This is the mail archive of the 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] PPC64: First in the series of patches implementing POWER8 vector math.

> Please include the patch description in the body of your message, not just
> in an attachment - "Empty Message" isn't helpful. When sending multiple
> patch versions, it's very helpful to describe what has changed since the
> previous version (after a "---" line or some other such indication,
> accepted by "git am", that the following text is not intended as part of
> the commit message).

The next patch sent will have a description of differences from the one we
are discussing.

> I think you need to have explicit buy-in from powerpc toolchain
> maintainers that the ABI you have chosen is the desired one for this
> purpose, agreed for use by any toolchain that wishes to be able to use
> this functionality.

Tulio and/or Bill:
Can you confirm that the POWER8 Vector Function ABI will match the
SSE ISA sections of x86_64 Vector Function ABI?

>Please also confirm the GCC version that implements
> the ABI in question (when given the pragmas / attributes in the header you
> add). To be clear, this information should be part of the commit message
> (in every version of the patch submission), not just in a one-off reply to
> this message.

Patches going foward will include information on the minimum GCC version
required for POWER8 libmvec.


Tulio and/or Bill:
Please affirm my understanding that at present GCC does NOT
yet implement the Vector Function ABI for POWER8.

> Please also reference somewhere other than Google Groups for the ABI (both
> because Google Groups requires non-free JavaScript, contrary to GNU
> principles, and because Google Groups has a history of breaking URLs that
> used to work for linking to messages, so somewhere with a more reliably
> stable URL is important). The glibc wiki has a copy of the x86_64 ABI at

Next patch will replace link to google groups with the glibc url you've

> Please describe in patch submissions how the patch was tested. In this
> case, running at least the libm tests for both powerpc64 big-endian and
> powerpc64 little-endian, and verifying there are no failures, would seem
> appropriate.

Testing will be described all patches.

Remaining issues you raised will be answered shortly.

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