void D3D12GSRender::upload_and_bind_scale_offset_matrix(size_t descriptorIndex)
{
size_t heap_offset = m_buffer_data.alloc<D3D12_CONSTANT_BUFFER_DATA_PLACEMENT_ALIGNMENT>(256);
// Scale offset buffer
// Separate constant buffer
void *mapped_buffer = m_buffer_data.map<void>(CD3DX12_RANGE(heap_offset, heap_offset + 256));
fill_scale_offset_data(mapped_buffer);
int is_alpha_tested = rsx::method_registers.alpha_test_enabled();
u8 alpha_ref_raw = rsx::method_registers.alpha_ref();
float alpha_ref = alpha_ref_raw / 255.f;
memcpy((char*)mapped_buffer + 16 * sizeof(float), &is_alpha_tested, sizeof(int));
memcpy((char*)mapped_buffer + 17 * sizeof(float), &alpha_ref, sizeof(float));
f32 fogp0 = rsx::method_registers.fog_params_0();
f32 fogp1 = rsx::method_registers.fog_params_1();
memcpy((char*)mapped_buffer + 18 * sizeof(float), &fogp0, sizeof(float));
memcpy((char*)mapped_buffer + 19 * sizeof(float), &fogp1, sizeof(float));
m_buffer_data.unmap(CD3DX12_RANGE(heap_offset, heap_offset + 256));
D3D12_CONSTANT_BUFFER_VIEW_DESC constant_buffer_view_desc = {
m_buffer_data.get_heap()->GetGPUVirtualAddress() + heap_offset,
256
};
m_device->CreateConstantBufferView(&constant_buffer_view_desc,
CD3DX12_CPU_DESCRIPTOR_HANDLE(get_current_resource_storage().descriptors_heap->GetCPUDescriptorHandleForHeapStart())
.Offset((INT)descriptorIndex, m_descriptor_stride_srv_cbv_uav));
}
std::tuple<D3D12_INDEX_BUFFER_VIEW, size_t> D3D12GSRender::generate_index_buffer_for_emulated_primitives_array(const std::vector<std::pair<u32, u32> > &vertex_ranges)
{
size_t index_count = 0;
for (const auto &pair : vertex_ranges)
index_count += get_index_count(draw_mode, pair.second);
// Alloc
size_t buffer_size = align(index_count * sizeof(u16), 64);
size_t heap_offset = m_buffer_data.alloc<D3D12_CONSTANT_BUFFER_DATA_PLACEMENT_ALIGNMENT>(buffer_size);
void *mapped_buffer = m_buffer_data.map<void>(CD3DX12_RANGE(heap_offset, heap_offset + buffer_size));
size_t first = 0;
for (const auto &pair : vertex_ranges)
{
size_t element_count = get_index_count(draw_mode, pair.second);
write_index_array_for_non_indexed_non_native_primitive_to_buffer((char*)mapped_buffer, draw_mode, (u32)first, (u32)pair.second);
mapped_buffer = (char*)mapped_buffer + element_count * sizeof(u16);
first += pair.second;
}
m_buffer_data.unmap(CD3DX12_RANGE(heap_offset, heap_offset + buffer_size));
D3D12_INDEX_BUFFER_VIEW index_buffer_view = {
m_buffer_data.get_heap()->GetGPUVirtualAddress() + heap_offset,
(UINT)buffer_size,
DXGI_FORMAT_R16_UINT
};
return std::make_tuple(index_buffer_view, index_count);
}
uint64 KGraphicsDevice::QueryTimeStamp(uint32 timestamp_query)
{
m_TimeStampQueryReadBackRes->Map(0, &CD3DX12_RANGE(0, 0), reinterpret_cast<void**>(&m_TimeStampQueryMem));
uint64 data = ((uint64*)m_TimeStampQueryMem)[timestamp_query];
m_TimeStampQueryReadBackRes->Unmap(0, nullptr);
return data;
}
void D3D12GSRender::upload_and_bind_vertex_shader_constants(size_t descriptor_index)
{
size_t buffer_size = 512 * 4 * sizeof(float);
size_t heap_offset = m_buffer_data.alloc<D3D12_CONSTANT_BUFFER_DATA_PLACEMENT_ALIGNMENT>(buffer_size);
void *mapped_buffer = m_buffer_data.map<void>(CD3DX12_RANGE(heap_offset, heap_offset + buffer_size));
fill_vertex_program_constants_data(mapped_buffer);
m_buffer_data.unmap(CD3DX12_RANGE(heap_offset, heap_offset + buffer_size));
D3D12_CONSTANT_BUFFER_VIEW_DESC constant_buffer_view_desc = {
m_buffer_data.get_heap()->GetGPUVirtualAddress() + heap_offset,
(UINT)buffer_size
};
m_device->CreateConstantBufferView(&constant_buffer_view_desc,
CD3DX12_CPU_DESCRIPTOR_HANDLE(get_current_resource_storage().descriptors_heap->GetCPUDescriptorHandleForHeapStart())
.Offset((INT)descriptor_index, m_descriptor_stride_srv_cbv_uav));
}
D3D12_CONSTANT_BUFFER_VIEW_DESC D3D12GSRender::upload_fragment_shader_constants()
{
// Get constant from fragment program
size_t buffer_size = m_pso_cache.get_fragment_constants_buffer_size(m_fragment_program);
// Multiple of 256 never 0
buffer_size = (buffer_size + 255) & ~255;
size_t heap_offset = m_buffer_data.alloc<D3D12_CONSTANT_BUFFER_DATA_PLACEMENT_ALIGNMENT>(buffer_size);
size_t offset = 0;
float *mapped_buffer = m_buffer_data.map<float>(CD3DX12_RANGE(heap_offset, heap_offset + buffer_size));
m_pso_cache.fill_fragment_constants_buffer({ mapped_buffer, ::narrow<int>(buffer_size) }, m_fragment_program);
m_buffer_data.unmap(CD3DX12_RANGE(heap_offset, heap_offset + buffer_size));
return {
m_buffer_data.get_heap()->GetGPUVirtualAddress() + heap_offset,
(UINT)buffer_size
};
}
void D3D12GSRender::upload_and_bind_fragment_shader_constants(size_t descriptor_index)
{
// Get constant from fragment program
size_t buffer_size = m_pso_cache.get_fragment_constants_buffer_size(m_fragment_program);
// Multiple of 256 never 0
buffer_size = (buffer_size + 255) & ~255;
size_t heap_offset = m_buffer_data.alloc<D3D12_CONSTANT_BUFFER_DATA_PLACEMENT_ALIGNMENT>(buffer_size);
size_t offset = 0;
float *mapped_buffer = m_buffer_data.map<float>(CD3DX12_RANGE(heap_offset, heap_offset + buffer_size));
m_pso_cache.fill_fragment_constans_buffer({ mapped_buffer, gsl::narrow<int>(buffer_size) }, m_fragment_program);
m_buffer_data.unmap(CD3DX12_RANGE(heap_offset, heap_offset + buffer_size));
D3D12_CONSTANT_BUFFER_VIEW_DESC constant_buffer_view_desc = {
m_buffer_data.get_heap()->GetGPUVirtualAddress() + heap_offset,
(UINT)buffer_size
};
m_device->CreateConstantBufferView(&constant_buffer_view_desc,
CD3DX12_CPU_DESCRIPTOR_HANDLE(get_current_resource_storage().descriptors_heap->GetCPUDescriptorHandleForHeapStart())
.Offset((INT)descriptor_index, m_descriptor_stride_srv_cbv_uav));
}
// Recalculate vertex positions for the color space triangles and then update the vertex buffer.
// The scene's command list must be in the recording state when this method is called.
void D3D12HDR::UpdateVertexBuffer()
{
const XMFLOAT2 primaries709[] =
{
{ 0.64f, 0.33f },
{ 0.30f, 0.60f },
{ 0.15f, 0.06f },
{ 0.3127f, 0.3290f }
};
const XMFLOAT2 primaries2020[] =
{
{ 0.708f, 0.292f },
{ 0.170f, 0.797f },
{ 0.131f, 0.046f },
{ 0.3127f, 0.3290f }
};
const XMFLOAT2 offset1 = { 0.2f, 0.0f };
const XMFLOAT2 offset2 = { 0.2f, -1.0f };
const XMFLOAT3 triangle709[] =
{
TransformVertex(primaries709[0], offset1), // R
TransformVertex(primaries709[1], offset1), // G
TransformVertex(primaries709[2], offset1), // B
TransformVertex(primaries709[3], offset1) // White
};
const XMFLOAT3 triangle2020[] =
{
TransformVertex(primaries2020[0], offset2), // R
TransformVertex(primaries2020[1], offset2), // G
TransformVertex(primaries2020[2], offset2), // B
TransformVertex(primaries2020[3], offset2) // White
};
TrianglesVertex triangleVertices[] =
{
// Upper triangles. Rec 709 primaries.
{ triangle709[2], primaries709[2] },
{ triangle709[1], primaries709[1] },
{ triangle709[3], primaries709[3] },
{ triangle709[1], primaries709[1] },
{ triangle709[0], primaries709[0] },
{ triangle709[3], primaries709[3] },
{ triangle709[0], primaries709[0] },
{ triangle709[2], primaries709[2] },
{ triangle709[3], primaries709[3] },
// Lower triangles. Rec 2020 primaries.
{ triangle2020[2], primaries2020[2] },
{ triangle2020[1], primaries2020[1] },
{ triangle2020[3], primaries2020[3] },
{ triangle2020[1], primaries2020[1] },
{ triangle2020[0], primaries2020[0] },
{ triangle2020[3], primaries2020[3] },
{ triangle2020[0], primaries2020[0] },
{ triangle2020[2], primaries2020[2] },
{ triangle2020[3], primaries2020[3] },
};
// Copy data to the intermediate upload heap and then schedule a copy
// from the upload heap to the vertex buffer.
UINT8* mappedUploadHeap = nullptr;
ThrowIfFailed(m_vertexBufferUpload->Map(0, &CD3DX12_RANGE(0, 0), reinterpret_cast<void**>(&mappedUploadHeap)));
mappedUploadHeap += m_gradientVertexBufferView.SizeInBytes;
memcpy(mappedUploadHeap, triangleVertices, sizeof(triangleVertices));
m_commandList->ResourceBarrier(1, &CD3DX12_RESOURCE_BARRIER::Transition(m_vertexBuffer.Get(), D3D12_RESOURCE_STATE_VERTEX_AND_CONSTANT_BUFFER, D3D12_RESOURCE_STATE_COPY_DEST));
m_commandList->CopyBufferRegion(m_vertexBuffer.Get(), 0, m_vertexBufferUpload.Get(), 0, m_vertexBuffer->GetDesc().Width);
m_commandList->ResourceBarrier(1, &CD3DX12_RESOURCE_BARRIER::Transition(m_vertexBuffer.Get(), D3D12_RESOURCE_STATE_COPY_DEST, D3D12_RESOURCE_STATE_VERTEX_AND_CONSTANT_BUFFER));
}
// Load the sample assets.
void D3D12HDR::LoadAssets()
{
// Create a root signature containing root constants for brightness information
// and the desired output curve as well as a SRV descriptor table pointing to the
// intermediate render targets.
{
CD3DX12_DESCRIPTOR_RANGE ranges[1];
ranges[0].Init(D3D12_DESCRIPTOR_RANGE_TYPE_SRV, 2, 0);
CD3DX12_ROOT_PARAMETER rootParameters[2];
rootParameters[0].InitAsConstants(4, 0);
rootParameters[1].InitAsDescriptorTable(1, &ranges[0]);
D3D12_STATIC_SAMPLER_DESC sampler = {};
sampler.Filter = D3D12_FILTER_MIN_MAG_MIP_POINT;
sampler.AddressU = D3D12_TEXTURE_ADDRESS_MODE_BORDER;
sampler.AddressV = D3D12_TEXTURE_ADDRESS_MODE_BORDER;
sampler.AddressW = D3D12_TEXTURE_ADDRESS_MODE_BORDER;
sampler.ComparisonFunc = D3D12_COMPARISON_FUNC_NEVER;
sampler.MaxLOD = D3D12_FLOAT32_MAX;
sampler.ShaderVisibility = D3D12_SHADER_VISIBILITY_ALL;
// Allow input layout and deny uneccessary access to certain pipeline stages.
D3D12_ROOT_SIGNATURE_FLAGS rootSignatureFlags =
D3D12_ROOT_SIGNATURE_FLAG_ALLOW_INPUT_ASSEMBLER_INPUT_LAYOUT |
D3D12_ROOT_SIGNATURE_FLAG_DENY_HULL_SHADER_ROOT_ACCESS |
D3D12_ROOT_SIGNATURE_FLAG_DENY_DOMAIN_SHADER_ROOT_ACCESS |
D3D12_ROOT_SIGNATURE_FLAG_DENY_GEOMETRY_SHADER_ROOT_ACCESS;
CD3DX12_ROOT_SIGNATURE_DESC rootSignatureDesc;
rootSignatureDesc.Init(_countof(rootParameters), rootParameters, 1, &sampler, rootSignatureFlags);
ComPtr<ID3DBlob> signature;
ComPtr<ID3DBlob> error;
ThrowIfFailed(D3D12SerializeRootSignature(&rootSignatureDesc, D3D_ROOT_SIGNATURE_VERSION_1, &signature, &error));
ThrowIfFailed(m_device->CreateRootSignature(0, signature->GetBufferPointer(), signature->GetBufferSize(), IID_PPV_ARGS(&m_rootSignature)));
}
// Create the pipeline state objects for the different views and render target formats
// as well as the intermediate blend step.
{
// Create the pipeline state for the scene geometry.
D3D12_INPUT_ELEMENT_DESC gradientElementDescs[] =
{
{ "POSITION", 0, DXGI_FORMAT_R32G32B32_FLOAT, 0, 0, D3D12_INPUT_CLASSIFICATION_PER_VERTEX_DATA, 0 },
{ "COLOR", 0, DXGI_FORMAT_R32G32B32_FLOAT, 0, D3D12_APPEND_ALIGNED_ELEMENT, D3D12_INPUT_CLASSIFICATION_PER_VERTEX_DATA, 0 },
};
// Describe and create the graphics pipeline state objects (PSO).
D3D12_GRAPHICS_PIPELINE_STATE_DESC psoDesc = {};
psoDesc.InputLayout = { gradientElementDescs, _countof(gradientElementDescs) };
psoDesc.pRootSignature = m_rootSignature.Get();
psoDesc.VS = CD3DX12_SHADER_BYTECODE(g_gradientVS, sizeof(g_gradientVS));
psoDesc.PS = CD3DX12_SHADER_BYTECODE(g_gradientPS, sizeof(g_gradientPS));
psoDesc.RasterizerState = CD3DX12_RASTERIZER_DESC(D3D12_DEFAULT);
psoDesc.BlendState = CD3DX12_BLEND_DESC(D3D12_DEFAULT);
psoDesc.DepthStencilState.DepthEnable = FALSE;
psoDesc.DepthStencilState.StencilEnable = FALSE;
psoDesc.SampleMask = UINT_MAX;
psoDesc.PrimitiveTopologyType = D3D12_PRIMITIVE_TOPOLOGY_TYPE_TRIANGLE;
psoDesc.NumRenderTargets = 1;
psoDesc.RTVFormats[0] = m_intermediateRenderTargetFormat;
psoDesc.SampleDesc.Count = 1;
ThrowIfFailed(m_device->CreateGraphicsPipelineState(&psoDesc, IID_PPV_ARGS(&m_pipelineStates[GradientPSO])));
// Create pipeline state for the color space triangles.
D3D12_INPUT_ELEMENT_DESC colorElementDescs[] =
{
{ "POSITION", 0, DXGI_FORMAT_R32G32B32_FLOAT, 0, 0, D3D12_INPUT_CLASSIFICATION_PER_VERTEX_DATA, 0 },
{ "TEXCOORD", 0, DXGI_FORMAT_R32G32_FLOAT, 0, D3D12_APPEND_ALIGNED_ELEMENT, D3D12_INPUT_CLASSIFICATION_PER_VERTEX_DATA, 0 },
};
psoDesc.InputLayout = { colorElementDescs, _countof(colorElementDescs) };
psoDesc.VS = CD3DX12_SHADER_BYTECODE(g_paletteVS, sizeof(g_paletteVS));
psoDesc.PS = CD3DX12_SHADER_BYTECODE(g_palettePS, sizeof(g_palettePS));
psoDesc.RTVFormats[0] = m_intermediateRenderTargetFormat;
ThrowIfFailed(m_device->CreateGraphicsPipelineState(&psoDesc, IID_PPV_ARGS(&m_pipelineStates[PalettePSO])));
// Create pipeline states for the final blend step.
// There will be one for each swap chain format the sample supports.
D3D12_INPUT_ELEMENT_DESC quadElementDescs[] =
{
{ "POSITION", 0, DXGI_FORMAT_R32G32B32_FLOAT, 0, 0, D3D12_INPUT_CLASSIFICATION_PER_VERTEX_DATA, 0 },
{ "TEXCOORD", 0, DXGI_FORMAT_R32G32_FLOAT, 0, D3D12_APPEND_ALIGNED_ELEMENT, D3D12_INPUT_CLASSIFICATION_PER_VERTEX_DATA, 0 },
};
psoDesc.InputLayout = { quadElementDescs, _countof(quadElementDescs) };
psoDesc.VS = CD3DX12_SHADER_BYTECODE(g_presentVS, sizeof(g_presentVS));
psoDesc.PS = CD3DX12_SHADER_BYTECODE(g_presentPS, sizeof(g_presentPS));
psoDesc.RTVFormats[0] = m_swapChainFormats[_8];
ThrowIfFailed(m_device->CreateGraphicsPipelineState(&psoDesc, IID_PPV_ARGS(&m_pipelineStates[Present8bitPSO])));
psoDesc.RTVFormats[0] = m_swapChainFormats[_10];
ThrowIfFailed(m_device->CreateGraphicsPipelineState(&psoDesc, IID_PPV_ARGS(&m_pipelineStates[Present10bitPSO])));
psoDesc.RTVFormats[0] = m_swapChainFormats[_16];
ThrowIfFailed(m_device->CreateGraphicsPipelineState(&psoDesc, IID_PPV_ARGS(&m_pipelineStates[Present16bitPSO])));
}
// Create the command list.
ThrowIfFailed(m_device->CreateCommandList(0, D3D12_COMMAND_LIST_TYPE_DIRECT, m_commandAllocators[m_frameIndex].Get(), nullptr, IID_PPV_ARGS(&m_commandList)));
// Create the vertex buffer.
{
// Create geometry for the different sections of the render target.
GradientVertex gradientVertices[] =
{
// Upper strip. SDR Gradient from [0,1].
{ { -1.0f, 0.45f, 0.0f }, { 0.0f, 0.0f, 0.0f } },
{ { -1.0f, 0.55f, 0.0f }, { 0.0f, 0.0f, 0.0f } },
{ { 0.0f, 0.45f, 0.0f }, { 1.0f, 1.0f, 1.0f } },
{ { 0.0f, 0.55f, 0.0f }, { 1.0f, 1.0f, 1.0f } },
// Lower strip. HDR Gradient from [0,9]. Perceptually, 9.0 is about 3 times as bright as 1.0. (See gradientPS.hlsl.)
{ { -1.0f, -0.55f, 0.0f }, { 0.0f, 0.0f, 0.0f } },
{ { -1.0f, -0.45f, 0.0f }, { 0.0f, 0.0f, 0.0f } },
{ { 0.0f, -0.55f, 0.0f }, { 3.0f, 3.0f, 3.0f } },
{ { 0.0f, -0.45f, 0.0f }, { 3.0f, 3.0f, 3.0f } },
};
// The vertices for the color space triangles are dependent on the size of the
// render target and will not be loaded at this time. We'll leave a gap in the
// buffer that will be filled in later.
const UINT triangleVerticesSize = sizeof(TrianglesVertex) * TrianglesVertexCount;
m_updateVertexBuffer = true;
PresentVertex presentVertices[] =
{
// 1 triangle that fills the entire render target.
{ { -1.0f, -3.0f, 0.0f }, { 0.0f, 2.0f } }, // Bottom left
{ { -1.0f, 1.0f, 0.0f }, { 0.0f, 0.0f } }, // Top left
{ { 3.0f, 1.0f, 0.0f }, { 2.0f, 0.0f } }, // Top right
};
const UINT vertexBufferSize = sizeof(gradientVertices) + triangleVerticesSize + sizeof(presentVertices);
ThrowIfFailed(m_device->CreateCommittedResource(
&CD3DX12_HEAP_PROPERTIES(D3D12_HEAP_TYPE_DEFAULT),
D3D12_HEAP_FLAG_NONE,
&CD3DX12_RESOURCE_DESC::Buffer(vertexBufferSize),
D3D12_RESOURCE_STATE_VERTEX_AND_CONSTANT_BUFFER,
nullptr,
IID_PPV_ARGS(&m_vertexBuffer)));
ThrowIfFailed(m_device->CreateCommittedResource(
&CD3DX12_HEAP_PROPERTIES(D3D12_HEAP_TYPE_UPLOAD),
D3D12_HEAP_FLAG_NONE,
&CD3DX12_RESOURCE_DESC::Buffer(vertexBufferSize),
D3D12_RESOURCE_STATE_GENERIC_READ,
nullptr,
IID_PPV_ARGS(&m_vertexBufferUpload)));
// Copy data to the intermediate upload heap. It will be uploaded to the
// DEFAULT buffer when the color space triangle vertices are updated.
UINT8* mappedUploadHeap = nullptr;
ThrowIfFailed(m_vertexBufferUpload->Map(0, &CD3DX12_RANGE(0, 0), reinterpret_cast<void**>(&mappedUploadHeap)));
memcpy(mappedUploadHeap, gradientVertices, sizeof(gradientVertices));
mappedUploadHeap += sizeof(gradientVertices) + triangleVerticesSize;
memcpy(mappedUploadHeap, presentVertices, sizeof(presentVertices));
m_vertexBufferUpload->Unmap(0, &CD3DX12_RANGE(0, 0));
// Initialize the vertex buffer views.
m_gradientVertexBufferView.BufferLocation = m_vertexBuffer->GetGPUVirtualAddress();
m_gradientVertexBufferView.StrideInBytes = sizeof(GradientVertex);
m_gradientVertexBufferView.SizeInBytes = sizeof(gradientVertices);
m_trianglesVertexBufferView.BufferLocation = m_gradientVertexBufferView.BufferLocation + m_gradientVertexBufferView.SizeInBytes;
m_trianglesVertexBufferView.StrideInBytes = sizeof(TrianglesVertex);
m_trianglesVertexBufferView.SizeInBytes = triangleVerticesSize;
m_presentVertexBufferView.BufferLocation = m_trianglesVertexBufferView.BufferLocation + m_trianglesVertexBufferView.SizeInBytes;
m_presentVertexBufferView.StrideInBytes = sizeof(PresentVertex);
m_presentVertexBufferView.SizeInBytes = sizeof(presentVertices);
}
LoadSizeDependentResources();
// Close the command list and execute it to begin the vertex buffer copy into
// the default heap.
ThrowIfFailed(m_commandList->Close());
ID3D12CommandList* ppCommandLists[] = { m_commandList.Get() };
m_commandQueue->ExecuteCommandLists(_countof(ppCommandLists), ppCommandLists);
// Create synchronization objects and wait until assets have been uploaded to the GPU.
{
ThrowIfFailed(m_device->CreateFence(m_fenceValues[m_frameIndex], D3D12_FENCE_FLAG_NONE, IID_PPV_ARGS(&m_fence)));
m_fenceValues[m_frameIndex]++;
// Create an event handle to use for frame synchronization.
m_fenceEvent = CreateEvent(nullptr, FALSE, FALSE, nullptr);
if (m_fenceEvent == nullptr)
{
ThrowIfFailed(HRESULT_FROM_WIN32(GetLastError()));
}
// Wait for the command list to execute; we are reusing the same command
// list in our main loop but for now, we just want to wait for setup to
// complete before continuing.
WaitForGpu();
}
}
void D3D12GSRender::flip(int buffer)
{
ID3D12Resource *resource_to_flip;
float viewport_w, viewport_h;
if (!is_flip_surface_in_global_memory(rsx::to_surface_target(rsx::method_registers[NV4097_SET_SURFACE_COLOR_TARGET])))
{
resource_storage &storage = get_current_resource_storage();
assert(storage.ram_framebuffer == nullptr);
size_t w = 0, h = 0, row_pitch = 0;
size_t offset = 0;
if (false)
{
CellGcmDisplayInfo* buffers = nullptr;// = vm::ps3::_ptr<CellGcmDisplayInfo>(m_gcm_buffers_addr);
u32 addr = rsx::get_address(gcm_buffers[gcm_current_buffer].offset, CELL_GCM_LOCATION_LOCAL);
w = gcm_buffers[gcm_current_buffer].width;
h = gcm_buffers[gcm_current_buffer].height;
u8 *src_buffer = vm::ps3::_ptr<u8>(addr);
row_pitch = align(w * 4, 256);
size_t texture_size = row_pitch * h; // * 4 for mipmap levels
size_t heap_offset = m_buffer_data.alloc<D3D12_TEXTURE_DATA_PLACEMENT_ALIGNMENT>(texture_size);
void *mapped_buffer = m_buffer_data.map<void>(heap_offset);
for (unsigned row = 0; row < h; row++)
memcpy((char*)mapped_buffer + row * row_pitch, (char*)src_buffer + row * w * 4, w * 4);
m_buffer_data.unmap(CD3DX12_RANGE(heap_offset, heap_offset + texture_size));
offset = heap_offset;
}
CHECK_HRESULT(
m_device->CreateCommittedResource(
&CD3DX12_HEAP_PROPERTIES(D3D12_HEAP_TYPE_DEFAULT),
D3D12_HEAP_FLAG_NONE,
&CD3DX12_RESOURCE_DESC::Tex2D(DXGI_FORMAT_R8G8B8A8_UNORM, (UINT)w, (UINT)h, 1, 1),
D3D12_RESOURCE_STATE_COPY_DEST,
nullptr,
IID_PPV_ARGS(storage.ram_framebuffer.GetAddressOf())
)
);
get_current_resource_storage().command_list->CopyTextureRegion(&CD3DX12_TEXTURE_COPY_LOCATION(storage.ram_framebuffer.Get(), 0), 0, 0, 0,
&CD3DX12_TEXTURE_COPY_LOCATION(m_buffer_data.get_heap(), { offset, { DXGI_FORMAT_R8G8B8A8_UNORM, (UINT)w, (UINT)h, 1, (UINT)row_pitch } }), nullptr);
get_current_resource_storage().command_list->ResourceBarrier(1, &CD3DX12_RESOURCE_BARRIER::Transition(storage.ram_framebuffer.Get(), D3D12_RESOURCE_STATE_COPY_DEST, D3D12_RESOURCE_STATE_GENERIC_READ));
resource_to_flip = storage.ram_framebuffer.Get();
viewport_w = (float)w, viewport_h = (float)h;
}
else
{
if (std::get<1>(m_rtts.m_bound_render_targets[0]) != nullptr)
{
get_current_resource_storage().command_list->ResourceBarrier(1, &CD3DX12_RESOURCE_BARRIER::Transition(std::get<1>(m_rtts.m_bound_render_targets[0]), D3D12_RESOURCE_STATE_RENDER_TARGET, D3D12_RESOURCE_STATE_GENERIC_READ));
resource_to_flip = std::get<1>(m_rtts.m_bound_render_targets[0]);
}
else if (std::get<1>(m_rtts.m_bound_render_targets[1]) != nullptr)
{
get_current_resource_storage().command_list->ResourceBarrier(1, &CD3DX12_RESOURCE_BARRIER::Transition(std::get<1>(m_rtts.m_bound_render_targets[1]), D3D12_RESOURCE_STATE_RENDER_TARGET, D3D12_RESOURCE_STATE_GENERIC_READ));
resource_to_flip = std::get<1>(m_rtts.m_bound_render_targets[1]);
}
else
resource_to_flip = nullptr;
}
get_current_resource_storage().command_list->ResourceBarrier(1, &CD3DX12_RESOURCE_BARRIER::Transition(m_backbuffer[m_swap_chain->GetCurrentBackBufferIndex()].Get(), D3D12_RESOURCE_STATE_PRESENT, D3D12_RESOURCE_STATE_RENDER_TARGET));
D3D12_VIEWPORT viewport =
{
0.f,
0.f,
(float)m_backbuffer[m_swap_chain->GetCurrentBackBufferIndex()]->GetDesc().Width,
(float)m_backbuffer[m_swap_chain->GetCurrentBackBufferIndex()]->GetDesc().Height,
0.f,
1.f
};
get_current_resource_storage().command_list->RSSetViewports(1, &viewport);
D3D12_RECT box =
{
0,
0,
(LONG)m_backbuffer[m_swap_chain->GetCurrentBackBufferIndex()]->GetDesc().Width,
(LONG)m_backbuffer[m_swap_chain->GetCurrentBackBufferIndex()]->GetDesc().Height,
};
get_current_resource_storage().command_list->RSSetScissorRects(1, &box);
get_current_resource_storage().command_list->SetGraphicsRootSignature(m_output_scaling_pass.root_signature);
get_current_resource_storage().command_list->SetPipelineState(m_output_scaling_pass.pso);
D3D12_SHADER_RESOURCE_VIEW_DESC shader_resource_view_desc = {};
// FIXME: Not always true
shader_resource_view_desc.Format = DXGI_FORMAT_R8G8B8A8_UNORM;
shader_resource_view_desc.ViewDimension = D3D12_SRV_DIMENSION_TEXTURE2D;
shader_resource_view_desc.Texture2D.MipLevels = 1;
if (is_flip_surface_in_global_memory(rsx::to_surface_target(rsx::method_registers[NV4097_SET_SURFACE_COLOR_TARGET])))
shader_resource_view_desc.Shader4ComponentMapping = D3D12_DEFAULT_SHADER_4_COMPONENT_MAPPING;
else
shader_resource_view_desc.Shader4ComponentMapping = D3D12_ENCODE_SHADER_4_COMPONENT_MAPPING(
D3D12_SHADER_COMPONENT_MAPPING_FROM_MEMORY_COMPONENT_1,
D3D12_SHADER_COMPONENT_MAPPING_FROM_MEMORY_COMPONENT_2,
D3D12_SHADER_COMPONENT_MAPPING_FROM_MEMORY_COMPONENT_3,
D3D12_SHADER_COMPONENT_MAPPING_FROM_MEMORY_COMPONENT_0
);
m_device->CreateShaderResourceView(resource_to_flip, &shader_resource_view_desc,
CD3DX12_CPU_DESCRIPTOR_HANDLE(m_output_scaling_pass.texture_descriptor_heap->GetCPUDescriptorHandleForHeapStart()).Offset(m_swap_chain->GetCurrentBackBufferIndex(), m_descriptor_stride_srv_cbv_uav));
D3D12_SAMPLER_DESC sampler_desc = {};
sampler_desc.Filter = D3D12_FILTER_MIN_MAG_LINEAR_MIP_POINT;
sampler_desc.AddressU = D3D12_TEXTURE_ADDRESS_MODE_WRAP;
sampler_desc.AddressV = D3D12_TEXTURE_ADDRESS_MODE_WRAP;
sampler_desc.AddressW = D3D12_TEXTURE_ADDRESS_MODE_WRAP;
m_device->CreateSampler(&sampler_desc,
CD3DX12_CPU_DESCRIPTOR_HANDLE(m_output_scaling_pass.sampler_descriptor_heap->GetCPUDescriptorHandleForHeapStart()).Offset(m_swap_chain->GetCurrentBackBufferIndex(), m_descriptor_stride_samplers));
ID3D12DescriptorHeap *descriptors_heaps[] =
{
m_output_scaling_pass.texture_descriptor_heap,
m_output_scaling_pass.sampler_descriptor_heap
};
get_current_resource_storage().command_list->SetDescriptorHeaps(2, descriptors_heaps);
get_current_resource_storage().command_list->SetGraphicsRootDescriptorTable(0,
CD3DX12_GPU_DESCRIPTOR_HANDLE(m_output_scaling_pass.texture_descriptor_heap->GetGPUDescriptorHandleForHeapStart()).Offset(m_swap_chain->GetCurrentBackBufferIndex(), m_descriptor_stride_srv_cbv_uav));
get_current_resource_storage().command_list->SetGraphicsRootDescriptorTable(1,
CD3DX12_GPU_DESCRIPTOR_HANDLE(m_output_scaling_pass.sampler_descriptor_heap->GetGPUDescriptorHandleForHeapStart()).Offset(m_swap_chain->GetCurrentBackBufferIndex(), m_descriptor_stride_samplers));
get_current_resource_storage().command_list->OMSetRenderTargets(1,
&CD3DX12_CPU_DESCRIPTOR_HANDLE(m_backbuffer_descriptor_heap[m_swap_chain->GetCurrentBackBufferIndex()]->GetCPUDescriptorHandleForHeapStart()),
true, nullptr);
D3D12_VERTEX_BUFFER_VIEW vertex_buffer_view = {};
vertex_buffer_view.BufferLocation = m_output_scaling_pass.vertex_buffer->GetGPUVirtualAddress();
vertex_buffer_view.StrideInBytes = 4 * sizeof(float);
vertex_buffer_view.SizeInBytes = 16 * sizeof(float);
get_current_resource_storage().command_list->IASetVertexBuffers(0, 1, &vertex_buffer_view);
get_current_resource_storage().command_list->IASetPrimitiveTopology(D3D_PRIMITIVE_TOPOLOGY_TRIANGLESTRIP);
if (resource_to_flip)
get_current_resource_storage().command_list->DrawInstanced(4, 1, 0, 0);
if (!rpcs3::config.rsx.d3d12.overlay.value())
get_current_resource_storage().command_list->ResourceBarrier(1, &CD3DX12_RESOURCE_BARRIER::Transition(m_backbuffer[m_swap_chain->GetCurrentBackBufferIndex()].Get(), D3D12_RESOURCE_STATE_RENDER_TARGET, D3D12_RESOURCE_STATE_PRESENT));
if (is_flip_surface_in_global_memory(rsx::to_surface_target(rsx::method_registers[NV4097_SET_SURFACE_COLOR_TARGET])) && resource_to_flip != nullptr)
get_current_resource_storage().command_list->ResourceBarrier(1, &CD3DX12_RESOURCE_BARRIER::Transition(resource_to_flip, D3D12_RESOURCE_STATE_GENERIC_READ, D3D12_RESOURCE_STATE_RENDER_TARGET));
CHECK_HRESULT(get_current_resource_storage().command_list->Close());
m_command_queue->ExecuteCommandLists(1, (ID3D12CommandList**)get_current_resource_storage().command_list.GetAddressOf());
if(rpcs3::config.rsx.d3d12.overlay.value())
render_overlay();
reset_timer();
std::chrono::time_point<std::chrono::system_clock> flip_start = std::chrono::system_clock::now();
CHECK_HRESULT(m_swap_chain->Present(rpcs3::state.config.rsx.vsync.value() ? 1 : 0, 0));
// Add an event signaling queue completion
resource_storage &storage = get_non_current_resource_storage();
m_command_queue->Signal(storage.frame_finished_fence.Get(), storage.fence_value);
storage.frame_finished_fence->SetEventOnCompletion(storage.fence_value, storage.frame_finished_handle);
storage.fence_value++;
storage.in_use = true;
storage.dirty_textures.merge(m_rtts.invalidated_resources);
m_rtts.invalidated_resources.clear();
// Get the put pos - 1. This way after cleaning we can set the get ptr to
// this value, allowing heap to proceed even if we cleant before allocating
// a new value (that's the reason of the -1)
storage.buffer_heap_get_pos = m_buffer_data.get_current_put_pos_minus_one();
storage.readback_heap_get_pos = m_readback_resources.get_current_put_pos_minus_one();
// Now get ready for next frame
resource_storage &new_storage = get_current_resource_storage();
new_storage.wait_and_clean();
if (new_storage.in_use)
{
m_buffer_data.m_get_pos = new_storage.buffer_heap_get_pos;
m_readback_resources.m_get_pos = new_storage.readback_heap_get_pos;
}
m_frame->flip(nullptr);
std::chrono::time_point<std::chrono::system_clock> flip_end = std::chrono::system_clock::now();
m_timers.flip_duration += std::chrono::duration_cast<std::chrono::microseconds>(flip_end - flip_start).count();
}
void DX12ClothSimulation::LoadSizeDependentResources()
{
m_viewport.Width = static_cast<float>( m_width );
m_viewport.Height = static_cast<float>( m_height );
m_viewport.MaxDepth = 1.0f;
m_scissorRect.right = static_cast<LONG>( m_width );
m_scissorRect.bottom = static_cast<LONG>( m_height );
// Create frame resources.
{
CD3DX12_CPU_DESCRIPTOR_HANDLE rtvHandle( m_rtvHeap->GetCPUDescriptorHandleForHeapStart() );
// Create a RTV for each frame.
for ( UINT n = 0; n < FrameCount; n++ )
{
ThrowIfFailed( m_swapChain->GetBuffer( n, IID_PPV_ARGS( &m_renderTargets[n] ) ) );
m_device->CreateRenderTargetView( m_renderTargets[n].Get(), nullptr, rtvHandle );
rtvHandle.Offset( 1, m_rtvDescriptorSize );
}
}
// Create/update the vertex buffer. When updating the vertex buffer it is important
// to ensure that the GPU is finished using the resource before it is released.
// The OnSizeChanged method waits for the GPU to be idle before this method is
// called.
{
// Define the geometry for a triangle that stays the same size regardless
// of the window size. This is not the recommended way to transform vertices.
// The same effect could be achieved by using constant buffers and
// transforming a static set of vertices in the vertex shader, but this
// sample merely demonstrates modifying a resource that is tied to the render
// target size.
// Other apps might also resize intermediate render targets or depth stencils
// at this time.
float x = TriangleWidth / m_viewport.Width;
float y = TriangleWidth / m_viewport.Height;
Vertex triangleVertices[] =
{
{ { 0.0f, y, 0.0f },{ 1.0f, 0.0f, 0.0f, 1.0f } },
{ { x, -y, 0.0f },{ 0.0f, 1.0f, 0.0f, 1.0f } },
{ { -x, -y, 0.0f },{ 0.0f, 0.0f, 1.0f, 1.0f } }
};
const UINT vertexBufferSize = sizeof( triangleVertices );
ThrowIfFailed( m_device->CreateCommittedResource(
&CD3DX12_HEAP_PROPERTIES( D3D12_HEAP_TYPE_DEFAULT ),
D3D12_HEAP_FLAG_NONE,
&CD3DX12_RESOURCE_DESC::Buffer( vertexBufferSize ),
D3D12_RESOURCE_STATE_COPY_DEST,
nullptr,
IID_PPV_ARGS( &m_vertexBuffer ) ) );
ThrowIfFailed( m_device->CreateCommittedResource(
&CD3DX12_HEAP_PROPERTIES( D3D12_HEAP_TYPE_UPLOAD ),
D3D12_HEAP_FLAG_NONE,
&CD3DX12_RESOURCE_DESC::Buffer( vertexBufferSize ),
D3D12_RESOURCE_STATE_GENERIC_READ,
nullptr,
IID_PPV_ARGS( &m_vertexBufferUpload ) ) );
// Copy data to the intermediate upload heap and then schedule a copy
// from the upload heap to the vertex buffer.
UINT8* pVertexDataBegin;
ThrowIfFailed( m_vertexBufferUpload->Map( 0, &CD3DX12_RANGE( 0, vertexBufferSize ), reinterpret_cast<void**>( &pVertexDataBegin ) ) );
memcpy( pVertexDataBegin, triangleVertices, sizeof( triangleVertices ) );
m_vertexBufferUpload->Unmap( 0, nullptr );
m_commandList->CopyBufferRegion( m_vertexBuffer.Get(), 0, m_vertexBufferUpload.Get(), 0, vertexBufferSize );
m_commandList->ResourceBarrier( 1, &CD3DX12_RESOURCE_BARRIER::Transition( m_vertexBuffer.Get(), D3D12_RESOURCE_STATE_COPY_DEST, D3D12_RESOURCE_STATE_VERTEX_AND_CONSTANT_BUFFER ) );
// Initialize the vertex buffer views.
m_vertexBufferView.BufferLocation = m_vertexBuffer->GetGPUVirtualAddress();
m_vertexBufferView.StrideInBytes = sizeof( Vertex );
m_vertexBufferView.SizeInBytes = vertexBufferSize;
}
m_resizeResources = false;
}
std::vector<D3D12_SHADER_RESOURCE_VIEW_DESC> D3D12GSRender::upload_vertex_attributes(
const std::vector<std::pair<u32, u32> > &vertex_ranges,
gsl::not_null<ID3D12GraphicsCommandList*> command_list)
{
std::vector<D3D12_SHADER_RESOURCE_VIEW_DESC> vertex_buffer_views;
command_list->ResourceBarrier(1, &CD3DX12_RESOURCE_BARRIER::Transition(m_vertex_buffer_data.Get(), D3D12_RESOURCE_STATE_VERTEX_AND_CONSTANT_BUFFER, D3D12_RESOURCE_STATE_COPY_DEST));
u32 vertex_count = get_vertex_count(vertex_ranges);
size_t offset_in_vertex_buffers_buffer = 0;
u32 input_mask = rsx::method_registers[NV4097_SET_VERTEX_ATTRIB_INPUT_MASK];
for (int index = 0; index < rsx::limits::vertex_count; ++index)
{
bool enabled = !!(input_mask & (1 << index));
if (!enabled)
continue;
if (vertex_arrays_info[index].size > 0)
{
// Active vertex array
const rsx::data_array_format_info &info = vertex_arrays_info[index];
u32 element_size = rsx::get_vertex_type_size_on_host(info.type, info.size);
UINT buffer_size = element_size * vertex_count;
size_t heap_offset = m_buffer_data.alloc<D3D12_CONSTANT_BUFFER_DATA_PLACEMENT_ALIGNMENT>(buffer_size);
void *mapped_buffer = m_buffer_data.map<void>(CD3DX12_RANGE(heap_offset, heap_offset + buffer_size));
for (const auto &range : vertex_ranges)
{
write_vertex_array_data_to_buffer(mapped_buffer, range.first, range.second, index, info);
mapped_buffer = (char*)mapped_buffer + range.second * element_size;
}
m_buffer_data.unmap(CD3DX12_RANGE(heap_offset, heap_offset + buffer_size));
command_list->CopyBufferRegion(m_vertex_buffer_data.Get(), offset_in_vertex_buffers_buffer, m_buffer_data.get_heap(), heap_offset, buffer_size);
vertex_buffer_views.emplace_back(get_vertex_attribute_srv(info, offset_in_vertex_buffers_buffer, buffer_size));
offset_in_vertex_buffers_buffer = get_next_multiple_of<48>(offset_in_vertex_buffers_buffer + buffer_size); // 48 is multiple of 2, 4, 6, 8, 12, 16
m_timers.buffer_upload_size += buffer_size;
}
else if (register_vertex_info[index].size > 0)
{
// In register vertex attribute
const rsx::data_array_format_info &info = register_vertex_info[index];
const std::vector<u8> &data = register_vertex_data[index];
u32 element_size = rsx::get_vertex_type_size_on_host(info.type, info.size);
UINT buffer_size = gsl::narrow<UINT>(data.size());
size_t heap_offset = m_buffer_data.alloc<D3D12_CONSTANT_BUFFER_DATA_PLACEMENT_ALIGNMENT>(buffer_size);
void *mapped_buffer = m_buffer_data.map<void>(CD3DX12_RANGE(heap_offset, heap_offset + buffer_size));
memcpy(mapped_buffer, data.data(), data.size());
m_buffer_data.unmap(CD3DX12_RANGE(heap_offset, heap_offset + buffer_size));
command_list->CopyBufferRegion(m_vertex_buffer_data.Get(), offset_in_vertex_buffers_buffer, m_buffer_data.get_heap(), heap_offset, buffer_size);
vertex_buffer_views.emplace_back(get_vertex_attribute_srv(info, offset_in_vertex_buffers_buffer, buffer_size));
offset_in_vertex_buffers_buffer = get_next_multiple_of<48>(offset_in_vertex_buffers_buffer + buffer_size); // 48 is multiple of 2, 4, 6, 8, 12, 16
}
}
command_list->ResourceBarrier(1, &CD3DX12_RESOURCE_BARRIER::Transition(m_vertex_buffer_data.Get(), D3D12_RESOURCE_STATE_COPY_DEST, D3D12_RESOURCE_STATE_VERTEX_AND_CONSTANT_BUFFER));
return vertex_buffer_views;
}
std::tuple<bool, size_t, std::vector<D3D12_SHADER_RESOURCE_VIEW_DESC>> D3D12GSRender::upload_and_set_vertex_index_data(ID3D12GraphicsCommandList *command_list)
{
if (draw_command == rsx::draw_command::inlined_array)
{
size_t vertex_count;
std::vector<D3D12_SHADER_RESOURCE_VIEW_DESC> vertex_buffer_view;
std::tie(vertex_buffer_view, vertex_count) = upload_inlined_vertex_array(
vertex_arrays_info,
{ (const gsl::byte*) inline_vertex_array.data(), gsl::narrow<int>(inline_vertex_array.size() * sizeof(uint)) },
m_buffer_data, m_vertex_buffer_data.Get(), command_list);
if (is_primitive_native(draw_mode))
return std::make_tuple(false, vertex_count, vertex_buffer_view);
D3D12_INDEX_BUFFER_VIEW index_buffer_view;
size_t index_count;
std::tie(index_buffer_view, index_count) = generate_index_buffer_for_emulated_primitives_array({ { 0, (u32)vertex_count } });
command_list->IASetIndexBuffer(&index_buffer_view);
return std::make_tuple(true, index_count, vertex_buffer_view);
}
if (draw_command == rsx::draw_command::array)
{
if (is_primitive_native(draw_mode))
{
size_t vertex_count = get_vertex_count(first_count_commands);
return std::make_tuple(false, vertex_count, upload_vertex_attributes(first_count_commands, command_list));
}
D3D12_INDEX_BUFFER_VIEW index_buffer_view;
size_t index_count;
std::tie(index_buffer_view, index_count) = generate_index_buffer_for_emulated_primitives_array(first_count_commands);
command_list->IASetIndexBuffer(&index_buffer_view);
return std::make_tuple(true, index_count, upload_vertex_attributes(first_count_commands, command_list));
}
assert(draw_command == rsx::draw_command::indexed);
// Index count
size_t index_count = get_index_count(draw_mode, gsl::narrow<int>(get_vertex_count(first_count_commands)));
rsx::index_array_type indexed_type = rsx::to_index_array_type(rsx::method_registers[NV4097_SET_INDEX_ARRAY_DMA] >> 4);
size_t index_size = get_index_type_size(indexed_type);
// Alloc
size_t buffer_size = align(index_count * index_size, 64);
size_t heap_offset = m_buffer_data.alloc<D3D12_CONSTANT_BUFFER_DATA_PLACEMENT_ALIGNMENT>(buffer_size);
void *mapped_buffer = m_buffer_data.map<void>(CD3DX12_RANGE(heap_offset, heap_offset + buffer_size));
u32 min_index, max_index;
if (indexed_type == rsx::index_array_type::u16)
{
gsl::span<u16> dst = { (u16*)mapped_buffer, gsl::narrow<int>(buffer_size / index_size) };
std::tie(min_index, max_index) = write_index_array_data_to_buffer(dst, draw_mode, first_count_commands);
}
if (indexed_type == rsx::index_array_type::u32)
{
gsl::span<u32> dst = { (u32*)mapped_buffer, gsl::narrow<int>(buffer_size / index_size) };
std::tie(min_index, max_index) = write_index_array_data_to_buffer(dst, draw_mode, first_count_commands);
}
m_buffer_data.unmap(CD3DX12_RANGE(heap_offset, heap_offset + buffer_size));
D3D12_INDEX_BUFFER_VIEW index_buffer_view = {
m_buffer_data.get_heap()->GetGPUVirtualAddress() + heap_offset,
(UINT)buffer_size,
get_index_type(indexed_type)
};
m_timers.buffer_upload_size += buffer_size;
command_list->IASetIndexBuffer(&index_buffer_view);
return std::make_tuple(true, index_count, upload_vertex_attributes({ std::make_pair(0, max_index + 1) }, command_list));
}
