Topics / Zero-copy

Performance topic

Zero-copy

Move references, not bytes: memory-mapped files, borrowed slices, scatter-gather I/O and shared buffers between stages.

1,179proven techniques
56pitfalls: tried, then slower
665repositories
86from papers

Examples from real commits

nodejs/node 5.4× Move object creation and arithmetic to JS, returning raw values via a pre-allocated Uint32Array. proof: commit → blender/blender 2× Let the memfile undo step take shared ownership of immutable data via implicit-sharing info instead of serializing and diffing it, with BLO_write_shared and BLO_read_shared. proof: commit →

Languages

  • C++520
  • Rust191
  • C132
  • Python109
  • Go47
  • CUDA38

Hardware

  • CPU243
  • NVIDIA GPU141
  • x8678
  • GPU75
  • ARM30

Fields

  • general381
  • data & compression282
  • GPU138
  • graphics127
  • networking108

Where these come from

Repositories with the most techniques on this topic in the database:

dora-rs/dora · 10apache/iggy · 8nodejs/node · 7apache/arrow-rs · 7Flaminis/Dalaran · 7HelixDB/helix-db · 7libbitcoin/libbitcoin-system · 6axmolengine/axmol · 6TigerVNC/tigervnc · 6vllm-project/vllm · 6

Ask Claude for the full answer

Each technique comes with its bottleneck, measured speedup, trade-offs, conditions and the commit or paper behind it, served to Claude through MCP.

claude mcp add --transport http adtomic https://www.adtomic.co/mcp

Connect Claude Pricing

Related topics

Branchless programmingInlining and devirtualizationLookup tablesCache locality and data reorderingMultithreading and work stealing