gitlab-com-suyu-emu/suyu
1
0
Fork 0
mirror of https://gitlab.com/suyu-emu/suyu.git synced 2024-03-15 23:15:44 +00:00
Code Issues Projects Releases Wiki Activity

Merge pull request #4364 from lioncash/desig5

Browse source 
vulkan: Make use of designated initializers where applicable
This commit is contained in:
bunnei 2020-07-18 00:12:43 -04:00 committed by GitHub
parent af1151b1b0 7785123b1c
commit 821d295f24
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
19 changed files with 758 additions and 659 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

63
src/video_core/renderer_vulkan/vk_buffer_cache.cpp
View file

@ -39,16 +39,17 @@ std::unique_ptr<VKStreamBuffer> CreateStreamBuffer(const VKDevice& device, VKSch
Buffer::Buffer(const VKDevice& device, VKMemoryManager& memory_manager, VKScheduler& scheduler_,
VKStagingBufferPool& staging_pool_, VAddr cpu_addr, std::size_t size)
: VideoCommon::BufferBlock{cpu_addr, size}, scheduler{scheduler_}, staging_pool{staging_pool_} {
VkBufferCreateInfo ci;
ci.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO;
ci.pNext = nullptr;
ci.flags = 0;
ci.size = static_cast<VkDeviceSize>(size);
ci.usage = BUFFER_USAGE | VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT;
ci.sharingMode = VK_SHARING_MODE_EXCLUSIVE;
ci.queueFamilyIndexCount = 0;
ci.pQueueFamilyIndices = nullptr;
: BufferBlock{cpu_addr, size}, scheduler{scheduler_}, staging_pool{staging_pool_} {
const VkBufferCreateInfo ci{
.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.size = static_cast<VkDeviceSize>(size),
.usage = BUFFER_USAGE | VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT,
.sharingMode = VK_SHARING_MODE_EXCLUSIVE,
.queueFamilyIndexCount = 0,
.pQueueFamilyIndices = nullptr,
};
buffer.handle = device.GetLogical().CreateBuffer(ci);
buffer.commit = memory_manager.Commit(buffer.handle, false);
@ -66,16 +67,17 @@ void Buffer::Upload(std::size_t offset, std::size_t size, const u8* data) {
scheduler.Record([staging = *staging.handle, handle, offset, size](vk::CommandBuffer cmdbuf) {
cmdbuf.CopyBuffer(staging, handle, VkBufferCopy{0, offset, size});
VkBufferMemoryBarrier barrier;
barrier.sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER;
barrier.pNext = nullptr;
barrier.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
barrier.dstAccessMask = UPLOAD_ACCESS_BARRIERS;
barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
barrier.buffer = handle;
barrier.offset = offset;
barrier.size = size;
const VkBufferMemoryBarrier barrier{
.sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER,
.pNext = nullptr,
.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
.dstAccessMask = UPLOAD_ACCESS_BARRIERS,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.buffer = handle,
.offset = offset,
.size = size,
};
cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, UPLOAD_PIPELINE_STAGE, 0, {},
barrier, {});
});
@ -87,16 +89,17 @@ void Buffer::Download(std::size_t offset, std::size_t size, u8* data) {
const VkBuffer handle = Handle();
scheduler.Record([staging = *staging.handle, handle, offset, size](vk::CommandBuffer cmdbuf) {
VkBufferMemoryBarrier barrier;
barrier.sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER;
barrier.pNext = nullptr;
barrier.srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT;
barrier.dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT;
barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
barrier.buffer = handle;
barrier.offset = offset;
barrier.size = size;
const VkBufferMemoryBarrier barrier{
.sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER,
.pNext = nullptr,
.srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT,
.dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.buffer = handle,
.offset = offset,
.size = size,
};
cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_VERTEX_SHADER_BIT |
VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT |

190
src/video_core/renderer_vulkan/vk_compute_pass.cpp
View file

@ -115,32 +115,32 @@ constexpr u8 quad_array[] = {
0xfd, 0x00, 0x01, 0x00, 0x38, 0x00, 0x01, 0x00};
VkDescriptorSetLayoutBinding BuildQuadArrayPassDescriptorSetLayoutBinding() {
VkDescriptorSetLayoutBinding binding;
binding.binding = 0;
binding.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER;
binding.descriptorCount = 1;
binding.stageFlags = VK_SHADER_STAGE_COMPUTE_BIT;
binding.pImmutableSamplers = nullptr;
return binding;
return {
.binding = 0,
.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
.descriptorCount = 1,
.stageFlags = VK_SHADER_STAGE_COMPUTE_BIT,
.pImmutableSamplers = nullptr,
};
}
VkDescriptorUpdateTemplateEntryKHR BuildQuadArrayPassDescriptorUpdateTemplateEntry() {
VkDescriptorUpdateTemplateEntryKHR entry;
entry.dstBinding = 0;
entry.dstArrayElement = 0;
entry.descriptorCount = 1;
entry.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER;
entry.offset = 0;
entry.stride = sizeof(DescriptorUpdateEntry);
return entry;
return {
.dstBinding = 0,
.dstArrayElement = 0,
.descriptorCount = 1,
.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
.offset = 0,
.stride = sizeof(DescriptorUpdateEntry),
};
}
VkPushConstantRange BuildComputePushConstantRange(std::size_t size) {
VkPushConstantRange range;
range.stageFlags = VK_SHADER_STAGE_COMPUTE_BIT;
range.offset = 0;
range.size = static_cast<u32>(size);
return range;
return {
.stageFlags = VK_SHADER_STAGE_COMPUTE_BIT,
.offset = 0,
.size = static_cast<u32>(size),
};
}
// Uint8 SPIR-V module. Generated from the "shaders/" directory.
@ -344,29 +344,33 @@ constexpr u8 QUAD_INDEXED_SPV[] = {
0xfd, 0x00, 0x01, 0x00, 0x38, 0x00, 0x01, 0x00};
std::array<VkDescriptorSetLayoutBinding, 2> BuildInputOutputDescriptorSetBindings() {
std::array<VkDescriptorSetLayoutBinding, 2> bindings;
bindings[0].binding = 0;
bindings[0].descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER;
bindings[0].descriptorCount = 1;
bindings[0].stageFlags = VK_SHADER_STAGE_COMPUTE_BIT;
bindings[0].pImmutableSamplers = nullptr;
bindings[1].binding = 1;
bindings[1].descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER;
bindings[1].descriptorCount = 1;
bindings[1].stageFlags = VK_SHADER_STAGE_COMPUTE_BIT;
bindings[1].pImmutableSamplers = nullptr;
return bindings;
return {{
{
.binding = 0,
.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
.descriptorCount = 1,
.stageFlags = VK_SHADER_STAGE_COMPUTE_BIT,
.pImmutableSamplers = nullptr,
},
{
.binding = 1,
.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
.descriptorCount = 1,
.stageFlags = VK_SHADER_STAGE_COMPUTE_BIT,
.pImmutableSamplers = nullptr,
},
}};
}
VkDescriptorUpdateTemplateEntryKHR BuildInputOutputDescriptorUpdateTemplate() {
VkDescriptorUpdateTemplateEntryKHR entry;
entry.dstBinding = 0;
entry.dstArrayElement = 0;
entry.descriptorCount = 2;
entry.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER;
entry.offset = 0;
entry.stride = sizeof(DescriptorUpdateEntry);
return entry;
return {
.dstBinding = 0,
.dstArrayElement = 0,
.descriptorCount = 2,
.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
.offset = 0,
.stride = sizeof(DescriptorUpdateEntry),
};
}
} // Anonymous namespace
@ -376,37 +380,37 @@ VKComputePass::VKComputePass(const VKDevice& device, VKDescriptorPool& descripto
vk::Span<VkDescriptorUpdateTemplateEntryKHR> templates,
vk::Span<VkPushConstantRange> push_constants, std::size_t code_size,
const u8* code) {
VkDescriptorSetLayoutCreateInfo descriptor_layout_ci;
descriptor_layout_ci.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO;
descriptor_layout_ci.pNext = nullptr;
descriptor_layout_ci.flags = 0;
descriptor_layout_ci.bindingCount = bindings.size();
descriptor_layout_ci.pBindings = bindings.data();
descriptor_set_layout = device.GetLogical().CreateDescriptorSetLayout(descriptor_layout_ci);
descriptor_set_layout = device.GetLogical().CreateDescriptorSetLayout({
.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.bindingCount = bindings.size(),
.pBindings = bindings.data(),
});
VkPipelineLayoutCreateInfo pipeline_layout_ci;
pipeline_layout_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO;
pipeline_layout_ci.pNext = nullptr;
pipeline_layout_ci.flags = 0;
pipeline_layout_ci.setLayoutCount = 1;
pipeline_layout_ci.pSetLayouts = descriptor_set_layout.address();
pipeline_layout_ci.pushConstantRangeCount = push_constants.size();
pipeline_layout_ci.pPushConstantRanges = push_constants.data();
layout = device.GetLogical().CreatePipelineLayout(pipeline_layout_ci);
layout = device.GetLogical().CreatePipelineLayout({
.sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.setLayoutCount = 1,
.pSetLayouts = descriptor_set_layout.address(),
.pushConstantRangeCount = push_constants.size(),
.pPushConstantRanges = push_constants.data(),
});
if (!templates.empty()) {
VkDescriptorUpdateTemplateCreateInfoKHR template_ci;
template_ci.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_UPDATE_TEMPLATE_CREATE_INFO_KHR;
template_ci.pNext = nullptr;
template_ci.flags = 0;
template_ci.descriptorUpdateEntryCount = templates.size();
template_ci.pDescriptorUpdateEntries = templates.data();
template_ci.templateType = VK_DESCRIPTOR_UPDATE_TEMPLATE_TYPE_DESCRIPTOR_SET_KHR;
template_ci.descriptorSetLayout = *descriptor_set_layout;
template_ci.pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS;
template_ci.pipelineLayout = *layout;
template_ci.set = 0;
descriptor_template = device.GetLogical().CreateDescriptorUpdateTemplateKHR(template_ci);
descriptor_template = device.GetLogical().CreateDescriptorUpdateTemplateKHR({
.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_UPDATE_TEMPLATE_CREATE_INFO_KHR,
.pNext = nullptr,
.flags = 0,
.descriptorUpdateEntryCount = templates.size(),
.pDescriptorUpdateEntries = templates.data(),
.templateType = VK_DESCRIPTOR_UPDATE_TEMPLATE_TYPE_DESCRIPTOR_SET_KHR,
.descriptorSetLayout = *descriptor_set_layout,
.pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS,
.pipelineLayout = *layout,
.set = 0,
});
descriptor_allocator.emplace(descriptor_pool, *descriptor_set_layout);
}
@ -414,32 +418,32 @@ VKComputePass::VKComputePass(const VKDevice& device, VKDescriptorPool& descripto
auto code_copy = std::make_unique<u32[]>(code_size / sizeof(u32) + 1);
std::memcpy(code_copy.get(), code, code_size);
VkShaderModuleCreateInfo module_ci;
module_ci.sType = VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO;
module_ci.pNext = nullptr;
module_ci.flags = 0;
module_ci.codeSize = code_size;
module_ci.pCode = code_copy.get();
module = device.GetLogical().CreateShaderModule(module_ci);
module = device.GetLogical().CreateShaderModule({
.sType = VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.codeSize = code_size,
.pCode = code_copy.get(),
});
VkComputePipelineCreateInfo pipeline_ci;
pipeline_ci.sType = VK_STRUCTURE_TYPE_COMPUTE_PIPELINE_CREATE_INFO;
pipeline_ci.pNext = nullptr;
pipeline_ci.flags = 0;
pipeline_ci.layout = *layout;
pipeline_ci.basePipelineHandle = nullptr;
pipeline_ci.basePipelineIndex = 0;
VkPipelineShaderStageCreateInfo& stage_ci = pipeline_ci.stage;
stage_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO;
stage_ci.pNext = nullptr;
stage_ci.flags = 0;
stage_ci.stage = VK_SHADER_STAGE_COMPUTE_BIT;
stage_ci.module = *module;
stage_ci.pName = "main";
stage_ci.pSpecializationInfo = nullptr;
pipeline = device.GetLogical().CreateComputePipeline(pipeline_ci);
pipeline = device.GetLogical().CreateComputePipeline({
.sType = VK_STRUCTURE_TYPE_COMPUTE_PIPELINE_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.stage =
{
.sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.stage = VK_SHADER_STAGE_COMPUTE_BIT,
.module = *module,
.pName = "main",
.pSpecializationInfo = nullptr,
},
.layout = *layout,
.basePipelineHandle = nullptr,
.basePipelineIndex = 0,
});
}
VKComputePass::~VKComputePass() = default;

129
src/video_core/renderer_vulkan/vk_compute_pipeline.cpp
View file

@ -43,12 +43,13 @@ vk::DescriptorSetLayout VKComputePipeline::CreateDescriptorSetLayout() const {
const auto add_bindings = [&](VkDescriptorType descriptor_type, std::size_t num_entries) {
// TODO(Rodrigo): Maybe make individual bindings here?
for (u32 bindpoint = 0; bindpoint < static_cast<u32>(num_entries); ++bindpoint) {
VkDescriptorSetLayoutBinding& entry = bindings.emplace_back();
entry.binding = binding++;
entry.descriptorType = descriptor_type;
entry.descriptorCount = 1;
entry.stageFlags = VK_SHADER_STAGE_COMPUTE_BIT;
entry.pImmutableSamplers = nullptr;
bindings.push_back({
.binding = binding++,
.descriptorType = descriptor_type,
.descriptorCount = 1,
.stageFlags = VK_SHADER_STAGE_COMPUTE_BIT,
.pImmutableSamplers = nullptr,
});
}
};
add_bindings(VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, entries.const_buffers.size());
@ -58,25 +59,25 @@ vk::DescriptorSetLayout VKComputePipeline::CreateDescriptorSetLayout() const {
add_bindings(VK_DESCRIPTOR_TYPE_STORAGE_TEXEL_BUFFER, entries.storage_texels.size());
add_bindings(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, entries.images.size());
VkDescriptorSetLayoutCreateInfo ci;
ci.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO;
ci.pNext = nullptr;
ci.flags = 0;
ci.bindingCount = static_cast<u32>(bindings.size());
ci.pBindings = bindings.data();
return device.GetLogical().CreateDescriptorSetLayout(ci);
return device.GetLogical().CreateDescriptorSetLayout({
.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.bindingCount = static_cast<u32>(bindings.size()),
.pBindings = bindings.data(),
});
}
vk::PipelineLayout VKComputePipeline::CreatePipelineLayout() const {
VkPipelineLayoutCreateInfo ci;
ci.sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO;
ci.pNext = nullptr;
ci.flags = 0;
ci.setLayoutCount = 1;
ci.pSetLayouts = descriptor_set_layout.address();
ci.pushConstantRangeCount = 0;
ci.pPushConstantRanges = nullptr;
return device.GetLogical().CreatePipelineLayout(ci);
return device.GetLogical().CreatePipelineLayout({
.sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.setLayoutCount = 1,
.pSetLayouts = descriptor_set_layout.address(),
.pushConstantRangeCount = 0,
.pPushConstantRanges = nullptr,
});
}
vk::DescriptorUpdateTemplateKHR VKComputePipeline::CreateDescriptorUpdateTemplate() const {
@ -89,59 +90,63 @@ vk::DescriptorUpdateTemplateKHR VKComputePipeline::CreateDescriptorUpdateTemplat
return {};
}
VkDescriptorUpdateTemplateCreateInfoKHR ci;
ci.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_UPDATE_TEMPLATE_CREATE_INFO_KHR;
ci.pNext = nullptr;
ci.flags = 0;
ci.descriptorUpdateEntryCount = static_cast<u32>(template_entries.size());
ci.pDescriptorUpdateEntries = template_entries.data();
ci.templateType = VK_DESCRIPTOR_UPDATE_TEMPLATE_TYPE_DESCRIPTOR_SET_KHR;
ci.descriptorSetLayout = *descriptor_set_layout;
ci.pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS;
ci.pipelineLayout = *layout;
ci.set = DESCRIPTOR_SET;
return device.GetLogical().CreateDescriptorUpdateTemplateKHR(ci);
return device.GetLogical().CreateDescriptorUpdateTemplateKHR({
.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_UPDATE_TEMPLATE_CREATE_INFO_KHR,
.pNext = nullptr,
.flags = 0,
.descriptorUpdateEntryCount = static_cast<u32>(template_entries.size()),
.pDescriptorUpdateEntries = template_entries.data(),
.templateType = VK_DESCRIPTOR_UPDATE_TEMPLATE_TYPE_DESCRIPTOR_SET_KHR,
.descriptorSetLayout = *descriptor_set_layout,
.pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS,
.pipelineLayout = *layout,
.set = DESCRIPTOR_SET,
});
}
vk::ShaderModule VKComputePipeline::CreateShaderModule(const std::vector<u32>& code) const {
device.SaveShader(code);
VkShaderModuleCreateInfo ci;
ci.sType = VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO;
ci.pNext = nullptr;
ci.flags = 0;
ci.codeSize = code.size() * sizeof(u32);
ci.pCode = code.data();
return device.GetLogical().CreateShaderModule(ci);
return device.GetLogical().CreateShaderModule({
.sType = VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.codeSize = code.size() * sizeof(u32),
.pCode = code.data(),
});
}
vk::Pipeline VKComputePipeline::CreatePipeline() const {
VkComputePipelineCreateInfo ci;
VkPipelineShaderStageCreateInfo& stage_ci = ci.stage;
stage_ci.sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO;
stage_ci.pNext = nullptr;
stage_ci.flags = 0;
stage_ci.stage = VK_SHADER_STAGE_COMPUTE_BIT;
stage_ci.module = *shader_module;
stage_ci.pName = "main";
stage_ci.pSpecializationInfo = nullptr;
VkPipelineShaderStageRequiredSubgroupSizeCreateInfoEXT subgroup_size_ci;
subgroup_size_ci.sType =
VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_REQUIRED_SUBGROUP_SIZE_CREATE_INFO_EXT;
subgroup_size_ci.pNext = nullptr;
subgroup_size_ci.requiredSubgroupSize = GuestWarpSize;
VkComputePipelineCreateInfo ci{
.sType = VK_STRUCTURE_TYPE_COMPUTE_PIPELINE_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.stage =
{
.sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.stage = VK_SHADER_STAGE_COMPUTE_BIT,
.module = *shader_module,
.pName = "main",
.pSpecializationInfo = nullptr,
},
.layout = *layout,
.basePipelineHandle = nullptr,
.basePipelineIndex = 0,
};
const VkPipelineShaderStageRequiredSubgroupSizeCreateInfoEXT subgroup_size_ci{
.sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_REQUIRED_SUBGROUP_SIZE_CREATE_INFO_EXT,
.pNext = nullptr,
.requiredSubgroupSize = GuestWarpSize,
};
if (entries.uses_warps && device.IsGuestWarpSizeSupported(VK_SHADER_STAGE_COMPUTE_BIT)) {
stage_ci.pNext = &subgroup_size_ci;
ci.stage.pNext = &subgroup_size_ci;
}
ci.sType = VK_STRUCTURE_TYPE_COMPUTE_PIPELINE_CREATE_INFO;
ci.pNext = nullptr;
ci.flags = 0;
ci.layout = *layout;
ci.basePipelineHandle = nullptr;
ci.basePipelineIndex = 0;
return device.GetLogical().CreateComputePipeline(ci);
}

31
src/video_core/renderer_vulkan/vk_descriptor_pool.cpp
View file

@ -43,27 +43,30 @@ vk::DescriptorPool* VKDescriptorPool::AllocateNewPool() {
{VK_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER, num_sets * 64},
{VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER, num_sets * 64},
{VK_DESCRIPTOR_TYPE_STORAGE_TEXEL_BUFFER, num_sets * 64},
{VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, num_sets * 40}};
{VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, num_sets * 40},
};
VkDescriptorPoolCreateInfo ci;
ci.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_POOL_CREATE_INFO;
ci.pNext = nullptr;
ci.flags = VK_DESCRIPTOR_POOL_CREATE_FREE_DESCRIPTOR_SET_BIT;
ci.maxSets = num_sets;
ci.poolSizeCount = static_cast<u32>(std::size(pool_sizes));
ci.pPoolSizes = std::data(pool_sizes);
const VkDescriptorPoolCreateInfo ci{
.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_POOL_CREATE_INFO,
.pNext = nullptr,
.flags = VK_DESCRIPTOR_POOL_CREATE_FREE_DESCRIPTOR_SET_BIT,
.maxSets = num_sets,
.poolSizeCount = static_cast<u32>(std::size(pool_sizes)),
.pPoolSizes = std::data(pool_sizes),
};
return &pools.emplace_back(device.GetLogical().CreateDescriptorPool(ci));
}
vk::DescriptorSets VKDescriptorPool::AllocateDescriptors(VkDescriptorSetLayout layout,
std::size_t count) {
const std::vector layout_copies(count, layout);
VkDescriptorSetAllocateInfo ai;
ai.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_ALLOCATE_INFO;
ai.pNext = nullptr;
ai.descriptorPool = **active_pool;
ai.descriptorSetCount = static_cast<u32>(count);
ai.pSetLayouts = layout_copies.data();
VkDescriptorSetAllocateInfo ai{
.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_ALLOCATE_INFO,
.pNext = nullptr,
.descriptorPool = **active_pool,
.descriptorSetCount = static_cast<u32>(count),
.pSetLayouts = layout_copies.data(),
};
vk::DescriptorSets sets = active_pool->Allocate(ai);
if (!sets.IsOutOfPoolMemory()) {

38
src/video_core/renderer_vulkan/vk_image.cpp
View file

@ -102,21 +102,29 @@ bool VKImage::HasChanged(u32 base_layer, u32 num_layers, u32 base_level, u32 num
void VKImage::CreatePresentView() {
// Image type has to be 2D to be presented.
VkImageViewCreateInfo image_view_ci;
image_view_ci.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO;
image_view_ci.pNext = nullptr;
image_view_ci.flags = 0;
image_view_ci.image = *image;
image_view_ci.viewType = VK_IMAGE_VIEW_TYPE_2D;
image_view_ci.format = format;
image_view_ci.components = {VK_COMPONENT_SWIZZLE_IDENTITY, VK_COMPONENT_SWIZZLE_IDENTITY,
VK_COMPONENT_SWIZZLE_IDENTITY, VK_COMPONENT_SWIZZLE_IDENTITY};
image_view_ci.subresourceRange.aspectMask = aspect_mask;
image_view_ci.subresourceRange.baseMipLevel = 0;
image_view_ci.subresourceRange.levelCount = 1;
image_view_ci.subresourceRange.baseArrayLayer = 0;
image_view_ci.subresourceRange.layerCount = 1;
present_view = device.GetLogical().CreateImageView(image_view_ci);
present_view = device.GetLogical().CreateImageView({
.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.image = *image,
.viewType = VK_IMAGE_VIEW_TYPE_2D,
.format = format,
.components =
{
.r = VK_COMPONENT_SWIZZLE_IDENTITY,
.g = VK_COMPONENT_SWIZZLE_IDENTITY,
.b = VK_COMPONENT_SWIZZLE_IDENTITY,
.a = VK_COMPONENT_SWIZZLE_IDENTITY,
},
.subresourceRange =
{
.aspectMask = aspect_mask,
.baseMipLevel = 0,
.levelCount = 1,
.baseArrayLayer = 0,
.layerCount = 1,
},
});
}
VKImage::SubrangeState& VKImage::GetSubrangeState(u32 layer, u32 level) noexcept {

13
src/video_core/renderer_vulkan/vk_memory_manager.cpp
View file

@ -178,13 +178,12 @@ bool VKMemoryManager::AllocMemory(VkMemoryPropertyFlags wanted_properties, u32 t
}();
// Try to allocate found type.
VkMemoryAllocateInfo memory_ai;
memory_ai.sType = VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO;
memory_ai.pNext = nullptr;
memory_ai.allocationSize = size;
memory_ai.memoryTypeIndex = type;
vk::DeviceMemory memory = device.GetLogical().TryAllocateMemory(memory_ai);
vk::DeviceMemory memory = device.GetLogical().TryAllocateMemory({
.sType = VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO,
.pNext = nullptr,
.allocationSize = size,
.memoryTypeIndex = type,
});
if (!memory) {
LOG_CRITICAL(Render_Vulkan, "Device allocation failed!");
return false;

66
src/video_core/renderer_vulkan/vk_pipeline_cache.cpp
View file

@ -88,12 +88,13 @@ void AddBindings(std::vector<VkDescriptorSetLayoutBinding>& bindings, u32& bindi
// Combined image samplers can be arrayed.
count = container[i].size;
}
VkDescriptorSetLayoutBinding& entry = bindings.emplace_back();
entry.binding = binding++;
entry.descriptorType = descriptor_type;
entry.descriptorCount = count;
entry.stageFlags = stage_flags;
entry.pImmutableSamplers = nullptr;
bindings.push_back({
.binding = binding++,
.descriptorType = descriptor_type,
.descriptorCount = count,
.stageFlags = stage_flags,
.pImmutableSamplers = nullptr,
});
}
}
@ -259,10 +260,10 @@ VKComputePipeline& VKPipelineCache::GetComputePipeline(const ComputePipelineCach
}
}
Specialization specialization;
specialization.workgroup_size = key.workgroup_size;
specialization.shared_memory_size = key.shared_memory_size;
const Specialization specialization{
.workgroup_size = key.workgroup_size,
.shared_memory_size = key.shared_memory_size,
};
const SPIRVShader spirv_shader{Decompile(device, shader->GetIR(), ShaderType::Compute,
shader->GetRegistry(), specialization),
shader->GetEntries()};
@ -370,13 +371,14 @@ void AddEntry(std::vector<VkDescriptorUpdateTemplateEntry>& template_entries, u3
if constexpr (descriptor_type == COMBINED_IMAGE_SAMPLER) {
for (u32 i = 0; i < count; ++i) {
const u32 num_samplers = container[i].size;
VkDescriptorUpdateTemplateEntry& entry = template_entries.emplace_back();
entry.dstBinding = binding;
entry.dstArrayElement = 0;
entry.descriptorCount = num_samplers;
entry.descriptorType = descriptor_type;
entry.offset = offset;
entry.stride = entry_size;
template_entries.push_back({
.dstBinding = binding,
.dstArrayElement = 0,
.descriptorCount = num_samplers,
.descriptorType = descriptor_type,
.offset = offset,
.stride = entry_size,
});
++binding;
offset += num_samplers * entry_size;
@ -389,22 +391,24 @@ void AddEntry(std::vector<VkDescriptorUpdateTemplateEntry>& template_entries, u3
// Nvidia has a bug where updating multiple texels at once causes the driver to crash.
// Note: Fixed in driver Windows 443.24, Linux 440.66.15
for (u32 i = 0; i < count; ++i) {
VkDescriptorUpdateTemplateEntry& entry = template_entries.emplace_back();
entry.dstBinding = binding + i;
entry.dstArrayElement = 0;
entry.descriptorCount = 1;
entry.descriptorType = descriptor_type;
entry.offset = static_cast<std::size_t>(offset + i * entry_size);
entry.stride = entry_size;
template_entries.push_back({
.dstBinding = binding + i,
.dstArrayElement = 0,
.descriptorCount = 1,
.descriptorType = descriptor_type,
.offset = static_cast<std::size_t>(offset + i * entry_size),
.stride = entry_size,
});
}
} else if (count > 0) {
VkDescriptorUpdateTemplateEntry& entry = template_entries.emplace_back();
entry.dstBinding = binding;
entry.dstArrayElement = 0;
entry.descriptorCount = count;
entry.descriptorType = descriptor_type;
entry.offset = offset;
entry.stride = entry_size;
template_entries.push_back({
.dstBinding = binding,
.dstArrayElement = 0,
.descriptorCount = count,
.descriptorType = descriptor_type,
.offset = offset,
.stride = entry_size,
});
}
offset += count * entry_size;
binding += count;

16
src/video_core/renderer_vulkan/vk_query_cache.cpp
View file

@ -47,14 +47,14 @@ std::pair<VkQueryPool, u32> QueryPool::Commit(VKFence& fence) {
void QueryPool::Allocate(std::size_t begin, std::size_t end) {
usage.resize(end);
VkQueryPoolCreateInfo query_pool_ci;
query_pool_ci.sType = VK_STRUCTURE_TYPE_QUERY_POOL_CREATE_INFO;
query_pool_ci.pNext = nullptr;
query_pool_ci.flags = 0;
query_pool_ci.queryType = GetTarget(type);
query_pool_ci.queryCount = static_cast<u32>(end - begin);
query_pool_ci.pipelineStatistics = 0;
pools.push_back(device->GetLogical().CreateQueryPool(query_pool_ci));
pools.push_back(device->GetLogical().CreateQueryPool({
.sType = VK_STRUCTURE_TYPE_QUERY_POOL_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.queryType = GetTarget(type),
.queryCount = static_cast<u32>(end - begin),
.pipelineStatistics = 0,
}));
}
void QueryPool::Reserve(std::pair<VkQueryPool, u32> query) {

90
src/video_core/renderer_vulkan/vk_rasterizer.cpp
View file

@ -64,20 +64,22 @@ VkViewport GetViewportState(const VKDevice& device, const Maxwell& regs, std::si
const auto& src = regs.viewport_transform[index];
const float width = src.scale_x * 2.0f;
const float height = src.scale_y * 2.0f;
VkViewport viewport;
viewport.x = src.translate_x - src.scale_x;
viewport.y = src.translate_y - src.scale_y;
viewport.width = width != 0.0f ? width : 1.0f;
viewport.height = height != 0.0f ? height : 1.0f;
const float reduce_z = regs.depth_mode == Maxwell::DepthMode::MinusOneToOne ? 1.0f : 0.0f;
viewport.minDepth = src.translate_z - src.scale_z * reduce_z;
viewport.maxDepth = src.translate_z + src.scale_z;
VkViewport viewport{
.x = src.translate_x - src.scale_x,
.y = src.translate_y - src.scale_y,
.width = width != 0.0f ? width : 1.0f,
.height = height != 0.0f ? height : 1.0f,
.minDepth = src.translate_z - src.scale_z * reduce_z,
.maxDepth = src.translate_z + src.scale_z,
};
if (!device.IsExtDepthRangeUnrestrictedSupported()) {
viewport.minDepth = std::clamp(viewport.minDepth, 0.0f, 1.0f);
viewport.maxDepth = std::clamp(viewport.maxDepth, 0.0f, 1.0f);
}
return viewport;
}
@ -508,10 +510,11 @@ void RasterizerVulkan::Clear() {
const u32 color_attachment = regs.clear_buffers.RT;
scheduler.Record([color_attachment, clear_value, clear_rect](vk::CommandBuffer cmdbuf) {
VkClearAttachment attachment;
attachment.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT;
attachment.colorAttachment = color_attachment;
attachment.clearValue = clear_value;
const VkClearAttachment attachment{
.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
.colorAttachment = color_attachment,
.clearValue = clear_value,
};
cmdbuf.ClearAttachments(attachment, clear_rect);
});
}
@ -551,13 +554,16 @@ void RasterizerVulkan::DispatchCompute(GPUVAddr code_addr) {
query_cache.UpdateCounters();
const auto& launch_desc = system.GPU().KeplerCompute().launch_description;
ComputePipelineCacheKey key;
key.shader = code_addr;
key.shared_memory_size = launch_desc.shared_alloc;
key.workgroup_size = {launch_desc.block_dim_x, launch_desc.block_dim_y,
launch_desc.block_dim_z};
auto& pipeline = pipeline_cache.GetComputePipeline(key);
auto& pipeline = pipeline_cache.GetComputePipeline({
.shader = code_addr,
.shared_memory_size = launch_desc.shared_alloc,
.workgroup_size =
{
launch_desc.block_dim_x,
launch_desc.block_dim_y,
launch_desc.block_dim_z,
},
});
// Compute dispatches can't be executed inside a renderpass
scheduler.RequestOutsideRenderPassOperationContext();
@ -841,17 +847,17 @@ std::tuple<VkFramebuffer, VkExtent2D> RasterizerVulkan::ConfigureFramebuffers(
const auto [fbentry, is_cache_miss] = framebuffer_cache.try_emplace(key);
auto& framebuffer = fbentry->second;
if (is_cache_miss) {
VkFramebufferCreateInfo framebuffer_ci;
framebuffer_ci.sType = VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO;
framebuffer_ci.pNext = nullptr;
framebuffer_ci.flags = 0;
framebuffer_ci.renderPass = key.renderpass;
framebuffer_ci.attachmentCount = static_cast<u32>(key.views.size());
framebuffer_ci.pAttachments = key.views.data();
framebuffer_ci.width = key.width;
framebuffer_ci.height = key.height;
framebuffer_ci.layers = key.layers;
framebuffer = device.GetLogical().CreateFramebuffer(framebuffer_ci);
framebuffer = device.GetLogical().CreateFramebuffer({
.sType = VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.renderPass = key.renderpass,
.attachmentCount = static_cast<u32>(key.views.size()),
.pAttachments = key.views.data(),
.width = key.width,
.height = key.height,
.layers = key.layers,
});
}
return {*framebuffer, VkExtent2D{key.width, key.height}};
@ -1553,17 +1559,17 @@ VkBuffer RasterizerVulkan::DefaultBuffer() {
return *default_buffer;
}
VkBufferCreateInfo ci;
ci.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO;
ci.pNext = nullptr;
ci.flags = 0;
ci.size = DEFAULT_BUFFER_SIZE;
ci.usage = VK_BUFFER_USAGE_TRANSFER_DST_BIT | VK_BUFFER_USAGE_VERTEX_BUFFER_BIT |
VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT;
ci.sharingMode = VK_SHARING_MODE_EXCLUSIVE;
ci.queueFamilyIndexCount = 0;
ci.pQueueFamilyIndices = nullptr;
default_buffer = device.GetLogical().CreateBuffer(ci);
default_buffer = device.GetLogical().CreateBuffer({
.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.size = DEFAULT_BUFFER_SIZE,
.usage = VK_BUFFER_USAGE_TRANSFER_DST_BIT | VK_BUFFER_USAGE_VERTEX_BUFFER_BIT |
VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT,
.sharingMode = VK_SHARING_MODE_EXCLUSIVE,
.queueFamilyIndexCount = 0,
.pQueueFamilyIndices = nullptr,
});
default_buffer_commit = memory_manager.Commit(default_buffer, false);
scheduler.RequestOutsideRenderPassOperationContext();

127
src/video_core/renderer_vulkan/vk_renderpass_cache.cpp
View file

@ -39,10 +39,14 @@ VkRenderPass VKRenderPassCache::GetRenderPass(const RenderPassParams& params) {
vk::RenderPass VKRenderPassCache::CreateRenderPass(const RenderPassParams& params) const {
using namespace VideoCore::Surface;
std::vector<VkAttachmentDescription> descriptors;
std::vector<VkAttachmentReference> color_references;
const std::size_t num_attachments = static_cast<std::size_t>(params.num_color_attachments);
std::vector<VkAttachmentDescription> descriptors;
descriptors.reserve(num_attachments);
std::vector<VkAttachmentReference> color_references;
color_references.reserve(num_attachments);
for (std::size_t rt = 0; rt < num_attachments; ++rt) {
const auto guest_format = static_cast<Tegra::RenderTargetFormat>(params.color_formats[rt]);
const PixelFormat pixel_format = PixelFormatFromRenderTargetFormat(guest_format);
@ -54,20 +58,22 @@ vk::RenderPass VKRenderPassCache::CreateRenderPass(const RenderPassParams& param
const VkImageLayout color_layout = ((params.texceptions >> rt) & 1) != 0
? VK_IMAGE_LAYOUT_GENERAL
: VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL;
VkAttachmentDescription& descriptor = descriptors.emplace_back();
descriptor.flags = VK_ATTACHMENT_DESCRIPTION_MAY_ALIAS_BIT;
descriptor.format = format.format;
descriptor.samples = VK_SAMPLE_COUNT_1_BIT;
descriptor.loadOp = VK_ATTACHMENT_LOAD_OP_LOAD;
descriptor.storeOp = VK_ATTACHMENT_STORE_OP_STORE;
descriptor.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
descriptor.stencilStoreOp = VK_ATTACHMENT_STORE_OP_DONT_CARE;
descriptor.initialLayout = color_layout;
descriptor.finalLayout = color_layout;
descriptors.push_back({
.flags = VK_ATTACHMENT_DESCRIPTION_MAY_ALIAS_BIT,
.format = format.format,
.samples = VK_SAMPLE_COUNT_1_BIT,
.loadOp = VK_ATTACHMENT_LOAD_OP_LOAD,
.storeOp = VK_ATTACHMENT_STORE_OP_STORE,
.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE,
.stencilStoreOp = VK_ATTACHMENT_STORE_OP_DONT_CARE,
.initialLayout = color_layout,
.finalLayout = color_layout,
});
VkAttachmentReference& reference = color_references.emplace_back();
reference.attachment = static_cast<u32>(rt);
reference.layout = color_layout;
color_references.push_back({
.attachment = static_cast<u32>(rt),
.layout = color_layout,
});
}
VkAttachmentReference zeta_attachment_ref;
@ -82,32 +88,36 @@ vk::RenderPass VKRenderPassCache::CreateRenderPass(const RenderPassParams& param
const VkImageLayout zeta_layout = params.zeta_texception != 0
? VK_IMAGE_LAYOUT_GENERAL
: VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL;
VkAttachmentDescription& descriptor = descriptors.emplace_back();
descriptor.flags = 0;
descriptor.format = format.format;
descriptor.samples = VK_SAMPLE_COUNT_1_BIT;
descriptor.loadOp = VK_ATTACHMENT_LOAD_OP_LOAD;
descriptor.storeOp = VK_ATTACHMENT_STORE_OP_STORE;
descriptor.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_LOAD;
descriptor.stencilStoreOp = VK_ATTACHMENT_STORE_OP_STORE;
descriptor.initialLayout = zeta_layout;
descriptor.finalLayout = zeta_layout;
descriptors.push_back({
.flags = 0,
.format = format.format,
.samples = VK_SAMPLE_COUNT_1_BIT,
.loadOp = VK_ATTACHMENT_LOAD_OP_LOAD,
.storeOp = VK_ATTACHMENT_STORE_OP_STORE,
.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_LOAD,
.stencilStoreOp = VK_ATTACHMENT_STORE_OP_STORE,
.initialLayout = zeta_layout,
.finalLayout = zeta_layout,
});
zeta_attachment_ref.attachment = static_cast<u32>(num_attachments);
zeta_attachment_ref.layout = zeta_layout;
zeta_attachment_ref = {
.attachment = static_cast<u32>(num_attachments),
.layout = zeta_layout,
};
}
VkSubpassDescription subpass_description;
subpass_description.flags = 0;
subpass_description.pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS;
subpass_description.inputAttachmentCount = 0;
subpass_description.pInputAttachments = nullptr;
subpass_description.colorAttachmentCount = static_cast<u32>(color_references.size());
subpass_description.pColorAttachments = color_references.data();
subpass_description.pResolveAttachments = nullptr;
subpass_description.pDepthStencilAttachment = has_zeta ? &zeta_attachment_ref : nullptr;
subpass_description.preserveAttachmentCount = 0;
subpass_description.pPreserveAttachments = nullptr;
const VkSubpassDescription subpass_description{
.flags = 0,
.pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS,
.inputAttachmentCount = 0,
.pInputAttachments = nullptr,
.colorAttachmentCount = static_cast<u32>(color_references.size()),
.pColorAttachments = color_references.data(),
.pResolveAttachments = nullptr,
.pDepthStencilAttachment = has_zeta ? &zeta_attachment_ref : nullptr,
.preserveAttachmentCount = 0,
.pPreserveAttachments = nullptr,
};
VkAccessFlags access = 0;
VkPipelineStageFlags stage = 0;
@ -122,26 +132,27 @@ vk::RenderPass VKRenderPassCache::CreateRenderPass(const RenderPassParams& param
stage |= VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT;
}
VkSubpassDependency subpass_dependency;
subpass_dependency.srcSubpass = VK_SUBPASS_EXTERNAL;
subpass_dependency.dstSubpass = 0;
subpass_dependency.srcStageMask = stage;
subpass_dependency.dstStageMask = stage;
subpass_dependency.srcAccessMask = 0;
subpass_dependency.dstAccessMask = access;
subpass_dependency.dependencyFlags = 0;
const VkSubpassDependency subpass_dependency{
.srcSubpass = VK_SUBPASS_EXTERNAL,
.dstSubpass = 0,
.srcStageMask = stage,
.dstStageMask = stage,
.srcAccessMask = 0,
.dstAccessMask = access,
.dependencyFlags = 0,
};
VkRenderPassCreateInfo ci;
ci.sType = VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO;
ci.pNext = nullptr;
ci.flags = 0;
ci.attachmentCount = static_cast<u32>(descriptors.size());
ci.pAttachments = descriptors.data();
ci.subpassCount = 1;
ci.pSubpasses = &subpass_description;
ci.dependencyCount = 1;
ci.pDependencies = &subpass_dependency;
return device.GetLogical().CreateRenderPass(ci);
return device.GetLogical().CreateRenderPass({
.sType = VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.attachmentCount = static_cast<u32>(descriptors.size()),
.pAttachments = descriptors.data(),
.subpassCount = 1,
.pSubpasses = &subpass_description,
.dependencyCount = 1,
.pDependencies = &subpass_dependency,
});
}
} // namespace Vulkan

29
src/video_core/renderer_vulkan/vk_resource_manager.cpp
View file

@ -18,33 +18,32 @@ namespace {
constexpr std::size_t COMMAND_BUFFER_POOL_SIZE = 0x1000;
constexpr std::size_t FENCES_GROW_STEP = 0x40;
VkFenceCreateInfo BuildFenceCreateInfo() {
VkFenceCreateInfo fence_ci;
fence_ci.sType = VK_STRUCTURE_TYPE_FENCE_CREATE_INFO;
fence_ci.pNext = nullptr;
fence_ci.flags = 0;
return fence_ci;
constexpr VkFenceCreateInfo BuildFenceCreateInfo() {
return {
.sType = VK_STRUCTURE_TYPE_FENCE_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
};
}
} // Anonymous namespace
class CommandBufferPool final : public VKFencedPool {
public:
CommandBufferPool(const VKDevice& device)
explicit CommandBufferPool(const VKDevice& device)
: VKFencedPool(COMMAND_BUFFER_POOL_SIZE), device{device} {}
void Allocate(std::size_t begin, std::size_t end) override {
// Command buffers are going to be commited, recorded, executed every single usage cycle.
// They are also going to be reseted when commited.
VkCommandPoolCreateInfo command_pool_ci;
command_pool_ci.sType = VK_STRUCTURE_TYPE_COMMAND_POOL_CREATE_INFO;
command_pool_ci.pNext = nullptr;
command_pool_ci.flags =
VK_COMMAND_POOL_CREATE_TRANSIENT_BIT | VK_COMMAND_POOL_CREATE_RESET_COMMAND_BUFFER_BIT;
command_pool_ci.queueFamilyIndex = device.GetGraphicsFamily();
Pool& pool = pools.emplace_back();
pool.handle = device.GetLogical().CreateCommandPool(command_pool_ci);
pool.handle = device.GetLogical().CreateCommandPool({
.sType = VK_STRUCTURE_TYPE_COMMAND_POOL_CREATE_INFO,
.pNext = nullptr,
.flags = VK_COMMAND_POOL_CREATE_TRANSIENT_BIT |
VK_COMMAND_POOL_CREATE_RESET_COMMAND_BUFFER_BIT,
.queueFamilyIndex = device.GetGraphicsFamily(),
});
pool.cmdbufs = pool.handle.Allocate(COMMAND_BUFFER_POOL_SIZE);
}

51
src/video_core/renderer_vulkan/vk_sampler_cache.cpp
View file

@ -44,32 +44,35 @@ vk::Sampler VKSamplerCache::CreateSampler(const Tegra::Texture::TSCEntry& tsc) c
const bool arbitrary_borders = device.IsExtCustomBorderColorSupported();
const std::array color = tsc.GetBorderColor();
VkSamplerCustomBorderColorCreateInfoEXT border;
border.sType = VK_STRUCTURE_TYPE_SAMPLER_CUSTOM_BORDER_COLOR_CREATE_INFO_EXT;
border.pNext = nullptr;
border.format = VK_FORMAT_UNDEFINED;
VkSamplerCustomBorderColorCreateInfoEXT border{
.sType = VK_STRUCTURE_TYPE_SAMPLER_CUSTOM_BORDER_COLOR_CREATE_INFO_EXT,
.pNext = nullptr,
.format = VK_FORMAT_UNDEFINED,
};
std::memcpy(&border.customBorderColor, color.data(), sizeof(color));
VkSamplerCreateInfo ci;
ci.sType = VK_STRUCTURE_TYPE_SAMPLER_CREATE_INFO;
ci.pNext = arbitrary_borders ? &border : nullptr;
ci.flags = 0;
ci.magFilter = MaxwellToVK::Sampler::Filter(tsc.mag_filter);
ci.minFilter = MaxwellToVK::Sampler::Filter(tsc.min_filter);
ci.mipmapMode = MaxwellToVK::Sampler::MipmapMode(tsc.mipmap_filter);
ci.addressModeU = MaxwellToVK::Sampler::WrapMode(device, tsc.wrap_u, tsc.mag_filter);
ci.addressModeV = MaxwellToVK::Sampler::WrapMode(device, tsc.wrap_v, tsc.mag_filter);
ci.addressModeW = MaxwellToVK::Sampler::WrapMode(device, tsc.wrap_p, tsc.mag_filter);
ci.mipLodBias = tsc.GetLodBias();
ci.anisotropyEnable = tsc.GetMaxAnisotropy() > 1.0f ? VK_TRUE : VK_FALSE;
ci.maxAnisotropy = tsc.GetMaxAnisotropy();
ci.compareEnable = tsc.depth_compare_enabled;
ci.compareOp = MaxwellToVK::Sampler::DepthCompareFunction(tsc.depth_compare_func);
ci.minLod = tsc.mipmap_filter == TextureMipmapFilter::None ? 0.0f : tsc.GetMinLod();
ci.maxLod = tsc.mipmap_filter == TextureMipmapFilter::None ? 0.25f : tsc.GetMaxLod();
ci.borderColor = arbitrary_borders ? VK_BORDER_COLOR_INT_CUSTOM_EXT : ConvertBorderColor(color);
ci.unnormalizedCoordinates = VK_FALSE;
return device.GetLogical().CreateSampler(ci);
return device.GetLogical().CreateSampler({
.sType = VK_STRUCTURE_TYPE_SAMPLER_CREATE_INFO,
.pNext = arbitrary_borders ? &border : nullptr,
.flags = 0,
.magFilter = MaxwellToVK::Sampler::Filter(tsc.mag_filter),
.minFilter = MaxwellToVK::Sampler::Filter(tsc.min_filter),
.mipmapMode = MaxwellToVK::Sampler::MipmapMode(tsc.mipmap_filter),
.addressModeU = MaxwellToVK::Sampler::WrapMode(device, tsc.wrap_u, tsc.mag_filter),
.addressModeV = MaxwellToVK::Sampler::WrapMode(device, tsc.wrap_v, tsc.mag_filter),
.addressModeW = MaxwellToVK::Sampler::WrapMode(device, tsc.wrap_p, tsc.mag_filter),
.mipLodBias = tsc.GetLodBias(),
.anisotropyEnable =
static_cast<VkBool32>(tsc.GetMaxAnisotropy() > 1.0f ? VK_TRUE : VK_FALSE),
.maxAnisotropy = tsc.GetMaxAnisotropy(),
.compareEnable = tsc.depth_compare_enabled,
.compareOp = MaxwellToVK::Sampler::DepthCompareFunction(tsc.depth_compare_func),
.minLod = tsc.mipmap_filter == TextureMipmapFilter::None ? 0.0f : tsc.GetMinLod(),
.maxLod = tsc.mipmap_filter == TextureMipmapFilter::None ? 0.25f : tsc.GetMaxLod(),
.borderColor =
arbitrary_borders ? VK_BORDER_COLOR_INT_CUSTOM_EXT : ConvertBorderColor(color),
.unnormalizedCoordinates = VK_FALSE,
});
}
VkSampler VKSamplerCache::ToSamplerType(const vk::Sampler& sampler) const {

57
src/video_core/renderer_vulkan/vk_scheduler.cpp
View file

@ -100,16 +100,19 @@ void VKScheduler::RequestRenderpass(VkRenderPass renderpass, VkFramebuffer frame
state.framebuffer = framebuffer;
state.render_area = render_area;
VkRenderPassBeginInfo renderpass_bi;
renderpass_bi.sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO;
renderpass_bi.pNext = nullptr;
renderpass_bi.renderPass = renderpass;
renderpass_bi.framebuffer = framebuffer;
renderpass_bi.renderArea.offset.x = 0;
renderpass_bi.renderArea.offset.y = 0;
renderpass_bi.renderArea.extent = render_area;
renderpass_bi.clearValueCount = 0;
renderpass_bi.pClearValues = nullptr;
const VkRenderPassBeginInfo renderpass_bi{
.sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO,
.pNext = nullptr,
.renderPass = renderpass,
.framebuffer = framebuffer,
.renderArea =
{
.offset = {.x = 0, .y = 0},
.extent = render_area,
},
.clearValueCount = 0,
.pClearValues = nullptr,
};
Record([renderpass_bi, end_renderpass](vk::CommandBuffer cmdbuf) {
if (end_renderpass) {
@ -157,16 +160,17 @@ void VKScheduler::SubmitExecution(VkSemaphore semaphore) {
current_cmdbuf.End();
VkSubmitInfo submit_info;
submit_info.sType = VK_STRUCTURE_TYPE_SUBMIT_INFO;
submit_info.pNext = nullptr;
submit_info.waitSemaphoreCount = 0;
submit_info.pWaitSemaphores = nullptr;
submit_info.pWaitDstStageMask = nullptr;
submit_info.commandBufferCount = 1;
submit_info.pCommandBuffers = current_cmdbuf.address();
submit_info.signalSemaphoreCount = semaphore ? 1 : 0;
submit_info.pSignalSemaphores = &semaphore;
const VkSubmitInfo submit_info{
.sType = VK_STRUCTURE_TYPE_SUBMIT_INFO,
.pNext = nullptr,
.waitSemaphoreCount = 0,
.pWaitSemaphores = nullptr,
.pWaitDstStageMask = nullptr,
.commandBufferCount = 1,
.pCommandBuffers = current_cmdbuf.address(),
.signalSemaphoreCount = semaphore ? 1U : 0U,
.pSignalSemaphores = &semaphore,
};
switch (const VkResult result = device.GetGraphicsQueue().Submit(submit_info, *current_fence)) {
case VK_SUCCESS:
break;
@ -181,19 +185,18 @@ void VKScheduler::SubmitExecution(VkSemaphore semaphore) {
void VKScheduler::AllocateNewContext() {
++ticks;
VkCommandBufferBeginInfo cmdbuf_bi;
cmdbuf_bi.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO;
cmdbuf_bi.pNext = nullptr;
cmdbuf_bi.flags = VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT;
cmdbuf_bi.pInheritanceInfo = nullptr;
std::unique_lock lock{mutex};
current_fence = next_fence;
next_fence = &resource_manager.CommitFence();
current_cmdbuf = vk::CommandBuffer(resource_manager.CommitCommandBuffer(*current_fence),
device.GetDispatchLoader());
current_cmdbuf.Begin(cmdbuf_bi);
current_cmdbuf.Begin({
.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO,
.pNext = nullptr,
.flags = VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT,
.pInheritanceInfo = nullptr,
});
// Enable counters once again. These are disabled when a command buffer is finished.
if (query_cache) {

14
src/video_core/renderer_vulkan/vk_shader_util.cpp
View file

@ -19,13 +19,13 @@ vk::ShaderModule BuildShader(const VKDevice& device, std::size_t code_size, cons
const auto data = std::make_unique<u32[]>(code_size / sizeof(u32));
std::memcpy(data.get(), code_data, code_size);
VkShaderModuleCreateInfo ci;
ci.sType = VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO;
ci.pNext = nullptr;
ci.flags = 0;
ci.codeSize = code_size;
ci.pCode = data.get();
return device.GetLogical().CreateShaderModule(ci);
return device.GetLogical().CreateShaderModule({
.sType = VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.codeSize = code_size,
.pCode = data.get(),
});
}
} // namespace Vulkan

25
src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp
View file

@ -71,20 +71,19 @@ VKBuffer* VKStagingBufferPool::TryGetReservedBuffer(std::size_t size, bool host_
VKBuffer& VKStagingBufferPool::CreateStagingBuffer(std::size_t size, bool host_visible) {
const u32 log2 = Common::Log2Ceil64(size);
VkBufferCreateInfo ci;
ci.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO;
ci.pNext = nullptr;
ci.flags = 0;
ci.size = 1ULL << log2;
ci.usage = VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT |
VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT | VK_BUFFER_USAGE_STORAGE_BUFFER_BIT |
VK_BUFFER_USAGE_INDEX_BUFFER_BIT | VK_BUFFER_USAGE_VERTEX_BUFFER_BIT;
ci.sharingMode = VK_SHARING_MODE_EXCLUSIVE;
ci.queueFamilyIndexCount = 0;
ci.pQueueFamilyIndices = nullptr;
auto buffer = std::make_unique<VKBuffer>();
buffer->handle = device.GetLogical().CreateBuffer(ci);
buffer->handle = device.GetLogical().CreateBuffer({
.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.size = 1ULL << log2,
.usage = VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT |
VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT | VK_BUFFER_USAGE_STORAGE_BUFFER_BIT |
VK_BUFFER_USAGE_INDEX_BUFFER_BIT | VK_BUFFER_USAGE_VERTEX_BUFFER_BIT,
.sharingMode = VK_SHARING_MODE_EXCLUSIVE,
.queueFamilyIndexCount = 0,
.pQueueFamilyIndices = nullptr,
});
buffer->commit = memory_manager.Commit(buffer->handle, host_visible);
auto& entries = GetCache(host_visible)[log2].entries;

35
src/video_core/renderer_vulkan/vk_stream_buffer.cpp
View file

@ -122,30 +122,27 @@ void VKStreamBuffer::CreateBuffers(VkBufferUsageFlags usage) {
// Substract from the preferred heap size some bytes to avoid getting out of memory.
const VkDeviceSize heap_size = memory_properties.memoryHeaps[preferred_heap].size;
const VkDeviceSize allocable_size = heap_size - 9 * 1024 * 1024;
VkBufferCreateInfo buffer_ci;
buffer_ci.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO;
buffer_ci.pNext = nullptr;
buffer_ci.flags = 0;
buffer_ci.size = std::min(PREFERRED_STREAM_BUFFER_SIZE, allocable_size);
buffer_ci.usage = usage;
buffer_ci.sharingMode = VK_SHARING_MODE_EXCLUSIVE;
buffer_ci.queueFamilyIndexCount = 0;
buffer_ci.pQueueFamilyIndices = nullptr;
buffer = device.GetLogical().CreateBuffer(buffer_ci);
buffer = device.GetLogical().CreateBuffer({
.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.size = std::min(PREFERRED_STREAM_BUFFER_SIZE, allocable_size),
.usage = usage,
.sharingMode = VK_SHARING_MODE_EXCLUSIVE,
.queueFamilyIndexCount = 0,
.pQueueFamilyIndices = nullptr,
});
const auto requirements = device.GetLogical().GetBufferMemoryRequirements(*buffer);
const u32 required_flags = requirements.memoryTypeBits;
stream_buffer_size = static_cast<u64>(requirements.size);
VkMemoryAllocateInfo memory_ai;
memory_ai.sType = VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO;
memory_ai.pNext = nullptr;
memory_ai.allocationSize = requirements.size;
memory_ai.memoryTypeIndex = GetMemoryType(memory_properties, required_flags);
memory = device.GetLogical().AllocateMemory(memory_ai);
memory = device.GetLogical().AllocateMemory({
.sType = VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO,
.pNext = nullptr,
.allocationSize = requirements.size,
.memoryTypeIndex = GetMemoryType(memory_properties, required_flags),
});
buffer.BindMemory(*memory, 0);
}

94
src/video_core/renderer_vulkan/vk_swapchain.cpp
View file

@ -95,15 +95,16 @@ bool VKSwapchain::Present(VkSemaphore render_semaphore, VKFence& fence) {
const auto present_queue{device.GetPresentQueue()};
bool recreated = false;
VkPresentInfoKHR present_info;
present_info.sType = VK_STRUCTURE_TYPE_PRESENT_INFO_KHR;
present_info.pNext = nullptr;
present_info.waitSemaphoreCount = render_semaphore ? 2U : 1U;
present_info.pWaitSemaphores = semaphores.data();
present_info.swapchainCount = 1;
present_info.pSwapchains = swapchain.address();
present_info.pImageIndices = &image_index;
present_info.pResults = nullptr;
const VkPresentInfoKHR present_info{
.sType = VK_STRUCTURE_TYPE_PRESENT_INFO_KHR,
.pNext = nullptr,
.waitSemaphoreCount = render_semaphore ? 2U : 1U,
.pWaitSemaphores = semaphores.data(),
.swapchainCount = 1,
.pSwapchains = swapchain.address(),
.pImageIndices = &image_index,
.pResults = nullptr,
};
switch (const VkResult result = present_queue.Present(present_info)) {
case VK_SUCCESS:
@ -147,24 +148,25 @@ void VKSwapchain::CreateSwapchain(const VkSurfaceCapabilitiesKHR& capabilities,
requested_image_count = capabilities.maxImageCount;
}
VkSwapchainCreateInfoKHR swapchain_ci;
swapchain_ci.sType = VK_STRUCTURE_TYPE_SWAPCHAIN_CREATE_INFO_KHR;
swapchain_ci.pNext = nullptr;
swapchain_ci.flags = 0;
swapchain_ci.surface = surface;
swapchain_ci.minImageCount = requested_image_count;
swapchain_ci.imageFormat = surface_format.format;
swapchain_ci.imageColorSpace = surface_format.colorSpace;
swapchain_ci.imageArrayLayers = 1;
swapchain_ci.imageUsage = VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT;
swapchain_ci.imageSharingMode = VK_SHARING_MODE_EXCLUSIVE;
swapchain_ci.queueFamilyIndexCount = 0;
swapchain_ci.pQueueFamilyIndices = nullptr;
swapchain_ci.preTransform = capabilities.currentTransform;
swapchain_ci.compositeAlpha = VK_COMPOSITE_ALPHA_OPAQUE_BIT_KHR;
swapchain_ci.presentMode = present_mode;
swapchain_ci.clipped = VK_FALSE;
swapchain_ci.oldSwapchain = nullptr;
VkSwapchainCreateInfoKHR swapchain_ci{
.sType = VK_STRUCTURE_TYPE_SWAPCHAIN_CREATE_INFO_KHR,
.pNext = nullptr,
.flags = 0,
.surface = surface,
.minImageCount = requested_image_count,
.imageFormat = surface_format.format,
.imageColorSpace = surface_format.colorSpace,
.imageArrayLayers = 1,
.imageUsage = VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT,
.imageSharingMode = VK_SHARING_MODE_EXCLUSIVE,
.queueFamilyIndexCount = 0,
.pQueueFamilyIndices = nullptr,
.preTransform = capabilities.currentTransform,
.compositeAlpha = VK_COMPOSITE_ALPHA_OPAQUE_BIT_KHR,
.presentMode = present_mode,
.clipped = VK_FALSE,
.oldSwapchain = nullptr,
};
const u32 graphics_family{device.GetGraphicsFamily()};
const u32 present_family{device.GetPresentFamily()};
@ -173,8 +175,6 @@ void VKSwapchain::CreateSwapchain(const VkSurfaceCapabilitiesKHR& capabilities,
swapchain_ci.imageSharingMode = VK_SHARING_MODE_CONCURRENT;
swapchain_ci.queueFamilyIndexCount = static_cast<u32>(queue_indices.size());
swapchain_ci.pQueueFamilyIndices = queue_indices.data();
} else {
swapchain_ci.imageSharingMode = VK_SHARING_MODE_EXCLUSIVE;
}
// Request the size again to reduce the possibility of a TOCTOU race condition.
@ -200,20 +200,28 @@ void VKSwapchain::CreateSemaphores() {
}
void VKSwapchain::CreateImageViews() {
VkImageViewCreateInfo ci;
ci.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO;
ci.pNext = nullptr;
ci.flags = 0;
// ci.image
ci.viewType = VK_IMAGE_VIEW_TYPE_2D;
ci.format = image_format;
ci.components = {VK_COMPONENT_SWIZZLE_IDENTITY, VK_COMPONENT_SWIZZLE_IDENTITY,
VK_COMPONENT_SWIZZLE_IDENTITY, VK_COMPONENT_SWIZZLE_IDENTITY};
ci.subresourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT;
ci.subresourceRange.baseMipLevel = 0;
ci.subresourceRange.levelCount = 1;
ci.subresourceRange.baseArrayLayer = 0;
ci.subresourceRange.layerCount = 1;
VkImageViewCreateInfo ci{
.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.viewType = VK_IMAGE_VIEW_TYPE_2D,
.format = image_format,
.components =
{
.r = VK_COMPONENT_SWIZZLE_IDENTITY,
.g = VK_COMPONENT_SWIZZLE_IDENTITY,
.b = VK_COMPONENT_SWIZZLE_IDENTITY,
.a = VK_COMPONENT_SWIZZLE_IDENTITY,
},
.subresourceRange =
{
.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
.baseMipLevel = 0,
.levelCount = 1,
.baseArrayLayer = 0,
.layerCount = 1,
},
};
image_views.resize(image_count);
for (std::size_t i = 0; i < image_count; i++) {

231
src/video_core/renderer_vulkan/vk_texture_cache.cpp
View file

@ -95,17 +95,18 @@ VkImageViewType GetImageViewType(SurfaceTarget target) {
vk::Buffer CreateBuffer(const VKDevice& device, const SurfaceParams& params,
std::size_t host_memory_size) {
// TODO(Rodrigo): Move texture buffer creation to the buffer cache
VkBufferCreateInfo ci;
ci.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO;
ci.pNext = nullptr;
ci.flags = 0;
ci.size = static_cast<VkDeviceSize>(host_memory_size);
ci.usage = VK_BUFFER_USAGE_UNIFORM_TEXEL_BUFFER_BIT | VK_BUFFER_USAGE_STORAGE_TEXEL_BUFFER_BIT |
VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT;
ci.sharingMode = VK_SHARING_MODE_EXCLUSIVE;
ci.queueFamilyIndexCount = 0;
ci.pQueueFamilyIndices = nullptr;
return device.GetLogical().CreateBuffer(ci);
return device.GetLogical().CreateBuffer({
.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.size = static_cast<VkDeviceSize>(host_memory_size),
.usage = VK_BUFFER_USAGE_UNIFORM_TEXEL_BUFFER_BIT |
VK_BUFFER_USAGE_STORAGE_TEXEL_BUFFER_BIT | VK_BUFFER_USAGE_TRANSFER_SRC_BIT |
VK_BUFFER_USAGE_TRANSFER_DST_BIT,
.sharingMode = VK_SHARING_MODE_EXCLUSIVE,
.queueFamilyIndexCount = 0,
.pQueueFamilyIndices = nullptr,
});
}
VkBufferViewCreateInfo GenerateBufferViewCreateInfo(const VKDevice& device,
@ -113,15 +114,16 @@ VkBufferViewCreateInfo GenerateBufferViewCreateInfo(const VKDevice& device,
std::size_t host_memory_size) {
ASSERT(params.IsBuffer());
VkBufferViewCreateInfo ci;
ci.sType = VK_STRUCTURE_TYPE_BUFFER_VIEW_CREATE_INFO;
ci.pNext = nullptr;
ci.flags = 0;
ci.buffer = buffer;
ci.format = MaxwellToVK::SurfaceFormat(device, FormatType::Buffer, params.pixel_format).format;
ci.offset = 0;
ci.range = static_cast<VkDeviceSize>(host_memory_size);
return ci;
return {
.sType = VK_STRUCTURE_TYPE_BUFFER_VIEW_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.buffer = buffer,
.format =
MaxwellToVK::SurfaceFormat(device, FormatType::Buffer, params.pixel_format).format,
.offset = 0,
.range = static_cast<VkDeviceSize>(host_memory_size),
};
}
VkImageCreateInfo GenerateImageCreateInfo(const VKDevice& device, const SurfaceParams& params) {
@ -130,23 +132,23 @@ VkImageCreateInfo GenerateImageCreateInfo(const VKDevice& device, const SurfaceP
const auto [format, attachable, storage] =
MaxwellToVK::SurfaceFormat(device, FormatType::Optimal, params.pixel_format);
VkImageCreateInfo ci;
ci.sType = VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO;
ci.pNext = nullptr;
ci.flags = 0;
ci.imageType = SurfaceTargetToImage(params.target);
ci.format = format;
ci.mipLevels = params.num_levels;
ci.arrayLayers = static_cast<u32>(params.GetNumLayers());
ci.samples = VK_SAMPLE_COUNT_1_BIT;
ci.tiling = VK_IMAGE_TILING_OPTIMAL;
ci.sharingMode = VK_SHARING_MODE_EXCLUSIVE;
ci.queueFamilyIndexCount = 0;
ci.pQueueFamilyIndices = nullptr;
ci.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED;
ci.usage = VK_IMAGE_USAGE_SAMPLED_BIT | VK_IMAGE_USAGE_TRANSFER_DST_BIT |
VK_IMAGE_USAGE_TRANSFER_SRC_BIT;
VkImageCreateInfo ci{
.sType = VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.imageType = SurfaceTargetToImage(params.target),
.format = format,
.mipLevels = params.num_levels,
.arrayLayers = static_cast<u32>(params.GetNumLayers()),
.samples = VK_SAMPLE_COUNT_1_BIT,
.tiling = VK_IMAGE_TILING_OPTIMAL,
.usage = VK_IMAGE_USAGE_SAMPLED_BIT | VK_IMAGE_USAGE_TRANSFER_DST_BIT |
VK_IMAGE_USAGE_TRANSFER_SRC_BIT,
.sharingMode = VK_SHARING_MODE_EXCLUSIVE,
.queueFamilyIndexCount = 0,
.pQueueFamilyIndices = nullptr,
.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED,
};
if (attachable) {
ci.usage |= params.IsPixelFormatZeta() ? VK_IMAGE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT
: VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT;
@ -321,22 +323,25 @@ void CachedSurface::UploadImage(const std::vector<u8>& staging_buffer) {
}
VkBufferImageCopy CachedSurface::GetBufferImageCopy(u32 level) const {
VkBufferImageCopy copy;
copy.bufferOffset = params.GetHostMipmapLevelOffset(level, is_converted);
copy.bufferRowLength = 0;
copy.bufferImageHeight = 0;
copy.imageSubresource.aspectMask = image->GetAspectMask();
copy.imageSubresource.mipLevel = level;
copy.imageSubresource.baseArrayLayer = 0;
copy.imageSubresource.layerCount = static_cast<u32>(params.GetNumLayers());
copy.imageOffset.x = 0;
copy.imageOffset.y = 0;
copy.imageOffset.z = 0;
copy.imageExtent.width = params.GetMipWidth(level);
copy.imageExtent.height = params.GetMipHeight(level);
copy.imageExtent.depth =
params.target == SurfaceTarget::Texture3D ? params.GetMipDepth(level) : 1;
return copy;
return {
.bufferOffset = params.GetHostMipmapLevelOffset(level, is_converted),
.bufferRowLength = 0,
.bufferImageHeight = 0,
.imageSubresource =
{
.aspectMask = image->GetAspectMask(),
.mipLevel = level,
.baseArrayLayer = 0,
.layerCount = static_cast<u32>(params.GetNumLayers()),
},
.imageOffset = {.x = 0, .y = 0, .z = 0},
.imageExtent =
{
.width = params.GetMipWidth(level),
.height = params.GetMipHeight(level),
.depth = params.target == SurfaceTarget::Texture3D ? params.GetMipDepth(level) : 1U,
},
};
}
VkImageSubresourceRange CachedSurface::GetImageSubresourceRange() const {
@ -416,20 +421,29 @@ VkImageView CachedSurfaceView::GetImageView(SwizzleSource x_source, SwizzleSourc
ASSERT(num_slices == params.depth);
}
VkImageViewCreateInfo ci;
ci.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO;
ci.pNext = nullptr;
ci.flags = 0;
ci.image = surface.GetImageHandle();
ci.viewType = image_view_type;
ci.format = surface.GetImage().GetFormat();
ci.components = {swizzle[0], swizzle[1], swizzle[2], swizzle[3]};
ci.subresourceRange.aspectMask = aspect;
ci.subresourceRange.baseMipLevel = base_level;
ci.subresourceRange.levelCount = num_levels;
ci.subresourceRange.baseArrayLayer = base_layer;
ci.subresourceRange.layerCount = num_layers;
image_view = device.GetLogical().CreateImageView(ci);
image_view = device.GetLogical().CreateImageView({
.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.image = surface.GetImageHandle(),
.viewType = image_view_type,
.format = surface.GetImage().GetFormat(),
.components =
{
.r = swizzle[0],
.g = swizzle[1],
.b = swizzle[2],
.a = swizzle[3],
},
.subresourceRange =
{
.aspectMask = aspect,
.baseMipLevel = base_level,
.levelCount = num_levels,
.baseArrayLayer = base_layer,
.layerCount = num_layers,
},
});
return last_image_view = *image_view;
}
@ -439,17 +453,26 @@ VkImageView CachedSurfaceView::GetAttachment() {
return *render_target;
}
VkImageViewCreateInfo ci;
ci.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO;
ci.pNext = nullptr;
ci.flags = 0;
ci.image = surface.GetImageHandle();
ci.format = surface.GetImage().GetFormat();
ci.components = {VK_COMPONENT_SWIZZLE_IDENTITY, VK_COMPONENT_SWIZZLE_IDENTITY,
VK_COMPONENT_SWIZZLE_IDENTITY, VK_COMPONENT_SWIZZLE_IDENTITY};
ci.subresourceRange.aspectMask = aspect_mask;
ci.subresourceRange.baseMipLevel = base_level;
ci.subresourceRange.levelCount = num_levels;
VkImageViewCreateInfo ci{
.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.image = surface.GetImageHandle(),
.format = surface.GetImage().GetFormat(),
.components =
{
.r = VK_COMPONENT_SWIZZLE_IDENTITY,
.g = VK_COMPONENT_SWIZZLE_IDENTITY,
.b = VK_COMPONENT_SWIZZLE_IDENTITY,
.a = VK_COMPONENT_SWIZZLE_IDENTITY,
},
.subresourceRange =
{
.aspectMask = aspect_mask,
.baseMipLevel = base_level,
.levelCount = num_levels,
},
};
if (image_view_type == VK_IMAGE_VIEW_TYPE_3D) {
ci.viewType = num_slices > 1 ? VK_IMAGE_VIEW_TYPE_2D_ARRAY : VK_IMAGE_VIEW_TYPE_2D;
ci.subresourceRange.baseArrayLayer = base_slice;
@ -502,24 +525,40 @@ void VKTextureCache::ImageCopy(Surface& src_surface, Surface& dst_surface,
VK_PIPELINE_STAGE_TRANSFER_BIT, VK_ACCESS_TRANSFER_WRITE_BIT,
VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL);
VkImageCopy copy;
copy.srcSubresource.aspectMask = src_surface->GetAspectMask();
copy.srcSubresource.mipLevel = copy_params.source_level;
copy.srcSubresource.baseArrayLayer = copy_params.source_z;
copy.srcSubresource.layerCount = num_layers;
copy.srcOffset.x = copy_params.source_x;
copy.srcOffset.y = copy_params.source_y;
copy.srcOffset.z = 0;
copy.dstSubresource.aspectMask = dst_surface->GetAspectMask();
copy.dstSubresource.mipLevel = copy_params.dest_level;
copy.dstSubresource.baseArrayLayer = dst_base_layer;
copy.dstSubresource.layerCount = num_layers;
copy.dstOffset.x = copy_params.dest_x;
copy.dstOffset.y = copy_params.dest_y;
copy.dstOffset.z = dst_offset_z;
copy.extent.width = copy_params.width;
copy.extent.height = copy_params.height;
copy.extent.depth = extent_z;
const VkImageCopy copy{
.srcSubresource =
{
.aspectMask = src_surface->GetAspectMask(),
.mipLevel = copy_params.source_level,
.baseArrayLayer = copy_params.source_z,
.layerCount = num_layers,
},
.srcOffset =
{
.x = static_cast<s32>(copy_params.source_x),
.y = static_cast<s32>(copy_params.source_y),
.z = 0,
},
.dstSubresource =
{
.aspectMask = dst_surface->GetAspectMask(),
.mipLevel = copy_params.dest_level,
.baseArrayLayer = dst_base_layer,
.layerCount = num_layers,
},
.dstOffset =
{
.x = static_cast<s32>(copy_params.dest_x),
.y = static_cast<s32>(copy_params.dest_y),
.z = static_cast<s32>(dst_offset_z),
},
.extent =
{
.width = copy_params.width,
.height = copy_params.height,
.depth = extent_z,
},
};
const VkImage src_image = src_surface->GetImageHandle();
const VkImage dst_image = dst_surface->GetImageHandle();

118
src/video_core/renderer_vulkan/wrapper.cpp
View file

@ -377,24 +377,26 @@ VkResult Free(VkDevice device, VkCommandPool handle, Span<VkCommandBuffer> buffe
Instance Instance::Create(Span<const char*> layers, Span<const char*> extensions,
InstanceDispatch& dld) noexcept {
VkApplicationInfo application_info;
application_info.sType = VK_STRUCTURE_TYPE_APPLICATION_INFO;
application_info.pNext = nullptr;
application_info.pApplicationName = "yuzu Emulator";
application_info.applicationVersion = VK_MAKE_VERSION(0, 1, 0);
application_info.pEngineName = "yuzu Emulator";
application_info.engineVersion = VK_MAKE_VERSION(0, 1, 0);
application_info.apiVersion = VK_API_VERSION_1_1;
static constexpr VkApplicationInfo application_info{
.sType = VK_STRUCTURE_TYPE_APPLICATION_INFO,
.pNext = nullptr,
.pApplicationName = "yuzu Emulator",
.applicationVersion = VK_MAKE_VERSION(0, 1, 0),
.pEngineName = "yuzu Emulator",
.engineVersion = VK_MAKE_VERSION(0, 1, 0),
.apiVersion = VK_API_VERSION_1_1,
};
VkInstanceCreateInfo ci;
ci.sType = VK_STRUCTURE_TYPE_INSTANCE_CREATE_INFO;
ci.pNext = nullptr;
ci.flags = 0;
ci.pApplicationInfo = &application_info;
ci.enabledLayerCount = layers.size();
ci.ppEnabledLayerNames = layers.data();
ci.enabledExtensionCount = extensions.size();
ci.ppEnabledExtensionNames = extensions.data();
const VkInstanceCreateInfo ci{
.sType = VK_STRUCTURE_TYPE_INSTANCE_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.pApplicationInfo = &application_info,
.enabledLayerCount = layers.size(),
.ppEnabledLayerNames = layers.data(),
.enabledExtensionCount = extensions.size(),
.ppEnabledExtensionNames = extensions.data(),
};
VkInstance instance;
if (dld.vkCreateInstance(&ci, nullptr, &instance) != VK_SUCCESS) {
@ -425,19 +427,20 @@ std::optional<std::vector<VkPhysicalDevice>> Instance::EnumeratePhysicalDevices(
DebugCallback Instance::TryCreateDebugCallback(
PFN_vkDebugUtilsMessengerCallbackEXT callback) noexcept {
VkDebugUtilsMessengerCreateInfoEXT ci;
ci.sType = VK_STRUCTURE_TYPE_DEBUG_UTILS_MESSENGER_CREATE_INFO_EXT;
ci.pNext = nullptr;
ci.flags = 0;
ci.messageSeverity = VK_DEBUG_UTILS_MESSAGE_SEVERITY_ERROR_BIT_EXT |
VK_DEBUG_UTILS_MESSAGE_SEVERITY_WARNING_BIT_EXT |
VK_DEBUG_UTILS_MESSAGE_SEVERITY_INFO_BIT_EXT |
VK_DEBUG_UTILS_MESSAGE_SEVERITY_VERBOSE_BIT_EXT;
ci.messageType = VK_DEBUG_UTILS_MESSAGE_TYPE_GENERAL_BIT_EXT |
VK_DEBUG_UTILS_MESSAGE_TYPE_VALIDATION_BIT_EXT |
VK_DEBUG_UTILS_MESSAGE_TYPE_PERFORMANCE_BIT_EXT;
ci.pfnUserCallback = callback;
ci.pUserData = nullptr;
const VkDebugUtilsMessengerCreateInfoEXT ci{
.sType = VK_STRUCTURE_TYPE_DEBUG_UTILS_MESSENGER_CREATE_INFO_EXT,
.pNext = nullptr,
.flags = 0,
.messageSeverity = VK_DEBUG_UTILS_MESSAGE_SEVERITY_ERROR_BIT_EXT |
VK_DEBUG_UTILS_MESSAGE_SEVERITY_WARNING_BIT_EXT |
VK_DEBUG_UTILS_MESSAGE_SEVERITY_INFO_BIT_EXT |
VK_DEBUG_UTILS_MESSAGE_SEVERITY_VERBOSE_BIT_EXT,
.messageType = VK_DEBUG_UTILS_MESSAGE_TYPE_GENERAL_BIT_EXT |
VK_DEBUG_UTILS_MESSAGE_TYPE_VALIDATION_BIT_EXT |
VK_DEBUG_UTILS_MESSAGE_TYPE_PERFORMANCE_BIT_EXT,
.pfnUserCallback = callback,
.pUserData = nullptr,
};
VkDebugUtilsMessengerEXT messenger;
if (dld->vkCreateDebugUtilsMessengerEXT(handle, &ci, nullptr, &messenger) != VK_SUCCESS) {
@ -468,12 +471,13 @@ DescriptorSets DescriptorPool::Allocate(const VkDescriptorSetAllocateInfo& ai) c
}
CommandBuffers CommandPool::Allocate(std::size_t num_buffers, VkCommandBufferLevel level) const {
VkCommandBufferAllocateInfo ai;
ai.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_ALLOCATE_INFO;
ai.pNext = nullptr;
ai.commandPool = handle;
ai.level = level;
ai.commandBufferCount = static_cast<u32>(num_buffers);
const VkCommandBufferAllocateInfo ai{
.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_ALLOCATE_INFO,
.pNext = nullptr,
.commandPool = handle,
.level = level,
.commandBufferCount = static_cast<u32>(num_buffers),
};
std::unique_ptr buffers = std::make_unique<VkCommandBuffer[]>(num_buffers);
switch (const VkResult result = dld->vkAllocateCommandBuffers(owner, &ai, buffers.get())) {
@ -497,17 +501,18 @@ std::vector<VkImage> SwapchainKHR::GetImages() const {
Device Device::Create(VkPhysicalDevice physical_device, Span<VkDeviceQueueCreateInfo> queues_ci,
Span<const char*> enabled_extensions, const void* next,
DeviceDispatch& dld) noexcept {
VkDeviceCreateInfo ci;
ci.sType = VK_STRUCTURE_TYPE_DEVICE_CREATE_INFO;
ci.pNext = next;
ci.flags = 0;
ci.queueCreateInfoCount = queues_ci.size();
ci.pQueueCreateInfos = queues_ci.data();
ci.enabledLayerCount = 0;
ci.ppEnabledLayerNames = nullptr;
ci.enabledExtensionCount = enabled_extensions.size();
ci.ppEnabledExtensionNames = enabled_extensions.data();
ci.pEnabledFeatures = nullptr;
const VkDeviceCreateInfo ci{
.sType = VK_STRUCTURE_TYPE_DEVICE_CREATE_INFO,
.pNext = next,
.flags = 0,
.queueCreateInfoCount = queues_ci.size(),
.pQueueCreateInfos = queues_ci.data(),
.enabledLayerCount = 0,
.ppEnabledLayerNames = nullptr,
.enabledExtensionCount = enabled_extensions.size(),
.ppEnabledExtensionNames = enabled_extensions.data(),
.pEnabledFeatures = nullptr,
};
VkDevice device;
if (dld.vkCreateDevice(physical_device, &ci, nullptr, &device) != VK_SUCCESS) {
@ -548,10 +553,11 @@ ImageView Device::CreateImageView(const VkImageViewCreateInfo& ci) const {
}
Semaphore Device::CreateSemaphore() const {
VkSemaphoreCreateInfo ci;
ci.sType = VK_STRUCTURE_TYPE_SEMAPHORE_CREATE_INFO;
ci.pNext = nullptr;
ci.flags = 0;
static constexpr VkSemaphoreCreateInfo ci{
.sType = VK_STRUCTURE_TYPE_SEMAPHORE_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
};
VkSemaphore object;
Check(dld->vkCreateSemaphore(handle, &ci, nullptr, &object));
@ -639,10 +645,12 @@ ShaderModule Device::CreateShaderModule(const VkShaderModuleCreateInfo& ci) cons
}
Event Device::CreateEvent() const {
VkEventCreateInfo ci;
ci.sType = VK_STRUCTURE_TYPE_EVENT_CREATE_INFO;
ci.pNext = nullptr;
ci.flags = 0;
static constexpr VkEventCreateInfo ci{
.sType = VK_STRUCTURE_TYPE_EVENT_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
};
VkEvent object;
Check(dld->vkCreateEvent(handle, &ci, nullptr, &object));
return Event(object, handle, *dld);