[Mlir-commits] [mlir] [mlir][ArmSME] Add rudimentary support for tile spills to the stack (PR #76086)

[Cullen Rhodes]

================
@@ -40,6 +132,209 @@ IntegerAttr getTileIdOrError(arm_sme::ArmSMETileOpInterface op) {
   return tileId;
 }
 
+/// Creates a alloca matching the size of tile used by `tileOp`. The alloca is
+/// placed in the first block of the function.
+static memref::AllocaOp
+createAllocaForTile(RewriterBase &rewriter, Location loc,
+                    FunctionOpInterface func,
+                    arm_sme::ArmSMETileOpInterface tileOp) {
+  RewriterBase::InsertionGuard g(rewriter);
+  // Move to the first operation in the function.
+  rewriter.setInsertionPoint(&func.getBlocks().front().front());
+  // Create an alloca matching the tile size of the `tileOp`.
+  auto vscale = rewriter.create<vector::VectorScaleOp>(loc);
+  auto tileElementType =
+      llvm::cast<VectorType>(tileOp.getTileType()).getElementType();
----------------
c-rhodes wrote:

cast is unnecessary
```suggestion
  auto tileElementType = tileOp.getTileType().getElementType();
```

https://github.com/llvm/llvm-project/pull/76086