The following signals are missing in the process sensitivity

[antonio bergnoli]

hi,
using xilinx xst i noticed that the message:

"The following signals are missing in the process sensitivity list:...."

and following a list of signals readed inside a process; but I _don't_
want to trigger the process with these signals, i just want to build
synchronous machine. Of course this is not a problem but this warning is
somewhat 'strange'. At least for me.

 

[Zara]

On 16 Dec 2005 11:30:16 +0100, antonio bergnoli <bergnoli@pd.infn.it>
wrote:

hi,
using xilinx xst i noticed that the message:

"The following signals are missing in the process sensitivity list:...."

and following a list of signals readed inside a process; but I _don't_
want to trigger the process with these signals, i just want to build
synchronous machine. Of course this is not a problem but this warning is
somewhat 'strange'. At least for me.

Without context, it is difficult to answer.

But let´s give it a try:

if you write

process(clock)
begin
if rising_edge(clock) then
output<=input;
end if;
end process;

everything is OK.

If you write

process(clock)
begin
output<=input;
end process;

then you will have the style of error/warning stated by the subject.

Putting a signal in the sensitivity list is information for the
compiler, but makes no default opeartion by itself

Hope this helps,

Zara

 

[Ralf Hildebrandt]

antonio bergnoli wrote:

"The following signals are missing in the process sensitivity list:...."

-> You modelled a latch or combinational logic. If this was your
intention, follow this hint and complete the sensitivity list. Otherwise
simulation and synthesis may differ.

but I _don't_
want to trigger the process with these signals, i just want to build
synchronous machine.

Then you did a mistake and the process is not synchronous. Use the
template for a flipflop:

process(reset,clock)
begin
if (reset='1') then
-- do some reset
elsif rising_edge(clock) then
-- do something synchronous
end if;
end process;

Of course this is not a problem but this warning is
somewhat 'strange'.

No, its not strange - its important! If you want to model a flipflop,
but get a latch, there is something seriously wrong.

Ralf

 

[Jean-Christophe Le Lann]

"The following signals are missing in the process sensitivity list:...."

To design a synchronous machine, you need to clearly distinguish between
combinational logic and registers. See
http://mikro.e-technik.uni-ulm.de/vhdl/anl-engl.syn/html/node6.htm for
instance.

By default, every signal entering the combinational logic should be listed
in the sensitivity list. That is essentially because of VHDL simulation
semantics : the kernel reactivates the evaluation of the processes content
each "time" ( it may be delta cycles) these signal change their value. If
you do not respect this, you can have mismatches between simulation and
synthesis.

I clearly understand that this is not so natural : as you said, you do
know that your machine will be synchronous, and having to be so precise
for signal changes is somehow odd in this context : but it is not if you
consider the wide spectrum of things you can do with VHDL, that may be
asynchronous.

For you information, there are other approaches to synchronous design,
much more clean, such as dedicated "synchronous languages" (Esterel,
Lustre, Signal), and in some extend "statecharts".

Hope this helps.

JCLL

 

[Weng Tianxiang]

Hi,
A tip for missing a few signals in sensitivity list in a combinational
process:
1. If you don't do simulation, everything is OK. After compilation, the
code can be used in FPGA or ACIS without error.

2. If you are doing simulation, the signal doesn't change in your
combinational process (like next state process when 2 processes are
used for a state machine), and it will cause you a lot of time
debugging why the equation in the process is wrong (because you miss a
signal in the sensitivity list).

In other words, sensitivity list is only used for simulation software,
not for final running code.

The sensitivity list is really an unnecessary burden for any serious
VHDL programmers. It was created in 1980' when related software and
technology were in an infant stage, and it will be used for next 9999
year. Every several months this subject arises somewhere by a newbie.

Weng

 

[Mike Treseler]

Jean-Christophe Le Lann wrote:

For you information, there are other approaches to synchronous design,
much more clean, such as dedicated "synchronous languages" (Esterel,
Lustre, Signal), and in some extend "statecharts".

It is possible to write very clean synchronous
VHDL code using a procedural, single-process
design style. For examples, see:

http://home.comcast.net/~mike_treseler/


-- Mike Treseler