v3d: add new flag dirty TMU cache at v3d_compiler

[mesa.git] / src / broadcom / compiler / vir.c

diff --git a/src/broadcom/compiler/vir.c b/src/broadcom/compiler/vir.c
index 78362a2949ce8382bc9c3418d748a03c00aa4b57..dc5d3fe3bed6d708124adddc70c1dccec20bde06 100644 (file)
--- a/src/broadcom/compiler/vir.c
+++ b/src/broadcom/compiler/vir.c
@@ -710,6 +710,7 @@ v3d_set_prog_data(struct v3d_compile *c,
         prog_data->threads = c->threads;
         prog_data->single_seg = !c->last_thrsw;
         prog_data->spill_size = c->spill_size;
+        prog_data->tmu_dirty_rcl = c->tmu_dirty_rcl;
 
         v3d_set_prog_data_uniforms(c, prog_data);
 
@@ -812,8 +813,7 @@ v3d_nir_lower_fs_early(struct v3d_compile *c)
          * enabling early_fragment_tests even if the user didn't.
          */
         if (!(c->s->info.num_images ||
-              c->s->info.num_ssbos ||
-              c->s->info.num_abos)) {
+              c->s->info.num_ssbos)) {
                 c->s->info.fs.early_fragment_tests = true;
         }
 }
@@ -826,7 +826,7 @@ v3d_nir_lower_vs_late(struct v3d_compile *c)
 
         if (c->key->ucp_enables) {
                 NIR_PASS_V(c->s, nir_lower_clip_vs, c->key->ucp_enables,
-                           false);
+                           false, false, NULL);
                 NIR_PASS_V(c->s, nir_lower_io_to_scalar,
                            nir_var_shader_out);
         }
@@ -844,14 +844,9 @@ v3d_nir_lower_fs_late(struct v3d_compile *c)
         if (c->fs_key->clamp_color)
                 NIR_PASS_V(c->s, nir_lower_clamp_color_outputs);
 
-        if (c->fs_key->alpha_test) {
-                NIR_PASS_V(c->s, nir_lower_alpha_test,
-                           c->fs_key->alpha_test_func,
-                           false);
-        }
-
         if (c->key->ucp_enables)
-                NIR_PASS_V(c->s, nir_lower_clip_fs, c->key->ucp_enables);
+                NIR_PASS_V(c->s, nir_lower_clip_fs, c->key->ucp_enables,
+                           false);
 
         /* Note: FS input scalarizing must happen after
          * nir_lower_two_sided_color, which only handles a vec4 at a time.
@@ -940,6 +935,22 @@ uint64_t *v3d_compile(const struct v3d_compiler *compiler,
         NIR_PASS_V(c->s, nir_lower_idiv);
 
         v3d_optimize_nir(c->s);
+
+        /* Do late algebraic optimization to turn add(a, neg(b)) back into
+         * subs, then the mandatory cleanup after algebraic.  Note that it may
+         * produce fnegs, and if so then we need to keep running to squash
+         * fneg(fneg(a)).
+         */
+        bool more_late_algebraic = true;
+        while (more_late_algebraic) {
+                more_late_algebraic = false;
+                NIR_PASS(more_late_algebraic, c->s, nir_opt_algebraic_late);
+                NIR_PASS_V(c->s, nir_opt_constant_folding);
+                NIR_PASS_V(c->s, nir_copy_prop);
+                NIR_PASS_V(c->s, nir_opt_dce);
+                NIR_PASS_V(c->s, nir_opt_cse);
+        }
+
         NIR_PASS_V(c->s, nir_lower_bool_to_int32);
         NIR_PASS_V(c->s, nir_convert_from_ssa, true);