Lines Matching refs:to
219 # Two values representable in the same format with sum very close to
221 # that there is no intermediate rounding to 64 bits when adding two
4849 # to half way between two representable values in a narrower format,
4854 # ratios of the form (1 + 2^-a) to the width of the wider format, but
4870 # Similar, for double rounding to 64-bit of a division of 53-bit values.
5474 # Bug 6801: errno setting may be missing. This applies to both plain
6031 # Values +/- 10ulp from overflow threshold. (Values very close to
6035 # or its absence to be unambiguous under glibc's accuracy standards).
6037 # the binary128 inputs for intel96 and m68k96, as too close to the
6303 # to produce integers +/- 1ulp for other formats).
6679 # close to half way between two representable values in a narrower
6698 # Similar, for double rounding to 64 bit of a product of 53-bit values.
7546 # close to half way between two representable values. In particular,
7547 # verify that there is no intermediate rounding to 64 bits when
7867 # IEEE semantics mean overflow very close to the threshold depends on
7869 # does not try to achieve this.