Re: Compiling fast on a VAX?
Lord Isildur <mrfusion <at> vaxpower.org>
2007-09-24 19:35:11 GMT
yummy, i'm going to have to play with this!
(been a long time since i've poked myhead above the surface here,
greetings to everyone! good to see familiar faces (well, names) about!)
so i think about 5 years ago we had a thread about pcc versus gcc. pcc's
optimization is more or less nonexistent, so it produces slower code,
but on the other hand, pcc's libc is much simpler (and somewhat less safe
or correct, mind you) and so it can work out either way. We were comparing
old BSD and NetBSD systems. Stuff compiled against the old libc from
CSRG-era bsd and pcc was noticeably faster if you did a lot of things like
printf, but slower if you were pushing structs around in core and doing
thinsg that a good optimizer can get a lot smarter at (especially with
plenty of registers).
The big win was in the speed of compilation. I'm not sure how much of
the slowness of gcc is in the useful parts of optimization versus what
parts are wasting our time.
I think at the time, we had also discussed the possibility of rolling back
to an older version of gcc (the 2.low numbers)- the VAX optimizer is
unlikely to have changed much in the past 15 years, so that part of gcc is
probably just about the same.. and most of the fancier stuff is
inapplicable to vax anyway.
pcc, however, is amazingly fast for compiling..
I can make a 4000/600 with 192 megs of core available for testing.
(thorium.vaxpower.org) . it's already been used for some gcc/vax
development some years back. currently running netbsd 2.0 but i could put
something newer on (from across the continent, i have a serial console!)
it's still living on CMU's network.