-#!/usr/bin/env perl
+#! /usr/bin/env perl
+# Copyright 2006-2016 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License"). You may not use
+# this file except in compliance with the License. You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+
# ====================================================================
# Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
# So far RSA *sign* performance improvement over pre-bn_mul_mont asm
# for 64-bit application running on PPC970/G5 is:
#
-# 512-bit +65%
+# 512-bit +65%
# 1024-bit +35%
# 2048-bit +18%
# 4096-bit +4%
$BNSZ= $BITS/8;
$SIZE_T=4;
$RZONE= 224;
- $FRAME= $SIZE_T*16;
$LD= "lwz"; # load
$LDU= "lwzu"; # load and update
$UMULL= "mullw"; # unsigned multiply low
$UMULH= "mulhwu"; # unsigned multiply high
$UCMP= "cmplw"; # unsigned compare
- $SHRI= "srwi"; # unsigned shift right by immediate
+ $SHRI= "srwi"; # unsigned shift right by immediate
$PUSH= $ST;
$POP= $LD;
} elsif ($flavour =~ /64/) {
$BNSZ= $BITS/8;
$SIZE_T=8;
$RZONE= 288;
- $FRAME= $SIZE_T*16;
# same as above, but 64-bit mnemonics...
$LD= "ld"; # load
$UMULL= "mulld"; # unsigned multiply low
$UMULH= "mulhdu"; # unsigned multiply high
$UCMP= "cmpld"; # unsigned compare
- $SHRI= "srdi"; # unsigned shift right by immediate
+ $SHRI= "srdi"; # unsigned shift right by immediate
$PUSH= $ST;
$POP= $LD;
} else { die "nonsense $flavour"; }
+$FRAME=8*$SIZE_T+$RZONE;
+$LOCALS=8*$SIZE_T;
+
$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
( $xlate="${dir}ppc-xlate.pl" and -f $xlate ) or
( $xlate="${dir}../../perlasm/ppc-xlate.pl" and -f $xlate) or
$sp="r1";
$toc="r2";
-$rp="r3"; $ovf="r3";
+$rp="r3";
$ap="r4";
$bp="r5";
$np="r6";
$n0="r7";
$num="r8";
-$rp="r9"; # $rp is reassigned
-$aj="r10";
-$nj="r11";
-$tj="r12";
+
+{
+my $ovf=$rp;
+my $rp="r9"; # $rp is reassigned
+my $aj="r10";
+my $nj="r11";
+my $tj="r12";
# non-volatile registers
-$i="r14";
-$j="r15";
-$tp="r16";
-$m0="r17";
-$m1="r18";
-$lo0="r19";
-$hi0="r20";
-$lo1="r21";
-$hi1="r22";
-$alo="r23";
-$ahi="r24";
-$nlo="r25";
+my $i="r20";
+my $j="r21";
+my $tp="r22";
+my $m0="r23";
+my $m1="r24";
+my $lo0="r25";
+my $hi0="r26";
+my $lo1="r27";
+my $hi1="r28";
+my $alo="r29";
+my $ahi="r30";
+my $nlo="r31";
#
-$nhi="r0";
+my $nhi="r0";
$code=<<___;
.machine "any"
$code.=<<___;
slwi $num,$num,`log($BNSZ)/log(2)`
li $tj,-4096
- addi $ovf,$num,`$FRAME+$RZONE`
+ addi $ovf,$num,$FRAME
subf $ovf,$ovf,$sp ; $sp-$ovf
and $ovf,$ovf,$tj ; minimize TLB usage
subf $ovf,$sp,$ovf ; $ovf-$sp
+ mr $tj,$sp
srwi $num,$num,`log($BNSZ)/log(2)`
$STUX $sp,$sp,$ovf
- $PUSH r14,`4*$SIZE_T`($sp)
- $PUSH r15,`5*$SIZE_T`($sp)
- $PUSH r16,`6*$SIZE_T`($sp)
- $PUSH r17,`7*$SIZE_T`($sp)
- $PUSH r18,`8*$SIZE_T`($sp)
- $PUSH r19,`9*$SIZE_T`($sp)
- $PUSH r20,`10*$SIZE_T`($sp)
- $PUSH r21,`11*$SIZE_T`($sp)
- $PUSH r22,`12*$SIZE_T`($sp)
- $PUSH r23,`13*$SIZE_T`($sp)
- $PUSH r24,`14*$SIZE_T`($sp)
- $PUSH r25,`15*$SIZE_T`($sp)
+ $PUSH r20,`-12*$SIZE_T`($tj)
+ $PUSH r21,`-11*$SIZE_T`($tj)
+ $PUSH r22,`-10*$SIZE_T`($tj)
+ $PUSH r23,`-9*$SIZE_T`($tj)
+ $PUSH r24,`-8*$SIZE_T`($tj)
+ $PUSH r25,`-7*$SIZE_T`($tj)
+ $PUSH r26,`-6*$SIZE_T`($tj)
+ $PUSH r27,`-5*$SIZE_T`($tj)
+ $PUSH r28,`-4*$SIZE_T`($tj)
+ $PUSH r29,`-3*$SIZE_T`($tj)
+ $PUSH r30,`-2*$SIZE_T`($tj)
+ $PUSH r31,`-1*$SIZE_T`($tj)
$LD $n0,0($n0) ; pull n0[0] value
addi $num,$num,-2 ; adjust $num for counter register
\f
$LD $m0,0($bp) ; m0=bp[0]
$LD $aj,0($ap) ; ap[0]
- addi $tp,$sp,$FRAME
+ addi $tp,$sp,$LOCALS
$UMULL $lo0,$aj,$m0 ; ap[0]*bp[0]
$UMULH $hi0,$aj,$m0
addi $j,$j,$BNSZ ; j++
addi $tp,$tp,$BNSZ ; tp++
- bdnz- L1st
+ bdnz L1st
;L1st
addc $lo0,$alo,$hi0
addze $hi0,$ahi
Louter:
$LDX $m0,$bp,$i ; m0=bp[i]
$LD $aj,0($ap) ; ap[0]
- addi $tp,$sp,$FRAME
- $LD $tj,$FRAME($sp) ; tp[0]
+ addi $tp,$sp,$LOCALS
+ $LD $tj,$LOCALS($sp); tp[0]
$UMULL $lo0,$aj,$m0 ; ap[0]*bp[i]
$UMULH $hi0,$aj,$m0
$LD $aj,$BNSZ($ap) ; ap[1]
addze $hi1,$hi1
$ST $lo1,0($tp) ; tp[j-1]
addi $tp,$tp,$BNSZ ; tp++
- bdnz- Linner
+ bdnz Linner
;Linner
$LD $tj,$BNSZ($tp) ; tp[j]
addc $lo0,$alo,$hi0
slwi $tj,$num,`log($BNSZ)/log(2)`
$UCMP $i,$tj
addi $i,$i,$BNSZ
- ble- Louter
+ ble Louter
\f
addi $num,$num,2 ; restore $num
subfc $j,$j,$j ; j=0 and "clear" XER[CA]
- addi $tp,$sp,$FRAME
+ addi $tp,$sp,$LOCALS
mtctr $num
.align 4
subfe $aj,$nj,$tj ; tp[j]-np[j]
$STX $aj,$rp,$j
addi $j,$j,$BNSZ
- bdnz- Lsub
+ bdnz Lsub
li $j,0
mtctr $num
$STX $tj,$rp,$j
$STX $j,$tp,$j ; zap at once
addi $j,$j,$BNSZ
- bdnz- Lcopy
-
- $POP r14,`4*$SIZE_T`($sp)
- $POP r15,`5*$SIZE_T`($sp)
- $POP r16,`6*$SIZE_T`($sp)
- $POP r17,`7*$SIZE_T`($sp)
- $POP r18,`8*$SIZE_T`($sp)
- $POP r19,`9*$SIZE_T`($sp)
- $POP r20,`10*$SIZE_T`($sp)
- $POP r21,`11*$SIZE_T`($sp)
- $POP r22,`12*$SIZE_T`($sp)
- $POP r23,`13*$SIZE_T`($sp)
- $POP r24,`14*$SIZE_T`($sp)
- $POP r25,`15*$SIZE_T`($sp)
- $POP $sp,0($sp)
+ bdnz Lcopy
+
+ $POP $tj,0($sp)
li r3,1
+ $POP r20,`-12*$SIZE_T`($tj)
+ $POP r21,`-11*$SIZE_T`($tj)
+ $POP r22,`-10*$SIZE_T`($tj)
+ $POP r23,`-9*$SIZE_T`($tj)
+ $POP r24,`-8*$SIZE_T`($tj)
+ $POP r25,`-7*$SIZE_T`($tj)
+ $POP r26,`-6*$SIZE_T`($tj)
+ $POP r27,`-5*$SIZE_T`($tj)
+ $POP r28,`-4*$SIZE_T`($tj)
+ $POP r29,`-3*$SIZE_T`($tj)
+ $POP r30,`-2*$SIZE_T`($tj)
+ $POP r31,`-1*$SIZE_T`($tj)
+ mr $sp,$tj
blr
.long 0
-.asciz "Montgomery Multiplication for PPC, CRYPTOGAMS by <appro\@fy.chalmers.se>"
+ .byte 0,12,4,0,0x80,12,6,0
+ .long 0
+.size .bn_mul_mont_int,.-.bn_mul_mont_int
+___
+}
+$code.=<<___;
+.asciz "Montgomery Multiplication for PPC, CRYPTOGAMS by <appro\@openssl.org>"
___
$code =~ s/\`([^\`]*)\`/eval $1/gem;
projects / openssl.git / blobdiff