benchmarks/dgemm/dgemm_gendata.scala

   1 #!/usr/bin/env scala
   2 !#
   3
   4 val size = args(0).toInt
   5
   6 def print_matrix(name: String, rows: Int, cols: Int, data: Array[Double]) = {
   7   println("const double " + name + "[DATA_SIZE*DATA_SIZE] = {")
   8   for (i <- 0 until rows) {
   9     println(data.slice(cols*i, cols*(i+1)).mkString(", ") + (if (i < rows-1) ", " else ""))
  10   }
  11   println("};")
  12 }
  13 def rand_matrix(rows: Int, cols: Int) = {
  14   var r = new scala.util.Random
  15   var m = new Array[Double](rows*cols)
  16   for (i <- 0 until rows*cols)
  17     m(i) = r.nextInt(1000)
  18   m
  19 }
  20 def matmul(a: Array[Double], b: Array[Double], m: Int, n: Int, k: Int) = {
  21   var c = new Array[Double](m*n)
  22   for (i <- 0 until m)
  23     for (j <- 0 until n)
  24       for (l <- 0 until k)
  25         c(i*n+j) += a(i*n+l)*b(l*k+j)
  26   c
  27 }
  28
  29 println("#define DATA_SIZE " + size)
  30
  31 val a = rand_matrix(size, size)
  32 val b = rand_matrix(size, size)
  33 val c = matmul(a, b, size, size, size)
  34
  35 print_matrix("input1_data", size, size, a)
  36 print_matrix("input2_data", size, size, b)
  37 print_matrix("verify_data", size, size, c)