STREAM results on SGI Indy

From: Chris Schoeneman (crs@lightscape.com)
Date: Wed Mar 13 1996 - 13:44:04 CST


Bill, I saw your post on comp.sys.sgi.hardware. I downloaded
STREAM to try on my Indy R4400 100MHz (64MB, 1MB 2ndary, 16KB I,
16KB D) running Irix 5.3. I believe the Indy has a 400MB/sec
memory to CPU bus.

The speeds were disappointing, around 45MB/sec. Out of curiosity,
I used cachectl() to turn off caching on the a, b, and c arrays.
The results showed no significant differences between cached and
uncached memory. For the cache enabled tests I also tried explicitly
enabling the cache, but the results were the same. I compiled using
-O3 optimization.

Cache enabled:
-------------------------------------------------------------
This system uses 8 bytes per DOUBLE PRECISION word.
-------------------------------------------------------------
Array size = 1000000, Offset = 0
Total memory required = 22.9 MB.
Each test is run 10 times, but only
the *best* time for each is used.
Enabling cache.
-------------------------------------------------------------
Your clock granularity/precision appears to be 9999 microseconds.
Each test below will take on the order of 210000 microseconds.
   (= 21 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
Assignment: 45.7145 0.3764 0.3500 0.4600
Scaling : 41.0257 0.4276 0.3900 0.4900
Summing : 44.4444 0.6199 0.5400 0.7200
SAXPYing : 50.0000 0.5069 0.4800 0.5900

Cache disabled:
Function Rate (MB/s) RMS time Min time Max time
Assignment: 45.7144 0.3570 0.3500 0.3600
Scaling : 42.1054 0.3881 0.3800 0.4000
Summing : 43.6364 0.5600 0.5500 0.5700
SAXPYing : 48.9796 0.4960 0.4900 0.5000

Cheers,
-chris



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