This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH 4/4] Remove powerpc64 strspn, strcspn, and strpbrk implementation
- From: Adhemerval Zanella <adhemerval dot zanella at linaro dot org>
- To: Tulio Magno Quites Machado Filho <tuliom at linux dot vnet dot ibm dot com>, libc-alpha at sourceware dot org
- Date: Wed, 30 Mar 2016 14:06:41 -0300
- Subject: Re: [PATCH 4/4] Remove powerpc64 strspn, strcspn, and strpbrk implementation
- Authentication-results: sourceware.org; auth=none
- References: <1459178389-14133-1-git-send-email-adhemerval dot zanella at linaro dot org> <1459178389-14133-5-git-send-email-adhemerval dot zanella at linaro dot org> <87r3es9kgb dot fsf at totoro dot br dot ibm dot com>
On 30-03-2016 10:14, Tulio Magno Quites Machado Filho wrote:
> Adhemerval Zanella <adhemerval.zanella@linaro.org> writes:
>
>> This patch removes the powerpc64 optimized strspn, strcspn, and
>> strpbrk assembly implementation now that the default C one
>> implements the same strategy. On internal glibc benchtests
>> current implementations shows similar performance with -O2.
>>
>> Tested on powerpc64le (POWER8).
>>
>> * sysdeps/powerpc/powerpc64/strspn.S: Likewise.
>
> I tested this on POWER7 as well. Although it isn't as good as in POWER8,
> it's competitive. I agree to remove it.
>
>> * sysdeps/powerpc/powerpc64/strcspn.S: Remove file.
>> * sysdeps/powerpc/powerpc64/strpbrk.S: Likewise.
>
> But the removal of these 2 brings some serious performance degradation.
> I think we should keep them for now.
Do you mean regression on POWER7 only or for POWER8 as well?
And what do you mean by 'serious'? I am asking because the
algorithm is essentially the same, with the slight difference
powerpc64 version does not align neither try to read four
bytes at time. I will expect some degradation for strpbrk, since
generic call now is calling strcsn.