8242848: Improve performance of InflaterOutputStream.write()

Reviewed-by: stuefe, vtewari, redestad, lancea

src/java.base/share/classes/java/util/zip/InflaterOutputStream.java

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2006, 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2006, 2020, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -236,16 +236,9 @@ public void write(byte[] b, int off, int len) throws IOException {
 
                 // Fill the decompressor buffer with output data
                 if (inf.needsInput()) {
-                    int part;
-
-                    if (len < 1) {
-                        break;
-                    }
-
-                    part = (len < 512 ? len : 512);
-                    inf.setInput(b, off, part);
-                    off += part;
-                    len -= part;
+                    inf.setInput(b, off, len);
+                    // Only use input buffer once.
+                    len = 0;
                 }
 
                 // Decompress and write blocks of output data
@@ -256,13 +249,14 @@ public void write(byte[] b, int off, int len) throws IOException {
                     }
                 } while (n > 0);
 
-                // Check the decompressor
-                if (inf.finished()) {
-                    break;
-                }
+                // Check for missing dictionary first
                 if (inf.needsDictionary()) {
                     throw new ZipException("ZLIB dictionary missing");
                 }
+                // Check the decompressor
+                if (inf.finished() || (len == 0)/* no more input */) {
+                    break;
+                }
             }
         } catch (DataFormatException ex) {
             // Improperly formatted compressed (ZIP) data

test/jdk/java/util/zip/DeflateIn_InflateOut.java

@@ -41,12 +41,29 @@ public class DeflateIn_InflateOut {
     private static ByteArrayOutputStream baos;
     private static InflaterOutputStream ios;
 
-    private static void reset() {
+    private static Inflater reset(byte[] dict) {
         bais = new ByteArrayInputStream(data);
-        dis = new DeflaterInputStream(bais);
+        if (dict == null) {
+            dis = new DeflaterInputStream(bais);
+        } else {
+            Deflater def = new Deflater();
+            def.setDictionary(dict);
+            dis = new DeflaterInputStream(bais, def);
+        }
 
         baos = new ByteArrayOutputStream();
-        ios = new InflaterOutputStream(baos);
+        if (dict == null) {
+            ios = new InflaterOutputStream(baos);
+            return null;
+        } else {
+            Inflater inf = new Inflater();
+            ios = new InflaterOutputStream(baos, inf);
+            return inf;
+        }
+    }
+
+    private static void reset() {
+        reset(null);
     }
 
     /** Check byte arrays read/write. */
@@ -214,6 +231,44 @@ private static void SkipBytes() throws Throwable {
         check(numNotSkipped + numSkipBytes == numReadable);
     }
 
+    /** Check "needsDictionary()". */
+    private static void NeedsDictionary() throws Throwable {
+        byte[] dict = {1, 2, 3, 4};
+        Adler32 adler32 = new Adler32();
+        adler32.update(dict);
+        long checksum = adler32.getValue();
+        byte[] buf = new byte[512];
+
+        Inflater inf = reset(dict);
+        check(dis.available() == 1);
+        boolean dictSet = false;
+        for (;;) {
+            int len = dis.read(buf, 0, buf.length);
+            if (len < 0) {
+                break;
+            } else {
+                try {
+                    ios.write(buf, 0, len);
+                    if (dictSet == false) {
+                        check(false, "Must throw ZipException without dictionary");
+                        return;
+                    }
+                } catch (ZipException ze) {
+                    check(dictSet == false, "Dictonary must be set only once");
+                    check(checksum == inf.getAdler(), "Incorrect dictionary");
+                    inf.setDictionary(dict);
+                    // After setting the dictionary, we have to flush the
+                    // InflaterOutputStream now in order to consume all the
+                    // pending input data from the last, failed call to "write()".
+                    ios.flush();
+                    dictSet = true;
+                }
+            }
+        }
+        check(dis.available() == 0);
+        ios.close();
+        check(Arrays.equals(data, baos.toByteArray()));
+    }
 
     public static void realMain(String[] args) throws Throwable {
         new Random(new Date().getTime()).nextBytes(data);
@@ -227,15 +282,24 @@ public static void realMain(String[] args) throws Throwable {
         ByteReadByteWrite();
 
         SkipBytes();
+
+        NeedsDictionary();
     }
 
     //--------------------- Infrastructure ---------------------------
     static volatile int passed = 0, failed = 0;
     static void pass() {passed++;}
-    static void fail() {failed++; Thread.dumpStack();}
-    static void fail(String msg) {System.out.println(msg); fail();}
+    static void fail() { fail(null); }
+    static void fail(String msg) {
+        failed++;
+        if (msg != null) {
+            System.err.println(msg);
+        }
+        Thread.dumpStack();
+    }
     static void unexpected(Throwable t) {failed++; t.printStackTrace();}
     static void check(boolean cond) {if (cond) pass(); else fail();}
+    static void check(boolean cond, String msg) {if (cond) pass(); else fail(msg);}
     static void equal(Object x, Object y) {
         if (x == null ? y == null : x.equals(y)) pass();
         else fail(x + " not equal to " + y);}

test/micro/org/openjdk/bench/java/util/zip/Streams.java

@@ -0,0 +1,137 @@
+/*
+ * Copyright (c) 2020, Amazon and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ */
+
+package org.openjdk.bench.java.util.zip;
+
+import org.openjdk.jmh.annotations.Benchmark;
+import org.openjdk.jmh.annotations.BenchmarkMode;
+import org.openjdk.jmh.annotations.Mode;
+import org.openjdk.jmh.annotations.OutputTimeUnit;
+import org.openjdk.jmh.annotations.Level;
+import org.openjdk.jmh.annotations.Param;
+import org.openjdk.jmh.annotations.Scope;
+import org.openjdk.jmh.annotations.Setup;
+import org.openjdk.jmh.annotations.State;
+
+import java.io.ByteArrayInputStream;
+import java.io.File;
+import java.io.FileInputStream;
+import java.io.FileOutputStream;
+import java.io.IOException;
+import java.util.Random;
+import java.util.concurrent.TimeUnit;
+import java.util.zip.Inflater;
+import java.util.zip.InflaterOutputStream;
+import java.util.zip.DeflaterOutputStream;
+
+/**
+ * Test the average execution time of "InflaterOutputStream.write()" depending
+ * on the size of the internal "InflaterOutputStream" byte buffer and the size
+ * of the compressed data input buffer passed to "write()".
+ *
+ * The size of the compressed data input buffer is controlled by the "size"
+ * parameter which runs from "512" to "65536".
+ *
+ * The size of the internal byte buffer is a multiple of "size" controlled by
+ * the "scale" paramter which runs from "1" to "8".
+ *
+ * For peak perfomance the internal buffer should be big enough to hold all
+ * the data decompressed from the input buffer. This of course depends on
+ * the compression rate of the input data. E.g. if the compression rate of
+ * the compressed input data is 4 (i.e. the original input data was compressed
+ * to 1/4 of its original size) the internal buffer should be four times bigger
+ * than the size of the compressed data buffer passed to "write()" because in
+ * that case one single call to the native zlib "inflate()" method is sufficent
+ * to decompress all data and store it in the output buffer from where it can
+ * be written to the output stream with one single call to the output streams
+ * "write()" method.
+ */
+@BenchmarkMode(Mode.AverageTime)
+@OutputTimeUnit(TimeUnit.MILLISECONDS)
+@State(Scope.Thread)
+public class Streams {
+
+    private FileInputStream in;
+    private FileOutputStream out;
+    @Param({"512", "1024", "2048", "4096", "8192", "16384", "32768", "65536"})
+    private int size;
+    @Param({"1", "2", "4", "8"})
+    private int scale;
+    private byte[] buf;
+
+    private static byte[] data = new byte[1024 * 1024];
+
+    @Setup(Level.Trial)
+    public void beforeRun() throws IOException {
+        // The reason for this whole dance is to programmatically create a one
+        // megabyte file which can be compressed by factor ~6. This will give
+        // us good results for the various scale factors (i.e. the relation
+        // between the deflated input buffer and the inflated output buffer).
+        // We achieve the desired compression factor by creating a 64 byte
+        // array of random data and than fill the final 1mb file with random
+        // 8-byte substrings of these 64 random bytes. This vaguely mimics
+        // a language with 8 character words over a set of 64 different characters.
+        final int characters = 64;
+        final int wordLength = 8;
+        buf = new byte[characters];
+        Random r = new Random(123456789);
+        r.nextBytes(buf);
+        for (int i = 0; i < data.length / wordLength; i++) {
+            System.arraycopy(buf, r.nextInt(characters - wordLength), data, i * wordLength, wordLength);
+        }
+        ByteArrayInputStream bais = new ByteArrayInputStream(data);
+
+        File deflated = File.createTempFile("inflaterOutputStreamWrite", ".deflated");
+        deflated.deleteOnExit();
+        FileOutputStream fout = new FileOutputStream(deflated);
+        DeflaterOutputStream defout = new DeflaterOutputStream(fout);
+        bais.transferTo(defout);
+        // We need to close the DeflaterOutputStream in order to flush all the
+        // compressed data in the Deflater and the underlying FileOutputStream.
+        defout.close();
+        in = new FileInputStream(deflated);
+        File inflated = File.createTempFile("inflaterOutputStreamWrite", ".inflated");
+        inflated.deleteOnExit();
+        out = new FileOutputStream(inflated);
+    }
+
+    @Setup(Level.Iteration)
+    public void beforeIteration() throws IOException {
+        in.getChannel().position(0);
+        out.getChannel().position(0);
+        buf = new byte[size];
+    }
+
+    @Benchmark
+    public void inflaterOutputStreamWrite() throws IOException {
+        in.getChannel().position(0);
+        out.getChannel().position(0);
+        InflaterOutputStream inflate = new InflaterOutputStream(out, new Inflater(), scale * size);
+        int len;
+        // buf.length == size
+        while ((len = in.read(buf)) != -1) {
+            inflate.write(buf, 0, len);
+        }
+        inflate.finish();
+    }
+}