On Fri, Sep 15, 2017 at 11:06:29PM +0200, Ingo Molnar wrote:
>
> Indeed, I suspect they should go through the crypto tree, these fixes are
> independent, they don't depend on anything in the x86 tree.
Sure I can pick them up through cryptodev.
Thanks,
--
Email: Herbert Xu
Home Page: http://gon
* Eric Biggers wrote:
> On Fri, Sep 15, 2017 at 07:34:31AM +0200, Ingo Molnar wrote:
> >
> > * Eric Biggers wrote:
> >
> > > Hi Josh,
> > >
> > > On Wed, Sep 13, 2017 at 05:33:03PM -0500, Josh Poimboeuf wrote:
> > > > And here's v2 of the sha512-avx2 patch. It should hopefully gain back
> >
On Fri, Sep 15, 2017 at 07:34:31AM +0200, Ingo Molnar wrote:
>
> * Eric Biggers wrote:
>
> > Hi Josh,
> >
> > On Wed, Sep 13, 2017 at 05:33:03PM -0500, Josh Poimboeuf wrote:
> > > And here's v2 of the sha512-avx2 patch. It should hopefully gain back
> > > most of the performance lost by v1.
>
* Josh Poimboeuf wrote:
> > So if this observation of mine is true we could go back to the old code for
> > the
> > hotpath, but use RDI for TBL and not reload it in the hotpath.
>
> Thanks for the excellent breakdown.
>
> When I looked at the patch again, I came to the same conclusion as yo
* Eric Biggers wrote:
> Hi Josh,
>
> On Wed, Sep 13, 2017 at 05:33:03PM -0500, Josh Poimboeuf wrote:
> > And here's v2 of the sha512-avx2 patch. It should hopefully gain back
> > most of the performance lost by v1.
> >
> > From: Josh Poimboeuf
> > Subject: [PATCH] x86/crypto: Fix RBP usage i
Hi Josh,
On Wed, Sep 13, 2017 at 05:33:03PM -0500, Josh Poimboeuf wrote:
> And here's v2 of the sha512-avx2 patch. It should hopefully gain back
> most of the performance lost by v1.
>
> From: Josh Poimboeuf
> Subject: [PATCH] x86/crypto: Fix RBP usage in sha512-avx2-asm.S
>
> Using RBP as a t
On Thu, Sep 14, 2017 at 11:16:12AM +0200, Ingo Molnar wrote:
>
> * Josh Poimboeuf wrote:
>
> > I'm still looking at the other one (sha512-avx2), but so far I haven't
> > found a way to speed it back up.
>
> Here's a couple of very quick observations with possible optimizations:
>
> AFAICS the
* Ingo Molnar wrote:
> 1)
>
> Note how R12 is used immediately, right in the next instruction:
>
> vpaddq (TBL), Y_0, XFER
>
> I.e. the RBP fixes lengthen the program order data dependencies - that's a
> new
> constraint and a few extra cycles per loop iteration if the workload is
* Josh Poimboeuf wrote:
> I'm still looking at the other one (sha512-avx2), but so far I haven't
> found a way to speed it back up.
Here's a couple of very quick observations with possible optimizations:
AFAICS the main effect of the RBP fixes is the introduction of a memory load
into
the cr
On Wed, Sep 13, 2017 at 04:24:28PM -0500, Josh Poimboeuf wrote:
> On Fri, Sep 08, 2017 at 10:57:05AM -0700, Eric Biggers wrote:
> > On Thu, Sep 07, 2017 at 11:26:47PM +0200, Ingo Molnar wrote:
> > >
> > > * Eric Biggers wrote:
> > >
> > > > On Thu, Sep 07, 2017 at 09:15:34AM +0200, Ingo Molnar w
On Fri, Sep 08, 2017 at 10:57:05AM -0700, Eric Biggers wrote:
> On Thu, Sep 07, 2017 at 11:26:47PM +0200, Ingo Molnar wrote:
> >
> > * Eric Biggers wrote:
> >
> > > On Thu, Sep 07, 2017 at 09:15:34AM +0200, Ingo Molnar wrote:
> > > >
> > > > * Eric Biggers wrote:
> > > >
> > > > > Thanks for
On Thu, Sep 07, 2017 at 11:26:47PM +0200, Ingo Molnar wrote:
>
> * Eric Biggers wrote:
>
> > On Thu, Sep 07, 2017 at 09:15:34AM +0200, Ingo Molnar wrote:
> > >
> > > * Eric Biggers wrote:
> > >
> > > > Thanks for fixing these! I don't have time to review these in detail,
> > > > but I ran
>
* Eric Biggers wrote:
> On Thu, Sep 07, 2017 at 09:15:34AM +0200, Ingo Molnar wrote:
> >
> > * Eric Biggers wrote:
> >
> > > Thanks for fixing these! I don't have time to review these in detail,
> > > but I ran
> > > the crypto self-tests on the affected algorithms, and they all pass. I
>
On Thu, Sep 07, 2017 at 09:15:34AM +0200, Ingo Molnar wrote:
>
> * Eric Biggers wrote:
>
> > Thanks for fixing these! I don't have time to review these in detail, but
> > I ran
> > the crypto self-tests on the affected algorithms, and they all pass. I also
> > benchmarked them before and afte
* Eric Biggers wrote:
> Thanks for fixing these! I don't have time to review these in detail, but I
> ran
> the crypto self-tests on the affected algorithms, and they all pass. I also
> benchmarked them before and after; the only noticable performance difference
> was
> that sha256-avx2 and
On Fri, Sep 01, 2017 at 05:09:19PM -0700, Eric Biggers wrote:
> Hi Josh,
>
> On Tue, Aug 29, 2017 at 01:05:33PM -0500, Josh Poimboeuf wrote:
> > Many of the x86 crypto functions use RBP as a temporary register. This
> > breaks frame pointer convention, and breaks stack traces when unwinding
> > f
Hi Josh,
On Tue, Aug 29, 2017 at 01:05:33PM -0500, Josh Poimboeuf wrote:
> Many of the x86 crypto functions use RBP as a temporary register. This
> breaks frame pointer convention, and breaks stack traces when unwinding
> from an interrupt in the crypto code.
>
> Convert most* of them to leave R
Many of the x86 crypto functions use RBP as a temporary register. This
breaks frame pointer convention, and breaks stack traces when unwinding
from an interrupt in the crypto code.
Convert most* of them to leave RBP alone.
These pass the crypto boot tests for me. Any further testing would be
ap
18 matches
Mail list logo