Homec4science
Diffusion COSOlver (llechot) Browse

cosolver/FM/src/SampleFM.f90
565bccac0d42llechot

Last Change
Raw File

SampleFM.f90
View Options

PROGRAM SAMPLE
! This is a sample program using the FMZM modules for doing arithmetic using the FM, IM, and ZM
! derived types.
! The program's output to the screen is also saved in file SampleFM.out.
! The program checks all the results and the last line of the output file should be
! "All results were ok."
! These examples show various ways to use the FM package, but the methods used are not always
! the most advanced for the sample problem.
USE FMZM
IMPLICIT NONE
! Declare the multiple precision variables. The three types are:
! (FM) for multiple precision real
! (IM) for multiple precision integer
! (ZM) for multiple precision complex
TYPE (FM), SAVE :: X1, X2, X3, X4
TYPE (FM), SAVE, ALLOCATABLE :: A(:,:), B(:,:), V1(:), V2(:)
TYPE (IM), SAVE :: I1, I2, I3
TYPE (ZM), SAVE :: Z1, Z2, Z3, Z4
! Declare the function name of a type (FM) function that will be passed as an argument
! to a subroutine called from this program.
TYPE (FM), EXTERNAL :: F
! Declare the other variables (not multiple precision).
CHARACTER(80) :: ST1
CHARACTER(175) :: FMT
INTEGER :: ITER, J, K, KOUT, NERROR
INTEGER :: SEED(7)
DOUBLE PRECISION :: VALUE
! Write output to the screen (unit *), and also to the file SampleFM.out.
KOUT = 18
OPEN (KOUT,FILE='SampleFM.out')
NERROR = 0
! 1. Find a root of the equation f(x) = x**5 - 3x**4 + x**3 - 4x**2 + x - 6 = 0.
! Set precision to give at least 60 significant digits.
CALL FM_SET(60)
! Use Newton's method with initial guess x = 3.12.
! Horner's rule is used to evaluate the function.
! X1 is the previous iterate.
! X2 is the current iterate.
! TO_FM is a function for converting other types of numbers to type FM. Note that
! TO_FM(3.12) converts the REAL constant to FM, but it is accurate only to single
! precision, since the number 3.12 cannot be represented exactly in binary and has
! already been rounded to single precision. Similarly, TO_FM(3.12D0) agrees with
! 3.12 to double precision accuracy, and TO_FM('3.12') or TO_FM(312)/TO_FM(100)
! agrees to full FM accuracy.
! Here, TO_FM(3.12) would be ok, since Newton iteration will correct the error
! coming from single precision, but it is a good habit to use the more accurate
! form.
X1 = TO_FM('3.12')
! Print the first iteration.
FMT = "(//' Sample 1. Real root of f(x) = x**5 - 3x**4 + x**3 - 4x**2 + x - 6 = 0.'//"// &
"' Iteration Newton approximation')"
WRITE (* ,FMT)
WRITE (KOUT,FMT)
! FM_FORM is a formatting subroutine.
CALL FM_FORM('F65.60',X1,ST1)
WRITE (* ,"(/I10,4X,A)") 0,TRIM(ST1)
WRITE (KOUT,"(/I10,4X,A)") 0,TRIM(ST1)
DO ITER = 1, 10
! X3 is f(X1).
X3 = ((((X1-3)*X1+1)*X1-4)*X1+1)*X1-6
! X4 is f'(X1).
X4 = (((5*X1-12)*X1+3)*X1-8)*X1+1
X2 = X1 - X3/X4
! Print each iteration.
CALL FM_FORM('F65.60',X2,ST1)
WRITE (* ,"(/I10,4X,A)") ITER,TRIM(ST1)
WRITE (KOUT,"(/I10,4X,A)") ITER,TRIM(ST1)
! Stop iterating if X1 and X2 agree to over 60 places.
X4 = ABS(X1-X2)
IF (X4 < 1.0D-61) EXIT
! Set X1 = X2 for the next iteration.
X1 = X2
ENDDO
! Check the answer.
X3 = TO_FM('3.1206562153267265004709560135237974846546239355990660149888284358')
! It is slightly safer to do this test with .NOT. instead of
! IF (ABS(X3-X2) >= 1.0D-61) THEN
! because if the result of ABS(X3-X2) is FM's UNKNOWN value,
! the comparison returns false for all comparisons.
IF (.NOT.(ABS(X3-X2) < 1.0D-61)) THEN
NERROR = NERROR + 1
WRITE (* ,"(/' Error in sample case number 1.'/)")
WRITE (KOUT,"(/' Error in sample case number 1.'/)")
ENDIF
! 2. Higher Precision. Compute the root above to 500 decimal places.
CALL FM_SET(500)
! It is tempting to just say X1 = X3 here to initialize the start of the higher
! precision iterations to be the check value defined above. That will not work,
! because precision has changed. Most of the digits of X3 may be undefined at
! the new precision.
! The most flexible way to pad a lower precision value with zeros when raising
! precision is to use subroutine FM_EQU, but here it is easier to re-define X1
! from scratch at the new precision.
X1 = TO_FM('3.1206562153267265004709560135237974846546239355990660149888284358')
DO ITER = 1, 10
! X3 is f(X1).
X3 = ((((X1-3)*X1+1)*X1-4)*X1+1)*X1-6
! X4 is f'(X1).
X4 = (((5*X1-12)*X1+3)*X1-8)*X1+1
X2 = X1 - X3/X4
! Stop iterating if X1 and X2 agree to over 500 places.
X4 = ABS(X1-X2)
! Compare this test to the similar one in case 1 above.
! For machines with 64-bit double precision, 1.0D-501 would be smaller than the
! smallest positive number. So this error tolerance is converted to an FM number
! from character form.
IF (X4 < TO_FM('1.0E-501')) EXIT
! Set X1 = X2 for the next iteration.
X1 = X2
ENDDO
! For very high precision output, it is sometimes more convenient to use FM_PRINT
! to format and print the numbers, since the line breaks are handled automatically.
! The unit number for the output, KW, and the format codes to be used, JFORM1 and
! JFORM2, are internal FM variables.
! Subroutine FM_SETVAR is used to re-define these, and the new values will remain in
! effect for any further calls to FM_PRINT.
! Other variables that can be changed and the options they control are listed in the
! documentation at the top of file FM.f95.
! Set the FM_PRINT format to F505.500
CALL FM_SETVAR(' JFORM1 = 2 ')
CALL FM_SETVAR(' JFORM2 = 500 ')
! Set the output screen width to 90 columns.
CALL FM_SETVAR(' KSWIDE = 90 ')
FMT = "(///' Sample 2. Find the root above to 500 decimal places.'/)"
WRITE (* ,FMT)
WRITE (KOUT,FMT)
! Write to the output file.
CALL FM_SETVAR(' KW = 18 ')
CALL FM_PRINT(X2)
! Write to the screen (unit 6).
CALL FM_SETVAR(' KW = 6 ')
CALL FM_PRINT(X2)
! Check the answer.
X3 = TO_FM('3.1206562153267265004709560135237974846546239355990660149888284358190264999' // &
'517954689783257450017151095811923431332682839420040840535954560118152245371'// &
'792881305271951017118938898212403662058307303983547376913282000110058273504'// &
'202838670709895619275413484521549282591891156945200789415818387529512010999'// &
'602155131321076797099026664236992803703462570149559724389392331827597552460'// &
'610612200485579529156910428115547013787714423708578161025641555097481179969'// &
'175028390105904786831680128384331143259309155577171683842444352768419176139060')
IF (.NOT.(ABS(X3-X2) < TO_FM('1.0E-501'))) THEN
NERROR = NERROR + 1
WRITE (* ,"(/' Error in sample case number 2.'/)")
WRITE (KOUT,"(/' Error in sample case number 2.'/)")
ENDIF
! 3. Compute the Riemann zeta function for s=3.
! Use Gosper's formula: zeta(3) =
! (5/4)*Sum[ (-1)**k * (k!)**2 / ((k+1)**2 * (2k+1)!) ]
! while k = 0, 1, ....
! X1 is the current partial sum.
! X2 is the current term.
! X3 is k!
! X4 is (2k+1)!
CALL FM_SET(60)
X1 = 1
X3 = 1
X4 = 1
DO K = 1, 200
X3 = K*X3
J = 2*K*(2*K+1)
X4 = J*X4
X2 = X3**2
J = (K+1)*(K+1)
X2 = (X2/J)/X4
IF (MOD(K,2) == 0) THEN
X1 = X1 + X2
ELSE
X1 = X1 - X2
ENDIF
! Test for convergence.
! Here the rate of convergence is much slower than in the Newton iterations above.
! Asking for 60 digits in the call to FM_SET will cause the internal precision to
! be set slightly higher than that, giving the user a few guard digits.
! X2 is the difference between the two most recent partial sums, so the test
! below will stop the sum when the last two partial sums agree to at least 65
! significant digits.
IF (ABS(X2/X1) < 1.0D-65) THEN
FMT = "(///' Sample 3.',2X,I5,' terms were added in the zeta sum.'/)"
WRITE (* ,FMT) K
WRITE (KOUT,FMT) K
EXIT
ENDIF
ENDDO
! Print the result.
X1 = (5*X1)/4
CALL FM_FORM('F63.60',X1,ST1)
WRITE (* ,"(' zeta(3) = ',A)") TRIM(ST1)
WRITE (KOUT,"(' zeta(3) = ',A)") TRIM(ST1)
! Check the answer.
X3 = TO_FM('1.202056903159594285399738161511449990764986292340498881792271555')
IF (.NOT.(ABS(X1-X3) < 1.0D-61)) THEN
NERROR = NERROR + 1
WRITE (* ,"(/' Error in sample case number 3.'/)")
WRITE (KOUT,"(/' Error in sample case number 3.'/)")
ENDIF
! 4. Integer multiple precision calculations.
! Fermat's theorem says x**(p-1) mod p = 1 when p is prime and x is not a
! multiple of p.
! If x**(p-1) mod p gives 1 for some p with several different x's, then it is
! very likely that p is prime (but it is not certain until further tests are done).
! Find a 70-digit number p that is "probably" prime.
! Use FM_RANDOM_NUMBER to generate a random 70-digit starting value and search for
! a prime from that point.
! Initialize the generator.
! Note that VALUE is double precision, unlike the similar Fortran intrinsic random
! number routine, which returns a single-precision result.
SEED = (/ 2718281,8284590,4523536,0287471,3526624,9775724,7093698 /)
CALL FM_RANDOM_SEED_PUT(SEED)
! I1 is the value p being tested.
I1 = 0
I3 = TO_IM(10)**13
DO J = 1, 6
CALL FM_RANDOM_NUMBER(VALUE)
I2 = 1.0D13*VALUE
I1 = I1*I3 + I2
ENDDO
I3 = TO_IM(10)**70
I1 = MOD(I1,I3)
! To speed up the search, test only values that are not
! multiples of 2, 3, 5, 7, 11, 13.
K = 2*3*5*7*11*13
I1 = (I1/K)*K + K + 1
I3 = 3
DO J = 1, 100
I2 = I1 - 1
! Compute 3**(p-1) mod p
I3 = POWER_MOD(I3,I2,I1)
IF (I3 == 1) THEN
! Check that 7**(p-1) mod p is also 1.
I3 = 7
I3 = POWER_MOD(I3,I2,I1)
IF (I3 == 1) THEN
FMT = "(///' Sample 4.',2X,I5,' values were checked before finding a prime p.'/)"
WRITE (* ,FMT) J
WRITE (KOUT,FMT) J
EXIT
ENDIF
ENDIF
I3 = 3
I1 = I1 + K
ENDDO
! Print the result.
CALL IM_FORM('I72',I1,ST1)
WRITE (* ,"(' p =',A)") TRIM(ST1)
WRITE (KOUT,"(' p =',A)") TRIM(ST1)
! Check the answer.
I3 = TO_IM('9552131129056058313103536357738804884840825498503088946760768419490591')
IF (.NOT.(I1 == I3)) THEN
NERROR = NERROR + 1
WRITE (* ,"(/' Error in sample case number 4.'/)")
WRITE (KOUT,"(/' Error in sample case number 4.'/)")
ENDIF
! 5. Log Integral function.
! Estimate the number of primes less than 10**30.
FMT = "(///' Sample 5. Log integral. Estimate the number of primes less than 10**30.'/"// &
"' It should be accurate to about 15 significant digits.'/)"
WRITE (* ,FMT)
WRITE (KOUT,FMT)
I2 = TO_IM(LOG_INTEGRAL(TO_FM('1.0E+30')))
! Print the result.
CALL IM_FORM('I30',I2,ST1)
WRITE (* ,"(' int(li(1.0e+30)) = ',A)") TRIM(ST1)
WRITE (KOUT,"(' int(li(1.0e+30)) = ',A)") TRIM(ST1)
! Check the answer.
I3 = TO_IM('14692398897720447639079087669')
IF (.NOT.(I2 == I3)) THEN
NERROR = NERROR + 1
WRITE (* ,"(/' Error in sample case number 5.'/)")
WRITE (KOUT,"(/' Error in sample case number 5.'/)")
ENDIF
! 6. Gamma function.
! Check that gamma(1/2) is sqrt(pi)
FMT = "(///' Sample 6. Check that gamma(1/2) = sqrt(pi).'/)"
WRITE (* ,FMT)
WRITE (KOUT,FMT)
X2 = GAMMA(TO_FM('0.5'))
! Print the result.
CALL FM_FORM('F63.60',X2,ST1)
WRITE (* ,"(' gamma(1/2) = ',A)") TRIM(ST1)
WRITE (KOUT,"(' gamma(1/2) = ',A)") TRIM(ST1)
! Check the answer.
X3 = SQRT(ACOS(TO_FM(-1)))
IF (.NOT.(ABS(X3-X2) < 1.0D-61)) THEN
NERROR = NERROR + 1
WRITE (* ,"(/' Error in sample case number 6.'/)")
WRITE (KOUT,"(/' Error in sample case number 6.'/)")
ENDIF
! 7. Psi and polygamma functions.
! Rational series can often be summed using these functions.
! Sum (n=1 to infinity) 1/(n**2 * (8n+1)**2) =
! 16*(psi(1) - psi(9/8)) + polygamma(1,1) + polygamma(1,9/8)
! Reference: Abramowitz & Stegun, Handbook of Mathematical Functions,
! chapter 6, Example 10.
FMT = "(///' Sample 7. Psi and polygamma functions.'/)"
WRITE (* ,FMT)
WRITE (KOUT,FMT)
X2 = 16*(PSI(TO_FM(1)) - PSI(TO_FM(9)/8)) + POLYGAMMA(1,TO_FM(1)) + POLYGAMMA(1,TO_FM(9)/8)
! Print the result.
CALL FM_FORM('F65.60',X2,ST1)
FMT = "(' Sum (n=1 to infinity) 1/(n**2 * (8n+1)**2) = '/9X,A)"
WRITE (* ,FMT) TRIM(ST1)
WRITE (KOUT,FMT) TRIM(ST1)
! Check the answer.
X3 = TO_FM('1.3499486145413024755107829105035147950644978635837270816327396M-2')
IF (.NOT.(ABS(X3-X2) < 1.0D-61)) THEN
NERROR = NERROR + 1
WRITE (* ,"(/' Error in sample case number 7.'/)")
WRITE (KOUT,"(/' Error in sample case number 7.'/)")
ENDIF
! 8. Incomplete gamma and gamma functions.
! Find the probability that an observed chi-square for a correct model should be
! less that 2.3 when the number of degrees of freedom is 5.
! Reference: Knuth, Volume 2, 3rd ed., Page 56, and Press, Flannery, Teukolsky,
! Vetterling, Numerical Recipes, 1st ed., Page 165.
FMT = "(///' Sample 8. Incomplete gamma and gamma functions.'/)"
WRITE (* ,FMT)
WRITE (KOUT,FMT)
X1 = TO_FM(5)/2
X2 = INCOMPLETE_GAMMA1(X1,TO_FM('2.3')/2) / GAMMA(X1)
! Print the result.
CALL FM_FORM('F62.60',X2,ST1)
WRITE (* ,"(' Probability = ',A)") TRIM(ST1)
WRITE (KOUT,"(' Probability = ',A)") TRIM(ST1)
! Check the answer.
X3 = TO_FM('0.19373313011487144632751025918250599953472318607121386973066283739')
IF (.NOT.(ABS(X3-X2) < 1.0D-61)) THEN
NERROR = NERROR + 1
WRITE (* ,"(/' Error in sample case number 8.'/)")
WRITE (KOUT,"(/' Error in sample case number 8.'/)")
ENDIF
! 9. Error function.
! Find the probability that a value drawn from a normal distribution is within
! 1 or 2 or 3 standard deviations from the mean.
FMT = "(///' Sample 9. Error function. Probability that a value drawn from a normal'/"// &
"' distribution is within k standard deviations from the mean.'/)"
WRITE (* ,FMT)
WRITE (KOUT,FMT)
DO K = 1, 3
X1 = K / SQRT(TO_FM(2))
X2 = ERF(X1)
! Print the results.
CALL FM_FORM('F52.50',X2,ST1)
WRITE (* ,"(' k = ',I2,', probability = ',A)") K,TRIM(ST1)
WRITE (KOUT,"(' k = ',I2,', probability = ',A)") K,TRIM(ST1)
! Check the answer.
IF (K == 1) THEN
X3 = TO_FM('0.68268949213708589717046509126407584495582593345320878197478890049')
ELSE IF (K == 2) THEN
X3 = TO_FM('0.95449973610364158559943472566693312505644755259664313203266799974')
ELSE
X3 = TO_FM('0.99730020393673981094669637046481004524434126368323870127155602929')
ENDIF
IF (.NOT.(ABS(X3-X2) < 1.0D-61)) THEN
NERROR = NERROR + 1
WRITE (* ,"(/' Error in sample case number 9.'/)")
WRITE (KOUT,"(/' Error in sample case number 9.'/)")
ENDIF
ENDDO
! 10. Array operations.
! Find the dominant eigenvalue and a corresponding eigenvector for this 5x5 matrix:
! 3 1 4 1 5
! 9 2 6 5 3
! A = 5 8 9 7 9
! 3 2 3 8 4
! 6 2 6 4 3
! Use the power method. Compute B = A**n. If v1 is an initial guess for the
! largest magnitude eigenvector, v2 = B*v1 should be a more accurate approximation.
! The ratio of the elements of v3 = A*v2 to those of v2 gives an estimate of the
! corresponding eigenvalue. By repeatedly squaring the matrix, each iteration uses
! the next higher power of 2 for n.
FMT = "(///' Sample 10. Eigenvalue from matrix powers.')"
WRITE (* ,FMT)
WRITE (KOUT,FMT)
! These type FM arrays were declared as allocatable. Allocate them now, and initialize.
ALLOCATE( A(5,5) )
ALLOCATE( B(5,5) )
ALLOCATE( V1(5) )
ALLOCATE( V2(5) )
! To initialize the matrix, we can use array sections to set one row at a time, and the
! FMZM interface will take care of converting from integer to type (FM). If the values
! were not integers, we could say A(1,1:5) = (/ TO_FM(' 3.7 '),TO_FM(' 4.2 '), etc.
! The initialization could also be done using A = RESHAPE( ....
A(1,1:5) = (/ 3, 1, 4, 1, 5 /)
A(2,1:5) = (/ 9, 2, 6, 5, 3 /)
A(3,1:5) = (/ 5, 8, 9, 7, 9 /)
A(4,1:5) = (/ 3, 2, 3, 8, 4 /)
A(5,1:5) = (/ 6, 2, 6, 4, 3 /)
! Initialize all elements of the initial guess vector to 1.
V1 = 1
B = A
WRITE (* ,"(/' Iteration eigenvalue approximation ')")
WRITE (KOUT,"(/' Iteration eigenvalue approximation ')")
DO J = 1, 7
B = MATMUL(B,B)
V1 = MATMUL(B,V1)
V2 = MATMUL(A,V1)
X1 = V2(1) / V1(1)
CALL FM_FORM('F64.57',X1,ST1)
WRITE (* ,"(/I10,A)") J,TRIM(ST1)
WRITE (KOUT,"(/I10,A)") J,TRIM(ST1)
ENDDO
! Normalize the eigenvector (L-2 norm).
V2 = V2 / NORM2(V2)
WRITE (* ,"(' ')")
WRITE (KOUT,"(' ')")
WRITE (* ,"(' The corresponding eigenvector is')")
WRITE (KOUT,"(' The corresponding eigenvector is')")
WRITE (* ,"(' ')")
WRITE (KOUT,"(' ')")
DO J = 1, 5
CALL FM_FORM('F61.57',V2(J),ST1)
WRITE (* ,"(A)") TRIM(ST1)
WRITE (KOUT,"(A)") TRIM(ST1)
ENDDO
! Check the answer.
X3 = TO_FM('23.91276717232132858935703922800330450554912919599927298216827247803204')
IF (.NOT.(ABS(X3-X1) < 1.0D-61)) THEN
NERROR = NERROR + 1
WRITE (* ,"(/' Error in sample case number 10.'/)")
WRITE (KOUT,"(/' Error in sample case number 10.'/)")
ENDIF
! 11. Function and subroutine example.
! Find the integral from 0 to 1/2 of 2*exp(-x**2)/sqrt(pi).
! The exact value of the integral is erf(1/2).
! Use a simple numerical integration routine to apply an integration rule
! using 100 intervals.
CALL FM_SET(40)
FMT = "(///' Sample 11. Function and subroutine example.'/)"
WRITE (* ,FMT)
WRITE (KOUT,FMT)
X1 = 0
X2 = TO_FM(' 0.5 ')
CALL PLAN_9(F,X1,X2,100,X3)
! Print the result.
CALL FM_FORM('F32.30',X3,ST1)
WRITE (* ,"(' Integral = ',A)") TRIM(ST1)
WRITE (KOUT,"(' Integral = ',A)") TRIM(ST1)
! Check the answer.
X4 = ERF(TO_FM('0.5'))
IF (.NOT.(ABS(X3-X4) < 1.0D-31)) THEN
NERROR = NERROR + 1
WRITE (* ,"(/' Error in sample case number 11.'/)")
WRITE (KOUT,"(/' Error in sample case number 11.'/)")
ENDIF
! Complex arithmetic.
! Set precision to give at least 30 significant digits.
CALL FM_SET(30)
! 12. Find a complex root of the equation
! f(x) = x**5 - 3x**4 + x**3 - 4x**2 + x - 6 = 0.
! Newton's method with initial guess x = .56 + 1.06 i.
! Z1 is the previous iterate.
! Z2 is the current iterate.
Z1 = TO_ZM('.56 + 1.06 i')
! Print the first iteration.
FMT = "(///' Sample 12. Complex root of f(x) = x**5 - 3x**4 + x**3 - 4x**2 + x - 6 = 0.'," &
//"//' Iteration Newton approximation')"
WRITE (* ,FMT)
WRITE (KOUT,FMT)
CALL ZM_FORM('F32.30','F32.30',Z1,ST1)
WRITE (* ,"(/I6,4X,A)") 0,TRIM(ST1)
WRITE (KOUT,"(/I6,4X,A)") 0,TRIM(ST1)
DO ITER = 1, 10
! Z3 is f(Z1).
Z3 = ((((Z1-3)*Z1+1)*Z1-4)*Z1+1)*Z1-6
! Z4 is f'(Z1).
Z4 = (((5*Z1-12)*Z1+3)*Z1-8)*Z1+1
Z2 = Z1 - Z3/Z4
! Print each iteration.
CALL ZM_FORM('F32.30','F32.30',Z2,ST1)
WRITE (* ,"(/I6,4X,A)") ITER,TRIM(ST1)
WRITE (KOUT,"(/I6,4X,A)") ITER,TRIM(ST1)
! Stop iterating if Z1 and Z2 agree to over 30 places.
IF (ABS(Z1-Z2) < 1.0D-31) EXIT
! Set Z1 = Z2 for the next iteration.
Z1 = Z2
ENDDO
! Check the answer.
Z3 = TO_ZM('0.561958308335403235498111195347453 + 1.061134679604332556983391239058885 i')
IF (.NOT.(ABS(Z3-Z2) < 1.0D-31)) THEN
NERROR = NERROR + 1
WRITE (* ,"(/' Error in sample case number 12.'/)")
WRITE (KOUT,"(/' Error in sample case number 12.'/)")
ENDIF
! 13. Compute exp(1.23-2.34i).
! Use the direct Taylor series.
! Z1 is x.
! Z2 is the current term, x**n/n!.
! Z3 is the current partial sum.
Z1 = TO_ZM('1.23-2.34i')
Z2 = 1
Z3 = 1
DO K = 1, 100
Z2 = Z2*Z1/K
Z4 = Z3 + Z2
! Test for convergence.
! This is a common way to check for series convergence -- wait until the term
! being added is so close to zero that the sum does not change. That is fine
! here, because we are using the default round-to-nearest rounding mode.
! There is a pitfall if we were to re-run the program with a different rounding
! mode. For example, if we change the rounding mode to round toward +infinity,
! then at 30-digit precision the addition 1.2 + 3.4e-100 rounds up to 1.200...001
! and so the test to see if the sum did not change might never be satisfied.
! This problem can occur with either type FM or ZM sums.
! For cases where other rounding modes might be used, doing the convergence check
! like we did in the zeta sum of example 3 above is better. Here that would be
! IF (ABS(Z2/Z3) < 1.0D-35) THEN
IF (Z4 == Z3) THEN
FMT = "(///' Sample 13.',2X,I5,' terms were added to get exp(1.23-2.34i).'/)"
WRITE (* ,FMT) K
WRITE (KOUT,FMT) K
EXIT
ENDIF
Z3 = Z4
ENDDO
! Print the result.
CALL ZM_FORM('F33.30','F32.30',Z3,ST1)
WRITE (* ,"(' Result= ',A)") TRIM(ST1)
WRITE (KOUT,"(' Result= ',A)") TRIM(ST1)
! Check the answer.
Z4 = TO_ZM('-2.379681796854777515745457977696745 - 2.458032970832342652397461908326042 i')
IF (.NOT.(ABS(Z4-Z3) < 1.0D-31)) THEN
NERROR = NERROR + 1
WRITE (* ,"(/' Error in sample case number 13.'/)")
WRITE (KOUT,"(/' Error in sample case number 13.'/)")
ENDIF
! 14. Exception handling.
! Iterate (real) exp(x) starting at 1.0 until overflow occurs.
! Testing to see if a type FM number is one of the special cases (+-overflow,
! +-underflow or unknown) by direct comparison can be tricky. When X1 is
! +overflow, the statement
! IF (X1 == TO_FM(' +OVERFLOW ')) THEN
! will return false, since the comparison routine cannot be sure that two
! different overflowed results would have been equal if the overflow threshold
! had been higher.
! Function IS_OVERFLOW can be used to directly check whether a number is + or -
! overflow, so that is a safer test.
! The FM warning message is written on unit KW, so in this test it appears on the
! screen and not in the output file.
CALL FM_SET(60)
X1 = TO_FM(1)
FMT = "(///' Sample 14. Exception handling.'//12X," // &
"' Iterate exp(x) starting at 1.0 until overflow occurs.'//" // &
"12X,' An FM warning message will be printed before the last iteration.')"
WRITE (*,FMT)
FMT = "(///' Sample 14. Exception handling.'//" // &
"12X,' Iterate exp(x) starting at 1.0 until overflow occurs.')"
WRITE (KOUT,FMT)
DO J = 1, 10
X1 = EXP(X1)
CALL FM_FORM('ES60.40',X1,ST1)
WRITE (* ,"(/' Iteration',I3,5X,A)") J,TRIM(ST1)
WRITE (KOUT,"(/' Iteration',I3,5X,A)") J,TRIM(ST1)
IF (IS_OVERFLOW(X1)) EXIT
ENDDO
! Check that the last result was +overflow.
IF (IS_OVERFLOW(X1)) THEN
WRITE (* ,"(/' Overflow was correctly detected.')")
WRITE (KOUT,"(/' Overflow was correctly detected.')")
ELSE
NERROR = NERROR + 1
WRITE (* ,"(/' Error in sample case number 14.'/)")
WRITE (* ,"(/' Overflow was not correctly detected.')")
WRITE (KOUT ,"(/' Error in sample case number 14.'/)")
WRITE (KOUT ,"(/' Overflow was not correctly detected.')")
ENDIF
IF (NERROR == 0) THEN
WRITE (* ,"(//A/)") ' All results were ok -- no errors were found.'
WRITE (KOUT,"(//A/)") ' All results were ok -- no errors were found.'
ELSE
WRITE (* ,"(//I3,A/)") NERROR,' error(s) found.'
WRITE (KOUT,"(//I3,A/)") NERROR,' error(s) found.'
ENDIF
STOP
END PROGRAM SAMPLE
SUBROUTINE PLAN_9(F,A,B,N,RESULT)
! Sample subroutine usage for FM.
! Integrate F(X) from A to B using N subintervals, and return the answer in RESULT.
! This does numerical integration using a 9-point rule.
! It is not a very good way to do high-precision integration, but it is a short routine
! and can often get 20 to 30 digits if f(x) is well-behaved and the interval of integration
! is not too big.
! Note that subroutines using FM variables (any of the three types) need a call to
! FM_ENTER_USER_ROUTINE upon entry to the routine and one to FM_EXIT_USER_ROUTINE upon exit.
! To keep from wasting memory, local variables like XJ should have the SAVE attribute.
USE FMZM
IMPLICIT NONE
TYPE (FM) :: A, B, RESULT
TYPE (FM), SAVE :: H, H8, XJ
TYPE (FM), EXTERNAL :: F
INTEGER :: N, J
INTENT (IN) :: N, A, B
INTENT (INOUT) :: RESULT
CALL FM_ENTER_USER_ROUTINE
H = (B - A)/N
H8 = H/8
RESULT = 0
DO J = 1, N
XJ = A + (J-1)*H
RESULT = RESULT + 989*F(XJ) + 5888*F(XJ+ H8) - 928*F(XJ+2*H8) + &
10496*F(XJ+3*H8) - 4540*F(XJ+4*H8) + 10496*F(XJ+5*H8) - &
928*F(XJ+6*H8) + 5888*F(XJ+7*H8) + 989*F(XJ+8*H8)
ENDDO
RESULT = H*RESULT/28350
CALL FM_EXIT_USER_ROUTINE
END SUBROUTINE PLAN_9
FUNCTION F(X)
! Sample function usage for FM.
! The test function for the integration subroutine is 2*exp(-x**2)/sqrt(pi).
! Note that functions returning an FM variable (any of the three types) need a call to
! FM_ENTER_USER_FUNCTION upon entry to the routine and one to FM_EXIT_USER_FUNCTION upon exit,
! where the argument in each case is the function name (F here).
! To keep from wasting memory, local variables like PI should have the SAVE attribute.
USE FMZM
IMPLICIT NONE
TYPE (FM) :: F, X
TYPE (FM), SAVE :: PI
CALL FM_ENTER_USER_FUNCTION(F)
! Compare the usage here with the SQRT(ACOS(TO_FM(-1))) usage in the gamma example
! in the main program. There pi was only used once, so ACOS(TO_FM(-1)) is more like
! what a non-multiple-precision program would do to get pi.
! If we need pi in a function like F that will be called hundreds of times, the acos
! call will be done every time. Here, since the argument is -1, the acos routine will
! recognize it as a special case and return the saved value of pi without needlessly
! making the program slower. But if another formula were used, like pi = 6*asin(1/2),
! it would be better to call FM_PI, since pi would be computed only once and later calls
! just use the saved value of pi.
! Another reason to call FM_PI instead of using a formula is that in case the calling
! program changed the trig function mode to degrees, instead of the default radians,
! then ACOS(TO_FM(-1)) would give 180, not pi.
! For this case the 2/sqrt(pi) could have been factored out of the integral so pi would
! not be needed every time F is called, but it was left in to illustrate similar but
! more complicated situations.
CALL FM_PI(PI)
F = 2*EXP(-X**2)/SQRT(PI)
CALL FM_EXIT_USER_FUNCTION(F)
END FUNCTION F
c4science · Help