The problem is that without JIT or AOT (broadly, “native”) compilation, even a 1 GHz monster wouldn’t be able to handle those tasks. Given native compilation, even the Cerb* and Hydra boards could handle those tasks. The CIL interpreter really is THAT slow. On top of that, no amount of GHz (or even native compilation) would free you from the non-preemptive 20ms thread quanta. That means jitter in signal generation, signal decoding, and all hardware interrupt handling.
100% agree. I believe it’s about volume. If only GHI could sell the millions that RaspPi does (and being subsidized by the SoC manufacturer doesn’t hurt either)
I use GHI controller boards and NETMF as a hobbyist. My main application is home automation and security. Currently in my home I use HomeSeer running on a small Windows XP computer. HomeSeer is not real-time, but it is quick enough to do everything I need to do in a home automation context. I would love to replace HomeSeer with an NETMF based controller board, running all my own software, but unfortunately the boards available to me aren’t quite powerful to do that yet.
At RaspPi prices? Absolutely, but I’ve never tried to do anything that my Cerb wouldn’t handle (or even my Panda II, for that matter…).
If I’m looking at it strictly from a performance (or even performance/$) perspective, not even RaspPi can beat the local university surplus, where I can pick up a 2.4 GHz Core 2 Duo with 2GB of RAM for ~$45. I could put Windows on that and run NETMF in the emulator, and have the cheapest, most powerful NETMF “board (even comes in a metal case!)” in the world, but that would be a bit silly (I love arguing ad absurdum).
I get a thrill by doing more with less, so it’s the bottom end that excites me, not the top end. Anyone can work magic with enough resources; I want to work magic using two sticks and a half-wad of bubble gum
If you want to program in C#, you can install mono.
My Raspberry PI has only 256MB and you can´t Develop on it.
Monodevelop is running, but… it is very slow. For small programs or asp.net webserver it should work. I try it in the near future.
I would like to be able to:
[ul]Load multiple runtime loadable plug-ins (yes, I know it can be done on NETMF, but there are limits regarding memory, etc)
Support multiple devices plugged into USB host port.
Support voice-recognition and speech output.
Support media server plug-in that can serve FLAC music files to multiple clients.
Support high-resolution monitors or other display device (not necessarily HDMI).
Load large files from SD card (e.g., music files).[/ul]
I know NETMF can do some of these things, but when you want to do all of them you quickly run up against speed and memory issues.
@ godefroi you are right, sometimes stuff we build doesn’t need to be a small form factor device and so a surplus computer is great fit in some cases. I’m a firm believer in the right tool for the job and don’t expect every tool to do every job but some people seem to think that is possible.
As for USB host, that’s a software limitation of GHI’s libraries, not a hardware limitation, I believe.
As for runtime loadable plug-ins, why? Why not simply build the functionality into the software? Embedded development and plug-ins are generally not used together (heck, the embedded purists believe that these systems should have no heap), but there’s no reason why you can’t. Hydra should have plenty of memory to load all the plug-ins you could ever want.
As for the rest, it sounds like what you want is a media player. Should be quite simple to implement. Grab yourself a VS1053 breakout and an SSD1289 touch LCD for the UI. Voice recognition and speech generation are going to require more horsepower than you have available, but like I said, even a Pi running NETMF likely couldn’t do it. The overhead of interpreted code is too heavy.
Certainly that’s not my argument. If you can pack more power, memory, and resources into the same physical size, cost, and power consumption, then by all means, go for broke. Ad absurdum, after all.
It’s just my belief that 90% of the limitations people run into with NETMF have nothing to do with CPU speed (or, at least, not enough to do with CPU speed that a 1GHz CPU would overcome them).
@ jasdev - We support multiple devices (HUBs) and everything else mentioned should run fine with existing offers. More speed will make it better but ChipworkX or faster board should run everything quite nice.
[ul]Eventually I would want to provide the ability for other developers to create open source plug-ins that myself and others can use on their home devices, so these plug-ins must be runtime loadable.
I don’t want a media player, I want a media server, capable of serving multiple FLAC music streams.
I need to be able to buffer several very large data files in memory as I’m streaming them out over the network.
I would like to be able to support multiple IP and/or USB cameras with the ability to capture, store, and forward these images over the network.[/ul]
When you say “streams” what do you mean? RTSP? SMB? HTTP?
But if they’re stored and read from flash (SD), then they’re already “buffered” in memory? Is there some particular reason that this “buffer” must be in SDRAM instead of flash?
That would definitely be doable, but you’d need native code if you were planning on implementing it from scratch.
