Sync up some perlasm license headers and easy fixes.
These files are otherwise up-to-date with OpenSSL master as of 50ea9d2b3521467a11559be41dcf05ee05feabd6, modulo a couple of spelling fixes which I've imported. I've also reverted the same-line label and instruction patch to x86_64-mont*.pl. The new delocate parser handles that fine. Change-Id: Ife35c671a8104c3cc2fb6c5a03127376fccc4402 Reviewed-on: https://boringssl-review.googlesource.com/25644 Reviewed-by: Adam Langley <agl@google.com>
This commit is contained in:
David Benjamin
Adam Langley
parent
0f4f6c2e02
commit
6dc994265e
@ -28,6 +28,7 @@
|
||||
# Denver 4.50/+82% 2.63 2.67(*)
|
||||
# X-Gene 9.50/+46% 8.82 8.89(*)
|
||||
# Mongoose 8.00/+44% 3.64 3.25
|
||||
# Kryo 8.17/+50% 4.83 4.65
|
||||
#
|
||||
# (*) it's expected that doubling interleave factor doesn't help
|
||||
# all processors, only those with higher NEON latency and
|
||||
|
||||
@ -1,4 +1,11 @@
|
||||
#!/usr/bin/env perl
|
||||
#! /usr/bin/env perl
|
||||
# Copyright 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
|
||||
|
||||
@ -1,7 +1,14 @@
|
||||
#!/usr/bin/env perl
|
||||
#! /usr/bin/env perl
|
||||
# Copyright 2004-2016 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
|
||||
# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
|
||||
# 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/.
|
||||
@ -32,7 +39,7 @@
|
||||
# for scaling too, I [try to] avoid the latter by favoring off-by-2
|
||||
# shifts and masking the result with 0xFF<<2 instead of "boring" 0xFF.
|
||||
#
|
||||
# As was shown by Dean Gaudet <dean@arctic.org>, the above note turned
|
||||
# As was shown by Dean Gaudet, the above note turned out to be
|
||||
# void. Performance improvement with off-by-2 shifts was observed on
|
||||
# intermediate implementation, which was spilling yet another register
|
||||
# to stack... Final offset*4 code below runs just a tad faster on P4,
|
||||
@ -48,8 +55,8 @@
|
||||
# better performance on most recent µ-archs...
|
||||
#
|
||||
# Third version adds AES_cbc_encrypt implementation, which resulted in
|
||||
# up to 40% performance imrovement of CBC benchmark results. 40% was
|
||||
# observed on P4 core, where "overall" imrovement coefficient, i.e. if
|
||||
# up to 40% performance improvement of CBC benchmark results. 40% was
|
||||
# observed on P4 core, where "overall" improvement coefficient, i.e. if
|
||||
# compared to PIC generated by GCC and in CBC mode, was observed to be
|
||||
# as large as 4x:-) CBC performance is virtually identical to ECB now
|
||||
# and on some platforms even better, e.g. 17.6 "small" cycles/byte on
|
||||
@ -152,7 +159,7 @@
|
||||
# combinations then attack becomes infeasible. This is why revised
|
||||
# AES_cbc_encrypt "dares" to switch to larger S-box when larger chunk
|
||||
# of data is to be processed in one stroke. The current size limit of
|
||||
# 512 bytes is chosen to provide same [diminishigly low] probability
|
||||
# 512 bytes is chosen to provide same [diminishingly low] probability
|
||||
# for cache-line to remain untouched in large chunk operation with
|
||||
# large S-box as for single block operation with compact S-box and
|
||||
# surely needs more careful consideration...
|
||||
@ -164,12 +171,12 @@
|
||||
# yield execution to process performing AES just before timer fires
|
||||
# off the scheduler, immediately regain control of CPU and analyze the
|
||||
# cache state. For this attack to be efficient attacker would have to
|
||||
# effectively slow down the operation by several *orders* of magnitute,
|
||||
# effectively slow down the operation by several *orders* of magnitude,
|
||||
# by ratio of time slice to duration of handful of AES rounds, which
|
||||
# unlikely to remain unnoticed. Not to mention that this also means
|
||||
# that he would spend correspondigly more time to collect enough
|
||||
# that he would spend correspondingly more time to collect enough
|
||||
# statistical data to mount the attack. It's probably appropriate to
|
||||
# say that if adeversary reckons that this attack is beneficial and
|
||||
# say that if adversary reckons that this attack is beneficial and
|
||||
# risks to be noticed, you probably have larger problems having him
|
||||
# mere opportunity. In other words suggested code design expects you
|
||||
# to preclude/mitigate this attack by overall system security design.
|
||||
@ -233,7 +240,7 @@ $small_footprint=1; # $small_footprint=1 code is ~5% slower [on
|
||||
# contention and in hope to "collect" 5% back
|
||||
# in real-life applications...
|
||||
|
||||
$vertical_spin=0; # shift "verticaly" defaults to 0, because of
|
||||
$vertical_spin=0; # shift "vertically" defaults to 0, because of
|
||||
# its proof-of-concept status...
|
||||
# Note that there is no decvert(), as well as last encryption round is
|
||||
# performed with "horizontal" shifts. This is because this "vertical"
|
||||
@ -1599,7 +1606,7 @@ sub decstep()
|
||||
# no instructions are reordered, as performance appears
|
||||
# optimal... or rather that all attempts to reorder didn't
|
||||
# result in better performance [which by the way is not a
|
||||
# bit lower than ecryption].
|
||||
# bit lower than encryption].
|
||||
if($i==3) { &mov ($key,$__key); }
|
||||
else { &mov ($out,$s[0]); }
|
||||
&and ($out,0xFF);
|
||||
|
||||
@ -1,4 +1,11 @@
|
||||
#!/usr/bin/env perl
|
||||
#! /usr/bin/env perl
|
||||
# Copyright 2009-2016 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
|
||||
@ -43,18 +50,22 @@
|
||||
# Add aesni_xts_[en|de]crypt. Westmere spends 1.50 cycles processing
|
||||
# one byte out of 8KB with 128-bit key, Sandy Bridge - 1.09.
|
||||
|
||||
# November 2015
|
||||
#
|
||||
# Add aesni_ocb_[en|de]crypt. [Removed in BoringSSL]
|
||||
|
||||
######################################################################
|
||||
# Current large-block performance in cycles per byte processed with
|
||||
# 128-bit key (less is better).
|
||||
#
|
||||
# CBC en-/decrypt CTR XTS ECB
|
||||
# CBC en-/decrypt CTR XTS ECB OCB
|
||||
# Westmere 3.77/1.37 1.37 1.52 1.27
|
||||
# * Bridge 5.07/0.98 0.99 1.09 0.91
|
||||
# Haswell 4.44/0.80 0.97 1.03 0.72
|
||||
# Skylake 2.68/0.65 0.65 0.66 0.64
|
||||
# Silvermont 5.77/3.56 3.67 4.03 3.46
|
||||
# Goldmont 3.84/1.39 1.39 1.63 1.31
|
||||
# Bulldozer 5.80/0.98 1.05 1.24 0.93
|
||||
# * Bridge 5.07/0.98 0.99 1.09 0.91 1.10
|
||||
# Haswell 4.44/0.80 0.97 1.03 0.72 0.76
|
||||
# Skylake 2.68/0.65 0.65 0.66 0.64 0.66
|
||||
# Silvermont 5.77/3.56 3.67 4.03 3.46 4.03
|
||||
# Goldmont 3.84/1.39 1.39 1.63 1.31 1.70
|
||||
# Bulldozer 5.80/0.98 1.05 1.24 0.93 1.23
|
||||
|
||||
$PREFIX="aesni"; # if $PREFIX is set to "AES", the script
|
||||
# generates drop-in replacement for
|
||||
@ -228,7 +239,7 @@ sub aesni_generate1 # fully unrolled loop
|
||||
# can schedule aes[enc|dec] every cycle optimal interleave factor
|
||||
# equals to corresponding instructions latency. 8x is optimal for
|
||||
# * Bridge, but it's unfeasible to accommodate such implementation
|
||||
# in XMM registers addreassable in 32-bit mode and therefore maximum
|
||||
# in XMM registers addressable in 32-bit mode and therefore maximum
|
||||
# of 6x is used instead...
|
||||
|
||||
sub aesni_generate2
|
||||
@ -2425,7 +2436,7 @@ if ($PREFIX eq "aesni") {
|
||||
&pxor ("xmm3","xmm3");
|
||||
&aesenclast ("xmm2","xmm3");
|
||||
|
||||
&movdqa ("xmm3","xmm1")
|
||||
&movdqa ("xmm3","xmm1");
|
||||
&pslldq ("xmm1",4);
|
||||
&pxor ("xmm3","xmm1");
|
||||
&pslldq ("xmm1",4);
|
||||
|
||||
@ -14,8 +14,7 @@
|
||||
# details see http://www.openssl.org/~appro/cryptogams/.
|
||||
#
|
||||
# Specific modes and adaptation for Linux kernel by Ard Biesheuvel
|
||||
# <ard.biesheuvel@linaro.org>. Permission to use under GPL terms is
|
||||
# granted.
|
||||
# of Linaro. Permission to use under GPL terms is granted.
|
||||
# ====================================================================
|
||||
|
||||
# Bit-sliced AES for ARM NEON
|
||||
@ -49,10 +48,7 @@
|
||||
# <appro@openssl.org>
|
||||
|
||||
# April-August 2013
|
||||
#
|
||||
# Add CBC, CTR and XTS subroutines, adapt for kernel use.
|
||||
#
|
||||
# <ard.biesheuvel@linaro.org>
|
||||
# Add CBC, CTR and XTS subroutines and adapt for kernel use; courtesy of Ard.
|
||||
|
||||
$flavour = shift;
|
||||
if ($flavour=~/\w[\w\-]*\.\w+$/) { $output=$flavour; undef $flavour; }
|
||||
|
||||
@ -1,4 +1,11 @@
|
||||
#!/usr/bin/env perl
|
||||
#! /usr/bin/env perl
|
||||
# Copyright 2011-2016 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
|
||||
|
||||
|
||||
######################################################################
|
||||
## Constant-time SSSE3 AES core implementation.
|
||||
|
||||
@ -1,4 +1,11 @@
|
||||
#!/usr/bin/env perl
|
||||
#! /usr/bin/env perl
|
||||
# Copyright 2011-2016 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
|
||||
|
||||
|
||||
######################################################################
|
||||
## Constant-time SSSE3 AES core implementation.
|
||||
|
||||
@ -1,4 +1,11 @@
|
||||
#!/usr/bin/env perl
|
||||
#! /usr/bin/env perl
|
||||
# Copyright 1995-2016 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
|
||||
|
||||
|
||||
$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
|
||||
push(@INC,"${dir}","${dir}../../../perlasm");
|
||||
|
||||
@ -1,4 +1,10 @@
|
||||
#!/usr/local/bin/perl
|
||||
#! /usr/bin/env perl
|
||||
# Copyright 1995-2016 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
|
||||
|
||||
$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
|
||||
push(@INC,"${dir}","${dir}../../../perlasm");
|
||||
|
||||
@ -1,7 +1,14 @@
|
||||
#!/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@fy.chalmers.se> for the OpenSSL
|
||||
# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
|
||||
# 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/.
|
||||
@ -71,7 +78,7 @@ $frame=32; # size of above frame rounded up to 16n
|
||||
&lea ("ebp",&DWP(-$frame,"esp","edi",4)); # future alloca($frame+4*(num+2))
|
||||
&neg ("edi");
|
||||
|
||||
# minimize cache contention by arraning 2K window between stack
|
||||
# minimize cache contention by arranging 2K window between stack
|
||||
# pointer and ap argument [np is also position sensitive vector,
|
||||
# but it's assumed to be near ap, as it's allocated at ~same
|
||||
# time].
|
||||
|
||||
@ -1,4 +1,11 @@
|
||||
#!/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
|
||||
@ -296,12 +303,11 @@ $code.=<<___;
|
||||
mov $num,$j # j=num
|
||||
jmp .Lsub
|
||||
.align 16
|
||||
.Lsub:
|
||||
sbb ($np,$i,8),%rax
|
||||
.Lsub: sbb ($np,$i,8),%rax
|
||||
mov %rax,($rp,$i,8) # rp[i]=tp[i]-np[i]
|
||||
mov 8($ap,$i,8),%rax # tp[i+1]
|
||||
lea 1($i),$i # i++
|
||||
dec $j # doesnn't affect CF!
|
||||
dec $j # doesn't affect CF!
|
||||
jnz .Lsub
|
||||
|
||||
sbb \$0,%rax # handle upmost overflow bit
|
||||
@ -732,7 +738,7 @@ $code.=<<___;
|
||||
mov 56($ap,$i,8),@ri[3]
|
||||
sbb 40($np,$i,8),@ri[1]
|
||||
lea 4($i),$i # i++
|
||||
dec $j # doesnn't affect CF!
|
||||
dec $j # doesn't affect CF!
|
||||
jnz .Lsub4x
|
||||
|
||||
mov @ri[0],0($rp,$i,8) # rp[i]=tp[i]-np[i]
|
||||
|
||||
@ -1,4 +1,11 @@
|
||||
#!/usr/bin/env perl
|
||||
#! /usr/bin/env perl
|
||||
# Copyright 2011-2016 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
|
||||
@ -396,12 +403,11 @@ $code.=<<___;
|
||||
mov $num,$j # j=num
|
||||
jmp .Lsub
|
||||
.align 16
|
||||
.Lsub:
|
||||
sbb ($np,$i,8),%rax
|
||||
.Lsub: sbb ($np,$i,8),%rax
|
||||
mov %rax,($rp,$i,8) # rp[i]=tp[i]-np[i]
|
||||
mov 8($ap,$i,8),%rax # tp[i+1]
|
||||
lea 1($i),$i # i++
|
||||
dec $j # doesnn't affect CF!
|
||||
dec $j # doesn't affect CF!
|
||||
jnz .Lsub
|
||||
|
||||
sbb \$0,%rax # handle upmost overflow bit
|
||||
@ -2405,7 +2411,7 @@ my $N=$STRIDE/4; # should match cache line size
|
||||
$code.=<<___;
|
||||
movdqa 0(%rax),%xmm0 # 00000001000000010000000000000000
|
||||
movdqa 16(%rax),%xmm1 # 00000002000000020000000200000002
|
||||
lea 88-112(%rsp,%r10),%r10 # place the mask after tp[num+1] (+ICache optimizaton)
|
||||
lea 88-112(%rsp,%r10),%r10 # place the mask after tp[num+1] (+ICache optimization)
|
||||
lea 128($bp),$bptr # size optimization
|
||||
|
||||
pshufd \$0,%xmm5,%xmm5 # broadcast index
|
||||
|
||||
@ -1,4 +1,11 @@
|
||||
#!/usr/bin/env perl
|
||||
#! /usr/bin/env perl
|
||||
# Copyright 2010-2016 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
|
||||
@ -96,14 +103,13 @@
|
||||
#
|
||||
# Does it make sense to increase Naggr? To start with it's virtually
|
||||
# impossible in 32-bit mode, because of limited register bank
|
||||
# capacity. Otherwise improvement has to be weighed agiainst slower
|
||||
# capacity. Otherwise improvement has to be weighed against slower
|
||||
# setup, as well as code size and complexity increase. As even
|
||||
# optimistic estimate doesn't promise 30% performance improvement,
|
||||
# there are currently no plans to increase Naggr.
|
||||
#
|
||||
# Special thanks to David Woodhouse <dwmw2@infradead.org> for
|
||||
# providing access to a Westmere-based system on behalf of Intel
|
||||
# Open Source Technology Centre.
|
||||
# Special thanks to David Woodhouse for providing access to a
|
||||
# Westmere-based system on behalf of Intel Open Source Technology Centre.
|
||||
|
||||
# January 2010
|
||||
#
|
||||
|
||||
@ -1,4 +1,11 @@
|
||||
#!/usr/bin/env perl
|
||||
#! /usr/bin/env perl
|
||||
# Copyright 1998-2016 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
|
||||
|
||||
|
||||
# ====================================================================
|
||||
# [Re]written by Andy Polyakov <appro@openssl.org> for the OpenSSL
|
||||
@ -28,10 +35,9 @@
|
||||
# P4 +85%(!) +45%
|
||||
#
|
||||
# As you can see Pentium came out as looser:-( Yet I reckoned that
|
||||
# improvement on P4 outweights the loss and incorporate this
|
||||
# improvement on P4 outweighs the loss and incorporate this
|
||||
# re-tuned code to 0.9.7 and later.
|
||||
# ----------------------------------------------------------------
|
||||
# <appro@fy.chalmers.se>
|
||||
|
||||
# August 2009.
|
||||
#
|
||||
@ -538,7 +544,7 @@ for($i=0;$i<20-4;$i+=2) {
|
||||
# being implemented in SSSE3). Once 8 quadruples or 32 elements are
|
||||
# collected, it switches to routine proposed by Max Locktyukhin.
|
||||
#
|
||||
# Calculations inevitably require temporary reqisters, and there are
|
||||
# Calculations inevitably require temporary registers, and there are
|
||||
# no %xmm registers left to spare. For this reason part of the ring
|
||||
# buffer, X[2..4] to be specific, is offloaded to 3 quadriples ring
|
||||
# buffer on the stack. Keep in mind that X[2] is alias X[-6], X[3] -
|
||||
@ -649,7 +655,7 @@ my $_ror=sub { &ror(@_) };
|
||||
&jmp (&label("loop"));
|
||||
|
||||
######################################################################
|
||||
# SSE instruction sequence is first broken to groups of indepentent
|
||||
# SSE instruction sequence is first broken to groups of independent
|
||||
# instructions, independent in respect to their inputs and shifter
|
||||
# (not all architectures have more than one). Then IALU instructions
|
||||
# are "knitted in" between the SSE groups. Distance is maintained for
|
||||
@ -658,14 +664,14 @@ my $_ror=sub { &ror(@_) };
|
||||
#
|
||||
# Temporary registers usage. X[2] is volatile at the entry and at the
|
||||
# end is restored from backtrace ring buffer. X[3] is expected to
|
||||
# contain current K_XX_XX constant and is used to caclulate X[-1]+K
|
||||
# contain current K_XX_XX constant and is used to calculate X[-1]+K
|
||||
# from previous round, it becomes volatile the moment the value is
|
||||
# saved to stack for transfer to IALU. X[4] becomes volatile whenever
|
||||
# X[-4] is accumulated and offloaded to backtrace ring buffer, at the
|
||||
# end it is loaded with next K_XX_XX [which becomes X[3] in next
|
||||
# round]...
|
||||
#
|
||||
sub Xupdate_ssse3_16_31() # recall that $Xi starts wtih 4
|
||||
sub Xupdate_ssse3_16_31() # recall that $Xi starts with 4
|
||||
{ use integer;
|
||||
my $body = shift;
|
||||
my @insns = (&$body,&$body,&$body,&$body); # 40 instructions
|
||||
@ -1188,7 +1194,7 @@ my $_ror=sub { &shrd(@_[0],@_) };
|
||||
&and (@T[0],@T[1]);
|
||||
&jmp (&label("loop"));
|
||||
|
||||
sub Xupdate_avx_16_31() # recall that $Xi starts wtih 4
|
||||
sub Xupdate_avx_16_31() # recall that $Xi starts with 4
|
||||
{ use integer;
|
||||
my $body = shift;
|
||||
my @insns = (&$body,&$body,&$body,&$body); # 40 instructions
|
||||
|
||||
@ -26,6 +26,7 @@
|
||||
# Denver 2.13 3.97 (+0%)(**)
|
||||
# X-Gene 8.80 (+200%)
|
||||
# Mongoose 2.05 6.50 (+160%)
|
||||
# Kryo 1.88 8.00 (+90%)
|
||||
#
|
||||
# (*) Software results are presented mostly for reference purposes.
|
||||
# (**) Keep in mind that Denver relies on binary translation, which
|
||||
|
||||
@ -1,4 +1,11 @@
|
||||
#!/usr/bin/env perl
|
||||
#! /usr/bin/env perl
|
||||
# Copyright 2007-2016 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
|
||||
@ -11,7 +18,7 @@
|
||||
#
|
||||
# Performance improvement over compiler generated code varies from
|
||||
# 10% to 40% [see below]. Not very impressive on some µ-archs, but
|
||||
# it's 5 times smaller and optimizies amount of writes.
|
||||
# it's 5 times smaller and optimizes amount of writes.
|
||||
#
|
||||
# May 2012.
|
||||
#
|
||||
|
||||
@ -1,4 +1,11 @@
|
||||
#!/usr/bin/env perl
|
||||
#! /usr/bin/env perl
|
||||
# Copyright 2007-2016 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
|
||||
@ -35,7 +42,7 @@
|
||||
# (*) whichever best applicable.
|
||||
# (**) x86_64 assembler performance is presented for reference
|
||||
# purposes, the results are for integer-only code.
|
||||
# (***) paddq is increadibly slow on Atom.
|
||||
# (***) paddq is incredibly slow on Atom.
|
||||
#
|
||||
# IALU code-path is optimized for elder Pentiums. On vanilla Pentium
|
||||
# performance improvement over compiler generated code reaches ~60%,
|
||||
|
||||
@ -526,7 +526,7 @@ my %globals;
|
||||
);
|
||||
|
||||
# Following constants are defined in x86_64 ABI supplement, for
|
||||
# example avaiable at https://www.uclibc.org/docs/psABI-x86_64.pdf,
|
||||
# example available at https://www.uclibc.org/docs/psABI-x86_64.pdf,
|
||||
# see section 3.7 "Stack Unwind Algorithm".
|
||||
my %DW_reg_idx = (
|
||||
"%rax"=>0, "%rdx"=>1, "%rcx"=>2, "%rbx"=>3,
|
||||
@ -539,7 +539,7 @@ my %globals;
|
||||
|
||||
# [us]leb128 format is variable-length integer representation base
|
||||
# 2^128, with most significant bit of each byte being 0 denoting
|
||||
# *last* most significat digit. See "Variable Length Data" in the
|
||||
# *last* most significant digit. See "Variable Length Data" in the
|
||||
# DWARF specification, numbered 7.6 at least in versions 3 and 4.
|
||||
sub sleb128 {
|
||||
use integer; # get right shift extend sign
|
||||
@ -1436,6 +1436,6 @@ close STDOUT;
|
||||
#
|
||||
# (*) Note that we're talking about run-time, not debug-time. Lack of
|
||||
# unwind information makes debugging hard on both Windows and
|
||||
# Unix. "Unlike" referes to the fact that on Unix signal handler
|
||||
# Unix. "Unlike" refers to the fact that on Unix signal handler
|
||||
# will always be invoked, core dumped and appropriate exit code
|
||||
# returned to parent (for user notification).
|
||||
|
||||
Loading…
Reference in New Issue
Block a user