Prev: PG 9.0 release timetable
Next: Is there anyway to get list of table name, beforeraw parser is analyze?
From: Andres Freund on 30 May 2010 15:43
On Sunday 30 May 2010 18:29:31 Greg Stark wrote:
> On Sun, May 30, 2010 at 4:54 AM, Tom Lane <tgl(a)sss.pgh.pa.us> wrote:
> > I read through that thread and couldn't find much discussion of
> > alternative CRC implementations --- we spent all our time on arguing
> > about whether we needed 64-bit CRC or not.
>
> Alright, how about this thread?
>
> http://thread.gmane.org/gmane.comp.db.postgresql.devel.general/71741
>
> This actually sounds like precisely the same algorithm. Perhaps this
> implementation is much better but your tests on the old one showed a
> big difference between smaller and larger data sequences.
I haven't yet had a chance to read the intel paper (I am in the train and
latency is 30s+ and the original link is dead), but I got the sf.net
implementation.

Seeing it I think I might know the reason why it wasn't as much faster as
promised - it introduces ordering constraints by avoiding shifts by using
term2. Not sure though.

Anybody got the implementation by Gurjeet? I couldn't find it online (within
the constraints of the connection).

Greetings,

Andres

--
Sent via pgsql-hackers mailing list (pgsql-hackers(a)postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers

From: Andres Freund on 30 May 2010 16:48
On Sunday 30 May 2010 18:43:12 Greg Stark wrote:
> On Sun, May 30, 2010 at 5:29 PM, Greg Stark <gsstark(a)mit.edu> wrote:
> > On Sun, May 30, 2010 at 4:54 AM, Tom Lane <tgl(a)sss.pgh.pa.us> wrote:
> >> I read through that thread and couldn't find much discussion of
> >> alternative CRC implementations --- we spent all our time on arguing
> >> about whether we needed 64-bit CRC or not.
> >
> > Alright, how about this thread?
> >
> > http://thread.gmane.org/gmane.comp.db.postgresql.devel.general/71741
>
> Huh, actually apparently this is right about on schedule for
> reconsidering this topic:
>
> http://article.gmane.org/gmane.comp.db.postgresql.devel.general/71903
Oh, and the first zlib version sporting the 4 separate shifted tables approach
was 1.2.0 (9 March 2003) ;-)

Andres

--
Sent via pgsql-hackers mailing list (pgsql-hackers(a)postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers

From: Florian Pflug on 7 Jun 2010 08:10
On Jun 7, 2010, at 12:45 , Andres Freund wrote:
> On Monday 07 June 2010 12:37:13 Pierre C wrote:
>>> On Sunday 30 May 2010 18:29:31 Greg Stark wrote:
>>>> On Sun, May 30, 2010 at 4:54 AM, Tom Lane <tgl(a)sss.pgh.pa.us> wrote:
>>>>> I read through that thread and couldn't find much discussion of
>>>>> alternative CRC implementations --- we spent all our time on arguing
>>>>> about whether we needed 64-bit CRC or not.
>>
>> SSE4.2 has a hardware CRC32 instruction, this might be interesting to
>> use...
> Different polynom unfortunately...

Since only the WAL uses CRC, I guess the polynomial could be changed though. pg_upgrade for example shouldn't care.

RFC3385 compares different checksumming methods for use in iSCSI, and CRC32c (which uses the same polynomial as the SSE4.2 instruction) wins. Here's
a link: http://www.faqs.org/rfcs/rfc3385.html

best regards,
Florian Pflug


--
Sent via pgsql-hackers mailing list (pgsql-hackers(a)postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers

From: Andres Freund on 7 Jun 2010 06:45
On Monday 07 June 2010 12:37:13 Pierre C wrote:
> > On Sunday 30 May 2010 18:29:31 Greg Stark wrote:
> >> On Sun, May 30, 2010 at 4:54 AM, Tom Lane <tgl(a)sss.pgh.pa.us> wrote:
> >> > I read through that thread and couldn't find much discussion of
> >> > alternative CRC implementations --- we spent all our time on arguing
> >> > about whether we needed 64-bit CRC or not.
>
> SSE4.2 has a hardware CRC32 instruction, this might be interesting to
> use...
Different polynom unfortunately...

Andres

--
Sent via pgsql-hackers mailing list (pgsql-hackers(a)postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers

First  |  Prev  | 
Pages: 1 2
Prev: PG 9.0 release timetable
Next: Is there anyway to get list of table name, beforeraw parser is analyze?