Occasional Contributor
Super Contributor"The lower levels have many LL regions and partitions of course. "
If you have LL regions in the lower levels, that's going to force the placement of them when you try to instantiate more than once a partition that contains them (in the same spot, as you say).
You can have separate LL regions in the top level, but you don't want to have, essentially, child regions forcing the placement of that lower-level logic. As you discovered, not having those lower level regions did allow the partitions to be placed.
It might be worthwhile to try creating the top level with the source code itself, instantiated multiple times, instead of trying to reuse the design block (.qdb), and not using any LL regions to give the Fitter the most flexibility for placement. You would need to compile the entire project from scratch, but there's a better chance everything will fit.
Exported synthesis partitions are a bit more restrictive now than they were back when you used the old incremental compilation feature in 14.0. Try a flat compile without reuse and see how it works.
#iwork4intel
Occasional ContributorIf you export the output of the synthesis step, then you shouldn't have the locations. But your right that apparently there is a bug that the placement is still exported in the syth step. When are you fixing this bug?
How are you supposed to use the lower partitions with multiple people? It sounds like your saying that the pro version doesn't support smaller projects as part of a larger project. If I just combine everything into one big flat project, then only one person can work on this and it is really a pain to write and debug.
I am trying to make a flat version, but there is some problem that it removes 95% of my design when it starts to place. It is all there in synthesis, but gone in place and route. There may have been some unseen error in copying the IP from all of the smaller projects to the big one. Probably 30+ individual IPs that I just copied into the flat project.
It is really disappointing that after years of being light years ahead of Xilinx, you have backtracked to be no better than quartus 9. That was a horrible release full of bugs and LL and partitions didn't work at all. Now, somehow, Vivado is better than quartus in partitions and reuse. The 2-4x improvement in processing time Quartus used to have over Vivado is now gone. Quartus crashes all the time now if there is anything at all complex. Very disappointing.......
