Topics / Loop unrolling and instruction-level parallelism

Performance topic

Loop unrolling and instruction-level parallelism

Break dependency chains so the CPU can run independent instructions in parallel: unrolling, multiple accumulators and interleaved streams.

705proven techniques
15pitfalls: tried, then slower
389repositories
17from papers

Languages

  • C++213
  • C184
  • Rust57
  • CUDA37
  • Go12
  • Python11

Hardware

  • x86269
  • ARM136
  • CPU131
  • NVIDIA GPU54
  • GPU24

Fields

  • math167
  • data & compression161
  • general127
  • CPU91
  • GPU75

Where these come from

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

torvalds/linux · 12openssl/openssl · 11weidai11/cryptopp · 9golang/go · 9gmp-mirror/gmp · 8facebook/zstd · 7OpenMathLib/OpenBLAS · 7wolfSSL/wolfssl · 7Cyan4973/xxHash · 7Legrandin/pycryptodome · 7

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

GPU kernel fusionArena and pool allocationI/O batchingFast math and approximationSIMD vectorization