home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Language/OS - Multiplatform Resource Library
/
LANGUAGE OS.iso
/
t3_1
/
nexttsrc.lha
/
nexttsources
/
sources
/
sys
/
big_fixnum.t
< prev
next >
Wrap
Text File
|
1988-02-05
|
3KB
|
84 lines
(herald big_fixnum
(env tsys))
;;; This file contains fixnum routines which require other number
;;; types to be present.
(define (fixnum-negative-expt x y)
(ratio 1 (fixnum-expt x (fx- 0 y))))
;;; Careful fixnum routines:
;;; Assume N bit arithmetic, machine result is N bits (what the hardware
;;; gives). The "true result" may be up to N+1 bits sign of result on
;;; overflow.
;;;
;;; if (A-B) overflows, A and B have different signs,
;;; true result is sign of A
;;; if (A+B) overflows, A and B have the same sign,
;;; true result is the same sign
;;;
;;; When the sign of the true result is positive, the magnitude of the
;;; true result is the machine result.
;;;
;;; When the sign of the true result is negative, the magnitude of the
;;; true result is the magnitude of the N+1 bit number
;;; formed by the carry bit and the machine result.
;;;
;;; Subtle issue:
;;; Taking the arithmetic complement of an N+1 bit number requires N+1 bits
;;; in all cases except for the most negative number which requires N+2 bits
;;; to complement. Fortunately we only want the magnitude, which is
;;; the arithmetic complement of the hardware result, except when the
;;; N+1 bit number is the most negative N+1 bit number (carry bit = 1,
;;; fixnum = 0), in which case the magnitude is that same N+1 bit number.
;;;
;;; Aren't you glad I figured this out?
;;;
;;; I'd be happier if you had also figured out the most negative number can
;;; only occur on an add overflow and not on a subtraction. Also, a
;;; subtraction cannot overflow and give a result of 0. Thus the CARRY?
;;; argument to FIXNUM-AFTERMATH->BIGNUM was unnecessary.
;;; This routine takes wreckage from an overflowed ADD or SUB and
;;; makes a bignum out of it. SIGN is computed as described above.
(define (fixnum-aftermath->bignum sign fixnum)
(cond ((fx= sign -1)
(cond ((fx= fixnum 0) ; min N+1 bit fixnum
(sign&magnitude->bignum -1 t 0))
(else
(sign&magnitude->bignum -1 nil (fixnum-negate fixnum)))))
(else
(sign&magnitude->bignum 1 nil fixnum))))
(define (fixnum-add-carefully x y)
(receive (sum overflow?)
(fixnum-add-with-overflow-xxx x y)
(cond (overflow?
(fixnum-aftermath->bignum (if (fx< x 0) -1 1) ; sign of either
sum))
(else sum))))
(define (fixnum-subtract-carefully x y)
(receive (diff overflow?)
(fixnum-subtract-with-overflow-xxx x y)
(cond (overflow?
(fixnum-aftermath->bignum (if (fx< x 0) -1 1) ; sign of first
diff))
(else diff))))
(define (fixnum-multiply-carefully x y)
(let ((res (fx+ (fixnum-howlong (fixnum-abs x))
(fixnum-howlong (fixnum-abs y)))))
(cond ((fx>= res *bits-per-fixnum*)
(bignum-multiply (fixnum->bignum x) (fixnum->bignum y)))
(else (fx* x y)))))
(define fixnum-divide-carefully ratio)
;(define (fixnum-negate-carefully fixnum)
; (cond ((fx= fixnum most-positive-fixnum) ;;; Huh? most-negative-fixnum is the problem.
; (fixnum->bignum fixnum))
; (else (fixnum-negate fixnum))))