Graphics Programming weekly - Issue 55 — September 23, 2018

posted in jendrikillner
Published September 23, 2018 Imported
Advertisement
  • new shader stages on Turing that allows the generation of primitives for direct consumption by the rasterizer
  • task shader: emits mesh shader workgroups
    • allows work on larger clusters, the selection of mesh resolution, …
  • mesh shader: generates primitives (index + vertex buffers)
  • OpenGL shader implementation that uses the new shader stages to allow culling on a per-cluster basis and generation of efficient mesh data
  • video recording
turing_mesh_shader.png

  • talks about the challenges of developing for open world games
  • covers water rendering, physically based time of day model, nighttime rendering
  • lighting, exposure, emissive effects, and tone mapping
  • material blending
fc5_open_world.png
  • a new algorithm to compute the intersection point and surface normal of an oriented 3D box without precomputations or spatial data structures
  • uses rasterizer pipeline to generate screen-space AABB to limit tests to only potentially visible pixels
  • provides GLSL shader implementation
ray_box_intersection.png
  • explains the evolution of the windows graphics driver model from win7 to win10
  • how WDDM 2.0 enables the features of D3D12
  • differences between D3D12 and Vulkan regarding queue submission design
  • D3D12 fences are designed to enable the OS to control command buffer scheduling
  • look at Nvidia and AMD hardware details
amd_command_processors.png
  • explains optimizations done to the hybrid shadow raytracer
  • change to the way the BVH layout is constructed (Surface Area Heuristic) and memory layout changes enable huge speedups (16ms to 0.8ms on the GTX970)
bvh-with-sah-heatmap.png
  • overview of the new Turing hardware architecture
  • description of new hardware features, exposed shader features, and improvements to existing technologies
turing.jpg
  • an extensive list of OpenGL and Vulkan extensions that expose the new Turing hardware features
  • with short descriptions of each feature
variable_rate_shading.png
  • Vulkan extension that makes it possible to skip parts of command buffers based on the content of a GPU buffer
  • the example uses this to pre-generate a single command buffer and controls visibility of mesh parts from the CPU
vulkan_conditional_rendering.png
  • overview of the interior mapping technique
  • showcase of different implementations and links to further information
Interior_Mapping.png
  • discussing of a technique that tries to preserve normal map variation by generation MIPs that preserve the variance in the roughness map
  • the comments contain links to further reading about this topic
roughness_normal_map.png
  • comparison of DXC (DirectX Shader Compiler) and FXC shader compiler in regards to performance and functional regressions
  • provides a docker file that allows FXC to run successfully
fxc_docker.png
  • provides a configuration that allows the DXC compiler to run within docker
  • explains the process that enables DXIL shaders to be signed after they have been compiled
  • provides code for detecting if a shader has already been signed and how to call the necessary libraries to sign the shader
  • provides a command line utility that can be used to sign DXIL shaders
dxil_signing.png
  • start of a series of posts that will describe the development of a ray tracer using Metal Performance Shaders
metal_ray_tracer.png
  • Khronos released provisional material on the memory model for Vulkan
  • including the specification, extensions for SPIR-V, GLSL and conformance tests
vulkan.png
  • presents a small cross-vendor library for D3D12 that allows collection of information to narrow down GPU hangs to specific draw call ranges
  • looks at the state of integration of the Vulkan swap chain with different OS and GPU vendors
  • explains the expected API, what is required for a good experience and how the implementations differ between platforms
  • discusses challenges and possible solutions to undersampling artifacts on curved geometry
cup_tesselation.png
  • links to the presentations from the machine learning and rendering course from Siggraph
  • an interview that explains what Nvidias DLSS is (Deep Learning Super Sampling)
  • a trained neural network that is able to infer a 64 sample per pixel supersampled image from a 1 sample per pixel source
  • the model is trained on Nvidia GPU clusters before the release of the game
DLSS1.png
  • explains how to generate value noise in a shader
value_noise.png
  • explains how to generate Perlin noise in a shader
perlin_noise.png
  • extends the Perlin noise implementation from the previous tutorial to support mixing of multiple layers of Perlin noise
  • uses this to generate a dynamic terrain in a vertex shader
layered_noise.png
  • explains how Unity command buffers enable the extension of the Unity graphics pipeline
  • uses this to implement selective bloom on objects
selective_bloom.png

If you are enjoying the series and getting value from it, please consider supporting this blog.

Support this blog

Read more

0 likes 0 comments

Comments

Nobody has left a comment. You can be the first!
You must log in to join the conversation.
Don't have a GameDev.net account? Sign up!
Advertisement