Erich Focht
The py-veo Python API to vector engine offloading VEO allows the easy use of Notebooks for interactively working on the SX-Aurora vector engine. This post describes how Jupyter notebook can be installed and used and shows an example notebook with code running offloaded on the SX-Aurora VE.
Jupyter notebooks are very handy because they integrate programming, comments and results of running code into one nice document. The document, including the output of running pieces of programs can be saved for documenting the work. It is the normal way of approaching interactive data analysis that requires coding.
The following instructions are part of the README file in my github project py-veo-perf. It comes with one Jupyter notebook ready to load and run. It shows how one can actually measure performance counters of the executed VE kernel and computed some meaningful metrics like MOPS, MFLOPS, etc.
Install prerequisites
sudo yum install -y python-devel python-virtualenv
Create and configure a virtualenv containing jupyter
virtualenv jupy
source jupy/bin/activate
pip install --upgrade pip
pip install jupyter
pip install py-veosinfo
pip install numpy
pip install py-veo
pip install psutil
Create a password
python -c 'from notebook.auth import passwd ; password = passwd() ; print password'
#-> Enter password:
#-> Verify password:
#-> sha1:1b1452c8da2c:fc4572b136f1d4e0c2735e5b21a2d1593c5eb884
Generate a config and edit password
jupyter notebook --generate-config
#-> Writing default config to: /home/focht/.jupyter/jupyter_notebook_config.py
Modify c.NotebookApp.password in generated config, put in password hash
vim /home/focht/.jupyter/jupyter_notebook_config.py
Run the jupyter notebook server
jupyter notebook
Copy jupyter notebook and load it
Copy the file py-veo-perf.ipynb to the jupy/ directory, open the web browser pointing to localhost:8888, authenticate with the password and open the notebook file.
Needless to say, the machine on which you run this must have a VE card installed.
The notebook looks like this.