UBC Theses and Dissertations
Accelerator compiler for the VENICE vector processor Liu, Zhiduo
This thesis describes the compiler design for VENICE, a new soft vector processor (SVP). The compiler is a new back-end target for the Microsoft Accelerator, a high-level data parallel library in C/C++ and C\#. This allows automatic compilation from high-level programs into VENICE assembly code, thus avoiding the process of writing assembly code used by previous SVPs. Experimental results show the compiler can generate scalable parallel code with execution times that are comparable to human-optimized VENICE assembly code. On data-parallel applications, VENICE at 100MHz on an Altera DE3 platform runs at speeds comparable to one core of a 2.53GHz Intel Xeon E5540 processor, beating it in performance on four of six benchmarks by up to 3.2x. The compiler also delivers near-linear scaling performance on five of six benchmarks, which exceed scalability of the Multi-core target of Accelerator.
Item Citations and Data
Attribution-NonCommercial-NoDerivatives 4.0 International