This project will no longer be maintained by Intel.
Intel will not provide or guarantee development of or support for this project, including but not limited to, maintenance, bug fixes, new releases or updates.
Patches to this project are no longer accepted by Intel.
If you have an ongoing need to use this project, are interested in independently developing it, or would like to maintain patches for the community, please create your own fork of the project.
The target of this project is to provide a time domain performance profiler to analysis high concurrent deep learning workloads.
The profiler provide apis to embedded log point into your workload and an automatic analyzer to visualize the performance and generate potential optimization suggestion.
We provide example codes of using the api to use the profiler, please see details in the examples directory
Please follow doc/Install.md.
Please follow examples/cpp_example/README.md for how to trace a C++ program
We provide an example processed timeline by the profiler for the DLRM workloads under examples/tensorflow/dlrm_example/example_log/.
The timeline is based on DLRM executed on Tensorflow 1.15.0. With original timeline captured by Tensorflow profiler, it expands to profiler Intra-op threads, Core threads mappings and Per intra-op task micro-architecture telemetry.
You can view the timeline by your self with perfetto or chrome://tracing/. Simply load the .json file directly.