----- Original Message -----
> On Sep 12, 2013, at 2:46 PM, Hal Finkel <[email protected]> wrote:
> 
> > ----- Original Message -----
> >> On Sep 12, 2013, at 2:14 PM, Hal Finkel <[email protected]> wrote:
> >> 
> >>> I've been thinking about constructing a patch to implement
> >>> __builtin_vectorconvert(type, value), do you think that is a
> >>> reasonable approach?
> >> 
> >> It would be nice if what you come up with is capable of being
> >> extended to include rounding and saturation, like the OpenCL
> >> convert_* ops.  Obviously that blows up the space enormously, so I
> >> wouldn’t expect anyone to do them all at once.
> > 
> > Do you have a recommended design? I could do something like this:
> >  __builtin_vectorconvert(type, value, rnd = 0, sat = 0)
> > where the last two parameters are integers with default values
> > specifying the rounding mode and a saturation parameter.
> 
> I am very much the wrong person to ask, but that seems reasonable to
> me.

Okay ;) -- Well, I'll leave off these parameters for the initial 
implementation; we can always add them on later.

Thanks again,
Hal

> 
> – Steve

-- 
Hal Finkel
Assistant Computational Scientist
Leadership Computing Facility
Argonne National Laboratory

_______________________________________________
cfe-commits mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits

Reply via email to