Closed Tang-Haojin closed 4 weeks ago
Some profiling evidences here. When I add -X 10
, the CPU usage can reach 100% approximately.
[Generated by IPC robot] commit: 9357c6eea1c4201ee311d0695a56a7e4f221839b | commit | astar | copy_and_run | coremark | gcc | gromacs | lbm | linux | mcf | microbench | milc | namd | povray | wrf | xalancbmk |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
9357c6e | 1.924 | 0.450 | 2.701 | 1.198 | 2.812 | 2.465 | 2.404 | 0.914 | 1.394 | 1.626 | 3.424 | 2.745 | 2.428 | 3.271 |
master branch: | commit | astar | copy_and_run | coremark | gcc | gromacs | lbm | linux | mcf | microbench | milc | namd | povray | wrf | xalancbmk |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
1eae6a3 | 1.924 | 0.450 | 2.701 | 1.198 | 2.812 | 2.465 | 2.404 | 0.914 | 1.394 | 1.626 | 3.424 | 2.745 | 2.428 | 3.271 | |
f20002a | 1.936 | 0.450 | 2.701 | 1.198 | 2.812 | 2.465 | 2.404 | 0.914 | 1.394 | 1.626 | 3.424 | 2.745 | 2.428 | 3.271 | |
b2d1865 | 1.936 | 0.450 | 2.701 | 1.198 | 2.812 | 2.465 | 2.404 | 0.914 | 1.394 | 1.626 | 3.424 | 2.745 | 2.428 | 3.271 | |
c1ebb15 | 1.936 | 0.450 | 2.701 | 1.199 | 2.812 | 2.465 | 2.404 | 0.914 | 1.394 | 1.626 | 3.424 | 2.755 | 2.429 | 3.271 | |
b2e314e | 1.936 | 0.450 | 2.701 | 1.199 | 2.812 | 2.465 | 2.404 | 0.914 | 1.394 | 1.626 | 3.424 | 2.755 | 2.429 | 3.271 | |
5e0a22e | 1.924 | 0.450 | 2.701 | 1.199 | 2.812 | 2.465 | 2.404 | 0.914 | 1.394 | 1.626 | 3.424 | 2.755 | 2.429 | 3.271 | |
ca0aa83 | 1.924 | 0.450 | 2.701 | 1.199 | 2.812 | 2.465 | 2.404 | 0.914 | 1.394 | 1.626 | 3.424 | 2.755 | 2.429 | 3.271 | |
b2564f6 | 1.936 | 0.450 | 2.701 | 1.199 | 2.812 | 2.465 | 2.404 | 0.914 | 1.394 | 1.626 | 3.424 | 2.755 | 2.429 | 3.271 | |
cabb9f4 | 1.934 | 0.451 | 2.704 | 1.193 | 2.812 | 2.465 | 2.404 | 0.913 | 1.401 | 1.626 | 3.420 | 2.750 | 2.429 | 3.271 |
This change can help us to capture the waveform of problem caused by hardware stuck. Besides, since we enlarged mmap memory, it may take much more time on
fork()
syscall. Enlarging the fork interval here also results better simulation speed in some circumstances.