Date: Tue, 20 Jul 1999 09:09:06 +0200
To: pgcc AT delorie DOT com
Cc: derrick AT cco DOT caltech DOT edu
Subject: Re: Do underflows cause a slowdown? Can I fix it?
Message-ID: <19990720090906.E263@cerebro.laendle>
Mail-Followup-To: pgcc AT delorie DOT com, derrick AT cco DOT caltech DOT edu
References: <378F9D37 DOT EC6D3B6B AT cco DOT caltech DOT edu>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
In-Reply-To: <378F9D37.EC6D3B6B@cco.caltech.edu>; from Derrick Bass on Fri, Jul 16, 1999 at 01:59:35PM -0700
X-Operating-System: Linux version 2.2.10 (root AT cerebro) (gcc driver version egcs-2.91.66 19990314 (egcs-1.1.2 release) executing gcc version 2.7.2.3) 
From: Marc Lehmann <pcg AT goof DOT com>
Reply-To: pgcc AT delorie DOT com
X-Mailing-List: pgcc AT delorie DOT com
X-Unsubscribes-To: listserv AT delorie DOT com
Precedence: bulk

On Fri, Jul 16, 1999 at 01:59:35PM -0700, Derrick Bass wrote:
> I am using a Pentium Pro (180 MHz, Redhat 5.2) and some of my FP
> intesive code occassionally slows by a factor of 4-6.  I think I've
> traced this to underflow. . .
> 
> Does handling underflow slow down floating point?

Yes, it can, however it does not do so on all steps of the cpu.

> switch or library call to make the underflows silently get replaced by
> zero?

No. You might experiment with setting the precision (see fpu_control.h) to
something else, but that might not help.

--  
      -----==-                                             |
      ----==-- _                                           |
      ---==---(_)__  __ ____  __       Marc Lehmann      +--
      --==---/ / _ \/ // /\ \/ /       pcg AT goof DOT com      |e|
      -=====/_/_//_/\_,_/ /_/\_\       XX11-RIPE         --+
    The choice of a GNU generation                       |
                                                         |