A bit of benchmarking ...

[Jeremy Allison]

On Fri, Mar 01, 2019 at 09:38:37AM +0100, Andreas Schneider wrote:
> On Thursday, February 28, 2019 5:52:26 PM CET Jeremy Allison wrote:
> > On Wed, Feb 27, 2019 at 03:50:45PM +0100, Andreas Schneider via samba-
> technical wrote:
> > > Hello,
> > > 
> > > as you might have noticed from my commits, I'm working on migrating Samba
> > > to GnuTLS. I've already done some work and there is still a lot ahead,
> > > but as I've migrated SMB3 encryption the last days I've did a bit of
> > > benchmarking ...
> > > 
> > > https://hackmd.io/s/S1bJnG4IN#
> > > 
> > > callgrind output as requested by metze is here:
> > > 
> > > https://xor.cryptomilk.org/samba/crypto/
> > > 
> > > The numbers for gnutls could be a bit better, I need to malloc buffers and
> > > memcpy vectors into them.
> > > 
> > > Lately they added a gnutls_aead_cipher_encryptv() so we could pass in the
> > > vector directly. However they don't work inplace and there is no decryptv
> > > function. So I've opened bugs so we get better APIs.
> > > 
> > > https://gitlab.com/gnutls/gnutls/issues/717
> > > https://gitlab.com/gnutls/gnutls/issues/718
> > > https://gitlab.com/gnutls/gnutls/issues/719
> > 
> > Oh, this looks really great Andreas - thanks !
> > 
> > It would be wonderful to move from any home-grown
> > crypto to GnuTLS for everything !
> 
> I've tested AES-GCM with Samba Crypto (Intel AES NI) and GnuTLS.
> 
> TL;DR
> 
> Samba Crypto (Intel AES NI):  10 min
> GnuTLS:                       12 sec
> 
> https://hackmd.io/s/S1bJnG4IN

Wow, that is a .. significant change. Any ideas
on why the difference is so great ?