OGR prerelease g ~15% slower than production on AMD K6-III/400

[Joe O]

[Dec 02 16:45:23 UTC] Summary: 9 OGR packets (3.06 Tnodes)
13.09:36:02.97 - [2.64 Mnodes/s]
[Dec 04 10:29:18 UTC] OGR: Summary: 2 packets (293.99 stats units)
1.11:47:27.33 - [2.28 Mnodes/s] <<<<<<<<<<<<<<<<<<<<
[Dec 09 02:19:47 UTC] Summary: 7 OGR packets (1.08 Tnodes)
4.15:51:07.42 - [2.68 Mnodes/s]

 

[Ken g6]

That's not exactly surprising. From what I understand, they leave a lot of debugging code and other junk in the prereleases that they will clean out for the actual release. Pre-releases are basically to determine correctness of code, not speed.

 

[Joe O]

Then explain this:
[Nov 28 14:45:26 UTC] OGR: Summary: 6 packets (986.23 stats units)
1.14:52:16.78 - [7.04 Mnodes/s] <<<<<<<<<<g
[Nov 06 08:59:27 UTC] Summary: 323 OGR packets (25.20 Tnodes)
55.13:47:42.34 - [5.24 Mnodes/s] <<<<<release
On a Dual Celeron 500.

 

[Joe O]

This was my first known double post!

 

[Riv]

I know that Celeron and P2/P3 processors have been experiencing 10-15% speedups. If AMD K6 processors are slower, well, not much to do about it (simply use the best client). In the latest released source (from May I think) the OGR code was all ANSI C so there were no processor specific cores and a change may very well affect different processors differently. Bug Cyrus (cyp@fb14.uni-mainz.de) about it.