OOMMF Variable Types and Macros

The following typedefs are defined in the oommf/pkg/oc/platform/ocport.h header file; this file is created by the pimake build process (see oommf/pkg/oc/procs.tcl), and contains platform and machine specific information.

• OC_BOOL Boolean type, unspecified width.
• OC_BYTE Unsigned integer type exactly one byte wide.
• OC_CHAR Character type, may be signed or unsigned.
• OC_UCHAR Unsigned character type.
• OC_SCHAR Signed character type. If signed char is not supported by a given compiler, then this falls back to a plain char, so use with caution.
• OC_INT2, OC_INT4 Signed integer with width of exactly 2, respectively 4, bytes.
• OC_INT2m, OC_INT4m Signed integer with width of at least 2, respectively 4, bytes. A type wider than the minimum may be specified if the wider type is handled faster by the particular machine.
• OC_UINT2, OC_UINT4, OC_UINT2m, OC_UINT4m Unsigned integer versions of the preceding.
• OC_REAL4, OC_REAL8 Four byte, respectively eight byte, floating point variable. Typically corresponds to C++ ``float'' and ``double'' types.
• OC_REAL4m, OC_REAL8m Floating point variable with width of at least 4, respectively 8, bytes. A type wider than the minimum may be specified if the wider type is handled faster by the particular machine.
• OC_REALWIDE Widest type natively supported by the underlying hardware. This is usually the C++ ``long double'' type, but may be overridden by the
  program_compiler_c++_typedef_realwide
  option in the oommf/config/platform/platform.tcl file.

The oommf/pkg/oc/platform/ocport.h header file also defines the following macros for use with the floating point variable types:

• OC_REAL8m_IS_DOUBLE True if OC_REAL8m type corresponds to the C++ ``double'' type.
• OC_REAL8m_IS_REAL8 True if OC_REAL8m and OC_REAL8 refer to the same type.
• OC_REAL4_EPSILON The smallest value that can be added to a OC_REAL4 value of ``1.0'' and yield a value different from ``1.0''. For IEEE 754 compatible floating point, this should be 1.1920929e-007.
• OC_SQRT_REAL4_EPSILON Square root of the preceding.
• OC_REAL8_EPSILON The smallest value that can be added to a OC_REAL8 value of ``1.0'' and yield a value different from ``1.0''. For IEEE 754 compatible floating point, this should be 2.2204460492503131e-016.
• OC_SQRT_REAL8_EPSILON, OC_CUBE_ROOT_REAL8_EPSILON Square and cube roots of the preceding.
• OC_FP_REGISTER_EXTRA_PRECISION True if intermediate floating point operations use a wider precision than the floating point variable type; notably, this occurs with some compilers on x86 hardware.

Note that all of the above macros have a leading ``OC_'' prefix. The prefix is intended to protect against possible name collisions with system header files. Versions of some of these macros are also defined without the prefix; these definitions represent backward support for existing OOMMF extensions. All new code should use the versions with the ``OC_'' prefix, and old code should be updated where possible. The complete list of deprecated macros is:

BOOL, UINT2m, INT4m, UINT4m, REAL4, REAL4m, REAL8, REAL8m, REALWIDE, REAL4_EPSILON, REAL8_EPSILON, SQRT_REAL8_EPSILON, CUBE_ROOT_REAL8_EPSILON, FP_REGISTER_EXTRA_PRECISION

Macros for system identification:

• OC_SYSTEM_TYPE One of OC_UNIX or OC_WINDOWS.
• OC_SYSTEM_SUBTYPE For unix systems, this is either OC_VANILLA (general unix) or OC_DARWIN (Mac OS X). For Windows systems, this is generally OC_WINNT, unless one is running out of a Cygwin shell, in which case the value is OC_CYGWIN.

Additional macros and typedefs:

• OC_POINTERWIDTH Width of pointer type, in bytes.
• OC_INDEX Typedef for signed array index type; typically the width of this (integer) type matches the width of the pointer type, but is in any event at least four bytes wide and not narrower than the pointer type.
• OC_UINDEX Typedef for unsigned version of OC_INDEX. It is intended for special-purpose use only. In general, use OC_INDEX where possible.
• OC_INDEX_WIDTH Width of OC_INDEX type.
• OC_BYTEORDER Either ``4321'' for little endian machines, or ``1234'' for big endian.
• OC_THROW(x) Throws a C++ exception with value ``x''.
• OOMMF_THREADS True threaded (multi-processing) builds.
• OC_USE_NUMA If true, then NUMA (non-uniform memory access) libraries are available.

OOMMF Documentation Team
September 27, 2019