*
*/
-#pragma once
+#ifndef NIR_H
+#define NIR_H
#include "util/hash_table.h"
#include "compiler/glsl/list.h"
#include "compiler/shader_info.h"
#include <stdio.h>
+#ifdef DEBUG
+#include "util/debug.h"
+#endif /* DEBUG */
+
#include "nir_opcodes.h"
+#if defined(_WIN32) && !defined(snprintf)
+#define snprintf _snprintf
+#endif
+
#ifdef __cplusplus
extern "C" {
#endif
typedef union {
float f32[4];
double f64[4];
+ int8_t i8[4];
+ uint8_t u8[4];
+ int16_t i16[4];
+ uint16_t u16[4];
int32_t i32[4];
uint32_t u32[4];
int64_t i64[4];
*/
bool is_packed;
- /** set of nir_src's where this register is used (read from) */
+ /** set of nir_srcs where this register is used (read from) */
struct list_head uses;
- /** set of nir_dest's where this register is defined (written to) */
+ /** set of nir_dests where this register is defined (written to) */
struct list_head defs;
- /** set of nir_if's where this register is used as a condition */
+ /** set of nir_ifs where this register is used as a condition */
struct list_head if_uses;
} nir_register;
+#define nir_foreach_register(reg, reg_list) \
+ foreach_list_typed(nir_register, reg, node, reg_list)
+#define nir_foreach_register_safe(reg, reg_list) \
+ foreach_list_typed_safe(nir_register, reg, node, reg_list)
+
typedef enum {
nir_instr_type_alu,
nir_instr_type_call,
}
static inline bool
-nir_instr_is_first(nir_instr *instr)
+nir_instr_is_first(const nir_instr *instr)
{
- return exec_node_is_head_sentinel(exec_node_get_prev(&instr->node));
+ return exec_node_is_head_sentinel(exec_node_get_prev_const(&instr->node));
}
static inline bool
-nir_instr_is_last(nir_instr *instr)
+nir_instr_is_last(const nir_instr *instr)
{
- return exec_node_is_tail_sentinel(exec_node_get_next(&instr->node));
+ return exec_node_is_tail_sentinel(exec_node_get_next_const(&instr->node));
}
typedef struct nir_ssa_def {
nir_instr *parent_instr;
- /** set of nir_instr's where this register is used (read from) */
+ /** set of nir_instrs where this register is used (read from) */
struct list_head uses;
- /** set of nir_if's where this register is used as a condition */
+ /** set of nir_ifs where this register is used as a condition */
struct list_head if_uses;
uint8_t num_components;
return type & NIR_ALU_TYPE_BASE_TYPE_MASK;
}
+static inline nir_alu_type
+nir_get_nir_type_for_glsl_base_type(enum glsl_base_type base_type)
+{
+ switch (base_type) {
+ case GLSL_TYPE_BOOL:
+ return nir_type_bool32;
+ break;
+ case GLSL_TYPE_UINT:
+ return nir_type_uint32;
+ break;
+ case GLSL_TYPE_INT:
+ return nir_type_int32;
+ break;
+ case GLSL_TYPE_UINT64:
+ return nir_type_uint64;
+ break;
+ case GLSL_TYPE_INT64:
+ return nir_type_int64;
+ break;
+ case GLSL_TYPE_FLOAT:
+ return nir_type_float32;
+ break;
+ case GLSL_TYPE_DOUBLE:
+ return nir_type_float64;
+ break;
+ default:
+ unreachable("unknown type");
+ }
+}
+
+static inline nir_alu_type
+nir_get_nir_type_for_glsl_type(const struct glsl_type *type)
+{
+ return nir_get_nir_type_for_glsl_base_type(glsl_get_base_type(type));
+}
+
+nir_op nir_type_conversion_op(nir_alu_type src, nir_alu_type dst);
+
typedef enum {
NIR_OP_IS_COMMUTATIVE = (1 << 0),
NIR_OP_IS_ASSOCIATIVE = (1 << 1),
/* is this source channel used? */
static inline bool
-nir_alu_instr_channel_used(nir_alu_instr *instr, unsigned src, unsigned channel)
+nir_alu_instr_channel_used(const nir_alu_instr *instr, unsigned src,
+ unsigned channel)
{
if (nir_op_infos[instr->op].input_sizes[src] > 0)
return channel < nir_op_infos[instr->op].input_sizes[src];
#define INTRINSIC_IDX_ACCESSORS(name, flag, type) \
static inline type \
-nir_intrinsic_##name(nir_intrinsic_instr *instr) \
+nir_intrinsic_##name(const nir_intrinsic_instr *instr) \
{ \
const nir_intrinsic_info *info = &nir_intrinsic_infos[instr->intrinsic]; \
assert(info->index_map[NIR_INTRINSIC_##flag] > 0); \
} nir_tex_instr;
static inline unsigned
-nir_tex_instr_dest_size(nir_tex_instr *instr)
+nir_tex_instr_dest_size(const nir_tex_instr *instr)
{
switch (instr->op) {
case nir_texop_txs: {
* rather than actually sampling it.
*/
static inline bool
-nir_tex_instr_is_query(nir_tex_instr *instr)
+nir_tex_instr_is_query(const nir_tex_instr *instr)
{
switch (instr->op) {
case nir_texop_txs:
}
static inline nir_alu_type
-nir_tex_instr_src_type(nir_tex_instr *instr, unsigned src)
+nir_tex_instr_src_type(const nir_tex_instr *instr, unsigned src)
{
switch (instr->src[src].src_type) {
case nir_tex_src_coord:
}
static inline unsigned
-nir_tex_instr_src_size(nir_tex_instr *instr, unsigned src)
+nir_tex_instr_src_size(const nir_tex_instr *instr, unsigned src)
{
if (instr->src[src].src_type == nir_tex_src_coord)
return instr->coord_components;
}
static inline int
-nir_tex_instr_src_index(nir_tex_instr *instr, nir_tex_src_type type)
+nir_tex_instr_src_index(const nir_tex_instr *instr, nir_tex_src_type type)
{
for (unsigned i = 0; i < instr->num_srcs; i++)
if (instr->src[i].src_type == type)
typedef struct {
nir_instr instr;
- /* A list of nir_parallel_copy_entry's. The sources of all of the
+ /* A list of nir_parallel_copy_entrys. The sources of all of the
* entries are copied to the corresponding destinations "in parallel".
* In other words, if we have two entries: a -> b and b -> a, the values
* get swapped.
unsigned num_dom_children;
struct nir_block **dom_children;
- /* Set of nir_block's on the dominance frontier of this block */
+ /* Set of nir_blocks on the dominance frontier of this block */
struct set *dom_frontier;
/*
bool lower_extract_byte;
bool lower_extract_word;
+ bool lower_vote_trivial;
+
/**
* Does the driver support real 32-bit integers? (Otherwise, integers
* are simulated by floats.)
const struct nir_shader_compiler_options *options;
/** Various bits of compile-time information about a given shader */
- struct shader_info *info;
+ struct shader_info info;
/** list of global variables in the shader (nir_variable) */
struct exec_list globals;
void nir_reg_remove(nir_register *reg);
-/** Adds a variable to the appropreate list in nir_shader */
+/** Adds a variable to the appropriate list in nir_shader */
void nir_shader_add_variable(nir_shader *shader, nir_variable *var);
static inline void
nir_deref_array *nir_deref_array_create(void *mem_ctx);
nir_deref_struct *nir_deref_struct_create(void *mem_ctx, unsigned field_index);
-nir_deref *nir_copy_deref(void *mem_ctx, nir_deref *deref);
-
typedef bool (*nir_deref_foreach_leaf_cb)(nir_deref_var *deref, void *state);
bool nir_deref_foreach_leaf(nir_deref_var *deref,
nir_deref_foreach_leaf_cb cb, void *state);
nir_function_impl *nir_function_impl_clone(const nir_function_impl *fi);
nir_constant *nir_constant_clone(const nir_constant *c, nir_variable *var);
nir_variable *nir_variable_clone(const nir_variable *c, nir_shader *shader);
+nir_deref *nir_deref_clone(const nir_deref *deref, void *mem_ctx);
+nir_deref_var *nir_deref_var_clone(const nir_deref_var *deref, void *mem_ctx);
#ifdef DEBUG
void nir_validate_shader(nir_shader *shader);
void nir_metadata_set_validation_flag(nir_shader *shader);
void nir_metadata_check_validation_flag(nir_shader *shader);
-#include "util/debug.h"
static inline bool
should_clone_nir(void)
{
return should_clone;
}
+
+static inline bool
+should_print_nir(void)
+{
+ static int should_print = -1;
+ if (should_print < 0)
+ should_print = env_var_as_boolean("NIR_PRINT", false);
+
+ return should_print;
+}
#else
static inline void nir_validate_shader(nir_shader *shader) { (void) shader; }
static inline void nir_metadata_set_validation_flag(nir_shader *shader) { (void) shader; }
static inline void nir_metadata_check_validation_flag(nir_shader *shader) { (void) shader; }
static inline bool should_clone_nir(void) { return false; }
+static inline bool should_print_nir(void) { return false; }
#endif /* DEBUG */
#define _PASS(nir, do_pass) do { \
#define NIR_PASS(progress, nir, pass, ...) _PASS(nir, \
nir_metadata_set_validation_flag(nir); \
+ if (should_print_nir()) \
+ printf("%s\n", #pass); \
if (pass(nir, ##__VA_ARGS__)) { \
progress = true; \
+ if (should_print_nir()) \
+ nir_print_shader(nir, stdout); \
nir_metadata_check_validation_flag(nir); \
} \
)
#define NIR_PASS_V(nir, pass, ...) _PASS(nir, \
+ if (should_print_nir()) \
+ printf("%s\n", #pass); \
pass(nir, ##__VA_ARGS__); \
+ if (should_print_nir()) \
+ nir_print_shader(nir, stdout); \
)
void nir_calc_dominance_impl(nir_function_impl *impl);
bool nir_propagate_invariant(nir_shader *shader);
-void nir_lower_var_copy_instr(nir_intrinsic_instr *copy, void *mem_ctx);
-void nir_lower_var_copies(nir_shader *shader);
+void nir_lower_var_copy_instr(nir_intrinsic_instr *copy, nir_shader *shader);
+bool nir_lower_var_copies(nir_shader *shader);
bool nir_lower_global_vars_to_local(nir_shader *shader);
*/
nir_lower_io_force_sample_interpolation = (1 << 1),
} nir_lower_io_options;
-void nir_lower_io(nir_shader *shader,
+bool nir_lower_io(nir_shader *shader,
nir_variable_mode modes,
int (*type_size)(const struct glsl_type *),
nir_lower_io_options);
nir_src *nir_get_io_offset_src(nir_intrinsic_instr *instr);
nir_src *nir_get_io_vertex_index_src(nir_intrinsic_instr *instr);
-bool nir_is_per_vertex_io(nir_variable *var, gl_shader_stage stage);
+bool nir_is_per_vertex_io(const nir_variable *var, gl_shader_stage stage);
void nir_lower_io_types(nir_shader *shader);
-void nir_lower_vars_to_ssa(nir_shader *shader);
+bool nir_lower_regs_to_ssa_impl(nir_function_impl *impl);
+bool nir_lower_regs_to_ssa(nir_shader *shader);
+bool nir_lower_vars_to_ssa(nir_shader *shader);
bool nir_remove_dead_variables(nir_shader *shader, nir_variable_mode modes);
bool nir_lower_constant_initializers(nir_shader *shader,
nir_variable_mode modes);
-void nir_move_vec_src_uses_to_dest(nir_shader *shader);
+bool nir_move_vec_src_uses_to_dest(nir_shader *shader);
bool nir_lower_vec_to_movs(nir_shader *shader);
bool nir_lower_alu_to_scalar(nir_shader *shader);
-void nir_lower_load_const_to_scalar(nir_shader *shader);
+bool nir_lower_load_const_to_scalar(nir_shader *shader);
bool nir_lower_phis_to_scalar(nir_shader *shader);
void nir_lower_io_to_scalar(nir_shader *shader, nir_variable_mode mask);
-void nir_lower_samplers(nir_shader *shader,
+bool nir_lower_samplers(nir_shader *shader,
const struct gl_shader_program *shader_program);
bool nir_lower_system_values(nir_shader *shader);
unsigned lower_y_uv_external;
unsigned lower_y_u_v_external;
unsigned lower_yx_xuxv_external;
+ unsigned lower_xy_uxvx_external;
/**
* To emulate certain texture wrap modes, this can be used
bool nir_lower_idiv(nir_shader *shader);
-void nir_lower_clip_vs(nir_shader *shader, unsigned ucp_enables);
-void nir_lower_clip_fs(nir_shader *shader, unsigned ucp_enables);
-void nir_lower_clip_cull_distance_arrays(nir_shader *nir);
+bool nir_lower_clip_vs(nir_shader *shader, unsigned ucp_enables);
+bool nir_lower_clip_fs(nir_shader *shader, unsigned ucp_enables);
+bool nir_lower_clip_cull_distance_arrays(nir_shader *nir);
void nir_lower_two_sided_color(nir_shader *shader);
-void nir_lower_clamp_color_outputs(nir_shader *shader);
+bool nir_lower_clamp_color_outputs(nir_shader *shader);
void nir_lower_passthrough_edgeflags(nir_shader *shader);
+void nir_lower_tes_patch_vertices(nir_shader *tes, unsigned patch_vertices);
typedef struct nir_lower_wpos_ytransform_options {
int state_tokens[5];
bool nir_lower_wpos_ytransform(nir_shader *shader,
const nir_lower_wpos_ytransform_options *options);
-bool nir_lower_wpos_center(nir_shader *shader);
+bool nir_lower_wpos_center(nir_shader *shader, const bool for_sample_shading);
typedef struct nir_lower_drawpixels_options {
int texcoord_state_tokens[5];
void nir_lower_bitmap(nir_shader *shader, const nir_lower_bitmap_options *options);
-void nir_lower_atomics(nir_shader *shader,
+bool nir_lower_atomics(nir_shader *shader,
const struct gl_shader_program *shader_program);
-void nir_lower_to_source_mods(nir_shader *shader);
+bool nir_lower_atomics_to_ssbo(nir_shader *shader, unsigned ssbo_offset);
+bool nir_lower_to_source_mods(nir_shader *shader);
bool nir_lower_gs_intrinsics(nir_shader *shader);
+typedef enum {
+ nir_lower_imul64 = (1 << 0),
+ nir_lower_isign64 = (1 << 1),
+ /** Lower all int64 modulus and division opcodes */
+ nir_lower_divmod64 = (1 << 2),
+} nir_lower_int64_options;
+
+bool nir_lower_int64(nir_shader *shader, nir_lower_int64_options options);
+
typedef enum {
nir_lower_drcp = (1 << 0),
nir_lower_dsqrt = (1 << 1),
nir_lower_dmod = (1 << 8)
} nir_lower_doubles_options;
-void nir_lower_doubles(nir_shader *shader, nir_lower_doubles_options options);
-void nir_lower_double_pack(nir_shader *shader);
+bool nir_lower_doubles(nir_shader *shader, nir_lower_doubles_options options);
+bool nir_lower_64bit_pack(nir_shader *shader);
bool nir_normalize_cubemap_coords(nir_shader *shader);
bool nir_ssa_defs_interfere(nir_ssa_def *a, nir_ssa_def *b);
-void nir_convert_to_ssa_impl(nir_function_impl *impl);
-void nir_convert_to_ssa(nir_shader *shader);
-
bool nir_repair_ssa_impl(nir_function_impl *impl);
bool nir_repair_ssa(nir_shader *shader);
* registers. If false, convert all values (even those not involved in a phi
* node) to registers.
*/
-void nir_convert_from_ssa(nir_shader *shader, bool phi_webs_only);
+bool nir_convert_from_ssa(nir_shader *shader, bool phi_webs_only);
bool nir_lower_phis_to_regs_block(nir_block *block);
bool nir_lower_ssa_defs_to_regs_block(nir_block *block);
bool nir_opt_algebraic(nir_shader *shader);
+bool nir_opt_algebraic_before_ffma(nir_shader *shader);
bool nir_opt_algebraic_late(nir_shader *shader);
bool nir_opt_constant_folding(nir_shader *shader);
bool nir_copy_prop(nir_shader *shader);
+bool nir_opt_copy_prop_vars(nir_shader *shader);
+
bool nir_opt_cse(nir_shader *shader);
bool nir_opt_dce(nir_shader *shader);
bool nir_opt_if(nir_shader *shader);
+bool nir_opt_intrinsics(nir_shader *shader);
+
bool nir_opt_loop_unroll(nir_shader *shader, nir_variable_mode indirect_mask);
+bool nir_opt_move_comparisons(nir_shader *shader);
+
bool nir_opt_peephole_select(nir_shader *shader, unsigned limit);
bool nir_opt_remove_phis(nir_shader *shader);
#ifdef __cplusplus
} /* extern "C" */
#endif
+
+#endif /* NIR_H */
projects / mesa.git / blobdiff