[canonical 4.2 harness] session=C2
[canonical 4.2 harness] protocol: 10 warmup + 100 continuous timed iters
[canonical 4.2 harness] inter-shape settle: 5.0s (NOT section-4 cooldown)
  lcsa_small_seq4k                 d=0.239 V2 p50= 1.143ms SDPA p50= 2.573ms ratio= 2.25x
  lcsa_small_seq4k_sparse          d=0.067 V2 p50= 0.663ms SDPA p50= 2.523ms ratio= 3.81x
  lcsa_mid_seq8k                   d=0.119 V2 p50= 1.220ms SDPA p50= 6.429ms ratio= 5.27x
  lcsa_mid_seq8k_sparse            d=0.030 V2 p50= 0.611ms SDPA p50= 6.439ms ratio=10.54x
  lcsa_large_seq16k                d=0.120 V2 p50= 2.027ms SDPA p50=12.715ms ratio= 6.27x
  lcsa_large_seq16k_sparse         d=0.030 V2 p50= 0.916ms SDPA p50=12.702ms ratio=13.86x
  lcsa_mid_seq8k_very_sparse       d=0.011 V2 p50= 0.484ms SDPA p50= 6.431ms ratio=13.27x

[canonical 4.2 harness] session 'C2' -> docs/methodology/canonical-bench-data.json
