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.

Xilinx synthesis problem

Status
Not open for further replies.

Zerox100

Full Member level 6
Full Member level 6
Joined
Mar 1, 2003
Messages
328
Helped
21
Reputation
42
Reaction score
10
Trophy points
1,298
Activity points
2,606
Dear my friends,

I have a created a soc project using vivado 2018. I have added an accelerator as an IP to main cpu of soc.
My problem is that when I synthesis the accelerator as a separate project (mode outofcontext) it uses 57000 LUT in synthesis. But when I add it to my design as IP it uses 65000 LUT in synthesis that is more than FPGA resources.
I have analyzed the synthesis details. Each part takes 10-15% more resources. Anybody knows any solution to my problem?

THX
 

synthesis report on resource usage is preliminary. It is finalised after placement/routing (implementation).
The tool tries to fit and pass timing on chosen fpga and as such may require more resource. You should check you are not over constraining design timing. There are also settings targeting resource optimisation at ip level or project level.

I have tried out-of-context Xilinx technology and as far as I noticed it is synthesis only. Not final placement/routing and I don't see it as New technology as Xilinx claims. It may just shorten tool processing time.
 

10 -15 % looks like normal range to me. There are many possible reasons why the IP takes more resources when implemented in the full design, e.g. datapath muxes, lack of dedicated multipliers, register duplication to keep timing constraints.
 

10 -15 % looks like normal range to me. There are many possible reasons why the IP takes more resources when implemented in the full design, e.g. datapath muxes, lack of dedicated multipliers, register duplication to keep timing constraints.
Thank you for your answer
Is there a way reduces the 10-15% to 5-7%?
 

Status
Not open for further replies.

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top