This is the mail archive of the
c++-embedded@sourceware.cygnus.com
mailing list .
Re: gcc for embedded: Worth it?
- To: Martin Boyer <gamin at videotron dot ca>
- Subject: Re: gcc for embedded: Worth it?
- From: Ken <shiva at compuserve dot com>
- Date: Fri, 26 Jun 1998 21:21:09 -0700
- CC: c++-embedded at cygnus dot com
- Newsgroups: comp.arch.embedded,comp.realtime,comp.sys.m68k
- References: <6l86kv$h9l@examiner.concentric.net> <357c8b45.9310007@news.mpx.com.au> <doenges.897372012@lpr.e-technik.tu-muenchen.de> <6mhdu7$fav$1@raven.inka.de> <jEOj1.3654$9Y1.32042@eagle> <3592d3ce.11558168@138.242.19.88> <6mvht3$orp$1@murmel.gams.at> !<x7emwcz757.fsf@carette.Foo.COM>
- Reply-To: shiva at well dot com
Martin Boyer wrote:
>
> "Martin Weiss" <martin.weiss@datentechnik.com> writes:
>
> > Who has ever tried to compile a crosscompiler under linux to run
> > under win32 and to target 68k? Experiences with c++ for embedded?
>
> I didn't have time to built the crosscompiler myself, so I had my
> company purchase GNUPro, from Cygnus Solutions. This is for an Intel
> i960 target on a Windows NT 4.0 host.
>
> After a few weeks of early problems, caused mainly by problems with
> segmentation faults from an older version of cygwin32, I can say this
> was a very good decision. The folks at Cygnus are very competent and
> diligent. The tools are good and easy to use (I'm a Unix old-timer).
>
> As for C++ for embedded, we had a protocol library that was using
> templates and the Standard Template Library. That never worked, most
> probably due to problems with constructors or structure
> initialization. I can't blame anyone; the protocol code was developed
> on Visual C++ 5.0 and used extensions that weren't in the C++ draft
> standard. To make a long story short, the parts that used templates
> were re-implemented without templates, and that worked. We have since
> added code which used templates in a limited way, and that worked as
> well. Go figure.
>
> Overall, I'd say that GNUPro can generate useful, embeddable, code
> from C++ sources, using inherited classes, constructors, templates (to
> some extent), etc. I would say that using C++ allowed for a very
> generic protocol library to be created (the same code runs in Visual
> C++ on the control station, and in GNUPro's C++ in the embedded
> system), at the cost of at least double the ROM and RAM requirements
> over a "low-tech" C implementation. That is a lot when you have 512k
> of each.
>
> Martin
What accounted for the doubling in size going from C to C++?
[I'm cc'ing this to the new c++-embedded list hosted at Cygnus. Send
subscription requests to mailto:c++-embedded-request@cygnus.com.]
--
Ken
mailto:shiva@well.com
mailto:shiva@CompuServe.COM
http://www.well.com/user/shiva/
http://sewnsurf.home.ml.org/
http://www.e-scrub.com/cgi-bin/wpoison/wpoison.cgi (Death to Spam!)