Spencer Bryngelson
Spencer Bryngelson
I found a few things: We are still using `dble`, which converts to double precision: https://gcc.gnu.org/onlinedocs/gfortran/DBLE.html you can get rid of the calls to `dble`: ```console [I]shb-m1pro: MFC/src $ grep...
@anshgupta1234 would you mind giving this PR a review as well, just to check for typos that might be around? I looked, but another set of eyes would be useful....
``` @sbryngelson What do you think about moving the contents of m_precision_select into m_constants ? ``` @henryleberre I don't have particular preference about this, seems fine. ``` I wonder how...
is this ready for merge? @belericant
we should also check performance on cpu and gpu, perhaps @wilfonba can pass along his script for that. i would also be curious to see the performance numbers on one...
What happens if you use wp=single and precision (output) = 1 or 2, and if you use wp=double and precision = 1 or 2? Do any of these cases break?...
Ok I will merge once that is complete.
> Numerical Schlieren doesn't seem to work with single working precision. When I run with double working precision and save single precision silo files it's nearly uniformly 1 after the...
I don't know if these are reasonable or not, I haven't used the feature in a while.
@wilfonba > It looked like on GPUs fftw3 was compiled after I got these errors, not sure what's up with that. I've seen this recently as well. It doesn't build...