Enabling lxrt in Makefile for ARM
mantegazza at aero.polimi.it
Fri Jan 16 12:48:26 CET 2004
Elder Costa wrote:
> Paolo Mantegazza wrote:
>> Since you are speaking of an LXRT directory that means you are
>> working on a 24.1.xx version. If you need to stay there for any
>> reasong please use NEWLXRT, without RTAI proper tasks. The LXRT
>> scheme is dead in 3.0.
> Yes. 24.1.13 plus fixes (cvs). We are using the divide and conquer
> approach: I am assuming porting to lxrt will be easier than to newlxrt
> as we have to learn RTAI/LXRT internals during the process and debug the
> generated code and involves only the processor specific functions to
> have LXRT working whereas NEWLXRT involves a new scheduler as well so
> there is two fronts to work with. As soon as I have LXRT working I
> presume it will easier to perform the changes to newlxrt.
> I intend using Paul Coene's patches as a reference but the work will be
> mostly porting x86 related code as there are too many changes since
> those patches have been released.
> If my approach is flawed I do appreciate comments and recommendations as
> we are in the early stages.
It is not flawed it is historically similar to what went on in my mind
at the very beginning of RTAI inception in user space, when we thought a
layered approach, i.e LXRT, was simpler. It turned out that NEWLXRT is
the simplest eventually. I agree that once the idea is understood
passing to NEWLXRT will not be difficult. What is wrong is to think that
NEWLXRT is another scheduler completely, while it is just the RTAI MUP
one, with another way of switching schedulable objects, as it switches
Linux tasks. or the rest is the same either if you use LXRT or NEWLXRT.
Recall also that under 3.0 there what is called LXRT is NEWLXRT.
BTW, from what I recall Paul's work was with NEWLXRT.
More information about the Rtai