Showing posts with label Memory Management. Show all posts
Showing posts with label Memory Management. Show all posts

31 January 2014

Process Stats: Understanding How Your App Uses RAM

Posted by Dianne Hackborn, Android framework team

Android 4.4 KitKat introduced a new system service called procstats that helps you better understand how your app is using the RAM resources on a device. Procstats makes it possible to see how your app is behaving over time — including how long it runs in the background and how much memory it uses during that time. It helps you quickly find inefficiencies and misbehaviors in your app that can affect how it performs, especially when running on low-RAM devices.

You can access procstats data using an adb shell command, but for convenience there is also a new Process Stats developer tool that provides a graphical front-end to that same data. You can find Process Stats in Settings > Developer options > Process Stats.

In this post we’ll first take a look at the Process Stats graphical tool, then dig into the details of the memory data behind it, how it's collected, and why it's so useful to you as you analyze your app.

Process Stats overview of memory used by background processes over time.

Looking at systemwide memory use and background processes

When you open Process Stats, you see a summary of systemwide memory conditions and details on how processes are using memory over time. The image at right gives you an example of what you might see on a typical device.

At the top of the screen we can see that:

  • We are looking at that data collected over the last ~3.5 hours.
  • Currently the device’s RAM is in good shape ("Device memory is currently normal").
  • During that entire time the memory state has been good — this is shown by the green bar. If device memory was getting low, you would see yellow and red regions on the left of the bar representing the amount of total time with low memory.

Below the green bar, we can see an overview of the processes running in the background and the memory load they've put on the system:

  • The percentage numbers on the right indicate the amount of time each process has spent running during the total duration.
  • The blue bars indicate the relative computed memory load of each process. (The memory load is runtime*avg_pss, which we will go into more detail on later.)
  • Some apps may be listed multiple times, since what is being shown is processes (for example, Google Play services runs in two processes). The memory load of these apps is the sum of the load of their individual processes.
  • There are a few processes at the top that have all been running for 100% of the time, but with different weights because of their relative memory use.

Analyzing memory for specific processes

The example shows some interesting data: we have a Clock app with a higher memory weight than Google Keyboard, even though it ran for less than half the time. We can dig into the details of these processes just by tapping on them:

Process Stats memory details for Clock and Keyboard processes over the past 3.5 hours.

The details for these two processes reveal that:

  • The reason that Clock has been running at all is because it is being used as the current screen saver when the device is idle.
  • Even though the Clock process ran for less than half the time of the Keyboard, its ram use was significantly larger (almost 3x), which is why its overall weight is larger.

Essentially, procstats provides a “memory use” gauge that's much like the storage use or data use gauges, showing how much RAM the apps running in the background are using. Unlike with storage or data, though, memory use is much harder to quantify and measure, and procstats uses some tricks to do so. To illustrate the complexity of measuring memory use, consider a related topic: task managers.

Understanding task managers and their memory info

We’ve had a long history of task managers on Android. Android has always deeply supported multitasking, which means the geeky of us will tend to want to have some kind of UI for seeing and controlling this multitasking like the traditional UI we are used to from the desktop. However, multitasking on Android is actually quite a bit more complicated and fundamentally different than on a traditional desktop operating system, as I previously covered in Multitasking the Android Way. This deeply impacts how we can show it to the user.

Multitasking and continuous process management

To get a feel for just how different process management is on Android, you can take a look at the output of an important system service, the activity manager, with adb shell dumpsys activity. The example below shows a snapshot of current application processes on Android 4.4, listing them from most important to least:

ACTIVITY MANAGER RUNNING PROCESSES (dumpsys activity processes)
Process LRU list (sorted by oom_adj, 22 total, non-act at 2, non-svc at 2):
  PERS #21: sys   F/ /P  trm: 0 23064:system/1000 (fixed)
  PERS #20: pers  F/ /P  trm: 0 23163:com.android.systemui/u0a12 (fixed)
  PERS #19: pers  F/ /P  trm: 0 23344:com.nuance.xt9.input/u0a77 (fixed)
  PERS #18: pers  F/ /P  trm: 0 23357:com.android.phone/1001 (fixed)
  PERS #17: pers  F/ /P  trm: 0 23371:com.android.nfc/1027 (fixed)
  Proc # 3: fore  F/ /IB trm: 0 13892:com.google.android.apps.magazines/u0a59 (service)
    com.google.android.apps.magazines/com.google.apps.dots.android.app.service.SyncService<=Proc{23064:system/1000}
  Proc # 2: fore  F/ /IB trm: 0 23513:com.google.process.gapps/u0a8 (provider)
    com.google.android.gsf/.gservices.GservicesProvider<=Proc{13892:com.google.android.apps.magazines/u0a59}
  Proc # 0: fore  F/A/T  trm: 0 24811:com.android.settings/1000 (top-activity)
  Proc # 4: vis   F/ /IF trm: 0 23472:com.google.process.location/u0a8 (service)
    com.google.android.backup/.BackupTransportService<=Proc{23064:system/1000}
  Proc #14: prcp  F/ /IF trm: 0 23298:com.google.android.inputmethod.latin/u0a57 (service)
    com.google.android.inputmethod.latin/com.android.inputmethod.latin.LatinIME<=Proc{23064:system/1000}
  Proc # 1: home  B/ /HO trm: 0 23395:com.android.launcher/u0a13 (home)
  Proc #16: cch   B/ /CA trm: 0 23966:com.google.android.deskclock/u0a36 (cch-act)
  Proc # 6: cch   B/ /CE trm: 0 7716:com.google.android.music:main/u0a62 (cch-empty)
  Proc # 5: cch   B/ /CE trm: 0 8644:com.google.android.apps.docs/u0a39 (cch-empty)
  Proc # 8: cch+2 B/ /CE trm: 0 5131:com.google.android.youtube/u0a78 (cch-empty)
  Proc # 7: cch+2 B/ /CE trm: 0 23338:com.google.android.gms/u0a8 (cch-empty)
  Proc #10: cch+4 B/ /CE trm: 0 8937:com.google.android.apps.walletnfcrel/u0a24 (cch-empty)
  Proc # 9: cch+4 B/ /CE trm: 0 24689:com.google.android.apps.plus/u0a70 (cch-empty)
  Proc #15: cch+6 B/ /S  trm: 0 23767:com.google.android.apps.currents/u0a35 (cch-started-services)
  Proc #13: cch+6 B/ /CE trm: 0 9115:com.google.android.gm/u0a44 (cch-empty)
  Proc #12: cch+6 B/ /S  trm: 0 7738:android.process.media/u0a6 (cch-started-services)
  Proc #11: cch+6 B/ /CE trm: 0 8922:com.google.android.setupwizard/u0a19 (cch-empty)

Example output of dumpsys activity command, showing all processes currently running.

There are a few major groups of processes here — persistent system processes, the foreground processes, background processes, and finally cached processes — and the category of a process is extremely important for understanding its impact on the system.

At the same time, processes on this list change all of the time. For example, in the snapshot above we can see that “com.google.android.gm” is currently an important process, but that is because it is doing a background sync, something the user would not generally be aware of or want to manage.

Snapshotting per-process RAM use

The traditional use of a task manager is closely tied to RAM use, and Android provides a tool called meminfo for looking at a snapshot of current per-process RAM use. You can access it with the command adb shell dumpsys meminfo. Here's an example of the output.

Total PSS by OOM adjustment:
    31841 kB: Native
               13173 kB: zygote (pid 23001)
                4372 kB: surfaceflinger (pid 23000)
                3721 kB: mediaserver (pid 126)
                3317 kB: glgps (pid 22993)
                1656 kB: drmserver (pid 125)
                 995 kB: wpa_supplicant (pid 23148)
                 786 kB: netd (pid 121)
                 518 kB: sdcard (pid 132)
                 475 kB: vold (pid 119)
                 458 kB: keystore (pid 128)
                 448 kB: /init (pid 1)
                 412 kB: adbd (pid 134)
                 254 kB: ueventd (pid 108)
                 238 kB: dhcpcd (pid 10617)
                 229 kB: tf_daemon (pid 130)
                 200 kB: installd (pid 127)
                 185 kB: dumpsys (pid 14207)
                 144 kB: healthd (pid 117)
                 139 kB: debuggerd (pid 122)
                 121 kB: servicemanager (pid 118)
    48217 kB: System
               48217 kB: system (pid 23064)
    49095 kB: Persistent
               34012 kB: com.android.systemui (pid 23163 / activities)
                7719 kB: com.android.phone (pid 23357)
                4676 kB: com.android.nfc (pid 23371)
                2688 kB: com.nuance.xt9.input (pid 23344)
    24945 kB: Foreground
               24945 kB: com.android.settings (pid 24811 / activities)
    17136 kB: Visible
               14026 kB: com.google.process.location (pid 23472)
                3110 kB: com.android.defcontainer (pid 13976)
     6911 kB: Perceptible
                6911 kB: com.google.android.inputmethod.latin (pid 23298)
    14277 kB: A Services
               14277 kB: com.google.process.gapps (pid 23513)
    26422 kB: Home
               26422 kB: com.android.launcher (pid 23395 / activities)
    21798 kB: B Services
               16242 kB: com.google.android.apps.currents (pid 23767)
                5556 kB: android.process.media (pid 7738)
   145869 kB: Cached
               41588 kB: com.google.android.apps.plus (pid 24689)
               21417 kB: com.google.android.deskclock (pid 23966 / activities)
               14463 kB: com.google.android.apps.docs (pid 8644)
               14303 kB: com.google.android.gm (pid 9115)
               11014 kB: com.google.android.music:main (pid 7716)
               10688 kB: com.google.android.apps.magazines (pid 13892)
               10240 kB: com.google.android.gms (pid 23338)
                9882 kB: com.google.android.youtube (pid 5131)
                8807 kB: com.google.android.apps.walletnfcrel (pid 8937)
                3467 kB: com.google.android.setupwizard (pid 8922)

Total RAM: 998096 kB
 Free RAM: 574945 kB (145869 cached pss + 393200 cached + 35876 free)
 Used RAM: 392334 kB (240642 used pss + 107196 buffers + 3856 shmem + 40640 slab)
 Lost RAM: 30817 kB
   Tuning: 64 (large 384), oom 122880 kB, restore limit 40960 kB (high-end-gfx)

Example output of dumpsys meminfo command, showing memory currently used by running processes.

We are now looking at the same processes as above, again organized by importance, but now with on their impact on RAM use.

Usually when we measure RAM use in Android, we do this with Linux’s PSS (Proportional Set Size) metric. This is the amount of RAM actually mapped into the process, but weighted by the amount it is shared across processes. So if there is a 4K page of RAM mapped in to two processes, its PSS amount for each process would be 2K.

The nice thing about using PSS is that you can add up this value across all processes to determine the actual total RAM use. This characteristic is used at the end of the meminfo report to compute how much RAM is in use (which comes in part from all non-cached processes), versus how much is "free" (which includes cached processes).

Task-manager style memory info, showing a snapshot of memory used by running apps.

Task manager UI based on PSS snapshot

Given the information we have so far, we can imagine various ways to present this in a somewhat traditional task manager UI. In fact, the UI you see in Settings > Apps > Running is derived from this information. It shows all processes running services (“svc” adjustment in the LRU list) and on behalf of the system (the processes with a “<=Proc{489:system/1000}” dependency), computing the PSS RAM for each of these and any other processes they have dependencies on.

The problem with visualizing memory use in this way is that it gives you the instantaneous state of the apps, without context over time. On Android, users don’t directly control the creation and removal of application processes — they may be kept for future use, removed when the system decides, or run in the background without the user explicitly launching them. So looking only at the instantaneous state of memory use only, you would be missing important information about what is actually going on over time.

For example, in our first look at the process state we see the com.google.android.apps.magazines process running for a sync, but when we collected the RAM use right after that it was no longer running in the background but just being kept around as an old cached process.

To address this problem, the new procstats tool continually monitors the state of all application processes over time, aggregating that information and collecting PSS samples from those processes while doing so. You can view the raw data being collected by procstats with the command adb shell dumpsys procstats.

Seeing memory use over time with procstats

Let’s now go back to procstats and take a look at the context it provides by showing memory use over time. We can use the command adb shell dumpsys procstats --hours 3 to output memory information collected over the last 3 hours. This is the same data as represented graphically in the first Process Stats example.

The output shows all of the processes that have run in the last 3 hours, sorted with the ones running the most first. (Processes in a cached state don’t count for the total time in this sort.) Like the initial graphical representation, we now clearly see a big group of processes that run all of the time, and then some that run occasionally — this includes the Magazines process, which we can now see ran for 3.6% of the time over the last 3 hours.

  * com.google.android.inputmethod.latin / u0a57:
           TOTAL: 100% (6.4MB-6.7MB-6.8MB/5.4MB-5.4MB-5.4MB over 21)
          Imp Fg: 100% (6.4MB-6.7MB-6.8MB/5.4MB-5.4MB-5.4MB over 21)
  * com.google.process.gapps / u0a8:
           TOTAL: 100% (12MB-13MB-14MB/10MB-11MB-12MB over 211)
          Imp Fg: 0.11%
          Imp Bg: 0.83% (13MB-13MB-13MB/11MB-11MB-11MB over 1)
         Service: 99% (12MB-13MB-14MB/10MB-11MB-12MB over 210)
  * com.android.systemui / u0a12:
           TOTAL: 100% (29MB-32MB-34MB/26MB-29MB-30MB over 21)
      Persistent: 100% (29MB-32MB-34MB/26MB-29MB-30MB over 21)
  * com.android.phone / 1001:
           TOTAL: 100% (6.5MB-7.1MB-7.6MB/5.4MB-5.9MB-6.4MB over 21)
      Persistent: 100% (6.5MB-7.1MB-7.6MB/5.4MB-5.9MB-6.4MB over 21)
  * com.nuance.xt9.input / u0a77:
           TOTAL: 100% (2.3MB-2.5MB-2.7MB/1.5MB-1.5MB-1.5MB over 21)
      Persistent: 100% (2.3MB-2.5MB-2.7MB/1.5MB-1.5MB-1.5MB over 21)
  * com.android.nfc / 1027:
           TOTAL: 100% (4.2MB-4.5MB-4.6MB/3.2MB-3.2MB-3.3MB over 21)
      Persistent: 100% (4.2MB-4.5MB-4.6MB/3.2MB-3.2MB-3.3MB over 21)
  * com.google.process.location / u0a8:
           TOTAL: 100% (13MB-13MB-14MB/10MB-11MB-11MB over 21)
          Imp Fg: 100% (13MB-13MB-14MB/10MB-11MB-11MB over 21)
  * system / 1000:
           TOTAL: 100% (42MB-46MB-56MB/39MB-42MB-48MB over 21)
      Persistent: 100% (42MB-46MB-56MB/39MB-42MB-48MB over 21)
  * com.google.android.apps.currents / u0a35:
           TOTAL: 100% (16MB-16MB-16MB/14MB-14MB-14MB over 17)
         Service: 100% (16MB-16MB-16MB/14MB-14MB-14MB over 17)
  * com.android.launcher / u0a13:
           TOTAL: 77% (25MB-26MB-27MB/22MB-23MB-24MB over 73)
             Top: 77% (25MB-26MB-27MB/22MB-23MB-24MB over 73)
          (Home): 23% (25MB-26MB-26MB/23MB-23MB-24MB over 12)
  * android.process.media / u0a6:
           TOTAL: 48% (5.0MB-5.3MB-5.5MB/4.0MB-4.2MB-4.2MB over 11)
          Imp Fg: 0.00%
          Imp Bg: 0.00%
         Service: 48% (5.0MB-5.3MB-5.5MB/4.0MB-4.2MB-4.2MB over 11)
        Receiver: 0.00%
        (Cached): 22% (4.1MB-4.5MB-4.8MB/3.0MB-3.5MB-3.8MB over 8)
  * com.google.android.deskclock / u0a36:
           TOTAL: 42% (20MB-21MB-21MB/18MB-19MB-19MB over 8)
          Imp Fg: 42% (20MB-21MB-21MB/18MB-19MB-19MB over 8)
         Service: 0.00%
        Receiver: 0.01%
        (Cached): 58% (17MB-20MB-21MB/16MB-18MB-19MB over 14)
  * com.android.settings / 1000:
           TOTAL: 23% (19MB-22MB-28MB/15MB-19MB-24MB over 31)
             Top: 23% (19MB-22MB-28MB/15MB-19MB-24MB over 31)
      (Last Act): 77% (9.7MB-14MB-20MB/7.5MB-11MB-18MB over 8)
        (Cached): 0.02%
  * com.google.android.apps.magazines / u0a59:
           TOTAL: 3.6% (10MB-10MB-10MB/8.7MB-9.0MB-9.0MB over 6)
          Imp Bg: 0.03%
         Service: 3.6% (10MB-10MB-10MB/8.7MB-9.0MB-9.0MB over 6)
        (Cached): 17% (9.9MB-10MB-10MB/8.7MB-8.9MB-9.0MB over 5)
  * com.android.defcontainer / u0a5:
           TOTAL: 1.4% (2.7MB-3.0MB-3.0MB/1.9MB-1.9MB-1.9MB over 7)
             Top: 1.2% (3.0MB-3.0MB-3.0MB/1.9MB-1.9MB-1.9MB over 6)
          Imp Fg: 0.19% (2.7MB-2.7MB-2.7MB/1.9MB-1.9MB-1.9MB over 1)
         Service: 0.00%
        (Cached): 15% (2.6MB-2.6MB-2.6MB/1.8MB-1.8MB-1.8MB over 1)
  * com.google.android.youtube / u0a78:
           TOTAL: 1.3% (9.0MB-9.0MB-9.0MB/7.8MB-7.8MB-7.8MB over 1)
          Imp Bg: 1.0% (9.0MB-9.0MB-9.0MB/7.8MB-7.8MB-7.8MB over 1)
         Service: 0.27%
      Service Rs: 0.01%
        Receiver: 0.00%
        (Cached): 99% (9.1MB-9.4MB-9.7MB/7.7MB-7.9MB-8.1MB over 24)
  * com.google.android.gms / u0a8:
           TOTAL: 0.91% (9.2MB-9.2MB-9.2MB/7.6MB-7.6MB-7.6MB over 1)
          Imp Bg: 0.79% (9.2MB-9.2MB-9.2MB/7.6MB-7.6MB-7.6MB over 1)
         Service: 0.11%
        Receiver: 0.00%
        (Cached): 99% (8.2MB-9.4MB-10MB/6.5MB-7.6MB-8.1MB over 25)
  * com.google.android.gm / u0a44:
           TOTAL: 0.56%
          Imp Bg: 0.55%
         Service: 0.01%
        Receiver: 0.00%
        (Cached): 99% (11MB-13MB-14MB/10MB-12MB-13MB over 24)
  * com.google.android.apps.plus / u0a70:
           TOTAL: 0.22%
          Imp Bg: 0.22%
         Service: 0.00%
        Receiver: 0.00%
        (Cached): 100% (38MB-40MB-41MB/36MB-38MB-39MB over 17)
  * com.google.android.apps.docs / u0a39:
           TOTAL: 0.15%
          Imp Bg: 0.09%
         Service: 0.06%
        (Cached): 54% (13MB-14MB-14MB/12MB-12MB-13MB over 17)
  * com.google.android.music:main / u0a62:
           TOTAL: 0.11%
          Imp Bg: 0.04%
         Service: 0.06%
        Receiver: 0.01%
        (Cached): 70% (7.7MB-10MB-11MB/6.4MB-9.0MB-9.3MB over 20)
  * com.google.android.apps.walletnfcrel / u0a24:
           TOTAL: 0.01%
        Receiver: 0.01%
        (Cached): 69% (8.1MB-8.4MB-8.6MB/7.0MB-7.1MB-7.1MB over 13)
  * com.google.android.setupwizard / u0a19:
           TOTAL: 0.00%
        Receiver: 0.00%
        (Cached): 69% (2.7MB-3.2MB-3.4MB/1.8MB-2.0MB-2.2MB over 13)

Run time Stats:
  SOff/Norm: +1h43m29s710ms
  SOn /Norm: +1h37m14s290ms
      TOTAL: +3h20m44s0ms

          Start time: 2013-11-06 07:24:27
  Total elapsed time: +3h42m23s56ms (partial) libdvm.so chromeview

Example output of dumpsys procstats --hours 3 command, showing memory details for processes running in the background over the past ~3 hours.

The percentages tell you how much of the overall time each process has spent in various key states. The memory numbers tell you about memory samples in those states, as minPss-avgPss-maxPss / minUss-avgUss-maxUss. The procstats tool also has a number of command line options to control its output — use adb shell dumpsys procstats -h to see a list of the available options.

Comparing this raw data from procstats with the visualization of its data we previously saw, we can see that it is showing only process run data from a subset of states: Imp Fg, Imp Bg, Service, Service Rs, and Receiver. These are the situations where the process is actively running in the background, for as long as it needs to complete the work it is doing. In terms of device memory use, these are the process states that tend to cause the most trouble: apps running in the background taking RAM from other things.

Getting started with procstats

We have already found the new procstats tool to be invaluable in better understanding the overall memory behavior of Android systems, and it has been a key part of the Project Svelte effort in Android 4.4.

As you develop your own applications, be sure to use procstats and the other tools mentioned here to help understand how your own app is behaving, especially how much it runs in the background and how much RAM it uses during that time.

More information about how to analyze and debug RAM use on Android is available on the developer page Investigating Your RAM Usage.

24 March 2011

Memory Analysis for Android Applications

[This post is by Patrick Dubroy, an Android engineer who writes about programming, usability, and interaction on his personal blog. — Tim Bray]

The Dalvik runtime may be garbage-collected, but that doesn't mean you can ignore memory management. You should be especially mindful of memory usage on mobile devices, where memory is more constrained. In this article, we're going to take a look at some of the memory profiling tools in the Android SDK that can help you trim your application's memory usage.

Some memory usage problems are obvious. For example, if your app leaks memory every time the user touches the screen, it will probably trigger an OutOfMemoryError eventually and crash your app. Other problems are more subtle, and may just degrade the performance of both your app (as garbage collections are more frequent and take longer) and the entire system.

Tools of the trade

The Android SDK provides two main ways of profiling the memory usage of an app: the Allocation Tracker tab in DDMS, and heap dumps. The Allocation Tracker is useful when you want to get a sense of what kinds of allocation are happening over a given time period, but it doesn't give you any information about the overall state of your application's heap. For more information about the Allocation Tracker, see the article on Tracking Memory Allocations. The rest of this article will focus on heap dumps, which are a more powerful memory analysis tool.

A heap dump is a snapshot of an application's heap, which is stored in a binary format called HPROF. Dalvik uses a format that is similar, but not identical, to the HPROF tool in Java. There are a few ways to generate a heap dump of a running Android app. One way is to use the Dump HPROF file button in DDMS. If you need to be more precise about when the dump is created, you can also create a heap dump programmatically by using the android.os.Debug.dumpHprofData() function.

To analyze a heap dump, you can use a standard tool like jhat or the Eclipse Memory Analyzer (MAT). However, first you'll need to convert the .hprof file from the Dalvik format to the J2SE HPROF format. You can do this using the hprof-conv tool provided in the Android SDK. For example:

hprof-conv dump.hprof converted-dump.hprof

Example: Debugging a memory leak

In the Dalvik runtime, the programmer doesn't explicitly allocate and free memory, so you can't really leak memory like you can in languages like C and C++. A "memory leak" in your code is when you keep a reference to an object that is no longer needed. Sometimes a single reference can prevent a large set of objects from being garbage collected.

Let's walk through an example using the Honeycomb Gallery sample app from the Android SDK. It's a simple photo gallery application that demonstrates how to use some of the new Honeycomb APIs. (To build and download the sample code, see the instructions.) We're going to deliberately add a memory leak to this app in order to demonstrate how it could be debugged.

Imagine that we want to modify this app to pull images from the network. In order to make it more responsive, we might decide to implement a cache which holds recently-viewed images. We can do that by making a few small changes to ContentFragment.java. At the top of the class, let's add a new static variable:

private static HashMap<String,Bitmap> sBitmapCache = new HashMap<String,Bitmap>();

This is where we'll cache the Bitmaps that we load. Now we can change the updateContentAndRecycleBitmap() method to check the cache before loading, and to add Bitmaps to the cache after they're loaded.

void updateContentAndRecycleBitmap(int category, int position) {
    if (mCurrentActionMode != null) {
        mCurrentActionMode.finish();
    }
 
    // Get the bitmap that needs to be drawn and update the ImageView.
 
    // Check if the Bitmap is already in the cache
    String bitmapId = "" + category + "." + position;
    mBitmap = sBitmapCache.get(bitmapId);
 
    if (mBitmap == null) {
        // It's not in the cache, so load the Bitmap and add it to the cache.
        // DANGER! We add items to this cache without ever removing any.
        mBitmap = Directory.getCategory(category).getEntry(position)
                .getBitmap(getResources());
        sBitmapCache.put(bitmapId, mBitmap);
    }
    ((ImageView) getView().findViewById(R.id.image)).setImageBitmap(mBitmap);
}

I've deliberately introduced a memory leak here: we add Bitmaps to the cache without ever removing them. In a real app, we'd probably want to limit the size of the cache in some way.

Examining heap usage in DDMS

The Dalvik Debug Monitor Server (DDMS) is one of the primary Android debugging tools. DDMS is part of the ADT Eclipse plug-in, and a standalone version can also be found in the tools/ directory of the Android SDK. For more information on DDMS, see Using DDMS.

Let's use DDMS to examine the heap usage of this app. You can start up DDMS in one of two ways:

  • from Eclipse: click Window > Open Perspective > Other... > DDMS
  • or from the command line: run ddms (or ./ddms on Mac/Linux) in the tools/ directory

Select the process com.example.android.hcgallery in the left pane, and then click the Show heap updates button in the toolbar. Then, switch to the VM Heap tab in DDMS. It shows some basic stats about our heap memory usage, updated after every GC. To see the first update, click the Cause GC button.

We can see that our live set (the Allocated column) is a little over 8MB. Now flip through the photos, and watch that number go up. Since there are only 13 photos in this app, the amount of memory we leak is bounded. In some ways, this is the worst kind of leak to have, because we never get an OutOfMemoryError indicating that we are leaking.

Creating a heap dump

Let's use a heap dump to track down the problem. Click the Dump HPROF file button in the DDMS toolbar, choose where you want to save the file, and then run hprof-conv on it. In this example, I'll be using the standalone version of MAT (version 1.0.1), available from the MAT download site.

If you're running ADT (which includes a plug-in version of DDMS) and have MAT installed in Eclipse as well, clicking the “dump HPROF” button will automatically do the conversion (using hprof-conv) and open the converted hprof file into Eclipse (which will be opened by MAT).

Analyzing heap dumps using MAT

Start up MAT and load the converted HPROF file we just created. MAT is a powerful tool, and it's beyond the scope of this article to explain all it's features, so I'm just going to show you one way you can use it to detect a leak: the Histogram view. The Histogram view shows a list of classes sortable by the number of instances, the shallow heap (total amount of memory used by all instances), or the retained heap (total amount of memory kept alive by all instances, including other objects that they have references to).

If we sort by shallow heap, we can see that instances of byte[] are at the top. As of Android 3.0 (Honeycomb), the pixel data for Bitmap objects is stored in byte arrays (previously it was not stored in the Dalvik heap), and based on the size of these objects, it's a safe bet that they are the backing memory for our leaked bitmaps.

Right-click on the byte[] class and select List Objects > with incoming references. This produces a list of all byte arrays in the heap, which we can sort based on Shallow Heap usage.

Pick one of the big objects, and drill down on it. This will show you the path from the root set to the object -- the chain of references that keeps this object alive. Lo and behold, there's our bitmap cache!

MAT can't tell us for sure that this is a leak, because it doesn't know whether these objects are needed or not -- only the programmer can do that. In this case, the cache is using a large amount of memory relative to the rest of the application, so we might consider limiting the size of the cache.

Comparing heap dumps with MAT

When debugging memory leaks, sometimes it's useful to compare the heap state at two different points in time. To do this, you'll need to create two separate HPROF files (don't forget to convert them using hprof-conv).

Here's how you can compare two heap dumps in MAT (it's a little complicated):

  1. Open the first HPROF file (using File > Open Heap Dump).
  2. Open the Histogram view.
  3. In the Navigation History view (use Window > Navigation History if it's not visible), right click on histogram and select Add to Compare Basket.
  4. Open the second HPROF file and repeat steps 2 and 3.
  5. Switch to the Compare Basket view, and click Compare the Results (the red "!" icon in the top right corner of the view).

Conclusion

In this article, I've shown how the Allocation Tracker and heap dumps can give you get a better sense of your application's memory usage. I also showed how The Eclipse Memory Analyzer (MAT) can help you track down memory leaks in your app. MAT is a powerful tool, and I've only scratched the surface of what you can do with it. If you'd like to learn more, I recommend reading some of these articles: