ci/bench.sh: Benchmark masking on arm64 with QEMU

Author: nhooyr

ci/bench.sh

@@ -6,4 +6,7 @@ go test --run=^$ --bench=. --benchmem --memprofile ci/out/prof.mem --cpuprofile
 (
   cd ./internal/thirdparty
   go test --run=^$ --bench=. --benchmem --memprofile ../../ci/out/prof-thirdparty.mem --cpuprofile ../../ci/out/prof-thirdparty.cpu -o ../../ci/out/thirdparty.test "$@" .
+
+  GOARCH=arm64 go test -c -o ../../ci/out/thirdparty-arm64.test .
+  qemu-aarch64 ../../ci/out/thirdparty-arm64.test --test.run=^$ --test.bench=Benchmark_mask --test.benchmem --test.memprofile ../../ci/out/prof-thirdparty-arm64.mem --test.cpuprofile ../../ci/out/prof-thirdparty-arm64.cpu .
 )

internal/thirdparty/frame_test.go

@@ -2,6 +2,7 @@ package thirdparty
 
 import (
 	"encoding/binary"
+	"runtime"
 	"strconv"
 	"testing"
 	_ "unsafe"
@@ -34,6 +35,10 @@ func nbioMaskBytes(b, key []byte) int
 func gorillaMaskBytes(key [4]byte, pos int, b []byte) int
 
 func Benchmark_mask(b *testing.B) {
+	b.Run(runtime.GOARCH, benchmark_mask)
+}
+
+func benchmark_mask(b *testing.B) {
 	sizes := []int{
 		8,
 		16,