[PATCH] D152766: AMDGPU: Use correct lowering for llvm.log2.f32

Matt Arsenault via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Tue Jun 13 11:04:00 PDT 2023


arsenm added inline comments.


================
Comment at: llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp:2463
+  if (VT == MVT::f16) {
+    // Nothing in half is a denormal when promoted to f32.
+    assert(!Subtarget->has16BitInsts());
----------------
rampitec wrote:
> Don't you need to produce v_log_f16 if denorm handling is not needed?
v_log_f16 is supposed to be fully correct. This is the promoted path for SI/CI that don't have it


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D152766/new/

https://reviews.llvm.org/D152766



More information about the llvm-commits mailing list