ShaderUtils.hpp

Go to the documentation of this file.

#pragma once
 
#include "vulkan/vulkan.hpp"
 
#include "GraphicsUtils.hpp"
 
#include "MayaFlux/Kakshya/NDData/VertexLayout.hpp"
 
namespace MayaFlux::Core {
struct VertexBinding;
struct VertexAttribute;
}
 
namespace MayaFlux::Portal::Graphics {
 
/**
 * @brief Resolve the extra vk::BufferUsageFlags a BufferUsageHint requires,
 *        on top of whatever base usage the caller already applies.
 */
[[nodiscard]] inline vk::BufferUsageFlags to_buffer_usage_flags(BufferUsageHint hint)
{
    switch (hint) {
    case BufferUsageHint::INDIRECT:
        return vk::BufferUsageFlagBits::eIndirectBuffer;
    case BufferUsageHint::COMPUTE_STORAGE:
        return vk::BufferUsageFlagBits::eStorageBuffer;
    default:
        break;
    }
 
    vk::BufferUsageFlags flags = vk::BufferUsageFlagBits::eTransferSrc | vk::BufferUsageFlagBits::eTransferDst;
 
    switch (hint) {
    case BufferUsageHint::STAGING:
        break;
    case BufferUsageHint::DEVICE:
    case BufferUsageHint::COMPUTE:
        flags |= vk::BufferUsageFlagBits::eStorageBuffer;
        break;
    case BufferUsageHint::VERTEX:
        flags |= vk::BufferUsageFlagBits::eVertexBuffer | vk::BufferUsageFlagBits::eStorageBuffer;
        break;
    case BufferUsageHint::INDEX:
        flags |= vk::BufferUsageFlagBits::eIndexBuffer;
        break;
    case BufferUsageHint::UNIFORM:
        flags |= vk::BufferUsageFlagBits::eUniformBuffer;
        break;
    case BufferUsageHint::UNIFORM_BDA:
        flags |= vk::BufferUsageFlagBits::eUniformBuffer
            | vk::BufferUsageFlagBits::eShaderDeviceAddress;
        break;
    case BufferUsageHint::STORAGE_BDA:
        flags |= vk::BufferUsageFlagBits::eStorageBuffer
            | vk::BufferUsageFlagBits::eShaderDeviceAddress;
        break;
    case BufferUsageHint::HOST_STORAGE:
        flags |= vk::BufferUsageFlagBits::eStorageBuffer;
        break;
    default:
        break;
    }
 
    return flags;
}
 
using ShaderID = uint64_t;
using DescriptorSetID = uint64_t;
using CommandBufferID = uint64_t;
using FenceID = uint64_t;
using SemaphoreID = uint64_t;
using RenderPipelineID = uint64_t;
using FramebufferID = uint64_t;
 
constexpr ShaderID INVALID_SHADER = 0;
constexpr DescriptorSetID INVALID_DESCRIPTOR_SET = 0;
constexpr CommandBufferID INVALID_COMMAND_BUFFER = 0;
constexpr FenceID INVALID_FENCE = 0;
constexpr SemaphoreID INVALID_SEMAPHORE = 0;
constexpr RenderPipelineID INVALID_RENDER_PIPELINE = 0;
constexpr FramebufferID INVALID_FRAMEBUFFER = 0;
 
/**
 * @struct ShaderCompilerConfig
 * @brief Configuration for shader compilation
 */
struct ShaderCompilerConfig {
    bool enable_optimization = true;
    bool enable_debug_info = false; ///< Include debug symbols (line numbers, variable names)
    bool enable_reflection = true; ///< Extract descriptor bindings and metadata
    bool enable_validation = true; ///< Validate SPIR-V after compilation
    std::vector<std::string> include_directories; ///< Paths for #include resolution
    std::unordered_map<std::string, std::string> defines; ///< Preprocessor macros
 
    ShaderCompilerConfig() = default;
};
 
/**
 * @struct GpuComputeConfig
 * @brief Plain-data description of the compute shader to dispatch.
 *
 * Either shader_path or shader_id must be set. If shader_id is not
 * INVALID_SHADER it takes precedence and shader_path is ignored,
 * allowing callers who already hold a compiled ShaderID (e.g. from
 * ShaderFoundry::load_shader(const ShaderSpec&)) to bypass file
 * resolution entirely.
 */
struct GpuComputeConfig {
    std::string shader_path;
    std::array<uint32_t, 3> workgroup_size { 256, 1, 1 };
    size_t push_constant_size { 0 };
    ShaderID shader_id { INVALID_SHADER };
};
 
/**
 * @struct ShaderSource
 * @brief Shader source descriptor for compilation
 */
struct ShaderSource {
    std::string content; ///< Shader source code, SPIR-V path, or SPIR-V assembly text
    ShaderStage stage;
    std::string entry_point = "main";
 
    enum class SourceType : uint8_t {
        GLSL_STRING, ///< In-memory GLSL source
        GLSL_FILE, ///< Path to .comp/.vert/.frag/etc
        SPIRV_FILE, ///< Path to .spv file
        SPIRV_ASM, ///< In-memory SPIR-V assembly text (assembled via SPIRV-Tools, no shaderc)
    } type = SourceType::GLSL_FILE;
 
    ShaderSource() = default;
    ShaderSource(std::string content_, ShaderStage stage_, SourceType type_)
        : content(std::move(content_))
        , stage(stage_)
        , type(type_)
    {
    }
};
 
/**
 * @struct PushConstantBindingInfo
 * @brief Self-describing push constant fragment staged on a VKBuffer.
 *
 * The push constant analogue of DescriptorBindingInfo. Where a descriptor
 * entry carries its own set and binding, a push constant entry carries its
 * own byte offset, so a shared list on the buffer merges into any
 * processor's own push constant block without a per-processor key.
 */
struct PushConstantBindingInfo {
    uint32_t offset {};
    std::vector<uint8_t> data;
    std::string name;
};
 
/**
 * @enum DescriptorRole
 * @brief Semantic descriptor type — maps to Vulkan descriptor types internally.
 *
 * Users declare what role a binding plays rather than which Vulkan type to use.
 * The mapping is:
 *   UNIFORM      → vk::DescriptorType::eUniformBuffer        (small, frequently updated values)
 *   STORAGE      → vk::DescriptorType::eStorageBuffer        (large or write-capable arrays)
 *   TEXTURE      → vk::DescriptorType::eCombinedImageSampler (sampled image + sampler)
 *   STORAGE_IMAGE → vk::DescriptorType::eStorageImage        (compute read/write image)
 */
enum class DescriptorRole : uint8_t {
    UNIFORM, ///< Small, read-only, frequently updated (scalars, param structs)
    STORAGE, ///< Large arrays or buffers the shader may write (SSBO)
    TEXTURE, ///< Sampled image + sampler combined
    STORAGE_IMAGE ///< Compute-writable image
};
 
/**
 * @brief Convert DescriptorRole to the corresponding vk::DescriptorType.
 */
[[nodiscard]] inline vk::DescriptorType to_vk_descriptor_type(DescriptorRole role)
{
    switch (role) {
    case DescriptorRole::UNIFORM:
        return vk::DescriptorType::eUniformBuffer;
    case DescriptorRole::STORAGE:
        return vk::DescriptorType::eStorageBuffer;
    case DescriptorRole::TEXTURE:
        return vk::DescriptorType::eCombinedImageSampler;
    case DescriptorRole::STORAGE_IMAGE:
        return vk::DescriptorType::eStorageImage;
    }
    return vk::DescriptorType::eStorageBuffer;
}
 
/**
 * @struct DescriptorBindingConfig
 * @brief Portal-level descriptor binding configuration
 */
struct DescriptorBindingInfo {
    uint32_t set { 1 };
    uint32_t binding {};
    vk::DescriptorType type = vk::DescriptorType::eStorageBuffer;
    vk::DescriptorBufferInfo buffer_info;
    std::string name;
    uint32_t count { 1 };
};
 
/**
 * @struct PushConstantRangeInfo
 * @brief Extracted push constant range from shader reflection
 */
struct PushConstantRangeInfo {
    uint32_t offset;
    uint32_t size;
};
 
/**
 * @struct ShaderReflectionInfo
 * @brief Extracted reflection information from compiled shader
 */
struct ShaderReflectionInfo {
    ShaderStage stage;
    std::string entry_point;
    std::optional<std::array<uint32_t, 3>> workgroup_size;
    std::vector<DescriptorBindingInfo> descriptor_bindings;
    std::vector<PushConstantRangeInfo> push_constant_ranges;
};
 
/**
 * @struct RasterizationConfig
 * @brief Rasterization state configuration
 */
struct RasterizationConfig {
    PolygonMode polygon_mode = PolygonMode::FILL;
    CullMode cull_mode = CullMode::BACK;
    bool front_face_ccw = true;
    float line_width = 1.0F;
    bool depth_clamp = false;
    bool depth_bias = false;
 
    RasterizationConfig() = default;
};
 
/**
 * @struct DepthStencilConfig
 * @brief Depth and stencil test configuration
 */
struct DepthStencilConfig {
    bool depth_test_enable = true;
    bool depth_write_enable = true;
    CompareOp depth_compare_op = CompareOp::NEVER;
    bool stencil_test_enable = false;
 
    DepthStencilConfig() = default;
};
 
/**
 * @struct BlendAttachmentConfig
 * @brief Per-attachment blend configuration
 */
struct BlendAttachmentConfig {
    bool blend_enable = false;
    BlendFactor src_color_factor = BlendFactor::ONE;
    BlendFactor dst_color_factor = BlendFactor::ZERO;
    BlendOp color_blend_op = BlendOp::ADD;
    BlendFactor src_alpha_factor = BlendFactor::ONE;
    BlendFactor dst_alpha_factor = BlendFactor::ZERO;
    BlendOp alpha_blend_op = BlendOp::ADD;
 
    BlendAttachmentConfig() = default;
 
    /// @brief Create standard alpha blending configuration
    static BlendAttachmentConfig alpha_blend()
    {
        BlendAttachmentConfig config;
        config.blend_enable = true;
        config.src_color_factor = BlendFactor::SRC_ALPHA;
        config.dst_color_factor = BlendFactor::ONE_MINUS_SRC_ALPHA;
        config.src_alpha_factor = BlendFactor::ONE;
        config.dst_alpha_factor = BlendFactor::ZERO;
        return config;
    }
};
 
/**
 * @struct RenderPipelineConfig
 * @brief Complete render pipeline configuration
 */
struct RenderPipelineConfig {
    // Shader stages
    ShaderID vertex_shader = INVALID_SHADER;
    ShaderID fragment_shader = INVALID_SHADER;
    ShaderID geometry_shader = INVALID_SHADER; ///< Optional
    ShaderID tess_control_shader = INVALID_SHADER; ///< Optional
    ShaderID tess_eval_shader = INVALID_SHADER; ///< Optional
    ShaderID mesh_shader = INVALID_SHADER; ///< Optional
    ShaderID task_shader = INVALID_SHADER; ///< Optional
 
    // Vertex input
    std::vector<Core::VertexBinding> vertex_bindings;
    std::vector<Core::VertexAttribute> vertex_attributes;
 
    // Input assembly
    PrimitiveTopology topology = PrimitiveTopology::TRIANGLE_LIST;
 
    // Optional semantic vertex layout
    std::optional<Kakshya::VertexLayout> semantic_vertex_layout;
 
    // Use reflection to auto-configure from vertex shader
    bool use_vertex_shader_reflection = true;
 
    // Rasterization
    RasterizationConfig rasterization;
 
    // Depth/stencil
    DepthStencilConfig depth_stencil;
 
    // Blend
    std::vector<BlendAttachmentConfig> blend_attachments;
 
    // Descriptor sets (similar to compute)
    std::vector<std::vector<DescriptorBindingInfo>> descriptor_sets;
 
    // Push constants
    size_t push_constant_size {};
 
    RenderPipelineConfig() = default;
};
 
} // namespace MayaFlux::Portal::Graphics