Pandora Gumstix Overo As A Dev Sandbox For Pandora?


sterling976

Member
Joined
Oct 7, 2008
Messages
217
I've been trying for a while now to get a sandbox set up in which to test code, cross-compile (or natively compile directly on the solution), and generally do a lot of the grunt work getting ready for my pre-ordered Pandora to arrive. I realized fairly early on that qemu won't work for this, since it doesn't emulate armv7a, and I don't trust that code which compiles and runs on armv5te will run properly on the newer chip.

Essentially, the main thing I'm trying to accomplish is get a MythTV Frontend up and running, whether the official version or a smaller client I write myself. The packages in Angstrom's repo aren't specific to armv7a, and some of the Frontend's dependencies seem to be broken unless you're running a very specific device, so I'm not convinced that it will be trivial to get this done.

So I was considering buying a Beagle Board after this last payday so I could use it as a sandbox to start crosscompiling software, make sure my armv7a toolchain was proper, and just generally getting some hands-on time with the architecture before my Pandora arrives. I pre-ordered on Oct. 4th, so I'm probably going to be one of the last of the first batch to get mine. Alas, a couple days before I was ready to buy, digikey sold out of Beagle Boards, so I'm left with money I'd budgeted and no device to spend it on. :(

I've put in my request for a pre-release board but don't know if my proposed project will be deemed worthy, or what timeframe I can expect for a response.

So I'm looking for alternatives. The Gumstix Overo seems to be the next-best option for getting my hands dirty (unless I want to spend a small fortune), but I've got some reservations. I looked around the forums here and only found one relevant post. My concern is that the processor on the Gumstix board is a 3503 vs. the Pandora's 3530. The main differences appear to be the lack of the DSP and SGX components, which aren't a huge deal to me, since I'm really just looking to make sure the applications I'm trying to port will run at all, and don't really care too much about performance.

I realize that the DSP module would help some with playback, but according to the 3503's spec sheet, it should be able to handle NTSC video display. Again, I'm not really concerned about performance -- if the video buffers and stutters, that's fine. I just want to make sure the code will run on the chipset, that I can get a program listing, and that the handoff to a media player works.

Does anyone here have experience with Gumstix or the OMAP 3503 processor? I believe that this would accomplish the goal, which is testing cross-compiled code against the armv7a instruction set. If there's anyone else who's considering this approach, let me know. Also, if I'm missing something in my research that would make the Overo an unsuitable solution, or there's a better solution that could be delivered and implemented before the new year, please speak up.
 
Last edited by a moderator:
_Sterling said:
The packages in Angstrom's repo aren't specific to armv7a, and some of the Frontend's dependencies seem to be broken unless you're running a very specific device, so I'm not convinced that it will be trivial to get this done.
What do you mean it isnt specific? At this page its says there are packages for armv7a?
 
Last edited by a moderator:
Pickle said:
What do you mean it isnt specific? At this page its says there are packages for armv7a?



Ah.. right. It's been a few days since I've looked at this, and I've misremembered some of the details, as they're not really relevant. To answer your question, that's not what I'm installing. This is, although I realize that package also has an armv7a version. If I recall correctly, I was trying to use qemu to create a sandbox, and couldn't get Angstrom running (have a google at that one and you'll see I'm not alone there) and had to install debian. I set up ipkg, and used the armv5te repos. I had a problem with some of the dependencies, too -- lirc and update-rc weren't available in the base repo. Note that my link is for armv7a even though that's not the base I was using at the time. At any rate, the packages did exist in the repos for specific machines, but mixing and matching repos can be extremely dangerous, even on more 'standard' distributions and architectures. You're often better off just compiling what you need yourself.

Unfortunately, this all just highlights the fact that I haven't found any documentation on how to get mythfrontend up and running on an OMAP35xx processor, or even evidence that anyone's done it. I've found documentation on getting it running for other ARM devices, such as the NSLU2 or the Nokia N810, but nothing specific to OMAP35xx, which is why I'm looking for hardware on which to create a sandbox, which was the original point of my post.

So... if we're done picking nits, does anyone have any insight as to the suitability of the Overo to the purpose I put forth? Or is there another option I may have overlooked?
 
Last edited by a moderator:
_Sterling said:
Alas, a couple days before I was ready to buy, digikey sold out of Beagle Boards, so I'm left with money I'd budgeted and no device to spend it on. :(
Now that truly is a completely unacceptable situation! ;)

However, I really posted to find out what you know about the overo? I have a couple of early gumstix boards (not OMAP based). The overo looks good - but I've been having a hard time finding out much about it - especially what daughter boards might be available.

Based on my experiences with gumstix I'd stick with the BeagleBoard route if I were you (for example I bought their LCD pack only to find that there's problems because it's an 18bpp display and there's not very good linux support - but that's what I get for not doing enough homework before buying). The gumstix crowd comes mostly from a robotics background and so things like LCDs aren't their priority. An important caveat is that I haven't been monitoring the gumstix for a good while now (probably 10 or 11 months) - so obviously a lot will have changed.

Y'know the LCD has always been my sticking point with all the projects I want to do. I want to create a simple PDA sized device with lots of USB ports and nothing more than an on/off button. I'd have lots of uses for such a device.

(Sorry this is all a bit OT).
 
Last edited by a moderator:
Ive come across some stuff googling where some have been running it on the beagleboard, which is probably the next best thing to dev on at the moment.
qemu is the only emu that ive heard of that simulates a ARM device.
 
jdh2550 said:
However, I really posted to find out what you know about the overo? I have a couple of early gumstix boards (not OMAP based). The overo looks good - but I've been having a hard time finding out much about it - especially what daughter boards might be available.

Based on my experiences with gumstix I'd stick with the BeagleBoard route if I were you (for example I bought their LCD pack only to find that there's problems because it's an 18bpp display and there's not very good linux support - but that's what I get for not doing enough homework before buying). The gumstix crowd comes mostly from a robotics background and so things like LCDs aren't their priority. An important caveat is that I haven't been monitoring the gumstix for a good while now (probably 10 or 11 months) - so obviously a lot will have changed.

Y'know the LCD has always been my sticking point with all the projects I want to do. I want to create a simple PDA sized device with lots of USB ports and nothing more than an on/off button. I'd have lots of uses for such a device.

(Sorry this is all a bit OT).
13.
Yeah.. I was hoping to get a Beagle Board to while away the weeks until the Pandoras ship. I fully agree that it's a better solution (and probably a better value for the money, to boot). Unfortunately, the BB mailing lists indicate there won't be any more available until December/January when the Rev C boards are released. By then, I may as well hunker down and wait for my Pandora to arrive. :D

I'd read somewhere that the 16-bit LCD screens were a safer bet with the gumstix boards. With the Overo, though, it looks like you wont' need one. The Summit expansion board is the only one available for the Overo on their web site, and it has a DVI-D jack. No word on drivers or resolution, but I'm still researching. Sadly, theovero and expansion boards are on backorder, and I'm fairly certain that they wouldn't get here before the Pandora if I ordered them now. :(

Oh, well. Back to waiting...


EDIT: Just when I'd given up waiting, Digikey has Beagle Boards in stock today, Nov 13.
 
Last edited by a moderator:
_Sterling said:
I'd read somewhere that the 16-bit LCD screens were a safer bet with the gumstix boards. With the Overo, though, it looks like you wont' need one. The Summit expansion board is the only one available for the Overo on their web site, and it has a DVI-D jack. No word on drivers or resolution, but I'm still researching. Sadly, theovero and expansion boards are on backorder, and I'm fairly certain that they wouldn't get here before the Pandora if I ordered them now. :(


That's an improvement - a few weeks back the Overo wasn't even available for order - so things are progressing. An Overo + Summit would be a nice development system - but not sure how it compares to the BB. I notice that gumstix now do a Linux Pro Pack now (for $419 based on the PXA270) - I need to find out what the differences between that and what I've got are. Maybe I can sell what I've got on ebay.

Maybe I should buy a BB.

Maybe I should just wait a little bit more (never been very good at that!)

So much cool stuff so little time! B)
 
Last edited by a moderator:
_Sterling said:
If I recall correctly, I was trying to use qemu to create a sandbox, and couldn't get Angstrom running (have a google at that one and you'll see I'm not alone there) and had to install debian.

I've had Angstrom running quite nicely on qemu, although it was only ARMv5 (kernels and root tarballs can be had here if they're of any use to anyone).

It looks like ARMv7 - even the Cortex A8 explicitly - is supported by qemu, but my problem was that there are no machines defined which make use of it (I base this purely on the output of qemu-system-arm -cpu ? and qemu-system-arm -M ? respectively). Then again, even if there was a machine defined, there's no guarantee of finding a pre-built Angstrom for it and, not being familiar with the whole bitbake thing, rolling my own isn't really an option just yet. I figured I might as well just wait for the real thing and therefore haven't looked any deeper into it since.
 
Last edited by a moderator:
Svartalf said:
Better get your order in, then... Those are worth having... ;)
Yeah. Did that before I edited my post. :D


@SteveM -- I think where I may have gone wrong was I downloaded the Beagleboard (armv7a) demo image. I extracted the kernel and initrd files, but they wouldn't boot. In retrospect, this was incredibly stupid of me, since I already knew I couldn't get qemu to emulate armv7a <_< Not sure what I was thinking that day. Could have just been that my frustration level was through the roof by the time I tried it and I wasn't thinking straight.

Oh, well. Problem solved with a quick trip to Digikey this morning. I should be happily tinkering away by early next week.
 
Last edited by a moderator:
jdh2550 said:
Based on my experiences with gumstix I'd stick with the BeagleBoard route if I were you (for example I bought their LCD pack only to find that there's problems because it's an 18bpp display and there's not very good linux support - but that's what I get for not doing enough homework before buying). The gumstix crowd comes mostly from a robotics background and so things like LCDs aren't their priority.
I've owned several gumstixs for robotics and other projects and have never had any problems - they do exactly what they say on the tin. I've had multiple LCDs connected as well, and have found the most difficult part being building a break-out board for the LCD connector and power regulation - some LCDs are really horrible to power. Why oh why must they have 6 different input voltages? That's a rhetorical question actually, many LCDs these days are much more friendly.

But seriously, buy a Beagle Board - gumstixs are extremely flexible but the BB is perfect for desktop development and testing...if I was doing an Overo project now I'd probably get a BB.

Personally I'm waiting for the Rev C BB which will have the LCD signals on a header and the USB issues resolved (actually they might be sorted already, I forget), though I'm torn between that and a bare-bones Pandora.
 
Last edited by a moderator:
Back
Top