MPIBlib: MPI Benchmark library
Detailed Description
Data structures describing collective communications to be measured, which are used as an argument of collective benchmark functions. Collective container allocators usually have an argument that points to the implementation of the collective communication operation, for example MPIB_Scatter_container_alloc has an argument MPIB_Scatter
, pointer to a scatter implementation. This provides three-level extension:
The same container can be used in different benchmarks. Container can call different algorithms of the same collective operation.
Function Documentation
void MPIB_coll_container_free | ( | MPIB_coll_container * | container | ) |
Frees collective container
MPIB_coll_container* MPIB_Scatter_container_alloc | ( | MPIB_Scatter | scatter | ) |
Allocates Scatter container
MPIB_coll_container* MPIB_Gather_container_alloc | ( | MPIB_Gather | gather | ) |
Allocates Gather container
MPIB_coll_container* MPIB_Bcast_container_alloc | ( | MPIB_Bcast | bcast | ) |
Allocates Bcast container
MPIB_coll_container* MPIB_Reduce_container_alloc | ( | MPIB_Reduce | reduce | ) |
Allocates Reduce container
MPIB_coll_container* MPIB_Comm_dup_free_container_alloc | ( | ) |
Allocates MPI_Comm_dup-MPI_Comm_free container
MPIB_coll_container* MPIB_Scatterv_container_alloc | ( | MPIB_Scatterv | scatterv, | |
const double * | factors | |||
) |
Allocates Scatterv container
MPIB_coll_container* MPIB_Gatherv_container_alloc | ( | MPIB_Gatherv | gatherv, | |
const double * | factors | |||
) |
Allocates Gatherv container
MPIB_coll_container* MPIB_Alltoall_container_alloc | ( | MPIB_Alltoall | alltoall | ) |
Allocates Gatherv container