A control unit module for a scalable floating-point-unit architecture

Abstract

The gap of execution time between software and hardware computing is significant and becomes more and more important when precision is required as it is the case for the floating point calculation. This paper presents the addition of a Floating Point Unit (FPU) module to an open-source processor called SecretBlaze. Besides the description of the chosen processor enhanced by FPU thanks to user instructions, this work focuses on the hardware method to add the scalable FPU system. As a first step, design implementations enable to compare chosen FPU with other FPU available in the open-source community. It also enables to evaluate the performance of the FPU added to the processor. As a second step, we present an FPU control unit added to the architecture in order to realize several functions with few resources. All proposed architectures have been implemented and tested on FPGA target.