VCMPEQUW V1, V9, V11        // compare index 2, 6, ... with sep
	VCMPEQUW V1, V10, V12       // compare index 3, 7, ... with sep
	VSEL     V6, V5, V29, V13   // merge index 0, 1, 4, 5, using mask
	VSEL     V12, V11, V30, V14 // merge index 2, 3, 6, 7, using mask
	VSEL     V14, V13, V31, V7  // final merge
	VCLZD    V7, V18            // Find first index for each half
	MFVSRD   V18, R25           // Isolate value

GLOBL ·kcon(SB), RODATA, $1312

#define SHA512ROUND0(a, b, c, d, e, f, g, h, xi, idx) \
	VSEL		g, f, e, FUNC; \
	VSHASIGMAD	$15, e, $1, S1; \
	VADDUDM		xi, h, h; \
	VSHASIGMAD	$0, a, $1, S0; \
	VADDUDM		FUNC, h, h; \
	VXOR		b, a, FUNC; \
	VADDUDM		S1, h, h; \
	VSEL		b, c, FUNC, FUNC; \
	VADDUDM		KI, g, g; \
	VADDUDM		h, d, d; \
	VADDUDM		FUNC, S0, S0; \
	LVX		(TBL)(idx), KI; \

#endif

GLOBL ·kcon(SB), RODATA, $1088

#define SHA256ROUND0(a, b, c, d, e, f, g, h, xi, idx) \
	VSEL		g, f, e, FUNC; \
	VSHASIGMAW	$15, e, $1, S1; \
	VADDUWM		xi, h, h; \
	VSHASIGMAW	$0, a, $1, S0; \
	VADDUWM		FUNC, h, h; \
	VXOR		b, a, FUNC; \
	VADDUWM		S1, h, h; \
	VSEL		b, c, FUNC, FUNC; \
	VADDUWM		KI, g, g; \
	VADDUWM		h, d, d; \
	VADDUWM		FUNC, S0, S0; \
	LVX		(TBL)(idx), KI; \

#endif

GLOBL ·kcon(SB), RODATA, $1088

#define SHA256ROUND0(a, b, c, d, e, f, g, h, xi, idx) \
	VSEL		g, f, e, FUNC; \
	VSHASIGMAW	$15, e, $1, S1; \
	VADDUWM		xi, h, h; \
	VSHASIGMAW	$0, a, $1, S0; \
	VADDUWM		FUNC, h, h; \
	VXOR		b, a, FUNC; \
	VADDUWM		S1, h, h; \
	VSEL		b, c, FUNC, FUNC; \
	VADDUWM		KI, g, g; \
	VADDUWM		h, d, d; \
	VADDUWM		FUNC, S0, S0; \
	LVX		(TBL)(idx), KI; \

	VLEIF	$2, $-43, V16           // e700ffd52843
	VLEIG	$1, $32767, V31         // e7f07fff1842
	VSLDB	$3, V1, V16, V18        // e72100030a77
	VERIMB	$2, V31, V1, V2         // e72f10020472
	VSEL	V1, V2, V3, V4          // e7412000308d
	VGFMAH	V21, V31, V24, V0       // e705f10087bc
	VFMADB	V16, V8, V9, V10        // e7a08300948f
	WFMADB	V17, V18, V19, V20      // e74123083f8f