From mboxrd@z Thu Jan  1 00:00:00 1970
From: Jeffrey A Law <law@cygnus.com>
To: Bernd Schmidt <crux@pool.informatik.rwth-aachen.de>
Cc: egcs@cygnus.com, scox@cygnus.com
Subject: Re: DImode speedup on i386 
Date: Thu, 19 Feb 1998 03:32:00 -0000
Message-id: <472.887875330@hurl.cygnus.com>
References: <Pine.SOL.3.90.980218113225.12315H-100000@bond.informatik.rwth-aachen.de>
X-SW-Source: 1998-02/msg00918.html

  In message < Pine.SOL.3.90.980218113225.12315H-100000@bond.informatik.rwth-aachen.de >you write:
  > 
  > The following patch improves performance of DImode operations on the i386.
  > For the chess program Crafty, which relies heavily upon those operations,
  > this patch improved the "nodes per second" count in the built-in benchmark
  > from 79558 to 93047. It also made the executable 12000 bytes smaller.
  > 
  > Bernd
  > 
  > 	* i386/i386.c (ix86_logical_operator): New function.
  > 	(split_di): When splitting a MEM, make sure the new MEMs have
  > 	SImode.
  > 	* i386/i386.md (xordi3, anddi3, iordi3): New patterns.
  > 	Add a define_split to implement them.
One general note -- you have to be *extremely* careful defining DImode
patterns on the x86 due to the bad interactions with register
reloading.

I would like Stan's opinion on these patches before they're installed
since he had the "pleasure" of dealing with the last round of
DImode problems on the x86.

jeff