Cactus

What is a good source of reset signal?

Recommended Posts

I'd like to use CLaSH (http://www.clash-lang.org/) with a Papilio Pro. By and large, it works; however, CLaSH requires an asynchronous, active-high reset spike to initialize registers. This is because CLaSH generates assignments on RESET only, instead of initializers; here's an example VHDL generated from CLaSH:

-- Automatically generated VHDL-93
library IEEE;
use IEEE.STD_LOGIC_1164.ALL;
use IEEE.NUMERIC_STD.ALL;
use IEEE.MATH_REAL.ALL;
use std.textio.all;
use work.all;
use work.blinkertop_types.all;

entity blinkerTop is
  port(-- clock
       CLK_32MHZ : in std_logic;
       -- asynchronous reset: active high
       RESET     : in std_logic;
       LED       : out std_logic);
end;

architecture structural of blinkerTop is
  signal \#tup_app_arg\       : unsigned(31 downto 0);
  signal \s'\                 : boolean;
  signal \#s'_case_alt\       : boolean;
  signal s                    : boolean;
  signal \#finished_case_alt\ : boolean;
  signal \#k'_case_alt\       : unsigned(31 downto 0);
  signal ds                   : blinkertop_types.tup2;
  signal \#finished_app_arg\  : signed(63 downto 0);
  signal x                    : unsigned(63 downto 0);
  signal x_0                  : blinkertop_types.tup2;
  signal \x#\                 : unsigned(63 downto 0);
  signal k                    : unsigned(31 downto 0);
  signal \#w\                 : unsigned(63 downto 0);
begin
  LED <= '1' when \s'\ else
         '0';

  \#tup_app_arg\ <= resize(to_unsigned(0,64),32) when \#finished_case_alt\ else
                    \#k'_case_alt\;

  \s'\ <= \#s'_case_alt\ when \#finished_case_alt\ else
          s;

  \#s'_case_alt\ <= false when s else
                    true;

  s <= ds.tup2_sel0;

  \#finished_case_alt\ <= tagToEnum(\#finished_app_arg\);

  \#w\ <= (\x#\ + to_unsigned(1,64));

  \#k'_case_alt\ <= resize((resize(\#w\(31 downto 0),64)),32);

  -- register begin 
  blinkertop_register : process(CLK_32MHZ,RESET)
  begin
    if RESET = '1' then
      ds <= ( tup2_sel0 => false, tup2_sel1 => resize(to_unsigned(0,64),32) )
      -- pragma translate_off
      after 1 ps
      -- pragma translate_on
      ;
    elsif rising_edge(CLK_32MHZ) then
      ds <= x_0
      -- pragma translate_off
      after 1 ps
      -- pragma translate_on
      ;
    end if;
  end process;
  -- register end

  \#finished_app_arg\ <= to_signed(1,64) when x = to_unsigned(32000000,64) else to_signed(0,64);

  x <= resize(\#k'_case_alt\,64);

  x_0 <= ( tup2_sel0 => \s'\
         , tup2_sel1 => \#tup_app_arg\ );

  \x#\ <= resize(k,64);

  k <= ds.tup2_sel1;
end;

(Note how `ds` is not initialized but set in the `blinkertop_register` process when RESET is high)

Of course, for a simple circuilt like above, where the initialization is for 0 anyway, just setting RESET to always low works; however, any slightly more complicated circuit will need register initialization to non-0 values as well. In these cases, I really need the spike.

Is there a pin on the Papilio Pro that I could use in my UCF file to get this reset spike? I was able to get it working, in more complicated circuits requiring non-0 initialization, by using a (negated) LogicStart joystick direction, but this approach has two problems:

  • I am getting a warning from the Xilinx tools that the joystick input shouldn't be used for reset or anything clock-like:
WARNING:Place:1109 - A clock IOB / BUFGMUX clock component pair have been found
   that are not placed at an optimal clock IOB / BUFGMUX site pair. The clock
   IOB component <RESET> is placed at site <P57>. The corresponding BUFG
   component <RESET_IBUF_BUFG> is placed at site <BUFGMUX_X3Y13>. There is only
   a select set of IOBs that can use the fast path to the Clocker buffer, and
   they are not being used. You may want to analyze why this problem exists and
   correct it. This is normally an ERROR but the CLOCK_DEDICATED_ROUTE
   constraint was applied on COMP.PIN <RESET.PAD> allowing your design to
   continue. This constraint disables all clock placer rules related to the
   specified COMP.PIN. The use of this override is highly discouraged as it may
   lead to very poor timing results. It is recommended that this error condition
   be corrected in the design.
  • I would like my circuit to start in an initialized state instead of requiring me to press a button

Share this post


Link to post
Share on other sites

Can you insert a small amount of VHDL (or Verilog) code into your Clash design, and use its output as a RESET, instead of taking RESET from an external pin?  Using initializers it'd fairly easy to generate a 1 for a few clock cycles after the FPGA is initialized, and 0 afterwards.  It can also be combined with a button so you'd have both ways of causing a reset.

It seems like something that would have already been done, and made available as a Clash primitive, by somebody.

As I think about it, you might be able to do it just in Clash, with two resets:

  • the "outer reset" is either always 0, or is driven by a button, and is only used by the reset logic
  • the "reset logic" has a few bits of state that's initially zero, and filled in with ones
  • the "inner reset" is driven by the reset logic, and is the NAND of the reset logic's state

Not sure of all the details, as I've never used Clash.

 

Share this post


Link to post
Share on other sites

Additional tip: The output of the reset logic should be synchronized by a register.  That's to avoid glitches which might be troublesome with an async reset.

 

Share this post


Link to post
Share on other sites
On 8/6/2018 at 3:41 AM, Jaxartes said:

Can you insert a small amount of VHDL (or Verilog) code into your Clash design, and use its output as a RESET, instead of taking RESET from an external pin?  Using initializers it'd fairly easy to generate a 1 for a few clock cycles after the FPGA is initialized, and 0 afterwards.

 

Thank you! I have now done just that, and it works. Wrapping the whole design into a toplevel VHDL module is a bit of a bummer (since I now have to manually "forward" all real IO from the real Clash code to the IO of this wrapper), but I think there should be a way around that by somehow getting Clash to automagically instantiate a VHDL part and thereby avoid this inversion of structure. 

Quote

 It can also be combined with a button so you'd have both ways of causing a reset.

Wouldn't that still have that problem of too much fanout, like in the original Xilinx warning?

Share this post


Link to post
Share on other sites
18 hours ago, Cactus said:

Wouldn't that still have that problem of too much fanout, like in the original Xilinx warning?

Not if you pass the result of all your reset logic through a register before using it as a reset -- which I recommended above (in my 2nd post) for other reasons.  Then it shouldn't make any difference, to that warning, what is controlling the register.

I think the cause of the warning is not with using the pin for something related to reset, but only with directly driving reset from it.  The Xilinx tools will try to figure out how long signals take to propagate from where they originate to wherever they are needed.  There can be a problem if they take too long, or if the time is too hard to predict.  The register improves both.

I don't see anything in the warning about fanout, either.  I've read about reset fanout being a problem, but haven't encountered it.  But then, my designs have been fairly small, have not demanded the utmost of performance, and have used synchronous rather than asynchronous reset.

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now