[llvm-commits] Patch: support half intrinsics

[Eric Christopher]

On Thu, Jan 10, 2013 at 1:43 PM, Evan Cheng <evan.cheng at apple.com> wrote:

>
> On Jan 10, 2013, at 1:29 PM, Michael Ilseman <milseman at apple.com> wrote:
>
> >
> > On Jan 10, 2013, at 1:01 PM, Chris Lattner <clattner at apple.com> wrote:
> >
> >>
> >> On Jan 10, 2013, at 10:23 AM, Michael Ilseman <milseman at apple.com>
> wrote:
> >>
> >>> Attached is a patch to allow support for half intrinsics.
> >>
> >> Ok.
> >>
> >>> This puts F16 into the fast encoding path, and pushes MMX down into
> the slow encoding path.
> >>
> >> How much bloat does this cause to the X86 backend?
> >>
> >
> > I measured the size of the libraries present in my build/Release/lib
> directory, and there was no difference in size (512-byte granularity). This
> is building on X86 for X86. Were you referring to something else?
>
> The encoding change would impact size of LLVM bitcode, right? But the
> change only impacts MMX intrinsics so I can't see it having much impact. Or
> perhaps I am misunderstanding.
>

That was my impression too and we pretty actively discourage mmx so I
wasn't too concerned either.

-eric
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20130110/c1155c83/attachment.html>