Software:
nbench http://gentoo-portage.com/app-benchmarks/nbench
gcc 4.4.0
binutils 2.19.1
glibc 2.9_p20090420
CFLAGS="-O2 -march=loongson2f -mabi=n32 -pipe -mplt"
Result:
zhangle@2f ~/oc (n32) $ 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 : 351 : 9.00 : 2.96
STRING SORT : 31.375 : 14.02 : 2.17
BITFIELD : 4.6277e+07 : 7.94 : 1.66
FP EMULATION : 20.335 : 9.76 : 2.25
FOURIER : 5078.2 : 5.78 : 3.24
ASSIGNMENT : 6.8794 : 26.18 : 6.79
IDEA : 1627 : 24.88 : 7.39
HUFFMAN : 515.87 : 14.31 : 4.57
NEURAL NET : 6.95 : 11.16 : 4.70
LU DECOMPOSITION : 114.96 : 5.96 : 4.30
==========================ORIGINAL BYTEMARK RESULTS==========================
INTEGER INDEX : 13.711
FLOATING-POINT INDEX: 7.268
Baseline (MSDOS*) : Pentium* 90, 256 KB L2-cache, Watcom* compiler 10.0
==============================LINUX DATA BELOW===============================
CPU :
L2 Cache :
OS : Linux 2.6.30.1-r0bertz
C compiler : mips64el-unknown-linux-gnu-gcc
libc :
MEMORY INDEX : 2.902
INTEGER INDEX : 3.871
FLOATING-POINT INDEX: 4.031
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.
--
Zhang, Le
Gentoo/Loongson Developer
http://zhangle.is-a-geek.org
0260 C902 B8F8 6506 6586 2B90 BC51 C808 1E4E 2973
yajin Wrote:
> Hi all,
>
> Following is the result on my gdium:
>
[snip]
>
> On 7��14��, ����3ʱ34��, Zhang Le <r0be...@gentoo.org> wrote:
>> Hardware:
>> Loongson 2F Fuloong box, overclocked to 1 GHz
>>
I am not familiar with nbench. Can anybody give a brief introduction for what
the meaning of these performance number ?
--
Thanks
Coly
It'll be much easier for me to reply if you have kept these number in your
reply. ;)
TEST : Iterations/sec. : Old Index : New Index
: : Pentium 90* : AMD K6/233*
--------------------:------------------:-------------:------------
NUMERIC SORT : 361.6 : 9.27 : 3.05
I believe here 9.27 means what speed the current cpu is if the speed of
Pentium 90 is 1. The same applies to 3.05 and AMD K6/233.
--
Zhang, Le
Gentoo/Loongson Developer
Zhang Le Wrote:
> On 18:33 Tue 14 Jul , Coly Li wrote:
>> I am not familiar with nbench. Can anybody give a brief introduction for what
>> the meaning of these performance number ?
>
> It'll be much easier for me to reply if you have kept these number in your
> reply. ;)
>
> TEST : Iterations/sec. : Old Index : New Index
> : : Pentium 90* : AMD K6/233*
> --------------------:------------------:-------------:------------
> NUMERIC SORT : 361.6 : 9.27 : 3.05
>
> I believe here 9.27 means what speed the current cpu is if the speed of
> Pentium 90 is 1. The same applies to 3.05 and AMD K6/233.
>
So it means current performance number is 927% fast as pentium 90 and 305%
faster as AMD K6/233 ?
It seems the performance number is not bad, am I right ?
--
Thanks
Coly
Zhang Le Wrote:
> On 22:43 Tue 14 Jul , Coly Li wrote:
>> Zhang Le Wrote:
>>> On 18:33 Tue 14 Jul , Coly Li wrote:
>>>> I am not familiar with nbench. Can anybody give a brief introduction for what
>>>> the meaning of these performance number ?
>>> It'll be much easier for me to reply if you have kept these number in your
>>> reply. ;)
>>>
>>> TEST : Iterations/sec. : Old Index : New Index
>>> : : Pentium 90* : AMD K6/233*
>>> --------------------:------------------:-------------:------------
>>> NUMERIC SORT : 361.6 : 9.27 : 3.05
>>>
>>> I believe here 9.27 means what speed the current cpu is if the speed of
>>> Pentium 90 is 1. The same applies to 3.05 and AMD K6/233.
>>>
>> So it means current performance number is 927% fast as pentium 90 and 305%
>> faster as AMD K6/233 ?
>>
>> It seems the performance number is not bad, am I right ?
>
> Yes, at least this is how I see it.
>
Great. Thanks for your patient explaining :-)
Coly
You are welcome, ;)
And I just got another result. I just change "-O2" in previous CFLAGS to
"-static -funroll-loops -O2" (thanks to jamesr's reminder). And the result is
much better, while still not as good as yajin's in NUMERIC SORT, BITFIELD,
IDEA and HUFFMAN test.
zhangle@2f ~/linux (n32) $ nbench
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 : 348.6 : 8.94 : 2.94
STRING SORT : 41.08 : 18.36 : 2.84
BITFIELD : 4.5627e+07 : 7.83 : 1.63
FP EMULATION : 85.372 : 40.97 : 9.45
FOURIER : 5209.6 : 5.92 : 3.33
ASSIGNMENT : 10.175 : 38.72 : 10.04
IDEA : 1538.8 : 23.54 : 6.99
HUFFMAN : 543.9 : 15.08 : 4.82
NEURAL NET : 9.1488 : 14.70 : 6.18
LU DECOMPOSITION : 194.32 : 10.07 : 7.27
==========================ORIGINAL BYTEMARK RESULTS==========================
INTEGER INDEX : 18.433
FLOATING-POINT INDEX: 9.570
Baseline (MSDOS*) : Pentium* 90, 256 KB L2-cache, Watcom* compiler 10.0
==============================LINUX DATA BELOW===============================
CPU :
L2 Cache :
OS : Linux 2.6.30.1-r0bertz
C compiler : mips64el-unknown-linux-gnu-gcc
libc : libc-2.9.so
MEMORY INDEX : 3.600
INTEGER INDEX : 5.528
FLOATING-POINT INDEX: 5.308
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.
--
Sorry, should be "-static -funroll-loops -O3".
Hello,
How do you do that ? :)
Cheers
Erwan
There is a register of loongson for tuning the cpu frequency, seems
somebody have written a kernel module to do that job.
Regards,
Wu Zhangjin
---------------
BTW:
here is the register and relative fields for tuning the cpu frequency
from the loongson user manual:
CR80: core_config
2:0 freq_scale RW 3'b111 Frequency scale control
if you want to tune it to the second high frequency, this works for you:
(*(u32 *)((char *)CKSEG1ADDR(0x1fe00000) + (0x80))) |= 0x06;
in the latest linux-loongson-2.6.30.1, you can do it like this:
#include <loongson.h>
...
LOONGSON_CHIPCFG0 |= 0x06;
http://www.gentoo-cn.org/~zhangle/oc.tar.bz2
I get it from lemote's bbs. Someone else wrote it.
$ insmod oc.ko fsb=3
This will overclock 2F to 1G.
You can see the result in benchmark performance.
However, since bogomips in /proc/cpuinfo is calculated only once during startup,
so the value remains unchanged while the CPU's frequency is actually increased.
After reboot, you can see cpuclock's value changed on the second line of dmesg's
output, bogomips changed too.
Note, the increased frequency can survive a reboot, but not a power off. If you
shut it down, then start it, you will find the frequency become 800 MHz again.