Paul Bauer [Wed, 13 Feb 2019 17:46:15 +0000 (18:46 +0100)]
Refactor t_molinfo to MoleculeInformation class
Refs #2833
Change-Id: I66a6afc8ade636fb3002492e6c4e1d30ae51019a
Berk Hess [Thu, 17 Jan 2019 13:25:50 +0000 (14:25 +0100)]
Clean up nbnxm enums
Converted the nbnxm kernel and ewald excl enums to enum classes.
Added resource, kernel setup and pairlist type enums.
Also made pairlistSets_ private.
Change-Id: Ic8ff8c59cf8a72d3462b8a0f7382cd1636136c40
Artem Zhmurov [Tue, 15 Jan 2019 15:04:49 +0000 (16:04 +0100)]
Test for LINCS and SHAKE constraints.
This version updates the tests making the selection of the
constraining algorithm more abstract. Makes it possible
to use the same test routines for new implementations (e.g.
CPU- or GPU-based) or (and) algorithms (e.g. LINCS or SHAKE).
Partly this is preparation for the GPU-based version of
the constraints (Refs #2816).
Change-Id: Ice7dfdcc6d86c04656b0a1dd4e328c5afdb8a263
Berk Hess [Fri, 11 Jan 2019 15:34:13 +0000 (16:34 +0100)]
Simplify make_pairlist() call signature
The nonbonded_verlet_group_t struct is no longer needed now the local
and non-local interactions always use the same kernel types.
This simplifies the make_pairlist() and put_on_grid() call signature.
Removed direct access to kernel and ewald exclusion types from outside
the nbnxn module.
Merged the pruning setup and the transfer of the pairlist to the GPU
into nbnxn_make_pairlist().
Also removed the ePBC argument from nbnxn_put_on_grid().
Change-Id: Id96d7e5aa6ce846e9d38614f7edcdced89687799
Berk Hess [Tue, 22 Jan 2019 13:29:15 +0000 (14:29 +0100)]
Add havePPDomainDecomposition()
When using 1 PP and 1 PME rank, non-local code paths were called
for with GPUs. Now the DOMAINDECOMP(cr) conditional for the paths
has been replaced by havePPDomainDecomposition().
Change-Id: I1b7036b0759a316c68d5c00bd5814b067c9fd9cd
Paul Bauer [Wed, 20 Feb 2019 18:09:31 +0000 (19:09 +0100)]
Merge "Merge branch 'release-2019' into master"
Berk Hess [Fri, 25 Jan 2019 09:58:28 +0000 (10:58 +0100)]
Refactor DLB region handling
The DLB region was controlled with two clunky enums and several
separate functions. These are now replaced by a single class that
handles the conditionals and has the open/close functions as methods.
This change also disabled the DLB regions with 1 PP and 1 PME rank,
since DLB is not applicable in this case.
Change-Id: I8a375a46b855ec1f1fcdf6573f44a3bf553410b9
Paul Bauer [Wed, 20 Feb 2019 12:50:49 +0000 (13:50 +0100)]
Merge branch 'release-2019' into master
Kept the way topologyinformation was removed from solvate and
insert_molecules in 2019.
Resolved Conflicts:
cmake/gmxVersionInfo.cmake
docs/CMakeLists.txt
docs/user-guide/mdp-options.rst
src/gromacs/gmxpreprocess/insert-molecules.cpp
src/gromacs/gmxpreprocess/solvate.cpp
src/gromacs/mdlib/nbnxn_cuda/nbnxn_cuda_kernel_utils.cuh
src/gromacs/mdlib/nbnxn_ocl/nbnxn_ocl_kernel_utils.clh
src/gromacs/mdlib/sim_util.cpp
src/gromacs/trajectoryanalysis/topologyinformation.cpp
Change-Id: I2576133c4bc7a3c55cedb29e9832bb5afa8fe76f
Paul Bauer [Fri, 15 Feb 2019 14:00:29 +0000 (15:00 +0100)]
Fix release notes formatting
Change-Id: Icea8712088068c14dbfcb20e403dd68f9aa11227
Paul Bauer [Wed, 13 Feb 2019 12:54:24 +0000 (13:54 +0100)]
Prepare for 2019.2
Reset doi strings.
Change-Id: I497ce4e5afe3a0f776dfba107e45d74bc51d4cc8
Artem Zhmurov [Tue, 15 Jan 2019 16:11:14 +0000 (17:11 +0100)]
PBC setup for GPU is moved into pbcutils.
The CUDA device function setPbcAiuc is moved from listed-forces
to pbcutils so that other parts of the code can use it.
Change-Id: I8bbbadc95a6bc72f9abeba481981b66b192a2804
Roland Schulz [Sat, 28 Apr 2018 01:20:33 +0000 (18:20 -0700)]
Add GMX_OCL_SHOW_DIAGNOSTICS env option
Change-Id: Ie886544d40ff78fccd301b87d610ae8c42483b14
Roland Schulz [Mon, 28 Jan 2019 04:49:02 +0000 (20:49 -0800)]
Improve libstd++ handling
clang and icc use libstdc++ under Linux by default.
Previously such dependencies were handled differently than other
dependencies, in that we relied on the compiler autodetection, and
required the user to use the compiler flags to specify a proper
version.
Now we discover libstdc++ using cmake by finding g++, unless the user
or the toolchain they chose have already managed which gcc toolchain
to use. Then the found version is provided to the compiler with the
proper flags.
This has the advantages:
- Makes it easy for us to check the required version is used.
Version check for libstdc++ 5 for C++14 is added.
- Makes it easy for the user to provide the correct version
because standard cmake variables work.
- Makes build reproducible because the instance of libstdc++ used
is constant (e.g. cached via GMX_GPLUSPLUS_PATH, or controlled
by the toolchain, but not if the user chose non-reproducibility
by setting CXXFLAGS environment variables). Previously, the
auto-detection depended on the PATH environment variable at
the time of build (not the time of the cmake run). Thus
rebuilding with the same cache file was not guranteed to
give the same result. It could cause cause confusing link
errors if parts were rebuilt differently.
Also managed libc++ for clang-analyzer builds a bit better.
Fixes #2842
Change-Id: I855e16a6d4bd670cfb7acd6ea5c740f3a1b226bf
David van der Spoel [Wed, 13 Feb 2019 19:13:58 +0000 (20:13 +0100)]
Precision fix for rescbt code.
The compute_restangles function was not sufficiently
precise. In fact it return completely different
results in single and double precision. By making part
of the function work in double the issue is fixed.
New tests added.
Part of #2795
Fixes #2862
Change-Id: Iad7f2bc45be996ba3e16358aab838c5427b157b8
Szilárd Páll [Fri, 15 Feb 2019 15:10:38 +0000 (16:10 +0100)]
Improve GPU target compatibility error message
Change-Id: I53bff22da766c2e8bc083ca6c0e5a08d7ed3b408
Berk Hess [Tue, 15 Jan 2019 08:35:04 +0000 (09:35 +0100)]
Use enum class for nbnxm locality
Converted the interaction and atom locality enum to enum class.
This exposed and fixed a few mixings of the two enums.
Move some files into the Nbnxm namespace.
Add local/non-local organization to the GPU timers.
Note that the use of the Nbnxm namespace outside the nbnxm module
is only temporary. This should all be replaced by methods.
Change-Id: I3c891b3b2b3e14d8175b63f4191f365a5cd64b18
Berk Hess [Tue, 8 Jan 2019 15:09:10 +0000 (16:09 +0100)]
Extract nbnxm grid.h and pairlistwork.h
Change-Id: Ia0da75a21e56d7e473f868cc3758f415ac9e4eb9
Berk Hess [Sat, 16 Feb 2019 17:24:13 +0000 (18:24 +0100)]
Fix compiler warnings
Change-Id: I179eb5a53b14d2bcdc340468d2215a44fcd6a3a4
Berk Hess [Thu, 10 Jan 2019 09:34:44 +0000 (10:34 +0100)]
Move nbnxm details from sim_util.cpp to module
The removes most nbnxm implementation details from sim_util.cpp.
The rest will be done when making nbnxn a class.
Change-Id: Iba411ce87b431a3b9aad656a22dd9114d5854546
Berk Hess [Fri, 4 Jan 2019 10:54:27 +0000 (11:54 +0100)]
Move nbnxn files to nbnxm directory
This change moves all files related to the nbnxn code to an new
nbnxm subdirectory. This change is only code motion and file renaming.
This is also the start of a renaming from nbnxn to nbnxm, which is
a more correct name, since the i- and j-cluster size can be different.
All nbnxn setup code in forcerec.cpp is moved to nbnxm_setup.cpp.
Removed nbnxn_ and _simd... pre- and postfixes from most files.
Renamed nbnxn_search.cpp to pairlist.cpp and nbnxn_search.h to
nbnxm_pairlistset.h. Although this is currently somewhat inconsistent,
future refactoring will make the contents consistent.
Renamed nbnxn_utility.h/cpp to nbnxm_geometry.h/cpp.
Change-Id: I39c17769c566f78b84211510c5ac17793fd131fa
David van der Spoel [Mon, 11 Feb 2019 19:25:50 +0000 (20:25 +0100)]
Tests of restrained listed potentials.
TODO:
- Tests of tabulated potentials.
- Tests of NMR related potentials.
- Tests of restrained angles.
Part of #2795
Change-Id: I27ebc7a5a8805cc92a8e00deb83e36ca3cd87a78
Paul Bauer [Thu, 14 Feb 2019 16:07:59 +0000 (17:07 +0100)]
Fix memory leaks in editconf
Fixed all leaks found by valgrind.
Change-Id: I87c02051e3be772c9706c0995fc0e94bcebd7add
Szilárd Páll [Sat, 10 Nov 2018 20:16:57 +0000 (21:16 +0100)]
Extend GpuEventSynchronizer to allow on-device sync
Added a method that allows inserting a dependency on the synchronizer's
even into a stream passed.
Change-Id: I982ad99b594aab6a6042749260479debb12042a5
Semen Yesylevskyy [Fri, 15 Feb 2019 11:15:26 +0000 (13:15 +0200)]
Fixed misleading wording for membed option.
It is currently wrritten "This approach is unsuitable" but it is only
unsuitable from the code organization point of view, not for conceptual
reasons.
Change-Id: I504bb2a78bdb85cc41aa67b492e2a0fac828d8e4
Semen Yesylevskyy [Fri, 15 Feb 2019 11:00:18 +0000 (13:00 +0200)]
Add new suggested external trajectory analysis tool
Also reorder the list alphabetically, to avoid appearance
of suggesting a preference.
Change-Id: I474d6064e8b119429e7a91e7a804bbe386339e05
Paul Bauer [Tue, 12 Feb 2019 09:12:22 +0000 (10:12 +0100)]
Fix memory issues in preprocess
Refs #2833
Change-Id: Ic4118958d98446d3a164fca10c5618ad0da9bed6
Paul Bauer [Wed, 13 Feb 2019 07:49:20 +0000 (08:49 +0100)]
Refactor vsite topology
Refs #2833
Change-Id: Ibde09f829a72cc5e6467e15cf8ee7e3c183a30ad
Paul Bauer [Wed, 13 Feb 2019 12:49:51 +0000 (13:49 +0100)]
Version 2019.1
Updated regression test hash.
Set DOI strings.
Change-Id: If56a4e81c853af49ebe6a908bb9cc8883d24a7bc
Mark Abraham [Thu, 14 Feb 2019 16:01:25 +0000 (17:01 +0100)]
Improved diagnostics when PME cannot run on GPUs
Fixes #2789, #2823
Change-Id: Iff7609a3774c11329d36bd8557b741a4d29f7c5a
Paul Bauer [Thu, 14 Feb 2019 15:38:09 +0000 (16:38 +0100)]
Fix tune_pme
There was an issue with the file option for the tpr file being empty
that caused a segmentation fault.
Fixes #2827
Change-Id: I2702898d071d6e9325fa2e7ebb223b9043b7109e
Berk Hess [Wed, 13 Feb 2019 08:55:35 +0000 (09:55 +0100)]
Fix mass Delta H double counting
With separate mass lambda parameters, the perturbed mass contribution
to Delta H (not dH/dlambda) was double counted. This was due to an
incorrect fix for issue #2703 for missing Delta H contribution.
Refs #2703
Fixes #2849
Change-Id: I1349260e1e90a17a6b7dbe0d239d9474dcfa928c
Szilárd Páll [Fri, 15 Feb 2019 01:09:39 +0000 (02:09 +0100)]
Extend performance considerations on bonded offload
Refs #2793
Change-Id: I4a8ae8554cf2aad540eb4eb485898f8cabeb3966
Mark Abraham [Tue, 29 Jan 2019 11:56:31 +0000 (12:56 +0100)]
Fix use of selections referring to index groups
Selections that refer only to groups in an index file do not require a
topology file. This (and other behaviours) was inadvertently broken
by the wrong implementation of the default constructor, because that
constructor is used in setting up the analysis environment.
The default constructor of TopologyInformation should establish the
intended invariant, namely that until a topology is loaded, topology
getters should return nullptr.
Fixed that constructor, and fixed the unit test that verifies that
behaviour. Since that test was added at the same time as a change to
the implementation of TopologyInformation in commit
264458c9, the
broken invariant wasn't noticed.
Added a test also to gmx rdf that verifies that the selection
behaviour is fixed.
Fixes #2847
Change-Id: I2149616e0248c6d5cdd5657a8292d2c3346614b9
Mark Abraham [Wed, 13 Feb 2019 14:24:04 +0000 (15:24 +0100)]
Revert changes to implementation of insert-molecules and solvate
This is a partial reversion of
5181f8d1, which made an inappropriate
dependency from preparation tools to TopologyInformation, which is
intended for analysis tools. This is needed in order to fix some
actual analysis tools that were broken by related changes to
TopologyInformation.
Unfortunately in
5181f8d1 readConformation was replaced with
TopologyInformation, and the former is no longer available. Instead,
readConfAndTopology is used in the same way that it was originally
used to implement readConformation. So the behaviour is reverted
even though the code looks different.
Refs #2847, #1862
Change-Id: I6e4900eea38b5c2e1b85b0d48427dabe26b09f9d
Mark Abraham [Wed, 13 Feb 2019 17:44:26 +0000 (18:44 +0100)]
Improved electric-field docs
Change-Id: Ib1c6aa260f54a3933b1d80e375dc6ee007317473
Szilárd Páll [Thu, 7 Feb 2019 13:12:17 +0000 (14:12 +0100)]
Disable instruction fusion on Power8
The -mpower8-fusion flag seems to be the source of incorrect code; not
confirmed, but likely a codegen issue that also affects Power9 with the
similar flag used.
Fixes #2747 #2746 #2734
Change-Id: I56f50e54db47f4fe30c42488f4c4f79ac474518a
Szilárd Páll [Sat, 10 Nov 2018 20:43:49 +0000 (21:43 +0100)]
Minor refactoring in mdrunnner
Simplified some runner.cpp code to reduce clutter and moved a few
variable declarations away from the beginning of the function.
Change-Id: I818196e3b775077f648415ed6fbbd6d2eb91521a
Paul Bauer [Tue, 12 Feb 2019 10:42:13 +0000 (11:42 +0100)]
Refactor rtp renaming
Refs #2833
Change-Id: I2407045adadb3b84810fb6bfd83d5b8e9daba6f5
Mark Abraham [Wed, 6 Feb 2019 09:35:55 +0000 (10:35 +0100)]
Make EnergyOutput wrapper class
Eventually we want a proper class for the energy accumulation and
output behaviours. In the meantime, this hides much about t_mdebin and
t_ebin from integrators, so that we can refactor with minimal
disruption to other work.
Made t_mdebin private to the new EnergyOutput implementation file.
Used bool rather than gmx_bool in the interface of EnergyOutput.
Noted several TODOs for improvements to code structure.
Removed dependency on ebin.h and enxio.h from the new header, so
various other places had to declare their own dependencies.
Change-Id: If09b2c82dd8c139c76cb5fc453d556c26a5eee89
Artem Zhmurov [Wed, 13 Feb 2019 10:14:47 +0000 (11:14 +0100)]
Ignore Visual Studio Code metadata.
Change-Id: Ic03e22eb3246d7d29c4063601cc86cd6a721e81d
Mark Abraham [Tue, 12 Feb 2019 14:57:08 +0000 (15:57 +0100)]
Reduce coupling to paddedvector.h
As we've improved this infrastructure, we no longer use PaddedVector
in so many places.
Change-Id: I996433f4e0f82b00ab579e158f3d71dac9daae0f
Paul Bauer [Tue, 5 Feb 2019 14:55:55 +0000 (15:55 +0100)]
Move simulated tempering default setting
Setting the default value for the simulated tempering when nstexpanded
was not specified in the mdp file was done before the values for the
temperature coupling are known. Moved this to a new location where the
value is known instead.
Added message to user in this case.
Refs #2754
Change-Id: I194829d5a817019da6d63a075e0dd7f4af8a4735
Szilárd Páll [Sat, 10 Nov 2018 20:43:49 +0000 (21:43 +0100)]
Minor refactor of gpuid / task mapping parsing
Added two separate functions that implement the user-provided GPU ID
string and task mapping parsing, respectively. The separation allows the
two to reflect the slight difrerence in requirements on the input (and
allow these to change in the future). Unit tests were updated as well as
change to explicitly reflect this difference between the -gpuid and
-gputasks strings.
Change-Id: I71144a810e38d2f45e965731e245c868d1d35332
Mark Abraham [Tue, 12 Feb 2019 16:35:29 +0000 (17:35 +0100)]
Quiet clang warning
Change-Id: I48358616bd1c3adf0a8d5c53451ec877a8067ed8
Paul Bauer [Tue, 12 Feb 2019 10:05:07 +0000 (11:05 +0100)]
Refactor special bonds
Refs #2833
Change-Id: I2e093d7fb5fa429f969a78eef7da639af3f7cd1e
Paul Bauer [Mon, 4 Feb 2019 11:47:11 +0000 (12:47 +0100)]
Refactor t_rbonded and t_rbondeds
Again, replaced raw allocations with std::container and changed function
signatures and loops to use new behaviour.
Refs #2833
Change-Id: I620866b823d9839cafafe357352e3e7a49033a69
Paul Bauer [Mon, 4 Feb 2019 09:29:17 +0000 (10:29 +0100)]
Refactor t_restp
Replaced raw allocations with std::containers.
Refs #2833
Change-Id: Ie74708fdf04082f37e69e000aef8f5bd7f577dba
Christian Blau [Tue, 15 Jan 2019 12:33:37 +0000 (13:33 +0100)]
3x3 matrices
Matrices with contiguous, aliged memory as C-style "matrix" replacement
Matrix elements are accessed efficiently with bracket notation m(x,y)
or, for compability reasons, with m[x][y] (though deprecated)
Matrices manage their own memory, but provide an implicit conversion to
views and const views on their data which should be used instead of a
copy.
Arithmetic operations will follow in a later patch.
refs #2834
Change-Id: Ia96037384d2e1d774559fbcac97f420c58106625
Mark Abraham [Sun, 10 Feb 2019 23:01:17 +0000 (00:01 +0100)]
Remove gmx::Regex
We now always use std::regex. The wrapper type added the value of
standardizing regex flags, which was useful while we had to use only
the subset that was usable on all platforms. But the wrapper type made
it harder to use than needed, because e.g. they could not be stored in
a container. Removing gmx::Regex will create new opportunities.
Refs #2881
Change-Id: I97974527052b52c17bb26304b214e8b152dafb4b
David van der Spoel [Mon, 11 Feb 2019 21:16:48 +0000 (22:16 +0100)]
Remove buggy debug output.
Code leftover from testing the SW model with model specifics
hardcoded.
Part of #2774
Change-Id: Ida59dbe1b23672d377e834f199c567d4a51aa0b9
David van der Spoel [Sun, 3 Feb 2019 07:49:19 +0000 (08:49 +0100)]
Tests for polarization functions.
Part of #2795
Change-Id: I30d0c67bac3e16412ca2afaa12ea7e28e8c01f6e
Christian Blau [Tue, 5 Feb 2019 18:12:14 +0000 (19:12 +0100)]
view on MultiDimArray
Access the view and a const view on a MultiDimArray.
refs #2834
Change-Id: Ia82983ac9e3bc8edebc53a1c9f3335101ad4c704
Paul Bauer [Fri, 1 Feb 2019 13:48:52 +0000 (14:48 +0100)]
Refactor t_hack to MoleculePatches
Replaced raw allocations with std::containers. Changed loops and
dependent functions to use new format.
Used hopefully better names for types and variables.
Part of preprocessing clean up.
Refs #2833
Change-Id: I518f5b0d5a7f20cbf8bf3636e88084c1daaab5d9
Roland Schulz [Fri, 8 Feb 2019 21:45:15 +0000 (13:45 -0800)]
Remove EmptyArrayRef
Make it more aligned with std::span.
Almost everywhere it wasn't needed anyhow.
Only exception: ternary conditional operator. But there the type
was already explict in all but one case and that one case
(src/gromacs/domdec/distribute.cpp) was confusing because of
multiple implicit conversions.
Related #2859
Change-Id: I0b61abdc2e60285a7ca17e3bdc7e53cb72c5cf6a
Paul Bauer [Wed, 6 Feb 2019 12:47:19 +0000 (13:47 +0100)]
Fix bug in residue type reading
I used the wrong comparison in a recent change, fixed by properly
checking iterators against each other.
Change-Id: Ia077899e84c7af779b873d183d49236d92296dd3
Kevin Boyd [Tue, 5 Feb 2019 04:20:15 +0000 (23:20 -0500)]
Use more vector in selection index groups structs
Removed unused functions
Change-Id: If8a90ef705c3a9c9ab641ddbb64d17607d7a610f
Berk Hess [Fri, 8 Feb 2019 09:57:48 +0000 (10:57 +0100)]
Disallow rerun with perturbed masses or constraints
Since rerun no longer integrates, free-energy contributions due to
changing masses and constraints are no longer computed. Now rerun
exits with a fatal error when such a system is provided.
Fixes #2849
Change-Id: Ic1550fce22f68c573b52d6895e1e61398b671e51
Paul Bauer [Fri, 8 Feb 2019 15:13:46 +0000 (16:13 +0100)]
Change nightly matrix label to not use hwloc
Done because the 1310 agent has issues with the installation.
Change-Id: I3bbe1b15bf4e1c53a12a4fdac2b759726172965a
David van der Spoel [Fri, 8 Feb 2019 22:08:22 +0000 (23:08 +0100)]
Further increase of tolerance for angles.
Part of #2795
Change-Id: I0d698b6c079cfd008a5419ac2bb7de54ea6887a4
David van der Spoel [Fri, 8 Feb 2019 16:04:47 +0000 (17:04 +0100)]
Increased tolerance for angle functions.
Increased the tolerance for urey-bradley and
quartic angles some more.
Part of #2795.
Change-Id: I929957f1a8de7ad78d174cc613c544234f1450d0
Mark Abraham [Fri, 8 Feb 2019 14:54:20 +0000 (15:54 +0100)]
Fix Intel compiler warning
Post-submit warning hopefully goes away now.
Change-Id: Iddbefff5ea884bccc5e3a9a9959cd7bd216f55f4
David van der Spoel [Fri, 8 Feb 2019 09:00:35 +0000 (10:00 +0100)]
Increase tolerance in Angle test.
One of the new listed-forces tests failed due to
precision in post-submit. Increasing the tolerance
should fix it.
Part of #2795
Change-Id: I1586a92642b4fc257ad65d94337e7caf99385f53
Thomas Ullmann [Thu, 7 Feb 2019 17:14:38 +0000 (18:14 +0100)]
separated .rst documentation of dihedral PCA and hydrogen-bonds
The section of the reference manual on hydrogen-bonds was included
in dihedral-pca.rst. This hindered insertion of a section on an
analysis method related to PCA right after the dihedral PCA section.
Change-Id: I4d7c88624a12fc516cb63465098060fc871cc987
Mark Abraham [Thu, 17 Jan 2019 20:01:02 +0000 (21:01 +0100)]
Remove copy_energy from energy-writing steps
Used range-based for to avoid yet another copy of energy data between
intermediate buffers before being written to the .edr file.
Renamed some variable from index to entryIndex for clarity.
This change is covered by the tests on t_mdebin functionality.
Change-Id: Iecaf28c8af48a74abaf2805feb353c99ab613b10
Mark Abraham [Thu, 7 Feb 2019 09:51:41 +0000 (10:51 +0100)]
Reform energy output tests
Separated tests of t_mdebin from tests of t_ebin to prepare for making
t_mdebin a proper class. Improved coverage of t_mdebin (and indirectly
t_ebin). Made done_ebin deallocate ebin, which means scoped_cptr is
now more usable.
Made some operations on t_mdebin more safe, by checking for nullptr or
reordering logic. Technically there are now two more branches when
writing energy output.
Change-Id: I4b056f660d16ce7af225e229030d6729a34acefe
Roland Schulz [Thu, 7 Feb 2019 22:58:05 +0000 (14:58 -0800)]
Fix sign warning
Caused by rebase of
55c76c88c4d (as part of merge) onto
43da62f5e3.
Change-Id: If2d4f2cc19df18297b689d91b7ce865e5b382773
David van der Spoel [Sat, 2 Feb 2019 17:35:19 +0000 (18:35 +0100)]
Tests for listed force potentials.
Make-over of previous bondedtests and update of code
to test everything.
TODO:
- Tests for restraints and polarization.
Part of #2795
Change-Id: I15e4e5053fb0d1dd5968614bd06f899f0d28dc3a
Szilárd Páll [Sun, 6 Jan 2019 10:51:26 +0000 (11:51 +0100)]
Require a 64-bit platform for OpenCL
GROMACS requires a 64-bit OpenCL implementation as the PME code assumes
8-byte size cl_mem. Therefore, we require a 64-bit platform,
where pointers and therefore cl_mem will be of the assumed size.
Fixes #2821
Change-Id: I5fa60419a948e2cfdb344e7611b197550f866262
Christian Blau [Mon, 4 Feb 2019 14:43:30 +0000 (15:43 +0100)]
mdspan - add bracket operator slicing
Add bracket operator that creates from a basic_mdspan using a bracket operator [].
Change-Id: I2727a7af42b169782b6d1542b6248f2fd23032ac
Roland Schulz [Wed, 6 Feb 2019 00:08:49 +0000 (16:08 -0800)]
Revert size to be unsigned for ArrayRef
Given the decision for std::span to be signed this became
inconsistent with future C++ while already being inconsistent
with std::vector. The goal of the original change, to have
signed variables for all arithmetic including loop indices,
can be achieve by using ssize everwhere arithmethic is used.
ssize is both available as free function (also for std::vector)
and member function.
Related #2826
Change-Id: Icc5d0f9561c610cbce34c575d268171e9890ca9c
Roland Schulz [Tue, 5 Feb 2019 23:51:26 +0000 (15:51 -0800)]
Add ssize and remove static_casts
Fixes #2826
Change-Id: I8cb3b64764798b164b355e71fb10874b49060a0a
Thomas Ullmann [Wed, 6 Feb 2019 21:59:02 +0000 (22:59 +0100)]
added a unit test for make_ndx to check generation of default protein index groups
Change-Id: I56b26a7eefeb818ec038696b8bb9e3a47e9ac536
Paul Bauer [Fri, 1 Feb 2019 08:01:16 +0000 (09:01 +0100)]
Refactor t_hackblock
Changed raw allocations to std::container, reworked functions that
depend on them to adapt to new format.
Part of work to change preprocessing to use enw topology datastructures.
Change-Id: Ifd2bf4b18affe7e41db6178a0a713376e7c87f12
Berk Hess [Wed, 6 Feb 2019 12:39:55 +0000 (13:39 +0100)]
Fix segfault with EM, DD and group scheme
Resetting to an old DD state during EM would leave the cg sorting array
used with the group scheme in an invalid state. This could cause
out of bounds vector access one DD step after rejecting an EM step.
When merging this into master branch, prefer to change to use ssize
rather than the static_cast.
Fixes #2813
Change-Id: I7f13b46d7ff5352ce41838b813c46f2e90c93b1c
Thomas Ullmann [Tue, 5 Feb 2019 18:51:59 +0000 (19:51 +0100)]
Fix bug in generating default protein index groups
The bug was introduced in
3c3dd54f00. When recording
atom indices for an index group in a temporary
vector, the indices of the previously processed index
group were not cleared resulting in index groups that
contained the indices of all previously processed
index groups besides the intended ones. This bug also
caused a crash of trajectory analyis tools with
selection options if the selection used such a
predefined index group.
Change-Id: I68fe19f0dfbca07618b882cf8509e33f570a8348
Mark Abraham [Mon, 28 Jan 2019 14:32:05 +0000 (15:32 +0100)]
Remove SIMD support from group scheme
It might take a bit longer to craft the patches that remove the rest
of the group scheme, but we may as well speed up our compilation times
first.
Refs #1852
Change-Id: I082e40e04678c744d9da8119333210878294a021
Paul Bauer [Mon, 4 Feb 2019 15:17:22 +0000 (16:17 +0100)]
Fix segmentation fault in minimize
Fixes #2858
Change-Id: I9cf391be031979607d09087fa851eb644fb7e56e
Roland Schulz [Sat, 2 Feb 2019 18:05:23 +0000 (10:05 -0800)]
Rename all source files from - to _ for consistency.
Refs #2839
Change-Id: I3a1af667b2b266b1fd65586445b5e7a8a1b6b618
Roland Schulz [Fri, 1 Feb 2019 01:43:05 +0000 (17:43 -0800)]
Fix cmake policy warning
Change-Id: I9cb247606b08cdace7880fd42b987f129c8f8dc6
Roland Schulz [Sat, 2 Feb 2019 19:34:02 +0000 (11:34 -0800)]
Replace compat::make_unique with std::make_unique
Change-Id: I6b0db007b8a7b9e1d72374dcee15f66587375f75
Christian Blau [Thu, 24 Jan 2019 22:01:08 +0000 (23:01 +0100)]
mdspan - free begin and end functions
Add functions to address contiguous mdspan data via begin and end function.
Allows range based looping over view elements.
Change-Id: I75dbf9be53d37fc80a023eaa1e8cda59254c0ebd
Christian Blau [Fri, 1 Feb 2019 12:48:47 +0000 (13:48 +0100)]
multidimensional array - bugfix swap
The swap operation for the multidimensional array was wrongly
implemented, because the views were only swapped without resetting them
onto the new data. The tests have been updated to better capture the
swapping behaviour.
Change-Id: Ic800a3ec0e059c8fcdfb4e5528747ca0c70afb82
Roland Schulz [Sat, 2 Feb 2019 17:41:02 +0000 (09:41 -0800)]
Fix Intel compiler on Windows build
Change-Id: I8a89a08c945bb088e4b896b1bc4fedef2320cd01
Roland Schulz [Tue, 22 Jan 2019 19:45:01 +0000 (11:45 -0800)]
Renamed listed-forces to listed_forces
There should be consistency in naming. _ is used in almost
all cases (2744 to 122 in src/gromacs not counting files
changed here). Follow up changes will change the remaining
files.
Change-Id: Iea57accfd94d7ede180b9becd7d607425a8df7f6
Roland Schulz [Sat, 2 Feb 2019 00:21:13 +0000 (16:21 -0800)]
Remove reliance on inconsistent cmake feature testing
Change-Id: I700b581d8e138328524c04cf9de9458e399d45d9
Roland Schulz [Sat, 2 Feb 2019 05:01:03 +0000 (21:01 -0800)]
Add recommendation of how to compile with MPI
Reason for recommendation: reproducible build.
Change-Id: Ib61e71c006cb0d6c852029667cf16696912a712f
Berk Hess [Thu, 31 Jan 2019 21:47:53 +0000 (22:47 +0100)]
Fix incorrect LJ repulsion force switching on GPUs
When using a CUDA or OpenCL GPU, the coefficient for the second order
term for the LJ repulsion in the force (not energy) switching function,
called 'A' in the manual, had the wrong sign in both the force-only
kernels and the force+energy kernels.
Note that the dispersion force switching was correct.
Fixes #2845
Change-Id: Ib5c250a1f370d4c0a4fd652bf6efa03e70e18748
Roland Schulz [Thu, 31 Jan 2019 17:00:24 +0000 (09:00 -0800)]
Fix clfft build
Change-Id: I9df2470aa02644aa4058ccce986b543374f10664
Mark Abraham [Wed, 30 Jan 2019 12:30:19 +0000 (13:30 +0100)]
Merge remote branch release-2018 into release-2019
Change-Id: Ie5a4add2fe010bc94c00418a728f01d3aac6cb9d
Szilárd Páll [Tue, 29 Jan 2019 14:49:18 +0000 (15:49 +0100)]
Fix double-increment of GPU NB launch subcounter
Only increment once per nonbonded task per step.
Fixes #2837
Change-Id: Ic68bf1af184f895f2fb769b026855fd3d48fa849
Mark Abraham [Wed, 30 Jan 2019 08:23:04 +0000 (09:23 +0100)]
Fix release notes issue links
The space means sphinx can't recognize the construct
Change-Id: I8c25114a1f45a7744345e2e8bc6862fdd4568f5f
Mark Abraham [Wed, 30 Jan 2019 08:20:31 +0000 (09:20 +0100)]
Fix release notes issue links
The space means sphinx can't recognize the construct
Change-Id: Ia0b96e8a1ecd24e2bac97d7bc9f176abbe30abc9
Roland Schulz [Wed, 30 Jan 2019 07:15:34 +0000 (23:15 -0800)]
Change all C to C++ outside of TNG
Change-Id: I43464b366cfdc1b816088cf8af117b17843f9359
Roland Schulz [Fri, 25 Jan 2019 21:29:15 +0000 (13:29 -0800)]
Add cmake C++14 feature check
Change-Id: Ia4e2185aca2a8e1ff1b07cc9b49cfee33a6de705
Teemu Murtola [Tue, 29 Jan 2019 07:05:44 +0000 (08:05 +0100)]
Avoid warnings with Apple Clang
The logic for -Wno-missing-prototypes for group kernels left the
warning enabled for Apple Clang, making it impossible to compile
the code with -Werror.
Change-Id: Ib9a0e23bd47f8cc1dd8f656aae49ba27bbf9b28b
Roland Schulz [Fri, 25 Jan 2019 04:23:15 +0000 (20:23 -0800)]
Resolve C++14 TODOs
Also simplify constructor from fixed sized container.
Change-Id: I4983020939b1569949c39f7bf715fe2dbe69ab0d
Mark Abraham [Thu, 10 Jan 2019 09:14:40 +0000 (10:14 +0100)]
Stop trying to check nvcc on Windows
The execute_process() call used in this check is not constructed to
work on Windows, so we should not run it there.
Refs #1616
Change-Id: I2103b78203f71d3f68b54898dd03c8fe0eb0fa4c
Mark Abraham [Fri, 25 Jan 2019 15:00:51 +0000 (16:00 +0100)]
Bump CUDA requirement to 9.0
Also bump requirement for clang CUDA build to clang 6.0
Adjusted test matrices to run the tests in an equivalent and
reasonably balanced way.
Updated developer guide
Change-Id: I577af6267dcf31572e1af73eb46162938ec63feb
David van der Spoel [Sat, 26 Jan 2019 16:50:00 +0000 (17:50 +0100)]
Fixed some typos in comments.
Change-Id: Id738f01c1eb21036cb7d6db680d597f289939fc1
Roland Schulz [Fri, 25 Jan 2019 22:58:09 +0000 (14:58 -0800)]
Fix clang warning
No need to capture proxy by reference.
Change-Id: I2a75a3395a7c7d8164bb308e5f2cdde03f4aeb55