skyjuiceOccasional ContributorJoined 7 years ago29 Posts1 LikeLikes received2 SolutionsView All Badges
ContributionsMost RecentMost LikesSolutionsRe: Is Agilex Synthesis is Timing Driven ? If you go to Quartus Settings --> Compiler Settings you will find a description of what each optimization mode does. The superior mode make changes to the synthesis stage to target better performance. Hope that answers your questions. Re: Registers with ERTM naming This is added by the early retimer, which happens at the end of the plan stage. If there is no further suffix after it then it is is an ALM reg, and not a hyper-register. Re: Very high interconnect delay The design has such a high interconnect delay is due to the huge clock skew. In order to meet hold time requirement, the router will need to add additional routing delays. Why huge clock skew you might ask? That's because this is a cross-clock transfer. Re: Compiling the s10_ref (20.2) on Quartus Prime Software Pro Do you mean this design is downloaded from Intel Design Store? Can you try to compile this design in the same Quartus version the design is developed in (i.e 20.2)? Re: How to reduce IC delay? Useful resources I saw that there are only 2 logic levels which are not huge. The IC delay is likely due to the source and destination nodes being placed far away. As a start, are there a lot of fanouts from the source nodes or the combination logic? Or could there be some sort of nodes beyond this path that is pulling them in opposite direction (if you know what I mean) Re: Module removal by compilation optimization Which Quartus version are you using? The latest 21.3 has a new report called 'Hierarchies Optimized Away During Sweep" that includes the details on what is optimized away alongside with the reasons. Re: Where is the document of pin function? The Pin Connection Guideline has the description on every pin in the device. For Stratix IV: https://www.intel.com/content/dam/www/programmable/us/en/pdfs/literature/dp/stratix4/pcg-01005.pdf Re: Change clock propagation through multiplexers Think of it this way - I have 2 inputs to a mux (clkA and clkB), and you're right only one clock will be selected. Let's assume for a specific scenario, clkA is selected and somewhere in the the downstream logic, it will need to perform a transfer to clkB and this will need to be timing-analyzed. If the tool where to cut all clkA <--> clkB paths, then the mentioned path will be affected, which is not what we want. Re: Controlling skew between clocks Thanks for the info. As expected, these are cross-clock transfers causing the huge clock skew. I suspect that the clock buffer for one of those clock networks must be located far away. You can see this through the timing report in Timing Analyzer (compare between Clock Arrival and Clock Required) and then locate them in Chip Planner for visualization. Can you try putting them into the clock networks whose the buffers are close to each other? Re: Controlling skew between clocks Can you give more context? Is this a common clock transfer? If yes, make sure you use global clock network and the skew will be minimal