AMDGPU/GlobalISel: Mark 32-bit G_ADD as legal

Reviewers: arsenm

Reviewed By: arsenm

Subscribers: kzhuravl, wdng, nhaehnle, yaxunl, rovka, kristof.beyls, igorb, dstuttard, tpr, llvm-commits, t-tye

Differential Revision: https://reviews.llvm.org/D33992

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@305232 91177308-0d34-0410-b5e6-96231b3b80d8

lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp

@@ -34,6 +34,8 @@ AMDGPULegalizerInfo::AMDGPULegalizerInfo() {
   const LLT P1 = LLT::pointer(1, 64);
   const LLT P2 = LLT::pointer(2, 64);
 
+  setAction({G_ADD, S32}, Legal);
+
   // FIXME: i1 operands to intrinsics should always be legal, but other i1
   // values may not be legal.  We need to figure out how to distinguish
   // between these two scenarios.

test/CodeGen/AMDGPU/GlobalISel/legalize-add.mir

@@ -0,0 +1,22 @@
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -O0 -run-pass=legalizer -global-isel %s -o - | FileCheck %s
+
+--- |
+  define void @test_add() { ret void }
+...
+
+---
+name:            test_add
+registers:
+  - { id: 0, class: _ }
+  - { id: 1, class: _ }
+  - { id: 2, class: _ }
+body: |
+  bb.0:
+    liveins: %vgpr0, %vgpr1
+    ; CHECK-LABEL: name: test_add
+    ; CHECK: %2(s32) = G_ADD %0, %1
+
+    %0(s32) = COPY %vgpr0
+    %1(s32) = COPY %vgpr1
+    %2(s32) = G_ADD %0, %1
+...