Use ByteProcessor in HpackHuffmanDecoder to reduce bound-checks and r… (

netty#9317)


Motivation:

ff0045e changed HpackHuffmanDecoder to use a lookup-table which greatly improved performance. We can squeeze out another 3% win by using an ByteProcessor which will reduce the number of bound-checks / reference-count-checks needed by processing byte-by-byte.

Modifications:

Implement logic with ByteProcessor

Result:

Another ~3% perf improvement which shows up when using h2load to simulate load.

`h2load -c 100 -m 100 --duration 60 --warm-up-time 10 http://127.0.0.1:8080`

Before:

```
finished in 70.02s, 620051.67 req/s, 20.70MB/s
requests: 37203100 total, 37203100 started, 37203100 done, 37203100 succeeded, 0 failed, 0 errored, 0 timeout
status codes: 37203100 2xx, 0 3xx, 0 4xx, 0 5xx
traffic: 1.21GB (1302108500) total, 41.84MB (43872600) headers (space savings 90.00%), 460.24MB (482598600) data
                     min         max         mean         sd        +/- sd
time for request:      404us     24.52ms     15.93ms      1.45ms    87.90%
time for connect:        0us         0us         0us         0us     0.00%
time to 1st byte:        0us         0us         0us         0us     0.00%
req/s           :    6186.64     6211.60     6199.00        5.18    65.00%
```

With this change:

```
finished in 70.02s, 642103.33 req/s, 21.43MB/s
requests: 38526200 total, 38526200 started, 38526200 done, 38526200 succeeded, 0 failed, 0 errored, 0 timeout
status codes: 38526200 2xx, 0 3xx, 0 4xx, 0 5xx
traffic: 1.26GB (1348417000) total, 42.39MB (44444900) headers (space savings 90.00%), 466.25MB (488893900) data
                     min         max         mean         sd        +/- sd
time for request:      370us     24.89ms     15.52ms      1.35ms    88.02%
time for connect:        0us         0us         0us         0us     0.00%
time to 1st byte:        0us         0us         0us         0us     0.00%
req/s           :    6407.06     6435.19     6419.74        5.62    67.00%
```

codec-http2/src/main/java/io/netty/handler/codec/http2/HpackDecoder.java

@@ -89,6 +89,7 @@ final class HpackDecoder {
     private static final byte READ_LITERAL_HEADER_VALUE_LENGTH = 8;
     private static final byte READ_LITERAL_HEADER_VALUE = 9;
 
+    private final HpackHuffmanDecoder huffmanDecoder = new HpackHuffmanDecoder();
     private final HpackDynamicTable hpackDynamicTable;
     private long maxHeaderListSize;
     private long maxDynamicTableSize;
@@ -445,7 +446,7 @@ private void insertHeader(Sink sink, CharSequence name, CharSequence value, Inde
 
     private CharSequence readStringLiteral(ByteBuf in, int length, boolean huffmanEncoded) throws Http2Exception {
         if (huffmanEncoded) {
-            return HpackHuffmanDecoder.decode(in, length);
+            return huffmanDecoder.decode(in, length);
         }
         byte[] buf = new byte[length];
         in.readBytes(buf);

codec-http2/src/main/java/io/netty/handler/codec/http2/HpackHuffmanDecoder.java

@@ -33,11 +33,12 @@
 
 import io.netty.buffer.ByteBuf;
 import io.netty.util.AsciiString;
+import io.netty.util.ByteProcessor;
 import io.netty.util.internal.ThrowableUtil;
 
 import static io.netty.handler.codec.http2.Http2Error.COMPRESSION_ERROR;
 
-final class HpackHuffmanDecoder {
+final class HpackHuffmanDecoder implements ByteProcessor {
 
     /* Scroll to the bottom! */
 
@@ -4668,50 +4669,77 @@ final class HpackHuffmanDecoder {
             Http2Exception.newStatic(COMPRESSION_ERROR, "HPACK - Bad Encoding",
                     Http2Exception.ShutdownHint.HARD_SHUTDOWN), HpackHuffmanDecoder.class, "decode(..)");
 
+    private byte[] dest;
+    private int k;
+    private int state;
+    private int flags;
+
+    HpackHuffmanDecoder() { }
+
     /**
      * Decompresses the given Huffman coded string literal.
      *
      * @param buf the string literal to be decoded
      * @return the output stream for the compressed data
      * @throws Http2Exception EOS Decoded
      */
-    public static AsciiString decode(ByteBuf buf, int length) throws Http2Exception {
+    public AsciiString decode(ByteBuf buf, int length) throws Http2Exception {
         if (length == 0) {
             return AsciiString.EMPTY_STRING;
         }
-        byte[] dest = new byte[length * 8 / 5];
-        int k = 0;
-        int state = 0;
-        int flags = 0;
-        for (int i = 0; i < length; i++) {
-            byte input = buf.readByte();
-            int index = (state << 4) | ((input & 0xFF) >>> 4);
-            int row = HUFFS[index];
-            flags = row & 0x00FF00;
-            if ((flags & HUFFMAN_FAIL_SHIFT) != 0) {
-                throw BAD_ENCODING;
+        dest = new byte[length * 8 / 5];
+        try {
+            int readerIndex = buf.readerIndex();
+            // Using ByteProcessor to reduce bounds-checking and reference-count checking during byte-by-byte
+            // processing of the ByteBuf.
+            int endIndex = buf.forEachByte(readerIndex, length, this);
+            if (endIndex == -1) {
+                // We did consume the requested length
+                buf.readerIndex(readerIndex + length);
+                if ((flags & HUFFMAN_COMPLETE_SHIFT) != HUFFMAN_COMPLETE_SHIFT) {
+                    throw BAD_ENCODING;
+                }
+                return new AsciiString(dest, 0, k, false);
             }
-            if ((flags & HUFFMAN_EMIT_SYMBOL_SHIFT) != 0) {
-                dest[k++] = (byte) (row & 0xFF);
-            }
-            state = row >> 16;
 
-            index = (state << 4) | (input & 0x0F);
-            row = HUFFS[index];
-            flags = row & 0x00FF00;
-            if ((flags & HUFFMAN_FAIL_SHIFT) != 0) {
-                throw BAD_ENCODING;
-            }
-            if ((flags & HUFFMAN_EMIT_SYMBOL_SHIFT) != 0) {
-                dest[k++] = (byte) (row & 0xFF);
-            }
-            state = row >> 16;
-        }
-        if ((flags & HUFFMAN_COMPLETE_SHIFT) != HUFFMAN_COMPLETE_SHIFT) {
+            // The process(...) method returned before the requested length was requested. This means there
+            // was a bad encoding detected.
+            buf.readerIndex(endIndex);
             throw BAD_ENCODING;
+        } finally {
+            dest = null;
+            k = 0;
+            state = 0;
+            flags = 0;
         }
-        return new AsciiString(dest, 0, k, false);
     }
 
-    private HpackHuffmanDecoder() { }
+    /**
+     * <strong>This should never be called from anything but this class itself!</strong>
+     */
+    @Override
+    public boolean process(byte input) {
+        int index = (state << 4) | ((input & 0xFF) >>> 4);
+        int row = HUFFS[index];
+        flags = row & 0x00FF00;
+        if ((flags & HUFFMAN_FAIL_SHIFT) != 0) {
+            return false;
+        }
+        if ((flags & HUFFMAN_EMIT_SYMBOL_SHIFT) != 0) {
+            dest[k++] = (byte) (row & 0xFF);
+        }
+        state = row >> 16;
+
+        index = (state << 4) | (input & 0x0F);
+        row = HUFFS[index];
+        flags = row & 0x00FF00;
+        if ((flags & HUFFMAN_FAIL_SHIFT) != 0) {
+            return false;
+        }
+        if ((flags & HUFFMAN_EMIT_SYMBOL_SHIFT) != 0) {
+            dest[k++] = (byte) (row & 0xFF);
+        }
+        state = row >> 16;
+        return true;
+    }
 }

codec-http2/src/test/java/io/netty/handler/codec/http2/HpackHuffmanTest.java

@@ -153,7 +153,7 @@ private static void roundTrip(HpackHuffmanEncoder encoder, byte[] buf)
     private static byte[] decode(byte[] bytes) throws Http2Exception {
         ByteBuf buffer = Unpooled.wrappedBuffer(bytes);
         try {
-            AsciiString decoded = HpackHuffmanDecoder.decode(buffer, buffer.readableBytes());
+            AsciiString decoded = new HpackHuffmanDecoder().decode(buffer, buffer.readableBytes());
             Assert.assertFalse(buffer.isReadable());
             return decoded.toByteArray();
         } finally {