Perfomance problems and main stream

[leonp at plris.com (Pollak Leon)]

At 09:51 23/02/99 +0100, VALETTE Eric wrote:
>I'm really tired to write drivers nearly from scratch for new boards and
>having to understand all the complexity of processor/system dependent
>interrupt mechanism, cache species, DMA handling...
>All this can/should be hidden by PROCESSOR API/BSP API.
	This is very desired but IMHO almost impossible situation in "REAL(!)
time" applications, where you always try to pump out maximum performance
(and minimum code). In these cases the only way that I know is to optimize
the BSP's environment tightly to your specific HW configuration and SW
application. And if you don't need to do this it means that you paid extra
money for the more powerful CPU you really need, which is not commercial
approach.


>Quality Quorum wrote:
>BSP should represent no more than common starting point for various
platforms...
	This is exactly my opinion. After I came to the situation when my
BSP+RTEMS is running I changed almost everything in BSP - all drivers,
interrupt prologue, context switch, even initialization structure.
	And I shall never agree to leave in my BSP any code that I, as minimum,
don't understand what it is doing. In almost all commercial cases I passed
through in my life, after studying the "demo" code, I found how to improve
it IN MY SPECIFIC CASE. I don't want to say that it was bad, but it was too
generic.


	A lot of thank to all of you for the discussion. I was afraid that this is
only my personal problem.


Dr.Leon M.Pollak
Director
PLR Information Systems Ltd.
E-mail: leonp at plris.com