Bootlin Wants to Bring Allwinner VPU Support to Mainline Linux (Crowdfunding)

I wrote about Free Electrons several times in the past due to their work on porting Arm SoCs to mainline Linux. Free Electrons is no more. But don’t be sad, as it’s not because they closed shop, but instead they changed their name to Bootlin due to trademark trolls.

Free Electrons Bootlin intensively worked on Allwinner processors, for example working with Next Things Co. to add CHIP board to mainline Linux, or more recently adding support for OpenGL ES in mainline Linux, albeit with closed-source user space binary blobs. The company normally work with partners for their work, but for their latest initiative aiming to bring Allwinner VPU (Video Processing Unit) to the official Linux kernel there’s asking funding from the community through a Kickstarter campaign.

For those who do not know, the VPU is the IP block within the SoC used for hardware video decoding of codecs such as MPEG2, MPEG4 or H264.Bootlin already showcased a demo with Cedrus open source VPU driver on C.H.I.P board in 2016, which supported and still only supports MPEG2 codec. Now as stated in their blog, they want to bring more complete support by working on the following if they can reach 17,600 Euros in funding:

  • Finish MPEG4 decoding support
  • Implementing H264 decoding
  • Optimize the rendering of video frames in cooperation with the display driver, and upstreaming the driver
  • Make sure the above works on Allwinner A10, A13, A20 and A33 SoC

They also have stretch goals:

  • 22,000 Euros – Support for newer SoCs such as such as the H3, H5 and A64
  • 30,800 Euros – H.265 video decoding support
  • 44,000 Euros – H.264 video encoding support

That’s the first time the company tries to fund upstream Linux kernel development through a crowdfunding campaign, so it’s unclear whether there will be interest. I’d assume it might be difficult to get it funded by individuals only, but if some companies are not satisfied with the closed-source Allwinner VPU driver that’s a good way to get open source software developed since costs are shared. Pledges start at 4 Euros, and go up to 1,024 Euros. The campaign will last for 45 days, and estimated delivery is scheduled for June 2018 for the main goals, and if all stretched goals are reached full development should be complete around December 2018. If the campaign is successful, a weekly progress report would be provided on the company website.

45
Leave a Reply

avatar
45 Comment threads
0 Thread replies
16 Followers
 
Most reacted comment
Hottest comment thread
18 Comment authors
willmoreJon SmirlTofChristianSander Recent comment authors
  Subscribe  
newest oldest most voted
Notify of
Daniel
Guest
Daniel

Pledged aswell, but I think it’s gonna be hard – Especially with the project description and the pretty similar pledges.

They should have asked someone who successfully did a kickstarter campain and followed their suggestions (even if it’s something like stickers at values below 99$, “having dinner” for pledges over 2000$ and so on).

theguyuk
Guest
theguyuk

The thing here is it is a shame all these companies can not pool knowledge, there are the folk working on Amlogic Linux, Orange Pi , Friendllyelec, iptv box sellers, tv box sellers ect All working on improving Arm in Linux

Harley
Guest
Harley

Which API do they plan on supporting? Video4Linux2 (V4L2) I hope!

Applications like Kodi are moving to only supporting Video4Linux2 (V4L2) on Linux in the future.

https://kodi.tv/article/fosdem-3-4-february-2018-brussels

mripard
Guest
mripard

@Harley
This is definitely our plan

Simos
Guest
Simos

@theguyuk

I am sure they would be interested.
I hope Bootlin makes an effort to approach them. Otherwise, it would be “volunteers” that may do an excellent job, or *uck it up.

Jerry
Guest
Jerry

Is this needed when RPi already provides free video with CEC?

willy
Guest
willy

theguyuk :
The thing here is it is a shame all these companies can not pool knowledge, there are the folk working on Amlogic Linux, Orange Pi , Friendllyelec, iptv box sellers, tv box sellers ect All working on improving Arm in Linux

In fact, when you see the amount they’re requesting, it’s lower than the cost of a single ad campaign for a company like ARM whose image would indirectly benefit from this.

BTW, I wasn’t aware that they had to change their name and feel sad for this. Their name was well thought and very well known in the Linux community, in part thanks to their lxr. Now they have to build a reputation again with a new name, that’s never easy. Good luck guys!

tkaiser
Guest
tkaiser

willy :
In fact, when you see the amount they’re requesting, it’s lower than the cost of a single ad campaign for a company like ARM whose image would indirectly benefit from this.

s/ARM/Allwinner/ 😉

But since the project will be funded within the next 24 hours we don’t need to worry anyway 🙂

tkaiser
Guest
tkaiser

@cnxsoft
Hehe, Olimex’ Tsvetan might be upset since Free Electrons* just recently added $2048 and $4096 pledge options and now Pine64 guys took the pole position (the 1st $4096 pledge is from them 🙂 )

*I refuse to call the great Free Electron guys Bootlin since it’s so annoying that those ISP trademark trolls are able to prevent others using the word ‘free’ 🙁

Mario
Guest
Mario

Backed 🙂

theguyuk
Guest
theguyuk

I think they should offer company badges of honour.

Ie the companies, large donation makers name or Logo on the page.

dave
Guest
dave

they like several processors they will work on but doesn’t include the new Allwinner H6 chip. I hope they work on that chip too.

blu
Guest
blu

How the heck does one trademark ‘free electrons’?

Anyhow, great to hear about this initiative. /goes to check the kickstarter page

blu
Guest
blu

@cnxsoft
That’s both irrational and said to hear. I guess no more FREE TICKETS in this world either. /s

willy
Guest
willy

s/ARM/Allwinner/

No, I was really meaning ARM. While omnipresent in smartphones, SoCs using their technology are widely known for always suffering from incomplete support in Linux, making it hard to build laptops or PCs for example, and often limiting them to short-lived devices never receiving any updates due to the SoC vendor not caring. This has improved a lot over the last years, in part thanks to Free Electrons and a few other companies, but there is still room for a lot of progress. We all know that whenever we see an ARM board, the first questions that comes to mind is “will it work out of the box, does it support linux at all, and will I be limited to the vendor’s BSP or not”. You never have to wonder this when you get, say, an Atom-based board. That’s why I’m saying that improving ARM-based SoCs support in general will definitely improve ARM’s image. And that’s really inexpensive.

roel
Guest
roel

They reached the 22000€!

roel
Guest
roel

@cnxsoft
Ah probably it was in $ and I was confused as when I pledged it was in €.

Sander
Guest
Sander

Although I do not like closed source hardware, I made a pledge, just to promote Linux kernel development.

theguyuk
Guest
theguyuk

You do need to consider the currency you are viewing it in

£16,028
pledged of £15,532 goal
133
backers
42 days to go

They hit the first target! 🙂

Now if only every reader of CNX Software pledged £1, $1,€1 or more wonder how much they would reach?

tkaiser
Guest
tkaiser

First stretch goal reached. FriendlyELEC joined the club in the meantime and chose one of the large ‘company pledges’ but top donations so far are from Pine64, Libre Computer and Xunlong / Orange Pi. Let’s hope the other two Stretch goals are reached soon.

Sander
Guest
Sander

@tkaiser

FriendlyELEC pledged? Cool! My money to FriendlyElec at work towards mainstream Linux … that’s how I like it.

tkaiser
Guest
tkaiser

@cnxsoft
I’m understanding the stretch goals differently than described above:

But the Kickstarter reads ‘Requires 8800 EUR in addition to our main goal and the first stretch goal’ and ‘Requires 13200 EUR in addition to our main goal and the first two stretch goals’ so it should be 22,000€, 30,800€ and 44,000€ instead?

Since now the main Allwinner board makers pledged larger amounts (though no idea what the Banana folks did or do) it might become a bit hard to reach the stretch goals especially if the amount needed is communicated as being lower than in reality?

And maybe it would be a good idea to exchange the remaining stretch goals since I think working and usable h.264 video encoding is a lot more important than HEVC decoding (see declining support for patent troll business models, dropped support for HEVC in products, Alliance for Open Media and finally AV1)

tkaiser
Guest
tkaiser

cnxsoft :
Yes, you’re right. I read that too fast on Friday.

Applies to me as well (and everybody else I was talking about the Kickstarter too). Would be good if the Kickstarter itself could be updated to clarify this.

Wrt additional HEVC decoding vs. H.264 encoding it’s just about a totally unrelated use case: being able to use all those cheap Allwinner boards with camera interface as such vs. ‘another codec’ that lost a lot of support recently in use cases that are relevant for SBC. But I might be wrong or biased (since good camera support in mainline kernel would finally allow to never ever buy Raspberries again 🙂 )

Christian
Guest
Christian

@tkaiser

tkaiser :
And maybe it would be a good idea to exchange the remaining stretch goals since I think working and usable h.264 video encoding is a lot more important than HEVC decoding (see declining support for patent troll business models, dropped support for HEVC in products, Alliance for Open Media and finally AV1)

While I appreciate your work a lot, I have to disagree here. Germany uses HEVC for DVB-T2, other European countries will follow. It’s also the codec for 4K-BD. Many Allwinner SBCs are the solution with the best price/performance ratio to get a Kodi device with HEVC support. I use a Orange Pi PC as such and eagerly wait to get that on mainline with all it’s advances.

While AV1 with Apple supporting it recently will be the future, it’s not even finished and there will be no chips supporting it before 2H 2019 the earliest.

theguyuk
Guest
theguyuk

@Christian
It is all to do with the game of life maths principles, life cycles growth, predation, decline then rebirth. It is not the end of the world, we are just in the mature growth phase. Eventually new birth ( new technology , hardware or software will come ). As CNX said H266 is already being worked on.

tkaiser
Guest
tkaiser

@Christian
Interesting. I didn’t thought Allwinner SoCs not being capable of 10-bit HEVC would be suitable for this use case. And given that stretch goals are expected to be delivered in Dec 2018 (only then ‘providing a user-space library easy to integrate in the popular open-source video players’ asking for some more work somewhere else?) I did not thought this would’ve much practical impact anyway given capabilities and progress with Amlogic/Rockchip SoCs…

Tof
Guest
Tof

What I found really sad, is that a billion dollars company (in capitalization) like Allwinner does not fund this kind of project directly.
People have to go through a Kickstarter campaigns which results will indirectly benefit in more Allwinner sales.

tkaiser
Guest
tkaiser

Tof :
What I found really sad, is that a billion dollars company (in capitalization) like Allwinner does not fund this kind of project directly.

Why should they? If they would consider missing Linux mainline kernel support for their SoCs a serious problem they would better follow the direction Rockchip, Amlogic or even MediaTek went: become part of the mainlining process and let their interns submit their code upstream so it becomes part of the official kernel when the hardware is still somewhat recent. Such Kickstarters (and the reverse engineering efforts as well as the massive delays between product launch and advanced features arriving upstream) shouldn’t be necessary in the first place:)

Member

I think we’ve seen the end of the h.26x series. The end was brought about by the excessive patent royalty demands of the various groups. h.265 is likely going nowhere in the broad market.

AV1 is well on it way to replacing the h.26x stuff.
http://www.streamingmedia.com/Articles/Editorial/Featured-Articles/AV1-A-Status-Update-120214.aspx

AV1 has the backing of Google, Netflix, Amazon, Hulu, Apple, Intel, etc. via the Alliance for Open Media. Google is also promising free hardware IP once the spec stabilizes. A stable spec should be happening right around now.

Member

This image from the MPEG chairman’s post shows the break down in h.265. There are three patent pools, each demanding separate royalties. Plus there are 15 more companies that have not declared which pool they are joining. There is also a pack of NPE/trolls in the mix too.

http://blog.chiariglione.org/wp-content/uploads/2018/01/HEVC.jpg

The problem with h.265 is that everyone is demanding at least a 100 to 1 return on the R&D dollars they spent. And of course they expect those billions in royalties to come from the pockets of YouTube, Netflix, Amazon Video, Hulu, etc. The content providers have responded by developing their own codec, AV1, for far, far less investment than the billions in royalties demands.

AV1 has been inspected by many patent lawyers trying to insure it does not infringe. But, as we all know, the company with h.265 IP will attack AV1 no matter what.

h.265 was recently given a huge gift by our much hated FCC chairman. Pai unilaterally approved ATSC 3 and authorized the shutdown of broadcast ATSC 1 about three years from now. By doing this every TV set sold in America is going to have to have an h.265 codec and all of the broadcasters are going to get stuck with large h.265 encoding fees. Note that when ATSC 1.0 initially came out each TV set paid around $27 in patent royalties. The total royalty demands for use of h.265 in ATSC 3.0 are unknown.

Christian
Guest
Christian

tkaiser :
@Christian
Interesting. I didn’t thought Allwinner SoCs not being capable of 10-bit HEVC would be suitable for this use case. And given that stretch goals are expected to be delivered in Dec 2018 (only then ‘providing a user-space library easy to integrate in the popular open-source video players’ asking for some more work somewhere else?) I did not thought this would’ve much practical impact anyway given capabilities and progress with Amlogic/Rockchip SoCs…

The German DVB-T2 HD standard uses only 8-bit, so it’s no problem 🙂 I know that Amlogic/Rockchip SoCs might be better for this, as e.g. they supported 4K at that time, but I personally don’t need 4k streaming and that 1080p 8bit will be the best German public television will offer at least for the next 5-7 years, so I think the Orange Pi PC offers the best best price/performance ratio for that use case.

Tof
Guest
Tof

tkaiser :
Why should they? If they would consider missing Linux mainline kernel support for their SoCs a serious problem they would better follow the direction Rockchip, Amlogic or even MediaTek went: become part of the mainlining process and let their interns submit their code upstream so it becomes part of the official kernel when the hardware is still somewhat recent. Such Kickstarters (and the reverse engineering efforts as well as the massive delays between product launch and advanced features arriving upstream) shouldn’t be necessary in the first place:)

I agree with you. And for that reason (the attitude of this company towards Linux support) for me Allwinner boards are a no go.
I would be really tempted by some board from OrangePi or FriendlyArm, but as long as it is powered by Allwinner chips, I won’t buy them.

Note: I recently converted an Amlogic TV Box to a debian server and I’m really thankful to Armbian efforts and the community.

Jerry
Guest
Jerry

@Tof
There’s always Broadcom & RPi 3. With CEC support, which is quite crucial for HTPC use.

willmore
Guest
willmore

@Jerry
He’s the H.265 decoding on the Rpi3? Do they support 4k?

tkaiser
Guest
tkaiser

@willmore
Not only this (2011 design lacking all recent features). According to this HTPC overview from FOSDEM at least within KODI Broadcom/RPi is still the most proprietary platform of all (starting at slide 25): https://fosdem.org/2018/schedule/event/kodi/attachments/slides/2166/export/events/attachments/kodi/slides/2166/FOSDEM_Presentation_2018___Lukas_Rusak.pdf

But maybe I misunderstood something, HTPC is nothing I’m interested in (but please give me h.264 encoding with mainline on Allwinner to never ever buy a Raspberry Pi again 😉 )

willmore
Guest
willmore

@tkaiser
That’s not surprising. Raspberry apologists like to overlook their platforms warts and focus on perceived warts of others.

Maybe it’s a case of “if all you have is a hammer, everything looks like nails”. Well, the whole world isn’t nails, there’s other things out there. Use the right tool for the job. If you consider that, then the Rpi family are rarely the right solution.

The two things the Rpi has going for it are:
1) There’s a lot of premade things using it for people who can’t solve problems themselves
2) If you have a reasonable sized city nearby, you can probably find one in a store.

Of course, you should outgrow #1 and #2 is solved by this little thing called ‘mail’.

theguyuk
Guest
theguyuk

@Jon Smirl

There is also as I stated on another thread, the fact of, what the Media is film on and edited in. Not only major companies make content any more and low end equipment often still uses H264, 4K Hevc.
Youtube use to convert to H264 on receiving a file ( if it is not already H264 ) , then into VP9 ( not sure if they are still doing that ).
Also a major market of online content is re-broadcast IPTV or KODI non official add-ons.