RLP Lite (observations)

[twospoons]

I like using bin files instead of ELF files. It requires a little more code to manually set up entry points, but they are a LOT smaller in size.
The downside to this is that the code is not portable between different CPUs.

Allowing for both would be a nice touch.

The requirment for Keil (or at least the headers and what not for keil) should be provided by GHI, so if you wanted to use a different compiler it would be easier get started.

Invoking functions without arguments work, but you have to pass in dummy arguments. And if you need more than 3 arguments then you are out of luck.

The performance of native on the Cerb seems a lot faster than on the fez panda II (using RLP). Not sure what the reason is; but a combination of a better compiler (keil tool chain), much faster CPU and less overhead since dynamic arguments are not support would be my guesses.

Eg;
c:
void test() {
return 5;
}

c#
var p1 = byte[0];
var p2 = int[0];
test.Invoke(p1,p2,p1);

Should be:
test.Invoke();

Nulls cannot be used since there is no overloaded function supporting that.

Wish list:
[ul]Add ELF support
Allow for dynamic arguments (like RLP)
Merge RLP with RLP Lite and just call it Native
Add support for yagarto[/ul]

[ianlee74]

@twospoons

And if you need more than 3 arguments then you are out of luck.

Not really. Each of those three arguments are arrays. So, you can have as many arguments as you want. The problem comes in mixing more than three [data] types of arguments.

The performance of native on the Cerb seems a lot faster than on the fez panda II (using RLP). Not sure what the reason is

72Mhz vs 168Mhz maybe?

Nulls cannot be used since there is no overloaded function supporting that.

Now that does seem unusual.

Wish list:
[ul]Add ELF support
Allow for dynamic arguments (like RLP)
Merge RLP with RLP Lite and just call it Native
Add support for yagarto[/ul]

If you’re going to be writing a lot of RLP, I suspect you’ll be a lot happier with the premium boards. It shouldn’t be too long before the G120 is available.

[twospoons]

Clock for clock the cortex beats the 72mhz panda II. So there is more to it than just raw clock speed.

Also, you can solve any problem with just having a byte array (one param), it just gets really inconvient when calling it.

[ianlee74]

@twospoons

Clock for clock the cortex beats the 72mhz panda II. So there is more to it than just raw clock speed.

It has some built-in floating point math features. So, if your test involves floating point math then I would expect it to be able to skip some cycles that the Panda couldn’t.