PSIM Support for MPC860

PSIM Support for MPC860

[Tom Taylor]

I'm attempting to use the GDB 5.0 PSIM simulator to analyze firmware
written for the MPC860T in the OEA mode.  I'm now aware of a
considerable amount of work that I apparently need to do to add at least
a partial simulation of the many added SPR registers, beginning with the
data cache control/status register DC_CST (#568).  Due to pre-C0 data
cache bug workarounds, it is not possible to complete even the basic
boot initialization including the DEC register setup without accessing
this SPR.

I have noticed that there is special code for handling problematic
MPC860 forward branches, which is referred to as option_mpc860c0.  This
implies to me that at least one person has used the simulator for 860
code, but I'm puzzled because there doesn't appear to be sufficient
support for this to work.  Is this because only application code was
being simulated, or is there additional support modules that haven't
been released as part of GDB?

If I undertake the task of adding MPC860 support to PSIM, is there any
interest in the GDB community for simulating this processor, and is
there any available code that could be used to reduce the time and
effort this task will require?

Tom Taylor
Taylor Consulting Services

Re: PSIM Support for MPC860

[Frank Ch. Eigler]

"Tom Taylor" <ttaylor@ateng.com> writes:

: I'm attempting to use the GDB 5.0 PSIM simulator to analyze firmware
: written for the MPC860T in the OEA mode.  [...]

You might have a big job ahead of you.

: I have noticed that there is special code for handling problematic
: MPC860 forward branches [...]  Is this because only application code
: was being simulated [...]

Yes.

: If I undertake the task of adding MPC860 support to PSIM, is there any
: interest in the GDB community for simulating this processor, 

Sure.

: and is there any available code that could be used to reduce the
: time and effort this task will require?

I don't know of any.  Building satisfactory simulations of complex
processors tends to be a big job.  (Red Hat does this sort of thing on
contract basis for example.)


- FChE

Re: PSIM Support for MPC860

[Tom Taylor]

"Frank Ch. Eigler" fche@redhat.com writes:
:
: "Tom Taylor" <ttaylor@ateng.com> writes:
:
: : I'm attempting to use the GDB 5.0 PSIM simulator to analyze firmware
: : written for the MPC860T in the OEA mode.  [...]
:
: You might have a big job ahead of you.
:
No doubt.  However,  I only need a limited simulation at first to verify
that
BSP, driver, and application code previously compiled and linked by my
client
using Diab Data tools will operate properly after compilation using GCC
and
linking with pSOS+ and pNA+ libraries.  My main concern has been
variable
length argument passing, since Diab Data uses a modified, two argument
version
of the sizeof() operator for walking the arg list.  Other concerns are
the correct
translation of Diab-specific '%' function-style operators for loading
32-bit
immediate and address values into more general at-sign suffix versions
supported
by GCC, e.g. using "symbol 'at' ha" instead of "%hiadj(symbol)", as well
as
replacement of extended section directives and the like.

CPM simulation will have to be severely limited.  Fortunately, UPM
emulation should be unimportant.  I believe that a comprehensive,
general
simulation of the MPC860T would be extremely difficult.  I'm more
interested in allowing analysis of interactions with code on
board-specific
peripherals such as the TI 5420 DSP through the HPI than in supporting
all possible 860T internal peripherals.  Chip-select simulation will be
required.  I guess I'll have to learn a lot about the device tree
approach;
unfortunately it appears that the IEEE 1275 OpenBoot standard has been
withdrawn.

: : and is there any available code that could be used to reduce the
: : time and effort this task will require?
:
: I don't know of any.  Building satisfactory simulations of complex
: processors tends to be a big job.  (Red Hat does this sort of thing on
: contract basis for example.)
:
Thanks, I'll keep the Red Hat contract services in mind.  I'll submit
any of
the more general-purpose code--such as I & D Cache SPR extensions--for
possible inclusion with PSIM when I have it working OK.

Tom Taylor

Re: PSIM Support for MPC860

[Frank Ch. Eigler]

"Tom Taylor" <ttaylor@ateng.com> writes:

: [...]  No doubt.  However, I only need a limited simulation at first
: to verify that BSP, driver, and application code previously compiled
: and linked by my client using Diab Data tools will operate properly
: after compilation using GCC and linking with pSOS+ and pNA+
: libraries.  [...]

Yes; these are compiler/assembler matters.


: CPM simulation will have to be severely limited.  Fortunately, UPM
: emulation should be unimportant.  I believe that a comprehensive,
: general simulation of the MPC860T would be extremely difficult.

Yes.


: I'm more interested in allowing analysis of interactions with code
: on board-specific peripherals such as the TI 5420 DSP through the
: HPI than in supporting all possible 860T internal peripherals.

If you are contemplating building models of a completely different
processor (the DSP chip), please be aware that psim is somewhat
limited in its integration capabilities with models of this rich
nature.  It might be worth your time to investigate other simulation
frameworks such as sid and modelling systems like cgen.  (Both these
are Red Hat projects on sources.redhat.com.)


: Chip-select simulation will be required.  I guess I'll have to learn
: a lot about the device tree approach; unfortunately it appears that
: the IEEE 1275 OpenBoot standard has been withdrawn.

The OpenBoot references in psim are somewhat illusory.  In psim, they
constitute merely a naming convention for the strings that constitute
configuration of the peripheral models.  There is no OpenBoot firmware
or anything like that.


: [...] I'll submit any of the more general-purpose code--such as I &
: D Cache SPR extensions--for possible inclusion with PSIM when I have
: it working OK.

Great.


- FChE