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
1169.48 ms
total CPU used
10302.8 ms
speedup
8.81×
vs serial
efficiency
88.1%
of 10× ideal
stream spawn ms spawned@ work start@ work end@ work ms reap wait ms
0 (main) 0 15.72 15.73 1150.24 1134.51 0
1 2.22 2.23 22.84 823.35 800.51 0.13
2 1.657 3.91 45.3 1156.7 1111.4 6.59
3 1.632 5.56 31.73 1026.4 994.67 0.19
4 1.647 7.22 30.03 1105.74 1075.71 0.2
5 1.719 8.96 43.41 1019.74 976.33 0.21
6 1.562 10.53 45.74 1097.54 1051.8 0.23
7 1.596 12.14 54.09 1090.25 1036.16 0.24
8 1.949 14.11 60.53 1079.16 1018.63 0.25
9 1.596 15.71 63.36 1166.44 1103.08 16.33
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.