[OPENMP 4.5] Fixed codegen for 'priority' and destructors in task-based

directives.

'kmp_task_t' record type added a new field for 'priority' clause and
changed the representation of pointer to destructors for privates used
within loop-based directives.
Old representation:

typedef struct kmp_task {                   /* GEH: Shouldn't this be
aligned somehow? */
  void *shareds;                            /**< pointer to block of
    pointers to shared vars   */
  kmp_routine_entry_t routine;              /**< pointer to routine
    to call for executing task */
  kmp_int32 part_id;                        /**< part id for the
    task                          */
  kmp_routine_entry_t destructors;        /* pointer to function to
  invoke deconstructors of firstprivate C++ objects */
  /*  private vars  */
} kmp_task_t;

New representation:

typedef struct kmp_task {                   /* GEH: Shouldn't this be
aligned somehow? */
  void *shareds;                            /**< pointer to block of
    pointers to shared vars   */
  kmp_routine_entry_t routine;              /**< pointer to routine
    to call for executing task */
  kmp_int32 part_id;                        /**< part id for the
    task                          */
  kmp_cmplrdata_t data1; /* Two known
optional additions: destructors and priority */
  kmp_cmplrdata_t data2; /* Process
destructors first, priority second */
/* future data */
  /*  private vars  */
} kmp_task_t;

Also excessive initialization of 'destructors' fields to 'null' was
removed from codegen if it is known that no destructors shal be used.
Currently a special bit is used in 'kmp_tasking_flags_t' bitfields
('destructors_thunk' bitfield).

llvm-svn: 271201

1 files changed, 32 insertions, 40 deletions

diff --git a/clang/test/OpenMP/taskloop_codegen.cpp b/clang/test/OpenMP/taskloop_codegen.cpp
index fff3ccd71d7..e585fcec10d 100644
--- a/clang/test/OpenMP/taskloop_codegen.cpp
+++ b/clang/test/OpenMP/taskloop_codegen.cpp
@@ -9,32 +9,28 @@
 // CHECK-LABEL: @main
 int main(int argc, char **argv) {
 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num(%ident_t* [[DEFLOC:@.+]])
-// CHECK: [[TASKV:%.+]] = call i8* @__kmpc_omp_task_alloc(%ident_t* [[DEFLOC]], i32 [[GTID]], i32 33, i64 64, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, [[TDP_TY:%.+]]*)* [[TASK1:@.+]] to i32 (i32, i8*)*))
+// CHECK: [[TASKV:%.+]] = call i8* @__kmpc_omp_task_alloc(%ident_t* [[DEFLOC]], i32 [[GTID]], i32 33, i64 72, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, [[TDP_TY:%.+]]*)* [[TASK1:@.+]] to i32 (i32, i8*)*))
 // CHECK: [[TASK:%.+]] = bitcast i8* [[TASKV]] to [[TDP_TY]]*
 // CHECK: [[TASK_DATA:%.+]] = getelementptr inbounds [[TDP_TY]], [[TDP_TY]]* [[TASK]], i32 0, i32 0
-// CHECK: getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 3
-// CHECK: store i32 (i32, i8*)* null, i32 (i32, i8*)** %{{.+}}
-// CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 4
+// CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 5
 // CHECK: store i64 0, i64* [[DOWN]],
-// CHECK: [[UP:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 5
+// CHECK: [[UP:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 6
 // CHECK: store i64 9, i64* [[UP]],
-// CHECK: [[ST:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 6
+// CHECK: [[ST:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 7
 // CHECK: store i64 1, i64* [[ST]],
 // CHECK: [[ST_VAL:%.+]] = load i64, i64* [[ST]],
 // CHECK: call void @__kmpc_taskloop(%ident_t* [[DEFLOC]], i32 [[GTID]], i8* [[TASKV]], i32 1, i64* [[DOWN]], i64* [[UP]], i64 [[ST_VAL]], i32 0, i32 0, i64 0, i8* null)
 #pragma omp taskloop priority(argc)
   for (int i = 0; i < 10; ++i)
     ;
-// CHECK: [[TASKV:%.+]] = call i8* @__kmpc_omp_task_alloc(%ident_t* [[DEFLOC]], i32 [[GTID]], i32 1, i64 64, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, [[TDP_TY:%.+]]*)* [[TASK2:@.+]] to i32 (i32, i8*)*))
+// CHECK: [[TASKV:%.+]] = call i8* @__kmpc_omp_task_alloc(%ident_t* [[DEFLOC]], i32 [[GTID]], i32 1, i64 72, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, [[TDP_TY:%.+]]*)* [[TASK2:@.+]] to i32 (i32, i8*)*))
 // CHECK: [[TASK:%.+]] = bitcast i8* [[TASKV]] to [[TDP_TY]]*
 // CHECK: [[TASK_DATA:%.+]] = getelementptr inbounds [[TDP_TY]], [[TDP_TY]]* [[TASK]], i32 0, i32 0
-// CHECK: getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 3
-// CHECK: store i32 (i32, i8*)* null, i32 (i32, i8*)** %{{.+}}
-// CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 4
+// CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 5
 // CHECK: store i64 0, i64* [[DOWN]],
-// CHECK: [[UP:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 5
+// CHECK: [[UP:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 6
 // CHECK: store i64 9, i64* [[UP]],
-// CHECK: [[ST:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 6
+// CHECK: [[ST:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 7
 // CHECK: store i64 1, i64* [[ST]],
 // CHECK: [[ST_VAL:%.+]] = load i64, i64* [[ST]],
 // CHECK: [[GRAINSIZE:%.+]] = zext i32 %{{.+}} to i64
@@ -42,18 +38,16 @@ int main(int argc, char **argv) {
 #pragma omp taskloop nogroup grainsize(argc)
   for (int i = 0; i < 10; ++i)
     ;
-// CHECK: [[TASKV:%.+]] = call i8* @__kmpc_omp_task_alloc(%ident_t* [[DEFLOC]], i32 [[GTID]], i32 1, i64 64, i64 24, i32 (i32, i8*)* bitcast (i32 (i32, [[TDP_TY:%.+]]*)* [[TASK3:@.+]] to i32 (i32, i8*)*))
+// CHECK: [[TASKV:%.+]] = call i8* @__kmpc_omp_task_alloc(%ident_t* [[DEFLOC]], i32 [[GTID]], i32 1, i64 72, i64 24, i32 (i32, i8*)* bitcast (i32 (i32, [[TDP_TY:%.+]]*)* [[TASK3:@.+]] to i32 (i32, i8*)*))
 // CHECK: [[TASK:%.+]] = bitcast i8* [[TASKV]] to [[TDP_TY]]*
 // CHECK: [[TASK_DATA:%.+]] = getelementptr inbounds [[TDP_TY]], [[TDP_TY]]* [[TASK]], i32 0, i32 0
-// CHECK: getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 3
-// CHECK: store i32 (i32, i8*)* null, i32 (i32, i8*)** %{{.+}}
 // CHECK: [[IF:%.+]] = icmp ne i32 %{{.+}}, 0
 // CHECK: [[IF_INT:%.+]] = sext i1 [[IF]] to i32
-// CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 4
+// CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 5
 // CHECK: store i64 0, i64* [[DOWN]],
-// CHECK: [[UP:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 5
+// CHECK: [[UP:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 6
 // CHECK: store i64 %{{.+}}, i64* [[UP]],
-// CHECK: [[ST:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 6
+// CHECK: [[ST:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 7
 // CHECK: store i64 1, i64* [[ST]],
 // CHECK: [[ST_VAL:%.+]] = load i64, i64* [[ST]],
 // CHECK: call void @__kmpc_taskloop(%ident_t* [[DEFLOC]], i32 [[GTID]], i8* [[TASKV]], i32 [[IF_INT]], i64* [[DOWN]], i64* [[UP]], i64 [[ST_VAL]], i32 0, i32 2, i64 4, i8* null)
@@ -65,13 +59,13 @@ int main(int argc, char **argv) {
 }
 
 // CHECK: define internal i32 [[TASK1]](
-// CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* %{{.+}}, i32 0, i32 4
+// CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* %{{.+}}, i32 0, i32 5
 // CHECK: [[DOWN_VAL:%.+]] = load i64, i64* [[DOWN]],
-// CHECK: [[UP:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 5
+// CHECK: [[UP:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 6
 // CHECK: [[UP_VAL:%.+]] = load i64, i64* [[UP]],
-// CHECK: [[ST:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 6
+// CHECK: [[ST:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 7
 // CHECK: [[ST_VAL:%.+]] = load i64, i64* [[ST]],
-// CHECK: [[LITER:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 7
+// CHECK: [[LITER:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 8
 // CHECK: [[LITER_VAL:%.+]] = load i32, i32* [[LITER]],
 // CHECK: store i64 [[DOWN_VAL]], i64* [[LB:%[^,]+]],
 // CHECK: store i64 [[UP_VAL]], i64* [[UB:%[^,]+]],
@@ -95,13 +89,13 @@ int main(int argc, char **argv) {
 // CHECK: ret i32 0
 
 // CHECK: define internal i32 [[TASK2]](
-// CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* %{{.+}}, i32 0, i32 4
+// CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* %{{.+}}, i32 0, i32 5
 // CHECK: [[DOWN_VAL:%.+]] = load i64, i64* [[DOWN]],
-// CHECK: [[UP:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 5
+// CHECK: [[UP:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 6
 // CHECK: [[UP_VAL:%.+]] = load i64, i64* [[UP]],
-// CHECK: [[ST:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 6
+// CHECK: [[ST:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 7
 // CHECK: [[ST_VAL:%.+]] = load i64, i64* [[ST]],
-// CHECK: [[LITER:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 7
+// CHECK: [[LITER:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 8
 // CHECK: [[LITER_VAL:%.+]] = load i32, i32* [[LITER]],
 // CHECK: store i64 [[DOWN_VAL]], i64* [[LB:%[^,]+]],
 // CHECK: store i64 [[UP_VAL]], i64* [[UB:%[^,]+]],
@@ -125,13 +119,13 @@ int main(int argc, char **argv) {
 // CHECK: ret i32 0
 
 // CHECK: define internal i32 [[TASK3]](
-// CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* %{{.+}}, i32 0, i32 4
+// CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* %{{.+}}, i32 0, i32 5
 // CHECK: [[DOWN_VAL:%.+]] = load i64, i64* [[DOWN]],
-// CHECK: [[UP:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 5
+// CHECK: [[UP:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 6
 // CHECK: [[UP_VAL:%.+]] = load i64, i64* [[UP]],
-// CHECK: [[ST:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 6
+// CHECK: [[ST:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 7
 // CHECK: [[ST_VAL:%.+]] = load i64, i64* [[ST]],
-// CHECK: [[LITER:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 7
+// CHECK: [[LITER:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 8
 // CHECK: [[LITER_VAL:%.+]] = load i32, i32* [[LITER]],
 // CHECK: store i64 [[DOWN_VAL]], i64* [[LB:%[^,]+]],
 // CHECK: store i64 [[UP_VAL]], i64* [[UB:%[^,]+]],
@@ -147,16 +141,14 @@ struct S {
   int a;
   S(int c) {
 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num(%ident_t* [[DEFLOC:@.+]])
-// CHECK: [[TASKV:%.+]] = call i8* @__kmpc_omp_task_alloc(%ident_t* [[DEFLOC]], i32 [[GTID]], i32 1, i64 64, i64 16, i32 (i32, i8*)* bitcast (i32 (i32, [[TDP_TY:%.+]]*)* [[TASK4:@.+]] to i32 (i32, i8*)*))
+// CHECK: [[TASKV:%.+]] = call i8* @__kmpc_omp_task_alloc(%ident_t* [[DEFLOC]], i32 [[GTID]], i32 1, i64 72, i64 16, i32 (i32, i8*)* bitcast (i32 (i32, [[TDP_TY:%.+]]*)* [[TASK4:@.+]] to i32 (i32, i8*)*))
 // CHECK: [[TASK:%.+]] = bitcast i8* [[TASKV]] to [[TDP_TY]]*
 // CHECK: [[TASK_DATA:%.+]] = getelementptr inbounds [[TDP_TY]], [[TDP_TY]]* [[TASK]], i32 0, i32 0
-// CHECK: getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 3
-// CHECK: store i32 (i32, i8*)* null, i32 (i32, i8*)** %{{.+}}
-// CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 4
+// CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 5
 // CHECK: store i64 0, i64* [[DOWN]],
-// CHECK: [[UP:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 5
+// CHECK: [[UP:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 6
 // CHECK: store i64 %{{.+}}, i64* [[UP]],
-// CHECK: [[ST:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 6
+// CHECK: [[ST:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 7
 // CHECK: store i64 1, i64* [[ST]],
 // CHECK: [[ST_VAL:%.+]] = load i64, i64* [[ST]],
 // CHECK: [[NUM_TASKS:%.+]] = zext i32 %{{.+}} to i64
@@ -168,13 +160,13 @@ struct S {
 } s(1);
 
 // CHECK: define internal i32 [[TASK4]](
-// CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* %{{.+}}, i32 0, i32 4
+// CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* %{{.+}}, i32 0, i32 5
 // CHECK: [[DOWN_VAL:%.+]] = load i64, i64* [[DOWN]],
-// CHECK: [[UP:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 5
+// CHECK: [[UP:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 6
 // CHECK: [[UP_VAL:%.+]] = load i64, i64* [[UP]],
-// CHECK: [[ST:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 6
+// CHECK: [[ST:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 7
 // CHECK: [[ST_VAL:%.+]] = load i64, i64* [[ST]],
-// CHECK: [[LITER:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 7
+// CHECK: [[LITER:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 8
 // CHECK: [[LITER_VAL:%.+]] = load i32, i32* [[LITER]],
 // CHECK: store i64 [[DOWN_VAL]], i64* [[LB:%[^,]+]],
 // CHECK: store i64 [[UP_VAL]], i64* [[UB:%[^,]+]],