On 1/2/19 5:04 am, Emil Velikov wrote:
On Thu, 31 Jan 2019 at 01:01, Timothy Arceri <tarc...@itsqueeze.com> wrote:
Without this we do not end up with a deterministic NIR. NIR must
be deterministic here because we use it to produce a sha for the
radeonsi backends disk cache.
This fixes the shader cache for a bunch of shaders.
I guess we'll need this for all branches since radeonsi started using
the nir lowerings and optimisations?
It's not a big deal as it just means the cache doesn't work as well as
it should rather than breaking anything and since NIR is disabled by
default I didn't bother with the fixes tag.
Also I've sent a V2 of this patch that moves this to the state tracker.
Moving it there also reduces the size of the NIR that gets compressed
and stored to disk.
Fixes: 41d4016e061 ("radeonsi/nir: perform radeonsi-specific lowering
and optimization passes")
---
src/gallium/drivers/radeonsi/si_shader_nir.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/src/gallium/drivers/radeonsi/si_shader_nir.c
b/src/gallium/drivers/radeonsi/si_shader_nir.c
index 7554f5b9f8b..96096abe4ea 100644
--- a/src/gallium/drivers/radeonsi/si_shader_nir.c
+++ b/src/gallium/drivers/radeonsi/si_shader_nir.c
@@ -848,6 +848,9 @@ si_lower_nir(struct si_shader_selector* sel)
} while (progress);
NIR_PASS_V(sel->nir, nir_lower_bool_to_int32);
+
+ nir_variable_mode mask = nir_var_function_temp;
+ nir_remove_dead_variables(sel->nir, mask);
Nit: I'd drop the temporary variable mask.
Regardless of the nit
Reviewed-by: Emil Velikov <emil.veli...@collabora.com>
-Emil
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/mesa-dev
