diff mbox series

x86-64: Fix SSE2 memcmp and SSSE3 memmove for x32

Message ID 20220422175800.1262332-1-hjl.tools@gmail.com
State Committed
Commit 8ea20ee5f6145de4bff9481d3e09ac36ba9df8f3
Headers
Series x86-64: Fix SSE2 memcmp and SSSE3 memmove for x32 |

Checks

Context Check Description
dj/TryBot-apply_patch success Patch applied to master at the time it was sent
dj/TryBot-32bit success Build for i686

Commit Message

H.J. Lu April 22, 2022, 5:58 p.m. UTC 
  Clear the upper 32 bits in RDX (memory size) for x32 to fix

FAIL: string/tst-size_t-memcmp
FAIL: string/tst-size_t-memcmp-2
FAIL: string/tst-size_t-memcpy
FAIL: wcsmbs/tst-size_t-wmemcmp

on x32 introduced by

8804157ad9 x86: Optimize memcmp SSE2 in memcmp.S
26b2478322 x86: Reduce code size of mem{move|pcpy|cpy}-ssse3
---
 sysdeps/x86_64/memcmp.S                  | 4 ++++
 sysdeps/x86_64/multiarch/memmove-ssse3.S | 4 ++++
 2 files changed, 8 insertions(+)

Comments

Noah Goldstein April 22, 2022, 6:06 p.m. UTC | #1 
On Fri, Apr 22, 2022 at 12:58 PM H.J. Lu <hjl.tools@gmail.com> wrote:
>
> Clear the upper 32 bits in RDX (memory size) for x32 to fix
>
> FAIL: string/tst-size_t-memcmp
> FAIL: string/tst-size_t-memcmp-2
> FAIL: string/tst-size_t-memcpy
> FAIL: wcsmbs/tst-size_t-wmemcmp
>
> on x32 introduced by
>
> 8804157ad9 x86: Optimize memcmp SSE2 in memcmp.S
> 26b2478322 x86: Reduce code size of mem{move|pcpy|cpy}-ssse3
> ---
>  sysdeps/x86_64/memcmp.S                  | 4 ++++
>  sysdeps/x86_64/multiarch/memmove-ssse3.S | 4 ++++
>  2 files changed, 8 insertions(+)
>
> diff --git a/sysdeps/x86_64/memcmp.S b/sysdeps/x86_64/memcmp.S
> index b153694048..5718a7da86 100644
> --- a/sysdeps/x86_64/memcmp.S
> +++ b/sysdeps/x86_64/memcmp.S
> @@ -46,6 +46,10 @@
>
>         .text
>  ENTRY(MEMCMP)
> +# ifdef __ILP32__
> +       /* Clear the upper 32 bits.  */
> +       movl    %edx, %edx
> +# endif
>  #ifdef USE_AS_WMEMCMP
>         /* Use 0xffff to test for mismatches on pmovmskb bitmask. Store
>            in ecx for code size. This is preferable to using `incw` as
> diff --git a/sysdeps/x86_64/multiarch/memmove-ssse3.S b/sysdeps/x86_64/multiarch/memmove-ssse3.S
> index 215583e7bd..310ff62b86 100644
> --- a/sysdeps/x86_64/multiarch/memmove-ssse3.S
> +++ b/sysdeps/x86_64/multiarch/memmove-ssse3.S
> @@ -27,6 +27,10 @@ ENTRY(MEMMOVE_CHK)
>  END(MEMMOVE_CHK)
>
>  ENTRY_P2ALIGN(MEMMOVE, 6)
> +# ifdef __ILP32__
> +       /* Clear the upper 32 bits.  */
> +       movl    %edx, %edx
> +# endif
>         movq    %rdi, %rax
>  L(start):
>         cmpq    $16, %rdx
> --
> 2.35.1
>

LGTM.
Sunil Pandey May 12, 2022, 8:12 p.m. UTC | #2 
On Fri, Apr 22, 2022 at 11:07 AM Noah Goldstein via Libc-alpha
<libc-alpha@sourceware.org> wrote:
>
> On Fri, Apr 22, 2022 at 12:58 PM H.J. Lu <hjl.tools@gmail.com> wrote:
> >
> > Clear the upper 32 bits in RDX (memory size) for x32 to fix
> >
> > FAIL: string/tst-size_t-memcmp
> > FAIL: string/tst-size_t-memcmp-2
> > FAIL: string/tst-size_t-memcpy
> > FAIL: wcsmbs/tst-size_t-wmemcmp
> >
> > on x32 introduced by
> >
> > 8804157ad9 x86: Optimize memcmp SSE2 in memcmp.S
> > 26b2478322 x86: Reduce code size of mem{move|pcpy|cpy}-ssse3
> > ---
> >  sysdeps/x86_64/memcmp.S                  | 4 ++++
> >  sysdeps/x86_64/multiarch/memmove-ssse3.S | 4 ++++
> >  2 files changed, 8 insertions(+)
> >
> > diff --git a/sysdeps/x86_64/memcmp.S b/sysdeps/x86_64/memcmp.S
> > index b153694048..5718a7da86 100644
> > --- a/sysdeps/x86_64/memcmp.S
> > +++ b/sysdeps/x86_64/memcmp.S
> > @@ -46,6 +46,10 @@
> >
> >         .text
> >  ENTRY(MEMCMP)
> > +# ifdef __ILP32__
> > +       /* Clear the upper 32 bits.  */
> > +       movl    %edx, %edx
> > +# endif
> >  #ifdef USE_AS_WMEMCMP
> >         /* Use 0xffff to test for mismatches on pmovmskb bitmask. Store
> >            in ecx for code size. This is preferable to using `incw` as
> > diff --git a/sysdeps/x86_64/multiarch/memmove-ssse3.S b/sysdeps/x86_64/multiarch/memmove-ssse3.S
> > index 215583e7bd..310ff62b86 100644
> > --- a/sysdeps/x86_64/multiarch/memmove-ssse3.S
> > +++ b/sysdeps/x86_64/multiarch/memmove-ssse3.S
> > @@ -27,6 +27,10 @@ ENTRY(MEMMOVE_CHK)
> >  END(MEMMOVE_CHK)
> >
> >  ENTRY_P2ALIGN(MEMMOVE, 6)
> > +# ifdef __ILP32__
> > +       /* Clear the upper 32 bits.  */
> > +       movl    %edx, %edx
> > +# endif
> >         movq    %rdi, %rax
> >  L(start):
> >         cmpq    $16, %rdx
> > --
> > 2.35.1
> >
>
> LGTM.

I would like to backport this patch to release branches.
Any comments or objections?

Conflict resolution patch attached.

--Sunil
diff mbox series

Patch

diff --git a/sysdeps/x86_64/memcmp.S b/sysdeps/x86_64/memcmp.S
index b153694048..5718a7da86 100644
--- a/sysdeps/x86_64/memcmp.S
+++ b/sysdeps/x86_64/memcmp.S
@@ -46,6 +46,10 @@ 
 
 	.text
 ENTRY(MEMCMP)
+# ifdef __ILP32__
+	/* Clear the upper 32 bits.  */
+	movl	%edx, %edx
+# endif
 #ifdef USE_AS_WMEMCMP
 	/* Use 0xffff to test for mismatches on pmovmskb bitmask. Store
 	   in ecx for code size. This is preferable to using `incw` as
diff --git a/sysdeps/x86_64/multiarch/memmove-ssse3.S b/sysdeps/x86_64/multiarch/memmove-ssse3.S
index 215583e7bd..310ff62b86 100644
--- a/sysdeps/x86_64/multiarch/memmove-ssse3.S
+++ b/sysdeps/x86_64/multiarch/memmove-ssse3.S
@@ -27,6 +27,10 @@  ENTRY(MEMMOVE_CHK)
 END(MEMMOVE_CHK)
 
 ENTRY_P2ALIGN(MEMMOVE, 6)
+# ifdef __ILP32__
+	/* Clear the upper 32 bits.  */
+	movl	%edx, %edx
+# endif
 	movq	%rdi, %rax
 L(start):
 	cmpq	$16, %rdx