[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Dynamic Priority



Dynamic Priority
~~~~~~~~~~~~~~~~

Starting from existing occam, we may as well leave PAR as it is.
Although  HARD PAR or SOFT PAR or TRANSPARENT PAR or PAR(HARD) may be
needed.

But we can have an optional CHAN parameter in PRI PAR: I think that this
is the more accurate representation.

PRI PAR(CHAN OF something priority)
  P0
  P1
  ...
  PN

Specifying the POR (Partial Order Relation)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

The syntax gives a reference order. My original idea of sending the ordered
image of a bijection as a table of indices is simple, and I hope lightweight
for the sender to prepare, and for the implementation to utilize. It is
pretty much a preprepared lookup table of offsets.

Gerald raises the idea of a more general order relation. Do we need this?
Will it ever be needed? Anyway, one more general possibility is
to have a list of ordered sets:

S0 > S! > S2....

For example {P0} > {P2,P3} > {P1,P4,P5} would correspond to
PRI PAR
  P0
  PAR
    P2
    P3
  PAR
    P1
    P4
    P5   .
Since each set Si could be represented by a bit list, this could be sent
as bit arrays (packed into INTs I guess, with a restriction of bits_in_word,
in a given set).  Of course, if we ever have real BIT arrays as I suggested a
few years ago, this is another application. :-)
So above would be specified as 00000001,00001100,00110010 in 8-bit words.

No doubt there are many ways to do this, and perhaps we could define specific
protocols to encode them.

But do we need any of this?

Adrian







-- 
A E Lawrence, MA., DPhil.  	adrian.lawrence@xxxxxxxxxxxxxx
MicroProcessor Unit, 13, Banbury Road, Oxford. OX2 6NN. UK.                
Voice: (+44)-1865-273274,  Fax: (+44)-1865-273275