If you're crunching any of the GFN numbers on PRPNet right now, and you're paying attention, you might notice there's a few unusually small numbers being crunched right now.

I noticed that approximately 160 candidates in those N ranges had returned obviously bad residual values. Those candidates have been re-inserted into the queue to be processed again.

In all but one case, this was caused by the typical GeneferCUDA "maxErr" error -- except that although a computation error did occur, it wasn't caught by the maxErr precision check. That happens sometimes. Since there's no double checking, this kind of error isn't caught.

Note that these are CUDA-capable tests, so if it's configured on your system these will run on your GPU.

The one exception is on GFN32768 where the error occurred in Genefer80.

There's 1 retest being done on GFN32768, 140 on GFN65536, and 17 on GFN262144. Until the retests are complete (which shouldn't take long), the server statistics may look a bit weird because of the low b numbers.
____________
My lucky number is 75898⁵²⁴²⁸⁸+1