SpaceHobo
Posts: 23
Joined: Fri Sep 09, 2011 8:31 pm

Re: Broadcom BCM2835 SoC

Sat Sep 10, 2011 9:33 am

Quote from hippy on September 10, 2011, 04:14
Quote from SpaceHobo on September 9, 2011, 22:46
Oh dear. This is highly disappointing. We can only use this device if we accept proprietary drivers? I fail to see how this would meet your goals for education. It seems counter-productive.

I don't really see how it's counter-productive but perhaps I'm just too pragmatic. At $25 with all the R-Pi offers I'd be happy if all the firmware were in ROM, entirely closed, and all I could run from SD Card were Perl scripts or similar. As it is the R-Pi is much more than that.

I find your definition of "pragmatic" confusing, and perhaps you could elaborate here. In this thread I and others have exhibited clear pragmatic reasons why the proprietary video driver in the OS is problematic for the continued usefulness of free operating systems on this device. These reasons are based not on ideological speculation, but on observed history and data gathered by OS developers and driver authors. This isn't about your capacity to surrender freedoms in the name of compromise, but about reliably predicted outcomes of this type of secrecy.

As to how secrecy could appear contrary to education, is that really a confusing idea? They're practically linguistic antonyms, aren't they?

obarthelemy
Posts: 1407
Joined: Tue Aug 09, 2011 10:53 pm

Re: Broadcom BCM2835 SoC

Sat Sep 10, 2011 9:39 am

@hobo: do you always ignore questions to which you don't have any answer like you're doing with my posts ? This smacks of trolling, on top of being useless.
- Re: firmware vs driver, there's nothing in this thread about what makes one different from the other, actually there's an argument at the beginning about the firmware needing to be open too for the exact same reasons the driver should be, so I'm waiting for your rationale... if you have any... on what makes graphics code either one or the other.
I'm also waiting on
- your explanation as to why proprietary is an impediment to learning or non-profit
- how you get out of the "practical vs nothing" dichotomy, with your theoretically pure $25 computer for November.

User avatar
Emanuele
Posts: 182
Joined: Wed Aug 03, 2011 5:28 pm
Contact: Website

Re: Broadcom BCM2835 SoC

Sat Sep 10, 2011 9:49 am

@SpaceHobo: To be honest, I've skipped most of the posts regarding these issues, so I'm not sure if it's already been said. I'm sure that's not what you mean when you say that


the proprietary video driver in the OS is problematic for the continued usefulness of free operating systems on this device.


but I don't want anyone to think you'll need to throw the device away in the future. You have the source of the whole system minus a binary blob. In the worst case scenario, the system will remain useful and you will always be able to backport important stuff (e.g. USB drivers for new hardware).

SpaceHobo
Posts: 23
Joined: Fri Sep 09, 2011 8:31 pm

Re: Broadcom BCM2835 SoC

Sat Sep 10, 2011 9:56 am

Quote from obarthelemy on September 10, 2011, 10:39
- Re: firmware vs driver, there's nothing in this thread about what makes one different from the other, actually there's an argument at the beginning about the firmware needing to be open too for the exact same reasons the driver should be, so I'm waiting for your rationale... if you have any... on what makes graphics code either one or the other.
I'm also waiting on
- your explanation as to why proprietary is an impediment to learning or non-profit
- how you get out of the "practical vs nothing" dichotomy, with your theoretically pure $25 computer for November.

1: there is indeed a statement that the firmware (while it absolutely should be free for ideological reasons) is not necessarily a required maintenance part for the OS kernel. Thus keeping ourselves to the pragmatic argument, the firmware is an API that the driver can code to.

Unfortunately the kernel is an ever-changing ABI that a proprietary driver must constantly adapt to. Thus the Linux community can easily update drivers to meet new APIs/ABIs for firmware, but it has been clearly demonstrated that no proprietary driver authors ever keep up with all ABI changes in the kernel. This is observed fact.

Equivocation between firmware and software drivers in this sort of discussion is usually only done by the advocates of software freedom. It's rare to see it posed by their opponents.

2: If you had a textbook on CPU design that said "This is what CPUs usually do, but we can't tell you how." would that be educational? Do you really see this point as confusing? Everyone else so far has reacted to this by claiming the realm of video drivers insignificant compared to the larger goals of the project. Are you claiming instead that secrecy will assist education in video drivers?

3: "practical vs nothing" already begs so many questions as to be useless. I and others have already demonstrated the impracticality of relying on a binary-only driver for a free operating system.

SpaceHobo
Posts: 23
Joined: Fri Sep 09, 2011 8:31 pm

Re: Broadcom BCM2835 SoC

Sat Sep 10, 2011 9:59 am

Quote from Emanuele on September 10, 2011, 10:49
@SpaceHobo: To be honest, I've skipped most of the posts regarding these issues, so I'm not sure if it's already been said. I'm sure that's not what you mean when you say that


the proprietary video driver in the OS is problematic for the continued usefulness of free operating systems on this device.


but I don't want anyone to think you'll need to throw the device away in the future. You have the source of the whole system minus a binary blob. In the worst case scenario, the system will remain useful and you will always be able to backport important stuff (e.g. USB drivers for new hardware).


Do you really think that the Linux kernel development community will go to the trouble of maintaining an out-of-date ABI for each piece of hardware that lacks free drivers?

If this proprietary piece is removed from the OS, will framebuffer consoles still be possible? What about unaccelerated X11?

User avatar
Emanuele
Posts: 182
Joined: Wed Aug 03, 2011 5:28 pm
Contact: Website

Re: Broadcom BCM2835 SoC

Sat Sep 10, 2011 10:07 am

Quote from SpaceHobo on September 10, 2011, 10:59
Quote from Emanuele on September 10, 2011, 10:49
@SpaceHobo: To be honest, I've skipped most of the posts regarding these issues, so I'm not sure if it's already been said. I'm sure that's not what you mean when you say that


the proprietary video driver in the OS is problematic for the continued usefulness of free operating systems on this device.


but I don't want anyone to think you'll need to throw the device away in the future. You have the source of the whole system minus a binary blob. In the worst case scenario, the system will remain useful and you will always be able to backport important stuff (e.g. USB drivers for new hardware).


Do you really think that the Linux kernel development community will go to the trouble of maintaining an out-of-date ABI for each piece of hardware that lacks free drivers?

If this proprietary piece is removed from the OS, will framebuffer consoles still be possible? What about unaccelerated X11?

I mean that you can keep using your own Raspberry Pi with your keyboard, mouse and TV until it breaks. That's by far the most important thing. If you have little money, a lot of talent and a lot of time you yourself will be able to backport the driver for your own new wifi dongle.

SpaceHobo
Posts: 23
Joined: Fri Sep 09, 2011 8:31 pm

Re: Broadcom BCM2835 SoC

Sat Sep 10, 2011 10:24 am

Quote from Emanuele on September 10, 2011, 11:07
Quote from SpaceHobo on September 10, 2011, 10:59
Do you really think that the Linux kernel development community will go to the trouble of maintaining an out-of-date ABI for each piece of hardware that lacks free drivers?

If this proprietary piece is removed from the OS, will framebuffer consoles still be possible? What about unaccelerated X11?

I mean that you can keep using your own Raspberry Pi with your keyboard, mouse and TV until it breaks. That's by far the most important thing. If you have little money, a lot of talent and a lot of time you yourself will be able to backport the driver for your own new wifi dongle.


I appreciate this perspective, but to me the thought of all these people running vulnerable old OS revisions with known-exploitable kernels on the Internet with their new wifi dongles... well it is one of the things about this proprietary driver situation that causes me much dismay.

obarthelemy
Posts: 1407
Joined: Tue Aug 09, 2011 10:53 pm

Re: Broadcom BCM2835 SoC

Sat Sep 10, 2011 10:37 am

@hobo
1- "the firmware is an API that the driver can code to". Same as "the driver is an API than the OS can code to" maybe ? The decision on what code goes where is arbitrary, a driver can write to a chip's registers, and firmware can implement OS API calls. Also, observed fact is that proprietary drivers are superior to open-source ones ^^
2- Would it prevent you from learning to program ? Millions of people program to very proprietary platforms. Also, having the latest kernel and its attending drivers is not really important to achieving Pi's goals. I'm fairly certain today's kernel and graphics stack will be perfectly OK to learn programming 5 yrs from now. And that's assuming Broadcom release no updates at all. Documentation is a much more important point than openness, by the way.
3- so your solution is ? at $25 for november ? If you don't have any, you're in the famous "realistic or nothing" dichotomy you don't "subscribe to" ? Not agreeing to a dichotomy must mean you have a 3rd way ?

Anyway, I'm bored. Signing off this topic.

Motley
Posts: 10
Joined: Fri Sep 09, 2011 9:59 pm

Re: Broadcom BCM2835 SoC

Sat Sep 10, 2011 10:41 am

Given its been said elsewhere that communication with the GPU is via a command stream, i.e. a data structure, then I don't see how API/ABI evolution of the kernel driver that creates that data structure will break the ability to communicate with the GPU? It would if that kernel driver was a binary blob but again I've read elsewhere that they don't know yet if that will be the case, or if it will be open source.

User avatar
Emanuele
Posts: 182
Joined: Wed Aug 03, 2011 5:28 pm
Contact: Website

Re: Broadcom BCM2835 SoC

Sat Sep 10, 2011 10:47 am

Quote from SpaceHobo on September 10, 2011, 11:24
I appreciate this perspective, but to me the thought of all these people running vulnerable old OS revisions with known-exploitable kernels on the Internet with their new wifi dongles... well it is one of the things about this proprietary driver situation that causes me much dismay.

The kernel security patches would be backported by the community while libc, userland, etc... would be kept updated.

By the way, I'm talking about worst case scenario here and I don't want to imply that this is what is likely to happen.

Quote from obarthelemy on September 10, 2011, 11:37
Signing off this topic.

Yes, probably the best thing to do. I made my point, so I better stop replying myself. I wouldn't want to cause this topic to close!

SpaceHobo
Posts: 23
Joined: Fri Sep 09, 2011 8:31 pm

Re: Broadcom BCM2835 SoC

Sat Sep 10, 2011 11:11 am

Quote from obarthelemy on September 10, 2011, 11:37
@hobo
1- "the firmware is an API that the driver can code to". Same as "the driver is an API than the OS can code to" maybe ? The decision on what code goes where is arbitrary, a driver can write to a chip's registers, and firmware can implement OS API calls. Also, observed fact is that proprietary drivers are superior to open-source ones ^^
2- Would it prevent you from learning to program ? Millions of people program to very proprietary platforms. Also, having the latest kernel and its attending drivers is not really important to achieving Pi's goals. I'm fairly certain today's kernel and graphics stack will be perfectly OK to learn programming 5 yrs from now. And that's assuming Broadcom release no updates at all. Documentation is a much more important point than openness, by the way.
3- so your solution is ? at $25 for november ? If you don't have any, you're in the famous "realistic or nothing" dichotomy you don't "subscribe to" ? Not agreeing to a dichotomy must mean you have a 3rd way ?

Anyway, I'm bored. Signing off this topic.

That's probably wise. I'm not here to argue against your ideological stance on this. You have some base axioms you're asserting from an emotional stance, and it's clear that they're deeply-held beliefs. I'm a pragmatist, and as such don't consider this style of conversation to be fruitful.

DanielSilva
Posts: 71
Joined: Sun Aug 28, 2011 7:07 pm

Re: Broadcom BCM2835 SoC

Sat Sep 10, 2011 12:57 pm

Free should be that, free to do what you want as long as it's legally viable. Not free as in "you're free to do whatever we think you should do" and this insistence on having to make everything under the sun "open source" even when it doesn't make sense to make it open source or being open source doesn't help at all to your case is clearly the later.

SpaceHobo
Posts: 23
Joined: Fri Sep 09, 2011 8:31 pm

Re: Broadcom BCM2835 SoC

Sat Sep 10, 2011 3:38 pm

Quote from DanielSilva on September 10, 2011, 13:57
Free should be that, free to do what you want as long as it's legally viable. Not free as in "you're free to do whatever we think you should do" and this insistence on having to make everything under the sun "open source" even when it doesn't make sense to make it open source or being open source doesn't help at all to your case is clearly the later.

Do you have the wrong thread? You seem to be addressing points made by no one who has participated in this discussion thus far.

User avatar
Lob0426
Posts: 2198
Joined: Fri Aug 05, 2011 4:30 pm
Location: Susanville CA.
Contact: Website

Re: Broadcom BCM2835 SoC

Sat Sep 10, 2011 4:34 pm

obarthelemy: that is true. If you are going to use it as a NAS there really no advantage to an overclock. There are many other considerations as to what could limit the overclock of this board. The GPU clock is not tied (from what I have seen) to the ARM clock in any way. So there will be no gains there as you usually get in an x86. The USB clock is separate. Even the memory clocks appear to be independent. It will be interesting to see what optimizations people find. In my experience you may see a 14% 'clock gain but the benchmarks will only show 2 or 3%. It would be nice if there were tweaks to the USB bus but it appears to be locked down to the standard. Has anyone seen any USB chipsets that are not set at pure standard?

SpaceHobo: at this point it is apparent that "they", the foundation, are not able to "open" the GPU. What we are trying to get across is that at this time it is a moot argument. There are many that have made this argument and the answer just is not going to change. I, and others will use it, as it is. Price/performance is the highest, will be the highest on the market at it's release. it will Work well as a NAS, it will work well as a programming tool/trainer, I will be interested to see how successful the people are at using it as a desktop or laptop replacement and it will be an excellent pocket size computing "device"! Not trying to be rude, but this may not be the device for you if you need direct access to the GPU. I do not know if any one of the boards on the market will give you direct access to the GPU. If you find any that will, please post them, as there are others that are interested also.

The Raspberry Pi "is what it is" and I believe it is more than enough. I have asked my own "what if's". And hopefully some of those will, in the future, be incorporated into the revisions. I still want one that has little or no GPU, but able to be mounted on this same board, for purely NAS or server projects. Or USB 3.0 for a real push in transfer. BUT, that is not the Raspberry Pi foundations goal at all. I'll have to settle for Quake 3. Oh so much sacrifice in this world. How can I possibly bear it. lol
512MB version 2.0 as WordPress Server
Motorola Lapdock with Pi2B
Modded Rev 1.0 with pin headers at USB

http://rich1.dyndns.tv/
(RS)Allied ships old stock to reward its Customers for long wait!

pixeltricks
Posts: 4
Joined: Sat Sep 10, 2011 3:49 pm

Re: Broadcom BCM2835 SoC

Sat Sep 10, 2011 5:08 pm

I signed up to post my view on the Raspberry Pi as an engineer. I'm not affiliated with the project or Broadcom in any way. First if someone had told me that Broadcom was going to help someone release an SOC that was open source in any form I would have called you a liar. I don't think most outsiders (meaning people that don't work with these companies), understand how big a deal this really is. This is a huge step for Broadcom to make. The normal procedure for companies like this is to say nothing at all until the documents are signed. It may be just the way things have always been done but it also has a ton to do with security. Broadcom is a major supplier of chips for cable and satellite tv boxes. These chips don't just control the video decoding or encoding but they also control the authorization and security for that content. They have to be extremely careful on what is made public. Leak the wrong information and a single exploit might cost the company billions in losses. I imagine that the gpu code on this chip is going to or has been used on other chips that are on the market used in secure devices. Also add in that they may have 'ip' licensing deals with other companies within that gpu code. They cannot just go to all these other companies and tell them they are open sourcing it all because someone who doesn't even register on profits wants them to.

I look forward to the Raspberry Pi release. There is so much potential in the device and the price point couldn't be better. I'm probably in the minority but I actually would prefer the device in its current form vs the shrink to smaller size because of accessing the individual pins on the board would be easier. My one wish is that we could get SATA support in some form for things like hard drives , that would really make it a contender for NAS, but I will be thrilled to buy one anyway without it.

I'm sure the SDK will provide everything needed to make it a very useful device for both hobbyist and education .

User avatar
liz
Raspberry Pi Foundation Employee & Forum Moderator
Raspberry Pi Foundation Employee & Forum Moderator
Posts: 5202
Joined: Thu Jul 28, 2011 7:22 pm
Contact: Website

Re: Broadcom BCM2835 SoC

Sat Sep 10, 2011 5:30 pm

Slashdot have just sent over the questions for Eben's Q&A next week, and there are a couple there on different aspects of the open issue. This means that as of Wednesday, we'll have something else we can point people asking these questions at (from what I can make out, he'll be repeating pretty much everything he said in last week's talk that I miserably failed to record in these answers too).

And yes, I'm buying a new video camera this week, so we've got something functional before Maker Faire!
Director of Communications, Raspberry Pi

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 25043
Joined: Sat Jul 30, 2011 7:41 pm

Re: Broadcom BCM2835 SoC

Sat Sep 10, 2011 5:43 pm

Quote from Lob0426 on September 10, 2011, 17:34
SpaceHobo: at this point it is apparent that "they", the foundation, are not able to "open" the GPU. What we are trying to get across is that at this time it is a moot argument. There are many that have made this argument and the answer just is not going to change. I, and others will use it, as it is. Price/performance is the highest, will be the highest on the market at it's release. it will Work well as a NAS, it will work well as a programming tool/trainer, I will be interested to see how successful the people are at using it as a desktop or laptop replacement and it will be an excellent pocket size computing "device"! Not trying to be rude, but this may not be the device for you if you need direct access to the GPU. I do not know if any one of the boards on the market will give you direct access to the GPU. If you find any that will, please post them, as there are others that are interested also.


Nobody needs direct access to the GPU except the people who are *able* to write code for it, and that is very few people indeed. And there is no need for access to the GPU blob in order to maintain compatibility with the Linux ABI. That's the job of the low level drivers on the Arm side.

The OSS status of the Arm side drivers is still under discussion I believe, but I may be a bit behind the times. It will depend on how much 'value' there is to Broadcom in those drivers - if publishing the code makes it easier for people to copy their designs/ideas (and therefore cost them money) then it seems obvious which way they will go. They have a duty to shareholders to maximise profits, and if it could be proven that they gave stuff away which cost the shareholders, then they would be in trouble.

My personal thoughts are that there isn't anything on the Arm side *at the moment* which could cause issues. But I am not management (and this stuff goes VERY high up)
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed. Here's an example...
“I own the world’s worst thesaurus. Not only is it awful, it’s awful."

gsh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 1474
Joined: Sat Sep 10, 2011 11:43 am

Re: Broadcom BCM2835 SoC

Sat Sep 10, 2011 7:01 pm

In reply about overclocking, I believe we have had fast chips running at 1GHz at nominal core voltage. Unfortunately there are many issues with doing this (the SMPS isn't rated high enough to provide the power for a start!)

_BUT_ you will not know whether you receive a fast or slow device (unless R-Pi decide to test and bin them)...

As for overvolting the devices, they are rated slightly above the nominal voltage, but it cannot be guaranteed that you won't be reducing the chip life and R-Pi will not be happy about returns in this case. Of course we can just fire off an OTP bit if you start playing around with over-volting which will then make your warranty invalid.
--
Gordon Hollingworth PhD
Raspberry Pi - Director of Software Engineering

User avatar
Lob0426
Posts: 2198
Joined: Fri Aug 05, 2011 4:30 pm
Location: Susanville CA.
Contact: Website

Re: Broadcom BCM2835 SoC

Sat Sep 10, 2011 8:26 pm

I break it, OC, I bought it. That's as it should be. So no ROG (Rise Of Gamer) version LMAO. I would check for over voltage use to be sure it is not a warranty issue. Since most people at this point seem to want multiple RasPii it should work out OK. Binning them would be kind of tedious and unless you are going to charge extra I would not waste the time. Overclocking will be an interesting experiment on an ARM. It may yield some data for future production also.

At 1Ghz were there any migration issues?
512MB version 2.0 as WordPress Server
Motorola Lapdock with Pi2B
Modded Rev 1.0 with pin headers at USB

http://rich1.dyndns.tv/
(RS)Allied ships old stock to reward its Customers for long wait!

DanielSilva
Posts: 71
Joined: Sun Aug 28, 2011 7:07 pm

Re: Broadcom BCM2835 SoC

Sat Sep 10, 2011 9:13 pm

Quote from SpaceHobo on September 10, 2011, 16:38
Do you have the wrong thread? You seem to be addressing points made by no one who has participated in this discussion thus far.

Ahh chucks then, i'll go there and cry at the corner because some company decided not to open source something that isn't even necessary to be open :(

Quote from Lob0426 on September 10, 2011, 21:26
Overclocking will be an interesting experiment on an ARM. It may yield some data for future production also.


Overclocking ARM cpu's is not uncommon though, even one of my wifi routers ( the old venerable wrt54g ) is overclocked and i even have friends who have their smartphones overclocked. Of course we're not talking about huge increments but they're overclocks anyway. I believe there are even Android apps that allow you to soft overclock your phons if i'm not mistaken.

hippy
Posts: 6775
Joined: Fri Sep 09, 2011 10:34 pm
Location: UK

Re: Broadcom BCM2835 SoC

Sat Sep 10, 2011 9:19 pm

Quote from SpaceHobo on September 10, 2011, 10:33
I find your definition of "pragmatic" confusing, and perhaps you could elaborate here. In this thread I and others have exhibited clear pragmatic reasons why the proprietary video driver in the OS is problematic for the continued usefulness of free operating systems on this device.

Put simply; I can build ( or learn to build ) a house from bricks without understanding how bricks are made and without knowing much else about bricks. Any proprietary nature of brick chosen doesn't limit the building or learning experience.

If the R-Pi were completely locked down, sealed in metre thick Kryptonite, impossible to ever alter, then it would limit the ability to change things but that does not necessarily mean it is either useless or unusable.

My pragmatism means I am prepared to accept the proprietary aspect given the low cost of the unit and what it does allow before any proprietary aspect becomes a problem. The proprietary aspect is so small and non-debilitating in my opinion that it's not a practical problem to me, does not detract from what I can and would ever likely want to do with an R-Pi. It's still a useful and potentially valuable educational tool and also in many other ways.

If the choice is between something which is cheap and powerful but has a proprietary aspect and something which is cheap, powerful, entirely open, but non-existent then I will choose the former. That's what I mean about being pragmatic.

SpaceHobo
Posts: 23
Joined: Fri Sep 09, 2011 8:31 pm

Re: Broadcom BCM2835 SoC

Sat Sep 10, 2011 9:19 pm

Quote from DanielSilva on September 10, 2011, 22:13
Quote from SpaceHobo on September 10, 2011, 16:38
Do you have the wrong thread? You seem to be addressing points made by no one who has participated in this discussion thus far.

Ahh chucks then, i'll go there and cry at the corner because some company decided not to open source something that isn't even necessary to be open :(


This is a stunning example of either helplessness or rudeness, but I'm at a loss as to which.

SpaceHobo
Posts: 23
Joined: Fri Sep 09, 2011 8:31 pm

Re: Broadcom BCM2835 SoC

Sat Sep 10, 2011 9:21 pm

Quote from hippy on September 10, 2011, 22:19
Quote from SpaceHobo on September 10, 2011, 10:33
I find your definition of "pragmatic" confusing, and perhaps you could elaborate here. In this thread I and others have exhibited clear pragmatic reasons why the proprietary video driver in the OS is problematic for the continued usefulness of free operating systems on this device.

If the choice is between something which is cheap and powerful but has a proprietary aspect and something which is cheap, powerful, entirely open, but non-existent then I will choose the former. That's what I mean about being pragmatic.

Ah, I see. You may have confused pragmatism with short-term thinking. The two concepts are not the same, and are, in fact, at odds.

User avatar
liz
Raspberry Pi Foundation Employee & Forum Moderator
Raspberry Pi Foundation Employee & Forum Moderator
Posts: 5202
Joined: Thu Jul 28, 2011 7:22 pm
Contact: Website

Re: Broadcom BCM2835 SoC

Sat Sep 10, 2011 9:26 pm

Less of the ad-hominems, please. You can contribute to this discussion by addressing the points made, not the people making them.
Director of Communications, Raspberry Pi

obarthelemy
Posts: 1407
Joined: Tue Aug 09, 2011 10:53 pm

Re: Broadcom BCM2835 SoC

Sat Sep 10, 2011 9:30 pm

Please everyone, stop feeding the troll.

Return to “General discussion”