GPU LL Testing FAQ

[fivemack]

pumpkin@pumpkin:~/cudalucas-build/cudalucas-code$ ./CUDALucas -d 1 -c 10000 1257787

------- DEVICE 1 -------
name                GeForce GTX 580
Compatibility       2.0
clockRate (MHz)     1566
memClockRate (MHz)  2010
totalGlobalMem      1607139328
totalConstMem       65536
l2CacheSize         786432
sharedMemPerBlock   49152
regsPerBlock        32768
warpSize            32
memPitch            2147483647
maxThreadsPerBlock  1024
maxThreadsPerMP     1536
multiProcessorCount 16
maxThreadsDim[3]    1024,1024,64
maxGridSize[3]      65535,65535,65535
textureAlignment    512
deviceOverlap       1

Using threads: square 256, splice 128.
Starting M1257787 fft length = 64K
|   Date     Time    |   Test Num     Iter        Residue        |    FFT   Error     ms/It     Time  |       ETA      Done   |
|  Nov 18  08:09:05  |   M1257787     10000  0x3f45bf9bea7213ea  |    64K  0.07812   0.0948    0.94s  |         1:58   0.79%  |
|  Nov 18  08:09:06  |   M1257787     20000  0x961d384390291afd  |    64K  0.07812   0.0962    0.96s  |         1:58   1.59%  |
...
|  Nov 18  08:11:05  |   M1257787   1250000  0x07869a6bb3f959a5  |    64K  0.07812   0.0977    0.97s  |         0:00  99.38%  |
M( 1257787 )P, n = 64K, CUDALucas v2.05.1, estimated total time = 2:02
pumpkin@pumpkin:~/cudalucas-build/cudalucas-code$ ./CUDALucas -d 0 -c 10000 1257787

------- DEVICE 0 -------
name                GeForce GTX 970
Compatibility       5.2
clockRate (MHz)     1215
memClockRate (MHz)  3505
totalGlobalMem      4294770688
totalConstMem       65536
l2CacheSize         1835008
sharedMemPerBlock   49152
regsPerBlock        65536
warpSize            32
memPitch            2147483647
maxThreadsPerBlock  1024
maxThreadsPerMP     2048
multiProcessorCount 13
maxThreadsDim[3]    1024,1024,64
maxGridSize[3]      2147483647,65535,65535
textureAlignment    512
deviceOverlap       1

Using threads: square 256, splice 128.
Starting M1257787 fft length = 64K
|   Date     Time    |   Test Num     Iter        Residue        |    FFT   Error     ms/It     Time  |       ETA      Done   |
|  Nov 18  08:13:08  |   M1257787     10000  0x0000000000000000  |    64K  0.00000   0.1163    1.16s  |         2:25   0.79%  |
|  Nov 18  08:13:09  |   M1257787     20000  0x0000000000000000  |    64K  0.00000   0.1171    1.17s  |         2:24   1.59%  |

So on the GTX580 it seems to be working OK (I am DCing M62046323 which should take four days and be a useful does-it-work test), and on the GTX970 it definitely isn't. Any hints?

[fivemack]

Originally Posted by Primeinator

Okay. Now that test week is over I am doing this. M(859433) is merrily being crunched by my GPU; however, it doesn't appear to be saving progress. I created a results.txt file in the same folder as CUDALucas. I am still obviously missing something

Edit: To clarify, pressing control+C does in fact close the program; however, upon restarting the program it starts the test from the beginning.

Edit 2: CUDALucas does not appear to write either to results.txt or create any new files to save results.

Edit 3: Running the program again I watched until the end of the test. When it finished CUDALucas closed. I checked the folder and alas, there was no edit made to the results.txt or any new file added displaying the result.

Have you installed CUDALucas in Program Files or something? This sounds as if it doesn't have permission to write to the place it's expecting to put its output.

[casmith789]

Um, I can't seem to download CUDALucas from its sourceforge page, it just creates a redirect loop without any download. I've tried it in Chrome, FF and Edge,

Is anyone else having this problem? I don't remember having this issue when I tried ~a year ago

[casmith789]

Well, I found my old backup of CUDALucas and it works fine on this (new) PC, so there isn't a rush. But it's weird that sourceforge doesn't work properly... :)

Tested it on a known prime and a known not prime so I know CUDALucas is working and I get these results (in case anyone was wondering):

CPU (Prime95, i7-4790, 4 cores running, 10.65 GHzd/d ish per core)
CPU (Prime95, i7-4790, 1 cores running, 13.83 GHzd/d ish per core)
GPU (nvidia gtx 970, CUDALucas 2.05.1, 20.75 GHzd/d ish)

GPU (nvidia gtx 970, mfaktc-0.21, 458 GHzd/d ish)

While these numbers will surely vary by exponent and the TFing is not done at the same range as I tested the LLs, it seems like LLing on the GPU is definitely some way off the TFing performance, and is in fact worse than the CPU for my machine if you take into account all my cores :)

[kladner]

The Sourceforge download just worked for me. Maybe try again?

[casmith789]

Originally Posted by kladner

The Sourceforge download just worked for me. Maybe try again?

Yeah, works now. Weird, I tried about 10 times with no success. Cheers

[kladner]

Originally Posted by casmith789

Yeah, works now. Weird, I tried about 10 times with no success. Cheers

I just now had multiple failures before I succeeded in downloading Ultradefrag from SourceForge. I was getting "Server Not Found" messages, on multiple mirrors. Then it cleared up. Something is a bit flaky over there.

[Blackadder]

Originally Posted by garo

Q. How do I get work for CUDALucas?
A. You will need to get manually get work from PrimeNet like mfaktc

You will need to get work manually get work from PrimeNet like mfaktc

[Blackadder]

Also, does this still work on Linux? I'm using Linux Mint 19 Xfce, but can't run the file: bash: ./CUDALucas-6.5: No such file or directory

[chris2be8]

1: Check CUDALucas-6.5 is marked executable.
2: Run file CUDALucas-6.5.
3: If it said its some sort of script check the first line. You may get that if the #! line points to a file that doesn't exist.
4: If file said it's an executable the run ldd CUDALucas-6.5 and check if it's looking for a run time library you don't have installed.

Chris

[Blackadder]

Originally Posted by chris2be8

1: Check CUDALucas-6.5 is marked executable.
2: Run file CUDALucas-6.5.
3: If it said its some sort of script check the first line. You may get that if the #! line points to a file that doesn't exist.
4: If file said it's an executable the run ldd CUDALucas-6.5 and check if it's looking for a run time library you don't have installed.

Chris

Aha:

linux-vdso.so.1 (0x00007ffdd19fe000)
libcufft.so.6.5 => not found
libcudart.so.6.5 => not found
libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007feb04e09000)
libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007feb04a18000)
/lib/ld-linux-x86-64.so.2 => /lib64/ld-linux-x86-64.so.2 (0x00007feb051a7000)

Thank you! I remember those two filenames: https://sourceforge.net/projects/cudalucas/files/

"Note that you need library files to run CUDALucas; these are "cudart.dll"
and "cufft.dll" for Windows, and "cudart.so" and "cufft.so" for Linux. In
Windows, it's sufficient to put the .dll files into the same directory as
the executable; in Linux, you have to set the LD_LIBRARY_PATH environment
varibale to include the directory where the .so's are located."

I have no idea how to follow those instructions though. Running mprime on Linux is as far as my Linux skills go. I have already downloaded and installed the CUDA toolkit using the proper NVIDIA instructions. Now I'm stuck, or too lazy to skim esoteric forums, to figure out how to configure the LD_LIBRARY_PATH environment.