mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Software (https://www.mersenneforum.org/forumdisplay.php?f=10)
-   -   CudaPm1 not working? (https://www.mersenneforum.org/showthread.php?t=23769)

robertfrost 2018-11-03 15:51

CudaPm1 not working?
 
When I run CudaPM1.exe I get:


Warning: ignoring line 1: "Test=332201" in "worktodo.txt". Reason: doesn't begin with Test= or DoubleCheck=.
No valid assignment found.


Which is something of a paradox! Any suggestions?

kriesel 2018-11-03 20:57

[QUOTE=robertfrost;499471]When I run CudaPM1.exe I get:

Warning: ignoring line 1: "Test=332201" in "worktodo.txt". Reason: doesn't begin with Test= or DoubleCheck=.
No valid assignment found.

Which is something of a paradox! Any suggestions?[/QUOTE]
CUDAPm1 was created from CUDALucas source.

You've hit an error message that did not get rewritten, in this code that the author described as alpha software (not even beta).
The worktodo assignment has a different form for CUDAPm1 than for CUDALucas.
See the CUDAPM1 readme. See also my reference threads at [url]https://www.mersenneforum.org/forumdisplay.php?f=154[/url]

robertfrost 2019-01-04 08:46

Setting BM1 and BM2
 
Thanks. I got it working! It wasn't clear at first I have to log in, before the entire set of posts you linked to are visible.


The key thing was: I had to put something different in the worktodo.txt file, but now it's run, it deleted that so I've lost what I entered to make it work! What's the syntax for an ECM p-1 test?

kriesel 2019-01-05 17:04

[QUOTE=robertfrost;504891]Thanks. ... What's the syntax for an ECM p-1 test?[/QUOTE]1) You're welcome, and welcome to the GIMPS forum.
2) I don't know; what color is an apple orange? :smile:

ECM and P-1 are two different factoring methods. P-1 is available for gpus. ECM is not available for gpus in implementations supporting enough bits for exponents of current GIMPS interest at the first test or double check wavefront.

If you're intending to ask about gmp-ecm's syntax for attempting P-1 factoring, I refer you to its documentation. Note, I tried it briefly long ago, and found it too slow to be useful, by comparison to prime95 and CUDAPm1, for exponents of current GIMPS wavefront size. It's my understanding it's intended for "smaller" numbers than GIMPS is concerned with currently. Every package has its strengths. GMP-ECM's usefulness lies elsewhere. [URL]https://www.mankier.com/1/gmp-ecm[/URL]

CUDAPm1 syntax for worktodo is shown in its readme. Or get a manual assignment from [URL]https://www.mersenne.org/manual_assignment/[/URL] (but only if you intend to complete it; if run time is too long for you, go to [URL]https://www.mersenne.org/workload/[/URL] to unreserve it) and copy and paste the assignment in. If you want a copy of the assignment saved, create your own separate assignment log text file, or put it in worktodo twice, and put # in front of one of them to make the line into a comment that will linger. Note, depending on what gpu you have, and choice of assignments, P-1 assignments can take hours, days, weeks, or months. (Years in extreme cases. Avoid those. Very long run times are unlikely to produce correct results. Also CUDAPm1 v0.20 fails to complete on highest exponents currently, above 176M to 430M depending on gpu model, and it can take days to months to hit the point of failure, wasting a long time, which is no fun. V0.22 may also have this issue.)


All times are UTC. The time now is 21:40.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.