Re: STREAM Result on 4x PPro200Mhz/512KCache, 128MB Mem

From: Pekka Hedqvist (pekka@erlang.ericsson.se)
Date: Tue Oct 22 1996 - 05:54:19 CDT


Problem solved. For Solaris to see the extra CPUs the APIC mode
moust be enabled in the BIOS. If not no extra CPUs. This mode
could be in two modes: "Fully Mapped" and "EISA".
We had "Fully Mapped", when changed to "EISA" I got the numbers
below which are more reasonable and within excpected range.
I checked the numbers with APIC non enabled (only one CPU seen
by OS) and the numbers were within 1% of the ones below.

Funny thing is that Linux 2.x (MP) needs APIC mode in "Fully
Mapped" to work properly. But I suppose this is the normal
situation the PC world..

Numbers for: Compaq Proliant 5000. 4x PPro 200Mhz w. 512K each,
128MB RAM, Solaris 2.5.1.

[~/work/stream]: gcc -O4 stream_d.c second_cpu.c -o str -lm
[~/work/stream]: ./str
-------------------------------------------------------------
This system uses 8 bytes per DOUBLE PRECISION word.
-------------------------------------------------------------
Array size = 2000000, Offset = 0
Total memory required = 45.8 MB.
Each test is run 10 times, but only
the *best* time for each is used.
-------------------------------------------------------------
Your clock granularity/precision appears to be 9999 microseconds.
Each test below will take on the order of 200000 microseconds.
   (= 20 clock ticks)
Increase the size of the arrays if this shows that
you are not getting at least 20 clock ticks per test.
-------------------------------------------------------------
WARNING -- The above is only a rough guideline.
For best results, please be sure you know the
precision of your system timer.
-------------------------------------------------------------
Function Rate (MB/s) RMS time Min time Max time
Copy: 123.0769 0.2620 0.2600 0.2700
Scale: 114.2857 0.2880 0.2800 0.2900
Add: 141.1765 0.3430 0.3400 0.3500
Triad: 126.3158 0.3800 0.3800 0.3800

>>>>> "John" == John McCalpin <mccalpin@frakir.asd.sgi.com> writes:

John> Do you have access to a Solaris compiler, or is gcc your only option?
John> What you have done certainly looks reasonable....

John> --
John> --
John> John D. McCalpin, Ph.D. Supercomputing Performance Analyst
John> Advanced Systems Division http://reality.sgi.com/employees/mccalpin
John> Silicon Graphics, Inc. mccalpin@asd.sgi.com 415-933-7407



This archive was generated by hypermail 2b29 : Tue Apr 18 2000 - 05:23:06 CDT