-qxlf2003

Category

Language element control

Purpose

Provides the ability to use language features specific to the Fortran 2003 standard when compiling with compiler invocations that conform to earlier Fortran standards, as well as the ability to disable these features when compiling with compiler invocations that conform to the Fortran 2003 standard.

Defaults

The default suboption depends on the invocation command that you specify.

The f2003, xlf2003, or xlf2003_r command has the following defaults:

autorealloc:bozlitargs:nodynamicacval:nooldnaninf:polymorphic:signdzerointr:

stopexcept:volatile

All other invocation commands have the following defaults:

noautorealloc:nobozlitargs:nodynamicacval:oldnaninf:nopolymorphic:

nosigndzerointr:nostopexcept:novolatile

Parameters

autorealloc | noautorealloc

Controls whether the compiler automatically reallocates the left-hand-side (LHS) with the shape of the right-hand-side (RHS) when assigning into an allocatable variable. If the LHS variable was not allocated before the assignment, it is allocated automatically. The default is autorealloc for the f2003, xlf2003, and xlf2003_r commands, and noautorealloc for all other commands. This suboption has no effect on reallocation when the values of length type parameters in the LHS and RHS differ.

bozlitargs | nobozlitargs

The bozlitargs suboption ensures that the passing of boz-literal constants as arguments to the INT, REAL, CMPLX, or DBLE intrinsic function conforms to the Fortran 2003 standard. The default is bozlitargs for the f2003, xlf2003, and xlf2003_r commands. The -qlanglvl=2003pure or -qlanglvl=2003std option must be specified, as well. If -qport=typlssarg and -qxlf2003=bozlitargs are specified, passing boz-literal constants to the CMPLX intrinsic will yield non-standard results.

dynamicacval | nodynamicacval

When dynamicacval is in effect, the dynamic types of array constructor values are used to determine the type of the array constructors and you can use unlimited polymorphic entities in array constructors. When nodynamicacval is in effect, the declared types of array constructor values are used to determine the type of the array constructors and you cannot use unlimited polymorphic entities in array constructors.

Note: To make the -qxlf2003=dynamicacval option effective, you must also specify -qxlf2003=polymorphic.

oldnaninf | nooldnaninf

The oldnaninf suboption controls the formatting of the output of IEEE NaN and infinity exceptional values. This suboption has no effect on input. When oldnaninf is in effect, the compiler uses the XL Fortran V10.1 (and earlier) behavior for output. That is, INF for infinity, NaNQ for a quiet NaN, and NANS for a signaling NaN.

When nooldnaninf is in effect, the compiler output for IEEE exceptional values is compliant with the Fortran 2003 standard. That is, Inf for infinity, NaN(Q) for a quiet NaN, and NaN(S) for a signaling NaN.

polymorphic | nopolymorphic

When polymorphic is in effect, the compiler allows polymorphic items in Fortran source files. You can specify the CLASS type specifier, the SELECT TYPE construct, and use polymorphic items in other Fortran statements. The use of the polymorphic argument also causes the compiler to produce runtime type information for each derived type definition.

When nopolymorphic is in effect, polymorphic items cannot be specified in Fortran source files and no runtime type information is generated.

signdzerointr | nosigndzerointr

When signdzerointr is in effect, the passing of signed zeros to the SQRT, LOG, and ATAN2 intrinsic functions returns results consistent with the Fortran 2003 standard. The -qxlf90=signedzero option must be in effect, as well. For the xlf, xlf_r, xlf_r7, f77, fort77, xlf90, xlf90_r, xlf90_r7, and f90 invocations, specify both options to have the Fortran 2003 behavior.

The following example shows the use of this suboption:

! If the Test program is compiled with -qxlf2003=signdzerointr
! and -qxlf90=signedzero, then Fortran 2003 behavior is seen.
! Otherwise, this program will demonstrate Fortran 95 behavior.

Program Test

real a, b
complex j, l
a = -0.0
j = sqrt(cmplx(-1.0,a))
b = atan2(a,-1.0)
l = log(cmplx(-1.0,a))
print *, 'j=', j
print *, 'b=', b
print *, 'l=', l
end

! Fortran 95 output:

 j= (-0.0000000000E+00,1.000000000)
 b= 3.141592741
 l= (0.0000000000E+00,3.141592741)

! Fortran 2003 output:

 j= (0.0000000000E+00,-1.000000000)
 b= -3.141592741
 l= (0.0000000000E+00,-3.141592741)

stopexcept | nostopexcept

When stopexcept is in effect, informational messages are displayed when IEEE floating-point exceptions are signaled by a STOP statement. Messages have the format:

STOP [stop-code]
(OVERFLOW, DIV-BY-ZERO, INVALID, UNDERFLOW, INEXACT)

where stop-code corresponds to the optional digit string or character constant specified in the STOP statement. OVERFLOW, DIV-BY-ZERO, INVALID, UNDERFLOW and INEXACT appear only if the corresponding flag is set.

The following example shows corresponding messages generated:

real :: rl1, rl2, rl3, rl4
logical :: l

rl1 = 1.3
rl2 = 0.0

rl3 = rl1 / rl2 ! divide by zero

rl4 = rl3 ! to make sure rl3 is actually used

rl4 = log(-rl1) ! invalid input for log

stop "The End"

end

Output:

STOP The End
(DIV-BY-ZERO, INVALID)

When nostopexcept is in effect, informational messages are suppressed.

volatile | novolatile

When volatile is in effect, a non-VOLATILE entity that is use- or host-associated can be specified as VOLATILE in inner or local scope.

Usage

If the application uses F2003 polymorphism, you must compile every unit with polymorphic specified. If the application does not use polymorphism, specify the nopolymorphic suboption; doing so may save compilation time and potentially improve runtime performance.

Related information

See the following information in XL Fortran Language Reference:

• Polymorphic entities
• Array constructors