Updated Benchmarks

[ghost]

Hello,

Quoting from Benchmark Results

ffjson results are from February 4th, 2016, using the latest ffjson and go1.6. go/codec results are from March 4th, 2016, using the latest go/codec and go1.6.

Would you guys mind updating these numbers? 4 years old benchmark is most likely not accurate any longer.

[luminosoda]

A late answer, but I think it's still relevant. I've launched a benchmark, the results of other libraries have improved, but easyjson is still the fastest. I also found a benchmark for json-iterator/go, don't know why they didn't add jsoniter to the results, but I did it.

Benchmark Results

Unmarshaling

lib	json size	MB/s	allocs/op	B/op
standard	regular	101.13	207	10000
standard	small	38.16	15	672
easyjson	regular	408.37	126	9512
easyjson	small	153.99	3	128
ffjson	regular	271.97	141	9882
ffjson	small	78.15	10	496
codec	regular	333.69	434	12840
codec	small	147.70	7	144
jsoniter	regular	206.20	449	28692
jsoniter	small	69.37	12	384

Marshaling, one goroutine.

lib	json size	MB/s	allocs/op	B/op
standard	regular	522.47	1	9473
standard	small	204.60	1	96
standard	large	508.09	1	450627
easyjson	regular	903.73	10	10303
easyjson*	regular	1085.43	9	721
easyjson	small	504.21	1	128
easyjson	large	934.43	28	457399
easyjson*	large	1157.64	25	2211
ffjson	regular	441.14	151	20815
ffjson**	regular	477.49	150	4415
ffjson	small	157.35	5	384
ffjson**	small	189.49	4	128
ffjson	large	450.38	7212	762209
ffjson**	large	431.58	7214	761326
codec	regular	373.13	9	35200
codec***	regular	511.02	0	0
codec	small	181.18	1	1024
codec***	small	340.84	0	0
codec	large	401.56	24	2298881
codec***	large	519.41	0	1659
jsoniter	regular	149.37	12	29211
jsoniter	small	79.86	4	328
jsoniter	large	152.12	40	1387533

* marshaling to a writer, ** using ffjson.Pool(), *** reusing output slice instead of resetting it to nil

Marshaling, concurrent.

lib	json size	MB/s	allocs/op	B/op
standard	regular	1339.92	1	9474
standard	small	733.19	1	96
standard	large	1579.63	1	450927
easyjson	regular	4519.22	9	10210
easyjson*	regular	4261.50	9	723
easyjson	small	1513.28	1	128
easyjson	large	2957.90	31	462411
easyjson*	large	4314.01	25	2271
ffjson	regular	1343.47	151	20837
ffjson**	regular	1724.99	150	4438
ffjson	small	430.05	5	384
ffjson**	small	563.40	4	128
ffjson	large	1078.97	7217	835397
ffjson**	large	434.00	7214	758387
codec	regular	1168.99	9	35200
codec***	regular	1863.45	0	0
codec	small	443.25	1	1024
codec***	small	1145.33	0	0
codec	large	948.98	24	2298897
codec***	large	1845.09	0	1791
jsoniter	regular	509.79	12	29271
jsoniter	small	277.85	4	328
jsoniter	large	501.28	38	1420690

* marshaling to a writer, ** using ffjson.Pool(), *** reusing output slice instead of resetting it to nil

Here is results of UltraJSON benchmark, I don't know how to translate them to MB/s:

Python ujson module, DECODE: 5000 loops, best of 5: 47.7 usec per loop
Python ujson module, ENCODE: 10000 loops, best of 5: 37.9 usec per loop

[ghost]

@luminosoda thanks for publishing this. Any chance you could add segmentio/encoding as well ?