Horton

Re: xlc atomics

Noah Misch <noah@leadboat.com>

View thread View on postgr.es
From: Noah Misch <noah@leadboat.com>
To: Andres Freund <andres@anarazel.de>
Cc: Pg Hackers <pgsql-hackers@postgresql.org>, Robert Haas <robertmhaas@gmail.com>, Tom Lane <tgl@sss.pgh.pa.us>, Peter Geoghegan <pg@heroku.com>, Ants Aasma <ants@cybertec.at>
Date: 2016-02-15T17:50:07Z
Lists: pgsql-hackers 
On Mon, Feb 15, 2016 at 06:33:42PM +0100, Andres Freund wrote:
> On 2016-02-15 12:11:29 -0500, Noah Misch wrote:
> > These suggested OidGenLock wasn't doing its job.  I've seen similar symptoms
> > around WALInsertLocks with "IBM XL C/C++ for Linux, V13.1.2 (5725-C73,
> > 5765-J08)" for ppc64le.  The problem is generic-xlc.h
> > pg_atomic_compare_exchange_u32_impl() issuing __isync() before
> > __compare_and_swap().  __isync() shall follow __compare_and_swap(); see our
> > own s_lock.h, its references, and other projects' usage:
> 
> Ugh. You're right! It's about not moving code before the stwcx...
> 
> Do you want to add the new test (no objection, curious), or is that more
> for testing?

The patch's test would join PostgreSQL indefinitely.

Commits

  1. Avoid the use of a separate spinlock to protect a LWLock's wait queue.

  2. Improve LWLock scalability.

  3. Reduce the number of semaphores used under --disable-spinlocks.

  4. Alter the configure script to fail immediately if the C compiler does not