Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Copy color to depth, defer depth buffer copies #15858

Merged
merged 7 commits into from
Aug 20, 2022
Merged

Copy color to depth, defer depth buffer copies #15858

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
a0ac2df
Defer depth copies until depth buffer is actually used.
hrydgard Aug 1, 2022
886679c
Remove the color-to-depth mode
hrydgard Aug 17, 2022
94ade8c
Extract depth copies to CopyToDepthFromOverlappingFramebuffers
hrydgard Aug 17, 2022
5d8fe4c
Implement copying color to depth on load. However, weird stuff is hap…
hrydgard Aug 17, 2022
97dbba0
Minor fixes. Everything seems fine now.
hrydgard Aug 17, 2022
12db0e5
Fix deferred-depth for bezier/spline. Move updating of last_frame_dep…
hrydgard Aug 18, 2022
29ea3ff
Restore the clearing optimization, avoiding unnecessary depth copies
hrydgard Aug 20, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 4 additions & 6 deletions GPU/Common/Draw2D.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -67,12 +67,8 @@ RasterChannel GenerateDraw2D565ToDepthFs(ShaderWriter &writer) {
// have to apply the scaling.
DepthScaleFactors factors = GetDepthScaleFactors();
writer.C(" vec3 rgb = ").SampleTexture2D("tex", "v_texcoord.xy").C(".xyz;\n");
writer.F(" highp float depthValue = (floor(rgb.x * 31.99) + floor(rgb.y * 63.99) * 32.0 + floor(rgb.z * 31.99) * 2048.0) / 65535.0; \n");
if (factors.scale != 1.0 || factors.offset != 0.0) {
writer.F(" gl_FragDepth = (depthValue / %f) + %f;\n", factors.scale / 65535.0f, factors.offset);
} else {
writer.C(" gl_FragDepth = depthValue;\n");
}
writer.F(" highp float depthValue = (floor(rgb.x * 31.99) + floor(rgb.y * 63.99) * 32.0 + floor(rgb.z * 31.99) * 2048.0); \n");
writer.F(" gl_FragDepth = (depthValue / %f) + %f;\n", factors.scale, factors.offset);
writer.EndFSMain("outColor", FSFLAG_WRITEDEPTH);
return RASTER_DEPTH;
}
Expand Down Expand Up @@ -200,6 +196,7 @@ void FramebufferManagerCommon::DrawStrip2D(Draw::Texture *tex, Draw2DVertex *ver
}
if (!draw2DPipelineDepth_) {
draw2DPipelineDepth_ = Create2DPipeline(&GenerateDraw2DDepthFs);
linearFilter = false;
}
draw_->BindPipeline(draw2DPipelineDepth_);
break;
Expand All @@ -211,6 +208,7 @@ void FramebufferManagerCommon::DrawStrip2D(Draw::Texture *tex, Draw2DVertex *ver
}
if (!draw2DPipeline565ToDepth_) {
draw2DPipeline565ToDepth_ = Create2DPipeline(&GenerateDraw2D565ToDepthFs);
linearFilter = false;
}
draw_->BindPipeline(draw2DPipeline565ToDepth_);
break;
Expand Down
18 changes: 10 additions & 8 deletions GPU/Common/FramebufferManagerCommon.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -405,7 +405,7 @@ VirtualFramebuffer *FramebufferManagerCommon::DoSetRenderFrameBuffer(const Frame
vfb->lastFrameNewSize = gpuStats.numFlips;
vfb->format = params.fmt;
vfb->drawnFormat = params.fmt;
vfb->usageFlags = FB_USAGE_RENDERTARGET;
vfb->usageFlags = FB_USAGE_RENDER_COLOR;

u32 byteSize = ColorBufferByteSize(vfb);
if (Memory::IsVRAMAddress(params.fb_address) && params.fb_address + byteSize > framebufRangeEnd_) {
Expand Down Expand Up @@ -454,7 +454,7 @@ VirtualFramebuffer *FramebufferManagerCommon::DoSetRenderFrameBuffer(const Frame
} else if (vfbs_[i]->z_stride != 0 && params.z_address == vfbs_[i]->z_address && params.fb_address != vfbs_[i]->fb_address && !sharingReported) {
// This happens a lot, but virtually always it's cleared.
// It's possible the other might not clear, but when every game is reported it's not useful.
if (params.isWritingDepth) {
if (params.isWritingDepth && (vfbs_[i]->usageFlags & FB_USAGE_RENDER_DEPTH)) {
WARN_LOG(SCEGE, "FBO reusing depthbuffer, c=%08x/d=%08x and c=%08x/d=%08x", params.fb_address, params.z_address, vfbs_[i]->fb_address, vfbs_[i]->z_address);
sharingReported = true;
}
Expand All @@ -465,7 +465,7 @@ VirtualFramebuffer *FramebufferManagerCommon::DoSetRenderFrameBuffer(const Frame
} else if (vfb != currentRenderVfb_) {
// Use it as a render target.
DEBUG_LOG(FRAMEBUF, "Switching render target to FBO for %08x: %d x %d x %d ", vfb->fb_address, vfb->width, vfb->height, vfb->format);
vfb->usageFlags |= FB_USAGE_RENDERTARGET;
vfb->usageFlags |= FB_USAGE_RENDER_COLOR;
vfb->last_frame_render = gpuStats.numFlips;
frameLastFramebufUsed_ = gpuStats.numFlips;
vfb->dirtyAfterDisplay = true;
Expand Down Expand Up @@ -505,6 +505,7 @@ void FramebufferManagerCommon::SetDepthFrameBuffer() {
// "Resolve" the depth buffer, by copying from any overlapping buffers with fresher content.
CopyToDepthFromOverlappingFramebuffers(currentRenderVfb_);

currentRenderVfb_->usageFlags |= FB_USAGE_RENDER_DEPTH;
currentRenderVfb_->depthBindSeq = GetBindSeqCount();
}

Expand All @@ -523,16 +524,17 @@ void FramebufferManagerCommon::CopyToDepthFromOverlappingFramebuffers(VirtualFra
};

std::vector<CopySource> sources;
for (auto src: vfbs_) {
for (auto src : vfbs_) {
if (src == dest)
continue;

if (src->fb_address == dest->z_address && src->fb_stride == dest->z_stride && src->format == GE_FORMAT_565) {
if (src->colorBindSeq < dest->depthBindSeq) {
if (src->colorBindSeq > dest->depthBindSeq) {
// Source has newer data than the current buffer, use it.
sources.push_back(CopySource{ src, RASTER_COLOR });
}
} else if (src->z_address == dest->z_address && src->z_stride == dest->z_stride && src->depthBindSeq > dest->depthBindSeq) {
// Don't bother if the buffer was from another frame. This heuristic is old.
hrydgard marked this conversation as resolved.
Show resolved Hide resolved
sources.push_back(CopySource{ src, RASTER_DEPTH });
} else {
// TODO: Do more detailed overlap checks here.
Expand Down Expand Up @@ -1572,7 +1574,7 @@ VirtualFramebuffer *FramebufferManagerCommon::CreateRAMFramebuffer(uint32_t fbAd
vfb->bufferHeight = vfb->height;
vfb->format = format;
vfb->drawnFormat = GE_FORMAT_8888;
vfb->usageFlags = FB_USAGE_RENDERTARGET;
vfb->usageFlags = FB_USAGE_RENDER_COLOR;
SetColorUpdated(vfb, 0);
char name[64];
snprintf(name, sizeof(name), "%08x_color_RAM", vfb->fb_address);
Expand Down Expand Up @@ -1640,7 +1642,7 @@ VirtualFramebuffer *FramebufferManagerCommon::FindDownloadTempBuffer(VirtualFram
UpdateDownloadTempBuffer(nvfb);
}

nvfb->usageFlags |= FB_USAGE_RENDERTARGET;
nvfb->usageFlags |= FB_USAGE_RENDER_COLOR;
nvfb->last_frame_render = gpuStats.numFlips;
nvfb->dirtyAfterDisplay = true;

Expand Down Expand Up @@ -1957,7 +1959,7 @@ void FramebufferManagerCommon::UpdateFramebufUsage(VirtualFramebuffer *vfb) {

checkFlag(FB_USAGE_DISPLAYED_FRAMEBUFFER, vfb->last_frame_displayed);
checkFlag(FB_USAGE_TEXTURE, vfb->last_frame_used);
checkFlag(FB_USAGE_RENDERTARGET, vfb->last_frame_render);
checkFlag(FB_USAGE_RENDER_COLOR, vfb->last_frame_render);
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Hm, what if it's only rendering depth? I could see it being useful to know that... hm. Maybe for render-to-texture selection too. Can't remember if I've seen games render a scene with RGB and stencil all masked out, just to establish depth... feel like I have...

-[Unknown]

Copy link
Owner Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yeah, this tracking isn't perfect yet. I don't think I have seen such a case though, but I also don't think it'll hurt very much if we get that wrong..

checkFlag(FB_USAGE_CLUT, vfb->last_frame_clut);
}

Expand Down
3 changes: 2 additions & 1 deletion GPU/Common/FramebufferManagerCommon.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -37,13 +37,14 @@

enum {
FB_USAGE_DISPLAYED_FRAMEBUFFER = 1,
FB_USAGE_RENDERTARGET = 2,
FB_USAGE_RENDER_COLOR = 2,
FB_USAGE_TEXTURE = 4,
FB_USAGE_CLUT = 8,
FB_USAGE_DOWNLOAD = 16,
FB_USAGE_DOWNLOAD_CLEAR = 32,
FB_USAGE_BLUE_TO_ALPHA = 64,
FB_USAGE_FIRST_FRAME_SAVED = 128,
FB_USAGE_RENDER_DEPTH = 256,
};

enum {
Expand Down
8 changes: 1 addition & 7 deletions GPU/Common/TextureCacheCommon.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -106,13 +106,7 @@ inline int dimHeight(u16 dim) {
// Vulkan color formats:
// TODO
TextureCacheCommon::TextureCacheCommon(Draw::DrawContext *draw)
: draw_(draw),
clutLastFormat_(0xFFFFFFFF),
clutTotalBytes_(0),
clutMaxBytes_(0),
clutRenderAddress_(0xFFFFFFFF),
clutAlphaLinear_(false),
isBgraBackend_(false) {
: draw_(draw) {
decimationCounter_ = TEXCACHE_DECIMATION_INTERVAL;

// TODO: Clamp down to 256/1KB? Need to check mipmapShareClut and clamp loadclut.
Expand Down
12 changes: 6 additions & 6 deletions GPU/Common/TextureCacheCommon.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -444,13 +444,13 @@ class TextureCacheCommon {
u32 *clutBufConverted_;
// This is the active one.
u32 *clutBuf_;
u32 clutLastFormat_;
u32 clutTotalBytes_;
u32 clutMaxBytes_;
u32 clutRenderAddress_;
u32 clutLastFormat_ = 0xFFFFFFFF;
u32 clutTotalBytes_ = 0;
u32 clutMaxBytes_ = 0;
u32 clutRenderAddress_ = 0xFFFFFFFF;
u32 clutRenderOffset_;
// True if the clut is just alpha values in the same order (RGBA4444-bit only.)
bool clutAlphaLinear_;
bool clutAlphaLinear_ = false;
u16 clutAlphaLinearColor_;

int standardScaleFactor_;
Expand All @@ -461,7 +461,7 @@ class TextureCacheCommon {
bool nextNeedsChange_;
bool nextNeedsRebuild_;

bool isBgraBackend_;
bool isBgraBackend_ = false;

u32 expandClut_[256];
};
Expand Down