[evla-sw-discuss] TC11IB development environment

[Bill Sahr]

Given below is a statement of what I see as the
requirements for the development environment
for the TC11IB boards, and a possible solution
which satisfies those requirements.  Please read
and comment.  If you see something that I've 
missed, or don't like the proposed solution
please contact me.  I will be writing P.O. Reqs
in just another day or two to purchase the
components for the develoment environment, so
please get back to me no later than the morning
of Fri, 6/21.  After that, we may have to live
with what we get.

Bill
-------------- next part --------------
   date:  06/18/2002

subject:  TC11IB Development Environment, MNT Simulation environment

 author:  Bill Sahr

Development Environment
By development environment I mean the toolset, a debugging connection,
a debugger, and the computing platform(s) to host these elements,
including a network connection between the computing platform and the
MIBs. 

I have a bit of additional information re the TASKING toolset, some 
of which is significant w.r.t. to the environment:

- The JTAG/OCDS interface is available only with the Windows version
  of the toolset.  It is impossible to download an application to the
  development board for on-board debugging via the JTAG/OCDS interface
  with the Unix/Solaris/Linux versions of the toolset unless one
  resorts to third party solutions.

- The PCP & PCP2 are supported for the assembler and the
  linker/locator, but they are not supported in the TASKING
  debugger.  I do not yet know if third party debuggers support
  the PCP & PCP2.

- The TC11IB is supported within the compiler, assembler, linker/
  locator, but not yet in the debugger.  Some manual setup & 
  configuration will be necessary to get the debugger to work
  correctly with our development board

- The TASKING debugger supports only OCDS level 1.

- The Linux version of the TASKING toolset runs under Red Hat,
  versions 6.1 and later.

Requirements 
The development environment must support: 
- The TASKING toolset, ie, the compilers, assemblers, linker/locator,
  and debugger.

- Debug connection for TASKING debugger.  This connection requires a
  parallel port on the laptop, and the "wiggler" connection on the
  development board, plus a straight through parallel port to parallel
  port cable.  (If you want to know the required gender of the
  parallel port connectors, ask George Peck.  I can't remember.)

- Debug connection and debugger, OCDS1 & OCDS2, for the MIB prototype
  board.  This capability would also give us what we need to support
  faster downloads, OCDS1, and OCDS2 for the development board.

- Support for the Nucleus MNT and MNT Binary Bundle simulation
  environment.  I do not want to preclude the possibility of using
  this simulation environment.

- The development & debugging tools must be usable both for bench
  testing and integration at the AOC and for work at an antenna.  It
  should also support working on a module at the VLA site that has
  been pulled from an antenna.

Have I missed anything in the requirements ?

Given all of the above, I am recommending a development environment
consisting of laptop computers running Windows 2000.  IF YOU WANT
A SET OF TOOLS FOR LINUX OR SOLARIS, IT IS IMPORTANT TO SPEAK WITH
ME OR TO SEND AN EMAIL STATING YOUR PREFERENCE.  I will take silence
as agreement with the recommendations.

More specifically, my recommendations are as follows:
- Three (3) laptop computers running Windows 2000 - 1 to be located
  downstairs, 1 for the second floor, and 1 floating unit

- Three (3) licenses for the TASKING toolset.  Unless I am told
  otherwise all three licenses will be for the Windows version of the
  toolset.  Originally, I was planing to order three floating licenses
  to allow for the possibility of running the toolset on machines
  other than the laptops.  I am now thinking that one or more of the
  licenses should be node-locked to give us a self-contained
  development environment on the laptop(s) that can be taken into an
  antenna and used without the need to acquire a license or to
  download toolset components (if they are not installed locally) from
  the test antenna.

  Question.  What is the best mix of floats & node-locked - 2 floats &
  1 node-locked, or 1 float & 2 node-locked ?  I can easily see the
  need to have one hardware & one software developer working in the
  test antenna at the same time, so I am inclined to go with 1 float
  and 2 node-locked.

- Two emulators to support direct connection to the 16-pin OCDS1/JTAG
  connector.

- Two trace modules to support runtime trace execution (OCDS level 2)

- A debugger, in addition to that supplied with the TASKING toolset,
  that will support both OCDS1 & OCDS2 for the TC11IB chip.  Two
  licenses for the debugger.  Again, floats or node-locked ?
  Node-locked seems to make sense here since the laptop must be cabled
  directly to the JTAG/OCDS connectors on the board.

- Additionally, I hope to have the laptops loaded with what is
  becoming the defacto standard for productivity software: MS
  Office (including MS Word, Excel, & Powerpoint), Adobe Acrobat,
  Visio, Exceed, Winzip, and SecureCRT.  VMWARE may be needed for the
  simulation environment (see below). 


Connecting laptops to the MIBs
  The only direct connection between a laptop and a MIB will be via
the OCDS1/JTAG & OCDS2 connectors.  For ethernet/network connections,
we have two cases - 1) connecting to the bench test setups, and 2)
connecting to actual antennas.  For now, I am assuming that the case
of working on a module at the VLA site that has been pulled from an
antenna will be done either via a direct OCDS connection or via the
same scheme that will be used for bench testing at the AOC.

Bench test setups 
James Robnett has been asked by Steve Durand to create three "antenna"
networks in the electronics area - two rack based systems and one
system that is spread across various work areas within the electronics
labs.  These three antenna networks will in turn be connected to the
AOC network.  To connect to a particular MIB, one would simply connect
a laptop to the general AOC network, and then address a particular MIB
using the antenna/slot based IP address, which will always be a
calculatable quantity.  I assume that James will do whatever is
necessary to interface the full duplex, fiber-based antenna networks
to the half duplex, twisted-pair based AOC network.

Will we eventually need a bench test setup, ie, an "antenna" network
in an electronics area at the site to work, at the site, on modules
pulled from an antenna ?

Connecting to actual antennas
This connection would work the same as the bench test setup connection
with the laptop connected to a spare switch port within the antenna.
The only gotcha here is that we must be certain that we can connect
the twisted pair RJ45 laptop connection to the full duplex fiber-based
network in the antenna.


Simulation environment
The simulation environment referred to is Nucleus MNT and the MNT
Binary Bundle.  There is an OS issue in that Nucleus MNT is known to
run under Windows 2000, but the networking components of the MNT
Binary Bundle have not yet been tested under Windows 2000.  Both
Nucleus MNT and the MNT Binary Bundle are known to run under Windows
NT 4.0.  To enable the possibility of using MNT and the Binary Bundle
for software development I suggest the following (developed in
cooperation with Marie Glendenning):

- We purchase laptops for which NT 4.0 and the drivers are still
  available.  The laptops will be purchased from Dell, and Dell
  maintains a site which offers the drivers for older versions of
  Windows, including NT 4.0.  This approach gives us a fallback
  position that allows us to run NT 4.0 and its drivers on the laptop
  should we be forced to do so.

- We initially run Windows 2000 on the laptops.  If, after reading the
  documentation, it appears halfway reasonable to attempt to install
  MNT and the Binary Bundle under Windows 2000, then we do so.

- If installation under Windows 2000 does not work, then we use VMWARE
  to run Windows NT 4.0.  VMWARE will be hosted by Windows 2000.
  Using VMWARE there are no NT 4.0 driver issues.  MNT and the Binary
  Bundle are then installed under Windows NT 4.0 which will be running
  under VMWARE.

- If the VMWARE approach does not work, then we fallback to a native
  Windows NT 4.0 installation on the laptops and install MNT and the
  Binary Bundle under Windows NT 4.0.


Comment and suggestions are welcome.  Please try to think of scenarios
(use cases ?!?) for the use of these laptops that I have not
envisioned.