Re: AU - CW code gen problems
Re: AU - CW code gen problems
- Subject: Re: AU - CW code gen problems
- From: Kurt Bigler <email@hidden>
- Date: Wed, 26 Mar 2003 16:02:11 -0800
on 3/26/03 12:33 PM, Marc Poirier <email@hidden> wrote:
>
Hi. I'm sorry for this frantic semi-offtopic message, but I'm (a) late to
>
go meet someone and (b) freaking out because I'm encountering ridiculous
>
code gen problems with my project for my job that I'm already way behind
>
schedule on. Consider this a desparate plea for help!
>
>
Has anyone experienced bad Mach-O code gen problems with CW? I've been
>
doing most of my development with PB/GCC, but with this project, CW builds
>
DSP code that performs about 30% better, and we *really need* that 30%, so
>
we have to use CW. But it's building a GUI component with horrible
>
crashing problems! They are completely erratic and weird. The crash logs
>
show totally different traces every time. It's hardly ever even in my
>
code. Making changes to my code is completely irregular with respect to
>
stopping crashes (sorry, I can't come up with a coherent way to say that
>
right now, but my main point is that I'm highly convinced that CW is at
>
fault). I just went back and rebuilt with GCC and all is well again. But
>
the DSP runs too slow...
I was sure I had some codegen problems last week, but it turned out to be
just a bug that was made hard to see by problems with the CodeWarrior Mach-O
debugger. Its a long story, but a series of problems with the debugger
ended up causing a cascade of incorrect conclusions. I'll spare you the
details, but CW Mach-O debugging is not a pretty thing, even with all
optimization turned off. Using mixed-listing display format combined with
watching the registers can make it work.
Are you sure it isn't a race condition? If 30% slower makes your feature
set break, then maybe you have never fully exercised it and the CW situation
is more representative.
In any case you will probably have to do something to simplify in order to
find this out. Changing optimization levels on a per-source basis might be
helpful. I have code in my config.h file which sets all the optimization
pragmas according to a #define OPTIMIZE flag which I set (or not) in each
source file before including config.h. This also avoids having to recompile
the world (or maintain separate targets) in order to change the optimization
level.
-Kurt Bigler
>
Anyway, I'm thinking that maybe this has something to do with how CW links
>
frameworks? Has anyone tried the different Mach-O linkers in CW and know
>
if one is better than others? Has anyone had to tool around with the
>
libraries that are linked against? I'm just working off of Emagic's
>
example CW projects. Anyway, I'm quite desparate, if anyone has any
>
wisdom to share with Mach-O building and/or linking of AUs in CW, please
>
let me know. If you solve the problem, you get a free license of SFX
>
Machine RT!
>
>
Thanks,
>
Marc
>
_______________________________________________
>
coreaudio-api mailing list | email@hidden
>
Help/Unsubscribe/Archives:
>
http://www.lists.apple.com/mailman/listinfo/coreaudio-api
>
Do not post admin requests to the list. They will be ignored.
_______________________________________________
coreaudio-api mailing list | email@hidden
Help/Unsubscribe/Archives:
http://www.lists.apple.com/mailman/listinfo/coreaudio-api
Do not post admin requests to the list. They will be ignored.