On January 18, 2007 12:00 am, twhk2007 wrote:
> Hi Curtis and other members,
>
> >Curtis wrote
> >You may look into downloading the Cirrus Logic GCC compiler with
> >Maverick Crunch enabled. This could double the speed of your code.
> >The only problem is the TS Linux kernel does not support Maverick
> >Crunch. So it isn't initialized and you can't have more than one
> >task using it, as the Crunch registers are not preserved between
> >tasks. You will also have trouble linking to libraries that were
> >not compiled with it.
> >Some people have ported Linux 2.6 kernels that should support
> >Crunch.
>
> I am new to the GCC complier. From the GCC 4.1.1 manual
> (http://gcc.gnu.org/onlinedocs/), I can only see the -mcpu option
> for EP9312. How about EP9302 and other EP93xx ? Is there anyone used
> the hardware floating point for EP9302 ? Does the current GCC
> complier (or other version) support the Maverick floating point co-
> processor in EP9302 ?
you use "-mcpu=ep9312" to compile with Maverick Crunch on the EP9302. There is
no "-mcpu=ep9302" (as of yet).
I had nothing but trouble trying to compile the vanilla GCC to crosscompile
Crunch. You might be able to find patches to help, but they were not enough
for me.
Cirrus Logic has GCC with Maverick Crunch posted on there website (This is the
easiest way to get Crunch):
http://arm.cirrus.com/forum/viewtopic.php?t=575
You may notice some problems if you use GAS (GNU Assembler), see:
http://groups.google.com/group/comp.sys.arm/browse_thread/thread/658d1365ad695a21/f44e276022a9277b?lnk=gst&q=maverick&rnum=2#f44e276022a9277b
There is a lot of errata for Maverick Crunch, see the cirrus webpage for the
processor you have.
-Curtis.
>
> Many thanks !
> James
>
> --- In "twhk2007" <> wrote:
> >
> > Thanks for your advice !
> >
> > --- In Curtis Monroe <curtis@> wrote:
> > >
> > > On January 17, 2007 08:55 am, Christopher Friedt wrote:
> > > > Hi James,
> > > >
> > > > Neither of these boards will execute any code from flash as
> > opposed to
> > > > ram, that would hinder the processor drastically. I don't even
> > think the
> > > > processor supports program addressing in flash-based devices.
> > >
> > > Technically NOR flash can execute code in-place (XIP). See
> article:
> > > http://www.edn.com/article-partner/CA6257144.html
> > >
> > > Nand Flash cannot XIP. It works more like a hard drive loading
> and
> > storing
> > > sectors. Nand also need to correct errors that occur in each
> pages
> > (ECC), NOR
> > > doesn't.
> > >
> > > I'm not sure if the TS-7200 with NOR flash uses XIP. But that
> > should not
> > > really matter. The EP9302 (or EP9301 on earlier TS-7200 boards)
> > has enough
> > > cache to store most of your critical code. Its the speed of the
> > cache that
> > > will be the biggest factor.
> > >
> > > So Nor and Nand will only make a difference in the speed of
> > loading code to
> > > your cache. Once the code is in the cache it will go at top
> speed!
> > >
> > >
> > > >
> > > > You're looking at nearly the exact same execution time. There
> > would only
> > > > be a very slight difference if one board had 32 MB sdram and
> one
> > had 64
> > > > MB sdram, since a board w/ 64 MB sdram has a dis-contiguous
> ram
> > layout
> > > > and thus would require a very very slight overhead in page
> table
> > > > translation from virtual addressing -> physical addressing.
> > >
> > > The 32MB sdram is discontinuous too. The EP9302 stores data and
> > code in the
> > > cache indexed by its virtual address, not its physical
> addresses.
> > So once the
> > > code is in the cache it does not need address translation to get
> > fetched
> > > again. So address translation will not slow it down.
> > >
> > >
> > > >
> > > > ~/Chris
> > > >
> > > > twhk2007 wrote:
> > > > > Hello all !
> > > > >
> > > > > TS-7200 uses NOR flash while TS-7260 uses NAND flash. I know
> > that the
> > > > > read time for NOR flash is faster than that of NAND flash. I
> > will use
> > > > > the board to do a lot of floating point calculations. Could
> > you anyone
> > >
> > > You may look into downloading the Cirrus Logic GCC compiler with
> > Maverick
> > > Crunch enabled. This could double the speed of your code. The
> only
> > problem is
> > > the TS Linux kernel does not support Maverick Crunch. So it
> isn't
> > initialized
> > > and you can't have more than one task using it, as the Crunch
> > registers are
> > > not preserved between tasks. You will also have trouble linking
> to
> > libraries
> > > that were not compiled with it.
> > >
> > > Some people have ported Linux 2.6 kernels that should support
> > Crunch.
> > >
> > > -Curtis.
> > >
> > >
> > > > > on the execution speed for these two boards ? Some people
> > mention that
> > > > > the code in NAND flash will be transfered to SDRAM during
> > excution, so
> > > > > the excution speed of a NAND flash-based board can be faster
> > than NOR
> > > > > flash-based board. Is this true for TS-7260 ?
> > > > >
> > > > > Best regards,
> > > > > James
> > > > >
> > > > >
> > > >
> > >
> >
>
>
>
Yahoo! Groups Links
<*> To visit your group on the web, go to:
http://groups.yahoo.com/group/ts-7000/
<*> Your email settings:
Individual Email | Traditional
<*> To change settings online go to:
http://groups.yahoo.com/group/ts-7000/join
(Yahoo! ID required)
<*> To change settings via email:
<*> To unsubscribe from this group, send an email to:
<*> Your use of Yahoo! Groups is subject to:
http://docs.yahoo.com/info/terms/
|