The Linux kernel is totally huge, several megabytes of compiled code.
So at a first glance you need a huge NISC chip or FPGA, assuming it's even possible because NISC maybe likes linear sequences of instructions. Something that can be branching often and all over the place might be difficult. Doing interrupts could be nigh impossible.
But let's imagine you've trimmed linux down (there's a lot of unneeded networking, file systems, drivers and other garbage), you can actually build it with the toolchain and you have a FPGA big enough (biggest FPGA in the world?).
It would only be useful to run other programs, so you have to bake them into that giant NISC spaghetti mess along with linux.
NISC seems to be made for running just one program, with no OS.
See also limitations
http://www.ics.uci.edu/~nisc/documents/FAQ.html#t2
Running any kind of OS would be a cool hack already. You would build something simple from scratch.
If you wanted linux you would build a real CPU in your FPGA, with stuff from opencores maybe, and run linux on that, you would then have NISC applications as coprocessors on the chip.
My two euro cents, from a limited and quick and dirty understanding.