Adaptive Multiclient Network-on-Chip Memory Core: Hardware Architecture, Software Abstraction Layer, and Application Exploration

Table 1

MPI functitons used for accessing the adaptive memory core.

int MPI_Alloc_mem (MPI_Aint size, MPI_Info info, void *baseptr)
MPI_Aint size	Size of memory to be allocated in byte
MPI_Info info	Usage depends on implementation; only the value 0 must be supported
void* baseptr	Pointer to the allocated memory
int MPI_Put (void *origin_addr, int origin_count, MPI_Datatype origin_datatype, int target_rank,
MPI Aint target_disp, int target_count, MPI Datatype_target datatype, MPI_Win win)
origin_*	Defines which data shall be sent to the target node
target_*	Defines where the data are to be sent: the node (by rank), in which local memory address (target_disp) and the amount of data
win	Defines the memory window on target
int MPI_Get (void *origin_addr, int origin_count, MPI_Datatype origin_datatype, int target_rank,
MPI_Aint target_disp, int target_count, MPI_Datatype target_datatype, MPI_Win win)
origin_*	Defines where the data which are read from target shall be stored
target_*	Defines which data shall be read: from which node (by rank), from which local memory address (target_disp), and the amount of data
win	Defines the memory window on target
int MPI_Free_mem(void *base)
void *base	Pointer to the memory that was allocated with MPI_Alloc_mem