RE: STREAM benchmark & MSCV++

From: James Van Buskirk (torsop@ix.netcom.com)
Date: Thu Sep 04 1997 - 17:33:42 CDT


----------
From: John McCalpin
Sent: Thursday, September 04, 1997 1:45 PM
To: James Van Buskirk
Subject: Re: STREAM benchmark & MSCV++

Sorry for the delay.....

I don't know much about PC compilers -- the issue is with the timers.

There is a binary of STREAM compiled for DOS that you can get from:

        http://www.ecf.utoronto.ca/~denlee/softbnch.htm

Good Luck!

Well, I don't think Dennis Lee's executable would have helped me much.
I did get the timing stuff to work, though.

Processor: 21164, 533 MHz (164LX motherboard)
OS: Windows NT 4.0
Compiler: MSVC++ (RISC Edition) 5.0
Mem/cache: 128 MB SDRAM/2MB
Compiler options:
/nologo /ML /Gt0 /W3 /GX /Ox /Ot /Oa /Og /Oi /D "WIN32" /D "NDEBUG" /D "_CONSOLE" /D "_MBCS" /Fp"AlphaRel/stream.pch" /YX /Fo"AlphaRel/" /Fd"AlphaRel/" /FD /QA21164A /QAOu16 /c

Output:
-------------------------------------------------------------
This system uses 8 bytes per DOUBLE PRECISION word.
-------------------------------------------------------------
Array size = 4000000, Offset = 0
Total memory required = 91.6 MB.
Each test is run 10 times, but only
the *best* time for each is used.
-------------------------------------------------------------
Your clock granularity/precision appears to be 6999 microseconds.
Each test below will take on the order of 171999 microseconds.
   (= 24 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: 241.5094 0.2733 0.2650 0.2820
Scale: 234.4322 0.2808 0.2730 0.3050
Add: 267.4095 0.3882 0.3590 0.4370
Triad: 256.0000 0.3879 0.3750 0.4220

Remark: This is slower than results for a similar system with no
L3 cache from your list. I suppose that's why DEC chose to report
results with no L3 cache.



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