bgfx/src/renderer_d3d12.cpp
2015-08-15 18:07:43 -07:00

4412 lines
137 KiB
C++

/*
* Copyright 2011-2015 Branimir Karadzic. All rights reserved.
* License: http://www.opensource.org/licenses/BSD-2-Clause
*/
#include "bgfx_p.h"
#if BGFX_CONFIG_RENDERER_DIRECT3D12
# include "renderer_d3d12.h"
# if !USE_D3D12_DYNAMIC_LIB
# pragma comment(lib, "D3D12.lib")
# endif // !USE_D3D12_DYNAMIC_LIB
namespace bgfx { namespace d3d12
{
static wchar_t s_viewNameW[BGFX_CONFIG_MAX_VIEWS][256];
struct PrimInfo
{
D3D_PRIMITIVE_TOPOLOGY m_toplogy;
D3D12_PRIMITIVE_TOPOLOGY_TYPE m_topologyType;
uint32_t m_min;
uint32_t m_div;
uint32_t m_sub;
};
static const PrimInfo s_primInfo[] =
{
{ D3D_PRIMITIVE_TOPOLOGY_TRIANGLELIST, D3D12_PRIMITIVE_TOPOLOGY_TYPE_TRIANGLE, 3, 3, 0 },
{ D3D_PRIMITIVE_TOPOLOGY_TRIANGLESTRIP, D3D12_PRIMITIVE_TOPOLOGY_TYPE_TRIANGLE, 3, 1, 2 },
{ D3D_PRIMITIVE_TOPOLOGY_LINELIST, D3D12_PRIMITIVE_TOPOLOGY_TYPE_LINE, 2, 2, 0 },
{ D3D_PRIMITIVE_TOPOLOGY_POINTLIST, D3D12_PRIMITIVE_TOPOLOGY_TYPE_POINT, 1, 1, 0 },
{ D3D_PRIMITIVE_TOPOLOGY_UNDEFINED, D3D12_PRIMITIVE_TOPOLOGY_TYPE_UNDEFINED, 0, 0, 0 },
};
static const char* s_primName[] =
{
"TriList",
"TriStrip",
"Line",
"Point",
};
BX_STATIC_ASSERT(BX_COUNTOF(s_primInfo) == BX_COUNTOF(s_primName)+1);
struct DrawIndirectCommand
{
// D3D12_GPU_VIRTUAL_ADDRESS srv;
D3D12_GPU_VIRTUAL_ADDRESS cbv;
D3D12_VERTEX_BUFFER_VIEW vbv;
D3D12_DRAW_ARGUMENTS draw;
};
struct DrawIndexedIndirectCommand
{
// D3D12_GPU_VIRTUAL_ADDRESS srv;
D3D12_GPU_VIRTUAL_ADDRESS cbv;
D3D12_VERTEX_BUFFER_VIEW vbv;
D3D12_INDEX_BUFFER_VIEW ibv;
D3D12_DRAW_INDEXED_ARGUMENTS drawIndexed;
};
static const uint32_t s_checkMsaa[] =
{
0,
2,
4,
8,
16,
};
static DXGI_SAMPLE_DESC s_msaa[] =
{
{ 1, 0 },
{ 2, 0 },
{ 4, 0 },
{ 8, 0 },
{ 16, 0 },
};
static const D3D12_BLEND s_blendFactor[][2] =
{
{ (D3D12_BLEND)0, (D3D12_BLEND)0 }, // ignored
{ D3D12_BLEND_ZERO, D3D12_BLEND_ZERO }, // ZERO
{ D3D12_BLEND_ONE, D3D12_BLEND_ONE }, // ONE
{ D3D12_BLEND_SRC_COLOR, D3D12_BLEND_SRC_ALPHA }, // SRC_COLOR
{ D3D12_BLEND_INV_SRC_COLOR, D3D12_BLEND_INV_SRC_ALPHA }, // INV_SRC_COLOR
{ D3D12_BLEND_SRC_ALPHA, D3D12_BLEND_SRC_ALPHA }, // SRC_ALPHA
{ D3D12_BLEND_INV_SRC_ALPHA, D3D12_BLEND_INV_SRC_ALPHA }, // INV_SRC_ALPHA
{ D3D12_BLEND_DEST_ALPHA, D3D12_BLEND_DEST_ALPHA }, // DST_ALPHA
{ D3D12_BLEND_INV_DEST_ALPHA, D3D12_BLEND_INV_DEST_ALPHA }, // INV_DST_ALPHA
{ D3D12_BLEND_DEST_COLOR, D3D12_BLEND_DEST_ALPHA }, // DST_COLOR
{ D3D12_BLEND_INV_DEST_COLOR, D3D12_BLEND_INV_DEST_ALPHA }, // INV_DST_COLOR
{ D3D12_BLEND_SRC_ALPHA_SAT, D3D12_BLEND_ONE }, // SRC_ALPHA_SAT
{ D3D12_BLEND_BLEND_FACTOR, D3D12_BLEND_BLEND_FACTOR }, // FACTOR
{ D3D12_BLEND_INV_BLEND_FACTOR, D3D12_BLEND_INV_BLEND_FACTOR }, // INV_FACTOR
};
static const D3D12_BLEND_OP s_blendEquation[] =
{
D3D12_BLEND_OP_ADD,
D3D12_BLEND_OP_SUBTRACT,
D3D12_BLEND_OP_REV_SUBTRACT,
D3D12_BLEND_OP_MIN,
D3D12_BLEND_OP_MAX,
};
static const D3D12_COMPARISON_FUNC s_cmpFunc[] =
{
D3D12_COMPARISON_FUNC(0), // ignored
D3D12_COMPARISON_FUNC_LESS,
D3D12_COMPARISON_FUNC_LESS_EQUAL,
D3D12_COMPARISON_FUNC_EQUAL,
D3D12_COMPARISON_FUNC_GREATER_EQUAL,
D3D12_COMPARISON_FUNC_GREATER,
D3D12_COMPARISON_FUNC_NOT_EQUAL,
D3D12_COMPARISON_FUNC_NEVER,
D3D12_COMPARISON_FUNC_ALWAYS,
};
static const D3D12_STENCIL_OP s_stencilOp[] =
{
D3D12_STENCIL_OP_ZERO,
D3D12_STENCIL_OP_KEEP,
D3D12_STENCIL_OP_REPLACE,
D3D12_STENCIL_OP_INCR,
D3D12_STENCIL_OP_INCR_SAT,
D3D12_STENCIL_OP_DECR,
D3D12_STENCIL_OP_DECR_SAT,
D3D12_STENCIL_OP_INVERT,
};
static const D3D12_CULL_MODE s_cullMode[] =
{
D3D12_CULL_MODE_NONE,
D3D12_CULL_MODE_FRONT,
D3D12_CULL_MODE_BACK,
};
static const D3D12_TEXTURE_ADDRESS_MODE s_textureAddress[] =
{
D3D12_TEXTURE_ADDRESS_MODE_WRAP,
D3D12_TEXTURE_ADDRESS_MODE_MIRROR,
D3D12_TEXTURE_ADDRESS_MODE_CLAMP,
};
/*
* D3D11_FILTER_MIN_MAG_MIP_POINT = 0x00,
* D3D11_FILTER_MIN_MAG_POINT_MIP_LINEAR = 0x01,
* D3D11_FILTER_MIN_POINT_MAG_LINEAR_MIP_POINT = 0x04,
* D3D11_FILTER_MIN_POINT_MAG_MIP_LINEAR = 0x05,
* D3D11_FILTER_MIN_LINEAR_MAG_MIP_POINT = 0x10,
* D3D11_FILTER_MIN_LINEAR_MAG_POINT_MIP_LINEAR = 0x11,
* D3D11_FILTER_MIN_MAG_LINEAR_MIP_POINT = 0x14,
* D3D11_FILTER_MIN_MAG_MIP_LINEAR = 0x15,
* D3D11_FILTER_ANISOTROPIC = 0x55,
*
* D3D11_COMPARISON_FILTERING_BIT = 0x80,
* D3D11_ANISOTROPIC_FILTERING_BIT = 0x40,
*
* According to D3D11_FILTER enum bits for mip, mag and mip are:
* 0x10 // MIN_LINEAR
* 0x04 // MAG_LINEAR
* 0x01 // MIP_LINEAR
*/
static const uint8_t s_textureFilter[3][3] =
{
{
0x10, // min linear
0x00, // min point
0x55, // anisotropic
},
{
0x04, // mag linear
0x00, // mag point
0x55, // anisotropic
},
{
0x01, // mip linear
0x00, // mip point
0x55, // anisotropic
},
};
struct TextureFormatInfo
{
DXGI_FORMAT m_fmt;
DXGI_FORMAT m_fmtSrv;
DXGI_FORMAT m_fmtDsv;
DXGI_FORMAT m_fmtSrgb;
};
static const TextureFormatInfo s_textureFormat[] =
{
{ DXGI_FORMAT_BC1_UNORM, DXGI_FORMAT_BC1_UNORM, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_BC1_UNORM_SRGB }, // BC1
{ DXGI_FORMAT_BC2_UNORM, DXGI_FORMAT_BC2_UNORM, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_BC2_UNORM_SRGB }, // BC2
{ DXGI_FORMAT_BC3_UNORM, DXGI_FORMAT_BC3_UNORM, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_BC3_UNORM_SRGB }, // BC3
{ DXGI_FORMAT_BC4_UNORM, DXGI_FORMAT_BC4_UNORM, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, // BC4
{ DXGI_FORMAT_BC5_UNORM, DXGI_FORMAT_BC5_UNORM, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, // BC5
{ DXGI_FORMAT_BC6H_SF16, DXGI_FORMAT_BC6H_SF16, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, // BC6H
{ DXGI_FORMAT_BC7_UNORM, DXGI_FORMAT_BC7_UNORM, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_BC7_UNORM_SRGB }, // BC7
{ DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, // ETC1
{ DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, // ETC2
{ DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, // ETC2A
{ DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, // ETC2A1
{ DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, // PTC12
{ DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, // PTC14
{ DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, // PTC12A
{ DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, // PTC14A
{ DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, // PTC22
{ DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, // PTC24
{ DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, // Unknown
{ DXGI_FORMAT_R1_UNORM, DXGI_FORMAT_R1_UNORM, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, // R1
{ DXGI_FORMAT_R8_UNORM, DXGI_FORMAT_R8_UNORM, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, // R8
{ DXGI_FORMAT_R16_UNORM, DXGI_FORMAT_R16_UNORM, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, // R16
{ DXGI_FORMAT_R16_FLOAT, DXGI_FORMAT_R16_FLOAT, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, // R16F
{ DXGI_FORMAT_R32_UINT, DXGI_FORMAT_R32_UINT, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, // R32
{ DXGI_FORMAT_R32_FLOAT, DXGI_FORMAT_R32_FLOAT, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, // R32F
{ DXGI_FORMAT_R8G8_UNORM, DXGI_FORMAT_R8G8_UNORM, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, // RG8
{ DXGI_FORMAT_R16G16_UNORM, DXGI_FORMAT_R16G16_UNORM, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, // RG16
{ DXGI_FORMAT_R16G16_FLOAT, DXGI_FORMAT_R16G16_FLOAT, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, // RG16F
{ DXGI_FORMAT_R32G32_UINT, DXGI_FORMAT_R32G32_UINT, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, // RG32
{ DXGI_FORMAT_R32G32_FLOAT, DXGI_FORMAT_R32G32_FLOAT, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, // RG32F
{ DXGI_FORMAT_B8G8R8A8_UNORM, DXGI_FORMAT_B8G8R8A8_UNORM, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_B8G8R8A8_UNORM_SRGB }, // BGRA8
{ DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_R8G8B8A8_UNORM_SRGB }, // RGBA8
{ DXGI_FORMAT_R16G16B16A16_UNORM, DXGI_FORMAT_R16G16B16A16_UNORM, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, // RGBA16
{ DXGI_FORMAT_R16G16B16A16_FLOAT, DXGI_FORMAT_R16G16B16A16_FLOAT, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, // RGBA16F
{ DXGI_FORMAT_R32G32B32A32_UINT, DXGI_FORMAT_R32G32B32A32_UINT, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, // RGBA32
{ DXGI_FORMAT_R32G32B32A32_FLOAT, DXGI_FORMAT_R32G32B32A32_FLOAT, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, // RGBA32F
{ DXGI_FORMAT_B5G6R5_UNORM, DXGI_FORMAT_B5G6R5_UNORM, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, // R5G6B5
{ DXGI_FORMAT_B4G4R4A4_UNORM, DXGI_FORMAT_B4G4R4A4_UNORM, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, // RGBA4
{ DXGI_FORMAT_B5G5R5A1_UNORM, DXGI_FORMAT_B5G5R5A1_UNORM, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, // RGB5A1
{ DXGI_FORMAT_R10G10B10A2_UNORM, DXGI_FORMAT_R10G10B10A2_UNORM, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, // RGB10A2
{ DXGI_FORMAT_R11G11B10_FLOAT, DXGI_FORMAT_R11G11B10_FLOAT, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, // R11G11B10F
{ DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, // UnknownDepth
{ DXGI_FORMAT_R16_TYPELESS, DXGI_FORMAT_R16_UNORM, DXGI_FORMAT_D16_UNORM, DXGI_FORMAT_UNKNOWN }, // D16
{ DXGI_FORMAT_R24G8_TYPELESS, DXGI_FORMAT_R24_UNORM_X8_TYPELESS, DXGI_FORMAT_D24_UNORM_S8_UINT, DXGI_FORMAT_UNKNOWN }, // D24
{ DXGI_FORMAT_R24G8_TYPELESS, DXGI_FORMAT_R24_UNORM_X8_TYPELESS, DXGI_FORMAT_D24_UNORM_S8_UINT, DXGI_FORMAT_UNKNOWN }, // D24S8
{ DXGI_FORMAT_R24G8_TYPELESS, DXGI_FORMAT_R24_UNORM_X8_TYPELESS, DXGI_FORMAT_D24_UNORM_S8_UINT, DXGI_FORMAT_UNKNOWN }, // D32
{ DXGI_FORMAT_R32_TYPELESS, DXGI_FORMAT_R32_FLOAT, DXGI_FORMAT_D32_FLOAT, DXGI_FORMAT_UNKNOWN }, // D16F
{ DXGI_FORMAT_R32_TYPELESS, DXGI_FORMAT_R32_FLOAT, DXGI_FORMAT_D32_FLOAT, DXGI_FORMAT_UNKNOWN }, // D24F
{ DXGI_FORMAT_R32_TYPELESS, DXGI_FORMAT_R32_FLOAT, DXGI_FORMAT_D32_FLOAT, DXGI_FORMAT_UNKNOWN }, // D32F
{ DXGI_FORMAT_R24G8_TYPELESS, DXGI_FORMAT_R24_UNORM_X8_TYPELESS, DXGI_FORMAT_D24_UNORM_S8_UINT, DXGI_FORMAT_UNKNOWN }, // D0S8
};
BX_STATIC_ASSERT(TextureFormat::Count == BX_COUNTOF(s_textureFormat) );
static const D3D12_INPUT_ELEMENT_DESC s_attrib[] =
{
{ "POSITION", 0, DXGI_FORMAT_R32G32B32_FLOAT, 0, D3D12_APPEND_ALIGNED_ELEMENT, D3D12_INPUT_CLASSIFICATION_PER_VERTEX_DATA, 0 },
{ "NORMAL", 0, DXGI_FORMAT_R32G32B32_FLOAT, 0, D3D12_APPEND_ALIGNED_ELEMENT, D3D12_INPUT_CLASSIFICATION_PER_VERTEX_DATA, 0 },
{ "TANGENT", 0, DXGI_FORMAT_R32G32B32_FLOAT, 0, D3D12_APPEND_ALIGNED_ELEMENT, D3D12_INPUT_CLASSIFICATION_PER_VERTEX_DATA, 0 },
{ "BITANGENT", 0, DXGI_FORMAT_R32G32B32_FLOAT, 0, D3D12_APPEND_ALIGNED_ELEMENT, D3D12_INPUT_CLASSIFICATION_PER_VERTEX_DATA, 0 },
{ "COLOR", 0, DXGI_FORMAT_R8G8B8A8_UINT, 0, D3D12_APPEND_ALIGNED_ELEMENT, D3D12_INPUT_CLASSIFICATION_PER_VERTEX_DATA, 0 },
{ "COLOR", 1, DXGI_FORMAT_R8G8B8A8_UINT, 0, D3D12_APPEND_ALIGNED_ELEMENT, D3D12_INPUT_CLASSIFICATION_PER_VERTEX_DATA, 0 },
{ "BLENDINDICES", 0, DXGI_FORMAT_R8G8B8A8_UINT, 0, D3D12_APPEND_ALIGNED_ELEMENT, D3D12_INPUT_CLASSIFICATION_PER_VERTEX_DATA, 0 },
{ "BLENDWEIGHT", 0, DXGI_FORMAT_R32G32B32_FLOAT, 0, D3D12_APPEND_ALIGNED_ELEMENT, 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 },
{ "TEXCOORD", 1, DXGI_FORMAT_R32G32_FLOAT, 0, D3D12_APPEND_ALIGNED_ELEMENT, D3D12_INPUT_CLASSIFICATION_PER_VERTEX_DATA, 0 },
{ "TEXCOORD", 2, DXGI_FORMAT_R32G32_FLOAT, 0, D3D12_APPEND_ALIGNED_ELEMENT, D3D12_INPUT_CLASSIFICATION_PER_VERTEX_DATA, 0 },
{ "TEXCOORD", 3, DXGI_FORMAT_R32G32_FLOAT, 0, D3D12_APPEND_ALIGNED_ELEMENT, D3D12_INPUT_CLASSIFICATION_PER_VERTEX_DATA, 0 },
{ "TEXCOORD", 4, DXGI_FORMAT_R32G32_FLOAT, 0, D3D12_APPEND_ALIGNED_ELEMENT, D3D12_INPUT_CLASSIFICATION_PER_VERTEX_DATA, 0 },
{ "TEXCOORD", 5, DXGI_FORMAT_R32G32_FLOAT, 0, D3D12_APPEND_ALIGNED_ELEMENT, D3D12_INPUT_CLASSIFICATION_PER_VERTEX_DATA, 0 },
{ "TEXCOORD", 6, DXGI_FORMAT_R32G32_FLOAT, 0, D3D12_APPEND_ALIGNED_ELEMENT, D3D12_INPUT_CLASSIFICATION_PER_VERTEX_DATA, 0 },
{ "TEXCOORD", 7, DXGI_FORMAT_R32G32_FLOAT, 0, D3D12_APPEND_ALIGNED_ELEMENT, D3D12_INPUT_CLASSIFICATION_PER_VERTEX_DATA, 0 },
};
BX_STATIC_ASSERT(Attrib::Count == BX_COUNTOF(s_attrib) );
static const DXGI_FORMAT s_attribType[][4][2] =
{
{ // Uint8
{ DXGI_FORMAT_R8_UINT, DXGI_FORMAT_R8_UNORM },
{ DXGI_FORMAT_R8G8_UINT, DXGI_FORMAT_R8G8_UNORM },
{ DXGI_FORMAT_R8G8B8A8_UINT, DXGI_FORMAT_R8G8B8A8_UNORM },
{ DXGI_FORMAT_R8G8B8A8_UINT, DXGI_FORMAT_R8G8B8A8_UNORM },
},
{ // Uint10
{ DXGI_FORMAT_R10G10B10A2_UINT, DXGI_FORMAT_R10G10B10A2_UNORM },
{ DXGI_FORMAT_R10G10B10A2_UINT, DXGI_FORMAT_R10G10B10A2_UNORM },
{ DXGI_FORMAT_R10G10B10A2_UINT, DXGI_FORMAT_R10G10B10A2_UNORM },
{ DXGI_FORMAT_R10G10B10A2_UINT, DXGI_FORMAT_R10G10B10A2_UNORM },
},
{ // Int16
{ DXGI_FORMAT_R16_SINT, DXGI_FORMAT_R16_SNORM },
{ DXGI_FORMAT_R16G16_SINT, DXGI_FORMAT_R16G16_SNORM },
{ DXGI_FORMAT_R16G16B16A16_SINT, DXGI_FORMAT_R16G16B16A16_SNORM },
{ DXGI_FORMAT_R16G16B16A16_SINT, DXGI_FORMAT_R16G16B16A16_SNORM },
},
{ // Half
{ DXGI_FORMAT_R16_FLOAT, DXGI_FORMAT_R16_FLOAT },
{ DXGI_FORMAT_R16G16_FLOAT, DXGI_FORMAT_R16G16_FLOAT },
{ DXGI_FORMAT_R16G16B16A16_FLOAT, DXGI_FORMAT_R16G16B16A16_FLOAT },
{ DXGI_FORMAT_R16G16B16A16_FLOAT, DXGI_FORMAT_R16G16B16A16_FLOAT },
},
{ // Float
{ DXGI_FORMAT_R32_FLOAT, DXGI_FORMAT_R32_FLOAT },
{ DXGI_FORMAT_R32G32_FLOAT, DXGI_FORMAT_R32G32_FLOAT },
{ DXGI_FORMAT_R32G32B32_FLOAT, DXGI_FORMAT_R32G32B32_FLOAT },
{ DXGI_FORMAT_R32G32B32A32_FLOAT, DXGI_FORMAT_R32G32B32A32_FLOAT },
},
};
BX_STATIC_ASSERT(AttribType::Count == BX_COUNTOF(s_attribType) );
static D3D12_INPUT_ELEMENT_DESC* fillVertexDecl(D3D12_INPUT_ELEMENT_DESC* _out, const VertexDecl& _decl)
{
D3D12_INPUT_ELEMENT_DESC* elem = _out;
for (uint32_t attr = 0; attr < Attrib::Count; ++attr)
{
if (UINT16_MAX != _decl.m_attributes[attr])
{
memcpy(elem, &s_attrib[attr], sizeof(D3D12_INPUT_ELEMENT_DESC) );
if (0 == _decl.m_attributes[attr])
{
elem->AlignedByteOffset = 0;
}
else
{
uint8_t num;
AttribType::Enum type;
bool normalized;
bool asInt;
_decl.decode(Attrib::Enum(attr), num, type, normalized, asInt);
elem->Format = s_attribType[type][num-1][normalized];
elem->AlignedByteOffset = _decl.m_offset[attr];
}
++elem;
}
}
return elem;
}
void setResourceBarrier(ID3D12GraphicsCommandList* _commandList, ID3D12Resource* _resource, D3D12_RESOURCE_STATES _stateBefore, D3D12_RESOURCE_STATES _stateAfter)
{
D3D12_RESOURCE_BARRIER barrier;
barrier.Type = D3D12_RESOURCE_BARRIER_TYPE_TRANSITION;
barrier.Flags = D3D12_RESOURCE_BARRIER_FLAG_NONE;
barrier.Transition.pResource = _resource;
barrier.Transition.Subresource = D3D12_RESOURCE_BARRIER_ALL_SUBRESOURCES;
barrier.Transition.StateBefore = _stateBefore;
barrier.Transition.StateAfter = _stateAfter;
_commandList->ResourceBarrier(1, &barrier);
}
struct HeapProperty
{
enum Enum
{
Default,
Upload,
ReadBack,
Count
};
D3D12_HEAP_PROPERTIES m_properties;
D3D12_RESOURCE_STATES m_state;
};
static const HeapProperty s_heapProperties[] =
{
{ { D3D12_HEAP_TYPE_DEFAULT, D3D12_CPU_PAGE_PROPERTY_UNKNOWN, D3D12_MEMORY_POOL_UNKNOWN, 1, 1 }, D3D12_RESOURCE_STATE_COMMON },
{ { D3D12_HEAP_TYPE_UPLOAD, D3D12_CPU_PAGE_PROPERTY_UNKNOWN, D3D12_MEMORY_POOL_UNKNOWN, 1, 1 }, D3D12_RESOURCE_STATE_GENERIC_READ },
{ { D3D12_HEAP_TYPE_READBACK, D3D12_CPU_PAGE_PROPERTY_UNKNOWN, D3D12_MEMORY_POOL_UNKNOWN, 1, 1 }, D3D12_RESOURCE_STATE_COPY_DEST },
};
BX_STATIC_ASSERT(BX_COUNTOF(s_heapProperties) == HeapProperty::Count);
ID3D12Resource* createCommittedResource(ID3D12Device* _device, HeapProperty::Enum _heapProperty, D3D12_RESOURCE_DESC* _resourceDesc, D3D12_CLEAR_VALUE* _clearValue)
{
const HeapProperty& heapProperty = s_heapProperties[_heapProperty];
ID3D12Resource* resource;
DX_CHECK(_device->CreateCommittedResource(&heapProperty.m_properties
, D3D12_HEAP_FLAG_NONE
, _resourceDesc
, heapProperty.m_state
, _clearValue
, __uuidof(ID3D12Resource)
, (void**)&resource
) );
return resource;
}
ID3D12Resource* createCommittedResource(ID3D12Device* _device, HeapProperty::Enum _heapProperty, uint64_t _size, D3D12_RESOURCE_FLAGS _flags = D3D12_RESOURCE_FLAG_NONE)
{
D3D12_RESOURCE_DESC resourceDesc;
resourceDesc.Dimension = D3D12_RESOURCE_DIMENSION_BUFFER;
resourceDesc.Alignment = 0;
resourceDesc.Width = _size;
resourceDesc.Height = 1;
resourceDesc.DepthOrArraySize = 1;
resourceDesc.MipLevels = 1;
resourceDesc.Format = DXGI_FORMAT_UNKNOWN;
resourceDesc.SampleDesc.Count = 1;
resourceDesc.SampleDesc.Quality = 0;
resourceDesc.Layout = D3D12_TEXTURE_LAYOUT_ROW_MAJOR;
resourceDesc.Flags = _flags;
return createCommittedResource(_device, _heapProperty, &resourceDesc, NULL);
}
BX_NO_INLINE void setDebugObjectName(ID3D12Object* _object, const char* _format, ...)
{
if (BX_ENABLED(BGFX_CONFIG_DEBUG_OBJECT_NAME) )
{
char temp[2048];
va_list argList;
va_start(argList, _format);
int size = bx::uint32_min(sizeof(temp)-1, vsnprintf(temp, sizeof(temp), _format, argList) );
va_end(argList);
temp[size] = '\0';
wchar_t* wtemp = (wchar_t*)alloca( (size+1)*2);
mbstowcs(wtemp, temp, size+1);
_object->SetName(wtemp);
}
}
#if USE_D3D12_DYNAMIC_LIB
static PFN_D3D12_CREATE_DEVICE D3D12CreateDevice;
static PFN_D3D12_GET_DEBUG_INTERFACE D3D12GetDebugInterface;
static PFN_D3D12_SERIALIZE_ROOT_SIGNATURE D3D12SerializeRootSignature;
static PFN_CREATE_DXGI_FACTORY CreateDXGIFactory1;
#endif // USE_D3D12_DYNAMIC_LIB
struct RendererContextD3D12 : public RendererContextI
{
RendererContextD3D12()
: m_wireframe(false)
, m_flags(BGFX_RESET_NONE)
, m_fsChanges(0)
, m_vsChanges(0)
, m_backBufferColorIdx(0)
, m_rtMsaa(false)
{
}
~RendererContextD3D12()
{
}
bool init()
{
uint32_t errorState = 0;
LUID luid;
m_fbh.idx = invalidHandle;
memset(m_uniforms, 0, sizeof(m_uniforms) );
memset(&m_resolution, 0, sizeof(m_resolution) );
#if USE_D3D12_DYNAMIC_LIB
m_d3d12dll = bx::dlopen("d3d12.dll");
BX_WARN(NULL != m_d3d12dll, "Failed to load d3d12.dll.");
if (NULL == m_d3d12dll)
{
goto error;
}
errorState = 1;
D3D12CreateDevice = (PFN_D3D12_CREATE_DEVICE)bx::dlsym(m_d3d12dll, "D3D12CreateDevice");
BX_WARN(NULL != D3D12CreateDevice, "Function D3D12CreateDevice not found.");
D3D12GetDebugInterface = (PFN_D3D12_GET_DEBUG_INTERFACE)bx::dlsym(m_d3d12dll, "D3D12GetDebugInterface");
BX_WARN(NULL != D3D12GetDebugInterface, "Function D3D12GetDebugInterface not found.");
D3D12SerializeRootSignature = (PFN_D3D12_SERIALIZE_ROOT_SIGNATURE)bx::dlsym(m_d3d12dll, "D3D12SerializeRootSignature");
BX_WARN(NULL != D3D12SerializeRootSignature, "Function D3D12SerializeRootSignature not found.");
if (NULL == D3D12CreateDevice
|| NULL == D3D12GetDebugInterface
|| NULL == D3D12SerializeRootSignature)
{
goto error;
}
m_dxgidll = bx::dlopen("dxgi.dll");
BX_WARN(NULL != m_dxgidll, "Failed to load dxgi.dll.");
if (NULL == m_dxgidll)
{
goto error;
}
errorState = 2;
CreateDXGIFactory1 = (PFN_CREATE_DXGI_FACTORY)bx::dlsym(m_dxgidll, "CreateDXGIFactory1");
BX_WARN(NULL != CreateDXGIFactory1, "Function CreateDXGIFactory1 not found.");
if (NULL == CreateDXGIFactory1)
{
goto error;
}
#else
errorState = 2;
#endif // USE_D3D12_DYNAMIC_LIB
HRESULT hr;
hr = CreateDXGIFactory1(__uuidof(m_factory), (void**)&m_factory);
BX_WARN(SUCCEEDED(hr), "Unable to create DXGI factory.");
if (FAILED(hr) )
{
goto error;
}
errorState = 3;
m_adapter = NULL;
m_driverType = D3D_DRIVER_TYPE_HARDWARE;
IDXGIAdapter* adapter;
for (uint32_t ii = 0; DXGI_ERROR_NOT_FOUND != m_factory->EnumAdapters(ii, &adapter); ++ii)
{
DXGI_ADAPTER_DESC desc;
hr = adapter->GetDesc(&desc);
if (SUCCEEDED(hr) )
{
BX_TRACE("Adapter #%d", ii);
char description[BX_COUNTOF(desc.Description)];
wcstombs(description, desc.Description, BX_COUNTOF(desc.Description) );
BX_TRACE("\tDescription: %s", description);
BX_TRACE("\tVendorId: 0x%08x, DeviceId: 0x%08x, SubSysId: 0x%08x, Revision: 0x%08x"
, desc.VendorId
, desc.DeviceId
, desc.SubSysId
, desc.Revision
);
BX_TRACE("\tMemory: %" PRIi64 " (video), %" PRIi64 " (system), %" PRIi64 " (shared)"
, desc.DedicatedVideoMemory
, desc.DedicatedSystemMemory
, desc.SharedSystemMemory
);
g_caps.gpu[ii].vendorId = (uint16_t)desc.VendorId;
g_caps.gpu[ii].deviceId = (uint16_t)desc.DeviceId;
++g_caps.numGPUs;
if ( (BGFX_PCI_ID_NONE != g_caps.vendorId || 0 != g_caps.deviceId)
&& (BGFX_PCI_ID_NONE == g_caps.vendorId || desc.VendorId == g_caps.vendorId)
&& (0 == g_caps.deviceId || desc.DeviceId == g_caps.deviceId) )
{
m_adapter = adapter;
m_adapter->AddRef();
m_driverType = D3D_DRIVER_TYPE_UNKNOWN;
}
if (BX_ENABLED(BGFX_CONFIG_DEBUG_PERFHUD)
&& 0 != strstr(description, "PerfHUD") )
{
m_adapter = adapter;
m_driverType = D3D_DRIVER_TYPE_REFERENCE;
}
}
DX_RELEASE(adapter, adapter == m_adapter ? 1 : 0);
}
if (BX_ENABLED(BGFX_CONFIG_DEBUG) )
{
ID3D12Debug* debug;
hr = D3D12GetDebugInterface(__uuidof(ID3D12Debug), (void**)&debug);
if (SUCCEEDED(hr) )
{
debug->EnableDebugLayer();
}
}
hr = D3D12CreateDevice(m_adapter
, D3D_FEATURE_LEVEL_11_0
, __uuidof(ID3D12Device)
, (void**)&m_device
);
BX_WARN(SUCCEEDED(hr), "Unable to create Direct3D12 device.");
if (NULL != m_adapter)
{
DX_RELEASE(m_adapter, 2);
}
if (FAILED(hr) )
{
goto error;
}
errorState = 4;
memset(&m_adapterDesc, 0, sizeof(m_adapterDesc) );
luid = m_device->GetAdapterLuid();
for (uint32_t ii = 0; DXGI_ERROR_NOT_FOUND != m_factory->EnumAdapters(ii, &adapter); ++ii)
{
adapter->GetDesc(&m_adapterDesc);
DX_RELEASE(adapter, 0);
if (m_adapterDesc.AdapterLuid.LowPart == luid.LowPart
&& m_adapterDesc.AdapterLuid.HighPart == luid.HighPart)
{
break;
}
}
g_caps.vendorId = (uint16_t)m_adapterDesc.VendorId;
g_caps.deviceId = (uint16_t)m_adapterDesc.DeviceId;
uint32_t numNodes = m_device->GetNodeCount();
BX_TRACE("D3D12 GPU Architecture (num nodes: %d):", numNodes);
for (uint32_t ii = 0; ii < numNodes; ++ii)
{
D3D12_FEATURE_DATA_ARCHITECTURE architecture;
architecture.NodeIndex = ii;
DX_CHECK(m_device->CheckFeatureSupport(D3D12_FEATURE_ARCHITECTURE, &architecture, sizeof(architecture) ) );
BX_TRACE("\tNode % 2d: TileBasedRenderer %d, UMA %d, CacheCoherentUMA %d"
, ii
, architecture.TileBasedRenderer
, architecture.UMA
, architecture.CacheCoherentUMA
);
if (0 == ii)
{
memcpy(&m_architecture, &architecture, sizeof(architecture) );
}
}
DX_CHECK(m_device->CheckFeatureSupport(D3D12_FEATURE_D3D12_OPTIONS, &m_options, sizeof(m_options) ) );
BX_TRACE("D3D12 options:")
BX_TRACE("\tTiledResourcesTier %d", m_options.TiledResourcesTier);
BX_TRACE("\tResourceBindingTier %d", m_options.ResourceBindingTier);
BX_TRACE("\tConservativeRasterizationTier %d", m_options.ConservativeRasterizationTier);
BX_TRACE("\tCrossNodeSharingTier %d", m_options.CrossNodeSharingTier);
BX_TRACE("\tResourceHeapTier %d", m_options.ResourceHeapTier);
m_cmd.init(m_device);
m_scd.BufferDesc.Width = BGFX_DEFAULT_WIDTH;
m_scd.BufferDesc.Height = BGFX_DEFAULT_HEIGHT;
m_scd.BufferDesc.Format = DXGI_FORMAT_R8G8B8A8_UNORM;
m_scd.BufferDesc.Scaling = DXGI_MODE_SCALING_STRETCHED;
m_scd.BufferDesc.ScanlineOrdering = DXGI_MODE_SCANLINE_ORDER_UNSPECIFIED;
m_scd.BufferDesc.RefreshRate.Numerator = 60;
m_scd.BufferDesc.RefreshRate.Denominator = 1;
m_scd.SampleDesc.Count = 1;
m_scd.SampleDesc.Quality = 0;
m_scd.BufferUsage = DXGI_USAGE_RENDER_TARGET_OUTPUT;
m_scd.BufferCount = bx::uint32_min(BX_COUNTOF(m_backBufferColor), 4);
m_scd.OutputWindow = (HWND)g_platformData.nwh;
m_scd.Windowed = true;
m_scd.SwapEffect = DXGI_SWAP_EFFECT_FLIP_SEQUENTIAL;
m_scd.Flags = DXGI_SWAP_CHAIN_FLAG_ALLOW_MODE_SWITCH;
BX_CHECK(m_scd.BufferCount <= BX_COUNTOF(m_backBufferColor), "Swap chain buffer count %d (max %d)."
, m_scd.BufferCount
, BX_COUNTOF(m_backBufferColor)
);
hr = m_factory->CreateSwapChain(m_cmd.m_commandQueue
, &m_scd
, reinterpret_cast<IDXGISwapChain**>(&m_swapChain)
);
BX_WARN(SUCCEEDED(hr), "Failed to create swap chain.");
if (FAILED(hr) )
{
goto error;
}
m_presentElapsed = 0;
{
m_resolution.m_width = BGFX_DEFAULT_WIDTH;
m_resolution.m_height = BGFX_DEFAULT_HEIGHT;
DX_CHECK(m_factory->MakeWindowAssociation( (HWND)g_platformData.nwh
, 0
| DXGI_MWA_NO_WINDOW_CHANGES
| DXGI_MWA_NO_ALT_ENTER
) );
m_numWindows = 1;
if (BX_ENABLED(BGFX_CONFIG_DEBUG) )
{
hr = m_device->QueryInterface(__uuidof(ID3D12InfoQueue), (void**)&m_infoQueue);
if (SUCCEEDED(hr) )
{
m_infoQueue->SetBreakOnSeverity(D3D12_MESSAGE_SEVERITY_CORRUPTION, true);
m_infoQueue->SetBreakOnSeverity(D3D12_MESSAGE_SEVERITY_ERROR, true);
m_infoQueue->SetBreakOnSeverity(D3D12_MESSAGE_SEVERITY_WARNING, false);
D3D12_INFO_QUEUE_FILTER filter;
memset(&filter, 0, sizeof(filter) );
D3D12_MESSAGE_CATEGORY catlist[] =
{
D3D12_MESSAGE_CATEGORY_STATE_CREATION,
D3D12_MESSAGE_CATEGORY_EXECUTION,
};
filter.DenyList.NumCategories = BX_COUNTOF(catlist);
filter.DenyList.pCategoryList = catlist;
m_infoQueue->PushStorageFilter(&filter);
DX_RELEASE_WARNONLY(m_infoQueue, 19);
}
}
D3D12_DESCRIPTOR_HEAP_DESC rtvDescHeap;
rtvDescHeap.NumDescriptors = 0
+ BX_COUNTOF(m_backBufferColor)
+ BGFX_CONFIG_MAX_FRAME_BUFFERS*BGFX_CONFIG_MAX_FRAME_BUFFER_ATTACHMENTS
;
rtvDescHeap.Type = D3D12_DESCRIPTOR_HEAP_TYPE_RTV;
rtvDescHeap.Flags = D3D12_DESCRIPTOR_HEAP_FLAG_NONE;
rtvDescHeap.NodeMask = 1;
DX_CHECK(m_device->CreateDescriptorHeap(&rtvDescHeap
, __uuidof(ID3D12DescriptorHeap)
, (void**)&m_rtvDescriptorHeap
) );
D3D12_DESCRIPTOR_HEAP_DESC dsvDescHeap;
dsvDescHeap.NumDescriptors = 0
+ 1 // reserved for depth backbuffer.
+ BGFX_CONFIG_MAX_FRAME_BUFFERS
;
dsvDescHeap.Type = D3D12_DESCRIPTOR_HEAP_TYPE_DSV;
dsvDescHeap.Flags = D3D12_DESCRIPTOR_HEAP_FLAG_NONE;
dsvDescHeap.NodeMask = 1;
DX_CHECK(m_device->CreateDescriptorHeap(&dsvDescHeap
, __uuidof(ID3D12DescriptorHeap)
, (void**)&m_dsvDescriptorHeap
) );
for (uint32_t ii = 0; ii < BX_COUNTOF(m_scratchBuffer); ++ii)
{
m_scratchBuffer[ii].create(BGFX_CONFIG_MAX_DRAW_CALLS*1024
, BGFX_CONFIG_MAX_TEXTURES + BGFX_CONFIG_MAX_SHADERS + BGFX_CONFIG_MAX_DRAW_CALLS
);
}
m_samplerAllocator.create(D3D12_DESCRIPTOR_HEAP_TYPE_SAMPLER
, 1024
, BGFX_CONFIG_MAX_TEXTURE_SAMPLERS
);
D3D12_DESCRIPTOR_RANGE descRange[] =
{
{ D3D12_DESCRIPTOR_RANGE_TYPE_SAMPLER, BGFX_CONFIG_MAX_TEXTURE_SAMPLERS, 0, 0, D3D12_DESCRIPTOR_RANGE_OFFSET_APPEND },
{ D3D12_DESCRIPTOR_RANGE_TYPE_SRV, BGFX_CONFIG_MAX_TEXTURE_SAMPLERS, 0, 0, D3D12_DESCRIPTOR_RANGE_OFFSET_APPEND },
{ D3D12_DESCRIPTOR_RANGE_TYPE_CBV, 1, 0, 0, D3D12_DESCRIPTOR_RANGE_OFFSET_APPEND },
{ D3D12_DESCRIPTOR_RANGE_TYPE_UAV, BGFX_CONFIG_MAX_TEXTURE_SAMPLERS, 0, 0, D3D12_DESCRIPTOR_RANGE_OFFSET_APPEND },
};
BX_STATIC_ASSERT(BX_COUNTOF(descRange) == Rdt::Count);
D3D12_ROOT_PARAMETER rootParameter[] =
{
{ D3D12_ROOT_PARAMETER_TYPE_DESCRIPTOR_TABLE, { 1, &descRange[Rdt::Sampler] }, D3D12_SHADER_VISIBILITY_ALL },
{ D3D12_ROOT_PARAMETER_TYPE_DESCRIPTOR_TABLE, { 1, &descRange[Rdt::SRV] }, D3D12_SHADER_VISIBILITY_ALL },
{ D3D12_ROOT_PARAMETER_TYPE_CBV, { 0, 0 }, D3D12_SHADER_VISIBILITY_ALL },
{ D3D12_ROOT_PARAMETER_TYPE_DESCRIPTOR_TABLE, { 1, &descRange[Rdt::UAV] }, D3D12_SHADER_VISIBILITY_ALL },
};
rootParameter[Rdt::CBV].Descriptor.RegisterSpace = 0;
rootParameter[Rdt::CBV].Descriptor.ShaderRegister = 0;
D3D12_ROOT_SIGNATURE_DESC descRootSignature;
descRootSignature.NumParameters = BX_COUNTOF(rootParameter);
descRootSignature.pParameters = rootParameter;
descRootSignature.NumStaticSamplers = 0;
descRootSignature.pStaticSamplers = NULL;
descRootSignature.Flags = D3D12_ROOT_SIGNATURE_FLAG_ALLOW_INPUT_ASSEMBLER_INPUT_LAYOUT;
ID3DBlob* outBlob;
ID3DBlob* errorBlob;
DX_CHECK(D3D12SerializeRootSignature(&descRootSignature
, D3D_ROOT_SIGNATURE_VERSION_1
, &outBlob
, &errorBlob
) );
DX_CHECK(m_device->CreateRootSignature(0
, outBlob->GetBufferPointer()
, outBlob->GetBufferSize()
, __uuidof(ID3D12RootSignature)
, (void**)&m_rootSignature
) );
D3D12_INDIRECT_ARGUMENT_DESC drawArgDesc[] =
{
// { D3D12_INDIRECT_ARGUMENT_TYPE_SHADER_RESOURCE_VIEW, { Rdt::SRV } },
{ D3D12_INDIRECT_ARGUMENT_TYPE_CONSTANT_BUFFER_VIEW, { Rdt::CBV } },
{ D3D12_INDIRECT_ARGUMENT_TYPE_VERTEX_BUFFER_VIEW, 0 },
{ D3D12_INDIRECT_ARGUMENT_TYPE_DRAW, 0 },
};
D3D12_COMMAND_SIGNATURE_DESC drawCommandSignature =
{
sizeof(DrawIndirectCommand),
BX_COUNTOF(drawArgDesc),
drawArgDesc,
1,
};
DX_CHECK(m_device->CreateCommandSignature(&drawCommandSignature
, m_rootSignature
, __uuidof(ID3D12CommandSignature)
, (void**)&m_commandSignature[0]
) );
D3D12_INDIRECT_ARGUMENT_DESC drawIndexedArgDesc[] =
{
// { D3D12_INDIRECT_ARGUMENT_TYPE_SHADER_RESOURCE_VIEW, { Rdt::SRV } },
{ D3D12_INDIRECT_ARGUMENT_TYPE_CONSTANT_BUFFER_VIEW, { Rdt::CBV } },
{ D3D12_INDIRECT_ARGUMENT_TYPE_VERTEX_BUFFER_VIEW, 0 },
{ D3D12_INDIRECT_ARGUMENT_TYPE_INDEX_BUFFER_VIEW, 0 },
{ D3D12_INDIRECT_ARGUMENT_TYPE_DRAW_INDEXED, 0 },
};
D3D12_COMMAND_SIGNATURE_DESC drawIndexedCommandSignature =
{
sizeof(DrawIndexedIndirectCommand),
BX_COUNTOF(drawIndexedArgDesc),
drawIndexedArgDesc,
1,
};
DX_CHECK(m_device->CreateCommandSignature(&drawIndexedCommandSignature
, m_rootSignature
, __uuidof(ID3D12CommandSignature)
, (void**)&m_commandSignature[1]
) );
UniformHandle handle = BGFX_INVALID_HANDLE;
for (uint32_t ii = 0; ii < PredefinedUniform::Count; ++ii)
{
m_uniformReg.add(handle, getPredefinedUniformName(PredefinedUniform::Enum(ii) ), &m_predefinedUniforms[ii]);
}
g_caps.supported |= ( 0
| BGFX_CAPS_TEXTURE_3D
| BGFX_CAPS_TEXTURE_COMPARE_ALL
| BGFX_CAPS_INSTANCING
| BGFX_CAPS_VERTEX_ATTRIB_HALF
| BGFX_CAPS_VERTEX_ATTRIB_UINT10
| BGFX_CAPS_FRAGMENT_DEPTH
| BGFX_CAPS_BLEND_INDEPENDENT
| BGFX_CAPS_COMPUTE
| BGFX_CAPS_FRAGMENT_ORDERING
// | BGFX_CAPS_SWAP_CHAIN
);
g_caps.maxTextureSize = 16384;
g_caps.maxFBAttachments = bx::uint32_min(16, BGFX_CONFIG_MAX_FRAME_BUFFER_ATTACHMENTS);
for (uint32_t ii = 0; ii < TextureFormat::Count; ++ii)
{
uint8_t support = BGFX_CAPS_FORMAT_TEXTURE_NONE;
const DXGI_FORMAT fmt = isDepth(TextureFormat::Enum(ii) )
? s_textureFormat[ii].m_fmtDsv
: s_textureFormat[ii].m_fmt
;
const DXGI_FORMAT fmtSrgb = s_textureFormat[ii].m_fmtSrgb;
if (DXGI_FORMAT_UNKNOWN != fmt)
{
D3D12_FEATURE_DATA_FORMAT_SUPPORT data;
data.Format = fmt;
hr = m_device->CheckFeatureSupport(D3D12_FEATURE_FORMAT_SUPPORT, &data, sizeof(data) );
if (SUCCEEDED(hr) )
{
support |= 0 != (data.Support1 & (0
| D3D12_FORMAT_SUPPORT1_TEXTURE2D
| D3D12_FORMAT_SUPPORT1_TEXTURE3D
| D3D12_FORMAT_SUPPORT1_TEXTURECUBE
) )
? BGFX_CAPS_FORMAT_TEXTURE_COLOR
: BGFX_CAPS_FORMAT_TEXTURE_NONE
;
support |= 0 != (data.Support1 & (0
| D3D12_FORMAT_SUPPORT1_BUFFER
| D3D12_FORMAT_SUPPORT1_IA_VERTEX_BUFFER
| D3D12_FORMAT_SUPPORT1_IA_INDEX_BUFFER
) )
? BGFX_CAPS_FORMAT_TEXTURE_VERTEX
: BGFX_CAPS_FORMAT_TEXTURE_NONE
;
support |= 0 != (data.Support1 & (0
| D3D12_FORMAT_SUPPORT1_SHADER_LOAD
) )
? BGFX_CAPS_FORMAT_TEXTURE_IMAGE
: BGFX_CAPS_FORMAT_TEXTURE_NONE
;
support |= 0 != (data.Support1 & (0
| D3D12_FORMAT_SUPPORT1_RENDER_TARGET
| D3D12_FORMAT_SUPPORT1_DEPTH_STENCIL
) )
? BGFX_CAPS_FORMAT_TEXTURE_FRAMEBUFFER
: BGFX_CAPS_FORMAT_TEXTURE_NONE
;
}
else
{
BX_TRACE("CheckFeatureSupport failed with %x for format %s.", hr, getName(TextureFormat::Enum(ii) ) );
}
if (0 != (support & BGFX_CAPS_FORMAT_TEXTURE_IMAGE) )
{
// clear image flag for additional testing
support &= ~BGFX_CAPS_FORMAT_TEXTURE_IMAGE;
data.Format = s_textureFormat[ii].m_fmt;
hr = m_device->CheckFeatureSupport(D3D12_FEATURE_FORMAT_SUPPORT, &data, sizeof(data) );
if (SUCCEEDED(hr) )
{
support |= 0 != (data.Support2 & (0
| D3D12_FORMAT_SUPPORT2_UAV_TYPED_LOAD
| D3D12_FORMAT_SUPPORT2_UAV_TYPED_STORE
) )
? BGFX_CAPS_FORMAT_TEXTURE_IMAGE
: BGFX_CAPS_FORMAT_TEXTURE_NONE
;
}
}
}
if (DXGI_FORMAT_UNKNOWN != fmtSrgb)
{
struct D3D11_FEATURE_DATA_FORMAT_SUPPORT
{
DXGI_FORMAT InFormat;
UINT OutFormatSupport;
};
D3D12_FEATURE_DATA_FORMAT_SUPPORT data;
data.Format = fmtSrgb;
hr = m_device->CheckFeatureSupport(D3D12_FEATURE_FORMAT_SUPPORT, &data, sizeof(data) );
if (SUCCEEDED(hr) )
{
support |= 0 != (data.Support1 & (0
| D3D12_FORMAT_SUPPORT1_TEXTURE2D
| D3D12_FORMAT_SUPPORT1_TEXTURE3D
| D3D12_FORMAT_SUPPORT1_TEXTURECUBE
) )
? BGFX_CAPS_FORMAT_TEXTURE_COLOR_SRGB
: BGFX_CAPS_FORMAT_TEXTURE_NONE
;
}
else
{
BX_TRACE("CheckFeatureSupport failed with %x for sRGB format %s.", hr, getName(TextureFormat::Enum(ii) ) );
}
}
g_caps.formats[ii] = support;
}
postReset();
}
return true;
error:
switch (errorState)
{
default:
case 4:
m_cmd.shutdown();
DX_RELEASE(m_device, 0);
case 3:
DX_RELEASE(m_factory, 0);
#if USE_D3D12_DYNAMIC_LIB
case 2:
bx::dlclose(m_dxgidll);
case 1:
bx::dlclose(m_d3d12dll);
#endif // USE_D3D12_DYNAMIC_LIB
case 0:
break;
}
return false;
}
void shutdown()
{
preReset();
m_samplerAllocator.destroy();
for (uint32_t ii = 0; ii < BX_COUNTOF(m_scratchBuffer); ++ii)
{
m_scratchBuffer[ii].destroy();
}
m_pipelineStateCache.invalidate();
for (uint32_t ii = 0; ii < BX_COUNTOF(m_indexBuffers); ++ii)
{
m_indexBuffers[ii].destroy();
}
for (uint32_t ii = 0; ii < BX_COUNTOF(m_vertexBuffers); ++ii)
{
m_vertexBuffers[ii].destroy();
}
for (uint32_t ii = 0; ii < BX_COUNTOF(m_shaders); ++ii)
{
m_shaders[ii].destroy();
}
for (uint32_t ii = 0; ii < BX_COUNTOF(m_textures); ++ii)
{
m_textures[ii].destroy();
}
DX_RELEASE(m_rtvDescriptorHeap, 0);
DX_RELEASE(m_dsvDescriptorHeap, 0);
DX_RELEASE(m_commandSignature[0], 0);
DX_RELEASE(m_commandSignature[1], 0);
DX_RELEASE(m_rootSignature, 0);
DX_RELEASE(m_swapChain, 0);
m_cmd.shutdown();
DX_RELEASE(m_device, 0);
DX_RELEASE(m_factory, 0);
#if USE_D3D12_DYNAMIC_LIB
bx::dlclose(m_d3d12dll);
bx::dlclose(m_dxgidll);
#endif // USE_D3D12_DYNAMIC_LIB
}
RendererType::Enum getRendererType() const BX_OVERRIDE
{
return RendererType::Direct3D12;
}
const char* getRendererName() const BX_OVERRIDE
{
return BGFX_RENDERER_DIRECT3D12_NAME;
}
static bool isLost(HRESULT _hr)
{
return DXGI_ERROR_DEVICE_REMOVED == _hr
|| DXGI_ERROR_DEVICE_HUNG == _hr
|| DXGI_ERROR_DEVICE_RESET == _hr
|| DXGI_ERROR_DRIVER_INTERNAL_ERROR == _hr
|| DXGI_ERROR_NOT_CURRENTLY_AVAILABLE == _hr
;
}
void flip(HMD& /*_hmd*/) BX_OVERRIDE
{
if (NULL != m_swapChain)
{
int64_t start = bx::getHPCounter();
HRESULT hr = 0;
uint32_t syncInterval = !!(m_flags & BGFX_RESET_VSYNC);
for (uint32_t ii = 1, num = m_numWindows; ii < num && SUCCEEDED(hr); ++ii)
{
hr = m_frameBuffers[m_windows[ii].idx].m_swapChain->Present(syncInterval, 0);
}
if (SUCCEEDED(hr) )
{
m_cmd.finish(m_backBufferColorFence[(m_backBufferColorIdx-1) % m_scd.BufferCount]);
hr = m_swapChain->Present(syncInterval, 0);
}
int64_t now = bx::getHPCounter();
m_presentElapsed = now - start;
if (FAILED(hr)
&& isLost(hr) )
{
++m_lost;
BGFX_FATAL(10 > m_lost, bgfx::Fatal::DeviceLost, "Device is lost. FAILED 0x%08x", hr);
}
else
{
m_lost = 0;
}
}
}
void createIndexBuffer(IndexBufferHandle _handle, Memory* _mem, uint16_t _flags) BX_OVERRIDE
{
m_indexBuffers[_handle.idx].create(_mem->size, _mem->data, _flags, false);
}
void destroyIndexBuffer(IndexBufferHandle _handle) BX_OVERRIDE
{
m_indexBuffers[_handle.idx].destroy();
}
void createVertexDecl(VertexDeclHandle _handle, const VertexDecl& _decl) BX_OVERRIDE
{
VertexDecl& decl = m_vertexDecls[_handle.idx];
memcpy(&decl, &_decl, sizeof(VertexDecl) );
dump(decl);
}
void destroyVertexDecl(VertexDeclHandle /*_handle*/) BX_OVERRIDE
{
}
void createVertexBuffer(VertexBufferHandle _handle, Memory* _mem, VertexDeclHandle _declHandle, uint16_t _flags) BX_OVERRIDE
{
m_vertexBuffers[_handle.idx].create(_mem->size, _mem->data, _declHandle, _flags);
}
void destroyVertexBuffer(VertexBufferHandle _handle) BX_OVERRIDE
{
m_vertexBuffers[_handle.idx].destroy();
}
void createDynamicIndexBuffer(IndexBufferHandle _handle, uint32_t _size, uint16_t _flags) BX_OVERRIDE
{
m_indexBuffers[_handle.idx].create(_size, NULL, _flags, false);
}
void updateDynamicIndexBuffer(IndexBufferHandle _handle, uint32_t _offset, uint32_t _size, Memory* _mem) BX_OVERRIDE
{
m_indexBuffers[_handle.idx].update(m_commandList, _offset, bx::uint32_min(_size, _mem->size), _mem->data);
}
void destroyDynamicIndexBuffer(IndexBufferHandle _handle) BX_OVERRIDE
{
m_indexBuffers[_handle.idx].destroy();
}
void createDynamicVertexBuffer(VertexBufferHandle _handle, uint32_t _size, uint16_t _flags) BX_OVERRIDE
{
VertexDeclHandle decl = BGFX_INVALID_HANDLE;
m_vertexBuffers[_handle.idx].create(_size, NULL, decl, _flags);
}
void updateDynamicVertexBuffer(VertexBufferHandle _handle, uint32_t _offset, uint32_t _size, Memory* _mem) BX_OVERRIDE
{
m_vertexBuffers[_handle.idx].update(m_commandList, _offset, bx::uint32_min(_size, _mem->size), _mem->data);
}
void destroyDynamicVertexBuffer(VertexBufferHandle _handle) BX_OVERRIDE
{
m_vertexBuffers[_handle.idx].destroy();
}
void createShader(ShaderHandle _handle, Memory* _mem) BX_OVERRIDE
{
m_shaders[_handle.idx].create(_mem);
}
void destroyShader(ShaderHandle _handle) BX_OVERRIDE
{
m_shaders[_handle.idx].destroy();
}
void createProgram(ProgramHandle _handle, ShaderHandle _vsh, ShaderHandle _fsh) BX_OVERRIDE
{
m_program[_handle.idx].create(&m_shaders[_vsh.idx], isValid(_fsh) ? &m_shaders[_fsh.idx] : NULL);
}
void destroyProgram(ProgramHandle _handle) BX_OVERRIDE
{
m_program[_handle.idx].destroy();
}
void createTexture(TextureHandle _handle, Memory* _mem, uint32_t _flags, uint8_t _skip) BX_OVERRIDE
{
m_textures[_handle.idx].create(_mem, _flags, _skip);
}
void updateTextureBegin(TextureHandle /*_handle*/, uint8_t /*_side*/, uint8_t /*_mip*/) BX_OVERRIDE
{
}
void updateTexture(TextureHandle _handle, uint8_t _side, uint8_t _mip, const Rect& _rect, uint16_t _z, uint16_t _depth, uint16_t _pitch, const Memory* _mem) BX_OVERRIDE
{
m_textures[_handle.idx].update(m_commandList, _side, _mip, _rect, _z, _depth, _pitch, _mem);
}
void updateTextureEnd() BX_OVERRIDE
{
}
void resizeTexture(TextureHandle _handle, uint16_t _width, uint16_t _height) BX_OVERRIDE
{
TextureD3D12& texture = m_textures[_handle.idx];
uint32_t size = sizeof(uint32_t) + sizeof(TextureCreate);
const Memory* mem = alloc(size);
bx::StaticMemoryBlockWriter writer(mem->data, mem->size);
uint32_t magic = BGFX_CHUNK_MAGIC_TEX;
bx::write(&writer, magic);
TextureCreate tc;
tc.m_flags = texture.m_flags;
tc.m_width = _width;
tc.m_height = _height;
tc.m_sides = 0;
tc.m_depth = 0;
tc.m_numMips = 1;
tc.m_format = texture.m_requestedFormat;
tc.m_cubeMap = false;
tc.m_mem = NULL;
bx::write(&writer, tc);
texture.destroy();
texture.create(mem, tc.m_flags, 0);
release(mem);
}
void destroyTexture(TextureHandle _handle) BX_OVERRIDE
{
m_textures[_handle.idx].destroy();
}
void createFrameBuffer(FrameBufferHandle _handle, uint8_t _num, const TextureHandle* _textureHandles) BX_OVERRIDE
{
m_frameBuffers[_handle.idx].create(_num, _textureHandles);
}
void createFrameBuffer(FrameBufferHandle _handle, void* _nwh, uint32_t _width, uint32_t _height, TextureFormat::Enum _depthFormat) BX_OVERRIDE
{
uint16_t denseIdx = m_numWindows++;
m_windows[denseIdx] = _handle;
m_frameBuffers[_handle.idx].create(denseIdx, _nwh, _width, _height, _depthFormat);
}
void destroyFrameBuffer(FrameBufferHandle _handle) BX_OVERRIDE
{
uint16_t denseIdx = m_frameBuffers[_handle.idx].destroy();
if (UINT16_MAX != denseIdx)
{
--m_numWindows;
if (m_numWindows > 1)
{
FrameBufferHandle handle = m_windows[m_numWindows];
m_windows[denseIdx] = handle;
m_frameBuffers[handle.idx].m_denseIdx = denseIdx;
}
}
}
void createUniform(UniformHandle _handle, UniformType::Enum _type, uint16_t _num, const char* _name) BX_OVERRIDE
{
if (NULL != m_uniforms[_handle.idx])
{
BX_FREE(g_allocator, m_uniforms[_handle.idx]);
}
uint32_t size = BX_ALIGN_16(g_uniformTypeSize[_type] * _num);
void* data = BX_ALLOC(g_allocator, size);
memset(data, 0, size);
m_uniforms[_handle.idx] = data;
m_uniformReg.add(_handle, _name, data);
}
void destroyUniform(UniformHandle _handle) BX_OVERRIDE
{
BX_FREE(g_allocator, m_uniforms[_handle.idx]);
m_uniforms[_handle.idx] = NULL;
}
void saveScreenShot(const char* _filePath) BX_OVERRIDE
{
uint32_t idx = (m_backBufferColorIdx-1) % m_scd.BufferCount;
m_cmd.finish(m_backBufferColorFence[idx]);
ID3D12Resource* backBuffer = m_backBufferColor[idx];
D3D12_RESOURCE_DESC desc = backBuffer->GetDesc();
const uint32_t width = (uint32_t)desc.Width;
const uint32_t height = (uint32_t)desc.Height;
D3D12_PLACED_SUBRESOURCE_FOOTPRINT layout;
uint32_t numRows;
uint64_t total;
uint64_t pitch;
m_device->GetCopyableFootprints(&desc
, 0
, 1
, 0
, &layout
, &numRows
, &pitch
, &total
);
ID3D12Resource* readback = createCommittedResource(m_device, HeapProperty::ReadBack, total);
D3D12_BOX box;
box.left = 0;
box.top = 0;
box.right = width;
box.bottom = height;
box.front = 0;
box.back = 1;
setResourceBarrier(m_commandList, backBuffer, D3D12_RESOURCE_STATE_PRESENT, D3D12_RESOURCE_STATE_COPY_SOURCE);
D3D12_TEXTURE_COPY_LOCATION dst = { readback, D3D12_TEXTURE_COPY_TYPE_PLACED_FOOTPRINT, layout };
D3D12_TEXTURE_COPY_LOCATION src = { backBuffer, D3D12_TEXTURE_COPY_TYPE_SUBRESOURCE_INDEX, { 0 } };
m_commandList->CopyTextureRegion(&dst, 0, 0, 0, &src, &box);
setResourceBarrier(m_commandList, backBuffer, D3D12_RESOURCE_STATE_COPY_SOURCE, D3D12_RESOURCE_STATE_PRESENT);
finish();
m_commandList = m_cmd.alloc();
void* data;
readback->Map(0, NULL, (void**)&data);
imageSwizzleBgra8(width
, height
, (uint32_t)pitch
, data
, data
);
g_callback->screenShot(_filePath
, width
, height
, (uint32_t)pitch
, data
, (uint32_t)total
, false
);
readback->Unmap(0, NULL);
DX_RELEASE(readback, 0);
}
void updateViewName(uint8_t /*_id*/, const char* /*_name*/) BX_OVERRIDE
{
}
void updateUniform(uint16_t _loc, const void* _data, uint32_t _size) BX_OVERRIDE
{
memcpy(m_uniforms[_loc], _data, _size);
}
void setMarker(const char* /*_marker*/, uint32_t /*_size*/) BX_OVERRIDE
{
}
void submit(Frame* _render, ClearQuad& _clearQuad, TextVideoMemBlitter& _textVideoMemBlitter) BX_OVERRIDE;
void blitSetup(TextVideoMemBlitter& _blitter) BX_OVERRIDE
{
const uint32_t width = m_scd.BufferDesc.Width;
const uint32_t height = m_scd.BufferDesc.Height;
FrameBufferHandle fbh = BGFX_INVALID_HANDLE;
setFrameBuffer(fbh, false);
D3D12_VIEWPORT vp;
vp.TopLeftX = 0;
vp.TopLeftY = 0;
vp.Width = (float)width;
vp.Height = (float)height;
vp.MinDepth = 0.0f;
vp.MaxDepth = 1.0f;
m_commandList->RSSetViewports(1, &vp);
const uint64_t state = 0
| BGFX_STATE_RGB_WRITE
| BGFX_STATE_ALPHA_WRITE
| BGFX_STATE_DEPTH_TEST_ALWAYS
;
m_currentProgram = &m_program[0];
ID3D12PipelineState* pso = getPipelineState(state
, packStencil(BGFX_STENCIL_DEFAULT, BGFX_STENCIL_DEFAULT)
, _blitter.m_vb->decl.idx
, _blitter.m_program.idx
, 0
);
m_commandList->SetPipelineState(pso);
m_commandList->SetGraphicsRootSignature(m_rootSignature);
float proj[16];
bx::mtxOrtho(proj, 0.0f, (float)width, (float)height, 0.0f, 0.0f, 1000.0f);
PredefinedUniform& predefined = m_program[_blitter.m_program.idx].m_predefined[0];
uint8_t flags = predefined.m_type;
setShaderUniform(flags, predefined.m_loc, proj, 4);
D3D12_GPU_VIRTUAL_ADDRESS gpuAddress;
commitShaderConstants(gpuAddress);
ScratchBufferD3D12& scratchBuffer = m_scratchBuffer[m_backBufferColorIdx];
ID3D12DescriptorHeap* heaps[] =
{
m_samplerAllocator.getHeap(),
scratchBuffer.getHeap(),
};
m_commandList->SetDescriptorHeaps(BX_COUNTOF(heaps), heaps);
m_commandList->SetGraphicsRootConstantBufferView(Rdt::CBV, gpuAddress);
TextureD3D12& texture = m_textures[_blitter.m_texture.idx];
uint32_t samplerFlags[BGFX_CONFIG_MAX_TEXTURE_SAMPLERS] = { texture.m_flags & BGFX_TEXTURE_SAMPLER_BITS_MASK };
uint16_t samplerStateIdx = getSamplerState(samplerFlags);
m_commandList->SetGraphicsRootDescriptorTable(Rdt::Sampler, m_samplerAllocator.get(samplerStateIdx) );
D3D12_GPU_DESCRIPTOR_HANDLE srvHandle;
scratchBuffer.alloc(srvHandle, texture);
m_commandList->SetGraphicsRootDescriptorTable(Rdt::SRV, srvHandle);
VertexBufferD3D12& vb = m_vertexBuffers[_blitter.m_vb->handle.idx];
const VertexDecl& vertexDecl = m_vertexDecls[_blitter.m_vb->decl.idx];
D3D12_VERTEX_BUFFER_VIEW viewDesc;
viewDesc.BufferLocation = vb.m_gpuVA;
viewDesc.StrideInBytes = vertexDecl.m_stride;
viewDesc.SizeInBytes = vb.m_size;
m_commandList->IASetVertexBuffers(0, 1, &viewDesc);
const BufferD3D12& ib = m_indexBuffers[_blitter.m_ib->handle.idx];
D3D12_INDEX_BUFFER_VIEW ibv;
ibv.Format = DXGI_FORMAT_R16_UINT;
ibv.BufferLocation = ib.m_gpuVA;
ibv.SizeInBytes = ib.m_size;
m_commandList->IASetIndexBuffer(&ibv);
m_commandList->IASetPrimitiveTopology(D3D_PRIMITIVE_TOPOLOGY_TRIANGLELIST);
}
void blitRender(TextVideoMemBlitter& _blitter, uint32_t _numIndices) BX_OVERRIDE
{
const uint32_t numVertices = _numIndices*4/6;
if (0 < numVertices)
{
m_indexBuffers [_blitter.m_ib->handle.idx].update(m_commandList, 0, _numIndices*2, _blitter.m_ib->data);
m_vertexBuffers[_blitter.m_vb->handle.idx].update(m_commandList, 0, numVertices*_blitter.m_decl.m_stride, _blitter.m_vb->data, true);
m_commandList->DrawIndexedInstanced(_numIndices
, 1
, 0
, 0
, 0
);
}
}
void preReset()
{
finishAll();
for (uint32_t ii = 0, num = m_scd.BufferCount; ii < num; ++ii)
{
DX_RELEASE(m_backBufferColor[ii], num-1-ii);
}
DX_RELEASE(m_backBufferDepthStencil, 0);
for (uint32_t ii = 0; ii < BX_COUNTOF(m_frameBuffers); ++ii)
{
m_frameBuffers[ii].preReset();
}
invalidateCache();
// capturePreReset();
}
void postReset()
{
memset(m_backBufferColorFence, 0, sizeof(m_backBufferColorFence) );
uint32_t rtvDescriptorSize = m_device->GetDescriptorHandleIncrementSize(D3D12_DESCRIPTOR_HEAP_TYPE_RTV);
for (uint32_t ii = 0, num = m_scd.BufferCount; ii < num; ++ii)
{
D3D12_CPU_DESCRIPTOR_HANDLE handle = m_rtvDescriptorHeap->GetCPUDescriptorHandleForHeapStart();
handle.ptr += ii * rtvDescriptorSize;
DX_CHECK(m_swapChain->GetBuffer(ii
, __uuidof(ID3D12Resource)
, (void**)&m_backBufferColor[ii]
) );
m_device->CreateRenderTargetView(m_backBufferColor[ii], NULL, handle);
}
D3D12_RESOURCE_DESC resourceDesc;
resourceDesc.Dimension = D3D12_RESOURCE_DIMENSION_TEXTURE2D;
resourceDesc.Alignment = 0;
resourceDesc.Width = bx::uint32_max(m_resolution.m_width, 1);
resourceDesc.Height = bx::uint32_max(m_resolution.m_height, 1);
resourceDesc.DepthOrArraySize = 1;
resourceDesc.MipLevels = 0;
resourceDesc.Format = DXGI_FORMAT_D24_UNORM_S8_UINT;
resourceDesc.SampleDesc.Count = 1;
resourceDesc.SampleDesc.Quality = 0;
resourceDesc.Layout = D3D12_TEXTURE_LAYOUT_UNKNOWN;
resourceDesc.Flags = D3D12_RESOURCE_FLAG_ALLOW_DEPTH_STENCIL;
D3D12_CLEAR_VALUE clearValue;
clearValue.Format = resourceDesc.Format;
clearValue.DepthStencil.Depth = 1.0f;
clearValue.DepthStencil.Stencil = 0;
m_backBufferDepthStencil = createCommittedResource(m_device, HeapProperty::Default, &resourceDesc, &clearValue);
D3D12_DEPTH_STENCIL_VIEW_DESC dsvDesc;
ZeroMemory(&dsvDesc, sizeof(dsvDesc) );
dsvDesc.Format = resourceDesc.Format;
dsvDesc.ViewDimension = D3D12_DSV_DIMENSION_TEXTURE2D;
dsvDesc.Flags = D3D12_DSV_FLAGS(0)
// | D3D12_DSV_FLAG_READ_ONLY_DEPTH
// | D3D12_DSV_FLAG_READ_ONLY_DEPTH
;
m_device->CreateDepthStencilView(m_backBufferDepthStencil
, &dsvDesc
, m_dsvDescriptorHeap->GetCPUDescriptorHandleForHeapStart()
);
for (uint32_t ii = 0; ii < BX_COUNTOF(m_frameBuffers); ++ii)
{
m_frameBuffers[ii].postReset();
}
m_commandList = m_cmd.alloc();
// capturePostReset();
}
void invalidateCache()
{
m_pipelineStateCache.invalidate();
m_samplerStateCache.invalidate();
m_samplerAllocator.reset();
}
void updateMsaa()
{
for (uint32_t ii = 1, last = 0; ii < BX_COUNTOF(s_msaa); ++ii)
{
uint32_t msaa = s_checkMsaa[ii];
D3D12_FEATURE_DATA_MULTISAMPLE_QUALITY_LEVELS data;
memset(&data, 0, sizeof(msaa) );
data.Format = m_scd.BufferDesc.Format;
data.SampleCount = msaa;
data.Flags = D3D12_MULTISAMPLE_QUALITY_LEVELS_FLAG_NONE;
HRESULT hr = m_device->CheckFeatureSupport(D3D12_FEATURE_MULTISAMPLE_QUALITY_LEVELS, &data, sizeof(data) );
data.NumQualityLevels = 0;
if (SUCCEEDED(hr)
&& 0 < data.NumQualityLevels)
{
s_msaa[ii].Count = data.SampleCount;
s_msaa[ii].Quality = data.NumQualityLevels - 1;
last = ii;
}
else
{
s_msaa[ii] = s_msaa[last];
}
}
}
void updateResolution(const Resolution& _resolution)
{
if ( (uint32_t)m_scd.BufferDesc.Width != _resolution.m_width
|| (uint32_t)m_scd.BufferDesc.Height != _resolution.m_height
|| m_flags != _resolution.m_flags)
{
bool resize = (m_flags&BGFX_RESET_MSAA_MASK) == (_resolution.m_flags&BGFX_RESET_MSAA_MASK);
m_flags = _resolution.m_flags;
m_textVideoMem.resize(false, _resolution.m_width, _resolution.m_height);
m_textVideoMem.clear();
m_resolution = _resolution;
m_scd.BufferDesc.Width = _resolution.m_width;
m_scd.BufferDesc.Height = _resolution.m_height;
preReset();
if (resize)
{
uint32_t nodeMask[] = { 1, 1, 1, 1 };
BX_STATIC_ASSERT(BX_COUNTOF(m_backBufferColor) == BX_COUNTOF(nodeMask) );
IUnknown* presentQueue[] ={ m_cmd.m_commandQueue, m_cmd.m_commandQueue, m_cmd.m_commandQueue, m_cmd.m_commandQueue };
BX_STATIC_ASSERT(BX_COUNTOF(m_backBufferColor) == BX_COUNTOF(presentQueue) );
DX_CHECK(m_swapChain->ResizeBuffers1(m_scd.BufferCount
, m_scd.BufferDesc.Width
, m_scd.BufferDesc.Height
, m_scd.BufferDesc.Format
, m_scd.Flags
, nodeMask
, presentQueue
) );
}
else
{
updateMsaa();
m_scd.SampleDesc = s_msaa[(m_flags&BGFX_RESET_MSAA_MASK)>>BGFX_RESET_MSAA_SHIFT];
DX_RELEASE(m_swapChain, 0);
HRESULT hr;
hr = m_factory->CreateSwapChain(m_cmd.m_commandQueue
, &m_scd
, reinterpret_cast<IDXGISwapChain**>(&m_swapChain)
);
BGFX_FATAL(SUCCEEDED(hr), bgfx::Fatal::UnableToInitialize, "Failed to create swap chain.");
}
postReset();
}
}
void setShaderUniform(uint8_t _flags, uint16_t _regIndex, const void* _val, uint16_t _numRegs)
{
if (_flags&BGFX_UNIFORM_FRAGMENTBIT)
{
memcpy(&m_fsScratch[_regIndex], _val, _numRegs*16);
m_fsChanges += _numRegs;
}
else
{
memcpy(&m_vsScratch[_regIndex], _val, _numRegs*16);
m_vsChanges += _numRegs;
}
}
void setShaderUniform4f(uint8_t _flags, uint16_t _regIndex, const void* _val, uint16_t _numRegs)
{
setShaderUniform(_flags, _regIndex, _val, _numRegs);
}
void setShaderUniform4x4f(uint8_t _flags, uint16_t _regIndex, const void* _val, uint16_t _numRegs)
{
setShaderUniform(_flags, _regIndex, _val, _numRegs);
}
void commitShaderConstants(D3D12_GPU_VIRTUAL_ADDRESS& _gpuAddress)
{
uint32_t total = bx::strideAlign(0
+ m_currentProgram->m_vsh->m_size
+ (NULL != m_currentProgram->m_fsh ? m_currentProgram->m_fsh->m_size : 0)
, D3D12_CONSTANT_BUFFER_DATA_PLACEMENT_ALIGNMENT
);
uint8_t* data = (uint8_t*)m_scratchBuffer[m_backBufferColorIdx].alloc(_gpuAddress, total);
{
uint32_t size = m_currentProgram->m_vsh->m_size;
memcpy(data, m_vsScratch, size);
data += size;
m_vsChanges = 0;
}
if (NULL != m_currentProgram->m_fsh)
{
memcpy(data, m_fsScratch, m_currentProgram->m_fsh->m_size);
m_fsChanges = 0;
}
}
void setFrameBuffer(FrameBufferHandle _fbh, bool _msaa = true)
{
if (isValid(m_fbh)
&& m_fbh.idx != _fbh.idx)
{
const FrameBufferD3D12& frameBuffer = m_frameBuffers[m_fbh.idx];
for (uint8_t ii = 0, num = frameBuffer.m_num; ii < num; ++ii)
{
TextureD3D12& texture = m_textures[frameBuffer.m_texture[ii].idx];
texture.setState(m_commandList, D3D12_RESOURCE_STATE_PIXEL_SHADER_RESOURCE);
}
if (isValid(frameBuffer.m_depth) )
{
TextureD3D12& texture = m_textures[frameBuffer.m_depth.idx];
const bool bufferOnly = 0 != (texture.m_flags&BGFX_TEXTURE_RT_BUFFER_ONLY);
if (!bufferOnly)
{
texture.setState(m_commandList, D3D12_RESOURCE_STATE_DEPTH_READ);
}
}
}
if (!isValid(_fbh) )
{
m_rtvHandle = m_rtvDescriptorHeap->GetCPUDescriptorHandleForHeapStart();
uint32_t rtvDescriptorSize = m_device->GetDescriptorHandleIncrementSize(D3D12_DESCRIPTOR_HEAP_TYPE_RTV);
m_rtvHandle.ptr += m_backBufferColorIdx * rtvDescriptorSize;
m_dsvHandle = m_dsvDescriptorHeap->GetCPUDescriptorHandleForHeapStart();
m_currentColor = &m_rtvHandle;
m_currentDepthStencil = &m_dsvHandle;
m_commandList->OMSetRenderTargets(1, m_currentColor, false, m_currentDepthStencil);
}
else
{
const FrameBufferD3D12& frameBuffer = m_frameBuffers[_fbh.idx];
if (0 < frameBuffer.m_num)
{
D3D12_CPU_DESCRIPTOR_HANDLE rtvDescriptor = m_rtvDescriptorHeap->GetCPUDescriptorHandleForHeapStart();
uint32_t rtvDescriptorSize = m_device->GetDescriptorHandleIncrementSize(D3D12_DESCRIPTOR_HEAP_TYPE_RTV);
m_rtvHandle.ptr = rtvDescriptor.ptr + (BX_COUNTOF(m_backBufferColor) + _fbh.idx * BGFX_CONFIG_MAX_FRAME_BUFFER_ATTACHMENTS) * rtvDescriptorSize;
m_currentColor = &m_rtvHandle;
}
else
{
m_currentColor = NULL;
}
if (isValid(frameBuffer.m_depth) )
{
D3D12_CPU_DESCRIPTOR_HANDLE dsvDescriptor = m_dsvDescriptorHeap->GetCPUDescriptorHandleForHeapStart();
uint32_t dsvDescriptorSize = m_device->GetDescriptorHandleIncrementSize(D3D12_DESCRIPTOR_HEAP_TYPE_DSV);
m_dsvHandle.ptr = dsvDescriptor.ptr + (1 + _fbh.idx) * dsvDescriptorSize;
m_currentDepthStencil = &m_dsvHandle;
}
else
{
m_currentDepthStencil = NULL;
}
for (uint8_t ii = 0, num = frameBuffer.m_num; ii < num; ++ii)
{
TextureD3D12& texture = m_textures[frameBuffer.m_texture[ii].idx];
texture.setState(m_commandList, D3D12_RESOURCE_STATE_RENDER_TARGET);
}
if (isValid(frameBuffer.m_depth) )
{
TextureD3D12& texture = m_textures[frameBuffer.m_depth.idx];
texture.setState(m_commandList, D3D12_RESOURCE_STATE_DEPTH_WRITE);
}
m_commandList->OMSetRenderTargets(frameBuffer.m_num
, m_currentColor
, true //NULL == m_currentDepthStencil
, m_currentDepthStencil
);
}
m_fbh = _fbh;
m_rtMsaa = _msaa;
}
void setBlendState(D3D12_BLEND_DESC& desc, uint64_t _state, uint32_t _rgba = 0)
{
memset(&desc, 0, sizeof(desc) );
desc.IndependentBlendEnable = !!(BGFX_STATE_BLEND_INDEPENDENT & _state);
D3D12_RENDER_TARGET_BLEND_DESC* drt = &desc.RenderTarget[0];
drt->BlendEnable = !!(BGFX_STATE_BLEND_MASK & _state);
{
const uint32_t blend = uint32_t( (_state & BGFX_STATE_BLEND_MASK ) >> BGFX_STATE_BLEND_SHIFT);
const uint32_t equation = uint32_t( (_state & BGFX_STATE_BLEND_EQUATION_MASK) >> BGFX_STATE_BLEND_EQUATION_SHIFT);
const uint32_t srcRGB = (blend ) & 0xf;
const uint32_t dstRGB = (blend >> 4) & 0xf;
const uint32_t srcA = (blend >> 8) & 0xf;
const uint32_t dstA = (blend >> 12) & 0xf;
const uint32_t equRGB = (equation ) & 0x7;
const uint32_t equA = (equation >> 3) & 0x7;
drt->SrcBlend = s_blendFactor[srcRGB][0];
drt->DestBlend = s_blendFactor[dstRGB][0];
drt->BlendOp = s_blendEquation[equRGB];
drt->SrcBlendAlpha = s_blendFactor[srcA][1];
drt->DestBlendAlpha = s_blendFactor[dstA][1];
drt->BlendOpAlpha = s_blendEquation[equA];
}
uint32_t writeMask = (_state & BGFX_STATE_ALPHA_WRITE)
? D3D12_COLOR_WRITE_ENABLE_ALPHA
: 0
;
writeMask |= (_state & BGFX_STATE_RGB_WRITE)
? D3D12_COLOR_WRITE_ENABLE_RED
| D3D12_COLOR_WRITE_ENABLE_GREEN
| D3D12_COLOR_WRITE_ENABLE_BLUE
: 0
;
drt->RenderTargetWriteMask = writeMask;
if (desc.IndependentBlendEnable)
{
for (uint32_t ii = 1, rgba = _rgba; ii < BGFX_CONFIG_MAX_FRAME_BUFFER_ATTACHMENTS; ++ii, rgba >>= 11)
{
drt = &desc.RenderTarget[ii];
drt->BlendEnable = 0 != (rgba & 0x7ff);
const uint32_t src = (rgba ) & 0xf;
const uint32_t dst = (rgba >> 4) & 0xf;
const uint32_t equation = (rgba >> 8) & 0x7;
drt->SrcBlend = s_blendFactor[src][0];
drt->DestBlend = s_blendFactor[dst][0];
drt->BlendOp = s_blendEquation[equation];
drt->SrcBlendAlpha = s_blendFactor[src][1];
drt->DestBlendAlpha = s_blendFactor[dst][1];
drt->BlendOpAlpha = s_blendEquation[equation];
drt->RenderTargetWriteMask = writeMask;
}
}
else
{
for (uint32_t ii = 1; ii < BGFX_CONFIG_MAX_FRAME_BUFFER_ATTACHMENTS; ++ii)
{
memcpy(&desc.RenderTarget[ii], drt, sizeof(D3D12_RENDER_TARGET_BLEND_DESC) );
}
}
}
void setRasterizerState(D3D12_RASTERIZER_DESC& desc, uint64_t _state, bool _wireframe = false)
{
const uint32_t cull = (_state&BGFX_STATE_CULL_MASK) >> BGFX_STATE_CULL_SHIFT;
desc.FillMode = _wireframe
? D3D12_FILL_MODE_WIREFRAME
: D3D12_FILL_MODE_SOLID
;
desc.CullMode = s_cullMode[cull];
desc.FrontCounterClockwise = false;
desc.DepthBias = 0;
desc.DepthBiasClamp = 0.0f;
desc.SlopeScaledDepthBias = 0.0f;
desc.DepthClipEnable = false;
desc.MultisampleEnable = !!(_state&BGFX_STATE_MSAA);
desc.AntialiasedLineEnable = false;
desc.ForcedSampleCount = 0;
desc.ConservativeRaster = D3D12_CONSERVATIVE_RASTERIZATION_MODE_OFF;
}
void setDepthStencilState(D3D12_DEPTH_STENCIL_DESC& desc, uint64_t _state, uint64_t _stencil = 0)
{
const uint32_t fstencil = unpackStencil(0, _stencil);
memset(&desc, 0, sizeof(desc) );
uint32_t func = (_state&BGFX_STATE_DEPTH_TEST_MASK)>>BGFX_STATE_DEPTH_TEST_SHIFT;
desc.DepthEnable = 0 != func;
desc.DepthWriteMask = !!(BGFX_STATE_DEPTH_WRITE & _state)
? D3D12_DEPTH_WRITE_MASK_ALL
: D3D12_DEPTH_WRITE_MASK_ZERO
;
desc.DepthFunc = s_cmpFunc[func];
uint32_t bstencil = unpackStencil(1, _stencil);
uint32_t frontAndBack = bstencil != BGFX_STENCIL_NONE && bstencil != fstencil;
bstencil = frontAndBack ? bstencil : fstencil;
desc.StencilEnable = 0 != _stencil;
desc.StencilReadMask = (fstencil & BGFX_STENCIL_FUNC_RMASK_MASK) >> BGFX_STENCIL_FUNC_RMASK_SHIFT;
desc.StencilWriteMask = 0xff;
desc.FrontFace.StencilFailOp = s_stencilOp[(fstencil & BGFX_STENCIL_OP_FAIL_S_MASK) >> BGFX_STENCIL_OP_FAIL_S_SHIFT];
desc.FrontFace.StencilDepthFailOp = s_stencilOp[(fstencil & BGFX_STENCIL_OP_FAIL_Z_MASK) >> BGFX_STENCIL_OP_FAIL_Z_SHIFT];
desc.FrontFace.StencilPassOp = s_stencilOp[(fstencil & BGFX_STENCIL_OP_PASS_Z_MASK) >> BGFX_STENCIL_OP_PASS_Z_SHIFT];
desc.FrontFace.StencilFunc = s_cmpFunc[(fstencil & BGFX_STENCIL_TEST_MASK) >> BGFX_STENCIL_TEST_SHIFT];
desc.BackFace.StencilFailOp = s_stencilOp[(bstencil & BGFX_STENCIL_OP_FAIL_S_MASK) >> BGFX_STENCIL_OP_FAIL_S_SHIFT];
desc.BackFace.StencilDepthFailOp = s_stencilOp[(bstencil & BGFX_STENCIL_OP_FAIL_Z_MASK) >> BGFX_STENCIL_OP_FAIL_Z_SHIFT];
desc.BackFace.StencilPassOp = s_stencilOp[(bstencil & BGFX_STENCIL_OP_PASS_Z_MASK) >> BGFX_STENCIL_OP_PASS_Z_SHIFT];
desc.BackFace.StencilFunc = s_cmpFunc[(bstencil&BGFX_STENCIL_TEST_MASK) >> BGFX_STENCIL_TEST_SHIFT];
}
uint32_t setInputLayout(D3D12_INPUT_ELEMENT_DESC* _vertexElements, const VertexDecl& _vertexDecl, const ProgramD3D12& _program, uint8_t _numInstanceData)
{
VertexDecl decl;
memcpy(&decl, &_vertexDecl, sizeof(VertexDecl) );
const uint16_t* attrMask = _program.m_vsh->m_attrMask;
for (uint32_t ii = 0; ii < Attrib::Count; ++ii)
{
uint16_t mask = attrMask[ii];
uint16_t attr = (decl.m_attributes[ii] & mask);
decl.m_attributes[ii] = attr == 0 ? UINT16_MAX : attr == UINT16_MAX ? 0 : attr;
}
D3D12_INPUT_ELEMENT_DESC* elem = fillVertexDecl(_vertexElements, decl);
uint32_t num = uint32_t(elem-_vertexElements);
const D3D12_INPUT_ELEMENT_DESC inst = { "TEXCOORD", 0, DXGI_FORMAT_R32G32B32A32_FLOAT, 0, D3D12_APPEND_ALIGNED_ELEMENT, D3D12_INPUT_CLASSIFICATION_PER_INSTANCE_DATA, 1 };
for (uint32_t ii = 0; ii < _numInstanceData; ++ii)
{
uint32_t index = 7 - ii; // TEXCOORD7 = i_data0, TEXCOORD6 = i_data1, etc.
uint32_t jj;
D3D12_INPUT_ELEMENT_DESC* curr = _vertexElements;
for (jj = 0; jj < num; ++jj)
{
curr = &_vertexElements[jj];
if (0 == strcmp(curr->SemanticName, "TEXCOORD")
&& curr->SemanticIndex == index)
{
break;
}
}
if (jj == num)
{
curr = elem;
++elem;
}
memcpy(curr, &inst, sizeof(D3D12_INPUT_ELEMENT_DESC) );
curr->InputSlot = 1;
curr->SemanticIndex = index;
curr->AlignedByteOffset = ii*16;
}
return uint32_t(elem-_vertexElements);
}
static void patchCb0(DxbcInstruction& _instruction, void* _userData)
{
union { void* ptr; uint32_t offset; } cast = { _userData };
for (uint32_t ii = 0; ii < _instruction.numOperands; ++ii)
{
DxbcOperand& operand = _instruction.operand[ii];
if (DxbcOperandType::ConstantBuffer == operand.type)
{
if (DxbcOperandAddrMode::Imm32 == operand.addrMode[0]
&& 0 == operand.regIndex[0]
&& DxbcOperandAddrMode::Imm32 == operand.addrMode[1])
{
operand.regIndex[1] += cast.offset;
}
}
}
}
ID3D12PipelineState* getPipelineState(uint16_t _programIdx)
{
ProgramD3D12& program = m_program[_programIdx];
bx::HashMurmur2A murmur;
murmur.begin();
murmur.add(program.m_vsh->m_hash);
const uint32_t hash = murmur.end();
ID3D12PipelineState* pso = m_pipelineStateCache.find(hash);
if(NULL != pso)
{
return pso;
}
D3D12_COMPUTE_PIPELINE_STATE_DESC desc;
memset(&desc, 0, sizeof(desc) );
desc.pRootSignature = m_rootSignature;
desc.CS.pShaderBytecode = program.m_vsh->m_code->data;
desc.CS.BytecodeLength = program.m_vsh->m_code->size;
DX_CHECK(m_device->CreateComputePipelineState(&desc
,__uuidof(ID3D12PipelineState)
,(void**)&pso
));
m_pipelineStateCache.add(hash, pso);
return pso;
}
ID3D12PipelineState* getPipelineState(uint64_t _state, uint64_t _stencil, uint16_t _declIdx, uint16_t _programIdx, uint8_t _numInstanceData)
{
ProgramD3D12& program = m_program[_programIdx];
_state &= 0
| BGFX_STATE_RGB_WRITE
| BGFX_STATE_ALPHA_WRITE
| BGFX_STATE_DEPTH_WRITE
| BGFX_STATE_DEPTH_TEST_MASK
| BGFX_STATE_BLEND_MASK
| BGFX_STATE_BLEND_EQUATION_MASK
| BGFX_STATE_BLEND_INDEPENDENT
| BGFX_STATE_CULL_MASK
| BGFX_STATE_MSAA
| BGFX_STATE_PT_MASK
;
_stencil &= packStencil(~BGFX_STENCIL_FUNC_REF_MASK, BGFX_STENCIL_MASK);
VertexDecl decl;
memcpy(&decl, &m_vertexDecls[_declIdx], sizeof(VertexDecl) );
const uint16_t* attrMask = program.m_vsh->m_attrMask;
for (uint32_t ii = 0; ii < Attrib::Count; ++ii)
{
uint16_t mask = attrMask[ii];
uint16_t attr = (decl.m_attributes[ii] & mask);
decl.m_attributes[ii] = attr == 0 ? UINT16_MAX : attr == UINT16_MAX ? 0 : attr;
}
bx::HashMurmur2A murmur;
murmur.begin();
murmur.add(_state);
murmur.add(_stencil);
murmur.add(program.m_vsh->m_hash);
murmur.add(program.m_vsh->m_attrMask, sizeof(program.m_vsh->m_attrMask) );
murmur.add(program.m_fsh->m_hash);
murmur.add(m_vertexDecls[_declIdx].m_hash);
murmur.add(decl.m_attributes, sizeof(decl.m_attributes) );
murmur.add(m_fbh.idx);
murmur.add(_numInstanceData);
const uint32_t hash = murmur.end();
ID3D12PipelineState* pso = m_pipelineStateCache.find(hash);
if (NULL != pso)
{
return pso;
}
D3D12_GRAPHICS_PIPELINE_STATE_DESC desc;
memset(&desc, 0, sizeof(desc) );
desc.pRootSignature = m_rootSignature;
desc.VS.pShaderBytecode = program.m_vsh->m_code->data;
desc.VS.BytecodeLength = program.m_vsh->m_code->size;
const Memory* temp = alloc(program.m_fsh->m_code->size);
memset(temp->data, 0, temp->size);
bx::MemoryReader rd(program.m_fsh->m_code->data, program.m_fsh->m_code->size);
bx::StaticMemoryBlockWriter wr(temp->data, temp->size);
DxbcContext dxbc;
read(&rd, dxbc);
bool patchShader = true;
if (BX_ENABLED(BGFX_CONFIG_DEBUG) )
{
union { uint32_t offset; void* ptr; } cast = { 0 };
filter(dxbc.shader, dxbc.shader, patchCb0, cast.ptr);
write(&wr, dxbc);
dxbcHash(temp->data + 20, temp->size - 20, temp->data + 4);
patchShader = 0 == memcmp(program.m_fsh->m_code->data, temp->data, 16);
BX_CHECK(patchShader, "DXBC fragment shader patching error (ShaderHandle: %d).", program.m_fsh - m_shaders);
if (!patchShader)
{
for (uint32_t ii = 20; ii < temp->size; ii += 16)
{
if (0 != memcmp(&program.m_fsh->m_code->data[ii], &temp->data[ii], 16) )
{
// dbgPrintfData(&program.m_fsh->m_code->data[ii], temp->size-ii, "");
// dbgPrintfData(&temp->data[ii], temp->size-ii, "");
break;
}
}
desc.PS.pShaderBytecode = program.m_fsh->m_code->data;
desc.PS.BytecodeLength = program.m_fsh->m_code->size;
}
}
if (patchShader)
{
memcpy(temp->data, program.m_fsh->m_code->data, program.m_fsh->m_code->size);
bx::seek(&wr, 0, bx::Whence::Begin);
union { uint32_t offset; void* ptr; } cast =
{
m_currentProgram->m_vsh->m_size/16
};
filter(dxbc.shader, dxbc.shader, patchCb0, cast.ptr);
write(&wr, dxbc);
dxbcHash(temp->data + 20, temp->size - 20, temp->data + 4);
desc.PS.pShaderBytecode = temp->data;
desc.PS.BytecodeLength = temp->size;
}
desc.DS.pShaderBytecode = NULL;
desc.DS.BytecodeLength = 0;
desc.HS.pShaderBytecode = NULL;
desc.HS.BytecodeLength = 0;
desc.GS.pShaderBytecode = NULL;
desc.GS.BytecodeLength = 0;
desc.StreamOutput.pSODeclaration = NULL;
desc.StreamOutput.NumEntries = 0;
desc.StreamOutput.pBufferStrides = NULL;
desc.StreamOutput.NumStrides = 0;
desc.StreamOutput.RasterizedStream = 0;
setBlendState(desc.BlendState, _state);
desc.SampleMask = 1;
setRasterizerState(desc.RasterizerState, _state);
setDepthStencilState(desc.DepthStencilState, _state, _stencil);
D3D12_INPUT_ELEMENT_DESC vertexElements[Attrib::Count + 1 + BGFX_CONFIG_MAX_INSTANCE_DATA_COUNT];
desc.InputLayout.NumElements = setInputLayout(vertexElements, m_vertexDecls[_declIdx], program, _numInstanceData);
desc.InputLayout.pInputElementDescs = vertexElements;
uint8_t primIndex = uint8_t( (_state&BGFX_STATE_PT_MASK) >> BGFX_STATE_PT_SHIFT);
desc.PrimitiveTopologyType = s_primInfo[primIndex].m_topologyType;
if (isValid(m_fbh) )
{
const FrameBufferD3D12& frameBuffer = m_frameBuffers[m_fbh.idx];
desc.NumRenderTargets = frameBuffer.m_num;
for (uint8_t ii = 0, num = frameBuffer.m_num; ii < num; ++ii)
{
desc.RTVFormats[ii] = m_textures[frameBuffer.m_texture[ii].idx].m_srvd.Format;
}
if (isValid(frameBuffer.m_depth) )
{
desc.DSVFormat = s_textureFormat[m_textures[frameBuffer.m_depth.idx].m_textureFormat].m_fmtDsv;
}
else
{
desc.DSVFormat = DXGI_FORMAT_UNKNOWN;
}
}
else
{
desc.NumRenderTargets = 1;
desc.RTVFormats[0] = DXGI_FORMAT_R8G8B8A8_UNORM;
desc.DSVFormat = DXGI_FORMAT_D24_UNORM_S8_UINT;
}
desc.SampleDesc.Count = 1;
desc.SampleDesc.Quality = 0;
uint32_t length = g_callback->cacheReadSize(hash);
bool cached = length > 0;
void* cachedData = NULL;
if (cached)
{
cachedData = BX_ALLOC(g_allocator, length);
if (g_callback->cacheRead(hash, cachedData, length) )
{
BX_TRACE("Loading chached PSO (size %d).", length);
bx::MemoryReader reader(cachedData, length);
// uint32_t format;
// bx::read(&reader, format);
desc.CachedPSO.pCachedBlob = reader.getDataPtr();
desc.CachedPSO.CachedBlobSizeInBytes = (size_t)reader.remaining();
HRESULT hr = m_device->CreateGraphicsPipelineState(&desc
, __uuidof(ID3D12PipelineState)
, (void**)&pso
);
if (FAILED(hr) )
{
BX_TRACE("Failed to load cached PSO (HRESULT 0x%08x).", hr);
memset(&desc.CachedPSO, 0, sizeof(desc.CachedPSO) );
}
}
}
if (NULL == pso)
{
DX_CHECK(m_device->CreateGraphicsPipelineState(&desc
, __uuidof(ID3D12PipelineState)
, (void**)&pso
) );
}
m_pipelineStateCache.add(hash, pso);
release(temp);
ID3DBlob* blob;
HRESULT hr = pso->GetCachedBlob(&blob);
if (SUCCEEDED(hr) )
{
void* data = blob->GetBufferPointer();
length = (uint32_t)blob->GetBufferSize();
g_callback->cacheWrite(hash, data, length);
DX_RELEASE(blob, 0);
}
if (NULL != cachedData)
{
BX_FREE(g_allocator, cachedData);
}
return pso;
}
uint16_t getSamplerState(const uint32_t* _flags, uint32_t _num = BGFX_CONFIG_MAX_TEXTURE_SAMPLERS)
{
bx::HashMurmur2A murmur;
murmur.begin();
murmur.add(_flags, _num * sizeof(uint32_t) );
uint32_t hash = murmur.end();
uint16_t sampler = m_samplerStateCache.find(hash);
if (UINT16_MAX == sampler)
{
sampler = m_samplerAllocator.alloc(_flags, _num);
m_samplerStateCache.add(hash, sampler);
}
return sampler;
}
void commit(ConstantBuffer& _constantBuffer)
{
_constantBuffer.reset();
for (;;)
{
uint32_t opcode = _constantBuffer.read();
if (UniformType::End == opcode)
{
break;
}
UniformType::Enum type;
uint16_t loc;
uint16_t num;
uint16_t copy;
ConstantBuffer::decodeOpcode(opcode, type, loc, num, copy);
const char* data;
if (copy)
{
data = _constantBuffer.read(g_uniformTypeSize[type]*num);
}
else
{
UniformHandle handle;
memcpy(&handle, _constantBuffer.read(sizeof(UniformHandle) ), sizeof(UniformHandle) );
data = (const char*)m_uniforms[handle.idx];
}
#define CASE_IMPLEMENT_UNIFORM(_uniform, _dxsuffix, _type) \
case UniformType::_uniform: \
case UniformType::_uniform|BGFX_UNIFORM_FRAGMENTBIT: \
{ \
setShaderUniform(type, loc, data, num); \
} \
break;
switch ( (int32_t)type)
{
case UniformType::Mat3:
case UniformType::Mat3|BGFX_UNIFORM_FRAGMENTBIT:
{
float* value = (float*)data;
for (uint32_t ii = 0, count = num/3; ii < count; ++ii, loc += 3*16, value += 9)
{
Matrix4 mtx;
mtx.un.val[ 0] = value[0];
mtx.un.val[ 1] = value[1];
mtx.un.val[ 2] = value[2];
mtx.un.val[ 3] = 0.0f;
mtx.un.val[ 4] = value[3];
mtx.un.val[ 5] = value[4];
mtx.un.val[ 6] = value[5];
mtx.un.val[ 7] = 0.0f;
mtx.un.val[ 8] = value[6];
mtx.un.val[ 9] = value[7];
mtx.un.val[10] = value[8];
mtx.un.val[11] = 0.0f;
setShaderUniform(type, loc, &mtx.un.val[0], 3);
}
}
break;
CASE_IMPLEMENT_UNIFORM(Int1, I, int);
CASE_IMPLEMENT_UNIFORM(Vec4, F, float);
CASE_IMPLEMENT_UNIFORM(Mat4, F, float);
case UniformType::End:
break;
default:
BX_TRACE("%4d: INVALID 0x%08x, t %d, l %d, n %d, c %d", _constantBuffer.getPos(), opcode, type, loc, num, copy);
break;
}
#undef CASE_IMPLEMENT_UNIFORM
}
}
void clear(const Clear& _clear, const float _palette[][4], const D3D12_RECT* _rect = NULL, uint32_t _num = 0)
{
if (isValid(m_fbh) )
{
FrameBufferD3D12& frameBuffer = m_frameBuffers[m_fbh.idx];
frameBuffer.clear(m_commandList, _clear, _palette);
}
else
{
if (NULL != m_currentColor
&& BGFX_CLEAR_COLOR & _clear.m_flags)
{
if (BGFX_CLEAR_COLOR_USE_PALETTE & _clear.m_flags)
{
uint8_t index = _clear.m_index[0];
if (UINT8_MAX != index)
{
m_commandList->ClearRenderTargetView(*m_currentColor
, _palette[index]
, _num
, _rect
);
}
}
else
{
float frgba[4] =
{
_clear.m_index[0] * 1.0f / 255.0f,
_clear.m_index[1] * 1.0f / 255.0f,
_clear.m_index[2] * 1.0f / 255.0f,
_clear.m_index[3] * 1.0f / 255.0f,
};
m_commandList->ClearRenderTargetView(*m_currentColor
, frgba
, _num
, _rect
);
}
}
if (NULL != m_currentDepthStencil
&& (BGFX_CLEAR_DEPTH | BGFX_CLEAR_STENCIL) & _clear.m_flags)
{
uint32_t flags = 0;
flags |= (_clear.m_flags & BGFX_CLEAR_DEPTH ) ? D3D12_CLEAR_FLAG_DEPTH : 0;
flags |= (_clear.m_flags & BGFX_CLEAR_STENCIL) ? D3D12_CLEAR_FLAG_STENCIL : 0;
m_commandList->ClearDepthStencilView(*m_currentDepthStencil
, D3D12_CLEAR_FLAGS(flags)
, _clear.m_depth
, _clear.m_stencil
, _num
, _rect
);
}
}
}
void clearQuad(ClearQuad& _clearQuad, const Rect& _rect, const Clear& _clear, const float _palette[][4])
{
BX_UNUSED(_clearQuad);
uint32_t width = m_scd.BufferDesc.Width;
uint32_t height = m_scd.BufferDesc.Height;
if (0 == _rect.m_x
&& 0 == _rect.m_y
&& width == _rect.m_width
&& height == _rect.m_height)
{
clear(_clear, _palette);
}
else
{
D3D12_RECT rect;
rect.left = _rect.m_x;
rect.top = _rect.m_y;
rect.right = _rect.m_x + _rect.m_width;
rect.bottom = _rect.m_y + _rect.m_height;
clear(_clear, _palette, &rect);
}
}
uint64_t kick()
{
uint64_t fence = m_cmd.kick();
m_commandList = m_cmd.alloc();
return fence;
}
void finish()
{
m_cmd.kick();
m_cmd.finish();
m_commandList = NULL;
}
void finishAll()
{
uint64_t fence = m_cmd.kick();
m_cmd.finish(fence, true);
m_commandList = NULL;
}
void* m_d3d12dll;
void* m_dxgidll;
D3D_DRIVER_TYPE m_driverType;
IDXGIAdapter* m_adapter;
DXGI_ADAPTER_DESC m_adapterDesc;
D3D12_FEATURE_DATA_ARCHITECTURE m_architecture;
D3D12_FEATURE_DATA_D3D12_OPTIONS m_options;
IDXGIFactory4* m_factory;
IDXGISwapChain3* m_swapChain;
int64_t m_presentElapsed;
uint16_t m_lost;
uint16_t m_numWindows;
FrameBufferHandle m_windows[BGFX_CONFIG_MAX_FRAME_BUFFERS];
ID3D12Device* m_device;
ID3D12InfoQueue* m_infoQueue;
ID3D12DescriptorHeap* m_rtvDescriptorHeap;
ID3D12DescriptorHeap* m_dsvDescriptorHeap;
D3D12_CPU_DESCRIPTOR_HANDLE m_rtvHandle;
D3D12_CPU_DESCRIPTOR_HANDLE m_dsvHandle;
D3D12_CPU_DESCRIPTOR_HANDLE* m_currentColor;
D3D12_CPU_DESCRIPTOR_HANDLE* m_currentDepthStencil;
ID3D12Resource* m_backBufferColor[4];
uint64_t m_backBufferColorFence[4];
ID3D12Resource* m_backBufferDepthStencil;
ScratchBufferD3D12 m_scratchBuffer[4];
DescriptorAllocator m_samplerAllocator;
ID3D12RootSignature* m_rootSignature;
ID3D12CommandSignature* m_commandSignature[2];
CommandQueue m_cmd;
ID3D12GraphicsCommandList* m_commandList;
Resolution m_resolution;
bool m_wireframe;
DXGI_SWAP_CHAIN_DESC m_scd;
uint32_t m_flags;
BufferD3D12 m_indexBuffers[BGFX_CONFIG_MAX_INDEX_BUFFERS];
VertexBufferD3D12 m_vertexBuffers[BGFX_CONFIG_MAX_VERTEX_BUFFERS];
ShaderD3D12 m_shaders[BGFX_CONFIG_MAX_SHADERS];
ProgramD3D12 m_program[BGFX_CONFIG_MAX_PROGRAMS];
TextureD3D12 m_textures[BGFX_CONFIG_MAX_TEXTURES];
VertexDecl m_vertexDecls[BGFX_CONFIG_MAX_VERTEX_DECLS];
FrameBufferD3D12 m_frameBuffers[BGFX_CONFIG_MAX_FRAME_BUFFERS];
void* m_uniforms[BGFX_CONFIG_MAX_UNIFORMS];
Matrix4 m_predefinedUniforms[PredefinedUniform::Count];
UniformRegistry m_uniformReg;
StateCacheT<ID3D12PipelineState> m_pipelineStateCache;
StateCache m_samplerStateCache;
TextVideoMem m_textVideoMem;
ProgramD3D12* m_currentProgram;
uint8_t m_fsScratch[64<<10];
uint8_t m_vsScratch[64<<10];
uint32_t m_fsChanges;
uint32_t m_vsChanges;
FrameBufferHandle m_fbh;
uint32_t m_backBufferColorIdx;
bool m_rtMsaa;
};
static RendererContextD3D12* s_renderD3D12;
RendererContextI* rendererCreate()
{
s_renderD3D12 = BX_NEW(g_allocator, RendererContextD3D12);
if (!s_renderD3D12->init() )
{
BX_DELETE(g_allocator, s_renderD3D12);
s_renderD3D12 = NULL;
}
return s_renderD3D12;
}
void rendererDestroy()
{
s_renderD3D12->shutdown();
BX_DELETE(g_allocator, s_renderD3D12);
s_renderD3D12 = NULL;
}
void ScratchBufferD3D12::create(uint32_t _size, uint32_t _maxDescriptors)
{
m_size = _size;
ID3D12Device* device = s_renderD3D12->m_device;
m_incrementSize = device->GetDescriptorHandleIncrementSize(D3D12_DESCRIPTOR_HEAP_TYPE_CBV_SRV_UAV);
D3D12_DESCRIPTOR_HEAP_DESC desc;
desc.NumDescriptors = _maxDescriptors;
desc.Type = D3D12_DESCRIPTOR_HEAP_TYPE_CBV_SRV_UAV;
desc.Flags = D3D12_DESCRIPTOR_HEAP_FLAG_SHADER_VISIBLE;
desc.NodeMask = 1;
DX_CHECK(device->CreateDescriptorHeap(&desc
, __uuidof(ID3D12DescriptorHeap)
, (void**)&m_heap
) );
m_upload = createCommittedResource(device, HeapProperty::Upload, desc.NumDescriptors * 1024);
m_gpuVA = m_upload->GetGPUVirtualAddress();
m_upload->Map(0, NULL, (void**)&m_data);
reset(m_gpuHandle);
}
void ScratchBufferD3D12::destroy()
{
m_upload->Unmap(0, NULL);
DX_RELEASE(m_upload, 0);
DX_RELEASE(m_heap, 0);
}
void ScratchBufferD3D12::reset(D3D12_GPU_DESCRIPTOR_HANDLE& gpuHandle)
{
m_pos = 0;
m_cpuHandle = m_heap->GetCPUDescriptorHandleForHeapStart();
m_gpuHandle = m_heap->GetGPUDescriptorHandleForHeapStart();
gpuHandle = m_gpuHandle;
}
void* ScratchBufferD3D12::alloc(D3D12_GPU_VIRTUAL_ADDRESS& _gpuAddress, uint32_t _size)
{
_gpuAddress = m_gpuVA + m_pos;
D3D12_CONSTANT_BUFFER_VIEW_DESC desc;
desc.BufferLocation = _gpuAddress;
desc.SizeInBytes = _size;
void* data = &m_data[m_pos];
m_pos += BX_ALIGN_256(_size);
ID3D12Device* device = s_renderD3D12->m_device;
device->CreateConstantBufferView(&desc
, m_cpuHandle
);
m_cpuHandle.ptr += m_incrementSize;
m_gpuHandle.ptr += m_incrementSize;
return data;
}
void ScratchBufferD3D12::alloc(D3D12_GPU_DESCRIPTOR_HANDLE& gpuHandle, TextureD3D12& _texture)
{
ID3D12Device* device = s_renderD3D12->m_device;
device->CreateShaderResourceView(_texture.m_ptr
, &_texture.m_srvd
, m_cpuHandle
);
m_cpuHandle.ptr += m_incrementSize;
gpuHandle = m_gpuHandle;
m_gpuHandle.ptr += m_incrementSize;
}
void ScratchBufferD3D12::allocUav(D3D12_GPU_DESCRIPTOR_HANDLE& gpuHandle, TextureD3D12& _texture, uint8_t _mip)
{
ID3D12Device* device = s_renderD3D12->m_device;
D3D12_UNORDERED_ACCESS_VIEW_DESC tmpUavd;
D3D12_UNORDERED_ACCESS_VIEW_DESC* uavd = &_texture.m_uavd;
if (0 != _mip)
{
memcpy(&tmpUavd, uavd, sizeof(tmpUavd) );
switch (_texture.m_uavd.ViewDimension)
{
default:
case D3D12_UAV_DIMENSION_TEXTURE2D:
uavd->Texture2D.MipSlice = _mip;
uavd->Texture2D.PlaneSlice = 0;
break;
case D3D12_UAV_DIMENSION_TEXTURE3D:
uavd->Texture3D.MipSlice = _mip;
break;
}
uavd = &tmpUavd;
}
device->CreateUnorderedAccessView(_texture.m_ptr
, NULL
, uavd
, m_cpuHandle
);
m_cpuHandle.ptr += m_incrementSize;
gpuHandle = m_gpuHandle;
m_gpuHandle.ptr += m_incrementSize;
}
void ScratchBufferD3D12::alloc(D3D12_GPU_DESCRIPTOR_HANDLE& gpuHandle, BufferD3D12& _buffer)
{
ID3D12Device* device = s_renderD3D12->m_device;
device->CreateShaderResourceView(_buffer.m_ptr
, &_buffer.m_srvd
, m_cpuHandle
);
m_cpuHandle.ptr += m_incrementSize;
gpuHandle = m_gpuHandle;
m_gpuHandle.ptr += m_incrementSize;
}
void ScratchBufferD3D12::allocUav(D3D12_GPU_DESCRIPTOR_HANDLE& gpuHandle, BufferD3D12& _buffer)
{
ID3D12Device* device = s_renderD3D12->m_device;
device->CreateUnorderedAccessView(_buffer.m_ptr
, NULL
, &_buffer.m_uavd
, m_cpuHandle
);
m_cpuHandle.ptr += m_incrementSize;
gpuHandle = m_gpuHandle;
m_gpuHandle.ptr += m_incrementSize;
}
void DescriptorAllocator::create(D3D12_DESCRIPTOR_HEAP_TYPE _type, uint32_t _maxDescriptors, uint16_t _numDescriptorsPerBlock)
{
m_handleAlloc = bx::createHandleAlloc(g_allocator, _maxDescriptors);
m_numDescriptorsPerBlock = _numDescriptorsPerBlock;
ID3D12Device* device = s_renderD3D12->m_device;
m_incrementSize = device->GetDescriptorHandleIncrementSize(_type);
D3D12_DESCRIPTOR_HEAP_DESC desc;
desc.NumDescriptors = _maxDescriptors;
desc.Type = _type;
desc.Flags = D3D12_DESCRIPTOR_HEAP_FLAG_SHADER_VISIBLE;
desc.NodeMask = 1;
DX_CHECK(device->CreateDescriptorHeap(&desc
, __uuidof(ID3D12DescriptorHeap)
, (void**)&m_heap
) );
m_cpuHandle = m_heap->GetCPUDescriptorHandleForHeapStart();
m_gpuHandle = m_heap->GetGPUDescriptorHandleForHeapStart();
}
void DescriptorAllocator::destroy()
{
bx::destroyHandleAlloc(g_allocator, m_handleAlloc);
DX_RELEASE(m_heap, 0);
}
uint16_t DescriptorAllocator::alloc(ID3D12Resource* _ptr, const D3D12_SHADER_RESOURCE_VIEW_DESC* _desc)
{
uint16_t idx = m_handleAlloc->alloc();
D3D12_CPU_DESCRIPTOR_HANDLE cpuHandle = { m_cpuHandle.ptr + idx * m_incrementSize };
ID3D12Device* device = s_renderD3D12->m_device;
device->CreateShaderResourceView(_ptr
, _desc
, cpuHandle
);
return idx;
}
uint16_t DescriptorAllocator::alloc(const uint32_t* _flags, uint32_t _num)
{
uint16_t idx = m_handleAlloc->alloc();
ID3D12Device* device = s_renderD3D12->m_device;
for (uint32_t ii = 0; ii < _num; ++ii)
{
uint32_t flags = _flags[ii];
const uint32_t cmpFunc = (flags&BGFX_TEXTURE_COMPARE_MASK)>>BGFX_TEXTURE_COMPARE_SHIFT;
const uint8_t minFilter = s_textureFilter[0][(flags&BGFX_TEXTURE_MIN_MASK)>>BGFX_TEXTURE_MIN_SHIFT];
const uint8_t magFilter = s_textureFilter[1][(flags&BGFX_TEXTURE_MAG_MASK)>>BGFX_TEXTURE_MAG_SHIFT];
const uint8_t mipFilter = s_textureFilter[2][(flags&BGFX_TEXTURE_MIP_MASK)>>BGFX_TEXTURE_MIP_SHIFT];
const uint8_t filter = 0 == cmpFunc ? 0 : D3D12_FILTER_COMPARISON_MIN_MAG_MIP_POINT;
D3D12_SAMPLER_DESC sd;
sd.Filter = (D3D12_FILTER)(filter|minFilter|magFilter|mipFilter);
sd.AddressU = s_textureAddress[(flags&BGFX_TEXTURE_U_MASK)>>BGFX_TEXTURE_U_SHIFT];
sd.AddressV = s_textureAddress[(flags&BGFX_TEXTURE_V_MASK)>>BGFX_TEXTURE_V_SHIFT];
sd.AddressW = s_textureAddress[(flags&BGFX_TEXTURE_W_MASK)>>BGFX_TEXTURE_W_SHIFT];
sd.MinLOD = 0;
sd.MaxLOD = D3D12_FLOAT32_MAX;
sd.MipLODBias = 0.0f;
sd.MaxAnisotropy = 1; //m_maxAnisotropy;
sd.ComparisonFunc = 0 == cmpFunc ? D3D12_COMPARISON_FUNC_NEVER : s_cmpFunc[cmpFunc];
D3D12_CPU_DESCRIPTOR_HANDLE cpuHandle =
{
m_cpuHandle.ptr + (idx * m_numDescriptorsPerBlock + ii) * m_incrementSize
};
device->CreateSampler(&sd, cpuHandle);
}
return idx;
}
void DescriptorAllocator::free(uint16_t _idx)
{
m_handleAlloc->free(_idx);
}
void DescriptorAllocator::reset()
{
uint16_t max = m_handleAlloc->getMaxHandles();
bx::destroyHandleAlloc(g_allocator, m_handleAlloc);
m_handleAlloc = bx::createHandleAlloc(g_allocator, max);
}
D3D12_GPU_DESCRIPTOR_HANDLE DescriptorAllocator::get(uint16_t _idx)
{
D3D12_GPU_DESCRIPTOR_HANDLE gpuHandle = { m_gpuHandle.ptr + _idx * m_numDescriptorsPerBlock * m_incrementSize };
return gpuHandle;
}
struct UavFormat
{
DXGI_FORMAT format[3];
uint32_t stride;
};
static const UavFormat s_uavFormat[] =
{ // BGFX_BUFFER_COMPUTE_TYPE_UINT, BGFX_BUFFER_COMPUTE_TYPE_INT, BGFX_BUFFER_COMPUTE_TYPE_FLOAT
{ { DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN }, 0 }, // ignored
{ { DXGI_FORMAT_R8_SINT, DXGI_FORMAT_R8_UINT, DXGI_FORMAT_UNKNOWN }, 1 }, // BGFX_BUFFER_COMPUTE_FORMAT_8x1
{ { DXGI_FORMAT_R8G8_SINT, DXGI_FORMAT_R8G8_UINT, DXGI_FORMAT_UNKNOWN }, 2 }, // BGFX_BUFFER_COMPUTE_FORMAT_8x2
{ { DXGI_FORMAT_R8G8B8A8_SINT, DXGI_FORMAT_R8G8B8A8_UINT, DXGI_FORMAT_UNKNOWN }, 4 }, // BGFX_BUFFER_COMPUTE_FORMAT_8x4
{ { DXGI_FORMAT_R16_SINT, DXGI_FORMAT_R16_UINT, DXGI_FORMAT_R16_FLOAT }, 2 }, // BGFX_BUFFER_COMPUTE_FORMAT_16x1
{ { DXGI_FORMAT_R16G16_SINT, DXGI_FORMAT_R16G16_UINT, DXGI_FORMAT_R16G16_FLOAT }, 4 }, // BGFX_BUFFER_COMPUTE_FORMAT_16x2
{ { DXGI_FORMAT_R16G16B16A16_SINT, DXGI_FORMAT_R16G16B16A16_UINT, DXGI_FORMAT_R16G16B16A16_FLOAT }, 8 }, // BGFX_BUFFER_COMPUTE_FORMAT_16x4
{ { DXGI_FORMAT_R32_SINT, DXGI_FORMAT_R32_UINT, DXGI_FORMAT_R32_FLOAT }, 4 }, // BGFX_BUFFER_COMPUTE_FORMAT_32x1
{ { DXGI_FORMAT_R32G32_SINT, DXGI_FORMAT_R32G32_UINT, DXGI_FORMAT_R32G32_FLOAT }, 8 }, // BGFX_BUFFER_COMPUTE_FORMAT_32x2
{ { DXGI_FORMAT_R32G32B32A32_SINT, DXGI_FORMAT_R32G32B32A32_UINT, DXGI_FORMAT_R32G32B32A32_FLOAT }, 16 }, // BGFX_BUFFER_COMPUTE_FORMAT_32x4
};
void BufferD3D12::create(uint32_t _size, void* _data, uint16_t _flags, bool _vertex, uint32_t _stride)
{
m_size = _size;
m_flags = _flags;
const bool needUav = 0 != (_flags & (BGFX_BUFFER_COMPUTE_WRITE|BGFX_BUFFER_DRAW_INDIRECT));
// const bool needSrv = 0 != (_flags & BGFX_BUFFER_COMPUTE_READ);
const bool drawIndirect = 0 != (_flags & BGFX_BUFFER_DRAW_INDIRECT);
m_dynamic = NULL == _data || needUav;
DXGI_FORMAT format;
uint32_t stride;
D3D12_RESOURCE_FLAGS flags = needUav
? D3D12_RESOURCE_FLAG_ALLOW_UNORDERED_ACCESS
: D3D12_RESOURCE_FLAG_NONE
;
if(drawIndirect)
{
format = DXGI_FORMAT_R32G32B32A32_UINT;
stride = 16;
}
else
{
uint32_t uavFormat = (_flags & BGFX_BUFFER_COMPUTE_FORMAT_MASK) >> BGFX_BUFFER_COMPUTE_FORMAT_SHIFT;
if (0 == uavFormat)
{
if (_vertex)
{
format = DXGI_FORMAT_R32G32B32A32_FLOAT;
stride = 16;
}
else
{
if (0 == (_flags & BGFX_BUFFER_INDEX32) )
{
format = DXGI_FORMAT_R16_UINT;
stride = 2;
}
else
{
format = DXGI_FORMAT_R32_UINT;
stride = 4;
}
}
}
else
{
const uint32_t uavType = bx::uint32_satsub( (_flags & BGFX_BUFFER_COMPUTE_TYPE_MASK) >> BGFX_BUFFER_COMPUTE_TYPE_SHIFT, 1);
format = s_uavFormat[uavFormat].format[uavType];
stride = s_uavFormat[uavFormat].stride;
}
}
stride = 0 == _stride ? stride : _stride;
m_srvd.Format = format;
m_srvd.ViewDimension = D3D12_SRV_DIMENSION_BUFFER;
m_srvd.Shader4ComponentMapping = D3D12_DEFAULT_SHADER_4_COMPONENT_MAPPING;
m_srvd.Buffer.FirstElement = 0;
m_srvd.Buffer.NumElements = m_size / stride;
m_srvd.Buffer.StructureByteStride = 0;
m_srvd.Buffer.Flags = D3D12_BUFFER_SRV_FLAG_NONE;
m_uavd.Format = format;
m_uavd.ViewDimension = D3D12_UAV_DIMENSION_BUFFER;
m_uavd.Buffer.FirstElement = 0;
m_uavd.Buffer.NumElements = m_size / stride;
m_uavd.Buffer.StructureByteStride = 0;
m_uavd.Buffer.CounterOffsetInBytes = 0;
m_uavd.Buffer.Flags = D3D12_BUFFER_UAV_FLAG_NONE;
ID3D12Device* device = s_renderD3D12->m_device;
ID3D12GraphicsCommandList* commandList = s_renderD3D12->m_commandList;
m_ptr = createCommittedResource(device, HeapProperty::Default, _size, flags);
m_gpuVA = m_ptr->GetGPUVirtualAddress();
setState(commandList, D3D12_RESOURCE_STATE_GENERIC_READ);
if (!m_dynamic)
{
update(commandList, 0, _size, _data);
}
}
void BufferD3D12::update(ID3D12GraphicsCommandList* _commandList, uint32_t _offset, uint32_t _size, void* _data, bool /*_discard*/)
{
ID3D12Resource* staging = createCommittedResource(s_renderD3D12->m_device, HeapProperty::Upload, _size);
uint8_t* data;
DX_CHECK(staging->Map(0, NULL, (void**)&data) );
memcpy(data, _data, _size);
staging->Unmap(0, NULL);
D3D12_RESOURCE_STATES state = setState(_commandList, D3D12_RESOURCE_STATE_COPY_DEST);
_commandList->CopyBufferRegion(m_ptr, _offset, staging, 0, _size);
setState(_commandList, state);
s_renderD3D12->m_cmd.release(staging);
}
D3D12_RESOURCE_STATES BufferD3D12::setState(ID3D12GraphicsCommandList* _commandList, D3D12_RESOURCE_STATES _state)
{
if (m_state != _state)
{
setResourceBarrier(_commandList
, m_ptr
, m_state
, _state
);
bx::xchg(m_state, _state);
}
return _state;
}
void VertexBufferD3D12::create(uint32_t _size, void* _data, VertexDeclHandle _declHandle, uint16_t _flags)
{
BufferD3D12::create(_size, _data, _flags, true);
m_decl = _declHandle;
}
void ShaderD3D12::create(const Memory* _mem)
{
bx::MemoryReader reader(_mem->data, _mem->size);
uint32_t magic;
bx::read(&reader, magic);
switch (magic)
{
case BGFX_CHUNK_MAGIC_CSH:
case BGFX_CHUNK_MAGIC_FSH:
case BGFX_CHUNK_MAGIC_VSH:
break;
default:
BGFX_FATAL(false, Fatal::InvalidShader, "Unknown shader format %x.", magic);
break;
}
bool fragment = BGFX_CHUNK_MAGIC_FSH == magic;
uint32_t iohash;
bx::read(&reader, iohash);
uint16_t count;
bx::read(&reader, count);
m_numPredefined = 0;
m_numUniforms = count;
BX_TRACE("%s Shader consts %d"
, BGFX_CHUNK_MAGIC_FSH == magic ? "Fragment" : BGFX_CHUNK_MAGIC_VSH == magic ? "Vertex" : "Compute"
, count
);
uint8_t fragmentBit = fragment ? BGFX_UNIFORM_FRAGMENTBIT : 0;
if (0 < count)
{
for (uint32_t ii = 0; ii < count; ++ii)
{
uint8_t nameSize;
bx::read(&reader, nameSize);
char name[256];
bx::read(&reader, &name, nameSize);
name[nameSize] = '\0';
uint8_t type;
bx::read(&reader, type);
uint8_t num;
bx::read(&reader, num);
uint16_t regIndex;
bx::read(&reader, regIndex);
uint16_t regCount;
bx::read(&reader, regCount);
const char* kind = "invalid";
PredefinedUniform::Enum predefined = nameToPredefinedUniformEnum(name);
if (PredefinedUniform::Count != predefined)
{
kind = "predefined";
m_predefined[m_numPredefined].m_loc = regIndex;
m_predefined[m_numPredefined].m_count = regCount;
m_predefined[m_numPredefined].m_type = predefined|fragmentBit;
m_numPredefined++;
}
else
{
const UniformInfo* info = s_renderD3D12->m_uniformReg.find(name);
if (NULL != info)
{
if (NULL == m_constantBuffer)
{
m_constantBuffer = ConstantBuffer::create(1024);
}
kind = "user";
m_constantBuffer->writeUniformHandle( (UniformType::Enum)(type|fragmentBit), regIndex, info->m_handle, regCount);
}
}
BX_TRACE("\t%s: %s (%s), num %2d, r.index %3d, r.count %2d"
, kind
, name
, getUniformTypeName(UniformType::Enum(type&~BGFX_UNIFORM_FRAGMENTBIT) )
, num
, regIndex
, regCount
);
BX_UNUSED(kind);
}
if (NULL != m_constantBuffer)
{
m_constantBuffer->finish();
}
}
uint16_t shaderSize;
bx::read(&reader, shaderSize);
const DWORD* code = (const DWORD*)reader.getDataPtr();
bx::skip(&reader, shaderSize+1);
m_code = copy(code, shaderSize);
uint8_t numAttrs;
bx::read(&reader, numAttrs);
memset(m_attrMask, 0, sizeof(m_attrMask) );
for (uint32_t ii = 0; ii < numAttrs; ++ii)
{
uint16_t id;
bx::read(&reader, id);
Attrib::Enum attr = idToAttrib(id);
if (Attrib::Count != attr)
{
m_attrMask[attr] = UINT16_MAX;
}
}
bx::HashMurmur2A murmur;
murmur.begin();
murmur.add(iohash);
murmur.add(code, shaderSize);
murmur.add(numAttrs);
murmur.add(m_attrMask, numAttrs);
m_hash = murmur.end();
bx::read(&reader, m_size);
}
void TextureD3D12::create(const Memory* _mem, uint32_t _flags, uint8_t _skip)
{
ImageContainer imageContainer;
if (imageParse(imageContainer, _mem->data, _mem->size) )
{
uint8_t numMips = imageContainer.m_numMips;
const uint32_t startLod = bx::uint32_min(_skip, numMips-1);
numMips -= startLod;
const ImageBlockInfo& blockInfo = getBlockInfo(TextureFormat::Enum(imageContainer.m_format) );
const uint32_t textureWidth = bx::uint32_max(blockInfo.blockWidth, imageContainer.m_width >>startLod);
const uint32_t textureHeight = bx::uint32_max(blockInfo.blockHeight, imageContainer.m_height>>startLod);
m_flags = _flags;
m_requestedFormat = (uint8_t)imageContainer.m_format;
m_textureFormat = (uint8_t)imageContainer.m_format;
const TextureFormatInfo& tfi = s_textureFormat[m_requestedFormat];
const bool convert = DXGI_FORMAT_UNKNOWN == tfi.m_fmt;
uint8_t bpp = getBitsPerPixel(TextureFormat::Enum(m_textureFormat) );
if (convert)
{
m_textureFormat = (uint8_t)TextureFormat::BGRA8;
bpp = 32;
}
if (imageContainer.m_cubeMap)
{
m_type = TextureCube;
}
else if (imageContainer.m_depth > 1)
{
m_type = Texture3D;
}
else
{
m_type = Texture2D;
}
m_numMips = numMips;
const uint32_t numSides = imageContainer.m_cubeMap ? 6 : 1;
uint32_t numSrd = numMips*numSides;
D3D12_SUBRESOURCE_DATA* srd = (D3D12_SUBRESOURCE_DATA*)alloca(numSrd*sizeof(D3D12_SUBRESOURCE_DATA) );
uint32_t kk = 0;
const bool compressed = isCompressed(TextureFormat::Enum(m_textureFormat) );
const bool swizzle = TextureFormat::BGRA8 == m_textureFormat && 0 != (m_flags&BGFX_TEXTURE_COMPUTE_WRITE);
uint32_t blockWidth = 1;
uint32_t blockHeight = 1;
if (convert && compressed)
{
blockWidth = blockInfo.blockWidth;
blockHeight = blockInfo.blockHeight;
}
const bool bufferOnly = 0 != (m_flags&BGFX_TEXTURE_RT_BUFFER_ONLY);
const bool computeWrite = 0 != (m_flags&BGFX_TEXTURE_COMPUTE_WRITE);
const bool renderTarget = 0 != (m_flags&BGFX_TEXTURE_RT_MASK);
BX_TRACE("Texture %3d: %s (requested: %s), %dx%d%s RT[%c], BO[%c], CW[%c]%s."
, this - s_renderD3D12->m_textures
, getName( (TextureFormat::Enum)m_textureFormat)
, getName( (TextureFormat::Enum)m_requestedFormat)
, textureWidth
, textureHeight
, imageContainer.m_cubeMap ? "x6" : ""
, renderTarget ? 'x' : ' '
, bufferOnly ? 'x' : ' '
, computeWrite ? 'x' : ' '
, swizzle ? " (swizzle BGRA8 -> RGBA8)" : ""
);
uint32_t totalSize = 0;
for (uint8_t side = 0; side < numSides; ++side)
{
uint32_t width = textureWidth;
uint32_t height = textureHeight;
uint32_t depth = imageContainer.m_depth;
for (uint32_t lod = 0; lod < numMips; ++lod)
{
width = bx::uint32_max(blockWidth, width);
height = bx::uint32_max(blockHeight, height);
depth = bx::uint32_max(1, depth);
ImageMip mip;
if (imageGetRawData(imageContainer, side, lod+startLod, _mem->data, _mem->size, mip) )
{
if (convert)
{
const uint32_t pitch = bx::strideAlign(width*bpp / 8, D3D12_TEXTURE_DATA_PITCH_ALIGNMENT);
const uint32_t slice = bx::strideAlign(pitch * height, D3D12_TEXTURE_DATA_PLACEMENT_ALIGNMENT);
uint8_t* temp = (uint8_t*)BX_ALLOC(g_allocator, slice);
imageDecodeToBgra8(temp
, mip.m_data
, mip.m_width
, mip.m_height
, pitch, mip.m_format
);
srd[kk].pData = temp;
srd[kk].RowPitch = pitch;
srd[kk].SlicePitch = slice;
totalSize += slice;
}
else if (compressed)
{
uint32_t pitch = bx::strideAlign( (mip.m_width /blockInfo.blockWidth )*mip.m_blockSize, D3D12_TEXTURE_DATA_PITCH_ALIGNMENT);
uint32_t slice = bx::strideAlign( (mip.m_height/blockInfo.blockHeight)*pitch, D3D12_TEXTURE_DATA_PLACEMENT_ALIGNMENT);
uint8_t* temp = (uint8_t*)BX_ALLOC(g_allocator, slice);
imageCopy(mip.m_height/blockInfo.blockHeight
, (mip.m_width /blockInfo.blockWidth )*mip.m_blockSize
, mip.m_data
, pitch
, temp
);
srd[kk].pData = temp;
srd[kk].RowPitch = pitch;
srd[kk].SlicePitch = slice;
totalSize += slice;
}
else
{
const uint32_t pitch = bx::strideAlign(mip.m_width*mip.m_bpp / 8, D3D12_TEXTURE_DATA_PITCH_ALIGNMENT);
const uint32_t slice = bx::strideAlign(pitch * mip.m_height, D3D12_TEXTURE_DATA_PLACEMENT_ALIGNMENT);
uint8_t* temp = (uint8_t*)BX_ALLOC(g_allocator, slice);
imageCopy(mip.m_height
, mip.m_width*mip.m_bpp / 8
, mip.m_data
, pitch
, temp
);
srd[kk].pData = temp;
srd[kk].RowPitch = pitch;
srd[kk].SlicePitch = slice;
totalSize += slice;
}
if (swizzle)
{
// imageSwizzleBgra8(width, height, mip.m_width*4, data, temp);
}
srd[kk].SlicePitch = mip.m_height*srd[kk].RowPitch;
++kk;
}
else
{
const uint32_t pitch = bx::strideAlign(width*bpp / 8, D3D12_TEXTURE_DATA_PITCH_ALIGNMENT);
const uint32_t slice = bx::strideAlign(pitch * height, D3D12_TEXTURE_DATA_PLACEMENT_ALIGNMENT);
totalSize += slice;
}
width >>= 1;
height >>= 1;
depth >>= 1;
}
}
BX_TRACE("texture total size: %d", totalSize);
const uint32_t msaaQuality = bx::uint32_satsub( (m_flags&BGFX_TEXTURE_RT_MSAA_MASK)>>BGFX_TEXTURE_RT_MSAA_SHIFT, 1);
const DXGI_SAMPLE_DESC& msaa = s_msaa[msaaQuality];
memset(&m_srvd, 0, sizeof(m_srvd) );
m_srvd.Shader4ComponentMapping = D3D12_DEFAULT_SHADER_4_COMPONENT_MAPPING;
m_srvd.Format = s_textureFormat[m_textureFormat].m_fmtSrv;
DXGI_FORMAT format = s_textureFormat[m_textureFormat].m_fmt;
if (swizzle)
{
format = DXGI_FORMAT_R8G8B8A8_UNORM;
m_srvd.Format = DXGI_FORMAT_R8G8B8A8_UNORM;
}
m_uavd.Format = m_srvd.Format;
ID3D12Device* device = s_renderD3D12->m_device;
ID3D12GraphicsCommandList* commandList = s_renderD3D12->m_commandList;
D3D12_RESOURCE_DESC resourceDesc;
resourceDesc.Alignment = 0;
resourceDesc.Width = textureWidth;
resourceDesc.Height = textureHeight;
resourceDesc.MipLevels = numMips;
resourceDesc.Format = format;
resourceDesc.SampleDesc = msaa;
resourceDesc.Layout = D3D12_TEXTURE_LAYOUT_UNKNOWN;
resourceDesc.Flags = D3D12_RESOURCE_FLAG_NONE;
resourceDesc.DepthOrArraySize = numSides;
D3D12_RESOURCE_STATES state = D3D12_RESOURCE_STATE_PIXEL_SHADER_RESOURCE;
D3D12_CLEAR_VALUE* clearValue = NULL;
if (isDepth(TextureFormat::Enum(m_textureFormat) ) )
{
resourceDesc.Format = s_textureFormat[m_textureFormat].m_fmt;
resourceDesc.Flags |= D3D12_RESOURCE_FLAG_ALLOW_DEPTH_STENCIL;
state |= D3D12_RESOURCE_STATE_DEPTH_WRITE;
state &= ~D3D12_RESOURCE_STATE_PIXEL_SHADER_RESOURCE;
clearValue = (D3D12_CLEAR_VALUE*)alloca(sizeof(D3D12_CLEAR_VALUE) );
clearValue->Format = s_textureFormat[m_textureFormat].m_fmtDsv;
clearValue->DepthStencil.Depth = 1.0f;
clearValue->DepthStencil.Stencil = 0;
}
else if (renderTarget)
{
clearValue = (D3D12_CLEAR_VALUE*)alloca(sizeof(D3D12_CLEAR_VALUE) );
clearValue->Format = resourceDesc.Format;
clearValue->Color[0] = 0.0f;
clearValue->Color[1] = 0.0f;
clearValue->Color[2] = 0.0f;
clearValue->Color[3] = 0.0f;
resourceDesc.Flags |= D3D12_RESOURCE_FLAG_ALLOW_RENDER_TARGET;
}
if (bufferOnly)
{
resourceDesc.Flags |= D3D12_RESOURCE_FLAG_DENY_SHADER_RESOURCE;
state &= ~D3D12_RESOURCE_STATE_PIXEL_SHADER_RESOURCE;
}
if (computeWrite)
{
resourceDesc.Flags &= ~D3D12_RESOURCE_FLAG_ALLOW_UNORDERED_ACCESS;
}
switch (m_type)
{
case Texture2D:
resourceDesc.Dimension = D3D12_RESOURCE_DIMENSION_TEXTURE2D;
m_srvd.ViewDimension = 1 < msaa.Count ? D3D12_SRV_DIMENSION_TEXTURE2DMS : D3D12_SRV_DIMENSION_TEXTURE2D;
m_srvd.Texture2D.MostDetailedMip = 0;
m_srvd.Texture2D.MipLevels = numMips;
m_srvd.Texture2D.ResourceMinLODClamp = 0.0f;
m_uavd.ViewDimension = D3D12_UAV_DIMENSION_TEXTURE2D;
m_uavd.Texture2D.MipSlice = 0;
m_uavd.Texture2D.PlaneSlice = 0;
break;
case Texture3D:
resourceDesc.Dimension = D3D12_RESOURCE_DIMENSION_TEXTURE3D;
m_srvd.ViewDimension = D3D12_SRV_DIMENSION_TEXTURE3D;
m_srvd.Texture3D.MostDetailedMip = 0;
m_srvd.Texture3D.MipLevels = numMips;
m_srvd.Texture3D.ResourceMinLODClamp = 0.0f;
m_uavd.ViewDimension = D3D12_UAV_DIMENSION_TEXTURE3D;
m_uavd.Texture3D.MipSlice = 0;
m_uavd.Texture3D.FirstWSlice = 0;
m_uavd.Texture3D.WSize = 0;
break;
case TextureCube:
resourceDesc.Dimension = D3D12_RESOURCE_DIMENSION_TEXTURE2D;
m_srvd.ViewDimension = D3D12_SRV_DIMENSION_TEXTURECUBE;
m_srvd.TextureCube.MostDetailedMip = 0;
m_srvd.TextureCube.MipLevels = numMips;
m_srvd.TextureCube.ResourceMinLODClamp = 0.0f;
m_uavd.ViewDimension = D3D12_UAV_DIMENSION_TEXTURE2D;
m_uavd.Texture2D.MipSlice = 0;
m_uavd.Texture2D.PlaneSlice = 0;
break;
}
m_ptr = createCommittedResource(device, HeapProperty::Default, &resourceDesc, clearValue);
{
uint64_t uploadBufferSize;
uint32_t* numRows = (uint32_t*)alloca(sizeof(uint32_t)*numSrd);
uint64_t* rowSizeInBytes = (uint64_t*)alloca(sizeof(uint64_t)*numSrd);
D3D12_PLACED_SUBRESOURCE_FOOTPRINT* layouts = (D3D12_PLACED_SUBRESOURCE_FOOTPRINT*)alloca(sizeof(D3D12_PLACED_SUBRESOURCE_FOOTPRINT)*numSrd);
device->GetCopyableFootprints(&resourceDesc
, 0
, numSrd
, 0
, layouts
, numRows
, rowSizeInBytes
, &uploadBufferSize
);
BX_WARN(uploadBufferSize == totalSize, "uploadBufferSize %d (totalSize %d), numRows %d, rowSizeInBytes %d"
, uploadBufferSize
, totalSize
, numRows[0]
, rowSizeInBytes[0]
);
}
if (kk != 0)
{
ID3D12Resource* staging = createCommittedResource(s_renderD3D12->m_device, HeapProperty::Upload, totalSize);
setState(commandList,D3D12_RESOURCE_STATE_COPY_DEST);
uint64_t result = UpdateSubresources(commandList
, m_ptr
, staging
, 0
, 0
, numSrd
, srd
);
BX_CHECK(0 != result, "Invalid size"); BX_UNUSED(result);
BX_TRACE("Update subresource %" PRId64, result);
setState(commandList, state);
s_renderD3D12->m_cmd.release(staging);
}
else
{
setState(commandList, state);
}
if (0 != kk)
{
kk = 0;
for (uint8_t side = 0; side < numSides; ++side)
{
for (uint32_t lod = 0, num = numMips; lod < num; ++lod)
{
BX_FREE(g_allocator, const_cast<void*>(srd[kk].pData) );
++kk;
}
}
}
}
}
void TextureD3D12::destroy()
{
if (NULL != m_ptr)
{
DX_RELEASE(m_ptr, 0);
m_ptr = NULL;
}
}
void TextureD3D12::update(ID3D12GraphicsCommandList* _commandList, uint8_t _side, uint8_t _mip, const Rect& _rect, uint16_t _z, uint16_t _depth, uint16_t _pitch, const Memory* _mem)
{
setState(_commandList, D3D12_RESOURCE_STATE_COPY_DEST);
const uint32_t subres = _mip + (_side * m_numMips);
const uint32_t bpp = getBitsPerPixel(TextureFormat::Enum(m_textureFormat) );
const uint32_t rectpitch = _rect.m_width*bpp/8;
const uint32_t srcpitch = UINT16_MAX == _pitch ? rectpitch : _pitch;
D3D12_RESOURCE_DESC desc = m_ptr->GetDesc();
desc.Height = _rect.m_height;
uint32_t numRows;
uint64_t rowPitch;
uint64_t totalBytes;
D3D12_PLACED_SUBRESOURCE_FOOTPRINT layout;
s_renderD3D12->m_device->GetCopyableFootprints(&desc
, subres
, 1
, 0
, &layout
, &numRows
, &rowPitch
, &totalBytes
);
ID3D12Resource* staging = createCommittedResource(s_renderD3D12->m_device, HeapProperty::Upload, totalBytes);
uint8_t* data;
DX_CHECK(staging->Map(0, NULL, (void**)&data) );
for (uint32_t ii = 0, height = _rect.m_height; ii < height; ++ii)
{
memcpy(&data[ii*rowPitch], &_mem->data[ii*srcpitch], srcpitch);
}
staging->Unmap(0, NULL);
D3D12_BOX box;
box.left = 0;
box.top = 0;
box.right = box.left + _rect.m_width;
box.bottom = box.top + _rect.m_height;
box.front = _z;
box.back = _z+_depth;
D3D12_TEXTURE_COPY_LOCATION dst = { m_ptr, D3D12_TEXTURE_COPY_TYPE_SUBRESOURCE_INDEX, { subres } };
D3D12_TEXTURE_COPY_LOCATION src = { staging, D3D12_TEXTURE_COPY_TYPE_PLACED_FOOTPRINT, layout };
_commandList->CopyTextureRegion(&dst, _rect.m_x, _rect.m_y, 0, &src, &box);
setState(_commandList, D3D12_RESOURCE_STATE_PIXEL_SHADER_RESOURCE);
s_renderD3D12->m_cmd.release(staging);
}
void TextureD3D12::commit(uint8_t _stage, uint32_t _flags)
{
BX_UNUSED(_stage, _flags);
}
void TextureD3D12::resolve()
{
}
D3D12_RESOURCE_STATES TextureD3D12::setState(ID3D12GraphicsCommandList* _commandList, D3D12_RESOURCE_STATES _state)
{
if (m_state != _state)
{
setResourceBarrier(_commandList
, m_ptr
, m_state
, _state
);
bx::xchg(m_state, _state);
}
return _state;
}
void FrameBufferD3D12::create(uint8_t _num, const TextureHandle* _handles)
{
m_numTh = _num;
memcpy(m_th, _handles, _num*sizeof(TextureHandle) );
postReset();
}
void FrameBufferD3D12::create(uint16_t /*_denseIdx*/, void* /*_nwh*/, uint32_t /*_width*/, uint32_t /*_height*/, TextureFormat::Enum /*_depthFormat*/)
{
}
void FrameBufferD3D12::preReset()
{
}
void FrameBufferD3D12::postReset()
{
if (m_numTh != 0)
{
ID3D12Device* device = s_renderD3D12->m_device;
D3D12_CPU_DESCRIPTOR_HANDLE rtvDescriptor = s_renderD3D12->m_rtvDescriptorHeap->GetCPUDescriptorHandleForHeapStart();
uint32_t rtvDescriptorSize = device->GetDescriptorHandleIncrementSize(D3D12_DESCRIPTOR_HEAP_TYPE_RTV);
uint32_t fbhIdx = (uint32_t)(this - s_renderD3D12->m_frameBuffers);
rtvDescriptor.ptr += (BX_COUNTOF(s_renderD3D12->m_backBufferColor) + fbhIdx * BGFX_CONFIG_MAX_FRAME_BUFFER_ATTACHMENTS) * rtvDescriptorSize;
m_depth.idx = bgfx::invalidHandle;
m_num = 0;
for (uint32_t ii = 0; ii < m_numTh; ++ii)
{
TextureHandle handle = m_th[ii];
if (isValid(handle) )
{
const TextureD3D12& texture = s_renderD3D12->m_textures[handle.idx];
if (isDepth( (TextureFormat::Enum)texture.m_textureFormat) )
{
BX_CHECK(!isValid(m_depth), "");
m_depth = handle;
D3D12_CPU_DESCRIPTOR_HANDLE dsvDescriptor = s_renderD3D12->m_dsvDescriptorHeap->GetCPUDescriptorHandleForHeapStart();
uint32_t dsvDescriptorSize = device->GetDescriptorHandleIncrementSize(D3D12_DESCRIPTOR_HEAP_TYPE_DSV);
dsvDescriptor.ptr += (1 + fbhIdx) * dsvDescriptorSize;
const ImageBlockInfo& blockInfo = getBlockInfo(TextureFormat::Enum(texture.m_textureFormat) );
BX_UNUSED(blockInfo);
D3D12_DEPTH_STENCIL_VIEW_DESC dsvDesc;
ZeroMemory(&dsvDesc, sizeof(dsvDesc) );
dsvDesc.Format = s_textureFormat[texture.m_textureFormat].m_fmtDsv;
dsvDesc.ViewDimension = D3D12_DSV_DIMENSION_TEXTURE2D;
dsvDesc.Flags = D3D12_DSV_FLAG_NONE
// | (blockInfo.depthBits > 0 ? D3D12_DSV_FLAG_READ_ONLY_DEPTH : D3D12_DSV_FLAG_NONE)
// | (blockInfo.stencilBits > 0 ? D3D12_DSV_FLAG_READ_ONLY_STENCIL : D3D12_DSV_FLAG_NONE)
;
device->CreateDepthStencilView(texture.m_ptr
, &dsvDesc
, dsvDescriptor
);
}
else
{
m_texture[m_num] = handle;
D3D12_CPU_DESCRIPTOR_HANDLE rtv = { rtvDescriptor.ptr + m_num * rtvDescriptorSize };
device->CreateRenderTargetView(texture.m_ptr
, NULL
, rtv
);
m_num++;
}
}
}
}
}
uint16_t FrameBufferD3D12::destroy()
{
m_numTh = 0;
m_depth.idx = bgfx::invalidHandle;
uint16_t denseIdx = m_denseIdx;
m_denseIdx = UINT16_MAX;
return denseIdx;
}
void FrameBufferD3D12::resolve()
{
}
void FrameBufferD3D12::clear(ID3D12GraphicsCommandList* _commandList, const Clear& _clear, const float _palette[][4], const D3D12_RECT* _rect, uint32_t _num)
{
ID3D12Device* device = s_renderD3D12->m_device;
const uint32_t fbhIdx = (uint32_t)(this - s_renderD3D12->m_frameBuffers);
if (BGFX_CLEAR_COLOR & _clear.m_flags
&& 0 != m_num)
{
D3D12_CPU_DESCRIPTOR_HANDLE rtvDescriptor = s_renderD3D12->m_rtvDescriptorHeap->GetCPUDescriptorHandleForHeapStart();
uint32_t rtvDescriptorSize = device->GetDescriptorHandleIncrementSize(D3D12_DESCRIPTOR_HEAP_TYPE_RTV);
rtvDescriptor.ptr += (BX_COUNTOF(s_renderD3D12->m_backBufferColor) + fbhIdx * BGFX_CONFIG_MAX_FRAME_BUFFER_ATTACHMENTS) * rtvDescriptorSize;
if (BGFX_CLEAR_COLOR_USE_PALETTE & _clear.m_flags)
{
for (uint32_t ii = 0, num = m_num; ii < num; ++ii)
{
uint8_t index = _clear.m_index[ii];
if (UINT8_MAX != index)
{
D3D12_CPU_DESCRIPTOR_HANDLE rtv = { rtvDescriptor.ptr + ii * rtvDescriptorSize };
_commandList->ClearRenderTargetView(rtv
, _palette[index]
, _num
, _rect
);
}
}
}
else
{
float frgba[4] =
{
_clear.m_index[0]*1.0f/255.0f,
_clear.m_index[1]*1.0f/255.0f,
_clear.m_index[2]*1.0f/255.0f,
_clear.m_index[3]*1.0f/255.0f,
};
for (uint32_t ii = 0, num = m_num; ii < num; ++ii)
{
D3D12_CPU_DESCRIPTOR_HANDLE rtv = { rtvDescriptor.ptr + ii * rtvDescriptorSize };
_commandList->ClearRenderTargetView(rtv
, frgba
, _num
, _rect
);
}
}
}
if (isValid(m_depth)
&& (BGFX_CLEAR_DEPTH|BGFX_CLEAR_STENCIL) & _clear.m_flags)
{
D3D12_CPU_DESCRIPTOR_HANDLE dsvDescriptor = s_renderD3D12->m_dsvDescriptorHeap->GetCPUDescriptorHandleForHeapStart();
uint32_t dsvDescriptorSize = device->GetDescriptorHandleIncrementSize(D3D12_DESCRIPTOR_HEAP_TYPE_DSV);
dsvDescriptor.ptr += (1 + fbhIdx) * dsvDescriptorSize;
DWORD flags = 0;
flags |= (_clear.m_flags & BGFX_CLEAR_DEPTH) ? D3D12_CLEAR_FLAG_DEPTH : 0;
flags |= (_clear.m_flags & BGFX_CLEAR_STENCIL) ? D3D12_CLEAR_FLAG_STENCIL : 0;
_commandList->ClearDepthStencilView(dsvDescriptor
, D3D12_CLEAR_FLAGS(flags)
, _clear.m_depth
, _clear.m_stencil
, _num
, _rect
);
}
}
void RendererContextD3D12::submit(Frame* _render, ClearQuad& _clearQuad, TextVideoMemBlitter& _textVideoMemBlitter)
{
// PIX_BEGINEVENT(D3DCOLOR_RGBA(0xff, 0x00, 0x00, 0xff), L"rendererSubmit");
updateResolution(_render->m_resolution);
int64_t elapsed = -bx::getHPCounter();
int64_t captureElapsed = 0;
if (0 < _render->m_iboffset)
{
TransientIndexBuffer* ib = _render->m_transientIb;
m_indexBuffers[ib->handle.idx].update(m_commandList, 0, _render->m_iboffset, ib->data);
}
if (0 < _render->m_vboffset)
{
TransientVertexBuffer* vb = _render->m_transientVb;
m_vertexBuffers[vb->handle.idx].update(m_commandList, 0, _render->m_vboffset, vb->data);
}
_render->sort();
RenderDraw currentState;
currentState.clear();
currentState.m_flags = BGFX_STATE_NONE;
currentState.m_stencil = packStencil(BGFX_STENCIL_NONE, BGFX_STENCIL_NONE);
_render->m_hmdInitialized = false;
const bool hmdEnabled = false;
ViewState viewState(_render, hmdEnabled);
viewState.reset(_render, hmdEnabled);
// bool wireframe = !!(_render->m_debug&BGFX_DEBUG_WIREFRAME);
// bool scissorEnabled = false;
// setDebugWireframe(wireframe);
uint16_t programIdx = invalidHandle;
uint16_t currentSamplerStateIdx = invalidHandle;
uint32_t currentBindHash = 0;
ID3D12PipelineState* currentPso = NULL;
SortKey key;
uint8_t view = 0xff;
FrameBufferHandle fbh = BGFX_INVALID_HANDLE;
float alphaRef = 0.0f;
uint32_t blendFactor = 0;
const uint64_t pt = _render->m_debug&BGFX_DEBUG_WIREFRAME ? BGFX_STATE_PT_LINES : 0;
uint8_t primIndex = uint8_t(pt >> BGFX_STATE_PT_SHIFT);
PrimInfo prim = s_primInfo[primIndex];
bool wasCompute = false;
bool viewHasScissor = false;
Rect viewScissorRect;
viewScissorRect.clear();
BX_UNUSED(alphaRef);
uint32_t statsNumPrimsSubmitted[BX_COUNTOF(s_primInfo)] = {};
uint32_t statsNumPrimsRendered[BX_COUNTOF(s_primInfo)] = {};
uint32_t statsNumInstances[BX_COUNTOF(s_primInfo)] = {};
uint32_t statsNumIndices = 0;
uint32_t statsKeyType[2] = {};
m_backBufferColorIdx = m_swapChain->GetCurrentBackBufferIndex();
const uint64_t f0 = BGFX_STATE_BLEND_FUNC(BGFX_STATE_BLEND_FACTOR, BGFX_STATE_BLEND_FACTOR);
const uint64_t f1 = BGFX_STATE_BLEND_FUNC(BGFX_STATE_BLEND_INV_FACTOR, BGFX_STATE_BLEND_INV_FACTOR);
D3D12_GPU_DESCRIPTOR_HANDLE gpuHandle;
ScratchBufferD3D12& scratchBuffer = m_scratchBuffer[m_backBufferColorIdx];
scratchBuffer.reset(gpuHandle);
D3D12_GPU_VIRTUAL_ADDRESS gpuAddress = {};
StateCacheLru<D3D12_GPU_DESCRIPTOR_HANDLE, 64> bindLru;
setResourceBarrier(m_commandList
, m_backBufferColor[m_backBufferColorIdx]
, D3D12_RESOURCE_STATE_PRESENT
, D3D12_RESOURCE_STATE_RENDER_TARGET
);
if (0 == (_render->m_debug&BGFX_DEBUG_IFH) )
{
// uint8_t eye = 0;
// uint8_t restartState = 0;
viewState.m_rect = _render->m_rect[0];
int32_t numItems = _render->m_num;
for (int32_t item = 0, restartItem = numItems; item < numItems || restartItem < numItems;)
{
const bool isCompute = key.decode(_render->m_sortKeys[item], _render->m_viewRemap);
statsKeyType[isCompute]++;
const bool viewChanged = 0
|| key.m_view != view
|| item == numItems
;
const RenderItem& renderItem = _render->m_renderItem[_render->m_sortValues[item] ];
++item;
if (viewChanged)
{
kick();
if (isCompute)
{
m_commandList->SetComputeRootSignature(m_rootSignature);
}
else
{
m_commandList->SetGraphicsRootSignature(m_rootSignature);
}
ID3D12DescriptorHeap* heaps[] = {
m_samplerAllocator.getHeap(),
scratchBuffer.getHeap(),
};
m_commandList->SetDescriptorHeaps(BX_COUNTOF(heaps), heaps);
view = key.m_view;
programIdx = invalidHandle;
currentPso = NULL;
fbh = _render->m_fb[view];
setFrameBuffer(fbh);
viewState.m_rect = _render->m_rect[view];
const Rect& rect = _render->m_rect[view];
const Rect& scissorRect = _render->m_scissor[view];
viewHasScissor = !scissorRect.isZero();
viewScissorRect = viewHasScissor ? scissorRect : rect;
D3D12_VIEWPORT vp;
vp.TopLeftX = rect.m_x;
vp.TopLeftY = rect.m_y;
vp.Width = rect.m_width;
vp.Height = rect.m_height;
vp.MinDepth = 0.0f;
vp.MaxDepth = 1.0f;
m_commandList->RSSetViewports(1, &vp);
D3D12_RECT rc;
rc.left = viewScissorRect.m_x;
rc.top = viewScissorRect.m_y;
rc.right = viewScissorRect.m_x + viewScissorRect.m_width;
rc.bottom = viewScissorRect.m_y + viewScissorRect.m_height;
m_commandList->RSSetScissorRects(1, &rc);
Clear& clear = _render->m_clear[view];
if (BGFX_CLEAR_NONE != clear.m_flags)
{
Rect clearRect = rect;
clearRect.intersect(rect, viewScissorRect);
clearQuad(_clearQuad, clearRect, clear, _render->m_clearColor);
}
prim = s_primInfo[BX_COUNTOF(s_primName)]; // Force primitive type update.
}
if (isCompute)
{
if (!wasCompute)
{
wasCompute = true;
}
const RenderCompute& compute = renderItem.compute;
bool programChanged = false;
bool constantsChanged = compute.m_constBegin < compute.m_constEnd;
rendererUpdateUniforms(this, _render->m_constantBuffer, compute.m_constBegin, compute.m_constEnd);
if (key.m_program != programIdx)
{
programIdx = key.m_program;
ProgramD3D12& program = m_program[key.m_program];
m_currentProgram = &program;
programChanged =
constantsChanged = true;
}
if (invalidHandle != programIdx)
{
ProgramD3D12& program = m_program[programIdx];
if (constantsChanged)
{
ConstantBuffer* vcb = program.m_vsh->m_constantBuffer;
if (NULL != vcb)
{
commit(*vcb);
}
}
viewState.setPredefined<4>(this, view, 0, program, _render, compute);
if (constantsChanged
|| program.m_numPredefined > 0)
{
commitShaderConstants(gpuAddress);
}
}
if (programChanged)
{
ID3D12PipelineState* pso = getPipelineState(programIdx);
m_commandList->SetPipelineState(pso);
}
D3D12_GPU_DESCRIPTOR_HANDLE srvHandle[BGFX_MAX_COMPUTE_BINDINGS] = {};
uint32_t samplerFlags[BGFX_MAX_COMPUTE_BINDINGS] = {};
for (uint32_t ii = 0; ii < BGFX_MAX_COMPUTE_BINDINGS; ++ii)
{
const Binding& bind = compute.m_bind[ii];
if (invalidHandle != bind.m_idx)
{
switch (bind.m_type)
{
case Binding::Image:
{
TextureD3D12& texture = m_textures[bind.m_idx];
if (Access::Read != bind.m_un.m_compute.m_access)
{
scratchBuffer.allocUav(srvHandle[ii], texture, bind.m_un.m_compute.m_mip);
}
else
{
scratchBuffer.alloc(srvHandle[ii], texture);
}
}
break;
case Binding::IndexBuffer:
case Binding::VertexBuffer:
{
BufferD3D12& buffer = Binding::IndexBuffer == bind.m_type
? m_indexBuffers[bind.m_idx]
: m_vertexBuffers[bind.m_idx]
;
if (Access::Read != bind.m_un.m_compute.m_access)
{
scratchBuffer.allocUav(srvHandle[ii], buffer);
}
else
{
scratchBuffer.alloc(srvHandle[ii], buffer);
}
}
break;
}
}
}
uint16_t samplerStateIdx = getSamplerState(samplerFlags, BGFX_MAX_COMPUTE_BINDINGS);
m_commandList->SetComputeRootDescriptorTable(Rdt::Sampler, m_samplerAllocator.get(samplerStateIdx));
m_commandList->SetComputeRootDescriptorTable(Rdt::SRV, srvHandle[0]);
m_commandList->SetComputeRootConstantBufferView(Rdt::CBV, gpuAddress);
m_commandList->SetComputeRootDescriptorTable(Rdt::UAV, srvHandle[0]);
if (isValid(compute.m_indirectBuffer) )
{
const VertexBufferD3D12& vb = m_vertexBuffers[compute.m_indirectBuffer.idx];
uint32_t numDrawIndirect = UINT16_MAX == compute.m_numIndirect
? vb.m_size/BGFX_CONFIG_DRAW_INDIRECT_STRIDE
: compute.m_numIndirect
;
uint32_t args = compute.m_startIndirect * BGFX_CONFIG_DRAW_INDIRECT_STRIDE;
for (uint32_t ii = 0; ii < numDrawIndirect; ++ii)
{
// deviceCtx->DispatchIndirect(ptr, args);
args += BGFX_CONFIG_DRAW_INDIRECT_STRIDE;
}
}
else
{
m_commandList->Dispatch(compute.m_numX, compute.m_numY, compute.m_numZ);
}
continue;
}
bool resetState = viewChanged || wasCompute;
if (wasCompute)
{
if (BX_ENABLED(BGFX_CONFIG_DEBUG_PIX) )
{
// wchar_t* viewNameW = s_viewNameW[view];
// viewNameW[3] = L' ';
// PIX_ENDEVENT();
// PIX_BEGINEVENT(D3DCOLOR_RGBA(0xff, 0x00, 0x00, 0xff), viewNameW);
}
wasCompute = false;
programIdx = invalidHandle;
m_currentProgram = NULL;
currentSamplerStateIdx = invalidHandle;
m_commandList->SetGraphicsRootSignature(m_rootSignature);
ID3D12DescriptorHeap* heaps[] = {
m_samplerAllocator.getHeap(),
scratchBuffer.getHeap(),
};
m_commandList->SetDescriptorHeaps(BX_COUNTOF(heaps), heaps);
// invalidateCompute();
}
const RenderDraw& draw = renderItem.draw;
const uint64_t newFlags = draw.m_flags;
uint64_t changedFlags = currentState.m_flags ^ draw.m_flags;
currentState.m_flags = newFlags;
const uint64_t newStencil = draw.m_stencil;
uint64_t changedStencil = (currentState.m_stencil ^ draw.m_stencil) & BGFX_STENCIL_FUNC_REF_MASK;
currentState.m_stencil = newStencil;
if (resetState)
{
currentState.clear();
currentState.m_scissor = !draw.m_scissor;
changedFlags = BGFX_STATE_MASK;
changedStencil = packStencil(BGFX_STENCIL_MASK, BGFX_STENCIL_MASK);
currentState.m_flags = newFlags;
currentState.m_stencil = newStencil;
const uint64_t pt = newFlags&BGFX_STATE_PT_MASK;
primIndex = uint8_t(pt>>BGFX_STATE_PT_SHIFT);
}
bool constantsChanged = draw.m_constBegin < draw.m_constEnd;
rendererUpdateUniforms(this, _render->m_constantBuffer, draw.m_constBegin, draw.m_constEnd);
if (0 != changedStencil)
{
const uint32_t fstencil = unpackStencil(0, draw.m_stencil);
const uint32_t ref = (fstencil&BGFX_STENCIL_FUNC_REF_MASK)>>BGFX_STENCIL_FUNC_REF_SHIFT;
m_commandList->OMSetStencilRef(ref);
}
const uint64_t state = draw.m_flags;
bool hasFactor = 0
|| f0 == (state & f0)
|| f1 == (state & f1)
;
if (hasFactor
&& blendFactor != draw.m_rgba)
{
blendFactor = draw.m_rgba;
float blendFactor[4] = { 1.0f, 1.0f, 1.0f, 1.0f };
blendFactor[0] = ( (draw.m_rgba>>24) )/255.0f;
blendFactor[1] = ( (draw.m_rgba>>16)&0xff)/255.0f;
blendFactor[2] = ( (draw.m_rgba>> 8)&0xff)/255.0f;
blendFactor[3] = ( (draw.m_rgba )&0xff)/255.0f;
m_commandList->OMSetBlendFactor(blendFactor);
}
if (BGFX_STATE_PT_MASK & changedFlags)
{
const uint64_t pt = draw.m_flags&BGFX_STATE_PT_MASK;
primIndex = uint8_t(pt >> BGFX_STATE_PT_SHIFT);
if (prim.m_toplogy != s_primInfo[primIndex].m_toplogy)
{
prim = s_primInfo[primIndex];
m_commandList->IASetPrimitiveTopology(prim.m_toplogy);
}
}
if (isValid(draw.m_vertexBuffer) )
{
bool programChanged = false;
if (key.m_program != programIdx)
{
programIdx = key.m_program;
if (invalidHandle == programIdx)
{
m_currentProgram = NULL;
}
else
{
ProgramD3D12& program = m_program[programIdx];
m_currentProgram = &program;
}
programChanged =
constantsChanged = true;
}
if (invalidHandle != programIdx)
{
ProgramD3D12& program = m_program[programIdx];
if (constantsChanged)
{
ConstantBuffer* vcb = program.m_vsh->m_constantBuffer;
if (NULL != vcb)
{
commit(*vcb);
}
ConstantBuffer* fcb = program.m_fsh->m_constantBuffer;
if (NULL != fcb)
{
commit(*fcb);
}
}
viewState.setPredefined<4>(this, view, 0, program, _render, draw);
constantsChanged |= program.m_numPredefined > 0;
if (constantsChanged)
{
commitShaderConstants(gpuAddress);
}
}
const VertexBufferD3D12& vb = m_vertexBuffers[draw.m_vertexBuffer.idx];
uint16_t declIdx = !isValid(vb.m_decl) ? draw.m_vertexDecl.idx : vb.m_decl.idx;
const VertexDecl& vertexDecl = m_vertexDecls[declIdx];
ID3D12PipelineState* pso =
getPipelineState(state
, draw.m_stencil
, declIdx
, programIdx
, draw.m_instanceDataStride/16
);
if (pso != currentPso)
{
currentPso = pso;
m_commandList->SetPipelineState(pso);
currentBindHash = 0;
}
for (uint32_t bindHash = bx::hashMurmur2A(draw.m_bind, sizeof(draw.m_bind) )
; currentBindHash != bindHash
; currentBindHash = bindHash
)
{
D3D12_GPU_DESCRIPTOR_HANDLE* srv = bindLru.find(bindHash);
if (NULL == srv)
{
D3D12_GPU_DESCRIPTOR_HANDLE srvHandle[BGFX_CONFIG_MAX_TEXTURE_SAMPLERS];
uint32_t samplerFlags[BGFX_CONFIG_MAX_TEXTURE_SAMPLERS];
{
srvHandle[0].ptr = 0;
for (uint32_t stage = 0; stage < BGFX_CONFIG_MAX_TEXTURE_SAMPLERS; ++stage)
{
const Binding& sampler = draw.m_bind[stage];
if (invalidHandle != sampler.m_idx)
{
TextureD3D12& texture = m_textures[sampler.m_idx];
scratchBuffer.alloc(srvHandle[stage], texture);
samplerFlags[stage] = (0 == (BGFX_SAMPLER_DEFAULT_FLAGS & sampler.m_un.m_draw.m_flags)
? sampler.m_un.m_draw.m_flags
: texture.m_flags
) & BGFX_TEXTURE_SAMPLER_BITS_MASK
;
}
else
{
memcpy(&srvHandle[stage], &srvHandle[0], sizeof(D3D12_GPU_DESCRIPTOR_HANDLE) );
samplerFlags[stage] = 0;
}
}
}
if (srvHandle[0].ptr != 0)
{
uint16_t samplerStateIdx = getSamplerState(samplerFlags);
if (samplerStateIdx != currentSamplerStateIdx)
{
currentSamplerStateIdx = samplerStateIdx;
m_commandList->SetGraphicsRootDescriptorTable(Rdt::Sampler, m_samplerAllocator.get(samplerStateIdx) );
}
m_commandList->SetGraphicsRootDescriptorTable(Rdt::SRV, srvHandle[0]);
bindLru.add(bindHash, srvHandle[0], 0);
}
}
else
{
m_commandList->SetGraphicsRootDescriptorTable(Rdt::SRV, *srv);
}
}
if (constantsChanged)
{
m_commandList->SetGraphicsRootConstantBufferView(Rdt::CBV, gpuAddress);
}
uint32_t numVertices = draw.m_numVertices;
if (UINT32_MAX == numVertices)
{
numVertices = vb.m_size / vertexDecl.m_stride;
}
if (currentState.m_vertexDecl.idx != draw.m_vertexDecl.idx
|| currentState.m_vertexBuffer.idx != draw.m_vertexBuffer.idx
|| currentState.m_instanceDataBuffer.idx != draw.m_instanceDataBuffer.idx
|| currentState.m_instanceDataOffset != draw.m_instanceDataOffset
|| currentState.m_instanceDataStride != draw.m_instanceDataStride)
{
currentState.m_vertexDecl = draw.m_vertexDecl;
currentState.m_vertexBuffer = draw.m_vertexBuffer;
currentState.m_instanceDataBuffer.idx = draw.m_instanceDataBuffer.idx;
currentState.m_instanceDataOffset = draw.m_instanceDataOffset;
currentState.m_instanceDataStride = draw.m_instanceDataStride;
D3D12_VERTEX_BUFFER_VIEW vbView[2];
uint32_t numVertexBuffers = 1;
vbView[0].BufferLocation = vb.m_gpuVA;
vbView[0].StrideInBytes = vertexDecl.m_stride;
vbView[0].SizeInBytes = vb.m_size;
if (isValid(draw.m_instanceDataBuffer) )
{
const VertexBufferD3D12& inst = m_vertexBuffers[draw.m_instanceDataBuffer.idx];
vbView[1].BufferLocation = inst.m_gpuVA + draw.m_instanceDataOffset;
vbView[1].StrideInBytes = draw.m_instanceDataStride;
vbView[1].SizeInBytes = draw.m_numInstances * draw.m_instanceDataStride;
++numVertexBuffers;
}
m_commandList->IASetVertexBuffers(0, numVertexBuffers, vbView);
}
uint32_t numIndices = 0;
uint32_t numPrimsSubmitted = 0;
uint32_t numInstances = 0;
uint32_t numPrimsRendered = 0;
if (isValid(draw.m_indexBuffer) )
{
const BufferD3D12& ib = m_indexBuffers[draw.m_indexBuffer.idx];
const bool hasIndex16 = 0 == (ib.m_flags & BGFX_BUFFER_INDEX32);
if (currentState.m_indexBuffer.idx != draw.m_indexBuffer.idx)
{
currentState.m_indexBuffer = draw.m_indexBuffer;
D3D12_INDEX_BUFFER_VIEW ibv;
ibv.Format = hasIndex16
? DXGI_FORMAT_R16_UINT
: DXGI_FORMAT_R32_UINT
;
ibv.BufferLocation = ib.m_gpuVA;
ibv.SizeInBytes = ib.m_size;
m_commandList->IASetIndexBuffer(&ibv);
}
if (UINT32_MAX == draw.m_numIndices)
{
const uint32_t indexSize = hasIndex16 ? 2 : 4;
numIndices = ib.m_size / indexSize;
numPrimsSubmitted = numIndices / prim.m_div - prim.m_sub;
numInstances = draw.m_numInstances;
numPrimsRendered = numPrimsSubmitted*draw.m_numInstances;
m_commandList->DrawIndexedInstanced(numIndices
, draw.m_numInstances
, draw.m_startIndex
, draw.m_startVertex
, 0
);
}
else if (prim.m_min <= draw.m_numIndices)
{
numIndices = draw.m_numIndices;
numPrimsSubmitted = numIndices / prim.m_div - prim.m_sub;
numInstances = draw.m_numInstances;
numPrimsRendered = numPrimsSubmitted*draw.m_numInstances;
m_commandList->DrawIndexedInstanced(numIndices
, draw.m_numInstances
, draw.m_startIndex
, draw.m_startVertex
, 0
);
}
}
else
{
numPrimsSubmitted = numVertices / prim.m_div - prim.m_sub;
numInstances = draw.m_numInstances;
numPrimsRendered = numPrimsSubmitted*draw.m_numInstances;
m_commandList->DrawInstanced(numVertices
, draw.m_numInstances
, draw.m_startVertex
, 0
);
}
statsNumPrimsSubmitted[primIndex] += numPrimsSubmitted;
statsNumPrimsRendered[primIndex] += numPrimsRendered;
statsNumInstances[primIndex] += numInstances;
statsNumIndices += numIndices;
}
}
}
int64_t now = bx::getHPCounter();
elapsed += now;
static int64_t last = now;
int64_t frameTime = now - last;
last = now;
static int64_t min = frameTime;
static int64_t max = frameTime;
min = bx::int64_min(min, frameTime);
max = bx::int64_max(max, frameTime);
static int64_t presentMin = m_presentElapsed;
static int64_t presentMax = m_presentElapsed;
presentMin = bx::int64_min(presentMin, m_presentElapsed);
presentMax = bx::int64_max(presentMax, m_presentElapsed);
if (_render->m_debug & (BGFX_DEBUG_IFH | BGFX_DEBUG_STATS) )
{
// PIX_BEGINEVENT(D3DCOLOR_RGBA(0x40, 0x40, 0x40, 0xff), L"debugstats");
TextVideoMem& tvm = m_textVideoMem;
static int64_t next = now;
if (now >= next)
{
next = now + bx::getHPFrequency();
double freq = double(bx::getHPFrequency() );
double toMs = 1000.0 / freq;
tvm.clear();
uint16_t pos = 0;
tvm.printf(0, pos++, BGFX_CONFIG_DEBUG ? 0x89 : 0x8f
, " %s / " BX_COMPILER_NAME " / " BX_CPU_NAME " / " BX_ARCH_NAME " / " BX_PLATFORM_NAME " "
, getRendererName()
);
const DXGI_ADAPTER_DESC& desc = m_adapterDesc;
char description[BX_COUNTOF(desc.Description)];
wcstombs(description, desc.Description, BX_COUNTOF(desc.Description) );
tvm.printf(0, pos++, 0x0f, " Device: %s", description);
char dedicatedVideo[16];
bx::prettify(dedicatedVideo, BX_COUNTOF(dedicatedVideo), desc.DedicatedVideoMemory);
char dedicatedSystem[16];
bx::prettify(dedicatedSystem, BX_COUNTOF(dedicatedSystem), desc.DedicatedSystemMemory);
char sharedSystem[16];
bx::prettify(sharedSystem, BX_COUNTOF(sharedSystem), desc.SharedSystemMemory);
tvm.printf(0, pos++, 0x0f, " Memory: %s (video), %s (system), %s (shared)"
, dedicatedVideo
, dedicatedSystem
, sharedSystem
);
pos = 10;
tvm.printf(10, pos++, 0x8e, " Frame: % 7.3f, % 7.3f \x1f, % 7.3f \x1e [ms] / % 6.2f FPS"
, double(frameTime)*toMs
, double(min)*toMs
, double(max)*toMs
, freq/frameTime
);
tvm.printf(10, pos++, 0x8e, " Present: % 7.3f, % 7.3f \x1f, % 7.3f \x1e [ms]"
, double(m_presentElapsed)*toMs
, double(presentMin)*toMs
, double(presentMax)*toMs
);
char hmd[16];
bx::snprintf(hmd, BX_COUNTOF(hmd), ", [%c] HMD ", hmdEnabled ? '\xfe' : ' ');
const uint32_t msaa = (m_resolution.m_flags&BGFX_RESET_MSAA_MASK)>>BGFX_RESET_MSAA_SHIFT;
tvm.printf(10, pos++, 0x8e, " Reset flags: [%c] vsync, [%c] MSAAx%d%s, [%c] MaxAnisotropy "
, !!(m_resolution.m_flags&BGFX_RESET_VSYNC) ? '\xfe' : ' '
, 0 != msaa ? '\xfe' : ' '
, 1<<msaa
, ", no-HMD "
, !!(m_resolution.m_flags&BGFX_RESET_MAXANISOTROPY) ? '\xfe' : ' '
);
double elapsedCpuMs = double(elapsed)*toMs;
tvm.printf(10, pos++, 0x8e, " Submitted: %4d (draw %4d, compute %4d) / CPU %3.4f [ms]"
, _render->m_num
, statsKeyType[0]
, statsKeyType[1]
, elapsedCpuMs
);
for (uint32_t ii = 0; ii < BX_COUNTOF(s_primName); ++ii)
{
tvm.printf(10, pos++, 0x8e, " %9s: %7d (#inst: %5d), submitted: %7d"
, s_primName[ii]
, statsNumPrimsRendered[ii]
, statsNumInstances[ii]
, statsNumPrimsSubmitted[ii]
);
}
// if (NULL != m_renderdocdll)
// {
// tvm.printf(tvm.m_width-27, 0, 0x1f, " [F11 - RenderDoc capture] ");
// }
tvm.printf(10, pos++, 0x8e, " Indices: %7d", statsNumIndices);
tvm.printf(10, pos++, 0x8e, " DVB size: %7d", _render->m_vboffset);
tvm.printf(10, pos++, 0x8e, " DIB size: %7d", _render->m_iboffset);
pos++;
tvm.printf(10, pos++, 0x8e, " State cache: ");
tvm.printf(10, pos++, 0x8e, " PSO | Sampler | Bind | Queued ");
tvm.printf(10, pos++, 0x8e, " %6d | %6d | %6d | %6d"
, m_pipelineStateCache.getCount()
, m_samplerStateCache.getCount()
, bindLru.getCount()
, m_cmd.m_control.available()
);
pos++;
double captureMs = double(captureElapsed)*toMs;
tvm.printf(10, pos++, 0x8e, " Capture: %3.4f [ms]", captureMs);
uint8_t attr[2] = { 0x89, 0x8a };
uint8_t attrIndex = _render->m_waitSubmit < _render->m_waitRender;
tvm.printf(10, pos++, attr[attrIndex&1], " Submit wait: %3.4f [ms]", _render->m_waitSubmit*toMs);
tvm.printf(10, pos++, attr[(attrIndex+1)&1], " Render wait: %3.4f [ms]", _render->m_waitRender*toMs);
min = frameTime;
max = frameTime;
presentMin = m_presentElapsed;
presentMax = m_presentElapsed;
}
blit(this, _textVideoMemBlitter, tvm);
// PIX_ENDEVENT();
}
else if (_render->m_debug & BGFX_DEBUG_TEXT)
{
// PIX_BEGINEVENT(D3DCOLOR_RGBA(0x40, 0x40, 0x40, 0xff), L"debugtext");
blit(this, _textVideoMemBlitter, _render->m_textVideoMem);
// PIX_ENDEVENT();
}
setResourceBarrier(m_commandList
, m_backBufferColor[m_backBufferColorIdx]
, D3D12_RESOURCE_STATE_RENDER_TARGET
, D3D12_RESOURCE_STATE_PRESENT
);
m_backBufferColorFence[m_backBufferColorIdx] = kick();
}
} /* namespace d3d12 */ } // namespace bgfx
#else
namespace bgfx { namespace d3d12
{
RendererContextI* rendererCreate()
{
return NULL;
}
void rendererDestroy()
{
}
} /* namespace d3d12 */ } // namespace bgfx
#endif // BGFX_CONFIG_RENDERER_DIRECT3D12