A system for evaluating performance and cost of SIMD array designs

SIMD arrays are likely to become increasingly important as coprocessors in domain specific systems as architects continue to leverage RAM technology i n their design. The problem this work addresses is the efficient evaluation of SIMD arrays with respect to complex applications while accounting for o perating frequency and chip area. The underlying issues include the size of the architecture space. the lack of portability of the test programs, and the inherent complexity of simulating up to hundreds of thousands of proces sing elements. The overall method we use is to combine architecture level a nd Electronic Design Automation (EDA) level modeling by using an EDA-based tool to calibrate architectural simulations. The resulting system retains m uch of the high throughput of the architecture level simulator but it also has accuracy similar to that of an early pass EDA synthesis and circuit sim ulation. The particular problem of computational cost of the architectural level simulation is addressed with a novel approach to trace-based simulati on (we call it trace compilation), which we find to be one to two orders of magnitude faster than instruction level simulation while still retaining m uch of the accuracy of the model. Furthermore, traces must be generated for only a small fraction of the possible parameter combinations. Using trace compilation also addresses program portability by allowing the user to code in a single data parallel language with a single compiler, regardless of t he target architecture. We have used our system to evaluate thousands of po tential SIMD array designs with respect to real applications and present so me Sample results. (C) 2000 Academic Press.