C/C++ Programmer's Guide (G06.27+, H06.03+)

HP C Implementation-Defined Behavior
HP C/C++ Programmer’s Guide for NonStop Systems—429301-010
A-4
G.3.7 Arrays and Pointers
The conversion behavior for IEEE floating-point types is:
The default IEEE rounding mode is to round to the nearest representable value. Where
two values are equally close, the even value (the one with a zero in the least significant
fraction bit) is chosen. For more details about the FP_IEEE_ROUND_GET_ and
FP_IEEE_ROUND_SET_ routines, see the Guardian Procedure Calls Reference
Manual.
Tandem floating-point format and IEEE floating-point format have different ranges and
precision. Therefore, a value that might convert exactly if one format is used might not
convert exactly if the other format is used. For more details about ranges and precision
of each format, see IEEE Floating-Point Arithmetic on page 4-7 and pragma
IEEE_FLOAT on page 13-44.
G.3.7 Arrays and Pointers
size_t is defined in stddef.h to be unsigned long.
Converting a pointer to a signed integer of equal length does not cause a
representation change. Converting an integer to a shorter signed integer causes a
representation change by discarding the high order bits.
ptrdiff_t is defined in stddef.h to be long.
G.3.8 Registers
The register storage class specifier cannot be used with structure or array
declarations. A register variable may be changed to a nonregister type or a
nonregister type changed to register by the optimizer.
G.3.9 Structures, Unions, Enumerations, and Bit Fields
A union is considered as a block of memory the size of the union. If a member of a
union has a value stored in it and is subsequently accessed using a member of a
different type the result is defined as the value of the accessed type at that block of
memory. If the size of the type stored is smaller than the accessed type and the
accessed value is beyond the stored type, the result is undefined. If the type stored is a
structure with holes, and the accessed value overlaps any of the holes, the value is
undefined.
Table A-1. Conversion Behavior for IEEE Floating-Point Format
Conversion IEEE Floating-Point
From fixed-point to floating-point Rounds to the nearest value according to the
current IEEE floating-point rounding mode
From floating-point to fixed-point Truncates to the nearest representable value
From a floating-point number to a
narrower floating-point number
Rounds to the nearest value according to the
current IEEE floating-point rounding mode