On Sun, 30 May 2021 17:30:56 GMT, Markus KARG <github.com+1701815+mk...@openjdk.org> wrote:
> This PR-*draft* is **work in progress** and an invitation to discuss a > possible solution for issue > [JDK-8265891](https://bugs.openjdk.java.net/browse/JDK-8265891). It is *not > yet* intended for a final review. > > As proposed in JDK-8265891, this PR provides an implementation for > `Channels.newInputStream().transferTo()` which provide superior performance > compared to the current implementation. The changes are: > * Prevents transfers through the JVM heap as much as possibly by offloading > to deeper levels via NIO, hence allowing the operating system to optimize the > transfer. > * Using more JRE heap in the fallback case when no NIO is possible (still > only KiBs, hence mostl ynegligible even on SBCs) to better perform on modern > hardware / fast I/O devides. > > Using JMH I have benchmarked both, the original implementation and this > implementation, and (depending on the used hardware and use case) performance > change was approx. doubled performance. So this PoC proofs that it makes > sense to finalize this work and turn it into an actual OpenJDK contribution. > > I encourage everybody to discuss this draft: > * Are there valid arguments for *not* doing this change? > * Is there a *better* way to improve performance of > `Channels.newInputStream().transferTo()`? > * How to go on from here: What is missing to get this ready for an actual > review? src/java.base/share/classes/sun/nio/ch/ChannelInputStream.java line 216: > 214: } > 215: > 216: return super.transferTo(out); (i am no reviewer) wouldn't this method be more intuitive if it wouldn't try to avoid using "else if" and "else"? If there are sequential if blocks with return in them, part of me is always trying to find the fall-through scenario, but in this case its all distinct code paths. Using branches would make that obvious + if there would be a fall through in future, the compiler would generate a "missing return" error right away. example: if (out instanceof ChannelOutputStream cos) { WritableByteChannel oc = cos.channel(); long i = 0L; if (ch instanceof FileChannel fc) { long pos = fc.position(); long size = fc.size(); try { for (long n = size - pos; i < n;) i += fc.transferTo(pos + i, Long.MAX_VALUE, oc); return i; } finally { fc.position(pos + i); } } else if (oc instanceof FileChannel fc) { long fcpos = fc.position(); if (ch instanceof SeekableByteChannel sbc) { long pos = sbc.position(); long size = sbc.size(); try { for (long n = size - pos; i < n;) i += fc.transferFrom(ch, fcpos + i, Long.MAX_VALUE); return i; } finally { fc.position(fcpos + i); } } else { ByteBuffer bb = Util.getTemporaryDirectBuffer(TRANSFER_SIZE); try { int r; do { i += fc.transferFrom(ch, fcpos + i, Long.MAX_VALUE); r = ch.read(bb); // detect end-of-stream if (r > -1) { bb.flip(); while (bb.hasRemaining()) oc.write(bb); bb.clear(); i += r; } } while (r > -1); return i; } finally { fc.position(fcpos + i); Util.releaseTemporaryDirectBuffer(bb); } } } else { ByteBuffer bb = Util.getTemporaryDirectBuffer(TRANSFER_SIZE); try { for (int r = ch.read(bb); r > -1; r = ch.read(bb)) { bb.flip(); while (bb.hasRemaining()) oc.write(bb); bb.clear(); i += r; } return i; } finally { Util.releaseTemporaryDirectBuffer(bb); } } } else { return super.transferTo(out); } ------------- PR: https://git.openjdk.java.net/jdk/pull/4263
