If Verilog is not portable across vendors I want to know it at the start, not a year later when an OEM chooses a different chip for their new design.
Given that the ARM cores themselves are written, tested and implemented in Verilog (and SystemVerilog), it can be very portable.
You *really* should focus on one FPGA and supporting it well. If this is your 1st go on a high end FPGA, you really need to think about your development strategy. Trying to target several FPGA platforms is going to lead to disappointment. Just look at other FOSS FPGA platforms (Red Pitaya). they have a hard enough time supporting *one* platform.
While it is true you can write Verilog that can synthesize under different toolchains, there are lots of features in the different vendors parts that cannot be inferred with generic HDL.
My experience over many FPGA projects (especially on the high end) is that purely generic HDL can be portable but is also not optimal. I completed FPGA project which was doing imaging processing for a space application. We were able to get a 5x improvement in processing efficiency by *thinking* about how the algorithm would map to available resources and directly instantiating hardware in the FPGA. Generic HDL can certainly get things to work but falls apart quickly when you are trying to push the speeds and density.
I'm starting to develop FOSS Verilog for DSOs and it would be *very* helpful to have a 3rd platform as it is much easier to get the vendor to take ownership of a bug in their development tools if you can tell them that the code works on two other systems.
There are only 2 players in the high end space: Xilinx and Altera. They are not going to care about your FOSS project and fixing bugs. Until you get to 6 or 7 figures in a purchase order, you have no leverage. That is just a reality.