intel: Rewrite the world of push/pull params

This moves us away to the array of pointers model and onto a model where
each param is represented by a generic uint32_t handle.  We reserve 2^16
of these handles for builtins that get generated by somewhere inside the
compiler and have well-defined meanings.  Generic params have handles
whose meanings are defined by the driver.

The primary downside to this new approach is that it moves a little bit
of the work that we would normally do at compile time to draw time.  On
my laptop this hurts OglBatch6 by no more than 1% and doesn't seem to
have any measurable affect on OglBatch7.  So, while this may come back
to bite us, it doesn't look too bad.

Reviewed-by: Jordan Justen <[email protected]>
Reviewed-by: Kenneth Graunke <[email protected]>

8 files changed, 90 insertions, 40 deletions

diff --git a/src/intel/compiler/brw_compiler.h b/src/intel/compiler/brw_compiler.h
index 038f3f95512..f2f9be750a0 100644
--- a/src/intel/compiler/brw_compiler.h
+++ b/src/intel/compiler/brw_compiler.h
@@ -36,7 +36,6 @@ extern "C" {
 struct ra_regs;
 struct nir_shader;
 struct brw_program;
-union gl_constant_value;
 
 struct brw_compiler {
    const struct gen_device_info *devinfo;
@@ -491,6 +490,66 @@ struct brw_ubo_range
    uint8_t length;
 };
 
+/* We reserve the first 2^16 values for builtins */
+#define BRW_PARAM_IS_BUILTIN(param) (((param) & 0xffff0000) == 0)
+
+enum brw_param_builtin {
+   BRW_PARAM_BUILTIN_ZERO,
+
+   BRW_PARAM_BUILTIN_CLIP_PLANE_0_X,
+   BRW_PARAM_BUILTIN_CLIP_PLANE_0_Y,
+   BRW_PARAM_BUILTIN_CLIP_PLANE_0_Z,
+   BRW_PARAM_BUILTIN_CLIP_PLANE_0_W,
+   BRW_PARAM_BUILTIN_CLIP_PLANE_1_X,
+   BRW_PARAM_BUILTIN_CLIP_PLANE_1_Y,
+   BRW_PARAM_BUILTIN_CLIP_PLANE_1_Z,
+   BRW_PARAM_BUILTIN_CLIP_PLANE_1_W,
+   BRW_PARAM_BUILTIN_CLIP_PLANE_2_X,
+   BRW_PARAM_BUILTIN_CLIP_PLANE_2_Y,
+   BRW_PARAM_BUILTIN_CLIP_PLANE_2_Z,
+   BRW_PARAM_BUILTIN_CLIP_PLANE_2_W,
+   BRW_PARAM_BUILTIN_CLIP_PLANE_3_X,
+   BRW_PARAM_BUILTIN_CLIP_PLANE_3_Y,
+   BRW_PARAM_BUILTIN_CLIP_PLANE_3_Z,
+   BRW_PARAM_BUILTIN_CLIP_PLANE_3_W,
+   BRW_PARAM_BUILTIN_CLIP_PLANE_4_X,
+   BRW_PARAM_BUILTIN_CLIP_PLANE_4_Y,
+   BRW_PARAM_BUILTIN_CLIP_PLANE_4_Z,
+   BRW_PARAM_BUILTIN_CLIP_PLANE_4_W,
+   BRW_PARAM_BUILTIN_CLIP_PLANE_5_X,
+   BRW_PARAM_BUILTIN_CLIP_PLANE_5_Y,
+   BRW_PARAM_BUILTIN_CLIP_PLANE_5_Z,
+   BRW_PARAM_BUILTIN_CLIP_PLANE_5_W,
+   BRW_PARAM_BUILTIN_CLIP_PLANE_6_X,
+   BRW_PARAM_BUILTIN_CLIP_PLANE_6_Y,
+   BRW_PARAM_BUILTIN_CLIP_PLANE_6_Z,
+   BRW_PARAM_BUILTIN_CLIP_PLANE_6_W,
+   BRW_PARAM_BUILTIN_CLIP_PLANE_7_X,
+   BRW_PARAM_BUILTIN_CLIP_PLANE_7_Y,
+   BRW_PARAM_BUILTIN_CLIP_PLANE_7_Z,
+   BRW_PARAM_BUILTIN_CLIP_PLANE_7_W,
+
+   BRW_PARAM_BUILTIN_TESS_LEVEL_OUTER_X,
+   BRW_PARAM_BUILTIN_TESS_LEVEL_OUTER_Y,
+   BRW_PARAM_BUILTIN_TESS_LEVEL_OUTER_Z,
+   BRW_PARAM_BUILTIN_TESS_LEVEL_OUTER_W,
+   BRW_PARAM_BUILTIN_TESS_LEVEL_INNER_X,
+   BRW_PARAM_BUILTIN_TESS_LEVEL_INNER_Y,
+};
+
+#define BRW_PARAM_BUILTIN_CLIP_PLANE(idx, comp) \
+   (BRW_PARAM_BUILTIN_CLIP_PLANE_0_X + ((idx) << 2) + (comp))
+
+#define BRW_PARAM_BUILTIN_IS_CLIP_PLANE(param)  \
+   ((param) >= BRW_PARAM_BUILTIN_CLIP_PLANE_0_X && \
+    (param) <= BRW_PARAM_BUILTIN_CLIP_PLANE_7_W)
+
+#define BRW_PARAM_BUILTIN_CLIP_PLANE_IDX(param) \
+   (((param) - BRW_PARAM_BUILTIN_CLIP_PLANE_0_X) >> 2)
+
+#define BRW_PARAM_BUILTIN_CLIP_PLANE_COMP(param) \
+   (((param) - BRW_PARAM_BUILTIN_CLIP_PLANE_0_X) & 0x3)
+
 struct brw_stage_prog_data {
    struct {
       /** size of our binding table. */
@@ -529,11 +588,14 @@ struct brw_stage_prog_data {
 
    bool use_alt_mode; /**< Use ALT floating point mode?  Otherwise, IEEE. */
 
-   /* Pointers to tracked values (only valid once
-    * _mesa_load_state_parameters has been called at runtime).
+   /* 32-bit identifiers for all push/pull parameters.  These can be anything
+    * the driver wishes them to be; the core of the back-end compiler simply
+    * re-arranges them.  The one restriction is that the bottom 2^16 values
+    * are reserved for builtins defined in the brw_param_builtin enum defined
+    * above.
     */
-   const union gl_constant_value **param;
-   const union gl_constant_value **pull_param;
+   uint32_t *param;
+   uint32_t *pull_param;
 
    /** Image metadata passed to the shader as uniforms. */
    struct brw_image_param *image_param;
@@ -1020,7 +1082,6 @@ brw_compile_vs(const struct brw_compiler *compiler, void *log_data,
                const struct brw_vs_prog_key *key,
                struct brw_vs_prog_data *prog_data,
                const struct nir_shader *shader,
-               gl_clip_plane *clip_planes,
                bool use_legacy_snorm_formula,
                int shader_time_index,
                unsigned *final_assembly_size,
diff --git a/src/intel/compiler/brw_fs.cpp b/src/intel/compiler/brw_fs.cpp
index a40b910c1a0..c1d67750a3a 100644
--- a/src/intel/compiler/brw_fs.cpp
+++ b/src/intel/compiler/brw_fs.cpp
@@ -2084,10 +2084,9 @@ fs_visitor::assign_constant_locations()
    /* As the uniforms are going to be reordered, take the data from a temporary
     * copy of the original param[].
     */
-   gl_constant_value **param = ralloc_array(NULL, gl_constant_value*,
-                                            stage_prog_data->nr_params);
+   uint32_t *param = ralloc_array(NULL, uint32_t, stage_prog_data->nr_params);
    memcpy(param, stage_prog_data->param,
-          sizeof(gl_constant_value*) * stage_prog_data->nr_params);
+          sizeof(uint32_t) * stage_prog_data->nr_params);
    stage_prog_data->nr_params = num_push_constants;
    stage_prog_data->nr_pull_params = num_pull_constants;
 
@@ -2115,8 +2114,7 @@ fs_visitor::assign_constant_locations()
     */
    int new_thread_local_id_index = -1;
    for (unsigned int i = 0; i < uniforms; i++) {
-      const gl_constant_value *value = param[i];
-
+      uint32_t value = param[i];
       if (pull_constant_loc[i] != -1) {
          stage_prog_data->pull_param[pull_constant_loc[i]] = value;
       } else if (push_constant_loc[i] != -1) {
@@ -5967,7 +5965,7 @@ fs_visitor::allocate_registers(bool allow_spilling)
 }
 
 bool
-fs_visitor::run_vs(gl_clip_plane *clip_planes)
+fs_visitor::run_vs()
 {
    assert(stage == MESA_SHADER_VERTEX);
 
@@ -5981,7 +5979,7 @@ fs_visitor::run_vs(gl_clip_plane *clip_planes)
    if (failed)
       return false;
 
-   compute_clip_distance(clip_planes);
+   compute_clip_distance();
 
    emit_urb_writes();
 
diff --git a/src/intel/compiler/brw_fs.h b/src/intel/compiler/brw_fs.h
index f1ba193de7e..20405750b78 100644
--- a/src/intel/compiler/brw_fs.h
+++ b/src/intel/compiler/brw_fs.h
@@ -80,8 +80,8 @@ public:
 
    fs_reg vgrf(const glsl_type *const type);
    void import_uniforms(fs_visitor *v);
-   void setup_uniform_clipplane_values(gl_clip_plane *clip_planes);
-   void compute_clip_distance(gl_clip_plane *clip_planes);
+   void setup_uniform_clipplane_values();
+   void compute_clip_distance();
 
    fs_inst *get_instruction_generating_reg(fs_inst *start,
 					   fs_inst *end,
@@ -95,7 +95,7 @@ public:
    void DEP_RESOLVE_MOV(const brw::fs_builder &bld, int grf);
 
    bool run_fs(bool allow_spilling, bool do_rep_send);
-   bool run_vs(gl_clip_plane *clip_planes);
+   bool run_vs();
    bool run_tcs_single_patch();
    bool run_tes();
    bool run_gs();
diff --git a/src/intel/compiler/brw_fs_visitor.cpp b/src/intel/compiler/brw_fs_visitor.cpp
index 758c8bf44a9..4ee1d4e0022 100644
--- a/src/intel/compiler/brw_fs_visitor.cpp
+++ b/src/intel/compiler/brw_fs_visitor.cpp
@@ -465,7 +465,7 @@ fs_visitor::emit_fb_writes()
 }
 
 void
-fs_visitor::setup_uniform_clipplane_values(gl_clip_plane *clip_planes)
+fs_visitor::setup_uniform_clipplane_values()
 {
    const struct brw_vs_prog_key *key =
       (const struct brw_vs_prog_key *) this->key;
@@ -474,7 +474,7 @@ fs_visitor::setup_uniform_clipplane_values(gl_clip_plane *clip_planes)
       this->userplane[i] = fs_reg(UNIFORM, uniforms);
       for (int j = 0; j < 4; ++j) {
          stage_prog_data->param[uniforms + j] =
-            (gl_constant_value *) &clip_planes[i][j];
+            BRW_PARAM_BUILTIN_CLIP_PLANE(i, j);
       }
       uniforms += 4;
    }
@@ -486,7 +486,7 @@ fs_visitor::setup_uniform_clipplane_values(gl_clip_plane *clip_planes)
  * This does nothing if the shader uses gl_ClipDistance or user clipping is
  * disabled altogether.
  */
-void fs_visitor::compute_clip_distance(gl_clip_plane *clip_planes)
+void fs_visitor::compute_clip_distance()
 {
    struct brw_vue_prog_data *vue_prog_data = brw_vue_prog_data(prog_data);
    const struct brw_vs_prog_key *key =
@@ -518,7 +518,7 @@ void fs_visitor::compute_clip_distance(gl_clip_plane *clip_planes)
    if (outputs[clip_vertex].file == BAD_FILE)
       return;
 
-   setup_uniform_clipplane_values(clip_planes);
+   setup_uniform_clipplane_values();
 
    const fs_builder abld = bld.annotate("user clip distances");
 
diff --git a/src/intel/compiler/brw_vec4.cpp b/src/intel/compiler/brw_vec4.cpp
index 9b9f5863721..8614886967c 100644
--- a/src/intel/compiler/brw_vec4.cpp
+++ b/src/intel/compiler/brw_vec4.cpp
@@ -698,10 +698,9 @@ vec4_visitor::pack_uniform_registers()
    /* As the uniforms are going to be reordered, take the data from a temporary
     * copy of the original param[].
     */
-   gl_constant_value **param = ralloc_array(NULL, gl_constant_value*,
-                                            stage_prog_data->nr_params);
+   uint32_t *param = ralloc_array(NULL, uint32_t, stage_prog_data->nr_params);
    memcpy(param, stage_prog_data->param,
-          sizeof(gl_constant_value*) * stage_prog_data->nr_params);
+          sizeof(uint32_t) * stage_prog_data->nr_params);
 
    /* Now, figure out a packing of the live uniform vectors into our
     * push constants. Start with dvec{3,4} because they are aligned to
@@ -907,7 +906,7 @@ vec4_visitor::move_push_constants_to_pull_constants()
       pull_constant_loc[i / 4] = -1;
 
       if (i >= max_uniform_components) {
-         const gl_constant_value **values = &stage_prog_data->param[i];
+         uint32_t *values = &stage_prog_data->param[i];
 
          /* Try to find an existing copy of this uniform in the pull
           * constants if it was part of an array access already.
@@ -1764,11 +1763,10 @@ vec4_visitor::setup_uniforms(int reg)
     */
    if (devinfo->gen < 6 && this->uniforms == 0) {
       stage_prog_data->param =
-         reralloc(NULL, stage_prog_data->param, const gl_constant_value *, 4);
+         reralloc(NULL, stage_prog_data->param, uint32_t, 4);
       for (unsigned int i = 0; i < 4; i++) {
 	 unsigned int slot = this->uniforms * 4 + i;
-	 static gl_constant_value zero = { 0.0 };
-	 stage_prog_data->param[slot] = &zero;
+	 stage_prog_data->param[slot] = BRW_PARAM_BUILTIN_ZERO;
       }
 
       this->uniforms++;
@@ -2742,7 +2740,6 @@ brw_compile_vs(const struct brw_compiler *compiler, void *log_data,
                const struct brw_vs_prog_key *key,
                struct brw_vs_prog_data *prog_data,
                const nir_shader *src_shader,
-               gl_clip_plane *clip_planes,
                bool use_legacy_snorm_formula,
                int shader_time_index,
                unsigned *final_assembly_size,
@@ -2866,7 +2863,7 @@ brw_compile_vs(const struct brw_compiler *compiler, void *log_data,
       fs_visitor v(compiler, log_data, mem_ctx, key, &prog_data->base.base,
                    NULL, /* prog; Only used for TEXTURE_RECTANGLE on gen < 8 */
                    shader, 8, shader_time_index);
-      if (!v.run_vs(clip_planes)) {
+      if (!v.run_vs()) {
          if (error_str)
             *error_str = ralloc_strdup(mem_ctx, v.fail_msg);
 
@@ -2895,7 +2892,7 @@ brw_compile_vs(const struct brw_compiler *compiler, void *log_data,
       prog_data->base.dispatch_mode = DISPATCH_MODE_4X2_DUAL_OBJECT;
 
       vec4_vs_visitor v(compiler, log_data, key, prog_data,
-                        shader, clip_planes, mem_ctx,
+                        shader, mem_ctx,
                         shader_time_index, use_legacy_snorm_formula);
       if (!v.run()) {
          if (error_str)
diff --git a/src/intel/compiler/brw_vec4_gs_visitor.cpp b/src/intel/compiler/brw_vec4_gs_visitor.cpp
index a8e445c473c..5df6d562ce6 100644
--- a/src/intel/compiler/brw_vec4_gs_visitor.cpp
+++ b/src/intel/compiler/brw_vec4_gs_visitor.cpp
@@ -890,10 +890,9 @@ brw_compile_gs(const struct brw_compiler *compiler, void *log_data,
           * values.
           */
          const unsigned param_count = prog_data->base.base.nr_params;
-         gl_constant_value **param = ralloc_array(NULL, gl_constant_value*,
-                                                  param_count);
+         uint32_t *param = ralloc_array(NULL, uint32_t, param_count);
          memcpy(param, prog_data->base.base.param,
-                sizeof(gl_constant_value*) * param_count);
+                sizeof(uint32_t) * param_count);
 
          if (v.run()) {
             /* Success! Backup is not needed */
@@ -910,7 +909,7 @@ brw_compile_gs(const struct brw_compiler *compiler, void *log_data,
              * FIXME: Could more variables be modified by this execution?
              */
             memcpy(prog_data->base.base.param, param,
-                   sizeof(gl_constant_value*) * param_count);
+                   sizeof(uint32_t) * param_count);
             prog_data->base.base.nr_params = param_count;
             prog_data->base.base.nr_pull_params = 0;
             ralloc_free(param);
diff --git a/src/intel/compiler/brw_vec4_vs.h b/src/intel/compiler/brw_vec4_vs.h
index cd07e0e99de..b2a862fdbde 100644
--- a/src/intel/compiler/brw_vec4_vs.h
+++ b/src/intel/compiler/brw_vec4_vs.h
@@ -36,7 +36,6 @@ public:
                    const struct brw_vs_prog_key *key,
                    struct brw_vs_prog_data *vs_prog_data,
                    const nir_shader *shader,
-                   gl_clip_plane *clip_planes,
                    void *mem_ctx,
                    int shader_time_index,
                    bool use_legacy_snorm_formula);
@@ -57,8 +56,6 @@ private:
    const struct brw_vs_prog_key *const key;
    struct brw_vs_prog_data * const vs_prog_data;
 
-   gl_clip_plane *clip_planes;
-
    bool use_legacy_snorm_formula;
 };
 
diff --git a/src/intel/compiler/brw_vec4_vs_visitor.cpp b/src/intel/compiler/brw_vec4_vs_visitor.cpp
index ad7f067ad60..86f365e2f0f 100644
--- a/src/intel/compiler/brw_vec4_vs_visitor.cpp
+++ b/src/intel/compiler/brw_vec4_vs_visitor.cpp
@@ -124,7 +124,7 @@ vec4_vs_visitor::setup_uniform_clipplane_values()
       this->userplane[i].type = BRW_REGISTER_TYPE_F;
       for (int j = 0; j < 4; ++j) {
          stage_prog_data->param[this->uniforms * 4 + j] =
-            (gl_constant_value *) &clip_planes[i][j];
+            BRW_PARAM_BUILTIN_CLIP_PLANE(i, j);
       }
       ++this->uniforms;
    }
@@ -164,7 +164,6 @@ vec4_vs_visitor::vec4_vs_visitor(const struct brw_compiler *compiler,
                                  const struct brw_vs_prog_key *key,
                                  struct brw_vs_prog_data *vs_prog_data,
                                  const nir_shader *shader,
-                                 gl_clip_plane *clip_planes,
                                  void *mem_ctx,
                                  int shader_time_index,
                                  bool use_legacy_snorm_formula)
@@ -172,7 +171,6 @@ vec4_vs_visitor::vec4_vs_visitor(const struct brw_compiler *compiler,
                   mem_ctx, false /* no_spills */, shader_time_index),
      key(key),
      vs_prog_data(vs_prog_data),
-     clip_planes(clip_planes),
      use_legacy_snorm_formula(use_legacy_snorm_formula)
 {
 }