TAG COLLECTION

GPU programming collection

Explore providers, articles, and companies related to GPU programming for practical research and comparison.

4 Providers 0 Articles 0 Companies

Providers

4 results

CUDA

United States

CUDA has turned NVIDIA GPUs into general parallel processors since 2006, using an SIMT model with 32-thread warps and backed by cuBLAS, cuDNN and TensorRT for AI and HPC; the performance ceiling is high, but it is fully tied to NVIDIA hardware.

GLSL

United States

GLSL arrived with the OpenGL 2.0 programmable pipeline and is maintained by Khronos, covering 6 shader stages with C-like syntax; its reach across OpenGL, WebGL and Vulkan is unmatched, while version fragmentation, no module system and thin debugging are the practical costs.

HLSL

United States

HLSL is Microsoft's shader language for DirectX, with Shader Model 6.x defining its limits across 7 shader stages and tight alignment with the DirectX 12 binding model; it is strongest on Windows and Xbox, at the cost of platform lock-in and hardware tier management.

Metal

United States

Metal has been Apple's unified graphics and compute interface since 2014, pairing the C++14-based MSL with unified memory that removes explicit copies and Metal Performance Shaders that supply ready-made kernels; integration and performance are excellent, but the platform is Apple only.

Related tags