On Tue, 13 Oct 2015, Michael Meissner wrote:

> I believe every non-NaN value that IBM extended double supports is
> representable in IEEE 754R 128-bit floating point, since IEEE has 112 bits of
> mantissa plus the hidden bit, while IBM extended double has 106 bits (52 bits
> of mantissa for each part plus 2 hidden bits). Even with all of the extra bits

No, because IBM long double can represent values with discontiguous 
mantissa bits (these are the values that 
libgcc/config/rs6000/ibm-ldouble-format describes as denormal but not 
subnormal).

> that you can hand craft into silent/signalling NaNs, you should be able
> represent those values in IEEE 128-bit floating point as similar NaNs.

The low part of a NaN in IBM long double is documented as don't-care (so 
all IBM long double NaNs are correctly converted to binary128 simply by 
converting the high part - as is also the case with zeroes).

-- 
Joseph S. Myers
jos...@codesourcery.com

Reply via email to