CUDALucas - Compile on CUDA10.0

[saviourz]

Hi folks,

I am compiling CUDALucas 2.06 from the sourceforge (https://sourceforge.net/projects/cudalucas/files/). After ran makefile, I used $./CUDALucas -r 1 to test it's reliable or not.

Unfortunately, I got all residue [0000000000000].

My OS is Ubuntu 18.04 and I have already changed CUDA path in makefile and also --generate-code arch=compute_60, code=sm_60.

Message showed on the top of ./CUDALucas:
binary compiled for CUDA 10.10
CUDA runtime version 10.0
CUDA driver version 10.20

GPU type is Tesla V100-PCIE and driver version is 440.33.01.

I have read all related posts to my question but none of them can solve my problem. What's more, I set worktodo.txt as "Test=79437629" and got
"Illegal residue: 0x0000000000000000. See mersenneforum.org for help.".

Thanks in advance for replies and sorry if I posted in wrong place.

[axn]

Originally Posted by saviourz

My OS is Ubuntu 18.04 and I have already changed CUDA path in makefile and also --generate-code arch=compute_60, code=sm_60.

Message showed on the top of ./CUDALucas:
binary compiled for CUDA 10.10
CUDA runtime version 10.0
CUDA driver version 10.20

GPU type is Tesla V100-PCIE and driver version is 440.33.01.

V100 or P100? V100 is CC7.0, not 6.0, so probably try recompiling with that. Ideally, even 6.0 should work, as long as a PTX is available, but ...

[saviourz]

Originally Posted by axn

V100 or P100? V100 is CC7.0, not 6.0, so probably try recompiling with that. Ideally, even 6.0 should work, as long as a PTX is available, but ...

It's V100. I tried CC7.0 by setting "--generate-code arch=compute_70, code=sm_70".

But I still got "Illegal Residue 0x0000000".

Am I right for understanding how to set the CC version to 7.0?

[axn]

Originally Posted by saviourz

Am I right for understanding how to set the CC version to 7.0?

That part is correct.

EDIT:- Next thing I'd look for is the fact that you have multiple CUDA versions in your system:

binary compiled for CUDA 10.10
CUDA runtime version 10.0
CUDA driver version 10.20

During compilation, it is picking 10.10, but during execution, it is picking 10.0, but the driver is 10.20. I'm not sure what it all means.

Would it be possible to nuke all nvidia stuff from the system and do a clean install of latest drivers + cuda toolkit?

[saviourz]

Originally Posted by axn

That part is correct.

EDIT:- Next thing I'd look for is the fact that you have multiple CUDA versions in your system:

binary compiled for CUDA 10.10
CUDA runtime version 10.0
CUDA driver version 10.20

During compilation, it is picking 10.10, but during execution, it is picking 10.0, but the driver is 10.20. I'm not sure what it all means.

Would it be possible to nuke all nvidia stuff from the system and do a clean install of latest drivers + cuda toolkit?

Thanks for the advice! I'll try to make the version of those three unified to see the results.

[saviourz]

Originally Posted by axn

That part is correct.

EDIT:- Next thing I'd look for is the fact that you have multiple CUDA versions in your system:

binary compiled for CUDA 10.10
CUDA runtime version 10.0
CUDA driver version 10.20

During compilation, it is picking 10.10, but during execution, it is picking 10.0, but the driver is 10.20. I'm not sure what it all means.

Would it be possible to nuke all nvidia stuff from the system and do a clean install of latest drivers + cuda toolkit?

Hi, I have already unified those three CUDA things to 10.20 as:

binary compiled for CUDA 10.20
CUDA runtime version 10.20
CUDA driver version 10.20

But the problem is still here. "Illegal Residue 0x0000000". I am now trying to run it on Tesla M6 with "--generate-code arch=compute_60,code=sm_60" in Makefile.

Which part should I suspect now? I checked nvidia-smi and saw ./CUDALucas was successfully on the GPU task queue.

[axn]

Somebody faced similar problem, and found that the pre-compiled binary worked. Try reading the next few posts starting here: https://www.mersenneforum.org/showth...460#post497460

But people have made it work with native compilation as well. Maybe you can try PM-ing those folks. Not sure what platform they were using.

[saviourz]

Originally Posted by axn

Somebody faced similar problem, and found that the pre-compiled binary worked. Try reading the next few posts starting here: https://www.mersenneforum.org/showth...460#post497460

But people have made it work with native compilation as well. Maybe you can try PM-ing those folks. Not sure what platform they were using.

Thanks for your advice! I changed CUDA version to 9.1 and used Linux pre-compiled CUDALucas from https://download.mersenne.ca/CUDALucas/old. The output seems good. No 0 residue appears in the middle of the looping.

But I am still trying to figure out my compilation problem. Open to any advice!

[saviourz]

FYI,

I succeed in compiling CUDALucas v2.06 on Tesla M6 with CUDA runtime 10.20. The test is correct with "./CUDALucas -r 1".

I don't understand clearly about how the previous error happened. Only changes I made are in Makefile:

CUDA = "path to CUDA10.20"

--generate-code arch=compute_50, code=sm_50

I attached this executive file. Hope it can help.

[kriesel]

Originally Posted by saviourz

FYI,

I succeed in compiling CUDALucas v2.06 on Tesla M6 with CUDA runtime 10.20. The test is correct with "./CUDALucas -r 1".

I don't understand clearly about how the previous error happened. Only changes I made are in Makefile:

CUDA = "path to CUDA10.20"

--generate-code arch=compute_50, code=sm_50

I attached this executive file. Hope it can help.

Outstanding! Thanks for posting the executable file.
Welcome to the forum.
If you haven't seen it yet, you might find some background and reference info interesting at https://www.mersenneforum.org/showthread.php?t=24607
Please run some double-checks with that executable to confirm it produces correct runs.

[saviourz]

Originally Posted by kriesel

Outstanding! Thanks for posting the executable file.
Welcome to the forum.
If you haven't seen it yet, you might find some background and reference info interesting at https://www.mersenneforum.org/showthread.php?t=24607

Thanks for those useful information!