CFP: RustINDIA Conference 2026

GPU programming has reached a state where developers are forced into vendor-specific ecosystems: CUDA for NVIDIA, Metal for Apple, DX12 for Windows, with minimal portability between them. WebGPU unifies this as a GPU API that targets Vulkan, DX12, and Metal as backends. wgpu is the Rust implementation of WebGPU, acting as the translation layer between your code and the underlying platform APIs. This talk begins with how different GPU architectures execute workloads, introducing wgpu and its significance in the GPU stack. We walk through the wgpu object model, bindings, and adapters, covering how wgpu’s Naga shader translation layer works, which takes WGSL and emits SPIR-V, MSL, or HLSL depending on the target backend, validating shaders for type correctness, memory safety, and undefined behavior before they ever reach the driver.

The second half shifts to what Rust uniquely contributes to GPU programming. We walk through how wgpu’s API encodes GPU resource lifetimes directly in the ownership system, preventing errors such as use-after-free on buffers at compile time, and making cross-thread resource sharing safe via the Send + Sync model. We walk through a concrete compute example in Rust and WGSL, benchmarking it against native Vulkan bindings via ash and vulkano, examining where the abstraction is effectively zero-cost and where it carries measurable overhead. The talk closes with an honest look at the current limits of wgpu, what native APIs expose that wgpu cannot, and a practical framework for deciding when wgpu is the right tool and when dropping to a native backend is worth the tradeoff.

Takeaways

• How wgpu fits in the GPU stack, how Naga translates WGSL to platform-specific backends, and how Rust’s ownership system eliminates real GPU programming bugs like use-after-free and cross-thread resource misuse at compile time rather than at runtime.

• An honest evaluation of the performance tradeoffs between wgpu and native Vulkan bindings, identifying the specific capabilities that native APIs expose that wgpu cannot.

• A look into the portability of GPU programs and open source development for GPU programming in Rust.

Target Audience

• Rust developers and systems programmers looking to get started with GPU compute without getting locked into a vendor-specific ecosystem.

• Engineers already working with native GPU APIs like Vulkan or CUDA who want to understand what wgpu offers and where its limits are.

Bio

Prajwal KP :- Open source contributer with experience in compilers. Contributions to LLVM, AFL++ and given co-authored a talk in BITS Pilani conference

Bhuvan B :- Open source contributer and systems enthusiast. Contributed to Open Robotics and RTEMS. Working on a wgpu port of a open robotics sonar CUDA simulation plugin to make it vendor agnostic using wgpu.
github: https://github.com/BhuvanB404/Dave-Sonar-WGPU-