A COMPARISON OF QUADRATURE RULES
by
Ren-Shiaw Yang
Abstract
的分法\lIJ 之一比較
In this paper
,
we sha11 develop different numerica1 methods to eva1uate a finite integra1
and then write computer programs to compare the efficiency of these methods. By the
comparisons and ana1yses of these
resu1ts、 we
can construct a decision tree for deciding
which ru1e to use for a particu1ar integrand.
(1)
INTRODUCTION
The numerica1 eva1uation of integra1s is one of the oldest prob1ems in mathematics. In
this paper
,
we sha11 examine the Romberg ru1e
,
adaptive Simpson's ru1e
,
Gaussian
Quadra-ture ru1e
(7
-points
,
15-points)
,
Gauss-Kronrod ru1e (7-points
,
15-points
,
21-points
,
31-points) and Newton-cotes ru1e (5+2 31-points). A1so
,
we sha11 review some of the mathematica1
background of the various approximation formu1ae and their accompanying
eηor
estimates.
In each numerica1 method we are given the integrand f(x)
,
the interva1 of integration
[a ,b] 徊, b
finite)
,
abso1ute error Ea
,
re1ative rrror Er
,
and we want to find an
approxima-tion S which hopefu
l1
y satisfied the fo11owing condition:
b
IJ f(x)dx -
51
a
b
手 max(Ea ,
ErlJ f(x)dxl)
a
The numerica1 methods considered in this paper can be used to approximate integra1s
whose integrands have no antiderivative in the a1gebra or e1ementary functions. A1so
,
these
methods can be used when the integrand has discontinuous or singu1ar points in the interva1
[a
,
b].
師大學報
第卅一wl
(11)
STATEMENTS
OF
THE QUADRATURE RULE
2.1
Romberg Integration Method
Romberg integration is a method that has wide application.
It
uses the trapezoidal
rule to give preliminary approximations and then applies the Richardson extrapolation
process to improve these approximations. The algorithm is as follows:
b
Suppose we want to evaluate an integral
,
J
f (x) dx
. (a
,
b are finite). The Tra
a
pezoidal rule says that
、
••
,
HUF Fι 司4-n
N7
.、 JX
F-h
t.-可-4-vb=
m-3
勻,-+
hu
s-‘
+
a
z--h-2
冒
X
AU
X
F-hu
rJa
where
a
<口〈 b , hzhp
and x.
==
a
+抖,
for each
j=O
,
1
,
2
,...,
m
3
Letting
hkz?丸,
b-a
consider the trapezoidal rules
k-1
b
IE
,
2 - 1 . b - a 2"
J
f(x)dx
=于 [f(a) + f(b) +
2
E
f(a+ih
k
)]- 了 hltf{υk)
,
i=1
a
where
a
<口 k
<
b
,
k =
1
,
2
,
3
,...,
Define
h.
z 」 [f(a}+f{b)1=2[f(a)+f{b}}
1
,
1
2
then
RJ屯, 1
bh
',
hu
t4
.l
+
a
',.、 ZL 司4-1ι
bhfuz
2.l
-gA
LA
hu
+
',辜,
L
RK
l-2
=
for each
k
=
2
,
3
,... ,
n
-
642 一
積分法則之一比較
Note how the error in the extended trapezo idal rule
1
)
-K
E
--..
.
l
+
a
{
z ﹒晶,
L
'斗, L-vb--ki
勻,-b
h.
J
f(x)dx
=去 [f(a)
+
f(b)
+
2
a
(b-a}h4
k
,,
(4)
一守主百
f
\
~
I(川)
+
[f'
(b) -
f'
(a)]
2hkE
υk
<
b
,
<
a
f or each k
=
1
,
2
, . .
.‘
and some
Rk - 1
,
1
b
.r
f(x)dx
~
a
Implíes that
)
A 崎 -bkh1
I
{
nυApplying Richardson extrapolation procedure
,
with the error in the above approximation
4 丸, 1
-
~-1 , 1
3
司 4,
-k
nuu
Define
--we obtain the approximations:
、.
-i
E
可﹒---
-l-R-1
-一
.••
2
--24 .14 司 •••• 司 •• 4.l-4.
R
'.
-.、 JAq
=
R
i
,
j
2
,
3
,... ,
i
0(h:j)
The approximations are often presented in a table of the following form:
=
2.,
3
,...,
n
--i
for each
with error
R1
,
1
R2
,
2
R2
,
1
R4
,
4
R3
,
3
R4
,
3
R3
,
2
R4
,
2
'-A
'
,
JR
R4
,
1
R
n
,
n
-
643 一
R
n
,
3
R
n
,
2
Rn
,
l
第卅一期
師大學報
2.2 Adaptive Simpson's rule
[事, b)
G
υ
for some
From the 3-point Simpson's
ru 峙, we
have
}
HV as-、}
aq
'
..
、 ZL5-o
hu-qd
}
hu
'
a
{
nhu--X
AU
}
x
(
ZLhu
a
pld
=
I
h
[f(a) + 4f(a+h) + f(b) )
h-3
=
S(a
,
b)
where
f(a+h)]
f
(b) ]
+
+
3h
4f(a+ 言.!.)
...
,
-n
需
4+
4f(a
+
+
(f(a+h)
‘ ..
,
a
(
6.-[
hu-ro
-u-6
--b)
a+b
S( 一γ ,
S(a
,
Let
and
term in the error is assumed to be constant
,
then for arbitrary
toler-)
HH(
)
A 唔 , z﹒、 SEAIf the
a+b
S
(一2'
a-恥 b 、
2 '
ance
é
the inequality
15ε
<
b)
I
S(a
,
IS(a
,
b)
(#)
a+b
S("'2~'
a+b 、
2
implies that
ε
(See reference [7]).
If
(#)
does not hold
,
apply the estimation procedure individually to the subintervals
<
b)
I
S(a
,
a+b
,
2 '
bJ
、 with a 忱mceof3
. If this tolerance is not obtained on
{a+b
-2-'
and
[a
,
each of the intervals. split each into two and repeat the above estimation procedure.
The interval splitting process can be represented by a binary tree of the form:
積分注目IJ之一比較
/
)'
'\
/\-8-'-4-/ \ b
z \ / \
「吃 γ 竿 b
] [
2.3 Gaussian
Qu
adrature rule (for Legendre)
Consider a quadrature of the form
1
n
J_f(x)dx
~
.E.wif(ai )
.
ι1
i=l
We say the above quadrature ru1e has degree of precision k if the ru1e is exact for a11
po1ynomia1s of degree 1ess than or equal to
k and is not exact for a11 po1ynomials of
degree k
+
1. Since we can choose the n nodes
,
a
1
,
a
2
' • • •
,
a
n
,
and n weights
,
W
, ,
w....
, . . . ,
w..
,
we can hope to obtain 2n-l as the degree Of precision.
1'"2
Let the nodes a" a....
l ' "'2 ' • • •
, . . . ,
,
a_
Qn
be chosen not equally spaced
,
but as the roots of
the Legendre po1ynomial L..(x) oforder n . Theweights W"w....
n''-'I -... _...-...
.
..&..-
'1'0...0.&...
"1 '"2'. . .
,...,
'"n
w_ arethen
determined by the system
0
1'
equations:
rw大學線
第卅一}的
W
1
+
W
2
+
W
3
+ •
.
• +
w
n
=
2
a1w 1
+ a
2
w
2
+ a
3
w
3
+ •
.
• + a
n n
W
=
0
2
2
2
2
2
a1w 1
+ a
2
w
2
+ ajw
3
+ •
.
• +
a~wn
n n
=
3"
.
.
+
可 JW
--n3
a
斗,
呵,w
唔,-n2
a
+
'L
W
',-n'L
a
ququ
.,占﹒'.-nnn
ed
EVE-d
-le--。
02-n
fjtlL
--n
w
'i
nn
a
4.
With the weights and nodes chosen as above
‘
the desired degree of precision
,
2n-l
,
will be obtained. For more details
,
see references [11 and [51.
2
.4
Gauss-Kronrod rule
Let L_
(x)
be the Legendre polynomial of degree n and define K
.,_ •
,
(x)
by the
2n+1
equation
K '"'_ •
2n+1'~1
,
(x)
=
-
L__ (
~n'~/~n+1
x)
P _ •
1
(x)
with pn+1(X)
,
a polynomial of degree n+1
,
chosen
SO
that k2rI+1(X}is orthogonal on
K
V'n
T _ . ___.1____._
. L . _ _ _ _ . _ _L"K..._ •
, (x)
[-1
,
11 to all powers
x"
,
K
=
O
,
l
,...,
n . Let us denote the roots of"2n+1 '^'by
‘毛,
1
,
x..
x
, . . . ,
X
.,_ •
1
'
and the weights of the corresponding quadrature rule by
2
,.. . ,
x
2n + 1
,
w 1
J
W
2'
, . .
,
w 2 n+ 1.
It
can then be shown (reference [11) that the quadrature rule
2n+1
1:
w.
i .- \ ""i
f ( x
, )
i=1
has degree of precision 3n+ 1 for even n
,
and 3n+2 for odd n. Note that since the n
K
(f)
=
nodes of the n point Gauss-Legendre rule are a subset of the nodes for this rule
,
we may
(without increasing the number of function evaluations) also compute the n-point Gaussian
approximation to the integral of f(x). The accuracy of the Gauss-Kronrod rule
approxi-mation can then be determined by a comparison with the Gauss-Legendre rule.
情分法則之一比較
2.5 The adaptive Newton
-c
otes Quadrature rule
Suppose that [a.b
1
is divided as follows:
a
h-2
+
nu
x
h2.4
4.
nu
X\
xO+h
Xo
+τ-XO+1-b {h=-z一)
3h
__
• 7h
._
,,_
b-a
11
11
11
"
"
11
11
X
o
x
l
x
2
x
3
x
4
x
5
x
6
The 5 point Newton-cotes rule on
徊, b
1
is given by
b
S
f(x)dx~A~[7('f(xn) + f(x
45'.'-'''0'
-'''6''
t:))
+ 32(f(x
--'-'--2.
.,)
+ f(x
-'--4
A ))
a
with er
r.o
r
+ 12f(x
3
)]
~
7
~(6)
h
f'VI (;)
15120
a
<已<
b
e
=
(6)
Using f ( XO) through f
(文 6)
to approximate f \
V I(f;),
and then subtracting the
cor-responding approximation of the error term
,
we obtain the 5+2 point rule approximation
b
lf{X} 心晶宮[圳的0) + f(x
6
)) + 2048(f(x
1
) + f(x
5
+ 2016 (f(x
2
) + f(x
4
--}) + 4004f(x
3
)] •
(111) STATEMENTS OF THE TEST RESULTS
In this paper
,
we use nine methods which are listed as follows:
Method
(1):
Romberg Method
Method (2): Adaptive Simpson's Method
Method (3): Gauss-Legendre Method
(7
points)
向1; 大咚!悟
;n月f 一朋
Method (4): Gauss-Kronrod Method (7 points)
Method (5): Gauss-Legendre Method
(1
5 points)
Method (6): Gauss-Kronrod Method
(1
5 points)
Method (7): Gauss-Kronrod Method (21 poings)
Method (8): Gauss-Kronrod Method (31 points)
Method (9): Newton-Cotes Method (5+2 points)
Numerical experiments have been conducted on Kahaner's test problems to investigate
the performance of various numerical methods. Eåch
0
1'
the nine methods above were
tested on eighteen different
1'
unctions over three di
1'
ferent intervals and three different
tolerances. The results of these tests are listed in the following form:
Integral
1'
orm a to b
EPSABS=.
Integrand F(x) = . . .
Method
(l):
Integral Approximation = . .
Number
0
1'
Function Evaluation = . . .
Tolerance
0
1'
the integral = . . .
Error code = . . .
Method (2):
Integral Approximation = . . .
Number of Function Evaluation = . . .
Tolerance of the integral = . . .
Error code = . . .
Method (3):
a:
lower limit
b:
upper limit
EPSABS: absolute error
Tolerance
0
1'
the integral: max
(E
PSABS
,
relative error)
Error code:
(l)
0: normal return
(2) 1: over 100 subdividions
0
1'
[a
,
b]
(3) 2: round
-ó
ff error flag.
( 4) 6: 1 nput to lerance erro
r.
的分法則之一比較
The sample results of these tests are listed in Appendix 1
,
and from these results
‘
we
pick the number of function evaluation to obtain the comparison tables listed in Appendix
2.
Analysis of the test result reveals the following:
(1)
The Gauss-Kronrod methods (7 points and 15 points) usually require the least number
of function evaluations. In 5 cases (out of a total of 36 cases) the 7-point
Gauss-Kronrod Method used the least number of function evaluations. In 9 cases
‘
the
15-point Gauss-Kronrod Method used the least number of functionevaluations.
(2) The adaptive Simpson's rule yields the least results for the integral f(x)
=
SQR
T(
x)
叫e叫吋吋叫
(ββ
圳
糾
3-
-1 , 3-2
圳
3-
只刊
-2
引
2)), f(你仲
x
f叫(忱川
x叫}
=
J
↓一
(Tabl
l
x~+l
e^回 1
(Table
(1
4-1
,
14-2))
,
f(x)
=
alog
(x)
(Table
15寸,
15-2)). Each of these
inte-grands is in some way ill-behaved near the origin
,
and well-behaved away.
It
is therefore
not surprising that an adaptive quadrature rule
,
which will naturally execute more
function evaluations near a singular point
,
yields the best results.
(3) ln theory
,
one should expect good results from the 21 and 31 point Gauss-Kronrod
rules
,
as these rules have a very high degree of precision. But as indicated by our test
results
,
a very large number of function evaluations may be needed to overcome
round
-o
ff error. However
,
this may not be as serious of a problem if double precision
arithmetic is used.
(4) The Gauss-
Kr
onrod rule always used less function evaluations than the corresponding
Gauss-Legendre rule. This is due to the previously mentioned fact that once the 2n+ 1
point Gauss-Kronrod rule has been computed
,
no more function evaluations are
needed to compute the n point Gauss-Legendre rule.
(5) The Romberg Method yields the best result for the uniformly well-behaved functions;
l
f(x)
=
EXP(x)
,
(T
able (201))
,
f (x)
=-1平Ëxp 忍了,
(Table
(1
0-1))
,
f(X)
=
Sin(x)
,
(Table 02-2))
,
and f(x)
=
Cos
(l
00 Sin(x))
,
(Table
(1
7-2)).
(6) In general
,
the Newton-cotes
quadra直u're
rule (5+2 points) does not yield the least
number of function evaluations.
(7)
If the integrand is undefined at an endpoint
,
we can't use the Romberg or adaptiv
,e
Simpson's method.
n,Ii 大 f~1f有
m Jtt一!的
(IV) CONCLUSION
Suppose that we want to evaluate the integral
b
J
f
(x) dx
,
where a and b are finite
a
By the previous discussion
,
we can construct a decision t
l"
ee as follows:
To evaluate
b
J
f (x)dx
a
Do you care about computer
time and are you willing
to do some analysis of this
problem?
no
Gauss-Kronrod (21 points)
or
Gauss-Kronrod
(31
points)
。、 d ρ-wvd
li--jIll-ll.,
Is the integrand uniformly
well-behaved on
峙, b)
一一一-IRomberg
ves
method
E
no
日
Is the integrand
ill-behaved in
some place and
well-behaved
away?
o
n
il--i 』Iz--••
Are the
discontin-uities or
singularities of
the integrand
within the
interval
,
and do
we know where
theyare?
o
n
--till--li--Has the integrand
end point
singularities?
0-
1
no
Gauss-Kronrod rule
(21 or 31 points).
If this routine
returns an error
flag ,、then
we use
Gauss-Kronrod rule
(7
or 15 points).
yes
..
yes
一一-yes
..
一-一一一-Adaptive Simpson's rule
Split Jbf into
a
c l C 2 n c 3 n b
J _
~f
+ L.-f + L_- f +...+ J
1;;
1
-';2
c n
where ci are the discontinuous or singular
points. Then on each subinterval use
Gauss-
Kr
onrod rule (7 or 15 points).
Does the integral transform
to an integral with no
sìngularities?
的分注目IJ 之一比較
Y郎,因
no
Gauss-Kronrod rule
(7
or 15
points).
-
651 一
師大學報
第卅一期
(V) PROGRAM LISTING
In this part
,
we summarize the subroutines as follows: .
Method
Subroutine
Number
Quadrature rule
Name
(1)
Romberg Method
ROMB
(2)
Adaptive Simpson's rule
ASIM
(3)
Gauss-Legendre rule
(7
points)
GS7
(4)
Gauss-Krontod rule (7 points)
GK7
(5)
Gauss-Legendre rule
(1
5 points)
GSl5
(6)
Gauss-
Kr
onrod rule
(1
5 points)
GK15
(7)
Gauss-Kronrod rule (21 points)
GK21
(8)
Gauss-Kronrod rule (31 points)
GK31
(9)
Adaptive Newton-cotes rule (5+2 points)
ANC7
We use the computer IBM 370 (Music) and coded in FORTRAN language.
的分法HlJ 之一比較
DATE • THU JUN 06
,
1985
t1
AIN
"""甜甜聶曲曲曲曲曲曲*曲曲曲曲""*"""壘*"*"甜甜食費,壘*"""曲曲曲食費*轟轟 """1、"飛禽食費 4、"*', "1‘
食費甜 食 *1、
最*晶
A
CDt1PA成 SIO"
OF
QUADRATURE
RULES
會£舞
曲食費
YANG
,REN-SHIAW
甜食告 ,、食由 于世會,可 貴""""負責甜甜"甜甜食*食費""""#1 11 繭 "111悟脅,甜甜甜甜甜 ft***""ftft 食費 """ftftftl悟*食,, "1悟,時 4、會會 Ir itRELEASE 2.0
食費""""負責曲曲曲賣會 *"""ftft""""""*""""""""*"""食費甜甜甜食*食費會負責舜,、由"*',由甜食,、 j,
"*"‘快,、
會曲曲 t1A 1 N
P R 0 G R A
t1 橫飛*"*"
"甜食
"""11""""""" 晶晶甜食晶晶#1"甜甜食費食費最 "ft"" 甜甜食青青侖,悟甜食,‘"*"晶晶晶負 *"*"1、 4、機"*食甜食食壘.
)
nv
nu
1
(
Hn
)
nu'
,
nu' 。 .,-'、, E',
nu
'、w-11
.,』 F." ,旬,,',
',
E 、 yo nu-nV-E-'
,
10
?'nu
p--,
1
•
•• 』 nu 勵自﹒' •.•
,
.
••
、,onu-nvnv
',‘',,
J
'tnu--h
',而 U'EP3.'"n
.,',、, 2 .LT'"VE a 肉,、',且',.,.
' ••• p3 、 ''"HR 自nv
••
"
nu
‘
FPa
',
nunr
'tnu'E
',','
, 3' ,、?',,
-L.
,
tom
-nn'
,.,
-L'
,
“呵,nURE--‘
.,
.•
R 's-RMH MHaH',
haan'
,‘
MHT'? •.
,
-nnvan
nwmu't
an
pa.p-MFLPLFEMP--' ,句 FLPLPLPUP-旬, LPLFORTRAN
IV 區 1
0001
0002
0003
PRINT THE HEADING
.
• A 唔,'.
• h 『 帥 "u'
..
.
•
nu
',
•
.
•
“呵.
"。',
-MHnu
"n
',
-L -R.
RR‘''
,
u'
',』‘.'?',
VA HHV <,1
., F 呵, '"r',
VA',
Eee'
,
IH
E--‘
J
."va 呵呵,‘ van '、"、•.
'
..
‘
"r ﹒'"苟,', FE',
E.,
r-mu
nn'
,'""
',
fm."
、',."',、 dnn 、 J' 、 1." 。 "uzd
••••••••••
肉, t',
..
F3?' ,。, 1 , b' ,、 nrM 間 , E 、','‘、', PE ﹒' , Ea 肉 Ev--" ,' ?'"H?'"H 自 '"H' , nnnnnu"nnu
wFWF
."‘'"
FLF-wFU3
OOO~
0005
0006
0007
CAll RO/\B SUBROUTINE AND PRINT OUT THE RESUlT
CUl RO/\B (F
, ^ ,B
,EP5A8S
,EPSR
E l ,II
t11 T
,RESUl T
,NEVAl
,1 ER)
CAllOUTPUT(RESUlT
,NEVAl
,EPSABS
,EPSREL
,IER)
PIMP-upzu0008
0009
CAll ASI
t1SU8ROUTINE AND PRINT OUT THE RESUlT
CAll AS 1 " (F
,^,閉, EÞSA8S , EPSREl , ll 州 IT , RESULT , NEVAl , IER)
曲曲曲曲,由PR
IIIT
THE RESUL T
甜甜食會曲曲CAllOUTPUl (RESUlT.NEVAl.EPSA8S
,EPSR
E l ,IER)
P-up--PL。010
E
。 011
CAll GS
,
GK AND ANC7 SUBROUTINE AND PRINT OUT THE RESUlT
D08KEY"I
,
7
CAll AGQ(F
,A
,8
,EPSA8S
,EPSREl
,KEY
,LI
t1IT
,RESUlT
,NEV^l
,IER)
由晶晶曲曲"
PRINT THE RESULT
自食晶晶晶,CALL OUTPUT(RESUlT
,NEVAL
,EPSA8S
,EPSREl
,IER)
CONT1NUE
WR
1
TE
(6
,
1
1)
WRITE(6
,
11)
F OR
/l
AT (' 1 ' )
S10P
END
FiwFLFtu8
11
E
。012
0013
。01~
0015
0016
0017
0018
0019
0020
NOTERt1, ID.E8CDIC , SOURCE , HOlIST , NOOECK , lOAD ,"。門 AP , NOTEST
NA/\E •
t1AIN
• llHECNT •
56
--
653 一
對 OPTIONS
IN
EFFECT 負
第卅一月1
師大學報
OATE • THU JUN 06
,
1385
甜食 ftlti悟,甜甜疊疊,食費曲曲曲聶曲曲曲,晶晶脅,轟轟轟 ftft 費甜甜,壘,食繭 ft 甜甜 ftftftft 聶食費禽,、食*費長青
甜甜 食費
甜食
SUBROUTINE ROMBERG HETHIO
食費
ft 費 由疊 疊 ft 轟轟轟 ftftftftftffftftft 甜甜食費由 ftft 轟轟曲甜甜 ftft 曲曲曲曲曲 ftftftftft 由晶晶晶食甜甜費 4、 ft"ftftftft"ft
MAIN
RELEASE 2.0
PIMP-旬, EMPIMP-up--PLFORTRAN
I'
V G 1
5UBROUT HIE
RO門自 (F , A , B , EPSABS , EP5REL , lIMIT , RE5UlT , HEVAl , IER)
。 IMENSION
R(2
,
100)
IER"。
IROFF-O
NEVAl-O
RESUlT-O.O
IF (EPSABS.lT.O.O .ANO. EPSRE
L.
LT.O.O)
IER胃6
IF(IER.EQ.6) GOTO 999
0001
{)002
0003
OOO~
0005
0006
0007
0008
"開 B-A
R (1,1)團"費 (F
(A)
+F
(B)) /2.0
00 515
1 個 2 ,lI MIT
NEVAl-NEVAL+l
sυ門﹒0.0
11-2 禽 ft
(1-2)
DO 525 K..I
,
ll
FX圖 F (A+ (K-0.5) 曲 H)
!:U門-SUK+ FX
R
(2, 1)
-0.5 ft (R (1
,
I)+Hft 5UM)
DQ"535 J.2
,
1
R t2
,
J) ﹒(圳青 (J-I)
ftR (2
,
J-
1)
-R (1
,
J-l)) /
(~負責 (J-
1)
-1)
RESUlT-R(2
,
1)
E
525
REO-' 唔,‘、 3. 句 PD' 旬,', noqdnu' ,0111111111122
nunununvnununununununununu
nunununununununununununvnu
535
TEST FOR ROUNDOFF ERROR ANO 5ET ERROR FlAG
.G
E.
ABS(R(2 , 1)-R(I , I-I)))IRO fF "IROFF~I
CI • L
T.
2) GOTO 520
(ABS (R (2
,
1) -R
(1,
1
-1))
(IROFF .G
E.
6) IER-2
'
•.••
,
•••
111
PEMF-",
.•
0022
0023
002~
n3
qJ
倆 wd)
‘
ynu
',?'
-LHU
HUFU
,且‘ F',
L-aHUHHMH
',、"目,
-nH
RE 個院 .",巴,"
.
• L',
',』..
』"n.
"、斗,‘.', "r 、', ',』•..
..
•
、 J •••• "。.
.
".'
,、 d' ,‘."rmm
',
L •••{}}
.,.,.,
-LVan--•
• m 鬥呵, •••••‘'‘
1(-{
f
.
l a 叭 "n••
‘"",
E
",.、 nu"Hu nυ ,3.PHJ' ‘', MH HHREq's-u 『 -L ﹒' "。.", rphJ' ,' "間, 2 、 Hn ﹒'"" "間, F 間 nu' ‘ .nupE
.
'"nnu"nplw
E
520
0025
0026
0027
0028
0029
0030
}
'
..
}
1
(
IER-I
tlEVAL-2 甜甜(II EVAL)
+ 1
WR ITE (6
,'
5
1)
FORMAT(/
,
15X
,
'METHOD
RETURN
END
5
"
5
515
E
999
51
0031
0032
0033
003"
0035
0036
ftOPTIOHS IN
EFFECT 曲
NOTERM ,
1 D
,
EBCD 1 C
,
50URCE
,
NOL 1 5T
,
NODE CK
,
LOAD
,
,
m"AP. HOTE5T
費 OPTION5
IN
EFFECT 晶
NAME
- ROIIB
,
LI
NE CIIT ..
56
費 STATISTICSft
SOURCE STATEIIEtlTS •
36
,
PROGRAH SIZE •
0008F 。
由 5T
ATI STI
C5 曲
NO
DIAGH05TICS GENERATED
積分 i去!IIJ 之一比較
1985
DATE -
THU JU" 06
,
Ir lrlrftlri悟 ft fr. ftftftfcftft 晶晶晶食費 Ir lr fr. lrflft 曲曲 Ir lrflfl 由 f.lrftflM.flftft 負責 lritftft 食,、 ft t:食費 It f, ft:':
It i、 台才有
Ir
lr
SUBROUTI"E ADAPTIVE SIMPSON RULE
吭吭
甜食 'Id、
最甜食 ftft* fr. i悟fr. ft fr.fr.fr. ft fr. ftft fr.fr. ftft 會食費 4、 4時,堅青青fr. ft 禽 ft fr.j、 fr.fr. i, fti、 {.flM、 4、fr.fr.會突如飛快 d堅決 ;d、,',
SUBROUTINE ASIH(F
,
A,
B.EPSABS.EPSREL
,
LIMIT
,
RE5ULT
,
"EVAl
,
IER)
DIHE"510N
FAlIST(IO叫.
FBlI ST
(1
00)
,
FCl I 5T
(1
00) • TOL (100) •
費
A lI ST(IOO)
.Rl I ST(IOO)
,
H(IOO)
IER興。
NEVH~。
RE5ULT"'0.0
IROFF~。
IF (EP5^B5.LT.0.0 .^ND. EP5RE
l .
L
T.
0.0)
IER~6
IF (IER.EQ.6) GOTO 999
1-1
TOL
(1)-1 。由^,,^ X
1 (EPSM5. EP5RELlrABS (RE5UL
T))
H(I)..(B-A) 晶。 .5
ALI ST (I)..A
FAlI ST
(1)
-F
(川
rcllST (I)-F
(A+刊 (1
))
FBlI ST (I)-r (B)
RlI ST
(1)
-H
(1)費 (FAlIST(I)+4 叫 ClIST(I)+FBlIST(I))/3.0
t1AIN
RHEASE 2.0
FEMF-MPLFLPL
FOnTR^N IV GI
0001
0002
0003
0004
0005
0006
0007
0008
0009
0010
0011
0012
0013
0014
0015
0016
'‘'‘fr.j、食
nvnv
..
,旬, b,
r
,
r
}}
))
11
{{
?','
,、",、呵,-L-L
PLHM
‘,',',FAR
O-?+ 膏 、',-hUPE 費‘.'‘Jhvr'?an
.,','‘ m 聶壘 , 1 、 1.q ﹒肉qdH1++
肉 "JAR',、、E'、., qd'pun--a 悶 T﹒魚, 1 , 1' , 1nunu司,TTAHHT'++RVP3"u
.,nu﹒'‘.'、',.,., ELFU4,.',,EL--h't、',、',、',••
L'‘、',‘aRFLHV、',••••
,.『、自' ,'、',."?'?',r,r-n﹒',‘、',、',、', -nuMVP3,耳,‘、,tp3t、?',fT'、 1 ,t u叭-PE.,.'‘“‘",',2p3,3.'?1 nu、“H-L-LE'、',=',.,.,、,',tpanUPEH.R.H.,.,.',L-LgL.,-L.,
nu--L'‘‘',、',、',、-n-La",LRU'‘‘nu-L﹒' ‘ 4. ,.叭,F',HHHnaH﹒叭,r,VPFHn',RH ﹒ 'twv"",胃。"四"間,闕,, nu'',EHU'E.,司 4. “.,啥,‘'、 Je ﹒旬,hd'。可', nu.,M",FPF 電 dpaauvMVHVMVMVMVMV',t1AIN DO-lOOP
u 青 ft fr.c
E
c
100
C
。 017
0018
0019
0020
0021
0022
0023
0024
0025
0026
00
2]
0028
0029
0030
0031
TE5T nOUNDOFF ERROR AND SET ERROR FlAG
IF(I.L
E.I)
GOTO 110
1 F (^BS (5 I+S2-V71 .GE. ABS (S
I+
S2-Rll 5T (1-
1)))
IF (InOFF .G
E.
6)IER-2
PLFLFL
。 032
0033
0034
IRorr..IROFF+1
1 F (^BS (5 1+52-V71 • L T • V6) GOTO 150
1-1+1
A
l I ST(I)..V
I+
V5
FAL 1 ST (1) "v3
FCLl 5T(I)-FE
FBlI 5T
(1)
..v4
-
655 一
E
110
。 035
0036
0037
0038
0039
00 1
,
0
師大學報
第卅一期
FORTRAN IV.Gl RElEASE 2.0
ASIH
DATE • THU JUN 06
,
1985
。0"1
00"2
00"3
H(I)-V5/2.0
TOl
(1)
-v6/2
.0
R
lI
ST
(I
)-S2
E
)
,
}
唔,‘{
‘ ..
,
唔,‘"u '、 enu+"
.'?'
,、凶, t‘',,
1
‘3MH
1++'
﹒?'嚕,‘.,amu-‘J.'
.•
L-Rwan
--HV,rHVE',
1.
,HU--‘‘FPE
M--圓圈,',-L'、d'、d."﹒',' ,‘',‘',‘F.O-EHVP虫,‘
y
.',',,.','‘',nunnnu'E.','
,', E 、,tf., 1••
'。'。“",旬,t 'E 、T'?','""、.',.‘﹒',,.,.,.、',“" .',',、dpap3••
',',,.,-L'E-RRR
&.,3.,.,
••
‘'',
E‘panUHURU-R
••
",'"HHU
-1.
,-L-L-L.,-L.'?',3?'"RHV.'"n?'HU
E-L-RPL"白,tnu-Lnu'Enu'E'E"nnuv'EMH .,.",r,『'『HnT'"nnu 侃"刊 "u心,MHHW',RH'Enununy
EJoqJ1
.,司,‘n3PD t 句EJ'。',89012缸,iqEJ6',由υqJO
'",.“
•.•
h 『 -h呵,-h 『 -uvphdphJphJphJphJP、 4 ,PHJP、dp、JP、d'hunvnununununununununununununvnununu
nununvnunununvnunvnunvnununununvnu
機 OPTIO"S
Ih ErrECTA
"OTER~ , ID , EBCDIC.SOURCE , NOlIST.NODECK.lOAD.NO"AP.NOTEST
~OPTIO"S
IN EFFECTA NA"E - ASI"
• LINECNT -
56
食 SHTI S Tl CS 晶
SOURCE STATE~ENTS
-
6o.PROGRA~
SIZE
.
00112"
~STATISTICSA
NO DIAGNOSTICS"GENERATED
f1'i 分法具IJ 之一比較
DATE - THU' JUH 06
,
1985
、 Jnu
RO
E-
--.',
-LRd
aH.
,
VL
amaπARAR袋,ER **ARA"恥",‘n&n.‘',
aH品"?'nu 食品"-Lnu*LHHU',
真 H&npb',、-RAn'tRU
**"nnnH
*禽,nuan&H',.'
‘RAn-'.
‘na“MH‘,',b
*M間*.,nu 曹 *﹒肉食-Lnunk‘“.'‘"
••
',且 ‘n'、4‘nv'',、.'.",34n't',‘',
-nHuanu叭,、dhu*-En*
•
•
•
•
*"u‘"-L-LhU
ARA",巴,E﹒ 禽,且袋"n﹒',‘mMV‘",3‘.,.L
-n﹒'。"nrnu--n',食,巴nu-L‘nhr‘“
••
'
,
E
‘的﹒肉食,、4,‘‘HR &""u-n 個 MUT',、d &En--n‘"﹒肉,、eRV *禽,3.',E ‘四",E*"r-L -RM間.",Emn﹒AR.'an--munE
&RT'‘""。、',MHRWJ'*HU‘m.o."nE
*nu‘"-mnu.
*nn*
••
'。 ‘四 B&nFS 、 OT -nHU。",‘、',-mu ‘npaa 伯伯 u、'、dnuFU**nu.,.
-n‘H.R-LRURV',、', 甜食 -m••••
‘,
b
*E 一 ooaBOO- **M 開 MH•••
",-nu' 這 nq ‘“‘",,nununu、',‘',‘',‘'',馳",E ‘阿*',.,nu--',,.,., -1 ‘',-an﹒ ‘"-nHU'、dRRUTEhn',、a'、,.、,巴、﹒',3"間 由 ARnvN0.L'.LnnT','?'?',1,『PZ .“‘nn刊,EE--",'"u'EP且,、dp3,、dnu'E., 也們 AH"。"HRHH"vp=pbp3.,.,.,.'"啊,.、'‘、 *&n*."*HU﹒',EPEa-",E" 也 -L-L-L-LRU',-PV .R.“‘"‘"‘",、dhu-'MH-LRR.R.-nnnm叭,E.,.,.,MAIN
R
El
EASE 2.0
piwpbpzupup--FORTRAN IV Gl
0001
0002
000)
OOO~
0005
0006
0007
0008
000'l
0010
0011
0012
001)
001"
flRST APPROXIMATION TO THE INTEGRAL
CALL
GS7(F ,_, B , RESULT , A 自 SERR , OEFABS , RESABS)
CALl
GK7(F.A , B , RESUlT , ABSERR , OEfA 肘, RESABS)
CAll GSI5(F.A
,
B
,
RESUlT
,
ABSEnR
,
OEFABS.RESABS)
CAll GKI5(F
,
A
,
B
,
RESUlT
,
ABSERR
,
OEfABS
,
RESABS)
CAll GK21 (f
,
A
,
B
,
RESULT
,
ABSERR
,
OEFABS
,
RESABS)
CAll GK)1 ".A
,
B
,
RESUlT
,
ABSERR
,
OEFABS
,
RESABS)
CAll ANC7(F
,
A.B
,
RE5UlT
,
ABSERR
,
OEFABS
,
RE5ABS)
kEYF-l
KEYF..)
kEYF"KEY
1F
(kEY .l
E.
O)
1F
(KEY .GT. 7)
C-kEYF
NEVAl"O
1F
(kEYF .EQ.
1)
1F
(KEYF .EQ.2)
IF (KEYF .EQ.))
IF (KEYF
.EQ.~)
IF (KEY
Fo
EQ.5)
IF (KEYF .EQ.6)
IF (kEYF .EQ.
71
lAST"1
RLI 5T
(1)
-RE5Ul T
El
1 5T (1) -AB5ERR
1 ORD
(1
)-1
FEMF-切,、0015
0016
0017
0018
0019
0020
0021
0022
002)
002~
0025
0026
0027
0028
0029
00)0
由壘,ERRBIIO-AMAXl
(EP5ABS , EP5REl 貴 AB5
(RE5
I1
l T))
1 F
(ll 川 T.EQ.lllER叫
IF (IER.N
E.
O .OR. (ABSERR.l
E.
ERRBIIO .AIIO. ABSERR.II
E.
RESABS)
A AB5ERR.EQ.0.0)GOTO 60
TE5T ON ACCURACY
11 壘,PUP--FEW
00)1
00)2
00))
.on
1111 壘INI Tl A
Ll
IAT10N
-
657 一
11""
ERR
/l
AX"ABSERR
MAXERR"'l
AREA"RESUlT
ERR5U
/I
"AB5ERR
NIUlAX.l
IROFF 1"0
FIWPLFL
。O)~
0035
00)6
0037
00)8
00)9
師大學報第卅一期
FORTR^N IV Gl
RElEA5E 2.0
AGQ
OATE _ THU JUN 0&. 1985
OO~O
IROFF2-0
E
c
1<費由們AIN
00 lOOP
最曲曲E
00~1
。o
10 lA5T-Z
,
lIHIT
c
E
費曲曲
BI5EtT THE 5UBINTERVAl WITH THE lARGE5T ERROR ESTIHATl
E
00'.2
Al 個All
5T
(tI
^XE
Il
R)
00~3
þl"O.5 1< (A 1I 5T(們AXERR)
+B
lI
5T
(們AXERR)
)
OO~I.
A2-Bl
。M5
自 2-B lI
5T
(I\
^XERR)
001.6
IF(KEYF.EQ.
l)
tAll G5
7(
F
,
Al
,
Bl
,
AREA1
,
ERROlll.RESABS.OHABll
00~7
IF(KEYF.EQ.2)tAll
GK7(F , Al ,因 I , AREA1 , ERnORI , RESABS , OEFAB1)
oM8
I
F(
KEYF .EQ.3
)t
All G5 15 (F
,
A
1 ,因 I , AREA1 , ERROR1 , RESABS.DEFAB1)
001'9
IF(KEYF.EQ.~)tAll
GKI5(F
,
Al
,
Bl
,
AREA1
,
ERnOR1
,
RESABS.DEFAB1)
。050
IF(KEYF.EQ.5)tAll GK21(F
,
Al
,
Bl
,
AREA1
,
ERRORI
,
RESABS.DHflBl)
0051
IF (KEYF.EQ.6)tAll GK31 (F
,
Al
,
Bl
,
AREA1
,
ERROR1
,
RESABS.DEFAB
o'
0052
IF (K
E-Y
F .EQ.
J)
tAll AN
C]
(F
,
A I
,
Bl
,
AREA1
,
ERROR1
,
RESABS
,
DEF AB
1)
c
0053
IF (KEYt .EQ.
l)
tAll G57 (F
,
A2
,
B2
,
AREA2
,
ERROR2
,
RESABS.DEFAB2)
005~
lF(KEYF.EQ.2)tAll GK7(F
,
A2
,
B2
,
AREA2
,
ERRO
Il
2
,
RES^BS
,
DEFAB21
0055
IF(KEYF.EQ.3)tAll G515(F
,
A2
,
B2
,
AREA2
,
ERRO
Il
2
,
RESABS.DEFAB2)
0056
IF(KEYF.EQ.~)tAll
GKI5(F
,
A2
,
B2
,
ARJA2
,
ERROR2
,
RESABS.DEFAB2)
0057
rr (KEYF .EQ.5) tAll GK21 (F
,
A2.B2.A
Il
EA2.ERROR2
,
RESABS DEFAB2)
0058
IF (KEYF .EQ.6) C^ll GK31 (F
,
A2.B2.AREA2.ERROR2.RESABS.DEFM21
。059
I F
(~EYF
• EQ.
J)
CAll ^N
C]
(F
,
A2
,
82
,
AREA2. ERROR2. RESABS. DEF AB 2)
E
E
IHPROVE PREVIOU5 APPROX. TO INTEGRAl AND ERROR
E
0060
0061
0062
0063
006~
0065
0066
NEVAl-NEV^l+1
A
Il
EAI2-^REAI+ARE^2
ERROI2-ERRORI+ERROR2
EnRSUH-ERRSUI件 ERROI2-E Il R /I^X
AREA-^REA+AREAI2-RlI5Tþ1AXERR)
IF (OEFAB1.EQ.ERRORl .OR. OEFAB2.EQ.ERROR2) GOTO 5
1 F (AB5 (Rll 5T
(/l
AXERR) -AREA 12) .lE. 1 .OE -5"AB5 (^RE^ 12)
1<
.AHO.
ERIl 012.GE.9.9E ﹒ 1 晶 E IlIl M^X)
1 ROF F
1 鹽 IRO rF l+1
IF (lA5T .G
T.
IO :ANO. ERROI2.GT .ERRHAXI IROFF2-IROFF
2+
1
5
R
lI
5T
(例AXERR)"AREAl
RlI5T(lA5T)-^REA2
ER
Il
BNO"A
/I
^X 1 (EP5AB5
,
EP5
1lEl
"'^,
'5 (AREA) )
1 F (ERR5U
I\
.l
E.
ERR8tto) GOTO 8
。067
0068
0069
0070
0071
PIMP-up--TE5T FOR ROUHOOFF ERROR AHO 5ET ERROR Fl^G
。072
00
]3
IF(IROFF1.G
E.
6 .OR. I
Il
OFF
2.
G
E.
20) IERs2
IF (lA5
T.
EQ.
lI
HIT) IER"I
007
"
0075
0076
F-up--F •• "HUIF(ERROR2.GT.ERRORI) GOTO 30
A
1I
5T (lAST) "^2
BlI5T(MAXE
Il
R)"Bl
曲曲曲
APPEND THE NEWlY-tREATED INTERVAlS TO THE lA5T
晶聶聶
-
658 一
的分法則之一比較
FORTHAN IV GI
RElEASE 2.0
AGQ
OATE - THU JUN 06
,
1985
。077
0078
0079
0080
0081
0082
0083
008
,.
0085
0086
0087
0088
0089
0090
0091
0092
0093
009"
0095
0096
0097
0098
0099
0100
0101
0102
0103
010"
0105
0106
0101
911 ST (lAST) "92
ElIST(MAXERR)-ERRORl
ElIST(lAST)-ERROR2
GOTO 20
30
AlIST(
I\
AXERR)"A2
E
All ST
(lAST) 吋 1
9l
1
ST (lAS
T)
"'9 1
RlIST(
I\
AXERR)-AREA2
RlIST(lAST)~AREAI
El
IST(IIAXERR)"ERROR2
ElIST(lAST)-ERRORI
C
CAll SORT TO
"AINTAINτUE
OESCEIIO I IIG OROER IIIG 111 111E
1I
51
C
OF ERROR ESTIMATES ANO SELECT THE 5UBINTERVAl WITH 1HE
C
lARGEST ERROR ESTIMATE
c
20
CAll SORT
(lI
M
IT
,
LAST
,
"AXERR
,
ERRI'AX
,
El
I ST • I ORO
,
IIRI'AX)
IF(IER.IIE.O .OR. ERRSUM.lE.ERRBNO)GOTO "0
10
CON
Tl
NUE
c
C
COllrUTEτHE
RESUlT
c
"0
RESUl T..O". 0
00 50
K 自 I , lAST
RESUlT"RESUlT+RlIST(K)
50
CONT IIIUE
ABSEnR~ERRSU們
60
IF(KEYF.EQ.I)
IIEVAl-18 脅 IIEVAl+9
IF(KEYF.EQ.2)
NEVAl-l"~NEVAl+1
IF (KEYF .EQ.3)
NEVAl 四 "2 叫nVAl+21
IF (KEYF.EQ.")
NEVAl"'30~NEVAl+I5
IF(KEYF.EQ.5)
NEVAl-42 甜 NEVAl+21
lF(KEYF.EQ.6)
NEVAl"62 曲 NEVAl+31
IF(KEYF.E~.1) NEVAl-l" 曲 NEVAL+7
I
COurlT ~KEY F+ 2
WRITE(6.11' ICOUNT
11
FORMAT{/
,
I5X
,
'METHOO
('.11.') :')
999
RETURN
END
壘。 P T1 0llS
It
l
E Fr ECT 費
NOTERM.10 , E9CDIC , SOURCE , NOL1ST , NOOECK , lOAD.II0 I\ AP , II0TEST
壘。P T1 ÒIIS'
IN
E Fr ECT費
NAI\ E
• AGQ
,
LIIIECIIT •
56
會 STATISTICSA
SOURCE STATEMENTS -
l01.PROGRAM SIZE • 001658
禽 STATI
STI
CS 曲
NO
DIAGNOSTICS GENERATED
師大學報
第卅一期
FORTRA" IV GI
RElEASE 2.0
I1
AIN
DATE .. THU JUN 06
,
1985
E
E
甜食""""""""甜甜食*"*"""""****""*曲曲""""**甜甜甜甜食"甜食費曲曲貴晶晶晶晶晶晶食費C
"*
"i悟
c
壘,SUBROUTINE SORT(ORDERING ROUTINE)
費 4悟E
t."
f.
t.
E
"食費 *1‘甜甜食*""""會聶*""食費曲曲曲,甜甜1r壘,脅""""甜食,貴 ""ft"""" 甜甜 4、食聶 4時會飛 fd聖*
。 001
SUBROUTINE
50RT(ll l1 IT , l^ST ,們AXERR , ERI1AX , ElIST , IORO , NRMAX)
0002
。 I I1ENSION
ElIST(IOO)
,
IORD(IOO)
c
c
---
FIRST EXECUTABlE STATE
I1
ENT
c
0003
If.(lAST.GT.2)GOTO 10
OOO~
10RD (J)吋
0005
1 ORO (2)-2
。006
GOTO 90
。007
10
ERR
I1
AX-
El
I 5T
(,
'AXERR)
0008
If(NR
t1
^X.EQ.I)GOTO 30
。 009
100-NR們AX-I
。010
。o
20 1-1
,
100
。011
ISUCC叫。RO (IIR州AX-I)
E
。012
IF(ERRM^X.lE.ElIST(ISUCC)~GOTO
30
0013
10RO(NRM^X)-ISUCC
001~
t
lR/1
AX-NR
/l
AX-1
。015
20
CONTINUE
c
。016
30
JUP8N四 lAST
。017
IFtlAST.GT. (ll
t1
IT/2+2))
JUP8N-ll 州1T+3-lAST
。018
ERR
t1
IN-ELIST(LAST)
E
E
。019
J8NO-JUPB
f/
-1
0020
• 18EG':NR
t1^X+
1
。021
If(IBEG.GT.J8ND)GOTO
SO
。022
。o
"0 I-IBEG
,
JBNO
00
2)
ISUCC 叫 ORO
(1)
002~
If(ERR
I1
AX.GE.ELIST(ISUCC))GOTO 60
。025
10RO(I-
I)
-ISUCC
0026
~O
CONTINUE
0027
50
IORO(J8NO)"
/l
AXERR
。028
10RO(JUPBN)-LAST
0029
GOTO 90
c
。030
60
1 ORO
(1-1)
""^XERR
0031
R ﹒ JBN。
0032
。o
70 J.I
,
JBtl。
0033
1 SUCC"I ORO (K)
003~
IF(ERR
t1
IN.LT.ELIST(ISUCC))GOTO 80
。035
10RO
(K+ I)叫 SUCC
0036
K-K-l
。037
70
CON
Tl
NUE
0038
10RO(
I)
"lAST
0039
GOTO 90
OO~O
80
10RO(K+I)"LAST
-
660 一
積分法則之一比較
DATE .. TIIU JUU oG
,
1985
SORT
RELEASE 2.0
FORT
1t
AN IV GI
SET MAXERR AND ERMAX
---們AXERR~IORD(NRMAX ,
ER"AXftElIST(MAXERR
,
RETURN
END
nu
ptupbpbny
O
Olt
l
00"2
0
0lt
3
00""
脅。 PTlOtlS
1"
EFFE r:T曲
NOTERM , ID , EBCPIC , SOURCE , NOlIST , NODECK , lOAD , NOHAP , HOTEST
會 OPTIOH5
IN EFFECT*
NAME - 50RT
,
ll"ECNT -
56
*STA
Tl
S
Tl
CS*
SOURCE
$T
ATE
IlE
NTS ..
"",
PROGRAM SIH - 00
0lt
E2
*STATISTICS*
NO DIAGNOSTICS GENERATED
DATE - THU JUN 06
,
1985
費 4、費由*************食費曲曲,甜食費 **ft*** It*1t甜甜食It ltlt 疊1t*1t曲 ftftft 曲 4、飛食,、寅貴聶甜 食 4、
It
ft
SUBROUTINE ADAPTIVE NEWTON-COTES
(7 個 PTS)
州
甜甜 狗食
晶晶費由甜食It i悟,食費壘 ftft 費負甜甜It ltlt 賣會 Mtftftltftlt 飛飛禽It ft 費食費It lti、會食食,悟 4、最*,:fd悟 fd.fti、
SUBROUTINE AHC7(F
,
A
,
B
,
RESUlT
,
ABSERR
,
RESABS
,
RESASC)
DIMENSION FV(])
MAI"
RELEASE 2.0
PLPLPLFzubL
FOI月 TR ^,'
IV G 1
1I
"0.5
1t
(B-A'
FV(I'''F( 的
FV (2' -F
(A+
II/". 0'
FV
(3,
-F
(A+1I
/2 .0'
FV ('" -F (A+II'
FV (5' -F
(A+3.0州 1 2. 0'
FV (G
,
-F
(A+'7.0 州 1".0)
FV(])-F (B)
E
0001
0002
0003
000"
0005
0006
0007
0008
0009
0010
COMPUTE THE INTEGRAl
n"
,-p3.+
.. ι 呵‘圖,,
1
••
,由 ‘ 1 也可 tnu 、 ynv , 1•
.
"."v ﹒旬,
1
,
b'
,
nuMV.',1nv
prnv,
3."
攪勻,‘""壘, Env41."
‘
1
.)()
、 4 •• , &R 、.',',bnUFHJ
ap{-(
‘1'"vnouv
、, Ezt 刊 rRJ+O{
, E. 、••
'‘
dpav2+B
Prf、。‘F-n+V-m'+
‘.',pphJ'、.'、',‘‘JSE‘
•••
唔,'‘.' ,.、‘",h",E、'‘ dMvnv'OM--'t
F.,fFV
'.、aMU‘', a ﹒‘,r 會."、'', 3 , t nvnv﹒肉"間,= ﹒呵,‘',‘-R" 回 2+V+ ﹒闕 ,‘J'‘.',r、'',1+‘',AR‘',
、 1 可 Inv.' 。 0){-{
..
',HV."",',唔,hJ,1,
••
nu',.,.,
MVAvnu,E‘nu',
',、',.", 3 旬,‘'。+1.TB+,f
‘'',.‘
••
,-nn‘.'‘',
﹒'",‘F,1、',‘F 'tprpp。"、',、 J MV'‘、',、nu'。﹒肉FARMV.'
••
,1
,10Fnpvuv
*.
+hFF
num苟,、',F 『 J',、,E、 •• HTZJ , 1 ,扭, 2 ,',phJ',、‘nRE'" ',、,巴、HVHH.R-R -n 膏,',Hnun,ts
••• 圓 "UURn-a 間 ,、",、 4. ,、J' ,且, EERHH"n"n
。"-m ‘"PLFLPL
0011
0012
0013
RESKH-RESK/(B- 的
RESASC"II 會 (5
"
9. 。由 (ABS
(FV
(1)
-RESKII
,
+ABS (FV (])
-RESK川 '+20
1
.8.0 份
青
(ABS
(FV
(2)
-RESK川 +ABS(FV(6)-RESKII"+2016.0ft(ABS(FV(3)-RESKII)~
It
ABS(FV(5)-RESKH))+"00".0州 BS(FV(")-RESKH')/6615.0
RESUlT"RESK
ABSERR'"ABS(RESK-RESG)
IF (RESASC.tl
E.
O.O .AIIO.
^目 SERR.N E. O.O,
* ABSERR-RESASC
It
A"INI
(I
.O
E+
OO
,
(2.0
E+
02
It
ABSERR/RESASC
,
UI .5
E+
00)
E
。01"
0015
。016
0017
0018
E
REτURN
END
機 OrTIO"S
IN
EfFECT 曲
"O'ER" , ID , EBCDlt , S~URCE.NOlIST , NOOECK , lO^O.NO"^P.NOTES ,
ftor'IONS IN
EFFECT 晶
NA"E
- ANC7
,
llNECNT ..
56
食 ST I\Tl STICSft
SOURCE STATE
I1Et
ITS -
20.PROGRA
I1
SI
lE ..
00053且
ftS'A'ISTICS*
HO DIAG"OSTICS GE"ERATED
- 661
0019
0020
第卅一期
師大學報
DATE • THU JUII 06
,
1985
曲曲曲晶***曲曲曲曲曲疊疊,脅,有曲曲曲*會曲曲,曲曲曲曲*禽,曲曲曲 ftft 甜食費**'悟聶聶食費*聶晶晶聶 4、 4時
ftft
**
膏 ft
SUBROUTINE GAUSSIAN
APPROXIMATION(l~PTS)
由*
ft 音 機會
ftft 晶晶晶食費 4、食 ft*ftftftftft 甜甜食費""甜甜食費 4、 ftft 由#費 """frfrfrft 食費由 frf.f.frf. 舜,﹒\j、繭 *:'1悟 fr