search

JLBLine / WODEN

CUDA code designed to simulate interferometric data using point, Gaussian, and shapelet sky models
https://woden.readthedocs.io/en/latest/
Mozilla Public License 2.0
5 stars 0 forks source link

JOSS review issue 9 (float vs double) #22

Closed JLBLine closed 2 years ago

JLBLine commented 2 years ago

Allow the user to choose between float and double precision. I'll do this by adding a switch that will assign float or double during compilation. This means the same code will always be used, just with different precision. I'll then get CMake to create two executables, woden_float and woden_double, that run_woden.py will be able to call depending on user arguments.

https://github.com/openjournals/joss-reviews/issues/3676

JLBLine commented 2 years ago

As of commit d745659cf36efe9ee6750fd075b5597b7ad0de00, the "double" or "float" choice is working in the code and unit testing, and has been written up into the JOSS paper. Still need to re-run all benchmarks, and need to update the Documentation in places.

JLBLine commented 2 years ago

@mkolopanis @plaplant as of commit 8b139527cbfbc934a2f92442a3b3778710c3ffff the documentation should be caught up to the float/double precision comparisons so I think this issue has been resolved now

mkolopanis commented 2 years ago

The documentation updates look good to me. Thanks for adding this feature.

plaplant commented 2 years ago

This is a very nice addition. Thanks for adding it!

JLBLine commented 2 years ago

Thanks all, I'll close this now.