[Mesa-dev] [PATCH 6/6] i965: Use nir_opt_trivial_continues and nir_opt_if

[Jason Ekstrand]

On Dec 21, 2016 9:17 PM, "Timothy Arceri" <timothy.arceri at collabora.com>
wrote:

On Mon, 2016-12-19 at 20:11 -0800, Jason Ekstrand wrote:
> ---
>  src/mesa/drivers/dri/i965/brw_nir.c | 2 ++
>  1 file changed, 2 insertions(+)
>
> diff --git a/src/mesa/drivers/dri/i965/brw_nir.c
> b/src/mesa/drivers/dri/i965/brw_nir.c
> index 0c1fb44..a091861 100644
> --- a/src/mesa/drivers/dri/i965/brw_nir.c
> +++ b/src/mesa/drivers/dri/i965/brw_nir.c
> @@ -429,6 +429,8 @@ nir_optimize(nir_shader *nir, const struct
> brw_compiler *compiler,
>        OPT(nir_opt_algebraic);
>        OPT(nir_opt_constant_folding);
>        OPT(nir_opt_dead_cf);
> +      OPT(nir_opt_trivial_continues);
> +      OPT(nir_opt_if);

I'm seeing regressions in my series for Vulkan (the Vulkan CTS was only
enabled for my branch today).  I wonder if we should be applying your
series before mine? Either way this or a similar patch is:


If you're seeing regressions, there's probably a real bug there.  I assume
you applied the else case fix?

Reviewed-by: Timothy Arceri <timothy.arceri at collabora.com>

>        if (nir->options->max_unroll_iterations != 0) {
>           OPT(nir_opt_loop_unroll, indirect_mask);
>        }
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/mesa-dev/attachments/20161221/7b6b6c3b/attachment.html>