Re: Fix fnspecs for math builtins

2020-10-27 Thread Richard Biener via Gcc-patches
On Tue, Oct 27, 2020 at 9:06 AM Jan Hubicka wrote: > > > On Mon, 26 Oct 2020, Jan Hubicka wrote: > > > > > Hi, > > > this patch makes us to use ".C" and ".P" fnspecs where > > > applicable. I also noticed that gamma and variants are > > > declared as storing to memory while they are not (gamma_r

Re: Fix fnspecs for math builtins

2020-10-27 Thread Jan Hubicka
> On Mon, 26 Oct 2020, Jan Hubicka wrote: > > > Hi, > > this patch makes us to use ".C" and ".P" fnspecs where > > applicable. I also noticed that gamma and variants are > > declared as storing to memory while they are not (gamma_r does) > > I think the point is that they store to the global

Re: Fix fnspecs for math builtins

2020-10-26 Thread Joseph Myers
On Mon, 26 Oct 2020, Jan Hubicka wrote: > Hi, > this patch makes us to use ".C" and ".P" fnspecs where > applicable. I also noticed that gamma and variants are > declared as storing to memory while they are not (gamma_r does) I think the point is that they store to the global signgam. --

Fix fnspecs for math builtins

2020-10-26 Thread Jan Hubicka
Hi, this patch makes us to use ".C" and ".P" fnspecs where applicable. I also noticed that gamma and variants are declared as storing to memory while they are not (gamma_r does) Bootstrapped/regtested x86_64-linux, OK? gcc/ChangeLog: 2020-10-26 Jan Hubicka * builtin-attrs.def