[Mlir-commits] [mlir] [mlir][arith.constant]Fix element type of the dense attributes in target attributes to be consistent with result type in LLVM::detail::oneToOneRewrite() (PR #149787)

[Mengmeng Sun]

MengmSun wrote:

I have updated my local changes to this branch. 
However, I met some problems locally with `make check-mlir` which confused me(these failures will be exposed in this CI as well): 

7 UT failed, and I have attached my local log below

[check.log](https://github.com/user-attachments/files/21503509/check.log)


```bash
********************
Testing:  0.. 10.. 20.. 30.. 40.. 50.. 60.. 70.. 80.. 90..
********************
Failed Tests (7):
  MLIR :: Conversion/FuncToLLVM/func-to-llvm.mlir
  MLIR :: Conversion/GPUCommon/lower-launch-func-to-gpu-runtime-calls.mlir
  MLIR :: Conversion/GPUToNVVM/wmma-ops-to-nvvm.mlir
  MLIR :: Conversion/GPUToROCDL/gpu-to-rocdl.mlir
  MLIR :: Conversion/MemRefToLLVM/expand-then-convert-to-llvm.mlir
  MLIR :: Dialect/LLVM/lower-to-llvm-e2e-with-target-tag.mlir
  MLIR :: Dialect/LLVM/lower-to-llvm-e2e-with-top-level-named-sequence.mlir


Testing Time: 57.32s

Total Discovered Tests: 3332
  Unsupported      :  458 (13.75%)
  Passed           : 2866 (86.01%)
  Expectedly Failed:    1 (0.03%)
  Failed           :    7 (0.21%)

```

These all about the conversion between index->i64. I'm confused because it looks like many places intend to treat `index` as `i32`. Is it intended? Should we treat `index` specifically(i.e. not convert `index` element type to `i64`?) 

https://github.com/llvm/llvm-project/pull/149787