maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/clang/test/OpenMP/parallel_copyin_codegen.cpp

527 lines
30 KiB
C++
Raw Normal View History

[OpenMP] Add TLS-based implementation for threadprivate directive. llvm-svn: 242080
2015-07-14 06:54:53 +08:00
// RUN: %clang_cc1 -verify -fopenmp -fnoopenmp-use-tls -x c++ -triple %itanium_abi_triple -emit-llvm %s -o - | FileCheck %s
// RUN: %clang_cc1 -fopenmp -fnoopenmp-use-tls -x c++ -std=c++11 -triple %itanium_abi_triple -emit-pch -o %t %s
// RUN: %clang_cc1 -fopenmp -fnoopenmp-use-tls -x c++ -triple %itanium_abi_triple -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s
// RUN: %clang_cc1 -verify -fopenmp -fnoopenmp-use-tls -x c++ -std=c++11 -DLAMBDA -triple %itanium_abi_triple -emit-llvm %s -o - | FileCheck -check-prefix=LAMBDA %s
// RUN: %clang_cc1 -verify -fopenmp -fnoopenmp-use-tls -x c++ -fblocks -DBLOCKS -triple %itanium_abi_triple -emit-llvm %s -o - | FileCheck -check-prefix=BLOCKS %s
// RUN: %clang_cc1 -verify -fopenmp -fnoopenmp-use-tls -x c++ -std=c++11 -DARRAY -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck -check-prefix=ARRAY %s
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
[OPENMP] Support for -fopenmp-simd option with compilation of simd loops only. Added support for -fopenmp-simd option that allows compilation of simd-based constructs without emission of OpenMP runtime calls. llvm-svn: 321560
2017-12-30 02:07:07 +08:00
// RUN: %clang_cc1 -verify -fopenmp-simd -fnoopenmp-use-tls -x c++ -triple %itanium_abi_triple -emit-llvm %s -o - | FileCheck --check-prefix SIMD-ONLY0 %s
// RUN: %clang_cc1 -fopenmp-simd -fnoopenmp-use-tls -x c++ -std=c++11 -triple %itanium_abi_triple -emit-pch -o %t %s
// RUN: %clang_cc1 -fopenmp-simd -fnoopenmp-use-tls -x c++ -triple %itanium_abi_triple -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck --check-prefix SIMD-ONLY0 %s
// RUN: %clang_cc1 -verify -fopenmp-simd -fnoopenmp-use-tls -x c++ -std=c++11 -DLAMBDA -triple %itanium_abi_triple -emit-llvm %s -o - | FileCheck --check-prefix SIMD-ONLY0 %s
// RUN: %clang_cc1 -verify -fopenmp-simd -fnoopenmp-use-tls -x c++ -fblocks -DBLOCKS -triple %itanium_abi_triple -emit-llvm %s -o - | FileCheck --check-prefix SIMD-ONLY0 %s
// RUN: %clang_cc1 -verify -fopenmp-simd -fnoopenmp-use-tls -x c++ -std=c++11 -DARRAY -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck --check-prefix SIMD-ONLY0 %s
// SIMD-ONLY0-NOT: {{__kmpc|__tgt}}
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// RUN: %clang_cc1 -verify -fopenmp -x c++ -triple %itanium_abi_triple -emit-llvm %s -o - | FileCheck %s -check-prefix=TLS-CHECK
// RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple %itanium_abi_triple -emit-pch -o %t %s
// RUN: %clang_cc1 -fopenmp -x c++ -triple %itanium_abi_triple -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s -check-prefix=TLS-CHECK
// RUN: %clang_cc1 -verify -fopenmp -x c++ -std=c++11 -DLAMBDA -triple %itanium_abi_triple -emit-llvm %s -o - | FileCheck -check-prefix=TLS-LAMBDA %s
// RUN: %clang_cc1 -verify -fopenmp -x c++ -fblocks -DBLOCKS -triple %itanium_abi_triple -emit-llvm %s -o - | FileCheck -check-prefix=TLS-BLOCKS %s
// RUN: %clang_cc1 -verify -fopenmp -x c++ -std=c++11 -DARRAY -triple x86_64-linux-gnu -emit-llvm %s -o - | FileCheck -check-prefix=TLS-ARRAY %s
[OPENMP] Support for -fopenmp-simd option with compilation of simd loops only. Added support for -fopenmp-simd option that allows compilation of simd-based constructs without emission of OpenMP runtime calls. llvm-svn: 321560
2017-12-30 02:07:07 +08:00
// RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -triple %itanium_abi_triple -emit-llvm %s -o - | FileCheck --check-prefix SIMD-ONLY1 %s
// RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple %itanium_abi_triple -emit-pch -o %t %s
// RUN: %clang_cc1 -fopenmp-simd -x c++ -triple %itanium_abi_triple -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck --check-prefix SIMD-ONLY1 %s
// RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -std=c++11 -DLAMBDA -triple %itanium_abi_triple -emit-llvm %s -o - | FileCheck --check-prefix SIMD-ONLY1 %s
// RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -fblocks -DBLOCKS -triple %itanium_abi_triple -emit-llvm %s -o - | FileCheck --check-prefix SIMD-ONLY1 %s
// RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -std=c++11 -DARRAY -triple x86_64-linux-gnu -emit-llvm %s -o - | FileCheck --check-prefix SIMD-ONLY1 %s
// SIMD-ONLY1-NOT: {{__kmpc|__tgt}}
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// expected-no-diagnostics
[OPENMP] Fixed codegen for copying/initialization of array variables/parameters. This modification generates proper copyin/initialization sequences for array variables/parameters. Before they were considered as pointers, not arrays. llvm-svn: 237691
2015-05-19 20:31:28 +08:00
#ifndef ARRAY
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
#ifndef HEADER
#define HEADER
[OPENMP] Propagate alignment from original variables to the private copies. Currently private copies of captured variables have default alignment. Patch makes private variables to have same alignment as original variables. llvm-svn: 247260
2015-09-10 17:48:30 +08:00
volatile int g __attribute__((aligned(128))) = 1212;
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
#pragma omp threadprivate(g)
template <class T>
struct S {
T f;
S(T a) : f(a + g) {}
S() : f(g) {}
S &operator=(const S &) { return *this; };
operator T() { return T(); }
~S() {}
};
// CHECK-DAG: [[S_FLOAT_TY:%.+]] = type { float }
// CHECK-DAG: [[S_INT_TY:%.+]] = type { i{{[0-9]+}} }
// CHECK-DAG: [[IMPLICIT_BARRIER_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 66, i32 0, i32 0, i8*
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-CHECK-DAG: [[S_FLOAT_TY:%.+]] = type { float }
// TLS-CHECK-DAG: [[S_INT_TY:%.+]] = type { i{{[0-9]+}} }
// TLS-CHECK-DAG: [[IMPLICIT_BARRIER_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 66, i32 0, i32 0, i8*
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// CHECK-DAG: [[T_VAR:@.+]] = internal global i{{[0-9]+}} 1122,
// CHECK-DAG: [[VEC:@.+]] = internal global [2 x i{{[0-9]+}}] [i{{[0-9]+}} 1, i{{[0-9]+}} 2],
// CHECK-DAG: [[S_ARR:@.+]] = internal global [2 x [[S_FLOAT_TY]]] zeroinitializer,
// CHECK-DAG: [[VAR:@.+]] = internal global [[S_FLOAT_TY]] zeroinitializer,
// CHECK-DAG: [[TMAIN_T_VAR:@.+]] = linkonce_odr global i{{[0-9]+}} 333,
// CHECK-DAG: [[TMAIN_VEC:@.+]] = linkonce_odr global [2 x i{{[0-9]+}}] [i{{[0-9]+}} 3, i{{[0-9]+}} 3],
// CHECK-DAG: [[TMAIN_S_ARR:@.+]] = linkonce_odr global [2 x [[S_INT_TY]]] zeroinitializer,
// CHECK-DAG: [[TMAIN_VAR:@.+]] = linkonce_odr global [[S_INT_TY]] zeroinitializer,
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-CHECK-DAG: [[T_VAR:@.+]] = internal thread_local global i{{[0-9]+}} 1122,
// TLS-CHECK-DAG: [[VEC:@.+]] = internal thread_local global [2 x i{{[0-9]+}}] [i{{[0-9]+}} 1, i{{[0-9]+}} 2],
// TLS-CHECK-DAG: [[S_ARR:@.+]] = internal thread_local global [2 x [[S_FLOAT_TY]]] zeroinitializer,
// TLS-CHECK-DAG: [[VAR:@.+]] = internal thread_local global [[S_FLOAT_TY]] zeroinitializer,
// TLS-CHECK-DAG: [[TMAIN_T_VAR:@.+]] = linkonce_odr thread_local global i{{[0-9]+}} 333,
// TLS-CHECK-DAG: [[TMAIN_VEC:@.+]] = linkonce_odr thread_local global [2 x i{{[0-9]+}}] [i{{[0-9]+}} 3, i{{[0-9]+}} 3],
// TLS-CHECK-DAG: [[TMAIN_S_ARR:@.+]] = linkonce_odr thread_local global [2 x [[S_INT_TY]]] zeroinitializer,
// TLS-CHECK-DAG: [[TMAIN_VAR:@.+]] = linkonce_odr thread_local global [[S_INT_TY]] zeroinitializer,
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
template <typename T>
T tmain() {
S<T> test;
test = S<T>();
[OPENMP] Propagate alignment from original variables to the private copies. Currently private copies of captured variables have default alignment. Patch makes private variables to have same alignment as original variables. llvm-svn: 247260
2015-09-10 17:48:30 +08:00
static T t_var __attribute__((aligned(128))) = 333;
static T vec[] __attribute__((aligned(128))) = {3, 3};
static S<T> s_arr[] __attribute__((aligned(128))) = {1, 2};
static S<T> var __attribute__((aligned(128))) (3);
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
#pragma omp threadprivate(t_var, vec, s_arr, var)
#pragma omp parallel copyin(t_var, vec, s_arr, var)
{
vec[0] = t_var;
s_arr[0] = var;
}
#pragma omp parallel copyin(t_var)
{}
return T();
}
int main() {
#ifdef LAMBDA
// LAMBDA: [[G:@.+]] = global i{{[0-9]+}} 1212,
// LAMBDA-LABEL: @main
Account for calling convention specifiers in function definitions in IR test cases Several tests wouldn't pass when executed on an armv7a_pc_linux triple due to the non-default arm_aapcs calling convention produced on the function definitions in the IR output. Account for this with the application of a little regex. Patch by Ying Yi. llvm-svn: 240971
2015-06-30 01:29:50 +08:00
// LAMBDA: call{{.*}} void [[OUTER_LAMBDA:@.+]](
[OpenMP] Fix copyin clause codegen regression caused by r243277. llvm-svn: 243285
2015-07-28 01:30:41 +08:00
// TLS-LAMBDA: [[G:@.+]] = {{.*}}thread_local {{.*}}global i{{[0-9]+}} 1212,
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-LAMBDA-LABEL: @main
// TLS-LAMBDA: call{{.*}} void [[OUTER_LAMBDA:@.+]](
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
[&]() {
// LAMBDA: define{{.*}} internal{{.*}} void [[OUTER_LAMBDA]](
[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
2015-09-10 16:12:02 +08:00
// LAMBDA: call {{.*}}void {{.+}} @__kmpc_fork_call({{.+}}, i32 0, {{.+}}* [[OMP_REGION:@.+]] to {{.+}})
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
[TLS on Darwin] use CXX_FAST_TLS calling convention for access functions. Also set nounwind attribute. rdar://problem/9001553 llvm-svn: 255860
2015-12-17 08:42:36 +08:00
// TLS-LAMBDA: [[G_CPY_VAL:%.+]] = call{{( cxx_fast_tlscc)?}} i{{[0-9]+}}* [[G_CTOR:@.+]]()
[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
2015-09-10 16:12:02 +08:00
// TLS-LAMBDA: call {{.*}}void {{.+}} @__kmpc_fork_call({{.+}}, i32 1, {{.+}}* [[OMP_REGION:@.+]] to {{.+}}, i32* [[G_CPY_VAL]])
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
#pragma omp parallel copyin(g)
{
[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
2015-09-10 16:12:02 +08:00
// LAMBDA: define{{.*}} internal{{.*}} void [[OMP_REGION]](i32* noalias %{{.+}}, i32* noalias %{{.+}})
// TLS-LAMBDA: define{{.*}} internal{{.*}} void [[OMP_REGION]](i32* noalias %{{.+}}, i32* noalias %{{.+}}, i32* dereferenceable(4) %{{.+}})
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// threadprivate_g = g;
Account for calling convention specifiers in function definitions in IR test cases Several tests wouldn't pass when executed on an armv7a_pc_linux triple due to the non-default arm_aapcs calling convention produced on the function definitions in the IR output. Account for this with the application of a little regex. Patch by Ying Yi. llvm-svn: 240971
2015-06-30 01:29:50 +08:00
// LAMBDA: call {{.*}}i8* @__kmpc_threadprivate_cached({{.+}} [[G]]
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// LAMBDA: ptrtoint i{{[0-9]+}}* %{{.+}} to i{{[0-9]+}}
// LAMBDA: icmp ne i{{[0-9]+}} ptrtoint (i{{[0-9]+}}* [[G]] to i{{[0-9]+}}), %{{.+}}
// LAMBDA: br i1 %{{.+}}, label %[[NOT_MASTER:.+]], label %[[DONE:.+]]
// LAMBDA: [[NOT_MASTER]]
[OPENMP] Propagate alignment from original variables to the private copies. Currently private copies of captured variables have default alignment. Patch makes private variables to have same alignment as original variables. llvm-svn: 247260
2015-09-10 17:48:30 +08:00
// LAMBDA: load i{{[0-9]+}}, i{{[0-9]+}}* [[G]], align 128
// LAMBDA: store volatile i{{[0-9]+}} %{{.+}}, i{{[0-9]+}}* %{{.+}}, align 128
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// LAMBDA: [[DONE]]
[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
2015-09-10 16:12:02 +08:00
// TLS-LAMBDA-DAG: [[G_CAPTURE_SRC:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** %
[TLS on Darwin] use CXX_FAST_TLS calling convention for access functions. Also set nounwind attribute. rdar://problem/9001553 llvm-svn: 255860
2015-12-17 08:42:36 +08:00
// TLS-LAMBDA-DAG: [[G_CAPTURE_DST:%.+]] = call{{( cxx_fast_tlscc)?}} i{{[0-9]+}}* [[G_CTOR]]()
[OpenMP] Fix copyin clause codegen regression caused by r243277. llvm-svn: 243285
2015-07-28 01:30:41 +08:00
// TLS-LAMBDA-DAG: [[G_CAPTURE_SRCC:%.+]] = ptrtoint i{{[0-9]+}}* [[G_CAPTURE_SRC]] to i{{[0-9]+}}
// TLS-LAMBDA-DAG: [[G_CAPTURE_DSTC:%.+]] = ptrtoint i{{[0-9]+}}* [[G_CAPTURE_DST]] to i{{[0-9]+}}
[OpenMP] Fix copyin codegen test regression in order used in compare instruction. llvm-svn: 243289
2015-07-28 01:49:18 +08:00
// TLS-LAMBDA: icmp ne i{{[0-9]+}} {{%.+}}, {{%.+}}
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-LAMBDA: br i1 %{{.+}}, label %[[NOT_MASTER:.+]], label %[[DONE:.+]]
// TLS-LAMBDA: [[NOT_MASTER]]
// TLS-LAMBDA: load i{{[0-9]+}}, i{{[0-9]+}}* [[G_CAPTURE_SRC]],
[OPENMP] Propagate alignment from original variables to the private copies. Currently private copies of captured variables have default alignment. Patch makes private variables to have same alignment as original variables. llvm-svn: 247260
2015-09-10 17:48:30 +08:00
// TLS-LAMBDA: store volatile i{{[0-9]+}} %{{.+}}, i{{[0-9]+}}* [[G_CAPTURE_DST]], align 128
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-LAMBDA: [[DONE]]
[OPENMP] Emit __kmpc_cancel_barrier() and code for 'cancellation point' only if 'cancel' is found. Patch improves codegen for OpenMP constructs. If the OpenMP region does not have internal 'cancel' construct, a call to 'void __kmpc_barrier()' runtime function is generated for all implicit/explicit barriers. If the region has inner 'cancel' directive, then ``` if (__kmpc_cancel_barrier()) exit from outer construct; ``` code is generated. Also, the code for 'canellation point' directive is not generated if parent directive does not have 'cancel' directive. llvm-svn: 247681
2015-09-15 20:52:43 +08:00
// LAMBDA: call {{.*}}void @__kmpc_barrier(
// TLS-LAMBDA: call {{.*}}void @__kmpc_barrier(
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
g = 1;
Account for calling convention specifiers in function definitions in IR test cases Several tests wouldn't pass when executed on an armv7a_pc_linux triple due to the non-default arm_aapcs calling convention produced on the function definitions in the IR output. Account for this with the application of a little regex. Patch by Ying Yi. llvm-svn: 240971
2015-06-30 01:29:50 +08:00
// LAMBDA: call{{.*}} void [[INNER_LAMBDA:@.+]](%{{.+}}*
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-LAMBDA: call{{.*}} void [[INNER_LAMBDA:@.+]](%{{.+}}*
[OPENMP] Allow runtime insert its own code inside OpenMP regions. Solution unifies interface of RegionCodeGenTy type to allow insert runtime-specific code before/after main codegen action defined in CGStmtOpenMP.cpp file. Runtime should not define its own RegionCodeGenTy for general OpenMP directives, but must be allowed to insert its own (required) code to support target specific codegen. llvm-svn: 264700
2016-03-29 13:34:15 +08:00
// TLS-LAMBDA: define {{.*}}i{{[0-9]+}}* [[G_CTOR]]()
// TLS-LAMBDA: ret i{{[0-9]+}}* [[G]]
// TLS-LAMBDA: }
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
[&]() {
// LAMBDA: define {{.+}} void [[INNER_LAMBDA]](%{{.+}}* [[ARG_PTR:%.+]])
// LAMBDA: store %{{.+}}* [[ARG_PTR]], %{{.+}}** [[ARG_PTR_REF:%.+]],
g = 2;
// LAMBDA: [[ARG_PTR:%.+]] = load %{{.+}}*, %{{.+}}** [[ARG_PTR_REF]]
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
[TLS on Darwin] use CXX_FAST_TLS calling convention for access functions. Also set nounwind attribute. rdar://problem/9001553 llvm-svn: 255860
2015-12-17 08:42:36 +08:00
// TLS-LAMBDA: [[G_CAPTURE_DST:%.+]] = call{{( cxx_fast_tlscc)?}} i{{[0-9]+}}* [[G_CTOR]]()
[OPENMP] Propagate alignment from original variables to the private copies. Currently private copies of captured variables have default alignment. Patch makes private variables to have same alignment as original variables. llvm-svn: 247260
2015-09-10 17:48:30 +08:00
// TLS-LAMBDA: store volatile i{{[0-9]+}} 2, i{{[0-9]+}}* [[G_CAPTURE_DST]], align 128
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
}();
}
}();
return 0;
#elif defined(BLOCKS)
// BLOCKS: [[G:@.+]] = global i{{[0-9]+}} 1212,
// BLOCKS-LABEL: @main
Account for calling convention specifiers in function definitions in IR test cases Several tests wouldn't pass when executed on an armv7a_pc_linux triple due to the non-default arm_aapcs calling convention produced on the function definitions in the IR output. Account for this with the application of a little regex. Patch by Ying Yi. llvm-svn: 240971
2015-06-30 01:29:50 +08:00
// BLOCKS: call {{.*}}void {{%.+}}(i8
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
[OpenMP] Fix copyin codegen test regression in order used in compare instruction. llvm-svn: 243289
2015-07-28 01:49:18 +08:00
// TLS-BLOCKS: [[G:@.+]] = {{.*}}thread_local {{.*}}global i{{[0-9]+}} 1212,
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-BLOCKS-LABEL: @main
// TLS-BLOCKS: call {{.*}}void {{%.+}}(i8
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
^{
// BLOCKS: define{{.*}} internal{{.*}} void {{.+}}(i8*
[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
2015-09-10 16:12:02 +08:00
// BLOCKS: call {{.*}}void {{.+}} @__kmpc_fork_call({{.+}}, i32 0, {{.+}}* [[OMP_REGION:@.+]] to {{.+}})
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
[TLS on Darwin] use CXX_FAST_TLS calling convention for access functions. Also set nounwind attribute. rdar://problem/9001553 llvm-svn: 255860
2015-12-17 08:42:36 +08:00
// TLS-BLOCKS: [[G_CPY_VAL:%.+]] = call{{( cxx_fast_tlscc)?}} i{{[0-9]+}}* [[G_CTOR:@.+]]()
[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
2015-09-10 16:12:02 +08:00
// TLS-BLOCKS: call {{.*}}void {{.+}} @__kmpc_fork_call({{.+}}, i32 1, {{.+}}* [[OMP_REGION:@.+]] to {{.+}}, i32* [[G_CPY_VAL]])
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
#pragma omp parallel copyin(g)
{
[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
2015-09-10 16:12:02 +08:00
// BLOCKS: define{{.*}} internal{{.*}} void [[OMP_REGION]](i32* noalias %{{.+}}, i32* noalias %{{.+}})
// TLS-BLOCKS: define{{.*}} internal{{.*}} void [[OMP_REGION]](i32* noalias %{{.+}}, i32* noalias %{{.+}}, i32* dereferenceable(4) %{{.+}})
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// threadprivate_g = g;
Account for calling convention specifiers in function definitions in IR test cases Several tests wouldn't pass when executed on an armv7a_pc_linux triple due to the non-default arm_aapcs calling convention produced on the function definitions in the IR output. Account for this with the application of a little regex. Patch by Ying Yi. llvm-svn: 240971
2015-06-30 01:29:50 +08:00
// BLOCKS: call {{.*}}i8* @__kmpc_threadprivate_cached({{.+}} [[G]]
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// BLOCKS: ptrtoint i{{[0-9]+}}* %{{.+}} to i{{[0-9]+}}
// BLOCKS: icmp ne i{{[0-9]+}} ptrtoint (i{{[0-9]+}}* [[G]] to i{{[0-9]+}}), %{{.+}}
// BLOCKS: br i1 %{{.+}}, label %[[NOT_MASTER:.+]], label %[[DONE:.+]]
// BLOCKS: [[NOT_MASTER]]
[OPENMP] Propagate alignment from original variables to the private copies. Currently private copies of captured variables have default alignment. Patch makes private variables to have same alignment as original variables. llvm-svn: 247260
2015-09-10 17:48:30 +08:00
// BLOCKS: load i{{[0-9]+}}, i{{[0-9]+}}* [[G]], align 128
// BLOCKS: store volatile i{{[0-9]+}} %{{.+}}, i{{[0-9]+}}* %{{.+}}, align 128
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// BLOCKS: [[DONE]]
[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
2015-09-10 16:12:02 +08:00
// TLS-BLOCKS-DAG: [[G_CAPTURE_SRC:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** %
[TLS on Darwin] use CXX_FAST_TLS calling convention for access functions. Also set nounwind attribute. rdar://problem/9001553 llvm-svn: 255860
2015-12-17 08:42:36 +08:00
// TLS-BLOCKS-DAG: [[G_CAPTURE_DST:%.+]] = call{{( cxx_fast_tlscc)?}} i{{[0-9]+}}* [[G_CTOR]]()
[OpenMP] Fix copyin clause codegen regression caused by r243277. llvm-svn: 243285
2015-07-28 01:30:41 +08:00
// TLS-BLOCKS-DAG: [[G_CAPTURE_SRCC:%.+]] = ptrtoint i{{[0-9]+}}* [[G_CAPTURE_SRC]] to i{{[0-9]+}}
// TLS-BLOCKS-DAG: [[G_CAPTURE_DSTC:%.+]] = ptrtoint i{{[0-9]+}}* [[G_CAPTURE_DST]] to i{{[0-9]+}}
[OpenMP] Fix copyin codegen test regression in order used in compare instruction. llvm-svn: 243289
2015-07-28 01:49:18 +08:00
// TLS-BLOCKS: icmp ne i{{[0-9]+}} {{%.+}}, {{%.+}}
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-BLOCKS: br i1 %{{.+}}, label %[[NOT_MASTER:.+]], label %[[DONE:.+]]
// TLS-BLOCKS: [[NOT_MASTER]]
// TLS-BLOCKS: load i{{[0-9]+}}, i{{[0-9]+}}* [[G_CAPTURE_SRC]],
[OPENMP] Propagate alignment from original variables to the private copies. Currently private copies of captured variables have default alignment. Patch makes private variables to have same alignment as original variables. llvm-svn: 247260
2015-09-10 17:48:30 +08:00
// TLS-BLOCKS: store volatile i{{[0-9]+}} %{{.+}}, i{{[0-9]+}}* [[G_CAPTURE_DST]], align 128
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-BLOCKS: [[DONE]]
[OPENMP] Emit __kmpc_cancel_barrier() and code for 'cancellation point' only if 'cancel' is found. Patch improves codegen for OpenMP constructs. If the OpenMP region does not have internal 'cancel' construct, a call to 'void __kmpc_barrier()' runtime function is generated for all implicit/explicit barriers. If the region has inner 'cancel' directive, then ``` if (__kmpc_cancel_barrier()) exit from outer construct; ``` code is generated. Also, the code for 'canellation point' directive is not generated if parent directive does not have 'cancel' directive. llvm-svn: 247681
2015-09-15 20:52:43 +08:00
// BLOCKS: call {{.*}}void @__kmpc_barrier(
// TLS-BLOCKS: call {{.*}}void @__kmpc_barrier(
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
g = 1;
// BLOCKS: store volatile i{{[0-9]+}} 1, i{{[0-9]+}}*
// BLOCKS-NOT: [[G]]{{[[^:word:]]}}
Account for calling convention specifiers in function definitions in IR test cases Several tests wouldn't pass when executed on an armv7a_pc_linux triple due to the non-default arm_aapcs calling convention produced on the function definitions in the IR output. Account for this with the application of a little regex. Patch by Ying Yi. llvm-svn: 240971
2015-06-30 01:29:50 +08:00
// BLOCKS: call {{.*}}void {{%.+}}(i8
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
[TLS on Darwin] use CXX_FAST_TLS calling convention for access functions. Also set nounwind attribute. rdar://problem/9001553 llvm-svn: 255860
2015-12-17 08:42:36 +08:00
// TLS-BLOCKS: [[G_CAPTURE_DST:%.+]] = call{{( cxx_fast_tlscc)?}} i{{[0-9]+}}* [[G_CTOR]]()
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-BLOCKS: store volatile i{{[0-9]+}} 1, i{{[0-9]+}}* [[G_CAPTURE_DST]]
// TLS-BLOCKS-NOT: [[G]]{{[[^:word:]]}}
// TLS-BLOCKS: call {{.*}}void {{%.+}}(i8
[OPENMP] Allow runtime insert its own code inside OpenMP regions. Solution unifies interface of RegionCodeGenTy type to allow insert runtime-specific code before/after main codegen action defined in CGStmtOpenMP.cpp file. Runtime should not define its own RegionCodeGenTy for general OpenMP directives, but must be allowed to insert its own (required) code to support target specific codegen. llvm-svn: 264700
2016-03-29 13:34:15 +08:00
// TLS-BLOCKS: define {{.*}}i{{[0-9]+}}* [[G_CTOR]]()
// TLS-BLOCKS: ret i{{[0-9]+}}* [[G]]
// TLS-BLOCKS: }
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
^{
// BLOCKS: define {{.+}} void {{@.+}}(i8*
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-BLOCKS: define {{.+}} void {{@.+}}(i8*
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
g = 2;
// BLOCKS-NOT: [[G]]{{[[^:word:]]}}
Account for calling convention specifiers in function definitions in IR test cases Several tests wouldn't pass when executed on an armv7a_pc_linux triple due to the non-default arm_aapcs calling convention produced on the function definitions in the IR output. Account for this with the application of a little regex. Patch by Ying Yi. llvm-svn: 240971
2015-06-30 01:29:50 +08:00
// BLOCKS: call {{.*}}i8* @__kmpc_threadprivate_cached({{.+}} [[G]]
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// BLOCKS: store volatile i{{[0-9]+}} 2, i{{[0-9]+}}*
// BLOCKS-NOT: [[G]]{{[[^:word:]]}}
// BLOCKS: ret
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-BLOCKS-NOT: [[G]]{{[[^:word:]]}}
[TLS on Darwin] use CXX_FAST_TLS calling convention for access functions. Also set nounwind attribute. rdar://problem/9001553 llvm-svn: 255860
2015-12-17 08:42:36 +08:00
// TLS-BLOCKS: [[G_CAPTURE_DST:%.+]] = call{{( cxx_fast_tlscc)?}} i{{[0-9]+}}* [[G_CTOR]]()
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-BLOCKS: store volatile i{{[0-9]+}} 2, i{{[0-9]+}}* [[G_CAPTURE_DST]]
// TLS-BLOCKS-NOT: [[G]]{{[[^:word:]]}}
// TLS-BLOCKS: ret
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
}();
}
}();
return 0;
#else
S<float> test;
test = S<float>();
static int t_var = 1122;
static int vec[] = {1, 2};
static S<float> s_arr[] = {1, 2};
static S<float> var(3);
#pragma omp threadprivate(t_var, vec, s_arr, var)
#pragma omp parallel copyin(t_var, vec, s_arr, var)
{
vec[0] = t_var;
s_arr[0] = var;
}
#pragma omp parallel copyin(t_var)
{}
return tmain<int>();
#endif
}
// CHECK-LABEL: @main
// CHECK: [[TEST:%.+]] = alloca [[S_FLOAT_TY]],
// CHECK: call {{.*}} [[S_FLOAT_TY_COPY_ASSIGN:@.+]]([[S_FLOAT_TY]]* [[TEST]], [[S_FLOAT_TY]]*
[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
2015-09-10 16:12:02 +08:00
// CHECK: call {{.*}}void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 0, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)* bitcast (void (i{{[0-9]+}}*, i{{[0-9]+}}*)* [[MAIN_MICROTASK:@.+]] to void (i32*, i32*, ...)*))
// CHECK: call {{.*}}void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 0, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)* bitcast (void (i{{[0-9]+}}*, i{{[0-9]+}}*)* [[MAIN_MICROTASK1:@.+]] to void (i32*, i32*, ...)*))
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// CHECK: = call {{.*}}i{{.+}} [[TMAIN_INT:@.+]]()
// CHECK: call {{.*}} [[S_FLOAT_TY_DESTR:@.+]]([[S_FLOAT_TY]]*
// CHECK: ret
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-CHECK-LABEL: @main
// TLS-CHECK: [[TEST:%.+]] = alloca [[S_FLOAT_TY]],
// TLS-CHECK: call {{.*}} [[S_FLOAT_TY_COPY_ASSIGN:@.+]]([[S_FLOAT_TY]]* [[TEST]], [[S_FLOAT_TY]]*
[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
2015-09-10 16:12:02 +08:00
// TLS-CHECK: call {{.*}}void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 4, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)* bitcast (void (i{{[0-9]+}}*, i{{[0-9]+}}*, i32*, [2 x i32]*, [2 x [[S_FLOAT_TY]]]*, [[S_FLOAT_TY]]*)* [[MAIN_MICROTASK:@.+]] to void (i32*, i32*, ...)*),
// TLS-CHECK: call {{.*}}void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 1, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)* bitcast (void (i{{[0-9]+}}*, i{{[0-9]+}}*, i32*)* [[MAIN_MICROTASK1:@.+]] to void (i32*, i32*, ...)*),
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-CHECK: = call {{.*}}i{{.+}} [[TMAIN_INT:@.+]]()
// TLS-CHECK: call {{.*}} [[S_FLOAT_TY_DESTR:@.+]]([[S_FLOAT_TY]]*
// TLS-CHECK: ret
[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
2015-09-10 16:12:02 +08:00
// CHECK: define internal {{.*}}void [[MAIN_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}})
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]],
// CHECK: [[GTID_ADDR:%.+]] = load i32*, i32** [[GTID_ADDR_ADDR]],
// CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_ADDR]],
[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
2015-09-10 16:12:02 +08:00
// TLS-CHECK: define internal {{.*}}void [[MAIN_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}},
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]],
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// threadprivate_t_var = t_var;
Account for calling convention specifiers in function definitions in IR test cases Several tests wouldn't pass when executed on an armv7a_pc_linux triple due to the non-default arm_aapcs calling convention produced on the function definitions in the IR output. Account for this with the application of a little regex. Patch by Ying Yi. llvm-svn: 240971
2015-06-30 01:29:50 +08:00
// CHECK: call {{.*}}i8* @__kmpc_threadprivate_cached({{.+}} [[T_VAR]]
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// CHECK: ptrtoint i{{[0-9]+}}* %{{.+}} to i{{[0-9]+}}
// CHECK: icmp ne i{{[0-9]+}} ptrtoint (i{{[0-9]+}}* [[T_VAR]] to i{{[0-9]+}}), %{{.+}}
// CHECK: br i1 %{{.+}}, label %[[NOT_MASTER:.+]], label %[[DONE:.+]]
// CHECK: [[NOT_MASTER]]
// CHECK: load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR]],
// CHECK: store i{{[0-9]+}} %{{.+}}, i{{[0-9]+}}* %{{.+}},
[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
2015-09-10 16:12:02 +08:00
// TLS-CHECK: [[MASTER_REF:%.+]] = load i32*, i32** %
// TLS-CHECK: [[MASTER_REF2:%.+]] = load [2 x i32]*, [2 x i32]** %
// TLS-CHECK: [[MASTER_REF3:%.+]] = load [2 x [[S_FLOAT_TY]]]*, [2 x [[S_FLOAT_TY]]]** %
// TLS-CHECK: [[MASTER_REF4:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** %
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-CHECK: [[MASTER_LONG:%.+]] = ptrtoint i32* [[MASTER_REF]] to i{{[0-9]+}}
// TLS-CHECK: icmp ne i{{[0-9]+}} [[MASTER_LONG]], ptrtoint (i{{[0-9]+}}* [[T_VAR]] to i{{[0-9]+}})
// TLS-CHECK: br i1 %{{.+}}, label %[[NOT_MASTER:.+]], label %[[DONE:.+]]
// TLS-CHECK: [[NOT_MASTER]]
// TLS-CHECK: [[MASTER_VAL:%.+]] = load i32, i32* [[MASTER_REF]]
// TLS-CHECK: store i32 [[MASTER_VAL]], i32* [[T_VAR]]
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// threadprivate_vec = vec;
Account for calling convention specifiers in function definitions in IR test cases Several tests wouldn't pass when executed on an armv7a_pc_linux triple due to the non-default arm_aapcs calling convention produced on the function definitions in the IR output. Account for this with the application of a little regex. Patch by Ying Yi. llvm-svn: 240971
2015-06-30 01:29:50 +08:00
// CHECK: call {{.*}}i8* @__kmpc_threadprivate_cached({{.+}} [[VEC]]
Change memcpy/memove/memset to have dest and source alignment attributes (Step 1). Summary: Upstream LLVM is changing the the prototypes of the @llvm.memcpy/memmove/memset intrinsics. This change updates the Clang tests for this change. The @llvm.memcpy/memmove/memset intrinsics currently have an explicit argument which is required to be a constant integer. It represents the alignment of the dest (and source), and so must be the minimum of the actual alignment of the two. This change removes the alignment argument in favour of placing the alignment attribute on the source and destination pointers of the memory intrinsic call. For example, code which used to read: call void @llvm.memcpy.p0i8.p0i8.i32(i8* %dest, i8* %src, i32 100, i32 4, i1 false) will now read call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %dest, i8* align 4 %src, i32 100, i1 false) At this time the source and destination alignments must be the same (Step 1). Step 2 of the change, to be landed shortly, will relax that contraint and allow the source and destination to have different alignments. llvm-svn: 322964
2018-01-20 01:12:54 +08:00
// CHECK: call void @llvm.memcpy{{.*}}(i8* align {{[0-9]+}} %{{.+}}, i8* align {{[0-9]+}} bitcast ([2 x i{{[0-9]+}}]* [[VEC]] to i8*),
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
[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
2015-09-10 16:12:02 +08:00
// TLS-CHECK: [[MASTER_CAST:%.+]] = bitcast [2 x i32]* [[MASTER_REF2]] to i8*
Change memcpy/memove/memset to have dest and source alignment attributes (Step 1). Summary: Upstream LLVM is changing the the prototypes of the @llvm.memcpy/memmove/memset intrinsics. This change updates the Clang tests for this change. The @llvm.memcpy/memmove/memset intrinsics currently have an explicit argument which is required to be a constant integer. It represents the alignment of the dest (and source), and so must be the minimum of the actual alignment of the two. This change removes the alignment argument in favour of placing the alignment attribute on the source and destination pointers of the memory intrinsic call. For example, code which used to read: call void @llvm.memcpy.p0i8.p0i8.i32(i8* %dest, i8* %src, i32 100, i32 4, i1 false) will now read call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %dest, i8* align 4 %src, i32 100, i1 false) At this time the source and destination alignments must be the same (Step 1). Step 2 of the change, to be landed shortly, will relax that contraint and allow the source and destination to have different alignments. llvm-svn: 322964
2018-01-20 01:12:54 +08:00
// TLS-CHECK: call void @llvm.memcpy{{.*}}(i8* align {{[0-9]+}} bitcast ([2 x i{{[0-9]+}}]* [[VEC]] to i8*), i8* align {{[0-9]+}} [[MASTER_CAST]]
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// threadprivate_s_arr = s_arr;
Account for calling convention specifiers in function definitions in IR test cases Several tests wouldn't pass when executed on an armv7a_pc_linux triple due to the non-default arm_aapcs calling convention produced on the function definitions in the IR output. Account for this with the application of a little regex. Patch by Ying Yi. llvm-svn: 240971
2015-06-30 01:29:50 +08:00
// CHECK: call {{.*}}i8* @__kmpc_threadprivate_cached({{.+}} [[S_ARR]]
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// CHECK: [[S_ARR_PRIV_BEGIN:%.+]] = getelementptr inbounds [2 x [[S_FLOAT_TY]]], [2 x [[S_FLOAT_TY]]]* {{%.+}}, i{{[0-9]+}} 0, i{{[0-9]+}} 0
// CHECK: [[S_ARR_PRIV_END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[S_ARR_PRIV_BEGIN]], i{{[0-9]+}} 2
// CHECK: [[IS_EMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[S_ARR_PRIV_BEGIN]], [[S_ARR_PRIV_END]]
// CHECK: br i1 [[IS_EMPTY]], label %[[S_ARR_BODY_DONE:.+]], label %[[S_ARR_BODY:.+]]
// CHECK: [[S_ARR_BODY]]
// CHECK: call {{.*}} [[S_FLOAT_TY_COPY_ASSIGN]]([[S_FLOAT_TY]]* {{.+}}, [[S_FLOAT_TY]]* {{.+}})
// CHECK: br i1 {{.+}}, label %{{.+}}, label %[[S_ARR_BODY]]
[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
2015-09-10 16:12:02 +08:00
// TLS-CHECK: [[MASTER_CAST:%.+]] = bitcast [2 x [[S_FLOAT_TY]]]* [[MASTER_REF3]] to [[S_FLOAT_TY]]*
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-CHECK-DAG: [[S_ARR_SRC_BEGIN:%.+]] = phi [[S_FLOAT_TY]]* {{.*}}[[MASTER_CAST]]
// TLS-CHECK-DAG: [[S_ARR_DST_BEGIN:%.+]] = phi [[S_FLOAT_TY]]* {{.*}}getelementptr inbounds ([2 x [[S_FLOAT_TY]]], [2 x [[S_FLOAT_TY]]]* [[S_ARR]], i{{[0-9]+}} 0, i{{[0-9]+}} 0)
// TLS-CHECK: call {{.*}} [[S_FLOAT_TY_COPY_ASSIGN]]([[S_FLOAT_TY]]* {{.+}}, [[S_FLOAT_TY]]* {{.+}})
// TLS-CHECK-DAG: [[S_ARR_SRC_END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[S_ARR_SRC_BEGIN]], i{{[0-9]+}} 1
// TLS-CHECK-DAG: [[S_ARR_DST_END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[S_ARR_DST_BEGIN]], i{{[0-9]+}} 1
// TLS-CHECK: icmp eq [[S_FLOAT_TY]]* [[S_ARR_DST_END]], getelementptr ([[S_FLOAT_TY]], [[S_FLOAT_TY]]* getelementptr inbounds ([2 x [[S_FLOAT_TY]]], [2 x [[S_FLOAT_TY]]]* [[S_ARR]], i{{[0-9]+}} 0, i{{[0-9]+}} 0), i{{[0-9]+}} 2)
// TLS-CHECK: br i1 %{{.*}}, label %[[ARR_DONE:.+]], label {{.*}}
// TLS-CHECK: [[ARR_DONE]]
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// threadprivate_var = var;
Account for calling convention specifiers in function definitions in IR test cases Several tests wouldn't pass when executed on an armv7a_pc_linux triple due to the non-default arm_aapcs calling convention produced on the function definitions in the IR output. Account for this with the application of a little regex. Patch by Ying Yi. llvm-svn: 240971
2015-06-30 01:29:50 +08:00
// CHECK: call {{.*}}i8* @__kmpc_threadprivate_cached({{.+}} [[VAR]]
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// CHECK: call {{.*}} [[S_FLOAT_TY_COPY_ASSIGN]]([[S_FLOAT_TY]]* {{%.+}}, [[S_FLOAT_TY]]* {{.*}}[[VAR]])
// CHECK: [[DONE]]
[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
2015-09-10 16:12:02 +08:00
// TLS-CHECK: call {{.*}} [[S_FLOAT_TY_COPY_ASSIGN]]([[S_FLOAT_TY]]* {{.*}}[[VAR]], [[S_FLOAT_TY]]* {{.*}}[[MASTER_REF4]])
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
[OPENMP] Emit __kmpc_cancel_barrier() and code for 'cancellation point' only if 'cancel' is found. Patch improves codegen for OpenMP constructs. If the OpenMP region does not have internal 'cancel' construct, a call to 'void __kmpc_barrier()' runtime function is generated for all implicit/explicit barriers. If the region has inner 'cancel' directive, then ``` if (__kmpc_cancel_barrier()) exit from outer construct; ``` code is generated. Also, the code for 'canellation point' directive is not generated if parent directive does not have 'cancel' directive. llvm-svn: 247681
2015-09-15 20:52:43 +08:00
// CHECK: call {{.*}}void @__kmpc_barrier(%{{.+}}* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]])
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// CHECK: ret void
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-CHECK: [[GTID_ADDR:%.+]] = load i32*, i32** [[GTID_ADDR_ADDR]],
// TLS-CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_ADDR]],
[OPENMP] Emit __kmpc_cancel_barrier() and code for 'cancellation point' only if 'cancel' is found. Patch improves codegen for OpenMP constructs. If the OpenMP region does not have internal 'cancel' construct, a call to 'void __kmpc_barrier()' runtime function is generated for all implicit/explicit barriers. If the region has inner 'cancel' directive, then ``` if (__kmpc_cancel_barrier()) exit from outer construct; ``` code is generated. Also, the code for 'canellation point' directive is not generated if parent directive does not have 'cancel' directive. llvm-svn: 247681
2015-09-15 20:52:43 +08:00
// TLS-CHECK: call {{.*}}void @__kmpc_barrier(%{{.+}}* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]])
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-CHECK: ret void
[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
2015-09-10 16:12:02 +08:00
// CHECK: define internal {{.*}}void [[MAIN_MICROTASK1]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}})
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]],
// CHECK: [[GTID_ADDR:%.+]] = load i32*, i32** [[GTID_ADDR_ADDR]],
// CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_ADDR]],
[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
2015-09-10 16:12:02 +08:00
// TLS-CHECK: define internal {{.*}}void [[MAIN_MICROTASK1]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}})
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]],
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// threadprivate_t_var = t_var;
Account for calling convention specifiers in function definitions in IR test cases Several tests wouldn't pass when executed on an armv7a_pc_linux triple due to the non-default arm_aapcs calling convention produced on the function definitions in the IR output. Account for this with the application of a little regex. Patch by Ying Yi. llvm-svn: 240971
2015-06-30 01:29:50 +08:00
// CHECK: call {{.*}}i8* @__kmpc_threadprivate_cached({{.+}} [[T_VAR]]
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// CHECK: ptrtoint i{{[0-9]+}}* %{{.+}} to i{{[0-9]+}}
// CHECK: icmp ne i{{[0-9]+}} ptrtoint (i{{[0-9]+}}* [[T_VAR]] to i{{[0-9]+}}), %{{.+}}
// CHECK: br i1 %{{.+}}, label %[[NOT_MASTER:.+]], label %[[DONE:.+]]
// CHECK: [[NOT_MASTER]]
// CHECK: load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR]],
// CHECK: store i{{[0-9]+}} %{{.+}}, i{{[0-9]+}}* %{{.+}},
// CHECK: [[DONE]]
[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
2015-09-10 16:12:02 +08:00
// TLS-CHECK: [[MASTER_REF:%.+]] = load i32*, i32** %
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-CHECK: [[MASTER_LONG:%.+]] = ptrtoint i32* [[MASTER_REF]] to i{{[0-9]+}}
// TLS-CHECK: icmp ne i{{[0-9]+}} [[MASTER_LONG]], ptrtoint (i{{[0-9]+}}* [[T_VAR]] to i{{[0-9]+}})
// TLS-CHECK: br i1 %{{.+}}, label %[[NOT_MASTER:.+]], label %[[DONE:.+]]
// TLS-CHECK: [[NOT_MASTER]]
// TLS-CHECK: [[MASTER_VAL:%.+]] = load i32, i32* [[MASTER_REF]]
// TLS-CHECK: store i32 [[MASTER_VAL]], i32* [[T_VAR]]
// TLS-CHECK: [[DONE]]
[OPENMP] Emit __kmpc_cancel_barrier() and code for 'cancellation point' only if 'cancel' is found. Patch improves codegen for OpenMP constructs. If the OpenMP region does not have internal 'cancel' construct, a call to 'void __kmpc_barrier()' runtime function is generated for all implicit/explicit barriers. If the region has inner 'cancel' directive, then ``` if (__kmpc_cancel_barrier()) exit from outer construct; ``` code is generated. Also, the code for 'canellation point' directive is not generated if parent directive does not have 'cancel' directive. llvm-svn: 247681
2015-09-15 20:52:43 +08:00
// CHECK: call {{.*}}void @__kmpc_barrier(%{{.+}}* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]])
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// CHECK: ret void
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-CHECK: [[GTID_ADDR:%.+]] = load i32*, i32** [[GTID_ADDR_ADDR]],
// TLS-CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_ADDR]],
[OPENMP] Emit __kmpc_cancel_barrier() and code for 'cancellation point' only if 'cancel' is found. Patch improves codegen for OpenMP constructs. If the OpenMP region does not have internal 'cancel' construct, a call to 'void __kmpc_barrier()' runtime function is generated for all implicit/explicit barriers. If the region has inner 'cancel' directive, then ``` if (__kmpc_cancel_barrier()) exit from outer construct; ``` code is generated. Also, the code for 'canellation point' directive is not generated if parent directive does not have 'cancel' directive. llvm-svn: 247681
2015-09-15 20:52:43 +08:00
// TLS-CHECK: call {{.*}}void @__kmpc_barrier(%{{.+}}* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]])
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-CHECK: ret void
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// CHECK: define {{.*}} i{{[0-9]+}} [[TMAIN_INT]]()
// CHECK: [[TEST:%.+]] = alloca [[S_INT_TY]],
// CHECK: call {{.*}} [[S_INT_TY_COPY_ASSIGN:@.+]]([[S_INT_TY]]* [[TEST]], [[S_INT_TY]]*
[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
2015-09-10 16:12:02 +08:00
// CHECK: call {{.*}}void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 0, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)* bitcast (void (i{{[0-9]+}}*, i{{[0-9]+}}*)* [[TMAIN_MICROTASK:@.+]] to void (i32*, i32*, ...)*))
// CHECK: call {{.*}}void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 0, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)* bitcast (void (i{{[0-9]+}}*, i{{[0-9]+}}*)* [[TMAIN_MICROTASK1:@.+]] to void (i32*, i32*, ...)*))
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// CHECK: call {{.*}} [[S_INT_TY_DESTR:@.+]]([[S_INT_TY]]*
// CHECK: ret
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-CHECK: define {{.*}} i{{[0-9]+}} [[TMAIN_INT]]()
// TLS-CHECK: [[TEST:%.+]] = alloca [[S_INT_TY]],
// TLS-CHECK: call {{.*}} [[S_INT_TY_COPY_ASSIGN:@.+]]([[S_INT_TY]]* [[TEST]], [[S_INT_TY]]*
[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
2015-09-10 16:12:02 +08:00
// TLS-CHECK: call {{.*}}void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 4, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)* bitcast (void (i{{[0-9]+}}*, i{{[0-9]+}}*, i32*, [2 x i32]*, [2 x [[S_INT_TY]]]*, [[S_INT_TY]]*)* [[TMAIN_MICROTASK:@.+]] to void (i32*, i32*, ...)*),
// TLS-CHECK: call {{.*}}void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 1, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)* bitcast (void (i{{[0-9]+}}*, i{{[0-9]+}}*, i32*)* [[TMAIN_MICROTASK1:@.+]] to void (i32*, i32*, ...)*),
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
//
[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
2015-09-10 16:12:02 +08:00
// CHECK: define internal {{.*}}void [[TMAIN_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}})
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]],
// CHECK: [[GTID_ADDR:%.+]] = load i32*, i32** [[GTID_ADDR_ADDR]],
// CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_ADDR]],
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
//
[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
2015-09-10 16:12:02 +08:00
// TLS-CHECK: define internal {{.*}}void [[TMAIN_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}})
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]],
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// threadprivate_t_var = t_var;
Account for calling convention specifiers in function definitions in IR test cases Several tests wouldn't pass when executed on an armv7a_pc_linux triple due to the non-default arm_aapcs calling convention produced on the function definitions in the IR output. Account for this with the application of a little regex. Patch by Ying Yi. llvm-svn: 240971
2015-06-30 01:29:50 +08:00
// CHECK: call {{.*}}i8* @__kmpc_threadprivate_cached({{.+}} [[TMAIN_T_VAR]]
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// CHECK: ptrtoint i{{[0-9]+}}* %{{.+}} to i{{[0-9]+}}
// CHECK: icmp ne i{{[0-9]+}} ptrtoint (i{{[0-9]+}}* [[TMAIN_T_VAR]] to i{{[0-9]+}}), %{{.+}}
// CHECK: br i1 %{{.+}}, label %[[NOT_MASTER:.+]], label %[[DONE:.+]]
// CHECK: [[NOT_MASTER]]
[OPENMP] Propagate alignment from original variables to the private copies. Currently private copies of captured variables have default alignment. Patch makes private variables to have same alignment as original variables. llvm-svn: 247260
2015-09-10 17:48:30 +08:00
// CHECK: load i{{[0-9]+}}, i{{[0-9]+}}* [[TMAIN_T_VAR]], align 128
// CHECK: store i{{[0-9]+}} %{{.+}}, i{{[0-9]+}}* %{{.+}}, align 128
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
[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
2015-09-10 16:12:02 +08:00
// TLS-CHECK: [[MASTER_REF:%.+]] = load i32*, i32** %
// TLS-CHECK: [[MASTER_REF1:%.+]] = load [2 x i32]*, [2 x i32]** %
// TLS-CHECK: [[MASTER_REF2:%.+]] = load [2 x [[S_INT_TY]]]*, [2 x [[S_INT_TY]]]** %
// TLS-CHECK: [[MASTER_REF3:%.+]] = load [[S_INT_TY]]*, [[S_INT_TY]]** %
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-CHECK: [[MASTER_LONG:%.+]] = ptrtoint i32* [[MASTER_REF]] to i{{[0-9]+}}
// TLS-CHECK: icmp ne i{{[0-9]+}} [[MASTER_LONG]], ptrtoint (i{{[0-9]+}}* [[TMAIN_T_VAR]] to i{{[0-9]+}})
// TLS-CHECK: br i1 %{{.+}}, label %[[NOT_MASTER:.+]], label %[[DONE:.+]]
// TLS-CHECK: [[NOT_MASTER]]
[OPENMP] Propagate alignment from original variables to the private copies. Currently private copies of captured variables have default alignment. Patch makes private variables to have same alignment as original variables. llvm-svn: 247260
2015-09-10 17:48:30 +08:00
// TLS-CHECK: [[MASTER_VAL:%.+]] = load i32, i32* [[MASTER_REF]],
// TLS-CHECK: store i32 [[MASTER_VAL]], i32* [[TMAIN_T_VAR]], align 128
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// threadprivate_vec = vec;
Account for calling convention specifiers in function definitions in IR test cases Several tests wouldn't pass when executed on an armv7a_pc_linux triple due to the non-default arm_aapcs calling convention produced on the function definitions in the IR output. Account for this with the application of a little regex. Patch by Ying Yi. llvm-svn: 240971
2015-06-30 01:29:50 +08:00
// CHECK: call {{.*}}i8* @__kmpc_threadprivate_cached({{.+}} [[TMAIN_VEC]]
Change memcpy/memove/memset to have dest and source alignment attributes (Step 1). Summary: Upstream LLVM is changing the the prototypes of the @llvm.memcpy/memmove/memset intrinsics. This change updates the Clang tests for this change. The @llvm.memcpy/memmove/memset intrinsics currently have an explicit argument which is required to be a constant integer. It represents the alignment of the dest (and source), and so must be the minimum of the actual alignment of the two. This change removes the alignment argument in favour of placing the alignment attribute on the source and destination pointers of the memory intrinsic call. For example, code which used to read: call void @llvm.memcpy.p0i8.p0i8.i32(i8* %dest, i8* %src, i32 100, i32 4, i1 false) will now read call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %dest, i8* align 4 %src, i32 100, i1 false) At this time the source and destination alignments must be the same (Step 1). Step 2 of the change, to be landed shortly, will relax that contraint and allow the source and destination to have different alignments. llvm-svn: 322964
2018-01-20 01:12:54 +08:00
// CHECK: call {{.*}}void @llvm.memcpy{{.*}}(i8* align {{[0-9]+}} %{{.+}}, i8* align {{[0-9]+}} bitcast ([2 x i{{[0-9]+}}]* [[TMAIN_VEC]] to i8*),
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
[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
2015-09-10 16:12:02 +08:00
// TLS-CHECK: [[MASTER_CAST:%.+]] = bitcast [2 x i32]* [[MASTER_REF1]] to i8*
Change memcpy/memove/memset to have dest and source alignment attributes (Step 1). Summary: Upstream LLVM is changing the the prototypes of the @llvm.memcpy/memmove/memset intrinsics. This change updates the Clang tests for this change. The @llvm.memcpy/memmove/memset intrinsics currently have an explicit argument which is required to be a constant integer. It represents the alignment of the dest (and source), and so must be the minimum of the actual alignment of the two. This change removes the alignment argument in favour of placing the alignment attribute on the source and destination pointers of the memory intrinsic call. For example, code which used to read: call void @llvm.memcpy.p0i8.p0i8.i32(i8* %dest, i8* %src, i32 100, i32 4, i1 false) will now read call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %dest, i8* align 4 %src, i32 100, i1 false) At this time the source and destination alignments must be the same (Step 1). Step 2 of the change, to be landed shortly, will relax that contraint and allow the source and destination to have different alignments. llvm-svn: 322964
2018-01-20 01:12:54 +08:00
// TLS-CHECK: call void @llvm.memcpy{{.*}}(i8* align {{[0-9]+}} bitcast ([2 x i{{[0-9]+}}]* [[TMAIN_VEC]] to i8*), i8* align {{[0-9]+}} [[MASTER_CAST]]
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// threadprivate_s_arr = s_arr;
Account for calling convention specifiers in function definitions in IR test cases Several tests wouldn't pass when executed on an armv7a_pc_linux triple due to the non-default arm_aapcs calling convention produced on the function definitions in the IR output. Account for this with the application of a little regex. Patch by Ying Yi. llvm-svn: 240971
2015-06-30 01:29:50 +08:00
// CHECK: call {{.*}}i8* @__kmpc_threadprivate_cached({{.+}} [[TMAIN_S_ARR]]
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// CHECK: [[S_ARR_PRIV_BEGIN:%.+]] = getelementptr inbounds [2 x [[S_INT_TY]]], [2 x [[S_INT_TY]]]* {{%.+}}, i{{[0-9]+}} 0, i{{[0-9]+}} 0
// CHECK: [[S_ARR_PRIV_END:%.+]] = getelementptr [[S_INT_TY]], [[S_INT_TY]]* [[S_ARR_PRIV_BEGIN]], i{{[0-9]+}} 2
// CHECK: [[IS_EMPTY:%.+]] = icmp eq [[S_INT_TY]]* [[S_ARR_PRIV_BEGIN]], [[S_ARR_PRIV_END]]
// CHECK: br i1 [[IS_EMPTY]], label %[[S_ARR_BODY_DONE:.+]], label %[[S_ARR_BODY:.+]]
// CHECK: [[S_ARR_BODY]]
// CHECK: call {{.*}} [[S_INT_TY_COPY_ASSIGN]]([[S_INT_TY]]* {{.+}}, [[S_INT_TY]]* {{.+}})
// CHECK: br i1 {{.+}}, label %{{.+}}, label %[[S_ARR_BODY]]
[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
2015-09-10 16:12:02 +08:00
// TLS-CHECK: [[MASTER_CAST:%.+]] = bitcast [2 x [[S_INT_TY]]]* [[MASTER_REF2]] to [[S_INT_TY]]*
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-CHECK-DAG: [[S_ARR_SRC_BEGIN:%.+]] = phi [[S_INT_TY]]* {{.*}}[[MASTER_CAST]]
// TLS-CHECK-DAG: [[S_ARR_DST_BEGIN:%.+]] = phi [[S_INT_TY]]* {{.*}}getelementptr inbounds ([2 x [[S_INT_TY]]], [2 x [[S_INT_TY]]]* [[TMAIN_S_ARR]], i{{[0-9]+}} 0, i{{[0-9]+}} 0)
// TLS-CHECK: call {{.*}} [[S_INT_TY_COPY_ASSIGN]]([[S_INT_TY]]* {{.+}}, [[S_INT_TY]]* {{.+}})
// TLS-CHECK-DAG: [[S_ARR_SRC_END:%.+]] = getelementptr [[S_INT_TY]], [[S_INT_TY]]* [[S_ARR_SRC_BEGIN]], i{{[0-9]+}} 1
// TLS-CHECK-DAG: [[S_ARR_DST_END:%.+]] = getelementptr [[S_INT_TY]], [[S_INT_TY]]* [[S_ARR_DST_BEGIN]], i{{[0-9]+}} 1
// TLS-CHECK: icmp eq [[S_INT_TY]]* [[S_ARR_DST_END]], getelementptr ([[S_INT_TY]], [[S_INT_TY]]* getelementptr inbounds ([2 x [[S_INT_TY]]], [2 x [[S_INT_TY]]]* [[TMAIN_S_ARR]], i{{[0-9]+}} 0, i{{[0-9]+}} 0), i{{[0-9]+}} 2)
// TLS-CHECK: br i1 %{{.*}}, label %[[ARR_DONE:.+]], label {{.*}}
// TLS-CHECK: [[ARR_DONE]]
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// threadprivate_var = var;
Account for calling convention specifiers in function definitions in IR test cases Several tests wouldn't pass when executed on an armv7a_pc_linux triple due to the non-default arm_aapcs calling convention produced on the function definitions in the IR output. Account for this with the application of a little regex. Patch by Ying Yi. llvm-svn: 240971
2015-06-30 01:29:50 +08:00
// CHECK: call {{.*}}i8* @__kmpc_threadprivate_cached({{.+}} [[TMAIN_VAR]]
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// CHECK: call {{.*}} [[S_INT_TY_COPY_ASSIGN]]([[S_INT_TY]]* {{%.+}}, [[S_INT_TY]]* {{.*}}[[TMAIN_VAR]])
// CHECK: [[DONE]]
[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
2015-09-10 16:12:02 +08:00
// TLS-CHECK: call {{.*}} [[S_INT_TY_COPY_ASSIGN]]([[S_INT_TY]]* {{.*}}[[TMAIN_VAR]], [[S_INT_TY]]* {{.*}}[[MASTER_REF3]])
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
[OPENMP] Emit __kmpc_cancel_barrier() and code for 'cancellation point' only if 'cancel' is found. Patch improves codegen for OpenMP constructs. If the OpenMP region does not have internal 'cancel' construct, a call to 'void __kmpc_barrier()' runtime function is generated for all implicit/explicit barriers. If the region has inner 'cancel' directive, then ``` if (__kmpc_cancel_barrier()) exit from outer construct; ``` code is generated. Also, the code for 'canellation point' directive is not generated if parent directive does not have 'cancel' directive. llvm-svn: 247681
2015-09-15 20:52:43 +08:00
// CHECK: call {{.*}}void @__kmpc_barrier(%{{.+}}* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]])
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// CHECK: ret void
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-CHECK: [[GTID_ADDR:%.+]] = load i32*, i32** [[GTID_ADDR_ADDR]],
// TLS-CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_ADDR]],
[OPENMP] Emit __kmpc_cancel_barrier() and code for 'cancellation point' only if 'cancel' is found. Patch improves codegen for OpenMP constructs. If the OpenMP region does not have internal 'cancel' construct, a call to 'void __kmpc_barrier()' runtime function is generated for all implicit/explicit barriers. If the region has inner 'cancel' directive, then ``` if (__kmpc_cancel_barrier()) exit from outer construct; ``` code is generated. Also, the code for 'canellation point' directive is not generated if parent directive does not have 'cancel' directive. llvm-svn: 247681
2015-09-15 20:52:43 +08:00
// TLS-CHECK: call {{.*}}void @__kmpc_barrier(%{{.+}}* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]])
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-CHECK: ret void
[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
2015-09-10 16:12:02 +08:00
// CHECK: define internal {{.*}}void [[TMAIN_MICROTASK1]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}})
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]],
// CHECK: [[GTID_ADDR:%.+]] = load i32*, i32** [[GTID_ADDR_ADDR]],
// CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_ADDR]],
[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
2015-09-10 16:12:02 +08:00
// TLS-CHECK: define internal {{.*}}void [[TMAIN_MICROTASK1]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}},
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]],
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// threadprivate_t_var = t_var;
Account for calling convention specifiers in function definitions in IR test cases Several tests wouldn't pass when executed on an armv7a_pc_linux triple due to the non-default arm_aapcs calling convention produced on the function definitions in the IR output. Account for this with the application of a little regex. Patch by Ying Yi. llvm-svn: 240971
2015-06-30 01:29:50 +08:00
// CHECK: call {{.*}}i8* @__kmpc_threadprivate_cached({{.+}} [[TMAIN_T_VAR]]
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// CHECK: ptrtoint i{{[0-9]+}}* %{{.+}} to i{{[0-9]+}}
// CHECK: icmp ne i{{[0-9]+}} ptrtoint (i{{[0-9]+}}* [[TMAIN_T_VAR]] to i{{[0-9]+}}), %{{.+}}
// CHECK: br i1 %{{.+}}, label %[[NOT_MASTER:.+]], label %[[DONE:.+]]
// CHECK: [[NOT_MASTER]]
// CHECK: load i{{[0-9]+}}, i{{[0-9]+}}* [[TMAIN_T_VAR]],
// CHECK: store i{{[0-9]+}} %{{.+}}, i{{[0-9]+}}* %{{.+}},
// CHECK: [[DONE]]
[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
2015-09-10 16:12:02 +08:00
// TLS-CHECK: [[MASTER_REF:%.+]] = load i32*, i32** %
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-CHECK: [[MASTER_LONG:%.+]] = ptrtoint i32* [[MASTER_REF]] to i{{[0-9]+}}
// TLS-CHECK: icmp ne i{{[0-9]+}} [[MASTER_LONG]], ptrtoint (i{{[0-9]+}}* [[TMAIN_T_VAR]] to i{{[0-9]+}})
// TLS-CHECK: br i1 %{{.+}}, label %[[NOT_MASTER:.+]], label %[[DONE:.+]]
// TLS-CHECK: [[NOT_MASTER]]
// TLS-CHECK: [[MASTER_VAL:%.+]] = load i32, i32* [[MASTER_REF]]
// TLS-CHECK: store i32 [[MASTER_VAL]], i32* [[TMAIN_T_VAR]]
// TLS-CHECK: [[DONE]]
[OPENMP] Emit __kmpc_cancel_barrier() and code for 'cancellation point' only if 'cancel' is found. Patch improves codegen for OpenMP constructs. If the OpenMP region does not have internal 'cancel' construct, a call to 'void __kmpc_barrier()' runtime function is generated for all implicit/explicit barriers. If the region has inner 'cancel' directive, then ``` if (__kmpc_cancel_barrier()) exit from outer construct; ``` code is generated. Also, the code for 'canellation point' directive is not generated if parent directive does not have 'cancel' directive. llvm-svn: 247681
2015-09-15 20:52:43 +08:00
// CHECK: call {{.*}}void @__kmpc_barrier(%{{.+}}* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]])
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
// CHECK: ret void
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-CHECK: [[GTID_ADDR:%.+]] = load i32*, i32** [[GTID_ADDR_ADDR]],
// TLS-CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_ADDR]],
[OPENMP] Emit __kmpc_cancel_barrier() and code for 'cancellation point' only if 'cancel' is found. Patch improves codegen for OpenMP constructs. If the OpenMP region does not have internal 'cancel' construct, a call to 'void __kmpc_barrier()' runtime function is generated for all implicit/explicit barriers. If the region has inner 'cancel' directive, then ``` if (__kmpc_cancel_barrier()) exit from outer construct; ``` code is generated. Also, the code for 'canellation point' directive is not generated if parent directive does not have 'cancel' directive. llvm-svn: 247681
2015-09-15 20:52:43 +08:00
// TLS-CHECK: call {{.*}}void @__kmpc_barrier(%{{.+}}* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]])
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-CHECK: ret void
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00
#endif
[OPENMP] Fixed codegen for copying/initialization of array variables/parameters. This modification generates proper copyin/initialization sequences for array variables/parameters. Before they were considered as pointers, not arrays. llvm-svn: 237691
2015-05-19 20:31:28 +08:00
#else
// ARRAY-LABEL: array_func
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-ARRAY-LABEL: array_func
[OPENMP] Fixed codegen for copying/initialization of array variables/parameters. This modification generates proper copyin/initialization sequences for array variables/parameters. Before they were considered as pointers, not arrays. llvm-svn: 237691
2015-05-19 20:31:28 +08:00
struct St {
int a, b;
St() : a(0), b(0) {}
St &operator=(const St &) { return *this; };
~St() {}
};
void array_func() {
static int a[2];
static St s[2];
// ARRAY: @__kmpc_fork_call(
// ARRAY: call i8* @__kmpc_threadprivate_cached(
Change memcpy/memove/memset to have dest and source alignment attributes (Step 1). Summary: Upstream LLVM is changing the the prototypes of the @llvm.memcpy/memmove/memset intrinsics. This change updates the Clang tests for this change. The @llvm.memcpy/memmove/memset intrinsics currently have an explicit argument which is required to be a constant integer. It represents the alignment of the dest (and source), and so must be the minimum of the actual alignment of the two. This change removes the alignment argument in favour of placing the alignment attribute on the source and destination pointers of the memory intrinsic call. For example, code which used to read: call void @llvm.memcpy.p0i8.p0i8.i32(i8* %dest, i8* %src, i32 100, i32 4, i1 false) will now read call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %dest, i8* align 4 %src, i32 100, i1 false) At this time the source and destination alignments must be the same (Step 1). Step 2 of the change, to be landed shortly, will relax that contraint and allow the source and destination to have different alignments. llvm-svn: 322964
2018-01-20 01:12:54 +08:00
// ARRAY: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %{{.+}}, i8* align 4 bitcast ([2 x i32]* @{{.+}} to i8*), i64 8, i1 false)
[OPENMP] Fixed codegen for copying/initialization of array variables/parameters. This modification generates proper copyin/initialization sequences for array variables/parameters. Before they were considered as pointers, not arrays. llvm-svn: 237691
2015-05-19 20:31:28 +08:00
// ARRAY: call dereferenceable(8) %struct.St* @{{.+}}(%struct.St* %{{.+}}, %struct.St* dereferenceable(8) %{{.+}})
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-ARRAY: @__kmpc_fork_call(
[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
2015-09-10 16:12:02 +08:00
// TLS-ARRAY: [[REFT:%.+]] = load [2 x i32]*, [2 x i32]** [[ADDR:%.+]],
// TLS-ARRAY: [[REF:%.+]] = bitcast [2 x i32]* [[REFT]] to i8*
Change memcpy/memove/memset to have dest and source alignment attributes (Step 1). Summary: Upstream LLVM is changing the the prototypes of the @llvm.memcpy/memmove/memset intrinsics. This change updates the Clang tests for this change. The @llvm.memcpy/memmove/memset intrinsics currently have an explicit argument which is required to be a constant integer. It represents the alignment of the dest (and source), and so must be the minimum of the actual alignment of the two. This change removes the alignment argument in favour of placing the alignment attribute on the source and destination pointers of the memory intrinsic call. For example, code which used to read: call void @llvm.memcpy.p0i8.p0i8.i32(i8* %dest, i8* %src, i32 100, i32 4, i1 false) will now read call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %dest, i8* align 4 %src, i32 100, i1 false) At this time the source and destination alignments must be the same (Step 1). Step 2 of the change, to be landed shortly, will relax that contraint and allow the source and destination to have different alignments. llvm-svn: 322964
2018-01-20 01:12:54 +08:00
// TLS-ARRAY: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 bitcast ([2 x i32]* @{{.+}} to i8*), i8* align 4 [[REF]], i64 8, i1 false)
[OpenMP] Add capture for threadprivate variables used in copyin clause if TLS is enabled in OpenMP code generation. llvm-svn: 243277
2015-07-28 00:38:06 +08:00
// TLS-ARRAY: call dereferenceable(8) %struct.St* @{{.+}}(%struct.St* %{{.+}}, %struct.St* dereferenceable(8) %{{.+}})
[OPENMP] Fixed codegen for copying/initialization of array variables/parameters. This modification generates proper copyin/initialization sequences for array variables/parameters. Before they were considered as pointers, not arrays. llvm-svn: 237691
2015-05-19 20:31:28 +08:00
#pragma omp threadprivate(a, s)
#pragma omp parallel copyin(a, s)
;
}
#endif
[OPENMP] Codegen for 'copyin' clause in 'parallel' directive. Emits the following code for the clause at the beginning of the outlined function for implicit threads: if (<not a master thread>) { ... <thread local copy of var> = <master thread local copy of var>; ... } <sync point>; Checking for a non-master thread is performed by comparing of the address of the thread local variable with the address of the master's variable. Master thread always uses original variables, so you always know the address of the variable in the master thread. Differential Revision: http://reviews.llvm.org/D9026 llvm-svn: 235075
2015-04-16 13:39:01 +08:00