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]>

4 files changed, 42 insertions, 28 deletions

diff --git a/src/intel/vulkan/anv_cmd_buffer.c b/src/intel/vulkan/anv_cmd_buffer.c
index 5eec67cb607..64d1417f5b1 100644
--- a/src/intel/vulkan/anv_cmd_buffer.c
+++ b/src/intel/vulkan/anv_cmd_buffer.c
@@ -629,6 +629,26 @@ anv_cmd_buffer_merge_dynamic(struct anv_cmd_buffer *cmd_buffer,
    return state;
 }
 
+static uint32_t
+anv_push_constant_value(struct anv_push_constants *data, uint32_t param)
+{
+   if (BRW_PARAM_IS_BUILTIN(param)) {
+      switch (param) {
+      case BRW_PARAM_BUILTIN_ZERO:
+         return 0;
+      default:
+         unreachable("Invalid param builtin");
+      }
+   } else {
+      uint32_t offset = ANV_PARAM_PUSH_OFFSET(param);
+      assert(offset % sizeof(uint32_t) == 0);
+      if (offset < data->size)
+         return *(uint32_t *)((uint8_t *)data + offset);
+      else
+         return 0;
+   }
+}
+
 struct anv_state
 anv_cmd_buffer_push_constants(struct anv_cmd_buffer *cmd_buffer,
                               gl_shader_stage stage)
@@ -653,10 +673,8 @@ anv_cmd_buffer_push_constants(struct anv_cmd_buffer *cmd_buffer,
 
    /* Walk through the param array and fill the buffer with data */
    uint32_t *u32_map = state.map;
-   for (unsigned i = 0; i < prog_data->nr_params; i++) {
-      uint32_t offset = (uintptr_t)prog_data->param[i];
-      u32_map[i] = *(uint32_t *)((uint8_t *)data + offset);
-   }
+   for (unsigned i = 0; i < prog_data->nr_params; i++)
+      u32_map[i] = anv_push_constant_value(data, prog_data->param[i]);
 
    anv_state_flush(cmd_buffer->device, state);
 
@@ -695,8 +713,7 @@ anv_cmd_buffer_cs_push_constants(struct anv_cmd_buffer *cmd_buffer)
       for (unsigned i = 0;
            i < cs_prog_data->push.cross_thread.dwords;
            i++) {
-         uint32_t offset = (uintptr_t)prog_data->param[i];
-         u32_map[i] = *(uint32_t *)((uint8_t *)data + offset);
+         u32_map[i] = anv_push_constant_value(data, prog_data->param[i]);
       }
    }
 
@@ -708,8 +725,8 @@ anv_cmd_buffer_cs_push_constants(struct anv_cmd_buffer *cmd_buffer)
          unsigned src = cs_prog_data->push.cross_thread.dwords;
          for ( ; src < prog_data->nr_params; src++, dst++) {
             if (src != cs_prog_data->thread_local_id_index) {
-               uint32_t offset = (uintptr_t)prog_data->param[src];
-               u32_map[dst] = *(uint32_t *)((uint8_t *)data + offset);
+               u32_map[dst] =
+                  anv_push_constant_value(data, prog_data->param[src]);
             } else {
                u32_map[dst] = t * cs_prog_data->simd_size;
             }
diff --git a/src/intel/vulkan/anv_nir_apply_pipeline_layout.c b/src/intel/vulkan/anv_nir_apply_pipeline_layout.c
index a35687379c8..f5a274d1a53 100644
--- a/src/intel/vulkan/anv_nir_apply_pipeline_layout.c
+++ b/src/intel/vulkan/anv_nir_apply_pipeline_layout.c
@@ -304,17 +304,13 @@ apply_pipeline_layout_block(nir_block *block,
 }
 
 static void
-setup_vec4_uniform_value(const union gl_constant_value **params,
-                         const union gl_constant_value *values,
-                         unsigned n)
+setup_vec4_uniform_value(uint32_t *params, uint32_t offset, unsigned n)
 {
-   static const gl_constant_value zero = { 0 };
-
    for (unsigned i = 0; i < n; ++i)
-      params[i] = &values[i];
+      params[i] = ANV_PARAM_PUSH(offset + i * sizeof(uint32_t));
 
    for (unsigned i = n; i < 4; ++i)
-      params[i] = &zero;
+      params[i] = BRW_PARAM_BUILTIN_ZERO;
 }
 
 void
@@ -478,22 +474,21 @@ anv_nir_apply_pipeline_layout(struct anv_pipeline *pipeline,
       }
 
       struct anv_push_constants *null_data = NULL;
-      const gl_constant_value **param =
-         prog_data->param + (shader->num_uniforms / 4);
+      uint32_t *param = prog_data->param + (shader->num_uniforms / 4);
       const struct brw_image_param *image_param = null_data->images;
       for (uint32_t i = 0; i < map->image_count; i++) {
          setup_vec4_uniform_value(param + BRW_IMAGE_PARAM_SURFACE_IDX_OFFSET,
-            (const union gl_constant_value *)&image_param->surface_idx, 1);
+                                  (uintptr_t)&image_param->surface_idx, 1);
          setup_vec4_uniform_value(param + BRW_IMAGE_PARAM_OFFSET_OFFSET,
-            (const union gl_constant_value *)image_param->offset, 2);
+                                  (uintptr_t)image_param->offset, 2);
          setup_vec4_uniform_value(param + BRW_IMAGE_PARAM_SIZE_OFFSET,
-            (const union gl_constant_value *)image_param->size, 3);
+                                  (uintptr_t)image_param->size, 3);
          setup_vec4_uniform_value(param + BRW_IMAGE_PARAM_STRIDE_OFFSET,
-            (const union gl_constant_value *)image_param->stride, 4);
+                                  (uintptr_t)image_param->stride, 4);
          setup_vec4_uniform_value(param + BRW_IMAGE_PARAM_TILING_OFFSET,
-            (const union gl_constant_value *)image_param->tiling, 3);
+                                  (uintptr_t)image_param->tiling, 3);
          setup_vec4_uniform_value(param + BRW_IMAGE_PARAM_SWIZZLING_OFFSET,
-            (const union gl_constant_value *)image_param->swizzling, 2);
+                                  (uintptr_t)image_param->swizzling, 2);
 
          param += BRW_IMAGE_PARAM_SIZE;
          image_param ++;
diff --git a/src/intel/vulkan/anv_pipeline.c b/src/intel/vulkan/anv_pipeline.c
index 3db9b9a6246..e6a7393a3d3 100644
--- a/src/intel/vulkan/anv_pipeline.c
+++ b/src/intel/vulkan/anv_pipeline.c
@@ -415,8 +415,7 @@ anv_pipeline_compile(struct anv_pipeline *pipeline,
 
    if (prog_data->nr_params > 0) {
       /* XXX: I think we're leaking this */
-      prog_data->param = (const union gl_constant_value **)
-         malloc(prog_data->nr_params * sizeof(union gl_constant_value *));
+      prog_data->param = malloc(prog_data->nr_params * sizeof(uint32_t));
 
       /* We now set the param values to be offsets into a
        * anv_push_constant_data structure.  Since the compiler doesn't
@@ -427,8 +426,8 @@ anv_pipeline_compile(struct anv_pipeline *pipeline,
       if (nir->num_uniforms > 0) {
          /* Fill out the push constants section of the param array */
          for (unsigned i = 0; i < MAX_PUSH_CONSTANTS_SIZE / sizeof(float); i++)
-            prog_data->param[i] = (const union gl_constant_value *)
-               &null_data->client_data[i * sizeof(float)];
+            prog_data->param[i] = ANV_PARAM_PUSH(
+               (uintptr_t)&null_data->client_data[i * sizeof(float)]);
       }
    }
 
@@ -540,7 +539,7 @@ anv_pipeline_compile_vs(struct anv_pipeline *pipeline,
       unsigned code_size;
       const unsigned *shader_code =
          brw_compile_vs(compiler, NULL, mem_ctx, &key, &prog_data, nir,
-                        NULL, false, -1, &code_size, NULL);
+                        false, -1, &code_size, NULL);
       if (shader_code == NULL) {
          ralloc_free(mem_ctx);
          return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
diff --git a/src/intel/vulkan/anv_private.h b/src/intel/vulkan/anv_private.h
index b33370c3da1..abc278b40b7 100644
--- a/src/intel/vulkan/anv_private.h
+++ b/src/intel/vulkan/anv_private.h
@@ -1567,6 +1567,9 @@ struct anv_vertex_binding {
    VkDeviceSize                                 offset;
 };
 
+#define ANV_PARAM_PUSH(offset)         ((1 << 16) | (uint32_t)(offset))
+#define ANV_PARAM_PUSH_OFFSET(param)   ((param) & 0xffff)
+
 struct anv_push_constants {
    /* Current allocated size of this push constants data structure.
     * Because a decent chunk of it may not be used (images on SKL, for