[cfe-dev] Adding a new pragma to Clang

Renato Golin renato.golin at linaro.org
Wed Feb 5 04:44:30 PST 2014


On 5 February 2014 12:27, Andrey Bokhanko <andreybokhanko at gmail.com> wrote:

> I agree -- "omp simd" pragma doesn't cover everything you proposed
> (certainly has nothing to do with unrolling). However, it is a part of an
> accepted and widely used standard. Thus, if we (meaning the whole
> community) want to introduce a method to control vectorization that we
> advocate to be employed by end users and not as a debug tool by llvm
> developers, this is, IMHO, the way to go.
>

Not a single argument against it. ;)


If you volunteer to do code review, we can prepare patches (that implement
> "omp simd" and rely on back-end support you implemented) and commit them.
>

I'm pretty sure others will as well, but feel free to copy me on the patch
reviews.

For an idea on what the metadata looks like, check the tests in
src/test/Transforms/LoopVectorize/
 * metadata-unroll.ll
 * metadata-width.ll
 * vectorize-once.ll
 * X86/metadata-enable.ll
 * X86/already-vectorized.ll

cheers,
--renato
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20140205/e74fba3a/attachment.html>


More information about the cfe-dev mailing list