Sergei,
Queue and Buffer managers were required because the list management
functions were taken into the controller. That is covered as part of item 1 I
mentioned.
Regards
swami
> -----Original Message-----
> From: Sergei Shtylyov [mailto:[email protected]]
> Sent: Sunday, March 22, 2009 4:43 PM
> To: Subbrathnam, Swaminathan
> Cc: David Brownell; [email protected]; linux-
> [email protected]
> Subject: Re: [PATCH] DaVinci: add support for CPPI 4.1
>
> Hello.
>
> Subbrathnam, Swaminathan wrote:
>
> > CPPI4.1 is an evolution over CPPI3.0.
>
> I would call the 3.0-to-4.1 changes rather revolutionary, than
> evolutionary...
>
> > As I see it the following features evolved compared to 3.0
> >
> > 1. List management functions are handled by the DMA engine in 4.1
> compared to 3.0 were the SW (driver) managed the same.
> >
>
> True, descriptor queues are handled by the queue manager now.
>
> > 2. New performance modes (Generic RNDIS) got introduced.
> >
>
> Those new modes don't have anything to do with CPPI 4.1 itself --
> they're part of USB XFER DMA. I think TI could well have kept using CPPI
> 3.0 and just added the new modes...
>
> > Beyond that the general processing of the DMA descriptors and
> functionality are very similar between CPPI3.0 and 4.1.
> >
>
> There's nothing similar in the descriptor formats, there was no DMA
> controller, and neither queue nor buffer managers in CPPI 3.0.
>
> > regards
> > swami
> >
> > ________________________________________
> > From: [email protected] [davinci-
> [email protected]] On Behalf Of Sergei
> Shtylyov [[email protected]]
> > Sent: Friday, March 20, 2009 11:08 PM
> > To: David Brownell
> > Cc: [email protected]; linux-arm-
> [email protected]
> > Subject: Re: [PATCH] DaVinci: add support for CPPI 4.1
> >
> > David Brownell wrote:
> >
> >>>> If this is only for the MUSB driver, it might be better to
> >>>> stash this code in driver/usb/musb and generalize the
> >>>> current CPPI glue there.
> >>>>
> >>> Generalize what? There's no way CPPI 3.0 and 4.1 can be "generalized",
> >>> they're too unlike.
> >>>
> >> If they're "too unlike" then they can't usefully connected
> >> to the musb engine and yet both work ...
> >>
> >
> > The difference being just 3.0 vs 4.1 might suggest that 4.1 is some
> sort
> > of superset to 3.0... but this is just not so -- everything is so
> completely
> > different between these specs, that I'm not sure why TI kept the same
> name.
> >
>
> WBR, Sergei
>
>
_______________________________________________
Davinci-linux-open-source mailing list
[email protected]
http://linux.davincidsp.com/mailman/listinfo/davinci-linux-open-source
