[llvm] 87ec6f4 - [OpenMPIRBuilder] Allocate temporary at the correct block in a nested parallel
William S. Moses via llvm-commits
llvm-commits at lists.llvm.org
Sun Mar 6 15:34:30 PST 2022
Author: William S. Moses
Date: 2022-03-06T18:34:25-05:00
New Revision: 87ec6f41bba6d72a3408e71cf19ae56feff523bc
URL: https://github.com/llvm/llvm-project/commit/87ec6f41bba6d72a3408e71cf19ae56feff523bc
DIFF: https://github.com/llvm/llvm-project/commit/87ec6f41bba6d72a3408e71cf19ae56feff523bc.diff
LOG: [OpenMPIRBuilder] Allocate temporary at the correct block in a nested parallel
The OpenMPIRBuilder has a bug. Specifically, suppose you have two nested openmp parallel regions (writing with MLIR for ease)
```
omp.parallel {
%a = ...
omp.parallel {
use(%a)
}
}
```
As OpenMP only permits pointer-like inputs, the builder will wrap all of the inputs into a stack allocation, and then pass this
allocation to the inner parallel. For example, we would want to get something like the following:
```
omp.parallel {
%a = ...
%tmp = alloc
store %tmp[] = %a
kmpc_fork(outlined, %tmp)
}
```
However, in practice, this is not what currently occurs in the context of nested parallel regions. Specifically to the OpenMPIRBuilder,
the entirety of the function (at the LLVM level) is currently inlined with blocks marking the corresponding start and end of each
region.
```
entry:
...
parallel1:
%a = ...
...
parallel2:
use(%a)
...
endparallel2:
...
endparallel1:
...
```
When the allocation is inserted, it presently inserted into the parent of the entire function (e.g. entry) rather than the parent
allocation scope to the function being outlined. If we were outlining parallel2, the corresponding alloca location would be parallel1.
This causes a variety of bugs, including https://github.com/llvm/llvm-project/issues/54165 as one example.
This PR allows the stack allocation to be created at the correct allocation block, and thus remedies such issues.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D121061
Added:
mlir/test/Target/LLVMIR/openmp-nested.mlir
Modified:
clang/test/OpenMP/irbuilder_nested_openmp_parallel_empty.c
clang/test/OpenMP/irbuilder_nested_parallel_for.c
llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h
llvm/include/llvm/Transforms/Utils/CodeExtractor.h
llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp
llvm/lib/Transforms/IPO/HotColdSplitting.cpp
llvm/lib/Transforms/IPO/IROutliner.cpp
llvm/lib/Transforms/Utils/CodeExtractor.cpp
Removed:
################################################################################
diff --git a/clang/test/OpenMP/irbuilder_nested_openmp_parallel_empty.c b/clang/test/OpenMP/irbuilder_nested_openmp_parallel_empty.c
index 56aa41b0db466..2c1139d7ef7df 100644
--- a/clang/test/OpenMP/irbuilder_nested_openmp_parallel_empty.c
+++ b/clang/test/OpenMP/irbuilder_nested_openmp_parallel_empty.c
@@ -33,8 +33,7 @@ void nested_parallel_0(void) {
// ALL-LABEL: @_Z17nested_parallel_1Pfid(
// ALL-NEXT: entry:
-// ALL-NEXT: [[STRUCTARG14:%.*]] = alloca { { i32*, double*, float** }*, i32*, double*, float** }, align 8
-// ALL-NEXT: [[STRUCTARG:%.*]] = alloca { i32*, double*, float** }, align 8
+// ALL-NEXT: [[STRUCTARG14:%.*]] = alloca { i32*, double*, float** }, align 8
// ALL-NEXT: [[R_ADDR:%.*]] = alloca float*, align 8
// ALL-NEXT: [[A_ADDR:%.*]] = alloca i32, align 4
// ALL-NEXT: [[B_ADDR:%.*]] = alloca double, align 8
@@ -44,15 +43,13 @@ void nested_parallel_0(void) {
// ALL-NEXT: [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// ALL-NEXT: br label [[OMP_PARALLEL:%.*]]
// ALL: omp_parallel:
-// ALL-NEXT: [[GEP_STRUCTARG:%.*]] = getelementptr { { i32*, double*, float** }*, i32*, double*, float** }, { { i32*, double*, float** }*, i32*, double*, float** }* [[STRUCTARG14]], i32 0, i32 0
-// ALL-NEXT: store { i32*, double*, float** }* [[STRUCTARG]], { i32*, double*, float** }** [[GEP_STRUCTARG]], align 8
-// ALL-NEXT: [[GEP_A_ADDR15:%.*]] = getelementptr { { i32*, double*, float** }*, i32*, double*, float** }, { { i32*, double*, float** }*, i32*, double*, float** }* [[STRUCTARG14]], i32 0, i32 1
+// ALL-NEXT: [[GEP_A_ADDR15:%.*]] = getelementptr { i32*, double*, float** }, { i32*, double*, float** }* [[STRUCTARG14]], i32 0, i32 0
// ALL-NEXT: store i32* [[A_ADDR]], i32** [[GEP_A_ADDR15]], align 8
-// ALL-NEXT: [[GEP_B_ADDR16:%.*]] = getelementptr { { i32*, double*, float** }*, i32*, double*, float** }, { { i32*, double*, float** }*, i32*, double*, float** }* [[STRUCTARG14]], i32 0, i32 2
+// ALL-NEXT: [[GEP_B_ADDR16:%.*]] = getelementptr { i32*, double*, float** }, { i32*, double*, float** }* [[STRUCTARG14]], i32 0, i32 1
// ALL-NEXT: store double* [[B_ADDR]], double** [[GEP_B_ADDR16]], align 8
-// ALL-NEXT: [[GEP_R_ADDR17:%.*]] = getelementptr { { i32*, double*, float** }*, i32*, double*, float** }, { { i32*, double*, float** }*, i32*, double*, float** }* [[STRUCTARG14]], i32 0, i32 3
+// ALL-NEXT: [[GEP_R_ADDR17:%.*]] = getelementptr { i32*, double*, float** }, { i32*, double*, float** }* [[STRUCTARG14]], i32 0, i32 2
// ALL-NEXT: store float** [[R_ADDR]], float*** [[GEP_R_ADDR17]], align 8
-// ALL-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, { { i32*, double*, float** }*, i32*, double*, float** }*)* @_Z17nested_parallel_1Pfid..omp_par.2 to void (i32*, i32*, ...)*), { { i32*, double*, float** }*, i32*, double*, float** }* [[STRUCTARG14]])
+// ALL-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, { i32*, double*, float** }*)* @_Z17nested_parallel_1Pfid..omp_par.2 to void (i32*, i32*, ...)*), { i32*, double*, float** }* [[STRUCTARG14]])
// ALL-NEXT: br label [[OMP_PAR_OUTLINED_EXIT13:%.*]]
// ALL: omp.par.outlined.exit13:
// ALL-NEXT: br label [[OMP_PAR_EXIT_SPLIT:%.*]]
@@ -71,9 +68,6 @@ void nested_parallel_1(float *r, int a, double b) {
// ALL-LABEL: @_Z17nested_parallel_2Pfid(
// ALL-NEXT: entry:
-// ALL-NEXT: [[STRUCTARG68:%.*]] = alloca { i32*, double*, float**, { i32*, double*, float**, { i32*, double*, float** }*, { i32*, double*, float** }* }*, { i32*, double*, float** }*, { i32*, double*, float** }* }, align 8
-// ALL-NEXT: [[STRUCTARG64:%.*]] = alloca { i32*, double*, float**, { i32*, double*, float** }*, { i32*, double*, float** }* }, align 8
-// ALL-NEXT: [[STRUCTARG59:%.*]] = alloca { i32*, double*, float** }, align 8
// ALL-NEXT: [[STRUCTARG:%.*]] = alloca { i32*, double*, float** }, align 8
// ALL-NEXT: [[R_ADDR:%.*]] = alloca float*, align 8
// ALL-NEXT: [[A_ADDR:%.*]] = alloca i32, align 4
@@ -84,19 +78,13 @@ void nested_parallel_1(float *r, int a, double b) {
// ALL-NEXT: [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// ALL-NEXT: br label [[OMP_PARALLEL:%.*]]
// ALL: omp_parallel:
-// ALL-NEXT: [[GEP_A_ADDR:%.*]] = getelementptr { i32*, double*, float**, { i32*, double*, float**, { i32*, double*, float** }*, { i32*, double*, float** }* }*, { i32*, double*, float** }*, { i32*, double*, float** }* }, { i32*, double*, float**, { i32*, double*, float**, { i32*, double*, float** }*, { i32*, double*, float** }* }*, { i32*, double*, float** }*, { i32*, double*, float** }* }* [[STRUCTARG68]], i32 0, i32 0
+// ALL-NEXT: [[GEP_A_ADDR:%.*]] = getelementptr { i32*, double*, float** }, { i32*, double*, float** }* [[STRUCTARG]], i32 0, i32 0
// ALL-NEXT: store i32* [[A_ADDR]], i32** [[GEP_A_ADDR]], align 8
-// ALL-NEXT: [[GEP_B_ADDR:%.*]] = getelementptr { i32*, double*, float**, { i32*, double*, float**, { i32*, double*, float** }*, { i32*, double*, float** }* }*, { i32*, double*, float** }*, { i32*, double*, float** }* }, { i32*, double*, float**, { i32*, double*, float**, { i32*, double*, float** }*, { i32*, double*, float** }* }*, { i32*, double*, float** }*, { i32*, double*, float** }* }* [[STRUCTARG68]], i32 0, i32 1
+// ALL-NEXT: [[GEP_B_ADDR:%.*]] = getelementptr { i32*, double*, float** }, { i32*, double*, float** }* [[STRUCTARG]], i32 0, i32 1
// ALL-NEXT: store double* [[B_ADDR]], double** [[GEP_B_ADDR]], align 8
-// ALL-NEXT: [[GEP_R_ADDR:%.*]] = getelementptr { i32*, double*, float**, { i32*, double*, float**, { i32*, double*, float** }*, { i32*, double*, float** }* }*, { i32*, double*, float** }*, { i32*, double*, float** }* }, { i32*, double*, float**, { i32*, double*, float**, { i32*, double*, float** }*, { i32*, double*, float** }* }*, { i32*, double*, float** }*, { i32*, double*, float** }* }* [[STRUCTARG68]], i32 0, i32 2
+// ALL-NEXT: [[GEP_R_ADDR:%.*]] = getelementptr { i32*, double*, float** }, { i32*, double*, float** }* [[STRUCTARG]], i32 0, i32 2
// ALL-NEXT: store float** [[R_ADDR]], float*** [[GEP_R_ADDR]], align 8
-// ALL-NEXT: [[GEP_STRUCTARG64:%.*]] = getelementptr { i32*, double*, float**, { i32*, double*, float**, { i32*, double*, float** }*, { i32*, double*, float** }* }*, { i32*, double*, float** }*, { i32*, double*, float** }* }, { i32*, double*, float**, { i32*, double*, float**, { i32*, double*, float** }*, { i32*, double*, float** }* }*, { i32*, double*, float** }*, { i32*, double*, float** }* }* [[STRUCTARG68]], i32 0, i32 3
-// ALL-NEXT: store { i32*, double*, float**, { i32*, double*, float** }*, { i32*, double*, float** }* }* [[STRUCTARG64]], { i32*, double*, float**, { i32*, double*, float** }*, { i32*, double*, float** }* }** [[GEP_STRUCTARG64]], align 8
-// ALL-NEXT: [[GEP_STRUCTARG69:%.*]] = getelementptr { i32*, double*, float**, { i32*, double*, float**, { i32*, double*, float** }*, { i32*, double*, float** }* }*, { i32*, double*, float** }*, { i32*, double*, float** }* }, { i32*, double*, float**, { i32*, double*, float**, { i32*, double*, float** }*, { i32*, double*, float** }* }*, { i32*, double*, float** }*, { i32*, double*, float** }* }* [[STRUCTARG68]], i32 0, i32 4
-// ALL-NEXT: store { i32*, double*, float** }* [[STRUCTARG]], { i32*, double*, float** }** [[GEP_STRUCTARG69]], align 8
-// ALL-NEXT: [[GEP_STRUCTARG5970:%.*]] = getelementptr { i32*, double*, float**, { i32*, double*, float**, { i32*, double*, float** }*, { i32*, double*, float** }* }*, { i32*, double*, float** }*, { i32*, double*, float** }* }, { i32*, double*, float**, { i32*, double*, float**, { i32*, double*, float** }*, { i32*, double*, float** }* }*, { i32*, double*, float** }*, { i32*, double*, float** }* }* [[STRUCTARG68]], i32 0, i32 5
-// ALL-NEXT: store { i32*, double*, float** }* [[STRUCTARG59]], { i32*, double*, float** }** [[GEP_STRUCTARG5970]], align 8
-// ALL-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, { i32*, double*, float**, { i32*, double*, float**, { i32*, double*, float** }*, { i32*, double*, float** }* }*, { i32*, double*, float** }*, { i32*, double*, float** }* }*)* @_Z17nested_parallel_2Pfid..omp_par.5 to void (i32*, i32*, ...)*), { i32*, double*, float**, { i32*, double*, float**, { i32*, double*, float** }*, { i32*, double*, float** }* }*, { i32*, double*, float** }*, { i32*, double*, float** }* }* [[STRUCTARG68]])
+// ALL-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, { i32*, double*, float** }*)* @_Z17nested_parallel_2Pfid..omp_par.5 to void (i32*, i32*, ...)*), { i32*, double*, float** }* [[STRUCTARG]])
// ALL-NEXT: br label [[OMP_PAR_OUTLINED_EXIT55:%.*]]
// ALL: omp.par.outlined.exit55:
// ALL-NEXT: br label [[OMP_PAR_EXIT_SPLIT:%.*]]
diff --git a/clang/test/OpenMP/irbuilder_nested_parallel_for.c b/clang/test/OpenMP/irbuilder_nested_parallel_for.c
index f4e24343eb52c..33712765e295d 100644
--- a/clang/test/OpenMP/irbuilder_nested_parallel_for.c
+++ b/clang/test/OpenMP/irbuilder_nested_parallel_for.c
@@ -44,8 +44,7 @@ void parallel_for_0(void) {
// CHECK-LABEL: @_Z14parallel_for_1Pfid(
// CHECK-NEXT: entry:
-// CHECK-NEXT: [[STRUCTARG17:%.*]] = alloca { { i32*, double*, float** }*, i32*, double*, float** }, align 8
-// CHECK-NEXT: [[STRUCTARG:%.*]] = alloca { i32*, double*, float** }, align 8
+// CHECK-NEXT: [[STRUCTARG17:%.*]] = alloca { i32*, double*, float** }, align 8
// CHECK-NEXT: [[R_ADDR:%.*]] = alloca float*, align 8
// CHECK-NEXT: [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK-NEXT: [[B_ADDR:%.*]] = alloca double, align 8
@@ -55,15 +54,13 @@ void parallel_for_0(void) {
// CHECK-NEXT: [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK-NEXT: br label [[OMP_PARALLEL:%.*]]
// CHECK: omp_parallel:
-// CHECK-NEXT: [[GEP_STRUCTARG:%.*]] = getelementptr { { i32*, double*, float** }*, i32*, double*, float** }, { { i32*, double*, float** }*, i32*, double*, float** }* [[STRUCTARG17]], i32 0, i32 0
-// CHECK-NEXT: store { i32*, double*, float** }* [[STRUCTARG]], { i32*, double*, float** }** [[GEP_STRUCTARG]], align 8
-// CHECK-NEXT: [[GEP_A_ADDR18:%.*]] = getelementptr { { i32*, double*, float** }*, i32*, double*, float** }, { { i32*, double*, float** }*, i32*, double*, float** }* [[STRUCTARG17]], i32 0, i32 1
+// CHECK-NEXT: [[GEP_A_ADDR18:%.*]] = getelementptr { i32*, double*, float** }, { i32*, double*, float** }* [[STRUCTARG17]], i32 0, i32 0
// CHECK-NEXT: store i32* [[A_ADDR]], i32** [[GEP_A_ADDR18]], align 8
-// CHECK-NEXT: [[GEP_B_ADDR19:%.*]] = getelementptr { { i32*, double*, float** }*, i32*, double*, float** }, { { i32*, double*, float** }*, i32*, double*, float** }* [[STRUCTARG17]], i32 0, i32 2
+// CHECK-NEXT: [[GEP_B_ADDR19:%.*]] = getelementptr { i32*, double*, float** }, { i32*, double*, float** }* [[STRUCTARG17]], i32 0, i32 1
// CHECK-NEXT: store double* [[B_ADDR]], double** [[GEP_B_ADDR19]], align 8
-// CHECK-NEXT: [[GEP_R_ADDR20:%.*]] = getelementptr { { i32*, double*, float** }*, i32*, double*, float** }, { { i32*, double*, float** }*, i32*, double*, float** }* [[STRUCTARG17]], i32 0, i32 3
+// CHECK-NEXT: [[GEP_R_ADDR20:%.*]] = getelementptr { i32*, double*, float** }, { i32*, double*, float** }* [[STRUCTARG17]], i32 0, i32 2
// CHECK-NEXT: store float** [[R_ADDR]], float*** [[GEP_R_ADDR20]], align 8
-// CHECK-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, { { i32*, double*, float** }*, i32*, double*, float** }*)* @_Z14parallel_for_1Pfid..omp_par.4 to void (i32*, i32*, ...)*), { { i32*, double*, float** }*, i32*, double*, float** }* [[STRUCTARG17]])
+// CHECK-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, { i32*, double*, float** }*)* @_Z14parallel_for_1Pfid..omp_par.4 to void (i32*, i32*, ...)*), { i32*, double*, float** }* [[STRUCTARG17]])
// CHECK-NEXT: br label [[OMP_PAR_OUTLINED_EXIT16:%.*]]
// CHECK: omp.par.outlined.exit16:
// CHECK-NEXT: br label [[OMP_PAR_EXIT_SPLIT:%.*]]
@@ -72,34 +69,31 @@ void parallel_for_0(void) {
//
// CHECK-DEBUG-LABEL: @_Z14parallel_for_1Pfid(
// CHECK-DEBUG-NEXT: entry:
-// CHECK-DEBUG-NEXT: [[STRUCTARG17:%.*]] = alloca { { i32*, double*, float** }*, i32*, double*, float** }, align 8
-// CHECK-DEBUG-NEXT: [[STRUCTARG:%.*]] = alloca { i32*, double*, float** }, align 8
+// CHECK-DEBUG-NEXT: [[STRUCTARG17:%.*]] = alloca { i32*, double*, float** }, align 8
// CHECK-DEBUG-NEXT: [[R_ADDR:%.*]] = alloca float*, align 8
// CHECK-DEBUG-NEXT: [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK-DEBUG-NEXT: [[B_ADDR:%.*]] = alloca double, align 8
// CHECK-DEBUG-NEXT: store float* [[R:%.*]], float** [[R_ADDR]], align 8
-// CHECK-DEBUG-NEXT: call void @llvm.dbg.declare(metadata float** [[R_ADDR]], metadata [[META73:![0-9]+]], metadata !DIExpression()), !dbg [[DBG74:![0-9]+]]
+// CHECK-DEBUG-NEXT: call void @llvm.dbg.declare(metadata float** [[R_ADDR]], metadata [[META72:![0-9]+]], metadata !DIExpression()), !dbg [[DBG73:![0-9]+]]
// CHECK-DEBUG-NEXT: store i32 [[A:%.*]], i32* [[A_ADDR]], align 4
-// CHECK-DEBUG-NEXT: call void @llvm.dbg.declare(metadata i32* [[A_ADDR]], metadata [[META75:![0-9]+]], metadata !DIExpression()), !dbg [[DBG76:![0-9]+]]
+// CHECK-DEBUG-NEXT: call void @llvm.dbg.declare(metadata i32* [[A_ADDR]], metadata [[META74:![0-9]+]], metadata !DIExpression()), !dbg [[DBG75:![0-9]+]]
// CHECK-DEBUG-NEXT: store double [[B:%.*]], double* [[B_ADDR]], align 8
-// CHECK-DEBUG-NEXT: call void @llvm.dbg.declare(metadata double* [[B_ADDR]], metadata [[META77:![0-9]+]], metadata !DIExpression()), !dbg [[DBG78:![0-9]+]]
-// CHECK-DEBUG-NEXT: [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6:[0-9]+]]), !dbg [[DBG79:![0-9]+]]
+// CHECK-DEBUG-NEXT: call void @llvm.dbg.declare(metadata double* [[B_ADDR]], metadata [[META76:![0-9]+]], metadata !DIExpression()), !dbg [[DBG77:![0-9]+]]
+// CHECK-DEBUG-NEXT: [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6:[0-9]+]]), !dbg [[DBG78:![0-9]+]]
// CHECK-DEBUG-NEXT: br label [[OMP_PARALLEL:%.*]]
// CHECK-DEBUG: omp_parallel:
-// CHECK-DEBUG-NEXT: [[GEP_STRUCTARG:%.*]] = getelementptr { { i32*, double*, float** }*, i32*, double*, float** }, { { i32*, double*, float** }*, i32*, double*, float** }* [[STRUCTARG17]], i32 0, i32 0
-// CHECK-DEBUG-NEXT: store { i32*, double*, float** }* [[STRUCTARG]], { i32*, double*, float** }** [[GEP_STRUCTARG]], align 8
-// CHECK-DEBUG-NEXT: [[GEP_A_ADDR18:%.*]] = getelementptr { { i32*, double*, float** }*, i32*, double*, float** }, { { i32*, double*, float** }*, i32*, double*, float** }* [[STRUCTARG17]], i32 0, i32 1
+// CHECK-DEBUG-NEXT: [[GEP_A_ADDR18:%.*]] = getelementptr { i32*, double*, float** }, { i32*, double*, float** }* [[STRUCTARG17]], i32 0, i32 0
// CHECK-DEBUG-NEXT: store i32* [[A_ADDR]], i32** [[GEP_A_ADDR18]], align 8
-// CHECK-DEBUG-NEXT: [[GEP_B_ADDR19:%.*]] = getelementptr { { i32*, double*, float** }*, i32*, double*, float** }, { { i32*, double*, float** }*, i32*, double*, float** }* [[STRUCTARG17]], i32 0, i32 2
+// CHECK-DEBUG-NEXT: [[GEP_B_ADDR19:%.*]] = getelementptr { i32*, double*, float** }, { i32*, double*, float** }* [[STRUCTARG17]], i32 0, i32 1
// CHECK-DEBUG-NEXT: store double* [[B_ADDR]], double** [[GEP_B_ADDR19]], align 8
-// CHECK-DEBUG-NEXT: [[GEP_R_ADDR20:%.*]] = getelementptr { { i32*, double*, float** }*, i32*, double*, float** }, { { i32*, double*, float** }*, i32*, double*, float** }* [[STRUCTARG17]], i32 0, i32 3
+// CHECK-DEBUG-NEXT: [[GEP_R_ADDR20:%.*]] = getelementptr { i32*, double*, float** }, { i32*, double*, float** }* [[STRUCTARG17]], i32 0, i32 2
// CHECK-DEBUG-NEXT: store float** [[R_ADDR]], float*** [[GEP_R_ADDR20]], align 8
-// CHECK-DEBUG-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB6]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, { { i32*, double*, float** }*, i32*, double*, float** }*)* @_Z14parallel_for_1Pfid..omp_par.4 to void (i32*, i32*, ...)*), { { i32*, double*, float** }*, i32*, double*, float** }* [[STRUCTARG17]]), !dbg [[DBG80:![0-9]+]]
+// CHECK-DEBUG-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB6]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, { i32*, double*, float** }*)* @_Z14parallel_for_1Pfid..omp_par.4 to void (i32*, i32*, ...)*), { i32*, double*, float** }* [[STRUCTARG17]]), !dbg [[DBG79:![0-9]+]]
// CHECK-DEBUG-NEXT: br label [[OMP_PAR_OUTLINED_EXIT16:%.*]]
// CHECK-DEBUG: omp.par.outlined.exit16:
// CHECK-DEBUG-NEXT: br label [[OMP_PAR_EXIT_SPLIT:%.*]]
// CHECK-DEBUG: omp.par.exit.split:
-// CHECK-DEBUG-NEXT: ret void, !dbg [[DBG82:![0-9]+]]
+// CHECK-DEBUG-NEXT: ret void, !dbg [[DBG81:![0-9]+]]
//
void parallel_for_1(float *r, int a, double b) {
#pragma omp parallel
@@ -116,9 +110,6 @@ void parallel_for_1(float *r, int a, double b) {
// CHECK-LABEL: @_Z14parallel_for_2Pfid(
// CHECK-NEXT: entry:
-// CHECK-NEXT: [[STRUCTARG218:%.*]] = alloca { { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }*, { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }, align 8
-// CHECK-NEXT: [[STRUCTARG214:%.*]] = alloca { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }, align 8
-// CHECK-NEXT: [[STRUCTARG209:%.*]] = alloca { i32*, double*, float** }, align 8
// CHECK-NEXT: [[STRUCTARG:%.*]] = alloca { i32*, double*, float** }, align 8
// CHECK-NEXT: [[R_ADDR:%.*]] = alloca float*, align 8
// CHECK-NEXT: [[A_ADDR:%.*]] = alloca i32, align 4
@@ -137,19 +128,13 @@ void parallel_for_1(float *r, int a, double b) {
// CHECK-NEXT: [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK-NEXT: br label [[OMP_PARALLEL:%.*]]
// CHECK: omp_parallel:
-// CHECK-NEXT: [[GEP_STRUCTARG214:%.*]] = getelementptr { { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }*, { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }, { { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }*, { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }* [[STRUCTARG218]], i32 0, i32 0
-// CHECK-NEXT: store { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }* [[STRUCTARG214]], { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }** [[GEP_STRUCTARG214]], align 8
-// CHECK-NEXT: [[GEP_STRUCTARG219:%.*]] = getelementptr { { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }*, { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }, { { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }*, { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }* [[STRUCTARG218]], i32 0, i32 1
-// CHECK-NEXT: store { i32*, double*, float** }* [[STRUCTARG]], { i32*, double*, float** }** [[GEP_STRUCTARG219]], align 8
-// CHECK-NEXT: [[GEP_A_ADDR:%.*]] = getelementptr { { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }*, { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }, { { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }*, { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }* [[STRUCTARG218]], i32 0, i32 2
+// CHECK-NEXT: [[GEP_A_ADDR:%.*]] = getelementptr { i32*, double*, float** }, { i32*, double*, float** }* [[STRUCTARG]], i32 0, i32 0
// CHECK-NEXT: store i32* [[A_ADDR]], i32** [[GEP_A_ADDR]], align 8
-// CHECK-NEXT: [[GEP_B_ADDR:%.*]] = getelementptr { { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }*, { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }, { { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }*, { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }* [[STRUCTARG218]], i32 0, i32 3
+// CHECK-NEXT: [[GEP_B_ADDR:%.*]] = getelementptr { i32*, double*, float** }, { i32*, double*, float** }* [[STRUCTARG]], i32 0, i32 1
// CHECK-NEXT: store double* [[B_ADDR]], double** [[GEP_B_ADDR]], align 8
-// CHECK-NEXT: [[GEP_R_ADDR:%.*]] = getelementptr { { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }*, { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }, { { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }*, { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }* [[STRUCTARG218]], i32 0, i32 4
+// CHECK-NEXT: [[GEP_R_ADDR:%.*]] = getelementptr { i32*, double*, float** }, { i32*, double*, float** }* [[STRUCTARG]], i32 0, i32 2
// CHECK-NEXT: store float** [[R_ADDR]], float*** [[GEP_R_ADDR]], align 8
-// CHECK-NEXT: [[GEP_STRUCTARG209220:%.*]] = getelementptr { { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }*, { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }, { { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }*, { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }* [[STRUCTARG218]], i32 0, i32 5
-// CHECK-NEXT: store { i32*, double*, float** }* [[STRUCTARG209]], { i32*, double*, float** }** [[GEP_STRUCTARG209220]], align 8
-// CHECK-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, { { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }*, { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }*)* @_Z14parallel_for_2Pfid..omp_par.23 to void (i32*, i32*, ...)*), { { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }*, { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }* [[STRUCTARG218]])
+// CHECK-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, { i32*, double*, float** }*)* @_Z14parallel_for_2Pfid..omp_par.23 to void (i32*, i32*, ...)*), { i32*, double*, float** }* [[STRUCTARG]])
// CHECK-NEXT: br label [[OMP_PAR_OUTLINED_EXIT184:%.*]]
// CHECK: omp.par.outlined.exit184:
// CHECK-NEXT: br label [[OMP_PAR_EXIT_SPLIT:%.*]]
@@ -205,9 +190,6 @@ void parallel_for_1(float *r, int a, double b) {
//
// CHECK-DEBUG-LABEL: @_Z14parallel_for_2Pfid(
// CHECK-DEBUG-NEXT: entry:
-// CHECK-DEBUG-NEXT: [[STRUCTARG218:%.*]] = alloca { { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }*, { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }, align 8
-// CHECK-DEBUG-NEXT: [[STRUCTARG214:%.*]] = alloca { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }, align 8
-// CHECK-DEBUG-NEXT: [[STRUCTARG209:%.*]] = alloca { i32*, double*, float** }, align 8
// CHECK-DEBUG-NEXT: [[STRUCTARG:%.*]] = alloca { i32*, double*, float** }, align 8
// CHECK-DEBUG-NEXT: [[R_ADDR:%.*]] = alloca float*, align 8
// CHECK-DEBUG-NEXT: [[A_ADDR:%.*]] = alloca i32, align 4
@@ -221,80 +203,74 @@ void parallel_for_1(float *r, int a, double b) {
// CHECK-DEBUG-NEXT: [[P_UPPERBOUND205:%.*]] = alloca i32, align 4
// CHECK-DEBUG-NEXT: [[P_STRIDE206:%.*]] = alloca i32, align 4
// CHECK-DEBUG-NEXT: store float* [[R:%.*]], float** [[R_ADDR]], align 8
-// CHECK-DEBUG-NEXT: call void @llvm.dbg.declare(metadata float** [[R_ADDR]], metadata [[META134:![0-9]+]], metadata !DIExpression()), !dbg [[DBG135:![0-9]+]]
+// CHECK-DEBUG-NEXT: call void @llvm.dbg.declare(metadata float** [[R_ADDR]], metadata [[META133:![0-9]+]], metadata !DIExpression()), !dbg [[DBG134:![0-9]+]]
// CHECK-DEBUG-NEXT: store i32 [[A:%.*]], i32* [[A_ADDR]], align 4
-// CHECK-DEBUG-NEXT: call void @llvm.dbg.declare(metadata i32* [[A_ADDR]], metadata [[META136:![0-9]+]], metadata !DIExpression()), !dbg [[DBG137:![0-9]+]]
+// CHECK-DEBUG-NEXT: call void @llvm.dbg.declare(metadata i32* [[A_ADDR]], metadata [[META135:![0-9]+]], metadata !DIExpression()), !dbg [[DBG136:![0-9]+]]
// CHECK-DEBUG-NEXT: store double [[B:%.*]], double* [[B_ADDR]], align 8
-// CHECK-DEBUG-NEXT: call void @llvm.dbg.declare(metadata double* [[B_ADDR]], metadata [[META138:![0-9]+]], metadata !DIExpression()), !dbg [[DBG139:![0-9]+]]
-// CHECK-DEBUG-NEXT: [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB13:[0-9]+]]), !dbg [[DBG140:![0-9]+]]
+// CHECK-DEBUG-NEXT: call void @llvm.dbg.declare(metadata double* [[B_ADDR]], metadata [[META137:![0-9]+]], metadata !DIExpression()), !dbg [[DBG138:![0-9]+]]
+// CHECK-DEBUG-NEXT: [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB13:[0-9]+]]), !dbg [[DBG139:![0-9]+]]
// CHECK-DEBUG-NEXT: br label [[OMP_PARALLEL:%.*]]
// CHECK-DEBUG: omp_parallel:
-// CHECK-DEBUG-NEXT: [[GEP_STRUCTARG214:%.*]] = getelementptr { { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }*, { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }, { { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }*, { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }* [[STRUCTARG218]], i32 0, i32 0
-// CHECK-DEBUG-NEXT: store { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }* [[STRUCTARG214]], { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }** [[GEP_STRUCTARG214]], align 8
-// CHECK-DEBUG-NEXT: [[GEP_STRUCTARG219:%.*]] = getelementptr { { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }*, { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }, { { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }*, { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }* [[STRUCTARG218]], i32 0, i32 1
-// CHECK-DEBUG-NEXT: store { i32*, double*, float** }* [[STRUCTARG]], { i32*, double*, float** }** [[GEP_STRUCTARG219]], align 8
-// CHECK-DEBUG-NEXT: [[GEP_A_ADDR:%.*]] = getelementptr { { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }*, { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }, { { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }*, { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }* [[STRUCTARG218]], i32 0, i32 2
+// CHECK-DEBUG-NEXT: [[GEP_A_ADDR:%.*]] = getelementptr { i32*, double*, float** }, { i32*, double*, float** }* [[STRUCTARG]], i32 0, i32 0
// CHECK-DEBUG-NEXT: store i32* [[A_ADDR]], i32** [[GEP_A_ADDR]], align 8
-// CHECK-DEBUG-NEXT: [[GEP_B_ADDR:%.*]] = getelementptr { { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }*, { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }, { { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }*, { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }* [[STRUCTARG218]], i32 0, i32 3
+// CHECK-DEBUG-NEXT: [[GEP_B_ADDR:%.*]] = getelementptr { i32*, double*, float** }, { i32*, double*, float** }* [[STRUCTARG]], i32 0, i32 1
// CHECK-DEBUG-NEXT: store double* [[B_ADDR]], double** [[GEP_B_ADDR]], align 8
-// CHECK-DEBUG-NEXT: [[GEP_R_ADDR:%.*]] = getelementptr { { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }*, { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }, { { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }*, { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }* [[STRUCTARG218]], i32 0, i32 4
+// CHECK-DEBUG-NEXT: [[GEP_R_ADDR:%.*]] = getelementptr { i32*, double*, float** }, { i32*, double*, float** }* [[STRUCTARG]], i32 0, i32 2
// CHECK-DEBUG-NEXT: store float** [[R_ADDR]], float*** [[GEP_R_ADDR]], align 8
-// CHECK-DEBUG-NEXT: [[GEP_STRUCTARG209220:%.*]] = getelementptr { { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }*, { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }, { { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }*, { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }* [[STRUCTARG218]], i32 0, i32 5
-// CHECK-DEBUG-NEXT: store { i32*, double*, float** }* [[STRUCTARG209]], { i32*, double*, float** }** [[GEP_STRUCTARG209220]], align 8
-// CHECK-DEBUG-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB13]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, { { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }*, { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }*)* @_Z14parallel_for_2Pfid..omp_par.23 to void (i32*, i32*, ...)*), { { { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }*, { i32*, double*, float** }*, i32*, double*, float**, { i32*, double*, float** }* }* [[STRUCTARG218]]), !dbg [[DBG141:![0-9]+]]
+// CHECK-DEBUG-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB13]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, { i32*, double*, float** }*)* @_Z14parallel_for_2Pfid..omp_par.23 to void (i32*, i32*, ...)*), { i32*, double*, float** }* [[STRUCTARG]]), !dbg [[DBG140:![0-9]+]]
// CHECK-DEBUG-NEXT: br label [[OMP_PAR_OUTLINED_EXIT184:%.*]]
// CHECK-DEBUG: omp.par.outlined.exit184:
// CHECK-DEBUG-NEXT: br label [[OMP_PAR_EXIT_SPLIT:%.*]]
// CHECK-DEBUG: omp.par.exit.split:
-// CHECK-DEBUG-NEXT: call void @llvm.dbg.declare(metadata i32* [[I185]], metadata [[META145:![0-9]+]], metadata !DIExpression()), !dbg [[DBG148:![0-9]+]]
-// CHECK-DEBUG-NEXT: store i32 0, i32* [[I185]], align 4, !dbg [[DBG148]]
-// CHECK-DEBUG-NEXT: [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_17]], %struct.anon.17* [[AGG_CAPTURED186]], i32 0, i32 0, !dbg [[DBG149:![0-9]+]]
-// CHECK-DEBUG-NEXT: store i32* [[I185]], i32** [[TMP0]], align 8, !dbg [[DBG149]]
-// CHECK-DEBUG-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_18]], %struct.anon.18* [[AGG_CAPTURED187]], i32 0, i32 0, !dbg [[DBG149]]
-// CHECK-DEBUG-NEXT: [[TMP2:%.*]] = load i32, i32* [[I185]], align 4, !dbg [[DBG150:![0-9]+]]
-// CHECK-DEBUG-NEXT: store i32 [[TMP2]], i32* [[TMP1]], align 4, !dbg [[DBG149]]
-// CHECK-DEBUG-NEXT: call void @__captured_stmt.19(i32* [[DOTCOUNT_ADDR188]], %struct.anon.17* [[AGG_CAPTURED186]]), !dbg [[DBG149]]
-// CHECK-DEBUG-NEXT: [[DOTCOUNT189:%.*]] = load i32, i32* [[DOTCOUNT_ADDR188]], align 4, !dbg [[DBG149]]
-// CHECK-DEBUG-NEXT: br label [[OMP_LOOP_PREHEADER190:%.*]], !dbg [[DBG149]]
+// CHECK-DEBUG-NEXT: call void @llvm.dbg.declare(metadata i32* [[I185]], metadata [[META144:![0-9]+]], metadata !DIExpression()), !dbg [[DBG147:![0-9]+]]
+// CHECK-DEBUG-NEXT: store i32 0, i32* [[I185]], align 4, !dbg [[DBG147]]
+// CHECK-DEBUG-NEXT: [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_17]], %struct.anon.17* [[AGG_CAPTURED186]], i32 0, i32 0, !dbg [[DBG148:![0-9]+]]
+// CHECK-DEBUG-NEXT: store i32* [[I185]], i32** [[TMP0]], align 8, !dbg [[DBG148]]
+// CHECK-DEBUG-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_18]], %struct.anon.18* [[AGG_CAPTURED187]], i32 0, i32 0, !dbg [[DBG148]]
+// CHECK-DEBUG-NEXT: [[TMP2:%.*]] = load i32, i32* [[I185]], align 4, !dbg [[DBG149:![0-9]+]]
+// CHECK-DEBUG-NEXT: store i32 [[TMP2]], i32* [[TMP1]], align 4, !dbg [[DBG148]]
+// CHECK-DEBUG-NEXT: call void @__captured_stmt.19(i32* [[DOTCOUNT_ADDR188]], %struct.anon.17* [[AGG_CAPTURED186]]), !dbg [[DBG148]]
+// CHECK-DEBUG-NEXT: [[DOTCOUNT189:%.*]] = load i32, i32* [[DOTCOUNT_ADDR188]], align 4, !dbg [[DBG148]]
+// CHECK-DEBUG-NEXT: br label [[OMP_LOOP_PREHEADER190:%.*]], !dbg [[DBG148]]
// CHECK-DEBUG: omp_loop.preheader190:
-// CHECK-DEBUG-NEXT: store i32 0, i32* [[P_LOWERBOUND204]], align 4, !dbg [[DBG149]]
-// CHECK-DEBUG-NEXT: [[TMP3:%.*]] = sub i32 [[DOTCOUNT189]], 1, !dbg [[DBG149]]
-// CHECK-DEBUG-NEXT: store i32 [[TMP3]], i32* [[P_UPPERBOUND205]], align 4, !dbg [[DBG149]]
-// CHECK-DEBUG-NEXT: store i32 1, i32* [[P_STRIDE206]], align 4, !dbg [[DBG149]]
-// CHECK-DEBUG-NEXT: [[OMP_GLOBAL_THREAD_NUM207:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB42:[0-9]+]]), !dbg [[DBG149]]
-// CHECK-DEBUG-NEXT: call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB42]], i32 [[OMP_GLOBAL_THREAD_NUM207]], i32 34, i32* [[P_LASTITER203]], i32* [[P_LOWERBOUND204]], i32* [[P_UPPERBOUND205]], i32* [[P_STRIDE206]], i32 1, i32 0), !dbg [[DBG149]]
-// CHECK-DEBUG-NEXT: [[TMP4:%.*]] = load i32, i32* [[P_LOWERBOUND204]], align 4, !dbg [[DBG149]]
-// CHECK-DEBUG-NEXT: [[TMP5:%.*]] = load i32, i32* [[P_UPPERBOUND205]], align 4, !dbg [[DBG149]]
-// CHECK-DEBUG-NEXT: [[TMP6:%.*]] = sub i32 [[TMP5]], [[TMP4]], !dbg [[DBG149]]
-// CHECK-DEBUG-NEXT: [[TMP7:%.*]] = add i32 [[TMP6]], 1, !dbg [[DBG149]]
-// CHECK-DEBUG-NEXT: br label [[OMP_LOOP_HEADER191:%.*]], !dbg [[DBG149]]
+// CHECK-DEBUG-NEXT: store i32 0, i32* [[P_LOWERBOUND204]], align 4, !dbg [[DBG148]]
+// CHECK-DEBUG-NEXT: [[TMP3:%.*]] = sub i32 [[DOTCOUNT189]], 1, !dbg [[DBG148]]
+// CHECK-DEBUG-NEXT: store i32 [[TMP3]], i32* [[P_UPPERBOUND205]], align 4, !dbg [[DBG148]]
+// CHECK-DEBUG-NEXT: store i32 1, i32* [[P_STRIDE206]], align 4, !dbg [[DBG148]]
+// CHECK-DEBUG-NEXT: [[OMP_GLOBAL_THREAD_NUM207:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB42:[0-9]+]]), !dbg [[DBG148]]
+// CHECK-DEBUG-NEXT: call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB42]], i32 [[OMP_GLOBAL_THREAD_NUM207]], i32 34, i32* [[P_LASTITER203]], i32* [[P_LOWERBOUND204]], i32* [[P_UPPERBOUND205]], i32* [[P_STRIDE206]], i32 1, i32 0), !dbg [[DBG148]]
+// CHECK-DEBUG-NEXT: [[TMP4:%.*]] = load i32, i32* [[P_LOWERBOUND204]], align 4, !dbg [[DBG148]]
+// CHECK-DEBUG-NEXT: [[TMP5:%.*]] = load i32, i32* [[P_UPPERBOUND205]], align 4, !dbg [[DBG148]]
+// CHECK-DEBUG-NEXT: [[TMP6:%.*]] = sub i32 [[TMP5]], [[TMP4]], !dbg [[DBG148]]
+// CHECK-DEBUG-NEXT: [[TMP7:%.*]] = add i32 [[TMP6]], 1, !dbg [[DBG148]]
+// CHECK-DEBUG-NEXT: br label [[OMP_LOOP_HEADER191:%.*]], !dbg [[DBG148]]
// CHECK-DEBUG: omp_loop.header191:
-// CHECK-DEBUG-NEXT: [[OMP_LOOP_IV197:%.*]] = phi i32 [ 0, [[OMP_LOOP_PREHEADER190]] ], [ [[OMP_LOOP_NEXT199:%.*]], [[OMP_LOOP_INC194:%.*]] ], !dbg [[DBG149]]
-// CHECK-DEBUG-NEXT: br label [[OMP_LOOP_COND192:%.*]], !dbg [[DBG149]]
+// CHECK-DEBUG-NEXT: [[OMP_LOOP_IV197:%.*]] = phi i32 [ 0, [[OMP_LOOP_PREHEADER190]] ], [ [[OMP_LOOP_NEXT199:%.*]], [[OMP_LOOP_INC194:%.*]] ], !dbg [[DBG148]]
+// CHECK-DEBUG-NEXT: br label [[OMP_LOOP_COND192:%.*]], !dbg [[DBG148]]
// CHECK-DEBUG: omp_loop.cond192:
-// CHECK-DEBUG-NEXT: [[OMP_LOOP_CMP198:%.*]] = icmp ult i32 [[OMP_LOOP_IV197]], [[TMP7]], !dbg [[DBG149]]
-// CHECK-DEBUG-NEXT: br i1 [[OMP_LOOP_CMP198]], label [[OMP_LOOP_BODY193:%.*]], label [[OMP_LOOP_EXIT195:%.*]], !dbg [[DBG149]]
+// CHECK-DEBUG-NEXT: [[OMP_LOOP_CMP198:%.*]] = icmp ult i32 [[OMP_LOOP_IV197]], [[TMP7]], !dbg [[DBG148]]
+// CHECK-DEBUG-NEXT: br i1 [[OMP_LOOP_CMP198]], label [[OMP_LOOP_BODY193:%.*]], label [[OMP_LOOP_EXIT195:%.*]], !dbg [[DBG148]]
// CHECK-DEBUG: omp_loop.body193:
-// CHECK-DEBUG-NEXT: [[TMP8:%.*]] = add i32 [[OMP_LOOP_IV197]], [[TMP4]], !dbg [[DBG151:![0-9]+]]
-// CHECK-DEBUG-NEXT: call void @__captured_stmt.20(i32* [[I185]], i32 [[TMP8]], %struct.anon.18* [[AGG_CAPTURED187]]), !dbg [[DBG149]]
-// CHECK-DEBUG-NEXT: [[TMP9:%.*]] = load i32, i32* [[A_ADDR]], align 4, !dbg [[DBG152:![0-9]+]]
-// CHECK-DEBUG-NEXT: [[CONV200:%.*]] = sitofp i32 [[TMP9]] to double, !dbg [[DBG152]]
-// CHECK-DEBUG-NEXT: [[TMP10:%.*]] = load double, double* [[B_ADDR]], align 8, !dbg [[DBG151]]
-// CHECK-DEBUG-NEXT: [[ADD201:%.*]] = fadd double [[CONV200]], [[TMP10]], !dbg [[DBG153:![0-9]+]]
-// CHECK-DEBUG-NEXT: [[CONV202:%.*]] = fptrunc double [[ADD201]] to float, !dbg [[DBG152]]
-// CHECK-DEBUG-NEXT: [[TMP11:%.*]] = load float*, float** [[R_ADDR]], align 8, !dbg [[DBG154:![0-9]+]]
-// CHECK-DEBUG-NEXT: store float [[CONV202]], float* [[TMP11]], align 4, !dbg [[DBG155:![0-9]+]]
-// CHECK-DEBUG-NEXT: br label [[OMP_LOOP_INC194]], !dbg [[DBG149]]
+// CHECK-DEBUG-NEXT: [[TMP8:%.*]] = add i32 [[OMP_LOOP_IV197]], [[TMP4]], !dbg [[DBG150:![0-9]+]]
+// CHECK-DEBUG-NEXT: call void @__captured_stmt.20(i32* [[I185]], i32 [[TMP8]], %struct.anon.18* [[AGG_CAPTURED187]]), !dbg [[DBG148]]
+// CHECK-DEBUG-NEXT: [[TMP9:%.*]] = load i32, i32* [[A_ADDR]], align 4, !dbg [[DBG151:![0-9]+]]
+// CHECK-DEBUG-NEXT: [[CONV200:%.*]] = sitofp i32 [[TMP9]] to double, !dbg [[DBG151]]
+// CHECK-DEBUG-NEXT: [[TMP10:%.*]] = load double, double* [[B_ADDR]], align 8, !dbg [[DBG150]]
+// CHECK-DEBUG-NEXT: [[ADD201:%.*]] = fadd double [[CONV200]], [[TMP10]], !dbg [[DBG152:![0-9]+]]
+// CHECK-DEBUG-NEXT: [[CONV202:%.*]] = fptrunc double [[ADD201]] to float, !dbg [[DBG151]]
+// CHECK-DEBUG-NEXT: [[TMP11:%.*]] = load float*, float** [[R_ADDR]], align 8, !dbg [[DBG153:![0-9]+]]
+// CHECK-DEBUG-NEXT: store float [[CONV202]], float* [[TMP11]], align 4, !dbg [[DBG154:![0-9]+]]
+// CHECK-DEBUG-NEXT: br label [[OMP_LOOP_INC194]], !dbg [[DBG148]]
// CHECK-DEBUG: omp_loop.inc194:
-// CHECK-DEBUG-NEXT: [[OMP_LOOP_NEXT199]] = add nuw i32 [[OMP_LOOP_IV197]], 1, !dbg [[DBG149]]
-// CHECK-DEBUG-NEXT: br label [[OMP_LOOP_HEADER191]], !dbg [[DBG149]]
+// CHECK-DEBUG-NEXT: [[OMP_LOOP_NEXT199]] = add nuw i32 [[OMP_LOOP_IV197]], 1, !dbg [[DBG148]]
+// CHECK-DEBUG-NEXT: br label [[OMP_LOOP_HEADER191]], !dbg [[DBG148]]
// CHECK-DEBUG: omp_loop.exit195:
-// CHECK-DEBUG-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB42]], i32 [[OMP_GLOBAL_THREAD_NUM207]]), !dbg [[DBG149]]
-// CHECK-DEBUG-NEXT: [[OMP_GLOBAL_THREAD_NUM208:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB42]]), !dbg [[DBG151]]
-// CHECK-DEBUG-NEXT: call void @__kmpc_barrier(%struct.ident_t* @[[GLOB43:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM208]]), !dbg [[DBG151]]
-// CHECK-DEBUG-NEXT: br label [[OMP_LOOP_AFTER196:%.*]], !dbg [[DBG149]]
+// CHECK-DEBUG-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB42]], i32 [[OMP_GLOBAL_THREAD_NUM207]]), !dbg [[DBG148]]
+// CHECK-DEBUG-NEXT: [[OMP_GLOBAL_THREAD_NUM208:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB42]]), !dbg [[DBG150]]
+// CHECK-DEBUG-NEXT: call void @__kmpc_barrier(%struct.ident_t* @[[GLOB43:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM208]]), !dbg [[DBG150]]
+// CHECK-DEBUG-NEXT: br label [[OMP_LOOP_AFTER196:%.*]], !dbg [[DBG148]]
// CHECK-DEBUG: omp_loop.after196:
-// CHECK-DEBUG-NEXT: ret void, !dbg [[DBG156:![0-9]+]]
+// CHECK-DEBUG-NEXT: ret void, !dbg [[DBG155:![0-9]+]]
//
void parallel_for_2(float *r, int a, double b) {
#pragma omp parallel
diff --git a/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h b/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h
index 8c85d9818cb4f..2ad79ddb051c3 100644
--- a/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h
+++ b/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h
@@ -784,7 +784,7 @@ class OpenMPIRBuilder {
struct OutlineInfo {
using PostOutlineCBTy = std::function<void(Function &)>;
PostOutlineCBTy PostOutlineCB;
- BasicBlock *EntryBB, *ExitBB;
+ BasicBlock *EntryBB, *ExitBB, *OuterAllocaBB;
SmallVector<Value *, 2> ExcludeArgsFromAggregate;
/// Collect all blocks in between EntryBB and ExitBB in both the given
diff --git a/llvm/include/llvm/Transforms/Utils/CodeExtractor.h b/llvm/include/llvm/Transforms/Utils/CodeExtractor.h
index 0cfa7ffe920b4..bb23cf4a9a3cb 100644
--- a/llvm/include/llvm/Transforms/Utils/CodeExtractor.h
+++ b/llvm/include/llvm/Transforms/Utils/CodeExtractor.h
@@ -92,6 +92,11 @@ class CodeExtractorAnalysisCache {
BranchProbabilityInfo *BPI;
AssumptionCache *AC;
+ // A block outside of the extraction set where any intermediate
+ // allocations will be placed inside. If this is null, allocations
+ // will be placed in the entry block of the function.
+ BasicBlock *AllocationBlock;
+
// If true, varargs functions can be extracted.
bool AllowVarArgs;
@@ -120,11 +125,15 @@ class CodeExtractorAnalysisCache {
/// code is extracted, including vastart. If AllowAlloca is true, then
/// extraction of blocks containing alloca instructions would be possible,
/// however code extractor won't validate whether extraction is legal.
+ /// Any new allocations will be placed in the AllocationBlock, unless
+ /// it is null, in which case it will be placed in the entry block of
+ /// the function from which the code is being extracted.
CodeExtractor(ArrayRef<BasicBlock *> BBs, DominatorTree *DT = nullptr,
bool AggregateArgs = false, BlockFrequencyInfo *BFI = nullptr,
BranchProbabilityInfo *BPI = nullptr,
- AssumptionCache *AC = nullptr,
- bool AllowVarArgs = false, bool AllowAlloca = false,
+ AssumptionCache *AC = nullptr, bool AllowVarArgs = false,
+ bool AllowAlloca = false,
+ BasicBlock *AllocationBlock = nullptr,
std::string Suffix = "");
/// Create a code extractor for a loop body.
diff --git a/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp b/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp
index 81cb41ecd57db..2dc369e12ed1b 100644
--- a/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp
+++ b/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp
@@ -300,6 +300,7 @@ void OpenMPIRBuilder::finalize(Function *Fn) {
/* AssumptionCache */ nullptr,
/* AllowVarArgs */ true,
/* AllowAlloca */ true,
+ /* AllocaBlock*/ OI.OuterAllocaBB,
/* Suffix */ ".omp_par");
LLVM_DEBUG(dbgs() << "Before outlining: " << *OuterFn << "\n");
@@ -878,6 +879,7 @@ IRBuilder<>::InsertPoint OpenMPIRBuilder::createParallel(
InsertPointTy PreFiniIP(PRegPreFiniBB, PRegPreFiniTI->getIterator());
FiniCB(PreFiniIP);
+ OI.OuterAllocaBB = OuterAllocaBlock;
OI.EntryBB = PRegEntryBB;
OI.ExitBB = PRegExitBB;
@@ -901,6 +903,7 @@ IRBuilder<>::InsertPoint OpenMPIRBuilder::createParallel(
/* AssumptionCache */ nullptr,
/* AllowVarArgs */ true,
/* AllowAlloca */ true,
+ /* AllocationBlock */ OuterAllocaBlock,
/* Suffix */ ".omp_par");
// Find inputs to, outputs from the code region.
diff --git a/llvm/lib/Transforms/IPO/HotColdSplitting.cpp b/llvm/lib/Transforms/IPO/HotColdSplitting.cpp
index a964fcde03966..8388f9bc15b69 100644
--- a/llvm/lib/Transforms/IPO/HotColdSplitting.cpp
+++ b/llvm/lib/Transforms/IPO/HotColdSplitting.cpp
@@ -352,7 +352,7 @@ Function *HotColdSplitting::extractColdRegion(
// TODO: Pass BFI and BPI to update profile information.
CodeExtractor CE(Region, &DT, /* AggregateArgs */ false, /* BFI */ nullptr,
/* BPI */ nullptr, AC, /* AllowVarArgs */ false,
- /* AllowAlloca */ false,
+ /* AllowAlloca */ false, /* AllocaBlock */ nullptr,
/* Suffix */ "cold." + std::to_string(Count));
// Perform a simple cost/benefit analysis to decide whether or not to permit
diff --git a/llvm/lib/Transforms/IPO/IROutliner.cpp b/llvm/lib/Transforms/IPO/IROutliner.cpp
index 2e3b22794d23a..142ce10449f31 100644
--- a/llvm/lib/Transforms/IPO/IROutliner.cpp
+++ b/llvm/lib/Transforms/IPO/IROutliner.cpp
@@ -2679,7 +2679,7 @@ unsigned IROutliner::doOutline(Module &M) {
OS->Candidate->getBasicBlocks(BlocksInRegion, BE);
OS->CE = new (ExtractorAllocator.Allocate())
CodeExtractor(BE, nullptr, false, nullptr, nullptr, nullptr, false,
- false, "outlined");
+ false, nullptr, "outlined");
findAddInputsOutputs(M, *OS, NotSame);
if (!OS->IgnoreRegion)
OutlinedRegions.push_back(OS);
@@ -2790,7 +2790,7 @@ unsigned IROutliner::doOutline(Module &M) {
OS->Candidate->getBasicBlocks(BlocksInRegion, BE);
OS->CE = new (ExtractorAllocator.Allocate())
CodeExtractor(BE, nullptr, false, nullptr, nullptr, nullptr, false,
- false, "outlined");
+ false, nullptr, "outlined");
bool FunctionOutlined = extractSection(*OS);
if (FunctionOutlined) {
unsigned StartIdx = OS->Candidate->getStartIdx();
diff --git a/llvm/lib/Transforms/Utils/CodeExtractor.cpp b/llvm/lib/Transforms/Utils/CodeExtractor.cpp
index 9bb0596f2a2d5..37689225b6465 100644
--- a/llvm/lib/Transforms/Utils/CodeExtractor.cpp
+++ b/llvm/lib/Transforms/Utils/CodeExtractor.cpp
@@ -246,9 +246,10 @@ CodeExtractor::CodeExtractor(ArrayRef<BasicBlock *> BBs, DominatorTree *DT,
bool AggregateArgs, BlockFrequencyInfo *BFI,
BranchProbabilityInfo *BPI, AssumptionCache *AC,
bool AllowVarArgs, bool AllowAlloca,
- std::string Suffix)
+ BasicBlock *AllocationBlock, std::string Suffix)
: DT(DT), AggregateArgs(AggregateArgs || AggregateArgsOpt), BFI(BFI),
- BPI(BPI), AC(AC), AllowVarArgs(AllowVarArgs),
+ BPI(BPI), AC(AC), AllocationBlock(AllocationBlock),
+ AllowVarArgs(AllowVarArgs),
Blocks(buildExtractionBlockSet(BBs, DT, AllowVarArgs, AllowAlloca)),
Suffix(Suffix) {}
@@ -257,7 +258,7 @@ CodeExtractor::CodeExtractor(DominatorTree &DT, Loop &L, bool AggregateArgs,
BranchProbabilityInfo *BPI, AssumptionCache *AC,
std::string Suffix)
: DT(&DT), AggregateArgs(AggregateArgs || AggregateArgsOpt), BFI(BFI),
- BPI(BPI), AC(AC), AllowVarArgs(false),
+ BPI(BPI), AC(AC), AllocationBlock(nullptr), AllowVarArgs(false),
Blocks(buildExtractionBlockSet(L.getBlocks(), &DT,
/* AllowVarArgs */ false,
/* AllowAlloca */ false)),
@@ -1189,9 +1190,10 @@ CallInst *CodeExtractor::emitCallAndSwitchStatement(Function *newFunction,
// Allocate a struct at the beginning of this function
StructArgTy = StructType::get(newFunction->getContext(), ArgTypes);
- Struct = new AllocaInst(StructArgTy, DL.getAllocaAddrSpace(), nullptr,
- "structArg",
- &codeReplacer->getParent()->front().front());
+ Struct = new AllocaInst(
+ StructArgTy, DL.getAllocaAddrSpace(), nullptr, "structArg",
+ AllocationBlock ? &*AllocationBlock->getFirstInsertionPt()
+ : &codeReplacer->getParent()->front().front());
params.push_back(Struct);
// Store aggregated inputs in the struct.
diff --git a/mlir/test/Target/LLVMIR/openmp-nested.mlir b/mlir/test/Target/LLVMIR/openmp-nested.mlir
new file mode 100644
index 0000000000000..6e900a302b09f
--- /dev/null
+++ b/mlir/test/Target/LLVMIR/openmp-nested.mlir
@@ -0,0 +1,41 @@
+// RUN: mlir-translate -mlir-to-llvmir -split-input-file %s | FileCheck %s
+
+module {
+ llvm.func @printf(!llvm.ptr<i8>, ...) -> i32
+ llvm.mlir.global internal constant @str0("WG size of kernel = %d X %d\0A\00")
+
+ llvm.func @main(%arg0: i32, %arg1: !llvm.ptr<ptr<i8>>) -> i32 {
+ omp.parallel {
+ %0 = llvm.mlir.constant(1 : index) : i64
+ %1 = llvm.mlir.constant(10 : index) : i64
+ %2 = llvm.mlir.constant(0 : index) : i64
+ %4 = llvm.mlir.constant(0 : i32) : i32
+ %12 = llvm.alloca %0 x i64 : (i64) -> !llvm.ptr<i64>
+ omp.wsloop (%arg2) : i64 = (%2) to (%1) step (%0) {
+ omp.parallel {
+ omp.wsloop (%arg3) : i64 = (%2) to (%0) step (%0) {
+ llvm.store %2, %12 : !llvm.ptr<i64>
+ omp.yield
+ }
+ omp.terminator
+ }
+ %19 = llvm.load %12 : !llvm.ptr<i64>
+ %20 = llvm.trunc %19 : i64 to i32
+ %5 = llvm.mlir.addressof @str0 : !llvm.ptr<array<29 x i8>>
+ %6 = llvm.getelementptr %5[%4, %4] : (!llvm.ptr<array<29 x i8>>, i32, i32) -> !llvm.ptr<i8>
+ %21 = llvm.call @printf(%6, %20, %20) : (!llvm.ptr<i8>, i32, i32) -> i32
+ omp.yield
+ }
+ omp.terminator
+ }
+ %a4 = llvm.mlir.constant(0 : i32) : i32
+ llvm.return %a4 : i32
+ }
+
+}
+
+// CHECK: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @1, i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @[[inner1:.+]] to void (i32*, i32*, ...)*))
+
+// CHECK: define internal void @[[inner1]]
+// CHECK: %[[structArg:.+]] = alloca { i64* }
+// CHECK: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @3, i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, { i64* }*)* @[[inner2:.+]] to void (i32*, i32*, ...)*), { i64* }* %[[structArg]])
More information about the llvm-commits
mailing list