{"schema_version":1,"rust_single_gate":{"description":"Criterion benchmark of one Hadamard gate applied to a statevector. Only the independent oracle, allocation-heavy reshape backend, and direct in-place bit-shift backend are retained for the article comparison.","measured_at":"2026-08-08","repository":"https://github.com/nez0b/rustatevec","commit":"05e720d9d7289b97516a2cb62f37d9f56c405185","benchmark_source":"crates/qsv-bench/benches/throughput.rs::single_gate","command":"env RAYON_NUM_THREADS=1 RUSTFLAGS='-C target-cpu=native' cargo bench -p qsv-bench --bench throughput -- single_h_gate","criterion":{"version":"0.5.1","samples":100,"warm_up_seconds":3.0,"confidence_level":0.95,"estimate":"arithmetic mean"},"environment":{"machine":"MacBook Pro (Mac15,6)","chip":"Apple M3 Pro","cores":"11 (5 performance, 6 efficiency)","memory":"36 GB","os":"macOS 15.4.1 (Darwin 24E263)","rustc":"rustc 1.86.0 (05f9846f8 2025-03-31), LLVM 19.1.7, aarch64-apple-darwin","cargo":"cargo 1.86.0 (adf9b6ad1 2025-02-28)","rustflags":"-C target-cpu=native","rayon_num_threads":1},"measurements":[{"backend":"oracle","display_name":"Independent oracle","qubits":12,"amplitudes":4096,"mean_ns":9251.184099,"ci95_lower_ns":9207.097474,"ci95_upper_ns":9301.286965},{"backend":"reshape","display_name":"Reshape + allocate","qubits":12,"amplitudes":4096,"mean_ns":5527.237843,"ci95_lower_ns":5503.344352,"ci95_upper_ns":5556.986294},{"backend":"bitshift","display_name":"In-place bit shift","qubits":12,"amplitudes":4096,"mean_ns":4035.173567,"ci95_lower_ns":4005.993141,"ci95_upper_ns":4070.156891},{"backend":"oracle","display_name":"Independent oracle","qubits":16,"amplitudes":65536,"mean_ns":161302.352361,"ci95_lower_ns":159093.876965,"ci95_upper_ns":164249.455175},{"backend":"reshape","display_name":"Reshape + allocate","qubits":16,"amplitudes":65536,"mean_ns":110847.803849,"ci95_lower_ns":109298.065757,"ci95_upper_ns":112936.223530},{"backend":"bitshift","display_name":"In-place bit shift","qubits":16,"amplitudes":65536,"mean_ns":64731.882688,"ci95_lower_ns":64433.750561,"ci95_upper_ns":65048.248962}]},"quest_bmi2_case_study":{"description":"Ranges at three progressively broader measurement levels from the work behind QuEST issue #717 and merged PR #796. The levels are intentionally not treated as one apples-to-apples benchmark.","measured_at":"2026-06","issue":"https://github.com/QuEST-Kit/QuEST/issues/717","pull_request":"https://github.com/QuEST-Kit/QuEST/pull/796","pull_request_merged_at":"2026-06-28","environment":{"cpu":"Intel Xeon Gold 6448H (Sapphire Rapids)","compiler":"GCC 11.2","notes":"Single-thread measurements unless stated otherwise; large statevectors and threaded kernels become memory-bandwidth-bound."},"levels":[{"key":"isolated","display_name":"Isolated PDEP/PEXT","scope":"Bit gather/scatter helper in a tight loop, with invariant masks hoisted","min_speedup":6.0,"max_speedup":12.0,"representative_speedup":8.9,"representative_note":"Sorted gather example: scalar 4.36 ns/call versus PEXT 0.49 ns/call","source":"PR.md, Benchmarks and AVX-512 comparison"},{"key":"kernel","display_name":"Inside a gate kernel","scope":"One-thread, L3-resident q=18 kernel including strided complex-amplitude traffic","min_speedup":2.53,"max_speedup":3.89,"representative_speedup":3.14,"representative_note":"PDEP scatter 2.53x; sorted PEXT gather 3.89x","source":"PR.md, Thread scaling table"},{"key":"circuit","display_name":"Whole circuits","scope":"QFT, random/supremacy, Grover, and VQE plus measurement at 12 and 16 qubits","min_speedup":1.01,"max_speedup":1.28,"representative_speedup":1.22,"representative_note":"Range across eight reported single-thread circuit measurements","samples":[1.26,1.01,1.16,1.27,1.17,1.27,1.15,1.28],"source":"PR.md, whole-circuit table"}]}}