This paper presents a custom processor designed to execute a time consuming function in a CFD application. The selected function implements the method TDMA (Tri-Diagonal Matrix Algorithm) for solving a tri-diagonal system of equations. The custom processor was implemented in a commercial PCI prototyping board based on Virtex4LX FPGAs and uses a dedicated memory cache system, address generators and a deep pipelined floating-point datapath. Running at 100MHz and assuming the input data already in the cache memories, the system reaches a throughput greater than 1.4GFLOPS. © 2008 Springer-Verlag Berlin Heidelberg.
CITATION STYLE
Oliveira, F., Santos, C. S., Castro, F. A., & Alves, J. C. (2008). A custom processor for a TDMA solver in a CFD application. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 4943 LNCS, pp. 63–74). https://doi.org/10.1007/978-3-540-78610-8_9
Mendeley helps you to discover research relevant for your work.