The compiler documentation covers these options ...

--cla_support
--float_support
--idiv_support
--tmu_support
--vcu_support

But it never clearly explains that these option settings must match what is supported by the device that will execute the code.  In particular, the compiler has no method by which it can verify these settings are correct.

In practice, this is enforced in CCS.  When a new project is started in CCS, the name of the variant of the processor is an input.  Based on that choice, CCS automatically chooses many option settings, including all of the above options.  

These details ought to be made clear in the compiler manual.