duplicate post. edited …
poor choice of words on my part. I meant that most criticisms I might make have already been stated in one form or another either here or on the mersenne forum.
I’m also quite “sceptical” of that.
I never thought I would get to know the purpose behind the mysterious 2.592e8 CPU year compute task. So this and the Mersenne forum thread made for an entertaining read, together with some other old forum threads over here.
I’d be curious to know what speedup has been achieved for the GPU version, as well as if any other algorithmic changes have shortened the expected runtime.
Also I’d be interested in the rationale for limiting the circle of potential contributors to owners of five CUDA-capable devices within the same box. I’d suspect the fraction of the CUDA compute power in such setups to be somewhat smaller than the total installed base.
In response to what Captain_Obvious stated on the previous page:
Putting up a project up on BOINC doesn’t guarentee that there are no malicious intentions by the project author. People simply have to trust that your BOINC project description exactly matches what your binaries perform and that there are no additional payloads included.
A BOINC admin can only intervene once malicious intent has been suspected and substantiated by some user. That would already be too late for people having joined the project previously.
I’m interested in trying out your code. Where is the .zip file available to download?
Added in edit:
Oops, silly me. I found your earlier Dropbox link.