Help with msieve CUDA (C1060)

toorandom · 2012-11-28, 03:15

Hello, i just downloaded msieve 1.50 and I have in my faculty a C1060

$ lspci | grep -i nv
01:00.0 3D controller: nVidia Corporation GT200 [Tesla C1060] (rev a1)

I just installed all the SDK and the problem is that when I compile it in this way

make x86_64 CUDA=1 NVCCFLAGS="-arch sm_13"

everything compiles fine and i can see that nvidia headers and libraries are being linked, in fact here are some relevant lines of the compilation

nvcc -m64 -arch sm_13 -ptx -o stage1_core.ptx \\ gnfs/poly/stage1/stage1_core_gpu/stage1_core.cu

gcc -D_FILE_OFFSET_BITS=64 -O3 -fomit-frame-pointer -DNDEBUG -D_LARGEFILE64_SOURCE -Wall -W -DMSIEVE_SVN_VERSION="\"exported\"" -I. -Iinclude -Ignfs -Ignfs/poly -Ignfs/poly/stage1 -I"/usr/local/cuda/include" -DHAVE_CUDA demo.c -o msieve \
libmsieve.a -lcuda -lz -lgmp -lm -lpthread

and the msieve binary ldd looks like

$ ldd msieve | grep -i cuda
libcuda.so.1 => /usr/lib64/libcuda.so.1 (0x00007fa1c2cfc000)

The big problem is that when I run it it doesnt use cuda, i have ran it with -g option and without , but when I run nvidia-smi -l it always says that
| No running compute processes found |

And if I run the nvidia samples, the process appears in the nvidia-smi command, here is how i run msieve and some output that shows that isnt using NVIDIA

$ ./msieve -g 0 -v 69986008711415694391421268580269058232048146719704518153244714221529713444447

Msieve v. 1.50 (SVN exported)
Tue Nov 27 21:17:09 2012
random seeds: 59e9351d e6fc711d
factoring 69986008711415694391421268580269058232048146719704518153244714221529713444447 (77 digits)
no P-1/P+1/ECM available, skipping
commencing quadratic sieve (77-digit input)
using multiplier of 1
using 32kb Intel Core sieve core
sieve interval: 12 blocks of size 32768
processing polynomials in batches of 17

Thanks

LaurV · 2012-11-28, 05:27

Quote:

Originally Posted by toorandom

Hello, i just downloaded msieve 1.50 and I have in my faculty a C1060

Man, say that is a Tesla C1060, you made my heart jump...

frmky · 2012-11-28, 05:43

Only the NFS polynomial selection is CUDA-enabled. QS nor NFS sieving supports CUDA.

frmky · 2012-11-28, 05:45

Quote:

Originally Posted by LaurV

Man, say that is a Tesla C1060, you made my heart jump...

Don't get too excited. It's no faster than a GTX 460. Just has more memory.

LaurV · 2012-11-28, 06:13

Quote:

Originally Posted by frmky

Don't get too excited. It's no faster than a GTX 460. Just has more memory.

No, I thought he wanna factorize a composite C1060.

(edit: owned two C2070 by myself, see my former posts one year ago)

frmky · 2012-11-28, 08:43

Owned? What happened to them?

Batalov · 2012-11-28, 09:11

Quote:

Originally Posted by LaurV

No, I thought he wanna factorize a composite C1060.

Here's a C1060. Completely factored, nothing to it, really.

jasonp · 2012-11-28, 17:37

frmky is correct, you are running the quadratic sieve and not the number field sieve. Use 'msieve -h' to get the complete list of command-line stuff you can use, but CUDA will only be useful for selecting NFS polynomials, and your current input is too small for the NFS code to handle (try an input larger than 85 digits, and specify an NFS task explicitly).

2012-11-28, 03:15	#1
toorandom Nov 2012 1 Posts	Help with msieve CUDA (C1060) Hello, i just downloaded msieve 1.50 and I have in my faculty a C1060 $ lspci \| grep -i nv 01:00.0 3D controller: nVidia Corporation GT200 [Tesla C1060] (rev a1) I just installed all the SDK and the problem is that when I compile it in this way make x86_64 CUDA=1 NVCCFLAGS="-arch sm_13" everything compiles fine and i can see that nvidia headers and libraries are being linked, in fact here are some relevant lines of the compilation nvcc -m64 -arch sm_13 -ptx -o stage1_core.ptx \\ gnfs/poly/stage1/stage1_core_gpu/stage1_core.cu gcc -D_FILE_OFFSET_BITS=64 -O3 -fomit-frame-pointer -DNDEBUG -D_LARGEFILE64_SOURCE -Wall -W -DMSIEVE_SVN_VERSION="\"exported\"" -I. -Iinclude -Ignfs -Ignfs/poly -Ignfs/poly/stage1 -I"/usr/local/cuda/include" -DHAVE_CUDA demo.c -o msieve \ libmsieve.a -lcuda -lz -lgmp -lm -lpthread and the msieve binary ldd looks like $ ldd msieve \| grep -i cuda libcuda.so.1 => /usr/lib64/libcuda.so.1 (0x00007fa1c2cfc000) The big problem is that when I run it it doesnt use cuda, i have ran it with -g option and without , but when I run nvidia-smi -l it always says that \| No running compute processes found \| And if I run the nvidia samples, the process appears in the nvidia-smi command, here is how i run msieve and some output that shows that isnt using NVIDIA $ ./msieve -g 0 -v 69986008711415694391421268580269058232048146719704518153244714221529713444447 Msieve v. 1.50 (SVN exported) Tue Nov 27 21:17:09 2012 random seeds: 59e9351d e6fc711d factoring 69986008711415694391421268580269058232048146719704518153244714221529713444447 (77 digits) no P-1/P+1/ECM available, skipping commencing quadratic sieve (77-digit input) using multiplier of 1 using 32kb Intel Core sieve core sieve interval: 12 blocks of size 32768 processing polynomials in batches of 17 Thanks

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
error when running msieve 1.53 with cuda	aein	Msieve	9	2019-02-25 14:09
Can anyone help me about msieve 1.53 with CUDA?	Seto	Msieve	8	2016-09-24 12:54
My msieve cuda setup as a newbie	pdazzl	Msieve	21	2016-08-20 11:58
Problem in running msieve with CUDA	mohamed	Msieve	20	2013-08-01 08:27
CUDA_ERROR_LAUNCH_OUT_OF_RESOURCES when running msieve 1.5.0 with CUDA	ryanp	Msieve	3	2012-06-12 03:27

2012-11-28, 05:43	#3
frmky Jul 2003 So Cal 19×137 Posts	Only the NFS polynomial selection is CUDA-enabled. QS nor NFS sieving supports CUDA.

2012-11-28, 08:43	#6
frmky Jul 2003 So Cal 101000101011₂ Posts	Owned? What happened to them?

2012-11-28, 17:37	#8
jasonp Tribal Bullet Oct 2004 3²×5×79 Posts	frmky is correct, you are running the quadratic sieve and not the number field sieve. Use 'msieve -h' to get the complete list of command-line stuff you can use, but CUDA will only be useful for selecting NFS polynomials, and your current input is too small for the NFS code to handle (try an input larger than 85 digits, and specify an NFS task explicitly).