Multiplication implementation

flake.nix

@@ -43,7 +43,7 @@
 
           alu-synth = pkgs.runCommandCC "alu-synth" {} ''
             mkdir -p "$out"
-            find ${./src} -name '*.v' -exec ${yosys}/bin/yosys -Q -p "synth_ice40 -top topmost -json $out/synth.json -dsp" {} +
+            find ${./src} -name '*.v' -exec ${yosys}/bin/yosys -f ' -sv' -Q -p "synth_ice40 -top topmost -json $out/synth.json -dsp" {} +
           '';
 
           alu-synth-view = pkgs.writeScriptBin "alu-synth-view" ''

fpga-files/rot.pcf

@@ -1,4 +1,5 @@
 set_io clk 39
+set_io en 38
 set_io op[0] 40
 set_io op[1] 41
 set_io op[2] 42

simulation/test_alu.cpp

@@ -27,7 +27,7 @@ struct alu_testcase {
     uint8_t op;
 
     // Outputs
-    uint32_t O;
+    uint64_t O; // {O_hi, O_lo}
     std::optional<bool> overflow, zero;
 
     std::optional<unsigned int> max_cycles;
@@ -104,7 +104,8 @@ void test_op(Tester *tester, alu_testcase test) {
     std::string o_name("O == ");
     o_name.append(fmt_hex(test.O));
 
-    subtester.assert_eq(o_name, test.state->valu->O, test.O);
+    uint64_t O = (((uint64_t) test.state->valu->O_hi) << 32) | (uint64_t) test.state->valu->O_lo;
+    subtester.assert_eq(o_name, O, test.O);
 
     if (test.overflow.has_value()) {
         if (*test.overflow)

src/alu/alu.v

@@ -19,20 +19,30 @@ module alu(
   input [31:0] A,
   input [31:0] B,
   input [2:0] op,
-  output reg [31:0] O,
+  output wire [31:0] O_lo,
-  output reg Fflow,
+  output wire [31:0] O_hi, // only used for OP_MUL
-  output reg Fzero
+  output wire Fflow,
+  output wire Fzero
 );
 
 // Constants
+/* verilator lint_off UNUSEDPARAM */
 localparam OP_ADD = 3'b000;
 localparam OP_SUB = 3'b001;
+localparam OP_MUL = 3'b010;
 localparam OP_AND = 3'b100;
 localparam OP_OR  = 3'b101;
 localparam OP_XOR = 3'b110;
 localparam OP_NOT = 3'b111;
 
 localparam ST_IDLE = 0;
+localparam ST_MULTIPLY_START = 1;
+localparam ST_MULTIPLY_END = 32;
+
+localparam OUT_ADD = 0;
+localparam OUT_SUB = 1;
+localparam OUT_BW  = 2;
+localparam OUT_MUL = 3;
 
 // State
 
@@ -43,18 +53,29 @@ assign RDY = state == ST_IDLE;
 
 reg [31:0] bitwise_out;
 
+// Multiplication
+
+reg [63:0] mult_out;
+reg [31:0] factorA, factorB; // factorA is static, factorB is shifted each cycle
+
 // Outputs
 
-assign O =
+assign O_lo =
-  (selected_out == OP_ADD || selected_out == OP_SUB) ? adder_out :
+  (selected_out == OUT_ADD || selected_out == OUT_SUB) ? adder_out :
+  selected_out == OUT_MUL ? mult_out[31:0] :
   bitwise_out;
 
-assign Fflow =
+assign O_hi =
-  selected_out == OP_ADD ? adder_carry_out :
+  selected_out == OUT_MUL ? mult_out[63:32] :
-  selected_out == OP_SUB ? ~adder_carry_out :
   0;
 
-assign Fzero = ~(| O);
+assign Fflow =
+  selected_out == OUT_ADD ? adder_carry_out :
+  selected_out == OUT_SUB ? ~adder_carry_out :
+  selected_out == OUT_MUL ? 0 :
+  0;
+
+assign Fzero = ~((| O_lo) | (| O_hi));
 
 // Modules
 
@@ -67,44 +88,68 @@ carry_select_adder adder(adder_A, adder_B, adder_carry_in, adder_out, adder_carr
 // Clocking
 
 always @(posedge CLK) begin
-  case (state)
+  reg [63:0] mult_out_tmp;
-    ST_IDLE: begin
+
-      if (EN) begin
+  if (state == ST_IDLE && EN) begin
     case (op)
       OP_ADD: begin
         adder_A <= A;
         adder_B <= B;
         adder_carry_in <= 0;
-            selected_out <= OP_ADD;
+        selected_out <= OUT_ADD;
       end
       OP_SUB: begin
         adder_A <= A;
         adder_B <= ~B;
         adder_carry_in <= 1;
-            selected_out <= OP_SUB;
+        selected_out <= OUT_SUB;
+      end
+      OP_MUL: begin
+        factorA <= A;
+        factorB <= B;
+        mult_out <= 0;
+        adder_A <= 0;
+        adder_B <= A;
+        adder_carry_in <= 0;
+        state <= ST_MULTIPLY_START;
+        selected_out <= OUT_MUL;
       end
       OP_AND: begin
         bitwise_out <= A & B;
-            selected_out <= OP_AND;
+        selected_out <= OUT_BW;
       end
       OP_OR: begin
         bitwise_out <= A | B;
-            selected_out <= OP_OR;
+        selected_out <= OUT_BW;
       end
       OP_XOR: begin
         bitwise_out <= A ^ B;
-            selected_out <= OP_XOR;
+        selected_out <= OUT_BW;
       end
       OP_NOT: begin
         bitwise_out <= ~A;
-            selected_out <= OP_NOT;
+        selected_out <= OUT_BW;
       end
       default: begin end // TODO: this should be $stop in verilator, no-op in synthesis
     endcase
   end
+
+  if (state >= ST_MULTIPLY_START && state <= ST_MULTIPLY_END) begin
+    mult_out_tmp = {
+      factorB[0] ? {adder_carry_out, adder_out} : {1'b0, mult_out[63:32]},
+      mult_out[31:1]
+    };
+    mult_out <= mult_out_tmp;
+
+    factorB <= {1'b0, factorB[31:1]};
+    adder_A <= {mult_out_tmp[63:32]};
+    adder_B <= factorA;
+    if (state < ST_MULTIPLY_END) begin
+      state <= state + 1;
+    end else begin
+      state <= ST_IDLE;
+    end
   end
-    default: $stop;
-  endcase
 end
 
 endmodule

src/topmost.v

@@ -1,11 +1,11 @@
 // Dummy module for connecting ALU and similar things, without having to break all inputs and outputs into separate pads
-module topmost(input clk, input [2:0] op, output xor_reduce);
+module topmost(input clk, input en, input [2:0] op, output xor_reduce);
 
 reg [31:0] A;
 reg [31:0] B;
-wire [31:0] O;
+wire [63:0] O;
 
-alu alu(.A(A), .B(B), .op(op), .O(O), .Fflow(), .Fzero());
+alu alu(.CLK(clk), .EN(en), .A(A), .B(B), .op(op), .O_lo(O[31:0]), .O_hi(O[63:32]), .Fflow(), .Fzero());
 
 always @(posedge clk) begin
   A <= A + 1;