Re: [avr-gcc-list] OT: Re: clarification on gcc

[Bob Paddock]

> If you're designing a commercial product, then it's long past time to
> add more RAM, if a bit of extra optimisation seems tempting. Putting a
> product out in the market, only to be unable to add features to the
> platform, to keep pace with competitors, is a seriously suboptimal
> strategy in many cases.

Problem is that Marketing never gets tired of adding new features, and
won't remove old ones.
To them a software change is 'free'.  Something that upper management
should address, and doesn't.

Redesigning the hardware is complicated by regulatory approvals that
can take years to get.
Even just putting in a part with more memory that is otherwise the
same, opens up the can of paperwork approvals.

Also in some markets the twenty cents or so you might spend on a part
with bigger RAM might have you lose out to those competitors that you
mention, if you are making a high volume product.  In some products
the profit margin is simply to low to put in a better part.

I have one Tiny2313 project that I use IAR for, because no amount of
code refactoring and GCC optimizations fits the code produced by any
version (3.3.x or later) of GCC in to the Tiny2313 for that project.