diff options
| author | Giorgis Georgakoudis <georgakoudis1@llnl.gov> | 2022-09-15 00:09:54 +0000 |
|---|---|---|
| committer | Dhruva Chakrabarti <Dhruva.Chakrabarti@amd.com> | 2022-09-15 00:54:05 +0000 |
| commit | 7539e9cf811e590d9f12ae39673ca789e26386b4 (patch) | |
| tree | 75e0c72276302bbdfee12daec93cfbb734d800c4 /clang/test/OpenMP/teams_distribute_simd_codegen.cpp | |
| parent | ef4b9c33f5dd096a4d7cd03654c17492f3802d89 (diff) | |
[OpenMP] Codegen aggregate for outlined function captures
Parallel regions are outlined as functions with capture variables explicitly generated as distinct parameters in the function's argument list. That complicates the fork_call interface in the OpenMP runtime: (1) the fork_call is variadic since there is a variable number of arguments to forward to the outlined function, (2) wrapping/unwrapping arguments happens in the OpenMP runtime, which is sub-optimal, has been a source of ABI bugs, and has a hardcoded limit (16) in the number of arguments, (3) forwarded arguments must cast to pointer types, which complicates debugging. This patch avoids those issues by aggregating captured arguments in a struct to pass to the fork_call.
Reviewed By: jdoerfert, jhuber6, ABataev
Differential Revision: https://reviews.llvm.org/D102107
Diffstat (limited to 'clang/test/OpenMP/teams_distribute_simd_codegen.cpp')
| -rw-r--r-- | clang/test/OpenMP/teams_distribute_simd_codegen.cpp | 1842 |
1 files changed, 977 insertions, 865 deletions
diff --git a/clang/test/OpenMP/teams_distribute_simd_codegen.cpp b/clang/test/OpenMP/teams_distribute_simd_codegen.cpp index 678735f33d0e..b98a727ad648 100644 --- a/clang/test/OpenMP/teams_distribute_simd_codegen.cpp +++ b/clang/test/OpenMP/teams_distribute_simd_codegen.cpp @@ -405,6 +405,7 @@ int main (int argc, char **argv) { // CHECK1-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 8 // CHECK1-NEXT: [[I_ADDR:%.*]] = alloca i64, align 8 // CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8 +// CHECK1-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8 // CHECK1-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]]) // CHECK1-NEXT: store i64 [[TE]], i64* [[TE_ADDR]], align 8 // CHECK1-NEXT: store i64 [[TH]], i64* [[TH_ADDR]], align 8 @@ -419,112 +420,118 @@ int main (int argc, char **argv) { // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[CONV]], align 4 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[CONV1]], align 4 // CHECK1-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]]) -// CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32*, [100 x i32]*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[CONV2]], i32* [[CONV3]], [100 x i32]* [[TMP1]]) +// CHECK1-NEXT: [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 +// CHECK1-NEXT: store i32* [[CONV2]], i32** [[TMP4]], align 8 +// CHECK1-NEXT: [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1 +// CHECK1-NEXT: store i32* [[CONV3]], i32** [[TMP5]], align 8 +// CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2 +// CHECK1-NEXT: store [100 x i32]* [[TMP1]], [100 x i32]** [[TMP6]], align 8 +// CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]]) // CHECK1-NEXT: ret void // // // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined. -// CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[I:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], [100 x i32]* noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] { +// CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.anon* noalias noundef [[__CONTEXT:%.*]]) #[[ATTR1]] { // CHECK1-NEXT: entry: // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 -// CHECK1-NEXT: [[I_ADDR:%.*]] = alloca i32*, align 8 -// CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8 -// CHECK1-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 8 +// CHECK1-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4 -// CHECK1-NEXT: [[I3:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 -// CHECK1-NEXT: [[I4:%.*]] = alloca i32, align 4 +// CHECK1-NEXT: [[I3:%.*]] = alloca i32, align 4 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 -// CHECK1-NEXT: store i32* [[I]], i32** [[I_ADDR]], align 8 -// CHECK1-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8 -// CHECK1-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 8 -// CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[I_ADDR]], align 8 -// CHECK1-NEXT: [[TMP1:%.*]] = load i32*, i32** [[N_ADDR]], align 8 -// CHECK1-NEXT: [[TMP2:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 8 -// CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP1]], align 4 -// CHECK1-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4 -// CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 -// CHECK1-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0 +// CHECK1-NEXT: store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8 +// CHECK1-NEXT: [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8 +// CHECK1-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0 +// CHECK1-NEXT: [[TMP2:%.*]] = load i32*, i32** [[TMP1]], align 8 +// CHECK1-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 1 +// CHECK1-NEXT: [[TMP4:%.*]] = load i32*, i32** [[TMP3]], align 8 +// CHECK1-NEXT: [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 2 +// CHECK1-NEXT: [[TMP6:%.*]] = load [100 x i32]*, [100 x i32]** [[TMP5]], align 8 +// CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP4]], align 4 +// CHECK1-NEXT: store i32 [[TMP7]], i32* [[DOTCAPTURE_EXPR_]], align 4 +// CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 +// CHECK1-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP8]], 0 // CHECK1-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 // CHECK1-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 // CHECK1-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4 -// CHECK1-NEXT: store i32 0, i32* [[I3]], align 4 -// CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 -// CHECK1-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]] +// CHECK1-NEXT: store i32 0, i32* [[I]], align 4 +// CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 +// CHECK1-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP9]] // CHECK1-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] // CHECK1: omp.precond.then: -// CHECK1-NEXT: [[ARRAYDECAY:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[TMP2]], i64 0, i64 0 +// CHECK1-NEXT: [[ARRAYDECAY:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[TMP6]], i64 0, i64 0 // CHECK1-NEXT: call void @llvm.assume(i1 true) [ "align"(i32* [[ARRAYDECAY]], i64 16) ] // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 -// CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 -// CHECK1-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4 +// CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK1-NEXT: store i32 [[TMP10]], i32* [[DOTOMP_UB]], align 4 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK1-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 -// CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4 -// CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) -// CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 -// CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 -// CHECK1-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]] -// CHECK1-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] +// CHECK1-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4 +// CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP12]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) +// CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK1-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK1-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP13]], [[TMP14]] +// CHECK1-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] // CHECK1: cond.true: -// CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK1-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 // CHECK1-NEXT: br label [[COND_END:%.*]] // CHECK1: cond.false: -// CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK1-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 // CHECK1-NEXT: br label [[COND_END]] // CHECK1: cond.end: -// CHECK1-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ] +// CHECK1-NEXT: [[COND:%.*]] = phi i32 [ [[TMP15]], [[COND_TRUE]] ], [ [[TMP16]], [[COND_FALSE]] ] // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 -// CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 -// CHECK1-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4 +// CHECK1-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 +// CHECK1-NEXT: store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] // CHECK1: omp.inner.for.cond: -// CHECK1-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 -// CHECK1-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !5 -// CHECK1-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]] -// CHECK1-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] +// CHECK1-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 +// CHECK1-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !5 +// CHECK1-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]] +// CHECK1-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] // CHECK1: omp.inner.for.body: -// CHECK1-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 -// CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1 +// CHECK1-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 +// CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP20]], 1 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] -// CHECK1-NEXT: store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !5 -// CHECK1-NEXT: [[TMP17:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !5 -// CHECK1-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64 -// CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[TMP2]], i64 0, i64 [[IDXPROM]] +// CHECK1-NEXT: store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !5 +// CHECK1-NEXT: [[TMP21:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !5 +// CHECK1-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP21]] to i64 +// CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[TMP6]], i64 0, i64 [[IDXPROM]] // CHECK1-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !5 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] // CHECK1: omp.body.continue: // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] // CHECK1: omp.inner.for.inc: -// CHECK1-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 -// CHECK1-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP18]], 1 -// CHECK1-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 +// CHECK1-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 +// CHECK1-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP22]], 1 +// CHECK1-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP6:![0-9]+]] // CHECK1: omp.inner.for.end: // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]] // CHECK1: omp.loop.exit: -// CHECK1-NEXT: [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 -// CHECK1-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4 -// CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP20]]) -// CHECK1-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK1-NEXT: [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0 -// CHECK1-NEXT: br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] +// CHECK1-NEXT: [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK1-NEXT: [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4 +// CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP24]]) +// CHECK1-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 +// CHECK1-NEXT: [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0 +// CHECK1-NEXT: br i1 [[TMP26]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] // CHECK1: .omp.final.then: -// CHECK1-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 -// CHECK1-NEXT: [[SUB8:%.*]] = sub nsw i32 [[TMP23]], 0 -// CHECK1-NEXT: [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1 -// CHECK1-NEXT: [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1 -// CHECK1-NEXT: [[ADD11:%.*]] = add nsw i32 0, [[MUL10]] -// CHECK1-NEXT: store i32 [[ADD11]], i32* [[TMP0]], align 4 +// CHECK1-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 +// CHECK1-NEXT: [[SUB7:%.*]] = sub nsw i32 [[TMP27]], 0 +// CHECK1-NEXT: [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1 +// CHECK1-NEXT: [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1 +// CHECK1-NEXT: [[ADD10:%.*]] = add nsw i32 0, [[MUL9]] +// CHECK1-NEXT: store i32 [[ADD10]], i32* [[TMP2]], align 4 // CHECK1-NEXT: br label [[DOTOMP_FINAL_DONE]] // CHECK1: .omp.final.done: // CHECK1-NEXT: br label [[OMP_PRECOND_END]] @@ -537,21 +544,25 @@ int main (int argc, char **argv) { // CHECK1-NEXT: entry: // CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8 // CHECK1-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 8 +// CHECK1-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 8 // CHECK1-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8 // CHECK1-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 8 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32* // CHECK1-NEXT: [[TMP0:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 8 -// CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, [100 x i32]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32* [[CONV]], [100 x i32]* [[TMP0]]) +// CHECK1-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 +// CHECK1-NEXT: store i32* [[CONV]], i32** [[TMP1]], align 8 +// CHECK1-NEXT: [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1 +// CHECK1-NEXT: store [100 x i32]* [[TMP0]], [100 x i32]** [[TMP2]], align 8 +// CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]]) // CHECK1-NEXT: ret void // // // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1 -// CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], [100 x i32]* noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] { +// CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias noundef [[__CONTEXT:%.*]]) #[[ATTR1]] { // CHECK1-NEXT: entry: // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 -// CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8 -// CHECK1-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 8 +// CHECK1-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 8 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 @@ -564,80 +575,82 @@ int main (int argc, char **argv) { // CHECK1-NEXT: [[I3:%.*]] = alloca i32, align 4 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 -// CHECK1-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8 -// CHECK1-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 8 -// CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8 -// CHECK1-NEXT: [[TMP1:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 8 -// CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4 -// CHECK1-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4 -// CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 -// CHECK1-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0 +// CHECK1-NEXT: store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 8 +// CHECK1-NEXT: [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 8 +// CHECK1-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0 +// CHECK1-NEXT: [[TMP2:%.*]] = load i32*, i32** [[TMP1]], align 8 +// CHECK1-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP0]], i32 0, i32 1 +// CHECK1-NEXT: [[TMP4:%.*]] = load [100 x i32]*, [100 x i32]** [[TMP3]], align 8 +// CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[TMP2]], align 4 +// CHECK1-NEXT: store i32 [[TMP5]], i32* [[DOTCAPTURE_EXPR_]], align 4 +// CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 +// CHECK1-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP6]], 0 // CHECK1-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 // CHECK1-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 // CHECK1-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4 // CHECK1-NEXT: store i32 0, i32* [[I]], align 4 -// CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 -// CHECK1-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP4]] +// CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 +// CHECK1-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP7]] // CHECK1-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] // CHECK1: omp.precond.then: // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 -// CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 -// CHECK1-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4 +// CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK1-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK1-NEXT: [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 -// CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4 -// CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP7]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) -// CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 -// CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 -// CHECK1-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]] +// CHECK1-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4 +// CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) +// CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK1-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]] // CHECK1-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] // CHECK1: cond.true: -// CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 // CHECK1-NEXT: br label [[COND_END:%.*]] // CHECK1: cond.false: -// CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK1-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 // CHECK1-NEXT: br label [[COND_END]] // CHECK1: cond.end: -// CHECK1-NEXT: [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ] +// CHECK1-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ] // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 -// CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 -// CHECK1-NEXT: store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4 +// CHECK1-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 +// CHECK1-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] // CHECK1: omp.inner.for.cond: -// CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 -// CHECK1-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 -// CHECK1-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]] +// CHECK1-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 +// CHECK1-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK1-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]] // CHECK1-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] // CHECK1: omp.inner.for.body: -// CHECK1-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 -// CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1 +// CHECK1-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 +// CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] // CHECK1-NEXT: store i32 [[ADD]], i32* [[I3]], align 4 -// CHECK1-NEXT: [[TMP16:%.*]] = load i32, i32* [[I3]], align 4 -// CHECK1-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64 -// CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[TMP1]], i64 0, i64 [[IDXPROM]] +// CHECK1-NEXT: [[TMP19:%.*]] = load i32, i32* [[I3]], align 4 +// CHECK1-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64 +// CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[TMP4]], i64 0, i64 [[IDXPROM]] // CHECK1-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] // CHECK1: omp.body.continue: // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] // CHECK1: omp.inner.for.inc: -// CHECK1-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 -// CHECK1-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP17]], 1 +// CHECK1-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 +// CHECK1-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1 // CHECK1-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]] // CHECK1: omp.inner.for.end: // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]] // CHECK1: omp.loop.exit: -// CHECK1-NEXT: [[TMP18:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 -// CHECK1-NEXT: [[TMP19:%.*]] = load i32, i32* [[TMP18]], align 4 -// CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP19]]) -// CHECK1-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK1-NEXT: [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0 -// CHECK1-NEXT: br i1 [[TMP21]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] +// CHECK1-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK1-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4 +// CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]]) +// CHECK1-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 +// CHECK1-NEXT: [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0 +// CHECK1-NEXT: br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] // CHECK1: .omp.final.then: -// CHECK1-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 -// CHECK1-NEXT: [[SUB7:%.*]] = sub nsw i32 [[TMP22]], 0 +// CHECK1-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 +// CHECK1-NEXT: [[SUB7:%.*]] = sub nsw i32 [[TMP25]], 0 // CHECK1-NEXT: [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1 // CHECK1-NEXT: [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1 // CHECK1-NEXT: [[ADD10:%.*]] = add nsw i32 0, [[MUL9]] @@ -845,6 +858,7 @@ int main (int argc, char **argv) { // CHECK3-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 4 // CHECK3-NEXT: [[I_ADDR:%.*]] = alloca i32, align 4 // CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4 // CHECK3-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]]) // CHECK3-NEXT: store i32 [[TE]], i32* [[TE_ADDR]], align 4 // CHECK3-NEXT: store i32 [[TH]], i32* [[TH_ADDR]], align 4 @@ -855,111 +869,117 @@ int main (int argc, char **argv) { // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TE_ADDR]], align 4 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[TH_ADDR]], align 4 // CHECK3-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]]) -// CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32*, [100 x i32]*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[I_ADDR]], i32* [[N_ADDR]], [100 x i32]* [[TMP1]]) +// CHECK3-NEXT: [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 +// CHECK3-NEXT: store i32* [[I_ADDR]], i32** [[TMP4]], align 4 +// CHECK3-NEXT: [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1 +// CHECK3-NEXT: store i32* [[N_ADDR]], i32** [[TMP5]], align 4 +// CHECK3-NEXT: [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2 +// CHECK3-NEXT: store [100 x i32]* [[TMP1]], [100 x i32]** [[TMP6]], align 4 +// CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]]) // CHECK3-NEXT: ret void // // // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined. -// CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[I:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], [100 x i32]* noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] { +// CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.anon* noalias noundef [[__CONTEXT:%.*]]) #[[ATTR1]] { // CHECK3-NEXT: entry: // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 -// CHECK3-NEXT: [[I_ADDR:%.*]] = alloca i32*, align 4 -// CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4 -// CHECK3-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 4 +// CHECK3-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 4 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4 -// CHECK3-NEXT: [[I3:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4 // CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 // CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 -// CHECK3-NEXT: [[I4:%.*]] = alloca i32, align 4 +// CHECK3-NEXT: [[I3:%.*]] = alloca i32, align 4 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 -// CHECK3-NEXT: store i32* [[I]], i32** [[I_ADDR]], align 4 -// CHECK3-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4 -// CHECK3-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 4 -// CHECK3-NEXT: [[TMP0:%.*]] = load i32*, i32** [[I_ADDR]], align 4 -// CHECK3-NEXT: [[TMP1:%.*]] = load i32*, i32** [[N_ADDR]], align 4 -// CHECK3-NEXT: [[TMP2:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 4 -// CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP1]], align 4 -// CHECK3-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4 -// CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 -// CHECK3-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0 +// CHECK3-NEXT: store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 4 +// CHECK3-NEXT: [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 4 +// CHECK3-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0 +// CHECK3-NEXT: [[TMP2:%.*]] = load i32*, i32** [[TMP1]], align 4 +// CHECK3-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 1 +// CHECK3-NEXT: [[TMP4:%.*]] = load i32*, i32** [[TMP3]], align 4 +// CHECK3-NEXT: [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 2 +// CHECK3-NEXT: [[TMP6:%.*]] = load [100 x i32]*, [100 x i32]** [[TMP5]], align 4 +// CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP4]], align 4 +// CHECK3-NEXT: store i32 [[TMP7]], i32* [[DOTCAPTURE_EXPR_]], align 4 +// CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 +// CHECK3-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP8]], 0 // CHECK3-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 // CHECK3-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 // CHECK3-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4 -// CHECK3-NEXT: store i32 0, i32* [[I3]], align 4 -// CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 -// CHECK3-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]] +// CHECK3-NEXT: store i32 0, i32* [[I]], align 4 +// CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 +// CHECK3-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP9]] // CHECK3-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] // CHECK3: omp.precond.then: -// CHECK3-NEXT: [[ARRAYDECAY:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[TMP2]], i32 0, i32 0 +// CHECK3-NEXT: [[ARRAYDECAY:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[TMP6]], i32 0, i32 0 // CHECK3-NEXT: call void @llvm.assume(i1 true) [ "align"(i32* [[ARRAYDECAY]], i32 16) ] // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 -// CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 -// CHECK3-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4 +// CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK3-NEXT: store i32 [[TMP10]], i32* [[DOTOMP_UB]], align 4 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK3-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 -// CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4 -// CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) -// CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 -// CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 -// CHECK3-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]] -// CHECK3-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] +// CHECK3-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 +// CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4 +// CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP12]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) +// CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK3-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK3-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP13]], [[TMP14]] +// CHECK3-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] // CHECK3: cond.true: -// CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK3-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 // CHECK3-NEXT: br label [[COND_END:%.*]] // CHECK3: cond.false: -// CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK3-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 // CHECK3-NEXT: br label [[COND_END]] // CHECK3: cond.end: -// CHECK3-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ] +// CHECK3-NEXT: [[COND:%.*]] = phi i32 [ [[TMP15]], [[COND_TRUE]] ], [ [[TMP16]], [[COND_FALSE]] ] // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 -// CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 -// CHECK3-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4 +// CHECK3-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 +// CHECK3-NEXT: store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] // CHECK3: omp.inner.for.cond: -// CHECK3-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6 -// CHECK3-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !6 -// CHECK3-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]] -// CHECK3-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] +// CHECK3-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6 +// CHECK3-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !6 +// CHECK3-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]] +// CHECK3-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] // CHECK3: omp.inner.for.body: -// CHECK3-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6 -// CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1 +// CHECK3-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6 +// CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP20]], 1 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] -// CHECK3-NEXT: store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !6 -// CHECK3-NEXT: [[TMP17:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !6 -// CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[TMP2]], i32 0, i32 [[TMP17]] +// CHECK3-NEXT: store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !6 +// CHECK3-NEXT: [[TMP21:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !6 +// CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[TMP6]], i32 0, i32 [[TMP21]] // CHECK3-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !6 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] // CHECK3: omp.body.continue: // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] // CHECK3: omp.inner.for.inc: -// CHECK3-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6 -// CHECK3-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP18]], 1 -// CHECK3-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6 +// CHECK3-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6 +// CHECK3-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP22]], 1 +// CHECK3-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]] // CHECK3: omp.inner.for.end: // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]] // CHECK3: omp.loop.exit: -// CHECK3-NEXT: [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 -// CHECK3-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4 -// CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP20]]) -// CHECK3-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK3-NEXT: [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0 -// CHECK3-NEXT: br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] +// CHECK3-NEXT: [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 +// CHECK3-NEXT: [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4 +// CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP24]]) +// CHECK3-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 +// CHECK3-NEXT: [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0 +// CHECK3-NEXT: br i1 [[TMP26]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] // CHECK3: .omp.final.then: -// CHECK3-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 -// CHECK3-NEXT: [[SUB8:%.*]] = sub nsw i32 [[TMP23]], 0 -// CHECK3-NEXT: [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1 -// CHECK3-NEXT: [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1 -// CHECK3-NEXT: [[ADD11:%.*]] = add nsw i32 0, [[MUL10]] -// CHECK3-NEXT: store i32 [[ADD11]], i32* [[TMP0]], align 4 +// CHECK3-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 +// CHECK3-NEXT: [[SUB7:%.*]] = sub nsw i32 [[TMP27]], 0 +// CHECK3-NEXT: [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1 +// CHECK3-NEXT: [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1 +// CHECK3-NEXT: [[ADD10:%.*]] = add nsw i32 0, [[MUL9]] +// CHECK3-NEXT: store i32 [[ADD10]], i32* [[TMP2]], align 4 // CHECK3-NEXT: br label [[DOTOMP_FINAL_DONE]] // CHECK3: .omp.final.done: // CHECK3-NEXT: br label [[OMP_PRECOND_END]] @@ -972,20 +992,24 @@ int main (int argc, char **argv) { // CHECK3-NEXT: entry: // CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4 // CHECK3-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 4 +// CHECK3-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 4 // CHECK3-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4 // CHECK3-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 4 // CHECK3-NEXT: [[TMP0:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 4 -// CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, [100 x i32]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], [100 x i32]* [[TMP0]]) +// CHECK3-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 +// CHECK3-NEXT: store i32* [[N_ADDR]], i32** [[TMP1]], align 4 +// CHECK3-NEXT: [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1 +// CHECK3-NEXT: store [100 x i32]* [[TMP0]], [100 x i32]** [[TMP2]], align 4 +// CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]]) // CHECK3-NEXT: ret void // // // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..1 -// CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], [100 x i32]* noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] { +// CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias noundef [[__CONTEXT:%.*]]) #[[ATTR1]] { // CHECK3-NEXT: entry: // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 -// CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4 -// CHECK3-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 4 +// CHECK3-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 4 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 @@ -998,79 +1022,81 @@ int main (int argc, char **argv) { // CHECK3-NEXT: [[I3:%.*]] = alloca i32, align 4 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 -// CHECK3-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4 -// CHECK3-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 4 -// CHECK3-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4 -// CHECK3-NEXT: [[TMP1:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 4 -// CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4 -// CHECK3-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4 -// CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 -// CHECK3-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0 +// CHECK3-NEXT: store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 4 +// CHECK3-NEXT: [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 4 +// CHECK3-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0 +// CHECK3-NEXT: [[TMP2:%.*]] = load i32*, i32** [[TMP1]], align 4 +// CHECK3-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP0]], i32 0, i32 1 +// CHECK3-NEXT: [[TMP4:%.*]] = load [100 x i32]*, [100 x i32]** [[TMP3]], align 4 +// CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[TMP2]], align 4 +// CHECK3-NEXT: store i32 [[TMP5]], i32* [[DOTCAPTURE_EXPR_]], align 4 +// CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 +// CHECK3-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP6]], 0 // CHECK3-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 // CHECK3-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 // CHECK3-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4 // CHECK3-NEXT: store i32 0, i32* [[I]], align 4 -// CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 -// CHECK3-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP4]] +// CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 +// CHECK3-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP7]] // CHECK3-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] // CHECK3: omp.precond.then: // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 -// CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 -// CHECK3-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4 +// CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK3-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK3-NEXT: [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 -// CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4 -// CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP7]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) -// CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 -// CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 -// CHECK3-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]] +// CHECK3-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 +// CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4 +// CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) +// CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK3-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]] // CHECK3-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] // CHECK3: cond.true: -// CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 // CHECK3-NEXT: br label [[COND_END:%.*]] // CHECK3: cond.false: -// CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK3-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 // CHECK3-NEXT: br label [[COND_END]] // CHECK3: cond.end: -// CHECK3-NEXT: [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ] +// CHECK3-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ] // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 -// CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 -// CHECK3-NEXT: store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4 +// CHECK3-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 +// CHECK3-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] // CHECK3: omp.inner.for.cond: -// CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 -// CHECK3-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 -// CHECK3-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]] +// CHECK3-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 +// CHECK3-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK3-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]] // CHECK3-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] // CHECK3: omp.inner.for.body: -// CHECK3-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 -// CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1 +// CHECK3-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 +// CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] // CHECK3-NEXT: store i32 [[ADD]], i32* [[I3]], align 4 -// CHECK3-NEXT: [[TMP16:%.*]] = load i32, i32* [[I3]], align 4 -// CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[TMP1]], i32 0, i32 [[TMP16]] +// CHECK3-NEXT: [[TMP19:%.*]] = load i32, i32* [[I3]], align 4 +// CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[TMP4]], i32 0, i32 [[TMP19]] // CHECK3-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] // CHECK3: omp.body.continue: // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] // CHECK3: omp.inner.for.inc: -// CHECK3-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 -// CHECK3-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP17]], 1 +// CHECK3-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 +// CHECK3-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1 // CHECK3-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]] // CHECK3: omp.inner.for.end: // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]] // CHECK3: omp.loop.exit: -// CHECK3-NEXT: [[TMP18:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 -// CHECK3-NEXT: [[TMP19:%.*]] = load i32, i32* [[TMP18]], align 4 -// CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP19]]) -// CHECK3-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK3-NEXT: [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0 -// CHECK3-NEXT: br i1 [[TMP21]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] +// CHECK3-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 +// CHECK3-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4 +// CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]]) +// CHECK3-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 +// CHECK3-NEXT: [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0 +// CHECK3-NEXT: br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] // CHECK3: .omp.final.then: -// CHECK3-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 -// CHECK3-NEXT: [[SUB7:%.*]] = sub nsw i32 [[TMP22]], 0 +// CHECK3-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 +// CHECK3-NEXT: [[SUB7:%.*]] = sub nsw i32 [[TMP25]], 0 // CHECK3-NEXT: [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1 // CHECK3-NEXT: [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1 // CHECK3-NEXT: [[ADD10:%.*]] = add nsw i32 0, [[MUL9]] @@ -1463,24 +1489,29 @@ int main (int argc, char **argv) { // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8 +// CHECK9-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8 // CHECK9-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32* // CHECK9-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8 // CHECK9-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8 -// CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]]) +// CHECK9-NEXT: [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 +// CHECK9-NEXT: store i32* [[CONV]], i32** [[TMP2]], align 8 +// CHECK9-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1 +// CHECK9-NEXT: store i64 [[TMP0]], i64* [[TMP3]], align 8 +// CHECK9-NEXT: [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2 +// CHECK9-NEXT: store i32* [[TMP1]], i32** [[TMP4]], align 8 +// CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]]) // CHECK9-NEXT: ret void // // // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined. -// CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] { +// CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.anon* noalias noundef [[__CONTEXT:%.*]]) #[[ATTR2]] { // CHECK9-NEXT: entry: // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 -// CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8 -// CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8 -// CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8 +// CHECK9-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 @@ -1493,82 +1524,84 @@ int main (int argc, char **argv) { // CHECK9-NEXT: [[I3:%.*]] = alloca i32, align 4 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 -// CHECK9-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8 -// CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8 -// CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8 -// CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8 -// CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8 -// CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8 -// CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4 -// CHECK9-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4 -// CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 -// CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0 +// CHECK9-NEXT: store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8 +// CHECK9-NEXT: [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8 +// CHECK9-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0 +// CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[TMP1]], align 8 +// CHECK9-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 1 +// CHECK9-NEXT: [[TMP4:%.*]] = load i64, i64* [[TMP3]], align 8 +// CHECK9-NEXT: [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 2 +// CHECK9-NEXT: [[TMP6:%.*]] = load i32*, i32** [[TMP5]], align 8 +// CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP2]], align 4 +// CHECK9-NEXT: store i32 [[TMP7]], i32* [[DOTCAPTURE_EXPR_]], align 4 +// CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 +// CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP8]], 0 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 // CHECK9-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 // CHECK9-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4 // CHECK9-NEXT: store i32 0, i32* [[I]], align 4 -// CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 -// CHECK9-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]] +// CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 +// CHECK9-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP9]] // CHECK9-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] // CHECK9: omp.precond.then: // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 -// CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 -// CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4 +// CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK9-NEXT: store i32 [[TMP10]], i32* [[DOTOMP_UB]], align 4 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK9-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 -// CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4 -// CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) -// CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 -// CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 -// CHECK9-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]] +// CHECK9-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4 +// CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP12]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) +// CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK9-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP13]], [[TMP14]] // CHECK9-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] // CHECK9: cond.true: -// CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 // CHECK9-NEXT: br label [[COND_END:%.*]] // CHECK9: cond.false: -// CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK9-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 // CHECK9-NEXT: br label [[COND_END]] // CHECK9: cond.end: -// CHECK9-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ] +// CHECK9-NEXT: [[COND:%.*]] = phi i32 [ [[TMP15]], [[COND_TRUE]] ], [ [[TMP16]], [[COND_FALSE]] ] // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 -// CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 -// CHECK9-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4 +// CHECK9-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 +// CHECK9-NEXT: store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] // CHECK9: omp.inner.for.cond: -// CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4 -// CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !4 -// CHECK9-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]] +// CHECK9-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4 +// CHECK9-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !4 +// CHECK9-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]] // CHECK9-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] // CHECK9: omp.inner.for.body: -// CHECK9-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4 -// CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1 +// CHECK9-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4 +// CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP20]], 1 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] // CHECK9-NEXT: store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !4 -// CHECK9-NEXT: [[TMP17:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !4 -// CHECK9-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64 -// CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]] +// CHECK9-NEXT: [[TMP21:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !4 +// CHECK9-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP21]] to i64 +// CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP6]], i64 [[IDXPROM]] // CHECK9-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !4 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] // CHECK9: omp.body.continue: // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] // CHECK9: omp.inner.for.inc: -// CHECK9-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4 -// CHECK9-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP18]], 1 +// CHECK9-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4 +// CHECK9-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP22]], 1 // CHECK9-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP5:![0-9]+]] // CHECK9: omp.inner.for.end: // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]] // CHECK9: omp.loop.exit: -// CHECK9-NEXT: [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 -// CHECK9-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4 -// CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP20]]) -// CHECK9-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK9-NEXT: [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0 -// CHECK9-NEXT: br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] +// CHECK9-NEXT: [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK9-NEXT: [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4 +// CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP24]]) +// CHECK9-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 +// CHECK9-NEXT: [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0 +// CHECK9-NEXT: br i1 [[TMP26]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] // CHECK9: .omp.final.then: -// CHECK9-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 -// CHECK9-NEXT: [[SUB7:%.*]] = sub nsw i32 [[TMP23]], 0 +// CHECK9-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 +// CHECK9-NEXT: [[SUB7:%.*]] = sub nsw i32 [[TMP27]], 0 // CHECK9-NEXT: [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1 // CHECK9-NEXT: [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1 // CHECK9-NEXT: [[ADD10:%.*]] = add nsw i32 0, [[MUL9]] @@ -1692,23 +1725,28 @@ int main (int argc, char **argv) { // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4 +// CHECK11-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4 // CHECK11-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4 // CHECK11-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4 // CHECK11-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4 -// CHECK11-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]]) +// CHECK11-NEXT: [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 +// CHECK11-NEXT: store i32* [[N_ADDR]], i32** [[TMP2]], align 4 +// CHECK11-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1 +// CHECK11-NEXT: store i32 [[TMP0]], i32* [[TMP3]], align 4 +// CHECK11-NEXT: [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2 +// CHECK11-NEXT: store i32* [[TMP1]], i32** [[TMP4]], align 4 +// CHECK11-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]]) // CHECK11-NEXT: ret void // // // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined. -// CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] { +// CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.anon* noalias noundef [[__CONTEXT:%.*]]) #[[ATTR2]] { // CHECK11-NEXT: entry: // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 -// CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4 -// CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4 -// CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4 +// CHECK11-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 4 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 @@ -1721,81 +1759,83 @@ int main (int argc, char **argv) { // CHECK11-NEXT: [[I3:%.*]] = alloca i32, align 4 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 -// CHECK11-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4 -// CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4 -// CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4 -// CHECK11-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4 -// CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4 -// CHECK11-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4 -// CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4 -// CHECK11-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4 -// CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 -// CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0 +// CHECK11-NEXT: store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 4 +// CHECK11-NEXT: [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 4 +// CHECK11-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0 +// CHECK11-NEXT: [[TMP2:%.*]] = load i32*, i32** [[TMP1]], align 4 +// CHECK11-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 1 +// CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4 +// CHECK11-NEXT: [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 2 +// CHECK11-NEXT: [[TMP6:%.*]] = load i32*, i32** [[TMP5]], align 4 +// CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP2]], align 4 +// CHECK11-NEXT: store i32 [[TMP7]], i32* [[DOTCAPTURE_EXPR_]], align 4 +// CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 +// CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP8]], 0 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 // CHECK11-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 // CHECK11-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4 // CHECK11-NEXT: store i32 0, i32* [[I]], align 4 -// CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 -// CHECK11-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]] +// CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 +// CHECK11-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP9]] // CHECK11-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] // CHECK11: omp.precond.then: // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 -// CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 -// CHECK11-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4 +// CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK11-NEXT: store i32 [[TMP10]], i32* [[DOTOMP_UB]], align 4 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK11-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 -// CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4 -// CHECK11-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) -// CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 -// CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 -// CHECK11-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]] +// CHECK11-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 +// CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4 +// CHECK11-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP12]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) +// CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK11-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK11-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP13]], [[TMP14]] // CHECK11-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] // CHECK11: cond.true: -// CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK11-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 // CHECK11-NEXT: br label [[COND_END:%.*]] // CHECK11: cond.false: -// CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK11-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 // CHECK11-NEXT: br label [[COND_END]] // CHECK11: cond.end: -// CHECK11-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ] +// CHECK11-NEXT: [[COND:%.*]] = phi i32 [ [[TMP15]], [[COND_TRUE]] ], [ [[TMP16]], [[COND_FALSE]] ] // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 -// CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 -// CHECK11-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4 +// CHECK11-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 +// CHECK11-NEXT: store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] // CHECK11: omp.inner.for.cond: -// CHECK11-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 -// CHECK11-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !5 -// CHECK11-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]] +// CHECK11-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 +// CHECK11-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !5 +// CHECK11-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]] // CHECK11-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] // CHECK11: omp.inner.for.body: -// CHECK11-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 -// CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1 +// CHECK11-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 +// CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP20]], 1 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] // CHECK11-NEXT: store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !5 -// CHECK11-NEXT: [[TMP17:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !5 -// CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP17]] +// CHECK11-NEXT: [[TMP21:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !5 +// CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP6]], i32 [[TMP21]] // CHECK11-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !5 // CHECK11-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] // CHECK11: omp.body.continue: // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] // CHECK11: omp.inner.for.inc: -// CHECK11-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 -// CHECK11-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP18]], 1 +// CHECK11-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 +// CHECK11-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP22]], 1 // CHECK11-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP6:![0-9]+]] // CHECK11: omp.inner.for.end: // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]] // CHECK11: omp.loop.exit: -// CHECK11-NEXT: [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 -// CHECK11-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4 -// CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP20]]) -// CHECK11-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK11-NEXT: [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0 -// CHECK11-NEXT: br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] +// CHECK11-NEXT: [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 +// CHECK11-NEXT: [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4 +// CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP24]]) +// CHECK11-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 +// CHECK11-NEXT: [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0 +// CHECK11-NEXT: br i1 [[TMP26]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] // CHECK11: .omp.final.then: -// CHECK11-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 -// CHECK11-NEXT: [[SUB7:%.*]] = sub nsw i32 [[TMP23]], 0 +// CHECK11-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 +// CHECK11-NEXT: [[SUB7:%.*]] = sub nsw i32 [[TMP27]], 0 // CHECK11-NEXT: [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1 // CHECK11-NEXT: [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1 // CHECK11-NEXT: [[ADD10:%.*]] = add nsw i32 0, [[MUL9]] @@ -2061,18 +2101,21 @@ int main (int argc, char **argv) { // CHECK17-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] { // CHECK17-NEXT: entry: // CHECK17-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8 +// CHECK17-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8 // CHECK17-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8 // CHECK17-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8 -// CHECK17-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]]) +// CHECK17-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 +// CHECK17-NEXT: store %struct.SS* [[TMP0]], %struct.SS** [[TMP1]], align 8 +// CHECK17-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]]) // CHECK17-NEXT: ret void // // // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined. -// CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] { +// CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.anon* noalias noundef [[__CONTEXT:%.*]]) #[[ATTR1]] { // CHECK17-NEXT: entry: // CHECK17-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 // CHECK17-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 -// CHECK17-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8 +// CHECK17-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8 // CHECK17-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 // CHECK17-NEXT: [[TMP:%.*]] = alloca i32, align 4 // CHECK17-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 @@ -2082,62 +2125,64 @@ int main (int argc, char **argv) { // CHECK17-NEXT: [[I:%.*]] = alloca i32, align 4 // CHECK17-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 // CHECK17-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 -// CHECK17-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8 -// CHECK17-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8 +// CHECK17-NEXT: store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8 +// CHECK17-NEXT: [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8 +// CHECK17-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0 +// CHECK17-NEXT: [[TMP2:%.*]] = load %struct.SS*, %struct.SS** [[TMP1]], align 8 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 // CHECK17-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4 // CHECK17-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK17-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 -// CHECK17-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4 -// CHECK17-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) -// CHECK17-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 -// CHECK17-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122 +// CHECK17-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK17-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4 +// CHECK17-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP4]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) +// CHECK17-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK17-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122 // CHECK17-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] // CHECK17: cond.true: // CHECK17-NEXT: br label [[COND_END:%.*]] // CHECK17: cond.false: -// CHECK17-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK17-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 // CHECK17-NEXT: br label [[COND_END]] // CHECK17: cond.end: -// CHECK17-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ] +// CHECK17-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ] // CHECK17-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 -// CHECK17-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 -// CHECK17-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4 +// CHECK17-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 +// CHECK17-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] // CHECK17: omp.inner.for.cond: -// CHECK17-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4 -// CHECK17-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !4 -// CHECK17-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]] +// CHECK17-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4 +// CHECK17-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !4 +// CHECK17-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]] // CHECK17-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] // CHECK17: omp.inner.for.body: -// CHECK17-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4 -// CHECK17-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1 +// CHECK17-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4 +// CHECK17-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1 // CHECK17-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] // CHECK17-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !4 -// CHECK17-NEXT: [[B:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 1 -// CHECK17-NEXT: [[TMP9:%.*]] = load float, float* [[B]], align 4, !llvm.access.group !4 -// CHECK17-NEXT: [[CONV:%.*]] = fptosi float [[TMP9]] to i32 -// CHECK17-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[TMP0]], i32 0, i32 0 -// CHECK17-NEXT: [[TMP10:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !4 -// CHECK17-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP10]] to i64 +// CHECK17-NEXT: [[B:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP2]], i32 0, i32 1 +// CHECK17-NEXT: [[TMP11:%.*]] = load float, float* [[B]], align 4, !llvm.access.group !4 +// CHECK17-NEXT: [[CONV:%.*]] = fptosi float [[TMP11]] to i32 +// CHECK17-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[TMP2]], i32 0, i32 0 +// CHECK17-NEXT: [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !4 +// CHECK17-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64 // CHECK17-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]] // CHECK17-NEXT: store i32 [[CONV]], i32* [[ARRAYIDX]], align 4, !llvm.access.group !4 // CHECK17-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] // CHECK17: omp.body.continue: // CHECK17-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] // CHECK17: omp.inner.for.inc: -// CHECK17-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4 -// CHECK17-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP11]], 1 +// CHECK17-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4 +// CHECK17-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1 // CHECK17-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP5:![0-9]+]] // CHECK17: omp.inner.for.end: // CHECK17-NEXT: br label [[OMP_LOOP_EXIT:%.*]] // CHECK17: omp.loop.exit: -// CHECK17-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]]) -// CHECK17-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK17-NEXT: [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0 -// CHECK17-NEXT: br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] +// CHECK17-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]]) +// CHECK17-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 +// CHECK17-NEXT: [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0 +// CHECK17-NEXT: br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] // CHECK17: .omp.final.then: // CHECK17-NEXT: store i32 123, i32* [[I]], align 4 // CHECK17-NEXT: br label [[DOTOMP_FINAL_DONE]] @@ -2247,18 +2292,21 @@ int main (int argc, char **argv) { // CHECK19-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] { // CHECK19-NEXT: entry: // CHECK19-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4 +// CHECK19-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4 // CHECK19-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4 // CHECK19-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4 -// CHECK19-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]]) +// CHECK19-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 +// CHECK19-NEXT: store %struct.SS* [[TMP0]], %struct.SS** [[TMP1]], align 4 +// CHECK19-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]]) // CHECK19-NEXT: ret void // // // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined. -// CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] { +// CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.anon* noalias noundef [[__CONTEXT:%.*]]) #[[ATTR1]] { // CHECK19-NEXT: entry: // CHECK19-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 // CHECK19-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 -// CHECK19-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4 +// CHECK19-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 4 // CHECK19-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 // CHECK19-NEXT: [[TMP:%.*]] = alloca i32, align 4 // CHECK19-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 @@ -2268,61 +2316,63 @@ int main (int argc, char **argv) { // CHECK19-NEXT: [[I:%.*]] = alloca i32, align 4 // CHECK19-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 // CHECK19-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 -// CHECK19-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4 -// CHECK19-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4 +// CHECK19-NEXT: store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 4 +// CHECK19-NEXT: [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 4 +// CHECK19-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0 +// CHECK19-NEXT: [[TMP2:%.*]] = load %struct.SS*, %struct.SS** [[TMP1]], align 4 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 // CHECK19-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4 // CHECK19-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK19-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 -// CHECK19-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4 -// CHECK19-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) -// CHECK19-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 -// CHECK19-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122 +// CHECK19-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 +// CHECK19-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4 +// CHECK19-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP4]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) +// CHECK19-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK19-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122 // CHECK19-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] // CHECK19: cond.true: // CHECK19-NEXT: br label [[COND_END:%.*]] // CHECK19: cond.false: -// CHECK19-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK19-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 // CHECK19-NEXT: br label [[COND_END]] // CHECK19: cond.end: -// CHECK19-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ] +// CHECK19-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ] // CHECK19-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 -// CHECK19-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 -// CHECK19-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4 +// CHECK19-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 +// CHECK19-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] // CHECK19: omp.inner.for.cond: -// CHECK19-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 -// CHECK19-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !5 -// CHECK19-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]] +// CHECK19-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 +// CHECK19-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !5 +// CHECK19-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]] // CHECK19-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] // CHECK19: omp.inner.for.body: -// CHECK19-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 -// CHECK19-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1 +// CHECK19-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 +// CHECK19-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1 // CHECK19-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] // CHECK19-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !5 -// CHECK19-NEXT: [[B:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 1 -// CHECK19-NEXT: [[TMP9:%.*]] = load float, float* [[B]], align 4, !llvm.access.group !5 -// CHECK19-NEXT: [[CONV:%.*]] = fptosi float [[TMP9]] to i32 -// CHECK19-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[TMP0]], i32 0, i32 0 -// CHECK19-NEXT: [[TMP10:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !5 -// CHECK19-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP10]] +// CHECK19-NEXT: [[B:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP2]], i32 0, i32 1 +// CHECK19-NEXT: [[TMP11:%.*]] = load float, float* [[B]], align 4, !llvm.access.group !5 +// CHECK19-NEXT: [[CONV:%.*]] = fptosi float [[TMP11]] to i32 +// CHECK19-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[TMP2]], i32 0, i32 0 +// CHECK19-NEXT: [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !5 +// CHECK19-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP12]] // CHECK19-NEXT: store i32 [[CONV]], i32* [[ARRAYIDX]], align 4, !llvm.access.group !5 // CHECK19-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] // CHECK19: omp.body.continue: // CHECK19-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] // CHECK19: omp.inner.for.inc: -// CHECK19-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 -// CHECK19-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP11]], 1 +// CHECK19-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 +// CHECK19-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1 // CHECK19-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP6:![0-9]+]] // CHECK19: omp.inner.for.end: // CHECK19-NEXT: br label [[OMP_LOOP_EXIT:%.*]] // CHECK19: omp.loop.exit: -// CHECK19-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]]) -// CHECK19-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK19-NEXT: [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0 -// CHECK19-NEXT: br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] +// CHECK19-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]]) +// CHECK19-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 +// CHECK19-NEXT: [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0 +// CHECK19-NEXT: br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] // CHECK19: .omp.final.then: // CHECK19-NEXT: store i32 123, i32* [[I]], align 4 // CHECK19-NEXT: br label [[DOTOMP_FINAL_DONE]] @@ -2433,7 +2483,7 @@ int main (int argc, char **argv) { // CHECK21-NEXT: entry: // CHECK21-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8 // CHECK21-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1 -// CHECK21-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8 +// CHECK21-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8 // CHECK21-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8 // CHECK21-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8 // CHECK21-NEXT: [[B:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 1 @@ -2441,23 +2491,24 @@ int main (int argc, char **argv) { // CHECK21-NEXT: [[TOBOOL:%.*]] = fcmp une float [[TMP1]], 0.000000e+00 // CHECK21-NEXT: [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8 // CHECK21-NEXT: store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1 -// CHECK21-NEXT: [[TMP2:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1 -// CHECK21-NEXT: [[TOBOOL1:%.*]] = trunc i8 [[TMP2]] to i1 -// CHECK21-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i8* +// CHECK21-NEXT: [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 +// CHECK21-NEXT: store %struct.SS* [[TMP0]], %struct.SS** [[TMP2]], align 8 +// CHECK21-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1 +// CHECK21-NEXT: [[TMP4:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1 +// CHECK21-NEXT: [[TOBOOL1:%.*]] = trunc i8 [[TMP4]] to i1 // CHECK21-NEXT: [[FROMBOOL2:%.*]] = zext i1 [[TOBOOL1]] to i8 -// CHECK21-NEXT: store i8 [[FROMBOOL2]], i8* [[CONV]], align 1 -// CHECK21-NEXT: [[TMP3:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8 -// CHECK21-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*, i64)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]], i64 [[TMP3]]) +// CHECK21-NEXT: store i8 [[FROMBOOL2]], i8* [[TMP3]], align 8 +// CHECK21-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]]) // CHECK21-NEXT: ret void // // // CHECK21-LABEL: define {{[^@]+}}@.omp_outlined. -// CHECK21-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] { +// CHECK21-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.anon* noalias noundef [[__CONTEXT:%.*]]) #[[ATTR1]] { // CHECK21-NEXT: entry: // CHECK21-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 // CHECK21-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 -// CHECK21-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8 -// CHECK21-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8 +// CHECK21-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8 +// CHECK21-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1 // CHECK21-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 // CHECK21-NEXT: [[TMP:%.*]] = alloca i32, align 4 // CHECK21-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 @@ -2467,59 +2518,64 @@ int main (int argc, char **argv) { // CHECK21-NEXT: [[I:%.*]] = alloca i32, align 4 // CHECK21-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 // CHECK21-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 -// CHECK21-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8 -// CHECK21-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8 -// CHECK21-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8 -// CHECK21-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i8* +// CHECK21-NEXT: store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8 +// CHECK21-NEXT: [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8 +// CHECK21-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0 +// CHECK21-NEXT: [[TMP2:%.*]] = load %struct.SS*, %struct.SS** [[TMP1]], align 8 +// CHECK21-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 1 +// CHECK21-NEXT: [[TMP4:%.*]] = load i8, i8* [[TMP3]], align 8 +// CHECK21-NEXT: [[TOBOOL:%.*]] = trunc i8 [[TMP4]] to i1 +// CHECK21-NEXT: [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8 +// CHECK21-NEXT: store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1 // CHECK21-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 // CHECK21-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4 // CHECK21-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 // CHECK21-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK21-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 -// CHECK21-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4 -// CHECK21-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) -// CHECK21-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 -// CHECK21-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122 +// CHECK21-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK21-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4 +// CHECK21-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP6]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) +// CHECK21-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK21-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 122 // CHECK21-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] // CHECK21: cond.true: // CHECK21-NEXT: br label [[COND_END:%.*]] // CHECK21: cond.false: -// CHECK21-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK21-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 // CHECK21-NEXT: br label [[COND_END]] // CHECK21: cond.end: -// CHECK21-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ] +// CHECK21-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ] // CHECK21-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 -// CHECK21-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 -// CHECK21-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4 -// CHECK21-NEXT: [[TMP6:%.*]] = load i8, i8* [[CONV]], align 1 -// CHECK21-NEXT: [[TOBOOL:%.*]] = trunc i8 [[TMP6]] to i1 -// CHECK21-NEXT: br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]] +// CHECK21-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 +// CHECK21-NEXT: store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4 +// CHECK21-NEXT: [[TMP10:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1 +// CHECK21-NEXT: [[TOBOOL1:%.*]] = trunc i8 [[TMP10]] to i1 +// CHECK21-NEXT: br i1 [[TOBOOL1]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]] // CHECK21: omp_if.then: // CHECK21-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] // CHECK21: omp.inner.for.cond: -// CHECK21-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4 -// CHECK21-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !4 -// CHECK21-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]] -// CHECK21-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] +// CHECK21-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4 +// CHECK21-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !4 +// CHECK21-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]] +// CHECK21-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] // CHECK21: omp.inner.for.body: -// CHECK21-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4 -// CHECK21-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1 +// CHECK21-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4 +// CHECK21-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP13]], 1 // CHECK21-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] // CHECK21-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !4 -// CHECK21-NEXT: [[B:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 1 -// CHECK21-NEXT: [[TMP10:%.*]] = load float, float* [[B]], align 4, !nontemporal !5, !llvm.access.group !4 -// CHECK21-NEXT: [[CONV2:%.*]] = fptosi float [[TMP10]] to i32 -// CHECK21-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[TMP0]], i32 0, i32 0 -// CHECK21-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !4 -// CHECK21-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64 +// CHECK21-NEXT: [[B:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP2]], i32 0, i32 1 +// CHECK21-NEXT: [[TMP14:%.*]] = load float, float* [[B]], align 4, !nontemporal !5, !llvm.access.group !4 +// CHECK21-NEXT: [[CONV:%.*]] = fptosi float [[TMP14]] to i32 +// CHECK21-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[TMP2]], i32 0, i32 0 +// CHECK21-NEXT: [[TMP15:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !4 +// CHECK21-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP15]] to i64 // CHECK21-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]] -// CHECK21-NEXT: store i32 [[CONV2]], i32* [[ARRAYIDX]], align 4, !llvm.access.group !4 +// CHECK21-NEXT: store i32 [[CONV]], i32* [[ARRAYIDX]], align 4, !llvm.access.group !4 // CHECK21-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] // CHECK21: omp.body.continue: // CHECK21-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] // CHECK21: omp.inner.for.inc: -// CHECK21-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4 -// CHECK21-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1 +// CHECK21-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4 +// CHECK21-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP16]], 1 // CHECK21-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4 // CHECK21-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP6:![0-9]+]] // CHECK21: omp.inner.for.end: @@ -2527,29 +2583,29 @@ int main (int argc, char **argv) { // CHECK21: omp_if.else: // CHECK21-NEXT: br label [[OMP_INNER_FOR_COND4:%.*]] // CHECK21: omp.inner.for.cond4: -// CHECK21-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 -// CHECK21-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 -// CHECK21-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]] +// CHECK21-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 +// CHECK21-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK21-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]] // CHECK21-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY6:%.*]], label [[OMP_INNER_FOR_END17:%.*]] // CHECK21: omp.inner.for.body6: -// CHECK21-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 -// CHECK21-NEXT: [[MUL7:%.*]] = mul nsw i32 [[TMP15]], 1 +// CHECK21-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 +// CHECK21-NEXT: [[MUL7:%.*]] = mul nsw i32 [[TMP19]], 1 // CHECK21-NEXT: [[ADD8:%.*]] = add nsw i32 0, [[MUL7]] // CHECK21-NEXT: store i32 [[ADD8]], i32* [[I]], align 4 -// CHECK21-NEXT: [[B9:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[TMP0]], i32 0, i32 1 -// CHECK21-NEXT: [[TMP16:%.*]] = load float, float* [[B9]], align 4 -// CHECK21-NEXT: [[CONV10:%.*]] = fptosi float [[TMP16]] to i32 -// CHECK21-NEXT: [[A11:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[TMP0]], i32 0, i32 0 -// CHECK21-NEXT: [[TMP17:%.*]] = load i32, i32* [[I]], align 4 -// CHECK21-NEXT: [[IDXPROM12:%.*]] = sext i32 [[TMP17]] to i64 +// CHECK21-NEXT: [[B9:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[TMP2]], i32 0, i32 1 +// CHECK21-NEXT: [[TMP20:%.*]] = load float, float* [[B9]], align 4 +// CHECK21-NEXT: [[CONV10:%.*]] = fptosi float [[TMP20]] to i32 +// CHECK21-NEXT: [[A11:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[TMP2]], i32 0, i32 0 +// CHECK21-NEXT: [[TMP21:%.*]] = load i32, i32* [[I]], align 4 +// CHECK21-NEXT: [[IDXPROM12:%.*]] = sext i32 [[TMP21]] to i64 // CHECK21-NEXT: [[ARRAYIDX13:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A11]], i64 0, i64 [[IDXPROM12]] // CHECK21-NEXT: store i32 [[CONV10]], i32* [[ARRAYIDX13]], align 4 // CHECK21-NEXT: br label [[OMP_BODY_CONTINUE14:%.*]] // CHECK21: omp.body.continue14: // CHECK21-NEXT: br label [[OMP_INNER_FOR_INC15:%.*]] // CHECK21: omp.inner.for.inc15: -// CHECK21-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 -// CHECK21-NEXT: [[ADD16:%.*]] = add nsw i32 [[TMP18]], 1 +// CHECK21-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 +// CHECK21-NEXT: [[ADD16:%.*]] = add nsw i32 [[TMP22]], 1 // CHECK21-NEXT: store i32 [[ADD16]], i32* [[DOTOMP_IV]], align 4 // CHECK21-NEXT: br label [[OMP_INNER_FOR_COND4]], !llvm.loop [[LOOP9:![0-9]+]] // CHECK21: omp.inner.for.end17: @@ -2557,10 +2613,10 @@ int main (int argc, char **argv) { // CHECK21: omp_if.end: // CHECK21-NEXT: br label [[OMP_LOOP_EXIT:%.*]] // CHECK21: omp.loop.exit: -// CHECK21-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]]) -// CHECK21-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK21-NEXT: [[TMP20:%.*]] = icmp ne i32 [[TMP19]], 0 -// CHECK21-NEXT: br i1 [[TMP20]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] +// CHECK21-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP6]]) +// CHECK21-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 +// CHECK21-NEXT: [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0 +// CHECK21-NEXT: br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] // CHECK21: .omp.final.then: // CHECK21-NEXT: store i32 123, i32* [[I]], align 4 // CHECK21-NEXT: br label [[DOTOMP_FINAL_DONE]] @@ -2671,7 +2727,7 @@ int main (int argc, char **argv) { // CHECK23-NEXT: entry: // CHECK23-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4 // CHECK23-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1 -// CHECK23-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4 +// CHECK23-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4 // CHECK23-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4 // CHECK23-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4 // CHECK23-NEXT: [[B:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 1 @@ -2679,23 +2735,24 @@ int main (int argc, char **argv) { // CHECK23-NEXT: [[TOBOOL:%.*]] = fcmp une float [[TMP1]], 0.000000e+00 // CHECK23-NEXT: [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8 // CHECK23-NEXT: store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1 -// CHECK23-NEXT: [[TMP2:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1 -// CHECK23-NEXT: [[TOBOOL1:%.*]] = trunc i8 [[TMP2]] to i1 -// CHECK23-NEXT: [[CONV:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__CASTED]] to i8* +// CHECK23-NEXT: [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 +// CHECK23-NEXT: store %struct.SS* [[TMP0]], %struct.SS** [[TMP2]], align 4 +// CHECK23-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1 +// CHECK23-NEXT: [[TMP4:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1 +// CHECK23-NEXT: [[TOBOOL1:%.*]] = trunc i8 [[TMP4]] to i1 // CHECK23-NEXT: [[FROMBOOL2:%.*]] = zext i1 [[TOBOOL1]] to i8 -// CHECK23-NEXT: store i8 [[FROMBOOL2]], i8* [[CONV]], align 1 -// CHECK23-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4 -// CHECK23-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*, i32)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]], i32 [[TMP3]]) +// CHECK23-NEXT: store i8 [[FROMBOOL2]], i8* [[TMP3]], align 4 +// CHECK23-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]]) // CHECK23-NEXT: ret void // // // CHECK23-LABEL: define {{[^@]+}}@.omp_outlined. -// CHECK23-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] { +// CHECK23-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.anon* noalias noundef [[__CONTEXT:%.*]]) #[[ATTR1]] { // CHECK23-NEXT: entry: // CHECK23-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 // CHECK23-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 -// CHECK23-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4 -// CHECK23-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4 +// CHECK23-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 4 +// CHECK23-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1 // CHECK23-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 // CHECK23-NEXT: [[TMP:%.*]] = alloca i32, align 4 // CHECK23-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 @@ -2705,58 +2762,63 @@ int main (int argc, char **argv) { // CHECK23-NEXT: [[I:%.*]] = alloca i32, align 4 // CHECK23-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 // CHECK23-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 -// CHECK23-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4 -// CHECK23-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4 -// CHECK23-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4 -// CHECK23-NEXT: [[CONV:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__ADDR]] to i8* +// CHECK23-NEXT: store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 4 +// CHECK23-NEXT: [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 4 +// CHECK23-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0 +// CHECK23-NEXT: [[TMP2:%.*]] = load %struct.SS*, %struct.SS** [[TMP1]], align 4 +// CHECK23-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 1 +// CHECK23-NEXT: [[TMP4:%.*]] = load i8, i8* [[TMP3]], align 4 +// CHECK23-NEXT: [[TOBOOL:%.*]] = trunc i8 [[TMP4]] to i1 +// CHECK23-NEXT: [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8 +// CHECK23-NEXT: store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1 // CHECK23-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 // CHECK23-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4 // CHECK23-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 // CHECK23-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK23-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 -// CHECK23-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4 -// CHECK23-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) -// CHECK23-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 -// CHECK23-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122 +// CHECK23-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 +// CHECK23-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4 +// CHECK23-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP6]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) +// CHECK23-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK23-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 122 // CHECK23-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] // CHECK23: cond.true: // CHECK23-NEXT: br label [[COND_END:%.*]] // CHECK23: cond.false: -// CHECK23-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK23-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 // CHECK23-NEXT: br label [[COND_END]] // CHECK23: cond.end: -// CHECK23-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ] +// CHECK23-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ] // CHECK23-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 -// CHECK23-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 -// CHECK23-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4 -// CHECK23-NEXT: [[TMP6:%.*]] = load i8, i8* [[CONV]], align 1 -// CHECK23-NEXT: [[TOBOOL:%.*]] = trunc i8 [[TMP6]] to i1 -// CHECK23-NEXT: br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]] +// CHECK23-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 +// CHECK23-NEXT: store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4 +// CHECK23-NEXT: [[TMP10:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1 +// CHECK23-NEXT: [[TOBOOL1:%.*]] = trunc i8 [[TMP10]] to i1 +// CHECK23-NEXT: br i1 [[TOBOOL1]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]] // CHECK23: omp_if.then: // CHECK23-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] // CHECK23: omp.inner.for.cond: -// CHECK23-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 -// CHECK23-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !5 -// CHECK23-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]] -// CHECK23-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] +// CHECK23-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 +// CHECK23-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !5 +// CHECK23-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]] +// CHECK23-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] // CHECK23: omp.inner.for.body: -// CHECK23-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 -// CHECK23-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1 +// CHECK23-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 +// CHECK23-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP13]], 1 // CHECK23-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] // CHECK23-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !5 -// CHECK23-NEXT: [[B:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 1 -// CHECK23-NEXT: [[TMP10:%.*]] = load float, float* [[B]], align 4, !nontemporal !6, !llvm.access.group !5 -// CHECK23-NEXT: [[CONV2:%.*]] = fptosi float [[TMP10]] to i32 -// CHECK23-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[TMP0]], i32 0, i32 0 -// CHECK23-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !5 -// CHECK23-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]] -// CHECK23-NEXT: store i32 [[CONV2]], i32* [[ARRAYIDX]], align 4, !llvm.access.group !5 +// CHECK23-NEXT: [[B:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP2]], i32 0, i32 1 +// CHECK23-NEXT: [[TMP14:%.*]] = load float, float* [[B]], align 4, !nontemporal !6, !llvm.access.group !5 +// CHECK23-NEXT: [[CONV:%.*]] = fptosi float [[TMP14]] to i32 +// CHECK23-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[TMP2]], i32 0, i32 0 +// CHECK23-NEXT: [[TMP15:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !5 +// CHECK23-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP15]] +// CHECK23-NEXT: store i32 [[CONV]], i32* [[ARRAYIDX]], align 4, !llvm.access.group !5 // CHECK23-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] // CHECK23: omp.body.continue: // CHECK23-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] // CHECK23: omp.inner.for.inc: -// CHECK23-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 -// CHECK23-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1 +// CHECK23-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 +// CHECK23-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP16]], 1 // CHECK23-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 // CHECK23-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]] // CHECK23: omp.inner.for.end: @@ -2764,28 +2826,28 @@ int main (int argc, char **argv) { // CHECK23: omp_if.else: // CHECK23-NEXT: br label [[OMP_INNER_FOR_COND4:%.*]] // CHECK23: omp.inner.for.cond4: -// CHECK23-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 -// CHECK23-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 -// CHECK23-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]] +// CHECK23-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 +// CHECK23-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK23-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]] // CHECK23-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY6:%.*]], label [[OMP_INNER_FOR_END16:%.*]] // CHECK23: omp.inner.for.body6: -// CHECK23-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 -// CHECK23-NEXT: [[MUL7:%.*]] = mul nsw i32 [[TMP15]], 1 +// CHECK23-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 +// CHECK23-NEXT: [[MUL7:%.*]] = mul nsw i32 [[TMP19]], 1 // CHECK23-NEXT: [[ADD8:%.*]] = add nsw i32 0, [[MUL7]] // CHECK23-NEXT: store i32 [[ADD8]], i32* [[I]], align 4 -// CHECK23-NEXT: [[B9:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[TMP0]], i32 0, i32 1 -// CHECK23-NEXT: [[TMP16:%.*]] = load float, float* [[B9]], align 4 -// CHECK23-NEXT: [[CONV10:%.*]] = fptosi float [[TMP16]] to i32 -// CHECK23-NEXT: [[A11:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[TMP0]], i32 0, i32 0 -// CHECK23-NEXT: [[TMP17:%.*]] = load i32, i32* [[I]], align 4 -// CHECK23-NEXT: [[ARRAYIDX12:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A11]], i32 0, i32 [[TMP17]] +// CHECK23-NEXT: [[B9:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[TMP2]], i32 0, i32 1 +// CHECK23-NEXT: [[TMP20:%.*]] = load float, float* [[B9]], align 4 +// CHECK23-NEXT: [[CONV10:%.*]] = fptosi float [[TMP20]] to i32 +// CHECK23-NEXT: [[A11:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[TMP2]], i32 0, i32 0 +// CHECK23-NEXT: [[TMP21:%.*]] = load i32, i32* [[I]], align 4 +// CHECK23-NEXT: [[ARRAYIDX12:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A11]], i32 0, i32 [[TMP21]] // CHECK23-NEXT: store i32 [[CONV10]], i32* [[ARRAYIDX12]], align 4 // CHECK23-NEXT: br label [[OMP_BODY_CONTINUE13:%.*]] // CHECK23: omp.body.continue13: // CHECK23-NEXT: br label [[OMP_INNER_FOR_INC14:%.*]] // CHECK23: omp.inner.for.inc14: -// CHECK23-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 -// CHECK23-NEXT: [[ADD15:%.*]] = add nsw i32 [[TMP18]], 1 +// CHECK23-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 +// CHECK23-NEXT: [[ADD15:%.*]] = add nsw i32 [[TMP22]], 1 // CHECK23-NEXT: store i32 [[ADD15]], i32* [[DOTOMP_IV]], align 4 // CHECK23-NEXT: br label [[OMP_INNER_FOR_COND4]], !llvm.loop [[LOOP10:![0-9]+]] // CHECK23: omp.inner.for.end16: @@ -2793,10 +2855,10 @@ int main (int argc, char **argv) { // CHECK23: omp_if.end: // CHECK23-NEXT: br label [[OMP_LOOP_EXIT:%.*]] // CHECK23: omp.loop.exit: -// CHECK23-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]]) -// CHECK23-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK23-NEXT: [[TMP20:%.*]] = icmp ne i32 [[TMP19]], 0 -// CHECK23-NEXT: br i1 [[TMP20]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] +// CHECK23-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP6]]) +// CHECK23-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 +// CHECK23-NEXT: [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0 +// CHECK23-NEXT: br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] // CHECK23: .omp.final.then: // CHECK23-NEXT: store i32 123, i32* [[I]], align 4 // CHECK23-NEXT: br label [[DOTOMP_FINAL_DONE]] @@ -3238,24 +3300,29 @@ int main (int argc, char **argv) { // CHECK33-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8 // CHECK33-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8 // CHECK33-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8 +// CHECK33-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8 // CHECK33-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8 // CHECK33-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8 // CHECK33-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8 // CHECK33-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32* // CHECK33-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8 // CHECK33-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8 -// CHECK33-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]]) +// CHECK33-NEXT: [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 +// CHECK33-NEXT: store i32* [[CONV]], i32** [[TMP2]], align 8 +// CHECK33-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1 +// CHECK33-NEXT: store i64 [[TMP0]], i64* [[TMP3]], align 8 +// CHECK33-NEXT: [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2 +// CHECK33-NEXT: store i32* [[TMP1]], i32** [[TMP4]], align 8 +// CHECK33-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]]) // CHECK33-NEXT: ret void // // // CHECK33-LABEL: define {{[^@]+}}@.omp_outlined. -// CHECK33-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] { +// CHECK33-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.anon* noalias noundef [[__CONTEXT:%.*]]) #[[ATTR2]] { // CHECK33-NEXT: entry: // CHECK33-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 // CHECK33-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 -// CHECK33-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8 -// CHECK33-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8 -// CHECK33-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8 +// CHECK33-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8 // CHECK33-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 // CHECK33-NEXT: [[TMP:%.*]] = alloca i32, align 4 // CHECK33-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 @@ -3268,82 +3335,84 @@ int main (int argc, char **argv) { // CHECK33-NEXT: [[I3:%.*]] = alloca i32, align 4 // CHECK33-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 // CHECK33-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 -// CHECK33-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8 -// CHECK33-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8 -// CHECK33-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8 -// CHECK33-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8 -// CHECK33-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8 -// CHECK33-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8 -// CHECK33-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4 -// CHECK33-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4 -// CHECK33-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 -// CHECK33-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0 +// CHECK33-NEXT: store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8 +// CHECK33-NEXT: [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8 +// CHECK33-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0 +// CHECK33-NEXT: [[TMP2:%.*]] = load i32*, i32** [[TMP1]], align 8 +// CHECK33-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 1 +// CHECK33-NEXT: [[TMP4:%.*]] = load i64, i64* [[TMP3]], align 8 +// CHECK33-NEXT: [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 2 +// CHECK33-NEXT: [[TMP6:%.*]] = load i32*, i32** [[TMP5]], align 8 +// CHECK33-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP2]], align 4 +// CHECK33-NEXT: store i32 [[TMP7]], i32* [[DOTCAPTURE_EXPR_]], align 4 +// CHECK33-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 +// CHECK33-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP8]], 0 // CHECK33-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 // CHECK33-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 // CHECK33-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4 // CHECK33-NEXT: store i32 0, i32* [[I]], align 4 -// CHECK33-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 -// CHECK33-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]] +// CHECK33-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 +// CHECK33-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP9]] // CHECK33-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] // CHECK33: omp.precond.then: // CHECK33-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 -// CHECK33-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 -// CHECK33-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4 +// CHECK33-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK33-NEXT: store i32 [[TMP10]], i32* [[DOTOMP_UB]], align 4 // CHECK33-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 // CHECK33-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK33-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 -// CHECK33-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4 -// CHECK33-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) -// CHECK33-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 -// CHECK33-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 -// CHECK33-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]] +// CHECK33-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK33-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4 +// CHECK33-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP12]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) +// CHECK33-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK33-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK33-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP13]], [[TMP14]] // CHECK33-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] // CHECK33: cond.true: -// CHECK33-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK33-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 // CHECK33-NEXT: br label [[COND_END:%.*]] // CHECK33: cond.false: -// CHECK33-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK33-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 // CHECK33-NEXT: br label [[COND_END]] // CHECK33: cond.end: -// CHECK33-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ] +// CHECK33-NEXT: [[COND:%.*]] = phi i32 [ [[TMP15]], [[COND_TRUE]] ], [ [[TMP16]], [[COND_FALSE]] ] // CHECK33-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 -// CHECK33-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 -// CHECK33-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4 +// CHECK33-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 +// CHECK33-NEXT: store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4 // CHECK33-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] // CHECK33: omp.inner.for.cond: -// CHECK33-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 -// CHECK33-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !5 -// CHECK33-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]] +// CHECK33-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 +// CHECK33-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !5 +// CHECK33-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]] // CHECK33-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] // CHECK33: omp.inner.for.body: -// CHECK33-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 -// CHECK33-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1 +// CHECK33-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 +// CHECK33-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP20]], 1 // CHECK33-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] // CHECK33-NEXT: store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !5 -// CHECK33-NEXT: [[TMP17:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !5 -// CHECK33-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64 -// CHECK33-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]] +// CHECK33-NEXT: [[TMP21:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !5 +// CHECK33-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP21]] to i64 +// CHECK33-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP6]], i64 [[IDXPROM]] // CHECK33-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !5 // CHECK33-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] // CHECK33: omp.body.continue: // CHECK33-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] // CHECK33: omp.inner.for.inc: -// CHECK33-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 -// CHECK33-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP18]], 1 +// CHECK33-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 +// CHECK33-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP22]], 1 // CHECK33-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 // CHECK33-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP6:![0-9]+]] // CHECK33: omp.inner.for.end: // CHECK33-NEXT: br label [[OMP_LOOP_EXIT:%.*]] // CHECK33: omp.loop.exit: -// CHECK33-NEXT: [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 -// CHECK33-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4 -// CHECK33-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP20]]) -// CHECK33-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK33-NEXT: [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0 -// CHECK33-NEXT: br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] +// CHECK33-NEXT: [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK33-NEXT: [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4 +// CHECK33-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP24]]) +// CHECK33-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 +// CHECK33-NEXT: [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0 +// CHECK33-NEXT: br i1 [[TMP26]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] // CHECK33: .omp.final.then: -// CHECK33-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 -// CHECK33-NEXT: [[SUB7:%.*]] = sub nsw i32 [[TMP23]], 0 +// CHECK33-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 +// CHECK33-NEXT: [[SUB7:%.*]] = sub nsw i32 [[TMP27]], 0 // CHECK33-NEXT: [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1 // CHECK33-NEXT: [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1 // CHECK33-NEXT: [[ADD10:%.*]] = add nsw i32 0, [[MUL9]] @@ -3441,6 +3510,7 @@ int main (int argc, char **argv) { // CHECK33-NEXT: [[TE_ADDR:%.*]] = alloca i64, align 8 // CHECK33-NEXT: [[TH_ADDR:%.*]] = alloca i64, align 8 // CHECK33-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8 +// CHECK33-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 8 // CHECK33-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]]) // CHECK33-NEXT: store i64 [[TE]], i64* [[TE_ADDR]], align 8 // CHECK33-NEXT: store i64 [[TH]], i64* [[TH_ADDR]], align 8 @@ -3451,16 +3521,18 @@ int main (int argc, char **argv) { // CHECK33-NEXT: [[TMP2:%.*]] = load i32, i32* [[CONV]], align 4 // CHECK33-NEXT: [[TMP3:%.*]] = load i32, i32* [[CONV1]], align 4 // CHECK33-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]]) -// CHECK33-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP1]]) +// CHECK33-NEXT: [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 +// CHECK33-NEXT: store [10 x i32]* [[TMP1]], [10 x i32]** [[TMP4]], align 8 +// CHECK33-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]]) // CHECK33-NEXT: ret void // // // CHECK33-LABEL: define {{[^@]+}}@.omp_outlined..1 -// CHECK33-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] { +// CHECK33-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias noundef [[__CONTEXT:%.*]]) #[[ATTR2]] { // CHECK33-NEXT: entry: // CHECK33-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 // CHECK33-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 -// CHECK33-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8 +// CHECK33-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 8 // CHECK33-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 // CHECK33-NEXT: [[TMP:%.*]] = alloca i32, align 4 // CHECK33-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 @@ -3470,58 +3542,60 @@ int main (int argc, char **argv) { // CHECK33-NEXT: [[I:%.*]] = alloca i32, align 4 // CHECK33-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 // CHECK33-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 -// CHECK33-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8 -// CHECK33-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8 +// CHECK33-NEXT: store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 8 +// CHECK33-NEXT: [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 8 +// CHECK33-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0 +// CHECK33-NEXT: [[TMP2:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP1]], align 8 // CHECK33-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 // CHECK33-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4 // CHECK33-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 // CHECK33-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK33-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 -// CHECK33-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4 -// CHECK33-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) -// CHECK33-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 -// CHECK33-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9 +// CHECK33-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK33-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4 +// CHECK33-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) +// CHECK33-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK33-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9 // CHECK33-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] // CHECK33: cond.true: // CHECK33-NEXT: br label [[COND_END:%.*]] // CHECK33: cond.false: -// CHECK33-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK33-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 // CHECK33-NEXT: br label [[COND_END]] // CHECK33: cond.end: -// CHECK33-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ] +// CHECK33-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ] // CHECK33-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 -// CHECK33-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 -// CHECK33-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4 +// CHECK33-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 +// CHECK33-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4 // CHECK33-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] // CHECK33: omp.inner.for.cond: -// CHECK33-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11 -// CHECK33-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !11 -// CHECK33-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]] +// CHECK33-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11 +// CHECK33-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !11 +// CHECK33-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]] // CHECK33-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] // CHECK33: omp.inner.for.body: -// CHECK33-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11 -// CHECK33-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1 +// CHECK33-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11 +// CHECK33-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1 // CHECK33-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] // CHECK33-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !11 -// CHECK33-NEXT: [[TMP9:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !11 -// CHECK33-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP9]] to i64 -// CHECK33-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]] +// CHECK33-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !11 +// CHECK33-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64 +// CHECK33-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP2]], i64 0, i64 [[IDXPROM]] // CHECK33-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !11 // CHECK33-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] // CHECK33: omp.body.continue: // CHECK33-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] // CHECK33: omp.inner.for.inc: -// CHECK33-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11 -// CHECK33-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1 +// CHECK33-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11 +// CHECK33-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1 // CHECK33-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11 // CHECK33-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]] // CHECK33: omp.inner.for.end: // CHECK33-NEXT: br label [[OMP_LOOP_EXIT:%.*]] // CHECK33: omp.loop.exit: -// CHECK33-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]]) -// CHECK33-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK33-NEXT: [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0 -// CHECK33-NEXT: br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] +// CHECK33-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]]) +// CHECK33-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 +// CHECK33-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0 +// CHECK33-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] // CHECK33: .omp.final.then: // CHECK33-NEXT: store i32 10, i32* [[I]], align 4 // CHECK33-NEXT: br label [[DOTOMP_FINAL_DONE]] @@ -3649,23 +3723,28 @@ int main (int argc, char **argv) { // CHECK35-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4 // CHECK35-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4 // CHECK35-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4 +// CHECK35-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4 // CHECK35-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4 // CHECK35-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4 // CHECK35-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4 // CHECK35-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4 // CHECK35-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4 -// CHECK35-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]]) +// CHECK35-NEXT: [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 +// CHECK35-NEXT: store i32* [[N_ADDR]], i32** [[TMP2]], align 4 +// CHECK35-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1 +// CHECK35-NEXT: store i32 [[TMP0]], i32* [[TMP3]], align 4 +// CHECK35-NEXT: [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2 +// CHECK35-NEXT: store i32* [[TMP1]], i32** [[TMP4]], align 4 +// CHECK35-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]]) // CHECK35-NEXT: ret void // // // CHECK35-LABEL: define {{[^@]+}}@.omp_outlined. -// CHECK35-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] { +// CHECK35-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.anon* noalias noundef [[__CONTEXT:%.*]]) #[[ATTR2]] { // CHECK35-NEXT: entry: // CHECK35-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 // CHECK35-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 -// CHECK35-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4 -// CHECK35-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4 -// CHECK35-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4 +// CHECK35-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 4 // CHECK35-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 // CHECK35-NEXT: [[TMP:%.*]] = alloca i32, align 4 // CHECK35-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 @@ -3678,81 +3757,83 @@ int main (int argc, char **argv) { // CHECK35-NEXT: [[I3:%.*]] = alloca i32, align 4 // CHECK35-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 // CHECK35-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 -// CHECK35-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4 -// CHECK35-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4 -// CHECK35-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4 -// CHECK35-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4 -// CHECK35-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4 -// CHECK35-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4 -// CHECK35-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4 -// CHECK35-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4 -// CHECK35-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 -// CHECK35-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0 +// CHECK35-NEXT: store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 4 +// CHECK35-NEXT: [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 4 +// CHECK35-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0 +// CHECK35-NEXT: [[TMP2:%.*]] = load i32*, i32** [[TMP1]], align 4 +// CHECK35-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 1 +// CHECK35-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4 +// CHECK35-NEXT: [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 2 +// CHECK35-NEXT: [[TMP6:%.*]] = load i32*, i32** [[TMP5]], align 4 +// CHECK35-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP2]], align 4 +// CHECK35-NEXT: store i32 [[TMP7]], i32* [[DOTCAPTURE_EXPR_]], align 4 +// CHECK35-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 +// CHECK35-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP8]], 0 // CHECK35-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 // CHECK35-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 // CHECK35-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4 // CHECK35-NEXT: store i32 0, i32* [[I]], align 4 -// CHECK35-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 -// CHECK35-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]] +// CHECK35-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 +// CHECK35-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP9]] // CHECK35-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] // CHECK35: omp.precond.then: // CHECK35-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 -// CHECK35-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 -// CHECK35-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4 +// CHECK35-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK35-NEXT: store i32 [[TMP10]], i32* [[DOTOMP_UB]], align 4 // CHECK35-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 // CHECK35-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK35-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 -// CHECK35-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4 -// CHECK35-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) -// CHECK35-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 -// CHECK35-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 -// CHECK35-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]] +// CHECK35-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 +// CHECK35-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4 +// CHECK35-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP12]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) +// CHECK35-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK35-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK35-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP13]], [[TMP14]] // CHECK35-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] // CHECK35: cond.true: -// CHECK35-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK35-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 // CHECK35-NEXT: br label [[COND_END:%.*]] // CHECK35: cond.false: -// CHECK35-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK35-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 // CHECK35-NEXT: br label [[COND_END]] // CHECK35: cond.end: -// CHECK35-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ] +// CHECK35-NEXT: [[COND:%.*]] = phi i32 [ [[TMP15]], [[COND_TRUE]] ], [ [[TMP16]], [[COND_FALSE]] ] // CHECK35-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 -// CHECK35-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 -// CHECK35-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4 +// CHECK35-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 +// CHECK35-NEXT: store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4 // CHECK35-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] // CHECK35: omp.inner.for.cond: -// CHECK35-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6 -// CHECK35-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !6 -// CHECK35-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]] +// CHECK35-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6 +// CHECK35-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !6 +// CHECK35-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]] // CHECK35-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] // CHECK35: omp.inner.for.body: -// CHECK35-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6 -// CHECK35-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1 +// CHECK35-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6 +// CHECK35-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP20]], 1 // CHECK35-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] // CHECK35-NEXT: store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !6 -// CHECK35-NEXT: [[TMP17:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !6 -// CHECK35-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP17]] +// CHECK35-NEXT: [[TMP21:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !6 +// CHECK35-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP6]], i32 [[TMP21]] // CHECK35-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !6 // CHECK35-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] // CHECK35: omp.body.continue: // CHECK35-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] // CHECK35: omp.inner.for.inc: -// CHECK35-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6 -// CHECK35-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP18]], 1 +// CHECK35-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6 +// CHECK35-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP22]], 1 // CHECK35-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6 // CHECK35-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]] // CHECK35: omp.inner.for.end: // CHECK35-NEXT: br label [[OMP_LOOP_EXIT:%.*]] // CHECK35: omp.loop.exit: -// CHECK35-NEXT: [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 -// CHECK35-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4 -// CHECK35-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP20]]) -// CHECK35-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK35-NEXT: [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0 -// CHECK35-NEXT: br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] +// CHECK35-NEXT: [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 +// CHECK35-NEXT: [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4 +// CHECK35-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP24]]) +// CHECK35-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 +// CHECK35-NEXT: [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0 +// CHECK35-NEXT: br i1 [[TMP26]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] // CHECK35: .omp.final.then: -// CHECK35-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 -// CHECK35-NEXT: [[SUB7:%.*]] = sub nsw i32 [[TMP23]], 0 +// CHECK35-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 +// CHECK35-NEXT: [[SUB7:%.*]] = sub nsw i32 [[TMP27]], 0 // CHECK35-NEXT: [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1 // CHECK35-NEXT: [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1 // CHECK35-NEXT: [[ADD10:%.*]] = add nsw i32 0, [[MUL9]] @@ -3848,6 +3929,7 @@ int main (int argc, char **argv) { // CHECK35-NEXT: [[TE_ADDR:%.*]] = alloca i32, align 4 // CHECK35-NEXT: [[TH_ADDR:%.*]] = alloca i32, align 4 // CHECK35-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4 +// CHECK35-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 4 // CHECK35-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]]) // CHECK35-NEXT: store i32 [[TE]], i32* [[TE_ADDR]], align 4 // CHECK35-NEXT: store i32 [[TH]], i32* [[TH_ADDR]], align 4 @@ -3856,16 +3938,18 @@ int main (int argc, char **argv) { // CHECK35-NEXT: [[TMP2:%.*]] = load i32, i32* [[TE_ADDR]], align 4 // CHECK35-NEXT: [[TMP3:%.*]] = load i32, i32* [[TH_ADDR]], align 4 // CHECK35-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]]) -// CHECK35-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP1]]) +// CHECK35-NEXT: [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 +// CHECK35-NEXT: store [10 x i32]* [[TMP1]], [10 x i32]** [[TMP4]], align 4 +// CHECK35-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]]) // CHECK35-NEXT: ret void // // // CHECK35-LABEL: define {{[^@]+}}@.omp_outlined..1 -// CHECK35-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] { +// CHECK35-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias noundef [[__CONTEXT:%.*]]) #[[ATTR2]] { // CHECK35-NEXT: entry: // CHECK35-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 // CHECK35-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 -// CHECK35-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4 +// CHECK35-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 4 // CHECK35-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 // CHECK35-NEXT: [[TMP:%.*]] = alloca i32, align 4 // CHECK35-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 @@ -3875,57 +3959,59 @@ int main (int argc, char **argv) { // CHECK35-NEXT: [[I:%.*]] = alloca i32, align 4 // CHECK35-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 // CHECK35-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 -// CHECK35-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4 -// CHECK35-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4 +// CHECK35-NEXT: store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 4 +// CHECK35-NEXT: [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 4 +// CHECK35-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0 +// CHECK35-NEXT: [[TMP2:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP1]], align 4 // CHECK35-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 // CHECK35-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4 // CHECK35-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 // CHECK35-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK35-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 -// CHECK35-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4 -// CHECK35-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) -// CHECK35-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 -// CHECK35-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9 +// CHECK35-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 +// CHECK35-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4 +// CHECK35-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) +// CHECK35-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK35-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9 // CHECK35-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] // CHECK35: cond.true: // CHECK35-NEXT: br label [[COND_END:%.*]] // CHECK35: cond.false: -// CHECK35-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK35-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 // CHECK35-NEXT: br label [[COND_END]] // CHECK35: cond.end: -// CHECK35-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ] +// CHECK35-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ] // CHECK35-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 -// CHECK35-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 -// CHECK35-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4 +// CHECK35-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 +// CHECK35-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4 // CHECK35-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] // CHECK35: omp.inner.for.cond: -// CHECK35-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12 -// CHECK35-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !12 -// CHECK35-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]] +// CHECK35-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12 +// CHECK35-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !12 +// CHECK35-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]] // CHECK35-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] // CHECK35: omp.inner.for.body: -// CHECK35-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12 -// CHECK35-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1 +// CHECK35-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12 +// CHECK35-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1 // CHECK35-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] // CHECK35-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !12 -// CHECK35-NEXT: [[TMP9:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !12 -// CHECK35-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP9]] +// CHECK35-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !12 +// CHECK35-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP2]], i32 0, i32 [[TMP11]] // CHECK35-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !12 // CHECK35-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] // CHECK35: omp.body.continue: // CHECK35-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] // CHECK35: omp.inner.for.inc: -// CHECK35-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12 -// CHECK35-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1 +// CHECK35-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12 +// CHECK35-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1 // CHECK35-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12 // CHECK35-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]] // CHECK35: omp.inner.for.end: // CHECK35-NEXT: br label [[OMP_LOOP_EXIT:%.*]] // CHECK35: omp.loop.exit: -// CHECK35-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]]) -// CHECK35-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK35-NEXT: [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0 -// CHECK35-NEXT: br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] +// CHECK35-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]]) +// CHECK35-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 +// CHECK35-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0 +// CHECK35-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] // CHECK35: .omp.final.then: // CHECK35-NEXT: store i32 10, i32* [[I]], align 4 // CHECK35-NEXT: br label [[DOTOMP_FINAL_DONE]] @@ -4069,7 +4155,7 @@ int main (int argc, char **argv) { // CHECK37-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8 // CHECK37-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8 // CHECK37-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1 -// CHECK37-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8 +// CHECK37-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8 // CHECK37-NEXT: store i64 [[ARGC]], i64* [[ARGC_ADDR]], align 8 // CHECK37-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8 // CHECK37-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8 @@ -4082,25 +4168,28 @@ int main (int argc, char **argv) { // CHECK37-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0 // CHECK37-NEXT: [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8 // CHECK37-NEXT: store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1 -// CHECK37-NEXT: [[TMP3:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1 -// CHECK37-NEXT: [[TOBOOL2:%.*]] = trunc i8 [[TMP3]] to i1 -// CHECK37-NEXT: [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i8* -// CHECK37-NEXT: [[FROMBOOL4:%.*]] = zext i1 [[TOBOOL2]] to i8 -// CHECK37-NEXT: store i8 [[FROMBOOL4]], i8* [[CONV3]], align 1 -// CHECK37-NEXT: [[TMP4:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8 -// CHECK37-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*, i64)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[CONV1]], i64 [[TMP0]], i32* [[TMP1]], i64 [[TMP4]]) +// CHECK37-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 +// CHECK37-NEXT: store i32* [[CONV1]], i32** [[TMP3]], align 8 +// CHECK37-NEXT: [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1 +// CHECK37-NEXT: store i64 [[TMP0]], i64* [[TMP4]], align 8 +// CHECK37-NEXT: [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2 +// CHECK37-NEXT: store i32* [[TMP1]], i32** [[TMP5]], align 8 +// CHECK37-NEXT: [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3 +// CHECK37-NEXT: [[TMP7:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1 +// CHECK37-NEXT: [[TOBOOL2:%.*]] = trunc i8 [[TMP7]] to i1 +// CHECK37-NEXT: [[FROMBOOL3:%.*]] = zext i1 [[TOBOOL2]] to i8 +// CHECK37-NEXT: store i8 [[FROMBOOL3]], i8* [[TMP6]], align 8 +// CHECK37-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]]) // CHECK37-NEXT: ret void // // // CHECK37-LABEL: define {{[^@]+}}@.omp_outlined. -// CHECK37-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] { +// CHECK37-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.anon* noalias noundef [[__CONTEXT:%.*]]) #[[ATTR2]] { // CHECK37-NEXT: entry: // CHECK37-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 // CHECK37-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 -// CHECK37-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8 -// CHECK37-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8 -// CHECK37-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8 -// CHECK37-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8 +// CHECK37-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8 +// CHECK37-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1 // CHECK37-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 // CHECK37-NEXT: [[TMP:%.*]] = alloca i32, align 4 // CHECK37-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4 @@ -4113,120 +4202,125 @@ int main (int argc, char **argv) { // CHECK37-NEXT: [[I4:%.*]] = alloca i32, align 4 // CHECK37-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 // CHECK37-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 -// CHECK37-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8 -// CHECK37-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8 -// CHECK37-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8 -// CHECK37-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8 -// CHECK37-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8 -// CHECK37-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8 -// CHECK37-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8 -// CHECK37-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i8* -// CHECK37-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4 -// CHECK37-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4 -// CHECK37-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 -// CHECK37-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0 +// CHECK37-NEXT: store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8 +// CHECK37-NEXT: [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8 +// CHECK37-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0 +// CHECK37-NEXT: [[TMP2:%.*]] = load i32*, i32** [[TMP1]], align 8 +// CHECK37-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 1 +// CHECK37-NEXT: [[TMP4:%.*]] = load i64, i64* [[TMP3]], align 8 +// CHECK37-NEXT: [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 2 +// CHECK37-NEXT: [[TMP6:%.*]] = load i32*, i32** [[TMP5]], align 8 +// CHECK37-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 3 +// CHECK37-NEXT: [[TMP8:%.*]] = load i8, i8* [[TMP7]], align 8 +// CHECK37-NEXT: [[TOBOOL:%.*]] = trunc i8 [[TMP8]] to i1 +// CHECK37-NEXT: [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8 +// CHECK37-NEXT: store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1 +// CHECK37-NEXT: [[TMP9:%.*]] = load i32, i32* [[TMP2]], align 4 +// CHECK37-NEXT: store i32 [[TMP9]], i32* [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK37-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK37-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP10]], 0 // CHECK37-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 // CHECK37-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1 // CHECK37-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4 // CHECK37-NEXT: store i32 0, i32* [[I]], align 4 -// CHECK37-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 -// CHECK37-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]] +// CHECK37-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK37-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP11]] // CHECK37-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] // CHECK37: omp.precond.then: // CHECK37-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 -// CHECK37-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4 -// CHECK37-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4 +// CHECK37-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4 +// CHECK37-NEXT: store i32 [[TMP12]], i32* [[DOTOMP_UB]], align 4 // CHECK37-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 // CHECK37-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK37-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 -// CHECK37-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4 -// CHECK37-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) -// CHECK37-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 -// CHECK37-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4 -// CHECK37-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]] +// CHECK37-NEXT: [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK37-NEXT: [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4 +// CHECK37-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP14]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) +// CHECK37-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK37-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4 +// CHECK37-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP15]], [[TMP16]] // CHECK37-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] // CHECK37: cond.true: -// CHECK37-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4 +// CHECK37-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4 // CHECK37-NEXT: br label [[COND_END:%.*]] // CHECK37: cond.false: -// CHECK37-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK37-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 // CHECK37-NEXT: br label [[COND_END]] // CHECK37: cond.end: -// CHECK37-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ] +// CHECK37-NEXT: [[COND:%.*]] = phi i32 [ [[TMP17]], [[COND_TRUE]] ], [ [[TMP18]], [[COND_FALSE]] ] // CHECK37-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 -// CHECK37-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 -// CHECK37-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4 -// CHECK37-NEXT: [[TMP14:%.*]] = load i8, i8* [[CONV]], align 1 -// CHECK37-NEXT: [[TOBOOL:%.*]] = trunc i8 [[TMP14]] to i1 -// CHECK37-NEXT: br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]] +// CHECK37-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 +// CHECK37-NEXT: store i32 [[TMP19]], i32* [[DOTOMP_IV]], align 4 +// CHECK37-NEXT: [[TMP20:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1 +// CHECK37-NEXT: [[TOBOOL6:%.*]] = trunc i8 [[TMP20]] to i1 +// CHECK37-NEXT: br i1 [[TOBOOL6]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]] // CHECK37: omp_if.then: // CHECK37-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] // CHECK37: omp.inner.for.cond: -// CHECK37-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 -// CHECK37-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !5 -// CHECK37-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]] -// CHECK37-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] +// CHECK37-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 +// CHECK37-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !5 +// CHECK37-NEXT: [[CMP7:%.*]] = icmp sle i32 [[TMP21]], [[TMP22]] +// CHECK37-NEXT: br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] // CHECK37: omp.inner.for.body: -// CHECK37-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 -// CHECK37-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP17]], 1 +// CHECK37-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 +// CHECK37-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP23]], 1 // CHECK37-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] // CHECK37-NEXT: store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !5 -// CHECK37-NEXT: [[TMP18:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !5 -// CHECK37-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP18]] to i64 -// CHECK37-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]] +// CHECK37-NEXT: [[TMP24:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !5 +// CHECK37-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP24]] to i64 +// CHECK37-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP6]], i64 [[IDXPROM]] // CHECK37-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !5 // CHECK37-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] // CHECK37: omp.body.continue: // CHECK37-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] // CHECK37: omp.inner.for.inc: -// CHECK37-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 -// CHECK37-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP19]], 1 -// CHECK37-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 +// CHECK37-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 +// CHECK37-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP25]], 1 +// CHECK37-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5 // CHECK37-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP6:![0-9]+]] // CHECK37: omp.inner.for.end: // CHECK37-NEXT: br label [[OMP_IF_END:%.*]] // CHECK37: omp_if.else: -// CHECK37-NEXT: br label [[OMP_INNER_FOR_COND8:%.*]] -// CHECK37: omp.inner.for.cond8: -// CHECK37-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 -// CHECK37-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 -// CHECK37-NEXT: [[CMP9:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]] -// CHECK37-NEXT: br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY10:%.*]], label [[OMP_INNER_FOR_END18:%.*]] -// CHECK37: omp.inner.for.body10: -// CHECK37-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 -// CHECK37-NEXT: [[MUL11:%.*]] = mul nsw i32 [[TMP22]], 1 -// CHECK37-NEXT: [[ADD12:%.*]] = add nsw i32 0, [[MUL11]] -// CHECK37-NEXT: store i32 [[ADD12]], i32* [[I4]], align 4 -// CHECK37-NEXT: [[TMP23:%.*]] = load i32, i32* [[I4]], align 4 -// CHECK37-NEXT: [[IDXPROM13:%.*]] = sext i32 [[TMP23]] to i64 -// CHECK37-NEXT: [[ARRAYIDX14:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM13]] -// CHECK37-NEXT: store i32 0, i32* [[ARRAYIDX14]], align 4 -// CHECK37-NEXT: br label [[OMP_BODY_CONTINUE15:%.*]] -// CHECK37: omp.body.continue15: -// CHECK37-NEXT: br label [[OMP_INNER_FOR_INC16:%.*]] -// CHECK37: omp.inner.for.inc16: -// CHECK37-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 -// CHECK37-NEXT: [[ADD17:%.*]] = add nsw i32 [[TMP24]], 1 -// CHECK37-NEXT: store i32 [[ADD17]], i32* [[DOTOMP_IV]], align 4 -// CHECK37-NEXT: br label [[OMP_INNER_FOR_COND8]], !llvm.loop [[LOOP9:![0-9]+]] -// CHECK37: omp.inner.for.end18: +// CHECK37-NEXT: br label [[OMP_INNER_FOR_COND9:%.*]] +// CHECK37: omp.inner.for.cond9: +// CHECK37-NEXT: [[TMP26:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 +// CHECK37-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK37-NEXT: [[CMP10:%.*]] = icmp sle i32 [[TMP26]], [[TMP27]] +// CHECK37-NEXT: br i1 [[CMP10]], label [[OMP_INNER_FOR_BODY11:%.*]], label [[OMP_INNER_FOR_END19:%.*]] +// CHECK37: omp.inner.for.body11: +// CHECK37-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 +// CHECK37-NEXT: [[MUL12:%.*]] = mul nsw i32 [[TMP28]], 1 +// CHECK37-NEXT: [[ADD13:%.*]] = add nsw i32 0, [[MUL12]] +// CHECK37-NEXT: store i32 [[ADD13]], i32* [[I4]], align 4 +// CHECK37-NEXT: [[TMP29:%.*]] = load i32, i32* [[I4]], align 4 +// CHECK37-NEXT: [[IDXPROM14:%.*]] = sext i32 [[TMP29]] to i64 +// CHECK37-NEXT: [[ARRAYIDX15:%.*]] = getelementptr inbounds i32, i32* [[TMP6]], i64 [[IDXPROM14]] +// CHECK37-NEXT: store i32 0, i32* [[ARRAYIDX15]], align 4 +// CHECK37-NEXT: br label [[OMP_BODY_CONTINUE16:%.*]] +// CHECK37: omp.body.continue16: +// CHECK37-NEXT: br label [[OMP_INNER_FOR_INC17:%.*]] +// CHECK37: omp.inner.for.inc17: +// CHECK37-NEXT: [[TMP30:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 +// CHECK37-NEXT: [[ADD18:%.*]] = add nsw i32 [[TMP30]], 1 +// CHECK37-NEXT: store i32 [[ADD18]], i32* [[DOTOMP_IV]], align 4 +// CHECK37-NEXT: br label [[OMP_INNER_FOR_COND9]], !llvm.loop [[LOOP9:![0-9]+]] +// CHECK37: omp.inner.for.end19: // CHECK37-NEXT: br label [[OMP_IF_END]] // CHECK37: omp_if.end: // CHECK37-NEXT: br label [[OMP_LOOP_EXIT:%.*]] // CHECK37: omp.loop.exit: -// CHECK37-NEXT: [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 -// CHECK37-NEXT: [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4 -// CHECK37-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]]) -// CHECK37-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK37-NEXT: [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0 -// CHECK37-NEXT: br i1 [[TMP28]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] +// CHECK37-NEXT: [[TMP31:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK37-NEXT: [[TMP32:%.*]] = load i32, i32* [[TMP31]], align 4 +// CHECK37-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP32]]) +// CHECK37-NEXT: [[TMP33:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 +// CHECK37-NEXT: [[TMP34:%.*]] = icmp ne i32 [[TMP33]], 0 +// CHECK37-NEXT: br i1 [[TMP34]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] // CHECK37: .omp.final.then: -// CHECK37-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 -// CHECK37-NEXT: [[SUB19:%.*]] = sub nsw i32 [[TMP29]], 0 -// CHECK37-NEXT: [[DIV20:%.*]] = sdiv i32 [[SUB19]], 1 -// CHECK37-NEXT: [[MUL21:%.*]] = mul nsw i32 [[DIV20]], 1 -// CHECK37-NEXT: [[ADD22:%.*]] = add nsw i32 0, [[MUL21]] -// CHECK37-NEXT: store i32 [[ADD22]], i32* [[I4]], align 4 +// CHECK37-NEXT: [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK37-NEXT: [[SUB20:%.*]] = sub nsw i32 [[TMP35]], 0 +// CHECK37-NEXT: [[DIV21:%.*]] = sdiv i32 [[SUB20]], 1 +// CHECK37-NEXT: [[MUL22:%.*]] = mul nsw i32 [[DIV21]], 1 +// CHECK37-NEXT: [[ADD23:%.*]] = add nsw i32 0, [[MUL22]] +// CHECK37-NEXT: store i32 [[ADD23]], i32* [[I4]], align 4 // CHECK37-NEXT: br label [[DOTOMP_FINAL_DONE]] // CHECK37: .omp.final.done: // CHECK37-NEXT: br label [[OMP_PRECOND_END]] @@ -4320,6 +4414,7 @@ int main (int argc, char **argv) { // CHECK37-NEXT: [[TE_ADDR:%.*]] = alloca i64, align 8 // CHECK37-NEXT: [[TH_ADDR:%.*]] = alloca i64, align 8 // CHECK37-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8 +// CHECK37-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 8 // CHECK37-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]]) // CHECK37-NEXT: store i64 [[TE]], i64* [[TE_ADDR]], align 8 // CHECK37-NEXT: store i64 [[TH]], i64* [[TH_ADDR]], align 8 @@ -4330,16 +4425,18 @@ int main (int argc, char **argv) { // CHECK37-NEXT: [[TMP2:%.*]] = load i32, i32* [[CONV]], align 4 // CHECK37-NEXT: [[TMP3:%.*]] = load i32, i32* [[CONV1]], align 4 // CHECK37-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]]) -// CHECK37-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP1]]) +// CHECK37-NEXT: [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 +// CHECK37-NEXT: store [10 x i32]* [[TMP1]], [10 x i32]** [[TMP4]], align 8 +// CHECK37-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]]) // CHECK37-NEXT: ret void // // // CHECK37-LABEL: define {{[^@]+}}@.omp_outlined..1 -// CHECK37-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] { +// CHECK37-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias noundef [[__CONTEXT:%.*]]) #[[ATTR2]] { // CHECK37-NEXT: entry: // CHECK37-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 // CHECK37-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 -// CHECK37-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8 +// CHECK37-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 8 // CHECK37-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 // CHECK37-NEXT: [[TMP:%.*]] = alloca i32, align 4 // CHECK37-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 @@ -4349,58 +4446,60 @@ int main (int argc, char **argv) { // CHECK37-NEXT: [[I:%.*]] = alloca i32, align 4 // CHECK37-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 // CHECK37-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 -// CHECK37-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8 -// CHECK37-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8 +// CHECK37-NEXT: store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 8 +// CHECK37-NEXT: [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 8 +// CHECK37-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0 +// CHECK37-NEXT: [[TMP2:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP1]], align 8 // CHECK37-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 // CHECK37-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4 // CHECK37-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 // CHECK37-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK37-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 -// CHECK37-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4 -// CHECK37-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) -// CHECK37-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 -// CHECK37-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9 +// CHECK37-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 +// CHECK37-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4 +// CHECK37-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) +// CHECK37-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK37-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9 // CHECK37-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] // CHECK37: cond.true: // CHECK37-NEXT: br label [[COND_END:%.*]] // CHECK37: cond.false: -// CHECK37-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK37-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 // CHECK37-NEXT: br label [[COND_END]] // CHECK37: cond.end: -// CHECK37-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ] +// CHECK37-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ] // CHECK37-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 -// CHECK37-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 -// CHECK37-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4 +// CHECK37-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 +// CHECK37-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4 // CHECK37-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] // CHECK37: omp.inner.for.cond: -// CHECK37-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13 -// CHECK37-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !13 -// CHECK37-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]] +// CHECK37-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13 +// CHECK37-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !13 +// CHECK37-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]] // CHECK37-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] // CHECK37: omp.inner.for.body: -// CHECK37-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13 -// CHECK37-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1 +// CHECK37-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13 +// CHECK37-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1 // CHECK37-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] // CHECK37-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !13 -// CHECK37-NEXT: [[TMP9:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !13 -// CHECK37-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP9]] to i64 -// CHECK37-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]] +// CHECK37-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !13 +// CHECK37-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64 +// CHECK37-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP2]], i64 0, i64 [[IDXPROM]] // CHECK37-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !13 // CHECK37-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] // CHECK37: omp.body.continue: // CHECK37-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] // CHECK37: omp.inner.for.inc: -// CHECK37-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13 -// CHECK37-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1 +// CHECK37-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13 +// CHECK37-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1 // CHECK37-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13 // CHECK37-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]] // CHECK37: omp.inner.for.end: // CHECK37-NEXT: br label [[OMP_LOOP_EXIT:%.*]] // CHECK37: omp.loop.exit: -// CHECK37-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]]) -// CHECK37-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK37-NEXT: [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0 -// CHECK37-NEXT: br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] +// CHECK37-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]]) +// CHECK37-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 +// CHECK37-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0 +// CHECK37-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] // CHECK37: .omp.final.then: // CHECK37-NEXT: store i32 10, i32* [[I]], align 4 // CHECK37-NEXT: br label [[DOTOMP_FINAL_DONE]] @@ -4542,7 +4641,7 @@ int main (int argc, char **argv) { // CHECK39-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4 // CHECK39-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4 // CHECK39-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1 -// CHECK39-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4 +// CHECK39-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4 // CHECK39-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4 // CHECK39-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4 // CHECK39-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4 @@ -4553,25 +4652,28 @@ int main (int argc, char **argv) { // CHECK39-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0 // CHECK39-NEXT: [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8 // CHECK39-NEXT: store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1 -// CHECK39-NEXT: [[TMP3:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1 -// CHECK39-NEXT: [[TOBOOL1:%.*]] = trunc i8 [[TMP3]] to i1 -// CHECK39-NEXT: [[CONV:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__CASTED]] to i8* +// CHECK39-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 +// CHECK39-NEXT: store i32* [[N_ADDR]], i32** [[TMP3]], align 4 +// CHECK39-NEXT: [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1 +// CHECK39-NEXT: store i32 [[TMP0]], i32* [[TMP4]], align 4 +// CHECK39-NEXT: [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2 +// CHECK39-NEXT: store i32* [[TMP1]], i32** [[TMP5]], align 4 +// CHECK39-NEXT: [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3 +// CHECK39-NEXT: [[TMP7:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1 +// CHECK39-NEXT: [[TOBOOL1:%.*]] = trunc i8 [[TMP7]] to i1 // CHECK39-NEXT: [[FROMBOOL2:%.*]] = zext i1 [[TOBOOL1]] to i8 -// CHECK39-NEXT: store i8 [[FROMBOOL2]], i8* [[CONV]], align 1 -// CHECK39-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4 -// CHECK39-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*, i32)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]], i32 [[TMP4]]) +// CHECK39-NEXT: store i8 [[FROMBOOL2]], i8* [[TMP6]], align 4 +// CHECK39-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]]) // CHECK39-NEXT: ret void // // // CHECK39-LABEL: define {{[^@]+}}@.omp_outlined. -// CHECK39-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] { +// CHECK39-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.anon* noalias noundef [[__CONTEXT:%.*]]) #[[ATTR2]] { // CHECK39-NEXT: entry: // CHECK39-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 // CHECK39-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 -// CHECK39-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4 -// CHECK39-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4 -// CHECK39-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4 -// CHECK39-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4 +// CHECK39-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 4 +// CHECK39-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1 // CHECK39-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 // CHECK39-NEXT: [[TMP:%.*]] = alloca i32, align 4 // CHECK39-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4 @@ -4584,118 +4686,123 @@ int main (int argc, char **argv) { // CHECK39-NEXT: [[I4:%.*]] = alloca i32, align 4 // CHECK39-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 // CHECK39-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 -// CHECK39-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4 -// CHECK39-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4 -// CHECK39-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4 -// CHECK39-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4 -// CHECK39-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4 -// CHECK39-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4 -// CHECK39-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4 -// CHECK39-NEXT: [[CONV:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__ADDR]] to i8* -// CHECK39-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4 -// CHECK39-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4 -// CHECK39-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 -// CHECK39-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0 +// CHECK39-NEXT: store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 4 +// CHECK39-NEXT: [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 4 +// CHECK39-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0 +// CHECK39-NEXT: [[TMP2:%.*]] = load i32*, i32** [[TMP1]], align 4 +// CHECK39-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 1 +// CHECK39-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4 +// CHECK39-NEXT: [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 2 +// CHECK39-NEXT: [[TMP6:%.*]] = load i32*, i32** [[TMP5]], align 4 +// CHECK39-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 3 +// CHECK39-NEXT: [[TMP8:%.*]] = load i8, i8* [[TMP7]], align 4 +// CHECK39-NEXT: [[TOBOOL:%.*]] = trunc i8 [[TMP8]] to i1 +// CHECK39-NEXT: [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8 +// CHECK39-NEXT: store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1 +// CHECK39-NEXT: [[TMP9:%.*]] = load i32, i32* [[TMP2]], align 4 +// CHECK39-NEXT: store i32 [[TMP9]], i32* [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK39-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK39-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP10]], 0 // CHECK39-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 // CHECK39-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1 // CHECK39-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4 // CHECK39-NEXT: store i32 0, i32* [[I]], align 4 -// CHECK39-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 -// CHECK39-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]] +// CHECK39-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK39-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP11]] // CHECK39-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] // CHECK39: omp.precond.then: // CHECK39-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 -// CHECK39-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4 -// CHECK39-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4 +// CHECK39-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4 +// CHECK39-NEXT: store i32 [[TMP12]], i32* [[DOTOMP_UB]], align 4 // CHECK39-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 // CHECK39-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK39-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 -// CHECK39-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4 -// CHECK39-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) -// CHECK39-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 -// CHECK39-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4 -// CHECK39-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]] +// CHECK39-NEXT: [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 +// CHECK39-NEXT: [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4 +// CHECK39-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP14]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) +// CHECK39-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK39-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4 +// CHECK39-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP15]], [[TMP16]] // CHECK39-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] // CHECK39: cond.true: -// CHECK39-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4 +// CHECK39-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4 // CHECK39-NEXT: br label [[COND_END:%.*]] // CHECK39: cond.false: -// CHECK39-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK39-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 // CHECK39-NEXT: br label [[COND_END]] // CHECK39: cond.end: -// CHECK39-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ] +// CHECK39-NEXT: [[COND:%.*]] = phi i32 [ [[TMP17]], [[COND_TRUE]] ], [ [[TMP18]], [[COND_FALSE]] ] // CHECK39-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 -// CHECK39-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 -// CHECK39-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4 -// CHECK39-NEXT: [[TMP14:%.*]] = load i8, i8* [[CONV]], align 1 -// CHECK39-NEXT: [[TOBOOL:%.*]] = trunc i8 [[TMP14]] to i1 -// CHECK39-NEXT: br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]] +// CHECK39-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 +// CHECK39-NEXT: store i32 [[TMP19]], i32* [[DOTOMP_IV]], align 4 +// CHECK39-NEXT: [[TMP20:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1 +// CHECK39-NEXT: [[TOBOOL6:%.*]] = trunc i8 [[TMP20]] to i1 +// CHECK39-NEXT: br i1 [[TOBOOL6]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]] // CHECK39: omp_if.then: // CHECK39-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] // CHECK39: omp.inner.for.cond: -// CHECK39-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6 -// CHECK39-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !6 -// CHECK39-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]] -// CHECK39-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] +// CHECK39-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6 +// CHECK39-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !6 +// CHECK39-NEXT: [[CMP7:%.*]] = icmp sle i32 [[TMP21]], [[TMP22]] +// CHECK39-NEXT: br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] // CHECK39: omp.inner.for.body: -// CHECK39-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6 -// CHECK39-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP17]], 1 +// CHECK39-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6 +// CHECK39-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP23]], 1 // CHECK39-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] // CHECK39-NEXT: store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !6 -// CHECK39-NEXT: [[TMP18:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !6 -// CHECK39-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP18]] +// CHECK39-NEXT: [[TMP24:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !6 +// CHECK39-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP6]], i32 [[TMP24]] // CHECK39-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !6 // CHECK39-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] // CHECK39: omp.body.continue: // CHECK39-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] // CHECK39: omp.inner.for.inc: -// CHECK39-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6 -// CHECK39-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP19]], 1 -// CHECK39-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6 +// CHECK39-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6 +// CHECK39-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP25]], 1 +// CHECK39-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6 // CHECK39-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]] // CHECK39: omp.inner.for.end: // CHECK39-NEXT: br label [[OMP_IF_END:%.*]] // CHECK39: omp_if.else: -// CHECK39-NEXT: br label [[OMP_INNER_FOR_COND8:%.*]] -// CHECK39: omp.inner.for.cond8: -// CHECK39-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 -// CHECK39-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 -// CHECK39-NEXT: [[CMP9:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]] -// CHECK39-NEXT: br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY10:%.*]], label [[OMP_INNER_FOR_END17:%.*]] -// CHECK39: omp.inner.for.body10: -// CHECK39-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 -// CHECK39-NEXT: [[MUL11:%.*]] = mul nsw i32 [[TMP22]], 1 -// CHECK39-NEXT: [[ADD12:%.*]] = add nsw i32 0, [[MUL11]] -// CHECK39-NEXT: store i32 [[ADD12]], i32* [[I4]], align 4 -// CHECK39-NEXT: [[TMP23:%.*]] = load i32, i32* [[I4]], align 4 -// CHECK39-NEXT: [[ARRAYIDX13:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP23]] -// CHECK39-NEXT: store i32 0, i32* [[ARRAYIDX13]], align 4 -// CHECK39-NEXT: br label [[OMP_BODY_CONTINUE14:%.*]] -// CHECK39: omp.body.continue14: -// CHECK39-NEXT: br label [[OMP_INNER_FOR_INC15:%.*]] -// CHECK39: omp.inner.for.inc15: -// CHECK39-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 -// CHECK39-NEXT: [[ADD16:%.*]] = add nsw i32 [[TMP24]], 1 -// CHECK39-NEXT: store i32 [[ADD16]], i32* [[DOTOMP_IV]], align 4 -// CHECK39-NEXT: br label [[OMP_INNER_FOR_COND8]], !llvm.loop [[LOOP10:![0-9]+]] -// CHECK39: omp.inner.for.end17: +// CHECK39-NEXT: br label [[OMP_INNER_FOR_COND9:%.*]] +// CHECK39: omp.inner.for.cond9: +// CHECK39-NEXT: [[TMP26:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 +// CHECK39-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK39-NEXT: [[CMP10:%.*]] = icmp sle i32 [[TMP26]], [[TMP27]] +// CHECK39-NEXT: br i1 [[CMP10]], label [[OMP_INNER_FOR_BODY11:%.*]], label [[OMP_INNER_FOR_END18:%.*]] +// CHECK39: omp.inner.for.body11: +// CHECK39-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 +// CHECK39-NEXT: [[MUL12:%.*]] = mul nsw i32 [[TMP28]], 1 +// CHECK39-NEXT: [[ADD13:%.*]] = add nsw i32 0, [[MUL12]] +// CHECK39-NEXT: store i32 [[ADD13]], i32* [[I4]], align 4 +// CHECK39-NEXT: [[TMP29:%.*]] = load i32, i32* [[I4]], align 4 +// CHECK39-NEXT: [[ARRAYIDX14:%.*]] = getelementptr inbounds i32, i32* [[TMP6]], i32 [[TMP29]] +// CHECK39-NEXT: store i32 0, i32* [[ARRAYIDX14]], align 4 +// CHECK39-NEXT: br label [[OMP_BODY_CONTINUE15:%.*]] +// CHECK39: omp.body.continue15: +// CHECK39-NEXT: br label [[OMP_INNER_FOR_INC16:%.*]] +// CHECK39: omp.inner.for.inc16: +// CHECK39-NEXT: [[TMP30:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 +// CHECK39-NEXT: [[ADD17:%.*]] = add nsw i32 [[TMP30]], 1 +// CHECK39-NEXT: store i32 [[ADD17]], i32* [[DOTOMP_IV]], align 4 +// CHECK39-NEXT: br label [[OMP_INNER_FOR_COND9]], !llvm.loop [[LOOP10:![0-9]+]] +// CHECK39: omp.inner.for.end18: // CHECK39-NEXT: br label [[OMP_IF_END]] // CHECK39: omp_if.end: // CHECK39-NEXT: br label [[OMP_LOOP_EXIT:%.*]] // CHECK39: omp.loop.exit: -// CHECK39-NEXT: [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 -// CHECK39-NEXT: [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4 -// CHECK39-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]]) -// CHECK39-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK39-NEXT: [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0 -// CHECK39-NEXT: br i1 [[TMP28]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] +// CHECK39-NEXT: [[TMP31:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 +// CHECK39-NEXT: [[TMP32:%.*]] = load i32, i32* [[TMP31]], align 4 +// CHECK39-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP32]]) +// CHECK39-NEXT: [[TMP33:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 +// CHECK39-NEXT: [[TMP34:%.*]] = icmp ne i32 [[TMP33]], 0 +// CHECK39-NEXT: br i1 [[TMP34]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] // CHECK39: .omp.final.then: -// CHECK39-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 -// CHECK39-NEXT: [[SUB18:%.*]] = sub nsw i32 [[TMP29]], 0 -// CHECK39-NEXT: [[DIV19:%.*]] = sdiv i32 [[SUB18]], 1 -// CHECK39-NEXT: [[MUL20:%.*]] = mul nsw i32 [[DIV19]], 1 -// CHECK39-NEXT: [[ADD21:%.*]] = add nsw i32 0, [[MUL20]] -// CHECK39-NEXT: store i32 [[ADD21]], i32* [[I4]], align 4 +// CHECK39-NEXT: [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 +// CHECK39-NEXT: [[SUB19:%.*]] = sub nsw i32 [[TMP35]], 0 +// CHECK39-NEXT: [[DIV20:%.*]] = sdiv i32 [[SUB19]], 1 +// CHECK39-NEXT: [[MUL21:%.*]] = mul nsw i32 [[DIV20]], 1 +// CHECK39-NEXT: [[ADD22:%.*]] = add nsw i32 0, [[MUL21]] +// CHECK39-NEXT: store i32 [[ADD22]], i32* [[I4]], align 4 // CHECK39-NEXT: br label [[DOTOMP_FINAL_DONE]] // CHECK39: .omp.final.done: // CHECK39-NEXT: br label [[OMP_PRECOND_END]] @@ -4787,6 +4894,7 @@ int main (int argc, char **argv) { // CHECK39-NEXT: [[TE_ADDR:%.*]] = alloca i32, align 4 // CHECK39-NEXT: [[TH_ADDR:%.*]] = alloca i32, align 4 // CHECK39-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4 +// CHECK39-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 4 // CHECK39-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]]) // CHECK39-NEXT: store i32 [[TE]], i32* [[TE_ADDR]], align 4 // CHECK39-NEXT: store i32 [[TH]], i32* [[TH_ADDR]], align 4 @@ -4795,16 +4903,18 @@ int main (int argc, char **argv) { // CHECK39-NEXT: [[TMP2:%.*]] = load i32, i32* [[TE_ADDR]], align 4 // CHECK39-NEXT: [[TMP3:%.*]] = load i32, i32* [[TH_ADDR]], align 4 // CHECK39-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]]) -// CHECK39-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP1]]) +// CHECK39-NEXT: [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 +// CHECK39-NEXT: store [10 x i32]* [[TMP1]], [10 x i32]** [[TMP4]], align 4 +// CHECK39-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]]) // CHECK39-NEXT: ret void // // // CHECK39-LABEL: define {{[^@]+}}@.omp_outlined..1 -// CHECK39-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] { +// CHECK39-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias noundef [[__CONTEXT:%.*]]) #[[ATTR2]] { // CHECK39-NEXT: entry: // CHECK39-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 // CHECK39-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 -// CHECK39-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4 +// CHECK39-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 4 // CHECK39-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 // CHECK39-NEXT: [[TMP:%.*]] = alloca i32, align 4 // CHECK39-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 @@ -4814,57 +4924,59 @@ int main (int argc, char **argv) { // CHECK39-NEXT: [[I:%.*]] = alloca i32, align 4 // CHECK39-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 // CHECK39-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 -// CHECK39-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4 -// CHECK39-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4 +// CHECK39-NEXT: store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 4 +// CHECK39-NEXT: [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 4 +// CHECK39-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0 +// CHECK39-NEXT: [[TMP2:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP1]], align 4 // CHECK39-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 // CHECK39-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4 // CHECK39-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 // CHECK39-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK39-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 -// CHECK39-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4 -// CHECK39-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) -// CHECK39-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 -// CHECK39-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9 +// CHECK39-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 +// CHECK39-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4 +// CHECK39-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) +// CHECK39-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK39-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9 // CHECK39-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] // CHECK39: cond.true: // CHECK39-NEXT: br label [[COND_END:%.*]] // CHECK39: cond.false: -// CHECK39-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 +// CHECK39-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 // CHECK39-NEXT: br label [[COND_END]] // CHECK39: cond.end: -// CHECK39-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ] +// CHECK39-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ] // CHECK39-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 -// CHECK39-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 -// CHECK39-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4 +// CHECK39-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 +// CHECK39-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4 // CHECK39-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] // CHECK39: omp.inner.for.cond: -// CHECK39-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14 -// CHECK39-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !14 -// CHECK39-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]] +// CHECK39-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14 +// CHECK39-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !14 +// CHECK39-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]] // CHECK39-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] // CHECK39: omp.inner.for.body: -// CHECK39-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14 -// CHECK39-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1 +// CHECK39-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14 +// CHECK39-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1 // CHECK39-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] // CHECK39-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !14 -// CHECK39-NEXT: [[TMP9:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !14 -// CHECK39-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP9]] +// CHECK39-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !14 +// CHECK39-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP2]], i32 0, i32 [[TMP11]] // CHECK39-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !14 // CHECK39-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] // CHECK39: omp.body.continue: // CHECK39-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] // CHECK39: omp.inner.for.inc: -// CHECK39-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14 -// CHECK39-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1 +// CHECK39-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14 +// CHECK39-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1 // CHECK39-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14 // CHECK39-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]] // CHECK39: omp.inner.for.end: // CHECK39-NEXT: br label [[OMP_LOOP_EXIT:%.*]] // CHECK39: omp.loop.exit: -// CHECK39-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]]) -// CHECK39-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 -// CHECK39-NEXT: [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0 -// CHECK39-NEXT: br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] +// CHECK39-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]]) +// CHECK39-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4 +// CHECK39-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0 +// CHECK39-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]] // CHECK39: .omp.final.then: // CHECK39-NEXT: store i32 10, i32* [[I]], align 4 // CHECK39-NEXT: br label [[DOTOMP_FINAL_DONE]] |