Dev. Units


Pickle

Mega GP Mania
Joined
May 30, 2006
Messages
5,518
Location
Detroit, Michigan
Website
Visit site
Ive been brewing with this in my mind for the last couple since the info was given. Maybe its worth getting a little discussion on. What level of developer should think about getting one?
Of course the obvious choices are the emulator dev's, but what about the more casual developer like myself?

For example if the toolchains and libraries were available I could for sure get my previous ports going like dosbox and wolf3d.
Other considerations ive been thinking of, what is the additional shippings costs? Are they coming from the UK only?
Will they be assembled? Ive read in the past they may need some assembly by the user.
What about the risks of buying the hw early, what kind of bugs will be in the dev units. For this reason im hesitant to get a dev unit, as I probally wont have enough to get a final version.
Maybe most of us will compile the apps without the hardware and have some else try it that does or wait.

Im interested on what some other opinions are. It would be interesting to know how many have shown interest in getting a dev unit.
 
No matter how good the designers and developers are (and they certainly are talented) there will certainly be HW and SW problems.
Will they prevent the machine from working? I bet not, or they won't send the units.
Will they be a pain? I bet some of the problems will be.

If all you want is compiling your apps and see them work, I think you should not take a dev kit, unless you are ready to help debug the kernel for instance, or wait a few days until someone else corrects the problem.

PS - This is a personal point of view sustained by no Pandora fact, just based on some previous experiences ;)
 
Laurent said:
No matter how good the designers and developers are (and they certainly are talented) there will certainly be HW and SW problems.
Will they prevent the machine from working? I bet not, or they won't send the units.
Will they be a pain? I bet some of the problems will be.
I realize that, but I think the problems from the 100 unit run to the 3000 run will be very small.

Laurent said:
If all you want is compiling your apps and see them work, I think you should not take a dev kit, unless you are ready to help debug the kernel for instance, or wait a few days until someone else corrects the problem.
Good point, I dont think I have the experience or the time to do kernel level development.
I dont think the time between runs will be long, I think the estimation were a few months?
 
Last edited by a moderator:
A related question is if snapshots of the boot loader and kernel source code will be made available to the first 100 developers.

I agree with Laurent and others that silicon is not perfect and that's exactly why I would like to see some code. Not for the Beagle, not for a generic OMAP 3530, but for the actual hardware, WiFi, audio, errata, warts, work arounds, TODO, WTF?!, KILL ME NOW comments and all the rest included.

I fully understand that snapshots might become obsolete the next day. That's why I use that term, not milestone or release.
 
netbsd said:
A related question is if snapshots of the boot loader and kernel source code will be made available to the first 100 developers.
They will be made available to everyone, not just the first 100. (The first 100 are most likely to be contributing towards kernel source anyway).
 
Last edited by a moderator:
Squidge said:
They will be made available to everyone, not just the first 100. (The first 100 are most likely to be contributing towards kernel source anyway).
That's great. I would like to be one of the 100, but I understand if Linux devs are preferred.

It also strongly depends on whether Pandora would be usable with only open sourced device drivers. In another thread Svartalf casts some doubt on that, in particular w.r.t. WiFi.

With "usable" I mean that scrolling on the framebuffer device without 2D accel is "fast enough" and that the keyboard, D-pad, sound (DAC), SDHC cards, USB host and WiFi can be used.

I hope someone is able to give a definitive answer to that soon.
 
Last edited by a moderator:
netbsd said:
That's great. I would like to be one of the 100, but I understand if Linux devs are preferred.

That was me telling it would help early developers, it's not any official statement :)
QUOTE
It also strongly depends on whether Pandora would be usable with only open sourced device drivers. In another thread Svartalf casts some doubt on that, in particular w.r.t. WiFi.

I don't know about WiFi, but the SGX driver will certainly not be open source. BTW can someone confirm it will even be available for the first 100 units release?

QUOTE
With "usable" I mean that scrolling on the framebuffer device without 2D accel is "fast enough"

The FB is part of the RAM. I guess we would need someone to benchmark memory bandwidth.
 
Last edited by a moderator:
Laurent said:
I don't know about WiFi, but the SGX driver will certainly not be open source. BTW can someone confirm it will even be available for the first 100 units release?

The FB is part of the RAM. I guess we would need someone to benchmark memory bandwidth.
I don't really expect scrolling to be "unbearably slow" at all.

What is this SGX driver? I assume (1) it's 2D/3D acceleration for X Svartalf was talking about and (2) that's it purely optional. It becomes a radically different story if there's secret sauce involved to do any screen output at all. Can't really imagine that.
 
Last edited by a moderator:
Laurent said:
I don't know about WiFi, but the SGX driver will certainly not be open source. BTW can someone confirm it will even be available for the first 100 units release?
I would assume so, it's available now. The question is whether or not they regard it as a public release.

QUOTE

What is this SGX driver? I assume (1) it's 2D/3D acceleration for X Svartalf was talking about and (2) that's it purely optional. It becomes a radically different story if there's secret sauce involved to do any screen output at all. Can't really imagine that.



You can do 2D easily, but the 2D/3D acceleration driver is SGX, so the source will not be released to that.
 
Last edited by a moderator:
netbsd said:
TODO, WTF?!, KILL ME NOW comments...
Heh... I've seen quite a few of those types of comments in my day. I've also seen a few of them
with coarser language than that (Heh... IN the Linux kernel even) and comments remarking on having
to do a workaround for "such-and-such C cowpiler". ;)

I concur... Can we start seeing the code? Pleeeease? :D

netbsd said:
It also strongly depends on whether Pandora would be usable with only open sourced device drivers. In another thread Svartalf casts some doubt on that, in particular w.r.t. WiFi.
Heh... I believe I cast some doubt a while back on the 2D/3D accel support based off of concerns of what they did to accomplish closed driver support- don't remember if I cast a cloud on the WiFi support's openness at all. The WiFi stuff will depend on if they're using a braindead Broadcom part or something more like what Atheros is selling these days.
 
Last edited by a moderator:
netbsd said:
Laurent said:
The FB is part of the RAM. I guess we would need someone to benchmark memory bandwidth.
I don't really expect scrolling to be "unbearably slow" at all.
Hum 800x480x4 bytes x 60 fps = 87.9 MB/s. Now if you scroll make that 87.9 MB/s of reading + 87.9 MB/s of writing.
Not much I agree, but definitely significant.

EDIT: that makes 46 million 32 bit transactions per second.

EDIT 2: the RAM will probably be 166 MHz and 32 bit wide if it's like Beagle OMAP3530. So that means the theoretical max is 333 million 32 bit transactions per second.
 
Last edited by a moderator:
I guess the pinned post CraigX just made settles it. Unless you're up to blazing the trail for the people that'll follow on, you'll want to wait for the next pass at them.
 
Svartalf said:
I guess the pinned post CraigX just made settles it. Unless you're up to blazing the trail for the people that'll follow on, you'll want to wait for the next pass at them.
yeah I would agree with that. Looks like I will be waiting.
 
Last edited by a moderator:
Back
Top