Forum Discussion

Altera_Forum's avatar
Altera_Forum
Icon for Honored Contributor rankHonored Contributor
12 years ago

incomplete sensitivity list check in Quartus II

I wrote some VHDL code with incomplete sensitivity list as following and compile it with Quartus II. But, Quartus II didn't give any warnings on this.

Does anybody know how to make quatus ii to check incomplete sensitivity list and give warnings?

Also how to make Modelsim to report the same warnings? 

p_3: process (iCLK)
   begin
        if rising_edge(iCLK) then
            if mDVAL = '1' then
                if iX_Cont(0)='1' and iY_Cont(0)='1' then
                    rRed <= taps_1;
                    rGreen <= taps_1_d0;
                    rBlue <= taps_0_d0;
                elsif iX_Cont(0)='0' and iY_Cont(0)='1' then
                    rRed <= taps_1_d0;
                    rGreen <= taps_0_d0;
                    rBlue <= taps_0_d1;
                elsif iX_Cont(0)='1' and iY_Cont(0)='0' then
                    rRed <= taps_1_d0;
                    rGreen <= taps_0_d0;
                    rBlue <= taps_0_d1;
                elsif iX_Cont(0)='0' and iY_Cont(0)='0' then
                    rRed <= taps_1_d0;
                    rGreen <= taps_0_d0;
                    rBlue <= taps_0_d1;
                end if;        
            end if;
        end if;
    end process;
end trans;

8 Replies

  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    12 years ago

    In this code there is no incomplete sensitivity list because the general description of a register is:

    process(clk)

    begin

    if rising_edge(clk) = '1' then

    q_reg <= q_next;

    end if;

    end process;

    The process only activates on clk change ( when you change the input of flip-flop D, the output don't change until active clock edge). So you don't need q_next or equivalent expression in the sensitivity list.
  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    12 years ago

    bertulus, thanks for your answer, what's the principle for adding signals to sensitivity list? For the code below, is the sensitivity list complete? 

       process (vin_clk,rst_n,frame_flag_vin,vin_de,data_cnt)
   begin
      if rising_edge(vin_clk) then
         if (rst_n = '0') then
            data_cnt <= (others=>'0');
         elsif (frame_flag_vin = '1') then
            data_cnt <= (others=>'0');
         elsif (vin_de = '1') then
            if (data_cnt = vin_width - 1) then
               data_cnt <= (others=>'0');
            else
               data_cnt <= data_cnt + 1;
            end if;
         else
            data_cnt <= data_cnt;
         end if;
      end if;
   end process;
  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    12 years ago

    You have over specified on this process, as it is a synchronous process. You only need to add signals to the sensitivity list that can cause a change in the output values. For a synchronous process, values can only change when you get a clock edge - this causes the process to re-run.

    Any signal in a sensitivity list will cause that process to be re-evaluated on signal change in simulation. So adding too many signals can cause simulations to slow down as it keeps re-evaluating the process unnecessarily. On the flip side, for a non-synchronous process, adding too few signals can cause simulation behaviour to missmatch the real logic. For example: 

    
process(clk)
begin
  if clk = '1' then
    op <= ip1 and ip2;
  end if;
end process;

    In simulation, the waveform for this process would make it look like a clocked process. But the synthesis tool would create a transparent latch. This would cause missmatches.
  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    12 years ago

    Dear Tricky, Thanks for your answer. If I understand correctly, in the sensitivity list of my code, I can delete all signals except for clk. And make it look like: process (vin_clk). Am I correct?

  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    12 years ago

    Is the consequence of over-specifying sensitivity list only limited to making simulation very slow? Is there any other consequence like making my real logic uncorrect?

  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    12 years ago

    Over speccing them wont really slow simulation down too much - modelsim et al. now should be clever enough to work out whats really needed. But its still potential for extra simulation burdon.

    A sensitivity list is just for simulation - and it is just a means to tell the simulator when a process should be executed. So you're unlikely to have correct logic and incorrect simulation because of a sensitivity list - its probably the code inside the process.

    You could cause errors in a testbench. Say you wanted to write something to a file every time a certain signal changed. Putting extra signals in the sensitivity list would make it write out to the file too often, but this is a bit contrived. 

    
--Output when A changes
process(a,b)
begin
  report "a is " & to_string(a) severity note;
  --oops, this also reports when B changes too
end process;
  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    12 years ago

    Dear Tricky, Thanks for your reply which makes me understand deeply and clearly.