[llvm] [AMDGPU][True16][MC] duplicate vop1 tests to fake16 and update real-true16 flags for GFX12 (PR #100849)

[Joe Nash via llvm-commits]

https://github.com/Sisyph commented:

We probably want a comment in each of the non-fake16 test files that the intention here is for the 'fake16 syntax' specified tests to be converted to 'real16 syntax' as the number of supported instructions progresses. 
There are a few of these already, such as v_ceil_f16, and I think it would be a logical patch to come soon to convert these already supported instructions.
Besides such a comment, the patch LGTM. Please wait for someone else to approve as well.

I don't believe we need to test the 'default' behavior of the assembler/disassembler with respect to the +real-true16 flag. So the changes here to have one file with +real-true16 and one with -real-true16 are a sufficient set. 

https://github.com/llvm/llvm-project/pull/100849