The VRAM usage / requirement is normally calculated from the memory requirement of the WU, which is put there by the workunit generator. Apparently there's something wrong with that. Until I can properly fix it next week, I now set a constant value of 3.2GB.
Note: I don't have time to look into each and every error, especially over the weekend. I'll run a statistic on the results and errors on Monday. In the meantime: it's Beta test. if you don't like what your machine does or does not - feel free to opt out.
You should now be able to
)
You should now be able to select the 'Multi-Directional Gravitational Wave search on O3 (GPU)' application in your project preferences.
BM
Bernd Machenschalk
)
And it worked for me, Thank You
For : O3 (GPU) v1.02
)
For : O3 (GPU) v1.02 (GW-opencl-nvidia) I got :
Computation error
...
(unknown error) - exit code 4294967295 (0xffffffff)
...
XLAL Error - XLALOpenCLExecuteKernel (/home/jenkins/workspace/workspace/EaH-GW-OpenCL-Testing/SLAVE/MinGW6.3/TARGET/windows-x64/EinsteinAtHome/source/lalsuite/lalpulsar/lib/GPUUtils/OpenCLUtils.c:547): Enqueue OpenCL kernel failed with OpenCL error: CL_MEM_OBJECT_ALLOCATION_FAILURE
XLAL Error - XLALOpenCLExecuteKernel (/home/jenkins/workspace/workspace/EaH-GW-OpenCL-Testing/SLAVE/MinGW6.3/TARGET/windows-x64/EinsteinAtHome/source/lalsuite/lalpulsar/lib/GPUUtils/OpenCLUtils.c:547): Generic failure
XLAL Error - XLALSemiCohStep_OpenCL (/home/jenkins/workspace/workspace/EaH-GW-OpenCL-Testing/SLAVE/MinGW6.3/TARGET/windows-x64/EinsteinAtHome/source/lalsuite/lalapps/src/pulsar/GCT/HierarchSearchGCT_OpenCL.c:137): Check failed: XLALOpenCLExecuteKernel ( &(GCTOpenCLKernels.kernel_SemiCohStep), &size, 1 ) == XLAL_SUCCESS
XLAL Error - XLALSemiCohStep_OpenCL (/home/jenkins/workspace/workspace/EaH-GW-OpenCL-Testing/SLAVE/MinGW6.3/TARGET/windows-x64/EinsteinAtHome/source/lalsuite/lalapps/src/pulsar/GCT/HierarchSearchGCT_OpenCL.c:137): Internal function call failed: Generic failure
XLAL Error - XLALSemiCohStep_GPU (/home/jenkins/workspace/workspace/EaH-GW-OpenCL-Testing/SLAVE/MinGW6.3/TARGET/windows-x64/EinsteinAtHome/source/lalsuite/lalapps/src/pulsar/GCT/HierarchSearchGCT.c:4633): Check failed: (*usefulparams->gct_gpu_funcs->SemiCohStep) ( coarsegrid, finegrid, stacks, NSegmentsInv, toplists_sortby, usefulparams->BSGLsetupGPU, usefulparams->computeBSGL, usefulparams->getMaxFperSeg, toplist1_last_entryGPU->data, toplist2_last_entryGPU->data, toplist3_last_entryGPU->data ) == XLAL_SUCCESS
XLAL Error - XLALSemiCohStep_GPU (/home/jenkins/workspace/workspace/EaH-GW-OpenCL-Testing/SLAVE/MinGW6.3/TARGET/windows-x64/EinsteinAtHome/source/lalsuite/lalapps/src/pulsar/GCT/HierarchSearchGCT.c:4633): Internal function call failed: Generic failure
XLAL Error - MAIN (/home/jenkins/workspace/workspace/EaH-GW-OpenCL-Testing/SLAVE/MinGW6.3/TARGET/windows-x64/EinsteinAtHome/source/lalsuite/lalapps/src/pulsar/GCT/HierarchSearchGCT.c:2242): Check failed: XLALSemiCohStep_GPU( &coarsegrid, &finegrid, nStacks, &usefulParams, NSegmentsInv, uvar->SortToplist, compartment, compartment2, compartment3) == XLAL_SUCCESS
XLAL Error - MAIN (/home/jenkins/workspace/workspace/EaH-GW-OpenCL-Testing/SLAVE/MinGW6.3/TARGET/windows-x64/EinsteinAtHome/source/lalsuite/lalapps/src/pulsar/GCT/HierarchSearchGCT.c:2242): Internal function call failed: Generic failure
2022-12-02 13:55:08.2799 (2908) [CRITICAL]: ERROR: MAIN() returned with error '-1'
thanks. I got some
)
thanks. I got some too.
are these tasks smaller than what production tasks will be? or are they full size?
_________________________________________________________________________
solling2 wrote: For : O3
)
here: CL_MEM_OBJECT_ALLOCATION_FAILURE
how much VRAM does your GPU have? these tasks use about 3.2GB of VRAM
_________________________________________________________________________
Ian&Steve C.
)
"3.072" is what my tool says. So: thanks for that hint! I'll stick to FGRPB since they require less than1.0.
Should we copy/paste
)
Should we copy/paste computation errors here? I am not sure how you see the errors on the backend and what's easiest.
I certainly see these. No
)
I certainly see these. No need to paste these here.
BM
Teh VRAM usage / requirement
)
The VRAM usage / requirement is normally calculated from the memory requirement of the WU, which is put there by the workunit generator. Apparently there's something wrong with that. Until I can properly fix it next week, I now set a constant value of 3.2GB.
BM
Note: I don't have time to
)
Note: I don't have time to look into each and every error, especially over the weekend. I'll run a statistic on the results and errors on Monday. In the meantime: it's Beta test. if you don't like what your machine does or does not - feel free to opt out.
BM