Download JXRay Memory Analyzer

 

IMPORTANT! BY DOWNLOADING THE SOFTWARE BELOW, YOU AGREE TO BE BOUND BY THE TERMS OF THE END USER LICENSE AGREEMENT.

 

 

Download
Zip file, all platforms
jxray-2.2.zip
Compressed archive in ZIP format [402.1 KB]

What's new in version 2.2

  • We now report off-heap (native) memory used by java.nio.DirectByteBuffer objects
  • Improved random samples of collections: contents of contained objects are printed where possible, capacity is printed for array-based collections, etc.
  • Improved "ergonomics": memory overhead is highlighted by different colors depending on the value; if a dump was generated as a result of OutOfMemoryError, the respective thread is printed in the Top-Level Stats section, etc.
  • Added support for some more collection types 
     

What's new in version 2.1

  • In the "where memory goes, by class" there are now random samples of objects, for each class and reference chain
  • Fixed some issues with parsing and analyzing Android heap dumps
  • Added new problem kind: primitive arrays where in each element the upper half (high bits), such as two high bytes in int, are zero
  • Performance and memory consumption for big heap dumps improved even more
     

What's new in version 2.0

  • "Where memory goes, by GC root" section is now presented as expandable trees, one tree for each GC root
  • All threads and stack traces are printed in a new report section
  • System.properties are printed in a new report section
  • The jxray.sh script now accepts JXRAY_JAVA_HOME and JXRAY_JAVA_FLAGS env variables, and/or the user can pass the JVM options directly to the script, using the -J prefix, e.g. -J-Xmx64g
  • Improved performance and reduced memory consumption, especially for big dumps

 

What's new in version 1.9

  • Significantly improved performance (at least 30 per cent faster than before) and reduced memory consumption
  • UI improvements: most tables are now collapsible; very big tables are scrollable; bytes in byte[] arrays with human-readable contents are presented as chars, etc.
  • Heap scanning order has been changed so that the more meaningful roots such as statics are scanned first, WeakHashMaps and weak references are scanned in the end, etc.
  • New problem kind: sparse collections

 

What's new in version 1.8

  • Semi-interactive HTML presentation
  • More detailed output, including the "Where memory goes, by class" section
  • New problem kind: sparse object arrays
  • New problem kind: sparse primitive arrays

 

What's new in version 1.7

  • New problem kind: objects that are unreachable, but still in the finalization queue
  • New problem kind: misconfigured maximum heap size, that makes the HotSpot JVM use a less compact object format (wide pointers) leading to higher than necessary memory consumption
  • Better presentation of reference chains with high retained memory
  • For duplicate strings, bad collections etc. we now report "expensive fields" when problematic objects are pointed by multiple different reference chains ending with the same data field, like MyClassReferencedByMany.almostAlwaysEmptyHashMap
  • The percentage of various "bad" objects wrt. the total number of such objects is now reported. For example, how many (what percentage) of HashMap instances are empty, etc.
  • Better lookup of possible sources of garbage objects. That is, finding live reference chains leading to objects of type Foo that are most similar to the given group of unreachable Foo objects.
  • Several performance, scalability and accuracy improvements

 

What's new in version 1.6

  • Improved calculation and presentation of reference chains with high retained memory when such chains stem from multiple concurrent threads and local variables.
  • Number of referenced objects is reported in addition to retained memory for each reference chain
  • Improved performance, especially for the heap dumps that have many small objects and strings
  • The real JVM pointer size (4 or 8 bytes) and object header size (12 or 16 bytes) is calculated correctly regardless of the heap dump file size
  • Linked lists of arbitrary type, such as Foo.next -> Foo.next -> ... are presented in a compact form, like {Foo.next}
  • Instances of StringBuilder and StringBuffer are now treated in the same way as Strings - their char[] buffer is handled as implementation details
  • Total number of threads in the dump is reported in addition to the total number of classes.

 

What's new in version 1.5

  • New problem detection: duplicate objects of any type
  • New problem detection: bad primitive arrays (size 0 and 1, empty (all zeroes), high number of trailing zeroes)
  • Better presentation: type of the last object(s) in every reference chain is now displayed
  • Some scalability issues fixed

 

Print Print | Sitemap
© JXRay.com