ShuriZma
/
duckstation
mirror of https://github.com/stenzek/duckstation
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

GPU/HW: Use GSVector instead of Rectangle

This commit is contained in:
Stenzek 2024-06-01 17:04:35 +10:00
parent ef152c47a6
commit 9ef7e8c5d0
No known key found for this signature in database
28 changed files with 586 additions and 538 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

78
src/core/gpu.cpp
View File

@ -21,6 +21,7 @@
#include "common/align.h"
#include "common/error.h"
#include "common/file_system.h"
#include "common/gsvector_formatter.h"
#include "common/heap_array.h"
#include "common/log.h"
#include "common/path.h"
@ -351,6 +352,7 @@ bool GPU::DoState(StateWrapper& sw, GPUTexture** host_texture, bool update_displ
m_draw_mode.texture_page_changed = true;
m_draw_mode.texture_window_changed = true;
m_drawing_area_changed = true;
SetClampedDrawingArea();
UpdateDMARequest();
}
@ -1089,12 +1091,11 @@ void GPU::UpdateCommandTickEvent()
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());
const GSVector4i draw_rc = CalculateDrawRect(g_gpu_device->GetWindowWidth(), g_gpu_device->GetWindowHeight(), true);
// 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());
const float scaled_display_x = (window_x - static_cast<float>(draw_rc.left)) / static_cast<float>(draw_rc.width());
const float scaled_display_y = (window_y - static_cast<float>(draw_rc.top)) / static_cast<float>(draw_rc.height());
// scale back to internal resolution
*display_x = scaled_display_x * static_cast<float>(m_crtc_state.display_width);
@ -1632,6 +1633,21 @@ void GPU::CopyVRAM(u32 src_x, u32 src_y, u32 dst_x, u32 dst_y, u32 width, u32 he
}
}
void GPU::SetClampedDrawingArea()
{
if (!IsDrawingAreaIsValid()) [[unlikely]]
{
m_clamped_drawing_area = GSVector4i::zero();
return;
}
const u32 right = std::min(m_drawing_area.right + 1, static_cast<u32>(VRAM_WIDTH));
const u32 left = std::min(m_drawing_area.left, std::min(m_drawing_area.right, VRAM_WIDTH - 1));
const u32 bottom = std::min(m_drawing_area.bottom + 1, static_cast<u32>(VRAM_HEIGHT));
const u32 top = std::min(m_drawing_area.top, std::min(m_drawing_area.bottom, VRAM_HEIGHT - 1));
m_clamped_drawing_area = GSVector4i(left, top, right, bottom);
}
void GPU::SetDrawMode(u16 value)
{
GPUDrawModeReg new_mode_reg{static_cast<u16>(value & GPUDrawModeReg::MASK)};
@ -1917,16 +1933,15 @@ bool GPU::PresentDisplay()
{
FlushRender();
const Common::Rectangle<s32> draw_rect =
m_display_texture ? CalculateDrawRect(g_gpu_device->GetWindowWidth(), g_gpu_device->GetWindowHeight()) :
Common::Rectangle<s32>();
const GSVector4i draw_rect = m_display_texture ?
CalculateDrawRect(g_gpu_device->GetWindowWidth(), g_gpu_device->GetWindowHeight()) :
GSVector4i::zero();
return RenderDisplay(nullptr, draw_rect, !g_settings.debugging.show_vram);
}
bool GPU::RenderDisplay(GPUTexture* target, const Common::Rectangle<s32>& draw_rect, bool postfx)
bool GPU::RenderDisplay(GPUTexture* target, const GSVector4i draw_rect, bool postfx)
{
GL_SCOPE_FMT("RenderDisplay: {}x{} at {},{}", draw_rect.GetWidth(), draw_rect.GetHeight(), draw_rect.left,
draw_rect.top);
GL_SCOPE_FMT("RenderDisplay: {}", draw_rect);
if (m_display_texture)
m_display_texture->MakeReadyForSampling();
@ -1947,8 +1962,8 @@ bool GPU::RenderDisplay(GPUTexture* target, const Common::Rectangle<s32>& draw_r
// Now we can apply the post chain.
GPUTexture* post_output_texture = PostProcessing::InternalChain.GetOutputTexture();
if (PostProcessing::InternalChain.Apply(display_texture, m_display_depth_buffer, post_output_texture, 0, 0,
display_texture_view_width, display_texture_view_height,
if (PostProcessing::InternalChain.Apply(display_texture, m_display_depth_buffer, post_output_texture,
GSVector4i(0, 0, display_texture_view_width, display_texture_view_height),
display_texture_view_width, display_texture_view_height,
m_crtc_state.display_width, m_crtc_state.display_height))
{
@ -1966,7 +1981,7 @@ bool GPU::RenderDisplay(GPUTexture* target, const Common::Rectangle<s32>& draw_r
(postfx && PostProcessing::DisplayChain.IsActive() && !g_gpu_device->GetWindowInfo().IsSurfaceless() &&
hdformat != GPUTexture::Format::Unknown && target_width > 0 && target_height > 0 &&
PostProcessing::DisplayChain.CheckTargets(hdformat, target_width, target_height));
const Common::Rectangle<s32> real_draw_rect =
const GSVector4i real_draw_rect =
g_gpu_device->UsesLowerLeftOrigin() ? GPUDevice::FlipToLowerLeft(draw_rect, target_height) : draw_rect;
if (really_postfx)
{
@ -2009,11 +2024,9 @@ bool GPU::RenderDisplay(GPUTexture* target, const Common::Rectangle<s32>& draw_r
{
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);
std::floor(static_cast<float>(draw_rect.width()) / 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);
std::floor(static_cast<float>(draw_rect.height()) / 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];
}
@ -2048,8 +2061,7 @@ bool GPU::RenderDisplay(GPUTexture* target, const Common::Rectangle<s32>& draw_r
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->SetViewportAndScissor(real_draw_rect);
g_gpu_device->Draw(3, 0);
}
@ -2066,14 +2078,11 @@ bool GPU::RenderDisplay(GPUTexture* target, const Common::Rectangle<s32>& draw_r
const s32 orig_height = static_cast<s32>(std::ceil(static_cast<float>(m_crtc_state.display_height) * upscale_y));
return PostProcessing::DisplayChain.Apply(PostProcessing::DisplayChain.GetInputTexture(), nullptr, target,
real_draw_rect.left, real_draw_rect.top, real_draw_rect.GetWidth(),
real_draw_rect.GetHeight(), orig_width, orig_height,
m_crtc_state.display_width, m_crtc_state.display_height);
real_draw_rect, orig_width, orig_height, m_crtc_state.display_width,
m_crtc_state.display_height);
}
else
{
return true;
}
}
void GPU::DestroyDeinterlaceTextures()
@ -2303,8 +2312,7 @@ bool GPU::ApplyChromaSmoothing()
return true;
}
Common::Rectangle<s32> GPU::CalculateDrawRect(s32 window_width, s32 window_height,
bool apply_aspect_ratio /* = true */) const
GSVector4i GPU::CalculateDrawRect(s32 window_width, s32 window_height, bool apply_aspect_ratio /* = true */) const
{
const bool integer_scale = (g_settings.display_scaling == DisplayScalingMode::NearestInteger ||
g_settings.display_scaling == DisplayScalingMode::BlinearInteger);
@ -2401,9 +2409,11 @@ Common::Rectangle<s32> GPU::CalculateDrawRect(s32 window_width, s32 window_heigh
}
// 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>(active_left * scale + left_padding), static_cast<s32>(active_top * scale + top_padding),
static_cast<s32>(active_width * scale), static_cast<s32>(active_height * scale));
const s32 left = static_cast<s32>(active_left * scale + left_padding);
const s32 top = static_cast<s32>(active_top * scale + top_padding);
const s32 right = left + static_cast<s32>(active_width * scale);
const s32 bottom = top + static_cast<s32>(active_height * scale);
return GSVector4i(left, top, right, bottom);
}
bool CompressAndWriteTextureToFile(u32 width, u32 height, std::string filename, FileSystem::ManagedCFilePtr fp,
@ -2575,7 +2585,7 @@ bool GPU::WriteDisplayTextureToFile(std::string filename, bool compress_on_threa
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,
bool GPU::RenderScreenshotToBuffer(u32 width, u32 height, const GSVector4i draw_rect, bool postfx,
std::vector<u32>* out_pixels, u32* out_stride, GPUTexture::Format* out_format)
{
const GPUTexture::Format hdformat =
@ -2627,15 +2637,15 @@ bool GPU::RenderScreenshotToFile(std::string filename, DisplayScreenshotMode mod
{
u32 width = g_gpu_device->GetWindowWidth();
u32 height = g_gpu_device->GetWindowHeight();
Common::Rectangle<s32> draw_rect = CalculateDrawRect(width, height);
GSVector4i draw_rect = CalculateDrawRect(width, height, true);
const bool internal_resolution = (mode != DisplayScreenshotMode::ScreenResolution || g_settings.debugging.show_vram);
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());
const u32 draw_width = static_cast<u32>(draw_rect.width());
const u32 draw_height = static_cast<u32>(draw_rect.height());
// 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.
@ -2679,7 +2689,7 @@ bool GPU::RenderScreenshotToFile(std::string filename, DisplayScreenshotMode mod
}
// Remove padding, it's not part of the framebuffer.
draw_rect.Set(0, 0, static_cast<s32>(width), static_cast<s32>(height));
draw_rect = GSVector4i(0, 0, static_cast<s32>(width), static_cast<s32>(height));
}
if (width == 0 || height == 0)
return false;

29
src/core/gpu.h
View File

@ -10,7 +10,6 @@
#include "common/bitfield.h"
#include "common/fifo_queue.h"
#include "common/rectangle.h"
#include "common/types.h"
#include <algorithm>
@ -208,13 +207,13 @@ public:
virtual void FlushRender() = 0;
/// Helper function for computing the draw rectangle in a larger window.
Common::Rectangle<s32> CalculateDrawRect(s32 window_width, s32 window_height, bool apply_aspect_ratio = true) const;
GSVector4i CalculateDrawRect(s32 window_width, s32 window_height, bool apply_aspect_ratio = true) const;
/// Helper function to save current display texture to PNG.
bool WriteDisplayTextureToFile(std::string filename, bool compress_on_thread = false);
/// Renders the display, optionally with postprocessing to the specified image.
bool RenderScreenshotToBuffer(u32 width, u32 height, const Common::Rectangle<s32>& draw_rect, bool postfx,
bool RenderScreenshotToBuffer(u32 width, u32 height, const GSVector4i draw_rect, bool postfx,
std::vector<u32>* out_pixels, u32* out_stride, GPUTexture::Format* out_format);
/// Helper function to save screenshot to PNG.
@ -276,6 +275,9 @@ protected:
/// Returns 0 if the currently-displayed field is on an even line in VRAM, otherwise 1.
ALWAYS_INLINE u32 GetActiveLineLSB() const { return ZeroExtend32(m_crtc_state.active_line_lsb); }
/// Updates drawing area that's suitablef or clamping.
void SetClampedDrawingArea();
/// Sets/decodes GP0(E1h) (set draw mode).
void SetDrawMode(u16 bits);
@ -323,23 +325,21 @@ protected:
virtual void DrawRendererStats();
virtual void OnBufferSwapped();
ALWAYS_INLINE_RELEASE void AddDrawTriangleTicks(s32 x1, s32 y1, s32 x2, s32 y2, s32 x3, s32 y3, bool shaded,
ALWAYS_INLINE_RELEASE void AddDrawTriangleTicks(GSVector4i v1, GSVector4i v2, GSVector4i v3, bool shaded,
bool textured, bool semitransparent)
{
// This will not produce the correct results for triangles which are partially outside the clip area.
// However, usually it'll undershoot not overshoot. If we wanted to make this more accurate, we'd need to intersect
// the edges with the clip rectangle.
// TODO: Coordinates are exclusive, so off by one here...
const s32 clip_right = static_cast<s32>(m_drawing_area.right) + 1;
const s32 clip_bottom = static_cast<s32>(m_drawing_area.bottom) + 1;
x1 = std::clamp(x1, static_cast<s32>(m_drawing_area.left), clip_right);
x2 = std::clamp(x2, static_cast<s32>(m_drawing_area.left), clip_right);
x3 = std::clamp(x3, static_cast<s32>(m_drawing_area.left), clip_right);
y1 = std::clamp(y1, static_cast<s32>(m_drawing_area.top), clip_bottom);
y2 = std::clamp(y2, static_cast<s32>(m_drawing_area.top), clip_bottom);
y3 = std::clamp(y3, static_cast<s32>(m_drawing_area.top), clip_bottom);
const GSVector4i clamp_min = m_clamped_drawing_area; // would be xyxy(), but zw isn't used.
const GSVector4i clamp_max = m_clamped_drawing_area.zwzw();
v1 = v1.sat_i32(clamp_min, clamp_max);
v2 = v2.sat_i32(clamp_min, clamp_max);
v3 = v3.sat_i32(clamp_min, clamp_max);
TickCount pixels = std::abs((x1 * y2 + x2 * y3 + x3 * y1 - x1 * y3 - x2 * y1 - x3 * y2) / 2);
TickCount pixels =
std::abs((v1.x * v2.y + v2.x * v3.y + v3.x * v1.y - v1.x * v3.y - v2.x * v1.y - v3.x * v2.y) / 2);
if (textured)
pixels += pixels;
if (semitransparent || m_GPUSTAT.check_mask_before_draw)
@ -479,6 +479,7 @@ protected:
GPUDrawingArea m_drawing_area = {};
GPUDrawingOffset m_drawing_offset = {};
GSVector4i m_clamped_drawing_area = {};
bool m_console_is_pal = false;
bool m_set_texture_disable_mask = false;
@ -609,7 +610,7 @@ protected:
void SetDisplayTexture(GPUTexture* texture, GPUTexture* depth_texture, s32 view_x, s32 view_y, s32 view_width,
s32 view_height);
bool RenderDisplay(GPUTexture* target, const Common::Rectangle<s32>& draw_rect, bool postfx);
bool RenderDisplay(GPUTexture* target, const GSVector4i draw_rect, bool postfx);
bool Deinterlace(u32 field, u32 line_skip);
bool DeinterlaceExtractField(u32 dst_bufidx, GPUTexture* src, u32 x, u32 y, u32 width, u32 height, u32 line_skip);

2
src/core/gpu_commands.cpp
View File

@ -250,6 +250,7 @@ bool GPU::HandleSetDrawingAreaTopLeftCommand()
m_drawing_area.left = left;
m_drawing_area.top = top;
m_drawing_area_changed = true;
SetClampedDrawingArea();
}
AddCommandTicks(1);
@ -271,6 +272,7 @@ bool GPU::HandleSetDrawingAreaBottomRightCommand()
m_drawing_area.right = right;
m_drawing_area.bottom = bottom;
m_drawing_area_changed = true;
SetClampedDrawingArea();
}
AddCommandTicks(1);

636
src/core/gpu_hw.cpp
View File

File diff suppressed because it is too large Load Diff

40
src/core/gpu_hw.h
View File

@ -1,4 +1,4 @@
// SPDX-FileCopyrightText: 2019-2022 Connor McLaughlin <stenzek@gmail.com>
// SPDX-FileCopyrightText: 2019-2024 Connor McLaughlin <stenzek@gmail.com>
// SPDX-License-Identifier: (GPL-3.0 OR CC-BY-NC-ND-4.0)
#pragma once
@ -9,10 +9,10 @@
#include "util/gpu_device.h"
#include "common/dimensional_array.h"
#include "common/gsvector.h"
#include "common/heap_array.h"
#include <sstream>
#include <string>
#include <limits>
#include <tuple>
#include <utility>
#include <vector>
@ -122,8 +122,7 @@ private:
struct BatchUBOData
{
u32 u_texture_window_and[2];
u32 u_texture_window_or[2];
u32 u_texture_window[4]; // and_x, and_y, or_x, or_y
float u_src_alpha_factor;
float u_dst_alpha_factor;
u32 u_interlaced_displayed_field;
@ -137,6 +136,11 @@ private:
u32 num_uniform_buffer_updates;
};
static constexpr GSVector4i VRAM_SIZE_RECT = GSVector4i::cxpr(0, 0, VRAM_WIDTH, VRAM_HEIGHT);
static constexpr GSVector4i INVALID_RECT =
GSVector4i::cxpr(std::numeric_limits<s32>::max(), std::numeric_limits<s32>::max(), std::numeric_limits<s32>::min(),
std::numeric_limits<s32>::min());
/// Returns true if a depth buffer should be created.
bool NeedsDepthBuffer() const;
GPUTexture::Format GetDepthBufferFormat() const;
@ -153,7 +157,6 @@ private:
void PrintSettingsToLog();
void CheckSettings();
void SetClampedDrawingArea();
void UpdateVRAMReadTexture(bool drawn, bool written);
void UpdateDepthBufferFromMaskBit();
void CopyAndClearDepthBuffer();
@ -172,9 +175,9 @@ private:
void SetFullVRAMDirtyRectangle();
void ClearVRAMDirtyRectangle();
void IncludeVRAMDirtyRectangle(Common::Rectangle<u32>& rect, const Common::Rectangle<u32>& new_rect);
void IncludeDrawnDirtyRectangle(s32 min_x, s32 min_y, s32 max_x, s32 max_y);
void CheckForTexPageOverlap(u32 texpage, u32 min_u, u32 min_v, u32 max_u, u32 max_v);
void IncludeVRAMDirtyRectangle(GSVector4i& rect, const GSVector4i new_rect);
void IncludeDrawnDirtyRectangle(const GSVector4i rect);
void CheckForTexPageOverlap(GSVector4i uv_rect);
bool IsFlushed() const;
void EnsureVertexBufferSpace(u32 required_vertices, u32 required_indices);
@ -205,7 +208,7 @@ private:
void OnBufferSwapped() override;
void UpdateVRAMOnGPU(u32 x, u32 y, u32 width, u32 height, const void* data, u32 data_pitch, bool set_mask,
bool check_mask, const Common::Rectangle<u32>& bounds);
bool check_mask, const GSVector4i bounds);
bool BlitVRAMReplacementTexture(const TextureReplacementTexture* tex, u32 dst_x, u32 dst_y, u32 width, u32 height);
/// Expands a line into two triangles.
@ -214,10 +217,10 @@ private:
/// Handles quads with flipped texture coordinate directions.
void HandleFlippedQuadTextureCoordinates(BatchVertex* vertices);
bool IsPossibleSpritePolygon(const BatchVertex* vertices) const;
void ExpandLineTriangles(BatchVertex* vertices, u32 base_vertex);
bool ExpandLineTriangles(BatchVertex* vertices);
/// Computes polygon U/V boundaries.
void ComputePolygonUVLimits(u32 texpage, BatchVertex* vertices, u32 num_vertices);
/// Computes polygon U/V boundaries, and for overlap with the current texture page.
void ComputePolygonUVLimits(BatchVertex* vertices, u32 num_vertices);
/// Sets the depth test flag for PGXP depth buffering.
void SetBatchDepthBuffer(bool enabled);
@ -271,9 +274,10 @@ private:
bool m_compute_uv_range : 1 = false;
bool m_allow_sprite_mode : 1 = false;
bool m_allow_shader_blend : 1 = false;
bool m_depth_was_copied : 1 = false;
bool m_texture_window_active : 1 = false;
u8 m_texpage_dirty = 0;
bool m_depth_was_copied = false;
BatchConfig m_batch;
@ -282,10 +286,10 @@ private:
BatchUBOData m_batch_ubo_data = {};
// Bounding box of VRAM area that the GPU has drawn into.
GPUDrawingArea m_clamped_drawing_area = {};
Common::Rectangle<u32> m_vram_dirty_draw_rect;
Common::Rectangle<u32> m_vram_dirty_write_rect;
Common::Rectangle<u32> m_current_uv_range;
GSVector4i m_vram_dirty_draw_rect = INVALID_RECT;
GSVector4i m_vram_dirty_write_rect = INVALID_RECT;
GSVector4i m_current_uv_range = INVALID_RECT;
GSVector2i m_current_texture_page_offset = {};
std::unique_ptr<GPUPipeline> m_wireframe_pipeline;

49
src/core/gpu_sw.cpp
View File

@ -523,6 +523,7 @@ void GPU_SW::DispatchRenderCommand()
GPUBackendDrawPolygonCommand* cmd = m_backend.NewDrawPolygonCommand(num_vertices);
FillDrawCommand(cmd, rc);
std::array<GSVector4i, 4> positions;
const u32 first_color = rc.color_for_first_vertex;
const bool shaded = rc.shading_enable;
const bool textured = rc.texture_enable;
@ -535,49 +536,55 @@ void GPU_SW::DispatchRenderCommand()
vert->x = m_drawing_offset.x + vp.x;
vert->y = m_drawing_offset.y + vp.y;
vert->texcoord = textured ? Truncate16(FifoPop()) : 0;
positions[i] = GSVector4i::loadl(&vert->x);
}
if (!IsDrawingAreaIsValid())
return;
// Cull polygons which are too large.
const auto [min_x_12, max_x_12] = MinMax(cmd->vertices[1].x, cmd->vertices[2].x);
const auto [min_y_12, max_y_12] = MinMax(cmd->vertices[1].y, cmd->vertices[2].y);
const s32 min_x = std::min(min_x_12, cmd->vertices[0].x);
const s32 max_x = std::max(max_x_12, cmd->vertices[0].x);
const s32 min_y = std::min(min_y_12, cmd->vertices[0].y);
const s32 max_y = std::max(max_y_12, cmd->vertices[0].y);
if ((max_x - min_x) >= MAX_PRIMITIVE_WIDTH || (max_y - min_y) >= MAX_PRIMITIVE_HEIGHT)
const GSVector4i min_pos_12 = positions[1].min_i32(positions[2]);
const GSVector4i max_pos_12 = positions[1].max_i32(positions[2]);
const GSVector4i draw_rect_012 =
min_pos_12.min_i32(positions[0]).upl64(max_pos_12.max_i32(positions[0])).add32(GSVector4i::cxpr(0, 0, 1, 1));
const bool first_tri_culled =
(draw_rect_012.width() > MAX_PRIMITIVE_WIDTH || draw_rect_012.height() > MAX_PRIMITIVE_HEIGHT ||
!m_clamped_drawing_area.rintersects(draw_rect_012));
if (first_tri_culled)
{
DEBUG_LOG("Culling too-large polygon: {},{} {},{} {},{}", cmd->vertices[0].x, cmd->vertices[0].y,
DEBUG_LOG("Culling off-screen/too-large polygon: {},{} {},{} {},{}", cmd->vertices[0].x, cmd->vertices[0].y,
cmd->vertices[1].x, cmd->vertices[1].y, cmd->vertices[2].x, cmd->vertices[2].y);
if (!rc.quad_polygon)
return;
}
else
{
AddDrawTriangleTicks(cmd->vertices[0].x, cmd->vertices[0].y, cmd->vertices[1].x, cmd->vertices[1].y,
cmd->vertices[2].x, cmd->vertices[2].y, rc.shading_enable, rc.texture_enable,
AddDrawTriangleTicks(positions[0], positions[1], positions[2], rc.shading_enable, rc.texture_enable,
rc.transparency_enable);
}
// quads
if (rc.quad_polygon)
{
const s32 min_x_123 = std::min(min_x_12, cmd->vertices[3].x);
const s32 max_x_123 = std::max(max_x_12, cmd->vertices[3].x);
const s32 min_y_123 = std::min(min_y_12, cmd->vertices[3].y);
const s32 max_y_123 = std::max(max_y_12, cmd->vertices[3].y);
const GSVector4i draw_rect_123 =
min_pos_12.min_i32(positions[3]).upl64(max_pos_12.max_i32(positions[3])).add32(GSVector4i::cxpr(0, 0, 1, 1));
// Cull polygons which are too large.
if ((max_x_123 - min_x_123) >= MAX_PRIMITIVE_WIDTH || (max_y_123 - min_y_123) >= MAX_PRIMITIVE_HEIGHT)
const bool second_tri_culled =
(draw_rect_123.width() > MAX_PRIMITIVE_WIDTH || draw_rect_123.height() > MAX_PRIMITIVE_HEIGHT ||
!m_clamped_drawing_area.rintersects(draw_rect_123));
if (second_tri_culled)
{
DEBUG_LOG("Culling too-large polygon (quad second half): {},{} {},{} {},{}", cmd->vertices[2].x,
cmd->vertices[2].y, cmd->vertices[1].x, cmd->vertices[1].y, cmd->vertices[0].x, cmd->vertices[0].y);
if (first_tri_culled)
return;
}
else
{
AddDrawTriangleTicks(cmd->vertices[2].x, cmd->vertices[2].y, cmd->vertices[1].x, cmd->vertices[1].y,
cmd->vertices[3].x, cmd->vertices[3].y, rc.shading_enable, rc.texture_enable,
AddDrawTriangleTicks(positions[2], positions[1], positions[3], rc.shading_enable, rc.texture_enable,
rc.transparency_enable);
}
}
@ -627,12 +634,6 @@ void GPU_SW::DispatchRenderCommand()
const u32 width_and_height = FifoPop();
cmd->width = static_cast<u16>(width_and_height & VRAM_WIDTH_MASK);
cmd->height = static_cast<u16>((width_and_height >> 16) & VRAM_HEIGHT_MASK);
if (cmd->width >= MAX_PRIMITIVE_WIDTH || cmd->height >= MAX_PRIMITIVE_HEIGHT)
{
DEBUG_LOG("Culling too-large rectangle: {},{} {}x{}", cmd->x, cmd->y, cmd->width, cmd->height);
return;
}
}
break;
}

21
src/core/gpu_types.h
View File

@ -4,7 +4,7 @@
#pragma once
#include "common/bitfield.h"
#include "common/bitutils.h"
#include "common/rectangle.h"
#include "common/gsvector.h"
#include "types.h"
#include <array>
@ -189,18 +189,19 @@ union GPUDrawModeReg
BitField<u16, bool, 12, 1> texture_x_flip;
BitField<u16, bool, 13, 1> texture_y_flip;
ALWAYS_INLINE u16 GetTexturePageBaseX() const { return ZeroExtend16(texture_page_x_base.GetValue()) * 64; }
ALWAYS_INLINE u16 GetTexturePageBaseY() const { return ZeroExtend16(texture_page_y_base.GetValue()) * 256; }
ALWAYS_INLINE u32 GetTexturePageBaseX() const { return ZeroExtend32(texture_page_x_base.GetValue()) * 64; }
ALWAYS_INLINE u32 GetTexturePageBaseY() const { return ZeroExtend32(texture_page_y_base.GetValue()) * 256; }
/// Returns true if the texture mode requires a palette.
ALWAYS_INLINE bool IsUsingPalette() const { return (bits & (2 << 7)) == 0; }
/// Returns a rectangle comprising the texture page area.
ALWAYS_INLINE_RELEASE Common::Rectangle<u32> GetTexturePageRectangle() const
ALWAYS_INLINE_RELEASE GSVector4i GetTexturePageRectangle() const
{
return Common::Rectangle<u32>::FromExtents(GetTexturePageBaseX(), GetTexturePageBaseY(),
texture_page_widths[static_cast<u8>(texture_mode.GetValue())],
TEXTURE_PAGE_HEIGHT);
const u32 base_x = GetTexturePageBaseX();
const u32 base_y = GetTexturePageBaseY();
return GSVector4i(base_x, base_y, base_x + texture_page_widths[static_cast<u8>(texture_mode.GetValue())],
base_y + TEXTURE_PAGE_HEIGHT);
}
};
@ -217,10 +218,12 @@ union GPUTexturePaletteReg
ALWAYS_INLINE u32 GetYBase() const { return static_cast<u32>(y); }
/// Returns a rectangle comprising the texture palette area.
ALWAYS_INLINE_RELEASE Common::Rectangle<u32> GetRectangle(GPUTextureMode mode) const
ALWAYS_INLINE_RELEASE GSVector4i GetRectangle(GPUTextureMode mode) const
{
static constexpr std::array<u32, 4> palette_widths = {{16, 256, 0, 0}};
return Common::Rectangle<u32>::FromExtents(GetXBase(), GetYBase(), palette_widths[static_cast<u8>(mode)], 1);
const u32 base_x = GetXBase();
const u32 base_y = GetYBase();
return GSVector4i(base_x, base_y, base_x + palette_widths[static_cast<u8>(mode)], base_y + 1);
}
};

4
src/core/system.cpp
View File

@ -2727,8 +2727,8 @@ bool System::SaveStateToStream(ByteStream* state, Error* error, u32 screenshot_s
u32 screenshot_stride;
GPUTexture::Format screenshot_format;
if (g_gpu->RenderScreenshotToBuffer(screenshot_width, screenshot_height,
Common::Rectangle<s32>::FromExtents(0, 0, screenshot_width, screenshot_height),
false, &screenshot_buffer, &screenshot_stride, &screenshot_format) &&
GSVector4i(0, 0, screenshot_width, screenshot_height), false,
&screenshot_buffer, &screenshot_stride, &screenshot_format) &&
GPUTexture::ConvertTextureDataToRGBA8(screenshot_width, screenshot_height, screenshot_buffer, screenshot_stride,
screenshot_format))
{

14
src/util/d3d11_device.cpp
View File

@ -14,7 +14,6 @@
#include "common/file_system.h"
#include "common/log.h"
#include "common/path.h"
#include "common/rectangle.h"
#include "common/string_util.h"
#include "fmt/format.h"
@ -1086,17 +1085,18 @@ void D3D11Device::UnbindTexture(D3D11Texture* tex)
}
}
void D3D11Device::SetViewport(s32 x, s32 y, s32 width, s32 height)
void D3D11Device::SetViewport(const GSVector4i rc)
{
const CD3D11_VIEWPORT vp(static_cast<float>(x), static_cast<float>(y), static_cast<float>(width),
static_cast<float>(height), 0.0f, 1.0f);
const CD3D11_VIEWPORT vp(static_cast<float>(rc.left), static_cast<float>(rc.top), static_cast<float>(rc.width()),
static_cast<float>(rc.height()), 0.0f, 1.0f);
m_context->RSSetViewports(1, &vp);
}
void D3D11Device::SetScissor(s32 x, s32 y, s32 width, s32 height)
void D3D11Device::SetScissor(const GSVector4i rc)
{
const CD3D11_RECT rc(x, y, x + width, y + height);
m_context->RSSetScissorRects(1, &rc);
alignas(16) D3D11_RECT drc;
GSVector4i::store<true>(&drc, rc);
m_context->RSSetScissorRects(1, &drc);
}
void D3D11Device::Draw(u32 vertex_count, u32 base_vertex)

4
src/util/d3d11_device.h
View File

@ -92,8 +92,8 @@ public:
void SetPipeline(GPUPipeline* pipeline) override;
void SetTextureSampler(u32 slot, GPUTexture* texture, GPUSampler* sampler) override;
void SetTextureBuffer(u32 slot, GPUTextureBuffer* buffer) override;
void SetViewport(s32 x, s32 y, s32 width, s32 height) override;
void SetScissor(s32 x, s32 y, s32 width, s32 height) override;
void SetViewport(const GSVector4i rc) override;
void SetScissor(const GSVector4i rc) override;
void Draw(u32 vertex_count, u32 base_vertex) override;
void DrawIndexed(u32 index_count, u32 base_index, u32 base_vertex) override;
void DrawIndexedWithBarrier(u32 index_count, u32 base_index, u32 base_vertex, DrawBarrier type) override;

19
src/util/d3d12_device.cpp
View File

@ -1927,8 +1927,8 @@ void D3D12Device::SetViewport(ID3D12GraphicsCommandList4* cmdlist)
{
const D3D12_VIEWPORT vp = {static_cast<float>(m_current_viewport.left),
static_cast<float>(m_current_viewport.top),
static_cast<float>(m_current_viewport.GetWidth()),
static_cast<float>(m_current_viewport.GetHeight()),
static_cast<float>(m_current_viewport.width()),
static_cast<float>(m_current_viewport.height()),
0.0f,
1.0f};
cmdlist->RSSetViewports(1, &vp);
@ -1936,9 +1936,8 @@ void D3D12Device::SetViewport(ID3D12GraphicsCommandList4* cmdlist)
void D3D12Device::SetScissor(ID3D12GraphicsCommandList4* cmdlist)
{
const D3D12_RECT rc = {static_cast<LONG>(m_current_scissor.left), static_cast<LONG>(m_current_scissor.top),
static_cast<LONG>(m_current_scissor.right), static_cast<LONG>(m_current_scissor.bottom)};
cmdlist->RSSetScissorRects(1, &rc);
static_assert(sizeof(GSVector4i) == sizeof(D3D12_RECT));
cmdlist->RSSetScissorRects(1, reinterpret_cast<const D3D12_RECT*>(&m_current_scissor));
}
void D3D12Device::SetTextureSampler(u32 slot, GPUTexture* texture, GPUSampler* sampler)
@ -2027,10 +2026,9 @@ void D3D12Device::UnbindTextureBuffer(D3D12TextureBuffer* buf)
m_dirty_flags |= DIRTY_FLAG_TEXTURES;
}
void D3D12Device::SetViewport(s32 x, s32 y, s32 width, s32 height)
void D3D12Device::SetViewport(const GSVector4i rc)
{
const Common::Rectangle<s32> rc = Common::Rectangle<s32>::FromExtents(x, y, width, height);
if (m_current_viewport == rc)
if (m_current_viewport.eq(rc))
return;
m_current_viewport = rc;
@ -2041,10 +2039,9 @@ void D3D12Device::SetViewport(s32 x, s32 y, s32 width, s32 height)
SetViewport(GetCommandList());
}
void D3D12Device::SetScissor(s32 x, s32 y, s32 width, s32 height)
void D3D12Device::SetScissor(const GSVector4i rc)
{
const Common::Rectangle<s32> rc = Common::Rectangle<s32>::FromExtents(x, y, width, height);
if (m_current_scissor == rc)
if (m_current_scissor.eq(rc))
return;
m_current_scissor = rc;

8
src/util/d3d12_device.h
View File

@ -114,8 +114,8 @@ public:
void SetPipeline(GPUPipeline* pipeline) override;
void SetTextureSampler(u32 slot, GPUTexture* texture, GPUSampler* sampler) override;
void SetTextureBuffer(u32 slot, GPUTextureBuffer* buffer) override;
void SetViewport(s32 x, s32 y, s32 width, s32 height) override;
void SetScissor(s32 x, s32 y, s32 width, s32 height) override;
void SetViewport(const GSVector4i rc) override;
void SetScissor(const GSVector4i rc) override;
void Draw(u32 vertex_count, u32 base_vertex) override;
void DrawIndexed(u32 index_count, u32 base_index, u32 base_vertex) override;
void DrawIndexedWithBarrier(u32 index_count, u32 base_index, u32 base_vertex, DrawBarrier type) override;
@ -344,6 +344,6 @@ private:
std::array<D3D12Texture*, MAX_TEXTURE_SAMPLERS> m_current_textures = {};
std::array<D3D12DescriptorHandle, MAX_TEXTURE_SAMPLERS> m_current_samplers = {};
D3D12TextureBuffer* m_current_texture_buffer = nullptr;
Common::Rectangle<s32> m_current_viewport{0, 0, 1, 1};
Common::Rectangle<s32> m_current_scissor{0, 0, 1, 1};
GSVector4i m_current_viewport = GSVector4i::cxpr(0, 0, 1, 1);
GSVector4i m_current_scissor = {};
};

11
src/util/d3d_common.cpp
View File

@ -7,8 +7,8 @@
#include "common/assert.h"
#include "common/error.h"
#include "common/file_system.h"
#include "common/gsvector.h"
#include "common/log.h"
#include "common/rectangle.h"
#include "common/string_util.h"
#include "fmt/format.h"
@ -179,7 +179,7 @@ bool D3DCommon::GetRequestedExclusiveFullscreenModeDesc(IDXGIFactory5* factory,
DXGI_MODE_DESC* fullscreen_mode, IDXGIOutput** output)
{
// We need to find which monitor the window is located on.
const Common::Rectangle<s32> client_rc_vec(window_rect.left, window_rect.top, window_rect.right, window_rect.bottom);
const GSVector4i client_rc_vec(window_rect.left, window_rect.top, window_rect.right, window_rect.bottom);
// The window might be on a different adapter to which we are rendering.. so we have to enumerate them all.
HRESULT hr;
@ -204,10 +204,9 @@ bool D3DCommon::GetRequestedExclusiveFullscreenModeDesc(IDXGIFactory5* factory,
else if (FAILED(hr) || FAILED(this_output->GetDesc(&output_desc)))
continue;
const Common::Rectangle<s32> output_rc(output_desc.DesktopCoordinates.left, output_desc.DesktopCoordinates.top,
output_desc.DesktopCoordinates.right,
output_desc.DesktopCoordinates.bottom);
if (!client_rc_vec.Intersects(output_rc))
const GSVector4i output_rc(output_desc.DesktopCoordinates.left, output_desc.DesktopCoordinates.top,
output_desc.DesktopCoordinates.right, output_desc.DesktopCoordinates.bottom);
if (!client_rc_vec.rintersects(output_rc))
{
intersecting_output = std::move(this_output);
break;

27
src/util/gpu_device.cpp
View File

@ -643,10 +643,25 @@ void GPUDevice::SetRenderTarget(GPUTexture* rt, GPUTexture* ds, GPUPipeline::Ren
SetRenderTargets(rt ? &rt : nullptr, rt ? 1 : 0, ds, render_pass_flags);
}
void GPUDevice::SetViewport(s32 x, s32 y, s32 width, s32 height)
{
SetViewport(GSVector4i(x, y, x + width, y + height));
}
void GPUDevice::SetScissor(s32 x, s32 y, s32 width, s32 height)
{
SetScissor(GSVector4i(x, y, x + width, y + height));
}
void GPUDevice::SetViewportAndScissor(s32 x, s32 y, s32 width, s32 height)
{
SetViewport(x, y, width, height);
SetScissor(x, y, width, height);
SetViewportAndScissor(GSVector4i(x, y, x + width, y + height));
}
void GPUDevice::SetViewportAndScissor(const GSVector4i rc)
{
SetViewport(rc);
SetScissor(rc);
}
void GPUDevice::ClearRenderTarget(GPUTexture* t, u32 c)
@ -818,11 +833,13 @@ bool GPUDevice::UsesLowerLeftOrigin() const
return (api == RenderAPI::OpenGL || api == RenderAPI::OpenGLES);
}
Common::Rectangle<s32> GPUDevice::FlipToLowerLeft(const Common::Rectangle<s32>& rc, s32 target_height)
GSVector4i GPUDevice::FlipToLowerLeft(GSVector4i rc, s32 target_height)
{
const s32 height = rc.GetHeight();
const s32 height = rc.height();
const s32 flipped_y = target_height - rc.top - height;
return Common::Rectangle<s32>(rc.left, flipped_y, rc.right, flipped_y + height);
rc.top = flipped_y;
rc.bottom = flipped_y + height;
return rc;
}
bool GPUDevice::IsTexturePoolType(GPUTexture::Type type)

11
src/util/gpu_device.h
View File

@ -8,8 +8,8 @@
#include "window_info.h"
#include "common/bitfield.h"
#include "common/gsvector.h"
#include "common/heap_array.h"
#include "common/rectangle.h"
#include "common/small_string.h"
#include "common/types.h"
@ -676,11 +676,14 @@ public:
virtual void SetPipeline(GPUPipeline* pipeline) = 0;
virtual void SetTextureSampler(u32 slot, GPUTexture* texture, GPUSampler* sampler) = 0;
virtual void SetTextureBuffer(u32 slot, GPUTextureBuffer* buffer) = 0;
virtual void SetViewport(s32 x, s32 y, s32 width, s32 height) = 0; // TODO: Rectangle
virtual void SetScissor(s32 x, s32 y, s32 width, s32 height) = 0;
virtual void SetViewport(const GSVector4i rc) = 0;
virtual void SetScissor(const GSVector4i rc) = 0;
void SetRenderTarget(GPUTexture* rt, GPUTexture* ds = nullptr,
GPUPipeline::RenderPassFlag render_pass_flags = GPUPipeline::NoRenderPassFlags);
void SetViewport(s32 x, s32 y, s32 width, s32 height);
void SetScissor(s32 x, s32 y, s32 width, s32 height);
void SetViewportAndScissor(s32 x, s32 y, s32 width, s32 height);
void SetViewportAndScissor(const GSVector4i rc);
// Drawing abstraction.
virtual void Draw(u32 vertex_count, u32 base_vertex) = 0;
@ -704,7 +707,7 @@ public:
bool UpdateImGuiFontTexture();
bool UsesLowerLeftOrigin() const;
static Common::Rectangle<s32> FlipToLowerLeft(const Common::Rectangle<s32>& rc, s32 target_height);
static GSVector4i FlipToLowerLeft(GSVector4i rc, s32 target_height);
bool ResizeTexture(std::unique_ptr<GPUTexture>* tex, u32 new_width, u32 new_height, GPUTexture::Type type,
GPUTexture::Format format, bool preserve = true);
bool ShouldSkipPresentingFrame();

7
src/util/gpu_texture.h
View File

@ -1,8 +1,9 @@
// SPDX-FileCopyrightText: 2019-2023 Connor McLaughlin <stenzek@gmail.com>
// SPDX-FileCopyrightText: 2019-2024 Connor McLaughlin <stenzek@gmail.com>
// SPDX-License-Identifier: (GPL-3.0 OR CC-BY-NC-ND-4.0)
#pragma once
#include "common/gsvector.h"
#include "common/types.h"
#include <algorithm>
@ -103,6 +104,10 @@ public:
ALWAYS_INLINE u32 GetSamples() const { return m_samples; }
ALWAYS_INLINE Type GetType() const { return m_type; }
ALWAYS_INLINE Format GetFormat() const { return m_format; }
ALWAYS_INLINE GSVector4i GetRect() const
{
return GSVector4i(0, 0, static_cast<s32>(m_width), static_cast<s32>(m_height));
}
ALWAYS_INLINE bool IsTextureArray() const { return m_layers > 1; }
ALWAYS_INLINE bool IsMultisampled() const { return m_samples > 1; }

9
src/util/metal_device.h
View File

@ -20,7 +20,6 @@
#include "metal_stream_buffer.h"
#include "window_info.h"
#include "common/rectangle.h"
#include "common/timer.h"
#include <atomic>
@ -255,8 +254,8 @@ public:
void SetPipeline(GPUPipeline* pipeline) override;
void SetTextureSampler(u32 slot, GPUTexture* texture, GPUSampler* sampler) override;
void SetTextureBuffer(u32 slot, GPUTextureBuffer* buffer) override;
void SetViewport(s32 x, s32 y, s32 width, s32 height) override;
void SetScissor(s32 x, s32 y, s32 width, s32 height) override;
void SetViewport(const GSVector4i rc) override;
void SetScissor(const GSVector4i rc) override;
void Draw(u32 vertex_count, u32 base_vertex) override;
void DrawIndexed(u32 index_count, u32 base_index, u32 base_vertex) override;
void DrawIndexedWithBarrier(u32 index_count, u32 base_index, u32 base_vertex, DrawBarrier type) override;
@ -404,8 +403,8 @@ private:
std::array<id<MTLTexture>, MAX_TEXTURE_SAMPLERS> m_current_textures = {};
std::array<id<MTLSamplerState>, MAX_TEXTURE_SAMPLERS> m_current_samplers = {};
id<MTLBuffer> m_current_ssbo = nil;
Common::Rectangle<s32> m_current_viewport = {};
Common::Rectangle<s32> m_current_scissor = {};
GSVector4i m_current_viewport = {};
GSVector4i m_current_scissor = {};
bool m_vsync_enabled = false;

18
src/util/opengl_device.cpp
View File

@ -766,14 +766,12 @@ bool OpenGLDevice::BeginPresent(bool skip_present)
m_last_blend_state.write_a);
glEnable(GL_SCISSOR_TEST);
const Common::Rectangle<s32> window_rc =
Common::Rectangle<s32>::FromExtents(0, 0, m_window_info.surface_width, m_window_info.surface_height);
m_current_fbo = 0;
m_num_current_render_targets = 0;
std::memset(m_current_render_targets.data(), 0, sizeof(m_current_render_targets));
m_current_depth_target = nullptr;
const GSVector4i window_rc = GSVector4i(0, 0, m_window_info.surface_width, m_window_info.surface_height);
m_last_viewport = window_rc;
m_last_scissor = window_rc;
UpdateViewport();
@ -1220,20 +1218,18 @@ void OpenGLDevice::SetTextureBuffer(u32 slot, GPUTextureBuffer* buffer)
}
}
void OpenGLDevice::SetViewport(s32 x, s32 y, s32 width, s32 height)
void OpenGLDevice::SetViewport(const GSVector4i rc)
{
const Common::Rectangle<s32> rc = Common::Rectangle<s32>::FromExtents(x, y, width, height);
if (m_last_viewport == rc)
if (m_last_viewport.eq(rc))
return;
m_last_viewport = rc;
UpdateViewport();
}
void OpenGLDevice::SetScissor(s32 x, s32 y, s32 width, s32 height)
void OpenGLDevice::SetScissor(const GSVector4i rc)
{
const Common::Rectangle<s32> rc = Common::Rectangle<s32>::FromExtents(x, y, width, height);
if (m_last_scissor == rc)
if (m_last_scissor.eq(rc))
return;
m_last_scissor = rc;
@ -1242,10 +1238,10 @@ void OpenGLDevice::SetScissor(s32 x, s32 y, s32 width, s32 height)
void OpenGLDevice::UpdateViewport()
{
glViewport(m_last_viewport.left, m_last_viewport.top, m_last_viewport.GetWidth(), m_last_viewport.GetHeight());
glViewport(m_last_viewport.left, m_last_viewport.top, m_last_viewport.width(), m_last_viewport.height());
}
void OpenGLDevice::UpdateScissor()
{
glScissor(m_last_scissor.left, m_last_scissor.top, m_last_scissor.GetWidth(), m_last_scissor.GetHeight());
glScissor(m_last_scissor.left, m_last_scissor.top, m_last_scissor.width(), m_last_scissor.height());
}

10
src/util/opengl_device.h
View File

@ -11,8 +11,6 @@
#include "opengl_pipeline.h"
#include "opengl_texture.h"
#include "common/rectangle.h"
#include <cstdio>
#include <memory>
#include <tuple>
@ -96,8 +94,8 @@ public:
void SetPipeline(GPUPipeline* pipeline) override;
void SetTextureSampler(u32 slot, GPUTexture* texture, GPUSampler* sampler) override;
void SetTextureBuffer(u32 slot, GPUTextureBuffer* buffer) override;
void SetViewport(s32 x, s32 y, s32 width, s32 height) override;
void SetScissor(s32 x, s32 y, s32 width, s32 height) override;
void SetViewport(const GSVector4i rc) override;
void SetScissor(const GSVector4i rc) override;
void Draw(u32 vertex_count, u32 base_vertex) override;
void DrawIndexed(u32 index_count, u32 base_index, u32 base_vertex) override;
void DrawIndexedWithBarrier(u32 index_count, u32 base_index, u32 base_vertex, DrawBarrier type) override;
@ -203,8 +201,8 @@ private:
u32 m_last_texture_unit = 0;
std::array<std::pair<GLuint, GLuint>, MAX_TEXTURE_SAMPLERS> m_last_samplers = {};
GLuint m_last_ssbo = 0;
Common::Rectangle<s32> m_last_viewport{0, 0, 1, 1};
Common::Rectangle<s32> m_last_scissor{0, 0, 1, 1};
GSVector4i m_last_viewport = {};
GSVector4i m_last_scissor = GSVector4i::cxpr(0, 0, 1, 1);
// Misc framebuffers
GLuint m_read_fbo = 0;

9
src/util/postprocessing.cpp
View File

@ -620,8 +620,8 @@ void PostProcessing::Chain::DestroyTextures()
}
bool PostProcessing::Chain::Apply(GPUTexture* input_color, GPUTexture* input_depth, GPUTexture* final_target,
s32 final_left, s32 final_top, s32 final_width, s32 final_height, s32 orig_width,
s32 orig_height, s32 native_width, s32 native_height)
GSVector4i final_rect, s32 orig_width, s32 orig_height, s32 native_width,
s32 native_height)
{
GL_SCOPE_FMT("{} Apply", m_section);
@ -634,9 +634,8 @@ bool PostProcessing::Chain::Apply(GPUTexture* input_color, GPUTexture* input_dep
{
const bool is_final = (stage.get() == m_stages.back().get());
if (!stage->Apply(input_color, input_depth, is_final ? final_target : output, final_left, final_top, final_width,
final_height, orig_width, orig_height, native_width, native_height, m_target_width,
m_target_height))
if (!stage->Apply(input_color, input_depth, is_final ? final_target : output, final_rect, orig_width, orig_height,
native_width, native_height, m_target_width, m_target_height))
{
return false;
}

4
src/util/postprocessing.h
View File

@ -134,8 +134,8 @@ public:
bool CheckTargets(GPUTexture::Format target_format, u32 target_width, u32 target_height,
ProgressCallback* progress = nullptr);
bool Apply(GPUTexture* input_color, GPUTexture* input_depth, GPUTexture* final_target, s32 final_left, s32 final_top,
s32 final_width, s32 final_height, s32 orig_width, s32 orig_height, s32 native_width, s32 native_height);
bool Apply(GPUTexture* input_color, GPUTexture* input_depth, GPUTexture* final_target, const GSVector4i final_rect,
s32 orig_width, s32 orig_height, s32 native_width, s32 native_height);
private:
void ClearStagesWithError(const Error& error);

8
src/util/postprocessing_shader.h
View File

@ -7,7 +7,7 @@
#include "gpu_texture.h"
#include "common/rectangle.h"
#include "common/gsvector.h"
#include "common/settings_interface.h"
#include "common/timer.h"
#include "common/types.h"
@ -49,9 +49,9 @@ public:
virtual bool CompilePipeline(GPUTexture::Format format, u32 width, u32 height, ProgressCallback* progress) = 0;
virtual bool Apply(GPUTexture* input_color, GPUTexture* input_depth, GPUTexture* final_target, s32 final_left,
s32 final_top, s32 final_width, s32 final_height, s32 orig_width, s32 orig_height,
s32 native_width, s32 native_height, u32 target_width, u32 target_height) = 0;
virtual bool Apply(GPUTexture* input_color, GPUTexture* input_depth, GPUTexture* final_target, GSVector4i final_rect,
s32 orig_width, s32 orig_height, s32 native_width, s32 native_height, u32 target_width,
u32 target_height) = 0;
protected:
static void ParseKeyValue(std::string_view line, std::string_view* key, std::string_view* value);

46
src/util/postprocessing_shader_fx.cpp
View File

@ -1490,16 +1490,15 @@ bool PostProcessing::ReShadeFXShader::ResizeOutput(GPUTexture::Format format, u3
}
bool PostProcessing::ReShadeFXShader::Apply(GPUTexture* input_color, GPUTexture* input_depth, GPUTexture* final_target,
s32 final_left, s32 final_top, s32 final_width, s32 final_height,
s32 orig_width, s32 orig_height, s32 native_width, s32 native_height,
u32 target_width, u32 target_height)
GSVector4i final_rect, s32 orig_width, s32 orig_height, s32 native_width,
s32 native_height, u32 target_width, u32 target_height)
{
GL_PUSH_FMT("PostProcessingShaderFX {}", m_name);
m_frame_count++;
// Reshade always draws at full size.
g_gpu_device->SetViewportAndScissor(0, 0, target_width, target_height);
g_gpu_device->SetViewportAndScissor(final_rect);
if (m_uniforms_size > 0)
{
@ -1675,84 +1674,85 @@ bool PostProcessing::ReShadeFXShader::Apply(GPUTexture* input_color, GPUTexture*
case SourceOptionType::ViewportX:
{
const float value = static_cast<float>(final_left);
const float value = static_cast<float>(final_rect.left);
std::memcpy(dst, &value, sizeof(value));
}
break;
case SourceOptionType::ViewportY:
{
const float value = static_cast<float>(final_top);
const float value = static_cast<float>(final_rect.top);
std::memcpy(dst, &value, sizeof(value));
}
break;
case SourceOptionType::ViewportWidth:
{
const float value = static_cast<float>(final_width);
const float value = static_cast<float>(final_rect.width());
std::memcpy(dst, &value, sizeof(value));
}
break;
case SourceOptionType::ViewportHeight:
{
const float value = static_cast<float>(final_height);
const float value = static_cast<float>(final_rect.height());
std::memcpy(dst, &value, sizeof(value));
}
break;
case SourceOptionType::ViewportOffset:
{
const float value[2] = {static_cast<float>(final_left), static_cast<float>(final_top)};
std::memcpy(dst, &value, sizeof(value));
GSVector4::storel(dst, GSVector4(final_rect));
}
break;
case SourceOptionType::ViewportSize:
{
const float value[2] = {static_cast<float>(final_width), static_cast<float>(final_height)};
const float value[2] = {static_cast<float>(final_rect.width()), static_cast<float>(final_rect.height())};
std::memcpy(dst, &value, sizeof(value));
}
break;
case SourceOptionType::InternalPixelSize:
{
const float value[2] = {static_cast<float>(final_width) / static_cast<float>(orig_width),
static_cast<float>(final_height) / static_cast<float>(orig_height)};
const float value[2] = {static_cast<float>(final_rect.width()) / static_cast<float>(orig_width),
static_cast<float>(final_rect.height()) / static_cast<float>(orig_height)};
std::memcpy(dst, value, sizeof(value));
}
break;
case SourceOptionType::InternalNormPixelSize:
{
const float value[2] = {
(static_cast<float>(final_width) / static_cast<float>(orig_width)) / static_cast<float>(target_width),
(static_cast<float>(final_height) / static_cast<float>(orig_height)) / static_cast<float>(target_height)};
const float value[2] = {(static_cast<float>(final_rect.width()) / static_cast<float>(orig_width)) /
static_cast<float>(target_width),
(static_cast<float>(final_rect.height()) / static_cast<float>(orig_height)) /
static_cast<float>(target_height)};
std::memcpy(dst, value, sizeof(value));
}
break;
case SourceOptionType::NativePixelSize:
{
const float value[2] = {static_cast<float>(final_width) / static_cast<float>(native_width),
static_cast<float>(final_height) / static_cast<float>(native_height)};
const float value[2] = {static_cast<float>(final_rect.width()) / static_cast<float>(native_width),
static_cast<float>(final_rect.height()) / static_cast<float>(native_height)};
std::memcpy(dst, value, sizeof(value));
}
break;
case SourceOptionType::NativeNormPixelSize:
{
const float value[2] = {
(static_cast<float>(final_width) / static_cast<float>(native_width)) / static_cast<float>(target_width),
(static_cast<float>(final_height) / static_cast<float>(native_height)) / static_cast<float>(target_height)};
const float value[2] = {(static_cast<float>(final_rect.width()) / static_cast<float>(native_width)) /
static_cast<float>(target_width),
(static_cast<float>(final_rect.height()) / static_cast<float>(native_height)) /
static_cast<float>(target_height)};
std::memcpy(dst, value, sizeof(value));
}
break;
case SourceOptionType::BufferToViewportRatio:
{
const float value[2] = {static_cast<float>(target_width) / static_cast<float>(final_width),
static_cast<float>(target_height) / static_cast<float>(final_height)};
const float value[2] = {static_cast<float>(target_width) / static_cast<float>(final_rect.width()),
static_cast<float>(target_height) / static_cast<float>(final_rect.height())};
std::memcpy(dst, value, sizeof(value));
}
break;

6
src/util/postprocessing_shader_fx.h
View File

@ -31,9 +31,9 @@ public:
bool ResizeOutput(GPUTexture::Format format, u32 width, u32 height) override;
bool CompilePipeline(GPUTexture::Format format, u32 width, u32 height, ProgressCallback* progress) override;
bool Apply(GPUTexture* input_color, GPUTexture* input_depth, GPUTexture* final_target, s32 final_left, s32 final_top,
s32 final_width, s32 final_height, s32 orig_width, s32 orig_height, s32 native_width, s32 native_height,
u32 target_width, u32 target_height) override;
bool Apply(GPUTexture* input_color, GPUTexture* input_depth, GPUTexture* final_target, GSVector4i final_rect,
s32 orig_width, s32 orig_height, s32 native_width, s32 native_height, u32 target_width,
u32 target_height) override;
private:
using TextureID = s32;

11
src/util/postprocessing_shader_glsl.cpp
View File

@ -168,9 +168,8 @@ bool PostProcessing::GLSLShader::CompilePipeline(GPUTexture::Format format, u32
}
bool PostProcessing::GLSLShader::Apply(GPUTexture* input_color, GPUTexture* input_depth, GPUTexture* final_target,
s32 final_left, s32 final_top, s32 final_width, s32 final_height, s32 orig_width,
s32 orig_height, s32 native_width, s32 native_height, u32 target_width,
u32 target_height)
GSVector4i final_rect, s32 orig_width, s32 orig_height, s32 native_width,
s32 native_height, u32 target_width, u32 target_height)
{
GL_SCOPE_FMT("GLSL Shader {}", m_name);
@ -188,12 +187,12 @@ bool PostProcessing::GLSLShader::Apply(GPUTexture* input_color, GPUTexture* inpu
g_gpu_device->SetPipeline(m_pipeline.get());
g_gpu_device->SetTextureSampler(0, input_color, m_sampler.get());
g_gpu_device->SetViewportAndScissor(final_left, final_top, final_width, final_height);
g_gpu_device->SetViewportAndScissor(final_rect);
const u32 uniforms_size = GetUniformsSize();
void* uniforms = g_gpu_device->MapUniformBuffer(uniforms_size);
FillUniformBuffer(uniforms, final_left, final_top, final_width, final_height, target_width, target_height, orig_width,
orig_height, native_width, native_height,
FillUniformBuffer(uniforms, final_rect.left, final_rect.top, final_rect.width(), final_rect.height(), target_width,
target_height, orig_width, orig_height, native_width, native_height,
static_cast<float>(PostProcessing::GetTimer().GetTimeSeconds()));
g_gpu_device->UnmapUniformBuffer(uniforms_size);
g_gpu_device->Draw(3, 0);

6
src/util/postprocessing_shader_glsl.h
View File

@ -24,9 +24,9 @@ public:
bool ResizeOutput(GPUTexture::Format format, u32 width, u32 height) override;
bool CompilePipeline(GPUTexture::Format format, u32 width, u32 height, ProgressCallback* progress) override;
bool Apply(GPUTexture* input_color, GPUTexture* input_depth, GPUTexture* final_target, s32 final_left, s32 final_top,
s32 final_width, s32 final_height, s32 orig_width, s32 orig_height, s32 native_width, s32 native_height,
u32 target_width, u32 target_height) override;
bool Apply(GPUTexture* input_color, GPUTexture* input_depth, GPUTexture* final_target, GSVector4i final_rect,
s32 orig_width, s32 orig_height, s32 native_width, s32 native_height, u32 target_width,
u32 target_height) override;
private:
struct CommonUniforms

29
src/util/vulkan_device.cpp
View File

@ -3611,15 +3611,14 @@ void VulkanDevice::SetInitialPipelineState()
const VkViewport vp = {static_cast<float>(m_current_viewport.left),
static_cast<float>(m_current_viewport.top),
static_cast<float>(m_current_viewport.GetWidth()),
static_cast<float>(m_current_viewport.GetHeight()),
static_cast<float>(m_current_viewport.width()),
static_cast<float>(m_current_viewport.height()),
0.0f,
1.0f};
vkCmdSetViewport(GetCurrentCommandBuffer(), 0, 1, &vp);
const VkRect2D vrc = {
{m_current_scissor.left, m_current_scissor.top},
{static_cast<u32>(m_current_scissor.GetWidth()), static_cast<u32>(m_current_scissor.GetHeight())}};
const VkRect2D vrc = {{m_current_scissor.left, m_current_scissor.top},
{static_cast<u32>(m_current_scissor.width()), static_cast<u32>(m_current_scissor.height())}};
vkCmdSetScissor(GetCurrentCommandBuffer(), 0, 1, &vrc);
}
@ -3706,10 +3705,9 @@ void VulkanDevice::UnbindTextureBuffer(VulkanTextureBuffer* buf)
m_dirty_flags |= DIRTY_FLAG_TEXTURES_OR_SAMPLERS;
}
void VulkanDevice::SetViewport(s32 x, s32 y, s32 width, s32 height)
void VulkanDevice::SetViewport(const GSVector4i rc)
{
const Common::Rectangle<s32> rc = Common::Rectangle<s32>::FromExtents(x, y, width, height);
if (m_current_viewport == rc)
if (m_current_viewport.eq(rc))
return;
m_current_viewport = rc;
@ -3717,15 +3715,18 @@ void VulkanDevice::SetViewport(s32 x, s32 y, s32 width, s32 height)
if (m_dirty_flags & DIRTY_FLAG_INITIAL)
return;
const VkViewport vp = {
static_cast<float>(x), static_cast<float>(y), static_cast<float>(width), static_cast<float>(height), 0.0f, 1.0f};
const VkViewport vp = {static_cast<float>(rc.x),
static_cast<float>(rc.y),
static_cast<float>(rc.width()),
static_cast<float>(rc.height()),
0.0f,
1.0f};
vkCmdSetViewport(GetCurrentCommandBuffer(), 0, 1, &vp);
}
void VulkanDevice::SetScissor(s32 x, s32 y, s32 width, s32 height)
void VulkanDevice::SetScissor(const GSVector4i rc)
{
const Common::Rectangle<s32> rc = Common::Rectangle<s32>::FromExtents(x, y, width, height);
if (m_current_scissor == rc)
if (m_current_scissor.eq(rc))
return;
m_current_scissor = rc;
@ -3733,7 +3734,7 @@ void VulkanDevice::SetScissor(s32 x, s32 y, s32 width, s32 height)
if (m_dirty_flags & DIRTY_FLAG_INITIAL)
return;
const VkRect2D vrc = {{x, y}, {static_cast<u32>(width), static_cast<u32>(height)}};
const VkRect2D vrc = {{rc.x, rc.y}, {static_cast<u32>(rc.width()), static_cast<u32>(rc.height())}};
vkCmdSetScissor(GetCurrentCommandBuffer(), 0, 1, &vrc);
}

8
src/util/vulkan_device.h
View File

@ -122,8 +122,8 @@ public:
void SetPipeline(GPUPipeline* pipeline) override;
void SetTextureSampler(u32 slot, GPUTexture* texture, GPUSampler* sampler) override;
void SetTextureBuffer(u32 slot, GPUTextureBuffer* buffer) override;
void SetViewport(s32 x, s32 y, s32 width, s32 height) override;
void SetScissor(s32 x, s32 y, s32 width, s32 height) override;
void SetViewport(const GSVector4i rc) override;
void SetScissor(const GSVector4i rc) override;
void Draw(u32 vertex_count, u32 base_vertex) override;
void DrawIndexed(u32 index_count, u32 base_index, u32 base_vertex) override;
void DrawIndexedWithBarrier(u32 index_count, u32 base_index, u32 base_vertex, DrawBarrier type) override;
@ -477,6 +477,6 @@ private:
std::array<VulkanTexture*, MAX_TEXTURE_SAMPLERS> m_current_textures = {};
std::array<VkSampler, MAX_TEXTURE_SAMPLERS> m_current_samplers = {};
VulkanTextureBuffer* m_current_texture_buffer = nullptr;
Common::Rectangle<s32> m_current_viewport{0, 0, 1, 1};
Common::Rectangle<s32> m_current_scissor{0, 0, 1, 1};
GSVector4i m_current_viewport = {};
GSVector4i m_current_scissor = GSVector4i::cxpr(0, 0, 1, 1);
};