[Mlir-commits] [mlir] [mlir][vector] Implement lowering for 1D vector.deinterleave operations (PR #93042)

[Benjamin Maxwell]

================
@@ -0,0 +1,30 @@
+// DEFINE: %{entry_point} = entry
+// DEFINE: %{compile} = mlir-opt %s -test-lower-to-llvm
+// DEFINE: %{run} = %mcr_aarch64_cmd -march=aarch64 -mattr=+sve \
+// DEFINE:  -e %{entry_point} -entry-point-result=void \
+// DEFINE:  -shared-libs=%mlir_runner_utils,%mlir_c_runner_utils,%mlir_arm_runner_utils
+
+// RUN: %{compile} | %{run} | FileCheck %s
+
+func.func @entry() {
+  // Scalable vector length set to 256 which is equivalent to vscale 2.
+  // The test assumes that the vscale is 2. As such the SVL is set accordingly.
----------------
MacDue wrote:

Note:
SVL = streaming vector length (not scalable vector length). This is just setting the (non-streaming) vector length. 

```suggestion
  // Set the vector length to 256-bit (equivalent to vscale=2).
  // This allows the checks (below) to look at the entire vectors. 
```

https://github.com/llvm/llvm-project/pull/93042