From gotoda@is.s.u-tokyo.ac.jp Mon Nov 25 06:25:17 1991
Received: from peoplesparc.Berkeley.EDU by centralsparc.Berkeley.EDU (4.1/1.42)
	id AA28415; Mon, 25 Nov 91 06:25:15 PST
Received: from utsun.s.u-tokyo.ac.jp ([133.11.11.11]) by peoplesparc.Berkeley.EDU (4.1/1.42)
	id AA03435; Mon, 25 Nov 91 06:20:32 PST
Received: from rabbit.is.s.u-tokyo.ac.jp by utsun.s.u-tokyo.ac.jp (5.65/TISN-1.2MD/1.7)
	id AA15191; Mon, 25 Nov 91 23:26:20 +0900
Received: from nami.is.s.u-tokyo.ac.jp by rabbit.is.s.u-tokyo.ac.jp (5.65/TISN-1.2/R1)
	id AA16441; Mon, 25 Nov 91 23:26:16 +0900
Received: by nami (4.1/TISN-1.0N/1.8)
	id AA09273; Mon, 25 Nov 91 23:25:55 JST
Date: Mon, 25 Nov 91 23:25:55 JST
From: gotoda@is.s.u-tokyo.ac.jp
Message-Id: <9111251425.AA09273@nami>
To: fateman@peoplesparc.Berkeley.EDU
Cc: fateman@peoplesparc.Berkeley.EDU
In-Reply-To: Richard Fateman's message of Fri, 22 Nov 91 07:58:21 PST <9111221558.AA02033@peoplesparc.Berkeley.EDU>
Subject: KCL version
Status: R

Dear Professor Fateman:

Here I list the changes that I made. (On KCL, Nov 13 1990 version, SUN-OS 4.1)

1. In math.lisp, I commented this out.

;;(deftype Integer() 'integer)
;;(deftype Rational() 'rational)

(This may be unnecessary. For safety reasons, I commented them out.)

2. In eval.lisp, I commented this out.

;(SetAttribute 'Set 'HoldFirst);; we don't use this -- Set is in Lisp

3. To use the KCL specific error-set, I commented this out in eval.lisp, 

;(import '(excl::errorset))  ;; your system may differ....

and made the changes below.

;;   (setq hin (multiple-value-bind
;;          (isnoerr val)
;;          (errorset (mma::p)t)
;;          (if isnoerr val (clear-input t))))

   (setq hin (multiple-value-bind
	  (iserr val)
	  (system:error-set '(mma::p))
	  (if iserr (clear-input t) val)))

;;   (setq h (multiple-value-bind
;;        (isnoerr val)
;;        (errorset (meval hin) t)
;;        (if isnoerr val (list 'Hold hin))))

   (setq h (multiple-value-bind
	(iserr val)
	(system:error-set (list 'meval (list 'quote hin)))
	(if iserr (list 'Hold hin) val)))

4. KCL's delete-if is non-destructive. I modified poly.lisp as follows:

         ;;(t (delete-if #'coefzerofunp x :start pos))))))

	  (t (setq x (delete-if #'coefzerofunp x :start pos)))))))

	;;(delete-if #' identity temp :start (+ 2 k))

	(setq temp (delete-if #' identity temp :start (+ 2 k)))

(These changes will be unnecessary in later versions.)

5. I modified init2.lisp as follows.


(defmacro declaim (x) `(proclaim ', x))
(or (find-package 'excl)
    (make-package 'excl))

((load "mma")
(use-package :mma)
(in-package :mma)
(shadow 'user::set)
(shadow 'Exp)
(shadow 'Log)
(shadow 'Sin)
(shadow 'Cos)
(shadow 'Tan)
(shadow 'Sinh)
(shadow 'Cosh)
(import 'user::declaim)
)
(To use built-in functions exp, log, sin, ...,
 one needs to redefine them, for instance, 

  (defun sin(x)
	(if (numberp x) (conv (lisp::sin x)) (ulist 'Sin x)))


  where "conv" should convert real number (1.2345s0) to the corresponding
  internal expression in mma ((Times (Real 1 2345) s0)). Under implementation.
)

6. Finally, I modified uconsalt.lisp as follows.

;;(defmacro ulist(&rest l)(cond ((null l)nil)
;;	          (t `(hcons ,(car l) (ulist ,@(cdr l)) *uniq-table*))))

(defmacro ulist(&rest l)(cond ((null l)nil)
    	  (t `(ucons ,(car l) (ulist ,@(cdr l))))))



Speed (On Solbone, 2 sparc chips, 32 Mbyte main memory)
-----

In[11] := Timing[RatExpand[(x+y+z)^15];done]

OUT[11] = 0.7666666666666666 SECOND[DONE]
In[12] :=  Timing[RatExpand[(x+y+z)^25];done]

OUT[12] = 4.266666666666667 SECOND[DONE]
In[13] :=  Timing[RatExpand[(x+y+z)^50];done]

OUT[13] = 46.2 SECOND[DONE]
In[14] :=  Timing[RatExpand[(x+y+z)^100];done]

OUT[14] = 447.8166666666667 SECOND[DONE]


---------------------------------------------------------------
When I wrote "without hcons, mma is too slow", I didn't know
how to compile files and functions. Blame me for my innocence.
Mma is very fast. With proper implementation of 'hcons', it 
will become much faster.

Thank you very much for your continuous help.

Sincerely yours, 

Hironobu Gotoda