Python on an embedded platform

[Derek]

Hi,

I am looking to port Python to an embedded platform (an ARM7 device with
fairly limited memory, capable of running an RTOS, but not an OS, such as
Linux). I came across DePython from a few years ago, but it seems to have
died a death.

Does anybody have advice? I am looking for any tricks, features I can
disable, etc so I can get the python core to be a small as possible (<100k
would be good).

Thanks
--
Derek

[Martin v. Löwis]

Derek wrote:

Does anybody have advice? I am looking for any tricks, features I can
disable, etc so I can get the python core to be a small as possible (<100k
would be good).

Please start with describing how far you got. What did you already do,
and how far did it get you?

My first recommendation is to disable all modules which you don't need...

Regards,
Martin

[Grant Edwards]

On 2006-01-19, Derek <dmsubs@_NOSPAM_dsl.pipex.com> wrote:

I am looking to port Python to an embedded platform (an ARM7
device with fairly limited memory, capable of running an RTOS,
but not an OS, such as Linux). I came across DePython from a
few years ago, but it seems to have died a death.
The other project you might want to take a look at is pippy:
http://pippy.sourceforge.net/. It's python 1.5.2 running on
PalmOS. The memory size is quite a bit less constrained that
the DePython project, but there might be some useful info.
Does anybody have advice? I am looking for any tricks,
features I can disable, etc so I can get the python core to be
a small as possible (<100k would be good).

At one point I was interested in doing something similar:
running Python on an ARM7 w/o virtual memory under eCos in
something on the order of 1MB. After looking at the DePython
and Pippy sites, I decided it was going to require more work
than I could justify.

--
Grant Edwards grante Yow! I guess we can live
at on his POT FARM in HADES!!
visi.com

[Kay Schluehr]

Derek wrote:

Hi,

I am looking to port Python to an embedded platform (an ARM7 device with
fairly limited memory, capable of running an RTOS, but not an OS, such as
Linux). I came across DePython from a few years ago, but it seems to have
died a death.

Does anybody have advice? I am looking for any tricks, features I can
disable, etc so I can get the python core to be a small as possible (<100k
would be good).

Thanks
--
Derek

Have you thought about using a JVM as the Python runtime? AFAIK the
PyPy team already uses a restricted subset of Python ( called RPython )
to translate type annotated RPython flowgraphs to several backends
including Java. Therefore it seems not necessary to write a new Jython
compiler or CPython runtime. JVMs are all around also for ARM
processors.

Kay

[Grant Edwards]

On 2006-01-19, Kay Schluehr <ka**********@gmx.net> wrote:

Does anybody have advice? I am looking for any tricks, features I can
disable, etc so I can get the python core to be a small as possible (<100k
would be good).

Have you thought about using a JVM as the Python runtime? AFAIK the
PyPy team already uses a restricted subset of Python ( called RPython )
to translate type annotated RPython flowgraphs to several backends
including Java. Therefore it seems not necessary to write a new Jython
compiler or CPython runtime. JVMs are all around also for ARM
processors.

With footprints in the 100K range?

--
Grant Edwards grante Yow! I decided to be JOHN
at TRAVOLTA instead!!
visi.com

[Paul Rubin]

Grant Edwards <gr****@visi.com> writes:

JVMs are all around also for ARM processors.

With footprints in the 100K range?

For the JVM itself, that's no big deal (look at J2ME). JVM plus
Python runtime might be harder.

[Kay Schluehr]

Grant Edwards wrote:

On 2006-01-19, Kay Schluehr <ka**********@gmx.net> wrote:

Does anybody have advice? I am looking for any tricks, features I can
disable, etc so I can get the python core to be a small as possible (<100k
would be good).
Have you thought about using a JVM as the Python runtime? AFAIK the
PyPy team already uses a restricted subset of Python ( called RPython )
to translate type annotated RPython flowgraphs to several backends
including Java. Therefore it seems not necessary to write a new Jython
compiler or CPython runtime. JVMs are all around also for ARM
processors.

With footprints in the 100K range?

Yes, definitely. The smallest JVM I've seen was dedicated for an 8-Bit
AVR with 8 KByte EEPROM ( or was it Flash? ) and a 768 byte sized heap.
JVMs are also a common technology on SmartCards but those are not open
source.

Kay


--
Grant Edwards grante Yow! I decided to be JOHN
at TRAVOLTA instead!!
visi.com

[Paul Rubin]

"Kay Schluehr" <ka**********@gmx.net> writes:

Yes, definitely. The smallest JVM I've seen was dedicated for an 8-Bit
AVR with 8 KByte EEPROM ( or was it Flash? ) and a 768 byte sized heap.

That sounds like Javacard, which is a subset of a JVM (e.g. a minimal
Javacard implementation doesn't need garbage collection). JVM's can
be quite small, but not THAT small.

Actually you might mean this:

http://www.harbaum.org/till/nanovm/