Newer milestone thread - Page 84

LaurV · 2013-12-02, 16:44

Quote:

Originally Posted by ET_

I'd like to do some of them too, but it seems that GPU72 can't handle them. I guess I should reserve them on Primenet...

Edit: No way

I should find another way to get them.

Luigi

BTW, a shared list of valid exponents would be nice to work with...

Here is a misfit list, the other (717 exponents) are already queued in my 4 copies of mfaktc. I won't do more of them (for the immediate future) so please feel free do divide the list between any potential contributors.

ET_ · 2013-12-02, 17:42

Now I have 2 distinct answers!

@Chalsall: If you plan to add small mersennes' trial-factoring, be sure to add a warning for the user, to choose a mfaktc version below 20 using "less classes".

@LaurV: I'm going to wait for Chalsall's response, then I'll do my part on this (and other) range(s).

A big thank you to both of you!

Luigi

chalsall · 2013-12-02, 17:58

Quote:

Originally Posted by ET_

@Chalsall: If you plan to add small mersennes' trial-factoring, be sure to add a warning for the user, to choose a mfaktc version below 20 using "less classes".

1. Copy. Could I further ask you to give exact language for the warning? Possibly with a link (perhaps to a specialized thread in this forum) explaining why, and how to optimize the environment for this specialized kind of work?

2. Could you, LaurV, and anyone else who wants to play in this space let me know what ranges you want to work, and to what TF level?

3. Could you and LaurV please continue to work the space you've assigned between yourselves? I will endeavor to bring online into GPU72 the ranges for this kind of work you provide from #2 above in the interim.

4. Sorry for the shortness. In multitasking mode.

kracker · 2013-12-02, 18:10

+1 on small ranges.

Would love to work on some.

ET_ · 2013-12-02, 19:52

Quote:

Originally Posted by chalsall

1. Copy. Could I further ask you to give exact language for the warning? Possibly with a link (perhaps to a specialized thread in this forum) explaining why, and how to optimize the environment for this specialized kind of work?

2. Could you, LaurV, and anyone else who wants to play in this space let me know what ranges you want to work, and to what TF level?

3. Could you and LaurV please continue to work the space you've assigned between yourselves? I will endeavor to bring online into GPU72 the ranges for this kind of work you provide from #2 above in the interim.

4. Sorry for the shortness. In multitasking mode.

1 - The idea came from this posts in the GPU Trial factoring FAQ thread:
http://www.mersenneforum.org/showpos...&postcount=363
http://www.mersenneforum.org/showpos...&postcount=364

Also, there is a brief interlude here.

I think LaurV and The Judger may have more infos about it.

As the range to be tested is short, it would be optimal to:
a) Compile (or get) mfaktc-019 with "MORE_CLASSES" undefined.
b) Run more than one instance of the program to maximize GPU throughput

2 - Once I get my hands on a working executable (I should have v0-.19 surce code to recompile) I will take the first 100 exponents of the list, just to be sure there is space for others.

3 - Of course! See section 2...

4 - Me too (C)

Edit: Forgot to ask: what should I do after completimg mmy first range? I.E. how to contact a server?
Edit 2: Never mind, I realized it is sufficient to send the result via the manual results page.

Luigi

chalsall · 2013-12-02, 20:06

Quote:

Originally Posted by ET_

2 - Once I get my hands on a working executable (I should have v0-.19 surce code to recompile) I will take the first 100 exponents of the list, just to be sure there is space for others.

No one has yet directly answered my question.

What range(s) do you guys want GPU72 to facilitate? And to what level?

Multiple ranges and depths can be worked simultaneously. But I need to know what you want.

Please present these requests in CSV format, or something similar. E.g. "10M,11M,65".

ET_ · 2013-12-02, 20:11

Quote:

Originally Posted by chalsall

No one has yet directly answered my question.

What range(s) do you guys want GPU72 to facilitate? And to what level?

Multiple ranges and depths can be worked simultaneously. But I need to know what you want.

Please present these requests in CSV format, or something similar. E.g. "10M,11M,65".

I have no particular requests, though I'm eager to contribute GPU time. From what I read, I think the actual step would be clearing the range 2M-3M up to 64 bits, but I have no plans ready to discuss.

Luigi

chalsall · 2013-12-02, 20:19

Quote:

Originally Posted by chalsall

Please present these requests in CSV format, or something similar. E.g. "10M,11M,65".

Sorry all, that was overly short...

At the same time I was typing that I was also trying to explain to a North American that Barbados is a sovereign state, and not a protectorate of the United States....

petrw1 · 2013-12-02, 20:42

My 2 cents ... that small range could be cleared faster by a big GPU than the time it takes to talk about it. I vote we keep GPU72 more focused and less special cases.

flashjh · 2013-12-02, 21:32

Quote:

Originally Posted by ET_

a) Compile (or get) mfaktc-019 with "MORE_CLASSES" undefined.

Why use .19 instead of .20? I can compile something if you let me know what you need.

ET_ · 2013-12-02, 21:53

Quote:

Originally Posted by flashjh

Why use .19 instead of .20? I can compile something if you let me know what you need.

I'm on Linux... :-)

I chose 0.19 because I read somewhere that you could not use v0.20 for this task. I'm glad I was found in error

Do you just undefine the MORE_CLASSES definitin on mfaktc and run the sieve on the GPU?

Edit: never mind, I figured it out by myself: it was on params.h, and there is still a typographic glitch on the screen.

Luigi

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Newer X64 build needed	Googulator	Msieve	73	2020-08-30 07:47
Performance of cuda-ecm on newer hardware?	fivemack	GMP-ECM	14	2015-02-12 20:10
Cause this don't belong in the milestone thread	bcp19	Data	30	2012-09-08 15:09
Newer msieves are slow on Core i7	mklasson	Msieve	9	2009-02-18 12:58
Use of large memory pages possible with newer linux kernels	Dresdenboy	Software	3	2003-12-08 14:47

2013-12-02, 17:42	#915
ET_ Banned "Luigi" Aug 2002 Team Italia 4819₁₀ Posts	Now I have 2 distinct answers! @Chalsall: If you plan to add small mersennes' trial-factoring, be sure to add a warning for the user, to choose a mfaktc version below 20 using "less classes". @LaurV: I'm going to wait for Chalsall's response, then I'll do my part on this (and other) range(s). A big thank you to both of you! Luigi

2013-12-02, 18:10	#917
kracker "Mr. Meeseeks" Jan 2012 California, USA 100001111000₂ Posts	+1 on small ranges. Would love to work on some.

2013-12-02, 20:42	#922
petrw1 1976 Toyota Corona years forever! "Wayne" Nov 2006 Saskatchewan, Canada 3×5×313 Posts	My 2 cents ... that small range could be cleared faster by a big GPU than the time it takes to talk about it. I vote we keep GPU72 more focused and less special cases.