mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   GPU Computing (https://www.mersenneforum.org/forumdisplay.php?f=92)
-   -   CUDALucas (a.k.a. MaclucasFFTW/CUDA 2.3/CUFFTW) (https://www.mersenneforum.org/showthread.php?t=12576)

msft 2012-02-24 02:07

Hi ,Karl M Johnson
[QUOTE=Karl M Johnson;290621]
The most important detail I forgot to mention - the device flag doesnt work along with "-r" flag, which is the residue test.
[/QUOTE]
$ CUDALucas -d 1 -r
Don't work ?

msft 2012-02-24 02:13

1 Attachment(s)
Ver 1.57
Add apsen's work.
Ctrl-C handler brought back.
checkpoint file write per 100000 iteration.

Brain 2012-02-24 04:47

[QUOTE=msft;290647]Ver 1.57
Add apsen's work.
Ctrl-C handler brought back.
checkpoint file write per 100000 iteration.[/QUOTE]
Neverending story... I'd prefer checkpoint every 10000 to 100000, best woulb be configurable, Will compile in 12 hours again.

msft 2012-02-24 05:13

[QUOTE=Brain;290674]Neverending story... I'd prefer checkpoint every 10000 to 100000, best woulb be configurable, Will compile in 12 hours again.[/QUOTE]
Anyway Ctrl-C signal make checkpoint file.
Only use every 10000 checkpoint file,When system crash!!!

msft 2012-02-24 05:44

[QUOTE=Brain;290674]Neverending story... I'd prefer checkpoint every 10000 to 100000, best woulb be configurable, Will compile in 12 hours again.[/QUOTE]
Understand.
I add -c option,and Brain compile 12 hour later.
PS. I catch sign from gpuLucas.

LaurV 2012-02-24 06:46

I don't get it, the v1.3alpha_eoc output format was PERFECT. The -c switch [B]worked nice[/B], with any number of iterations, it shows the [B]ETA[/B], the [B]REAL spent time[/B], the ms/iter with 4 ([B]FOUR[/B]) decimals... etc. (a small problem scrambling residues, however).

Why can't you guys copy the i/o stuff from there?:surprised

Karl M Johnson 2012-02-24 07:12

[QUOTE=msft;290645]Hi ,Karl M Johnson

$ CUDALucas -d 1 -r
Don't work ?[/QUOTE]

That works.
So the program wants device flag first, then everything else.
It's perfectly fine, since it's stated if you run CUDALucas without any flags.
Case closed.

msft 2012-02-24 09:19

1 Attachment(s)
Ver 1.58
LaurV's Functional specification.

LaurV 2012-02-24 10:10

[QUOTE=msft;290705]Ver 1.58
LaurV's Functional specification.[/QUOTE]

Billion thanks! ([URL="http://www.youtube.com/watch?v=1dkhZhCryX0&feature=related"]link[/URL]).

I am dying to reach home (in few hours) after work to give it a try!

Brain 2012-02-24 12:22

[QUOTE=LaurV;290693]I don't get it, the v1.3alpha_eoc output format was PERFECT. The -c switch [B]worked nice[/B], with any number of iterations, it shows the [B]ETA[/B], the [B]REAL spent time[/B], the ms/iter with 4 ([B]FOUR[/B]) decimals... etc. (a small problem scrambling residues, however).

Why can't you guys copy the i/o stuff from there?:surprised[/QUOTE]
@LaurV: You sound a bit rude, without msft, we would have nothing! Ahh, the thanks compensates it....
@msft: Your enduring work is very appreciated...

Case closed.

LaurV 2012-02-24 14:15

[QUOTE=Brain;290715]@LaurV: You sound a bit rude.[/QUOTE]
You should know me better after an year of "forumming" around... My rudeness is usually expressed by completely ignoring the threads. If I started arguing about something, that means I care about it, and I want to see it a better product... For example yafu evolved to the actual form partially due to my rudeness in the yafu thread. I stepped on bsquared toes until he got pissed of me :D

How can you be offended by a guy from the internet, which you don't know, and he doesn't know you for real? People tell a lot of sheesh...

by the way, any binaries? :P:P


All times are UTC. The time now is 23:09.

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