automatic strongpoolstrong allocation improving performance by.pdfVIP

Automatic Pool Allocation: Improving Performance by Controlling Data Structure Layout in the Heap Chris Lattner Vikram Adve University of Illinois at Urbana-Champaign Abstract alyzing and controlling memory access patterns for dense arrays This paper describes Automatic Pool Allocation , a transformation than for pointer-based data structures. A key difference between framework that segregates distinct instances of heap-based data the two is that compilers have precise knowledge of the runtime structures into seperate memory pools and allows heuristics to be layout of arrays in memory, whereas they have much less informa- used to partially control the internal layout of those data structures. tion about complex data structures allocated on the heap. In such The primary goal of this work is performance improvement, not (pointer-based) data structures, both the relative layout of distinct automatic memory management, and the paper makes several new data structures in memory (which affects working set sizes) and contributions. The key contribution is a new compiler algorithm the relative layout of nodes within a single data structure (which for partitioning heap objects in imperative programs based on a affects memory traversal patterns) are difficult to predict. One di- context-sensitive pointer analysis, including a novel strategy for rect consequence is that irregular memory traversal patterns often correct handling of indirect (and potentially unsafe) function calls.