Re: [HACKERS] Should we cacheline align PGXACT?

On Mon, Feb 13, 2017 at 8:02 PM, Ashutosh Sharma <ashu.coek88@gmail.com> wrote:

I too have performed benchmarking of this patch on a large machine
(with 128 CPU(s), 520GB RAM, intel x86-64 architecture) and would like
to share my observations for the same (Please note that, as I had to
reverify readings on few client counts, it did take some time for me
to share these test-results.)

 

Great! Thank you very much for testing.

Case3: Data fits in shared buffer, Read-write workload:
-----------------------------------------------------------------------------
In this case, I could see that the tps on head and patch are very
close to each other with a small variation of (+-)3-4% which i assume
is a run-to-run variation. PFA result sheet
'results-readwrite-300-1000-SF' containing the test-results.

I wouldn't say it's just a variation.  It looks like relatively small but noticeable regression in the patch.

According to Andres comment [1] I made a version of patch (pgxact-align-3.patch) which align PGXACT to 16 bytes.

That excludes situation when single PGXACT is spread over 2 cache lines.

Results of read-only tests are attached.  We can see that 16-byte alignment gives speedup in read-only tests, but it's a bit less than speedup of cache line alignment version.

Read-write tests are now running.  Hopefully 16-byte alignment version of patch wouldn't cause regression in read-write benchmark.

1. https://www.postgresql.org/message-id/20160822021747.u5bqx2xwwjzac5u5@alap3.anarazel.de