Back to PBT run

input.mlir

dataflow.thread private @thread_0 domain(#dataflow.thread_domain<dense>)(
    %limit: index, %memory: memref<?xindex>) ctrl (%ctrl: none) {
  "loom.spatial_region"(%limit, %memory)
      <{operandSegmentSizes = array<i32: 1, 0, 1, 0>,
        resultSegmentSizes = array<i32: 0, 0>}> ({
    ^bb0(%lim: index, %target: memref<?xindex>):
      %zero = arith.constant 0 : index
      %one = arith.constant 1 : index
      %two = arith.constant 2 : index
      %flag = arith.cmpi ult, %zero, %two : index
      %wr0 = scf.while (%wa0 = %zero) : (index) -> index {
        %wc0 = arith.cmpi ult, %wa0, %two : index
        scf.condition(%wc0) %wa0 : index
      } do {
      ^bb0(%wb0: index):
      %wr1 = scf.while (%wa1 = %zero) : (index) -> index {
        %wc1 = arith.cmpi ult, %wa1, %two : index
        scf.condition(%wc1) %wa1 : index
      } do {
      ^bb0(%wb1: index):
      memref.store %one, %target[%zero] : memref<?xindex>
      %ld3 = memref.load %target[%zero] : memref<?xindex>
      %ad3 = arith.addi %ld3, %one : index
      memref.store %ad3, %target[%zero] : memref<?xindex>
        %wn1 = arith.addi %wb1, %one : index
        scf.yield %wn1 : index
      }
      %wr4 = scf.while (%wa4 = %zero) : (index) -> index {
        %wc4 = arith.cmpi ult, %wa4, %two : index
        scf.condition(%wc4) %wa4 : index
      } do {
      ^bb0(%wb4: index):
      memref.store %one, %target[%zero] : memref<?xindex>
      %ld6 = memref.load %target[%zero] : memref<?xindex>
      %ad6 = arith.addi %ld6, %one : index
      memref.store %ad6, %target[%zero] : memref<?xindex>
        %wn4 = arith.addi %wb4, %one : index
        scf.yield %wn4 : index
      }
        %wn0 = arith.addi %wb0, %one : index
        scf.yield %wn0 : index
      }
      %wr7 = scf.while (%wa7 = %zero) : (index) -> index {
        %wc7 = arith.cmpi ult, %wa7, %two : index
        scf.condition(%wc7) %wa7 : index
      } do {
      ^bb0(%wb7: index):
      scf.parallel (%iv8) = (%zero) to (%two) step (%one) {
        %mx8 = arith.muli %zero, %two : index
        %ix8 = arith.addi %mx8, %iv8 : index
      memref.store %one, %target[%ix8] : memref<?xindex>
        scf.reduce
      }
      memref.store %one, %target[%zero] : memref<?xindex>
        %wn7 = arith.addi %wb7, %one : index
        scf.yield %wn7 : index
      }
      "loom.spatial_yield"()
          <{operandSegmentSizes = array<i32: 0, 0>}> : () -> ()
  }) {graph_name = "g_thread_0", source_maps = []} :
      (index, memref<?xindex>) -> ()
  dataflow.thread.yield
}
dataflow.thread private @thread_1 domain(#dataflow.thread_domain<dense>)(
    %limit: index, %memory: memref<?xindex>) ctrl (%ctrl: none) {
  "loom.spatial_region"(%limit, %memory)
      <{operandSegmentSizes = array<i32: 1, 0, 1, 0>,
        resultSegmentSizes = array<i32: 0, 0>}> ({
    ^bb0(%lim: index, %target: memref<?xindex>):
      %zero = arith.constant 0 : index
      %one = arith.constant 1 : index
      %two = arith.constant 2 : index
      %flag = arith.cmpi ult, %zero, %two : index
      scf.forall (%iv11) in (2) {
        %mx11 = arith.muli %zero, %two : index
        %ix11 = arith.addi %mx11, %iv11 : index
      scf.for %iv12 = %zero to %two step %one {
      memref.store %one, %target[%ix11] : memref<?xindex>
      }
      }
      scf.parallel (%iv14) = (%zero) to (%two) step (%one) {
        %mx14 = arith.muli %zero, %two : index
        %ix14 = arith.addi %mx14, %iv14 : index
      memref.store %one, %target[%ix14] : memref<?xindex>
      memref.store %one, %target[%ix14] : memref<?xindex>
        scf.reduce
      }
      "loom.spatial_yield"()
          <{operandSegmentSizes = array<i32: 0, 0>}> : () -> ()
  }) {graph_name = "g_thread_1", source_maps = []} :
      (index, memref<?xindex>) -> ()
  dataflow.thread.yield
}