Dr. Gustafson describes this paper on the "Benchmark 6-Pack" as follows:
"DARPA motivated this benchmark suite development while we were working on the T Series at FPS. No one had any idea how to benchmark parallel computers back then. Barton and Bailey were working concurrently on the NAS Parallel Benchmarks, and arrived at many of the same ideas presented in this paper.
"We tried to find the theoretical "best" serial version of each task so we could define the operation count... but we knew we were on shaky ground with respect to linear equation solving (Strassen methods), N-body problems (clustering methods) and FFT (mixed radix methods). The effort to fix this chronic problem in benchmark definitions is what led to SLALOM and HINT many years later."