View Single Post
Old 2020-03-12, 19:20   #11
kuratkull's Avatar
Mar 2007

2·67 Posts


* fixed issue with gwstartnextfft false-negatives
* selftest success at least up to n=150000 (for k<300)
* print LLR64 compatible residue
* automatic niceness 19 (not configurable)
* extra logging when errors happen (won't catch all)
* result output says 'maybe' if errors were seen
* speed still on par with LLR64

So basically it seems to work - though I still wouldn't bet my head on it working in all cases (yet), though I currently do not know a case where it would fail silently. It currently crashes for large n's (>9M).

Also the argument format has changed, and setting threads to 0 causes RPT to auto-benchmark to find the fastest number of threads.

Currently my guarantee is that I --selftest 150000, which means all k<300 are tested upto known primes n=150000, and a negative case for each n is generated (n - 1 if it is not a known prime).
kuratkull is offline   Reply With Quote