blob: 49e2b33d32423a0469e32f81bc666e1cdeb704b0 (
plain) (
blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
|
// SPDX-FileCopyrightText: Copyright 2019 yuzu Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later
#pragma once
#include <condition_variable>
#include <cstddef>
#include <memory>
#include <thread>
#include <utility>
#include <queue>
#include "common/alignment.h"
#include "common/common_types.h"
#include "video_core/renderer_vulkan/vk_master_semaphore.h"
#include "video_core/vulkan_common/vulkan_wrapper.h"
namespace Vulkan {
class CommandPool;
class Device;
class Framebuffer;
class GraphicsPipeline;
class StateTracker;
class VKQueryCache;
/// The scheduler abstracts command buffer and fence management with an interface that's able to do
/// OpenGL-like operations on Vulkan command buffers.
class Scheduler {
public:
explicit Scheduler(const Device& device, StateTracker& state_tracker);
~Scheduler();
/// Sends the current execution context to the GPU.
void Flush(VkSemaphore signal_semaphore = nullptr, VkSemaphore wait_semaphore = nullptr);
/// Sends the current execution context to the GPU and waits for it to complete.
void Finish(VkSemaphore signal_semaphore = nullptr, VkSemaphore wait_semaphore = nullptr);
/// Waits for the worker thread to finish executing everything. After this function returns it's
/// safe to touch worker resources.
void WaitWorker();
/// Sends currently recorded work to the worker thread.
void DispatchWork();
/// Requests to begin a renderpass.
void RequestRenderpass(const Framebuffer* framebuffer);
/// Requests the current executino context to be able to execute operations only allowed outside
/// of a renderpass.
void RequestOutsideRenderPassOperationContext();
/// Update the pipeline to the current execution context.
bool UpdateGraphicsPipeline(GraphicsPipeline* pipeline);
/// Update the rescaling state. Returns true if the state has to be updated.
bool UpdateRescaling(bool is_rescaling);
/// Invalidates current command buffer state except for render passes
void InvalidateState();
/// Assigns the query cache.
void SetQueryCache(VKQueryCache& query_cache_) {
query_cache = &query_cache_;
}
/// Send work to a separate thread.
template <typename T>
void Record(T&& command) {
if (chunk->Record(command)) {
return;
}
DispatchWork();
(void)chunk->Record(command);
}
/// Returns the current command buffer tick.
[[nodiscard]] u64 CurrentTick() const noexcept {
return master_semaphore->CurrentTick();
}
/// Returns true when a tick has been triggered by the GPU.
[[nodiscard]] bool IsFree(u64 tick) const noexcept {
return master_semaphore->IsFree(tick);
}
/// Waits for the given tick to trigger on the GPU.
void Wait(u64 tick) {
if (tick >= master_semaphore->CurrentTick()) {
// Make sure we are not waiting for the current tick without signalling
Flush();
}
master_semaphore->Wait(tick);
}
/// Returns the master timeline semaphore.
[[nodiscard]] MasterSemaphore& GetMasterSemaphore() const noexcept {
return *master_semaphore;
}
private:
class Command {
public:
virtual ~Command() = default;
virtual void Execute(vk::CommandBuffer cmdbuf) const = 0;
Command* GetNext() const {
return next;
}
void SetNext(Command* next_) {
next = next_;
}
private:
Command* next = nullptr;
};
template <typename T>
class TypedCommand final : public Command {
public:
explicit TypedCommand(T&& command_) : command{std::move(command_)} {}
~TypedCommand() override = default;
TypedCommand(TypedCommand&&) = delete;
TypedCommand& operator=(TypedCommand&&) = delete;
void Execute(vk::CommandBuffer cmdbuf) const override {
command(cmdbuf);
}
private:
T command;
};
class CommandChunk final {
public:
void ExecuteAll(vk::CommandBuffer cmdbuf);
template <typename T>
bool Record(T& command) {
using FuncType = TypedCommand<T>;
static_assert(sizeof(FuncType) < sizeof(data), "Lambda is too large");
recorded_counts++;
command_offset = Common::AlignUp(command_offset, alignof(FuncType));
if (command_offset > sizeof(data) - sizeof(FuncType)) {
return false;
}
Command* const current_last = last;
last = new (data.data() + command_offset) FuncType(std::move(command));
if (current_last) {
current_last->SetNext(last);
} else {
first = last;
}
command_offset += sizeof(FuncType);
return true;
}
void MarkSubmit() {
submit = true;
}
bool Empty() const {
return recorded_counts == 0;
}
bool HasSubmit() const {
return submit;
}
private:
Command* first = nullptr;
Command* last = nullptr;
size_t recorded_counts = 0;
size_t command_offset = 0;
bool submit = false;
alignas(std::max_align_t) std::array<u8, 0x8000> data{};
};
struct State {
VkRenderPass renderpass = nullptr;
VkFramebuffer framebuffer = nullptr;
VkExtent2D render_area = {0, 0};
GraphicsPipeline* graphics_pipeline = nullptr;
bool is_rescaling = false;
bool rescaling_defined = false;
};
void WorkerThread(std::stop_token stop_token);
void AllocateWorkerCommandBuffer();
void SubmitExecution(VkSemaphore signal_semaphore, VkSemaphore wait_semaphore);
void AllocateNewContext();
void EndPendingOperations();
void EndRenderPass();
void AcquireNewChunk();
const Device& device;
StateTracker& state_tracker;
std::unique_ptr<MasterSemaphore> master_semaphore;
std::unique_ptr<CommandPool> command_pool;
VKQueryCache* query_cache = nullptr;
vk::CommandBuffer current_cmdbuf;
std::unique_ptr<CommandChunk> chunk;
State state;
u32 num_renderpass_images = 0;
std::array<VkImage, 9> renderpass_images{};
std::array<VkImageSubresourceRange, 9> renderpass_image_ranges{};
std::queue<std::unique_ptr<CommandChunk>> work_queue;
std::vector<std::unique_ptr<CommandChunk>> chunk_reserve;
std::mutex reserve_mutex;
std::mutex work_mutex;
std::condition_variable_any work_cv;
std::condition_variable wait_cv;
std::jthread worker_thread;
};
} // namespace Vulkan
|