* </ul>
   *
   * <p>The overall principle is to continue to treat every checked exception as a checked
   * exception, every unchecked exception as an unchecked exception, and every error as an error. In
   * addition, the cause of any {@code ExecutionException} is wrapped in order to ensure that the
   * new stack trace matches that of the current thread.
   *

# the library which loads libtensorflow_framework.so
# (e.g. _pywrap_tensorflow_internal.so for Python, libtensorflow.so for the C
# API) is responsible for registering ops with libtensorflow_framework.so. In
# addition to this core set of ops, user libraries which are loaded (via
# TF_LoadLibrary/tf.load_op_library) register their ops and kernels with this
# shared object directly.
#

# Select supported compute capabilities (supported graphics cards).
# This is the same as the official TensorFlow builds.
# See https://developer.nvidia.com/cuda-gpus#compute
# `compute_XY` enables PTX embedding in addition to SASS. PTX
# is forward compatible beyond the current compute capability major
# release while SASS is only forward compatible inside the current
# major release. Example: sm_80 kernels can run on sm_89 GPUs but

 *   System.out.println(url.queryParameterName(i) + ": " + url.queryParameterValue(i));
 * }
 * ```
 *
 * which prints:
 *
 * ```
 * q: cute #puppies
 * f: images
 * ```
 *
 * In addition to composing URLs from their component parts and decomposing URLs into their
 * component parts, this class implements relative URL resolution: what address you'd reach by
 * clicking a relative link on a specified page. For example:

//
// ``legacyObject``: This is the legacy object in xlV1 format, preserved until its overwritten
//
// The most recently updated element in the array is considered the latest version.

// In addition to these we have a special kind called free-version. This is represented
// using a delete-marker and MetaSys entries. It's used to track tiered content of a

   * higher than performing just the operation without enforcing the capacity constraint.
   *
   * This implementation uses a per-segment queue to record a memento of the additions, removals,
   * and accesses that were performed on the map. The queue is drained on writes and when it exceeds
   * its capacity threshold.
   *

   * higher than performing just the operation without enforcing the capacity constraint.
   *
   * This implementation uses a per-segment queue to record a memento of the additions, removals,
   * and accesses that were performed on the map. The queue is drained on writes and when it exceeds
   * its capacity threshold.
   *