Re: End of optimization...

The free and most of the open source compilers for small micro
processor tend to be based on old compiler technology that was often
originally targeted to workstation processors with optimizations not
well suited to embedded small micros. Small micro applications are
bi-modal either small one or few of projects or volume applications
with wide distribution. In the later case compiler optimizations can
significantly help reduce part costs, power consumption and help with
FCC and other regulatory approvals.


The embedded system industry is overshadowed by the performance
perceptions of the public domain and low cost silicon company
development tools. Good processor and application specific compiler
technology for this part of the industry has a large impact on the
ability to use a specific processor in an application.


w..


Barak Zalstein wrote:


> > Open source is seductive, for retargeting processors of similar
> > architecture and applications it works quite well. For retargeting to
> > other application area's and architectures it saves little and costs a
> > lot, always competing with the earlier software design constraints. As
> > the previous comment points out the software architecture may prevent
> > newer optimizations and compiler implementations.
> > Walter Banks
> >
> Usually these projects infrastructure can be stretched into application specific domains with low enough cost and
> high enough quality/performance.
> Nowadays, when free or open (or whatever they call it now) projects supply such a large arsenal of optimizations,
> it would probably make sense to write a compiler from scratch only if your architecture doesn't involve something like instruction set, registers, pipeline, and units.
>
> Barak.

Return to the comp.compilers page.
Search the comp.compilers archives again.