01-06-2016, 05:07 PM
(This post was last modified: 01-06-2016, 05:13 PM by paulieg.)
(01-06-2016, 04:42 PM)nomadewolf Wrote: I people want Linux with accelerated graphics, this board is a no go, no matter what! Unfortunately this is my case 
This is inaccurate given the information we have as of right now. If people want accelerated graphics and linux > 3.0.x *now*, it is a no go. Otherwise, with a bit of dev and elbow grease, it's got a decent chance of working. I accept that there may be some people who want accelerated graphics and new kernel features like containers, but I expect that number to be fairly low. Do you *need* a kernel > 3.0.x? Is it a showstopper for you? I'm always curious to know what other people want to do, so this is genuine interest rather than sarcasm.
Keep in mind that Allwinner claim to have a timeline for releasing newer kernels as they release Android updates. True, this is not as good as mainline, they aren't available immediately etc, but this is a fairly new SoC so we are guaranteed that it'll happen because no one is going to buy an SoC to make Android devices that can't run L or M.
Granted, some people say the Allwinner blobs may have issues given past performance. I can't argue since I don't have experience with them, but technically there is no reason they should given the same kernel+libc. In any case, we haven't tried them yet and so it is inaccurate to make any sort of definite statement about their usability. It's not as if open source software doesn't have stability issues that are impossible to track down, speaking as someone who at one point in his life spent half a year trying to make a particular, fully open source, filesystem implementation stable enough for production use and failed (not for lack of trying or competence).
I used the nvidia and amd blobs as an analogy in an argument about open source vs blobs. What I specifically did *not* do is equate the engineering complexity of their GPUs with Mali. You know why? Because that would've made me sound a complete idiot. Mali is to those GPUs as a Trabant is to a Veyron. As an aside, I run open source drivers for amd (my desktop) and nvidia (my laptop) exclusively and like them better, although I don't really do anything particularly demanding. I would certainly consider them good enough and that's a much more difficult problem than Mali. In any case, we will have blobs available and there's a decent chance they will work.
There are lots of unknowns and variables at play and I certainly share everyone's concerns about the open source side of things, but I see no factual and logical support for claiming that this board is a definite no go for any type of hardware acceleration at any time and under all possible circumstances.
-p
paulieg
(01-06-2016, 04:42 PM)nomadewolf Wrote: I people want Linux with accelerated graphics, this board is a no go, no matter what! Unfortunately this is my case
This is inaccurate given the information we have as of right now. If people want accelerated graphics and linux > 3.0.x *now*, it is a no go. Otherwise, with a bit of dev and elbow grease, it's got a decent chance of working. I accept that there may be some people who want accelerated graphics and new kernel features like containers, but I expect that number to be fairly low. Do you *need* a kernel > 3.0.x? Is it a showstopper for you? I'm always curious to know what other people want to do, so this is genuine interest rather than sarcasm.
Keep in mind that Allwinner claim to have a timeline for releasing newer kernels as they release Android updates. True, this is not as good as mainline, they aren't available immediately etc, but this is a fairly new SoC so we are guaranteed that it'll happen because no one is going to buy an SoC to make Android devices that can't run L or M.
Granted, some people say the Allwinner blobs may have issues given past performance. I can't argue since I don't have experience with them, but technically there is no reason they should given the same kernel+libc. In any case, we haven't tried them yet and so it is inaccurate to make any sort of definite statement about their usability. It's not as if open source software doesn't have stability issues that are impossible to track down, speaking as someone who at one point in his life spent half a year trying to make a particular, fully open source, filesystem implementation stable enough for production use and failed (not for lack of trying or competence).
I used the nvidia and amd blobs as an analogy in an argument about open source vs blobs. What I specifically did *not* do is equate the engineering complexity of their GPUs with Mali. You know why? Because that would've made me sound a complete idiot. Mali is to those GPUs as a Trabant is to a Veyron. As an aside, I run open source drivers for amd (my desktop) and nvidia (my laptop) exclusively and like them better, although I don't really do anything particularly demanding. I would certainly consider them good enough and that's a much more difficult problem than Mali. In any case, we will have blobs available and there's a decent chance they will work.
There are lots of unknowns and variables at play and I certainly share everyone's concerns about the open source side of things, but I see no factual and logical support for claiming that this board is a definite no go for any type of hardware acceleration at any time and under all possible circumstances.
-p
By 'a bit of dev and elbow grease' you're definitely being optimistic. There's absolute no guarantee on that department. While it is entirely possible, it doesn't mean anything else. We just need to wait and see.
And though the AMD/Nvidia GPUs are on a completely different level of complexity, at least on the AMD side there is documentation and cooperation and code contribution. And that is a BIG deal and a big difference. I would much rather have the schematics for a Veyron than redesign a Trabant from scratch.
As for using an older kernel, no i'm not so square that i wouldn't use a kernel 'just' beacause it is old. But with age come the bugs and more importantly security flaws. So yeah, it is a showstopper for me, on account of security concerns which is a serious issue, at least for me.
Even if Android didn't use the Linux kernel (which would make the whole thing much more unlikely), there would still be a possibility. Nothing is impossible if there is enough work involved. But we have to factor reality in our judgement: there are a number of good boards out there, with Allwinner SoCs and the same possibility as the Pine64 to have decent Linux support.
Do you know any one that has it?
Why should we believe that it will be any different with the Pine64?
That said, myself, i will just wait for the board to come out and see how it goes on the Linux side. And if it goes well, i'll GLADLY buy the 2GB version. I want it to go well. I want to be wrong. That's why i bitch about it here. But (given the facts) i'm sceptic...
(01-07-2016, 02:44 PM)nomadewolf Wrote: By 'a bit of dev and elbow grease' you're definitely being optimistic. There's absolute no guarantee on that department. While it is entirely possible, it doesn't mean anything else. We just need to wait and see.
Allwinner's uboot booting into an Allwinner 3.0.x kernel with a linux rootfs is definitely just a matter of doing the work. Getting a mainstream distro image from that is just a matter of dealing with things that depend on kernel > 3.0.x like systemd. This is not a matter of opinion, it is a fact that inevitably follows from technical understanding of what's required for that. If the Allwinner blobs work, this same setup gives you acceleration. If they, for some mysterious reason, work in android but not in linux, it might get tricky (or not).
I'm Russian, so I can't possibly be an optimist - it's entirely alien to our national character  It's also fatal in engineering. I never said it'll take a couple of hours to do the above, just that it is entirely doable with a reasonable amount of work because it's technically possible and not rocket surgery in the grand scheme of things.
Quote:And though the AMD/Nvidia GPUs are on a completely different level of complexity, at least on the AMD side there is documentation and cooperation and code contribution. And that is a BIG deal and a big difference.
No argument there, but this is ARM IP iiuc and, if so, not going to happen.
Quote:As for using an older kernel, no i'm not so square that i wouldn't use a kernel 'just' beacause it is old. But with age come the bugs and more importantly security flaws. So yeah, it is a showstopper for me, on account of security concerns which is a serious issue, at least for me.
I made a very extensive post on what 3.0.x means for security either in this thread or the dev board request thread. On a non-multiuser system, it's not a problem at all. In fact, if you're running openwrt say, you're likely running kernel <= 3 (my devices are on 2.6 actually) and sleeping peacefully. It's not ideal, but it is an issue in a miniscule number of cases.
Quote:Even if Android didn't use the Linux kernel (which would make the whole thing much more unlikely), there would still be a possibility. Nothing is impossible if there is enough work involved. But we have to factor reality in our judgement: there are a number of good boards out there, with Allwinner SoCs and the same possibility as the Pine64 to have decent Linux support.
Do you know any one that has it?
Why should we believe that it will be any different with the Pine64?
You're making different arguments. Those boards have linux support. Pine64 will have linux support. That is a given. If by 'decent' support you mean mainline uboot + mainline kernel + full hw acceleration all at the same time, I'd say at least not for a while for the first two and possibly never for the latter. I would, however, argue that that would be 'excellent' rather than 'decent' and a very high standard well above the level of support required to make the board useful to a large number of people. Most people want to be able to slap debian on and go do things with it and that should be available in the not too distant future.
Quote:That said, myself, i will just wait for the board to come out and see how it goes on the Linux side. And if it goes well, i'll GLADLY buy the 2GB version. I want it to go well. I want to be wrong. That's why i bitch about it here. But (given the facts) i'm sceptic...
Perfectly reasonable attitude wrt waiting. I will, however, keep correcting these 'facts' when they, in fact, are not
-p
(01-07-2016, 03:56 PM)paulieg Wrote: (01-07-2016, 02:44 PM)nomadewolf Wrote: By 'a bit of dev and elbow grease' you're definitely being optimistic. There's absolute no guarantee on that department. While it is entirely possible, it doesn't mean anything else. We just need to wait and see.
Allwinner's uboot booting into an Allwinner 3.0.x kernel with a linux rootfs is definitely just a matter of doing the work. Getting a mainstream distro image from that is just a matter of dealing with things that depend on kernel > 3.0.x like systemd. This is not a matter of opinion, it is a fact that inevitably follows from technical understanding of what's required for that. If the Allwinner blobs work, this same setup gives you acceleration. If they, for some mysterious reason, work in android but not in linux, it might get tricky (or not).
I'm Russian, so I can't possibly be an optimist - it's entirely alien to our national character It's also fatal in engineering. I never said it'll take a couple of hours to do the above, just that it is entirely doable with a reasonable amount of work because it's technically possible and not rocket surgery in the grand scheme of things.
Quote:And though the AMD/Nvidia GPUs are on a completely different level of complexity, at least on the AMD side there is documentation and cooperation and code contribution. And that is a BIG deal and a big difference.
No argument there, but this is ARM IP iiuc and, if so, not going to happen.
Quote:As for using an older kernel, no i'm not so square that i wouldn't use a kernel 'just' beacause it is old. But with age come the bugs and more importantly security flaws. So yeah, it is a showstopper for me, on account of security concerns which is a serious issue, at least for me.
I made a very extensive post on what 3.0.x means for security either in this thread or the dev board request thread. On a non-multiuser system, it's not a problem at all. In fact, if you're running openwrt say, you're likely running kernel <= 3 (my devices are on 2.6 actually) and sleeping peacefully. It's not ideal, but it is an issue in a miniscule number of cases.
Quote:Even if Android didn't use the Linux kernel (which would make the whole thing much more unlikely), there would still be a possibility. Nothing is impossible if there is enough work involved. But we have to factor reality in our judgement: there are a number of good boards out there, with Allwinner SoCs and the same possibility as the Pine64 to have decent Linux support.
Do you know any one that has it?
Why should we believe that it will be any different with the Pine64?
You're making different arguments. Those boards have linux support. Pine64 will have linux support. That is a given. If by 'decent' support you mean mainline uboot + mainline kernel + full hw acceleration all at the same time, I'd say at least not for a while for the first two and possibly never for the latter. I would, however, argue that that would be 'excellent' rather than 'decent' and a very high standard well above the level of support required to make the board useful to a large number of people. Most people want to be able to slap debian on and go do things with it and that should be available in the not too distant future.
Quote:That said, myself, i will just wait for the board to come out and see how it goes on the Linux side. And if it goes well, i'll GLADLY buy the 2GB version. I want it to go well. I want to be wrong. That's why i bitch about it here. But (given the facts) i'm sceptic...
Perfectly reasonable attitude wrt waiting. I will, however, keep correcting these 'facts' when they, in fact, are not
-p
You obviously know way more about the Linux kernel than me, and i don't want to be pushy but i don't really think that there is decent support for any other board, or at least for the ones i researched (HummingBoard, Banana Pi, Orange Pi, O-Droid, BeagleBoard, etc) and i do know that the O-Droid actually runs on a Samsung SoC, for instance.
Just saying that neither of them has it.
And by decent support, i mean hardware acceleration that works.
There is the case of the Raspberry Pi 2, but i want at least 2GB RAM, Gigabit Ethernet, USB 3.0 and a Quad-Core processor that's above the Gigahertz barrier.
Just meet up with Allwinner folks this afternoon, their president still voice support on open source community and currently assign dedicate resource to do so. In pass, Allwinner rely on third parties to support open source community and no direct involve. Hopefully this time with Allwinner direct involvement, the situation can getting better. Please allow me to have some time to make this arrangement. I try to help up, good for open source community and also good for Pine64. Lets work together, cheers.
@ tllim,
That is good news; thank you for the update. Perhaps you could encourage some of the Allwinner OSS folks to participate here? It does seem like we have some super knowledgeable folks on this thread and perhaps connecting them with the right Allwinner people could get a jump start on the process....
(01-08-2016, 02:15 AM)tllim Wrote: Just meet up with Allwinner folks this afternoon, their president still voice support on open source community and currently assign dedicate resource to do so. In pass, Allwinner rely on third parties to support open source community and no direct involve. Hopefully this time with Allwinner direct involvement, the situation can getting better. Please allow me to have some time to make this arrangement. I try to help up, good for open source community and also good for Pine64. Lets work together, cheers.
Thank you TL for having a chat with Allwinner.
(01-08-2016, 07:27 AM)beard Wrote: Thank you TL for having a chat with Allwinner.
Do you trust in chats more than in 'measurable' efforts how to support the open source community? For example do you find A64 here: https://github.com/allwinner-zh/documents
Of course not, you'll have to visit Olimex github since they're the OSHW people publishing every stuff even if they had to sign NDA: https://github.com/OLIMEX/OLINUXINO/tree...r/HARDWARE
01-08-2016, 07:57 AM
(This post was last modified: 01-08-2016, 08:08 AM by joe.)
Andrew2
Do you trust in chats more than in 'measurable' efforts how to support the open source community? For example do you find A64 here: https://github.com/allwinner-zh/documents
Of course not, you'll have to visit Olimex github since they're the OSHW people publishing every stuff even if they had to sign NDA: https://github.com/OLIMEX/OLINUXINO/tree...r/HARDWARE
---------------------------------------------------------------------------------------------------------------
Why you know have sign NDA?
this is Non-disclosure agreement or Open disclosure agreement?
(01-08-2016, 07:45 AM)Andrew2 Wrote: (01-08-2016, 07:27 AM)beard Wrote: Thank you TL for having a chat with Allwinner.
Do you trust in chats more than in 'measurable' efforts how to support the open source community? For example do you find A64 here: https://github.com/allwinner-zh/documents
Of course not, you'll have to visit Olimex github since they're the OSHW people publishing every stuff even if they had to sign NDA: https://github.com/OLIMEX/OLINUXINO/tree...r/HARDWARE
Are you on crack? What's wrong with you?
We need Allwinner to cooperate with the open source community. TL had several meetings with them, including the CEO, and got them to commit to doing so in the first instance and now to dedicate engineering resources to it. Is this a guarantee that they will cooperate? No. However, it is a necessary first step. What is there to criticise about TL/Pine64 doing this? This commitment is further than anyone has gotten before with Allwinner.
A64 is irrelevant. Pine64 uses the A53 SoC. The entire BSP for A53, which is exactly what you were pointing to on the olinuxino github for the A64, is available for download from the pine64 wiki. No NDA required. Just some RTFW, tar and gzip. If it makes you feel better, you can put it up on github.
-p
|
