#include "brw_program.h"
#ifdef __cplusplus
+#include "brw_ir_vec4.h"
+
extern "C" {
#endif
GLuint last_scratch; /**< measured in 32-byte (register size) units */
};
-
-struct brw_vec4_prog_key {
- GLuint program_string_id;
-
- /**
- * True if at least one clip flag is enabled, regardless of whether the
- * shader uses clip planes or gl_ClipDistance.
- */
- GLuint userclip_active:1;
-
- /**
- * How many user clipping planes are being uploaded to the vertex shader as
- * push constants.
- */
- GLuint nr_userclip_plane_consts:4;
-
- GLuint clamp_vertex_color:1;
-
- struct brw_sampler_prog_key_data tex;
-};
-
-
#ifdef __cplusplus
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);
+brw_vue_setup_prog_key_for_precompile(struct gl_context *ctx,
+ struct brw_vue_prog_key *key,
+ GLuint id, struct gl_program *prog);
#ifdef __cplusplus
} /* extern "C" */
namespace brw {
-class dst_reg;
-
-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(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_vec4_prog_data *prog_data, int i);
-
- dst_reg dst;
- src_reg src[3];
-
- bool shadow_compare;
-
- enum brw_urb_write_flags urb_write_flags;
- bool header_present;
-
- 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;
- }
-};
+class vec4_live_variables;
/**
* The vertex shader front-end.
vec4_visitor(struct brw_context *brw,
struct brw_vec4_compile *c,
struct gl_program *prog,
- const struct brw_vec4_prog_key *key,
- struct brw_vec4_prog_data *prog_data,
+ const struct brw_vue_prog_key *key,
+ struct brw_vue_prog_data *prog_data,
struct gl_shader_program *shader_prog,
gl_shader_stage stage,
void *mem_ctx,
}
struct brw_vec4_compile * const c;
- const struct brw_vec4_prog_key * const key;
- struct brw_vec4_prog_data * const prog_data;
+ const struct brw_vue_prog_key * const key;
+ struct brw_vue_prog_data * const prog_data;
unsigned int sanity_param_count;
char *fail_msg;
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;
int *virtual_grf_end;
+ 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;
-
- bool live_intervals_valid;
-
dst_reg *variable_storage(ir_variable *var);
void reladdr_to_temp(ir_instruction *ir, src_reg *reg, int *num_reladdr);
bool run(void);
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);
void calculate_live_intervals();
void invalidate_live_intervals();
void split_virtual_grfs();
+ bool opt_vector_float();
bool opt_reduce_swizzle();
bool dead_code_eliminate();
bool virtual_grf_interferes(int a, int b);
- bool opt_copy_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 is_dep_ctrl_unsafe(const vec4_instruction *inst);
void opt_set_dependency_control();
void opt_schedule_instructions();
vec4_instruction *emit(vec4_instruction *inst);
vec4_instruction *emit(enum opcode opcode);
-
- vec4_instruction *emit(enum opcode opcode, dst_reg dst);
-
- vec4_instruction *emit(enum opcode opcode, dst_reg dst, src_reg src0);
-
- vec4_instruction *emit(enum opcode opcode, dst_reg dst,
- src_reg src0, src_reg src1);
-
- vec4_instruction *emit(enum opcode opcode, dst_reg dst,
- src_reg src0, src_reg src1, src_reg src2);
+ vec4_instruction *emit(enum opcode opcode, const dst_reg &dst);
+ vec4_instruction *emit(enum opcode opcode, const dst_reg &dst,
+ const src_reg &src0);
+ vec4_instruction *emit(enum opcode opcode, const dst_reg &dst,
+ const src_reg &src0, const src_reg &src1);
+ vec4_instruction *emit(enum opcode opcode, const dst_reg &dst,
+ const src_reg &src0, const src_reg &src1,
+ const src_reg &src2);
vec4_instruction *emit_before(bblock_t *block,
vec4_instruction *inst,
vec4_instruction *new_inst);
- vec4_instruction *MOV(const dst_reg &dst, const src_reg &src0);
- vec4_instruction *NOT(const dst_reg &dst, const src_reg &src0);
- vec4_instruction *RNDD(const dst_reg &dst, const src_reg &src0);
- vec4_instruction *RNDE(const dst_reg &dst, const src_reg &src0);
- vec4_instruction *RNDZ(const dst_reg &dst, const src_reg &src0);
- vec4_instruction *FRC(const dst_reg &dst, const src_reg &src0);
- vec4_instruction *F32TO16(const dst_reg &dst, const src_reg &src0);
- vec4_instruction *F16TO32(const dst_reg &dst, const src_reg &src0);
- vec4_instruction *ADD(const dst_reg &dst, const src_reg &src0,
- const src_reg &src1);
- vec4_instruction *MUL(const dst_reg &dst, const src_reg &src0,
- const src_reg &src1);
- vec4_instruction *MACH(const dst_reg &dst, const src_reg &src0,
- const src_reg &src1);
- vec4_instruction *MAC(const dst_reg &dst, const src_reg &src0,
- const src_reg &src1);
- vec4_instruction *AND(const dst_reg &dst, const src_reg &src0,
- const src_reg &src1);
- vec4_instruction *OR(const dst_reg &dst, const src_reg &src0,
- const src_reg &src1);
- vec4_instruction *XOR(const dst_reg &dst, const src_reg &src0,
- const src_reg &src1);
- vec4_instruction *DP3(const dst_reg &dst, const src_reg &src0,
- const src_reg &src1);
- vec4_instruction *DP4(const dst_reg &dst, const src_reg &src0,
- const src_reg &src1);
- vec4_instruction *DPH(const dst_reg &dst, const src_reg &src0,
- const src_reg &src1);
- vec4_instruction *SHL(const dst_reg &dst, const src_reg &src0,
- const src_reg &src1);
- vec4_instruction *SHR(const dst_reg &dst, const src_reg &src0,
- const src_reg &src1);
- vec4_instruction *ASR(const dst_reg &dst, const src_reg &src0,
- const src_reg &src1);
+#define EMIT1(op) vec4_instruction *op(const dst_reg &, const src_reg &);
+#define EMIT2(op) vec4_instruction *op(const dst_reg &, const src_reg &, const src_reg &);
+#define EMIT3(op) vec4_instruction *op(const dst_reg &, const src_reg &, const src_reg &, const src_reg &);
+ EMIT1(MOV)
+ EMIT1(NOT)
+ EMIT1(RNDD)
+ EMIT1(RNDE)
+ EMIT1(RNDZ)
+ EMIT1(FRC)
+ EMIT1(F32TO16)
+ EMIT1(F16TO32)
+ EMIT2(ADD)
+ EMIT2(MUL)
+ EMIT2(MACH)
+ EMIT2(MAC)
+ EMIT2(AND)
+ EMIT2(OR)
+ EMIT2(XOR)
+ EMIT2(DP3)
+ EMIT2(DP4)
+ EMIT2(DPH)
+ EMIT2(SHL)
+ EMIT2(SHR)
+ EMIT2(ASR)
vec4_instruction *CMP(dst_reg dst, src_reg src0, src_reg src1,
enum brw_conditional_mod condition);
vec4_instruction *IF(src_reg src0, src_reg src1,
enum brw_conditional_mod condition);
vec4_instruction *IF(enum brw_predicate predicate);
- vec4_instruction *PULL_CONSTANT_LOAD(const dst_reg &dst,
- const src_reg &index);
- vec4_instruction *SCRATCH_READ(const dst_reg &dst, const src_reg &index);
- vec4_instruction *SCRATCH_WRITE(const dst_reg &dst, const src_reg &src,
- const src_reg &index);
- vec4_instruction *LRP(const dst_reg &dst, const src_reg &a,
- const src_reg &y, const src_reg &x);
- vec4_instruction *BFREV(const dst_reg &dst, const src_reg &value);
- vec4_instruction *BFE(const dst_reg &dst, const src_reg &bits,
- const src_reg &offset, const src_reg &value);
- vec4_instruction *BFI1(const dst_reg &dst, const src_reg &bits,
- const src_reg &offset);
- vec4_instruction *BFI2(const dst_reg &dst, const src_reg &bfi1_dst,
- const src_reg &insert, const src_reg &base);
- vec4_instruction *FBH(const dst_reg &dst, const src_reg &value);
- vec4_instruction *FBL(const dst_reg &dst, const src_reg &value);
- vec4_instruction *CBIT(const dst_reg &dst, const src_reg &value);
- vec4_instruction *MAD(const dst_reg &dst, const src_reg &c,
- const src_reg &b, const src_reg &a);
- vec4_instruction *ADDC(const dst_reg &dst, const src_reg &src0,
- const src_reg &src1);
- vec4_instruction *SUBB(const dst_reg &dst, const src_reg &src0,
- const src_reg &src1);
+ EMIT1(PULL_CONSTANT_LOAD)
+ EMIT1(SCRATCH_READ)
+ EMIT2(SCRATCH_WRITE)
+ EMIT3(LRP)
+ EMIT1(BFREV)
+ EMIT3(BFE)
+ EMIT2(BFI1)
+ EMIT3(BFI2)
+ EMIT1(FBH)
+ EMIT1(FBL)
+ EMIT1(CBIT)
+ EMIT3(MAD)
+ EMIT2(ADDC)
+ EMIT2(SUBB)
+#undef EMIT1
+#undef EMIT2
+#undef EMIT3
int implied_mrf_writes(vec4_instruction *inst);
src_reg fix_3src_operand(src_reg src);
- void emit_math1_gen6(enum opcode opcode, dst_reg dst, src_reg src);
- void emit_math1_gen4(enum opcode opcode, dst_reg dst, src_reg src);
- void emit_math(enum opcode opcode, dst_reg dst, src_reg src);
- void emit_math2_gen6(enum opcode opcode, dst_reg dst, src_reg src0, src_reg src1);
- void emit_math2_gen4(enum opcode opcode, dst_reg dst, src_reg src0, src_reg src1);
- void emit_math(enum opcode opcode, dst_reg dst, src_reg src0, src_reg src1);
+ 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);
void emit_pack_half_2x16(dst_reg dst, src_reg src0);
void emit_unpack_half_2x16(dst_reg dst, src_reg src0);
+ void emit_unpack_unorm_4x8(const dst_reg &dst, src_reg src0);
+ void emit_unpack_snorm_4x8(const 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_ndc_computation();
void emit_psiz_and_flags(dst_reg reg);
void emit_clip_distances(dst_reg reg, int offset);
- void emit_generic_urb_slot(dst_reg reg, int varying);
+ vec4_instruction *emit_generic_urb_slot(dst_reg reg, int varying);
void emit_urb_slot(dst_reg reg, int varying);
void emit_shader_time_begin();
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();
vec4_generator(struct brw_context *brw,
struct gl_shader_program *shader_prog,
struct gl_program *prog,
- struct brw_vec4_prog_data *prog_data,
+ struct brw_vue_prog_data *prog_data,
void *mem_ctx,
- bool debug_flag);
+ bool debug_flag,
+ const char *stage_name,
+ const char *stage_abbrev);
~vec4_generator();
const unsigned *generate_assembly(const cfg_t *cfg, unsigned *asm_size);
struct gl_shader_program *shader_prog;
const struct gl_program *prog;
- struct brw_vec4_prog_data *prog_data;
+ struct brw_vue_prog_data *prog_data;
void *mem_ctx;
+ const char *stage_name;
+ const char *stage_abbrev;
const bool debug_flag;
};
projects / mesa.git / blobdiff