Acceleration Technology, The Gpu-based Eda

EDA technology plays a critical part in modern IC industry. Along with the increasing of complexity of IC chips, the demanding of EDA tools is getting higher. The speed of solving N class linear equations is the critical factor of circuit simulation and synthesis.Meanwhile, GPGPU has become a novel field in massive parallelism computation. With introducing shader model 3.0 and shader model 4.0 as well as CUDA and OpenCL platform technology, can GPU not only nearly 10 times faster than CPU does in the same Paradigm, the developing and evolution speed are much more faster than CPU.Taking the advantage of the parallelism computing power, the core algorithm of EDA technology is re-architecting and optimizing to fit the GPU architecture. The Mathematic basis of EDA system is linear algebra. And matrix calculation is the basic algorithm in linear algebra. The acceleration of matrix computation in GPU architecture is greatly control the acceleration of EDA system. My paper proposal an optimal method to product sparse matrix in EDA system, and design a GPU based quadratic placement algorithm. The speedup is nearly 10 times on average. The paper introduced a Monte-Carlo based statistical static timing analysis, and applies it to GPU architecture, which approved to be data parallelism, and gain nearly 90 times speedup. Also the paper introduced breadth-first search based algorithm to calculate critical path in graph algorithm, applying matrix fast production, and get nearly 12 times speedup on average. The advantages and disadvantages of this proposal are fully discussed and analyzed. Paper also introduced a Monte-calor based static timing analysis accelerating technique. Te algorithm is accelerated by nearly 90 times than conventional algorithm.