Sat Jan 17 17:11:59 EST 2015
I've been using the STM32F103C8 for work and I am very pleased with
the chip and the workflow. I am considering (again) to move the
(commercial) synth development to C on ARM. For the simple reason
that I can't do both: develop a reasonably functional experimental
idiosyncratic Forth system and at the same time get a real synth out
So let's look at that again. What are the real reasons?
- Math efficiency
- Not distracted by premature optimization
- Faster turnaround. C really is easier to use for me.
It pains me. The efficiency is a platform thing I can't do anything
about. The latter two are realities of my workflow that I don't like
I like C because I've found a way to work with it that doesn't suck so
much. Names are important for organizing code: C structs, functions
and (a limited number of) static variables per file work well enough
for embedded work. Take this together with GDB being able to call
functions and print data structures, and that all hardware is
optimized for running C well, this is a lost cause.
So where is the purpose for Staapl? I find I need a little bit of
meaning, even if it is a delusion that's only temporary. It has to be
in heavily factored, slow running compact code.
So embrace the slow?
And focus on protocol-oriented programming? ( An old idea of
designing protocols such that they are easy to parse in a single pass
from left to right, avoiding the need for serialized data buffers. ).