Started working on the pipelined Montgomery modular multiplier. Currently can

do the "square" part of the multiplication, i.e. compute the twice larger
intermediate product AB = A * B.

1 files changed, 384 insertions, 0 deletions

diff --git a/rtl/dev/temp.txt b/rtl/dev/temp.txt
new file mode 100644
index 0000000..987bd86
--- /dev/null
+++ b/rtl/dev/temp.txt
@@ -0,0 +1,384 @@
+    //
+    // Helper Functions
+    //
+    /*
+    function  [INDEX_WIDTH-1:0] calc_preset_a_index;
+        input [INDEX_WIDTH-4:0] col_in;
+        input integer           x_in;
+              integer           index_out;
+        begin
+            index_out = col_in * NUM_MULTS + x_in;
+            calc_preset_a_index = index_out[INDEX_WIDTH-1:0];
+        end
+    endfunction
+
+    function  [INDEX_WIDTH-1:0] calc_rotate_a_index;
+        input [INDEX_WIDTH-1:0] current_index_in;
+        input [INDEX_WIDTH-1:0] last_index_in;
+        begin
+            if (current_index_in > {INDEX_WIDTH{1'b0}})
+                calc_rotate_a_index = current_index_in - 1'b1;
+            else
+                calc_rotate_a_index = last_index_in;
+        end
+    endfunction
+    */
+
+ /*
+    //
+    // Narrow Counters
+    //
+    reg        [INDEX_WIDTH-1:0] din_addr_narrow_reg;
+    reg        [INDEX_WIDTH-1:0] din_addr_narrow_dly;
+    localparam [INDEX_WIDTH-1:0] din_addr_narrow_zero = {INDEX_WIDTH{1'b0}};
+    wire       [INDEX_WIDTH-1:0] din_addr_narrow_next = (din_addr_narrow_reg < index_last) ?
+        din_addr_narrow_reg + 1'b1 : din_addr_narrow_zero;
+    wire                         din_addr_narrow_done = din_addr_narrow_reg == index_last;
+    
+    assign din_addr_narrow = din_addr_narrow_reg;
+    
+    always @(posedge clk)
+        //
+        din_addr_narrow_dly <= din_addr_narrow_reg;
+    
+    always @(posedge clk)
+        //
+        case (fsm_state_next)
+            FSM_STATE_MULT_SQUARE_COL_0_TRIG:   din_addr_narrow_reg <= din_addr_narrow_zero;
+            FSM_STATE_MULT_SQUARE_COL_0_BUSY:   din_addr_narrow_reg <= din_addr_narrow_next;
+            FSM_STATE_MULT_SQUARE_COL_N_TRIG:   din_addr_narrow_reg <= din_addr_narrow_zero;
+            FSM_STATE_MULT_SQUARE_COL_N_BUSY:   din_addr_narrow_reg <= din_addr_narrow_next;
+        endcase
+    
+
+    //
+    // Helper Functions
+    //
+    function  [NUM_MULTS-1:0] calc_mac_clear_bitmask;
+        input [2:0] t;
+        begin
+            case (t)
+                3'd0: calc_mac_clear_bitmask = 8'b00000001;
+                3'd1: calc_mac_clear_bitmask = 8'b00000010;
+                3'd2: calc_mac_clear_bitmask = 8'b00000100;
+                3'd3: calc_mac_clear_bitmask = 8'b00001000;
+                3'd4: calc_mac_clear_bitmask = 8'b00010000;
+                3'd5: calc_mac_clear_bitmask = 8'b00100000;
+                3'd6: calc_mac_clear_bitmask = 8'b01000000;
+                3'd7: calc_mac_clear_bitmask = 8'b10000000;
+            endcase
+        end
+    endfunction
+    
+    function  [NUM_MULTS:0] calc_mac_clear_square;
+        input [INDEX_WIDTH-4:0] current_col_index;
+        input [INDEX_WIDTH-1:0] b_addr_prev;
+        begin
+            if (b_addr_prev[INDEX_WIDTH-1:3] == current_col_index)
+                calc_mac_clear_square = {1'b0, calc_mac_clear_bitmask(b_addr_prev[2:0])};
+            else
+                calc_mac_clear_square = {1'b0, {NUM_MULTS{1'b0}}};
+        end
+    endfunction
+        
+        
+    //
+    // Wide Counters
+    //
+    reg [INDEX_WIDTH-1:0] din_addr_wide_reg[0:NUM_MULTS-1];
+ 
+    integer xi;
+    always @(posedge clk)
+        //
+        for (xi=0; xi<NUM_MULTS; xi=xi+1)
+            //
+            case (fsm_state_next)
+                //
+                FSM_STATE_MULT_SQUARE_COL_0_TRIG: din_addr_wide_reg[xi] <= calc_preset_a_index(0, xi);
+                FSM_STATE_MULT_SQUARE_COL_N_TRIG: din_addr_wide_reg[xi] <= calc_preset_a_index(col_index + 1'b1, xi);
+                //
+                FSM_STATE_MULT_SQUARE_COL_0_BUSY,
+                FSM_STATE_MULT_SQUARE_COL_N_BUSY: din_addr_wide_reg[xi] <= calc_rotate_a_index(din_addr_wide_reg[xi], index_last);
+                //
+            endcase
+        
+ 
+    //
+    // Enables
+    //
+    reg                 din_ena_narrow_reg = 1'b0;
+    reg [NUM_MULTS-1:0] din_ena_wide_reg = {NUM_MULTS{1'b0}};
+    
+    assign din_ena_narrow = din_ena_narrow_reg;
+    assign din_ena_wide   = din_ena_wide_reg;
+    
+    always @(posedge clk or negedge rst_n)
+        //
+        if (rst_n == 1'b0) din_ena_narrow_reg <= 1'b0;
+        else case (fsm_state_next)
+            FSM_STATE_MULT_SQUARE_COL_0_TRIG,
+            FSM_STATE_MULT_SQUARE_COL_0_BUSY,
+            FSM_STATE_MULT_SQUARE_COL_N_TRIG,
+            FSM_STATE_MULT_SQUARE_COL_N_BUSY: din_ena_narrow_reg <= 1'b1;
+            default:                          din_ena_narrow_reg <= 1'b0;
+        endcase
+
+    always @(posedge clk or negedge rst_n)
+        //
+        if (rst_n == 1'b0) din_ena_wide_reg <= {NUM_MULTS{1'b0}};
+        else case (fsm_state_next)
+            FSM_STATE_MULT_SQUARE_COL_0_TRIG,
+            FSM_STATE_MULT_SQUARE_COL_0_BUSY,
+            FSM_STATE_MULT_SQUARE_COL_N_TRIG,
+            FSM_STATE_MULT_SQUARE_COL_N_BUSY: din_ena_wide_reg <= {NUM_MULTS{1'b1}};
+            default:                          din_ena_wide_reg <= {NUM_MULTS{1'b0}};
+        endcase
+        
+        
+    //
+    // Modes
+    //
+    reg [2-1:0] din_mode_wide_reg;
+    reg [2-1:0] din_mode_narrow_reg;
+    reg [2-1:0] dout_mode_wide_reg;
+    reg [2-1:0] dout_mode_narrow_reg;
+
+    assign din_mode_wide = din_mode_wide_reg;
+    assign din_mode_narrow = din_mode_narrow_reg;
+
+    always @(posedge clk)
+        //
+        case (fsm_state_next)
+            FSM_STATE_MULT_SQUARE_COL_0_TRIG,
+            FSM_STATE_MULT_SQUARE_COL_0_BUSY,
+            FSM_STATE_MULT_SQUARE_COL_N_TRIG,
+            FSM_STATE_MULT_SQUARE_COL_N_BUSY: din_mode_wide_reg <= MODEXPNG_MODE_A;
+            default:                          din_mode_wide_reg <= 2'bXX;
+        endcase
+
+    always @(posedge clk)
+        //
+        case (fsm_state_next)
+            FSM_STATE_MULT_SQUARE_COL_0_TRIG,
+            FSM_STATE_MULT_SQUARE_COL_0_BUSY,
+            FSM_STATE_MULT_SQUARE_COL_N_TRIG,
+            FSM_STATE_MULT_SQUARE_COL_N_BUSY: din_mode_narrow_reg <= MODEXPNG_MODE_B;
+            default:                          din_mode_narrow_reg <= 2'bXX;
+        endcase
+
+
+    //
+    // MAC Array
+    //
+    wire [MODEXPNG_WORD_WIDTH-1:0] mac_din_a[0:NUM_MULTS];
+    wire [MODEXPNG_WORD_WIDTH-1:0] mac_din_b;
+    reg  [          NUM_MULTS  :0] mac_ce;
+    reg  [          NUM_MULTS  :0] mac_clr;
+    wire [ MODEXPNG_MAC_WIDTH-1:0] mac_p[0:NUM_MULTS];
+    reg  [          NUM_MULTS  :0] mac_rdy_lsb;
+    reg  [          NUM_MULTS  :0] mac_rdy_lsb_dly[MODEXPNG_MAC_LATENCY-1:0];
+    
+    //reg  [          NUM_MULTS  :0] mac_ce_dly[MODEXPNG_MAC_LATENCY-1:0];
+    //wire [          NUM_MULTS  :0] mac_rdy;
+
+
+
+
+
+    assign mac_din_b = din_narrow;
+    
+    
+    genvar x;
+    generate for (x=0; x<=NUM_MULTS; x=x+1)
+        begin : gen_macs
+            //
+            //assign mac_rdy[x] = mac_ce_dly[MODEXPNG_MAC_LATENCY-1][x];
+            //
+            modexpng_mac mac_inst
+            (
+                .clk    (clk),
+                .ce     (mac_ce[x]),
+                .clr    (mac_clr[x]),
+                .a      (mac_din_a[x]),
+                .b      (mac_din_b),
+                .p      (mac_p[x])
+            );
+            //
+        end
+        //
+    endgenerate
+    
+    generate for (x=0; x<NUM_MULTS; x=x+1)
+        begin : gen_mac_din_a
+            //
+            assign mac_din_a[x] = din_wide[x*MODEXPNG_WORD_WIDTH+:MODEXPNG_WORD_WIDTH];
+            //
+        end
+    endgenerate
+
+    generate for (x=0; x<NUM_MULTS; x=x+1)
+        begin : gen_din_addr_wide
+            //
+            assign din_addr_wide[x*INDEX_WIDTH+:INDEX_WIDTH] = din_addr_wide_reg[x];
+            //
+        end
+    endgenerate
+ 
+ 
+    //
+    // MAC Clock Enable Logic
+    //
+    always @(posedge clk or negedge rst_n)
+        //
+        if (rst_n == 1'b0) mac_ce <= {1'b0, {NUM_MULTS{1'b0}}};
+        else case (fsm_state)
+            FSM_STATE_MULT_SQUARE_COL_0_TRIG,
+            FSM_STATE_MULT_SQUARE_COL_0_BUSY,
+            FSM_STATE_MULT_SQUARE_COL_N_TRIG,
+            FSM_STATE_MULT_SQUARE_COL_N_BUSY: mac_ce <= {1'b0, {NUM_MULTS{1'b1}}};
+            default:                          mac_ce <= {1'b0, {NUM_MULTS{1'b0}}};
+        endcase
+    
+    
+    //
+    // MAC Valid Logic
+    //
+    integer y;
+
+    always @(posedge clk)
+        //
+        for (xi=0; xi<=NUM_MULTS; xi=xi+1) begin
+            mac_rdy_lsb_dly[0][xi] <= mac_rdy_lsb[xi];
+            for (y=1; y<MODEXPNG_MAC_LATENCY; y=y+1)
+                mac_rdy_lsb_dly[y][xi] <= mac_rdy_lsb_dly[y-1][xi];
+        end
+
+    always @(posedge clk) begin
+        //
+        fsm_state_dly[0] <= fsm_state;
+        for (y=1; y<=MODEXPNG_MAC_LATENCY; y=y+1)
+            fsm_state_dly[y] <= fsm_state_dly[y-1];
+    end
+
+    */
+
+    /*
+    always @(posedge clk)
+        //
+        for (xi=0; xi<=NUM_MULTS; xi=xi+1) begin
+            mac_ce_dly[0][xi] <= mac_ce[xi];
+            for (y=1; y<MODEXPNG_MAC_LATENCY; y=y+1)
+                mac_ce_dly[y][xi] <= mac_ce_dly[y-1][xi];
+        end
+    */
+    /*
+    always @(posedge clk)
+        //
+        for (xi=0; xi<=NUM_MULTS; xi=xi+1) begin
+            mac_clr_dly[0][xi] <= mac_clr[xi];
+            for (y=1; y<MODEXPNG_MAC_LATENCY; y=y+1)
+                mac_clr_dly[y][xi] <= mac_clr_dly[y-1][xi];
+        end
+    */
+
+  /*  
+    //
+    // MAC Clear Logic
+    //
+    always @(posedge clk)
+        //
+        case (fsm_state)
+            FSM_STATE_MULT_SQUARE_COL_0_TRIG,
+            FSM_STATE_MULT_SQUARE_COL_N_TRIG: mac_clr <= {1'b0, {NUM_MULTS{1'b1}}};
+            FSM_STATE_MULT_SQUARE_COL_0_BUSY,
+            FSM_STATE_MULT_SQUARE_COL_N_BUSY: mac_clr <= calc_mac_clear_square(col_index, din_addr_narrow_dly);
+            default:                          mac_clr <= {1'bX, {NUM_MULTS{1'bX}}};
+        endcase
+
+
+    //
+    // MAC Ready Logic
+    //
+    always @(posedge clk)
+        //
+        case (fsm_state)
+            FSM_STATE_MULT_SQUARE_COL_0_TRIG,
+            FSM_STATE_MULT_SQUARE_COL_N_TRIG,
+            FSM_STATE_MULT_SQUARE_COL_0_BUSY,
+            FSM_STATE_MULT_SQUARE_COL_N_BUSY: mac_rdy_lsb <= calc_mac_clear_square(col_index, din_addr_narrow);
+            default:                          mac_rdy_lsb <= {1'bX, {NUM_MULTS{1'bX}}};
+        endcase
+ 
+ 
+ 
+    //
+    // Recombinators
+    //
+    reg          rcmb_lsb_ce;
+    reg          rcmb_lsb_clr;
+    reg  [MODEXPNG_MAC_WIDTH-1: 0] rcmb_lsb_din;
+    wire [15: 0] rcmb_lsb_dout;
+    
+    modexpng_part_recombinator recomb_lsb
+    (
+        .clk    (clk),
+        .ce     (rcmb_lsb_ce),
+        .clr    (rcmb_lsb_clr),
+        .din    (rcmb_lsb_din),
+        .dout   (rcmb_lsb_dout)
+    );
+ 
+ 
+    reg calc_rcmb_lsb_ce;
+    always @*
+        //
+        calc_rcmb_lsb_ce = | mac_rdy_lsb_dly[MODEXPNG_MAC_LATENCY-1][NUM_MULTS-1:0];
+
+    reg [MODEXPNG_MAC_WIDTH-1:0] calc_rcmb_lsb_din;
+    
+    always @*
+        //
+        casez (mac_rdy_lsb_dly[MODEXPNG_MAC_LATENCY-1][NUM_MULTS-1:0])
+            8'b00000001: calc_rcmb_lsb_din = mac_p[0];
+            8'b00000010: calc_rcmb_lsb_din = mac_p[1];
+            8'b00000100: calc_rcmb_lsb_din = mac_p[2];
+            8'b00001000: calc_rcmb_lsb_din = mac_p[3];
+            8'b00010000: calc_rcmb_lsb_din = mac_p[4];
+            8'b00100000: calc_rcmb_lsb_din = mac_p[5];
+            8'b01000000: calc_rcmb_lsb_din = mac_p[6];
+            8'b10000000: calc_rcmb_lsb_din = mac_p[7];
+            default:     calc_rcmb_lsb_din = {MODEXPNG_MAC_WIDTH{1'bX}};
+        endcase
+ 
+    always @(posedge clk or negedge rst_n)
+        //
+        if (rst_n == 1'b0)
+            rcmb_lsb_ce <= 1'b0;
+        else case (fsm_state_dly[MODEXPNG_MAC_LATENCY])
+            FSM_STATE_MULT_SQUARE_COL_0_TRIG,
+            FSM_STATE_MULT_SQUARE_COL_N_TRIG,
+            FSM_STATE_MULT_SQUARE_COL_0_BUSY,
+            FSM_STATE_MULT_SQUARE_COL_N_BUSY: rcmb_lsb_ce <= calc_rcmb_lsb_ce;
+            default:                          rcmb_lsb_ce <= 1'b0;
+        endcase
+
+    always @(posedge clk)
+        //
+        case (fsm_state_dly[MODEXPNG_MAC_LATENCY])
+            FSM_STATE_MULT_SQUARE_COL_0_TRIG: rcmb_lsb_clr <= 1'b1;
+            default:                          rcmb_lsb_clr <= 1'b0;
+        endcase
+
+    always @(posedge clk)
+        //
+        case (fsm_state_dly[MODEXPNG_MAC_LATENCY])
+            FSM_STATE_MULT_SQUARE_COL_0_TRIG,
+            FSM_STATE_MULT_SQUARE_COL_N_TRIG,
+            FSM_STATE_MULT_SQUARE_COL_0_BUSY,
+            FSM_STATE_MULT_SQUARE_COL_N_BUSY: rcmb_lsb_din <= calc_rcmb_lsb_din;
+            default:                          rcmb_lsb_din <= {MODEXPNG_MAC_WIDTH{1'bX}};
+        endcase
+ 
+ 
+
+*/