Question: What do the __config__.py settings effect?

[sef43]

Hello,
Would you be able to explain about the configuration settings here:
https://github.com/pyscf/gpu4pyscf/blob/master/gpu4pyscf/__config__.py

Specifically what changes between:

elif props['totalGlobalMem'] >= 16 * GB:
    min_ao_blksize = 128
    min_grid_blksize = 128*128
    ao_aligned = 32
    grid_aligned = 256
    mem_fraction = 0.9
    number_of_threads = 1024 * 80
# other gaming cards
else:
    min_ao_blksize = 64
    min_grid_blksize = 64*64
    ao_aligned = 32
    grid_aligned = 128
    mem_fraction = 0.9
    number_of_threads = 1024 * 80

I have been encountering intermittent out of memory issues: #75

I was running calculations fine on a 4090 24GB. And also mostly fine on a 12GB titan V (a few OOMS). I then tried running on V100 16GB and get many more of these errors: CUDA Error of GINTint2e_jk_kernel: out of memory

I am wondering if these settings are related?