HDF5 automatically performs necessary conversions between different types of floats. For example if you are reading a file on a UNIX computer which represents floats with big-endian notation and the floats in the file are represented using little-endian notation, HDF5 will convert to big-endian for you. This conversion takes extra time when reading the data. Since data is likely going to be read more times than it was written (you only write once), it is generally preferable to write the data in the format that you will be post-processing the data on. The XMDF library provides a function that you can use to specify how you want the floats represented in the file (little-endian is the default). Call the function and pass XFALSE to write floating point numbers in big-endian notation.

C/C++: int xfWriteFloatsAsLittleEndian(xbool LittleEndian); ???

FORTRAN: SUBROUTINE XF_WRITE_FLOATS_AS_LITTLE_ENDIAN(LittleEndian, Error)

CHARACTER(LEN=*), INTENT(IN):: Name

LOGICAL, INTENT(IN) :: LittleEndian

INTEGER, INTENT(OUT) :: Error