crypto/aes: linux/arm64 Go 1.9 performance is +20X slower than OpenSSL

[williamweixiao]

Please answer these questions before submitting your issue. Thanks!

What version of Go are you using (go version)?

go version go1.9.2 linux/arm64

Does this issue reproduce with the latest release?

yes

What operating system and processor architecture are you using (go env)?

GOARCH="arm64"
GOBIN=""
GOEXE=""
GOHOSTARCH="arm64"
GOHOSTOS="linux"
GOOS="linux"
GOPATH=""
GORACE=""
GOROOT="/usr/lib/go-1.6"
GOTOOLDIR="/usr/lib/go-1.6/pkg/tool/linux_arm64"
GO15VENDOREXPERIMENT="1"
CC="gcc"
GOGCCFLAGS="-fPIC -pthread -fmessage-length=0"
CXX="g++"
CGO_ENABLED="1"

What did you do?

go test crypto/cipher -bench GCM

What did you expect to see?

Performance can be on par with OpenSSL (https://blog.cloudflare.com/content/images/2017/11/sym_key_1_core.png)

What did you see instead?

+20X slower than OpenSSL( https://blog.cloudflare.com/content/images/2017/11/go_sym_key_1_core.png)

[as]

C:\Users\as>go test crypto/cipher -bench GCM
goos: windows
goarch: amd64
pkg: crypto/cipher
BenchmarkAESGCMSeal1K-4          5000000               314 ns/op        3260.87 MB/s
BenchmarkAESGCMOpen1K-4          5000000               276 ns/op        3705.65 MB/s
BenchmarkAESGCMSeal8K-4          1000000              1732 ns/op        4728.25 MB/s
BenchmarkAESGCMOpen8K-4          1000000              1721 ns/op        4758.42 MB/s
PASS
ok      crypto/cipher   7.189s

12.5MBps is exactly 100Mbps. This number seems too round to ignore, any thoughts on why it won't exceed the speed of a "fast ethernet" network interface? It would help to see the benchmark's output from your system as well.

[williamweixiao]

Following is the result I got from some typical arm64 platform:

go test crypto/cipher -bench GCM
goos: linux
goarch: arm64
pkg: crypto/cipher
BenchmarkAESGCMSeal1K-32 50000 35735 ns/op 28.65 MB/s
BenchmarkAESGCMOpen1K-32 50000 35858 ns/op 28.56 MB/s
BenchmarkAESGCMSeal8K-32 5000 285102 ns/op 28.73 MB/s
BenchmarkAESGCMOpen8K-32 5000 278226 ns/op 29.44 MB/s
PASS
ok crypto/cipher 7.199s

[as]

Taking a second look at the results and seeing arm64 instead of amd64, I realize the reason has already been stated in prior discussions: no assembly implementations for the platform. It's probably safe to disregard my previous inquisition.

[matt2909]

This looks to be a duplicate of #18498

[titanous]

Duplicate of #18498. Closing.

[gopherbot]

Change https://golang.org/cl/64490 mentions this issue: crypto/aes: optimize arm64 AES implementation

[vielmetti]

On arm64, Packet Type 2A / c1.large.arm Cavium ThunderX:

ed@ed-2a-bcc-llvm:~$ go test crypto/cipher -bench GCM

goos: linux
goarch: arm64
pkg: crypto/cipher
BenchmarkAESGCMSeal1K-96           20000             68788 ns/op          14.89 MB/s
BenchmarkAESGCMOpen1K-96           20000             68697 ns/op          14.91 MB/s
BenchmarkAESGCMSign8K-96           10000            182114 ns/op          44.98 MB/s
BenchmarkAESGCMSeal8K-96            3000            536359 ns/op          15.27 MB/s
BenchmarkAESGCMOpen8K-96            3000            537432 ns/op          15.24 MB/s
PASS
ok      crypto/cipher   9.404s
ed@ed-2a-bcc-llvm:~$ 
ed@ed-2a-bcc-llvm:~$ go version
go version go1.10.2 linux/arm64
ed@ed-2a-bcc-llvm:~$ ~/go/bin/go1.11beta1 test crypto/cipher -bench GCM
goos: linux
goarch: arm64
pkg: crypto/cipher
BenchmarkAESGCMSeal1K-96           50000             37520 ns/op          27.29 MB/s
BenchmarkAESGCMOpen1K-96           50000             37550 ns/op          27.27 MB/s
BenchmarkAESGCMSign8K-96           10000            172278 ns/op          47.55 MB/s
BenchmarkAESGCMSeal8K-96            5000            289794 ns/op          28.27 MB/s
BenchmarkAESGCMOpen8K-96            5000            288511 ns/op          28.39 MB/s
PASS
ok      crypto/cipher   9.274s

1.11beta1 is substantially faster than 1.10.2.