[llvm] Adding more vector calls for -fveclib=AMDLIBM (PR #109662)

[Simon Pilgrim via llvm-commits]

================
@@ -1600,4 +1653,163 @@ for.end:
   ret void
 }
 
+define void @exp10_f64(ptr nocapture %varray) {
+; CHECK-LABEL: @exp10_f64(
+; CHECK-VF2:    [[TMP5:%.*]] = call <2 x double> @amd_vrd2_exp10(<2 x double> [[TMP4:%.*]])
+; CHECK-VF4-NOT:    call <4 x double> @amd_vrd4_exp10(<4 x double> [[TMP4:%.*]])
----------------
RKSimon wrote:

Are we just missing exp10 handling in isTriviallyVectorizable?

https://github.com/llvm/llvm-project/pull/109662