An Efficient Multi-Core SIMD Implementation for H.264/AVC Encoder

Algorithm 3

Unaligned load SIMD implementation without concatenate instruction.

ui32_t PackCurr0 = *(orig_line);

ui32_t PackCurr1 = *(orig_line+1);

/* Pack to 128 bits */

TmpVectArray[0] = PackCurr0;

TmpVectArray[1] = PackCurr1;

Pack128In = ldqi(Pack128In, TmpVectArray,0);

/* Reorganize pixels */

Va = vmrgbeh(Va,Pack128In,VZero,permute0);

Vb = vmrgboh(Vb,Pack128In,VZero, permute1);

VPackCurr = vaddh(VPackCurr,Va,Vb,0);