Merge pull request #117 from davidharrishmc/dev

ZMMUL support and MMU cleanup

config/buildroot/wally-config.vh

@@ -136,6 +136,7 @@
 
 
 `define SVADU_SUPPORTED 1
+`define ZMMUL_SUPPORTED 0
 
 // FPU division architecture
 `define RADIX 32'h4

config/fpga/wally-config.vh

@@ -145,6 +145,7 @@
 
 
 `define SVADU_SUPPORTED 1
+`define ZMMUL_SUPPORTED 0
 
 // FPU division architecture
 `define RADIX 32'h4

config/rv32e/wally-config.vh

@@ -139,6 +139,7 @@
 `define BTB_SIZE 10
 
 `define SVADU_SUPPORTED 0
+`define ZMMUL_SUPPORTED 0
 
 // FPU division architecture
 `define RADIX 32'h4

config/rv32gc/wally-config.vh

@@ -138,6 +138,7 @@
 `define BTB_SIZE 10
 
 `define SVADU_SUPPORTED 0
+`define ZMMUL_SUPPORTED 0
 
 // FPU division architecture
 `define RADIX 32'h4

config/rv32i/wally-config.vh

@@ -139,6 +139,7 @@
 `define BTB_SIZE 10
 
 `define SVADU_SUPPORTED 0
+`define ZMMUL_SUPPORTED 0
 
 // FPU division architecture
 `define RADIX 32'h4

config/rv32imc/wally-config.vh

@@ -138,6 +138,7 @@
 `define BTB_SIZE 10
 
 `define SVADU_SUPPORTED 0
+`define ZMMUL_SUPPORTED 0
 
 // FPU division architecture
 `define RADIX 32'h4

config/rv64fpquad/wally-config.vh

@@ -141,6 +141,7 @@
 `define BTB_SIZE 10
 
 `define SVADU_SUPPORTED 0
+`define ZMMUL_SUPPORTED 0
 
 // FPU division architecture
 `define RADIX 32'h4

config/rv64gc/wally-config.vh

@@ -141,6 +141,7 @@
 `define BTB_SIZE 10
 
 `define SVADU_SUPPORTED 0
+`define ZMMUL_SUPPORTED 0
 
 // FPU division architecture
 `define RADIX 32'h4

config/rv64i/wally-config.vh

@@ -141,6 +141,7 @@
 `define BTB_SIZE 10
 
 `define SVADU_SUPPORTED 0
+`define ZMMUL_SUPPORTED 0
 
 // FPU division architecture
 `define RADIX 32'h4

src/ieu/controller.sv

@@ -150,14 +150,14 @@ module controller(
                       ControlsD = `CTRLW'b0_000_00_00_000_0_0_0_0_0_0_0_0_0_00_1; // Non-implemented instruction
       7'b0110011: if (Funct7D == 7'b0000000 | Funct7D == 7'b0100000)
                       ControlsD = `CTRLW'b1_000_00_00_000_0_1_0_0_0_0_0_0_0_00_0; // R-type 
-                  else if (Funct7D == 7'b0000001 & `M_SUPPORTED)
+                  else if (Funct7D == 7'b0000001 & (`M_SUPPORTED | (`ZMMUL_SUPPORTED & ~Funct3D[2])))
                       ControlsD = `CTRLW'b1_000_00_00_011_0_0_0_0_0_0_0_0_1_00_0; // Multiply/divide
                   else
                       ControlsD = `CTRLW'b0_000_00_00_000_0_0_0_0_0_0_0_0_0_00_1; // Non-implemented instruction
       7'b0110111:     ControlsD = `CTRLW'b1_100_01_00_000_0_0_0_1_0_0_0_0_0_00_0; // lui
       7'b0111011: if ((Funct7D == 7'b0000000 | Funct7D == 7'b0100000) & `XLEN == 64)
                       ControlsD = `CTRLW'b1_000_00_00_000_0_1_0_0_1_0_0_0_0_00_0; // R-type W instructions for RV64i
-                  else if (Funct7D == 7'b0000001 & `M_SUPPORTED & `XLEN == 64)
+                  else if (Funct7D == 7'b0000001 & (`M_SUPPORTED | (`ZMMUL_SUPPORTED & ~Funct3D[2])) & `XLEN == 64)
                       ControlsD = `CTRLW'b1_000_00_00_011_0_0_0_0_1_0_0_0_1_00_0; // W-type Multiply/Divide
                   else
                       ControlsD = `CTRLW'b0_000_00_00_000_0_0_0_0_0_0_0_0_0_00_1; // Non-implemented instruction

src/ifu/ifu.sv

@@ -88,7 +88,7 @@ module ifu (
   input logic [1:0] 		STATUS_MPP,                               // Status CSR: previous machine privilege level
   input logic               sfencevmaM,                               // Virtual memory address fence, invalidate TLB entries
   output logic 				ITLBMissF,                                // ITLB miss causes HPTW (hardware pagetable walker) walk
-  output logic              InstrDAPageFaultF,                        // ITLB hit needs to update dirty or access bits
+  output logic              InstrUpdateDAF,                        // ITLB hit needs to update dirty or access bits
   input  var logic [7:0] PMPCFG_ARRAY_REGW[`PMP_ENTRIES-1:0],         // PMP configuration from privileged unit
   input  var logic [`XLEN-1:0] PMPADDR_ARRAY_REGW[`PMP_ENTRIES-1:0],  // PMP address from privileged unit
   output logic 				InstrAccessFaultF,                        // Instruction access fault 
@@ -145,7 +145,7 @@ module ifu (
 
   if(`C_SUPPORTED) begin : Spill
     spill #(`ICACHE_SUPPORTED) spill(.clk, .reset, .StallD, .FlushD, .PCF, .PCPlus4F, .PCNextF, .InstrRawF,
-      .InstrDAPageFaultF, .IFUCacheBusStallD, .ITLBMissF, .PCNextFSpill, .PCFSpill, .SelNextSpillF, .PostSpillInstrRawF, .CompressedF);
+      .InstrUpdateDAF, .IFUCacheBusStallD, .ITLBMissF, .PCNextFSpill, .PCFSpill, .SelNextSpillF, .PostSpillInstrRawF, .CompressedF);
   end else begin : NoSpill
     assign PCNextFSpill = PCNextF;
     assign PCFSpill = PCF;
@@ -185,12 +185,12 @@ module ifu (
          .InstrAccessFaultF, .LoadAccessFaultM(), .StoreAmoAccessFaultM(),
          .InstrPageFaultF, .LoadPageFaultM(), .StoreAmoPageFaultM(),
          .LoadMisalignedFaultM(), .StoreAmoMisalignedFaultM(),
-         .DAPageFault(InstrDAPageFaultF),
+         .UpdateDA(InstrUpdateDAF),
          .AtomicAccessM(1'b0),.ExecuteAccessF(1'b1), .WriteAccessM(1'b0), .ReadAccessM(1'b0),
          .PMPCFG_ARRAY_REGW, .PMPADDR_ARRAY_REGW);
 
   end else begin
-    assign {ITLBMissF, InstrAccessFaultF, InstrPageFaultF, InstrDAPageFaultF} = '0;
+    assign {ITLBMissF, InstrAccessFaultF, InstrPageFaultF, InstrUpdateDAF} = '0;
     assign PCPF = PCFExt[`PA_BITS-1:0];
     assign CacheableF = '1;
     assign SelIROM = '0;

src/ifu/spill.sv

@@ -42,7 +42,7 @@ module spill #(
   input logic [31:0] 	   InstrRawF,         // Instruction from the IROM, I$, or bus. Used to check if the instruction if compressed
   input logic 			   IFUCacheBusStallD, // I$ or bus are stalled. Transition to second fetch of spill after the first is fetched
   input logic 			   ITLBMissF,         // ITLB miss, ignore memory request
-  input logic 			   InstrDAPageFaultF, // Ignore memory request if the hptw support write and a DA page fault occurs (hptw is still active)
+  input logic 			   InstrUpdateDAF, // Ignore memory request if the hptw support write and a DA page fault occurs (hptw is still active)
   output logic [`XLEN-1:0] PCNextFSpill,      // The next PCF for one of the two memory addresses of the spill
   output logic [`XLEN-1:0] PCFSpill,          // PCF for one of the two memory addresses of the spill
   output logic 			   SelNextSpillF,     // During the transition between the two spill operations, the IFU should stall the pipeline
@@ -77,7 +77,7 @@ module spill #(
   ////////////////////////////////////////////////////////////////////////////////////////////////////
 
   assign SpillF = &PCF[$clog2(SPILLTHRESHOLD)+1:1];
-  assign TakeSpillF = SpillF & ~IFUCacheBusStallD & ~(ITLBMissF | (`SVADU_SUPPORTED & InstrDAPageFaultF));
+  assign TakeSpillF = SpillF & ~IFUCacheBusStallD & ~(ITLBMissF | (`SVADU_SUPPORTED & InstrUpdateDAF));
   
   always_ff @(posedge clk)
     if (reset | FlushD)    CurrState <= #1 STATE_READY;

src/lsu/lsu.sv

@@ -81,7 +81,7 @@ module lsu (
   input  logic [1:0]          STATUS_MPP,                           // Machine previous privilege mode
   input  logic [`XLEN-1:0]    PCFSpill,                                  // Fetch PC 
   input  logic                ITLBMissF,                            // ITLB miss causes HPTW (hardware pagetable walker) walk
-  input  logic                InstrDAPageFaultF,                    // ITLB hit needs to update dirty or access bits
+  input  logic                InstrUpdateDAF,                    // ITLB hit needs to update dirty or access bits
   output logic [`XLEN-1:0]    PTE,                                  // Page table entry write to ITLB
   output logic [1:0]          PageType,                             // Type of page table entry to write to ITLB
   output logic                ITLBWriteF,                           // Write PTE to ITLB
@@ -127,7 +127,7 @@ module lsu (
 
   logic                     DTLBMissM;                              // DTLB miss causes HPTW walk
   logic                     DTLBWriteM;                             // Writes PTE and PageType to DTLB
-  logic                     DataDAPageFaultM;                       // DTLB hit needs to update dirty or access bits
+  logic                     DataUpdateDAM;                       // DTLB hit needs to update dirty or access bits
   logic                     LSULoadAccessFaultM;                    // Load acces fault
   logic 					LSUStoreAmoAccessFaultM;                // Store access fault
   logic                     IgnoreRequestTLB;                       // On either ITLB or DTLB miss, ignore miss so HPTW can handle
@@ -151,7 +151,7 @@ module lsu (
 
   if(`VIRTMEM_SUPPORTED) begin : VIRTMEM_SUPPORTED
     hptw hptw(.clk, .reset, .MemRWM, .AtomicM, .ITLBMissF, .ITLBWriteF,
-      .DTLBMissM, .DTLBWriteM, .InstrDAPageFaultF, .DataDAPageFaultM,
+      .DTLBMissM, .DTLBWriteM, .InstrUpdateDAF, .DataUpdateDAM,
       .FlushW, .DCacheStallM, .SATP_REGW, .PCFSpill,
       .STATUS_MXR, .STATUS_SUM, .STATUS_MPRV, .STATUS_MPP, .PrivilegeModeW,
       .ReadDataM(ReadDataM[`XLEN-1:0]), // ReadDataM is LLEN, but HPTW only needs XLEN
@@ -196,7 +196,7 @@ module lsu (
       .StoreAmoAccessFaultM(LSUStoreAmoAccessFaultM), .InstrPageFaultF(), .LoadPageFaultM, 
 	  .StoreAmoPageFaultM,
       .LoadMisalignedFaultM, .StoreAmoMisalignedFaultM,   // *** these faults need to be supressed during hptw.
-      .DAPageFault(DataDAPageFaultM),
+      .UpdateDA(DataUpdateDAM),
       .AtomicAccessM(|LSUAtomicM), .ExecuteAccessF(1'b0), 
       .WriteAccessM(PreLSURWM[0]), .ReadAccessM(PreLSURWM[1]),
       .PMPCFG_ARRAY_REGW, .PMPADDR_ARRAY_REGW);

src/mdu/mdu.sv

@@ -51,16 +51,18 @@ module mdu(
 	// Divider
 	// Start a divide when a new division instruction is received and the divider isn't already busy or finishing
 	// When IDIV_ON_FPU is set, use the FPU divider instead
-	if (`IDIV_ON_FPU) begin  
+	// In ZMMUL, with M_SUPPORTED = 0, omit the divider
+	if ((`IDIV_ON_FPU) || (!`M_SUPPORTED)) begin:nodiv  
 	  assign QuotM = 0;
 	  assign RemM = 0;
 	  assign DivBusyE = 0;
-	end else begin
+	end else begin:div
 		intdivrestoring div(.clk, .reset, .StallM, .FlushE, .DivSignedE(~Funct3E[0]), .W64E, .IntDivE, 
 							.ForwardedSrcAE, .ForwardedSrcBE, .DivBusyE, .QuotM, .RemM);
 	end
 		
 	// Result multiplexer
+	// For ZMMUL, QuotM and RemM are tied to 0, so the mux automatically simplifies
 	always_comb
 		case (Funct3M)	   
 			3'b000: PrelimResultM = ProdM[`XLEN-1:0];					// mul

src/mmu/hptw.sv

@@ -49,8 +49,8 @@ module hptw (
 	input  logic                ITLBMissF,
 	input  logic                DTLBMissM,
 	input  logic                FlushW,
-	input  logic                InstrDAPageFaultF,
-	input  logic                DataDAPageFaultM,
+	input  logic                InstrUpdateDAF,
+	input  logic                DataUpdateDAM,
 	output logic [`XLEN-1:0]    PTE, 								// page table entry to TLBs
 	output logic [1:0]          PageType, 					// page type to TLBs
 	output logic ITLBWriteF, DTLBWriteM, // write TLB with new entry
@@ -87,21 +87,23 @@ module hptw (
   logic [`XLEN-1:0] 	   TranslationVAdr;
   logic [`XLEN-1:0] 	   NextPTE;
   logic 				   UpdatePTE;
-  logic 				   HPTWDAPageFault;
+  logic 				   HPTWUpdateDA;
   logic [`PA_BITS-1:0] 	   HPTWReadAdr;
   logic 				   SelHPTWAdr;
   logic [`XLEN+1:0] 	   HPTWAdrExt;
   logic 				   ITLBMissOrDAFaultF;
   logic 				   DTLBMissOrDAFaultM;
+  logic                    LSUAccessFaultM;
   logic [`PA_BITS-1:0] 	   HPTWAdr;
   logic [1:0] 			   HPTWRW;
   logic [2:0] 			   HPTWSize; // 32 or 64 bit access
   statetype WalkerState, NextWalkerState, InitialWalkerState;
 
   // map hptw access faults onto either the original LSU load/store fault or instruction access fault
-  assign LoadAccessFaultM 		 = WalkerState == IDLE ? LSULoadAccessFaultM : (LSULoadAccessFaultM | LSUStoreAmoAccessFaultM) & DTLBWalk & MemRWM[1] & ~MemRWM[0];
-  assign StoreAmoAccessFaultM	 = WalkerState == IDLE ? LSUStoreAmoAccessFaultM : (LSULoadAccessFaultM | LSUStoreAmoAccessFaultM) & DTLBWalk & MemRWM[0];
-  assign HPTWInstrAccessFaultM = WalkerState == IDLE ? 1'b0: (LSUStoreAmoAccessFaultM | LSULoadAccessFaultM) & ~DTLBWalk;
+  assign LSUAccessFaultM         = LSULoadAccessFaultM | LSUStoreAmoAccessFaultM;
+  assign LoadAccessFaultM 		 = WalkerState == IDLE ? LSULoadAccessFaultM : LSUAccessFaultM & DTLBWalk & MemRWM[1] & ~MemRWM[0];
+  assign StoreAmoAccessFaultM	 = WalkerState == IDLE ? LSUStoreAmoAccessFaultM : LSUAccessFaultM & DTLBWalk & MemRWM[0];
+  assign HPTWInstrAccessFaultM   = WalkerState == IDLE ? 1'b0: LSUAccessFaultM & ~DTLBWalk;
 
 	// Extract bits from CSRs and inputs
 	assign SvMode = SATP_REGW[`XLEN-1:`XLEN-`SVMODE_BITS];
@@ -127,8 +129,8 @@ module hptw (
 
   if(`SVADU_SUPPORTED) begin : hptwwrites
     logic                     ReadAccess, WriteAccess;
-    logic                     InvalidRead, InvalidWrite;
-    logic                     UpperBitsUnequalPageFault; 
+    logic                     InvalidRead, InvalidWrite, InvalidOp;
+    logic                     UpperBitsUnequal; 
     logic                     OtherPageFault;
     logic [1:0]               EffectivePrivilegeMode;
     logic                     ImproperPrivilege;
@@ -147,7 +149,7 @@ module hptw (
     mux2 #(`PA_BITS) HPTWWriteAdrMux(HPTWReadAdr, HPTWWriteAdr, SelHPTWWriteAdr, HPTWAdr); 
 
     assign {Dirty, Accessed} = PTE[7:6];
-    assign WriteAccess = MemRWM[0] | (|AtomicM);
+    assign WriteAccess = MemRWM[0]; // implies | (|AtomicM);
     assign SetDirty = ~Dirty & DTLBWalk & WriteAccess;
     assign ReadAccess = MemRWM[1];
 
@@ -157,24 +159,24 @@ module hptw (
 
     // Check for page faults
 		vm64check vm64check(.SATP_MODE(SATP_REGW[`XLEN-1:`XLEN-`SVMODE_BITS]), .VAdr(TranslationVAdr), 
-	  	.SV39Mode(), .UpperBitsUnequalPageFault);
+	  	.SV39Mode(), .UpperBitsUnequal);
     assign InvalidRead = ReadAccess & ~Readable & (~STATUS_MXR | ~Executable);
     assign InvalidWrite = WriteAccess & ~Writable;
-    assign OtherPageFault = DTLBWalk? ImproperPrivilege | InvalidRead | InvalidWrite | UpperBitsUnequalPageFault | Misaligned | ~Valid :
-                            ImproperPrivilege | ~Executable | UpperBitsUnequalPageFault | Misaligned | ~Valid;
+	assign InvalidOp = DTLBWalk ? (InvalidRead | InvalidWrite) : ~Executable;
+    assign OtherPageFault = ImproperPrivilege | InvalidOp | UpperBitsUnequal | Misaligned | ~Valid;
 
     // hptw needs to know if there is a Dirty or Access fault occuring on this
     // memory access.  If there is the PTE needs to be updated seting Access
     // and possibly also Dirty.  Dirty is set if the operation is a store/amo.
     // However any other fault should not cause the update.
-    assign HPTWDAPageFault = ValidLeafPTE & (~Accessed | SetDirty) & ~OtherPageFault;
+    assign HPTWUpdateDA = ValidLeafPTE & (~Accessed | SetDirty) & ~OtherPageFault;
 
     assign HPTWRW[0] = (WalkerState == UPDATE_PTE);
-    assign UpdatePTE = (WalkerState == LEAF) & HPTWDAPageFault;
+    assign UpdatePTE = (WalkerState == LEAF) & HPTWUpdateDA;
   end else begin // block: hptwwrites
     assign NextPTE = ReadDataM;
     assign HPTWAdr = HPTWReadAdr;
-    assign HPTWDAPageFault = '0;
+    assign HPTWUpdateDA = '0;
     assign UpdatePTE = '0;
     assign HPTWRW[0] = '0;
   end
@@ -182,8 +184,8 @@ module hptw (
 	// Enable and select signals based on states
 	assign StartWalk = (WalkerState == IDLE) & TLBMiss;
 	assign HPTWRW[1] = (WalkerState == L3_RD) | (WalkerState == L2_RD) | (WalkerState == L1_RD) | (WalkerState == L0_RD);
-	assign DTLBWriteM = (WalkerState == LEAF & ~HPTWDAPageFault) & DTLBWalk;
-	assign ITLBWriteF = (WalkerState == LEAF & ~HPTWDAPageFault) & ~DTLBWalk;
+	assign DTLBWriteM = (WalkerState == LEAF & ~HPTWUpdateDA) & DTLBWalk;
+	assign ITLBWriteF = (WalkerState == LEAF & ~HPTWUpdateDA) & ~DTLBWalk;
   
 	// FSM to track PageType based on the levels of the page table traversed
 	flopr #(2) PageTypeReg(clk, reset, NextPageType, PageType);
@@ -262,7 +264,7 @@ module hptw (
 					else                                 										NextWalkerState = LEAF;
 			L0_RD: if (DCacheStallM)                     								NextWalkerState = L0_RD;
 				   else                                     							NextWalkerState = LEAF;
-			LEAF: if (`SVADU_SUPPORTED & HPTWDAPageFault)             NextWalkerState = UPDATE_PTE;
+			LEAF: if (`SVADU_SUPPORTED & HPTWUpdateDA)             NextWalkerState = UPDATE_PTE;
 				  else 																										NextWalkerState = IDLE;
 			UPDATE_PTE: if(DCacheStallM) 		                        		NextWalkerState = UPDATE_PTE;
 						else 																									NextWalkerState = LEAF;
@@ -273,8 +275,8 @@ module hptw (
   assign SelHPTW = WalkerState != IDLE;
   assign HPTWStall = (WalkerState != IDLE) | (WalkerState == IDLE & TLBMiss);
 
-  assign ITLBMissOrDAFaultF = ITLBMissF | (`SVADU_SUPPORTED & InstrDAPageFaultF);
-  assign DTLBMissOrDAFaultM = DTLBMissM | (`SVADU_SUPPORTED & DataDAPageFaultM);  
+  assign ITLBMissOrDAFaultF = ITLBMissF | (`SVADU_SUPPORTED & InstrUpdateDAF);
+  assign DTLBMissOrDAFaultM = DTLBMissM | (`SVADU_SUPPORTED & DataUpdateDAM);  
 
   // HTPW address/data/control muxing
 

src/mmu/mmu.sv

@@ -51,7 +51,7 @@ module mmu #(parameter TLB_ENTRIES = 8, IMMU = 0) (
   // Faults
   output logic                InstrAccessFaultF, LoadAccessFaultM, StoreAmoAccessFaultM,  // access fault sources
   output logic                InstrPageFaultF, LoadPageFaultM, StoreAmoPageFaultM,        // page fault sources
-  output logic                DAPageFault,                                                // page fault due to setting dirty or access bit
+  output logic                UpdateDA,                                                // page fault due to setting dirty or access bit
   output logic                LoadMisalignedFaultM, StoreAmoMisalignedFaultM,             // misaligned fault sources
   // PMA checker signals
   input  logic                 AtomicAccessM, ExecuteAccessF, WriteAccessM, ReadAccessM,  // access type
@@ -70,6 +70,7 @@ module mmu #(parameter TLB_ENTRIES = 8, IMMU = 0) (
   logic                       Translate;                // Translation occurs when virtual memory is active and DisableTranslation is off
   logic                       TLBHit;                   // Hit in TLB
   logic                       TLBPageFault;             // Page fault from TLB
+  logic                       ReadNoAmoAccessM;         // Read that is not part of atomic operation causes Load faults.  Otherwise StoreAmo faults
   
   // only instantiate TLB if Virtual Memory is supported
   if (`VIRTMEM_SUPPORTED) begin:tlb
@@ -84,7 +85,7 @@ module mmu #(parameter TLB_ENTRIES = 8, IMMU = 0) (
           .PrivilegeModeW, .ReadAccess, .WriteAccess,
           .DisableTranslation, .PTE, .PageTypeWriteVal,
           .TLBWrite, .TLBFlush, .TLBPAdr, .TLBMiss, .TLBHit, 
-          .Translate, .TLBPageFault, .DAPageFault);
+          .Translate, .TLBPageFault, .UpdateDA);
   end else begin:tlb// just pass address through as physical
     assign Translate = 0;
     assign TLBMiss = 0;
@@ -118,6 +119,8 @@ module mmu #(parameter TLB_ENTRIES = 8, IMMU = 0) (
     assign PMPLoadAccessFaultM      = 0;
   end
 
+  assign ReadNoAmoAccessM = ReadAccessM & ~WriteAccessM;// AMO causes StoreAmo rather than Load fault
+
   // Access faults
   // If TLB miss and translating we want to not have faults from the PMA and PMP checkers.
   assign InstrAccessFaultF    = (PMAInstrAccessFaultF    | PMPInstrAccessFaultF)    & ~TLBMiss;
@@ -132,11 +135,11 @@ module mmu #(parameter TLB_ENTRIES = 8, IMMU = 0) (
       2'b10:  DataMisalignedM = VAdr[1] | VAdr[0]; // lw, sw, flw, fsw, lwu
       2'b11:  DataMisalignedM = |VAdr[2:0];        // ld, sd, fld, fsd
     endcase 
-  assign LoadMisalignedFaultM     = DataMisalignedM & ReadAccessM;
-  assign StoreAmoMisalignedFaultM = DataMisalignedM & (WriteAccessM | AtomicAccessM);
+  assign LoadMisalignedFaultM     = DataMisalignedM & ReadNoAmoAccessM;
+  assign StoreAmoMisalignedFaultM = DataMisalignedM & WriteAccessM;
 
   // Specify which type of page fault is occurring
   assign InstrPageFaultF    = TLBPageFault & ExecuteAccessF;
-  assign LoadPageFaultM     = TLBPageFault & ReadAccessM;
-  assign StoreAmoPageFaultM = TLBPageFault & (WriteAccessM | AtomicAccessM);
+  assign LoadPageFaultM     = TLBPageFault & ReadNoAmoAccessM; 
+  assign StoreAmoPageFaultM = TLBPageFault & WriteAccessM;
 endmodule

src/mmu/tlb/tlb.sv

@@ -72,7 +72,7 @@ module tlb #(parameter TLB_ENTRIES = 8, ITLB = 0) (
   output logic                    TLBHit,
   output logic                    Translate,
   output logic                    TLBPageFault,
-  output logic                    DAPageFault
+  output logic                    UpdateDA
 );
 
   logic [TLB_ENTRIES-1:0]         Matches, WriteEnables, PTE_Gs; // used as the one-hot encoding of WriteIndex
@@ -105,7 +105,7 @@ module tlb #(parameter TLB_ENTRIES = 8, ITLB = 0) (
   tlbcontrol #(ITLB) tlbcontrol(.SATP_MODE, .VAdr, .STATUS_MXR, .STATUS_SUM, .STATUS_MPRV, .STATUS_MPP,
     .PrivilegeModeW, .ReadAccess, .WriteAccess, .DisableTranslation, .TLBFlush,
     .PTEAccessBits, .CAMHit, .Misaligned, .TLBMiss, .TLBHit, .TLBPageFault, 
-    .DAPageFault, .SV39Mode, .Translate);
+    .UpdateDA, .SV39Mode, .Translate);
 
   tlblru #(TLB_ENTRIES) lru(.clk, .reset, .TLBWrite, .TLBFlush, .Matches, .CAMHit, .WriteEnables);
   tlbcam #(TLB_ENTRIES, `VPN_BITS + `ASID_BITS, `VPN_SEGMENT_BITS) 

src/mmu/tlb/tlbcontrol.sv

@@ -43,7 +43,7 @@ module tlbcontrol #(parameter ITLB = 0) (
   output logic                    TLBMiss,
   output logic                    TLBHit,
   output logic                    TLBPageFault,
-  output logic                    DAPageFault,
+  output logic                    UpdateDA,
   output logic                    SV39Mode,
   output logic                    Translate
 );
@@ -52,7 +52,7 @@ module tlbcontrol #(parameter ITLB = 0) (
   logic [1:0]                     EffectivePrivilegeMode;
 
   logic                           PTE_D, PTE_A, PTE_U, PTE_X, PTE_W, PTE_R, PTE_V; // Useful PTE Control Bits
-  logic                           UpperBitsUnequalPageFault;
+  logic                           UpperBitsUnequal;
   logic                           TLBAccess;
   logic                           ImproperPrivilege;
 
@@ -64,7 +64,7 @@ module tlbcontrol #(parameter ITLB = 0) (
   assign TLBAccess = ReadAccess | WriteAccess;
 
   // Check that upper bits are legal (all 0s or all 1s)
-  vm64check vm64check(.SATP_MODE, .VAdr, .SV39Mode, .UpperBitsUnequalPageFault);
+  vm64check vm64check(.SATP_MODE, .VAdr, .SV39Mode, .UpperBitsUnequal);
 
   // unswizzle useful PTE bits
   assign {PTE_D, PTE_A} = PTEAccessBits[7:6];
@@ -77,12 +77,12 @@ module tlbcontrol #(parameter ITLB = 0) (
     assign ImproperPrivilege = ((EffectivePrivilegeMode == `U_MODE) & ~PTE_U) |
       ((EffectivePrivilegeMode == `S_MODE) & PTE_U);
     if(`SVADU_SUPPORTED) begin : hptwwrites
-      assign DAPageFault = Translate & TLBHit & ~PTE_A & ~TLBPageFault;
-      assign TLBPageFault = (Translate  & TLBHit & (ImproperPrivilege | ~PTE_X | UpperBitsUnequalPageFault | Misaligned | ~PTE_V));
+      assign UpdateDA = Translate & TLBHit & ~PTE_A & ~TLBPageFault;
+      assign TLBPageFault = Translate  & TLBHit & (ImproperPrivilege | ~PTE_X | UpperBitsUnequal | Misaligned | ~PTE_V);
     end else begin
       // fault for software handling if access bit is off
-      assign DAPageFault = ~PTE_A;
-      assign TLBPageFault = (Translate  & TLBHit & (ImproperPrivilege | ~PTE_X | DAPageFault | UpperBitsUnequalPageFault | Misaligned | ~PTE_V));
+      assign UpdateDA = ~PTE_A;
+      assign TLBPageFault = Translate  & TLBHit & (ImproperPrivilege | ~PTE_X | UpdateDA | UpperBitsUnequal | Misaligned | ~PTE_V);
     end
   end else begin:dtlb // Data TLB fault checking
     logic InvalidRead, InvalidWrite;
@@ -99,12 +99,12 @@ module tlbcontrol #(parameter ITLB = 0) (
     // low.
     assign InvalidWrite = WriteAccess & ~PTE_W;
     if(`SVADU_SUPPORTED) begin : hptwwrites
-      assign DAPageFault = Translate & TLBHit & (~PTE_A | WriteAccess & ~PTE_D) & ~TLBPageFault; 
-      assign TLBPageFault =  (Translate & TLBHit & (ImproperPrivilege | InvalidRead | InvalidWrite | UpperBitsUnequalPageFault | Misaligned | ~PTE_V));
+      assign UpdateDA = Translate & TLBHit & (~PTE_A | WriteAccess & ~PTE_D) & ~TLBPageFault; 
+      assign TLBPageFault =  (Translate & TLBHit & (ImproperPrivilege | InvalidRead | InvalidWrite | UpperBitsUnequal | Misaligned | ~PTE_V));
     end else begin
       // Fault for software handling if access bit is off or writing a page with dirty bit off
-      assign DAPageFault = ~PTE_A | WriteAccess & ~PTE_D; 
-      assign TLBPageFault = (Translate & TLBHit & (ImproperPrivilege | InvalidRead | InvalidWrite | DAPageFault | UpperBitsUnequalPageFault | Misaligned | ~PTE_V));
+      assign UpdateDA = ~PTE_A | WriteAccess & ~PTE_D; 
+      assign TLBPageFault = (Translate & TLBHit & (ImproperPrivilege | InvalidRead | InvalidWrite | UpdateDA | UpperBitsUnequal | Misaligned | ~PTE_V));
     end
   end
 

src/mmu/tlb/vm64check.sv

@@ -32,7 +32,7 @@ module vm64check (
   input  logic [`SVMODE_BITS-1:0] SATP_MODE,
   input  logic [`XLEN-1:0]        VAdr,
   output logic                    SV39Mode, 
-  output logic                    UpperBitsUnequalPageFault
+  output logic                    UpperBitsUnequal
 );
 
   if (`XLEN == 64) begin
@@ -42,9 +42,9 @@ module vm64check (
     logic                           eq_63_47, eq_46_38;
     assign eq_46_38 = &(VAdr[46:38]) | ~|(VAdr[46:38]);
     assign eq_63_47 = &(VAdr[63:47]) | ~|(VAdr[63:47]); 
-    assign UpperBitsUnequalPageFault = SV39Mode ? ~(eq_63_47 & eq_46_38) : ~eq_63_47;
+    assign UpperBitsUnequal = SV39Mode ? ~(eq_63_47 & eq_46_38) : ~eq_63_47;
   end else begin
     assign SV39Mode = 0;
-    assign UpperBitsUnequalPageFault = 0;
+    assign UpperBitsUnequal = 0;
   end
 endmodule

src/wally/wallypipelinedcore.sv

@@ -156,7 +156,7 @@ module wallypipelinedcore (
   logic                          ICacheMiss;
   logic                          ICacheAccess;
   logic                          BreakpointFaultM, EcallFaultM;
-  logic                          InstrDAPageFaultF;
+  logic                          InstrUpdateDAF;
   logic                          BigEndianM;
   logic                          FCvtIntE;
   logic                          CommittedF;
@@ -184,7 +184,7 @@ module wallypipelinedcore (
     .PrivilegeModeW, .PTE, .PageType, .SATP_REGW, .STATUS_MXR, .STATUS_SUM, .STATUS_MPRV,
     .STATUS_MPP, .ITLBWriteF, .sfencevmaM, .ITLBMissF,
     // pmp/pma (inside mmu) signals. 
-    .PMPCFG_ARRAY_REGW,  .PMPADDR_ARRAY_REGW, .InstrAccessFaultF, .InstrDAPageFaultF); 
+    .PMPCFG_ARRAY_REGW,  .PMPADDR_ARRAY_REGW, .InstrAccessFaultF, .InstrUpdateDAF); 
     
   // integer execution unit: integer register file, datapath and controller
   ieu ieu(.clk, .reset,
@@ -238,7 +238,7 @@ module wallypipelinedcore (
     .HPTWInstrAccessFaultM,         // connects to privilege
     .StoreAmoMisalignedFaultM, // connects to privilege
     .StoreAmoAccessFaultM,     // connects to privilege
-    .InstrDAPageFaultF,
+    .InstrUpdateDAF,
     .PCFSpill, .ITLBMissF, .PTE, .PageType, .ITLBWriteF, .SelHPTW,
     .LSUStallM);                    
 
@@ -313,7 +313,7 @@ module wallypipelinedcore (
   end
 
   // multiply/divide unit
-  if (`M_SUPPORTED) begin:mdu
+  if (`M_SUPPORTED | `ZMMUL_SUPPORTED) begin:mdu
     mdu mdu(.clk, .reset, .StallM, .StallW, .FlushE, .FlushM, .FlushW,
       .ForwardedSrcAE, .ForwardedSrcBE, 
       .Funct3E, .Funct3M, .IntDivE, .W64E,

testbench/common/riscvassertions.sv

@@ -23,40 +23,42 @@
 
 module riscvassertions;
   initial begin
-    assert (`PMP_ENTRIES == 0 | `PMP_ENTRIES==16 | `PMP_ENTRIES==64) else $error("Illegal number of PMP entries: PMP_ENTRIES must be 0, 16, or 64");
-    assert (`S_SUPPORTED | `VIRTMEM_SUPPORTED == 0) else $error("Virtual memory requires S mode support");
-    assert (`IDIV_BITSPERCYCLE == 1 | `IDIV_BITSPERCYCLE==2 | `IDIV_BITSPERCYCLE==4) else $error("Illegal number of divider bits/cycle: IDIV_BITSPERCYCLE must be 1, 2, or 4");
-    assert (`F_SUPPORTED | ~`D_SUPPORTED) else $error("Can't support double fp (D) without supporting float (F)");
-    assert (`D_SUPPORTED | ~`Q_SUPPORTED) else $error("Can't support quad fp (Q) without supporting double (D)");
-    assert (`F_SUPPORTED | ~`ZFH_SUPPORTED) else $error("Can't support half-precision fp (ZFH) without supporting float (F)");
-    assert (`DCACHE_SUPPORTED | ~`F_SUPPORTED | `FLEN <= `XLEN) else $error("Data cache required to support FLEN > XLEN because AHB bus width is XLEN");
+    $display("IDIV_ON_FPU = %b M_SUPPORTED %b comb %b\n", `IDIV_ON_FPU, `M_SUPPORTED, ((`IDIV_ON_FPU) || (!`M_SUPPORTED)));
+    assert (`PMP_ENTRIES == 0 || `PMP_ENTRIES==16 || `PMP_ENTRIES==64) else $error("Illegal number of PMP entries: PMP_ENTRIES must be 0, 16, or 64");
+    assert (`S_SUPPORTED || `VIRTMEM_SUPPORTED == 0) else $error("Virtual memory requires S mode support");
+    assert (`IDIV_BITSPERCYCLE == 1 || `IDIV_BITSPERCYCLE==2 || `IDIV_BITSPERCYCLE==4) else $error("Illegal number of divider bits/cycle: IDIV_BITSPERCYCLE must be 1, 2, or 4");
+    assert (`F_SUPPORTED || ~`D_SUPPORTED) else $error("Can't support double fp (D) without supporting float (F)");
+    assert (`D_SUPPORTED || ~`Q_SUPPORTED) else $error("Can't support quad fp (Q) without supporting double (D)");
+    assert (`F_SUPPORTED || ~`ZFH_SUPPORTED) else $error("Can't support half-precision fp (ZFH) without supporting float (F)");
+    assert (`DCACHE_SUPPORTED || ~`F_SUPPORTED || `FLEN <= `XLEN) else $error("Data cache required to support FLEN > XLEN because AHB bus width is XLEN");
     assert (`I_SUPPORTED ^ `E_SUPPORTED) else $error("Exactly one of I and E must be supported");
-    assert (`FLEN<=`XLEN | `DCACHE_SUPPORTED | `DTIM_SUPPORTED) else $error("Wally does not support FLEN > XLEN unleses data cache or DTIM is supported");
-    assert (`DCACHE_WAYSIZEINBYTES <= 4096 | (!`DCACHE_SUPPORTED) | `VIRTMEM_SUPPORTED == 0) else $error("DCACHE_WAYSIZEINBYTES cannot exceed 4 KiB when caches and vitual memory is enabled (to prevent aliasing)");
-    assert (`DCACHE_LINELENINBITS >= 128 | (!`DCACHE_SUPPORTED)) else $error("DCACHE_LINELENINBITS must be at least 128 when caches are enabled");
+    assert (`FLEN<=`XLEN || `DCACHE_SUPPORTED || `DTIM_SUPPORTED) else $error("Wally does not support FLEN > XLEN unleses data cache or DTIM is supported");
+    assert (`DCACHE_WAYSIZEINBYTES <= 4096 || (!`DCACHE_SUPPORTED) || `VIRTMEM_SUPPORTED == 0) else $error("DCACHE_WAYSIZEINBYTES cannot exceed 4 KiB when caches and vitual memory is enabled (to prevent aliasing)");
+    assert (`DCACHE_LINELENINBITS >= 128 || (!`DCACHE_SUPPORTED)) else $error("DCACHE_LINELENINBITS must be at least 128 when caches are enabled");
     assert (`DCACHE_LINELENINBITS < `DCACHE_WAYSIZEINBYTES*8) else $error("DCACHE_LINELENINBITS must be smaller than way size");
-    assert (`ICACHE_WAYSIZEINBYTES <= 4096 | (!`ICACHE_SUPPORTED) | `VIRTMEM_SUPPORTED == 0) else $error("ICACHE_WAYSIZEINBYTES cannot exceed 4 KiB when caches and vitual memory is enabled (to prevent aliasing)");
-    assert (`ICACHE_LINELENINBITS >= 32 | (!`ICACHE_SUPPORTED)) else $error("ICACHE_LINELENINBITS must be at least 32 when caches are enabled");
+    assert (`ICACHE_WAYSIZEINBYTES <= 4096 || (!`ICACHE_SUPPORTED) || `VIRTMEM_SUPPORTED == 0) else $error("ICACHE_WAYSIZEINBYTES cannot exceed 4 KiB when caches and vitual memory is enabled (to prevent aliasing)");
+    assert (`ICACHE_LINELENINBITS >= 32 || (!`ICACHE_SUPPORTED)) else $error("ICACHE_LINELENINBITS must be at least 32 when caches are enabled");
     assert (`ICACHE_LINELENINBITS < `ICACHE_WAYSIZEINBYTES*8) else $error("ICACHE_LINELENINBITS must be smaller than way size");
-    assert (2**$clog2(`DCACHE_LINELENINBITS) == `DCACHE_LINELENINBITS | (!`DCACHE_SUPPORTED)) else $error("DCACHE_LINELENINBITS must be a power of 2");
-    assert (2**$clog2(`DCACHE_WAYSIZEINBYTES) == `DCACHE_WAYSIZEINBYTES | (!`DCACHE_SUPPORTED)) else $error("DCACHE_WAYSIZEINBYTES must be a power of 2");
-    assert (2**$clog2(`ICACHE_LINELENINBITS) == `ICACHE_LINELENINBITS | (!`ICACHE_SUPPORTED)) else $error("ICACHE_LINELENINBITS must be a power of 2");
-    assert (2**$clog2(`ICACHE_WAYSIZEINBYTES) == `ICACHE_WAYSIZEINBYTES | (!`ICACHE_SUPPORTED)) else $error("ICACHE_WAYSIZEINBYTES must be a power of 2");
-    assert (2**$clog2(`ITLB_ENTRIES) == `ITLB_ENTRIES | `VIRTMEM_SUPPORTED==0) else $error("ITLB_ENTRIES must be a power of 2");
-    assert (2**$clog2(`DTLB_ENTRIES) == `DTLB_ENTRIES | `VIRTMEM_SUPPORTED==0) else $error("DTLB_ENTRIES must be a power of 2");
+    assert (2**$clog2(`DCACHE_LINELENINBITS) == `DCACHE_LINELENINBITS || (!`DCACHE_SUPPORTED)) else $error("DCACHE_LINELENINBITS must be a power of 2");
+    assert (2**$clog2(`DCACHE_WAYSIZEINBYTES) == `DCACHE_WAYSIZEINBYTES || (!`DCACHE_SUPPORTED)) else $error("DCACHE_WAYSIZEINBYTES must be a power of 2");
+    assert (2**$clog2(`ICACHE_LINELENINBITS) == `ICACHE_LINELENINBITS || (!`ICACHE_SUPPORTED)) else $error("ICACHE_LINELENINBITS must be a power of 2");
+    assert (2**$clog2(`ICACHE_WAYSIZEINBYTES) == `ICACHE_WAYSIZEINBYTES || (!`ICACHE_SUPPORTED)) else $error("ICACHE_WAYSIZEINBYTES must be a power of 2");
+    assert (2**$clog2(`ITLB_ENTRIES) == `ITLB_ENTRIES || `VIRTMEM_SUPPORTED==0) else $error("ITLB_ENTRIES must be a power of 2");
+    assert (2**$clog2(`DTLB_ENTRIES) == `DTLB_ENTRIES || `VIRTMEM_SUPPORTED==0) else $error("DTLB_ENTRIES must be a power of 2");
     assert (`UNCORE_RAM_RANGE >= 56'h07FFFFFF) else $warning("Some regression tests will fail if UNCORE_RAM_RANGE is less than 56'h07FFFFFF");
-	  assert (`ZICSR_SUPPORTED == 1 | (`PMP_ENTRIES == 0 & `VIRTMEM_SUPPORTED == 0)) else $error("PMP_ENTRIES and VIRTMEM_SUPPORTED must be zero if ZICSR not supported.");
-    assert (`ZICSR_SUPPORTED == 1 | (`S_SUPPORTED == 0 & `U_SUPPORTED == 0)) else $error("S and U modes not supported if ZICSR not supported");
-    assert (`U_SUPPORTED | (`S_SUPPORTED == 0)) else $error ("S mode only supported if U also is supported");
-    assert (`VIRTMEM_SUPPORTED == 0 | (`DTIM_SUPPORTED == 0 & `IROM_SUPPORTED == 0)) else $error("Can't simultaneously have virtual memory and DTIM_SUPPORTED/IROM_SUPPORTED because local memories don't translate addresses");
-    assert (`DCACHE_SUPPORTED | `VIRTMEM_SUPPORTED ==0) else $error("Virtual memory needs dcache");
-    assert (`ICACHE_SUPPORTED | `VIRTMEM_SUPPORTED ==0) else $error("Virtual memory needs icache");
-    assert ((`DCACHE_SUPPORTED == 0 & `ICACHE_SUPPORTED == 0) | `BUS_SUPPORTED) else $error("Dcache and Icache requires DBUS_SUPPORTED.");
-    assert (`DCACHE_LINELENINBITS <= `XLEN*16 | (!`DCACHE_SUPPORTED)) else $error("DCACHE_LINELENINBITS must not exceed 16 words because max AHB burst size is 1");
+	  assert (`ZICSR_SUPPORTED == 1 || (`PMP_ENTRIES == 0 && `VIRTMEM_SUPPORTED == 0)) else $error("PMP_ENTRIES and VIRTMEM_SUPPORTED must be zero if ZICSR not supported.");
+    assert (`ZICSR_SUPPORTED == 1 || (`S_SUPPORTED == 0 && `U_SUPPORTED == 0)) else $error("S and U modes not supported if ZICSR not supported");
+    assert (`U_SUPPORTED || (`S_SUPPORTED == 0)) else $error ("S mode only supported if U also is supported");
+    assert (`VIRTMEM_SUPPORTED == 0 || (`DTIM_SUPPORTED == 0 && `IROM_SUPPORTED == 0)) else $error("Can't simultaneously have virtual memory and DTIM_SUPPORTED/IROM_SUPPORTED because local memories don't translate addresses");
+    assert (`DCACHE_SUPPORTED || `VIRTMEM_SUPPORTED ==0) else $error("Virtual memory needs dcache");
+    assert (`ICACHE_SUPPORTED || `VIRTMEM_SUPPORTED ==0) else $error("Virtual memory needs icache");
+    assert ((`DCACHE_SUPPORTED == 0 && `ICACHE_SUPPORTED == 0) || `BUS_SUPPORTED) else $error("Dcache and Icache requires DBUS_SUPPORTED.");
+    assert (`DCACHE_LINELENINBITS <= `XLEN*16 || (!`DCACHE_SUPPORTED)) else $error("DCACHE_LINELENINBITS must not exceed 16 words because max AHB burst size is 1");
     assert (`DCACHE_LINELENINBITS % 4 == 0) else $error("DCACHE_LINELENINBITS must hold 4, 8, or 16 words");
-    assert (`DCACHE_SUPPORTED | `A_SUPPORTED == 0) else $error("Atomic extension (A) requires cache on Wally.");
-    assert (`IDIV_ON_FPU == 0 | `F_SUPPORTED) else $error("IDIV on FPU needs F_SUPPORTED");
-    assert (`SSTC_SUPPORTED == 0 | (`S_SUPPORTED)) else $error("SSTC requires S_SUPPORTED");
+    assert (`DCACHE_SUPPORTED || (`A_SUPPORTED == 0)) else $error("Atomic extension (A) requires cache on Wally.");
+    assert (`IDIV_ON_FPU == 0 || `F_SUPPORTED) else $error("IDIV on FPU needs F_SUPPORTED");
+    assert (`SSTC_SUPPORTED == 0 || (`S_SUPPORTED)) else $error("SSTC requires S_SUPPORTED");
+    assert ((`ZMMUL_SUPPORTED == 0) || (`M_SUPPORTED ==0)) else $error("At most one of ZMMUL_SUPPORTED and M_SUPPORTED can be enabled");
   end
 
 endmodule