Re: [Qemu-devel] [PATCH v5 09/15] hw: Added generic FIFO API.

[Peter Crosthwaite]

On Mon, Aug 6, 2012 at 7:48 PM, Peter Maydell <address@hidden> wrote:
> On 6 August 2012 03:16, Peter A. G. Crosthwaite
> <address@hidden> wrote:
>> Added a FIFO API that can be used to create and operate byte FIFOs.
>
> I'm not asking for actual conversions, but it would be nice to see a
> list of some devices that could in principle be moved to using this FIFO,
> as an indication of its general utility.
>

The two device models in this series. mcf_uart.c, strongarm.c
(StrongARMUARTState), xilinx_uartlite, cadence_uart.c, sh_serial.c,
tsc210x.c to name the ones Ive just had a brief look over. There are
more. grep -i fifo hw | grep uint8_t gives you a short list of files
to look for potential clients.

Dominated by uarts as they are the ones that need 8-bit fifos.
Usability is greatly expanded if the fifo can be an arbitrary data
type (uint32_t, uint16_t, struct foo), but that is a much harder
problem to solve that id like to say is out of scope of this series. I
think this is what Igor is getting at RE his PL330 comments on his
continuation of this thread. Can we get the ball rolling with uint8_t
then talk about generalisation strategy?

> Would it make sense for the FIFO to be a QOM object, or is that a
> silly idea?

There could be merit in making the data entries some form of QOM
object. It would hurt performance, but could solve the arbitrary data
problem.

Regards,
Peter

>
> -- PMM