CUDA

CUDA is NVIDIA's platform to run parallel computations across thousands of GPU cores. It's an extension of C/C++ and the foundation of AI model training, scientific simulations and large-scale data processing.

• Creado por NVIDIA
• Año 2007
• Paradigma: Parallel (GPU) · compiled
• Extensiones: .cu, .cuh

Para qué se usa

• Artificial intelligence
• Scientific simulation
• Massive GPU computing

A favor

• Brutal GPU acceleration
• Key in deep learning
• Mature ecosystem

En contra

• NVIDIA GPUs only
• Hard parallel programming
• Proprietary

Ecosistema

• cuDNN
• cuBLAS
• Thrust
• Nsight

Ejemplo

__global__ void hola() {
  printf("¡Hola, mundo!\n");
}