[Mesa-dev] fp/int64 on gen11/12?

[Jason Ekstrand]

On Tue, Nov 17, 2020 at 6:10 PM Jordan Justen <jordan.l.justen at intel.com> wrote:
>
> On 2020-11-17 16:03:31, Brian Paul wrote:
> > Another Intel question:  It looks like gen11/gen12 don't have fp/int64
> > enabled in the Vulkan driver. From gen_device_info.c:
> >
> > #define GEN11_FEATURES(_gt, _slices, _subslices, _l3) \
> >     GEN8_FEATURES,                                     \
> >     GEN11_HW_INFO,                                     \
> >     .has_64bit_float = false,                          \
> >     .has_64bit_int = false,
> > ...
> >
> > #define GEN12_FEATURES(_gt, _slices, _l3)                       \
> >     GEN8_FEATURES,                                               \
> >     GEN12_HW_INFO,                                               \
> >     .has_64bit_float = false,                                    \
> >     .has_64bit_int = false,
> >
> > But gen8/9 do support it.  Is this a driver and/or hardware issue?
> >
>
> It matches the hardware.

That ^^

Expounding a bit, we do have emulated int64 and fp64 for OpenGL.  I've
got an MR open (!7329) to provide emulated int64 support in Vulkan.
We have no plans for emulated fp64 in Vulkan.  It's so painful that it
really doesn't play well with the Vulkan philosophy of "expose what's
in hardware".  Also, the lowering code we have in Mesa today really
requires the GLSL compiler.

--Jason