Gromacs  2027.0-dev-20261113-fb41eee
 All Classes Namespaces Files Functions Variables Typedefs Enumerations Enumerator Friends Macros Groups Pages
List of all members | Public Member Functions | Public Attributes
gmx::NbnxmGpu Struct Reference

#include <gromacs/nbnxm/cuda/nbnxm_cuda_types.h>

+ Collaboration diagram for gmx::NbnxmGpu:

Description

Main data structure for CUDA nonbonded force calculations.

Public Member Functions

 NbnxmGpu (const DeviceStreamManager &deviceStreamManager, std::optional< size_t > nLambda)
 

Public Attributes

const DeviceContext & deviceContext
 GPU device context.
 
const HostAllocationPolicy hostAllocationPolicy
 Host allocation policy for buffers for possible GPU transfers.
 
bool bUseTwoStreams = false
 true if doing both local/non-local NB work on GPU
 
bool bNonLocalStreamDoneMarked = false
 true indicates that the nonlocal_done event was marked
 
NBAtomDataGpu * atdat = nullptr
 atom data
 
int * atomIndices = nullptr
 array of atom indices
 
int atomIndicesSize = 0
 size of atom indices
 
int atomIndicesSize_alloc = 0
 size of atom indices allocated in device buffer
 
int * numAtomsPerCell = nullptr
 x buf ops num of atoms
 
int numAtomsPerCellSize = 0
 number of elements in numAtomsPerCell
 
int numAtomsPerCellAlloc = 0
 number of elements allocated in device buffer
 
int * cellToBin = nullptr
 x buf ops bin index mapping
 
int cellToBinSize = 0
 number of elements in cellToBin
 
int cellToBinAlloc = 0
 number of elements allocated in device buffer
 
NBParamGpu * nbparam = nullptr
 parameters required for the non-bonded calc.
 
EnumerationArray
< InteractionLocality,
std::unique_ptr< GpuPairlist > > 
plist = { { nullptr } }
 pair-list data structures (local and non-local)
 
EnumerationArray
< InteractionLocality,
std::unique_ptr< GpuFeplist > > 
feplist = { { nullptr } }
 fep-list data structures (local and non-local)
 
GpuFepHostData * fephostdata = nullptr
 host buffers required for the FEP H2D copies
 
NBStagingData nbst
 staging area where fshift/energies get downloaded
 
EnumerationArray
< InteractionLocality, const
DeviceStream * > 
deviceStreams = { { nullptr } }
 local and non-local GPU streams
 
GpuEventSynchronizer nonlocal_done
 Event triggered when the non-local non-bonded kernel is done (and the local transfer can proceed)
 
GpuEventSynchronizer misc_ops_and_local_H2D_done
 Event triggered when the tasks issued in the local stream that need to precede the non-local force or buffer operation calculations are done (e.g. f buffer 0-ing, local x/q H2D, buffer op initialization in local stream that is required also by nonlocal stream )
 
EnumerationArray
< InteractionLocality, bool > 
haveWork = { { false } }
 True if there is work for the current domain in the respective locality. More...
 
bool bDoTime = false
 True if event-based timing is enabled.
 
GpuTimers * timers = nullptr
 CUDA event-based timers.
 
std::unique_ptr
< gmx_wallclock_gpu_nbnxm_t > 
timings
 Timing data. TODO: deprecate this and query timers for accumulated data instead.
 
EnumerationArray
< InteractionLocality,
FusedXToXqLaunchParams > 
xToXqLaunchParams
 Kernel launch parameters per interaction locality, computed once at pair-list setup.
 

Member Data Documentation

EnumerationArray<InteractionLocality, bool> gmx::NbnxmGpu::haveWork = { { false } }

True if there is work for the current domain in the respective locality.

This includes local/nonlocal GPU work, either bonded or nonbonded, scheduled to be executed in the current domain. As long as bonded work is not split up into local/nonlocal, if there is bonded GPU work, both flags will be true.


The documentation for this struct was generated from the following file: