[9fans] new 8c

[9fans] new 8c

[Bruce Ellis]

I've improved the vlong support in 8c.

The missing ops are in, load/store optimizations,
and things like add, sub etc. aren't function calls.

A simple test (summing a large array of random
vlongs) shows a factor of 3 speed up on my
machine (braindead PII) so maybe I haven't been
just wasting time.

It's not ready for public consumption yet but I
have thrown a copy at presto and jmk.

I'll put it on my web-site in a few days.
Reading x86 assemb, let alone generating it is
starting to get to me.

brucee

Re: [9fans] new 8c

[Sape Mullender]

> I've improved the vlong support in 8c.

Brucee, you're my hero.

Re: [9fans] new 8c

[Bruce Ellis]

ahh, i forgot that edf.c was vlong bound.  you'll be
glad to know that i get a factor of 6 in a complex
expression - they tend to cluster around vlongs.

brucee
----- Original Message ----- 
From: "Sape Mullender" <sape@research.bell-labs.com>
To: <9fans@cse.psu.edu>
Sent: Wednesday, March 17, 2004 8:56 AM
Subject: Re: [9fans] new 8c


> > I've improved the vlong support in 8c.
> 
> Brucee, you're my hero.

Re: [9fans] new 8c

[Sape Mullender]

> ahh, i forgot that edf.c was vlong bound.  you'll be
> glad to know that i get a factor of 6 in a complex
> expression - they tend to cluster around vlongs.

I am, I am!

	Sape

Re: [9fans] new 8c

[boyd, rounin]

> Reading x86 assemb, let alone generating it is
> starting to get to me.

did you run out of ice?

[9fans] new 8c

[Bruce Ellis]

here is an 8c with native vlong support. comments welcome.

http://www.chunder.com/plan9/

brucee

[9fans] new 8c

[Bruce Ellis]

update to vlong 8c at ...

http://www.chunder.com/plan9/

thanks to jmk (and others) for testing.

brucee