feat (encoder): add encoder option NoNullSliceOrMap

[AsterDY]

Benchmark

• before

goos: darwin
goarch: amd64
pkg: github.com/bytedance/sonic/encoder
cpu: Intel(R) Core(TM) i9-9880H CPU @ 2.30GHz
BenchmarkOptionSliceOrMapNoNull/true-16                 10000000               246.6 ns/op           286 B/op          5 allocs/op
BenchmarkOptionSliceOrMapNoNull/false-16                10000000               247.4 ns/op           288 B/op          5 allocs/op

• after

goos: darwin
goarch: amd64
pkg: github.com/bytedance/sonic/encoder
cpu: Intel(R) Core(TM) i9-9880H CPU @ 2.30GHz
BenchmarkOptionSliceOrMapNoNull/true-16                 10000000               247.4 ns/op           282 B/op          5 allocs/op
BenchmarkOptionSliceOrMapNoNull/false-16                10000000               236.5 ns/op           283 B/op          5 allocs/op

• conclusion
  no significant effects on performance

• interesting discovery
  writing 4 chars (instruction MOVL) is faster than writing 2 chars (instruction MOVW)?

More Benchmarks

go: added golang.org/x/perf v0.0.0-20220411212318-84e58bfe0a7e
name                                     old time/op    new time/op    delta
MarshalConcrete/CanadaGeometry_Sonic-12     724µs ±10%     712µs ± 7%     ~     (p=0.842 n=10+9)
MarshalConcrete/CitmCatalog_Sonic-12        450µs ± 8%     462µs ±21%     ~     (p=0.968 n=9+10)
MarshalConcrete/StringEscaped_Sonic-12     4.53µs ±23%    4.84µs ±12%     ~     (p=0.095 n=9+10)
MarshalConcrete/StringUnicode_Sonic-12     4.95µs ±30%    5.77µs ±40%     ~     (p=0.063 n=10+10)
MarshalConcrete/TwitterStatus_Sonic-12      309µs ±19%     281µs ±20%     ~     (p=0.075 n=10+10)
MarshalConcrete/SyntheaFhir_Sonic-12       8.96ms ± 3%    8.82ms ± 2%   -1.55%  (p=0.036 n=8+9)
MarshalConcrete/GolangSource_Sonic-12      2.29ms ± 7%    2.06ms ± 9%  -10.06%  (p=0.000 n=9+10)

name                                     old speed      new speed      delta
MarshalConcrete/CanadaGeometry_Sonic-12   375MB/s ± 9%   380MB/s ± 6%     ~     (p=0.842 n=10+9)
MarshalConcrete/CitmCatalog_Sonic-12     3.78GB/s ±16%  3.78GB/s ±18%     ~     (p=0.853 n=10+10)
MarshalConcrete/StringEscaped_Sonic-12   9.10GB/s ±29%  8.74GB/s ±12%     ~     (p=0.247 n=10+10)
MarshalConcrete/StringUnicode_Sonic-12   3.74GB/s ±25%  3.24GB/s ±31%     ~     (p=0.063 n=10+10)
MarshalConcrete/TwitterStatus_Sonic-12   2.07GB/s ±17%  2.26GB/s ±17%     ~     (p=0.075 n=10+10)
MarshalConcrete/SyntheaFhir_Sonic-12      224MB/s ± 3%   228MB/s ± 2%   +1.57%  (p=0.036 n=8+9)
MarshalConcrete/GolangSource_Sonic-12     849MB/s ± 8%   945MB/s ± 9%  +11.26%  (p=0.000 n=9+10)