AMDGPU/GlobalISel: Handle llvm.amdgcn.reloc.constant

Author: arsenm

llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp

@@ -904,6 +904,8 @@ bool AMDGPUInstructionSelector::selectG_INTRINSIC(MachineInstr &I) const {
     return selectIntrinsicIcmp(I);
   case Intrinsic::amdgcn_ballot:
     return selectBallot(I);
+  case Intrinsic::amdgcn_reloc_constant:
+    return selectRelocConstant(I);
   default:
     return selectImpl(I, *CoverageInfo);
   }
@@ -1084,6 +1086,31 @@ bool AMDGPUInstructionSelector::selectBallot(MachineInstr &I) const {
   return true;
 }
 
+bool AMDGPUInstructionSelector::selectRelocConstant(MachineInstr &I) const {
+  Register DstReg = I.getOperand(0).getReg();
+  const RegisterBank *DstBank = RBI.getRegBank(DstReg, *MRI, TRI);
+  const TargetRegisterClass *DstRC =
+    TRI.getRegClassForSizeOnBank(32, *DstBank, *MRI);
+  if (!DstRC || !RBI.constrainGenericRegister(DstReg, *DstRC, *MRI))
+    return false;
+
+  const bool IsVALU = DstBank->getID() == AMDGPU::VGPRRegBankID;
+
+  Module *M = MF->getFunction().getParent();
+  const MDNode *Metadata = I.getOperand(2).getMetadata();
+  auto SymbolName = cast<MDString>(Metadata->getOperand(0))->getString();
+  auto RelocSymbol = cast<GlobalVariable>(
+    M->getOrInsertGlobal(SymbolName, Type::getInt32Ty(M->getContext())));
+
+  MachineBasicBlock *BB = I.getParent();
+  BuildMI(*BB, &I, I.getDebugLoc(),
+          TII.get(IsVALU ? AMDGPU::V_MOV_B32_e32 : AMDGPU::S_MOV_B32), DstReg)
+    .addGlobalAddress(RelocSymbol, 0, SIInstrInfo::MO_ABS32_LO);
+
+  I.eraseFromParent();
+  return true;
+}
+
 bool AMDGPUInstructionSelector::selectEndCfIntrinsic(MachineInstr &MI) const {
   // FIXME: Manually selecting to avoid dealiing with the SReg_1 trick
   // SelectionDAG uses for wave32 vs wave64.

llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.h

@@ -108,6 +108,7 @@ class AMDGPUInstructionSelector final : public InstructionSelector {
   bool selectDivScale(MachineInstr &MI) const;
   bool selectIntrinsicIcmp(MachineInstr &MI) const;
   bool selectBallot(MachineInstr &I) const;
+  bool selectRelocConstant(MachineInstr &I) const;
   bool selectG_INTRINSIC(MachineInstr &I) const;
 
   bool selectEndCfIntrinsic(MachineInstr &MI) const;

llvm/lib/Target/AMDGPU/AMDGPURegisterBankInfo.cpp

@@ -4023,7 +4023,8 @@ AMDGPURegisterBankInfo::getInstrMapping(const MachineInstr &MI) const {
       return getDefaultMappingAllVGPR(MI);
     case Intrinsic::amdgcn_kernarg_segment_ptr:
     case Intrinsic::amdgcn_s_getpc:
-    case Intrinsic::amdgcn_groupstaticsize: {
+    case Intrinsic::amdgcn_groupstaticsize:
+    case Intrinsic::amdgcn_reloc_constant: {
       unsigned Size = MRI.getType(MI.getOperand(0).getReg()).getSizeInBits();
       OpdsMapping[0] = AMDGPU::getValueMapping(AMDGPU::SGPRRegBankID, Size);
       break;

llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-amdgcn.reloc.constant.mir

@@ -0,0 +1,50 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -march=amdgcn -mcpu=tahiti -run-pass=instruction-select -verify-machineinstrs -o - %s | FileCheck -check-prefix=GCN %s
+
+--- |
+
+  define void @reloc_constant_sgpr32() { ret void }
+  define void @reloc_constant_vgpr32() { ret void }
+
+  declare i32 @llvm.amdgcn.reloc.constant(metadata)
+
+  !0 = !{!"arst"}
+
+...
+
+---
+name:            reloc_constant_sgpr32
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+body:             |
+  bb.0:
+
+    ; GCN-LABEL: name: reloc_constant_sgpr32
+    ; GCN: [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 target-flags(amdgpu-abs32-lo) @arst
+    ; GCN: $sgpr0 = COPY [[S_MOV_B32_]]
+    ; GCN: S_ENDPGM 0, implicit $sgpr0
+    %0:sgpr(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.reloc.constant), !0
+    $sgpr0 = COPY %0
+    S_ENDPGM 0, implicit $sgpr0
+
+...
+
+---
+name:            reloc_constant_vgpr32
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+body:             |
+  bb.0:
+
+    ; GCN-LABEL: name: reloc_constant_vgpr32
+    ; GCN: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 target-flags(amdgpu-abs32-lo) @arst, implicit $exec
+    ; GCN: $vgpr0 = COPY [[V_MOV_B32_e32_]]
+    ; GCN: S_ENDPGM 0, implicit $vgpr0
+    %0:vgpr(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.reloc.constant), !0
+    $vgpr0 = COPY %0
+    S_ENDPGM 0, implicit $vgpr0
+
+...
+

llvm/test/CodeGen/AMDGPU/amdgpu-reloc-const.ll

@@ -1,5 +1,9 @@
 ; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
 ; RUN: llc -mtriple=amdgcn--amdpal -mcpu=gfx900 -filetype=obj -o %t.o < %s && llvm-readobj -relocations %t.o | FileCheck --check-prefix=ELF %s
+
+; RUN: llc -global-isel -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
+; RUN: llc -global-isel -mtriple=amdgcn--amdpal -mcpu=gfx900 -filetype=obj -o %t.o < %s && llvm-readobj -relocations %t.o | FileCheck --check-prefix=ELF %s
+
 ; GCN-LABEL: {{^}}ps_main:
 ; GCN: v_mov_b32_{{.*}} v[[relocreg:[0-9]+]], doff_0_0_b@abs32@lo
 ; GCN-NEXT: exp {{.*}} v[[relocreg]], {{.*}}