A parallel processing approach to proximity correction

We present an approach to proximity effect computation in which the program is hosted on a parallel processor to accelerate the computation. Our requirements are: (i) factor of 100 speed improvement over a VAX‐11/780 (106 instructions/s), (ii) ability to handle dense patterns with 0.5‐μm design rules, (iii) ability to handle arbitrarily sized chips, and (iv) affordable processing power. We have concluded that parallel processing is a viable approach to realizing these goals. A parallel processing implementation will permit proximity correction of complex very large scale integrated (VLSI) patterns in a few hours instead of hundreds of hours. The proximity correction program is the functional equivalent of the one we have described previously. The program maintains pattern hierarchy insofar as possible, partitions features to permit edge placement control to 0.025 μm, and performs dose computations based on a multiple Gaussian scattering model. We have estimated that the best aggregate computation time per...