rocm_jax/jax/experimental/pjit.py

# Copyright 2021 Google LLC
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     https://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

from enum import IntEnum
import numpy as np
from collections import OrderedDict, Counter
from typing import Callable, Sequence, Tuple, Union
from warnings import warn
import itertools as it
from functools import partial

from . import maps
from . import PartitionSpec
from .. import core
from .. import linear_util as lu
from .._src.api import _check_callable, _check_arg
from .._src import source_info_util
from ..api_util import (argnums_partial_except, flatten_axes,
                        flatten_fun_nokwargs, _ensure_index_tuple,
                        donation_vector, rebase_donate_argnums)
from ..errors import JAXTypeError
from ..interpreters import ad
from ..interpreters import pxla
from ..interpreters import xla
from ..interpreters import batching
from ..interpreters import partial_eval as pe
from ..lib import xla_bridge as xb
from ..lib import xla_client as xc
from ..tree_util import tree_flatten, tree_unflatten
from .._src.util import (extend_name_stack, HashableFunction, safe_zip,
                         wrap_name, wraps, distributed_debug_log,
                         split_list, cache, tuple_insert)
xops = xc._xla.ops

def pjit(fun: Callable,
         in_axis_resources,
         out_axis_resources,
         static_argnums: Union[int, Sequence[int]] = (),
         donate_argnums: Union[int, Sequence[int]] = ()):
  """Makes ``fun`` compiled and automatically partitioned across multiple devices.

  The returned function has semantics equivalent to those of ``fun``, but is
  compiled to an XLA computation that runs across multiple devices
  (e.g. multiple GPUs or multiple TPU cores). This can be useful if the jitted
  version of ``fun`` would not fit in a single device's memory, or to speed up
  ``fun`` by running each operation in parallel across multiple devices.

  The partitioning over devices happens automatically based on
  propagation of input partitioning specified in ``in_axis_resources`` and
  output partitioning specified in ``out_axis_resources``. The resources
  specified in those two arguments must refer to mesh axes, as defined by
  the :py:func:`jax.experimental.maps.mesh` context manager. Note that the mesh
  definition at ``pjit`` application time is ignored, and the returned function
  will use the mesh definition available at each call site.

  Inputs to a pjit'd function will be automatically partitioned across devices
  if they're not already correctly partitioned based on ``in_axis_resources``.
  In some scenarios, ensuring that the inputs are already correctly pre-partitioned
  can increase performance. For example, if passing the output of one pjit'd function
  to another pjit’d function (or the same pjit’d function in a loop), make sure the
  relevant ``out_axis_resources`` match the corresponding ``in_axis_resources``.

  .. note::
    **Multi-process platforms:** On multi-process platforms such as TPU pods,
    ``pjit`` can be used to run computations across all available devices across
    processes. To achieve this, ``pjit`` is designed to be used in SPMD Python
    programs, where every process is running the same Python code such that all
    processes run the same pjit'd function in the same order.

    When running in this configuration, the mesh should contain devices across
    all processes. However, any input argument dimensions partitioned over
    multi-process mesh axes should be of size equal to the corresponding *local*
    mesh axis size, and outputs will be similarly sized according to the local
    mesh. ``fun`` will still be executed across *all* devices in the mesh,
    including those from other processes, and will be given a global view of the
    data spread accross multiple processes as a single array. However, outside
    of ``pjit`` every process only "sees" its local piece of the input and output,
    corresponding to its local sub-mesh.

    The SPMD model requires that the same multi-process ``pjit``'d functions must
    be run in the same order on all processes, but they can be interspersed with
    arbitrary operations running in a single process.

  Args:
    fun: Function to be compiled. Should be a pure function, as side-effects may
      only be executed once. Its arguments and return value should be arrays,
      scalars, or (nested) standard Python containers (tuple/list/dict) thereof.
      Positional arguments indicated by ``static_argnums`` can be anything at
      all, provided they are hashable and have an equality operation defined.
      Static arguments are included as part of a compilation cache key, which is
      why hash and equality operators must be defined.
    in_axis_resources: Pytree of structure matching that of arguments to ``fun``,
      with all actual arguments replaced by resource assignment specifications.
      It is also valid to specify a pytree prefix (e.g. one value in place of a
      whole subtree), in which case the leaves get broadcast to all values in
      that subtree.

      The valid resource assignment specifications are:
        - :py:obj:`None`, in which case the value will be replicated on all devices
        - :py:class:`PartitionSpec`, a tuple of length at most equal to the rank
          of the partitioned value. Each element can be a :py:obj:`None`, a mesh
          axis or a tuple of mesh axes, and specifies the set of resources assigned
          to partition the value's dimension matching its position in the spec.

      The size of every dimension has to be a multiple of the total number of
      resources assigned to it.
    out_axis_resources: Like ``in_axis_resources``, but specifies resource
      assignment for function outputs.
    static_argnums: An optional int or collection of ints that specify which
      positional arguments to treat as static (compile-time constant).
      Operations that only depend on static arguments will be constant-folded in
      Python (during tracing), and so the corresponding argument values can be
      any Python object.

      Static arguments should be hashable, meaning both ``__hash__`` and
      ``__eq__`` are implemented, and immutable. Calling the jitted function
      with different values for these constants will trigger recompilation.
      Arguments that are not arrays or containers thereof must be marked as
      static.

      If ``static_argnums`` is not provided, no arguments are treated as static.
    donate_argnums: Specify which arguments are "donated" to the computation.
      It is safe to donate arguments if you no longer need them once the
      computation has finished. In some cases XLA can make use of donated
      buffers to reduce the amount of memory needed to perform a computation,
      for example recycling one of your input buffers to store a result. You
      should not reuse buffers that you donate to a computation, JAX will raise
      an error if you try to.

  Returns:
    A wrapped version of ``fun``, set up for just-in-time compilation and
    automatic partitioned by the mesh available at each call site.

  For example, a convolution operator can be automatically partitioned over
  an arbitrary set of devices by a single ```pjit`` application:

  >>> import jax
  >>> import jax.numpy as jnp
  >>> from jax.experimental.maps import mesh
  >>> from jax.experimental.pjit import PartitionSpec, pjit
  >>>
  >>> x = jnp.arange(8, dtype=jnp.float32)
  >>> f = pjit(lambda x: jax.numpy.convolve(x, jnp.asarray([0.5, 1.0, 0.5]), 'same'),
  ...         in_axis_resources=None, out_axis_resources=PartitionSpec('devices'))
  >>> with mesh(jax.devices(), ('devices',)):
  ...   print(f(x))  # doctest: +SKIP
  [ 0.5  2.   4.   6.   8.  10.  12.  10. ]
  """
  warn("pjit is an experimental feature and probably has bugs!")
  _check_callable(fun)

  # To be a tree prefix of the positional args tuple, in_axes can never be a
  # list: if in_axes is not a leaf, it must be a tuple of trees. However,
  # in cases like these users expect tuples and lists to be treated
  # essentially interchangeably, so we canonicalize lists to tuples here
  # rather than raising an error. https://github.com/google/jax/issues/2367
  if isinstance(in_axis_resources, list):
    in_axis_resources = tuple(in_axis_resources)
  if isinstance(out_axis_resources, list):
    out_axis_resources = tuple(out_axis_resources)

  in_axis_resources, in_axis_resources_entries, _ = \
      _prepare_axis_resources(in_axis_resources, "in_axis_resources")
  out_axis_resources, out_axis_resources_entries, out_axis_treedef = \
      _prepare_axis_resources(out_axis_resources, "out_axis_resources")
  out_axis_resources_entries = tuple(out_axis_resources_entries)

  static_argnums = _ensure_index_tuple(static_argnums)
  donate_argnums = _ensure_index_tuple(donate_argnums)
  donate_argnums = rebase_donate_argnums(donate_argnums, static_argnums)

  @wraps(fun)
  def wrapped(*args, **kwargs):
    if kwargs:
      raise NotImplementedError("pjit over kwargs not yet supported")
    if max(static_argnums + donate_argnums, default=-1) >= len(args):
      raise ValueError(f"jitted function has static_argnums={static_argnums}, "
                       f"donate_argnums={donate_argnums} but "
                       f"was called with only {len(args)} positional arguments.")

    # Putting this outside of wrapped would make resources lexically scoped
    resource_env = maps.thread_resources.env
    mesh = resource_env.physical_mesh
    if mesh.empty:
      raise RuntimeError("pjit requires a non-empty mesh! Are you sure that "
                         "it's defined at the call site?")

    f = lu.wrap_init(fun)
    if static_argnums:
      f, dyn_args = argnums_partial_except(
          f, static_argnums, args, allow_invalid=False)
    else:
      dyn_args = args

    args_flat, in_tree = tree_flatten(args)
    for arg in args_flat: _check_arg(arg)
    flat_fun, out_tree = flatten_fun_nokwargs(f, in_tree)
    if donate_argnums:
      donated_invars = donation_vector(donate_argnums, dyn_args, ())
    else:
      donated_invars = (False,) * len(args_flat)

    local_in_avals = tuple(core.raise_to_shaped(core.get_aval(a)) for a in args_flat)
    jaxpr, in_axis_resources_flat, out_axis_resources_flat = \
        _pjit_jaxpr(flat_fun, mesh, local_in_avals,
                    in_tree, hashable_pytree(in_axis_resources),
                    HashableFunction(out_tree, closure=()), hashable_pytree(out_axis_resources))

    out = pjit_p.bind(
        *args_flat,
        jaxpr=jaxpr,
        in_axis_resources=in_axis_resources_flat,
        out_axis_resources=out_axis_resources_flat,
        resource_env=resource_env,
        donated_invars=donated_invars,
        name=flat_fun.__name__)
    return tree_unflatten(out_tree(), out)

  return wrapped

class _ListWithW(list):
  __slots__ = ('__weakref__',)

def hashable_pytree(pytree):
  vals, treedef = tree_flatten(pytree)
  vals = tuple(vals)
  return HashableFunction(lambda: tree_unflatten(treedef, vals),
                          closure=(treedef, vals))

@lu.cache
def _pjit_jaxpr(fun, mesh, local_in_avals,
                in_tree, in_axis_resources_thunk,
                out_tree, out_axis_resources_thunk):
  in_axis_resources_flat = tuple(flatten_axes("pjit in_axis_resources",
                                              in_tree, in_axis_resources_thunk()))
  _check_shapes_against_resources("pjit arguments", False, mesh.local_mesh.shape,
                                  local_in_avals, in_axis_resources_flat)
  global_in_avals = local_to_global(mesh, local_in_avals, in_axis_resources_flat)

  jaxpr, global_out_avals, consts = pe.trace_to_jaxpr_dynamic(fun, global_in_avals)
  jaxpr = core.ClosedJaxpr(jaxpr, consts)

  out_axis_resources_flat = tuple(flatten_axes("pjit out_axis_resources",
                                               out_tree(), out_axis_resources_thunk()))
  _check_shapes_against_resources("pjit outputs", mesh.is_multi_process, mesh.shape,
                                  global_out_avals, out_axis_resources_flat)
  # lu.cache needs to be able to create weakrefs to outputs, so we can't return a plain tuple
  return _ListWithW([jaxpr, in_axis_resources_flat, out_axis_resources_flat])


class SpecSync(IntEnum):
  """Encodes how much out of sync the real value of partitions is compared to the user specified one.

  We use this to make sure we don't show garbage modified values while claiming
  that the users have specified them like that.
  """
  DIM_PERMUTE = 1  # Dimensions permuted, but no new sharding axes
  IN_SYNC = 2  # Entirely in sync

class ParsedPartitionSpec:
  __slots__ = ('partitions', 'unsafe_user_spec', 'sync')

  def __init__(self, user_spec, partitions, sync=SpecSync.IN_SYNC):
    self.partitions = tuple(partitions)
    self.unsafe_user_spec = user_spec
    self.sync = sync

  @property
  def user_spec(self):
    return self.unsynced_user_spec(SpecSync.IN_SYNC)

  def unsynced_user_spec(self, min_sync):
    if self.sync < min_sync:
      raise AssertionError(f"Please open a bug report! ({self.sync} >= {min_sync})")
    return self.unsafe_user_spec

  def insert_axis_partitions(self, dim, val):
    parts = self.partitions
    too_short = dim - len(parts)
    if too_short > 0:
      parts += ((),) * too_short
    new_partitions = tuple_insert(parts, dim, val)
    new_sync = SpecSync.DIM_PERMUTE if val == () else SpecSync.IN_SYNC
    return ParsedPartitionSpec(self.unsafe_user_spec, new_partitions, sync=new_sync)

  @classmethod
  def from_user_input(cls, entry, arg_name):
    if entry is None:
      return cls(entry, ())
    if not isinstance(entry, PartitionSpec):
      raise TypeError(f"{arg_name} are expected to be "
                      f"PartitionSpec instances or None, but got {entry}")
    axis_specs = []
    for axis_spec in entry:
      if axis_spec is None:
        axis_spec = ()
      elif isinstance(axis_spec, (list, tuple)):
        axis_spec = tuple(axis_spec)
      else:
        axis_spec = (axis_spec,)
      axis_specs.append(axis_spec)
    return cls(entry, axis_specs)

  def __hash__(self):
    return hash((self.partitions, self.sync))

  def __eq__(self, other):
    return (self.partitions == other.partitions and
            self.unsafe_user_spec == other.unsafe_user_spec and
            self.sync == other.sync)

  def __len__(self):
    return len(self.partitions)

  def __getitem__(self, i):
    return self.partitions[i]

  def __iter__(self):
    return iter(self.partitions)

REPLICATED = ParsedPartitionSpec(None, ())


def _prepare_axis_resources(axis_resources, arg_name):
  # PyTrees don't treat None values as leaves, so we explicitly need
  # to explicitly declare them as such
  entries, treedef = tree_flatten(axis_resources, is_leaf=lambda x: x is None)
  what = f"{arg_name} leaf specifications"
  entries = [ParsedPartitionSpec.from_user_input(entry, what) for entry in entries]
  _check_unique_resources(entries, arg_name)
  return tree_unflatten(treedef, entries), entries, treedef

def _check_unique_resources(axis_resources, arg_name):
  for arg_axis_resources in axis_resources:
    if not arg_axis_resources: continue
    resource_counts = Counter(it.chain.from_iterable(arg_axis_resources))
    if not resource_counts: continue
    if resource_counts.most_common(1)[0][1] > 1:
      multiple_uses = [r for r, c in resource_counts.items() if c > 1]
      if multiple_uses:
        raise ValueError(f"A single {arg_name} specification can map every mesh axis "
                         f"to at most one positional dimension, but {arg_axis_resources.user_spec} "
                         f"has duplicate entries for {maps.show_axes(multiple_uses)}")

def _check_shapes_against_resources(what: str, is_global_shape: bool, mesh_shape, flat_avals, flat_axis_resources):
  global_str = " global" if is_global_shape else ""
  for aval, aval_axis_resources in zip(flat_avals, flat_axis_resources):
    shape = aval.shape
    if len(shape) < len(aval_axis_resources):
      raise ValueError(f"One of {what} was given the resource assignment "
                       f"of {aval_axis_resources.user_spec}, which implies that "
                       f"it has a rank of at least {len(aval_axis_resources)}, "
                       f"but it is {len(shape)}")
    for i, axis_resources in enumerate(aval_axis_resources):
      try:
        size = int(np.prod([mesh_shape[resource] for resource in axis_resources], dtype=np.int64))
      except KeyError as e:
        raise ValueError(f"One of {what} was given the resource assignment "
                         f"of {aval_axis_resources.user_spec}, but resource axis "
                         f"{e.args[0]} is undefined. Did you forget to declare the mesh?") from None
      if shape[i] % size != 0:
        raise ValueError(f"One of {what} was given the resource assignment "
                         f"of {aval_axis_resources.user_spec}, which implies that "
                         f"the{global_str} size of its dimension {i} should be "
                         f"divisible by {size}, but it is equal to {shape[i]}")

# -------------------- pjit rules --------------------

pjit_p = core.Primitive("pjit")
pjit_p.multiple_results = True


def _pjit_call_impl(*args, jaxpr,
                    in_axis_resources, out_axis_resources,
                    resource_env, donated_invars, name):
  compiled = pjit_callable(
      jaxpr, in_axis_resources, out_axis_resources,
      resource_env, donated_invars, name)
  distributed_debug_log(("Running pjit'd function", name),
                        ("mesh", resource_env.physical_mesh))
  return compiled(*args)
pjit_p.def_impl(_pjit_call_impl)

@cache()
def pjit_callable(
    jaxpr: core.ClosedJaxpr,
    in_axis_resources: Tuple[ParsedPartitionSpec, ...],
    out_axis_resources: Tuple[ParsedPartitionSpec, ...],
    resource_env,
    donated_invars,
    name: str):

  in_axes = [get_array_mapping(axes) for axes in in_axis_resources]
  out_axes = [get_array_mapping(axes) for axes in out_axis_resources]
  fun = lu.wrap_init(core.jaxpr_as_fun(jaxpr))
  local_in_avals = global_to_local(resource_env.physical_mesh,
                                   jaxpr.in_avals, in_axis_resources)
  # TODO(skye): allow for using a submesh of physical_mesh
  return pxla.mesh_callable(fun, name, None, resource_env.physical_mesh,
                            in_axes, out_axes, donated_invars,
                            True, *local_in_avals, tile_by_mesh_axes=False,
                            do_resource_typecheck="pjit")



def _pjit_abstract_eval(*args, jaxpr, out_axis_resources, resource_env, **_):
  return global_to_local(resource_env.physical_mesh,
                         jaxpr.out_avals, out_axis_resources)
pjit_p.def_abstract_eval(_pjit_abstract_eval)


def _pjit_translation_rule(c, axis_env, in_nodes, name_stack, backend, name,
                           jaxpr, in_axis_resources, out_axis_resources,
                           resource_env, donated_invars):
  mesh = resource_env.physical_mesh
  subc = xc.XlaBuilder(f"pjit_{name}")

  args = []
  for i, (n, axis_resources) in enumerate(safe_zip(in_nodes, in_axis_resources)):
    # N.B. inlined calls shouldn't have shardings set directly on the inputs or
    # outputs (set_sharding_proto adds an identity operation).
    arg = xb.parameter(subc, i, c.GetShape(n))
    args.append(xb.set_sharding_proto(subc, arg,
                                      get_sharding_proto(c, n, axis_resources, mesh)))

  # TODO: Think about how to avoid duplicating constants with the outer jaxpr
  out_nodes = xla.jaxpr_subcomp(
      subc, jaxpr.jaxpr, backend, axis_env, xla._xla_consts(subc, jaxpr.consts),
      extend_name_stack(name_stack, wrap_name(name, "pjit")), *args)
  out_nodes = [
      xb.set_sharding_proto(subc, out,
                            get_sharding_proto(subc, out, axis_resources, mesh))
      for out, axis_resources in safe_zip(out_nodes, out_axis_resources)
  ]

  subc = subc.build(xops.Tuple(subc, out_nodes))
  return xops.Call(c, subc, list(in_nodes))
xla.call_translations[pjit_p] = _pjit_translation_rule


def _pjit_batcher(vals_in, dims_in,
                  axis_name, main_type,
                  jaxpr, in_axis_resources, out_axis_resources,
                  resource_env, donated_invars, name):
  axis_size, = {x.shape[d] for x, d in zip(vals_in, dims_in) if d is not batching.not_mapped}
  # batch_jaxpr expects all batching dimensions to be equal to 0
  vals_in = [batching.moveaxis(x, d, 0) if d is not batching.not_mapped and d != 0
             else x for x, d in zip(vals_in, dims_in)]
  is_mapped_in = [d is not batching.not_mapped for d in dims_in]
  new_jaxpr, is_mapped_out = batching.batch_jaxpr(
      jaxpr, axis_size, is_mapped_in,
      instantiate=False, axis_name=axis_name, main_type=main_type)

  in_axis_resources = tuple(
      spec.insert_axis_partitions(0, ()) if is_mapped else spec
      for is_mapped, spec in zip(is_mapped_in, in_axis_resources))
  out_axis_resources = tuple(
      spec.insert_axis_partitions(0, ()) if is_mapped else spec
      for is_mapped, spec in zip(is_mapped_out, out_axis_resources))
  vals_out = pjit_p.bind(
    *vals_in,
    jaxpr=new_jaxpr,
    in_axis_resources=in_axis_resources,
    out_axis_resources=out_axis_resources,
    resource_env=resource_env,
    donated_invars=donated_invars,
    name=name)
  dims_out = [0 if batched else batching.not_mapped for batched in is_mapped_out]
  return vals_out, dims_out
batching.initial_style_batchers[pjit_p] = _pjit_batcher


def _pjit_jvp(primals_in, tangents_in,
              jaxpr, in_axis_resources, out_axis_resources,
              resource_env, donated_invars, name):
  is_nz_tangents_in = [type(t) is not ad.Zero for t in tangents_in]
  jaxpr_jvp, is_nz_tangents_out = ad.jvp_jaxpr(
      jaxpr, is_nz_tangents_in, instantiate=False)

  def _filter_zeros(is_nz_l, l):
    return (x for nz, x in zip(is_nz_l, l) if nz)
  _filter_zeros_in = partial(_filter_zeros, is_nz_tangents_in)
  _filter_zeros_out = partial(_filter_zeros, is_nz_tangents_out)
  outputs = pjit_p.bind(
      *primals_in, *_filter_zeros_in(tangents_in),
      jaxpr=jaxpr_jvp,
      in_axis_resources=(*in_axis_resources, *_filter_zeros_in(in_axis_resources)),
      out_axis_resources=(*out_axis_resources, *_filter_zeros_out(out_axis_resources)),
      resource_env=resource_env,
      donated_invars=(*donated_invars, *_filter_zeros_in(donated_invars)),
      name=wrap_name(name, 'jvp'))

  primals_out, tangents_out = split_list(outputs, [-len(is_nz_tangents_out)])
  assert len(primals_out) == len(jaxpr.jaxpr.outvars)
  tangents_out_it = iter(tangents_out)
  return primals_out, [next(tangents_out_it) if nz else ad.Zero(aval)
                       for nz, aval in zip(is_nz_tangents_out, jaxpr.out_avals)]
ad.primitive_jvps[pjit_p] = _pjit_jvp


def _check_resources_against_named_axes(what, aval, pos_axis_resources, named_axis_resources):
  pjit_resources = set(it.chain.from_iterable(pos_axis_resources))
  aval_resources = set(it.chain.from_iterable(
    named_axis_resources[a] for a in aval.named_shape))
  overlap = pjit_resources & aval_resources
  if overlap:
    raise JAXTypeError(
        f"{what} has an axis resources specification of "
        f"{pos_axis_resources.unsynced_user_spec(SpecSync.DIM_PERMUTE)} "
        f"that uses one or more mesh axes already used by xmap to partition "
        f"a named axis appearing in its named_shape (both use mesh axes "
        f"{maps.show_axes(overlap)})")

def _resource_typing_pjit(avals, params, source_info, named_axis_resources):
  jaxpr = params["jaxpr"]
  what = "pjit input"
  for aval, pos_axis_resources in zip(jaxpr.in_avals, params['in_axis_resources']):
    _check_resources_against_named_axes(what, aval, pos_axis_resources, named_axis_resources)
  pxla.resource_typecheck(
      jaxpr.jaxpr, named_axis_resources,
      lambda: (f"a pjit'ed function {params['name']} "
               f"(pjit called at {source_info_util.summarize(source_info)})"))
  what = "pjit output"
  for aval, pos_axis_resources in zip(jaxpr.out_avals, params['out_axis_resources']):
    _check_resources_against_named_axes(what, aval, pos_axis_resources, named_axis_resources)
pxla.custom_resource_typing_rules[pjit_p] = _resource_typing_pjit


# -------------------- with_sharding_constraint --------------------

def with_sharding_constraint(x, axis_resources):
  x_flat, tree = tree_flatten(x)
  parsed_axis_resources, entries, _ = _prepare_axis_resources(axis_resources, "axis_resources")
  axis_resources_flat = tuple(
      flatten_axes("with_sharding_constraint axis_resources",
                   tree, parsed_axis_resources))
  resource_env = maps.thread_resources.env
  mesh = resource_env.physical_mesh
  _check_shapes_against_resources(
      "with_sharding_constraint arguments",
      mesh.is_multi_process, mesh.shape,
      x_flat, axis_resources_flat)
  outs = [sharding_constraint_p.bind(y, axis_resources=r, resource_env=resource_env)
          for y, r in safe_zip(x_flat, axis_resources_flat)]
  return tree_unflatten(tree, outs)

def _sharding_constraint_impl(x, axis_resources, resource_env):
  # TODO(skye): can we also prevent this from being called in other
  # non-pjit contexts? (e.g. pmap, control flow)
  raise NotImplementedError(
      "with_sharding_constraint() should only be called inside pjit()")

def _sharding_constraint_translation_rule(c, x_node, axis_resources, resource_env):
  mesh = resource_env.physical_mesh
  return xb.set_sharding_proto(c, x_node,
                               get_sharding_proto(c, x_node, axis_resources, mesh))

sharding_constraint_p = core.Primitive("sharding_constraint")
sharding_constraint_p.def_impl(_sharding_constraint_impl)
sharding_constraint_p.def_abstract_eval(lambda x, **unused: x)
ad.deflinear2(sharding_constraint_p,
              lambda ct, _, axis_resources, resource_env: (
                  sharding_constraint_p.bind(
                      ct, axis_resources=axis_resources, resource_env=resource_env),))
xla.translations[sharding_constraint_p] = _sharding_constraint_translation_rule

def _resource_typing_sharding_constraint(avals, params, source_info, named_axis_resources):
  aval, = avals
  _check_resources_against_named_axes(
    "with_sharding_constraint input", aval,
    params['axis_resources'], named_axis_resources)
pxla.custom_resource_typing_rules[sharding_constraint_p] = \
    _resource_typing_sharding_constraint

# -------------------- helpers --------------------

def get_array_mapping(axis_resources: ParsedPartitionSpec) -> pxla.ArrayMapping:
  return OrderedDict((axis, i)
                     for i, axes in enumerate(axis_resources)
                     for axis in axes)

def get_sharding_proto(c, xla_op, axis_resources: ParsedPartitionSpec,
                       mesh: maps.Mesh) -> xc.OpSharding:
  xla_shape = c.GetShape(xla_op)
  if xla_shape.is_token():
    aval = core.abstract_token
    assert axis_resources is REPLICATED
  else:
    aval = core.ShapedArray(xla_shape.dimensions(), xla_shape.element_type())
  return get_aval_sharding_proto(aval, axis_resources, mesh)


def get_aval_sharding_proto(aval: core.AbstractValue,
                            axis_resources: ParsedPartitionSpec,
                            mesh: maps.Mesh) -> xc.OpSharding:
  array_mapping = get_array_mapping(axis_resources)
  sharding_spec = pxla.mesh_sharding_specs(mesh.shape, mesh.axis_names)(
      aval, array_mapping)
  return sharding_spec.sharding_proto()

def global_to_local(mesh, avals, axes):
  return [mesh.global_to_local(get_array_mapping(aval_axes), aval)
          for aval, aval_axes in zip(avals, axes)]

def local_to_global(mesh, avals, axes):
  return [mesh.local_to_global(get_array_mapping(aval_axes), aval)
          for aval, aval_axes in zip(avals, axes)]