RE: syscall improvement patch [0/12]

From: David Mosberger <davidm_at_napali.hpl.hp.com>
Date: 2005-03-25 09:25:51
>>>>> On Thu, 24 Mar 2005 14:07:42 -0800, "Luck, Tony" <tony.luck@intel.com> said:

  >> That's very strange.  The branch-prediction may or may not help,
  >> depending on your test program.  The srlz.d elimination
  >> definitely should help, though.  What test-program are you using?
  >> Also, are you using McKinley or Madison?

  Tony> Madison ... but it might be an oddball stepping (it is very
  Tony> old).  I have a home-grown test that reads ar.itc before/after
  Tony> the getpid() and saves the difference in an array.  The
  Tony> program repeats 1000 times, then prints the min, max and
  Tony> median number of cycles.

Oh, getpid() probably doesn't do anything in your case.  You can
verify with strace.  If you see only a single getpid() syscall, your
libc is caching the getpid return value (I don't think glibc _ought_
to be caching getpid(), but it does...).  Try getuid() instead.

	--david
-
To unsubscribe from this list: send the line "unsubscribe linux-ia64" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Received on Thu Mar 24 17:26:13 2005

This archive was generated by hypermail 2.1.8 : 2005-08-02 09:20:37 EST