GlobalISel: Implement narrowScalar for vector extract/insert indexes

[aemerson]

LGTM.

[arsenm]

r366113

lib/CodeGen/GlobalISel/LegalizerHelper.cpp

for (unsigned i = 0; i < NumParts; ++i) {
         MIB.addUse(SrcRegs[j / 2][i]).add(MI.getOperand(j + 1));
     }
     MIRBuilder.setInsertPt(MBB, --MBB.getFirstNonPHI());
     MIRBuilder.buildMerge(MI.getOperand(0).getReg(), DstRegs);
     Observer.changedInstr(MI);
     MI.eraseFromParent();
     return Legalized;
   }
+  case TargetOpcode::G_EXTRACT_VECTOR_ELT:
+  case TargetOpcode::G_INSERT_VECTOR_ELT: {
+    if (TypeIdx != 2)
+      return UnableToLegalize;
+
+    int OpIdx = MI.getOpcode() == TargetOpcode::G_EXTRACT_VECTOR_ELT ? 2 : 3;
+    Observer.changingInstr(MI);
+    narrowScalarSrc(MI, NarrowTy, OpIdx);
+    Observer.changedInstr(MI);
+    return Legalized;
+  }
   }
 }
 
 void LegalizerHelper::widenScalarSrc(MachineInstr &MI, LLT WideTy,
                                      unsigned OpIdx, unsigned ExtOpcode) {
   MachineOperand &MO = MI.getOperand(OpIdx);
   auto ExtB = MIRBuilder.buildInstr(ExtOpcode, {WideTy}, {MO.getReg()});
   MO.setReg(ExtB->getOperand(0).getReg());

test/CodeGen/AMDGPU/GlobalISel/legalize-extract-vector-elt.mir

bb.0:
     ; CHECK: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr8
     ; CHECK: [[EVEC:%[0-9]+]]:_(s128) = G_EXTRACT_VECTOR_ELT [[COPY]](<2 x s128>), [[COPY1]](s32)
     ; CHECK: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[EVEC]](s128)
     %0:_(<2 x s128>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7
     %1:_(s32) = COPY $vgpr8
     %2:_(s128) = G_EXTRACT_VECTOR_ELT %0, %1
     $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %2
 ...
+
+---
+name: extract_vector_elt_v2i32_varidx_i64
+
+body: |
+  bb.0:
+    liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+
+    ; CHECK-LABEL: name: extract_vector_elt_v2i32_varidx_i64
+    ; CHECK: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr0_vgpr1
+    ; CHECK: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
+    ; CHECK: [[TRUNC:%[0-9]+]]:_(s32) = G_TRUNC [[COPY1]](s64)
+    ; CHECK: [[EVEC:%[0-9]+]]:_(s32) = G_EXTRACT_VECTOR_ELT [[COPY]](<2 x s32>), [[TRUNC]](s32)
+    ; CHECK: $vgpr0 = COPY [[EVEC]](s32)
+    %0:_(<2 x s32>) = COPY $vgpr0_vgpr1
+    %1:_(s64) = COPY $vgpr2_vgpr3
+    %2:_(s32) = G_EXTRACT_VECTOR_ELT %0, %1
+    $vgpr0 = COPY %2
+...
+

test/CodeGen/AMDGPU/GlobalISel/legalize-insert-vector-elt.mir

 # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
 # RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -run-pass=legalizer %s -o - | FileCheck %s
 
 ---
-name: insert_vector_elt_0_v2i32
+name: insert_vector_elt_0_v2s32
 
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2
-    ; CHECK-LABEL: name: insert_vector_elt_0_v2i32
+    ; CHECK-LABEL: name: insert_vector_elt_0_v2s32
     ; CHECK: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr0_vgpr1
     ; CHECK: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
     ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
     ; CHECK: [[IVEC:%[0-9]+]]:_(<2 x s32>) = G_INSERT_VECTOR_ELT [[COPY]], [[COPY1]](s32), [[C]](s32)
     ; CHECK: $vgpr0_vgpr1 = COPY [[IVEC]](<2 x s32>)
     %0:_(<2 x s32>) = COPY $vgpr0_vgpr1
     %1:_(s32) = COPY $vgpr2
     %2:_(s32) = G_CONSTANT i32 0
     %3:_(<2 x s32>) = G_INSERT_VECTOR_ELT %0, %1, %2
     $vgpr0_vgpr1 = COPY %3
 ...
+
+---
+name: insert_vector_elt_v2s32_varidx_i64
+
+body: |
+  bb.0:
+    liveins: $vgpr0_vgpr1, $vgpr2, $vgpr3_vgpr4
+
+    ; CHECK-LABEL: name: insert_vector_elt_v2s32_varidx_i64
+    ; CHECK: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr0_vgpr1
+    ; CHECK: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+    ; CHECK: [[COPY2:%[0-9]+]]:_(s64) = COPY $vgpr3_vgpr4
+    ; CHECK: [[TRUNC:%[0-9]+]]:_(s32) = G_TRUNC [[COPY2]](s64)
+    ; CHECK: [[IVEC:%[0-9]+]]:_(<2 x s32>) = G_INSERT_VECTOR_ELT [[COPY]], [[COPY1]](s32), [[TRUNC]](s32)
+    ; CHECK: $vgpr0_vgpr1 = COPY [[IVEC]](<2 x s32>)
+    %0:_(<2 x s32>) = COPY $vgpr0_vgpr1
+    %1:_(s32) = COPY $vgpr2
+    %2:_(s64) = COPY $vgpr3_vgpr4
+    %3:_(<2 x s32>) = G_INSERT_VECTOR_ELT %0, %1, %2
+    $vgpr0_vgpr1 = COPY %3
+...
+
+---
+name: insert_vector_elt_v16s32_varidx_i64
+
+body: |
+  bb.0:
+    liveins: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15, $vgpr16, $vgpr17_vgpr18
+
+    ; CHECK-LABEL: name: insert_vector_elt_v16s32_varidx_i64
+    ; CHECK: [[COPY:%[0-9]+]]:_(<16 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15
+    ; CHECK: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr16
+    ; CHECK: [[COPY2:%[0-9]+]]:_(s64) = COPY $vgpr17_vgpr18
+    ; CHECK: [[TRUNC:%[0-9]+]]:_(s32) = G_TRUNC [[COPY2]](s64)
+    ; CHECK: [[IVEC:%[0-9]+]]:_(<16 x s32>) = G_INSERT_VECTOR_ELT [[COPY]], [[COPY1]](s32), [[TRUNC]](s32)
+    ; CHECK: S_ENDPGM 0, implicit [[IVEC]](<16 x s32>)
+    %0:_(<16 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15
+    %1:_(s32) = COPY $vgpr16
+    %2:_(s64) = COPY $vgpr17_vgpr18
+    %3:_(<16 x s32>) = G_INSERT_VECTOR_ELT %0, %1, %2
+    S_ENDPGM 0, implicit %3
+...