Re: linux-2.1.60/i586,glibc-2.0.5c,egcs-971114 snapshot :)

[teunis <teunis at mauve dot computersupportcentre dot com>]

On Tue, 18 Nov 1997, Jeffrey A Law wrote:

>   In message <Pine.LNX.3.96.971117143148.2661A-100000@sigil.computersupportcentre.com>you write:
>   > binutils-latest + support for additional wanted platforms (eg. Win32 :)
>   > libraries?  I don't know where to obtain them.
>   > Same for simulator....  (why a simulator?)
> Well, for many you can use the latest newlib release.

???? newlib release ????

> You'd want a instruction set simulator so that you can run the "execution"
> tests -- particularly useful if you don't have a target board, or it has
> a terribly slow (less than ethernet) speed link.

To be quite honest:
	I don't have immediate access to some of the hardware.....
	... and there's no Win32 emulator available... that works.

So how do I build a cross-compiler?

>   > I also get random "exception 11"s - where I didn't used to with previous
>   > compilers.  [they're quite consistent actually - but I haven't yet tracked
>   > down what's causing them]
> You might want to look at the g77 info files for "Signal 11 and Friends".  This
> might be a hardware failure.

No - this is definitely egcs-971114
	(ref someone else's post about linux source causing this)

gcc-2.7.2.1 - works perfectly on all files
egcs-2.90.10p (pentium-gcc mods) - works perfectly... most of the time...
	(older egcs - some bugs present then)
egcs-2.90.16 - some problems, nothing serious.

my test code is ~3M of VR/math-related work.  VERY cpu-intensive (and
memory-intensive) to compile.  Makes an excellent test of optimizations
too - for instance: pgcc's output code is 30-50% faster than gcc-2.7.2.1's
output code with this project.

... egcs (plus other changes to the computer) have slowed it down - so
egcs is not yet as efficient for i586 as pentium-gcc.

G'day, eh? :)
	- Teunis