GitLab

It allows to use a faster float-to-int64 in the HIP platform.

* Minor optimizations to computing single pairs

* Adjusted MAX_BITS_FOR_PAIRS on Ampere

* DrudeLangevinIntegrator has getSystemTemperature()

* DrudeNoseHooverIntegrator has getSystemTemperature()

* StateDataReporter reports system temperature for Drude systems

* Fixed incorrect return type

* Fixed potential invalid memory access

* Fixed exception

* Optimized CudaSort for non-uniformly distributed data

* Optimized OpenCLSort for non-uniformly distributed data

* Further tuned distributing elements between buckets

* Copied optimizations over to OpenCL

* Use multiple streams to broadcast positions

* Use multiple streams to reduce forces

* Adds sync between default stream and peer-copy

* Minor cleanup
Co-authored-by: David Clark <daclark@nvidia.com>

* Use cuCtxPushCurrent() and cuCtxPopCurrent() for selecting CUDA context

* Fixed errors in amoeba coda

* Fixed more errors in context selection

* Allow querying current step count

* Fixed error building Python wrapper

* Began converting AMOEBA to common platform

* Beginning of OpenCL platform for AMOEBA

* Converted AmoebaVdwForce to common platform

* Cleaned up reference AMOEBA tests

* Began converting AmoebaMultipoleForce to common platform

* Continue converting AmoebaMultipoleForce to common platform

* Bug fixes

* Bug fix

* Continue converting AmoebaMultipoleForce to common platform

* Converting AmoebaMultipoleForce and AmoebaGeneralizedKirkwoodForce to common platform

* Converting AmoebaMultipoleForce and AmoebaGeneralizedKirkwoodForce to common platform

* Creating OpenCL version of AmoebaMultipoleForce and AmoebaGeneralizedKirkwoodForce

* Creating OpenCL version of AmoebaMultipoleForce and AmoebaGeneralizedKirkwoodForce

* Creating OpenCL version of AmoebaMultipoleForce and AmoebaGeneralizedKirkwoodForce

* Converted arrays from real3 to real

* Bug fix to OpenCL AmoebaGeneralizedKirkwoodForce

* Fixes for AMD GPUs

* Began converting HippoNonbondedForce to common platform

* Continuing to convert HippoNonbondedForce to common platform

* Continuing to convert HippoNonbondedForce to common platform

* Working on unifying PME kernels

* Fixed error on devices without 64 bit atomics

* Unified PME kernels

* Converted HippoNonbondedForce to common platform

* Creating OpenCL implementation of HippoNonbondedForce

* Continuing OpenCL implementation of HippoNonbondedForce

* Mostly finished OpenCL implementation of HippoNonbondedForce

* Eliminated three component vector types in host code

* Fix errors on CPU OpenCL

* Skip double precision tests for AMOEBA on OpenCL

* Bug fixes

* Bug fixes

* Fixed compilation error

* Converted RPMD plugin to common platform

* Merged RPMD tests for different platforms

* Try to fix errors on CPU OpenCL

* Converted more code to common platform

* Converted more code to common platform

Co-authored-by: David Clark <daclark@nvidia.com>

* Began implementing geometric functions on points

* Started common implementation of point functions

* Completed implementation of point functions for CustomCompoundBondForce

* Implemented point functions for CustomCentroidBondForce

* Implemented point functions for CustomManyParticleForce

* Use point functions to simplify implementation of custom forces

* Removed unnecessary code

* Fixed typo

* Improve selection of architecture to compile for

* Fixed compilation errors on older CUDA versions

* Use command line compiler if it is explicitly specified

* Don't set OPENMM_CUDA_COMPILER on Jenkins

* Frames distance calculation as matrix multiplciation

* Adds comment explaining distance calculation

* Tunes launch bound for cuda11.2

* Simplifies the effective matrix multiplication
Co-authored-by: David Clark <daclark@nvidia.com>

* Changes name of NVRTC program

* Adds launch bounds for findInteractingBlocks

* Replaces launch bound parameter with named constant
Co-authored-by: David Clark <daclark@nvidia.com>

* Adds cuFFT error checking for execution

* Removes tabs
Co-authored-by: David Clark <daclark@nvidia.com>

* CCMA with a small number of constraints uses a single kernel

* Fixed compilation errors in kernel

* Fixed compilation errors in kernel

* Further optimizations to CCMA with few constraints

* Fixed range overflow with very large numbers of atoms

* More fixes to overflow with large numbers of atoms

* Fix test failures