predicate predictor calls added; LSQ logic was changed to keep stores longer so as to allow greater chance for LSQ forwarding; TAGE-SC-L added; micro op cache added; issue with widths of different stages fixed; bug fix related to looking for FMA unit before looking for floatALU/ floatMul/ floatVectorALU, floatVectorMul; causalityTool now works with java-11, ubuntu 20.04, pin-98332

src/emulator/pin/causalityTool.cpp

@@ -24,7 +24,7 @@
 #include <stdlib.h>
 #include <sys/types.h>
 #include <sys/stat.h>
-
+using namespace std;
 #include <fcntl.h>
 
 #ifndef _WIN32
@@ -688,7 +688,7 @@ VOID Instruction(INS ins, VOID *v) {
 
 	UINT32 memOperands = INS_MemoryOperandCount(ins);
 
-	if (INS_IsBranchOrCall(ins))//INS_IsIndirectBranchOrCall(ins))
+	if (INS_IsControlFlow(ins))//INS_IsBranchOrCall(ins))//INS_IsIndirectBranchOrCall(ins))
 	{
 		INS_InsertCall(ins, IPOINT_BEFORE, (AFUNPTR) BrnFun, IARG_THREAD_ID,
 				IARG_BRANCH_TARGET_ADDR, IARG_BRANCH_TAKEN, IARG_INST_PTR,

src/emulator/pin/makefile_linux_mac

@@ -1,4 +1,4 @@
-PIN_KIT ?=/home/rajshekar/softwares/pin-97554/
+PIN_KIT ?=/home/rajshekar/softwares/pin-98332/
 CXX=$(shell make PIN_ROOT=$(PIN_KIT) VAR=CXX -f pin_makefile print_var)
 LINKER=$(shell make PIN_ROOT=$(PIN_KIT) VAR=LINKER -f pin_makefile print_var)
 TOOL_CXXFLAGS=$(shell make PIN_ROOT=$(PIN_KIT) VAR=TOOL_CXXFLAGS -f pin_makefile print_var)
@@ -50,6 +50,8 @@ $(BINDIR)/causalityTool.$(LIB_EXTENSION): $(BINDIR)/causalityTool.$(OBJ_EXTENSIO
 
 $(BINDIR)/causalityTool.$(OBJ_EXTENSION): causalityTool.cpp $(COMMDIR)/IPCBase.h $(COMMDIR)/shm/shmem.h  $(COMMDIR)/filePacket/filePacket.h $(COMMDIR)/shm/shmem.cc
 	$(CXX) $(TOOL_CXXFLAGS) $(COMM_INCLUDE) -c causalityTool.cpp ../../simulator/emulatorinterface/communication/shm/shmem.cc
+	mkdir $(JNIBINDIR)
+	mkdir $(BINDIR)
 	mv causalityTool.$(OBJ_EXTENSION) $(BINDIR)/causalityTool.$(OBJ_EXTENSION)
 	mv shmem.$(OBJ_EXTENSION) $(BINDIR)/shmem.$(OBJ_EXTENSION)
 
@@ -58,16 +60,21 @@ $(BINDIR)/shmem.$(OBJ_EXTENSION): $(COMMDIR)/IPCBase.h $(COMMDIR)/shm/shmem.h  $
 
 
 ################################ JNI stuff comes here ############################################
-JNIPACKAGE = emulatorinterface.communication.shm.SharedMem
+#JNIPACKAGE = emulatorinterface.communication.shm.SharedMem #use this for java-8
-JNINCLUDE =-I/usr/lib/jvm/java-8-openjdk-amd64/include/linux -I/usr/lib/jvm/java-8-openjdk-amd64/include 
+JNIPACKAGE = ../../simulator/emulatorinterface/communication/shm/SharedMem.java #use this for java-11
+#JNINCLUDE =-I/usr/lib/jvm/java-8-openjdk-amd64/include/linux -I/usr/lib/jvm/java-8-openjdk-amd64/include #use this for java-8
+JNINCLUDE =-I/usr/lib/jvm/java-11-openjdk-amd64/include/linux -I/usr/lib/jvm/java-11-openjdk-amd64/include #use this for java-11
 JNILinkingFlags = -shared -Wall $(POSITION_INDEPENDENCE)
-JAVAH = javah -jni
+#JAVAH = javah -jni #use this for java-8
+JAVAH = javac #use this for java-11
 
 $(JNIBINDIR)/libshmlib.$(LIB_EXTENSION): $(JNIBINDIR)/SharedMem.h $(COMMDIR)/shm/JNIShm.c $(COMMDIR)/common.h
 	$(shell $(JNICOMMAND))
 
 $(JNIBINDIR)/SharedMem.h: $(TOPBINDIR)/emulatorinterface/communication/shm/SharedMem.class
-	$(JAVAH) -classpath $(TOPBINDIR) -o $(JNIBINDIR)/SharedMem.h $(JNIPACKAGE)
+	#$(JAVAH) -classpath $(TOPBINDIR) -o $(JNIBINDIR)/SharedMem.h $(JNIPACKAGE) #use this for java-8
+	$(JAVAH) -classpath $(TOPBINDIR) -h $(JNIBINDIR) $(JNIPACKAGE) #use this line and the next for java-11
+	mv $(JNIBINDIR)/emulatorinterface_communication_shm_SharedMem.h $(JNIBINDIR)/SharedMem.h
 
 clean:
-	rm -rf $(BINDIR)/* $(JNIBINDIR)/*
+	rm -rf $(BINDIR) $(JNIBINDIR)

src/simulator/config/BranchPredictorConfig.java

@@ -5,8 +5,9 @@ public class BranchPredictorConfig {
 	public int BHRsize;
 	public int saturating_bits;
 	public BP predictorMode;
+    public String TAGESCLLibDirectory;
 	
 	public static enum BP {
-		NoPredictor, PerfectPredictor, AlwaysTaken, AlwaysNotTaken, Tournament, Bimodal, GShare, GAg, GAp, PAg, PAp,TAGE
+		NoPredictor, PerfectPredictor, AlwaysTaken, AlwaysNotTaken, Tournament, Bimodal, GShare, GAg, GAp, PAg, PAp,TAGE, TAGE_SC_L,
 	}
 }

src/simulator/config/CoreConfig.java

@@ -58,7 +58,9 @@ public class CoreConfig
 	public int STLBAccessPorts;
 	public int STLBPortOccupancy;
 
+	public int NoOfMicroOpCacheEntries;
 	public int DecodeWidth;
+	public int RenameWidth;
 	public int IssueWidth;
 	public int RetireWidth;
 	public int ROBSize;

src/simulator/config/XMLParser.java

@@ -481,6 +481,9 @@ public class XMLParser
 				core.STLBPortOccupancy = Integer.parseInt(getImmediateString("PortOccupancy", sTLBElmnt));
 				core.sTLBPower = getEnergyConfig(sTLBElmnt);
 	
+				Element microOpCacheElmnt = (Element)(coreElmnt.getElementsByTagName("MicroOpCache")).item(0);
+				core.NoOfMicroOpCacheEntries = Integer.parseInt(getImmediateString("NumberOfMicroOps", microOpCacheElmnt));
+	
 				Element decodeElmnt = (Element)(coreElmnt.getElementsByTagName("Decode")).item(0);
 				core.DecodeWidth = Integer.parseInt(getImmediateString("Width", decodeElmnt));
 				core.decodePower = getEnergyConfig(decodeElmnt);
@@ -499,6 +502,7 @@ public class XMLParser
 				core.resultsBroadcastBusPower = getEnergyConfig(resultsBroadcastBusElmnt);
 				
 				Element renameElmnt = (Element)(coreElmnt.getElementsByTagName("Rename")).item(0);
+				core.RenameWidth = Integer.parseInt(getImmediateString("Width", renameElmnt));
 				
 				Element ratElmnt = (Element)(renameElmnt.getElementsByTagName("RAT")).item(0);
 				core.intRATPower = getEnergyConfig((Element)ratElmnt.getElementsByTagName("Integer").item(0));
@@ -1066,9 +1070,14 @@ public class XMLParser
 		{
                         branchPredictor.predictorMode = BP.TAGE;
 		}
+		else if(tempStr.equalsIgnoreCase("TAGE-SC-L"))
+		{
+			branchPredictor.predictorMode = BP.TAGE_SC_L;
+		}
 		branchPredictor.PCBits = Integer.parseInt(getImmediateString("PCBits", predictorElmnt));
 		branchPredictor.BHRsize = Integer.parseInt(getImmediateString("BHRsize", predictorElmnt));
 		branchPredictor.saturating_bits = Integer.parseInt(getImmediateString("SaturatingBits", predictorElmnt));
+        branchPredictor.TAGESCLLibDirectory = getImmediateString("TAGESCLLibDirectory", predictorElmnt);
 	}
 	
 	private static boolean setDirectoryCoherent(String immediateString) {

src/simulator/config/config.xml

@@ -47,12 +47,12 @@ TDP = 15W
 		<!-- If store packets in a file option is set to true, this parameter indicates the basename for the trace files -->
 		<!--One trace file is maintained for each store. The name of trace file for core n is basename_n.gz-->
 		<!--We do not allow overwriting of trace files. So if a tracefile with same name is pre-existing, kindly rename it-->
-		<BasenameForTraceFiles>/home/rajshekar/tmp/gcc_trace</BasenameForTraceFiles>
+		<BasenameForTraceFiles>/home/rajshekar/projects/tejas/tests/test1_trace</BasenameForTraceFiles>
 		
-		<PinTool>/home/rajshekar/softwares/pin-97554/</PinTool>
+		<PinTool>/home/rajshekar/softwares/pin-98332/</PinTool>
 		<PinInstrumentor>/home/rajshekar/projects/tejas/workspace/Tejas/src/emulator/pin/obj-pin/causalityTool.so</PinInstrumentor>
 		<QemuTool>TODO/home/prathmesh/workspace/qemu/x86_64-linux-user/qemu-x86_64 /home/prathmesh/tmp/testQemu.o</QemuTool>
-		<ShmLibDirectory>/home/rajshekar/resources/tejas_configs/</ShmLibDirectory>
+		<ShmLibDirectory>/home/rajshekar/projects/tejas/workspace/Tejas/src/emulator/pin/obj-comm/</ShmLibDirectory>
 		<GetBenchmarkPIDScript>/home/rajshekar_resources/tejas_configs/getBenchmarkPID.sh</GetBenchmarkPIDScript>
 		<KillEmulatorScript>/home/rajshekar/resources/tejas_configs/killAllDescendents.sh</KillEmulatorScript>
 	</Emulator>
@@ -112,11 +112,12 @@ TDP = 15W
 			<PipelineType>outOfOrder</PipelineType>	<!--inOrder,outOfOrder(set issue width for multi-issue in-order)-->
 			
 			<BranchPredictor>
-				<Predictor_Mode>TAGE</Predictor_Mode>	<!-- Legal Values are NoPredictor, PerfectPredictor, AlwaysTaken, AlwaysNotTaken, Tournament, Bimodal, GAg, GAp, GShare, PAg, PAp, TAGE -->
+				<Predictor_Mode>TAGE</Predictor_Mode>	<!-- Legal Values are NoPredictor, PerfectPredictor, AlwaysTaken, AlwaysNotTaken, Tournament, Bimodal, GAg, GAp, GShare, PAg, PAp, TAGE, TAGE-SC-L -->
 				<PCBits>8</PCBits>
 				<BHRsize>16</BHRsize>
 				<BranchMispredPenalty>17</BranchMispredPenalty>	<!--Branch misprediction penalty--><!-- https://www.7-cpu.com/cpu/Skylake.html -->
 				<SaturatingBits>2</SaturatingBits>
+				<TAGESCLLibDirectory>/home/rajshekar/projects/tejas/workspace/Tejas/src/simulator/pipeline/branchpredictor/TAGESCL/</TAGESCLLibDirectory>
 				<LeakageEnergy>0.0178</LeakageEnergy>
 				<DynamicEnergy>0.0962</DynamicEnergy>
 			</BranchPredictor>			
@@ -165,6 +166,10 @@ TDP = 15W
 				<DynamicEnergy>0.06792852941</DynamicEnergy>	<!-- TODO this number isn't right -->
 			</STLB>
 
+            <MicroOpCache>
+                    <NumberOfMicroOps>2304</NumberOfMicroOps>
+            </MicroOpCache>
+			
 			<Decode>								<!--Instruction decode-->
 				<Width>6</Width>			
 				<LeakageEnergy>0.0598</LeakageEnergy>
@@ -172,6 +177,7 @@ TDP = 15W
 			</Decode>
 			
 			<Rename>
+				<Width>6</Width>
 				<RAT>
 					<Integer>
 						<LeakageEnergy>0.0045</LeakageEnergy>

src/simulator/generic/Core.java

@@ -207,6 +207,10 @@ public class Core extends SimulationElement{
 		return coreConfig.DecodeWidth;
 	}
 	
+	public int getRenameWidth() {
+		return coreConfig.RenameWidth;
+	}
+
 	public int getVectorRegisterFileSize() {
 		return coreConfig.VectorRegFileSize;
 	}

src/simulator/generic/Statistics.java

@@ -22,6 +22,7 @@ import memorysystem.nuca.NucaCache;
 import memorysystem.nuca.NucaCache.NucaType;
 import net.NocInterface;
 import net.Router;
+import pipeline.outoforder.OutOrderExecutionEngine;
 import config.CoreConfig;
 import config.EmulatorConfig;
 import config.EnergyConfig;
@@ -33,6 +34,7 @@ import emulatorinterface.translator.qemuTranslationCache.TranslatedInstructionCa
 
 import dram.MainMemoryDRAMController;
 import config.MainMemoryConfig;
+import config.PipelineType;
 
 public class Statistics {
 	
@@ -199,12 +201,22 @@ public class Statistics {
 				outputFileWriter.write("time taken\t=\t" + formatDouble((double)coreCyclesTaken[i]/GlobalClock.effectiveGlobalClockFrequency) + " microseconds\n");
 				outputFileWriter.write("\n");
 				
+				if(cores[i].getCoreConfig().pipelineType == PipelineType.outOfOrder)
+				{
+					outputFileWriter.write("number of micro-op cache accesses = " + ((OutOrderExecutionEngine)cores[i].getExecEngine()).getMicroOpCache().numSearches + "\n");
+					outputFileWriter.write("micro-op cache hit rate = " + formatDouble((double)((OutOrderExecutionEngine)cores[i].getExecEngine()).getMicroOpCache().numHits/(double)((OutOrderExecutionEngine)cores[i].getExecEngine()).getMicroOpCache().numSearches) + "\n");
+					outputFileWriter.write("\n");
+				}
+				
 				outputFileWriter.write("number of branches\t=\t" + cores[i].getExecEngine().getNumberOfBranches() + "\n");
 				outputFileWriter.write("number of mispredicted branches\t=\t" + cores[i].getExecEngine().getNumberOfMispredictedBranches() + "\n");
 				outputFileWriter.write("branch predictor accuracy\t=\t" + formatDouble((double)((double)(1.0 - (double)cores[i].getExecEngine().getNumberOfMispredictedBranches()/(double)cores[i].getExecEngine().getNumberOfBranches())*100.0)) + " %\n");
 				outputFileWriter.write("number of jumps\t=\t" + cores[i].getExecEngine().getNumberOfJumps() + "\n");
 				outputFileWriter.write("number of mispredicted jump targets\t=\t" + cores[i].getExecEngine().getNumberOfMispredictedTargets() + "\n");
 				outputFileWriter.write("target predictor accuracy\t=\t" + formatDouble((double)((double)(1.0 - (double)cores[i].getExecEngine().getNumberOfMispredictedTargets()/(double)cores[i].getExecEngine().getNumberOfJumps())*100.0)) + " %\n");
+				outputFileWriter.write("number of predicate instructions\t=\t" + ((OutOrderExecutionEngine)cores[i].getExecEngine()).getReorderBuffer().predicateCount + "\n");
+				outputFileWriter.write("number of mispredicted predicate instructions\t=\t" + ((OutOrderExecutionEngine)cores[i].getExecEngine()).getReorderBuffer().predicateMispredCount + "\n");
+				outputFileWriter.write("predicate predictor accuracy\t=\t" + formatDouble((double)((double)(1.0 - (double)((OutOrderExecutionEngine)cores[i].getExecEngine()).getReorderBuffer().predicateMispredCount/(double)((OutOrderExecutionEngine)cores[i].getExecEngine()).getReorderBuffer().predicateCount)*100.0)) + " %\n");
 				outputFileWriter.write("\n");
 				
 				outputFileWriter.write("predictor type = " + coreConfig.branchPredictor.predictorMode + "\n");
@@ -213,6 +225,14 @@ public class Statistics {
 				outputFileWriter.write("Saturating bits = " + coreConfig.branchPredictor.saturating_bits + "\n");
 				outputFileWriter.write("\n");
 				
+				outputFileWriter.write("\nIW Full stall = " + ((OutOrderExecutionEngine)cores[i].getExecEngine()).getReorderBuffer().getStall1Count());
+				outputFileWriter.write("\nrename stall = " + ((OutOrderExecutionEngine)cores[i].getExecEngine()).getReorderBuffer().getStall2Count());
+				outputFileWriter.write("\nLSQ Full stall = " + ((OutOrderExecutionEngine)cores[i].getExecEngine()).getReorderBuffer().getStall3Count());
+				outputFileWriter.write("\nROB Full stall = " + ((OutOrderExecutionEngine)cores[i].getExecEngine()).getReorderBuffer().getStall4Count());
+				outputFileWriter.write("\nMispred stall = " + ((OutOrderExecutionEngine)cores[i].getExecEngine()).getReorderBuffer().getStall5Count());
+				outputFileWriter.write("\nSerialization instruction stall = " + ((OutOrderExecutionEngine)cores[i].getExecEngine()).getReorderBuffer().getStall6Count());
+				outputFileWriter.write("\n");
+				
 			}
 			outputFileWriter.write("\n");
 		}

src/simulator/memorysystem/LSQ.java

@@ -301,6 +301,13 @@ public class LSQ extends SimulationElement
 	}
 
 	public boolean isFull(boolean isLoad)
+	{
+		if(privIsFull(isLoad))
+			freeOneEntry(isLoad);
+		return privIsFull(isLoad);
+	}
+	
+	private boolean privIsFull(boolean isLoad)
 	{
 		if(isLoad)
 		{
@@ -495,52 +502,10 @@ committed
 			LSQEntry tmpEntry = lsqueue[i];
 
 			// if it is a store, send the request to the cache
-			if(tmpEntry.getType() == LSQEntry.LSQEntryType.STORE) 
+			if(tmpEntry.getType() == LSQEntry.LSQEntryType.STORE
+					|| tmpEntry.getType() == LSQEntryType.LOAD && tmpEntry.isForwarded()) 
 			{
-				if(tmpEntry.isValid() == false)
+				tmpEntry.setCanBeRemoved(true);
-				{
-					misc.Error.showErrorAndExit("store not ready to be committed");
-				}
-
-				boolean requestIssued =
-						containingMemSys.issueRequestToL1Cache(RequestType.Cache_Write,
-								tmpEntry.getAddr());
-
-				if(requestIssued == false)
-				{
-					event.addEventTime(1);
-					event.getEventQ().addEvent(event);
-					break; //removals must be in-order : if u can't commit the operation at the head, u can't commit the ones that follow it
-				}
-
-				else
-				{
-					if(head == tail)
-					{
-						head = tail = -1;
-					}
-					else
-					{
-						this.head = this.incrementQ(this.head);
-					}
-					this.curNumStoresInQ--;
-					tmpEntry.setRemoved(true);
-				}
-			}
-
-			//If it is a LOAD which has received its value
-			else if (tmpEntry.isForwarded())
-			{
-				if(head == tail)
-				{
-					head = tail = -1;
-				}
-				else
-				{
-					this.head = this.incrementQ(this.head);
-				}
-				this.curNumLoadsInQ--;
-				tmpEntry.setRemoved(true);
 			}
 
 			//If it is a LOAD which has not yet received its value
@@ -559,6 +524,82 @@ committed
 		//incrementNumAccesses(1);
 	}
 
+	public void freeOneEntry(boolean isLoadToBeRemoved)
+	{
+		boolean removedEnough = false;
+		
+		while(removedEnough == false)
+		{
+			LSQEntry tmpEntry = lsqueue[head];
+			
+			if(tmpEntry.isCanBeRemoved() == false)
+			{
+				return;
+			}
+	
+			// if it is a store, send the request to the cache
+			if(tmpEntry.getType() == LSQEntry.LSQEntryType.STORE) 
+			{
+				if(tmpEntry.isValid() == false)
+				{
+					misc.Error.showErrorAndExit("store not ready to be committed");
+				}
+	
+				boolean requestIssued =
+						containingMemSys.issueRequestToL1Cache(RequestType.Cache_Write,
+								tmpEntry.getAddr());
+	
+				if(requestIssued == false)
+				{
+					return; //removals must be in-order : if u can't commit the operation at the head, u can't commit the ones that follow it
+				}
+	
+				else
+				{
+					if(head == tail)
+					{
+						head = tail = -1;
+					}
+					else
+					{
+						this.head = this.incrementQ(this.head);
+					}
+					this.curNumStoresInQ--;
+					tmpEntry.setRemoved(true);
+					
+					if(isLoadToBeRemoved == false)
+						removedEnough = true;
+				}
+			}
+	
+			//If it is a LOAD which has received its value
+			else if (tmpEntry.isForwarded())
+			{
+				if(head == tail)
+				{
+					head = tail = -1;
+				}
+				else
+				{
+					this.head = this.incrementQ(this.head);
+				}
+				this.curNumLoadsInQ--;
+				tmpEntry.setRemoved(true);
+				
+				if(isLoadToBeRemoved == true)
+					removedEnough = true;
+			}
+	
+			//If it is a LOAD which has not yet received its value
+			else
+			{
+				System.err.println("Error in LSQ " +this.containingMemSys.coreID+ " :  ROB sent commit for a load which has not received its value");
+				misc.Error.showErrorAndExit(tmpEntry.getIndexInQ() + " : load : " + tmpEntry.getAddr());
+			}
+			//incrementNumAccesses(1);
+		}
+	}
+
 	void incrementNumAccesses(int incrementBy)
 	{
 		numAccesses += incrementBy;

src/simulator/memorysystem/LSQEntry.java

@@ -33,7 +33,7 @@ public class LSQEntry
         private boolean valid;
         private boolean issued;
         private boolean forwarded;//Whether the load has got its value or not
-
+        private boolean canBeRemoved;
         private boolean removed; //If the entry has been committed and removed from the LSQ
         
         public enum LSQEntryType {LOAD, STORE};
@@ -45,6 +45,7 @@ public class LSQEntry
                 valid = false;
                 issued = false;
                 forwarded = false;
+                canBeRemoved = false;
                 removed = true;
         }
         
@@ -54,6 +55,7 @@ public class LSQEntry
                 valid = false;
                 issued = false;
                 forwarded = false;
+                canBeRemoved = false;
                 removed = false;
         }
 
@@ -108,6 +110,14 @@ public class LSQEntry
                 this.forwarded = forwarded;
         }
         
+        public boolean isCanBeRemoved() {
+			return canBeRemoved;
+		}
+
+		public void setCanBeRemoved(boolean canBeRemoved) {
+			this.canBeRemoved = canBeRemoved;
+		}
+        
         protected boolean isRemoved() {
                 return removed;
         }

src/simulator/pipeline/ExecutionEngine.java

@@ -20,6 +20,7 @@ import pipeline.branchpredictor.PApPredictor;
 import pipeline.branchpredictor.PerfectPredictor;
 import pipeline.branchpredictor.TournamentPredictor;
 import pipeline.branchpredictor.TAGE;
+import pipeline.branchpredictor.TAGESCL.TAGESCL;
 import pipeline.branchpredictor.BTB;
 import generic.Core;
 import generic.GenericCircularQueue;
@@ -88,6 +89,8 @@ public abstract class ExecutionEngine {
                         this.branchPredictor = new TAGE(this,
                                         coreConfig.branchPredictor.PCBits,
                                         coreConfig.branchPredictor.saturating_bits);
+		else if(coreConfig.branchPredictor.predictorMode == BP.TAGE_SC_L)
+                                        	this.branchPredictor = new TAGESCL(this);
 
 		BTB = new BTB(coreConfig.branchPredictor.PCBits, coreConfig.branchPredictor.BHRsize);
 	}

src/simulator/pipeline/OpTypeToFUTypeMapping.java

@@ -10,13 +10,13 @@ public class OpTypeToFUTypeMapping {
 	public static FunctionalUnitType[] intALUFUs = {FunctionalUnitType.integerALU};
 	public static FunctionalUnitType[] intMulFUs = {FunctionalUnitType.integerMul};
 	public static FunctionalUnitType[] intDivFUs = {FunctionalUnitType.integerDiv};
-	public static FunctionalUnitType[] floatALUFUs = {FunctionalUnitType.floatALU, FunctionalUnitType.FMA};
+	public static FunctionalUnitType[] floatALUFUs = {FunctionalUnitType.FMA, FunctionalUnitType.floatALU};
-	public static FunctionalUnitType[] floatMulFUs = {FunctionalUnitType.floatMul, FunctionalUnitType.FMA};
+	public static FunctionalUnitType[] floatMulFUs = {FunctionalUnitType.FMA, FunctionalUnitType.floatMul};
 	public static FunctionalUnitType[] floatDivFUs = {FunctionalUnitType.floatDiv};
 	public static FunctionalUnitType[] intVectorALUFUs = {FunctionalUnitType.integerVectorALU};
 	public static FunctionalUnitType[] intVectorMulFUs = {FunctionalUnitType.integerVectorMul};
-	public static FunctionalUnitType[] floatVectorALUFUs = {FunctionalUnitType.floatVectorALU, FunctionalUnitType.FMA};
+	public static FunctionalUnitType[] floatVectorALUFUs = {FunctionalUnitType.FMA, FunctionalUnitType.floatVectorALU};
-	public static FunctionalUnitType[] floatVectorMulFUs = {FunctionalUnitType.floatVectorMul, FunctionalUnitType.FMA};
+	public static FunctionalUnitType[] floatVectorMulFUs = {FunctionalUnitType.FMA, FunctionalUnitType.floatVectorMul};
 	public static FunctionalUnitType[] FMAFUs = {FunctionalUnitType.FMA};
 	public static FunctionalUnitType[] VectorFMAFUs = {FunctionalUnitType.FMA};
 	public static FunctionalUnitType[] loadFUs = {FunctionalUnitType.load};

src/simulator/pipeline/branchpredictor/TAGESCL/Makefile

@@ -0,0 +1,8 @@
+all:
+	javac -h . TAGESCLInvoker.java
+	#gcc -fPIC -I/usr/lib/jvm/java-8-openjdk-amd64/include/ -I/usr/lib/jvm/java-8-openjdk-amd64/include/linux/ -shared -o libnative.so pipeline_branchpredictor_TAGESCL_TAGESCLInvoker.cc
+	gcc -fPIC -I/usr/lib/jvm/java-11-openjdk-amd64/include/ -I/usr/lib/jvm/java-11-openjdk-amd64/include/linux/ -shared -o libnative.so pipeline_branchpredictor_TAGESCL_TAGESCLInvoker.cc
+clean:
+	rm -f *.h
+	rm -f *.class
+	rm -f *.so

src/simulator/pipeline/branchpredictor/TAGESCL/TAGESCL.java

@@ -0,0 +1,32 @@
+package pipeline.branchpredictor.TAGESCL;
+
+import pipeline.ExecutionEngine;
+import pipeline.branchpredictor.BranchPredictor;
+
+public class TAGESCL extends BranchPredictor {
+
+	TAGESCLInvoker ti;	
+	public TAGESCL(ExecutionEngine containingExecEngine)
+	{
+		super(containingExecEngine);
+		ti = new TAGESCLInvoker(containingExecEngine.getContainingCore().getCoreConfig().branchPredictor.TAGESCLLibDirectory);
+	}
+	
+	public boolean predict(long address, boolean outcome)
+	{
+		return ti.invokerPredict(address);
+	}
+	
+	public void Train(long address, boolean outcome, boolean predict)
+	{
+		misc.Error.showErrorAndExit("use the other Train() function");
+		//don't use this!!
+		//use Train(long PC, int opType, boolean resolveDir, boolean predDir, long branchTarget)
+	}
+	
+	public void Train(long PC, int opType, boolean resolveDir, boolean predDir, long branchTarget)
+	{
+		ti.invokerTrain (PC, opType, resolveDir, predDir, branchTarget);
+	}
+	
+}

src/simulator/pipeline/branchpredictor/TAGESCL/TAGESCLInvoker.java

@@ -0,0 +1,25 @@
+package pipeline.branchpredictor.TAGESCL;
+
+public class TAGESCLInvoker {
+	
+	private native void initialize();
+	private native boolean predict (long PC);
+	private native void train (long PC, int opType, boolean resolveDir, boolean predDir, long branchTarget);
+	
+	public TAGESCLInvoker(String TAGESCLLibDirectory)
+	{
+		System.load(TAGESCLLibDirectory + "/libnative.so");
+		initialize();
+	}
+	
+	public boolean invokerPredict(long address)
+	{
+		return predict(address);
+	}
+	
+	public void invokerTrain(long PC, int opType, boolean resolveDir, boolean predDir, long branchTarget)
+	{
+		train (PC, opType, resolveDir, predDir, branchTarget);
+	}
+	
+}

src/simulator/pipeline/outoforder/DecodeLogic.java

@@ -39,9 +39,7 @@ public class DecodeLogic extends SimulationElement {
 	
 	public void performDecode()
 	{
-		if(containingExecutionEngine.isToStall5() == true /*pipeline stalled due to branch mis-prediction*/
+		if(containingExecutionEngine.isToStall5() == true /*pipeline stalled due to branch mis-prediction*/)
-				|| containingExecutionEngine.isToStall1() == true /*IW full*/
-				|| containingExecutionEngine.isToStall2() == true /*rename stall*/)
 		{
 			return;
 		}

src/simulator/pipeline/outoforder/FetchLogic.java

@@ -1,6 +1,7 @@
 package pipeline.outoforder;
 
 import config.SimulationConfig;
+import config.SystemConfig;
 import emulatorinterface.translator.x86.registers.Registers;
 import main.ArchitecturalComponent;
 import main.CustomObjectPool;
@@ -11,6 +12,7 @@ import generic.Core;
 import generic.Event;
 import generic.EventQueue;
 import generic.GenericCircularQueue;
+import generic.GlobalClock;
 import generic.Instruction;
 import generic.OperationType;
 import generic.PortType;
@@ -25,12 +27,14 @@ public class FetchLogic extends SimulationElement {
 	GenericCircularQueue<Instruction>[] inputToPipeline;
 	int inputPipeToReadNext;
 	ICacheBuffer iCacheBuffer;
+	MicroOpCache microOpCache;
 	GenericCircularQueue<Instruction> fetchBuffer;	
 	int fetchWidth;
 	OperationType[] instructionsToBeDropped;
 	boolean sleep;
 	
 	long serialNo;
+	long lastValidIPSeen = -1;
 
 	public FetchLogic(Core core, OutOrderExecutionEngine execEngine)
 	{
@@ -38,6 +42,7 @@ public class FetchLogic extends SimulationElement {
 		this.core = core;
 		this.execEngine = execEngine;
 		fetchBuffer = execEngine.getFetchBuffer();
+		microOpCache = execEngine.getMicroOpCache();
 		fetchWidth = core.getDecodeWidth();
 		inputPipeToReadNext = 0;
 		sleep = false;
@@ -77,13 +82,9 @@ public class FetchLogic extends SimulationElement {
 		
 		Instruction newInstruction;
 		
-		if(!execEngine.isToStall1() &&
+		if(execEngine.isToStall5())
-				!execEngine.isToStall2() &&
+			return;
-				!execEngine.isToStall3() &&
+		
-				!execEngine.isToStall4() &&
-				!execEngine.isToStall5() &&
-				!execEngine.isToStall6())
-		{
 		//add instructions, for whom "fetch" from iCache has completed, to fetch buffer
 		//decode stage reads from this buffer
 		for(int i = 0; i < fetchWidth; i++)
@@ -97,6 +98,11 @@ public class FetchLogic extends SimulationElement {
 			if(newInstruction != null)
 			{
 				fetchBuffer.enqueue(newInstruction);
+				
+				if(SimulationConfig.debugMode)
+				{
+					System.out.println("fetched : " + GlobalClock.getCurrentTime()/core.getStepSize() + " : "  + newInstruction);
+				}
 			}
 			else
 			{
@@ -104,7 +110,6 @@ public class FetchLogic extends SimulationElement {
 				break;
 			}
 		}
-		}
 		
 		//this loop reads from inputToPipeline and places the instruction in iCacheBuffer
 		//fetch of the instruction is also issued to the iCache
@@ -191,10 +196,32 @@ public class FetchLogic extends SimulationElement {
 				{
 						// The first micro-operation of an instruction has a valid CISC IP. All the subsequent 
 					  	// micro-ops will have IP = -1(meaning invalid). We must not forward this requests to iCache.
-						if(newInstruction.getCISCProgramCounter()!=-1)
+						// If the micro-ops are available in the micro-op cache, we don't need to access the i-cache
+						if(newInstruction.getCISCProgramCounter()!=-1 && newInstruction.getCISCProgramCounter() != lastValidIPSeen
+								&& microOpCache.isPresentInCache(newInstruction.getCISCProgramCounter()) == false)
 						{
 							execEngine.getCoreMemorySystem().issueRequestToInstrCache(newInstruction.getCISCProgramCounter());							
 						}
+						else
+						{
+							iCacheBuffer.fetchComplete[iCacheBuffer.tail] = true;
+							if(newInstruction.getCISCProgramCounter()==-1 || newInstruction.getCISCProgramCounter() == lastValidIPSeen)
+								microOpCache.isPresentInCache(lastValidIPSeen); //accessing micro-op cache just to get the micro-op cache LRU and counters right
+						}
+						
+						if(newInstruction.getCISCProgramCounter()!=-1)
+						{
+							lastValidIPSeen = newInstruction.getCISCProgramCounter();
+						}
+						else
+						{
+							newInstruction.setCISCProgramCounter(lastValidIPSeen);
+						}
+						
+						if(SimulationConfig.debugMode)
+						{
+							System.out.println("fetch_initiated : " + GlobalClock.getCurrentTime()/core.getStepSize() + " : "  + newInstruction);
+						}
 				}
 			}
 			else
@@ -235,7 +262,8 @@ public class FetchLogic extends SimulationElement {
 	
 	public void processCompletionOfMemRequest(long address)
 	{
-		iCacheBuffer.updateFetchComplete(address);
+		int numberOfMicroOpsUpdated = iCacheBuffer.updateFetchComplete(address);
+		microOpCache.addToCache(address, numberOfMicroOpsUpdated);
 	}
 	
 	public GenericCircularQueue<Instruction>[] getInputToPipeline() {

src/simulator/pipeline/outoforder/ICacheBuffer.java

@@ -78,21 +78,26 @@ public class ICacheBuffer {
 		return toBeReturned;
 	}
 	
-	public void updateFetchComplete(long programCounter)
+	public int updateFetchComplete(long programCounter)
 	{
+		int numberOfMicroOpsUpdated = 0;
+		
 		if(head == -1)
-			return;
+			return numberOfMicroOpsUpdated;
 		
 		for(int i = head; ; i = (i + 1)%size)
 		{
 			if(buffer[i] != null && buffer[i].getCISCProgramCounter() == programCounter)
 			{
 				fetchComplete[i] = true;
+				numberOfMicroOpsUpdated++;
 			}
 			
 			if(i == tail)
 				break;
 		}
+		
+		return numberOfMicroOpsUpdated;
 	}
 	
 	public boolean isFull()

src/simulator/pipeline/outoforder/IWPushLogic.java

@@ -16,7 +16,7 @@ public class IWPushLogic extends SimulationElement {
 	OutOrderExecutionEngine execEngine;
 	GenericCircularQueue<ReorderBufferEntry> renameBuffer;
 	InstructionWindow IW;
-	int decodeWidth;	
+	int renameWidth;	
 	
 	public IWPushLogic(Core core, OutOrderExecutionEngine execEngine)
 	{
@@ -25,7 +25,7 @@ public class IWPushLogic extends SimulationElement {
 		this.execEngine = execEngine;
 		renameBuffer = execEngine.getRenameBuffer();		
 		IW = execEngine.getInstructionWindow();
-		decodeWidth = core.getDecodeWidth();
+		renameWidth = core.getRenameWidth();
 	}
 	
 	/*
@@ -39,7 +39,7 @@ public class IWPushLogic extends SimulationElement {
 			return;
 		}
 		
-		for(int i = 0; i < decodeWidth; i++)
+		for(int i = 0; i < renameWidth; i++)
 		{
 			ReorderBufferEntry headROBEntry = renameBuffer.peek(0);
 			if(headROBEntry != null)

src/simulator/pipeline/outoforder/MicroOpCache.java

@@ -0,0 +1,107 @@
+package pipeline.outoforder;
+
+import java.util.HashMap;
+import java.util.Map;
+import java.util.Vector;
+
+import config.SimulationConfig;
+import generic.Event;
+import generic.EventQueue;
+import generic.GlobalClock;
+import generic.PortType;
+import generic.SimulationElement;
+
+public class MicroOpCache extends SimulationElement {
+	
+	int maxSize; //in terms of number of micro-ops
+	int curSize;
+	HashMap<Long, MicroOpCacheEntry> uopCache;
+	
+	public long numAdditions;
+	public long numSearches;
+	public long numHits;
+
+	public MicroOpCache(int maxSize) {
+		super(PortType.Unlimited, -1, -1, -1, -1);
+		this.maxSize = maxSize;
+		uopCache = new HashMap<Long, MicroOpCacheEntry>();
+	}
+
+	@Override
+	public void handleEvent(EventQueue eventQ, Event event) {
+		// TODO Auto-generated method stub
+
+	}
+	
+	public boolean isPresentInCache(long searchPC) //will be called for each micro-op (and not each CISC instruction)
+	{
+		numSearches++;
+		
+		MicroOpCacheEntry entry = uopCache.get(searchPC);
+		if(entry != null)
+		{
+			entry.timeLastUsed = GlobalClock.getCurrentTime();
+			numHits++;
+			if(SimulationConfig.debugMode)
+			{
+				System.out.println("hit in microp-cache : " + GlobalClock.getCurrentTime()/24 + " : "  + Long.toHexString(searchPC));
+			}
+			return true;
+		}
+		
+		return false;
+	}
+	
+	public void addToCache(long newPC, int numberOfMicroOps)
+	{
+		if(uopCache.containsKey(newPC) == false)
+		{
+			//remove old entries to make place for the new one
+			while(curSize + numberOfMicroOps > maxSize)
+			{
+				//find LRU PC
+				long LRU_PC = -1;
+				MicroOpCacheEntry LRUEntry = null;
+				for(Map.Entry<Long, MicroOpCacheEntry> entry : uopCache.entrySet())
+				{
+					if(LRUEntry == null)
+					{
+						LRUEntry = entry.getValue();
+						LRU_PC = entry.getKey();
+					}
+					else
+					{
+						if(entry.getValue().timeLastUsed < LRUEntry.timeLastUsed)
+						{
+							LRUEntry = entry.getValue();
+							LRU_PC = entry.getKey();
+						}
+					}
+				}
+				
+				//remove all micro-ops corresponding to LRU PC
+				uopCache.remove(LRU_PC);
+				curSize -= LRUEntry.numberOfMicroOps;
+			}
+			
+			//add new micro-ops
+			MicroOpCacheEntry newEntry = new MicroOpCacheEntry();
+			newEntry.numberOfMicroOps = numberOfMicroOps;
+			newEntry.timeLastUsed = GlobalClock.getCurrentTime();
+			uopCache.put(newPC, newEntry);
+			curSize += numberOfMicroOps;
+			
+			numAdditions += numberOfMicroOps;
+			if(SimulationConfig.debugMode)
+			{
+				System.out.println("add to microp-cache : " + GlobalClock.getCurrentTime()/24 + " : "  + Long.toHexString(newPC));
+			}
+		}
+	}
+}
+
+class MicroOpCacheEntry
+{
+	int numberOfMicroOps;
+	long timeLastUsed;
+}

src/simulator/pipeline/outoforder/OutOrderExecutionEngine.java

@@ -22,6 +22,7 @@ public class OutOrderExecutionEngine extends ExecutionEngine {
 	//components of the execution engine
 	private ICacheBuffer iCacheBuffer;
 	private FetchLogic fetcher;
+	private MicroOpCache microOpCache;
 	private GenericCircularQueue<Instruction> fetchBuffer;
 	private DecodeLogic decoder;
 	private GenericCircularQueue<ReorderBufferEntry> decodeBuffer;
@@ -81,11 +82,12 @@ public class OutOrderExecutionEngine extends ExecutionEngine {
 		vectorRegisterFile = new RegisterFile(core, core.getVectorRegisterFileSize());
 		vectorRenameTable = new RenameTable(this, core.getNVectorArchitecturalRegisters(), core.getVectorRegisterFileSize(), vectorRegisterFile, core.getNo_of_input_pipes());
 
-		fetchBuffer = new GenericCircularQueue(Instruction.class, core.getDecodeWidth());
+		fetchBuffer = new GenericCircularQueue(Instruction.class, (core.getDecodeWidth()>core.getRenameWidth()?core.getDecodeWidth():core.getRenameWidth()));
+		microOpCache = new MicroOpCache(core.getCoreConfig().NoOfMicroOpCacheEntries);
 		fetcher = new FetchLogic(core, this);
-		decodeBuffer = new GenericCircularQueue(ReorderBufferEntry.class, core.getDecodeWidth());
+		decodeBuffer = new GenericCircularQueue(ReorderBufferEntry.class, (core.getDecodeWidth()>core.getRenameWidth()?core.getDecodeWidth():core.getRenameWidth()));
 		decoder = new DecodeLogic(core, this);
-		renameBuffer = new GenericCircularQueue(ReorderBufferEntry.class, core.getDecodeWidth());
+		renameBuffer = new GenericCircularQueue(ReorderBufferEntry.class, (core.getDecodeWidth()>core.getRenameWidth()?core.getDecodeWidth():core.getRenameWidth()));
 		renamer = new RenameLogic(core, this);
 		IWPusher = new IWPushLogic(core, this);
 		selector = new SelectLogic(core, this);
@@ -202,6 +204,10 @@ public class OutOrderExecutionEngine extends ExecutionEngine {
 		return renameBuffer;
 	}
 	
+	public MicroOpCache getMicroOpCache() {
+		return microOpCache;
+	}
+
 	public FetchLogic getFetcher() {
 		return fetcher;
 	}

src/simulator/pipeline/outoforder/RenameLogic.java

@@ -23,7 +23,7 @@ public class RenameLogic extends SimulationElement {
 	OutOrderExecutionEngine execEngine;
 	GenericCircularQueue<ReorderBufferEntry> decodeBuffer;
 	GenericCircularQueue<ReorderBufferEntry> renameBuffer;
-	int decodeWidth;
+	int renameWidth;
 	
 	int threadID;
 	Instruction instruction;
@@ -37,18 +37,17 @@ public class RenameLogic extends SimulationElement {
 		this.execEngine = execEngine;
 		decodeBuffer = execEngine.getDecodeBuffer();
 		renameBuffer = execEngine.getRenameBuffer();
-		decodeWidth = core.getDecodeWidth();
+		renameWidth = core.getRenameWidth();
 	}
 	
 	public void performRename()
 	{
-		if(execEngine.isToStall5() == true /*pipeline stalled due to branch mis-prediction*/
+		if(execEngine.isToStall5() == true /*pipeline stalled due to branch mis-prediction*/)
-				|| execEngine.isToStall1() == true /*IW full*/)
 		{
 			return;
 		}
 		
-		for(int i = 0; i < decodeWidth; i++)
+		for(int i = 0; i < renameWidth; i++)
 		{
 			if(renameBuffer.isFull() == true)
 			{

src/simulator/pipeline/outoforder/ReorderBuffer.java

@@ -21,9 +21,11 @@ import java.io.OutputStreamWriter;
 import java.util.zip.GZIPOutputStream;
 
 import main.CustomObjectPool;
+import pipeline.branchpredictor.TAGESCL.TAGESCL;
 import config.EmulatorConfig;
 import config.EnergyConfig;
 import config.SimulationConfig;
+import config.BranchPredictorConfig.BP;
 
 public class ReorderBuffer extends SimulationElement{
 	
@@ -49,6 +51,8 @@ public class ReorderBuffer extends SimulationElement{
 	int stall6Count;
 	long branchCount;
 	long mispredCount;
+	public long predicateCount;
+	public long predicateMispredCount;
 	long jumpCount;
 	long targetMispredCount;
 	long lastValidIPSeen;
@@ -249,25 +253,44 @@ public class ReorderBuffer extends SimulationElement{
 				if(firstOpType == OperationType.branch)
 				{
 					//perform prediction
-					boolean prediction = execEngine.getBranchPredictor().predict(
+					boolean prediction;
+					prediction = execEngine.getBranchPredictor().predict(
 							lastValidIPSeen,
 							first.getInstruction().isBranchTaken());
+					
 					if(prediction != first.getInstruction().isBranchTaken())
 					{
+						if(SimulationConfig.debugMode)
+						{
+							System.out.println("branch mispredicted : " + firstInstruction.getSerialNo());
+						}
+						
 						anyMispredictedBranch = true;
 						mispredCount++;
 					}
-					this.execEngine.getBranchPredictor().incrementNumAccesses(1);
 					
 					//train predictor
+					if(core.getCoreConfig().branchPredictor.predictorMode != BP.TAGE_SC_L)
+					{
 						execEngine.getBranchPredictor().Train(
 								lastValidIPSeen,
 								firstInstruction.isBranchTaken(),
 								prediction
 								);
-					this.execEngine.getBranchPredictor().incrementNumAccesses(1);
+					}
+					else
+					{
+						((TAGESCL)execEngine.getBranchPredictor()).Train(
+								lastValidIPSeen,
+								9, //OPTYPE_JMP_DIRECT_COND,
+								firstInstruction.isBranchTaken(),
+								prediction,
+								firstInstruction.getBranchTargetAddress()
+								);
+					}
 					
 					this.execEngine.getBTB().GHRTrain(firstInstruction.isBranchTaken());
+					this.execEngine.getBranchPredictor().incrementNumAccesses(2);
 					
 					branchCount++;
 				}
@@ -275,19 +298,75 @@ public class ReorderBuffer extends SimulationElement{
 				//jump operation
 				if(firstOpType == OperationType.jump)
 				{
-					long actualTarget = firstInstruction.getBranchTargetAddress();					
+					long actualTarget = first.getInstruction().getBranchTargetAddress();					
 					long predictedTarget = this.execEngine.getBTB().BTBPredict(lastValidIPSeen);
 					
 					if(actualTarget != predictedTarget)
 					{
+						if(SimulationConfig.debugMode)
+						{
+							System.out.println("jump target mispredicted : " + firstInstruction.getSerialNo());
+						}
+						
 						anyMispredictedBranch = true;
 						targetMispredCount++;
 					}
 				
 					this.execEngine.getBTB().BTBTrain(lastValidIPSeen, actualTarget);
+					
+					if(core.getCoreConfig().branchPredictor.predictorMode == BP.TAGE_SC_L)
+					{
+						((TAGESCL)execEngine.getBranchPredictor()).Train(
+								lastValidIPSeen,
+								4, //OPTYPE_JMP_DIRECT_UNCOND,
+								true,
+								true,
+								firstInstruction.getBranchTargetAddress()
+								);
+					}
 					jumpCount++;
 				}
 				
+				//predicate prediction
+				if(firstInstruction.isPredicate())
+				{
+					//perform prediction
+					boolean prediction = execEngine.getBranchPredictor().predict(
+																		lastValidIPSeen,
+																		!first.getInstruction().isPredicateAndNotExecuted());
+					if(prediction != !first.getInstruction().isPredicateAndNotExecuted())
+					{	
+						if(SimulationConfig.debugMode)
+						{
+							System.out.println("predicate mispredicted : " + firstInstruction.getSerialNo());
+						}
+						
+						anyMispredictedBranch = true;
+						predicateMispredCount++;
+					}
+				
+					//train predictor
+					if(core.getCoreConfig().branchPredictor.predictorMode != BP.TAGE_SC_L)
+					{
+						execEngine.getBranchPredictor().Train(
+								lastValidIPSeen,
+								!firstInstruction.isPredicateAndNotExecuted(),
+								prediction
+								);
+					}
+					else
+					{
+						((TAGESCL)execEngine.getBranchPredictor()).Train(
+								lastValidIPSeen,
+								9, //OPTYPE_JMP_DIRECT_COND,
+								!firstInstruction.isPredicateAndNotExecuted(),
+								prediction,
+								firstInstruction.getBranchTargetAddress()
+								);
+					}
+					predicateCount++;
+				}
+				
 				//Signal LSQ for committing the Instruction at the queue head
 				if(firstOpType == OperationType.load || firstOpType == OperationType.store)
 				{
@@ -648,6 +727,10 @@ public class ReorderBuffer extends SimulationElement{
 		return stall5Count;
 	}
 
+	public int getStall6Count() {
+		return stall6Count;
+	}
+
 	public long getBranchCount() {
 		return branchCount;
 	}