cpu test: single thread and multi-threads get the same result

[chengweiv5]

Hi there, I was using sysbench 0.4.12 to do cpu testing, use below command in an 8 vcpu VM

1. for single thread
   # sysbench --test=cpu --num-threads=1 --cpu-max-prime=50000 run

2. for 8 threads
   # sysbench --test=cpu --num-threads=8 --cpu-max-prime=50000 run

and from the total time: , I could see that 8 threads is nearly 8 times less than 1 threads, e.g. 2s VS 15s.

However, when I upgrade to sysbench 1.0.5, the results for 1 thread and 8 threads are seems the same, e.g. 10.0117s VS 10.0321s

Is this a problem?

[chengweiv5]

In addition, I found that, for sysbench 1.0.5, 10000 primes to 100000 primes just get the same result, see below

10000 primes

WARNING: the --test option is deprecated. You can pass a script name or path on the command line without any options.
sysbench 1.0.5 (using system LuaJIT 2.0.4)

Running the test with following options:
Number of threads: 1
Initializing random number generator from current time


Prime numbers limit: 10000

Initializing worker threads...

Threads started!


General statistics:
    total time:                          10.0009s
    total number of events:              7384

Latency (ms):
         min:                                  1.21
         avg:                                  1.35
         max:                                  1.76
         95th percentile:                      1.44
         sum:                               9992.83

Threads fairness:
    events (avg/stddev):           7384.0000/0.00
    execution time (avg/stddev):   9.9928/0.00

100000 primes

WARNING: the --test option is deprecated. You can pass a script name or path on the command line without any options.
sysbench 1.0.5 (using system LuaJIT 2.0.4)

Running the test with following options:
Number of threads: 1
Initializing random number generator from current time


Prime numbers limit: 100000

Initializing worker threads...

Threads started!


General statistics:
    total time:                          10.0313s
    total number of events:              296

Latency (ms):
         min:                                 32.92
         avg:                                 33.89
         max:                                 35.20
         95th percentile:                     34.95
         sum:                              10030.37

Threads fairness:
    events (avg/stddev):           296.0000/0.00
    execution time (avg/stddev):   10.0304/0.00

[chengweiv5]

even 1000000 primes

WARNING: the --test option is deprecated. You can pass a script name or path on the command line without any options.
sysbench 1.0.5 (using system LuaJIT 2.0.4)

Running the test with following options:
Number of threads: 1
Initializing random number generator from current time


Prime numbers limit: 1000000

Initializing worker threads...

Threads started!


General statistics:
    total time:                          10.0336s
    total number of events:              12

Latency (ms):
         min:                                807.62
         avg:                                836.11
         max:                                860.82
         95th percentile:                    861.95
         sum:                              10033.26

Threads fairness:
    events (avg/stddev):           12.0000/0.00
    execution time (avg/stddev):   10.0333/0.00

[akopytov]

Hi, I answered a similar question recently in #140. Do my comments in that issue answer your question?

I'm closing this one, feel free to reopen if this question looks different from #140.

[Willz01]

2 years lol... this just saved me my lecturer gave this ""sysbench --test=cpu --cpu -max -prime =1000 –num -threads =1 run" seem he misplaced the command, thanks!