A Custom Processor for a TDMA Solver in a CFD Application
International Workshop on Applied Reconfigurable Computing, 2008•Springer
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 …
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 …
Abstract
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.
Springer
以上显示的是最相近的搜索结果。 查看全部搜索结果