will get you quite far.
We run with a 20MB syscache and a 10MB relcache with 100k+ objects and hundreds
of backends
A dumb LRU is plenty good for the purpose.
That being said I would love to see these caches shared. :-)
Cheers
Serge
Salesforce
Yeah, our approach had to mergeable. You can rip that bandaid.
Big +1 on this one.
Here is what we did. It's very crude, but minimized the amount of pain:
It helps that the C compiler treats arrays and pointers the same.
I can dig for the complete patch if you want...
Cheers
Serge
/*
* This struct is the data actually passed to an fmgr-c
rrep->stats != context->stats)
sAtomicAdd(context->stats, size);
}
}
}
I'll try myself on peeling out a patch for community for the stats and ask he
developer responsible for hash spilling to engage.
Cheers
Serge
--
Tomas Vondra http://www.2ndQ
Context, ..)
That's quite trivial, actually.
Also we have fixed all those missing hash spills - albeit based on the 9.6 hash
table design I think.
If there is interest by the community we are very willing to share.
Cheers
Serge
Salesforce
hand it’s
conceptually pretty straight forward.
Cheers Serge Rielau SFDC
