*/
#include "glsl_to_nir.h"
+#include "nir_control_flow.h"
+#include "nir_builder.h"
#include "ir_visitor.h"
#include "ir_hierarchical_visitor.h"
#include "ir.h"
+#include "main/imports.h"
/*
* pass to lower GLSL IR to NIR
class nir_visitor : public ir_visitor
{
public:
- nir_visitor(nir_shader *shader, bool supports_ints);
+ nir_visitor(nir_shader *shader);
~nir_visitor();
virtual void visit(ir_variable *);
virtual void visit(ir_dereference_variable *);
virtual void visit(ir_dereference_record *);
virtual void visit(ir_dereference_array *);
+ virtual void visit(ir_barrier *);
void create_function(ir_function *ir);
private:
void create_overload(ir_function_signature *ir, nir_function *function);
void add_instr(nir_instr *instr, unsigned num_components);
- nir_src evaluate_rvalue(ir_rvalue *ir);
+ nir_ssa_def *evaluate_rvalue(ir_rvalue *ir);
- nir_alu_instr *emit(nir_op op, unsigned dest_size, nir_src *srcs);
- nir_alu_instr *emit(nir_op op, unsigned dest_size, nir_src src1);
- nir_alu_instr *emit(nir_op op, unsigned dest_size, nir_src src1,
- nir_src src2);
- nir_alu_instr *emit(nir_op op, unsigned dest_size, nir_src src1,
- nir_src src2, nir_src src3);
+ nir_alu_instr *emit(nir_op op, unsigned dest_size, nir_ssa_def **srcs);
+ nir_alu_instr *emit(nir_op op, unsigned dest_size, nir_ssa_def *src1);
+ nir_alu_instr *emit(nir_op op, unsigned dest_size, nir_ssa_def *src1,
+ nir_ssa_def *src2);
+ nir_alu_instr *emit(nir_op op, unsigned dest_size, nir_ssa_def *src1,
+ nir_ssa_def *src2, nir_ssa_def *src3);
bool supports_ints;
nir_shader *shader;
nir_function_impl *impl;
- exec_list *cf_node_list;
- nir_instr *result; /* result of the expression tree last visited */
+ nir_builder b;
+ nir_ssa_def *result; /* result of the expression tree last visited */
+
+ nir_deref_var *evaluate_deref(nir_instr *mem_ctx, ir_instruction *ir);
/* the head of the dereference chain we're creating */
nir_deref_var *deref_head;
}; /* end of anonymous namespace */
-static const nir_shader_compiler_options default_options = {
-};
-
nir_shader *
-glsl_to_nir(exec_list *ir, _mesa_glsl_parse_state *state,
- bool native_integers)
+glsl_to_nir(const struct gl_shader_program *shader_prog,
+ gl_shader_stage stage,
+ const nir_shader_compiler_options *options)
{
- const nir_shader_compiler_options *options;
+ struct gl_shader *sh = shader_prog->_LinkedShaders[stage];
- if (state) {
- struct gl_context *ctx = state->ctx;
- struct gl_shader_compiler_options *gl_options =
- &ctx->Const.ShaderCompilerOptions[state->stage];
+ nir_shader *shader = nir_shader_create(NULL, stage, options);
- if (!gl_options->NirOptions) {
- nir_shader_compiler_options *new_options =
- rzalloc(ctx, nir_shader_compiler_options);
- options = gl_options->NirOptions = new_options;
- } else {
- options = gl_options->NirOptions;
- }
- } else {
- options = &default_options;
+ nir_visitor v1(shader);
+ nir_function_visitor v2(&v1);
+ v2.run(sh->ir);
+ visit_exec_list(sh->ir, &v1);
+
+ nir_lower_outputs_to_temporaries(shader);
+
+ shader->info.name = ralloc_asprintf(shader, "GLSL%d", shader_prog->Name);
+ if (shader_prog->Label)
+ shader->info.label = ralloc_strdup(shader, shader_prog->Label);
+ shader->info.num_textures = _mesa_fls(sh->Program->SamplersUsed);
+ shader->info.num_ubos = sh->NumUniformBlocks;
+ shader->info.num_abos = shader_prog->NumAtomicBuffers;
+ shader->info.num_ssbos = sh->NumShaderStorageBlocks;
+ shader->info.num_images = sh->NumImages;
+ shader->info.inputs_read = sh->Program->InputsRead;
+ shader->info.outputs_written = sh->Program->OutputsWritten;
+ shader->info.patch_inputs_read = sh->Program->PatchInputsRead;
+ shader->info.patch_outputs_written = sh->Program->PatchOutputsWritten;
+ shader->info.system_values_read = sh->Program->SystemValuesRead;
+ shader->info.uses_texture_gather = sh->Program->UsesGather;
+ shader->info.uses_clip_distance_out =
+ sh->Program->ClipDistanceArraySize != 0;
+ shader->info.separate_shader = shader_prog->SeparateShader;
+ shader->info.has_transform_feedback_varyings =
+ shader_prog->TransformFeedback.NumVarying > 0;
+
+ switch (stage) {
+ case MESA_SHADER_GEOMETRY:
+ shader->info.gs.vertices_in = shader_prog->Geom.VerticesIn;
+ shader->info.gs.output_primitive = sh->Geom.OutputType;
+ shader->info.gs.vertices_out = sh->Geom.VerticesOut;
+ shader->info.gs.invocations = sh->Geom.Invocations;
+ shader->info.gs.uses_end_primitive = shader_prog->Geom.UsesEndPrimitive;
+ shader->info.gs.uses_streams = shader_prog->Geom.UsesStreams;
+ break;
+
+ case MESA_SHADER_FRAGMENT: {
+ struct gl_fragment_program *fp =
+ (struct gl_fragment_program *)sh->Program;
+
+ shader->info.fs.uses_discard = fp->UsesKill;
+ shader->info.fs.early_fragment_tests = sh->EarlyFragmentTests;
+ shader->info.fs.depth_layout = fp->FragDepthLayout;
+ break;
}
- nir_shader *shader = nir_shader_create(NULL, options);
-
- if (state) {
- shader->num_user_structures = state->num_user_structures;
- shader->user_structures = ralloc_array(shader, glsl_type *,
- shader->num_user_structures);
- memcpy(shader->user_structures, state->user_structures,
- shader->num_user_structures * sizeof(glsl_type *));
- } else {
- shader->num_user_structures = 0;
- shader->user_structures = NULL;
+ case MESA_SHADER_COMPUTE: {
+ struct gl_compute_program *cp = (struct gl_compute_program *)sh->Program;
+ shader->info.cs.local_size[0] = cp->LocalSize[0];
+ shader->info.cs.local_size[1] = cp->LocalSize[1];
+ shader->info.cs.local_size[2] = cp->LocalSize[2];
+ break;
}
- nir_visitor v1(shader, native_integers);
- nir_function_visitor v2(&v1);
- v2.run(ir);
- visit_exec_list(ir, &v1);
+ default:
+ break; /* No stage-specific info */
+ }
return shader;
}
-nir_visitor::nir_visitor(nir_shader *shader, bool supports_ints)
+nir_visitor::nir_visitor(nir_shader *shader)
{
- this->supports_ints = supports_ints;
+ this->supports_ints = shader->options->native_integers;
this->shader = shader;
this->is_global = true;
this->var_table = _mesa_hash_table_create(NULL, _mesa_hash_pointer,
_mesa_hash_table_destroy(this->overload_table, NULL);
}
+nir_deref_var *
+nir_visitor::evaluate_deref(nir_instr *mem_ctx, ir_instruction *ir)
+{
+ ir->accept(this);
+ ralloc_steal(mem_ctx, this->deref_head);
+ return this->deref_head;
+}
+
static nir_constant *
constant_copy(ir_constant *ir, void *mem_ctx)
{
var->data.read_only = ir->data.read_only;
var->data.centroid = ir->data.centroid;
var->data.sample = ir->data.sample;
+ var->data.patch = ir->data.patch;
var->data.invariant = ir->data.invariant;
var->data.location = ir->data.location;
break;
case ir_var_shader_in:
- if (ir->data.location == VARYING_SLOT_FACE) {
+ if (shader->stage == MESA_SHADER_FRAGMENT &&
+ ir->data.location == VARYING_SLOT_FACE) {
/* For whatever reason, GLSL IR makes gl_FrontFacing an input */
var->data.location = SYSTEM_VALUE_FRONT_FACE;
var->data.mode = nir_var_system_value;
+ } else if (shader->stage == MESA_SHADER_GEOMETRY &&
+ ir->data.location == VARYING_SLOT_PRIMITIVE_ID) {
+ /* For whatever reason, GLSL IR makes gl_PrimitiveIDIn an input */
+ var->data.location = SYSTEM_VALUE_PRIMITIVE_ID;
+ var->data.mode = nir_var_system_value;
} else {
var->data.mode = nir_var_shader_in;
}
var->data.mode = nir_var_uniform;
break;
+ case ir_var_shader_storage:
+ var->data.mode = nir_var_shader_storage;
+ break;
case ir_var_system_value:
var->data.mode = nir_var_system_value;
var->data.index = ir->data.index;
var->data.binding = ir->data.binding;
- /* XXX Get rid of buffer_index */
- var->data.atomic.buffer_index = ir->data.binding;
var->data.atomic.offset = ir->data.atomic.offset;
var->data.image.read_only = ir->data.image_read_only;
var->data.image.write_only = ir->data.image_write_only;
var->interface_type = ir->get_interface_type();
- switch (var->data.mode) {
- case nir_var_local:
- exec_list_push_tail(&impl->locals, &var->node);
- break;
-
- case nir_var_global:
- exec_list_push_tail(&shader->globals, &var->node);
- break;
-
- case nir_var_shader_in:
- _mesa_hash_table_insert(shader->inputs, var->name, var);
- break;
-
- case nir_var_shader_out:
- _mesa_hash_table_insert(shader->outputs, var->name, var);
- break;
-
- case nir_var_uniform:
- _mesa_hash_table_insert(shader->uniforms, var->name, var);
- break;
-
- case nir_var_system_value:
- exec_list_push_tail(&shader->system_values, &var->node);
- break;
-
- default:
- unreachable("not reached");
- }
+ if (var->data.mode == nir_var_local)
+ nir_function_impl_add_variable(impl, var);
+ else
+ nir_shader_add_variable(shader, var);
_mesa_hash_table_insert(var_table, ir, var);
this->var = var;
this->is_global = false;
- this->cf_node_list = &impl->body;
+ nir_builder_init(&b, impl);
+ b.cursor = nir_after_cf_list(&impl->body);
visit_exec_list(&ir->body, this);
this->is_global = true;
void
nir_visitor::visit(ir_loop *ir)
{
- exec_list *old_list = this->cf_node_list;
-
nir_loop *loop = nir_loop_create(this->shader);
- nir_cf_node_insert_end(old_list, &loop->cf_node);
- this->cf_node_list = &loop->body;
- visit_exec_list(&ir->body_instructions, this);
+ nir_builder_cf_insert(&b, &loop->cf_node);
- this->cf_node_list = old_list;
+ b.cursor = nir_after_cf_list(&loop->body);
+ visit_exec_list(&ir->body_instructions, this);
+ b.cursor = nir_after_cf_node(&loop->cf_node);
}
void
nir_visitor::visit(ir_if *ir)
{
- nir_src condition = evaluate_rvalue(ir->condition);
-
- exec_list *old_list = this->cf_node_list;
+ nir_src condition =
+ nir_src_for_ssa(evaluate_rvalue(ir->condition));
nir_if *if_stmt = nir_if_create(this->shader);
if_stmt->condition = condition;
- nir_cf_node_insert_end(old_list, &if_stmt->cf_node);
+ nir_builder_cf_insert(&b, &if_stmt->cf_node);
- this->cf_node_list = &if_stmt->then_list;
+ b.cursor = nir_after_cf_list(&if_stmt->then_list);
visit_exec_list(&ir->then_instructions, this);
- this->cf_node_list = &if_stmt->else_list;
+ b.cursor = nir_after_cf_list(&if_stmt->else_list);
visit_exec_list(&ir->else_instructions, this);
- this->cf_node_list = old_list;
+ b.cursor = nir_after_cf_node(&if_stmt->cf_node);
}
void
* discards will be immediately followed by a return.
*/
- nir_intrinsic_instr *discard =
- nir_intrinsic_instr_create(this->shader, nir_intrinsic_discard);
- nir_instr_insert_after_cf_list(this->cf_node_list, &discard->instr);
+ nir_intrinsic_instr *discard;
+ if (ir->condition) {
+ discard = nir_intrinsic_instr_create(this->shader,
+ nir_intrinsic_discard_if);
+ discard->src[0] =
+ nir_src_for_ssa(evaluate_rvalue(ir->condition));
+ } else {
+ discard = nir_intrinsic_instr_create(this->shader, nir_intrinsic_discard);
+ }
+
+ nir_builder_instr_insert(&b, &discard->instr);
}
void
nir_intrinsic_instr *instr =
nir_intrinsic_instr_create(this->shader, nir_intrinsic_emit_vertex);
instr->const_index[0] = ir->stream_id();
- nir_instr_insert_after_cf_list(this->cf_node_list, &instr->instr);
+ nir_builder_instr_insert(&b, &instr->instr);
}
void
nir_intrinsic_instr *instr =
nir_intrinsic_instr_create(this->shader, nir_intrinsic_end_primitive);
instr->const_index[0] = ir->stream_id();
- nir_instr_insert_after_cf_list(this->cf_node_list, &instr->instr);
+ nir_builder_instr_insert(&b, &instr->instr);
}
void
}
nir_jump_instr *instr = nir_jump_instr_create(this->shader, type);
- nir_instr_insert_after_cf_list(this->cf_node_list, &instr->instr);
+ nir_builder_instr_insert(&b, &instr->instr);
}
void
nir_visitor::visit(ir_return *ir)
{
if (ir->value != NULL) {
- ir->value->accept(this);
nir_intrinsic_instr *copy =
nir_intrinsic_instr_create(this->shader, nir_intrinsic_copy_var);
- copy->variables[0] = nir_deref_var_create(this->shader,
- this->impl->return_var);
- copy->variables[1] = this->deref_head;
+ copy->variables[0] = nir_deref_var_create(copy, this->impl->return_var);
+ copy->variables[1] = evaluate_deref(©->instr, ir->value);
}
nir_jump_instr *instr = nir_jump_instr_create(this->shader, nir_jump_return);
- nir_instr_insert_after_cf_list(this->cf_node_list, &instr->instr);
+ nir_builder_instr_insert(&b, &instr->instr);
}
void
op = nir_intrinsic_atomic_counter_inc_var;
} else if (strcmp(ir->callee_name(), "__intrinsic_atomic_predecrement") == 0) {
op = nir_intrinsic_atomic_counter_dec_var;
+ } else if (strcmp(ir->callee_name(), "__intrinsic_image_load") == 0) {
+ op = nir_intrinsic_image_load;
+ } else if (strcmp(ir->callee_name(), "__intrinsic_image_store") == 0) {
+ op = nir_intrinsic_image_store;
+ } else if (strcmp(ir->callee_name(), "__intrinsic_image_atomic_add") == 0) {
+ op = nir_intrinsic_image_atomic_add;
+ } else if (strcmp(ir->callee_name(), "__intrinsic_image_atomic_min") == 0) {
+ op = nir_intrinsic_image_atomic_min;
+ } else if (strcmp(ir->callee_name(), "__intrinsic_image_atomic_max") == 0) {
+ op = nir_intrinsic_image_atomic_max;
+ } else if (strcmp(ir->callee_name(), "__intrinsic_image_atomic_and") == 0) {
+ op = nir_intrinsic_image_atomic_and;
+ } else if (strcmp(ir->callee_name(), "__intrinsic_image_atomic_or") == 0) {
+ op = nir_intrinsic_image_atomic_or;
+ } else if (strcmp(ir->callee_name(), "__intrinsic_image_atomic_xor") == 0) {
+ op = nir_intrinsic_image_atomic_xor;
+ } else if (strcmp(ir->callee_name(), "__intrinsic_image_atomic_exchange") == 0) {
+ op = nir_intrinsic_image_atomic_exchange;
+ } else if (strcmp(ir->callee_name(), "__intrinsic_image_atomic_comp_swap") == 0) {
+ op = nir_intrinsic_image_atomic_comp_swap;
+ } else if (strcmp(ir->callee_name(), "__intrinsic_memory_barrier") == 0) {
+ op = nir_intrinsic_memory_barrier;
+ } else if (strcmp(ir->callee_name(), "__intrinsic_image_size") == 0) {
+ op = nir_intrinsic_image_size;
+ } else if (strcmp(ir->callee_name(), "__intrinsic_image_samples") == 0) {
+ op = nir_intrinsic_image_samples;
+ } else if (strcmp(ir->callee_name(), "__intrinsic_store_ssbo") == 0) {
+ op = nir_intrinsic_store_ssbo;
+ } else if (strcmp(ir->callee_name(), "__intrinsic_load_ssbo") == 0) {
+ op = nir_intrinsic_load_ssbo;
+ } else if (strcmp(ir->callee_name(), "__intrinsic_ssbo_atomic_add_internal") == 0) {
+ op = nir_intrinsic_ssbo_atomic_add;
+ } else if (strcmp(ir->callee_name(), "__intrinsic_ssbo_atomic_and_internal") == 0) {
+ op = nir_intrinsic_ssbo_atomic_and;
+ } else if (strcmp(ir->callee_name(), "__intrinsic_ssbo_atomic_or_internal") == 0) {
+ op = nir_intrinsic_ssbo_atomic_or;
+ } else if (strcmp(ir->callee_name(), "__intrinsic_ssbo_atomic_xor_internal") == 0) {
+ op = nir_intrinsic_ssbo_atomic_xor;
+ } else if (strcmp(ir->callee_name(), "__intrinsic_ssbo_atomic_min_internal") == 0) {
+ assert(ir->return_deref);
+ if (ir->return_deref->type == glsl_type::int_type)
+ op = nir_intrinsic_ssbo_atomic_imin;
+ else if (ir->return_deref->type == glsl_type::uint_type)
+ op = nir_intrinsic_ssbo_atomic_umin;
+ else
+ unreachable("Invalid type");
+ } else if (strcmp(ir->callee_name(), "__intrinsic_ssbo_atomic_max_internal") == 0) {
+ assert(ir->return_deref);
+ if (ir->return_deref->type == glsl_type::int_type)
+ op = nir_intrinsic_ssbo_atomic_imax;
+ else if (ir->return_deref->type == glsl_type::uint_type)
+ op = nir_intrinsic_ssbo_atomic_umax;
+ else
+ unreachable("Invalid type");
+ } else if (strcmp(ir->callee_name(), "__intrinsic_ssbo_atomic_exchange_internal") == 0) {
+ op = nir_intrinsic_ssbo_atomic_exchange;
+ } else if (strcmp(ir->callee_name(), "__intrinsic_ssbo_atomic_comp_swap_internal") == 0) {
+ op = nir_intrinsic_ssbo_atomic_comp_swap;
+ } else if (strcmp(ir->callee_name(), "__intrinsic_shader_clock") == 0) {
+ op = nir_intrinsic_shader_clock;
+ } else if (strcmp(ir->callee_name(), "__intrinsic_group_memory_barrier") == 0) {
+ op = nir_intrinsic_group_memory_barrier;
+ } else if (strcmp(ir->callee_name(), "__intrinsic_memory_barrier_atomic_counter") == 0) {
+ op = nir_intrinsic_memory_barrier_atomic_counter;
+ } else if (strcmp(ir->callee_name(), "__intrinsic_memory_barrier_buffer") == 0) {
+ op = nir_intrinsic_memory_barrier_buffer;
+ } else if (strcmp(ir->callee_name(), "__intrinsic_memory_barrier_image") == 0) {
+ op = nir_intrinsic_memory_barrier_image;
+ } else if (strcmp(ir->callee_name(), "__intrinsic_memory_barrier_shared") == 0) {
+ op = nir_intrinsic_memory_barrier_shared;
} else {
unreachable("not reached");
}
nir_intrinsic_instr *instr = nir_intrinsic_instr_create(shader, op);
- ir_dereference *param =
- (ir_dereference *) ir->actual_parameters.get_head();
- param->accept(this);
- instr->variables[0] = this->deref_head;
- nir_ssa_dest_init(&instr->instr, &instr->dest, 1, NULL);
+ nir_dest *dest = &instr->dest;
+
+ switch (op) {
+ case nir_intrinsic_atomic_counter_read_var:
+ case nir_intrinsic_atomic_counter_inc_var:
+ case nir_intrinsic_atomic_counter_dec_var: {
+ ir_dereference *param =
+ (ir_dereference *) ir->actual_parameters.get_head();
+ instr->variables[0] = evaluate_deref(&instr->instr, param);
+ nir_ssa_dest_init(&instr->instr, &instr->dest, 1, NULL);
+ nir_builder_instr_insert(&b, &instr->instr);
+ break;
+ }
+ case nir_intrinsic_image_load:
+ case nir_intrinsic_image_store:
+ case nir_intrinsic_image_atomic_add:
+ case nir_intrinsic_image_atomic_min:
+ case nir_intrinsic_image_atomic_max:
+ case nir_intrinsic_image_atomic_and:
+ case nir_intrinsic_image_atomic_or:
+ case nir_intrinsic_image_atomic_xor:
+ case nir_intrinsic_image_atomic_exchange:
+ case nir_intrinsic_image_atomic_comp_swap:
+ case nir_intrinsic_image_samples:
+ case nir_intrinsic_image_size: {
+ nir_ssa_undef_instr *instr_undef =
+ nir_ssa_undef_instr_create(shader, 1);
+ nir_builder_instr_insert(&b, &instr_undef->instr);
+
+ /* Set the image variable dereference. */
+ exec_node *param = ir->actual_parameters.get_head();
+ ir_dereference *image = (ir_dereference *)param;
+ const glsl_type *type =
+ image->variable_referenced()->type->without_array();
+
+ instr->variables[0] = evaluate_deref(&instr->instr, image);
+ param = param->get_next();
+
+ /* Set the intrinsic destination. */
+ if (ir->return_deref) {
+ const nir_intrinsic_info *info =
+ &nir_intrinsic_infos[instr->intrinsic];
+ nir_ssa_dest_init(&instr->instr, &instr->dest,
+ info->dest_components, NULL);
+ }
+
+ if (op == nir_intrinsic_image_size ||
+ op == nir_intrinsic_image_samples) {
+ nir_builder_instr_insert(&b, &instr->instr);
+ break;
+ }
+
+ /* Set the address argument, extending the coordinate vector to four
+ * components.
+ */
+ nir_ssa_def *src_addr =
+ evaluate_rvalue((ir_dereference *)param);
+ nir_ssa_def *srcs[4];
+
+ for (int i = 0; i < 4; i++) {
+ if (i < type->coordinate_components())
+ srcs[i] = nir_channel(&b, src_addr, i);
+ else
+ srcs[i] = &instr_undef->def;
+ }
+
+ instr->src[0] = nir_src_for_ssa(nir_vec(&b, srcs, 4));
+ param = param->get_next();
+
+ /* Set the sample argument, which is undefined for single-sample
+ * images.
+ */
+ if (type->sampler_dimensionality == GLSL_SAMPLER_DIM_MS) {
+ instr->src[1] =
+ nir_src_for_ssa(evaluate_rvalue((ir_dereference *)param));
+ param = param->get_next();
+ } else {
+ instr->src[1] = nir_src_for_ssa(&instr_undef->def);
+ }
+
+ /* Set the intrinsic parameters. */
+ if (!param->is_tail_sentinel()) {
+ instr->src[2] =
+ nir_src_for_ssa(evaluate_rvalue((ir_dereference *)param));
+ param = param->get_next();
+ }
+
+ if (!param->is_tail_sentinel()) {
+ instr->src[3] =
+ nir_src_for_ssa(evaluate_rvalue((ir_dereference *)param));
+ param = param->get_next();
+ }
+ nir_builder_instr_insert(&b, &instr->instr);
+ break;
+ }
+ case nir_intrinsic_memory_barrier:
+ case nir_intrinsic_group_memory_barrier:
+ case nir_intrinsic_memory_barrier_atomic_counter:
+ case nir_intrinsic_memory_barrier_buffer:
+ case nir_intrinsic_memory_barrier_image:
+ case nir_intrinsic_memory_barrier_shared:
+ nir_builder_instr_insert(&b, &instr->instr);
+ break;
+ case nir_intrinsic_shader_clock:
+ nir_ssa_dest_init(&instr->instr, &instr->dest, 1, NULL);
+ nir_builder_instr_insert(&b, &instr->instr);
+ break;
+ case nir_intrinsic_store_ssbo: {
+ exec_node *param = ir->actual_parameters.get_head();
+ ir_rvalue *block = ((ir_instruction *)param)->as_rvalue();
+
+ param = param->get_next();
+ ir_rvalue *offset = ((ir_instruction *)param)->as_rvalue();
+
+ param = param->get_next();
+ ir_rvalue *val = ((ir_instruction *)param)->as_rvalue();
+
+ param = param->get_next();
+ ir_constant *write_mask = ((ir_instruction *)param)->as_constant();
+ assert(write_mask);
+
+ /* Check if we need the indirect version */
+ ir_constant *const_offset = offset->as_constant();
+ if (!const_offset) {
+ op = nir_intrinsic_store_ssbo_indirect;
+ ralloc_free(instr);
+ instr = nir_intrinsic_instr_create(shader, op);
+ instr->src[2] = nir_src_for_ssa(evaluate_rvalue(offset));
+ instr->const_index[0] = 0;
+ } else {
+ instr->const_index[0] = const_offset->value.u[0];
+ }
- nir_instr_insert_after_cf_list(this->cf_node_list, &instr->instr);
+ instr->const_index[1] = write_mask->value.u[0];
- nir_intrinsic_instr *store_instr =
- nir_intrinsic_instr_create(shader, nir_intrinsic_store_var);
- store_instr->num_components = 1;
+ instr->src[0] = nir_src_for_ssa(evaluate_rvalue(val));
+ instr->num_components = val->type->vector_elements;
- ir->return_deref->accept(this);
- store_instr->variables[0] = this->deref_head;
- store_instr->src[0].is_ssa = true;
- store_instr->src[0].ssa = &instr->dest.ssa;
+ instr->src[1] = nir_src_for_ssa(evaluate_rvalue(block));
+ nir_builder_instr_insert(&b, &instr->instr);
+ break;
+ }
+ case nir_intrinsic_load_ssbo: {
+ exec_node *param = ir->actual_parameters.get_head();
+ ir_rvalue *block = ((ir_instruction *)param)->as_rvalue();
+
+ param = param->get_next();
+ ir_rvalue *offset = ((ir_instruction *)param)->as_rvalue();
+
+ /* Check if we need the indirect version */
+ ir_constant *const_offset = offset->as_constant();
+ if (!const_offset) {
+ op = nir_intrinsic_load_ssbo_indirect;
+ ralloc_free(instr);
+ instr = nir_intrinsic_instr_create(shader, op);
+ instr->src[1] = nir_src_for_ssa(evaluate_rvalue(offset));
+ instr->const_index[0] = 0;
+ dest = &instr->dest;
+ } else {
+ instr->const_index[0] = const_offset->value.u[0];
+ }
- nir_instr_insert_after_cf_list(this->cf_node_list, &store_instr->instr);
+ instr->src[0] = nir_src_for_ssa(evaluate_rvalue(block));
+
+ const glsl_type *type = ir->return_deref->var->type;
+ instr->num_components = type->vector_elements;
+
+ /* Setup destination register */
+ nir_ssa_dest_init(&instr->instr, &instr->dest,
+ type->vector_elements, NULL);
+
+ /* Insert the created nir instruction now since in the case of boolean
+ * result we will need to emit another instruction after it
+ */
+ nir_builder_instr_insert(&b, &instr->instr);
+
+ /*
+ * In SSBO/UBO's, a true boolean value is any non-zero value, but we
+ * consider a true boolean to be ~0. Fix this up with a != 0
+ * comparison.
+ */
+ if (type->base_type == GLSL_TYPE_BOOL) {
+ nir_alu_instr *load_ssbo_compare =
+ nir_alu_instr_create(shader, nir_op_ine);
+ load_ssbo_compare->src[0].src.is_ssa = true;
+ load_ssbo_compare->src[0].src.ssa = &instr->dest.ssa;
+ load_ssbo_compare->src[1].src =
+ nir_src_for_ssa(nir_imm_int(&b, 0));
+ for (unsigned i = 0; i < type->vector_elements; i++)
+ load_ssbo_compare->src[1].swizzle[i] = 0;
+ nir_ssa_dest_init(&load_ssbo_compare->instr,
+ &load_ssbo_compare->dest.dest,
+ type->vector_elements, NULL);
+ load_ssbo_compare->dest.write_mask = (1 << type->vector_elements) - 1;
+ nir_builder_instr_insert(&b, &load_ssbo_compare->instr);
+ dest = &load_ssbo_compare->dest.dest;
+ }
+ break;
+ }
+ case nir_intrinsic_ssbo_atomic_add:
+ case nir_intrinsic_ssbo_atomic_imin:
+ case nir_intrinsic_ssbo_atomic_umin:
+ case nir_intrinsic_ssbo_atomic_imax:
+ case nir_intrinsic_ssbo_atomic_umax:
+ case nir_intrinsic_ssbo_atomic_and:
+ case nir_intrinsic_ssbo_atomic_or:
+ case nir_intrinsic_ssbo_atomic_xor:
+ case nir_intrinsic_ssbo_atomic_exchange:
+ case nir_intrinsic_ssbo_atomic_comp_swap: {
+ int param_count = ir->actual_parameters.length();
+ assert(param_count == 3 || param_count == 4);
+
+ /* Block index */
+ exec_node *param = ir->actual_parameters.get_head();
+ ir_instruction *inst = (ir_instruction *) param;
+ instr->src[0] = nir_src_for_ssa(evaluate_rvalue(inst->as_rvalue()));
+
+ /* Offset */
+ param = param->get_next();
+ inst = (ir_instruction *) param;
+ instr->src[1] = nir_src_for_ssa(evaluate_rvalue(inst->as_rvalue()));
+
+ /* data1 parameter (this is always present) */
+ param = param->get_next();
+ inst = (ir_instruction *) param;
+ instr->src[2] = nir_src_for_ssa(evaluate_rvalue(inst->as_rvalue()));
+
+ /* data2 parameter (only with atomic_comp_swap) */
+ if (param_count == 4) {
+ assert(op == nir_intrinsic_ssbo_atomic_comp_swap);
+ param = param->get_next();
+ inst = (ir_instruction *) param;
+ instr->src[3] = nir_src_for_ssa(evaluate_rvalue(inst->as_rvalue()));
+ }
+
+ /* Atomic result */
+ assert(ir->return_deref);
+ nir_ssa_dest_init(&instr->instr, &instr->dest,
+ ir->return_deref->type->vector_elements, NULL);
+ nir_builder_instr_insert(&b, &instr->instr);
+ break;
+ }
+ default:
+ unreachable("not reached");
+ }
+
+ if (ir->return_deref) {
+ nir_intrinsic_instr *store_instr =
+ nir_intrinsic_instr_create(shader, nir_intrinsic_store_var);
+ store_instr->num_components = ir->return_deref->type->vector_elements;
+
+ store_instr->variables[0] =
+ evaluate_deref(&store_instr->instr, ir->return_deref);
+ store_instr->src[0] = nir_src_for_ssa(&dest->ssa);
+
+ nir_builder_instr_insert(&b, &store_instr->instr);
+ }
return;
}
unsigned i = 0;
foreach_in_list(ir_dereference, param, &ir->actual_parameters) {
- param->accept(this);
- instr->params[i] = this->deref_head;
+ instr->params[i] = evaluate_deref(&instr->instr, param);
i++;
}
- ir->return_deref->accept(this);
- instr->return_deref = this->deref_head;
- nir_instr_insert_after_cf_list(this->cf_node_list, &instr->instr);
+ instr->return_deref = evaluate_deref(&instr->instr, ir->return_deref);
+ nir_builder_instr_insert(&b, &instr->instr);
}
void
nir_intrinsic_instr *copy =
nir_intrinsic_instr_create(this->shader, nir_intrinsic_copy_var);
- ir->lhs->accept(this);
- copy->variables[0] = this->deref_head;
-
- ir->rhs->accept(this);
- copy->variables[1] = this->deref_head;
-
+ copy->variables[0] = evaluate_deref(©->instr, ir->lhs);
+ copy->variables[1] = evaluate_deref(©->instr, ir->rhs);
if (ir->condition) {
nir_if *if_stmt = nir_if_create(this->shader);
- if_stmt->condition = evaluate_rvalue(ir->condition);
- nir_cf_node_insert_end(this->cf_node_list, &if_stmt->cf_node);
+ if_stmt->condition = nir_src_for_ssa(evaluate_rvalue(ir->condition));
+ nir_builder_cf_insert(&b, &if_stmt->cf_node);
nir_instr_insert_after_cf_list(&if_stmt->then_list, ©->instr);
+ b.cursor = nir_after_cf_node(&if_stmt->cf_node);
} else {
- nir_instr_insert_after_cf_list(this->cf_node_list, ©->instr);
+ nir_builder_instr_insert(&b, ©->instr);
}
return;
}
ir->lhs->accept(this);
nir_deref_var *lhs_deref = this->deref_head;
- nir_src src = evaluate_rvalue(ir->rhs);
+ nir_ssa_def *src = evaluate_rvalue(ir->rhs);
if (ir->write_mask != (1 << num_components) - 1 && ir->write_mask != 0) {
/*
load->num_components = ir->lhs->type->vector_elements;
nir_ssa_dest_init(&load->instr, &load->dest, num_components, NULL);
load->variables[0] = lhs_deref;
- nir_instr_insert_after_cf_list(this->cf_node_list, &load->instr);
-
- nir_op vec_op;
- switch (ir->lhs->type->vector_elements) {
- case 1: vec_op = nir_op_imov; break;
- case 2: vec_op = nir_op_vec2; break;
- case 3: vec_op = nir_op_vec3; break;
- case 4: vec_op = nir_op_vec4; break;
- default: unreachable("Invalid number of components"); break;
- }
- nir_alu_instr *vec = nir_alu_instr_create(this->shader, vec_op);
- nir_ssa_dest_init(&vec->instr, &vec->dest.dest, num_components, NULL);
- vec->dest.write_mask = (1 << num_components) - 1;
+ ralloc_steal(load, load->variables[0]);
+ nir_builder_instr_insert(&b, &load->instr);
+
+ nir_ssa_def *srcs[4];
unsigned component = 0;
for (unsigned i = 0; i < ir->lhs->type->vector_elements; i++) {
if (ir->write_mask & (1 << i)) {
- vec->src[i].src = src;
-
/* GLSL IR will give us the input to the write-masked assignment
* in a single packed vector. So, for example, if the
* writemask is xzw, then we have to swizzle x -> x, y -> z,
* and z -> w and get the y component from the load.
*/
- vec->src[i].swizzle[0] = component++;
+ srcs[i] = nir_channel(&b, src, component++);
} else {
- vec->src[i].src.is_ssa = true;
- vec->src[i].src.ssa = &load->dest.ssa;
- vec->src[i].swizzle[0] = i;
+ srcs[i] = nir_channel(&b, &load->dest.ssa, i);
}
}
- nir_instr_insert_after_cf_list(this->cf_node_list, &vec->instr);
-
- src.is_ssa = true;
- src.ssa = &vec->dest.dest.ssa;
+ src = nir_vec(&b, srcs, ir->lhs->type->vector_elements);
}
nir_intrinsic_instr *store =
nir_intrinsic_instr_create(this->shader, nir_intrinsic_store_var);
store->num_components = ir->lhs->type->vector_elements;
- nir_deref *store_deref = nir_copy_deref(this->shader, &lhs_deref->deref);
+ nir_deref *store_deref = nir_copy_deref(store, &lhs_deref->deref);
store->variables[0] = nir_deref_as_var(store_deref);
- store->src[0] = src;
+ store->src[0] = nir_src_for_ssa(src);
if (ir->condition) {
nir_if *if_stmt = nir_if_create(this->shader);
- if_stmt->condition = evaluate_rvalue(ir->condition);
- nir_cf_node_insert_end(this->cf_node_list, &if_stmt->cf_node);
+ if_stmt->condition = nir_src_for_ssa(evaluate_rvalue(ir->condition));
+ nir_builder_cf_insert(&b, &if_stmt->cf_node);
nir_instr_insert_after_cf_list(&if_stmt->then_list, &store->instr);
+ b.cursor = nir_after_cf_node(&if_stmt->cf_node);
} else {
- nir_instr_insert_after_cf_list(this->cf_node_list, &store->instr);
+ nir_builder_instr_insert(&b, &store->instr);
}
}
{
nir_dest *dest = get_instr_dest(instr);
- nir_ssa_dest_init(instr, dest, num_components, NULL);
+ if (dest)
+ nir_ssa_dest_init(instr, dest, num_components, NULL);
+
+ nir_builder_instr_insert(&b, instr);
- nir_instr_insert_after_cf_list(this->cf_node_list, instr);
- this->result = instr;
+ if (dest) {
+ assert(dest->is_ssa);
+ this->result = &dest->ssa;
+ }
}
-nir_src
+nir_ssa_def *
nir_visitor::evaluate_rvalue(ir_rvalue* ir)
{
ir->accept(this);
nir_intrinsic_instr_create(this->shader, nir_intrinsic_load_var);
load_instr->num_components = ir->type->vector_elements;
load_instr->variables[0] = this->deref_head;
+ ralloc_steal(load_instr, load_instr->variables[0]);
add_instr(&load_instr->instr, ir->type->vector_elements);
}
- nir_dest *dest = get_instr_dest(this->result);
-
- assert(dest->is_ssa);
- nir_src src;
- src.is_ssa = true;
- src.ssa = &dest->ssa;
-
- return src;
-}
-
-nir_alu_instr *
-nir_visitor::emit(nir_op op, unsigned dest_size, nir_src *srcs)
-{
- nir_alu_instr *instr = nir_alu_instr_create(this->shader, op);
- for (unsigned i = 0; i < nir_op_infos[op].num_inputs; i++)
- instr->src[i].src = srcs[i];
- instr->dest.write_mask = (1 << dest_size) - 1;
- add_instr(&instr->instr, dest_size);
- return instr;
-}
-
-nir_alu_instr *
-nir_visitor::emit(nir_op op, unsigned dest_size, nir_src src1)
-{
- assert(nir_op_infos[op].num_inputs == 1);
- return emit(op, dest_size, &src1);
-}
-
-nir_alu_instr *
-nir_visitor::emit(nir_op op, unsigned dest_size, nir_src src1,
- nir_src src2)
-{
- assert(nir_op_infos[op].num_inputs == 2);
- nir_src srcs[] = { src1, src2 };
- return emit(op, dest_size, srcs);
-}
-
-nir_alu_instr *
-nir_visitor::emit(nir_op op, unsigned dest_size, nir_src src1,
- nir_src src2, nir_src src3)
-{
- assert(nir_op_infos[op].num_inputs == 3);
- nir_src srcs[] = { src1, src2, src3 };
- return emit(op, dest_size, srcs);
+ return this->result;
}
void
nir_intrinsic_instr *load = nir_intrinsic_instr_create(this->shader, op);
load->num_components = ir->type->vector_elements;
load->const_index[0] = const_index ? const_index->value.u[0] : 0; /* base offset */
- load->const_index[1] = 1; /* number of vec4's */
- load->src[0] = evaluate_rvalue(ir->operands[0]);
+ load->src[0] = nir_src_for_ssa(evaluate_rvalue(ir->operands[0]));
if (!const_index)
- load->src[1] = evaluate_rvalue(ir->operands[1]);
+ load->src[1] = nir_src_for_ssa(evaluate_rvalue(ir->operands[1]));
add_instr(&load->instr, ir->type->vector_elements);
/*
* a true boolean to be ~0. Fix this up with a != 0 comparison.
*/
- if (ir->type->base_type == GLSL_TYPE_BOOL) {
- nir_load_const_instr *const_zero = nir_load_const_instr_create(shader, 1);
- const_zero->value.u[0] = 0;
- nir_instr_insert_after_cf_list(this->cf_node_list, &const_zero->instr);
-
- nir_alu_instr *compare = nir_alu_instr_create(shader, nir_op_ine);
- compare->src[0].src.is_ssa = true;
- compare->src[0].src.ssa = &load->dest.ssa;
- compare->src[1].src.is_ssa = true;
- compare->src[1].src.ssa = &const_zero->def;
- for (unsigned i = 0; i < ir->type->vector_elements; i++)
- compare->src[1].swizzle[i] = 0;
- compare->dest.write_mask = (1 << ir->type->vector_elements) - 1;
-
- add_instr(&compare->instr, ir->type->vector_elements);
- }
+ if (ir->type->base_type == GLSL_TYPE_BOOL)
+ this->result = nir_ine(&b, &load->dest.ssa, nir_imm_int(&b, 0));
return;
}
nir_intrinsic_instr *intrin = nir_intrinsic_instr_create(shader, op);
intrin->num_components = deref->type->vector_elements;
intrin->variables[0] = this->deref_head;
+ ralloc_steal(intrin, intrin->variables[0]);
if (intrin->intrinsic == nir_intrinsic_interp_var_at_offset ||
intrin->intrinsic == nir_intrinsic_interp_var_at_sample)
- intrin->src[0] = evaluate_rvalue(ir->operands[1]);
+ intrin->src[0] = nir_src_for_ssa(evaluate_rvalue(ir->operands[1]));
add_instr(&intrin->instr, deref->type->vector_elements);
if (swizzle) {
- nir_alu_instr *mov = nir_alu_instr_create(shader, nir_op_imov);
- mov->dest.write_mask = (1 << swizzle->type->vector_elements) - 1;
- mov->src[0].src.is_ssa = true;
- mov->src[0].src.ssa = &intrin->dest.ssa;
-
- mov->src[0].swizzle[0] = swizzle->mask.x;
- mov->src[0].swizzle[1] = swizzle->mask.y;
- mov->src[0].swizzle[2] = swizzle->mask.z;
- mov->src[0].swizzle[3] = swizzle->mask.w;
- for (unsigned i = deref->type->vector_elements; i < 4; i++)
- mov->src[0].swizzle[i] = 0;
-
- add_instr(&mov->instr, swizzle->type->vector_elements);
+ unsigned swiz[4] = {
+ swizzle->mask.x, swizzle->mask.y, swizzle->mask.z, swizzle->mask.w
+ };
+
+ result = nir_swizzle(&b, result, swiz,
+ swizzle->type->vector_elements, false);
}
return;
break;
}
- nir_src srcs[4];
+ nir_ssa_def *srcs[4];
for (unsigned i = 0; i < ir->get_num_operands(); i++)
srcs[i] = evaluate_rvalue(ir->operands[i]);
else
out_type = GLSL_TYPE_FLOAT;
- unsigned dest_size = ir->type->vector_elements;
-
- nir_alu_instr *instr;
- nir_op op;
-
switch (ir->operation) {
- case ir_unop_bit_not: emit(nir_op_inot, dest_size, srcs); break;
+ case ir_unop_bit_not: result = nir_inot(&b, srcs[0]); break;
case ir_unop_logic_not:
- emit(supports_ints ? nir_op_inot : nir_op_fnot, dest_size, srcs);
+ result = supports_ints ? nir_inot(&b, srcs[0]) : nir_fnot(&b, srcs[0]);
break;
case ir_unop_neg:
- instr = emit(types[0] == GLSL_TYPE_FLOAT ? nir_op_fneg : nir_op_ineg,
- dest_size, srcs);
+ result = (types[0] == GLSL_TYPE_FLOAT) ? nir_fneg(&b, srcs[0])
+ : nir_ineg(&b, srcs[0]);
break;
case ir_unop_abs:
- instr = emit(types[0] == GLSL_TYPE_FLOAT ? nir_op_fabs : nir_op_iabs,
- dest_size, srcs);
+ result = (types[0] == GLSL_TYPE_FLOAT) ? nir_fabs(&b, srcs[0])
+ : nir_iabs(&b, srcs[0]);
break;
case ir_unop_saturate:
assert(types[0] == GLSL_TYPE_FLOAT);
- instr = emit(nir_op_fsat, dest_size, srcs);
+ result = nir_fsat(&b, srcs[0]);
break;
case ir_unop_sign:
- emit(types[0] == GLSL_TYPE_FLOAT ? nir_op_fsign : nir_op_isign,
- dest_size, srcs);
- break;
- case ir_unop_rcp: emit(nir_op_frcp, dest_size, srcs); break;
- case ir_unop_rsq: emit(nir_op_frsq, dest_size, srcs); break;
- case ir_unop_sqrt: emit(nir_op_fsqrt, dest_size, srcs); break;
- case ir_unop_exp: emit(nir_op_fexp, dest_size, srcs); break;
- case ir_unop_log: emit(nir_op_flog, dest_size, srcs); break;
- case ir_unop_exp2: emit(nir_op_fexp2, dest_size, srcs); break;
- case ir_unop_log2: emit(nir_op_flog2, dest_size, srcs); break;
+ result = (types[0] == GLSL_TYPE_FLOAT) ? nir_fsign(&b, srcs[0])
+ : nir_isign(&b, srcs[0]);
+ break;
+ case ir_unop_rcp: result = nir_frcp(&b, srcs[0]); break;
+ case ir_unop_rsq: result = nir_frsq(&b, srcs[0]); break;
+ case ir_unop_sqrt: result = nir_fsqrt(&b, srcs[0]); break;
+ case ir_unop_exp: unreachable("ir_unop_exp should have been lowered");
+ case ir_unop_log: unreachable("ir_unop_log should have been lowered");
+ case ir_unop_exp2: result = nir_fexp2(&b, srcs[0]); break;
+ case ir_unop_log2: result = nir_flog2(&b, srcs[0]); break;
case ir_unop_i2f:
- emit(supports_ints ? nir_op_i2f : nir_op_fmov, dest_size, srcs);
+ result = supports_ints ? nir_i2f(&b, srcs[0]) : nir_fmov(&b, srcs[0]);
break;
case ir_unop_u2f:
- emit(supports_ints ? nir_op_u2f : nir_op_fmov, dest_size, srcs);
+ result = supports_ints ? nir_u2f(&b, srcs[0]) : nir_fmov(&b, srcs[0]);
break;
case ir_unop_b2f:
- emit(supports_ints ? nir_op_b2f : nir_op_fmov, dest_size, srcs);
+ result = supports_ints ? nir_b2f(&b, srcs[0]) : nir_fmov(&b, srcs[0]);
break;
- case ir_unop_f2i: emit(nir_op_f2i, dest_size, srcs); break;
- case ir_unop_f2u: emit(nir_op_f2u, dest_size, srcs); break;
- case ir_unop_f2b: emit(nir_op_f2b, dest_size, srcs); break;
- case ir_unop_i2b: emit(nir_op_i2b, dest_size, srcs); break;
- case ir_unop_b2i: emit(nir_op_b2i, dest_size, srcs); break;
+ case ir_unop_f2i: result = nir_f2i(&b, srcs[0]); break;
+ case ir_unop_f2u: result = nir_f2u(&b, srcs[0]); break;
+ case ir_unop_f2b: result = nir_f2b(&b, srcs[0]); break;
+ case ir_unop_i2b: result = nir_i2b(&b, srcs[0]); break;
+ case ir_unop_b2i: result = nir_b2i(&b, srcs[0]); break;
case ir_unop_i2u:
case ir_unop_u2i:
case ir_unop_bitcast_i2f:
case ir_unop_bitcast_f2i:
case ir_unop_bitcast_u2f:
case ir_unop_bitcast_f2u:
+ case ir_unop_subroutine_to_int:
/* no-op */
- emit(nir_op_imov, dest_size, srcs);
+ result = nir_imov(&b, srcs[0]);
break;
case ir_unop_any:
switch (ir->operands[0]->type->vector_elements) {
case 2:
- emit(supports_ints ? nir_op_bany2 : nir_op_fany2,
- dest_size, srcs);
+ result = supports_ints ? nir_bany2(&b, srcs[0])
+ : nir_fany2(&b, srcs[0]);
break;
case 3:
- emit(supports_ints ? nir_op_bany3 : nir_op_fany3,
- dest_size, srcs);
+ result = supports_ints ? nir_bany3(&b, srcs[0])
+ : nir_fany3(&b, srcs[0]);
break;
case 4:
- emit(supports_ints ? nir_op_bany4 : nir_op_fany4,
- dest_size, srcs);
+ result = supports_ints ? nir_bany4(&b, srcs[0])
+ : nir_fany4(&b, srcs[0]);
break;
default:
unreachable("not reached");
}
break;
- case ir_unop_trunc: emit(nir_op_ftrunc, dest_size, srcs); break;
- case ir_unop_ceil: emit(nir_op_fceil, dest_size, srcs); break;
- case ir_unop_floor: emit(nir_op_ffloor, dest_size, srcs); break;
- case ir_unop_fract: emit(nir_op_ffract, dest_size, srcs); break;
- case ir_unop_round_even: emit(nir_op_fround_even, dest_size, srcs); break;
- case ir_unop_sin: emit(nir_op_fsin, dest_size, srcs); break;
- case ir_unop_cos: emit(nir_op_fcos, dest_size, srcs); break;
- case ir_unop_sin_reduced:
- emit(nir_op_fsin_reduced, dest_size, srcs);
- break;
- case ir_unop_cos_reduced:
- emit(nir_op_fcos_reduced, dest_size, srcs);
- break;
- case ir_unop_dFdx: emit(nir_op_fddx, dest_size, srcs); break;
- case ir_unop_dFdy: emit(nir_op_fddy, dest_size, srcs); break;
- case ir_unop_dFdx_fine: emit(nir_op_fddx_fine, dest_size, srcs); break;
- case ir_unop_dFdy_fine: emit(nir_op_fddy_fine, dest_size, srcs); break;
- case ir_unop_dFdx_coarse: emit(nir_op_fddx_coarse, dest_size, srcs); break;
- case ir_unop_dFdy_coarse: emit(nir_op_fddy_coarse, dest_size, srcs); break;
+ case ir_unop_trunc: result = nir_ftrunc(&b, srcs[0]); break;
+ case ir_unop_ceil: result = nir_fceil(&b, srcs[0]); break;
+ case ir_unop_floor: result = nir_ffloor(&b, srcs[0]); break;
+ case ir_unop_fract: result = nir_ffract(&b, srcs[0]); break;
+ case ir_unop_round_even: result = nir_fround_even(&b, srcs[0]); break;
+ case ir_unop_sin: result = nir_fsin(&b, srcs[0]); break;
+ case ir_unop_cos: result = nir_fcos(&b, srcs[0]); break;
+ case ir_unop_dFdx: result = nir_fddx(&b, srcs[0]); break;
+ case ir_unop_dFdy: result = nir_fddy(&b, srcs[0]); break;
+ case ir_unop_dFdx_fine: result = nir_fddx_fine(&b, srcs[0]); break;
+ case ir_unop_dFdy_fine: result = nir_fddy_fine(&b, srcs[0]); break;
+ case ir_unop_dFdx_coarse: result = nir_fddx_coarse(&b, srcs[0]); break;
+ case ir_unop_dFdy_coarse: result = nir_fddy_coarse(&b, srcs[0]); break;
case ir_unop_pack_snorm_2x16:
- emit(nir_op_pack_snorm_2x16, dest_size, srcs);
+ result = nir_pack_snorm_2x16(&b, srcs[0]);
break;
case ir_unop_pack_snorm_4x8:
- emit(nir_op_pack_snorm_4x8, dest_size, srcs);
+ result = nir_pack_snorm_4x8(&b, srcs[0]);
break;
case ir_unop_pack_unorm_2x16:
- emit(nir_op_pack_unorm_2x16, dest_size, srcs);
+ result = nir_pack_unorm_2x16(&b, srcs[0]);
break;
case ir_unop_pack_unorm_4x8:
- emit(nir_op_pack_unorm_4x8, dest_size, srcs);
+ result = nir_pack_unorm_4x8(&b, srcs[0]);
break;
case ir_unop_pack_half_2x16:
- emit(nir_op_pack_half_2x16, dest_size, srcs);
+ result = nir_pack_half_2x16(&b, srcs[0]);
break;
case ir_unop_unpack_snorm_2x16:
- emit(nir_op_unpack_snorm_2x16, dest_size, srcs);
+ result = nir_unpack_snorm_2x16(&b, srcs[0]);
break;
case ir_unop_unpack_snorm_4x8:
- emit(nir_op_unpack_snorm_4x8, dest_size, srcs);
+ result = nir_unpack_snorm_4x8(&b, srcs[0]);
break;
case ir_unop_unpack_unorm_2x16:
- emit(nir_op_unpack_unorm_2x16, dest_size, srcs);
+ result = nir_unpack_unorm_2x16(&b, srcs[0]);
break;
case ir_unop_unpack_unorm_4x8:
- emit(nir_op_unpack_unorm_4x8, dest_size, srcs);
+ result = nir_unpack_unorm_4x8(&b, srcs[0]);
break;
case ir_unop_unpack_half_2x16:
- emit(nir_op_unpack_half_2x16, dest_size, srcs);
+ result = nir_unpack_half_2x16(&b, srcs[0]);
break;
case ir_unop_unpack_half_2x16_split_x:
- emit(nir_op_unpack_half_2x16_split_x, dest_size, srcs);
+ result = nir_unpack_half_2x16_split_x(&b, srcs[0]);
break;
case ir_unop_unpack_half_2x16_split_y:
- emit(nir_op_unpack_half_2x16_split_y, dest_size, srcs);
+ result = nir_unpack_half_2x16_split_y(&b, srcs[0]);
break;
case ir_unop_bitfield_reverse:
- emit(nir_op_bitfield_reverse, dest_size, srcs);
+ result = nir_bitfield_reverse(&b, srcs[0]);
break;
case ir_unop_bit_count:
- emit(nir_op_bit_count, dest_size, srcs);
+ result = nir_bit_count(&b, srcs[0]);
break;
case ir_unop_find_msb:
switch (types[0]) {
case GLSL_TYPE_UINT:
- emit(nir_op_ufind_msb, dest_size, srcs);
+ result = nir_ufind_msb(&b, srcs[0]);
break;
case GLSL_TYPE_INT:
- emit(nir_op_ifind_msb, dest_size, srcs);
+ result = nir_ifind_msb(&b, srcs[0]);
break;
default:
unreachable("Invalid type for findMSB()");
}
break;
case ir_unop_find_lsb:
- emit(nir_op_find_lsb, dest_size, srcs);
+ result = nir_find_lsb(&b, srcs[0]);
break;
case ir_unop_noise:
switch (ir->type->vector_elements) {
case 1:
switch (ir->operands[0]->type->vector_elements) {
- case 1: emit(nir_op_fnoise1_1, dest_size, srcs); break;
- case 2: emit(nir_op_fnoise1_2, dest_size, srcs); break;
- case 3: emit(nir_op_fnoise1_3, dest_size, srcs); break;
- case 4: emit(nir_op_fnoise1_4, dest_size, srcs); break;
+ case 1: result = nir_fnoise1_1(&b, srcs[0]); break;
+ case 2: result = nir_fnoise1_2(&b, srcs[0]); break;
+ case 3: result = nir_fnoise1_3(&b, srcs[0]); break;
+ case 4: result = nir_fnoise1_4(&b, srcs[0]); break;
default: unreachable("not reached");
}
break;
case 2:
switch (ir->operands[0]->type->vector_elements) {
- case 1: emit(nir_op_fnoise2_1, dest_size, srcs); break;
- case 2: emit(nir_op_fnoise2_2, dest_size, srcs); break;
- case 3: emit(nir_op_fnoise2_3, dest_size, srcs); break;
- case 4: emit(nir_op_fnoise2_4, dest_size, srcs); break;
+ case 1: result = nir_fnoise2_1(&b, srcs[0]); break;
+ case 2: result = nir_fnoise2_2(&b, srcs[0]); break;
+ case 3: result = nir_fnoise2_3(&b, srcs[0]); break;
+ case 4: result = nir_fnoise2_4(&b, srcs[0]); break;
default: unreachable("not reached");
}
break;
case 3:
switch (ir->operands[0]->type->vector_elements) {
- case 1: emit(nir_op_fnoise3_1, dest_size, srcs); break;
- case 2: emit(nir_op_fnoise3_2, dest_size, srcs); break;
- case 3: emit(nir_op_fnoise3_3, dest_size, srcs); break;
- case 4: emit(nir_op_fnoise3_4, dest_size, srcs); break;
+ case 1: result = nir_fnoise3_1(&b, srcs[0]); break;
+ case 2: result = nir_fnoise3_2(&b, srcs[0]); break;
+ case 3: result = nir_fnoise3_3(&b, srcs[0]); break;
+ case 4: result = nir_fnoise3_4(&b, srcs[0]); break;
default: unreachable("not reached");
}
break;
case 4:
switch (ir->operands[0]->type->vector_elements) {
- case 1: emit(nir_op_fnoise4_1, dest_size, srcs); break;
- case 2: emit(nir_op_fnoise4_2, dest_size, srcs); break;
- case 3: emit(nir_op_fnoise4_3, dest_size, srcs); break;
- case 4: emit(nir_op_fnoise4_4, dest_size, srcs); break;
+ case 1: result = nir_fnoise4_1(&b, srcs[0]); break;
+ case 2: result = nir_fnoise4_2(&b, srcs[0]); break;
+ case 3: result = nir_fnoise4_3(&b, srcs[0]); break;
+ case 4: result = nir_fnoise4_4(&b, srcs[0]); break;
default: unreachable("not reached");
}
break;
unreachable("not reached");
}
break;
+ case ir_unop_get_buffer_size: {
+ nir_intrinsic_instr *load = nir_intrinsic_instr_create(
+ this->shader,
+ nir_intrinsic_get_buffer_size);
+ load->num_components = ir->type->vector_elements;
+ load->src[0] = nir_src_for_ssa(evaluate_rvalue(ir->operands[0]));
+ add_instr(&load->instr, ir->type->vector_elements);
+ return;
+ }
+
case ir_binop_add:
+ result = (out_type == GLSL_TYPE_FLOAT) ? nir_fadd(&b, srcs[0], srcs[1])
+ : nir_iadd(&b, srcs[0], srcs[1]);
+ break;
case ir_binop_sub:
+ result = (out_type == GLSL_TYPE_FLOAT) ? nir_fsub(&b, srcs[0], srcs[1])
+ : nir_isub(&b, srcs[0], srcs[1]);
+ break;
case ir_binop_mul:
+ result = (out_type == GLSL_TYPE_FLOAT) ? nir_fmul(&b, srcs[0], srcs[1])
+ : nir_imul(&b, srcs[0], srcs[1]);
+ break;
case ir_binop_div:
+ if (out_type == GLSL_TYPE_FLOAT)
+ result = nir_fdiv(&b, srcs[0], srcs[1]);
+ else if (out_type == GLSL_TYPE_INT)
+ result = nir_idiv(&b, srcs[0], srcs[1]);
+ else
+ result = nir_udiv(&b, srcs[0], srcs[1]);
+ break;
case ir_binop_mod:
+ result = (out_type == GLSL_TYPE_FLOAT) ? nir_fmod(&b, srcs[0], srcs[1])
+ : nir_umod(&b, srcs[0], srcs[1]);
+ break;
case ir_binop_min:
+ if (out_type == GLSL_TYPE_FLOAT)
+ result = nir_fmin(&b, srcs[0], srcs[1]);
+ else if (out_type == GLSL_TYPE_INT)
+ result = nir_imin(&b, srcs[0], srcs[1]);
+ else
+ result = nir_umin(&b, srcs[0], srcs[1]);
+ break;
case ir_binop_max:
- case ir_binop_pow:
- case ir_binop_bit_and:
- case ir_binop_bit_or:
- case ir_binop_bit_xor:
- case ir_binop_lshift:
+ if (out_type == GLSL_TYPE_FLOAT)
+ result = nir_fmax(&b, srcs[0], srcs[1]);
+ else if (out_type == GLSL_TYPE_INT)
+ result = nir_imax(&b, srcs[0], srcs[1]);
+ else
+ result = nir_umax(&b, srcs[0], srcs[1]);
+ break;
+ case ir_binop_pow: result = nir_fpow(&b, srcs[0], srcs[1]); break;
+ case ir_binop_bit_and: result = nir_iand(&b, srcs[0], srcs[1]); break;
+ case ir_binop_bit_or: result = nir_ior(&b, srcs[0], srcs[1]); break;
+ case ir_binop_bit_xor: result = nir_ixor(&b, srcs[0], srcs[1]); break;
+ case ir_binop_logic_and:
+ result = supports_ints ? nir_iand(&b, srcs[0], srcs[1])
+ : nir_fand(&b, srcs[0], srcs[1]);
+ break;
+ case ir_binop_logic_or:
+ result = supports_ints ? nir_ior(&b, srcs[0], srcs[1])
+ : nir_for(&b, srcs[0], srcs[1]);
+ break;
+ case ir_binop_logic_xor:
+ result = supports_ints ? nir_ixor(&b, srcs[0], srcs[1])
+ : nir_fxor(&b, srcs[0], srcs[1]);
+ break;
+ case ir_binop_lshift: result = nir_ishl(&b, srcs[0], srcs[1]); break;
case ir_binop_rshift:
- switch (ir->operation) {
- case ir_binop_add:
- if (out_type == GLSL_TYPE_FLOAT)
- op = nir_op_fadd;
- else
- op = nir_op_iadd;
- break;
- case ir_binop_sub:
- if (out_type == GLSL_TYPE_FLOAT)
- op = nir_op_fsub;
- else
- op = nir_op_isub;
- break;
- case ir_binop_mul:
- if (out_type == GLSL_TYPE_FLOAT)
- op = nir_op_fmul;
- else
- op = nir_op_imul;
- break;
- case ir_binop_div:
- if (out_type == GLSL_TYPE_FLOAT)
- op = nir_op_fdiv;
- else if (out_type == GLSL_TYPE_INT)
- op = nir_op_idiv;
- else
- op = nir_op_udiv;
- break;
- case ir_binop_mod:
- if (out_type == GLSL_TYPE_FLOAT)
- op = nir_op_fmod;
- else
- op = nir_op_umod;
- break;
- case ir_binop_min:
- if (out_type == GLSL_TYPE_FLOAT)
- op = nir_op_fmin;
- else if (out_type == GLSL_TYPE_INT)
- op = nir_op_imin;
- else
- op = nir_op_umin;
- break;
- case ir_binop_max:
- if (out_type == GLSL_TYPE_FLOAT)
- op = nir_op_fmax;
- else if (out_type == GLSL_TYPE_INT)
- op = nir_op_imax;
- else
- op = nir_op_umax;
- break;
- case ir_binop_bit_and:
- op = nir_op_iand;
- break;
- case ir_binop_bit_or:
- op = nir_op_ior;
- break;
- case ir_binop_bit_xor:
- op = nir_op_ixor;
- break;
- case ir_binop_lshift:
- op = nir_op_ishl;
- break;
- case ir_binop_rshift:
- if (out_type == GLSL_TYPE_INT)
- op = nir_op_ishr;
- else
- op = nir_op_ushr;
- break;
- case ir_binop_pow:
- op = nir_op_fpow;
- break;
-
- default:
- unreachable("not reached");
- }
-
- instr = emit(op, dest_size, srcs);
-
- if (ir->operands[0]->type->vector_elements != 1 &&
- ir->operands[1]->type->vector_elements == 1) {
- for (unsigned i = 0; i < ir->operands[0]->type->vector_elements;
- i++) {
- instr->src[1].swizzle[i] = 0;
- }
- }
-
- if (ir->operands[1]->type->vector_elements != 1 &&
- ir->operands[0]->type->vector_elements == 1) {
- for (unsigned i = 0; i < ir->operands[1]->type->vector_elements;
- i++) {
- instr->src[0].swizzle[i] = 0;
- }
- }
-
+ result = (out_type == GLSL_TYPE_INT) ? nir_ishr(&b, srcs[0], srcs[1])
+ : nir_ushr(&b, srcs[0], srcs[1]);
break;
case ir_binop_imul_high:
- emit(out_type == GLSL_TYPE_UINT ? nir_op_umul_high : nir_op_imul_high,
- dest_size, srcs);
+ result = (out_type == GLSL_TYPE_INT) ? nir_imul_high(&b, srcs[0], srcs[1])
+ : nir_umul_high(&b, srcs[0], srcs[1]);
break;
- case ir_binop_carry: emit(nir_op_uadd_carry, dest_size, srcs); break;
- case ir_binop_borrow: emit(nir_op_usub_borrow, dest_size, srcs); break;
+ case ir_binop_carry: result = nir_uadd_carry(&b, srcs[0], srcs[1]); break;
+ case ir_binop_borrow: result = nir_usub_borrow(&b, srcs[0], srcs[1]); break;
case ir_binop_less:
if (supports_ints) {
if (types[0] == GLSL_TYPE_FLOAT)
- emit(nir_op_flt, dest_size, srcs);
+ result = nir_flt(&b, srcs[0], srcs[1]);
else if (types[0] == GLSL_TYPE_INT)
- emit(nir_op_ilt, dest_size, srcs);
+ result = nir_ilt(&b, srcs[0], srcs[1]);
else
- emit(nir_op_ult, dest_size, srcs);
+ result = nir_ult(&b, srcs[0], srcs[1]);
} else {
- emit(nir_op_slt, dest_size, srcs);
+ result = nir_slt(&b, srcs[0], srcs[1]);
}
break;
case ir_binop_greater:
if (supports_ints) {
if (types[0] == GLSL_TYPE_FLOAT)
- emit(nir_op_flt, dest_size, srcs[1], srcs[0]);
+ result = nir_flt(&b, srcs[1], srcs[0]);
else if (types[0] == GLSL_TYPE_INT)
- emit(nir_op_ilt, dest_size, srcs[1], srcs[0]);
+ result = nir_ilt(&b, srcs[1], srcs[0]);
else
- emit(nir_op_ult, dest_size, srcs[1], srcs[0]);
+ result = nir_ult(&b, srcs[1], srcs[0]);
} else {
- emit(nir_op_slt, dest_size, srcs[1], srcs[0]);
+ result = nir_slt(&b, srcs[1], srcs[0]);
}
break;
case ir_binop_lequal:
if (supports_ints) {
if (types[0] == GLSL_TYPE_FLOAT)
- emit(nir_op_fge, dest_size, srcs[1], srcs[0]);
+ result = nir_fge(&b, srcs[1], srcs[0]);
else if (types[0] == GLSL_TYPE_INT)
- emit(nir_op_ige, dest_size, srcs[1], srcs[0]);
+ result = nir_ige(&b, srcs[1], srcs[0]);
else
- emit(nir_op_uge, dest_size, srcs[1], srcs[0]);
+ result = nir_uge(&b, srcs[1], srcs[0]);
} else {
- emit(nir_op_slt, dest_size, srcs[1], srcs[0]);
+ result = nir_slt(&b, srcs[1], srcs[0]);
}
break;
case ir_binop_gequal:
if (supports_ints) {
if (types[0] == GLSL_TYPE_FLOAT)
- emit(nir_op_fge, dest_size, srcs);
+ result = nir_fge(&b, srcs[0], srcs[1]);
else if (types[0] == GLSL_TYPE_INT)
- emit(nir_op_ige, dest_size, srcs);
+ result = nir_ige(&b, srcs[0], srcs[1]);
else
- emit(nir_op_uge, dest_size, srcs);
+ result = nir_uge(&b, srcs[0], srcs[1]);
} else {
- emit(nir_op_slt, dest_size, srcs);
+ result = nir_slt(&b, srcs[0], srcs[1]);
}
break;
case ir_binop_equal:
if (supports_ints) {
if (types[0] == GLSL_TYPE_FLOAT)
- emit(nir_op_feq, dest_size, srcs);
+ result = nir_feq(&b, srcs[0], srcs[1]);
else
- emit(nir_op_ieq, dest_size, srcs);
+ result = nir_ieq(&b, srcs[0], srcs[1]);
} else {
- emit(nir_op_seq, dest_size, srcs);
+ result = nir_seq(&b, srcs[0], srcs[1]);
}
break;
case ir_binop_nequal:
if (supports_ints) {
if (types[0] == GLSL_TYPE_FLOAT)
- emit(nir_op_fne, dest_size, srcs);
+ result = nir_fne(&b, srcs[0], srcs[1]);
else
- emit(nir_op_ine, dest_size, srcs);
+ result = nir_ine(&b, srcs[0], srcs[1]);
} else {
- emit(nir_op_sne, dest_size, srcs);
+ result = nir_sne(&b, srcs[0], srcs[1]);
}
break;
case ir_binop_all_equal:
if (supports_ints) {
if (types[0] == GLSL_TYPE_FLOAT) {
switch (ir->operands[0]->type->vector_elements) {
- case 1: emit(nir_op_feq, dest_size, srcs); break;
- case 2: emit(nir_op_ball_fequal2, dest_size, srcs); break;
- case 3: emit(nir_op_ball_fequal3, dest_size, srcs); break;
- case 4: emit(nir_op_ball_fequal4, dest_size, srcs); break;
+ case 1: result = nir_feq(&b, srcs[0], srcs[1]); break;
+ case 2: result = nir_ball_fequal2(&b, srcs[0], srcs[1]); break;
+ case 3: result = nir_ball_fequal3(&b, srcs[0], srcs[1]); break;
+ case 4: result = nir_ball_fequal4(&b, srcs[0], srcs[1]); break;
default:
unreachable("not reached");
}
} else {
switch (ir->operands[0]->type->vector_elements) {
- case 1: emit(nir_op_ieq, dest_size, srcs); break;
- case 2: emit(nir_op_ball_iequal2, dest_size, srcs); break;
- case 3: emit(nir_op_ball_iequal3, dest_size, srcs); break;
- case 4: emit(nir_op_ball_iequal4, dest_size, srcs); break;
+ case 1: result = nir_ieq(&b, srcs[0], srcs[1]); break;
+ case 2: result = nir_ball_iequal2(&b, srcs[0], srcs[1]); break;
+ case 3: result = nir_ball_iequal3(&b, srcs[0], srcs[1]); break;
+ case 4: result = nir_ball_iequal4(&b, srcs[0], srcs[1]); break;
default:
unreachable("not reached");
}
}
} else {
switch (ir->operands[0]->type->vector_elements) {
- case 1: emit(nir_op_seq, dest_size, srcs); break;
- case 2: emit(nir_op_fall_equal2, dest_size, srcs); break;
- case 3: emit(nir_op_fall_equal3, dest_size, srcs); break;
- case 4: emit(nir_op_fall_equal4, dest_size, srcs); break;
+ case 1: result = nir_seq(&b, srcs[0], srcs[1]); break;
+ case 2: result = nir_fall_equal2(&b, srcs[0], srcs[1]); break;
+ case 3: result = nir_fall_equal3(&b, srcs[0], srcs[1]); break;
+ case 4: result = nir_fall_equal4(&b, srcs[0], srcs[1]); break;
default:
unreachable("not reached");
}
if (supports_ints) {
if (types[0] == GLSL_TYPE_FLOAT) {
switch (ir->operands[0]->type->vector_elements) {
- case 1: emit(nir_op_fne, dest_size, srcs); break;
- case 2: emit(nir_op_bany_fnequal2, dest_size, srcs); break;
- case 3: emit(nir_op_bany_fnequal3, dest_size, srcs); break;
- case 4: emit(nir_op_bany_fnequal4, dest_size, srcs); break;
+ case 1: result = nir_fne(&b, srcs[0], srcs[1]); break;
+ case 2: result = nir_bany_fnequal2(&b, srcs[0], srcs[1]); break;
+ case 3: result = nir_bany_fnequal3(&b, srcs[0], srcs[1]); break;
+ case 4: result = nir_bany_fnequal4(&b, srcs[0], srcs[1]); break;
default:
unreachable("not reached");
}
} else {
switch (ir->operands[0]->type->vector_elements) {
- case 1: emit(nir_op_ine, dest_size, srcs); break;
- case 2: emit(nir_op_bany_inequal2, dest_size, srcs); break;
- case 3: emit(nir_op_bany_inequal3, dest_size, srcs); break;
- case 4: emit(nir_op_bany_inequal4, dest_size, srcs); break;
+ case 1: result = nir_ine(&b, srcs[0], srcs[1]); break;
+ case 2: result = nir_bany_inequal2(&b, srcs[0], srcs[1]); break;
+ case 3: result = nir_bany_inequal3(&b, srcs[0], srcs[1]); break;
+ case 4: result = nir_bany_inequal4(&b, srcs[0], srcs[1]); break;
default:
unreachable("not reached");
}
}
} else {
switch (ir->operands[0]->type->vector_elements) {
- case 1: emit(nir_op_sne, dest_size, srcs); break;
- case 2: emit(nir_op_fany_nequal2, dest_size, srcs); break;
- case 3: emit(nir_op_fany_nequal3, dest_size, srcs); break;
- case 4: emit(nir_op_fany_nequal4, dest_size, srcs); break;
+ case 1: result = nir_sne(&b, srcs[0], srcs[1]); break;
+ case 2: result = nir_fany_nequal2(&b, srcs[0], srcs[1]); break;
+ case 3: result = nir_fany_nequal3(&b, srcs[0], srcs[1]); break;
+ case 4: result = nir_fany_nequal4(&b, srcs[0], srcs[1]); break;
default:
unreachable("not reached");
}
}
break;
- case ir_binop_logic_and:
- if (supports_ints)
- emit(nir_op_iand, dest_size, srcs);
- else
- emit(nir_op_fand, dest_size, srcs);
- break;
- case ir_binop_logic_or:
- if (supports_ints)
- emit(nir_op_ior, dest_size, srcs);
- else
- emit(nir_op_for, dest_size, srcs);
- break;
- case ir_binop_logic_xor:
- if (supports_ints)
- emit(nir_op_ixor, dest_size, srcs);
- else
- emit(nir_op_fxor, dest_size, srcs);
- break;
case ir_binop_dot:
switch (ir->operands[0]->type->vector_elements) {
- case 2: emit(nir_op_fdot2, dest_size, srcs); break;
- case 3: emit(nir_op_fdot3, dest_size, srcs); break;
- case 4: emit(nir_op_fdot4, dest_size, srcs); break;
+ case 2: result = nir_fdot2(&b, srcs[0], srcs[1]); break;
+ case 3: result = nir_fdot3(&b, srcs[0], srcs[1]); break;
+ case 4: result = nir_fdot4(&b, srcs[0], srcs[1]); break;
default:
unreachable("not reached");
}
break;
case ir_binop_pack_half_2x16_split:
- emit(nir_op_pack_half_2x16_split, dest_size, srcs);
+ result = nir_pack_half_2x16_split(&b, srcs[0], srcs[1]);
break;
- case ir_binop_bfm: emit(nir_op_bfm, dest_size, srcs); break;
- case ir_binop_ldexp: emit(nir_op_ldexp, dest_size, srcs); break;
- case ir_triop_fma: emit(nir_op_ffma, dest_size, srcs); break;
+ case ir_binop_bfm: result = nir_bfm(&b, srcs[0], srcs[1]); break;
+ case ir_binop_ldexp: result = nir_ldexp(&b, srcs[0], srcs[1]); break;
+ case ir_triop_fma:
+ result = nir_ffma(&b, srcs[0], srcs[1], srcs[2]);
+ break;
case ir_triop_lrp:
- instr = emit(nir_op_flrp, dest_size, srcs);
- if (ir->operands[0]->type->vector_elements != 1 &&
- ir->operands[2]->type->vector_elements == 1) {
- for (unsigned i = 0; i < ir->operands[0]->type->vector_elements;
- i++) {
- instr->src[2].swizzle[i] = 0;
- }
- }
+ result = nir_flrp(&b, srcs[0], srcs[1], srcs[2]);
break;
case ir_triop_csel:
if (supports_ints)
- emit(nir_op_bcsel, dest_size, srcs);
+ result = nir_bcsel(&b, srcs[0], srcs[1], srcs[2]);
else
- emit(nir_op_fcsel, dest_size, srcs);
+ result = nir_fcsel(&b, srcs[0], srcs[1], srcs[2]);
break;
case ir_triop_bfi:
- instr = emit(nir_op_bfi, dest_size, srcs);
- for (unsigned i = 0; i < ir->operands[1]->type->vector_elements; i++) {
- instr->src[0].swizzle[i] = 0;
- }
+ result = nir_bfi(&b, srcs[0], srcs[1], srcs[2]);
break;
case ir_triop_bitfield_extract:
- instr = emit(out_type == GLSL_TYPE_INT ? nir_op_ibitfield_extract :
- nir_op_ubitfield_extract, dest_size, srcs);
- for (unsigned i = 0; i < ir->operands[0]->type->vector_elements; i++) {
- instr->src[1].swizzle[i] = 0;
- instr->src[2].swizzle[i] = 0;
- }
+ result = (out_type == GLSL_TYPE_INT) ?
+ nir_ibitfield_extract(&b, srcs[0], srcs[1], srcs[2]) :
+ nir_ubitfield_extract(&b, srcs[0], srcs[1], srcs[2]);
break;
case ir_quadop_bitfield_insert:
- instr = emit(nir_op_bitfield_insert, dest_size, srcs);
- for (unsigned i = 0; i < ir->operands[0]->type->vector_elements; i++) {
- instr->src[2].swizzle[i] = 0;
- instr->src[3].swizzle[i] = 0;
- }
+ result = nir_bitfield_insert(&b, srcs[0], srcs[1], srcs[2], srcs[3]);
break;
case ir_quadop_vector:
- switch (ir->type->vector_elements) {
- case 2: emit(nir_op_vec2, dest_size, srcs); break;
- case 3: emit(nir_op_vec3, dest_size, srcs); break;
- case 4: emit(nir_op_vec4, dest_size, srcs); break;
- default: unreachable("not reached");
- }
+ result = nir_vec(&b, srcs, ir->type->vector_elements);
break;
default:
void
nir_visitor::visit(ir_swizzle *ir)
{
- nir_alu_instr *instr = emit(supports_ints ? nir_op_imov : nir_op_fmov,
- ir->type->vector_elements,
- evaluate_rvalue(ir->val));
-
unsigned swizzle[4] = { ir->mask.x, ir->mask.y, ir->mask.z, ir->mask.w };
- for (unsigned i = 0; i < ir->type->vector_elements; i++)
- instr->src[0].swizzle[i] = swizzle[i];
+ result = nir_swizzle(&b, evaluate_rvalue(ir->val), swizzle,
+ ir->type->vector_elements, !supports_ints);
}
void
num_srcs = 0;
break;
+ case ir_texture_samples:
+ op = nir_texop_texture_samples;
+ num_srcs = 0;
+ break;
+
default:
unreachable("not reached");
}
unreachable("not reached");
}
- ir->sampler->accept(this);
- instr->sampler = this->deref_head;
+ instr->sampler = evaluate_deref(&instr->instr, ir->sampler);
unsigned src_number = 0;
if (ir->coordinate != NULL) {
instr->coord_components = ir->coordinate->type->vector_elements;
- instr->src[src_number].src = evaluate_rvalue(ir->coordinate);
+ instr->src[src_number].src =
+ nir_src_for_ssa(evaluate_rvalue(ir->coordinate));
instr->src[src_number].src_type = nir_tex_src_coord;
src_number++;
}
if (ir->projector != NULL) {
- instr->src[src_number].src = evaluate_rvalue(ir->projector);
+ instr->src[src_number].src =
+ nir_src_for_ssa(evaluate_rvalue(ir->projector));
instr->src[src_number].src_type = nir_tex_src_projector;
src_number++;
}
if (ir->shadow_comparitor != NULL) {
- instr->src[src_number].src = evaluate_rvalue(ir->shadow_comparitor);
+ instr->src[src_number].src =
+ nir_src_for_ssa(evaluate_rvalue(ir->shadow_comparitor));
instr->src[src_number].src_type = nir_tex_src_comparitor;
src_number++;
}
for (unsigned i = 0; i < const_offset->type->vector_elements; i++)
instr->const_offset[i] = const_offset->value.i[i];
} else {
- instr->src[src_number].src = evaluate_rvalue(ir->offset);
+ instr->src[src_number].src =
+ nir_src_for_ssa(evaluate_rvalue(ir->offset));
instr->src[src_number].src_type = nir_tex_src_offset;
src_number++;
}
switch (ir->op) {
case ir_txb:
- instr->src[src_number].src = evaluate_rvalue(ir->lod_info.bias);
+ instr->src[src_number].src =
+ nir_src_for_ssa(evaluate_rvalue(ir->lod_info.bias));
instr->src[src_number].src_type = nir_tex_src_bias;
src_number++;
break;
case ir_txf:
case ir_txs:
if (ir->lod_info.lod != NULL) {
- instr->src[src_number].src = evaluate_rvalue(ir->lod_info.lod);
+ instr->src[src_number].src =
+ nir_src_for_ssa(evaluate_rvalue(ir->lod_info.lod));
instr->src[src_number].src_type = nir_tex_src_lod;
src_number++;
}
break;
case ir_txd:
- instr->src[src_number].src = evaluate_rvalue(ir->lod_info.grad.dPdx);
+ instr->src[src_number].src =
+ nir_src_for_ssa(evaluate_rvalue(ir->lod_info.grad.dPdx));
instr->src[src_number].src_type = nir_tex_src_ddx;
src_number++;
- instr->src[src_number].src = evaluate_rvalue(ir->lod_info.grad.dPdy);
+ instr->src[src_number].src =
+ nir_src_for_ssa(evaluate_rvalue(ir->lod_info.grad.dPdy));
instr->src[src_number].src_type = nir_tex_src_ddy;
src_number++;
break;
case ir_txf_ms:
- instr->src[src_number].src = evaluate_rvalue(ir->lod_info.sample_index);
+ instr->src[src_number].src =
+ nir_src_for_ssa(evaluate_rvalue(ir->lod_info.sample_index));
instr->src[src_number].src_type = nir_tex_src_ms_index;
src_number++;
break;
* constant initializer and return a dereference.
*/
- nir_variable *var = ralloc(this->shader, nir_variable);
- var->name = ralloc_strdup(var, "const_temp");
- var->type = ir->type;
- var->data.mode = nir_var_local;
+ nir_variable *var =
+ nir_local_variable_create(this->impl, ir->type, "const_temp");
var->data.read_only = true;
var->constant_initializer = constant_copy(ir, var);
- exec_list_push_tail(&this->impl->locals, &var->node);
this->deref_head = nir_deref_var_create(this->shader, var);
this->deref_tail = &this->deref_head->deref;
int field_index = this->deref_tail->type->field_index(ir->field);
assert(field_index >= 0);
- nir_deref_struct *deref = nir_deref_struct_create(this->shader, field_index);
+ nir_deref_struct *deref = nir_deref_struct_create(this->deref_tail, field_index);
deref->deref.type = ir->type;
this->deref_tail->child = &deref->deref;
this->deref_tail = &deref->deref;
deref->base_offset = const_index->value.u[0];
} else {
deref->deref_array_type = nir_deref_array_type_indirect;
- deref->indirect = evaluate_rvalue(ir->array_index);
+ deref->indirect =
+ nir_src_for_ssa(evaluate_rvalue(ir->array_index));
}
ir->array->accept(this);
this->deref_tail->child = &deref->deref;
+ ralloc_steal(this->deref_tail, deref);
this->deref_tail = &deref->deref;
}
+
+void
+nir_visitor::visit(ir_barrier *ir)
+{
+ nir_intrinsic_instr *instr =
+ nir_intrinsic_instr_create(this->shader, nir_intrinsic_barrier);
+ nir_builder_instr_insert(&b, &instr->instr);
+}