==============================================================================*/

// This transformation pass prepares for legalization to the TFLite dialect by
// converting operations in TensorFlow dialect into operations that can be
// legalized to TensorFlow Lite dialect with simple replacements.  The newly
// created operations are in the TensorFlow dialect if the operation can be
// represented using a TensorFlow op.  Otherwise, TensorFlow Lite dialect op is

// This transformation pass prepares for legalization to the TFLite dialect by
// converting Tensorlist operations in TensorFlow dialect into operations that
// can be legalized to TensorFlow Lite dialect with simple replacements.  The
// newly created operations are in the TensorFlow dialect if the operation can
// be represented using a TensorFlow op. Otherwise, TensorFlow Lite dialect op
// is used.

   (HasOneUse $relu_out),
   (HasOneUse $mul_out),
   (HasOneUse $input_neg_out)]>;

// The constant folding in this pass might produce constant in the tf dialect.
// This rule is to legalize these constant to the tfl dialect.
def LegalizeConstOp : Pat<
  (TF_ConstOp ElementsAttr:$value), (TFL_ConstOp $value)>;

// Reorders adds to allow constant folding.
// Add --> Add $input, $constantA

[source,kotlin]
----
tasks.test {
    // lazy configuration
}

// Lazy reference
val testProvider: TaskProvider<Test> = tasks.test

testProvider {
    // lazy configuration
}

// Eagerly realized Test task, defeat configuration avoidance if done out of a lazy context
val test: Test = tasks.test.get()
----

This requires inspecting all configurations and artifacts.
Processing artifacts created by tasks causes those tasks to be realized and configured.

This internal representation is now created more lazily, which can change the order in which tasks are configured.
Some tasks may never be configured.