An integrated floating point vector processor for DSP and scientific computing
D.W. Spaderna, P. Green, Kenway W. Tam, T. Datta, Manoj Kumar · 2003
The architecture of a high-performance floating-point microprocessor intended for computation-intensive applications is presented. Utilizing a massively parallel execution unit, very high bandwidth onchip memory, and extensive parallel address calculation hardware, the processor is expected to achieve peak execution rates of 400 MFLOPS for single-precision floating-point numbers. This type of performance is competitive with that of popular mainframe vector processors.>