This is the mail archive of the libc-ports@sources.redhat.com mailing list for the libc-ports 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, v6] ARM: Add Cortex-A15 optimized NEON and VFP memcpy routines, with IFUNC.


On 4 May 2013 14:13, Joseph S. Myers <joseph@codesourcery.com> wrote:

Hi Joseph,

> On Fri, 3 May 2013, Will Newton wrote:
>
>> On 3 May 2013 21:18, Joseph S. Myers <joseph@codesourcery.com> wrote:
>>
>> Hi Joseph,
>>
>> >> This was tested on armv7l-unknown-linux-gnueabihf. One new testsuite
>> >
>> > I think string functions also need big-endian testing (which should be
>> > possible with userspace QEMU - you should be able to run the string tests
>> > that way, though not the whole testsuite).  If such testing passes, this
>> > version is OK (though as Roland notes there's scope for further
>> > optimization in the __ARM_NEON__ case).
>>
>> I have tested the same code using a big-endian qemu with the
>> cortex-strings tests which are taken from glibc:
>>
>> http://bazaar.launchpad.net/~linaro-toolchain-dev/cortex-strings/trunk/view/head:/tests/test-memcpy.c
>>
>> I haven't built glibc big-endian yet, but I can do that if you think
>> it would be valuable.
>
> Yes, I believe this should be tested with the current glibc string tests
> for a big-endian build of glibc.

I can confirm test-memcpy and test-memcpy-ifunc both pass under big endian qemu.


--
Will Newton
Toolchain Working Group, Linaro


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