{"$~15", "$-16"},
	{"(((8)&0xf)*4)(SP)", "32(SP)"},
	{"(((8-14)&0xf)*4)(SP)", "40(SP)"},
	{"(6+8)(AX)", "14(AX)"},
	{"(8*4)(BP)", "32(BP)"},
	{"(AX)", "(AX)"},
	{"(AX)(CX*8)", "(AX)(CX*8)"},
	{"(BP)(CX*4)", "(BP)(CX*4)"},
	{"(BP)(DX*4)", "(BP)(DX*4)"},
	{"(BP)(R8*4)", "(BP)(R8*4)"},
	{"(BX)", "(BX)"},
	{"(DI)", "(DI)"},
	{"(DI)(BX*1)", "(DI)(BX*1)"},
	{"(DX)", "(DX)"},
	{"(R9)", "(R9)"},

</p>

<pre>
get_tls(CX)
MOVQ	g(CX), AX     // Move g into AX.
MOVQ	g_m(AX), BX   // Move g.m into BX.
</pre>

<p>
Register <code>BP</code> is callee-save.
The assembler automatically inserts <code>BP</code> save/restore when frame size is larger than zero.
Using <code>BP</code> as a general purpose register is allowed,
however it can interfere with sampling-based profiling.
</p>

<h3 id="arm">ARM</h3>

<p>

	for _, v := range s {
		if len(v) > maxInt-n {
			panic("bytes: Join output length overflow")
		}
		n += len(v)
	}

	b := bytealg.MakeNoZero(n)[:n:n]
	bp := copy(b, s[0])
	for _, v := range s[1:] {
		bp += copy(b[bp:], sep)
		bp += copy(b[bp:], v)
	}
	return b
}

// HasPrefix reports whether the byte slice s begins with prefix.
func HasPrefix(s, prefix []byte) bool {

	VPMADD52HUQ X0, X11, K1, X18                       // 62e2a509b5d0
	VPMADD52HUQ 17(SP)(BP*2), X11, K1, X18             // 62e2a509b5946c11000000
	VPMADD52HUQ -7(DI)(R8*4), X11, K1, X18             // 62a2a509b59487f9ffffff
	VPMADD52HUQ X7, X31, K1, X18                       // 62e28501b5d7
	VPMADD52HUQ X0, X31, K1, X18                       // 62e28501b5d0
	VPMADD52HUQ 17(SP)(BP*2), X31, K1, X18             // 62e28501b5946c11000000

	VGF2P8AFFINEINVQB $64, -17(BP), X31, K3, X26       // 62638503cf95efffffff40
	VGF2P8AFFINEINVQB $64, -15(R14)(R15*8), X31, K3, X26 // 62038503cf94fef1ffffff40
	VGF2P8AFFINEINVQB $64, X8, X16, K3, X26            // 6243fd03cfd040
	VGF2P8AFFINEINVQB $64, X1, X16, K3, X26            // 6263fd03cfd140
	VGF2P8AFFINEINVQB $64, X0, X16, K3, X26            // 6263fd03cfd040
	VGF2P8AFFINEINVQB $64, -17(BP), X16, K3, X26       // 6263fd03cf95efffffff40

	VPLZCNTD (BX), K3, Y3                              // 62f27d2b441b
	VPLZCNTD -17(BP)(SI*1), K3, Y3                     // 62f27d2b449c35efffffff
	VPLZCNTD Z21, K3, Z3                               // 62b27d4b44dd
	VPLZCNTD Z13, K3, Z3                               // 62d27d4b44dd
	VPLZCNTD 17(SP)(BP*8), K3, Z3                      // 62f27d4b449cec11000000
	VPLZCNTD 17(SP)(BP*4), K3, Z3                      // 62f27d4b449cac11000000

	VPERMB 17(SP)(BP*1), Y31, K7, Y22                  // 62e205278db42c11000000
	VPERMB -7(CX)(DX*8), Y31, K7, Y22                  // 62e205278db4d1f9ffffff
	VPERMB Y5, Y5, K7, Y22                             // 62e2552f8df5
	VPERMB Y19, Y5, K7, Y22                            // 62a2552f8df3
	VPERMB Y31, Y5, K7, Y22                            // 6282552f8df7

	VPDPBUSDS 17(SP)(BP*8), Z13, K2, Z21               // 62e2154a51acec11000000
	VPDPBUSDS 17(SP)(BP*4), Z13, K2, Z21               // 62e2154a51acac11000000
	VPDPBUSDS Z12, Z16, K2, Z5                         // 62d27d4251ec
	VPDPBUSDS Z27, Z16, K2, Z5                         // 62927d4251eb
	VPDPBUSDS 17(SP)(BP*8), Z16, K2, Z5                // 62f27d4251acec11000000

	VPOPCNTB Z18, K7, Z13                              // 62327d4f54ea
	VPOPCNTB Z8, K7, Z13                               // 62527d4f54e8
	VPOPCNTB 17(SP)(BP*8), K7, Z13                     // 62727d4f54acec11000000
	VPOPCNTB 17(SP)(BP*4), K7, Z13                     // 62727d4f54acac11000000
	VPOPCNTW X20, K3, X11                              // 6232fd0b54dc
	VPOPCNTW X5, K3, X11                               // 6272fd0b54dd

	VRSQRT28PS 17(SP)(BP*8), K1, Z12                   // 62727d49cca4ec11000000
	VRSQRT28PS 17(SP)(BP*4), K1, Z12                   // 62727d49cca4ac11000000
	VRSQRT28PS Z11, K1, Z22                            // 62c27d49ccf3
	VRSQRT28PS Z5, K1, Z22                             // 62e27d49ccf5
	VRSQRT28PS 17(SP)(BP*8), K1, Z22                   // 62e27d49ccb4ec11000000