On Mon, Jul 25, 2016 at 3:02 PM, Thomas Munro
<thomas.munro@enterprisedb.com> wrote:
> I measured the following times for unpatched master, on my 4 core laptop:
>
> 16 workers = 73.067s, 74.869s, 75.338s
> 8 workers = 65.846s, 67.622s, 68.039s
> 4 workers = 68.763s, 68.980s, 69.035s <-- curiously slower here
> 3 workers = 59.701s, 59.991s, 60.133s
> 2 workers = 53.620s, 55.300s, 55.790s
> 1 worker = 21.578s, 21.535s, 21.598s
>
> With the attached patched I got:
>
> 16 workers = 75.341s, 77.445s, 77.635s <- +3.4%
> 8 workers = 67.462s, 68.622s, 68.851s <- +1.4%
> 4 workers = 64.983s, 65.021s, 65.496s <- -5.7%
> 3 workers = 60.247s, 60.425s, 60.492s <- +0.7%
> 2 workers = 57.544s, 57.626s, 58.133s <- +2.3%
> 1 worker = 21.403s, 21.486s, 21.661s <- -0.2%
Correction, that +2.3% for 2 workers should be +4.2%. And to clarify,
I ran the test 3 times as shown and those percentage changes are based
on the middle times.
--
Thomas Munro
http://www.enterprisedb.com