Vulkan: Implement Dynamic States 2

This commit is contained in:
Fernando Sahmkow 2022-12-05 15:44:10 +01:00
parent cb1497d0d7
commit c897c55e3c
13 changed files with 315 additions and 66 deletions

View file

@ -48,23 +48,16 @@ void RefreshXfbState(VideoCommon::TransformFeedbackState& state, const Maxwell&
} }
} // Anonymous namespace } // Anonymous namespace
void FixedPipelineState::Refresh(Tegra::Engines::Maxwell3D& maxwell3d, void FixedPipelineState::Refresh(Tegra::Engines::Maxwell3D& maxwell3d, DynamicFeatures& features) {
bool has_extended_dynamic_state, bool has_dynamic_vertex_input) {
const Maxwell& regs = maxwell3d.regs; const Maxwell& regs = maxwell3d.regs;
const auto topology_ = maxwell3d.draw_manager->GetDrawState().topology; const auto topology_ = maxwell3d.draw_manager->GetDrawState().topology;
const std::array enabled_lut{
regs.polygon_offset_point_enable,
regs.polygon_offset_line_enable,
regs.polygon_offset_fill_enable,
};
const u32 topology_index = static_cast<u32>(topology_);
raw1 = 0; raw1 = 0;
extended_dynamic_state.Assign(has_extended_dynamic_state ? 1 : 0); extended_dynamic_state.Assign(features.has_extended_dynamic_state ? 1 : 0);
dynamic_vertex_input.Assign(has_dynamic_vertex_input ? 1 : 0); extended_dynamic_state_2.Assign(features.has_extended_dynamic_state_2 ? 1 : 0);
extended_dynamic_state_3.Assign(features.has_extended_dynamic_state_3 ? 1 : 0);
dynamic_vertex_input.Assign(features.has_dynamic_vertex_input ? 1 : 0);
xfb_enabled.Assign(regs.transform_feedback_enabled != 0); xfb_enabled.Assign(regs.transform_feedback_enabled != 0);
primitive_restart_enable.Assign(regs.primitive_restart.enabled != 0 ? 1 : 0);
depth_bias_enable.Assign(enabled_lut[POLYGON_OFFSET_ENABLE_LUT[topology_index]] != 0 ? 1 : 0);
depth_clamp_disabled.Assign(regs.viewport_clip_control.geometry_clip == depth_clamp_disabled.Assign(regs.viewport_clip_control.geometry_clip ==
Maxwell::ViewportClipControl::GeometryClip::Passthrough || Maxwell::ViewportClipControl::GeometryClip::Passthrough ||
regs.viewport_clip_control.geometry_clip == regs.viewport_clip_control.geometry_clip ==
@ -84,7 +77,7 @@ void FixedPipelineState::Refresh(Tegra::Engines::Maxwell3D& maxwell3d,
msaa_mode.Assign(regs.anti_alias_samples_mode); msaa_mode.Assign(regs.anti_alias_samples_mode);
raw2 = 0; raw2 = 0;
rasterize_enable.Assign(regs.rasterize_enable != 0 ? 1 : 0);
const auto test_func = const auto test_func =
regs.alpha_test_enabled != 0 ? regs.alpha_test_func : Maxwell::ComparisonOp::Always_GL; regs.alpha_test_enabled != 0 ? regs.alpha_test_func : Maxwell::ComparisonOp::Always_GL;
alpha_test_func.Assign(PackComparisonOp(test_func)); alpha_test_func.Assign(PackComparisonOp(test_func));
@ -106,7 +99,7 @@ void FixedPipelineState::Refresh(Tegra::Engines::Maxwell3D& maxwell3d,
point_size = Common::BitCast<u32>(regs.point_size); point_size = Common::BitCast<u32>(regs.point_size);
if (maxwell3d.dirty.flags[Dirty::VertexInput]) { if (maxwell3d.dirty.flags[Dirty::VertexInput]) {
if (has_dynamic_vertex_input) { if (features.has_dynamic_vertex_input) {
// Dirty flag will be reset by the command buffer update // Dirty flag will be reset by the command buffer update
static constexpr std::array LUT{ static constexpr std::array LUT{
0u, // Invalid 0u, // Invalid
@ -158,9 +151,17 @@ void FixedPipelineState::Refresh(Tegra::Engines::Maxwell3D& maxwell3d,
return static_cast<u16>(viewport.swizzle.raw); return static_cast<u16>(viewport.swizzle.raw);
}); });
} }
dynamic_state.raw1 = 0;
dynamic_state.raw2 = 0;
if (!extended_dynamic_state) { if (!extended_dynamic_state) {
dynamic_state.Refresh(regs); dynamic_state.Refresh(regs);
} }
if (!extended_dynamic_state_2) {
dynamic_state.Refresh2(regs, topology);
}
if (!extended_dynamic_state_3) {
dynamic_state.Refresh3(regs);
}
if (xfb_enabled) { if (xfb_enabled) {
RefreshXfbState(xfb_state, regs); RefreshXfbState(xfb_state, regs);
} }
@ -212,8 +213,6 @@ void FixedPipelineState::DynamicState::Refresh(const Maxwell& regs) {
packed_front_face = 1 - packed_front_face; packed_front_face = 1 - packed_front_face;
} }
raw1 = 0;
raw2 = 0;
front.action_stencil_fail.Assign(PackStencilOp(regs.stencil_front_op.fail)); front.action_stencil_fail.Assign(PackStencilOp(regs.stencil_front_op.fail));
front.action_depth_fail.Assign(PackStencilOp(regs.stencil_front_op.zfail)); front.action_depth_fail.Assign(PackStencilOp(regs.stencil_front_op.zfail));
front.action_depth_pass.Assign(PackStencilOp(regs.stencil_front_op.zpass)); front.action_depth_pass.Assign(PackStencilOp(regs.stencil_front_op.zpass));
@ -242,6 +241,21 @@ void FixedPipelineState::DynamicState::Refresh(const Maxwell& regs) {
}); });
} }
void FixedPipelineState::DynamicState::Refresh2(const Maxwell& regs, Maxwell::PrimitiveTopology topology_) {
const std::array enabled_lut{
regs.polygon_offset_point_enable,
regs.polygon_offset_line_enable,
regs.polygon_offset_fill_enable,
};
const u32 topology_index = static_cast<u32>(topology_);
rasterize_enable.Assign(regs.rasterize_enable != 0 ? 1 : 0);
primitive_restart_enable.Assign(regs.primitive_restart.enabled != 0 ? 1 : 0);
depth_bias_enable.Assign(enabled_lut[POLYGON_OFFSET_ENABLE_LUT[topology_index]] != 0 ? 1 : 0);
}
void FixedPipelineState::DynamicState::Refresh3(const Maxwell&) {}
size_t FixedPipelineState::Hash() const noexcept { size_t FixedPipelineState::Hash() const noexcept {
const u64 hash = Common::CityHash64(reinterpret_cast<const char*>(this), Size()); const u64 hash = Common::CityHash64(reinterpret_cast<const char*>(this), Size());
return static_cast<size_t>(hash); return static_cast<size_t>(hash);

View file

@ -17,6 +17,14 @@ namespace Vulkan {
using Maxwell = Tegra::Engines::Maxwell3D::Regs; using Maxwell = Tegra::Engines::Maxwell3D::Regs;
struct DynamicFeatures {
bool has_extended_dynamic_state;
bool has_extended_dynamic_state_2;
bool has_extended_dynamic_state_2_extra;
bool has_extended_dynamic_state_3;
bool has_dynamic_vertex_input;
};
struct FixedPipelineState { struct FixedPipelineState {
static u32 PackComparisonOp(Maxwell::ComparisonOp op) noexcept; static u32 PackComparisonOp(Maxwell::ComparisonOp op) noexcept;
static Maxwell::ComparisonOp UnpackComparisonOp(u32 packed) noexcept; static Maxwell::ComparisonOp UnpackComparisonOp(u32 packed) noexcept;
@ -133,6 +141,14 @@ struct FixedPipelineState {
struct DynamicState { struct DynamicState {
union { union {
u32 raw1; u32 raw1;
BitField<0, 2, u32> cull_face;
BitField<2, 1, u32> cull_enable;
BitField<3, 1, u32> primitive_restart_enable;
BitField<4, 1, u32> depth_bias_enable;
BitField<5, 1, u32> rasterize_enable;
};
union {
u32 raw2;
StencilFace<0> front; StencilFace<0> front;
StencilFace<12> back; StencilFace<12> back;
BitField<24, 1, u32> stencil_enable; BitField<24, 1, u32> stencil_enable;
@ -142,15 +158,12 @@ struct FixedPipelineState {
BitField<28, 1, u32> front_face; BitField<28, 1, u32> front_face;
BitField<29, 3, u32> depth_test_func; BitField<29, 3, u32> depth_test_func;
}; };
union {
u32 raw2;
BitField<0, 2, u32> cull_face;
BitField<2, 1, u32> cull_enable;
};
// Vertex stride is a 12 bits value, we have 4 bits to spare per element // Vertex stride is a 12 bits value, we have 4 bits to spare per element
std::array<u16, Maxwell::NumVertexArrays> vertex_strides; std::array<u16, Maxwell::NumVertexArrays> vertex_strides;
void Refresh(const Maxwell& regs); void Refresh(const Maxwell& regs);
void Refresh2(const Maxwell& regs, Maxwell::PrimitiveTopology topology);
void Refresh3(const Maxwell& regs);
Maxwell::ComparisonOp DepthTestFunc() const noexcept { Maxwell::ComparisonOp DepthTestFunc() const noexcept {
return UnpackComparisonOp(depth_test_func); return UnpackComparisonOp(depth_test_func);
@ -168,10 +181,10 @@ struct FixedPipelineState {
union { union {
u32 raw1; u32 raw1;
BitField<0, 1, u32> extended_dynamic_state; BitField<0, 1, u32> extended_dynamic_state;
BitField<1, 1, u32> dynamic_vertex_input; BitField<1, 1, u32> extended_dynamic_state_2;
BitField<2, 1, u32> xfb_enabled; BitField<2, 1, u32> extended_dynamic_state_3;
BitField<3, 1, u32> primitive_restart_enable; BitField<3, 1, u32> dynamic_vertex_input;
BitField<4, 1, u32> depth_bias_enable; BitField<4, 1, u32> xfb_enabled;
BitField<5, 1, u32> depth_clamp_disabled; BitField<5, 1, u32> depth_clamp_disabled;
BitField<6, 1, u32> ndc_minus_one_to_one; BitField<6, 1, u32> ndc_minus_one_to_one;
BitField<7, 2, u32> polygon_mode; BitField<7, 2, u32> polygon_mode;
@ -186,7 +199,6 @@ struct FixedPipelineState {
}; };
union { union {
u32 raw2; u32 raw2;
BitField<0, 1, u32> rasterize_enable;
BitField<1, 3, u32> alpha_test_func; BitField<1, 3, u32> alpha_test_func;
BitField<4, 1, u32> early_z; BitField<4, 1, u32> early_z;
BitField<5, 1, u32> depth_enabled; BitField<5, 1, u32> depth_enabled;
@ -215,8 +227,7 @@ struct FixedPipelineState {
DynamicState dynamic_state; DynamicState dynamic_state;
VideoCommon::TransformFeedbackState xfb_state; VideoCommon::TransformFeedbackState xfb_state;
void Refresh(Tegra::Engines::Maxwell3D& maxwell3d, bool has_extended_dynamic_state, void Refresh(Tegra::Engines::Maxwell3D& maxwell3d, DynamicFeatures& features);
bool has_dynamic_vertex_input);
size_t Hash() const noexcept; size_t Hash() const noexcept;
@ -231,14 +242,18 @@ struct FixedPipelineState {
// When transform feedback is enabled, use the whole struct // When transform feedback is enabled, use the whole struct
return sizeof(*this); return sizeof(*this);
} }
if (dynamic_vertex_input) { if (dynamic_vertex_input && extended_dynamic_state_2) {
// Exclude dynamic state and attributes // Exclude dynamic state and attributes
return offsetof(FixedPipelineState, attributes); return offsetof(FixedPipelineState, attributes);
} }
if (extended_dynamic_state) { if (extended_dynamic_state_2) {
// Exclude dynamic state // Exclude dynamic state
return offsetof(FixedPipelineState, dynamic_state); return offsetof(FixedPipelineState, dynamic_state);
} }
if (extended_dynamic_state) {
// Exclude dynamic state
return offsetof(FixedPipelineState, dynamic_state.raw2);
}
// Default // Default
return offsetof(FixedPipelineState, xfb_state); return offsetof(FixedPipelineState, xfb_state);
} }

View file

@ -525,6 +525,9 @@ void GraphicsPipeline::MakePipeline(VkRenderPass render_pass) {
if (!key.state.extended_dynamic_state) { if (!key.state.extended_dynamic_state) {
dynamic = key.state.dynamic_state; dynamic = key.state.dynamic_state;
} }
if (!key.state.extended_dynamic_state_2) {
dynamic.raw2 = key.state.dynamic_state.raw2;
}
static_vector<VkVertexInputBindingDescription, 32> vertex_bindings; static_vector<VkVertexInputBindingDescription, 32> vertex_bindings;
static_vector<VkVertexInputBindingDivisorDescriptionEXT, 32> vertex_binding_divisors; static_vector<VkVertexInputBindingDivisorDescriptionEXT, 32> vertex_binding_divisors;
static_vector<VkVertexInputAttributeDescription, 32> vertex_attributes; static_vector<VkVertexInputAttributeDescription, 32> vertex_attributes;
@ -625,7 +628,7 @@ void GraphicsPipeline::MakePipeline(VkRenderPass render_pass) {
.pNext = nullptr, .pNext = nullptr,
.flags = 0, .flags = 0,
.topology = input_assembly_topology, .topology = input_assembly_topology,
.primitiveRestartEnable = key.state.primitive_restart_enable != 0 && .primitiveRestartEnable = key.state.dynamic_state.primitive_restart_enable != 0 &&
((input_assembly_topology != VK_PRIMITIVE_TOPOLOGY_PATCH_LIST && ((input_assembly_topology != VK_PRIMITIVE_TOPOLOGY_PATCH_LIST &&
device.IsTopologyListPrimitiveRestartSupported()) || device.IsTopologyListPrimitiveRestartSupported()) ||
SupportsPrimitiveRestart(input_assembly_topology) || SupportsPrimitiveRestart(input_assembly_topology) ||
@ -674,13 +677,13 @@ void GraphicsPipeline::MakePipeline(VkRenderPass render_pass) {
.depthClampEnable = .depthClampEnable =
static_cast<VkBool32>(key.state.depth_clamp_disabled == 0 ? VK_TRUE : VK_FALSE), static_cast<VkBool32>(key.state.depth_clamp_disabled == 0 ? VK_TRUE : VK_FALSE),
.rasterizerDiscardEnable = .rasterizerDiscardEnable =
static_cast<VkBool32>(key.state.rasterize_enable == 0 ? VK_TRUE : VK_FALSE), static_cast<VkBool32>(dynamic.rasterize_enable == 0 ? VK_TRUE : VK_FALSE),
.polygonMode = .polygonMode =
MaxwellToVK::PolygonMode(FixedPipelineState::UnpackPolygonMode(key.state.polygon_mode)), MaxwellToVK::PolygonMode(FixedPipelineState::UnpackPolygonMode(key.state.polygon_mode)),
.cullMode = static_cast<VkCullModeFlags>( .cullMode = static_cast<VkCullModeFlags>(
dynamic.cull_enable ? MaxwellToVK::CullFace(dynamic.CullFace()) : VK_CULL_MODE_NONE), dynamic.cull_enable ? MaxwellToVK::CullFace(dynamic.CullFace()) : VK_CULL_MODE_NONE),
.frontFace = MaxwellToVK::FrontFace(dynamic.FrontFace()), .frontFace = MaxwellToVK::FrontFace(dynamic.FrontFace()),
.depthBiasEnable = key.state.depth_bias_enable, .depthBiasEnable = (dynamic.depth_bias_enable == 0 ? VK_TRUE : VK_FALSE),
.depthBiasConstantFactor = 0.0f, .depthBiasConstantFactor = 0.0f,
.depthBiasClamp = 0.0f, .depthBiasClamp = 0.0f,
.depthBiasSlopeFactor = 0.0f, .depthBiasSlopeFactor = 0.0f,
@ -788,7 +791,7 @@ void GraphicsPipeline::MakePipeline(VkRenderPass render_pass) {
.pAttachments = cb_attachments.data(), .pAttachments = cb_attachments.data(),
.blendConstants = {}, .blendConstants = {},
}; };
static_vector<VkDynamicState, 19> dynamic_states{ static_vector<VkDynamicState, 22> dynamic_states{
VK_DYNAMIC_STATE_VIEWPORT, VK_DYNAMIC_STATE_SCISSOR, VK_DYNAMIC_STATE_VIEWPORT, VK_DYNAMIC_STATE_SCISSOR,
VK_DYNAMIC_STATE_DEPTH_BIAS, VK_DYNAMIC_STATE_BLEND_CONSTANTS, VK_DYNAMIC_STATE_DEPTH_BIAS, VK_DYNAMIC_STATE_BLEND_CONSTANTS,
VK_DYNAMIC_STATE_DEPTH_BOUNDS, VK_DYNAMIC_STATE_STENCIL_COMPARE_MASK, VK_DYNAMIC_STATE_DEPTH_BOUNDS, VK_DYNAMIC_STATE_STENCIL_COMPARE_MASK,
@ -811,6 +814,14 @@ void GraphicsPipeline::MakePipeline(VkRenderPass render_pass) {
dynamic_states.push_back(VK_DYNAMIC_STATE_VERTEX_INPUT_EXT); dynamic_states.push_back(VK_DYNAMIC_STATE_VERTEX_INPUT_EXT);
} }
dynamic_states.insert(dynamic_states.end(), extended.begin(), extended.end()); dynamic_states.insert(dynamic_states.end(), extended.begin(), extended.end());
if (key.state.extended_dynamic_state_2) {
static constexpr std::array extended2{
VK_DYNAMIC_STATE_DEPTH_BIAS_ENABLE_EXT,
VK_DYNAMIC_STATE_PRIMITIVE_RESTART_ENABLE_EXT,
VK_DYNAMIC_STATE_RASTERIZER_DISCARD_ENABLE_EXT,
};
dynamic_states.insert(dynamic_states.end(), extended2.begin(), extended2.end());
}
} }
const VkPipelineDynamicStateCreateInfo dynamic_state_ci{ const VkPipelineDynamicStateCreateInfo dynamic_state_ci{
.sType = VK_STRUCTURE_TYPE_PIPELINE_DYNAMIC_STATE_CREATE_INFO, .sType = VK_STRUCTURE_TYPE_PIPELINE_DYNAMIC_STATE_CREATE_INFO,

View file

@ -351,6 +351,14 @@ PipelineCache::PipelineCache(RasterizerVulkan& rasterizer_, const Device& device
LOG_WARNING(Render_Vulkan, "maxVertexInputBindings is too low: {} < {}", LOG_WARNING(Render_Vulkan, "maxVertexInputBindings is too low: {} < {}",
device.GetMaxVertexInputBindings(), Maxwell::NumVertexArrays); device.GetMaxVertexInputBindings(), Maxwell::NumVertexArrays);
} }
dynamic_features = DynamicFeatures{
.has_extended_dynamic_state = device.IsExtExtendedDynamicStateSupported(),
.has_extended_dynamic_state_2 = device.IsExtExtendedDynamicState2Supported(),
.has_extended_dynamic_state_2_extra = device.IsExtExtendedDynamicState2ExtrasSupported(),
.has_extended_dynamic_state_3 = device.IsExtExtendedDynamicState3Supported(),
.has_dynamic_vertex_input = device.IsExtVertexInputDynamicStateSupported(),
};
} }
PipelineCache::~PipelineCache() = default; PipelineCache::~PipelineCache() = default;
@ -362,8 +370,7 @@ GraphicsPipeline* PipelineCache::CurrentGraphicsPipeline() {
current_pipeline = nullptr; current_pipeline = nullptr;
return nullptr; return nullptr;
} }
graphics_key.state.Refresh(*maxwell3d, device.IsExtExtendedDynamicStateSupported(), graphics_key.state.Refresh(*maxwell3d, dynamic_features);
device.IsExtVertexInputDynamicStateSupported());
if (current_pipeline) { if (current_pipeline) {
GraphicsPipeline* const next{current_pipeline->Next(graphics_key)}; GraphicsPipeline* const next{current_pipeline->Next(graphics_key)};
@ -439,14 +446,17 @@ void PipelineCache::LoadDiskResources(u64 title_id, std::stop_token stop_loading
}); });
++state.total; ++state.total;
}}; }};
const bool extended_dynamic_state = device.IsExtExtendedDynamicStateSupported();
const bool dynamic_vertex_input = device.IsExtVertexInputDynamicStateSupported();
const auto load_graphics{[&](std::ifstream& file, std::vector<FileEnvironment> envs) { const auto load_graphics{[&](std::ifstream& file, std::vector<FileEnvironment> envs) {
GraphicsPipelineCacheKey key; GraphicsPipelineCacheKey key;
file.read(reinterpret_cast<char*>(&key), sizeof(key)); file.read(reinterpret_cast<char*>(&key), sizeof(key));
if ((key.state.extended_dynamic_state != 0) != extended_dynamic_state || if ((key.state.extended_dynamic_state != 0) !=
(key.state.dynamic_vertex_input != 0) != dynamic_vertex_input) { dynamic_features.has_extended_dynamic_state ||
(key.state.extended_dynamic_state_2 != 0) !=
dynamic_features.has_extended_dynamic_state_2 ||
(key.state.extended_dynamic_state_3 != 0) !=
dynamic_features.has_extended_dynamic_state_3 ||
(key.state.dynamic_vertex_input != 0) != dynamic_features.has_dynamic_vertex_input) {
return; return;
} }
workers.QueueWork([this, key, envs = std::move(envs), &state, &callback]() mutable { workers.QueueWork([this, key, envs = std::move(envs), &state, &callback]() mutable {

View file

@ -160,6 +160,7 @@ private:
Common::ThreadWorker workers; Common::ThreadWorker workers;
Common::ThreadWorker serialization_thread; Common::ThreadWorker serialization_thread;
DynamicFeatures dynamic_features;
}; };
} // namespace Vulkan } // namespace Vulkan

View file

@ -680,16 +680,26 @@ void RasterizerVulkan::UpdateDynamicStates() {
UpdateLineWidth(regs); UpdateLineWidth(regs);
if (device.IsExtExtendedDynamicStateSupported()) { if (device.IsExtExtendedDynamicStateSupported()) {
UpdateCullMode(regs); UpdateCullMode(regs);
UpdateDepthBoundsTestEnable(regs);
UpdateDepthTestEnable(regs);
UpdateDepthWriteEnable(regs);
UpdateDepthCompareOp(regs); UpdateDepthCompareOp(regs);
UpdateFrontFace(regs); UpdateFrontFace(regs);
UpdateStencilOp(regs); UpdateStencilOp(regs);
UpdateStencilTestEnable(regs);
if (device.IsExtVertexInputDynamicStateSupported()) { if (device.IsExtVertexInputDynamicStateSupported()) {
UpdateVertexInput(regs); UpdateVertexInput(regs);
} }
if (state_tracker.TouchStateEnable()) {
UpdateDepthBoundsTestEnable(regs);
UpdateDepthTestEnable(regs);
UpdateDepthWriteEnable(regs);
UpdateStencilTestEnable(regs);
if (device.IsExtExtendedDynamicState2Supported()) {
UpdatePrimitiveRestartEnable(regs);
UpdateRasterizerDiscardEnable(regs);
UpdateDepthBiasEnable(regs);
}
}
} }
} }
@ -909,6 +919,58 @@ void RasterizerVulkan::UpdateDepthWriteEnable(Tegra::Engines::Maxwell3D::Regs& r
}); });
} }
void RasterizerVulkan::UpdatePrimitiveRestartEnable(Tegra::Engines::Maxwell3D::Regs& regs) {
if (!state_tracker.TouchPrimitiveRestartEnable()) {
return;
}
scheduler.Record([enable = regs.primitive_restart.enabled](vk::CommandBuffer cmdbuf) {
cmdbuf.SetPrimitiveRestartEnableEXT(enable);
});
}
void RasterizerVulkan::UpdateRasterizerDiscardEnable(Tegra::Engines::Maxwell3D::Regs& regs) {
if (!state_tracker.TouchRasterizerDiscardEnable()) {
return;
}
scheduler.Record([disable = regs.rasterize_enable](vk::CommandBuffer cmdbuf) {
cmdbuf.SetRasterizerDiscardEnableEXT(disable == 0);
});
}
void RasterizerVulkan::UpdateDepthBiasEnable(Tegra::Engines::Maxwell3D::Regs& regs) {
if (!state_tracker.TouchDepthBiasEnable()) {
return;
}
constexpr size_t POINT = 0;
constexpr size_t LINE = 1;
constexpr size_t POLYGON = 2;
constexpr std::array POLYGON_OFFSET_ENABLE_LUT = {
POINT, // Points
LINE, // Lines
LINE, // LineLoop
LINE, // LineStrip
POLYGON, // Triangles
POLYGON, // TriangleStrip
POLYGON, // TriangleFan
POLYGON, // Quads
POLYGON, // QuadStrip
POLYGON, // Polygon
LINE, // LinesAdjacency
LINE, // LineStripAdjacency
POLYGON, // TrianglesAdjacency
POLYGON, // TriangleStripAdjacency
POLYGON, // Patches
};
const std::array enabled_lut{
regs.polygon_offset_point_enable,
regs.polygon_offset_line_enable,
regs.polygon_offset_fill_enable,
};
const u32 topology_index = static_cast<u32>(maxwell3d->draw_manager->GetDrawState().topology);
const u32 enable = enabled_lut[POLYGON_OFFSET_ENABLE_LUT[topology_index]];
scheduler.Record([enable](vk::CommandBuffer cmdbuf) { cmdbuf.SetDepthBiasEnableEXT(enable); });
}
void RasterizerVulkan::UpdateDepthCompareOp(Tegra::Engines::Maxwell3D::Regs& regs) { void RasterizerVulkan::UpdateDepthCompareOp(Tegra::Engines::Maxwell3D::Regs& regs) {
if (!state_tracker.TouchDepthCompareOp()) { if (!state_tracker.TouchDepthCompareOp()) {
return; return;

View file

@ -139,6 +139,9 @@ private:
void UpdateDepthTestEnable(Tegra::Engines::Maxwell3D::Regs& regs); void UpdateDepthTestEnable(Tegra::Engines::Maxwell3D::Regs& regs);
void UpdateDepthWriteEnable(Tegra::Engines::Maxwell3D::Regs& regs); void UpdateDepthWriteEnable(Tegra::Engines::Maxwell3D::Regs& regs);
void UpdateDepthCompareOp(Tegra::Engines::Maxwell3D::Regs& regs); void UpdateDepthCompareOp(Tegra::Engines::Maxwell3D::Regs& regs);
void UpdatePrimitiveRestartEnable(Tegra::Engines::Maxwell3D::Regs& regs);
void UpdateRasterizerDiscardEnable(Tegra::Engines::Maxwell3D::Regs& regs);
void UpdateDepthBiasEnable(Tegra::Engines::Maxwell3D::Regs& regs);
void UpdateFrontFace(Tegra::Engines::Maxwell3D::Regs& regs); void UpdateFrontFace(Tegra::Engines::Maxwell3D::Regs& regs);
void UpdateStencilOp(Tegra::Engines::Maxwell3D::Regs& regs); void UpdateStencilOp(Tegra::Engines::Maxwell3D::Regs& regs);
void UpdateStencilTestEnable(Tegra::Engines::Maxwell3D::Regs& regs); void UpdateStencilTestEnable(Tegra::Engines::Maxwell3D::Regs& regs);

View file

@ -27,10 +27,27 @@ using Flags = Maxwell3D::DirtyState::Flags;
Flags MakeInvalidationFlags() { Flags MakeInvalidationFlags() {
static constexpr int INVALIDATION_FLAGS[]{ static constexpr int INVALIDATION_FLAGS[]{
Viewports, Scissors, DepthBias, BlendConstants, DepthBounds, Viewports,
StencilProperties, LineWidth, CullMode, DepthBoundsEnable, DepthTestEnable, Scissors,
DepthWriteEnable, DepthCompareOp, FrontFace, StencilOp, StencilTestEnable, DepthBias,
VertexBuffers, VertexInput, BlendConstants,
DepthBounds,
StencilProperties,
LineWidth,
CullMode,
DepthBoundsEnable,
DepthTestEnable,
DepthWriteEnable,
DepthCompareOp,
FrontFace,
StencilOp,
StencilTestEnable,
VertexBuffers,
VertexInput,
StateEnable,
PrimitiveRestartEnable,
RasterizerDiscardEnable,
DepthBiasEnable,
}; };
Flags flags{}; Flags flags{};
for (const int flag : INVALIDATION_FLAGS) { for (const int flag : INVALIDATION_FLAGS) {
@ -96,16 +113,20 @@ void SetupDirtyCullMode(Tables& tables) {
table[OFF(gl_cull_test_enabled)] = CullMode; table[OFF(gl_cull_test_enabled)] = CullMode;
} }
void SetupDirtyDepthBoundsEnable(Tables& tables) { void SetupDirtyStateEnable(Tables& tables) {
tables[0][OFF(depth_bounds_enable)] = DepthBoundsEnable; const auto setup = [&](size_t position, u8 flag) {
} tables[0][position] = flag;
tables[1][position] = StateEnable;
void SetupDirtyDepthTestEnable(Tables& tables) { };
tables[0][OFF(depth_test_enable)] = DepthTestEnable; setup(OFF(depth_bounds_enable), DepthBoundsEnable);
} setup(OFF(depth_test_enable), DepthTestEnable);
setup(OFF(depth_write_enabled), DepthWriteEnable);
void SetupDirtyDepthWriteEnable(Tables& tables) { setup(OFF(stencil_enable), StencilTestEnable);
tables[0][OFF(depth_write_enabled)] = DepthWriteEnable; setup(OFF(primitive_restart.enabled), PrimitiveRestartEnable);
setup(OFF(rasterize_enable), RasterizerDiscardEnable);
setup(OFF(polygon_offset_point_enable), DepthBiasEnable);
setup(OFF(polygon_offset_line_enable), DepthBiasEnable);
setup(OFF(polygon_offset_fill_enable), DepthBiasEnable);
} }
void SetupDirtyDepthCompareOp(Tables& tables) { void SetupDirtyDepthCompareOp(Tables& tables) {
@ -133,10 +154,6 @@ void SetupDirtyStencilOp(Tables& tables) {
tables[1][OFF(stencil_two_side_enable)] = StencilOp; tables[1][OFF(stencil_two_side_enable)] = StencilOp;
} }
void SetupDirtyStencilTestEnable(Tables& tables) {
tables[0][OFF(stencil_enable)] = StencilTestEnable;
}
void SetupDirtyBlending(Tables& tables) { void SetupDirtyBlending(Tables& tables) {
tables[0][OFF(color_mask_common)] = Blending; tables[0][OFF(color_mask_common)] = Blending;
tables[0][OFF(blend_per_target_enabled)] = Blending; tables[0][OFF(blend_per_target_enabled)] = Blending;
@ -185,13 +202,10 @@ void StateTracker::SetupTables(Tegra::Control::ChannelState& channel_state) {
SetupDirtyStencilProperties(tables); SetupDirtyStencilProperties(tables);
SetupDirtyLineWidth(tables); SetupDirtyLineWidth(tables);
SetupDirtyCullMode(tables); SetupDirtyCullMode(tables);
SetupDirtyDepthBoundsEnable(tables); SetupDirtyStateEnable(tables);
SetupDirtyDepthTestEnable(tables);
SetupDirtyDepthWriteEnable(tables);
SetupDirtyDepthCompareOp(tables); SetupDirtyDepthCompareOp(tables);
SetupDirtyFrontFace(tables); SetupDirtyFrontFace(tables);
SetupDirtyStencilOp(tables); SetupDirtyStencilOp(tables);
SetupDirtyStencilTestEnable(tables);
SetupDirtyBlending(tables); SetupDirtyBlending(tables);
SetupDirtyViewportSwizzles(tables); SetupDirtyViewportSwizzles(tables);
SetupDirtyVertexAttributes(tables); SetupDirtyVertexAttributes(tables);

View file

@ -45,6 +45,10 @@ enum : u8 {
FrontFace, FrontFace,
StencilOp, StencilOp,
StencilTestEnable, StencilTestEnable,
PrimitiveRestartEnable,
RasterizerDiscardEnable,
DepthBiasEnable,
StateEnable,
Blending, Blending,
ViewportSwizzles, ViewportSwizzles,
@ -111,6 +115,10 @@ public:
return Exchange(Dirty::CullMode, false); return Exchange(Dirty::CullMode, false);
} }
bool TouchStateEnable() {
return Exchange(Dirty::StateEnable, false);
}
bool TouchDepthBoundsTestEnable() { bool TouchDepthBoundsTestEnable() {
return Exchange(Dirty::DepthBoundsEnable, false); return Exchange(Dirty::DepthBoundsEnable, false);
} }
@ -123,6 +131,18 @@ public:
return Exchange(Dirty::DepthWriteEnable, false); return Exchange(Dirty::DepthWriteEnable, false);
} }
bool TouchPrimitiveRestartEnable() {
return Exchange(Dirty::PrimitiveRestartEnable, false);
}
bool TouchRasterizerDiscardEnable() {
return Exchange(Dirty::RasterizerDiscardEnable, false);
}
bool TouchDepthBiasEnable() {
return Exchange(Dirty::DepthBiasEnable, false);
}
bool TouchDepthCompareOp() { bool TouchDepthCompareOp() {
return Exchange(Dirty::DepthCompareOp, false); return Exchange(Dirty::DepthCompareOp, false);
} }

View file

@ -569,6 +569,34 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR
LOG_INFO(Render_Vulkan, "Device doesn't support extended dynamic state"); LOG_INFO(Render_Vulkan, "Device doesn't support extended dynamic state");
} }
VkPhysicalDeviceExtendedDynamicState2FeaturesEXT dynamic_state2;
if (ext_extended_dynamic_state2) {
dynamic_state2 = {
.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_EXTENDED_DYNAMIC_STATE_2_FEATURES_EXT,
.pNext = nullptr,
.extendedDynamicState2 = VK_TRUE,
.extendedDynamicState2LogicOp = ext_extended_dynamic_state2_extra ? VK_TRUE : VK_FALSE,
.extendedDynamicState2PatchControlPoints =
ext_extended_dynamic_state2_extra ? VK_TRUE : VK_FALSE,
};
SetNext(next, dynamic_state2);
} else {
LOG_INFO(Render_Vulkan, "Device doesn't support extended dynamic state 2");
}
VkPhysicalDeviceExtendedDynamicState3FeaturesEXT dynamic_state3;
if (ext_extended_dynamic_state3) {
dynamic_state3 = {
.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_EXTENDED_DYNAMIC_STATE_3_FEATURES_EXT,
.pNext = nullptr,
.extendedDynamicState3ColorBlendEnable = VK_TRUE,
.extendedDynamicState3ColorBlendEquation = VK_TRUE,
};
SetNext(next, dynamic_state3);
} else {
LOG_INFO(Render_Vulkan, "Device doesn't support extended dynamic state 3");
}
VkPhysicalDeviceLineRasterizationFeaturesEXT line_raster; VkPhysicalDeviceLineRasterizationFeaturesEXT line_raster;
if (ext_line_rasterization) { if (ext_line_rasterization) {
line_raster = { line_raster = {
@ -1091,6 +1119,8 @@ std::vector<const char*> Device::LoadExtensions(bool requires_surface) {
bool has_ext_transform_feedback{}; bool has_ext_transform_feedback{};
bool has_ext_custom_border_color{}; bool has_ext_custom_border_color{};
bool has_ext_extended_dynamic_state{}; bool has_ext_extended_dynamic_state{};
bool has_ext_extended_dynamic_state2{};
bool has_ext_extended_dynamic_state3{};
bool has_ext_shader_atomic_int64{}; bool has_ext_shader_atomic_int64{};
bool has_ext_provoking_vertex{}; bool has_ext_provoking_vertex{};
bool has_ext_vertex_input_dynamic_state{}; bool has_ext_vertex_input_dynamic_state{};
@ -1135,6 +1165,10 @@ std::vector<const char*> Device::LoadExtensions(bool requires_surface) {
test(has_ext_transform_feedback, VK_EXT_TRANSFORM_FEEDBACK_EXTENSION_NAME, false); test(has_ext_transform_feedback, VK_EXT_TRANSFORM_FEEDBACK_EXTENSION_NAME, false);
test(has_ext_custom_border_color, VK_EXT_CUSTOM_BORDER_COLOR_EXTENSION_NAME, false); test(has_ext_custom_border_color, VK_EXT_CUSTOM_BORDER_COLOR_EXTENSION_NAME, false);
test(has_ext_extended_dynamic_state, VK_EXT_EXTENDED_DYNAMIC_STATE_EXTENSION_NAME, false); test(has_ext_extended_dynamic_state, VK_EXT_EXTENDED_DYNAMIC_STATE_EXTENSION_NAME, false);
test(has_ext_extended_dynamic_state2, VK_EXT_EXTENDED_DYNAMIC_STATE_2_EXTENSION_NAME,
false);
test(has_ext_extended_dynamic_state3, VK_EXT_EXTENDED_DYNAMIC_STATE_3_EXTENSION_NAME,
false);
test(has_ext_subgroup_size_control, VK_EXT_SUBGROUP_SIZE_CONTROL_EXTENSION_NAME, true); test(has_ext_subgroup_size_control, VK_EXT_SUBGROUP_SIZE_CONTROL_EXTENSION_NAME, true);
test(has_ext_provoking_vertex, VK_EXT_PROVOKING_VERTEX_EXTENSION_NAME, false); test(has_ext_provoking_vertex, VK_EXT_PROVOKING_VERTEX_EXTENSION_NAME, false);
test(has_ext_vertex_input_dynamic_state, VK_EXT_VERTEX_INPUT_DYNAMIC_STATE_EXTENSION_NAME, test(has_ext_vertex_input_dynamic_state, VK_EXT_VERTEX_INPUT_DYNAMIC_STATE_EXTENSION_NAME,
@ -1284,6 +1318,36 @@ std::vector<const char*> Device::LoadExtensions(bool requires_surface) {
ext_extended_dynamic_state = true; ext_extended_dynamic_state = true;
} }
} }
if (has_ext_extended_dynamic_state2) {
VkPhysicalDeviceExtendedDynamicState2FeaturesEXT extended_dynamic_state2;
extended_dynamic_state2.sType =
VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_EXTENDED_DYNAMIC_STATE_2_FEATURES_EXT;
extended_dynamic_state2.pNext = nullptr;
features.pNext = &extended_dynamic_state2;
physical.GetFeatures2(features);
if (extended_dynamic_state2.extendedDynamicState2) {
extensions.push_back(VK_EXT_EXTENDED_DYNAMIC_STATE_2_EXTENSION_NAME);
ext_extended_dynamic_state2 = true;
ext_extended_dynamic_state2_extra =
extended_dynamic_state2.extendedDynamicState2LogicOp &&
extended_dynamic_state2.extendedDynamicState2PatchControlPoints;
}
}
if (has_ext_extended_dynamic_state3) {
VkPhysicalDeviceExtendedDynamicState3FeaturesEXT extended_dynamic_state3;
extended_dynamic_state3.sType =
VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_EXTENDED_DYNAMIC_STATE_3_FEATURES_EXT;
extended_dynamic_state3.pNext = nullptr;
features.pNext = &extended_dynamic_state3;
physical.GetFeatures2(features);
if (extended_dynamic_state3.extendedDynamicState3ColorBlendEnable &&
extended_dynamic_state3.extendedDynamicState3ColorBlendEquation) {
extensions.push_back(VK_EXT_EXTENDED_DYNAMIC_STATE_3_EXTENSION_NAME);
ext_extended_dynamic_state3 = true;
}
}
if (has_ext_line_rasterization) { if (has_ext_line_rasterization) {
VkPhysicalDeviceLineRasterizationFeaturesEXT line_raster; VkPhysicalDeviceLineRasterizationFeaturesEXT line_raster;
line_raster.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_LINE_RASTERIZATION_FEATURES_EXT; line_raster.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_LINE_RASTERIZATION_FEATURES_EXT;

View file

@ -286,6 +286,20 @@ public:
return ext_extended_dynamic_state; return ext_extended_dynamic_state;
} }
/// Returns true if the device supports VK_EXT_extended_dynamic_state2.
bool IsExtExtendedDynamicState2Supported() const {
return ext_extended_dynamic_state2;
}
bool IsExtExtendedDynamicState2ExtrasSupported() const {
return ext_extended_dynamic_state2_extra;
}
/// Returns true if the device supports VK_EXT_extended_dynamic_state3.
bool IsExtExtendedDynamicState3Supported() const {
return ext_extended_dynamic_state3;
}
/// Returns true if the device supports VK_EXT_line_rasterization. /// Returns true if the device supports VK_EXT_line_rasterization.
bool IsExtLineRasterizationSupported() const { bool IsExtLineRasterizationSupported() const {
return ext_line_rasterization; return ext_line_rasterization;
@ -468,6 +482,9 @@ private:
bool ext_transform_feedback{}; ///< Support for VK_EXT_transform_feedback. bool ext_transform_feedback{}; ///< Support for VK_EXT_transform_feedback.
bool ext_custom_border_color{}; ///< Support for VK_EXT_custom_border_color. bool ext_custom_border_color{}; ///< Support for VK_EXT_custom_border_color.
bool ext_extended_dynamic_state{}; ///< Support for VK_EXT_extended_dynamic_state. bool ext_extended_dynamic_state{}; ///< Support for VK_EXT_extended_dynamic_state.
bool ext_extended_dynamic_state2{}; ///< Support for VK_EXT_extended_dynamic_state2.
bool ext_extended_dynamic_state2_extra{}; ///< Support for VK_EXT_extended_dynamic_state2.
bool ext_extended_dynamic_state3{}; ///< Support for VK_EXT_extended_dynamic_state3.
bool ext_line_rasterization{}; ///< Support for VK_EXT_line_rasterization. bool ext_line_rasterization{}; ///< Support for VK_EXT_line_rasterization.
bool ext_vertex_input_dynamic_state{}; ///< Support for VK_EXT_vertex_input_dynamic_state. bool ext_vertex_input_dynamic_state{}; ///< Support for VK_EXT_vertex_input_dynamic_state.
bool ext_shader_stencil_export{}; ///< Support for VK_EXT_shader_stencil_export. bool ext_shader_stencil_export{}; ///< Support for VK_EXT_shader_stencil_export.

View file

@ -122,6 +122,9 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept {
X(vkCmdSetDepthCompareOpEXT); X(vkCmdSetDepthCompareOpEXT);
X(vkCmdSetDepthTestEnableEXT); X(vkCmdSetDepthTestEnableEXT);
X(vkCmdSetDepthWriteEnableEXT); X(vkCmdSetDepthWriteEnableEXT);
X(vkCmdSetPrimitiveRestartEnableEXT);
X(vkCmdSetRasterizerDiscardEnableEXT);
X(vkCmdSetDepthBiasEnableEXT);
X(vkCmdSetFrontFaceEXT); X(vkCmdSetFrontFaceEXT);
X(vkCmdSetLineWidth); X(vkCmdSetLineWidth);
X(vkCmdSetPrimitiveTopologyEXT); X(vkCmdSetPrimitiveTopologyEXT);

View file

@ -234,6 +234,9 @@ struct DeviceDispatch : InstanceDispatch {
PFN_vkCmdSetDepthCompareOpEXT vkCmdSetDepthCompareOpEXT{}; PFN_vkCmdSetDepthCompareOpEXT vkCmdSetDepthCompareOpEXT{};
PFN_vkCmdSetDepthTestEnableEXT vkCmdSetDepthTestEnableEXT{}; PFN_vkCmdSetDepthTestEnableEXT vkCmdSetDepthTestEnableEXT{};
PFN_vkCmdSetDepthWriteEnableEXT vkCmdSetDepthWriteEnableEXT{}; PFN_vkCmdSetDepthWriteEnableEXT vkCmdSetDepthWriteEnableEXT{};
PFN_vkCmdSetPrimitiveRestartEnableEXT vkCmdSetPrimitiveRestartEnableEXT{};
PFN_vkCmdSetRasterizerDiscardEnableEXT vkCmdSetRasterizerDiscardEnableEXT{};
PFN_vkCmdSetDepthBiasEnableEXT vkCmdSetDepthBiasEnableEXT{};
PFN_vkCmdSetEvent vkCmdSetEvent{}; PFN_vkCmdSetEvent vkCmdSetEvent{};
PFN_vkCmdSetFrontFaceEXT vkCmdSetFrontFaceEXT{}; PFN_vkCmdSetFrontFaceEXT vkCmdSetFrontFaceEXT{};
PFN_vkCmdSetLineWidth vkCmdSetLineWidth{}; PFN_vkCmdSetLineWidth vkCmdSetLineWidth{};
@ -1219,6 +1222,18 @@ public:
dld->vkCmdSetDepthWriteEnableEXT(handle, enable ? VK_TRUE : VK_FALSE); dld->vkCmdSetDepthWriteEnableEXT(handle, enable ? VK_TRUE : VK_FALSE);
} }
void SetPrimitiveRestartEnableEXT(bool enable) const noexcept {
dld->vkCmdSetPrimitiveRestartEnableEXT(handle, enable ? VK_TRUE : VK_FALSE);
}
void SetRasterizerDiscardEnableEXT(bool enable) const noexcept {
dld->vkCmdSetRasterizerDiscardEnableEXT(handle, enable ? VK_TRUE : VK_FALSE);
}
void SetDepthBiasEnableEXT(bool enable) const noexcept {
dld->vkCmdSetDepthBiasEnableEXT(handle, enable ? VK_TRUE : VK_FALSE);
}
void SetFrontFaceEXT(VkFrontFace front_face) const noexcept { void SetFrontFaceEXT(VkFrontFace front_face) const noexcept {
dld->vkCmdSetFrontFaceEXT(handle, front_face); dld->vkCmdSetFrontFaceEXT(handle, front_face);
} }