Assume a vector length equal to 4. Two elements are conflict-free; two elements conflict (these values are stored to the same memory address). The vpconflict instruction checks the conflict and creates a mask for the vector register that is applied during SIMD processing. This mask ensures only two conflict-free elements are computed during vector processing; the other two elements are processed separately (and ignored during vector processing) to avoid unsafe behavior or an incorrect result. The loop is re-executed for any masked-out lanes.