[MLIR][Affine] Fix crash in loop unswitching/hoistAffineIfOp #130401
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
@llvm/pr-subscribers-mlir Author: Uday Bondhugula (bondhugula) ChangesFix obvious crash as a result of missing affine.parallel handling. Also, Fixes: #62323 Full diff: https://github.com/llvm/llvm-project/pull/130401.diff 2 Files Affected:
diff --git a/mlir/lib/Dialect/Affine/Utils/Utils.cpp b/mlir/lib/Dialect/Affine/Utils/Utils.cpp
index 7ef016f88be37..dbad46821d039 100644
--- a/mlir/lib/Dialect/Affine/Utils/Utils.cpp
+++ b/mlir/lib/Dialect/Affine/Utils/Utils.cpp
@@ -263,16 +263,17 @@ static void promoteIfBlock(AffineIfOp ifOp, bool elseBlock) {
static Operation *getOutermostInvariantForOp(AffineIfOp ifOp) {
// Walk up the parents past all for op that this conditional is invariant on.
auto ifOperands = ifOp.getOperands();
- auto *res = ifOp.getOperation();
- while (!isa<func::FuncOp>(res->getParentOp())) {
+ Operation *res = ifOp;
+ while (!res->getParentOp()->hasTrait<OpTrait::IsIsolatedFromAbove>()) {
auto *parentOp = res->getParentOp();
if (auto forOp = dyn_cast<AffineForOp>(parentOp)) {
if (llvm::is_contained(ifOperands, forOp.getInductionVar()))
break;
} else if (auto parallelOp = dyn_cast<AffineParallelOp>(parentOp)) {
- for (auto iv : parallelOp.getIVs())
- if (llvm::is_contained(ifOperands, iv))
- break;
+ if (llvm::any_of(parallelOp.getIVs(), [&](Value iv) {
+ return llvm::is_contained(ifOperands, iv);
+ }))
+ break;
} else if (!isa<AffineIfOp>(parentOp)) {
// Won't walk up past anything other than affine.for/if ops.
break;
@@ -442,7 +443,7 @@ LogicalResult mlir::affine::hoistAffineIfOp(AffineIfOp ifOp, bool *folded) {
// canonicalization is missing composition of affine.applys into it.
assert(llvm::all_of(ifOp.getOperands(),
[](Value v) {
- return isTopLevelValue(v) || isAffineForInductionVar(v);
+ return isTopLevelValue(v) || isAffineInductionVar(v);
}) &&
"operands not composed");
diff --git a/mlir/test/Dialect/Affine/loop-unswitch.mlir b/mlir/test/Dialect/Affine/loop-unswitch.mlir
index 5a58941937bf5..ea94d0b72d823 100644
--- a/mlir/test/Dialect/Affine/loop-unswitch.mlir
+++ b/mlir/test/Dialect/Affine/loop-unswitch.mlir
@@ -254,3 +254,31 @@ func.func @multiple_if(%N : index) {
// CHECK-NEXT: return
func.func private @external()
+
+// Check if affine.parallel ops don't lead to any breakage.
+
+#set = affine_set<(d0) : (-d0 + 3 >= 0)>
+// CHECK-LABEL: affine_parallel
+func.func @affine_parallel(%arg0: memref<35xf32>) -> memref<35xf32> {
+ %0 = llvm.mlir.constant(1.000000e+00 : f32) : f32
+ %alloc = memref.alloc() {alignment = 64 : i64} : memref<35xf32>
+ // CHECK: affine.parallel
+ affine.parallel (%arg1) = (0) to (35) step (32) {
+ // This can't be hoisted further.
+ // CHECK-NEXT: affine.if
+ affine.if #set(%arg1) {
+ affine.parallel (%arg2) = (%arg1) to (%arg1 + 32) {
+ %1 = affine.load %arg0[%arg2] : memref<35xf32>
+ %2 = llvm.fdiv %0, %1 : f32
+ affine.store %2, %alloc[%arg2] : memref<35xf32>
+ }
+ } else {
+ affine.parallel (%arg2) = (%arg1) to (min(%arg1 + 32, 35)) {
+ %1 = affine.load %arg0[%arg2] : memref<35xf32>
+ %2 = llvm.fdiv %0, %1 : f32
+ affine.store %2, %alloc[%arg2] : memref<35xf32>
+ }
+ }
+ }
+ return %alloc : memref<35xf32>
+}
|
|
@llvm/pr-subscribers-mlir-affine Author: Uday Bondhugula (bondhugula) ChangesFix obvious crash as a result of missing affine.parallel handling. Also, Fixes: #62323 Full diff: https://github.com/llvm/llvm-project/pull/130401.diff 2 Files Affected:
diff --git a/mlir/lib/Dialect/Affine/Utils/Utils.cpp b/mlir/lib/Dialect/Affine/Utils/Utils.cpp
index 7ef016f88be37..dbad46821d039 100644
--- a/mlir/lib/Dialect/Affine/Utils/Utils.cpp
+++ b/mlir/lib/Dialect/Affine/Utils/Utils.cpp
@@ -263,16 +263,17 @@ static void promoteIfBlock(AffineIfOp ifOp, bool elseBlock) {
static Operation *getOutermostInvariantForOp(AffineIfOp ifOp) {
// Walk up the parents past all for op that this conditional is invariant on.
auto ifOperands = ifOp.getOperands();
- auto *res = ifOp.getOperation();
- while (!isa<func::FuncOp>(res->getParentOp())) {
+ Operation *res = ifOp;
+ while (!res->getParentOp()->hasTrait<OpTrait::IsIsolatedFromAbove>()) {
auto *parentOp = res->getParentOp();
if (auto forOp = dyn_cast<AffineForOp>(parentOp)) {
if (llvm::is_contained(ifOperands, forOp.getInductionVar()))
break;
} else if (auto parallelOp = dyn_cast<AffineParallelOp>(parentOp)) {
- for (auto iv : parallelOp.getIVs())
- if (llvm::is_contained(ifOperands, iv))
- break;
+ if (llvm::any_of(parallelOp.getIVs(), [&](Value iv) {
+ return llvm::is_contained(ifOperands, iv);
+ }))
+ break;
} else if (!isa<AffineIfOp>(parentOp)) {
// Won't walk up past anything other than affine.for/if ops.
break;
@@ -442,7 +443,7 @@ LogicalResult mlir::affine::hoistAffineIfOp(AffineIfOp ifOp, bool *folded) {
// canonicalization is missing composition of affine.applys into it.
assert(llvm::all_of(ifOp.getOperands(),
[](Value v) {
- return isTopLevelValue(v) || isAffineForInductionVar(v);
+ return isTopLevelValue(v) || isAffineInductionVar(v);
}) &&
"operands not composed");
diff --git a/mlir/test/Dialect/Affine/loop-unswitch.mlir b/mlir/test/Dialect/Affine/loop-unswitch.mlir
index 5a58941937bf5..ea94d0b72d823 100644
--- a/mlir/test/Dialect/Affine/loop-unswitch.mlir
+++ b/mlir/test/Dialect/Affine/loop-unswitch.mlir
@@ -254,3 +254,31 @@ func.func @multiple_if(%N : index) {
// CHECK-NEXT: return
func.func private @external()
+
+// Check if affine.parallel ops don't lead to any breakage.
+
+#set = affine_set<(d0) : (-d0 + 3 >= 0)>
+// CHECK-LABEL: affine_parallel
+func.func @affine_parallel(%arg0: memref<35xf32>) -> memref<35xf32> {
+ %0 = llvm.mlir.constant(1.000000e+00 : f32) : f32
+ %alloc = memref.alloc() {alignment = 64 : i64} : memref<35xf32>
+ // CHECK: affine.parallel
+ affine.parallel (%arg1) = (0) to (35) step (32) {
+ // This can't be hoisted further.
+ // CHECK-NEXT: affine.if
+ affine.if #set(%arg1) {
+ affine.parallel (%arg2) = (%arg1) to (%arg1 + 32) {
+ %1 = affine.load %arg0[%arg2] : memref<35xf32>
+ %2 = llvm.fdiv %0, %1 : f32
+ affine.store %2, %alloc[%arg2] : memref<35xf32>
+ }
+ } else {
+ affine.parallel (%arg2) = (%arg1) to (min(%arg1 + 32, 35)) {
+ %1 = affine.load %arg0[%arg2] : memref<35xf32>
+ %2 = llvm.fdiv %0, %1 : f32
+ affine.store %2, %alloc[%arg2] : memref<35xf32>
+ }
+ }
+ }
+ return %alloc : memref<35xf32>
+}
|
bondhugula
force-pushed
the
uday/hoist_affine_if_crash
branch
from
278e06d to
44d9d33
Compare
Fix obvious crash as a result of missing affine.parallel handling. Also, fix bug exposed in a helper method used by hoistAffineIfOp. Also, remove the specialcasing for FuncOp's. Fixes: llvm#62323
bondhugula
force-pushed
the
uday/hoist_affine_if_crash
branch
from
44d9d33 to
cb46134
Compare
bondhugula
changed the title
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Fix obvious crash as a result of missing affine.parallel handling. Also,
fix bug exposed in a helper method used by hoistAffineIfOp.
Fixes: #62323