Motivation
JMH 1.22 was released recently, we might as well use the latest when
running benchmarks.
Summary of changes:
https://mail.openjdk.java.net/pipermail/jmh-dev/2019-November/002879.html
Modifications
Update jmh dependencies in microbench module from version 1.21 to 1.22.
Result
Benchmarks run using latest JMH
Motivation
Currently when future tasks are scheduled via EventExecutors from a
different thread, at least two allocations are performed - the
ScheduledFutureTask wrapping the to-be-run task, and a Runnable wrapping
the action to add to the scheduled task priority queue. The latter can
be avoided by incorporating this logic into the former.
Modification
- When scheduling or cancelling a future task from outside the event
loop, enqueue the task itself rather than wrapping in a Runnable
- Have ScheduledFutureTask#run first verify the task's deadline has
passed and if not add or remove it from the scheduledTaskQueue depending
on its cancellation state
- Add new outside-event-loop benchmarks to ScheduleFutureTaskBenchmark
Result
Fewer allocations when scheduling/cancelling future tasks
Motivation
Currently a static AtomicLong is used to allocate a unique id whenever a
task is scheduled to any event loop. This could be a source of
contention if delayed tasks are scheduled at a high frequency and can be
easily avoided by having a non-volatile id counter per queue.
Modifications
- Replace static AtomicLong ScheduledFutureTask#nextTaskId with a long
field in AbstractScheduledExecutorService
- Set ScheduledFutureTask#id based on this when adding the task to the
queue (in event loop) instead of at construction time
- Add simple benchmark
Result
Less contention / cache-miss possibility when scheduling future tasks
Before:
Benchmark (num) Mode Cnt Score Error Units
scheduleLots 100000 thrpt 20 346.008 ± 21.931 ops/s
Benchmark (num) Mode Cnt Score Error Units
scheduleLots 100000 thrpt 20 654.824 ± 22.064 ops/s
Motivation:
Netty homepage(netty.io) serves both "http" and "https".
It's recommended to use https than http.
Modification:
I changed from "http://netty.io" to "https://netty.io"
Result:
No effects.
Motivation:
The previous used maxHeaderListSize was too low which resulted in exceptions during the benchmark run:
```
io.netty.handler.codec.http2.Http2Exception: Header size exceeded max allowed size (8192)
at io.netty.handler.codec.http2.Http2Exception.connectionError(Http2Exception.java:103)
at io.netty.handler.codec.http2.Http2Exception.headerListSizeError(Http2Exception.java:188)
at io.netty.handler.codec.http2.Http2CodecUtil.headerListSizeExceeded(Http2CodecUtil.java:231)
at io.netty.handler.codec.http2.HpackDecoder$Http2HeadersSink.finish(HpackDecoder.java:545)
at io.netty.handler.codec.http2.HpackDecoder.decode(HpackDecoder.java:132)
at io.netty.handler.codec.http2.HpackDecoderBenchmark.decode(HpackDecoderBenchmark.java:85)
at io.netty.handler.codec.http2.generated.HpackDecoderBenchmark_decode_jmhTest.decode_thrpt_jmhStub(HpackDecoderBenchmark_decode_jmhTest.java:120)
at io.netty.handler.codec.http2.generated.HpackDecoderBenchmark_decode_jmhTest.decode_Throughput(HpackDecoderBenchmark_decode_jmhTest.java:83)
at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
at java.lang.reflect.Method.invoke(Method.java:498)
at org.openjdk.jmh.runner.BenchmarkHandler$BenchmarkTask.call(BenchmarkHandler.java:453)
at org.openjdk.jmh.runner.BenchmarkHandler$BenchmarkTask.call(BenchmarkHandler.java:437)
at java.util.concurrent.FutureTask.run(FutureTask.java:266)
at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511)
at java.util.concurrent.FutureTask.run(FutureTask.java:266)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
at io.netty.util.concurrent.FastThreadLocalRunnable.run(FastThreadLocalRunnable.java:30)
at java.lang.Thread.run(Thread.java:748)
```
Also we should ensure we only use ascii for header names.
Modifications:
Just use Integer.MAX_VALUE as limit
Result:
Be able to run benchmark without exceptions
Motivation:
Some methods that either override others or are implemented as part of implementation an interface did miss the `@Override` annotation
Modifications:
Add missing `@Override`s
Result:
Code cleanup
Motivation:
SpotJMHBugs reports that accumulating a value as a way of eliding dead code
elimination may be inadvisable, as discussed in
`JMHSample_34_SafeLooping::measureWrong_2`. Change the test so that it consumes
the response with `Blackhole::consume` instead.
Modifications:
- Replace addition of results with explicit `blackhole.consume()` call
Result:
Tests work as before, but with different benchmark numbers.
Motivation:
Some JMH benchmarks need additional explanations to motivate
specific code choices.
Modifications:
Introduced comment to explai why calling BlackHole::consume
in a loop is not always the right choice for some benchmark.
Result:
The relevant method shows a comment that warn about changing
the code to introduce BlackHole::consume in the loop.
Motivation:
Resolve the issue highlighted by SpotJMHBugs that the creation of the RecyclableArrayList may be elided by the JIT since the result isn't consumed or returned.
Modifications:
Return the result of `list.recycle()` so that the list isn't elided.
Result:
The JMH benchmark shows a change in performance indicating that the prior results of this may be unsound.
Motivation:
The wakeup logic in EpollEventLoop is overly complex
Modification:
* Simplify the race to wakeup the loop
* Dont let the event loop wake up itself (it's already awake!)
* Make event loop check if there are any more tasks after preparing to
sleep. There is small window where the non-eventloop writers can issue
eventfd writes here, but that is okay.
Result:
Cleaner wakeup logic.
Benchmarks:
```
BEFORE
Benchmark Mode Cnt Score Error Units
EpollSocketChannelBenchmark.executeMulti thrpt 20 408381.411 ± 2857.498 ops/s
EpollSocketChannelBenchmark.executeSingle thrpt 20 157022.360 ± 1240.573 ops/s
EpollSocketChannelBenchmark.pingPong thrpt 20 60571.704 ± 331.125 ops/s
Benchmark Mode Cnt Score Error Units
EpollSocketChannelBenchmark.executeMulti thrpt 20 440546.953 ± 1652.823 ops/s
EpollSocketChannelBenchmark.executeSingle thrpt 20 168114.751 ± 1176.609 ops/s
EpollSocketChannelBenchmark.pingPong thrpt 20 61231.878 ± 520.108 ops/s
```
Motivation
Pipeline handlers are free to "take control" of input buffers if they have singular refcount - in particular to mutate their raw data if non-readonly via discarding of read bytes, etc.
However there are various places (primarily unit tests) where a wrapped byte-array buffer is passed in and the wrapped array is assumed not to change (used after the wrapped buffer is passed to EmbeddedChannel.writeInbound()). This invalid assumption could result in unexpected errors, such as those exposed by #8931.
Modifications
Anywhere that the data passed to writeInbound() might be used again, ensure that either:
- A copy is used rather than wrapping a shared byte array, or
- The buffer is otherwise protected from modification by making it read-only
For the tests, copying is preferred since it still allows the "mutating" optimizations to be exercised.
Results
Avoid possible errors when pipeline assumes it has full control of input buffer.
Motivation:
Results are just wrong for small delays.
Modifications:
Switching to AvarageTime avoid to rely on OS nanoTime granularity.
Result:
Uncontended low delay results are not reliable
Motivation:
Invoking ChannelHandlers is not free and can result in some overhead when the ChannelPipeline becomes very long. This is especially true if most handlers will just forward the call to the next handler in the pipeline. When the user extends Channel*HandlerAdapter we can easily detect if can just skip the handler and invoke the next handler in the pipeline directly. This reduce the overhead of dispatch but also reduce the call-stack in many cases.
This backports https://github.com/netty/netty/pull/8723 and https://github.com/netty/netty/pull/8987 to 4.1
Modifications:
Detect if we can skip the handler when walking the pipeline.
Result:
Reduce overhead for long pipelines.
Benchmark (extraHandlers) Mode Cnt Score Error Units
DefaultChannelPipelineBenchmark.propagateEventOld 4 thrpt 10 267313.031 ± 9131.140 ops/s
DefaultChannelPipelineBenchmark.propagateEvent 4 thrpt 10 824825.673 ± 12727.594 ops/s
Motivation:
Netty is very widely used which can lead to a lot of pain when we break API / ABI. We should make use japicmp-maven-plugin during the build to verify we do not introduce breakage by mistake.
Modifications:
- Add japicmp-maven-plugin to the build process
- Fix a method signature change in HttpProxyHandler that was flagged as a possible problem.
Result:
Ensure no API/ABI breakage accour between releases.
Motivation:
Just was looking through code and found 1 interesting place DateFormatter.tryParseMonth that was not very effective, so I decided to optimize it a bit.
Modification:
Changed DateFormatter.tryParseMonth method. Instead of invocation regionMatch() for every month - compare chars one by one.
Result:
DateFormatter.parseHttpDate method performance improved from ~3% to ~15%.
Benchmark (DATE_STRING) Mode Cnt Score Error Units
DateFormatter2Benchmark.parseHttpHeaderDateFormatter Sun, 27 Jan 2016 19:18:46 GMT thrpt 6 4142781.221 ± 82155.002 ops/s
DateFormatter2Benchmark.parseHttpHeaderDateFormatter Sun, 27 Dec 2016 19:18:46 GMT thrpt 6 3781810.558 ± 38679.061 ops/s
DateFormatter2Benchmark.parseHttpHeaderDateFormatterNew Sun, 27 Jan 2016 19:18:46 GMT thrpt 6 4372569.705 ± 30257.537 ops/s
DateFormatter2Benchmark.parseHttpHeaderDateFormatterNew Sun, 27 Dec 2016 19:18:46 GMT thrpt 6 4339785.100 ± 57542.660 ops/s
Motivation:
We need to update to a new checkstyle plugin to allow the usage of lambdas.
Modifications:
- Update to new plugin version.
- Fix checkstyle problems.
Result:
Be able to use checkstyle plugin which supports new Java syntax.
Motivation:
Netty executors doesn't have yet any means to compare with each others
nor to compare with the j.u.c. executors
Modifications:
A new benchmark measuring execute burst cost is being added
Result:
It's now possible to compare some of Netty executors with each others
and with the j.u.c. executors
Motivation:
CompositeByteBuf is a powerful and versatile abstraction, allowing for
manipulation of large data without copying bytes. There is still a
non-negligible cost to reading/writing however relative to "singular"
ByteBufs, and this can be mostly eliminated with some rework of the
internals.
My use case is message modification/transformation while zero-copy
proxying. For example replacing a string within a large message with one
of a different length
Modifications:
- No longer slice added buffers and unwrap added slices
- Components store target buf offset relative to position in
composite buf
- Less allocations, object footprint, pointer indirection, offset
arithmetic
- Use Component[] rather than ArrayList<Component>
- Avoid pointer indirection and duplicate bounds check, more
efficient backing array growth
- Facilitates optimization when doing bulk-inserts - inserting n
ByteBufs behind m is now O(m + n) instead of O(mn)
- Avoid unnecessary casting and method call indirection via superclass
- Eliminate some duplicate range/ref checks via non-checking versions of
toComponentIndex and findComponent
- Add simple fast-path for toComponentIndex(0); add racy cache of
last-accessed Component to findComponent(int)
- Override forEachByte0(...) and forEachByteDesc0(...) methods
- Make use of RecyclableArrayList in nioBuffers(int, int) (in line with
FasterCompositeByteBuf impl)
- Modify addComponents0(boolean,int,Iterable) to use the Iterable
directly rather than copy to an array first (and possibly to an
ArrayList before that)
- Optimize addComponents0(boolean,int,ByteBuf[],int) to not perform
repeated array insertions and avoid second loop for offset updates
- Simplify other logic in various places, in particular the general
pattern used where a sub-range is iterated over
- Add benchmarks to demonstrate some improvements
While refactoring I also came across a couple of clear bugs. They are
fixed in these changes but I will open another PR with unit tests and
fixes to the current version.
Result:
Much faster creation, manipulation, and access; many fewer allocations
and smaller footprint. Benchmark results to follow.
* Optimize AbstractByteBuf.getCharSequence() in US_ASCII case
Motivation:
Inspired by https://github.com/netty/netty/pull/8388, I noticed this
simple optimization to avoid char[] allocation (also suggested in a TODO
here).
Modifications:
Return an AsciiString from AbstractByteBuf.getCharSequence() if
requested charset is US_ASCII or ISO_8859_1 (latter thanks to
@Scottmitch's suggestion). Also tweak unit tests not to require Strings
and include a new benchmark to demonstrate the speedup.
Result:
Speed-up of AbstractByteBuf.getCharSequence() in ascii and iso 8859/1
cases
Motiviation:
At the moment whenever ensureAccessible() is called in our ByteBuf implementations (which is basically on each operation) we will do a volatile read. That per-se is not such a bad thing but the problem here is that it will also reduce the the optimizations that the compiler / jit can do. For example as these are volatile it can not eliminate multiple loads of it when inline the methods of ByteBuf which happens quite frequently because most of them a quite small and very hot. That is especially true for all the methods that act on primitives.
It gets even worse as people often call a lot of these after each other in the same method or even use method chaining here.
The idea of the change is basically just ue a non-volatile read for the ensureAccessible() check as its a best-effort implementation to detect acting on already released buffers anyway as even with a volatile read it could happen that the user will release it in another thread before we actual access the buffer after the reference check.
Modifications:
- Try to do a non-volatile read using sun.misc.Unsafe if we can use it.
- Add a benchmark
Result:
Big performance win when multiple ByteBuf methods are called from a method.
With the change:
UnsafeByteBufBenchmark.setGetLongUnsafeByteBuf thrpt 20 281395842,128 ± 5050792,296 ops/s
Before the change:
UnsafeByteBufBenchmark.setGetLongUnsafeByteBuf thrpt 20 217419832,801 ± 5080579,030 ops/s
Motivation:
We should use the latest jmh version which also supports -prof dtraceasm on MacOS.
Modifications:
Update to latest jmh version.
Result:
Better benchmark / profiling support on MacOS.
Motivation:
Some of the flags we used are not supported anymore on more recent JDK versions. We should just remove all of them and only keep what we really need. This may also reflect better what people use in production.
Modifications:
Remove some flags when running the benchmarks.
Result:
Benchmarks also run with JDK11.
Motivation:
It is sometimes useful to be able to run benchmarks easily from the commandline and passs different arguments / options here. We should support this.
Modifications:
Add the benchmark-jar profile which allows to generate such an "uber-jar" that can be used directly to run benchmarks as documented at http://openjdk.java.net/projects/code-tools/jmh/.
Result:
More flexible way to run benchmarks.
Motivation:
Optimizing the Epoll channel needs an objective measure of how fast
it is.
Modification:
Add a simple, closed loop, ping-pong benchmark.
Result:
Benchmark can be used to measure #7816
Initial numbers:
```
Result "io.netty.microbench.channel.epoll.EpollSocketChannelBenchmark.pingPong":
22614.403 ±(99.9%) 797.263 ops/s [Average]
(min, avg, max) = (21093.160, 22614.403, 24977.387), stdev = 918.130
CI (99.9%): [21817.140, 23411.666] (assumes normal distribution)
Benchmark Mode Cnt Score Error Units
EpollSocketChannelBenchmark.pingPong thrpt 20 22614.403 ± 797.263 ops/s
```
Motivation:
The JVM isn't always able to hoist out/reduce bounds checking (due to ref counting operations etc etc) hence making it configurable could improve performances for most CPU intensive use cases.
Modifications:
Each AbstractByteBuf bounds check has been tested against a new static final configuration property similar to checkAccessible ie io.netty.buffer.bytebuf.checkBounds.
Result:
Any user could disable ByteBuf bounds checking in order to get extra performances.
Motivation:
The usage of Invocation level for JMH fixture methods (setup/teardown) inccurs in a significant overhead
in the benchmark time (see org.openjdk.jmh.annotations.Level documentation).
In the case of CodecInputListBenchmark, benchmarks are far too small (less than 50ns) and the Invocation
level setup offsets the measurement considerably.
On such cases, the recommended fix patch is to include the setup/teardown code in the benchmark method.
Modifications:
Include the setup/teardown code in the relevant benchmark methods.
Remove the setup/teardown methods from the benchmark class.
Result:
We run the entire benchmark 10 times with default parameters we observed:
- ArrayList benchmark affected directly by JMH overhead is now from 15-80% faster.
- CodecList benchmark is now 50% faster than original (even with the setup code being measured).
- Recyclable ArrayList is ~30% slower.
- All benchmarks have significant different means (ANOVA) and medians (Moore)
Mode: Throughput (Higher the better)
Method Full params Factor Modified (Median) Original (Median)
recyclableArrayList (elements = 1) 0.615520967 21719082.75 35285691.2
recyclableArrayList (elements = 4) 0.699553431 17149442.76 24514843.31
arrayList (elements = 4) 1.152666631 27120407.18 23528404.88
codecOutList (elements = 1) 1.527275908 67251089.04 44033359.47
codecOutList (elements = 4) 1.596917095 59174088.78 37055204.03
arrayList (elements = 1) 1.878616889 62188238.24 33103204.06
Environment:
Tests run on a Computational server with CPU: E5-1660-3.3GHZ (6 cores + HT), 64 GB RAM.
Motivation:
The usage of Invocation level for JMH fixture methods (setup/teardown) inccurs in a significant impact in
in the benchmark time (see org.openjdk.jmh.annotations.Level documentation).
When the benchmark and the setup/teardown is too small (less than a milisecond) the Invocation level might saturate the system with
timestamp requests and iteration synchronizations which introduce artificial latency, throughput, and scalability bottlenecks.
In the HeadersBenchmark, all benchmarks take less than 100ns and the Invocation level setup offsets the measurement considerably.
As fixture methods is defined for the entire class, this overhead also impacts every single benchmark in this class, not only
the ones that use the emptyHttpHeaders object (cleaned in the setup).
The recommended fix patch here is to include the setup/teardown code in the benchmark where the object is used.
Modifications:
Include the setup/teardown code in the relevant benchmark methods.
Remove the setup/teardown method of Invocation level from the benchmark class.
Result:
We run all benchmarks from HeadersBenchmark 10 times with default parameter, we observe:
- Benchmarks that were not directly affected by the fix patch, improved execution time.
For instance, http2Remove with (exampleHeader = THREE) had its median reported as 2x faster than the original version.
- Benchmarks that had the setup code inserted (eg. http2AddAllFastest) did not suffer a significant punch in the execution time,
as the benchmarks are not dominated by the clear().
Environment:
Tests run on a Computational server with CPU: E5-1660-3.3GHZ (6 cores + HT), 64 GB RAM.
Motivation:
NetUtilBenchmark is using out of date data, throws an exception in the benchmark, and allocates a Set on each run.
Modifications:
- Update the benchmark and reduce each run's overhead
Result:
NetUtilBenchmark is updated.
Motivation:
Read-only heap ByteBuffer doesn't expose array: the existent method to perform copies to direct ByteBuf involves the creation of a (maybe pooled) additional heap ByteBuf instance and copy
Modifications:
To avoid stressing the allocator with additional (and stealth) heap ByteBuf allocations is provided a method to perform copies using the (pooled) internal NIO buffer
Result:
Copies from read-only heap ByteBuffer to direct ByteBuf won't create any intermediate ByteBuf
Motivation:
According to the spec:
All pseudo-header fields MUST appear in the header block before regular
header fields. Any request or response that contains a pseudo-header
field that appears in a header block after
a regular header field MUST be treated as malformed (Section 8.1.2.6).
Pseudo-header fields are only valid in the context in which they are defined.
Pseudo-header fields defined for requests MUST NOT appear in responses;
pseudo-header fields defined for responses MUST NOT appear in requests.
Pseudo-header fields MUST NOT appear in trailers.
Endpoints MUST treat a request or response that contains undefined or
invalid pseudo-header fields as malformed (Section 8.1.2.6).
Clients MUST NOT accept a malformed response. Note that these requirements
are intended to protect against several types of common attacks against HTTP;
they are deliberately strict because being permissive can expose
implementations to these vulnerabilities.
Modifications:
- Introduce validation in HPackDecoder
Result:
- Requests with unknown pseudo-field headers are rejected
- Requests with containing response specific pseudo-headers are rejected
- Requests where pseudo-header appear after regular header are rejected
- h2spec 8.1.2.1 pass
Motivation:
We used Recycler for the CodecOutputList which is not optimized for the use-case of access only from the same Thread all the time.
Modifications:
- Use FastThreadLocal for CodecOutputList
- Add benchmark
Result:
Less overhead in our codecs.
Motivation:
The JMH doc suggests to use BlackHoles to avoid dead code elimination hence would be better to follow this best practice.
Modifications:
Each benchmark method is returning the ByteBuf/ByteBuffer to avoid the JVM to perform any dead code elimination.
Result:
The results are more reliable and comparable to the others provided by other ByteBuf benchmarks (eg HeapByteBufBenchmark)
Motivation:
HttpMethod#valueOf shows up on profiler results in the top set of
results. Since it is a relatively simple operation it can be improved in
isolation.
Modifications:
- Introduce a special case map which assigns each HttpMethod to a unique
index in an array and provides constant time lookup from a hash code
algorithm. When the bucket is matched we can then directly do equality
comparison instead of potentially following a linked structure when
HashMap has hash collisions.
Result:
~10% improvement in benchmark results for HttpMethod#valueOf
Benchmark Mode Cnt Score Error Units
HttpMethodMapBenchmark.newMapKnownMethods thrpt 16 31.831 ± 0.928 ops/us
HttpMethodMapBenchmark.newMapMixMethods thrpt 16 25.568 ± 0.400 ops/us
HttpMethodMapBenchmark.newMapUnknownMethods thrpt 16 51.413 ± 1.824 ops/us
HttpMethodMapBenchmark.oldMapKnownMethods thrpt 16 29.226 ± 0.330 ops/us
HttpMethodMapBenchmark.oldMapMixMethods thrpt 16 21.073 ± 0.247 ops/us
HttpMethodMapBenchmark.oldMapUnknownMethods thrpt 16 49.081 ± 0.577 ops/us
Motivation:
DefaultHttp2FrameWriter#writeData allocates a DataFrameHeader for each write operation. DataFrameHeader maintains internal state and allocates multiple slices of a buffer which is a maximum of 30 bytes. This 30 byte buffer may not always be necessary and the additional slice operations can utilize retainedSlice to take advantage of pooled objects. We can also save computation and object allocations if there is no padding which is a common case in practice.
Modifications:
- Remove DataFrameHeader
- Add a fast path for padding == 0
Result:
Less object allocation in DefaultHttp2FrameWriter
Motivation:
`AbstractScheduledEventExecutor` uses a standard `java.util.PriorityQueue` to keep track of task deadlines. `ScheduledFuture.cancel` removes tasks from this `PriorityQueue`. Unfortunately, `PriorityQueue.remove` has `O(n)` performance since it must search for the item in the entire queue before removing it. This is fast when the future is at the front of the queue (e.g., already triggered) but not when it's randomly located in the queue.
Many servers will use `ScheduledFuture.cancel` on all requests, e.g., to manage a request timeout. As these cancellations will be happen in arbitrary order, when there are many scheduled futures, `PriorityQueue.remove` is a bottleneck and greatly hurts performance with many concurrent requests (>10K).
Modification:
Use netty's `DefaultPriorityQueue` for scheduling futures instead of the JDK. `DefaultPriorityQueue` is almost identical to the JDK version except it is able to remove futures without searching for them in the queue. This means `DefaultPriorityQueue.remove` has `O(log n)` performance.
Result:
Before - cancelling futures has varying performance, capped at `O(n)`
After - cancelling futures has stable performance, capped at `O(log n)`
Benchmark results
After - cancelling in order and in reverse order have similar performance within `O(log n)` bounds
```
Benchmark (num) Mode Cnt Score Error Units
ScheduledFutureTaskBenchmark.cancelInOrder 100 thrpt 20 137779.616 ± 7709.751 ops/s
ScheduledFutureTaskBenchmark.cancelInOrder 1000 thrpt 20 11049.448 ± 385.832 ops/s
ScheduledFutureTaskBenchmark.cancelInOrder 10000 thrpt 20 943.294 ± 12.391 ops/s
ScheduledFutureTaskBenchmark.cancelInOrder 100000 thrpt 20 64.210 ± 1.824 ops/s
ScheduledFutureTaskBenchmark.cancelInReverseOrder 100 thrpt 20 167531.096 ± 9187.865 ops/s
ScheduledFutureTaskBenchmark.cancelInReverseOrder 1000 thrpt 20 33019.786 ± 4737.770 ops/s
ScheduledFutureTaskBenchmark.cancelInReverseOrder 10000 thrpt 20 2976.955 ± 248.555 ops/s
ScheduledFutureTaskBenchmark.cancelInReverseOrder 100000 thrpt 20 362.654 ± 45.716 ops/s
```
Before - cancelling in order and in reverse order have significantly different performance at higher queue size, orders of magnitude worse than the new implementation.
```
Benchmark (num) Mode Cnt Score Error Units
ScheduledFutureTaskBenchmark.cancelInOrder 100 thrpt 20 139968.586 ± 12951.333 ops/s
ScheduledFutureTaskBenchmark.cancelInOrder 1000 thrpt 20 12274.420 ± 337.800 ops/s
ScheduledFutureTaskBenchmark.cancelInOrder 10000 thrpt 20 958.168 ± 15.350 ops/s
ScheduledFutureTaskBenchmark.cancelInOrder 100000 thrpt 20 53.381 ± 13.981 ops/s
ScheduledFutureTaskBenchmark.cancelInReverseOrder 100 thrpt 20 123918.829 ± 3642.517 ops/s
ScheduledFutureTaskBenchmark.cancelInReverseOrder 1000 thrpt 20 5099.810 ± 206.992 ops/s
ScheduledFutureTaskBenchmark.cancelInReverseOrder 10000 thrpt 20 72.335 ± 0.443 ops/s
ScheduledFutureTaskBenchmark.cancelInReverseOrder 100000 thrpt 20 0.743 ± 0.003 ops/s
```
Motivation:
Highly retained and released objects have contention on their ref
count. Currently, the ref count is updated using compareAndSet
with care to make sure the count doesn't overflow, double free, or
revive the object.
Profiling has shown that a non trivial (~1%) of CPU time on gRPC
latency benchmarks is from the ref count updating.
Modification:
Rather than pessimistically assuming the ref count will be invalid,
optimistically update it assuming it will be. If the update was
wrong, then use the slow path to revert the change and throw an
execption. Most of the time, the ref counts are correct.
This changes from using compareAndSet to getAndAdd, which emits a
different CPU instruction on x86 (CMPXCHG to XADD). Because the
CPU knows it will modifiy the memory, it can avoid contention.
On a highly contended machine, this can be about 2x faster.
There is a downside to the new approach. The ref counters can
temporarily enter invalid states if over retained or over released.
The code does handle these overflow and underflow scenarios, but it
is possible that another concurrent access may push the failure to
a different location. For example:
Time 1 Thread 1: obj.retain(INT_MAX - 1)
Time 2 Thread 1: obj.retain(2)
Time 2 Thread 2: obj.retain(1)
Previously Thread 2 would always succeed and Thread 1 would always
fail on the second access. Now, thread 2 could fail while thread 1
is rolling back its change.
====
There are a few reasons why I think this is okay:
1. Buggy code is going to have bugs. An exception _is_ going to be
thrown. This just causes the other threads to notice the state
is messed up and stop early.
2. If high retention counts are a use case, then ref count should
be a long rather than an int.
3. The critical section is greatly reduced compared to the previous
version, so the likelihood of this happening is lower
4. On error, the code always rollsback the change atomically, so
there is no possibility of corruption.
Result:
Faster refcounting
```
BEFORE:
Benchmark (delay) Mode Cnt Score Error Units
AbstractReferenceCountedByteBufBenchmark.retainRelease_contended 1 sample 2901361 804.579 ± 1.835 ns/op
AbstractReferenceCountedByteBufBenchmark.retainRelease_contended 10 sample 3038729 785.376 ± 16.471 ns/op
AbstractReferenceCountedByteBufBenchmark.retainRelease_contended 100 sample 2899401 817.392 ± 6.668 ns/op
AbstractReferenceCountedByteBufBenchmark.retainRelease_contended 1000 sample 3650566 2077.700 ± 0.600 ns/op
AbstractReferenceCountedByteBufBenchmark.retainRelease_contended 10000 sample 3005467 19949.334 ± 4.243 ns/op
AbstractReferenceCountedByteBufBenchmark.retainRelease_uncontended 1 sample 456091 48.610 ± 1.162 ns/op
AbstractReferenceCountedByteBufBenchmark.retainRelease_uncontended 10 sample 732051 62.599 ± 0.815 ns/op
AbstractReferenceCountedByteBufBenchmark.retainRelease_uncontended 100 sample 778925 228.629 ± 1.205 ns/op
AbstractReferenceCountedByteBufBenchmark.retainRelease_uncontended 1000 sample 633682 2002.987 ± 2.856 ns/op
AbstractReferenceCountedByteBufBenchmark.retainRelease_uncontended 10000 sample 506442 19735.345 ± 12.312 ns/op
AFTER:
Benchmark (delay) Mode Cnt Score Error Units
AbstractReferenceCountedByteBufBenchmark.retainRelease_contended 1 sample 3761980 383.436 ± 1.315 ns/op
AbstractReferenceCountedByteBufBenchmark.retainRelease_contended 10 sample 3667304 474.429 ± 1.101 ns/op
AbstractReferenceCountedByteBufBenchmark.retainRelease_contended 100 sample 3039374 479.267 ± 0.435 ns/op
AbstractReferenceCountedByteBufBenchmark.retainRelease_contended 1000 sample 3709210 2044.603 ± 0.989 ns/op
AbstractReferenceCountedByteBufBenchmark.retainRelease_contended 10000 sample 3011591 19904.227 ± 18.025 ns/op
AbstractReferenceCountedByteBufBenchmark.retainRelease_uncontended 1 sample 494975 52.269 ± 8.345 ns/op
AbstractReferenceCountedByteBufBenchmark.retainRelease_uncontended 10 sample 771094 62.290 ± 0.795 ns/op
AbstractReferenceCountedByteBufBenchmark.retainRelease_uncontended 100 sample 763230 235.044 ± 1.552 ns/op
AbstractReferenceCountedByteBufBenchmark.retainRelease_uncontended 1000 sample 634037 2006.578 ± 3.574 ns/op
AbstractReferenceCountedByteBufBenchmark.retainRelease_uncontended 10000 sample 506284 19742.605 ± 13.729 ns/op
```
