GitHub - rdbuf/istrstr: SIMD-accelerated string searching / proof of concept

Branches Tags

Name		Name	Last commit message	Last commit date
Latest commit History 20 Commits
.idea		.idea
src		src
CMakeLists.txt		CMakeLists.txt
LICENSE		LICENSE
README		README
conanfile.txt		conanfile.txt
testgen.py		testgen.py

Repository files navigation

Problem: search fast in multiple GB's of text.
Concept: perform a 2-pass search consisting of
    1) a SIMD-powered filter for 4-byte prefixes
    2) and a slower full string search.
Current SOTA for 1:
    - (case-sensitive)   4.465356282457702 bytes/cycle, 13.333333333333334 GB/s
    - (case-insensitive) 3.668801996256802 bytes/cycle, 10 GB/s
Measured on MBP 15 2019.