Polymorphous computer architectures (PCA) are new computer architectures being developed under a DARPA/IPTO program to support mission agility for future high performance DoD embedded applications. These new architectures will have the ability to "morph" into different modes of execution with the goal of delivering uniform, high performance across a large variety of different processing types and workload compositions. Examples of these architectures include the MIT RAW machine, the Stanford Smart ...
This presentation focuses on understanding when a stream algorithm exists for a given kernel. We do so by considering the directed acyclic graph (DAG) for a particular implementation of the kernel. Nodes in the DAG represent inputs, outputs, or intermediate products of the algorithm, and edges from node A to node B in the DAG show that A is used to compute B. We can characterize the DAG for an ...
In this paper we describe and benchmark an implementation of the matrix exponential function. The implementation is based on the standard technique of scaling and squaring from the literature. The major kernels in this technique are matrix multiplication and Gaussian elimination. In the matrix multiply kernel, the implementation makes use of SIMD vector extensions present on the PowerPC G4 (Altivec) and the Intel Xeon (SSE-2). Although the use of the ...
Reports by Author
