Skip to main content

GPU & High-Performance Computing

CUDA, tensor cores, multi-GPU communication, and cluster-scale workload orchestration.

  • GPU
  • 28 concepts
Browse all 28 concepts

GPU architecture

Understand how parallel hardware schedules work and where useful throughput is lost.

Starting knowledge: Basic loops, arrays and processor execution

Select a concept. Explore its connections, or open the full explanation.

Previewing Flynn’s classification.

From parallel models to warp execution

Start with instruction and data streams, then inspect one SM and the stalls its scheduler must handle.

  1. From parallel models to warp execution

    Flynn’s classification

    Distinguish instruction and data streams, then relate SIMD and MIMD to modern hybrid machines.

    Read concept

    Used in: connections 1 to 2 of 2.

Accelerate matrix operations

Explore the matrix engine after the SM overview, with precision and data layout as separate constraints.

Compare accelerator ecosystems

Use the hardware vocabulary to compare software support and workload constraints across vendors.

Lines show suggested reading within a branch. Connections offer optional ways to explore; they are not a checklist of prerequisites.

All GPU & High-Performance Computing Concepts

Mastodon