func (bp *BytePoolCap) Populate() {
	for _, buf := range reedsolomon.AllocAligned(cap(bp.c), bp.wcap) {
		bp.Put(buf[:bp.w])
	}
}

// Get gets a []byte from the BytePool, or creates a new one if none are
// available in the pool.
func (bp *BytePoolCap) Get() (b []byte) {
	if bp == nil {
		return nil
	}
	select {
	case b = <-bp.c:
		// reuse existing buffer
	default:

	VPGATHERQQ (BP)(X2*2), K1, X2   // ERROR "index and destination registers should be distinct"
	VPGATHERQQ (BP)(Y15*2), K1, Y15 // ERROR "index and destination registers should be distinct"
	VPGATHERQQ (BP)(Z20*2), K1, Z20 // ERROR "index and destination registers should be distinct"
	VPGATHERDQ (BP)(X2*2), K1, X2   // ERROR "index and destination registers should be distinct"

			"#define KEYROUND(xmm, load, off, r1, r2, index) \\",
			"\tMOVBLZX	(BP)(DX*4),	R8 \\",
			"\tload((off+1), r2) \\",
			"\tMOVB	R8,		(off*4)(R12) \\",
			"\tPINSRW	$index, (BP)(R8*4), xmm",
			"#define LOAD(off, reg) \\",
			"\tMOVBLZX	(off*4)(R12),	reg \\",
			"\tADDB	reg,		DX",
			"KEYROUND(X0, LOAD, 8, AX, BX, 0)",
		),