* oostore on x86
@ 2012-04-07 19:48 Xin Tong
2012-04-08 21:49 ` H. Peter Anvin
0 siblings, 1 reply; 6+ messages in thread
From: Xin Tong @ 2012-04-07 19:48 UTC (permalink / raw)
To: linux-kernel
how can one enable oostore on x86, does one need to recompile the kernel ?
Thanks
Xin
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: oostore on x86
2012-04-07 19:48 oostore on x86 Xin Tong
@ 2012-04-08 21:49 ` H. Peter Anvin
2012-04-08 23:27 ` Dave Jones
0 siblings, 1 reply; 6+ messages in thread
From: H. Peter Anvin @ 2012-04-08 21:49 UTC (permalink / raw)
To: Xin Tong; +Cc: linux-kernel
On 04/07/2012 12:48 PM, Xin Tong wrote:
> how can one enable oostore on x86, does one need to recompile the kernel ?
Yes, and expect the code to have seriously bitrotted.
-hpa
--
H. Peter Anvin, Intel Open Source Technology Center
I work for Intel. I don't speak on their behalf.
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: oostore on x86
2012-04-08 21:49 ` H. Peter Anvin
@ 2012-04-08 23:27 ` Dave Jones
2012-04-09 2:15 ` Xin Tong
2012-04-09 9:37 ` Alan Cox
0 siblings, 2 replies; 6+ messages in thread
From: Dave Jones @ 2012-04-08 23:27 UTC (permalink / raw)
To: H. Peter Anvin; +Cc: Xin Tong, linux-kernel
On Sun, Apr 08, 2012 at 02:49:56PM -0700, H. Peter Anvin wrote:
> On 04/07/2012 12:48 PM, Xin Tong wrote:
> > how can one enable oostore on x86, does one need to recompile the kernel ?
>
> Yes, and expect the code to have seriously bitrotted.
It also only ever worked on the early IDT Winchip processors.
They weren't particularly great when they came out in the late 90s,
and I doubt they've improved with age.
I'd estimate the number of people running anything newer than a
2.4 kernel on those chips to be approximating zero, largely because
those old systems rarely had a lot of memory, and modern 2.6 capable
distros probably run like a dog on them.
I've suggested just removing that code in the past a few times.
Even if it does still work, anyone complaining about performance loss
of a modern kernel on something of that era can't be taken seriously.
Dave
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: oostore on x86
2012-04-08 23:27 ` Dave Jones
@ 2012-04-09 2:15 ` Xin Tong
2012-04-09 9:43 ` Alan Cox
2012-04-09 9:37 ` Alan Cox
1 sibling, 1 reply; 6+ messages in thread
From: Xin Tong @ 2012-04-09 2:15 UTC (permalink / raw)
To: Dave Jones, H. Peter Anvin, Xin Tong, linux-kernel
Is there a way to relax the amd64 memory model. does the latest intel
processor micro-architecture support it ?
Xin
On Sun, Apr 8, 2012 at 7:27 PM, Dave Jones <davej@redhat.com> wrote:
> On Sun, Apr 08, 2012 at 02:49:56PM -0700, H. Peter Anvin wrote:
> > On 04/07/2012 12:48 PM, Xin Tong wrote:
> > > how can one enable oostore on x86, does one need to recompile the kernel ?
> >
> > Yes, and expect the code to have seriously bitrotted.
>
> It also only ever worked on the early IDT Winchip processors.
> They weren't particularly great when they came out in the late 90s,
> and I doubt they've improved with age.
>
> I'd estimate the number of people running anything newer than a
> 2.4 kernel on those chips to be approximating zero, largely because
> those old systems rarely had a lot of memory, and modern 2.6 capable
> distros probably run like a dog on them.
>
> I've suggested just removing that code in the past a few times.
> Even if it does still work, anyone complaining about performance loss
> of a modern kernel on something of that era can't be taken seriously.
>
> Dave
>
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: oostore on x86
2012-04-08 23:27 ` Dave Jones
2012-04-09 2:15 ` Xin Tong
@ 2012-04-09 9:37 ` Alan Cox
1 sibling, 0 replies; 6+ messages in thread
From: Alan Cox @ 2012-04-09 9:37 UTC (permalink / raw)
To: Dave Jones; +Cc: H. Peter Anvin, Xin Tong, linux-kernel
> I've suggested just removing that code in the past a few times.
> Even if it does still work, anyone complaining about performance loss
> of a modern kernel on something of that era can't be taken seriously.
I'm not sure removing it makes sense given what a tiny spec of code is
involved. Also in practice most of the same bits are actually used with
different fencing operators on PPro for errata and on PIII and higher for
temporal stores. Of the tree tiny changes it causes two are identical to
PPro...
Alan
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: oostore on x86
2012-04-09 2:15 ` Xin Tong
@ 2012-04-09 9:43 ` Alan Cox
0 siblings, 0 replies; 6+ messages in thread
From: Alan Cox @ 2012-04-09 9:43 UTC (permalink / raw)
To: Xin Tong; +Cc: Dave Jones, H. Peter Anvin, linux-kernel
On Sun, 8 Apr 2012 22:15:53 -0400
Xin Tong <xerox.time.tech@gmail.com> wrote:
> Is there a way to relax the amd64 memory model. does the latest intel
> processor micro-architecture support it ?
No, nor is it clear it would make any sense to do so. The newer
processors are capable of all sorts of clever ordering of their own. The
IDT WinChip wasn't.
Alan
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2012-04-09 9:41 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2012-04-07 19:48 oostore on x86 Xin Tong
2012-04-08 21:49 ` H. Peter Anvin
2012-04-08 23:27 ` Dave Jones
2012-04-09 2:15 ` Xin Tong
2012-04-09 9:43 ` Alan Cox
2012-04-09 9:37 ` Alan Cox
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.