cwpearson/astaroth
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Device id is now properly set when using MPI and there are multiple visible GPUs per node

Browse Source
This commit is contained in:
jpekkila
2019-11-26 16:54:56 +02:00
parent 53695d66a3
commit 5e3caf086e
1 changed files with 3 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
src/core/device.cu
View File

@@ -1242,7 +1242,9 @@ acDeviceRunMPITest(void)
//////////////////////////////////////////////////////////////////////////////////////////////// ////////////////////////////////////////////////////////////////////////////////////////////////
Device device; Device device;
acDeviceCreate(0, submesh_info, &device); int devices_per_node = -1;
cudaGetDeviceCount(&devices_per_node);
acDeviceCreate(pid % devices_per_node, submesh_info, &device);
acDeviceLoadMesh(device, STREAM_DEFAULT, submesh); acDeviceLoadMesh(device, STREAM_DEFAULT, submesh);
// Benchmark // Benchmark