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

Re: Some comments relating to ICFP contest



On Wed, 26 Jul 2006, Michael Sperber wrote:

For sure, SRFI 77 doesn't have 32-bit or 64-bit integers *as types*.
(And that would cause major pain with the way Scheme's arithmetic is
generally set up.)

However, it does have the necessary operations to
easily implement 32-bit or 64-bit unsigned or two's complement
arithmetic in the form of `div+mod' and `div0+mod0' operations.  As
such, it also has bit operations on them.

Conceded. But the context of my comment was efficiency. How likely is it that implementations will compile the expression (exact-and (* x y) #xFFFFFFFF) to the single machine instruction available on many architectures?

Given that SRFI-74 already mentions ranges u32, s32, ..., operations such as exact-u32*, exact-s32-div, ..., would seem natural and could address the efficiency concern.

SRFI 77 doesn't have, but R6RS will have, support for "unboxed arrays"
of numbers in the form of a variation on SRFI 74.  I'm not sure they
qualify as "uniform," but is the uniformity crucial to your
application?

Oh, they would probably do the job.

Andre