Commit Graph

11767 Commits

Author SHA1 Message Date
Stan Moore 4f195254af Fixing bug with Kokkos and reading restart files 2016-12-28 10:59:30 -07:00
Dan Ibanez 9a0007a13f rename region_block_kokkos inside()
this prevents compiler confusion with
the inside() function provided by
the normal region_block, as the two
should be completely separate.
2016-12-28 10:56:16 -07:00
Dan Ibanez 994f36bc6f silence "implicit dereference" warning
see kokkos/kokkos#177 for detailed
discussion of the issue and fix
2016-12-28 10:56:06 -07:00
Dan Ibanez b3557bfbf5 add missing return in comm_tiled_kokkos 2016-12-28 10:55:51 -07:00
Dan Ibanez 371df8ea72 repair sync bugs in fix_momentum_kokkos
Since the Group class is completely
unaware of Kokkos, the direct calls from
FixMomentumKokkos to Group methods
need to be preceded by atomKK->sync calls
for every atom variable that Group intends
to use.
fix_momentum_kokkos definitely does not
work on GPUs prior to this commit.
2016-12-28 10:51:59 -07:00
Dan Ibanez 06ae2804f6 ensure velocity is marked as modified before syncing
it worked before, but this seems more reliable
2016-12-28 10:51:20 -07:00
Efrem Braun 68814d4fc8 Made documentation changes to fix nvk 2016-12-28 16:26:05 +01:00
Efrem Braun 616ca1de03 Fix nvk implemented. 2016-12-28 16:17:07 +01:00
Giacomo Fiorin b0263e87bb Fix missing force with extended-Lagrangian mass 2016-12-27 17:16:32 -05:00
Giacomo Fiorin 925f42727f Fix typo 2016-12-27 14:26:43 -05:00
Giacomo Fiorin f553e230db Update Colvars library to version 2016-12-22
Significant code cleanup and several fixes (walls + extended Lagrangian)

New harmonicWalls bias to apply confining boundaries with time-dependent force
constant & integration
2016-12-27 13:17:34 -05:00
Giacomo Fiorin 6ab716164b Fix seg fault for gyration collective variable 2016-12-27 13:17:02 -05:00
Giacomo Fiorin 7a45c72b97 Allow extended Lagrangian on non-scalar collective variables 2016-12-27 12:35:30 -05:00
Axel Kohlmeyer 634eb357d2 synchronize pppm/cg with pppm and remove block on triclinic 2016-12-24 16:09:18 -05:00
Tim Mattox a1036f2d74 USER-DPD: bugfix for new PairExp6rx::polynomialScaling() function. 2016-12-23 11:36:59 -05:00
Tim Mattox c301d70333 USER-DPD: update example input and output (see commits 3faa57 and eff7238) 2016-12-22 17:32:41 -05:00
Tim Mattox 781daad2a0 USER-DPD: update documentation for changes in commit 3faa57 2016-12-22 17:31:17 -05:00
Tim Mattox 3faa57a413 USER-DPD: Several updates to *_rx files:
1) Added MY_EPSILON to handle machine precision checks
2) Removed error checks for DPD-RX; enabled use with DPD-E simulations
3) Expanded the EOS functional form to allow corrections
   in the thermo file or on the command line
4) Updated naming convention from fraction to mixWtSite*
5) Changed the name of getParams() method to getMixingWeights()
6) getMixingWeights() now handles fractional and molecular weighting
7) Added optional argument (fractional or molecular) to pair_style command
8) Added argument to specify the exp6 parameter scaling method
   NOTE: Requires additional arguments in the pair coefficients,
   thus command line areguments are NOT backward-compatible.
2016-12-22 17:15:09 -05:00
Tim Mattox fa435fb514 USER-DPD: remove unused variable, and convert some constants to MY_EPSILON 2016-12-22 17:05:49 -05:00
Tim Mattox ba96fcc15a USER-DPD: update reference output to reflect changes from commit eff7238 2016-12-22 16:21:05 -05:00
stanmoore1 304f65b164 Merge pull request #2 from ibaned/fix-momentum-kokkos
Fix momentum kokkos from Dan Ibanez
2016-12-21 12:23:32 -07:00
Dan Ibanez 4c33f31265 Merge branch 'debug-cuda' into fix-momentum-kokkos 2016-12-21 10:57:44 -07:00
Dan Ibanez ae8d882b03 need to sync new velocities back to host 2016-12-21 10:57:18 -07:00
Dan Ibanez 7559bc9c5f workaround CUDA View::reference_type
it is not an lvalue reference in CUDA.
also, the previous code assumed contiguous
entries for one atom; now it should be
robust in the case of LayoutLeft.
2016-12-21 10:21:55 -07:00
Steve Plimpton 62dea1bb63 21Dec16 patch 2016-12-21 09:53:32 -07:00
sjplimp 800ff43413 Merge pull request #304 from timattox/USER-DPD_whitespace
USER-DPD: Whitespace cleanup to pair_dpd_fdt_energy.cpp
2016-12-21 09:51:10 -07:00
Steve Plimpton 9161bd98bf fixed bug with pair hybrid/overaly and manybody potentials finding the right skip neighbor method 2016-12-21 09:50:29 -07:00
Dan Ibanez f3327ca214 allow constructing Few from array on device 2016-12-21 09:21:11 -07:00
Dan Ibanez 54963ba7da allow fix momentum angular with CUDA 2016-12-21 08:55:33 -07:00
Dan Ibanez ea76041803 CUDA-friendly fix_momentum_kokkos angular 2016-12-21 08:54:40 -07:00
Dan Ibanez 7fb4faa439 draft CUDA-callable version of Domain::unmap 2016-12-21 08:53:57 -07:00
Dan Ibanez 41c9357dde allow constructing Few from C array 2016-12-21 08:53:27 -07:00
Dan Ibanez d1a55ad2e0 add kokkos_few.h to .gitignore for /src 2016-12-21 08:07:19 -07:00
Dan Ibanez d9a0f575f6 get fix_momentum_kokkos to compile on White 2016-12-21 08:01:19 -07:00
Dan Ibanez 01e3a31639 put Few in the global namespace... 2016-12-21 08:01:06 -07:00
Dan Ibanez 992becc75f silence warning about partial x2lamda overload 2016-12-21 08:00:42 -07:00
Dan Ibanez 8b5e15e979 add a Makefile for the White testbed 2016-12-20 16:30:50 -07:00
Dan Ibanez b2b33cca16 start working on fix_momentum Kokkos+CUDA
it doesn't compile anymore,
all in good time...
2016-12-20 16:30:21 -07:00
Dan Ibanez 2ceee6b9be install kokkos_few, and remove Int 2016-12-20 16:19:17 -07:00
Dan Ibanez 386c12c970 start porting my Few class into this code 2016-12-20 16:09:29 -07:00
Dan Ibanez 590f317550 fix_momentum_kokkos: don't override init() 2016-12-20 11:03:16 -07:00
Tim Mattox c4e02a5d2b USER-DPD: more whitespace fixes 2016-12-20 11:17:11 -05:00
Dan Ibanez c7ac9e79cb preemptive changes for Kokkos+CUDA
I haven't compiled in that mode yet
(don't know how), but these are some
changes I suspect the compiler
will require.
2016-12-19 13:17:22 -07:00
Tim Mattox 2ba424e1a3 USER-DPD: Whitespace cleanup to pair_dpd_fdt_energy.cpp 2016-12-19 15:11:52 -05:00
Dan Ibanez ca30c1ec88 got fix_momentum_kokkos to compile
there are likely still some compile
errors for Kokkos+CUDA...
2016-12-19 13:08:09 -07:00
Dan Ibanez a1b441a71f draft the parallel_* constructs based on lambdas
LAMMPS_LAMBDA was added to kokkos_type.h to
facilitate this.

some member variables will likely need local copies
in the fix_momentum code.
2016-12-19 12:02:12 -07:00
Dan Ibanez f6f2170369 first draft
copied fix_momentum, work on the Kokkos
View types for the arrays.
the next step is parallel_for and parallel_reduce
transformations.
2016-12-19 11:28:24 -07:00
Steve Plimpton 81a2db8a0c 17Dec16 patch 2016-12-16 11:36:54 -07:00
Steve Plimpton 0a176841e7 extra python_wrapper change needed for last patch 2016-12-16 11:35:42 -07:00
Steve Plimpton 3027ac9250 patch 16Dec16 2016-12-16 10:30:57 -07:00