This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Inline across -ffast-math boundary
- From: Jan Hubicka <hubicka at ucw dot cz>
- To: Szabolcs Nagy <szabolcs dot nagy at arm dot com>, aldyh at redhat dot com
- Cc: Jan Hubicka <hubicka at ucw dot cz>, gcc-patches at gcc dot gnu dot org, rguenther at suse dot de, nd at arm dot com
- Date: Tue, 3 May 2016 16:01:25 +0200
- Subject: Re: Inline across -ffast-math boundary
- Authentication-results: sourceware.org; auth=none
- References: <20160421114548 dot GA22377 at kam dot mff dot cuni dot cz> <5728AA79 dot 2070409 at arm dot com>
> On 21/04/16 12:45, Jan Hubicka wrote:
> > this patch implements the long promised logic to inline across -ffast-math
> > boundary when eitehr caller or callee has no fp operations in it. This is
> > needed to resolve code quality regression on Firefox with LTO where
> > -O3/-O2/-Ofast flags are combined and we fail to inline a lot of comdats
> > otherwise.
>
> -Ofast turns allow-store-data-races param on.
> are such params handled by lto?
No, allow-store-data-races is ignored by LTO option handling and whatever is
used at final linktime wins. THis is because it is --param parameter rather
than optimization option.
Given the nature of the option, I wonder why it was implemented as --param
thing? It looks like typical optimization option to me and this way we lose
mechanizm to set it with function granuality like we can do other optimization
options.
Honza