This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Unreviewed math builtins patch (3x)
On Thu, 2003-08-28 at 15:09, Kaveh R. Ghazi wrote:
> The reason I treated sincos different from sin and cos is that sincos
> writes into the memory referenced by it's pointer parameters.
OK, I see now. The FPROUNDING/ERROR attributes are just macros that end
up at one of the base attributes nothow/const_nothrow/pure_nothrow, so
since these set memory, we have to use the nothrow choice directly.
This could cause maintenance problems in the future if the list of base
attributes changes. Maybe we could have something like
ATTR_MATHFN_FPROUNDING_ERRNO_STORE which maps directly to ATTR_NOTHROW,
so that we don't lose track of the fact that this function sets error
and depends on the rounding mode..
--
Jim Wilson, GNU Tools Support, http://www.SpecifixInc.com