// Last dim smaller than expected_dim, zero-pad the input
    } else if (input_ty.getShape().back() < expected_dim) {
      SmallVector<int64_t, 4> no_padding(input_shape.size(), 0);
      SmallVector<int64_t, 4> padding(input_shape.size() - 1, 0);
      padding.push_back(expected_dim - input_shape.back());
      Value zero =
          GetScalarConstOfType(input_ty.getElementType(), loc, 0, &rewriter);

                type: string
              bpfExtToServiceConnmark:
                description: 'BPFExtToServiceConnmark in BPF mode, control a 32bit
                  mark that is set on connections from an external client to a local
                  service. This mark allows us to control how packets of that connection
                  are routed within the host and how is routing interpreted by RPF
                  check. [Default: 0]'

					// If s.pp.Pos already has a statement mark, then it was set here (below) for
					// the previous value.  If an actual instruction had been emitted for that
					// value, then the statement mark would have been reset.  Since the statement
					// mark of s.pp.Pos was not reset, this position (file/line) still needs a
					// statement mark on an instruction.  If file and line for this value are

- The job controller back-off logic is now decoupled from workqueue. In case of parallelism > 1, if there are multiple new failures in a reconciliation cycle, all the failures are taken into account to compute the back-off. Previously, the back-off kicked in for all types of failures; with this change, only pod failures are taken into account. If the back-off limits exceeds, the job is marked as failed immediately; before this change, the...

stability and performance improvements in the 1.13 release.  A new alpha feature is introduced to improve the mechanism that nodes heartbeat back to the control plane.  The `NodeLease` feature results in the node using a `Lease` resource in the `kube-node-lease` namespace that is renewed periodically.  The `NodeStatus` that was used previously to heartbeat back to the control plane is only updated when it changes.  This reduces load on the control plane for large clusters. The Kubelet plugin registration...

### Other (Cleanup or Flake)

- When the go-client reflector relists, the ResourceVersion list option is set to the reflector's latest synced resource version to ensure the reflector does not "go back in time" and reprocess events older than it has already processed. If the server responds with an HTTP 410 (Gone) status code response, the relist falls back to using `resourceVersion=""`. ([#83520](https://github.com/kubernetes/kubernetes/pull/83520), [@jpbetz](https://github.com/jpbetz))

    * OverrideSeparator string Delimiter to use to separate region of occurrence, url and signing region for each override Defaults to ","
    * ServiceOverrides string example: s3|region1, https://s3.foo.bar, some signing_region & ec2|region2, https://ec2.foo.bar, signing_region

* Update Cluster Autoscaler to version 1.3.7. ([#74136](https://github.com/kubernetes/kubernetes/pull/74136), [@jkaniuk](https://github.com/jkaniuk))
* Fix watch to not send the same set of events multiple times causing watcher to go back in time ([#73845](https://github.com/kubernetes/kubernetes/pull/73845), [@wojtek-t](https://github.com/wojtek-t))

- The `IPTablesOwnershipCleanup` feature (KEP-3178) is now GA; kubelet no longer
  creates the `KUBE-MARK-DROP` chain (which has been unused for several releases)
  or the `KUBE-MARK-MASQ` chain (which is now only created by kube-proxy). ([#119374](https://github.com/kubernetes/kubernetes/pull/119374), [@danwinship](https://github.com/danwinship))