Date: Wed, 14 Oct 1998 12:10:22 -0500
From: Eric Rudd <rudd AT cyberoptics DOT com>
Subject: Re: libc math function upgrade work
To: djgpp-workers AT delorie DOT com
Message-id: <3624DAFE.384CA7B9@cyberoptics.com>
Organization: CyberOptics
MIME-version: 1.0
X-Mailer: Mozilla 4.05 [en] (Win95; U)
Content-type: text/plain; charset=us-ascii
Content-transfer-encoding: 7bit
References: <Pine DOT SUN DOT 3 DOT 91 DOT 981014174655 DOT 10409C-100000 AT is>
Reply-To: djgpp-workers AT delorie DOT com

Eli Zaretskii wrote:

> On Wed, 14 Oct 1998, Eric Rudd wrote:
>
> > I observed about 10-20% variation on various compiles, probably due to code
> > alignment differences, but on the average the new routines are about three
> > times faster than libm.
>
> Was that with libm.a from DJGPP v2.01 or from the alpha release of
> v2.02?  Libm was replaced by a new and a very different version in the
> last alpha release of v2.02.  (I expect the new version be slower than
> the old, but you can never know until you test.)

I used the libm.a dated 98-07-26 from a recent 2.02 alpha release.  As I recall,
the old libm had some assembler routines that were about as fast as the existing
libc routines, but also had accuracy problems.

-Eric Rudd