#include "brw_program.h"
#ifdef __cplusplus
+#include "brw_ir_vec4.h"
+
extern "C" {
#endif
#endif
#include "glsl/ir.h"
+#include "glsl/nir/nir.h"
-struct brw_vec4_compile {
- GLuint last_scratch; /**< measured in 32-byte (register size) units */
-};
-
#ifdef __cplusplus
extern "C" {
#endif
-void
-brw_vue_setup_prog_key_for_precompile(struct gl_context *ctx,
- struct brw_vue_prog_key *key,
- GLuint id, struct gl_program *prog);
+const unsigned *
+brw_vec4_generate_assembly(const struct brw_compiler *compiler,
+ void *log_data,
+ void *mem_ctx,
+ const nir_shader *nir,
+ struct brw_vue_prog_data *prog_data,
+ const struct cfg_t *cfg,
+ unsigned *out_assembly_size);
#ifdef __cplusplus
} /* extern "C" */
namespace brw {
-class dst_reg;
-
class vec4_live_variables;
-unsigned
-swizzle_for_size(int size);
-
-class src_reg : public backend_reg
-{
-public:
- DECLARE_RALLOC_CXX_OPERATORS(src_reg)
-
- void init();
-
- src_reg(register_file file, int reg, const glsl_type *type);
- src_reg();
- src_reg(float f);
- src_reg(uint32_t u);
- src_reg(int32_t i);
- src_reg(uint8_t vf[4]);
- src_reg(uint8_t vf0, uint8_t vf1, uint8_t vf2, uint8_t vf3);
- src_reg(struct brw_reg reg);
-
- bool equals(const src_reg &r) const;
-
- src_reg(class vec4_visitor *v, const struct glsl_type *type);
- src_reg(class vec4_visitor *v, const struct glsl_type *type, int size);
-
- explicit src_reg(dst_reg reg);
-
- GLuint swizzle; /**< BRW_SWIZZLE_XYZW macros from brw_reg.h. */
-
- src_reg *reladdr;
-};
-
-static inline src_reg
-retype(src_reg reg, enum brw_reg_type type)
-{
- reg.fixed_hw_reg.type = reg.type = type;
- return reg;
-}
-
-static inline src_reg
-offset(src_reg reg, unsigned delta)
-{
- assert(delta == 0 || (reg.file != HW_REG && reg.file != IMM));
- reg.reg_offset += delta;
- return reg;
-}
-
-/**
- * Reswizzle a given source register.
- * \sa brw_swizzle().
- */
-static inline src_reg
-swizzle(src_reg reg, unsigned swizzle)
-{
- assert(reg.file != HW_REG);
- reg.swizzle = BRW_SWIZZLE4(
- BRW_GET_SWZ(reg.swizzle, BRW_GET_SWZ(swizzle, 0)),
- BRW_GET_SWZ(reg.swizzle, BRW_GET_SWZ(swizzle, 1)),
- BRW_GET_SWZ(reg.swizzle, BRW_GET_SWZ(swizzle, 2)),
- BRW_GET_SWZ(reg.swizzle, BRW_GET_SWZ(swizzle, 3)));
- return reg;
-}
-
-static inline src_reg
-negate(src_reg reg)
-{
- assert(reg.file != HW_REG && reg.file != IMM);
- reg.negate = !reg.negate;
- return reg;
-}
-
-class dst_reg : public backend_reg
-{
-public:
- DECLARE_RALLOC_CXX_OPERATORS(dst_reg)
-
- void init();
-
- dst_reg();
- dst_reg(register_file file, int reg);
- dst_reg(register_file file, int reg, const glsl_type *type, int writemask);
- dst_reg(struct brw_reg reg);
- dst_reg(class vec4_visitor *v, const struct glsl_type *type);
-
- explicit dst_reg(src_reg reg);
-
- int writemask; /**< Bitfield of WRITEMASK_[XYZW] */
-
- src_reg *reladdr;
-};
-
-static inline dst_reg
-retype(dst_reg reg, enum brw_reg_type type)
-{
- reg.fixed_hw_reg.type = reg.type = type;
- return reg;
-}
-
-static inline dst_reg
-offset(dst_reg reg, unsigned delta)
-{
- assert(delta == 0 || (reg.file != HW_REG && reg.file != IMM));
- reg.reg_offset += delta;
- return reg;
-}
-
-static inline dst_reg
-writemask(dst_reg reg, unsigned mask)
-{
- assert(reg.file != HW_REG && reg.file != IMM);
- assert((reg.writemask & mask) != 0);
- reg.writemask &= mask;
- return reg;
-}
-
-class vec4_instruction : public backend_instruction {
-public:
- DECLARE_RALLOC_CXX_OPERATORS(vec4_instruction)
-
- vec4_instruction(vec4_visitor *v, enum opcode opcode,
- const dst_reg &dst = dst_reg(),
- const src_reg &src0 = src_reg(),
- const src_reg &src1 = src_reg(),
- const src_reg &src2 = src_reg());
-
- struct brw_reg get_dst(void);
- struct brw_reg get_src(const struct brw_vue_prog_data *prog_data, int i);
-
- dst_reg dst;
- src_reg src[3];
-
- enum brw_urb_write_flags urb_write_flags;
-
- unsigned sol_binding; /**< gen6: SOL binding table index */
- bool sol_final_write; /**< gen6: send commit message */
- unsigned sol_vertex; /**< gen6: used for setting dst index in SVB header */
-
- bool is_send_from_grf();
- bool can_reswizzle(int dst_writemask, int swizzle, int swizzle_mask);
- void reswizzle(int dst_writemask, int swizzle);
- bool can_do_source_mods(struct brw_context *brw);
-
- bool reads_flag()
- {
- return predicate || opcode == VS_OPCODE_UNPACK_FLAGS_SIMD4X2;
- }
-
- bool writes_flag()
- {
- return (conditional_mod && (opcode != BRW_OPCODE_SEL &&
- opcode != BRW_OPCODE_IF &&
- opcode != BRW_OPCODE_WHILE));
- }
-};
-
/**
* The vertex shader front-end.
*
* Translates either GLSL IR or Mesa IR (for ARB_vertex_program and
* fixed-function) into VS IR.
*/
-class vec4_visitor : public backend_visitor
+class vec4_visitor : public backend_shader
{
public:
- vec4_visitor(struct brw_context *brw,
- struct brw_vec4_compile *c,
- struct gl_program *prog,
- const struct brw_vue_prog_key *key,
+ vec4_visitor(const struct brw_compiler *compiler,
+ void *log_data,
+ const struct brw_sampler_prog_key_data *key,
struct brw_vue_prog_data *prog_data,
- struct gl_shader_program *shader_prog,
- gl_shader_stage stage,
+ const nir_shader *shader,
void *mem_ctx,
- 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();
+ int shader_time_index);
+ virtual ~vec4_visitor();
dst_reg dst_null_f()
{
return dst_reg(retype(brw_null_reg(), BRW_REGISTER_TYPE_UD));
}
- struct brw_vec4_compile * const c;
- const struct brw_vue_prog_key * const key;
+ const struct brw_sampler_prog_key_data * const key_tex;
struct brw_vue_prog_data * const prog_data;
- unsigned int sanity_param_count;
-
char *fail_msg;
bool failed;
const void *base_ir;
const char *current_annotation;
- int *virtual_grf_sizes;
- int virtual_grf_count;
- int virtual_grf_array_size;
int first_non_payload_grf;
unsigned int max_grf;
int *virtual_grf_start;
brw::vec4_live_variables *live_intervals;
dst_reg userplane[MAX_CLIP_PLANES];
- /**
- * This is the size to be used for an array with an element per
- * reg_offset
- */
- int virtual_grf_reg_count;
- /** Per-virtual-grf indices into an array of size virtual_grf_reg_count */
- int *virtual_grf_reg_map;
-
- dst_reg *variable_storage(ir_variable *var);
-
- void reladdr_to_temp(ir_instruction *ir, src_reg *reg, int *num_reladdr);
-
bool need_all_constants_in_pull_buffer;
- /**
- * \name Visit methods
- *
- * As typical for the visitor pattern, there must be one \c visit method for
- * each concrete subclass of \c ir_instruction. Virtual base classes within
- * the hierarchy should not have \c visit methods.
- */
- /*@{*/
- virtual void visit(ir_variable *);
- virtual void visit(ir_loop *);
- virtual void visit(ir_loop_jump *);
- virtual void visit(ir_function_signature *);
- virtual void visit(ir_function *);
- virtual void visit(ir_expression *);
- virtual void visit(ir_swizzle *);
- virtual void visit(ir_dereference_variable *);
- virtual void visit(ir_dereference_array *);
- virtual void visit(ir_dereference_record *);
- virtual void visit(ir_assignment *);
- virtual void visit(ir_constant *);
- virtual void visit(ir_call *);
- virtual void visit(ir_return *);
- virtual void visit(ir_discard *);
- virtual void visit(ir_texture *);
- virtual void visit(ir_if *);
- virtual void visit(ir_emit_vertex *);
- virtual void visit(ir_end_primitive *);
- /*@}*/
-
- src_reg result;
-
/* Regs for vertex results. Generated at ir_variable visiting time
* for the ir->location's used.
*/
dst_reg output_reg[BRW_VARYING_SLOT_COUNT];
const char *output_reg_annotation[BRW_VARYING_SLOT_COUNT];
int *uniform_size;
- int *uniform_vector_size;
- int uniform_array_size; /*< Size of uniform_[vector_]size arrays */
+ int uniform_array_size; /*< Size of the uniform_size array */
int uniforms;
src_reg shader_start_time;
- struct hash_table *variable_ht;
-
- bool run(void);
+ bool run();
void fail(const char *msg, ...);
- int virtual_grf_alloc(int size);
- void setup_uniform_clipplane_values();
- void setup_uniform_values(ir_variable *ir);
- void setup_builtin_uniform_values(ir_variable *ir);
int setup_uniforms(int payload_reg);
+
bool reg_allocate_trivial();
bool reg_allocate();
void evaluate_spill_costs(float *spill_costs, bool *no_spill);
bool opt_vector_float();
bool opt_reduce_swizzle();
bool dead_code_eliminate();
+ int var_range_start(unsigned v, unsigned n) const;
+ int var_range_end(unsigned v, unsigned n) const;
bool virtual_grf_interferes(int a, int b);
- bool opt_copy_propagation();
+ bool opt_cmod_propagation();
+ bool opt_copy_propagation(bool do_constant_prop = true);
bool opt_cse_local(bblock_t *block);
bool opt_cse();
bool opt_algebraic();
bool opt_register_coalesce();
+ bool eliminate_find_live_channel();
bool is_dep_ctrl_unsafe(const vec4_instruction *inst);
void opt_set_dependency_control();
void opt_schedule_instructions();
+ void convert_to_hw_regs();
vec4_instruction *emit(vec4_instruction *inst);
vec4_instruction *IF(src_reg src0, src_reg src1,
enum brw_conditional_mod condition);
vec4_instruction *IF(enum brw_predicate predicate);
- EMIT1(PULL_CONSTANT_LOAD)
EMIT1(SCRATCH_READ)
EMIT2(SCRATCH_WRITE)
EMIT3(LRP)
int implied_mrf_writes(vec4_instruction *inst);
- bool try_rewrite_rhs_to_dst(ir_assignment *ir,
- dst_reg dst,
- src_reg src,
- vec4_instruction *pre_rhs_inst,
- vec4_instruction *last_rhs_inst);
-
- /** Walks an exec_list of ir_instruction and sends it through this visitor. */
- void visit_instructions(const exec_list *list);
-
- void emit_vp_sop(enum brw_conditional_mod condmod, dst_reg dst,
- src_reg src0, src_reg src1, src_reg one);
-
- void emit_bool_to_cond_code(ir_rvalue *ir, enum brw_predicate *predicate);
- void emit_if_gen6(ir_if *ir);
-
- void emit_minmax(enum brw_conditional_mod conditionalmod, dst_reg dst,
- src_reg src0, src_reg src1);
-
- void emit_lrp(const dst_reg &dst,
- const src_reg &x, const src_reg &y, const src_reg &a);
+ vec4_instruction *emit_minmax(enum brw_conditional_mod conditionalmod, dst_reg dst,
+ src_reg src0, src_reg src1);
- void emit_block_move(dst_reg *dst, src_reg *src,
- const struct glsl_type *type, brw_predicate predicate);
-
- void emit_constant_values(dst_reg *dst, ir_constant *value);
+ vec4_instruction *emit_lrp(const dst_reg &dst, const src_reg &x,
+ const src_reg &y, const src_reg &a);
/**
- * Emit the correct dot-product instruction for the type of arguments
+ * Copy any live channel from \p src to the first channel of the
+ * result.
*/
- void emit_dp(dst_reg dst, src_reg src0, src_reg src1, unsigned elements);
-
- void emit_scalar(ir_instruction *ir, enum prog_opcode op,
- dst_reg dst, src_reg src0);
-
- void emit_scalar(ir_instruction *ir, enum prog_opcode op,
- dst_reg dst, src_reg src0, src_reg src1);
+ src_reg emit_uniformize(const src_reg &src);
- void emit_scs(ir_instruction *ir, enum prog_opcode op,
- dst_reg dst, const src_reg &src);
+ src_reg fix_3src_operand(const src_reg &src);
+ src_reg resolve_source_modifiers(const src_reg &src);
- src_reg fix_3src_operand(src_reg src);
+ vec4_instruction *emit_math(enum opcode opcode, const dst_reg &dst, const src_reg &src0,
+ const src_reg &src1 = src_reg());
- void emit_math(enum opcode opcode, const dst_reg &dst, const src_reg &src0,
- const src_reg &src1 = src_reg());
- src_reg fix_math_operand(src_reg src);
+ src_reg fix_math_operand(const src_reg &src);
void emit_pack_half_2x16(dst_reg dst, src_reg src0);
void emit_unpack_half_2x16(dst_reg dst, src_reg src0);
void emit_pack_unorm_4x8(const dst_reg &dst, const src_reg &src0);
void emit_pack_snorm_4x8(const dst_reg &dst, const src_reg &src0);
- uint32_t gather_channel(ir_texture *ir, uint32_t sampler);
- src_reg emit_mcs_fetch(ir_texture *ir, src_reg coordinate, src_reg sampler);
+ void emit_texture(ir_texture_opcode op,
+ dst_reg dest,
+ const glsl_type *dest_type,
+ src_reg coordinate,
+ int coord_components,
+ src_reg shadow_comparitor,
+ src_reg lod, src_reg lod2,
+ src_reg sample_index,
+ uint32_t constant_offset,
+ src_reg offset_value,
+ src_reg mcs,
+ bool is_cube_array,
+ uint32_t surface, src_reg surface_reg,
+ uint32_t sampler, src_reg sampler_reg);
+
+ src_reg emit_mcs_fetch(const glsl_type *coordinate_type, src_reg coordinate,
+ src_reg sampler);
void emit_gen6_gather_wa(uint8_t wa, dst_reg dst);
- void swizzle_result(ir_texture *ir, src_reg orig_val, uint32_t sampler);
void emit_ndc_computation();
void emit_psiz_and_flags(dst_reg reg);
- void emit_clip_distances(dst_reg reg, int offset);
vec4_instruction *emit_generic_urb_slot(dst_reg reg, int varying);
- void emit_urb_slot(dst_reg reg, int varying);
+ virtual void emit_urb_slot(dst_reg reg, int varying);
void emit_shader_time_begin();
void emit_shader_time_end();
- void emit_shader_time_write(enum shader_time_shader_type type,
- src_reg value);
+ void emit_shader_time_write(int shader_time_subindex, src_reg value);
void emit_untyped_atomic(unsigned atomic_op, unsigned surf_index,
dst_reg dst, src_reg offset, src_reg src0,
dst_reg dst,
src_reg orig_src,
int base_offset);
+ void emit_pull_constant_load_reg(dst_reg dst,
+ src_reg surf_index,
+ src_reg offset,
+ bblock_t *before_block,
+ vec4_instruction *before_inst);
+ src_reg emit_resolve_reladdr(int scratch_loc[], bblock_t *block,
+ vec4_instruction *inst, src_reg src);
- bool try_emit_mad(ir_expression *ir);
- bool try_emit_b2f_of_compare(ir_expression *ir);
void resolve_ud_negate(src_reg *reg);
- void resolve_bool_comparison(ir_rvalue *rvalue, src_reg *reg);
src_reg get_timestamp();
- bool process_move_condition(ir_rvalue *ir);
-
void dump_instruction(backend_instruction *inst);
void dump_instruction(backend_instruction *inst, FILE *file);
- void visit_atomic_counter_intrinsic(ir_call *ir);
+ bool is_high_sampler(src_reg sampler);
+
+ virtual void emit_nir_code();
+ virtual void nir_setup_inputs();
+ virtual void nir_setup_uniforms();
+ virtual void nir_setup_system_value_intrinsic(nir_intrinsic_instr *instr);
+ virtual void nir_setup_system_values();
+ virtual void nir_emit_impl(nir_function_impl *impl);
+ virtual void nir_emit_cf_list(exec_list *list);
+ virtual void nir_emit_if(nir_if *if_stmt);
+ virtual void nir_emit_loop(nir_loop *loop);
+ virtual void nir_emit_block(nir_block *block);
+ virtual void nir_emit_instr(nir_instr *instr);
+ virtual void nir_emit_load_const(nir_load_const_instr *instr);
+ virtual void nir_emit_intrinsic(nir_intrinsic_instr *instr);
+ virtual void nir_emit_alu(nir_alu_instr *instr);
+ virtual void nir_emit_jump(nir_jump_instr *instr);
+ virtual void nir_emit_texture(nir_tex_instr *instr);
+ virtual void nir_emit_undef(nir_ssa_undef_instr *instr);
+ virtual void nir_emit_ssbo_atomic(int op, nir_intrinsic_instr *instr);
+
+ dst_reg get_nir_dest(nir_dest dest, enum brw_reg_type type);
+ dst_reg get_nir_dest(nir_dest dest, nir_alu_type type);
+ dst_reg get_nir_dest(nir_dest dest);
+ src_reg get_nir_src(nir_src src, enum brw_reg_type type,
+ unsigned num_components = 4);
+ src_reg get_nir_src(nir_src src, nir_alu_type type,
+ unsigned num_components = 4);
+ src_reg get_nir_src(nir_src src,
+ unsigned num_components = 4);
+
+ virtual dst_reg *make_reg_for_system_value(int location,
+ const glsl_type *type) = 0;
+
+ dst_reg *nir_locals;
+ dst_reg *nir_ssa_values;
+ src_reg *nir_inputs;
+ dst_reg *nir_system_values;
protected:
void emit_vertex();
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;
- virtual void assign_binding_table_offsets();
virtual void setup_payload() = 0;
virtual void emit_prolog() = 0;
- virtual void emit_program_code() = 0;
virtual void emit_thread_end() = 0;
virtual void emit_urb_write_header(int mrf) = 0;
virtual vec4_instruction *emit_urb_write_opcode(bool complete) = 0;
- virtual int compute_array_stride(ir_dereference_array *ir);
-
- const bool debug_flag;
+ virtual void gs_emit_vertex(int stream_id);
+ virtual void gs_end_primitive();
private:
/**
*/
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;
-};
-
-
-/**
- * The vertex shader code generator.
- *
- * Translates VS IR to actual i965 assembly code.
- */
-class vec4_generator
-{
-public:
- vec4_generator(struct brw_context *brw,
- struct gl_shader_program *shader_prog,
- struct gl_program *prog,
- struct brw_vue_prog_data *prog_data,
- void *mem_ctx,
- bool debug_flag);
- ~vec4_generator();
+ int shader_time_index;
- const unsigned *generate_assembly(const cfg_t *cfg, unsigned *asm_size);
-
-private:
- void generate_code(const cfg_t *cfg);
-
- void generate_math1_gen4(vec4_instruction *inst,
- struct brw_reg dst,
- struct brw_reg src);
- void generate_math2_gen4(vec4_instruction *inst,
- struct brw_reg dst,
- struct brw_reg src0,
- struct brw_reg src1);
- void generate_math_gen6(vec4_instruction *inst,
- struct brw_reg dst,
- struct brw_reg src0,
- struct brw_reg src1);
-
- void generate_tex(vec4_instruction *inst,
- struct brw_reg dst,
- struct brw_reg src,
- struct brw_reg sampler_index);
-
- void generate_vs_urb_write(vec4_instruction *inst);
- void generate_gs_urb_write(vec4_instruction *inst);
- void generate_gs_urb_write_allocate(vec4_instruction *inst);
- void generate_gs_thread_end(vec4_instruction *inst);
- 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_svb_write(vec4_instruction *inst,
- struct brw_reg dst,
- struct brw_reg src0,
- struct brw_reg src1);
- void generate_gs_svb_set_destination_index(vec4_instruction *inst,
- struct brw_reg dst,
- struct brw_reg src);
- void generate_gs_set_dword_2(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_gs_get_instance_id(struct brw_reg dst);
- void generate_gs_ff_sync_set_primitives(struct brw_reg dst,
- struct brw_reg src0,
- struct brw_reg src1,
- struct brw_reg src2);
- void generate_gs_ff_sync(vec4_instruction *inst,
- struct brw_reg dst,
- struct brw_reg src0,
- struct brw_reg src1);
- void generate_gs_set_primitive_id(struct brw_reg dst);
- 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 generate_pull_constant_load_gen7(vec4_instruction *inst,
- struct brw_reg dst,
- struct brw_reg surf_index,
- struct brw_reg offset);
- 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);
-
- struct brw_context *brw;
-
- struct brw_compile *p;
-
- struct gl_shader_program *shader_prog;
- const struct gl_program *prog;
-
- struct brw_vue_prog_data *prog_data;
-
- void *mem_ctx;
- const bool debug_flag;
+ unsigned last_scratch; /**< measured in 32-byte (register size) units */
};
} /* namespace brw */
projects / mesa.git / blobdiff