ConceptKernel / GPU Kernel

Compute Kernel

The smallest executable computational unit running on a GPU. A single Transformer layer is typically split into dozens of kernels executed in sequence. Kernel fusion is a common approach to inference optimization.

Timeline (last 90 days)

Oct 13

A compute kernel is the smallest executable compute unit running on a GPU, and a single Transformer layer is often split into dozens of kernels executed in sequence.

Unverified50%

All Facts (1)

Unverified

A compute kernel is the smallest executable compute unit running on a GPU, and a single Transformer layer is often split into dozens of kernels executed in sequence.

50%

Source Articles

vLLM New Release: DeepSeek Inference Performance Soars 1.9x to 5.3xOct 13