The sieve was done via BOINC and the post-processing was done on 576 cores of the Lonestar cluster with msieve MPI version. Linear algebra phase took 170 hours with matrix size of ~57.6M (~25 GB).
Log file in the attachment for the curious. You can see how much time filtering took place, peak of memory, number of processors used on LA phase, etc...hope now you can understand how hard is to factor some big integers. I will ask Greg how much time sieve took place via BOINC, probably 3-5 months on ~1000 cores.
Last edited by pinhodecarlos; 10-13-12 at 06:07 AM.
Reply With Quote