Sign inSign in
ConceptAI & ML

Kernel fusion

Combining multiple separate kernels into a single kernel to reduce memory roundtrips and avoid redundant global memory accesses.
Watch it explained

Know first

Named as a prerequisite by the lectures that define this. The number is how many say so. A grey one has no page yet.

CUDA kernels1GPU architecture1GPU computing1GPU kernels1Kernel execution in PyTorch1Memory hierarchy1Race condition1

Opens up

Concepts that name this one as a prerequisite.

Lecturers

2

Each in their own terms, from the easiest level.

  • Andrej Karpathy

    Graduate

    An optimization technique where multiple element-wise operations are combined into a single kernel execution to avoid repeated round-trips to the global memory.
    • 1:55:09Let's Reproduce GPT-2 (124M): Distributed PyTorch, FlashAttention & FineWeb

Connections

4

The lectures that define it or assume it, and the concepts next to it.

2 lecturers. Definitions are written by AI from each lecture’s transcript. A time with a play mark starts the lecture at the moment the lecturer says it.