Implement MSAA support for desktop GPUs in Vulkan #16458

hrydgard · 2022-11-28T19:49:16Z

Partially solves #6218

This is an implementation of MSAA antialiasing, for desktop GPUs only, and Vulkan only for now.

There are some glitches ~~(like the sun in Burnout Legends)~~ but it mostly works in a lot of games.

It detects available MSAA modes and even allows runtime switching between them.

(In theory this could work on mobile but would be terribly expensive on tiler GPUs because it "stores back" the MSAA framebuffers to memory, so that it can continue rendering to the same render target when needed. On mobile, we should use STORE_OP = DONT_CARE, but then we also need to do a blit from the resolved buffer to the MSAA framebuffer in order to continue rendering, but that's work for another day.)

Remaining work before merge:

Correctly handle blits and copies (copies might need conversion to blits), and any related image layout transitions
- To be safe, we might have to copy to both the msaa and the normal image of the destination framebuffer... ugh.
- Now doing correct depth copies ~~Not fully correct yet, but switched to raster copies, which do work, except we lose some sampling resolution.~~

Postponed quality setting until after merge.

hrydgard · 2022-11-28T22:26:12Z

Heh, getting the basics up was not so hard, but there's a lot of image layout transition logic in QueueRunner that's kinda bogus or incomplete in multisample mode. All that stuff only works because we don't do copies correctly so never transition away the msaa targets from ATTACHMENT_*_OPTIMAL :)

Doing all copies through raster might be the easy way out... Though, won't be perfect for depth buffers where we'll lose valuable resolution (by texturing from a resolved target, drawing to a multisampled one)

hrydgard · 2022-11-28T23:16:54Z

Raster copies fixes all the common cases. We do lose depth resolution which might matter in cases like Jeanne D'Arc, but mostly it works just fine.

I actually think we could get this in pretty soon. Should probably make a configuration option to select how much sample shading we want - right now we enable full per-sample shading only when DISCARD is used in a shader, which smoothes out fences and stuff very nicely.

hrydgard · 2022-11-28T23:30:59Z

Hm, there seems to be some shutdown race condition related to shader destruction, not sure why it appeared here. Doesn't always happen.

hrydgard · 2022-11-29T14:45:21Z

There's a related race between the pipeline compiler thread and forced recompilation of pipelines with the wrong msaa depth.

Think I'll have to purge the compiled pipelines completely on MSAA level switch somehow to get around this..

…s to support.

…esolve.

… since we don't texture from stencil, but whatever).

…ty setting.

… mode.

…on't use it while not multisampling.

…rt for sample rate shading.

hrydgard · 2022-12-01T22:49:28Z

This seems fairly solid-ish now, and runs validation-clean throughout, switching on and off, and changing MSAA level - although there are things I want to improve (add quality setting, make actual copies work instead of raster copies), everything I throw at it seems to work, and it generally looks absolutely fantastic, at least on GPUs that support sampleRateShading. So even though it's late in the process, would be cool to get in for 1.14.

My reluctance to put multiple MSAA levels into the renderpasstype key resulted in the need for some rather ugly invalidation hacks, but I think it's worth it. There's only one additional renderpass key bit I foresee us ever needing now (#16327), while adding 3 additional more bits to represent the various MSAA levels would start to get to the point where we should switch to hashmaps or something instead of plain arrays, and it's nice to avoid that... I think.

EDIT: Actually I've found two problems:

~~Cars Race-o-Rama doesn't work, not entirely surprising~~
~~Ratchet & Clank's depth trickery is partially broken~~ (never mind, seems unrelated, but should look into.. can fix by changing MSAA mode)

Hmm..

Common/GPU/Vulkan/VulkanFramebuffer.cpp

unknownbrackets · 2022-12-02T00:34:52Z

Common/GPU/Vulkan/thin3d_vulkan.cpp

-		caps_.multiSampleLevelsMask = (limits.framebufferColorSampleCounts & limits.framebufferDepthSampleCounts & limits.framebufferStencilSampleCounts);
+		// Check for depth stencil resolve. Without it, depth textures won't work, and we don't want that mess
+		// of compatibility reports, so we'll just disable stencil in this case for now.
+		// There are potential workarounds for devices that don't support it, but those are nearly non-existent now.


*MSAA, not stencil. Also, meant to check resolveProperties, like supportedStencilResolveModes?

-[Unknown]

Oops, yeah.

Not checking resolveModes, we just use "sample0" resolve mode for now which is required to be supported if resolve is supported at all, it's fine. It won't make very much of a difference to use the others, and I think doing average for example could result in artifacts in some casese.

Although just to be safe, I'll throw in a check for that resolve mode.

iota97 · 2022-12-02T11:23:40Z

This SEGFAULT on Linux, will provide a stack trace when I can :/

hrydgard · 2022-12-02T11:41:07Z

@iota97 thanks, please do.

Getting this in 1.14 might be ambitious, I may push it to later.

iota97 · 2022-12-02T11:59:57Z

Ok, kinda good news is this actually work on Linux using Vulkan, but crash on OpenGL.

Thread 29 "Emu" received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7fffc5ffb6c0 (LWP 12150)]
0x0000555555ad2e17 in GenerateVertexShader (id=..., buffer=0x7fffa05fffe0 "", compat=..., bugs=..., attrMask=0x7fffc5ffa744, uniformMask=0x7fffc5ffa748, vertexShaderFlags=0x0, errorString=0x7fffc5ffa770) at /home/iota/Documents/Git/ppsspp/GPU/Common/VertexShaderGenerator.cpp:133
133		*vertexShaderFlags = (VertexShaderFlags)0;
(gdb) bt
#0  0x0000555555ad2e17 in GenerateVertexShader (id=..., buffer=0x7fffa05fffe0 "", compat=..., bugs=..., attrMask=0x7fffc5ffa744, uniformMask=0x7fffc5ffa748, vertexShaderFlags=0x0, 
    errorString=0x7fffc5ffa770) at /home/iota/Documents/Git/ppsspp/GPU/Common/VertexShaderGenerator.cpp:133
#1  0x0000555555c1faec in ShaderManagerGLES::CompileVertexShader (this=0x7fffa0006b60, VSID=...) at /home/iota/Documents/Git/ppsspp/GPU/GLES/ShaderManagerGLES.cpp:769
#2  0x0000555555c21176 in ShaderManagerGLES::ContinuePrecompile (this=0x7fffa0006b60, sliceTime=0.0166666675) at /home/iota/Documents/Git/ppsspp/GPU/GLES/ShaderManagerGLES.cpp:1062
#3  0x0000555555c14741 in GPU_GLES::IsReady (this=0x7fffa0230310) at /home/iota/Documents/Git/ppsspp/GPU/GLES/GPU_GLES.cpp:213
#4  0x0000555555b73fbc in GPU_IsReady () at /home/iota/Documents/Git/ppsspp/GPU/GPU.cpp:57
#5  0x0000555555a695f5 in PSP_InitUpdate (error_string=0x7fffc5ffaa30) at /home/iota/Documents/Git/ppsspp/Core/System.cpp:485
#6  0x00005555556b9b91 in EmuScreen::bootGame (this=0x7fffa0002f20, filename=...) at /home/iota/Documents/Git/ppsspp/UI/EmuScreen.cpp:237
#7  0x00005555556bef9e in EmuScreen::update (this=0x7fffa0002f20) at /home/iota/Documents/Git/ppsspp/UI/EmuScreen.cpp:1037
#8  0x0000555555e0417a in ScreenManager::update (this=0x555557207ce0) at /home/iota/Documents/Git/ppsspp/Common/UI/Screen.cpp:51
#9  0x0000555555681d68 in NativeUpdate () at /home/iota/Documents/Git/ppsspp/UI/NativeApp.cpp:1279
#10 0x00005555557d6c44 in UpdateRunLoop () at /home/iota/Documents/Git/ppsspp/Core/Core.cpp:215
#11 0x0000555555e41c80 in EmuThreadFunc (graphicsContext=0x5555571dacd0) at /home/iota/Documents/Git/ppsspp/SDL/SDLMain.cpp:494
#12 0x0000555555e46b76 in std::__invoke_impl<void, void (*)(GraphicsContext*), GraphicsContext*> (__f=@0x5555572a15c0: 0x555555e41c3e <EmuThreadFunc(GraphicsContext*)>)
    at /usr/include/c++/12.2.0/bits/invoke.h:61
#13 0x0000555555e46af9 in std::__invoke<void (*)(GraphicsContext*), GraphicsContext*> (__fn=@0x5555572a15c0: 0x555555e41c3e <EmuThreadFunc(GraphicsContext*)>)
    at /usr/include/c++/12.2.0/bits/invoke.h:96
#14 0x0000555555e46a69 in std::thread::_Invoker<std::tuple<void (*)(GraphicsContext*), GraphicsContext*> >::_M_invoke<0ul, 1ul> (this=0x5555572a15b8)
    at /usr/include/c++/12.2.0/bits/std_thread.h:252
#15 0x0000555555e46a22 in std::thread::_Invoker<std::tuple<void (*)(GraphicsContext*), GraphicsContext*> >::operator() (this=0x5555572a15b8) at /usr/include/c++/12.2.0/bits/std_thread.h:259
#16 0x0000555555e46a06 in std::thread::_State_impl<std::thread::_Invoker<std::tuple<void (*)(GraphicsContext*), GraphicsContext*> > >::_M_run (this=0x5555572a15b0)
    at /usr/include/c++/12.2.0/bits/std_thread.h:210
#17 0x00007ffff7b122f3 in std::execute_native_thread_routine (__p=0x5555572a15b0) at /usr/src/debug/gcc/libstdc++-v3/src/c++11/thread.cc:82
#18 0x00007ffff77d18fd in ?? () from /usr/lib/libc.so.6
#19 0x00007ffff7853a60 in ?? () from /usr/lib/libc.so.6

You removed quite a few null pointer checks there, does this crash on Windows too with OpenGL?

hrydgard · 2022-12-02T12:29:26Z

Oops, I'll fix that up. Meant to always pass a valid pointer in, removing the need for the null checks, but apparently failed to update some of the other backends.

…Needed to keep all the information.

hrydgard · 2022-12-03T10:22:35Z

It works well enough now. I'm gonna get it in, and deal with any consequences :)

hrydgard added Vulkan GE emulation Backend-independent GPU issues labels Nov 28, 2022

hrydgard marked this pull request as ready for review November 28, 2022 23:16

hrydgard mentioned this pull request Nov 28, 2022

Massive performance drop in DX9 under very specific circumstances with an Nvidia GPU and certain driver settings #16332

Closed

5 tasks

hrydgard marked this pull request as draft November 29, 2022 08:33

hrydgard force-pushed the desktop-friendly-msaa branch from bffe96c to 0ef03bf Compare November 29, 2022 10:19

hrydgard mentioned this pull request Nov 29, 2022

Fragment shader generator: Move the framebuffer fetch bit to the shader ID. #16462

Merged

hrydgard added 14 commits December 1, 2022 22:48

Add check for tiling GPUs. We'll use this to inform on what MSAA mode…

6daecb4

…s to support.

Multisampling groundwork

06af304

Basically working MSAA on desktop GPUs! Some glitches remain.

4dfce4f

Fix switching between MSAA levels

e3679df

Add path that uses CreateRenderpass2, we're gonna need it for depth r…

8b9b80f

…esolve.

Implement depth/stencil resolve. (We probably only need depth resolve…

a0f27c7

… since we don't texture from stencil, but whatever).

Add a pipeline flag for USES_DISCARD. Will be used for the MSAA quali…

8208768

…ty setting.

Use raster copies to work around lack of working copy support in MSAA…

b9ea8de

… mode.

Implement "enhanced" multisample quality setting

4990683

Remove quality setting for now, we default to a balanced setting.

f758cc0

Headless buildfix. Crashfix in shader generator test.

7950a00

Correctly handle input attachments with CreateRenderpass2. But also d…

29a2157

…on't use it while not multisampling.

Detect Apple/MoltenVK and allow multisampling. Unfortunately no suppo…

28e1b53

…rt for sample rate shading.

Not pretty, but with this, you can switch MSAA level at runtime.

8a3e92a

hrydgard force-pushed the desktop-friendly-msaa branch from 588d50a to 8a3e92a Compare December 1, 2022 22:42

hrydgard added this to the v1.14.0 milestone Dec 1, 2022

hrydgard marked this pull request as ready for review December 1, 2022 22:44

Assert improvement

5610734

unknownbrackets approved these changes Dec 2, 2022

View reviewed changes

hrydgard added 5 commits December 2, 2022 13:49

Crashfix the other backends

cd0acec

Address feedback

75b558e

Check resolve mode, just to be safe.

2d6c632

Implement copy operation properly for the multisampled case

d554085

Use proper copies for depth blits between multisampled framebuffers. …

35777d5

…Needed to keep all the information.

hrydgard changed the title ~~Implement MSAA support for desktop GPUs~~ Implement MSAA support for desktop GPUs in Vulkan Dec 3, 2022

hrydgard merged commit 02b8bf3 into master Dec 3, 2022

hrydgard deleted the desktop-friendly-msaa branch December 3, 2022 10:22

ghost mentioned this pull request Dec 3, 2022

Feature Request: Add support for MSAA #6218

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Implement MSAA support for desktop GPUs in Vulkan #16458

Implement MSAA support for desktop GPUs in Vulkan #16458

hrydgard commented Nov 28, 2022 •

edited

Loading

hrydgard commented Nov 28, 2022 •

edited

Loading

hrydgard commented Nov 28, 2022

hrydgard commented Nov 28, 2022

hrydgard commented Nov 29, 2022

hrydgard commented Dec 1, 2022 •

edited

Loading

unknownbrackets Dec 2, 2022

hrydgard Dec 2, 2022

hrydgard Dec 2, 2022

iota97 commented Dec 2, 2022

hrydgard commented Dec 2, 2022

iota97 commented Dec 2, 2022

hrydgard commented Dec 2, 2022

hrydgard commented Dec 3, 2022

Implement MSAA support for desktop GPUs in Vulkan #16458

Implement MSAA support for desktop GPUs in Vulkan #16458

Conversation

hrydgard commented Nov 28, 2022 • edited Loading

hrydgard commented Nov 28, 2022 • edited Loading

hrydgard commented Nov 28, 2022

hrydgard commented Nov 28, 2022

hrydgard commented Nov 29, 2022

hrydgard commented Dec 1, 2022 • edited Loading

unknownbrackets Dec 2, 2022

Choose a reason for hiding this comment

hrydgard Dec 2, 2022

Choose a reason for hiding this comment

hrydgard Dec 2, 2022

Choose a reason for hiding this comment

iota97 commented Dec 2, 2022

hrydgard commented Dec 2, 2022

iota97 commented Dec 2, 2022

hrydgard commented Dec 2, 2022

hrydgard commented Dec 3, 2022

hrydgard commented Nov 28, 2022 •

edited

Loading

hrydgard commented Nov 28, 2022 •

edited

Loading

hrydgard commented Dec 1, 2022 •

edited

Loading