erichkeane wrote:

So I've done everything you've asked except for the 'add more attribute tests', 
including the tablegen changes.

I ALSO did a 'merge' off of main, which brought in 3 way comparisons (and 
something OMP related?), which found a few different problems!
1- The generic `matchAndRewrite` function was giving up if the operands/region 
was all 'legal', and not considering the attributes.  We do now.

2- The new 3-way-Attr type needed the TableGen changes I made above, so that 
was added.

3- For some reason the OMP Dialect attributes were now trying to be 
transformed!  I don't know what changed to make that so (or if I just didn't 
test 'enough' after my large cleanup before submitting?), but I added it the 
same as ACC: Just consider them legal.  They will fail the verifier if anyone 
ends up getting an illegal type in them.

Next Steps:
Tomorrow is my last day before WG21.  I intend to make an effort to get all of 
the 'transformed' attributes tested in some way, though some might just not be 
possible (as I see at least 1 doesn't have the ability to have a 'record' type 
in it, despite having a type!).  I'll likely have to figure out how to get 
`cir-opt` to let me run just the cxx-abi-lowering in order to be able to 
hand-write IR for these though.

Even if I DO manage to get it done/accepted, I likely won't merge this until I 
return from WG21 (about a week and a half from now), so I anticipate I'll have 
some cleanups at one point to get that fixed up too.

https://github.com/llvm/llvm-project/pull/187327
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to