Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!utcs!mnetor!seismo!gatech!cuae2!ihnp4!inuxc!pur-ee!uiucdcs!okstate.UUCP!uokvax.UUCP!emjej
From: emjej@uokvax.UUCP.UUCP
Newsgroups: net.micro.6809
Subject: clib.l, and Dhrystone
Message-ID: <3500160@uokvax.UUCP>
Date: Wed, 16-Jul-86 16:48:00 EDT
Article-I.D.: uokvax.3500160
Posted: Wed Jul 16 16:48:00 1986
Date-Received: Sat, 19-Jul-86 02:04:58 EDT
Lines: 21
Nf-ID: #N:uokvax.UUCP:3500160:000:1010
Nf-From: uokvax.UUCP!emjej    Jul 16 15:48:00 1986


Those folks running OS-9/6809 and hacking C, I have a recommendation:
grab Carl Kreider's hacks of clib.l.  Mr. Kreider has rewritten a fair
number of library routines in assembumbler, to respectable advantage in
speed and size.

To give you an example--you'll recall my previous posting about running
dhry.c on a 2MHz 6809, in which I said that with register variables and
with global scalars forced to the direct page it ran something like
247 Dhrystones/second?  Well, I recompiled and ran with the revised clib.l,
and under the same conditions (register and direct page) I got 278
Dhrystones/second.  (Turning off stack overflow checking got it up to
303 Dhrystones/second.)

I think the major reason for the speedup on the Dhrystone benchmark is
that Microware C doesn't do structure assignment inline--you have to call
a function, and Mr. Kreider did hack _strass().  If the compiler supported
structure assignment, the 6809 would make an even better showing.  (eh, eh,
nudge, nudge?)

						James Jones