nir/gather_info: fix per-vertex handling in try_mask_partial_io

pipeline-db (Navi, ACO):
Totals from affected shaders:
SGPRS: 6432 -> 6432 (0.00 %)
VGPRS: 11924 -> 11924 (0.00 %)
Spilled SGPRs: 0 -> 0 (0.00 %)
Spilled VGPRs: 0 -> 0 (0.00 %)
Scratch size: 1596 -> 1596 (0.00 %) dwords per thread
Code Size: 575524 -> 518620 (-9.89 %) bytes
LDS: 12187 -> 12187 (0.00 %) blocks
Max Waves: 2695 -> 2695 (0.00 %)

Helps a few hundred Dark Souls 3 shaders.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
CC: <mesa-stable@lists.freedesktop.org>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4190>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4190>

diff --git a/src/compiler/nir/nir_gather_info.c b/src/compiler/nir/nir_gather_info.c
index 124f355f733028dfeb3f0b8a76b35ead1fe815a4..455b4e5f4925bfb584bdaee519f4a019702f6640 100644 (file)
--- a/src/compiler/nir/nir_gather_info.c
+++ b/src/compiler/nir/nir_gather_info.c
@@ -102,12 +102,15 @@ mark_whole_variable(nir_shader *shader, nir_variable *var, bool is_output_read)
 }
 
 static unsigned
-get_io_offset(nir_deref_instr *deref, bool is_vertex_input)
+get_io_offset(nir_deref_instr *deref, bool is_vertex_input, bool per_vertex)
 {
    unsigned offset = 0;
 
    for (nir_deref_instr *d = deref; d; d = nir_deref_instr_parent(d)) {
       if (d->deref_type == nir_deref_type_array) {
+         if (per_vertex && nir_deref_instr_parent(d)->deref_type == nir_deref_type_var)
+            break;
+
          if (!nir_src_is_const(d->arr.index))
             return -1;
 
@@ -132,8 +135,9 @@ try_mask_partial_io(nir_shader *shader, nir_variable *var,
                     nir_deref_instr *deref, bool is_output_read)
 {
    const struct glsl_type *type = var->type;
+   bool per_vertex = nir_is_per_vertex_io(var, shader->info.stage);
 
-   if (nir_is_per_vertex_io(var, shader->info.stage)) {
+   if (per_vertex) {
       assert(glsl_type_is_array(type));
       type = glsl_get_array_element(type);
    }
@@ -157,7 +161,7 @@ try_mask_partial_io(nir_shader *shader, nir_variable *var,
       return false;
    }
 
-   unsigned offset = get_io_offset(deref, false);
+   unsigned offset = get_io_offset(deref, false, per_vertex);
    if (offset == -1)
       return false;