ChanServ changed the topic of #linux-sunxi to: Allwinner/sunxi development - Did you try looking at our wiki? https://linux-sunxi.org - Don't ask to ask. Just ask and wait for an answer! - This channel is logged at https://oftc.irclog.whitequark.org/linux-sunxi
Mangy_Dog has quit [Ping timeout: 480 seconds]
sunshavi_ has joined #linux-sunxi
sunshavi has quit [Ping timeout: 480 seconds]
montjoie has joined #linux-sunxi
montjoie_ has quit [Ping timeout: 480 seconds]
cmeerw has joined #linux-sunxi
cmeerw has quit [Ping timeout: 480 seconds]
hlauer has joined #linux-sunxi
hlauer has quit [Ping timeout: 480 seconds]
warpme_ has joined #linux-sunxi
<aperezdc> Hello! I'm new here, just got a Nezha board which has an All winner D1 (the RISC-V one)
<aperezdc> I was wondering if it would be welcome to add information about it to the wiki
<aperezdc> The page at https://linux-sunxi.org/D1 looks quite a bit like a stub
<apritzel> aperezdc: sure, go ahead, but create a new page for the *board*
<apritzel> aperezdc: there is some guidance on new device pages here: http://linux-sunxi.org/New_Device_page
<aperezdc> Ah, thanks for the tip about making a board page
<aperezdc> I'll rear through the new device page and start documenting the boars after work 👌
<apritzel> aperezdc: thanks!
<aperezdc> No problem at all
<aperezdc> s/rear/read
<aperezdc> s/boars/board
* aperezdc looks like they have fat fingers today
* apritzel is glad that aperezdc is not dealing with rears of boars after work ;-)
<aperezdc> XD
Mangy_Dog has joined #linux-sunxi
<gediz0x539> jernej: ndufresne: i see. now i've compiled gstreamer and all plugins from master branch. thanks a lot.
<gediz0x539> i still could not get a proper video play but maybe its about my test video. i'll give it a try with sintel trailer.
tnovotny has joined #linux-sunxi
<aperezdc> here goes an initial skeleton with the infobox practically complete, serial port and FEL sections; pictures later in the evening
<apritzel> aperezdc: thanks! make sure to keep in touch with smaeul and MoeIcenowy here, both have boards for a while and have done quite some work on them already
<apritzel> mmh, SPI NAND, does that chip take care of those nasty raw NAND sensitivities, like read and write disturb?
hanetzer1 has joined #linux-sunxi
hanetzer has quit [Ping timeout: 480 seconds]
hanetzer1 is now known as hanetzer
<aperezdc> I am not sure to be honest... I don't know much about flash/NAND memories, I'm more of a software guy
<aperezdc> but the preflashed system uses UBIFS, if knowing that helps
<aperezdc> And the uboot that came with the board has a sunxi_nand command
<aperezdc> If there is some way to check that I can do it later
<smaeul> fyi, the nezha SBC is made by Allwinner, not Sipeed. Sipeed is just one of a few distributors
<aperezdc> Ah, I got confused with that, in suppose renaming the page is fine
<apritzel> not sure renaming is easy, if you cannot do it, libv might be able to help out
<smaeul> apritzel: with just compatible = "spi-nand"; mainline cannot detect/read the UBI image used by the BSP, so that is a bit disappointing
choozy has joined #linux-sunxi
<gediz0x539> ndufresne: i've tried some example commands you've provided before and iiuc decoding works but kmssink wont.
<gediz0x539> when i terminate pipeline with fakesink, everything looks ok except some "using DMABuf without synchronization" and "Dropping frame due to QoS" warnings. they do not sound bad, i assume.
<gediz0x539> there's the output log: https://hastebin.com/jihuqapere.rb
<aperezdc> ...aaand moved, easier than expected: https://linux-sunxi.org/Allwinner_Nezha
<aperezdc> (easy once one notices that “move page” means ”rename”)
<gediz0x539> ndufresne: thanks. i will try it immediately.
<ndufresne> gediz0x539: I'm very suprised of "Using DMABuf without synchronization", but I doubt it's related
<ndufresne> for the drops, switch from fakesink to fakevideosink
<gediz0x539> this is what happens when i try to use kmssink btw: https://hastebin.com/nifijivira.yaml
<ndufresne> hmm, nevermind, you are using that
<gediz0x539> oh sorry, i've already used fakevideosink but used wrong naming there
<ndufresne> gediz0x539: as I tried to explained yesterday, you need videoconvert before kmssink
<ndufresne> good note to myself that I should try in the error log to differentiate the case the driver does not support it, vs that display (downstream) does not
<gediz0x539> oh I think I understand now. videoconvert will convert the decoded frames to the kmssink compatible format using CPU, right?
<gediz0x539> not actually frames, but, i think i kind of get that :)
<ndufresne> correct
<ndufresne> so until we have kmssink support for the tiled format used by this HW, videoconvert an compensate (slowly) in software
<ndufresne> but the problem I had last time, was testing on A20, was that everything was green, and didn't figure-out yet what was going on
<libv> aperezdc, apritzel: should the page be moved?
<libv> ah, done
<libv> ok :)
<aperezdc> I moved it already, yes
<libv> that was easy ;p
<libv> allwinner did an indiegogo?
<libv> oh wow.
<libv> they cannot be trusted with lgpl or gpl license, but we still trust them with crowdfunding?
<libv> humans.
<libv> heh, 111 backers for 12k, well, that's called a marketing campaign
<libv> and they are still pushing tina linux?
<aperezdc> yup 🤷‍♂️
<libv> chip makers pushing their own "OS" has always gone well
<libv> *cough* meego *cough*
<aperezdc> for me it was more of a “hey, I want to get into RISC-V, this looks like a small nice thing I can put on my desk for a price inside my budget”
<aperezdc> so I did the Indiegogo thing
<libv> sure, and here you are adding a cool page to our wiki and going through the motions, and do the actual work :)
<libv> which is cool, but is shows where allwinner is going wrong :)
<aperezdc> I'm more interested in bare metal than running Linux, but I also enjoy Linux, free softwaree and probably it's about time to contribute back a bit... I've used the sunxi wiki a lot :]
<aperezdc> used = used to read info and get things done at work
<libv> a wiki, when done right, is a big multiplier
<aperezdc> yeah, I agree that it would have been better that Allwinner would support mainline development themselves
<aperezdc> instead of pushing their own thing
<libv> aperezdc: i've marked you as a human being, so you can create your own user page
<aperezdc> yay!
* aperezdc has successfully proved not being a robot
<libv> at least close enough to have your own user page :)
<libv> aperezdc: example user page: https://linux-sunxi.org/User:Libv
<libv> contains like a hw list, and personal cheat sheet for things that should not be general
<libv> and the reason why we need to elevate to people is because of multi-year spamming we had or have
<aperezdc> I wrote down a super quick skeleton user page for myself
<aperezdc> no worries about having to verify, it's understandable :)
jernej_ has joined #linux-sunxi
jernej has quit [Remote host closed the connection]
ftg has quit [Ping timeout: 480 seconds]
<gediz0x539> ndufresne: i get green stripes like interlaced at startup of some sample videos too when i use playbin3
choozy has quit [Remote host closed the connection]
<libv> is it me, or is the D1 some sort of descendant of the a10/a20/r40?
<libv> it claims to have g2d?
<aperezdc> yes, there is G2D
<libv> and apparently, the fact that the SDK is open source is special...
<libv> ok.
<aperezdc> I haven't checked the a[124]0 pages for comparison, but from the devicetree it does look like it reuses quite a few components which already existed
<gediz0x539> libv: i've heard a rumour that there's also a Cortex-A7 core in the same die with the D1
<aperezdc> I haven't done through the registration to download the opensource sdk yet
<libv> if the SDK is available somewhere, then we will put it up on dl.
<libv> requiring registration is very much in the spirit of open source
<libv> gediz0x539: that would be rather weird
<libv> there is a tensilica dsp on it though
<aperezdc> registration was a reason why I am not downloading it yet :]
<libv> so for whatever reason, they did eschewed anything arm
<libv> so i would expect there to not be an a7 on it
<gediz0x539> yeah it sounds really strange
<libv> it seems allwinner has not really learned anything this past (almost) decade
<apritzel> libv: aperezdc: the D1 is apparently the same as the R329, which is one of the newer dual-A7 SoCs (just with those two cores disabled)
<aperezdc> interesting
<libv> an npu from arm china?
<apritzel> so yeah, of course they recycled all of their IP, in its current revision, so there are a lot of similarities (H616 also has G2D again, for instance)
<libv> oh, ok, i was not aware
faruk has quit []
<apritzel> but they also updated some stuff, so the pin controller is slightly different, for instance, but nothing is really RV specific
<libv> right, so it definitely fits amongst the rest of the sunxi support
gediz0x539 has quit [Quit: Leaving]
<apritzel> and omitting 3D might be just due to missing (closed-source) driver support for both Mali and PowerVR on TV, or they just don't have 3D in those simple SoCs in the first place (smart speaker market?)
<apritzel> s/on TV/on RISC-V/
* aperezdc just uploaded the front/back/uart pics for the Nezha
vagrantc has joined #linux-sunxi
<apritzel> aperezdc: can you identify and write down the labels on the chips?
<apritzel> The DRAM is readable on the pictures, but on the flash it's hard to read, and the one chip on the front (USB-C controller?) is completely unreadable
<aperezdc> apritzel: sure thing, give me a moment
<aperezdc> they are hard to read even by eye
<aperezdc> another day I can try and make better pics, too
<apritzel> well, that's not easy (especially in one shot for all chips, tried that myself many times), so either you do separate close ups, or just copy the inscription into the Wiki
<aperezdc> should this go in the board wiki page?
<aperezdc> of course I didn't take the metallic cover of the BT/Wifi chip, but I suppose it's not as interesting ;-)
<aperezdc> I gotta go now, more later (or tomorrow) :)
<apritzel> aperezdc: yes, onto the boards Wiki page, ideally into the info box (check other pages)
<aperezdc> okay, I'll search later another board as example and move the info
<apritzel> a PCF? For what? running out of GPIOs?
cmeerw has joined #linux-sunxi
tnovotny has quit []
<libv> apritzel: so what is the thing with sipeed and allwinner?
<libv> they are one and the same?
<apritzel> no clue, it seems like Sipeed is a distributor
<libv> ah, smaeul had that info
<apritzel> from what I know the chips are in short supply, so AW is not really selling them freely, but only ships them on their *own* boards (whatever that means)
<libv> curious
warpme_ has quit [Quit: Connection closed for inactivity]
warpme_ has joined #linux-sunxi
jernej_ is now known as jernej
Danct12 has quit [Remote host closed the connection]
Danct12 has joined #linux-sunxi
hlauer has joined #linux-sunxi
cmeerw has quit [Ping timeout: 480 seconds]
hlauer has quit [Ping timeout: 480 seconds]
<smaeul> apritzel: R329 is dual A53, and is not the same die as the D1, but is indeed quite similar. The Cortex-A7 variant of the D1 die is the R528, aka sun8iw20
<apritzel> smaeul: ah, thanks, and sorry for mixing that up
<smaeul> apritzel: my understanding is that ARM won't license Mali for a non-ARM-CPU SoC, so that leaves PowerVR as the main GPU option for RISC-V SoCs. So a lot of them have that or no GPU at all.
<apritzel> now looking at the numbering again it is obvious that 329 is 64-bit, while 528 is 32 ;-)
<smaeul> there is indeed a GPIO expander for some reason, probably needed due to complete absence of R_PIO and allocating a bunch of pins to LCD/LVDS
<apritzel> smaeul: I don't know about licensing details, but it seems plausible, and I think you typically buy bundles of ARM IPs
<apritzel> and if you think about Allwinner, then the lack of a binary Mali driver compiled for RISC-V might already be a showstopper for them
<smaeul> then thankfully the lack of a binary PowerVR driver might have stopped them from putting one of those in ;)
<swiftgeek> apritzel: there are tons of misc/generic IP from ARM, and having something tied to CPU arch sounds at the very least fishy
<swiftgeek> if anything i would expect them not willing to prepare blobs for RISC-V
<swiftgeek> but the soft/hard IPs from them shouldn't have any inherent limitations
<apritzel> swiftgeek: not tying, but bundling, so you get a GIC suitable for your Cortex core, for instance
<swiftgeek> that could still come from somewhere else
<apritzel> swiftgeek: and again: I have no clue how the sales people work, so don't read anything into my statements
<swiftgeek> and then licenses are usually kinda perpetual with arm lately aren't they?
<swiftgeek> lately as in since cortex
<apritzel> swiftgeek: you can of course mix and match, but you save yourself a lot of trouble if you don't try to slap a GICv2 on a 16 core system ;-)
<swiftgeek> apritzel: and there are more bits than that, that arm still can provide, like sram compilers
<swiftgeek> nearly everyone provides that, and so does arm
<swiftgeek> apritzel: i was more thinking about alternative vendor
<swiftgeek> > As a SRAM Compiler Engineer at Facebook (...)
<swiftgeek> that's not what i expected to see in my search results
<apritzel> last time I checked upselling was totally legit and common, also many people are actually happy to get multiple things from one vendor ;-)
<swiftgeek> if i were to pick vendor with everything it would probably be synopsys, though it sucks that they don't have RISC-V offering yet
<apritzel> well, doesn't need to be *everything*, but it makes sense to buy your core IP in a tested combination, doesn't it?
<swiftgeek> that would put stronger preference on synopsys cores :P
<apritzel> they offer ARM cores?
<swiftgeek> i guess this is why they are so big, besides all those other companies they acquired in past decade
<swiftgeek> apritzel: they have more things tested in a combination at particular process
<swiftgeek> especially in arm ecosystems
<swiftgeek> cpu wise, i think they just have ARC cores
<swiftgeek> unless we count dw8051 too
<apritzel> indeed, but those don't really compete against, say a Cortex-A710?
<swiftgeek> yeah