Skip to content

[AMDGPU] Use correct SlotIndex to calculate live-out register set. #161720

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

Merged
merged 3 commits into from
Oct 10, 2025
Merged

[AMDGPU] Use correct SlotIndex to calculate live-out register set. #161720

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
d27a6de
[AMDGPU] Use correct SlotIndex to calculate live-out register set.
vpykhtin Oct 2, 2025
4874306
Use getMBBEndIdx(MBB).getPrevSlot() as getMBBLastIdx result.
vpykhtin Oct 3, 2025
9bf2639
Remove unrelated changes, use MBBLastSlot for live-throughs as well.
vpykhtin Oct 3, 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
17 changes: 15 additions & 2 deletions llvm/include/llvm/CodeGen/SlotIndexes.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -467,16 +467,29 @@ class raw_ostream;
return getMBBRange(mbb).first;
}

/// Returns the last index in the given basic block number.
/// Returns the index past the last valid index in the given basic block.
SlotIndex getMBBEndIdx(unsigned Num) const {
return getMBBRange(Num).second;
}

/// Returns the last index in the given basic block.
/// Returns the index past the last valid index in the given basic block.
SlotIndex getMBBEndIdx(const MachineBasicBlock *mbb) const {
return getMBBRange(mbb).second;
}

/// Returns the last valid index in the given basic block.
/// This index corresponds to the dead slot of the last non-debug
/// instruction and can be used to find live-out ranges of the block. Note
/// that getMBBEndIdx returns the start index of the next block, which is
/// also used as the start index for segments with phi-def values. Returns
/// an invalid SlotIndex if the block has no non-debug instructions.
SlotIndex getMBBLastIdx(const MachineBasicBlock *MBB) const {
auto LastInstrI = MBB->getLastNonDebugInstr();
return LastInstrI == MBB->end()
? SlotIndex()
: getInstructionIndex(*LastInstrI).getDeadSlot();
}

/// Iterator over the idx2MBBMap (sorted pairs of slot index of basic block
/// begin and basic block)
using MBBIndexIterator = SmallVectorImpl<IdxMBBPair>::const_iterator;
Expand Down
39 changes: 20 additions & 19 deletions llvm/lib/Target/AMDGPU/GCNRegPressure.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -461,6 +461,8 @@ GCNRPTracker::LiveRegSet llvm::getLiveRegs(SlotIndex SI,
const LiveIntervals &LIS,
const MachineRegisterInfo &MRI) {
GCNRPTracker::LiveRegSet LiveRegs;
if (!SI.isValid())
return LiveRegs;
for (unsigned I = 0, E = MRI.getNumVirtRegs(); I != E; ++I) {
auto Reg = Register::index2VirtReg(I);
if (!LIS.hasInterval(Reg))
Expand Down Expand Up @@ -906,32 +908,31 @@ bool GCNRegPressurePrinter::runOnMachineFunction(MachineFunction &MF) {

SlotIndex MBBStartSlot = LIS.getSlotIndexes()->getMBBStartIdx(&MBB);
SlotIndex MBBEndSlot = LIS.getSlotIndexes()->getMBBEndIdx(&MBB);
SlotIndex MBBLastSlot = LIS.getSlotIndexes()->getMBBLastIdx(&MBB);

GCNRPTracker::LiveRegSet LiveIn, LiveOut;
GCNRegPressure RPAtMBBEnd;

if (UseDownwardTracker) {
if (MBB.empty()) {
LiveIn = LiveOut = getLiveRegs(MBBStartSlot, LIS, MRI);
RPAtMBBEnd = getRegPressure(MRI, LiveIn);
} else {
GCNDownwardRPTracker RPT(LIS);
RPT.reset(MBB.front());
if (!MBBLastSlot.isValid()) { // MBB doesn't have any non-debug instrs.
LiveIn = LiveOut = getLiveRegs(MBBStartSlot, LIS, MRI);
RPAtMBBEnd = getRegPressure(MRI, LiveIn);
} else if (UseDownwardTracker) {
GCNDownwardRPTracker RPT(LIS);
RPT.reset(MBB.front());

LiveIn = RPT.getLiveRegs();

while (!RPT.advanceBeforeNext()) {
GCNRegPressure RPBeforeMI = RPT.getPressure();
RPT.advanceToNext();
RP.emplace_back(RPBeforeMI, RPT.getPressure());
}
LiveIn = RPT.getLiveRegs();

LiveOut = RPT.getLiveRegs();
RPAtMBBEnd = RPT.getPressure();
while (!RPT.advanceBeforeNext()) {
GCNRegPressure RPBeforeMI = RPT.getPressure();
RPT.advanceToNext();
RP.emplace_back(RPBeforeMI, RPT.getPressure());
}

LiveOut = RPT.getLiveRegs();
RPAtMBBEnd = RPT.getPressure();
} else {
GCNUpwardRPTracker RPT(LIS);
RPT.reset(MRI, MBBEndSlot);
RPT.reset(MRI, MBBLastSlot);

LiveOut = RPT.getLiveRegs();
RPAtMBBEnd = RPT.getPressure();
Expand Down Expand Up @@ -965,8 +966,8 @@ bool GCNRegPressurePrinter::runOnMachineFunction(MachineFunction &MF) {
OS << printRP(RPAtMBBEnd) << '\n';

OS << PFX " Live-out:" << llvm::print(LiveOut, MRI);
if (UseDownwardTracker)
ReportLISMismatchIfAny(LiveOut, getLiveRegs(MBBEndSlot, LIS, MRI));
if (UseDownwardTracker && MBBLastSlot.isValid())
ReportLISMismatchIfAny(LiveOut, getLiveRegs(MBBLastSlot, LIS, MRI));

GCNRPTracker::LiveRegSet LiveThrough;
for (auto [Reg, Mask] : LiveIn) {
Expand Down
4 changes: 2 additions & 2 deletions llvm/lib/Target/AMDGPU/GCNRegPressure.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -313,8 +313,8 @@ class GCNUpwardRPTracker : public GCNRPTracker {

/// reset tracker to the end of the \p MBB.
void reset(const MachineBasicBlock &MBB) {
reset(MBB.getParent()->getRegInfo(),
LIS.getSlotIndexes()->getMBBEndIdx(&MBB));
SlotIndex MBBLastSlot = LIS.getSlotIndexes()->getMBBLastIdx(&MBB);
reset(MBB.getParent()->getRegInfo(), MBBLastSlot);
}

/// reset tracker to the point just after \p MI (in program order).
Expand Down
18 changes: 8 additions & 10 deletions llvm/test/CodeGen/AMDGPU/regpressure_printer.mir
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -510,14 +510,14 @@ body: |
; RPU-NEXT: 0 0 $sgpr0 = S_BUFFER_LOAD_DWORD_IMM $sgpr0_sgpr1_sgpr2_sgpr3, 0, 0
; RPU-NEXT: 0 0
; RPU-NEXT: 0 1 undef %0.sub5:vreg_512 = V_MOV_B32_e32 5, implicit $exec
; RPU-NEXT: 0 0
; RPU-NEXT: 0 0 S_CMP_GT_U32 $sgpr0, 15, implicit-def $scc
; RPU-NEXT: 0 0
; RPU-NEXT: 0 0 S_CBRANCH_SCC1 %bb.2, implicit $scc
; RPU-NEXT: 0 0
; RPU-NEXT: 0 0 S_BRANCH %bb.1
; RPU-NEXT: 0 0
; RPU-NEXT: Live-out:
; RPU-NEXT: 0 1
; RPU-NEXT: 0 1 S_CMP_GT_U32 $sgpr0, 15, implicit-def $scc
; RPU-NEXT: 0 1
; RPU-NEXT: 0 1 S_CBRANCH_SCC1 %bb.2, implicit $scc
; RPU-NEXT: 0 1
; RPU-NEXT: 0 1 S_BRANCH %bb.1
; RPU-NEXT: 0 1
; RPU-NEXT: Live-out: %0:0000000000000C00
; RPU-NEXT: Live-thr:
; RPU-NEXT: 0 0
; RPU-NEXT: bb.1:
Expand Down Expand Up @@ -571,8 +571,6 @@ body: |
; RPD-NEXT: 0 1 S_BRANCH %bb.1
; RPD-NEXT: 0 1
; RPD-NEXT: Live-out: %0:0000000000000C00
; RPD-NEXT: mis LIS:
; RPD-NEXT: %0:L0000000000000C00 isn't found in LIS reported set
; RPD-NEXT: Live-thr:
; RPD-NEXT: 0 0
; RPD-NEXT: bb.1:
Expand Down