forked from ShuriZma/suyu
1
0
Fork 0

Vulkan: Implement Dynamic State 3

This commit is contained in:
Fernando Sahmkow 2022-12-06 00:40:01 +01:00
parent f800e485c9
commit d33251db93
13 changed files with 315 additions and 107 deletions

View File

@ -124,7 +124,6 @@ void Maxwell3D::InitializeRegisterDefaults() {
regs.gl_front_face = Maxwell3D::Regs::FrontFace::ClockWise;
regs.polygon_mode_back = Maxwell3D::Regs::PolygonMode::Fill;
regs.polygon_mode_front = Maxwell3D::Regs::PolygonMode::Fill;
regs.logic_op.op = Maxwell3D::Regs::LogicOp::Op::Clear;
shadow_state = regs;
}

View File

@ -56,22 +56,16 @@ void FixedPipelineState::Refresh(Tegra::Engines::Maxwell3D& maxwell3d, DynamicFe
extended_dynamic_state.Assign(features.has_extended_dynamic_state ? 1 : 0);
extended_dynamic_state_2.Assign(features.has_extended_dynamic_state_2 ? 1 : 0);
extended_dynamic_state_2_extra.Assign(features.has_extended_dynamic_state_2_extra ? 1 : 0);
extended_dynamic_state_3.Assign(features.has_extended_dynamic_state_3 ? 1 : 0);
extended_dynamic_state_3_blend.Assign(features.has_extended_dynamic_state_3_blend ? 1 : 0);
extended_dynamic_state_3_enables.Assign(features.has_extended_dynamic_state_3_enables ? 1 : 0);
dynamic_vertex_input.Assign(features.has_dynamic_vertex_input ? 1 : 0);
xfb_enabled.Assign(regs.transform_feedback_enabled != 0);
depth_clamp_disabled.Assign(regs.viewport_clip_control.geometry_clip ==
Maxwell::ViewportClipControl::GeometryClip::Passthrough ||
regs.viewport_clip_control.geometry_clip ==
Maxwell::ViewportClipControl::GeometryClip::FrustumXYZ ||
regs.viewport_clip_control.geometry_clip ==
Maxwell::ViewportClipControl::GeometryClip::FrustumZ);
ndc_minus_one_to_one.Assign(regs.depth_mode == Maxwell::DepthMode::MinusOneToOne ? 1 : 0);
polygon_mode.Assign(PackPolygonMode(regs.polygon_mode_front));
tessellation_primitive.Assign(static_cast<u32>(regs.tessellation.params.domain_type.Value()));
tessellation_spacing.Assign(static_cast<u32>(regs.tessellation.params.spacing.Value()));
tessellation_clockwise.Assign(regs.tessellation.params.output_primitives.Value() ==
Maxwell::Tessellation::OutputPrimitives::Triangles_CW);
logic_op_enable.Assign(regs.logic_op.enable != 0 ? 1 : 0);
patch_control_points_minus_one.Assign(regs.patch_vertices - 1);
topology.Assign(topology_);
msaa_mode.Assign(regs.anti_alias_samples_mode);
@ -138,12 +132,6 @@ void FixedPipelineState::Refresh(Tegra::Engines::Maxwell3D& maxwell3d, DynamicFe
}
}
}
if (maxwell3d.dirty.flags[Dirty::Blending]) {
maxwell3d.dirty.flags[Dirty::Blending] = false;
for (size_t index = 0; index < attachments.size(); ++index) {
attachments[index].Refresh(regs, index);
}
}
if (maxwell3d.dirty.flags[Dirty::ViewportSwizzles]) {
maxwell3d.dirty.flags[Dirty::ViewportSwizzles] = false;
const auto& transform = regs.viewport_transform;
@ -155,11 +143,22 @@ void FixedPipelineState::Refresh(Tegra::Engines::Maxwell3D& maxwell3d, DynamicFe
dynamic_state.raw2 = 0;
if (!extended_dynamic_state) {
dynamic_state.Refresh(regs);
std::ranges::transform(regs.vertex_streams, vertex_strides.begin(), [](const auto& array) {
return static_cast<u16>(array.stride.Value());
});
}
if (!extended_dynamic_state_2_extra) {
dynamic_state.Refresh2(regs, topology, extended_dynamic_state_2);
}
if (!extended_dynamic_state_3) {
if (!extended_dynamic_state_3_blend) {
if (maxwell3d.dirty.flags[Dirty::Blending]) {
maxwell3d.dirty.flags[Dirty::Blending] = false;
for (size_t index = 0; index < attachments.size(); ++index) {
attachments[index].Refresh(regs, index);
}
}
}
if (!extended_dynamic_state_3_enables) {
dynamic_state.Refresh3(regs);
}
if (xfb_enabled) {
@ -177,12 +176,11 @@ void FixedPipelineState::BlendingAttachment::Refresh(const Maxwell& regs, size_t
mask_a.Assign(mask.A);
// TODO: C++20 Use templated lambda to deduplicate code
if (!regs.blend.enable[index]) {
return;
}
if (!regs.blend_per_target_enabled) {
if (!regs.blend.enable[index]) {
return;
}
const auto& src = regs.blend;
const auto setup_blend = [&]<typename T>(const T& src) {
equation_rgb.Assign(PackBlendEquation(src.color_op));
equation_a.Assign(PackBlendEquation(src.alpha_op));
factor_source_rgb.Assign(PackBlendFactor(src.color_source));
@ -190,20 +188,13 @@ void FixedPipelineState::BlendingAttachment::Refresh(const Maxwell& regs, size_t
factor_source_a.Assign(PackBlendFactor(src.alpha_source));
factor_dest_a.Assign(PackBlendFactor(src.alpha_dest));
enable.Assign(1);
return;
}
};
if (!regs.blend.enable[index]) {
if (!regs.blend_per_target_enabled) {
setup_blend(regs.blend);
return;
}
const auto& src = regs.blend_per_target[index];
equation_rgb.Assign(PackBlendEquation(src.color_op));
equation_a.Assign(PackBlendEquation(src.alpha_op));
factor_source_rgb.Assign(PackBlendFactor(src.color_source));
factor_dest_rgb.Assign(PackBlendFactor(src.color_dest));
factor_source_a.Assign(PackBlendFactor(src.alpha_source));
factor_dest_a.Assign(PackBlendFactor(src.alpha_dest));
enable.Assign(1);
setup_blend(regs.blend_per_target[index]);
}
void FixedPipelineState::DynamicState::Refresh(const Maxwell& regs) {
@ -236,12 +227,11 @@ void FixedPipelineState::DynamicState::Refresh(const Maxwell& regs) {
depth_test_func.Assign(PackComparisonOp(regs.depth_test_func));
cull_face.Assign(PackCullFace(regs.gl_cull_face));
cull_enable.Assign(regs.gl_cull_test_enabled != 0 ? 1 : 0);
std::ranges::transform(regs.vertex_streams, vertex_strides.begin(), [](const auto& array) {
return static_cast<u16>(array.stride.Value());
});
}
void FixedPipelineState::DynamicState::Refresh2(const Maxwell& regs, Maxwell::PrimitiveTopology topology_, bool base_feautures_supported) {
void FixedPipelineState::DynamicState::Refresh2(const Maxwell& regs,
Maxwell::PrimitiveTopology topology_,
bool base_feautures_supported) {
logic_op.Assign(PackLogicOp(regs.logic_op.op));
if (base_feautures_supported) {
@ -260,7 +250,15 @@ void FixedPipelineState::DynamicState::Refresh2(const Maxwell& regs, Maxwell::Pr
depth_bias_enable.Assign(enabled_lut[POLYGON_OFFSET_ENABLE_LUT[topology_index]] != 0 ? 1 : 0);
}
void FixedPipelineState::DynamicState::Refresh3(const Maxwell&) {}
void FixedPipelineState::DynamicState::Refresh3(const Maxwell& regs) {
logic_op_enable.Assign(regs.logic_op.enable != 0 ? 1 : 0);
depth_clamp_disabled.Assign(regs.viewport_clip_control.geometry_clip ==
Maxwell::ViewportClipControl::GeometryClip::Passthrough ||
regs.viewport_clip_control.geometry_clip ==
Maxwell::ViewportClipControl::GeometryClip::FrustumXYZ ||
regs.viewport_clip_control.geometry_clip ==
Maxwell::ViewportClipControl::GeometryClip::FrustumZ);
}
size_t FixedPipelineState::Hash() const noexcept {
const u64 hash = Common::CityHash64(reinterpret_cast<const char*>(this), Size());

View File

@ -21,7 +21,8 @@ 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_extended_dynamic_state_3_blend;
bool has_extended_dynamic_state_3_enables;
bool has_dynamic_vertex_input;
};
@ -147,6 +148,8 @@ struct FixedPipelineState {
BitField<4, 1, u32> depth_bias_enable;
BitField<5, 1, u32> rasterize_enable;
BitField<6, 4, u32> logic_op;
BitField<10, 1, u32> logic_op_enable;
BitField<11, 1, u32> depth_clamp_disabled;
};
union {
u32 raw2;
@ -159,8 +162,6 @@ struct FixedPipelineState {
BitField<28, 1, u32> front_face;
BitField<29, 3, u32> depth_test_func;
};
// Vertex stride is a 12 bits value, we have 4 bits to spare per element
std::array<u16, Maxwell::NumVertexArrays> vertex_strides;
void Refresh(const Maxwell& regs);
void Refresh2(const Maxwell& regs, Maxwell::PrimitiveTopology topology, bool base_feautures_supported);
@ -184,17 +185,16 @@ struct FixedPipelineState {
BitField<0, 1, u32> extended_dynamic_state;
BitField<1, 1, u32> extended_dynamic_state_2;
BitField<2, 1, u32> extended_dynamic_state_2_extra;
BitField<3, 1, u32> extended_dynamic_state_3;
BitField<4, 1, u32> dynamic_vertex_input;
BitField<5, 1, u32> xfb_enabled;
BitField<6, 1, u32> depth_clamp_disabled;
BitField<3, 1, u32> extended_dynamic_state_3_blend;
BitField<4, 1, u32> extended_dynamic_state_3_enables;
BitField<5, 1, u32> dynamic_vertex_input;
BitField<6, 1, u32> xfb_enabled;
BitField<7, 1, u32> ndc_minus_one_to_one;
BitField<8, 2, u32> polygon_mode;
BitField<10, 2, u32> tessellation_primitive;
BitField<12, 2, u32> tessellation_spacing;
BitField<14, 1, u32> tessellation_clockwise;
BitField<15, 1, u32> logic_op_enable;
BitField<16, 5, u32> patch_control_points_minus_one;
BitField<15, 5, u32> patch_control_points_minus_one;
BitField<24, 4, Maxwell::PrimitiveTopology> topology;
BitField<28, 4, Tegra::Texture::MsaaMode> msaa_mode;
@ -217,16 +217,19 @@ struct FixedPipelineState {
u32 alpha_test_ref;
u32 point_size;
std::array<BlendingAttachment, Maxwell::NumRenderTargets> attachments;
std::array<u16, Maxwell::NumViewports> viewport_swizzles;
union {
u64 attribute_types; // Used with VK_EXT_vertex_input_dynamic_state
u64 enabled_divisors;
};
std::array<VertexAttribute, Maxwell::NumVertexAttributes> attributes;
std::array<u32, Maxwell::NumVertexArrays> binding_divisors;
DynamicState dynamic_state;
std::array<BlendingAttachment, Maxwell::NumRenderTargets> attachments;
std::array<VertexAttribute, Maxwell::NumVertexAttributes> attributes;
std::array<u32, Maxwell::NumVertexArrays> binding_divisors;
// Vertex stride is a 12 bits value, we have 4 bits to spare per element
std::array<u16, Maxwell::NumVertexArrays> vertex_strides;
VideoCommon::TransformFeedbackState xfb_state;
void Refresh(Tegra::Engines::Maxwell3D& maxwell3d, DynamicFeatures& features);
@ -244,17 +247,17 @@ struct FixedPipelineState {
// When transform feedback is enabled, use the whole struct
return sizeof(*this);
}
if (dynamic_vertex_input && extended_dynamic_state_2) {
if (dynamic_vertex_input && extended_dynamic_state_3_blend) {
// Exclude dynamic state and attributes
return offsetof(FixedPipelineState, attributes);
}
if (extended_dynamic_state_2_extra) {
// Exclude dynamic state
return offsetof(FixedPipelineState, dynamic_state);
}
if (dynamic_vertex_input) {
// Exclude dynamic state
return offsetof(FixedPipelineState, attributes);
}
if (extended_dynamic_state) {
// Exclude dynamic state
return offsetof(FixedPipelineState, dynamic_state.raw2);
return offsetof(FixedPipelineState, vertex_strides);
}
// Default
return offsetof(FixedPipelineState, xfb_state);

View File

@ -489,9 +489,11 @@ void GraphicsPipeline::ConfigureDraw(const RescalingPushConstant& rescaling,
if (bind_pipeline) {
cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, *pipeline);
}
cmdbuf.PushConstants(*pipeline_layout, VK_SHADER_STAGE_ALL_GRAPHICS,
RESCALING_LAYOUT_WORDS_OFFSET, sizeof(rescaling_data),
rescaling_data.data());
if (is_rescaling) {
cmdbuf.PushConstants(*pipeline_layout, VK_SHADER_STAGE_ALL_GRAPHICS,
RESCALING_LAYOUT_WORDS_OFFSET, sizeof(rescaling_data),
rescaling_data.data());
}
if (update_rescaling) {
const f32 config_down_factor{Settings::values.resolution_info.down_factor};
const f32 scale_down_factor{is_rescaling ? config_down_factor : 1.0f};
@ -524,9 +526,8 @@ void GraphicsPipeline::MakePipeline(VkRenderPass render_pass) {
FixedPipelineState::DynamicState dynamic{};
if (!key.state.extended_dynamic_state) {
dynamic = key.state.dynamic_state;
}
if (!key.state.extended_dynamic_state_2) {
dynamic.raw2 = key.state.dynamic_state.raw2;
} else {
dynamic.raw1 = key.state.dynamic_state.raw1;
}
static_vector<VkVertexInputBindingDescription, 32> vertex_bindings;
static_vector<VkVertexInputBindingDivisorDescriptionEXT, 32> vertex_binding_divisors;
@ -564,7 +565,7 @@ void GraphicsPipeline::MakePipeline(VkRenderPass render_pass) {
instanced ? VK_VERTEX_INPUT_RATE_INSTANCE : VK_VERTEX_INPUT_RATE_VERTEX;
vertex_bindings.push_back({
.binding = static_cast<u32>(index),
.stride = dynamic.vertex_strides[index],
.stride = key.state.vertex_strides[index],
.inputRate = rate,
});
if (instanced) {
@ -675,7 +676,7 @@ void GraphicsPipeline::MakePipeline(VkRenderPass render_pass) {
.pNext = nullptr,
.flags = 0,
.depthClampEnable =
static_cast<VkBool32>(key.state.depth_clamp_disabled == 0 ? VK_TRUE : VK_FALSE),
static_cast<VkBool32>(dynamic.depth_clamp_disabled == 0 ? VK_TRUE : VK_FALSE),
.rasterizerDiscardEnable =
static_cast<VkBool32>(dynamic.rasterize_enable == 0 ? VK_TRUE : VK_FALSE),
.polygonMode =
@ -785,13 +786,13 @@ void GraphicsPipeline::MakePipeline(VkRenderPass render_pass) {
.sType = VK_STRUCTURE_TYPE_PIPELINE_COLOR_BLEND_STATE_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.logicOpEnable = key.state.logic_op_enable != 0,
.logicOpEnable = dynamic.logic_op_enable != 0,
.logicOp = static_cast<VkLogicOp>(dynamic.logic_op.Value()),
.attachmentCount = static_cast<u32>(cb_attachments.size()),
.pAttachments = cb_attachments.data(),
.blendConstants = {},
};
static_vector<VkDynamicState, 23> dynamic_states{
static_vector<VkDynamicState, 28> dynamic_states{
VK_DYNAMIC_STATE_VIEWPORT, VK_DYNAMIC_STATE_SCISSOR,
VK_DYNAMIC_STATE_DEPTH_BIAS, VK_DYNAMIC_STATE_BLEND_CONSTANTS,
VK_DYNAMIC_STATE_DEPTH_BOUNDS, VK_DYNAMIC_STATE_STENCIL_COMPARE_MASK,
@ -825,6 +826,21 @@ void GraphicsPipeline::MakePipeline(VkRenderPass render_pass) {
if (key.state.extended_dynamic_state_2_extra) {
dynamic_states.push_back(VK_DYNAMIC_STATE_LOGIC_OP_EXT);
}
if (key.state.extended_dynamic_state_3_blend) {
static constexpr std::array extended3{
VK_DYNAMIC_STATE_COLOR_BLEND_ENABLE_EXT,
VK_DYNAMIC_STATE_COLOR_BLEND_EQUATION_EXT,
VK_DYNAMIC_STATE_COLOR_WRITE_MASK_EXT,
};
dynamic_states.insert(dynamic_states.end(), extended3.begin(), extended3.end());
}
if (key.state.extended_dynamic_state_3_enables) {
static constexpr std::array extended3{
VK_DYNAMIC_STATE_DEPTH_CLAMP_ENABLE_EXT,
VK_DYNAMIC_STATE_LOGIC_OP_ENABLE_EXT,
};
dynamic_states.insert(dynamic_states.end(), extended3.begin(), extended3.end());
}
}
const VkPipelineDynamicStateCreateInfo dynamic_state_ci{
.sType = VK_STRUCTURE_TYPE_PIPELINE_DYNAMIC_STATE_CREATE_INFO,

View File

@ -356,7 +356,8 @@ PipelineCache::PipelineCache(RasterizerVulkan& rasterizer_, const Device& device
.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_extended_dynamic_state_3_blend = device.IsExtExtendedDynamicState3BlendingSupported(),
.has_extended_dynamic_state_3_enables = device.IsExtExtendedDynamicState3EnablesSupported(),
.has_dynamic_vertex_input = device.IsExtVertexInputDynamicStateSupported(),
};
}
@ -456,8 +457,10 @@ void PipelineCache::LoadDiskResources(u64 title_id, std::stop_token stop_loading
dynamic_features.has_extended_dynamic_state_2 ||
(key.state.extended_dynamic_state_2_extra != 0) !=
dynamic_features.has_extended_dynamic_state_2_extra ||
(key.state.extended_dynamic_state_3 != 0) !=
dynamic_features.has_extended_dynamic_state_3 ||
(key.state.extended_dynamic_state_3_blend != 0) !=
dynamic_features.has_extended_dynamic_state_3_blend ||
(key.state.extended_dynamic_state_3_enables != 0) !=
dynamic_features.has_extended_dynamic_state_3_enables ||
(key.state.dynamic_vertex_input != 0) != dynamic_features.has_dynamic_vertex_input) {
return;
}

View File

@ -698,10 +698,17 @@ void RasterizerVulkan::UpdateDynamicStates() {
UpdateRasterizerDiscardEnable(regs);
UpdateDepthBiasEnable(regs);
}
if (device.IsExtExtendedDynamicState3EnablesSupported()) {
UpdateLogicOpEnable(regs);
UpdateDepthClampEnable(regs);
}
}
if (device.IsExtExtendedDynamicState2ExtrasSupported()) {
UpdateLogicOp(regs);
}
if (device.IsExtExtendedDynamicState3Supported()) {
UpdateBlending(regs);
}
}
}
@ -970,7 +977,30 @@ void RasterizerVulkan::UpdateDepthBiasEnable(Tegra::Engines::Maxwell3D::Regs& re
};
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); });
scheduler.Record([enable](vk::CommandBuffer cmdbuf) { cmdbuf.SetDepthBiasEnableEXT(enable != 0); });
}
void RasterizerVulkan::UpdateLogicOpEnable(Tegra::Engines::Maxwell3D::Regs& regs) {
if (!state_tracker.TouchLogicOpEnable()) {
return;
}
scheduler.Record([enable = regs.logic_op.enable](vk::CommandBuffer cmdbuf) {
cmdbuf.SetLogicOpEnableEXT(enable != 0);
});
}
void RasterizerVulkan::UpdateDepthClampEnable(Tegra::Engines::Maxwell3D::Regs& regs) {
if (!state_tracker.TouchDepthClampEnable()) {
return;
}
bool is_enabled = !(regs.viewport_clip_control.geometry_clip ==
Maxwell::ViewportClipControl::GeometryClip::Passthrough ||
regs.viewport_clip_control.geometry_clip ==
Maxwell::ViewportClipControl::GeometryClip::FrustumXYZ ||
regs.viewport_clip_control.geometry_clip ==
Maxwell::ViewportClipControl::GeometryClip::FrustumZ);
scheduler.Record(
[is_enabled](vk::CommandBuffer cmdbuf) { cmdbuf.SetDepthClampEnableEXT(is_enabled); });
}
void RasterizerVulkan::UpdateDepthCompareOp(Tegra::Engines::Maxwell3D::Regs& regs) {
@ -1041,6 +1071,68 @@ void RasterizerVulkan::UpdateLogicOp(Tegra::Engines::Maxwell3D::Regs& regs) {
scheduler.Record([op](vk::CommandBuffer cmdbuf) { cmdbuf.SetLogicOpEXT(op); });
}
void RasterizerVulkan::UpdateBlending(Tegra::Engines::Maxwell3D::Regs& regs) {
if (!state_tracker.TouchBlending()) {
return;
}
if (state_tracker.TouchColorMask()) {
std::array<VkColorComponentFlags, Maxwell::NumRenderTargets> setup_masks{};
for (size_t index = 0; index < Maxwell::NumRenderTargets; index++) {
const auto& mask = regs.color_mask[regs.color_mask_common ? 0 : index];
auto& current = setup_masks[index];
if (mask.R) {
current |= VK_COLOR_COMPONENT_R_BIT;
}
if (mask.G) {
current |= VK_COLOR_COMPONENT_G_BIT;
}
if (mask.B) {
current |= VK_COLOR_COMPONENT_B_BIT;
}
if (mask.A) {
current |= VK_COLOR_COMPONENT_A_BIT;
}
}
scheduler.Record([setup_masks](vk::CommandBuffer cmdbuf) {
cmdbuf.SetColorWriteMaskEXT(0, setup_masks);
});
}
if (state_tracker.TouchBlendEnable()) {
std::array<VkBool32, Maxwell::NumRenderTargets> setup_enables{};
std::ranges::transform(
regs.blend.enable, setup_enables.begin(),
[&](const auto& is_enabled) { return is_enabled != 0 ? VK_TRUE : VK_FALSE; });
scheduler.Record([setup_enables](vk::CommandBuffer cmdbuf) {
cmdbuf.SetColorBlendEnableEXT(0, setup_enables);
});
}
if (state_tracker.TouchBlendEquations()) {
std::array<VkColorBlendEquationEXT, Maxwell::NumRenderTargets> setup_blends{};
for (size_t index = 0; index < Maxwell::NumRenderTargets; index++) {
const auto blend_setup = [&]<typename T>(const T& guest_blend) {
auto& host_blend = setup_blends[index];
host_blend.srcColorBlendFactor = MaxwellToVK::BlendFactor(guest_blend.color_source);
host_blend.dstColorBlendFactor = MaxwellToVK::BlendFactor(guest_blend.color_dest);
host_blend.colorBlendOp = MaxwellToVK::BlendEquation(guest_blend.color_op);
host_blend.srcAlphaBlendFactor = MaxwellToVK::BlendFactor(guest_blend.alpha_source);
host_blend.dstAlphaBlendFactor = MaxwellToVK::BlendFactor(guest_blend.alpha_dest);
host_blend.alphaBlendOp = MaxwellToVK::BlendEquation(guest_blend.alpha_op);
};
if (!regs.blend_per_target_enabled) {
blend_setup(regs.blend);
continue;
}
blend_setup(regs.blend_per_target[index]);
}
scheduler.Record([setup_blends](vk::CommandBuffer cmdbuf) {
cmdbuf.SetColorBlendEquationEXT(0, setup_blends);
});
}
}
void RasterizerVulkan::UpdateStencilTestEnable(Tegra::Engines::Maxwell3D::Regs& regs) {
if (!state_tracker.TouchStencilTestEnable()) {
return;

View File

@ -142,10 +142,13 @@ private:
void UpdatePrimitiveRestartEnable(Tegra::Engines::Maxwell3D::Regs& regs);
void UpdateRasterizerDiscardEnable(Tegra::Engines::Maxwell3D::Regs& regs);
void UpdateDepthBiasEnable(Tegra::Engines::Maxwell3D::Regs& regs);
void UpdateLogicOpEnable(Tegra::Engines::Maxwell3D::Regs& regs);
void UpdateDepthClampEnable(Tegra::Engines::Maxwell3D::Regs& regs);
void UpdateFrontFace(Tegra::Engines::Maxwell3D::Regs& regs);
void UpdateStencilOp(Tegra::Engines::Maxwell3D::Regs& regs);
void UpdateStencilTestEnable(Tegra::Engines::Maxwell3D::Regs& regs);
void UpdateLogicOp(Tegra::Engines::Maxwell3D::Regs& regs);
void UpdateBlending(Tegra::Engines::Maxwell3D::Regs& regs);
void UpdateVertexInput(Tegra::Engines::Maxwell3D::Regs& regs);

View File

@ -48,7 +48,13 @@ Flags MakeInvalidationFlags() {
PrimitiveRestartEnable,
RasterizerDiscardEnable,
DepthBiasEnable,
LogicOpEnable,
DepthClampEnable,
LogicOp,
Blending,
ColorMask,
BlendEquations,
BlendEnable,
};
Flags flags{};
for (const int flag : INVALIDATION_FLAGS) {
@ -128,6 +134,8 @@ void SetupDirtyStateEnable(Tables& tables) {
setup(OFF(polygon_offset_point_enable), DepthBiasEnable);
setup(OFF(polygon_offset_line_enable), DepthBiasEnable);
setup(OFF(polygon_offset_fill_enable), DepthBiasEnable);
setup(OFF(logic_op.enable), LogicOpEnable);
setup(OFF(viewport_clip_control.geometry_clip), DepthClampEnable);
}
void SetupDirtyDepthCompareOp(Tables& tables) {
@ -157,10 +165,16 @@ void SetupDirtyStencilOp(Tables& tables) {
void SetupDirtyBlending(Tables& tables) {
tables[0][OFF(color_mask_common)] = Blending;
tables[1][OFF(color_mask_common)] = ColorMask;
tables[0][OFF(blend_per_target_enabled)] = Blending;
tables[1][OFF(blend_per_target_enabled)] = BlendEquations;
FillBlock(tables[0], OFF(color_mask), NUM(color_mask), Blending);
FillBlock(tables[1], OFF(color_mask), NUM(color_mask), ColorMask);
FillBlock(tables[0], OFF(blend), NUM(blend), Blending);
FillBlock(tables[1], OFF(blend), NUM(blend), BlendEquations);
FillBlock(tables[1], OFF(blend.enable), NUM(blend.enable), BlendEnable);
FillBlock(tables[0], OFF(blend_per_target), NUM(blend_per_target), Blending);
FillBlock(tables[1], OFF(blend_per_target), NUM(blend_per_target), BlendEquations);
}
void SetupDirtySpecialOps(Tables& tables) {

View File

@ -50,8 +50,13 @@ enum : u8 {
DepthBiasEnable,
StateEnable,
LogicOp,
LogicOpEnable,
DepthClampEnable,
Blending,
BlendEnable,
BlendEquations,
ColorMask,
ViewportSwizzles,
Last,
@ -144,6 +149,14 @@ public:
return Exchange(Dirty::DepthBiasEnable, false);
}
bool TouchLogicOpEnable() {
return Exchange(Dirty::LogicOpEnable, false);
}
bool TouchDepthClampEnable() {
return Exchange(Dirty::DepthClampEnable, false);
}
bool TouchDepthCompareOp() {
return Exchange(Dirty::DepthCompareOp, false);
}
@ -156,6 +169,22 @@ public:
return Exchange(Dirty::StencilOp, false);
}
bool TouchBlending() {
return Exchange(Dirty::Blending, false);
}
bool TouchBlendEnable() {
return Exchange(Dirty::BlendEnable, false);
}
bool TouchBlendEquations() {
return Exchange(Dirty::BlendEquations, false);
}
bool TouchColorMask() {
return Exchange(Dirty::ColorMask, false);
}
bool TouchStencilTestEnable() {
return Exchange(Dirty::StencilTestEnable, false);
}

View File

@ -569,28 +569,31 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR
LOG_INFO(Render_Vulkan, "Device doesn't support extended dynamic state");
}
VkPhysicalDeviceExtendedDynamicState2FeaturesEXT dynamic_state2;
if (ext_extended_dynamic_state2) {
dynamic_state2 = {
VkPhysicalDeviceExtendedDynamicState2FeaturesEXT dynamic_state_2;
if (ext_extended_dynamic_state_2) {
dynamic_state_2 = {
.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,
.extendedDynamicState2LogicOp = ext_extended_dynamic_state_2_extra ? VK_TRUE : VK_FALSE,
};
SetNext(next, dynamic_state2);
SetNext(next, dynamic_state_2);
} else {
LOG_INFO(Render_Vulkan, "Device doesn't support extended dynamic state 2");
}
VkPhysicalDeviceExtendedDynamicState3FeaturesEXT dynamic_state3;
if (ext_extended_dynamic_state3) {
dynamic_state3 = {
VkPhysicalDeviceExtendedDynamicState3FeaturesEXT dynamic_state_3;
if (ext_extended_dynamic_state_3) {
dynamic_state_3 = {
.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_EXTENDED_DYNAMIC_STATE_3_FEATURES_EXT,
.pNext = nullptr,
.extendedDynamicState3ColorBlendEnable = VK_TRUE,
.extendedDynamicState3ColorBlendEquation = VK_TRUE,
.extendedDynamicState3DepthClampEnable = ext_extended_dynamic_state_3_enables ? VK_TRUE : VK_FALSE,
.extendedDynamicState3LogicOpEnable = ext_extended_dynamic_state_3_enables ? VK_TRUE : VK_FALSE,
.extendedDynamicState3ColorBlendEnable = ext_extended_dynamic_state_3_blend ? VK_TRUE : VK_FALSE,
.extendedDynamicState3ColorBlendEquation = ext_extended_dynamic_state_3_blend ? VK_TRUE : VK_FALSE,
.extendedDynamicState3ColorWriteMask = ext_extended_dynamic_state_3_blend ? VK_TRUE : VK_FALSE,
};
SetNext(next, dynamic_state3);
SetNext(next, dynamic_state_3);
} else {
LOG_INFO(Render_Vulkan, "Device doesn't support extended dynamic state 3");
}
@ -1117,8 +1120,8 @@ std::vector<const char*> Device::LoadExtensions(bool requires_surface) {
bool has_ext_transform_feedback{};
bool has_ext_custom_border_color{};
bool has_ext_extended_dynamic_state{};
bool has_ext_extended_dynamic_state2{};
bool has_ext_extended_dynamic_state3{};
bool has_ext_extended_dynamic_state_2{};
bool has_ext_extended_dynamic_state_3{};
bool has_ext_shader_atomic_int64{};
bool has_ext_provoking_vertex{};
bool has_ext_vertex_input_dynamic_state{};
@ -1163,9 +1166,9 @@ std::vector<const char*> Device::LoadExtensions(bool requires_surface) {
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_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,
test(has_ext_extended_dynamic_state_2, VK_EXT_EXTENDED_DYNAMIC_STATE_2_EXTENSION_NAME,
false);
test(has_ext_extended_dynamic_state3, VK_EXT_EXTENDED_DYNAMIC_STATE_3_EXTENSION_NAME,
test(has_ext_extended_dynamic_state_3, 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_provoking_vertex, VK_EXT_PROVOKING_VERTEX_EXTENSION_NAME, false);
@ -1316,33 +1319,39 @@ std::vector<const char*> Device::LoadExtensions(bool requires_surface) {
ext_extended_dynamic_state = true;
}
}
if (has_ext_extended_dynamic_state2) {
VkPhysicalDeviceExtendedDynamicState2FeaturesEXT extended_dynamic_state2;
extended_dynamic_state2.sType =
if (has_ext_extended_dynamic_state_2) {
VkPhysicalDeviceExtendedDynamicState2FeaturesEXT extended_dynamic_state_2;
extended_dynamic_state_2.sType =
VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_EXTENDED_DYNAMIC_STATE_2_FEATURES_EXT;
extended_dynamic_state2.pNext = nullptr;
features.pNext = &extended_dynamic_state2;
extended_dynamic_state_2.pNext = nullptr;
features.pNext = &extended_dynamic_state_2;
physical.GetFeatures2(features);
if (extended_dynamic_state2.extendedDynamicState2) {
if (extended_dynamic_state_2.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;
ext_extended_dynamic_state_2 = true;
ext_extended_dynamic_state_2_extra =
extended_dynamic_state_2.extendedDynamicState2LogicOp;
}
}
if (has_ext_extended_dynamic_state3) {
VkPhysicalDeviceExtendedDynamicState3FeaturesEXT extended_dynamic_state3;
extended_dynamic_state3.sType =
if (has_ext_extended_dynamic_state_3) {
VkPhysicalDeviceExtendedDynamicState3FeaturesEXT extended_dynamic_state_3;
extended_dynamic_state_3.sType =
VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_EXTENDED_DYNAMIC_STATE_3_FEATURES_EXT;
extended_dynamic_state3.pNext = nullptr;
features.pNext = &extended_dynamic_state3;
extended_dynamic_state_3.pNext = nullptr;
features.pNext = &extended_dynamic_state_3;
physical.GetFeatures2(features);
if (extended_dynamic_state3.extendedDynamicState3ColorBlendEnable &&
extended_dynamic_state3.extendedDynamicState3ColorBlendEquation) {
ext_extended_dynamic_state_3_blend = extended_dynamic_state_3.extendedDynamicState3ColorBlendEnable &&
extended_dynamic_state_3.extendedDynamicState3ColorBlendEquation &&
extended_dynamic_state_3.extendedDynamicState3ColorWriteMask;
ext_extended_dynamic_state_3_enables = extended_dynamic_state_3.extendedDynamicState3DepthClampEnable &&
extended_dynamic_state_3.extendedDynamicState3LogicOpEnable;
ext_extended_dynamic_state_3 = ext_extended_dynamic_state_3_blend || ext_extended_dynamic_state_3_enables;
if (ext_extended_dynamic_state_3) {
extensions.push_back(VK_EXT_EXTENDED_DYNAMIC_STATE_3_EXTENSION_NAME);
ext_extended_dynamic_state3 = true;
}
}
if (has_ext_line_rasterization) {

View File

@ -288,16 +288,26 @@ public:
/// Returns true if the device supports VK_EXT_extended_dynamic_state2.
bool IsExtExtendedDynamicState2Supported() const {
return ext_extended_dynamic_state2;
return ext_extended_dynamic_state_2;
}
bool IsExtExtendedDynamicState2ExtrasSupported() const {
return ext_extended_dynamic_state2_extra;
return ext_extended_dynamic_state_2_extra;
}
/// Returns true if the device supports VK_EXT_extended_dynamic_state3.
bool IsExtExtendedDynamicState3Supported() const {
return ext_extended_dynamic_state3;
return ext_extended_dynamic_state_3;
}
/// Returns true if the device supports VK_EXT_extended_dynamic_state3.
bool IsExtExtendedDynamicState3BlendingSupported() const {
return ext_extended_dynamic_state_3_blend;
}
/// Returns true if the device supports VK_EXT_extended_dynamic_state3.
bool IsExtExtendedDynamicState3EnablesSupported() const {
return ext_extended_dynamic_state_3_enables;
}
/// Returns true if the device supports VK_EXT_line_rasterization.
@ -482,9 +492,11 @@ private:
bool ext_transform_feedback{}; ///< Support for VK_EXT_transform_feedback.
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_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_extended_dynamic_state_2{}; ///< Support for VK_EXT_extended_dynamic_state2.
bool ext_extended_dynamic_state_2_extra{}; ///< Support for VK_EXT_extended_dynamic_state2.
bool ext_extended_dynamic_state_3{}; ///< Support for VK_EXT_extended_dynamic_state3.
bool ext_extended_dynamic_state_3_blend{}; ///< Support for VK_EXT_extended_dynamic_state3.
bool ext_extended_dynamic_state_3_enables{}; ///< Support for VK_EXT_extended_dynamic_state3.
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_shader_stencil_export{}; ///< Support for VK_EXT_shader_stencil_export.

View File

@ -125,6 +125,8 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept {
X(vkCmdSetPrimitiveRestartEnableEXT);
X(vkCmdSetRasterizerDiscardEnableEXT);
X(vkCmdSetDepthBiasEnableEXT);
X(vkCmdSetLogicOpEnableEXT);
X(vkCmdSetDepthClampEnableEXT);
X(vkCmdSetFrontFaceEXT);
X(vkCmdSetLogicOpEXT);
X(vkCmdSetPatchControlPointsEXT);
@ -133,6 +135,9 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept {
X(vkCmdSetStencilOpEXT);
X(vkCmdSetStencilTestEnableEXT);
X(vkCmdSetVertexInputEXT);
X(vkCmdSetColorWriteMaskEXT);
X(vkCmdSetColorBlendEnableEXT);
X(vkCmdSetColorBlendEquationEXT);
X(vkCmdResolveImage);
X(vkCreateBuffer);
X(vkCreateBufferView);

View File

@ -237,6 +237,8 @@ struct DeviceDispatch : InstanceDispatch {
PFN_vkCmdSetPrimitiveRestartEnableEXT vkCmdSetPrimitiveRestartEnableEXT{};
PFN_vkCmdSetRasterizerDiscardEnableEXT vkCmdSetRasterizerDiscardEnableEXT{};
PFN_vkCmdSetDepthBiasEnableEXT vkCmdSetDepthBiasEnableEXT{};
PFN_vkCmdSetLogicOpEnableEXT vkCmdSetLogicOpEnableEXT{};
PFN_vkCmdSetDepthClampEnableEXT vkCmdSetDepthClampEnableEXT{};
PFN_vkCmdSetEvent vkCmdSetEvent{};
PFN_vkCmdSetFrontFaceEXT vkCmdSetFrontFaceEXT{};
PFN_vkCmdSetPatchControlPointsEXT vkCmdSetPatchControlPointsEXT{};
@ -251,6 +253,9 @@ struct DeviceDispatch : InstanceDispatch {
PFN_vkCmdSetStencilWriteMask vkCmdSetStencilWriteMask{};
PFN_vkCmdSetVertexInputEXT vkCmdSetVertexInputEXT{};
PFN_vkCmdSetViewport vkCmdSetViewport{};
PFN_vkCmdSetColorWriteMaskEXT vkCmdSetColorWriteMaskEXT{};
PFN_vkCmdSetColorBlendEnableEXT vkCmdSetColorBlendEnableEXT{};
PFN_vkCmdSetColorBlendEquationEXT vkCmdSetColorBlendEquationEXT{};
PFN_vkCmdWaitEvents vkCmdWaitEvents{};
PFN_vkCreateBuffer vkCreateBuffer{};
PFN_vkCreateBufferView vkCreateBufferView{};
@ -1236,6 +1241,14 @@ public:
dld->vkCmdSetDepthBiasEnableEXT(handle, enable ? VK_TRUE : VK_FALSE);
}
void SetLogicOpEnableEXT(bool enable) const noexcept {
dld->vkCmdSetLogicOpEnableEXT(handle, enable ? VK_TRUE : VK_FALSE);
}
void SetDepthClampEnableEXT(bool enable) const noexcept {
dld->vkCmdSetDepthClampEnableEXT(handle, enable ? VK_TRUE : VK_FALSE);
}
void SetFrontFaceEXT(VkFrontFace front_face) const noexcept {
dld->vkCmdSetFrontFaceEXT(handle, front_face);
}
@ -1248,6 +1261,18 @@ public:
dld->vkCmdSetPatchControlPointsEXT(handle, patch_control_points);
}
void SetColorWriteMaskEXT(u32 first, Span<VkColorComponentFlags> masks) const noexcept {
dld->vkCmdSetColorWriteMaskEXT(handle, first, masks.size(), masks.data());
}
void SetColorBlendEnableEXT(u32 first, Span<VkBool32> enables) const noexcept {
dld->vkCmdSetColorBlendEnableEXT(handle, first, enables.size(), enables.data());
}
void SetColorBlendEquationEXT(u32 first, Span<VkColorBlendEquationEXT> equations) const noexcept {
dld->vkCmdSetColorBlendEquationEXT(handle, first, equations.size(), equations.data());
}
void SetLineWidth(float line_width) const noexcept {
dld->vkCmdSetLineWidth(handle, line_width);
}