Closed p15-git-acc closed 2 years ago
With the improved K=K/2
timings, n=1e24
was successfully completed in only 14 hours (prec=256, 64 threads/32 cores
).
At 1e24
the number of zeros found starts to slightly decline, so K/2
could run out of steam here (or it is the prec
-setting). EDIT: at 1e25
it seems to pick up its pace and the number of zeros isolated increases again.
Note that just doubling B
only had adverse effects, so additional tuning of the other parameters seems required for this case.
see for example https://github.com/fredrik-johansson/arb/issues/335#issuecomment-701663788