atom feed18 messages in org.freebsd.freebsd-currentRe: Pentium optimizations
FromSent OnAttachments
AlexDec 16, 1997 6:59 pm 
Tim LiddelowDec 16, 1997 8:05 pm 
John S. DysonDec 16, 1997 8:37 pm 
AlexDec 16, 1997 9:17 pm 
Tim LiddelowDec 16, 1997 9:36 pm 
Scott MichelDec 16, 1997 10:02 pm 
John S. DysonDec 16, 1997 10:23 pm 
Brian HandyDec 16, 1997 10:47 pm 
John S. DysonDec 16, 1997 11:04 pm 
Warner LoshDec 16, 1997 11:49 pm 
John S. DysonDec 17, 1997 12:04 am 
Poul-Henning KampDec 17, 1997 2:55 am 
Warner LoshDec 17, 1997 7:09 am 
Russell L. CarterDec 17, 1997 7:42 am 
Eivind EklundDec 17, 1997 10:13 am 
Tim LiddelowDec 17, 1997 2:26 pm 
Doug RabsonDec 18, 1997 12:35 pm 
John PolstraDec 21, 1997 1:35 pm 
Subject:Re: Pentium optimizations
From:Tim Liddelow (TLid@cybec.com.au)
Date:Dec 16, 1997 8:05:18 pm
List:org.freebsd.freebsd-current

Alex wrote:

I've been lurking on the egcs list, and a few Linux/x86ites have been tossing around various Pentium or K6 enhanced functions (strlen and memcpy come to mind). I was wondering what the chance of this actually being merged into the -current cc? I was hopin we could get some Pentium optimizations without having to use the (as of now) fairly incompatable egcs. Or maybe get some of the dynamic library gunk merged into egcs.

I'd love to see egcs as a package for both -current and also -stable... anyone interested in doing it? I would do it if I had the time... (I know, you've heard that before). I'm not really familiar with the grokery/hackery that has been involved in merging gcc into the FreeBSD tree anyway. When gcc changes, how are these changes munged into FreeBSD's gcc ? (Not that gcc has changed much over the last eon!). I wonder if anyone has ever thought about "unbundling" cc(1) like some of the commercial unixen do...and just making it a package...then you could select the cc you wanted from sysinstall... for example, developers may select egcs, standard users may select gcc, other users may select pgcc, others a simple C compiler. Some users won't ever use C++, so why should they get the extra bloat of g++ ?

Of course, this "unbundling" isn't really unbundling, because you can simply pick the compiler you want. It also means 3rd party vendors may be more inclined to provide a compiler one day.

Just some thoughts.

Tim.

& waits for flames on unbundling &