Re: 11706 vs ([lno] Canonical iv creation)

[Gabriel Dos Reis <gdr at integrable-solutions dot net>]

Paolo Carlini <pcarlini@suse.de> writes:

| Paolo Carlini wrote:
| 
| > Why not removing it, the version taking integer arguments (and
| > __pow_helper), and adding the above instead?
| 
| ... well, however there is libstdc++/3181, and we don't want to
| regress on that, therefore, probably, we must keep 'pow(double __x,
| int __i)'

pow(double, int) is a standard mandated function.  There are other
versions of pow() that are templates.

| and so on, but I don't see what's wrong with using
| __builtin_pow in the implementation!

You don't want to do that globally.  You want to do that selectively.
And in the past, I've argued for having _builtin_powfi (or something
like that).  I think Roger implemented them or agreed they should be
implemented. I have not checked recently.

| Notice that the very same thing (integer overload using __builtin_* in
| the implementation) is /already/ happening for sin, cos, tan, and so
| on!!!!

Those overloads for integers are in some sense necessary.


-- Gaby