[CCC DEV] REAL32 next steps

[Matt Jadud]

Hi Adam, all,

I spent yesterday poking around the compiler sources, and it is not
yet obvious to me where the REAL32 error would live. This is the
second time in a long time that I've dug around the sources, though,
so it isn't surprising that a day spent spellunking didn't yield
immediate results.

A simple short-term solution would be to wrap up a few calls out to
math.h (I probably only need a dozen functions) in "math.module" or
similar. The UAV team would write code that looked like:

REAL32 a, b, result:
SEQ
  a := 3.0
  b := 5.0
  math.pow(a, b, result)

This gets them moving, and buys time to:

1. Implement calls out from FORALL to avr-libc, or
2. Find the back-end bug in the compiler w.r.t. 16-bit REAL32 generation.

Updating their code to then use FORALL shouldn't be an issue.

Does that sound like a reasonable way forward? I'm not trying to be
impatient -- exactly the opposite, actually. I figure this is a "good
enough" solution that lets us roll along while we look for a "better"
solution.

Thoughts?

Cheers,
Matt