[PATCH] D156414: [AMDGPU] Break Large PHIs: Take whole PHI chains into account

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added inline comments.


================
Comment at: llvm/lib/Target/AMDGPU/AMDGPUCodeGenPrepare.cpp:1828
+  // This threshold has been determined through testing.
+  const auto Threshold = ceil((WorkList.size() / 3.0) * 2);
+  unsigned NumBreakablePHIs = 0;
----------------
Try to avoid FP for this?


================
Comment at: llvm/lib/Target/AMDGPU/AMDGPUCodeGenPrepare.cpp:1833
+    // Metadata used by tests.
+    if (Cur->getMetadata("amdgpu.unbreakablephi"))
+      continue;
----------------
don't understand the point of the metadata, adding metadata only for a single test seems like a bad idea


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D156414/new/

https://reviews.llvm.org/D156414