Predicting Lifetimes in Dynamically Allocated Memory

Predictions oflifetimes of dynamically allocated objects can be used to improve time and space efficiency of dynamic memory management in computer programs. Barrett and Zorn [1993] used a simple lifetime predictor and demonstrated this improvement on a variety of computer programs. In this paper, we use decision trees to do lifetime prediction on the same programs and show significantly better prediction . Our method also has the advantage that during training we can use a large number of features and let the decision tree automatically choose the relevant subset . 1 INTELLIGENT MEMORY ALLOCATION Dynamic memory allocation is used in many computer applications. The application requests blocks of memory from the operating system or from a memory manager when needed and explicitly frees them up after use. Typically, all of these requests are handled in the same way, without any regard for how or for how long the requested block will be used . Sometimes programmers use runtime profiles to analyze the typical behavior of their program and write special purpose memory management routines specifically tuned to dominant classes of allocation events . Machine learning methods offer the opportunity to automate the process of tuning memory management systems. In a recent study, Barrett and Zorn [1993] used two allocators: a special allocator for objects that are short-lived, and a default allocator for everything else. They tried a simple prediction method on a number of public-domain , allocation-intensive programs and got mixed results on the lifetime prediction problem. Nevertheless, they showed that for all the cases where they were able to predict well, their strategy of assigning objects predicted to be short-lived to the special allocator led to savings 940 D. A. Cohn and S. Singh in program running times. Their results imply that if we could predict well in all cases we could get similar savings for all programs. We concentrate on the lifetime prediction task in this paper and show that using axis-parallel decision trees does indeed lead to significantly better prediction on all the programs studied by Zorn and Grunwald and some others that we included. Another advantage of our approach is that we can use a large number of features about the allocation requests and let the decision tree decide on their relevance. There are a number of advantages of using lifetime predictions for intelligent memory management. It can improve CPU usage, by using special-purpose allocators, e.g., short-lived objects can be allocated in small spaces by incrementing a pointer and deallocated together when they are all dead. It can decrease memory fragmentation, because the short-lived objects do not pollute the address space of long lived objects. Finally, it can improve program locality, and thus program speed, because the short-lived objects are all allocated in a small part of the heap. The advantages of prediction must be weighed against the time required to examine each request and make that prediction about its intended use. It is frequently argued that, as computers and memory become faster and cheaper, we need to be less concerned about the speed and efficiency of machine learning algorithms. When the purpose of the algorithm is to save space and computation, however, these concerns are paramount.