[geeks] galeon Doesn't Suck that much....

[Greg A. Woods]

[ On Tuesday, April 16, 2002 at 08:51:56 (+0100), David Cantrell wrote: ]
> Subject: Re: [geeks] galeon Doesn't Suck that much....
>
> On Mon, Apr 15, 2002 at 08:27:14PM -0400, Greg A. Woods wrote:
> > [ On Monday, April 15, 2002 at 23:07:42 (+0200), Jochen Kunz wrote: ]
> > > GCC is knowen to make bad, or at least suboptimal, code on all RISC
> > > platforms. Installing it to make my live easier is no choice, as I am no
> > > whimp. I _want_ to die hard. ;-)
> > All of the nearly 50-odd supported NetBSD platforms, running on 10-16
> > different CPU families to date (depending on how you count them), use
> > only GCC, and at least four of those are strictly RISC.
> 
> I fail to see any contradiction here.

Jochen suggested gcc may generate bad (and "at least suboptimal") code
on "all" RISC platforms.  NetBSD runs fine on RISC platforms (definitely
better than "bad", and at least optimally enough to be chosen by very
careful engineers (eg. NASA) in place of supported vendor sytems even
when said vendor systems claim to have better compilers).  NetBSD is
always compiled with GCC on all platforms (and usually even most
compilers are compiled with the host GCC too, though LCC is now
available for NetBSD and supposedly can target some of the RISC
platforms NetBSD runs on, including Alpha and MIPS, though I've only
tried it on i386 and I don't know if it generates more optimal code than
GCC or not (I didn't test it for performance).

> > I haven't heard too many people complain about NetBSD's performance.
> 
> No, but then, people run NetBSD because it's often the only choice other
> than $vendor_OS, so they are usually choosing it for reasons other than
> performance.

Some of the people (eg. NASA) "choosing" NetBSD on RISC platforms are
not only using it in place of vendor operating sytems (and sometimes
even compilers) that they have partly already paid for anyway.  They are
(or at least were) also spending a great deal on contributing core
development efforts to NetBSD (and thus indirectly to OpenBSD :-).  You
don't do that to get a final product that has "bad code"!  :-)

-- 
								Greg A. Woods

+1 416 218-0098;  <gwoods at acm.org>;  <g.a.woods at ieee.org>;  <woods at robohack.ca>
Planix, Inc. <woods at planix.com>; VE3TCP; Secrets of the Weird <woods at weird.com>