Re: Parallel bitmap heap scan

On Wed, Feb 8, 2017 at 8:07 AM, Robert Haas <robertmhaas(at)gmail(dot)com> wrote:

.Thanks for your input, I have few queries about these comments.

> 2. Add a new function dsa_pointer tbm_prepare_shared_iterate(TIDBitmap
> *tbm) which allocates shared iteration arrays using the DSA passed to
> tbm_create() and returns a dsa_pointer to the result. Arrange this so
> that if it's called more than once, each call returns a separate
> iterator, so that you can call it once to get the main iterator and a
> second time for the prefetch iterator, but have both of those point to
> the same underlying iteration arrays.
>
> 3. Add a new function TBMSharedIterator
> *tbm_attach_shared_iterate(dsa_area *dsa, dsa_pointer) which is called
> once per backend and gets passed the dsa_pointer from the previous
> step.

IIUC, tbm_prepare_shared_iterate will be called only by the leader,
for tbmiterator as well as for the prefetch_iterator. And,
tbm_attach_shared_iterate will be called by each backend and for both
the iterators.

IMHO, tbm_attach_shared_iterate should return TBMIterator with
reference to TBMSharedIterator inside it. The reason behind same is
that we can not keep TBMIterateResult inside TBMSharedIterator
otherwise, results will also go in shared memory but we want to have
local result memory for each worker so that other worker doesn't
disturb it.

Another option can be that we change the tbm_shared_iterate as explained below

TBMIterateResult * tbm_shared_iterate(TBMSharedIterator *,
TBMIterateResult *result).

Now, if result passed to this API is NULL then it will allocate the
memory for the result and that way we will have local result memory,
and if it's not NULL we will use this memory to store our results.
BitmapHeapNode already having a reference to the TBMIterateResult so
we should not have any problem in passing this reference to the
tbm_shared_iterate. I think this one looks better than what I
explained above.

Please suggest.

>
> 4. Add a new function TBMIterateResult
> *tbm_shared_iterate(TBMSharedIterator *) to fetch the next result.

--
Regards,
Dilip Kumar
EnterpriseDB: http://www.enterprisedb.com