1 / 18

Enhancing Reference Counting for C++ with Data Structure Annotations

This document explores the improvement of reference counting in C++ to address issues with garbage collection, specifically concerning cycles that cannot be freed automatically. The proposed solution utilizes user-specified data structure annotations to optimize performance by minimizing unnecessary marking and sweeping, and preventing memory leaks. Examples include differentiating between primary and secondary pointers in complex data structures like doubly linked lists and binary trees. Techniques, analyses, and related works are discussed to enhance reference counting and memory management.

erol
Télécharger la présentation

Enhancing Reference Counting for C++ with Data Structure Annotations

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Specialized Reference Counting Garbage Collection using Data Structure Annotations By Eric Watkins and Dzin Avots for CS 343 Spring 2002

  2. Motivation • Improve use of ref counting for C++ • Problem: cycles can’t be freed automatically, expensive to determine what is unreachable • Solution: Use user specified data structure annotations to improve performance

  3. Goals & Techniques • Avoid unnecessary mark & sweeps • Acyclic Data Structure Optimizations • Avoid some memory leaks • Reference Removal Invariant Model

  4. Reference Counting using Smart Pointers

  5. Garbage Example #1 • Doubly Linked List • Everything is strongly connected • Nothing can be freed until all external refs are gone • How can we recognize this data structure? ext ptr ext ptr

  6. Abstract Description of Data Structures • The compiler can’t differentiate these two even though they will obviously be connected in different ways. • ADDS can tell us how to customize smart pointer usage for certain data structures

  7. Doubly Linked List ADDS

  8. Binary Tree w/ Parents ADDS

  9. ADDS spec • Dimensions (1-d, 2-d, etc.) • Directionality (forward and backward) • Uniqueness of forward references (DAGs vs. lists) • Cyclicity

  10. Linked List Optimization • The entire dimension is strongly connected, forwards and backwards • Summarize with single ref counter for all external references to dimension

  11. Dimensional Independence • Dimensions X and Y are dependent if one traversal along X and another traversal along Y may lead to a common node by different paths

  12. Other ADDS issues • What about describing relations between different acyclic data structures? • Could avoid marking & sweeping this structure etc.

  13. Primary & Secondary Pointers • Primary pointers – if these point to an object, then object is still in use • Secondary pointers – point to object, but don’t affect use status • Memory leak – Primary pointers are all removed, but a secondary pointer remains

  14. Garbage Example #2 • Programs can contain complicated, interconnected relationships between objects BugManager Grid Secondary refs Primary refs Bug Bug

  15. Garbage Example #2 • If the user forgets to remove the secondary reference, the Bug won’t be deleted. BugManager Grid Secondary refs Primary ref Bug Bug

  16. Primary & Secondary Pointers • User specifies primary & secondary pointers in ADDS specification • Primary & Secondary pointer invariant can be expressed as instrumentation predicate in 3-valued logic • Generate shape analysis to detect leak condition

  17. Plan of Action • Investigate use of ADDS specs to improve refcount performance • Generate interprocedural shape analysis from primary & secondary edge specs • Look into other improvements from user specifications

  18. Related Work • ADDS work: L. Hendren and J. Hummel, Abstractions for Recursive Pointer Data Structures: Improving the Analysis and Transformation of Imperative Programs • Shape Analysis: M. Sagiv, T. Reps and R. Wilhelm, Parametric Shape Analysis via 3-Valued Logic

More Related