What means the symbol and the principle of creating index_maps for iree_linalg_ext.CustomOp ?

[shoveller86]

Like in this example, it shows two linalg.matmul + one linalg.generic are fused into one iree_linalg_ext.custom_op.
Q1: I'm curious about how the indexing_maps and iterator_types are created ?
Q2: what means the symbols here (like s0, s1), why need them ?

    %0:2 = iree_linalg_ext.custom_op {
        indexing_maps = [affine_map<(d0, d1)[s0, s1] -> (d0, s0)>,
                         affine_map<(d0, d1)[s0, s1] -> (s0, s1)>,
                         affine_map<(d0, d1)[s0, s1] -> (s1, d1)>,
                         affine_map<(d0, d1)[s0, s1] -> (d0, s1)>,
                         affine_map<(d0, d1)[s0, s1] -> (d0, d1)],
        iterator_types = ["parallel", "parallel"]}
        ins(%lhs1, %rhs1, %rhs2
            : tensor<1000000x?xf32>, tensor<?x?xf32>, tensor<?x?xf32>)
        outs(%outs1, %outs2 : tensor<1000000x?xf32>, tensor<1000000x?xf32>) {
      ^bb0(%t0 : tensor<?x?xf32>, %t1 : tensor<?x?xf32>, %t2 : tensor<?x?xf32>,
           %t3 : tensor<?x?xf32>, %t4 : tensor<?x?xf32>) :
        %0 = linalg.matmul ins(%t0, %t1 : tensor<?x?xf32>, tensor<?x?xf32>)
            outs(%t3 : tensor<?x?xf32>) -> tensor<?x?xf32>
        %1 = linalg.matmul ins(%0, %t2 : tensor<?x?xf32>, tensor<?x?xf32>)
            outs(%t4 : tensor<?x?xf32>) -> tensor<?x?xf32>
        iree_linalg_ext.yield %0, %1 : tensor<?x?xf32>, tensor<?x?xf32>
    } -> tensor<1000000x?xf32>, tensor<x?xf32>