linalg_transpose_2d_generic.mlir

#map = affine_map<(d0, d1) -> (d0, d1)>
func.func @linalg_transpose_2d(%arg0: tensor<?x?xf32>, %permutation: tensor<2xindex>, %1 : tensor<?x?xf32>) -> tensor<?x?xf32> {
  %2 = linalg.generic {indexing_maps = [#map, #map], iterator_types = ["parallel", "parallel"]} ins(%arg0 : tensor<?x?xf32>) outs(%1 : tensor<?x?xf32>) {
  ^bb0(%in: f32, %out: f32):
    %idx0 = linalg.index 0 : index
    %idx1 = linalg.index 1 : index
    %indexes = tensor.from_elements  %idx0, %idx1 : tensor<2xindex>
    %cst0 = arith.constant 0 : index
    %cst1 = arith.constant 1 : index
    %permutationIdx0 = tensor.extract %permutation[%cst0] : tensor<2xindex>
    %permutationIdx1 = tensor.extract %permutation[%cst1] : tensor<2xindex>
    %outputIdx0 = tensor.extract %indexes[%permutationIdx0] : tensor<2xindex> 
    %outputIdx1 = tensor.extract %indexes[%permutationIdx1] : tensor<2xindex> 
    %3 = tensor.extract %arg0[%outputIdx0, %outputIdx1] : tensor<?x?xf32>
    linalg.yield %3 : f32 
  } -> tensor<?x?xf32>
  return %2 : tensor<?x?xf32>
}