What′s New in dotMemory

dotMemory can now get sampled data about memory allocation based on ETW events. Compared to the traditional (statistical) way of collecting allocation data, sampling is less accurate but provides a number of advantages:

• The sampled allocation data is collected without performance penalties.
• You don’t need to enable this data collection explicitly; it is always active.
• Sampled allocation data can also be collected when you attach the profiler to an application that is already running.

Note that this feature is available only on Windows.

You can now use the Subsystems view to analyze memory allocation data. A subsystem groups all methods belonging to the same type, namespace, or assembly. The resulting view displays objects created by the subsystems and a merged call tree for each subsystem.

In this release, we have continued to improve the way you analyze memory allocation. Two new tabs have been added to the Memory Allocation view:

• The Methods tab lets you analyze a plain list of methods that have allocated memory.
• The Call Tree tab lets you analyze the allocation call tree. Each node in the tree shows not only the called method but also the objects allocated by the call.

The search bar at the top of dotMemory views is now more flexible and easier to use:

• The filters have additional options for narrowing the search scope: #struct (search only in value types), #c (only in types), #m (only in methods), #ns (only in namespaces).
• You can get help for all filters possible in a given context.

It’s now possible to:

• Specify the traversal depth when exporting object data.
• Select one or more types and export their data to a single file.
• Export object data from the Instances, Outgoing References, and Unreachable Objects views.

We’ve completely reworked the dotMemory Home screen – it is much easier to configure and start new profiling sessions, work with snapshots, and perform other basic operations.

Now, you can analyze dumps of .NET Core applications collected on the Linux systems using gcore.

Now, dotMemory lets you analyze memory allocation on an arbitrary time interval. Just select the interval on the timeline and the Memory Allocation view will show you the objects allocated on this interval, as well as the stack trace that allocated them.

dotMemory uses the timeline data to provide automatic inspections right on the Timeline graph.

Currently, there are three inspections available:

• High Garbage Collection pressure
• Pinned objects in Gen0
• Fast Large Objects Heap growth rate

.NET 5 introduced a separate heap segment for storing pinned objects - objects that cannot be moved in a heap which leads to heap fragmentation. dotMemory shows the Pinned Object Heap in the Heap Fragmentation View. Also, now dotMemory shows memory allocated to the Pinned Object Heap in real-time during profiling.

• Garbage collections are shown on a separate graph below the main timeline.
• Now, you can select a particular time range on the timeline: The information about GC time and allocated memory will be shown for the selected time range.
• Performance improvements.

The dotMemory command-line profiler for Linux (any distribution with GLIBC_2.23 or later) lets you profile applications on ARM64 systems.

You can now open pinned objects as a separate object set.

You can now open objects that are not reachable from GC roots as a separate object set.