Re: [PATCH] Fix truncf for sNaN input
Joseph Myers <[email protected]>
| Newsgroups | gmane.comp.lib.newlib |
|---|---|
| Message-ID | <[email protected]> |
On Tue, 17 Mar 2020, Fabian Schriever wrote: > On implementations that support the IEC60559: 1989 standard floating point, > functions with signaling NaN argument(s) shall be treated as if the function > were called with an argument that is a required domain error and shall return > a quiet NaN result, except where stated otherwise. This fails to allow for functions such as fabs where the corresponding IEEE operations pass through a signaling NaN rather than returning a quiet NaN with "invalid" raised. Furthermore, TS 18661-1 makes it implementation-defined whether there is a domain error. And in practice it's more convenient for implementations not to treat it as a domain error (not to set errno) as that means they can just e.g. add a NaN argument to itself rather than rather than explicitly testing for a signaling NaN. -- Joseph S. Myers [email protected]