EXPLOITING DIRECT ACCESS SHARED MEMORY FOR MPI ON MULTI-CORE PROCESSORS

In 2008, the Catamount lightweight kernel was extended to support direct access shared memory between processes running on the same compute node. This extension, called SMARTMAP, allows each process read/write access to another process' memory by extending the virtual address of each process. Simple virtual address bit manipulation can be used to access the same virtual address in a different process' address space. In this paper we describe an implementation of the Message Passing Interface (MPI) that uses SMARTMAP for intra-node message passing. SMARTMAP has several advantages over POSIX shared memory techniques for implementing MPI. We present performance results comparing MPI using SMARTMAP to the existing MPI transport layer on a quad-core Cray XT platform.