Donate To Seti@HomeSeti@Home optimized science apps and information
 
Welcome, Guest. Please login or register.
25 Nov 2014, 10:42:55 pm

Login with username, password and session length
 
» Home
» Forums
» Downloads
» FAQ
» News

» Search site
 
 
 
If you've registered already but never got your activation email, please click here.
 
 
 1   Optimized Seti@Home apps / Alternate Hardware platforms / Re: OSX Multibeam OpenCL question  on: Today at 04:31:29 PM 
Started by Joe Fox | Last post by Joe Fox
And to clarify, the duplication was due to a mistake that I made. I have cleaned up my project and removed the duplicate copy of OpenCL_FFT, and now I am getting a successful run using the ati version on the mac pro. I will also check the nvidia version later today.

Thanks again for the assistance.

 2   Optimized Seti@Home apps / Alternate Hardware platforms / Re: OSX Multibeam OpenCL question  on: Today at 04:01:53 PM 
Started by Joe Fox | Last post by Joe Fox
Ok, I see now. There are two copies of fft_execute. The shared one in src/OpenCL_FFT and another one in AKv8/mac_build/OpenCL_FFT/ which is out of date. I will update my project to point to the shared one. Sorry for the trouble and thanks for the help.

 3   Optimized Seti@Home apps / Alternate Hardware platforms / Re: OSX Multibeam OpenCL question  on: Today at 03:27:29 PM 
Started by Joe Fox | Last post by Raistmer
The advise is quite simple

that's how really current code looks:

Quote
          err |= clSetKernelArg(kernelInfo->kernel, 3, sizeof(cl_int), &s);
#if USE_OPENCL_INTEL //R: only iGPU uses Taylor trig approx and LuT. Old Ati drivers can't accept NULL as valid buf
         err |= clSetKernelArg(kernelInfo->kernel, 4, sizeof(cl_mem), &(plan->cossin_LUT_d1));
            err |= clSetKernelArg(kernelInfo->kernel, 5, sizeof(cl_mem), &(plan->cossin_LUT_d2));
#endif
          err |= clEnqueueNDRangeKernel(queue,  kernelInfo->kernel, 1, NULL, &gWorkItems, &lWorkItems, 0, NULL,NULL/*&fft_event*/);

So, do update to the latest revision and don't forget this time that SETI_opt source tree includes not only AKv8 directory but also src one that common between both MB and AP. Seems you use outdated src now.

 4   Optimized Seti@Home apps / Alternate Hardware platforms / Re: OSX Multibeam OpenCL question  on: Today at 02:53:58 PM 
Started by Joe Fox | Last post by Joe Fox
Finally got some time to trace down to the KernelArg that is throwing the error.

Here is the relevant code:
fft_execute.cpp line 583-603

Code:
    getKernelWorkDimensions(plan, kernelInfo, &s, &gWorkItems, &lWorkItems);
        fprintf(stderr,"After getKernelWorkDimensions:\nbatchSize s =%d, gWorkItems =%lu, lWorkItems =%lu, dir =%d\n", s, gWorkItems, lWorkItems, dir);
    err |= clSetKernelArg(kernelInfo->kernel, 0, sizeof(cl_mem), &memObj[currRead]);
            fprintf(stderr, "after param 0: err=%d\n", err);
    err |= clSetKernelArg(kernelInfo->kernel, 1, sizeof(cl_mem), &memObj[currWrite]);
            fprintf(stderr, "after param 1: err=%d\n", err);
    err |= clSetKernelArg(kernelInfo->kernel, 2, sizeof(cl_int), &dir);
            fprintf(stderr, "after param 2: err=%d\n", err);
    err |= clSetKernelArg(kernelInfo->kernel, 3, sizeof(cl_int), &s);
            fprintf(stderr, "after param 3: err=%d\n", err);
            err |= clSetKernelArg(kernelInfo->kernel, 4, sizeof(cl_mem), &(plan->cossin_LUT_d1));
            fprintf(stderr, "after param 4: err=%d\n", err);
            err |= clSetKernelArg(kernelInfo->kernel, 5, sizeof(cl_mem), &(plan->cossin_LUT_d2));
            fprintf(stderr, "after param 5: err=%d\n", err);

    err |= clEnqueueNDRangeKernel(queue,  kernelInfo->kernel, 1, NULL, &gWorkItems, &lWorkItems, 0, NULL,NULL/*&fft_event*/);
            fprintf(stderr, "after enqueue: err=%d\n", err);
if(err){
fprintf(stderr,"Error in mb oclFFT_2: %d\n",err);
return err;
}

and the resulting output:
Code:
INFO: oclFFT done no strip. plan_ptr: 0x7fff500f0060, FtfNum=0, fftlen=8, NumBlockFfts=131072, chirplen=1048576
After getKernelWorkDimensions:
batchSize s =131072, gWorkItems =131072, lWorkItems =64, dir =1
after param 0: err=0
after param 1: err=0
after param 2: err=0
after param 3: err=0
after param 4: err=-49
after param 5: err=-49
after enqueue: err=-49
Error in mb oclFFT_2: -49
ERROR: OpenCL kernel/call 'non-strip fft' call failed (-49) in file /Users/joe/projects/seti/sah_v7_opt/AKv8/client/analyzeFuncs.cpp near line 3823.

So it looks like line 593
Code:
            err |= clSetKernelArg(kernelInfo->kernel, 4, sizeof(cl_mem), &(plan->cossin_LUT_d1));
is the culprit.

This is all on svn revision 2760 which appears to be the latest.

And advice on next steps?

 5   Optimized Seti@Home apps / GPU crunching / Re: Latest drivers (NVidia and ATI)  on: Today at 11:57:10 AM 
Started by Sutaru Tsureku | Last post by benool
Try Cat 13.12.
I had them installed and situation was the same.
This when I deceide to go forward instead of backwards and give 14.x driver series chance.

Try deleting the compilations for the opencl_ati_cat132 app, and let the app generate new ones.
You mean all the .bin files have a name like MB_clFFTplan_Turks_512_r1831.bin ?
That's something I haven't tried.

Quote from: Claggy
Have you tried restarting the host? It was working with that app version two weeks ago.
Yes host is set to auto-start and shutdown everyday giving it about 10 hours of computation time per day.

Quote from: Claggy
Have you freed any CPU cores?
Yes, host is a dual core with BOINC set to use only 50% of avail CPU cores.


 6   Optimized Seti@Home apps / GPU crunching / Re: Latest drivers (NVidia and ATI)  on: Yesterday at 03:13:58 PM 
Started by Sutaru Tsureku | Last post by Claggy
Anyone still crunching on an ATI 6670 card ? What drivers are you running?

I'm having issues with recent Catalyst and Seti@home beta that insists on sending me tasks labeled with _cat132 (like opencl_ati5_cat132).
These tasks usually crash video card driver in a matter of minutes and causes the BOINC job to just sit there and wait forever.
Some of them do complete successfully but it's far too unpredictable.

Lunactic's very own optimized apps run just fine on this card. For now I have disable AMD task in my SETI@home beta preferences.

Host ID: http://setiweb.ssl.berkeley.edu/beta/results.php?hostid=71860
(and yes this is host mixing intel iGPU, Nvidia and AMD Wink )


Try deleting the compilations for the opencl_ati_cat132 app, and let the app generate new ones.

Have you tried restarting the host? It was working with that app version two weeks ago.

Have you freed any CPU cores?

Claggy

 7   Optimized Seti@Home apps / GPU crunching / Re: Latest drivers (NVidia and ATI)  on: Yesterday at 08:48:08 AM 
Started by Sutaru Tsureku | Last post by Mike
Anyone still crunching on an ATI 6670 card ? What drivers are you running?

I'm having issues with recent Catalyst and Seti@home beta that insists on sending me tasks labeled with _cat132 (like opencl_ati5_cat132).
These tasks usually crash video card driver in a matter of minutes and causes the BOINC job to just sit there and wait forever.
Some of them do complete successfully but it's far too unpredictable.

Lunactic's very own optimized apps run just fine on this card. For now I have disable AMD task in my SETI@home beta preferences.

Host ID: http://setiweb.ssl.berkeley.edu/beta/results.php?hostid=71860
(and yes this is host mixing intel iGPU, Nvidia and AMD Wink )



Try Cat 13.12.

 8   Optimized Seti@Home apps / GPU crunching / Re: Latest drivers (NVidia and ATI)  on: Yesterday at 05:42:18 AM 
Started by Sutaru Tsureku | Last post by benool
Anyone still crunching on an ATI 6670 card ? What drivers are you running?

I'm having issues with recent Catalyst and Seti@home beta that insists on sending me tasks labeled with _cat132 (like opencl_ati5_cat132).
These tasks usually crash video card driver in a matter of minutes and causes the BOINC job to just sit there and wait forever.
Some of them do complete successfully but it's far too unpredictable.

Lunactic's very own optimized apps run just fine on this card. For now I have disable AMD task in my SETI@home beta preferences.

Host ID: http://setiweb.ssl.berkeley.edu/beta/results.php?hostid=71860
(and yes this is host mixing intel iGPU, Nvidia and AMD Wink )


 9   Optimized Seti@Home apps / Alternate Hardware platforms / Re: OSX Multibeam OpenCL question  on: 20 Nov 2014, 04:33:09 pm 
Started by Joe Fox | Last post by Joe Fox
ah, ok. Sorry for the misunderstanding.

 10   Optimized Seti@Home apps / Alternate Hardware platforms / Re: OSX Multibeam OpenCL question  on: 20 Nov 2014, 12:57:44 pm 
Started by Joe Fox | Last post by Raistmer
Code:
      //fprintf(stderr,"After getKernelWorkDimensions:\nbatchSize s =%d, gWorkItems =%d, lWorkItems =%d, dir =%d\n", s, gWorkItems, lWorkItems, dir);
err |= clSetKernelArg(kernelInfo->kernel, 0, sizeof(cl_mem), &memObj[currRead]);
err |= clSetKernelArg(kernelInfo->kernel, 1, sizeof(cl_mem), &memObj[currWrite]);
err |= clSetKernelArg(kernelInfo->kernel, 2, sizeof(cl_int), &dir);
err |= clSetKernelArg(kernelInfo->kernel, 3, sizeof(cl_int), &s);
#if USE_OPENCL_INTEL //R: only iGPU uses Taylor trig approx and LuT. Old Ati drivers can't accept NULL as valid buf
err |= clSetKernelArg(kernelInfo->kernel, 4, sizeof(cl_mem), &(plan->cossin_LUT_d1));
            err |= clSetKernelArg(kernelInfo->kernel, 5, sizeof(cl_mem), &(plan->cossin_LUT_d2));
#endif
err |= clEnqueueNDRangeKernel(queue,  kernelInfo->kernel, 1, NULL, &gWorkItems, &lWorkItems, 0, NULL,NULL/*&fft_event*/);
if(err){
fprintf(stderr,"Error in mb oclFFT_1: %d\n",err);
return err;
}
#if OCL_VERBOSE
fprintf(stderr,"INFO: in mb oclFFT_1 ok\n");
#endif

and

Code:
        //fprintf(stderr,"After getKernelWorkDimensions:\nbatchSize s =%d, gWorkItems =%d, lWorkItems =%d, dir =%d\n", s, gWorkItems, lWorkItems, dir);
    err |= clSetKernelArg(kernelInfo->kernel, 0, sizeof(cl_mem), &memObj[currRead]);
    err |= clSetKernelArg(kernelInfo->kernel, 1, sizeof(cl_mem), &memObj[currWrite]);
    err |= clSetKernelArg(kernelInfo->kernel, 2, sizeof(cl_int), &dir);
    err |= clSetKernelArg(kernelInfo->kernel, 3, sizeof(cl_int), &s);
#if USE_OPENCL_INTEL //R: only iGPU uses Taylor trig approx and LuT. Old Ati drivers can't accept NULL as valid buf
err |= clSetKernelArg(kernelInfo->kernel, 4, sizeof(cl_mem), &(plan->cossin_LUT_d1));
            err |= clSetKernelArg(kernelInfo->kernel, 5, sizeof(cl_mem), &(plan->cossin_LUT_d2));
#endif
    err |= clEnqueueNDRangeKernel(queue,  kernelInfo->kernel, 1, NULL, &gWorkItems, &lWorkItems, 0, NULL,NULL/*&fft_event*/);
if(err){
fprintf(stderr,"Error in mb oclFFT_2: %d\n",err);
return err;
}
#if OCL_VERBOSE
fprintf(stderr,"INFO: in mb oclFFT_2 ok\n");
#endif

Return to the board index.

Quote!
No act of kindness, no matter how small, is ever wasted.
- Aesop, The Lion and the Mouse

 
Site Statistics
Total Members:96
Total Posts:56,013
Total Topics:1,581
Downloads
..Some PHP stuff ToDo
Pages served
Today:4,468
Total:20,160,577
(since 6/26/2006)
Latest Member:
Just Will Lite
 
 
Seti@Home optimized science apps and information | Powered by Enigma 2.0 (RC1).
© 2003-2014, LSP Dev Team. All Rights Reserved.
Seti@Home optimized science apps and information Forums | Powered by SMF.
© 2005, Simple Machines LLC. All Rights Reserved.
Powered by MySQL Powered by PHP Valid XHTML 1.0! Valid CSS!