Re: svn commit: r272906 - in head/sys: conf libkern netpfil/pf sys

2014-10-18 Thread Dag-Erling Smørgrav
Gleb Smirnoff writes: > Dag-Erling Smørgrav writes: > > It's not intended as a "general kernel hashing function". It is > > intended solely for pf. > Why? If so, it should live in pf and not compiled uncoditionally into > all kernels. Anyway, I don't see reason for doing it non-generic. OK. I

Re: svn commit: r272906 - in head/sys: conf libkern netpfil/pf sys

2014-10-16 Thread Gleb Smirnoff
On Fri, Oct 17, 2014 at 01:30:35AM +0200, Dag-Erling Smørgrav wrote: D> Gleb Smirnoff writes: D> > The function should also be documented in hash(9). D> D> It's not intended as a "general kernel hashing function". It is D> intended solely for pf. Why? If so, it should live in pf and not compile

Re: svn commit: r272906 - in head/sys: conf libkern netpfil/pf sys

2014-10-16 Thread Gleb Smirnoff
Luigi, On Fri, Oct 17, 2014 at 01:04:46AM +0200, Luigi Rizzo wrote: L> > Thanks a lot, Ed. I rerun the tests. Results are fine. Distribution L> > is equal (plots attached). L> > L> > Murmur is 9% +/- 1.5% faster. Sorry for noise. L> L> is that 9% on the hash alone ? Right. L> If so it is sur

Re: svn commit: r272906 - in head/sys: conf libkern netpfil/pf sys

2014-10-16 Thread Dag-Erling Smørgrav
Gleb Smirnoff writes: > The function should also be documented in hash(9). It's not intended as a "general kernel hashing function". It is intended solely for pf. DES -- Dag-Erling Smørgrav - d...@des.no ___ svn-src-head@freebsd.org mailing list http

Re: svn commit: r272906 - in head/sys: conf libkern netpfil/pf sys

2014-10-16 Thread Luigi Rizzo
On Wed, Oct 15, 2014 at 08:14:19AM +0400, Gleb Smirnoff wrote: > On Tue, Oct 14, 2014 at 11:38:23PM -0400, Ed Maste wrote: > E> On 14 October 2014 23:04, Gleb Smirnoff wrote: > E> > > E> > Look at results at the end of email. Guess what -j and -m mean. I > E> > want either proving me doing the tes

Re: svn commit: r272906 - in head/sys: conf libkern netpfil/pf sys

2014-10-16 Thread Gleb Smirnoff
On Wed, Oct 15, 2014 at 11:21:52AM +0200, Dag-Erling Smørgrav wrote: D> Ed Maste writes: D> > murmur3's size argument is bytes, not uint32_ts, so the test is only D> > hashing the first 1/4 of the keys. D> D> That's my fault, and I'm going to change it, because it's illogical. D> The function wor

Re: svn commit: r272906 - in head/sys: conf libkern netpfil/pf sys

2014-10-15 Thread Dag-Erling Smørgrav
Ed Maste writes: > Dag-Erling Smørgrav writes: > > That's my fault, and I'm going to change it, because it's illogical. > > The function works on aligned arrays of int32s and should take a count, > > not a size. I just didn't want to delay the patch even further. > The fact that it works only on

Re: svn commit: r272906 - in head/sys: conf libkern netpfil/pf sys

2014-10-15 Thread Ed Maste
On 15 October 2014 05:21, Dag-Erling Smørgrav wrote: > > That's my fault, and I'm going to change it, because it's illogical. > The function works on aligned arrays of int32s and should take a count, > not a size. I just didn't want to delay the patch even further. The fact that it works only on

Re: svn commit: r272906 - in head/sys: conf libkern netpfil/pf sys

2014-10-15 Thread Dag-Erling Smørgrav
Ed Maste writes: > murmur3's size argument is bytes, not uint32_ts, so the test is only > hashing the first 1/4 of the keys. That's my fault, and I'm going to change it, because it's illogical. The function works on aligned arrays of int32s and should take a count, not a size. I just didn't want

Re: svn commit: r272906 - in head/sys: conf libkern netpfil/pf sys

2014-10-14 Thread Gleb Smirnoff
On Tue, Oct 14, 2014 at 11:38:23PM -0400, Ed Maste wrote: E> On 14 October 2014 23:04, Gleb Smirnoff wrote: E> > E> > Look at results at the end of email. Guess what -j and -m mean. I E> > want either proving me doing the test wrong, or backing the change E> > out ASAP. E> E> It looks like there

Re: svn commit: r272906 - in head/sys: conf libkern netpfil/pf sys

2014-10-14 Thread Ed Maste
On 14 October 2014 23:04, Gleb Smirnoff wrote: > > Look at results at the end of email. Guess what -j and -m mean. I > want either proving me doing the test wrong, or backing the change > out ASAP. It looks like there is indeed an error in the test: > h = murmur3_aligned_32((uint32_t *)&key[i],

Re: svn commit: r272906 - in head/sys: conf libkern netpfil/pf sys

2014-10-14 Thread Gleb Smirnoff
On Wed, Oct 15, 2014 at 07:04:55AM +0400, Gleb Smirnoff wrote: T> On Mon, Oct 13, 2014 at 06:42:11PM +0200, Olivier Cochard-Labbé wrote: T> O> > Author: gnn T> O> > Date: Fri Oct 10 19:26:26 2014 T> O> > New Revision: 272906 T> O> > URL: https://svnweb.freebsd.org/changeset/base/272906 T> O> > T> O

Re: svn commit: r272906 - in head/sys: conf libkern netpfil/pf sys

2014-10-14 Thread Gleb Smirnoff
On Mon, Oct 13, 2014 at 06:42:11PM +0200, Olivier Cochard-Labbé wrote: O> > Author: gnn O> > Date: Fri Oct 10 19:26:26 2014 O> > New Revision: 272906 O> > URL: https://svnweb.freebsd.org/changeset/base/272906 O> > O> > Log: O> > Change the PF hash from Jenkins to Murmur3. In forwarding tests O>

Re: svn commit: r272906 - in head/sys: conf libkern netpfil/pf sys

2014-10-13 Thread Olivier Cochard-Labbé
On Fri, Oct 10, 2014 at 9:26 PM, George V. Neville-Neil wrote: > Author: gnn > Date: Fri Oct 10 19:26:26 2014 > New Revision: 272906 > URL: https://svnweb.freebsd.org/changeset/base/272906 > > Log: > Change the PF hash from Jenkins to Murmur3. In forwarding tests > this showed a conservative

svn commit: r272906 - in head/sys: conf libkern netpfil/pf sys

2014-10-10 Thread George V. Neville-Neil
Author: gnn Date: Fri Oct 10 19:26:26 2014 New Revision: 272906 URL: https://svnweb.freebsd.org/changeset/base/272906 Log: Change the PF hash from Jenkins to Murmur3. In forwarding tests this showed a conservative 3% incrase in PPS. Differential Revision:https://reviews.freebsd.o