Move to tryConstantFoldOp

Change-Id: I9b14559b4b5dc9c4bb383ebd517edcdc094a2e6c

Author: jrbyrnes

llvm/lib/Target/AMDGPU/SIFoldOperands.cpp

@@ -1684,6 +1684,14 @@ bool SIFoldOperandsImpl::tryConstantFoldOp(MachineInstr *MI) const {
     }
   }
 
+  if (Opc == AMDGPU::V_ADD_U32_e64 || Opc == AMDGPU::V_ADD_U32_e32) {
+    if (Src1Val == 0) {
+      // y = add x, 0 -> y = copy x
+      MI->removeOperand(Src1Idx);
+      mutateCopyOp(*MI, TII->get(AMDGPU::COPY));
+    }
+  }
+
   return false;
 }
 

llvm/test/CodeGen/AMDGPU/groupstaticsize-zero.ll

@@ -0,0 +1,20 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx950 < %s | FileCheck -check-prefixes=GCN %s
+
+@global_smem = external addrspace(3) global [0 x i8]
+
+define amdgpu_kernel void @addzero() {
+; GCN-LABEL: addzero:
+; GCN:       ; %bb.0: ; %.lr.ph
+; GCN-NEXT:    v_mov_b32_e32 v2, 0
+; GCN-NEXT:    v_and_b32_e32 v0, 1, v0
+; GCN-NEXT:    v_mov_b32_e32 v3, v2
+; GCN-NEXT:    ds_write_b64 v0, v[2:3]
+; GCN-NEXT:    s_endpgm
+.lr.ph:
+  %0 = tail call i32 @llvm.amdgcn.workitem.id.x()
+  %1 = and i32 %0, 1
+  %2 = getelementptr i8, ptr addrspace(3) @global_smem, i32 %1
+  store <4 x bfloat> zeroinitializer, ptr addrspace(3) %2, align 8
+  ret void
+}