Jason
Clamping to +- 10 sounds good to me. I have sent a patch to the mesa-dev
list.
Please port spirv_to_nir in a separate patch if you could, I'm afraid I'm
not really familiar with that part and don't know how to test it.
On Thu, Dec 8, 2016 at 4:54 PM, Jason Ekstrand wrote:
> On Thu, Dec 8
On Thu, Dec 8, 2016 at 3:31 PM, Haixia Shi wrote:
> We're encountering failures in the latest version of dEQP (specifically,
> dEQP-GLES3.functional.shaders.builtin_functions.precision.tanh.highp_*)
> on all Intel devices.
>
> The problem is that when the abs value of input is too large (say 80),
We're encountering failures in the latest version of dEQP (specifically,
dEQP-GLES3.functional.shaders.builtin_functions.precision.tanh.highp_*) on
all Intel devices.
The problem is that when the abs value of input is too large (say 80), the
function should return +/-1, but the actual shader outpu
