Re: [Linux-ia64] Average Bogomips for quad itanium 2?

From: David Mosberger <davidm_at_napali.hpl.hp.com>
Date: 2003-06-03 05:15:38
>>>>> On Mon, 2 Jun 2003 13:58:06 -0500 (CDT), "Brent M. Clements" <bclem@rice.edu> said:

  Brent> The reason I'm asking is because I am getting the following

  Brent> family     : Itanium 2
  Brent> model      : 0
  Brent> revision   : 7
  Brent> archrev    : 0
  Brent> features   : branchlong
  Brent> cpu number : 0
  Brent> cpu regs   : 4
  Brent> cpu MHz    : 900.000000
  Brent> itc MHz    : 900.000000
  Brent> BogoMIPS   : 1346.37

  Brent> Is this the normal bogomips for this processor or should I be worried?

This looks normal to me.  On McKinley, a "single-cycle loop" usually
take a bit more than a cycle to execute on average.  If we assume that
4 iterations execute in 3 cycles and consider that BogoMIPS counts
each iteration as 2 instructions (decrement & branch), you'd expect to
see 2*900e6/(4/3)=1350, which is close to what see.

Of course, you could unroll the loop, but there is a reason it's
called _Bogo_MIPS... ;-)

	--david
Received on Mon Jun 02 12:17:15 2003

This archive was generated by hypermail 2.1.8 : 2005-08-02 09:20:15 EST