[Date Prev][Date Next][Thread Prev][Thread Next][Author Index][Date Index][Thread Index]
Re: Var is dead! Long Live CheckedPtrVar and IntegerVar!
- To: <michael>, <vlad!mark>, <xtech>
- Subject: Re: Var is dead! Long Live CheckedPtrVar and IntegerVar!
- From: Roger Gregory <roger>
- Date: Wed, 15 Aug 90 08:24:47 PDT
>From michael Wed Aug 15 07:13:36 1990
They have another property: They must have a fixed set of possible
values, regardless of the platform. We must decide now, for all time,
whether they are four-byte twos-complement binary numbers (or equivalent),
This happens because they carry no flavoring that would allow them to
be of arbitrary size, yet they:
- Participate in code which must have identical behavior regardless
of the platform's archetecture.
- May be members of objects whose execution spans platforms of varying
- Are encoded in formats that cross platforms on comm links or
Each of these might be worked around separately, but the first two
would require enormous effort.
Not so. IntegerVars DO have to do bignum stuff so all that behaviou shit
can be rolled into the bignum stuff. The only types of machines that I
see this as being important for are the 64 bit int machines where we will
want the bignum stuff to roll in at a different point. It is possible that
16 bit machines would have performance charistics that would make bignums
want to happen at 64K but that is of limited interest.
There is little reason that the disk and communication formats of IntegerVars
should be the same as the internal format, certainly the communication formats
can vary with the rev of the communication protocol.