ShuriZma
/
dolphin
mirror of https://github.com/dolphin-emu/dolphin.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

VideoBackends: Move SamplerState to common

This commit is contained in:
Stenzek 2017-09-09 18:30:15 +10:00
parent 340aabbb06
commit 24ddea04ce
21 changed files with 352 additions and 401 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

101
Source/Core/VideoBackends/D3D/D3DState.cpp
View File

@ -12,7 +12,7 @@
#include "VideoBackends/D3D/D3DBase.h" #include "VideoBackends/D3D/D3DBase.h"
#include "VideoBackends/D3D/D3DState.h" #include "VideoBackends/D3D/D3DState.h"
#include "VideoCommon/SamplerCommon.h" #include "VideoCommon/VideoConfig.h"
namespace DX11 namespace DX11
{ {
@ -265,90 +265,55 @@ void StateManager::SetTextureByMask(u32 textureSlotMask, ID3D11ShaderResourceVie
ID3D11SamplerState* StateCache::Get(SamplerState state) ID3D11SamplerState* StateCache::Get(SamplerState state)
{ {
auto it = m_sampler.find(state.packed); auto it = m_sampler.find(state.hex);
if (it != m_sampler.end()) if (it != m_sampler.end())
{
return it->second; return it->second;
}
const unsigned int d3dMipFilters[4] = {
TexMode0::TEXF_NONE, TexMode0::TEXF_POINT, TexMode0::TEXF_LINEAR,
TexMode0::TEXF_NONE, // reserved
};
const D3D11_TEXTURE_ADDRESS_MODE d3dClamps[4] = {
D3D11_TEXTURE_ADDRESS_CLAMP, D3D11_TEXTURE_ADDRESS_WRAP, D3D11_TEXTURE_ADDRESS_MIRROR,
D3D11_TEXTURE_ADDRESS_WRAP // reserved
};
D3D11_SAMPLER_DESC sampdc = CD3D11_SAMPLER_DESC(CD3D11_DEFAULT()); D3D11_SAMPLER_DESC sampdc = CD3D11_SAMPLER_DESC(CD3D11_DEFAULT());
if (state.mipmap_filter == SamplerState::Filter::Linear)
unsigned int mip = d3dMipFilters[state.min_filter & 3];
if (state.max_anisotropy > 1 && !SamplerCommon::IsBpTexMode0PointFiltering(state))
{ {
sampdc.Filter = D3D11_FILTER_ANISOTROPIC; if (state.min_filter == SamplerState::Filter::Linear)
sampdc.MaxAnisotropy = (u32)state.max_anisotropy; sampdc.Filter = (state.mag_filter == SamplerState::Filter::Linear) ?
D3D11_FILTER_MIN_MAG_MIP_LINEAR :
D3D11_FILTER_MIN_LINEAR_MAG_POINT_MIP_LINEAR;
else
sampdc.Filter = (state.mag_filter == SamplerState::Filter::Linear) ?
D3D11_FILTER_MIN_POINT_MAG_MIP_LINEAR :
D3D11_FILTER_MIN_MAG_POINT_MIP_LINEAR;
} }
else if (state.min_filter & 4) // linear min filter else
{ {
if (state.mag_filter) // linear mag filter if (state.min_filter == SamplerState::Filter::Linear)
{ sampdc.Filter = (state.mag_filter == SamplerState::Filter::Linear) ?
if (mip == TexMode0::TEXF_NONE) D3D11_FILTER_MIN_MAG_LINEAR_MIP_POINT :
sampdc.Filter = D3D11_FILTER_MIN_MAG_LINEAR_MIP_POINT; D3D11_FILTER_MIN_LINEAR_MAG_MIP_POINT;
else if (mip == TexMode0::TEXF_POINT) else
sampdc.Filter = D3D11_FILTER_MIN_MAG_LINEAR_MIP_POINT; sampdc.Filter = (state.mag_filter == SamplerState::Filter::Linear) ?
else if (mip == TexMode0::TEXF_LINEAR) D3D11_FILTER_MIN_POINT_MAG_LINEAR_MIP_POINT :
sampdc.Filter = D3D11_FILTER_MIN_MAG_MIP_LINEAR; D3D11_FILTER_MIN_MAG_MIP_POINT;
}
else // point mag filter
{
if (mip == TexMode0::TEXF_NONE)
sampdc.Filter = D3D11_FILTER_MIN_LINEAR_MAG_MIP_POINT;
else if (mip == TexMode0::TEXF_POINT)
sampdc.Filter = D3D11_FILTER_MIN_LINEAR_MAG_MIP_POINT;
else if (mip == TexMode0::TEXF_LINEAR)
sampdc.Filter = D3D11_FILTER_MIN_LINEAR_MAG_POINT_MIP_LINEAR;
}
}
else // point min filter
{
if (state.mag_filter) // linear mag filter
{
if (mip == TexMode0::TEXF_NONE)
sampdc.Filter = D3D11_FILTER_MIN_POINT_MAG_LINEAR_MIP_POINT;
else if (mip == TexMode0::TEXF_POINT)
sampdc.Filter = D3D11_FILTER_MIN_POINT_MAG_LINEAR_MIP_POINT;
else if (mip == TexMode0::TEXF_LINEAR)
sampdc.Filter = D3D11_FILTER_MIN_POINT_MAG_MIP_LINEAR;
}
else // point mag filter
{
if (mip == TexMode0::TEXF_NONE)
sampdc.Filter = D3D11_FILTER_MIN_MAG_MIP_POINT;
else if (mip == TexMode0::TEXF_POINT)
sampdc.Filter = D3D11_FILTER_MIN_MAG_MIP_POINT;
else if (mip == TexMode0::TEXF_LINEAR)
sampdc.Filter = D3D11_FILTER_MIN_MAG_POINT_MIP_LINEAR;
}
} }
sampdc.AddressU = d3dClamps[state.wrap_s]; static constexpr std::array<D3D11_TEXTURE_ADDRESS_MODE, 3> address_modes = {
sampdc.AddressV = d3dClamps[state.wrap_t]; {D3D11_TEXTURE_ADDRESS_CLAMP, D3D11_TEXTURE_ADDRESS_WRAP, D3D11_TEXTURE_ADDRESS_MIRROR}};
sampdc.AddressU = address_modes[static_cast<u32>(state.wrap_u.Value())];
sampdc.MaxLOD = SamplerCommon::AreBpTexMode0MipmapsEnabled(state) ? state.max_lod / 16.f : 0.f; sampdc.AddressV = address_modes[static_cast<u32>(state.wrap_v.Value())];
sampdc.MinLOD = std::min(state.min_lod / 16.f, sampdc.MaxLOD); sampdc.MaxLOD = state.max_lod / 16.f;
sampdc.MinLOD = state.min_lod / 16.f;
sampdc.MipLODBias = (s32)state.lod_bias / 32.0f; sampdc.MipLODBias = (s32)state.lod_bias / 32.0f;
ID3D11SamplerState* res = nullptr; if (state.anisotropic_filtering)
{
sampdc.Filter = D3D11_FILTER_ANISOTROPIC;
sampdc.MaxAnisotropy = 1u << g_ActiveConfig.iMaxAnisotropy;
}
ID3D11SamplerState* res = nullptr;
HRESULT hr = D3D::device->CreateSamplerState(&sampdc, &res); HRESULT hr = D3D::device->CreateSamplerState(&sampdc, &res);
if (FAILED(hr)) if (FAILED(hr))
PanicAlert("Fail %s %d\n", __FILE__, __LINE__); PanicAlert("Fail %s %d\n", __FILE__, __LINE__);
D3D::SetDebugObjectName(res, "sampler state used to emulate the GX pipeline"); D3D::SetDebugObjectName(res, "sampler state used to emulate the GX pipeline");
m_sampler.emplace(state.packed, res); m_sampler.emplace(state.hex, res);
return res; return res;
} }

16
Source/Core/VideoBackends/D3D/D3DState.h
View File

@ -20,20 +20,6 @@ struct ID3D11RasterizerState;
namespace DX11 namespace DX11
{ {
union SamplerState
{
BitField<0, 3, u64> min_filter;
BitField<3, 1, u64> mag_filter;
BitField<4, 8, u64> min_lod;
BitField<12, 8, u64> max_lod;
BitField<20, 8, s64> lod_bias;
BitField<28, 2, u64> wrap_s;
BitField<30, 2, u64> wrap_t;
BitField<32, 5, u64> max_anisotropy;
u64 packed;
};
class StateCache class StateCache
{ {
public: public:
@ -54,7 +40,7 @@ private:
std::unordered_map<u32, ID3D11DepthStencilState*> m_depth; std::unordered_map<u32, ID3D11DepthStencilState*> m_depth;
std::unordered_map<u32, ID3D11RasterizerState*> m_raster; std::unordered_map<u32, ID3D11RasterizerState*> m_raster;
std::unordered_map<u32, ID3D11BlendState*> m_blend; std::unordered_map<u32, ID3D11BlendState*> m_blend;
std::unordered_map<u64, ID3D11SamplerState*> m_sampler; std::unordered_map<u32, ID3D11SamplerState*> m_sampler;
}; };
namespace D3D namespace D3D

41
Source/Core/VideoBackends/D3D/Render.cpp
View File

@ -249,7 +249,7 @@ Renderer::Renderer() : ::Renderer(D3D::GetBackBufferWidth(), D3D::GetBackBufferH
// Setup GX pipeline state // Setup GX pipeline state
for (auto& sampler : s_gx_state.samplers) for (auto& sampler : s_gx_state.samplers)
sampler.packed = 0; sampler.hex = RenderState::GetPointSamplerState().hex;
s_gx_state.zmode.testenable = false; s_gx_state.zmode.testenable = false;
s_gx_state.zmode.updateenable = false; s_gx_state.zmode.updateenable = false;
@ -870,12 +870,8 @@ void Renderer::ApplyState()
StateCache::GetPrimitiveTopology(s_gx_state.raster.primitive)); StateCache::GetPrimitiveTopology(s_gx_state.raster.primitive));
FramebufferManager::SetIntegerEFBRenderTarget(s_gx_state.blend.logicopenable); FramebufferManager::SetIntegerEFBRenderTarget(s_gx_state.blend.logicopenable);
for (size_t stage = 0; stage < s_gx_state.samplers.size(); stage++) for (u32 stage = 0; stage < static_cast<u32>(s_gx_state.samplers.size()); stage++)
{
// TODO: cache SamplerState directly, not d3d object
s_gx_state.samplers[stage].max_anisotropy = UINT64_C(1) << g_ActiveConfig.iMaxAnisotropy;
D3D::stateman->SetSampler(stage, s_gx_state_cache.Get(s_gx_state.samplers[stage])); D3D::stateman->SetSampler(stage, s_gx_state_cache.Get(s_gx_state.samplers[stage]));
}
ID3D11Buffer* vertexConstants = VertexShaderCache::GetConstantBuffer(); ID3D11Buffer* vertexConstants = VertexShaderCache::GetConstantBuffer();
@ -902,38 +898,9 @@ void Renderer::SetDepthState(const DepthState& state)
s_gx_state.zmode.hex = state.hex; s_gx_state.zmode.hex = state.hex;
} }
void Renderer::SetSamplerState(int stage, int texindex, bool custom_tex) void Renderer::SetSamplerState(u32 index, const SamplerState& state)
{ {
const FourTexUnits& tex = bpmem.tex[texindex]; s_gx_state.samplers[index].hex = state.hex;
const TexMode0& tm0 = tex.texMode0[stage];
const TexMode1& tm1 = tex.texMode1[stage];
if (texindex)
stage += 4;
if (g_ActiveConfig.bForceFiltering)
{
// Only use mipmaps if the game says they are available.
s_gx_state.samplers[stage].min_filter = SamplerCommon::AreBpTexMode0MipmapsEnabled(tm0) ? 6 : 4;
s_gx_state.samplers[stage].mag_filter = 1; // linear mag
}
else
{
s_gx_state.samplers[stage].min_filter = (u32)tm0.min_filter;
s_gx_state.samplers[stage].mag_filter = (u32)tm0.mag_filter;
}
s_gx_state.samplers[stage].wrap_s = (u32)tm0.wrap_s;
s_gx_state.samplers[stage].wrap_t = (u32)tm0.wrap_t;
s_gx_state.samplers[stage].max_lod = (u32)tm1.max_lod;
s_gx_state.samplers[stage].min_lod = (u32)tm1.min_lod;
s_gx_state.samplers[stage].lod_bias = (s32)tm0.lod_bias;
// custom textures may have higher resolution, so disable the max_lod
if (custom_tex)
{
s_gx_state.samplers[stage].max_lod = 255;
}
} }
void Renderer::SetInterlacingMode() void Renderer::SetInterlacingMode()

2
Source/Core/VideoBackends/D3D/Render.h
View File

@ -23,7 +23,7 @@ public:
void SetScissorRect(const EFBRectangle& rc) override; void SetScissorRect(const EFBRectangle& rc) override;
void SetRasterizationState(const RasterizationState& state) override; void SetRasterizationState(const RasterizationState& state) override;
void SetDepthState(const DepthState& state) override; void SetDepthState(const DepthState& state) override;
void SetSamplerState(int stage, int texindex, bool custom_tex) override; void SetSamplerState(u32 index, const SamplerState& state) override;
void SetInterlacingMode() override; void SetInterlacingMode() override;
void SetViewport() override; void SetViewport() override;
void SetFullscreen(bool enable_fullscreen) override; void SetFullscreen(bool enable_fullscreen) override;

6
Source/Core/VideoBackends/OGL/ProgramShaderCache.cpp
View File

@ -1119,7 +1119,8 @@ bool ProgramShaderCache::ShaderCompileWorkItem::Compile()
gcode = GenerateGeometryShaderCode(APIType::OpenGL, host_config, m_uid.guid.GetUidData()); gcode = GenerateGeometryShaderCode(APIType::OpenGL, host_config, m_uid.guid.GetUidData());
CompileShader(m_program, vcode.GetBuffer(), pcode.GetBuffer(), gcode.GetBuffer()); CompileShader(m_program, vcode.GetBuffer(), pcode.GetBuffer(), gcode.GetBuffer());
DrawPrerenderArray(m_program, m_uid.guid.GetUidData()->primitive_type); DrawPrerenderArray(m_program,
static_cast<PrimitiveType>(m_uid.guid.GetUidData()->primitive_type));
return true; return true;
} }
@ -1157,7 +1158,8 @@ bool ProgramShaderCache::UberShaderCompileWorkItem::Compile()
gcode = GenerateGeometryShaderCode(APIType::OpenGL, host_config, m_uid.guid.GetUidData()); gcode = GenerateGeometryShaderCode(APIType::OpenGL, host_config, m_uid.guid.GetUidData());
CompileShader(m_program, vcode.GetBuffer(), pcode.GetBuffer(), gcode.GetBuffer()); CompileShader(m_program, vcode.GetBuffer(), pcode.GetBuffer(), gcode.GetBuffer());
DrawPrerenderArray(m_program, m_uid.guid.GetUidData()->primitive_type); DrawPrerenderArray(m_program,
static_cast<PrimitiveType>(m_uid.guid.GetUidData()->primitive_type));
return true; return true;
} }

12
Source/Core/VideoBackends/OGL/Render.cpp
View File

@ -1503,9 +1503,13 @@ void Renderer::SwapImpl(u32 xfbAddr, u32 fbWidth, u32 fbStride, u32 fbHeight,
g_Config.iSaveTargetId = 0; g_Config.iSaveTargetId = 0;
int old_anisotropy = g_ActiveConfig.iMaxAnisotropy;
UpdateActiveConfig(); UpdateActiveConfig();
g_texture_cache->OnConfigChanged(g_ActiveConfig); g_texture_cache->OnConfigChanged(g_ActiveConfig);
if (old_anisotropy != g_ActiveConfig.iMaxAnisotropy)
g_sampler_cache->Clear();
// Invalidate shader cache when the host config changes. // Invalidate shader cache when the host config changes.
if (CheckForHostConfigChanges()) if (CheckForHostConfigChanges())
ProgramShaderCache::Reload(); ProgramShaderCache::Reload();
@ -1834,13 +1838,9 @@ void Renderer::SetDepthState(const DepthState& state)
} }
} }
void Renderer::SetSamplerState(int stage, int texindex, bool custom_tex) void Renderer::SetSamplerState(u32 index, const SamplerState& state)
{ {
auto const& tex = bpmem.tex[texindex]; g_sampler_cache->SetSamplerState(index, state);
auto const& tm0 = tex.texMode0[stage];
auto const& tm1 = tex.texMode1[stage];
g_sampler_cache->SetSamplerState((texindex * 4) + stage, tm0, tm1, custom_tex);
} }
void Renderer::SetInterlacingMode() void Renderer::SetInterlacingMode()

2
Source/Core/VideoBackends/OGL/Render.h
View File

@ -81,7 +81,7 @@ public:
void SetScissorRect(const EFBRectangle& rc) override; void SetScissorRect(const EFBRectangle& rc) override;
void SetRasterizationState(const RasterizationState& state) override; void SetRasterizationState(const RasterizationState& state) override;
void SetDepthState(const DepthState& state) override; void SetDepthState(const DepthState& state) override;
void SetSamplerState(int stage, int texindex, bool custom_tex) override; void SetSamplerState(u32 index, const SamplerState& state) override;
void SetInterlacingMode() override; void SetInterlacingMode() override;
void SetViewport() override; void SetViewport() override;

163
Source/Core/VideoBackends/OGL/SamplerCache.cpp
View File

@ -15,141 +15,106 @@ namespace OGL
{ {
std::unique_ptr<SamplerCache> g_sampler_cache; std::unique_ptr<SamplerCache> g_sampler_cache;
SamplerCache::SamplerCache() : m_last_max_anisotropy() SamplerCache::SamplerCache()
{ {
glGenSamplers(2, m_sampler_id); glGenSamplers(1, &m_point_sampler);
glSamplerParameteri(m_sampler_id[0], GL_TEXTURE_MIN_FILTER, GL_NEAREST); glGenSamplers(1, &m_linear_sampler);
glSamplerParameteri(m_sampler_id[0], GL_TEXTURE_MAG_FILTER, GL_NEAREST); glSamplerParameteri(m_point_sampler, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
glSamplerParameteri(m_sampler_id[0], GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE); glSamplerParameteri(m_point_sampler, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
glSamplerParameteri(m_sampler_id[0], GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE); glSamplerParameteri(m_point_sampler, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
glSamplerParameteri(m_sampler_id[1], GL_TEXTURE_MIN_FILTER, GL_LINEAR); glSamplerParameteri(m_point_sampler, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
glSamplerParameteri(m_sampler_id[1], GL_TEXTURE_MAG_FILTER, GL_LINEAR); glSamplerParameteri(m_linear_sampler, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
glSamplerParameteri(m_sampler_id[1], GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE); glSamplerParameteri(m_linear_sampler, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
glSamplerParameteri(m_sampler_id[1], GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE); glSamplerParameteri(m_linear_sampler, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
glSamplerParameteri(m_linear_sampler, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
} }
SamplerCache::~SamplerCache() SamplerCache::~SamplerCache()
{ {
Clear(); Clear();
glDeleteSamplers(2, m_sampler_id); glDeleteSamplers(1, &m_point_sampler);
glDeleteSamplers(1, &m_linear_sampler);
} }
void SamplerCache::BindNearestSampler(int stage) void SamplerCache::BindNearestSampler(int stage)
{ {
glBindSampler(stage, m_sampler_id[0]); glBindSampler(stage, m_point_sampler);
} }
void SamplerCache::BindLinearSampler(int stage) void SamplerCache::BindLinearSampler(int stage)
{ {
glBindSampler(stage, m_sampler_id[1]); glBindSampler(stage, m_linear_sampler);
} }
void SamplerCache::SetSamplerState(int stage, const TexMode0& tm0, const TexMode1& tm1, void SamplerCache::SetSamplerState(u32 stage, const SamplerState& state)
bool custom_tex)
{ {
// TODO: can this go somewhere else? if (m_active_samplers[stage].first == state && m_active_samplers[stage].second != 0)
if (m_last_max_anisotropy != g_ActiveConfig.iMaxAnisotropy) return;
auto it = m_cache.find(state);
if (it == m_cache.end())
{ {
m_last_max_anisotropy = g_ActiveConfig.iMaxAnisotropy; GLuint sampler;
Clear(); glGenSamplers(1, &sampler);
SetParameters(sampler, state);
it = m_cache.emplace(state, sampler).first;
} }
Params params(tm0, tm1); m_active_samplers[stage].first = state;
m_active_samplers[stage].second = it->second;
// take equivalent forced linear when bForceFiltering glBindSampler(stage, it->second);
if (g_ActiveConfig.bForceFiltering)
{
params.tm0.min_filter = SamplerCommon::AreBpTexMode0MipmapsEnabled(tm0) ? 6 : 4;
params.tm0.mag_filter = 1;
}
// custom textures may have higher resolution, so disable the max_lod
if (custom_tex)
{
params.tm1.max_lod = 255;
}
// TODO: Should keep a circular buffer for each stage of recently used samplers.
auto& active_sampler = m_active_samplers[stage];
if (active_sampler.first != params || !active_sampler.second.sampler_id)
{
// Active sampler does not match parameters (or is invalid), bind the proper one.
active_sampler.first = params;
active_sampler.second = GetEntry(params);
glBindSampler(stage, active_sampler.second.sampler_id);
}
} }
SamplerCache::Value& SamplerCache::GetEntry(const Params& params) void SamplerCache::InvalidateBinding(u32 stage)
{ {
auto& val = m_cache[params]; m_active_samplers[stage].second = 0;
if (!val.sampler_id)
{
// Sampler not found in cache, create it.
glGenSamplers(1, &val.sampler_id);
SetParameters(val.sampler_id, params);
// TODO: Maybe kill old samplers if the cache gets huge. It doesn't seem to get huge though.
// ERROR_LOG(VIDEO, "Sampler cache size is now %ld.", m_cache.size());
}
return val;
} }
void SamplerCache::SetParameters(GLuint sampler_id, const Params& params) void SamplerCache::SetParameters(GLuint sampler_id, const SamplerState& params)
{ {
static const GLint min_filters[8] = { GLenum min_filter;
GL_NEAREST, GL_NEAREST_MIPMAP_NEAREST, GL_NEAREST_MIPMAP_LINEAR, GL_NEAREST, GLenum mag_filter = (params.mag_filter == SamplerState::Filter::Point) ? GL_NEAREST : GL_LINEAR;
GL_LINEAR, GL_LINEAR_MIPMAP_NEAREST, GL_LINEAR_MIPMAP_LINEAR, GL_LINEAR, if (params.mipmap_filter == SamplerState::Filter::Linear)
};
static const GLint wrap_settings[4] = {
GL_CLAMP_TO_EDGE, GL_REPEAT, GL_MIRRORED_REPEAT, GL_REPEAT,
};
auto& tm0 = params.tm0;
auto& tm1 = params.tm1;
glSamplerParameteri(sampler_id, GL_TEXTURE_WRAP_S, wrap_settings[tm0.wrap_s]);
glSamplerParameteri(sampler_id, GL_TEXTURE_WRAP_T, wrap_settings[tm0.wrap_t]);
glSamplerParameterf(sampler_id, GL_TEXTURE_MIN_LOD, tm1.min_lod / 16.f);
glSamplerParameterf(sampler_id, GL_TEXTURE_MAX_LOD, tm1.max_lod / 16.f);
if (GLInterface->GetMode() == GLInterfaceMode::MODE_OPENGL)
glSamplerParameterf(sampler_id, GL_TEXTURE_LOD_BIAS, (s32)tm0.lod_bias / 32.f);
GLint min_filter = min_filters[tm0.min_filter];
GLint mag_filter = tm0.mag_filter ? GL_LINEAR : GL_NEAREST;
if (g_ActiveConfig.iMaxAnisotropy > 0 && g_ogl_config.bSupportsAniso &&
!SamplerCommon::IsBpTexMode0PointFiltering(tm0))
{ {
// https://www.opengl.org/registry/specs/EXT/texture_filter_anisotropic.txt min_filter = (params.min_filter == SamplerState::Filter::Point) ? GL_NEAREST_MIPMAP_LINEAR :
// For predictable results on all hardware/drivers, only use one of: GL_LINEAR_MIPMAP_LINEAR;
// GL_LINEAR + GL_LINEAR (No Mipmaps [Bilinear]) }
// GL_LINEAR + GL_LINEAR_MIPMAP_LINEAR (w/ Mipmaps [Trilinear]) else
// Letting the game set other combinations will have varying arbitrary results; {
// possibly being interpreted as equal to bilinear/trilinear, implicitly min_filter = (params.min_filter == SamplerState::Filter::Point) ? GL_NEAREST_MIPMAP_NEAREST :
// disabling anisotropy, or changing the anisotropic algorithm employed. GL_LINEAR_MIPMAP_NEAREST;
min_filter =
SamplerCommon::AreBpTexMode0MipmapsEnabled(tm0) ? GL_LINEAR_MIPMAP_LINEAR : GL_LINEAR;
mag_filter = GL_LINEAR;
glSamplerParameterf(sampler_id, GL_TEXTURE_MAX_ANISOTROPY_EXT,
(float)(1 << g_ActiveConfig.iMaxAnisotropy));
} }
glSamplerParameteri(sampler_id, GL_TEXTURE_MIN_FILTER, min_filter); glSamplerParameteri(sampler_id, GL_TEXTURE_MIN_FILTER, min_filter);
glSamplerParameteri(sampler_id, GL_TEXTURE_MAG_FILTER, mag_filter); glSamplerParameteri(sampler_id, GL_TEXTURE_MAG_FILTER, mag_filter);
static constexpr std::array<GLenum, 3> address_modes = {
{GL_CLAMP_TO_EDGE, GL_REPEAT, GL_MIRRORED_REPEAT}};
glSamplerParameteri(sampler_id, GL_TEXTURE_WRAP_S,
address_modes[static_cast<u32>(params.wrap_u.Value())]);
glSamplerParameteri(sampler_id, GL_TEXTURE_WRAP_T,
address_modes[static_cast<u32>(params.wrap_v.Value())]);
glSamplerParameterf(sampler_id, GL_TEXTURE_MIN_LOD, params.min_lod / 16.f);
glSamplerParameterf(sampler_id, GL_TEXTURE_MAX_LOD, params.max_lod / 16.f);
if (GLInterface->GetMode() == GLInterfaceMode::MODE_OPENGL)
glSamplerParameterf(sampler_id, GL_TEXTURE_LOD_BIAS, params.lod_bias / 32.f);
if (params.anisotropic_filtering && g_ogl_config.bSupportsAniso)
{
glSamplerParameterf(sampler_id, GL_TEXTURE_MAX_ANISOTROPY_EXT,
static_cast<float>(1 << g_ActiveConfig.iMaxAnisotropy));
}
} }
void SamplerCache::Clear() void SamplerCache::Clear()
{ {
for (auto& p : m_cache) for (auto& p : m_cache)
{ glDeleteSamplers(1, &p.second);
glDeleteSamplers(1, &p.second.sampler_id); for (auto& p : m_active_samplers)
} p.second = 0;
m_cache.clear(); m_cache.clear();
} }
} }

45
Source/Core/VideoBackends/OGL/SamplerCache.h
View File

@ -4,6 +4,7 @@
#pragma once #pragma once
#include <array>
#include <map> #include <map>
#include <memory> #include <memory>
@ -24,49 +25,21 @@ public:
SamplerCache(SamplerCache&&) = delete; SamplerCache(SamplerCache&&) = delete;
SamplerCache& operator=(SamplerCache&&) = delete; SamplerCache& operator=(SamplerCache&&) = delete;
void SetSamplerState(int stage, const TexMode0& tm0, const TexMode1& tm1, bool custom_tex); void SetSamplerState(u32 stage, const SamplerState& state);
void InvalidateBinding(u32 stage);
void Clear(); void Clear();
void BindNearestSampler(int stage); void BindNearestSampler(int stage);
void BindLinearSampler(int stage); void BindLinearSampler(int stage);
private: private:
struct Params static void SetParameters(GLuint sampler_id, const SamplerState& params);
{
union
{
struct
{
TexMode0 tm0;
TexMode1 tm1;
};
u64 hex; std::map<SamplerState, GLuint> m_cache;
}; std::array<std::pair<SamplerState, GLuint>, 8> m_active_samplers{};
Params() : hex() {} GLuint m_point_sampler;
Params(const TexMode0& _tm0, const TexMode1& _tm1) : tm0(_tm0), tm1(_tm1) GLuint m_linear_sampler;
{
static_assert(sizeof(Params) == 8, "Assuming I can treat this as a 64bit int.");
}
bool operator<(const Params& other) const { return hex < other.hex; }
bool operator!=(const Params& other) const { return hex != other.hex; }
};
struct Value
{
Value() : sampler_id() {}
GLuint sampler_id;
};
void SetParameters(GLuint sampler_id, const Params& params);
Value& GetEntry(const Params& params);
std::map<Params, Value> m_cache;
std::pair<Params, Value> m_active_samplers[8];
int m_last_max_anisotropy;
u32 m_sampler_id[2];
}; };
extern std::unique_ptr<SamplerCache> g_sampler_cache; extern std::unique_ptr<SamplerCache> g_sampler_cache;

16
Source/Core/VideoBackends/Vulkan/Constants.h
View File

@ -132,20 +132,4 @@ union MultisamplingState
u32 hex; u32 hex;
}; };
// Sampler info
union SamplerState
{
BitField<0, 1, VkFilter> min_filter;
BitField<1, 1, VkFilter> mag_filter;
BitField<2, 1, VkSamplerMipmapMode> mipmap_mode;
BitField<3, 2, VkSamplerAddressMode> wrap_u;
BitField<5, 2, VkSamplerAddressMode> wrap_v;
BitField<7, 8, u32> min_lod;
BitField<15, 8, u32> max_lod;
BitField<23, 8, s32> lod_bias;
BitField<31, 1, u32> enable_anisotropic_filtering;
u32 bits;
};
} // namespace Vulkan } // namespace Vulkan

25
Source/Core/VideoBackends/Vulkan/ObjectCache.cpp
View File

@ -312,29 +312,36 @@ VkSampler ObjectCache::GetSampler(const SamplerState& info)
if (iter != m_sampler_cache.end()) if (iter != m_sampler_cache.end())
return iter->second; return iter->second;
static constexpr std::array<VkFilter, 4> filters = {{VK_FILTER_NEAREST, VK_FILTER_LINEAR}};
static constexpr std::array<VkSamplerMipmapMode, 2> mipmap_modes = {
{VK_SAMPLER_MIPMAP_MODE_NEAREST, VK_SAMPLER_MIPMAP_MODE_LINEAR}};
static constexpr std::array<VkSamplerAddressMode, 4> address_modes = {
{VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE, VK_SAMPLER_ADDRESS_MODE_REPEAT,
VK_SAMPLER_ADDRESS_MODE_MIRRORED_REPEAT}};
VkSamplerCreateInfo create_info = { VkSamplerCreateInfo create_info = {
VK_STRUCTURE_TYPE_SAMPLER_CREATE_INFO, // VkStructureType sType VK_STRUCTURE_TYPE_SAMPLER_CREATE_INFO, // VkStructureType sType
nullptr, // const void* pNext nullptr, // const void* pNext
0, // VkSamplerCreateFlags flags 0, // VkSamplerCreateFlags flags
info.mag_filter, // VkFilter magFilter filters[static_cast<u32>(info.mag_filter.Value())], // VkFilter magFilter
info.min_filter, // VkFilter minFilter filters[static_cast<u32>(info.min_filter.Value())], // VkFilter minFilter
info.mipmap_mode, // VkSamplerMipmapMode mipmapMode mipmap_modes[static_cast<u32>(info.mipmap_filter.Value())], // VkSamplerMipmapMode mipmapMode
info.wrap_u, // VkSamplerAddressMode addressModeU address_modes[static_cast<u32>(info.wrap_u.Value())], // VkSamplerAddressMode addressModeU
info.wrap_v, // VkSamplerAddressMode addressModeV address_modes[static_cast<u32>(info.wrap_v.Value())], // VkSamplerAddressMode addressModeV
VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE, // VkSamplerAddressMode addressModeW VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE, // VkSamplerAddressMode addressModeW
static_cast<float>(info.lod_bias / 32.0f), // float mipLodBias info.lod_bias / 32.0f, // float mipLodBias
VK_FALSE, // VkBool32 anisotropyEnable VK_FALSE, // VkBool32 anisotropyEnable
0.0f, // float maxAnisotropy 0.0f, // float maxAnisotropy
VK_FALSE, // VkBool32 compareEnable VK_FALSE, // VkBool32 compareEnable
VK_COMPARE_OP_ALWAYS, // VkCompareOp compareOp VK_COMPARE_OP_ALWAYS, // VkCompareOp compareOp
static_cast<float>(info.min_lod / 16.0f), // float minLod info.min_lod / 16.0f, // float minLod
static_cast<float>(info.max_lod / 16.0f), // float maxLod info.max_lod / 16.0f, // float maxLod
VK_BORDER_COLOR_FLOAT_TRANSPARENT_BLACK, // VkBorderColor borderColor VK_BORDER_COLOR_FLOAT_TRANSPARENT_BLACK, // VkBorderColor borderColor
VK_FALSE // VkBool32 unnormalizedCoordinates VK_FALSE // VkBool32 unnormalizedCoordinates
}; };
// Can we use anisotropic filtering with this sampler? // Can we use anisotropic filtering with this sampler?
if (info.enable_anisotropic_filtering && g_vulkan_context->SupportsAnisotropicFiltering()) if (info.anisotropic_filtering && g_vulkan_context->SupportsAnisotropicFiltering())
{ {
// Cap anisotropy to device limits. // Cap anisotropy to device limits.
create_info.anisotropyEnable = VK_TRUE; create_info.anisotropyEnable = VK_TRUE;

59
Source/Core/VideoBackends/Vulkan/Renderer.cpp
View File

@ -52,10 +52,8 @@ Renderer::Renderer(std::unique_ptr<SwapChain> swap_chain)
m_swap_chain(std::move(swap_chain)) m_swap_chain(std::move(swap_chain))
{ {
UpdateActiveConfig(); UpdateActiveConfig();
// Set to something invalid, forcing all states to be re-initialized.
for (size_t i = 0; i < m_sampler_states.size(); i++) for (size_t i = 0; i < m_sampler_states.size(); i++)
m_sampler_states[i].bits = std::numeric_limits<decltype(m_sampler_states[i].bits)>::max(); m_sampler_states[i].hex = RenderState::GetPointSamplerState().hex;
} }
Renderer::~Renderer() Renderer::~Renderer()
@ -1282,65 +1280,22 @@ void Renderer::SetBlendingState(const BlendingState& state)
StateTracker::GetInstance()->SetBlendState(state); StateTracker::GetInstance()->SetBlendState(state);
} }
void Renderer::SetSamplerState(int stage, int texindex, bool custom_tex) void Renderer::SetSamplerState(u32 index, const SamplerState& state)
{ {
const FourTexUnits& tex = bpmem.tex[texindex];
const TexMode0& tm0 = tex.texMode0[stage];
const TexMode1& tm1 = tex.texMode1[stage];
SamplerState new_state = {};
if (g_ActiveConfig.bForceFiltering)
{
new_state.min_filter = VK_FILTER_LINEAR;
new_state.mag_filter = VK_FILTER_LINEAR;
new_state.mipmap_mode = SamplerCommon::AreBpTexMode0MipmapsEnabled(tm0) ?
VK_SAMPLER_MIPMAP_MODE_LINEAR :
VK_SAMPLER_MIPMAP_MODE_NEAREST;
}
else
{
// Constants for these?
new_state.min_filter = (tm0.min_filter & 4) != 0 ? VK_FILTER_LINEAR : VK_FILTER_NEAREST;
new_state.mipmap_mode = SamplerCommon::AreBpTexMode0MipmapsEnabled(tm0) ?
VK_SAMPLER_MIPMAP_MODE_LINEAR :
VK_SAMPLER_MIPMAP_MODE_NEAREST;
new_state.mag_filter = tm0.mag_filter != 0 ? VK_FILTER_LINEAR : VK_FILTER_NEAREST;
}
// If mipmaps are disabled, clamp min/max lod
new_state.max_lod = SamplerCommon::AreBpTexMode0MipmapsEnabled(tm0) ? tm1.max_lod : 0;
new_state.min_lod = std::min(new_state.max_lod.Value(), tm1.min_lod);
new_state.lod_bias = SamplerCommon::AreBpTexMode0MipmapsEnabled(tm0) ? tm0.lod_bias : 0;
// Custom textures may have a greater number of mips
if (custom_tex)
new_state.max_lod = 255;
// Address modes
static const VkSamplerAddressMode address_modes[] = {
VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE, VK_SAMPLER_ADDRESS_MODE_REPEAT,
VK_SAMPLER_ADDRESS_MODE_MIRRORED_REPEAT, VK_SAMPLER_ADDRESS_MODE_REPEAT};
new_state.wrap_u = address_modes[tm0.wrap_s];
new_state.wrap_v = address_modes[tm0.wrap_t];
// Only use anisotropic filtering for textures that would be linearly filtered.
new_state.enable_anisotropic_filtering = SamplerCommon::IsBpTexMode0PointFiltering(tm0) ? 0 : 1;
// Skip lookup if the state hasn't changed. // Skip lookup if the state hasn't changed.
size_t bind_index = (texindex * 4) + stage; if (m_sampler_states[index].hex == state.hex)
if (m_sampler_states[bind_index].bits == new_state.bits)
return; return;
// Look up new state and replace in state tracker. // Look up new state and replace in state tracker.
VkSampler sampler = g_object_cache->GetSampler(new_state); VkSampler sampler = g_object_cache->GetSampler(state);
if (sampler == VK_NULL_HANDLE) if (sampler == VK_NULL_HANDLE)
{ {
ERROR_LOG(VIDEO, "Failed to create sampler"); ERROR_LOG(VIDEO, "Failed to create sampler");
sampler = g_object_cache->GetPointSampler(); sampler = g_object_cache->GetPointSampler();
} }
StateTracker::GetInstance()->SetSampler(bind_index, sampler); StateTracker::GetInstance()->SetSampler(index, sampler);
m_sampler_states[bind_index].bits = new_state.bits; m_sampler_states[index].hex = state.hex;
} }
void Renderer::ResetSamplerStates() void Renderer::ResetSamplerStates()
@ -1352,7 +1307,7 @@ void Renderer::ResetSamplerStates()
// Invalidate all sampler states, next draw will re-initialize them. // Invalidate all sampler states, next draw will re-initialize them.
for (size_t i = 0; i < m_sampler_states.size(); i++) for (size_t i = 0; i < m_sampler_states.size(); i++)
{ {
m_sampler_states[i].bits = std::numeric_limits<decltype(m_sampler_states[i].bits)>::max(); m_sampler_states[i].hex = RenderState::GetPointSamplerState().hex;
StateTracker::GetInstance()->SetSampler(i, g_object_cache->GetPointSampler()); StateTracker::GetInstance()->SetSampler(i, g_object_cache->GetPointSampler());
} }

2
Source/Core/VideoBackends/Vulkan/Renderer.h
View File

@ -60,7 +60,7 @@ public:
void SetScissorRect(const EFBRectangle& rc) override; void SetScissorRect(const EFBRectangle& rc) override;
void SetRasterizationState(const RasterizationState& state) override; void SetRasterizationState(const RasterizationState& state) override;
void SetDepthState(const DepthState& state) override; void SetDepthState(const DepthState& state) override;
void SetSamplerState(int stage, int texindex, bool custom_tex) override; void SetSamplerState(u32 index, const SamplerState& state) override;
void SetInterlacingMode() override; void SetInterlacingMode() override;
void SetViewport() override; void SetViewport() override;

35
Source/Core/VideoBackends/Vulkan/ShaderCache.cpp
View File

@ -977,18 +977,6 @@ std::string ShaderCache::GetUtilityShaderHeader() const
return ss.str(); return ss.str();
} }
// Comparison operators for PipelineInfos
// Since these all boil down to POD types, we can just memcmp the entire thing for speed
// The is_trivially_copyable check fails on MSVC due to BitField.
// TODO: Can we work around this any way?
#if defined(__GNUC__) && !defined(__clang__) && __GNUC__ < 5 && !defined(_MSC_VER)
static_assert(std::has_trivial_copy_constructor<PipelineInfo>::value,
"PipelineInfo is trivially copyable");
#elif !defined(_MSC_VER)
static_assert(std::is_trivially_copyable<PipelineInfo>::value,
"PipelineInfo is trivially copyable");
#endif
std::size_t PipelineInfoHash::operator()(const PipelineInfo& key) const std::size_t PipelineInfoHash::operator()(const PipelineInfo& key) const
{ {
return static_cast<std::size_t>(XXH64(&key, sizeof(key), 0)); return static_cast<std::size_t>(XXH64(&key, sizeof(key), 0));
@ -1014,26 +1002,6 @@ bool operator>(const PipelineInfo& lhs, const PipelineInfo& rhs)
return std::memcmp(&lhs, &rhs, sizeof(lhs)) > 0; return std::memcmp(&lhs, &rhs, sizeof(lhs)) > 0;
} }
bool operator==(const SamplerState& lhs, const SamplerState& rhs)
{
return lhs.bits == rhs.bits;
}
bool operator!=(const SamplerState& lhs, const SamplerState& rhs)
{
return !operator==(lhs, rhs);
}
bool operator>(const SamplerState& lhs, const SamplerState& rhs)
{
return lhs.bits > rhs.bits;
}
bool operator<(const SamplerState& lhs, const SamplerState& rhs)
{
return lhs.bits < rhs.bits;
}
std::size_t ComputePipelineInfoHash::operator()(const ComputePipelineInfo& key) const std::size_t ComputePipelineInfoHash::operator()(const ComputePipelineInfo& key) const
{ {
return static_cast<std::size_t>(XXH64(&key, sizeof(key), 0)); return static_cast<std::size_t>(XXH64(&key, sizeof(key), 0));
@ -1214,7 +1182,8 @@ void ShaderCache::CreateDummyPipeline(const UberShader::VertexShaderUid& vuid,
pinfo.depth_state.hex = RenderState::GetNoDepthTestingDepthStencilState().hex; pinfo.depth_state.hex = RenderState::GetNoDepthTestingDepthStencilState().hex;
pinfo.blend_state.hex = RenderState::GetNoBlendingBlendState().hex; pinfo.blend_state.hex = RenderState::GetNoBlendingBlendState().hex;
pinfo.multisampling_state.hex = FramebufferManager::GetInstance()->GetEFBMultisamplingState().hex; pinfo.multisampling_state.hex = FramebufferManager::GetInstance()->GetEFBMultisamplingState().hex;
pinfo.rasterization_state.primitive = guid.GetUidData()->primitive_type; pinfo.rasterization_state.primitive =
static_cast<PrimitiveType>(guid.GetUidData()->primitive_type);
GetPipelineWithCacheResultAsync(pinfo); GetPipelineWithCacheResultAsync(pinfo);
} }

4
Source/Core/VideoBackends/Vulkan/ShaderCache.h
View File

@ -62,10 +62,6 @@ bool operator==(const PipelineInfo& lhs, const PipelineInfo& rhs);
bool operator!=(const PipelineInfo& lhs, const PipelineInfo& rhs); bool operator!=(const PipelineInfo& lhs, const PipelineInfo& rhs);
bool operator<(const PipelineInfo& lhs, const PipelineInfo& rhs); bool operator<(const PipelineInfo& lhs, const PipelineInfo& rhs);
bool operator>(const PipelineInfo& lhs, const PipelineInfo& rhs); bool operator>(const PipelineInfo& lhs, const PipelineInfo& rhs);
bool operator==(const SamplerState& lhs, const SamplerState& rhs);
bool operator!=(const SamplerState& lhs, const SamplerState& rhs);
bool operator>(const SamplerState& lhs, const SamplerState& rhs);
bool operator<(const SamplerState& lhs, const SamplerState& rhs);
struct ComputePipelineInfo struct ComputePipelineInfo
{ {

17
Source/Core/VideoCommon/GeometryShaderGen.cpp
View File

@ -22,7 +22,7 @@ bool geometry_shader_uid_data::IsPassthrough() const
{ {
const bool stereo = g_ActiveConfig.iStereoMode > 0; const bool stereo = g_ActiveConfig.iStereoMode > 0;
const bool wireframe = g_ActiveConfig.bWireFrame; const bool wireframe = g_ActiveConfig.bWireFrame;
return primitive_type >= PrimitiveType::Triangles && !stereo && !wireframe; return primitive_type >= static_cast<u32>(PrimitiveType::Triangles) && !stereo && !wireframe;
} }
GeometryShaderUid GetGeometryShaderUid(PrimitiveType primitive_type) GeometryShaderUid GetGeometryShaderUid(PrimitiveType primitive_type)
@ -31,7 +31,7 @@ GeometryShaderUid GetGeometryShaderUid(PrimitiveType primitive_type)
geometry_shader_uid_data* uid_data = out.GetUidData<geometry_shader_uid_data>(); geometry_shader_uid_data* uid_data = out.GetUidData<geometry_shader_uid_data>();
memset(uid_data, 0, sizeof(geometry_shader_uid_data)); memset(uid_data, 0, sizeof(geometry_shader_uid_data));
uid_data->primitive_type = primitive_type; uid_data->primitive_type = static_cast<u32>(primitive_type);
uid_data->numTexGens = xfmem.numTexGen.numTexGens; uid_data->numTexGens = xfmem.numTexGen.numTexGens;
return out; return out;
@ -56,9 +56,10 @@ ShaderCode GenerateGeometryShaderCode(APIType ApiType, const ShaderHostConfig& h
const bool msaa = host_config.msaa; const bool msaa = host_config.msaa;
const bool ssaa = host_config.ssaa; const bool ssaa = host_config.ssaa;
const bool stereo = host_config.stereo; const bool stereo = host_config.stereo;
const PrimitiveType primitive_type = static_cast<PrimitiveType>(uid_data->primitive_type);
const unsigned primitive_type_index = static_cast<unsigned>(uid_data->primitive_type); const unsigned primitive_type_index = static_cast<unsigned>(uid_data->primitive_type);
const unsigned vertex_in = std::min(static_cast<unsigned>(primitive_type_index) + 1, 3u); const unsigned vertex_in = std::min(static_cast<unsigned>(primitive_type_index) + 1, 3u);
unsigned vertex_out = uid_data->primitive_type == PrimitiveType::TriangleStrip ? 3 : 4; unsigned vertex_out = primitive_type == PrimitiveType::TriangleStrip ? 3 : 4;
if (wireframe) if (wireframe)
vertex_out++; vertex_out++;
@ -146,7 +147,7 @@ ShaderCode GenerateGeometryShaderCode(APIType ApiType, const ShaderHostConfig& h
out.Write("\tVertexData ps;\n"); out.Write("\tVertexData ps;\n");
} }
if (uid_data->primitive_type == PrimitiveType::Lines) if (primitive_type == PrimitiveType::Lines)
{ {
if (ApiType == APIType::OpenGL || ApiType == APIType::Vulkan) if (ApiType == APIType::OpenGL || ApiType == APIType::Vulkan)
{ {
@ -177,7 +178,7 @@ ShaderCode GenerateGeometryShaderCode(APIType ApiType, const ShaderHostConfig& h
"\t\toffset = float2(0, -" I_LINEPTPARAMS ".z / " I_LINEPTPARAMS ".y);\n" "\t\toffset = float2(0, -" I_LINEPTPARAMS ".z / " I_LINEPTPARAMS ".y);\n"
"\t}\n"); "\t}\n");
} }
else if (uid_data->primitive_type == PrimitiveType::Points) else if (primitive_type == PrimitiveType::Points)
{ {
if (ApiType == APIType::OpenGL || ApiType == APIType::Vulkan) if (ApiType == APIType::OpenGL || ApiType == APIType::Vulkan)
{ {
@ -247,7 +248,7 @@ ShaderCode GenerateGeometryShaderCode(APIType ApiType, const ShaderHostConfig& h
out.Write("\tf.pos.x += hoffset * (f.pos.w - " I_STEREOPARAMS ".z);\n"); out.Write("\tf.pos.x += hoffset * (f.pos.w - " I_STEREOPARAMS ".z);\n");
} }
if (uid_data->primitive_type == PrimitiveType::Lines) if (primitive_type == PrimitiveType::Lines)
{ {
out.Write("\tVS_OUTPUT l = f;\n" out.Write("\tVS_OUTPUT l = f;\n"
"\tVS_OUTPUT r = f;\n"); "\tVS_OUTPUT r = f;\n");
@ -268,7 +269,7 @@ ShaderCode GenerateGeometryShaderCode(APIType ApiType, const ShaderHostConfig& h
EmitVertex(out, host_config, uid_data, "l", ApiType, wireframe, pixel_lighting, true); EmitVertex(out, host_config, uid_data, "l", ApiType, wireframe, pixel_lighting, true);
EmitVertex(out, host_config, uid_data, "r", ApiType, wireframe, pixel_lighting); EmitVertex(out, host_config, uid_data, "r", ApiType, wireframe, pixel_lighting);
} }
else if (uid_data->primitive_type == PrimitiveType::Points) else if (primitive_type == PrimitiveType::Points)
{ {
out.Write("\tVS_OUTPUT ll = f;\n" out.Write("\tVS_OUTPUT ll = f;\n"
"\tVS_OUTPUT lr = f;\n" "\tVS_OUTPUT lr = f;\n"
@ -376,7 +377,7 @@ void EnumerateGeometryShaderUids(const std::function<void(const GeometryShaderUi
for (PrimitiveType primitive : primitive_lut) for (PrimitiveType primitive : primitive_lut)
{ {
auto* guid = uid.GetUidData<geometry_shader_uid_data>(); auto* guid = uid.GetUidData<geometry_shader_uid_data>();
guid->primitive_type = primitive; guid->primitive_type = static_cast<u32>(primitive);
for (u32 texgens = 0; texgens <= 8; texgens++) for (u32 texgens = 0; texgens <= 8; texgens++)
{ {

2
Source/Core/VideoCommon/GeometryShaderGen.h
View File

@ -20,7 +20,7 @@ struct geometry_shader_uid_data
bool IsPassthrough() const; bool IsPassthrough() const;
u32 numTexGens : 4; u32 numTexGens : 4;
PrimitiveType primitive_type : 2; u32 primitive_type : 2;
}; };
#pragma pack() #pragma pack()

2
Source/Core/VideoCommon/RenderBase.h
View File

@ -68,7 +68,7 @@ public:
virtual void SetScissorRect(const EFBRectangle& rc) {} virtual void SetScissorRect(const EFBRectangle& rc) {}
virtual void SetRasterizationState(const RasterizationState& state) {} virtual void SetRasterizationState(const RasterizationState& state) {}
virtual void SetDepthState(const DepthState& state) {} virtual void SetDepthState(const DepthState& state) {}
virtual void SetSamplerState(int stage, int texindex, bool custom_tex) {} virtual void SetSamplerState(u32 index, const SamplerState& state) {}
virtual void SetInterlacingMode() {} virtual void SetInterlacingMode() {}
virtual void SetViewport() {} virtual void SetViewport() {}
virtual void SetFullscreen(bool enable_fullscreen) {} virtual void SetFullscreen(bool enable_fullscreen) {}

82
Source/Core/VideoCommon/RenderState.cpp
View File

@ -3,6 +3,9 @@
// Refer to the license.txt file included. // Refer to the license.txt file included.
#include "VideoCommon/RenderState.h" #include "VideoCommon/RenderState.h"
#include <algorithm>
#include <array>
#include "VideoCommon/SamplerCommon.h"
void RasterizationState::Generate(const BPMemory& bp, PrimitiveType primitive_type) void RasterizationState::Generate(const BPMemory& bp, PrimitiveType primitive_type)
{ {
@ -14,6 +17,12 @@ void RasterizationState::Generate(const BPMemory& bp, PrimitiveType primitive_ty
cullmode = GenMode::CULL_NONE; cullmode = GenMode::CULL_NONE;
} }
RasterizationState& RasterizationState::operator=(const RasterizationState& rhs)
{
hex = rhs.hex;
return *this;
}
void DepthState::Generate(const BPMemory& bp) void DepthState::Generate(const BPMemory& bp)
{ {
testenable = bp.zmode.testenable.Value(); testenable = bp.zmode.testenable.Value();
@ -21,6 +30,12 @@ void DepthState::Generate(const BPMemory& bp)
func = bp.zmode.func.Value(); func = bp.zmode.func.Value();
} }
DepthState& DepthState::operator=(const DepthState& rhs)
{
hex = rhs.hex;
return *this;
}
// If the framebuffer format has no alpha channel, it is assumed to // If the framebuffer format has no alpha channel, it is assumed to
// ONE on blending. As the backends may emulate this framebuffer // ONE on blending. As the backends may emulate this framebuffer
// configuration with an alpha channel, we just drop all references // configuration with an alpha channel, we just drop all references
@ -145,6 +160,43 @@ void BlendingState::Generate(const BPMemory& bp)
} }
} }
BlendingState& BlendingState::operator=(const BlendingState& rhs)
{
hex = rhs.hex;
return *this;
}
void SamplerState::Generate(const BPMemory& bp, u32 index)
{
const FourTexUnits& tex = bpmem.tex[index / 4];
const TexMode0& tm0 = tex.texMode0[index % 4];
const TexMode1& tm1 = tex.texMode1[index % 4];
// GX can configure the mip filter to none. However, D3D and Vulkan can't express this in their
// sampler states. Therefore, we set the min/max LOD to zero if this option is used.
min_filter = (tm0.min_filter & 4) != 0 ? Filter::Linear : Filter::Point;
mipmap_filter = (tm0.min_filter & 3) == TexMode0::TEXF_LINEAR ? Filter::Linear : Filter::Point;
mag_filter = tm0.mag_filter != 0 ? Filter::Linear : Filter::Point;
// If mipmaps are disabled, clamp min/max lod
max_lod = SamplerCommon::AreBpTexMode0MipmapsEnabled(tm0) ? tm1.max_lod : 0;
min_lod = std::min(max_lod.Value(), tm1.min_lod);
lod_bias = SamplerCommon::AreBpTexMode0MipmapsEnabled(tm0) ? tm0.lod_bias : 0;
// Address modes
static constexpr std::array<AddressMode, 4> address_modes = {
{AddressMode::Clamp, AddressMode::Repeat, AddressMode::MirroredRepeat, AddressMode::Repeat}};
wrap_u = address_modes[tm0.wrap_s];
wrap_v = address_modes[tm0.wrap_t];
anisotropic_filtering = 0;
}
SamplerState& SamplerState::operator=(const SamplerState& rhs)
{
hex = rhs.hex;
return *this;
}
namespace RenderState namespace RenderState
{ {
RasterizationState GetNoCullRasterizationState() RasterizationState GetNoCullRasterizationState()
@ -177,4 +229,34 @@ BlendingState GetNoBlendingBlendState()
state.alphaupdate = true; state.alphaupdate = true;
return state; return state;
} }
SamplerState GetPointSamplerState()
{
SamplerState state = {};
state.min_filter = SamplerState::Filter::Point;
state.mag_filter = SamplerState::Filter::Point;
state.mipmap_filter = SamplerState::Filter::Point;
state.wrap_u = SamplerState::AddressMode::Clamp;
state.wrap_v = SamplerState::AddressMode::Clamp;
state.min_lod = 0;
state.max_lod = 255;
state.lod_bias = 0;
state.anisotropic_filtering = false;
return state;
}
SamplerState GetLinearSamplerState()
{
SamplerState state = {};
state.min_filter = SamplerState::Filter::Linear;
state.mag_filter = SamplerState::Filter::Linear;
state.mipmap_filter = SamplerState::Filter::Linear;
state.wrap_u = SamplerState::AddressMode::Clamp;
state.wrap_v = SamplerState::AddressMode::Clamp;
state.min_lod = 0;
state.max_lod = 255;
state.lod_bias = 0;
state.anisotropic_filtering = false;
return state;
}
} }

52
Source/Core/VideoCommon/RenderState.h
View File

@ -21,6 +21,11 @@ union RasterizationState
{ {
void Generate(const BPMemory& bp, PrimitiveType primitive_type); void Generate(const BPMemory& bp, PrimitiveType primitive_type);
RasterizationState& operator=(const RasterizationState& rhs);
bool operator==(const RasterizationState& rhs) const { return hex == rhs.hex; }
bool operator!=(const RasterizationState& rhs) const { return hex != rhs.hex; }
bool operator<(const RasterizationState& rhs) const { return hex < rhs.hex; }
BitField<0, 2, GenMode::CullMode> cullmode; BitField<0, 2, GenMode::CullMode> cullmode;
BitField<3, 2, PrimitiveType> primitive; BitField<3, 2, PrimitiveType> primitive;
@ -31,6 +36,11 @@ union DepthState
{ {
void Generate(const BPMemory& bp); void Generate(const BPMemory& bp);
DepthState& operator=(const DepthState& rhs);
bool operator==(const DepthState& rhs) const { return hex == rhs.hex; }
bool operator!=(const DepthState& rhs) const { return hex != rhs.hex; }
bool operator<(const DepthState& rhs) const { return hex < rhs.hex; }
BitField<0, 1, u32> testenable; BitField<0, 1, u32> testenable;
BitField<1, 1, u32> updateenable; BitField<1, 1, u32> updateenable;
BitField<2, 3, ZMode::CompareMode> func; BitField<2, 3, ZMode::CompareMode> func;
@ -42,6 +52,11 @@ union BlendingState
{ {
void Generate(const BPMemory& bp); void Generate(const BPMemory& bp);
BlendingState& operator=(const BlendingState& rhs);
bool operator==(const BlendingState& rhs) const { return hex == rhs.hex; }
bool operator!=(const BlendingState& rhs) const { return hex != rhs.hex; }
bool operator<(const BlendingState& rhs) const { return hex < rhs.hex; }
BitField<0, 1, u32> blendenable; BitField<0, 1, u32> blendenable;
BitField<1, 1, u32> logicopenable; BitField<1, 1, u32> logicopenable;
BitField<2, 1, u32> dstalpha; BitField<2, 1, u32> dstalpha;
@ -59,9 +74,46 @@ union BlendingState
u32 hex; u32 hex;
}; };
union SamplerState
{
enum class Filter : u32
{
Point,
Linear
};
enum class AddressMode : u32
{
Clamp,
Repeat,
MirroredRepeat
};
void Generate(const BPMemory& bp, u32 index);
SamplerState& operator=(const SamplerState& rhs);
bool operator==(const SamplerState& rhs) const { return hex == rhs.hex; }
bool operator!=(const SamplerState& rhs) const { return hex != rhs.hex; }
bool operator<(const SamplerState& rhs) const { return hex < rhs.hex; }
BitField<0, 1, Filter> min_filter;
BitField<1, 1, Filter> mag_filter;
BitField<2, 1, Filter> mipmap_filter;
BitField<3, 2, AddressMode> wrap_u;
BitField<5, 2, AddressMode> wrap_v;
BitField<7, 8, u32> min_lod; // multiplied by 16
BitField<15, 8, u32> max_lod; // multiplied by 16
BitField<23, 8, s32> lod_bias; // multiplied by 32
BitField<31, 1, u32> anisotropic_filtering;
u32 hex;
};
namespace RenderState namespace RenderState
{ {
RasterizationState GetNoCullRasterizationState(); RasterizationState GetNoCullRasterizationState();
DepthState GetNoDepthTestingDepthStencilState(); DepthState GetNoDepthTestingDepthStencilState();
BlendingState GetNoBlendingBlendState(); BlendingState GetNoBlendingBlendState();
SamplerState GetPointSamplerState();
SamplerState GetLinearSamplerState();
} }

49
Source/Core/VideoCommon/VertexManagerBase.cpp
View File

@ -24,6 +24,7 @@
#include "VideoCommon/PerfQueryBase.h" #include "VideoCommon/PerfQueryBase.h"
#include "VideoCommon/PixelShaderManager.h" #include "VideoCommon/PixelShaderManager.h"
#include "VideoCommon/RenderBase.h" #include "VideoCommon/RenderBase.h"
#include "VideoCommon/SamplerCommon.h"
#include "VideoCommon/TextureCacheBase.h" #include "VideoCommon/TextureCacheBase.h"
#include "VideoCommon/VertexLoaderManager.h" #include "VideoCommon/VertexLoaderManager.h"
#include "VideoCommon/VertexShaderManager.h" #include "VideoCommon/VertexShaderManager.h"
@ -208,6 +209,52 @@ std::pair<size_t, size_t> VertexManagerBase::ResetFlushAspectRatioCount()
return val; return val;
} }
static void SetSamplerState(u32 index, bool custom_tex)
{
const FourTexUnits& tex = bpmem.tex[index / 4];
const TexMode0& tm0 = tex.texMode0[index % 4];
SamplerState state = {};
state.Generate(bpmem, index);
// Force texture filtering config option.
if (g_ActiveConfig.bForceFiltering)
{
state.min_filter = SamplerState::Filter::Linear;
state.mag_filter = SamplerState::Filter::Linear;
state.mipmap_filter = SamplerCommon::AreBpTexMode0MipmapsEnabled(tm0) ?
SamplerState::Filter::Linear :
SamplerState::Filter::Point;
}
// Custom textures may have a greater number of mips
if (custom_tex)
state.max_lod = 255;
// Anisotropic filtering option.
if (g_ActiveConfig.iMaxAnisotropy != 0 && !SamplerCommon::IsBpTexMode0PointFiltering(tm0))
{
// https://www.opengl.org/registry/specs/EXT/texture_filter_anisotropic.txt
// For predictable results on all hardware/drivers, only use one of:
// GL_LINEAR + GL_LINEAR (No Mipmaps [Bilinear])
// GL_LINEAR + GL_LINEAR_MIPMAP_LINEAR (w/ Mipmaps [Trilinear])
// Letting the game set other combinations will have varying arbitrary results;
// possibly being interpreted as equal to bilinear/trilinear, implicitly
// disabling anisotropy, or changing the anisotropic algorithm employed.
state.min_filter = SamplerState::Filter::Linear;
state.mag_filter = SamplerState::Filter::Linear;
if (SamplerCommon::AreBpTexMode0MipmapsEnabled(tm0))
state.mipmap_filter = SamplerState::Filter::Linear;
state.anisotropic_filtering = 1;
}
else
{
state.anisotropic_filtering = 0;
}
g_renderer->SetSamplerState(index, state);
}
void VertexManagerBase::Flush() void VertexManagerBase::Flush()
{ {
if (m_is_flushed) if (m_is_flushed)
@ -276,7 +323,7 @@ void VertexManagerBase::Flush()
if (tentry) if (tentry)
{ {
g_renderer->SetSamplerState(i & 3, i >> 2, tentry->is_custom_tex); SetSamplerState(i, tentry->is_custom_tex);
PixelShaderManager::SetTexDims(i, tentry->native_width, tentry->native_height); PixelShaderManager::SetTexDims(i, tentry->native_width, tentry->native_height);
} }
else else