From mboxrd@z Thu Jan  1 00:00:00 1970
From: Jeffrey A Law <law@cygnus.com>
To: Jim Wilson <wilson@cygnus.com>
Cc: hjl@lucon.org (H.J. Lu), scox@cygnus.com, crux@pool.informatik.rwth-aachen.de, egcs@cygnus.com
Subject: Re: More fp bug in egcs 
Date: Tue, 05 May 1998 05:46:00 -0000
Message-id: <23019.894353515@hurl.cygnus.com>
References: <199805032254.PAA08319@rtl.cygnus.com>
X-SW-Source: 1998-05/msg00125.html

  In message < 199805032254.PAA08319@rtl.cygnus.com >you write:
  > 	How does this patch look? It works for my test case.
  > 
  > A reload patch is not a safe way to fix this problem.  Your reload patch
  > affects all programs for all targets.  My patch affects only x86 programs
  > that currently fail.  Since my patch has much more limited scope, it is a
  > much better solution for a egcs-1.0.3 bug fix release.
Agreed!

  > 	I think it is a reload bug. I don't know what the purpose to output the
  > 	last reload for a dead register. I don't what the best solution is and
  > 	I am not sure if my patch covers all cases.
  > 
  > I think it is more accurate to call it a `missed optimization' than a `bug',
  > as the output is correct.  The only reason it fails on the x86 is because of
  > a i386.md bug which my patch fixes.
Agreed again.

HJ -- you're making the same fundamental mistake as you did with the
"disable regmove" bug which really turned out to be a bug in i386.md.

Jim's patch is more correct to fix the bug and several orders of
magnitude safer.

And I think both Jim and I agree that you can make a case for your
patch as an optimization -- in which case it would belong in the
development branch, not the release branch.

jeff