develooper Front page | perl.perl6.internals.bignum | Postings from December 2001

Re: Standard Decimal Arithmetic

Thread Previous | Thread Next
From:
Uri Guttman
Date:
December 17, 2001 09:43
Subject:
Re: Standard Decimal Arithmetic
Message ID:
200112171743.MAA27891@home.sysarch.com.
>>>>> "AG" == Alex Gough <alexander.gough@st-hughs.oxford.ac.uk> writes:

  AG> Can I propose that we aim to support the Standard Decimal Arithmetic,
  AG> this will allow us to make use of an extensive test suite and claim
  AG> Enterprise-Bonus-Points at the appropriate time:

  AG> http://www2.hursley.ibm.com/decimal/

i will look at it.

  AG> Looking through the current code, I wonder why we are bothering to
  AG> have a distinction between the fractional and integer parts of
  AG> each number, to me it seems simpler to use a single string of
  AG> digits with a signed exponent as multiplying two numbers then
  AG> merely requires multiplying the digit strings and adding the
  AG> exponents directly.

there is only a distinction between those parts so we can locate the
decimal point. once two numbers are being processed (added, multiplied),
they look like one long decimal string. there are two pieces of scaling
info, the number of integer/fraction digits (their related word counts
are for efficiency) and the signed external exponent. note that we are
supporting a fixed (but growable) precision which is not well understood
by most bignum advocates. a given bignum will have a maximum number of
actual digits stored which is the precision. within that range could be
the decimal point (which is always on a word boundary for
efficiency). if the precision is shorter than the scale of the number
(in either direction) the exponent is used to make up for the extra
scaling needed.

there will be a need to take a number (assigned or calculated) and find
its first (highest) and last (lowest) non-zero digit so we can store its
current integer/fraction digits info. this will effectively be our
'normalization' operation.

i need to get back into code hacking on this. i have been off of it for
several weeks. someone kick me regularly so i get some work done. once i
get scaled addition going and some precision support (copying with
changed precision), then multiplication will follow quickly. most of the
design is done in my head (as i have done decimal math libs before) but
the coding is different with the new precision which is changeable.

uri

-- 
Uri Guttman  ------  uri@stemsystems.com  -------- http://www.stemsystems.com
-- Stem is an Open Source Network Development Toolkit and Application Suite -
----- Stem and Perl Development, Systems Architecture, Design and Coding ----
Search or Offer Perl Jobs  ----------------------------  http://jobs.perl.org

Thread Previous | Thread Next


nntp.perl.org: Perl Programming lists via nntp and http.
Comments to Ask Bjørn Hansen at ask@perl.org | Group listing | About