SML / MLton: naive implementation for BLAS-based matrix multiplication

mlton-blas-naive.sml

  local
      val call = _import "cblas_dgemm" public: int * int * int * int * int * int * real * real Vector.vector * int * real Vector.vector * int * real * real Array.array * int -> unit;
      datatype cblasTranspose = NoTrans | Trans | ConjTrans | ConjNoTrans
      fun cblasOrder Array2.RowMajor = 101
    | cblasOrder Array2.ColMajor = 102
      fun cblasTranspose NoTrans = 111
    | cblasTranspose Trans = 112
    | cblasTranspose ConjTrans = 113
    | cblasTranspose ConjNoTrans = 114
      fun getVector a =
      let open Array2
          val (m,n) = dimensions a
          val a'= Array.array(m*n, 0.0)
          val _ = appi RowMajor (fn (i,j,aij) => Array.update(a',i+m*j,aij)) {base=a, row=0, col=0, nrows=NONE, ncols=NONE}
      in
          Array.vector a'
      end
      fun makeArray(a,m,n) =
      Array2.tabulate Array2.RowMajor (m, n, fn (i,j) => Array.sub(a,i+m*j))
  in
  fun matmul2(a, b) =
      let
      open Array2
      val ((m,k), (k',n)) = (dimensions a, dimensions b)
      val () = if k <> k' then raise General.Size else ()
      val arrayc = Array.array(m*n,0.0)
      val _ = call(cblasOrder ColMajor, cblasTranspose NoTrans, cblasTranspose NoTrans, m, n, k, 1.0, getVector a, m, getVector b, k, 0.0, arrayc, m)
      in
      makeArray (arrayc, m, n)
      end
  end;