src/render/gpu/SDL_pipeline_gpu.c

/*
  Simple DirectMedia Layer
  Copyright (C) 1997-2025 Sam Lantinga <slouken@libsdl.org>

  This software is provided 'as-is', without any express or implied
  warranty.  In no event will the authors be held liable for any damages
  arising from the use of this software.

  Permission is granted to anyone to use this software for any purpose,
  including commercial applications, and to alter it and redistribute it
  freely, subject to the following restrictions:

  1. The origin of this software must not be misrepresented; you must not
     claim that you wrote the original software. If you use this software
     in a product, an acknowledgment in the product documentation would be
     appreciated but is not required.
  2. Altered source versions must be plainly marked as such, and must not be
     misrepresented as being the original software.
  3. This notice may not be removed or altered from any source distribution.
*/
#include "SDL_internal.h"

#ifdef SDL_VIDEO_RENDER_GPU

#include "SDL_gpu_util.h"
#include "SDL_pipeline_gpu.h"

#include "../SDL_sysrender.h"

struct GPU_PipelineCacheKeyStruct
{
    Uint64 blend_mode : 28;
    Uint64 frag_shader : 4;
    Uint64 vert_shader : 4;
    Uint64 attachment_format : 6;
    Uint64 primitive_type : 3;
};

typedef union GPU_PipelineCacheKeyConverter
{
    struct GPU_PipelineCacheKeyStruct as_struct;
    Uint64 as_uint64;
} GPU_PipelineCacheKeyConverter;

SDL_COMPILE_TIME_ASSERT(GPU_PipelineCacheKeyConverter_Size, sizeof(GPU_PipelineCacheKeyConverter) <= sizeof(Uint64));

static Uint32 SDLCALL HashPipelineCacheKey(void *userdata, const void *key)
{
    const GPU_PipelineParameters *params = (const GPU_PipelineParameters *) key;
    GPU_PipelineCacheKeyConverter cvt;
    cvt.as_uint64 = 0;
    cvt.as_struct.blend_mode = params->blend_mode;
    cvt.as_struct.frag_shader = params->frag_shader;
    cvt.as_struct.vert_shader = params->vert_shader;
    cvt.as_struct.attachment_format = params->attachment_format;
    cvt.as_struct.primitive_type = params->primitive_type;

    // 64-bit uint hash function stolen from taisei (which stole it from somewhere else)
    Uint64 x = cvt.as_uint64;
    x = (x ^ (x >> 30)) * UINT64_C(0xbf58476d1ce4e5b9);
    x = (x ^ (x >> 27)) * UINT64_C(0x94d049bb133111eb);
    x = x ^ (x >> 31);
    return (Uint32)(x & 0xffffffff);
}

static bool SDLCALL MatchPipelineCacheKey(void *userdata, const void *a, const void *b)
{
    return (SDL_memcmp(a, b, sizeof (GPU_PipelineParameters)) == 0);
}

static void SDLCALL DestroyPipelineCacheHashItem(void *userdata, const void *key, const void *value)
{
    SDL_GPUGraphicsPipeline *pipeline = (SDL_GPUGraphicsPipeline *) value;
    SDL_GPUDevice *device = (SDL_GPUDevice *) userdata;
    SDL_ReleaseGPUGraphicsPipeline(device, pipeline);
    SDL_free((GPU_PipelineParameters *) key);
}

bool GPU_InitPipelineCache(GPU_PipelineCache *cache, SDL_GPUDevice *device)
{
    cache->table = SDL_CreateHashTable(0, false, HashPipelineCacheKey, MatchPipelineCacheKey, DestroyPipelineCacheHashItem, device);
    return (cache->table != NULL);
}

void GPU_DestroyPipelineCache(GPU_PipelineCache *cache)
{
    SDL_DestroyHashTable(cache->table);
}

static SDL_GPUGraphicsPipeline *MakePipeline(SDL_GPUDevice *device, GPU_Shaders *shaders, const GPU_PipelineParameters *params)
{
    SDL_GPUColorTargetDescription ad;
    SDL_zero(ad);
    ad.format = params->attachment_format;

    SDL_BlendMode blend = params->blend_mode;
    ad.blend_state.enable_blend = blend != 0;
    ad.blend_state.color_write_mask = 0xF;
    ad.blend_state.alpha_blend_op = GPU_ConvertBlendOperation(SDL_GetBlendModeAlphaOperation(blend));
    ad.blend_state.dst_alpha_blendfactor = GPU_ConvertBlendFactor(SDL_GetBlendModeDstAlphaFactor(blend));
    ad.blend_state.src_alpha_blendfactor = GPU_ConvertBlendFactor(SDL_GetBlendModeSrcAlphaFactor(blend));
    ad.blend_state.color_blend_op = GPU_ConvertBlendOperation(SDL_GetBlendModeColorOperation(blend));
    ad.blend_state.dst_color_blendfactor = GPU_ConvertBlendFactor(SDL_GetBlendModeDstColorFactor(blend));
    ad.blend_state.src_color_blendfactor = GPU_ConvertBlendFactor(SDL_GetBlendModeSrcColorFactor(blend));

    SDL_GPUGraphicsPipelineCreateInfo pci;
    SDL_zero(pci);
    pci.target_info.has_depth_stencil_target = false;
    pci.target_info.num_color_targets = 1;
    pci.target_info.color_target_descriptions = &ad;
    pci.vertex_shader = GPU_GetVertexShader(shaders, params->vert_shader);
    pci.fragment_shader = GPU_GetFragmentShader(shaders, params->frag_shader);
    pci.multisample_state.sample_count = SDL_GPU_SAMPLECOUNT_1;
    pci.multisample_state.enable_mask = false;
    pci.primitive_type = params->primitive_type;

    pci.rasterizer_state.cull_mode = SDL_GPU_CULLMODE_NONE;
    pci.rasterizer_state.fill_mode = SDL_GPU_FILLMODE_FILL;
    pci.rasterizer_state.front_face = SDL_GPU_FRONTFACE_COUNTER_CLOCKWISE;

    SDL_GPUVertexBufferDescription vertex_buffer_desc;
    SDL_zero(vertex_buffer_desc);

    Uint32 num_attribs = 0;
    SDL_GPUVertexAttribute attribs[4];
    SDL_zero(attribs);

    bool have_attr_color = false;
    bool have_attr_uv = false;

    switch (params->vert_shader) {
    case VERT_SHADER_TRI_TEXTURE:
        have_attr_uv = true;
        SDL_FALLTHROUGH;
    case VERT_SHADER_TRI_COLOR:
        have_attr_color = true;
        SDL_FALLTHROUGH;
    default:
        break;
    }

    // Position
    attribs[num_attribs].location = num_attribs;
    attribs[num_attribs].format = SDL_GPU_VERTEXELEMENTFORMAT_FLOAT2;
    attribs[num_attribs].offset = vertex_buffer_desc.pitch;
    vertex_buffer_desc.pitch += 2 * sizeof(float);
    num_attribs++;

    if (have_attr_color) {
        // Color
        attribs[num_attribs].location = num_attribs;
        attribs[num_attribs].format = SDL_GPU_VERTEXELEMENTFORMAT_FLOAT4;
        attribs[num_attribs].offset = vertex_buffer_desc.pitch;
        vertex_buffer_desc.pitch += 4 * sizeof(float);
        num_attribs++;
    }

    if (have_attr_uv) {
        // UVs
        attribs[num_attribs].location = num_attribs;
        attribs[num_attribs].format = SDL_GPU_VERTEXELEMENTFORMAT_FLOAT2;
        attribs[num_attribs].offset = vertex_buffer_desc.pitch;
        vertex_buffer_desc.pitch += 2 * sizeof(float);
        num_attribs++;
    }

    pci.vertex_input_state.num_vertex_attributes = num_attribs;
    pci.vertex_input_state.vertex_attributes = attribs;
    pci.vertex_input_state.num_vertex_buffers = 1;
    pci.vertex_input_state.vertex_buffer_descriptions = &vertex_buffer_desc;

    return SDL_CreateGPUGraphicsPipeline(device, &pci);
}

SDL_GPUGraphicsPipeline *GPU_GetPipeline(GPU_PipelineCache *cache, GPU_Shaders *shaders, SDL_GPUDevice *device, const GPU_PipelineParameters *params)
{
    SDL_GPUGraphicsPipeline *pipeline = NULL;
    if (!SDL_FindInHashTable(cache->table, params, (const void **) &pipeline)) {
        bool inserted = false;
        // !!! FIXME: why don't we have an SDL_alloc_copy function/macro?
        GPU_PipelineParameters *paramscpy = (GPU_PipelineParameters *) SDL_malloc(sizeof (*paramscpy));
        if (paramscpy) {
            SDL_copyp(paramscpy, params);
            pipeline = MakePipeline(device, shaders, params);
            if (pipeline) {
                inserted = SDL_InsertIntoHashTable(cache->table, paramscpy, pipeline, false);
            }
        }

        if (!inserted) {
            SDL_free(paramscpy);
            if (pipeline) {
                SDL_ReleaseGPUGraphicsPipeline(device, pipeline);
                pipeline = NULL;
            }
        }
    }

    return pipeline;
}

#endif // SDL_VIDEO_RENDER_GPU