[MLIR][Linalg] mlir-linalg-ods-yaml-gen generates code relying on textual parsing

[joker-eph]

ArrayAttr Mmt4DOp::getIndexingMaps() {
  static const char memoizeAttr[] = "linalg.memoized_indexing_maps";
  ArrayAttr cached = getOperation()->getAttrOfType<ArrayAttr>(memoizeAttr);
  if (cached)
    return cached;

  MLIRContext *context = getContext();
  auto symbolBindings = getSymbolBindings(*this);
  SmallVector<AffineMap> maps;
  maps.push_back(llvm::cast<AffineMapAttr>(mlir::parseAttribute("affine_map<(d0, d1, d2, d3, d4, d5)[s0, s1, s2, s3, s4, s5] -> (d0, d2, d3, d5)>", context)).getValue());
  maps.back() = simplifyAffineMap(maps.back().replaceDimsAndSymbols({}, symbolBindings, 6, 0));
  maps.push_back(llvm::cast<AffineMapAttr>(mlir::parseAttribute("affine_map<(d0, d1, d2, d3, d4, d5)[s0, s1, s2, s3, s4, s5] -> (d1, d2, d4, d5)>", context)).getValue());
  maps.back() = simplifyAffineMap(maps.back().replaceDimsAndSymbols({}, symbolBindings, 6, 0));
  maps.push_back(llvm::cast<AffineMapAttr>(mlir::parseAttribute("affine_map<(d0, d1, d2, d3, d4, d5)[s0, s1, s2, s3, s4, s5] -> (d0, d1, d3, d4)>", context)).getValue());
  maps.back() = simplifyAffineMap(maps.back().replaceDimsAndSymbols({}, symbolBindings, 6, 0));
  cached = Builder(context).getAffineMapArrayAttr(maps);
  getOperation()->setAttr(memoizeAttr, cached);
  return cached;
}

The generated code for getIndexingMaps() relies on textual parsing to build affine maps: that is close to a layering violation to me: the C++ code involved outside of strict textual parsing phase should never rely on textual parsing for building IR entities.

[llvmbot]

@llvm/issue-subscribers-mlir-linalg

Author: Mehdi Amini (joker-eph)

``` ArrayAttr Mmt4DOp::getIndexingMaps() { static const char memoizeAttr[] = "linalg.memoized_indexing_maps"; ArrayAttr cached = getOperation()->getAttrOfType<ArrayAttr>(memoizeAttr); if (cached) return cached;

MLIRContext *context = getContext();
auto symbolBindings = getSymbolBindings(*this);
SmallVector<AffineMap> maps;
maps.push_back(llvm::cast<AffineMapAttr>(mlir::parseAttribute("affine_map<(d0, d1, d2, d3, d4, d5)[s0, s1, s2, s3, s4, s5] -> (d0, d2, d3, d5)>", context)).getValue());
maps.back() = simplifyAffineMap(maps.back().replaceDimsAndSymbols({}, symbolBindings, 6, 0));
maps.push_back(llvm::cast<AffineMapAttr>(mlir::parseAttribute("affine_map<(d0, d1, d2, d3, d4, d5)[s0, s1, s2, s3, s4, s5] -> (d1, d2, d4, d5)>", context)).getValue());
maps.back() = simplifyAffineMap(maps.back().replaceDimsAndSymbols({}, symbolBindings, 6, 0));
maps.push_back(llvm::cast<AffineMapAttr>(mlir::parseAttribute("affine_map<(d0, d1, d2, d3, d4, d5)[s0, s1, s2, s3, s4, s5] -> (d0, d1, d3, d4)>", context)).getValue());
maps.back() = simplifyAffineMap(maps.back().replaceDimsAndSymbols({}, symbolBindings, 6, 0));
cached = Builder(context).getAffineMapArrayAttr(maps);
getOperation()->setAttr(memoizeAttr, cached);
return cached;
}

The generated code for `getIndexingMaps()` relies on textual parsing to build affine maps: that is close to a layering violation to me: the C++ code involved outside of strict textual parsing phase should never rely on textual parsing for building IR entities. 

</details>

[llvmbot]

@llvm/issue-subscribers-mlir

Author: Mehdi Amini (joker-eph)

``` ArrayAttr Mmt4DOp::getIndexingMaps() { static const char memoizeAttr[] = "linalg.memoized_indexing_maps"; ArrayAttr cached = getOperation()->getAttrOfType<ArrayAttr>(memoizeAttr); if (cached) return cached;

MLIRContext *context = getContext();
auto symbolBindings = getSymbolBindings(*this);
SmallVector<AffineMap> maps;
maps.push_back(llvm::cast<AffineMapAttr>(mlir::parseAttribute("affine_map<(d0, d1, d2, d3, d4, d5)[s0, s1, s2, s3, s4, s5] -> (d0, d2, d3, d5)>", context)).getValue());
maps.back() = simplifyAffineMap(maps.back().replaceDimsAndSymbols({}, symbolBindings, 6, 0));
maps.push_back(llvm::cast<AffineMapAttr>(mlir::parseAttribute("affine_map<(d0, d1, d2, d3, d4, d5)[s0, s1, s2, s3, s4, s5] -> (d1, d2, d4, d5)>", context)).getValue());
maps.back() = simplifyAffineMap(maps.back().replaceDimsAndSymbols({}, symbolBindings, 6, 0));
maps.push_back(llvm::cast<AffineMapAttr>(mlir::parseAttribute("affine_map<(d0, d1, d2, d3, d4, d5)[s0, s1, s2, s3, s4, s5] -> (d0, d1, d3, d4)>", context)).getValue());
maps.back() = simplifyAffineMap(maps.back().replaceDimsAndSymbols({}, symbolBindings, 6, 0));
cached = Builder(context).getAffineMapArrayAttr(maps);
getOperation()->setAttr(memoizeAttr, cached);
return cached;
}

The generated code for `getIndexingMaps()` relies on textual parsing to build affine maps: that is close to a layering violation to me: the C++ code involved outside of strict textual parsing phase should never rely on textual parsing for building IR entities. 

</details>