mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   GMP-ECM (https://www.mersenneforum.org/forumdisplay.php?f=55)
-   -   GPU_ECM seems to not use GPU (https://www.mersenneforum.org/showthread.php?t=20550)

Gordon 2015-10-17 19:14

GPU_ECM seems to not use GPU
 
I managed to find and download an apparently gpu enabled version, found the missing nvivia dll and tried it out with a prime95 stage 1 curve.

Result? - it did [U][B]ALL of the calculation on the cpu[/B][/U], didn't touch either gtx card once, even specifiying -gpudevice 0 made no difference at all...and it's about 25% slower than gmp-ecm v7...oh well

wombatman 2015-10-17 20:25

GPU-ECM uses the CPU for stage 2. It uses the gpu (with flag -gpu) for stage 1 only. The GPU-enabled stage 1 is also limited to numbers less than or equal to 2^1018-1 for the time being.

Gordon 2015-10-17 22:34

[QUOTE=wombatman;412953]GPU-ECM uses the CPU for stage 2. It uses the gpu (with flag -gpu) for stage 1 only. The GPU-enabled stage 1 is also limited to numbers less than or equal to 2^1018-1 for the time being.[/QUOTE]

So, as everything below M1019 has been factored, there isn't in fact any use for gpu_ecm?

wombatman 2015-10-17 22:58

For Mersenne, no. For other numbers, it works brilliantly.

MatWur-S530113 2015-10-21 10:10

[QUOTE=Gordon;412963]So, as everything below M1019 has been factored, there isn't in fact any use for gpu_ecm?[/QUOTE]

in addition to wombatman's answer:
the following exponents of M-numbers with already known factors have composite cofactors with 1018 or less bits (atm, of course):
1213, 1217, 1229, 1237, 1297

see f.e. factordb.com .

but I don't think it's worth to run stages 1 with the gpu on them with B1 less than 1e9 :wink:


All times are UTC. The time now is 06:50.

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