[clang] [X86][AVX10.2] Use 's_' for saturate-convert intrinsics (PR #131592)

Phoebe Wang via cfe-commits cfe-commits at lists.llvm.org
Thu Mar 20 02:02:19 PDT 2025


phoebewang wrote:

> > > > Went through the patch with my eyes, I'm not sure how useful it was but wasn't able to spot anything.
> > > > > Here is GCC patch: https://gcc.gnu.org/pipermail/gcc-patches/2025-March/678227.html
> > > > 
> > > > 
> > > > I haven't found any changes regarding `vcvt2ph2` and no other similar threads. Have I missed it? Or it will be next step?
> > > 
> > > 
> > > No idea. @Rejur @jianghc724
> > 
> > 
> > That is missing since it is not in satcvt, GCC will need to add that.
> 
> I have a question, do we need to modify `VCVTBIASPH2BF8S`'s intrin name to cvtbiass_ph_bf8.

Good question! Discussed with @jianghc724, we prefer to `cvts_biassph_hf8`

https://github.com/llvm/llvm-project/pull/131592


More information about the cfe-commits mailing list