andreacavalli/netty5
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Reduce the memory copies in JdkZlibEncoder

Browse Source 
Motivation:

At the moment we use a lot of unnecessary memory copies in JdkZlibEncoder. This is caused by either allocate a to small ByteBuf and expand it later or using a temporary byte array.
Beside this the memory footprint of JdkZlibEncoder is pretty high because of the byte[] used for compressing.

Modification:

- Override allocateBuffer(...) and calculate the estimatedsize in there, this reduce expanding of the ByteBuf later
- Not use byte[] in the instance itself but allocate a heap ByteBuf and write directly into the byte array

Result:

Less memory copies and smaller memory footprint
This commit is contained in:
Norman Maurer 2014-06-26 10:57:27 +02:00
parent 937f790f70
commit 4d2b78ca3c
1 changed files with 30 additions and 19 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

49
codec/src/main/java/io/netty/handler/codec/compression/JdkZlibEncoder.java
View File

@ -34,7 +34,6 @@ import java.util.zip.Deflater;
public class JdkZlibEncoder extends ZlibEncoder {
private final ZlibWrapper wrapper;
private final byte[] encodeBuf = new byte[8192];
private final Deflater deflater;
private volatile boolean finished;
private volatile ChannelHandlerContext ctx;
@ -211,23 +210,11 @@ public class JdkZlibEncoder extends ZlibEncoder {
offset = 0;
}
int sizeEstimate = (int) Math.ceil(inAry.length * 1.001) + 12;
if (writeHeader) {
writeHeader = false;
switch (wrapper) {
case GZIP:
out.ensureWritable(sizeEstimate + gzipHeader.length);
out.writeBytes(gzipHeader);
break;
case ZLIB:
out.ensureWritable(sizeEstimate + 2); // first two magic bytes
break;
default:
out.ensureWritable(sizeEstimate);
if (wrapper == ZlibWrapper.GZIP) {
out.writeBytes(gzipHeader);
}
} else {
out.ensureWritable(sizeEstimate);
}
if (wrapper == ZlibWrapper.GZIP) {
@ -240,6 +227,23 @@ public class JdkZlibEncoder extends ZlibEncoder {
}
}
@Override
protected final ByteBuf allocateBuffer(ChannelHandlerContext ctx, ByteBuf msg,
boolean preferDirect) throws Exception {
int sizeEstimate = (int) Math.ceil(msg.readableBytes() * 1.001) + 12;
if (writeHeader) {
switch (wrapper) {
case GZIP:
sizeEstimate += gzipHeader.length;
break;
case ZLIB:
sizeEstimate += 2; // first two magic bytes
break;
}
}
return ctx.alloc().heapBuffer(sizeEstimate);
}
@Override
public void close(final ChannelHandlerContext ctx, final ChannelPromise promise) throws Exception {
ChannelFuture f = finishEncode(ctx, ctx.newPromise());
@ -268,8 +272,7 @@ public class JdkZlibEncoder extends ZlibEncoder {
}
finished = true;
ByteBuf footer = ctx.alloc().buffer();
ByteBuf footer = ctx.alloc().heapBuffer();
if (writeHeader && wrapper == ZlibWrapper.GZIP) {
// Write the GZIP header first if not written yet. (i.e. user wrote nothing.)
writeHeader = false;
@ -277,8 +280,14 @@ public class JdkZlibEncoder extends ZlibEncoder {
}
deflater.finish();
while (!deflater.finished()) {
deflate(footer);
if (!footer.isWritable()) {
// no more space so write it to the channel and continue
ctx.write(footer);
footer = ctx.alloc().heapBuffer();
}
}
if (wrapper == ZlibWrapper.GZIP) {
int crcValue = (int) crc.getValue();
@ -299,8 +308,10 @@ public class JdkZlibEncoder extends ZlibEncoder {
private void deflate(ByteBuf out) {
int numBytes;
do {
numBytes = deflater.deflate(encodeBuf, 0, encodeBuf.length, Deflater.SYNC_FLUSH);
out.writeBytes(encodeBuf, 0, numBytes);
int writerIndex = out.writerIndex();
numBytes = deflater.deflate(
out.array(), out.arrayOffset() + writerIndex, out.writableBytes(), Deflater.SYNC_FLUSH);
out.writerIndex(writerIndex + numBytes);
} while (numBytes > 0);
}