Shreya and I found a bug with the exeuction of JAL and JALR instructions. The link was only set in the writeback stage. Once the branch predictor started correctly predicting JAL(R)s the ALU and forwarding logic need to have the PCLinkE at the execution stage in case an instruction in the next two clocks need the data.

wally-pipelined/regression/wave.do

@@ -115,7 +115,7 @@ add wave -noupdate -expand -group {alu execution stage} /testbench/dut/hart/ieu/
 add wave -noupdate -expand -group {alu execution stage} /testbench/dut/hart/ieu/dp/SrcBE
 add wave -noupdate /testbench/dut/hart/ieu/dp/ALUResultM
 TreeUpdate [SetDefaultTree]
-WaveRestoreCursors {{Cursor 2} {231033 ns} 0} {{Cursor 3} {1276133 ns} 0}
+WaveRestoreCursors {{Cursor 2} {231033 ns} 0} {{Cursor 3} {1276117 ns} 0}
 quietly wave cursor active 2
 configure wave -namecolwidth 250
 configure wave -valuecolwidth 518

wally-pipelined/src/ieu/controller.sv

@@ -44,6 +44,7 @@ module controller(
   output logic 	     ALUSrcAE, ALUSrcBE,
   output logic       TargetSrcE,
   output logic       MemReadE,  // for Hazard Unit
+  output logic       JumpE,
   // Memory stage control signals
   input  logic       FlushM,
   input  logic       DataMisalignedM,
@@ -64,7 +65,7 @@ module controller(
   logic 	    RegWriteD, RegWriteE;
   logic [1:0] ResultSrcD, ResultSrcE, ResultSrcM;
   logic [1:0] MemRWD, MemRWE;
-  logic		    JumpD, JumpE;
+  logic		    JumpD;
   logic		    BranchD, BranchE;
   logic	[1:0] ALUOpD;
   logic [4:0] ALUControlD;

wally-pipelined/src/ieu/datapath.sv

@@ -38,7 +38,9 @@ module datapath (
   input  logic [4:0]       ALUControlE,
   input  logic             ALUSrcAE, ALUSrcBE,
   input  logic             TargetSrcE, 
+  input  logic             JumpE,
   input  logic [`XLEN-1:0] PCE,
+  input  logic [`XLEN-1:0] PCLinkE,
   output logic [2:0]       FlagsE,
   output logic [`XLEN-1:0] PCTargetE,
   // Memory stage signals
@@ -67,7 +69,7 @@ module datapath (
   // Execute stage signals
   logic [`XLEN-1:0] RD1E, RD2E;
   logic [`XLEN-1:0] ExtImmE;
-  logic [`XLEN-1:0] PreSrcAE, SrcAE, SrcBE;
+  logic [`XLEN-1:0] PreSrcAE, SrcAE, SrcBE, SrcAE2, SrcBE2;
   logic [`XLEN-1:0] ALUResultE;
   logic [`XLEN-1:0] WriteDataE;
   logic [`XLEN-1:0] TargetBaseE;
@@ -95,8 +97,10 @@ module datapath (
   mux3  #(`XLEN)  faemux(RD1E, ResultW, ALUResultM, ForwardAE, PreSrcAE);
   mux3  #(`XLEN)  fbemux(RD2E, ResultW, ALUResultM, ForwardBE, WriteDataE);
   mux2  #(`XLEN)  srcamux(PreSrcAE, PCE, ALUSrcAE, SrcAE);
+  mux2  #(`XLEN)  srcamux2(SrcAE, PCLinkE, JumpE, SrcAE2);  
   mux2  #(`XLEN)  srcbmux(WriteDataE, ExtImmE, ALUSrcBE, SrcBE);
-  alu   #(`XLEN)  alu(SrcAE, SrcBE, ALUControlE, ALUResultE, FlagsE);
+  mux2  #(`XLEN)  srcbmux2(SrcBE, {`XLEN{1'b0}}, JumpE, SrcBE2); // *** May be able to remove this mux.
+  alu   #(`XLEN)  alu(SrcAE2, SrcBE2, ALUControlE, ALUResultE, FlagsE);
   mux2  #(`XLEN)  targetsrcmux(PCE, SrcAE, TargetSrcE, TargetBaseE);
   assign  PCTargetE = ExtImmE + TargetBaseE;
 
@@ -111,5 +115,7 @@ module datapath (
   floprc #(`XLEN) ALUResultWReg(clk, reset, FlushW, ALUResultM, ALUResultW);
   floprc #(5)    RdWEg(clk, reset, FlushW, RdM, RdW);
 
+  // This mux4:1 no longer needs to include PCLinkW.  This is set correctly in the execution stage.
+  // *** need to look at how the decoder is coded to fix.
   mux4  #(`XLEN) resultmux(ALUResultW, ReadDataW, PCLinkW, CSRReadValW, ResultSrcW, ResultW);	
 endmodule

wally-pipelined/src/ieu/ieu.sv

@@ -33,6 +33,7 @@ module ieu (
   output logic             IllegalBaseInstrFaultD,
   // Execute Stage interface
   input  logic [`XLEN-1:0] PCE, 
+  input  logic [`XLEN-1:0] PCLinkE,
   output logic [`XLEN-1:0] PCTargetE,
   // Memory stage interface
   input  logic             DataMisalignedM,
@@ -68,6 +69,7 @@ module ieu (
   logic [1:0]       ForwardAE, ForwardBE;
   logic             RegWriteM, RegWriteW;
   logic             MemReadE;
+  logic             JumpE;
            
   controller c(.OpD(InstrD[6:0]), .Funct3D(InstrD[14:12]), .Funct7b5D(InstrD[30]), .*);
   datapath   dp(.*);             

wally-pipelined/src/ifu/bpred.sv

@@ -78,6 +78,7 @@ module bpred
   // Part 2 branch direction prediction
 
   twoBitPredictor DirPredictor(.clk(clk),
+			       .reset(reset),
 			       .LookUpPC(PCNextF),
 			       .Prediction(BPPredF),
 			       // update
@@ -110,8 +111,8 @@ module bpred
 			       .UpdateTarget(PCTargetE));
 
   // need to forward when updating to the same address as reading.
-  assign CorrectPCE = PCSrcE ? PCTargetE : PCLinkE;
+  //assign CorrectPCE = PCSrcE ? PCTargetE : PCLinkE;
-  assign TargetPC = (PCE == PCNextF) ? CorrectPCE : BTBPredPCF;
+  //assign TargetPC = (PCE == PCNextF) ? CorrectPCE : BTBPredPCF;
 
   // Part 4 RAS
   // *** need to add the logic to restore RAS on flushes.  We will use incr for this.

wally-pipelined/src/ifu/ifu.sv

@@ -36,6 +36,7 @@ module ifu (
   // Decode  
   output logic 		   InstrStall,
   // Execute
+  output logic [`XLEN-1:0] PCLinkE,
   input logic 		   PCSrcE, 
   input logic [`XLEN-1:0]  PCTargetE,
   output logic [`XLEN-1:0] PCE,
@@ -58,7 +59,7 @@ module ifu (
   logic misaligned, BranchMisalignedFaultE, BranchMisalignedFaultM, TrapMisalignedFaultM;
   logic StallExceptResolveBranchesF, PrivilegedChangePCM;
   logic IllegalCompInstrD;
-  logic [`XLEN-1:0] PCPlusUpperF, PCPlus2or4F, PCD, PCW, PCLinkD, PCLinkE, PCLinkM;
+  logic [`XLEN-1:0] PCPlusUpperF, PCPlus2or4F, PCD, PCW, PCLinkD, PCLinkM;
   logic        CompressedF;
   logic [31:0]     InstrRawD, InstrE;
   logic [31:0]     nop = 32'h00000013; // instruction for NOP

wally-pipelined/src/wally/wallypipelinedhart.sv

@@ -57,7 +57,7 @@ module wallypipelinedhart (
   logic [`XLEN-1:0] SrcAM;
 //  logic [31:0] InstrF;
   logic [31:0] InstrD, InstrM;
-  logic [`XLEN-1:0] PCE, PCM, PCLinkW;
+  logic [`XLEN-1:0] PCE, PCM, PCLinkE, PCLinkW;
   logic [`XLEN-1:0] PCTargetE;
   logic [`XLEN-1:0] CSRReadValW;
   logic [`XLEN-1:0] PrivilegedNextPCM;

wally-pipelined/testbench/testbench-imperas.sv

@@ -408,7 +408,7 @@ string tests32i[] = {
     end // always @ (negedge clk)
 
   // track the current function or label
-  function_rfunction_radix function_radix();
+  //function_rfunction_radix function_radix();
   
 endmodule