summaryrefslogtreecommitdiffstats
path: root/src/core/hw/gpu.cpp
blob: cfba82e51cba2976e6b03522ef20d8b551f48100 (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
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
// Copyright 2014 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.

#include <cstring>
#include <numeric>
#include <type_traits>
#include "common/color.h"
#include "common/common_types.h"
#include "common/logging/log.h"
#include "common/math_util.h"
#include "common/microprofile.h"
#include "common/thread.h"
#include "common/timer.h"
#include "common/vector_math.h"
#include "core/core_timing.h"
#include "core/hle/service/gsp_gpu.h"
#include "core/hle/service/hid/hid.h"
#include "core/hw/gpu.h"
#include "core/hw/hw.h"
#include "core/memory.h"
#include "core/settings.h"
#include "core/tracer/recorder.h"
#include "video_core/command_processor.h"
#include "video_core/debug_utils/debug_utils.h"
#include "video_core/rasterizer_interface.h"
#include "video_core/renderer_base.h"
#include "video_core/utils.h"
#include "video_core/video_core.h"

namespace GPU {

Regs g_regs;

/// 268MHz CPU clocks / 60Hz frames per second
const u64 frame_ticks = 268123480ull / 60;
/// Event id for CoreTiming
static int vblank_event;
/// Total number of frames drawn
static u64 frame_count;
/// Start clock for frame limiter
static u32 time_point;
/// Total delay caused by slow frames
static float time_delay;
constexpr float FIXED_FRAME_TIME = 1000.0f / 60;
// Max lag caused by slow frames. Can be adjusted to compensate for too many slow frames. Higher
// values increases time needed to limit frame rate after spikes
constexpr float MAX_LAG_TIME = 18;

template <typename T>
inline void Read(T& var, const u32 raw_addr) {
    u32 addr = raw_addr - HW::VADDR_GPU;
    u32 index = addr / 4;

    // Reads other than u32 are untested, so I'd rather have them abort than silently fail
    if (index >= Regs::NumIds() || !std::is_same<T, u32>::value) {
        LOG_ERROR(HW_GPU, "unknown Read%lu @ 0x%08X", sizeof(var) * 8, addr);
        return;
    }

    var = g_regs[addr / 4];
}

static Math::Vec4<u8> DecodePixel(Regs::PixelFormat input_format, const u8* src_pixel) {
    switch (input_format) {
    case Regs::PixelFormat::RGBA8:
        return Color::DecodeRGBA8(src_pixel);

    case Regs::PixelFormat::RGB8:
        return Color::DecodeRGB8(src_pixel);

    case Regs::PixelFormat::RGB565:
        return Color::DecodeRGB565(src_pixel);

    case Regs::PixelFormat::RGB5A1:
        return Color::DecodeRGB5A1(src_pixel);

    case Regs::PixelFormat::RGBA4:
        return Color::DecodeRGBA4(src_pixel);

    default:
        LOG_ERROR(HW_GPU, "Unknown source framebuffer format %x", input_format);
        return {0, 0, 0, 0};
    }
}

MICROPROFILE_DEFINE(GPU_DisplayTransfer, "GPU", "DisplayTransfer", MP_RGB(100, 100, 255));
MICROPROFILE_DEFINE(GPU_CmdlistProcessing, "GPU", "Cmdlist Processing", MP_RGB(100, 255, 100));

static void MemoryFill(const Regs::MemoryFillConfig& config) {
    const PAddr start_addr = config.GetStartAddress();
    const PAddr end_addr = config.GetEndAddress();

    // TODO: do hwtest with these cases
    if (!Memory::IsValidPhysicalAddress(start_addr)) {
        LOG_CRITICAL(HW_GPU, "invalid start address 0x%08X", start_addr);
        return;
    }

    if (!Memory::IsValidPhysicalAddress(end_addr)) {
        LOG_CRITICAL(HW_GPU, "invalid end address 0x%08X", end_addr);
        return;
    }

    if (end_addr <= start_addr) {
        LOG_CRITICAL(HW_GPU, "invalid memory range from 0x%08X to 0x%08X", start_addr, end_addr);
        return;
    }

    u8* start = Memory::GetPhysicalPointer(start_addr);
    u8* end = Memory::GetPhysicalPointer(end_addr);

    // TODO: Consider always accelerating and returning vector of
    //       regions that the accelerated fill did not cover to
    //       reduce/eliminate the fill that the cpu has to do.
    //       This would also mean that the flush below is not needed.
    //       Fill should first flush all surfaces that touch but are
    //       not completely within the fill range.
    //       Then fill all completely covered surfaces, and return the
    //       regions that were between surfaces or within the touching
    //       ones for cpu to manually fill here.
    if (VideoCore::g_renderer->Rasterizer()->AccelerateFill(config))
        return;

    Memory::RasterizerFlushAndInvalidateRegion(config.GetStartAddress(),
                                               config.GetEndAddress() - config.GetStartAddress());

    if (config.fill_24bit) {
        // fill with 24-bit values
        for (u8* ptr = start; ptr < end; ptr += 3) {
            ptr[0] = config.value_24bit_r;
            ptr[1] = config.value_24bit_g;
            ptr[2] = config.value_24bit_b;
        }
    } else if (config.fill_32bit) {
        // fill with 32-bit values
        if (end > start) {
            u32 value = config.value_32bit;
            size_t len = (end - start) / sizeof(u32);
            for (size_t i = 0; i < len; ++i)
                memcpy(&start[i * sizeof(u32)], &value, sizeof(u32));
        }
    } else {
        // fill with 16-bit values
        u16 value_16bit = config.value_16bit.Value();
        for (u8* ptr = start; ptr < end; ptr += sizeof(u16))
            memcpy(ptr, &value_16bit, sizeof(u16));
    }
}

static void DisplayTransfer(const Regs::DisplayTransferConfig& config) {
    const PAddr src_addr = config.GetPhysicalInputAddress();
    const PAddr dst_addr = config.GetPhysicalOutputAddress();

    // TODO: do hwtest with these cases
    if (!Memory::IsValidPhysicalAddress(src_addr)) {
        LOG_CRITICAL(HW_GPU, "invalid input address 0x%08X", src_addr);
        return;
    }

    if (!Memory::IsValidPhysicalAddress(dst_addr)) {
        LOG_CRITICAL(HW_GPU, "invalid output address 0x%08X", dst_addr);
        return;
    }

    if (config.input_width == 0) {
        LOG_CRITICAL(HW_GPU, "zero input width");
        return;
    }

    if (config.input_height == 0) {
        LOG_CRITICAL(HW_GPU, "zero input height");
        return;
    }

    if (config.output_width == 0) {
        LOG_CRITICAL(HW_GPU, "zero output width");
        return;
    }

    if (config.output_height == 0) {
        LOG_CRITICAL(HW_GPU, "zero output height");
        return;
    }

    if (VideoCore::g_renderer->Rasterizer()->AccelerateDisplayTransfer(config))
        return;

    u8* src_pointer = Memory::GetPhysicalPointer(src_addr);
    u8* dst_pointer = Memory::GetPhysicalPointer(dst_addr);

    if (config.scaling > config.ScaleXY) {
        LOG_CRITICAL(HW_GPU, "Unimplemented display transfer scaling mode %u",
                     config.scaling.Value());
        UNIMPLEMENTED();
        return;
    }

    if (config.input_linear && config.scaling != config.NoScale) {
        LOG_CRITICAL(HW_GPU, "Scaling is only implemented on tiled input");
        UNIMPLEMENTED();
        return;
    }

    int horizontal_scale = config.scaling != config.NoScale ? 1 : 0;
    int vertical_scale = config.scaling == config.ScaleXY ? 1 : 0;

    u32 output_width = config.output_width >> horizontal_scale;
    u32 output_height = config.output_height >> vertical_scale;

    u32 input_size =
        config.input_width * config.input_height * GPU::Regs::BytesPerPixel(config.input_format);
    u32 output_size = output_width * output_height * GPU::Regs::BytesPerPixel(config.output_format);

    Memory::RasterizerFlushRegion(config.GetPhysicalInputAddress(), input_size);
    Memory::RasterizerFlushAndInvalidateRegion(config.GetPhysicalOutputAddress(), output_size);

    for (u32 y = 0; y < output_height; ++y) {
        for (u32 x = 0; x < output_width; ++x) {
            Math::Vec4<u8> src_color;

            // Calculate the [x,y] position of the input image
            // based on the current output position and the scale
            u32 input_x = x << horizontal_scale;
            u32 input_y = y << vertical_scale;

            u32 output_y;
            if (config.flip_vertically) {
                // Flip the y value of the output data,
                // we do this after calculating the [x,y] position of the input image
                // to account for the scaling options.
                output_y = output_height - y - 1;
            } else {
                output_y = y;
            }

            u32 dst_bytes_per_pixel = GPU::Regs::BytesPerPixel(config.output_format);
            u32 src_bytes_per_pixel = GPU::Regs::BytesPerPixel(config.input_format);
            u32 src_offset;
            u32 dst_offset;

            if (config.input_linear) {
                if (!config.dont_swizzle) {
                    // Interpret the input as linear and the output as tiled
                    u32 coarse_y = output_y & ~7;
                    u32 stride = output_width * dst_bytes_per_pixel;

                    src_offset = (input_x + input_y * config.input_width) * src_bytes_per_pixel;
                    dst_offset = VideoCore::GetMortonOffset(x, output_y, dst_bytes_per_pixel) +
                                 coarse_y * stride;
                } else {
                    // Both input and output are linear
                    src_offset = (input_x + input_y * config.input_width) * src_bytes_per_pixel;
                    dst_offset = (x + output_y * output_width) * dst_bytes_per_pixel;
                }
            } else {
                if (!config.dont_swizzle) {
                    // Interpret the input as tiled and the output as linear
                    u32 coarse_y = input_y & ~7;
                    u32 stride = config.input_width * src_bytes_per_pixel;

                    src_offset = VideoCore::GetMortonOffset(input_x, input_y, src_bytes_per_pixel) +
                                 coarse_y * stride;
                    dst_offset = (x + output_y * output_width) * dst_bytes_per_pixel;
                } else {
                    // Both input and output are tiled
                    u32 out_coarse_y = output_y & ~7;
                    u32 out_stride = output_width * dst_bytes_per_pixel;

                    u32 in_coarse_y = input_y & ~7;
                    u32 in_stride = config.input_width * src_bytes_per_pixel;

                    src_offset = VideoCore::GetMortonOffset(input_x, input_y, src_bytes_per_pixel) +
                                 in_coarse_y * in_stride;
                    dst_offset = VideoCore::GetMortonOffset(x, output_y, dst_bytes_per_pixel) +
                                 out_coarse_y * out_stride;
                }
            }

            const u8* src_pixel = src_pointer + src_offset;
            src_color = DecodePixel(config.input_format, src_pixel);
            if (config.scaling == config.ScaleX) {
                Math::Vec4<u8> pixel =
                    DecodePixel(config.input_format, src_pixel + src_bytes_per_pixel);
                src_color = ((src_color + pixel) / 2).Cast<u8>();
            } else if (config.scaling == config.ScaleXY) {
                Math::Vec4<u8> pixel1 =
                    DecodePixel(config.input_format, src_pixel + 1 * src_bytes_per_pixel);
                Math::Vec4<u8> pixel2 =
                    DecodePixel(config.input_format, src_pixel + 2 * src_bytes_per_pixel);
                Math::Vec4<u8> pixel3 =
                    DecodePixel(config.input_format, src_pixel + 3 * src_bytes_per_pixel);
                src_color = (((src_color + pixel1) + (pixel2 + pixel3)) / 4).Cast<u8>();
            }

            u8* dst_pixel = dst_pointer + dst_offset;
            switch (config.output_format) {
            case Regs::PixelFormat::RGBA8:
                Color::EncodeRGBA8(src_color, dst_pixel);
                break;

            case Regs::PixelFormat::RGB8:
                Color::EncodeRGB8(src_color, dst_pixel);
                break;

            case Regs::PixelFormat::RGB565:
                Color::EncodeRGB565(src_color, dst_pixel);
                break;

            case Regs::PixelFormat::RGB5A1:
                Color::EncodeRGB5A1(src_color, dst_pixel);
                break;

            case Regs::PixelFormat::RGBA4:
                Color::EncodeRGBA4(src_color, dst_pixel);
                break;

            default:
                LOG_ERROR(HW_GPU, "Unknown destination framebuffer format %x",
                          config.output_format.Value());
                break;
            }
        }
    }
}

static void TextureCopy(const Regs::DisplayTransferConfig& config) {
    const PAddr src_addr = config.GetPhysicalInputAddress();
    const PAddr dst_addr = config.GetPhysicalOutputAddress();

    // TODO: do hwtest with these cases
    if (!Memory::IsValidPhysicalAddress(src_addr)) {
        LOG_CRITICAL(HW_GPU, "invalid input address 0x%08X", src_addr);
        return;
    }

    if (!Memory::IsValidPhysicalAddress(dst_addr)) {
        LOG_CRITICAL(HW_GPU, "invalid output address 0x%08X", dst_addr);
        return;
    }

    if (config.texture_copy.input_width == 0) {
        LOG_CRITICAL(HW_GPU, "zero input width");
        return;
    }

    if (config.texture_copy.output_width == 0) {
        LOG_CRITICAL(HW_GPU, "zero output width");
        return;
    }

    if (config.texture_copy.size == 0) {
        LOG_CRITICAL(HW_GPU, "zero size");
        return;
    }

    if (VideoCore::g_renderer->Rasterizer()->AccelerateTextureCopy(config))
        return;

    u8* src_pointer = Memory::GetPhysicalPointer(src_addr);
    u8* dst_pointer = Memory::GetPhysicalPointer(dst_addr);

    u32 input_width = config.texture_copy.input_width * 16;
    u32 input_gap = config.texture_copy.input_gap * 16;
    u32 output_width = config.texture_copy.output_width * 16;
    u32 output_gap = config.texture_copy.output_gap * 16;

    size_t contiguous_input_size =
        config.texture_copy.size / input_width * (input_width + input_gap);
    Memory::RasterizerFlushRegion(config.GetPhysicalInputAddress(),
                                  static_cast<u32>(contiguous_input_size));

    size_t contiguous_output_size =
        config.texture_copy.size / output_width * (output_width + output_gap);
    Memory::RasterizerFlushAndInvalidateRegion(config.GetPhysicalOutputAddress(),
                                               static_cast<u32>(contiguous_output_size));

    u32 remaining_size = config.texture_copy.size;
    u32 remaining_input = input_width;
    u32 remaining_output = output_width;
    while (remaining_size > 0) {
        u32 copy_size = std::min({remaining_input, remaining_output, remaining_size});

        std::memcpy(dst_pointer, src_pointer, copy_size);
        src_pointer += copy_size;
        dst_pointer += copy_size;

        remaining_input -= copy_size;
        remaining_output -= copy_size;
        remaining_size -= copy_size;

        if (remaining_input == 0) {
            remaining_input = input_width;
            src_pointer += input_gap;
        }
        if (remaining_output == 0) {
            remaining_output = output_width;
            dst_pointer += output_gap;
        }
    }
}

template <typename T>
inline void Write(u32 addr, const T data) {
    addr -= HW::VADDR_GPU;
    u32 index = addr / 4;

    // Writes other than u32 are untested, so I'd rather have them abort than silently fail
    if (index >= Regs::NumIds() || !std::is_same<T, u32>::value) {
        LOG_ERROR(HW_GPU, "unknown Write%lu 0x%08X @ 0x%08X", sizeof(data) * 8, (u32)data, addr);
        return;
    }

    g_regs[index] = static_cast<u32>(data);

    switch (index) {

    // Memory fills are triggered once the fill value is written.
    case GPU_REG_INDEX_WORKAROUND(memory_fill_config[0].trigger, 0x00004 + 0x3):
    case GPU_REG_INDEX_WORKAROUND(memory_fill_config[1].trigger, 0x00008 + 0x3): {
        const bool is_second_filler = (index != GPU_REG_INDEX(memory_fill_config[0].trigger));
        auto& config = g_regs.memory_fill_config[is_second_filler];

        if (config.trigger) {
            MemoryFill(config);
            LOG_TRACE(HW_GPU, "MemoryFill from 0x%08x to 0x%08x", config.GetStartAddress(),
                      config.GetEndAddress());

            // It seems that it won't signal interrupt if "address_start" is zero.
            // TODO: hwtest this
            if (config.GetStartAddress() != 0) {
                if (!is_second_filler) {
                    GSP_GPU::SignalInterrupt(GSP_GPU::InterruptId::PSC0);
                } else {
                    GSP_GPU::SignalInterrupt(GSP_GPU::InterruptId::PSC1);
                }
            }

            // Reset "trigger" flag and set the "finish" flag
            // NOTE: This was confirmed to happen on hardware even if "address_start" is zero.
            config.trigger.Assign(0);
            config.finished.Assign(1);
        }
        break;
    }

    case GPU_REG_INDEX(display_transfer_config.trigger): {
        MICROPROFILE_SCOPE(GPU_DisplayTransfer);

        const auto& config = g_regs.display_transfer_config;
        if (config.trigger & 1) {

            if (Pica::g_debug_context)
                Pica::g_debug_context->OnEvent(Pica::DebugContext::Event::IncomingDisplayTransfer,
                                               nullptr);

            if (config.is_texture_copy) {
                TextureCopy(config);
                LOG_TRACE(HW_GPU, "TextureCopy: 0x%X bytes from 0x%08X(%u+%u)-> "
                                  "0x%08X(%u+%u), flags 0x%08X",
                          config.texture_copy.size, config.GetPhysicalInputAddress(),
                          config.texture_copy.input_width * 16, config.texture_copy.input_gap * 16,
                          config.GetPhysicalOutputAddress(), config.texture_copy.output_width * 16,
                          config.texture_copy.output_gap * 16, config.flags);
            } else {
                DisplayTransfer(config);
                LOG_TRACE(HW_GPU, "DisplayTransfer: 0x%08x(%ux%u)-> "
                                  "0x%08x(%ux%u), dst format %x, flags 0x%08X",
                          config.GetPhysicalInputAddress(), config.input_width.Value(),
                          config.input_height.Value(), config.GetPhysicalOutputAddress(),
                          config.output_width.Value(), config.output_height.Value(),
                          config.output_format.Value(), config.flags);
            }

            g_regs.display_transfer_config.trigger = 0;
            GSP_GPU::SignalInterrupt(GSP_GPU::InterruptId::PPF);
        }
        break;
    }

    // Seems like writing to this register triggers processing
    case GPU_REG_INDEX(command_processor_config.trigger): {
        const auto& config = g_regs.command_processor_config;
        if (config.trigger & 1) {
            MICROPROFILE_SCOPE(GPU_CmdlistProcessing);

            u32* buffer = (u32*)Memory::GetPhysicalPointer(config.GetPhysicalAddress());

            if (Pica::g_debug_context && Pica::g_debug_context->recorder) {
                Pica::g_debug_context->recorder->MemoryAccessed(
                    (u8*)buffer, config.size * sizeof(u32), config.GetPhysicalAddress());
            }

            Pica::CommandProcessor::ProcessCommandList(buffer, config.size);

            g_regs.command_processor_config.trigger = 0;
        }
        break;
    }

    default:
        break;
    }

    // Notify tracer about the register write
    // This is happening *after* handling the write to make sure we properly catch all memory reads.
    if (Pica::g_debug_context && Pica::g_debug_context->recorder) {
        // addr + GPU VBase - IO VBase + IO PBase
        Pica::g_debug_context->recorder->RegisterWritten<T>(
            addr + 0x1EF00000 - 0x1EC00000 + 0x10100000, data);
    }
}

// Explicitly instantiate template functions because we aren't defining this in the header:

template void Read<u64>(u64& var, const u32 addr);
template void Read<u32>(u32& var, const u32 addr);
template void Read<u16>(u16& var, const u32 addr);
template void Read<u8>(u8& var, const u32 addr);

template void Write<u64>(u32 addr, const u64 data);
template void Write<u32>(u32 addr, const u32 data);
template void Write<u16>(u32 addr, const u16 data);
template void Write<u8>(u32 addr, const u8 data);

static void FrameLimiter() {
    time_delay += FIXED_FRAME_TIME;
    time_delay = MathUtil::Clamp(time_delay, -MAX_LAG_TIME, MAX_LAG_TIME);
    s32 desired_time = static_cast<s32>(time_delay);
    s32 elapsed_time = static_cast<s32>(Common::Timer::GetTimeMs() - time_point);

    if (elapsed_time < desired_time) {
        Common::SleepCurrentThread(desired_time - elapsed_time);
    }

    u32 frame_time = Common::Timer::GetTimeMs() - time_point;

    time_delay -= frame_time;
}

/// Update hardware
static void VBlankCallback(u64 userdata, int cycles_late) {
    frame_count++;
    VideoCore::g_renderer->SwapBuffers();

    // Signal to GSP that GPU interrupt has occurred
    // TODO(yuriks): hwtest to determine if PDC0 is for the Top screen and PDC1 for the Sub
    // screen, or if both use the same interrupts and these two instead determine the
    // beginning and end of the VBlank period. If needed, split the interrupt firing into
    // two different intervals.
    GSP_GPU::SignalInterrupt(GSP_GPU::InterruptId::PDC0);
    GSP_GPU::SignalInterrupt(GSP_GPU::InterruptId::PDC1);

    // Check for user input updates
    Service::HID::Update();

    if (!Settings::values.use_vsync && Settings::values.toggle_framelimit) {
        FrameLimiter();
    }

    time_point = Common::Timer::GetTimeMs();

    // Reschedule recurrent event
    CoreTiming::ScheduleEvent(frame_ticks - cycles_late, vblank_event);
}

/// Initialize hardware
void Init() {
    memset(&g_regs, 0, sizeof(g_regs));

    auto& framebuffer_top = g_regs.framebuffer_config[0];
    auto& framebuffer_sub = g_regs.framebuffer_config[1];

    // Setup default framebuffer addresses (located in VRAM)
    // .. or at least these are the ones used by system applets.
    // There's probably a smarter way to come up with addresses
    // like this which does not require hardcoding.
    framebuffer_top.address_left1 = 0x181E6000;
    framebuffer_top.address_left2 = 0x1822C800;
    framebuffer_top.address_right1 = 0x18273000;
    framebuffer_top.address_right2 = 0x182B9800;
    framebuffer_sub.address_left1 = 0x1848F000;
    framebuffer_sub.address_left2 = 0x184C7800;

    framebuffer_top.width.Assign(240);
    framebuffer_top.height.Assign(400);
    framebuffer_top.stride = 3 * 240;
    framebuffer_top.color_format.Assign(Regs::PixelFormat::RGB8);
    framebuffer_top.active_fb = 0;

    framebuffer_sub.width.Assign(240);
    framebuffer_sub.height.Assign(320);
    framebuffer_sub.stride = 3 * 240;
    framebuffer_sub.color_format.Assign(Regs::PixelFormat::RGB8);
    framebuffer_sub.active_fb = 0;

    frame_count = 0;
    time_point = Common::Timer::GetTimeMs();

    vblank_event = CoreTiming::RegisterEvent("GPU::VBlankCallback", VBlankCallback);
    CoreTiming::ScheduleEvent(frame_ticks, vblank_event);

    LOG_DEBUG(HW_GPU, "initialized OK");
}

/// Shutdown hardware
void Shutdown() {
    LOG_DEBUG(HW_GPU, "shutdown OK");
}

} // namespace