Vulkan renderer backend

[adrian-cojocaru]

Rough implementation of a Vulkan renderer backend (tested on Windows).

Notes:

• Vulkan 1.0
• validation layers are populated from the environment (VulkanSDK, vkconfig, etc)
• Vulkan-Headers C++ bindings with dynamic function loader
• VulkanMemoryAllocator
• runtime shader compilation using glslang
• vsync is forcefully enabled
• requires a valid surface (no offscreen rendering at the moment)
• a common pipeline layout is used for compatibility
• one command buffer submitted per frame

List of working shaders (and associated functionality):

• BackgroundShader
• BackgroundPatternShader
• CircleShader
• ClippingMaskProgram
• CollisionBoxShader
• CollisionCircleShader
• CustomSymbolIconShader
• DebugShader
• FillShader
• FillOutlineShader
• FillPatternShader
• FillOutlinePatternShader
• FillOutlineTriangulatedShader
• FillExtrusionShader
• FillExtrusionPatternShader
• HeatmapShader
• HeatmapTextureShader
• HillshadeShader
• HillshadePrepareShader
• LineShader
• LineGradientShader
• LineSDFShader
• LinePatternShader
• RasterShader
• SymbolIconShader
• SymbolSDFIconShader
• SymbolTextAndIconShader
• WideVectorShader

[github-actions]

Bloaty Results (iOS) 🐋

Compared to main

    FILE SIZE        VM SIZE    
 --------------  -------------- 
  +0.0% +1.27Ki  [ = ]       0    TOTAL

Full report: https://maplibre-native.s3.eu-central-1.amazonaws.com/bloaty-results-ios/pr-2564-compared-to-main.txt

[github-actions]

Bloaty Results 🐋

Compared to main

    FILE SIZE        VM SIZE    
 --------------  -------------- 
  +0.0% +20.7Ki  +0.0% +3.88Ki    TOTAL

Full report: https://maplibre-native.s3.eu-central-1.amazonaws.com/bloaty-results/pr-2564-compared-to-main.txt

Compared to d387090 (legacy)

    FILE SIZE        VM SIZE    
 --------------  -------------- 
   +21% +24.7Mi  +411% +24.6Mi    TOTAL

Full report: https://maplibre-native.s3.eu-central-1.amazonaws.com/bloaty-results/pr-2564-compared-to-legacy.txt

[github-actions]

Benchmark Results ⚡

Benchmark                                                     Time             CPU      Time Old      Time New       CPU Old       CPU New
------------------------------------------------------------------------------------------------------------------------------------------
OVERALL_GEOMEAN                                            +0.0019         +0.0017             0             0             0             0

Full report: https://maplibre-native.s3.eu-central-1.amazonaws.com/benchmark-results/pr-2564-compared-to-main.txt

[louwers]

As discussed, let's not add any tests to CI yet for the Vulkan backend, but make sure it builds for Linux on CI.

@adrian-cojocaru If you can share some build instructions I am happy to add it to linux-ci.yml.

[louwers]

Happy we got a build going on CI!

There's not a lot I can comment on except for C++ nitpicking. Left some comments but there's not really any blockers in that regard from my side.

[jwinarske]

I am able to build and run mbgl-glfw built with clang 10 as well as GCC 9.4 with attached patch.

Lots of clang-tidy issues; some are directly related to the compiler errors.

Build flags:

-DMLN_WITH_VULKAN=ON
-DMLN_WITH_X11=OFF
-DMLN_WITH_WAYLAND=OFF
-DMLN_DRAWABLE_RENDERER=ON
-DMLN_LEGACY_RENDERER=OFF
-DMLN_WITH_VULKAN=ON
-DMLN_WITH_OPENGL=OFF
-DMLN_WITH_EGL=OFF

0001-Changes-to-build-GCC-9.4-Clang-10.zip

[louwers]

@jwinarske Could you share your clang-tidy config? Here is ours: https://github.com/maplibre/maplibre-native/blob/main/.clang-tidy

Any chance you could upstream your patch as a PR?

[adrian-cojocaru]

@jwinarske Seems to compile without any issue on clang 10/GCC 9.4. As @louwers suggested the difference is probably in clang-tidy config.

[birkskyum]

Can/will this also be used for validating the opengl GLSL shaders, or is it only relevant for vulkan/SPIR-V?

[adrian-cojocaru]

It can be used for validation/reflection but it's relevant only for Vulkan at the moment. Ideally even there it should end up as a build time step to avoid linking.