vadd.go

package vadd

func addGeneric(dst, src []uint8) {
	for i, s := range src {
		dst[i] += s
	}
}

func addVector(dst, src []uint8)

vadd_arm64.s

TEXT ·addVector(SB),$56
	MOVD	dstp+0(FP), R1
	MOVD	srcp+24(FP), R3

	VLD1	(R1), [V0.B16, V1.B16, V2.B16, V3.B16]

	VLD1	(R3), [V4.B16, V5.B16, V6.B16, V7.B16]

	VADD	V0.B16, V4.B16, V4.B16
	VADD	V1.B16, V5.B16, V5.B16
	VADD	V2.B16, V6.B16, V6.B16
	VADD	V3.B16, V7.B16, V7.B16

	VST1	[V4.B16, V5.B16, V6.B16, V7.B16], (R1)

	RET

vadd_test.go

package vadd

import (
	"testing"
)

func TestAddVector(t *testing.T) {
	var src, dst []uint8
	for i := uint8(0); i < 64; i++ {
		src = append(src, i)
		dst = append(dst, 64-i)
	}
	addVector(dst, src)
	for i, c := range dst {
		if c != 64 {
			t.Error(i, c)
		}
	}
}

func BenchmarkAddGeneric(b *testing.B) {
	src := make([]uint8, 64)
	dst := make([]uint8, 64)
	b.SetBytes(64)
	b.ResetTimer()
	for i := 0; i < b.N; i++ {
		addGeneric(dst, src)
	}
}
func BenchmarkAddVector(b *testing.B) {
	src := make([]uint8, 64)
	dst := make([]uint8, 64)
	b.SetBytes(64)
	b.ResetTimer()
	for i := 0; i < b.N; i++ {
		addVector(dst, src)
	}
}