openssl/crypto/bn/asm/x86_64-mont.pl
2005-10-07 15:18:16 +00:00

206 lines
4.1 KiB
Raku
Executable file

#!/usr/bin/env perl
# ====================================================================
# 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.
# ====================================================================
# October 2005.
#
# Montgomery multiplication routine for x86_64. While it gives modest
# 9% improvement of rsa4096 sign on Opteron, rsa512 sign runs more
# than twice, >2x, as fast. Most common rsa1024 sign is improved by
# respectful 50%. It remains to be seen if loop unrolling and
# dedicated squaring routine can provide further improvement...
$output=shift;
open STDOUT,"| $^X ../perlasm/x86_64-xlate.pl $output";
# int bn_mul_mont(
$rp="%rdi"; # BN_ULONG *rp,
$ap="%rsi"; # const BN_ULONG *ap,
$bp="%rdx"; # const BN_ULONG *bp,
$np="%rcx"; # const BN_ULONG *np,
$n0="%r8"; # BN_ULONG n0,
$num="%r9"; # int num);
$lo0="%r10";
$hi0="%r11";
$bp="%r12"; # reassign $bp
$hi1="%r13";
$i="%r14";
$j="%r15";
$m0="%rbx";
$m1="%rbp";
$code=<<___;
.text
.globl bn_mul_mont
.type bn_mul_mont,\@function,6
.align 16
bn_mul_mont:
push %rbx
push %rbp
push %r12
push %r13
push %r14
push %r15
lea 2($num),%rax
mov %rsp,%rbp
neg %rax
lea (%rsp,%rax,8),%rsp # tp=alloca(8*(num+2))
and \$-1024,%rsp # minimize TLB usage
mov %rbp,8(%rsp,$num,8) # tp[num+1]=%rsp
mov %rdx,$bp # $bp reassigned, remember?
xor $i,$i # i=0
xor $j,$j # j=0
mov ($bp),$m0 # m0=bp[0]
mov ($ap),%rax
mulq $m0 # ap[0]*bp[0]
mov %rax,$lo0
mov %rdx,$hi0
imulq $n0,%rax # "tp[0]"*n0
mov %rax,$m1
mulq ($np) # np[0]*m1
add $lo0,%rax # discarded
adc \$0,%rdx
mov %rdx,$hi1
lea 1($j),$j # j++
.L1st:
mov ($ap,$j,8),%rax
mulq $m0 # ap[j]*bp[0]
add $hi0,%rax
adc \$0,%rdx
mov %rax,$lo0
mov %rdx,$hi0
mov ($np,$j,8),%rax
mulq $m1 # np[j]*m1
add $hi1,%rax
adc \$0,%rdx
add $lo0,%rax # np[j]*m1+ap[j]*bp[0]
adc \$0,%rdx
mov %rax,-8(%rsp,$j,8) # tp[j-1]
mov %rdx,$hi1
lea 1($j),$j # j++
cmp $num,$j
jl .L1st
xor %rdx,%rdx
add $hi0,$hi1
adc \$0,%rdx
mov $hi1,-8(%rsp,$num,8)
mov %rdx,(%rsp,$num,8) # store upmost overflow bit
lea 1($i),$i # i++
.align 4
.Louter:
xor $j,$j # j=0
mov ($bp,$i,8),$m0 # m0=bp[i]
mov ($ap),%rax # ap[0]
mulq $m0 # ap[0]*bp[i]
add (%rsp),%rax # ap[0]*bp[i]+tp[0]
adc \$0,%rdx
mov %rax,$lo0
mov %rdx,$hi0
imulq $n0,%rax # tp[0]*n0
mov %rax,$m1
mulq ($np,$j,8) # np[0]*m1
add $lo0,%rax # discarded
adc \$0,%rdx
mov %rdx,$hi1
lea 1($j),$j # j++
.align 4
.Linner:
mov ($ap,$j,8),%rax
mulq $m0 # ap[j]*bp[i]
add $hi0,%rax
adc \$0,%rdx
add (%rsp,$j,8),%rax # ap[j]*bp[i]+tp[j]
adc \$0,%rdx
mov %rax,$lo0
mov %rdx,$hi0
mov ($np,$j,8),%rax
mulq $m1 # np[j]*m1
add $hi1,%rax
adc \$0,%rdx
add $lo0,%rax # np[j]*m1+ap[j]*bp[i]+tp[j]
adc \$0,%rdx
mov %rax,-8(%rsp,$j,8) # tp[j-1]
mov %rdx,$hi1
lea 1($j),$j # j++
cmp $num,$j
jl .Linner
xor %rdx,%rdx
add $hi0,$hi1
adc \$0,%rdx
add (%rsp,$num,8),$hi1 # pull upmost overflow bit
adc \$0,%rdx
mov $hi1,-8(%rsp,$num,8)
mov %rdx,(%rsp,$num,8) # store upmost overflow bit
lea 1($i),$i # i++
cmp $num,$i
jl .Louter
xor $i,$i # i=0
lea -1($num),$j # j=num-1
cmp \$0,%rdx # %rdx still holds upmost overflow bit
jnz .Lsub # CF is cleared by compare with 0
mov (%rsp,$j,8),%rax
cmp ($np,$j,8),%rax # tp[num-1]-np[num-1]
jae .Lsub # if taken CF was cleared by above cmp
.align 4
.Lcopy:
mov (%rsp,$j,8),%rax
mov %rax,($rp,$j,8) # rp[i]=tp[i]
mov $i,(%rsp,$j,8) # zap temporary vector
dec $j
jge .Lcopy
.align 4
.Lexit:
mov 8(%rsp,$num,8),%rsp # restore %rsp
mov \$1,%rax
pop %r15
pop %r14
pop %r13
pop %r12
pop %rbp
pop %rbx
ret
.align 16
.Lsub: mov (%rsp,$i,8),%rax
sbb ($np,$i,8),%rax
mov %rax,($rp,$i,8) # rp[i]=tp[i]-np[j]
lea 1($i),$i # i++
dec $j # doesn't affect CF!
jge .Lsub
lea -1($num),$j # j=num-1
sbb \$0,%rdx
jc .Lcopy # tp was less than np
.align 4
.Lzap: mov $i,(%rsp,$j,8) # zap temporary vector
dec $j
jge .Lzap
jmp .Lexit
.size bn_mul_mont,.-bn_mul_mont
___
print $code;
close STDOUT;