gcc: Floating point implementation
1
1 4.6 Floating Point
1 ==================
1
1 * 'The accuracy of the floating-point operations and of the library
1 functions in '<math.h>' and '<complex.h>' that return
1 floating-point results (C90, C99 and C11 5.2.4.2.2).'
1
1 The accuracy is unknown.
1
1 * 'The rounding behaviors characterized by non-standard values of
1 'FLT_ROUNDS' (C90, C99 and C11 5.2.4.2.2).'
1
1 GCC does not use such values.
1
1 * 'The evaluation methods characterized by non-standard negative
1 values of 'FLT_EVAL_METHOD' (C99 and C11 5.2.4.2.2).'
1
1 GCC does not use such values.
1
1 * 'The direction of rounding when an integer is converted to a
1 floating-point number that cannot exactly represent the original
1 value (C90 6.2.1.3, C99 and C11 6.3.1.4).'
1
1 C99 Annex F is followed.
1
1 * 'The direction of rounding when a floating-point number is
1 converted to a narrower floating-point number (C90 6.2.1.4, C99 and
1 C11 6.3.1.5).'
1
1 C99 Annex F is followed.
1
1 * 'How the nearest representable value or the larger or smaller
1 representable value immediately adjacent to the nearest
1 representable value is chosen for certain floating constants (C90
1 6.1.3.1, C99 and C11 6.4.4.2).'
1
1 C99 Annex F is followed.
1
1 * 'Whether and how floating expressions are contracted when not
1 disallowed by the 'FP_CONTRACT' pragma (C99 and C11 6.5).'
1
1 Expressions are currently only contracted if '-ffp-contract=fast',
1 '-funsafe-math-optimizations' or '-ffast-math' are used. This is
1 subject to change.
1
1 * 'The default state for the 'FENV_ACCESS' pragma (C99 and C11
1 7.6.1).'
1
1 This pragma is not implemented, but the default is to "off" unless
1 '-frounding-math' is used in which case it is "on".
1
1 * 'Additional floating-point exceptions, rounding modes,
1 environments, and classifications, and their macro names (C99 and
1 C11 7.6, C99 and C11 7.12).'
1
1 This is dependent on the implementation of the C library, and is
1 not defined by GCC itself.
1
1 * 'The default state for the 'FP_CONTRACT' pragma (C99 and C11
1 7.12.2).'
1
1 This pragma is not implemented. Expressions are currently only
1 contracted if '-ffp-contract=fast', '-funsafe-math-optimizations'
1 or '-ffast-math' are used. This is subject to change.
1
1 * 'Whether the "inexact" floating-point exception can be raised when
1 the rounded result actually does equal the mathematical result in
1 an IEC 60559 conformant implementation (C99 F.9).'
1
1 This is dependent on the implementation of the C library, and is
1 not defined by GCC itself.
1
1 * 'Whether the "underflow" (and "inexact") floating-point exception
1 can be raised when a result is tiny but not inexact in an IEC 60559
1 conformant implementation (C99 F.9).'
1
1 This is dependent on the implementation of the C library, and is
1 not defined by GCC itself.
1