Subversion Repositories openrisc_me

#

#  $Id: README,v 1.2 2001-09-27 12:00:16 chris Exp $

#

Notes about the MVME147S bsp

MVME147 port for TNI - Telecom Bretagne

by Dominique LE CAMPION (Dominique.LECAMPION@enst-bretagne.fr)

May 1996

This bsp is essentially based on the mvme136 bsp,

and is only an extention of the MVME147 bsp. There

are slight differences in the VMEchips used in the MVME147S and

the MVME147 that should prevent the shmsupport and the startup

code from running on a MVME147.

Summary of the modifications that were made:

 * include

   - bsp.h

      Peripheral Channel Controller memory mapping

      Z8530 memory mapping

      VMEchip memory mapping

 * startup

   - bspstart.c

      main () setup for VME roundrobin mode

              setup for the PCC interrupt vector base

              setup of the VME shared memory

   - bspclean.c

      bsp_cleanup () disable timer 1 & 2 interruptions

   - linkcmds set the RAM start (0x7000) and size (4Meg - 0x7000)

   - setvec.c unchanged

   - sbrk.c unchanged

 * console

   - console.c taken from the dmv152 bsp (Zilog Z8530)

               with no modification

 * clock

   - ckinit.c entirely rewritten for the PCC tick timer 2

 * timer

   - timerisr.s and timer.c

     entirely rewritten for the PCC tick timer 1

     now gives results un 6.25 us units (mininum timer delay,

     suprising big grain)

 * times

   - updated results for the mvme147 (beware of the 6.25 us grain)

 * shmsupp :

   Specific to the S version of the MVME147

   Only tested with 2 boards, in interrupt mode.

   Uses the top 128k of the VME system controller board RAM

   as the shared space.

   - mpisr.c : uses the SIGLP interruption

   - Makefile : unchanged

   - getcfg.c : rewritten

   - lock.c unchanged

   - addrconv.c unchanged

 * Makefiles

   - unchanged

To be done:

 * update the overheads in coverhead.h

 * add support for serial ports 2,3 and 4.

Other notes:

 * All the timing tests and sp tests have been run except tmoverhd.

   The test hangs during the pause (where the task should be suspended

   until a return). Maybe the rtems_initialize_executive is no more

   reentrant with this bsp.