Looks like BRP4 SSE tasks using v1.33 validates against v1.34 intel_opencl tasks. Giving that SSE app uses only a little bit more than 200 Mb, so I can use this platform on my old rigs with 512 Mb of memory and return results in more reasonable time (in no more than several hours per WU). But I still can't build an app_info.xml that can be used to mix there CUDA and OpenCL tasks.
After 30 minutes. This host where I'm working on app_info.xml just downloaded Nvidia GPU task but it errored out right after start with the message
This host where I'm working on app_info.xml just downloaded Nvidia GPU task
Yep, we fixed a problem related to newer clients and anonymous CUDA/NVidia app versions on the server side about an hour ago.
BM
Edit: The BRP GPU plan classes don't have any restriction regarding CPU features, so why are you trying anonymous platform there at all? If your graphics card or driver doesn't match the requirements for our standard GPU apps I'm not surprised at all about errors of any kind.
If your graphics card or driver doesn't match the requirements for our standard GPU apps I'm not surprised at all about errors of any kind.
It's my hobby to tune computers up so they start to do what they can't to by the
nature. Joking :)
I have Intel HD 4000 GPU and NVidia GTX 560 on this host. And both of them run quite smoothly in default mode (without app_info).
I'm trying to run BRP4 SSE tasks on this machine for several reasons:
1) This host is Win7 32 with 2.5 Gb of useful memory (others of 8 Gb are eaten by Microsoft with their idea of 64-bit, Intel GPU and NVidia shared memory and lots of browser windows, because I'm "living" on it), so if I want to make 8-core CPU busy, I need to use smaller apps than FGRP4. BRP4SSE is the best solution.
2) I try to build a complex/universal app_info that can be then used with small changes in old Celerons and P4s (I'll only will have to cut out everything else).
3) I want to make run side by side both OpenCL BRP4 (not G) app on Intel GPU and on CPU just to see the difference in productivity.
4) I want to concentrate my efforts on BRP4 help you keep up with the incoming data,
so I want my Nvidia GTX 560 to run fully BRP4 (G tasks is the only solution).
For now I stuck with two problems:
1) I can not understand how can I make BRP4 OpenCL app running - currently it doesn't even try to download work for this app, instead it reports 0 instances of OpenCL Intel GPU;
2) I have the trouble running BRP4G CUDA tasks described here.
I read carefully about building app_info, understand what I'm doing and how to write this XML, but yet can not understand what info should I provide there to make things work as I want them to.
P.S. Maybe it's better to move this thread to the crunchers corner?
Ok. I now made an app_info.xml download BRP4 SSE and BRP4 opencl tasks.
But now I have 3 questions that I can not find answer for:
1) Why does BOINC log says this? File is in the project folder. I downloded and checked it twice. No virus or antivirus activity. No other harmed files. Einstein@Home | [error] State file error: missing application file cudart_xp32_55_22.dll
2) I still can not make BOINC download CUDA tasks and can not find reason why.
BOINC says Einstein@Home | Not requesting tasks: don't need (CPU: job cache full; NVIDIA GPU: ; Intel GPU: job cache full)
3) BOINC constanly deletes einsteinbinary_BRP4G_1.52_windows_intelx86__BRP4G-Beta-cuda32.exe after restart. I have to manually copy this file to BOINC project folder each time before running.
Looks like BRP4 SSE tasks
)
Looks like BRP4 SSE tasks using v1.33 validates against v1.34 intel_opencl tasks. Giving that SSE app uses only a little bit more than 200 Mb, so I can use this platform on my old rigs with 512 Mb of memory and return results in more reasonable time (in no more than several hours per WU). But I still can't build an app_info.xml that can be used to mix there CUDA and OpenCL tasks.
After 30 minutes. This host where I'm working on app_info.xml just downloaded Nvidia GPU task but it errored out right after start with the message
[handle] [RESULT#524548296] [WU#229966934]: client_state 3 exit_status -1073741515; setting outcome ERRORAm I missed something in app_info?
RE: This host where I'm
)
Yep, we fixed a problem related to newer clients and anonymous CUDA/NVidia app versions on the server side about an hour ago.
BM
Edit: The BRP GPU plan classes don't have any restriction regarding CPU features, so why are you trying anonymous platform there at all? If your graphics card or driver doesn't match the requirements for our standard GPU apps I'm not surprised at all about errors of any kind.
BM
RE: If your graphics card
)
It's my hobby to tune computers up so they start to do what they can't to by the
nature. Joking :)
I have Intel HD 4000 GPU and NVidia GTX 560 on this host. And both of them run quite smoothly in default mode (without app_info).
I'm trying to run BRP4 SSE tasks on this machine for several reasons:
1) This host is Win7 32 with 2.5 Gb of useful memory (others of 8 Gb are eaten by Microsoft with their idea of 64-bit, Intel GPU and NVidia shared memory and lots of browser windows, because I'm "living" on it), so if I want to make 8-core CPU busy, I need to use smaller apps than FGRP4. BRP4SSE is the best solution.
2) I try to build a complex/universal app_info that can be then used with small changes in old Celerons and P4s (I'll only will have to cut out everything else).
3) I want to make run side by side both OpenCL BRP4 (not G) app on Intel GPU and on CPU just to see the difference in productivity.
4) I want to concentrate my efforts on BRP4 help you keep up with the incoming data,
so I want my Nvidia GTX 560 to run fully BRP4 (G tasks is the only solution).
For now I stuck with two problems:
1) I can not understand how can I make BRP4 OpenCL app running - currently it doesn't even try to download work for this app, instead it reports 0 instances of OpenCL Intel GPU;
2) I have the trouble running BRP4G CUDA tasks described here.
I read carefully about building app_info, understand what I'm doing and how to write this XML, but yet can not understand what info should I provide there to make things work as I want them to.
P.S. Maybe it's better to move this thread to the crunchers corner?
Ok. I now made an
)
Ok. I now made an app_info.xml download BRP4 SSE and BRP4 opencl tasks.
But now I have 3 questions that I can not find answer for:
1) Why does BOINC log says this? File is in the project folder. I downloded and checked it twice. No virus or antivirus activity. No other harmed files.
Einstein@Home | [error] State file error: missing application file cudart_xp32_55_22.dll2) I still can not make BOINC download CUDA tasks and can not find reason why.
BOINC says
Einstein@Home | Not requesting tasks: don't need (CPU: job cache full; NVIDIA GPU: ; Intel GPU: job cache full)3) BOINC constanly deletes einsteinbinary_BRP4G_1.52_windows_intelx86__BRP4G-Beta-cuda32.exe after restart. I have to manually copy this file to BOINC project folder each time before running.
einsteinbinary_BRP4
einsteinbinary_BRP4G
einsteinbinary_BRP4_1.33_windows_intelx86__BRP4SSE.exe
einsteinbinary_BRP4_1.34_windows_intelx86__opencl-intel_gpu.exe
einsteinbinary_BRP4G_1.52_windows_intelx86__BRP4G-Beta-cuda32.exe
einsteinbinary_BRP4
windows_intelx86
098
einsteinbinary_BRP4_1.33_windows_intelx86__BRP4SSE.exe
einsteinbinary_BRP4
opencl-intel_gpu
windows_intelx86
099
intel_gpu
1
einsteinbinary_BRP4_1.34_windows_intelx86__opencl-intel_gpu.exe
einsteinbinary_BRP4G
windows_intelx86
CUDA
099
CUDA
1
einsteinbinary_BRP4G_1.52_windows_intelx86__BRP4G-Beta-cuda32.exe
cudart_xp32_55_22.dll
cufft_xp32_55_22.dll
cufft32_55.dll
einsteinbinary_BRP4G_1.52_windows_intelx86__BRP4G-Beta-cuda32.exe-db.dev
db.dev
einstein.phys.uwm.edu\einsteinbinary_BRP4G_1.52_windows_intelx86__BRP4G-Beta-cuda32.exe-dbhs.dev
dbhs.dev
einstein.phys.uwm.edu\einsteinbinary_BRP4_1.00_graphics_windows_intelx86.exe
graphics_app
cudart_xp32_55_22.dll cuff
)
cufft_xp32_55_22.dll
cufft32_55.dll
The is missing for the cudart file. I would guess the expected one is
cudart32_55.dllBM
BM