Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

place and route in soce with/without latch

Status
Not open for further replies.

ludan

Newbie level 5
Joined
May 24, 2009
Messages
9
Helped
0
Reputation
0
Reaction score
0
Trophy points
1,281
Activity points
1,377
Hi guys,

I am facing the following problem since a while: blockA <--> blockB,
at the output of blockA there is a FF and a wire that goes as input to a latch
in blockB.
If I place these two blocks in soc encounter with a certain distance (5mm)
I see a certain delay (~3ns) on the link interconnecting the two blocks.

Now if I replace the latch in blockB with a FF and I place the blocks at the
same distance of 5mm, the delay I see on the interconnection link is 1.7ns

Looking at the timing report, it is clear that using on both sides the FFs,
the tool optimizes the link inferring buffers with a big drive strength.
In the example with 1ff and 1 latch, the buffers inferred are smaller and the
delay is much higher.

Now my question: is there any specific reason why the tool does not recognize
such a long link between A and B as the critical path in case I use a latch?
It is clearly the critical path (as shown by the experiment with 2 FFs), but
for some reason when I use a latch that path is not recognized as critical,
therefore the effort to optimize the link is low and the final interconnection
delay quite high (and disappointing) :-(

Ideas?
 

Do you have any setup violations with either approach? It is possible that the tool is using time borrowing when using the latch. Check the tool variables to see how to turn latch time borrowing on/off and run some more tests.
 

In both cases I have violations the only difference is how big they are:
- both FF: 1.7ns
- 1 FF and 1 latch: 3ns

As far as I can see from the report_timing in prime time:

Point Incr Path
...
..
time borrowed from endpoint 0.00 0.64
...
...

which means no time has been borrowed :-(
the other funny thing is that the tool does not seem to recognize this path as critical due to the fact that the sampling happens with the same clock sent from blockB. Indeed, this is something I forgot to say in the previous post: blockA send the clock to the latch/FF in blockB along with the data. Can be this the reason why this path is not seen as critical?

Cheers
 

Status
Not open for further replies.

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top