On Thu, 2014-08-14 at 12:52 -0400, Robert Haas wrote:
> It appears to me that the performance characteristics for this version
> are not significantly different from version 1. I have not looked at
> the code.
While trying to reproduce your results, I noticed what might be around a
1% regression just from adding the 3 fields to MemoryContextData. If I
cut it down to adding just one field, the regression disappears.
The results are fairly noisy, so I could be chasing the wrong thing. But
one reason to believe it is that I pushed the size of MemoryContextData
above 64, which sounds like it might be an important threshold.
Regards,Jeff Davis