tikv_util: introduce future channel #13407
Conversation
It's a super set of batch channel. It can be used as a batch channel or just a future channel. Signed-off-by: Jay Lee <BusyJayLee@gmail.com>
|
[REVIEW NOTIFICATION] This pull request has been approved by:
To complete the pull request process, please ask the reviewers in the list to review by filling The full list of commands accepted by this bot can be found here. Reviewer can indicate their review by submitting an approval review. |
|
/release |
| return Poll::Ready(Some(t)); | ||
| } | ||
| queue.waker.register(cx.waker()); | ||
| if let Some(t) = queue.queue.pop() { |
There was a problem hiding this comment.
add a comment for why try pop twice
| if let Some(t) = queue.queue.pop() { | ||
| return Poll::Ready(Some(t)); | ||
| } | ||
| queue.waker.register(cx.waker()); |
There was a problem hiding this comment.
What if the below pop consume the element, then the next wake would be unnecessary.
There was a problem hiding this comment.
I don't get it. What do you mean "consume the element"? User is expected to keep calling poll_next until None is returned.
There was a problem hiding this comment.
I mean, should it be
if let Some(t) = queue.queue.pop() {
return Poll::Ready(Some(t));
}
if let Some(t) = queue.queue.pop() {
return Poll::Ready(Some(t));
}
queue.waker.register(cx.waker());
``
There was a problem hiding this comment.
It's for edge cases only, which should hardly happen. If that happens, clear the waker may cause unnecessary contention.
Signed-off-by: Jay Lee <BusyJayLee@gmail.com>
|
Benched with sysbench read_write, insert, update index, point select, the differences in througput and latency are within 3%. PTAL |
| let elem = ctx.elem.take(); | ||
| if let Some(m) = received { | ||
| let collection = ctx.elem.get_or_insert_with(&ctx.initializer); | ||
| let _received = ctx.collector.collect(collection, m); |
There was a problem hiding this comment.
Otherwise received will be dropped and message is lost.
There was a problem hiding this comment.
The logic is simplified, now it won't collect again.
| } | ||
| } | ||
|
|
||
| unsafe impl<T: Send> Send for Sender<T> {} |
There was a problem hiding this comment.
What about also Sync? Sometimes non-Sync types may cause difficulties when Sync bounds are required.
| break false; | ||
| } | ||
| count += 1; | ||
| if count >= ctx.max_batch_size { |
There was a problem hiding this comment.
If max_batch_size belongs to BatchReceiver, in which case do we expect BatchCollector::collect returns Some and split?
There was a problem hiding this comment.
max_batch_size is for the number of message, BatchCollector can add other limits like data size.
There was a problem hiding this comment.
The limit is for raft client in the past. Now raft client uses its own batch algorithm, so the API is unnecessary. I remove it for simplicity.
Signed-off-by: Jay Lee <BusyJayLee@gmail.com>
| /// `initializer` is used to generate a initial value, and `collector` | ||
| /// will collect every (at most `max_batch_size`) raw items into the | ||
| /// batched value. | ||
| pub fn new(rx: Receiver<T>, max_batch_size: usize, initializer: I, collector: C) -> Self { |
There was a problem hiding this comment.
What's the use case for the collector? Isn't vec adequate?
There was a problem hiding this comment.
Check
Lines 2357 to 2369 in 8e311cd
ti-chi-bot
added
status/LGT2
|
/merge |
|
@BusyJay: It seems you want to merge this PR, I will help you trigger all the tests: /run-all-tests You only need to trigger If you have any questions about the PR merge process, please refer to pr process. Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the ti-community-infra/tichi repository. |
|
This pull request has been accepted and is ready to merge. Commit hash: 74d4d82
|
|
@BusyJay: Your PR was out of date, I have automatically updated it for you. At the same time I will also trigger all tests for you: /run-all-tests If the CI test fails, you just re-trigger the test that failed and the bot will merge the PR for you after the CI passes. Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the ti-community-infra/tichi repository. |
This reverts commit 7382d4e. Signed-off-by: tabokie <xy.tao@outlook.com>
This reverts commit 7382d4e.
ref tikv#13394, ref tikv#13407 Signed-off-by: tabokie <xy.tao@outlook.com>
ref tikv#13394, ref tikv#13407 Signed-off-by: tabokie <xy.tao@outlook.com>
What is changed and how it works?
Issue Number: Ref #12842
What's Changed:
I bench it with the old implementation, the result is highly unstable as if the receiver goes into sleep state, it will take more time to be woken up. Anyway, I post the result as below. There is no big difference between implementations.
Check List
Tests
Release note