Detailed explanation of several garbage collection principles in Java

In Java, except for basic types such as integers and references, all objects are allocated in the heap area instead of the stack area. This design eliminates the need for programmers to pay attention to the life cycle of variables, but at the cost of generating more garbage.

Reachability

Data that can be directly accessed by the program by dereferencing any pointer is reachable.

Locality Principle

If the storage location of a program is likely to be accessed again in a short period of time, the program is said to have temporal locality. A program is spatially localized if a location adjacent to an accessed memory location is likely to be accessed within a short period of time.

It is generally believed that a program spends 90% of its time executing 10% of the code.

The principles of several garbage collectors

Mark-sweep collector

This collector first traverses the object graph and marks reachable objects, and then scans Stack to find unmarked objects and release their memory. This type of collector generally uses a single thread to work and stops other operations. Moreover, because it only clears those unmarked objects without compressing the marked objects, a large amount of memory fragmentation will be generated, thus wasting memory.

Mark-Compact Collector

Sometimes also called Mark-Sweep-Compact Collector, it has the same marking phase as the Mark-Sweep Collector. In the second phase, the marked object is copied to a new area of ​​the stack in order to compress the stack. This collector also stops other operations.

Copy collector

This collector divides the stack into two domains, often called half-space. Only half of the space is used each time, and new objects generated by the JVM are placed in the other half of the space. When the GC runs, it copies the reachable objects to the other half of the space, thus compressing the stack. This method is suitable for short-lived objects. Continuous copying of long-lived objects will lead to reduced efficiency. And for a given size heap, twice the amount of memory is required because only half of it is used at any time.

Incremental collector

The incremental collector divides the stack into multiple domains and only collects garbage from one domain at a time. It can also be understood as dividing the stack into small pieces. , only one block is garbage collected at a time. This results in a small application interruption time, so that the user is generally not aware that the garbage collector is working.

Partial Recycling Principle

Usually 80% to 90% of newly allocated objects are within a few million instructions or have been reallocated.

Generational garbage collector (Generational garbage coolection)

It is based on the principle of copy collector and partial recycling.

An effective way to take advantage of the "die young" property of most objects.

Divide the heap into a series of small areas and number them with 0,1,2...n. The smaller the number, the younger the object stored in the area. The object is first stored in the 0 area. After creation and filling, the garbage is recycled, and the reachable objects are moved to area 1. Each round of garbage collection is performed for areas with serial numbers less than or equal to i, where i is the highest number of the currently filled area.

As long as i is recycled, all areas with serial number Xiaoyu i will also be garbage collected. This is because younger generations tend to contain more garbage, that is, they are recycled more frequently.

The oldest generation saves the most mature objects, and the recycling of these objects is the most expensive, equivalent to a complete recycling. Can cause longer pauses. The solution is to use the train algorithm.

HotSpot's four GC collectors:
Serial recycling serial collector:
Features: The application will be suspended during recycling.
Young area: Combine Eden and a Survivor area Surviving objects are copied to another Survivor area (set to TO) (large objects are placed directly in the old area). If the TO area is full, they are copied directly to the old area.
old area: use mark-sweep-compact GC Algorithm, that is, first mark the surviving objects, then clear the discarded objects, and then move the surviving objects to an area to free up a larger free space.
Scope of application: Most client applications can use this kind of recycling Algorithm, this is also the default recycling algorithm of HotSpot. On the current machine (2006), a complete recycling of a 64MB area takes less than half a second.

Parallel recycling parallel collector :
Features: Multi-core CPU can be used.
Young area: The application still needs to be paused. The basic mechanism is similar to serialization, but multi-threading is used. This can speed up efficiency.
Old area: Same string Linearization.
Can be used on multi-core computers.

Parallel compacting collector:
Compared with parallel recycling, it mainly has a new algorithm in the old area. At the same time, according to the white paper, this kind of recycling will eventually replace parallel recycling.
young Area: Same as parallel recycling.
old area: First, divide old into several consecutive areas. Then, check each area in parallel and mark the alive objects (first mark the objects that can be directly referenced, Then all). Then start checking these areas to get the density (the area on the left is definitely denser than the area on the right), starting from an area that is not very dense, and compress the area on the right in parallel.
Scope of application: For environments with multi-core and pause time requirements, it is better to use parallel compression and recycling than parallel recycling. However, for servers with a high sharing rate (that is, one server runs multiple applications), due to the old area The collection is slow and multi-threaded, so the GC of one application will affect other applications. The corresponding solution: you can configure to reduce the number of threads in parallel.

Parallel mark clearing and recycling Concurrent Mark Sweep collector :
young area: Same as parallel recycling.
old area: divided into several steps.
Initialmark: When GC needs to be executed, first pause the application, and then mark all directly referenced objects.
Concurrentmark: Then continue the application and check the marked objects at the same time to get all surviving objects.
remark: Pause the application again and check the objects modified by the application during the duration of Concurrent mark (new, Abandoned), and mark the surviving objects. This phase lasts for a long time, so multi-threading is used. After the phase is over, all surviving objects are marked, and unmarked objects are garbage objects.
sweep: Stop pausing the application, and then release the space of all garbage objects.

Differences from other algorithms:
First: no compression is performed. However, it will be calculated by calculating possible future memory requirements Merge/split certain memory blocks.
Second: The GC is not executed until the old area is full, but starts when the space is less than a certain level.
Third: Since compression is not performed, fragmentation will occur.

In addition, CMS can also use the incremental operation mode, that is, only perform part of the work in the Concurrentmark phase, and then return the resources to the application. The work of the recycler will be divided into several parts and scheduled in two young The idle phase of regional recycling is completed. This mode is generally used when pause time is required and the number of processors is not large (single-core or dual-core).
In general, compared with parallel recycling, CMS is reduced Reduce the pause time of old GC (sometimes the effect is very significant), slightly lengthen the time of young GC (because the time for objects to move from the young area to the old area will be longer: no compression is performed, so a suitable area must be found first), reducing This improves some execution efficiency of the entire system and greatly enhances the demand for memory space.

The above is the detailed content of Detailed explanation of several garbage collection principles in Java. For more information, please follow other related articles on the PHP Chinese website!