-#!/usr/bin/env perl
+#! /usr/bin/env perl
+# Copyright 2005-2016 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@openssl.org> for the OpenSSL
# level parallelism, on a given CPU implementation in this case.
#
# Special note on Intel EM64T. While Opteron CPU exhibits perfect
-# perfromance ratio of 1.5 between 64- and 32-bit flavors [see above],
+# performance ratio of 1.5 between 64- and 32-bit flavors [see above],
# [currently available] EM64T CPUs apparently are far from it. On the
# contrary, 64-bit version, sha512_block, is ~30% *slower* than 32-bit
# sha256_block:-( This is presumably because 64-bit shifts/rotates
# Haswell 12.2 9.28(+31%) 7.80(+56%) 7.66 5.40(+42%)
# Skylake 11.4 9.03(+26%) 7.70(+48%) 7.25 5.20(+40%)
# Bulldozer 21.1 13.6(+54%) 13.6(+54%(***)) 13.5 8.58(+57%)
+# Ryzen 11.0 9.02(+22%) 2.05(+440%) 7.05 5.67(+20%)
# VIA Nano 23.0 16.5(+39%) - 14.7 -
# Atom 23.0 18.9(+22%) - 14.7 -
# Silvermont 27.4 20.6(+33%) - 17.5 -
+# Knights L 27.4 21.0(+30%) 19.6(+40%) 17.5 12.8(+37%)
+# Goldmont 18.9 14.3(+32%) 4.16(+350%) 12.0 -
#
-# (*) whichever best applicable;
+# (*) whichever best applicable, including SHAEXT;
# (**) switch from ror to shrd stands for fair share of improvement;
# (***) execution time is fully determined by remaining integer-only
# part, body_00_15; reducing the amount of SIMD instructions
$shaext=1; ### set to zero if compiling for 1.0.1
$avx=1 if (!$shaext && $avx);
-open OUT,"| \"$^X\" $xlate $flavour $output";
+open OUT,"| \"$^X\" \"$xlate\" $flavour \"$output\"";
*STDOUT=*OUT;
if ($output =~ /512/) {
$_ctx="16*$SZ+0*8(%rsp)";
$_inp="16*$SZ+1*8(%rsp)";
$_end="16*$SZ+2*8(%rsp)";
-$_rsp="16*$SZ+3*8(%rsp)";
+$_rsp="`16*$SZ+3*8`(%rsp)";
$framesz="16*$SZ+4*8";
.type $func,\@function,3
.align 16
$func:
+.cfi_startproc
___
$code.=<<___ if ($SZ==4 || $avx);
lea OPENSSL_ia32cap_P(%rip),%r11
jnz .Lssse3_shortcut
___
$code.=<<___;
+ mov %rsp,%rax # copy %rsp
+.cfi_def_cfa_register %rax
push %rbx
+.cfi_push %rbx
push %rbp
+.cfi_push %rbp
push %r12
+.cfi_push %r12
push %r13
+.cfi_push %r13
push %r14
+.cfi_push %r14
push %r15
- mov %rsp,%r11 # copy %rsp
+.cfi_push %r15
shl \$4,%rdx # num*16
sub \$$framesz,%rsp
lea ($inp,%rdx,$SZ),%rdx # inp+num*16*$SZ
mov $ctx,$_ctx # save ctx, 1st arg
mov $inp,$_inp # save inp, 2nd arh
mov %rdx,$_end # save end pointer, "3rd" arg
- mov %r11,$_rsp # save copy of %rsp
+ mov %rax,$_rsp # save copy of %rsp
+.cfi_cfa_expression $_rsp,deref,+8
.Lprologue:
mov $SZ*0($ctx),$A
jb .Lloop
mov $_rsp,%rsi
- mov (%rsi),%r15
- mov 8(%rsi),%r14
- mov 16(%rsi),%r13
- mov 24(%rsi),%r12
- mov 32(%rsi),%rbp
- mov 40(%rsi),%rbx
- lea 48(%rsi),%rsp
+.cfi_def_cfa %rsi,8
+ mov -48(%rsi),%r15
+.cfi_restore %r15
+ mov -40(%rsi),%r14
+.cfi_restore %r14
+ mov -32(%rsi),%r13
+.cfi_restore %r13
+ mov -24(%rsi),%r12
+.cfi_restore %r12
+ mov -16(%rsi),%rbp
+.cfi_restore %rbp
+ mov -8(%rsi),%rbx
+.cfi_restore %rbx
+ lea (%rsi),%rsp
+.cfi_def_cfa_register %rsp
.Lepilogue:
ret
+.cfi_endproc
.size $func,.-$func
___
.type ${func}_ssse3,\@function,3
.align 64
${func}_ssse3:
+.cfi_startproc
.Lssse3_shortcut:
+ mov %rsp,%rax # copy %rsp
+.cfi_def_cfa_register %rax
push %rbx
+.cfi_push %rbx
push %rbp
+.cfi_push %rbp
push %r12
+.cfi_push %r12
push %r13
+.cfi_push %r13
push %r14
+.cfi_push %r14
push %r15
- mov %rsp,%r11 # copy %rsp
+.cfi_push %r15
shl \$4,%rdx # num*16
sub \$`$framesz+$win64*16*4`,%rsp
lea ($inp,%rdx,$SZ),%rdx # inp+num*16*$SZ
mov $ctx,$_ctx # save ctx, 1st arg
mov $inp,$_inp # save inp, 2nd arh
mov %rdx,$_end # save end pointer, "3rd" arg
- mov %r11,$_rsp # save copy of %rsp
+ mov %rax,$_rsp # save copy of %rsp
+.cfi_cfa_expression $_rsp,deref,+8
___
$code.=<<___ if ($win64);
movaps %xmm6,16*$SZ+32(%rsp)
jb .Lloop_ssse3
mov $_rsp,%rsi
+.cfi_def_cfa %rsi,8
___
$code.=<<___ if ($win64);
movaps 16*$SZ+32(%rsp),%xmm6
movaps 16*$SZ+80(%rsp),%xmm9
___
$code.=<<___;
- mov (%rsi),%r15
- mov 8(%rsi),%r14
- mov 16(%rsi),%r13
- mov 24(%rsi),%r12
- mov 32(%rsi),%rbp
- mov 40(%rsi),%rbx
- lea 48(%rsi),%rsp
+ mov -48(%rsi),%r15
+.cfi_restore %r15
+ mov -40(%rsi),%r14
+.cfi_restore %r14
+ mov -32(%rsi),%r13
+.cfi_restore %r13
+ mov -24(%rsi),%r12
+.cfi_restore %r12
+ mov -16(%rsi),%rbp
+.cfi_restore %rbp
+ mov -8(%rsi),%rbx
+.cfi_restore %rbx
+ lea (%rsi),%rsp
+.cfi_def_cfa_register %rsp
.Lepilogue_ssse3:
ret
+.cfi_endproc
.size ${func}_ssse3,.-${func}_ssse3
___
}
.type ${func}_xop,\@function,3
.align 64
${func}_xop:
+.cfi_startproc
.Lxop_shortcut:
+ mov %rsp,%rax # copy %rsp
+.cfi_def_cfa_register %rax
push %rbx
+.cfi_push %rbx
push %rbp
+.cfi_push %rbp
push %r12
+.cfi_push %r12
push %r13
+.cfi_push %r13
push %r14
+.cfi_push %r14
push %r15
- mov %rsp,%r11 # copy %rsp
+.cfi_push %r15
shl \$4,%rdx # num*16
sub \$`$framesz+$win64*16*($SZ==4?4:6)`,%rsp
lea ($inp,%rdx,$SZ),%rdx # inp+num*16*$SZ
mov $ctx,$_ctx # save ctx, 1st arg
mov $inp,$_inp # save inp, 2nd arh
mov %rdx,$_end # save end pointer, "3rd" arg
- mov %r11,$_rsp # save copy of %rsp
+ mov %rax,$_rsp # save copy of %rsp
+.cfi_cfa_expression $_rsp,deref,+8
___
$code.=<<___ if ($win64);
movaps %xmm6,16*$SZ+32(%rsp)
jb .Lloop_xop
mov $_rsp,%rsi
+.cfi_def_cfa %rsi,8
vzeroupper
___
$code.=<<___ if ($win64);
movaps 16*$SZ+112(%rsp),%xmm11
___
$code.=<<___;
- mov (%rsi),%r15
- mov 8(%rsi),%r14
- mov 16(%rsi),%r13
- mov 24(%rsi),%r12
- mov 32(%rsi),%rbp
- mov 40(%rsi),%rbx
- lea 48(%rsi),%rsp
+ mov -48(%rsi),%r15
+.cfi_restore %r15
+ mov -40(%rsi),%r14
+.cfi_restore %r14
+ mov -32(%rsi),%r13
+.cfi_restore %r13
+ mov -24(%rsi),%r12
+.cfi_restore %r12
+ mov -16(%rsi),%rbp
+.cfi_restore %rbp
+ mov -8(%rsi),%rbx
+.cfi_restore %rbx
+ lea (%rsi),%rsp
+.cfi_def_cfa_register %rsp
.Lepilogue_xop:
ret
+.cfi_endproc
.size ${func}_xop,.-${func}_xop
___
}
.type ${func}_avx,\@function,3
.align 64
${func}_avx:
+.cfi_startproc
.Lavx_shortcut:
+ mov %rsp,%rax # copy %rsp
+.cfi_def_cfa_register %rax
push %rbx
+.cfi_push %rbx
push %rbp
+.cfi_push %rbp
push %r12
+.cfi_push %r12
push %r13
+.cfi_push %r13
push %r14
+.cfi_push %r14
push %r15
- mov %rsp,%r11 # copy %rsp
+.cfi_push %r15
shl \$4,%rdx # num*16
sub \$`$framesz+$win64*16*($SZ==4?4:6)`,%rsp
lea ($inp,%rdx,$SZ),%rdx # inp+num*16*$SZ
mov $ctx,$_ctx # save ctx, 1st arg
mov $inp,$_inp # save inp, 2nd arh
mov %rdx,$_end # save end pointer, "3rd" arg
- mov %r11,$_rsp # save copy of %rsp
+ mov %rax,$_rsp # save copy of %rsp
+.cfi_cfa_expression $_rsp,deref,+8
___
$code.=<<___ if ($win64);
movaps %xmm6,16*$SZ+32(%rsp)
jb .Lloop_avx
mov $_rsp,%rsi
+.cfi_def_cfa %rsi,8
vzeroupper
___
$code.=<<___ if ($win64);
movaps 16*$SZ+112(%rsp),%xmm11
___
$code.=<<___;
- mov (%rsi),%r15
- mov 8(%rsi),%r14
- mov 16(%rsi),%r13
- mov 24(%rsi),%r12
- mov 32(%rsi),%rbp
- mov 40(%rsi),%rbx
- lea 48(%rsi),%rsp
+ mov -48(%rsi),%r15
+.cfi_restore %r15
+ mov -40(%rsi),%r14
+.cfi_restore %r14
+ mov -32(%rsi),%r13
+.cfi_restore %r13
+ mov -24(%rsi),%r12
+.cfi_restore %r12
+ mov -16(%rsi),%rbp
+.cfi_restore %rbp
+ mov -8(%rsi),%rbx
+.cfi_restore %rbx
+ lea (%rsi),%rsp
+.cfi_def_cfa_register %rsp
.Lepilogue_avx:
ret
+.cfi_endproc
.size ${func}_avx,.-${func}_avx
___
######################################################################
# AVX2+BMI code path
#
-my $a5=$SZ==4?"%esi":"%rsi"; # zap $inp
+my $a5=$SZ==4?"%esi":"%rsi"; # zap $inp
my $PUSH8=8*2*$SZ;
use integer;
.type ${func}_avx2,\@function,3
.align 64
${func}_avx2:
+.cfi_startproc
.Lavx2_shortcut:
+ mov %rsp,%rax # copy %rsp
+.cfi_def_cfa_register %rax
push %rbx
+.cfi_push %rbx
push %rbp
+.cfi_push %rbp
push %r12
+.cfi_push %r12
push %r13
+.cfi_push %r13
push %r14
+.cfi_push %r14
push %r15
- mov %rsp,%r11 # copy %rsp
+.cfi_push %r15
sub \$`2*$SZ*$rounds+4*8+$win64*16*($SZ==4?4:6)`,%rsp
shl \$4,%rdx # num*16
and \$-256*$SZ,%rsp # align stack frame
mov $ctx,$_ctx # save ctx, 1st arg
mov $inp,$_inp # save inp, 2nd arh
mov %rdx,$_end # save end pointer, "3rd" arg
- mov %r11,$_rsp # save copy of %rsp
+ mov %rax,$_rsp # save copy of %rsp
+.cfi_cfa_expression $_rsp,deref,+8
___
$code.=<<___ if ($win64);
movaps %xmm6,16*$SZ+32(%rsp)
.Ldone_avx2:
lea ($Tbl),%rsp
mov $_rsp,%rsi
+.cfi_def_cfa %rsi,8
vzeroupper
___
$code.=<<___ if ($win64);
movaps 16*$SZ+112(%rsp),%xmm11
___
$code.=<<___;
- mov (%rsi),%r15
- mov 8(%rsi),%r14
- mov 16(%rsi),%r13
- mov 24(%rsi),%r12
- mov 32(%rsi),%rbp
- mov 40(%rsi),%rbx
- lea 48(%rsi),%rsp
+ mov -48(%rsi),%r15
+.cfi_restore %r15
+ mov -40(%rsi),%r14
+.cfi_restore %r14
+ mov -32(%rsi),%r13
+.cfi_restore %r13
+ mov -24(%rsi),%r12
+.cfi_restore %r12
+ mov -16(%rsi),%rbp
+.cfi_restore %rbp
+ mov -8(%rsi),%rbx
+.cfi_restore %rbx
+ lea (%rsi),%rsp
+.cfi_def_cfa_register %rsp
.Lepilogue_avx2:
ret
+.cfi_endproc
.size ${func}_avx2,.-${func}_avx2
___
}}
$code.=<<___;
mov %rax,%rsi # put aside Rsp
mov 16*$SZ+3*8(%rax),%rax # pull $_rsp
- lea 48(%rax),%rax
mov -8(%rax),%rbx
mov -16(%rax),%rbp