performance(Drawable): Adjusts Turret Positioning, Recoil and Muzzle for Model Draw to Update Only when Necessary by IamInnocent3X · Pull Request #2046 · TheSuperHackers/GeneralsGameCode

IamInnocent3X · 2026-01-01T16:42:57Z

Modifies the GameLogic to tell handleClientTurretPositioning() and handleClientRecoil() function in W3DModelDraw to only update when Necessary.

Skyaero42

Any indication on the performance gain?

This may need extensive testing for mismatching.

Skyaero42 · 2026-01-02T18:25:06Z

+//-------------------------------------------------------------------------------------------------
+void W3DModelDraw::setNeedUpdateTurretPositioning(Bool set)
+{
+	m_needUpdateTurretPosition = set; // A simple function with the dangers of an atomic bomb, misuse and it'll cause desync


Comment not necessary.

Skyaero42 · 2026-01-02T18:26:23Z

 	if( xfer->getXferMode() == XFER_LOAD && m_subObjectVec.empty() == FALSE )
 		updateSubObjects();

+#if !RETAIL_COMPATIBLE_CRC


Does this require a xfer version change?

IamInnocent3X · 2026-01-02T23:58:50Z

Any indication on the performance gain?

I am not sure, but the devs does mention it being not a major problem.

Line 2494 @todo fix me someday (srj)

handleClientRecoil function are majority just For Loop Checks, each time checks with for loops based on how many Maximum Weapons are in game, (In this case, 3), and checks for each Weapon Slot whether there are any recoil to handle. If it does, it handles the Model Bone Recoil and handles their Muzzle from showing. Every Barrel Bone present will require checks.

handleClientTurretPositioning however, is O(n) based. It calls for all Turrets for each Weapon Slot to update Every Frame regardless if there are any changes. For each turret in the runtime it calls for AIUpdateInterface to get their Turret Pitch and Angle, then Transforms the Matrix up to 2 Times (One for Angle, One for Pitch). This is called every frame.

The changes just add a go/stop variable at the beginning of both functions, (m_doHandleRecoil, and m_needUpdateTurretPosition) to inform W3DModelDraw to allow the calling of the respective functions.

stephanmeesters · 2026-01-03T12:04:34Z

I think this can be simplified by removing the changes to xfer and instead forcing a recompute right after you load the savegame, like this (do test it):

diff --git a/GeneralsMD/Code/GameEngine/Source/GameLogic/Object/Object.cpp b/GeneralsMD/Code/GameEngine/Source/GameLogic/Object/Object.cpp
  @@ void Object::loadPostProcess()
   	if( m_xferContainedByID != INVALID_ID )
   		m_containedBy = TheGameLogic->findObjectByID(m_xferContainedByID);
   	else
   		m_containedBy = NULL;
  +
  +	setNeedUpdateTurretPositioning(TRUE);

diff --git a/GeneralsMD/Code/GameEngineDevice/Source/W3DDevice/GameClient/Drawable/Draw/W3DModelDraw.cpp b/GeneralsMD/Code/GameEngineDevice/
  Source/W3DDevice/GameClient/Drawable/Draw/W3DModelDraw.cpp
  @@ void W3DModelDraw::loadPostProcess( void )
   	// extend base class
   	DrawModule::loadPostProcess();
  +
  +	m_needUpdateTurretPosition = TRUE;
  +	m_doHandleRecoil = TRUE;

This may change the CRC right after loading a savegame but I don't see how that would be a problem

IamInnocent3X · 2026-01-04T12:16:30Z

I think this can be simplified by removing the changes to xfer and instead forcing a recompute right after you load the savegame, like this (do test it):

diff --git a/GeneralsMD/Code/GameEngine/Source/GameLogic/Object/Object.cpp b/GeneralsMD/Code/GameEngine/Source/GameLogic/Object/Object.cpp
  @@ void Object::loadPostProcess()
   	if( m_xferContainedByID != INVALID_ID )
   		m_containedBy = TheGameLogic->findObjectByID(m_xferContainedByID);
   	else
   		m_containedBy = NULL;
  +
  +	setNeedUpdateTurretPositioning(TRUE);

diff --git a/GeneralsMD/Code/GameEngineDevice/Source/W3DDevice/GameClient/Drawable/Draw/W3DModelDraw.cpp b/GeneralsMD/Code/GameEngineDevice/
  Source/W3DDevice/GameClient/Drawable/Draw/W3DModelDraw.cpp
  @@ void W3DModelDraw::loadPostProcess( void )
   	// extend base class
   	DrawModule::loadPostProcess();
  +
  +	m_needUpdateTurretPosition = TRUE;
  +	m_doHandleRecoil = TRUE;

This may change the CRC right after loading a savegame but I don't see how that would be a problem

This change seems logical, tested on my end from loading a retail compatible save and it seems fine. I'll adjust and update my commit.

Caball009

I don't think the turret positioning implementation is right.

I built a couple of firebases. I moved the camera somewhere else and then back to the firebases and the turrets snap to a certain angle. It looks like a purely visual thing, though, because it didn't cause a mismatch with a client that didn't have PR.

IamInnocent3X · 2026-06-30T09:15:52Z

Alright, thanks for reporting. Pushed an update, it should fix it.

Edit: Nevermind, the issue still persist, just less noticable. I'm looking into it.

greptile-apps · 2026-06-30T09:30:21Z

Greptile Summary

This PR introduces a guard mechanism to skip handleClientTurretPositioning() and handleClientRecoil() when no turret or recoil animation is actively changing, reducing unnecessary bone updates each frame.

Adds m_needUpdateTurretPosition / m_lastNeedUpdateTurretPosition to W3DModelDraw with a one-trailing-frame pattern so bones settle to their final position before updates are gated off; TurretAI state machine sites (AimTurret, RecenterTurret, IdleScanState, setTurretTargetObject/Position, setTurretEnabled) now call Object::setNeedUpdateTurretPositioning to drive the flag.
Adds m_doHandleRecoil to W3DModelDraw, reset to FALSE at the top of handleClientRecoil() and re-armed to TRUE by any barrel still in RECOIL_START/RECOIL/SETTLE state, ensuring the function exits early only when every barrel is truly idle.
Three new if statements place their body on the same line as the condition, violating the team's style rule (rule 16b9b669); these are flagged as suggestions.

Confidence Score: 5/5

This PR is safe to merge. The optimization is well-scoped: updates only skip when every turret and every barrel is idle, and a one-trailing-frame mechanism ensures bones settle to their final position before gating off.
The flag propagation chain (TurretAI → Object → Drawable → W3DModelDraw) is correctly wired, the recoil accumulation pattern covers all active barrel states, and the change-detection dedup at the Object layer avoids redundant virtual dispatch. The only findings are style nits on three new if-statements.
No files require special attention beyond the three same-line if-statement style fixes already suggested.

Important Files Changed

Filename	Overview
Core/GameEngineDevice/Source/W3DDevice/GameClient/Drawable/Draw/W3DModelDraw.cpp	Core of the optimization: adds dual-flag early-exit to handleClientTurretPositioning (one trailing frame to settle bones) and a reset-then-accumulate pattern for m_doHandleRecoil in handleClientRecoil. The recoil logic correctly covers RECOIL_START/RECOIL (via fall-through) and SETTLE states. Two style violations (same-line if bodies) in the new code.
GeneralsMD/Code/GameEngine/Source/GameLogic/AI/TurretAI.cpp	Inserts setNeedUpdateTurretPositioning calls in every state that causes bone movement: setTurretTargetObject, setTurretTargetPosition, recenterTurret, setTurretEnabled, TurretAIAimTurretState, TurretAIRecenterTurretState, and TurretAIIdleScanState. One style rule violation (same-line if in setTurretEnabled).
GeneralsMD/Code/GameEngine/Source/GameLogic/Object/Object.cpp	Adds m_turretNeedPositioning (initialized FALSE) with a change-detection gate so the drawable is only notified when the boolean actually flips, avoiding redundant per-frame virtual dispatch through the draw module chain.
GeneralsMD/Code/GameEngine/Source/GameClient/Drawable.cpp	Adds setNeedUpdateTurretPositioning that fans out to all draw modules via ObjectDrawInterface, guarded by the null-check on di. Straightforward delegation, no issues.
GeneralsMD/Code/GameEngine/Include/Common/DrawModule.h	Adds setNeedUpdateTurretPositioning as a pure virtual on ObjectDrawInterface. Only W3DModelDraw inherits this interface, so no other concrete class is broken.
Core/GameEngineDevice/Include/W3DDevice/GameClient/Module/W3DModelDraw.h	Adds the three new member booleans and the virtual override declaration. Clean header change.
GeneralsMD/Code/GameEngine/Include/GameLogic/Object.h	Adds m_turretNeedPositioning member and setNeedUpdateTurretPositioning declaration. Aligned with surrounding member declarations.
GeneralsMD/Code/GameEngine/Include/GameClient/Drawable.h	Adds setNeedUpdateTurretPositioning declaration. Clean header change, no issues.

Sequence Diagram

sequenceDiagram
    participant TAI as TurretAI State Machine
    participant Obj as Object
    participant Drw as Drawable
    participant MDraw as W3DModelDraw

    Note over TAI,MDraw: Turret starts moving (e.g. new target)
    TAI->>Obj: setNeedUpdateTurretPositioning(TRUE)
    Obj->>Obj: m_turretNeedPositioning FALSE→TRUE (changed)
    Obj->>Drw: setNeedUpdateTurretPositioning(TRUE)
    Drw->>MDraw: setNeedUpdateTurretPositioning(TRUE)
    MDraw->>MDraw: "m_needUpdateTurretPosition=true, m_lastNeedUpdateTurretPosition=true"

    loop Each client frame (turret moving)
        MDraw->>MDraw: "handleClientTurretPositioning()<br/>m_need=true → runs update"
    end

    Note over TAI,MDraw: Turret reaches target angle
    TAI->>Obj: setNeedUpdateTurretPositioning(FALSE)
    Obj->>Obj: m_turretNeedPositioning TRUE→FALSE (changed)
    Obj->>Drw: setNeedUpdateTurretPositioning(FALSE)
    Drw->>MDraw: setNeedUpdateTurretPositioning(FALSE)
    MDraw->>MDraw: "m_needUpdateTurretPosition=false<br/>(m_last stays true)"

    MDraw->>MDraw: "handleClientTurretPositioning()<br/>m_need=false, m_last=true → runs trailing update<br/>sets m_last=false"

    MDraw->>MDraw: "handleClientTurretPositioning()<br/>m_need=false, m_last=false → early return ✓"

    Note over TAI,MDraw: Weapon fires — recoil path
    TAI->>MDraw: handleWeaponFireFX()
    MDraw->>MDraw: "m_doHandleRecoil=TRUE, state=RECOIL_START"

    loop Each frame while barrel active
        MDraw->>MDraw: "handleClientRecoil()<br/>reset m_doHandleRecoil=FALSE<br/>barrel in RECOIL→sets TRUE<br/>barrel in SETTLE→sets TRUE"
    end

    MDraw->>MDraw: "handleClientRecoil()<br/>SETTLE→IDLE: m_doHandleRecoil stays FALSE<br/>Next frame: early return ✓"

_{Reviews (5): Last reviewed commit: "Updated handleRecoil" | Re-trigger Greptile}

greptile-apps · 2026-06-30T09:30:25Z

+	if(!m_needUpdateTurretPosition && !m_lastNeedUpdateTurretPosition)
+		return;
+
+	m_lastNeedUpdateTurretPosition = m_needUpdateTurretPosition;


Single Object flag spans all turret slots

m_needUpdateTurretPosition is one Bool for the entire object, but handleClientTurretPositioning loops over MAX_TURRETS. Any path that clears the flag (e.g. a two-turret unit where RecenterTurretState finishes for slot 1) suppresses updates for every other turret slot. A per-slot flag array, or a reference-counted active-turret counter, would prevent one turret's quiescence from silencing a sibling that is still rotating.

Prompt To Fix With AI

This is a comment left during a code review. Path: GeneralsMD/Code/GameEngineDevice/Source/W3DDevice/GameClient/Drawable/Draw/W3DModelDraw.cpp Line: 2426-2429 Comment: **Single Object flag spans all turret slots** `m_needUpdateTurretPosition` is one `Bool` for the entire object, but `handleClientTurretPositioning` loops over `MAX_TURRETS`. Any path that clears the flag (e.g. a two-turret unit where `RecenterTurretState` finishes for slot 1) suppresses updates for every other turret slot. A per-slot flag array, or a reference-counted active-turret counter, would prevent one turret's quiescence from silencing a sibling that is still rotating. How can I resolve this? If you propose a fix, please make it concise.

IamInnocent3X · 2026-06-30T12:07:21Z

Alright, thanks for reporting. Pushed an update, it should fix it.

Edit: Nevermind, the issue still persist, just less noticable. I'm looking into it.

Ignore the above, the issue is linked towards my GameClient Drawable Iteration Issue in another PR. That being said, without that Iteration feature even before the fix attempt, I cannot replicate the bug by moving the camera either by scrolling or clicking from the minimap. @Caball009 , able to send a video on replicating it?

Caball009 · 2026-06-30T15:01:47Z

Here you go.

firebase_turret_angle_snap.mp4

IamInnocent3X · 2026-07-01T10:40:25Z

Thank you, I have reattempted the fix as TurretAI's function expands beyond AIUpdate. I have also tested the fix on the current updated version of the SH Repo. Should be fine now.

Caball009 · 2026-07-01T14:58:02Z

Nice, I'll try to make some time to test it later.

Can you rebase this branch?

IamInnocent3X · 2026-07-06T07:35:58Z

Nice, I'll try to make some time to test it later.

Can you rebase this branch?

It's done. Sorry for the wait.

…Update when Necessary

…a on Client's side

Skyaero42

It's done. Sorry for the wait.

It's not. Issue still persist. Please test your changes before submitting them

IamInnocent3X · 2026-08-12T10:15:25Z

It's done. Sorry for the wait.

It's not. Issue still persist. Please test your changes before submitting them

Sorry about that. It should be fixed now.

https://youtu.be/riLCfKjc_fU

CryoTheRenegade · 2026-08-28T00:15:31Z

 {
 	const W3DModelDrawModuleData* d = getW3DModelDrawModuleData();
-	if (!(m_curState->m_validStuff & ModelConditionInfo::BARRELS_VALID))
+	if (!(m_curState->m_validStuff & ModelConditionInfo::BARRELS_VALID) || !m_doHandleRecoil)


Needs a null guard for m_curState

stephanmeesters · 2026-08-28T08:25:52Z

I think the current solution can be bug prone and hard to maintain because if you forget to set the update flag at some point then you can break visuals in a suble way. The extra complexity can be worth it if the performance benefit is there but we don’t have raw numbers yet.

Perhaps use the Tracy profiler to get measurements of the handleClientRecoil and handleClientTurretPositioning functions (put a PROFILER_SECTION; at the top of the function and see project README how to enable profiler)

IamInnocent3X · 2026-08-28T15:46:03Z

I think the current solution can be bug prone and hard to maintain because if you forget to set the update flag at some point then you can break visuals in a suble way. The extra complexity can be worth it if the performance benefit is there but we don’t have raw numbers yet.

Perhaps use the Tracy profiler to get measurements of the handleClientRecoil and handleClientTurretPositioning functions (put a PROFILER_SECTION; at the top of the function and see project README how to enable profiler)

Yeah, it would be hard if someone were to modify the direction of the code and building upon existing ones, I could see it becoming an issue towards oversight. Though if it could be further simplified...

One of the troublesome solution is to use define headers to separate the features, but it could further complicate things.

Though the profiler will be very useful to identify whether the tradeoffs would be worth it for this and #2044

Skyaero42 requested changes Jan 2, 2026

View reviewed changes

IamInnocent3X requested a review from Skyaero42 January 4, 2026 12:51

Caball009 requested changes Jun 29, 2026

View reviewed changes

IamInnocent3X force-pushed the W3DModelDrawPerfImprovements branch from 3ea78f3 to 3e96ab1 Compare June 30, 2026 09:24

greptile-apps Bot reviewed Jun 30, 2026

View reviewed changes

IamInnocent3X force-pushed the W3DModelDrawPerfImprovements branch from 955c9e1 to b402a9b Compare July 6, 2026 07:34

IamInnocent3X added 5 commits July 6, 2026 15:38

Adjusted handleClientTurretPositioning() and handleClientRecoil() to …

2e73a71

…Update when Necessary

Adjusted according to Changes Proposed

957cc83

Changes retail xfer to loadPostProcess

032da9b

Fixed Turret snapping into other Positions when moved away from Camer…

d9014c8

…a on Client's side

Redo Fix for Turret Snapping

ab42a7e

IamInnocent3X force-pushed the W3DModelDrawPerfImprovements branch from b402a9b to ab42a7e Compare July 6, 2026 07:38

Skyaero42 requested changes Aug 11, 2026

View reviewed changes

Fixed Turret Position Snapping when going offscreen

16c7055

Skyaero42 reviewed Aug 13, 2026

View reviewed changes

Comment thread GeneralsReplays

Revert Submodule Changes for GeneralsReplay

52760fe

CryoTheRenegade reviewed Aug 28, 2026

View reviewed changes

greptile-apps Bot reviewed Aug 28, 2026

View reviewed changes

Comment thread Core/GameEngineDevice/Source/W3DDevice/GameClient/Drawable/Draw/W3DModelDraw.cpp

Updated handleRecoil

0815b77

IamInnocent3X force-pushed the W3DModelDrawPerfImprovements branch from ae2357e to 0815b77 Compare August 28, 2026 04:27

Uh oh!

Conversation

IamInnocent3X commented Jan 1, 2026

Uh oh!

Skyaero42 left a comment

Choose a reason for hiding this comment

Uh oh!

Skyaero42 Jan 2, 2026

Choose a reason for hiding this comment

Uh oh!

Skyaero42 Jan 2, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

IamInnocent3X commented Jan 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

stephanmeesters commented Jan 3, 2026

Uh oh!

IamInnocent3X commented Jan 4, 2026

Uh oh!

Caball009 left a comment

Choose a reason for hiding this comment

Uh oh!

IamInnocent3X commented Jun 30, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

greptile-apps Bot commented Jun 30, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Greptile Summary

Confidence Score: 5/5

Important Files Changed

Sequence Diagram

Uh oh!

greptile-apps Bot Jun 30, 2026

Choose a reason for hiding this comment

Uh oh!

IamInnocent3X commented Jun 30, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Caball009 commented Jun 30, 2026

Uh oh!

IamInnocent3X commented Jul 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Caball009 commented Jul 1, 2026

Uh oh!

IamInnocent3X commented Jul 6, 2026

Uh oh!

Skyaero42 left a comment

Choose a reason for hiding this comment

Uh oh!

IamInnocent3X commented Aug 12, 2026

Uh oh!

Uh oh!

CryoTheRenegade Aug 28, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

stephanmeesters commented Aug 28, 2026

Uh oh!

IamInnocent3X commented Aug 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

IamInnocent3X commented Jan 2, 2026 •

edited

Loading

IamInnocent3X commented Jun 30, 2026 •

edited

Loading

greptile-apps Bot commented Jun 30, 2026 •

edited

Loading

IamInnocent3X commented Jun 30, 2026 •

edited

Loading

IamInnocent3X commented Jul 1, 2026 •

edited

Loading

IamInnocent3X commented Aug 28, 2026 •

edited

Loading