initial, untested, arm64

countlines_arm64.s

@@ -0,0 +1,73 @@
+// +build arm64,!gccgo,!appengine
+
+#include "textflag.h"
+
+TEXT ·countNewlinesASM(SB),NOSPLIT,$0
+	MOVQ src+0(FP), R1
+	MOVQ len+8(FP), R2
+
+	MOVD ZR, R0
+  VMOVI $0x0A V23.B16
+
+	CMP $16, R2
+	BLT tail
+
+	CMP $64, R2
+	BLT loop
+
+bigloop:
+	VLD1.P 64(R1), [V16.B16, V17.B16, V18.B16, V19.B16]
+	
+	VCMEQ V16.B16, V23.B16, V16.B16
+	VCMEQ V17.B16, V23.B16, V17.B16
+	VCMEQ V18.B16, V23.B16, V18.B16
+	VCMEQ V19.B16, V23.B16, V19.B16
+
+	VADD V16.B16, V17.B16, V17.B16
+	VADD V18.B16, V19.B16, V19.B16
+	VADD V17.B16, V19.B16, V7.B16
+	WORD $0x6E20B8E7   // VNEG V7.B16, V7.B16
+
+	VUADDLV V7.B16, V8
+	VMOV    V8.H[0], R3
+	ADD R3, R0, R0
+
+	SUB $64, R2
+	CMP ZR, R2
+	BEQ ret
+
+	CMP $64, R2
+	BGE bigloop
+
+	CMP $16, R2
+	BLT tail
+
+loop:
+	VLD1.P  16(R1), [V7.B16]
+	WORD $0x6E20B8E7   // VNEG V7.B16, V7.B16
+	VUADDLV V7.B16, V8
+	VMOV    V8.H[0], R3
+	ADD R3, R0, R0
+
+	SUB $16, R2
+	CMP ZR, R2
+	BEQ ret
+
+	CMP $16, R2
+	BGE loop
+
+tail:
+	MOVBU 1(R1), R3
+	CMP $0x0A, R3
+	BNE next
+
+	ADD $1, R0
+next:
+
+	SUB $1, R2
+	JNZ tail
+
+
+ret:
+	MOVD R0, ret+16(FP)
+	RET