Hello ng

* Is it possible to write kernel modules in D?

* Is it possible to write realtime apps in D?
  For example using the rtai for linux.

* If any, what are the missing requirements?

Frank

In article <dn67ub$1u3c$1@digitaldaemon.com>, Frank Benoit says...
>
>Hello ng
>
>* Is it possible to write kernel modules in D?
>
>* Is it possible to write realtime apps in D?
>  For example using the rtai for linux.
>
>* If any, what are the missing requirements?
>
>Frank

Perhaps these will help:

http://digitalmars.com/d/iasm.html http://digitalmars.com/d/memory.html http://digitalmars.com/d/interfaceToC.html

Frank Benoit schrieb am 2005-12-07:
> Hello ng
>
> * Is it possible to write kernel modules in D?

http://dkernel.kuehne.cn/hello.d

Thomas

Thanks for the replies.


it seems to me, D can be an alternative to C/C++ in realtime programming.

From http://www.digitalmars.com/d/memory.html#realtime
"The most reliable way to guarantee latency is to preallocate all data
that will be needed by the time critical portion."

What are the things to avoid?
- every new
- dynamic arrays in general?
- slicing arrays
- concatenating strings?
- ...?

Can I make a dynamic allocation protection. A mechanism that throws an exception if the programm makes a direkt or indirekt allocation.

The disable/enable of the gc, does only throw outofmemory. Is there a possibility to get the exception at the first unwanted allocation try?

void InterruptServiceRoutine()
{
	std.gc.disable();

	try
	{
		/// all the realtime things
	}
	catch( OutOfMemory oom )
	{
	}
	finally
	{
		std.gc.enable();
	}
}

Does the gc and D work with RTAI Lxrt? http://www.rtai.org/documentation/magma/html/api/whatis_lxrt.html There is this:

"They must be POSIX real time Linux processes locked into memory using SCHED_FIFO. Thus their memory must be pre grown to its maximum extension and completely locked in memory."

I do not understand that completely, especially not the consequences for D and the gc.

Frank

In article <dndh3a$289t$1@digitaldaemon.com>, Frank Benoit says...
>
>Thanks for the replies.
>
>
>it seems to me, D can be an alternative to C/C++ in realtime programming.
>
>From http://www.digitalmars.com/d/memory.html#realtime
>"The most reliable way to guarantee latency is to preallocate all data
>that will be needed by the time critical portion."
>
>What are the things to avoid?
>- every new
>- dynamic arrays in general?
>- slicing arrays
>- concatenating strings?
>- ...?
>

I would say the basic rules would be:

- avoid the built-in new (you can override it - that's covered in the docs.).
- use static arrays instead of the built-in dynamic arrays. The compiler
disallows concatenation and
- don't use .dup on any type of array.

Slices return a reference to the array and, for D style arrays, also modify the length member. But they should not allocate memory.

>Can I make a dynamic allocation protection. A mechanism that throws an exception if the programm makes a direkt or indirekt allocation.
>
>The disable/enable of the gc, does only throw outofmemory. Is there a possibility to get the exception at the first unwanted allocation try?
>
>void InterruptServiceRoutine()
>{
>	std.gc.disable();
>
>	try
>	{
>		/// all the realtime things
>	}
>	catch( OutOfMemory oom )
>	{
>	}
>	finally
>	{
>		std.gc.enable();
>	}
>}
>
>Does the gc and D work with RTAI Lxrt? http://www.rtai.org/documentation/magma/html/api/whatis_lxrt.html There is this:
>
>"They must be POSIX real time Linux processes locked into memory using SCHED_FIFO. Thus their memory must be pre grown to its maximum extension and completely locked in memory."
>
>I do not understand that completely, especially not the consequences for D and the gc.
>
>Frank