Created by: baiyfbupt

我在动态图中用profileer

with profiler.profiler('All', 'total', '/tmp/profile', 'AllOpDetail') as prof:

结果如下：

------------------------->     Profiling Report     <-------------------------

Note! This Report merge all thread info into one.
Place: All
Time unit: ms
Sorted by total time in descending order in the same thread

Total time: 85.3118
  Computation time       Total: 0           Ratio: 0%
  Framework overhead     Total: 85.3118     Ratio: 100%

-------------------------     GpuMemCpy Summary     -------------------------

GpuMemcpy                Calls: 1592        Total: 67.659      Ratio: 79.3078%
  GpuMemcpyAsync         Calls: 1581        Total: 67.3125     Ratio: 78.9017%
  GpuMemcpySync          Calls: 11          Total: 0.346447    Ratio: 0.406095%

-------------------------       Event Summary       -------------------------

Event                                    Calls       Total       CPU Time (Ratio)        GPU Time (Ratio)        Min.        Max.        Ave.        Ratio.
GpuMemcpyAsync:CPU->GPU                  1507        60.6811     59.008674 (0.972439)    1.672417 (0.027561)     0.016008    10.3689     0.0402662   0.711286
BufferedReader:MemoryCopy                26          21.8826     19.813381 (0.905441)    2.069192 (0.094559)     0.563868    1.1884      0.841637    0.256501
  GpuMemcpyAsync:CUDAPinned->GPU         52          4.22972     2.160526 (0.510797)     2.069192 (0.489203)     0.01446     0.225421    0.0813407   0.0495795
GpuMemcpyAsync(same_gpu):GPU->GPU        22          2.40172     2.373141 (0.988102)     0.028576 (0.011898)     0.039795    1.46456     0.109169    0.0281522
GpuMemcpySync(same_gpu):GPU->GPU         11          0.346447    0.330703 (0.954556)     0.015744 (0.045444)     0.028282    0.043144    0.0314952   0.00406095

------------------------->     Profiling Report     <-------------------------

Place: All
Time unit: ms
Sorted by total time in descending order in the same thread


-------------------------       Event Summary       -------------------------

Event                                             Calls       Total       CPU Time (Ratio)        GPU Time (Ratio)        Min.        Max.        Ave.        Ratio.
thread14::BufferedReader:MemoryCopy               13          10.5248     9.455296 (0.898379)     1.069541 (0.101621)     0.713474    0.976148    0.809603    1
  GpuMemcpyAsync:CUDAPinned->GPU                  26          2.0956      1.026063 (0.489626)     1.069541 (0.510374)     0.01446     0.225421    0.0806002   0.19911
thread13::BufferedReader:MemoryCopy               13          11.3577     10.358085 (0.911985)    0.999651 (0.088015)     0.563868    1.1884      0.873672    1
  GpuMemcpyAsync:CUDAPinned->GPU                  26          2.13411     1.134463 (0.531585)     0.999651 (0.468415)     0.016397    0.202192    0.0820813   0.1879
thread0::GpuMemcpyAsync:CPU->GPU                  1507        60.6811     59.008674 (0.972439)    1.672417 (0.027561)     0.016008    10.3689     0.0402662   0.956674
thread0::GpuMemcpyAsync(same_gpu):GPU->GPU        22          2.40172     2.373141 (0.988102)     0.028576 (0.011898)     0.039795    1.46456     0.109169    0.0378645
thread0::GpuMemcpySync(same_gpu):GPU->GPU         11          0.346447    0.330703 (0.954556)     0.015744 (0.045444)     0.028282    0.043144    0.0314952   0.00546194

只能看到memory copy/sync的结果，没有各个OP的，想问下是否动态图还不支持用profiler呢，如果不支持还有什么性能分析工具可用吗？