Programmable Devices
CPLDs, FPGAs, SoC FPGAs, Configuration, and Transceivers
20777 Discussions

SOPC Generates Failing Paths

Altera_Forum
Honored Contributor II
1,168 Views

Hi all, 

I generated 8 Master templates (read and write together) and one slave sdram component. When i compile the device gives the warning timing requirements not met.  

 

In the timingsadvisor i made alle the recommende settings except : 

 

- Duplicate logic for fan-out control 

- Use location asignments & back-annotation 

- Use fast input, fast output, and fast output enable registers 

 

What else can i do, to fit timing?
0 Kudos
7 Replies
Altera_Forum
Honored Contributor II
488 Views

I would put some pipeline bridges into those paths. Perhaps connect four of the masters to on bridge, the other four master to another bridge, then connect those two bridges to the SDRAM controller. If that's not enough then perhaps change the bridge topology slightly. Here is a document that talks about this in more detail: http://www.altera.com/literature/hb/nios2/edh_ed51007.pdf

0 Kudos
Altera_Forum
Honored Contributor II
488 Views

Thx, 

Reading it*
0 Kudos
Altera_Forum
Honored Contributor II
488 Views

If u have several masters, is there an arbiter added by the sopc builder? Or do you have to add one? And i'll try to add pipelined bridges, thx!

0 Kudos
Altera_Forum
Honored Contributor II
488 Views

The arbiters are automatically added by SOPC builder, when several masters are connected to the same slave.

0 Kudos
Altera_Forum
Honored Contributor II
488 Views

Then i have another question. as said above i use 8 Master Templates thas uses 1 Slave SDRAM controller. 

If all templates do an action read or write at the same time...  

Will the arbiter handle every action sequentially or will only one action be completed and the others lost? 

 

Or what components would u use, within the sopc? Is it enough with one  

Pipelined bridge ?
0 Kudos
Altera_Forum
Honored Contributor II
488 Views

No action will be lost. The mandatory waitrequest signal on the Avalon master interface will freeze 7 masters while the last one will perform its action. 

I think the arbitrator uses a more or less round robin system to handle the arbitration, and you can control how many shares each master has if you want some of them to have more frequent access than the others to the slave.
0 Kudos
Altera_Forum
Honored Contributor II
488 Views

Thx Daixiwen, 

 

I still know so less of FPGA but your help is usefull !
0 Kudos
Reply