glsl: add texture gather changes

V2 [Chris Forbes]:
   - Add new pattern, fixup parameter reading.

V3: Rebase onto new builtins machinery

Reviewed-by: Kenneth Graunke <[email protected]>
Reviewed-by: Ian Romanick <[email protected]>

13 files changed, 57 insertions, 4 deletions

diff --git a/src/glsl/builtin_functions.cpp b/src/glsl/builtin_functions.cpp
index 72054e0fe93..df735ef31c9 100644
--- a/src/glsl/builtin_functions.cpp
+++ b/src/glsl/builtin_functions.cpp
@@ -262,6 +262,13 @@ texture_query_lod(const _mesa_glsl_parse_state *state)
           state->ARB_texture_query_lod_enable;
 }
 
+static bool
+texture_gather(const _mesa_glsl_parse_state *state)
+{
+   return state->is_version(400, 0) ||
+          state->ARB_texture_gather_enable;
+}
+
 /* Desktop GL or OES_standard_derivatives + fragment shader only */
 static bool
 fs_oes_derivatives(const _mesa_glsl_parse_state *state)
@@ -1816,6 +1823,34 @@ builtin_builder::create_builtins()
                 _texture(ir_txd, shader_texture_lod_and_rect, glsl_type::vec4_type,  glsl_type::sampler2DRectShadow_type, glsl_type::vec4_type, TEX_PROJECT),
                 NULL);
 
+   add_function("textureGather",
+                _texture(ir_tg4, texture_gather, glsl_type::vec4_type, glsl_type::sampler2D_type, glsl_type::vec2_type),
+                _texture(ir_tg4, texture_gather, glsl_type::ivec4_type, glsl_type::isampler2D_type, glsl_type::vec2_type),
+                _texture(ir_tg4, texture_gather, glsl_type::uvec4_type, glsl_type::usampler2D_type, glsl_type::vec2_type),
+
+                _texture(ir_tg4, texture_gather, glsl_type::vec4_type, glsl_type::sampler2DArray_type, glsl_type::vec3_type),
+                _texture(ir_tg4, texture_gather, glsl_type::ivec4_type, glsl_type::isampler2DArray_type, glsl_type::vec3_type),
+                _texture(ir_tg4, texture_gather, glsl_type::uvec4_type, glsl_type::usampler2DArray_type, glsl_type::vec3_type),
+
+                _texture(ir_tg4, texture_gather, glsl_type::vec4_type, glsl_type::samplerCube_type, glsl_type::vec3_type),
+                _texture(ir_tg4, texture_gather, glsl_type::ivec4_type, glsl_type::isamplerCube_type, glsl_type::vec3_type),
+                _texture(ir_tg4, texture_gather, glsl_type::uvec4_type, glsl_type::usamplerCube_type, glsl_type::vec3_type),
+
+                _texture(ir_tg4, texture_gather, glsl_type::vec4_type, glsl_type::samplerCubeArray_type, glsl_type::vec4_type),
+                _texture(ir_tg4, texture_gather, glsl_type::ivec4_type, glsl_type::isamplerCubeArray_type, glsl_type::vec4_type),
+                _texture(ir_tg4, texture_gather, glsl_type::uvec4_type, glsl_type::usamplerCubeArray_type, glsl_type::vec4_type),
+                NULL);
+
+   add_function("textureGatherOffset",
+                _texture(ir_tg4, texture_gather, glsl_type::vec4_type, glsl_type::sampler2D_type, glsl_type::vec2_type, TEX_OFFSET),
+                _texture(ir_tg4, texture_gather, glsl_type::ivec4_type, glsl_type::isampler2D_type, glsl_type::vec2_type, TEX_OFFSET),
+                _texture(ir_tg4, texture_gather, glsl_type::uvec4_type, glsl_type::usampler2D_type, glsl_type::vec2_type, TEX_OFFSET),
+
+                _texture(ir_tg4, texture_gather, glsl_type::vec4_type, glsl_type::sampler2DArray_type, glsl_type::vec3_type, TEX_OFFSET),
+                _texture(ir_tg4, texture_gather, glsl_type::ivec4_type, glsl_type::isampler2DArray_type, glsl_type::vec3_type, TEX_OFFSET),
+                _texture(ir_tg4, texture_gather, glsl_type::uvec4_type, glsl_type::usampler2DArray_type, glsl_type::vec3_type, TEX_OFFSET),
+                NULL);
+
    F(dFdx)
    F(dFdy)
    F(fwidth)
diff --git a/src/glsl/glcpp/glcpp-parse.y b/src/glsl/glcpp/glcpp-parse.y
index 6eaa5f95ef1..c7ad3e95864 100644
--- a/src/glsl/glcpp/glcpp-parse.y
+++ b/src/glsl/glcpp/glcpp-parse.y
@@ -1248,6 +1248,9 @@ glcpp_parser_create (const struct gl_extensions *extensions, int api)
 
 	      if (extensions->EXT_shader_integer_mix)
 	         add_builtin_define(parser, "GL_EXT_shader_integer_mix", 1);
+
+	      if (extensions->ARB_texture_gather)
+	         add_builtin_define(parser, "GL_ARB_texture_gather", 1);
 	   }
 	}
 
diff --git a/src/glsl/glsl_parser_extras.cpp b/src/glsl/glsl_parser_extras.cpp
index 4f2f2893a90..7368a715688 100644
--- a/src/glsl/glsl_parser_extras.cpp
+++ b/src/glsl/glsl_parser_extras.cpp
@@ -522,6 +522,7 @@ static const _mesa_glsl_extension _mesa_glsl_supported_extensions[] = {
    EXT(ARB_gpu_shader5,                true,  false,     ARB_gpu_shader5),
    EXT(AMD_vertex_shader_layer,        true,  false,     AMD_vertex_shader_layer),
    EXT(EXT_shader_integer_mix,         true,  true,      EXT_shader_integer_mix),
+   EXT(ARB_texture_gather,             true,  false,     ARB_texture_gather),
 };
 
 #undef EXT
diff --git a/src/glsl/glsl_parser_extras.h b/src/glsl/glsl_parser_extras.h
index d557f8ba69d..9c3e88721f8 100644
--- a/src/glsl/glsl_parser_extras.h
+++ b/src/glsl/glsl_parser_extras.h
@@ -274,6 +274,8 @@ struct _mesa_glsl_parse_state {
    bool ARB_fragment_coord_conventions_warn;
    bool ARB_texture_rectangle_enable;
    bool ARB_texture_rectangle_warn;
+   bool ARB_texture_gather_enable;
+   bool ARB_texture_gather_warn;
    bool EXT_texture_array_enable;
    bool EXT_texture_array_warn;
    bool ARB_shader_texture_lod_enable;
diff --git a/src/glsl/ir.cpp b/src/glsl/ir.cpp
index b0f92cb3f53..81c6380cad0 100644
--- a/src/glsl/ir.cpp
+++ b/src/glsl/ir.cpp
@@ -1374,7 +1374,7 @@ ir_dereference::is_lvalue() const
 }
 
 
-static const char *tex_opcode_strs[] = { "tex", "txb", "txl", "txd", "txf", "txf_ms", "txs", "lod" };
+static const char *tex_opcode_strs[] = { "tex", "txb", "txl", "txd", "txf", "txf_ms", "txs", "lod", "tg4" };
 
 const char *ir_texture::opcode_string()
 {
diff --git a/src/glsl/ir.h b/src/glsl/ir.h
index 6c5630b0965..885837a7ab7 100644
--- a/src/glsl/ir.h
+++ b/src/glsl/ir.h
@@ -1559,7 +1559,8 @@ enum ir_texture_opcode {
    ir_txf,		/**< Texel fetch with explicit LOD */
    ir_txf_ms,           /**< Multisample texture fetch */
    ir_txs,		/**< Texture size */
-   ir_lod		/**< Texture lod query */
+   ir_lod,		/**< Texture lod query */
+   ir_tg4		/**< Texture gather */
 };
 
 
@@ -1584,6 +1585,7 @@ enum ir_texture_opcode {
  *      <type> <sampler> <coordinate>         <sample_index>)
  * (txs <type> <sampler> <lod>)
  * (lod <type> <sampler> <coordinate>)
+ * (tg4 <type> <sampler> <coordinate> 0)
  */
 class ir_texture : public ir_rvalue {
 public:
diff --git a/src/glsl/ir_clone.cpp b/src/glsl/ir_clone.cpp
index fb303b0b75c..fe531a8e306 100644
--- a/src/glsl/ir_clone.cpp
+++ b/src/glsl/ir_clone.cpp
@@ -248,6 +248,7 @@ ir_texture::clone(void *mem_ctx, struct hash_table *ht) const
    switch (this->op) {
    case ir_tex:
    case ir_lod:
+   case ir_tg4:
       break;
    case ir_txb:
       new_tex->lod_info.bias = this->lod_info.bias->clone(mem_ctx, ht);
diff --git a/src/glsl/ir_hv_accept.cpp b/src/glsl/ir_hv_accept.cpp
index 76a607d17cb..3877e721c24 100644
--- a/src/glsl/ir_hv_accept.cpp
+++ b/src/glsl/ir_hv_accept.cpp
@@ -214,6 +214,7 @@ ir_texture::accept(ir_hierarchical_visitor *v)
    switch (this->op) {
    case ir_tex:
    case ir_lod:
+   case ir_tg4:
       break;
    case ir_txb:
       s = this->lod_info.bias->accept(v);
diff --git a/src/glsl/ir_print_visitor.cpp b/src/glsl/ir_print_visitor.cpp
index b5183108d6b..2c06b95f61a 100644
--- a/src/glsl/ir_print_visitor.cpp
+++ b/src/glsl/ir_print_visitor.cpp
@@ -266,7 +266,7 @@ void ir_print_visitor::visit(ir_texture *ir)
       printf(" ");
    }
 
-   if (ir->op != ir_txf && ir->op != ir_txf_ms && ir->op != ir_txs) {
+   if (ir->op != ir_txf && ir->op != ir_txf_ms && ir->op != ir_txs && ir->op != ir_tg4) {
       if (ir->projector)
 	 ir->projector->accept(this);
       else
@@ -285,6 +285,7 @@ void ir_print_visitor::visit(ir_texture *ir)
    {
    case ir_tex:
    case ir_lod:
+   case ir_tg4:
       break;
    case ir_txb:
       ir->lod_info.bias->accept(this);
diff --git a/src/glsl/ir_reader.cpp b/src/glsl/ir_reader.cpp
index f0318ea210a..8038b8f5db2 100644
--- a/src/glsl/ir_reader.cpp
+++ b/src/glsl/ir_reader.cpp
@@ -947,6 +947,8 @@ ir_reader::read_texture(s_expression *expr)
       { "txf_ms", s_type, s_sampler, s_coord, s_sample_index };
    s_pattern txs_pattern[] =
       { "txs", s_type, s_sampler, s_lod };
+   s_pattern tg4_pattern[] =
+      { "tg4", s_type, s_sampler, s_coord, s_offset };
    s_pattern other_pattern[] =
       { tag, s_type, s_sampler, s_coord, s_offset, s_proj, s_shadow, s_lod };
 
@@ -960,6 +962,8 @@ ir_reader::read_texture(s_expression *expr)
       op = ir_txf_ms;
    } else if (MATCH(expr, txs_pattern)) {
       op = ir_txs;
+   } else if (MATCH(expr, tg4_pattern)) {
+      op = ir_tg4;
    } else if (MATCH(expr, other_pattern)) {
       op = ir_texture::get_opcode(tag->value());
       if (op == -1)
@@ -1010,7 +1014,7 @@ ir_reader::read_texture(s_expression *expr)
       }
    }
 
-   if (op != ir_txf && op != ir_txf_ms && op != ir_txs && op != ir_lod) {
+   if (op != ir_txf && op != ir_txf_ms && op != ir_txs && op != ir_lod && op != ir_tg4) {
       s_int *proj_as_int = SX_AS_INT(s_proj);
       if (proj_as_int && proj_as_int->value() == 1) {
 	 tex->projector = NULL;
diff --git a/src/glsl/ir_rvalue_visitor.cpp b/src/glsl/ir_rvalue_visitor.cpp
index 8eb1c62c30a..4c3d813206c 100644
--- a/src/glsl/ir_rvalue_visitor.cpp
+++ b/src/glsl/ir_rvalue_visitor.cpp
@@ -57,6 +57,7 @@ ir_rvalue_base_visitor::rvalue_visit(ir_texture *ir)
    switch (ir->op) {
    case ir_tex:
    case ir_lod:
+   case ir_tg4:
       break;
    case ir_txb:
       handle_rvalue(&ir->lod_info.bias);
diff --git a/src/glsl/opt_tree_grafting.cpp b/src/glsl/opt_tree_grafting.cpp
index 9aceb134def..03b920d8a2f 100644
--- a/src/glsl/opt_tree_grafting.cpp
+++ b/src/glsl/opt_tree_grafting.cpp
@@ -275,6 +275,7 @@ ir_tree_grafting_visitor::visit_enter(ir_texture *ir)
    switch (ir->op) {
    case ir_tex:
    case ir_lod:
+   case ir_tg4:
       break;
    case ir_txb:
       if (do_graft(&ir->lod_info.bias))
diff --git a/src/glsl/standalone_scaffolding.cpp b/src/glsl/standalone_scaffolding.cpp
index 6dec205f7ab..b03734c0693 100644
--- a/src/glsl/standalone_scaffolding.cpp
+++ b/src/glsl/standalone_scaffolding.cpp
@@ -105,6 +105,7 @@ void initialize_context_to_defaults(struct gl_context *ctx, gl_api api)
    ctx->Extensions.ARB_texture_multisample = true;
    ctx->Extensions.ARB_texture_query_lod = true;
    ctx->Extensions.ARB_gpu_shader5 = true;
+   ctx->Extensions.ARB_texture_gather = true;
 
    ctx->Const.GLSLVersion = 120;