[mw section4 harness] session=M1
[mw section4 harness] warmup gap: 50ms (downclock threshold: < 100ms)
[mw section4 harness] warmup workload: sparse_attention_nax B=1 H=4 qL=kL=2048 D=64 BT=16 density=0.1
[mw section4 harness] correctness smoke...
  smoke: rmse=5.0998e-08 -> PASS
[mw section4 harness] warmup density_actual=0.101
[mw section4 harness] single warmup dispatch: 3198.5us (target <= 10000us for <= 20% duty cycle)
[mw section4 harness] priming GPU (100 matched-workload dispatches)...
[mw section4 harness] initial cooldown 180.0s (matched-workload-family)
  fired 2995 warmup dispatches during initial cooldown
  lcsa_small_seq4k                 d=0.239 V2=  1.383ms SDPA=  2.704ms ratio= 1.95x drift= 1.3%
  lcsa_small_seq4k_sparse          d=0.067 V2=  1.517ms SDPA=  2.644ms ratio= 1.74x drift= 8.2%
  lcsa_mid_seq8k                   d=0.119 V2=  3.144ms SDPA=  6.621ms ratio= 2.11x drift= 6.4%
  lcsa_mid_seq8k_sparse            d=0.030 V2=  1.758ms SDPA=  6.727ms ratio= 3.83x drift=44.3%
  lcsa_large_seq16k                d=0.120 V2=  3.098ms SDPA= 13.200ms ratio= 4.26x drift=27.0%
  lcsa_large_seq16k_sparse         d=0.030 V2=  1.608ms SDPA= 13.037ms ratio= 8.11x drift=127.5%
  lcsa_mid_seq8k_very_sparse       d=0.011 V2=  1.669ms SDPA=  6.700ms ratio= 4.01x drift=102.6%

[mw section4 harness] session 'M1' -> docs/methodology/matched-workload-data.json
[mw section4 harness] total warmup dispatches: 29962 across 21 cooldown intervals
