38using namespace blender::gpu::shader;
49 return stream << (input->
is_zone_io ?
"zone" :
"tmp") << input->
id;
51 return stream << (input->
is_zone_io ?
"zone" :
"cons") << input->
id;
53 return stream <<
"node_tree.u" << input->
id << (input->
is_duplicate ?
"b" :
"");
55 return stream <<
"var_attrs.v" << input->attr->
id;
57 return stream <<
"UNI_ATTR(unf_attrs[resource_id].attr" << input->uniform_attr->
id <<
")";
59 return stream <<
"attr_load_layer(kg, " << input->layer_attr->hash_code <<
")";
61 return stream << (input->
is_zone_io ?
"zone" :
"strct") << input->
id;
63 return stream <<
"sd";
65 return stream <<
"kg";
67 return stream << input->texture->sampler_name;
69 return stream << input->texture->tiled_mapping_name;
78 return stream << (
output->is_zone_io ?
"zone" :
"tmp") <<
output->id;
89 char formatted_float[32];
91 stream << formatted_float;
104 for (
const int &
element : span) {
119 switch (input->
type) {
146using blender::operator<<;
169 create_info = MEM_new<GPUCodegenCreateInfo>(__func__, debug_name);
176 MEM_SAFE_DELETE(cryptomatte_input_);
177 MEM_SAFE_DELETE(thickness_input_);
179 ubo_inputs_.free_no_destruct();
186 bool do_optimize = (nodes_total_ >= 60 || textures_total_ >= 4 || uniforms_total_ >= 64) &&
187 (textures_total_ >= 1 && uniforms_total_ >= 8 && nodes_total_ >= 4);
200 info.
interface_generated = MEM_new<StageInterfaceInfo>(__func__,
"codegen_iface",
"var_attrs");
205 std::stringstream load_ss;
221 GPUType input_type, iface_type;
223 load_ss <<
"var_attrs." << var_name;
226 load_ss <<
" = attr_load_" << input_type <<
"(domain, " << attr_name <<
", " << attr_n
235 load_ss <<
" = attr_load_orco(domain, " << attr_name <<
", " << attr_n <<
");\n";
239 load_ss <<
" = attr_load_tangent(domain, " << attr_name <<
", " << attr_n <<
");\n";
243 load_ss <<
" = attr_load_" << input_type <<
"(domain, " << attr_name <<
", " << attr_n
261 std::stringstream ss;
263 ss <<
"#line 1 \"" __FILE__
"\"\n";
291 textures_total_ = slot;
293 if (!ubo_inputs_.is_empty()) {
294 const char *linted_struct_suffix =
"_host_shared_";
296 ss <<
"struct NodeTree {\n";
297 for (
LinkData &link : ubo_inputs_) {
300 ss <<
"bool32_t thickness_mode;\n";
303 ss << input->
type <<
" crypto_hash;\n";
313 ss <<
" u" << input->
id << (input->
is_duplicate ?
"b" :
"") <<
";\n";
317 ss <<
"#define NodeTree" << linted_struct_suffix <<
" NodeTree\n";
318 ss <<
"#define NodeTree" << linted_struct_suffix <<
"uniform_ NodeTree\n";
325 ss <<
"struct UniformAttrs {\n";
327 ss <<
"vec4 attr" << attr.
id <<
";\n";
340 std::stringstream &eval_ss,
345 auto source_reference = [&](
GPUInput *input) {
350 input->link->
output->type;
353 eval_ss << to <<
"_from_" << from <<
"(";
360 eval_ss << input->link->output;
366 float coefficients[3];
375 for (GPUInput &input :
node->inputs) {
378 std::stringstream ss;
386 eval_ss << type() <<
" " << &input <<
"; " << input.function_call << &input <<
");\n";
393 eval_ss << input.
type <<
" " << &input <<
" = CLOSURE_DEFAULT;\n";
397 eval_ss << type() <<
" " << &input <<
" = " <<
static_cast<GPUConstant *
>(&input)
404 eval_ss << type() <<
" " << &input <<
" = ";
405 source_reference(&input);
411 eval_ss << type() <<
" zone" << input.
id <<
" = " << &input <<
";\n";
433 eval_ss <<
node->name <<
"(";
435 for (GPUInput &input :
node->inputs) {
442 source_reference(&input);
487 std::ranges::sort(source_files);
493 const char *output_default)
495 if (output_link ==
nullptr && output_default ==
nullptr) {
499 Set<StringRefNull> used_libraries;
500 std::stringstream eval_ss;
501 bool has_nodes =
false;
505 if ((
node.tag & tree_tag) == 0) {
508 node_serialize(used_libraries, eval_ss, &
node);
517 eval_ss <<
"return " << output_link->output <<
";\n";
521 eval_ss <<
"return " << output_default <<
";\n";
524 std::string
str = eval_ss.str();
529GPUGraphOutput GPUCodegen::graph_serialize(
GPUNodeTag tree_tag)
531 std::stringstream eval_ss;
532 Set<StringRefNull> used_libraries;
534 if (
node.tag & tree_tag) {
535 node_serialize(used_libraries, eval_ss, &
node);
538 std::string
str = eval_ss.str();
546 cryptomatte_input_ = MEM_new<GPUInput>(__func__);
550 float material_hash = 0.0f;
555 material_hash =
hash.float_encoded();
557 cryptomatte_input_->constant_data = material_hash;
562 thickness_input_ = MEM_new<GPUInput>(__func__);
569 thickness_mode = material->thickness_mode;
571 thickness_input_->constant_data = bool(thickness_mode);
593 if (!ubo_inputs_.is_empty()) {
601 if (!ubo_inputs_.is_empty()) {
608void GPUCodegen::set_unique_ids()
621 if (
node.zone_index != -1) {
622 auto &map =
node.is_zone_end ? zone_ends : zone_starts;
627 auto find_zone_io = [](
auto first) {
628 while (first && !first->is_zone_io && first->next) {
635 for (GPUNode *end : zone_ends.
values()) {
637 GPUInput *end_input = find_zone_io(end->inputs.first());
638 GPUOutput *end_output = find_zone_io(end->outputs.first());
640 GPUNode *start = zone_starts.
lookup(end->zone_index);
642 GPUInput *start_input = find_zone_io(start->
inputs.first());
643 GPUOutput *start_output = find_zone_io(start->
outputs.first());
645 for (; start_input; start_input = start_input->
next,
646 start_output = start_output->
next,
647 end_input = end_input->
next,
648 end_output = end_output->
next)
650 start_output->
id = start_input->
id;
651 end_input->
id = start_input->
id;
652 end_output->
id = start_input->
id;
661 output.surface = graph_serialize(
664 output.displacement = graph_serialize(
673 std::stringstream eval_ss;
681 eval_ss <<
"float " << func_link.
name <<
"(KernelGlobals kg, ShadingData sd) {\n"
#define SNPRINTF(dst, format,...)
Read Blender memory allocation and freeing.
#define BLI_assert_unreachable()
#define BLI_assert_msg(a, msg)
ValueIterator values() const &
bool add(const Key &key, const Value &value)
const Value & lookup(const Key &key) const
Span< NewT > constexpr cast() const
constexpr int64_t size() const
constexpr const T & last(const int64_t n=0) const
void append(const T &value)
GPUCodegenCreateInfo * create_info
void generate_uniform_buffer()
bool should_optimize_heuristic() const
GPUCodegen(GPUMaterial *mat_, GPUNodeGraph *graph_, const char *debug_name)
void generate_material_props()
void extract_uniform_list()
void generate_resources()
void sort_uniform_inputs()
std::ostream & operator<<(std::ostream &stream, const AssetCatalogPath &path_to_append)
BLI_assert(slot->slot_subtype.map==BMO_OP_SLOT_SUBTYPE_MAP_FLT)
uint32_t BLI_hash_mm2a_end(BLI_HashMurmur2A *mm2)
char * STRNCPY(char(&dst)[N], const char *src)
constexpr GPUType gpu_int_type_from_element_count(const int count)
void BLI_hash_mm2a_add(BLI_HashMurmur2A *mm2, const unsigned char *data, size_t len)
Span< float > gpu_constant_to_float_span(const GPUInputConstantData &data, const GPUType type)
Span< int > gpu_constant_to_int_span(const GPUInputConstantData &data, const GPUType type)
constexpr GPUType gpu_float_type_from_element_count(const int count)
BLI_INLINE void IMB_colormanagement_get_luminance_coefficients(float r_rgb[3])
void gpu_material_library_use_function(Set< StringRefNull > &used_libraries, const char *name)
void gpu_nodes_tag(GPUNodeGraph *graph, GPUNodeLink *link_start, GPUNodeTag tag)
void GPU_material_uniform_buffer_create(GPUMaterial *material, ListBaseT< LinkData > *inputs)
eGPUMaterialFlag GPU_material_flag(const GPUMaterial *mat)
bool gpu_constant_to_bool(const GPUInputConstantData &data)
@ GPU_NODE_TAG_COMPOSITOR
@ GPU_NODE_TAG_DISPLACEMENT
int char char int int int int size_t BLI_strnlen(const char *str, size_t maxlen) ATTR_WARN_UNUSED_RESULT ATTR_NONNULL(1)
LinkData * BLI_genericNodeN(void *data)
void BLI_hash_mm2a_add_int(BLI_HashMurmur2A *mm2, int data)
void BLI_hash_mm2a_init(BLI_HashMurmur2A *mm2, uint32_t seed)
@ GPU_SOURCE_TEX_TILED_MAPPING
@ GPU_SOURCE_SHADING_DATA
@ GPU_SOURCE_FUNCTION_CALL
@ GPU_SOURCE_THICKNESS_MODE
@ GPU_SOURCE_UNIFORM_ATTR
@ GPU_SOURCE_KERNEL_GLOBALS
void BLI_addtail(ListBase *listbase, void *vlink) ATTR_NONNULL(1)
uint64_t GPU_material_uuid_get(GPUMaterial *mat)
Material * GPU_material_get_material(GPUMaterial *material)
void GPU_uniformbuf_inputs_sort(ListBaseT< LinkData > *inputs)
ATTR_WARN_UNUSED_RESULT const void * element
static Vector< StringRefNull > set_to_vector_stable(Set< StringRefNull > &set)
GPUShaderCreateInfo * create_info
Vector< StringRefNull > dependencies
char tiled_mapping_name[32]
gpu::Texture ** colorband
ListBaseT< GPUMaterialAttribute > attributes
ListBaseT< GPUNodeGraphFunctionLink > material_functions
GPUNodeLink * outlink_surface
GPUNodeLink * outlink_displacement
GPUNodeLink * outlink_thickness
ListBaseT< GPUNode > nodes
ListBaseT< GPUNodeGraphOutputLink > outlink_compositor
GPUNodeLink * outlink_volume
ListBaseT< GPUMaterialTexture > textures
GPUUniformAttrList uniform_attrs
ListBaseT< GPUOutput > outputs
ListBaseT< GPUInput > inputs
char attr_names[16][GPU_MAX_SAFE_ATTR_NAME+1]
Vector< std::unique_ptr< NameEntry >, 16 > sampler_names
const char * append_sampler_name(const char name[32])
StageInterfaceInfo * interface_generated
Describe inputs & outputs, stage interfaces, resources and sources of a shader. If all data is correc...
Self & uniform_buf(int slot, StringRefNull type_name, StringRefNull name, Frequency freq=Frequency::PASS, ConditionFn cond=nullptr)
Self & vertex_in(int slot, Type type, StringRefNull name, ConditionFn cond=nullptr)
std::string typedef_source_generated
Self & vertex_out(StageInterfaceInfo &interface, ConditionFn cond=nullptr)
Self & sampler(int slot, ImageType type, StringRefNull name, Frequency freq=Frequency::PASS, ConditionFn cond=nullptr)
Self & smooth(Type type, StringRefNull _name)