diff --git a/Source/Core/Core/Src/PowerPC/Jit64/Jit.cpp b/Source/Core/Core/Src/PowerPC/Jit64/Jit.cpp
index d3baf4edd7..e7d4ed3c43 100644
--- a/Source/Core/Core/Src/PowerPC/Jit64/Jit.cpp
+++ b/Source/Core/Core/Src/PowerPC/Jit64/Jit.cpp
@@ -437,10 +437,11 @@ const u8* Jit64::DoJit(u32 em_address, PPCAnalyst::CodeBuffer *code_buf, JitBloc
 	//Analyze the block, collect all instructions it is made of (including inlining,
 	//if that is enabled), reorder instructions for optimal performance, and join joinable instructions.
 	u32 nextPC = em_address;
+	std::vector<u32> merged_addresses;
 	if (!memory_exception)
 	{
 		// If there is a memory exception inside a block (broken_block==true), compile up to that instruction.
-		nextPC = PPCAnalyst::Flatten(em_address, &size, &js.st, &js.gpa, &js.fpa, broken_block, code_buf, blockSize);
+		nextPC = PPCAnalyst::Flatten(em_address, &size, &js.st, &js.gpa, &js.fpa, broken_block, code_buf, blockSize, merged_addresses);
 	}
 
 	PPCAnalyst::CodeOp *ops = code_buf->codebuffer;
@@ -496,7 +497,13 @@ const u8* Jit64::DoJit(u32 em_address, PPCAnalyst::CodeBuffer *code_buf, JitBloc
 
 	js.downcountAmount = 0;
 	if (!Core::g_CoreStartupParameter.bEnableDebugging)
-		js.downcountAmount += PatchEngine::GetSpeedhackCycles(em_address);
+	{
+		for (int i = 0; i < merged_addresses.size(); ++i)
+		{
+			const u32 address = merged_addresses[i];
+			js.downcountAmount += PatchEngine::GetSpeedhackCycles(address);
+		}
+	}
 
 	js.skipnext = false;
 	js.blockSize = size;
diff --git a/Source/Core/Core/Src/PowerPC/Jit64IL/JitIL.cpp b/Source/Core/Core/Src/PowerPC/Jit64IL/JitIL.cpp
index 42b82a532e..da0f5eb51b 100644
--- a/Source/Core/Core/Src/PowerPC/Jit64IL/JitIL.cpp
+++ b/Source/Core/Core/Src/PowerPC/Jit64IL/JitIL.cpp
@@ -429,10 +429,11 @@ const u8* JitIL::DoJit(u32 em_address, PPCAnalyst::CodeBuffer *code_buf, JitBloc
 	//Analyze the block, collect all instructions it is made of (including inlining,
 	//if that is enabled), reorder instructions for optimal performance, and join joinable instructions.
 	b->exitAddress[0] = em_address;
+	std::vector<u32> merged_addresses;
 	if (!memory_exception)
 	{
 		// If there is a memory exception inside a block (broken_block==true), compile up to that instruction.
-		b->exitAddress[0] = PPCAnalyst::Flatten(em_address, &size, &js.st, &js.gpa, &js.fpa, broken_block, code_buf, blockSize);
+		b->exitAddress[0] = PPCAnalyst::Flatten(em_address, &size, &js.st, &js.gpa, &js.fpa, broken_block, code_buf, blockSize, merged_addresses);
 	}
 	PPCAnalyst::CodeOp *ops = code_buf->codebuffer;
 
@@ -468,10 +469,15 @@ const u8* JitIL::DoJit(u32 em_address, PPCAnalyst::CodeBuffer *code_buf, JitBloc
 	// instruction processed by the JIT routines)
 	ibuild.Reset();
 
-	
 	js.downcountAmount = 0;
 	if (!Core::g_CoreStartupParameter.bEnableDebugging)
-		js.downcountAmount += PatchEngine::GetSpeedhackCycles(em_address);
+	{
+		for (int i = 0; i < merged_addresses.size(); ++i)
+		{
+			const u32 address = merged_addresses[i];
+			js.downcountAmount += PatchEngine::GetSpeedhackCycles(address);
+		}
+	}
 
 	// Translate instructions
 	for (int i = 0; i < (int)size; i++)
diff --git a/Source/Core/Core/Src/PowerPC/PPCAnalyst.cpp b/Source/Core/Core/Src/PowerPC/PPCAnalyst.cpp
index ccc9472333..56276507f2 100644
--- a/Source/Core/Core/Src/PowerPC/PPCAnalyst.cpp
+++ b/Source/Core/Core/Src/PowerPC/PPCAnalyst.cpp
@@ -285,8 +285,9 @@ bool CanSwapAdjacentOps(const CodeOp &a, const CodeOp &b)
 
 // Does not yet perform inlining - although there are plans for that.
 // Returns the exit address of the next PC
-u32 Flatten(u32 address, int *realsize, BlockStats *st, BlockRegStats *gpa, BlockRegStats *fpa, bool &broken_block, CodeBuffer *buffer, int blockSize)
+u32 Flatten(u32 address, int *realsize, BlockStats *st, BlockRegStats *gpa, BlockRegStats *fpa, bool &broken_block, CodeBuffer *buffer, int blockSize, std::vector<u32>& merged_addresses)
 {
+	merged_addresses.push_back(address);
 	memset(st, 0, sizeof(st));
 	
 	// Disabled the following optimization in preference of FAST_ICACHE
@@ -474,6 +475,7 @@ u32 Flatten(u32 address, int *realsize, BlockStats *st, BlockRegStats *gpa, Bloc
 				// because bx may store a certain value to the link register.
 				// Instead, we skip a part of bx in Jit**::bx().
 				address = destination;
+				merged_addresses.push_back(address);
 			}
 		}
 		else
diff --git a/Source/Core/Core/Src/PowerPC/PPCAnalyst.h b/Source/Core/Core/Src/PowerPC/PPCAnalyst.h
index 174b0ec128..3eecc263c9 100644
--- a/Source/Core/Core/Src/PowerPC/PPCAnalyst.h
+++ b/Source/Core/Core/Src/PowerPC/PPCAnalyst.h
@@ -108,7 +108,7 @@ public:
 
 };
 
-u32 Flatten(u32 address, int *realsize, BlockStats *st, BlockRegStats *gpa, BlockRegStats *fpa, bool &broken_block, CodeBuffer *buffer, int blockSize);
+u32 Flatten(u32 address, int *realsize, BlockStats *st, BlockRegStats *gpa, BlockRegStats *fpa, bool &broken_block, CodeBuffer *buffer, int blockSize, std::vector<u32>& merged_addresses);
 void LogFunctionCall(u32 addr);
 void FindFunctions(u32 startAddr, u32 endAddr, PPCSymbolDB *func_db);
 bool AnalyzeFunction(u32 startAddr, Symbol &func, int max_size = 0);