. In 2004 BODY_*
 # improvement on P4 outweights the loss and incorporate this
 # re-tuned code to 0.9.7 and later.
 # ----------------------------------------------------------------
-# Those who for any particular reason absolutely must score on
-# Pentium can replace this module with one from 0.9.6 distribution.
-# This "offer" shall be revoked the moment programming interface to
-# this module is changed, in which case this paragraph should be
-# removed.
-# ----------------------------------------------------------------
 #                                      <appro@fy.chalmers.se>
 
-push(@INC,"perlasm","../../perlasm");
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+push(@INC,"${dir}","${dir}../../perlasm");
 require "x86asm.pl";
 
 &asm_init($ARGV[0],"sha1-586.pl",$ARGV[$#ARGV] eq "386");
@@ -189,10 +185,6 @@ sub BODY_40_59
        &mov($D,&DWP(12,$tmp1));
        # E is pre-loaded
 
-       &set_label("shortcut");
-       # keep a note of shortcut label so it can be used outside the block.
-       $shortcut = &label("shortcut");
-
        for($i=0;$i<16;$i++)    { &BODY_00_15($i,@V); unshift(@V,pop(@V)); }
        for(;$i<20;$i++)        { &BODY_16_19($i,@V); unshift(@V,pop(@V)); }
        for(;$i<40;$i++)        { &BODY_20_39($i,@V); unshift(@V,pop(@V)); }
@@ -223,35 +215,6 @@ sub BODY_40_59
 
        &stack_pop(16);
 &function_end("sha1_block_asm_data_order");
-
-&function_begin("sha1_block_asm_host_order",16);
-       &mov($tmp1,&wparam(0)); # SHA_CTX *c
-       &mov($T,&wparam(1));    # const void *input
-       #&mov($A,&wparam(2));   # size_t num, always 1
-       &stack_push(16);        # allocate X[16]
-       &lea($A,&DWP(64,$T));   # this works, because num is always 1 here
-       &mov(&wparam(2),$A);    # pointer beyond the end of input
-
-       # just copy input to X
-       for ($i=0; $i<16; $i+=4)
-               {
-               &mov($A,&DWP(4*($i+0),$T));
-               &mov($B,&DWP(4*($i+1),$T));
-               &mov($C,&DWP(4*($i+2),$T));
-               &mov($D,&DWP(4*($i+3),$T));
-               &mov(&swtmp($i+0),$A);
-               &mov(&swtmp($i+1),$B);
-               &mov(&swtmp($i+2),$C);
-               &mov(&swtmp($i+3),$D);
-               }
-
-       &mov($A,&DWP(0,$tmp1)); # load SHA_CTX
-       &mov($B,&DWP(4,$tmp1));
-       &mov($C,&DWP(8,$tmp1));
-       &mov($D,&DWP(12,$tmp1));
-       &mov($E,&DWP(16,$tmp1));
-
-       &jmp($shortcut);        # this works, because num is always 1
-&function_end_B("sha1_block_asm_host_order");
+&asciz("SHA1 block transform for x86, CRYPTOGAMS by <appro\@openssl.org>");
 
 &asm_finish();
index 9478f5d..a5d8197 100644 (file)
@@ -2,8 +2,9 @@
 #
 # ====================================================================
 # Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
-# project. Rights for redistribution and usage in source and binary
-# forms are granted according to the OpenSSL license.
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
 # ====================================================================
 #
 # Eternal question is what's wrong with compiler generated code? The
 # to perform rotations by maintaining copy of 32-bit value in upper
 # bits of 64-bit register. Just follow mux2 and shrp instructions...
 # Performance under big-endian OS such as HP-UX is 179MBps*1GHz, which
-# is >50% better than HP C and >2x better than gcc. As of this moment
-# performance under little-endian OS such as Linux and Windows will be
-# a bit lower, because data has to be picked in reverse byte-order.
-# It's possible to resolve this issue by implementing third function,
-# sha1_block_asm_data_order_aligned, which would temporarily flip
-# BE field in User Mask register...
+# is >50% better than HP C and >2x better than gcc.
 
 $code=<<___;
-.ident  \"sha1-ia64.s, version 1.0\"
+.ident  \"sha1-ia64.s, version 1.1\"
 .ident  \"IA-64 ISA artwork by Andy Polyakov <appro\@fy.chalmers.se>\"
 .explicit
 
@@ -245,164 +241,11 @@ tmp3=r11;
 ctx=r32;       // in0
 inp=r33;       // in1
 
-// void sha1_block_asm_host_order(SHA_CTX *c,const void *p,size_t num);
-.global        sha1_block_asm_host_order#
-.proc  sha1_block_asm_host_order#
-.align 32
-sha1_block_asm_host_order:
-       .prologue
-{ .mmi;        alloc   tmp1=ar.pfs,3,15,0,0
-       $ADDP   tmp0=4,ctx
-       .save   ar.lc,r3
-       mov     r3=ar.lc                }
-{ .mmi;        $ADDP   ctx=0,ctx
-       $ADDP   inp=0,inp
-       mov     r2=pr                   };;
-tmp4=in2;
-tmp5=loc13;
-tmp6=loc14;
-       .body
-{ .mlx;        ld4     $h0=[ctx],8
-       movl    $K_00_19=0x5a827999     }
-{ .mlx;        ld4     $h1=[tmp0],8
-       movl    $K_20_39=0x6ed9eba1     };;
-{ .mlx;        ld4     $h2=[ctx],8
-       movl    $K_40_59=0x8f1bbcdc     }
-{ .mlx;        ld4     $h3=[tmp0]
-       movl    $K_60_79=0xca62c1d6     };;
-{ .mmi;        ld4     $h4=[ctx],-16
-       add     in2=-1,in2                  // adjust num for ar.lc
-       mov     ar.ec=1                 };;
-{ .mmi;        ld4     $X[0]=[inp],4               // prefetch
-       cmp.ne  p16,p0=r0,in2               // prefecth at loop end
-       mov     ar.lc=in2               };; // brp.loop.imp: too far
-
-.Lhtop:
-{ .mmi;        mov     $A=$h0
-       mov     $B=$h1
-       mux2    tmp6=$h1,0x44           }
-{ .mmi;        mov     $C=$h2
-       mov     $D=$h3
-       mov     $E=$h4                  };;
-
-___
-
-       &BODY_00_15(\$code, 0,$A,$B,$C,$D,$E,$T);
-       &BODY_00_15(\$code, 1,$T,$A,$B,$C,$D,$E);
-       &BODY_00_15(\$code, 2,$E,$T,$A,$B,$C,$D);
-       &BODY_00_15(\$code, 3,$D,$E,$T,$A,$B,$C);
-       &BODY_00_15(\$code, 4,$C,$D,$E,$T,$A,$B);
-       &BODY_00_15(\$code, 5,$B,$C,$D,$E,$T,$A);
-       &BODY_00_15(\$code, 6,$A,$B,$C,$D,$E,$T);
-       &BODY_00_15(\$code, 7,$T,$A,$B,$C,$D,$E);
-       &BODY_00_15(\$code, 8,$E,$T,$A,$B,$C,$D);
-       &BODY_00_15(\$code, 9,$D,$E,$T,$A,$B,$C);
-       &BODY_00_15(\$code,10,$C,$D,$E,$T,$A,$B);
-       &BODY_00_15(\$code,11,$B,$C,$D,$E,$T,$A);
-       &BODY_00_15(\$code,12,$A,$B,$C,$D,$E,$T);
-       &BODY_00_15(\$code,13,$T,$A,$B,$C,$D,$E);
-       &BODY_00_15(\$code,14,$E,$T,$A,$B,$C,$D);
-       &BODY_00_15(\$code,15,$D,$E,$T,$A,$B,$C);
-
-       &BODY_16_19(\$code,16,$C,$D,$E,$T,$A,$B);
-       &BODY_16_19(\$code,17,$B,$C,$D,$E,$T,$A);
-       &BODY_16_19(\$code,18,$A,$B,$C,$D,$E,$T);
-       &BODY_16_19(\$code,19,$T,$A,$B,$C,$D,$E);
-
-       &BODY_20_39(\$code,20,$E,$T,$A,$B,$C,$D);
-       &BODY_20_39(\$code,21,$D,$E,$T,$A,$B,$C);
-       &BODY_20_39(\$code,22,$C,$D,$E,$T,$A,$B);
-       &BODY_20_39(\$code,23,$B,$C,$D,$E,$T,$A);
-       &BODY_20_39(\$code,24,$A,$B,$C,$D,$E,$T);
-       &BODY_20_39(\$code,25,$T,$A,$B,$C,$D,$E);
-       &BODY_20_39(\$code,26,$E,$T,$A,$B,$C,$D);
-       &BODY_20_39(\$code,27,$D,$E,$T,$A,$B,$C);
-       &BODY_20_39(\$code,28,$C,$D,$E,$T,$A,$B);
-       &BODY_20_39(\$code,29,$B,$C,$D,$E,$T,$A);
-       &BODY_20_39(\$code,30,$A,$B,$C,$D,$E,$T);
-       &BODY_20_39(\$code,31,$T,$A,$B,$C,$D,$E);
-       &BODY_20_39(\$code,32,$E,$T,$A,$B,$C,$D);
-       &BODY_20_39(\$code,33,$D,$E,$T,$A,$B,$C);
-       &BODY_20_39(\$code,34,$C,$D,$E,$T,$A,$B);
-       &BODY_20_39(\$code,35,$B,$C,$D,$E,$T,$A);
-       &BODY_20_39(\$code,36,$A,$B,$C,$D,$E,$T);
-       &BODY_20_39(\$code,37,$T,$A,$B,$C,$D,$E);
-       &BODY_20_39(\$code,38,$E,$T,$A,$B,$C,$D);
-       &BODY_20_39(\$code,39,$D,$E,$T,$A,$B,$C);
-
-       &BODY_40_59(\$code,40,$C,$D,$E,$T,$A,$B);
-       &BODY_40_59(\$code,41,$B,$C,$D,$E,$T,$A);
-       &BODY_40_59(\$code,42,$A,$B,$C,$D,$E,$T);
-       &BODY_40_59(\$code,43,$T,$A,$B,$C,$D,$E);
-       &BODY_40_59(\$code,44,$E,$T,$A,$B,$C,$D);
-       &BODY_40_59(\$code,45,$D,$E,$T,$A,$B,$C);
-       &BODY_40_59(\$code,46,$C,$D,$E,$T,$A,$B);
-       &BODY_40_59(\$code,47,$B,$C,$D,$E,$T,$A);
-       &BODY_40_59(\$code,48,$A,$B,$C,$D,$E,$T);
-       &BODY_40_59(\$code,49,$T,$A,$B,$C,$D,$E);
-       &BODY_40_59(\$code,50,$E,$T,$A,$B,$C,$D);
-       &BODY_40_59(\$code,51,$D,$E,$T,$A,$B,$C);
-       &BODY_40_59(\$code,52,$C,$D,$E,$T,$A,$B);
-       &BODY_40_59(\$code,53,$B,$C,$D,$E,$T,$A);
-       &BODY_40_59(\$code,54,$A,$B,$C,$D,$E,$T);
-       &BODY_40_59(\$code,55,$T,$A,$B,$C,$D,$E);
-       &BODY_40_59(\$code,56,$E,$T,$A,$B,$C,$D);
-       &BODY_40_59(\$code,57,$D,$E,$T,$A,$B,$C);
-       &BODY_40_59(\$code,58,$C,$D,$E,$T,$A,$B);
-       &BODY_40_59(\$code,59,$B,$C,$D,$E,$T,$A);
-
-       &BODY_60_79(\$code,60,$A,$B,$C,$D,$E,$T);
-       &BODY_60_79(\$code,61,$T,$A,$B,$C,$D,$E);
-       &BODY_60_79(\$code,62,$E,$T,$A,$B,$C,$D);
-       &BODY_60_79(\$code,63,$D,$E,$T,$A,$B,$C);
-       &BODY_60_79(\$code,64,$C,$D,$E,$T,$A,$B);
-       &BODY_60_79(\$code,65,$B,$C,$D,$E,$T,$A);
-       &BODY_60_79(\$code,66,$A,$B,$C,$D,$E,$T);
-       &BODY_60_79(\$code,67,$T,$A,$B,$C,$D,$E);
-       &BODY_60_79(\$code,68,$E,$T,$A,$B,$C,$D);
-       &BODY_60_79(\$code,69,$D,$E,$T,$A,$B,$C);
-       &BODY_60_79(\$code,70,$C,$D,$E,$T,$A,$B);
-       &BODY_60_79(\$code,71,$B,$C,$D,$E,$T,$A);
-       &BODY_60_79(\$code,72,$A,$B,$C,$D,$E,$T);
-       &BODY_60_79(\$code,73,$T,$A,$B,$C,$D,$E);
-       &BODY_60_79(\$code,74,$E,$T,$A,$B,$C,$D);
-       &BODY_60_79(\$code,75,$D,$E,$T,$A,$B,$C);
-       &BODY_60_79(\$code,76,$C,$D,$E,$T,$A,$B);
-       &BODY_60_79(\$code,77,$B,$C,$D,$E,$T,$A);
-       &BODY_60_79(\$code,78,$A,$B,$C,$D,$E,$T);
-       &BODY_60_79(\$code,79,$T,$A,$B,$C,$D,$E);
-
-$code.=<<___;
-{ .mmb;        add     $h0=$h0,$E
-       nop.m   0
-       br.ctop.dptk.many       .Lhtop  };;
-.Lhend:
-{ .mmi;        add     tmp0=4,ctx
-       mov     ar.lc=r3                };;
-{ .mmi;        st4     [ctx]=$h0,8
-       st4     [tmp0]=$h1,8            };;
-{ .mmi;        st4     [ctx]=$h2,8
-       st4     [tmp0]=$h3              };;
-{ .mib;        st4     [ctx]=$h4,-16
-       mov     pr=r2,0x1ffff
-       br.ret.sptk.many        b0      };;
-.endp  sha1_block_asm_host_order#
-___
-
-
-$code.=<<___;
 // void sha1_block_asm_data_order(SHA_CTX *c,const void *p,size_t num);
 .global        sha1_block_asm_data_order#
 .proc  sha1_block_asm_data_order#
 .align 32
 sha1_block_asm_data_order:
-___
-$code.=<<___ if ($big_endian);
-{ .mmi;        and     r2=3,inp                                };;
-{ .mib;        cmp.eq  p6,p0=r0,r2
-(p6)   br.dptk.many    sha1_block_asm_host_order       };;
-___
-$code.=<<___;
        .prologue
 { .mmi;        alloc   tmp1=ar.pfs,3,15,0,0
        $ADDP   tmp0=4,ctx
@@ -440,90 +283,16 @@ tmp6=loc14;
 
 ___
 
-       &BODY_00_15(\$code, 0,$A,$B,$C,$D,$E,$T,1);
-       &BODY_00_15(\$code, 1,$T,$A,$B,$C,$D,$E,1);
-       &BODY_00_15(\$code, 2,$E,$T,$A,$B,$C,$D,1);
-       &BODY_00_15(\$code, 3,$D,$E,$T,$A,$B,$C,1);
-       &BODY_00_15(\$code, 4,$C,$D,$E,$T,$A,$B,1);
-       &BODY_00_15(\$code, 5,$B,$C,$D,$E,$T,$A,1);
-       &BODY_00_15(\$code, 6,$A,$B,$C,$D,$E,$T,1);
-       &BODY_00_15(\$code, 7,$T,$A,$B,$C,$D,$E,1);
-       &BODY_00_15(\$code, 8,$E,$T,$A,$B,$C,$D,1);
-       &BODY_00_15(\$code, 9,$D,$E,$T,$A,$B,$C,1);
-       &BODY_00_15(\$code,10,$C,$D,$E,$T,$A,$B,1);
-       &BODY_00_15(\$code,11,$B,$C,$D,$E,$T,$A,1);
-       &BODY_00_15(\$code,12,$A,$B,$C,$D,$E,$T,1);
-       &BODY_00_15(\$code,13,$T,$A,$B,$C,$D,$E,1);
-       &BODY_00_15(\$code,14,$E,$T,$A,$B,$C,$D,1);
-       &BODY_00_15(\$code,15,$D,$E,$T,$A,$B,$C,1);
+{ my $i,@V=($A,$B,$C,$D,$E,$T);
 
-       &BODY_16_19(\$code,16,$C,$D,$E,$T,$A,$B);
-       &BODY_16_19(\$code,17,$B,$C,$D,$E,$T,$A);
-       &BODY_16_19(\$code,18,$A,$B,$C,$D,$E,$T);
-       &BODY_16_19(\$code,19,$T,$A,$B,$C,$D,$E);
+       for($i=0;$i<16;$i++)    { &BODY_00_15(\$code,$i,@V,1); unshift(@V,pop(@V)); }
+       for(;$i<20;$i++)        { &BODY_16_19(\$code,$i,@V); unshift(@V,pop(@V)); }
+       for(;$i<40;$i++)        { &BODY_20_39(\$code,$i,@V); unshift(@V,pop(@V)); }
+       for(;$i<60;$i++)        { &BODY_40_59(\$code,$i,@V); unshift(@V,pop(@V)); }
+       for(;$i<80;$i++)        { &BODY_60_79(\$code,$i,@V); unshift(@V,pop(@V)); }
 
-       &BODY_20_39(\$code,20,$E,$T,$A,$B,$C,$D);
-       &BODY_20_39(\$code,21,$D,$E,$T,$A,$B,$C);
-       &BODY_20_39(\$code,22,$C,$D,$E,$T,$A,$B);
-       &BODY_20_39(\$code,23,$B,$C,$D,$E,$T,$A);
-       &BODY_20_39(\$code,24,$A,$B,$C,$D,$E,$T);
-       &BODY_20_39(\$code,25,$T,$A,$B,$C,$D,$E);
-       &BODY_20_39(\$code,26,$E,$T,$A,$B,$C,$D);
-       &BODY_20_39(\$code,27,$D,$E,$T,$A,$B,$C);
-       &BODY_20_39(\$code,28,$C,$D,$E,$T,$A,$B);
-       &BODY_20_39(\$code,29,$B,$C,$D,$E,$T,$A);
-       &BODY_20_39(\$code,30,$A,$B,$C,$D,$E,$T);
-       &BODY_20_39(\$code,31,$T,$A,$B,$C,$D,$E);
-       &BODY_20_39(\$code,32,$E,$T,$A,$B,$C,$D);
-       &BODY_20_39(\$code,33,$D,$E,$T,$A,$B,$C);
-       &BODY_20_39(\$code,34,$C,$D,$E,$T,$A,$B);
-       &BODY_20_39(\$code,35,$B,$C,$D,$E,$T,$A);
-       &BODY_20_39(\$code,36,$A,$B,$C,$D,$E,$T);
-       &BODY_20_39(\$code,37,$T,$A,$B,$C,$D,$E);
-       &BODY_20_39(\$code,38,$E,$T,$A,$B,$C,$D);
-       &BODY_20_39(\$code,39,$D,$E,$T,$A,$B,$C);
-
-       &BODY_40_59(\$code,40,$C,$D,$E,$T,$A,$B);
-       &BODY_40_59(\$code,41,$B,$C,$D,$E,$T,$A);
-       &BODY_40_59(\$code,42,$A,$B,$C,$D,$E,$T);
-       &BODY_40_59(\$code,43,$T,$A,$B,$C,$D,$E);
-       &BODY_40_59(\$code,44,$E,$T,$A,$B,$C,$D);
-       &BODY_40_59(\$code,45,$D,$E,$T,$A,$B,$C);
-       &BODY_40_59(\$code,46,$C,$D,$E,$T,$A,$B);
-       &BODY_40_59(\$code,47,$B,$C,$D,$E,$T,$A);
-       &BODY_40_59(\$code,48,$A,$B,$C,$D,$E,$T);
-       &BODY_40_59(\$code,49,$T,$A,$B,$C,$D,$E);
-       &BODY_40_59(\$code,50,$E,$T,$A,$B,$C,$D);
-       &BODY_40_59(\$code,51,$D,$E,$T,$A,$B,$C);
-       &BODY_40_59(\$code,52,$C,$D,$E,$T,$A,$B);
-       &BODY_40_59(\$code,53,$B,$C,$D,$E,$T,$A);
-       &BODY_40_59(\$code,54,$A,$B,$C,$D,$E,$T);
-       &BODY_40_59(\$code,55,$T,$A,$B,$C,$D,$E);
-       &BODY_40_59(\$code,56,$E,$T,$A,$B,$C,$D);
-       &BODY_40_59(\$code,57,$D,$E,$T,$A,$B,$C);
-       &BODY_40_59(\$code,58,$C,$D,$E,$T,$A,$B);
-       &BODY_40_59(\$code,59,$B,$C,$D,$E,$T,$A);
-
-       &BODY_60_79(\$code,60,$A,$B,$C,$D,$E,$T);
-       &BODY_60_79(\$code,61,$T,$A,$B,$C,$D,$E);
-       &BODY_60_79(\$code,62,$E,$T,$A,$B,$C,$D);
-       &BODY_60_79(\$code,63,$D,$E,$T,$A,$B,$C);
-       &BODY_60_79(\$code,64,$C,$D,$E,$T,$A,$B);
-       &BODY_60_79(\$code,65,$B,$C,$D,$E,$T,$A);
-       &BODY_60_79(\$code,66,$A,$B,$C,$D,$E,$T);
-       &BODY_60_79(\$code,67,$T,$A,$B,$C,$D,$E);
-       &BODY_60_79(\$code,68,$E,$T,$A,$B,$C,$D);
-       &BODY_60_79(\$code,69,$D,$E,$T,$A,$B,$C);
-       &BODY_60_79(\$code,70,$C,$D,$E,$T,$A,$B);
-       &BODY_60_79(\$code,71,$B,$C,$D,$E,$T,$A);
-       &BODY_60_79(\$code,72,$A,$B,$C,$D,$E,$T);
-       &BODY_60_79(\$code,73,$T,$A,$B,$C,$D,$E);
-       &BODY_60_79(\$code,74,$E,$T,$A,$B,$C,$D);
-       &BODY_60_79(\$code,75,$D,$E,$T,$A,$B,$C);
-       &BODY_60_79(\$code,76,$C,$D,$E,$T,$A,$B);
-       &BODY_60_79(\$code,77,$B,$C,$D,$E,$T,$A);
-       &BODY_60_79(\$code,78,$A,$B,$C,$D,$E,$T);
-       &BODY_60_79(\$code,79,$T,$A,$B,$C,$D,$E);
+       (($V[5] eq $D) and ($V[0] eq $E)) or die;       # double-check
+}
 
 $code.=<<___;
 { .mmb;        add     $h0=$h0,$E
index 704887b..5c5c7d8 100755 (executable)
@@ -2,8 +2,9 @@
 
 # ====================================================================
 # Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
-# project. Rights for redistribution and usage in source and binary
-# forms are granted according to the OpenSSL license.
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
 # ====================================================================
 
 # I let hardware handle unaligned input(*), except on page boundaries
@@ -34,8 +35,13 @@ if ($output =~ /64\.s/) {
        $PUSH   ="stw";
 } else { die "nonsense $output"; }
 
-( defined shift || open STDOUT,"| $^X ../perlasm/ppc-xlate.pl $output" ) ||
-       die "can't call ../perlasm/ppc-xlate.pl: $!";
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+( $xlate="${dir}ppc-xlate.pl" and -f $xlate ) or
+( $xlate="${dir}../../perlasm/ppc-xlate.pl" and -f $xlate) or
+die "can't locate ppc-xlate.pl";
+
+( defined shift || open STDOUT,"| $^X $xlate $output" ) ||
+       die "can't call $xlate: $!";
 
 $FRAME=24*$SIZE_T;
 
@@ -303,6 +309,9 @@ $code.=<<___;
        bdnz-   Lsha1_block_private
        blr
 ___
+$code.=<<___;
+.asciz "SHA1 block transform for PPC, CRYPTOGAMS by <appro\@fy.chalmers.se>"
+___
 
 $code =~ s/\`([^\`]*)\`/eval $1/gem;
 print $code;
index 433e2f9..857db2d 100755 (executable)
@@ -2,8 +2,9 @@
 #
 # ====================================================================
 # Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
-# project. Rights for redistribution and usage in source and binary
-# forms are granted according to the OpenSSL license.
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
 # ====================================================================
 #
 # sha1_block procedure for x86_64.
 # Xeon         +65%            +0%             9.9
 
 $output=shift;
-open STDOUT,"| $^X ../perlasm/x86_64-xlate.pl $output";
+
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or
+( $xlate="${dir}../../perlasm/x86_64-xlate.pl" and -f $xlate) or
+die "can't locate x86_64-xlate.pl";
+
+open STDOUT,"| $^X $xlate $output";
 
 $ctx="%rdi";   # 1st arg
 $inp="%rsi";   # 2nd arg
@@ -112,7 +119,6 @@ $code.=<<___ if ($i<15);
        rol     \$30,$b
        add     $t0,$f
 ___
-$code.=".Lshortcut:\n" if ($i==15);
 $code.=<<___ if ($i>=15);
        lea     0x5a827999($xi,$e),$f
        mov     `4*($j%16)`(%rsp),$xi
@@ -222,18 +228,12 @@ $code.=<<___;
        jnz     .Lloop
 ___
 &EPILOGUE("sha1_block_asm_data_order");
-
-####################################################################
-
-@V=($A,$B,$C,$D,$E,$T);
-
-&PROLOGUE("sha1_block_asm_host_order");
-for($i=0;$i<15;$i++)   { &BODY_00_19($i,@V,1); unshift(@V,pop(@V)); }
 $code.=<<___;
-       jmp     .Lshortcut
-.size  sha1_block_asm_host_order,.-sha1_block_asm_host_order
+.asciz "SHA1 block transform for x86_64, CRYPTOGAMS by <appro\@openssl.org>"
 ___
 
+####################################################################
+
 $code =~ s/\`([^\`]*)\`/eval $1/gem;
 print $code;
 close STDOUT;