Skip to content

[AMDGPU] Extending wave reduction intrinsics for i64 types - 3 #151310

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 2 commits into
base: users/easyonaadit/amdgpu/wave-reduce-intrinsics-arithmetic
Choose a base branch
from
Open

[AMDGPU] Extending wave reduction intrinsics for i64 types - 3 #151310

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
6409f93
[AMDGPU] Extending wave reduction intrinsics for `i64` types - 3
easyonaadit Jul 19, 2025
7301cfb
Removing Redundant Instructions
easyonaadit Jul 31, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
65 changes: 59 additions & 6 deletions llvm/lib/Target/AMDGPU/SIISelLowering.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5111,9 +5111,12 @@ static uint32_t getIdentityValueForWaveReduction(unsigned Opc) {
case AMDGPU::S_SUB_I32:
case AMDGPU::S_SUB_U64_PSEUDO:
case AMDGPU::S_OR_B32:
case AMDGPU::S_OR_B64:
case AMDGPU::S_XOR_B32:
case AMDGPU::S_XOR_B64:
return std::numeric_limits<uint32_t>::min();
case AMDGPU::S_AND_B32:
case AMDGPU::S_AND_B64:
return std::numeric_limits<uint32_t>::max();
default:
llvm_unreachable("Unexpected opcode in getIdentityValueForWaveReduction");
Expand Down Expand Up @@ -5146,14 +5149,17 @@ static MachineBasicBlock *lowerWaveReduce(MachineInstr &MI,
case AMDGPU::S_MAX_I32:
case AMDGPU::V_CMP_GT_I64_e64: /*max*/
case AMDGPU::S_AND_B32:
case AMDGPU::S_OR_B32: {
case AMDGPU::S_AND_B64:
case AMDGPU::S_OR_B32:
case AMDGPU::S_OR_B64: {
// Idempotent operations.
unsigned movOpc = is32BitOpc ? AMDGPU::S_MOV_B32 : AMDGPU::S_MOV_B64;
BuildMI(BB, MI, DL, TII->get(movOpc), DstReg).addReg(SrcReg);
RetBB = &BB;
break;
}
case AMDGPU::S_XOR_B32:
case AMDGPU::S_XOR_B64:
case AMDGPU::S_ADD_I32:
case AMDGPU::S_ADD_U64_PSEUDO:
case AMDGPU::S_SUB_I32:
Expand All @@ -5177,7 +5183,8 @@ static MachineBasicBlock *lowerWaveReduce(MachineInstr &MI,
.addReg(ExecMask);

switch (Opc) {
case AMDGPU::S_XOR_B32: {
case AMDGPU::S_XOR_B32:
case AMDGPU::S_XOR_B64: {
// Performing an XOR operation on a uniform value
// depends on the parity of the number of active lanes.
// For even parity, the result will be 0, for odd
Expand All @@ -5189,10 +5196,41 @@ static MachineBasicBlock *lowerWaveReduce(MachineInstr &MI,
.addReg(NewAccumulator->getOperand(0).getReg())
.addImm(1)
.setOperandDead(3); // Dead scc
BuildMI(BB, MI, DL, TII->get(AMDGPU::S_MUL_I32), DstReg)
.addReg(SrcReg)
.addReg(ParityRegister);
break;
if (is32BitOpc) {
BuildMI(BB, MI, DL, TII->get(AMDGPU::S_MUL_I32), DstReg)
.addReg(SrcReg)
.addReg(ParityRegister);
break;
} else {
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

No else after break

Register DestSub0 =
MRI.createVirtualRegister(&AMDGPU::SReg_32RegClass);
Register DestSub1 =
MRI.createVirtualRegister(&AMDGPU::SReg_32RegClass);

const TargetRegisterClass *SrcRC = MRI.getRegClass(SrcReg);
const TargetRegisterClass *SrcSubRC =
TRI->getSubRegisterClass(SrcRC, AMDGPU::sub0);

MachineOperand Op1L = TII->buildExtractSubRegOrImm(
MI, MRI, MI.getOperand(1), SrcRC, AMDGPU::sub0, SrcSubRC);
MachineOperand Op1H = TII->buildExtractSubRegOrImm(
MI, MRI, MI.getOperand(1), SrcRC, AMDGPU::sub1, SrcSubRC);

BuildMI(BB, MI, DL, TII->get(AMDGPU::S_MUL_I32), DestSub0)
.add(Op1L)
.addReg(ParityRegister);

BuildMI(BB, MI, DL, TII->get(AMDGPU::S_MUL_I32), DestSub1)
.add(Op1H)
.addReg(ParityRegister);

BuildMI(BB, MI, DL, TII->get(TargetOpcode::REG_SEQUENCE), DstReg)
.addReg(DestSub0)
.addImm(AMDGPU::sub0)
.addReg(DestSub1)
.addImm(AMDGPU::sub1);
break;
}
}
case AMDGPU::S_SUB_I32: {
Register NegatedVal = MRI.createVirtualRegister(DstRegClass);
Expand Down Expand Up @@ -5407,6 +5445,15 @@ static MachineBasicBlock *lowerWaveReduce(MachineInstr &MI,
.addReg(LaneValueHiReg)
.addImm(AMDGPU::sub1);
switch (Opc) {
case ::AMDGPU::S_OR_B64:
case ::AMDGPU::S_AND_B64:
case ::AMDGPU::S_XOR_B64: {
NewAccumulator = BuildMI(*ComputeLoop, I, DL, TII->get(Opc), DstReg)
.addReg(Accumulator->getOperand(0).getReg())
.addReg(LaneValue->getOperand(0).getReg())
.setOperandDead(3); // Dead scc
break;
}
case AMDGPU::V_CMP_GT_I64_e64:
case AMDGPU::V_CMP_GT_U64_e64:
case AMDGPU::V_CMP_LT_I64_e64:
Expand Down Expand Up @@ -5538,10 +5585,16 @@ SITargetLowering::EmitInstrWithCustomInserter(MachineInstr &MI,
return lowerWaveReduce(MI, *BB, *getSubtarget(), AMDGPU::S_SUB_U64_PSEUDO);
case AMDGPU::WAVE_REDUCE_AND_PSEUDO_B32:
return lowerWaveReduce(MI, *BB, *getSubtarget(), AMDGPU::S_AND_B32);
case AMDGPU::WAVE_REDUCE_AND_PSEUDO_B64:
return lowerWaveReduce(MI, *BB, *getSubtarget(), AMDGPU::S_AND_B64);
case AMDGPU::WAVE_REDUCE_OR_PSEUDO_B32:
return lowerWaveReduce(MI, *BB, *getSubtarget(), AMDGPU::S_OR_B32);
case AMDGPU::WAVE_REDUCE_OR_PSEUDO_B64:
return lowerWaveReduce(MI, *BB, *getSubtarget(), AMDGPU::S_OR_B64);
case AMDGPU::WAVE_REDUCE_XOR_PSEUDO_B32:
return lowerWaveReduce(MI, *BB, *getSubtarget(), AMDGPU::S_XOR_B32);
case AMDGPU::WAVE_REDUCE_XOR_PSEUDO_B64:
return lowerWaveReduce(MI, *BB, *getSubtarget(), AMDGPU::S_XOR_B64);
case AMDGPU::S_UADDO_PSEUDO:
case AMDGPU::S_USUBO_PSEUDO: {
const DebugLoc &DL = MI.getDebugLoc();
Expand Down
3 changes: 3 additions & 0 deletions llvm/lib/Target/AMDGPU/SIInstructions.td
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -347,6 +347,9 @@ defvar Operations = [
WaveReduceOp<"max", "I64", i64, SGPR_64, VSrc_b64>,
WaveReduceOp<"add", "U64", i64, SGPR_64, VSrc_b64>,
WaveReduceOp<"sub", "U64", i64, SGPR_64, VSrc_b64>,
WaveReduceOp<"and", "B64", i64, SGPR_64, VSrc_b64>,
WaveReduceOp<"or", "B64", i64, SGPR_64, VSrc_b64>,
WaveReduceOp<"xor", "B64", i64, SGPR_64, VSrc_b64>,
];

foreach Op = Operations in {
Expand Down
Loading