Forcing loop iterations to execute sequentially

I am afraid I do not know of any pragmas that could force serial execution. Altera has many undocumented pargmas, they could also have one for this, but rarely anybody outside of Altera knows about them.

If you managed to create a simple example that reproduces the problem, it would be interesting to take a look at it; there could be a bug in the compiler after all.

P.S. Try to test your code with the latest version of Quartus (17.1 was released yesterday), even if your board does not have a BSP for it, to see how the optimization report changes. They might have fixed the issue already.