================
@@ -0,0 +1,3 @@
+_CLC_OVERLOAD _CLC_DEF __CLC_GENTYPE __clc_native_log(__CLC_GENTYPE val) {
+ return __clc_native_log2(val) * (1.0f / M_LOG2E_F);
----------------
frasercrmck wrote:
We now apply `-fapprox-func`, as of #133119.
If I change this PR to remove the AMDGPU definitions of `__clc_native_log`, and
instead rely on the generic one, we see, e.g., in `amdgcn--amdhsa.bc`:
``` llvm
; Function Attrs: alwaysinline mustprogress nofree norecurse nosync nounwind
willreturn memory(none)
define linkonce_odr hidden noundef <2 x float> @_Z10native_logDv2_f(<2 x float>
noundef %a) local_unnamed_addr #13 {
entry:
%elt.log.i = tail call afn noundef <2 x float> @llvm.log.v2f32(<2 x float> %a)
ret <2 x float> %elt.log.i
}
```
Is this what you had in mind?
https://github.com/llvm/llvm-project/pull/129679
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
