提交 a8f302e5 编写于 作者: A Andy Polyakov

poly1305/asm/poly1305-x86_64.pl: switch to pure AVX512F.

Convert AVX512F+VL+BW code path to pure AVX512F, so that it can be
executed even on Knights Landing. Trigger for modification was
observation that AVX512 code paths can negatively affect overall
Skylake-X system performance. Since we are likely to suppress
AVX512F capability flag [at least on Skylake-X], conversion serves
as kind of "investment protection".
Reviewed-by: NRich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4758)
上级 10a3195f
...@@ -24,6 +24,16 @@ ...@@ -24,6 +24,16 @@
# #
# Add AVX512F+VL+BW code path. # Add AVX512F+VL+BW code path.
# #
# November 2017
#
# Convert AVX512F+VL+BW code path to pure AVX512F, so that it can be
# executed even on Knights Landing. Trigger for modification was
# observation that AVX512 code paths can negatively affect overall
# Skylake-X system performance. Since we are likely to suppress
# AVX512F capability flag [at least on Skylake-X], conversion serves
# as kind of "investment protection". Note that next *lake processor,
# Cannolake, has AVX512IFMA code path to execute...
#
# Numbers are cycles per processed byte with poly1305_blocks alone, # Numbers are cycles per processed byte with poly1305_blocks alone,
# measured with rdtsc at fixed clock frequency. # measured with rdtsc at fixed clock frequency.
# #
...@@ -35,7 +45,7 @@ ...@@ -35,7 +45,7 @@
# Haswell 1.14/+175% 1.11 0.65 # Haswell 1.14/+175% 1.11 0.65
# Skylake[-X] 1.13/+120% 0.96 0.51 [0.35] # Skylake[-X] 1.13/+120% 0.96 0.51 [0.35]
# Silvermont 2.83/+95% - # Silvermont 2.83/+95% -
# Knights L 3.60/- 1.65 1.10 (***) # Knights L 3.60/? 1.65 1.10 ?
# Goldmont 1.70/+180% - # Goldmont 1.70/+180% -
# VIA Nano 1.82/+150% - # VIA Nano 1.82/+150% -
# Sledgehammer 1.38/+160% - # Sledgehammer 1.38/+160% -
...@@ -50,8 +60,6 @@ ...@@ -50,8 +60,6 @@
# Core processors, 50-30%, less newer processor is, but slower on # Core processors, 50-30%, less newer processor is, but slower on
# contemporary ones, for example almost 2x slower on Atom, and as # contemporary ones, for example almost 2x slower on Atom, and as
# former are naturally disappearing, SSE2 is deemed unnecessary; # former are naturally disappearing, SSE2 is deemed unnecessary;
# (***) Current AVX-512 code requires BW and VL extensions and can not
# execute on Knights Landing;
$flavour = shift; $flavour = shift;
$output = shift; $output = shift;
...@@ -1685,7 +1693,6 @@ poly1305_blocks_avx2: ...@@ -1685,7 +1693,6 @@ poly1305_blocks_avx2:
.Leven_avx2: .Leven_avx2:
.cfi_startproc .cfi_startproc
mov OPENSSL_ia32cap_P+8(%rip),%r10d mov OPENSSL_ia32cap_P+8(%rip),%r10d
mov \$`(1<<31|1<<30|1<<16)`,%r11d
vmovd 4*0($ctx),%x#$H0 # load hash value base 2^26 vmovd 4*0($ctx),%x#$H0 # load hash value base 2^26
vmovd 4*1($ctx),%x#$H1 vmovd 4*1($ctx),%x#$H1
vmovd 4*2($ctx),%x#$H2 vmovd 4*2($ctx),%x#$H2
...@@ -1698,8 +1705,8 @@ $code.=<<___ if ($avx>2); ...@@ -1698,8 +1705,8 @@ $code.=<<___ if ($avx>2);
cmp \$512,$len cmp \$512,$len
jb .Lskip_avx512 jb .Lskip_avx512
and %r11d,%r10d and %r11d,%r10d
cmp %r11d,%r10d # check for AVX512F+BW+VL test \$`1<<16`,%r10d # check for AVX512F
je .Lblocks_avx512 jnz .Lblocks_avx512
.Lskip_avx512: .Lskip_avx512:
___ ___
$code.=<<___ if (!$win64); $code.=<<___ if (!$win64);
...@@ -2109,10 +2116,14 @@ if ($avx>2) { ...@@ -2109,10 +2116,14 @@ if ($avx>2) {
# reason stack layout is kept identical to poly1305_blocks_avx2. If not # reason stack layout is kept identical to poly1305_blocks_avx2. If not
# for this tail, we wouldn't have to even allocate stack frame... # for this tail, we wouldn't have to even allocate stack frame...
my ($R0,$R1,$R2,$R3,$R4, $S1,$S2,$S3,$S4) = map("%ymm$_",(16..24)); my ($R0,$R1,$R2,$R3,$R4, $S1,$S2,$S3,$S4) = map("%zmm$_",(16..24));
my ($M0,$M1,$M2,$M3,$M4) = map("%ymm$_",(25..29)); my ($M0,$M1,$M2,$M3,$M4) = map("%zmm$_",(25..29));
my $PADBIT="%zmm30"; my $PADBIT="%zmm30";
my $GATHER="%ymm31";
map(s/%y/%z/,($T4,$T0,$T1,$T2,$T3)); # switch to %zmm domain
map(s/%y/%z/,($D0,$D1,$D2,$D3,$D4));
map(s/%y/%z/,($H0,$H1,$H2,$H3,$H4));
map(s/%y/%z/,($MASK));
$code.=<<___; $code.=<<___;
.type poly1305_blocks_avx512,\@function,4 .type poly1305_blocks_avx512,\@function,4
...@@ -2120,7 +2131,8 @@ $code.=<<___; ...@@ -2120,7 +2131,8 @@ $code.=<<___;
poly1305_blocks_avx512: poly1305_blocks_avx512:
.cfi_startproc .cfi_startproc
.Lblocks_avx512: .Lblocks_avx512:
vzeroupper mov \$15,%eax
kmovw %eax,%k2
___ ___
$code.=<<___ if (!$win64); $code.=<<___ if (!$win64);
lea -8(%rsp),%r11 lea -8(%rsp),%r11
...@@ -2133,52 +2145,53 @@ $code.=<<___ if ($win64); ...@@ -2133,52 +2145,53 @@ $code.=<<___ if ($win64);
vmovdqa %xmm6,0x50(%r11) vmovdqa %xmm6,0x50(%r11)
vmovdqa %xmm7,0x60(%r11) vmovdqa %xmm7,0x60(%r11)
vmovdqa %xmm8,0x70(%r11) vmovdqa %xmm8,0x70(%r11)
vmovdqa32 %xmm9,0x80(%r11) vmovdqa %xmm9,0x80(%r11)
vmovdqa32 %xmm10,0x90(%r11) vmovdqa %xmm10,0x90(%r11)
vmovdqa32 %xmm11,0xa0(%r11) vmovdqa %xmm11,0xa0(%r11)
vmovdqa32 %xmm12,0xb0(%r11) vmovdqa %xmm12,0xb0(%r11)
vmovdqa32 %xmm13,0xc0(%r11) vmovdqa %xmm13,0xc0(%r11)
vmovdqa32 %xmm14,0xd0(%r11) vmovdqa %xmm14,0xd0(%r11)
vmovdqa32 %xmm15,0xe0(%r11) vmovdqa %xmm15,0xe0(%r11)
.Ldo_avx512_body: .Ldo_avx512_body:
___ ___
$code.=<<___; $code.=<<___;
lea .Lconst(%rip),%rcx lea .Lconst(%rip),%rcx
lea 48+64($ctx),$ctx # size optimization lea 48+64($ctx),$ctx # size optimization
vmovdqa 96(%rcx),$T2 # .Lpermd_avx2 vmovdqa 96(%rcx),%y#$T2 # .Lpermd_avx2
# expand pre-calculated table # expand pre-calculated table
vmovdqu32 `16*0-64`($ctx),%x#$R0 vmovdqu32 `16*0-64`($ctx),${R0}{%k2}{z}
and \$-512,%rsp and \$-512,%rsp
vmovdqu32 `16*1-64`($ctx),%x#$R1 vmovdqu32 `16*1-64`($ctx),${R1}{%k2}{z}
vmovdqu32 `16*2-64`($ctx),%x#$S1 mov \$0x20,%rax
vmovdqu32 `16*3-64`($ctx),%x#$R2 vmovdqu32 `16*2-64`($ctx),${S1}{%k2}{z}
vmovdqu32 `16*4-64`($ctx),%x#$S2 vmovdqu32 `16*3-64`($ctx),${R2}{%k2}{z}
vmovdqu32 `16*5-64`($ctx),%x#$R3 vmovdqu32 `16*4-64`($ctx),${S2}{%k2}{z}
vmovdqu32 `16*6-64`($ctx),%x#$S3 vmovdqu32 `16*5-64`($ctx),${R3}{%k2}{z}
vmovdqu32 `16*7-64`($ctx),%x#$R4 vmovdqu32 `16*6-64`($ctx),${S3}{%k2}{z}
vmovdqu32 `16*8-64`($ctx),%x#$S4 vmovdqu32 `16*7-64`($ctx),${R4}{%k2}{z}
vmovdqu32 `16*8-64`($ctx),${S4}{%k2}{z}
vpermd $R0,$T2,$R0 # 00003412 -> 14243444 vpermd $R0,$T2,$R0 # 00003412 -> 14243444
vmovdqa64 64(%rcx),$MASK # .Lmask26 vpbroadcastq 64(%rcx),$MASK # .Lmask26
vpermd $R1,$T2,$R1 vpermd $R1,$T2,$R1
vpermd $S1,$T2,$S1 vpermd $S1,$T2,$S1
vpermd $R2,$T2,$R2 vpermd $R2,$T2,$R2
vmovdqa32 $R0,0x00(%rsp) # save in case $len%128 != 0 vmovdqa64 $R0,0x00(%rsp){%k2} # save in case $len%128 != 0
vpsrlq \$32,$R0,$T0 # 14243444 -> 01020304 vpsrlq \$32,$R0,$T0 # 14243444 -> 01020304
vpermd $S2,$T2,$S2 vpermd $S2,$T2,$S2
vmovdqa32 $R1,0x20(%rsp) vmovdqu64 $R1,0x00(%rsp,%rax){%k2}
vpsrlq \$32,$R1,$T1 vpsrlq \$32,$R1,$T1
vpermd $R3,$T2,$R3 vpermd $R3,$T2,$R3
vmovdqa32 $S1,0x40(%rsp) vmovdqa64 $S1,0x40(%rsp){%k2}
vpermd $S3,$T2,$S3 vpermd $S3,$T2,$S3
vpermd $R4,$T2,$R4 vpermd $R4,$T2,$R4
vmovdqa32 $R2,0x60(%rsp) vmovdqu64 $R2,0x40(%rsp,%rax){%k2}
vpermd $S4,$T2,$S4 vpermd $S4,$T2,$S4
vmovdqa32 $S2,0x80(%rsp) vmovdqa64 $S2,0x80(%rsp){%k2}
vmovdqa32 $R3,0xa0(%rsp) vmovdqu64 $R3,0x80(%rsp,%rax){%k2}
vmovdqa32 $S3,0xc0(%rsp) vmovdqa64 $S3,0xc0(%rsp){%k2}
vmovdqa32 $R4,0xe0(%rsp) vmovdqu64 $R4,0xc0(%rsp,%rax){%k2}
vmovdqa32 $S4,0x100(%rsp) vmovdqa64 $S4,0x100(%rsp){%k2}
################################################################ ################################################################
# calculate 5th through 8th powers of the key # calculate 5th through 8th powers of the key
...@@ -2282,14 +2295,6 @@ $code.=<<___; ...@@ -2282,14 +2295,6 @@ $code.=<<___;
vpandq $MASK,$D3,$D3 vpandq $MASK,$D3,$D3
vpaddq $M3,$D4,$D4 # d3 -> d4 vpaddq $M3,$D4,$D4 # d3 -> d4
___
map(s/%y/%z/,($T4,$T0,$T1,$T2,$T3)); # switch to %zmm domain
map(s/%y/%z/,($M4,$M0,$M1,$M2,$M3));
map(s/%y/%z/,($D0,$D1,$D2,$D3,$D4));
map(s/%y/%z/,($R0,$R1,$R2,$R3,$R4, $S1,$S2,$S3,$S4));
map(s/%y/%z/,($H0,$H1,$H2,$H3,$H4));
map(s/%y/%z/,($MASK));
$code.=<<___;
################################################################ ################################################################
# at this point we have 14243444 in $R0-$S4 and 05060708 in # at this point we have 14243444 in $R0-$S4 and 05060708 in
# $D0-$D4, ... # $D0-$D4, ...
...@@ -2327,7 +2332,6 @@ $code.=<<___; ...@@ -2327,7 +2332,6 @@ $code.=<<___;
vpaddd $R3,$S3,$S3 vpaddd $R3,$S3,$S3
vpaddd $R4,$S4,$S4 vpaddd $R4,$S4,$S4
vpbroadcastq %x#$MASK,$MASK
vpbroadcastq 32(%rcx),$PADBIT # .L129 vpbroadcastq 32(%rcx),$PADBIT # .L129
vpsrlq \$52,$T0,$T2 # splat input vpsrlq \$52,$T0,$T2 # splat input
...@@ -2345,7 +2349,7 @@ $code.=<<___; ...@@ -2345,7 +2349,7 @@ $code.=<<___;
vpaddq $H2,$T2,$H2 # accumulate input vpaddq $H2,$T2,$H2 # accumulate input
sub \$192,$len sub \$192,$len
jbe .Ltail_avx512 jbe .Ltail_avx512
#jmp .Loop_avx512 jmp .Loop_avx512
.align 32 .align 32
.Loop_avx512: .Loop_avx512:
...@@ -2532,7 +2536,7 @@ $code.=<<___; ...@@ -2532,7 +2536,7 @@ $code.=<<___;
vpaddq $H3,$T3,$H3 vpaddq $H3,$T3,$H3
vpaddq $H4,$T4,$H4 vpaddq $H4,$T4,$H4
vmovdqu64 16*0($inp),%x#$T0 vmovdqu 16*0($inp),%x#$T0
vpmuludq $H0,$R3,$M3 vpmuludq $H0,$R3,$M3
vpmuludq $H0,$R4,$M4 vpmuludq $H0,$R4,$M4
vpmuludq $H0,$R0,$M0 vpmuludq $H0,$R0,$M0
...@@ -2542,7 +2546,7 @@ $code.=<<___; ...@@ -2542,7 +2546,7 @@ $code.=<<___;
vpaddq $M0,$D0,$D0 # d0 += h0*r0 vpaddq $M0,$D0,$D0 # d0 += h0*r0
vpaddq $M1,$D1,$D1 # d1 += h0*r1 vpaddq $M1,$D1,$D1 # d1 += h0*r1
vmovdqu64 16*1($inp),%x#$T1 vmovdqu 16*1($inp),%x#$T1
vpmuludq $H1,$R2,$M3 vpmuludq $H1,$R2,$M3
vpmuludq $H1,$R3,$M4 vpmuludq $H1,$R3,$M4
vpmuludq $H1,$S4,$M0 vpmuludq $H1,$S4,$M0
...@@ -2552,7 +2556,7 @@ $code.=<<___; ...@@ -2552,7 +2556,7 @@ $code.=<<___;
vpaddq $M0,$D0,$D0 # d0 += h1*s4 vpaddq $M0,$D0,$D0 # d0 += h1*s4
vpaddq $M2,$D2,$D2 # d2 += h0*r2 vpaddq $M2,$D2,$D2 # d2 += h0*r2
vinserti64x2 \$1,16*2($inp),$T0,$T0 vinserti128 \$1,16*2($inp),%y#$T0,%y#$T0
vpmuludq $H3,$R0,$M3 vpmuludq $H3,$R0,$M3
vpmuludq $H3,$R1,$M4 vpmuludq $H3,$R1,$M4
vpmuludq $H1,$R0,$M1 vpmuludq $H1,$R0,$M1
...@@ -2562,7 +2566,7 @@ $code.=<<___; ...@@ -2562,7 +2566,7 @@ $code.=<<___;
vpaddq $M1,$D1,$D1 # d1 += h1*r0 vpaddq $M1,$D1,$D1 # d1 += h1*r0
vpaddq $M2,$D2,$D2 # d2 += h1*r1 vpaddq $M2,$D2,$D2 # d2 += h1*r1
vinserti64x2 \$1,16*3($inp),$T1,$T1 vinserti128 \$1,16*3($inp),%y#$T1,%y#$T1
vpmuludq $H4,$S4,$M3 vpmuludq $H4,$S4,$M3
vpmuludq $H4,$R0,$M4 vpmuludq $H4,$R0,$M4
vpmuludq $H3,$S2,$M0 vpmuludq $H3,$S2,$M0
...@@ -2585,11 +2589,11 @@ $code.=<<___; ...@@ -2585,11 +2589,11 @@ $code.=<<___;
# horizontal addition # horizontal addition
mov \$1,%eax mov \$1,%eax
vpsrldq \$8,$H3,$D3 vpermq \$0xb1,$H3,$D3
vpsrldq \$8,$D4,$H4 vpermq \$0xb1,$D4,$H4
vpsrldq \$8,$H0,$D0 vpermq \$0xb1,$H0,$D0
vpsrldq \$8,$H1,$D1 vpermq \$0xb1,$H1,$D1
vpsrldq \$8,$H2,$D2 vpermq \$0xb1,$H2,$D2
vpaddq $D3,$H3,$H3 vpaddq $D3,$H3,$H3
vpaddq $D4,$H4,$H4 vpaddq $D4,$H4,$H4
vpaddq $D0,$H0,$H0 vpaddq $D0,$H0,$H0
...@@ -2626,23 +2630,23 @@ $code.=<<___; ...@@ -2626,23 +2630,23 @@ $code.=<<___;
# lazy reduction (interleaved with input splat) # lazy reduction (interleaved with input splat)
vpsrlq \$26,$H3,$D3 vpsrlq \$26,$H3,$D3
vpandq $MASK,$H3,$H3 vpand $MASK,$H3,$H3
vpsrldq \$6,$T0,$T2 # splat input vpsrldq \$6,$T0,$T2 # splat input
vpsrldq \$6,$T1,$T3 vpsrldq \$6,$T1,$T3
vpunpckhqdq $T1,$T0,$T4 # 4 vpunpckhqdq $T1,$T0,$T4 # 4
vpaddq $D3,$H4,$H4 # h3 -> h4 vpaddq $D3,$H4,$H4 # h3 -> h4
vpsrlq \$26,$H0,$D0 vpsrlq \$26,$H0,$D0
vpandq $MASK,$H0,$H0 vpand $MASK,$H0,$H0
vpunpcklqdq $T3,$T2,$T2 # 2:3 vpunpcklqdq $T3,$T2,$T2 # 2:3
vpunpcklqdq $T1,$T0,$T0 # 0:1 vpunpcklqdq $T1,$T0,$T0 # 0:1
vpaddq $D0,$H1,$H1 # h0 -> h1 vpaddq $D0,$H1,$H1 # h0 -> h1
vpsrlq \$26,$H4,$D4 vpsrlq \$26,$H4,$D4
vpandq $MASK,$H4,$H4 vpand $MASK,$H4,$H4
vpsrlq \$26,$H1,$D1 vpsrlq \$26,$H1,$D1
vpandq $MASK,$H1,$H1 vpand $MASK,$H1,$H1
vpsrlq \$30,$T2,$T3 vpsrlq \$30,$T2,$T3
vpsrlq \$4,$T2,$T2 vpsrlq \$4,$T2,$T2
vpaddq $D1,$H2,$H2 # h1 -> h2 vpaddq $D1,$H2,$H2 # h1 -> h2
...@@ -2654,21 +2658,21 @@ $code.=<<___; ...@@ -2654,21 +2658,21 @@ $code.=<<___;
vpaddq $D4,$H0,$H0 # h4 -> h0 vpaddq $D4,$H0,$H0 # h4 -> h0
vpsrlq \$26,$H2,$D2 vpsrlq \$26,$H2,$D2
vpandq $MASK,$H2,$H2 vpand $MASK,$H2,$H2
vpandq $MASK,$T2,$T2 # 2 vpand $MASK,$T2,$T2 # 2
vpandq $MASK,$T0,$T0 # 0 vpand $MASK,$T0,$T0 # 0
vpaddq $D2,$H3,$H3 # h2 -> h3 vpaddq $D2,$H3,$H3 # h2 -> h3
vpsrlq \$26,$H0,$D0 vpsrlq \$26,$H0,$D0
vpandq $MASK,$H0,$H0 vpand $MASK,$H0,$H0
vpaddq $H2,$T2,$H2 # accumulate input for .Ltail_avx2 vpaddq $H2,$T2,$H2 # accumulate input for .Ltail_avx2
vpandq $MASK,$T1,$T1 # 1 vpand $MASK,$T1,$T1 # 1
vpaddq $D0,$H1,$H1 # h0 -> h1 vpaddq $D0,$H1,$H1 # h0 -> h1
vpsrlq \$26,$H3,$D3 vpsrlq \$26,$H3,$D3
vpandq $MASK,$H3,$H3 vpand $MASK,$H3,$H3
vpandq $MASK,$T3,$T3 # 3 vpand $MASK,$T3,$T3 # 3
vporq $PADBIT,$T4,$T4 # padbit, yes, always vpor 32(%rcx),$T4,$T4 # padbit, yes, always
vpaddq $D3,$H4,$H4 # h3 -> h4 vpaddq $D3,$H4,$H4 # h3 -> h4
lea 0x90(%rsp),%rax # size optimization for .Ltail_avx2 lea 0x90(%rsp),%rax # size optimization for .Ltail_avx2
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册