Added support for ZMMUL

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/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/wally/wallypipelinedcore.sv

@@ -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");
+    $display("IDIV_ON_FPU = %b M_SUPPORTED %b comb %b\n", `IDIV_ON_FPU, `M_SUPPORTED, ((`IDIV_ON_FPU) || (!`M_SUPPORTED)));
-    assert (`S_SUPPORTED | `VIRTMEM_SUPPORTED == 0) else $error("Virtual memory requires S mode support");
+    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 (`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 (`S_SUPPORTED || `VIRTMEM_SUPPORTED == 0) else $error("Virtual memory requires S mode support");
-    assert (`F_SUPPORTED | ~`D_SUPPORTED) else $error("Can't support double fp (D) without supporting float (F)");
+    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 (`D_SUPPORTED | ~`Q_SUPPORTED) else $error("Can't support quad fp (Q) without supporting double (D)");
+    assert (`F_SUPPORTED || ~`D_SUPPORTED) else $error("Can't support double fp (D) without supporting float (F)");
-    assert (`F_SUPPORTED | ~`ZFH_SUPPORTED) else $error("Can't support half-precision fp (ZFH) without supporting float (F)");
+    assert (`D_SUPPORTED || ~`Q_SUPPORTED) else $error("Can't support quad fp (Q) without supporting double (D)");
-    assert (`DCACHE_SUPPORTED | ~`F_SUPPORTED | `FLEN <= `XLEN) else $error("Data cache required to support FLEN > XLEN because AHB bus width is XLEN");
+    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 (`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_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 >= 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_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 >= 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_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(`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_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(`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(`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(`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 || (`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 (`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 (`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 (`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 (`DCACHE_SUPPORTED || `VIRTMEM_SUPPORTED ==0) else $error("Virtual memory needs dcache");
-    assert (`ICACHE_SUPPORTED | `VIRTMEM_SUPPORTED ==0) else $error("Virtual memory needs icache");
+    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_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 <= `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 (`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 (`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 (`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