Compare commits

..

4 Commits

Author SHA1 Message Date
lizzie 9bd8924aeb 2026-09-05 05:55:38
Signed-off-by: lizzie <lizzie@eden-emu.dev>
2026-09-05 23:57:00 +02:00
lizzie 702fb672c2 2026-09-05 05:53:11
Signed-off-by: lizzie <lizzie@eden-emu.dev>
2026-09-05 23:57:00 +02:00
lizzie 4dca6ccdb1 2026-09-04 23:59:41
Signed-off-by: lizzie <lizzie@eden-emu.dev>
2026-09-05 23:57:00 +02:00
lizzie 06dc97a290 2026-09-04 23:57:52
Signed-off-by: lizzie <lizzie@eden-emu.dev>
2026-09-05 23:57:00 +02:00
29 changed files with 70 additions and 1112 deletions
@@ -30,7 +30,6 @@ enum class BooleanSetting(override val key: String) : AbstractBooleanSetting {
RENDERER_REACTIVE_FLUSHING("use_reactive_flushing"), RENDERER_REACTIVE_FLUSHING("use_reactive_flushing"),
ENABLE_BUFFER_HISTORY("enable_buffer_history"), ENABLE_BUFFER_HISTORY("enable_buffer_history"),
USE_OPTIMIZED_VERTEX_BUFFERS("use_optimized_vertex_buffers"), USE_OPTIMIZED_VERTEX_BUFFERS("use_optimized_vertex_buffers"),
ENABLE_SHADER_PHI_TRACKING("enable_shader_phi_tracking"),
ENABLE_GPU_BUFFER_READBACK("enable_gpu_buffer_readback"), ENABLE_GPU_BUFFER_READBACK("enable_gpu_buffer_readback"),
SYNC_MEMORY_OPERATIONS("sync_memory_operations"), SYNC_MEMORY_OPERATIONS("sync_memory_operations"),
BUFFER_REORDER_DISABLE("disable_buffer_reorder"), BUFFER_REORDER_DISABLE("disable_buffer_reorder"),
@@ -927,13 +927,6 @@ abstract class SettingsItem(
descriptionId = R.string.use_optimized_vertex_buffers_description descriptionId = R.string.use_optimized_vertex_buffers_description
) )
) )
put(
SwitchSetting(
BooleanSetting.ENABLE_SHADER_PHI_TRACKING,
titleId = R.string.enable_shader_phi_tracking,
descriptionId = R.string.enable_shader_phi_tracking_description
)
)
put( put(
SwitchSetting( SwitchSetting(
BooleanSetting.SYNC_MEMORY_OPERATIONS, BooleanSetting.SYNC_MEMORY_OPERATIONS,
@@ -338,7 +338,6 @@ class SettingsFragmentPresenter(
add(BooleanSetting.ENABLE_BUFFER_HISTORY.key) add(BooleanSetting.ENABLE_BUFFER_HISTORY.key)
add(BooleanSetting.ENABLE_GPU_BUFFER_READBACK.key) add(BooleanSetting.ENABLE_GPU_BUFFER_READBACK.key)
add(BooleanSetting.USE_OPTIMIZED_VERTEX_BUFFERS.key) add(BooleanSetting.USE_OPTIMIZED_VERTEX_BUFFERS.key)
add(BooleanSetting.ENABLE_SHADER_PHI_TRACKING.key)
add(HeaderSetting(R.string.hacks)) add(HeaderSetting(R.string.hacks))
@@ -570,8 +570,6 @@
<string name="enable_gpu_buffer_readback_description">Preserves GPU-modified buffer data by reading it back before uploads. Some games require this to render certain effects properly. May cause issues if the hardware cannot handle the additional workload.</string> <string name="enable_gpu_buffer_readback_description">Preserves GPU-modified buffer data by reading it back before uploads. Some games require this to render certain effects properly. May cause issues if the hardware cannot handle the additional workload.</string>
<string name="use_optimized_vertex_buffers">Optimized Vertex Buffers</string> <string name="use_optimized_vertex_buffers">Optimized Vertex Buffers</string>
<string name="use_optimized_vertex_buffers_description">Enables optimized vertex buffer binding for improved performance. Requires Mesa 26.0+ Turnip drivers/ QCOM drivers. Will crash on older Turnip drivers (25.3 and below).</string> <string name="use_optimized_vertex_buffers_description">Enables optimized vertex buffer binding for improved performance. Requires Mesa 26.0+ Turnip drivers/ QCOM drivers. Will crash on older Turnip drivers (25.3 and below).</string>
<string name="enable_shader_phi_tracking">Shader Phi Tracking</string>
<string name="enable_shader_phi_tracking_description">toggle for test on shader phi tracking.</string>
<string name="hacks">Hacks</string> <string name="hacks">Hacks</string>
-8
View File
@@ -592,14 +592,6 @@ struct Values {
true, true,
true}; true};
SwitchableSetting<bool> enable_shader_phi_tracking{linkage,
true,
"enable_shader_phi_tracking",
Category::RendererAdvanced,
Specialization::Default,
true,
true};
#ifdef __ANDROID__ #ifdef __ANDROID__
SwitchableSetting<bool> use_optimized_vertex_buffers{linkage, SwitchableSetting<bool> use_optimized_vertex_buffers{linkage,
false, false,
+11 -13
View File
@@ -4,17 +4,16 @@
// SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project // SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later // SPDX-License-Identifier: GPL-2.0-or-later
#include <thread> #include <chrono>
#include <fmt/ranges.h> #include <fmt/ranges.h>
#include <math.h> #include <math.h>
#include "common/param_package.h" #include "common/param_package.h"
#include "common/settings.h" #include "common/settings.h"
#include "common/thread.h" #include "common/steady_clock.h"
#include "input_common/drivers/mouse.h" #include "input_common/drivers/mouse.h"
namespace InputCommon { namespace InputCommon {
constexpr int update_time = 10;
constexpr float default_panning_sensitivity = 0.0010f; constexpr float default_panning_sensitivity = 0.0010f;
constexpr float default_stick_sensitivity = 0.0006f; constexpr float default_stick_sensitivity = 0.0006f;
constexpr float default_deadzone_counterweight = 0.01f; constexpr float default_deadzone_counterweight = 0.01f;
@@ -74,7 +73,7 @@ Mouse::Mouse(std::string input_engine_) : InputEngine(std::move(input_engine_))
last_motion_change = {}; last_motion_change = {};
} }
void Mouse::UpdateStickInput() { void Mouse::UpdateStickInput(Common::SteadyClock::time_point timestamp) {
if (!IsMousePanningEnabled()) { if (!IsMousePanningEnabled()) {
return; return;
} }
@@ -100,12 +99,9 @@ void Mouse::UpdateStickInput() {
last_mouse_change *= clamped_decay; last_mouse_change *= clamped_decay;
} }
void Mouse::UpdateMotionInput() { void Mouse::UpdateMotionInput(Common::SteadyClock::time_point timestamp) {
const float sensitivity = const float sensitivity = IsMousePanningEnabled() ? default_motion_panning_sensitivity : default_motion_sensitivity;
IsMousePanningEnabled() ? default_motion_panning_sensitivity : default_motion_sensitivity; const float rotation_velocity = std::sqrt(last_motion_change.x * last_motion_change.x + last_motion_change.y * last_motion_change.y);
const float rotation_velocity = std::sqrt(last_motion_change.x * last_motion_change.x +
last_motion_change.y * last_motion_change.y);
// Clamp rotation speed // Clamp rotation speed
if (rotation_velocity > maximum_rotation_speed / sensitivity) { if (rotation_velocity > maximum_rotation_speed / sensitivity) {
@@ -121,7 +117,7 @@ void Mouse::UpdateMotionInput() {
.accel_x = 0, .accel_x = 0,
.accel_y = 0, .accel_y = 0,
.accel_z = 0, .accel_z = 0,
.delta_timestamp = update_time * 1000, .delta_timestamp = u64(std::chrono::duration_cast<std::chrono::microseconds>(timestamp - last_notify_timestamp).count()),
}; };
if (IsMousePanningEnabled()) { if (IsMousePanningEnabled()) {
@@ -177,8 +173,10 @@ void Mouse::Move(int x, int y, int center_x, int center_y) {
} }
void Mouse::NotifyChanged() { void Mouse::NotifyChanged() {
UpdateStickInput(); auto const timestamp = Common::SteadyClock::Now();
UpdateMotionInput(); UpdateStickInput(timestamp);
UpdateMotionInput(timestamp);
last_notify_timestamp = Common::SteadyClock::Now();
} }
void Mouse::MouseMove(f32 touch_x, f32 touch_y) { void Mouse::MouseMove(f32 touch_x, f32 touch_y) {
+10 -7
View File
@@ -7,7 +7,9 @@
#pragma once #pragma once
#include <thread> #include <thread>
#include <chrono>
#include "common/steady_clock.h"
#include "common/polyfill_thread.h" #include "common/polyfill_thread.h"
#include "common/vector_math.h" #include "common/vector_math.h"
#include "input_common/input_engine.h" #include "input_common/input_engine.h"
@@ -101,17 +103,18 @@ public:
Common::Input::ButtonNames GetUIName(const Common::ParamPackage& params) const override; Common::Input::ButtonNames GetUIName(const Common::ParamPackage& params) const override;
private: private:
void UpdateStickInput(); void UpdateStickInput(Common::SteadyClock::time_point timestamp);
void UpdateMotionInput(); void UpdateMotionInput(Common::SteadyClock::time_point timestamp);
bool IsMousePanningEnabled(); bool IsMousePanningEnabled();
Common::Input::ButtonNames GetUIButtonName(const Common::ParamPackage& params) const; Common::Input::ButtonNames GetUIButtonName(const Common::ParamPackage& params) const;
Common::Vec2<int> mouse_origin; Common::Vec2<int> mouse_origin{};
Common::Vec2<int> last_mouse_position; Common::Vec2<int> last_mouse_position{};
Common::Vec2<float> last_mouse_change; Common::Vec2<float> last_mouse_change{};
Common::Vec3<float> last_motion_change; Common::Vec3<float> last_motion_change{};
Common::Vec2<int> wheel_position; Common::Vec2<int> wheel_position{};
Common::SteadyClock::time_point last_notify_timestamp{};
bool button_pressed = false; bool button_pressed = false;
}; };
@@ -267,8 +267,6 @@ std::unique_ptr<TranslationMap> InitializeTranslations(QObject* parent) {
INSERT(Settings, enable_buffer_history, tr("Enable buffer history"), INSERT(Settings, enable_buffer_history, tr("Enable buffer history"),
tr("Enables access to previous buffer states.\nThis option may improve rendering " tr("Enables access to previous buffer states.\nThis option may improve rendering "
"quality and performance consistency in some games.")); "quality and performance consistency in some games."));
INSERT(Settings, enable_shader_phi_tracking, tr("Shader phi tracking"),
tr("toggle for test on shader phi tracking."));
INSERT(Settings, fix_bloom_effects, tr("Fix bloom effects"), tr("Removes bloom in Burnout.")); INSERT(Settings, fix_bloom_effects, tr("Fix bloom effects"), tr("Removes bloom in Burnout."));
INSERT(Settings, rescale_hack, tr("Enable Legacy Rescale Pass"), INSERT(Settings, rescale_hack, tr("Enable Legacy Rescale Pass"),
+15 -124
View File
@@ -299,128 +299,23 @@ static inline bool IsTexturePixelFormatIntegerCached(Environment& env,
} }
constexpr size_t PHI_TRACK_MAX_DEPTH = 3; std::optional<ConstBufferAddr> Track(const IR::Value& value, Environment& env, const HostTranslateInfo& host_info);
static inline std::optional<ConstBufferAddr> TrackCached(const IR::Value& v, Environment& env, const HostTranslateInfo& host_info) {
struct PhiTrackState {
boost::container::small_vector<const IR::Inst*, 8> active;
size_t depth{};
};
std::optional<ConstBufferAddr> Track(const IR::Value& value, Environment& env,
const HostTranslateInfo& host_info, PhiTrackState& state);
static inline std::optional<ConstBufferAddr> TrackCached(const IR::Value& v, Environment& env,
const HostTranslateInfo& host_info,
PhiTrackState& state) {
if (const IR::Inst* key = v.InstRecursive()) { if (const IR::Inst* key = v.InstRecursive()) {
if (auto it = env.track_cache.find(key); it != env.track_cache.end()) return it->second; if (auto it = env.track_cache.find(key); it != env.track_cache.end()) return it->second;
auto found = Track(v, env, host_info, state); auto found = Track(v, env, host_info);
if (found) env.track_cache.emplace(key, *found); if (found) env.track_cache.emplace(key, *found);
return found; return found;
} }
return Track(v, env, host_info, state); return Track(v, env, host_info);
} }
std::optional<ConstBufferAddr> TryGetConstBuffer(const IR::Inst* inst, Environment& env, std::optional<ConstBufferAddr> TryGetConstBuffer(const IR::Inst* inst, Environment& env, const HostTranslateInfo& host_info);
const HostTranslateInfo& host_info,
PhiTrackState& state);
bool IsSameConstBufferAddr(const ConstBufferAddr& lhs, const ConstBufferAddr& rhs) { std::optional<ConstBufferAddr> Track(const IR::Value& value, Environment& env, const HostTranslateInfo& host_info) {
return lhs.index == rhs.index && lhs.offset == rhs.offset && return IR::BreadthFirstSearch(value, [&env, &host_info](const IR::Inst* inst) {
lhs.shift_left == rhs.shift_left && lhs.secondary_index == rhs.secondary_index && return TryGetConstBuffer(inst, env, host_info);
lhs.secondary_offset == rhs.secondary_offset && });
lhs.secondary_shift_left == rhs.secondary_shift_left && lhs.count == rhs.count &&
lhs.has_secondary == rhs.has_secondary && lhs.dynamic_offset == rhs.dynamic_offset;
}
std::optional<ConstBufferAddr> TrackUncached(const IR::Value& value, Environment& env,
const HostTranslateInfo& host_info,
PhiTrackState& state, bool& ambiguous);
std::optional<ConstBufferAddr> TrackPhi(const IR::Inst* phi, Environment& env,
const HostTranslateInfo& host_info, PhiTrackState& state,
bool& ambiguous) {
if (state.depth >= PHI_TRACK_MAX_DEPTH) {
ambiguous = true;
return std::nullopt;
}
if (std::ranges::find(state.active, phi) != state.active.end()) {
return std::nullopt;
}
state.active.push_back(phi);
++state.depth;
std::optional<ConstBufferAddr> agreed;
bool failed = false;
const size_t num_args{phi->NumArgs()};
for (size_t index = 0; index < num_args; ++index) {
const IR::Value arg{phi->Arg(index).Resolve()};
if (arg.IsImmediate()) {
failed = true;
break;
}
const IR::Inst* arg_inst{arg.InstRecursive()};
if (arg_inst == phi) {
continue;
}
if (std::ranges::find(state.active, arg_inst) != state.active.end()) {
continue;
}
bool operand_ambiguous = false;
const std::optional<ConstBufferAddr> operand{
TrackUncached(arg, env, host_info, state, operand_ambiguous)};
if (!operand || operand_ambiguous) {
failed = true;
break;
}
if (!agreed) {
agreed = operand;
continue;
}
if (!IsSameConstBufferAddr(*agreed, *operand)) {
failed = true;
break;
}
}
--state.depth;
state.active.pop_back();
if (failed || !agreed) {
ambiguous = true;
return std::nullopt;
}
return agreed;
}
std::optional<ConstBufferAddr> TrackUncached(const IR::Value& value, Environment& env,
const HostTranslateInfo& host_info,
PhiTrackState& state, bool& ambiguous) {
return IR::BreadthFirstSearch(
value, [&env, &host_info, &state, &ambiguous](
const IR::Inst* inst) -> std::optional<ConstBufferAddr> {
if (inst->GetOpcode() == IR::Opcode::Phi) {
return TrackPhi(inst, env, host_info, state, ambiguous);
}
return TryGetConstBuffer(inst, env, host_info, state);
});
}
std::optional<ConstBufferAddr> Track(const IR::Value& value, Environment& env,
const HostTranslateInfo& host_info, PhiTrackState& state) {
if (!Settings::values.enable_shader_phi_tracking.GetValue()) {
return IR::BreadthFirstSearch(
value, [&env, &host_info, &state](
const IR::Inst* inst) -> std::optional<ConstBufferAddr> {
return TryGetConstBuffer(inst, env, host_info, state);
});
}
bool ambiguous = false;
const std::optional<ConstBufferAddr> result{
TrackUncached(value, env, host_info, state, ambiguous)};
if (ambiguous) {
return std::nullopt;
}
return result;
} }
std::optional<u32> TryGetConstant(IR::Value& value, Environment& env) { std::optional<u32> TryGetConstant(IR::Value& value, Environment& env) {
@@ -444,15 +339,13 @@ std::optional<u32> TryGetConstant(IR::Value& value, Environment& env) {
return ReadCbufCached(env, index_number, offset_number); return ReadCbufCached(env, index_number, offset_number);
} }
std::optional<ConstBufferAddr> TryGetConstBuffer(const IR::Inst* inst, Environment& env, std::optional<ConstBufferAddr> TryGetConstBuffer(const IR::Inst* inst, Environment& env, const HostTranslateInfo& host_info) {
const HostTranslateInfo& host_info,
PhiTrackState& state) {
switch (inst->GetOpcode()) { switch (inst->GetOpcode()) {
default: default:
return std::nullopt; return std::nullopt;
case IR::Opcode::BitwiseOr32: { case IR::Opcode::BitwiseOr32: {
std::optional lhs{TrackCached(inst->Arg(0), env, host_info, state)}; std::optional lhs{TrackCached(inst->Arg(0), env, host_info)};
std::optional rhs{TrackCached(inst->Arg(1), env, host_info, state)}; std::optional rhs{TrackCached(inst->Arg(1), env, host_info)};
if (!lhs || !rhs) { if (!lhs || !rhs) {
return std::nullopt; return std::nullopt;
} }
@@ -482,7 +375,7 @@ std::optional<ConstBufferAddr> TryGetConstBuffer(const IR::Inst* inst, Environme
if (!shift.IsImmediate()) { if (!shift.IsImmediate()) {
return std::nullopt; return std::nullopt;
} }
std::optional lhs{TrackCached(inst->Arg(0), env, host_info, state)}; std::optional lhs{TrackCached(inst->Arg(0), env, host_info)};
if (lhs) { if (lhs) {
lhs->shift_left = shift.U32(); lhs->shift_left = shift.U32();
} }
@@ -510,7 +403,7 @@ std::optional<ConstBufferAddr> TryGetConstBuffer(const IR::Inst* inst, Environme
return std::nullopt; return std::nullopt;
} while (false); } while (false);
} }
std::optional lhs{TrackCached(op1, env, host_info, state)}; std::optional lhs{TrackCached(op1, env, host_info)};
if (lhs) { if (lhs) {
lhs->shift_left = static_cast<u32>(std::countr_zero(op2.U32())); lhs->shift_left = static_cast<u32>(std::countr_zero(op2.U32()));
} }
@@ -576,9 +469,7 @@ std::optional<ConstBufferAddr> TryGetConstBuffer(const IR::Inst* inst, Environme
TextureInst MakeInst(Environment& env, IR::Block* block, IR::Inst& inst, const HostTranslateInfo& host_info) { TextureInst MakeInst(Environment& env, IR::Block* block, IR::Inst& inst, const HostTranslateInfo& host_info) {
ConstBufferAddr addr; ConstBufferAddr addr;
if (IsBindless(inst)) { if (IsBindless(inst)) {
PhiTrackState state; const std::optional<ConstBufferAddr> track_addr{TrackCached(inst.Arg(0), env, host_info)};
const std::optional<ConstBufferAddr> track_addr{
TrackCached(inst.Arg(0), env, host_info, state)};
if (!track_addr) { if (!track_addr) {
throw NotImplementedException("Failed to track bindless texture constant buffer"); throw NotImplementedException("Failed to track bindless texture constant buffer");
-3
View File
@@ -20,7 +20,6 @@ add_library(video_core STATIC
buffer_cache/buffer_cache.h buffer_cache/buffer_cache.h
buffer_cache/memory_tracker_base.h buffer_cache/memory_tracker_base.h
buffer_cache/usage_tracker.h buffer_cache/usage_tracker.h
buffer_cache/virtual_range_cache.h
buffer_cache/word_manager.h buffer_cache/word_manager.h
cache_types.h cache_types.h
capture.h capture.h
@@ -167,8 +166,6 @@ add_library(video_core STATIC
renderer_vulkan/vk_fence_manager.h renderer_vulkan/vk_fence_manager.h
renderer_vulkan/vk_graphics_pipeline.cpp renderer_vulkan/vk_graphics_pipeline.cpp
renderer_vulkan/vk_graphics_pipeline.h renderer_vulkan/vk_graphics_pipeline.h
renderer_vulkan/vk_multi_range_buffer.cpp
renderer_vulkan/vk_multi_range_buffer.h
renderer_vulkan/vk_master_semaphore.cpp renderer_vulkan/vk_master_semaphore.cpp
renderer_vulkan/vk_master_semaphore.h renderer_vulkan/vk_master_semaphore.h
renderer_vulkan/vk_pipeline_cache.cpp renderer_vulkan/vk_pipeline_cache.cpp
+9 -117
View File
@@ -112,13 +112,6 @@ void BufferCache<P>::TickFrame() {
async_buffers_death_ring.clear(); async_buffers_death_ring.clear();
} }
template <class P>
void BufferCache<P>::UnmapGPUMemory(size_t as_id, GPUVAddr gpu_addr, size_t size) {
if constexpr (requires { runtime.BindMultiRangeStorageBuffer(u64{}); }) {
virtual_ranges.Unmap(as_id, gpu_addr, size);
}
}
template <class P> template <class P>
void BufferCache<P>::WriteMemory(DAddr device_addr, u64 size) { void BufferCache<P>::WriteMemory(DAddr device_addr, u64 size) {
if (memory_tracker.IsRegionGpuModified(device_addr, size)) { if (memory_tracker.IsRegionGpuModified(device_addr, size)) {
@@ -1005,85 +998,11 @@ void BufferCache<P>::BindHostGraphicsUniformBuffer(size_t stage, u32 index, u32
channel_state->fast_bound_uniform_buffers[stage] &= ~(1u << binding_index); channel_state->fast_bound_uniform_buffers[stage] &= ~(1u << binding_index);
} }
template <class P>
void BufferCache<P>::ResolveMultiRangeStorage(Binding& binding, bool is_written,
std::vector<MultiRangeSegment>& pool) {
binding.segment_first = 0;
binding.segment_count = 0;
if constexpr (requires { runtime.BindMultiRangeStorageBuffer(u64{}); }) {
if (binding.gpu_addr == 0 || binding.size == 0) {
return;
}
if (is_written && !runtime.PrefersSparseSources()) {
return;
}
const VirtualSegments* found =
virtual_ranges.Query(*gpu_memory, binding.gpu_addr, binding.size);
if (!found || found->size() < 2) {
return;
}
const VirtualSegments segments = *found;
const u32 first = static_cast<u32>(pool.size());
const bool prefer_sparse = runtime.PrefersSparseSources();
for (const VirtualSegment& segment : segments) {
const BufferId buffer_id =
FindBuffer(segment.device_addr, segment.size, prefer_sparse);
if (!buffer_id) {
pool.resize(first);
return;
}
pool.push_back(MultiRangeSegment{
.buffer_id = buffer_id,
.device_addr = segment.device_addr,
.size = segment.size,
});
}
binding.segment_first = first;
binding.segment_count = static_cast<u32>(segments.size());
}
}
template <class P>
bool BufferCache<P>::BindMultiRangeStorage(const Binding& binding, bool is_written,
std::span<const MultiRangeSegment> pool) {
if constexpr (requires { runtime.BindMultiRangeStorageBuffer(u64{}); }) {
if (binding.segment_count < 2) {
return false;
}
if (binding.segment_first + binding.segment_count > pool.size()) {
return false;
}
const u64 key = (static_cast<u64>(gpu_memory->GetID()) << 48) ^ binding.gpu_addr;
runtime.ResetMultiRange();
for (u32 index = 0; index < binding.segment_count; ++index) {
const MultiRangeSegment& segment = pool[binding.segment_first + index];
Buffer& buffer = slot_buffers[segment.buffer_id];
TouchBuffer(buffer, segment.buffer_id);
if (SynchronizeBuffer(buffer, segment.device_addr, segment.size)) {
runtime.InvalidateMultiRange(key);
}
const u32 offset = buffer.Offset(segment.device_addr);
buffer.MarkUsage(offset, segment.size);
if (is_written) {
MarkWrittenBuffer(segment.buffer_id, segment.device_addr, segment.size);
}
runtime.PushMultiRangeSource(buffer, offset, segment.size);
}
return runtime.BindMultiRangeStorageBuffer(key);
} else {
return false;
}
}
template <class P> template <class P>
void BufferCache<P>::BindHostGraphicsStorageBuffers(size_t stage) { void BufferCache<P>::BindHostGraphicsStorageBuffers(size_t stage) {
u32 binding_index = 0; u32 binding_index = 0;
ForEachEnabledBit(channel_state->enabled_storage_buffers[stage], [&](u32 index) { ForEachEnabledBit(channel_state->enabled_storage_buffers[stage], [&](u32 index) {
const Binding& binding = channel_state->storage_buffers[stage][index]; const Binding& binding = channel_state->storage_buffers[stage][index];
const bool is_written = ((channel_state->written_storage_buffers[stage] >> index) & 1) != 0;
if (BindMultiRangeStorage(binding, is_written, graphics_segments)) {
return;
}
Buffer& buffer = slot_buffers[binding.buffer_id]; Buffer& buffer = slot_buffers[binding.buffer_id];
TouchBuffer(buffer, binding.buffer_id); TouchBuffer(buffer, binding.buffer_id);
const u32 size = binding.size; const u32 size = binding.size;
@@ -1091,6 +1010,7 @@ void BufferCache<P>::BindHostGraphicsStorageBuffers(size_t stage) {
const u32 offset = buffer.Offset(binding.device_addr); const u32 offset = buffer.Offset(binding.device_addr);
buffer.MarkUsage(offset, size); buffer.MarkUsage(offset, size);
const bool is_written = ((channel_state->written_storage_buffers[stage] >> index) & 1) != 0;
if (is_written) { if (is_written) {
MarkWrittenBuffer(binding.buffer_id, binding.device_addr, size); MarkWrittenBuffer(binding.buffer_id, binding.device_addr, size);
@@ -1219,11 +1139,6 @@ void BufferCache<P>::BindHostComputeStorageBuffers() {
u32 binding_index = 0; u32 binding_index = 0;
ForEachEnabledBit(channel_state->enabled_compute_storage_buffers, [&](u32 index) { ForEachEnabledBit(channel_state->enabled_compute_storage_buffers, [&](u32 index) {
const Binding& binding = channel_state->compute_storage_buffers[index]; const Binding& binding = channel_state->compute_storage_buffers[index];
const bool is_written =
((channel_state->written_compute_storage_buffers >> index) & 1) != 0;
if (BindMultiRangeStorage(binding, is_written, compute_segments)) {
return;
}
Buffer& buffer = slot_buffers[binding.buffer_id]; Buffer& buffer = slot_buffers[binding.buffer_id];
TouchBuffer(buffer, binding.buffer_id); TouchBuffer(buffer, binding.buffer_id);
const u32 size = binding.size; const u32 size = binding.size;
@@ -1231,6 +1146,8 @@ void BufferCache<P>::BindHostComputeStorageBuffers() {
const u32 offset = buffer.Offset(binding.device_addr); const u32 offset = buffer.Offset(binding.device_addr);
buffer.MarkUsage(offset, size); buffer.MarkUsage(offset, size);
const bool is_written =
((channel_state->written_compute_storage_buffers >> index) & 1) != 0;
if (is_written) { if (is_written) {
MarkWrittenBuffer(binding.buffer_id, binding.device_addr, size); MarkWrittenBuffer(binding.buffer_id, binding.device_addr, size);
@@ -1276,7 +1193,6 @@ void BufferCache<P>::BindHostComputeTextureBuffers() {
template <class P> template <class P>
void BufferCache<P>::DoUpdateGraphicsBuffers(bool is_indexed) { void BufferCache<P>::DoUpdateGraphicsBuffers(bool is_indexed) {
graphics_segments.clear();
BufferOperations([&]() { BufferOperations([&]() {
if (is_indexed) { if (is_indexed) {
UpdateIndexBuffer(); UpdateIndexBuffer();
@@ -1296,7 +1212,6 @@ void BufferCache<P>::DoUpdateGraphicsBuffers(bool is_indexed) {
template <class P> template <class P>
void BufferCache<P>::DoUpdateComputeBuffers() { void BufferCache<P>::DoUpdateComputeBuffers() {
compute_segments.clear();
BufferOperations([&]() { BufferOperations([&]() {
UpdateComputeUniformBuffers(); UpdateComputeUniformBuffers();
UpdateComputeStorageBuffers(); UpdateComputeStorageBuffers();
@@ -1439,8 +1354,6 @@ void BufferCache<P>::UpdateStorageBuffers(size_t stage) {
Binding& binding = channel_state->storage_buffers[stage][index]; Binding& binding = channel_state->storage_buffers[stage][index];
const BufferId buffer_id = FindBuffer(binding.device_addr, binding.size); const BufferId buffer_id = FindBuffer(binding.device_addr, binding.size);
binding.buffer_id = buffer_id; binding.buffer_id = buffer_id;
const bool is_written = ((channel_state->written_storage_buffers[stage] >> index) & 1) != 0;
ResolveMultiRangeStorage(binding, is_written, graphics_segments);
}); });
} }
@@ -1504,9 +1417,6 @@ void BufferCache<P>::UpdateComputeStorageBuffers() {
// Resolve buffer // Resolve buffer
Binding& binding = channel_state->compute_storage_buffers[index]; Binding& binding = channel_state->compute_storage_buffers[index];
binding.buffer_id = FindBuffer(binding.device_addr, binding.size); binding.buffer_id = FindBuffer(binding.device_addr, binding.size);
const bool is_written =
((channel_state->written_compute_storage_buffers >> index) & 1) != 0;
ResolveMultiRangeStorage(binding, is_written, compute_segments);
}); });
} }
@@ -1530,7 +1440,7 @@ void BufferCache<P>::MarkWrittenBuffer(BufferId buffer_id, DAddr device_addr, u3
} }
template <class P> template <class P>
BufferId BufferCache<P>::FindBuffer(DAddr device_addr, u32 size, bool sparse_compatible) { BufferId BufferCache<P>::FindBuffer(DAddr device_addr, u32 size) {
if (device_addr == 0) { if (device_addr == 0) {
return NULL_BUFFER_ID; return NULL_BUFFER_ID;
} }
@@ -1540,18 +1450,10 @@ BufferId BufferCache<P>::FindBuffer(DAddr device_addr, u32 size, bool sparse_com
Buffer& buffer = slot_buffers[buffer_id]; Buffer& buffer = slot_buffers[buffer_id];
WaitForGpuFenceIfNeeded(buffer); WaitForGpuFenceIfNeeded(buffer);
if (buffer.IsInBounds(device_addr, size)) { if (buffer.IsInBounds(device_addr, size)) {
bool usable = true; return buffer_id;
if constexpr (requires { buffer.IsSparseCompatible(); }) {
if (sparse_compatible && !buffer.IsSparseCompatible()) {
usable = false;
}
}
if (usable) {
return buffer_id;
}
} }
} }
return CreateBuffer(device_addr, size, sparse_compatible); return CreateBuffer(device_addr, size);
} }
template <class P> template <class P>
@@ -1673,15 +1575,13 @@ void BufferCache<P>::JoinOverlap(BufferId new_buffer_id, BufferId overlap_id,
} }
template <class P> template <class P>
BufferId BufferCache<P>::CreateBuffer(DAddr device_addr, u32 wanted_size, BufferId BufferCache<P>::CreateBuffer(DAddr device_addr, u32 wanted_size) {
bool sparse_compatible) {
DAddr device_addr_end = Common::AlignUp(device_addr + wanted_size, CACHING_PAGESIZE); DAddr device_addr_end = Common::AlignUp(device_addr + wanted_size, CACHING_PAGESIZE);
device_addr = Common::AlignDown(device_addr, CACHING_PAGESIZE); device_addr = Common::AlignDown(device_addr, CACHING_PAGESIZE);
wanted_size = static_cast<u32>(device_addr_end - device_addr); wanted_size = static_cast<u32>(device_addr_end - device_addr);
const OverlapResult overlap = ResolveOverlaps(device_addr, wanted_size); const OverlapResult overlap = ResolveOverlaps(device_addr, wanted_size);
const u32 size = static_cast<u32>(overlap.end - overlap.begin); const u32 size = static_cast<u32>(overlap.end - overlap.begin);
const BufferId new_buffer_id = const BufferId new_buffer_id = slot_buffers.insert(runtime, overlap.begin, size);
slot_buffers.insert(runtime, overlap.begin, size, sparse_compatible);
auto& new_buffer = slot_buffers[new_buffer_id]; auto& new_buffer = slot_buffers[new_buffer_id];
const size_t size_bytes = new_buffer.SizeBytes(); const size_t size_bytes = new_buffer.SizeBytes();
runtime.ClearBuffer(new_buffer, 0, size_bytes, 0); runtime.ClearBuffer(new_buffer, 0, size_bytes, 0);
@@ -1931,9 +1831,6 @@ void BufferCache<P>::DownloadBufferMemory(Buffer& buffer, DAddr device_addr, u64
template <class P> template <class P>
void BufferCache<P>::DeleteBuffer(BufferId buffer_id, bool do_not_mark) { void BufferCache<P>::DeleteBuffer(BufferId buffer_id, bool do_not_mark) {
if constexpr (requires { runtime.OnBufferDeleted(slot_buffers[buffer_id]); }) {
runtime.OnBufferDeleted(slot_buffers[buffer_id]);
}
bool dirty_index{false}; bool dirty_index{false};
boost::container::small_vector<u64, NUM_VERTEX_BUFFERS> dirty_vertex_buffers; boost::container::small_vector<u64, NUM_VERTEX_BUFFERS> dirty_vertex_buffers;
const auto scalar_replace = [buffer_id](Binding& binding) { const auto scalar_replace = [buffer_id](Binding& binding) {
@@ -2037,15 +1934,10 @@ Binding BufferCache<P>::StorageBufferBinding(GPUVAddr ssbo_addr, u32 cbuf_index,
// The end address used for size calculation does not need to be aligned // The end address used for size calculation does not need to be aligned
const DAddr cpu_end = Common::AlignUp(*device_addr + size, Core::DEVICE_PAGESIZE); const DAddr cpu_end = Common::AlignUp(*device_addr + size, Core::DEVICE_PAGESIZE);
u32 binding_size = static_cast<u32>(cpu_end - *aligned_device_addr);
if (is_written) {
binding_size = aligned_size;
}
const Binding binding{ const Binding binding{
.device_addr = *aligned_device_addr, .device_addr = *aligned_device_addr,
.size = binding_size, .size = is_written ? aligned_size : static_cast<u32>(cpu_end - *aligned_device_addr),
.buffer_id = BufferId{}, .buffer_id = BufferId{},
.gpu_addr = aligned_gpu_addr,
}; };
return binding; return binding;
} }
@@ -29,7 +29,6 @@
#include "common/settings.h" #include "common/settings.h"
#include "common/slot_vector.h" #include "common/slot_vector.h"
#include "video_core/buffer_cache/buffer_base.h" #include "video_core/buffer_cache/buffer_base.h"
#include "video_core/buffer_cache/virtual_range_cache.h"
#include "video_core/control/channel_state_cache.h" #include "video_core/control/channel_state_cache.h"
#include "video_core/delayed_destruction_ring.h" #include "video_core/delayed_destruction_ring.h"
#include "video_core/dirty_flags.h" #include "video_core/dirty_flags.h"
@@ -84,15 +83,6 @@ struct Binding {
DAddr device_addr{}; DAddr device_addr{};
u32 size{}; u32 size{};
BufferId buffer_id; BufferId buffer_id;
GPUVAddr gpu_addr{};
u32 segment_first{};
u32 segment_count{};
};
struct MultiRangeSegment {
BufferId buffer_id;
DAddr device_addr{};
u32 size{};
}; };
struct TextureBufferBinding : Binding { struct TextureBufferBinding : Binding {
@@ -225,14 +215,6 @@ public:
void TickFrame(); void TickFrame();
bool BindMultiRangeStorage(const Binding& binding, bool is_written,
std::span<const MultiRangeSegment> pool);
void ResolveMultiRangeStorage(Binding& binding, bool is_written,
std::vector<MultiRangeSegment>& pool);
void UnmapGPUMemory(size_t as_id, GPUVAddr gpu_addr, size_t size);
void WriteMemory(DAddr device_addr, u64 size); void WriteMemory(DAddr device_addr, u64 size);
void CachedWriteMemory(DAddr device_addr, u64 size); void CachedWriteMemory(DAddr device_addr, u64 size);
@@ -432,8 +414,7 @@ private:
void MarkWrittenBuffer(BufferId buffer_id, DAddr device_addr, u32 size); void MarkWrittenBuffer(BufferId buffer_id, DAddr device_addr, u32 size);
[[nodiscard]] BufferId FindBuffer(DAddr device_addr, u32 size, [[nodiscard]] BufferId FindBuffer(DAddr device_addr, u32 size);
bool sparse_compatible = false);
void WaitForGpuFenceIfNeeded(Buffer& buffer); void WaitForGpuFenceIfNeeded(Buffer& buffer);
@@ -441,8 +422,7 @@ private:
void JoinOverlap(BufferId new_buffer_id, BufferId overlap_id, bool accumulate_stream_score); void JoinOverlap(BufferId new_buffer_id, BufferId overlap_id, bool accumulate_stream_score);
[[nodiscard]] BufferId CreateBuffer(DAddr device_addr, u32 wanted_size, [[nodiscard]] BufferId CreateBuffer(DAddr device_addr, u32 wanted_size);
bool sparse_compatible = false);
void Register(BufferId buffer_id); void Register(BufferId buffer_id);
@@ -534,9 +514,6 @@ private:
}; };
Common::LeastRecentlyUsedCache<LRUItemParams> lru_cache; Common::LeastRecentlyUsedCache<LRUItemParams> lru_cache;
u64 frame_tick = 0; u64 frame_tick = 0;
VirtualRangeCache virtual_ranges;
std::vector<MultiRangeSegment> graphics_segments;
std::vector<MultiRangeSegment> compute_segments;
u64 total_used_memory = 0; u64 total_used_memory = 0;
u64 minimum_memory = 0; u64 minimum_memory = 0;
u64 critical_memory = 0; u64 critical_memory = 0;
@@ -1,163 +0,0 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
#pragma once
#include <atomic>
#include <limits>
#include <mutex>
#include <optional>
#include <unordered_map>
#include <vector>
#include <boost/container/small_vector.hpp>
#include "common/common_types.h"
#include "video_core/memory_manager.h"
namespace VideoCommon {
struct VirtualSegment {
GPUVAddr gpu_addr;
DAddr device_addr;
u32 size;
};
using VirtualSegments = boost::container::small_vector<VirtualSegment, 8>;
class VirtualRangeCache {
public:
const VirtualSegments* Query(Tegra::MemoryManager& memory, GPUVAddr gpu_addr, u32 size) {
if (has_deferred.load(std::memory_order_acquire)) {
ApplyDeferred();
}
const size_t as_id = memory.GetID();
const u64 key = MakeKey(as_id, gpu_addr);
const auto it = entries.find(key);
if (it != entries.end() && it->second.as_id == as_id &&
it->second.gpu_addr == gpu_addr && it->second.size == size) {
return &it->second.segments;
}
Entry entry;
entry.as_id = as_id;
entry.gpu_addr = gpu_addr;
entry.size = size;
const auto ranges = memory.GetSubmappedRange(gpu_addr, size);
GPUVAddr expected = gpu_addr;
bool contiguous = true;
for (const auto& [range_addr, range_size] : ranges) {
if (range_addr != expected || range_size == 0) {
contiguous = false;
break;
}
const std::optional<DAddr> device_addr = memory.GpuToCpuAddress(range_addr);
if (!device_addr || *device_addr == 0) {
contiguous = false;
break;
}
if (range_size > static_cast<size_t>((std::numeric_limits<u32>::max)())) {
contiguous = false;
break;
}
entry.segments.push_back(VirtualSegment{
.gpu_addr = range_addr,
.device_addr = *device_addr,
.size = static_cast<u32>(range_size),
});
expected += range_size;
}
if (!contiguous || expected != gpu_addr + size) {
entry.segments.clear();
}
const auto result = entries.insert_or_assign(key, std::move(entry));
return &result.first->second.segments;
}
void Unmap(size_t as_id, GPUVAddr gpu_addr, u64 size) {
if (size == 0) {
return;
}
{
std::scoped_lock lock{deferred_mutex};
if (!deferred.empty()) {
DeferredUnmap& last = deferred.back();
if (last.as_id == as_id && last.gpu_addr + last.size == gpu_addr) {
last.size += size;
has_deferred.store(true, std::memory_order_release);
return;
}
}
deferred.push_back(DeferredUnmap{
.as_id = as_id,
.gpu_addr = gpu_addr,
.size = size,
});
}
has_deferred.store(true, std::memory_order_release);
}
void Clear() {
{
std::scoped_lock lock{deferred_mutex};
deferred.clear();
}
has_deferred.store(false, std::memory_order_release);
entries.clear();
}
private:
struct Entry {
size_t as_id{};
GPUVAddr gpu_addr{};
u32 size{};
VirtualSegments segments;
};
struct DeferredUnmap {
size_t as_id;
GPUVAddr gpu_addr;
u64 size;
};
static u64 MakeKey(size_t as_id, GPUVAddr gpu_addr) {
return (static_cast<u64>(as_id) << 48) ^ gpu_addr;
}
void ApplyDeferred() {
std::vector<DeferredUnmap> pending;
{
std::scoped_lock lock{deferred_mutex};
has_deferred.store(false, std::memory_order_release);
pending.swap(deferred);
}
if (pending.empty() || entries.empty()) {
return;
}
for (auto it = entries.begin(); it != entries.end();) {
const Entry& entry = it->second;
const GPUVAddr entry_end = entry.gpu_addr + entry.size;
bool overlaps = false;
for (const DeferredUnmap& unmap : pending) {
if (unmap.as_id != entry.as_id) {
continue;
}
if (entry.gpu_addr < unmap.gpu_addr + unmap.size && unmap.gpu_addr < entry_end) {
overlaps = true;
break;
}
}
if (overlaps) {
it = entries.erase(it);
} else {
++it;
}
}
}
std::unordered_map<u64, Entry> entries;
std::vector<DeferredUnmap> deferred;
std::mutex deferred_mutex;
std::atomic<bool> has_deferred{false};
};
} // namespace VideoCommon
@@ -52,7 +52,7 @@ constexpr std::array PROGRAM_LUT{
Buffer::Buffer(BufferCacheRuntime&, VideoCommon::NullBufferParams null_params) Buffer::Buffer(BufferCacheRuntime&, VideoCommon::NullBufferParams null_params)
: VideoCommon::BufferBase(null_params) {} : VideoCommon::BufferBase(null_params) {}
Buffer::Buffer(BufferCacheRuntime& runtime, DAddr cpu_addr_, u64 size_bytes_, bool) Buffer::Buffer(BufferCacheRuntime& runtime, DAddr cpu_addr_, u64 size_bytes_)
: VideoCommon::BufferBase(cpu_addr_, size_bytes_) { : VideoCommon::BufferBase(cpu_addr_, size_bytes_) {
buffer.Create(); buffer.Create();
if (runtime.device.HasDebuggingToolAttached()) { if (runtime.device.HasDebuggingToolAttached()) {
@@ -23,8 +23,7 @@ class BufferCacheRuntime;
class Buffer : public VideoCommon::BufferBase { class Buffer : public VideoCommon::BufferBase {
public: public:
explicit Buffer(BufferCacheRuntime&, DAddr cpu_addr, u64 size_bytes, explicit Buffer(BufferCacheRuntime&, DAddr cpu_addr, u64 size_bytes);
bool sparse_compatible = false);
explicit Buffer(BufferCacheRuntime&, VideoCommon::NullBufferParams); explicit Buffer(BufferCacheRuntime&, VideoCommon::NullBufferParams);
void ImmediateUpload(size_t offset, std::span<const u8> data) noexcept; void ImmediateUpload(size_t offset, std::span<const u8> data) noexcept;
@@ -56,8 +56,7 @@ size_t BytesPerIndex(VkIndexType index_type) {
} }
} }
vk::Buffer CreateBuffer(const Device& device, const MemoryAllocator& memory_allocator, u64 size, vk::Buffer CreateBuffer(const Device& device, const MemoryAllocator& memory_allocator, u64 size) {
VkDeviceSize sparse_alignment) {
VkBufferUsageFlags flags = VkBufferUsageFlags flags =
VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT | VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT |
VK_BUFFER_USAGE_UNIFORM_TEXEL_BUFFER_BIT | VK_BUFFER_USAGE_STORAGE_TEXEL_BUFFER_BIT | VK_BUFFER_USAGE_UNIFORM_TEXEL_BUFFER_BIT | VK_BUFFER_USAGE_STORAGE_TEXEL_BUFFER_BIT |
@@ -83,9 +82,6 @@ vk::Buffer CreateBuffer(const Device& device, const MemoryAllocator& memory_allo
.queueFamilyIndexCount = 0, .queueFamilyIndexCount = 0,
.pQueueFamilyIndices = nullptr, .pQueueFamilyIndices = nullptr,
}; };
if (sparse_alignment > 1) {
return memory_allocator.CreateBuffer(buffer_ci, MemoryUsage::DeviceLocal, sparse_alignment);
}
return memory_allocator.CreateBuffer(buffer_ci, MemoryUsage::DeviceLocal); return memory_allocator.CreateBuffer(buffer_ci, MemoryUsage::DeviceLocal);
} }
} // Anonymous namespace } // Anonymous namespace
@@ -103,14 +99,10 @@ Buffer::Buffer(BufferCacheRuntime& runtime, VideoCommon::NullBufferParams null_p
} }
} }
Buffer::Buffer(BufferCacheRuntime& runtime, DAddr cpu_addr_, u64 size_bytes_, Buffer::Buffer(BufferCacheRuntime& runtime, DAddr cpu_addr_, u64 size_bytes_)
bool sparse_compatible_)
: VideoCommon::BufferBase(cpu_addr_, size_bytes_), device{&runtime.device}, : VideoCommon::BufferBase(cpu_addr_, size_bytes_), device{&runtime.device},
scheduler{&runtime.scheduler}, scheduler{&runtime.scheduler},
buffer{CreateBuffer(*device, runtime.memory_allocator, SizeBytes(), buffer{CreateBuffer(*device, runtime.memory_allocator, SizeBytes())}, tracker{SizeBytes()} {
runtime.SparseAlignmentFor(sparse_compatible_))},
tracker{SizeBytes()} {
sparse_compatible = sparse_compatible_;
if (runtime.device.HasDebuggingToolAttached()) { if (runtime.device.HasDebuggingToolAttached()) {
buffer.SetObjectNameEXT(fmt::format("Buffer {:#x}", CpuAddr()).c_str()); buffer.SetObjectNameEXT(fmt::format("Buffer {:#x}", CpuAddr()).c_str());
} }
@@ -356,8 +348,7 @@ BufferCacheRuntime::BufferCacheRuntime(const Device& device_, MemoryAllocator& m
: device{device_}, memory_allocator{memory_allocator_}, scheduler{scheduler_}, : device{device_}, memory_allocator{memory_allocator_}, scheduler{scheduler_},
staging_pool{staging_pool_}, guest_descriptor_queue{guest_descriptor_queue_}, staging_pool{staging_pool_}, guest_descriptor_queue{guest_descriptor_queue_},
quad_index_pass(device, scheduler, descriptor_pool, staging_pool, quad_index_pass(device, scheduler, descriptor_pool, staging_pool,
compute_pass_descriptor_queue), compute_pass_descriptor_queue) {
multi_range_buffers(device_, memory_allocator_, scheduler_) {
const VkDriverIdKHR driver_id = device.GetDriverID(); const VkDriverIdKHR driver_id = device.GetDriverID();
limit_dynamic_storage_buffers = driver_id == VK_DRIVER_ID_QUALCOMM_PROPRIETARY || limit_dynamic_storage_buffers = driver_id == VK_DRIVER_ID_QUALCOMM_PROPRIETARY ||
driver_id == VK_DRIVER_ID_ARM_PROPRIETARY; driver_id == VK_DRIVER_ID_ARM_PROPRIETARY;
@@ -545,33 +536,6 @@ void BufferCacheRuntime::ClearBuffer(VkBuffer dest_buffer, u32 offset, size_t si
}); });
} }
bool BufferCacheRuntime::BindMultiRangeStorageBuffer(u64 key) {
if (multi_range_sources.empty() || multi_range_total == 0) {
return false;
}
const MultiRangeRef ref = multi_range_buffers.Get(key, multi_range_sources, multi_range_total);
if (ref.handle == VK_NULL_HANDLE) {
return false;
}
if (ref.needs_gather) {
PreCopyBarrier();
VkDeviceSize dst_offset = 0;
for (const MultiRangeSource& source : multi_range_sources) {
const std::array<VideoCommon::BufferCopy, 1> copy{VideoCommon::BufferCopy{
.src_offset = static_cast<u64>(source.offset),
.dst_offset = static_cast<u64>(dst_offset),
.size = static_cast<size_t>(source.size),
}};
CopyBuffer(ref.handle, source.handle, copy, false);
dst_offset += source.size;
}
PostCopyBarrier();
multi_range_buffers.MarkGathered(key);
}
guest_descriptor_queue.AddBuffer(ref.handle, ref.address, 0, ref.size);
return true;
}
void BufferCacheRuntime::BindIndexBuffer(PrimitiveTopology topology, IndexFormat index_format, void BufferCacheRuntime::BindIndexBuffer(PrimitiveTopology topology, IndexFormat index_format,
u32 base_vertex, u32 num_indices, VkBuffer buffer, u32 base_vertex, u32 num_indices, VkBuffer buffer,
u32 offset, [[maybe_unused]] u32 size) { u32 offset, [[maybe_unused]] u32 size) {
@@ -8,14 +8,11 @@
#include <limits> #include <limits>
#include <boost/container/small_vector.hpp>
#include "video_core/buffer_cache/buffer_cache_base.h" #include "video_core/buffer_cache/buffer_cache_base.h"
#include "video_core/buffer_cache/memory_tracker_base.h" #include "video_core/buffer_cache/memory_tracker_base.h"
#include "video_core/buffer_cache/usage_tracker.h" #include "video_core/buffer_cache/usage_tracker.h"
#include "video_core/engines/maxwell_3d.h" #include "video_core/engines/maxwell_3d.h"
#include "video_core/renderer_vulkan/vk_compute_pass.h" #include "video_core/renderer_vulkan/vk_compute_pass.h"
#include "video_core/renderer_vulkan/vk_multi_range_buffer.h"
#include "video_core/renderer_vulkan/vk_staging_buffer_pool.h" #include "video_core/renderer_vulkan/vk_staging_buffer_pool.h"
#include "video_core/renderer_vulkan/vk_update_descriptor.h" #include "video_core/renderer_vulkan/vk_update_descriptor.h"
#include "video_core/surface.h" #include "video_core/surface.h"
@@ -34,8 +31,7 @@ class BufferCacheRuntime;
class Buffer : public VideoCommon::BufferBase { class Buffer : public VideoCommon::BufferBase {
public: public:
explicit Buffer(BufferCacheRuntime&, VideoCommon::NullBufferParams null_params); explicit Buffer(BufferCacheRuntime&, VideoCommon::NullBufferParams null_params);
explicit Buffer(BufferCacheRuntime& runtime, VAddr cpu_addr_, u64 size_bytes_, explicit Buffer(BufferCacheRuntime& runtime, VAddr cpu_addr_, u64 size_bytes_);
bool sparse_compatible_ = false);
[[nodiscard]] VkBufferView View(u32 offset, u32 size, VideoCore::Surface::PixelFormat format); [[nodiscard]] VkBufferView View(u32 offset, u32 size, VideoCore::Surface::PixelFormat format);
@@ -47,14 +43,6 @@ public:
return device_address; return device_address;
} }
[[nodiscard]] bool IsSparseCompatible() const noexcept {
return sparse_compatible;
}
[[nodiscard]] vk::MemoryLocation Location() const noexcept {
return buffer.Location();
}
[[nodiscard]] bool IsRegionUsed(u64 offset, u64 size) const noexcept { [[nodiscard]] bool IsRegionUsed(u64 offset, u64 size) const noexcept {
return tracker.IsUsed(offset, size); return tracker.IsUsed(offset, size);
} }
@@ -89,7 +77,6 @@ private:
VkDeviceAddress device_address{}; VkDeviceAddress device_address{};
u64 last_usage_tick{}; u64 last_usage_tick{};
bool is_null{}; bool is_null{};
bool sparse_compatible{};
}; };
class QuadArrayIndexBuffer; class QuadArrayIndexBuffer;
@@ -138,7 +125,7 @@ public:
void PreCopyBarrier(); void PreCopyBarrier();
void CopyBuffer(VkBuffer dst_buffer, VkBuffer src_buffer, void CopyBuffer(VkBuffer src_buffer, VkBuffer dst_buffer,
std::span<const VideoCommon::BufferCopy> copies, bool barrier, std::span<const VideoCommon::BufferCopy> copies, bool barrier,
bool can_reorder_upload = false); bool can_reorder_upload = false);
@@ -168,45 +155,6 @@ public:
return ref.mapped_span; return ref.mapped_span;
} }
[[nodiscard]] VkDeviceSize SparseAlignmentFor(bool sparse_compatible) const noexcept {
if (!sparse_compatible || !multi_range_buffers.UsesSparse()) {
return 0;
}
return multi_range_buffers.BlockSize();
}
[[nodiscard]] bool PrefersSparseSources() const noexcept {
return multi_range_buffers.UsesSparse();
}
void ResetMultiRange() noexcept {
multi_range_sources.clear();
multi_range_total = 0;
}
void PushMultiRangeSource(const Buffer& buffer, u32 offset, u32 size) {
const vk::MemoryLocation location = buffer.Location();
multi_range_sources.push_back(MultiRangeSource{
.handle = buffer.Handle(),
.memory = location.memory,
.memory_offset = location.offset,
.offset = offset,
.size = size,
.memory_type = location.memory_type,
});
multi_range_total += size;
}
bool BindMultiRangeStorageBuffer(u64 key);
void InvalidateMultiRange(u64 key) {
multi_range_buffers.Invalidate(key);
}
void OnBufferDeleted(const Buffer& buffer) {
multi_range_buffers.DropOwner(buffer.Handle());
}
void BindUniformBuffer(const Buffer& buffer, u32 offset, u32 size) { void BindUniformBuffer(const Buffer& buffer, u32 offset, u32 size) {
BindBuffer(buffer, offset, size); BindBuffer(buffer, offset, size);
} }
@@ -260,10 +208,6 @@ private:
std::unique_ptr<Uint8Pass> uint8_pass; std::unique_ptr<Uint8Pass> uint8_pass;
QuadIndexedPass quad_index_pass; QuadIndexedPass quad_index_pass;
MultiRangeBufferCache multi_range_buffers;
boost::container::small_vector<MultiRangeSource, 16> multi_range_sources;
VkDeviceSize multi_range_total{};
bool limit_dynamic_storage_buffers = false; bool limit_dynamic_storage_buffers = false;
u32 max_dynamic_storage_buffers = (std::numeric_limits<u32>::max)(); u32 max_dynamic_storage_buffers = (std::numeric_limits<u32>::max)();
}; };
@@ -1,341 +0,0 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
#include <mutex>
#include "video_core/renderer_vulkan/vk_multi_range_buffer.h"
#include "video_core/renderer_vulkan/vk_scheduler.h"
#include "video_core/vulkan_common/vulkan_device.h"
namespace Vulkan {
MultiRangeBufferCache::MultiRangeBufferCache(const Device& device_,
MemoryAllocator& memory_allocator_,
Scheduler& scheduler_)
: device{device_}, memory_allocator{memory_allocator_}, scheduler{scheduler_} {
sparse_usage = VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT |
VK_BUFFER_USAGE_STORAGE_BUFFER_BIT;
if (device.IsBufferDeviceAddressSupported()) {
sparse_usage |= VK_BUFFER_USAGE_SHADER_DEVICE_ADDRESS_BIT;
}
if (!device.IsSparseBindingSupported()) {
return;
}
u32 memory_type_bits = 0;
const VkDeviceSize queried = QueryBlockSize(memory_type_bits);
if (queried == 0 || memory_type_bits == 0) {
return;
}
block_size = queried;
sparse_memory_type_bits = memory_type_bits;
use_sparse = true;
}
MultiRangeBufferCache::~MultiRangeBufferCache() {
const VkDevice logical = *device.GetLogical();
const auto& dld = device.GetDispatchLoader();
for (auto& [key, entry] : entries) {
if (entry.sparse_handle != VK_NULL_HANDLE) {
dld.vkDestroyBuffer(logical, entry.sparse_handle, nullptr);
}
}
entries.clear();
for (const Retired& item : retired) {
dld.vkDestroyBuffer(logical, item.handle, nullptr);
}
retired.clear();
}
VkDeviceSize MultiRangeBufferCache::QueryBlockSize(u32& memory_type_bits) const {
const VkDevice logical = *device.GetLogical();
const auto& dld = device.GetDispatchLoader();
const VkBufferCreateInfo probe_ci{
.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
.pNext = nullptr,
.flags = VK_BUFFER_CREATE_SPARSE_BINDING_BIT | VK_BUFFER_CREATE_SPARSE_ALIASED_BIT,
.size = DEFAULT_BLOCK_SIZE,
.usage = sparse_usage,
.sharingMode = VK_SHARING_MODE_EXCLUSIVE,
.queueFamilyIndexCount = 0,
.pQueueFamilyIndices = nullptr,
};
VkBuffer probe{};
if (dld.vkCreateBuffer(logical, &probe_ci, nullptr, &probe) != VK_SUCCESS) {
return 0;
}
const VkBufferMemoryRequirementsInfo2 reqs_info{
.sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_REQUIREMENTS_INFO_2,
.pNext = nullptr,
.buffer = probe,
};
VkMemoryRequirements2 reqs2{
.sType = VK_STRUCTURE_TYPE_MEMORY_REQUIREMENTS_2,
.pNext = nullptr,
.memoryRequirements = {},
};
dld.vkGetBufferMemoryRequirements2(logical, &reqs_info, &reqs2);
dld.vkDestroyBuffer(logical, probe, nullptr);
memory_type_bits = reqs2.memoryRequirements.memoryTypeBits;
return reqs2.memoryRequirements.alignment;
}
u64 MultiRangeBufferCache::HashSources(std::span<const MultiRangeSource> sources) const {
u64 hash = 0xcbf29ce484222325ULL;
const auto mix = [&hash](u64 value) {
hash ^= value;
hash *= 0x100000001b3ULL;
};
for (const MultiRangeSource& source : sources) {
mix(reinterpret_cast<u64>(source.handle));
mix(static_cast<u64>(source.offset));
mix(static_cast<u64>(source.size));
}
return hash;
}
bool MultiRangeBufferCache::CanBindSparse(std::span<const MultiRangeSource> sources) const {
if (!UsesSparse()) {
return false;
}
for (const MultiRangeSource& source : sources) {
if (source.memory == VK_NULL_HANDLE) {
return false;
}
if (source.memory_type >= 32) {
return false;
}
if (((sparse_memory_type_bits >> source.memory_type) & 1) == 0) {
return false;
}
const VkDeviceSize memory_offset = source.memory_offset + source.offset;
if ((memory_offset % block_size) != 0) {
return false;
}
if ((source.size % block_size) != 0) {
return false;
}
}
return true;
}
VkBuffer MultiRangeBufferCache::CreateSparse(std::span<const MultiRangeSource> sources,
VkDeviceSize total) {
const VkDevice logical = *device.GetLogical();
const auto& dld = device.GetDispatchLoader();
const VkBufferCreateInfo buffer_ci{
.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
.pNext = nullptr,
.flags = VK_BUFFER_CREATE_SPARSE_BINDING_BIT | VK_BUFFER_CREATE_SPARSE_ALIASED_BIT,
.size = total,
.usage = sparse_usage,
.sharingMode = VK_SHARING_MODE_EXCLUSIVE,
.queueFamilyIndexCount = 0,
.pQueueFamilyIndices = nullptr,
};
VkBuffer handle{};
if (dld.vkCreateBuffer(logical, &buffer_ci, nullptr, &handle) != VK_SUCCESS) {
return VK_NULL_HANDLE;
}
std::vector<VkSparseMemoryBind> binds;
binds.reserve(sources.size());
VkDeviceSize resource_offset = 0;
for (const MultiRangeSource& source : sources) {
binds.push_back(VkSparseMemoryBind{
.resourceOffset = resource_offset,
.size = source.size,
.memory = source.memory,
.memoryOffset = source.memory_offset + source.offset,
.flags = 0,
});
resource_offset += source.size;
}
const VkSparseBufferMemoryBindInfo buffer_bind{
.buffer = handle,
.bindCount = static_cast<u32>(binds.size()),
.pBinds = binds.data(),
};
const VkBindSparseInfo bind_info{
.sType = VK_STRUCTURE_TYPE_BIND_SPARSE_INFO,
.pNext = nullptr,
.waitSemaphoreCount = 0,
.pWaitSemaphores = nullptr,
.bufferBindCount = 1,
.pBufferBinds = &buffer_bind,
.imageOpaqueBindCount = 0,
.pImageOpaqueBinds = nullptr,
.imageBindCount = 0,
.pImageBinds = nullptr,
.signalSemaphoreCount = 0,
.pSignalSemaphores = nullptr,
};
const VkFenceCreateInfo fence_ci{
.sType = VK_STRUCTURE_TYPE_FENCE_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
};
vk::Fence fence = device.GetLogical().CreateFence(fence_ci);
VkResult bind_result = VK_ERROR_UNKNOWN;
{
std::scoped_lock lock{scheduler.submit_mutex};
bind_result = device.GetGraphicsQueue().BindSparse(bind_info, *fence);
}
if (bind_result != VK_SUCCESS) {
dld.vkDestroyBuffer(logical, handle, nullptr);
return VK_NULL_HANDLE;
}
fence.Wait();
return handle;
}
void MultiRangeBufferCache::DestroySparse(VkBuffer handle) {
if (handle == VK_NULL_HANDLE) {
return;
}
retired.push_back(Retired{
.handle = handle,
.tick = scheduler.CurrentTick(),
});
}
void MultiRangeBufferCache::DrainRetired() {
const VkDevice logical = *device.GetLogical();
const auto& dld = device.GetDispatchLoader();
size_t index = 0;
while (index < retired.size()) {
if (scheduler.IsFree(retired[index].tick)) {
dld.vkDestroyBuffer(logical, retired[index].handle, nullptr);
retired[index] = retired.back();
retired.pop_back();
} else {
++index;
}
}
}
MultiRangeRef MultiRangeBufferCache::Get(u64 key, std::span<const MultiRangeSource> sources,
VkDeviceSize total) {
if (sources.empty() || total == 0) {
return MultiRangeRef{};
}
if (!retired.empty()) {
DrainRetired();
}
const u64 geometry = HashSources(sources);
const auto it = entries.find(key);
if (it != entries.end() && it->second.geometry == geometry && it->second.size == total) {
Entry& entry = it->second;
MultiRangeRef ref{
.handle = entry.sparse_handle,
.address = entry.address,
.size = entry.size,
.needs_gather = false,
};
if (entry.sparse_handle == VK_NULL_HANDLE) {
ref.handle = *entry.gathered;
ref.needs_gather = entry.dirty;
}
return ref;
}
if (it != entries.end()) {
DestroySparse(it->second.sparse_handle);
entries.erase(it);
}
Entry entry;
entry.geometry = geometry;
entry.size = total;
if (CanBindSparse(sources)) {
entry.sparse_handle = CreateSparse(sources, total);
if (entry.sparse_handle != VK_NULL_HANDLE) {
entry.owners.reserve(sources.size());
for (const MultiRangeSource& source : sources) {
entry.owners.push_back(source.handle);
}
}
}
if (entry.sparse_handle == VK_NULL_HANDLE) {
VkBufferUsageFlags flags = VK_BUFFER_USAGE_TRANSFER_SRC_BIT |
VK_BUFFER_USAGE_TRANSFER_DST_BIT |
VK_BUFFER_USAGE_STORAGE_BUFFER_BIT;
if (device.IsBufferDeviceAddressSupported()) {
flags |= VK_BUFFER_USAGE_SHADER_DEVICE_ADDRESS_BIT;
}
const VkBufferCreateInfo gather_ci{
.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.size = total,
.usage = flags,
.sharingMode = VK_SHARING_MODE_EXCLUSIVE,
.queueFamilyIndexCount = 0,
.pQueueFamilyIndices = nullptr,
};
entry.gathered = memory_allocator.CreateBuffer(gather_ci, MemoryUsage::DeviceLocal);
entry.dirty = true;
}
if (device.IsBufferDeviceAddressSupported()) {
VkBuffer address_handle = entry.sparse_handle;
if (address_handle == VK_NULL_HANDLE) {
address_handle = *entry.gathered;
}
entry.address = device.GetLogical().GetBufferDeviceAddress(address_handle);
}
MultiRangeRef ref{
.handle = entry.sparse_handle,
.address = entry.address,
.size = entry.size,
.needs_gather = false,
};
if (entry.sparse_handle == VK_NULL_HANDLE) {
ref.handle = *entry.gathered;
ref.needs_gather = true;
}
entries.emplace(key, std::move(entry));
return ref;
}
void MultiRangeBufferCache::MarkGathered(u64 key) {
const auto it = entries.find(key);
if (it != entries.end()) {
it->second.dirty = false;
}
}
void MultiRangeBufferCache::DropOwner(VkBuffer owner) {
if (owner == VK_NULL_HANDLE) {
return;
}
for (auto it = entries.begin(); it != entries.end();) {
Entry& entry = it->second;
bool owned = false;
for (const VkBuffer handle : entry.owners) {
if (handle == owner) {
owned = true;
break;
}
}
if (owned) {
DestroySparse(entry.sparse_handle);
it = entries.erase(it);
} else {
++it;
}
}
}
void MultiRangeBufferCache::Invalidate(u64 key) {
const auto it = entries.find(key);
if (it != entries.end()) {
it->second.dirty = true;
}
}
void MultiRangeBufferCache::Clear() {
for (auto& [key, entry] : entries) {
DestroySparse(entry.sparse_handle);
}
entries.clear();
}
} // namespace Vulkan
@@ -1,105 +0,0 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
#pragma once
#include <span>
#include <unordered_map>
#include <vector>
#include "common/common_types.h"
#include "video_core/vulkan_common/vulkan_memory_allocator.h"
#include "video_core/vulkan_common/vulkan_wrapper.h"
namespace Vulkan {
class Device;
class Scheduler;
struct MultiRangeSource {
VkBuffer handle{};
VkDeviceMemory memory{};
VkDeviceSize memory_offset{};
VkDeviceSize offset{};
VkDeviceSize size{};
u32 memory_type{};
};
struct MultiRangeRef {
VkBuffer handle{};
VkDeviceAddress address{};
VkDeviceSize size{};
bool needs_gather{};
};
class MultiRangeBufferCache final {
public:
static constexpr VkDeviceSize DEFAULT_BLOCK_SIZE = 64 * 1024;
explicit MultiRangeBufferCache(const Device& device_, MemoryAllocator& memory_allocator_,
Scheduler& scheduler_);
~MultiRangeBufferCache();
MultiRangeBufferCache(const MultiRangeBufferCache&) = delete;
MultiRangeBufferCache& operator=(const MultiRangeBufferCache&) = delete;
[[nodiscard]] bool UsesSparse() const noexcept {
return use_sparse;
}
[[nodiscard]] VkDeviceSize BlockSize() const noexcept {
return block_size;
}
[[nodiscard]] MultiRangeRef Get(u64 key, std::span<const MultiRangeSource> sources,
VkDeviceSize total);
void MarkGathered(u64 key);
void Invalidate(u64 key);
void DropOwner(VkBuffer owner);
void Clear();
private:
struct Retired {
VkBuffer handle{};
u64 tick{};
};
struct Entry {
vk::Buffer gathered;
VkBuffer sparse_handle{};
VkDeviceAddress address{};
VkDeviceSize size{};
u64 geometry{};
bool dirty{true};
std::vector<VkBuffer> owners;
};
[[nodiscard]] u64 HashSources(std::span<const MultiRangeSource> sources) const;
[[nodiscard]] bool CanBindSparse(std::span<const MultiRangeSource> sources) const;
[[nodiscard]] VkBuffer CreateSparse(std::span<const MultiRangeSource> sources,
VkDeviceSize total);
[[nodiscard]] VkDeviceSize QueryBlockSize(u32& memory_type_bits) const;
void DestroySparse(VkBuffer handle);
void DrainRetired();
const Device& device;
MemoryAllocator& memory_allocator;
Scheduler& scheduler;
bool use_sparse{};
VkDeviceSize block_size{DEFAULT_BLOCK_SIZE};
u32 sparse_memory_type_bits{};
VkBufferUsageFlags sparse_usage{};
std::unordered_map<u64, Entry> entries;
std::vector<Retired> retired;
};
} // namespace Vulkan
@@ -819,7 +819,6 @@ void RasterizerVulkan::ModifyGPUMemory(size_t as_id, GPUVAddr addr, u64 size) {
std::scoped_lock lock{texture_cache.mutex}; std::scoped_lock lock{texture_cache.mutex};
texture_cache.UnmapGPUMemory(as_id, addr, size); texture_cache.UnmapGPUMemory(as_id, addr, size);
} }
buffer_cache.UnmapGPUMemory(as_id, addr, size);
} }
void RasterizerVulkan::SignalFence(std::function<void()>&& func) { void RasterizerVulkan::SignalFence(std::function<void()>&& func) {
@@ -1570,8 +1570,6 @@ void Device::SetupFamilies(VkSurfaceKHR surface) {
} }
if (graphics) { if (graphics) {
graphics_family = *graphics; graphics_family = *graphics;
graphics_family_sparse_binding =
(queue_family_properties[*graphics].queueFlags & VK_QUEUE_SPARSE_BINDING_BIT) != 0;
} }
if (present) { if (present) {
present_family = *present; present_family = *present;
@@ -317,10 +317,6 @@ public:
return properties.driver.driverID; return properties.driver.driverID;
} }
bool IsSparseBindingSupported() const {
return features.features.sparseBinding && graphics_family_sparse_binding;
}
/// Returns true for tile-based deferred renderers. /// Returns true for tile-based deferred renderers.
bool IsTiler() const { bool IsTiler() const {
switch (GetDriverID()) { switch (GetDriverID()) {
@@ -1150,7 +1146,6 @@ private:
bool owns_static_pipeline_cache{}; bool owns_static_pipeline_cache{};
u32 instance_version{}; ///< Vulkan instance version. u32 instance_version{}; ///< Vulkan instance version.
u32 graphics_family{}; ///< Main graphics queue family index. u32 graphics_family{}; ///< Main graphics queue family index.
bool graphics_family_sparse_binding{};
u32 present_family{}; ///< Main present queue family index. u32 present_family{}; ///< Main present queue family index.
struct Extensions { struct Extensions {
@@ -280,51 +280,6 @@ vk::Buffer MemoryAllocator::CreateBuffer(const VkBufferCreateInfo &ci, MemoryUsa
device.GetDispatchLoader()); device.GetDispatchLoader());
} }
vk::Buffer MemoryAllocator::CreateBuffer(const VkBufferCreateInfo &ci, MemoryUsage usage,
VkDeviceSize min_alignment) const {
if (min_alignment <= 1) {
return CreateBuffer(ci, usage);
}
VkMemoryPropertyFlags anv_flags = 0;
if (usage == MemoryUsage::Stream &&
device.GetDriverID() == VK_DRIVER_ID_INTEL_OPEN_SOURCE_MESA) {
anv_flags = VK_MEMORY_PROPERTY_HOST_CACHED_BIT;
}
u32 memory_type_bits = valid_memory_types;
if (usage == MemoryUsage::Stream) {
memory_type_bits = 0u;
}
const VmaAllocationCreateInfo alloc_ci = {
.flags = VMA_ALLOCATION_CREATE_WITHIN_BUDGET_BIT | MemoryUsageVmaFlags(usage),
.usage = MemoryUsageVma(usage),
.requiredFlags = 0,
.preferredFlags = MemoryUsagePreferredVmaFlags(usage) | anv_flags,
.memoryTypeBits = memory_type_bits,
.pool = VK_NULL_HANDLE,
.pUserData = nullptr,
.priority = 0.f,
};
VkBuffer handle{};
VmaAllocationInfo alloc_info{};
VmaAllocation allocation{};
VkMemoryPropertyFlags property_flags{};
vk::Check(vmaCreateBufferWithAlignment(allocator, &ci, &alloc_ci, min_alignment, &handle,
&allocation, &alloc_info));
vmaGetAllocationMemoryProperties(allocator, allocation, &property_flags);
u8 *data = reinterpret_cast<u8 *>(alloc_info.pMappedData);
std::span<u8> mapped_data{};
if (data) {
mapped_data = std::span<u8>{data, ci.size};
}
const bool is_coherent = (property_flags & VK_MEMORY_PROPERTY_HOST_COHERENT_BIT) != 0;
return vk::Buffer(handle, *device.GetLogical(), allocator, allocation, mapped_data, is_coherent,
device.GetDispatchLoader());
}
MemoryCommit MemoryAllocator::Commit(const VkMemoryRequirements &reqs, MemoryUsage usage) MemoryCommit MemoryAllocator::Commit(const VkMemoryRequirements &reqs, MemoryUsage usage)
{ {
const auto vma_usage = MemoryUsageVma(usage); const auto vma_usage = MemoryUsageVma(usage);
@@ -1,4 +1,4 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project // SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later // SPDX-License-Identifier: GPL-3.0-or-later
// SPDX-FileCopyrightText: Copyright 2019 yuzu Emulator Project // SPDX-FileCopyrightText: Copyright 2019 yuzu Emulator Project
@@ -107,9 +107,6 @@ namespace Vulkan {
vk::Buffer CreateBuffer(const VkBufferCreateInfo &ci, MemoryUsage usage) const; vk::Buffer CreateBuffer(const VkBufferCreateInfo &ci, MemoryUsage usage) const;
vk::Buffer CreateBuffer(const VkBufferCreateInfo &ci, MemoryUsage usage,
VkDeviceSize min_alignment) const;
/** /**
* Commits a memory with the specified requirements. * Commits a memory with the specified requirements.
* *
@@ -229,7 +229,6 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept {
X(vkGetPipelineExecutableStatisticsKHR); X(vkGetPipelineExecutableStatisticsKHR);
X(vkGetSemaphoreCounterValue); X(vkGetSemaphoreCounterValue);
X(vkMapMemory); X(vkMapMemory);
X(vkQueueBindSparse);
X(vkQueueSubmit); X(vkQueueSubmit);
X(vkQueueSubmit2); X(vkQueueSubmit2);
X(vkResetFences); X(vkResetFences);
@@ -540,19 +539,6 @@ void Buffer::SetObjectNameEXT(const char* name) const {
SetObjectName(dld, owner, handle, VK_OBJECT_TYPE_BUFFER, name); SetObjectName(dld, owner, handle, VK_OBJECT_TYPE_BUFFER, name);
} }
MemoryLocation Buffer::Location() const noexcept {
if (!allocation) {
return MemoryLocation{};
}
VmaAllocationInfo info{};
vmaGetAllocationInfo(allocator, allocation, &info);
return MemoryLocation{
.memory = info.deviceMemory,
.offset = info.offset,
.memory_type = info.memoryType,
};
}
void Buffer::Release() const noexcept { void Buffer::Release() const noexcept {
if (handle) { if (handle) {
vmaDestroyBuffer(allocator, handle, allocation); vmaDestroyBuffer(allocator, handle, allocation);
@@ -345,7 +345,6 @@ struct DeviceDispatch : InstanceDispatch {
PFN_vkGetQueryPoolResults vkGetQueryPoolResults{}; PFN_vkGetQueryPoolResults vkGetQueryPoolResults{};
PFN_vkGetSemaphoreCounterValue vkGetSemaphoreCounterValue{}; PFN_vkGetSemaphoreCounterValue vkGetSemaphoreCounterValue{};
PFN_vkMapMemory vkMapMemory{}; PFN_vkMapMemory vkMapMemory{};
PFN_vkQueueBindSparse vkQueueBindSparse{};
PFN_vkQueueSubmit vkQueueSubmit{}; PFN_vkQueueSubmit vkQueueSubmit{};
PFN_vkQueueSubmit2 vkQueueSubmit2{}; PFN_vkQueueSubmit2 vkQueueSubmit2{};
PFN_vkResetFences vkResetFences{}; PFN_vkResetFences vkResetFences{};
@@ -741,12 +740,6 @@ private:
const DeviceDispatch* dld = nullptr; const DeviceDispatch* dld = nullptr;
}; };
struct MemoryLocation {
VkDeviceMemory memory{};
VkDeviceSize offset{};
u32 memory_type{};
};
class Buffer { class Buffer {
public: public:
explicit Buffer(VkBuffer handle_, VkDevice owner_, VmaAllocator allocator_, explicit Buffer(VkBuffer handle_, VkDevice owner_, VmaAllocator allocator_,
@@ -818,8 +811,6 @@ public:
void SetObjectNameEXT(const char* name) const; void SetObjectNameEXT(const char* name) const;
MemoryLocation Location() const noexcept;
private: private:
void Release() const noexcept; void Release() const noexcept;
@@ -852,11 +843,6 @@ public:
return dld->vkQueueSubmit2(queue, submit_infos.size(), submit_infos.data(), fence); return dld->vkQueueSubmit2(queue, submit_infos.size(), submit_infos.data(), fence);
} }
VkResult BindSparse(Span<VkBindSparseInfo> bind_infos,
VkFence fence = VK_NULL_HANDLE) const noexcept {
return dld->vkQueueBindSparse(queue, bind_infos.size(), bind_infos.data(), fence);
}
VkResult Present(const VkPresentInfoKHR& present_info) const noexcept { VkResult Present(const VkPresentInfoKHR& present_info) const noexcept {
return dld->vkQueuePresentKHR(queue, &present_info); return dld->vkQueuePresentKHR(queue, &present_info);
} }
+5 -7
View File
@@ -538,6 +538,7 @@ void GRenderWindow::mouseReleaseEvent(QMouseEvent* event) {
} }
void GRenderWindow::ConstrainMouse() { void GRenderWindow::ConstrainMouse() {
input_subsystem->GetMouse()->NotifyChanged(); // required to reset mouse once it's no longer moved
if (QtCommon::emu_thread == nullptr || !Settings::values.mouse_panning) { if (QtCommon::emu_thread == nullptr || !Settings::values.mouse_panning) {
mouse_constrain_timer.stop(); mouse_constrain_timer.stop();
return; return;
@@ -552,15 +553,12 @@ void GRenderWindow::ConstrainMouse() {
const auto pos = mapFromGlobal(QCursor::pos()); const auto pos = mapFromGlobal(QCursor::pos());
const int new_pos_x = std::clamp(pos.x(), 0, width()); const int new_pos_x = std::clamp(pos.x(), 0, width());
const int new_pos_y = std::clamp(pos.y(), 0, height()); const int new_pos_y = std::clamp(pos.y(), 0, height());
QCursor::setPos(mapToGlobal(QPoint{new_pos_x, new_pos_y})); QCursor::setPos(mapToGlobal(QPoint{new_pos_x, new_pos_y}));
return; } else {
const int center_x = width() / 2;
const int center_y = height() / 2;
QCursor::setPos(mapToGlobal(QPoint{center_x, center_y}));
} }
const int center_x = width() / 2;
const int center_y = height() / 2;
QCursor::setPos(mapToGlobal(QPoint{center_x, center_y}));
} }
void GRenderWindow::wheelEvent(QWheelEvent* event) { void GRenderWindow::wheelEvent(QWheelEvent* event) {
@@ -37,10 +37,16 @@ EmuWindow_SDL3::EmuWindow_SDL3(InputCommon::InputSubsystem* input_subsystem_, Co
SDL_SetWindowTitle(this_->render_window, title.c_str()); SDL_SetWindowTitle(this_->render_window, title.c_str());
return 2000; return 2000;
}, this); }, this);
mouse_timer = SDL_AddTimer(100, [](void *userdata, SDL_TimerID, Uint32) -> Uint32 {
auto* this_ = (EmuWindow_SDL3*)userdata;
this_->input_subsystem->GetMouse()->NotifyChanged();
return 100;
}, this);
} }
EmuWindow_SDL3::~EmuWindow_SDL3() { EmuWindow_SDL3::~EmuWindow_SDL3() {
SDL_RemoveTimer(titlebar_timer); SDL_RemoveTimer(titlebar_timer);
SDL_RemoveTimer(mouse_timer);
system.HIDCore().UnloadInputDevices(); system.HIDCore().UnloadInputDevices();
input_subsystem->Shutdown(); input_subsystem->Shutdown();
SDL_Quit(); SDL_Quit();
@@ -84,6 +84,9 @@ protected:
/// Periodic changer of titlebar (independent of event loop) /// Periodic changer of titlebar (independent of event loop)
SDL_TimerID titlebar_timer; SDL_TimerID titlebar_timer;
// Mouse resetter once it
SDL_TimerID mouse_timer;
/// Is the window still open? /// Is the window still open?
bool is_open = true; bool is_open = true;