|
Does this subject appeal to the masses here?, and would anyone be interested in it?
I have it on my site and it gets a lot of hits but don't know if it would be an interesting addition here.
|
|
|
|
|
I don't do much assembler, but I'd be interested in reading articles on the AVR.
|
|
|
|
|
I know there are a few AVR enthusiasts here on CP but didn't know if anyone used assembler.
I don't use it very ofter as I'm trying to push C++ now that the Atmel Studio supports it but it is handy to know and the more you know about the hardware the better.
|
|
|
|
|
C++ is a hard push on any embedded system, most are still stuck in C. I tried to convince an old developer to use C++ on an ARM once, he nearly had me crucified.
|
|
|
|
|
It's hard to get people to change I've been in a couple of positions where I've pushed C++ even to the point of writing a framework, giving classes and such but they were set in there ways and I finally just did my own thing and left them to there's.
I'm using the ATMega1280 and ATMega2560 boards for my C++ projects and the have 128K and 256K flash memory respectively so there's no "limited memory" excuse and since C++ runs just about as fast as C that shouldn't be a problem. The IDE has finally started to support it and I've been writing a lot of libraries in C++.
|
|
|
|
|
Mike Hankey wrote: I'm using the ATMega1280 and ATMega2560 boards for my C++ projects and the have 128K and 256K flash memory respectively so there's no "limited memory" excuse and since C++ runs just about as fast as C that shouldn't be a problem. The IDE has finally started to support it and I've been writing a lot of libraries in C++.
The "memory problem" rather lies in the fact that even on a Mega2560, you still have only 8KB of RAM. So you have to still be fairly conscious about any dynamic memory allocation (avoid virtual methods at all cost) on top of the fact that AVR MCUs are prone to lock up on heavy memory fragmentation...
And the Arduino IDE was always using C++, the "sketch" that you write is nothing more than an include file for a main() file wrapper and the "Arduino language" isn't actually more than using methods of the default libraries (which in turn are written in C++). After all, the compiler underneath is nothing more than a AVR targeted GCC (ok, there's now a ARM option to support the Due and Intel has their own version of the IDE to support the x86 in their Galileo board).
Ralf 
|
|
|
|
|
Bitbeisser wrote: The "memory problem" rather lies in the fact that even on a Mega2560, you still have only 8KB of RAM
Agreed but that is always a consideration no matter what language you are using.
Bitbeisser wrote: (avoid virtual methods at all cost)
Why? It just uses a virtual table right?
I tend to stay away from doing to much dynamic allocation because as you say memory does get fragmented easily and the memory management sucks.
|
|
|
|
|
Mike Hankey wrote: Bitbeisser wrote: The "memory problem" rather lies in the fact that even on a Mega2560, you still have only 8KB of RAM
Agreed but that is always a consideration no matter what language you are using.
But with C++ and the craze of a lot of people to inherit the heck out of it, it is much more likely becoming a problem as it isn't as obvious how much RAM a certain object/method is using...
Mike Hankey wrote: Bitbeisser wrote: (avoid virtual methods at all cost)
Why? It just uses a virtual table right?
Correct. And where do you think those VMTs are created?
Ralf
|
|
|
|
|
I guess my point is in my Init method I create say an CSpi object and use it for the lifetime of the run I don't create and destroy every time I use it.
I agree when you're working with limited resources of any kind you need to be conscious of what you're doing but I believe the price is worth it for the extra benefit gained from C++.
I believe we're saying the same thing!
|
|
|
|
|
Bitbeisser wrote: (avoid virtual methods at all cost)
Why? It just uses a virtual table right?
That is the point, those VT are using up (a lot!) of the valuable RAM space...
Ralf
|
|
|
|
|
I for one would be most interested in an article of that nature. With the 328p having such limited program/memory space, I'd love to see such an article.
|
|
|
|
|
I have done a couple of project using the 328P and C++ and it works very well.
|
|
|
|
|
Indeed. I've done a couple too. But as with desktops, sometimes a c/c++ just doesn't seem to cut the mustard. (using avr-gcc here)
One cant be sure that a register value wont be chucked away or that a register wont be loaded twice. Specifically, I'm playing with (a) a musical instrument synthesizer with waveform display and am looking to add the ability to add tracking info - i.e Make an (amiga) .mod player. (b) a cheap and nasty DSO/bus sniffer (c) a 6, 7 or 8 channel RX compatible with the protocols used by both the Turnigy 9x and Husban transmitters. Every byte left is well appreciated, sine they each use a channel-hopping protocol and have rather short frame-times. I need a couple of largish static buffers and few wasted cyles. To maintain a lock and avoid dropping packets, one needs to be rather efficient with the 328. I'm able to successfully change channel between each packet, but still cant capture more than about 5 or 6 packets in a row before I drop one. Interference isn't the problem, it's just a matter of timing. So memory for the channel hopping tables and speed to avoid packet loss are my main interests. This also means that an a7105 2.4ghz module driven by the 328 doesn't satisfy the binding requirements of either a commercial Turnigy/Hobby-King RX.
|
|
|
|
|
|
A lot of hobbyists use the Arduino IDE which has a lot of C++ code and I don't use it so haven't delved to deep in it. But most people that use it seriously use Asm or C for production!
|
|
|
|
|
It certainly will be. I am learning AVR Assembler as a part of my Apprenticeship, using the STK 600 Board[^]. We are currently implementing a Real Time Clock, being triggered by the interrupt signal. Maybe I am going to put it up here when it is finished.
Veni, vidi, caecus | Everything summarizes to Assembly code
|
|
|
|
|
|
I'd be very interested and what's to loose if you publish it also here 
|
|
|
|
|
As you say nothing ventured nothing gained. I'm working on converting it now and should be ready in a couple of hours.
|
|
|
|
|
|
I don't know AVR (I'm on the Microchip side) but would be interested.
Veni, vidi, vici.
|
|
|
|
|
|
Great! I am really looking forward to studying that over the weekend.
|
|
|
|
|
Cool let me know what you think!
Enjoy!
|
|
|
|
|
I'm working on AVR so wouldn't mind you posting on it. But whether anyone else would be interested on CP I don't know.
"Program testing can be used to show the presence of bugs, but never to show their absence."
<< please vote!! >>
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
