VADDSS X11, X9, X11                     // c4413258db
	VADDSUBPD (BX), X9, X2                  // c4e131d013 or c5b1d013
	VADDSUBPD (R11), X9, X2                 // c4c131d013
	VADDSUBPD X2, X9, X2                    // c4e131d0d2 or c5b1d0d2
	VADDSUBPD X11, X9, X2                   // c4c131d0d3
	VADDSUBPD (BX), X9, X11                 // c46131d01b or c531d01b

MOVOU bswapMask<>+0(SB), X0 MOVQ SI, X1 PINSRQ $0x01, DI, X1 MOVAPS X1, X8 PSHUFB X0, X1 MOVQ SI, R8 ADDQ $0x07, R8 JC ctr8_slow XORQ R8, R8 INCQ R8 PXOR X9, X9 PINSRQ $0x00, R8, X9 PADDQ X9, X8 MOVAPS X8, X2 PADDQ X9, X8 MOVAPS X8, X3 PADDQ X9, X8 MOVAPS X8, X4 PADDQ X9, X8 MOVAPS X8, X5 PADDQ X9, X8 MOVAPS X8, X6 PADDQ X9, X8 MOVAPS X8, X7 PADDQ X9, X8 MOVAPS X8, X8 JMP ctr8_done ctr8_slow: ADDQ $0x01, SI ADCQ $0x00, DI MOVQ SI, X2 PINSRQ $0x01, DI, X2 ADDQ $0x01, SI ADCQ $0x00, DI MOVQ SI, X3 PINSRQ...

:= arg1[5] x6 := arg1[6] x7 := arg1[7] x8 := arg1[8] x9 := arg1[0] var x10 uint64 var x11 uint64 x11, x10 = bits.Mul64(x9, arg2[8]) var x12 uint64 var x13 uint64 x13, x12 = bits.Mul64(x9, arg2[7]) var x14 uint64 var x15 uint64 x15, x14 = bits.Mul64(x9, arg2[6]) var x16 uint64 var x17 uint64 x17, x16 = bits.Mul64(x9, arg2[5]) var x18 uint64 var x19 uint64 x19, x18 = bits.Mul64(x9, arg2[4]) var x20 uint64 var x21 uint64 x21, x20 = bits.Mul64(x9, arg2[3]) var x22 uint64 var x23 uint64 x23, x22 = bits.Mul64(x9,...