Skip to main content

Loading the code workspace…

LC

Structure of Arrays vs Array of Structures Benchmark

medium3/8

Fix False Sharing Bug: Two Threads Updating Adjacent Array Elements Implement Cache-Oblivious Matrix Transpose

Language

Compiler

2.0s Limit 256MB

Submit and pass all tests to enable saving to GitHub.

Implementation

Benchmark Structure of Arrays vs Array of Structures.

Requirements:

Implement Particle structure (AoS): x, y, z, vx, vy, vz
Implement Particles structure (SoA): separate arrays for each component
Implement update_positions_aos() for AoS
Implement update_positions_soa() for SoA
Benchmark both with 1M particles
Measure cache performance with perf stat
Calculate and print speedup
Show cache hit rate difference
Demonstrate SIMD auto-vectorization benefit
Print layout comparison and analysis

Why This Matters

Game engines and physics simulators (Unreal's particle systems, DOD-style ECS frameworks like EnTT) switched from AoS to SoA specifically to let the compiler auto-vectorize with AVX/AVX2, since scattered per-particle fields defeat SIMD entirely. This same layout decision is why columnar databases like ClickHouse and Parquet outperform row-oriented storage by an order of magnitude on analytical scans.

Hints

0 / 3 revealed

Validation Suite

main.c

Workspace 1.0

← Fix False Sharing Bug: Two Threads Updating Adjacent Array ElementsImplement Cache-Oblivious Matrix Transpose →