Обсуждение: Fw: bad performance on irix

----- Original Message -----
From: "Robert E. Bruccoleri" <bruc@stone.congenomics.com>
To: "Luis Alberto Amigo Navarro" <lamigo@atc.unican.es>
Sent: Monday, March 18, 2002 4:08 PM
Subject: Re: [HACKERS] bad performance on irix


> Dear Luis,
> >
> > Dear Bob:
> > I've removed ifdefs from s_lock.h trying if semop using was define
problem,
> > but it's still using semop
> > any suggest?
>
> No, I see the same compilation as you do with 7.2. It's using the
spinlocks
> for some locks, but semaphores for others. I don't know what to
> do next. Alas... --Bob
>
> +-----------------------------+------------------------------------+
> | Robert E. Bruccoleri, Ph.D. | email: bruc@acm.org                |
> | P.O. Box 314                | URL:   http://www.congen.com/~bruc |
> | Pennington, NJ 08534        |                                    |
> +-----------------------------+------------------------------------+
>

"Luis Alberto Amigo Navarro" <lamigo@atc.unican.es> forwards:
>> It's using the spinlocks
>> for some locks, but semaphores for others.

That doesn't make any sense to me.  For one thing, if HAS_TEST_AND_SET
is defined in the config header, the executable will just plain fail to
build if there's no tas implementation, because lmgr/spin.c won't be
compiled.  And I sure don't see how some of the locks might be
implemented one way and some the other.

Which ones do you think are being implemented as semaphores, and what's
your evidence?
        regards, tom lane