I seem to be the only person that agrees with davidthefat on this issue. I don't believe many people here have tried to do anything they "weren't supposed to" yet.
Quote:
Originally Posted by davidalln
Your statement above is akin to joining the football team, and being disappointed that you didn't end the season a basketball star. There are many different subsections in computer science; FRC focuses on robotic manipulation, computer sensing, and HCI. All of these are rich fields that I have found are able to be explored through my experience working with the bot.
|
I think you are reading too far into his response (although, true, it's the only thing we have to go on) and trying to disregard the problem of closed source, albeit widely used, systems on the cRIO. If I know where he's coming from, that is the most logical train of thought to tell people (not saying this isn't what he actually wants to do); in essence, however, no matter how "open" VxWorks is supposed to be, it's still rather limited.
Example: I wanted to bypass NIVision and skip the hassle of trying to train it to work in variable conditions (having it actually recognize something here, and then at the competition, are two different things) and substitute in a much simpler, static algorithm. On the presumption I can access raw images received from the camera, I start looking at the dashboard code to see if I can markup the image and send a modified one.
I come to find out I can't directly access the image sent to the dashboard (and what I wish to do is more than draw boxes on it), so I put a custom dashboard in the back of my mind as something I needed to do. I can live without the feed for now.
Right, next step, getting the image. After a day and a half of looking, I come to find out I can't actually bypass NIVision's access to the image. I looked around a while and found out the new radio is a combo bridge/access point. I plug the camera into the router instead of the cRIO and access it as such. After searching for a JPEG decoding routine, it turns out to be too slow (the cRIO doesn't use a very current VM at all).
I end up moving all the image processing to the computer, and am developing from there. Anyway, this was all a
really fun process, the cRIO is totally open and accessible!
Quote:
Originally Posted by davidalln
Mess with multithreading and optimization!
|
Those are considered rather high-level concepts. Optimization can take place in machine code, but it is usually done at the source level (and the only application I can see of optimization to FRC is source-level). The implementations of these concepts, however, are what he mentions, and there is no way to implement either (or need to) for the FRC.
Quote:
Originally Posted by davidalln
FIRST hasn't supplied some dumbed down OS designed just for the competition. Interested in compilers? Windriver compiles using gcc, the most widely used completely open source C compiler out there.
|
Arguably, they have. There is no way to access the internals of
anything (should you so desire). You are also not free to see how they do it.
I'm assuming the second part about compilers is a separate thought, but either way, it's still a tenuous link at best.
Too long; didn't read it? "It just bugs me."
@davidthefat: Hope my presumptions about you don't offend, I'll explain them in more detail if you want. Personally, the only thing I can tell people (that makes sense to them) when asked about why I want to know how something works is similar response to your own.