Odin inserts undriven signals in netlist

[GoogleCodeExporter]

For the following example, Odin inserts undriven signals (named  
top^BUF_NODE~...) in a netlist.  This appears to have something to do with the 
black box instantiation.

Here is the source Verilog:

  module test (a, x, y);
  input  a;
  output x;
  output y;

  wire t1;
  wire t2;

  split split_a (.a(a), .x(t1), .y(t2));  // Blackbox specified in arch file.

  assign x = t1;
  assign y = t2;

  endmodule

And here is the BLIF that Oden generates:

  .model test
  .inputs top^a
  .outputs top^x top^y

  .names gnd
  .names unconn
  .names vcc
  1

  .names top^BUF_NODE~1 top^x
  1 1

  .names top^BUF_NODE~2 top^y
  1 1

  .end

  .model split
  .inputs a
  .outputs y x
  .blackbox
  .end

Note that top^BUF_NODE~1/2 signals are not defined.

Any help to resolve this bug would be greatly appreciated.  This is currently 
holding up our project.

Thanks!

-Tom

Original issue reported on code.google.com by tomahawkins on 11 Mar 2015 at 12:23

[GoogleCodeExporter]

It looks like your hardware primitive "split" is not being instantiated.  Can 
you attach your architecture file that contains the hardware primitive "split" 
so that we can see why this is happening?

Ken, I think we should error out in Odin II when a user defines a blackbox that 
we cannot instantiate.  I hypothesize that Odin II recognized that "split" is a 
black box but couldn't find "split" in the architecture file so ignored 
instantiating it.

Original comment by JasonKai...@gmail.com on 11 Mar 2015 at 8:49

[GoogleCodeExporter]

I have also observed this issue, and I think it might be related to the width 
of the output. The attached example is a simple memory connected to a module's 
I/O.
I'm using blackboxes for memories instead of vtr memories.
When I use a width of 2 (or more), it works fine and the output of the memory 
is attached to the output of the module. However, when the width drops to 1, 
Odin inserts a BUF_NODE to drive the output, and doesn't use the output of the 
memory.

Thank you for your help!

Original comment by edin.kad...@gmail.com on 31 May 2015 at 1:33

Attachments:

• buf_node.tar.gz