koopman@a.gp.cs.cmu.edu (Philip Koopman) (07/09/88)
I am in the unenviable position of having to determine the performance of an unconventional processor compared to other processors. In particular, the Harris RTX stack-machine processors compared to things like MIPSco processors, AMD29000, VAXen, SUNs, etc. (please don't take it personally if I left your particular processor out.) Since I well know that this is a generally problem that has no completely satisfactory solution, I propose to only look at performance running the C language on a variety of benchmarks: Dhrystone, Stanford integer, etc. What I want is not "native MIPS", which is pretty meaningless in this case, but rather "average MIPS" normalized to other processors. Can anyone out there tell me how (or where to find) the most common ways of computing "average MIPS"? I would think that an obvious approach would be to use a weighted average of several performance benchmark programs. For me, the point is not what a "MIPS" is, but rather getting a crude approximation of performance on at least some tasks compared to other machines that the outside world will understand. By the way, yes I know that this exercise is fraught with peril, but I did say that my task was unenviable. Thanks for any help, Phil Koopman koopman@faraday.ece.cmu.edu (sometime consultant to Harris Semiconductor) 5551 Beacon St. Pittsburgh, PA 15217