not satisfied with the benchmark result

[xtaci]

$ go test -run=^$ -bench  .
beginning tests, encryption:salsa20, fec:10/3
goos: darwin
goarch: amd64
pkg: github.com/xtaci/kcp-go
BenchmarkSM4-4             	   20000	     92234 ns/op	  32.53 MB/s	    6016 B/op	     376 allocs/op
BenchmarkAES128-4          	  200000	      7268 ns/op	 412.74 MB/s	       0 B/op	       0 allocs/op
BenchmarkAES192-4          	  200000	      8190 ns/op	 366.28 MB/s	       0 B/op	       0 allocs/op
BenchmarkAES256-4          	  200000	      9055 ns/op	 331.30 MB/s	       0 B/op	       0 allocs/op
BenchmarkTEA-4             	  100000	     19463 ns/op	 154.13 MB/s	       0 B/op	       0 allocs/op
BenchmarkXOR-4             	20000000	       108 ns/op	27679.97 MB/s	       0 B/op	       0 allocs/op
BenchmarkBlowfish-4        	   50000	     34005 ns/op	  88.22 MB/s	       0 B/op	       0 allocs/op
BenchmarkNone-4            	30000000	        56.5 ns/op	53066.58 MB/s	       0 B/op	       0 allocs/op
BenchmarkCast5-4           	   30000	     41978 ns/op	  71.47 MB/s	       0 B/op	       0 allocs/op
Benchmark3DES-4            	   10000	    149801 ns/op	  20.03 MB/s	       0 B/op	       0 allocs/op
BenchmarkTwofish-4         	   30000	     48963 ns/op	  61.27 MB/s	       0 B/op	       0 allocs/op
BenchmarkXTEA-4            	   30000	     59826 ns/op	  50.15 MB/s	       0 B/op	       0 allocs/op
BenchmarkSalsa20-4         	  500000	      4010 ns/op	 748.07 MB/s	       0 B/op	       0 allocs/op

SM4 is slow compared to other block ciphers, and it has many memory allocations when crypting.

am I misused this library or may be any improvements can be taken?

[tjfoc]

thanks.

similar to DES, SM4 is slow compared with other block ciphers, especially AES.

Will keep this issue open and take some time to optimize this. And pull requests are welcome.

[QwertyJack]

Speed up approx 70 %.
See #11 .

[tjfoc]

#11 merged. thanks

[xtaci]

the speed is comparable to twofish now. 4x faster than 3des.

[xtaci]

the next step is to add support for _amd64.s , optimize the core computation with assembly.

[tjfoc]

performance improved in 1.1.0