[PATCH] D63206: [AMDGPU] gfx1010 wavefrontsize intrinsic folding

Stanislav Mekhanoshin via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Wed Jun 12 08:33:18 PDT 2019


rampitec marked an inline comment as done.
rampitec added inline comments.


================
Comment at: lib/Target/AMDGPU/AMDGPULibCalls.cpp:1392-1414
+bool AMDGPULibCalls::fold_wavefrontsize(CallInst *CI, IRBuilder<> &B) {
+  if (!TM)
+    return false;
+
+  StringRef CPU = TM->getTargetCPU();
+  StringRef Features = TM->getTargetFeatureString();
+  if ((CPU.empty() || CPU.equals_lower("generic")) &&
----------------
arsenm wrote:
> arsenm wrote:
> > This doesn't really belong in this pass
> This might make more sense In AMDGPULowerIntrinsics?
The point is to give folding a chance as early as possible. It also belongs to libcalls because library is the primary user.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D63206/new/

https://reviews.llvm.org/D63206





More information about the llvm-commits mailing list