Skip to content

[MLIR][SCF] fix loop pipelining pass use of uninitialized value #146991

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 1 commit into from
Jul 5, 2025
Merged

[MLIR][SCF] fix loop pipelining pass use of uninitialized value #146991

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
29 changes: 15 additions & 14 deletions mlir/lib/Dialect/SCF/Transforms/LoopPipelining.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -106,6 +106,20 @@ bool LoopPipelinerInternal::initializeLoopInfo(
lb = forOp.getLowerBound();
step = forOp.getStep();

std::vector<std::pair<Operation *, unsigned>> schedule;
options.getScheduleFn(forOp, schedule);
if (schedule.empty()) {
LDBG("--empty schedule -> BAIL");
return false;
}

opOrder.reserve(schedule.size());
Copy link

Copilot AI Jul 4, 2025

Choose a reason for hiding this comment

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

Before reserving and populating opOrder and stages, ensure both containers are cleared (e.g., opOrder.clear(); stages.clear();) to avoid carrying over any stale entries when initializeLoopInfo is called multiple times.

Copilot uses AI. Check for mistakes.
for (auto &opSchedule : schedule) {
maxStage = std::max(maxStage, opSchedule.second);
stages[opSchedule.first] = opSchedule.second;
opOrder.push_back(opSchedule.first);
}

dynamicLoop = true;
auto upperBoundCst = getConstantIntValue(ub);
auto lowerBoundCst = getConstantIntValue(lb);
Expand All @@ -124,7 +138,7 @@ bool LoopPipelinerInternal::initializeLoopInfo(
return false;
}
int64_t numIteration = llvm::divideCeilSigned(ubImm - lbImm, stepImm);
if (numIteration > maxStage) {
if (numIteration >= maxStage) {
dynamicLoop = false;
} else if (!options.supportDynamicLoops) {
LDBG("--fewer loop iterations than pipeline stages -> BAIL");
Expand All @@ -137,19 +151,6 @@ bool LoopPipelinerInternal::initializeLoopInfo(
LDBG("--no epilogue or predicate set -> BAIL");
return false;
}
std::vector<std::pair<Operation *, unsigned>> schedule;
options.getScheduleFn(forOp, schedule);
if (schedule.empty()) {
LDBG("--empty schedule -> BAIL");
return false;
}

opOrder.reserve(schedule.size());
for (auto &opSchedule : schedule) {
maxStage = std::max(maxStage, opSchedule.second);
stages[opSchedule.first] = opSchedule.second;
opOrder.push_back(opSchedule.first);
}

// All operations need to have a stage.
for (Operation &op : forOp.getBody()->without_terminator()) {
Expand Down
36 changes: 36 additions & 0 deletions mlir/test/Dialect/SCF/loop-pipelining.mlir
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -34,6 +34,42 @@ func.func @simple_pipeline(%A: memref<?xf32>, %result: memref<?xf32>) {
return
}

// -----

// A static loop does not satisfy `numIteration >= maxStage`

// CHECK-LABEL: func.func @iteration_lt_stage(
// CHECK-DAG: %[[C0:.*]] = arith.constant 0 : index
// CHECK-DAG: %[[C1:.*]] = arith.constant 1 : index
// CHECK-DAG: %[[C_NEG1:.*]] = arith.constant -1 : index
// CHECK-DAG: %[[TRUE:.*]] = arith.constant true
// CHECK-DAG: %[[FALSE:.*]] = arith.constant false
// Prologue:
// CHECK: scf.if %[[TRUE]]
// CHECK: scf.if %[[TRUE]]
// CHECK: scf.if %[[FALSE]]
// Kernel:
// CHECK: scf.for %[[IV:.*]] = %[[C0:.*]] to %[[C_NEG1:.*]] step %[[C1:.*]]
// Epilogue:
// CHECK: scf.if %[[TRUE]]
// CHECK: scf.if %[[TRUE]]
// CHECK: scf.if %[[TRUE]]
// CHECK: scf.if %[[TRUE]]
// CHECK: scf.if %[[FALSE]]
// CHECK: scf.if %[[FALSE]]
func.func @iteration_lt_stage(%A: memref<?xf32>, %result: memref<?xf32>) {
%c0 = arith.constant 0 : index
%c1 = arith.constant 1 : index
%c2 = arith.constant 2 : index
%cf = arith.constant 1.0 : f32
scf.for %i0 = %c0 to %c2 step %c1 {
%A_elem = memref.load %A[%i0] { __test_pipelining_stage__ = 0, __test_pipelining_op_order__ = 2 } : memref<?xf32>
%A1_elem = arith.addf %A_elem, %cf { __test_pipelining_stage__ = 3, __test_pipelining_op_order__ = 0 } : f32
memref.store %A1_elem, %result[%i0] { __test_pipelining_stage__ = 3, __test_pipelining_op_order__ = 1 } : memref<?xf32>
} { __test_pipelining_loop__ }
return
}


// -----

Expand Down
Loading