nbench-byte-2.1 for RISC OS

This is a quick port of nbench-byte-2/1, i.e. BYTE's benchmarking tool to RISC OS. It is pretty much untouched from the original source.

Points to note :

It's 32 bit ready, and so needs SharedUnixLibrary 1.02 (from http://www.prowl.org/peter/system.zip)
Results from some tests are terrible, usually because of FP code.
It was compiled using gcc 2.95.4 20010319 (prerelease) [gccsdk20020112] with -O3. Future compiler improvements may improve results. Compiling with -O2 rather than -O3 may improve some results, and -msoft-float may improve FP speeds, but you pays your money you takes your chance.
It asks for a 2MB wimpslot - it will fail if you run it from the CLI with too small an allocation.
This source has additional features when compiled under linux. As you might expect, they don't apply to RISC OS.

Typical results (StrongARM 242.8MHz RiscPC with VRAM):

BBBBBB   YYY   Y  TTTTTTT  EEEEEEE
BBB   B  YYY   Y    TTT    EEE
BBB   B  YYY   Y    TTT    EEE
BBBBBB    YYY Y     TTT    EEEEEEE
BBB   B    YYY      TTT    EEE
BBB   B    YYY      TTT    EEE
BBBBBB     YYY      TTT    EEEEEEE


BYTEmark (tm) Native Mode Benchmark ver. 2 (10/95)
NUMERIC SORT        :  Iterations/sec.:         51.80  Index:   1.33
STRING SORT         :  Iterations/sec.:          5.55  Index:   2.48
BITFIELD            :  Iterations/sec.:   17438684.03  Index:   2.99
FP EMULATION        :  Iterations/sec.:          6.33  Index:   3.04
FOURIER             :  Iterations/sec.:         39.39  Index:   0.04
ASSIGNMENT          :  Iterations/sec.:          0.31  Index:   1.17
IDEA                :  Iterations/sec.:        215.23  Index:   3.29
HUFFMAN             :  Iterations/sec.:         45.58  Index:   1.26
NEURAL NET          :  Iterations/sec.:          0.03  Index:   0.04
LU DECOMPOSITION    :  Iterations/sec.:          0.81  Index:   0.04
==========================ORIGINAL BYTEMARK RESULTS==========================
INTEGER INDEX       : 2.038
FLOATING-POINT INDEX: 0.043
Baseline (MSDOS*)   : Pentium* 90, 256 KB L2-cache, Watcom* compiler 10.0
* Trademarks are property of their respective holder.

Initial comparison of speed indexes

Remember the index of 1 is a Pentium 90, 256 KB L2-cache, Watcom compiler 10.0. The test systems are as follows:

242.8MHz (Overclocked) StrongARM RiscPC
Virtual Acorn 1.3 Demo (PC Acorn A5000 emulator with JIT) running under a loaded Windows XP Athlon 1800XP in full screen mode (GF3MX200/256Mb DDR RAM).
600MHz Intel® XScale 80321 Iyonix
Duron system executable was compiled using gcc 2.96, and executed on a loaded system running X, just to make the results ultra useful (sarcasm).

Test
242.8MHz StrongARM RiscPC
Athlon 1800XP emulating ARM3 A5000
600MHz Intel® XScale 80321 Iyonix
1.1GHz AMD Duron SDRAM based PC

Numeric Sort
1.33
1.85
4.94
13.03

String Sort
2.48
2.82
5.09
31.00

Bitfield
2.99
3.26
6.09
28.38

FP Emulation
3.04
2.14
6.32
23.57

Fourier
0.04
0.01
0.10
11.46

Assignment
1.17
3.75
4.88
23.36

Idea
3.29
2.20
6.11
23.84

Huffman
1.27
0.43
2.62
15.29

Neural Net
0.04
0.01
0.09
20.46

LU Decomposition
0.04
0.01
0.10
27.34

Integer Index
2.038
2.005
5.066
21.748

FP Index
0.043
0.011
0.096
18.575

Interpret these results as you like.

I'm not sure why I did this. Please feel free to toy with it, the source etc as how you see fit.

Download nbench.zip including source code (234k)

Tony Howat
02/03/2003

Test	242.8MHz StrongARM RiscPC	Athlon 1800XP emulating ARM3 A5000	600MHz Intel® XScale 80321 Iyonix	1.1GHz AMD Duron SDRAM based PC
Numeric Sort	1.33	1.85	4.94	13.03
String Sort	2.48	2.82	5.09	31.00
Bitfield	2.99	3.26	6.09	28.38
FP Emulation	3.04	2.14	6.32	23.57
Fourier	0.04	0.01	0.10	11.46
Assignment	1.17	3.75	4.88	23.36
Idea	3.29	2.20	6.11	23.84
Huffman	1.27	0.43	2.62	15.29
Neural Net	0.04	0.01	0.09	20.46
LU Decomposition	0.04	0.01	0.10	27.34
Integer Index	2.038	2.005	5.066	21.748
FP Index	0.043	0.011	0.096	18.575