"Pavan Deolasee" <pavan.deolasee@gmail.com> writes:
> On 1/31/07, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> The toast code takes pains to ensure that the tuples it creates won't be
>> subject to re-toasting. Else it'd be an infinite recursion.
> I think I found it. The toast_insert_or_update() function gets into an
> unnecessary recursion because of alignment issues. It thus toasts
> already toasted data. This IMHO might be causing unnecessary
> overheads for each toast operation.
Interesting --- I'd never seen this because both of my usual development
machines have MAXALIGN 8, and it works out that that makes
TOAST_MAX_CHUNK_SIZE 1986, which makes the actual toasted tuple size
2030, which maxaligns to 2032, which is still less than
TOAST_TUPLE_THRESHOLD. I think the coding was implicitly assuming that
TOAST_TUPLE_THRESHOLD would itself be a maxalign'd value, but it's not
necessarily (and in fact not, with the current page header size ---
I wonder whether the bug was originally masked because the page header
size was different??)
We can't change TOAST_MAX_CHUNK_SIZE without forcing an initdb, but I
think that it would be safe to remove the MAXALIGN'ing of the tuple
size in the tests in heapam.c, that is
if (HeapTupleHasExternal(tup) || (MAXALIGN(tup->t_len) > TOAST_TUPLE_THRESHOLD)) heaptup =
toast_insert_or_update(relation,tup, NULL); else heaptup = tup;
becomes
if (HeapTupleHasExternal(tup) || (tup->t_len > TOAST_TUPLE_THRESHOLD)) heaptup =
toast_insert_or_update(relation,tup, NULL); else heaptup = tup;
which'll save a cycle or two as well as avoid this corner case.
It seems like a number of the uses of MAXALIGN in tuptoaster.c
are useless/bogus as well. Comments?
regards, tom lane