This repository has been archived on 2023-10-09. You can view files and clone it. You cannot open issues or pull requests or push a commit.
Files
blender-archive/source/blender/gpu/intern/gpu_node_graph.h
Jason Fielder 7b9d1cb51f Eevee: GPU Material node graph optimization.
Certain material node graphs can be very expensive to run. This feature aims to produce secondary GPUPass shaders within a GPUMaterial which provide optimal runtime performance. Such optimizations include baking constant data into the shader source directly, allowing the compiler to propogate constants and perform aggressive optimization upfront.

As optimizations can result in reduction of shader editor and animation interactivity, optimized pass generation and compilation is deferred until all outstanding compilations have completed. Optimization is also delayed util a material has remained unmodified for a set period of time, to reduce excessive compilation. The original variant of the material shader is kept to maintain interactivity.

Also adding a new concept to gpu::Shader allowing assignment of a parent shader from which a shader can pull PSO descriptors and any required metadata for asynchronous shader cache warming. This enables fully asynchronous shader optimization, without runtime hitching, while also reducing runtime hitching for standard materials, by using PSO descriptors from default materials, ahead of rendering.

Further shader graph optimizations are likely also possible with this architecture. Certain scenes, such as Wanderer benefit significantly. Viewport performance for this scene is 2-3x faster on Apple-silicon based GPUs.

Authored by Apple: Michael Parkin-White

Ref T96261
Pull Request #104536
2023-02-14 21:51:03 +01:00

234 lines
6.0 KiB
C++

/* SPDX-License-Identifier: GPL-2.0-or-later
* Copyright 2005 Blender Foundation. All rights reserved. */
/** \file
* \ingroup gpu
*
* Intermediate node graph for generating GLSL shaders.
*/
#pragma once
#include "DNA_customdata_types.h"
#include "DNA_listBase.h"
#include "BLI_ghash.h"
#include "GPU_material.h"
#include "GPU_shader.h"
#ifdef __cplusplus
extern "C" {
#endif
struct GPUNode;
struct GPUOutput;
struct ListBase;
typedef enum eGPUDataSource {
GPU_SOURCE_OUTPUT,
GPU_SOURCE_CONSTANT,
GPU_SOURCE_UNIFORM,
GPU_SOURCE_ATTR,
GPU_SOURCE_UNIFORM_ATTR,
GPU_SOURCE_LAYER_ATTR,
GPU_SOURCE_STRUCT,
GPU_SOURCE_TEX,
GPU_SOURCE_TEX_TILED_MAPPING,
GPU_SOURCE_FUNCTION_CALL,
GPU_SOURCE_CRYPTOMATTE,
} eGPUDataSource;
typedef enum {
GPU_NODE_LINK_NONE = 0,
GPU_NODE_LINK_ATTR,
GPU_NODE_LINK_UNIFORM_ATTR,
GPU_NODE_LINK_LAYER_ATTR,
GPU_NODE_LINK_COLORBAND,
GPU_NODE_LINK_CONSTANT,
GPU_NODE_LINK_IMAGE,
GPU_NODE_LINK_IMAGE_TILED,
GPU_NODE_LINK_IMAGE_TILED_MAPPING,
GPU_NODE_LINK_IMAGE_SKY,
GPU_NODE_LINK_OUTPUT,
GPU_NODE_LINK_UNIFORM,
GPU_NODE_LINK_DIFFERENTIATE_FLOAT_FN,
} GPUNodeLinkType;
typedef enum {
GPU_NODE_TAG_NONE = 0,
GPU_NODE_TAG_SURFACE = (1 << 0),
GPU_NODE_TAG_VOLUME = (1 << 1),
GPU_NODE_TAG_DISPLACEMENT = (1 << 2),
GPU_NODE_TAG_THICKNESS = (1 << 3),
GPU_NODE_TAG_AOV = (1 << 4),
GPU_NODE_TAG_FUNCTION = (1 << 5),
GPU_NODE_TAG_COMPOSITOR = (1 << 6),
} eGPUNodeTag;
ENUM_OPERATORS(eGPUNodeTag, GPU_NODE_TAG_COMPOSITOR)
struct GPUNode {
struct GPUNode *next, *prev;
const char *name;
/* Internal flag to mark nodes during pruning */
eGPUNodeTag tag;
ListBase inputs;
ListBase outputs;
};
struct GPUNodeLink {
GPUNodeStack *socket;
GPUNodeLinkType link_type;
int users; /* Refcount */
union {
/* GPU_NODE_LINK_CONSTANT | GPU_NODE_LINK_UNIFORM */
const float *data;
/* GPU_NODE_LINK_COLORBAND */
struct GPUTexture **colorband;
/* GPU_NODE_LINK_OUTPUT */
struct GPUOutput *output;
/* GPU_NODE_LINK_ATTR */
struct GPUMaterialAttribute *attr;
/* GPU_NODE_LINK_UNIFORM_ATTR */
struct GPUUniformAttr *uniform_attr;
/* GPU_NODE_LINK_LAYER_ATTR */
struct GPULayerAttr *layer_attr;
/* GPU_NODE_LINK_IMAGE_BLENDER */
struct GPUMaterialTexture *texture;
/* GPU_NODE_LINK_DIFFERENTIATE_FLOAT_FN */
const char *function_name;
};
};
typedef struct GPUOutput {
struct GPUOutput *next, *prev;
GPUNode *node;
eGPUType type; /* data type = length of vector/matrix */
GPUNodeLink *link; /* output link */
int id; /* unique id as created by code generator */
} GPUOutput;
typedef struct GPUInput {
struct GPUInput *next, *prev;
GPUNode *node;
eGPUType type; /* data-type. */
GPUNodeLink *link;
int id; /* unique id as created by code generator */
eGPUDataSource source; /* data source */
/* Content based on eGPUDataSource */
union {
/* GPU_SOURCE_CONSTANT | GPU_SOURCE_UNIFORM */
float vec[16]; /* vector data */
/* GPU_SOURCE_TEX | GPU_SOURCE_TEX_TILED_MAPPING */
struct GPUMaterialTexture *texture;
/* GPU_SOURCE_ATTR */
struct GPUMaterialAttribute *attr;
/* GPU_SOURCE_UNIFORM_ATTR */
struct GPUUniformAttr *uniform_attr;
/* GPU_SOURCE_LAYER_ATTR */
struct GPULayerAttr *layer_attr;
/* GPU_SOURCE_FUNCTION_CALL */
char function_call[64];
};
} GPUInput;
typedef struct GPUNodeGraphOutputLink {
struct GPUNodeGraphOutputLink *next, *prev;
int hash;
GPUNodeLink *outlink;
} GPUNodeGraphOutputLink;
typedef struct GPUNodeGraphFunctionLink {
struct GPUNodeGraphFunctionLink *next, *prev;
char name[16];
GPUNodeLink *outlink;
} GPUNodeGraphFunctionLink;
typedef struct GPUNodeGraph {
/* Nodes */
ListBase nodes;
/* Main Outputs. */
GPUNodeLink *outlink_surface;
GPUNodeLink *outlink_volume;
GPUNodeLink *outlink_displacement;
GPUNodeLink *outlink_thickness;
/* List of GPUNodeGraphOutputLink */
ListBase outlink_aovs;
/* List of GPUNodeGraphFunctionLink */
ListBase material_functions;
/* List of GPUNodeGraphOutputLink */
ListBase outlink_compositor;
/* Requested attributes and textures. */
ListBase attributes;
ListBase textures;
/* The list of uniform attributes. */
GPUUniformAttrList uniform_attrs;
/* The list of layer attributes. */
ListBase layer_attrs;
/** Set of all the GLSL lib code blocks . */
GSet *used_libraries;
} GPUNodeGraph;
/* Node Graph */
void gpu_nodes_tag(GPUNodeLink *link, eGPUNodeTag tag);
void gpu_node_graph_prune_unused(GPUNodeGraph *graph);
void gpu_node_graph_finalize_uniform_attrs(GPUNodeGraph *graph);
/**
* Optimize node graph for optimized material shader path.
* Once the base material has been generated, we can modify the shader
* node graph to create one which will produce an optimally performing shader.
* This currently involves baking uniform data into constant data to enable
* aggressive constant folding by the compiler in order to reduce complexity and
* shader core memory pressure.
*
* NOTE: Graph optimizations will produce a shader which needs to be re-compiled
* more frequently, however, the default material pass will always exist to fall
* back on. */
void gpu_node_graph_optimize(GPUNodeGraph *graph);
/**
* Free intermediate node graph.
*/
void gpu_node_graph_free_nodes(GPUNodeGraph *graph);
/**
* Free both node graph and requested attributes and textures.
*/
void gpu_node_graph_free(GPUNodeGraph *graph);
/* Material calls */
struct GPUNodeGraph *gpu_material_node_graph(struct GPUMaterial *material);
/**
* Returns the address of the future pointer to coba_tex.
*/
struct GPUTexture **gpu_material_ramp_texture_row_set(struct GPUMaterial *mat,
int size,
float *pixels,
float *row);
/**
* Returns the address of the future pointer to sky_tex
*/
struct GPUTexture **gpu_material_sky_texture_layer_set(
struct GPUMaterial *mat, int width, int height, const float *pixels, float *row);
#ifdef __cplusplus
}
#endif