To start, look at the memory maps produced by the compiler,
then some of the compiled assembly files and try to identify the
biggest culprits and those that would return the biggest savings
if optimized. I'm not familiar with C535 target but pointer usage
is often an issue with small devices/architectures so, as a guess,
you might want to look at those.
I know the C535 quite well and I have seen that the sdcc wants to put local vars onto the internal iram.
But there I have quite a few timing critical variables so that flows over. Also the memory models back and forth have allways some shortcommings when it comes to locals.
Often the biggest gains come from a good strategical moves, not
from smaller tactical tweakings, but this is not alway feasible.
True, but before I start to really change my code for that kind of big change I wanted to try to get some opinions if it is worth the work. Of course all that happens in a seperate branch (using svn) ;-)
I work on a Mac and want to get rid of PC software, Keil is the last I
Ah, fellow Mac users, if I can only get rid of PICKit flash burning utility
I'd be PC free...
Yes, I love my Mac setup, just downloaded the latest svn version to get the merge capabillities ... but now svnX is broken ... which client do you use (in case you use svn at all) ?
Out of curiosity, do you use Netbeans as an IDE for C/sdcc?
My poison is Eclipse/CDT.
I use netbeans as C and Java IDE. Since my gadet also comes with a java editor that setup works quite well.
You might want to look at some stuff on my site athttp://www.sparetimelabs.com
See "printf revisited" for some write on experience of sdcc versus
one commercial compiler. And "fun with hc08" for setting up
Eclipse to use sdcc (or vice versa).
just my 2 snt worth, Kusti
Will do soon ;-)