Functions
std::vector< GpuTaskAssignment >	buildTaskAssignment (const GpuTasksOnRanks &gpuTasksOnRanksOfThisNode, ArrayRef< const int > gpuIds)
	Build the GPU task assignment for the ranks of this node. More...

bool	isAnyGpuSharedBetweenRanks (ArrayRef< const GpuTaskAssignment > gpuTaskAssignments)
	Return whether a GPU device is shared between any ranks. More...

size_t	countGpuTasksOnThisNode (const GpuTasksOnRanks &gpuTasksOnRanksOfThisNode)
	Counts all the GPU tasks on this node.

int	countOverAllRanks (MPI_Comm comm, int countOnThisRank)
	Return on each rank the total count over all ranks of all simulations.

void	barrierOverAllRanks (MPI_Comm comm)
	Barrier over all rank in `comm`.

Function Documentation

std::vector<GpuTaskAssignment> gmx::anonymous_namespace{taskassignment.cpp}::buildTaskAssignment	(	const GpuTasksOnRanks &	gpuTasksOnRanksOfThisNode,
		ArrayRef< const int >	gpuIds
	)

Build the GPU task assignment for the ranks of this node.

Parameters

[in]	gpuTasksOnRanksOfThisNode	For each rank on this node, the set of tasks that are eligible to run on GPUs.
[in]	gpuIds	The GPU IDs for the tasks on this node, supplied either by the user or the automatic assignment.

Returns: A vector with elements for each rank on this node that describes the GPU tasks and the assigned device ID.

Exceptions

InvalidInputError when the user GPU assignment requests multiple devices on a rank

bool gmx::anonymous_namespace{taskassignment.cpp}::isAnyGpuSharedBetweenRanks ( ArrayRef< const GpuTaskAssignment > gpuTaskAssignments )

Return whether a GPU device is shared between any ranks.

Sharing GPUs among multiple ranks is possible via either user or automated selection.

Functions