The branch master has been updated
via 6d3f798cba8075e700003aaf34f1e72bb930086c (commit)
from b938544969577e3b74da6f8c689c87c90ceced22 (commit)
- Log -----------------------------------------------------------------
commit 6d3f798cba8075e700003aaf34f1e72bb930086c
Author: Jean-Philippe Boivin <jpboi...@me.com>
Date: Mon May 17 16:38:14 2021 -0400
Properly restore XMM registers in ChaCha20's AVX-512(VL) assembly
Reviewed-by: Paul Dale <pa...@openssl.org>
Reviewed-by: Tomas Mraz <to...@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15315)
-----------------------------------------------------------------------
Summary of changes:
crypto/chacha/asm/chacha-x86_64.pl | 52 ++++++++++++++++++++++++++++++--------
1 file changed, 42 insertions(+), 10 deletions(-)
diff --git a/crypto/chacha/asm/chacha-x86_64.pl
b/crypto/chacha/asm/chacha-x86_64.pl
index 964c926fde..45f0084767 100755
--- a/crypto/chacha/asm/chacha-x86_64.pl
+++ b/crypto/chacha/asm/chacha-x86_64.pl
@@ -473,7 +473,7 @@ sub SSSE3ROUND { # critical path is 20 "SIMD ticks" per
round
&por ($b,$t);
}
-my $xframe = $win64 ? 32+8 : 8;
+my $xframe = $win64 ? 160+8 : 8;
$code.=<<___;
.type ChaCha20_ssse3,\@function,5
@@ -2501,7 +2501,7 @@ sub AVX512ROUND { # critical path is 14 "SIMD ticks" per
round
&vprold ($b,$b,7);
}
-my $xframe = $win64 ? 32+8 : 8;
+my $xframe = $win64 ? 160+8 : 8;
$code.=<<___;
.type ChaCha20_avx512,\@function,5
@@ -2517,8 +2517,16 @@ ChaCha20_avx512:
sub \$64+$xframe,%rsp
___
$code.=<<___ if ($win64);
- movaps %xmm6,-0x28(%r9)
- movaps %xmm7,-0x18(%r9)
+ movaps %xmm6,-0xa8(%r9)
+ movaps %xmm7,-0x98(%r9)
+ movaps %xmm8,-0x88(%r9)
+ movaps %xmm9,-0x78(%r9)
+ movaps %xmm10,-0x68(%r9)
+ movaps %xmm11,-0x58(%r9)
+ movaps %xmm12,-0x48(%r9)
+ movaps %xmm13,-0x38(%r9)
+ movaps %xmm14,-0x28(%r9)
+ movaps %xmm15,-0x18(%r9)
.Lavx512_body:
___
$code.=<<___;
@@ -2685,8 +2693,16 @@ $code.=<<___;
vzeroall
___
$code.=<<___ if ($win64);
- movaps -0x28(%r9),%xmm6
- movaps -0x18(%r9),%xmm7
+ movaps -0xa8(%r9),%xmm6
+ movaps -0x98(%r9),%xmm7
+ movaps -0x88(%r9),%xmm8
+ movaps -0x78(%r9),%xmm9
+ movaps -0x68(%r9),%xmm10
+ movaps -0x58(%r9),%xmm11
+ movaps -0x48(%r9),%xmm12
+ movaps -0x38(%r9),%xmm13
+ movaps -0x28(%r9),%xmm14
+ movaps -0x18(%r9),%xmm15
___
$code.=<<___;
lea (%r9),%rsp
@@ -2713,8 +2729,16 @@ ChaCha20_avx512vl:
sub \$64+$xframe,%rsp
___
$code.=<<___ if ($win64);
- movaps %xmm6,-0x28(%r9)
- movaps %xmm7,-0x18(%r9)
+ movaps %xmm6,-0xa8(%r9)
+ movaps %xmm7,-0x98(%r9)
+ movaps %xmm8,-0x88(%r9)
+ movaps %xmm9,-0x78(%r9)
+ movaps %xmm10,-0x68(%r9)
+ movaps %xmm11,-0x58(%r9)
+ movaps %xmm12,-0x48(%r9)
+ movaps %xmm13,-0x38(%r9)
+ movaps %xmm14,-0x28(%r9)
+ movaps %xmm15,-0x18(%r9)
.Lavx512vl_body:
___
$code.=<<___;
@@ -2838,8 +2862,16 @@ $code.=<<___;
vzeroall
___
$code.=<<___ if ($win64);
- movaps -0x28(%r9),%xmm6
- movaps -0x18(%r9),%xmm7
+ movaps -0xa8(%r9),%xmm6
+ movaps -0x98(%r9),%xmm7
+ movaps -0x88(%r9),%xmm8
+ movaps -0x78(%r9),%xmm9
+ movaps -0x68(%r9),%xmm10
+ movaps -0x58(%r9),%xmm11
+ movaps -0x48(%r9),%xmm12
+ movaps -0x38(%r9),%xmm13
+ movaps -0x28(%r9),%xmm14
+ movaps -0x18(%r9),%xmm15
___
$code.=<<___;
lea (%r9),%rsp
