void StateCache::LoadFromDisk()
{
std::string cache_filename = GetDiskShaderCacheFileName(API_D3D11, "pso", true, true);
PipelineStateCacheInserter inserter;
s_pso_disk_cache.OpenAndRead(cache_filename, inserter);
if (s_cache_is_corrupted)
{
// If a PSO fails to create, that means either:
// - The file itself is corrupt.
// - A driver/HW change has occured, causing the existing cache blobs to be invalid.
//
// In either case, we want to re-create the disk cache. This should not be a frequent occurence.
s_pso_disk_cache.Close();
s_gx_state_cache.m_small_pso_map.clear();
File::Delete(cache_filename);
s_pso_disk_cache.OpenAndRead(cache_filename, inserter);
s_cache_is_corrupted = false;
}
}
void VertexShaderCache::LoadShaderCache()
{
VertexShaderCacheInserter<VertexShaderUid> inserter;
g_vs_disk_cache.OpenAndRead(GetDiskShaderCacheFileName(APIType::D3D, "VS", true, true), inserter);
VertexShaderCacheInserter<UberShader::VertexShaderUid> uber_inserter;
g_uber_vs_disk_cache.OpenAndRead(GetDiskShaderCacheFileName(APIType::D3D, "UberVS", false, true),
uber_inserter);
}
void VertexShaderCache::Init()
{
const D3D11_INPUT_ELEMENT_DESC simpleelems[2] = {
{"POSITION", 0, DXGI_FORMAT_R32G32B32_FLOAT, 0, 0, D3D11_INPUT_PER_VERTEX_DATA, 0},
{"TEXCOORD", 0, DXGI_FORMAT_R32G32B32_FLOAT, 0, 12, D3D11_INPUT_PER_VERTEX_DATA, 0},
};
const D3D11_INPUT_ELEMENT_DESC clearelems[2] = {
{"POSITION", 0, DXGI_FORMAT_R32G32B32_FLOAT, 0, 0, D3D11_INPUT_PER_VERTEX_DATA, 0},
{"COLOR", 0, DXGI_FORMAT_R8G8B8A8_UNORM, 0, 12, D3D11_INPUT_PER_VERTEX_DATA, 0},
};
unsigned int cbsize = ROUND_UP(sizeof(VertexShaderConstants), 16); // must be a multiple of 16
D3D11_BUFFER_DESC cbdesc = CD3D11_BUFFER_DESC(cbsize, D3D11_BIND_CONSTANT_BUFFER,
D3D11_USAGE_DYNAMIC, D3D11_CPU_ACCESS_WRITE);
HRESULT hr = D3D::device->CreateBuffer(&cbdesc, nullptr, &vscbuf);
CHECK(hr == S_OK, "Create vertex shader constant buffer (size=%u)", cbsize);
D3D::SetDebugObjectName((ID3D11DeviceChild*)vscbuf,
"vertex shader constant buffer used to emulate the GX pipeline");
D3DBlob* blob;
D3D::CompileVertexShader(simple_shader_code, &blob);
D3D::device->CreateInputLayout(simpleelems, 2, blob->Data(), blob->Size(), &SimpleLayout);
SimpleVertexShader = D3D::CreateVertexShaderFromByteCode(blob);
if (SimpleLayout == nullptr || SimpleVertexShader == nullptr)
PanicAlert("Failed to create simple vertex shader or input layout at %s %d\n", __FILE__,
__LINE__);
blob->Release();
D3D::SetDebugObjectName((ID3D11DeviceChild*)SimpleVertexShader, "simple vertex shader");
D3D::SetDebugObjectName((ID3D11DeviceChild*)SimpleLayout, "simple input layout");
D3D::CompileVertexShader(clear_shader_code, &blob);
D3D::device->CreateInputLayout(clearelems, 2, blob->Data(), blob->Size(), &ClearLayout);
ClearVertexShader = D3D::CreateVertexShaderFromByteCode(blob);
if (ClearLayout == nullptr || ClearVertexShader == nullptr)
PanicAlert("Failed to create clear vertex shader or input layout at %s %d\n", __FILE__,
__LINE__);
blob->Release();
D3D::SetDebugObjectName((ID3D11DeviceChild*)ClearVertexShader, "clear vertex shader");
D3D::SetDebugObjectName((ID3D11DeviceChild*)ClearLayout, "clear input layout");
Clear();
if (!File::Exists(File::GetUserPath(D_SHADERCACHE_IDX)))
File::CreateDir(File::GetUserPath(D_SHADERCACHE_IDX));
SETSTAT(stats.numVertexShadersCreated, 0);
SETSTAT(stats.numVertexShadersAlive, 0);
std::string cache_filename =
StringFromFormat("%sdx11-%s-vs.cache", File::GetUserPath(D_SHADERCACHE_IDX).c_str(),
SConfig::GetInstance().m_strUniqueID.c_str());
VertexShaderCacheInserter inserter;
g_vs_disk_cache.OpenAndRead(cache_filename, inserter);
if (g_Config.bEnableShaderDebugging)
Clear();
last_entry = nullptr;
}
void GeometryShaderCache::Init()
{
unsigned int gbsize = ROUND_UP(sizeof(GeometryShaderConstants), 16); // must be a multiple of 16
D3D11_BUFFER_DESC gbdesc = CD3D11_BUFFER_DESC(gbsize, D3D11_BIND_CONSTANT_BUFFER, D3D11_USAGE_DYNAMIC, D3D11_CPU_ACCESS_WRITE);
HRESULT hr = D3D::device->CreateBuffer(&gbdesc, nullptr, &gscbuf);
CHECK(hr == S_OK, "Create geometry shader constant buffer (size=%u)", gbsize);
D3D::SetDebugObjectName((ID3D11DeviceChild*)gscbuf, "geometry shader constant buffer used to emulate the GX pipeline");
// used when drawing clear quads
ClearGeometryShader = D3D::CompileAndCreateGeometryShader(clear_shader_code);
CHECK(ClearGeometryShader != nullptr, "Create clear geometry shader");
D3D::SetDebugObjectName((ID3D11DeviceChild*)ClearGeometryShader, "clear geometry shader");
// used for buffer copy
CopyGeometryShader = D3D::CompileAndCreateGeometryShader(copy_shader_code);
CHECK(CopyGeometryShader != nullptr, "Create copy geometry shader");
D3D::SetDebugObjectName((ID3D11DeviceChild*)CopyGeometryShader, "copy geometry shader");
Clear();
if (!File::Exists(File::GetUserPath(D_SHADERCACHE_IDX)))
File::CreateDir(File::GetUserPath(D_SHADERCACHE_IDX));
std::string cache_filename = StringFromFormat("%sdx11-%s-gs.cache", File::GetUserPath(D_SHADERCACHE_IDX).c_str(),
SConfig::GetInstance().m_strUniqueID.c_str());
GeometryShaderCacheInserter inserter;
g_gs_disk_cache.OpenAndRead(cache_filename, inserter);
if (g_Config.bEnableShaderDebugging)
Clear();
last_entry = nullptr;
}
void StateCache::Init()
{
// Root signature isn't available at time of StateCache construction, so fill it in now.
gx_state_cache.m_current_pso_desc.pRootSignature = D3D::default_root_signature;
// Multi-sample configuration isn't available at time of StateCache construction, so fille it in
// now.
gx_state_cache.m_current_pso_desc.SampleDesc.Count = g_ActiveConfig.iMultisamples;
gx_state_cache.m_current_pso_desc.SampleDesc.Quality = 0;
if (!File::Exists(File::GetUserPath(D_SHADERCACHE_IDX)))
File::CreateDir(File::GetUserPath(D_SHADERCACHE_IDX));
std::string cache_filename =
StringFromFormat("%sdx12-%s-pso.cache", File::GetUserPath(D_SHADERCACHE_IDX).c_str(),
SConfig::GetInstance().m_strUniqueID.c_str());
PipelineStateCacheInserter inserter;
s_pso_disk_cache.OpenAndRead(cache_filename, inserter);
if (s_cache_is_corrupted)
{
// If a PSO fails to create, that means either:
// - The file itself is corrupt.
// - A driver/HW change has occurred, causing the existing cache blobs to be invalid.
//
// In either case, we want to re-create the disk cache. This should not be a frequent occurence.
s_pso_disk_cache.Close();
for (auto it : gx_state_cache.m_small_pso_map)
{
SAFE_RELEASE(it.second);
}
gx_state_cache.m_small_pso_map.clear();
File::Delete(cache_filename);
s_pso_disk_cache.OpenAndRead(cache_filename, inserter);
s_cache_is_corrupted = false;
}
}
void ProgramShaderCache::Init(void)
{
// We have to get the UBO alignment here because
// if we generate a buffer that isn't aligned
// then the UBO will fail.
if (g_ActiveConfig.backend_info.bSupportsGLSLUBO)
{
GLint Align;
glGetIntegerv(GL_UNIFORM_BUFFER_OFFSET_ALIGNMENT, &Align);
s_ps_data_size = C_PENVCONST_END * sizeof(float) * 4;
s_vs_data_size = C_VENVCONST_END * sizeof(float) * 4;
s_vs_data_offset = ROUND_UP(s_ps_data_size, Align);
s_ubo_buffer_size = ROUND_UP(s_ps_data_size, Align) + ROUND_UP(s_vs_data_size, Align);
// We multiply by *4*4 because we need to get down to basic machine units.
// So multiply by four to get how many floats we have from vec4s
// Then once more to get bytes
s_buffer = new StreamBuffer(GL_UNIFORM_BUFFER, UBO_LENGTH);
s_ubo_buffer = new u8[s_ubo_buffer_size];
memset(s_ubo_buffer, 0, s_ubo_buffer_size);
s_ubo_dirty = true;
}
// Read our shader cache, only if supported
if (g_ogl_config.bSupportsGLSLCache)
{
GLint Supported;
glGetIntegerv(GL_NUM_PROGRAM_BINARY_FORMATS, &Supported);
if(!Supported)
{
ERROR_LOG(VIDEO, "GL_ARB_get_program_binary is supported, but no binary format is known. So disable shader cache.");
g_ogl_config.bSupportsGLSLCache = false;
}
else
{
if (!File::Exists(File::GetUserPath(D_SHADERCACHE_IDX)))
File::CreateDir(File::GetUserPath(D_SHADERCACHE_IDX).c_str());
char cache_filename[MAX_PATH];
sprintf(cache_filename, "%sogl-%s-shaders.cache", File::GetUserPath(D_SHADERCACHE_IDX).c_str(),
SConfig::GetInstance().m_LocalCoreStartupParameter.m_strUniqueID.c_str());
ProgramShaderCacheInserter inserter;
g_program_disk_cache.OpenAndRead(cache_filename, inserter);
}
SETSTAT(stats.numPixelShadersAlive, pshaders.size());
}
CreateHeader();
CurrentProgram = 0;
last_entry = NULL;
}
void ShaderCache::Init()
{
// This class intentionally shares its shader cache files with DX11, as the shaders are (right
// now) identical.
// Reduces unnecessary compilation when switching between APIs.
s_last_geometry_shader_bytecode = {};
s_last_pixel_shader_bytecode = {};
s_last_vertex_shader_bytecode = {};
s_last_geometry_shader_uid = {};
s_last_pixel_shader_uid = {};
s_last_vertex_shader_uid = {};
// Ensure shader cache directory exists..
std::string shader_cache_path = File::GetUserPath(D_SHADERCACHE_IDX);
if (!File::Exists(shader_cache_path))
File::CreateDir(File::GetUserPath(D_SHADERCACHE_IDX));
std::string title_game_id = SConfig::GetInstance().m_strGameID.c_str();
std::string gs_cache_filename =
StringFromFormat("%sdx11-%s-gs.cache", shader_cache_path.c_str(), title_game_id.c_str());
std::string ps_cache_filename =
StringFromFormat("%sdx11-%s-ps.cache", shader_cache_path.c_str(), title_game_id.c_str());
std::string vs_cache_filename =
StringFromFormat("%sdx11-%s-vs.cache", shader_cache_path.c_str(), title_game_id.c_str());
ShaderCacheInserter<GeometryShaderUid, GsBytecodeCache, &s_gs_bytecode_cache> gs_inserter;
s_gs_disk_cache.OpenAndRead(gs_cache_filename, gs_inserter);
ShaderCacheInserter<PixelShaderUid, PsBytecodeCache, &s_ps_bytecode_cache> ps_inserter;
s_ps_disk_cache.OpenAndRead(ps_cache_filename, ps_inserter);
ShaderCacheInserter<VertexShaderUid, VsBytecodeCache, &s_vs_bytecode_cache> vs_inserter;
s_vs_disk_cache.OpenAndRead(vs_cache_filename, vs_inserter);
SETSTAT(stats.numPixelShadersAlive, static_cast<int>(s_ps_bytecode_cache.size()));
SETSTAT(stats.numPixelShadersCreated, static_cast<int>(s_ps_bytecode_cache.size()));
SETSTAT(stats.numVertexShadersAlive, static_cast<int>(s_vs_bytecode_cache.size()));
SETSTAT(stats.numVertexShadersCreated, static_cast<int>(s_vs_bytecode_cache.size()));
}
void PixelShaderCache::Init()
{
unsigned int cbsize = Common::AlignUp(static_cast<unsigned int>(sizeof(PixelShaderConstants)),
16); // must be a multiple of 16
D3D11_BUFFER_DESC cbdesc = CD3D11_BUFFER_DESC(cbsize, D3D11_BIND_CONSTANT_BUFFER,
D3D11_USAGE_DYNAMIC, D3D11_CPU_ACCESS_WRITE);
D3D::device->CreateBuffer(&cbdesc, nullptr, &pscbuf);
CHECK(pscbuf != nullptr, "Create pixel shader constant buffer");
D3D::SetDebugObjectName((ID3D11DeviceChild*)pscbuf,
"pixel shader constant buffer used to emulate the GX pipeline");
// used when drawing clear quads
s_ClearProgram = D3D::CompileAndCreatePixelShader(clear_program_code);
CHECK(s_ClearProgram != nullptr, "Create clear pixel shader");
D3D::SetDebugObjectName((ID3D11DeviceChild*)s_ClearProgram, "clear pixel shader");
// used for anaglyph stereoscopy
s_AnaglyphProgram = D3D::CompileAndCreatePixelShader(anaglyph_program_code);
CHECK(s_AnaglyphProgram != nullptr, "Create anaglyph pixel shader");
D3D::SetDebugObjectName((ID3D11DeviceChild*)s_AnaglyphProgram, "anaglyph pixel shader");
// used when copying/resolving the color buffer
s_ColorCopyProgram[0] = D3D::CompileAndCreatePixelShader(color_copy_program_code);
CHECK(s_ColorCopyProgram[0] != nullptr, "Create color copy pixel shader");
D3D::SetDebugObjectName((ID3D11DeviceChild*)s_ColorCopyProgram[0], "color copy pixel shader");
// used for color conversion
s_ColorMatrixProgram[0] = D3D::CompileAndCreatePixelShader(color_matrix_program_code);
CHECK(s_ColorMatrixProgram[0] != nullptr, "Create color matrix pixel shader");
D3D::SetDebugObjectName((ID3D11DeviceChild*)s_ColorMatrixProgram[0], "color matrix pixel shader");
// used for depth copy
s_DepthMatrixProgram[0] = D3D::CompileAndCreatePixelShader(depth_matrix_program);
CHECK(s_DepthMatrixProgram[0] != nullptr, "Create depth matrix pixel shader");
D3D::SetDebugObjectName((ID3D11DeviceChild*)s_DepthMatrixProgram[0], "depth matrix pixel shader");
Clear();
if (!File::Exists(File::GetUserPath(D_SHADERCACHE_IDX)))
File::CreateDir(File::GetUserPath(D_SHADERCACHE_IDX));
SETSTAT(stats.numPixelShadersCreated, 0);
SETSTAT(stats.numPixelShadersAlive, 0);
std::string cache_filename =
StringFromFormat("%sdx11-%s-ps.cache", File::GetUserPath(D_SHADERCACHE_IDX).c_str(),
SConfig::GetInstance().m_strGameID.c_str());
PixelShaderCacheInserter inserter;
g_ps_disk_cache.OpenAndRead(cache_filename, inserter);
last_entry = nullptr;
}
void PixelShaderCache::Init()
{
unsigned int cbsize = ((sizeof(PixelShaderConstants))&(~0xf))+0x10; // must be a multiple of 16
D3D11_BUFFER_DESC cbdesc = CD3D11_BUFFER_DESC(cbsize, D3D11_BIND_CONSTANT_BUFFER, D3D11_USAGE_DYNAMIC, D3D11_CPU_ACCESS_WRITE);
D3D::device->CreateBuffer(&cbdesc, NULL, &pscbuf);
CHECK(pscbuf!=NULL, "Create pixel shader constant buffer");
D3D::SetDebugObjectName((ID3D11DeviceChild*)pscbuf, "pixel shader constant buffer used to emulate the GX pipeline");
// used when drawing clear quads
s_ClearProgram = D3D::CompileAndCreatePixelShader(clear_program_code, sizeof(clear_program_code));
CHECK(s_ClearProgram!=NULL, "Create clear pixel shader");
D3D::SetDebugObjectName((ID3D11DeviceChild*)s_ClearProgram, "clear pixel shader");
// used when copying/resolving the color buffer
s_ColorCopyProgram[0] = D3D::CompileAndCreatePixelShader(color_copy_program_code, sizeof(color_copy_program_code));
CHECK(s_ColorCopyProgram[0]!=NULL, "Create color copy pixel shader");
D3D::SetDebugObjectName((ID3D11DeviceChild*)s_ColorCopyProgram[0], "color copy pixel shader");
// used for color conversion
s_ColorMatrixProgram[0] = D3D::CompileAndCreatePixelShader(color_matrix_program_code, sizeof(color_matrix_program_code));
CHECK(s_ColorMatrixProgram[0]!=NULL, "Create color matrix pixel shader");
D3D::SetDebugObjectName((ID3D11DeviceChild*)s_ColorMatrixProgram[0], "color matrix pixel shader");
// used for depth copy
s_DepthMatrixProgram[0] = D3D::CompileAndCreatePixelShader(depth_matrix_program, sizeof(depth_matrix_program));
CHECK(s_DepthMatrixProgram[0]!=NULL, "Create depth matrix pixel shader");
D3D::SetDebugObjectName((ID3D11DeviceChild*)s_DepthMatrixProgram[0], "depth matrix pixel shader");
Clear();
if (!File::Exists(File::GetUserPath(D_SHADERCACHE_IDX)))
File::CreateDir(File::GetUserPath(D_SHADERCACHE_IDX).c_str());
SETSTAT(stats.numPixelShadersCreated, 0);
SETSTAT(stats.numPixelShadersAlive, 0);
char cache_filename[MAX_PATH];
sprintf(cache_filename, "%sdx11-%s-ps.cache", File::GetUserPath(D_SHADERCACHE_IDX).c_str(),
SConfig::GetInstance().m_LocalCoreStartupParameter.m_strUniqueID.c_str());
PixelShaderCacheInserter inserter;
g_ps_disk_cache.OpenAndRead(cache_filename, inserter);
if (g_Config.bEnableShaderDebugging)
Clear();
last_entry = NULL;
}
void ProgramShaderCache::Init()
{
// We have to get the UBO alignment here because
// if we generate a buffer that isn't aligned
// then the UBO will fail.
glGetIntegerv(GL_UNIFORM_BUFFER_OFFSET_ALIGNMENT, &s_ubo_align);
s_ubo_buffer_size = ROUND_UP(PixelShaderManager::ConstantBufferSize * sizeof(float), s_ubo_align)
+ ROUND_UP(VertexShaderManager::ConstantBufferSize * sizeof(float), s_ubo_align)
+ ROUND_UP(sizeof(GeometryShaderConstants), s_ubo_align);
// We multiply by *4*4 because we need to get down to basic machine units.
// So multiply by four to get how many floats we have from vec4s
// Then once more to get bytes
s_buffer = StreamBuffer::Create(GL_UNIFORM_BUFFER, UBO_LENGTH);
// Read our shader cache, only if supported
if (g_ogl_config.bSupportsGLSLCache && !g_Config.bEnableShaderDebugging)
{
GLint Supported;
glGetIntegerv(GL_NUM_PROGRAM_BINARY_FORMATS, &Supported);
if (!Supported)
{
ERROR_LOG(VIDEO, "GL_ARB_get_program_binary is supported, but no binary format is known. So disable shader cache.");
g_ogl_config.bSupportsGLSLCache = false;
}
else
{
if (!File::Exists(File::GetUserPath(D_SHADERCACHE_IDX)))
File::CreateDir(File::GetUserPath(D_SHADERCACHE_IDX));
std::string cache_filename = StringFromFormat("%sIOGL-%s-shaders.cache", File::GetUserPath(D_SHADERCACHE_IDX).c_str(),
SConfig::GetInstance().m_strUniqueID.c_str());
ProgramShaderCacheInserter inserter;
g_program_disk_cache.OpenAndRead(cache_filename, inserter);
}
SETSTAT(stats.numPixelShadersAlive, pshaders.size());
}
CreateHeader();
CurrentProgram = 0;
last_entry = nullptr;
}
void GeometryShaderCache::Init()
{
unsigned int gscbuf12sizeInBytes = gscbuf12paddedSize * gscbuf12Slots;
CheckHR(
D3D::device12->CreateCommittedResource(
&CD3DX12_HEAP_PROPERTIES(D3D12_HEAP_TYPE_UPLOAD),
D3D12_HEAP_FLAG_NONE,
&CD3DX12_RESOURCE_DESC::Buffer(gscbuf12sizeInBytes),
D3D12_RESOURCE_STATE_GENERIC_READ,
nullptr,
IID_PPV_ARGS(&gscbuf12)
)
);
D3D::SetDebugObjectName12(gscbuf12, "vertex shader constant buffer used to emulate the GX pipeline");
// Obtain persistent CPU pointer to GS Constant Buffer
CheckHR(gscbuf12->Map(0, nullptr, &gscbuf12data));
// used when drawing clear quads
D3D::CompileGeometryShader(clear_shader_code, &ClearGeometryShaderBlob);
// used for buffer copy
D3D::CompileGeometryShader(copy_shader_code, &CopyGeometryShaderBlob);
Clear();
if (!File::Exists(File::GetUserPath(D_SHADERCACHE_IDX)))
File::CreateDir(File::GetUserPath(D_SHADERCACHE_IDX));
// Intentionally share the same cache as DX11, as the shaders are identical. Reduces recompilation when switching APIs.
std::string cache_filename = StringFromFormat("%sdx11-%s-gs.cache", File::GetUserPath(D_SHADERCACHE_IDX).c_str(),
SConfig::GetInstance().m_strUniqueID.c_str());
GeometryShaderCacheInserter inserter;
g_gs_disk_cache.OpenAndRead(cache_filename, inserter);
if (g_Config.bEnableShaderDebugging)
Clear();
last_entry = nullptr;
last_uid = {};
}
void VertexShaderCache::Init()
{
Compiler = &HLSLAsyncCompiler::getInstance();
vshaderslock.unlock();
const char* code = "struct VSOUTPUT\n"
"{\n"
"float4 vPosition : POSITION;\n"
"float2 vTexCoord : TEXCOORD0;\n"
"float vTexCoord1 : TEXCOORD1;\n"
"};\n"
"VSOUTPUT main(float4 inPosition : POSITION,float2 inTEX0 : TEXCOORD0,float2 inTEX1 : TEXCOORD1,float inTEX2 : TEXCOORD2)\n"
"{\n"
"VSOUTPUT OUT;\n"
"OUT.vPosition = inPosition;\n"
"OUT.vTexCoord = inTEX0;\n"
"OUT.vTexCoord1 = inTEX2;\n"
"return OUT;\n"
"}\0";
SimpleVertexShader[0] = D3D::CompileAndCreateVertexShader(code, (int)strlen(code));
code = "struct VSOUTPUT\n"
"{\n"
"float4 vPosition : POSITION;\n"
"float4 vColor0 : COLOR0;\n"
"};\n"
"VSOUTPUT main(float4 inPosition : POSITION,float4 inColor0: COLOR0)\n"
"{\n"
"VSOUTPUT OUT;\n"
"OUT.vPosition = inPosition;\n"
"OUT.vColor0 = inColor0;\n"
"return OUT;\n"
"}\0";
ClearVertexShader = D3D::CompileAndCreateVertexShader(code, (int)strlen(code));
code = "struct VSOUTPUT\n"
"{\n"
"float4 vPosition : POSITION;\n"
"float4 vTexCoord : TEXCOORD0;\n"
"float vTexCoord1 : TEXCOORD1;\n"
"float4 vTexCoord2 : TEXCOORD2;\n"
"float4 vTexCoord3 : TEXCOORD3;\n"
"};\n"
"VSOUTPUT main(float4 inPosition : POSITION,float2 inTEX0 : TEXCOORD0,float2 inTEX1 : TEXCOORD1,float inTEX2 : TEXCOORD2)\n"
"{\n"
"VSOUTPUT OUT;"
"OUT.vPosition = inPosition;\n"
"OUT.vTexCoord = inTEX0.xyyx;\n"
"OUT.vTexCoord1 = inTEX2.x;\n"
"OUT.vTexCoord2 = inTEX0.xyyx + (float4(-0.375f,-0.125f,-0.375f, 0.125f) * inTEX1.xyyx);\n"
"OUT.vTexCoord3 = inTEX0.xyyx + (float4( 0.375f, 0.125f, 0.375f,-0.125f) * inTEX1.xyyx);\n"
"return OUT;\n"
"}\0";
SimpleVertexShader[1] = D3D::CompileAndCreateVertexShader(code, (int)strlen(code));
Clear();
if (!File::Exists(File::GetUserPath(D_SHADERCACHE_IDX)))
File::CreateDir(File::GetUserPath(D_SHADERCACHE_IDX).c_str());
SETSTAT(stats.numVertexShadersCreated, 0);
SETSTAT(stats.numVertexShadersAlive, 0);
char cache_filename[MAX_PATH];
sprintf(cache_filename, "%sIDX9-%s-vs.cache", File::GetUserPath(D_SHADERCACHE_IDX).c_str(),
SConfig::GetInstance().m_strUniqueID.c_str());
VertexShaderCacheInserter inserter;
vshaderslock.lock();
g_vs_disk_cache.OpenAndRead(cache_filename, inserter);
vshaderslock.unlock();
if (g_Config.bEnableShaderDebugging)
Clear();
last_entry = NULL;
}
void GeometryShaderCache::LoadShaderCache()
{
GeometryShaderCacheInserter inserter;
g_gs_disk_cache.OpenAndRead(GetDiskShaderCacheFileName(APIType::D3D, "GS", true, true), inserter);
}
void VertexShaderCache::Init()
{
char* vProg = new char[2048];
sprintf(vProg,"struct VSOUTPUT\n"
"{\n"
"float4 vPosition : POSITION;\n"
"float2 vTexCoord : TEXCOORD0;\n"
"float vTexCoord1 : TEXCOORD1;\n"
"};\n"
"VSOUTPUT main(float4 inPosition : POSITION,float2 inTEX0 : TEXCOORD0,float2 inTEX1 : TEXCOORD1,float inTEX2 : TEXCOORD2)\n"
"{\n"
"VSOUTPUT OUT;\n"
"OUT.vPosition = inPosition;\n"
"OUT.vTexCoord = inTEX0;\n"
"OUT.vTexCoord1 = inTEX2;\n"
"return OUT;\n"
"}\n");
SimpleVertexShader[0] = D3D::CompileAndCreateVertexShader(vProg, (int)strlen(vProg));
sprintf(vProg,"struct VSOUTPUT\n"
"{\n"
"float4 vPosition : POSITION;\n"
"float4 vColor0 : COLOR0;\n"
"};\n"
"VSOUTPUT main(float4 inPosition : POSITION,float4 inColor0: COLOR0)\n"
"{\n"
"VSOUTPUT OUT;\n"
"OUT.vPosition = inPosition;\n"
"OUT.vColor0 = inColor0;\n"
"return OUT;\n"
"}\n");
ClearVertexShader = D3D::CompileAndCreateVertexShader(vProg, (int)strlen(vProg));
sprintf(vProg, "struct VSOUTPUT\n"
"{\n"
"float4 vPosition : POSITION;\n"
"float2 vTexCoord : TEXCOORD0;\n"
"float vTexCoord1 : TEXCOORD1;\n"
"};\n"
"VSOUTPUT main(float4 inPosition : POSITION,float2 inTEX0 : TEXCOORD0,float2 inInvTexSize : TEXCOORD1,float inTEX2 : TEXCOORD2)\n"
"{\n"
"VSOUTPUT OUT;"
"OUT.vPosition = inPosition;\n"
// HACK: Scale the texture coordinate range from (0,width) to (0,width-1), otherwise the linear filter won't average our samples correctly
"OUT.vTexCoord = inTEX0 * (float2(1.f,1.f) / inInvTexSize - float2(1.f,1.f)) * inInvTexSize;\n"
"OUT.vTexCoord1 = inTEX2;\n"
"return OUT;\n"
"}\n");
SimpleVertexShader[1] = D3D::CompileAndCreateVertexShader(vProg, (int)strlen(vProg));
sprintf(vProg, "struct VSOUTPUT\n"
"{\n"
"float4 vPosition : POSITION;\n"
"float4 vTexCoord : TEXCOORD0;\n"
"float vTexCoord1 : TEXCOORD1;\n"
"float4 vTexCoord2 : TEXCOORD2;\n"
"float4 vTexCoord3 : TEXCOORD3;\n"
"};\n"
"VSOUTPUT main(float4 inPosition : POSITION,float2 inTEX0 : TEXCOORD0,float2 inTEX1 : TEXCOORD1,float inTEX2 : TEXCOORD2)\n"
"{\n"
"VSOUTPUT OUT;"
"OUT.vPosition = inPosition;\n"
"OUT.vTexCoord = inTEX0.xyyx;\n"
"OUT.vTexCoord1 = inTEX2.x;\n"
"OUT.vTexCoord2 = inTEX0.xyyx + (float4(-1.0f,-0.5f, 1.0f,-0.5f) * inTEX1.xyyx);\n"
"OUT.vTexCoord3 = inTEX0.xyyx + (float4( 1.0f, 0.5f,-1.0f, 0.5f) * inTEX1.xyyx);\n"
"return OUT;\n"
"}\n");
SimpleVertexShader[2] = D3D::CompileAndCreateVertexShader(vProg, (int)strlen(vProg));
Clear();
delete [] vProg;
if (!File::Exists(File::GetUserPath(D_SHADERCACHE_IDX)))
File::CreateDir(File::GetUserPath(D_SHADERCACHE_IDX).c_str());
SETSTAT(stats.numVertexShadersCreated, 0);
SETSTAT(stats.numVertexShadersAlive, 0);
char cache_filename[MAX_PATH];
sprintf(cache_filename, "%sdx9-%s-vs.cache", File::GetUserPath(D_SHADERCACHE_IDX).c_str(),
SConfig::GetInstance().m_LocalCoreStartupParameter.m_strUniqueID.c_str());
VertexShaderCacheInserter inserter;
g_vs_disk_cache.OpenAndRead(cache_filename, inserter);
}
void ShaderCache::Init()
{
s_compiler = &HLSLAsyncCompiler::getInstance();
s_shaders_lock.unlock();
s_pass_entry.m_compiled = true;
s_pass_entry.m_initialized.test_and_set();
// This class intentionally shares its shader cache files with DX11, as the shaders are (right now) identical.
// Reduces unnecessary compilation when switching between APIs.
s_last_domain_shader_bytecode = &s_pass_entry;
s_last_hull_shader_bytecode = &s_pass_entry;
s_last_geometry_shader_bytecode = &s_pass_entry;
s_last_pixel_shader_bytecode = nullptr;
s_last_vertex_shader_bytecode = nullptr;
s_last_geometry_shader_uid = {};
s_last_pixel_shader_uid = {};
s_last_vertex_shader_uid = {};
s_last_tessellation_shader_uid = {};
s_last_cpu_geometry_shader_uid = {};
s_last_cpu_pixel_shader_uid = {};
s_last_cpu_vertex_shader_uid = {};
s_last_cpu_tessellation_shader_uid = {};
// Ensure shader cache directory exists..
std::string shader_cache_path = File::GetUserPath(D_SHADERCACHE_IDX);
if (!File::Exists(shader_cache_path))
File::CreateDir(File::GetUserPath(D_SHADERCACHE_IDX));
std::string title_unique_id = SConfig::GetInstance().m_strGameID;
std::string ds_cache_filename = StringFromFormat("%sIDX11-%s-ds.cache", shader_cache_path.c_str(), title_unique_id.c_str());
std::string hs_cache_filename = StringFromFormat("%sIDX11-%s-hs.cache", shader_cache_path.c_str(), title_unique_id.c_str());
std::string gs_cache_filename = StringFromFormat("%sIDX11-%s-gs.cache", shader_cache_path.c_str(), title_unique_id.c_str());
std::string ps_cache_filename = StringFromFormat("%sIDX11-%s-ps.cache", shader_cache_path.c_str(), title_unique_id.c_str());
std::string vs_cache_filename = StringFromFormat("%sIDX11-%s-vs.cache", shader_cache_path.c_str(), title_unique_id.c_str());
pKey_t gameid = (pKey_t)GetMurmurHash3(reinterpret_cast<const u8*>(SConfig::GetInstance().m_strGameID.data()), (u32)SConfig::GetInstance().m_strGameID.size(), 0);
vs_bytecode_cache = VsBytecodeCache::Create(
gameid,
VERTEXSHADERGEN_UID_VERSION,
"Ishiiruka.vs",
StringFromFormat("%s.vs", title_unique_id.c_str())
);
ps_bytecode_cache = PsBytecodeCache::Create(
gameid,
PIXELSHADERGEN_UID_VERSION,
"Ishiiruka.ps",
StringFromFormat("%s.ps", title_unique_id.c_str())
);
gs_bytecode_cache = GsBytecodeCache::Create(
gameid,
GEOMETRYSHADERGEN_UID_VERSION,
"Ishiiruka.gs",
StringFromFormat("%s.gs", title_unique_id.c_str())
);
ts_bytecode_cache = TsBytecodeCache::Create(
gameid,
TESSELLATIONSHADERGEN_UID_VERSION,
"Ishiiruka.ts",
StringFromFormat("%s.ts", title_unique_id.c_str())
);
DShaderCacheInserter ds_inserter;
s_ds_disk_cache.OpenAndRead(ds_cache_filename, ds_inserter);
HShaderCacheInserter hs_inserter;
s_hs_disk_cache.OpenAndRead(hs_cache_filename, hs_inserter);
ShaderCacheInserter<GeometryShaderUid, GsBytecodeCache, &gs_bytecode_cache> gs_inserter;
s_gs_disk_cache.OpenAndRead(gs_cache_filename, gs_inserter);
ShaderCacheInserter<PixelShaderUid, PsBytecodeCache, &ps_bytecode_cache> ps_inserter;
s_ps_disk_cache.OpenAndRead(ps_cache_filename, ps_inserter);
ShaderCacheInserter<VertexShaderUid, VsBytecodeCache, &vs_bytecode_cache> vs_inserter;
s_vs_disk_cache.OpenAndRead(vs_cache_filename, vs_inserter);
// Clear out disk cache when debugging shaders to ensure stale ones don't stick around..
SETSTAT(stats.numGeometryShadersAlive, static_cast<int>(gs_bytecode_cache->size()));
SETSTAT(stats.numGeometryShadersCreated, 0);
SETSTAT(stats.numPixelShadersAlive, static_cast<int>(ps_bytecode_cache->size()));
SETSTAT(stats.numPixelShadersCreated, 0);
SETSTAT(stats.numVertexShadersAlive, static_cast<int>(vs_bytecode_cache->size()));
SETSTAT(stats.numVertexShadersCreated, 0);
if (g_ActiveConfig.bCompileShaderOnStartup)
{
size_t shader_count = 0;
ps_bytecode_cache->ForEachMostUsedByCategory(gameid,
[&](const PixelShaderUid& it, size_t total)
{
PixelShaderUid item = it;
item.ClearHASH();
item.CalculateUIDHash();
HandlePSUIDChange(item, true);
shader_count++;
if ((shader_count & 7) == 0)
{
Host_UpdateTitle(StringFromFormat("Compiling Pixel Shaders %i %% (%i/%i)", (shader_count * 100) / total, shader_count, total));
s_compiler->WaitForFinish();
}
},
[](ByteCodeCacheEntry& entry)
{
return !entry.m_shader_bytecode.pShaderBytecode;
}
, true);
shader_count = 0;
vs_bytecode_cache->ForEachMostUsedByCategory(gameid,
[&](const VertexShaderUid& it, size_t total)
{
VertexShaderUid item = it;
item.ClearHASH();
item.CalculateUIDHash();
HandleVSUIDChange(item, true);
shader_count++;
if ((shader_count & 31) == 0)
{
Host_UpdateTitle(StringFromFormat("Compiling Vertex Shaders %i %% (%i/%i)", (shader_count * 100) / total, shader_count, total));
s_compiler->WaitForFinish();
}
},
[](ByteCodeCacheEntry& entry)
{
return !entry.m_shader_bytecode.pShaderBytecode;
}
, true);
shader_count = 0;
gs_bytecode_cache->ForEachMostUsedByCategory(gameid,
[&](const GeometryShaderUid& it, size_t total)
{
GeometryShaderUid item = it;
item.ClearHASH();
item.CalculateUIDHash();
HandleGSUIDChange(item, true);
shader_count++;
if ((shader_count & 7) == 0)
{
Host_UpdateTitle(StringFromFormat("Compiling Geometry Shaders %i %% (%i/%i)", (shader_count * 100) / total, shader_count, total));
s_compiler->WaitForFinish();
}
},
[](ByteCodeCacheEntry& entry)
{
return !entry.m_shader_bytecode.pShaderBytecode;
}
, true);
shader_count = 0;
ts_bytecode_cache->ForEachMostUsedByCategory(gameid,
[&](const TessellationShaderUid& it, size_t total)
{
TessellationShaderUid item = it;
item.ClearHASH();
item.CalculateUIDHash();
HandleTSUIDChange(item, true);
shader_count++;
if ((shader_count & 31) == 0)
{
Host_UpdateTitle(StringFromFormat("Compiling Tessellation Shaders %i %% (%i/%i)", (shader_count * 100) / total, shader_count, total));
s_compiler->WaitForFinish();
}
},
[](std::pair<ByteCodeCacheEntry, ByteCodeCacheEntry>& entry)
{
return !entry.first.m_shader_bytecode.pShaderBytecode;
}
, true);
s_compiler->WaitForFinish();
}
}
void VertexShaderCache::Init()
{
char* vProg = new char[2048];
sprintf(vProg,"struct VSOUTPUT\n"
"{\n"
"float4 vPosition : POSITION;\n"
"float2 vTexCoord : TEXCOORD0;\n"
"float vTexCoord1 : TEXCOORD1;\n"
"};\n"
"VSOUTPUT main(float4 inPosition : POSITION,float2 inTEX0 : TEXCOORD0,float2 inTEX1 : TEXCOORD1,float inTEX2 : TEXCOORD2)\n"
"{\n"
"VSOUTPUT OUT;\n"
"OUT.vPosition = inPosition;\n"
"OUT.vTexCoord = inTEX0;\n"
"OUT.vTexCoord1 = inTEX2;\n"
"return OUT;\n"
"}\n");
SimpleVertexShader[0] = D3D::CompileAndCreateVertexShader(vProg, (int)strlen(vProg));
sprintf(vProg,"struct VSOUTPUT\n"
"{\n"
"float4 vPosition : POSITION;\n"
"float4 vColor0 : COLOR0;\n"
"};\n"
"VSOUTPUT main(float4 inPosition : POSITION,float4 inColor0: COLOR0)\n"
"{\n"
"VSOUTPUT OUT;\n"
"OUT.vPosition = inPosition;\n"
"OUT.vColor0 = inColor0;\n"
"return OUT;\n"
"}\n");
ClearVertexShader = D3D::CompileAndCreateVertexShader(vProg, (int)strlen(vProg));
sprintf(vProg, "struct VSOUTPUT\n"
"{\n"
"float4 vPosition : POSITION;\n"
"float4 vTexCoord : TEXCOORD0;\n"
"float vTexCoord1 : TEXCOORD1;\n"
"float4 vTexCoord2 : TEXCOORD2;\n"
"float4 vTexCoord3 : TEXCOORD3;\n"
"};\n"
"VSOUTPUT main(float4 inPosition : POSITION,float2 inTEX0 : TEXCOORD0,float2 inTEX1 : TEXCOORD1,float inTEX2 : TEXCOORD2)\n"
"{\n"
"VSOUTPUT OUT;"
"OUT.vPosition = inPosition;\n"
"OUT.vTexCoord = inTEX0.xyyx;\n"
"OUT.vTexCoord1 = inTEX2.x;\n"
"OUT.vTexCoord2 = inTEX0.xyyx + (float4(-0.495f,-0.495f, 0.495f,-0.495f) * inTEX1.xyyx);\n"
"OUT.vTexCoord3 = inTEX0.xyyx + (float4( 0.495f, 0.495f,-0.495f, 0.495f) * inTEX1.xyyx);\n"
"return OUT;\n"
"}\n");
SimpleVertexShader[1] = D3D::CompileAndCreateVertexShader(vProg, (int)strlen(vProg));
sprintf(vProg, "struct VSOUTPUT\n"
"{\n"
"float4 vPosition : POSITION;\n"
"float4 vTexCoord : TEXCOORD0;\n"
"float vTexCoord1 : TEXCOORD1;\n"
"float4 vTexCoord2 : TEXCOORD2;\n"
"float4 vTexCoord3 : TEXCOORD3;\n"
"};\n"
"VSOUTPUT main(float4 inPosition : POSITION,float2 inTEX0 : TEXCOORD0,float2 inTEX1 : TEXCOORD1,float inTEX2 : TEXCOORD2)\n"
"{\n"
"VSOUTPUT OUT;"
"OUT.vPosition = inPosition;\n"
"OUT.vTexCoord = inTEX0.xyyx;\n"
"OUT.vTexCoord1 = inTEX2.x;\n"
"OUT.vTexCoord2 = inTEX0.xyyx + (float4(-0.9f,-0.45f, 0.9f,-0.45f) * inTEX1.xyyx);\n"
"OUT.vTexCoord3 = inTEX0.xyyx + (float4( 0.9f, 0.45f,-0.9f, 0.45f) * inTEX1.xyyx);\n"
"return OUT;\n"
"}\n");
SimpleVertexShader[2] = D3D::CompileAndCreateVertexShader(vProg, (int)strlen(vProg));
Clear();
delete [] vProg;
if (!File::Exists(File::GetUserPath(D_SHADERCACHE_IDX)))
File::CreateDir(File::GetUserPath(D_SHADERCACHE_IDX).c_str());
SETSTAT(stats.numVertexShadersCreated, 0);
SETSTAT(stats.numVertexShadersAlive, 0);
char cache_filename[MAX_PATH];
sprintf(cache_filename, "%sdx9-%s-vs.cache", File::GetUserPath(D_SHADERCACHE_IDX).c_str(),
SConfig::GetInstance().m_LocalCoreStartupParameter.m_strUniqueID.c_str());
VertexShaderCacheInserter inserter;
g_vs_disk_cache.OpenAndRead(cache_filename, inserter);
if (g_Config.bEnableShaderDebugging)
Clear();
last_entry = NULL;
}
void VertexShaderCache::Init()
{
const D3D11_INPUT_ELEMENT_DESC simpleelems[2] =
{
{ "POSITION", 0, DXGI_FORMAT_R32G32B32_FLOAT, 0, 0, D3D11_INPUT_PER_VERTEX_DATA, 0 },
{ "TEXCOORD", 0, DXGI_FORMAT_R32G32B32_FLOAT, 0, 12, D3D11_INPUT_PER_VERTEX_DATA, 0 },
};
const D3D11_INPUT_ELEMENT_DESC clearelems[2] =
{
{ "POSITION", 0, DXGI_FORMAT_R32G32B32_FLOAT, 0, 0, D3D11_INPUT_PER_VERTEX_DATA, 0 },
{ "COLOR", 0, DXGI_FORMAT_R8G8B8A8_UNORM, 0, 12, D3D11_INPUT_PER_VERTEX_DATA, 0 },
};
unsigned int cbsize = ((sizeof(vsconstants))&(~0xf))+0x10; // must be a multiple of 16
D3D11_BUFFER_DESC cbdesc = CD3D11_BUFFER_DESC(cbsize, D3D11_BIND_CONSTANT_BUFFER, D3D11_USAGE_DYNAMIC, D3D11_CPU_ACCESS_WRITE);
HRESULT hr = D3D::device->CreateBuffer(&cbdesc, NULL, &vscbuf);
CHECK(hr==S_OK, "Create vertex shader constant buffer (size=%u)", cbsize);
D3D::SetDebugObjectName((ID3D11DeviceChild*)vscbuf, "vertex shader constant buffer used to emulate the GX pipeline");
D3DBlob* blob;
D3D::CompileVertexShader(simple_shader_code, sizeof(simple_shader_code), &blob);
D3D::device->CreateInputLayout(simpleelems, 2, blob->Data(), blob->Size(), &SimpleLayout);
SimpleVertexShader = D3D::CreateVertexShaderFromByteCode(blob);
if (SimpleLayout == NULL || SimpleVertexShader == NULL) PanicAlert("Failed to create simple vertex shader or input layout at %s %d\n", __FILE__, __LINE__);
blob->Release();
D3D::SetDebugObjectName((ID3D11DeviceChild*)SimpleVertexShader, "simple vertex shader");
D3D::SetDebugObjectName((ID3D11DeviceChild*)SimpleLayout, "simple input layout");
D3D::CompileVertexShader(clear_shader_code, sizeof(clear_shader_code), &blob);
D3D::device->CreateInputLayout(clearelems, 2, blob->Data(), blob->Size(), &ClearLayout);
ClearVertexShader = D3D::CreateVertexShaderFromByteCode(blob);
if (ClearLayout == NULL || ClearVertexShader == NULL) PanicAlert("Failed to create clear vertex shader or input layout at %s %d\n", __FILE__, __LINE__);
blob->Release();
D3D::SetDebugObjectName((ID3D11DeviceChild*)ClearVertexShader, "clear vertex shader");
D3D::SetDebugObjectName((ID3D11DeviceChild*)ClearLayout, "clear input layout");
Clear();
// these values are hardcoded, they depend on internal D3DCompile behavior
// TODO: Do this with D3DReflect or something instead
unsigned int k;
for (k = 0;k < 6;k++) vs_constant_offset_table[C_POSNORMALMATRIX+k] = 0+4*k;
for (k = 0;k < 4;k++) vs_constant_offset_table[C_PROJECTION+k] = 24+4*k;
for (k = 0;k < 4;k++) vs_constant_offset_table[C_MATERIALS+k] = 40+4*k;
for (k = 0;k < 40;k++) vs_constant_offset_table[C_LIGHTS+k] = 56+4*k;
for (k = 0;k < 24;k++) vs_constant_offset_table[C_TEXMATRICES+k] = 216+4*k;
for (k = 0;k < 64;k++) vs_constant_offset_table[C_TRANSFORMMATRICES+k] = 312+4*k;
for (k = 0;k < 32;k++) vs_constant_offset_table[C_NORMALMATRICES+k] = 568+4*k;
for (k = 0;k < 64;k++) vs_constant_offset_table[C_POSTTRANSFORMMATRICES+k] = 696+4*k;
for (k = 0;k < 4;k++) vs_constant_offset_table[C_DEPTHPARAMS+k] = 952+4*k;
if (!File::Exists(File::GetUserPath(D_SHADERCACHE_IDX)))
File::CreateDir(File::GetUserPath(D_SHADERCACHE_IDX).c_str());
SETSTAT(stats.numVertexShadersCreated, 0);
SETSTAT(stats.numVertexShadersAlive, 0);
char cache_filename[MAX_PATH];
sprintf(cache_filename, "%sdx11-%s-vs.cache", File::GetUserPath(D_SHADERCACHE_IDX).c_str(),
SConfig::GetInstance().m_LocalCoreStartupParameter.m_strUniqueID.c_str());
VertexShaderCacheInserter inserter;
g_vs_disk_cache.OpenAndRead(cache_filename, inserter);
if (g_Config.bEnableShaderDebugging)
Clear();
last_entry = NULL;
}