Version: Unity 6 Preview (6000.0)
Language : English
Web native plug-ins for Emscripten
Cache behavior in Web

Memory in Unity Web

Memory constraints in Unity Web can restrict the complexity of the content you run.

Web content runs in the browser. The browser allocates the memory in its memory space that your application needs to run your content. The amount of available memory varies depending on:

  • The device you use
  • The operating system you use
  • The browser you use, and whether it runs on a 32 or 64 processor
  • How much memory the browser’s JavaScript engine requires to parse your code
  • Whether the browser uses separate processes for each tab, or your content needs to share a memory space with all other open tabs.

Note: For information on security risks related to Web memory, refer to Security and Memory Resources.

Memory usage in Unity Web

The following areas of Unity Web content require the browser to allocate significant amounts of memory.

Unity heap

Unity uses a memory heap to store all Unity engine runtime objects. These include managed and native objects, loaded Assets, ScenesA Scene contains the environments and menus of your game. Think of each unique Scene file as a unique level. In each Scene, you place your environments, obstacles, and decorations, essentially designing and building your game in pieces. More info
See in Glossary
, and shadersA program that runs on the GPU. More info
See in Glossary
. This is like the memory that Unity Players use on any other platform.

The Unity heap is a contiguous block of allocated memory. Unity supports automatic resizing for the heap to suit the needs of the application. The heap size expands as an application runs, and can expand up to 2GB. Unity creates this memory heap as a Memory object. The Memory object’s buffer property is a re-sizable ArrayBuffer that holds the raw bytes of memory accessed by WebAssembly code.

Automatic resizing of the heap can cause your application to crash if the browser fails to allocate a contiguous memory block in the address space. For this reason, it’s important to keep the Unity heap size as small as possible. Therefore, be mindful when you are planning the memory usage of your application. If you want to test the size of your Unity heap, you can use the ProfilerA window that helps you to optimize your game. It shows how much time is spent in the various areas of your game. For example, it can report the percentage of time spent rendering, animating, or in your game logic. More info
See in Glossary
to profile the contents of the memory block.

You can control the initial size and growth of the heap by using the Memory Growth Mode options in the Web Player Settings. The default options are configured to work well for all desktop use cases. However, for mobile browsers you need to use the advanced tuning options. For mobile browsers, it’s recommended to configure the Initial Memory Size to the typical heap usage of the application.

Asset data

When you create a Unity Web build, Unity generates a .data file. This contains all the Scenes and Assets the application needs to launch. Because Unity Web doesn’t have access to the real file system, it creates a virtual memory file system, and the browser unpacks the .data file here. The Emscripten framework (JavaScript) allocates this memory file system in the browser memory space. While your content runs, the browser memory keeps the uncompressed data. To keep both download times and memory usage low, try to keep this uncompressed data as small as possible.

To reduce memory use, you can pack your Asset data into AssetBundles. AssetBundles offer full control over your asset downloads. You can control when your application downloads an asset, and when the runtime unloads it. You can unload unused assets to free up memory.

AssetBundles are downloaded directly into the Unity heap, so these don’t result in extra allocation by the browser.

Enable Data Caching to automatically cache the Asset data in your content on the user’s machine. This means you don’t need to re-download that data during later runs. The Unity Web loader implements Data Caching with the IndexedDB API. This option lets you to cache files which are too large for the browser to cache natively.

Data caching enables the browser to store application data on the user’s machine. Browsers often limit the amount you can store in their cache and the maximum file size that can be cached. This is often not enough for an application to run smoothly. The Unity Web loader Caching with the IndexedDB API that lets Unity store the data in the IndexedDB instead of the browser cache.

To enable the Data Caching option go to File > Build Settings > Player Settings > Publishing Settings.

Garbage collection

Garbage collection is the process of locating and freeing up unused memory. For an overview on how Unity garbage collection works, refer to Automatic Memory Management. To debug the garbage collection process, use the Unity Profiler.

Due to a security limitation of WebAssembly, user programs are not allowed to examine the native execution stack to prevent possible exploits.

This means that on the Web platform, the GC can only run when no managed code is executing (which could potentially reference live C# objects). This occurs at the end of every rendered game frame.

In other words, on the Web platform, the garbage collector can’t run in the middle of executing C# code, and only runs at the end of each program frame. This discrepancy causes some differences in garbage collection behavior on Web compared to other platforms.

Because of these differences, pay close attention to code that performs a lot of temporary allocations per-frame, especially if these allocations might exhibit a sequence of linear size growth. Such allocations may cause a temporary quadratic memory growth pressure for the garbage collector.

As an example, if you have a long-running loop, the following code might fail when you run it on Web because the garbage collector doesn’t run between iterations of the for loop. In this case, the garbage collector can’t free up memory that the intermediate string objects use and will run out of memory in the Unity heap.

string hugeString = "";
 
for (int i = 0; i < 100000; i++)
{
   hugeString += "foo";
}

In the above example, the length of hugeString at the end of the loop is 3 * 100000 = 300000 characters. The code, however, generates a hundred thousand temporary strings before producing the final string. The total allocated memory throughout the loop is 3 * (1 + 2 + 3 + … + 100000) = 3 * (100000 * 100001 / 2) = 15 gigabytes.

On native platforms, the garbage collector continuously cleans up previous temporary copies of the string while the loop executes. So the above code doesn’t require 15 GB of RAM in total to run.

On the Web platform, the garbage collector doesn’t reclaim the temporary string copies until the end of the frame. As a result, the above code runs out of memory attempting to allocate 15 GB of RAM.

The following code shows a second example where this type of temporary quadratic memory pressure can occur:

byte[] data;
 
for (int i = 0; i < 100000; i++)
{
   data = new byte[i];
   // do something temporary with data[]
} 

Here the code allocates 1 + 2 + 3 + … + 100000 bytes = 5 GB worth of bytes temporarily, even though only the last 100 KB array is preserved. This causes the program to seemingly run out of memory on the Web platform even though only 100 KB are necessary in the final output.

To limit these types of issues, avoid code constructs that perform quadratically increasing amounts of temporary memory allocations. Instead, either pre-allocate the final needed data size, or use a List<T> or similar data structure that performs geometrically increasing capacity reservations that mitigate the temporary memory pressure.

For example, with the List<T> container, consider using the List<T>.ReserveCapacity() function that enables pre-allocating the needed capacity, if you know the final size of the data structure. Likewise, consider using the List<T>.TrimExcess() function when shrinking the size of a container that previously held several megabytes of memory.

Note: When you use C# delegates or events such as Delegate, Action, Func, these classes internally utilize similar linear growth allocations as above. Avoid excessive amounts of per-frame delegate registrations and unregistrations with these classes to minimize the temporary memory pressure for the garbage collector on the Web platform.

Additional resources

Web native plug-ins for Emscripten
Cache behavior in Web