Optimize RPN execution to reduce FPS overhead via deduplication #33

cpuwolf · 2025-11-24T23:27:12Z

Description:

This PR introduces a performance enhancement for handling multiple clients subscribing to identical SimVars. Currently, if multiple clients add the same RPN code (e.g., via MF.SimVars.Add), the WASM module executes the code redundantly for every instance within a single frame.

Changes:

Implemented a "calculate once, distribute to many" logic. The WASM module will now detect duplicate RPN codes and execute them only once per frame.
Introduced a new ReadRPNCode structure to manage the mapping of single code executions to multiple SimVars/StringSimVars.

//RPN code execution for reading values in every frame
struct ReadRPNCode {
	std::string Code;
	int RetType; //RetType: 0:float 1:integer 2:string
	std::vector<SimVar> SimVars;
	std::vector<StringSimVar> StringSimVars;
};

Stability:

This patch is highly stable and production-ready. It has been deployed to over 1,000 customers for more than a year with no reported issues.

While multiple clients Add(MF.SimVars.Add.xxxxRPNcodexxx) the exact same SimVars to WASM, From MSFS2020 perspective, it is unnecessary to execute the same RPNcode several times in each frame. This is an enhancement change to reduce FPS impact WASM will smartly execute duplicated SimVars only once in each frame. introduce new data structure: //RPN code execution for reading values in every frame struct ReadRPNCode { std::string Code; int RetType; //RetType: 0:float 1:integer 2:string std::vector<SimVar> SimVars; std::vector<StringSimVar> StringSimVars; }; -- This patch is stable because we have 1000+ customers who have been using these changes over a year

Copilot

Pull request overview

This PR optimizes RPN (Reverse Polish Notation) code execution in the WASM module by implementing a deduplication mechanism. When multiple clients subscribe to identical SimVars with the same RPN code, the module now executes each unique code only once per frame and distributes the result to all subscribers, reducing computational overhead.

Key Changes

Introduced ReadRPNCode structure to map single RPN code executions to multiple SimVar/StringSimVar subscribers
Removed Name field from SimVar and StringSimVar structs, adding client pointer instead for tracking ownership
Moved rolling read index from per-client to global scope to track position in deduplicated RPN code list
Refactored ReadSimVar functions to execute RPN code once and update all associated SimVars

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2025-12-04T10:14:00Z

src/Sources/Code/Module.cpp

 		WriteSimVar(simVar, client);
 	}
 	client->StringSimVars.clear();
+	// clear RNP code list


Typo in comment: "RNP" should be "RPN" to match the actual data structure name (ReadRPNCode).

Suggested change

// clear RNP code list

// clear RPN code list

Copilot · 2025-12-04T10:14:00Z

src/Sources/Code/Module.cpp

+		if ((simVar.Value > floatVal) && (simVar.Value - floatVal < 0.00001F)) {
+			continue;
+		} else if ((simVar.Value < floatVal) && (floatVal - simVar.Value  < 0.00001F)) {
+			continue;
+		}


The floating-point comparison logic is flawed. The condition (simVar.Value > floatVal) && (simVar.Value - floatVal < 0.00001F) will skip updates when values are nearly equal AND the old value is greater. However, if simVar.Value is exactly equal to floatVal, or if they differ by more than 0.00001, the update proceeds. The logic should use std::abs(simVar.Value - floatVal) < 0.00001F to properly check if values are approximately equal regardless of which is larger. Also note the epsilon threshold may be too small for some float ranges.

Copilot · 2025-12-04T10:14:00Z

src/Sources/Code/Module.cpp

 	std::cout << "MobiFlight[" << client->Name.c_str() << "]: Cleared SimVar tracking." << std::endl;
-	//client->RollingClientDataReadIndex = client->SimVars.begin();
-	client->RollingClientDataReadIndex = 0;
+	//client->RollingDataReadIndex = client->SimVars.begin();


Outdated comment reference. The comment refers to client->RollingDataReadIndex but the code now uses the global RollingDataReadIndex. Either update the comment to reflect the current implementation or remove it as it may cause confusion.

Suggested change

//client->RollingDataReadIndex = client->SimVars.begin();

Copilot · 2025-12-04T10:14:01Z

src/Sources/Code/Module.cpp

 	SimVars->push_back(newSimVar);

+	//duplicated SimVar
+	if (pdupRpn) {
+		pdupRpn->SimVars.push_back(newSimVar);
+	} else {
+		ReadRPNCode rpnCode;
+		rpnCode.Code = code;
+		rpnCode.RetType = 0;//hardcoded type id
+		rpnCode.SimVars.push_back(newSimVar);
+		RPNCodelist.push_back(rpnCode);
+	}


Data synchronization issue: newSimVar is pushed to both client->SimVars (line 323) and rpn.SimVars (line 327 or 332). These are separate copies of the struct. When ReadSimVarFloat updates simVar.Value in rpn.SimVars (line 484), the corresponding entry in client->SimVars becomes stale and won't reflect the current value. While SimConnect receives updates via WriteSimVar, the client->SimVars vector will contain outdated values, which could cause issues if this vector is read elsewhere in the codebase or during debugging.

Copilot · 2025-12-04T10:14:01Z

src/Sources/Code/Module.cpp

+	for (int i=0; i < maxVarsPerFrame; ++i) {
+		ReadSimVar(RPNCodelist.at(RollingDataReadIndex));

-		for (int i=0; i < maxVarsPerFrame; ++i) {
-			if(client->RollingClientDataReadIndex < SimVars->size() ) {
-				ReadSimVar(SimVars->at(client->RollingClientDataReadIndex), client);
-			}
-			else {
-				ReadSimVar(StringSimVars->at(client->RollingClientDataReadIndex - SimVars->size()), client);
-			}
-			client->RollingClientDataReadIndex++;
-			if (client->RollingClientDataReadIndex >= totalSimVars)
-				client->RollingClientDataReadIndex = 0;
-		}
+		RollingDataReadIndex++;
+		if (RollingDataReadIndex >= totalSimVars)
+			RollingDataReadIndex = 0;
 	}


Potential out-of-bounds access. If RPNCodelist is empty (totalSimVars == 0), maxVarsPerFrame will be 0, and the loop won't execute. However, if the list becomes empty during execution (e.g., all clients disconnect between frames), and RollingDataReadIndex is non-zero from a previous frame, the next frame could attempt to access RPNCodelist.at(RollingDataReadIndex) with an invalid index. Add a check: if (totalSimVars == 0) { RollingDataReadIndex = 0; return; } before the loop.

Copilot · 2025-12-04T10:14:02Z

src/Sources/Code/Module.cpp

-	uint16_t RollingClientDataReadIndex;
+};
+
+// Runtime Rolling CLient Data reading Index


Typo in comment: "CLient" should be "Client".

Suggested change

// Runtime Rolling CLient Data reading Index

// Runtime Rolling Client Data reading Index

Copilot · 2025-12-04T10:14:03Z

src/Sources/Code/Module.cpp

+				++sit;
+			}
+		}
+		//remove empty RNP code


Typo in comment: "RNP" should be "RPN" to match the actual data structure name (ReadRPNCode).

Suggested change

//remove empty RNP code

//remove empty RPN code

Copilot · 2025-12-04T10:14:03Z

src/Sources/Code/Module.cpp

+	} else {
+		ReadRPNCode rpnCode;
+		rpnCode.Code = code;
+		rpnCode.RetType = 0;//hardcoded type id


Hardcoded RetType value (0 for float) reduces maintainability. Consider defining an enum or constants like RPN_TYPE_FLOAT = 0, RPN_TYPE_INTEGER = 1, RPN_TYPE_STRING = 2 to make the code more readable and maintainable.

Copilot · 2025-12-04T10:14:03Z

src/Sources/Code/Module.cpp

+	} else {
+		ReadRPNCode rpnCode;
+		rpnCode.Code = code;
+		rpnCode.RetType = 2;//hardcoded type id


Hardcoded RetType value (2 for string) reduces maintainability. Consider defining an enum or constants like RPN_TYPE_FLOAT = 0, RPN_TYPE_INTEGER = 1, RPN_TYPE_STRING = 2 to make the code more readable and maintainable.

Copilot · 2025-12-04T10:14:04Z

src/Sources/Code/Module.cpp

 	StringSimVars->push_back(newStringSimVar);

+	//duplicated SimVar
+	if (pdupRpn) {
+		pdupRpn->StringSimVars.push_back(newStringSimVar);
+	} else {
+		ReadRPNCode rpnCode;
+		rpnCode.Code = code;
+		rpnCode.RetType = 2;//hardcoded type id
+		rpnCode.StringSimVars.push_back(newStringSimVar);
+		RPNCodelist.push_back(rpnCode);
+	}


Data synchronization issue: newStringSimVar is pushed to both client->StringSimVars (line 380) and rpn.StringSimVars (line 384 or 389). These are separate copies of the struct. When ReadSimVarString updates simVar.Value in rpn.StringSimVars (line 504), the corresponding entry in client->StringSimVars becomes stale and won't reflect the current value. While SimConnect receives updates via WriteSimVar, the client->StringSimVars vector will contain outdated values, which could cause issues if this vector is read elsewhere in the codebase or during debugging.

cpuwolf changed the title ~~Avoid duplicated reading RPN code execution across multiple clients~~ Optimize RPN execution to reduce FPS overhead via deduplication Nov 24, 2025

DocMoebiuz requested a review from Copilot December 4, 2025 10:08

Copilot started reviewing on behalf of DocMoebiuz December 4, 2025 10:09 View session

Copilot finished reviewing on behalf of DocMoebiuz December 4, 2025 10:12

Copilot AI reviewed Dec 4, 2025

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

Optimize RPN execution to reduce FPS overhead via deduplication #33

Optimize RPN execution to reduce FPS overhead via deduplication #33

cpuwolf commented Nov 24, 2025 •

edited

Loading

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Dec 4, 2025

Uh oh!

Copilot AI Dec 4, 2025

Uh oh!

Copilot AI Dec 4, 2025

Uh oh!

Copilot AI Dec 4, 2025

Uh oh!

Copilot AI Dec 4, 2025

Uh oh!

Copilot AI Dec 4, 2025

Uh oh!

Copilot AI Dec 4, 2025

Uh oh!

Copilot AI Dec 4, 2025

Uh oh!

Copilot AI Dec 4, 2025

Uh oh!

Copilot AI Dec 4, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

	// Runtime Rolling CLient Data reading Index
	// Runtime Rolling Client Data reading Index

Uh oh!

Optimize RPN execution to reduce FPS overhead via deduplication #33

Are you sure you want to change the base?

Optimize RPN execution to reduce FPS overhead via deduplication #33

Conversation

cpuwolf commented Nov 24, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description:

Changes:

Stability:

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Key Changes

Uh oh!

Copilot AI Dec 4, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 4, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 4, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 4, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 4, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 4, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 4, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 4, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 4, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 4, 2025

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

cpuwolf commented Nov 24, 2025 •

edited

Loading