[Mlir-commits] [mlir] [MLIR] Graceful handling of unint sparse tensor encodings with `sparse-tensor-codegen` (PR #181145)

[Aart Bik]

================
@@ -0,0 +1,20 @@
+// RUN: mlir-opt %s -sparse-tensor-codegen -verify-diagnostics
+
+#SparseVector = #sparse_tensor.encoding<{
+  map = (d0) -> (d0 : compressed)
+}>
+
+module {
+  func.func @main() -> tensor<8xf32, #SparseVector> {
+    %dense = arith.constant dense<[1.0, 0.0, 0.0, 2.0, 0.0, 3.0, 0.0, 0.0]>
+      : tensor<8xf32>
+
+    // expected-error at +1 {{failed to legalize operation 'sparse_tensor.convert' that was explicitly marked illegal}}
+    %sparse = sparse_tensor.convert %dense
----------------
aartbik wrote:

The construct is valid under the right passes. Even though you say that in the comment, it would be more clear if you  have to RUN command for this file, one that lowers correct (--stage-sparse-ops --lower-sparse-ops-to-foreach or something like that) and one that generates this error when sparse codegen is called too soon

https://github.com/llvm/llvm-project/pull/181145