We present a fine grained, massively parallel SIMD architecture called the data structure accelerator and demonstrate its use in a number of problems in computational geometry. This architecture is extremely dense and highly scalable. Systems of 10$^6$ processing elements can be feasibly embedded in work stations. We advocate that this architecture be used in tandem with conventional single sequence machines and with small scale, shared memory multiprocessors. We present a language for programming such heterogeneous systems that smoothly encorporates the SIMD instructions of the data structure accelerator with conventional single sequence code. [footnote: This is an expanded version of a paper that was presented at the Jerusalem Conference on Information Technology, October 1990.]