[BZ,#20115] Extra alignment in memset-vec-unaligned-erms.S

X86-64 memset-vec-unaligned-erms.S aligns many jump targets, which
increases code sizes, but not necessarily improve performance.  As
memset benchtest data of align vs no align on various Intel and AMD
processors

https://sourceware.org/bugzilla/attachment.cgi?id=9277

shows that aligning jump targets isn't necessary.

Any comments, feedbacks?


H.J.
---
	[BZ #20115]
	* sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S (__memset):
	Remove alignments on jump targets.
---
 .../x86_64/multiarch/memset-vec-unaligned-erms.S   | 37 +++-------------------
 1 file changed, 5 insertions(+), 32 deletions(-)

On Wed, May 18, 2016 at 1:54 PM, H.J. Lu <hongjiu.lu@intel.com> wrote:
> X86-64 memset-vec-unaligned-erms.S aligns many jump targets, which
> increases code sizes, but not necessarily improve performance.  As
> memset benchtest data of align vs no align on various Intel and AMD
> processors
>
> https://sourceware.org/bugzilla/attachment.cgi?id=9277
>
> shows that aligning jump targets isn't necessary.
>
> Any comments, feedbacks?
>
>
> H.J.
> ---
>         [BZ #20115]
>         * sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S (__memset):
>         Remove alignments on jump targets.
> ---
>  .../x86_64/multiarch/memset-vec-unaligned-erms.S   | 37 +++-------------------
>  1 file changed, 5 insertions(+), 32 deletions(-)
>
> diff --git a/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S b/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S
> index 578a5ae..b1df228 100644
> --- a/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S
> +++ b/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S
> @@ -18,12 +18,10 @@
>
>  /* memset is implemented as:
>     1. Use overlapping store to avoid branch.
> -   2. Force 32-bit displacement for branches to avoid long nop between
> -      instructions.
> -   3. If size is less than VEC, use integer register stores.
> -   4. If size is from VEC_SIZE to 2 * VEC_SIZE, use 2 VEC stores.
> -   5. If size is from 2 * VEC_SIZE to 4 * VEC_SIZE, use 4 VEC stores.
> -   6. If size is more to 4 * VEC_SIZE, align to 4 * VEC_SIZE with
> +   2. If size is less than VEC, use integer register stores.
> +   3. If size is from VEC_SIZE to 2 * VEC_SIZE, use 2 VEC stores.
> +   4. If size is from 2 * VEC_SIZE to 4 * VEC_SIZE, use 4 VEC stores.
> +   5. If size is more to 4 * VEC_SIZE, align to 4 * VEC_SIZE with
>        4 VEC stores and store 4 * VEC at a time until done.  */
>
>  #include <sysdep.h>
> @@ -143,14 +141,10 @@ ENTRY (MEMSET_SYMBOL (__memset, unaligned_erms))
>         VZEROUPPER
>         ret
>
> -       .p2align 4
>  L(stosb_more_2x_vec):
>         cmpq    $REP_STOSB_THRESHOLD, %rdx
> -       /* Force 32-bit displacement to avoid long nop between
> -          instructions.  */
> -       ja.d32  L(stosb)
> +       ja      L(stosb)
>  #endif
> -       .p2align 4
>  L(more_2x_vec):
>         cmpq  $(VEC_SIZE * 4), %rdx
>         ja      L(loop_start)
> @@ -162,26 +156,12 @@ L(return):
>         VZEROUPPER
>         ret
>
> -       .p2align 4
>  L(loop_start):
>         leaq    (VEC_SIZE * 4)(%rdi), %rcx
> -# if VEC_SIZE == 32 || VEC_SIZE == 64
> -       /* Force 32-bit displacement to avoid long nop between
> -          instructions.  */
> -       VMOVU.d32 %VEC(0), (%rdi)
> -# else
>         VMOVU   %VEC(0), (%rdi)
> -# endif
>         andq    $-(VEC_SIZE * 4), %rcx
> -# if VEC_SIZE == 32
> -       /* Force 32-bit displacement to avoid long nop between
> -          instructions.  */
> -       VMOVU.d32 %VEC(0), -VEC_SIZE(%rdi,%rdx)
> -       VMOVU.d32 %VEC(0), VEC_SIZE(%rdi)
> -# else
>         VMOVU   %VEC(0), -VEC_SIZE(%rdi,%rdx)
>         VMOVU   %VEC(0), VEC_SIZE(%rdi)
> -# endif
>         VMOVU   %VEC(0), -(VEC_SIZE * 2)(%rdi,%rdx)
>         VMOVU   %VEC(0), (VEC_SIZE * 2)(%rdi)
>         VMOVU   %VEC(0), -(VEC_SIZE * 3)(%rdi,%rdx)
> @@ -190,14 +170,7 @@ L(loop_start):
>         addq    %rdi, %rdx
>         andq    $-(VEC_SIZE * 4), %rdx
>         cmpq    %rdx, %rcx
> -# if VEC_SIZE == 32 || VEC_SIZE == 64
> -       /* Force 32-bit displacement to avoid long nop between
> -          instructions.  */
> -       je.d32  L(return)
> -# else
>         je      L(return)
> -# endif
> -       .p2align 4
>  L(loop):
>         VMOVA   %VEC(0), (%rcx)
>         VMOVA   %VEC(0), VEC_SIZE(%rcx)
> --
> 2.5.5
>

I am checking in this.