On 2 February 2018 at 03:40, Roland Scheidegger wrote:
> I suppose this is a bit inefficient if the hw actually supports some
> double ops?
> I mean for uint->double you could just convert the low 8 bits and high
> 24 bits (or different split) separately to float, convert both to double
> and add
Quoting Elie Tournier (2018-02-01 12:13:53)
> On Thu, Feb 01, 2018 at 10:00:12AM -0800, Dylan Baker wrote:
> > There are changes from Makefile.sources that need be ported to corresponding
> > meson.build files as well.
> Oh yes, nice spot. I will do it on my local fp64 branch too.
> >
> > It looks
On Thu, Feb 01, 2018 at 10:00:12AM -0800, Dylan Baker wrote:
> There are changes from Makefile.sources that need be ported to corresponding
> meson.build files as well.
Oh yes, nice spot. I will do it on my local fp64 branch too.
>
> It looks like builtin_float64.h is generated from float64.glsl,
There are changes from Makefile.sources that need be ported to corresponding
meson.build files as well.
It looks like builtin_float64.h is generated from float64.glsl, is that right?
I'm not very happy about checking in a generated file for all of the reasons
that we don't check generated files in
I suppose this is a bit inefficient if the hw actually supports some
double ops?
I mean for uint->double you could just convert the low 8 bits and high
24 bits (or different split) separately to float, convert both to double
and add them. (I think should work for int->double exactly the same
despit
From: Elie Tournier
This adds the first two fp64 builtins.
They are for int/uint to fp64 conversions.
cayman hw can't do it, and this avoids me having to lower it in
the backend, and also starts to provide some of the soft fp64
infrastructure.
[airlied: extracted from Elie's tree]
Signed-off-by