Stream benchmark

From: Daniel Loyd Paull (paull-dl@ee.uwa.edu.au)
Date: Tue Aug 04 1998 - 04:55:59 CDT


John,

In case no-one else has commented on using Stream under Linux, I can
assure you that it compiled out of the box.

As this was done on an Intel based system, I ran the DOS version to
confirm the results (included below). Note: I bumped up the size of the
array under Linux to get more than 20 clock ticks per test.

Copiled with:
        $ cc -o stream stream_d.c second_cpu.c -lm

The system:
        P120
        PC Partner Mother Board (I dunno what release... lost the manual)
        64MB EDORAM (60ns I think)
        256KB Cache

As you can see, the results are almost identical.

Cheers,

Daniel Paull
4th Year BE(IT)
University of Western Australia

The DOS version output:

STREAM for DOS v2 by Dennis Lee
===============================
1 MB = 1000000 Bytes in the following measurements.

For accurate results, this benchmark should be executed
in a true DOS session, and not a DOS shell under another OS.

   Time Operation Mem Speed Error
   ---- --------- --------- -----
  7.41 sec COPY32 86.37 MB/s 0.8%
  7.42 sec COPY64 86.25 MB/s 0.8%
  6.37 sec SCALE 100.47 MB/s 1.0%
  8.68 sec ADD 110.60 MB/s 0.7%
  8.73 sec TRIAD 109.97 MB/s 0.7%

These results are comparable with those on the STREAM website.
See <http://www.cs.virginia.edu/stream> for info on STREAM.

------------------------------------------------------------------------

When compiled under linux:

-------------------------------------------------------------
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 320000 microseconds.
   (= 32 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: 86.4865 0.3770 0.3700 0.3800
Scale: 100.0000 0.3210 0.3200 0.3300
Add: 111.6279 0.4390 0.4300 0.4400
Triad: 109.0909 0.4450 0.4400 0.4500

-----------------------------------------------------------------------



This archive was generated by hypermail 2b29 : Tue Apr 18 2000 - 05:23:07 CDT