Oops. I took a look at your page of results, and noticed that the
OMAP2420 didn't really perform all that well at floating point tests,
despite its having a hardware VFP (vector floating point) unit.
It then dawned on me that although I'd compiled nbench with hardware
fp, it had been linked against a soft-float libm.
If nothing else, comparing the differences between these new results
and the ones I previously posted shows how much of a difference the
floating point libm makes.
So here are the new results with everything hardware floating point
(I've also flashed an updated image since I posted the other results,
but that should make little difference):
Nokia-N800-26:~# ./nbench
BYTEmark* Native Mode Benchmark ver. 2 (10/95)
Index-split by Andrew D. Balsa (11/97)
Linux/Unix* port by Uwe F. Mayer (12/96,11/97)
TEST: Iterations/sec. : Old Index : New Index
: : Pentium 90* : AMD K6/233*
:--:-:
NUMERIC SORT: 116.56 : 2.99 : 0.98
STRING SORT : 6.5519 : 2.93 : 0.45
BITFIELD: 3.2551e+07 : 5.58 : 1.17
FP EMULATION: 13.204 : 6.34 : 1.46
FOURIER : 996.52 : 1.13 : 0.64
ASSIGNMENT : 1.0108 : 3.85 : 1.00
IDEA: 284.63 : 4.35 : 1.29
HUFFMAN : 119.99 : 3.33 : 1.06
NEURAL NET : 1.1723 : 1.88 : 0.79
LU DECOMPOSITION: 39.681 : 2.06 : 1.48
==ORIGINAL BYTEMARK RESULTS==
INTEGER INDEX : 4.029
FLOATING-POINT INDEX: 1.637
Baseline (MSDOS*) : Pentium* 90, 256 KB L2-cache, Watcom* compiler 10.0
==LINUX DATA BELOW===
CPU :
L2 Cache:
OS : Linux 2.6.18-omap1
C compiler : gcc version 3.4.4 (release) (CodeSourcery ARM 2005q3-2)
libc: ld-2.3.6.so
MEMORY INDEX: 0.808
INTEGER INDEX : 1.185
FLOATING-POINT INDEX: 0.908
Baseline (LINUX): AMD K6/233*, 512 KB L2-cache, gcc 2.7.2.3, libc-5.4.38
* Trademarks are property of their respective holder.
Cheers,
Simon
___
OpenMoko community mailing list
community@lists.openmoko.org
http://lists.openmoko.org/mailman/listinfo/community
