summaryrefslogtreecommitdiffstats
path: root/src/video_core/gpu.cpp
blob: 6ad3b94f85dfebf7409edb1e79de1eb4829fb2fa (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
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
// SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later

#include <array>
#include <atomic>
#include <chrono>
#include <condition_variable>
#include <list>
#include <memory>

#include "common/assert.h"
#include "common/microprofile.h"
#include "common/settings.h"
#include "core/core.h"
#include "core/core_timing.h"
#include "core/frontend/emu_window.h"
#include "core/frontend/graphics_context.h"
#include "core/hle/service/nvdrv/nvdata.h"
#include "core/perf_stats.h"
#include "video_core/cdma_pusher.h"
#include "video_core/control/channel_state.h"
#include "video_core/control/scheduler.h"
#include "video_core/dma_pusher.h"
#include "video_core/engines/fermi_2d.h"
#include "video_core/engines/kepler_compute.h"
#include "video_core/engines/kepler_memory.h"
#include "video_core/engines/maxwell_3d.h"
#include "video_core/engines/maxwell_dma.h"
#include "video_core/gpu.h"
#include "video_core/gpu_thread.h"
#include "video_core/host1x/host1x.h"
#include "video_core/host1x/syncpoint_manager.h"
#include "video_core/memory_manager.h"
#include "video_core/renderer_base.h"
#include "video_core/shader_notify.h"

#pragma optimize("", off)

namespace Tegra {

struct GPU::Impl {
    explicit Impl(GPU& gpu_, Core::System& system_, bool is_async_, bool use_nvdec_)
        : gpu{gpu_}, system{system_}, host1x{system.Host1x()}, use_nvdec{use_nvdec_},
          shader_notify{std::make_unique<VideoCore::ShaderNotify>()}, is_async{is_async_},
          gpu_thread{system_, is_async_}, scheduler{std::make_unique<Control::Scheduler>(gpu)} {}

    ~Impl() = default;

    std::shared_ptr<Control::ChannelState> CreateChannel(s32 channel_id) {
        auto channel_state = std::make_shared<Tegra::Control::ChannelState>(channel_id);
        channels.emplace(channel_id, channel_state);
        scheduler->DeclareChannel(channel_state);
        return channel_state;
    }

    void BindChannel(s32 channel_id) {
        if (bound_channel == channel_id) {
            return;
        }
        auto it = channels.find(channel_id);
        ASSERT(it != channels.end());
        bound_channel = channel_id;
        current_channel = it->second.get();

        rasterizer->BindChannel(*current_channel);
    }

    std::shared_ptr<Control::ChannelState> AllocateChannel() {
        return CreateChannel(new_channel_id++);
    }

    void InitChannel(Control::ChannelState& to_init) {
        to_init.Init(system, gpu);
        to_init.BindRasterizer(rasterizer);
        rasterizer->InitializeChannel(to_init);
    }

    void InitAddressSpace(Tegra::MemoryManager& memory_manager) {
        memory_manager.BindRasterizer(rasterizer);
    }

    void ReleaseChannel(Control::ChannelState& to_release) {
        UNIMPLEMENTED();
    }

    /// Binds a renderer to the GPU.
    void BindRenderer(std::unique_ptr<VideoCore::RendererBase> renderer_) {
        renderer = std::move(renderer_);
        rasterizer = renderer->ReadRasterizer();
        host1x.MemoryManager().BindInterface(rasterizer);
        host1x.GMMU().BindRasterizer(rasterizer);
    }

    /// Flush all current written commands into the host GPU for execution.
    void FlushCommands() {
        rasterizer->FlushCommands();
    }

    /// Synchronizes CPU writes with Host GPU memory.
    void InvalidateGPUCache() {
        std::function<void(PAddr, size_t)> callback_writes(
            [this](PAddr address, size_t size) { rasterizer->OnCacheInvalidation(address, size); });
        system.GatherGPUDirtyMemory(callback_writes);
    }

    /// Signal the ending of command list.
    void OnCommandListEnd() {
        rasterizer->ReleaseFences(false);
        Settings::UpdateGPUAccuracy();
    }

    /// Request a host GPU memory flush from the CPU.
    template <typename Func>
    [[nodiscard]] u64 RequestSyncOperation(Func&& action) {
        std::unique_lock lck{sync_request_mutex};
        const u64 fence = ++last_sync_fence;
        sync_requests.emplace_back(action);
        return fence;
    }

    /// Obtains current flush request fence id.
    [[nodiscard]] u64 CurrentSyncRequestFence() const {
        return current_sync_fence.load(std::memory_order_relaxed);
    }

    void WaitForSyncOperation(const u64 fence) {
        std::unique_lock lck{sync_request_mutex};
        sync_request_cv.wait(lck, [this, fence] { return CurrentSyncRequestFence() >= fence; });
    }

    /// Tick pending requests within the GPU.
    void TickWork() {
        std::unique_lock lck{sync_request_mutex};
        while (!sync_requests.empty()) {
            auto request = std::move(sync_requests.front());
            sync_requests.pop_front();
            sync_request_mutex.unlock();
            request();
            current_sync_fence.fetch_add(1, std::memory_order_release);
            sync_request_mutex.lock();
            sync_request_cv.notify_all();
        }
    }

    /// Returns a reference to the Maxwell3D GPU engine.
    [[nodiscard]] Engines::Maxwell3D& Maxwell3D() {
        ASSERT(current_channel);
        return *current_channel->maxwell_3d;
    }

    /// Returns a const reference to the Maxwell3D GPU engine.
    [[nodiscard]] const Engines::Maxwell3D& Maxwell3D() const {
        ASSERT(current_channel);
        return *current_channel->maxwell_3d;
    }

    /// Returns a reference to the KeplerCompute GPU engine.
    [[nodiscard]] Engines::KeplerCompute& KeplerCompute() {
        ASSERT(current_channel);
        return *current_channel->kepler_compute;
    }

    /// Returns a reference to the KeplerCompute GPU engine.
    [[nodiscard]] const Engines::KeplerCompute& KeplerCompute() const {
        ASSERT(current_channel);
        return *current_channel->kepler_compute;
    }

    /// Returns a reference to the GPU DMA pusher.
    [[nodiscard]] Tegra::DmaPusher& DmaPusher() {
        ASSERT(current_channel);
        return *current_channel->dma_pusher;
    }

    /// Returns a const reference to the GPU DMA pusher.
    [[nodiscard]] const Tegra::DmaPusher& DmaPusher() const {
        ASSERT(current_channel);
        return *current_channel->dma_pusher;
    }

    /// Returns a reference to the underlying renderer.
    [[nodiscard]] VideoCore::RendererBase& Renderer() {
        return *renderer;
    }

    /// Returns a const reference to the underlying renderer.
    [[nodiscard]] const VideoCore::RendererBase& Renderer() const {
        return *renderer;
    }

    /// Returns a reference to the shader notifier.
    [[nodiscard]] VideoCore::ShaderNotify& ShaderNotify() {
        return *shader_notify;
    }

    /// Returns a const reference to the shader notifier.
    [[nodiscard]] const VideoCore::ShaderNotify& ShaderNotify() const {
        return *shader_notify;
    }

    [[nodiscard]] u64 GetTicks() const {
        u64 gpu_tick = system.CoreTiming().GetGPUTicks();

        if (Settings::values.use_fast_gpu_time.GetValue()) {
            gpu_tick /= 256;
        }

        return gpu_tick;
    }

    [[nodiscard]] bool IsAsync() const {
        return is_async;
    }

    [[nodiscard]] bool UseNvdec() const {
        return use_nvdec;
    }

    void RendererFrameEndNotify() {
        system.GetPerfStats().EndGameFrame();
    }

    /// Performs any additional setup necessary in order to begin GPU emulation.
    /// This can be used to launch any necessary threads and register any necessary
    /// core timing events.
    void Start() {
        Settings::UpdateGPUAccuracy();
        gpu_thread.StartThread(*renderer, renderer->Context(), *scheduler);
    }

    void NotifyShutdown() {
        std::unique_lock lk{sync_mutex};
        shutting_down.store(true, std::memory_order::relaxed);
        sync_cv.notify_all();
    }

    /// Obtain the CPU Context
    void ObtainContext() {
        if (!cpu_context) {
            cpu_context = renderer->GetRenderWindow().CreateSharedContext();
        }
        cpu_context->MakeCurrent();
    }

    /// Release the CPU Context
    void ReleaseContext() {
        cpu_context->DoneCurrent();
    }

    /// Push GPU command entries to be processed
    void PushGPUEntries(s32 channel, Tegra::CommandList&& entries) {
        gpu_thread.SubmitList(channel, std::move(entries));
    }

    /// Push GPU command buffer entries to be processed
    void PushCommandBuffer(u32 id, Tegra::ChCommandHeaderList& entries) {
        if (!use_nvdec) {
            return;
        }

        if (!cdma_pushers.contains(id)) {
            cdma_pushers.insert_or_assign(id, std::make_unique<Tegra::CDmaPusher>(host1x));
        }

        // SubmitCommandBuffer would make the nvdec operations async, this is not currently working
        // TODO(ameerj): RE proper async nvdec operation
        // gpu_thread.SubmitCommandBuffer(std::move(entries));
        cdma_pushers[id]->ProcessEntries(std::move(entries));
    }

    /// Frees the CDMAPusher instance to free up resources
    void ClearCdmaInstance(u32 id) {
        const auto iter = cdma_pushers.find(id);
        if (iter != cdma_pushers.end()) {
            cdma_pushers.erase(iter);
        }
    }

    /// Swap buffers (render frame)
    void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
        gpu_thread.SwapBuffers(framebuffer);
    }

    /// Notify rasterizer that any caches of the specified region should be flushed to Switch memory
    void FlushRegion(DAddr addr, u64 size) {
        gpu_thread.FlushRegion(addr, size);
    }

    VideoCore::RasterizerDownloadArea OnCPURead(DAddr addr, u64 size) {
        auto raster_area = rasterizer->GetFlushArea(addr, size);
        if (raster_area.preemtive) {
            return raster_area;
        }
        raster_area.preemtive = true;
        const u64 fence = RequestSyncOperation([this, &raster_area]() {
            rasterizer->FlushRegion(raster_area.start_address,
                                    raster_area.end_address - raster_area.start_address);
        });
        gpu_thread.TickGPU();
        WaitForSyncOperation(fence);
        return raster_area;
    }

    /// Notify rasterizer that any caches of the specified region should be invalidated
    void InvalidateRegion(DAddr addr, u64 size) {
        gpu_thread.InvalidateRegion(addr, size);
    }

    bool OnCPUWrite(DAddr addr, u64 size) {
        return rasterizer->OnCPUWrite(addr, size);
    }

    /// Notify rasterizer that any caches of the specified region should be flushed and invalidated
    void FlushAndInvalidateRegion(DAddr addr, u64 size) {
        gpu_thread.FlushAndInvalidateRegion(addr, size);
    }

    void RequestSwapBuffers(const Tegra::FramebufferConfig* framebuffer,
                            std::array<Service::Nvidia::NvFence, 4>& fences, size_t num_fences) {
        size_t current_request_counter{};
        {
            std::unique_lock<std::mutex> lk(request_swap_mutex);
            if (free_swap_counters.empty()) {
                current_request_counter = request_swap_counters.size();
                request_swap_counters.emplace_back(num_fences);
            } else {
                current_request_counter = free_swap_counters.front();
                request_swap_counters[current_request_counter] = num_fences;
                free_swap_counters.pop_front();
            }
        }
        const auto wait_fence =
            RequestSyncOperation([this, current_request_counter, framebuffer, fences, num_fences] {
                auto& syncpoint_manager = host1x.GetSyncpointManager();
                if (num_fences == 0) {
                    renderer->SwapBuffers(framebuffer);
                }
                const auto executer = [this, current_request_counter,
                                       framebuffer_copy = *framebuffer]() {
                    {
                        std::unique_lock<std::mutex> lk(request_swap_mutex);
                        if (--request_swap_counters[current_request_counter] != 0) {
                            return;
                        }
                        free_swap_counters.push_back(current_request_counter);
                    }
                    renderer->SwapBuffers(&framebuffer_copy);
                };
                for (size_t i = 0; i < num_fences; i++) {
                    syncpoint_manager.RegisterGuestAction(fences[i].id, fences[i].value, executer);
                }
            });
        gpu_thread.TickGPU();
        WaitForSyncOperation(wait_fence);
    }

    GPU& gpu;
    Core::System& system;
    Host1x::Host1x& host1x;

    std::map<u32, std::unique_ptr<Tegra::CDmaPusher>> cdma_pushers;
    std::unique_ptr<VideoCore::RendererBase> renderer;
    VideoCore::RasterizerInterface* rasterizer = nullptr;
    const bool use_nvdec;

    s32 new_channel_id{1};
    /// Shader build notifier
    std::unique_ptr<VideoCore::ShaderNotify> shader_notify;
    /// When true, we are about to shut down emulation session, so terminate outstanding tasks
    std::atomic_bool shutting_down{};

    std::array<std::atomic<u32>, Service::Nvidia::MaxSyncPoints> syncpoints{};

    std::array<std::list<u32>, Service::Nvidia::MaxSyncPoints> syncpt_interrupts;

    std::mutex sync_mutex;
    std::mutex device_mutex;

    std::condition_variable sync_cv;

    std::list<std::function<void()>> sync_requests;
    std::atomic<u64> current_sync_fence{};
    u64 last_sync_fence{};
    std::mutex sync_request_mutex;
    std::condition_variable sync_request_cv;

    const bool is_async;

    VideoCommon::GPUThread::ThreadManager gpu_thread;
    std::unique_ptr<Core::Frontend::GraphicsContext> cpu_context;

    std::unique_ptr<Tegra::Control::Scheduler> scheduler;
    std::unordered_map<s32, std::shared_ptr<Tegra::Control::ChannelState>> channels;
    Tegra::Control::ChannelState* current_channel;
    s32 bound_channel{-1};

    std::deque<size_t> free_swap_counters;
    std::deque<size_t> request_swap_counters;
    std::mutex request_swap_mutex;
};

GPU::GPU(Core::System& system, bool is_async, bool use_nvdec)
    : impl{std::make_unique<Impl>(*this, system, is_async, use_nvdec)} {}

GPU::~GPU() = default;

std::shared_ptr<Control::ChannelState> GPU::AllocateChannel() {
    return impl->AllocateChannel();
}

void GPU::InitChannel(Control::ChannelState& to_init) {
    impl->InitChannel(to_init);
}

void GPU::BindChannel(s32 channel_id) {
    impl->BindChannel(channel_id);
}

void GPU::ReleaseChannel(Control::ChannelState& to_release) {
    impl->ReleaseChannel(to_release);
}

void GPU::InitAddressSpace(Tegra::MemoryManager& memory_manager) {
    impl->InitAddressSpace(memory_manager);
}

void GPU::BindRenderer(std::unique_ptr<VideoCore::RendererBase> renderer) {
    impl->BindRenderer(std::move(renderer));
}

void GPU::FlushCommands() {
    impl->FlushCommands();
}

void GPU::InvalidateGPUCache() {
    impl->InvalidateGPUCache();
}

void GPU::OnCommandListEnd() {
    impl->OnCommandListEnd();
}

u64 GPU::RequestFlush(DAddr addr, std::size_t size) {
    return impl->RequestSyncOperation(
        [this, addr, size]() { impl->rasterizer->FlushRegion(addr, size); });
}

u64 GPU::CurrentSyncRequestFence() const {
    return impl->CurrentSyncRequestFence();
}

void GPU::WaitForSyncOperation(u64 fence) {
    return impl->WaitForSyncOperation(fence);
}

void GPU::TickWork() {
    impl->TickWork();
}

/// Gets a mutable reference to the Host1x interface
Host1x::Host1x& GPU::Host1x() {
    return impl->host1x;
}

/// Gets an immutable reference to the Host1x interface.
const Host1x::Host1x& GPU::Host1x() const {
    return impl->host1x;
}

Engines::Maxwell3D& GPU::Maxwell3D() {
    return impl->Maxwell3D();
}

const Engines::Maxwell3D& GPU::Maxwell3D() const {
    return impl->Maxwell3D();
}

Engines::KeplerCompute& GPU::KeplerCompute() {
    return impl->KeplerCompute();
}

const Engines::KeplerCompute& GPU::KeplerCompute() const {
    return impl->KeplerCompute();
}

Tegra::DmaPusher& GPU::DmaPusher() {
    return impl->DmaPusher();
}

const Tegra::DmaPusher& GPU::DmaPusher() const {
    return impl->DmaPusher();
}

VideoCore::RendererBase& GPU::Renderer() {
    return impl->Renderer();
}

const VideoCore::RendererBase& GPU::Renderer() const {
    return impl->Renderer();
}

VideoCore::ShaderNotify& GPU::ShaderNotify() {
    return impl->ShaderNotify();
}

const VideoCore::ShaderNotify& GPU::ShaderNotify() const {
    return impl->ShaderNotify();
}

void GPU::RequestSwapBuffers(const Tegra::FramebufferConfig* framebuffer,
                             std::array<Service::Nvidia::NvFence, 4>& fences, size_t num_fences) {
    impl->RequestSwapBuffers(framebuffer, fences, num_fences);
}

u64 GPU::GetTicks() const {
    return impl->GetTicks();
}

bool GPU::IsAsync() const {
    return impl->IsAsync();
}

bool GPU::UseNvdec() const {
    return impl->UseNvdec();
}

void GPU::RendererFrameEndNotify() {
    impl->RendererFrameEndNotify();
}

void GPU::Start() {
    impl->Start();
}

void GPU::NotifyShutdown() {
    impl->NotifyShutdown();
}

void GPU::ObtainContext() {
    impl->ObtainContext();
}

void GPU::ReleaseContext() {
    impl->ReleaseContext();
}

void GPU::PushGPUEntries(s32 channel, Tegra::CommandList&& entries) {
    impl->PushGPUEntries(channel, std::move(entries));
}

void GPU::PushCommandBuffer(u32 id, Tegra::ChCommandHeaderList& entries) {
    impl->PushCommandBuffer(id, entries);
}

void GPU::ClearCdmaInstance(u32 id) {
    impl->ClearCdmaInstance(id);
}

void GPU::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
    impl->SwapBuffers(framebuffer);
}

VideoCore::RasterizerDownloadArea GPU::OnCPURead(PAddr addr, u64 size) {
    return impl->OnCPURead(addr, size);
}

void GPU::FlushRegion(DAddr addr, u64 size) {
    impl->FlushRegion(addr, size);
}

void GPU::InvalidateRegion(DAddr addr, u64 size) {
    impl->InvalidateRegion(addr, size);
}

bool GPU::OnCPUWrite(DAddr addr, u64 size) {
    return impl->OnCPUWrite(addr, size);
}

void GPU::FlushAndInvalidateRegion(DAddr addr, u64 size) {
    impl->FlushAndInvalidateRegion(addr, size);
}

} // namespace Tegra