Horton

Re: Avoiding repeated snapshot computation

Ants Aasma <ants.aasma@eesti.ee>

View thread
From: Ants Aasma <ants.aasma@eesti.ee>
To: Pavan Deolasee <pavan.deolasee@gmail.com>
Cc: Tom Lane <tgl@sss.pgh.pa.us>, Robert Haas <robertmhaas@gmail.com>, pgsql-hackers <pgsql-hackers@postgresql.org>
Date: 2011-11-30T00:03:23Z
Lists: pgsql-hackers

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

  1. Slightly reorganize struct SnapshotData.

Attachments

On Tue, Nov 29, 2011 at 7:12 AM, Pavan Deolasee
<pavan.deolasee@gmail.com> wrote:
> I think that a good idea. We need a representation that needs minimum
> processing to derive the snapshot.

I was looking over the generated code for GetSnapshotData to see if there
is any low hanging fruit for micro-optimization. The assembly mostly looks
pretty tight, but there are 3 function calls to TransactionIdPrecedes and
TransactionIdFollowsOrEquals. All the parameters are known to be normal
xids, so there are duplicated checks for that and a lot of movs for the calling
convention. I wonder if replacing them with special case macros would be
a good idea. In that case the whole check will compile down to one cmp
instruction. I'm running a set of benchmarks now on my laptop, but I guess
the difference will mostly become noticeable on beefier hardware when
ProcArray lock is heavily contended. Attached is a patch, if anyone wishes
to give it a go.

--
Ants Aasma