[Mlir-commits] [mlir] [mlir][sparse] use a consistent order between [dis]assembleOp and sto… (PR #84079)

[Aart Bik]

================
@@ -738,12 +738,6 @@ class SparseTensorDisassembleConverter
     auto stt = getSparseTensorType(op.getTensor());
     SmallVector<Value> retVal;
     SmallVector<Value> retLen;
-    // Get the values buffer first.
-    auto vals = genValuesCall(rewriter, loc, stt, adaptor.getTensor());
-    auto valLenTp = op.getValLen().getType();
-    auto valLen = linalg::createOrFoldDimOp(rewriter, loc, vals, 0);
-    retVal.push_back(vals);
-    retLen.push_back(genScalarToTensor(rewriter, loc, valLen, valLenTp));
     // Then get the positions and coordinates buffers.
----------------
aartbik wrote:

Remove the "Then" now that we don't have the first Get

https://github.com/llvm/llvm-project/pull/84079