[PATCH] D33212: AMDGPU/GlobalISel: Mark 32-bit float constants as legal

Matt Arsenault via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Tue May 16 06:53:39 PDT 2017


arsenm added inline comments.


================
Comment at: lib/Target/AMDGPU/SIInstrInfo.cpp:2118-2120
+
+  if (SIInstrInfo::isGenericOpcode(MI.getOpcode()))
+    return true;
----------------
Why is this necessary? It looks like a separate change


================
Comment at: test/CodeGen/AMDGPU/GlobalISel/legalize-constant.mir:34-36
+    ; CHECK: %0(s32) = G_FCONSTANT  float 1.000000e+00
+
+    %0(s32) = G_FCONSTANT float 1.0
----------------
It might be worthwhile to test a non-inline immediate value too


https://reviews.llvm.org/D33212





More information about the llvm-commits mailing list