%0 = tfl.add %arg0, %arg1 {fused_activation_function = "RELU6", tac.device = "GPU"} : tensor<100xf32>
    func.return %0 : tensor<100xf32>
  }
  func.func @func_2_GPU_FLOAT(%arg0: tensor<100xf32>, %arg1: tensor<100xf32>) -> tensor<2x100xf32> attributes {tac.cost = 8.040000e+01 : f32, tac.device = "GPU", tac.inference_type = "FLOAT", tac.interface_name = "func_2"} {

// CHECK:           %[[VAL_3:.*]] = tfl.add %[[VAL_0]], %[[VAL_1]] {fused_activation_function = "RELU6", tac.device = "GPU", tac.inference_type = "FLOAT"} : tensor<1xf32>
// CHECK:           %[[VAL_4:.*]] = tfl.mul %[[VAL_3]], %[[VAL_2]] {fused_activation_function = "RELU6", tac.device = "GPU", tac.inference_type = "FLOAT"} : tensor<1xf32>

// CHECK:           %[[VAL_5:.*]] = tfl.add %[[VAL_0]], %[[VAL_1]] {fused_activation_function = "RELU6", tac.device = "GPU", tac.inference_type = "FLOAT"} : tensor<1xf32>
// CHECK:           %[[VAL_6:.*]] = tfl.mul %[[VAL_5]], %[[VAL_2]] {fused_activation_function = "RELU6", tac.device = "GPU", tac.inference_type = "FLOAT"} : tensor<1xf32>

// Test for partial folding: only fold i32 constants.
// RUN: tac-opt-all-backends -tfl-fold-constants-to-subgraph='fold-all-constants=false' %s -split-input-file -verify-diagnostics | FileCheck --check-prefix=PARTIAL %s

// Test for fold all constants.
// RUN: tac-opt-all-backends -tfl-fold-constants-to-subgraph='fold-all-constants=true' %s -split-input-file -verify-diagnostics | FileCheck --check-prefix=ALL %s

module {

tf_cc_binary(
    name = "tac-opt-all-backends",
    testonly = True,
    deps = [
        ":tac-opt_lib",
        "//tensorflow/compiler/mlir/lite/experimental/common:outline_operations",
        "//tensorflow/compiler/mlir/lite/experimental/tac/hardwares:all-target-hardwares",
    ],
)

cc_library(
    name = "tac-translate-lib",
    srcs = [
        "tac_translate.cc",
    ],
    deps = [

}

// CHECK: %[[CST:.*]] = arith.constant dense<1> : tensor<4xi32>
// CHECK:  [[VAL_0:%.*]] = "tfl.reshape"(%1, %[[CST]]) {tac.device = "GPU",  tac.inference_type = "FLOAT"} : (tensor<1xf32>, tensor<4xi32>) -> tensor<1x1x1x1xf32>
// CHECK:  [[VAL_1:%.*]] = "tfl.reshape"(%2, %[[CST]]) {tac.device = "GPU",  tac.inference_type = "FLOAT"} : (tensor<1xf32>, tensor<4xi32>) -> tensor<1x1x1x1xf32>

#include "tensorflow/compiler/mlir/lite/experimental/tac/common/targets.h"
#include "tensorflow/compiler/mlir/lite/experimental/tac/tac_filter.pb.h"
#include "tensorflow/compiler/mlir/lite/experimental/tac/transforms/passes.h"

namespace mlir {
namespace TFL {
namespace tac {
namespace {

using ::third_party::tensorflow::compiler::mlir::lite::experimental::tac::
    FunctionFilter;

#include "tensorflow/compiler/mlir/lite/experimental/tac/common/targets.h"
#include "tensorflow/compiler/mlir/lite/experimental/tac/common/utils.h"
#include "tensorflow/compiler/mlir/lite/experimental/tac/hardwares/target_hardware.h"
#include "tensorflow/compiler/mlir/lite/experimental/tac/transforms/device_transform_gpu.h"
#include "tensorflow/compiler/mlir/lite/ir/tfl_ops.h"

#include "tensorflow/compiler/mlir/lite/experimental/tac/common/cost.h"
#include "tensorflow/compiler/mlir/lite/experimental/tac/common/targets.h"
#include "tensorflow/compiler/mlir/lite/experimental/tac/common/utils.h"
#include "tensorflow/compiler/mlir/lite/experimental/tac/hardwares/target_hardware.h"
#include "tensorflow/compiler/mlir/lite/ir/tfl_ops.h"

namespace mlir {
namespace TFL {
namespace tac {
namespace {

  %1 = "tfl.mul"(%0, %arg2) {fused_activation_function = "RELU6", per_device_costs = {CPU = 5.0 : f32, GPU = 1.0 : f32}, tac.device = "GPU"} : (tensor<1xf32>, tensor<1xf32>) -> tensor<1xf32>