Trailing-Edge
-
PDP-10 Archives
-
decus_20tap2_198111
-
decus/20-0026/acfi.ssp
There are 2 other files named acfi.ssp in the archive. Click here to see a list.
C ACFI 10
C ..................................................................ACFI 20
C ACFI 30
C SUBROUTINE ACFI ACFI 40
C ACFI 50
C PURPOSE ACFI 60
C TO INTERPOLATE FUNCTION VALUE Y FOR A GIVEN ARGUMENT VALUE ACFI 70
C X USING A GIVEN TABLE (ARG,VAL) OF ARGUMENT AND FUNCTION ACFI 80
C VALUES. ACFI 90
C ACFI 100
C USAGE ACFI 110
C CALL ACFI (X,ARG,VAL,Y,NDIM,EPS,IER) ACFI 120
C ACFI 130
C DESCRIPTION OF PARAMETERS ACFI 140
C X - THE ARGUMENT VALUE SPECIFIED BY INPUT. ACFI 150
C ARG - THE INPUT VECTOR (DIMENSION NDIM) OF ARGUMENT ACFI 160
C VALUES OF THE TABLE (POSSIBLY DESTROYED). ACFI 170
C VAL - THE INPUT VECTOR (DIMENSION NDIM) OF FUNCTION ACFI 180
C VALUES OF THE TABLE (DESTROYED). ACFI 190
C Y - THE RESULTING INTERPOLATED FUNCTION VALUE. ACFI 200
C NDIM - AN INPUT VALUE WHICH SPECIFIES THE NUMBER OF ACFI 210
C POINTS IN TABLE (ARG,VAL). ACFI 220
C EPS - AN INPUT CONSTANT WHICH IS USED AS UPPER BOUND ACFI 230
C FOR THE ABSOLUTE ERROR. ACFI 240
C IER - A RESULTING ERROR PARAMETER. ACFI 250
C ACFI 260
C REMARKS ACFI 270
C (1) TABLE (ARG,VAL) SHOULD REPRESENT A SINGLE-VALUED ACFI 280
C FUNCTION AND SHOULD BE STORED IN SUCH A WAY, THAT THE ACFI 290
C DISTANCES ABS(ARG(I)-X) INCREASE WITH INCREASING ACFI 300
C SUBSCRIPT I. TO GENERATE THIS ORDER IN TABLE (ARG,VAL), ACFI 310
C SUBROUTINES ATSG, ATSM OR ATSE COULD BE USED IN A ACFI 320
C PREVIOUS STAGE. ACFI 330
C (2) NO ACTION BESIDES ERROR MESSAGE IN CASE NDIM LESS ACFI 340
C THAN 1. ACFI 350
C (3) INTERPOLATION IS TERMINATED EITHER IF THE DIFFERENCE ACFI 360
C BETWEEN TWO SUCCESSIVE INTERPOLATED VALUES IS ACFI 370
C ABSOLUTELY LESS THAN TOLERANCE EPS, OR IF THE ABSOLUTE ACFI 380
C VALUE OF THIS DIFFERENCE STOPS DIMINISHING, OR AFTER ACFI 390
C (NDIM-1) STEPS (THE NUMBER OF POSSIBLE STEPS IS ACFI 400
C DIMINISHED IF AT ANY STAGE INFINITY ELEMENT APPEARS IN ACFI 410
C THE DOWNWARD DIAGONAL OF INVERTED-DIFFERENCES-SCHEME ACFI 420
C AND IF IT IS IMPOSSIBLE TO ELIMINATE THIS INFINITY ACFI 430
C ELEMENT BY INTERCHANGING OF TABLE POINTS). ACFI 440
C FURTHER IT IS TERMINATED IF THE PROCEDURE DISCOVERS TWO ACFI 450
C ARGUMENT VALUES IN VECTOR ARG WHICH ARE IDENTICAL. ACFI 460
C DEPENDENT ON THESE FOUR CASES, ERROR PARAMETER IER IS ACFI 470
C CODED IN THE FOLLOWING FORM ACFI 480
C IER=0 - IT WAS POSSIBLE TO REACH THE REQUIRED ACFI 490
C ACCURACY (NO ERROR). ACFI 500
C IER=1 - IT WAS IMPOSSIBLE TO REACH THE REQUIRED ACFI 510
C ACCURACY BECAUSE OF ROUNDING ERRORS. ACFI 520
C IER=2 - IT WAS IMPOSSIBLE TO CHECK ACCURACY BECAUSE ACFI 530
C NDIM IS LESS THAN 2, OR THE REQUIRED ACCURACY ACFI 540
C COULD NOT BE REACHED BY MEANS OF THE GIVEN ACFI 550
C TABLE. NDIM SHOULD BE INCREASED. ACFI 560
C IER=3 - THE PROCEDURE DISCOVERED TWO ARGUMENT VALUES ACFI 570
C IN VECTOR ARG WHICH ARE IDENTICAL. ACFI 580
C ACFI 590
C SUBROUTINES AND FUNCTION SUBPROGRAMS REQUIRED ACFI 600
C NONE ACFI 610
C ACFI 620
C METHOD ACFI 630
C INTERPOLATION IS DONE BY CONTINUED FRACTIONS AND INVERTED- ACFI 640
C DIFFERENCES-SCHEME. ON RETURN Y CONTAINS AN INTERPOLATED ACFI 650
C FUNCTION VALUE AT POINT X, WHICH IS IN THE SENSE OF REMARK ACFI 660
C (3) OPTIMAL WITH RESPECT TO GIVEN TABLE. FOR REFERENCE, SEE ACFI 670
C F.B.HILDEBRAND, INTRODUCTION TO NUMERICAL ANALYSIS, ACFI 680
C MCGRAW-HILL, NEW YORK/TORONTO/LONDON, 1956, PP.395-406. ACFI 690
C ACFI 700
C ..................................................................ACFI 710
C ACFI 720
SUBROUTINE ACFI(X,ARG,VAL,Y,NDIM,EPS,IER) ACFI 730
C ACFI 740
C ACFI 750
DIMENSION ARG(1),VAL(1) ACFI 760
IER=2 ACFI 770
IF(NDIM)20,20,1 ACFI 780
1 Y=VAL(1) ACFI 790
DELT2=0. ACFI 800
IF(NDIM-1)20,20,2 ACFI 810
C ACFI 820
C PREPARATIONS FOR INTERPOLATION LOOP ACFI 830
2 P2=1. ACFI 840
P3=Y ACFI 850
Q2=0. ACFI 860
Q3=1. ACFI 870
C ACFI 880
C ACFI 890
C START INTERPOLATION LOOP ACFI 900
DO 16 I=2,NDIM ACFI 910
II=0 ACFI 920
P1=P2 ACFI 930
P2=P3 ACFI 940
Q1=Q2 ACFI 950
Q2=Q3 ACFI 960
Z=Y ACFI 970
DELT1=DELT2 ACFI 980
JEND=I-1 ACFI 990
C ACFI1000
C COMPUTATION OF INVERTED DIFFERENCES ACFI1010
3 AUX=VAL(I) ACFI1020
DO 10 J=1,JEND ACFI1030
H=VAL(I)-VAL(J) ACFI1040
IF(ABS(H)-1.E-6*ABS(VAL(I)))4,4,9 ACFI1050
4 IF(ARG(I)-ARG(J))5,17,5 ACFI1060
5 IF(J-JEND)8,6,6 ACFI1070
C ACFI1080
C INTERCHANGE ROW I WITH ROW I+II ACFI1090
6 II=II+1 ACFI1100
III=I+II ACFI1110
IF(III-NDIM)7,7,19 ACFI1120
7 VAL(I)=VAL(III) ACFI1130
VAL(III)=AUX ACFI1140
AUX=ARG(I) ACFI1150
ARG(I)=ARG(III) ACFI1160
ARG(III)=AUX ACFI1170
GOTO 3 ACFI1180
C ACFI1190
C COMPUTATION OF VAL(I) IN CASE VAL(I)=VAL(J) AND J LESS THAN I-1 ACFI1200
8 VAL(I)=1.7E38 ACFI1210
GOTO 10 ACFI1220
C ACFI1230
C COMPUTATION OF VAL(I) IN CASE VAL(I) NOT EQUAL TO VAL(J) ACFI1240
9 VAL(I)=(ARG(I)-ARG(J))/H ACFI1250
10 CONTINUE ACFI1260
C INVERTED DIFFERENCES ARE COMPUTED ACFI1270
C ACFI1280
C COMPUTATION OF NEW Y ACFI1290
P3=VAL(I)*P2+(X-ARG(I-1))*P1 ACFI1300
Q3=VAL(I)*Q2+(X-ARG(I-1))*Q1 ACFI1310
IF(Q3)11,12,11 ACFI1320
11 Y=P3/Q3 ACFI1330
GOTO 13 ACFI1340
12 Y=1.7E38 ACFI1350
13 DELT2=ABS(Z-Y) ACFI1360
IF(DELT2-EPS)19,19,14 ACFI1370
14 IF(I-8)16,15,15 ACFI1380
15 IF(DELT2-DELT1)16,18,18 ACFI1390
16 CONTINUE ACFI1400
C END OF INTERPOLATION LOOP ACFI1410
C ACFI1420
C ACFI1430
RETURN ACFI1440
C ACFI1450
C THERE ARE TWO IDENTICAL ARGUMENT VALUES IN VECTOR ARG ACFI1460
17 IER=3 ACFI1470
RETURN ACFI1480
C ACFI1490
C TEST VALUE DELT2 STARTS OSCILLATING ACFI1500
18 Y=Z ACFI1510
IER=1 ACFI1520
RETURN ACFI1530
C ACFI1540
C THERE IS SATISFACTORY ACCURACY WITHIN NDIM-1 STEPS ACFI1550
19 IER=0 ACFI1560
20 RETURN ACFI1570
END ACFI1580