Peter A. Daly wrote:
>
> Tom Lane wrote:
>
>>
>> Btree index build is primarily a sort, so cannot have better than
>> O(n*log(n)) performance for random data. Not sure why you'd expect
>> linearity.
>>
>> Increasing SORT_MEM would help the constant factor, however...
>>
> What is the most amount of SORT_MEM it makes sense to allocate?
> Pretend the ONLY thing I care about is BTREE index creation time. (2
> gigs of RAM on my DB machine.) Disk IO is not a bottleneck.
Another though. If postgres has one of my CPU's at near 100%, does that
mean I can't get any more performance out of it? Still, how big can I
make sort_mem?
-Pete