All TechSheets

#algorithms

7 articles

Case-Folding at 45 GiB/s: How Branchless Byte-Space Arithmetic Reaches Memory Speed
PerformanceEngineering

Case-Folding at 45 GiB/s: How Branchless Byte-Space Arithmetic Reaches Memory Speed

Discover how GitHub eliminated CPU branch mispredictions and achieved 45+ GiB/s single-core case-folding throughput using SIMD, SWAR, and byte-space arithmetic.

T
Thanga·
9 min read
Case-Folding Source Code at 45+ GiB/s: Branchless Loops and Byte-Space Arithmetic
PerformanceSystems

Case-Folding Source Code at 45+ GiB/s: Branchless Loops and Byte-Space Arithmetic

An architectural deep-dive into how modern search engines case-fold billions of bytes of source code at hardware memory limits using branch-free SWAR and byte-space arithmetic.

T
Thanga·
8 min read
Case-Folding Code at Memory Speed: How Branch-Free Loops Hit 45 GiB/s
PerformanceSystems

Case-Folding Code at Memory Speed: How Branch-Free Loops Hit 45 GiB/s

An architectural deep-dive into branchless byte-space arithmetic, SWAR, and SIMD techniques for ultra-high-throughput string normalization.

T
Thanga·
7 min read
Case-Folding Source Code at 45 GiB/s: Branchless Bitwise Arithmetic for Ultra-Fast Search
PerformanceArchitecture

Case-Folding Source Code at 45 GiB/s: Branchless Bitwise Arithmetic for Ultra-Fast Search

Learn how branchless bit arithmetic and SIMD vectorization enable case-folding source code at memory bandwidth limits without CPU branch mispredictions.

T
Thanga·
8 min read
Case-Folding Source Code at Memory Speed: Branchless Loops and Byte-Space Arithmetic
PerformanceAlgorithms

Case-Folding Source Code at Memory Speed: Branchless Loops and Byte-Space Arithmetic

Learn how branch-free loops, bitwise arithmetic, and SWAR/SIMD vectorization enable source code case-folding at over 45 GiB/s on a single CPU core.

T
Thanga·
9 min read
Case-Folding Code at 45 GiB/s: Branch-Free Arithmetic and SIMD Search
PerformanceAlgorithms

Case-Folding Code at 45 GiB/s: Branch-Free Arithmetic and SIMD Search

Learn how branch-free byte-space arithmetic and SIMD vectorization push case-insensitive code search to the physical limits of memory bandwidth.

T
Thanga·
9 min read
Case-Folding Source Code at 45 GiB/s: Branch-Free Byte Arithmetic and SWAR
PerformanceSystems

Case-Folding Source Code at 45 GiB/s: Branch-Free Byte Arithmetic and SWAR

Learn how branch-free loops and SWAR byte-space arithmetic eliminate CPU pipeline stalls to case-fold source code at memory bandwidth speeds.

T
Thanga·
8 min read