Closed fwyzard closed 2 years ago
Tested on patatrack02
with multiple GPUs:
$ ./alpaka --cuda --numberOfThreads 20 --numberOfStreams 20 --maxEvents 20000 --validation
Found 5 devices:
- NVIDIA A10
- NVIDIA A10
- Tesla V100-PCIE-32GB
- Tesla T4
- Tesla T4
Processing 20000 events, of which 20 concurrently, with 20 threads.
CountValidator: all 20000 events passed validation
Average relative track difference 0.000893301 (all within tolerance)
Average absolute vertex difference 0.0005 (all within tolerance)
Processed 20000 events in 6.907347e+00 seconds, throughput 2895.47 events/s, CPU usage per thread: 52.4%
Update the version of alpaka to the HEAD of the develop branch as of 2022.03.10, corresponding to the commit 4189bd1e2d .
Relevant changes:
alpaka::extent
namespace;alpaka::getNativeHandle
instead of data members.Other changes:
getDevIndex
togetDeviceIndex
;getDeviceIndex
.