As the thread title says, in linux, running Genefer on Nvidia GPU uses a full CPU. Is there anything in the works to fix this?
Judging from Iain's post here, he was looking into it.
But that post was 60 days ago, so I don't know if it's moving along. Either way, Iain is the person to implement this.
Ok. Maybe we need input from creator(s) of AP27. That one uses very little CPU. PPS-sieve also, but I think that one is CUDA.
EDIT:- GFNSvOCL as well with the help of W1 flag (I don't have direct experience running it, but saw it in the forums)
there is a fix described in the complete thread (maybe read it in total); I tried to gave the most important post here
Ok, I haven't looked at that in any detail. Any reason why the fix can't be done in the program itself? The other OpenCL apps are well behaved. Is there anything fundamental about Genefer that prevents the same behavior?
4)
Message boards :
Generalized Fermat Prime Search :
Is there any plan to fix 100% cpu usage on linux?
(Message 123085)
Posted 1761 days ago by axn
As the thread title says, in linux, running Genefer on Nvidia GPU uses a full CPU. Is there anything in the works to fix this?
6)
Message boards :
Sieving :
Bad n value
(Message 119971)
Posted 1872 days ago by axn
Must be an older version of the program with different n limits. Just running the program without any parameter will print the help. What does it say in terms of supported n ?
Something wrong with users "Team" column.
Mine is showing "The Knights Who Say Ni!", should be "PrimeSearchTeam".
And I noticed errors in "Team" for other users too.
--
Lumiukko
Yep. The pattern of error is a simple off-by-one error, based on the very first challenge. The n-th user in challenge 1 is associated with (n+1)-th user's team.
PPS-Sieve is the least important of the GPU projects. We'd therefore prefer people run AP27 or any of the GFN tasks, with larger GFN tasks prefered over the shorter tasks. PPS-Sieve is very far ahead of its associated LLR projects and the sieving currently being done won't be useful for a while.
Unfortunately, due to the _huge_ credit difference, this is unlikely to happen for the vast majority of people, whose only/main focus is credits.
