* Upgrade the default etcd server version to 3.1.12 to pick up critical etcd "mvcc "unsynced" watcher restore operation" fix. ([#60998](https://github.com/kubernetes/kubernetes/pull/60998), [@jpbetz](https://github.com/jpbetz))

# Build with nvcc for CUDA and clang for host
build:nvcc_clang --config=cuda
# Unfortunately, cuda_configure.bzl demands this for using nvcc + clang
build:nvcc_clang --action_env=TF_CUDA_CLANG="1"
build:nvcc_clang --action_env=TF_NVCC_CLANG="1"
build:nvcc_clang --@local_config_cuda//:cuda_compiler=nvcc


# Debug config
build:dbg -c dbg

	NC	$8, (R15), n-8(SP)       // d407f010f000
	OC	$8, (R15), n-8(SP)       // d607f010f000
	MVC	$8, (R15), n-8(SP)       // d207f010f000
	MVCIN	$8, (R15), n-8(SP)       // e807f010f000
	CLC	$8, (R15), n-8(SP)       // d507f000f010
	XC	$256, -8(R15), -8(R15)   // b90400afc2a8fffffff8d7ffa000a000
	MVC	$256, 8192(R1), 8192(R2) // b90400a2c2a800002000b90400b1c2b800002000d2ffa000b000

	CMP	R1, R2                 // b9200012

  Args:
    environ_cp: copy of the os.environ.
  """
  question = 'Do you want to use clang as CUDA compiler?'
  yes_reply = 'Clang will be used as CUDA compiler.'
  no_reply = 'nvcc will be used as CUDA compiler.'
  set_action_env_var(
      environ_cp,
      'TF_CUDA_CLANG',
      None,
      True,
      question=question,
      yes_reply=yes_reply,
      no_reply=no_reply,

    ),
)

# Config setting that is satisfied when CUDA device code should be compiled
# with nvcc. It does not imply that CUDA support has been enabled.
alias(
    name = "is_cuda_compiler_nvcc",
    actual = if_oss(
        "@local_config_cuda//:is_cuda_compiler_nvcc",
        "@local_config_cuda//cuda:FALSE",
    ),
)

<code>R5</code>, <code>R6</code> and <code>R7</code> with the 64-bit values at
<code>0(R9)</code>, <code>8(R9)</code> and <code>16(R9)</code> respectively.
</p>

<p>
Storage-and-storage instructions such as <code>MVC</code> and <code>XC</code> are written
with the length as the first argument.
For example, <code>XC</code> <code>$8,</code> <code>(R9),</code> <code>(R9)</code> would clear
eight bytes at the address specified in <code>R9</code>.

    *   Add documentation for ops supported by the `tf2xla` bridge.
    *   Fix minor typos in the doc of `SpaceToDepth` and `DepthToSpace`.
    *   Updated documentation comments in `mfcc_mel_filterbank.h` and `mfcc.h`
        to clarify that the input domain is squared magnitude spectra and the
        weighting is done on linear magnitude spectra (sqrt of inputs).