Re: HINT: JOHN GUSTAFSON'S LATEST MEASURE OF PERFORMANCE

[Penio Penev <penev@xxxxxxx>]

On Thu, 17 Nov 1994, Stephen Pelc wrote:

> To run these benchmarks, surely a C compiler is needed for real 
> comparison?

These are benchmarks for what can be achieved on a computer system and 
are irrelevenat to the compiler. There is a discussion about that in the 
paper. The statement was, that when looking for further oprimisation 
after the compiler, not much could be done. Not that it shouldn't be 
done.

The discussion was, that the benchmark tries to measure the overall progress
towards the quality of the answer, which includes the processor, caches,
memory subsystem, the disk subsystem, the OS, the compiler, and all other
relevant factors. 

The fact, that humans can be measured also clearly shows, that a C 
compiler is not needed :-)

--
Penio Penev <Penev@venezia.Rockefeller.edu> 1-212-327-7423