The standard approach for exploiting SIMD instructions would be to execute k independent set intersections in par- allel instead of a single scalar set intersection. This could be achieved by exchanging scalar instructions with SIMD in- structions. Scalar comparison instructions as well as scalar arithmetic instructions could be exchanged since SSE pro- vides SIMD counterparts for these instructions (see Sec- tion 2.3). However, the main problem of this approach is that available load and store instructions are restricted to loading and storing continuous memory chunks.