[Mlir-commits] [mlir] [mlir][test] Test conversion of TOSA to EmitC via LinAlg (PR #94640)

[Paul Kirth]

ilovepi wrote:

@mtrofin @simon-camp, sorry that this is about a week behind schedule. I had most of this done last week, but I was trying to figure out why some of the transforms weren't working as expected. I believe that using the `tosa-to-linalg.mlir` test as a basis may have been a mistake.

Once we settle on the precise input MLIR, alls that's left to do is to validate the generated header via CHECK lines.

@simon-camp perhaps you have an idea why the second RUN line is failing? These are all basically translations from your script, and there shouldn't be any material deviations from the original.

```
# .---command stderr------------
# | /usr/local/google/home/paulkirth/llvm-fork/build/tools/mlir/test/Conversion/TosaToEmitC/Output/tosa-to-linalg.mlir.tmp.model.linalg.mlir:19:1: error: redefinition of attribute alias id 'map'
# | #map = affine_map<(d0) -> (d0)>
# | ^
# `-----------------------------
# error: command failed with exit status: 1
```


https://github.com/llvm/llvm-project/pull/94640