frasercrmck wrote: > The amdgpu customization should be removed. This is legacy code from before > we had the ldexp intrinsic. Regular __builtin_ldexp* should work now
Good to know, thanks. No one else in libclc is using `__builtin_ldexp`, so in its current form it'd still have to be some sort of a customization. Everyone is opting for the software implementation. We could have `__builtin_ldexp*` in the default implementation, but guard it behind a macro that only AMD defines. Would we ideally have an `__builtin_elementwise_ldexp` to target the vector intrinsics directly? https://github.com/llvm/llvm-project/pull/126078 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
