class lower_packed_varyings_visitor
{
public:
- lower_packed_varyings_visitor(void *mem_ctx, unsigned locations_used,
+ lower_packed_varyings_visitor(void *mem_ctx,
+ unsigned locations_used,
+ const uint8_t *components,
ir_variable_mode mode,
unsigned gs_input_vertices,
exec_list *out_instructions,
bool disable_varying_packing,
bool xfb_enabled);
- void run(struct gl_shader *shader);
+ void run(struct gl_linked_shader *shader);
private:
void bitwise_assign_pack(ir_rvalue *lhs, ir_rvalue *rhs);
*/
const unsigned locations_used;
+ const uint8_t* components;
+
/**
* Array of pointers to the packed varyings that have been created for each
* generic varying slot. NULL entries in this array indicate varying slots
} /* anonymous namespace */
lower_packed_varyings_visitor::lower_packed_varyings_visitor(
- void *mem_ctx, unsigned locations_used, ir_variable_mode mode,
+ void *mem_ctx, unsigned locations_used, const uint8_t *components,
+ ir_variable_mode mode,
unsigned gs_input_vertices, exec_list *out_instructions,
exec_list *out_variables, bool disable_varying_packing,
bool xfb_enabled)
: mem_ctx(mem_ctx),
locations_used(locations_used),
+ components(components),
packed_varyings((ir_variable **)
rzalloc_array_size(mem_ctx, sizeof(*packed_varyings),
locations_used)),
}
void
-lower_packed_varyings_visitor::run(struct gl_shader *shader)
+lower_packed_varyings_visitor::run(struct gl_linked_shader *shader)
{
foreach_in_list(ir_instruction, node, shader->ir) {
ir_variable *var = node->as_variable();
continue;
/* This lowering pass is only capable of packing floats and ints
- * together when their interpolation mode is "flat". Therefore, to be
- * safe, caller should ensure that integral varyings always use flat
- * interpolation, even when this is not required by GLSL.
+ * together when their interpolation mode is "flat". Treat integers as
+ * being flat when the interpolation mode is none.
*/
- assert(var->data.interpolation == INTERP_QUALIFIER_FLAT ||
+ assert(var->data.interpolation == INTERP_MODE_FLAT ||
+ var->data.interpolation == INTERP_MODE_NONE ||
!var->type->contains_integer());
/* Clone the variable for program resource list before
bool gs_input_toplevel,
unsigned vertex_index)
{
- unsigned dmul = rvalue->type->is_double() ? 2 : 1;
+ unsigned dmul = rvalue->type->is_64bit() ? 2 : 1;
/* When gs_input_toplevel is set, we should be looking at a geometry shader
* input array.
*/
char right_swizzle_name[4] = { 0, 0, 0, 0 };
left_components = 4 - fine_location % 4;
- if (rvalue->type->is_double()) {
+ if (rvalue->type->is_64bit()) {
/* We might actually end up with 0 left components! */
left_components /= 2;
}
ir_dereference *packed_deref =
this->get_packed_varying_deref(location, unpacked_var, name,
vertex_index);
+ if (unpacked_var->data.stream != 0) {
+ assert(unpacked_var->data.stream < 4);
+ ir_variable *packed_var = packed_deref->variable_referenced();
+ for (unsigned i = 0; i < components; ++i) {
+ packed_var->data.stream |=
+ unpacked_var->data.stream << (2 * (location_frac + i));
+ }
+ }
ir_swizzle *swizzle = new(this->mem_ctx)
ir_swizzle(packed_deref, swizzle_values, components);
if (this->mode == ir_var_shader_out) {
if (this->packed_varyings[slot] == NULL) {
char *packed_name = ralloc_asprintf(this->mem_ctx, "packed:%s", name);
const glsl_type *packed_type;
- if (unpacked_var->data.interpolation == INTERP_QUALIFIER_FLAT)
- packed_type = glsl_type::ivec4_type;
+ assert(components[slot] != 0);
+ if (unpacked_var->is_interpolation_flat())
+ packed_type = glsl_type::get_instance(GLSL_TYPE_INT, components[slot], 1);
else
- packed_type = glsl_type::vec4_type;
+ packed_type = glsl_type::get_instance(GLSL_TYPE_FLOAT, components[slot], 1);
if (this->gs_input_vertices != 0) {
packed_type =
glsl_type::get_array_instance(packed_type,
packed_var->data.centroid = unpacked_var->data.centroid;
packed_var->data.sample = unpacked_var->data.sample;
packed_var->data.patch = unpacked_var->data.patch;
- packed_var->data.interpolation = unpacked_var->data.interpolation;
+ packed_var->data.interpolation = packed_type == glsl_type::ivec4_type
+ ? unsigned(INTERP_MODE_FLAT) : unpacked_var->data.interpolation;
packed_var->data.location = location;
packed_var->data.precision = unpacked_var->data.precision;
packed_var->data.always_active_io = unpacked_var->data.always_active_io;
+ packed_var->data.stream = 1u << 31;
unpacked_var->insert_before(packed_var);
this->packed_varyings[slot] = packed_var;
} else {
* first time we visit each component.
*/
if (this->gs_input_vertices == 0 || vertex_index == 0) {
- ralloc_asprintf_append((char **) &this->packed_varyings[slot]->name,
- ",%s", name);
+ ir_variable *var = this->packed_varyings[slot];
+
+ if (var->is_name_ralloced())
+ ralloc_asprintf_append((char **) &var->name, ",%s", name);
+ else
+ var->name = ralloc_asprintf(var, "%s,%s", var->name, name);
}
}
return false;
type = type->without_array();
- if (type->vector_elements == 4 && !type->is_double())
+ if (type->vector_elements == 4 && !type->is_64bit())
return false;
return true;
}
return visit_continue;
}
+/**
+ * Visitor that splices varying packing code before every return.
+ */
+class lower_packed_varyings_return_splicer : public ir_hierarchical_visitor
+{
+public:
+ explicit lower_packed_varyings_return_splicer(void *mem_ctx,
+ const exec_list *instructions);
+
+ virtual ir_visitor_status visit_leave(ir_return *ret);
+
+private:
+ /**
+ * Memory context used to allocate new instructions for the shader.
+ */
+ void * const mem_ctx;
+
+ /**
+ * Instructions that should be spliced into place before each return.
+ */
+ const exec_list *instructions;
+};
+
+
+lower_packed_varyings_return_splicer::lower_packed_varyings_return_splicer(
+ void *mem_ctx, const exec_list *instructions)
+ : mem_ctx(mem_ctx), instructions(instructions)
+{
+}
+
+
+ir_visitor_status
+lower_packed_varyings_return_splicer::visit_leave(ir_return *ret)
+{
+ foreach_in_list(ir_instruction, ir, this->instructions) {
+ ret->insert_before(ir->clone(this->mem_ctx, NULL));
+ }
+ return visit_continue;
+}
void
lower_packed_varyings(void *mem_ctx, unsigned locations_used,
+ const uint8_t *components,
ir_variable_mode mode, unsigned gs_input_vertices,
- gl_shader *shader, bool disable_varying_packing,
+ gl_linked_shader *shader, bool disable_varying_packing,
bool xfb_enabled)
{
exec_list *instructions = shader->ir;
ir_function_signature *main_func_sig
= main_func->matching_signature(NULL, &void_parameters, false);
exec_list new_instructions, new_variables;
- lower_packed_varyings_visitor visitor(mem_ctx, locations_used, mode,
+ lower_packed_varyings_visitor visitor(mem_ctx,
+ locations_used,
+ components,
+ mode,
gs_input_vertices,
&new_instructions,
&new_variables,
lower_packed_varyings_gs_splicer splicer(mem_ctx, &new_instructions);
/* Add all the variables in first. */
- main_func_sig->body.head->insert_before(&new_variables);
+ main_func_sig->body.get_head_raw()->insert_before(&new_variables);
/* Now update all the EmitVertex instances */
splicer.run(instructions);
} else {
- /* For other shader types, outputs need to be lowered at the end of
- * main()
+ /* For other shader types, outputs need to be lowered before each
+ * return statement and at the end of main()
+ */
+
+ lower_packed_varyings_return_splicer splicer(mem_ctx, &new_instructions);
+
+ main_func_sig->body.get_head_raw()->insert_before(&new_variables);
+
+ splicer.run(instructions);
+
+ /* Lower outputs at the end of main() if the last instruction is not
+ * a return statement
*/
- main_func_sig->body.append_list(&new_variables);
- main_func_sig->body.append_list(&new_instructions);
+ if (((ir_instruction*)instructions->get_tail())->ir_type != ir_type_return) {
+ main_func_sig->body.append_list(&new_instructions);
+ }
}
} else {
/* Shader inputs need to be lowered at the beginning of main() */
- main_func_sig->body.head->insert_before(&new_instructions);
- main_func_sig->body.head->insert_before(&new_variables);
+ main_func_sig->body.get_head_raw()->insert_before(&new_instructions);
+ main_func_sig->body.get_head_raw()->insert_before(&new_variables);
}
}
projects / mesa.git / blobdiff