 Test of EXP(Z) vs EXP(Z-DEL)EXP(DEL), DEL = (1/16,1/16)


   2000 Random arguments were tested from the box
      (     0.0000E+00,     0.1000E+01)
      (     0.6250E-01,     0.1000E+01)


 First, report the vector error


 Test function was larger     0 times,
                   agreed   519 times, and
              was smaller  1481 times.


 There are  53 base   2 significant digits in a floating-point number.


 The maximum relative error of     0.3247E-15 =    2 ** -51.45
    occurred for Z =     0.313000E+00     0.625000E-01
    with a value of      0.136485E+01     0.854145E-01
 The estimated loss of base   2 significant digits is   1.55


 The root mean square relative error was     0.1077E-15 =    2 ** -53.04
 The estimated loss of base   2 significant digits is   0.00


 Then, report the error in the real component.


 Test function was larger   569 times,
                   agreed  1206 times, and
              was smaller   225 times.


 There are  53 base   2 significant digits in a floating-point number.


 The maximum relative error of     0.3254E-15 =    2 ** -51.45
    occurred for Z =     0.313000E+00     0.625000E-01
    with a value of      0.136485E+01     0.854145E-01
 The estimated loss of base   2 significant digits is   1.55


 The root mean square relative error was     0.1076E-15 =    2 ** -53.05
 The estimated loss of base   2 significant digits is   0.00


 Then, report the error in the imaginary component.


 Test function was larger    63 times,
                   agreed   965 times, and
              was smaller   972 times.


 There are  53 base   2 significant digits in a floating-point number.


 The maximum relative error of     0.4148E-15 =    2 ** -51.10
    occurred for Z =     0.762000E+00     0.625000E-01
    with a value of      0.213837E+01     0.133823E+00
 The estimated loss of base   2 significant digits is   1.90


 The root mean square relative error was     0.1294E-15 =    2 ** -52.78
 The estimated loss of base   2 significant digits is   0.22


 Test of EXP(Z) vs EXP(Z-DEL)EXP(DEL), DEL = (1/16,1/16)


   2000 Random arguments were tested from the box
      (     0.1625E+01,     0.3000E+01)
      (     0.1625E+01,     0.3000E+01)


 First, report the vector error


 Test function was larger     0 times,
                   agreed   186 times, and
              was smaller  1814 times.


 There are  53 base   2 significant digits in a floating-point number.


 The maximum relative error of     0.4034E-15 =    2 ** -51.14
    occurred for Z =     0.286869E+01     0.162500E+01
    with a value of     -0.954270E+00     0.175880E+02
 The estimated loss of base   2 significant digits is   1.86


 The root mean square relative error was     0.1055E-15 =    2 ** -53.07
 The estimated loss of base   2 significant digits is   0.00


 Then, report the error in the real component.


 Test function was larger    33 times,
                   agreed   310 times, and
              was smaller  1657 times.


 There are  53 base   2 significant digits in a floating-point number.


 The maximum relative error of     0.6672E-15 =    2 ** -50.41
    occurred for Z =     0.250844E+01     0.162500E+01
    with a value of     -0.665605E+00     0.122677E+02
 The estimated loss of base   2 significant digits is   2.59


 The root mean square relative error was     0.2425E-15 =    2 ** -51.87
 The estimated loss of base   2 significant digits is   1.13


 Then, report the error in the imaginary component.


 Test function was larger   301 times,
                   agreed  1205 times, and
              was smaller   494 times.


 There are  53 base   2 significant digits in a floating-point number.


 The maximum relative error of     0.4040E-15 =    2 ** -51.14
    occurred for Z =     0.286869E+01     0.162500E+01
    with a value of     -0.954270E+00     0.175880E+02
 The estimated loss of base   2 significant digits is   1.86


 The root mean square relative error was     0.1048E-15 =    2 ** -53.08
 The estimated loss of base   2 significant digits is   0.00


 Test of EXP(Z) vs EXP(Z-DEL)EXP(DEL), DEL = (1/16,1/16)


   2000 Random arguments were tested from the box
      (     0.1600E+02,     0.1700E+02)
      (     0.1600E+02,     0.1700E+02)


 First, report the vector error


 Test function was larger     0 times,
                   agreed   604 times, and
              was smaller  1396 times.


 There are  53 base   2 significant digits in a floating-point number.


 The maximum relative error of     0.3885E-15 =    2 ** -51.19
    occurred for Z =     0.167995E+02     0.160000E+02
    with a value of     -0.189296E+08    -0.569085E+07
 The estimated loss of base   2 significant digits is   1.81


 The root mean square relative error was     0.1088E-15 =    2 ** -53.03
 The estimated loss of base   2 significant digits is   0.00


 Then, report the error in the real component.


 Test function was larger   321 times,
                   agreed  1236 times, and
              was smaller   443 times.


 There are  53 base   2 significant digits in a floating-point number.


 The maximum relative error of     0.3947E-15 =    2 ** -51.17
    occurred for Z =     0.161035E+02     0.160000E+02
    with a value of     -0.943783E+07    -0.283732E+07
 The estimated loss of base   2 significant digits is   1.83


 The root mean square relative error was     0.1067E-15 =    2 ** -53.06
 The estimated loss of base   2 significant digits is   0.00


 Then, report the error in the imaginary component.


 Test function was larger    91 times,
                   agreed   934 times, and
              was smaller   975 times.


 There are  53 base   2 significant digits in a floating-point number.


 The maximum relative error of     0.4403E-15 =    2 ** -51.01
    occurred for Z =     0.169085E+02     0.160000E+02
    with a value of     -0.211096E+08    -0.634622E+07
 The estimated loss of base   2 significant digits is   1.99


 The root mean square relative error was     0.1303E-15 =    2 ** -52.77
 The estimated loss of base   2 significant digits is   0.23


 Special tests


 The identity  EXP(Z)*EXP(-Z) = 1.0  will be tested.

        Z                        F(Z)*F(-Z) - 1

  0.0000000E+00  0.0000000E+00  0.0000000E+00  0.0000000E+00

  0.0000000E+00  0.0000000E+00  0.0000000E+00  0.0000000E+00

  0.0000000E+00  0.0000000E+00  0.0000000E+00  0.0000000E+00

  0.0000000E+00  0.0000000E+00  0.0000000E+00  0.0000000E+00

  0.0000000E+00  0.0000000E+00  0.0000000E+00  0.0000000E+00



 Test of special arguments


 EXP(0.0,0.0) - 1.0E0 = (  0.0000000E+00  0.0000000E+00)



 Test cos(y), sin(y), and exp(x)


 [REAL(EXP(0,10)) - COS(10)]/COS(10) =   0.0000000E+00

 [IMAG(EXP(0,10)) - SIN(10)]/SIN(10) =   0.0000000E+00

 [REAL(EXP(10,0)) - EXP(10)]/EXP(10) =   0.0000000E+00



 Tests near extreme arguments


 The real part of the first result below may underflow.

 EXP(-0.708000E+03 0.100000E+01) = ( 0.178708-307 0.278321-307)

 EXP( 0.709000E+03 0.100000E+01) = ( 0.444042+308 0.691555+308)


 If EXP( 0.354500E+03 0.100000E+02) = (-0.760664+154-0.493185+154) is not about
 EXP( 0.177250E+03 0.500000E+01)**2 = (-0.760664+154-0.493185+154), there is an
 argument reduction error


 Test of error returns


 EXP will be called with the argument (    -0.6704+154     0.1000E+01)

 The real component of the argument is so negative
 that the real exponential should underflow to zero.
 Thus, the result should be (0,0).

 EXP returned the value (     0.0000E+00     0.0000E+00)



 EXP will be called with the argument (     0.1000E+02     0.1341+155)

 The imaginary component of the argument is too large
 for the sin and cos computations to make sense.

 EXP returned the value (     0.1823E+05    -0.1236E+05)



 EXP will be called with the argument (     0.6704+154    -0.1000E+01)

 The real component of the argument is so large
 that the real exponential should overflow.  The
 result should be (inf,inf) with an error return.

 EXP returned the value (+++++++++++++++---------------)


 This concludes the tests
