From: Vladislav Vaintroub
Date: September 22 2010 3:52pm
Subject: RE: bzr commit into mysql-5.5-bugfixing branch (vvaintroub:3207)	Bug#56585
List-Archive: http://lists.mysql.com/commits/118848
Message-Id: <000c01cb5a6e$19c897f0$4d59c7d0$@oracle.com>
MIME-Version: 1.0
Content-Type: text/plain;
	charset="us-ascii"
Content-Transfer-Encoding: 7bit



> -----Original Message-----
> From: Dmitri.Lenev@stripped [mailto:Dmitri.Lenev@stripped] On Behalf Of Dmitry Lenev
> Sent: Wednesday, September 15, 2010 12:54 PM
> To: Vladislav Vaintroub
> Cc: commits@stripped
> Subject: Re: bzr commit into mysql-5.5-bugfixing branch (vvaintroub:3207) Bug#56585
> 
> Hello Vlad!
> 
> Here are my comments about new version of your patch:

Hi Dmitri!
Meanwhile there is yet another version of the patch I which I think is superior (thanks to Davi for the idea of giving up on exactly
this patch). 
Thanks a lot for your comments, if you can find time and do not mind there is http://lists.mysql.com/commits/118829   and I'm sorry
for throwing more work for you:)  This is a bit better commented and the bug reports includes 
benchmarks for all cases, for 3 different implementation of rwlock.  I  done the benchmarks myself, since it takes a while if I ask
QA :)

The problem with this exact  patch you reviewed was that it solved a corner case situation pretty well  (mostly-writer access it is
not corner case in MySQL, but for general rwlock they are), while leaving out what would be "standard" case( multiple readers,
single writer). You  are wondering why  I implemented fast-writer prlock rather than rwlock, and this was exactly the reason - it
won't perform that well as general purpose rwlock.

Thanks,
Vlad