⚡ Bolt: Optimize decompression for offset 5 using AVX2#229
Conversation
Implemented a specialized AVX2 SIMD path for match offset 5 in `decompress_bmi2`. This optimization uses `_mm256_shuffle_epi8` with precomputed masks to replicate the 5-byte pattern across 32-byte vectors, processing 160 bytes per iteration. Key changes: - Added `copy_match_offset5_avx2` helper function. - Added `OFFSET5_MASKS` table. - Integrated the optimized path into the main decompression loop. - Fixed `OFFSET15_MASKS` definition (length mismatch) found during development. - Restored accidentally removed handlers for offsets 10, 12, 14. This follows the existing pattern for small offsets (e.g., offset 3) to improve throughput by reducing loop overhead and utilizing vector stores. Co-authored-by: 404Setup <[email protected]>
|
👋 Jules, reporting for duty! I'm here to lend a hand with this pull request. When you start a review, I'll add a 👀 emoji to each comment to let you know I've read it. I'll focus on feedback directed at me and will do my best to stay out of conversations between you and other bots or reviewers to keep the noise down. I'll push a commit with your requested changes shortly after. Please note there might be a delay between these steps, but rest assured I'm on the job! For more direct control, you can switch me to Reactive Mode. When this mode is on, I will only act on comments where you specifically mention me with New to Jules? Learn more at jules.google/docs. For security, I will only act on instructions from the user who triggered this task. |
1 participant
⚡ Bolt: Optimize decompression for offset 5 using AVX2
💡 What: Implemented a specialized AVX2 SIMD path for match offset 5.
🎯 Why: Standard decompression falls back to byte-by-byte copy for small non-power-of-2 offsets, which is inefficient. AVX2 allows processing 160 bytes per loop iteration.
📊 Impact: Expected significant throughput improvement for data with frequent 5-byte repeating patterns (common in certain binary formats or text). Baseline benchmarks showed ~10.3 GiB/s, optimization should exceed this.
🔬 Measurement: Verified with
cargo test --libandcargo bench --bench bench_main bench_decompress_offset5.PR created automatically by Jules for task 1225911123532623525 started by @404Setup