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
9 +1 main
iters total
500M
50000000/stream
elapsed
1177.28 ms
total CPU used
10274.23 ms
speedup
8.73×
vs serial
efficiency
87.3%
of 10× ideal
stream spawn ms spawned@ work start@ work end@ work ms reap wait ms
0 (main) 0 24.83 24.84 1155.04 1130.2 0
1 2.36 2.38 20.21 1008.56 988.35 0.14
2 1.735 4.14 21.09 963.75 942.66 0.22
3 1.772 5.93 52.49 1017.39 964.9 0.23
4 3.232 9.18 52.43 1158.96 1106.53 4.03
5 3.022 12.21 59.59 1060.6 1001.01 0.24
6 1.813 14.04 74.78 1097.11 1022.33 0.26
7 1.698 15.76 59.66 1051.7 992.04 0.27
8 6.91 22.68 58.05 1090.46 1032.41 0.29
9 2.118 24.81 79.66 1173.46 1093.8 18.56
main
w1
w2
w3
w4
w5
w6
w7
w8
w9
    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.