Research Article
LTTng CLUST: A System-Wide Unified CPU and GPU Tracing Tool for OpenCL Applications
Table 4
Real application tracing timing.
| Width (pixel) | Height (pixel) | Base ave. (ns/iter.) | Base Std. dev. (ns/iter.) | Preload ave. (ns/iter.) | Preload Std. dev. (ns/iter.) | CLUST ave. (ns/iter.) | CLUST Std. dev. (ns/iter.) | CLUST + LTTng ave. (ns/iter.) | CLUST + LTTng Std. dev. (ns/iter.) |
| 1 | 1 | 35198 | 2162 | 36399 | 1941 | 50890 | 2297 | 58838 | 3364 | 10 | 1 | 35183 | 1916 | 35702 | 821 | 51883 | 2315 | 58265 | 3135 | 10 | 10 | 36031 | 1936 | 36890 | 1941 | 50758 | 1931 | 59619 | 3531 | 10 | 100 | 37937 | 1868 | 39067 | 169 | 55820 | 2731 | 61108 | 2645 | 100 | 100 | 56770 | 2440 | 59709 | 2135 | 75073 | 2661 | 84746 | 2232 | 1000 | 100 | 250694 | 3371 | 251165 | 3268 | 268726 | 3388 | 280299 | 4534 | 1280 | 720 | 1951826 | 4104 | 1951965 | 4443 | 1976916 | 4528 | 1988445 | 4747 | 1920 | 1080 | 4466096 | 6345 | 4466777 | 5597 | 4491589 | 5636 | 4511394 | 5509 |
|
|
Sample size = 1000; loop size = 100.
|