CPU scaling benchmark

W=0 W=1 W=2 W=3 W=4 W=6 W=8 W=9 W=10 W=12 W=16 iters: 100M 500M 1B 2B
workers
16 +1 main
iters total
500M
29411764/stream
elapsed
1209.99 ms
total CPU used
16671.39 ms
speedup
13.78×
vs serial
efficiency
81.1%
of 17× ideal
stream spawn ms spawned@ work start@ work end@ work ms reap wait ms
0 (main) 0 116.8 116.81 1132.91 1016.1 0
1 2.114 2.13 20.55 1164.65 1144.1 31.85
2 1.59 3.74 18.04 1071.46 1053.42 0.16
3 1.578 5.34 24.59 1160.93 1136.34 28.14
4 5.487 10.85 36.7 1178.37 1141.67 49.62
5 1.707 12.58 53.65 1199.64 1145.99 66.84
6 1.472 14.07 45.45 1143.33 1097.88 13.75
7 1.468 15.56 49.87 1075.51 1025.64 0.23
8 8.94 24.52 68.93 748.21 679.28 0.24
9 9.66 34.21 91.37 1206.07 1114.7 73.28
10 9.263 43.49 99.11 1206.46 1107.35 76.25
11 2.015 45.52 96.49 1177.18 1080.69 50.95
12 12.118 57.66 134.68 994.47 859.79 0.25
13 2.187 59.87 99.35 873.67 774.32 0.27
14 28.281 88.17 131.17 1105.58 974.41 0.29
15 19.059 107.25 140.12 562.99 422.87 0.32
16 9.503 116.77 166.67 1063.51 896.84 0.33
main
w1
w2
w3
w4
w5
w6
w7
w8
w9
w10
w11
w12
w13
w14
w15
w16
    fork+handshake      CPU work      parent reap wait
what this measures
Each stream runs a tight integer LCG loop — working set is one CPU register, no memory access, no shared data. Speedup = sum(stream CPU time) / wall-clock elapsed. Efficiency = speedup / (workers+1). 100% efficiency means perfect linear scaling; less than 100% is the cost of serial fork setup, reap tail, SMT/core contention.