[RFC] Example use of NAMD CudaGlobalMaster interface#783

Merged

giacomofiorin merged 34 commits into

Colvars:masterfrom

HanatoK:api_cudagm

May 12, 2025

Commits on May 12, 2025

Example use of NAMD CudaGlobalMaster interface
HanatoK
committed
refactor: use CUDA to transpose the positions
HanatoK
committed
fix: correct the atom id setup and update the masses and charges
HanatoK
committed
chore: cleanups
HanatoK
committed
chore: remove the debugging log
HanatoK
committed
fix: use the correct path for Colvars source code
HanatoK
committed
prof: use nvtx to profile Colvars
HanatoK
committed
refactor: using custom allocator for CUDA
HanatoK
committed
opt: remove the unnecessary stream synchronization
HanatoK
committed
opt: transpose to the device arrays at first and then copy to pinned
HanatoK
committed
Fix the deallocate of CudaHostAllocator
HanatoK
committed
Fix the compilation with C++11
HanatoK
committed
opt: use the onBuffersUpdated interface to separate the data moving from
HanatoK
committed
feature: support NAMD_TCL
HanatoK
committed
Squashed commit of the following:
HanatoK
committed
fix: fix the support of cv reset
HanatoK
committed
fix: fix the setup and initialization
HanatoK
committed
feat: add TCL and LEPTON to the CMakeLists.txt file
HanatoK
committed
fix: implement TCL callback functions
HanatoK
committed
workaround: use an independent TCL interpreter for Colvars
HanatoK
committed
fix: setup the constants that are missing
HanatoK
committed
fix: move the setup of initial step out of setup()
HanatoK
committed
fix: check the last CUDA error after launching the kernels
HanatoK
committed
fix: fix the total force calculation
HanatoK
committed
Mark this plugin to be incompatible with the Colvars bundled with NAMD
HanatoK
committed
doc: add README.md for the plugin
HanatoK
committed
opt: make sure the DtoH memcpy overlaps with the NB kernel
HanatoK
committed
Track Charm++ include directory

giacomofiorin
authored and
HanatoK
committed
Add missing test input file

giacomofiorin
authored and
HanatoK
committed
refactor: rename the vector types with host-pinned memory allocator
HanatoK
committed
Add missing instruction in README

jhenin
authored and
HanatoK
committed
Move the CudaGlobalMaster plugin to namd/
HanatoK
committed
doc: correct the path in README.md of CudaGlobalMaster
HanatoK
committed
doc: CudaGlobalMaster is currently not compatible with minimization
HanatoK
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[RFC] Example use of NAMD CudaGlobalMaster interface#783

[RFC] Example use of NAMD CudaGlobalMaster interface#783
giacomofiorin merged 34 commits into
Colvars:masterfrom
HanatoK:api_cudagm

Commits on May 12, 2025

Example use of NAMD CudaGlobalMaster interface

refactor: use CUDA to transpose the positions

fix: correct the atom id setup and update the masses and charges

chore: cleanups

chore: remove the debugging log

fix: use the correct path for Colvars source code

prof: use nvtx to profile Colvars

refactor: using custom allocator for CUDA

opt: remove the unnecessary stream synchronization

opt: transpose to the device arrays at first and then copy to pinned

Fix the deallocate of CudaHostAllocator

Fix the compilation with C++11

opt: use the onBuffersUpdated interface to separate the data moving from

feature: support NAMD_TCL

Squashed commit of the following:

fix: fix the support of cv reset

fix: fix the setup and initialization

feat: add TCL and LEPTON to the CMakeLists.txt file

fix: implement TCL callback functions

workaround: use an independent TCL interpreter for Colvars

fix: setup the constants that are missing

fix: move the setup of initial step out of setup()

fix: check the last CUDA error after launching the kernels

fix: fix the total force calculation

Mark this plugin to be incompatible with the Colvars bundled with NAMD

doc: add README.md for the plugin

opt: make sure the DtoH memcpy overlaps with the NB kernel

Track Charm++ include directory

Add missing test input file

refactor: rename the vector types with host-pinned memory allocator

Add missing instruction in README

Move the CudaGlobalMaster plugin to namd/

doc: correct the path in README.md of CudaGlobalMaster

doc: CudaGlobalMaster is currently not compatible with minimization