llrCUDA - Page 34

VBCurtis · 2021-01-04, 03:30

No such program exists.

Stargate38 · 2021-01-04, 18:26

Why not? Technical limitations, or lack of interest? The only reason that I'm asking is because I recently got a new PC with a CUDA-capable GPU, and I want to see if I can speed up my searches with it, as GPUs are faster than CPUs when it comes to TF, LL, and N+/-1. Anyone willing to write such a program, preferably for free, even if I have to run it in Windows Bash? I'm not good at coding. Also, I'm hoping that Batalov's DivPhi gets released to the public someday (see https://www.mersenneforum.org/showthread.php?t=19725). I've been wanting to run that as well, and I'm sure others want to run it too.

VBCurtis · 2021-01-04, 18:34

Who knows? "why hasn't something been done yet?" is a matter of opinion. If you read this entire thread, you'll get a sense for how hard it is to get performant code on CUDA. That's the best answer you'll get, I think.

But, hey, may as well be yourself and ask someone to do dozens of hours of coding "For Free" because you want it. I mean, nobody was motivated to do it the last 5 years, but now that you've asked for free labor maybe they'll jump up and get coding.

Batalov · 2021-01-04, 18:35

Quote:

Originally Posted by Stargate38

...I'm not good at coding. Also, I'm hoping that DivPhi gets released to the public someday (see https://www.mersenneforum.org/showthread.php?t=19725). I've been wanting to run that as well.

Do you want it spoon-fed to you personally? Fedexed to you personally?
It was released half a year ago!

Stargate38 · 2021-01-04, 18:59

Oops, sorry. I didn't know it was integrated into a public build of LLR.

Jean Penné · 2021-02-12, 20:26

Hi All,

I released now a new GPU version of the LLR program on my personal page : jpenne.free.fr

No much new feature, but some improvements related to reliability and speed.
- By default, all tests on base two numbers use Gerbicz error checking.
This is the case for PRP Fermat and SPRP tests as in Prime95 or Mprime,
but also for the deterministic prime tests of Proth numbers.

LLR tests on Riesel numbers are only done after a positive Fermat PRP result.
Also, if b==2, k==+1 and abs(c)==1, a random shift on the PRP base is done.
It is especially interesting for the prime test of Gaussian Mersenne norms.
Like the previous versions, this code is fully C and C++ written, no Assembler code.

Large numbers (at least 1 mega digits) benefit more from the GPU
parallelism, but this program may also be used on smaller positive
results for verification...

For more details, would you see the Readme.txt file.

Please, let me know if you have any problem to run the binary on Linux and/or to build it on your system.

I wish you many successes in prime hunting!
Best Regards,
Jean

diep · 2021-02-12, 20:41

couldn't have chosen better time to release it was last night -14C outside here and had to turn up the heat in this office instead of profit from the watercooled gpu cooling the office!

Jean Penné · 2021-02-15, 20:07

Hi All,

While Fermat testing k*b^n+c large numbers with c<0, and Gerbicz error checking activated, the final computation of the residue used a call to the invg() function in the giants.c code.
This code being CPU only, it is very time consuming for mega_digits numbers and so, its use must be avoided.

That is the only fix done in this new build of llrCUDA.

Would you excuse me for this drawback, and Best Regards,
Jean

paulunderwood · 2021-02-15, 20:24

Quote:

Originally Posted by Jean Penné

Hi All,

While Fermat testing k*b^n+c large numbers with c<0, and Gerbicz error checking activated, the final computation of the residue used a call to the invg() function in the giants.c code.
This code being CPU only, it is very time consuming for mega_digits numbers and so, its use must be avoided.

That is the only fix done in this new build of llrCUDA.

Would you excuse me for this drawback, and Best Regards,
Jean

Jean, GMP has this function and should be fairly rapid,

Jean Penné · 2021-02-16, 06:00

Quote:

Originally Posted by paulunderwood

Jean, GMP has this function and should be fairly rapid,

I know that, Paul, but it will require some conversions...
Thanks!
Jean

Jean Penné · 2021-03-10, 20:35

Hi All,

When a too large round off error was encountered, the test was restarted from the beginning in "cufft only" mode, which was very penalizing about elapsed time.
I fixed that in this new build, so the restart is now from the last saved intermediate file.

Sorry for this drawback and Best Regards,

Jean

2021-01-04, 18:26	#365
Stargate38 "Daniel Jackson" May 2011 14285714285714285714 769 Posts	Why not? Technical limitations, or lack of interest? The only reason that I'm asking is because I recently got a new PC with a CUDA-capable GPU, and I want to see if I can speed up my searches with it, as GPUs are faster than CPUs when it comes to TF, LL, and N+/-1. Anyone willing to write such a program, preferably for free, even if I have to run it in Windows Bash? I'm not good at coding. Also, I'm hoping that Batalov's DivPhi gets released to the public someday (see https://www.mersenneforum.org/showthread.php?t=19725). I've been wanting to run that as well, and I'm sure others want to run it too. Last fiddled with by Stargate38 on 2021-01-04 at 18:32

2021-02-12, 20:26	#369
Jean Penné May 2004 FRANCE 270₁₆ Posts	llrCUDA version 3.8.3 is released! Hi All, I released now a new GPU version of the LLR program on my personal page : jpenne.free.fr No much new feature, but some improvements related to reliability and speed. - By default, all tests on base two numbers use Gerbicz error checking. This is the case for PRP Fermat and SPRP tests as in Prime95 or Mprime, but also for the deterministic prime tests of Proth numbers. LLR tests on Riesel numbers are only done after a positive Fermat PRP result. Also, if b==2, k==+1 and abs(c)==1, a random shift on the PRP base is done. It is especially interesting for the prime test of Gaussian Mersenne norms. Like the previous versions, this code is fully C and C++ written, no Assembler code. Large numbers (at least 1 mega digits) benefit more from the GPU parallelism, but this program may also be used on smaller positive results for verification... For more details, would you see the Readme.txt file. Please, let me know if you have any problem to run the binary on Linux and/or to build it on your system. I wish you many successes in prime hunting! Best Regards, Jean

2021-02-15, 20:07	#371
Jean Penné May 2004 FRANCE 2⁴·3·13 Posts	llrCUDA version 3.8.3 Build 2 is released! Hi All, While Fermat testing k*b^n+c large numbers with c<0, and Gerbicz error checking activated, the final computation of the residue used a call to the invg() function in the giants.c code. This code being CPU only, it is very time consuming for mega_digits numbers and so, its use must be avoided. That is the only fix done in this new build of llrCUDA. Would you excuse me for this drawback, and Best Regards, Jean

2021-03-10, 20:35	#374
Jean Penné May 2004 FRANCE 2⁴×3×13 Posts	llrCUDA version 3.8.3 Build 3 is released! Hi All, When a too large round off error was encountered, the test was restarted from the beginning in "cufft only" mode, which was very penalizing about elapsed time. I fixed that in this new build, so the restart is now from the last saved intermediate file. Sorry for this drawback and Best Regards, Jean

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
LLRcuda	shanecruise	Riesel Prime Search	8	2014-09-16 02:09
LLRCUDA - getting it to work	diep	GPU Computing	1	2013-10-02 12:12

2021-01-04, 03:30	#364
VBCurtis "Curtis" Feb 2005 Riverside, CA 5854₁₀ Posts	No such program exists.

2021-01-04, 18:34	#366
VBCurtis "Curtis" Feb 2005 Riverside, CA 2×2,927 Posts	Who knows? "why hasn't something been done yet?" is a matter of opinion. If you read this entire thread, you'll get a sense for how hard it is to get performant code on CUDA. That's the best answer you'll get, I think. But, hey, may as well be yourself and ask someone to do dozens of hours of coding "For Free" because you want it. I mean, nobody was motivated to do it the last 5 years, but now that you've asked for free labor maybe they'll jump up and get coding.

2021-01-04, 18:59	#368
Stargate38 "Daniel Jackson" May 2011 14285714285714285714 1100000001₂ Posts	Oops, sorry. I didn't know it was integrated into a public build of LLR.

2021-02-12, 20:41	#370
diep Sep 2006 The Netherlands 807₁₀ Posts	couldn't have chosen better time to release it was last night -14C outside here and had to turn up the heat in this office instead of profit from the watercooled gpu cooling the office!