Apple Inc.

David Vernon Payne

Daniel Mark Delwood

Katherine Blake Stone

Bertrand Serlet

Darin Benjamin Adler

Maciej Stachowiak

Christina Elizabeth Warren

Christopher A Wolf

Maxwell Oliver Drukman

Andreas W Wendker

Daniel Andreas Steffen

Stephen Richard Lewallen

Kenneth S Orr

Han Ming Ong

Jason Molenda

The present technology monitors events that allocate and deallocate virtual memory regions in a device, wherein the events include system calls from user space. The system can generate a log of events, and based on the log of events, track regions of virtual memory allocated and deallocated via the events. The system can also record events with corresponding stack traces. Next, the system can group recorded events having matching stack traces to yield event groupings, and instrument functions in a compiled code associated with the process to determine retain counts of respective events associated with the functions. The system can then automatically pair at least one of a first portion of the events and a second portion of the respective events based on the event groupings and the retain counts of the respective events to yield paired events.

Memory stack trace management tool

Described is a tool for allowing developers to analyze objects in to process virtual address space. The system includes a technique for identifying references between objects within memory and refining these references with runtime metadata and analysis heuristics. These objects may then be displayed in the form of a graph that employs a progressive disclosure to more effectively allow a developer debug memory. For example, the graph may organize objects within a coordinate grid system that allows a developer to concentrate or expand reference paths as necessary when analyzing objects or detecting potentially leaked or abandoned sources of memory.

Memory debugging tool for displaying objects within a navigable graph

Memory management tools

When queuing work items for asynchronous execution, the approaches set forth herein enable a debugger to provide more meaningful and complete stack trace information for enqueued work items. An example system identifies a work item to be enqueued for asynchronous execution as a separate thread during runtime of a program. Upon enqueuing the work item in an execution queue, the system captures pre-queue stack trace information associated with the work item, and associates the pre-queue stack trace information with the work item. The system removes the work item from the execution queue for execution, wherein the execution is associated with current stack trace information separate from the pre-queue stack trace information. Upon detecting an event associated with executing the work item, the system merges the pre-queue stack trace information with the current stack trace information to yield a merged stack trace.

Queue debugging using stored backtrace information

A tool for analyzing software is enhanced to provide multiple views of a stack trace, with each view having a different level of detail. Different views may be lightly simplified, moderately simplified, or heavily simplified. The display of a complete stack trace includes entries for all stack frames in the stack trace. The display of a simplified stack trace includes entries for fewer than all stack frames in the stack trace, thereby 'hiding' or 'collapsing' entries for some of the stack frames, relative to the complete stack trace display. After a user specifies a level of complexity with which to show a stack trace, the enhanced analysis tool GUI updates the stack trace display according to the specified level. A complexity level is associated with a set of heuristics that is used to create a view of a stack trace at that complexity level.

Collapsible stack trace

A system, method, and computer program for analyzing code execution and software performance characteristics are disclosed. Samples of code during run-time or from crash logs can be displayed as both a top-down tree of calls from a particular function and a bottom-up tree of calls to the particular function. Data mining techniques, including filtering and focusing, remove code that is not of interest and display code that is more relevant to performance analysis. In particular, functions contained in libraries such as system libraries can be stripped from view. The cost associated with an excluded function can be propagated to a calling function, in either the top-down or bottom-up tree.

Software performance analysis using data mining

Growing community of inventors

Los Altos Hills, CA, United States of America

David Vernon Payne