Is there any good arm9 chip contains excellent float multiple performance?

    A full functional SOC is prefered, LCD controller, A/D, PS/2
keyboards & mouse, SDRAM, etc. Any suggestion?

Readon Shaw napisał(a):
>     A full functional SOC is prefered, LCD controller, A/D, PS/2
> keyboards & mouse, SDRAM, etc. Any suggestion?
> 

Have a look at cirrus ARM SOCs.

Maciek.

Sawik =E5=86=99=E9=81=93=EF=BC=9A

> Readon Shaw napisa=C5=82(a):
> >     A full functional SOC is prefered, LCD controller, A/D, PS/2
> > keyboards & mouse, SDRAM, etc. Any suggestion?
> >
>
> Have a look at cirrus ARM SOCs.
>
> Maciek.
The most important is the float multiple performance. i don't know
whether the ARM9 would satisfied.

"Sawik" <sawik_remove_this@o2.pl> skrev i meddelandet 
news:eelrb5$sj9$1@atlantis.news.tpi.pl...
> Readon Shaw napisal(a):
>>     A full functional SOC is prefered, LCD controller, A/D, PS/2
>> keyboards & mouse, SDRAM, etc. Any suggestion?
>>
>
> Have a look at cirrus ARM SOCs.
>
> Maciek.

If you can live with an ARM7, the Magic Coprocessor in the Atmel AT572D740 
Diopsys will run up to 1 GFlops.

-- 
Best Regards,
Ulf Samuelsson
This is intended to be my personal opinion which may,
or may not be shared by my employer Atmel Nordic AB 

Readon Shaw napisał(a):
> Sawik 写道：
> 
>> Readon Shaw napisał(a):
>>>     A full functional SOC is prefered, LCD controller, A/D, PS/2
>>> keyboards & mouse, SDRAM, etc. Any suggestion?
>>>
>> Have a look at cirrus ARM SOCs.
>>
>> Maciek.
> The most important is the float multiple performance. i don't know
> whether the ARM9 would satisfied.
> 

It has something they call "math crunch engine" attached to ARM9 core.
Will that be good enough for your purposes, well, that's something you 
will have to decide for yourself in the end. Multiplication/accumulation 
operations are supported. No division. (At least it's not mentioned 
anywhere)
If you find any useful google links regarding it's performance, I'll be 
glad to look it over myself as I am taking this chip into consideration 
for the new project myself. Performance/benchmark data for those chips 
seems to be hard to find on net.

Maciek.

Sawik wrote:
> Readon Shaw napisa=C5=82(a):
> > Sawik =E5=86=99=E9=81=93=EF=BC=9A
> >
> >> Readon Shaw napisa=C5=82(a):
> >>>     A full functional SOC is prefered, LCD controller, A/D, PS/2
> >>> keyboards & mouse, SDRAM, etc. Any suggestion?
> >>>
> >> Have a look at cirrus ARM SOCs.
> >>
> >> Maciek.
> > The most important is the float multiple performance. i don't know
> > whether the ARM9 would satisfied.
> >
>
> It has something they call "math crunch engine" attached to ARM9 core.
> Will that be good enough for your purposes, well, that's something you
> will have to decide for yourself in the end. Multiplication/accumulation
> operations are supported. No division. (At least it's not mentioned
> anywhere)
> If you find any useful google links regarding it's performance, I'll be
> glad to look it over myself as I am taking this chip into consideration
> for the new project myself. Performance/benchmark data for those chips
> seems to be hard to find on net.
>
> Maciek.
I have not find anything about performance comparison. there is only a
summary on debian's website.There are two hard fp accelerations techs -
vfp & math crunch engine.

Readon Shaw <xydarcher@163.com> wrote:
> I have not find anything about performance comparison. there is only a
> summary on debian's website.There are two hard fp accelerations techs -
> vfp & math crunch engine.

Actually there are more than that.

See http://wiki.debian.org/ArmEabiPort for details.

The current Debian port uses FPA which was the original floating point ISA
for ARM. However FPA is no longer supported and newer chips do not implement
the instructions.

The current floating point ISA is VFP - "Vector Floating Point". There is
currently work going on in Debian to transition the ARM port to a new ABI
(the EABI for ARM) amongst other things this allows mixing of hard VFP and
SoftVFP code.

The "Math Crunch Engine" is a thirdparty co-processor and not part of the
ARM architecture. Currently in Debian the only way to use it is by
programming in assembler. In the EABI port you can generate code for it via
passing "-mfpu=maverick" to GCC however you will be unable to link such code
against the standard GCC startup files or libraries as they use VFP+SoftVFP. 

The EABI port makes it possible to have multiple versions of the libraries
packaged for example it would be possible to have libc6-vfp and
libc6-maverick however I'm not such if anyone is working on such packages.

-p
-- 
"Unix is user friendly, it's just picky about who its friends are."
 - Anonymous
--------------------------------------------------------------------