Differentiable Slang: A Shading Language for Renderers That Learn (NVIDIA)

Scott_Arm · Oct 24, 2023

Differentiable Slang: A Shading Language for Renderers That Learn | NVIDIA Technical Blog

NVIDIA just released a SIGGRAPH Asia 2023 research paper, SLANG.D: Fast, Modular and Differentiable Shader Programming. The paper shows how a single language can serve as a unified platform for real…

developer.nvidia.com

SLANG.D: Fast, Modular and Differentiable Shader Programming | NVIDIA Real-Time Graphics Research

We introduce SLANG.D, a shading language that incorporates first-class automatic differentiation support derived from the Slang language. The new shading language allows us to transform a Direct3D-based path tracer to be fully differentiable with minor modifications to existing code. SLANG.D...

research.nvidia.com

Differentiable Slang: Example Applications | NVIDIA Technical Blog

This is part of a series on Differentiable Slang. For more information about differential programming and automatic gradient computation in the Slang language, see Differentiable Slang: A Shading…

developer.nvidia.com

Just starting to read through. Seems interesting.

Scott_Arm · 2025-01-31T16:46:35+0000

I was reading through the blackwell architecture doc and come across this part about Neural shaders. Seems like Slang is the way Nvidia expects devs to write "neural" shaders.

Up until this point, NVIDIA has been using neural
shading for DLSS, using CUDA to harness the
Tensor Cores. With the new Cooperative Vectors
API for DX12 and Vulkan, Tensor Cores can be
accessed through any type of shader, including
pixel and ray tracing, in a graphics application
allowing for a host of neural technologies. NVIDIA
has worked with Microsoft to create the new
Cooperative Vectors API. When combined with
differentiable shading language features in
Slang, Cooperative Vectors unlock the ability for
game developers to use neural techniques in
their games including neural texture
compression, that provides up to seven-to-one
VRAM compression over block compressed
formats, and other techniques such as RTX
Neural Materials, Neural Radiance Cache, RTX
Skin, and RTX Neural Faces.

Because Slang is an IR, it can generate HLSL or whatever shader format is needed for your program. Slang allows for automatic-differentiation, which I guess is the key part of neural shaders.

A key pillar of ML methods is gradient-based optimization. Specifically, most ML algorithms are powered by reverse-mode automatic differentiation, an efficient way to propagate derivatives through a series of computations.