go-mc
go-mc copied to clipboard
Published
20 hours ago •
Tnze
Tnze
some enhancement for Mojang login and some performance enhancement in packet level
As mentioned in #209,
I add KeyManager for caching RSA keypair,
not yet add the option for prevent-proxy-connections: true right now.
And this PR also include some performance enhancement in packet level, which:
- use
sync.Poolforbytes.NewReader()inPacket.Scan() - switch standard library
compress/zlibto third partygithub.com/klauspost/compress/zlib - add option for setting compress level
I don't believe the first change can improve the performance.
bytes.NewReader doesn't allocate any memory on heap. Using sync.Pool in this case only cause the variables escape from stack(or registers) to heap, which decreases the performance actually.
And I don't like to switch the zlib implementation, unless you can prove it's necessary and the third-party library is reliable.
(I means may be it has performance improvement but also vulnerabilities)
I don't believe the first change can improve the performance.
bytes.NewReaderdoesn't allocate any memory on heap. Usingsync.Poolin this case only cause the variables escape from stack(or registers) to heap, which decreases the performance actually.
I will try to make a benchmark, and check this one.
And I don't like to switch the
zlibimplementation, unless you can prove it's necessary and the third-party library is reliable.(I means may be it has performance improvement but also vulnerabilities)
I had add a little benchmark code for this,
and I think if we only use zlib in standard library,
then we need to use BestSpeed (level 1) as default.
ps. I'm building a L7 proxy which can filter out malicious traffic (like WAF in web), need low level access/modify on packet level, neither BungeeCord nor Velocity can do this. So the performance in packet level are important to me.
here is the benchmark result:
standard library with default compress level (5):
$ go test -bench=. -benchmem -cpu=1,2,4,8 -run=^#
goos: linux
goarch: amd64
pkg: github.com/Tnze/go-mc/net
cpu: Intel(R) Core(TM) i3-10100 CPU @ 3.60GHz
BenchmarkConnTCPCompress 2467 475211 ns/op 275.82 MB/s 40945 B/op 13 allocs/op
BenchmarkConnTCPCompress-2 3016 396766 ns/op 330.35 MB/s 40888 B/op 13 allocs/op
BenchmarkConnTCPCompress-4 2979 404220 ns/op 324.26 MB/s 40912 B/op 13 allocs/op
BenchmarkConnTCPCompress-8 2907 406397 ns/op 322.52 MB/s 40940 B/op 13 allocs/op
BenchmarkConnTCPCompressLevelDefault 2462 469518 ns/op 279.16 MB/s 40945 B/op 13 allocs/op
BenchmarkConnTCPCompressLevelDefault-2 3058 412957 ns/op 317.40 MB/s 40885 B/op 13 allocs/op
BenchmarkConnTCPCompressLevelDefault-4 2978 405987 ns/op 322.85 MB/s 40913 B/op 13 allocs/op
BenchmarkConnTCPCompressLevelDefault-8 2928 402088 ns/op 325.98 MB/s 40939 B/op 13 allocs/op
BenchmarkConnTCPCompressLevelBestSpeed 5551 193154 ns/op 678.59 MB/s 40833 B/op 13 allocs/op
BenchmarkConnTCPCompressLevelBestSpeed-2 9980 116114 ns/op 1128.82 MB/s 40741 B/op 13 allocs/op
BenchmarkConnTCPCompressLevelBestSpeed-4 9666 129276 ns/op 1013.89 MB/s 40768 B/op 13 allocs/op
BenchmarkConnTCPCompressLevelBestSpeed-8 9777 130696 ns/op 1002.88 MB/s 40788 B/op 13 allocs/op
BenchmarkConnTCPNoCompress 50786 23776 ns/op 5512.86 MB/s 14 B/op 6 allocs/op
BenchmarkConnTCPNoCompress-2 51469 22816 ns/op 5744.70 MB/s 14 B/op 6 allocs/op
BenchmarkConnTCPNoCompress-4 52388 23631 ns/op 5546.65 MB/s 14 B/op 6 allocs/op
BenchmarkConnTCPNoCompress-8 51853 22698 ns/op 5774.52 MB/s 14 B/op 6 allocs/op
BenchmarkConnPipeCompress 2565 464185 ns/op 282.37 MB/s 40932 B/op 13 allocs/op
BenchmarkConnPipeCompress-2 2551 467944 ns/op 280.10 MB/s 40954 B/op 13 allocs/op
BenchmarkConnPipeCompress-4 2520 466229 ns/op 281.13 MB/s 40988 B/op 13 allocs/op
BenchmarkConnPipeCompress-8 2454 468174 ns/op 279.96 MB/s 41026 B/op 13 allocs/op
BenchmarkConnPipeCompressLevelDefault 2503 462964 ns/op 283.12 MB/s 40940 B/op 13 allocs/op
BenchmarkConnPipeCompressLevelDefault-2 2505 493891 ns/op 265.39 MB/s 40954 B/op 13 allocs/op
BenchmarkConnPipeCompressLevelDefault-4 2379 495828 ns/op 264.35 MB/s 41007 B/op 13 allocs/op
BenchmarkConnPipeCompressLevelDefault-8 2395 493343 ns/op 265.68 MB/s 41034 B/op 13 allocs/op
BenchmarkConnPipeCompressLevelBestSpeed 6310 189067 ns/op 693.26 MB/s 40809 B/op 13 allocs/op
BenchmarkConnPipeCompressLevelBestSpeed-2 6751 170305 ns/op 769.63 MB/s 40807 B/op 13 allocs/op
BenchmarkConnPipeCompressLevelBestSpeed-4 7082 159598 ns/op 821.27 MB/s 40829 B/op 13 allocs/op
BenchmarkConnPipeCompressLevelBestSpeed-8 6976 161709 ns/op 810.54 MB/s 40845 B/op 13 allocs/op
BenchmarkConnPipeNoCompress 161882 7238 ns/op 18107.81 MB/s 16 B/op 6 allocs/op
BenchmarkConnPipeNoCompress-2 151468 7982 ns/op 16421.27 MB/s 16 B/op 6 allocs/op
BenchmarkConnPipeNoCompress-4 154296 7718 ns/op 16983.31 MB/s 16 B/op 6 allocs/op
BenchmarkConnPipeNoCompress-8 150906 7898 ns/op 16595.25 MB/s 16 B/op 6 allocs/op
PASS
ok github.com/Tnze/go-mc/net 40.928s
$ go test -bench=. -benchmem -cpu=1,2,4,8,16 -run=^#
goos: linux
goarch: amd64
pkg: github.com/Tnze/go-mc/net
cpu: AMD Ryzen 7 5700G with Radeon Graphics
BenchmarkConnTCPCompress 2791 395030 ns/op 331.80 MB/s 40907 B/op 13 allocs/op
BenchmarkConnTCPCompress-2 2863 446081 ns/op 293.83 MB/s 40903 B/op 13 allocs/op
BenchmarkConnTCPCompress-4 2594 468913 ns/op 279.52 MB/s 40958 B/op 13 allocs/op
BenchmarkConnTCPCompress-8 3438 368327 ns/op 355.86 MB/s 40894 B/op 13 allocs/op
BenchmarkConnTCPCompress-16 3142 372094 ns/op 352.26 MB/s 40949 B/op 13 allocs/op
BenchmarkConnTCPCompressLevelDefault 2961 391298 ns/op 334.97 MB/s 40890 B/op 13 allocs/op
BenchmarkConnTCPCompressLevelDefault-2 3219 363205 ns/op 360.88 MB/s 40871 B/op 13 allocs/op
BenchmarkConnTCPCompressLevelDefault-4 3363 366926 ns/op 357.22 MB/s 40885 B/op 13 allocs/op
BenchmarkConnTCPCompressLevelDefault-8 2874 358037 ns/op 366.08 MB/s 40939 B/op 13 allocs/op
BenchmarkConnTCPCompressLevelDefault-16 3571 373549 ns/op 350.88 MB/s 40920 B/op 13 allocs/op
BenchmarkConnTCPCompressLevelBestSpeed 6598 156948 ns/op 835.13 MB/s 40798 B/op 13 allocs/op
BenchmarkConnTCPCompressLevelBestSpeed-2 9990 103465 ns/op 1266.82 MB/s 40743 B/op 13 allocs/op
BenchmarkConnTCPCompressLevelBestSpeed-4 10189 115331 ns/op 1136.49 MB/s 40762 B/op 13 allocs/op
BenchmarkConnTCPCompressLevelBestSpeed-8 9100 120232 ns/op 1090.16 MB/s 40796 B/op 13 allocs/op
BenchmarkConnTCPCompressLevelBestSpeed-16 10000 124585 ns/op 1052.07 MB/s 40823 B/op 13 allocs/op
BenchmarkConnTCPNoCompress 40892 29392 ns/op 4459.49 MB/s 14 B/op 6 allocs/op
BenchmarkConnTCPNoCompress-2 41454 33862 ns/op 3870.81 MB/s 14 B/op 6 allocs/op
BenchmarkConnTCPNoCompress-4 34868 29148 ns/op 4496.77 MB/s 14 B/op 6 allocs/op
BenchmarkConnTCPNoCompress-8 42120 32508 ns/op 4032.00 MB/s 14 B/op 6 allocs/op
BenchmarkConnTCPNoCompress-16 33949 35599 ns/op 3681.88 MB/s 14 B/op 6 allocs/op
BenchmarkConnPipeCompress 2709 394220 ns/op 332.48 MB/s 40915 B/op 13 allocs/op
BenchmarkConnPipeCompress-2 2755 436481 ns/op 300.29 MB/s 40918 B/op 13 allocs/op
BenchmarkConnPipeCompress-4 2432 495604 ns/op 264.47 MB/s 40995 B/op 13 allocs/op
BenchmarkConnPipeCompress-8 2388 515409 ns/op 254.31 MB/s 41026 B/op 13 allocs/op
BenchmarkConnPipeCompress-16 2497 515602 ns/op 254.21 MB/s 41048 B/op 13 allocs/op
BenchmarkConnPipeCompressLevelDefault 2829 388350 ns/op 337.51 MB/s 40903 B/op 13 allocs/op
BenchmarkConnPipeCompressLevelDefault-2 2659 427471 ns/op 306.62 MB/s 40929 B/op 13 allocs/op
BenchmarkConnPipeCompressLevelDefault-4 2377 501661 ns/op 261.28 MB/s 40998 B/op 13 allocs/op
BenchmarkConnPipeCompressLevelDefault-8 2415 502988 ns/op 260.59 MB/s 41024 B/op 13 allocs/op
BenchmarkConnPipeCompressLevelDefault-16 2451 503161 ns/op 260.50 MB/s 41052 B/op 13 allocs/op
BenchmarkConnPipeCompressLevelBestSpeed 6332 160025 ns/op 819.07 MB/s 40809 B/op 13 allocs/op
BenchmarkConnPipeCompressLevelBestSpeed-2 6945 160985 ns/op 814.19 MB/s 40803 B/op 13 allocs/op
BenchmarkConnPipeCompressLevelBestSpeed-4 5392 187860 ns/op 697.71 MB/s 40886 B/op 13 allocs/op
BenchmarkConnPipeCompressLevelBestSpeed-8 6722 196734 ns/op 666.24 MB/s 40866 B/op 13 allocs/op
BenchmarkConnPipeCompressLevelBestSpeed-16 5944 198636 ns/op 659.86 MB/s 40929 B/op 13 allocs/op
BenchmarkConnPipeNoCompress 209176 4978 ns/op 26330.02 MB/s 16 B/op 6 allocs/op
BenchmarkConnPipeNoCompress-2 166504 6128 ns/op 21388.26 MB/s 15 B/op 6 allocs/op
BenchmarkConnPipeNoCompress-4 175752 6893 ns/op 19016.30 MB/s 16 B/op 6 allocs/op
BenchmarkConnPipeNoCompress-8 183888 6701 ns/op 19558.66 MB/s 16 B/op 6 allocs/op
BenchmarkConnPipeNoCompress-16 174505 6399 ns/op 20484.39 MB/s 16 B/op 6 allocs/op
PASS
ok github.com/Tnze/go-mc/net 51.743s
third-party library with default compress level (5):
$ go test -bench=. -benchmem -cpu=1,2,4,8 -run=^#
goos: linux
goarch: amd64
pkg: github.com/Tnze/go-mc/net
cpu: Intel(R) Core(TM) i3-10100 CPU @ 3.60GHz
BenchmarkConnTCPCompress 7154 165390 ns/op 792.50 MB/s 38317 B/op 18 allocs/op
BenchmarkConnTCPCompress-2 13526 88940 ns/op 1473.71 MB/s 38250 B/op 18 allocs/op
BenchmarkConnTCPCompress-4 12187 97644 ns/op 1342.34 MB/s 38275 B/op 18 allocs/op
BenchmarkConnTCPCompress-8 12138 101844 ns/op 1286.98 MB/s 38292 B/op 18 allocs/op
BenchmarkConnTCPCompressLevelDefault 7302 165953 ns/op 789.82 MB/s 38314 B/op 18 allocs/op
BenchmarkConnTCPCompressLevelDefault-2 13072 96233 ns/op 1362.03 MB/s 38253 B/op 18 allocs/op
BenchmarkConnTCPCompressLevelDefault-4 12362 101187 ns/op 1295.35 MB/s 38274 B/op 18 allocs/op
BenchmarkConnTCPCompressLevelDefault-8 10000 100154 ns/op 1308.71 MB/s 38309 B/op 18 allocs/op
BenchmarkConnTCPCompressLevelBestSpeed 6423 170282 ns/op 769.74 MB/s 38293 B/op 18 allocs/op
BenchmarkConnTCPCompressLevelBestSpeed-2 12848 92922 ns/op 1410.56 MB/s 38233 B/op 18 allocs/op
BenchmarkConnTCPCompressLevelBestSpeed-4 10000 103129 ns/op 1270.95 MB/s 38266 B/op 18 allocs/op
BenchmarkConnTCPCompressLevelBestSpeed-8 10000 103585 ns/op 1265.36 MB/s 38282 B/op 18 allocs/op
BenchmarkConnTCPNoCompress 50139 24010 ns/op 5459.06 MB/s 14 B/op 6 allocs/op
BenchmarkConnTCPNoCompress-2 53199 23249 ns/op 5637.85 MB/s 14 B/op 6 allocs/op
BenchmarkConnTCPNoCompress-4 53641 22466 ns/op 5834.22 MB/s 14 B/op 6 allocs/op
BenchmarkConnTCPNoCompress-8 52851 22399 ns/op 5851.59 MB/s 14 B/op 6 allocs/op
BenchmarkConnPipeCompress 7042 164710 ns/op 795.77 MB/s 38320 B/op 18 allocs/op
BenchmarkConnPipeCompress-2 7897 142726 ns/op 918.34 MB/s 38319 B/op 18 allocs/op
BenchmarkConnPipeCompress-4 8050 142635 ns/op 918.93 MB/s 38329 B/op 18 allocs/op
BenchmarkConnPipeCompress-8 7992 145220 ns/op 902.58 MB/s 38349 B/op 18 allocs/op
BenchmarkConnPipeCompressLevelDefault 7137 164062 ns/op 798.92 MB/s 38318 B/op 18 allocs/op
BenchmarkConnPipeCompressLevelDefault-2 7773 141402 ns/op 926.94 MB/s 38320 B/op 18 allocs/op
BenchmarkConnPipeCompressLevelDefault-4 7947 142868 ns/op 917.43 MB/s 38332 B/op 18 allocs/op
BenchmarkConnPipeCompressLevelDefault-8 7916 146142 ns/op 896.88 MB/s 38346 B/op 18 allocs/op
BenchmarkConnPipeCompressLevelBestSpeed 7095 165301 ns/op 792.93 MB/s 38282 B/op 18 allocs/op
BenchmarkConnPipeCompressLevelBestSpeed-2 7699 144025 ns/op 910.06 MB/s 38289 B/op 18 allocs/op
BenchmarkConnPipeCompressLevelBestSpeed-4 7833 142418 ns/op 920.33 MB/s 38296 B/op 18 allocs/op
BenchmarkConnPipeCompressLevelBestSpeed-8 7958 145914 ns/op 898.28 MB/s 38313 B/op 18 allocs/op
BenchmarkConnPipeNoCompress 160188 7353 ns/op 17825.17 MB/s 16 B/op 6 allocs/op
BenchmarkConnPipeNoCompress-2 154214 7493 ns/op 17492.17 MB/s 16 B/op 6 allocs/op
BenchmarkConnPipeNoCompress-4 159274 7474 ns/op 17538.11 MB/s 16 B/op 6 allocs/op
BenchmarkConnPipeNoCompress-8 153895 7903 ns/op 16585.19 MB/s 16 B/op 6 allocs/op
PASS
ok github.com/Tnze/go-mc/net 44.223s
$ go test -bench=. -benchmem -cpu=1,2,4,8,16 -run=^#
goos: linux
goarch: amd64
pkg: github.com/Tnze/go-mc/net
cpu: AMD Ryzen 7 5700G with Radeon Graphics
BenchmarkConnTCPCompress 8784 133766 ns/op 979.86 MB/s 38288 B/op 18 allocs/op
BenchmarkConnTCPCompress-2 14545 83160 ns/op 1576.14 MB/s 38245 B/op 18 allocs/op
BenchmarkConnTCPCompress-4 12916 88339 ns/op 1483.74 MB/s 38264 B/op 18 allocs/op
BenchmarkConnTCPCompress-8 12808 91455 ns/op 1433.19 MB/s 38282 B/op 18 allocs/op
BenchmarkConnTCPCompress-16 12381 94965 ns/op 1380.21 MB/s 38316 B/op 18 allocs/op
BenchmarkConnTCPCompressLevelDefault 8824 133798 ns/op 979.63 MB/s 38288 B/op 18 allocs/op
BenchmarkConnTCPCompressLevelDefault-2 14373 80770 ns/op 1622.79 MB/s 38246 B/op 18 allocs/op
BenchmarkConnTCPCompressLevelDefault-4 13879 89441 ns/op 1465.46 MB/s 38261 B/op 18 allocs/op
BenchmarkConnTCPCompressLevelDefault-8 12514 93520 ns/op 1401.54 MB/s 38287 B/op 18 allocs/op
BenchmarkConnTCPCompressLevelDefault-16 12661 95047 ns/op 1379.03 MB/s 38317 B/op 18 allocs/op
BenchmarkConnTCPCompressLevelBestSpeed 8616 138932 ns/op 943.42 MB/s 38260 B/op 18 allocs/op
BenchmarkConnTCPCompressLevelBestSpeed-2 13503 81945 ns/op 1599.52 MB/s 38230 B/op 18 allocs/op
BenchmarkConnTCPCompressLevelBestSpeed-4 12320 94841 ns/op 1382.02 MB/s 38250 B/op 18 allocs/op
BenchmarkConnTCPCompressLevelBestSpeed-8 10000 104422 ns/op 1255.21 MB/s 38281 B/op 18 allocs/op
BenchmarkConnTCPCompressLevelBestSpeed-16 10000 106485 ns/op 1230.90 MB/s 38311 B/op 18 allocs/op
BenchmarkConnTCPNoCompress 42501 29066 ns/op 4509.52 MB/s 14 B/op 6 allocs/op
BenchmarkConnTCPNoCompress-2 32374 32985 ns/op 3973.70 MB/s 14 B/op 6 allocs/op
BenchmarkConnTCPNoCompress-4 34548 29363 ns/op 4463.91 MB/s 14 B/op 6 allocs/op
BenchmarkConnTCPNoCompress-8 38563 27726 ns/op 4727.41 MB/s 14 B/op 6 allocs/op
BenchmarkConnTCPNoCompress-16 38089 27333 ns/op 4795.42 MB/s 14 B/op 6 allocs/op
BenchmarkConnPipeCompress 8316 138462 ns/op 946.63 MB/s 38298 B/op 18 allocs/op
BenchmarkConnPipeCompress-2 6968 150723 ns/op 869.62 MB/s 38340 B/op 18 allocs/op
BenchmarkConnPipeCompress-4 7059 167589 ns/op 782.11 MB/s 38359 B/op 18 allocs/op
BenchmarkConnPipeCompress-8 6238 175174 ns/op 748.24 MB/s 38400 B/op 18 allocs/op
BenchmarkConnPipeCompress-16 6588 176193 ns/op 743.91 MB/s 38426 B/op 18 allocs/op
BenchmarkConnPipeCompressLevelDefault 7267 141580 ns/op 925.78 MB/s 38317 B/op 18 allocs/op
BenchmarkConnPipeCompressLevelDefault-2 7839 144355 ns/op 907.99 MB/s 38322 B/op 18 allocs/op
BenchmarkConnPipeCompressLevelDefault-4 6847 171854 ns/op 762.69 MB/s 38365 B/op 18 allocs/op
BenchmarkConnPipeCompressLevelDefault-8 5984 174048 ns/op 753.08 MB/s 38408 B/op 18 allocs/op
BenchmarkConnPipeCompressLevelDefault-16 7562 172656 ns/op 759.15 MB/s 38404 B/op 18 allocs/op
BenchmarkConnPipeCompressLevelBestSpeed 7066 148212 ns/op 884.35 MB/s 38282 B/op 18 allocs/op
BenchmarkConnPipeCompressLevelBestSpeed-2 7628 148497 ns/op 882.66 MB/s 38287 B/op 18 allocs/op
BenchmarkConnPipeCompressLevelBestSpeed-4 7167 169025 ns/op 775.46 MB/s 38321 B/op 18 allocs/op
BenchmarkConnPipeCompressLevelBestSpeed-8 6714 174141 ns/op 752.68 MB/s 38349 B/op 18 allocs/op
BenchmarkConnPipeCompressLevelBestSpeed-16 7598 177958 ns/op 736.53 MB/s 38363 B/op 18 allocs/op
BenchmarkConnPipeNoCompress 221833 5095 ns/op 25723.46 MB/s 16 B/op 6 allocs/op
BenchmarkConnPipeNoCompress-2 168592 6419 ns/op 20418.38 MB/s 16 B/op 6 allocs/op
BenchmarkConnPipeNoCompress-4 178712 6580 ns/op 19918.43 MB/s 16 B/op 6 allocs/op
BenchmarkConnPipeNoCompress-8 167635 7047 ns/op 18600.16 MB/s 16 B/op 6 allocs/op
BenchmarkConnPipeNoCompress-16 160638 6837 ns/op 19172.32 MB/s 16 B/op 6 allocs/op
PASS
ok github.com/Tnze/go-mc/net 61.470s
And I don't like to switch the
zlibimplementation, unless you can prove it's necessary and the third-party library is reliable.(I means may be it has performance improvement but also vulnerabilities)
In theory, we cannot 100% prove/promise that a library will not has vulnerability, even already done code review/fuzz test...etc. And "klauspost/compress" seems using the same test code in standard library (even more test). So I think it should be almost same reliable as standard library if we could trust standard library.
Or we can add option for using third-party library or standard library? for building server, use third-party library for the packet send to client, use standard library read from client. for building client, use third-party library for the packet send to server, use standard library read from sever.
I don't believe the first change can improve the performance.
bytes.NewReaderdoesn't allocate any memory on heap. Usingsync.Poolin this case only cause the variables escape from stack(or registers) to heap, which decreases the performance actually.
Actually only save 1 allocate, and seems performance nearly same...
I add this code in net/packet/packet_test.go
func BenchmarkPacketScan(b *testing.B) {
p := pk.Packet{ID: 0x24, Data: testJoinGameData}
var (
EID pk.Int
Hardcore pk.Boolean
Gamemode pk.UnsignedByte
PreGamemode pk.Byte
WorldNames = make([]pk.Identifier, 0) // This cannot replace with "var DimensionNames []pk.Identifier" because "nil" has no type information
DimensionCodec struct {
DimensionType interface{} `nbt:"minecraft:dimension_type"`
WorldgenBiome interface{} `nbt:"minecraft:worldgen/biome"`
}
Dimension interface{}
WorldName pk.Identifier
HashedSeed pk.Long
MaxPlayers pk.VarInt
ViewDistance pk.VarInt
RDI, ERS, IsDebug, IsFlat pk.Boolean
)
b.SetBytes(int64(len(p.Data)))
b.ResetTimer()
b.ReportAllocs()
for i := 0; i < b.N; i++ {
err := p.Scan(
&EID,
&Hardcore,
&Gamemode,
&PreGamemode,
pk.Array(&WorldNames),
pk.NBT(&DimensionCodec),
pk.NBT(&Dimension),
&WorldName,
&HashedSeed,
&MaxPlayers,
&ViewDistance,
&RDI, &ERS, &IsDebug, &IsFlat,
)
if err != nil {
b.Fatalf("err = %d; want <nil>", err)
}
}
}
benchmark result:
without pool:
$ go test -bench=. -benchmem -cpu=1,2,4,8 -run=^#
goos: linux
goarch: amd64
pkg: github.com/Tnze/go-mc/net/packet
cpu: Intel(R) Core(TM) i3-10100 CPU @ 3.60GHz
BenchmarkPacketScan 2385 489661 ns/op 63.00 MB/s 224318 B/op 12713 allocs/op
BenchmarkPacketScan-2 2406 487766 ns/op 63.24 MB/s 224336 B/op 12713 allocs/op
BenchmarkPacketScan-4 2433 496988 ns/op 62.07 MB/s 224325 B/op 12713 allocs/op
BenchmarkPacketScan-8 2402 493389 ns/op 62.52 MB/s 224336 B/op 12713 allocs/op
PASS
ok github.com/Tnze/go-mc/net/packet 4.945s
$ go test -bench=. -benchmem -cpu=1,2,4,8,16 -run=^#
goos: linux
goarch: amd64
pkg: github.com/Tnze/go-mc/net/packet
cpu: AMD Ryzen 7 5700G with Radeon Graphics
BenchmarkPacketScan 2367 465953 ns/op 66.20 MB/s 224331 B/op 12713 allocs/op
BenchmarkPacketScan-2 2185 486443 ns/op 63.41 MB/s 224340 B/op 12713 allocs/op
BenchmarkPacketScan-4 2139 498789 ns/op 61.84 MB/s 224335 B/op 12713 allocs/op
BenchmarkPacketScan-8 2175 524318 ns/op 58.83 MB/s 224333 B/op 12713 allocs/op
BenchmarkPacketScan-16 2128 529007 ns/op 58.31 MB/s 224338 B/op 12713 allocs/op
PASS
ok github.com/Tnze/go-mc/net/packet 5.789s
with pool:
$ go test -bench=. -benchmem -cpu=1,2,4,8 -run=^#
goos: linux
goarch: amd64
pkg: github.com/Tnze/go-mc/net/packet
cpu: Intel(R) Core(TM) i3-10100 CPU @ 3.60GHz
BenchmarkPacketScan 2392 491476 ns/op 62.76 MB/s 224283 B/op 12712 allocs/op
BenchmarkPacketScan-2 2421 485727 ns/op 63.51 MB/s 224301 B/op 12712 allocs/op
BenchmarkPacketScan-4 2383 493056 ns/op 62.56 MB/s 224308 B/op 12712 allocs/op
BenchmarkPacketScan-8 2386 505003 ns/op 61.08 MB/s 224354 B/op 12712 allocs/op
PASS
ok github.com/Tnze/go-mc/net/packet 4.942s
$ go test -bench=. -benchmem -cpu=1,2,4,8,16 -run=^#
goos: linux
goarch: amd64
pkg: github.com/Tnze/go-mc/net/packet
cpu: AMD Ryzen 7 5700G with Radeon Graphics
BenchmarkPacketScan 2390 456629 ns/op 67.55 MB/s 224288 B/op 12712 allocs/op
BenchmarkPacketScan-2 2342 482222 ns/op 63.97 MB/s 224307 B/op 12712 allocs/op
BenchmarkPacketScan-4 2216 522766 ns/op 59.01 MB/s 224313 B/op 12712 allocs/op
BenchmarkPacketScan-8 2210 515063 ns/op 59.89 MB/s 224344 B/op 12712 allocs/op
BenchmarkPacketScan-16 2276 523909 ns/op 58.88 MB/s 224412 B/op 12712 allocs/op
PASS
ok github.com/Tnze/go-mc/net/packet 5.986s
Maybe I should split into 2 PR.... should I squash the commits before merge?
Let's do what we talk about in #209 first.
(By simply store the server's key in the LoginHandler. There is no need to use interface if we have only one implementation.
I kept my opinion of switching the zlib lib, and you can open a new issue for that if you like.