Lpc2000

Hi!
I recently studied the LPC MAC instruction set and suspect
that the 21xx would be very good at running many DSP routines.

Infact so efficent that it would make many low cost DSP's
quite obsolete! Knowing that most DSP dev tools are costly!

Have anyone any experience into this?
What do you think about the LPC's DSP kapability?

Reg
KD

----- Original Message ----- 

From: "karldalen" <karldalen@...>
To: <lpc2100@yahoogroups.com>
Sent: Wednesday, January 14, 2004 1:52 PM
Subject: [lpc2100] DSP'ing with the LPC21xx


> Hi!
> I recently studied the LPC MAC instruction set and suspect
> that the 21xx would be very good at running many DSP routines.
>
> Infact so efficent that it would make many low cost DSP's
> quite obsolete! Knowing that most DSP dev tools are costly!
>
> Have anyone any experience into this?
> What do you think about the LPC's DSP kapability?

I'm interested in trying this myself. The performance should be reasonable,
approaching 60 MMACs. You'll have to use assembler as the C compilers don't
support the MAC operation AFAIK.

It won't be anything like as good as the ADI Blackfin of course. I've got
the Blackfin EZKit - it offers up to 1.2 GMACs from its two ALUs operating
in parallel at 600 MHz!

Leon
--
Leon Heller, G1HSM
Email: aqzf13@...
My low-cost Philips LPC210x ARM development system:
http://www.geocities.com/leon_heller/lpc2104.html

Maybe not as good as ADI's DSPs, but probably
better than some low-end DSPs (for example Microchip's
rfPIC). There are some drawbacks, of course (no
hardware loops, lack of division, no special
addressing modes), yet I believe it should perform
quite well for medium complexity DSP algorithms.

    - 
--- Leon Heller <leon_heller@...> wrote:
> 
> ----- Original Message ----- 
> From: "karldalen" <karldalen@...>
> To: <lpc2100@yahoogroups.com>
> Sent: Wednesday, January 14, 2004 1:52 PM
> Subject: [lpc2100] DSP'ing with the LPC21xx
> 
> 
> > Hi!
> > I recently studied the LPC MAC instruction set and
> suspect
> > that the 21xx would be very good at running many
> DSP routines.
> >
> > Infact so efficent that it would make many low
> cost DSP's
> > quite obsolete! Knowing that most DSP dev tools
> are costly!
> >
> > Have anyone any experience into this?
> > What do you think about the LPC's DSP kapability?
> 
> I'm interested in trying this myself. The
> performance should be reasonable,
> approaching 60 MMACs. You'll have to use assembler
> as the C compilers don't
> support the MAC operation AFAIK.
> 
> It won't be anything like as good as the ADI
> Blackfin of course. I've got
> the Blackfin EZKit - it offers up to 1.2 GMACs from
> its two ALUs operating
> in parallel at 600 MHz!
> 
> Leon
> --
> Leon Heller, G1HSM
> Email: aqzf13@...
> My low-cost Philips LPC210x ARM development system:
> http://www.geocities.com/leon_heller/lpc2104.html
> 
>  
> 
> Yahoo! Groups Links
> 
> To visit your group on the web, go to:
>  http://groups.yahoo.com/group/lpc2100/
> 
> To unsubscribe from this group, send an email to:
>  lpc2100-unsubscribe@yahoogroups.com
> 
> Your use of Yahoo! Groups is subject to:
>  http://docs.yahoo.com/info/terms/ 
> 
> 


__________________________________
Do you Yahoo!?
Yahoo! Hotjobs: Enter the "Signing Bonus" Sweepstakes
http://hotjobs.sweepstakes.yahoo.com/signingbonus

I apologize for my mistake; rfPIC should be read
as dsPIC, of course :)

--- Bogdan Marinescu <a_bogdan_marinescu@...>
wrote:
>     Maybe not as good as ADI's DSPs, but probably
> better than some low-end DSPs (for example
> Microchip's
> rfPIC). There are some drawbacks, of course (no
> hardware loops, lack of division, no special
> addressing modes), yet I believe it should perform
> quite well for medium complexity DSP algorithms.
> 
>     - 
> --- Leon Heller <leon_heller@...> wrote:
> > 
> > ----- Original Message ----- 
> > From: "karldalen" <karldalen@...>
> > To: <lpc2100@yahoogroups.com>
> > Sent: Wednesday, January 14, 2004 1:52 PM
> > Subject: [lpc2100] DSP'ing with the LPC21xx
> > 
> > 
> > > Hi!
> > > I recently studied the LPC MAC instruction set
> and
> > suspect
> > > that the 21xx would be very good at running many
> > DSP routines.
> > >
> > > Infact so efficent that it would make many low
> > cost DSP's
> > > quite obsolete! Knowing that most DSP dev tools
> > are costly!
> > >
> > > Have anyone any experience into this?
> > > What do you think about the LPC's DSP
> kapability?
> > 
> > I'm interested in trying this myself. The
> > performance should be reasonable,
> > approaching 60 MMACs. You'll have to use assembler
> > as the C compilers don't
> > support the MAC operation AFAIK.
> > 
> > It won't be anything like as good as the ADI
> > Blackfin of course. I've got
> > the Blackfin EZKit - it offers up to 1.2 GMACs
> from
> > its two ALUs operating
> > in parallel at 600 MHz!
> > 
> > Leon
> > --
> > Leon Heller, G1HSM
> > Email: aqzf13@...
> > My low-cost Philips LPC210x ARM development
> system:
> > http://www.geocities.com/leon_heller/lpc2104.html
> > 
> >  
> > 
> > Yahoo! Groups Links
> > 
> > To visit your group on the web, go to:
> >  http://groups.yahoo.com/group/lpc2100/
> > 
> > To unsubscribe from this group, send an email to:
> >  lpc2100-unsubscribe@yahoogroups.com
> > 
> > Your use of Yahoo! Groups is subject to:
> >  http://docs.yahoo.com/info/terms/ 
> > 
> > 
> 
> 
> __________________________________
> Do you Yahoo!?
> Yahoo! Hotjobs: Enter the "Signing Bonus"
> Sweepstakes
> http://hotjobs.sweepstakes.yahoo.com/signingbonus
> 


__________________________________
Do you Yahoo!?
Yahoo! Hotjobs: Enter the "Signing Bonus" Sweepstakes
http://hotjobs.sweepstakes.yahoo.com/signingbonus

> > > I recently studied the LPC MAC instruction set
> > and
> > > suspect
> > > > that the 21xx would be very good at running many
> > > DSP routines.

I did a feasibility study on line-echo cancellation based on an LMS 
filter for a DECT phone handset back in '95 on ARM7TDMI. The echo 
cancellation functions were broadly compliant to ETS 300 175-2. I 
got fairly good results using C compiled with the then current ARM 
tools (SDT 2.0 I think), but had to go to assembler to minimise the 
data shuffling and work on 4 data terms at a time in parallel. The 
final code required just 4MHz for the LMS filter and another 1MHz 
total for two non-linear processing functions, assuming zero 
waitstate memory. Sorry I can't remember details of the filter I 
used, itwas a long time ago :)

Lack of division hardware is a pain, but if you're doing repeated 
divisions *by the same divisor*, there is a nice alternative using 
the 7TDMI's UMULL instruction to multiply by 0xFFFFFFFF/divisor and 
take the top word of the result. You do have to check the result by 
multiplying and add one if there has been a rounding error, but this 
is still much faster than a straight division.

Peter.