i965: Use sample barycentric coordinates with per sample shading
[mesa.git] / src / mesa / drivers / dri / i965 / brw_vec4.h
index 689040b933640529c40b2d4b9b815acff95cd1ab..4a5b57775c30abea3bb29aaacbf96267b7b49176 100644 (file)
@@ -39,6 +39,7 @@ extern "C" {
 
 #ifdef __cplusplus
 }; /* extern "C" */
+#include "gen8_generator.h"
 #endif
 
 #include "glsl/ir.h"
@@ -64,12 +65,6 @@ struct brw_vec4_prog_key {
     */
    GLuint nr_userclip_plane_consts:4;
 
-   /**
-    * True if the shader uses gl_ClipDistance, regardless of whether any clip
-    * flags are enabled.
-    */
-   GLuint uses_clip_distance:1;
-
    GLuint clamp_vertex_color:1;
 
    struct brw_sampler_prog_key_data tex;
@@ -80,6 +75,10 @@ struct brw_vec4_prog_key {
 extern "C" {
 #endif
 
+void
+brw_vec4_setup_prog_key_for_precompile(struct gl_context *ctx,
+                                       struct brw_vec4_prog_key *key,
+                                       GLuint id, struct gl_program *prog);
 bool brw_vec4_prog_data_compare(const struct brw_vec4_prog_data *a,
                                 const struct brw_vec4_prog_data *b);
 void brw_vec4_prog_data_free(const struct brw_vec4_prog_data *prog_data);
@@ -97,7 +96,7 @@ swizzle_for_size(int size);
 class reg
 {
 public:
-   /** Register file: ARF, GRF, MRF, IMM. */
+   /** Register file: GRF, MRF, IMM. */
    enum register_file file;
    /** virtual register number.  0 = fixed hw reg */
    int reg;
@@ -168,7 +167,7 @@ with_writemask(dst_reg const &r, int mask);
 
 class vec4_instruction : public backend_instruction {
 public:
-   DECLARE_RZALLOC_CXX_OPERATORS(vec4_instruction)
+   DECLARE_RALLOC_CXX_OPERATORS(vec4_instruction)
 
    vec4_instruction(vec4_visitor *v, enum opcode opcode,
                    dst_reg dst = dst_reg(),
@@ -232,7 +231,11 @@ public:
                struct gl_shader_program *shader_prog,
                struct brw_shader *shader,
                void *mem_ctx,
-                bool debug_flag);
+                bool debug_flag,
+                bool no_spills,
+                shader_time_shader_type st_base,
+                shader_time_shader_type st_written,
+                shader_time_shader_type st_reset);
    ~vec4_visitor();
 
    dst_reg dst_null_f()
@@ -245,7 +248,11 @@ public:
       return dst_reg(retype(brw_null_reg(), BRW_REGISTER_TYPE_D));
    }
 
-   struct gl_program *prog;
+   dst_reg dst_null_ud()
+   {
+      return dst_reg(retype(brw_null_reg(), BRW_REGISTER_TYPE_UD));
+   }
+
    struct brw_vec4_compile *c;
    const struct brw_vec4_prog_key *key;
    struct brw_vec4_prog_data *prog_data;
@@ -349,6 +356,7 @@ public:
    void split_uniform_registers();
    void pack_uniform_registers();
    void calculate_live_intervals();
+   void invalidate_live_intervals();
    void split_virtual_grfs();
    bool dead_code_eliminate();
    bool virtual_grf_interferes(int a, int b);
@@ -414,6 +422,8 @@ public:
    vec4_instruction *FBL(dst_reg dst, src_reg value);
    vec4_instruction *CBIT(dst_reg dst, src_reg value);
    vec4_instruction *MAD(dst_reg dst, src_reg c, src_reg b, src_reg a);
+   vec4_instruction *ADDC(dst_reg dst, src_reg src0, src_reg src1);
+   vec4_instruction *SUBB(dst_reg dst, src_reg src0, src_reg src1);
 
    int implied_mrf_writes(vec4_instruction *inst);
 
@@ -470,6 +480,8 @@ public:
    void emit_pack_half_2x16(dst_reg dst, src_reg src0);
    void emit_unpack_half_2x16(dst_reg dst, src_reg src0);
 
+   uint32_t gather_channel(ir_texture *ir, int sampler);
+   src_reg emit_mcs_fetch(ir_texture *ir, src_reg coordinate, int sampler);
    void swizzle_result(ir_texture *ir, src_reg orig_val, int sampler);
 
    void emit_ndc_computation();
@@ -483,6 +495,13 @@ public:
    void emit_shader_time_write(enum shader_time_shader_type type,
                                src_reg value);
 
+   void emit_untyped_atomic(unsigned atomic_op, unsigned surf_index,
+                            dst_reg dst, src_reg offset, src_reg src0,
+                            src_reg src1);
+
+   void emit_untyped_surface_read(unsigned surf_index, dst_reg dst,
+                                  src_reg offset);
+
    src_reg get_scratch_offset(vec4_instruction *inst,
                              src_reg *reladdr, int reg_offset);
    src_reg get_pull_constant_offset(vec4_instruction *inst,
@@ -508,9 +527,12 @@ public:
 
    void dump_instruction(backend_instruction *inst);
 
+   void visit_atomic_counter_intrinsic(ir_call *ir);
+
 protected:
    void emit_vertex();
-   void lower_attributes_to_hw_regs(const int *attribute_map);
+   void lower_attributes_to_hw_regs(const int *attribute_map,
+                                    bool interleaved);
    void setup_payload_interference(struct ra_graph *g, int first_payload_node,
                                    int reg_node_count);
    virtual dst_reg *make_reg_for_system_value(ir_variable *ir) = 0;
@@ -523,6 +545,16 @@ protected:
    virtual int compute_array_stride(ir_dereference_array *ir);
 
    const bool debug_flag;
+
+private:
+   /**
+    * If true, then register allocation should fail instead of spilling.
+    */
+   const bool no_spills;
+
+   const shader_time_shader_type st_base;
+   const shader_time_shader_type st_written;
+   const shader_time_shader_type st_reset;
 };
 
 
@@ -604,6 +636,15 @@ private:
    void generate_unpack_flags(vec4_instruction *inst,
                               struct brw_reg dst);
 
+   void generate_untyped_atomic(vec4_instruction *inst,
+                                struct brw_reg dst,
+                                struct brw_reg atomic_op,
+                                struct brw_reg surf_index);
+
+   void generate_untyped_surface_read(vec4_instruction *inst,
+                                      struct brw_reg dst,
+                                      struct brw_reg surf_index);
+
    void mark_surface_used(unsigned surf_index);
 
    struct brw_context *brw;
@@ -620,6 +661,66 @@ private:
    const bool debug_flag;
 };
 
+/**
+ * The vertex shader code generator.
+ *
+ * Translates VS IR to actual i965 assembly code.
+ */
+class gen8_vec4_generator : public gen8_generator
+{
+public:
+   gen8_vec4_generator(struct brw_context *brw,
+                       struct gl_shader_program *shader_prog,
+                       struct gl_program *prog,
+                       struct brw_vec4_prog_data *prog_data,
+                       void *mem_ctx,
+                       bool debug_flag);
+   ~gen8_vec4_generator();
+
+   const unsigned *generate_assembly(exec_list *insts, unsigned *asm_size);
+
+private:
+   void generate_code(exec_list *instructions);
+   void generate_vec4_instruction(vec4_instruction *inst,
+                                  struct brw_reg dst,
+                                  struct brw_reg *src);
+
+   void generate_tex(vec4_instruction *inst,
+                     struct brw_reg dst);
+
+   void generate_urb_write(vec4_instruction *ir, bool copy_g0);
+   void generate_gs_thread_end(vec4_instruction *ir);
+   void generate_gs_set_write_offset(struct brw_reg dst,
+                                     struct brw_reg src0,
+                                     struct brw_reg src1);
+   void generate_gs_set_vertex_count(struct brw_reg dst,
+                                     struct brw_reg src);
+   void generate_gs_set_dword_2_immed(struct brw_reg dst, struct brw_reg src);
+   void generate_gs_prepare_channel_masks(struct brw_reg dst);
+   void generate_gs_set_channel_masks(struct brw_reg dst, struct brw_reg src);
+
+   void generate_oword_dual_block_offsets(struct brw_reg m1,
+                                          struct brw_reg index);
+   void generate_scratch_write(vec4_instruction *inst,
+                               struct brw_reg dst,
+                               struct brw_reg src,
+                               struct brw_reg index);
+   void generate_scratch_read(vec4_instruction *inst,
+                              struct brw_reg dst,
+                              struct brw_reg index);
+   void generate_pull_constant_load(vec4_instruction *inst,
+                                    struct brw_reg dst,
+                                    struct brw_reg index,
+                                    struct brw_reg offset);
+
+   void mark_surface_used(unsigned surf_index);
+
+   struct brw_vec4_prog_data *prog_data;
+
+   const bool debug_flag;
+};
+
+
 } /* namespace brw */
 #endif /* __cplusplus */