Is this a true statement?
Tim Daneliuk
tundra at tundraware.com
Tue Jun 26 14:50:03 EDT 2001
Andrew Dalke wrote:
>
> Software is used for more things than "Computational Power."
> Take again the example I posted a few days ago on allocating in
> shared memory.
>
> Suppose I have a hardware environment where there is special
> shared memory hardware which can be used to pass data between
> otherwise different processors. One example is the CAVE
> visualization system, which is an immersive reality system
> where images are projected on the walls of a room to make it
> look like you are inside of a synthetic environment -- the
> closest we can get to Star Trek's holodeck. One processor
> is used to generate the visualization model into shared memory,
> then other CPUs (one for each wall) gets the model to generate
> the appropate image with the correct viewpoint.
>
> In C++ I can define how 'new' and 'delete' work on a given
> class, so that I can allocate and deallocate the objects
> to be shared from the shared memory arena while leaving
> the other objects (like file I/O) in local non-shared (and
> more plentiful and cheaper) memory.
>
> There is no way nor even a proposal of a way to use special
> memory allocation routines for specific object, at least not
> in a very Pythonic fashion. So using different memory arenas
> is an example where you can "do" something in C++ which is
> much harder to do in Python. It is not a limitation or
> restriction of the implementation of Python - it's in the
> language itself.
>
> To bring it back to Turing machines, the concept of "shared
> memory" or "I/O ports" or "interrupts" or "signals" does not
> exist in a TM. To talk about them in a TM context requires
> emulating their physical behaviour, and it is that emulation -
> that translation from a language description to a physical
> device - that some languages do not provide.
>
> Andrew
> dalke at acm.org
What prevents me from writing out a byte string from Python which is a set
of primitives for accessing the SM hardware and then building up higher
level Python interfaces from these primitives?
--
------------------------------------------------------------------------------
Tim Daneliuk
tundra at tundraware.com
More information about the Python-list
mailing list