bradburts
Posts: 341
Joined: Sun Oct 02, 2011 7:07 am

Re: Broadcom reaction to distructive entrepreneurship

Mon Dec 05, 2011 12:49 pm

[quote]But not Windows.[/quote]
I know that feeling ;)

What do you do with LambdaPi, whats it best at?

The TI processor looked interesting, got beat to the post on comparison of features.
The $99 eval board would hurt me, tend to kill too many of them!

Neil
Posts: 98
Joined: Thu Sep 29, 2011 7:10 am
Contact: Website

Re: Broadcom reaction to distructive entrepreneurship

Mon Dec 05, 2011 12:52 pm

[quote]Quote from ozone on December 5, 2011, 11:58
What about TI A8 AM335x ? launched at 5$ with all ti ecosystem...
[/quote]

Nice of you to fall into the marketing trap. Check the TI product page:
http://www.ti.com/product/am3359#samples

The 1,000-off price is $17.80. If you want the $5 price you need to be buying them in serious volumes (say, 100k/wk). If you buy them from distros then you need to include their markup as well (lets say 40%).

Scribe
Posts: 98
Joined: Wed Sep 28, 2011 12:00 pm
Contact: Website

Re: Broadcom reaction to distructive entrepreneurship

Mon Dec 05, 2011 12:53 pm

[quote]Quote from bradburts on December 5, 2011, 12:49
[quote]But not Windows.[/quote]
I know that feeling ;)

What do you do with LambdaPi, whats it best at?

The TI processor looked interesting, got beat to the post on comparison of features.
The $99 eval board would hurt me, tend to kill too many of them![/quote]

That\'s exactly why I ordered 2 along with enough 150MHz RF isolators to allow me to isolate each channel to 5KV!

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 23682
Joined: Sat Jul 30, 2011 7:41 pm

Re: Broadcom reaction to distructive entrepreneurship

Mon Dec 05, 2011 1:17 pm

Not bad prices though, looks like a good buy for some of the people above who need better documentation and ability to buy in very low numbers. Probably a better fit than the BRCM2835 anyway, if you don\'t need the graphics horsepower. It\'s a bit more expensive, but not prohibitively so.
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed. Here's an example...
"My grief counseller just died, luckily, he was so good, I didn't care."

bradburts
Posts: 341
Joined: Sun Oct 02, 2011 7:07 am

Re: Broadcom reaction to distructive entrepreneurship

Mon Dec 05, 2011 2:15 pm

[quote]That\'s exactly why I ordered 2 along with enough 150MHz RF isolators to allow me to isolate each channel to 5KV![/quote]

Only 2???? You must be very careful :)
Last project I got through 10 boards and most of the interfaces were farmed off to SPI devices!
Its the PSU which gets me every time.

$17 is a lot, hope soon to get a RPI for $25 connectors and all.
I don\'t need the GPU much either.
You can see why the foundation choose the SoC though, its a good all rounder and perhaps Broadcom were the only ones who showed much of an interest.
Wish there was a fast ethernet and SATA .......

@Storris
Posts: 11
Joined: Mon Nov 28, 2011 3:59 pm

Re: Broadcom reaction to distructive entrepreneurship

Mon Dec 05, 2011 2:25 pm

The first 10 pages of this thread are hugely amusing, but Conspiracy Theories are best if documented on YouTube with links to relevant pastors and texts so that we can follow the whole line of non-thinking.

As for Broadcom not releasing the information that is needed to secure their product, how happy do we think their commercial consumers be if 10 year olds with an #RPi were able to hack Sesame Street videos into all of their products?

There are options available to you if you think you need the data,
#1 Ask for it
#2 Go to another chip that does have open-access
#3 Make your own Chip.

The options are limited, but there they are, like it or lump it the choice is yours and no-one elses.

bradburts
Posts: 341
Joined: Sun Oct 02, 2011 7:07 am

Re: Broadcom reaction to distructive entrepreneurship

Mon Dec 05, 2011 3:30 pm

[quote]
#1 Ask for it
[/quote]
We did. As I understand it the standard stuff is documented in its drivers.
No-one can document the bugs, so driver source is fine.
GPU no dice, but then who can program a GPU? The device\'s USP seems to be the GPU, HDMI etc so the GPU driver will just work.

You\'re missing #4
#4 We add the information to the wiki as it is figured out.
I am going with that one.

Isn\'t your hacking point the same as the OP\'s original thought? Lets not get back to conspiracy theories please!
I don\'t think that argument holds up.
I guess that the SoC is meant for set top media players.
Be easy enough to roll your own RPI media player with MPlayer or whatever. Dunno about 10 year olds but a teen if they would get off their mobile phone for more than a few minutes !

Scribe
Posts: 98
Joined: Wed Sep 28, 2011 12:00 pm
Contact: Website

Re: Broadcom reaction to distructive entrepreneurship

Mon Dec 05, 2011 3:59 pm

[quote]Quote from bradburts on December 5, 2011, 15:30
[quote]
#1 Ask for it
[/quote]
We did. As I understand it the standard stuff is documented in its drivers.
No-one can document the bugs, so driver source is fine.
GPU no dice, but then who can program a GPU? The device\'s USP seems to be the GPU, HDMI etc so the GPU driver will just work.

You\'re missing #4
#4 We add the information to the wiki as it is figured out.
I am going with that one.

Isn\'t your hacking point the same as the OP\'s original thought? Lets not get back to conspiracy theories please!
I don\'t think that argument holds up.
I guess that the SoC is meant for set top media players.
Be easy enough to roll your own RPI media player with MPlayer or whatever. Dunno about 10 year olds but a teen if they would get off their mobile phone for more than a few minutes !

[/quote]

Will have to have a gander at the drivers at some point, but if it\'s anything like typical Linux Kernel coding style I\'d hang on to the idea of a programming guide as well lol. Old-style coding and C89 == migraine. I\'m yet to see a flurry of commenting in Linux driver code as well, though it\'s technically down to the individual contributor.

I think perhaps an example of why I like programming guides (excerpt from AM335X guide):

[quote]
4.6.2.2.5 Interrupt Status Clearing
After servicing the interrupt (after execution of the ISR), interrupt status is to be cleared. If a system interrupt status is not cleared, then another host interrupt may not be triggered or another host interrupt may be triggered incorrectly. It is also essential to clear all system interrupts before the PRU is halted as the PRU does not power down unless all the interrupt status are cleared. For clearing the status of an interrupt, whose interrupt number is N, write a 1 to the Nth bit position in the system interrupt status enabled/clear registers (SECR1-SECR2). System interrupt N can also be cleared by writing the value N into the system interrupt status indexed clear register (SICR).
[/quote]

It lets you know important information about the functionality of the device that could lead to issues or unexplained behaviour on the OS level, like why\'s my power consumption not going down? Why\'s this interrupt not working?

Unless there\'re many many more comments than code, this kind of information is lost in drivers. With no comments, you can work out what\'s going on (after you cope with the typically unfriendly variable names) but not why, increasing the chance of problematic modifications.

Not a shot at RaspPi by the way, just my experiences with Linux.

bradburts
Posts: 341
Joined: Sun Oct 02, 2011 7:07 am

Re: Broadcom reaction to distructive entrepreneurship

Mon Dec 05, 2011 5:12 pm

You\'re right Scribe.
I was thinking about the peripheral side and assumed that the USB etc cells were standard and that the source code would lead to the other references, if needs must.
I don\'t know enough about today\'s ARMs to know what\'s in the core (& so have documents elsewhere) or not. Would hope that the basics; interrupts, timers, i/o etc have their ARM equivelants to 8259, 8254, etc and assumed so as the source for these drivers was made available.
I don\'t know though.

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 23682
Joined: Sat Jul 30, 2011 7:41 pm

Re: Broadcom reaction to distructive entrepreneurship

Mon Dec 05, 2011 7:23 pm

[quote]Quote from bradburts on December 5, 2011, 15:30
[quote]
#1 Ask for it
[/quote]
We did. As I understand it the standard stuff is documented in its drivers.
No-one can document the bugs, so driver source is fine.
[/quote]

Well, technically you asked Raspberry Pi I think, not the chip supplier who are the people who have that information.

And Scribe is unfortunately right - unless the driver source is very well documented it\'s still not easy. His example programming guide paragraph is probably the best I have ever seen btw - but going to that level of detail on the GPU would result in a 10k page guide just for the registers, never mind all the libraries that run on the GPU.
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed. Here's an example...
"My grief counseller just died, luckily, he was so good, I didn't care."

bradburts
Posts: 341
Joined: Sun Oct 02, 2011 7:07 am

Re: Broadcom reaction to distructive entrepreneurship

Mon Dec 05, 2011 7:46 pm

Lets put that down to how helpful the PI guys are viewed as being then.

Drivers are never easy, even with the datasheet. I live in hope that most bits come from standard cells though or something BC have done before.

Svartalf
Posts: 596
Joined: Fri Jul 29, 2011 6:50 pm

Re: Broadcom reaction to distructive entrepreneurship

Mon Dec 05, 2011 7:54 pm

[quote]Quote from johnbeetem on December 4, 2011, 23:02
1. No documentation about GPU internals or register interface is going to be published. Nada. Zip. None of the other SoCs publish this so it\'s unreasonable to expect Broadcom to.
[/quote]

Yet, that is... And for the time being, they\'re not in a position, right at this time, to realistically do so. Moreover, WE are not Broadcom\'s customers here. R-Pi, Roku and others are. I am surprised that this keeps coming up time and time again. The parties doing it aren\'t doing the FOSS community ANY favors by the way they\'re doing this- the industry does not work the way they think it does. All you\'re doing is raising your blood pressure and basically biting the hand that feeds you.

(It should be noted that there\'s a vendor that could pretty much open up their tech data on their GPU- but they\'ve not had their customers ask for that and it\'s way low on their list of priorities (Obviously. However, they\'re in a bit of a unique position to do that very thing- can\'t say whom...but you might be able to guess, if you look at the whole GPU landscape, who that might be.) You know, if you\'re clever and could get their customers (again, you\'re not this vendor\'s customer...) to ask for this, you never know, might cause a landslide of openings... ;) )

Svartalf
Posts: 596
Joined: Fri Jul 29, 2011 6:50 pm

Re: Broadcom reaction to distructive entrepreneurship

Mon Dec 05, 2011 8:05 pm

[quote]Quote from jamesh on December 5, 2011, 19:23
And Scribe is unfortunately right - unless the driver source is very well documented it\'s still not easy. His example programming guide paragraph is probably the best I have ever seen btw - but going to that level of detail on the GPU would result in a 10k page guide just for the registers, never mind all the libraries that run on the GPU.

[/quote]

I\'ll state, beings as I\'m a driver developer for hire, that this is the God\'s Truth on the matter. It\'s NOT easy. It\'s even less so for a GPU these days. Long gone are the days when you could just jam values into a register and basically make it go. It\'s more like a specialized version of C coding to drive stream processors to their peak capabilities in the case of modern GPUs. I\'ve coded drivers for both classes of GPUs. It\'s not simple. I think I might be in the class to be able to eventually grok what they\'ve done and be able to do something with it- but I know enough to know James isn\'t kidding on this.

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 23682
Joined: Sat Jul 30, 2011 7:41 pm

Re: Broadcom reaction to distructive entrepreneurship

Mon Dec 05, 2011 8:27 pm

[quote]Quote from Svartalf on December 5, 2011, 20:05
I\'ve coded drivers for both classes of GPUs. It\'s not simple. I think I might be in the class to be able to eventually grok what they\'ve done and be able to do something with it- but I know enough to know James isn\'t kidding on this. [/quote]

No, I don\'t kid. As anyone who has heard any of my jokes would willingly confirm.

(I might exaggerate slightly sometimes, mainly to females, just on the off chance...hold on, back to the kidding bit again)
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed. Here's an example...
"My grief counseller just died, luckily, he was so good, I didn't care."

mikeh
Posts: 1
Joined: Mon Dec 05, 2011 10:41 pm

Re: Broadcom reaction to distructive entrepreneurship

Mon Dec 05, 2011 10:52 pm

HI,

I think the Pi is a great opportunity for developers, however there are applications that do not require any graphics but could benefit from using features of the GPU. Does anyone have any idea if there might be an OpenCL toolset available at sometime ?

mikeh

Scribe
Posts: 98
Joined: Wed Sep 28, 2011 12:00 pm
Contact: Website

Re: Broadcom reaction to distructive entrepreneurship

Tue Dec 06, 2011 1:15 am

[quote]Quote from mikeh on December 5, 2011, 22:52
HI,

I think the Pi is a great opportunity for developers, however there are applications that do not require any graphics but could benefit from using features of the GPU. Does anyone have any idea if there might be an OpenCL toolset available at sometime ?

mikeh[/quote]

Hmm, perhaps, though from what I gather we\'ve been told the unit is made from many smaller separate elements with many lines of ASM, would likely take a bit of effort/time to wrap them up nicely for general compute.

Unlocking user-friendly access to the 24GFLOPs general purpose power would clearly be the key to the Pi\'s success as this is what differentiates it from the competition, but it\'s of no use if you can\'t get at it, ideally with a standardised API. It seems to be a fairly untapped market sector, TI do have DSPs with general compute cores of 24GFLOPS + 48GMACs but I can only find silicon in the high-end employing 2-8 of these cores in a single chip at whopping prices of £100 to £168 (single unit), though the prices scale well as you add cores, there\'s nothing in the low-end price bracket currently on offer that reaches 24GFLOPS, at least that I can find.

OpenMax support is there from what I gather however for h/w media codec work.

bradburts
Posts: 341
Joined: Sun Oct 02, 2011 7:07 am

Re: Broadcom reaction to distructive entrepreneurship

Tue Dec 06, 2011 9:04 am

24 GFLOPS for a couple of watts. Could add a lot of eco. friendly folding if the Pi takes off (EDIT) and ....

Scribe, there\'s a polling thread http://www.raspberrypi.org/forum?mingle ... stid-20092 addressing some of the sensitivity issues raised here.

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 23682
Joined: Sat Jul 30, 2011 7:41 pm

Re: Broadcom reaction to distructive entrepreneurship

Tue Dec 06, 2011 9:25 am

Problem is that OpenCL or indeed anything that gives access to the multiple cores on the Videocore (Vector cores plus others) is not an easy task. Some of the cores requires a \'distinct\' mindset to program, even the relatively (!) simple vector cores requires serious brainpower. And using those brains to implement OpenCL would be expensive, and as there is no real market as yet for this sort of device (in the volumes needed to pay for the development) it\'s not an easy call to make.

I\'m looking at doing something but its still in the very early stages.
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed. Here's an example...
"My grief counseller just died, luckily, he was so good, I didn't care."

bradburts
Posts: 341
Joined: Sun Oct 02, 2011 7:07 am

Re: Broadcom reaction to distructive entrepreneurship

Tue Dec 06, 2011 9:38 am

Good luck with that.
Although hold on, I have just figured out what the rift in space/time (the double post) is about.
Your future self is trying to warn me that if OpenCL gets written then all those Pi nodes will become sentient. And we all know what happens next......

PS
Would like to know more if you are able. Used to program DSP processors years ago. Was one of the most complicated (most fun) jobs I have done, getting all the buses running at full speed. Back then you could run n columns of opcodes with only some bus operations allowed in each column, more like MIMD than SIMD though.

Neil
Posts: 98
Joined: Thu Sep 29, 2011 7:10 am
Contact: Website

Re: Broadcom reaction to distructive entrepreneurship

Tue Dec 06, 2011 1:57 pm

Programming a traditional DSP (TI TMS320, ADI Sharc, Mot DSP56k) is childs-play compared to the VideoCore.

Return to “General discussion”