[patch] important performance bug fix

[David Mosberger]

>>>>> On 17 Jul 2002 15:50:16 -0400, Jim Wilson <wilson@redhat.com> said:

  Jim> The patch was posted late on Friday.  I didn't see it until
  Jim> Monday.  I didn't have time to look at it until today because
  Jim> of various problems: a power outage, a network router failure,
  Jim> an internal RH mail/news reorg, meetings, and of course the
  Jim> work I have to do to get paid.

No problem.  Like I said, I just wanted to make sure someone is going
to take a look at it (before I forget about it...).

  Jim> There is one problem with what HJ did.  The 2.13 binutils
  Jim> branch was created over the weekend.  Presumably we want this
  Jim> patch in the next binutils release.

Yes, the sooner the better.  It's not good to emit invalid opcodes,
even though in practice it doesn't make any difference on today's CPUs
(e.g., there could be tools that rely on the proper encodings).  [In
the original mail, I reported that on McKinley the wrong encoding can
make a significant performance difference; this turns out to be not
the case, so this is strictly an architectural correctness
issue---neither Merced nor McKinley are affected by the bug.]

  Jim> I can take care of that.  I'll have to check out the branch,
  Jim> apply the patch, and test it, so it might not get checked in
  Jim> until tomorrow.

That would be great!

Thanks!

	--david