Re: [ECOS] c++ ecos performance

[Mike Moran <mnmoran at bellsouth dot net>]

Grant Edwards wrote:
> I've compared timing info for eCos, ThreadX, Nucleus, and one
> other I don't remember.  All but eCos were written in C.  eCos
> times for context switching and whatnot were all pretty much
> the same as the others.

Thanks for this validation of what embedded C++ programmers already
understand. However, I must say that the chosen language does not
affect the performance of the implementation as much as the way
in which the language is used. E.g. I can write a poorly performing
kernel in any suitable language. IMHO, it is the libraries intended
for application programming, especially those that allocate memory
behind the back of the programmer for "simplicity", that are most
unsuitable for embedded systems programming.

> [I don't like C++ much as a language (in the abstract) as I
> think it's too complex and too unsafe.  But, I must say that
> the guys who wrote eCos used it well.]

Yikes! "Too complex"? "too unsafe"? Excuse me for reacting to this
flame bait, but this is pure bunk, and I get tired of this unjustified
attitude. C++ provides many features as a means to solve various types
of problems encountered in software engineering. Best of all, if you
don't want the feature, you don't have to use it.


Sorry for the knee jerk,
mike

-- 
Michael N. Moran
5009 Old Field Ct.
Kennesaw, GA 30144
(h) 770 516 7918
(c) 678 521 5460

The Beatles were wrong: 1 & 1 & 1 is 1

-- 
Before posting, please read the FAQ: http://sources.redhat.com/fom/ecos
and search the list archive: http://sources.redhat.com/ml/ecos-discuss