Msieve benchmarking

[storm5510]

Originally Posted by frmky

Then recompile with

make clean
make all CUDA=61 VBITS=256

Well, it didn't fare so well:

40 error detected in the compilation of "/tmp/tmp_xft_0000099f_00000000-8_sort_engine.ccpl.ii"
make[1]: *** [Makefile:106: sort_engine.so] Error 1
make[1]: *** [Makefile:360: cub/built] Error 2

It appears "clean" may have damaged the setup. I couldn't get it back to CUDA=75.

Rather than waste a lot of people's time, I will let this go. I learned a few things and this is what matters most.

Thank you all again!

[frmky]

Did you remember the

git checkout update_cub

[storm5510]

Originally Posted by frmky

Did you remember the

git checkout update_cub

I did. I wanted to test the GPU so I installed gpuOwl on it and it took right off. I looked at it about 40 minutes later and it is sitting on its BIOS setup screen. It had rebooted itself. That particular system has always been a little unstable at times. I have never been a big fan of UEFI. It has this and it can be disabled in the BIOS. I may try a clean setup with it disabled. It is not like I can hurt anything.

Thanks!

Edit: It has been my experience that the GPU can cause these reboots. It is underpowered in this setup. So, I have to reduce the maximum power draw. It is capable of 200W. I have had good luck running it at 160W. This made for a bit of performance drop-off, but not all that much. nvidia-smi is handy for making these adjustments.

[storm5510]

I installed 20.04 again, went through all the "apt" installs and run the compile. No errors!

I am still getting "error (line 1062): unexpected error when I try to run most of the "n" switches. "nc2" produces error generating for reading polynomials. I have seen this in Windows builds as well.

In any case, I feel like I accomplished something I was not sure I could do.

Edit: I do not know if this means anything. In the first line the program writes to the screen, it says "SVN unknown" I had problems with Windows builds which reported this.

[chris2be8]

Originally Posted by storm5510

Edit: I do not know if this means anything. In the first line the program writes to the screen, it says "SVN unknown" I had problems with Windows builds which reported this.

That's not a problem. From the makefile:

SVN_VERSION := $(shell svnversion .)
ifeq ($(SVN_VERSION),)
	SVN_VERSION := unknown
endif

If you did a svn checkout to fetch msieve that would set it to the version you checked out. All "SVN unknown" means is that you didn't get msieve that way. The code will work just as well without it.

[storm5510]

Originally Posted by chris2be8

That's not a problem. From the makefile:

SVN_VERSION := $(shell svnversion .)
ifeq ($(SVN_VERSION),)
	SVN_VERSION := unknown
endif

If you did a svn checkout to fetch msieve that would set it to the version you checked out. All "SVN unknown" means is that you didn't get msieve that way. The code will work just as well without it.

Thanks! I followed the web trail on Github to the author. I opened an "issue" regarding the errors I mentioned. I may, or may not, get a response.

[charybdis]

Originally Posted by storm5510

Thanks! I followed the web trail on Github to the author. I opened an "issue" regarding the errors I mentioned. I may, or may not, get a response.

You've already been talking to the author. He is "frmky". (Edit for clarification: he's only the author of that github repo with the GPU linear algebra and multithreaded square root. The main author of msieve is "jasonp".)

It looks like you're trying to run msieve NFS postprocessing without a polynomial file. What number are you trying to run it on? If you're trying to run it on the benchmarking job, have you copied the msieve.fb, msieve.dat and worktodo.ini files across to the directory that has the msieve binary?

[storm5510]

Originally Posted by charybdis

You've already been talking to the author. He is "frmky". (Edit for clarification: he's only the author of that github repo with the GPU linear algebra and multithreaded square root. The main author of msieve is "jasonp".)

It looks like you're trying to run msieve NFS postprocessing without a polynomial file. What number are you trying to run it on? If you're trying to run it on the benchmarking job, have you copied the msieve.fb, msieve.dat and worktodo.ini files across to the directory that has the msieve binary?

What I am trying to run is a number from James Heinrich's Factoring Project. It is a 128 digits in length. msieve never runs long enough to generate the msieve.??? files you mention. "Error line 1062." All I have is the worktodo.ini. All other files are in the same folder as the main program.

I know the GPU setup is correct. It ran gpuOwl overnight last night without problems.

[frmky]

Originally Posted by storm5510

"Error line 1062."

That would be the line

CUDA_TRY(cuModuleLoad(&t->gpu_module, "stage1_core.ptx"))

Make sure you have the files stage1_core.ptx and lanczos_kernel.ptx in the same directory in which you are running msieve as well as sort_engine.so and spmv_engine.so in the cub subdirectory.

If you do have stage1_core.ptx in the same directory as msieve, then what is the output of

head -n 15 stage1_core.ptx

[storm5510]

Originally Posted by frmky

...Make sure you have the files stage1_core.ptx and lanczos_kernel.ptx in the same directory in which you are running msieve as well as sort_engine.so and spmv_engine.so in the cub subdirectory.

If you do have stage1_core.ptx in the same directory as msieve, then what is the output of

head -n 15 stage1_core.ptx

All files mentioned above are present and in their proper place.

$ ./msieve head -n 15 stage1_core.ptx

deadline: 8640000 CPU-seconds per coefficient
error (line 1062): unexpected error

[kruoli]

In the case of the head ... command, please leave out the ./msieve part. Thanks!