From: Russell Wallace <manorsof AT iol DOT ie>
Newsgroups: comp.os.msdos.djgpp
Subject: Performance of -lemu
Date: Fri, 19 Jun 1998 21:46:27 +0100
Organization: Ireland On-Line
Lines: 22
Message-ID: <358ACE23.384B@iol.ie>
Reply-To: manorsof AT iol DOT ie
NNTP-Posting-Host: dialup-882.dublin.iol.ie
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
To: djgpp AT delorie DOT com
DJ-Gateway: from newsgroup comp.os.msdos.djgpp
Precedence: bulk

I'm working on a program which will make extensive use of floating
point.  I'd like to link with -lemu so it'll work as-is on 386 and 486SX
machines.

However, when I checked the performance of generated code on a simple
benchmark program (multiplying two 100 by 100 matrices, double
precision), I get the following figures on a Pentium-133:

-O3		21 megaflops
-O3 -lemu	13 megaflops

My understanding had been that -lemu should make no difference at all if
the program does in fact end up running on a machine with FP hardware. 
Can anyone explain the reason for this discrepancy, and whether there's
anything that can be done about it?  (Short of forgetting about working
as-is and requiring users of 386s to run one of the .exe emulators,
which is what I'm planning to do as things currently stand.)

-- 
"To summarize the summary of the summary: people are a problem."
Russell Wallace
manorsof AT iol DOT ie