[OpenMP] Fix calculation of dependencies for multi-dimensional iteration space #99347
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
llvmbot
added
clang
|
@llvm/pr-subscribers-clang Author: Joachim (jprotze) ChangesThe expectation for multiple iterators used in a single depend clause ( Full diff: https://github.com/llvm/llvm-project/pull/99347.diff 3 Files Affected:
diff --git a/clang/lib/CodeGen/CGOpenMPRuntime.cpp b/clang/lib/CodeGen/CGOpenMPRuntime.cpp
index 652fb700fc6af..a3ade3780b39a 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntime.cpp
+++ b/clang/lib/CodeGen/CGOpenMPRuntime.cpp
@@ -4259,14 +4259,18 @@ std::pair<llvm::Value *, Address> CGOpenMPRuntime::emitDependClause(
// Include number of iterations, if any.
if (const auto *IE = cast_or_null<OMPIteratorExpr>(D.IteratorExpr)) {
+ llvm::Value *ClauseIteratorSpace =
+ llvm::ConstantInt::get(CGF.IntPtrTy, 1);
for (unsigned I = 0, E = IE->numOfIterators(); I < E; ++I) {
llvm::Value *Sz = CGF.EmitScalarExpr(IE->getHelper(I).Upper);
Sz = CGF.Builder.CreateIntCast(Sz, CGF.IntPtrTy, /*isSigned=*/false);
- llvm::Value *NumClauseDeps = CGF.Builder.CreateNUWMul(
- Sz, llvm::ConstantInt::get(CGF.IntPtrTy, D.DepExprs.size()));
- NumOfRegularWithIterators =
- CGF.Builder.CreateNUWAdd(NumOfRegularWithIterators, NumClauseDeps);
+ ClauseIteratorSpace = CGF.Builder.CreateNUWMul(Sz, ClauseIteratorSpace);
}
+ llvm::Value *NumClauseDeps =
+ CGF.Builder.CreateNUWMul(ClauseIteratorSpace,
+ llvm::ConstantInt::get(CGF.IntPtrTy, D.DepExprs.size()));
+ NumOfRegularWithIterators =
+ CGF.Builder.CreateNUWAdd(NumOfRegularWithIterators, NumClauseDeps);
HasRegularWithIterators = true;
continue;
}
diff --git a/clang/test/OpenMP/depend_iterator_bug.c b/clang/test/OpenMP/depend_iterator_bug.c
index b4aaaac08374f..ff11d8a5d0e40 100644
--- a/clang/test/OpenMP/depend_iterator_bug.c
+++ b/clang/test/OpenMP/depend_iterator_bug.c
@@ -5,6 +5,7 @@
int x[100];
int y[100];
+int z[100][100];
// CHECK-LABEL: @many_iterators_single_clause(
// CHECK: [[VLA:%.*]] = alloca [[STRUCT_KMP_DEPEND_INFO:%.*]], i64 10, align 16
@@ -24,3 +25,31 @@ void many_iterators_many_clauses(void) {
{
}
}
+
+// CHECK-LABEL: @multidim_iterators_clause1(
+// CHECK: [[VLA:%.*]] = alloca [[STRUCT_KMP_DEPEND_INFO:%.*]], i64 1, align 16
+// CHECK: = call i32 @__kmpc_omp_task_with_deps(ptr {{.*}}, i32 {{.*}}, ptr {{.*}}, i32 1, ptr {{.*}}, i32 0, ptr null)
+void multidim_iterators_clause1(void) {
+ #pragma omp task depend(iterator(i=0:1, j=0:1), in: z[i][j])
+ {
+ }
+}
+
+// CHECK-LABEL: @multidim_iterators_offset_clause(
+// CHECK: [[VLA:%.*]] = alloca [[STRUCT_KMP_DEPEND_INFO:%.*]], i64 1, align 16
+// CHECK: = call i32 @__kmpc_omp_task_with_deps(ptr {{.*}}, i32 {{.*}}, ptr {{.*}}, i32 1, ptr {{.*}}, i32 0, ptr null)
+void multidim_iterators_offset_clause(void) {
+ #pragma omp task depend(iterator(i=5:6, j=10:11), in: z[i][j])
+ {
+ }
+}
+
+// CHECK-LABEL: @multidim_iterators_clause25(
+// CHECK: [[VLA:%.*]] = alloca [[STRUCT_KMP_DEPEND_INFO:%.*]], i64 25, align 16
+// CHECK: = call i32 @__kmpc_omp_task_with_deps(ptr {{.*}}, i32 {{.*}}, ptr {{.*}}, i32 25, ptr {{.*}}, i32 0, ptr null)
+void multidim_iterators_clause25(void) {
+ #pragma omp task depend(iterator(i=0:5, j=0:5), in: z[i][j])
+ {
+ }
+}
+
diff --git a/clang/test/OpenMP/task_codegen.c b/clang/test/OpenMP/task_codegen.c
index be404827ce901..0ca815c0eccf1 100644
--- a/clang/test/OpenMP/task_codegen.c
+++ b/clang/test/OpenMP/task_codegen.c
@@ -139,7 +139,8 @@ for (int i = 0; i < 10; ++i)
// CHECK: [[EB_SUB_2_ADD_1_SUB:%.+]] = sub i32 [[EB_SUB_2_ADD]], 1
// CHECK: [[EB_SUB_2_ADD_1_SUB_2_DIV:%.+]] = udiv i32 [[EB_SUB_2_ADD_1_SUB]], 2
// CHECK: [[ELEMS:%.+]] = zext i32 [[EB_SUB_2_ADD_1_SUB_2_DIV]] to i64
- // CHECK: [[NELEMS:%.+]] = mul nuw i64 [[ELEMS]], 1
+ // CHECK: [[ELEMS2:%.+]] = mul nuw i64 [[ELEMS]], 1
+ // CHECK: [[NELEMS:%.+]] = mul nuw i64 [[ELEMS2]], 1
// ITERATOR_TOTAL = NELEMS + 0;
// CHECK: [[ITERATOR_TOTAL:%.+]] = add nuw i64 0, [[NELEMS]]
|
|
@llvm/pr-subscribers-clang-codegen Author: Joachim (jprotze) ChangesThe expectation for multiple iterators used in a single depend clause ( Full diff: https://github.com/llvm/llvm-project/pull/99347.diff 3 Files Affected:
diff --git a/clang/lib/CodeGen/CGOpenMPRuntime.cpp b/clang/lib/CodeGen/CGOpenMPRuntime.cpp
index 652fb700fc6af..a3ade3780b39a 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntime.cpp
+++ b/clang/lib/CodeGen/CGOpenMPRuntime.cpp
@@ -4259,14 +4259,18 @@ std::pair<llvm::Value *, Address> CGOpenMPRuntime::emitDependClause(
// Include number of iterations, if any.
if (const auto *IE = cast_or_null<OMPIteratorExpr>(D.IteratorExpr)) {
+ llvm::Value *ClauseIteratorSpace =
+ llvm::ConstantInt::get(CGF.IntPtrTy, 1);
for (unsigned I = 0, E = IE->numOfIterators(); I < E; ++I) {
llvm::Value *Sz = CGF.EmitScalarExpr(IE->getHelper(I).Upper);
Sz = CGF.Builder.CreateIntCast(Sz, CGF.IntPtrTy, /*isSigned=*/false);
- llvm::Value *NumClauseDeps = CGF.Builder.CreateNUWMul(
- Sz, llvm::ConstantInt::get(CGF.IntPtrTy, D.DepExprs.size()));
- NumOfRegularWithIterators =
- CGF.Builder.CreateNUWAdd(NumOfRegularWithIterators, NumClauseDeps);
+ ClauseIteratorSpace = CGF.Builder.CreateNUWMul(Sz, ClauseIteratorSpace);
}
+ llvm::Value *NumClauseDeps =
+ CGF.Builder.CreateNUWMul(ClauseIteratorSpace,
+ llvm::ConstantInt::get(CGF.IntPtrTy, D.DepExprs.size()));
+ NumOfRegularWithIterators =
+ CGF.Builder.CreateNUWAdd(NumOfRegularWithIterators, NumClauseDeps);
HasRegularWithIterators = true;
continue;
}
diff --git a/clang/test/OpenMP/depend_iterator_bug.c b/clang/test/OpenMP/depend_iterator_bug.c
index b4aaaac08374f..ff11d8a5d0e40 100644
--- a/clang/test/OpenMP/depend_iterator_bug.c
+++ b/clang/test/OpenMP/depend_iterator_bug.c
@@ -5,6 +5,7 @@
int x[100];
int y[100];
+int z[100][100];
// CHECK-LABEL: @many_iterators_single_clause(
// CHECK: [[VLA:%.*]] = alloca [[STRUCT_KMP_DEPEND_INFO:%.*]], i64 10, align 16
@@ -24,3 +25,31 @@ void many_iterators_many_clauses(void) {
{
}
}
+
+// CHECK-LABEL: @multidim_iterators_clause1(
+// CHECK: [[VLA:%.*]] = alloca [[STRUCT_KMP_DEPEND_INFO:%.*]], i64 1, align 16
+// CHECK: = call i32 @__kmpc_omp_task_with_deps(ptr {{.*}}, i32 {{.*}}, ptr {{.*}}, i32 1, ptr {{.*}}, i32 0, ptr null)
+void multidim_iterators_clause1(void) {
+ #pragma omp task depend(iterator(i=0:1, j=0:1), in: z[i][j])
+ {
+ }
+}
+
+// CHECK-LABEL: @multidim_iterators_offset_clause(
+// CHECK: [[VLA:%.*]] = alloca [[STRUCT_KMP_DEPEND_INFO:%.*]], i64 1, align 16
+// CHECK: = call i32 @__kmpc_omp_task_with_deps(ptr {{.*}}, i32 {{.*}}, ptr {{.*}}, i32 1, ptr {{.*}}, i32 0, ptr null)
+void multidim_iterators_offset_clause(void) {
+ #pragma omp task depend(iterator(i=5:6, j=10:11), in: z[i][j])
+ {
+ }
+}
+
+// CHECK-LABEL: @multidim_iterators_clause25(
+// CHECK: [[VLA:%.*]] = alloca [[STRUCT_KMP_DEPEND_INFO:%.*]], i64 25, align 16
+// CHECK: = call i32 @__kmpc_omp_task_with_deps(ptr {{.*}}, i32 {{.*}}, ptr {{.*}}, i32 25, ptr {{.*}}, i32 0, ptr null)
+void multidim_iterators_clause25(void) {
+ #pragma omp task depend(iterator(i=0:5, j=0:5), in: z[i][j])
+ {
+ }
+}
+
diff --git a/clang/test/OpenMP/task_codegen.c b/clang/test/OpenMP/task_codegen.c
index be404827ce901..0ca815c0eccf1 100644
--- a/clang/test/OpenMP/task_codegen.c
+++ b/clang/test/OpenMP/task_codegen.c
@@ -139,7 +139,8 @@ for (int i = 0; i < 10; ++i)
// CHECK: [[EB_SUB_2_ADD_1_SUB:%.+]] = sub i32 [[EB_SUB_2_ADD]], 1
// CHECK: [[EB_SUB_2_ADD_1_SUB_2_DIV:%.+]] = udiv i32 [[EB_SUB_2_ADD_1_SUB]], 2
// CHECK: [[ELEMS:%.+]] = zext i32 [[EB_SUB_2_ADD_1_SUB_2_DIV]] to i64
- // CHECK: [[NELEMS:%.+]] = mul nuw i64 [[ELEMS]], 1
+ // CHECK: [[ELEMS2:%.+]] = mul nuw i64 [[ELEMS]], 1
+ // CHECK: [[NELEMS:%.+]] = mul nuw i64 [[ELEMS2]], 1
// ITERATOR_TOTAL = NELEMS + 0;
// CHECK: [[ITERATOR_TOTAL:%.+]] = add nuw i64 0, [[NELEMS]]
|
|
✅ With the latest revision this PR passed the C/C++ code formatter. |
jprotze
force-pushed
the
fix-omp-multi-iterators
branch
from
655acc6 to
b45379e
Compare
jprotze
force-pushed
the
fix-omp-multi-iterators
branch
from
b45379e to
033b032
Compare
yuxuanchen1997
pushed a commit
that referenced
this pull request
Jul 25, 2024
…ion space (#99347) The expectation for multiple iterators used in a single depend clause (`depend(iterator(i=0:5,j=0:5), in:x[i][j])`) is that the iterator space is the product of the iteration vectors (25 in that case). The current codeGen only works correctly, if `numIterators() = 1`. For more iterators, the execution results in runtime assertions or segfaults. The modified codeGen first calculates the iteration space, then multiplies to the number of dependencies in the depend clause and finally adds to the total number of iterator dependencies.
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.
The expectation for multiple iterators used in a single depend clause (
depend(iterator(i=0:5,j=0:5), in:x[i][j])) is that the iterator space is the product of the iteration vectors (25 in that case). The current codeGen only works correctly, ifnumIterators() = 1. For more iterators, the execution results in runtime assertions or segfaults.The modified codeGen first calculates the iteration space, then multiplies to the number of dependencies in the depend clause and finally adds to the total number of iterator dependencies.