[mlir][Conversion] Allow lowering to fixed arrays of scalable vectors

This allows lowering vector types like: vector<3x[4]> or vector<3x2x[4]>
to LLVM IR, i.e. vectors where the trailing dim is scalable.

This is contingent on:
https://discourse.llvm.org/t/rfc-enable-arrays-of-scalable-vector-types/72935

More tests will be added in later patches, however, some MLIR fixes are
needed first.

Depends on: D158517

Reviewed By: awarzynski

Differential Revision: https://reviews.llvm.org/D158752

Author: MacDue

mlir/lib/Conversion/LLVMCommon/TypeConverter.cpp

@@ -507,7 +507,8 @@ FailureOr<Type> LLVMTypeConverter::convertVectorType(VectorType type) const {
                                     type.getScalableDims().back());
   assert(LLVM::isCompatibleVectorType(vectorType) &&
          "expected vector type compatible with the LLVM dialect");
-  if (type.isScalable() && (type.getRank() > 1))
+  // Only the trailing dimension can be scalable.
+  if (llvm::is_contained(type.getScalableDims().drop_back(), true))
     return failure();
   auto shape = type.getShape();
   for (int i = shape.size() - 2; i >= 0; --i)

mlir/test/Conversion/VectorToLLVM/vector-to-llvm.mlir

@@ -2260,3 +2260,13 @@ func.func @vector_scalable_extract(%vec: vector<[4]xf32>) -> vector<8xf32> {
   %0 = vector.scalable.extract %vec[0] : vector<8xf32> from vector<[4]xf32>
   return %0 : vector<8xf32>
 }
+
+// -----
+
+// CHECK-LABEL: @make_fixed_vector_of_scalable_vector
+func.func @make_fixed_vector_of_scalable_vector(%f : f64) -> vector<3x[2]xf64>
+{
+  // CHECK: %{{.*}} = llvm.mlir.undef : !llvm.array<3 x vector<[2]xf64>>
+  %res = vector.broadcast %f : f64 to vector<3x[2]xf64>
+  return %res : vector<3x[2]xf64>
+}