This is the mail archive of the libc-alpha@sourceware.org mailing list for the glibc project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] BZ #14059 - HAS_FMA4 check needs to also check for AVX


On Mon, May 14, 2012 at 4:12 AM, Andreas Jaeger <aj@suse.com> wrote:
> On Monday, May 14, 2012 09:52:17 Andreas Jaeger wrote:
>> Btw. here's a small test program to show that the values are setup
>> correctly.
>>
>> Currently it fails on my system (/proc/cpuinfo contains avx but
>> glibc test reports it does not).
>>
>> Hope it's usefull,
>
> Yes, was helpfull for me;)
>
> I found the problem, the assignments to set the bits are wrong.
>
> Instead of
> __cpu_features.feature[index_AVX_Usable] = 1;
>
> use
> __cpu_features.feature[index_AVX_Usable]
> ? ? ? ? ? ?|= bit_AVX_Usable;
>
> The same for the FMA4 test.
>
> Carlos, I suggest to add the test to glibc as well,

OK, so the next step is:

* Fix the mutual recursion issue.
* Fix the |= vs. = issue with AVX_Usable.
* Move the FMA4 test up into the AVX test we already have.
* Roll the extra test into the testsuite.

I'll do this, retest on an system with and and without AVX and report back.

Cheers,
Carlos.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]