diff options
author | Andy Polyakov <appro@openssl.org> | 2011-05-29 12:39:48 +0000 |
---|---|---|
committer | Andy Polyakov <appro@openssl.org> | 2011-05-29 12:39:48 +0000 |
commit | 0c149802a299ec7f448278ed90e53ac6c2dac88e (patch) | |
tree | 353c82d297d0bc06672d0a28fb534ce05b6870c7 /crypto/sha/asm | |
parent | c33066900ca7639797d8722ec9c7cc716ced5d0e (diff) |
sha1-586|x86_64.pl: add SSSE3 and AVX code paths.
Diffstat (limited to 'crypto/sha/asm')
-rw-r--r-- | crypto/sha/asm/sha1-586.pl | 1007 | ||||
-rwxr-xr-x | crypto/sha/asm/sha1-x86_64.pl | 919 |
2 files changed, 1915 insertions, 11 deletions
diff --git a/crypto/sha/asm/sha1-586.pl b/crypto/sha/asm/sha1-586.pl index dc855ef557..ac1c9e9ded 100644 --- a/crypto/sha/asm/sha1-586.pl +++ b/crypto/sha/asm/sha1-586.pl @@ -42,7 +42,61 @@ # degradation on other CPUs:-). Also, the code was revised to maximize # "distance" between instructions producing input to 'lea' instruction # and the 'lea' instruction itself, which is essential for Intel Atom -# core. +# core and resulted in ~15% improvement. + +# October 2010. +# +# Add SSSE3, Supplemental[!] SSE3, implementation. The idea behind it +# is to offload message schedule denoted by Wt in NIST specification, +# or Xupdate in OpenSSL source, to SIMD unit. The idea is not novel, +# and in SSE2 context was first explored by Dean Gaudet in 2004, see +# http://arctic.org/~dean/crypto/sha1.html. Since then several things +# have changed that made it interesting again: +# +# a) XMM units became faster and wider; +# b) instruction set became more versatile; +# c) an important observation was made by Max Locktykhin, which made +# it possible to reduce amount of instructions required to perform +# the operation in question, for further details see +# http://software.intel.com/en-us/articles/improving-the-performance-of-the-secure-hash-algorithm-1/. + +# April 2011. +# +# Add AVX code path, probably most controversial... The thing is that +# switch to AVX alone improves performance by as little as 4% in +# comparison to SSSE3 code path. But below result doesn't look like +# 4% improvement... Trouble is that Sandy Bridge decodes 'ro[rl]' as +# pair of µ-ops, and it's the additional µ-ops, two per round, that +# make it run slower than Core2 and Westmere. But 'sh[rl]d' is decoded +# as single µ-op by Sandy Bridge and it's replacing 'ro[rl]' with +# equivalent 'sh[rl]d' that is responsible for the impressive 5.1 +# cycles per processed byte. But 'sh[rl]d' is not something that used +# to be fast, nor does it appear to be fast in upcoming Bulldozer +# [according to its optimization manual]. Which is why AVX code path +# is guarded by *both* AVX and synthetic bit denoting Intel CPUs. +# One can argue that it's unfair to AMD, but without 'sh[rl]d' it +# makes no sense to keep the AVX code path. If somebody feels that +# strongly, it's probably more appropriate to discuss possibility of +# using vector rotate XOP on AMD... + +###################################################################### +# Current performance is summarized in following table. Numbers are +# CPU clock cycles spent to process single byte (less is better). +# +# x86 SSSE3 AVX +# Pentium 15.7 - +# PIII 11.5 - +# P4 10.6 - +# AMD K8 7.1 - +# Core2 7.3 6.1/+20% - +# Atom 12.5 9.5(*)/+32% - +# Westmere 7.3 5.6/+30% - +# Sandy Bridge 8.8 6.2/+40% 5.1(**)/+70% +# +# (*) Loop is 1056 instructions long and expected result is ~8.25. +# It remains mystery [to me] why ILP is limited to 1.7. +# +# (**) As per above comment, the result is for AVX *plus* sh[rl]d. $0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; push(@INC,"${dir}","${dir}../../perlasm"); @@ -50,6 +104,21 @@ require "x86asm.pl"; &asm_init($ARGV[0],"sha1-586.pl",$ARGV[$#ARGV] eq "386"); +$xmm=$ymm=0; +for (@ARGV) { $xmm=1 if (/-DOPENSSL_IA32_SSE2/); } + +$ymm=1 if ($xmm && + `$ENV{CC} -Wa,-V -c -o /dev/null -x assembler /dev/null 2>&1` + =~ /GNU assembler version ([2-9]\.[0-9]+)/ && + $1>=2.19); # first version supporting AVX + +$ymm=1 if ($xmm && !$ymm && $ARGV[0] eq "win32n" && + `nasm -v 2>&1` =~ /NASM version ([2-9]\.[0-9]+)/ && + $1>=2.03); # first version supporting AVX + +&external_label("OPENSSL_ia32cap_P") if ($xmm); + + $A="eax"; $B="ebx"; $C="ecx"; @@ -60,6 +129,10 @@ $tmp1="ebp"; @V=($A,$B,$C,$D,$E,$T); +$alt=0; # 1 denotes alternative IALU implementation, which performs + # 8% *worse* on P4, same on Westmere and Atom, 2% better on + # Sandy Bridge... + sub BODY_00_15 { local($n,$a,$b,$c,$d,$e,$f)=@_; @@ -83,6 +156,7 @@ sub BODY_00_15 if ($n==15) { &mov($e,&swtmp(($n+1)%16));# pre-fetch f for next round &add($f,$tmp1); } # f+=tmp1 else { &add($tmp1,$f); } # f becomes a in next round + &mov($tmp1,$a) if ($alt && $n==15); } sub BODY_16_19 @@ -91,6 +165,24 @@ sub BODY_16_19 &comment("16_19 $n"); +if ($alt) { + &xor($c,$d); + &xor($f,&swtmp(($n+2)%16)); # f to hold Xupdate(xi,xa,xb,xc,xd) + &and($tmp1,$c); # tmp1 to hold F_00_19(b,c,d), b&=c^d + &xor($f,&swtmp(($n+8)%16)); + &xor($tmp1,$d); # tmp1=F_00_19(b,c,d) + &xor($f,&swtmp(($n+13)%16)); # f holds xa^xb^xc^xd + &rotl($f,1); # f=ROTATE(f,1) + &add($e,$tmp1); # e+=F_00_19(b,c,d) + &xor($c,$d); # restore $c + &mov($tmp1,$a); # b in next round + &rotr($b,$n==16?2:7); # b=ROTATE(b,30) + &mov(&swtmp($n%16),$f); # xi=f + &rotl($a,5); # ROTATE(a,5) + &lea($f,&DWP(0x5a827999,$f,$e));# f+=F_00_19(b,c,d)+e + &mov($e,&swtmp(($n+1)%16)); # pre-fetch f for next round + &add($f,$a); # f+=ROTATE(a,5) +} else { &mov($tmp1,$c); # tmp1 to hold F_00_19(b,c,d) &xor($f,&swtmp(($n+2)%16)); # f to hold Xupdate(xi,xa,xb,xc,xd) &xor($tmp1,$d); @@ -107,6 +199,7 @@ sub BODY_16_19 &lea($f,&DWP(0x5a827999,$f,$e));# f+=F_00_19(b,c,d)+e &mov($e,&swtmp(($n+1)%16)); # pre-fetch f for next round &add($f,$tmp1); # f+=ROTATE(a,5) +} } sub BODY_20_39 @@ -116,6 +209,25 @@ sub BODY_20_39 &comment("20_39 $n"); +if ($alt) { + &xor($tmp1,$c); # tmp1 to hold F_20_39(b,c,d), b^=c + &xor($f,&swtmp(($n+2)%16)); # f to hold Xupdate(xi,xa,xb,xc,xd) + &xor($tmp1,$d); # tmp1 holds F_20_39(b,c,d) + &xor($f,&swtmp(($n+8)%16)); + &add($e,$tmp1); # e+=F_20_39(b,c,d) + &xor($f,&swtmp(($n+13)%16)); # f holds xa^xb^xc^xd + &rotl($f,1); # f=ROTATE(f,1) + &mov($tmp1,$a); # b in next round + &rotr($b,7); # b=ROTATE(b,30) + &mov(&swtmp($n%16),$f) if($n<77);# xi=f + &rotl($a,5); # ROTATE(a,5) + &xor($b,$c) if($n==39);# warm up for BODY_40_59 + &and($tmp1,$b) if($n==39); + &lea($f,&DWP($K,$f,$e)); # f+=e+K_XX_YY + &mov($e,&swtmp(($n+1)%16)) if($n<79);# pre-fetch f for next round + &add($f,$a); # f+=ROTATE(a,5) + &rotr($a,5) if ($n==79); +} else { &mov($tmp1,$b); # tmp1 to hold F_20_39(b,c,d) &xor($f,&swtmp(($n+2)%16)); # f to hold Xupdate(xi,xa,xb,xc,xd) &xor($tmp1,$c); @@ -131,6 +243,7 @@ sub BODY_20_39 &lea($f,&DWP($K,$f,$e)); # f+=e+K_XX_YY &mov($e,&swtmp(($n+1)%16)) if($n<79);# pre-fetch f for next round &add($f,$tmp1); # f+=ROTATE(a,5) +} } sub BODY_40_59 @@ -139,6 +252,26 @@ sub BODY_40_59 &comment("40_59 $n"); +if ($alt) { + &add($e,$tmp1); # e+=b&(c^d) + &xor($f,&swtmp(($n+2)%16)); # f to hold Xupdate(xi,xa,xb,xc,xd) + &mov($tmp1,$d); + &xor($f,&swtmp(($n+8)%16)); + &xor($c,$d); # restore $c + &xor($f,&swtmp(($n+13)%16)); # f holds xa^xb^xc^xd + &rotl($f,1); # f=ROTATE(f,1) + &and($tmp1,$c); + &rotr($b,7); # b=ROTATE(b,30) + &add($e,$tmp1); # e+=c&d + &mov($tmp1,$a); # b in next round + &mov(&swtmp($n%16),$f); # xi=f + &rotl($a,5); # ROTATE(a,5) + &xor($b,$c) if ($n<59); + &and($tmp1,$b) if ($n<59);# tmp1 to hold F_40_59(b,c,d) + &lea($f,&DWP(0x8f1bbcdc,$f,$e));# f+=K_40_59+e+(b&(c^d)) + &mov($e,&swtmp(($n+1)%16)); # pre-fetch f for next round + &add($f,$a); # f+=ROTATE(a,5) +} else { &mov($tmp1,$c); # tmp1 to hold F_40_59(b,c,d) &xor($f,&swtmp(($n+2)%16)); # f to hold Xupdate(xi,xa,xb,xc,xd) &xor($tmp1,$d); @@ -157,19 +290,48 @@ sub BODY_40_59 &and($tmp1,$d); &mov($e,&swtmp(($n+1)%16)); # pre-fetch f for next round &add($f,$tmp1); # f+=c&d +} } &function_begin("sha1_block_data_order"); +if ($xmm) { + &static_label("ssse3_shortcut"); + &static_label("avx_shortcut") if ($ymm); + &static_label("K_XX_XX"); + + &call (&label("pic_point")); # make it PIC! + &set_label("pic_point"); + &blindpop($tmp1); + &picmeup($T,"OPENSSL_ia32cap_P",$tmp1,&label("pic_point")); + &lea ($tmp1,&DWP(&label("K_XX_XX")."-".&label("pic_point"),$tmp1)); + + &mov ($A,&DWP(0,$T)); + &mov ($D,&DWP(4,$T)); + &test ($D,1<<9); # check SSSE3 bit + &jz (&label("x86")); + &test ($A,1<<24); # check FXSR bit + &jz (&label("x86")); + if ($ymm) { + &and ($D,1<<28); # mask AVX bit + &and ($A,1<<30); # mask "Intel CPU" bit + &or ($A,$D); + &cmp ($A,1<<28|1<<30); + &je (&label("avx_shortcut")); + } + &jmp (&label("ssse3_shortcut")); + &set_label("x86",16); +} &mov($tmp1,&wparam(0)); # SHA_CTX *c &mov($T,&wparam(1)); # const void *input &mov($A,&wparam(2)); # size_t num - &stack_push(16); # allocate X[16] + &stack_push(16+3); # allocate X[16] &shl($A,6); &add($A,$T); &mov(&wparam(2),$A); # pointer beyond the end of input &mov($E,&DWP(16,$tmp1));# pre-load E + &jmp(&label("loop")); - &set_label("loop",16); +&set_label("loop",16); # copy input chunk to X, but reversing byte order! for ($i=0; $i<16; $i+=4) @@ -223,8 +385,845 @@ sub BODY_40_59 &mov(&DWP(16,$tmp1),$C); &jb(&label("loop")); - &stack_pop(16); + &stack_pop(16+3); &function_end("sha1_block_data_order"); + +if ($xmm) { +###################################################################### +# The SSSE3 implementation. +# +# %xmm[0-7] are used as ring @X[] buffer containing quadruples of last +# 32 elements of the message schedule or Xupdate outputs. First 4 +# quadruples are simply byte-swapped input, next 4 are calculated +# according to method originally suggested by Dean Gaudet (modulo +# 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 +# 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] - +# X[-5], and X[4] - X[-4]... +# +# Another notable optimization is aggressive stack frame compression +# aiming to minimize amount of 9-byte instructions... +# +# Yet another notable optimization is "jumping" $B variable. It means +# that there is no register permanently allocated for $B value. This +# allowed to eliminate one instruction from body_20_39... +# +my $Xi=4; # 4xSIMD Xupdate round, start pre-seeded +my @X=map("xmm$_",(4..7,0..3)); # pre-seeded for $Xi=4 +my @V=($A,$B,$C,$D,$E); +my $j=0; # hash round +my @T=($T,$tmp1); +my $inp; + +my $_rol=sub { &rol(@_) }; +my $_ror=sub { &ror(@_) }; + +&function_begin("_sha1_block_data_order_ssse3"); + &call (&label("pic_point")); # make it PIC! + &set_label("pic_point"); + &blindpop($tmp1); + &lea ($tmp1,&DWP(&label("K_XX_XX")."-".&label("pic_point"),$tmp1)); +&set_label("ssse3_shortcut"); + + &movdqa (@X[3],&QWP(0,$tmp1)); # K_00_19 + &movdqa (@X[4],&QWP(16,$tmp1)); # K_20_39 + &movdqa (@X[5],&QWP(32,$tmp1)); # K_40_59 + &movdqa (@X[6],&QWP(48,$tmp1)); # K_60_79 + &movdqa (@X[2],&QWP(64,$tmp1)); # pbswap mask + + &mov ($E,&wparam(0)); # load argument block + &mov ($inp=@T[1],&wparam(1)); + &mov ($D,&wparam(2)); + &mov (@T[0],"esp"); + + # stack frame layout + # + # +0 X[0]+K X[1]+K X[2]+K X[3]+K # XMM->IALU xfer area + # X[4]+K X[5]+K X[6]+K X[7]+K + # X[8]+K X[9]+K X[10]+K X[11]+K + # X[12]+K X[13]+K X[14]+K X[15]+K + # + # +64 X[0] X[1] X[2] X[3] # XMM->XMM backtrace area + # X[4] X[5] X[6] X[7] + # X[8] X[9] X[10] X[11] # even borrowed for K_00_19 + # + # +112 K_20_39 K_20_39 K_20_39 K_20_39 # constants + # K_40_59 K_40_59 K_40_59 K_40_59 + # K_60_79 K_60_79 K_60_79 K_60_79 + # K_00_19 K_00_19 K_00_19 K_00_19 + # pbswap mask + # + # +192 ctx # argument block + # +196 inp + # +200 end + # +204 esp + &sub ("esp",208); + &and ("esp",-64); + + &movdqa (&QWP(112+0,"esp"),@X[4]); # copy constants + &movdqa (&QWP(112+16,"esp"),@X[5]); + &movdqa (&QWP(112+32,"esp"),@X[6]); + &shl ($D,6); # len*64 + &movdqa (&QWP(112+48,"esp"),@X[3]); + &add ($D,$inp); # end of input + &movdqa (&QWP(112+64,"esp"),@X[2]); + &add ($inp,64); + &mov (&DWP(192+0,"esp"),$E); # save argument block + &mov (&DWP(192+4,"esp"),$inp); + &mov (&DWP(192+8,"esp"),$D); + &mov (&DWP(192+12,"esp"),@T[0]); # save original %esp + + &mov ($A,&DWP(0,$E)); # load context + &mov ($B,&DWP(4,$E)); + &mov ($C,&DWP(8,$E)); + &mov ($D,&DWP(12,$E)); + &mov ($E,&DWP(16,$E)); + &mov (@T[0],$B); # magic seed + + &movdqu (@X[-4&7],&QWP(-64,$inp)); # load input to %xmm[0-3] + &movdqu (@X[-3&7],&QWP(-48,$inp)); + &movdqu (@X[-2&7],&QWP(-32,$inp)); + &movdqu (@X[-1&7],&QWP(-16,$inp)); + &pshufb (@X[-4&7],@X[2]); # byte swap + &pshufb (@X[-3&7],@X[2]); + &pshufb (@X[-2&7],@X[2]); + &movdqa (&QWP(112-16,"esp"),@X[3]); # borrow last backtrace slot + &pshufb (@X[-1&7],@X[2]); + &paddd (@X[-4&7],@X[3]); # add K_00_19 + &paddd (@X[-3&7],@X[3]); + &paddd (@X[-2&7],@X[3]); + &movdqa (&QWP(0,"esp"),@X[-4&7]); # X[]+K xfer to IALU + &psubd (@X[-4&7],@X[3]); # restore X[] + &movdqa (&QWP(0+16,"esp"),@X[-3&7]); + &psubd (@X[-3&7],@X[3]); + &movdqa (&QWP(0+32,"esp"),@X[-2&7]); + &psubd (@X[-2&7],@X[3]); + &movdqa (@X[0],@X[-3&7]); + &jmp (&label("loop")); + +###################################################################### +# SSE instruction sequence is first broken to groups of indepentent +# 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 +# SSE latency of 2 in hope that it fits better upcoming AMD Bulldozer +# [which allegedly also implements SSSE3]... +# +# 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 +# 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 +{ use integer; + my $body = shift; + my @insns = (&$body,&$body,&$body,&$body); # 40 instructions + my ($a,$b,$c,$d,$e); + + eval(shift(@insns)); + eval(shift(@insns)); + &palignr(@X[0],@X[-4&7],8); # compose "X[-14]" in "X[0]" + &movdqa (@X[2],@X[-1&7]); + eval(shift(@insns)); + eval(shift(@insns)); + + &paddd (@X[3],@X[-1&7]); + &movdqa (&QWP(64+16*(($Xi-4)%3),"esp"),@X[-4&7]);# save X[] to backtrace buffer + eval(shift(@insns)); + eval(shift(@insns)); + &psrldq (@X[2],4); # "X[-3]", 3 dwords + eval(shift(@insns)); + eval(shift(@insns)); + &pxor (@X[0],@X[-4&7]); # "X[0]"^="X[-16]" + eval(shift(@insns)); + eval(shift(@insns)); + + &pxor (@X[2],@X[-2&7]); # "X[-3]"^"X[-8]" + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + + &pxor (@X[0],@X[2]); # "X[0]"^="X[-3]"^"X[-8]" + eval(shift(@insns)); + eval(shift(@insns)); + &movdqa (&QWP(0+16*(($Xi-1)&3),"esp"),@X[3]); # X[]+K xfer to IALU + eval(shift(@insns)); + eval(shift(@insns)); + + &movdqa (@X[4],@X[0]); + &movdqa (@X[2],@X[0]); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + + &pslldq (@X[4],12); # "X[0]"<<96, extract one dword + &paddd (@X[0],@X[0]); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + + &psrld (@X[2],31); + eval(shift(@insns)); + eval(shift(@insns)); + &movdqa (@X[3],@X[4]); + eval(shift(@insns)); + eval(shift(@insns)); + + &psrld (@X[4],30); + &por (@X[0],@X[2]); # "X[0]"<<<=1 + eval(shift(@insns)); + eval(shift(@insns)); + &movdqa (@X[2],&QWP(64+16*(($Xi-6)%3),"esp")) if ($Xi>5); # restore X[] from backtrace buffer + eval(shift(@insns)); + eval(shift(@insns)); + + &pslld (@X[3],2); + &pxor (@X[0],@X[4]); + eval(shift(@insns)); + eval(shift(@insns)); + &movdqa (@X[4],&QWP(112-16+16*(($Xi)/5),"esp")); # K_XX_XX + eval(shift(@insns)); + eval(shift(@insns)); + + &pxor (@X[0],@X[3]); # "X[0]"^=("X[0]"<<96)<<<2 + &movdqa (@X[1],@X[-2&7]) if ($Xi<7); + eval(shift(@insns)); + eval(shift(@insns)); + + foreach (@insns) { eval; } # remaining instructions [if any] + + $Xi++; push(@X,shift(@X)); # "rotate" X[] +} + +sub Xupdate_ssse3_32_79() +{ use integer; + my $body = shift; + my @insns = (&$body,&$body,&$body,&$body); # 32 to 48 instructions + my ($a,$b,$c,$d,$e); + + &movdqa (@X[2],@X[-1&7]) if ($Xi==8); + eval(shift(@insns)); # body_20_39 + &pxor (@X[0],@X[-4&7]); # "X[0]"="X[-32]"^"X[-16]" + &palignr(@X[2],@X[-2&7],8); # compose "X[-6]" + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); # rol + + &pxor (@X[0],@X[-7&7]); # "X[0]"^="X[-28]" + &movdqa (&QWP(64+16*(($Xi-4)%3),"esp"),@X[-4&7]); # save X[] to backtrace buffer + eval(shift(@insns)); + eval(shift(@insns)); + if ($Xi%5) { + &movdqa (@X[4],@X[3]); # "perpetuate" K_XX_XX... + } else { # ... or load next one + &movdqa (@X[4],&QWP(112-16+16*($Xi/5),"esp")); + } + &paddd (@X[3],@X[-1&7]); + eval(shift(@insns)); # ror + eval(shift(@insns)); + + &pxor (@X[0],@X[2]); # "X[0]"^="X[-6]" + eval(shift(@insns)); # body_20_39 + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); # rol + + &movdqa (@X[2],@X[0]); + &movdqa (&QWP(0+16*(($Xi-1)&3),"esp"),@X[3]); # X[]+K xfer to IALU + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); # ror + eval(shift(@insns)); + + &pslld (@X[0],2); + eval(shift(@insns)); # body_20_39 + eval(shift(@insns)); + &psrld (@X[2],30); + eval(shift(@insns)); + eval(shift(@insns)); # rol + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); # ror + eval(shift(@insns)); + + &por (@X[0],@X[2]); # "X[0]"<<<=2 + eval(shift(@insns)); # body_20_39 + eval(shift(@insns)); + &movdqa (@X[2],&QWP(64+16*(($Xi-6)%3),"esp")) if($Xi<19); # restore X[] from backtrace buffer + eval(shift(@insns)); + eval(shift(@insns)); # rol + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); # ror + &movdqa (@X[3],@X[0]) if ($Xi<19); + eval(shift(@insns)); + + foreach (@insns) { eval; } # remaining instructions + + $Xi++; push(@X,shift(@X)); # "rotate" X[] +} + +sub Xuplast_ssse3_80() +{ use integer; + my $body = shift; + my @insns = (&$body,&$body,&$body,&$body); # 32 instructions + my ($a,$b,$c,$d,$e); + + eval(shift(@insns)); + &paddd (@X[3],@X[-1&7]); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + + &movdqa (&QWP(0+16*(($Xi-1)&3),"esp"),@X[3]); # X[]+K xfer IALU + + foreach (@insns) { eval; } # remaining instructions + + &mov ($inp=@T[1],&DWP(192+4,"esp")); + &cmp ($inp,&DWP(192+8,"esp")); + &je (&label("done")); + + &movdqa (@X[3],&QWP(112+48,"esp")); # K_00_19 + &movdqa (@X[2],&QWP(112+64,"esp")); # pbswap mask + &movdqu (@X[-4&7],&QWP(0,$inp)); # load input + &movdqu (@X[-3&7],&QWP(16,$inp)); + &movdqu (@X[-2&7],&QWP(32,$inp)); + &movdqu (@X[-1&7],&QWP(48,$inp)); + &add ($inp,64); + &pshufb (@X[-4&7],@X[2]); # byte swap + &mov (&DWP(192+4,"esp"),$inp); + &movdqa (&QWP(112-16,"esp"),@X[3]); # borrow last backtrace slot + + $Xi=0; +} + +sub Xloop_ssse3() +{ use integer; + my $body = shift; + my @insns = (&$body,&$body,&$body,&$body); # 32 instructions + my ($a,$b,$c,$d,$e); + + eval(shift(@insns)); + eval(shift(@insns)); + &pshufb (@X[($Xi-3)&7],@X[2]); + eval(shift(@insns)); + eval(shift(@insns)); + &paddd (@X[($Xi-4)&7],@X[3]); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + &movdqa (&QWP(0+16*$Xi,"esp"),@X[($Xi-4)&7]); # X[]+K xfer to IALU + eval(shift(@insns)); + eval(shift(@insns)); + &psubd (@X[($Xi-4)&7],@X[3]); + + foreach (@insns) { eval; } + $Xi++; +} + +sub Xtail_ssse3() +{ use integer; + my $body = shift; + my @insns = (&$body,&$body,&$body,&$body); # 32 instructions + my ($a,$b,$c,$d,$e); + + foreach (@insns) { eval; } +} + +sub body_00_19 () { + ( + '($a,$b,$c,$d,$e)=@V;'. + '&add ($e,&DWP(4*($j&15),"esp"));', # X[]+K xfer + '&xor ($c,$d);', + '&mov (@T[1],$a);', # $b in next round + '&$_rol ($a,5);', + '&and (@T[0],$c);', # ($b&($c^$d)) + '&xor ($c,$d);', # restore $c + '&xor (@T[0],$d);', + '&add ($e,$a);', + '&$_ror ($b,$j?7:2);', # $b>>>2 + '&add ($e,@T[0]);' .'$j++; unshift(@V,pop(@V)); unshift(@T,pop(@T));' + ); +} + +sub body_20_39 () { + ( + '($a,$b,$c,$d,$e)=@V;'. + '&add ($e,&DWP(4*($j++&15),"esp"));', # X[]+K xfer + '&xor (@T[0],$d);', # ($b^$d) + '&mov (@T[1],$a);', # $b in next round + '&$_rol ($a,5);', + '&xor (@T[0],$c);', # ($b^$d^$c) + '&add ($e,$a);', + '&$_ror ($b,7);', # $b>>>2 + '&add ($e,@T[0]);' .'unshift(@V,pop(@V)); unshift(@T,pop(@T));' + ); +} + +sub body_40_59 () { + ( + '($a,$b,$c,$d,$e)=@V;'. + '&mov (@T[1],$c);', + '&xor ($c,$d);', + '&add ($e,&DWP(4*($j++&15),"esp"));', # X[]+K xfer + '&and (@T[1],$d);', + '&and (@T[0],$c);', # ($b&($c^$d)) + '&$_ror ($b,7);', # $b>>>2 + '&add ($e,@T[1]);', + '&mov (@T[1],$a);', # $b in next round + '&$_rol ($a,5);', + '&add ($e,@T[0]);', + '&xor ($c,$d);', # restore $c + '&add ($e,$a);' .'unshift(@V,pop(@V)); unshift(@T,pop(@T));' + ); +} + +&set_label("loop",16); + &Xupdate_ssse3_16_31(\&body_00_19); + &Xupdate_ssse3_16_31(\&body_00_19); + &Xupdate_ssse3_16_31(\&body_00_19); + &Xupdate_ssse3_16_31(\&body_00_19); + &Xupdate_ssse3_32_79(\&body_00_19); + &Xupdate_ssse3_32_79(\&body_20_39); + &Xupdate_ssse3_32_79(\&body_20_39); + &Xupdate_ssse3_32_79(\&body_20_39); + &Xupdate_ssse3_32_79(\&body_20_39); + &Xupdate_ssse3_32_79(\&body_20_39); + &Xupdate_ssse3_32_79(\&body_40_59); + &Xupdate_ssse3_32_79(\&body_40_59); + &Xupdate_ssse3_32_79(\&body_40_59); + &Xupdate_ssse3_32_79(\&body_40_59); + &Xupdate_ssse3_32_79(\&body_40_59); + &Xupdate_ssse3_32_79(\&body_20_39); + &Xuplast_ssse3_80(\&body_20_39); # can jump to "done" + + $saved_j=$j; @saved_V=@V; + + &Xloop_ssse3(\&body_20_39); + &Xloop_ssse3(\&body_20_39); + &Xloop_ssse3(\&body_20_39); + + &mov (@T[1],&DWP(192,"esp")); # update context + &add ($A,&DWP(0,@T[1])); + &add (@T[0],&DWP(4,@T[1])); # $b + &add ($C,&DWP(8,@T[1])); + &mov (&DWP(0,@T[1]),$A); + &add ($D,&DWP(12,@T[1])); + &mov (&DWP(4,@T[1]),@T[0]); + &add ($E,&DWP(16,@T[1])); + &mov (&DWP(8,@T[1]),$C); + &mov ($B,@T[0]); + &mov (&DWP(12,@T[1]),$D); + &mov (&DWP(16,@T[1]),$E); + &movdqa (@X[0],@X[-3&7]); + + &jmp (&label("loop")); + +&set_label("done",16); $j=$saved_j; @V=@saved_V; + + &Xtail_ssse3(\&body_20_39); + &Xtail_ssse3(\&body_20_39); + &Xtail_ssse3(\&body_20_39); + + &mov (@T[1],&DWP(192,"esp")); # update context + &add ($A,&DWP(0,@T[1])); + &mov ("esp",&DWP(192+12,"esp")); # restore %esp + &add (@T[0],&DWP(4,@T[1])); # $b + &add ($C,&DWP(8,@T[1])); + &mov (&DWP(0,@T[1]),$A); + &add ($D,&DWP(12,@T[1])); + &mov (&DWP(4,@T[1]),@T[0]); + &add ($E,&DWP(16,@T[1])); + &mov (&DWP(8,@T[1]),$C); + &mov (&DWP(12,@T[1]),$D); + &mov (&DWP(16,@T[1]),$E); + +&function_end("_sha1_block_data_order_ssse3"); + +if ($ymm) { +my $Xi=4; # 4xSIMD Xupdate round, start pre-seeded +my @X=map("xmm$_",(4..7,0..3)); # pre-seeded for $Xi=4 +my @V=($A,$B,$C,$D,$E); +my $j=0; # hash round +my @T=($T,$tmp1); +my $inp; + +my $_rol=sub { &shld(@_[0],@_) }; +my $_ror=sub { &shrd(@_[0],@_) }; + +&function_begin("_sha1_block_data_order_avx"); + &call (&label("pic_point")); # make it PIC! + &set_label("pic_point"); + &blindpop($tmp1); + &lea ($tmp1,&DWP(&label("K_XX_XX")."-".&label("pic_point"),$tmp1)); +&set_label("avx_shortcut"); + &vzeroall(); + + &vmovdqa(@X[3],&QWP(0,$tmp1)); # K_00_19 + &vmovdqa(@X[4],&QWP(16,$tmp1)); # K_20_39 + &vmovdqa(@X[5],&QWP(32,$tmp1)); # K_40_59 + &vmovdqa(@X[6],&QWP(48,$tmp1)); # K_60_79 + &vmovdqa(@X[2],&QWP(64,$tmp1)); # pbswap mask + + &mov ($E,&wparam(0)); # load argument block + &mov ($inp=@T[1],&wparam(1)); + &mov ($D,&wparam(2)); + &mov (@T[0],"esp"); + + # stack frame layout + # + # +0 X[0]+K X[1]+K X[2]+K X[3]+K # XMM->IALU xfer area + # X[4]+K X[5]+K X[6]+K X[7]+K + # X[8]+K X[9]+K X[10]+K X[11]+K + # X[12]+K X[13]+K X[14]+K X[15]+K + # + # +64 X[0] X[1] X[2] X[3] # XMM->XMM backtrace area + # X[4] X[5] X[6] X[7] + # X[8] X[9] X[10] X[11] # even borrowed for K_00_19 + # + # +112 K_20_39 K_20_39 K_20_39 K_20_39 # constants + # K_40_59 K_40_59 K_40_59 K_40_59 + # K_60_79 K_60_79 K_60_79 K_60_79 + # K_00_19 K_00_19 K_00_19 K_00_19 + # pbswap mask + # + # +192 ctx # argument block + # +196 inp + # +200 end + # +204 esp + &sub ("esp",208); + &and ("esp",-64); + + &vmovdqa(&QWP(112+0,"esp"),@X[4]); # copy constants + &vmovdqa(&QWP(112+16,"esp"),@X[5]); + &vmovdqa(&QWP(112+32,"esp"),@X[6]); + &shl ($D,6); # len*64 + &vmovdqa(&QWP(112+48,"esp"),@X[3]); + &add ($D,$inp); # end of input + &vmovdqa(&QWP(112+64,"esp"),@X[2]); + &add ($inp,64); + &mov (&DWP(192+0,"esp"),$E); # save argument block + &mov (&DWP(192+4,"esp"),$inp); + &mov (&DWP(192+8,"esp"),$D); + &mov (&DWP(192+12,"esp"),@T[0]); # save original %esp + + &mov ($A,&DWP(0,$E)); # load context + &mov ($B,&DWP(4,$E)); + &mov ($C,&DWP(8,$E)); + &mov ($D,&DWP(12,$E)); + &mov ($E,&DWP(16,$E)); + &mov (@T[0],$B); # magic seed + + &vmovdqu(@X[-4&7],&QWP(-64,$inp)); # load input to %xmm[0-3] + &vmovdqu(@X[-3&7],&QWP(-48,$inp)); + &vmovdqu(@X[-2&7],&QWP(-32,$inp)); + &vmovdqu(@X[-1&7],&QWP(-16,$inp)); + &vpshufb(@X[-4&7],@X[-4&7],@X[2]); # byte swap + &vpshufb(@X[-3&7],@X[-3&7],@X[2]); + &vpshufb(@X[-2&7],@X[-2&7],@X[2]); + &vmovdqa(&QWP(112-16,"esp"),@X[3]); # borrow last backtrace slot + &vpshufb(@X[-1&7],@X[-1&7],@X[2]); + &vpaddd (@X[0],@X[-4&7],@X[3]); # add K_00_19 + &vpaddd (@X[1],@X[-3&7],@X[3]); + &vpaddd (@X[2],@X[-2&7],@X[3]); + &vmovdqa(&QWP(0,"esp"),@X[0]); # X[]+K xfer to IALU + &vmovdqa(&QWP(0+16,"esp"),@X[1]); + &vmovdqa(&QWP(0+32,"esp"),@X[2]); + &jmp (&label("loop")); + +sub Xupdate_avx_16_31() # recall that $Xi starts wtih 4 +{ use integer; + my $body = shift; + my @insns = (&$body,&$body,&$body,&$body); # 40 instructions + my ($a,$b,$c,$d,$e); + + eval(shift(@insns)); + eval(shift(@insns)); + &vpalignr(@X[0],@X[-3&7],@X[-4&7],8); # compose "X[-14]" in "X[0]" + eval(shift(@insns)); + eval(shift(@insns)); + + &vpaddd (@X[3],@X[3],@X[-1&7]); + &vmovdqa (&QWP(64+16*(($Xi-4)%3),"esp"),@X[-4&7]);# save X[] to backtrace buffer + eval(shift(@insns)); + eval(shift(@insns)); + &vpsrldq(@X[2],@X[-1&7],4); # "X[-3]", 3 dwords + eval(shift(@insns)); + eval(shift(@insns)); + &vpxor (@X[0],@X[0],@X[-4&7]); # "X[0]"^="X[-16]" + eval(shift(@insns)); + eval(shift(@insns)); + + &vpxor (@X[2],@X[2],@X[-2&7]); # "X[-3]"^"X[-8]" + eval(shift(@insns)); + eval(shift(@insns)); + &vmovdqa (&QWP(0+16*(($Xi-1)&3),"esp"),@X[3]); # X[]+K xfer to IALU + eval(shift(@insns)); + eval(shift(@insns)); + + &vpxor (@X[0],@X[0],@X[2]); # "X[0]"^="X[-3]"^"X[-8]" + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + + &vpsrld (@X[2],@X[0],31); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + + &vpslldq(@X[4],@X[0],12); # "X[0]"<<96, extract one dword + &vpaddd (@X[0],@X[0],@X[0]); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + + &vpsrld (@X[3],@X[4],30); + &vpor (@X[0],@X[0],@X[2]); # "X[0]"<<<=1 + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + + &vpslld (@X[4],@X[4],2); + &vmovdqa (@X[2],&QWP(64+16*(($Xi-6)%3),"esp")) if ($Xi>5); # restore X[] from backtrace buffer + eval(shift(@insns)); + eval(shift(@insns)); + &vpxor (@X[0],@X[0],@X[3]); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + + &vpxor (@X[0],@X[0],@X[4]); # "X[0]"^=("X[0]"<<96)<<<2 + eval(shift(@insns)); + eval(shift(@insns)); + &vmovdqa (@X[4],&QWP(112-16+16*(($Xi)/5),"esp")); # K_XX_XX + eval(shift(@insns)); + eval(shift(@insns)); + + foreach (@insns) { eval; } # remaining instructions [if any] + + $Xi++; push(@X,shift(@X)); # "rotate" X[] +} + +sub Xupdate_avx_32_79() +{ use integer; + my $body = shift; + my @insns = (&$body,&$body,&$body,&$body); # 32 to 48 instructions + my ($a,$b,$c,$d,$e); + + &vpalignr(@X[2],@X[-1&7],@X[-2&7],8); # compose "X[-6]" + &vpxor (@X[0],@X[0],@X[-4&7]); # "X[0]"="X[-32]"^"X[-16]" + eval(shift(@insns)); # body_20_39 + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); # rol + + &vpxor (@X[0],@X[0],@X[-7&7]); # "X[0]"^="X[-28]" + &vmovdqa (&QWP(64+16*(($Xi-4)%3),"esp"),@X[-4&7]); # save X[] to backtrace buffer + eval(shift(@insns)); + eval(shift(@insns)); + if ($Xi%5) { + &vmovdqa (@X[4],@X[3]); # "perpetuate" K_XX_XX... + } else { # ... or load next one + &vmovdqa (@X[4],&QWP(112-16+16*($Xi/5),"esp")); + } + &vpaddd (@X[3],@X[3],@X[-1&7]); + eval(shift(@insns)); # ror + eval(shift(@insns)); + + &vpxor (@X[0],@X[0],@X[2]); # "X[0]"^="X[-6]" + eval(shift(@insns)); # body_20_39 + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); # rol + + &vpsrld (@X[2],@X[0],30); + &vmovdqa (&QWP(0+16*(($Xi-1)&3),"esp"),@X[3]); # X[]+K xfer to IALU + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); # ror + eval(shift(@insns)); + + &vpslld (@X[0],@X[0],2); + eval(shift(@insns)); # body_20_39 + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); # rol + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); # ror + eval(shift(@insns)); + + &vpor (@X[0],@X[0],@X[2]); # "X[0]"<<<=2 + eval(shift(@insns)); # body_20_39 + eval(shift(@insns)); + &vmovdqa (@X[2],&QWP(64+16*(($Xi-6)%3),"esp")) if($Xi<19); # restore X[] from backtrace buffer + eval(shift(@insns)); + eval(shift(@insns)); # rol + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); # ror + eval(shift(@insns)); + + foreach (@insns) { eval; } # remaining instructions + + $Xi++; push(@X,shift(@X)); # "rotate" X[] +} + +sub Xuplast_avx_80() +{ use integer; + my $body = shift; + my @insns = (&$body,&$body,&$body,&$body); # 32 instructions + my ($a,$b,$c,$d,$e); + + eval(shift(@insns)); + &vpaddd (@X[3],@X[3],@X[-1&7]); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + eval(shift(@insns)); + + &vmovdqa (&QWP(0+16*(($Xi-1)&3),"esp"),@X[3]); # X[]+K xfer IALU + + foreach (@insns) { eval; } # remaining instructions + + &mov ($inp=@T[1],&DWP(192+4,"esp")); + &cmp ($inp,&DWP(192+8,"esp")); + &je (&label("done")); + + &vmovdqa(@X[3],&QWP(112+48,"esp")); # K_00_19 + &vmovdqa(@X[2],&QWP(112+64,"esp")); # pbswap mask + &vmovdqu(@X[-4&7],&QWP(0,$inp)); # load input + &vmovdqu(@X[-3&7],&QWP(16,$inp)); + &vmovdqu(@X[-2&7],&QWP(32,$inp)); + &vmovdqu(@X[-1&7],&QWP(48,$inp)); + &add ($inp,64); + &vpshufb(@X[-4&7],@X[-4&7],@X[2]); # byte swap + &mov (&DWP(192+4,"esp"),$inp); + &vmovdqa(&QWP(112-16,"esp"),@X[3]); # borrow last backtrace slot + + $Xi=0; +} + +sub Xloop_avx() +{ use integer; + my $body = shift; + my @insns = (&$body,&$body,&$body,&$body); # 32 instructions |