CompuServe Messages

#No 68060?

    02-Jan-95 01:56:17
Sb: #180083-#No 68060?
Fm: daniel wolf 70250,626
To: Dale Larson 76702,654
From what I saw of Amiga C compiler output a while back I'd have to doubt your optimism on 'fixing' C programs, but I've heard the SAS compiler has come a long way so I hope you're right. Profiling should be fairly obvious to many programs, but isn't. I'd suggest just examining loops with lots of computation in 'em! At least experimenting with some table lookups, etc. is worth a try and a few experiments might reveal big speedups without much work. I think many programmers 'quit' when the code works at all. Any program which does computation-intensive stuff like rendering is a worthy subject for profiling and optimization. It would save Amiga owners time and money spent on faster clock-speed accelerators. I do think much is to be gained by improving the programs. I know specific programs, the programmers of which have told me they'd 'like to get a look back at that loop….' or they 'wish they had done that loop in assembly' or some such thing. Sad thing is they never do. More features get added and those who need performance have to go buy accelerators, but the original program could easily have been 50% to 200% faster with a bit more work.