Prepacking problem

[GoogleCodeExporter]

What steps will reproduce the problem?
1.I've developed an architecture file that uses a hard logic instead of LUT.
2.I also mapped MCNC benchmarks on the library that implements this hard logic.
3.This architecture works properly for combinational benchmarks like "alu4", 
but when I run siquential benchmarks it results VPR to crash and show following 
error message:

"ERROR(1): Can not find any logic block that can implement molecule.
 ERROR(2):  Atom n_n2770"

What is the expected output? What do you see instead?

What version of the product are you using? On what operating system?
I use VPR 7.0

Please provide any additional information below.
I've used prepacking option to pack hard logic and FF inside the cluster, but 
it results VPR to core dumped.
The architecture blif file are attached.

Thanks

Original issue reported on code.google.com by 7thi...@gmail.com on 3 Jul 2014 at 10:28

Attachments:

• my_arch_HL.xml
• bigkey.aic3.blif

[GoogleCodeExporter]

There is a problem with the design of your architecture.  Lets assume that the 
input circuit is just inpad --> flip-flop --> outpad.  Your architecture, as 
specified, cannot implement this circuit because the flip-flops in your 
architecture cannot be reached without first going through a gate.  To fix this 
problem, you can do one of two things:

1.  Add a bypass mux so that the flip-flip can be reached without going through 
your cone.

or

2.  If your HL cone has the ability to implement an interconnect wire, then you 
must explicitly state that by creating a mode of operation that implements just 
a wire.

Original comment by JasonKai...@gmail.com on 3 Jul 2014 at 4:35

• Changed state: Accepted

[GoogleCodeExporter]

You are right.
I added a bypass mux and now it works.
I truly appreciate your help in resolving the problem.

Original comment by 7thi...@gmail.com on 3 Jul 2014 at 5:46

[GoogleCodeExporter]

Original comment by JasonKai...@gmail.com on 3 Jul 2014 at 5:50

• Changed state: Done