Chuck Yount

Default optimization

5

The README states that the default build for the OpenMP iso3dfd code creates the baseline w/o any optimizations. It appears that the default build is actually level-3 optimization. To build...

Enable streaming writes automatically

1

Enable streaming writes when a stencil writes an element w/o a preceding read.

After auto-tuner finishes on all ranks, test best setting on each rank and keep it if better.

Use non-temporal load when beneficial

1

For points that are not shared between neighbor stencils

Provide an interface to a package such as https://www.ospray.org/.

Template parameter should specify size.

Using https://github.com/LLNL/zfp or something similar.

Rewrite yask_main.cpp to work with APIs only

1

Will require adding APIs, removing functionality, and/or moving code from kernel/lib to yask_main.cpp. This is not critical, but it would showcase the API if we actually used it!