MENUDescription
Let’s face it, analyzing large amounts of data can take seemingly forever. Animating images to a movie takes an entire weekend to complete. Why can’t Tecplot just do it for me?
This Webinar will show you how the power of Python can take advantage of multi-process parallelism. Using the Tecplot 360 Python API, PyTecplot, you can get your results up to six times faster. We’ll show results of two datasets:
- Deep water asteroid strike. 476 timesteps, each containing ~27 million grid points (structured). 707 Gb on disk. Data courtesy Los Alamos National Lab (https://oceans11.lanl.gov/deepwaterimpact/).
- Flow around a cylinder: 1434 timesteps, each containing ~1.3 million grid points (structured), 4 Gb on disk, Plot3D format.
Webinar Outline
The Problem
- Animating to a movie takes a long time and consumes too much RAM.
- Why can’t Tecplot 360 just do it for me? Much of the data pipeline is inherently single-threaded.
The Solution
- Minimize Memory Use (use if you’re ok with the time, but unhappy with the RAM consumption). Also consider changing $!FILECONFIG TEMPFILEPATH or updating your TEMP/TMP environment variable.
- Script with PyTecplot.
- Use Python multiprocessing.
The Results
- Given enough CPUs, get your results 6x faster.
- Here are a few Caveats.
- Speed is hardware dependent.
- Know your constraints – Disk I/O, CPU, RAM.
- Need enough licenses.
- Stylesheets need to work with the datasets.
- Future work includes testing on a distributed HPC system.
Thanks for watching!