[dpdk-dev] [PATCH v4 5/8] baseband/la12xx: add queue and modem config support

[Chautru, Nicolas]

> -----Original Message-----
> From: Nipun Gupta <nipun.gupta at nxp.com>
> Sent: Monday, May 17, 2021 10:00 AM
> 
> Hi,
> 
> > -----Original Message-----
> > From: Chautru, Nicolas <nicolas.chautru at intel.com>
> > Sent: Thursday, May 13, 2021 8:21 PM
> > To: Nipun Gupta <nipun.gupta at nxp.com>; Hemant Agrawal
> > <hemant.agrawal at nxp.com>; dev at dpdk.org; gakhil at marvell.com
> > Cc: david.marchand at redhat.com
> > Subject: RE: [PATCH v4 5/8] baseband/la12xx: add queue and modem
> > config support
> >
> > Hi Nipun,
> >
> > > -----Original Message-----
> > > From: Nipun Gupta <nipun.gupta at nxp.com>
> > > Sent: Thursday, May 13, 2021 4:02 AM
> > > To: Chautru, Nicolas <nicolas.chautru at intel.com>; Hemant Agrawal
> > > <hemant.agrawal at nxp.com>; dev at dpdk.org; gakhil at marvell.com
> > > Cc: david.marchand at redhat.com
> > > Subject: RE: [PATCH v4 5/8] baseband/la12xx: add queue and modem
> > > config support
> > >
> > >
> > >
> > > > > +		.type   = RTE_BBDEV_OP_LDPC_DEC,
> > > > > +		.cap.ldpc_dec = {
> > > > > +			.capability_flags =
> > > > > +
> 	RTE_BBDEV_LDPC_CRC_TYPE_24A_CHECK |
> > > > > +
> 	RTE_BBDEV_LDPC_LLR_COMPRESSION |
> > > >
> > > > Are you sure you support the 6bits LLR compression option? I don't
> > > > see it being used in the PMD below. Nor in doc.
> > > > Let me know if you are unclear what that flag really mean.
> > >
> > > We support LLR compression in the Hardware by default.
> > > We do not need to call ' ldpc_input_llr_scaling' API in testbbdev.
> > >
> >
> > So you always assume to pack up the input data as 6 bits back-to-back?
> > You understand what that feature mean?
> > My concern is that you would not support current API and many unit
> > test would fail then for you PMD.
> > Are you able to run the bbdev-test and see failures (all tests should
> > be either skipped : missing capability) or passing
> 
> Our test vectors pass with our driver (Performance, validation, Latency and
> offload cost tests), provided we do not call 'ldpc_input_llr_scaling' API.
> I am not sure why this API is being called in testbbdev. Can you provide more
> insight on why this API is doing and why does it modifies the input data?

What LLR numerical representation are you doing for your PMD/HW.
Ie how many integer bits, how many decimals bits. 
This is to be able to support seamless different decoder implementation which have different assumptions or implementation. 

> 
> >
> >
> > > Regards,
> > > Nipun