During europython one Intel guy show how to profile python code with the new vtune amplifier. I didn't manage to reproduce the example, I think I missed something during the project configuration (and I didn't find the video of the talk online).
Can you provide a simple step by step tutorial to profile python code (and see the python call not the python-c-api call).
More precisely does it requires to install a specific python version ? (like the one provided by intel?)
Do I need to add some specific entry in Binary/Symbol Search configuration ?
Do I need to add some specific entry in Source Search configuration ?
My current project configuration:
Application parameters: path_to_a_pythonfile.py
Managed code profiling mode: mixed
Analysis type: Advanced HotSpot
I managed (after digging a bit in the config xml file) to get the python hospot (using AMPLXE_EXPERIMENTAL=python-hotspots ./amplxe-gui then creating a new analysis based on python-hostpots.. I get what I expected). Now I am going to play a bit with it, thanks.
in order to enable Python profiler feature, do the following:
1. Enable the feature environment variable.
- On the Windows* OS: set AMPLXE_EXPERIMENTAL=python-hotspots
- On the Linux* OS: export AMPLXE_EXPERIMENTAL=python-hotspots
2. Launch the VTune Amplifier XE GUI: amplxe-gui
3. Choose Launch application or Attach to application
4. Select an application with Python* code. For example: application = python.exe, arguments=path/to/somescript.py
5. Set MRTE mode to Managed
6. Choose Algorithm Analysis > Python Hotspots to profile in pure-Python mode.
For Python related questions with VTune Amplifier XE, please see the Intel(R) Distribution for Python Forum (https://software.intel.com/en-us/forums/intel-distribution-for-python).