There is a benchmarks suite included in the repository.
(GNU-like compilers only at time of writing.)
Take these benchmarks with a large grain of salt, as actual performance will depend greatly on success of branch prediction / whether the instructions in the jump table are prefetched, and compiler inline decisions will be affected by what the actual visitor is doing.
(From experience, these microbenchmarks are extremely sensitive to the benchmark framework -- marking different components at different steps as opaque to the optimizer has a very large effect. Feel free to poke it and see what I'm talking about.)
But with that in mind, here are some benchmark numbers for visiting ten thousand variants in randomly distributed states, with varying numbers of types in the variant.
Benchmark numbers represent average number of nanoseconds per visit.
gcc 5.4.0
|
Number of types |
2 |
3 |
4 |
5 |
6 |
8 |
10 |
12 |
15 |
18 |
20 |
50 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
|
|
6.481400 |
7.102700 |
7.851000 |
8.150300 |
8.705700 |
9.607100 |
9.010800 |
9.151600 |
9.319300 |
9.435100 |
9.369900 |
N/A |
|
libcxx (dev) |
8.157500 |
10.038500 |
11.234200 |
12.202000 |
12.081600 |
15.154700 |
15.588100 |
13.664700 |
13.287600 |
20.551800 |
19.929300 |
24.191200 |
|
|
0.538400 |
3.010500 |
5.066300 |
7.719000 |
9.092100 |
10.049600 |
11.931200 |
13.531000 |
14.647300 |
14.891000 |
16.392000 |
25.794200 |
|
|
0.676200 |
2.958400 |
3.465800 |
4.423300 |
5.006200 |
6.020200 |
6.966400 |
7.431300 |
7.585400 |
8.796200 |
9.485000 |
13.482700 |
|
|
8.331400 |
9.701500 |
10.550000 |
11.724400 |
12.618300 |
15.397400 |
18.612200 |
16.798500 |
19.801800 |
21.027400 |
21.903400 |
26.065400 |
|
|
6.805700 |
8.315000 |
8.948000 |
9.379700 |
10.019700 |
10.797000 |
10.483200 |
10.682700 |
10.854300 |
10.828700 |
10.955000 |
11.177400 |
|
|
6.744500 |
8.357300 |
9.757800 |
10.487500 |
10.113300 |
10.634800 |
11.095100 |
11.595900 |
12.160900 |
19.551400 |
19.727200 |
22.055700 |
clang 3.8.0
|
Number of types |
2 |
3 |
4 |
5 |
6 |
8 |
10 |
12 |
15 |
18 |
20 |
50 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
|
|
7.477200 |
6.645700 |
6.545900 |
6.628300 |
6.568700 |
6.564100 |
6.889100 |
6.581600 |
6.545600 |
7.651300 |
0.915300 |
N/A |
|
libcxx (dev) |
8.080700 |
10.121900 |
11.084000 |
12.564200 |
12.486700 |
13.248000 |
14.108300 |
14.250700 |
14.377300 |
14.378900 |
14.913100 |
14.945100 |
|
|
0.837900 |
1.889300 |
2.091400 |
4.484100 |
5.044800 |
5.205900 |
7.585900 |
8.154700 |
8.149400 |
10.217100 |
12.028300 |
16.273400 |
|
|
0.865100 |
1.212400 |
2.521900 |
4.265600 |
3.997600 |
5.940000 |
6.685200 |
7.643000 |
8.871700 |
9.998200 |
10.028600 |
14.105200 |
|
|
9.606700 |
10.345300 |
10.603600 |
10.897700 |
13.203500 |
14.954600 |
15.090100 |
17.874800 |
19.452000 |
19.677100 |
19.533400 |
25.147900 |
|
|
6.944900 |
8.718300 |
9.555800 |
10.104900 |
10.443600 |
11.002000 |
11.322000 |
11.809000 |
11.509700 |
11.656900 |
11.696600 |
12.044800 |
|
|
7.009600 |
8.675900 |
9.564000 |
10.066800 |
10.524500 |
11.320100 |
11.098200 |
11.244600 |
11.586100 |
11.549700 |
11.686800 |
11.972000 |
The settings used for these numbers are:
seq_length = 10000 repeat_num = 1000
Test subjects:
boost::variant at version 1.58
eggs::variant from this
repository at commit 1692cb849311cd8dfe77146ad21b4f00299a68cd
juice::variant from this
repository at commit 5572c1b9017f50e9b39d9bf33b1a6719e1983476
mapbox::variant from this
repository at commit 388376ac9f0102feba2d2122873b08e15a66a879
std::experimental::variant from this
repository
std::variant draft at variant
branch from this
repository at commit d93edff042e7b9d333eb5b6f16145953e00fd182
strict_variant from this
repository at commit 94f80fe1cd0d34f8b49268e223f3dfca66605c3a
Test machine /proc/cpuinfo
looks like this:
$ cat /proc/cpuinfo processor : 0 vendor_id : GenuineIntel cpu family : 6 model : 78 model name : Intel(R) Core(TM) i7-6600U CPU @ 2.60GHz stepping : 3 microcode : 0x6a cpu MHz : 438.484 cache size : 4096 KB physical id : 0 siblings : 4 core id : 0 cpu cores : 2 apicid : 0 initial apicid : 0 fpu : yes fpu_exception : yes cpuid level : 22 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc aperfmperf eagerfpu pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch epb intel_pt tpr_shadow vnmi flexpriority ept vpid fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm mpx rdseed adx smap clflushopt xsaveopt xsavec xgetbv1 dtherm ida arat pln pts hwp hwp_notify hwp_act_window hwp_epp bugs : bogomips : 5615.78 clflush size : 64 cache_alignment : 64 address sizes : 39 bits physical, 48 bits virtual power management:
with three additional cores identical to that one.