ORRW R13<<4, R8, R26                       // 1a110d2a
	ORR R3<<22, R5, R6                         // a65803aa
	PRFM (R8), $25                             // 190180f9
	PRFM (R2), PLDL1KEEP                       // 400080f9
	//TODO PRFM (R27)(R30.SXTW<<3), PLDL2STRM  // 63dbbff8
	//TODO PRFUM 22(R16), PSTL1KEEP            // 106281f8
	RBITW R9, R22                              // 3601c05a

	VTRN2	V5.D2, V4.D2, V23.D2            // 9768c54e


// special
	PRFM	(R2), PLDL1KEEP                 // 400080f9
	PRFM	16(R2), PLDL1KEEP               // 400880f9
	PRFM	48(R6), PSTL2STRM               // d31880f9
	PRFM	8(R12), PLIL3STRM               // 8d0580f9
	PRFM	(R8), $25                       // 190180f9
	PRFM	8(R9), $30                      // 3e0580f9
	NOOP                                    // 1f2003d5

V7.D2, V6, i3 \ VADD i3.D2, i1.D2, i4.D2 \ SHA512H2 i0.D2, i1, i3 // func blockSHA512(dig *Digest, p []byte) TEXT ·blockSHA512(SB),NOSPLIT,$0 MOVD dig+0(FP), R0 MOVD p_base+8(FP), R1 MOVD p_len+16(FP), R2 MOVD $·_K+0(SB), R3 // long enough to prefetch PRFM (R3), PLDL3KEEP // load digest VLD1 (R0), [V8.D2, V9.D2, V10.D2, V11.D2] loop: // load digest in V0-V3 keeping original in V8-V11 VMOV V8.B16, V0.B16 VMOV V9.B16, V1.B16 VMOV V10.B16, V2.B16 VMOV V11.B16, V3.B16 // load message data in V12-V19 VLD1.P...