Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly
Appearance settings

[mlir][OpenMP] Fix broken insertion point for charbox with omp task #143112

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

Jump to bottom
Open
wants to merge 1 commit into
base: main
Choose a base branch
Loading
from
Open

[mlir][OpenMP] Fix broken insertion point for charbox with omp task #143112

wants to merge 1 commit into from

Conversation

tblah
Copy link
Contributor

@tblah tblah commented Jun 6, 2025

Fixes #142365

@llvmbot
Copy link
Member

llvmbot commented Jun 6, 2025 •
edited
Loading

@llvm/pr-subscribers-mlir-openmp
@llvm/pr-subscribers-flang-openmp

@llvm/pr-subscribers-mlir-llvm

Author: Tom Eccles (tblah)

Changes

Fixes #142365

Full diff: https://github.com/llvm/llvm-project/pull/143112.diff

2 Files Affected:

  • (modified) mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp (+1-2)
  • (added) mlir/test/Target/LLVMIR/openmp-task-charbox.mlir (+87)
diff --git a/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp b/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
index ec1a9c909c94a..069a5cbb9112d 100644
--- a/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
+++ b/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
@@ -2293,8 +2293,7 @@ convertOmpTaskOp(omp::TaskOp taskOp, llvm::IRBuilderBase &builder,
     if (!privateVarOrErr)
       return handleError(privateVarOrErr, *taskOp.getOperation());
 
-    llvm::IRBuilderBase::InsertPointGuard guard(builder);
-    builder.SetInsertPoint(builder.GetInsertBlock()->getTerminator());
+    setInsertPointForPossiblyEmptyBlock(builder);
 
     // TODO: this is a bit of a hack for Fortran character boxes.
     // Character boxes are passed by value into the init region and then the
diff --git a/mlir/test/Target/LLVMIR/openmp-task-charbox.mlir b/mlir/test/Target/LLVMIR/openmp-task-charbox.mlir
new file mode 100644
index 0000000000000..7a448f74ed648
--- /dev/null
+++ b/mlir/test/Target/LLVMIR/openmp-task-charbox.mlir
@@ -0,0 +1,87 @@
+// RUN: mlir-translate --mlir-to-llvmir %s | FileCheck %s
+
+// Regression test for a compiler crash. Ensure that the insertion point is set
+// correctly when triggering the charbox hack multiple times.
+// Nonsense test code to minimally reproduce the issue.
+
+module {
+  llvm.func @free(!llvm.ptr)
+  llvm.func @malloc(i64) -> !llvm.ptr
+  omp.private {type = private} @_QFEc2_private_box_heap_c8xU : !llvm.struct<(ptr, i64, i32, i8, i8, i8, i8)> init {
+  ^bb0(%arg0: !llvm.ptr, %arg1: !llvm.ptr):
+    %0 = llvm.mlir.constant(24 : i32) : i32
+    %1 = llvm.mlir.constant(0 : i64) : i64
+    %2 = llvm.mlir.constant(1 : i32) : i32
+    %3 = llvm.alloca %2 x !llvm.struct<(ptr, i64, i32, i8, i8, i8, i8)> {alignment = 8 : i64} : (i32) -> !llvm.ptr
+    "llvm.intr.memcpy"(%3, %arg0, %0) <{isVolatile = false}> : (!llvm.ptr, !llvm.ptr, i32) -> ()
+    %6 = llvm.ptrtoint %arg0 : !llvm.ptr to i64
+    %7 = llvm.icmp "eq" %6, %1 : i64
+    llvm.cond_br %7, ^bb1, ^bb2
+  ^bb1:  // pred: ^bb0
+    llvm.br ^bb3
+  ^bb2:  // pred: ^bb0
+    llvm.br ^bb3
+  ^bb3:  // 2 preds: ^bb1, ^bb2
+    omp.yield(%arg1 : !llvm.ptr)
+  } dealloc {
+  ^bb0(%arg0: !llvm.ptr):
+    omp.yield
+  }
+  omp.private {type = private} @_QFEc1_private_box_ptr_c8xU : !llvm.struct<(ptr, i64, i32, i8, i8, i8, i8)> init {
+  ^bb0(%arg0: !llvm.ptr, %arg1: !llvm.ptr):
+    %0 = llvm.mlir.constant(24 : i32) : i32
+    %1 = llvm.mlir.constant(1 : i32) : i32
+    %2 = llvm.alloca %1 x !llvm.struct<(ptr, i64, i32, i8, i8, i8, i8)> {alignment = 8 : i64} : (i32) -> !llvm.ptr
+    "llvm.intr.memcpy"(%2, %arg0, %0) <{isVolatile = false}> : (!llvm.ptr, !llvm.ptr, i32) -> ()
+    omp.yield(%arg1 : !llvm.ptr)
+  }
+  llvm.func @_QQmain() {
+    %0 = llvm.mlir.constant(1 : i64) : i64
+    %1 = llvm.alloca %0 x !llvm.struct<(ptr, i64, i32, i8, i8, i8, i8)> {bindc_name = "c2"} : (i64) -> !llvm.ptr
+    %2 = llvm.alloca %0 x !llvm.struct<(ptr, i64, i32, i8, i8, i8, i8)> {bindc_name = "c1"} : (i64) -> !llvm.ptr
+    omp.task private(@_QFEc1_private_box_ptr_c8xU %2 -> %arg0, @_QFEc2_private_box_heap_c8xU %1 -> %arg1 : !llvm.ptr, !llvm.ptr) {
+      omp.terminator
+    }
+    llvm.return
+  }
+}
+
+// CHECK-LABEL: @_QQmain() {
+// CHECK:         %[[STRUCTARG:.*]] = alloca { ptr }, align 8
+// CHECK:         %[[VAL_0:.*]] = alloca { ptr, i64, i32, i8, i8, i8, i8 }, i64 1, align 8
+// CHECK:         br label %[[VAL_2:.*]]
+// CHECK:       entry:                                            ; preds = %[[VAL_3:.*]]
+// CHECK:         br label %[[VAL_4:.*]]
+// CHECK:       omp.private.init:                                 ; preds = %[[VAL_2]]
+// CHECK:         %[[VAL_5:.*]] = tail call ptr @malloc(i64 ptrtoint (ptr getelementptr ({ { ptr, i64, i32, i8, i8, i8, i8 }, { ptr, i64, i32, i8, i8, i8, i8 } }, ptr null, i32 1) to i64))
+// CHECK:         %[[VAL_6:.*]] = getelementptr { { ptr, i64, i32, i8, i8, i8, i8 }, { ptr, i64, i32, i8, i8, i8, i8 } }, ptr %[[VAL_5]], i32 0, i32 0
+// CHECK:         %[[VAL_7:.*]] = getelementptr { { ptr, i64, i32, i8, i8, i8, i8 }, { ptr, i64, i32, i8, i8, i8, i8 } }, ptr %[[VAL_5]], i32 0, i32 1
+// ...
+// CHECK:         br label %[[VAL_9:.*]]
+// CHECK:       omp.private.init4:                                ; preds = %[[VAL_10:.*]], %[[VAL_11:.*]]
+// CHECK:         br label %[[VAL_12:.*]]
+// CHECK:       omp.private.init3:                                ; preds = %[[VAL_9]]
+// CHECK:         br label %[[VAL_13:.*]]
+// CHECK:       omp.private.init2:                                ; preds = %[[VAL_9]]
+// CHECK:         br label %[[VAL_13]]
+// CHECK:       omp.private.init1:                                ; preds = %[[VAL_4]]
+// CHECK:         %[[VAL_14:.*]] = alloca { ptr, i64, i32, i8, i8, i8, i8 }, align 8
+// CHECK:         call void @llvm.memcpy.p0.p0.i32(ptr %[[VAL_14]], ptr %[[VAL_0]], i32 24, i1 false)
+// CHECK:         %[[VAL_15:.*]] = ptrtoint ptr %[[VAL_0]] to i64
+// CHECK:         %[[VAL_16:.*]] = icmp eq i64 %[[VAL_15]], 0
+// CHECK:         br i1 %[[VAL_16]], label %[[VAL_10]], label %[[VAL_11]]
+// CHECK:       omp.region.cont:                                  ; preds = %[[VAL_13]]
+// CHECK:         %[[VAL_17:.*]] = phi ptr [ %[[VAL_7]], %[[VAL_13]] ]
+// CHECK:         br label %[[VAL_18:.*]]
+// CHECK:       omp.private.copy:                                 ; preds = %[[VAL_12]]
+// CHECK:         br label %[[VAL_19:.*]]
+// CHECK:       omp.task.start:                                   ; preds = %[[VAL_18]]
+// CHECK:         br label %[[VAL_20:.*]]
+// CHECK:       codeRepl:                                         ; preds = %[[VAL_19]]
+// CHECK:         %[[VAL_21:.*]] = getelementptr { ptr }, ptr %[[STRUCTARG]], i32 0, i32 0
+// CHECK:         store ptr %[[VAL_5]], ptr %[[VAL_21]], align 8
+// CHECK:         %[[VAL_22:.*]] = call i32 @__kmpc_global_thread_num(ptr @1)
+// CHECK:         %[[VAL_23:.*]] = call ptr @__kmpc_omp_task_alloc(ptr @1, i32 %[[VAL_22]], i32 1, i64 40, i64 8, ptr @_QQmain..omp_par)
+// CHECK:         %[[VAL_24:.*]] = load ptr, ptr %[[VAL_23]], align 8
+// CHECK:         call void @llvm.memcpy.p0.p0.i64(ptr align 1 %[[VAL_24]], ptr align 1 %[[STRUCTARG]], i64 8, i1 false)
+// CHECK:         %[[VAL_25:.*]] = call i32 @__kmpc_omp_task(ptr @1, i32 %[[VAL_22]], ptr %[[VAL_23]])

@llvmbot
Copy link
Member

llvmbot commented Jun 6, 2025

@llvm/pr-subscribers-mlir

Author: Tom Eccles (tblah)

Changes

Fixes #142365

Full diff: https://github.com/llvm/llvm-project/pull/143112.diff

2 Files Affected:

  • (modified) mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp (+1-2)
  • (added) mlir/test/Target/LLVMIR/openmp-task-charbox.mlir (+87)
diff --git a/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp b/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
index ec1a9c909c94a..069a5cbb9112d 100644
--- a/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
+++ b/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
@@ -2293,8 +2293,7 @@ convertOmpTaskOp(omp::TaskOp taskOp, llvm::IRBuilderBase &builder,
     if (!privateVarOrErr)
       return handleError(privateVarOrErr, *taskOp.getOperation());
 
-    llvm::IRBuilderBase::InsertPointGuard guard(builder);
-    builder.SetInsertPoint(builder.GetInsertBlock()->getTerminator());
+    setInsertPointForPossiblyEmptyBlock(builder);
 
     // TODO: this is a bit of a hack for Fortran character boxes.
     // Character boxes are passed by value into the init region and then the
diff --git a/mlir/test/Target/LLVMIR/openmp-task-charbox.mlir b/mlir/test/Target/LLVMIR/openmp-task-charbox.mlir
new file mode 100644
index 0000000000000..7a448f74ed648
--- /dev/null
+++ b/mlir/test/Target/LLVMIR/openmp-task-charbox.mlir
@@ -0,0 +1,87 @@
+// RUN: mlir-translate --mlir-to-llvmir %s | FileCheck %s
+
+// Regression test for a compiler crash. Ensure that the insertion point is set
+// correctly when triggering the charbox hack multiple times.
+// Nonsense test code to minimally reproduce the issue.
+
+module {
+  llvm.func @free(!llvm.ptr)
+  llvm.func @malloc(i64) -> !llvm.ptr
+  omp.private {type = private} @_QFEc2_private_box_heap_c8xU : !llvm.struct<(ptr, i64, i32, i8, i8, i8, i8)> init {
+  ^bb0(%arg0: !llvm.ptr, %arg1: !llvm.ptr):
+    %0 = llvm.mlir.constant(24 : i32) : i32
+    %1 = llvm.mlir.constant(0 : i64) : i64
+    %2 = llvm.mlir.constant(1 : i32) : i32
+    %3 = llvm.alloca %2 x !llvm.struct<(ptr, i64, i32, i8, i8, i8, i8)> {alignment = 8 : i64} : (i32) -> !llvm.ptr
+    "llvm.intr.memcpy"(%3, %arg0, %0) <{isVolatile = false}> : (!llvm.ptr, !llvm.ptr, i32) -> ()
+    %6 = llvm.ptrtoint %arg0 : !llvm.ptr to i64
+    %7 = llvm.icmp "eq" %6, %1 : i64
+    llvm.cond_br %7, ^bb1, ^bb2
+  ^bb1:  // pred: ^bb0
+    llvm.br ^bb3
+  ^bb2:  // pred: ^bb0
+    llvm.br ^bb3
+  ^bb3:  // 2 preds: ^bb1, ^bb2
+    omp.yield(%arg1 : !llvm.ptr)
+  } dealloc {
+  ^bb0(%arg0: !llvm.ptr):
+    omp.yield
+  }
+  omp.private {type = private} @_QFEc1_private_box_ptr_c8xU : !llvm.struct<(ptr, i64, i32, i8, i8, i8, i8)> init {
+  ^bb0(%arg0: !llvm.ptr, %arg1: !llvm.ptr):
+    %0 = llvm.mlir.constant(24 : i32) : i32
+    %1 = llvm.mlir.constant(1 : i32) : i32
+    %2 = llvm.alloca %1 x !llvm.struct<(ptr, i64, i32, i8, i8, i8, i8)> {alignment = 8 : i64} : (i32) -> !llvm.ptr
+    "llvm.intr.memcpy"(%2, %arg0, %0) <{isVolatile = false}> : (!llvm.ptr, !llvm.ptr, i32) -> ()
+    omp.yield(%arg1 : !llvm.ptr)
+  }
+  llvm.func @_QQmain() {
+    %0 = llvm.mlir.constant(1 : i64) : i64
+    %1 = llvm.alloca %0 x !llvm.struct<(ptr, i64, i32, i8, i8, i8, i8)> {bindc_name = "c2"} : (i64) -> !llvm.ptr
+    %2 = llvm.alloca %0 x !llvm.struct<(ptr, i64, i32, i8, i8, i8, i8)> {bindc_name = "c1"} : (i64) -> !llvm.ptr
+    omp.task private(@_QFEc1_private_box_ptr_c8xU %2 -> %arg0, @_QFEc2_private_box_heap_c8xU %1 -> %arg1 : !llvm.ptr, !llvm.ptr) {
+      omp.terminator
+    }
+    llvm.return
+  }
+}
+
+// CHECK-LABEL: @_QQmain() {
+// CHECK:         %[[STRUCTARG:.*]] = alloca { ptr }, align 8
+// CHECK:         %[[VAL_0:.*]] = alloca { ptr, i64, i32, i8, i8, i8, i8 }, i64 1, align 8
+// CHECK:         br label %[[VAL_2:.*]]
+// CHECK:       entry:                                            ; preds = %[[VAL_3:.*]]
+// CHECK:         br label %[[VAL_4:.*]]
+// CHECK:       omp.private.init:                                 ; preds = %[[VAL_2]]
+// CHECK:         %[[VAL_5:.*]] = tail call ptr @malloc(i64 ptrtoint (ptr getelementptr ({ { ptr, i64, i32, i8, i8, i8, i8 }, { ptr, i64, i32, i8, i8, i8, i8 } }, ptr null, i32 1) to i64))
+// CHECK:         %[[VAL_6:.*]] = getelementptr { { ptr, i64, i32, i8, i8, i8, i8 }, { ptr, i64, i32, i8, i8, i8, i8 } }, ptr %[[VAL_5]], i32 0, i32 0
+// CHECK:         %[[VAL_7:.*]] = getelementptr { { ptr, i64, i32, i8, i8, i8, i8 }, { ptr, i64, i32, i8, i8, i8, i8 } }, ptr %[[VAL_5]], i32 0, i32 1
+// ...
+// CHECK:         br label %[[VAL_9:.*]]
+// CHECK:       omp.private.init4:                                ; preds = %[[VAL_10:.*]], %[[VAL_11:.*]]
+// CHECK:         br label %[[VAL_12:.*]]
+// CHECK:       omp.private.init3:                                ; preds = %[[VAL_9]]
+// CHECK:         br label %[[VAL_13:.*]]
+// CHECK:       omp.private.init2:                                ; preds = %[[VAL_9]]
+// CHECK:         br label %[[VAL_13]]
+// CHECK:       omp.private.init1:                                ; preds = %[[VAL_4]]
+// CHECK:         %[[VAL_14:.*]] = alloca { ptr, i64, i32, i8, i8, i8, i8 }, align 8
+// CHECK:         call void @llvm.memcpy.p0.p0.i32(ptr %[[VAL_14]], ptr %[[VAL_0]], i32 24, i1 false)
+// CHECK:         %[[VAL_15:.*]] = ptrtoint ptr %[[VAL_0]] to i64
+// CHECK:         %[[VAL_16:.*]] = icmp eq i64 %[[VAL_15]], 0
+// CHECK:         br i1 %[[VAL_16]], label %[[VAL_10]], label %[[VAL_11]]
+// CHECK:       omp.region.cont:                                  ; preds = %[[VAL_13]]
+// CHECK:         %[[VAL_17:.*]] = phi ptr [ %[[VAL_7]], %[[VAL_13]] ]
+// CHECK:         br label %[[VAL_18:.*]]
+// CHECK:       omp.private.copy:                                 ; preds = %[[VAL_12]]
+// CHECK:         br label %[[VAL_19:.*]]
+// CHECK:       omp.task.start:                                   ; preds = %[[VAL_18]]
+// CHECK:         br label %[[VAL_20:.*]]
+// CHECK:       codeRepl:                                         ; preds = %[[VAL_19]]
+// CHECK:         %[[VAL_21:.*]] = getelementptr { ptr }, ptr %[[STRUCTARG]], i32 0, i32 0
+// CHECK:         store ptr %[[VAL_5]], ptr %[[VAL_21]], align 8
+// CHECK:         %[[VAL_22:.*]] = call i32 @__kmpc_global_thread_num(ptr @1)
+// CHECK:         %[[VAL_23:.*]] = call ptr @__kmpc_omp_task_alloc(ptr @1, i32 %[[VAL_22]], i32 1, i64 40, i64 8, ptr @_QQmain..omp_par)
+// CHECK:         %[[VAL_24:.*]] = load ptr, ptr %[[VAL_23]], align 8
+// CHECK:         call void @llvm.memcpy.p0.p0.i64(ptr align 1 %[[VAL_24]], ptr align 1 %[[STRUCTARG]], i64 8, i1 false)
+// CHECK:         %[[VAL_25:.*]] = call i32 @__kmpc_omp_task(ptr @1, i32 %[[VAL_22]], ptr %[[VAL_23]])

@tblah tblah mentioned this pull request Jun 6, 2025
Open
91 tasks
@tblah
Copy link
Contributor Author

tblah commented Jun 11, 2025

Ping for review

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@ergawy ergawy Awaiting requested review from ergawy

@Meinersbur Meinersbur Awaiting requested review from Meinersbur

@skatrak skatrak Awaiting requested review from skatrak

@bhandarkar-pranav bhandarkar-pranav Awaiting requested review from bhandarkar-pranav

Assignees
No one assigned
Labels
flang:openmp mlir:llvm mlir:openmp mlir
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.

[Flang][OpenMP] crashes when using task private with character pointer/allocatable
2 participants
@tblah @llvmbot
Morty Proxy This is a proxified and sanitized view of the page, visit original site.