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
1170.37 ms
total CPU used
9996.75 ms
speedup
8.54×
vs serial
efficiency
85.4%
of 10× ideal
stream spawn ms spawned@ work start@ work end@ work ms reap wait ms
0 (main) 0 15.16 15.39 1033.56 1018.17 0
1 1.956 1.97 19.4 1136.95 1117.55 103.53
2 1.624 3.62 28.59 820.08 791.49 0.14
3 1.781 5.42 20.17 913.5 893.33 0.2
4 1.708 7.15 61.07 1043.14 982.07 9.71
5 1.628 8.8 46.71 1102.95 1056.24 72.38
6 1.522 10.33 51.06 971.38 920.32 0.22
7 1.621 11.97 48.14 1121.71 1073.57 90.51
8 1.636 13.62 78.29 1167.44 1089.15 134.01
9 1.515 15.15 61.13 1115.99 1054.86 85.25
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.