GPU/HW: Tidying up

This commit is contained in:
Stenzek 2023-09-20 18:53:29 +10:00
parent 3cd28f433b
commit 18ae01bedb
2 changed files with 104 additions and 77 deletions

View file

@ -146,6 +146,49 @@ GPU_HW::~GPU_HW()
}
}
ALWAYS_INLINE void GPU_HW::BatchVertex::Set(float x_, float y_, float z_, float w_, u32 color_, u32 texpage_,
u16 packed_texcoord, u32 uv_limits_)
{
Set(x_, y_, z_, w_, color_, texpage_, packed_texcoord & 0xFF, (packed_texcoord >> 8), uv_limits_);
}
ALWAYS_INLINE void GPU_HW::BatchVertex::Set(float x_, float y_, float z_, float w_, u32 color_, u32 texpage_, u16 u_,
u16 v_, u32 uv_limits_)
{
x = x_;
y = y_;
z = z_;
w = w_;
color = color_;
texpage = texpage_;
u = u_;
v = v_;
uv_limits = uv_limits_;
}
ALWAYS_INLINE u32 GPU_HW::BatchVertex::PackUVLimits(u32 min_u, u32 max_u, u32 min_v, u32 max_v)
{
return min_u | (min_v << 8) | (max_u << 16) | (max_v << 24);
}
ALWAYS_INLINE void GPU_HW::BatchVertex::SetUVLimits(u32 min_u, u32 max_u, u32 min_v, u32 max_v)
{
uv_limits = PackUVLimits(min_u, max_u, min_v, max_v);
}
ALWAYS_INLINE void GPU_HW::AddVertex(const BatchVertex& v)
{
std::memcpy(m_batch_current_vertex_ptr, &v, sizeof(BatchVertex));
m_batch_current_vertex_ptr++;
}
template<typename... Args>
ALWAYS_INLINE void GPU_HW::AddNewVertex(Args&&... args)
{
m_batch_current_vertex_ptr->Set(std::forward<Args>(args)...);
m_batch_current_vertex_ptr++;
}
const Threading::Thread* GPU_HW::GetSWThread() const
{
return m_sw_renderer ? m_sw_renderer->GetThread() : nullptr;
@ -1160,6 +1203,12 @@ void GPU_HW::DestroyPipelines()
m_display_pipelines.enumerate(destroy);
}
GPU_HW::BatchRenderMode GPU_HW::BatchConfig::GetRenderMode() const
{
return transparency_mode == GPUTransparencyMode::Disabled ? BatchRenderMode::TransparencyDisabled :
BatchRenderMode::TransparentAndOpaque;
}
void GPU_HW::UpdateVRAMReadTexture()
{
GL_SCOPE("UpdateVRAMReadTexture()");
@ -1969,6 +2018,11 @@ void GPU_HW::IncludeVRAMDirtyRectangle(const Common::Rectangle<u32>& rect)
}
}
ALWAYS_INLINE bool GPU_HW::IsFlushed() const
{
return m_batch_current_vertex_ptr == m_batch_start_vertex_ptr;
}
GPU_HW::InterlacedRenderMode GPU_HW::GetInterlacedRenderMode() const
{
if (IsInterlacedDisplayEnabled())
@ -1982,6 +2036,27 @@ GPU_HW::InterlacedRenderMode GPU_HW::GetInterlacedRenderMode() const
}
}
ALWAYS_INLINE bool GPU_HW::NeedsTwoPassRendering() const
{
// We need two-pass rendering when using BG-FG blending and texturing, as the transparency can be enabled
// on a per-pixel basis, and the opaque pixels shouldn't be blended at all.
// TODO: see if there's a better way we can do this. definitely can with fbfetch.
return (m_batch.texture_mode != GPUTextureMode::Disabled &&
(m_batch.transparency_mode == GPUTransparencyMode::BackgroundMinusForeground ||
(!m_supports_dual_source_blend && m_batch.transparency_mode != GPUTransparencyMode::Disabled)));
}
ALWAYS_INLINE u32 GPU_HW::GetBatchVertexSpace() const
{
return static_cast<u32>(m_batch_end_vertex_ptr - m_batch_current_vertex_ptr);
}
ALWAYS_INLINE u32 GPU_HW::GetBatchVertexCount() const
{
return static_cast<u32>(m_batch_current_vertex_ptr - m_batch_start_vertex_ptr);
}
void GPU_HW::EnsureVertexBufferSpace(u32 required_vertices)
{
if (m_batch_current_vertex_ptr)
@ -2041,6 +2116,11 @@ void GPU_HW::ResetBatchVertexDepth()
m_current_depth = 1;
}
ALWAYS_INLINE float GPU_HW::GetCurrentNormalizedVertexDepth() const
{
return 1.0f - (static_cast<float>(m_current_depth) / 65535.0f);
}
void GPU_HW::UpdateSoftwareRenderer(bool copy_vram_from_hw)
{
const bool current_enabled = (m_sw_renderer != nullptr);

View file

@ -56,6 +56,9 @@ public:
std::tuple<u32, u32> GetEffectiveDisplayResolution(bool scaled = true) override final;
std::tuple<u32, u32> GetFullDisplayResolution(bool scaled = true) override final;
void ClearDisplay() override;
void UpdateDisplay() override;
private:
enum : u32
{
@ -78,35 +81,10 @@ private:
u16 v;
u32 uv_limits;
ALWAYS_INLINE void Set(float x_, float y_, float z_, float w_, u32 color_, u32 texpage_, u16 packed_texcoord,
u32 uv_limits_)
{
Set(x_, y_, z_, w_, color_, texpage_, packed_texcoord & 0xFF, (packed_texcoord >> 8), uv_limits_);
}
ALWAYS_INLINE void Set(float x_, float y_, float z_, float w_, u32 color_, u32 texpage_, u16 u_, u16 v_,
u32 uv_limits_)
{
x = x_;
y = y_;
z = z_;
w = w_;
color = color_;
texpage = texpage_;
u = u_;
v = v_;
uv_limits = uv_limits_;
}
ALWAYS_INLINE static u32 PackUVLimits(u32 min_u, u32 max_u, u32 min_v, u32 max_v)
{
return min_u | (min_v << 8) | (max_u << 16) | (max_v << 24);
}
ALWAYS_INLINE void SetUVLimits(u32 min_u, u32 max_u, u32 min_v, u32 max_v)
{
uv_limits = PackUVLimits(min_u, max_u, min_v, max_v);
}
void Set(float x_, float y_, float z_, float w_, u32 color_, u32 texpage_, u16 packed_texcoord, u32 uv_limits_);
void Set(float x_, float y_, float z_, float w_, u32 color_, u32 texpage_, u16 u_, u16 v_, u32 uv_limits_);
static u32 PackUVLimits(u32 min_u, u32 max_u, u32 min_v, u32 max_v);
void SetUVLimits(u32 min_u, u32 max_u, u32 min_v, u32 max_v);
};
struct BatchConfig
@ -120,11 +98,7 @@ private:
bool use_depth_buffer = false;
// Returns the render mode for this batch.
BatchRenderMode GetRenderMode() const
{
return transparency_mode == GPUTransparencyMode::Disabled ? BatchRenderMode::TransparencyDisabled :
BatchRenderMode::TransparentAndOpaque;
}
BatchRenderMode GetRenderMode() const;
};
struct BatchUBOData
@ -151,6 +125,16 @@ private:
bool CompilePipelines();
void DestroyPipelines();
void LoadVertices();
void AddVertex(const BatchVertex& v);
template<typename... Args>
void AddNewVertex(Args&&... args);
void PrintSettingsToLog();
void CheckSettings();
void UpdateVRAMReadTexture();
void UpdateDepthBufferFromMaskBit();
void ClearDepthBuffer();
@ -158,8 +142,6 @@ private:
void MapBatchVertexPointer(u32 required_vertices);
void UnmapBatchVertexPointer(u32 used_vertices);
void DrawBatchVertices(BatchRenderMode render_mode, u32 num_vertices, u32 base_vertex);
void ClearDisplay() override;
void UpdateDisplay() override;
u32 CalculateResolutionScale() const;
GPUDownsampleMode GetDownsampleMode(u32 resolution_scale) const;
@ -171,37 +153,21 @@ private:
void ClearVRAMDirtyRectangle();
void IncludeVRAMDirtyRectangle(const Common::Rectangle<u32>& rect);
ALWAYS_INLINE bool IsFlushed() const { return m_batch_current_vertex_ptr == m_batch_start_vertex_ptr; }
ALWAYS_INLINE u32 GetBatchVertexSpace() const
{
return static_cast<u32>(m_batch_end_vertex_ptr - m_batch_current_vertex_ptr);
}
ALWAYS_INLINE u32 GetBatchVertexCount() const
{
return static_cast<u32>(m_batch_current_vertex_ptr - m_batch_start_vertex_ptr);
}
bool IsFlushed() const;
u32 GetBatchVertexSpace() const;
u32 GetBatchVertexCount() const;
void EnsureVertexBufferSpace(u32 required_vertices);
void EnsureVertexBufferSpaceForCurrentCommand();
void ResetBatchVertexDepth();
/// Returns the value to be written to the depth buffer for the current operation for mask bit emulation.
ALWAYS_INLINE float GetCurrentNormalizedVertexDepth() const
{
return 1.0f - (static_cast<float>(m_current_depth) / 65535.0f);
}
float GetCurrentNormalizedVertexDepth() const;
/// Returns the interlaced mode to use when scanning out/displaying.
InterlacedRenderMode GetInterlacedRenderMode() const;
/// We need two-pass rendering when using BG-FG blending and texturing, as the transparency can be enabled
/// on a per-pixel basis, and the opaque pixels shouldn't be blended at all.
ALWAYS_INLINE bool NeedsTwoPassRendering() const
{
// TODO: see if there's a better way we can do this. definitely can with fbfetch.
return (m_batch.texture_mode != GPUTextureMode::Disabled &&
(m_batch.transparency_mode == GPUTransparencyMode::BackgroundMinusForeground ||
(!m_supports_dual_source_blend && m_batch.transparency_mode != GPUTransparencyMode::Disabled)));
}
/// Returns if the draw needs to be broken into opaque/transparent passes.
bool NeedsTwoPassRendering() const;
void FillBackendCommandParameters(GPUBackendCommand* cmd) const;
void FillDrawCommand(GPUBackendDrawCommand* cmd, GPURenderCommand rc) const;
@ -329,23 +295,4 @@ private:
// Statistics
RendererStats m_renderer_stats = {};
RendererStats m_last_renderer_stats = {};
private:
void LoadVertices();
ALWAYS_INLINE void AddVertex(const BatchVertex& v)
{
std::memcpy(m_batch_current_vertex_ptr, &v, sizeof(BatchVertex));
m_batch_current_vertex_ptr++;
}
template<typename... Args>
ALWAYS_INLINE void AddNewVertex(Args&&... args)
{
m_batch_current_vertex_ptr->Set(std::forward<Args>(args)...);
m_batch_current_vertex_ptr++;
}
void PrintSettingsToLog();
void CheckSettings();
};