[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <e67149df1e72ccdaf7e9f51a0b2f81ac6602bfa2.camel@collabora.com>
Date: Fri, 23 Dec 2022 20:20:10 -0300
From: Daniel Almeida <daniel.almeida@...labora.com>
To: Benjamin Gaignard <benjamin.gaignard@...labora.com>,
Nicolas Dufresne <nicolas@...fresne.ca>,
ezequiel@...guardiasur.com.ar, p.zabel@...gutronix.de,
mchehab@...nel.org, robh+dt@...nel.org,
krzysztof.kozlowski+dt@...aro.org, heiko@...ech.de,
nicolas.dufresne@...labora.co.uk
Cc: linux-media@...r.kernel.org, linux-rockchip@...ts.infradead.org,
devicetree@...r.kernel.org, linux-arm-kernel@...ts.infradead.org,
linux-kernel@...r.kernel.org, kernel@...labora.com
Subject: Re: [PATCH v1 7/9] media: verisilicon: Add Rockchip AV1 decoder
Ah, I was too quick on that answer :/
I see that they have REFS_PER_FRAME (7), TOTAL_REFS_PER_FRAME (8) _and_
NUM_REF_FRAMES (8), in which case it is redundant indeed. I will remove
that on v4.
-- Daniel
On Fri, 2022-12-23 at 20:15 -0300, Daniel Almeida wrote:
> > > + int cur_offset[V4L2_AV1_NUM_REF_FRAMES - 1];
> > > + int cur_roffset[V4L2_AV1_NUM_REF_FRAMES - 1];
> >
> > This looks like V4L2_AV1_REFS_PER_FRAME. Daniel, should be remove
> > this
> > V4L2_AV1_NUM_REF_FRAMES ? Its redundant with
> > V4L2_AV1_TOTAL_REFS_PER_FRAME ...
>
> Hi. These are different. NUM_REF_FRAMES is the size of the "DPB"
> while
> TOTAL_REFS_PER_FRAME is the maximum number of references a frame can
> use. It just so happens that in AV1 these two are close in absolute
> value (i.e. 7 vs 8).
>
> Using VP9 as a comparison, the DPB size is still 8, but
> REFS_PER_FRAME
> is 3 (meaning a frame can specificy LAST, GOLDEN and ALTREF values).
>
> As this is per spec and a mere convenience, I vote for keeping it.
>
> -- Daniel
Powered by blists - more mailing lists