Forum > Linux
SETI MB CUDA for Linux
Metod, S56RKO:
Here's link: http://setiathome.berkeley.edu/show_host_detail.php?hostid=129668.
I'm using Crunch3r's setiathome-CUDA-6.08.i686.
For living this machine does some GIS stuff with interface to MS SQL server. We've found out that Unix ODBC is broken on 64-bit while 32-bit one is OK. Hence 32-bit OS on this machine. My heart bleeds, but one can not have everything. :(
Hummm ... on the other hand, it seems to proceed somehow. state.sah has this line in it:
--- Code: ---<prog>0.20329889</prog>
--- End code ---
My concern (not primary though) is about why BOINC doesn't run GPU task in addition to 4 CPU tasks (quad core processor).
Metod, S56RKO:
Seems that those lines are somewhat benign for the application itself. GPU finished one WU: http://setiathome.berkeley.edu/result.php?resultid=1690649422. Run time is nothing special (half of normal CPU time). What remains is to see result verified. If it verifies, then those lines in stderr make major aesthetic problem.
If the above is fine, then the issue of BOINC not running GPU task concurrent with 4 CPU tasks becomes primary concern.
sunu:
I guess you tried a 32bit chroot for your work and didn't work?
For the seti part:
-256.44 is the latest nvidia driver. Get it.
- If you don't run a x server make sure that you initialize your card properly. This is from the cuda release notes:
o In order to run CUDA applications, the CUDA module must be
loaded and the entries in /dev created. This may be achieved
by initializing X Windows, or by creating a script to load the
kernel module and create the entries.
An example script (to be run at boot time):
#!/bin/bash
/sbin/modprobe nvidia
if [ "$?" -eq 0 ]; then
# Count the number of NVIDIA controllers found.
N3D=`/sbin/lspci | grep -i NVIDIA | grep "3D controller" | wc -l`
NVGA=`/sbin/lspci | grep -i NVIDIA | grep "VGA compatible controller" | wc -l`
N=`expr $N3D + $NVGA - 1`
for i in `seq 0 $N`; do
mknod -m 666 /dev/nvidia$i c 195 $i;
done
mknod -m 666 /dev/nvidiactl c 195 255
else
exit 1
fi
- Latest boinc is 6.10.56. Get it.
Let's try to reinstall the whole thing:
- Close boinc and delete everything in projects/setiathome.berkeley.edu directory. Delete client_state.xml and client_state_prev.xml in your boinc directory. Delete everything in the slots directory.
- Get http://calbe.dw70.de/AKV8/REL/V4/AK_V8_SSSE3_LX32.tar.bz2 and put the executable in projects/setiathome.berkeley.edu directory
- Get http://calbe.dw70.de/cuda/rel/setiathome-CUDA_2.2_6.08.x86_vlarkill.tar.bz2 and put the executable in projects/setiathome.berkeley.edu directory.
- Get http://lunatics.kwsn.net/index.php?module=Downloads;sa=dlview;id=258 and put the executable in projects/setiathome.berkeley.edu
- Get the 2.3 toolkit (delete everything from the 3.1 you have) and put
libcudart.so
libcudart.so.2
libcudart.so.2.3
libcufft.so
libcufft.so.2
libcufft.so.2.3
in the projects/setiathome.berkeley.edu directory.
- Edit accordingly your ld.so.conf or the corresponding ld-something file of your distro with the above location of the cuda libs. Do a ldconfig to update the cache.
- Do a ldd... to the cuda app to see if all dependencies are met.
-Make a file named app_info.xml in the projects/setiathome.berkeley.edu directory and put the following:
<app_info>
<app>
<name>astropulse_v505</name>
</app>
<file_info>
<name>ap_5.06r411_sse3_x86_64-beta4</name>
<executable/>
</file_info>
<app_version>
<app_name>astropulse_v505</app_name>
<version_num>506</version_num>
<flops>41430637443.0638666</flops>
<avg_ncpus>1.0000</avg_ncpus>
<max_ncpus>1.0000</max_ncpus>
<file_ref>
<file_name>ap_5.06r411_sse3_x86_64-beta4</file_name>
<main_program/>
</file_ref>
</app_version>
<app>
<name>setiathome_enhanced</name>
</app>
<file_info>
<name>AK_V8_linux64_ssse3</name>
<executable/>
</file_info>
<file_info>
<name>setiathome-6.08.CUDA_2.2_x86_64-pc-linux-gnu</name>
<executable/>
</file_info>
<file_info>
<name>libcudart.so.2</name>
<executable/>
</file_info>
<file_info>
<name>libcufft.so.2</name>
<executable/>
</file_info>
<app_version>
<app_name>setiathome_enhanced</app_name>
<version_num>603</version_num>
<flops>24463044489.9277371</flops>
<avg_ncpus>1.0000</avg_ncpus>
<max_ncpus>1.0000</max_ncpus>
<file_ref>
<file_name>AK_V8_linux64_ssse3</file_name>
<main_program/>
</file_ref>
</app_version>
<app_version>
<app_name>setiathome_enhanced</app_name>
<version_num>608</version_num>
<plan_class>cuda</plan_class>
<flops>244630444899.2773714</flops>
<avg_ncpus>0.1400</avg_ncpus>
<max_ncpus>0.1400</max_ncpus>
<coproc>
<type>CUDA</type>
<count>1</count>
</coproc>
<file_ref>
<file_name>setiathome-6.08.CUDA_2.2_x86_64-pc-linux-gnu</file_name>
<main_program/>
</file_ref>
<file_ref>
<file_name>libcudart.so.2</file_name>
</file_ref>
<file_ref>
<file_name>libcufft.so.2</file_name>
</file_ref>
</app_version>
</app_info>
-Change the yellow filenames with the ones you've got.
- Change the red number with something much lower, I don't know what a proper number for your card will be, maybe half of it.
- You have to make the cuda app run at 0 priority. I use the following script that runs in infinite loop:
#!/bin/sh
while true
do
for arg in `ps -Lo lwp --no-headers -p $(pgrep setiathome-6)`
do
renice 0 -p ${arg}
done
sleep 5
done
for that script to work you'll need the procps and bsdutils packages installed.
- Fire up boinc and see what we've done :D
Metod, S56RKO:
--- Quote from: sunu on 23 Aug 2010, 12:38:18 pm ---- Change the red number with something much lower, I don't know what a proper number for your card will be, maybe half of it.
--- End quote ---
I assume that this flops stuff is more or less aesthetics.
Other than few differences in appinfo.xml (missing flops stanza, different values for avg_ncpus and max_ncpus - was 0.04) I've managed quite a different setup. I've left toolkit 3.1 while installing 2.0 libraries systemwide as well. ldd shows that correct (2.0) libraries are then loaded by seti/cuda app.
I've updated toolkit to 2.3 now and Crunch3r's app to the one you proposed. I'll check and see how it works out.
--- Quote from: sunu on 23 Aug 2010, 12:38:18 pm ---- You have to make the cuda app run at 0 priority. I use the following script that runs in infinite loop:
--- End quote ---
Could this be reason for quite long run times on my card? Meanwhile two results got verified and peers were running stock GPU app for Windows. One peer had GTX295, the other GTX285, both had from 5 to 8 times shorter run time. Or is such a speed difference expected between more modern and almost ancient GPUs?
sunu:
--- Quote from: Metod, S56RKO on 25 Aug 2010, 10:20:08 am ---I assume that this flops stuff is more or less aesthetics.
--- End quote ---
Absolutely NOT! Check http://lunatics.kwsn.net/3-linux/seti-mb-cuda-for-linux.msg19829.html#msg19829 and http://lunatics.kwsn.net/5-windows/appinfo-flops-question.0.html
--- Quote from: Metod, S56RKO on 25 Aug 2010, 10:20:08 am ---I've left toolkit 3.1 while installing 2.0 libraries systemwide as well. ldd shows that correct (2.0) libraries are then loaded by seti/cuda app.
I've updated toolkit to 2.3 now and Crunch3r's app to the one you proposed. I'll check and see how it works out.
--- End quote ---
If you don't need them for other things DELETE the 3.1 and 2.0 and KEEP the 2.3.
--- Quote from: Metod, S56RKO on 25 Aug 2010, 10:20:08 am ---
--- Quote from: sunu on 23 Aug 2010, 12:38:18 pm ---- You have to make the cuda app run at 0 priority. I use the following script that runs in infinite loop:
--- End quote ---
Could this be reason for quite long run times on my card?
--- End quote ---
Yes, you ABSOLUTELY have to run the cuda app at priority 0.
--- Quote from: Metod, S56RKO on 25 Aug 2010, 10:20:08 am ---Meanwhile two results got verified and peers were running stock GPU app for Windows. One peer had GTX295, the other GTX285, both had from 5 to 8 times shorter run time. Or is such a speed difference expected between more modern and almost ancient GPUs?
--- End quote ---
Of course a gtx295 or gtx 285 will be multiple times faster than a 8600gt. If they weren't, why people then would fork out multiple hundred euros/dollars to buy them?
Also note that the linux cuda app is slower in general than the windows one.
Navigation
[0] Message Index
[#] Next page
[*] Previous page
Go to full version