News for hackers. Stuff that matters.

A comment on R49081 Is Prime. Parent comment

Re:R49081 Is Prime by pclmulqdq on Monday November 21, 2022 @02:32PM (#33696824)
To be clear, this problem (prime searching) works well on a single compute node. It doesn't work well if you have to scale to many nodes, because communication latency quickly bounds your speed.

That's why GIMPS is the way it is: with volunteers offering computers that are separated and do a siloed calculation.

[ Parent | on HN ]
Re:R49081 Is Prime by dragontamer on Monday November 21, 2022 @05:49PM (#33699248)
I can believe that.

But the fact that this was done on a Threadripper 3990x, a high-core count (but slow interconnect / core-to-core communication) kind of computer, makes me believe that there's a degree of parallelism in this problem.

Otherwise, another computer (ie: higher GHz / less core count) could have been used instead.

---------

I really don't know this specific algorithm: ECPP. It comes down to how that works... especially its parallelization.

[ Parent | on HN ]
Re:R49081 Is Prime by pclmulqdq on Monday November 21, 2022 @07:02PM (#33699984)
AFAIK it comes down to efficient FFTs for multiplying the large numbers, which are well-suited to GPUs, and this could likely have been done a lot faster on a GPU.

I'm guessing there just isn't a good CUDA implementation out there.

[ Parent | on HN ]