DirectByteBuffer直接内存的使用场景和作用

生命周期长的大对象,

减少java堆GC, 减少内存copy

http://www.importnew.com/26334.html

 


 1 public class DirectByteBufferTest {
 2 
 3 
 4     @Test
 5     public void test_copyFile() throws IOException {
 6         ByteBuffer byteBuffer = ByteBuffer.allocateDirect(10);//100kbytes
 7         FileChannel readChannel = FileChannel.open(new File("D:/in.txt").toPath());
 8         //out.txt必须已经存在, writeChannel必须以WRITE方式打开
 9         FileChannel writeChannel = FileChannel.open(new File("D:/out.txt").toPath(), StandardOpenOption.WRITE);
10         int read;
11         while ((read = readChannel.read(byteBuffer)) != -1) {
12             //buffer从读切换到写
13             byteBuffer.flip();
14             // 打印信息必须放在flip后面, 否则decode出来的是上次read的结果.根据in.txt的字符编码修改下面的ISO_8859_1
15             //System.out.println(read + "--" + StandardCharsets.ISO_8859_1.decode(byteBuffer));
16             writeChannel.write(byteBuffer);
17             // 写完之后清空缓冲区,否则read=0一直死循环
18             byteBuffer.clear();
19         }
20         writeChannel.close();
21         readChannel.close();
22     }
23 }

 测试其他总结:

ByteBuffer.allocateDirect缓冲区大小根据输入文件的大小调整,但是太大时输出性能也提高不了多少, 对于大文件,1M的缓存区应该差不多了。
对于超大文件, 应该换成其他读取方式

benchmark: https://developer.ibm.com/articles/j-zerocopy


相关文章: