public inbox for gcc@gcc.gnu.org
 help / color / mirror / Atom feed
* Re: K6 patches for review
@ 1998-08-24 18:48 Mike Stump
  0 siblings, 0 replies; 12+ messages in thread
From: Mike Stump @ 1998-08-24 18:48 UTC (permalink / raw)
  To: davem, law; +Cc: egcs, jfc

> Date: Sun, 23 Aug 1998 20:53:53 -0600
> From: Jeffrey A Law <law@cygnus.com>

> It was for a target which is still covered by an NDA.

We all at times have to deal with NDAs.

> It probably would have been better if we had included at least one
> example for a public target.

Well, we can understand that this is hard to do in general.  Another
possibility might have been to just give us all a short primer on it,
send any docs in the code to the list once (for new code), and just be
a a little more responsive (if time permits) than usual on the new
feature.  I must admit, I was sleeping and didn't even know about the
new flexibility.

^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: K6 patches for review
@ 1998-09-06  1:09 John Wehle
  0 siblings, 0 replies; 12+ messages in thread
From: John Wehle @ 1998-09-06  1:09 UTC (permalink / raw)
  To: law; +Cc: egcs, egcs-patches, jfc

> They looked fine to me.  Note that John Wehle had some changes to
> improve scheduling on Pentiums that needed minor updating and may
> conflict with your patch.

Actually my patch was for improving scheduling on the Pentium Pro /
Pentium II.

> I'd like to get both the Pentium improvements and your K6 stuff into
> the tree.  I don't know your schedule for benchmarking this stuff
> (presumably you'd want to do so before installing :-) and I don't
> know John's schedule for updating his patch.  So it's hard to figure
> out what to do next.

I'm about to start looking at improving the scheduling for the Pentium Pro /
Pentium II again.  Among other things I want to see if I can use the
MD_SCHED macros to teach Haifa more about scheduling for the Pentium Pro /
Pentium II.

> I also played with Haifa on my PPro a while back and was getting
> reasonable results (not great, but better than the old scheduler).
> So we might want to think about doing any necessary tuning for Haifa
> and enabling haifa by default on the x86 machines.

Yep, that's what I'm shooting for.

-- John
-------------------------------------------------------------------------
|   Feith Systems  |   Voice: 1-215-646-8000  |  Email: john@feith.com  |
|    John Wehle    |     Fax: 1-215-540-5495  |                         |
-------------------------------------------------------------------------


^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: K6 patches for review
  1998-08-22  6:46 John Carr
  1998-08-22 11:00 ` David S. Miller
@ 1998-09-05 20:30 ` Jeffrey A Law
  1 sibling, 0 replies; 12+ messages in thread
From: Jeffrey A Law @ 1998-09-05 20:30 UTC (permalink / raw)
  To: John Carr; +Cc: egcs, egcs-patches, John Wehle

  In message <199808221346.JAA09718@pc.>you write:
  > 
  > Please comment.
  > 
  > This patch adds AMD K6 optimizations, based on AMD publication 21924B
  > for the K6-2.  I don't know if the K6 pipeline is significantly
  > different.  I have not run benchmarks yet.
They looked fine to me.  Note that John Wehle had some changes to
improve scheduling on Pentiums that needed minor updating and may
conflict with your patch.

I'd like to get both the Pentium improvements and your K6 stuff into
the tree.  I don't know your schedule for benchmarking this stuff
(presumably you'd want to do so before installing :-) and I don't
know John's schedule for updating his patch.  So it's hard to figure
out what to do next.

I also played with Haifa on my PPro a while back and was getting
reasonable results (not great, but better than the old scheduler).
So we might want to think about doing any necessary tuning for Haifa
and enabling haifa by default on the x86 machines.


jeff

^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: K6 patches for review
  1998-08-25  5:21     ` Joern Rennecke
@ 1998-08-25 18:14       ` David S. Miller
  0 siblings, 0 replies; 12+ messages in thread
From: David S. Miller @ 1998-08-25 18:14 UTC (permalink / raw)
  To: amylaar; +Cc: law, jfc, egcs

   From: Joern Rennecke <amylaar@cygnus.co.uk>
   Date: Mon, 24 Aug 1998 19:24:26 +0100 (BST)

   It is crucial to note that while you can adjust a cost downwards,
   adjusting it upwards doesn't work.  So if you have an exception
   where a cost is higher, you have to set the default cost to the
   higher value, and use ADJUST_COST to lower it in the
   non-exceptional case.

Or just use the MD_SCHED_* facilties and avoid all of these issues :-)

Later,
David S. Miller
davem@dm.cobaltmicro.com

^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: K6 patches for review
  1998-08-24 18:48         ` David S. Miller
@ 1998-08-25 14:32           ` Jeffrey A Law
  0 siblings, 0 replies; 12+ messages in thread
From: Jeffrey A Law @ 1998-08-25 14:32 UTC (permalink / raw)
  To: David S. Miller; +Cc: jfc, egcs

  In message < 199808250013.RAA28708@dm.cobaltmicro.com >you write:
  >    Date: Sun, 23 Aug 1998 20:53:53 -0600
  >    From: Jeffrey A Law <law@cygnus.com>
  > 
  >    It was for a target which is still covered by an NDA.
  > 
  >    Cygnus figured the MD_SCHED stuff would be useful to others and
  >    contributing it would not violate the NDA, so Cygnus went ahead and
  >    contributed the code.
  > 
  > I heard this chip was released recently, does this mean the MD_SCHED_*
  > support for it will be soon integrated into the tree?
Possibly.  I don't have the contract in front of me to review for
confidentiality releases.  And in many cases that's just the first
step that has to be taken before Cygnus contributes a port.

Contributions from Cygnus go through a conceptually simple process --
customers get the code first, after our customers have had it for
a while Cygnus contributes the code to egcs.  That's the basic model.
We do make exceptions, for example infrastructure changes like MD_SCHED
can often go directly to egcs.


jeff

^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: K6 patches for review
  1998-08-22 18:06   ` Jeffrey A Law
  1998-08-23 19:12     ` David S. Miller
@ 1998-08-25  5:21     ` Joern Rennecke
  1998-08-25 18:14       ` David S. Miller
  1 sibling, 1 reply; 12+ messages in thread
From: Joern Rennecke @ 1998-08-25  5:21 UTC (permalink / raw)
  To: law; +Cc: davem, jfc, egcs

>   > scheme, now for FPU code Haifa gets quad issue often.  ADJUST_COST is
>   > yet another hack that doesn't work at all...
> I'd argue with "doesn't work at all...".   It's value as diminished
> over time as superscalar machines have become popular.

It is crucial to note that while you can adjust a cost downwards, adjusting
it upwards doesn't work.  So if you have an exception where a cost is higher,
you have to set the default cost to the higher value, and use ADJUST_COST
to lower it in the non-exceptional case.

^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: K6 patches for review
  1998-08-23 19:54       ` Jeffrey A Law
@ 1998-08-24 18:48         ` David S. Miller
  1998-08-25 14:32           ` Jeffrey A Law
  0 siblings, 1 reply; 12+ messages in thread
From: David S. Miller @ 1998-08-24 18:48 UTC (permalink / raw)
  To: law; +Cc: jfc, egcs

   Date: Sun, 23 Aug 1998 20:53:53 -0600
   From: Jeffrey A Law <law@cygnus.com>

   It was for a target which is still covered by an NDA.

   Cygnus figured the MD_SCHED stuff would be useful to others and
   contributing it would not violate the NDA, so Cygnus went ahead and
   contributed the code.

I heard this chip was released recently, does this mean the MD_SCHED_*
support for it will be soon integrated into the tree?

Later,
David S. Miller
davem@dm.cobaltmicro.com

^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: K6 patches for review
  1998-08-23 19:12     ` David S. Miller
@ 1998-08-23 19:54       ` Jeffrey A Law
  1998-08-24 18:48         ` David S. Miller
  0 siblings, 1 reply; 12+ messages in thread
From: Jeffrey A Law @ 1998-08-23 19:54 UTC (permalink / raw)
  To: David S. Miller; +Cc: jfc, egcs

  In message < 199808231632.JAA15715@dm.cobaltmicro.com >you write:
  >    Date: Sat, 22 Aug 1998 18:23:29 -0600
  >    From: Jeffrey A Law <law@hurl.cygnus.com>
  > 
  >    Presumably as we get a better feel for the kinds of problems
  >    everyone uses MD_SCHED to solve we'll have a better clue what kinds
  >    of things we need to do in the generic scheduler :-)
  > 
  > One thing which ate me was this interface went in without any example
  > usage in any backend (yes I know what it was intended for).  So I had
  > to fly by the seat of my pants when I tried to take advantage of it.
It was for a target which is still covered by an NDA.

Cygnus figured the MD_SCHED stuff would be useful to others and contributing
it would not violate the NDA, so Cygnus went ahead and contributed the
code.

It probably would have been better if we had included at least one
example for a public target.

jeff

^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: K6 patches for review
  1998-08-22 18:06   ` Jeffrey A Law
@ 1998-08-23 19:12     ` David S. Miller
  1998-08-23 19:54       ` Jeffrey A Law
  1998-08-25  5:21     ` Joern Rennecke
  1 sibling, 1 reply; 12+ messages in thread
From: David S. Miller @ 1998-08-23 19:12 UTC (permalink / raw)
  To: law; +Cc: jfc, egcs, egcs-patches

   Date: Sat, 22 Aug 1998 18:23:29 -0600
   From: Jeffrey A Law <law@hurl.cygnus.com>

   Presumably as we get a better feel for the kinds of problems
   everyone uses MD_SCHED to solve we'll have a better clue what kinds
   of things we need to do in the generic scheduler :-)

One thing which ate me was this interface went in without any example
usage in any backend (yes I know what it was intended for).  So I had
to fly by the seat of my pants when I tried to take advantage of it.

Later,
David S. Miller
davem@dm.cobaltmicro.com

^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: K6 patches for review
  1998-08-22 11:00 ` David S. Miller
@ 1998-08-22 18:06   ` Jeffrey A Law
  1998-08-23 19:12     ` David S. Miller
  1998-08-25  5:21     ` Joern Rennecke
  0 siblings, 2 replies; 12+ messages in thread
From: Jeffrey A Law @ 1998-08-22 18:06 UTC (permalink / raw)
  To: David S. Miller; +Cc: jfc, egcs, egcs-patches

  In message < 199808221758.KAA06341@dm.cobaltmicro.com >you write:
  >    From: John Carr <jfc@tiac.net>
  >    Date: Sat, 22 Aug 1998 09:46:34 EDT
  > 
  >    Please comment.
  > 
  >    This patch adds AMD K6 optimizations, based on AMD publication 21924B
  >    for the K6-2.  I don't know if the K6 pipeline is significantly
  >    different.  I have not run benchmarks yet.
  > 
  > You will get more profitable results if you use a slotting scheme like
  > I did on the UltraSparc using the MD_SCHED macros.  Haifa is really
  > dumb.  Haifa never got quad issue on Ultra before my hacks, I simply
  > couldn't describe the chip using the current define_function_unit
  > scheme, now for FPU code Haifa gets quad issue often.  ADJUST_COST is
  > yet another hack that doesn't work at all...
I'd argue with "doesn't work at all...".   It's value as diminished
over time as superscalar machines have become popular.

MD_SCHED is the way to go right now.  There's serveral things I want
to describe for the PA using MD_SCHED when I get the chance.  Presumably
as we get a better feel for the kinds of problems everyone uses MD_SCHED 
to solve we'll have a better clue what kinds of things we need to do
in the generic scheduler :-)

jeff

^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: K6 patches for review
  1998-08-22  6:46 John Carr
@ 1998-08-22 11:00 ` David S. Miller
  1998-08-22 18:06   ` Jeffrey A Law
  1998-09-05 20:30 ` Jeffrey A Law
  1 sibling, 1 reply; 12+ messages in thread
From: David S. Miller @ 1998-08-22 11:00 UTC (permalink / raw)
  To: jfc; +Cc: egcs, egcs-patches

   From: John Carr <jfc@tiac.net>
   Date: Sat, 22 Aug 1998 09:46:34 EDT

   Please comment.

   This patch adds AMD K6 optimizations, based on AMD publication 21924B
   for the K6-2.  I don't know if the K6 pipeline is significantly
   different.  I have not run benchmarks yet.

You will get more profitable results if you use a slotting scheme like
I did on the UltraSparc using the MD_SCHED macros.  Haifa is really
dumb.  Haifa never got quad issue on Ultra before my hacks, I simply
couldn't describe the chip using the current define_function_unit
scheme, now for FPU code Haifa gets quad issue often.  ADJUST_COST is
yet another hack that doesn't work at all...

Later,
David S. Miller
davem@dm.cobaltmicro.com

^ permalink raw reply	[flat|nested] 12+ messages in thread

* K6 patches for review
@ 1998-08-22  6:46 John Carr
  1998-08-22 11:00 ` David S. Miller
  1998-09-05 20:30 ` Jeffrey A Law
  0 siblings, 2 replies; 12+ messages in thread
From: John Carr @ 1998-08-22  6:46 UTC (permalink / raw)
  To: egcs, egcs-patches

Please comment.

This patch adds AMD K6 optimizations, based on AMD publication 21924B
for the K6-2.  I don't know if the K6 pipeline is significantly
different.  I have not run benchmarks yet.

begin 644 k6-patch.tar.gz
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-US?[]7\7>(@E  @! +F[
 
end

^ permalink raw reply	[flat|nested] 12+ messages in thread

end of thread, other threads:[~1998-09-06  1:09 UTC | newest]

Thread overview: 12+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
1998-08-24 18:48 K6 patches for review Mike Stump
  -- strict thread matches above, loose matches on Subject: below --
1998-09-06  1:09 John Wehle
1998-08-22  6:46 John Carr
1998-08-22 11:00 ` David S. Miller
1998-08-22 18:06   ` Jeffrey A Law
1998-08-23 19:12     ` David S. Miller
1998-08-23 19:54       ` Jeffrey A Law
1998-08-24 18:48         ` David S. Miller
1998-08-25 14:32           ` Jeffrey A Law
1998-08-25  5:21     ` Joern Rennecke
1998-08-25 18:14       ` David S. Miller
1998-09-05 20:30 ` Jeffrey A Law

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).