Ammerhai/Mikemon
2
0
Fork 0
Code Issues 8 Pull Requests Packages Projects 2 Releases 1 Wiki Activity

Update SDL3 from 3.2.4 to 3.2.20

Browse Source
This commit is contained in:
Sven Balzer
2025-08-27 21:24:05 +02:00
parent 6283160467
commit ad651462df
332 changed files with 20334 additions and 4852 deletions
Download Patch File Download Diff File Expand all files Collapse all files
libs/SDL3/src/gpu/SDL_gpu.c
+446 -38
View File
@@ -56,15 +56,47 @@
}
#define CHECK_RENDERPASS \
if (!((Pass *)render_pass)->in_progress) { \
if (!((RenderPass *)render_pass)->in_progress) { \
SDL_assert_release(!"Render pass not in progress!"); \
return; \
}
#define CHECK_GRAPHICS_PIPELINE_BOUND \
if (!((CommandBufferCommonHeader *)RENDERPASS_COMMAND_BUFFER)->graphics_pipeline_bound) { \
SDL_assert_release(!"Graphics pipeline not bound!"); \
return; \
#define CHECK_SAMPLER_TEXTURES \
RenderPass *rp = (RenderPass *)render_pass; \
for (Uint32 color_target_index = 0; color_target_index < rp->num_color_targets; color_target_index += 1) { \
for (Uint32 texture_sampler_index = 0; texture_sampler_index < num_bindings; texture_sampler_index += 1) { \
if (rp->color_targets[color_target_index] == texture_sampler_bindings[texture_sampler_index].texture) { \
SDL_assert_release(!"Texture cannot be simultaneously bound as a color target and a sampler!"); \
} \
} \
} \
\
for (Uint32 texture_sampler_index = 0; texture_sampler_index < num_bindings; texture_sampler_index += 1) { \
if (rp->depth_stencil_target != NULL && rp->depth_stencil_target == texture_sampler_bindings[texture_sampler_index].texture) { \
SDL_assert_release(!"Texture cannot be simultaneously bound as a depth stencil target and a sampler!"); \
} \
}
#define CHECK_STORAGE_TEXTURES \
RenderPass *rp = (RenderPass *)render_pass; \
for (Uint32 color_target_index = 0; color_target_index < rp->num_color_targets; color_target_index += 1) { \
for (Uint32 texture_sampler_index = 0; texture_sampler_index < num_bindings; texture_sampler_index += 1) { \
if (rp->color_targets[color_target_index] == storage_textures[texture_sampler_index]) { \
SDL_assert_release(!"Texture cannot be simultaneously bound as a color target and a storage texture!"); \
} \
} \
} \
\
for (Uint32 texture_sampler_index = 0; texture_sampler_index < num_bindings; texture_sampler_index += 1) { \
if (rp->depth_stencil_target != NULL && rp->depth_stencil_target == storage_textures[texture_sampler_index]) { \
SDL_assert_release(!"Texture cannot be simultaneously bound as a depth stencil target and a storage texture!"); \
} \
}
#define CHECK_GRAPHICS_PIPELINE_BOUND \
if (!((RenderPass *)render_pass)->graphics_pipeline) { \
SDL_assert_release(!"Graphics pipeline not bound!"); \
return; \
}
#define CHECK_COMPUTEPASS \
@@ -74,7 +106,7 @@
}
#define CHECK_COMPUTE_PIPELINE_BOUND \
if (!((CommandBufferCommonHeader *)COMPUTEPASS_COMMAND_BUFFER)->compute_pipeline_bound) { \
if (!((ComputePass *)compute_pass)->compute_pipeline) { \
SDL_assert_release(!"Compute pipeline not bound!"); \
return; \
}
@@ -137,27 +169,144 @@
((CommandBufferCommonHeader *)command_buffer)->device
#define RENDERPASS_COMMAND_BUFFER \
((Pass *)render_pass)->command_buffer
((RenderPass *)render_pass)->command_buffer
#define RENDERPASS_DEVICE \
((CommandBufferCommonHeader *)RENDERPASS_COMMAND_BUFFER)->device
#define RENDERPASS_BOUND_PIPELINE \
((RenderPass *)render_pass)->graphics_pipeline
#define COMPUTEPASS_COMMAND_BUFFER \
((Pass *)compute_pass)->command_buffer
#define COMPUTEPASS_DEVICE \
((CommandBufferCommonHeader *)COMPUTEPASS_COMMAND_BUFFER)->device
#define COMPUTEPASS_BOUND_PIPELINE \
((ComputePass *)compute_pass)->compute_pipeline
#define COPYPASS_COMMAND_BUFFER \
((Pass *)copy_pass)->command_buffer
#define COPYPASS_DEVICE \
((CommandBufferCommonHeader *)COPYPASS_COMMAND_BUFFER)->device
static bool TextureFormatIsComputeWritable[] = {
false, // INVALID
false, // A8_UNORM
true, // R8_UNORM
true, // R8G8_UNORM
true, // R8G8B8A8_UNORM
true, // R16_UNORM
true, // R16G16_UNORM
true, // R16G16B16A16_UNORM
true, // R10G10B10A2_UNORM
false, // B5G6R5_UNORM
false, // B5G5R5A1_UNORM
false, // B4G4R4A4_UNORM
false, // B8G8R8A8_UNORM
false, // BC1_UNORM
false, // BC2_UNORM
false, // BC3_UNORM
false, // BC4_UNORM
false, // BC5_UNORM
false, // BC7_UNORM
false, // BC6H_FLOAT
false, // BC6H_UFLOAT
true, // R8_SNORM
true, // R8G8_SNORM
true, // R8G8B8A8_SNORM
true, // R16_SNORM
true, // R16G16_SNORM
true, // R16G16B16A16_SNORM
true, // R16_FLOAT
true, // R16G16_FLOAT
true, // R16G16B16A16_FLOAT
true, // R32_FLOAT
true, // R32G32_FLOAT
true, // R32G32B32A32_FLOAT
true, // R11G11B10_UFLOAT
true, // R8_UINT
true, // R8G8_UINT
true, // R8G8B8A8_UINT
true, // R16_UINT
true, // R16G16_UINT
true, // R16G16B16A16_UINT
true, // R32_UINT
true, // R32G32_UINT
true, // R32G32B32A32_UINT
true, // R8_INT
true, // R8G8_INT
true, // R8G8B8A8_INT
true, // R16_INT
true, // R16G16_INT
true, // R16G16B16A16_INT
true, // R32_INT
true, // R32G32_INT
true, // R32G32B32A32_INT
false, // R8G8B8A8_UNORM_SRGB
false, // B8G8R8A8_UNORM_SRGB
false, // BC1_UNORM_SRGB
false, // BC3_UNORM_SRGB
false, // BC3_UNORM_SRGB
false, // BC7_UNORM_SRGB
false, // D16_UNORM
false, // D24_UNORM
false, // D32_FLOAT
false, // D24_UNORM_S8_UINT
false, // D32_FLOAT_S8_UINT
false, // ASTC_4x4_UNORM
false, // ASTC_5x4_UNORM
false, // ASTC_5x5_UNORM
false, // ASTC_6x5_UNORM
false, // ASTC_6x6_UNORM
false, // ASTC_8x5_UNORM
false, // ASTC_8x6_UNORM
false, // ASTC_8x8_UNORM
false, // ASTC_10x5_UNORM
false, // ASTC_10x6_UNORM
false, // ASTC_10x8_UNORM
false, // ASTC_10x10_UNORM
false, // ASTC_12x10_UNORM
false, // ASTC_12x12_UNORM
false, // ASTC_4x4_UNORM_SRGB
false, // ASTC_5x4_UNORM_SRGB
false, // ASTC_5x5_UNORM_SRGB
false, // ASTC_6x5_UNORM_SRGB
false, // ASTC_6x6_UNORM_SRGB
false, // ASTC_8x5_UNORM_SRGB
false, // ASTC_8x6_UNORM_SRGB
false, // ASTC_8x8_UNORM_SRGB
false, // ASTC_10x5_UNORM_SRGB
false, // ASTC_10x6_UNORM_SRGB
false, // ASTC_10x8_UNORM_SRGB
false, // ASTC_10x10_UNORM_SRGB
false, // ASTC_12x10_UNORM_SRGB
false, // ASTC_12x12_UNORM_SRGB
false, // ASTC_4x4_FLOAT
false, // ASTC_5x4_FLOAT
false, // ASTC_5x5_FLOAT
false, // ASTC_6x5_FLOAT
false, // ASTC_6x6_FLOAT
false, // ASTC_8x5_FLOAT
false, // ASTC_8x6_FLOAT
false, // ASTC_8x8_FLOAT
false, // ASTC_10x5_FLOAT
false, // ASTC_10x6_FLOAT
false, // ASTC_10x8_FLOAT
false, // ASTC_10x10_FLOAT
false, // ASTC_12x10_FLOAT
false // ASTC_12x12_FLOAT
};
// Drivers
#ifndef SDL_GPU_DISABLED
static const SDL_GPUBootstrap *backends[] = {
#ifdef SDL_GPU_PRIVATE
&PrivateGPUDriver,
#endif
#ifdef SDL_GPU_METAL
&MetalDriver,
#endif
@@ -368,6 +517,73 @@ void SDL_GPU_BlitCommon(
SDL_EndGPURenderPass(render_pass);
}
static void SDL_GPU_CheckGraphicsBindings(SDL_GPURenderPass *render_pass)
{
RenderPass *rp = (RenderPass *)render_pass;
GraphicsPipelineCommonHeader *pipeline = (GraphicsPipelineCommonHeader *)RENDERPASS_BOUND_PIPELINE;
for (Uint32 i = 0; i < pipeline->num_vertex_samplers; i += 1) {
if (!rp->vertex_sampler_bound[i]) {
SDL_assert_release(!"Missing vertex sampler binding!");
}
}
for (Uint32 i = 0; i < pipeline->num_vertex_storage_textures; i += 1) {
if (!rp->vertex_storage_texture_bound[i]) {
SDL_assert_release(!"Missing vertex storage texture binding!");
}
}
for (Uint32 i = 0; i < pipeline->num_vertex_storage_buffers; i += 1) {
if (!rp->vertex_storage_buffer_bound[i]) {
SDL_assert_release(!"Missing vertex storage buffer binding!");
}
}
for (Uint32 i = 0; i < pipeline->num_fragment_samplers; i += 1) {
if (!rp->fragment_sampler_bound[i]) {
SDL_assert_release(!"Missing fragment sampler binding!");
}
}
for (Uint32 i = 0; i < pipeline->num_fragment_storage_textures; i += 1) {
if (!rp->fragment_storage_texture_bound[i]) {
SDL_assert_release(!"Missing fragment storage texture binding!");
}
}
for (Uint32 i = 0; i < pipeline->num_fragment_storage_buffers; i += 1) {
if (!rp->fragment_storage_buffer_bound[i]) {
SDL_assert_release(!"Missing fragment storage buffer binding!");
}
}
}
static void SDL_GPU_CheckComputeBindings(SDL_GPUComputePass *compute_pass)
{
ComputePass *cp = (ComputePass *)compute_pass;
ComputePipelineCommonHeader *pipeline = (ComputePipelineCommonHeader *)COMPUTEPASS_BOUND_PIPELINE;
for (Uint32 i = 0; i < pipeline->numSamplers; i += 1) {
if (!cp->sampler_bound[i]) {
SDL_assert_release(!"Missing compute sampler binding!");
}
}
for (Uint32 i = 0; i < pipeline->numReadonlyStorageTextures; i += 1) {
if (!cp->read_only_storage_texture_bound[i]) {
SDL_assert_release(!"Missing compute readonly storage texture binding!");
}
}
for (Uint32 i = 0; i < pipeline->numReadonlyStorageBuffers; i += 1) {
if (!cp->read_only_storage_buffer_bound[i]) {
SDL_assert_release(!"Missing compute readonly storage buffer binding!");
}
}
for (Uint32 i = 0; i < pipeline->numReadWriteStorageTextures; i += 1) {
if (!cp->read_write_storage_texture_bound[i]) {
SDL_assert_release(!"Missing compute read-write storage texture binding!");
}
}
for (Uint32 i = 0; i < pipeline->numReadWriteStorageBuffers; i += 1) {
if (!cp->read_write_storage_buffer_bound[i]) {
SDL_assert_release(!"Missing compute read-write storage buffer bbinding!");
}
}
}
// Driver Functions
#ifndef SDL_GPU_DISABLED
@@ -529,7 +745,6 @@ SDL_GPUDevice *SDL_CreateGPUDeviceWithProperties(SDL_PropertiesID props)
result = selectedBackend->CreateDevice(debug_mode, preferLowPower, props);
if (result != NULL) {
result->backend = selectedBackend->name;
result->shader_formats = selectedBackend->shader_formats;
result->debug_mode = debug_mode;
}
}
@@ -605,6 +820,7 @@ Uint32 SDL_GPUTextureFormatTexelBlockSize(
case SDL_GPU_TEXTUREFORMAT_R8_SNORM:
case SDL_GPU_TEXTUREFORMAT_A8_UNORM:
case SDL_GPU_TEXTUREFORMAT_R8_UINT:
case SDL_GPU_TEXTUREFORMAT_R8_INT:
return 1;
case SDL_GPU_TEXTUREFORMAT_B5G6R5_UNORM:
case SDL_GPU_TEXTUREFORMAT_B4G4R4A4_UNORM:
@@ -613,9 +829,11 @@ Uint32 SDL_GPUTextureFormatTexelBlockSize(
case SDL_GPU_TEXTUREFORMAT_R8G8_SNORM:
case SDL_GPU_TEXTUREFORMAT_R8G8_UNORM:
case SDL_GPU_TEXTUREFORMAT_R8G8_UINT:
case SDL_GPU_TEXTUREFORMAT_R8G8_INT:
case SDL_GPU_TEXTUREFORMAT_R16_UNORM:
case SDL_GPU_TEXTUREFORMAT_R16_SNORM:
case SDL_GPU_TEXTUREFORMAT_R16_UINT:
case SDL_GPU_TEXTUREFORMAT_R16_INT:
case SDL_GPU_TEXTUREFORMAT_D16_UNORM:
return 2;
case SDL_GPU_TEXTUREFORMAT_R8G8B8A8_UNORM:
@@ -628,11 +846,15 @@ Uint32 SDL_GPUTextureFormatTexelBlockSize(
case SDL_GPU_TEXTUREFORMAT_R8G8B8A8_SNORM:
case SDL_GPU_TEXTUREFORMAT_R10G10B10A2_UNORM:
case SDL_GPU_TEXTUREFORMAT_R8G8B8A8_UINT:
case SDL_GPU_TEXTUREFORMAT_R8G8B8A8_INT:
case SDL_GPU_TEXTUREFORMAT_R16G16_UINT:
case SDL_GPU_TEXTUREFORMAT_R16G16_INT:
case SDL_GPU_TEXTUREFORMAT_R16G16_UNORM:
case SDL_GPU_TEXTUREFORMAT_R16G16_SNORM:
case SDL_GPU_TEXTUREFORMAT_D24_UNORM:
case SDL_GPU_TEXTUREFORMAT_D32_FLOAT:
case SDL_GPU_TEXTUREFORMAT_R32_UINT:
case SDL_GPU_TEXTUREFORMAT_R32_INT:
case SDL_GPU_TEXTUREFORMAT_D24_UNORM_S8_UINT:
return 4;
case SDL_GPU_TEXTUREFORMAT_D32_FLOAT_S8_UINT:
@@ -640,10 +862,15 @@ Uint32 SDL_GPUTextureFormatTexelBlockSize(
case SDL_GPU_TEXTUREFORMAT_R16G16B16A16_FLOAT:
case SDL_GPU_TEXTUREFORMAT_R16G16B16A16_UNORM:
case SDL_GPU_TEXTUREFORMAT_R16G16B16A16_SNORM:
case SDL_GPU_TEXTUREFORMAT_R32G32_FLOAT:
case SDL_GPU_TEXTUREFORMAT_R16G16B16A16_UINT:
case SDL_GPU_TEXTUREFORMAT_R16G16B16A16_INT:
case SDL_GPU_TEXTUREFORMAT_R32G32_FLOAT:
case SDL_GPU_TEXTUREFORMAT_R32G32_UINT:
case SDL_GPU_TEXTUREFORMAT_R32G32_INT:
return 8;
case SDL_GPU_TEXTUREFORMAT_R32G32B32A32_FLOAT:
case SDL_GPU_TEXTUREFORMAT_R32G32B32A32_INT:
case SDL_GPU_TEXTUREFORMAT_R32G32B32A32_UINT:
return 16;
case SDL_GPU_TEXTUREFORMAT_ASTC_4x4_UNORM:
case SDL_GPU_TEXTUREFORMAT_ASTC_5x4_UNORM:
@@ -706,6 +933,13 @@ bool SDL_GPUTextureSupportsFormat(
CHECK_TEXTUREFORMAT_ENUM_INVALID(format, false)
}
if ((usage & SDL_GPU_TEXTUREUSAGE_COMPUTE_STORAGE_WRITE) ||
(usage & SDL_GPU_TEXTUREUSAGE_COMPUTE_STORAGE_SIMULTANEOUS_READ_WRITE)) {
if (!TextureFormatIsComputeWritable[format]) {
return false;
}
}
return device->SupportsTextureFormat(
device->driverData,
format,
@@ -842,6 +1076,12 @@ SDL_GPUGraphicsPipeline *SDL_CreateGPUGraphicsPipeline(
SDL_assert_release(!"The number of vertex attributes in a vertex input state must not exceed 16!");
return NULL;
}
for (Uint32 i = 0; i < graphicsPipelineCreateInfo->vertex_input_state.num_vertex_buffers; i += 1) {
if (graphicsPipelineCreateInfo->vertex_input_state.vertex_buffer_descriptions[i].instance_step_rate != 0) {
SDL_assert_release(!"For all vertex buffer descriptions, instance_step_rate must be 0!");
return NULL;
}
}
Uint32 locations[MAX_VERTEX_ATTRIBUTES];
for (Uint32 i = 0; i < graphicsPipelineCreateInfo->vertex_input_state.num_vertex_attributes; i += 1) {
CHECK_VERTEXELEMENTFORMAT_ENUM_INVALID(graphicsPipelineCreateInfo->vertex_input_state.vertex_attributes[i].format, NULL);
@@ -850,9 +1090,18 @@ SDL_GPUGraphicsPipeline *SDL_CreateGPUGraphicsPipeline(
for (Uint32 j = 0; j < i; j += 1) {
if (locations[j] == locations[i]) {
SDL_assert_release(!"Each vertex attribute location in a vertex input state must be unique!");
return NULL;
}
}
}
if (graphicsPipelineCreateInfo->multisample_state.enable_mask) {
SDL_assert_release(!"For multisample states, enable_mask must be false!");
return NULL;
}
if (graphicsPipelineCreateInfo->multisample_state.sample_mask != 0) {
SDL_assert_release(!"For multisample states, sample_mask must be 0!");
return NULL;
}
if (graphicsPipelineCreateInfo->depth_stencil_state.enable_depth_test) {
CHECK_COMPAREOP_ENUM_INVALID(graphicsPipelineCreateInfo->depth_stencil_state.compare_op, NULL)
}
@@ -1313,15 +1562,30 @@ SDL_GPUCommandBuffer *SDL_AcquireGPUCommandBuffer(
commandBufferHeader = (CommandBufferCommonHeader *)command_buffer;
commandBufferHeader->device = device;
commandBufferHeader->render_pass.command_buffer = command_buffer;
commandBufferHeader->render_pass.in_progress = false;
commandBufferHeader->graphics_pipeline_bound = false;
commandBufferHeader->compute_pass.command_buffer = command_buffer;
commandBufferHeader->compute_pass.in_progress = false;
commandBufferHeader->compute_pipeline_bound = false;
commandBufferHeader->copy_pass.command_buffer = command_buffer;
commandBufferHeader->copy_pass.in_progress = false;
commandBufferHeader->swapchain_texture_acquired = false;
commandBufferHeader->submitted = false;
if (device->debug_mode) {
commandBufferHeader->render_pass.in_progress = false;
commandBufferHeader->render_pass.graphics_pipeline = NULL;
commandBufferHeader->compute_pass.in_progress = false;
commandBufferHeader->compute_pass.compute_pipeline = NULL;
commandBufferHeader->copy_pass.in_progress = false;
commandBufferHeader->swapchain_texture_acquired = false;
commandBufferHeader->submitted = false;
commandBufferHeader->ignore_render_pass_texture_validation = false;
SDL_zeroa(commandBufferHeader->render_pass.vertex_sampler_bound);
SDL_zeroa(commandBufferHeader->render_pass.vertex_storage_texture_bound);
SDL_zeroa(commandBufferHeader->render_pass.vertex_storage_buffer_bound);
SDL_zeroa(commandBufferHeader->render_pass.fragment_sampler_bound);
SDL_zeroa(commandBufferHeader->render_pass.fragment_storage_texture_bound);
SDL_zeroa(commandBufferHeader->render_pass.fragment_storage_buffer_bound);
SDL_zeroa(commandBufferHeader->compute_pass.sampler_bound);
SDL_zeroa(commandBufferHeader->compute_pass.read_only_storage_texture_bound);
SDL_zeroa(commandBufferHeader->compute_pass.read_only_storage_buffer_bound);
SDL_zeroa(commandBufferHeader->compute_pass.read_write_storage_texture_bound);
SDL_zeroa(commandBufferHeader->compute_pass.read_write_storage_buffer_bound);
}
return command_buffer;
}
@@ -1439,30 +1703,47 @@ SDL_GPURenderPass *SDL_BeginGPURenderPass(
if (color_target_infos[i].cycle && color_target_infos[i].load_op == SDL_GPU_LOADOP_LOAD) {
SDL_assert_release(!"Cannot cycle color target when load op is LOAD!");
return NULL;
}
if (color_target_infos[i].store_op == SDL_GPU_STOREOP_RESOLVE || color_target_infos[i].store_op == SDL_GPU_STOREOP_RESOLVE_AND_STORE) {
if (color_target_infos[i].resolve_texture == NULL) {
SDL_assert_release(!"Store op is RESOLVE or RESOLVE_AND_STORE but resolve_texture is NULL!");
return NULL;
} else {
TextureCommonHeader *resolveTextureHeader = (TextureCommonHeader *)color_target_infos[i].resolve_texture;
if (textureHeader->info.sample_count == SDL_GPU_SAMPLECOUNT_1) {
SDL_assert_release(!"Store op is RESOLVE or RESOLVE_AND_STORE but texture is not multisample!");
return NULL;
}
if (resolveTextureHeader->info.sample_count != SDL_GPU_SAMPLECOUNT_1) {
SDL_assert_release(!"Resolve texture must have a sample count of 1!");
return NULL;
}
if (resolveTextureHeader->info.format != textureHeader->info.format) {
SDL_assert_release(!"Resolve texture must have the same format as its corresponding color target!");
return NULL;
}
if (resolveTextureHeader->info.type == SDL_GPU_TEXTURETYPE_3D) {
SDL_assert_release(!"Resolve texture must not be of TEXTURETYPE_3D!");
return NULL;
}
if (!(resolveTextureHeader->info.usage & SDL_GPU_TEXTUREUSAGE_COLOR_TARGET)) {
SDL_assert_release(!"Resolve texture usage must include COLOR_TARGET!");
return NULL;
}
}
}
if (color_target_infos[i].layer_or_depth_plane >= textureHeader->info.layer_count_or_depth) {
SDL_assert_release(!"Color target layer index must be less than the texture's layer count!");
return NULL;
}
if (color_target_infos[i].mip_level >= textureHeader->info.num_levels) {
SDL_assert_release(!"Color target mip level must be less than the texture's level count!");
return NULL;
}
}
if (depth_stencil_target_info != NULL) {
@@ -1470,10 +1751,12 @@ SDL_GPURenderPass *SDL_BeginGPURenderPass(
TextureCommonHeader *textureHeader = (TextureCommonHeader *)depth_stencil_target_info->texture;
if (!(textureHeader->info.usage & SDL_GPU_TEXTUREUSAGE_DEPTH_STENCIL_TARGET)) {
SDL_assert_release(!"Depth target must have been created with the DEPTH_STENCIL_TARGET usage flag!");
return NULL;
}
if (depth_stencil_target_info->cycle && (depth_stencil_target_info->load_op == SDL_GPU_LOADOP_LOAD || depth_stencil_target_info->stencil_load_op == SDL_GPU_LOADOP_LOAD)) {
SDL_assert_release(!"Cannot cycle depth target when load op or stencil load op is LOAD!");
return NULL;
}
if (depth_stencil_target_info->store_op == SDL_GPU_STOREOP_RESOLVE ||
@@ -1481,6 +1764,7 @@ SDL_GPURenderPass *SDL_BeginGPURenderPass(
depth_stencil_target_info->store_op == SDL_GPU_STOREOP_RESOLVE_AND_STORE ||
depth_stencil_target_info->stencil_store_op == SDL_GPU_STOREOP_RESOLVE_AND_STORE) {
SDL_assert_release(!"RESOLVE store ops are not supported for depth-stencil targets!");
return NULL;
}
}
}
@@ -1492,7 +1776,18 @@ SDL_GPURenderPass *SDL_BeginGPURenderPass(
depth_stencil_target_info);
commandBufferHeader = (CommandBufferCommonHeader *)command_buffer;
commandBufferHeader->render_pass.in_progress = true;
if (COMMAND_BUFFER_DEVICE->debug_mode) {
commandBufferHeader->render_pass.in_progress = true;
for (Uint32 i = 0; i < num_color_targets; i += 1) {
commandBufferHeader->render_pass.color_targets[i] = color_target_infos[i].texture;
}
commandBufferHeader->render_pass.num_color_targets = num_color_targets;
if (depth_stencil_target_info != NULL) {
commandBufferHeader->render_pass.depth_stencil_target = depth_stencil_target_info->texture;
}
}
return (SDL_GPURenderPass *)&(commandBufferHeader->render_pass);
}
@@ -1500,8 +1795,6 @@ void SDL_BindGPUGraphicsPipeline(
SDL_GPURenderPass *render_pass,
SDL_GPUGraphicsPipeline *graphics_pipeline)
{
CommandBufferCommonHeader *commandBufferHeader;
if (render_pass == NULL) {
SDL_InvalidParamError("render_pass");
return;
@@ -1515,8 +1808,10 @@ void SDL_BindGPUGraphicsPipeline(
RENDERPASS_COMMAND_BUFFER,
graphics_pipeline);
commandBufferHeader = (CommandBufferCommonHeader *)RENDERPASS_COMMAND_BUFFER;
commandBufferHeader->graphics_pipeline_bound = true;
if (RENDERPASS_DEVICE->debug_mode) {
RENDERPASS_BOUND_PIPELINE = graphics_pipeline;
}
}
void SDL_SetGPUViewport(
@@ -1666,6 +1961,15 @@ void SDL_BindGPUVertexSamplers(
if (RENDERPASS_DEVICE->debug_mode) {
CHECK_RENDERPASS
if (!((CommandBufferCommonHeader*)RENDERPASS_COMMAND_BUFFER)->ignore_render_pass_texture_validation)
{
CHECK_SAMPLER_TEXTURES
}
for (Uint32 i = 0; i < num_bindings; i += 1) {
((RenderPass *)render_pass)->vertex_sampler_bound[first_slot + i] = true;
}
}
RENDERPASS_DEVICE->BindVertexSamplers(
@@ -1692,6 +1996,11 @@ void SDL_BindGPUVertexStorageTextures(
if (RENDERPASS_DEVICE->debug_mode) {
CHECK_RENDERPASS
CHECK_STORAGE_TEXTURES
for (Uint32 i = 0; i < num_bindings; i += 1) {
((RenderPass *)render_pass)->vertex_storage_texture_bound[first_slot + i] = true;
}
}
RENDERPASS_DEVICE->BindVertexStorageTextures(
@@ -1718,6 +2027,10 @@ void SDL_BindGPUVertexStorageBuffers(
if (RENDERPASS_DEVICE->debug_mode) {
CHECK_RENDERPASS
for (Uint32 i = 0; i < num_bindings; i += 1) {
((RenderPass *)render_pass)->vertex_storage_buffer_bound[first_slot + i] = true;
}
}
RENDERPASS_DEVICE->BindVertexStorageBuffers(
@@ -1744,6 +2057,14 @@ void SDL_BindGPUFragmentSamplers(
if (RENDERPASS_DEVICE->debug_mode) {
CHECK_RENDERPASS
if (!((CommandBufferCommonHeader*)RENDERPASS_COMMAND_BUFFER)->ignore_render_pass_texture_validation) {
CHECK_SAMPLER_TEXTURES
}
for (Uint32 i = 0; i < num_bindings; i += 1) {
((RenderPass *)render_pass)->fragment_sampler_bound[first_slot + i] = true;
}
}
RENDERPASS_DEVICE->BindFragmentSamplers(
@@ -1770,6 +2091,11 @@ void SDL_BindGPUFragmentStorageTextures(
if (RENDERPASS_DEVICE->debug_mode) {
CHECK_RENDERPASS
CHECK_STORAGE_TEXTURES
for (Uint32 i = 0; i < num_bindings; i += 1) {
((RenderPass *)render_pass)->fragment_storage_texture_bound[first_slot + i] = true;
}
}
RENDERPASS_DEVICE->BindFragmentStorageTextures(
@@ -1796,6 +2122,10 @@ void SDL_BindGPUFragmentStorageBuffers(
if (RENDERPASS_DEVICE->debug_mode) {
CHECK_RENDERPASS
for (Uint32 i = 0; i < num_bindings; i += 1) {
((RenderPass *)render_pass)->fragment_storage_buffer_bound[first_slot + i] = true;
}
}
RENDERPASS_DEVICE->BindFragmentStorageBuffers(
@@ -1821,6 +2151,7 @@ void SDL_DrawGPUIndexedPrimitives(
if (RENDERPASS_DEVICE->debug_mode) {
CHECK_RENDERPASS
CHECK_GRAPHICS_PIPELINE_BOUND
SDL_GPU_CheckGraphicsBindings(render_pass);
}
RENDERPASS_DEVICE->DrawIndexedPrimitives(
@@ -1847,6 +2178,7 @@ void SDL_DrawGPUPrimitives(
if (RENDERPASS_DEVICE->debug_mode) {
CHECK_RENDERPASS
CHECK_GRAPHICS_PIPELINE_BOUND
SDL_GPU_CheckGraphicsBindings(render_pass);
}
RENDERPASS_DEVICE->DrawPrimitives(
@@ -1875,6 +2207,7 @@ void SDL_DrawGPUPrimitivesIndirect(
if (RENDERPASS_DEVICE->debug_mode) {
CHECK_RENDERPASS
CHECK_GRAPHICS_PIPELINE_BOUND
SDL_GPU_CheckGraphicsBindings(render_pass);
}
RENDERPASS_DEVICE->DrawPrimitivesIndirect(
@@ -1902,6 +2235,7 @@ void SDL_DrawGPUIndexedPrimitivesIndirect(
if (RENDERPASS_DEVICE->debug_mode) {
CHECK_RENDERPASS
CHECK_GRAPHICS_PIPELINE_BOUND
SDL_GPU_CheckGraphicsBindings(render_pass);
}
RENDERPASS_DEVICE->DrawIndexedPrimitivesIndirect(
@@ -1914,13 +2248,14 @@ void SDL_DrawGPUIndexedPrimitivesIndirect(
void SDL_EndGPURenderPass(
SDL_GPURenderPass *render_pass)
{
CommandBufferCommonHeader *commandBufferCommonHeader;
if (render_pass == NULL) {
SDL_InvalidParamError("render_pass");
return;
}
CommandBufferCommonHeader *commandBufferCommonHeader;
commandBufferCommonHeader = (CommandBufferCommonHeader *)RENDERPASS_COMMAND_BUFFER;
if (RENDERPASS_DEVICE->debug_mode) {
CHECK_RENDERPASS
}
@@ -1928,9 +2263,22 @@ void SDL_EndGPURenderPass(
RENDERPASS_DEVICE->EndRenderPass(
RENDERPASS_COMMAND_BUFFER);
commandBufferCommonHeader = (CommandBufferCommonHeader *)RENDERPASS_COMMAND_BUFFER;
commandBufferCommonHeader->render_pass.in_progress = false;
commandBufferCommonHeader->graphics_pipeline_bound = false;
if (RENDERPASS_DEVICE->debug_mode) {
commandBufferCommonHeader->render_pass.in_progress = false;
for (Uint32 i = 0; i < MAX_COLOR_TARGET_BINDINGS; i += 1)
{
commandBufferCommonHeader->render_pass.color_targets[i] = NULL;
}
commandBufferCommonHeader->render_pass.num_color_targets = 0;
commandBufferCommonHeader->render_pass.depth_stencil_target = NULL;
commandBufferCommonHeader->render_pass.graphics_pipeline = NULL;
SDL_zeroa(commandBufferCommonHeader->render_pass.vertex_sampler_bound);
SDL_zeroa(commandBufferCommonHeader->render_pass.vertex_storage_texture_bound);
SDL_zeroa(commandBufferCommonHeader->render_pass.vertex_storage_buffer_bound);
SDL_zeroa(commandBufferCommonHeader->render_pass.fragment_sampler_bound);
SDL_zeroa(commandBufferCommonHeader->render_pass.fragment_storage_texture_bound);
SDL_zeroa(commandBufferCommonHeader->render_pass.fragment_storage_buffer_bound);
}
}
// Compute Pass
@@ -1974,6 +2322,16 @@ SDL_GPUComputePass *SDL_BeginGPUComputePass(
SDL_assert_release(!"Texture must be created with COMPUTE_STORAGE_WRITE or COMPUTE_STORAGE_SIMULTANEOUS_READ_WRITE flag");
return NULL;
}
if (storage_texture_bindings[i].layer >= header->info.layer_count_or_depth) {
SDL_assert_release(!"Storage texture layer index must be less than the texture's layer count!");
return NULL;
}
if (storage_texture_bindings[i].mip_level >= header->info.num_levels) {
SDL_assert_release(!"Storage texture mip level must be less than the texture's level count!");
return NULL;
}
}
// TODO: validate buffer usage?
@@ -1987,7 +2345,19 @@ SDL_GPUComputePass *SDL_BeginGPUComputePass(
num_storage_buffer_bindings);
commandBufferHeader = (CommandBufferCommonHeader *)command_buffer;
commandBufferHeader->compute_pass.in_progress = true;
if (COMMAND_BUFFER_DEVICE->debug_mode) {
commandBufferHeader->compute_pass.in_progress = true;
for (Uint32 i = 0; i < num_storage_texture_bindings; i += 1) {
commandBufferHeader->compute_pass.read_write_storage_texture_bound[i] = true;
}
for (Uint32 i = 0; i < num_storage_buffer_bindings; i += 1) {
commandBufferHeader->compute_pass.read_write_storage_buffer_bound[i] = true;
}
}
return (SDL_GPUComputePass *)&(commandBufferHeader->compute_pass);
}
@@ -1995,8 +2365,6 @@ void SDL_BindGPUComputePipeline(
SDL_GPUComputePass *compute_pass,
SDL_GPUComputePipeline *compute_pipeline)
{
CommandBufferCommonHeader *commandBufferHeader;
if (compute_pass == NULL) {
SDL_InvalidParamError("compute_pass");
return;
@@ -2014,8 +2382,10 @@ void SDL_BindGPUComputePipeline(
COMPUTEPASS_COMMAND_BUFFER,
compute_pipeline);
commandBufferHeader = (CommandBufferCommonHeader *)COMPUTEPASS_COMMAND_BUFFER;
commandBufferHeader->compute_pipeline_bound = true;
if (COMPUTEPASS_DEVICE->debug_mode) {
COMPUTEPASS_BOUND_PIPELINE = compute_pipeline;
}
}
void SDL_BindGPUComputeSamplers(
@@ -2035,6 +2405,10 @@ void SDL_BindGPUComputeSamplers(
if (COMPUTEPASS_DEVICE->debug_mode) {
CHECK_COMPUTEPASS
for (Uint32 i = 0; i < num_bindings; i += 1) {
((ComputePass *)compute_pass)->sampler_bound[first_slot + i] = true;
}
}
COMPUTEPASS_DEVICE->BindComputeSamplers(
@@ -2061,6 +2435,10 @@ void SDL_BindGPUComputeStorageTextures(
if (COMPUTEPASS_DEVICE->debug_mode) {
CHECK_COMPUTEPASS
for (Uint32 i = 0; i < num_bindings; i += 1) {
((ComputePass *)compute_pass)->read_only_storage_texture_bound[first_slot + i] = true;
}
}
COMPUTEPASS_DEVICE->BindComputeStorageTextures(
@@ -2087,6 +2465,10 @@ void SDL_BindGPUComputeStorageBuffers(
if (COMPUTEPASS_DEVICE->debug_mode) {
CHECK_COMPUTEPASS
for (Uint32 i = 0; i < num_bindings; i += 1) {
((ComputePass *)compute_pass)->read_only_storage_buffer_bound[first_slot + i] = true;
}
}
COMPUTEPASS_DEVICE->BindComputeStorageBuffers(
@@ -2110,6 +2492,7 @@ void SDL_DispatchGPUCompute(
if (COMPUTEPASS_DEVICE->debug_mode) {
CHECK_COMPUTEPASS
CHECK_COMPUTE_PIPELINE_BOUND
SDL_GPU_CheckComputeBindings(compute_pass);
}
COMPUTEPASS_DEVICE->DispatchCompute(
@@ -2132,6 +2515,7 @@ void SDL_DispatchGPUComputeIndirect(
if (COMPUTEPASS_DEVICE->debug_mode) {
CHECK_COMPUTEPASS
CHECK_COMPUTE_PIPELINE_BOUND
SDL_GPU_CheckComputeBindings(compute_pass);
}
COMPUTEPASS_DEVICE->DispatchComputeIndirect(
@@ -2157,9 +2541,16 @@ void SDL_EndGPUComputePass(
COMPUTEPASS_DEVICE->EndComputePass(
COMPUTEPASS_COMMAND_BUFFER);
commandBufferCommonHeader = (CommandBufferCommonHeader *)COMPUTEPASS_COMMAND_BUFFER;
commandBufferCommonHeader->compute_pass.in_progress = false;
commandBufferCommonHeader->compute_pipeline_bound = false;
if (COMPUTEPASS_DEVICE->debug_mode) {
commandBufferCommonHeader = (CommandBufferCommonHeader *)COMPUTEPASS_COMMAND_BUFFER;
commandBufferCommonHeader->compute_pass.in_progress = false;
commandBufferCommonHeader->compute_pass.compute_pipeline = NULL;
SDL_zeroa(commandBufferCommonHeader->compute_pass.sampler_bound);
SDL_zeroa(commandBufferCommonHeader->compute_pass.read_only_storage_texture_bound);
SDL_zeroa(commandBufferCommonHeader->compute_pass.read_only_storage_buffer_bound);
SDL_zeroa(commandBufferCommonHeader->compute_pass.read_write_storage_texture_bound);
SDL_zeroa(commandBufferCommonHeader->compute_pass.read_write_storage_buffer_bound);
}
}
// TransferBuffer Data
@@ -2217,7 +2608,11 @@ SDL_GPUCopyPass *SDL_BeginGPUCopyPass(
command_buffer);
commandBufferHeader = (CommandBufferCommonHeader *)command_buffer;
commandBufferHeader->copy_pass.in_progress = true;
if (COMMAND_BUFFER_DEVICE->debug_mode) {
commandBufferHeader->copy_pass.in_progress = true;
}
return (SDL_GPUCopyPass *)&(commandBufferHeader->copy_pass);
}
@@ -2468,7 +2863,9 @@ void SDL_EndGPUCopyPass(
COPYPASS_DEVICE->EndCopyPass(
COPYPASS_COMMAND_BUFFER);
((CommandBufferCommonHeader *)COPYPASS_COMMAND_BUFFER)->copy_pass.in_progress = false;
if (COPYPASS_DEVICE->debug_mode) {
((CommandBufferCommonHeader *)COPYPASS_COMMAND_BUFFER)->copy_pass.in_progress = false;
}
}
void SDL_GenerateMipmapsForGPUTexture(
@@ -2498,11 +2895,19 @@ void SDL_GenerateMipmapsForGPUTexture(
SDL_assert_release(!"GenerateMipmaps texture must be created with SAMPLER and COLOR_TARGET usage flags!");
return;
}
CommandBufferCommonHeader *commandBufferHeader = (CommandBufferCommonHeader *)command_buffer;
commandBufferHeader->ignore_render_pass_texture_validation = true;
}
COMMAND_BUFFER_DEVICE->GenerateMipmaps(
command_buffer,
texture);
if (COMMAND_BUFFER_DEVICE->debug_mode) {
CommandBufferCommonHeader *commandBufferHeader = (CommandBufferCommonHeader *)command_buffer;
commandBufferHeader->ignore_render_pass_texture_validation = false;
}
}
void SDL_BlitGPUTexture(
@@ -2616,8 +3021,11 @@ bool SDL_ClaimWindowForGPUDevice(
{
CHECK_DEVICE_MAGIC(device, false);
if (window == NULL) {
SDL_InvalidParamError("window");
return false;
return SDL_InvalidParamError("window");
}
if ((window->flags & SDL_WINDOW_TRANSPARENT) != 0) {
return SDL_SetError("The GPU API doesn't support transparent windows");
}
return device->ClaimWindow(
libs/SDL3/src/gpu/SDL_sysgpu.h
+92 -20
View File
@@ -24,6 +24,21 @@
#ifndef SDL_GPU_DRIVER_H
#define SDL_GPU_DRIVER_H
// GraphicsDevice Limits
#define MAX_TEXTURE_SAMPLERS_PER_STAGE 16
#define MAX_STORAGE_TEXTURES_PER_STAGE 8
#define MAX_STORAGE_BUFFERS_PER_STAGE 8
#define MAX_UNIFORM_BUFFERS_PER_STAGE 4
#define MAX_COMPUTE_WRITE_TEXTURES 8
#define MAX_COMPUTE_WRITE_BUFFERS 8
#define UNIFORM_BUFFER_SIZE 32768
#define MAX_VERTEX_BUFFERS 16
#define MAX_VERTEX_ATTRIBUTES 16
#define MAX_COLOR_TARGET_BINDINGS 4
#define MAX_PRESENT_COUNT 16
#define MAX_FRAMES_IN_FLIGHT 3
// Common Structs
typedef struct Pass
@@ -32,16 +47,51 @@ typedef struct Pass
bool in_progress;
} Pass;
typedef struct ComputePass
{
SDL_GPUCommandBuffer *command_buffer;
bool in_progress;
SDL_GPUComputePipeline *compute_pipeline;
bool sampler_bound[MAX_TEXTURE_SAMPLERS_PER_STAGE];
bool read_only_storage_texture_bound[MAX_STORAGE_TEXTURES_PER_STAGE];
bool read_only_storage_buffer_bound[MAX_STORAGE_BUFFERS_PER_STAGE];
bool read_write_storage_texture_bound[MAX_COMPUTE_WRITE_TEXTURES];
bool read_write_storage_buffer_bound[MAX_COMPUTE_WRITE_BUFFERS];
} ComputePass;
typedef struct RenderPass
{
SDL_GPUCommandBuffer *command_buffer;
bool in_progress;
SDL_GPUTexture *color_targets[MAX_COLOR_TARGET_BINDINGS];
Uint32 num_color_targets;
SDL_GPUTexture *depth_stencil_target;
SDL_GPUGraphicsPipeline *graphics_pipeline;
bool vertex_sampler_bound[MAX_TEXTURE_SAMPLERS_PER_STAGE];
bool vertex_storage_texture_bound[MAX_STORAGE_TEXTURES_PER_STAGE];
bool vertex_storage_buffer_bound[MAX_STORAGE_BUFFERS_PER_STAGE];
bool fragment_sampler_bound[MAX_TEXTURE_SAMPLERS_PER_STAGE];
bool fragment_storage_texture_bound[MAX_STORAGE_TEXTURES_PER_STAGE];
bool fragment_storage_buffer_bound[MAX_STORAGE_BUFFERS_PER_STAGE];
} RenderPass;
typedef struct CommandBufferCommonHeader
{
SDL_GPUDevice *device;
Pass render_pass;
bool graphics_pipeline_bound;
Pass compute_pass;
bool compute_pipeline_bound;
RenderPass render_pass;
ComputePass compute_pass;
Pass copy_pass;
bool swapchain_texture_acquired;
bool submitted;
// used to avoid tripping assert on GenerateMipmaps
bool ignore_render_pass_texture_validation;
} CommandBufferCommonHeader;
typedef struct TextureCommonHeader
@@ -49,6 +99,29 @@ typedef struct TextureCommonHeader
SDL_GPUTextureCreateInfo info;
} TextureCommonHeader;
typedef struct GraphicsPipelineCommonHeader
{
Uint32 num_vertex_samplers;
Uint32 num_vertex_storage_textures;
Uint32 num_vertex_storage_buffers;
Uint32 num_vertex_uniform_buffers;
Uint32 num_fragment_samplers;
Uint32 num_fragment_storage_textures;
Uint32 num_fragment_storage_buffers;
Uint32 num_fragment_uniform_buffers;
} GraphicsPipelineCommonHeader;
typedef struct ComputePipelineCommonHeader
{
Uint32 numSamplers;
Uint32 numReadonlyStorageTextures;
Uint32 numReadonlyStorageBuffers;
Uint32 numReadWriteStorageTextures;
Uint32 numReadWriteStorageBuffers;
Uint32 numUniformBuffers;
} ComputePipelineCommonHeader;
typedef struct BlitFragmentUniforms
{
// texcoord space
@@ -136,6 +209,7 @@ static inline Sint32 Texture_GetBlockWidth(
case SDL_GPU_TEXTUREFORMAT_BC6H_RGB_FLOAT:
case SDL_GPU_TEXTUREFORMAT_BC6H_RGB_UFLOAT:
case SDL_GPU_TEXTUREFORMAT_BC1_RGBA_UNORM_SRGB:
case SDL_GPU_TEXTUREFORMAT_BC2_RGBA_UNORM_SRGB:
case SDL_GPU_TEXTUREFORMAT_BC3_RGBA_UNORM_SRGB:
case SDL_GPU_TEXTUREFORMAT_BC7_RGBA_UNORM_SRGB:
case SDL_GPU_TEXTUREFORMAT_ASTC_4x4_UNORM:
@@ -173,12 +247,18 @@ static inline Sint32 Texture_GetBlockWidth(
case SDL_GPU_TEXTUREFORMAT_R16_UINT:
case SDL_GPU_TEXTUREFORMAT_R16G16_UINT:
case SDL_GPU_TEXTUREFORMAT_R16G16B16A16_UINT:
case SDL_GPU_TEXTUREFORMAT_R32_UINT:
case SDL_GPU_TEXTUREFORMAT_R32G32_UINT:
case SDL_GPU_TEXTUREFORMAT_R32G32B32A32_UINT:
case SDL_GPU_TEXTUREFORMAT_R8_INT:
case SDL_GPU_TEXTUREFORMAT_R8G8_INT:
case SDL_GPU_TEXTUREFORMAT_R8G8B8A8_INT:
case SDL_GPU_TEXTUREFORMAT_R16_INT:
case SDL_GPU_TEXTUREFORMAT_R16G16_INT:
case SDL_GPU_TEXTUREFORMAT_R16G16B16A16_INT:
case SDL_GPU_TEXTUREFORMAT_R32_INT:
case SDL_GPU_TEXTUREFORMAT_R32G32_INT:
case SDL_GPU_TEXTUREFORMAT_R32G32B32A32_INT:
case SDL_GPU_TEXTUREFORMAT_R8G8B8A8_UNORM_SRGB:
case SDL_GPU_TEXTUREFORMAT_B8G8R8A8_UNORM_SRGB:
case SDL_GPU_TEXTUREFORMAT_D16_UNORM:
@@ -247,6 +327,7 @@ static inline Sint32 Texture_GetBlockHeight(
case SDL_GPU_TEXTUREFORMAT_BC6H_RGB_FLOAT:
case SDL_GPU_TEXTUREFORMAT_BC6H_RGB_UFLOAT:
case SDL_GPU_TEXTUREFORMAT_BC1_RGBA_UNORM_SRGB:
case SDL_GPU_TEXTUREFORMAT_BC2_RGBA_UNORM_SRGB:
case SDL_GPU_TEXTUREFORMAT_BC3_RGBA_UNORM_SRGB:
case SDL_GPU_TEXTUREFORMAT_BC7_RGBA_UNORM_SRGB:
case SDL_GPU_TEXTUREFORMAT_ASTC_5x4_UNORM:
@@ -287,12 +368,18 @@ static inline Sint32 Texture_GetBlockHeight(
case SDL_GPU_TEXTUREFORMAT_R16_UINT:
case SDL_GPU_TEXTUREFORMAT_R16G16_UINT:
case SDL_GPU_TEXTUREFORMAT_R16G16B16A16_UINT:
case SDL_GPU_TEXTUREFORMAT_R32_UINT:
case SDL_GPU_TEXTUREFORMAT_R32G32_UINT:
case SDL_GPU_TEXTUREFORMAT_R32G32B32A32_UINT:
case SDL_GPU_TEXTUREFORMAT_R8_INT:
case SDL_GPU_TEXTUREFORMAT_R8G8_INT:
case SDL_GPU_TEXTUREFORMAT_R8G8B8A8_INT:
case SDL_GPU_TEXTUREFORMAT_R16_INT:
case SDL_GPU_TEXTUREFORMAT_R16G16_INT:
case SDL_GPU_TEXTUREFORMAT_R16G16B16A16_INT:
case SDL_GPU_TEXTUREFORMAT_R32_INT:
case SDL_GPU_TEXTUREFORMAT_R32G32_INT:
case SDL_GPU_TEXTUREFORMAT_R32G32B32A32_INT:
case SDL_GPU_TEXTUREFORMAT_R8G8B8A8_UNORM_SRGB:
case SDL_GPU_TEXTUREFORMAT_B8G8R8A8_UNORM_SRGB:
case SDL_GPU_TEXTUREFORMAT_D16_UNORM:
@@ -373,21 +460,6 @@ static inline Uint32 BytesPerRow(
return blocksPerRow * SDL_GPUTextureFormatTexelBlockSize(format);
}
// GraphicsDevice Limits
#define MAX_TEXTURE_SAMPLERS_PER_STAGE 16
#define MAX_STORAGE_TEXTURES_PER_STAGE 8
#define MAX_STORAGE_BUFFERS_PER_STAGE 8
#define MAX_UNIFORM_BUFFERS_PER_STAGE 4
#define MAX_COMPUTE_WRITE_TEXTURES 8
#define MAX_COMPUTE_WRITE_BUFFERS 8
#define UNIFORM_BUFFER_SIZE 32768
#define MAX_VERTEX_BUFFERS 16
#define MAX_VERTEX_ATTRIBUTES 16
#define MAX_COLOR_TARGET_BINDINGS 4
#define MAX_PRESENT_COUNT 16
#define MAX_FRAMES_IN_FLIGHT 3
// Internal Macros
#define EXPAND_ARRAY_IF_NEEDED(arr, elementType, newCount, capacity, newCapacity) \
@@ -977,7 +1049,7 @@ extern "C" {
extern SDL_GPUBootstrap VulkanDriver;
extern SDL_GPUBootstrap D3D12Driver;
extern SDL_GPUBootstrap MetalDriver;
extern SDL_GPUBootstrap PS5Driver;
extern SDL_GPUBootstrap PrivateGPUDriver;
#ifdef __cplusplus
}
libs/SDL3/src/gpu/d3d12/SDL_gpu_d3d12.c
+351 -192
View File
File diff suppressed because it is too large Load Diff
libs/SDL3/src/gpu/metal/SDL_gpu_metal.m
+72 -85
View File
@@ -476,35 +476,21 @@ typedef struct MetalShader
typedef struct MetalGraphicsPipeline
{
id<MTLRenderPipelineState> handle;
GraphicsPipelineCommonHeader header;
Uint32 sample_mask;
id<MTLRenderPipelineState> handle;
SDL_GPURasterizerState rasterizerState;
SDL_GPUPrimitiveType primitiveType;
id<MTLDepthStencilState> depth_stencil_state;
Uint32 vertexSamplerCount;
Uint32 vertexUniformBufferCount;
Uint32 vertexStorageBufferCount;
Uint32 vertexStorageTextureCount;
Uint32 fragmentSamplerCount;
Uint32 fragmentUniformBufferCount;
Uint32 fragmentStorageBufferCount;
Uint32 fragmentStorageTextureCount;
} MetalGraphicsPipeline;
typedef struct MetalComputePipeline
{
ComputePipelineCommonHeader header;
id<MTLComputePipelineState> handle;
Uint32 numSamplers;
Uint32 numReadonlyStorageTextures;
Uint32 numReadWriteStorageTextures;
Uint32 numReadonlyStorageBuffers;
Uint32 numReadWriteStorageBuffers;
Uint32 numUniformBuffers;
Uint32 threadcountX;
Uint32 threadcountY;
Uint32 threadcountZ;
@@ -842,6 +828,10 @@ static MetalLibraryFunction METAL_INTERNAL_CompileShader(
dispatch_data_t data;
id<MTLFunction> function;
if (!entrypoint) {
entrypoint = "main0";
}
if (format == SDL_GPU_SHADERFORMAT_MSL) {
NSString *codeString = [[NSString alloc]
initWithBytes:code
@@ -856,7 +846,7 @@ static MetalLibraryFunction METAL_INTERNAL_CompileShader(
code,
codeSize,
dispatch_get_global_queue(0, 0),
^{ /* do nothing */ });
DISPATCH_DATA_DESTRUCTOR_DEFAULT);
library = [renderer->device newLibraryWithData:data error:&error];
} else {
SDL_assert(!"SDL_gpu.c should have already validated this!");
@@ -898,6 +888,7 @@ static void METAL_INTERNAL_DestroyTextureContainer(
container->textures[i]->handle = nil;
SDL_free(container->textures[i]);
}
SDL_DestroyProperties(container->header.info.props);
if (container->debugName != NULL) {
SDL_free(container->debugName);
}
@@ -1057,12 +1048,12 @@ static SDL_GPUComputePipeline *METAL_CreateComputePipeline(
pipeline = SDL_calloc(1, sizeof(MetalComputePipeline));
pipeline->handle = handle;
pipeline->numSamplers = createinfo->num_samplers;
pipeline->numReadonlyStorageTextures = createinfo->num_readonly_storage_textures;
pipeline->numReadWriteStorageTextures = createinfo->num_readwrite_storage_textures;
pipeline->numReadonlyStorageBuffers = createinfo->num_readonly_storage_buffers;
pipeline->numReadWriteStorageBuffers = createinfo->num_readwrite_storage_buffers;
pipeline->numUniformBuffers = createinfo->num_uniform_buffers;
pipeline->header.numSamplers = createinfo->num_samplers;
pipeline->header.numReadonlyStorageTextures = createinfo->num_readonly_storage_textures;
pipeline->header.numReadWriteStorageTextures = createinfo->num_readwrite_storage_textures;
pipeline->header.numReadonlyStorageBuffers = createinfo->num_readonly_storage_buffers;
pipeline->header.numReadWriteStorageBuffers = createinfo->num_readwrite_storage_buffers;
pipeline->header.numUniformBuffers = createinfo->num_uniform_buffers;
pipeline->threadcountX = createinfo->threadcount_x;
pipeline->threadcountY = createinfo->threadcount_y;
pipeline->threadcountZ = createinfo->threadcount_z;
@@ -1181,9 +1172,7 @@ static SDL_GPUGraphicsPipeline *METAL_CreateGraphicsPipeline(
for (Uint32 i = 0; i < createinfo->vertex_input_state.num_vertex_buffers; i += 1) {
binding = METAL_FIRST_VERTEX_BUFFER_SLOT + createinfo->vertex_input_state.vertex_buffer_descriptions[i].slot;
vertexDescriptor.layouts[binding].stepFunction = SDLToMetal_StepFunction[createinfo->vertex_input_state.vertex_buffer_descriptions[i].input_rate];
vertexDescriptor.layouts[binding].stepRate = (createinfo->vertex_input_state.vertex_buffer_descriptions[i].input_rate == SDL_GPU_VERTEXINPUTRATE_INSTANCE)
? createinfo->vertex_input_state.vertex_buffer_descriptions[i].instance_step_rate
: 1;
vertexDescriptor.layouts[binding].stepRate = 1;
vertexDescriptor.layouts[binding].stride = createinfo->vertex_input_state.vertex_buffer_descriptions[i].pitch;
}
@@ -1202,24 +1191,19 @@ static SDL_GPUGraphicsPipeline *METAL_CreateGraphicsPipeline(
SET_ERROR_AND_RETURN("Creating render pipeline failed: %s", [[error description] UTF8String], NULL);
}
Uint32 sampleMask = createinfo->multisample_state.enable_mask ?
createinfo->multisample_state.sample_mask :
0xFFFFFFFF;
result = SDL_calloc(1, sizeof(MetalGraphicsPipeline));
result->handle = pipelineState;
result->sample_mask = sampleMask;
result->depth_stencil_state = depthStencilState;
result->rasterizerState = createinfo->rasterizer_state;
result->primitiveType = createinfo->primitive_type;
result->vertexSamplerCount = vertexShader->numSamplers;
result->vertexUniformBufferCount = vertexShader->numUniformBuffers;
result->vertexStorageBufferCount = vertexShader->numStorageBuffers;
result->vertexStorageTextureCount = vertexShader->numStorageTextures;
result->fragmentSamplerCount = fragmentShader->numSamplers;
result->fragmentUniformBufferCount = fragmentShader->numUniformBuffers;
result->fragmentStorageBufferCount = fragmentShader->numStorageBuffers;
result->fragmentStorageTextureCount = fragmentShader->numStorageTextures;
result->header.num_vertex_samplers = vertexShader->numSamplers;
result->header.num_vertex_uniform_buffers = vertexShader->numUniformBuffers;
result->header.num_vertex_storage_buffers = vertexShader->numStorageBuffers;
result->header.num_vertex_storage_textures = vertexShader->numStorageTextures;
result->header.num_fragment_samplers = fragmentShader->numSamplers;
result->header.num_fragment_uniform_buffers = fragmentShader->numUniformBuffers;
result->header.num_fragment_storage_buffers = fragmentShader->numStorageBuffers;
result->header.num_fragment_storage_textures = fragmentShader->numStorageTextures;
return (SDL_GPUGraphicsPipeline *)result;
}
}
@@ -1503,7 +1487,9 @@ static SDL_GPUTexture *METAL_CreateTexture(
// Copy properties so we don't lose information when the client destroys them
container->header.info = *createinfo;
container->header.info.props = SDL_CreateProperties();
SDL_CopyProperties(createinfo->props, container->header.info.props);
if (createinfo->props) {
SDL_CopyProperties(createinfo->props, container->header.info.props);
}
container->activeTexture = texture;
container->textureCapacity = 1;
@@ -2415,14 +2401,14 @@ static void METAL_BindGraphicsPipeline(
metalCommandBuffer->needFragmentUniformBufferBind[i] = true;
}
for (i = 0; i < pipeline->vertexUniformBufferCount; i += 1) {
for (i = 0; i < pipeline->header.num_vertex_uniform_buffers; i += 1) {
if (metalCommandBuffer->vertexUniformBuffers[i] == NULL) {
metalCommandBuffer->vertexUniformBuffers[i] = METAL_INTERNAL_AcquireUniformBufferFromPool(
metalCommandBuffer);
}
}
for (i = 0; i < pipeline->fragmentUniformBufferCount; i += 1) {
for (i = 0; i < pipeline->header.num_fragment_uniform_buffers; i += 1) {
if (metalCommandBuffer->fragmentUniformBuffers[i] == NULL) {
metalCommandBuffer->fragmentUniformBuffers[i] = METAL_INTERNAL_AcquireUniformBufferFromPool(
metalCommandBuffer);
@@ -2653,11 +2639,11 @@ static void METAL_INTERNAL_BindGraphicsResources(
// Vertex Samplers+Textures
if (commandBuffer->needVertexSamplerBind) {
if (graphicsPipeline->vertexSamplerCount > 0) {
if (graphicsPipeline->header.num_vertex_samplers > 0) {
[commandBuffer->renderEncoder setVertexSamplerStates:commandBuffer->vertexSamplers
withRange:NSMakeRange(0, graphicsPipeline->vertexSamplerCount)];
withRange:NSMakeRange(0, graphicsPipeline->header.num_vertex_samplers)];
[commandBuffer->renderEncoder setVertexTextures:commandBuffer->vertexTextures
withRange:NSMakeRange(0, graphicsPipeline->vertexSamplerCount)];
withRange:NSMakeRange(0, graphicsPipeline->header.num_vertex_samplers)];
}
commandBuffer->needVertexSamplerBind = false;
}
@@ -2665,10 +2651,10 @@ static void METAL_INTERNAL_BindGraphicsResources(
// Vertex Storage Textures
if (commandBuffer->needVertexStorageTextureBind) {
if (graphicsPipeline->vertexStorageTextureCount > 0) {
if (graphicsPipeline->header.num_vertex_storage_textures > 0) {
[commandBuffer->renderEncoder setVertexTextures:commandBuffer->vertexStorageTextures
withRange:NSMakeRange(graphicsPipeline->vertexSamplerCount,
graphicsPipeline->vertexStorageTextureCount)];
withRange:NSMakeRange(graphicsPipeline->header.num_vertex_samplers,
graphicsPipeline->header.num_vertex_storage_textures)];
}
commandBuffer->needVertexStorageTextureBind = false;
}
@@ -2676,20 +2662,20 @@ static void METAL_INTERNAL_BindGraphicsResources(
// Vertex Storage Buffers
if (commandBuffer->needVertexStorageBufferBind) {
if (graphicsPipeline->vertexStorageBufferCount > 0) {
if (graphicsPipeline->header.num_vertex_storage_buffers > 0) {
[commandBuffer->renderEncoder setVertexBuffers:commandBuffer->vertexStorageBuffers
offsets:offsets
withRange:NSMakeRange(graphicsPipeline->vertexUniformBufferCount,
graphicsPipeline->vertexStorageBufferCount)];
withRange:NSMakeRange(graphicsPipeline->header.num_vertex_uniform_buffers,
graphicsPipeline->header.num_vertex_storage_buffers)];
}
commandBuffer->needVertexStorageBufferBind = false;
}
// Vertex Uniform Buffers
for (Uint32 i = 0; i < graphicsPipeline->vertexUniformBufferCount; i += 1) {
for (Uint32 i = 0; i < graphicsPipeline->header.num_vertex_uniform_buffers; i += 1) {
if (commandBuffer->needVertexUniformBufferBind[i]) {
if (graphicsPipeline->vertexUniformBufferCount > i) {
if (graphicsPipeline->header.num_vertex_uniform_buffers > i) {
[commandBuffer->renderEncoder
setVertexBuffer:commandBuffer->vertexUniformBuffers[i]->handle
offset:commandBuffer->vertexUniformBuffers[i]->drawOffset
@@ -2702,11 +2688,11 @@ static void METAL_INTERNAL_BindGraphicsResources(
// Fragment Samplers+Textures
if (commandBuffer->needFragmentSamplerBind) {
if (graphicsPipeline->fragmentSamplerCount > 0) {
if (graphicsPipeline->header.num_fragment_samplers > 0) {
[commandBuffer->renderEncoder setFragmentSamplerStates:commandBuffer->fragmentSamplers
withRange:NSMakeRange(0, graphicsPipeline->fragmentSamplerCount)];
withRange:NSMakeRange(0, graphicsPipeline->header.num_fragment_samplers)];
[commandBuffer->renderEncoder setFragmentTextures:commandBuffer->fragmentTextures
withRange:NSMakeRange(0, graphicsPipeline->fragmentSamplerCount)];
withRange:NSMakeRange(0, graphicsPipeline->header.num_fragment_samplers)];
}
commandBuffer->needFragmentSamplerBind = false;
}
@@ -2714,10 +2700,10 @@ static void METAL_INTERNAL_BindGraphicsResources(
// Fragment Storage Textures
if (commandBuffer->needFragmentStorageTextureBind) {
if (graphicsPipeline->fragmentStorageTextureCount > 0) {
if (graphicsPipeline->header.num_fragment_storage_textures > 0) {
[commandBuffer->renderEncoder setFragmentTextures:commandBuffer->fragmentStorageTextures
withRange:NSMakeRange(graphicsPipeline->fragmentSamplerCount,
graphicsPipeline->fragmentStorageTextureCount)];
withRange:NSMakeRange(graphicsPipeline->header.num_fragment_samplers,
graphicsPipeline->header.num_fragment_storage_textures)];
}
commandBuffer->needFragmentStorageTextureBind = false;
}
@@ -2725,20 +2711,20 @@ static void METAL_INTERNAL_BindGraphicsResources(
// Fragment Storage Buffers
if (commandBuffer->needFragmentStorageBufferBind) {
if (graphicsPipeline->fragmentStorageBufferCount > 0) {
if (graphicsPipeline->header.num_fragment_storage_buffers > 0) {
[commandBuffer->renderEncoder setFragmentBuffers:commandBuffer->fragmentStorageBuffers
offsets:offsets
withRange:NSMakeRange(graphicsPipeline->fragmentUniformBufferCount,
graphicsPipeline->fragmentStorageBufferCount)];
withRange:NSMakeRange(graphicsPipeline->header.num_fragment_uniform_buffers,
graphicsPipeline->header.num_fragment_storage_buffers)];
}
commandBuffer->needFragmentStorageBufferBind = false;
}
// Fragment Uniform Buffers
for (Uint32 i = 0; i < graphicsPipeline->fragmentUniformBufferCount; i += 1) {
for (Uint32 i = 0; i < graphicsPipeline->header.num_fragment_uniform_buffers; i += 1) {
if (commandBuffer->needFragmentUniformBufferBind[i]) {
if (graphicsPipeline->fragmentUniformBufferCount > i) {
if (graphicsPipeline->header.num_fragment_uniform_buffers > i) {
[commandBuffer->renderEncoder
setFragmentBuffer:commandBuffer->fragmentUniformBuffers[i]->handle
offset:commandBuffer->fragmentUniformBuffers[i]->drawOffset
@@ -2757,38 +2743,38 @@ static void METAL_INTERNAL_BindComputeResources(
NSUInteger offsets[MAX_STORAGE_BUFFERS_PER_STAGE] = { 0 };
if (commandBuffer->needComputeSamplerBind) {
if (computePipeline->numSamplers > 0) {
if (computePipeline->header.numSamplers > 0) {
[commandBuffer->computeEncoder setTextures:commandBuffer->computeSamplerTextures
withRange:NSMakeRange(0, computePipeline->numSamplers)];
withRange:NSMakeRange(0, computePipeline->header.numSamplers)];
[commandBuffer->computeEncoder setSamplerStates:commandBuffer->computeSamplers
withRange:NSMakeRange(0, computePipeline->numSamplers)];
withRange:NSMakeRange(0, computePipeline->header.numSamplers)];
}
commandBuffer->needComputeSamplerBind = false;
}
if (commandBuffer->needComputeReadOnlyStorageTextureBind) {
if (computePipeline->numReadonlyStorageTextures > 0) {
if (computePipeline->header.numReadonlyStorageTextures > 0) {
[commandBuffer->computeEncoder setTextures:commandBuffer->computeReadOnlyTextures
withRange:NSMakeRange(
computePipeline->numSamplers,
computePipeline->numReadonlyStorageTextures)];
computePipeline->header.numSamplers,
computePipeline->header.numReadonlyStorageTextures)];
}
commandBuffer->needComputeReadOnlyStorageTextureBind = false;
}
if (commandBuffer->needComputeReadOnlyStorageBufferBind) {
if (computePipeline->numReadonlyStorageBuffers > 0) {
if (computePipeline->header.numReadonlyStorageBuffers > 0) {
[commandBuffer->computeEncoder setBuffers:commandBuffer->computeReadOnlyBuffers
offsets:offsets
withRange:NSMakeRange(computePipeline->numUniformBuffers,
computePipeline->numReadonlyStorageBuffers)];
withRange:NSMakeRange(computePipeline->header.numUniformBuffers,
computePipeline->header.numReadonlyStorageBuffers)];
}
commandBuffer->needComputeReadOnlyStorageBufferBind = false;
}
for (Uint32 i = 0; i < MAX_UNIFORM_BUFFERS_PER_STAGE; i += 1) {
if (commandBuffer->needComputeUniformBufferBind[i]) {
if (computePipeline->numUniformBuffers > i) {
if (computePipeline->header.numUniformBuffers > i) {
[commandBuffer->computeEncoder
setBuffer:commandBuffer->computeUniformBuffers[i]->handle
offset:commandBuffer->computeUniformBuffers[i]->drawOffset
@@ -3136,7 +3122,7 @@ static void METAL_BindComputePipeline(
metalCommandBuffer->needComputeUniformBufferBind[i] = true;
}
for (Uint32 i = 0; i < pipeline->numUniformBuffers; i += 1) {
for (Uint32 i = 0; i < pipeline->header.numUniformBuffers; i += 1) {
if (metalCommandBuffer->computeUniformBuffers[i] == NULL) {
metalCommandBuffer->computeUniformBuffers[i] = METAL_INTERNAL_AcquireUniformBufferFromPool(
metalCommandBuffer);
@@ -3144,22 +3130,22 @@ static void METAL_BindComputePipeline(
}
// Bind write-only resources
if (pipeline->numReadWriteStorageTextures > 0) {
if (pipeline->header.numReadWriteStorageTextures > 0) {
[metalCommandBuffer->computeEncoder setTextures:metalCommandBuffer->computeReadWriteTextures
withRange:NSMakeRange(
pipeline->numSamplers +
pipeline->numReadonlyStorageTextures,
pipeline->numReadWriteStorageTextures)];
pipeline->header.numSamplers +
pipeline->header.numReadonlyStorageTextures,
pipeline->header.numReadWriteStorageTextures)];
}
NSUInteger offsets[MAX_COMPUTE_WRITE_BUFFERS] = { 0 };
if (pipeline->numReadWriteStorageBuffers > 0) {
if (pipeline->header.numReadWriteStorageBuffers > 0) {
[metalCommandBuffer->computeEncoder setBuffers:metalCommandBuffer->computeReadWriteBuffers
offsets:offsets
withRange:NSMakeRange(
pipeline->numUniformBuffers +
pipeline->numReadonlyStorageBuffers,
pipeline->numReadWriteStorageBuffers)];
pipeline->header.numUniformBuffers +
pipeline->header.numReadonlyStorageBuffers,
pipeline->header.numReadWriteStorageBuffers)];
}
}
}
@@ -4573,6 +4559,7 @@ static SDL_GPUDevice *METAL_CreateDevice(bool debugMode, bool preferLowPower, SD
SDL_GPUDevice *result = SDL_calloc(1, sizeof(SDL_GPUDevice));
ASSIGN_DRIVER(METAL)
result->driverData = (SDL_GPURenderer *)renderer;
result->shader_formats = SDL_GPU_SHADERFORMAT_MSL | SDL_GPU_SHADERFORMAT_METALLIB;
renderer->sdlGPUDevice = result;
return result;
libs/SDL3/src/gpu/vulkan/SDL_gpu_vulkan.c
+470 -388
View File
File diff suppressed because it is too large Load Diff