On Fri, May 19, 2023 at 4:04 PM Andres Freund <andres@anarazel.de> wrote:
With "yet to see any significant changes" do you mean that the runs are comparable with earlier runs, showing the same regression? Or that the regression vanished? Or ...?
I mean that I might be chasing noise and the mean+stddev for throughput in version 16 pre-beta so far appears to be similar to 15.2. When I ran the insert benchmark a few times, I focused on the cases where 16 pre-beta was worse than 15.2 while ignoring the cases where it was better. Big regressions are easy to document, small ones not so much.
Regardless, I am repeating tests from both the insert benchmark and sysbench for version 16 (pre-beta, and soon beta1).