[Sound-open-firmware] Static pipeline and dsp only use case

Daniel Baluta daniel.baluta at gmail.com
Mon Mar 23 11:44:50 CET 2020


On Wed, Mar 18, 2020 at 10:38 PM Liam Girdwood
<liam.r.girdwood at linux.intel.com> wrote:
>
> On Wed, 2020-03-18 at 10:08 +0000, Daniel Baluta wrote:
> > Hi Liam,
> >
> > We have now basic support for our i.MX8 boards in Qemu.
> >
> > Do you know if we can continue development/testing having
> > only the DSP side?
>
> Yes you could, the host side of qemu is really useful for integration
> with the driver (and driver is mature now). You could hard code your
> unit tests in the init sequence if needed.
>
> If you are working on processing algos then I would recommend the
> testbench, sinec it can be all debugged on host (qemu is slow and does
> not support xtensa SIMD).
>
> >
> > So far none of the i.MX8 have qemu support.
> >
> > Looking at src/schedule/task.c I noticed that we can create a static
> > pipeline
> > but how do we trigger processing?
> >
>
> The fuzzer could be used to load a pipeline, obviously you would not
> want it to mangle the IPC in this case.

Thanks, the fuzzer like interface/utility is what we are looking for.

>
> > Our goal for qemu right now is to complete the DSP part and  validate
> > some
> > processing algorithms.
>
> Best to validate any processing algos that use HiFi 2,3,4 on teh DSP HW
> after you are happy with infrastructure on qemu and testbench.


More information about the Sound-open-firmware mailing list