wondered at this, but at the time it all seemed quite natural);
but when the Rabbit actually TOOK A WATCH OUT OF ITS WAISTCOAT-
POCKET, and looked at it, and then hurried on, Alice started to
her feet, for it flashed across her mind that she had never
before seen a rabbit with either a waistcoat-pocket, or a watch to
take out of it, and burning with curiosity, she ran across the
field after it, and fortunately was just in time to see it pop

wondered at this, but at the time it all seemed quite natural);
but when the Rabbit actually TOOK A WATCH OUT OF ITS WAISTCOAT-
POCKET, and looked at it, and then hurried on, Alice started to
her feet, for it flashed across her mind that she had never
before seen a rabbit with either a waistcoat-pocket, or a watch to
take out of it, and burning with curiosity, she ran across the
field after it, and fortunately was just in time to see it pop

	//
	// The 1s sleep in the test allows GetDeletableResources and
	// gc.resyncMonitors to run ~5 times to ensure the changes to the
	// fakeDiscoveryClient are picked up.
	go gc.Sync(tCtx, fakeDiscoveryClient, 200*time.Millisecond)

	// Wait until the sync discovers the initial resources
	time.Sleep(1 * time.Second)

		pp, _ = pidleget(0)
	}
	var locked bool
	if pp == nil {
		globrunqput(gp)

		// Below, we stoplockedm if gp is locked. globrunqput releases
		// ownership of gp, so we must check if gp is locked prior to
		// committing the release by unlocking sched.lock, otherwise we
		// could race with another M transitioning gp from unlocked to
		// locked.
		locked = gp.lockedm != 0

  // batch size.

  // Pack operation should pack at least two values.
  if (getValues().size() < 2) return {};

  // Dimensions packed along axis = 0 (pack scalars into vector).
  if (getAxis() != 0) return {};

  // First packed value is defined by a strided slice operation.
  auto slice_op =
      dyn_cast_or_null<StridedSliceOp>(getValues()[0].getDefiningOp());
  if (!slice_op) return {};

	// added in z13
	op_CXPT   uint32 = 0xEDAF // 	RSL-b	CONVERT FROM PACKED (to extended DFP)
	op_CDPT   uint32 = 0xEDAE // 	RSL-b	CONVERT FROM PACKED (to long DFP)
	op_CPXT   uint32 = 0xEDAD // 	RSL-b	CONVERT TO PACKED (from extended DFP)
	op_CPDT   uint32 = 0xEDAC // 	RSL-b	CONVERT TO PACKED (from long DFP)
	op_LZRF   uint32 = 0xE33B // 	RXY-a	LOAD AND ZERO RIGHTMOST BYTE (32)

        return entry.getValue();
      }
    };
  }

  /**
   * Returns an immutable map instance containing the given entries. Internally, the returned map
   * will be backed by an {@link EnumMap}.
   *
   * <p>The iteration order of the returned map follows the enum's iteration order, not the order in
   * which the elements appear in the given map.
   *

        return entry.getValue();
      }
    };
  }

  /**
   * Returns an immutable map instance containing the given entries. Internally, the returned map
   * will be backed by an {@link EnumMap}.
   *
   * <p>The iteration order of the returned map follows the enum's iteration order, not the order in
   * which the elements appear in the given map.
   *

    only supports resource variable read or read + write. For all TPU computations,
    resource variables are required to be initialized prior to execution. Write only
    resource variable uses can be generated currently via packed tensor uses.

    For example, the following:

    ```mlir
    func @write_only_resource(%value: tensor<i32>, %resource: tensor<*x!tf_type.resource<tensor<i32>>>) {

  %9 = "tf.Pack"(%8, %d1, %d2) {axis = 0 : i64} : (tensor<i32>, tensor<i32>, tensor<i32>) -> tensor<3xi32>
  %10 = "tf.Reshape"(%arg1, %9) : (tensor<?x1x2xf32>, tensor<3xi32>) -> tensor<?x1x2xf32>

  // Packed dimensions have different order from the reshape operand:
  //   [?, 1, 2] vs [?, 2, 1]