[net2-wg] 600 pps, high throughput

[Jonathan Hui]

Yes, as Phil mentioned, the next iteration of the USART stack does
include support for SPI over DMA. It will be committed in the next day
or two.

--
Jonathan W. Hui
jwhui at cs.berkeley.edu
http://www.cs.berkeley.edu/~jwhui/

On 3/14/06, henri dubois-ferriere <henridf at gmail.com> wrote:
> Jonathan, do you have a working SPI over DMA  implementation under 2.x
> and if so are you planning to commit it?
>
> I'd be very interested to use it on tinynode: not so much to increase
> throughput as with the cc2420, but simply to stay within the very
> tight timing constraints to run the xe1205 radio at high bitrates.
>
> thanks
> henri
>
> On 14/03/06, Jonathan Hui <jwhui at cs.berkeley.edu> wrote:
> > On 3/13/06, Philip Levis <pal at cs.stanford.edu> wrote:
> > > As Joe's pointed out in the past, the best solution is to use DMA
> > > (assuming you have a free DMA channel). If you have CSMA enabled, none
> > > of this makes a big difference, though, as the CSMA latency is greater
> > > than the loading time. I think Jonathan was planning on checking in a
> > > DMA version at some point, based on the re-architecture of the MSP430
> > > USART system Vlado, he, and I came up with after the TTX. He's been
> > > doing some pretty serious performance tradeoff comparisons, so probably
> > > has some good thoughts on the matter.
> > >
> > > Jonathan?
> >
> > The tradeoffs are pretty obvious. As Phil pointed out, the expected
> > backoff time can far exceed any lost time by not using the DMA
> > controller. Utilizing DMA would actually be more useful to pull
> > packets out of the radio, but contention under CSMA often keeps
> > packets from coming in back to back. If backoffs are turned off,
> > utilizing DMA would definitely have a greater impact on overall packet
> > throughput (both transmitting and receiving).
> >
> > One issue is that the msp430 only provides three DMA channels, two of
> > which would be required for full DMA support on the SPI. While, the
> > DMA can be helpful for packet training optimizations it also limits
> > their use for other potentially crucial tasks as sampling the ADC.
> >
> > --
> > Jonathan W. Hui
> > jwhui at cs.berkeley.edu
> > http://www.cs.berkeley.edu/~jwhui/
> >
>