[PATCH] D157874: AMDGPU: Replace some libcalls with intrinsics

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added inline comments.


================
Comment at: llvm/lib/Target/AMDGPU/AMDGPULibCalls.cpp:601-615
+    case AMDGPULibFunc::EI_LOG:
+      if (FMF.none())
+        return false;
+      return tryReplaceLibcallWithSimpleIntrinsic(CI, Intrinsic::log,
+                                                  FMF.approxFunc());
+    case AMDGPULibFunc::EI_LOG2:
+      if (FMF.none())
----------------
deferring the log handling since this exposed some failures when using fast math for denormal values (apparently the conformance test enforces some accuracy limit with fast math?)


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D157874/new/

https://reviews.llvm.org/D157874