CUDA_ERROR_LAUNCH_OUT_OF_RESOURCES when running msieve 1.5.0 with CUDA

ryanp · 2012-06-12, 00:58

Hi,

I'm just getting started with trying to run msieve 1.5.0, which I built as:

make x86_64 ECM=1 CUDA=1

This is on Ubuntu 10.04, x86-64, with a Quadro FX 580.

Here's some sample output; I'm consistently getting this error:

Code:

$ ./msieve -v -n -g 0 -t 4 121451298068529844233553416568714964256622520646270917226600348303197298080156462274524325121

Msieve v. 1.50 (SVN exported)
Mon Jun 11 17:53:44 2012
random seeds: fbe091d5 8759ed26
factoring 121451298068529844233553416568714964256622520646270917226600348303197298080156462274524325121 (93 digits)
searching for 15-digit factors
commencing number field sieve (93-digit input)
commencing number field sieve polynomial selection
time limit set to 0.15 CPU-hours
expecting poly E from 3.59e-08 to 4.12e-08
searching leading coefficients from 1 to 15724575
using GPU 0 (Quadro FX 580)
deadline: 5 CPU-seconds per coefficient
coeff 12 specialq 1 - 13109 other 4623 - 11095
aprogs: 281 entries, 726 roots
batch size 12
error (line 433): CUDA_ERROR_LAUNCH_OUT_OF_RESOURCES

Anyone know what's going on? Do I have a bad CUDA library, or a graphics card that's too old to take advantage of CUDA?

In case it's relevant, msieve's dynamic library dependencies are:

Code:

$ ldd ./msieve
	linux-vdso.so.1 =>  (0x00007fffacdac000)
	libecm.so.0 => /usr/lib/libecm.so.0 (0x00007ff716ea2000)
	libcuda.so.1 => /usr/lib/nvidia-current/libcuda.so.1 (0x00007ff7164f9000)
	libz.so.1 => /lib/libz.so.1 (0x00007ff7162e1000)
	libgmp.so.3 => /usr/lib/libgmp.so.3 (0x00007ff716081000)
	libm.so.6 => /lib/libm.so.6 (0x00007ff715dfe000)
	libpthread.so.0 => /lib/libpthread.so.0 (0x00007ff715be0000)
	libc.so.6 => /lib/libc.so.6 (0x00007ff71585d000)
	libdl.so.2 => /lib/libdl.so.2 (0x00007ff715659000)
	/lib64/ld-linux-x86-64.so.2 (0x00007ff717113000)

Any help is appreciated -- thanks!

jrk · 2012-06-12, 02:58

See here for a likely cause and resolution: http://www.mersenneforum.org/showthr...334#post296334

Please copy the output of:

Code:

grep "Pointer Size" stage1_core.ptx

ryanp · 2012-06-12, 03:18

Quote:

Originally Posted by jrk

See here for a likely cause and resolution: http://www.mersenneforum.org/showthr...334#post296334

Please copy the output of:

Code:

grep "Pointer Size" stage1_core.ptx

Thanks for the help, it was:

Code:

//  Target:ptx, ISA:sm_10, Endian:little, Pointer Size:32

I've added "-m 64" to the NVCCFLAGS, and now it looks better:

Code:

//  Target:ptx, ISA:sm_10, Endian:little, Pointer Size:64

Unfortunately, after a recompile, mprime now just segfaults. :(

Code:

$ ./msieve -v -n -g 0 121451298068529844233553416568714964256622520646270917226600348303197298080156462274524325121


Msieve v. 1.50 (SVN exported)
Mon Jun 11 20:14:29 2012
random seeds: 124cc441 3d470019
factoring 121451298068529844233553416568714964256622520646270917226600348303197298080156462274524325121 (93 digits)
searching for 15-digit factors
commencing number field sieve (93-digit input)
R0: -36038659760682026114607
R1: 153341814101
A0: -1303508358934043619358360475
A1: -911356210288725902720
A2: -365874123099670
A3: -189443842
A4: 72
skew 3014459.96, size 1.058e-12, alpha -4.644, combined = 3.357e-08 rroots = 2
factor base loaded:
153713 rational ideals (max prime = 2069227)
153330 algebraic ideals (max prime = 2069213)
a range: [-6153846, 6153846]
b range: [1, 4294967295]
number of hash buckets: 32
sieve block size: 65536

Segmentation fault

jrk · 2012-06-12, 03:27

You are trying to use msieve's built-in line sieve for NFS. In msieve 1.50 there was a bug where the sieve was not initialized properly and it would crash. This has been fixed in SVN (revision 716). But, you should not really be using the line sieve anyway and instead are encouraged to use the lattice siever from the GGNFS project.

2012-06-12, 02:58	#2
jrk May 2008 447₁₆ Posts	See here for a likely cause and resolution: http://www.mersenneforum.org/showthr...334#post296334 Please copy the output of: Code: grep "Pointer Size" stage1_core.ptx Last fiddled with by jrk on 2012-06-12 at 02:59

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
Problem in running msieve with CUDA	mohamed	Msieve	20	2013-08-01 08:27
Help with msieve CUDA (C1060)	toorandom	Msieve	7	2012-11-28 17:37
Do I have CUDA running?	jasong	GPU Computing	5	2012-11-28 00:36

2012-06-12, 03:27	#4
jrk May 2008 3×5×73 Posts	You are trying to use msieve's built-in line sieve for NFS. In msieve 1.50 there was a bug where the sieve was not initialized properly and it would crash. This has been fixed in SVN (revision 716). But, you should not really be using the line sieve anyway and instead are encouraged to use the lattice siever from the GGNFS project.