glsl: use separate varying slots for patch varyings

The idea is to allow 32 normal varyings and 32 patch varyings,
a total of 64. Previously, only a total of 32 was allowed.

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

diff --git a/src/glsl/ast_to_hir.cpp b/src/glsl/ast_to_hir.cpp
index 1cc0731e93247697aeed875bb967f93647bb9922..a45b6e8f026ea449fcaa19a9f395970645bba77c 100644 (file)
--- a/src/glsl/ast_to_hir.cpp
+++ b/src/glsl/ast_to_hir.cpp
@@ -2367,7 +2367,10 @@ validate_explicit_location(const struct ast_type_qualifier *qual,
          case MESA_SHADER_TESS_CTRL:
          case MESA_SHADER_TESS_EVAL:
          case MESA_SHADER_GEOMETRY:
-            var->data.location = qual->location + VARYING_SLOT_VAR0;
+            if (var->data.patch)
+               var->data.location = qual->location + VARYING_SLOT_PATCH0;
+            else
+               var->data.location = qual->location + VARYING_SLOT_VAR0;
             break;
 
          case MESA_SHADER_FRAGMENT:

diff --git a/src/glsl/ir_set_program_inouts.cpp b/src/glsl/ir_set_program_inouts.cpp
index fe4d6da45db57bd20b00b00e5f5fd8082c0ada98..b7a0f6e95baf0da51f9fcd343e7208e00044f040 100644 (file)
--- a/src/glsl/ir_set_program_inouts.cpp
+++ b/src/glsl/ir_set_program_inouts.cpp
@@ -103,10 +103,26 @@ mark(struct gl_program *prog, ir_variable *var, int offset, int len,
    for (int i = 0; i < len; i++) {
       bool dual_slot = is_dual_slot(var);
       int idx = var->data.location + var->data.index + offset + i;
-      GLbitfield64 bitfield = BITFIELD64_BIT(idx);
+      bool is_patch_generic = var->data.patch &&
+                              idx != VARYING_SLOT_TESS_LEVEL_INNER &&
+                              idx != VARYING_SLOT_TESS_LEVEL_OUTER;
+      GLbitfield64 bitfield;
+
+      if (is_patch_generic) {
+         assert(idx >= VARYING_SLOT_PATCH0 && idx < VARYING_SLOT_TESS_MAX);
+         bitfield = BITFIELD64_BIT(idx - VARYING_SLOT_PATCH0);
+      }
+      else {
+         assert(idx < VARYING_SLOT_MAX);
+         bitfield = BITFIELD64_BIT(idx);
+      }
 
       if (var->data.mode == ir_var_shader_in) {
-         prog->InputsRead |= bitfield;
+         if (is_patch_generic)
+            prog->PatchInputsRead |= bitfield;
+         else
+            prog->InputsRead |= bitfield;
+
          if (dual_slot)
             prog->DoubleInputsRead |= bitfield;
          if (is_fragment_shader) {
@@ -122,7 +138,10 @@ mark(struct gl_program *prog, ir_variable *var, int offset, int len,
          prog->SystemValuesRead |= bitfield;
       } else {
          assert(var->data.mode == ir_var_shader_out);
-        prog->OutputsWritten |= bitfield;
+         if (is_patch_generic)
+            prog->PatchOutputsWritten |= bitfield;
+         else
+            prog->OutputsWritten |= bitfield;
       }
    }
 }
@@ -406,6 +425,8 @@ do_set_program_inouts(exec_list *instructions, struct gl_program *prog,
 
    prog->InputsRead = 0;
    prog->OutputsWritten = 0;
+   prog->PatchInputsRead = 0;
+   prog->PatchOutputsWritten = 0;
    prog->SystemValuesRead = 0;
    if (shader_stage == MESA_SHADER_FRAGMENT) {
       gl_fragment_program *fprog = (gl_fragment_program *) prog;

diff --git a/src/glsl/link_varyings.cpp b/src/glsl/link_varyings.cpp
index 1104872ca3c4fcf0678096a9813d10be1dcb00a2..1c52ff35f44a9a2339fc653507e3333b9f438b40 100644 (file)
--- a/src/glsl/link_varyings.cpp
+++ b/src/glsl/link_varyings.cpp
@@ -985,8 +985,17 @@ varying_matches::assign_locations()
          &varying_matches::match_comparator);
 
    unsigned generic_location = 0;
+   unsigned generic_patch_location = MAX_VARYING*4;
 
    for (unsigned i = 0; i < this->num_matches; i++) {
+      unsigned *location = &generic_location;
+
+      if ((this->matches[i].consumer_var &&
+           this->matches[i].consumer_var->data.patch) ||
+          (this->matches[i].producer_var &&
+           this->matches[i].producer_var->data.patch))
+         location = &generic_patch_location;
+
       /* Advance to the next slot if this varying has a different packing
        * class than the previous one, and we're not already on a slot
        * boundary.
@@ -994,12 +1003,12 @@ varying_matches::assign_locations()
       if (i > 0 &&
           this->matches[i - 1].packing_class
           != this->matches[i].packing_class) {
-         generic_location = ALIGN(generic_location, 4);
+         *location = ALIGN(*location, 4);
       }
 
-      this->matches[i].generic_location = generic_location;
+      this->matches[i].generic_location = *location;
 
-      generic_location += this->matches[i].num_components;
+      *location += this->matches[i].num_components;
    }
 
    return (generic_location + 3) / 4;
@@ -1213,11 +1222,11 @@ bool
 populate_consumer_input_sets(void *mem_ctx, exec_list *ir,
                              hash_table *consumer_inputs,
                              hash_table *consumer_interface_inputs,
-                             ir_variable *consumer_inputs_with_locations[VARYING_SLOT_MAX])
+                             ir_variable *consumer_inputs_with_locations[VARYING_SLOT_TESS_MAX])
 {
    memset(consumer_inputs_with_locations,
           0,
-          sizeof(consumer_inputs_with_locations[0]) * VARYING_SLOT_MAX);
+          sizeof(consumer_inputs_with_locations[0]) * VARYING_SLOT_TESS_MAX);
 
    foreach_in_list(ir_instruction, node, ir) {
       ir_variable *const input_var = node->as_variable();
@@ -1273,7 +1282,7 @@ get_matching_input(void *mem_ctx,
                    const ir_variable *output_var,
                    hash_table *consumer_inputs,
                    hash_table *consumer_interface_inputs,
-                   ir_variable *consumer_inputs_with_locations[VARYING_SLOT_MAX])
+                   ir_variable *consumer_inputs_with_locations[VARYING_SLOT_TESS_MAX])
 {
    ir_variable *input_var;
 
@@ -1414,7 +1423,7 @@ assign_varying_locations(struct gl_context *ctx,
       = hash_table_ctor(0, hash_table_string_hash, hash_table_string_compare);
    hash_table *consumer_interface_inputs
       = hash_table_ctor(0, hash_table_string_hash, hash_table_string_compare);
-   ir_variable *consumer_inputs_with_locations[VARYING_SLOT_MAX] = {
+   ir_variable *consumer_inputs_with_locations[VARYING_SLOT_TESS_MAX] = {
       NULL,
    };
 

diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h
index 85e04b5f18b972e676036ae0e78626dfa6aaefc9..b13271bbd06da1f0e6e76f0f1b38910c0337305b 100644 (file)
--- a/src/mesa/main/mtypes.h
+++ b/src/mesa/main/mtypes.h
@@ -243,7 +243,9 @@ typedef enum
    VARYING_SLOT_TESS_LEVEL_OUTER, /* Only appears as TCS output. */
    VARYING_SLOT_TESS_LEVEL_INNER, /* Only appears as TCS output. */
    VARYING_SLOT_VAR0, /* First generic varying slot */
-   VARYING_SLOT_MAX = VARYING_SLOT_VAR0 + MAX_VARYING
+   VARYING_SLOT_MAX = VARYING_SLOT_VAR0 + MAX_VARYING,
+   VARYING_SLOT_PATCH0 = VARYING_SLOT_MAX,
+   VARYING_SLOT_TESS_MAX = VARYING_SLOT_PATCH0 + MAX_VARYING
 } gl_varying_slot;
 
 
@@ -2103,6 +2105,8 @@ struct gl_program
    GLbitfield64 InputsRead;     /**< Bitmask of which input regs are read */
    GLbitfield64 DoubleInputsRead;     /**< Bitmask of which input regs are read  and are doubles */
    GLbitfield64 OutputsWritten; /**< Bitmask of which output regs are written */
+   GLbitfield PatchInputsRead;  /**< VAR[0..31] usage for patch inputs (user-defined only) */
+   GLbitfield PatchOutputsWritten; /**< VAR[0..31] usage for patch outputs (user-defined only) */
    GLbitfield SystemValuesRead;   /**< Bitmask of SYSTEM_VALUE_x inputs used */
    GLbitfield TexturesUsed[MAX_COMBINED_TEXTURE_IMAGE_UNITS];  /**< TEXTURE_x_BIT bitmask */
    GLbitfield SamplersUsed;   /**< Bitfield of which samplers are used */