GitLab

Apply heuristics for HIP neighbor-list construction:
use fewer nonbonded force blocks for small neighbor-list systems, use two
tiles per batch for larger atom-block counts, and increase the
findBlocksWithInteractions thread block size for small atom-block counts.

Standard concurrent validation shows no clear per-case regression and a
small geomean throughput improvement over the current blocksPerCU baseline.

- Use bitwise prefix accounting when storing sparse interactions as single pairs in the HIP pair-list kernel. This reduces the number of ballot operations needed to compute per-lane single-pair offsets.
- For HIP CDNA single precision, raise MAX_BITS_FOR_PAIRS to 8 so more sparse interactions are emitted as single pairs instead of full tiles. Keep the existing double precision and RDNA thresholds unchanged.
- Also simplify the HIP LJPME direct correction by computing alpha^2*r2

Avoid forcing Coulomb PME to re-sort whenever LJ-PME is enabled, and give dispersion PME its own atom-grid index and sort state so the performance impact can be measured independently.

* Created ReplicaExchangeSampler

* Improvements to ReplicaExchangeSampler

* PythonForce can be restricted to a subset of particles

* Fix exception with CUDA

* Infrastructure for multistate sampling

* Added computeRelativeEnergies()

* optimization, switch to unordered map

* Remove emplace, reduce reserve allocation

* Cache coefficients for long range correction

* updateParametersInContext() clears cache

* fix 5234 and add test

* clean up docstring and standardize test name

* Fixed issue that caused inefficient sorting when a block contained only one atom

* Add the fix to OpenCL and HIP

* Added two more synonyms for HOH

* Change H20 to H2O

* Update patch documentation

Clarified the definition and purpose of patches.

* Fix typo in RemoveExternalBond tag description

* Update wording in patches

* A few edits to the description of patches

---------
Co-authored-by: Peter Eastman <peter.eastman@gmail.com>

* Make reference/CPU minimizer into a kernel

* Add per-platform support for GPU minimization

* Initial implementation of GPU minimization

* Fixes

* Increase robustness when initial gradient is huge

* Handle overflow leading to non-finite values gracefully

* Handle large forces in single precision more robustly

* Optimize kernels

* Fix kernel launch size

* Update banner years

* Don't create MinimizeKernel until first minimization requested

* Make some compile-time constants into kernel arguments

* Consolidate scale calculation kernel

* Condense alpha/beta reduction kernels using atomics

* Condense line search dot kernels with reductions

* Remove a download, and download grad norm separately

* Asynchronously check lbfgs convergence condition

* Restructure line search to avoid download waiting

* Start line search preemptively in case CPU evaluation is not needed

* In rare cases, constraint error might not decrease after one optimization round

* Better handling of unsupported 64-bit atomics, use FLT_MAX

* Pick gradient mode based on GPU vs. CPU evaluation

* Rework getDiff/getScale reduction, remove reduceBuffer

* Older CUDA might not like float hex literals

* Fix error in a comment

* Residue templates can specify constraints

* Patched template generation preserves constraints

* API for querying devices

* CUDA and HIP implementations of getDevices()

* Fix test failures

* Fix test failures

* CUDA returns correct devices even if no context has been created

* Return a single device for Reference and CPU

* Fix CI failure

* Templates can match whole molecules

* addExtraParticles() supports molecule templates

* Documentation on molecule templates

* Bug fix

* Remove std::enable_if, warpRotateLeft is always used with TILE_SIZE

* Do not use built-in warpSize in constexpr contexts

Starting from ROCm 7 warpSize is no longer constexpr.
findInteractingBlocks.hip uses it for sizes of __shared__ arrays.

* Check if hipHostMallocNumaUser is allowed before using it