Petri, how about posting your Current CUDA app_info.xml section so we can see what You are using? Similar to what I posted. I believe that's what Keith is looking for.

    <file_info>
      <name></name>
    </file_info>
    <file_info>
      <name></name>
    </file_info>

Hi Petri, then color me very CONFUSED. I downloaded the file from your link. It is named setiV0.97.linux_x64_10x0. So that is what I changed the appname to setiV0.97.linux_x64_10x0 in the file info and app version defines.

I look at your app_info example and see no sign of the 0.97 app name setiV0.97.linux_x64_10x0. I see instead setiathome_x41zc_x86_64-pc-linux-gnu_cuda65_v8.

So is the real name of the 0.97 app setiathome_x41zc_x86_64-pc-linux-gnu_cuda65_v8?

No wonder I dumped all the work if the name definition is not the name of the real executable name.

Also should I remove the libcudart.so.9.0 and libcufft.so.9.0 file defines for the zi3v app? Or just leave them in there and they will cause no harm. I expect they have to be removed. Also I see the -nobs still in there. I though someone said that is no longer effective for the static linked apps?

More clarification please.

<app_info>
  <app>
    <name>astropulse_v7</name>
  </app>
  
  <file_info>
    <name>ap_7.01r2793_sse3_clGPU_x86_64</name>
    <executable/>
  </file_info>
  
 
  <app_version>
    <app_name>astropulse_v7</app_name>
    <version_num>708</version_num>
    <platform>linux_x86_64</platform>
    <plan_class>opencl_nvidia_100</plan_class> 
    <cmdline> -verb -st -nog -unroll 80 -ffa_block 2304 -ffa_block_fetch 1152 -oclFFT_plan 256 16 256 </cmdline> 
    <coproc> 
      <type>NVIDIA</type> 
      <count>0.33</count> 
    </coproc>
    <file_ref>
      <file_name>ap_7.01r2793_sse3_clGPU_x86_64</file_name>
      <main_program/>
    </file_ref>
  </app_version>

  
  <app>
    <name>setiathome_v8</name>
  </app>
  
  <file_info>
    <name>MBv8_8.22r3712_avx2_x86_64-pc-linux-gnu</name>
    <namex>MBv8_8.05r3345_avx_linux64</namex>
    <executable/>
  </file_info>
  
  <file_info>
    <name>setiathome_x41zc_x86_64-pc-linux-gnu_cuda65_v8</name>
    <executable/>
  </file_info>
  
  <app_version>
    <app_name>setiathome_v8</app_name>
    <version_num>800</version_num>
    <platform>x86_64-pc-linux-gnu</platform>
    <avg_ncpus>1.000000</avg_ncpus>
    <max_ncpus>1.000000</max_ncpus>
    <file_ref>
      <file_name>MBv8_8.22r3712_avx2_x86_64-pc-linux-gnu</file_name>
      <main_program/>
    </file_ref>
  </app_version>

  <app_version>
    <app_name>setiathome_v8</app_name>
    <version_num>801</version_num>
    <platform>x86_64-pc-linux-gnu</platform>
    <avg_ncpus>1.000000</avg_ncpus>
    <max_ncpus>1.000000</max_ncpus>
    <file_ref>
      <file_name>MBv8_8.22r3712_avx2_x86_64-pc-linux-gnu</file_name>
      <main_program/>
    </file_ref>
  </app_version>
  
  <app_version>
    <app_name>setiathome_v8</app_name>
    <version_num>804</version_num>
    <platform>x86_64-pc-linux-gnu</platform>
    <avg_ncpus>1.000000</avg_ncpus>
    <max_ncpus>1.000000</max_ncpus>
    <file_ref>
      <file_name>MBv8_8.22r3712_avx2_x86_64-pc-linux-gnu</file_name>
      <main_program/>
    </file_ref>
  </app_version>   

  <app_version>
    <app_name>setiathome_v8</app_name>
    <version_num>808</version_num>
    <platform>x86_64-pc-linux-gnu</platform>
    <avg_ncpus>0.1</avg_ncpus> 
    <max_ncpus>0.1</max_ncpus> 
    <plan_class>nvidia_gpu</plan_class> 
    <cmdline> -nobs -pfb 32 </cmdline> 
    <coproc> 
      <type>NVIDIA</type> 
      <count>1.00</count> 
    </coproc>
    <file_ref>
      <file_name>setiathome_x41zc_x86_64-pc-linux-gnu_cuda65_v8</file_name>
      <main_program/>
    </file_ref>
  </app_version>

  <app_version>
    <app_name>setiathome_v8</app_name>
    <version_num>809</version_num>
    <platform>x86_64-pc-linux-gnu</platform>
    <avg_ncpus>0.1</avg_ncpus> 
    <max_ncpus>0.1</max_ncpus> 
    <plan_class>opencl_nvidia_sah</plan_class> 
    <cmdline> -nobs -pfb 32  </cmdline> 
    <coproc> 
      <type>NVIDIA</type> 
      <count>1.00</count> 
    </coproc>
    <file_ref>
      <file_name>setiathome_x41zc_x86_64-pc-linux-gnu_cuda65_v8</file_name>
      <main_program/>
    </file_ref>
  </app_version>
  
</app_info>

<app_config>
  <xproject_max_concurrent>10</xproject_max_concurrent>
  
  <app>
    <name>astropulse_v7</name>
    <max_concurrent>8</max_concurrent>
    <gpu_versions>
      <gpu_usage>0.50</gpu_usage>
      <cpu_usage>0.125</cpu_usage>
    </gpu_versions>
  </app>

  <app>
    <name>setiathome_v8</name>
    <xmax_concurrent>10</xmax_concurrent>
    <gpu_versions>
      <gpu_usage>1.0</gpu_usage>
      <cpu_usage>0.1</cpu_usage>
    </gpu_versions>
  </app>
  
  <app_version>
    <app_name>astropulse_v7</app_name>
    <plan_class>opencl_nvidia_100</plan_class>
    <cmdline> -verb -st -nog -unroll 80 -sbs 2048 -ffa_block 2304 -ffa_block_fetch 1152 -oclFFT_plan 256 16 256 </cmdline>
  </app_version>

  <app_version>
    <app_name>setiathome_v8</app_name>
    <plan_class>opencl_nvidia_sah</plan_class>
    <cmdline> -pfb 32 -nobs -pfl 64 </cmdline>
  </app_version>

  <app_version>
    <app_name>setiathome_v8</app_name>
    <plan_class>nvidia_gpu</plan_class>
    <cmdline> -pfb 32 -nobs -pfl 64 </cmdline>
  </app_version>
  
 </app_config>

Happy B-Day, Petri! :-)

Mine was June 30th, I'm 51 now... BUT, I'm in the "Forever 16 Club!" ;-)

1984 was a VERY good year. :-)


TL

Happy B-Day Petri, you make me feel like an olde man at 73!

Well you are, I'll be 73 next month.

Not all that old, I am still crunching at 77.
Happy B-day Petri

I still measure time in years. Today is my 50th birthday.
Petri

Happy Birthday! All the best!!!

Is it possible that you are having a heat issue? That is something that will cause errors like this.
Hoping you are not OverClocking???

I see I could have used Petri's new utility just now. I just trashed all 300 gpu tasks on one system because the Compute Cache got corrupted. I was too aggressive on a memory overclock I guess and it trashed the work and froze the system probably on the transition for P2 to P3. Backed off my memory overclock to what was stable previously. Lesson learned.

Well! Happy Birthday Petri! Just wanted to say how much we crunchers appreciate our volunteer app developers. That goes for Raistmer and TBar too and all the supporting volunteers who test and suggest improvements.

If anything is found that part is done sequentially on one SM to ensure correctness.

Petri

Hi, Petri

Same I use in OpenCL. If this allow to pass test cases collected so far I would say bugfix applied. Of course sequential scan slows things down but it's not a bug as long as valid results are provided.

Regarding signal collecting on GPU - pulses are hard die at least for me :) If I recall right, pulses and triplets are the ones that not "on GPU" in Signals on GPU (SoG) build.
So I wish you have more time too! Would be glad to port CUDA solution to OpenCL if any ;)
Perhaps the key is reordering on GPU too as you say. In SoG I just gather signals and final reordering is done on CPU. And that's not quite appropriate for Pulses. Maybe smth else too but at time I actively worked on SoG Pulses resisted %).

If anything is found that part is done sequentially on one SM to ensure correctness.

Petri

Hi, Petri

Same I use in OpenCL. If this allow to pass test cases collected so far I would say bugfix applied. Of course sequential scan slows things down but it's not a bug as long as valid results are provided.

Regarding signal collecting on GPU - pulses are hard die at least for me :) If I recall right, pulses and triplets are the ones that not "on GPU" in Signals on GPU (SoG) build.
So I wish you have more time too! Would be glad to port CUDA solution to OpenCL if any ;)
Perhaps the key is reordering on GPU too as you say. In SoG I just gather signals and final reordering is done on CPU. And that's not quite appropriate for Pulses. Maybe smth else too but at time I actively worked on SoG Pulses resisted %).

But based on comments alluded by Stephen and what I remember, our Windows app developer Raistmer said it would be almost impossible to use the exiting CUDA9 Linux app code source for any Windows app simply because the mechanisms the Linux app uses isn't available in the Windows environment.

Ups, hardly I said so :)
What I could say it's no Windows build until sync bug would be resolved (is it?).
CUDA is quite OS-independent AFAIK though I dealt only with older versions.

Just a quick, little bit out in the weeds question, what is the issue running P0 in Linux on the latest 10x0 series cards? Is it a driver issue, an OS issue, or a software utility issue which won't allow it to be set? And, what is needed for a solution, Nvidia, Linus, or the utility developer to get on it? And the likelihood of it happening any time soon? Don't know much about it, other than I also thought I heard that P0 is always the best way to go, obviously if you can set it, that is.