Mesa (master): i965/fs: Add support for 16-wide dispatch with uniforms in use.

[Eric Anholt]

Module: Mesa
Branch: master
Commit: 3b20f999bb7e9056e83ca09a842a9747d4ac1674
URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=3b20f999bb7e9056e83ca09a842a9747d4ac1674

Author: Eric Anholt <eric at anholt.net>
Date:   Wed Mar 23 12:50:53 2011 -0700

i965/fs: Add support for 16-wide dispatch with uniforms in use.

This is glued in in a bit of an ugly way -- we rely on the uniforms
having been set up by 8-wide dispatch, and we just reuse them without
the ability to add new uniforms for any reason, since the 8-wide
compile is already completed.  Today, this all works out because our
optimization passes are effectively the same for both and even if they
weren't, we don't reduce the set of uniforms pushed after
optimization.

Reviewed-by: Kenneth Graunke <kenneth at whitecape.org>

---

 src/mesa/drivers/dri/i965/brw_fs.cpp |   60 +++++++++++++++++++++++++++++----
 src/mesa/drivers/dri/i965/brw_fs.h   |    1 +
 2 files changed, 53 insertions(+), 8 deletions(-)

diff --git a/src/mesa/drivers/dri/i965/brw_fs.cpp b/src/mesa/drivers/dri/i965/brw_fs.cpp
index 5a643a9..b0fe0c7 100644
--- a/src/mesa/drivers/dri/i965/brw_fs.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs.cpp
@@ -339,6 +339,31 @@ fs_visitor::variable_storage(ir_variable *var)
    return (fs_reg *)hash_table_find(this->variable_ht, var);
 }
 
+void
+import_uniforms_callback(const void *key,
+			 void *data,
+			 void *closure)
+{
+   struct hash_table *dst_ht = (struct hash_table *)closure;
+   const fs_reg *reg = (const fs_reg *)data;
+
+   if (reg->file != UNIFORM)
+      return;
+
+   hash_table_insert(dst_ht, data, key);
+}
+
+/* For 16-wide, we need to follow from the uniform setup of 8-wide dispatch.
+ * This brings in those uniform definitions
+ */
+void
+fs_visitor::import_uniforms(struct hash_table *src_variable_ht)
+{
+   hash_table_call_foreach(src_variable_ht,
+			   import_uniforms_callback,
+			   variable_ht);
+}
+
 /* Our support for uniforms is piggy-backed on the struct
  * gl_fragment_program, because that's where the values actually
  * get stored, rather than in some global gl_shader_program uniform
@@ -715,6 +740,13 @@ fs_visitor::visit(ir_variable *ir)
    if (ir->mode == ir_var_uniform) {
       int param_index = c->prog_data.nr_params;
 
+      if (c->dispatch_width == 16) {
+	 if (!variable_storage(ir)) {
+	    fail("Failed to find uniform '%s' in 16-wide\n", ir->name);
+	 }
+	 return;
+      }
+
       if (!strncmp(ir->name, "gl_", 3)) {
 	 setup_builtin_uniform_values(ir);
       } else {
@@ -1388,6 +1420,12 @@ fs_visitor::visit(ir_texture *ir)
 	 0
       };
 
+      if (c->dispatch_width == 16) {
+	 fail("rectangle scale uniform setup not supported on 16-wide\n");
+	 this->result = fs_reg(this, ir->type);
+	 return;
+      }
+
       c->prog_data.param_convert[c->prog_data.nr_params] =
 	 PARAM_NO_CONVERT;
       c->prog_data.param_convert[c->prog_data.nr_params + 1] =
@@ -2708,6 +2746,9 @@ fs_visitor::generate_pull_constant_load(fs_inst *inst, struct brw_reg dst)
 void
 fs_visitor::setup_paramvalues_refs()
 {
+   if (c->dispatch_width != 8)
+      return;
+
    /* Set up the pointers to ParamValues now that that array is finalized. */
    for (unsigned int i = 0; i < c->prog_data.nr_params; i++) {
       c->prog_data.param[i] =
@@ -2910,6 +2951,11 @@ fs_visitor::setup_pull_constants()
    if (c->prog_data.nr_params <= max_uniform_components)
       return;
 
+   if (c->dispatch_width == 16) {
+      fail("Pull constants not supported in 16-wide\n");
+      return;
+   }
+
    /* Just demote the end of the list.  We could probably do better
     * here, demoting things that are rarely used in the program first.
     */
@@ -3937,17 +3983,11 @@ bool
 fs_visitor::run()
 {
    uint32_t prog_offset_16 = 0;
+   uint32_t orig_nr_params = c->prog_data.nr_params;
 
    brw_wm_payload_setup(brw, c);
 
    if (c->dispatch_width == 16) {
-      if (c->prog_data.curb_read_length) {
-	 /* Haven't hooked in support for uniforms through the 16-wide
-	  * version yet.
-	  */
-	 return false;
-      }
-
       /* align to 64 byte boundary. */
       while ((c->func.nr_insn * sizeof(struct brw_instruction)) % 64) {
 	 brw_NOP(p);
@@ -4031,6 +4071,9 @@ fs_visitor::run()
    } else {
       c->prog_data.total_grf_16 = grf_used;
       c->prog_data.prog_offset_16 = prog_offset_16;
+
+      /* Make sure we didn't try to sneak in an extra uniform */
+      assert(orig_nr_params == c->prog_data.nr_params);
    }
 
    return !failed;
@@ -4068,9 +4111,10 @@ brw_wm_fs_emit(struct brw_context *brw, struct brw_wm_compile *c)
       return false;
    }
 
-   if (intel->gen >= 5) {
+   if (intel->gen >= 5 && c->prog_data.nr_pull_params == 0) {
       c->dispatch_width = 16;
       fs_visitor v2(c, shader);
+      v2.import_uniforms(v.variable_ht);
       v2.run();
    }
 
diff --git a/src/mesa/drivers/dri/i965/brw_fs.h b/src/mesa/drivers/dri/i965/brw_fs.h
index 60398ac..518d091 100644
--- a/src/mesa/drivers/dri/i965/brw_fs.h
+++ b/src/mesa/drivers/dri/i965/brw_fs.h
@@ -420,6 +420,7 @@ public:
 
    fs_reg *variable_storage(ir_variable *var);
    int virtual_grf_alloc(int size);
+   void import_uniforms(struct hash_table *src_variable_ht);
 
    void visit(ir_variable *ir);
    void visit(ir_assignment *ir);