FazBrowse GitHub Viewer | Trending |
URL:
| Home
Tools: [Download Repo ZIP]   [Original HTTPS Page]

modified benchmark · fcccode/cpp-taskflow@61c650d · GitHub

/ cpp-taskflow Public
forked from taskflow/taskflow

Commit 61c650d

Browse files
Cong Hao
committed
modified benchmark
1 parent 28e90f5 commit 61c650d

11 files changed

Lines changed: 60 additions & 47 deletions

File tree

‎benchmark/graph_traversal/levelgraph.hpp‎

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -203,9 +203,9 @@ class LevelGraph {
203203

204204
};
205205

206-
std::chrono::microseconds measure_time_taskflow(LevelGraph&);
207-
std::chrono::microseconds measure_time_omp(LevelGraph&);
208-
std::chrono::microseconds measure_time_tbb(LevelGraph&);
206+
std::chrono::microseconds measure_time_taskflow(LevelGraph&, unsigned);
207+
std::chrono::microseconds measure_time_omp(LevelGraph&, unsigned);
208+
std::chrono::microseconds measure_time_tbb(LevelGraph&, unsigned);
209209

210210

211211

‎benchmark/graph_traversal/main.cpp‎

Lines changed: 12 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -1,30 +1,36 @@
11
#include "levelgraph.hpp"
22

3-
int main() {
3+
int main(int argc, char* argv[]) {
4+
5+
unsigned num_threads = std::thread::hardware_concurrency();
6+
7+
if(argc > 1) {
8+
num_threads = std::atoi(argv[1]);
9+
}
410

511
double omp_time {0.0};
612
double tbb_time {0.0};
713
double tf_time {0.0};
814
int rounds {5};
9-
15+
1016
std::cout << std::setw(12) << "|V|+|E|"
1117
<< std::setw(12) << "OpenMP"
1218
<< std::setw(12) << "TBB"
1319
<< std::setw(12) << "Taskflow"
1420
<< std::endl;
1521

16-
for(int i=1; i<=401; i += 4) {
22+
for(int i=1; i<=401; i += 10) {
1723

1824
LevelGraph graph(i, i);
1925

2026
for(int j=0; j<rounds; ++j) {
21-
omp_time += measure_time_omp(graph).count();
27+
omp_time += measure_time_omp(graph, num_threads).count();
2228
graph.clear_graph();
2329

24-
tbb_time += measure_time_tbb(graph).count();
30+
tbb_time += measure_time_tbb(graph, num_threads).count();
2531
graph.clear_graph();
2632

27-
tf_time += measure_time_taskflow(graph).count();
33+
tf_time += measure_time_taskflow(graph, num_threads).count();
2834
graph.clear_graph();
2935
}
3036

‎benchmark/graph_traversal/omp.cpp‎

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -4,9 +4,9 @@
44

55
#include "levelgraph.hpp"
66

7-
void traverse_regular_graph_omp(LevelGraph& graph){
7+
void traverse_regular_graph_omp(LevelGraph& graph, unsigned num_threads){
88

9-
omp_set_num_threads(std::thread::hardware_concurrency());
9+
omp_set_num_threads(num_threads);
1010

1111
#pragma omp parallel
1212
{
@@ -262,9 +262,9 @@ void traverse_regular_graph_omp(LevelGraph& graph){
262262
}
263263
}
264264

265-
std::chrono::microseconds measure_time_omp(LevelGraph& graph){
265+
std::chrono::microseconds measure_time_omp(LevelGraph& graph, unsigned num_threads){
266266
auto beg = std::chrono::high_resolution_clock::now();
267-
traverse_regular_graph_omp(graph);
267+
traverse_regular_graph_omp(graph, num_threads);
268268
auto end = std::chrono::high_resolution_clock::now();
269269
return std::chrono::duration_cast<std::chrono::microseconds>(end - beg);
270270
}

‎benchmark/graph_traversal/taskflow.cpp‎

Lines changed: 5 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -6,7 +6,7 @@
66

77
struct TF {
88

9-
TF(LevelGraph& graph) {
9+
TF(LevelGraph& graph, unsigned num_threads) : tf {num_threads} {
1010

1111
tasks.resize(graph.level());
1212
for(size_t i=0; i<tasks.size(); ++i) {
@@ -38,14 +38,14 @@ struct TF {
3838

3939
};
4040

41-
void traverse_level_graph_taskflow(LevelGraph& graph){
42-
TF tf(graph);
41+
void traverse_level_graph_taskflow(LevelGraph& graph, unsigned num_threads){
42+
TF tf(graph, num_threads);
4343
tf.run();
4444
}
4545

46-
std::chrono::microseconds measure_time_taskflow(LevelGraph& graph){
46+
std::chrono::microseconds measure_time_taskflow(LevelGraph& graph, unsigned num_threads){
4747
auto beg = std::chrono::high_resolution_clock::now();
48-
traverse_level_graph_taskflow(graph);
48+
traverse_level_graph_taskflow(graph, num_threads);
4949
auto end = std::chrono::high_resolution_clock::now();
5050
return std::chrono::duration_cast<std::chrono::microseconds>(end - beg);
5151
}

‎benchmark/graph_traversal/tbb.cpp‎

Lines changed: 6 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -11,9 +11,9 @@ using namespace tbb::flow;
1111

1212
struct TBB {
1313

14-
TBB(LevelGraph& graph) {
14+
TBB(LevelGraph& graph, unsigned num_threads) {
1515

16-
tbb::task_scheduler_init init(std::thread::hardware_concurrency());
16+
tbb::task_scheduler_init init(num_threads);
1717

1818
tasks.resize(graph.level());
1919
for(size_t i=0; i<tasks.size(); ++i) {
@@ -60,14 +60,14 @@ struct TBB {
6060
std::unique_ptr<continue_node<continue_msg>> source;
6161
};
6262

63-
void traverse_regular_graph_tbb(LevelGraph& graph){
64-
TBB tbb(graph);
63+
void traverse_regular_graph_tbb(LevelGraph& graph, unsigned num_threads){
64+
TBB tbb(graph, num_threads);
6565
tbb.run();
6666
}
6767

68-
std::chrono::microseconds measure_time_tbb(LevelGraph& graph){
68+
std::chrono::microseconds measure_time_tbb(LevelGraph& graph, unsigned num_threads){
6969
auto beg = std::chrono::high_resolution_clock::now();
70-
traverse_regular_graph_tbb(graph);
70+
traverse_regular_graph_tbb(graph, num_threads);
7171
auto end = std::chrono::high_resolution_clock::now();
7272
return std::chrono::duration_cast<std::chrono::microseconds>(end - beg);
7373
}

‎benchmark/wavefront/main.cpp‎

Lines changed: 10 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -1,6 +1,12 @@
11
#include "matrix.hpp"
22

3-
int main() {
3+
int main(int argc, char* argv[]) {
4+
5+
unsigned num_threads = std::thread::hardware_concurrency();
6+
7+
if(argc > 1) {
8+
num_threads = std::atoi(argv[1]);
9+
}
410

511
double omp_time {0.0};
612
double tbb_time {0.0};
@@ -23,9 +29,9 @@ int main() {
2329
init_matrix();
2430

2531
for(int j=0; j<rounds; ++j) {
26-
omp_time += measure_time_omp().count();
27-
tbb_time += measure_time_tbb().count();
28-
tf_time += measure_time_taskflow().count();
32+
omp_time += measure_time_omp(num_threads).count();
33+
tbb_time += measure_time_tbb(num_threads).count();
34+
tf_time += measure_time_taskflow(num_threads).count();
2935
}
3036

3137
destroy_matrix();

‎benchmark/wavefront/matrix.hpp‎

Lines changed: 5 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -17,8 +17,8 @@ inline double **matrix {nullptr};
1717

1818
// nominal operations
1919
inline double calc(double v0, double v1) {
20-
return (v0 == v1) ? std::pow(v0/v1, 4.0f) : std::max(v0,v1);
21-
//return std::max(v0, v1);
20+
//return (v0 == v1) ? std::pow(v0/v1, 4.0f) : std::max(v0,v1);
21+
return std::max(v0, v1);
2222
}
2323

2424
// initialize the matrix
@@ -56,8 +56,8 @@ inline void block_computation(int i, int j){
5656
}
5757

5858

59-
std::chrono::microseconds measure_time_taskflow();
60-
std::chrono::microseconds measure_time_omp();
61-
std::chrono::microseconds measure_time_tbb();
59+
std::chrono::microseconds measure_time_taskflow(unsigned);
60+
std::chrono::microseconds measure_time_omp(unsigned);
61+
std::chrono::microseconds measure_time_tbb(unsigned);
6262

6363

‎benchmark/wavefront/omp.cpp‎

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -5,7 +5,7 @@
55
int **D {nullptr};
66

77
// wavefront computation
8-
void wavefront_omp() {
8+
void wavefront_omp(unsigned num_threads) {
99

1010
// set up the dependency matrix
1111
D = new int *[MB];
@@ -16,7 +16,7 @@ void wavefront_omp() {
1616
}
1717
}
1818

19-
omp_set_num_threads(std::thread::hardware_concurrency());
19+
omp_set_num_threads(num_threads);
2020

2121
#pragma omp parallel
2222
{
@@ -73,9 +73,9 @@ void wavefront_omp() {
7373
delete [] D;
7474
}
7575

76-
std::chrono::microseconds measure_time_omp() {
76+
std::chrono::microseconds measure_time_omp(unsigned num_threads) {
7777
auto beg = std::chrono::high_resolution_clock::now();
78-
wavefront_omp();
78+
wavefront_omp(num_threads);
7979
auto end = std::chrono::high_resolution_clock::now();
8080
return std::chrono::duration_cast<std::chrono::milliseconds>(end - beg);
8181
}

‎benchmark/wavefront/taskflow.cpp‎

Lines changed: 6 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -2,9 +2,9 @@
22
#include <taskflow/taskflow.hpp>
33

44
// wavefront computing
5-
void wavefront_taskflow() {
5+
void wavefront_taskflow(unsigned num_threads) {
66

7-
tf::Taskflow tf;
7+
tf::Taskflow tf{num_threads};
88

99
std::vector<std::vector<tf::Task>> node(MB);
1010

@@ -22,17 +22,18 @@ void wavefront_taskflow() {
2222
block_computation(i, j);
2323
}
2424
);
25-
if(j+1 < NB) node[i][j].precede(node[i][j+1]);
25+
2626
if(i+1 < MB) node[i][j].precede(node[i+1][j]);
27+
if(j+1 < NB) node[i][j].precede(node[i][j+1]);
2728
}
2829
}
2930

3031
tf.wait_for_all();
3132
}
3233

33-
std::chrono::microseconds measure_time_taskflow() {
34+
std::chrono::microseconds measure_time_taskflow(unsigned num_threads) {
3435
auto beg = std::chrono::high_resolution_clock::now();
35-
wavefront_taskflow();
36+
wavefront_taskflow(num_threads);
3637
auto end = std::chrono::high_resolution_clock::now();
3738
return std::chrono::duration_cast<std::chrono::milliseconds>(end - beg);
3839
}

‎benchmark/wavefront/tbb.cpp‎

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -3,12 +3,12 @@
33
#include <tbb/flow_graph.h>
44

55
// the wavefront computation
6-
void wavefront_tbb() {
6+
void wavefront_tbb(unsigned num_threads) {
77

88
using namespace tbb;
99
using namespace tbb::flow;
1010

11-
tbb::task_scheduler_init init(std::thread::hardware_concurrency());
11+
tbb::task_scheduler_init init(num_threads);
1212

1313
continue_node<continue_msg> ***node = new continue_node<continue_msg> **[MB];
1414

@@ -42,9 +42,9 @@ void wavefront_tbb() {
4242
}
4343
}
4444

45-
std::chrono::microseconds measure_time_tbb() {
45+
std::chrono::microseconds measure_time_tbb(unsigned num_threads) {
4646
auto beg = std::chrono::high_resolution_clock::now();
47-
wavefront_tbb();
47+
wavefront_tbb(num_threads);
4848
auto end = std::chrono::high_resolution_clock::now();
4949
return std::chrono::duration_cast<std::chrono::milliseconds>(end - beg);
5050
}

0 commit comments

Comments
 (0)
Back | FazBrowse Home | New Git URL