Re: COPY locking

Tom Lane wrote:
>
> John Coers <coers(at)intrinsity(dot)com> writes:
> > Ah so! So after each tuple is written to shared memory there is a giant
> > scramble for the lock. Explains all the semops and timer sets and sleeps
> > in the truss output that I've been seeing.
>
> Hmm ... semops? What platform are you on, anyway, and which PG version?
> It sounds like you might be running with the fallback SysV-semaphore-
> based implementation of spinlocks, which is orders of magnitude more
> horrid than even the normal test-and-set-based implementation. If there
> is TAS support for your machine then I'd only expect to see select(2)
> calls during spinlock operations. Does your version of src/include/os.h
> define HAS_TEST_AND_SET?
>
> regards, tom lane

Solaris 5.7, postgres 7.1

I see this in src/include/os.h:

#elif defined(__sparc__)
typedef unsigned char slock_t;

#define HAS_TEST_AND_SET

I've attached a little facsinating truss output. 7687 is the postmaster and the
rest are backends.

--
John Coers Intrinsity, Inc.
coers(at)intrinsity(dot)com Austin, Texas