[PATCH] D73893: [MLIR][GPU] Implement initial mapping from loop.parallel to gpu.launch.

[Mahesh Ravishankar via Phabricator via llvm-commits]

mravishankar resigned from this revision.
mravishankar added a comment.

Thanks Stephan! I see overall where this is headed. This is nice to have indeed.



================
Comment at: mlir/lib/Conversion/LoopsToGPU/LoopsToGPU.cpp:504
+std::tuple<unsigned, AffineMap, AffineMap>
+extractMapAndOperand(Attribute attribute) {
+  DictionaryAttr dict = attribute.dyn_cast<DictionaryAttr>();
----------------
herhut wrote:
> mravishankar wrote:
> > If this is really a dictionary_attribute, this lookup might be simplified using a StructAttr
> I had a struct attribute before but I did not want to specify the attribute on the `loop.parallel` operation, as it is GPU dialect specific. So I went with an unspecified optional attribute with custom accessor.
You can add it as an attribute to GPU dialect. There is no harm in adding this attribute "while" lowering to GPU dialect, or a pre-pass before lowering to GPU dialect right? 


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D73893/new/

https://reviews.llvm.org/D73893