Added stalls for pending SFENCE.VMA and FENCE.I in hazard unit

pipelined/src/hazard/hazard.sv

@@ -32,13 +32,13 @@
 
 module hazard(
   // Detect hazards
-(* mark_debug = "true" *)	      input logic  BPPredWrongE, CSRWritePendingDEM, RetM, TrapM,
+(* mark_debug = "true" *)	      input logic  BPPredWrongE, CSRWriteFencePendingDEM, RetM, TrapM,
 (* mark_debug = "true" *)	      input logic  LoadStallD, StoreStallD, MDUStallD, CSRRdStallD,
 (* mark_debug = "true" *)	      input logic  LSUStallM, IFUStallF,
 (* mark_debug = "true" *)              input logic  FPUStallD, FStallD,
 (* mark_debug = "true" *)	      input logic  DivBusyE,FDivBusyE,
 (* mark_debug = "true" *)	      input logic  EcallFaultM, BreakpointFaultM,
-(* mark_debug = "true" *)        input logic  InvalidateICacheM, sfencevmaM, wfiM, IntPendingM,
+(* mark_debug = "true" *)        input logic  wfiM, IntPendingM,
   // Stall & flush outputs
 (* mark_debug = "true" *)	      output logic StallF, StallD, StallE, StallM, StallW,
 (* mark_debug = "true" *)	      output logic FlushF, FlushD, FlushE, FlushM, FlushW
@@ -46,7 +46,6 @@ module hazard(
 
   logic StallFCause, StallDCause, StallECause, StallMCause, StallWCause;
   logic FirstUnstalledD, FirstUnstalledE, FirstUnstalledM, FirstUnstalledW;
-  logic Fence, PrivilegedFlush;
 
   // stalls and flushes
   // loads: stall for one cycle if the subsequent instruction depends on the load
@@ -62,7 +61,7 @@ module hazard(
 
   // *** can stalls be pushed into earlier stages (e.g. no stall after Decode?)
 
-  assign StallFCause = CSRWritePendingDEM & ~(TrapM | RetM | BPPredWrongE);
+  assign StallFCause = CSRWriteFencePendingDEM & ~(TrapM | RetM | BPPredWrongE);
   // stall in decode if instruction is a load/mul/csr dependent on previous
   assign StallDCause = (LoadStallD | StoreStallD | MDUStallD | CSRRdStallD | FPUStallD | FStallD) & ~(TrapM | RetM | BPPredWrongE);    
   assign StallECause = (DivBusyE | FDivBusyE) & ~(TrapM);  // *** can we move to decode stage (KP?)
@@ -82,12 +81,10 @@ module hazard(
   assign FirstUnstalledW = ~StallW & StallM;
   
   // Each stage flushes if the previous stage is the last one stalled (for cause) or the system has reason to flush
-  assign Fence = InvalidateICacheM; // | sfencevmaM; // Fences should flush the pipeline and restart *** sfence not working
-  assign PrivilegedFlush = TrapM | RetM | Fence; // privileged stage change and fences flush pipeline
-  assign FlushF = BPPredWrongE | Fence;
-  assign FlushD = FirstUnstalledD | PrivilegedFlush | BPPredWrongE; 
-  assign FlushE = FirstUnstalledE | PrivilegedFlush | BPPredWrongE; // *** why is BPPredWrongE here, but not needed in simple processor 
-  assign FlushM = FirstUnstalledM | PrivilegedFlush;
+  assign FlushF = BPPredWrongE;
+  assign FlushD = FirstUnstalledD | TrapM | RetM | BPPredWrongE; 
+  assign FlushE = FirstUnstalledE | TrapM | RetM | BPPredWrongE; // *** why is BPPredWrongE here, but not needed in simple processor 
+  assign FlushM = FirstUnstalledM | TrapM | RetM;
   // on Trap the memory stage should be flushed going into the W stage,
   // except if the instruction causing the Trap is an ecall or ebreak.
   assign FlushW = FirstUnstalledW | (TrapM & ~(BreakpointFaultM | EcallFaultM));

pipelined/src/ieu/controller.sv

@@ -67,7 +67,7 @@ module controller(
   output logic 	     RegWriteW,     // for datapath and Hazard Unit
   output logic [2:0] ResultSrcW,
   // Stall during CSRs
-  output logic       CSRWritePendingDEM,
+  output logic       CSRWriteFencePendingDEM,
   output logic       StoreStallD
 );
 
@@ -107,6 +107,8 @@ module controller(
   logic        IEURegWriteE;
   logic        IllegalERegAdrD;
   logic [1:0]  AtomicE;
+   logic       FencePendingD, FencePendingE, FencePendingM;
+   
 
   // Extract fields
   assign OpD = InstrD[6:0];
@@ -177,6 +179,7 @@ module controller(
 
   assign CSRZeroSrcD = InstrD[14] ? (InstrD[19:15] == 0) : (Rs1D == 0); // Is a CSR instruction using zero as the source?
   assign CSRWriteD = CSRReadD & !(CSRZeroSrcD & InstrD[13]); // Don't write if setting or clearing zeros
+  assign FencePendingD = PrivilegedD & (InstrD[31:25] ==  7'b0001001) | FenceD; // possible sfence.vma or fence.i
 
   // ALU Decoding is lazy, only using func7[5] to distinguish add/sub and srl/sra
   assign sltD = (Funct3D == 3'b010);
@@ -203,9 +206,9 @@ module controller(
   flopenrc #(1)  controlregD(clk, reset, FlushD, ~StallD, 1'b1, InstrValidD);
 
   // Execute stage pipeline control register and logic
-  flopenrc #(27) controlregE(clk, reset, FlushE, ~StallE,
-                           {RegWriteD, ResultSrcD, MemRWD, JumpD, BranchD, ALUControlD, ALUSrcAD, ALUSrcBD, ALUResultSrcD, CSRReadD, CSRWriteD, PrivilegedD, Funct3D, W64D, MDUD, AtomicD, InvalidateICacheD, FlushDCacheD, InstrValidD},
-                           {IEURegWriteE, ResultSrcE, MemRWE, JumpE, BranchE, ALUControlE, ALUSrcAE, ALUSrcBE, ALUResultSrcE, CSRReadE, CSRWriteE, PrivilegedE, Funct3E, W64E, MDUE, AtomicE, InvalidateICacheE, FlushDCacheE, InstrValidE});
+  flopenrc #(28) controlregE(clk, reset, FlushE, ~StallE,
+                           {RegWriteD, ResultSrcD, MemRWD, JumpD, BranchD, ALUControlD, ALUSrcAD, ALUSrcBD, ALUResultSrcD, CSRReadD, CSRWriteD, PrivilegedD, Funct3D, W64D, MDUD, AtomicD, InvalidateICacheD, FlushDCacheD, FencePendingD, InstrValidD},
+                           {IEURegWriteE, ResultSrcE, MemRWE, JumpE, BranchE, ALUControlE, ALUSrcAE, ALUSrcBE, ALUResultSrcE, CSRReadE, CSRWriteE, PrivilegedE, Funct3E, W64E, MDUE, AtomicE, InvalidateICacheE, FlushDCacheE, FencePendingE, InstrValidE});
 
   // Branch Logic
   assign {eqE, ltE, ltuE} = FlagsE;
@@ -219,16 +222,17 @@ module controller(
   assign RegWriteE = IEURegWriteE | FWriteIntE; // IRF register writes could come from IEU or FPU controllers
   
   // Memory stage pipeline control register
-  flopenrc #(18) controlregM(clk, reset, FlushM, ~StallM,
-                         {RegWriteE, ResultSrcE, MemRWE, CSRReadE, CSRWriteE, PrivilegedE, Funct3E, FWriteIntE, AtomicE, InvalidateICacheE, FlushDCacheE, InstrValidE},
-                         {RegWriteM, ResultSrcM, MemRWM, CSRReadM, CSRWriteM, PrivilegedM, Funct3M, FWriteIntM, AtomicM, InvalidateICacheM, FlushDCacheM, InstrValidM});
+  flopenrc #(19) controlregM(clk, reset, FlushM, ~StallM,
+                         {RegWriteE, ResultSrcE, MemRWE, CSRReadE, CSRWriteE, PrivilegedE, Funct3E, FWriteIntE, AtomicE, InvalidateICacheE, FlushDCacheE, FencePendingE, InstrValidE},
+                         {RegWriteM, ResultSrcM, MemRWM, CSRReadM, CSRWriteM, PrivilegedM, Funct3M, FWriteIntM, AtomicM, InvalidateICacheM, FlushDCacheM, FencePendingM, InstrValidM});
   
   // Writeback stage pipeline control register
   flopenrc #(4) controlregW(clk, reset, FlushW, ~StallW,
                          {RegWriteM, ResultSrcM},
                          {RegWriteW, ResultSrcW});  
 
-  assign CSRWritePendingDEM = CSRWriteD | CSRWriteE | CSRWriteM;
+  // Stall pipeline at Fetch if a CSR Write or Fence is pending in the subsequent stages
+  assign CSRWriteFencePendingDEM = CSRWriteD | CSRWriteE | CSRWriteM | FencePendingD | FencePendingE | FencePendingM;
 
   assign StoreStallD = MemRWE[0] & ((|MemRWD) | (|AtomicD));
 endmodule

pipelined/src/ieu/ieu.sv

@@ -71,7 +71,7 @@ module ieu (
   output logic 		   FPUStallD, LoadStallD, MDUStallD, CSRRdStallD,
   output logic 		   PCSrcE,
   output logic 		   CSRReadM, CSRWriteM, PrivilegedM,
-  output logic 		   CSRWritePendingDEM,
+  output logic 		   CSRWriteFencePendingDEM,
   output logic             StoreStallD
 );
 
@@ -99,7 +99,7 @@ module ieu (
     .Funct3E, .MDUE, .W64E, .JumpE, .StallM, .FlushM, .MemRWM,
     .CSRReadM, .CSRWriteM, .PrivilegedM, .SCE, .AtomicM, .Funct3M,
     .RegWriteM, .InvalidateICacheM, .FlushDCacheM, .InstrValidM, .FWriteIntM,
-    .StallW, .FlushW, .RegWriteW, .ResultSrcW, .CSRWritePendingDEM, .StoreStallD);
+    .StallW, .FlushW, .RegWriteW, .ResultSrcW, .CSRWriteFencePendingDEM, .StoreStallD);
 
   datapath   dp(
     .clk, .reset, .ImmSrcD, .InstrD, .StallE, .FlushE, .ForwardAE, .ForwardBE,

pipelined/src/wally/wallypipelinedcore.sv

@@ -82,7 +82,7 @@ module wallypipelinedcore (
   logic             StoreAmoMisalignedFaultM, StoreAmoAccessFaultM;
   logic       InvalidateICacheM, FlushDCacheM;
   logic             PCSrcE;
-  logic             CSRWritePendingDEM;
+  logic             CSRWriteFencePendingDEM;
   logic             DivBusyE;
   logic             LoadStallD, StoreStallD, MDUStallD, CSRRdStallD;
   logic             SquashSCW;
@@ -231,7 +231,7 @@ module wallypipelinedcore (
      .FPUStallD, .LoadStallD, .MDUStallD, .CSRRdStallD,
      .PCSrcE,
      .CSRReadM, .CSRWriteM, .PrivilegedM,
-     .CSRWritePendingDEM, .StoreStallD
+     .CSRWriteFencePendingDEM, .StoreStallD
 
   ); // integer execution unit: integer register file, datapath and controller
 
@@ -294,13 +294,13 @@ module wallypipelinedcore (
 
   
    hazard     hzu(
-     .BPPredWrongE, .CSRWritePendingDEM, .RetM, .TrapM,
+     .BPPredWrongE, .CSRWriteFencePendingDEM, .RetM, .TrapM,
      .LoadStallD, .StoreStallD, .MDUStallD, .CSRRdStallD,
      .LSUStallM, .IFUStallF,
      .FPUStallD, .FStallD,
     .DivBusyE, .FDivBusyE,
     .EcallFaultM, .BreakpointFaultM,
-     .InvalidateICacheM, .sfencevmaM, .wfiM, .IntPendingM,
+     .wfiM, .IntPendingM,
      // Stall & flush outputs
     .StallF, .StallD, .StallE, .StallM, .StallW,
     .FlushF, .FlushD, .FlushE, .FlushM, .FlushW