AMDGPU/GlobalISel: Implement select for 32-bit G_ADD

Reviewers: arsenm

Reviewed By: arsenm

Subscribers: hiraditya, kzhuravl, jvesely, wdng, nhaehnle, yaxunl, rovka, kristof.beyls, dstuttard, tpr, t-tye, Petar.Avramovic, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D58804

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@364797 91177308-0d34-0410-b5e6-96231b3b80d8

Author: tstellar

lib/Target/AMDGPU/AMDGPUGISel.td

@@ -128,6 +128,9 @@ multiclass GISelVop2IntrPat <
 def : GISelSop2Pat <or, S_OR_B32, i32>;
 def : GISelVop2Pat <or, V_OR_B32_e32, i32>;
 
+def : GISelSop2Pat <add, S_ADD_I32, i32>;
+def : GISelVop2Pat <add, V_ADD_I32_e32, i32>;
+
 def : GISelSop2Pat <sra, S_ASHR_I32, i32>;
 let AddedComplexity = 100 in {
 let SubtargetPredicate = isGFX6GFX7 in {

lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp

@@ -1056,10 +1056,12 @@ bool AMDGPUInstructionSelector::select(MachineInstr &I,
   }
 
   switch (I.getOpcode()) {
+  case TargetOpcode::G_ADD:
+    if (selectG_ADD(I))
+      return true;
+    LLVM_FALLTHROUGH;
   default:
     return selectImpl(I, CoverageInfo);
-  case TargetOpcode::G_ADD:
-    return selectG_ADD(I);
   case TargetOpcode::G_INTTOPTR:
   case TargetOpcode::G_BITCAST:
     return selectCOPY(I);

test/CodeGen/AMDGPU/GlobalISel/inst-select-add.mir

@@ -0,0 +1,43 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -march=amdgcn -run-pass=instruction-select -verify-machineinstrs -global-isel %s -o - | FileCheck %s -check-prefixes=GCN
+
+---
+name:            add_i32
+legalized:       true
+regBankSelected: true
+
+body: |
+  bb.0:
+    liveins: $sgpr0, $sgpr1, $vgpr0, $vgpr3_vgpr4
+    ; GCN-LABEL: name: add_i32
+    ; GCN: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr0
+    ; GCN: [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr1
+    ; GCN: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+    ; GCN: [[COPY3:%[0-9]+]]:vreg_64 = COPY $vgpr3_vgpr4
+    ; GCN: [[S_ADD_I32_:%[0-9]+]]:sreg_32_xm0 = S_ADD_I32 [[COPY]], [[COPY1]], implicit-def $scc
+    ; GCN: [[V_ADD_I32_e32_:%[0-9]+]]:vgpr_32 = V_ADD_I32_e32 [[S_ADD_I32_]], [[COPY2]], implicit-def $vcc, implicit $exec
+    ; GCN: [[V_ADD_I32_e32_1:%[0-9]+]]:vgpr_32 = V_ADD_I32_e32 [[S_ADD_I32_]], [[V_ADD_I32_e32_]], implicit-def $vcc, implicit $exec
+    ; GCN: [[V_ADD_I32_e32_2:%[0-9]+]]:vgpr_32 = V_ADD_I32_e32 [[V_ADD_I32_e32_1]], [[COPY2]], implicit-def $vcc, implicit $exec
+    ; GCN: FLAT_STORE_DWORD [[COPY3]], [[V_ADD_I32_e32_2]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr
+    %0:sgpr(s32) = COPY $sgpr0
+    %1:sgpr(s32) = COPY $sgpr1
+    %2:vgpr(s32) = COPY $vgpr0
+    %3:vgpr(p1) = COPY $vgpr3_vgpr4
+    %4:sgpr(s32) = G_CONSTANT i32 1
+    %5:sgpr(s32) = G_CONSTANT i32 4096
+
+    ; add ss
+    %6:sgpr(s32) = G_ADD %0, %1
+
+    ; add vs
+    %7:vgpr(s32) = G_ADD %2, %6
+
+    ; add sv
+    %8:vgpr(s32) = G_ADD %6, %7
+
+    ; add vv
+    %9:vgpr(s32) = G_ADD %8, %2
+
+    G_STORE %9, %3 :: (store 4, addrspace 1)
+
+...