[ELDK] Bogomips difference between 2.6.25 and 2.6.30
Philippe Villet
Philippe.Villet at jdsu.com
Mon Sep 7 11:52:07 CEST 2009
Dear Wolfgang,
If I force the timer clock to be CPU clock (By setting the 3 parameter to zero) :
ibm440ep_fixup_clocks(sysclk, 0, 0);
I get ;
Calibrating delay loop... 790.52 BogoMIPS (lpj=1581056)
In this case I confirm that :
timebase = clock = bogomips / 2
But I want to use the TMRCLK input clock (25MHz on my custom board), so I wrote :
ibm440ep_fixup_clocks(sysclk, 0, 25000000);
and got :
Calibrating delay loop... 49.92 BogoMIPS (lpj=99840)
I've checked the timebase_period_ns variable defined in devtree.c, generic_calibrate_decr function in time.c and everything seems correct.
Any idea?
Thanks.
Philippe
-----Message d'origine-----
De : Wolfgang Denk [mailto:wd at denx.de]
Envoyé : vendredi 4 septembre 2009 17:24
À : Philippe Villet
Cc : eldk at lists.denx.de
Objet : Re: [ELDK] Bogomips difference between 2.6.25 and 2.6.30
Dear "Philippe Villet",
In message <F9D1A9212590B640B744B50EF9E01686044E130C at ENIEXCH01.ds.jdsu.net> you wrote:
>
> On our custom board (based on AMCC 440EPx) , we have 592 Bogomips with a
> 2.6.25 kernel and only 50 bogomips with a 2.6.30 (moving from a ppc to
> powerpc architecture).
>
> Is there any reason for such a difference?????
Do you just see such a difference in the bogmips numbers, or also in
the real system performance?
> /proc/cpuinfo on 2.3.30
> processor : 0
> cpu : 440EPX
> clock : 396.200004MHz
> revision : 24.212 (pvr 2162 18d4)
> bogomips : 49.92
> timebase : 25000000
> platform : PowerPC 44x Platform
> model : jdsu,mts4000
> Memory : 255 MB
Where is the difference between clock and timebase coming from?
Normally we see timebase = clock = bogomips / 2.
I cannot confirm such behaviour from our systems; just tested out
current development kernel (v2.6.31-rc8-1360-g39b19e6) on two Sequoia
boards (an old one at 528 MHz and a more recent one at 666 MHz):
U-Boot 2009.06 (Jul 22 2009 - 15:21:57)
CPU: AMCC PowerPC 440EPx Rev. A at 528 MHz (PLB=132, OPB=66, EBC=66 MHz)
Security/Kasumi support
Bootstrap Option H - Boot ROM Location I2C (Addr 0x52)
Internal PCI arbiter enabled, PCI async ext clock used
32 kB I-Cache 32 kB D-Cache
Board: Sequoia - AMCC PPC440EPx Evaluation Board, Rev. F, PCI=66 MHz
...
## Booting kernel from Legacy Image at 00400000 ...
Image Name: Linux-2.6.31-rc8-01360-g39b19e6
Image Type: PowerPC Linux Kernel Image (gzip compressed)
Data Size: 1456635 Bytes = 1.4 MB
Load Address: 00000000
Entry Point: 00000000
Verifying Checksum ... OK
...
-bash-3.2# cat /proc/cpuinfo
processor : 0
cpu : 440EPX
clock : 528.000008MHz
revision : 24.208 (pvr 2162 18d0)
bogomips : 1056.00
timebase : 528000008
platform : PowerPC 44x Platform
model : amcc,sequoia
Memory : 255 MB
And:
CPU: AMCC PowerPC 440EPx Rev. A at 666.667 MHz (PLB=166, OPB=83, EBC=55 MHz)
Security/Kasumi support
Bootstrap Option H - Boot ROM Location I2C (Addr 0x52)
Internal PCI arbiter enabled, PCI async ext clock used
32 kB I-Cache 32 kB D-Cache
Board: Sequoia - AMCC PPC440EPx Evaluation Board, Rev. F, PCI=33 MHz
...
## Booting kernel from Legacy Image at 00400000 ...
Image Name: Linux-2.6.31-rc8-01360-g39b19e6
Image Type: PowerPC Linux Kernel Image (gzip compressed)
Data Size: 1456635 Bytes = 1.4 MB
Load Address: 00000000
Entry Point: 00000000
Verifying Checksum ... OK
...
-bash-3.2# cat /proc/cpuinfo
processor : 0
cpu : 440EPX
clock : 666.666670MHz
revision : 24.208 (pvr 2162 18d0)
bogomips : 1333.33
timebase : 666666670
platform : PowerPC 44x Platform
model : amcc,sequoia
Memory : 255 MB
Best regards,
Wolfgang Denk
--
DENX Software Engineering GmbH, MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
The price of curiosity is a terminal experience.
- Terry Pratchett, _The Dark Side of the Sun_
More information about the eldk
mailing list