Shulou(Shulou.com)06/01 Report--

This article mainly explains "Stream API performance Analysis of Java8". The content of the explanation is simple and clear, and it is easy to learn and understand. Please follow the editor's train of thought to study and learn "Stream API performance Analysis of Java8".

Stream Performance

Enough hype has been made about the use of Stream API, which is simple and intuitive, but what about the performance? Will there be a high performance loss? In this section, we take a look at the performance of Stream API.

To ensure the authenticity and credibility of the test results, we run JVM in-server mode, and the test data is of the order of GB. The test machine uses a common commercial server, and the configuration is as follows:

OSCentOS 6.7 x86_64CPUIntel Xeon X5675, 12m Cache 3.06 GHz, 6 Cores 12 Threads memory 96GBJDKjava version 1.8.0mm 91, Java HotSpot (TM) 64-Bit Server VM test method and test data

Performance testing is not easy, and Java performance testing is more laborious, because virtual machines have a great impact on performance, and JVM affects performance in two ways:

The influence of GC. The behavior of GC is a very difficult piece of Java to control. To increase certainty, we manually specify the use of the CMS collector and use 10GB to fix the size of heap memory. Specific to the JVM parameter is-XX:+UseConcMarkSweepGC-Xms10G-Xmx10G

JIT (Just-In-Time) real-time compilation technology. Just-in-time compilation technology will compile the hot code into local code when JVM is running. During the test, we will warm up the program and trigger the real-time compilation of the test function. The relevant JVM parameter is-XX:CompileThreshold=10000.

The thread pool obtained by ForkJoinPool.commonPool () is used for Stream parallel execution, and we use Linux's taskset command to specify the number of cores available for JVM to control parallelism.

The test data is randomly generated by the program. In order to prevent the jitter caused by a test, the average time is calculated as the running time for 4 times.

Experiment 1 basic type iteration

Test content: find the minimum value in the integer array. Compare the external iteration performance of the for loop with the internal iteration performance of the Stream API.

The test program IntTest, and the test results are shown below:

The figure shows the time-consuming ratio of the external iterations of the for loop. The analysis is as follows:

For basic types of Stream serial iterations, the performance overhead is significantly higher than that of external iterations (twice)

Stream parallel iterations perform better than serial iterations and external iterations.

The performance of parallel iteration is related to the number of cores available. All 12 cores are used in the parallel iteration in the figure above. In order to investigate the effect of using the number of cores on performance, we specifically tested the effect of Stream parallel iteration under different kernels:

Analysis, for basic types:

Using Stream parallel API has poor performance in a single core, which is worse than that of Stream serial API.

As the number of cores used increases, the effect of Stream parallelism gets better, which is better than using external iterations of for loops.

The above two tests show that for basic types of simple iterations, the performance of Stream serial iterations is worse, but the performance of Stream iterations is better in the case of multi-core.

Experiment 2 object iteration

Let's look at the iterative effect of the object.

Test content: find the smallest element in the string list (natural order) and compare the performance of the external iteration of the for loop with the internal iteration of Stream API.

The test program StringTest, and the test results are shown below:

The results are analyzed as follows:

The performance overhead of Stream serial iterations for object types is still higher than that of external iterations (1.5 times), but the gap is not as large as that of basic types.

Stream parallel iterations perform better than serial iterations and external iterations.

Let's examine the effect of Stream parallel iteration separately:

Analysis, for object types:

The performance of parallel API using Stream is worse than that of external iterations of for loop in single core case.

With the increase of the number of cores used, the parallel effect of Stream becomes better and better, and the effect of multi-core is obvious.

The above two tests show that for simple iterations of object types, the performance of Stream serial iterations is worse, but the performance of Stream iterations is better in multi-core cases.

Experiment 3 reduction of complex objects

From the results of experiment 1 and 2, the effect of Stream serial execution is worse than that of external iterations (much), does it mean that Stream really fails? Don't jump to conclusions, let's take a look at the more complex operations.

Test content: given the list of orders, count the total transaction volume of each user. Compare the performance between the manual implementation using external iterations and Stream API.

We simplify the order to a constituent tuple and represent it as an Order object. The test program ReductionTest, and the test results are shown below:

Analysis, for complex reduction operations:

The performance of Stream API is generally better than that of external manual iteration, and the effect of parallel Stream is better.

Then examine the influence of parallelism on the parallel effect, and the test results are as follows:

Analysis, for complex reduction operations:

The performance of Stream parallel reduction is worse than that of serial reduction and manual reduction in the case of single core, which is simply the worst.

With the increase of the number of cores used, the parallel effect of Stream becomes better and better, and the effect of multi-core is obvious.

The above two experiments show that for complex reduction operations, the effect of Stream serial reduction is better than that of manual reduction, and in multi-core cases, the effect of parallel reduction is better. We have reason to believe that Stream API can show similar performance for other complex operations.

Thank you for reading, the above is the content of "Stream API performance Analysis of Java8". After the study of this article, I believe you have a deeper understanding of the Stream API performance analysis of Java8, and the specific use needs to be verified in practice. Here is, the editor will push for you more related knowledge points of the article, welcome to follow!

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.