Gromacs  2025.0-dev-20241009-5c23d5f
 All Classes Namespaces Files Functions Variables Typedefs Enumerations Enumerator Friends Macros Groups Pages
List of all members | Public Member Functions | Public Attributes
gmx::GpuPairlist Class Reference

#include <gromacs/nbnxm/gpu_types_common.h>

+ Collaboration diagram for gmx::GpuPairlist:

Description

GPU pair list structure.

Public Member Functions

 GpuPairlist (const GpuPairlist &)=delete
 Do not allow copy construct.
 
 GpuPairlist (GpuPairlist &&)=delete
 Do not allow move construct until device buffers have ownership semantics.
 
GpuPairlistoperator= (const GpuPairlist &)=delete
 Do not allow copy assign.
 
GpuPairlistoperator= (GpuPairlist &&)=delete
 Do not allow move assign until device buffers have ownership semantics.
 

Public Attributes

int numAtomsPerCluster = -1
 number of atoms per cluster
 
int numSci = -1
 size of sci, # of i clusters in the list
 
int sciAllocationSize = -1
 allocation size of sci
 
DeviceBuffer< nbnxn_sci_tsci = nullptr
 list of i-cluster ("super-clusters")
 
GpuPairlistSorting sorting
 sorted pair list and data used for sorting
 
int numPackedJClusters = -1
 total # of packed j clusters
 
int packedJClustersAllocationSize = -1
 allocation size of cjPacked
 
DeviceBuffer< nbnxn_cj_packed_tcjPacked = nullptr
 Packed j cluster list, contains j cluster number and index into the i cluster list.
 
int numIMask = -1
 

of packed j clusters * # of warps


 
int iMaskAllocationSize = -1
 allocation size of imask
 
DeviceBuffer< unsigned int > imask = nullptr
 imask for 2 warps for each 4*j cluster group
 
DeviceBuffer< nbnxn_excl_texcl = nullptr
 atom interaction bits
 
int numExcl = 1
 count for excl
 
int exclAllocationSize = -1
 allocation size of excl
 
bool haveFreshList = false
 true after search, indicates that initial pruning with outer pruning is needed
 
int rollingPruningNumParts = 0
 the number of parts/steps over which one cycle of rolling pruning takes places
 
int rollingPruningPart = 0
 the next part to which the rolling pruning needs to be applied
 
DeviceBuffer< int > d_rollingPruningPart = nullptr
 device memory buffer (1 value per thread block) for next part to which the rolling pruning needs to be applied
 
int d_rollingPruningPartSize = -1
 size of rolling pruning part buffer on device
 
int d_rollingPruningPartAllocationSize = -1
 allocated size of rolling pruning part buffer on device
 

The documentation for this class was generated from the following files: