Floating point on IBM PPC405 (266MHz)

Do you have a question? Post it now! No Registration Necessary

Translate This Thread From English to

Threaded View
Hi guys,

Is anyone out there using linux on a board with IBM PPC405 (266MHz) CPU?
Are there any serious performance problems with the floating point operations?

Best Regards,
Alexander

Re: Floating point on IBM PPC405 (266MHz)
On 18 Sep 2003 06:56:34 -0700, s snipped-for-privacy@prosyst.bg (Alexander Popov)

Quoted text here. Click to load it

Yes. PPC405GPr.


Yes, because the PPC405 has no floating point support.

Hope that helps
JB
--
Juergen Beisert        email: jbeisert _AT_ eurodsn _DOT_ de
EuroDesign embedded technologies GmbH  http://www.eurodsn.de

Re: Floating point on IBM PPC405 (266MHz)
On Fri, 19 Sep 2003 06:44:39 GMT,
Quoted text here. Click to load it

In any case, you should make sure you are using a proper toolchain that does
not generate floating point instructions (ie. implies a -fsoft-float). These
can be handled by the kernel's math emulation, but this will slow things
down even more.

Regards,
Marius.

--
Marius Groeger        SYSGO Real-Time Solutions GmbH    www.sysgo.de
Software Engineering  Embedded and Real-Time Software   www.osek.de
We've slightly trimmed the long signature. Click to see the full one.
Re: Floating point on IBM PPC405 (266MHz)
Quoted text here. Click to load it

We're using the PPC405, too.  I never got the kernel's floating point
emulation to work, so we're 100% soft-float.  I have a toolchain build
script at http://kegel.com/crosstool that handles that right.
No idea about performance, though; soft float probably sucks.
- Dan


Re: Floating point on IBM PPC405 (266MHz)
Quoted text here. Click to load it

Well, compared to kernel emulation it doesn't suck.  It would
suck to have to trap every time you needed to emulate a FP
instruction. :)

If somebody needs good FP performance on 4xx then they could buy
(in huge volume) a 405+FPU ASSP/CSSP like one company I've heard
of or get the announced 440EP (that has the FPU macro cell).

-Matt

Re: Floating point on IBM PPC405 (266MHz)
On Fri, 19 Sep 2003 06:44:39 GMT snipped-for-privacy@netscape.net (Juergen Beisert)

10x Juergen,

I'm asking because I have QNX (6.2.1a) running on this baby and it's extreamly
slow
with floating point stuff, so I was wondering if anyone could help me check the
performance on linux
( I'm trying to convice our client to switch to linux )...

I'm also dissapointed of the integer performance too... On a StrongARM 1110 (
206MHz ) with linux
the integer ops are 50% faster and the floating point is 150 TIMES faster!!!

Is anyone willing to run this test on PPC405 with linux (it's very simple. I'll
send you the
source.)?

Quoted text here. Click to load it

Best Regards,
--
Alexander Popov                  ProSyst Bulgaria Inc.
RTOS Leader                      48 Vladajska Str.
We've slightly trimmed the long signature. Click to see the full one.
Re: Floating point on IBM PPC405 (266MHz)
On Tue, 30 Sep 2003 10:48:43 +0300, Alexander Popov

Quoted text here. Click to load it
send you the
Quoted text here. Click to load it

Ok. Send me the code. I need some stress test...
--
Juergen Beisert        email: jbeisert _AT_ eurodsn _DOT_ de
EuroDesign embedded technologies GmbH  http://www.eurodsn.de

Site Timeline