benchmark.bin fails silently for large grid size

[lgarrison]

Per the title, the benchmark seems to complete instantly if the fine grid gets too big:

lgarrison@workergpu142:~/spiner/bin/test$ ./benchmark.bin 64 6000
# ncoarse = 64
# nfine time/point (ns) L2 error
6000    2.03796296296296e-07    0.00000000000000e+00
lgarrison@workergpu142:~/spiner/bin/test$ echo $?
0

x-ref: openjournals/joss-reviews#4367, but this is low-priority and doesn't need to be resolved as part of the review.

[Yurlungur]

Interesting. I'll take a look at this. In what mode did you run this? On GPU? My guess is that the GPU is running out of memory when the benchmark attempts to allocate the arrays. But the fact there's no runtime error is disturbing.

[lgarrison]

Yes, Kokkos CUDA.

…

On Wed, Jun 15, 2022 at 2:43 PM Jonah Miller ***@***.***> wrote: Interesting. I'll take a look at this. In what mode did you run this? On GPU? My guess is that the GPU is running out of memory when the benchmark attempts to allocate the arrays. But the fact there's no runtime error is disturbing. — Reply to this email directly, view it on GitHub <#38 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABLA7S3VSBD7VGIVQLJY233VPIP3ZANCNFSM5Y4ILP7A> . You are receiving this because you authored the thread.Message ID: ***@***.***>

-- Lehman Garrison Flatiron Research Fellow, Cosmology X Data Science Group Center for Computational Astrophysics, Flatiron Institute lgarrison.github.io

[Yurlungur]

Cool, thanks for the pointer. I'll look at this.