[MIPS GlobalISel] Select all instructions required to load and store i32 on stack

[Petar.Avramovic]

Summary

Add code for selection of G_LOAD, G_STORE, G_GEP, G_FRAMEINDEX and G_CONSTANT.
Support loads and stores of i32 values.

[Petar.Avramovic]

Does anybody have any comment?

[atanasyan]

lib/Target/Mips/MipsInstructionSelector.cpp

135 ↗

(On Diff #154172)

Redundant empty line.

143 ↗

(On Diff #154172)

Ditto

152 ↗

(On Diff #154172)

I suggest to remove the selectLoadStoreOpcode function and inline the code as follow. I think it explicitly show when we do nothing and when need to call BuildMM.

if (DestRegBank != Mips::GPRBRegBankID || OpSize != 32)
  return false;

const unsigned NewOpc =
    I.getOpcode() == G_STORE ? Mips::SW : Mips::LW;

162 ↗

(On Diff #154172)

Redundant empty line.

[Petar.Avramovic]

Addressed review comments

[atanasyan]

LGTM

llvm/trunk/lib/Target/Mips/MipsInstructionSelector.cpp

if (I.isCopy())
       return selectCopy(I, TII, MRI, TRI, RBI);
 
     return true;
   }
 
   if (selectImpl(I, CoverageInfo)) {
     return true;
   }
-  // We didn't select anything.
+
+  MachineInstr *MI = nullptr;
+  using namespace TargetOpcode;
+
+  switch (I.getOpcode()) {
+  case G_GEP: {
+    MI = BuildMI(MBB, I, I.getDebugLoc(), TII.get(Mips::ADDu))
+             .add(I.getOperand(0))
+             .add(I.getOperand(1))
+             .add(I.getOperand(2));
+    break;
+  }
+  case G_FRAME_INDEX: {
+    MI = BuildMI(MBB, I, I.getDebugLoc(), TII.get(Mips::ADDiu))
+             .add(I.getOperand(0))
+             .add(I.getOperand(1))
+             .addImm(0);
+    break;
+  }
+  case G_STORE:
+  case G_LOAD: {
+    const unsigned DestReg = I.getOperand(0).getReg();
+    const unsigned DestRegBank = RBI.getRegBank(DestReg, MRI, TRI)->getID();
+    const unsigned OpSize = MRI.getType(DestReg).getSizeInBits();
+
+    if (DestRegBank != Mips::GPRBRegBankID || OpSize != 32)
+      return false;
+
+    const unsigned NewOpc = I.getOpcode() == G_STORE ? Mips::SW : Mips::LW;
+
+    MI = BuildMI(MBB, I, I.getDebugLoc(), TII.get(NewOpc))
+             .add(I.getOperand(0))
+             .add(I.getOperand(1))
+             .addImm(0)
+             .addMemOperand(*I.memoperands_begin());
+    break;
+  }
+  case G_CONSTANT: {
+    int Imm = I.getOperand(1).getCImm()->getValue().getLimitedValue();
+    unsigned LUiReg = MRI.createVirtualRegister(&Mips::GPR32RegClass);
+    MachineInstr *LUi, *ORi;
+
+    LUi = BuildMI(MBB, I, I.getDebugLoc(), TII.get(Mips::LUi))
+              .addDef(LUiReg)
+              .addImm(Imm >> 16);
+
+    ORi = BuildMI(MBB, I, I.getDebugLoc(), TII.get(Mips::ORi))
+              .addDef(I.getOperand(0).getReg())
+              .addUse(LUiReg)
+              .addImm(Imm & 0xFFFF);
+
+    if (!constrainSelectedInstRegOperands(*LUi, TII, TRI, RBI))
+      return false;
+    if (!constrainSelectedInstRegOperands(*ORi, TII, TRI, RBI))
       return false;
+
+    I.eraseFromParent();
+    return true;
+  }
+
+  default:
+    return false;
+  }
+
+  I.eraseFromParent();
+  return constrainSelectedInstRegOperands(*MI, TII, TRI, RBI);
 }
 
 namespace llvm {
 InstructionSelector *createMipsInstructionSelector(const MipsTargetMachine &TM,
                                                    MipsSubtarget &Subtarget,
                                                    MipsRegisterBankInfo &RBI) {
   return new MipsInstructionSelector(TM, Subtarget, RBI);
 }
 } // end namespace llvm

llvm/trunk/lib/Target/Mips/MipsLegalizerInfo.cpp

 #include "MipsTargetMachine.h"
 
 using namespace llvm;
 
 MipsLegalizerInfo::MipsLegalizerInfo(const MipsSubtarget &ST) {
   using namespace TargetOpcode;
 
   const LLT s32 = LLT::scalar(32);
+  const LLT p0 = LLT::pointer(0, 32);
 
   getActionDefinitionsBuilder(G_ADD).legalFor({s32});
 
+  getActionDefinitionsBuilder({G_LOAD, G_STORE})
+      .legalFor({{s32, p0}});
+
+  getActionDefinitionsBuilder(G_CONSTANT)
+      .legalFor({s32});
+
+  getActionDefinitionsBuilder(G_GEP)
+      .legalFor({{p0, s32}});
+
+  getActionDefinitionsBuilder(G_FRAME_INDEX)
+      .legalFor({p0});
+
   computeTables();
   verify(*ST.getInstrInfo());
 }

llvm/trunk/lib/Target/Mips/MipsRegisterBankInfo.cpp

 const RegisterBank &MipsRegisterBankInfo::getRegBankFromRegClass(
     const TargetRegisterClass &RC) const {
   using namespace Mips;
 
   switch (RC.getID()) {
   case Mips::GPR32RegClassID:
   case Mips::CPU16Regs_and_GPRMM16ZeroRegClassID:
   case Mips::GPRMM16MoveP_and_CPU16Regs_and_GPRMM16ZeroRegClassID:
+  case Mips::SP32RegClassID:
     return getRegBank(Mips::GPRBRegBankID);
   default:
     llvm_unreachable("Register class not supported");
   }
 }
 
 const RegisterBankInfo::InstructionMapping &
 MipsRegisterBankInfo::getInstrMapping(const MachineInstr &MI) const {
 
   unsigned Opc = MI.getOpcode();
 
   const RegisterBankInfo::InstructionMapping &Mapping = getInstrMappingImpl(MI);
   if (Mapping.isValid())
     return Mapping;
 
   using namespace TargetOpcode;
 
   unsigned NumOperands = MI.getNumOperands();
   const ValueMapping *OperandsMapping = &Mips::ValueMappings[Mips::GPRIdx];
 
   switch (Opc) {
   case G_ADD:
+  case G_LOAD:
+  case G_STORE:
+  case G_GEP:
     OperandsMapping = &Mips::ValueMappings[Mips::GPRIdx];
     break;
+  case G_CONSTANT:
+  case G_FRAME_INDEX:
+    OperandsMapping =
+        getOperandsMapping({&Mips::ValueMappings[Mips::GPRIdx], nullptr});
+    break;
   default:
     return getInvalidInstructionMapping();
   }
 
   return getInstructionMapping(DefaultMappingID, /*Cost=*/1, OperandsMapping,
                                NumOperands);
 }

llvm/trunk/test/CodeGen/Mips/GlobalISel/instruction-select/stack_args.mir

+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -O0 -mtriple=mipsel-linux-gnu -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck %s -check-prefixes=MIPS32
+--- |
+
+  declare i32 @f(i32, i32, i32, i32, i32)
+  define void @g(i32, i32, i32, i32, i32) {entry: ret void}
+
+...
+---
+name:            g
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+fixedStack:
+  - { id: 0, offset: 16, size: 4, alignment: 8, stack-id: 0, isImmutable: true }
+body:             |
+  bb.1.entry:
+    liveins: $a0, $a1, $a2, $a3
+
+    ; MIPS32-LABEL: name: g
+    ; MIPS32: liveins: $a0, $a1, $a2, $a3
+    ; MIPS32: [[COPY:%[0-9]+]]:gpr32 = COPY $a0
+    ; MIPS32: [[COPY1:%[0-9]+]]:gpr32 = COPY $a1
+    ; MIPS32: [[COPY2:%[0-9]+]]:gpr32 = COPY $a2
+    ; MIPS32: [[COPY3:%[0-9]+]]:gpr32 = COPY $a3
+    ; MIPS32: [[ADDiu:%[0-9]+]]:gpr32 = ADDiu %fixed-stack.0, 0
+    ; MIPS32: [[LW:%[0-9]+]]:gpr32 = LW [[ADDiu]], 0 :: (load 4 from %fixed-stack.0, align 0)
+    ; MIPS32: ADJCALLSTACKDOWN 24, 0, implicit-def $sp, implicit $sp
+    ; MIPS32: $a0 = COPY [[COPY]]
+    ; MIPS32: $a1 = COPY [[COPY1]]
+    ; MIPS32: $a2 = COPY [[COPY2]]
+    ; MIPS32: $a3 = COPY [[COPY3]]
+    ; MIPS32: [[COPY4:%[0-9]+]]:gpr32 = COPY $sp
+    ; MIPS32: [[LUi:%[0-9]+]]:gpr32 = LUi 0
+    ; MIPS32: [[ORi:%[0-9]+]]:gpr32 = ORi [[LUi]], 16
+    ; MIPS32: [[ADDu:%[0-9]+]]:gpr32 = ADDu [[COPY4]], [[ORi]]
+    ; MIPS32: SW [[LW]], [[ADDu]], 0 :: (store 4 into stack + 16, align 0)
+    ; MIPS32: JAL @f, csr_o32, implicit-def $ra, implicit-def $sp, implicit $a0, implicit $a1, implicit $a2, implicit $a3, implicit-def $v0
+    ; MIPS32: [[COPY5:%[0-9]+]]:gpr32 = COPY $v0
+    ; MIPS32: ADJCALLSTACKUP 24, 0, implicit-def $sp, implicit $sp
+    ; MIPS32: $v0 = COPY [[COPY5]]
+    ; MIPS32: RetRA implicit $v0
+    %0:gprb(s32) = COPY $a0
+    %1:gprb(s32) = COPY $a1
+    %2:gprb(s32) = COPY $a2
+    %3:gprb(s32) = COPY $a3
+    %5:gprb(p0) = G_FRAME_INDEX %fixed-stack.0
+    %4:gprb(s32) = G_LOAD %5(p0) :: (load 4 from %fixed-stack.0, align 0)
+    ADJCALLSTACKDOWN 24, 0, implicit-def $sp, implicit $sp
+    $a0 = COPY %0(s32)
+    $a1 = COPY %1(s32)
+    $a2 = COPY %2(s32)
+    $a3 = COPY %3(s32)
+    %7:gprb(p0) = COPY $sp
+    %8:gprb(s32) = G_CONSTANT i32 16
+    %9:gprb(p0) = G_GEP %7, %8(s32)
+    G_STORE %4(s32), %9(p0) :: (store 4 into stack + 16, align 0)
+    JAL @f, csr_o32, implicit-def $ra, implicit-def $sp, implicit $a0, implicit $a1, implicit $a2, implicit $a3, implicit-def $v0
+    %6:gprb(s32) = COPY $v0
+    ADJCALLSTACKUP 24, 0, implicit-def $sp, implicit $sp
+    $v0 = COPY %6(s32)
+    RetRA implicit $v0
+
+...

llvm/trunk/test/CodeGen/Mips/GlobalISel/legalizer/stack_args.mir

+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -O0 -mtriple=mipsel-linux-gnu -run-pass=legalizer -verify-machineinstrs %s -o - | FileCheck %s -check-prefixes=MIPS32
+--- |
+
+  declare i32 @f(i32, i32, i32, i32, i32)
+  define void @g(i32, i32, i32, i32, i32) {entry: ret void}
+
+...
+---
+name:            g
+alignment:       2
+tracksRegLiveness: true
+fixedStack:
+  - { id: 0, offset: 16, size: 4, alignment: 8, stack-id: 0, isImmutable: true }
+body:             |
+  bb.1.entry:
+    liveins: $a0, $a1, $a2, $a3
+
+    ; MIPS32-LABEL: name: g
+    ; MIPS32: liveins: $a0, $a1, $a2, $a3
+    ; MIPS32: [[COPY:%[0-9]+]]:_(s32) = COPY $a0
+    ; MIPS32: [[COPY1:%[0-9]+]]:_(s32) = COPY $a1
+    ; MIPS32: [[COPY2:%[0-9]+]]:_(s32) = COPY $a2
+    ; MIPS32: [[COPY3:%[0-9]+]]:_(s32) = COPY $a3
+    ; MIPS32: [[FRAME_INDEX:%[0-9]+]]:_(p0) = G_FRAME_INDEX %fixed-stack.0
+    ; MIPS32: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[FRAME_INDEX]](p0) :: (load 4 from %fixed-stack.0, align 0)
+    ; MIPS32: ADJCALLSTACKDOWN 24, 0, implicit-def $sp, implicit $sp
+    ; MIPS32: $a0 = COPY [[COPY]](s32)
+    ; MIPS32: $a1 = COPY [[COPY1]](s32)
+    ; MIPS32: $a2 = COPY [[COPY2]](s32)
+    ; MIPS32: $a3 = COPY [[COPY3]](s32)
+    ; MIPS32: [[COPY4:%[0-9]+]]:_(p0) = COPY $sp
+    ; MIPS32: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; MIPS32: [[GEP:%[0-9]+]]:_(p0) = G_GEP [[COPY4]], [[C]](s32)
+    ; MIPS32: G_STORE [[LOAD]](s32), [[GEP]](p0) :: (store 4 into stack + 16, align 0)
+    ; MIPS32: JAL @f, csr_o32, implicit-def $ra, implicit-def $sp, implicit $a0, implicit $a1, implicit $a2, implicit $a3, implicit-def $v0
+    ; MIPS32: [[COPY5:%[0-9]+]]:_(s32) = COPY $v0
+    ; MIPS32: ADJCALLSTACKUP 24, 0, implicit-def $sp, implicit $sp
+    ; MIPS32: $v0 = COPY [[COPY5]](s32)
+    ; MIPS32: RetRA implicit $v0
+    %0:_(s32) = COPY $a0
+    %1:_(s32) = COPY $a1
+    %2:_(s32) = COPY $a2
+    %3:_(s32) = COPY $a3
+    %5:_(p0) = G_FRAME_INDEX %fixed-stack.0
+    %4:_(s32) = G_LOAD %5(p0) :: (load 4 from %fixed-stack.0, align 0)
+    ADJCALLSTACKDOWN 24, 0, implicit-def $sp, implicit $sp
+    $a0 = COPY %0(s32)
+    $a1 = COPY %1(s32)
+    $a2 = COPY %2(s32)
+    $a3 = COPY %3(s32)
+    %7:_(p0) = COPY $sp
+    %8:_(s32) = G_CONSTANT i32 16
+    %9:_(p0) = G_GEP %7, %8(s32)
+    G_STORE %4(s32), %9(p0) :: (store 4 into stack + 16, align 0)
+    JAL @f, csr_o32, implicit-def $ra, implicit-def $sp, implicit $a0, implicit $a1, implicit $a2, implicit $a3, implicit-def $v0
+    %6:_(s32) = COPY $v0
+    ADJCALLSTACKUP 24, 0, implicit-def $sp, implicit $sp
+    $v0 = COPY %6(s32)
+    RetRA implicit $v0
+
+...

llvm/trunk/test/CodeGen/Mips/GlobalISel/llvm-ir/stack_args.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc  -O0 -mtriple=mipsel-linux-gnu -global-isel  -verify-machineinstrs %s -o -| FileCheck %s -check-prefixes=MIPS32
+
+declare i32 @f(i32, i32, i32, i32, i32)
+
+define  i32 @g(i32  %x1, i32 %x2, i32 %x3, i32 %x4, i32 %x5){
+; MIPS32-LABEL: g:
+; MIPS32:       # %bb.0: # %entry
+; MIPS32-NEXT:    addiu $sp, $sp, -32
+; MIPS32-NEXT:    .cfi_def_cfa_offset 32
+; MIPS32-NEXT:    sw $ra, 28($sp) # 4-byte Folded Spill
+; MIPS32-NEXT:    .cfi_offset 31, -4
+; MIPS32-NEXT:    addiu $1, $sp, 48
+; MIPS32-NEXT:    lw $1, 0($1)
+; MIPS32-NEXT:    move $2, $sp
+; MIPS32-NEXT:    lui $3, 0
+; MIPS32-NEXT:    ori $3, $3, 16
+; MIPS32-NEXT:    addu $2, $2, $3
+; MIPS32-NEXT:    sw $1, 0($2)
+; MIPS32-NEXT:    jal f
+; MIPS32-NEXT:    nop
+; MIPS32-NEXT:    lw $ra, 28($sp) # 4-byte Folded Reload
+; MIPS32-NEXT:    addiu $sp, $sp, 32
+; MIPS32-NEXT:    jr $ra
+; MIPS32-NEXT:    nop
+entry:
+  %call = call i32 @f(i32 %x1, i32 %x2, i32 %x3, i32 %x4, i32 %x5)
+  ret i32 %call
+}

llvm/trunk/test/CodeGen/Mips/GlobalISel/regbankselect/stack_args.mir

+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -O0 -mtriple=mipsel-linux-gnu -run-pass=regbankselect -verify-machineinstrs %s -o - | FileCheck %s -check-prefixes=MIPS32
+--- |
+
+  declare i32 @f(i32, i32, i32, i32, i32)
+  define void @g(i32, i32, i32, i32, i32) {entry: ret void}
+
+...
+---
+name:            g
+alignment:       2
+legalized:       true
+tracksRegLiveness: true
+fixedStack:
+  - { id: 0, offset: 16, size: 4, alignment: 8, stack-id: 0, isImmutable: true }
+body:             |
+ bb.1.entry:
+    liveins: $a0, $a1, $a2, $a3
+
+    ; MIPS32-LABEL: name: g
+    ; MIPS32: liveins: $a0, $a1, $a2, $a3
+    ; MIPS32: [[COPY:%[0-9]+]]:gprb(s32) = COPY $a0
+    ; MIPS32: [[COPY1:%[0-9]+]]:gprb(s32) = COPY $a1
+    ; MIPS32: [[COPY2:%[0-9]+]]:gprb(s32) = COPY $a2
+    ; MIPS32: [[COPY3:%[0-9]+]]:gprb(s32) = COPY $a3
+    ; MIPS32: [[FRAME_INDEX:%[0-9]+]]:gprb(p0) = G_FRAME_INDEX %fixed-stack.0
+    ; MIPS32: [[LOAD:%[0-9]+]]:gprb(s32) = G_LOAD [[FRAME_INDEX]](p0) :: (load 4 from %fixed-stack.0, align 0)
+    ; MIPS32: ADJCALLSTACKDOWN 24, 0, implicit-def $sp, implicit $sp
+    ; MIPS32: $a0 = COPY [[COPY]](s32)
+    ; MIPS32: $a1 = COPY [[COPY1]](s32)
+    ; MIPS32: $a2 = COPY [[COPY2]](s32)
+    ; MIPS32: $a3 = COPY [[COPY3]](s32)
+    ; MIPS32: [[COPY4:%[0-9]+]]:gprb(p0) = COPY $sp
+    ; MIPS32: [[C:%[0-9]+]]:gprb(s32) = G_CONSTANT i32 16
+    ; MIPS32: [[GEP:%[0-9]+]]:gprb(p0) = G_GEP [[COPY4]], [[C]](s32)
+    ; MIPS32: G_STORE [[LOAD]](s32), [[GEP]](p0) :: (store 4 into stack + 16, align 0)
+    ; MIPS32: JAL @f, csr_o32, implicit-def $ra, implicit-def $sp, implicit $a0, implicit $a1, implicit $a2, implicit $a3, implicit-def $v0
+    ; MIPS32: [[COPY5:%[0-9]+]]:gprb(s32) = COPY $v0
+    ; MIPS32: ADJCALLSTACKUP 24, 0, implicit-def $sp, implicit $sp
+    ; MIPS32: $v0 = COPY [[COPY5]](s32)
+    ; MIPS32: RetRA implicit $v0
+    %0:_(s32) = COPY $a0
+    %1:_(s32) = COPY $a1
+    %2:_(s32) = COPY $a2
+    %3:_(s32) = COPY $a3
+    %5:_(p0) = G_FRAME_INDEX %fixed-stack.0
+    %4:_(s32) = G_LOAD %5(p0) :: (load 4 from %fixed-stack.0, align 0)
+    ADJCALLSTACKDOWN 24, 0, implicit-def $sp, implicit $sp
+    $a0 = COPY %0(s32)
+    $a1 = COPY %1(s32)
+    $a2 = COPY %2(s32)
+    $a3 = COPY %3(s32)
+    %7:_(p0) = COPY $sp
+    %8:_(s32) = G_CONSTANT i32 16
+    %9:_(p0) = G_GEP %7, %8(s32)
+    G_STORE %4(s32), %9(p0) :: (store 4 into stack + 16, align 0)
+    JAL @f, csr_o32, implicit-def $ra, implicit-def $sp, implicit $a0, implicit $a1, implicit $a2, implicit $a3, implicit-def $v0
+    %6:_(s32) = COPY $v0
+    ADJCALLSTACKUP 24, 0, implicit-def $sp, implicit $sp
+    $v0 = COPY %6(s32)
+    RetRA implicit $v0
+
+...