[WIP]Performance[mqbi::DispatcherEvent]: replace multiple inheritance with variant by 678098 · Pull Request #396 · bloomberg/blazingmq

678098 · 2024-08-09T23:58:45Z

Before

sizeof(mqbi::DispatcherEvent): 880
mqbi::DispatcherEvent::reset():
       total: 708.62 ms (100000000 iterations)
    per call: 7 ns

After

sizeof(mqbi::DispatcherEvent): 320
sizeof(mqbi::DispatcherPutEvent): 128
sizeof(mqbi::DispatcherPushEvent): 288
sizeof(mqbi::DispatcherDispatcherEvent): 152
sizeof(mqbi::DispatcherCallbackEvent): 80
sizeof(mqbi::DispatcherReceiptEvent): 32
sizeof(mqbi::DispatcherAckEvent): 80
sizeof(mqbi::DispatcherControlMessageEvent): 192
mqbi::DispatcherEvent::reset():
       total: 51.14 ms (100000000 iterations)
    per call: 0 ns

hallfox · 2024-08-12T16:23:07Z

src/groups/mqb/mqbi/mqbi_dispatcher.h

-    DispatcherEvent& setType(DispatcherEventType::Enum value);
+    inline DispatcherPutEvent& makePutEvent()
+    {
+        d_type = DispatcherEventType::e_PUT;


Do we even need to set this type anymore?

hallfox · 2024-08-12T16:25:58Z

src/groups/mqb/mqbi/mqbi_dispatcher.h

+    bsl::shared_ptr<bdlbb::Blob> d_blob_sp;
+    // Blob of data embedded in this
+    // event. Refer to the corresponding
+    // accessor on the various
+    // DispatcherEvent view interfaces
+    // for more specific information.


Suggested change

bsl::shared_ptr<bdlbb::Blob> d_blob_sp;

// Blob of data embedded in this

// event. Refer to the corresponding

// accessor on the various

// DispatcherEvent view interfaces

// for more specific information.

/// Blob of data embedded in this

/// event. Refer to the corresponding

/// accessor on the various

/// DispatcherEvent view interfaces

/// for more specific information.

bsl::shared_ptr<bdlbb::Blob> d_blob_sp;

Docstring style

hallfox · 2024-08-12T16:26:29Z

src/groups/mqb/mqbi/mqbi_dispatcher.h

+    }
+
    /// Destructor.
    virtual ~DispatcherAckEvent();


Do we need virtual destructors still?

No, will clean up this

hallfox · 2024-08-12T16:28:01Z

src/groups/mqb/mqbc/mqbc_clusternodesession.h

+    // TODO refactor
+    if (mqbi::DispatcherEventType::e_PUSH == ev.type()) {
+        const_cast<mqbi::DispatcherPushEvent*>(ev.asPushEvent())
+            ->setClusterNode(d_clusterNode_p);


Is the refactoring here to make asPushEvent and family to have a const and non-const overload?

hallfox · 2024-08-12T16:34:51Z

src/groups/mqb/mqba/mqba_dispatcher.cpp

-        .setType(mqbi::DispatcherEventType::e_CALLBACK)
-        .setCallback(mqbi::Dispatcher::voidToProcessorFunctor(f))
-        .setDestination(const_cast<mqbi::DispatcherClient*>(d_client_p));
+        .setDestination(const_cast<mqbi::DispatcherClient*>(d_client_p))


Little wary of this const cast, what's the reason this is okay?

678098 · 2024-08-12T17:04:03Z

src/groups/mqb/mqbblp/mqbblp_queue.cpp

+        .makePushEvent()
        .setBlob(appData)
        .setOptions(options)
        .setGuid(msgGUID)


Note that initially we are building DispatcherEvent, and after makePushEvent() we are calling setters for DispatcherPushEvent. So we switch between objects amidst these setters calls. This interface is not perfect, but I have plans to replace it by calling constructors with all args instead.

So instead of this

(*dispEvent) .setSource(this) // DispatcherEvent .makePushEvent() // !!! DispatcherPushEvent .setBlob(appData) // DispatcherPushEvent .setOptions(options) // DispatcherPushEvent .setGuid(msgGUID) // DispatcherPushEvent

We will have smth like this:

(*dispEvent) .setSource(this) .makePushEvent(appData, options, msgGUID, ...)

I plan to do it in a separate PR, with performance benchmarks, to limit the number of changes in this PR.

bmq-oss-ci

Build 2891 of commit 4968e09 has completed with FAILURE

Signed-off-by: Evgeny Malygin <emalygin@bloomberg.net>

Signed-off-by: Taylor Foxhall <tfoxhall@bloomberg.net>

bmq-oss-ci

Build 2977 of commit 7895323 has completed with FAILURE

678098 · 2026-03-04T20:20:19Z

Replaced by #1116

678098 requested a review from a team as a code owner August 9, 2024 23:58

678098 force-pushed the t2463_DispatcherEvent_opt branch from e16b562 to ef96296 Compare August 10, 2024 00:19

hallfox reviewed Aug 12, 2024

View reviewed changes

678098 commented Aug 12, 2024

View reviewed changes

678098 force-pushed the t2463_DispatcherEvent_opt branch from 5870886 to adf9611 Compare August 14, 2024 18:14

678098 force-pushed the t2463_DispatcherEvent_opt branch 2 times, most recently from 9c1186a to 4f7f605 Compare September 18, 2024 16:29

678098 force-pushed the t2463_DispatcherEvent_opt branch from 4f7f605 to 2f56f75 Compare October 8, 2024 19:37

678098 mentioned this pull request Oct 8, 2024

Refactor[mqbblp::Cluster]: use concrete event types #448

Merged

678098 force-pushed the t2463_DispatcherEvent_opt branch 2 times, most recently from 6f1e149 to 4968e09 Compare October 24, 2024 17:43

bmq-oss-ci bot reviewed Jul 11, 2025

View reviewed changes

678098 force-pushed the t2463_DispatcherEvent_opt branch 2 times, most recently from 86ec727 to 615c13a Compare August 22, 2025 15:28

678098 and others added 3 commits August 22, 2025 15:38

Perf[mqbi::DispatcherEvent]: replace multiple inheritance with variant

c8766cf

Signed-off-by: Evgeny Malygin <emalygin@bloomberg.net>

Update to use new managed callback object

a1bf506

Signed-off-by: Taylor Foxhall <tfoxhall@bloomberg.net>

Refactor dispatcher types to out of line definitions

7895323

Signed-off-by: Taylor Foxhall <tfoxhall@bloomberg.net>

678098 force-pushed the t2463_DispatcherEvent_opt branch from 615c13a to 7895323 Compare August 22, 2025 15:42

bmq-oss-ci bot reviewed Aug 22, 2025

View reviewed changes

678098 mentioned this pull request Nov 25, 2025

Refactor[bmqc::MQTP]: ObjectPool -> SharedObjectPool #1037

Merged

678098 closed this Mar 4, 2026

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[WIP]Performance[mqbi::DispatcherEvent]: replace multiple inheritance with variant#396

[WIP]Performance[mqbi::DispatcherEvent]: replace multiple inheritance with variant#396
678098 wants to merge 3 commits intobloomberg:mainfrom
678098:t2463_DispatcherEvent_opt

678098 commented Aug 9, 2024 •

edited

Loading

Uh oh!

hallfox Aug 12, 2024

Uh oh!

hallfox Aug 12, 2024

Uh oh!

hallfox Aug 12, 2024

Uh oh!

678098 Aug 12, 2024

Uh oh!

hallfox Aug 12, 2024

Uh oh!

hallfox Aug 12, 2024

Uh oh!

678098 Aug 12, 2024 •

edited

Loading

Uh oh!

bmq-oss-ci bot left a comment

Uh oh!

bmq-oss-ci bot left a comment

Uh oh!

678098 commented Mar 4, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

678098 commented Aug 9, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

hallfox Aug 12, 2024

Choose a reason for hiding this comment

Uh oh!

hallfox Aug 12, 2024

Choose a reason for hiding this comment

Uh oh!

hallfox Aug 12, 2024

Choose a reason for hiding this comment

Uh oh!

678098 Aug 12, 2024

Choose a reason for hiding this comment

Uh oh!

hallfox Aug 12, 2024

Choose a reason for hiding this comment

Uh oh!

hallfox Aug 12, 2024

Choose a reason for hiding this comment

Uh oh!

678098 Aug 12, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

bmq-oss-ci bot left a comment

Choose a reason for hiding this comment

Uh oh!

bmq-oss-ci bot left a comment

Choose a reason for hiding this comment

Uh oh!

678098 commented Mar 4, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

678098 commented Aug 9, 2024 •

edited

Loading

678098 Aug 12, 2024 •

edited

Loading