Name		Name	Last commit message	Last commit date
parent directory ..
images		images
README.md		README.md
__init__.py		__init__.py
profile_utils.py		profile_utils.py
third-party-programs.txt		third-party-programs.txt

README.md

oneDNN verbose log parser

This parser tool helps users to analyze oneDNN verbose logs with some statistic charts and a timeline chart.

A pie chart of oneDNN primitive types breakdown will be generated as the below diagram.

A pie chart of oneDNN JIT kernels breakdown will be generated as the below diagram.

A oneDNN.json file will be generated, and users could view the timeline result by loading oneDNN.json fie into chrome://tracing tool as the below diagram.

Prerequisites

users need to get a oneDNN verbose log with display timestamp enabled from their workloads first.
Please enable both oneDNN verobse log with timestamp by following commands.
$export DNNL_VERBOSE=1
$export DNNL_VERBOSE_TIMESTAMP=1

How to parse logs

Raw log from frameworks like tensorflow or pytorch

parse a raw log "log.txt" from workload : $python profile_utils.py log.txt
- users will see output from console
- users will also get some pie chart diagram PNG files like typeTime Breakdown.png
- users will also get a parsed output mkldnn_log.csv which only contains onednn logs

Pure oneDNN log or parsed ouput 'mkldnn_log.csv'

parse a onednn log "mkldnn_log.csv" : $python profile_utils.py mkldnn_log.csv
- users will see output from console
- users will also get some pie chart diagram PNG files like typeTime Breakdown.png

Compare two pure oneDNN logs

compare two onednn log "a.csv" and "b.csv" : $python profile_utils.py a.csv b.csv
- users will see output from console
- users will also get a bar chart diagram PNG files like typeTime Comparison.png