Field-Programmable Custom Computing Machines, Annual IEEE Symposium on (2009)
Apr. 5, 2009 to Apr. 7, 2009
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/FCCM.2009.14
Single-FPGA spatial implementations can provide an order of magnitude speedup over sequential microprocessor implementations for data-parallel, floating-point computation in SPICE model-evaluation. Model-evaluation is a key component of the SPICE circuit simulator and it is characterized by large irregular floating-point compute graphs. We show how to exploit the parallelism available in these graphs on single-FPGA designs with a low-overhead VLIW-scheduled architecture. Our architecture uses spatial floating-point operators coupled to local high-bandwidth memories and interconnected by a time-shared network. We retime operation inputs in the model-evaluation to allow independent scheduling of computation and communication. With this approach, we demonstrate speedups of 2–18×over a dual-core 3GHz Intel Xeon 5160 when using a XilinxVirtex 5 LX330T for a variety of SPICE device models.
Spice, Analog Circuit Simulator, Spatial Computation, VLIW Scheduling, Loop Unrolling, Floating-Point
André DeHon, Nachiket Kapre, "Accelerating SPICE Model-Evaluation using FPGAs", Field-Programmable Custom Computing Machines, Annual IEEE Symposium on, vol. 00, no. , pp. 37-44, 2009, doi:10.1109/FCCM.2009.14