Mesa (main): turnip: fix register_index calculations of xfb outputs

[GitLab Mirror]

Module: Mesa
Branch: main
Commit: b71e27ea84d8260dd534067872cb83016bf97e02
URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=b71e27ea84d8260dd534067872cb83016bf97e02

Author: Danylo Piliaiev <dpiliaiev at igalia.com>
Date:   Tue Jun  1 16:27:27 2021 +0300

turnip: fix register_index calculations of xfb outputs

nir_assign_io_var_locations() does not use outputs_written when
assigning driver locations. Use driver_location to avoid incorrectly
guessing what locations it assigned.

Copied from lavapipe 8731a1beb72da8f4165e5596519a56d16299c3f0

Will fix provoking vertex tf tests when VK_EXT_provoking_vertex
would be enabled:
 dEQP-VK.rasterization.provoking_vertex.transform_feedback.*

Signed-off-by: Danylo Piliaiev <dpiliaiev at igalia.com>
Reviewed-by: Samuel Iglesias Gonsálvez <siglesias at igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11111>

---

 src/freedreno/vulkan/tu_shader.c | 38 ++++++++++++++++++--------------------
 1 file changed, 18 insertions(+), 20 deletions(-)

diff --git a/src/freedreno/vulkan/tu_shader.c b/src/freedreno/vulkan/tu_shader.c
index 84389417af4..0991ec35240 100644
--- a/src/freedreno/vulkan/tu_shader.c
+++ b/src/freedreno/vulkan/tu_shader.c
@@ -737,18 +737,15 @@ tu_gather_xfb_info(nir_shader *nir, struct ir3_stream_output_info *info)
    if (!xfb)
       return;
 
-   /* creating a map from VARYING_SLOT_* enums to consecutive index */
-   uint8_t num_outputs = 0;
-   uint64_t outputs_written = 0;
-   for (int i = 0; i < xfb->output_count; i++)
-      outputs_written |= BITFIELD64_BIT(xfb->outputs[i].location);
-
    uint8_t output_map[VARYING_SLOT_TESS_MAX];
    memset(output_map, 0, sizeof(output_map));
 
-   for (unsigned attr = 0; attr < VARYING_SLOT_MAX; attr++) {
-      if (outputs_written & BITFIELD64_BIT(attr))
-         output_map[attr] = num_outputs++;
+   nir_foreach_shader_out_variable(var, nir) {
+      unsigned slots =
+         var->data.compact ? DIV_ROUND_UP(glsl_get_length(var->type), 4)
+                           : glsl_count_attribute_slots(var->type, false);
+      for (unsigned i = 0; i < slots; i++)
+         output_map[var->data.location + i] = var->data.driver_location + i;
    }
 
    assert(xfb->output_count < IR3_MAX_SO_OUTPUTS);
@@ -790,17 +787,6 @@ tu_shader_create(struct tu_device *dev,
    if (!shader)
       return NULL;
 
-   /* Gather information for transform feedback.
-    * This should be called after nir_split_per_member_structs.
-    * Also needs to be called after nir_remove_dead_variables with varyings,
-    * so that we could align stream outputs correctly.
-    */
-   struct ir3_stream_output_info so_info = {};
-   if (nir->info.stage == MESA_SHADER_VERTEX ||
-         nir->info.stage == MESA_SHADER_TESS_EVAL ||
-         nir->info.stage == MESA_SHADER_GEOMETRY)
-      tu_gather_xfb_info(nir, &so_info);
-
    if (nir->info.stage == MESA_SHADER_FRAGMENT) {
       NIR_PASS_V(nir, nir_lower_input_attachments,
                  &(nir_input_attachment_options) {
@@ -845,6 +831,18 @@ tu_shader_create(struct tu_device *dev,
    nir_assign_io_var_locations(nir, nir_var_shader_in, &nir->num_inputs, nir->info.stage);
    nir_assign_io_var_locations(nir, nir_var_shader_out, &nir->num_outputs, nir->info.stage);
 
+  /* Gather information for transform feedback. This should be called after:
+    * - nir_split_per_member_structs.
+    * - nir_remove_dead_variables with varyings, so that we could align
+    *   stream outputs correctly.
+    * - nir_assign_io_var_locations - to have valid driver_location
+    */
+   struct ir3_stream_output_info so_info = {};
+   if (nir->info.stage == MESA_SHADER_VERTEX ||
+         nir->info.stage == MESA_SHADER_TESS_EVAL ||
+         nir->info.stage == MESA_SHADER_GEOMETRY)
+      tu_gather_xfb_info(nir, &so_info);
+
    NIR_PASS_V(nir, tu_lower_io, shader, layout);
 
    NIR_PASS_V(nir, tu_lower_image_size);