ISE Bug?

[Florian Heinz]

Hello

I found some strange behaviour in ISE WebPack 8.2.03i when trying to
synthesize the following VHDL-Code for an XC9572 CPLD:


Port-modes: OUT: buffer, IN1/2: in

OUT <= '0' when IN1 = '1' else
'1' when IN2 = '1' else
OUT;

ISE generates the following warning:
Removing unused input(s) 'IN1'. The input(s) are unused

And indeed, Signal IN1 does not work.
The RTL/Technology-Schematic shows everything as expected, IN1 is
connected properly, but the fitter seems to "optimize" it away later
in the process.
I am aware that there are other, better ways to achieve the intended
behaviour, but I wonder if ISE is right with what it does here and if
so, why?

Thanks for listening,

Florian

 

[Ray Andraka]

Florian Heinz wrote:

Hello

I found some strange behaviour in ISE WebPack 8.2.03i when trying to
synthesize the following VHDL-Code for an XC9572 CPLD:


Port-modes: OUT: buffer, IN1/2: in

OUT <= '0' when IN1 = '1' else
'1' when IN2 = '1' else
OUT;

ISE generates the following warning:
Removing unused input(s) 'IN1'. The input(s) are unused

And indeed, Signal IN1 does not work.
The RTL/Technology-Schematic shows everything as expected, IN1 is
connected properly, but the fitter seems to "optimize" it away later
in the process.
I am aware that there are other, better ways to achieve the intended
behaviour, but I wonder if ISE is right with what it does here and if
so, why?

Thanks for listening,

Florian

You've got a combinatorial loop here involving the signal OUT. ISE is
probably getting messed up because of it (I'm not sure what you
intention is either). Also, I'm surprised it let you use OUT as a signal
name, since 'out' is a reserved word.

perhaps:

output_driver <= '0' when IN1='1' else '1' when IN2='1' else 'Z';

was more like what you intended? this drives the output to hi_Z when
neither IN1 nor IN2 are high. Or were you really looking to have a
circuit with memory here?

Combinatorial loops are poor form, and are prone to timing problems even
if you coax the tool into dealing with them correctly.

 

[sky]

On 2006-11-21, Ray Andraka <ray@andraka.com> wrote:

You've got a combinatorial loop here involving the signal OUT. ISE is
probably getting messed up because of it (I'm not sure what you
intention is either).

It's a Flipflop.

Also, I'm surprised it let you use OUT as a signal
name, since 'out' is a reserved word.

Right, read "OUT1".

perhaps:

output_driver <= '0' when IN1='1' else '1' when IN2='1' else 'Z';

was more like what you intended? this drives the output to hi_Z when
neither IN1 nor IN2 are high. Or were you really looking to have a
circuit with memory here?

Yes, for example this works as expected:

OUT1 <= IN1 nor OUT2;
OUT2 <= IN2 nor OUT1;

Combinatorial loops are poor form, and are prone to timing problems even
if you coax the tool into dealing with them correctly.

As I wrote, I am aware of these problems. But do you really think, it's
ok that ISE behaves like that?

Florian

 

[Ralf Hildebrandt]

sky schrieb:

[quote extended]

OUT <= '0' when IN1 = '1' else
'1' when IN2 = '1' else
OUT;

(I'm not sure what you
intention is either).

It's a Flipflop.

No, it is a latch, reset with IN1, set with IN2 and retaining its value
otherwise.

Why do you model it this may? It would be equivalent to do it this way
(note that I have changed the name OUT to OUT1):

OUT1 <= '0' when IN1 = '1' else
'1' when IN2 = '1';

So you don't need OUT1 to be an output signal of type buffer.


Seems to be an RS-latch without the forbidden state. Am I right?


Ralf