[Mlir-commits] [mlir] [mlir][emitc] Add ArrayType (PR #83386)

[Matthias Gehre]

mgehre-amd wrote:

> This looks good now besides the two nits. It would be helpful though to see how this is going to be used with the subscript op for reading/writing elements. Especially as there are ongoing discussions around the distinction between SSA values and C variables [here](https://discourse.llvm.org/t/rfc-separate-variables-from-ssa-values-in-emitc/75224/5).

You can see the following PRs here:
https://github.com/Xilinx/llvm-project/pull/118: Add emitc.subscript
https://github.com/Xilinx/llvm-project/pull/113: Initial MemRefToEmitC pass (type conversions of memref to emitc.array))
https://github.com/Xilinx/llvm-project/pull/119: Convert memref.load/store to emitc.subscript
https://github.com/Xilinx/llvm-project/pull/120: Lower memref.alloca to emitc.variable

Regarding the discussion of SSA vs C variables, I believe the subscript op follows the current design of the emitc.variable. If we decide to change the design of emitc.variable after the ongoing discussions, it should be easy to modify the emitc.subscript op design, too. I don't think we need to block emitc.subscript on concluding that discussions.


https://github.com/llvm/llvm-project/pull/83386