25th Anniversary of the Personal Computer" 4214

Around 1984, one of my tasks was to select a small computer system to replace the over-burdened HP87XM calculator-pretending-to-be-a- computer being used extensively for Engineering design calculations. The program I had in mind had an estimated runtime of about a day; based on a simulation of the number-crunching and amount of data required.

There were a number of possible systems, from PDP-11 family, through 80x86 and 68k with operating systems including Digital's, Unix-Xenix and M-PM. I devised my benchmark equivalent to the simulation, in FORTRAN and C... it was about 60 lines of code mainly because I realized from the outset that I'd have to probably key the program myself on the target system and deal with compiler-grammar problems building the benchmark.

All the 80x86 systems proved to be about one tenth of the speed of the 68010-based system. This was a little puzzling until a friend pointed out that the array declarations defined 80 kilobytes and that that wasn't handled efficiently on the 80x86 architecture at the time. Reducing the array sizes for a data segment of less than 64kbytes produced a stunning performance improvement.

I settled for the 68010-based system because the data requirements for the real program(s) were likely to be greater than 80 kbytes. -- "Bernd Felsche - Innovative Reckoning, Perth, Western Australia ASCII ribbon campaign "Laws do not persuade just because X against HTML mail they threaten." and postings Lucius Annaeus Seneca, c. 4BC - 65AD.