Nitsan Wakart

Use shaded dependency on JCTools instead of copy and paste

7

Currently this library uses some queue implementations lifted in spirit and to some extent code out of JCTools. The 2 projects share a contributor, @akarnokd. I would recommend this project...

Add support for color scheme controls

1

Thanks for a great tool! In B.Gregg reference implementation we have the --color option where users can pick a color scheme, and a preset color map which can be used...

Improve documentation on queue progress guarantees

3

Currently the java doc does not state, or sometimes incorrectly states, the progress guarantees (wait free, lock free, blocking etc) and edge cases on JCTools queues. This can lead to...

[Channels] NOTE: Channels unaligned memory access considerations

3

The SPSC fixed sized ring buffer is using an int marker for publication: ``` private boolean isMessageReady(long offset) { return UNSAFE.getIntVolatile(null, offset) == READY_MESSAGE_INDICATOR; } private void busyIndicator(long offset) {...

[Channels] Add a set of benchmarks for comparing Channel vs. Disruptor vs. queue

1

The premise of the Channel approach is that inlining the data into the ring buffer is good for locality and should therefore improve performance. We should be able to demonstrate...

Release test artifact for shaders

1

see https://github.com/ben-manes/caffeine/issues/127#issuecomment-257176295

Implement intrusive linked queue MPSC

14

This was found to be a useful usecase in Netty for internal uses where long/medium lived events are passed between threads and we wish to avoid the allocation of nodes...

Actively encourage proper dependency instead of C&P from other libs

3

This is meant to track instances of other frameworks/libraries/applications using JCTools but instead of depending on a particular tested and released version relying on copy and pasted source code. Currently...

RFE: Implement single writer map

1

Splitting from #31

Use runtime code generation to create implementations

3

This allows: 1. minimal boilerplate for padding (+ runtime decision on pad size) 2. parametrized constants 3. code optimization based on runtime info The trade off is code bloat.