32c301e3cf
This adds a function that can turn an existing `bNodeTree` into an inlined one. The new node tree has all node groups, repeat zones, closures and bundles inlined. So it's just a flat tree that ideally can be consumed easily by render engines. As part of the process, it also does constant folding. The goal is to support more advanced features from geometry nodes (repeat zones, etc.) in shader nodes which the evaluator is more limited because it has to be able to run on the GPU. Creating an inlined `bNodeTree` is likely the most direct way to get but may also be limiting in the future. Since this is a fairly local change, it's likely still worth it to support these features in all render engines without having to make their evaluators significantly more complex. Some limitations apply here that do not apply in Geometry Nodes. For example, the iterations count in a repeat zone has to be a constant after constant folding. There is also a `Test Inlining Shader Nodes` operator that creates the inlined tree and creates a group node for it. This is just for testing purposes. #145811 will make this functionality available to the Python API as well so that external renderers can use it too.
98 lines
4.6 KiB
C++
98 lines
4.6 KiB
C++
/* SPDX-FileCopyrightText: 2025 Blender Authors
|
|
*
|
|
* SPDX-License-Identifier: GPL-2.0-or-later */
|
|
|
|
#pragma once
|
|
|
|
#include "BKE_compute_context_cache_fwd.hh"
|
|
#include "BKE_compute_contexts.hh"
|
|
|
|
#include "BLI_linear_allocator.hh"
|
|
#include "BLI_map.hh"
|
|
#include "BLI_vector.hh"
|
|
|
|
namespace blender::bke {
|
|
|
|
/**
|
|
* When traversing the computation of a node tree (like in `socket_usage_inference.cc` or
|
|
* `partial_eval.cc`) one often enters and exists the same compute contexts. The cache implemented
|
|
* here avoids re-creating the same compute contexts over and over again. While requiring less
|
|
* memory and having potentially better performance, it can also be used to ensure that the same
|
|
* compute context will always have the same pointer, even if it's created in two different places.
|
|
*
|
|
* Constructing compute contexts through this cache can also be a bit more convenient.
|
|
*/
|
|
class ComputeContextCache {
|
|
/** Allocator used to allocate the compute contexts. */
|
|
LinearAllocator<> allocator_;
|
|
/** The allocated computed contexts that need to be destructed in the end. */
|
|
Vector<destruct_ptr<ComputeContext>> cache_;
|
|
|
|
Map<std::pair<const ComputeContext *, int>, const ModifierComputeContext *>
|
|
modifier_contexts_cache_;
|
|
Map<const ComputeContext *, const OperatorComputeContext *> operator_contexts_cache_;
|
|
Map<const ComputeContext *, const ShaderComputeContext *> shader_contexts_cache_;
|
|
Map<std::pair<const ComputeContext *, int32_t>, const GroupNodeComputeContext *>
|
|
group_node_contexts_cache_;
|
|
Map<std::pair<const ComputeContext *, int32_t>, const SimulationZoneComputeContext *>
|
|
simulation_zone_contexts_cache_;
|
|
Map<std::pair<const ComputeContext *, std::pair<int32_t, int>>, const RepeatZoneComputeContext *>
|
|
repeat_zone_contexts_cache_;
|
|
Map<std::pair<const ComputeContext *, std::pair<int32_t, int>>,
|
|
const ForeachGeometryElementZoneComputeContext *>
|
|
foreach_geometry_element_zone_contexts_cache_;
|
|
Map<std::pair<const ComputeContext *, int32_t>, const EvaluateClosureComputeContext *>
|
|
evaluate_closure_contexts_cache_;
|
|
|
|
public:
|
|
const ModifierComputeContext &for_modifier(const ComputeContext *parent,
|
|
const NodesModifierData &nmd);
|
|
const ModifierComputeContext &for_modifier(const ComputeContext *parent, int modifier_uid);
|
|
|
|
const OperatorComputeContext &for_operator(const ComputeContext *parent);
|
|
const OperatorComputeContext &for_operator(const ComputeContext *parent, const bNodeTree &tree);
|
|
|
|
const ShaderComputeContext &for_shader(const ComputeContext *parent, const bNodeTree *tree);
|
|
|
|
const GroupNodeComputeContext &for_group_node(const ComputeContext *parent,
|
|
int32_t node_id,
|
|
const bNodeTree *tree = nullptr);
|
|
|
|
const SimulationZoneComputeContext &for_simulation_zone(const ComputeContext *parent,
|
|
int output_node_id);
|
|
const SimulationZoneComputeContext &for_simulation_zone(const ComputeContext *parent,
|
|
const bNode &output_node);
|
|
|
|
const RepeatZoneComputeContext &for_repeat_zone(const ComputeContext *parent,
|
|
int32_t output_node_id,
|
|
int iteration);
|
|
const RepeatZoneComputeContext &for_repeat_zone(const ComputeContext *parent,
|
|
const bNode &output_node,
|
|
int iteration);
|
|
|
|
const ForeachGeometryElementZoneComputeContext &for_foreach_geometry_element_zone(
|
|
const ComputeContext *parent, int32_t output_node_id, int index);
|
|
const ForeachGeometryElementZoneComputeContext &for_foreach_geometry_element_zone(
|
|
const ComputeContext *parent, const bNode &output_node, int index);
|
|
|
|
const EvaluateClosureComputeContext &for_evaluate_closure(
|
|
const ComputeContext *parent,
|
|
int32_t node_id,
|
|
const bNodeTree *tree = nullptr,
|
|
const std::optional<nodes::ClosureSourceLocation> &closure_source_location = std::nullopt);
|
|
|
|
/**
|
|
* A fallback that does not use caching and can be used for any compute context.
|
|
* More constructors like the ones above can be added as they become necessary.
|
|
*/
|
|
template<typename T, typename... Args> const T &for_any_uncached(Args &&...args)
|
|
{
|
|
destruct_ptr<T> compute_context = allocator_.construct<T>(std::forward<Args>(args)...);
|
|
const T &result = *compute_context;
|
|
cache_.append(std::move(compute_context));
|
|
return result;
|
|
}
|
|
};
|
|
|
|
} // namespace blender::bke
|