commit 2377fe95c6beeba5cbba356fee7fba32257bbce2
author Alexey Bataev <a.bataev@hotmail.com> Thu Sep 10 08:12:02 2015 +0000
committer Alexey Bataev <a.bataev@hotmail.com> Thu Sep 10 08:12:02 2015 +0000
tree b853c3ba0847ca5371ae261d956726aa9d99223e
parent f054eca167ea834a168704ff7065a149d75e8f0f

[OPENMP] Outlined function for parallel and other regions with list of captured variables.
Currently all variables used in OpenMP regions are captured into a record and passed to outlined functions in this record. It may result in some poor performance because of too complex analysis later in optimization passes. Patch makes to emit outlined functions for parallel-based regions with a list of captured variables. It reduces code for 2*n GEPs, stores and loads at least.
Codegen for task-based regions remains unchanged because runtime requires that all captured variables are passed in captured record.

llvm-svn: 247251

clang/test/OpenMP/task_private_codegen.cpp

diff --git a/clang/test/OpenMP/task_private_codegen.cpp b/clang/test/OpenMP/task_private_codegen.cpp
index b29d0d3..63ca915 100644
--- a/clang/test/OpenMP/task_private_codegen.cpp
+++ b/clang/test/OpenMP/task_private_codegen.cpp
@@ -69,7 +69,7 @@
     // LAMBDA: [[G_REF:%.+]] = load double*, double** [[G_PTR_REF]]
     // LAMBDA: store double 2.0{{.+}}, double* [[G_REF]]
 
-    // LAMBDA: define internal i32 [[TASK_ENTRY]](i32, %{{.+}}*)
+    // LAMBDA: define internal i32 [[TASK_ENTRY]](i32, %{{.+}}* noalias)
     g = 1;
     // LAMBDA: store double 1.0{{.+}}, double* %{{.+}},
     // LAMBDA: call void [[INNER_LAMBDA]](%
@@ -99,7 +99,7 @@
     // BLOCKS-NOT: [[G]]{{[[^:word:]]}}
     // BLOCKS: ret
 
-    // BLOCKS: define internal i32 [[TASK_ENTRY]](i32, %{{.+}}*)
+    // BLOCKS: define internal i32 [[TASK_ENTRY]](i32, %{{.+}}* noalias)
     g = 1;
     // BLOCKS: store double 1.0{{.+}}, double* %{{.+}},
     // BLOCKS-NOT: [[G]]{{[[^:word:]]}}
@@ -199,7 +199,7 @@
 // CHECK: store [2 x i32]* [[PRIV_VEC]], [2 x i32]** [[ARG4]],
 // CHECK: ret void
 
-// CHECK: define internal i32 [[TASK_ENTRY]](i32, [[KMP_TASK_MAIN_TY]]*)
+// CHECK: define internal i32 [[TASK_ENTRY]](i32, [[KMP_TASK_MAIN_TY]]* noalias)
 
 // CHECK: [[PRIV_VAR_ADDR:%.+]] = alloca [[S_DOUBLE_TY]]*,
 // CHECK: [[PRIV_T_VAR_ADDR:%.+]] = alloca i32*,
@@ -221,7 +221,7 @@
 
 // CHECK: ret
 
-// CHECK: define internal i32 [[DESTRUCTORS]](i32, [[KMP_TASK_MAIN_TY]]*)
+// CHECK: define internal i32 [[DESTRUCTORS]](i32, [[KMP_TASK_MAIN_TY]]* noalias)
 // CHECK: [[PRIVATES:%.+]] = getelementptr inbounds [[KMP_TASK_MAIN_TY]], [[KMP_TASK_MAIN_TY]]* [[RES_KMP_TASK:%.+]], i{{[0-9]+}} 0, i{{[0-9]+}} 1
 // CHECK: [[PRIVATE_VAR_REF:%.+]] = getelementptr inbounds [[PRIVATES_MAIN_TY]], [[PRIVATES_MAIN_TY]]* [[PRIVATES]], i{{.+}} 0, i{{.+}} 0
 // CHECK: [[PRIVATE_S_ARR_REF:%.+]] = getelementptr inbounds [[PRIVATES_MAIN_TY]], [[PRIVATES_MAIN_TY]]* [[PRIVATES]], i{{.+}} 0, i{{.+}} 1
@@ -306,7 +306,7 @@
 // CHECK: store [[S_INT_TY]]* [[PRIV_VAR]], [[S_INT_TY]]** [[ARG4]],
 // CHECK: ret void
 
-// CHECK: define internal i32 [[TASK_ENTRY]](i32, [[KMP_TASK_TMAIN_TY]]*)
+// CHECK: define internal i32 [[TASK_ENTRY]](i32, [[KMP_TASK_TMAIN_TY]]* noalias)
 
 // CHECK: [[PRIV_T_VAR_ADDR:%.+]] = alloca i32*,
 // CHECK: [[PRIV_VEC_ADDR:%.+]] = alloca [2 x i32]*,
@@ -328,7 +328,7 @@
 
 // CHECK: ret
 
-// CHECK: define internal i32 [[DESTRUCTORS]](i32, [[KMP_TASK_TMAIN_TY]]*)
+// CHECK: define internal i32 [[DESTRUCTORS]](i32, [[KMP_TASK_TMAIN_TY]]* noalias)
 // CHECK: [[PRIVATES:%.+]] = getelementptr inbounds [[KMP_TASK_TMAIN_TY]], [[KMP_TASK_TMAIN_TY]]* [[RES_KMP_TASK:%.+]], i{{[0-9]+}} 0, i{{[0-9]+}} 1
 // CHECK: [[PRIVATE_S_ARR_REF:%.+]] = getelementptr inbounds [[PRIVATES_TMAIN_TY]], [[PRIVATES_TMAIN_TY]]* [[PRIVATES]], i{{.+}} 0, i{{.+}} 2
 // CHECK: [[PRIVATE_VAR_REF:%.+]] = getelementptr inbounds [[PRIVATES_TMAIN_TY]], [[PRIVATES_TMAIN_TY]]* [[PRIVATES]], i{{.+}} 0, i{{.+}} 3