Re: Various silliness in heap_getnext and related routines

Dear Tom,
>
>
> bruc(at)stone(dot)congenomics(dot)com (Robert E. Bruccoleri) writes:
> > For my immediate problem, would removing the spinlock acquisition
> > be OK?
>
> It'd be interesting to remove the marked lines:
>
> bufHdr = &BufferDescriptors[buffer - 1];
> - SpinAcquire(BufMgrLock);
> if (bufHdr->tag.blockNum == blockNumber &&
> RelFileNodeEquals(bufHdr->tag.rnode, relation->rd_node))
> {
> - SpinRelease(BufMgrLock);
> return buffer;
> }
> - return ReadBufferWithBufferLock(relation, blockNumber, true);
>
> and see how that affects your performance issue, if at all.
>

I have made those changes, ran the regression tests, and then issued
eight simultaneous retrieval jobs against the database. Performance is
now greatly improved with all the jobs making quick progress through
the sequential query and completing in very reasonable times.

Thanks very much for your help.

BTW, given the high level of support that you provide to the PostgreSQL
community, it's very accurate to state that support for PostgreSQL
is far superior to that of Oracle, especially for SGI systems.

Sincerely,
Bob

+----------------------------------+------------------------------------+
| Robert E. Bruccoleri, Ph.D. | Phone: 609 737 6383 |
| President, Congenomics, Inc. | Fax: 609 737 7528 |
| 114 W Franklin Ave, Suite K1,4,5 | email: bruc(at)acm(dot)org |
| P.O. Box 314 | URL: http://www.congen.com/~bruc |
| Pennington, NJ 08534 | |
+----------------------------------+------------------------------------+