#include "util/macros.h"
#include "compiler/nir_types.h"
#include "compiler/shader_enums.h"
+#include "compiler/shader_info.h"
#include <stdio.h>
#include "nir_opcodes.h"
* Note that you have to be a bit careful as the generated cast function
* destroys constness.
*/
-#define NIR_DEFINE_CAST(name, in_type, out_type, field) \
-static inline out_type * \
-name(const in_type *parent) \
-{ \
- return exec_node_data(out_type, parent, field); \
+#define NIR_DEFINE_CAST(name, in_type, out_type, field, \
+ type_field, type_value) \
+static inline out_type * \
+name(const in_type *parent) \
+{ \
+ assert(parent && parent->type_field == type_value); \
+ return exec_node_data(out_type, parent, field); \
}
struct nir_function;
/**
* Interpolation mode for shader inputs / outputs
*
- * \sa glsl_interp_qualifier
+ * \sa glsl_interp_mode
*/
unsigned interpolation:2;
*/
unsigned location_frac:2;
+ /**
+ * If true, this variable represents an array of scalars that should
+ * be tightly packed. In other words, consecutive array elements
+ * should be stored one component apart, rather than one slot apart.
+ */
+ bool compact:1;
+
+ /**
+ * Whether this is a fragment shader output implicitly initialized with
+ * the previous contents of the specified render target at the
+ * framebuffer location corresponding to this shader invocation.
+ */
+ unsigned fb_fetch_output:1;
+
/**
* \brief Layout qualifier for gl_FragDepth.
*
return instr->dest.dest.ssa.num_components;
}
+bool nir_alu_srcs_equal(const nir_alu_instr *alu1, const nir_alu_instr *alu2,
+ unsigned src1, unsigned src2);
+
typedef enum {
nir_deref_type_var,
nir_deref_type_array,
} nir_deref_var;
/* This enum describes how the array is referenced. If the deref is
- * direct then the base_offset is used. If the deref is indirect then then
+ * direct then the base_offset is used. If the deref is indirect then
* offset is given by base_offset + indirect. If the deref is a wildcard
* then the deref refers to all of the elements of the array at the same
* time. Wildcard dereferences are only ever allowed in copy_var
unsigned index;
} nir_deref_struct;
-NIR_DEFINE_CAST(nir_deref_as_var, nir_deref, nir_deref_var, deref)
-NIR_DEFINE_CAST(nir_deref_as_array, nir_deref, nir_deref_array, deref)
-NIR_DEFINE_CAST(nir_deref_as_struct, nir_deref, nir_deref_struct, deref)
+NIR_DEFINE_CAST(nir_deref_as_var, nir_deref, nir_deref_var, deref,
+ deref_type, nir_deref_type_var)
+NIR_DEFINE_CAST(nir_deref_as_array, nir_deref, nir_deref_array, deref,
+ deref_type, nir_deref_type_array)
+NIR_DEFINE_CAST(nir_deref_as_struct, nir_deref, nir_deref_struct, deref,
+ deref_type, nir_deref_type_struct)
/* Returns the last deref in the chain. */
static inline nir_deref *
nir_num_intrinsics = nir_last_intrinsic + 1
} nir_intrinsic_op;
-#undef INTRINSIC
-#undef LAST_INTRINSIC
-
#define NIR_INTRINSIC_MAX_CONST_INDEX 3
/** Represents an intrinsic
*/
NIR_INTRINSIC_BINDING = 7,
+ /**
+ * Component offset.
+ */
+ NIR_INTRINSIC_COMPONENT = 8,
+
+ /**
+ * Interpolation mode (only meaningful for FS inputs).
+ */
+ NIR_INTRINSIC_INTERP_MODE = 9,
+
NIR_INTRINSIC_NUM_INDEX_FLAGS,
} nir_intrinsic_index_flag;
INTRINSIC_IDX_ACCESSORS(range, RANGE, unsigned)
INTRINSIC_IDX_ACCESSORS(desc_set, DESC_SET, unsigned)
INTRINSIC_IDX_ACCESSORS(binding, BINDING, unsigned)
+INTRINSIC_IDX_ACCESSORS(component, COMPONENT, unsigned)
+INTRINSIC_IDX_ACCESSORS(interp_mode, INTERP_MODE, unsigned)
/**
* \group texture information
case GLSL_SAMPLER_DIM_MS:
case GLSL_SAMPLER_DIM_RECT:
case GLSL_SAMPLER_DIM_EXTERNAL:
+ case GLSL_SAMPLER_DIM_SUBPASS:
ret = 2;
break;
case GLSL_SAMPLER_DIM_3D:
}
}
+static inline nir_alu_type
+nir_tex_instr_src_type(nir_tex_instr *instr, unsigned src)
+{
+ switch (instr->src[src].src_type) {
+ case nir_tex_src_coord:
+ switch (instr->op) {
+ case nir_texop_txf:
+ case nir_texop_txf_ms:
+ case nir_texop_txf_ms_mcs:
+ case nir_texop_samples_identical:
+ return nir_type_int;
+
+ default:
+ return nir_type_float;
+ }
+
+ case nir_tex_src_lod:
+ switch (instr->op) {
+ case nir_texop_txs:
+ case nir_texop_txf:
+ return nir_type_int;
+
+ default:
+ return nir_type_float;
+ }
+
+ case nir_tex_src_projector:
+ case nir_tex_src_comparitor:
+ case nir_tex_src_bias:
+ case nir_tex_src_ddx:
+ case nir_tex_src_ddy:
+ return nir_type_float;
+
+ case nir_tex_src_offset:
+ case nir_tex_src_ms_index:
+ case nir_tex_src_texture_offset:
+ case nir_tex_src_sampler_offset:
+ return nir_type_int;
+
+ default:
+ unreachable("Invalid texture source type");
+ }
+}
+
static inline unsigned
nir_tex_instr_src_size(nir_tex_instr *instr, unsigned src)
{
return -1;
}
+void nir_tex_instr_remove_src(nir_tex_instr *tex, unsigned src_idx);
+
typedef union {
float f32[4];
double f64[4];
struct exec_list entries;
} nir_parallel_copy_instr;
-NIR_DEFINE_CAST(nir_instr_as_alu, nir_instr, nir_alu_instr, instr)
-NIR_DEFINE_CAST(nir_instr_as_call, nir_instr, nir_call_instr, instr)
-NIR_DEFINE_CAST(nir_instr_as_jump, nir_instr, nir_jump_instr, instr)
-NIR_DEFINE_CAST(nir_instr_as_tex, nir_instr, nir_tex_instr, instr)
-NIR_DEFINE_CAST(nir_instr_as_intrinsic, nir_instr, nir_intrinsic_instr, instr)
-NIR_DEFINE_CAST(nir_instr_as_load_const, nir_instr, nir_load_const_instr, instr)
-NIR_DEFINE_CAST(nir_instr_as_ssa_undef, nir_instr, nir_ssa_undef_instr, instr)
-NIR_DEFINE_CAST(nir_instr_as_phi, nir_instr, nir_phi_instr, instr)
+NIR_DEFINE_CAST(nir_instr_as_alu, nir_instr, nir_alu_instr, instr,
+ type, nir_instr_type_alu)
+NIR_DEFINE_CAST(nir_instr_as_call, nir_instr, nir_call_instr, instr,
+ type, nir_instr_type_call)
+NIR_DEFINE_CAST(nir_instr_as_jump, nir_instr, nir_jump_instr, instr,
+ type, nir_instr_type_jump)
+NIR_DEFINE_CAST(nir_instr_as_tex, nir_instr, nir_tex_instr, instr,
+ type, nir_instr_type_tex)
+NIR_DEFINE_CAST(nir_instr_as_intrinsic, nir_instr, nir_intrinsic_instr, instr,
+ type, nir_instr_type_intrinsic)
+NIR_DEFINE_CAST(nir_instr_as_load_const, nir_instr, nir_load_const_instr, instr,
+ type, nir_instr_type_load_const)
+NIR_DEFINE_CAST(nir_instr_as_ssa_undef, nir_instr, nir_ssa_undef_instr, instr,
+ type, nir_instr_type_ssa_undef)
+NIR_DEFINE_CAST(nir_instr_as_phi, nir_instr, nir_phi_instr, instr,
+ type, nir_instr_type_phi)
NIR_DEFINE_CAST(nir_instr_as_parallel_copy, nir_instr,
- nir_parallel_copy_instr, instr)
+ nir_parallel_copy_instr, instr,
+ type, nir_instr_type_parallel_copy)
/*
* Control flow
struct exec_list else_list; /** < list of nir_cf_node */
} nir_if;
-static inline nir_cf_node *
-nir_if_first_then_node(nir_if *if_stmt)
-{
- struct exec_node *head = exec_list_get_head(&if_stmt->then_list);
- return exec_node_data(nir_cf_node, head, node);
-}
-
-static inline nir_cf_node *
-nir_if_last_then_node(nir_if *if_stmt)
-{
- struct exec_node *tail = exec_list_get_tail(&if_stmt->then_list);
- return exec_node_data(nir_cf_node, tail, node);
-}
-
-static inline nir_cf_node *
-nir_if_first_else_node(nir_if *if_stmt)
-{
- struct exec_node *head = exec_list_get_head(&if_stmt->else_list);
- return exec_node_data(nir_cf_node, head, node);
-}
-
-static inline nir_cf_node *
-nir_if_last_else_node(nir_if *if_stmt)
-{
- struct exec_node *tail = exec_list_get_tail(&if_stmt->else_list);
- return exec_node_data(nir_cf_node, tail, node);
-}
-
typedef struct {
nir_cf_node cf_node;
struct exec_list body; /** < list of nir_cf_node */
} nir_loop;
-static inline nir_cf_node *
-nir_loop_first_cf_node(nir_loop *loop)
-{
- return exec_node_data(nir_cf_node, exec_list_get_head(&loop->body), node);
-}
-
-static inline nir_cf_node *
-nir_loop_last_cf_node(nir_loop *loop)
-{
- return exec_node_data(nir_cf_node, exec_list_get_tail(&loop->body), node);
-}
-
/**
* Various bits of metadata that can may be created or required by
* optimization and analysis passes
ATTRIBUTE_RETURNS_NONNULL static inline nir_block *
nir_start_block(nir_function_impl *impl)
{
- return (nir_block *) impl->body.head;
+ return (nir_block *) impl->body.head_sentinel.next;
}
ATTRIBUTE_RETURNS_NONNULL static inline nir_block *
nir_impl_last_block(nir_function_impl *impl)
{
- return (nir_block *) impl->body.tail_pred;
+ return (nir_block *) impl->body.tail_sentinel.prev;
}
static inline nir_cf_node *
return exec_node_is_tail_sentinel(node->node.next);
}
-NIR_DEFINE_CAST(nir_cf_node_as_block, nir_cf_node, nir_block, cf_node)
-NIR_DEFINE_CAST(nir_cf_node_as_if, nir_cf_node, nir_if, cf_node)
-NIR_DEFINE_CAST(nir_cf_node_as_loop, nir_cf_node, nir_loop, cf_node)
-NIR_DEFINE_CAST(nir_cf_node_as_function, nir_cf_node, nir_function_impl, cf_node)
+NIR_DEFINE_CAST(nir_cf_node_as_block, nir_cf_node, nir_block, cf_node,
+ type, nir_cf_node_block)
+NIR_DEFINE_CAST(nir_cf_node_as_if, nir_cf_node, nir_if, cf_node,
+ type, nir_cf_node_if)
+NIR_DEFINE_CAST(nir_cf_node_as_loop, nir_cf_node, nir_loop, cf_node,
+ type, nir_cf_node_loop)
+NIR_DEFINE_CAST(nir_cf_node_as_function, nir_cf_node,
+ nir_function_impl, cf_node, type, nir_cf_node_function)
+
+static inline nir_block *
+nir_if_first_then_block(nir_if *if_stmt)
+{
+ struct exec_node *head = exec_list_get_head(&if_stmt->then_list);
+ return nir_cf_node_as_block(exec_node_data(nir_cf_node, head, node));
+}
+
+static inline nir_block *
+nir_if_last_then_block(nir_if *if_stmt)
+{
+ struct exec_node *tail = exec_list_get_tail(&if_stmt->then_list);
+ return nir_cf_node_as_block(exec_node_data(nir_cf_node, tail, node));
+}
+
+static inline nir_block *
+nir_if_first_else_block(nir_if *if_stmt)
+{
+ struct exec_node *head = exec_list_get_head(&if_stmt->else_list);
+ return nir_cf_node_as_block(exec_node_data(nir_cf_node, head, node));
+}
+
+static inline nir_block *
+nir_if_last_else_block(nir_if *if_stmt)
+{
+ struct exec_node *tail = exec_list_get_tail(&if_stmt->else_list);
+ return nir_cf_node_as_block(exec_node_data(nir_cf_node, tail, node));
+}
+
+static inline nir_block *
+nir_loop_first_block(nir_loop *loop)
+{
+ struct exec_node *head = exec_list_get_head(&loop->body);
+ return nir_cf_node_as_block(exec_node_data(nir_cf_node, head, node));
+}
+
+static inline nir_block *
+nir_loop_last_block(nir_loop *loop)
+{
+ struct exec_node *tail = exec_list_get_tail(&loop->body);
+ return nir_cf_node_as_block(exec_node_data(nir_cf_node, tail, node));
+}
typedef enum {
nir_parameter_in,
bool vertex_id_zero_based;
bool lower_cs_local_index_from_id;
-} nir_shader_compiler_options;
-typedef struct nir_shader_info {
- const char *name;
-
- /* Descriptive name provided by the client; may be NULL */
- const char *label;
-
- /* Number of textures used by this shader */
- unsigned num_textures;
- /* Number of uniform buffers used by this shader */
- unsigned num_ubos;
- /* Number of atomic buffers used by this shader */
- unsigned num_abos;
- /* Number of shader storage buffers used by this shader */
- unsigned num_ssbos;
- /* Number of images used by this shader */
- unsigned num_images;
-
- /* Which inputs are actually read */
- uint64_t inputs_read;
- /* Which inputs are actually read and are double */
- uint64_t double_inputs_read;
- /* Which outputs are actually written */
- uint64_t outputs_written;
- /* Which system values are actually read */
- uint64_t system_values_read;
-
- /* Which patch inputs are actually read */
- uint32_t patch_inputs_read;
- /* Which patch outputs are actually written */
- uint32_t patch_outputs_written;
-
- /* Whether or not this shader ever uses textureGather() */
- bool uses_texture_gather;
-
- /** Whether or not this shader uses nir_intrinsic_interp_var_at_offset */
- bool uses_interp_var_at_offset;
-
- /* Whether or not this shader uses the gl_ClipDistance output */
- bool uses_clip_distance_out;
-
- /* Whether or not separate shader objects were used */
- bool separate_shader;
-
- /** Was this shader linked with any transform feedback varyings? */
- bool has_transform_feedback_varyings;
-
- union {
- struct {
- /** The number of vertices recieves per input primitive */
- unsigned vertices_in;
-
- /** The output primitive type (GL enum value) */
- unsigned output_primitive;
-
- /** The maximum number of vertices the geometry shader might write. */
- unsigned vertices_out;
-
- /** 1 .. MAX_GEOMETRY_SHADER_INVOCATIONS */
- unsigned invocations;
-
- /** Whether or not this shader uses EndPrimitive */
- bool uses_end_primitive;
-
- /** Whether or not this shader uses non-zero streams */
- bool uses_streams;
- } gs;
-
- struct {
- bool uses_discard;
-
- /**
- * Whether any inputs are declared with the "sample" qualifier.
- */
- bool uses_sample_qualifier;
-
- /**
- * Whether early fragment tests are enabled as defined by
- * ARB_shader_image_load_store.
- */
- bool early_fragment_tests;
-
- /** gl_FragDepth layout for ARB_conservative_depth. */
- enum gl_frag_depth_layout depth_layout;
- } fs;
-
- struct {
- unsigned local_size[3];
- } cs;
-
- struct {
- /** The number of vertices in the TCS output patch. */
- unsigned vertices_out;
- } tcs;
- };
-} nir_shader_info;
+ /**
+ * Should nir_lower_io() create load_interpolated_input intrinsics?
+ *
+ * If not, it generates regular load_input intrinsics and interpolation
+ * information must be inferred from the list of input nir_variables.
+ */
+ bool use_interpolated_input_intrinsics;
+} nir_shader_compiler_options;
typedef struct nir_shader {
/** list of uniforms (nir_variable) */
const struct nir_shader_compiler_options *options;
/** Various bits of compile-time information about a given shader */
- struct nir_shader_info info;
+ struct shader_info *info;
/** list of global variables in the shader (nir_variable) */
struct exec_list globals;
gl_shader_stage stage;
} nir_shader;
-static inline nir_function *
+static inline nir_function_impl *
nir_shader_get_entrypoint(nir_shader *shader)
{
assert(exec_list_length(&shader->functions) == 1);
nir_function *func = exec_node_data(nir_function, func_node, node);
assert(func->return_type == glsl_void_type());
assert(func->num_params == 0);
- return func;
+ assert(func->impl);
+ return func->impl;
}
#define nir_foreach_function(func, shader) \
nir_shader *nir_shader_create(void *mem_ctx,
gl_shader_stage stage,
- const nir_shader_compiler_options *options);
+ const nir_shader_compiler_options *options,
+ shader_info *si);
/** creates a register, including assigning it an index and adding it to the list */
nir_register *nir_global_reg_create(nir_shader *shader);
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_load_const_instr *
nir_deref_get_const_initializer_load(nir_shader *shader, nir_deref_var *deref);
return nir_before_block(nir_cf_node_as_block(nir_cf_node_next(node)));
}
+static inline nir_cursor
+nir_after_phis(nir_block *block)
+{
+ nir_foreach_instr(instr, block) {
+ if (instr->type != nir_instr_type_phi)
+ return nir_before_instr(instr);
+ }
+ return nir_after_block(block);
+}
+
static inline nir_cursor
nir_after_cf_node_and_phis(nir_cf_node *node)
{
return nir_after_block(nir_cf_node_as_block(node));
nir_block *block = nir_cf_node_as_block(nir_cf_node_next(node));
- assert(block->cf_node.type == nir_cf_node_block);
- nir_foreach_instr(instr, block) {
- if (instr->type != nir_instr_type_phi)
- return nir_before_instr(instr);
- }
- return nir_after_block(block);
+ return nir_after_phis(block);
}
static inline nir_cursor
bool nir_inline_functions(nir_shader *shader);
+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);
bool nir_lower_locals_to_regs(nir_shader *shader);
-void nir_lower_io_to_temporaries(nir_shader *shader, nir_function *entrypoint,
+void nir_lower_io_to_temporaries(nir_shader *shader,
+ nir_function_impl *entrypoint,
bool outputs, bool inputs);
void nir_shader_gather_info(nir_shader *shader, nir_function_impl *entrypoint);
-void nir_assign_var_locations(struct exec_list *var_list,
- unsigned *size,
+void nir_assign_var_locations(struct exec_list *var_list, unsigned *size,
int (*type_size)(const struct glsl_type *));
+typedef enum {
+ /* If set, this forces all non-flat fragment shader inputs to be
+ * interpolated as if with the "sample" qualifier. This requires
+ * nir_shader_compiler_options::use_interpolated_input_intrinsics.
+ */
+ nir_lower_io_force_sample_interpolation = (1 << 1),
+} nir_lower_io_options;
void nir_lower_io(nir_shader *shader,
nir_variable_mode modes,
- int (*type_size)(const struct glsl_type *));
+ 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);
+
void nir_lower_io_types(nir_shader *shader);
void nir_lower_vars_to_ssa(nir_shader *shader);
void nir_move_vec_src_uses_to_dest(nir_shader *shader);
bool nir_lower_vec_to_movs(nir_shader *shader);
-void nir_lower_alu_to_scalar(nir_shader *shader);
+bool nir_lower_alu_to_scalar(nir_shader *shader);
void nir_lower_load_const_to_scalar(nir_shader *shader);
-void nir_lower_phis_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,
const struct gl_shader_program *shader_program);
*/
unsigned lower_txp;
+ /**
+ * If true, lower away nir_tex_src_offset for all texelfetch instructions.
+ */
+ bool lower_txf_offset;
+
+ /**
+ * If true, lower away nir_tex_src_offset for all rect textures.
+ */
+ bool lower_rect_offset;
+
/**
* If true, lower rect textures to 2D, using txs to fetch the
* texture dimensions and dividing the texture coords by the
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);
void nir_lower_two_sided_color(nir_shader *shader);
bool nir_opt_dead_cf(nir_shader *shader);
-void nir_opt_gcm(nir_shader *shader);
+bool nir_opt_gcm(nir_shader *shader, bool value_number);
-bool nir_opt_peephole_select(nir_shader *shader);
+bool nir_opt_peephole_select(nir_shader *shader, unsigned limit);
bool nir_opt_remove_phis(nir_shader *shader);
bool nir_opt_undef(nir_shader *shader);
+bool nir_opt_conditional_discard(nir_shader *shader);
+
void nir_sweep(nir_shader *shader);
nir_intrinsic_op nir_intrinsic_from_system_value(gl_system_value val);