New Contributor
New ContributorHi sstrell:
Thanks for your reply, but it's not the answer what I approach for.
I want to forbid Quartus Prime to use a particular resource, because I find some particular LAB delay is 10x larger than a normal LAB in Timing Analyzer. After I use chip planner for post-netlist ECO to change the LAB resource to a adjacent LAB, the problem is solved. So, I think the 10x delay LAB may be something wrong.
I appreciate it if you could kindly provide a solution for my request.
Super ContributorI don't understand how a particular LAB could be slower than any other. They are identical. Can you share your .sdc and the timing report where you're seeing this? There is some reason why the Fitter is selecting this LAB for you (super full design perhaps?) and causing the issue. The extra delay may be getting a signal to a particular physical location, not the physical resource itself. If that's the case, you need to figure out why the Fitter is doing this.
The ECO fixes it, but it might be useful going forward to understand the cause of the issue.
New ContributorHi Sstrell:
I don't understand that either, but it happens.
I provide the Chip Planner layout and Timing report here.
The utilization of my design is 27%, and I think it's not an issue for FPGA fitter.
BTW, if you know how to avoid using a specific LAB in Quartus Prime, please tell me how. Thank you.
Super ContributorAs I suspected, IC is interconnect delay where you're seeing the problem, not LAB (cell) delay. For some reason, the Fitter is using long routing wires which doesn't make sense as you say since they're in the same LAB.
When you run Report Timing in the Timing Analyzer, turn on the Show Routing option. That breaks down IC delay into RE (routing element) delay.
There must be some physical reason why the Fitter is doing this. Again, it would also be useful to see your timing constraints, at least what speed you're trying to run this at.
New ContributorHi Sstrell:
Thanks you.
Your suspicion is right, and the large IC delay is caused by lots of RE delay.
Though the data path is within the same LAB cell, Quartus II still deliver this long long routing path.
Do you have any idea why Quartus II does it?
My design is a simple image processor. The PLL clock in is 64MHz, and my design runs at 32MHz.
Utilization = 27% on Cyclone V-E FPGA.
