Re: udp scheduling semantics

From: Gordon Vreugdenhil (gvreugde@synopsys.com)
Date: Mon Sep 15 2003 - 16:45:08 PDT

  • Next message: Michael McNamara: "Re: udp scheduling semantics"

    Precedence: bulk

    Steven Sharp wrote:
    >
    [...]
    >
    > I have appended a test which uses a ripple-counter-like structure, where
    > the output of one sequential UDP triggers the clock of the next UDP. It
    > triggers the clock, schedules an NBA, waits for the NBA to complete, then
    > prints the output of the chain of UDPs. The output will be 1 instead of x
    > if and only if five UDP delays are still faster than one NBA delay.

    VCS also produces a "1" on this example (with arbitrary long UDP chains).

    > My original version used arrays of instances to allow an arbitrarily
    > specified number of UDP delays. Since some simulators might still not be
    > supporting those, I changed it to just use 5 hand-instantiated UDP delays
    > in series.

    You could use V2K generate.... :-)

    module top;
      parameter p = 20;

      reg rclk, dclk;
      wire clk = rclk;
      wire q[p:0];

      generate
        genvar i;
        u_dff ff0(q[0], 1'b1, clk);
        for (i = 1; i <= p; i = i + 1) begin:chain
          u_dff ff1(q[i], 1'b1, q[i-1]);
        end
      endgenerate

      initial begin
            #1 rclk = 1;
            dclk <= 1;
            @(dclk) $display("%b", q[p]);
      end
    endmodule

    Gord 

    -- 
----------------------------------------------------------------------
Gord Vreugdenhil                                 gvreugde@synopsys.com
Staff Engineer, VCS (Verification Tech. Group)   (503) 547-6054
Synopsys Inc., Beaverton OR

    This archive was generated by hypermail 2.1.4 : Mon Sep 15 2003 - 15:56:45 PDT and
    sponsored by Boyd Technology, Inc.