| commit | 3623153a2518fed5a284bec5a531e651d3b74c7d | |
| author | Szilárd Páll <pall.szilard@gmail.com> | |
| Thu, 28 Sep 2017 14:22:17 +0000 (28 16:22 +0200) | ||
| committer | Berk Hess <hess@kth.se> | |
| Thu, 28 Sep 2017 17:42:34 +0000 (28 19:42 +0200) | ||
| tree | 263a9a0b1a1d1fdb68d21578904650257afeaf59 | treesnapshot (tar.gz zip) |
| parent | 9795725750b49fd304f7897b977bbd97367d59bd | commitdiff |
Correct misplaced CUDA timing event record
The event record happens after a cudaStreamWaitEvent is placed in the
non-local stream and therefore, in that stream it would include the wait
time in the measurement. However, as with DD / two streams timing can
not be performed due to the limitations of CUDA events, in practice this
was never an issue.
Change-Id: I2ca89c7acd461e480a324d40911dd4c6f5aac478
The event record happens after a cudaStreamWaitEvent is placed in the
non-local stream and therefore, in that stream it would include the wait
time in the measurement. However, as with DD / two streams timing can
not be performed due to the limitations of CUDA events, in practice this
was never an issue.
Change-Id: I2ca89c7acd461e480a324d40911dd4c6f5aac478
| src/gromacs/mdlib/nbnxn_cuda/nbnxn_cuda.cu | diffblobblamehistory |