@MISC{Virginia_comparativestudy, author = {Arcilio Jaime-raul Virginia}, title = {Comparative Study of VHDL Generators}, year = {} }
Share
OpenURL
Abstract
The Molen Polymorphic Instruction-Set Computer, is a reconfigurable processor consisting of a GPP which provides the software programmability and the FPGA that provides hardware flexibility. Creating applications for such a reconfigurable processor requires both hardware and software knowledge. The Delft Workbench hardware/software co-design environment aims at eliminating the necessity for hardware knowledge through automated hardware generation from ANSI-C. Generated VHDL designs however, do not perform as well as manual designs. To improve the performance of generated VHDL designs C2VHDL compilers apply different strategies involving specialized intermediate representations and optimizing transformations. As part of the Delft Workbench research group, the aim of this project was to evaluate DWARV, ROCCC and Spark C2VHDL compilers through qualitative and quantitative measures to ascertain what compilation strategy leads to the best performing VHDL. To compare these compilers a literature study was performed and empirical data was gathered from the synthesis results of 33 VHDL designs generated by DWARV and Spark. From these results the throughput/slice was calculated and revealed that Spark out performs DWARV, which does not perform optimizing transformations.