Subject: Re: Benchmarks
To: Ale Terlevich <A.I.Terlevich@durham.ac.uk>
From: Peter Berg <Peter.Berg@ww.tu-freiberg.de>
List: port-arm32
Date: 11/11/1996 12:50:33
>
>
> Did you use the same cc opts as me?
> ( OPTON = -O3 -pipe -fomit-frame-pointer -finline-functions)
>
> It seems quite amazing that some of the tests on the SA110 are almost
> half the speed, and how can the Dc: sqrt(2) to 99 decimal places be
> 2.5 times quicker on the Arm7? I would have assumed this to be a purely
> computational thing (Like Dhrystone and Arithmetic test) and so be
> quicker on the SA
>
> Also, I don't believe my results for FileRead. I'm running on the
> internal 420Mb IDE and 65MB per sec seems a bit much! Dunno why this
> should produce such weird results!
I think this is caused by the good file caching mechanism of the ffs system.
What you get seems now to be the transfer speed in the RAM, because the ffs
cache as much as possible.
The options I used were only -O.
I've done it again with your options and here is the result.
<fixed> BYTE UNIX Benchmarks (Version 3.11)
System -- NetBSD line9 1.1B NetBSD 1.1B (ATAPI) #4193: RiscPC 700
/ARM710-40Mhz
Start Benchmark Run: Mon Nov 11 11:10:56 MET 1996
1 interactive users.
Dhrystone 2 without register variables 43958.5 lps (10 secs, 6 samples)
Dhrystone 2 using register variables 43995.7 lps (10 secs, 6 samples)
Arithmetic Test (type = arithoh) 76275.0 lps (10 secs, 6 samples)
Arithmetic Test (type = register) 7442.2 lps (10 secs, 6 samples)
Arithmetic Test (type = short) 6689.7 lps (10 secs, 6 samples)
Arithmetic Test (type = int) 7443.5 lps (10 secs, 6 samples)
Arithmetic Test (type = long) 7443.3 lps (10 secs, 6 samples)
Arithmetic Test (type = float) 104.8 lps (10 secs, 6 samples)
Arithmetic Test (type = double) 107.1 lps (10 secs, 6 samples)
System Call Overhead Test 8345.4 lps (10 secs, 6 samples)
Pipe Throughput Test 1362.9 lps (10 secs, 6 samples)
Pipe-based Context Switching Test 644.2 lps (10 secs, 6 samples)
Process Creation Test 47.7 lps (10 secs, 6 samples)
Execl Throughput Test 52.5 lps (9 secs, 6 samples)
File Read (10 seconds) 10957.0 KBps (10 secs, 6 samples)
File Write (10 seconds) 600.0 KBps (10 secs, 6 samples)
File Copy (10 seconds) 396.0 KBps (10 secs, 6 samples)
File Read (30 seconds) 12267.0 KBps (30 secs, 6 samples)
File Write (30 seconds) 533.0 KBps (30 secs, 6 samples)
File Copy (30 seconds) 499.0 KBps (30 secs, 6 samples)
C Compiler Test 16.2 lpm (60 secs, 3 samples)
Shell scripts (1 concurrent) 32.6 lpm (60 secs, 3 samples)
Shell scripts (2 concurrent) 13.0 lpm (60 secs, 3 samples)
Shell scripts (4 concurrent) 6.0 lpm (60 secs, 3 samples)
Shell scripts (8 concurrent) 2.6 lpm (60 secs, 3 samples)
Dc: sqrt(2) to 99 decimal places 1254.1 lpm (70 secs, 6 samples)
Recursion Test--Tower of Hanoi 713.1 lps (10 secs, 6 samples)
INDEX VALUES
TEST BASELINE RESULT INDEX
Arithmetic Test (type = double) 2541.7 107.1 0.0
Dhrystone 2 without register variables 22366.3 43958.5 2.0
Execl Throughput Test 16.5 52.5 3.2
File Copy (30 seconds) 179.0 499.0 2.8
Pipe-based Context Switching Test 1318.5 644.2 0.5
Shell scripts (8 concurrent) 4.0 2.6 0.6
=========
SUM of 6 items 9.1
AVERAGE 1.5
</fixed>
It seems, everything has turned into a little bit better for my system with
this options then for yours ;)
Ciao, Peter
_____ _ ____ _____ _____
| __ \(_) | _ \ / ____| __ \
| |__) |_ ___ ___| |_) | (___ | | | |
| _ /| / __|/ __| _ < \___ \| | | |
| | \ \| \__ \ (__| |_) |____) | |__| |
|_| \_\_|___/\___|____/|_____/|_____/
E-Mail: Peter.Berg@ww.tu-freiberg.de
WWW : http://www.ww.tu-freiberg.de/~pberg