Abstract

Strictness analysis techniques can be classified along four different dimensions: firstorder vs. higher-order, flat vs. non-flat, low fidelity vs. high fidelity, and forward vs. backward. Plotting a table of the positions of known techniques within this space reveals that certain regions are densely occupied while others are empty. In particular, techniques for high-fidelity forward and low-fidelity backward analysis are well known, while those for low-fidelity forward and high-fidelity backward analysis are lacking. This paper fills in the gaps: the low-fidelity forward methods provide faster analyses than the high-fidelity forward methods, at the cost of accuracy, while the high-fidelity backward methods provide more information than the low-fidelity backward methods, at the cost of time. 1 Introduction Strictness analysis is an important part of many compilers for lazy functional languages, and a wide variety of strictness analysis techniques have been proposed. It is not clear how...

Citations