* Re: [PATCH] crypto/x86: Use XORL r32,32 in poly1305-x86_64-cryptogams.pl
2020-08-27 17:38 [PATCH] crypto/x86: Use XORL r32,32 in poly1305-x86_64-cryptogams.pl Uros Bizjak
@ 2020-09-01 19:16 ` Jason A. Donenfeld
2020-09-02 5:52 ` Uros Bizjak
2020-09-07 13:16 ` Jason A. Donenfeld
2020-09-11 6:56 ` Herbert Xu
2 siblings, 1 reply; 5+ messages in thread
From: Jason A. Donenfeld @ 2020-09-01 19:16 UTC (permalink / raw)
To: Uros Bizjak, Andy Polyakov; +Cc: linux-crypto, x86, Herbert Xu, David S. Miller
Hi Uros,
Any benchmarks for this? Seems like it's all in initialization code,
right? I'm CC'ing Andy into this.
Jason
On Thu, Aug 27, 2020 at 07:38:31PM +0200, Uros Bizjak wrote:
> x86_64 zero extends 32bit operations, so for 64bit operands,
> XORL r32,r32 is functionally equal to XORQ r64,r64, but avoids
> a REX prefix byte when legacy registers are used.
>
> Signed-off-by: Uros Bizjak <ubizjak@gmail.com>
> Cc: Herbert Xu <herbert@gondor.apana.org.au>
> Cc: "David S. Miller" <davem@davemloft.net>
> ---
> arch/x86/crypto/poly1305-x86_64-cryptogams.pl | 8 ++++----
> 1 file changed, 4 insertions(+), 4 deletions(-)
>
> diff --git a/arch/x86/crypto/poly1305-x86_64-cryptogams.pl b/arch/x86/crypto/poly1305-x86_64-cryptogams.pl
> index 137edcf038cb..7d568012cc15 100644
> --- a/arch/x86/crypto/poly1305-x86_64-cryptogams.pl
> +++ b/arch/x86/crypto/poly1305-x86_64-cryptogams.pl
> @@ -246,7 +246,7 @@ $code.=<<___ if (!$kernel);
> ___
> &declare_function("poly1305_init_x86_64", 32, 3);
> $code.=<<___;
> - xor %rax,%rax
> + xor %eax,%eax
> mov %rax,0($ctx) # initialize hash value
> mov %rax,8($ctx)
> mov %rax,16($ctx)
> @@ -2853,7 +2853,7 @@ $code.=<<___;
> .type poly1305_init_base2_44,\@function,3
> .align 32
> poly1305_init_base2_44:
> - xor %rax,%rax
> + xor %eax,%eax
> mov %rax,0($ctx) # initialize hash value
> mov %rax,8($ctx)
> mov %rax,16($ctx)
> @@ -3947,7 +3947,7 @@ xor128_decrypt_n_pad:
> mov \$16,$len
> sub %r10,$len
> xor %eax,%eax
> - xor %r11,%r11
> + xor %r11d,%r11d
> .Loop_dec_byte:
> mov ($inp,$otp),%r11b
> mov ($otp),%al
> @@ -4085,7 +4085,7 @@ avx_handler:
> .long 0xa548f3fc # cld; rep movsq
>
> mov $disp,%rsi
> - xor %rcx,%rcx # arg1, UNW_FLAG_NHANDLER
> + xor %ecx,%ecx # arg1, UNW_FLAG_NHANDLER
> mov 8(%rsi),%rdx # arg2, disp->ImageBase
> mov 0(%rsi),%r8 # arg3, disp->ControlPc
> mov 16(%rsi),%r9 # arg4, disp->FunctionEntry
> --
> 2.26.2
>
--
Jason A. Donenfeld
Deep Space Explorer
fr: +33 6 51 90 82 66
us: +1 513 476 1200
www.jasondonenfeld.com
www.zx2c4.com
zx2c4.com/keys/AB9942E6D4A4CFC3412620A749FC7012A5DE03AE.asc
^ permalink raw reply [flat|nested] 5+ messages in thread* Re: [PATCH] crypto/x86: Use XORL r32,32 in poly1305-x86_64-cryptogams.pl
2020-09-01 19:16 ` Jason A. Donenfeld
@ 2020-09-02 5:52 ` Uros Bizjak
0 siblings, 0 replies; 5+ messages in thread
From: Uros Bizjak @ 2020-09-02 5:52 UTC (permalink / raw)
To: Jason A. Donenfeld
Cc: Andy Polyakov, Linux Crypto List, X86 ML, Herbert Xu, David S. Miller
On Tue, Sep 1, 2020 at 9:16 PM Jason A. Donenfeld <Jason@zx2c4.com> wrote:
>
> Hi Uros,
>
> Any benchmarks for this? Seems like it's all in initialization code,
> right? I'm CC'ing Andy into this.
This patch should have no performance effect, it saves REX prefix byte
when the optimization is applied to legacy registers.
Uros.
> Jason
>
> On Thu, Aug 27, 2020 at 07:38:31PM +0200, Uros Bizjak wrote:
> > x86_64 zero extends 32bit operations, so for 64bit operands,
> > XORL r32,r32 is functionally equal to XORQ r64,r64, but avoids
> > a REX prefix byte when legacy registers are used.
> >
> > Signed-off-by: Uros Bizjak <ubizjak@gmail.com>
> > Cc: Herbert Xu <herbert@gondor.apana.org.au>
> > Cc: "David S. Miller" <davem@davemloft.net>
> > ---
> > arch/x86/crypto/poly1305-x86_64-cryptogams.pl | 8 ++++----
> > 1 file changed, 4 insertions(+), 4 deletions(-)
> >
> > diff --git a/arch/x86/crypto/poly1305-x86_64-cryptogams.pl b/arch/x86/crypto/poly1305-x86_64-cryptogams.pl
> > index 137edcf038cb..7d568012cc15 100644
> > --- a/arch/x86/crypto/poly1305-x86_64-cryptogams.pl
> > +++ b/arch/x86/crypto/poly1305-x86_64-cryptogams.pl
> > @@ -246,7 +246,7 @@ $code.=<<___ if (!$kernel);
> > ___
> > &declare_function("poly1305_init_x86_64", 32, 3);
> > $code.=<<___;
> > - xor %rax,%rax
> > + xor %eax,%eax
> > mov %rax,0($ctx) # initialize hash value
> > mov %rax,8($ctx)
> > mov %rax,16($ctx)
> > @@ -2853,7 +2853,7 @@ $code.=<<___;
> > .type poly1305_init_base2_44,\@function,3
> > .align 32
> > poly1305_init_base2_44:
> > - xor %rax,%rax
> > + xor %eax,%eax
> > mov %rax,0($ctx) # initialize hash value
> > mov %rax,8($ctx)
> > mov %rax,16($ctx)
> > @@ -3947,7 +3947,7 @@ xor128_decrypt_n_pad:
> > mov \$16,$len
> > sub %r10,$len
> > xor %eax,%eax
> > - xor %r11,%r11
> > + xor %r11d,%r11d
> > .Loop_dec_byte:
> > mov ($inp,$otp),%r11b
> > mov ($otp),%al
> > @@ -4085,7 +4085,7 @@ avx_handler:
> > .long 0xa548f3fc # cld; rep movsq
> >
> > mov $disp,%rsi
> > - xor %rcx,%rcx # arg1, UNW_FLAG_NHANDLER
> > + xor %ecx,%ecx # arg1, UNW_FLAG_NHANDLER
> > mov 8(%rsi),%rdx # arg2, disp->ImageBase
> > mov 0(%rsi),%r8 # arg3, disp->ControlPc
> > mov 16(%rsi),%r9 # arg4, disp->FunctionEntry
> > --
> > 2.26.2
> >
>
> --
> Jason A. Donenfeld
> Deep Space Explorer
> fr: +33 6 51 90 82 66
> us: +1 513 476 1200
> www.jasondonenfeld.com
> www.zx2c4.com
> zx2c4.com/keys/AB9942E6D4A4CFC3412620A749FC7012A5DE03AE.asc
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] crypto/x86: Use XORL r32,32 in poly1305-x86_64-cryptogams.pl
2020-08-27 17:38 [PATCH] crypto/x86: Use XORL r32,32 in poly1305-x86_64-cryptogams.pl Uros Bizjak
2020-09-01 19:16 ` Jason A. Donenfeld
@ 2020-09-07 13:16 ` Jason A. Donenfeld
2020-09-11 6:56 ` Herbert Xu
2 siblings, 0 replies; 5+ messages in thread
From: Jason A. Donenfeld @ 2020-09-07 13:16 UTC (permalink / raw)
To: Uros Bizjak, herbert; +Cc: linux-crypto, x86, Herbert Xu, David S. Miller
Hi Uros, Herbert,
On Thu, Aug 27, 2020 at 07:38:31PM +0200, Uros Bizjak wrote:
> x86_64 zero extends 32bit operations, so for 64bit operands,
> XORL r32,r32 is functionally equal to XORQ r64,r64, but avoids
> a REX prefix byte when legacy registers are used.
>
> Signed-off-by: Uros Bizjak <ubizjak@gmail.com>
> Cc: Herbert Xu <herbert@gondor.apana.org.au>
> Cc: "David S. Miller" <davem@davemloft.net>
> ---
> arch/x86/crypto/poly1305-x86_64-cryptogams.pl | 8 ++++----
> 1 file changed, 4 insertions(+), 4 deletions(-)
>
> diff --git a/arch/x86/crypto/poly1305-x86_64-cryptogams.pl b/arch/x86/crypto/poly1305-x86_64-cryptogams.pl
> index 137edcf038cb..7d568012cc15 100644
> --- a/arch/x86/crypto/poly1305-x86_64-cryptogams.pl
> +++ b/arch/x86/crypto/poly1305-x86_64-cryptogams.pl
> @@ -246,7 +246,7 @@ $code.=<<___ if (!$kernel);
> ___
> &declare_function("poly1305_init_x86_64", 32, 3);
> $code.=<<___;
> - xor %rax,%rax
> + xor %eax,%eax
> mov %rax,0($ctx) # initialize hash value
> mov %rax,8($ctx)
> mov %rax,16($ctx)
> @@ -2853,7 +2853,7 @@ $code.=<<___;
> .type poly1305_init_base2_44,\@function,3
> .align 32
> poly1305_init_base2_44:
> - xor %rax,%rax
> + xor %eax,%eax
> mov %rax,0($ctx) # initialize hash value
> mov %rax,8($ctx)
> mov %rax,16($ctx)
> @@ -3947,7 +3947,7 @@ xor128_decrypt_n_pad:
> mov \$16,$len
> sub %r10,$len
> xor %eax,%eax
> - xor %r11,%r11
> + xor %r11d,%r11d
> .Loop_dec_byte:
> mov ($inp,$otp),%r11b
> mov ($otp),%al
> @@ -4085,7 +4085,7 @@ avx_handler:
> .long 0xa548f3fc # cld; rep movsq
>
> mov $disp,%rsi
> - xor %rcx,%rcx # arg1, UNW_FLAG_NHANDLER
> + xor %ecx,%ecx # arg1, UNW_FLAG_NHANDLER
> mov 8(%rsi),%rdx # arg2, disp->ImageBase
> mov 0(%rsi),%r8 # arg3, disp->ControlPc
> mov 16(%rsi),%r9 # arg4, disp->FunctionEntry
> --
> 2.26.2
>
Per the discussion elsewhere,
Acked-by: Jason A. Donenfeld <Jason@zx2c4.com>
for cryptodev-2.6.git, rather than crypto-2.6.git
Thanks,
Jason
^ permalink raw reply [flat|nested] 5+ messages in thread* Re: [PATCH] crypto/x86: Use XORL r32,32 in poly1305-x86_64-cryptogams.pl
2020-08-27 17:38 [PATCH] crypto/x86: Use XORL r32,32 in poly1305-x86_64-cryptogams.pl Uros Bizjak
2020-09-01 19:16 ` Jason A. Donenfeld
2020-09-07 13:16 ` Jason A. Donenfeld
@ 2020-09-11 6:56 ` Herbert Xu
2 siblings, 0 replies; 5+ messages in thread
From: Herbert Xu @ 2020-09-11 6:56 UTC (permalink / raw)
To: Uros Bizjak; +Cc: linux-crypto, x86, David S. Miller
On Thu, Aug 27, 2020 at 07:38:31PM +0200, Uros Bizjak wrote:
> x86_64 zero extends 32bit operations, so for 64bit operands,
> XORL r32,r32 is functionally equal to XORQ r64,r64, but avoids
> a REX prefix byte when legacy registers are used.
>
> Signed-off-by: Uros Bizjak <ubizjak@gmail.com>
> Cc: Herbert Xu <herbert@gondor.apana.org.au>
> Cc: "David S. Miller" <davem@davemloft.net>
> ---
> arch/x86/crypto/poly1305-x86_64-cryptogams.pl | 8 ++++----
> 1 file changed, 4 insertions(+), 4 deletions(-)
Patch applied. Thanks.
--
Email: Herbert Xu <herbert@gondor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
^ permalink raw reply [flat|nested] 5+ messages in thread