? Optimization ?

[future]

While doing some changes to my code, I noticed this:

[Neutone]

When you are making a copy of an array of bytes, the optimizer should switch to from using indirect addressing to direct memory addressing when the number of bytes copied is greater than 11. That is true for the example you posted at least. That will optimize for code size.

[future]

But if 5 MOVFF's were used, the code would be faster/smaller, dont you think?

[bluetooth]

Future:

Grab any decent text on compiler design - you'll discover there's a HUGE distance between code optimized in an application specific environment by a professional assembly programmer and code from a generic set of statements fed to an optimizer that (obviously) must work after optimization - no matter how complex the (unknown to the compiler author) statements are!

CCS does an amazingly good job for a multi-hundred $$ compiler - we've used Greenhills in the past for 68xxx and it was better - but we paid a WHOLE lot more for that environment.

If you want "optimal" optimization, write the code in assembler. But remember, it'll probably take longer to write and debug, there's always one more byte that can be squeezed out by some wiseguy

The real question to be asked is: does the code work under the conditions required? And did using the compiler get you there quicker than writing it in assembler? If both answers are yes, then mission accomplished.

Just my $0.02....

[ckielstra]