- Registered on: 12/21/2018
- Last connection: 04/30/2019
- 01:40 AM GROMACS Revision 42b343b9: Position buffer ops in CUDA
- On all but search steps the buffer ops transform can now be done on a
CUDA GPU. If PME runs on the same GPU the alrea...
- 05:02 PM GROMACS Feature #2934 (New): GPU X Buffer ops
- Implement and improve the GPU version of position buffer operations. Gerrit change 9169 implements the functionality,...
- 05:37 PM GROMACS Feature #2915: GPU direct communications
Szilard wrote (copying from issue 2891):
>How do we provide fallbacks for when i) no MPI is used ii) no CUDA-aware...
- 05:30 PM GROMACS Feature #2891: PME/PP GPU communications
>As noted in CR, we should initiate the PP->PME send exactly at the same location where the CPU path does it; the c...
- 05:29 PM GROMACS Feature #2915 (New): GPU direct communications
- Part of this family of developments involves enablement of direct communications between GPU memory spaces. This issu...
- 05:06 PM GROMACS Revision 8fa8a779: NVIDIA Prototype (nondraft): GPU update&constraints plus D2D comms
- This patch does not require review. It is a (slightly updated) copy of
draft patch 8858, but committed as non-draft p...
- 12:24 PM GROMACS Feature #2817: GPU X/F buffer ops
- Now rebased such that (haveSpecialForces || haveCpuBondedWork) is available. But we can't use it just yet to make the...
- 12:19 PM GROMACS Revision 3cced793: F buffer operations in CUDA
- TODO: split out reduction
Implements part of #2817
- 12:17 PM GROMACS Revision 20934303: Position buffer ops in CUDA
- improve the CUDA kernel
- waits for X copy on the PME stream to finish, need to implement sync
- 05:30 PM GROMACS Revision ad0ca36c: test
- Change-Id: I48aee4be8a9a195c9f6ec8ff65c2f20956093f23
Also available in: Atom