An optimized schedule for a simulated VLIW/SIMD CPU kernel executing a parallel tree traversal & custom hashing workload using Python. Achieves a 16.74x speedup (reducing CPU cycles from 18532 to 1107), surpassing the Claude Opus 4.5 baseline.
performance-engineering hpc simd vectorization compiler-optimization cpu-simulator vliw instruction-scheduling anthropic
-
Updated
May 24, 2026