Compare commits

...

18 Commits

Author SHA1 Message Date
ReinUsesLisp
b9e3f5eb36 fixed_pipeline_state: Define symetric operator!= and mark as noexcept
Marks as noexcept Hash, operator== and operator!= for consistency.
2019-12-24 18:24:08 -03:00
ReinUsesLisp
4a3026b16b fixed_pipeline_state: Define structure and loaders
The intention behind this hasheable structure is to describe the state
of fixed function pipeline state that gets compiled to a single graphics
pipeline state object. This is all dynamic state in OpenGL but Vulkan
wants it in an immutable state, even if hardware can edit it freely.

In this commit the structure is defined in an optimized state (it uses
booleans, has paddings and many data entries that can be packed to
single integers). This is intentional as an initial implementation that
is easier to debug, implement and review. It will be optimized in later
stages, or it might change if Vulkan gets more dynamic states.
2019-12-22 22:59:11 -03:00
ReinUsesLisp
5770418fb3 maxwell_3d: Add depth bounds registers 2019-12-22 22:55:06 -03:00
bunnei
e976d0e924 Merge pull request #3241 from ReinUsesLisp/gl-shader-cache
gl_shader_cache: Style changes
2019-12-22 16:23:46 -05:00
bunnei
1e76655f83 Merge pull request #3238 from ReinUsesLisp/vk-resource-manager
vk_resource_manager: Catch device losses and other changes
2019-12-22 15:57:16 -05:00
bunnei
0f3ac9cfeb Merge pull request #3203 from FernandoS27/tex-cache-fixes
Texture Cache: Add HLE methods for building 3D textures
2019-12-22 14:25:13 -05:00
Fernando Sahmkow
3dc585d011 Merge pull request #3237 from ReinUsesLisp/vk-shader-decompiler
vk_shader_decompiler: Misc changes
2019-12-22 12:36:56 -04:00
ReinUsesLisp
1e16023d60 gl_shader_cache: Update commentary for shared memory
Remove false commentary. Not dividing by 4 the size of shared memory is
not a hack; it describes the number of integers, not bytes.

While we are at it sort the generated code to put preprocessor lines on
the top.
2019-12-20 22:51:21 -03:00
ReinUsesLisp
486c6a5316 gl_shader_cache: Remove unused entry in GetPrimitiveDescription 2019-12-20 22:49:30 -03:00
ReinUsesLisp
e41da22c8d vk_resource_manager: Add entry to VKFence to test its usage 2019-12-19 16:31:34 -03:00
ReinUsesLisp
ec983a2451 vk_reosurce_manager: Add assert for releasing fences
Notify the programmer when a request to release a fence is invalid
because the fence is already free.
2019-12-19 16:31:34 -03:00
ReinUsesLisp
6ddffa010a vk_resource_manager: Implement VKFenceWatch move constructor
This allows us to put VKFenceWatch inside a std::vector without storing
it in heap. On move we have to signal the fences where the new protected
resource is, adding some overhead.
2019-12-19 16:31:34 -03:00
ReinUsesLisp
54747d60bc vk_device: Add entry to catch device losses
VK_NV_device_diagnostic_checkpoints allows us to push data to a Vulkan
queue and then query it even after a device loss. This allows us to push
the current pipeline object and see what was the call that killed the
device.
2019-12-19 16:31:33 -03:00
ReinUsesLisp
2a63b3bdb9 vk_shader_decompiler: Fix full decompilation
When full decompilation was enabled, labels were not being inserted and
instructions were misused. Fix these bugs.
2019-12-19 16:24:45 -03:00
ReinUsesLisp
de918ebeb0 vk_shader_decompiler: Skip NDC correction when it is native
Avoid changing gl_Position when the NDC used by the game is [0, 1]
(Vulkan's native).
2019-12-19 16:24:45 -03:00
ReinUsesLisp
485c21eac3 vk_shader_decompiler: Normalize output fragment attachments
Some games write from fragment shaders to an unexistant framebuffer
attachment or they don't write to one when it exists in the framebuffer.
Fix this by skipping writes or adding zeroes.
2019-12-19 16:24:45 -03:00
ReinUsesLisp
f4a25f854c vk_device: Add query for RGBA8Uint 2019-12-19 02:08:29 -03:00
ReinUsesLisp
abb33d4aec vk_shader_decompiler: Update sirit and implement Texture AOFFI 2019-12-19 01:42:13 -03:00
12 changed files with 742 additions and 67 deletions

View File

@@ -151,6 +151,8 @@ add_library(video_core STATIC
if (ENABLE_VULKAN)
target_sources(video_core PRIVATE
renderer_vulkan/declarations.h
renderer_vulkan/fixed_pipeline_state.cpp
renderer_vulkan/fixed_pipeline_state.h
renderer_vulkan/maxwell_to_vk.cpp
renderer_vulkan/maxwell_to_vk.h
renderer_vulkan/vk_buffer_cache.cpp

View File

@@ -707,13 +707,15 @@ public:
u32 color_mask_common;
INSERT_UNION_PADDING_WORDS(0x6);
u32 rt_separate_frag_data;
INSERT_UNION_PADDING_WORDS(0x2);
f32 depth_bounds[2];
INSERT_UNION_PADDING_WORDS(0xA);
INSERT_UNION_PADDING_WORDS(0x2);
u32 rt_separate_frag_data;
INSERT_UNION_PADDING_WORDS(0xC);
struct {
u32 address_high;
@@ -1030,7 +1032,12 @@ public:
BitField<4, 1, u32> depth_clamp_far;
} view_volume_clip_control;
INSERT_UNION_PADDING_WORDS(0x21);
INSERT_UNION_PADDING_WORDS(0x1F);
u32 depth_bounds_enable;
INSERT_UNION_PADDING_WORDS(1);
struct {
u32 enable;
LogicOperation operation;
@@ -1439,7 +1446,7 @@ ASSERT_REG_POSITION(stencil_back_func_mask, 0x3D6);
ASSERT_REG_POSITION(stencil_back_mask, 0x3D7);
ASSERT_REG_POSITION(color_mask_common, 0x3E4);
ASSERT_REG_POSITION(rt_separate_frag_data, 0x3EB);
ASSERT_REG_POSITION(depth_bounds, 0x3EC);
ASSERT_REG_POSITION(depth_bounds, 0x3E7);
ASSERT_REG_POSITION(zeta, 0x3F8);
ASSERT_REG_POSITION(clear_flags, 0x43E);
ASSERT_REG_POSITION(vertex_attrib_format, 0x458);
@@ -1495,6 +1502,7 @@ ASSERT_REG_POSITION(cull, 0x646);
ASSERT_REG_POSITION(pixel_center_integer, 0x649);
ASSERT_REG_POSITION(viewport_transform_enabled, 0x64B);
ASSERT_REG_POSITION(view_volume_clip_control, 0x64F);
ASSERT_REG_POSITION(depth_bounds_enable, 0x66F);
ASSERT_REG_POSITION(logic_op, 0x671);
ASSERT_REG_POSITION(clear_buffers, 0x674);
ASSERT_REG_POSITION(color_mask, 0x680);

View File

@@ -112,25 +112,25 @@ constexpr GLenum GetGLShaderType(ShaderType shader_type) {
}
/// Describes primitive behavior on geometry shaders
constexpr std::tuple<const char*, const char*, u32> GetPrimitiveDescription(GLenum primitive_mode) {
constexpr std::pair<const char*, u32> GetPrimitiveDescription(GLenum primitive_mode) {
switch (primitive_mode) {
case GL_POINTS:
return {"points", "Points", 1};
return {"points", 1};
case GL_LINES:
case GL_LINE_STRIP:
return {"lines", "Lines", 2};
return {"lines", 2};
case GL_LINES_ADJACENCY:
case GL_LINE_STRIP_ADJACENCY:
return {"lines_adjacency", "LinesAdj", 4};
return {"lines_adjacency", 4};
case GL_TRIANGLES:
case GL_TRIANGLE_STRIP:
case GL_TRIANGLE_FAN:
return {"triangles", "Triangles", 3};
return {"triangles", 3};
case GL_TRIANGLES_ADJACENCY:
case GL_TRIANGLE_STRIP_ADJACENCY:
return {"triangles_adjacency", "TrianglesAdj", 6};
return {"triangles_adjacency", 6};
default:
return {"points", "Invalid", 1};
return {"points", 1};
}
}
@@ -264,30 +264,25 @@ CachedProgram BuildShader(const Device& device, u64 unique_identifier, ShaderTyp
"#extension GL_NV_shader_thread_group : require\n"
"#extension GL_NV_shader_thread_shuffle : require\n";
}
source += '\n';
if (shader_type == ShaderType::Geometry) {
const auto [glsl_topology, debug_name, max_vertices] =
GetPrimitiveDescription(variant.primitive_mode);
source += fmt::format("layout ({}) in;\n\n", glsl_topology);
const auto [glsl_topology, max_vertices] = GetPrimitiveDescription(variant.primitive_mode);
source += fmt::format("#define MAX_VERTEX_INPUT {}\n", max_vertices);
source += fmt::format("layout ({}) in;\n", glsl_topology);
}
if (shader_type == ShaderType::Compute) {
if (variant.local_memory_size > 0) {
source += fmt::format("#define LOCAL_MEMORY_SIZE {}\n",
Common::AlignUp(variant.local_memory_size, 4) / 4);
}
source +=
fmt::format("layout (local_size_x = {}, local_size_y = {}, local_size_z = {}) in;\n",
variant.block_x, variant.block_y, variant.block_z);
if (variant.shared_memory_size > 0) {
// TODO(Rodrigo): We should divide by four here, but having a larger shared memory pool
// avoids out of bound stores. Find out why shared memory size is being invalid.
// shared_memory_size is described in number of words
source += fmt::format("shared uint smem[{}];\n", variant.shared_memory_size);
}
if (variant.local_memory_size > 0) {
source += fmt::format("#define LOCAL_MEMORY_SIZE {}\n",
Common::AlignUp(variant.local_memory_size, 4) / 4);
}
}
source += '\n';

View File

@@ -0,0 +1,296 @@
// Copyright 2019 yuzu Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include <tuple>
#include <boost/functional/hash.hpp>
#include "common/common_types.h"
#include "video_core/renderer_vulkan/fixed_pipeline_state.h"
namespace Vulkan {
namespace {
constexpr FixedPipelineState::DepthStencil GetDepthStencilState(const Maxwell& regs) {
const FixedPipelineState::StencilFace front_stencil(
regs.stencil_front_op_fail, regs.stencil_front_op_zfail, regs.stencil_front_op_zpass,
regs.stencil_front_func_func);
const FixedPipelineState::StencilFace back_stencil =
regs.stencil_two_side_enable
? FixedPipelineState::StencilFace(regs.stencil_back_op_fail, regs.stencil_back_op_zfail,
regs.stencil_back_op_zpass,
regs.stencil_back_func_func)
: front_stencil;
return FixedPipelineState::DepthStencil(
regs.depth_test_enable == 1, regs.depth_write_enabled == 1, regs.depth_bounds_enable == 1,
regs.stencil_enable == 1, regs.depth_test_func, front_stencil, back_stencil);
}
constexpr FixedPipelineState::InputAssembly GetInputAssemblyState(const Maxwell& regs) {
return FixedPipelineState::InputAssembly(
regs.draw.topology, regs.primitive_restart.enabled,
regs.draw.topology == Maxwell::PrimitiveTopology::Points ? regs.point_size : 0.0f);
}
constexpr FixedPipelineState::BlendingAttachment GetBlendingAttachmentState(
const Maxwell& regs, std::size_t render_target) {
const auto& mask = regs.color_mask[regs.color_mask_common ? 0 : render_target];
const std::array components = {mask.R != 0, mask.G != 0, mask.B != 0, mask.A != 0};
const FixedPipelineState::BlendingAttachment default_blending(
false, Maxwell::Blend::Equation::Add, Maxwell::Blend::Factor::One,
Maxwell::Blend::Factor::Zero, Maxwell::Blend::Equation::Add, Maxwell::Blend::Factor::One,
Maxwell::Blend::Factor::Zero, components);
if (render_target >= regs.rt_control.count) {
return default_blending;
}
if (!regs.independent_blend_enable) {
const auto& src = regs.blend;
if (!src.enable[render_target]) {
return default_blending;
}
return FixedPipelineState::BlendingAttachment(
true, src.equation_rgb, src.factor_source_rgb, src.factor_dest_rgb, src.equation_a,
src.factor_source_a, src.factor_dest_a, components);
}
if (!regs.blend.enable[render_target]) {
return default_blending;
}
const auto& src = regs.independent_blend[render_target];
return FixedPipelineState::BlendingAttachment(
true, src.equation_rgb, src.factor_source_rgb, src.factor_dest_rgb, src.equation_a,
src.factor_source_a, src.factor_dest_a, components);
}
constexpr FixedPipelineState::ColorBlending GetColorBlendingState(const Maxwell& regs) {
return FixedPipelineState::ColorBlending(
{regs.blend_color.r, regs.blend_color.g, regs.blend_color.b, regs.blend_color.a},
regs.rt_control.count,
{GetBlendingAttachmentState(regs, 0), GetBlendingAttachmentState(regs, 1),
GetBlendingAttachmentState(regs, 2), GetBlendingAttachmentState(regs, 3),
GetBlendingAttachmentState(regs, 4), GetBlendingAttachmentState(regs, 5),
GetBlendingAttachmentState(regs, 6), GetBlendingAttachmentState(regs, 7)});
}
constexpr FixedPipelineState::Tessellation GetTessellationState(const Maxwell& regs) {
return FixedPipelineState::Tessellation(regs.patch_vertices, regs.tess_mode.prim,
regs.tess_mode.spacing, regs.tess_mode.cw != 0);
}
constexpr std::size_t Point = 0;
constexpr std::size_t Line = 1;
constexpr std::size_t Polygon = 2;
constexpr std::array PolygonOffsetEnableLUT = {
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
};
constexpr FixedPipelineState::Rasterizer GetRasterizerState(const Maxwell& regs) {
const std::array enabled_lut = {regs.polygon_offset_point_enable,
regs.polygon_offset_line_enable,
regs.polygon_offset_fill_enable};
const auto topology = static_cast<std::size_t>(regs.draw.topology.Value());
const bool depth_bias_enabled = enabled_lut[PolygonOffsetEnableLUT[topology]];
Maxwell::Cull::FrontFace front_face = regs.cull.front_face;
if (regs.screen_y_control.triangle_rast_flip != 0 &&
regs.viewport_transform[0].scale_y > 0.0f) {
if (front_face == Maxwell::Cull::FrontFace::CounterClockWise)
front_face = Maxwell::Cull::FrontFace::ClockWise;
else if (front_face == Maxwell::Cull::FrontFace::ClockWise)
front_face = Maxwell::Cull::FrontFace::CounterClockWise;
}
const bool gl_ndc = regs.depth_mode == Maxwell::DepthMode::MinusOneToOne;
return FixedPipelineState::Rasterizer(regs.cull.enabled, depth_bias_enabled, gl_ndc,
regs.cull.cull_face, front_face);
}
} // Anonymous namespace
std::size_t FixedPipelineState::VertexBinding::Hash() const noexcept {
return (index << stride) ^ divisor;
}
bool FixedPipelineState::VertexBinding::operator==(const VertexBinding& rhs) const noexcept {
return std::tie(index, stride, divisor) == std::tie(rhs.index, rhs.stride, rhs.divisor);
}
std::size_t FixedPipelineState::VertexAttribute::Hash() const noexcept {
return static_cast<std::size_t>(index) ^ (static_cast<std::size_t>(buffer) << 13) ^
(static_cast<std::size_t>(type) << 22) ^ (static_cast<std::size_t>(size) << 31) ^
(static_cast<std::size_t>(offset) << 36);
}
bool FixedPipelineState::VertexAttribute::operator==(const VertexAttribute& rhs) const noexcept {
return std::tie(index, buffer, type, size, offset) ==
std::tie(rhs.index, rhs.buffer, rhs.type, rhs.size, rhs.offset);
}
std::size_t FixedPipelineState::StencilFace::Hash() const noexcept {
return static_cast<std::size_t>(action_stencil_fail) ^
(static_cast<std::size_t>(action_depth_fail) << 4) ^
(static_cast<std::size_t>(action_depth_fail) << 20) ^
(static_cast<std::size_t>(action_depth_pass) << 36);
}
bool FixedPipelineState::StencilFace::operator==(const StencilFace& rhs) const noexcept {
return std::tie(action_stencil_fail, action_depth_fail, action_depth_pass, test_func) ==
std::tie(rhs.action_stencil_fail, rhs.action_depth_fail, rhs.action_depth_pass,
rhs.test_func);
}
std::size_t FixedPipelineState::BlendingAttachment::Hash() const noexcept {
return static_cast<std::size_t>(enable) ^ (static_cast<std::size_t>(rgb_equation) << 5) ^
(static_cast<std::size_t>(src_rgb_func) << 10) ^
(static_cast<std::size_t>(dst_rgb_func) << 15) ^
(static_cast<std::size_t>(a_equation) << 20) ^
(static_cast<std::size_t>(src_a_func) << 25) ^
(static_cast<std::size_t>(dst_a_func) << 30) ^
(static_cast<std::size_t>(components[0]) << 35) ^
(static_cast<std::size_t>(components[1]) << 36) ^
(static_cast<std::size_t>(components[2]) << 37) ^
(static_cast<std::size_t>(components[3]) << 38);
}
bool FixedPipelineState::BlendingAttachment::operator==(const BlendingAttachment& rhs) const
noexcept {
return std::tie(enable, rgb_equation, src_rgb_func, dst_rgb_func, a_equation, src_a_func,
dst_a_func, components) ==
std::tie(rhs.enable, rhs.rgb_equation, rhs.src_rgb_func, rhs.dst_rgb_func,
rhs.a_equation, rhs.src_a_func, rhs.dst_a_func, rhs.components);
}
std::size_t FixedPipelineState::VertexInput::Hash() const noexcept {
std::size_t hash = num_bindings ^ (num_attributes << 32);
for (std::size_t i = 0; i < num_bindings; ++i) {
boost::hash_combine(hash, bindings[i].Hash());
}
for (std::size_t i = 0; i < num_attributes; ++i) {
boost::hash_combine(hash, attributes[i].Hash());
}
return hash;
}
bool FixedPipelineState::VertexInput::operator==(const VertexInput& rhs) const noexcept {
return std::equal(bindings.begin(), bindings.begin() + num_bindings, rhs.bindings.begin(),
rhs.bindings.begin() + rhs.num_bindings) &&
std::equal(attributes.begin(), attributes.begin() + num_attributes,
rhs.attributes.begin(), rhs.attributes.begin() + rhs.num_attributes);
}
std::size_t FixedPipelineState::InputAssembly::Hash() const noexcept {
std::size_t point_size_int = 0;
std::memcpy(&point_size_int, &point_size, sizeof(point_size));
return (static_cast<std::size_t>(topology) << 24) ^ (point_size_int << 32) ^
static_cast<std::size_t>(primitive_restart_enable);
}
bool FixedPipelineState::InputAssembly::operator==(const InputAssembly& rhs) const noexcept {
return std::tie(topology, primitive_restart_enable, point_size) ==
std::tie(rhs.topology, rhs.primitive_restart_enable, rhs.point_size);
}
std::size_t FixedPipelineState::Tessellation::Hash() const noexcept {
return static_cast<std::size_t>(patch_control_points) ^
(static_cast<std::size_t>(primitive) << 6) ^ (static_cast<std::size_t>(spacing) << 8) ^
(static_cast<std::size_t>(clockwise) << 10);
}
bool FixedPipelineState::Tessellation::operator==(const Tessellation& rhs) const noexcept {
return std::tie(patch_control_points, primitive, spacing, clockwise) ==
std::tie(rhs.patch_control_points, rhs.primitive, rhs.spacing, rhs.clockwise);
}
std::size_t FixedPipelineState::Rasterizer::Hash() const noexcept {
return static_cast<std::size_t>(cull_enable) ^
(static_cast<std::size_t>(depth_bias_enable) << 1) ^
(static_cast<std::size_t>(ndc_minus_one_to_one) << 2) ^
(static_cast<std::size_t>(cull_face) << 24) ^
(static_cast<std::size_t>(front_face) << 48);
}
bool FixedPipelineState::Rasterizer::operator==(const Rasterizer& rhs) const noexcept {
return std::tie(cull_enable, depth_bias_enable, ndc_minus_one_to_one, cull_face, front_face) ==
std::tie(rhs.cull_enable, rhs.depth_bias_enable, rhs.ndc_minus_one_to_one, rhs.cull_face,
rhs.front_face);
}
std::size_t FixedPipelineState::DepthStencil::Hash() const noexcept {
std::size_t hash = static_cast<std::size_t>(depth_test_enable) ^
(static_cast<std::size_t>(depth_write_enable) << 1) ^
(static_cast<std::size_t>(depth_bounds_enable) << 2) ^
(static_cast<std::size_t>(stencil_enable) << 3) ^
(static_cast<std::size_t>(depth_test_function) << 4);
boost::hash_combine(hash, front_stencil.Hash());
boost::hash_combine(hash, back_stencil.Hash());
return hash;
}
bool FixedPipelineState::DepthStencil::operator==(const DepthStencil& rhs) const noexcept {
return std::tie(depth_test_enable, depth_write_enable, depth_bounds_enable, depth_test_function,
stencil_enable, front_stencil, back_stencil) ==
std::tie(rhs.depth_test_enable, rhs.depth_write_enable, rhs.depth_bounds_enable,
rhs.depth_test_function, rhs.stencil_enable, rhs.front_stencil,
rhs.back_stencil);
}
std::size_t FixedPipelineState::ColorBlending::Hash() const noexcept {
std::size_t hash = attachments_count << 13;
for (std::size_t rt = 0; rt < static_cast<std::size_t>(attachments_count); ++rt) {
boost::hash_combine(hash, attachments[rt].Hash());
}
return hash;
}
bool FixedPipelineState::ColorBlending::operator==(const ColorBlending& rhs) const noexcept {
return std::equal(attachments.begin(), attachments.begin() + attachments_count,
rhs.attachments.begin(), rhs.attachments.begin() + rhs.attachments_count);
}
std::size_t FixedPipelineState::Hash() const noexcept {
std::size_t hash = 0;
boost::hash_combine(hash, vertex_input.Hash());
boost::hash_combine(hash, input_assembly.Hash());
boost::hash_combine(hash, tessellation.Hash());
boost::hash_combine(hash, rasterizer.Hash());
boost::hash_combine(hash, depth_stencil.Hash());
boost::hash_combine(hash, color_blending.Hash());
return hash;
}
bool FixedPipelineState::operator==(const FixedPipelineState& rhs) const noexcept {
return std::tie(vertex_input, input_assembly, tessellation, rasterizer, depth_stencil,
color_blending) == std::tie(rhs.vertex_input, rhs.input_assembly,
rhs.tessellation, rhs.rasterizer, rhs.depth_stencil,
rhs.color_blending);
}
FixedPipelineState GetFixedPipelineState(const Maxwell& regs) {
FixedPipelineState fixed_state;
fixed_state.input_assembly = GetInputAssemblyState(regs);
fixed_state.tessellation = GetTessellationState(regs);
fixed_state.rasterizer = GetRasterizerState(regs);
fixed_state.depth_stencil = GetDepthStencilState(regs);
fixed_state.color_blending = GetColorBlendingState(regs);
return fixed_state;
}
} // namespace Vulkan

View File

@@ -0,0 +1,282 @@
// Copyright 2019 yuzu Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#pragma once
#include <array>
#include <type_traits>
#include "common/common_types.h"
#include "video_core/engines/maxwell_3d.h"
#include "video_core/surface.h"
namespace Vulkan {
using Maxwell = Tegra::Engines::Maxwell3D::Regs;
// TODO(Rodrigo): Optimize this structure.
struct FixedPipelineState {
using PixelFormat = VideoCore::Surface::PixelFormat;
struct VertexBinding {
constexpr VertexBinding(u32 index, u32 stride, u32 divisor)
: index{index}, stride{stride}, divisor{divisor} {}
VertexBinding() = default;
u32 index;
u32 stride;
u32 divisor;
std::size_t Hash() const noexcept;
bool operator==(const VertexBinding& rhs) const noexcept;
bool operator!=(const VertexBinding& rhs) const noexcept {
return !operator==(rhs);
}
};
struct VertexAttribute {
constexpr VertexAttribute(u32 index, u32 buffer, Maxwell::VertexAttribute::Type type,
Maxwell::VertexAttribute::Size size, u32 offset)
: index{index}, buffer{buffer}, type{type}, size{size}, offset{offset} {}
VertexAttribute() = default;
u32 index;
u32 buffer;
Maxwell::VertexAttribute::Type type;
Maxwell::VertexAttribute::Size size;
u32 offset;
std::size_t Hash() const noexcept;
bool operator==(const VertexAttribute& rhs) const noexcept;
bool operator!=(const VertexAttribute& rhs) const noexcept {
return !operator==(rhs);
}
};
struct StencilFace {
constexpr StencilFace(Maxwell::StencilOp action_stencil_fail,
Maxwell::StencilOp action_depth_fail,
Maxwell::StencilOp action_depth_pass, Maxwell::ComparisonOp test_func)
: action_stencil_fail{action_stencil_fail}, action_depth_fail{action_depth_fail},
action_depth_pass{action_depth_pass}, test_func{test_func} {}
StencilFace() = default;
Maxwell::StencilOp action_stencil_fail;
Maxwell::StencilOp action_depth_fail;
Maxwell::StencilOp action_depth_pass;
Maxwell::ComparisonOp test_func;
std::size_t Hash() const noexcept;
bool operator==(const StencilFace& rhs) const noexcept;
bool operator!=(const StencilFace& rhs) const noexcept {
return !operator==(rhs);
}
};
struct BlendingAttachment {
constexpr BlendingAttachment(bool enable, Maxwell::Blend::Equation rgb_equation,
Maxwell::Blend::Factor src_rgb_func,
Maxwell::Blend::Factor dst_rgb_func,
Maxwell::Blend::Equation a_equation,
Maxwell::Blend::Factor src_a_func,
Maxwell::Blend::Factor dst_a_func,
std::array<bool, 4> components)
: enable{enable}, rgb_equation{rgb_equation}, src_rgb_func{src_rgb_func},
dst_rgb_func{dst_rgb_func}, a_equation{a_equation}, src_a_func{src_a_func},
dst_a_func{dst_a_func}, components{components} {}
BlendingAttachment() = default;
bool enable;
Maxwell::Blend::Equation rgb_equation;
Maxwell::Blend::Factor src_rgb_func;
Maxwell::Blend::Factor dst_rgb_func;
Maxwell::Blend::Equation a_equation;
Maxwell::Blend::Factor src_a_func;
Maxwell::Blend::Factor dst_a_func;
std::array<bool, 4> components;
std::size_t Hash() const noexcept;
bool operator==(const BlendingAttachment& rhs) const noexcept;
bool operator!=(const BlendingAttachment& rhs) const noexcept {
return !operator==(rhs);
}
};
struct VertexInput {
std::size_t num_bindings = 0;
std::size_t num_attributes = 0;
std::array<VertexBinding, Maxwell::NumVertexArrays> bindings;
std::array<VertexAttribute, Maxwell::NumVertexAttributes> attributes;
std::size_t Hash() const noexcept;
bool operator==(const VertexInput& rhs) const noexcept;
bool operator!=(const VertexInput& rhs) const noexcept {
return !operator==(rhs);
}
};
struct InputAssembly {
constexpr InputAssembly(Maxwell::PrimitiveTopology topology, bool primitive_restart_enable,
float point_size)
: topology{topology}, primitive_restart_enable{primitive_restart_enable},
point_size{point_size} {}
InputAssembly() = default;
Maxwell::PrimitiveTopology topology;
bool primitive_restart_enable;
float point_size;
std::size_t Hash() const noexcept;
bool operator==(const InputAssembly& rhs) const noexcept;
bool operator!=(const InputAssembly& rhs) const noexcept {
return !operator==(rhs);
}
};
struct Tessellation {
constexpr Tessellation(u32 patch_control_points, Maxwell::TessellationPrimitive primitive,
Maxwell::TessellationSpacing spacing, bool clockwise)
: patch_control_points{patch_control_points}, primitive{primitive}, spacing{spacing},
clockwise{clockwise} {}
Tessellation() = default;
u32 patch_control_points;
Maxwell::TessellationPrimitive primitive;
Maxwell::TessellationSpacing spacing;
bool clockwise;
std::size_t Hash() const noexcept;
bool operator==(const Tessellation& rhs) const noexcept;
bool operator!=(const Tessellation& rhs) const noexcept {
return !operator==(rhs);
}
};
struct Rasterizer {
constexpr Rasterizer(bool cull_enable, bool depth_bias_enable, bool ndc_minus_one_to_one,
Maxwell::Cull::CullFace cull_face, Maxwell::Cull::FrontFace front_face)
: cull_enable{cull_enable}, depth_bias_enable{depth_bias_enable},
ndc_minus_one_to_one{ndc_minus_one_to_one}, cull_face{cull_face}, front_face{
front_face} {}
Rasterizer() = default;
bool cull_enable;
bool depth_bias_enable;
bool ndc_minus_one_to_one;
Maxwell::Cull::CullFace cull_face;
Maxwell::Cull::FrontFace front_face;
std::size_t Hash() const noexcept;
bool operator==(const Rasterizer& rhs) const noexcept;
bool operator!=(const Rasterizer& rhs) const noexcept {
return !operator==(rhs);
}
};
struct DepthStencil {
constexpr DepthStencil(bool depth_test_enable, bool depth_write_enable,
bool depth_bounds_enable, bool stencil_enable,
Maxwell::ComparisonOp depth_test_function, StencilFace front_stencil,
StencilFace back_stencil)
: depth_test_enable{depth_test_enable}, depth_write_enable{depth_write_enable},
depth_bounds_enable{depth_bounds_enable}, stencil_enable{stencil_enable},
depth_test_function{depth_test_function}, front_stencil{front_stencil},
back_stencil{back_stencil} {}
DepthStencil() = default;
bool depth_test_enable;
bool depth_write_enable;
bool depth_bounds_enable;
bool stencil_enable;
Maxwell::ComparisonOp depth_test_function;
StencilFace front_stencil;
StencilFace back_stencil;
std::size_t Hash() const noexcept;
bool operator==(const DepthStencil& rhs) const noexcept;
bool operator!=(const DepthStencil& rhs) const noexcept {
return !operator==(rhs);
}
};
struct ColorBlending {
constexpr ColorBlending(
std::array<float, 4> blend_constants, std::size_t attachments_count,
std::array<BlendingAttachment, Maxwell::NumRenderTargets> attachments)
: attachments_count{attachments_count}, attachments{attachments} {}
ColorBlending() = default;
std::size_t attachments_count;
std::array<BlendingAttachment, Maxwell::NumRenderTargets> attachments;
std::size_t Hash() const noexcept;
bool operator==(const ColorBlending& rhs) const noexcept;
bool operator!=(const ColorBlending& rhs) const noexcept {
return !operator==(rhs);
}
};
std::size_t Hash() const noexcept;
bool operator==(const FixedPipelineState& rhs) const noexcept;
bool operator!=(const FixedPipelineState& rhs) const noexcept {
return !operator==(rhs);
}
VertexInput vertex_input;
InputAssembly input_assembly;
Tessellation tessellation;
Rasterizer rasterizer;
DepthStencil depth_stencil;
ColorBlending color_blending;
};
static_assert(std::is_trivially_copyable_v<FixedPipelineState::VertexBinding>);
static_assert(std::is_trivially_copyable_v<FixedPipelineState::VertexAttribute>);
static_assert(std::is_trivially_copyable_v<FixedPipelineState::StencilFace>);
static_assert(std::is_trivially_copyable_v<FixedPipelineState::BlendingAttachment>);
static_assert(std::is_trivially_copyable_v<FixedPipelineState::VertexInput>);
static_assert(std::is_trivially_copyable_v<FixedPipelineState::InputAssembly>);
static_assert(std::is_trivially_copyable_v<FixedPipelineState::Tessellation>);
static_assert(std::is_trivially_copyable_v<FixedPipelineState::Rasterizer>);
static_assert(std::is_trivially_copyable_v<FixedPipelineState::DepthStencil>);
static_assert(std::is_trivially_copyable_v<FixedPipelineState::ColorBlending>);
static_assert(std::is_trivially_copyable_v<FixedPipelineState>);
FixedPipelineState GetFixedPipelineState(const Maxwell& regs);
} // namespace Vulkan
namespace std {
template <>
struct hash<Vulkan::FixedPipelineState> {
std::size_t operator()(const Vulkan::FixedPipelineState& k) const noexcept {
return k.Hash();
}
};
} // namespace std

View File

@@ -3,12 +3,15 @@
// Refer to the license.txt file included.
#include <bitset>
#include <chrono>
#include <cstdlib>
#include <optional>
#include <set>
#include <string_view>
#include <thread>
#include <vector>
#include "common/assert.h"
#include "core/settings.h"
#include "video_core/renderer_vulkan/declarations.h"
#include "video_core/renderer_vulkan/vk_device.h"
@@ -201,6 +204,22 @@ vk::Format VKDevice::GetSupportedFormat(vk::Format wanted_format,
return wanted_format;
}
void VKDevice::ReportLoss() const {
LOG_CRITICAL(Render_Vulkan, "Device loss occured!");
// Wait some time to let the log flush
std::this_thread::sleep_for(std::chrono::seconds{1});
if (!nv_device_diagnostic_checkpoints) {
return;
}
[[maybe_unused]] const std::vector data = graphics_queue.getCheckpointDataNV(dld);
// Catch here in debug builds (or with optimizations disabled) the last graphics pipeline to be
// executed. It can be done on a debugger by evaluating the expression:
// *(VKGraphicsPipeline*)data[0]
}
bool VKDevice::IsOptimalAstcSupported(const vk::PhysicalDeviceFeatures& features,
const vk::DispatchLoaderDynamic& dldi) const {
// Disable for now to avoid converting ASTC twice.
@@ -381,6 +400,8 @@ std::vector<const char*> VKDevice::LoadExtensions(const vk::DispatchLoaderDynami
VK_EXT_SHADER_VIEWPORT_INDEX_LAYER_EXTENSION_NAME, true);
Test(extension, ext_subgroup_size_control, VK_EXT_SUBGROUP_SIZE_CONTROL_EXTENSION_NAME,
false);
Test(extension, nv_device_diagnostic_checkpoints,
VK_NV_DEVICE_DIAGNOSTIC_CHECKPOINTS_EXTENSION_NAME, true);
}
if (khr_shader_float16_int8) {
@@ -464,6 +485,7 @@ std::vector<vk::DeviceQueueCreateInfo> VKDevice::GetDeviceQueueCreateInfos() con
std::unordered_map<vk::Format, vk::FormatProperties> VKDevice::GetFormatProperties(
const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical) {
static constexpr std::array formats{vk::Format::eA8B8G8R8UnormPack32,
vk::Format::eA8B8G8R8UintPack32,
vk::Format::eA8B8G8R8SnormPack32,
vk::Format::eA8B8G8R8SrgbPack32,
vk::Format::eB5G6R5UnormPack16,

View File

@@ -39,6 +39,9 @@ public:
vk::Format GetSupportedFormat(vk::Format wanted_format, vk::FormatFeatureFlags wanted_usage,
FormatType format_type) const;
/// Reports a device loss.
void ReportLoss() const;
/// Returns the dispatch loader with direct function pointers of the device.
const vk::DispatchLoaderDynamic& GetDispatchLoader() const {
return dld;
@@ -159,6 +162,11 @@ public:
return ext_shader_viewport_index_layer;
}
/// Returns true if the device supports VK_NV_device_diagnostic_checkpoints.
bool IsNvDeviceDiagnosticCheckpoints() const {
return nv_device_diagnostic_checkpoints;
}
/// Returns the vendor name reported from Vulkan.
std::string_view GetVendorName() const {
return vendor_name;
@@ -218,6 +226,7 @@ private:
bool ext_index_type_uint8{}; ///< Support for VK_EXT_index_type_uint8.
bool ext_depth_range_unrestricted{}; ///< Support for VK_EXT_depth_range_unrestricted.
bool ext_shader_viewport_index_layer{}; ///< Support for VK_EXT_shader_viewport_index_layer.
bool nv_device_diagnostic_checkpoints{}; ///< Support for VK_NV_device_diagnostic_checkpoints.
// Telemetry parameters
std::string vendor_name; ///< Device's driver name.

View File

@@ -72,12 +72,22 @@ VKFence::VKFence(const VKDevice& device, UniqueFence handle)
VKFence::~VKFence() = default;
void VKFence::Wait() {
static constexpr u64 timeout = std::numeric_limits<u64>::max();
const auto dev = device.GetLogical();
const auto& dld = device.GetDispatchLoader();
dev.waitForFences({*handle}, true, std::numeric_limits<u64>::max(), dld);
switch (const auto result = dev.waitForFences(1, &*handle, true, timeout, dld)) {
case vk::Result::eSuccess:
return;
case vk::Result::eErrorDeviceLost:
device.ReportLoss();
[[fallthrough]];
default:
vk::throwResultException(result, "vk::waitForFences");
}
}
void VKFence::Release() {
ASSERT(is_owned);
is_owned = false;
}
@@ -133,8 +143,32 @@ void VKFence::Unprotect(VKResource* resource) {
protected_resources.erase(it);
}
void VKFence::RedirectProtection(VKResource* old_resource, VKResource* new_resource) noexcept {
std::replace(std::begin(protected_resources), std::end(protected_resources), old_resource,
new_resource);
}
VKFenceWatch::VKFenceWatch() = default;
VKFenceWatch::VKFenceWatch(VKFence& initial_fence) {
Watch(initial_fence);
}
VKFenceWatch::VKFenceWatch(VKFenceWatch&& rhs) noexcept {
fence = std::exchange(rhs.fence, nullptr);
if (fence) {
fence->RedirectProtection(&rhs, this);
}
}
VKFenceWatch& VKFenceWatch::operator=(VKFenceWatch&& rhs) noexcept {
fence = std::exchange(rhs.fence, nullptr);
if (fence) {
fence->RedirectProtection(&rhs, this);
}
return *this;
}
VKFenceWatch::~VKFenceWatch() {
if (fence) {
fence->Unprotect(this);

View File

@@ -65,6 +65,9 @@ public:
/// Removes protection for a resource.
void Unprotect(VKResource* resource);
/// Redirects one protected resource to a new address.
void RedirectProtection(VKResource* old_resource, VKResource* new_resource) noexcept;
/// Retreives the fence.
operator vk::Fence() const {
return *handle;
@@ -97,8 +100,13 @@ private:
class VKFenceWatch final : public VKResource {
public:
explicit VKFenceWatch();
VKFenceWatch(VKFence& initial_fence);
VKFenceWatch(VKFenceWatch&&) noexcept;
VKFenceWatch(const VKFenceWatch&) = delete;
~VKFenceWatch() override;
VKFenceWatch& operator=(VKFenceWatch&&) noexcept;
/// Waits for the fence to be released.
void Wait();
@@ -116,6 +124,14 @@ public:
void OnFenceRemoval(VKFence* signaling_fence) override;
/**
* Do not use it paired with Watch. Use TryWatch instead.
* Returns true when the watch is free.
*/
bool IsUsed() const {
return fence != nullptr;
}
private:
VKFence* fence{}; ///< Fence watching this resource. nullptr when the watch is free.
};

View File

@@ -543,7 +543,7 @@ private:
}
for (u32 rt = 0; rt < static_cast<u32>(frag_colors.size()); ++rt) {
if (!IsRenderTargetUsed(rt)) {
if (!specialization.enabled_rendertargets[rt]) {
continue;
}
@@ -1555,26 +1555,11 @@ private:
Expression Texture(Operation operation) {
const auto& meta = std::get<MetaTexture>(operation.GetMeta());
UNIMPLEMENTED_IF(!meta.aoffi.empty());
const bool can_implicit = stage == ShaderType::Fragment;
const Id sampler = GetTextureSampler(operation);
const Id coords = GetCoordinates(operation, Type::Float);
if (meta.depth_compare) {
// Depth sampling
UNIMPLEMENTED_IF(meta.bias);
const Id dref = AsFloat(Visit(meta.depth_compare));
if (can_implicit) {
return {OpImageSampleDrefImplicitLod(t_float, sampler, coords, dref, {}),
Type::Float};
} else {
return {OpImageSampleDrefExplicitLod(t_float, sampler, coords, dref,
spv::ImageOperandsMask::Lod, v_float_zero),
Type::Float};
}
}
std::vector<Id> operands;
spv::ImageOperandsMask mask{};
if (meta.bias) {
@@ -1582,13 +1567,36 @@ private:
operands.push_back(AsFloat(Visit(meta.bias)));
}
if (!can_implicit) {
mask = mask | spv::ImageOperandsMask::Lod;
operands.push_back(v_float_zero);
}
if (!meta.aoffi.empty()) {
mask = mask | spv::ImageOperandsMask::Offset;
operands.push_back(GetOffsetCoordinates(operation));
}
if (meta.depth_compare) {
// Depth sampling
UNIMPLEMENTED_IF(meta.bias);
const Id dref = AsFloat(Visit(meta.depth_compare));
if (can_implicit) {
return {
OpImageSampleDrefImplicitLod(t_float, sampler, coords, dref, mask, operands),
Type::Float};
} else {
return {
OpImageSampleDrefExplicitLod(t_float, sampler, coords, dref, mask, operands),
Type::Float};
}
}
Id texture;
if (can_implicit) {
texture = OpImageSampleImplicitLod(t_float4, sampler, coords, mask, operands);
} else {
texture = OpImageSampleExplicitLod(t_float4, sampler, coords,
mask | spv::ImageOperandsMask::Lod, v_float_zero,
operands);
texture = OpImageSampleExplicitLod(t_float4, sampler, coords, mask, operands);
}
return GetTextureElement(operation, texture, Type::Float);
}
@@ -1601,7 +1609,8 @@ private:
const Id lod = AsFloat(Visit(meta.lod));
spv::ImageOperandsMask mask = spv::ImageOperandsMask::Lod;
std::vector<Id> operands;
std::vector<Id> operands{lod};
if (!meta.aoffi.empty()) {
mask = mask | spv::ImageOperandsMask::Offset;
operands.push_back(GetOffsetCoordinates(operation));
@@ -1609,11 +1618,10 @@ private:
if (meta.sampler.IsShadow()) {
const Id dref = AsFloat(Visit(meta.depth_compare));
return {
OpImageSampleDrefExplicitLod(t_float, sampler, coords, dref, mask, lod, operands),
Type::Float};
return {OpImageSampleDrefExplicitLod(t_float, sampler, coords, dref, mask, operands),
Type::Float};
}
const Id texture = OpImageSampleExplicitLod(t_float4, sampler, coords, mask, lod, operands);
const Id texture = OpImageSampleExplicitLod(t_float4, sampler, coords, mask, operands);
return GetTextureElement(operation, texture, Type::Float);
}
@@ -1722,7 +1730,7 @@ private:
const std::vector grad = {dx, dy};
static constexpr auto mask = spv::ImageOperandsMask::Grad;
const Id texture = OpImageSampleImplicitLod(t_float4, sampler, coords, mask, grad);
const Id texture = OpImageSampleExplicitLod(t_float4, sampler, coords, mask, grad);
return GetTextureElement(operation, texture, Type::Float);
}
@@ -1833,7 +1841,7 @@ private:
}
void PreExit() {
if (stage == ShaderType::Vertex) {
if (stage == ShaderType::Vertex && specialization.ndc_minus_one_to_one) {
const u32 position_index = out_indices.position.value();
const Id z_pointer = AccessElement(t_out_float, out_vertex, position_index, 2U);
const Id w_pointer = AccessElement(t_out_float, out_vertex, position_index, 3U);
@@ -1860,12 +1868,18 @@ private:
// rendertargets/components are skipped in the register assignment.
u32 current_reg = 0;
for (u32 rt = 0; rt < Maxwell::NumRenderTargets; ++rt) {
if (!specialization.enabled_rendertargets[rt]) {
// Skip rendertargets that are not enabled
continue;
}
// TODO(Subv): Figure out how dual-source blending is configured in the Switch.
for (u32 component = 0; component < 4; ++component) {
const Id pointer = AccessElement(t_out_float, frag_colors.at(rt), component);
if (header.ps.IsColorComponentOutputEnabled(rt, component)) {
OpStore(AccessElement(t_out_float, frag_colors.at(rt), component),
SafeGetRegister(current_reg));
OpStore(pointer, SafeGetRegister(current_reg));
++current_reg;
} else {
OpStore(pointer, component == 3 ? v_float_one : v_float_zero);
}
}
}
@@ -1995,15 +2009,6 @@ private:
return DeclareBuiltIn(builtin, spv::StorageClass::Input, type, std::move(name));
}
bool IsRenderTargetUsed(u32 rt) const {
for (u32 component = 0; component < 4; ++component) {
if (header.ps.IsColorComponentOutputEnabled(rt, component)) {
return true;
}
}
return false;
}
template <typename... Args>
Id AccessElement(Id pointer_type, Id composite, Args... elements_) {
std::vector<Id> members;
@@ -2567,7 +2572,7 @@ public:
const Id target = decomp.Constant(decomp.t_uint, expr.value);
Id gpr = decomp.OpLoad(decomp.t_float, decomp.registers.at(expr.gpr));
gpr = decomp.OpBitcast(decomp.t_uint, gpr);
return decomp.OpLogicalEqual(decomp.t_uint, gpr, target);
return decomp.OpIEqual(decomp.t_bool, gpr, target);
}
Id Visit(const Expr& node) {
@@ -2637,11 +2642,11 @@ public:
const Id loop_label = decomp.OpLabel();
const Id endloop_label = decomp.OpLabel();
const Id loop_start_block = decomp.OpLabel();
const Id loop_end_block = decomp.OpLabel();
const Id loop_continue_block = decomp.OpLabel();
current_loop_exit = endloop_label;
decomp.OpBranch(loop_label);
decomp.AddLabel(loop_label);
decomp.OpLoopMerge(endloop_label, loop_end_block, spv::LoopControlMask::MaskNone);
decomp.OpLoopMerge(endloop_label, loop_continue_block, spv::LoopControlMask::MaskNone);
decomp.OpBranch(loop_start_block);
decomp.AddLabel(loop_start_block);
ASTNode current = ast.nodes.GetFirst();
@@ -2649,6 +2654,8 @@ public:
Visit(current);
current = current->GetNext();
}
decomp.OpBranch(loop_continue_block);
decomp.AddLabel(loop_continue_block);
ExprDecompiler expr_parser{decomp};
const Id condition = expr_parser.Visit(ast.condition);
decomp.OpBranchConditional(condition, loop_label, endloop_label);

View File

@@ -94,6 +94,7 @@ struct Specialization final {
Maxwell::PrimitiveTopology primitive_topology{};
std::optional<float> point_size{};
std::array<Maxwell::VertexAttribute::Type, Maxwell::NumVertexAttributes> attribute_types{};
bool ndc_minus_one_to_one{};
// Tessellation specific
struct {
@@ -101,6 +102,9 @@ struct Specialization final {
Maxwell::TessellationSpacing spacing{};
bool clockwise{};
} tessellation;
// Fragment specific
std::bitset<8> enabled_rendertargets;
};
// Old gcc versions don't consider this trivially copyable.
// static_assert(std::is_trivially_copyable_v<Specialization>);