AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

API and Usage Documentation

Browse Source 
Change-Id: If34e497ff7938a352eecd4c1cd972cca5df6c5a3


[ROCm/rocprofiler commit: 24286477af]
This commit is contained in:
gobhardw
2023-05-29 23:59:59 +05:30
committed by Gopesh Bhardwaj
parent 7c3aa38dac
commit 1d95e00954
6 changed files with 2859 additions and 76 deletions
Download Patch File Download Diff File Expand all files Collapse all files
projects/rocprofiler/doc/rocprof_tool.md
+1 -1
View File
@@ -6,7 +6,7 @@ This command line tool is implemented as a script which is setting up the enviro
## 2. Profiling Modes
‘rocprof’ can be used for GPU profiling using HW counters and application tracing
### 2.1. GPU profiling
GPU profiling is controlled with input file which defines a list of metrics/counters and a profiling scope. An input file is provided using option ‘-i <input file>’. Output CSV file with a line per submitted kernel is generated. Each line has kernel name, kernel parameters and counter values. By option ‘—stats’ the kernel execution stats can be generated in CSV format. Currently profiling has limitation of serializing submitted kernels.
GPU profiling is controlled with input file which defines a list of metrics/counters and a profiling scope. An input file is provided using option ‘-i [input file]’. Output CSV file with a line per submitted kernel is generated. Each line has kernel name, kernel parameters and counter values. By option ‘—stats’ the kernel execution stats can be generated in CSV format. Currently profiling has limitation of serializing submitted kernels.
An example of input file:
```
# Perf counters group 1