Обсуждение: Re: [Fwd: QNX6 port (need some assistance from developers)]

Tom Lane wrote:
>
> Igor Kovalenko <Igor.Kovalenko@motorola.com> writes:
> > The patch is made against 72b2 tarball (not CVS). Do you want me to just
> > post diffs or what do I do to get it merged? And where do I post it,
> > PORTS or PATCHES (as it might be interesting for many platforms)?
>
> Send it to patches.  Note that I doubt we will have any interest in
> applying such a major change to 7.2 --- but we'll take a look at it
> for 7.3.

Why? The change only affects anything when HAVE_POSIX_IPC is defined and
that will only be defined on QNX6 (unless someone goes ahead and does it
for their platforms) which is not currently supported anyway. It still
builds cleanly and passes all tests when that constand is not defined
either.

- igor

> > > The patch is made against 72b2 tarball (not CVS). Do you want me to just
> > > post diffs or what do I do to get it merged? And where do I post it,
> > > PORTS or PATCHES (as it might be interesting for many platforms)?
> > Send it to patches.  Note that I doubt we will have any interest in
> > applying such a major change to 7.2 --- but we'll take a look at it
> > for 7.3.
> Why? The change only affects anything when HAVE_POSIX_IPC is defined and
> that will only be defined on QNX6 (unless someone goes ahead and does it
> for their platforms) which is not currently supported anyway. It still
> builds cleanly and passes all tests when that constand is not defined
> either.

It should be an interesting patch for several platforms. We are right in
the middle of wrapping up the 7.2 release, and the concern is in
reviewing the code and ensuring that it has no side effects on other
platforms.

If this were understood to be a simple patch to get a platform running
on 7.2, then it would go into the tree right now. Since it is likely
more complicated than that (have you posted it yet?), it will certainly
need more discussion to get in.

I'm willing to go either way, and if someone were to verify a
significant performance boost for, say, Linux with the patch enabled
then it would likely get more immediate interest.

                      - Thomas

> > > > The patch is made against 72b2 tarball (not CVS). Do you want me to
just
> > > > post diffs or what do I do to get it merged? And where do I post it,
> > > > PORTS or PATCHES (as it might be interesting for many platforms)?
> > > Send it to patches.  Note that I doubt we will have any interest in
> > > applying such a major change to 7.2 --- but we'll take a look at it
> > > for 7.3.
> > Why? The change only affects anything when HAVE_POSIX_IPC is defined and
> > that will only be defined on QNX6 (unless someone goes ahead and does it
> > for their platforms) which is not currently supported anyway. It still
> > builds cleanly and passes all tests when that constand is not defined
> > either.
>
> It should be an interesting patch for several platforms. We are right in
> the middle of wrapping up the 7.2 release, and the concern is in
> reviewing the code and ensuring that it has no side effects on other
> platforms.
>
> If this were understood to be a simple patch to get a platform running
> on 7.2, then it would go into the tree right now. Since it is likely
> more complicated than that (have you posted it yet?), it will certainly
> need more discussion to get in.

I posted it last night, but there seems to be big propagation delay. I only
saw my posting into PORTS few days after doing the post, at first I though
list is not working at all. Strangely, first posting into PATCHES went
through fast enough ;)

> I'm willing to go either way, and if someone were to verify a
> significant performance boost for, say, Linux with the patch enabled
> then it would likely get more immediate interest.

Take a look at that article:
http://www-106.ibm.com/developernetworks/linux/library/l-rt5

With such a large difference in speed between posix & sysV semaphores,
there's bound to be some boost, although it might be mostly visible in
situations when lot of backends process requests at the same time. But on
platforms which don't have hardware TAS, like MIPS, this could be real
lifesaver.

- igor