Karel Zak <zakkr@zf.jcu.cz> writes:
> I again a little look at aset code and I probably found small performance
> reserve in small chunks (chunk <= ALLOC_CHUNK_LIMIT) reallocation.
Hmm. I wouldn't have thought that realloc got called often enough to be
worth optimizing, but it does seem to get called a few hundred times
during the regress tests, so maybe it's worth a little more code to do
this. (Looks like most of the realloc calls come from enlargeStringInfo
while dealing with long query strings --- since in this case the string
buffer is the only thing yet allocated in QueryContext, the special-case
check wins.)
I've committed this change. Thanks for the suggestion!
regards, tom lane