17.05.2023 04:25, Michael Paquier wrote:
The numbers between f193883fc~1 and HEAD+patch are close to each
other. It does not seem to make the whole difference with 15.3, but
most of it. The difference can also be explained with some noise,
based on the number patterns of the third runs?
I have now applied the revert, ready for beta1. Thanks for the
feedback!
Thank you for paying attention to it!
Yes, I ran the benchmark on my workstation, so numbers could vary due parallel
activity. Now I've compared 15.3 (8382864eb) with d8c3106bb and 1d369c9e9,
this time with the CPU boost mode disabled:
1d369c9e9:
10007.130326, 10047.722122, 9920.612426, 10016.053846, 10060.606408
d8c3106bb:
10492.100485, 10505.326827, 10535.918137, 10625.904871, 10573.608859
15.3:
10458.752330, 10308.677192, 10366.496526, 10489.395275, 10319.458041
Best "1d369c9e9" worse than "15.3" by 4.1 percents (10060.61 < 10489.40)
Average "1d369c9e9" worse than "15.3" by 3.6 percents (10010.43 < 10388.56)
Best "d8c3106bb" better than "15.3" by 1.3 percents (10625.90 > 10489.40)
Average "d8c3106bb" better than "15.3" by 1.5 percents (10546.57 > 10388.56)
So it seems that there is nothing left on this plate.
Best regards,
Alexander
