llrCUDA Version 3.8.5 is released!

[Jean Penné]

Hi All,

This GPU version of the LLR program has the same features as the previous 3.8.3, but I think that the updates made from it deserve to change the version number.
Here are some of these updates :

- In previous Version 3.8.4, one call to free() function was missing in
Gerbicz error checking code ; this caused an important memory leak...
This issue is now fixed here!

- K*2^n+c numbers can now be tested using IBDWT, even if abs(c) != 1
- All 64 bits residues match with those obtained with LLR 3.8.24
- several issues found on llrCUDA 3.8.3b3 are now fixed.

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

[ET_]

Originally Posted by Jean Penné

Hi All,

This GPU version of the LLR program has the same features as the previous 3.8.3, but I think that the updates made from it deserve to change the version number.
Here are some of these updates :

- In previous Version 3.8.4, one call to free() function was missing in
Gerbicz error checking code ; this caused an important memory leak...
This issue is now fixed here!

- K*2^n+c numbers can now be tested using IBDWT, even if abs(c) != 1
- All 64 bits residues match with those obtained with LLR 3.8.24
- several issues found on llrCUDA 3.8.3b3 are now fixed.

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

Where is the download link?

[sdbardwick]

Originally Posted by ET_

Where is the download link?

Not to step on Jean's toes, but this was provided in his other post: http://jpenne.free.fr/

[ET_]

Originally Posted by sdbardwick

Not to step on Jean's toes, but this was provided in his other post: http://jpenne.free.fr/

Thanks... my bad, I didn't look below the newest link

[Magellan3s]

I was attempting to use this software and got the following error "jesus@Magellan:~/CudaLucas/2/llrcuda385src$ ./llrCUDA -a333333991
*** buffer overflow detected ***: terminated
Aborted (core dumped)
"

[paulunderwood]

Originally Posted by Magellan3s

I was attempting to use this software and got the following error "jesus@Magellan:~/CudaLucas/2/llrcuda385src$ ./llrCUDA -a333333991
*** buffer overflow detected ***: terminated
Aborted (core dumped)
"

What did you expect? From the readme.txt: - With -a, you choose another set of .ini and results file.

What you should do is run the program with -m (or possibly -q if it is just one candidate)

[Magellan3s]

Originally Posted by paulunderwood

What did you expect? From the readme.txt: - With -a, you choose another set of .ini and results file.

What you should do is run the program with -m (or possibly -q if it is just one candidate)

Sorry, I got it running but I was unable to change the FFT length or test 100 million digit prime numbers.

[paulunderwood]

Originally Posted by Magellan3s

Sorry, I got it running but I was unable to change the FFT length or test 100 million digit prime numbers.

I am glad you got it running. Why do you want to change the FFT length? What is the form of the 100 million digit candidate prime numbers you have in mind?

[Magellan3s]

Originally Posted by paulunderwood

I am glad you got it running. Why do you want to change the FFT length? What is the form of the 100 million digit candidate prime numbers you have in mind?

The current FFT length for a wavefront Mersenne number I attempted to test ~ 113,687,731 selected an FFT of 15M which is much higher than the 6M which should be used for that particular exponent.

jesus@Magellan:~/CudaLLR/llrcuda385src$ ./llrCUDA -d -q"1*2^113687731-1"
Starting Fermat PRP test of 1*2^113687731-1
Using complex rational base DWT and generic reduction, FFT length = 15728640, a = 3
Iteration: 1 / 113687731 [0.00%].

I then attempted to test another number M333333991 but the program said the number was too big.

jesus@Magellan:~/CudaLLR/llrcuda385src$ ./llrCUDA -d -q"1*2^333333991-1"
Requested giant too big.

[paulunderwood]

Originally Posted by Magellan3s

The current FFT length for a wavefront Mersenne number I attempted to test ~ 113,687,731 selected an FFT of 15M which is much higher than the 6M which should be used for that particular exponent.

jesus@Magellan:~/CudaLLR/llrcuda385src$ ./llrCUDA -d -q"1*2^113687731-1"
Starting Fermat PRP test of 1*2^113687731-1
Using complex rational base DWT and generic reduction, FFT length = 15728640, a = 3
Iteration: 1 / 113687731 [0.00%].

I then attempted to test another number M333333991 but the program said the number was too big.

jesus@Magellan:~/CudaLLR/llrcuda385src$ ./llrCUDA -d -q"1*2^333333991-1"
Requested giant too big.

While it is admirable to attempt testing Mersenne numbers with llrCUDA, most people run gpuOwl for them. Timing comparisons would be interesting.

[Magellan3s]

Originally Posted by paulunderwood

While it is admirable to attempt testing Mersenne numbers with llrCUDA, most people run gpuOwl for them. Timing comparisons would be interesting.

Yes, this program interested me because it was specifically coded with CUDA!