When i tried to run mpm_test_oreo i got warning about illegal memory access:
❯ ./build/tests/mpm_test_oreo
[Cuda] Initialize device index to 0
[Cuda] Device Property:
GPU Device: 0
Global Memory: 12488343552 bytes
Shared Memory: 49152 bytes
Register Per SM: 65536
Multi-processor count: 28
SM compute capabilities: 8.6.
[Cuda] Created 32 streams for device 0.
[MPM Engine info] [10:43:27 +07:00] [thread 82916] Set Default dt = 1.0673906e-05
[MPM Engine info] [10:43:27 +07:00] [thread 82916] Start initializing particles.
[MPM Engine info] [10:43:27 +07:00] [thread 82916] Initializing model with particle count 1615062
[MPM Engine info] [10:43:27 +07:00] [thread 82916] Initializing model with particle count 658581
[MPM Engine info] [10:43:27 +07:00] [thread 82916] Finished initializating particles.
GPUassert: an illegal memory access was encountered /run/media/overload/Transcend/Project/CKMPM/include/mpm_engine.cuh 228I also ran compute-sanitizer to check what was wrong and got the following errors:
- Invalid atomic size
- Out of bounds memory address access
- And mismatching particle initialization
Here is the complete compute-sanitizer output:
❯ compute-sanitizer ./build/tests/mpm_test_oreo
========= COMPUTE-SANITIZER
[Cuda] Initialize device index to 0
[Cuda] Device Property:
GPU Device: 0
Global Memory: 12488343552 bytes
Shared Memory: 49152 bytes
Register Per SM: 65536
Multi-processor count: 28
SM compute capabilities: 8.6.
[Cuda] Created 32 streams for device 0.
[MPM Engine info] [10:52:14 +07:00] [thread 90740] Set Default dt = 1.0673906e-05
[MPM Engine info] [10:52:14 +07:00] [thread 90740] Start initializing particles.
[MPM Engine info] [10:52:14 +07:00] [thread 90740] Initializing model with particle count 1615062
[MPM Engine info] [10:52:14 +07:00] [thread 90740] Initializing model with particle count 658581
[MPM Engine info] [10:52:14 +07:00] [thread 90740] Finished initializating particles.
========= Invalid __global__ atomic of size 4 bytes
========= at void mpm::ActivateBlocksWithParticles<mpm::test::MPMTestScene::MPMTestOreoConfig, mpm::MPMPartition<mpm::MPMGridConfig<mpm::MPMDomain<mpm::MPMDomainRange<(int)64, (int)64, (int)64>, mpm::MPMDomainOffset<(int)0, (int)0, (int)0>, (int)-1>, mpm::meta::Empty>>, mpm::Matrix<float, (unsigned long)3, (unsigned long)1> *>(T1, unsigned int, T3, T2)+0x280
========= by thread (5,0,0) in block (0,0,0)
========= Access to 0x7feec87fc530 is out of bounds
========= and is 15.056 bytes before the nearest allocation at 0x7feec8800000 of size 1.048.576 bytes
========= Saved host backtrace up to driver entry point at kernel launch time
========= Host Frame: cudaLaunchKernel [0x101c77] in mpm_test_oreo
========= Host Frame: void mpm::ActivateBlocksWithParticles<mpm::test::MPMTestScene::MPMTestOreoConfig, mpm::MPMPartition<mpm::MPMGridConfig<mpm::MPMDomain<mpm::MPMDomainRange<64, 64, 64>, mpm::MPMDomainOffset<0, 0, 0>, -1>, mpm::meta::Empty> >, mpm::Matrix<float, 3ul, 1ul>*>(mpm::test::MPMTestScene::MPMTestOreoConfig, unsigned int, mpm::Matrix<float, 3ul, 1ul>*, mpm::MPMPartition<mpm::MPMGridConfig<mpm::MPMDomain<mpm::MPMDomainRange<64, 64, 64>, mpm::MPMDomainOffset<0, 0, 0>, -1>, mpm::meta::Empty> >) [0x16227] in mpm_test_oreo
========= Host Frame: void mpm::MPMEngine<mpm::MPMGridConfig<mpm::MPMDomain<mpm::MPMDomainRange<64, 64, 64>, mpm::MPMDomainOffset<0, 0, 0>, -1>, mpm::meta::Empty> >::InitialSetup<mpm::test::MPMTestScene::MPMTestOreoConfig>(mpm::test::MPMTestScene::MPMTestOreoConfig const&) [0x838b2] in mpm_test_oreo
========= Host Frame: main [0x15be4] in mpm_test_oreo
=========
========= Program hit cudaErrorLaunchFailure (error 719) due to "unspecified launch failure" on CUDA API call to cudaMemcpyAsync.
========= Saved host backtrace up to driver entry point at error
========= Host Frame: void mpm::MPMEngine<mpm::MPMGridConfig<mpm::MPMDomain<mpm::MPMDomainRange<64, 64, 64>, mpm::MPMDomainOffset<0, 0, 0>, -1>, mpm::meta::Empty> >::InitialSetup<mpm::test::MPMTestScene::MPMTestOreoConfig>(mpm::test::MPMTestScene::MPMTestOreoConfig const&) [0x83926] in mpm_test_oreo
========= Host Frame: main [0x15be4] in mpm_test_oreo
=========
GPUassert: unspecified launch failure /run/media/overload/Transcend/Project/CKMPM/include/mpm_engine.cuh 228
========= Target application returned an error
========= ERROR SUMMARY: 2 errorsI assume compilation error was in the play so i recompile it to no avail, the issue persists. Is there any way to adjust the simulation parameters to find the cause of error?
When i tried to run
mpm_test_oreoi got warning about illegal memory access:❯ ./build/tests/mpm_test_oreo [Cuda] Initialize device index to 0 [Cuda] Device Property: GPU Device: 0 Global Memory: 12488343552 bytes Shared Memory: 49152 bytes Register Per SM: 65536 Multi-processor count: 28 SM compute capabilities: 8.6. [Cuda] Created 32 streams for device 0. [MPM Engine info] [10:43:27 +07:00] [thread 82916] Set Default dt = 1.0673906e-05 [MPM Engine info] [10:43:27 +07:00] [thread 82916] Start initializing particles. [MPM Engine info] [10:43:27 +07:00] [thread 82916] Initializing model with particle count 1615062 [MPM Engine info] [10:43:27 +07:00] [thread 82916] Initializing model with particle count 658581 [MPM Engine info] [10:43:27 +07:00] [thread 82916] Finished initializating particles. GPUassert: an illegal memory access was encountered /run/media/overload/Transcend/Project/CKMPM/include/mpm_engine.cuh 228I also ran
compute-sanitizerto check what was wrong and got the following errors:Here is the complete
compute-sanitizeroutput:❯ compute-sanitizer ./build/tests/mpm_test_oreo ========= COMPUTE-SANITIZER [Cuda] Initialize device index to 0 [Cuda] Device Property: GPU Device: 0 Global Memory: 12488343552 bytes Shared Memory: 49152 bytes Register Per SM: 65536 Multi-processor count: 28 SM compute capabilities: 8.6. [Cuda] Created 32 streams for device 0. [MPM Engine info] [10:52:14 +07:00] [thread 90740] Set Default dt = 1.0673906e-05 [MPM Engine info] [10:52:14 +07:00] [thread 90740] Start initializing particles. [MPM Engine info] [10:52:14 +07:00] [thread 90740] Initializing model with particle count 1615062 [MPM Engine info] [10:52:14 +07:00] [thread 90740] Initializing model with particle count 658581 [MPM Engine info] [10:52:14 +07:00] [thread 90740] Finished initializating particles. ========= Invalid __global__ atomic of size 4 bytes ========= at void mpm::ActivateBlocksWithParticles<mpm::test::MPMTestScene::MPMTestOreoConfig, mpm::MPMPartition<mpm::MPMGridConfig<mpm::MPMDomain<mpm::MPMDomainRange<(int)64, (int)64, (int)64>, mpm::MPMDomainOffset<(int)0, (int)0, (int)0>, (int)-1>, mpm::meta::Empty>>, mpm::Matrix<float, (unsigned long)3, (unsigned long)1> *>(T1, unsigned int, T3, T2)+0x280 ========= by thread (5,0,0) in block (0,0,0) ========= Access to 0x7feec87fc530 is out of bounds ========= and is 15.056 bytes before the nearest allocation at 0x7feec8800000 of size 1.048.576 bytes ========= Saved host backtrace up to driver entry point at kernel launch time ========= Host Frame: cudaLaunchKernel [0x101c77] in mpm_test_oreo ========= Host Frame: void mpm::ActivateBlocksWithParticles<mpm::test::MPMTestScene::MPMTestOreoConfig, mpm::MPMPartition<mpm::MPMGridConfig<mpm::MPMDomain<mpm::MPMDomainRange<64, 64, 64>, mpm::MPMDomainOffset<0, 0, 0>, -1>, mpm::meta::Empty> >, mpm::Matrix<float, 3ul, 1ul>*>(mpm::test::MPMTestScene::MPMTestOreoConfig, unsigned int, mpm::Matrix<float, 3ul, 1ul>*, mpm::MPMPartition<mpm::MPMGridConfig<mpm::MPMDomain<mpm::MPMDomainRange<64, 64, 64>, mpm::MPMDomainOffset<0, 0, 0>, -1>, mpm::meta::Empty> >) [0x16227] in mpm_test_oreo ========= Host Frame: void mpm::MPMEngine<mpm::MPMGridConfig<mpm::MPMDomain<mpm::MPMDomainRange<64, 64, 64>, mpm::MPMDomainOffset<0, 0, 0>, -1>, mpm::meta::Empty> >::InitialSetup<mpm::test::MPMTestScene::MPMTestOreoConfig>(mpm::test::MPMTestScene::MPMTestOreoConfig const&) [0x838b2] in mpm_test_oreo ========= Host Frame: main [0x15be4] in mpm_test_oreo ========= ========= Program hit cudaErrorLaunchFailure (error 719) due to "unspecified launch failure" on CUDA API call to cudaMemcpyAsync. ========= Saved host backtrace up to driver entry point at error ========= Host Frame: void mpm::MPMEngine<mpm::MPMGridConfig<mpm::MPMDomain<mpm::MPMDomainRange<64, 64, 64>, mpm::MPMDomainOffset<0, 0, 0>, -1>, mpm::meta::Empty> >::InitialSetup<mpm::test::MPMTestScene::MPMTestOreoConfig>(mpm::test::MPMTestScene::MPMTestOreoConfig const&) [0x83926] in mpm_test_oreo ========= Host Frame: main [0x15be4] in mpm_test_oreo ========= GPUassert: unspecified launch failure /run/media/overload/Transcend/Project/CKMPM/include/mpm_engine.cuh 228 ========= Target application returned an error ========= ERROR SUMMARY: 2 errorsI assume compilation error was in the play so i recompile it to no avail, the issue persists. Is there any way to adjust the simulation parameters to find the cause of error?