Sparse linear algebra is a great framework to tackle a wide variety of problems but in practice, its implementations suffer from being memory-bound. Thus a number of optimizations have been proposed towards the reduction of the number of memory accesses to increase the arithmetics-to-memory-accesses ratio with fusion optimization being the major one. This work proposes an approach to employ distillation to support fusion. It provides some benchmarks and describes the directions of future work on how to achieve even better performance.