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 c18318a62957eb72a7e546eb9fe49414b3769472..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);
 
@@ -758,9 +759,16 @@ v3d_nir_lower_vs_early(struct v3d_compile *c)
         NIR_PASS_V(c->s, nir_lower_global_vars_to_local);
         v3d_optimize_nir(c->s);
         NIR_PASS_V(c->s, nir_remove_dead_variables, nir_var_shader_in);
+
+        /* This must go before nir_lower_io */
+        if (c->vs_key->per_vertex_point_size)
+                NIR_PASS_V(c->s, nir_lower_point_size, 1.0f, 0.0f);
+
         NIR_PASS_V(c->s, nir_lower_io, nir_var_shader_in | nir_var_shader_out,
                    type_size_vec4,
                    (nir_lower_io_options)0);
+        /* clean up nir_lower_io's deref_var remains */
+        NIR_PASS_V(c->s, nir_opt_dce);
 }
 
 static void
@@ -805,8 +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;
         }
 }
@@ -819,7 +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);
         }
@@ -837,14 +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.
@@ -933,6 +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);
 
@@ -945,7 +963,8 @@ uint64_t *v3d_compile(const struct v3d_compiler *compiler,
         char *shaderdb;
         int ret = asprintf(&shaderdb,
                            "%s shader: %d inst, %d threads, %d loops, "
-                           "%d uniforms, %d max-temps, %d:%d spills:fills",
+                           "%d uniforms, %d max-temps, %d:%d spills:fills, "
+                           "%d sfu-stalls, %d inst-and-stalls",
                            vir_get_stage_name(c),
                            c->qpu_inst_count,
                            c->threads,
@@ -953,7 +972,9 @@ uint64_t *v3d_compile(const struct v3d_compiler *compiler,
                            c->num_uniforms,
                            vir_get_max_temps(c),
                            c->spills,
-                           c->fills);
+                           c->fills,
+                           c->qpu_inst_stalled_count,
+                           c->qpu_inst_count + c->qpu_inst_stalled_count);
         if (ret >= 0) {
                 if (V3D_DEBUG & V3D_DEBUG_SHADERDB)
                         fprintf(stderr, "SHADER-DB: %s\n", shaderdb);