Duckstation/src/core/gpu.cpp

// SPDX-FileCopyrightText: 2019-2024 Connor McLaughlin <stenzek@gmail.com>
// SPDX-License-Identifier: (GPL-3.0 OR CC-BY-NC-ND-4.0)

#include "gpu.h"
#include "dma.h"
#include "gpu_shadergen.h"
#include "host.h"
#include "imgui.h"
#include "interrupt_controller.h"
#include "settings.h"
#include "system.h"
#include "timers.h"

#include "util/gpu_device.h"
#include "util/image.h"
#include "util/imgui_manager.h"
#include "util/postprocessing.h"
#include "util/shadergen.h"
#include "util/state_wrapper.h"

#include "common/align.h"
#include "common/file_system.h"
#include "common/heap_array.h"
#include "common/log.h"
#include "common/path.h"
#include "common/small_string.h"
#include "common/string_util.h"

#include "IconsFontAwesome5.h"
#include "fmt/format.h"

#include <cmath>
#include <thread>

Log_SetChannel(GPU);

std::unique_ptr<GPU> g_gpu;
alignas(HOST_PAGE_SIZE) u16 g_vram[VRAM_SIZE / sizeof(u16)];

const GPU::GP0CommandHandlerTable GPU::s_GP0_command_handler_table = GPU::GenerateGP0CommandHandlerTable();

static bool CompressAndWriteTextureToFile(u32 width, u32 height, std::string filename, FileSystem::ManagedCFilePtr fp,
                                          u8 quality, bool clear_alpha, bool flip_y, std::vector<u32> texture_data,
                                          u32 texture_data_stride, GPUTexture::Format texture_format,
                                          bool display_osd_message, bool use_thread);
static void JoinScreenshotThreads();

static std::deque<std::thread> s_screenshot_threads;
static std::mutex s_screenshot_threads_mutex;

GPU::GPU()
{
  ResetStatistics();
}

GPU::~GPU()
{
  JoinScreenshotThreads();
  DestroyDeinterlaceTextures();
  g_gpu_device->RecycleTexture(std::move(m_chroma_smoothing_texture));

  if (g_gpu_device)
    g_gpu_device->SetGPUTimingEnabled(false);
}

bool GPU::Initialize()
{
  m_force_progressive_scan = g_settings.gpu_disable_interlacing;
  m_force_ntsc_timings = g_settings.gpu_force_ntsc_timings;
  m_crtc_tick_event = TimingEvents::CreateTimingEvent(
    "GPU CRTC Tick", 1, 1,
    [](void* param, TickCount ticks, TickCount ticks_late) { static_cast<GPU*>(param)->CRTCTickEvent(ticks); }, this,
    true);
  m_command_tick_event = TimingEvents::CreateTimingEvent(
    "GPU Command Tick", 1, 1,
    [](void* param, TickCount ticks, TickCount ticks_late) { static_cast<GPU*>(param)->CommandTickEvent(ticks); }, this,
    true);
  m_fifo_size = g_settings.gpu_fifo_size;
  m_max_run_ahead = g_settings.gpu_max_run_ahead;
  m_console_is_pal = System::IsPALRegion();
  UpdateCRTCConfig();

  if (!CompileDisplayPipelines(true, true, g_settings.gpu_24bit_chroma_smoothing))
  {
    Host::ReportErrorAsync("Error", "Failed to compile base GPU pipelines.");
    return false;
  }

  g_gpu_device->SetGPUTimingEnabled(g_settings.display_show_gpu_usage);

  return true;
}

void GPU::UpdateSettings(const Settings& old_settings)
{
  FlushRender();

  m_force_progressive_scan = g_settings.gpu_disable_interlacing;
  m_fifo_size = g_settings.gpu_fifo_size;
  m_max_run_ahead = g_settings.gpu_max_run_ahead;

  if (m_force_ntsc_timings != g_settings.gpu_force_ntsc_timings || m_console_is_pal != System::IsPALRegion())
  {
    m_force_ntsc_timings = g_settings.gpu_force_ntsc_timings;
    m_console_is_pal = System::IsPALRegion();
    UpdateCRTCConfig();
  }

  // Crop mode calls this, so recalculate the display area
  UpdateCRTCDisplayParameters();

  if (g_settings.display_scaling != old_settings.display_scaling ||
      g_settings.display_deinterlacing_mode != old_settings.display_deinterlacing_mode ||
      g_settings.gpu_24bit_chroma_smoothing != old_settings.gpu_24bit_chroma_smoothing)
  {
    // Toss buffers on mode change.
    if (g_settings.display_deinterlacing_mode != old_settings.display_deinterlacing_mode)
      DestroyDeinterlaceTextures();

    if (!CompileDisplayPipelines(g_settings.display_scaling != old_settings.display_scaling,
                                 g_settings.display_deinterlacing_mode != old_settings.display_deinterlacing_mode,
                                 g_settings.gpu_24bit_chroma_smoothing != old_settings.gpu_24bit_chroma_smoothing))
    {
      Panic("Failed to compile display pipeline on settings change.");
    }
  }

  g_gpu_device->SetGPUTimingEnabled(g_settings.display_show_gpu_usage);
}

void GPU::CPUClockChanged()
{
  UpdateCRTCConfig();
}

void GPU::UpdateResolutionScale()
{
}

std::tuple<u32, u32> GPU::GetEffectiveDisplayResolution(bool scaled /* = true */)
{
  return std::tie(m_crtc_state.display_vram_width, m_crtc_state.display_vram_height);
}

std::tuple<u32, u32> GPU::GetFullDisplayResolution(bool scaled /* = true */)
{
  return std::tie(m_crtc_state.display_width, m_crtc_state.display_height);
}

void GPU::Reset(bool clear_vram)
{
  m_GPUSTAT.bits = 0x14802000;
  m_set_texture_disable_mask = false;
  m_GPUREAD_latch = 0;
  m_crtc_state.fractional_ticks = 0;
  m_crtc_state.fractional_dot_ticks = 0;
  m_crtc_state.current_tick_in_scanline = 0;
  m_crtc_state.current_scanline = 0;
  m_crtc_state.in_hblank = false;
  m_crtc_state.in_vblank = false;
  m_crtc_state.interlaced_field = 0;
  m_crtc_state.interlaced_display_field = 0;

  if (clear_vram)
    std::memset(g_vram, 0, sizeof(g_vram));

  SoftReset();
  UpdateDisplay();
}

void GPU::SoftReset()
{
  FlushRender();
  if (m_blitter_state == BlitterState::WritingVRAM)
    FinishVRAMWrite();

  m_GPUSTAT.texture_page_x_base = 0;
  m_GPUSTAT.texture_page_y_base = 0;
  m_GPUSTAT.semi_transparency_mode = GPUTransparencyMode::HalfBackgroundPlusHalfForeground;
  m_GPUSTAT.texture_color_mode = GPUTextureMode::Palette4Bit;
  m_GPUSTAT.dither_enable = false;
  m_GPUSTAT.draw_to_displayed_field = false;
  m_GPUSTAT.set_mask_while_drawing = false;
  m_GPUSTAT.check_mask_before_draw = false;
  m_GPUSTAT.reverse_flag = false;
  m_GPUSTAT.texture_disable = false;
  m_GPUSTAT.horizontal_resolution_2 = 0;
  m_GPUSTAT.horizontal_resolution_1 = 0;
  m_GPUSTAT.vertical_resolution = false;
  m_GPUSTAT.pal_mode = System::IsPALRegion();
  m_GPUSTAT.display_area_color_depth_24 = false;
  m_GPUSTAT.vertical_interlace = false;
  m_GPUSTAT.display_disable = true;
  m_GPUSTAT.dma_direction = DMADirection::Off;
  m_drawing_area.Set(0, 0, 0, 0);
  m_drawing_area_changed = true;
  m_drawing_offset = {};
  std::memset(&m_crtc_state.regs, 0, sizeof(m_crtc_state.regs));
  m_crtc_state.regs.horizontal_display_range = 0xC60260;
  m_crtc_state.regs.vertical_display_range = 0x3FC10;
  m_blitter_state = BlitterState::Idle;
  m_pending_command_ticks = 0;
  m_command_total_words = 0;
  m_vram_transfer = {};
  m_fifo.Clear();
  m_blit_buffer.clear();
  m_blit_remaining_words = 0;
  m_draw_mode.texture_window_value = 0xFFFFFFFFu;
  SetDrawMode(0);
  SetTexturePalette(0);
  SetTextureWindow(0);
  UpdateDMARequest();
  UpdateCRTCConfig();
  UpdateCRTCTickEvent();
  UpdateCommandTickEvent();
  UpdateGPUIdle();
}

bool GPU::DoState(StateWrapper& sw, GPUTexture** host_texture, bool update_display)
{
  FlushRender();

  if (sw.IsReading())
  {
    // perform a reset to discard all pending draws/fb state
    Reset(host_texture == nullptr);
  }

  sw.Do(&m_GPUSTAT.bits);

  sw.Do(&m_draw_mode.mode_reg.bits);
  sw.Do(&m_draw_mode.palette_reg.bits);
  sw.Do(&m_draw_mode.texture_window_value);

  if (sw.GetVersion() < 62)
  {
    // texture_page_x, texture_page_y, texture_palette_x, texture_palette_y
    DebugAssert(sw.IsReading());
    sw.SkipBytes(sizeof(u32) * 4);
  }

  sw.Do(&m_draw_mode.texture_window.and_x);
  sw.Do(&m_draw_mode.texture_window.and_y);
  sw.Do(&m_draw_mode.texture_window.or_x);
  sw.Do(&m_draw_mode.texture_window.or_y);
  sw.Do(&m_draw_mode.texture_x_flip);
  sw.Do(&m_draw_mode.texture_y_flip);

  sw.Do(&m_drawing_area.left);
  sw.Do(&m_drawing_area.top);
  sw.Do(&m_drawing_area.right);
  sw.Do(&m_drawing_area.bottom);
  sw.Do(&m_drawing_offset.x);
  sw.Do(&m_drawing_offset.y);
  sw.Do(&m_drawing_offset.x);

  sw.Do(&m_console_is_pal);
  sw.Do(&m_set_texture_disable_mask);

  sw.Do(&m_crtc_state.regs.display_address_start);
  sw.Do(&m_crtc_state.regs.horizontal_display_range);
  sw.Do(&m_crtc_state.regs.vertical_display_range);
  sw.Do(&m_crtc_state.dot_clock_divider);
  sw.Do(&m_crtc_state.display_width);
  sw.Do(&m_crtc_state.display_height);
  sw.Do(&m_crtc_state.display_origin_left);
  sw.Do(&m_crtc_state.display_origin_top);
  sw.Do(&m_crtc_state.display_vram_left);
  sw.Do(&m_crtc_state.display_vram_top);
  sw.Do(&m_crtc_state.display_vram_width);
  sw.Do(&m_crtc_state.display_vram_height);
  sw.Do(&m_crtc_state.horizontal_total);
  sw.Do(&m_crtc_state.horizontal_visible_start);
  sw.Do(&m_crtc_state.horizontal_visible_end);
  sw.Do(&m_crtc_state.horizontal_display_start);
  sw.Do(&m_crtc_state.horizontal_display_end);
  sw.Do(&m_crtc_state.vertical_total);
  sw.Do(&m_crtc_state.vertical_visible_start);
  sw.Do(&m_crtc_state.vertical_visible_end);
  sw.Do(&m_crtc_state.vertical_display_start);
  sw.Do(&m_crtc_state.vertical_display_end);
  sw.Do(&m_crtc_state.fractional_ticks);
  sw.Do(&m_crtc_state.current_tick_in_scanline);
  sw.Do(&m_crtc_state.current_scanline);
  sw.DoEx(&m_crtc_state.fractional_dot_ticks, 46, 0);
  sw.Do(&m_crtc_state.in_hblank);
  sw.Do(&m_crtc_state.in_vblank);
  sw.Do(&m_crtc_state.interlaced_field);
  sw.Do(&m_crtc_state.interlaced_display_field);
  sw.Do(&m_crtc_state.active_line_lsb);

  sw.Do(&m_blitter_state);
  sw.Do(&m_pending_command_ticks);
  sw.Do(&m_command_total_words);
  sw.Do(&m_GPUREAD_latch);

  sw.Do(&m_vram_transfer.x);
  sw.Do(&m_vram_transfer.y);
  sw.Do(&m_vram_transfer.width);
  sw.Do(&m_vram_transfer.height);
  sw.Do(&m_vram_transfer.col);
  sw.Do(&m_vram_transfer.row);

  sw.Do(&m_fifo);
  sw.Do(&m_blit_buffer);
  sw.Do(&m_blit_remaining_words);
  sw.Do(&m_render_command.bits);

  sw.Do(&m_max_run_ahead);
  sw.Do(&m_fifo_size);

  if (sw.IsReading())
  {
    m_draw_mode.texture_page_changed = true;
    m_draw_mode.texture_window_changed = true;
    m_drawing_area_changed = true;
    UpdateDMARequest();
  }

  if (!host_texture)
  {
    if (!sw.DoMarker("GPU-VRAM"))
      return false;

    if (sw.IsReading())
    {
      // Still need a temporary here.
      FixedHeapArray<u16, VRAM_WIDTH * VRAM_HEIGHT> temp;
      sw.DoBytes(temp.data(), VRAM_WIDTH * VRAM_HEIGHT * sizeof(u16));
      UpdateVRAM(0, 0, VRAM_WIDTH, VRAM_HEIGHT, temp.data(), false, false);
    }
    else
    {
      ReadVRAM(0, 0, VRAM_WIDTH, VRAM_HEIGHT);
      sw.DoBytes(g_vram, VRAM_WIDTH * VRAM_HEIGHT * sizeof(u16));
    }
  }

  if (sw.IsReading())
  {
    UpdateCRTCConfig();
    if (update_display)
      UpdateDisplay();

    UpdateCRTCTickEvent();
    UpdateCommandTickEvent();
  }

  return !sw.HasError();
}

void GPU::RestoreDeviceContext()
{
}

void GPU::UpdateDMARequest()
{
  switch (m_blitter_state)
  {
    case BlitterState::Idle:
      m_GPUSTAT.ready_to_send_vram = false;
      m_GPUSTAT.ready_to_recieve_dma = (m_fifo.IsEmpty() || m_fifo.GetSize() < m_command_total_words);
      break;

    case BlitterState::WritingVRAM:
      m_GPUSTAT.ready_to_send_vram = false;
      m_GPUSTAT.ready_to_recieve_dma = (m_fifo.GetSize() < m_fifo_size);
      break;

    case BlitterState::ReadingVRAM:
      m_GPUSTAT.ready_to_send_vram = true;
      m_GPUSTAT.ready_to_recieve_dma = m_fifo.IsEmpty();
      break;

    case BlitterState::DrawingPolyLine:
      m_GPUSTAT.ready_to_send_vram = false;
      m_GPUSTAT.ready_to_recieve_dma = (m_fifo.GetSize() < m_fifo_size);
      break;

    default:
      UnreachableCode();
      break;
  }

  bool dma_request;
  switch (m_GPUSTAT.dma_direction)
  {
    case DMADirection::Off:
      dma_request = false;
      break;

    case DMADirection::FIFO:
      dma_request = m_GPUSTAT.ready_to_recieve_dma;
      break;

    case DMADirection::CPUtoGP0:
      dma_request = m_GPUSTAT.ready_to_recieve_dma;
      break;

    case DMADirection::GPUREADtoCPU:
      dma_request = m_GPUSTAT.ready_to_send_vram;
      break;

    default:
      dma_request = false;
      break;
  }
  m_GPUSTAT.dma_data_request = dma_request;
  DMA::SetRequest(DMA::Channel::GPU, dma_request);
}

void GPU::UpdateGPUIdle()
{
  switch (m_blitter_state)
  {
    case BlitterState::Idle:
      m_GPUSTAT.gpu_idle = (m_pending_command_ticks <= 0 && m_fifo.IsEmpty());
      break;

    case BlitterState::WritingVRAM:
      m_GPUSTAT.gpu_idle = false;
      break;

    case BlitterState::ReadingVRAM:
      m_GPUSTAT.gpu_idle = false;
      break;

    case BlitterState::DrawingPolyLine:
      m_GPUSTAT.gpu_idle = false;
      break;

    default:
      UnreachableCode();
      break;
  }
}

u32 GPU::ReadRegister(u32 offset)
{
  switch (offset)
  {
    case 0x00:
      return ReadGPUREAD();

    case 0x04:
    {
      // code can be dependent on the odd/even bit, so update the GPU state when reading.
      // we can mitigate this slightly by only updating when the raster is actually hitting a new line
      if (IsCRTCScanlinePending())
        SynchronizeCRTC();
      if (IsCommandCompletionPending())
        m_command_tick_event->InvokeEarly();

      return m_GPUSTAT.bits;
    }

    default:
      Log_ErrorPrintf("Unhandled register read: %02X", offset);
      return UINT32_C(0xFFFFFFFF);
  }
}

void GPU::WriteRegister(u32 offset, u32 value)
{
  switch (offset)
  {
    case 0x00:
      m_fifo.Push(value);
      ExecuteCommands();
      UpdateCommandTickEvent();
      return;

    case 0x04:
      WriteGP1(value);
      return;

    default:
      Log_ErrorPrintf("Unhandled register write: %02X <- %08X", offset, value);
      return;
  }
}

void GPU::DMARead(u32* words, u32 word_count)
{
  if (m_GPUSTAT.dma_direction != DMADirection::GPUREADtoCPU)
  {
    Log_ErrorPrintf("Invalid DMA direction from GPU DMA read");
    std::fill_n(words, word_count, UINT32_C(0xFFFFFFFF));
    return;
  }

  for (u32 i = 0; i < word_count; i++)
    words[i] = ReadGPUREAD();
}

void GPU::EndDMAWrite()
{
  m_fifo_pushed = true;
  if (!m_syncing)
  {
    ExecuteCommands();
    UpdateCommandTickEvent();
  }
  else
  {
    UpdateDMARequest();
  }
}

/**
 * NTSC GPU clock 53.693175 MHz
 * PAL GPU clock 53.203425 MHz
 * courtesy of @ggrtk
 *
 * NTSC - sysclk * 715909 / 451584
 * PAL - sysclk * 709379 / 451584
 */

TickCount GPU::GetCRTCFrequency() const
{
  return m_console_is_pal ? 53203425 : 53693175;
}

TickCount GPU::CRTCTicksToSystemTicks(TickCount gpu_ticks, TickCount fractional_ticks) const
{
  // convert to master clock, rounding up as we want to overshoot not undershoot
  if (!m_console_is_pal)
    return static_cast<TickCount>((u64(gpu_ticks) * u64(451584) + fractional_ticks + u64(715908)) / u64(715909));
  else
    return static_cast<TickCount>((u64(gpu_ticks) * u64(451584) + fractional_ticks + u64(709378)) / u64(709379));
}

TickCount GPU::SystemTicksToCRTCTicks(TickCount sysclk_ticks, TickCount* fractional_ticks) const
{
  u64 mul = u64(sysclk_ticks);
  mul *= !m_console_is_pal ? u64(715909) : u64(709379);
  mul += u64(*fractional_ticks);

  const TickCount ticks = static_cast<TickCount>(mul / u64(451584));
  *fractional_ticks = static_cast<TickCount>(mul % u64(451584));
  return ticks;
}

void GPU::AddCommandTicks(TickCount ticks)
{
  m_pending_command_ticks += ticks;
}

void GPU::SynchronizeCRTC()
{
  m_crtc_tick_event->InvokeEarly();
}

float GPU::ComputeHorizontalFrequency() const
{
  const CRTCState& cs = m_crtc_state;
  TickCount fractional_ticks = 0;
  return static_cast<float>(
    static_cast<double>(SystemTicksToCRTCTicks(System::GetTicksPerSecond(), &fractional_ticks)) /
    static_cast<double>(cs.horizontal_total));
}

float GPU::ComputeVerticalFrequency() const
{
  const CRTCState& cs = m_crtc_state;
  const TickCount ticks_per_frame = cs.horizontal_total * cs.vertical_total;
  TickCount fractional_ticks = 0;
  return static_cast<float>(
    static_cast<double>(SystemTicksToCRTCTicks(System::GetTicksPerSecond(), &fractional_ticks)) /
    static_cast<double>(ticks_per_frame));
}

float GPU::ComputeDisplayAspectRatio() const
{
  if (g_settings.display_force_4_3_for_24bit && m_GPUSTAT.display_area_color_depth_24)
  {
    return 4.0f / 3.0f;
  }
  else if (g_settings.display_aspect_ratio == DisplayAspectRatio::Auto)
  {
    const CRTCState& cs = m_crtc_state;
    float relative_width = static_cast<float>(cs.horizontal_visible_end - cs.horizontal_visible_start);
    float relative_height = static_cast<float>(cs.vertical_visible_end - cs.vertical_visible_start);

    if (relative_width <= 0 || relative_height <= 0)
      return 4.0f / 3.0f;

    if (m_GPUSTAT.pal_mode)
    {
      relative_width /= static_cast<float>(PAL_HORIZONTAL_ACTIVE_END - PAL_HORIZONTAL_ACTIVE_START);
      relative_height /= static_cast<float>(PAL_VERTICAL_ACTIVE_END - PAL_VERTICAL_ACTIVE_START);
    }
    else
    {
      relative_width /= static_cast<float>(NTSC_HORIZONTAL_ACTIVE_END - NTSC_HORIZONTAL_ACTIVE_START);
      relative_height /= static_cast<float>(NTSC_VERTICAL_ACTIVE_END - NTSC_VERTICAL_ACTIVE_START);
    }
    return (relative_width / relative_height) * (4.0f / 3.0f);
  }
  else if (g_settings.display_aspect_ratio == DisplayAspectRatio::PAR1_1)
  {
    if (m_crtc_state.display_width == 0 || m_crtc_state.display_height == 0)
      return 4.0f / 3.0f;

    return static_cast<float>(m_crtc_state.display_width) / static_cast<float>(m_crtc_state.display_height);
  }
  else
  {
    return g_settings.GetDisplayAspectRatioValue();
  }
}

void GPU::UpdateCRTCConfig()
{
  static constexpr std::array<u16, 8> dot_clock_dividers = {{10, 8, 5, 4, 7, 7, 7, 7}};
  CRTCState& cs = m_crtc_state;

  if (m_GPUSTAT.pal_mode)
  {
    cs.vertical_total = PAL_TOTAL_LINES;
    cs.current_scanline %= PAL_TOTAL_LINES;
    cs.horizontal_total = PAL_TICKS_PER_LINE;
    cs.horizontal_sync_start = PAL_HSYNC_TICKS;
    cs.current_tick_in_scanline %= System::ScaleTicksToOverclock(PAL_TICKS_PER_LINE);
  }
  else
  {
    cs.vertical_total = NTSC_TOTAL_LINES;
    cs.current_scanline %= NTSC_TOTAL_LINES;
    cs.horizontal_total = NTSC_TICKS_PER_LINE;
    cs.horizontal_sync_start = NTSC_HSYNC_TICKS;
    cs.current_tick_in_scanline %= System::ScaleTicksToOverclock(NTSC_TICKS_PER_LINE);
  }

  cs.in_hblank = (cs.current_tick_in_scanline >= cs.horizontal_sync_start);

  const u8 horizontal_resolution_index = m_GPUSTAT.horizontal_resolution_1 | (m_GPUSTAT.horizontal_resolution_2 << 2);
  cs.dot_clock_divider = dot_clock_dividers[horizontal_resolution_index];
  cs.horizontal_display_start =
    (std::min<u16>(cs.regs.X1, cs.horizontal_total) / cs.dot_clock_divider) * cs.dot_clock_divider;
  cs.horizontal_display_end =
    (std::min<u16>(cs.regs.X2, cs.horizontal_total) / cs.dot_clock_divider) * cs.dot_clock_divider;
  cs.vertical_display_start = std::min<u16>(cs.regs.Y1, cs.vertical_total);
  cs.vertical_display_end = std::min<u16>(cs.regs.Y2, cs.vertical_total);

  if (m_GPUSTAT.pal_mode && m_force_ntsc_timings)
  {
    // scale to NTSC parameters
    cs.horizontal_display_start =
      static_cast<u16>((static_cast<u32>(cs.horizontal_display_start) * NTSC_TICKS_PER_LINE) / PAL_TICKS_PER_LINE);
    cs.horizontal_display_end = static_cast<u16>(
      ((static_cast<u32>(cs.horizontal_display_end) * NTSC_TICKS_PER_LINE) + (PAL_TICKS_PER_LINE - 1)) /
      PAL_TICKS_PER_LINE);
    cs.vertical_display_start =
      static_cast<u16>((static_cast<u32>(cs.vertical_display_start) * NTSC_TOTAL_LINES) / PAL_TOTAL_LINES);
    cs.vertical_display_end = static_cast<u16>(
      ((static_cast<u32>(cs.vertical_display_end) * NTSC_TOTAL_LINES) + (PAL_TOTAL_LINES - 1)) / PAL_TOTAL_LINES);

    cs.vertical_total = NTSC_TOTAL_LINES;
    cs.current_scanline %= NTSC_TOTAL_LINES;
    cs.horizontal_total = NTSC_TICKS_PER_LINE;
    cs.current_tick_in_scanline %= NTSC_TICKS_PER_LINE;
  }

  cs.horizontal_display_start =
    static_cast<u16>(System::ScaleTicksToOverclock(static_cast<TickCount>(cs.horizontal_display_start)));
  cs.horizontal_display_end =
    static_cast<u16>(System::ScaleTicksToOverclock(static_cast<TickCount>(cs.horizontal_display_end)));
  cs.horizontal_total = static_cast<u16>(System::ScaleTicksToOverclock(static_cast<TickCount>(cs.horizontal_total)));

  System::SetThrottleFrequency(ComputeVerticalFrequency());

  UpdateCRTCDisplayParameters();
  UpdateCRTCTickEvent();
}

void GPU::UpdateCRTCDisplayParameters()
{
  CRTCState& cs = m_crtc_state;
  const DisplayCropMode crop_mode = g_settings.display_crop_mode;

  const u16 horizontal_total = m_GPUSTAT.pal_mode ? PAL_TICKS_PER_LINE : NTSC_TICKS_PER_LINE;
  const u16 vertical_total = m_GPUSTAT.pal_mode ? PAL_TOTAL_LINES : NTSC_TOTAL_LINES;
  const u16 horizontal_display_start =
    (std::min<u16>(cs.regs.X1, horizontal_total) / cs.dot_clock_divider) * cs.dot_clock_divider;
  const u16 horizontal_display_end =
    (std::min<u16>(cs.regs.X2, horizontal_total) / cs.dot_clock_divider) * cs.dot_clock_divider;
  const u16 vertical_display_start = std::min<u16>(cs.regs.Y1, vertical_total);
  const u16 vertical_display_end = std::min<u16>(cs.regs.Y2, vertical_total);

  if (m_GPUSTAT.pal_mode)
  {
    // TODO: Verify PAL numbers.
    switch (crop_mode)
    {
      case DisplayCropMode::None:
        cs.horizontal_visible_start = PAL_HORIZONTAL_ACTIVE_START;
        cs.horizontal_visible_end = PAL_HORIZONTAL_ACTIVE_END;
        cs.vertical_visible_start = PAL_VERTICAL_ACTIVE_START;
        cs.vertical_visible_end = PAL_VERTICAL_ACTIVE_END;
        break;

      case DisplayCropMode::Overscan:
        cs.horizontal_visible_start = static_cast<u16>(std::max<int>(0, 628 + g_settings.display_active_start_offset));
        cs.horizontal_visible_end =
          static_cast<u16>(std::max<int>(cs.horizontal_visible_start, 3188 + g_settings.display_active_end_offset));
        cs.vertical_visible_start = static_cast<u16>(std::max<int>(0, 30 + g_settings.display_line_start_offset));
        cs.vertical_visible_end =
          static_cast<u16>(std::max<int>(cs.vertical_visible_start, 298 + g_settings.display_line_end_offset));
        break;

      case DisplayCropMode::Borders:
      default:
        cs.horizontal_visible_start = horizontal_display_start;
        cs.horizontal_visible_end = horizontal_display_end;
        cs.vertical_visible_start = vertical_display_start;
        cs.vertical_visible_end = vertical_display_end;
        break;
    }
    cs.horizontal_visible_start =
      std::clamp<u16>(cs.horizontal_visible_start, PAL_HORIZONTAL_ACTIVE_START, PAL_HORIZONTAL_ACTIVE_END);
    cs.horizontal_visible_end =
      std::clamp<u16>(cs.horizontal_visible_end, cs.horizontal_visible_start, PAL_HORIZONTAL_ACTIVE_END);
    cs.vertical_visible_start =
      std::clamp<u16>(cs.vertical_visible_start, PAL_VERTICAL_ACTIVE_START, PAL_VERTICAL_ACTIVE_END);
    cs.vertical_visible_end =
      std::clamp<u16>(cs.vertical_visible_end, cs.vertical_visible_start, PAL_VERTICAL_ACTIVE_END);
  }
  else
  {
    switch (crop_mode)
    {
      case DisplayCropMode::None:
        cs.horizontal_visible_start = NTSC_HORIZONTAL_ACTIVE_START;
        cs.horizontal_visible_end = NTSC_HORIZONTAL_ACTIVE_END;
        cs.vertical_visible_start = NTSC_VERTICAL_ACTIVE_START;
        cs.vertical_visible_end = NTSC_VERTICAL_ACTIVE_END;
        break;

      case DisplayCropMode::Overscan:
        cs.horizontal_visible_start = static_cast<u16>(std::max<int>(0, 608 + g_settings.display_active_start_offset));
        cs.horizontal_visible_end =
          static_cast<u16>(std::max<int>(cs.horizontal_visible_start, 3168 + g_settings.display_active_end_offset));
        cs.vertical_visible_start = static_cast<u16>(std::max<int>(0, 24 + g_settings.display_line_start_offset));
        cs.vertical_visible_end =
          static_cast<u16>(std::max<int>(cs.vertical_visible_start, 248 + g_settings.display_line_end_offset));
        break;

      case DisplayCropMode::Borders:
      default:
        cs.horizontal_visible_start = horizontal_display_start;
        cs.horizontal_visible_end = horizontal_display_end;
        cs.vertical_visible_start = vertical_display_start;
        cs.vertical_visible_end = vertical_display_end;
        break;
    }
    cs.horizontal_visible_start =
      std::clamp<u16>(cs.horizontal_visible_start, NTSC_HORIZONTAL_ACTIVE_START, NTSC_HORIZONTAL_ACTIVE_END);
    cs.horizontal_visible_end =
      std::clamp<u16>(cs.horizontal_visible_end, cs.horizontal_visible_start, NTSC_HORIZONTAL_ACTIVE_END);
    cs.vertical_visible_start =
      std::clamp<u16>(cs.vertical_visible_start, NTSC_VERTICAL_ACTIVE_START, NTSC_VERTICAL_ACTIVE_END);
    cs.vertical_visible_end =
      std::clamp<u16>(cs.vertical_visible_end, cs.vertical_visible_start, NTSC_VERTICAL_ACTIVE_END);
  }

  // If force-progressive is enabled, we only double the height in 480i mode. This way non-interleaved 480i framebuffers
  // won't be broken when displayed.
  const u8 y_shift = BoolToUInt8(m_GPUSTAT.vertical_interlace && m_GPUSTAT.vertical_resolution);
  const u8 height_shift = m_force_progressive_scan ? y_shift : BoolToUInt8(m_GPUSTAT.vertical_interlace);

  // Determine screen size.
  cs.display_width = (cs.horizontal_visible_end - cs.horizontal_visible_start) / cs.dot_clock_divider;
  cs.display_height = (cs.vertical_visible_end - cs.vertical_visible_start) << height_shift;

  // Determine number of pixels outputted from VRAM (in general, round to 4-pixel multiple).
  // TODO: Verify behavior if values are outside of the active video portion of scanline.
  const u16 horizontal_display_ticks =
    (horizontal_display_end < horizontal_display_start) ? 0 : (horizontal_display_end - horizontal_display_start);

  const u16 horizontal_display_pixels = horizontal_display_ticks / cs.dot_clock_divider;
  if (horizontal_display_pixels == 1u)
    cs.display_vram_width = 4u;
  else
    cs.display_vram_width = (horizontal_display_pixels + 2u) & ~3u;

  // Determine if we need to adjust the VRAM rectangle (because the display is starting outside the visible area) or add
  // padding.
  u16 horizontal_skip_pixels;
  if (horizontal_display_start >= cs.horizontal_visible_start)
  {
    cs.display_origin_left = (horizontal_display_start - cs.horizontal_visible_start) / cs.dot_clock_divider;
    cs.display_vram_left = cs.regs.X;
    horizontal_skip_pixels = 0;
  }
  else
  {
    horizontal_skip_pixels = (cs.horizontal_visible_start - horizontal_display_start) / cs.dot_clock_divider;
    cs.display_origin_left = 0;
    cs.display_vram_left = (cs.regs.X + horizontal_skip_pixels) % VRAM_WIDTH;
  }

  // apply the crop from the start (usually overscan)
  cs.display_vram_width -= std::min(cs.display_vram_width, horizontal_skip_pixels);

  // Apply crop from the end by shrinking VRAM rectangle width if display would end outside the visible area.
  cs.display_vram_width = std::min<u16>(cs.display_vram_width, cs.display_width - cs.display_origin_left);

  if (vertical_display_start >= cs.vertical_visible_start)
  {
    cs.display_origin_top = (vertical_display_start - cs.vertical_visible_start) << y_shift;
    cs.display_vram_top = cs.regs.Y;
  }
  else
  {
    cs.display_origin_top = 0;
    cs.display_vram_top = (cs.regs.Y + ((cs.vertical_visible_start - vertical_display_start) << y_shift)) % VRAM_HEIGHT;
  }

  if (vertical_display_end <= cs.vertical_visible_end)
  {
    cs.display_vram_height =
      (vertical_display_end -
       std::min(vertical_display_end, std::max(vertical_display_start, cs.vertical_visible_start)))
      << height_shift;
  }
  else
  {
    cs.display_vram_height =
      (cs.vertical_visible_end -
       std::min(cs.vertical_visible_end, std::max(vertical_display_start, cs.vertical_visible_start)))
      << height_shift;
  }
}

TickCount GPU::GetPendingCRTCTicks() const
{
  const TickCount pending_sysclk_ticks = m_crtc_tick_event->GetTicksSinceLastExecution();
  TickCount fractional_ticks = m_crtc_state.fractional_ticks;
  return SystemTicksToCRTCTicks(pending_sysclk_ticks, &fractional_ticks);
}

TickCount GPU::GetPendingCommandTicks() const
{
  if (!m_command_tick_event->IsActive())
    return 0;

  return SystemTicksToGPUTicks(m_command_tick_event->GetTicksSinceLastExecution());
}

void GPU::UpdateCRTCTickEvent()
{
  // figure out how many GPU ticks until the next vblank or event
  TickCount lines_until_event;
  if (Timers::IsSyncEnabled(HBLANK_TIMER_INDEX))
  {
    // when the timer sync is enabled we need to sync at vblank start and end
    lines_until_event =
      (m_crtc_state.current_scanline >= m_crtc_state.vertical_display_end) ?
        (m_crtc_state.vertical_total - m_crtc_state.current_scanline + m_crtc_state.vertical_display_start) :
        (m_crtc_state.vertical_display_end - m_crtc_state.current_scanline);
  }
  else
  {
    lines_until_event =
      (m_crtc_state.current_scanline >= m_crtc_state.vertical_display_end ?
         (m_crtc_state.vertical_total - m_crtc_state.current_scanline + m_crtc_state.vertical_display_end) :
         (m_crtc_state.vertical_display_end - m_crtc_state.current_scanline));
  }
  if (Timers::IsExternalIRQEnabled(HBLANK_TIMER_INDEX))
    lines_until_event = std::min(lines_until_event, Timers::GetTicksUntilIRQ(HBLANK_TIMER_INDEX));

  TickCount ticks_until_event =
    lines_until_event * m_crtc_state.horizontal_total - m_crtc_state.current_tick_in_scanline;
  if (Timers::IsExternalIRQEnabled(DOT_TIMER_INDEX))
  {
    const TickCount dots_until_irq = Timers::GetTicksUntilIRQ(DOT_TIMER_INDEX);
    const TickCount ticks_until_irq =
      (dots_until_irq * m_crtc_state.dot_clock_divider) - m_crtc_state.fractional_dot_ticks;
    ticks_until_event = std::min(ticks_until_event, std::max<TickCount>(ticks_until_irq, 0));
  }

#if 0
  const TickCount ticks_until_hblank =
    (m_crtc_state.current_tick_in_scanline >= m_crtc_state.horizontal_display_end) ?
    (m_crtc_state.horizontal_total - m_crtc_state.current_tick_in_scanline + m_crtc_state.horizontal_display_end) :
    (m_crtc_state.horizontal_display_end - m_crtc_state.current_tick_in_scanline);
#endif

  m_crtc_tick_event->Schedule(CRTCTicksToSystemTicks(ticks_until_event, m_crtc_state.fractional_ticks));
}

bool GPU::IsCRTCScanlinePending() const
{
  const TickCount ticks = (GetPendingCRTCTicks() + m_crtc_state.current_tick_in_scanline);
  return (ticks >= (m_crtc_state.in_hblank ? m_crtc_state.horizontal_total : m_crtc_state.horizontal_sync_start));
}

bool GPU::IsCommandCompletionPending() const
{
  return (m_pending_command_ticks > 0 && GetPendingCommandTicks() >= m_pending_command_ticks);
}

void GPU::CRTCTickEvent(TickCount ticks)
{
  // convert cpu/master clock to GPU ticks, accounting for partial cycles because of the non-integer divider
  {
    const TickCount gpu_ticks = SystemTicksToCRTCTicks(ticks, &m_crtc_state.fractional_ticks);
    m_crtc_state.current_tick_in_scanline += gpu_ticks;

    if (Timers::IsUsingExternalClock(DOT_TIMER_INDEX))
    {
      m_crtc_state.fractional_dot_ticks += gpu_ticks;
      const TickCount dots = m_crtc_state.fractional_dot_ticks / m_crtc_state.dot_clock_divider;
      m_crtc_state.fractional_dot_ticks = m_crtc_state.fractional_dot_ticks % m_crtc_state.dot_clock_divider;
      if (dots > 0)
        Timers::AddTicks(DOT_TIMER_INDEX, dots);
    }
  }

  if (m_crtc_state.current_tick_in_scanline < m_crtc_state.horizontal_total)
  {
    // short path when we execute <1 line.. this shouldn't occur often.
    const bool old_hblank = m_crtc_state.in_hblank;
    const bool new_hblank = (m_crtc_state.current_tick_in_scanline >= m_crtc_state.horizontal_sync_start);
    m_crtc_state.in_hblank = new_hblank;
    if (!old_hblank && new_hblank && Timers::IsUsingExternalClock(HBLANK_TIMER_INDEX))
      Timers::AddTicks(HBLANK_TIMER_INDEX, 1);

    UpdateCRTCTickEvent();
    return;
  }

  u32 lines_to_draw = m_crtc_state.current_tick_in_scanline / m_crtc_state.horizontal_total;
  m_crtc_state.current_tick_in_scanline %= m_crtc_state.horizontal_total;
#if 0
  Log_WarningPrintf("Old line: %u, new line: %u, drawing %u", m_crtc_state.current_scanline,
                    m_crtc_state.current_scanline + lines_to_draw, lines_to_draw);
#endif

  const bool old_hblank = m_crtc_state.in_hblank;
  const bool new_hblank = (m_crtc_state.current_tick_in_scanline >= m_crtc_state.horizontal_sync_start);
  m_crtc_state.in_hblank = new_hblank;
  if (Timers::IsUsingExternalClock(HBLANK_TIMER_INDEX))
  {
    const u32 hblank_timer_ticks = BoolToUInt32(!old_hblank) + BoolToUInt32(new_hblank) + (lines_to_draw - 1);
    Timers::AddTicks(HBLANK_TIMER_INDEX, static_cast<TickCount>(hblank_timer_ticks));
  }

  while (lines_to_draw > 0)
  {
    const u32 lines_to_draw_this_loop =
      std::min(lines_to_draw, m_crtc_state.vertical_total - m_crtc_state.current_scanline);
    const u32 prev_scanline = m_crtc_state.current_scanline;
    m_crtc_state.current_scanline += lines_to_draw_this_loop;
    DebugAssert(m_crtc_state.current_scanline <= m_crtc_state.vertical_total);
    lines_to_draw -= lines_to_draw_this_loop;

    // clear the vblank flag if the beam would pass through the display area
    if (prev_scanline < m_crtc_state.vertical_display_start &&
        m_crtc_state.current_scanline >= m_crtc_state.vertical_display_end)
    {
      Timers::SetGate(HBLANK_TIMER_INDEX, false);
      InterruptController::SetLineState(InterruptController::IRQ::VBLANK, false);
      m_crtc_state.in_vblank = false;
    }

    const bool new_vblank = m_crtc_state.current_scanline < m_crtc_state.vertical_display_start ||
                            m_crtc_state.current_scanline >= m_crtc_state.vertical_display_end;
    if (m_crtc_state.in_vblank != new_vblank)
    {
      if (new_vblank)
      {
        Log_DebugPrintf("Now in v-blank");

        // flush any pending draws and "scan out" the image
        // TODO: move present in here I guess
        FlushRender();
        UpdateDisplay();
        TimingEvents::SetFrameDone();

        // switch fields early. this is needed so we draw to the correct one.
        if (m_GPUSTAT.InInterleaved480iMode())
          m_crtc_state.interlaced_display_field = m_crtc_state.interlaced_field ^ 1u;
        else
          m_crtc_state.interlaced_display_field = 0;
      }

      Timers::SetGate(HBLANK_TIMER_INDEX, new_vblank);
      InterruptController::SetLineState(InterruptController::IRQ::VBLANK, new_vblank);
      m_crtc_state.in_vblank = new_vblank;
    }

    // past the end of vblank?
    if (m_crtc_state.current_scanline == m_crtc_state.vertical_total)
    {
      // start the new frame
      m_crtc_state.current_scanline = 0;
      if (m_GPUSTAT.vertical_interlace)
      {
        m_crtc_state.interlaced_field ^= 1u;
        m_GPUSTAT.interlaced_field = !m_crtc_state.interlaced_field;
      }
      else
      {
        m_crtc_state.interlaced_field = 0;
        m_GPUSTAT.interlaced_field = 0u; // new GPU = 1, old GPU = 0
      }
    }
  }

  // alternating even line bit in 240-line mode
  if (m_GPUSTAT.InInterleaved480iMode())
  {
    m_crtc_state.active_line_lsb =
      Truncate8((m_crtc_state.regs.Y + BoolToUInt32(m_crtc_state.interlaced_display_field)) & u32(1));
    m_GPUSTAT.display_line_lsb = ConvertToBoolUnchecked(
      (m_crtc_state.regs.Y + (BoolToUInt8(!m_crtc_state.in_vblank) & m_crtc_state.interlaced_display_field)) & u32(1));
  }
  else
  {
    m_crtc_state.active_line_lsb = 0;
    m_GPUSTAT.display_line_lsb = ConvertToBoolUnchecked((m_crtc_state.regs.Y + m_crtc_state.current_scanline) & u32(1));
  }

  UpdateCRTCTickEvent();
}

void GPU::CommandTickEvent(TickCount ticks)
{
  m_pending_command_ticks -= SystemTicksToGPUTicks(ticks);
  m_command_tick_event->Deactivate();

  // we can be syncing if this came from a DMA write. recursively executing commands would be bad.
  if (!m_syncing)
    ExecuteCommands();

  UpdateGPUIdle();

  if (m_pending_command_ticks <= 0)
    m_pending_command_ticks = 0;
  else
    m_command_tick_event->SetIntervalAndSchedule(GPUTicksToSystemTicks(m_pending_command_ticks));
}

void GPU::UpdateCommandTickEvent()
{
  if (m_pending_command_ticks <= 0)
    m_command_tick_event->Deactivate();
  else if (!m_command_tick_event->IsActive())
    m_command_tick_event->SetIntervalAndSchedule(GPUTicksToSystemTicks(m_pending_command_ticks));
}

void GPU::ConvertScreenCoordinatesToDisplayCoordinates(float window_x, float window_y, float* display_x,
                                                       float* display_y) const
{
  const Common::Rectangle<s32> draw_rc =
    CalculateDrawRect(g_gpu_device->GetWindowWidth(), g_gpu_device->GetWindowHeight());

  // convert coordinates to active display region, then to full display region
  const float scaled_display_x = (window_x - static_cast<float>(draw_rc.left)) / static_cast<float>(draw_rc.GetWidth());
  const float scaled_display_y = (window_y - static_cast<float>(draw_rc.top)) / static_cast<float>(draw_rc.GetHeight());

  // scale back to internal resolution
  *display_x = scaled_display_x * static_cast<float>(m_crtc_state.display_width);
  *display_y = scaled_display_y * static_cast<float>(m_crtc_state.display_height);

  Log_DevPrintf("win %.0f,%.0f -> local %.0f,%.0f, disp %.2f,%.2f (size %u,%u frac %f,%f)", window_x, window_y,
                window_x - draw_rc.left, window_y - draw_rc.top, *display_x, *display_y, m_crtc_state.display_width,
                m_crtc_state.display_height, *display_x / static_cast<float>(m_crtc_state.display_width),
                *display_y / static_cast<float>(m_crtc_state.display_height));
}

bool GPU::ConvertDisplayCoordinatesToBeamTicksAndLines(float display_x, float display_y, float x_scale, u32* out_tick,
                                                       u32* out_line) const
{
  if (x_scale != 1.0f)
  {
    const float dw = static_cast<float>(m_crtc_state.display_width);
    float scaled_x = ((display_x / dw) * 2.0f) - 1.0f; // 0..1 -> -1..1
    scaled_x *= x_scale;
    display_x = (((scaled_x + 1.0f) * 0.5f) * dw); // -1..1 -> 0..1
  }

  if (display_x < 0 || static_cast<u32>(display_x) >= m_crtc_state.display_width || display_y < 0 ||
      static_cast<u32>(display_y) >= m_crtc_state.display_height)
  {
    return false;
  }

  *out_line = (static_cast<u32>(std::round(display_y)) >> BoolToUInt8(m_GPUSTAT.vertical_interlace)) +
              m_crtc_state.vertical_visible_start;
  *out_tick = static_cast<u32>(std::round(display_x * static_cast<float>(m_crtc_state.dot_clock_divider))) +
              m_crtc_state.horizontal_visible_start;
  return true;
}

u32 GPU::ReadGPUREAD()
{
  if (m_blitter_state != BlitterState::ReadingVRAM)
    return m_GPUREAD_latch;

  // Read two pixels out of VRAM and combine them. Zero fill odd pixel counts.
  u32 value = 0;
  for (u32 i = 0; i < 2; i++)
  {
    // Read with correct wrap-around behavior.
    const u16 read_x = (m_vram_transfer.x + m_vram_transfer.col) % VRAM_WIDTH;
    const u16 read_y = (m_vram_transfer.y + m_vram_transfer.row) % VRAM_HEIGHT;
    value |= ZeroExtend32(g_vram[read_y * VRAM_WIDTH + read_x]) << (i * 16);

    if (++m_vram_transfer.col == m_vram_transfer.width)
    {
      m_vram_transfer.col = 0;

      if (++m_vram_transfer.row == m_vram_transfer.height)
      {
        Log_DebugPrintf("End of VRAM->CPU transfer");
        m_vram_transfer = {};
        m_blitter_state = BlitterState::Idle;

        // end of transfer, catch up on any commands which were written (unlikely)
        ExecuteCommands();
        UpdateCommandTickEvent();
        break;
      }
    }
  }

  m_GPUREAD_latch = value;
  return value;
}

void GPU::WriteGP1(u32 value)
{
  const u32 command = (value >> 24) & 0x3Fu;
  const u32 param = value & UINT32_C(0x00FFFFFF);
  switch (command)
  {
    case 0x00: // Reset GPU
    {
      Log_DebugPrintf("GP1 reset GPU");
      m_command_tick_event->InvokeEarly();
      SynchronizeCRTC();
      SoftReset();
    }
    break;

    case 0x01: // Clear FIFO
    {
      Log_DebugPrintf("GP1 clear FIFO");
      m_command_tick_event->InvokeEarly();
      SynchronizeCRTC();

      // flush partial writes
      if (m_blitter_state == BlitterState::WritingVRAM)
        FinishVRAMWrite();

      m_blitter_state = BlitterState::Idle;
      m_command_total_words = 0;
      m_vram_transfer = {};
      m_fifo.Clear();
      m_blit_buffer.clear();
      m_blit_remaining_words = 0;
      m_pending_command_ticks = 0;
      m_command_tick_event->Deactivate();
      UpdateDMARequest();
      UpdateGPUIdle();
    }
    break;

    case 0x02: // Acknowledge Interrupt
    {
      Log_DebugPrintf("Acknowledge interrupt");
      m_GPUSTAT.interrupt_request = false;
    }
    break;

    case 0x03: // Display on/off
    {
      const bool disable = ConvertToBoolUnchecked(value & 0x01);
      Log_DebugPrintf("Display %s", disable ? "disabled" : "enabled");
      SynchronizeCRTC();

      if (!m_GPUSTAT.display_disable && disable && m_GPUSTAT.vertical_interlace && !m_force_progressive_scan)
        ClearDisplay();

      m_GPUSTAT.display_disable = disable;
    }
    break;

    case 0x04: // DMA Direction
    {
      Log_DebugPrintf("DMA direction <- 0x%02X", static_cast<u32>(param));
      if (m_GPUSTAT.dma_direction != static_cast<DMADirection>(param))
      {
        m_GPUSTAT.dma_direction = static_cast<DMADirection>(param);
        UpdateDMARequest();
      }
    }
    break;

    case 0x05: // Set display start address
    {
      const u32 new_value = param & CRTCState::Regs::DISPLAY_ADDRESS_START_MASK;
      Log_DebugPrintf("Display address start <- 0x%08X", new_value);

      System::IncrementInternalFrameNumber();
      if (m_crtc_state.regs.display_address_start != new_value)
      {
        SynchronizeCRTC();
        m_crtc_state.regs.display_address_start = new_value;
        UpdateCRTCDisplayParameters();
      }
    }
    break;

    case 0x06: // Set horizontal display range
    {
      const u32 new_value = param & CRTCState::Regs::HORIZONTAL_DISPLAY_RANGE_MASK;
      Log_DebugPrintf("Horizontal display range <- 0x%08X", new_value);

      if (m_crtc_state.regs.horizontal_display_range != new_value)
      {
        SynchronizeCRTC();
        m_crtc_state.regs.horizontal_display_range = new_value;
        UpdateCRTCConfig();
      }
    }
    break;

    case 0x07: // Set vertical display range
    {
      const u32 new_value = param & CRTCState::Regs::VERTICAL_DISPLAY_RANGE_MASK;
      Log_DebugPrintf("Vertical display range <- 0x%08X", new_value);

      if (m_crtc_state.regs.vertical_display_range != new_value)
      {
        SynchronizeCRTC();
        m_crtc_state.regs.vertical_display_range = new_value;
        UpdateCRTCConfig();
      }
    }
    break;

    case 0x08: // Set display mode
    {
      union GP1_08h
      {
        u32 bits;

        BitField<u32, u8, 0, 2> horizontal_resolution_1;
        BitField<u32, bool, 2, 1> vertical_resolution;
        BitField<u32, bool, 3, 1> pal_mode;
        BitField<u32, bool, 4, 1> display_area_color_depth;
        BitField<u32, bool, 5, 1> vertical_interlace;
        BitField<u32, bool, 6, 1> horizontal_resolution_2;
        BitField<u32, bool, 7, 1> reverse_flag;
      };

      const GP1_08h dm{param};
      GPUSTAT new_GPUSTAT{m_GPUSTAT.bits};
      new_GPUSTAT.horizontal_resolution_1 = dm.horizontal_resolution_1;
      new_GPUSTAT.vertical_resolution = dm.vertical_resolution;
      new_GPUSTAT.pal_mode = dm.pal_mode;
      new_GPUSTAT.display_area_color_depth_24 = dm.display_area_color_depth;
      new_GPUSTAT.vertical_interlace = dm.vertical_interlace;
      new_GPUSTAT.horizontal_resolution_2 = dm.horizontal_resolution_2;
      new_GPUSTAT.reverse_flag = dm.reverse_flag;
      Log_DebugPrintf("Set display mode <- 0x%08X", dm.bits);

      if (!m_GPUSTAT.vertical_interlace && dm.vertical_interlace && !m_force_progressive_scan)
      {
        // bit of a hack, technically we should pull the previous frame in, but this may not exist anymore
        ClearDisplay();
      }

      if (m_GPUSTAT.bits != new_GPUSTAT.bits)
      {
        // Have to be careful when setting this because Synchronize() can modify GPUSTAT.
        static constexpr u32 SET_MASK = UINT32_C(0b00000000011111110100000000000000);
        m_command_tick_event->InvokeEarly();
        SynchronizeCRTC();
        m_GPUSTAT.bits = (m_GPUSTAT.bits & ~SET_MASK) | (new_GPUSTAT.bits & SET_MASK);
        UpdateCRTCConfig();
      }
    }
    break;

    case 0x09: // Allow texture disable
    {
      m_set_texture_disable_mask = ConvertToBoolUnchecked(param & 0x01);
      Log_DebugPrintf("Set texture disable mask <- %s", m_set_texture_disable_mask ? "allowed" : "ignored");
    }
    break;

    case 0x10:
    case 0x11:
    case 0x12:
    case 0x13:
    case 0x14:
    case 0x15:
    case 0x16:
    case 0x17:
    case 0x18:
    case 0x19:
    case 0x1A:
    case 0x1B:
    case 0x1C:
    case 0x1D:
    case 0x1E:
    case 0x1F:
    {
      HandleGetGPUInfoCommand(value);
    }
    break;

    default:
      Log_ErrorPrintf("Unimplemented GP1 command 0x%02X", command);
      break;
  }
}

void GPU::HandleGetGPUInfoCommand(u32 value)
{
  const u8 subcommand = Truncate8(value & 0x07);
  switch (subcommand)
  {
    case 0x00:
    case 0x01:
    case 0x06:
    case 0x07:
      // leave GPUREAD intact
      break;

    case 0x02: // Get Texture Window
    {
      Log_DebugPrintf("Get texture window");
      m_GPUREAD_latch = m_draw_mode.texture_window_value;
    }
    break;

    case 0x03: // Get Draw Area Top Left
    {
      Log_DebugPrintf("Get drawing area top left");
      m_GPUREAD_latch =
        ((m_drawing_area.left & UINT32_C(0b1111111111)) | ((m_drawing_area.top & UINT32_C(0b1111111111)) << 10));
    }
    break;

    case 0x04: // Get Draw Area Bottom Right
    {
      Log_DebugPrintf("Get drawing area bottom right");
      m_GPUREAD_latch =
        ((m_drawing_area.right & UINT32_C(0b1111111111)) | ((m_drawing_area.bottom & UINT32_C(0b1111111111)) << 10));
    }
    break;

    case 0x05: // Get Drawing Offset
    {
      Log_DebugPrintf("Get drawing offset");
      m_GPUREAD_latch =
        ((m_drawing_offset.x & INT32_C(0b11111111111)) | ((m_drawing_offset.y & INT32_C(0b11111111111)) << 11));
    }
    break;

    default:
      Log_WarningPrintf("Unhandled GetGPUInfo(0x%02X)", ZeroExtend32(subcommand));
      break;
  }
}

void GPU::ClearDisplay()
{
  ClearDisplayTexture();

  // Just recycle the textures, it'll get re-fetched.
  DestroyDeinterlaceTextures();
}

void GPU::UpdateDisplay()
{
}

void GPU::ReadVRAM(u32 x, u32 y, u32 width, u32 height)
{
}

void GPU::FillVRAM(u32 x, u32 y, u32 width, u32 height, u32 color)
{
  const u16 color16 = VRAMRGBA8888ToRGBA5551(color);
  if ((x + width) <= VRAM_WIDTH && !IsInterlacedRenderingEnabled())
  {
    for (u32 yoffs = 0; yoffs < height; yoffs++)
    {
      const u32 row = (y + yoffs) % VRAM_HEIGHT;
      std::fill_n(&g_vram[row * VRAM_WIDTH + x], width, color16);
    }
  }
  else if (IsInterlacedRenderingEnabled())
  {
    // Hardware tests show that fills seem to break on the first two lines when the offset matches the displayed field.
    if (IsCRTCScanlinePending())
      SynchronizeCRTC();

    const u32 active_field = GetActiveLineLSB();
    for (u32 yoffs = 0; yoffs < height; yoffs++)
    {
      const u32 row = (y + yoffs) % VRAM_HEIGHT;
      if ((row & u32(1)) == active_field)
        continue;

      u16* row_ptr = &g_vram[row * VRAM_WIDTH];
      for (u32 xoffs = 0; xoffs < width; xoffs++)
      {
        const u32 col = (x + xoffs) % VRAM_WIDTH;
        row_ptr[col] = color16;
      }
    }
  }
  else
  {
    for (u32 yoffs = 0; yoffs < height; yoffs++)
    {
      const u32 row = (y + yoffs) % VRAM_HEIGHT;
      u16* row_ptr = &g_vram[row * VRAM_WIDTH];
      for (u32 xoffs = 0; xoffs < width; xoffs++)
      {
        const u32 col = (x + xoffs) % VRAM_WIDTH;
        row_ptr[col] = color16;
      }
    }
  }
}

void GPU::UpdateVRAM(u32 x, u32 y, u32 width, u32 height, const void* data, bool set_mask, bool check_mask)
{
  // Fast path when the copy is not oversized.
  if ((x + width) <= VRAM_WIDTH && (y + height) <= VRAM_HEIGHT && !set_mask && !check_mask)
  {
    const u16* src_ptr = static_cast<const u16*>(data);
    u16* dst_ptr = &g_vram[y * VRAM_WIDTH + x];
    for (u32 yoffs = 0; yoffs < height; yoffs++)
    {
      std::copy_n(src_ptr, width, dst_ptr);
      src_ptr += width;
      dst_ptr += VRAM_WIDTH;
    }
  }
  else
  {
    // Slow path when we need to handle wrap-around.
    // During transfer/render operations, if ((dst_pixel & mask_and) == 0) { pixel = src_pixel | mask_or }
    const u16* src_ptr = static_cast<const u16*>(data);
    const u16 mask_and = check_mask ? 0x8000 : 0;
    const u16 mask_or = set_mask ? 0x8000 : 0;

    for (u32 row = 0; row < height;)
    {
      u16* dst_row_ptr = &g_vram[((y + row++) % VRAM_HEIGHT) * VRAM_WIDTH];
      for (u32 col = 0; col < width;)
      {
        // TODO: Handle unaligned reads...
        u16* pixel_ptr = &dst_row_ptr[(x + col++) % VRAM_WIDTH];
        if (((*pixel_ptr) & mask_and) == 0)
          *pixel_ptr = *(src_ptr++) | mask_or;
      }
    }
  }
}

void GPU::CopyVRAM(u32 src_x, u32 src_y, u32 dst_x, u32 dst_y, u32 width, u32 height)
{
  // Break up oversized copies. This behavior has not been verified on console.
  if ((src_x + width) > VRAM_WIDTH || (dst_x + width) > VRAM_WIDTH)
  {
    u32 remaining_rows = height;
    u32 current_src_y = src_y;
    u32 current_dst_y = dst_y;
    while (remaining_rows > 0)
    {
      const u32 rows_to_copy =
        std::min<u32>(remaining_rows, std::min<u32>(VRAM_HEIGHT - current_src_y, VRAM_HEIGHT - current_dst_y));

      u32 remaining_columns = width;
      u32 current_src_x = src_x;
      u32 current_dst_x = dst_x;
      while (remaining_columns > 0)
      {
        const u32 columns_to_copy =
          std::min<u32>(remaining_columns, std::min<u32>(VRAM_WIDTH - current_src_x, VRAM_WIDTH - current_dst_x));
        CopyVRAM(current_src_x, current_src_y, current_dst_x, current_dst_y, columns_to_copy, rows_to_copy);
        current_src_x = (current_src_x + columns_to_copy) % VRAM_WIDTH;
        current_dst_x = (current_dst_x + columns_to_copy) % VRAM_WIDTH;
        remaining_columns -= columns_to_copy;
      }

      current_src_y = (current_src_y + rows_to_copy) % VRAM_HEIGHT;
      current_dst_y = (current_dst_y + rows_to_copy) % VRAM_HEIGHT;
      remaining_rows -= rows_to_copy;
    }

    return;
  }

  // This doesn't have a fast path, but do we really need one? It's not common.
  const u16 mask_and = m_GPUSTAT.GetMaskAND();
  const u16 mask_or = m_GPUSTAT.GetMaskOR();

  // Copy in reverse when src_x < dst_x, this is verified on console.
  if (src_x < dst_x || ((src_x + width - 1) % VRAM_WIDTH) < ((dst_x + width - 1) % VRAM_WIDTH))
  {
    for (u32 row = 0; row < height; row++)
    {
      const u16* src_row_ptr = &g_vram[((src_y + row) % VRAM_HEIGHT) * VRAM_WIDTH];
      u16* dst_row_ptr = &g_vram[((dst_y + row) % VRAM_HEIGHT) * VRAM_WIDTH];

      for (s32 col = static_cast<s32>(width - 1); col >= 0; col--)
      {
        const u16 src_pixel = src_row_ptr[(src_x + static_cast<u32>(col)) % VRAM_WIDTH];
        u16* dst_pixel_ptr = &dst_row_ptr[(dst_x + static_cast<u32>(col)) % VRAM_WIDTH];
        if ((*dst_pixel_ptr & mask_and) == 0)
          *dst_pixel_ptr = src_pixel | mask_or;
      }
    }
  }
  else
  {
    for (u32 row = 0; row < height; row++)
    {
      const u16* src_row_ptr = &g_vram[((src_y + row) % VRAM_HEIGHT) * VRAM_WIDTH];
      u16* dst_row_ptr = &g_vram[((dst_y + row) % VRAM_HEIGHT) * VRAM_WIDTH];

      for (u32 col = 0; col < width; col++)
      {
        const u16 src_pixel = src_row_ptr[(src_x + col) % VRAM_WIDTH];
        u16* dst_pixel_ptr = &dst_row_ptr[(dst_x + col) % VRAM_WIDTH];
        if ((*dst_pixel_ptr & mask_and) == 0)
          *dst_pixel_ptr = src_pixel | mask_or;
      }
    }
  }
}

void GPU::DispatchRenderCommand()
{
}

void GPU::FlushRender()
{
}

void GPU::SetDrawMode(u16 value)
{
  GPUDrawModeReg new_mode_reg{static_cast<u16>(value & GPUDrawModeReg::MASK)};
  if (!m_set_texture_disable_mask)
    new_mode_reg.texture_disable = false;

  if (new_mode_reg.bits == m_draw_mode.mode_reg.bits)
    return;

  m_draw_mode.texture_page_changed |= ((new_mode_reg.bits & GPUDrawModeReg::TEXTURE_PAGE_MASK) !=
                                       (m_draw_mode.mode_reg.bits & GPUDrawModeReg::TEXTURE_PAGE_MASK));
  m_draw_mode.mode_reg.bits = new_mode_reg.bits;

  if (m_GPUSTAT.draw_to_displayed_field != new_mode_reg.draw_to_displayed_field)
    FlushRender();

  // Bits 0..10 are returned in the GPU status register.
  m_GPUSTAT.bits = (m_GPUSTAT.bits & ~(GPUDrawModeReg::GPUSTAT_MASK)) |
                   (ZeroExtend32(new_mode_reg.bits) & GPUDrawModeReg::GPUSTAT_MASK);
  m_GPUSTAT.texture_disable = m_draw_mode.mode_reg.texture_disable;
}

void GPU::SetTexturePalette(u16 value)
{
  value &= DrawMode::PALETTE_MASK;
  if (m_draw_mode.palette_reg.bits == value)
    return;

  m_draw_mode.palette_reg.bits = value;
  m_draw_mode.texture_page_changed = true;
}

void GPU::SetTextureWindow(u32 value)
{
  value &= DrawMode::TEXTURE_WINDOW_MASK;
  if (m_draw_mode.texture_window_value == value)
    return;

  FlushRender();

  const u8 mask_x = Truncate8(value & UINT32_C(0x1F));
  const u8 mask_y = Truncate8((value >> 5) & UINT32_C(0x1F));
  const u8 offset_x = Truncate8((value >> 10) & UINT32_C(0x1F));
  const u8 offset_y = Truncate8((value >> 15) & UINT32_C(0x1F));
  Log_DebugPrintf("Set texture window %02X %02X %02X %02X", mask_x, mask_y, offset_x, offset_y);

  m_draw_mode.texture_window.and_x = ~(mask_x * 8);
  m_draw_mode.texture_window.and_y = ~(mask_y * 8);
  m_draw_mode.texture_window.or_x = (offset_x & mask_x) * 8u;
  m_draw_mode.texture_window.or_y = (offset_y & mask_y) * 8u;
  m_draw_mode.texture_window_value = value;
  m_draw_mode.texture_window_changed = true;
}

bool GPU::CompileDisplayPipelines(bool display, bool deinterlace, bool chroma_smoothing)
{
  GPUShaderGen shadergen(g_gpu_device->GetRenderAPI(), g_gpu_device->GetFeatures().dual_source_blend,
                         g_gpu_device->GetFeatures().framebuffer_fetch);

  GPUPipeline::GraphicsConfig plconfig;
  plconfig.input_layout.vertex_stride = 0;
  plconfig.primitive = GPUPipeline::Primitive::Triangles;
  plconfig.rasterization = GPUPipeline::RasterizationState::GetNoCullState();
  plconfig.depth = GPUPipeline::DepthState::GetNoTestsState();
  plconfig.blend = GPUPipeline::BlendState::GetNoBlendingState();
  plconfig.depth_format = GPUTexture::Format::Unknown;
  plconfig.samples = 1;
  plconfig.per_sample_shading = false;
  plconfig.geometry_shader = nullptr;

  if (display)
  {
    plconfig.layout = GPUPipeline::Layout::SingleTextureAndPushConstants;
    plconfig.SetTargetFormats(g_gpu_device->HasSurface() ? g_gpu_device->GetWindowFormat() : GPUTexture::Format::RGBA8);

    std::string vs = shadergen.GenerateDisplayVertexShader();
    std::string fs;
    switch (g_settings.display_scaling)
    {
      case DisplayScalingMode::BilinearSharp:
        fs = shadergen.GenerateDisplaySharpBilinearFragmentShader();
        break;

      case DisplayScalingMode::BilinearSmooth:
        fs = shadergen.GenerateDisplayFragmentShader(true);
        break;

      case DisplayScalingMode::Nearest:
      case DisplayScalingMode::NearestInteger:
      default:
        fs = shadergen.GenerateDisplayFragmentShader(false);
        break;
    }

    std::unique_ptr<GPUShader> vso = g_gpu_device->CreateShader(GPUShaderStage::Vertex, vs);
    std::unique_ptr<GPUShader> fso = g_gpu_device->CreateShader(GPUShaderStage::Fragment, fs);
    if (!vso || !fso)
      return false;
    GL_OBJECT_NAME(vso, "Display Vertex Shader");
    GL_OBJECT_NAME_FMT(fso, "Display Fragment Shader [{}]",
                       Settings::GetDisplayScalingName(g_settings.display_scaling));

    plconfig.vertex_shader = vso.get();
    plconfig.fragment_shader = fso.get();
    if (!(m_display_pipeline = g_gpu_device->CreatePipeline(plconfig)))
      return false;
    GL_OBJECT_NAME_FMT(m_display_pipeline, "Display Pipeline [{}]",
                       Settings::GetDisplayScalingName(g_settings.display_scaling));
  }

  if (deinterlace)
  {
    plconfig.SetTargetFormats(GPUTexture::Format::RGBA8);

    std::unique_ptr<GPUShader> vso =
      g_gpu_device->CreateShader(GPUShaderStage::Vertex, shadergen.GenerateScreenQuadVertexShader());
    if (!vso)
      return false;
    GL_OBJECT_NAME(vso, "Deinterlace Vertex Shader");

    std::unique_ptr<GPUShader> fso;
    if (!(fso = g_gpu_device->CreateShader(GPUShaderStage::Fragment,
                                           shadergen.GenerateInterleavedFieldExtractFragmentShader())))
    {
      return false;
    }

    GL_OBJECT_NAME(fso, "Deinterlace Field Extract Fragment Shader");

    plconfig.layout = GPUPipeline::Layout::SingleTextureAndPushConstants;
    plconfig.vertex_shader = vso.get();
    plconfig.fragment_shader = fso.get();
    if (!(m_deinterlace_extract_pipeline = g_gpu_device->CreatePipeline(plconfig)))
      return false;

    GL_OBJECT_NAME(m_deinterlace_extract_pipeline, "Deinterlace Field Extract Pipeline");

    switch (g_settings.display_deinterlacing_mode)
    {
      case DisplayDeinterlacingMode::Disabled:
        break;

      case DisplayDeinterlacingMode::Weave:
      {
        if (!(fso = g_gpu_device->CreateShader(GPUShaderStage::Fragment,
                                               shadergen.GenerateDeinterlaceWeaveFragmentShader())))
        {
          return false;
        }

        GL_OBJECT_NAME(fso, "Weave Deinterlace Fragment Shader");

        plconfig.layout = GPUPipeline::Layout::SingleTextureAndPushConstants;
        plconfig.vertex_shader = vso.get();
        plconfig.fragment_shader = fso.get();
        if (!(m_deinterlace_pipeline = g_gpu_device->CreatePipeline(plconfig)))
          return false;

        GL_OBJECT_NAME(m_deinterlace_pipeline, "Weave Deinterlace Pipeline");
      }
      break;

      case DisplayDeinterlacingMode::Blend:
      {
        if (!(fso = g_gpu_device->CreateShader(GPUShaderStage::Fragment,
                                               shadergen.GenerateDeinterlaceBlendFragmentShader())))
        {
          return false;
        }

        GL_OBJECT_NAME(fso, "Blend Deinterlace Fragment Shader");

        plconfig.layout = GPUPipeline::Layout::MultiTextureAndPushConstants;
        plconfig.vertex_shader = vso.get();
        plconfig.fragment_shader = fso.get();
        if (!(m_deinterlace_pipeline = g_gpu_device->CreatePipeline(plconfig)))
          return false;

        GL_OBJECT_NAME(m_deinterlace_pipeline, "Blend Deinterlace Pipeline");
      }
      break;

      case DisplayDeinterlacingMode::Adaptive:
      {
        fso =
          g_gpu_device->CreateShader(GPUShaderStage::Fragment, shadergen.GenerateFastMADReconstructFragmentShader());
        if (!fso)
          return false;

        GL_OBJECT_NAME(fso, "FastMAD Reconstruct Fragment Shader");

        plconfig.layout = GPUPipeline::Layout::MultiTextureAndPushConstants;
        plconfig.fragment_shader = fso.get();
        if (!(m_deinterlace_pipeline = g_gpu_device->CreatePipeline(plconfig)))
          return false;

        GL_OBJECT_NAME(m_deinterlace_pipeline, "FastMAD Reconstruct Pipeline");
      }
      break;

      default:
        UnreachableCode();
    }
  }

  if (chroma_smoothing)
  {
    m_chroma_smoothing_pipeline.reset();
    g_gpu_device->RecycleTexture(std::move(m_chroma_smoothing_texture));

    if (g_settings.gpu_24bit_chroma_smoothing)
    {
      plconfig.layout = GPUPipeline::Layout::SingleTextureAndPushConstants;
      plconfig.SetTargetFormats(GPUTexture::Format::RGBA8);

      std::unique_ptr<GPUShader> vso =
        g_gpu_device->CreateShader(GPUShaderStage::Vertex, shadergen.GenerateScreenQuadVertexShader());
      std::unique_ptr<GPUShader> fso =
        g_gpu_device->CreateShader(GPUShaderStage::Fragment, shadergen.GenerateChromaSmoothingFragmentShader());
      if (!vso || !fso)
        return false;
      GL_OBJECT_NAME(vso, "Chroma Smoothing Vertex Shader");
      GL_OBJECT_NAME(fso, "Chroma Smoothing Fragment Shader");

      plconfig.vertex_shader = vso.get();
      plconfig.fragment_shader = fso.get();
      if (!(m_chroma_smoothing_pipeline = g_gpu_device->CreatePipeline(plconfig)))
        return false;
      GL_OBJECT_NAME(m_chroma_smoothing_pipeline, "Chroma Smoothing Pipeline");
    }
  }

  return true;
}

void GPU::ClearDisplayTexture()
{
  m_display_texture = nullptr;
  m_display_texture_view_x = 0;
  m_display_texture_view_y = 0;
  m_display_texture_view_width = 0;
  m_display_texture_view_height = 0;
}

void GPU::SetDisplayTexture(GPUTexture* texture, s32 view_x, s32 view_y, s32 view_width, s32 view_height)
{
  DebugAssert(texture);
  m_display_texture = texture;
  m_display_texture_view_x = view_x;
  m_display_texture_view_y = view_y;
  m_display_texture_view_width = view_width;
  m_display_texture_view_height = view_height;
}

void GPU::SetDisplayTextureRect(s32 view_x, s32 view_y, s32 view_width, s32 view_height)
{
  m_display_texture_view_x = view_x;
  m_display_texture_view_y = view_y;
  m_display_texture_view_width = view_width;
  m_display_texture_view_height = view_height;
}

void GPU::SetDisplayParameters(s32 display_width, s32 display_height, s32 active_left, s32 active_top, s32 active_width,
                               s32 active_height, float display_aspect_ratio)
{
  m_display_width = display_width;
  m_display_height = display_height;
  m_display_active_left = active_left;
  m_display_active_top = active_top;
  m_display_active_width = active_width;
  m_display_active_height = active_height;
  m_display_aspect_ratio = display_aspect_ratio;
}

bool GPU::PresentDisplay()
{
  FlushRender();

  if (!HasDisplayTexture())
    return g_gpu_device->BeginPresent(false);

  const Common::Rectangle<s32> draw_rect =
    CalculateDrawRect(g_gpu_device->GetWindowWidth(), g_gpu_device->GetWindowHeight());
  return RenderDisplay(nullptr, draw_rect, true);
}

bool GPU::RenderDisplay(GPUTexture* target, const Common::Rectangle<s32>& draw_rect, bool postfx)
{
  GL_SCOPE_FMT("RenderDisplay: {}x{} at {},{}", draw_rect.left, draw_rect.top, draw_rect.GetWidth(),
               draw_rect.GetHeight());

  if (m_display_texture)
    m_display_texture->MakeReadyForSampling();

  bool texture_filter_linear = false;

  struct Uniforms
  {
    float src_rect[4];
    float src_size[4];
    float clamp_rect[4];
    float params[4];
  } uniforms;
  std::memset(uniforms.params, 0, sizeof(uniforms.params));

  switch (g_settings.display_scaling)
  {
    case DisplayScalingMode::Nearest:
    case DisplayScalingMode::NearestInteger:
      break;

    case DisplayScalingMode::BilinearSmooth:
      texture_filter_linear = true;
      break;

    case DisplayScalingMode::BilinearSharp:
    {
      texture_filter_linear = true;
      uniforms.params[0] = std::max(
        std::floor(static_cast<float>(draw_rect.GetWidth()) / static_cast<float>(m_display_texture_view_width)), 1.0f);
      uniforms.params[1] = std::max(
        std::floor(static_cast<float>(draw_rect.GetHeight()) / static_cast<float>(m_display_texture_view_height)),
        1.0f);
      uniforms.params[2] = 0.5f - 0.5f / uniforms.params[0];
      uniforms.params[3] = 0.5f - 0.5f / uniforms.params[1];
    }
    break;

    default:
      UnreachableCode();
      break;
  }

  const GPUTexture::Format hdformat = target ? target->GetFormat() : g_gpu_device->GetWindowFormat();
  const u32 target_width = target ? target->GetWidth() : g_gpu_device->GetWindowWidth();
  const u32 target_height = target ? target->GetHeight() : g_gpu_device->GetWindowHeight();
  const bool really_postfx =
    (postfx && HasDisplayTexture() && PostProcessing::IsActive() && !g_gpu_device->GetWindowInfo().IsSurfaceless() &&
     hdformat != GPUTexture::Format::Unknown && target_width > 0 && target_height > 0 &&
     PostProcessing::CheckTargets(hdformat, target_width, target_height));
  const Common::Rectangle<s32> real_draw_rect =
    g_gpu_device->UsesLowerLeftOrigin() ? GPUDevice::FlipToLowerLeft(draw_rect, target_height) : draw_rect;
  if (really_postfx)
  {
    g_gpu_device->ClearRenderTarget(PostProcessing::GetInputTexture(), 0);
    g_gpu_device->SetRenderTarget(PostProcessing::GetInputTexture());
  }
  else
  {
    if (target)
      g_gpu_device->SetRenderTarget(target);
    else if (!g_gpu_device->BeginPresent(false))
      return false;
  }

  if (!HasDisplayTexture())
    return true;

  g_gpu_device->SetPipeline(m_display_pipeline.get());
  g_gpu_device->SetTextureSampler(
    0, m_display_texture, texture_filter_linear ? g_gpu_device->GetLinearSampler() : g_gpu_device->GetNearestSampler());

  // For bilinear, clamp to 0.5/SIZE-0.5 to avoid bleeding from the adjacent texels in VRAM. This is because
  // 1.0 in UV space is not the bottom-right texel, but a mix of the bottom-right and wrapped/next texel.
  const float rcp_width = 1.0f / static_cast<float>(m_display_texture->GetWidth());
  const float rcp_height = 1.0f / static_cast<float>(m_display_texture->GetHeight());
  uniforms.src_rect[0] = static_cast<float>(m_display_texture_view_x) * rcp_width;
  uniforms.src_rect[1] = static_cast<float>(m_display_texture_view_y) * rcp_height;
  uniforms.src_rect[2] = static_cast<float>(m_display_texture_view_width) * rcp_width;
  uniforms.src_rect[3] = static_cast<float>(m_display_texture_view_height) * rcp_height;
  uniforms.clamp_rect[0] = (static_cast<float>(m_display_texture_view_x) + 0.5f) * rcp_width;
  uniforms.clamp_rect[1] = (static_cast<float>(m_display_texture_view_y) + 0.5f) * rcp_height;
  uniforms.clamp_rect[2] =
    (static_cast<float>(m_display_texture_view_x + m_display_texture_view_width) - 0.5f) * rcp_width;
  uniforms.clamp_rect[3] =
    (static_cast<float>(m_display_texture_view_y + m_display_texture_view_height) - 0.5f) * rcp_height;
  uniforms.src_size[0] = static_cast<float>(m_display_texture->GetWidth());
  uniforms.src_size[1] = static_cast<float>(m_display_texture->GetHeight());
  uniforms.src_size[2] = rcp_width;
  uniforms.src_size[3] = rcp_height;
  g_gpu_device->PushUniformBuffer(&uniforms, sizeof(uniforms));

  g_gpu_device->SetViewportAndScissor(real_draw_rect.left, real_draw_rect.top, real_draw_rect.GetWidth(),
                                      real_draw_rect.GetHeight());
  g_gpu_device->Draw(3, 0);

  if (really_postfx)
  {
    return PostProcessing::Apply(target, real_draw_rect.left, real_draw_rect.top, real_draw_rect.GetWidth(),
                                 real_draw_rect.GetHeight(), m_display_texture_view_width,
                                 m_display_texture_view_height);
  }
  else
  {
    return true;
  }
}

void GPU::DestroyDeinterlaceTextures()
{
  for (std::unique_ptr<GPUTexture>& tex : m_deinterlace_buffers)
    g_gpu_device->RecycleTexture(std::move(tex));
  g_gpu_device->RecycleTexture(std::move(m_deinterlace_texture));
  m_current_deinterlace_buffer = 0;
}

bool GPU::Deinterlace(GPUTexture* src, u32 x, u32 y, u32 width, u32 height, u32 field, u32 line_skip)
{
  switch (g_settings.display_deinterlacing_mode)
  {
    case DisplayDeinterlacingMode::Disabled:
    {
      if (line_skip == 0)
      {
        SetDisplayTexture(src, x, y, width, height);
        return true;
      }

      // Still have to extract the field.
      if (!DeinterlaceExtractField(0, src, x, y, width, height, line_skip)) [[unlikely]]
        return false;

      SetDisplayTexture(m_deinterlace_buffers[0].get(), 0, 0, width, height);
      return true;
    }

    case DisplayDeinterlacingMode::Weave:
    {
      GL_SCOPE_FMT("DeinterlaceWeave({{{},{}}}, {}x{}, field={}, line_skip={})", x, y, width, height, field, line_skip);

      const u32 full_height = height * 2;
      if (!DeinterlaceSetTargetSize(width, full_height, true)) [[unlikely]]
      {
        ClearDisplayTexture();
        return false;
      }

      src->MakeReadyForSampling();

      g_gpu_device->SetRenderTarget(m_deinterlace_texture.get());
      g_gpu_device->SetPipeline(m_deinterlace_pipeline.get());
      g_gpu_device->SetTextureSampler(0, src, g_gpu_device->GetNearestSampler());
      const u32 uniforms[] = {x, y, field, line_skip};
      g_gpu_device->PushUniformBuffer(uniforms, sizeof(uniforms));
      g_gpu_device->SetViewportAndScissor(0, 0, width, full_height);
      g_gpu_device->Draw(3, 0);

      m_deinterlace_texture->MakeReadyForSampling();
      SetDisplayTexture(m_deinterlace_texture.get(), 0, 0, width, full_height);
      return true;
    }

    case DisplayDeinterlacingMode::Blend:
    {
      constexpr u32 NUM_BLEND_BUFFERS = 2;

      GL_SCOPE_FMT("DeinterlaceBlend({{{},{}}}, {}x{}, field={}, line_skip={})", x, y, width, height, field, line_skip);

      const u32 this_buffer = m_current_deinterlace_buffer;
      m_current_deinterlace_buffer = (m_current_deinterlace_buffer + 1u) % NUM_BLEND_BUFFERS;
      GL_INS_FMT("Current buffer: {}", this_buffer);
      if (!DeinterlaceExtractField(this_buffer, src, x, y, width, height, line_skip) ||
          !DeinterlaceSetTargetSize(width, height, false)) [[unlikely]]
      {
        ClearDisplayTexture();
        return false;
      }

      // TODO: could be implemented with alpha blending instead..

      g_gpu_device->InvalidateRenderTarget(m_deinterlace_texture.get());
      g_gpu_device->SetRenderTarget(m_deinterlace_texture.get());
      g_gpu_device->SetPipeline(m_deinterlace_pipeline.get());
      g_gpu_device->SetTextureSampler(0, m_deinterlace_buffers[this_buffer].get(), g_gpu_device->GetNearestSampler());
      g_gpu_device->SetTextureSampler(1, m_deinterlace_buffers[(this_buffer - 1) % NUM_BLEND_BUFFERS].get(),
                                      g_gpu_device->GetNearestSampler());
      g_gpu_device->SetViewportAndScissor(0, 0, width, height);
      g_gpu_device->Draw(3, 0);

      m_deinterlace_texture->MakeReadyForSampling();
      SetDisplayTexture(m_deinterlace_texture.get(), 0, 0, width, height);
      return true;
    }

    case DisplayDeinterlacingMode::Adaptive:
    {
      GL_SCOPE_FMT("DeinterlaceAdaptive({{{},{}}}, {}x{}, field={}, line_skip={})", x, y, width, height, field,
                   line_skip);

      const u32 full_height = height * 2;
      const u32 this_buffer = m_current_deinterlace_buffer;
      m_current_deinterlace_buffer = (m_current_deinterlace_buffer + 1u) % DEINTERLACE_BUFFER_COUNT;
      GL_INS_FMT("Current buffer: {}", this_buffer);
      if (!DeinterlaceExtractField(this_buffer, src, x, y, width, height, line_skip) ||
          !DeinterlaceSetTargetSize(width, full_height, false)) [[unlikely]]
      {
        ClearDisplayTexture();
        return false;
      }

      g_gpu_device->SetRenderTarget(m_deinterlace_texture.get());
      g_gpu_device->SetPipeline(m_deinterlace_pipeline.get());
      g_gpu_device->SetTextureSampler(0, m_deinterlace_buffers[this_buffer].get(), g_gpu_device->GetNearestSampler());
      g_gpu_device->SetTextureSampler(1, m_deinterlace_buffers[(this_buffer - 1) % DEINTERLACE_BUFFER_COUNT].get(),
                                      g_gpu_device->GetNearestSampler());
      g_gpu_device->SetTextureSampler(2, m_deinterlace_buffers[(this_buffer - 2) % DEINTERLACE_BUFFER_COUNT].get(),
                                      g_gpu_device->GetNearestSampler());
      g_gpu_device->SetTextureSampler(3, m_deinterlace_buffers[(this_buffer - 3) % DEINTERLACE_BUFFER_COUNT].get(),
                                      g_gpu_device->GetNearestSampler());
      const u32 uniforms[] = {field, full_height};
      g_gpu_device->PushUniformBuffer(uniforms, sizeof(uniforms));
      g_gpu_device->SetViewportAndScissor(0, 0, width, full_height);
      g_gpu_device->Draw(3, 0);

      m_deinterlace_texture->MakeReadyForSampling();
      SetDisplayTexture(m_deinterlace_texture.get(), 0, 0, width, full_height);
      return true;
    }

    default:
      UnreachableCode();
  }
}

bool GPU::DeinterlaceExtractField(u32 dst_bufidx, GPUTexture* src, u32 x, u32 y, u32 width, u32 height, u32 line_skip)
{
  if (!m_deinterlace_buffers[dst_bufidx] || m_deinterlace_buffers[dst_bufidx]->GetWidth() != width ||
      m_deinterlace_buffers[dst_bufidx]->GetHeight() != height)
  {
    if (!g_gpu_device->ResizeTexture(&m_deinterlace_buffers[dst_bufidx], width, height, GPUTexture::Type::RenderTarget,
                                     GPUTexture::Format::RGBA8, false)) [[unlikely]]
    {
      return false;
    }

    GL_OBJECT_NAME_FMT(m_deinterlace_buffers[dst_bufidx], "Blend Deinterlace Buffer {}", dst_bufidx);
  }

  GPUTexture* dst = m_deinterlace_buffers[dst_bufidx].get();
  g_gpu_device->InvalidateRenderTarget(dst);

  // If we're not skipping lines, then we can simply copy the texture.
  if (line_skip == 0 && src->GetFormat() == dst->GetFormat())
  {
    GL_INS_FMT("DeinterlaceExtractField({{{},{}}} {}x{} line_skip={}) => copy direct", x, y, width, height, line_skip);
    g_gpu_device->CopyTextureRegion(dst, 0, 0, 0, 0, src, x, y, 0, 0, width, height);
  }
  else
  {
    GL_SCOPE_FMT("DeinterlaceExtractField({{{},{}}} {}x{} line_skip={}) => shader copy", x, y, width, height,
                 line_skip);

    // Otherwise, we need to extract every other line from the texture.
    src->MakeReadyForSampling();
    g_gpu_device->SetRenderTarget(dst);
    g_gpu_device->SetPipeline(m_deinterlace_extract_pipeline.get());
    g_gpu_device->SetTextureSampler(0, src, g_gpu_device->GetNearestSampler());
    const u32 uniforms[] = {x, y, line_skip};
    g_gpu_device->PushUniformBuffer(uniforms, sizeof(uniforms));
    g_gpu_device->SetViewportAndScissor(0, 0, width, height);
    g_gpu_device->Draw(3, 0);

    GL_POP();
  }

  dst->MakeReadyForSampling();
  return true;
}

bool GPU::DeinterlaceSetTargetSize(u32 width, u32 height, bool preserve)
{
  if (!m_deinterlace_texture || m_deinterlace_texture->GetWidth() != width ||
      m_deinterlace_texture->GetHeight() != height)
  {
    if (!g_gpu_device->ResizeTexture(&m_deinterlace_texture, width, height, GPUTexture::Type::RenderTarget,
                                     GPUTexture::Format::RGBA8, preserve)) [[unlikely]]
    {
      return false;
    }

    GL_OBJECT_NAME(m_deinterlace_texture, "Deinterlace target texture");
  }

  return true;
}

bool GPU::ApplyChromaSmoothing(GPUTexture* src, u32 x, u32 y, u32 width, u32 height)
{
  if (!m_chroma_smoothing_texture || m_chroma_smoothing_texture->GetWidth() != width ||
      m_chroma_smoothing_texture->GetHeight() != height)
  {
    if (!g_gpu_device->ResizeTexture(&m_chroma_smoothing_texture, width, height, GPUTexture::Type::RenderTarget,
                                     GPUTexture::Format::RGBA8, false))
    {
      ClearDisplayTexture();
      return false;
    }

    GL_OBJECT_NAME(m_chroma_smoothing_texture, "Chroma smoothing texture");
  }

  GL_SCOPE_FMT("ApplyChromaSmoothing({{{},{}}}, {}x{})", x, y, width, height);

  src->MakeReadyForSampling();
  g_gpu_device->InvalidateRenderTarget(m_chroma_smoothing_texture.get());
  g_gpu_device->SetRenderTarget(m_chroma_smoothing_texture.get());
  g_gpu_device->SetPipeline(m_chroma_smoothing_pipeline.get());
  g_gpu_device->SetTextureSampler(0, src, g_gpu_device->GetNearestSampler());
  const u32 uniforms[] = {x, y, width - 1, height - 1};
  g_gpu_device->PushUniformBuffer(uniforms, sizeof(uniforms));
  g_gpu_device->SetViewportAndScissor(0, 0, width, height);
  g_gpu_device->Draw(3, 0);

  m_chroma_smoothing_texture->MakeReadyForSampling();
  SetDisplayTexture(m_chroma_smoothing_texture.get(), 0, 0, width, height);
  return true;
}

Common::Rectangle<float> GPU::CalculateDrawRect(s32 window_width, s32 window_height, float* out_left_padding,
                                                float* out_top_padding, float* out_scale, float* out_x_scale,
                                                bool apply_aspect_ratio /* = true */) const
{
  const float window_ratio = static_cast<float>(window_width) / static_cast<float>(window_height);
  const float x_scale =
    apply_aspect_ratio ?
      (m_display_aspect_ratio / (static_cast<float>(m_display_width) / static_cast<float>(m_display_height))) :
      1.0f;
  const float display_width = g_settings.display_stretch_vertically ? static_cast<float>(m_display_width) :
                                                                      static_cast<float>(m_display_width) * x_scale;
  const float display_height = g_settings.display_stretch_vertically ? static_cast<float>(m_display_height) / x_scale :
                                                                       static_cast<float>(m_display_height);
  const float active_left = g_settings.display_stretch_vertically ? static_cast<float>(m_display_active_left) :
                                                                    static_cast<float>(m_display_active_left) * x_scale;
  const float active_top = g_settings.display_stretch_vertically ? static_cast<float>(m_display_active_top) / x_scale :
                                                                   static_cast<float>(m_display_active_top);
  const float active_width = g_settings.display_stretch_vertically ?
                               static_cast<float>(m_display_active_width) :
                               static_cast<float>(m_display_active_width) * x_scale;
  const float active_height = g_settings.display_stretch_vertically ?
                                static_cast<float>(m_display_active_height) / x_scale :
                                static_cast<float>(m_display_active_height);
  if (out_x_scale)
    *out_x_scale = x_scale;

  // now fit it within the window
  float scale;
  if ((display_width / display_height) >= window_ratio)
  {
    // align in middle vertically
    scale = static_cast<float>(window_width) / display_width;
    if (g_settings.display_scaling == DisplayScalingMode::NearestInteger)
      scale = std::max(std::floor(scale), 1.0f);

    if (out_left_padding)
    {
      if (g_settings.display_scaling == DisplayScalingMode::NearestInteger)
        *out_left_padding = std::max<float>((static_cast<float>(window_width) - display_width * scale) / 2.0f, 0.0f);
      else
        *out_left_padding = 0.0f;
    }
    if (out_top_padding)
    {
      switch (g_settings.display_alignment)
      {
        case DisplayAlignment::RightOrBottom:
          *out_top_padding = std::max<float>(static_cast<float>(window_height) - (display_height * scale), 0.0f);
          break;

        case DisplayAlignment::Center:
          *out_top_padding =
            std::max<float>((static_cast<float>(window_height) - (display_height * scale)) / 2.0f, 0.0f);
          break;

        case DisplayAlignment::LeftOrTop:
        default:
          *out_top_padding = 0.0f;
          break;
      }
    }
  }
  else
  {
    // align in middle horizontally
    scale = static_cast<float>(window_height) / display_height;
    if (g_settings.display_scaling == DisplayScalingMode::NearestInteger)
      scale = std::max(std::floor(scale), 1.0f);

    if (out_left_padding)
    {
      switch (g_settings.display_alignment)
      {
        case DisplayAlignment::RightOrBottom:
          *out_left_padding = std::max<float>(static_cast<float>(window_width) - (display_width * scale), 0.0f);
          break;

        case DisplayAlignment::Center:
          *out_left_padding =
            std::max<float>((static_cast<float>(window_width) - (display_width * scale)) / 2.0f, 0.0f);
          break;

        case DisplayAlignment::LeftOrTop:
        default:
          *out_left_padding = 0.0f;
          break;
      }
    }

    if (out_top_padding)
    {
      if (g_settings.display_scaling == DisplayScalingMode::NearestInteger)
        *out_top_padding = std::max<float>((static_cast<float>(window_height) - (display_height * scale)) / 2.0f, 0.0f);
      else
        *out_top_padding = 0.0f;
    }
  }

  if (out_scale)
    *out_scale = scale;

  return Common::Rectangle<float>::FromExtents(active_left * scale, active_top * scale, active_width * scale,
                                               active_height * scale);
}

Common::Rectangle<s32> GPU::CalculateDrawRect(s32 window_width, s32 window_height,
                                              bool apply_aspect_ratio /* = true */) const
{
  float left_padding, top_padding;
  const Common::Rectangle<float> draw_rc =
    CalculateDrawRect(window_width, window_height, &left_padding, &top_padding, nullptr, nullptr, apply_aspect_ratio);

  // TODO: This should be a float rectangle. But because GL is lame, it only has integer viewports...
  return Common::Rectangle<s32>::FromExtents(
    static_cast<s32>(draw_rc.left + left_padding), static_cast<s32>(draw_rc.top + top_padding),
    static_cast<s32>(draw_rc.GetWidth()), static_cast<s32>(draw_rc.GetHeight()));
}

bool CompressAndWriteTextureToFile(u32 width, u32 height, std::string filename, FileSystem::ManagedCFilePtr fp,
                                   u8 quality, bool clear_alpha, bool flip_y, std::vector<u32> texture_data,
                                   u32 texture_data_stride, GPUTexture::Format texture_format, bool display_osd_message,
                                   bool use_thread)
{
  std::string osd_key;
  if (display_osd_message)
  {
    // Use a 60 second timeout to give it plenty of time to actually save.
    osd_key = fmt::format("ScreenshotSaver_{}", filename);
    Host::AddIconOSDMessage(osd_key, ICON_FA_CAMERA,
                            fmt::format(TRANSLATE_FS("GPU", "Saving screenshot to '{}'."), Path::GetFileName(filename)),
                            60.0f);
  }

  static constexpr auto proc = [](u32 width, u32 height, std::string filename, FileSystem::ManagedCFilePtr fp,
                                  u8 quality, bool clear_alpha, bool flip_y, std::vector<u32> texture_data,
                                  u32 texture_data_stride, GPUTexture::Format texture_format, std::string osd_key,
                                  bool use_thread) {
    bool result;

    const char* extension = std::strrchr(filename.c_str(), '.');
    if (extension)
    {
      if (GPUTexture::ConvertTextureDataToRGBA8(width, height, texture_data, texture_data_stride, texture_format))
      {
        if (clear_alpha)
        {
          for (u32& pixel : texture_data)
            pixel |= 0xFF000000u;
        }

        if (flip_y)
          GPUTexture::FlipTextureDataRGBA8(width, height, reinterpret_cast<u8*>(texture_data.data()),
                                           texture_data_stride);

        Assert(texture_data_stride == sizeof(u32) * width);
        RGBA8Image image(width, height, std::move(texture_data));
        if (image.SaveToFile(filename.c_str(), fp.get(), quality))
        {
          result = true;
        }
        else
        {
          Log_ErrorPrintf("Unknown extension in filename '%s' or save error: '%s'", filename.c_str(), extension);
          result = false;
        }
      }
      else
      {
        result = false;
      }
    }
    else
    {
      Log_ErrorPrintf("Unable to determine file extension for '%s'", filename.c_str());
      result = false;
    }

    if (!osd_key.empty())
    {
      Host::AddIconOSDMessage(std::move(osd_key), ICON_FA_CAMERA,
                              fmt::format(result ? TRANSLATE_FS("GS", "Saved screenshot to '{}'.") :
                                                   TRANSLATE_FS("GPU", "Failed to save screenshot to '{}'."),
                                          Path::GetFileName(filename),
                                          result ? Host::OSD_INFO_DURATION : Host::OSD_ERROR_DURATION));
    }

    if (use_thread)
    {
      // remove ourselves from the list, if the GS thread is waiting for us, we won't be in there
      const auto this_id = std::this_thread::get_id();
      std::unique_lock lock(s_screenshot_threads_mutex);
      for (auto it = s_screenshot_threads.begin(); it != s_screenshot_threads.end(); ++it)
      {
        if (it->get_id() == this_id)
        {
          it->detach();
          s_screenshot_threads.erase(it);
          break;
        }
      }
    }

    return result;
  };

  if (!use_thread)
  {
    return proc(width, height, std::move(filename), std::move(fp), quality, clear_alpha, flip_y,
                std::move(texture_data), texture_data_stride, texture_format, std::move(osd_key), use_thread);
  }

  std::thread thread(proc, width, height, std::move(filename), std::move(fp), quality, clear_alpha, flip_y,
                     std::move(texture_data), texture_data_stride, texture_format, std::move(osd_key), use_thread);
  std::unique_lock lock(s_screenshot_threads_mutex);
  s_screenshot_threads.push_back(std::move(thread));
  return true;
}

void JoinScreenshotThreads()
{
  std::unique_lock lock(s_screenshot_threads_mutex);
  while (!s_screenshot_threads.empty())
  {
    std::thread save_thread(std::move(s_screenshot_threads.front()));
    s_screenshot_threads.pop_front();
    lock.unlock();
    save_thread.join();
    lock.lock();
  }
}

bool GPU::WriteDisplayTextureToFile(std::string filename, bool compress_on_thread /* = false */)
{
  if (!m_display_texture)
    return false;

  const u32 read_x = static_cast<u32>(m_display_texture_view_x);
  const u32 read_y = static_cast<u32>(m_display_texture_view_y);
  const u32 read_width = static_cast<u32>(m_display_texture_view_width);
  const u32 read_height = static_cast<u32>(m_display_texture_view_height);

  const u32 texture_data_stride =
    Common::AlignUpPow2(GPUTexture::GetPixelSize(m_display_texture->GetFormat()) * read_width, 4);
  std::vector<u32> texture_data((texture_data_stride * read_height) / sizeof(u32));

  std::unique_ptr<GPUDownloadTexture> dltex;
  if (g_gpu_device->GetFeatures().memory_import)
  {
    dltex =
      g_gpu_device->CreateDownloadTexture(read_width, read_height, m_display_texture->GetFormat(), texture_data.data(),
                                          texture_data.size() * sizeof(u32), texture_data_stride);
  }
  if (!dltex)
  {
    if (!(dltex = g_gpu_device->CreateDownloadTexture(read_width, read_height, m_display_texture->GetFormat())))
    {
      Log_ErrorFmt("Failed to create {}x{} {} download texture", read_width, read_height,
                   GPUTexture::GetFormatName(m_display_texture->GetFormat()));
      return false;
    }
  }

  dltex->CopyFromTexture(0, 0, m_display_texture, read_x, read_y, read_width, read_height, 0, 0, !dltex->IsImported());
  if (!dltex->ReadTexels(0, 0, read_width, read_height, texture_data.data(), texture_data_stride))
  {
    RestoreDeviceContext();
    return false;
  }

  RestoreDeviceContext();

  auto fp = FileSystem::OpenManagedCFile(filename.c_str(), "wb");
  if (!fp)
  {
    Log_ErrorPrintf("Can't open file '%s': errno %d", filename.c_str(), errno);
    return false;
  }

  constexpr bool clear_alpha = true;
  const bool flip_y = g_gpu_device->UsesLowerLeftOrigin();

  return CompressAndWriteTextureToFile(
    read_width, read_height, std::move(filename), std::move(fp), g_settings.display_screenshot_quality, clear_alpha,
    flip_y, std::move(texture_data), texture_data_stride, m_display_texture->GetFormat(), false, compress_on_thread);
}

bool GPU::RenderScreenshotToBuffer(u32 width, u32 height, const Common::Rectangle<s32>& draw_rect, bool postfx,
                                   std::vector<u32>* out_pixels, u32* out_stride, GPUTexture::Format* out_format)
{
  const GPUTexture::Format hdformat =
    g_gpu_device->HasSurface() ? g_gpu_device->GetWindowFormat() : GPUTexture::Format::RGBA8;

  auto render_texture =
    g_gpu_device->FetchAutoRecycleTexture(width, height, 1, 1, 1, GPUTexture::Type::RenderTarget, hdformat);
  if (!render_texture)
    return false;

  g_gpu_device->ClearRenderTarget(render_texture.get(), 0);

  // TODO: this should use copy shader instead.
  RenderDisplay(render_texture.get(), draw_rect, postfx);

  const u32 stride = Common::AlignUpPow2(GPUTexture::GetPixelSize(hdformat) * width, sizeof(u32));
  out_pixels->resize((height * stride) / sizeof(u32));

  std::unique_ptr<GPUDownloadTexture> dltex;
  if (g_gpu_device->GetFeatures().memory_import)
  {
    dltex = g_gpu_device->CreateDownloadTexture(width, height, hdformat, out_pixels->data(),
                                                out_pixels->size() * sizeof(u32), stride);
  }
  if (!dltex)
  {
    if (!(dltex = g_gpu_device->CreateDownloadTexture(width, height, hdformat)))
    {
      Log_ErrorFmt("Failed to create {}x{} download texture", width, height);
      return false;
    }
  }

  dltex->CopyFromTexture(0, 0, render_texture.get(), 0, 0, width, height, 0, 0, false);
  if (!dltex->ReadTexels(0, 0, width, height, out_pixels->data(), stride))
  {
    RestoreDeviceContext();
    return false;
  }

  *out_stride = stride;
  *out_format = hdformat;
  RestoreDeviceContext();
  return true;
}

bool GPU::RenderScreenshotToFile(std::string filename, DisplayScreenshotMode mode, u8 quality, bool compress_on_thread,
                                 bool show_osd_message)
{
  u32 width = g_gpu_device->GetWindowWidth();
  u32 height = g_gpu_device->GetWindowHeight();
  Common::Rectangle<s32> draw_rect = CalculateDrawRect(width, height);

  const bool internal_resolution = (mode != DisplayScreenshotMode::ScreenResolution);
  if (internal_resolution && m_display_texture_view_width != 0 && m_display_texture_view_height != 0)
  {
    if (mode == DisplayScreenshotMode::InternalResolution)
    {
      const u32 draw_width = static_cast<u32>(draw_rect.GetWidth());
      const u32 draw_height = static_cast<u32>(draw_rect.GetHeight());

      // If internal res, scale the computed draw rectangle to the internal res.
      // We re-use the draw rect because it's already been AR corrected.
      const float sar =
        static_cast<float>(m_display_texture_view_width) / static_cast<float>(m_display_texture_view_height);
      const float dar = static_cast<float>(draw_width) / static_cast<float>(draw_height);
      if (sar >= dar)
      {
        // stretch height, preserve width
        const float scale = static_cast<float>(m_display_texture_view_width) / static_cast<float>(draw_width);
        width = m_display_texture_view_width;
        height = static_cast<u32>(std::round(static_cast<float>(draw_height) * scale));
      }
      else
      {
        // stretch width, preserve height
        const float scale = static_cast<float>(m_display_texture_view_height) / static_cast<float>(draw_height);
        width = static_cast<u32>(std::round(static_cast<float>(draw_width) * scale));
        height = m_display_texture_view_height;
      }

      // DX11 won't go past 16K texture size.
      const u32 max_texture_size = g_gpu_device->GetMaxTextureSize();
      if (width > max_texture_size)
      {
        height = static_cast<u32>(static_cast<float>(height) /
                                  (static_cast<float>(width) / static_cast<float>(max_texture_size)));
        width = max_texture_size;
      }
      if (height > max_texture_size)
      {
        height = max_texture_size;
        width = static_cast<u32>(static_cast<float>(width) /
                                 (static_cast<float>(height) / static_cast<float>(max_texture_size)));
      }
    }
    else // if (mode == DisplayScreenshotMode::UncorrectedInternalResolution)
    {
      width = m_display_texture_view_width;
      height = m_display_texture_view_height;
    }

    // Remove padding, it's not part of the framebuffer.
    draw_rect.Set(0, 0, static_cast<s32>(width), static_cast<s32>(height));
  }
  if (width == 0 || height == 0)
    return false;

  std::vector<u32> pixels;
  u32 pixels_stride;
  GPUTexture::Format pixels_format;
  if (!RenderScreenshotToBuffer(width, height, draw_rect, !internal_resolution, &pixels, &pixels_stride,
                                &pixels_format))
  {
    Log_ErrorPrintf("Failed to render %ux%u screenshot", width, height);
    return false;
  }

  // These filenames tend to be fairly long, so remove any MAX_PATH limit.
  auto fp = FileSystem::OpenManagedCFile(Path::RemoveLengthLimits(filename).c_str(), "wb");
  if (!fp)
  {
    Log_ErrorPrintf("Can't open file '%s': errno %d", filename.c_str(), errno);
    return false;
  }

  return CompressAndWriteTextureToFile(width, height, std::move(filename), std::move(fp), quality, true,
                                       g_gpu_device->UsesLowerLeftOrigin(), std::move(pixels), pixels_stride,
                                       pixels_format, show_osd_message, compress_on_thread);
}

bool GPU::DumpVRAMToFile(const char* filename)
{
  ReadVRAM(0, 0, VRAM_WIDTH, VRAM_HEIGHT);

  const char* extension = std::strrchr(filename, '.');
  if (extension && StringUtil::Strcasecmp(extension, ".png") == 0)
  {
    return DumpVRAMToFile(filename, VRAM_WIDTH, VRAM_HEIGHT, sizeof(u16) * VRAM_WIDTH, g_vram, true);
  }
  else if (extension && StringUtil::Strcasecmp(extension, ".bin") == 0)
  {
    return FileSystem::WriteBinaryFile(filename, g_vram, VRAM_WIDTH * VRAM_HEIGHT * sizeof(u16));
  }
  else
  {
    Log_ErrorPrintf("Unknown extension: '%s'", filename);
    return false;
  }
}

bool GPU::DumpVRAMToFile(const char* filename, u32 width, u32 height, u32 stride, const void* buffer, bool remove_alpha)
{
  RGBA8Image image(width, height);

  const char* ptr_in = static_cast<const char*>(buffer);
  for (u32 row = 0; row < height; row++)
  {
    const char* row_ptr_in = ptr_in;
    u32* ptr_out = image.GetRowPixels(row);

    for (u32 col = 0; col < width; col++)
    {
      u16 src_col;
      std::memcpy(&src_col, row_ptr_in, sizeof(u16));
      row_ptr_in += sizeof(u16);
      *(ptr_out++) = VRAMRGBA5551ToRGBA8888(remove_alpha ? (src_col | u16(0x8000)) : src_col);
    }

    ptr_in += stride;
  }

  return image.SaveToFile(filename);
}

void GPU::DrawDebugStateWindow()
{
  const float framebuffer_scale = Host::GetOSDScale();

  ImGui::SetNextWindowSize(ImVec2(450.0f * framebuffer_scale, 550.0f * framebuffer_scale), ImGuiCond_FirstUseEver);
  if (!ImGui::Begin("GPU", nullptr))
  {
    ImGui::End();
    return;
  }

  DrawRendererStats();

  if (ImGui::CollapsingHeader("GPU", ImGuiTreeNodeFlags_DefaultOpen))
  {
    static constexpr std::array<const char*, 5> state_strings = {
      {"Idle", "Reading VRAM", "Writing VRAM", "Drawing Polyline"}};

    ImGui::Text("State: %s", state_strings[static_cast<u8>(m_blitter_state)]);
    ImGui::Text("Dither: %s", m_GPUSTAT.dither_enable ? "Enabled" : "Disabled");
    ImGui::Text("Draw To Displayed Field: %s", m_GPUSTAT.draw_to_displayed_field ? "Enabled" : "Disabled");
    ImGui::Text("Draw Set Mask Bit: %s", m_GPUSTAT.set_mask_while_drawing ? "Yes" : "No");
    ImGui::Text("Draw To Masked Pixels: %s", m_GPUSTAT.check_mask_before_draw ? "Yes" : "No");
    ImGui::Text("Reverse Flag: %s", m_GPUSTAT.reverse_flag ? "Yes" : "No");
    ImGui::Text("Texture Disable: %s", m_GPUSTAT.texture_disable ? "Yes" : "No");
    ImGui::Text("PAL Mode: %s", m_GPUSTAT.pal_mode ? "Yes" : "No");
    ImGui::Text("Interrupt Request: %s", m_GPUSTAT.interrupt_request ? "Yes" : "No");
    ImGui::Text("DMA Request: %s", m_GPUSTAT.dma_data_request ? "Yes" : "No");
  }

  if (ImGui::CollapsingHeader("CRTC", ImGuiTreeNodeFlags_DefaultOpen))
  {
    const auto& cs = m_crtc_state;
    ImGui::Text("Clock: %s", (m_console_is_pal ? (m_GPUSTAT.pal_mode ? "PAL-on-PAL" : "NTSC-on-PAL") :
                                                 (m_GPUSTAT.pal_mode ? "PAL-on-NTSC" : "NTSC-on-NTSC")));
    ImGui::Text("Horizontal Frequency: %.3f KHz", ComputeHorizontalFrequency() / 1000.0f);
    ImGui::Text("Vertical Frequency: %.3f Hz", ComputeVerticalFrequency());
    ImGui::Text("Dot Clock Divider: %u", cs.dot_clock_divider);
    ImGui::Text("Vertical Interlace: %s (%s field)", m_GPUSTAT.vertical_interlace ? "Yes" : "No",
                cs.interlaced_field ? "odd" : "even");
    ImGui::Text("Current Scanline: %u (tick %u)", cs.current_scanline, cs.current_tick_in_scanline);
    ImGui::Text("Display Disable: %s", m_GPUSTAT.display_disable ? "Yes" : "No");
    ImGui::Text("Displaying Odd Lines: %s", cs.active_line_lsb ? "Yes" : "No");
    ImGui::Text("Color Depth: %u-bit", m_GPUSTAT.display_area_color_depth_24 ? 24 : 15);
    ImGui::Text("Start Offset in VRAM: (%u, %u)", cs.regs.X.GetValue(), cs.regs.Y.GetValue());
    ImGui::Text("Display Total: %u (%u) horizontal, %u vertical", cs.horizontal_total,
                cs.horizontal_total / cs.dot_clock_divider, cs.vertical_total);
    ImGui::Text("Configured Display Range: %u-%u (%u-%u), %u-%u", cs.regs.X1.GetValue(), cs.regs.X2.GetValue(),
                cs.regs.X1.GetValue() / cs.dot_clock_divider, cs.regs.X2.GetValue() / cs.dot_clock_divider,
                cs.regs.Y1.GetValue(), cs.regs.Y2.GetValue());
    ImGui::Text("Output Display Range: %u-%u (%u-%u), %u-%u", cs.horizontal_display_start, cs.horizontal_display_end,
                cs.horizontal_display_start / cs.dot_clock_divider, cs.horizontal_display_end / cs.dot_clock_divider,
                cs.vertical_display_start, cs.vertical_display_end);
    ImGui::Text("Cropping: %s", Settings::GetDisplayCropModeName(g_settings.display_crop_mode));
    ImGui::Text("Visible Display Range: %u-%u (%u-%u), %u-%u", cs.horizontal_visible_start, cs.horizontal_visible_end,
                cs.horizontal_visible_start / cs.dot_clock_divider, cs.horizontal_visible_end / cs.dot_clock_divider,
                cs.vertical_visible_start, cs.vertical_visible_end);
    ImGui::Text("Display Resolution: %ux%u", cs.display_width, cs.display_height);
    ImGui::Text("Display Origin: %u, %u", cs.display_origin_left, cs.display_origin_top);
    ImGui::Text("Displayed/Visible VRAM Portion: %ux%u @ (%u, %u)", cs.display_vram_width, cs.display_vram_height,
                cs.display_vram_left, cs.display_vram_top);
    ImGui::Text("Padding: Left=%d, Top=%d, Right=%d, Bottom=%d", cs.display_origin_left, cs.display_origin_top,
                cs.display_width - cs.display_vram_width - cs.display_origin_left,
                cs.display_height - cs.display_vram_height - cs.display_origin_top);
  }

  ImGui::End();
}

void GPU::DrawRendererStats()
{
}

void GPU::GetStatsString(SmallStringBase& str)
{
  if (IsHardwareRenderer())
  {
    str.format("{} HW | {} P | {} DC | {} RP | {} RB | {} C | {} W",
               GPUDevice::RenderAPIToString(g_gpu_device->GetRenderAPI()), m_stats.num_primitives,
               m_stats.host_num_draws, m_stats.host_num_render_passes, m_stats.num_reads, m_stats.num_copies,
               m_stats.num_writes);
  }
  else
  {
    str.format("{} SW | {} P | {} R | {} C | {} W", GPUDevice::RenderAPIToString(g_gpu_device->GetRenderAPI()),
               m_stats.num_primitives, m_stats.num_reads, m_stats.num_copies, m_stats.num_writes);
  }
}

void GPU::GetMemoryStatsString(SmallStringBase& str)
{
  const u32 vram_usage_mb = static_cast<u32>((g_gpu_device->GetVRAMUsage() + (1048576 - 1)) / 1048576);
  const u32 stream_kb = static_cast<u32>((m_stats.host_buffer_streamed + (1024 - 1)) / 1024);

  str.format("{} MB VRAM | {} KB STR | {} TC | {} TU", vram_usage_mb, stream_kb, m_stats.host_num_copies,
             m_stats.host_num_uploads);
}

void GPU::ResetStatistics()
{
  m_counters = {};
  g_gpu_device->ResetStatistics();
}

void GPU::UpdateStatistics(u32 frame_count)
{
  const GPUDevice::Statistics& stats = g_gpu_device->GetStatistics();
  const u32 round = (frame_count - 1);

#define UPDATE_COUNTER(x) m_stats.x = (m_counters.x + round) / frame_count
#define UPDATE_GPU_STAT(x) m_stats.host_##x = (stats.x + round) / frame_count

  UPDATE_COUNTER(num_reads);
  UPDATE_COUNTER(num_writes);
  UPDATE_COUNTER(num_copies);
  UPDATE_COUNTER(num_vertices);
  UPDATE_COUNTER(num_primitives);

  // UPDATE_COUNTER(num_read_texture_updates);
  // UPDATE_COUNTER(num_ubo_updates);

  UPDATE_GPU_STAT(buffer_streamed);
  UPDATE_GPU_STAT(num_draws);
  UPDATE_GPU_STAT(num_render_passes);
  UPDATE_GPU_STAT(num_copies);
  UPDATE_GPU_STAT(num_downloads);
  UPDATE_GPU_STAT(num_uploads);

#undef UPDATE_GPU_STAT
#undef UPDATE_COUNTER

  ResetStatistics();
}