Still fighting with the launch bound fix:
dumpGPUDefParam:void dumpGPUDefParam(const char* outputfile = "parameters.out"){auto param = o2::gpu::internal::GPUDefParametersLoad();printf("Loaded params:\n%s\nWriting them to %s\n", o2::gpu::internal::GPUDefParametersExport(param, false).c_str(), outputfile);FILE* fp = fopen(outputfile, "w+b");fwrite(¶m, 1, sizeof(param), fp);fclose(fp);}minBlockFactor to the call to GPUDefParametersExport and dump that export somehow 29797 | extern "C" {__attribute__((global)) void GPUCA_KRNL_REG_DEFAULT((GPUCA_LB_GPUTPCNeighboursFinder)) krnl_GPUTPCNeighboursFinder( int32_t _iSector_internal ) { __attribute__((shared)) typename GPUTPCNeighboursFinder::GPUSharedMemory smem; GPUTPCNeighboursFinder::template Thread<GPUTPCNeighboursFinder::defaultKernel>((gridDim.x), (blockDim.x), (blockIdx.x), (threadIdx.x), smem, GPUTPCNeighboursFinder::Processor((gGPUConstantMemBuffer.v))[_iSector_internal] ); }}
build.sh . To build, just run docker run -v /cvmfs:/cvmfs alice_gpu_hs23 --backend HIP --arch gfx908CMake Warning (dev) at /opt/rocm/lib/cmake/hip/hip-config-amd.cmake:91 (message): AMDGPU_TARGETS was not set, and system GPU detection was unsuccsesful.
The amdgpu-arch tool failed: Error: 'Failed to get device count' Output: ''
As a result, --offload-arch will not be set for subsuqent compilations, and the default architecture (gfx906 for dynamic build / gfx942 for static build) will be usedCall Stack (most recent call first): /opt/rocm/lib/cmake/hip/hip-config.cmake:149 (include) /alice_hs23/O2/dependencies/FindO2GPU.cmake:258 (find_package) CMakeLists.txt:130 (find_package)This warning is for project developers. Use -Wno-dev to suppress it.
But then:-- Building GPUTracking with HIP support (GPU Target gfx908)-- Using optimized HIP settings for MI100 GPU
And in the actual compilation commands appears --offload-arch=gfx908
run.sh will use dataset within the container