On Sat, Sep 29, 2007 at 09:07:30AM -0700, Linus Torvalds wrote:
>
>
> On Sat, 29 Sep 2007, Nick Piggin wrote:
> > >
> > > The non-temporal stores should be basically considered to be "IO", not
> > > any
> > > normal memory operation.
> >
> > Maybe you're thinking of uncached / WC? Non-tempora
On Fri, Sep 28, 2007 at 05:07:19PM +0100, Alan Cox wrote:
> > Winchip: can any of these CPUs with ooostores do SMP? If not, then smp_wmb
> > can also be a simple barrier on i386 too.
>
> The IDT Winchip can do SMP apparently.
>From the Winchip3 (which was the final winchip) specs..
"The ID
On Sat, 29 Sep 2007, Nick Piggin wrote:
> >
> > The non-temporal stores should be basically considered to be "IO", not any
> > normal memory operation.
>
> Maybe you're thinking of uncached / WC? Non-temporal stores to cacheable
> RAM apparently can go out of order too, and they are being used
On Fri, Sep 28, 2007 at 06:18:31PM +0100, Alan Cox wrote:
> > on the broken ppro stores config option if you just tell me what should
> > be there (again, remember that my patch isn't actually changing anything
> > already there except for smp_rmb side).
>
> The PPro needs rmb to ensure a store do
On Fri, Sep 28, 2007 at 09:15:06AM -0700, Linus Torvalds wrote:
>
>
> On Fri, 28 Sep 2007, Alan Cox wrote:
> >
> > However
> > - You've not shown the patch has any performance gain
>
> It would be nice to see this.
Actually, in a userspace test I have (which actually does enough
work to trigg
> on the broken ppro stores config option if you just tell me what should
> be there (again, remember that my patch isn't actually changing anything
> already there except for smp_rmb side).
The PPro needs rmb to ensure a store doesn't go for a walk on the wild
side and pass the read especially wh
On Fri, Sep 28, 2007 at 05:07:19PM +0100, Alan Cox wrote:
> > The only alternative is to assume a weak memory model, and add the required
> > barriers to spin_unlock -- something that has been explicitly avoided, but
>
> We have the barriers in spin_unlock already for Pentium Pro and IDT
> Winchip
On Fri, 28 Sep 2007, Alan Cox wrote:
>
> However
> - You've not shown the patch has any performance gain
It would be nice to see this.
> - You've probably broken Pentium Pro
Probably not a big deal, but yeah, we should have that broken-ppro option.
> - and for modern processors its still no
> The only alternative is to assume a weak memory model, and add the required
> barriers to spin_unlock -- something that has been explicitly avoided, but
We have the barriers in spin_unlock already for Pentium Pro and IDT
Winchip systems. The Winchip explicitly supports out of order store (and
wa
According to latest memory ordering specification documents from Intel and
AMD, both manufacturers are committed to in-order loads from cacheable memory
for the x86 architecture. Hence, smp_rmb() may be a simple barrier.
Also according to those documents, and according to existing practice in Linu
10 matches
Mail list logo
