20 Oct
2015
20 Oct
'15
6:50 p.m.
At 06:47 AM 10/20/2015, Noel Chiappa wrote:
Yup. Lots of work. That's kind of why I find the implication that I
don't know what's going on here somewhat disconcerting. :)
From: Rick
Murphy
Lots and lots of fun like that. What an space
optimizer would do if
there was one. :)
Wow. You did all that by hand? If so, with a program that big, wouldn't it have
made sense to try and
write
a separate post-processor to do all those optimizations? Surely it
wouldn't
have been that much work, compared to going through the entire compiler
output!
As with most things, it initially looked to be more work to write
something to parse the code and optimize it. Some of the work was
pretty much rote, but a lot of it took some judgement to find more
compact code. It wasn't until I described the effort that I actually
considered calling it "space optimization". Oh, and some of this was
painfully impacted by the limits of the compiler symbol table sizes,
thus a lot of constants don't get meaningful names.
Ugh. I wouldn't have volunteered for this. I still want to get it to
the point where 28K is possible, but that'll take a new optimization
that I haven't yet thought of. :) Basically, at this point, it's really
a FPP assembler version of Adventure slightly assisted by the Fortran
compiler. I'm pretty proud of the idle loop blinkenlights, though.
Looks good on my PiDP-8.
-Rick