* Re: Expose interfaces to allow declaring SIMD kernels
[not found] <Sid0gjWNkgnLAKOzvwf5w-yNAzjM-XrmESFPuLicgEyXwTjDVYZjZ7yr393uWwMApOqTIVrhelnVUNFCbompwsMyQmah3Bo0b1fxDTQVUyc=@zhuyi.fan>
@ 2023-10-24 21:37 ` Antoni Boucher
0 siblings, 0 replies; only message in thread
From: Antoni Boucher @ 2023-10-24 21:37 UTC (permalink / raw)
To: Schrodinger ZHU Yifan, jit; +Cc: Guillaume Gomez
Hi.
We do have a plan to add support for function attributes soon.
The base of the patch can be seen here:
https://github.com/antoyo/gcc/commit/96eed4fe2949535c05473a9c11fcc267ba94bca5
Would this interface be enough for your use case?
Thanks.
On Tue, 2023-10-24 at 21:33 +0000, Schrodinger ZHU Yifan wrote:
> Hi,
>
> Currently, there is no proper way in libgccjit to get/declare
> vectorized variants of a function. I have implemented a working patch
> to allow user to add “omp declare simd” attribute via
> gcc_jit_function_set_bool_declare_simd.
>
> I wonder if this a good approach? To me, this approach is a little
> bit too ad-hoc. Should we provide interfaces that allow user to set
> any attributes to a function instead?
>
> If you think it is fine to
> have gcc_jit_function_set_bool_declare_simd, could you please offer
> me some pointers to the guideline of patch submission?
>
> Best,
> Yifan
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2023-10-24 21:37 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
[not found] <Sid0gjWNkgnLAKOzvwf5w-yNAzjM-XrmESFPuLicgEyXwTjDVYZjZ7yr393uWwMApOqTIVrhelnVUNFCbompwsMyQmah3Bo0b1fxDTQVUyc=@zhuyi.fan>
2023-10-24 21:37 ` Expose interfaces to allow declaring SIMD kernels Antoni Boucher
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).