# Goroutine is not running nor in syscall, so use the info in goroutine
	if status != G_RUNNING and status != G_SYSCALL:
		return pc.cast(vp), sp.cast(vp)

	# If the goroutine is in a syscall, use syscallpc/sp.
	pc, sp = ptr['syscallpc'], ptr['syscallsp']
	if sp != 0:
		return pc.cast(vp), sp.cast(vp)
	# Otherwise, the goroutine is running, so it doesn't have
	# saved scheduler state. Find G's OS thread.
	m = ptr['m']

      /*effective_hidden_scale_intermediate=*/mlir::TypeAttr());

  // Cast the static shaped lstm result to FuncOp's signature -
  // Ranked but unknown 2nd dimension to support stacking these.
  SmallVector<int64_t, 2> func_output_shape = {1, tensorflow::kTFDynamicSize};
  auto func_result_type = tensorflow::GetTypeFromTFTensorShape(
      func_output_shape,
      mlir::cast<RankedTensorType>(input_.getType()).getElementType());

  // CHECK: [[cast:%.*]] = tf_mlrt.executeop(
  // CHECK-SAME: ReadVariableOp
  %v2 = "tf.ReadVariableOp"(%arg3) {__op_key = 2: i32, device = "/CPU:0"} : (tensor<*x!tf_type.resource>) -> tensor<i32>
  // CHECK: [[cast:%.*]] = tf_mlrt.executeop.device
  // CHECK-SAME: Cast
  %0 = "tf.Cast"(%arg0) {__op_key = 3: i32, device = "/device:CPU:0"} : (tensor<i32>) -> tensor<f32>

// TFR function call defines the signatures of the TF ops.
//
namespace mlir {
namespace TFR {

namespace {

// This pattern is to rewrite the "tfr.call" op and the "tfr.cast" ops on the
// operands by a TF op with "tfr.cast" ops on the results. The result type of
// the new TF op is an unranked tensor with element type derived.
class RewriteTFRCallOp : public OpRewritePattern<CallOp> {

      // Expand/Squeeze op must come in pair.
      return rewriter.notifyMatchFailure(
          op, "ExpandDimsOp and SqueezeOp should come in pair");
    }
    expand_op = llvm::cast<TF::ExpandDimsOp>(producer_op);
    squeeze_op = llvm::cast<TF::SqueezeOp>(consumer_op);
    if (!expand_op.getResult().hasOneUse()) {
      return rewriter.notifyMatchFailure(
          expand_op, "result for current op has more than 1 use");
    }

// CHECK-DAG: %[[KEEP:.*]] = "tf.Const"() <{value = dense<false> : tensor<i1>}> : () -> tensor<i1>
// CHECK: %[[CAST:.*]] = "tf.Cast"(%arg2) <{Truncate = false}> : (tensor<*xi32>) -> tensor<*xf32>
// CHECK: %[[RET:.*]] = "tf.MapAndBatchDatasetV0"(%arg0, %[[BATCH]], %[[PARAL]], %[[KEEP]], %arg1, %[[CAST]])

    }

    private Compiler<SwiftCompileSpec> createCompiler() {
        NativePlatformInternal targetPlatform = Cast.cast(NativePlatformInternal.class, this.targetPlatform.get());
        NativeToolChainInternal toolChain = Cast.cast(NativeToolChainInternal.class, getToolChain().get());
        PlatformToolProvider toolProvider = toolChain.select(targetPlatform);

     */
    public boolean isClass() {
        return wrapper instanceof ClassTypeWrapper;
    }

    public Class<? super T> getRawClass() {
        return Cast.uncheckedCast(wrapper.getRawClass());
    }

    public Class<T> getConcreteClass() {
        return Cast.uncheckedCast(wrapper.getRawClass());
    }

    public boolean isRawClassOfParameterizedType() {

LogicalResult VerifyTensorTypesCompatible(Type t1, Type t2) {
  if (!mlir::isa<TensorType>(t1) || !mlir::isa<TensorType>(t2)) {
    return failure();
  }
  return verifyCompatibleShape(mlir::cast<TensorType>(t1),
                               mlir::cast<TensorType>(t2));
}

LogicalResult GlobalTensorOp::verify() {
  GlobalTensorOp global_tensor = *this;
  if (global_tensor.getValue()) {
    if (failed(VerifyTensorTypesCompatible(

  "SliceDenseIntElementsAttrColumn2D($0.cast<ElementsAttr>(), " # column # " )">;

class SliceDenseIntElementsAttr<string index, string axis> : NativeCodeCall<
  "SliceDenseIntElementsAttr($0.cast<ElementsAttr>(), " # index # ", " # axis # ")">;

// Interior padding attribute based on the TF padding.
def GetInteriorPadding : NativeCodeCall <
  "GetInteriorPadding($0.cast<ElementsAttr>())">;