Fix comment about promising tuples.
Oversight in commit d168b66682
, which added bottom-up index deletion.
This commit is contained in:
parent
65445469d6
commit
fbe4cb3bd4
|
@ -7879,8 +7879,10 @@ bottomup_sort_and_shrink(TM_IndexDeleteOp *delstate)
|
|||
/*
|
||||
* We're about ready to sort block groups to determine the optimal order
|
||||
* for visiting heap blocks. But before we do, round the number of
|
||||
* promising tuples for each block group up to the nearest power-of-two
|
||||
* (except for block groups where npromisingtids is already 0).
|
||||
* promising tuples for each block group up to the next power-of-two,
|
||||
* unless it is very low (less than 4), in which case we round up to 4.
|
||||
* npromisingtids is far too noisy to trust when choosing between a pair
|
||||
* of block groups that both have very low values.
|
||||
*
|
||||
* This scheme divides heap blocks/block groups into buckets. Each bucket
|
||||
* contains blocks that have _approximately_ the same number of promising
|
||||
|
|
Loading…
Reference in New Issue