Module mpi_cuda 

Explicit host-staged MPI plus rank-local CUDA composition adapter.

Structs

AcceptedMpiCudaLinearExecution

Complete host result accepted by MPI and its inseparable rank-local CUDA composition evidence.

MpiCudaLinearExecutionEvidence

Rank-local CUDA observations paired with one common accepted MPI result.

Constants

MPI_CUDA_ADAPTER_VERSION

Exact Eqiora composition-adapter package version compiled into this binary.

MPI_CUDA_LOCAL_ACTION_EXECUTION

Exact rank-local CUDA action selected by the host-staged composition.

MPI_CUDA_LOCAL_ACTION_EXECUTION_PROVIDER

Exact declared partition-local CUDA action release used by the MPI composition.

Traits

MpiCudaAdmittedExecutionAdapter

L3 seam consuming one exact distributed-CUDA L2 admission.