dace.transformation.interstate package

Submodules

dace.transformation.interstate.fpga_transform_sdfg module

Contains inter-state transformations of an SDFG to run on an FPGA.

class dace.transformation.interstate.fpga_transform_sdfg.FPGATransformSDFG(*args, **kwargs)

Bases: MultiStateTransformation

Implements the FPGATransformSDFG transformation, which takes an entire SDFG and transforms it into an FPGA-capable SDFG.

static annotates_memlets()

Indicates whether the transformation annotates the edges it creates or modifies with the appropriate memlets. This determines whether to apply memlet propagation after the transformation.

apply(*args, **kwargs)

Applies this transformation instance on the matched pattern graph.

Parameters:

sdfg – The SDFG to apply the transformation to.

Returns:

A transformation-defined return value, which could be used to pass analysis data out, or nothing.

apply_pass(*args, **kwargs)

Applies the pass to the given SDFG.

Parameters:

• sdfg – The SDFG to apply the pass to.

• pipeline_results – If in the context of a Pipeline, a dictionary that is populated with prior Pass results as {Pass subclass name: returned object from pass}. If not run in a pipeline, an empty dictionary is expected.

Returns:

Some object if pass was applied, or None if nothing changed.

apply_pattern(*args, **kwargs)

Applies this transformation on the given SDFG, using the transformation instance to find the right control flow graph object (based on control flow graph ID), and applying memlet propagation as necessary.

Parameters:

• append – If True, appends the transformation to the SDFG transformation history.

• annotate – If True, applies memlet propagation as necessary.

Returns:

A transformation-defined return value, which could be used to pass analysis data out, or nothing.

apply_to(*args, **kwargs)

Applies this transformation to a given subgraph, defined by a set of nodes. Raises an error if arguments are invalid or transformation is not applicable.

The subgraph is defined by the where dictionary, where each key is taken from the PatternNode fields of the transformation. For example, applying MapCollapse on two maps can pe performed as follows:

` MapCollapse.apply_to(sdfg, outer_map_entry=map_a, inner_map_entry=map_b) `

Parameters:

• sdfg – The SDFG to apply the transformation to.

• options – A set of parameters to use for applying the transformation.

• expr_index – The pattern expression index to try to match with.

• verify – Check that can_be_applied returns True before applying.

• annotate – Run memlet propagation after application if necessary.

• permissive – Apply transformation in permissive mode.

• save – Save transformation as part of the SDFG file. Set to False if composing transformations.

• where – A dictionary of node names (from the transformation) to nodes in the SDFG or a single state.

can_be_applied(*args, **kwargs)

Returns True if this transformation can be applied on the candidate matched subgraph.

Parameters:

• graph – SDFG object in which the match was found.

• candidate – A mapping between node IDs returned from PatternTransformation.expressions and the nodes in graph.

• expr_index – The list index from PatternTransformation.expressions that was matched.

• sdfg – The SDFG in which the match was found (equal to graph).

• permissive – Whether transformation should run in permissive mode.

Returns:

True if the transformation can be applied.

classmethod expressions()

Returns a list of SDFG subgraphs that will be matched in the subgraph isomorphism phase. Used as a pre-pass before calling can_be_applied.

promote_global_trans

If True, transient arrays that are fully internal are pulled out so that they can be allocated on the host.

properties()

setup_match(*args, **kwargs)

Sets the transformation to a given subgraph pattern.

Parameters:

• cfg_id – A unique ID of the SDFG.

• state_id – The node ID of the SDFG state, if applicable. If transformation does not operate on a single state, the value should be -1.

• subgraph – A mapping between node IDs returned from PatternTransformation.expressions and the nodes in graph.

• expr_index – The list index from PatternTransformation.expressions that was matched.

• override – If True, accepts the subgraph dictionary as-is (mostly for internal use).

• options – An optional dictionary of transformation properties

Raises:

• TypeError – When transformation is not subclass of PatternTransformation.

• TypeError – When state_id is not instance of int.

• TypeError – When subgraph is not a dict of {PatternNode: int}.

dace.transformation.interstate.fpga_transform_state module

Contains inter-state transformations of an SDFG to run on an FPGA.

class dace.transformation.interstate.fpga_transform_state.FPGATransformState(*args, **kwargs)

Bases: MultiStateTransformation

Implements the FPGATransformState transformation.

apply(*args, **kwargs)

Applies this transformation instance on the matched pattern graph.

Parameters:

sdfg – The SDFG to apply the transformation to.

Returns:

A transformation-defined return value, which could be used to pass analysis data out, or nothing.

apply_pass(*args, **kwargs)

Applies the pass to the given SDFG.

Parameters:

• sdfg – The SDFG to apply the pass to.

• pipeline_results – If in the context of a Pipeline, a dictionary that is populated with prior Pass results as {Pass subclass name: returned object from pass}. If not run in a pipeline, an empty dictionary is expected.

Returns:

Some object if pass was applied, or None if nothing changed.

apply_pattern(*args, **kwargs)

Applies this transformation on the given SDFG, using the transformation instance to find the right control flow graph object (based on control flow graph ID), and applying memlet propagation as necessary.

Parameters:

• append – If True, appends the transformation to the SDFG transformation history.

• annotate – If True, applies memlet propagation as necessary.

Returns:

A transformation-defined return value, which could be used to pass analysis data out, or nothing.

apply_to(*args, **kwargs)

Applies this transformation to a given subgraph, defined by a set of nodes. Raises an error if arguments are invalid or transformation is not applicable.

The subgraph is defined by the where dictionary, where each key is taken from the PatternNode fields of the transformation. For example, applying MapCollapse on two maps can pe performed as follows:

` MapCollapse.apply_to(sdfg, outer_map_entry=map_a, inner_map_entry=map_b) `

Parameters:

• sdfg – The SDFG to apply the transformation to.

• options – A set of parameters to use for applying the transformation.

• expr_index – The pattern expression index to try to match with.

• verify – Check that can_be_applied returns True before applying.

• annotate – Run memlet propagation after application if necessary.

• permissive – Apply transformation in permissive mode.

• save – Save transformation as part of the SDFG file. Set to False if composing transformations.

• where – A dictionary of node names (from the transformation) to nodes in the SDFG or a single state.

can_be_applied(*args, **kwargs)

Returns True if this transformation can be applied on the candidate matched subgraph.

Parameters:

• graph – SDFG object in which the match was found.

• candidate – A mapping between node IDs returned from PatternTransformation.expressions and the nodes in graph.

• expr_index – The list index from PatternTransformation.expressions that was matched.

• sdfg – The SDFG in which the match was found (equal to graph).

• permissive – Whether transformation should run in permissive mode.

Returns:

True if the transformation can be applied.

classmethod expressions()

Returns a list of SDFG subgraphs that will be matched in the subgraph isomorphism phase. Used as a pre-pass before calling can_be_applied.

setup_match(*args, **kwargs)

Sets the transformation to a given subgraph pattern.

Parameters:

• cfg_id – A unique ID of the SDFG.

• state_id – The node ID of the SDFG state, if applicable. If transformation does not operate on a single state, the value should be -1.

• subgraph – A mapping between node IDs returned from PatternTransformation.expressions and the nodes in graph.

• expr_index – The list index from PatternTransformation.expressions that was matched.

• override – If True, accepts the subgraph dictionary as-is (mostly for internal use).

• options – An optional dictionary of transformation properties

Raises:

• TypeError – When transformation is not subclass of PatternTransformation.

• TypeError – When state_id is not instance of int.

• TypeError – When subgraph is not a dict of {PatternNode: int}.

state

Static field wrapper of a node or an SDFG state that designates it as part of a subgraph pattern. These objects are used in subclasses of PatternTransformation to represent the subgraph patterns.

Example use:

class MyTransformation(SingleStateTransformation):
    some_map_node = PatternNode(nodes.MapEntry)
    array = PatternNode(nodes.AccessNode)

The two nodes can then be used in the transformation static methods (e.g., expressions, can_be_applied) to represent the nodes, and in the instance methods to point to the nodes in the parent SDFG.

dace.transformation.interstate.fpga_transform_state.fpga_update(sdfg, state, depth)

dace.transformation.interstate.gpu_transform_sdfg module

Contains inter-state transformations of an SDFG to run on the GPU.

class dace.transformation.interstate.gpu_transform_sdfg.GPUTransformSDFG(*args, **kwargs)

Bases: MultiStateTransformation

Implements the GPUTransformSDFG transformation.

Transforms a whole SDFG to run on the GPU:

1. Acquire metadata about SDFG and arrays

2. Replace all non-transients with their GPU counterparts

3. Copy-in state from host to GPU

4. Copy-out state from GPU to host

5. Re-store Default-top/CPU_Heap transients as GPU_Global

6. Global tasklets are wrapped with a map of size 1

7. Global Maps are re-scheduled to use the GPU

8. Make data ready for interstate edges that use them

9. Re-apply simplification to get rid of extra states and transients

static annotates_memlets()

Indicates whether the transformation annotates the edges it creates or modifies with the appropriate memlets. This determines whether to apply memlet propagation after the transformation.

apply(*args, **kwargs)

Applies this transformation instance on the matched pattern graph.

Parameters:

sdfg – The SDFG to apply the transformation to.

Returns:

A transformation-defined return value, which could be used to pass analysis data out, or nothing.

apply_pass(*args, **kwargs)

Applies the pass to the given SDFG.

Parameters:

• sdfg – The SDFG to apply the pass to.

• pipeline_results – If in the context of a Pipeline, a dictionary that is populated with prior Pass results as {Pass subclass name: returned object from pass}. If not run in a pipeline, an empty dictionary is expected.

Returns:

Some object if pass was applied, or None if nothing changed.

apply_pattern(*args, **kwargs)

Applies this transformation on the given SDFG, using the transformation instance to find the right control flow graph object (based on control flow graph ID), and applying memlet propagation as necessary.

Parameters:

• append – If True, appends the transformation to the SDFG transformation history.

• annotate – If True, applies memlet propagation as necessary.

Returns:

A transformation-defined return value, which could be used to pass analysis data out, or nothing.

apply_to(*args, **kwargs)

Applies this transformation to a given subgraph, defined by a set of nodes. Raises an error if arguments are invalid or transformation is not applicable.

The subgraph is defined by the where dictionary, where each key is taken from the PatternNode fields of the transformation. For example, applying MapCollapse on two maps can pe performed as follows:

` MapCollapse.apply_to(sdfg, outer_map_entry=map_a, inner_map_entry=map_b) `

Parameters:

• sdfg – The SDFG to apply the transformation to.

• options – A set of parameters to use for applying the transformation.

• expr_index – The pattern expression index to try to match with.

• verify – Check that can_be_applied returns True before applying.

• annotate – Run memlet propagation after application if necessary.

• permissive – Apply transformation in permissive mode.

• save – Save transformation as part of the SDFG file. Set to False if composing transformations.

• where – A dictionary of node names (from the transformation) to nodes in the SDFG or a single state.

can_be_applied(*args, **kwargs)

Returns True if this transformation can be applied on the candidate matched subgraph.

Parameters:

• graph – SDFG object in which the match was found.

• candidate – A mapping between node IDs returned from PatternTransformation.expressions and the nodes in graph.

• expr_index – The list index from PatternTransformation.expressions that was matched.

• sdfg – The SDFG in which the match was found (equal to graph).

• permissive – Whether transformation should run in permissive mode.

Returns:

True if the transformation can be applied.

exclude_copyin

Exclude these arrays from being copied into the device (comma-separated)

exclude_copyout

Exclude these arrays from being copied out of the device (comma-separated)

exclude_tasklets

Exclude these tasklets from being processed as CPU tasklets (comma-separated)

classmethod expressions()

Returns a list of SDFG subgraphs that will be matched in the subgraph isomorphism phase. Used as a pre-pass before calling can_be_applied.

properties()

register_trans

Make all transients inside GPU maps registers

sequential_innermaps

Make all internal maps Sequential

setup_match(*args, **kwargs)

Sets the transformation to a given subgraph pattern.

Parameters:

• cfg_id – A unique ID of the SDFG.

• state_id – The node ID of the SDFG state, if applicable. If transformation does not operate on a single state, the value should be -1.

• subgraph – A mapping between node IDs returned from PatternTransformation.expressions and the nodes in graph.

• expr_index – The list index from PatternTransformation.expressions that was matched.

• override – If True, accepts the subgraph dictionary as-is (mostly for internal use).

• options – An optional dictionary of transformation properties

Raises:

• TypeError – When transformation is not subclass of PatternTransformation.

• TypeError – When state_id is not instance of int.

• TypeError – When subgraph is not a dict of {PatternNode: int}.

simplify

Reapply simplification after modifying graph

skip_scalar_tasklets

If True, does not transform tasklets that manipulate (Default-stored) scalars

toplevel_trans

Make all GPU transients top-level

dace.transformation.interstate.loop_detection module

Loop detection transformation

class dace.transformation.interstate.loop_detection.DetectLoop(*args, **kwargs)

Bases: PatternTransformation

Detects a for-loop construct from an SDFG.

apply(_, sdfg)

Applies this transformation instance on the matched pattern graph.

Parameters:

sdfg – The SDFG to apply the transformation to.

Returns:

A transformation-defined return value, which could be used to pass analysis data out, or nothing.

can_be_applied(graph, expr_index, sdfg, permissive=False)

Returns True if this transformation can be applied on the candidate matched subgraph.

Parameters:

• graph (ControlFlowRegion) – SDFGState object if this transformation is single-state, or ControlFlowRegion object otherwise.

• expr_index (int) – The list index from PatternTransformation.expressions that was matched.

• sdfg (SDFG) – If graph is an SDFGState, its parent SDFG. Otherwise should be equal to graph.

• permissive (bool) – Whether transformation should run in permissive mode.

Return type:

bool

Returns:

True if the transformation can be applied.

detect_loop(graph, multistate_loop, accept_missing_itvar=False)

Detects a loop of the form:

           ----------------
           |              v
entry -> guard -> body    exit
           ^        |
           ----------

Parameters:

• graph (ControlFlowRegion) – The graph to look for the loop.

• multistate_loop (bool) – Whether the loop contains multiple states.

Return type:

Optional[str]

Returns:

The loop variable or None if not detected.

detect_rotated_loop(graph, multistate_loop, accept_missing_itvar=False, separate_latch=False)

Detects a loop of the form:

entry -> body -> latch -> exit
           ^        |
           ----------

Parameters:

• graph (ControlFlowRegion) – The graph to look for the loop.

• multistate_loop (bool) – Whether the loop contains multiple states.

Return type:

Optional[str]

Returns:

The loop variable or None if not detected.

detect_self_loop(graph, accept_missing_itvar=False)

Detects a loop of the form:

entry -> body state -> exit
           ^    |
           ------

Parameters:

graph (ControlFlowRegion) – The graph to look for the loop.

Return type:

Optional[str]

Returns:

The loop variable or None if not detected.

entry_state

Static field wrapper of a node or an SDFG state that designates it as part of a subgraph pattern. These objects are used in subclasses of PatternTransformation to represent the subgraph patterns.

Example use:

class MyTransformation(SingleStateTransformation):
    some_map_node = PatternNode(nodes.MapEntry)
    array = PatternNode(nodes.AccessNode)

The two nodes can then be used in the transformation static methods (e.g., expressions, can_be_applied) to represent the nodes, and in the instance methods to point to the nodes in the parent SDFG.

exit_state

Static field wrapper of a node or an SDFG state that designates it as part of a subgraph pattern. These objects are used in subclasses of PatternTransformation to represent the subgraph patterns.

Example use:

class MyTransformation(SingleStateTransformation):
    some_map_node = PatternNode(nodes.MapEntry)
    array = PatternNode(nodes.AccessNode)

The two nodes can then be used in the transformation static methods (e.g., expressions, can_be_applied) to represent the nodes, and in the instance methods to point to the nodes in the parent SDFG.

classmethod expressions()

Returns a list of Graph objects that will be matched in the subgraph isomorphism phase. Used as a pre-pass before calling can_be_applied.

See:

PatternTransformation.can_be_applied

property first_loop_block: ControlFlowBlock

The first control flow block executed in each loop iteration.

property inverted: bool

Whether the loop matched a pattern of an inverted (do-while style) loop.

loop_begin

Static field wrapper of a node or an SDFG state that designates it as part of a subgraph pattern. These objects are used in subclasses of PatternTransformation to represent the subgraph patterns.

Example use:

class MyTransformation(SingleStateTransformation):
    some_map_node = PatternNode(nodes.MapEntry)
    array = PatternNode(nodes.AccessNode)

The two nodes can then be used in the transformation static methods (e.g., expressions, can_be_applied) to represent the nodes, and in the instance methods to point to the nodes in the parent SDFG.

loop_body()

Returns a list of all control flow blocks (or states) contained in the loop.

Return type:

List[ControlFlowBlock]

loop_break

Static field wrapper of a node or an SDFG state that designates it as part of a subgraph pattern. These objects are used in subclasses of PatternTransformation to represent the subgraph patterns.

Example use:

class MyTransformation(SingleStateTransformation):
    some_map_node = PatternNode(nodes.MapEntry)
    array = PatternNode(nodes.AccessNode)

The two nodes can then be used in the transformation static methods (e.g., expressions, can_be_applied) to represent the nodes, and in the instance methods to point to the nodes in the parent SDFG.

loop_condition_edge()

Returns the positive condition edge that (re-)enters the loop after the bound check.

Return type:

Edge[InterstateEdge]

loop_exit_edge()

Returns the negative condition edge that exits the loop.

Return type:

Edge[InterstateEdge]

loop_guard

Static field wrapper of a node or an SDFG state that designates it as part of a subgraph pattern. These objects are used in subclasses of PatternTransformation to represent the subgraph patterns.

Example use:

class MyTransformation(SingleStateTransformation):
    some_map_node = PatternNode(nodes.MapEntry)
    array = PatternNode(nodes.AccessNode)

The two nodes can then be used in the transformation static methods (e.g., expressions, can_be_applied) to represent the nodes, and in the instance methods to point to the nodes in the parent SDFG.

loop_increment_edge()

Returns the back-edge that increments the loop induction variable.

Return type:

Edge[InterstateEdge]

loop_information(itervar=None)

Return type:

Optional[Tuple[AnyStr, Tuple[Union[Basic, SymExpr], Union[Basic, SymExpr], Union[Basic, SymExpr]], Tuple[List[SDFGState], SDFGState]]]

loop_init_edge()

Returns the initialization edge of the loop (assignment to the beginning of the loop range).

Return type:

Edge[InterstateEdge]

loop_latch

Static field wrapper of a node or an SDFG state that designates it as part of a subgraph pattern. These objects are used in subclasses of PatternTransformation to represent the subgraph patterns.

Example use:

class MyTransformation(SingleStateTransformation):
    some_map_node = PatternNode(nodes.MapEntry)
    array = PatternNode(nodes.AccessNode)

The two nodes can then be used in the transformation static methods (e.g., expressions, can_be_applied) to represent the nodes, and in the instance methods to point to the nodes in the parent SDFG.

loop_meta_states()

Returns the non-body control-flow blocks of this loop (e.g., guard, latch).

Return type:

List[ControlFlowBlock]

dace.transformation.interstate.loop_detection.find_for_loop(graph, guard, entry, itervar=None)

Finds loop range from state machine.

Parameters:

• guard (SDFGState) – State from which the outgoing edges detect whether to exit the loop or not.

• entry (SDFGState) – First state in the loop body.

• itervar (Optional[str]) – An optional field that overrides the analyzed iteration variable.

Return type:

Optional[Tuple[AnyStr, Tuple[Union[Basic, SymExpr], Union[Basic, SymExpr], Union[Basic, SymExpr]], Tuple[List[SDFGState], SDFGState]]]

Returns:

(iteration variable, (start, end, stride), (start_states, last_loop_state)), or None if proper for-loop was not detected. end is inclusive.

dace.transformation.interstate.loop_detection.find_rotated_for_loop(graph, latch, entry, itervar=None, separate_latch=False)

Finds rotated loop range from state machine.

Parameters:

• latch (SDFGState) – State from which the outgoing edges detect whether to reenter the loop or not.

• entry (SDFGState) – First state in the loop body.

• itervar (Optional[str]) – An optional field that overrides the analyzed iteration variable.

Return type:

Optional[Tuple[AnyStr, Tuple[Union[Basic, SymExpr], Union[Basic, SymExpr], Union[Basic, SymExpr]], Tuple[List[SDFGState], SDFGState]]]

Returns:

(iteration variable, (start, end, stride), (start_states, last_loop_state)), or None if proper for-loop was not detected. end is inclusive.

dace.transformation.interstate.loop_detection.rotated_loop_find_itvar(begin_inedges, latch_inedges, backedge, latch, accept_missing_itvar=False)

Return type:

Tuple[Optional[str], Optional[Edge[InterstateEdge]]]

dace.transformation.interstate.loop_to_map module

Loop to map transformation

class dace.transformation.interstate.loop_to_map.LoopToMap(*args, **kwargs)

Bases: DetectLoop, MultiStateTransformation

Convert a control flow loop into a dataflow map. Currently only supports the simple case where there is no overlap between inputs and outputs in the body of the loop, and where the loop body only consists of a single state.

apply(*args, **kwargs)

Applies this transformation instance on the matched pattern graph.

Parameters:

sdfg – The SDFG to apply the transformation to.

Returns:

A transformation-defined return value, which could be used to pass analysis data out, or nothing.

apply_pass(*args, **kwargs)

Applies the pass to the given SDFG.

Parameters:

• sdfg – The SDFG to apply the pass to.

• pipeline_results – If in the context of a Pipeline, a dictionary that is populated with prior Pass results as {Pass subclass name: returned object from pass}. If not run in a pipeline, an empty dictionary is expected.

Returns:

Some object if pass was applied, or None if nothing changed.

apply_pattern(*args, **kwargs)

Applies this transformation on the given SDFG, using the transformation instance to find the right control flow graph object (based on control flow graph ID), and applying memlet propagation as necessary.

Parameters:

• append – If True, appends the transformation to the SDFG transformation history.

• annotate – If True, applies memlet propagation as necessary.

Returns:

A transformation-defined return value, which could be used to pass analysis data out, or nothing.

apply_to(*args, **kwargs)

Applies this transformation to a given subgraph, defined by a set of nodes. Raises an error if arguments are invalid or transformation is not applicable.

The subgraph is defined by the where dictionary, where each key is taken from the PatternNode fields of the transformation. For example, applying MapCollapse on two maps can pe performed as follows:

` MapCollapse.apply_to(sdfg, outer_map_entry=map_a, inner_map_entry=map_b) `

Parameters:

• sdfg – The SDFG to apply the transformation to.

• options – A set of parameters to use for applying the transformation.

• expr_index – The pattern expression index to try to match with.

• verify – Check that can_be_applied returns True before applying.

• annotate – Run memlet propagation after application if necessary.

• permissive – Apply transformation in permissive mode.

• save – Save transformation as part of the SDFG file. Set to False if composing transformations.

• where – A dictionary of node names (from the transformation) to nodes in the SDFG or a single state.

can_be_applied(*args, **kwargs)

Returns True if this transformation can be applied on the candidate matched subgraph.

Parameters:

• graph – SDFG object in which the match was found.

• candidate – A mapping between node IDs returned from PatternTransformation.expressions and the nodes in graph.

• expr_index – The list index from PatternTransformation.expressions that was matched.

• sdfg – The SDFG in which the match was found (equal to graph).

• permissive – Whether transformation should run in permissive mode.

Returns:

True if the transformation can be applied.

itervar

The name of the iteration variable (optional).

properties()

setup_match(*args, **kwargs)

Sets the transformation to a given subgraph pattern.

Parameters:

• cfg_id – A unique ID of the SDFG.

• state_id – The node ID of the SDFG state, if applicable. If transformation does not operate on a single state, the value should be -1.

• subgraph – A mapping between node IDs returned from PatternTransformation.expressions and the nodes in graph.

• expr_index – The list index from PatternTransformation.expressions that was matched.

• override – If True, accepts the subgraph dictionary as-is (mostly for internal use).

• options – An optional dictionary of transformation properties

Raises:

• TypeError – When transformation is not subclass of PatternTransformation.

• TypeError – When state_id is not instance of int.

• TypeError – When subgraph is not a dict of {PatternNode: int}.

test_read_memlet(sdfg, state, edge, itersym, itervar, start, end, step, write_memlets, mmlt, src_subset)

dace.transformation.interstate.move_loop_into_map module

Moves a loop around a map into the map

class dace.transformation.interstate.move_loop_into_map.MoveLoopIntoMap(*args, **kwargs)

Bases: DetectLoop, MultiStateTransformation

Moves a loop around a map into the map

apply(*args, **kwargs)

Applies this transformation instance on the matched pattern graph.

Parameters:

sdfg – The SDFG to apply the transformation to.

Returns:

A transformation-defined return value, which could be used to pass analysis data out, or nothing.

apply_pass(*args, **kwargs)

Applies the pass to the given SDFG.

Parameters:

• sdfg – The SDFG to apply the pass to.

• pipeline_results – If in the context of a Pipeline, a dictionary that is populated with prior Pass results as {Pass subclass name: returned object from pass}. If not run in a pipeline, an empty dictionary is expected.

Returns:

Some object if pass was applied, or None if nothing changed.

apply_pattern(*args, **kwargs)

Applies this transformation on the given SDFG, using the transformation instance to find the right control flow graph object (based on control flow graph ID), and applying memlet propagation as necessary.

Parameters:

• append – If True, appends the transformation to the SDFG transformation history.

• annotate – If True, applies memlet propagation as necessary.

Returns:

A transformation-defined return value, which could be used to pass analysis data out, or nothing.

apply_to(*args, **kwargs)

Applies this transformation to a given subgraph, defined by a set of nodes. Raises an error if arguments are invalid or transformation is not applicable.

The subgraph is defined by the where dictionary, where each key is taken from the PatternNode fields of the transformation. For example, applying MapCollapse on two maps can pe performed as follows:

` MapCollapse.apply_to(sdfg, outer_map_entry=map_a, inner_map_entry=map_b) `

Parameters:

• sdfg – The SDFG to apply the transformation to.

• options – A set of parameters to use for applying the transformation.

• expr_index – The pattern expression index to try to match with.

• verify – Check that can_be_applied returns True before applying.

• annotate – Run memlet propagation after application if necessary.

• permissive – Apply transformation in permissive mode.

• save – Save transformation as part of the SDFG file. Set to False if composing transformations.

• where – A dictionary of node names (from the transformation) to nodes in the SDFG or a single state.

can_be_applied(*args, **kwargs)

Returns True if this transformation can be applied on the candidate matched subgraph.

Parameters:

• graph – SDFG object in which the match was found.

• candidate – A mapping between node IDs returned from PatternTransformation.expressions and the nodes in graph.

• expr_index – The list index from PatternTransformation.expressions that was matched.

• sdfg – The SDFG in which the match was found (equal to graph).

• permissive – Whether transformation should run in permissive mode.

Returns:

True if the transformation can be applied.

setup_match(*args, **kwargs)

Sets the transformation to a given subgraph pattern.

Parameters:

• cfg_id – A unique ID of the SDFG.

• state_id – The node ID of the SDFG state, if applicable. If transformation does not operate on a single state, the value should be -1.

• subgraph – A mapping between node IDs returned from PatternTransformation.expressions and the nodes in graph.

• expr_index – The list index from PatternTransformation.expressions that was matched.

• override – If True, accepts the subgraph dictionary as-is (mostly for internal use).

• options – An optional dictionary of transformation properties

Raises:

• TypeError – When transformation is not subclass of PatternTransformation.

• TypeError – When state_id is not instance of int.

• TypeError – When subgraph is not a dict of {PatternNode: int}.

dace.transformation.interstate.move_loop_into_map.fold(memlet_subset_ranges, itervar, lower, upper)

dace.transformation.interstate.move_loop_into_map.offset(memlet_subset_ranges, value)

dace.transformation.interstate.loop_peeling module

Loop unroll transformation

class dace.transformation.interstate.loop_peeling.LoopPeeling(*args, **kwargs)

Bases: LoopUnroll

Splits the first count iterations of a state machine for-loop into multiple, separate states.

apply(graph, sdfg)

Applies this transformation instance on the matched pattern graph.

Parameters:

sdfg (SDFG) – The SDFG to apply the transformation to.

Returns:

A transformation-defined return value, which could be used to pass analysis data out, or nothing.

begin

If True, peels loop from beginning (first count iterations), otherwise peels last count iterations.

can_be_applied(graph, expr_index, sdfg, permissive=False)

Returns True if this transformation can be applied on the candidate matched subgraph.

Parameters:

• graph – SDFG object in which the match was found.

• candidate – A mapping between node IDs returned from PatternTransformation.expressions and the nodes in graph.

• expr_index – The list index from PatternTransformation.expressions that was matched.

• sdfg – The SDFG in which the match was found (equal to graph).

• permissive – Whether transformation should run in permissive mode.

Returns:

True if the transformation can be applied.

properties()

dace.transformation.interstate.loop_unroll module

Loop unroll transformation

class dace.transformation.interstate.loop_unroll.LoopUnroll(*args, **kwargs)

Bases: DetectLoop, MultiStateTransformation

Unrolls a state machine for-loop into multiple states

apply(graph, sdfg)

Applies this transformation instance on the matched pattern graph.

Parameters:

sdfg – The SDFG to apply the transformation to.

Returns:

A transformation-defined return value, which could be used to pass analysis data out, or nothing.

can_be_applied(graph, expr_index, sdfg, permissive=False)

Returns True if this transformation can be applied on the candidate matched subgraph.

Parameters:

• graph – SDFG object in which the match was found.

• candidate – A mapping between node IDs returned from PatternTransformation.expressions and the nodes in graph.

• expr_index – The list index from PatternTransformation.expressions that was matched.

• sdfg – The SDFG in which the match was found (equal to graph).

• permissive – Whether transformation should run in permissive mode.

Returns:

True if the transformation can be applied.

count

Number of iterations to unroll, or zero for all iterations (loop must be constant-sized for 0)

instantiate_loop(sdfg, loop_states, loop_subgraph, itervar, value, state_suffix=None)

properties()

dace.transformation.interstate.sdfg_nesting module

SDFG nesting transformation.

class dace.transformation.interstate.sdfg_nesting.ASTRefiner(to_refine, refine_subset, sdfg, indices=None)

Bases: NodeTransformer

Python AST transformer used in RefineNestedAccess to reduce (refine) the subscript ranges based on the specification given in the transformation.

visit_Subscript(node)

Return type:

Subscript

class dace.transformation.interstate.sdfg_nesting.InlineSDFG(*args, **kwargs)

Bases: SingleStateTransformation

Inlines a single-state nested SDFG into a top-level SDFG.

In particular, the steps taken are:

1. All transient arrays become transients of the parent

2. If a source/sink node is one of the inputs/outputs:

   1. Remove it

   2. Reconnect through external edges (map/accessnode)

   3. Replace and reoffset memlets with external data descriptor

3. If other nodes carry the names of inputs/outputs:

   1. Replace data with external data descriptor

   2. Replace and reoffset memlets with external data descriptor

4. If source/sink node is not connected to a source/destination, and the nested SDFG is in a scope, connect to scope with empty memlets

5. Remove all unused external inputs/output memlet paths

6. Remove isolated nodes resulting from previous step

static annotates_memlets()

Indicates whether the transformation annotates the edges it creates or modifies with the appropriate memlets. This determines whether to apply memlet propagation after the transformation.

apply(*args, **kwargs)

Applies this transformation instance on the matched pattern graph.

Parameters:

sdfg – The SDFG to apply the transformation to.

Returns:

A transformation-defined return value, which could be used to pass analysis data out, or nothing.

apply_pass(*args, **kwargs)

Applies the pass to the given SDFG.

Parameters:

• sdfg – The SDFG to apply the pass to.

• pipeline_results – If in the context of a Pipeline, a dictionary that is populated with prior Pass results as {Pass subclass name: returned object from pass}. If not run in a pipeline, an empty dictionary is expected.

Returns:

Some object if pass was applied, or None if nothing changed.

apply_pattern(*args, **kwargs)

Applies this transformation on the given SDFG, using the transformation instance to find the right control flow graph object (based on control flow graph ID), and applying memlet propagation as necessary.

Parameters:

• append – If True, appends the transformation to the SDFG transformation history.

• annotate – If True, applies memlet propagation as necessary.

Returns:

A transformation-defined return value, which could be used to pass analysis data out, or nothing.

apply_to(*args, **kwargs)

Applies this transformation to a given subgraph, defined by a set of nodes. Raises an error if arguments are invalid or transformation is not applicable.

The subgraph is defined by the where dictionary, where each key is taken from the PatternNode fields of the transformation. For example, applying MapCollapse on two maps can pe performed as follows:

` MapCollapse.apply_to(sdfg, outer_map_entry=map_a, inner_map_entry=map_b) `

Parameters:

• sdfg – The SDFG to apply the transformation to.

• options – A set of parameters to use for applying the transformation.

• expr_index – The pattern expression index to try to match with.

• verify – Check that can_be_applied returns True before applying.

• annotate – Run memlet propagation after application if necessary.

• permissive – Apply transformation in permissive mode.

• save – Save transformation as part of the SDFG file. Set to False if composing transformations.

• where – A dictionary of node names (from the transformation) to nodes in the SDFG or a single state.

can_be_applied(*args, **kwargs)

Returns True if this transformation can be applied on the candidate matched subgraph.

Parameters:

• graph – SDFGState object in which the match was found.

• candidate – A mapping between node IDs returned from PatternTransformation.expressions and the nodes in graph.

• expr_index – The list index from PatternTransformation.expressions that was matched.

• sdfg – The parent SDFG of the matched state.

• permissive – Whether transformation should run in permissive mode.

Returns:

True if the transformation can be applied.

classmethod expressions()

Returns a list of SDFG state subgraphs that will be matched in the subgraph isomorphism phase. Used as a pre-pass before calling can_be_applied.

nested_sdfg

Static field wrapper of a node or an SDFG state that designates it as part of a subgraph pattern. These objects are used in subclasses of PatternTransformation to represent the subgraph patterns.

Example use:

class MyTransformation(SingleStateTransformation):
    some_map_node = PatternNode(nodes.MapEntry)
    array = PatternNode(nodes.AccessNode)

The two nodes can then be used in the transformation static methods (e.g., expressions, can_be_applied) to represent the nodes, and in the instance methods to point to the nodes in the parent SDFG.

properties()

setup_match(*args, **kwargs)

Sets the transformation to a given subgraph pattern.

Parameters:

• cfg_id – A unique ID of the SDFG.

• state_id – The node ID of the SDFG state, if applicable. If transformation does not operate on a single state, the value should be -1.

• subgraph – A mapping between node IDs returned from PatternTransformation.expressions and the nodes in graph.

• expr_index – The list index from PatternTransformation.expressions that was matched.

• override – If True, accepts the subgraph dictionary as-is (mostly for internal use).

• options – An optional dictionary of transformation properties

Raises:

• TypeError – When transformation is not subclass of PatternTransformation.

• TypeError – When state_id is not instance of int.

• TypeError – When subgraph is not a dict of {PatternNode: int}.

class dace.transformation.interstate.sdfg_nesting.InlineTransients(*args, **kwargs)

Bases: SingleStateTransformation

Inlines all transient arrays that are not used anywhere else into a nested SDFG.

static annotates_memlets()

Indicates whether the transformation annotates the edges it creates or modifies with the appropriate memlets. This determines whether to apply memlet propagation after the transformation.

apply(*args, **kwargs)

Applies this transformation instance on the matched pattern graph.

Parameters:

sdfg – The SDFG to apply the transformation to.

Returns:

A transformation-defined return value, which could be used to pass analysis data out, or nothing.

apply_pass(*args, **kwargs)

Applies the pass to the given SDFG.

Parameters:

• sdfg – The SDFG to apply the pass to.

• pipeline_results – If in the context of a Pipeline, a dictionary that is populated with prior Pass results as {Pass subclass name: returned object from pass}. If not run in a pipeline, an empty dictionary is expected.

Returns:

Some object if pass was applied, or None if nothing changed.

apply_pattern(*args, **kwargs)

Applies this transformation on the given SDFG, using the transformation instance to find the right control flow graph object (based on control flow graph ID), and applying memlet propagation as necessary.

Parameters:

• append – If True, appends the transformation to the SDFG transformation history.

• annotate – If True, applies memlet propagation as necessary.

Returns:

A transformation-defined return value, which could be used to pass analysis data out, or nothing.

apply_to(*args, **kwargs)

Applies this transformation to a given subgraph, defined by a set of nodes. Raises an error if arguments are invalid or transformation is not applicable.

The subgraph is defined by the where dictionary, where each key is taken from the PatternNode fields of the transformation. For example, applying MapCollapse on two maps can pe performed as follows:

` MapCollapse.apply_to(sdfg, outer_map_entry=map_a, inner_map_entry=map_b) `

Parameters:

• sdfg – The SDFG to apply the transformation to.

• options – A set of parameters to use for applying the transformation.

• expr_index – The pattern expression index to try to match with.

• verify – Check that can_be_applied returns True before applying.

• annotate – Run memlet propagation after application if necessary.

• permissive – Apply transformation in permissive mode.

• save – Save transformation as part of the SDFG file. Set to False if composing transformations.

• where – A dictionary of node names (from the transformation) to nodes in the SDFG or a single state.

can_be_applied(*args, **kwargs)

Returns True if this transformation can be applied on the candidate matched subgraph.

Parameters:

• graph – SDFGState object in which the match was found.

• candidate – A mapping between node IDs returned from PatternTransformation.expressions and the nodes in graph.

• expr_index – The list index from PatternTransformation.expressions that was matched.

• sdfg – The parent SDFG of the matched state.

• permissive – Whether transformation should run in permissive mode.

Returns:

True if the transformation can be applied.

classmethod expressions()

Returns a list of SDFG state subgraphs that will be matched in the subgraph isomorphism phase. Used as a pre-pass before calling can_be_applied.

nsdfg

Static field wrapper of a node or an SDFG state that designates it as part of a subgraph pattern. These objects are used in subclasses of PatternTransformation to represent the subgraph patterns.

Example use:

class MyTransformation(SingleStateTransformation):
    some_map_node = PatternNode(nodes.MapEntry)
    array = PatternNode(nodes.AccessNode)

The two nodes can then be used in the transformation static methods (e.g., expressions, can_be_applied) to represent the nodes, and in the instance methods to point to the nodes in the parent SDFG.

properties()

setup_match(*args, **kwargs)

Sets the transformation to a given subgraph pattern.

Parameters:

• cfg_id – A unique ID of the SDFG.

• state_id – The node ID of the SDFG state, if applicable. If transformation does not operate on a single state, the value should be -1.

• subgraph – A mapping between node IDs returned from PatternTransformation.expressions and the nodes in graph.

• expr_index – The list index from PatternTransformation.expressions that was matched.

• override – If True, accepts the subgraph dictionary as-is (mostly for internal use).

• options – An optional dictionary of transformation properties

Raises:

• TypeError – When transformation is not subclass of PatternTransformation.

• TypeError – When state_id is not instance of int.

• TypeError – When subgraph is not a dict of {PatternNode: int}.

class dace.transformation.interstate.sdfg_nesting.NestSDFG(*args, **kwargs)

Bases: MultiStateTransformation

Implements SDFG Nesting, taking an SDFG as an input and creating a nested SDFG node from it.

static annotates_memlets()

Indicates whether the transformation annotates the edges it creates or modifies with the appropriate memlets. This determines whether to apply memlet propagation after the transformation.

apply(*args, **kwargs)

Applies this transformation instance on the matched pattern graph.

Parameters:

sdfg – The SDFG to apply the transformation to.

Returns:

A transformation-defined return value, which could be used to pass analysis data out, or nothing.

apply_pass(*args, **kwargs)

Applies the pass to the given SDFG.

Parameters:

• sdfg – The SDFG to apply the pass to.

• pipeline_results – If in the context of a Pipeline, a dictionary that is populated with prior Pass results as {Pass subclass name: returned object from pass}. If not run in a pipeline, an empty dictionary is expected.

Returns:

Some object if pass was applied, or None if nothing changed.

apply_pattern(*args, **kwargs)

Applies this transformation on the given SDFG, using the transformation instance to find the right control flow graph object (based on control flow graph ID), and applying memlet propagation as necessary.

Parameters:

• append – If True, appends the transformation to the SDFG transformation history.

• annotate – If True, applies memlet propagation as necessary.

Returns:

A transformation-defined return value, which could be used to pass analysis data out, or nothing.

apply_to(*args, **kwargs)

Applies this transformation to a given subgraph, defined by a set of nodes. Raises an error if arguments are invalid or transformation is not applicable.

The subgraph is defined by the where dictionary, where each key is taken from the PatternNode fields of the transformation. For example, applying MapCollapse on two maps can pe performed as follows:

` MapCollapse.apply_to(sdfg, outer_map_entry=map_a, inner_map_entry=map_b) `

Parameters:

• sdfg – The SDFG to apply the transformation to.

• options – A set of parameters to use for applying the transformation.

• expr_index – The pattern expression index to try to match with.

• verify – Check that can_be_applied returns True before applying.

• annotate – Run memlet propagation after application if necessary.

• permissive – Apply transformation in permissive mode.

• save – Save transformation as part of the SDFG file. Set to False if composing transformations.

• where – A dictionary of node names (from the transformation) to nodes in the SDFG or a single state.

can_be_applied(*args, **kwargs)

Returns True if this transformation can be applied on the candidate matched subgraph.

Parameters:

• graph – SDFG object in which the match was found.

• candidate – A mapping between node IDs returned from PatternTransformation.expressions and the nodes in graph.

• expr_index – The list index from PatternTransformation.expressions that was matched.

• sdfg – The SDFG in which the match was found (equal to graph).

• permissive – Whether transformation should run in permissive mode.

Returns:

True if the transformation can be applied.

classmethod expressions()

Returns a list of SDFG subgraphs that will be matched in the subgraph isomorphism phase. Used as a pre-pass before calling can_be_applied.

promote_global_trans

Promotes transients to be allocated once

properties()

setup_match(*args, **kwargs)

Sets the transformation to a given subgraph pattern.

Parameters:

• cfg_id – A unique ID of the SDFG.

• state_id – The node ID of the SDFG state, if applicable. If transformation does not operate on a single state, the value should be -1.

• subgraph – A mapping between node IDs returned from PatternTransformation.expressions and the nodes in graph.

• expr_index – The list index from PatternTransformation.expressions that was matched.

• override – If True, accepts the subgraph dictionary as-is (mostly for internal use).

• options – An optional dictionary of transformation properties

Raises:

• TypeError – When transformation is not subclass of PatternTransformation.

• TypeError – When state_id is not instance of int.

• TypeError – When subgraph is not a dict of {PatternNode: int}.

class dace.transformation.interstate.sdfg_nesting.RefineNestedAccess(*args, **kwargs)

Bases: SingleStateTransformation

Reduces memlet shape when a memlet is connected to a nested SDFG, but not using all of the contents. Makes the outer memlet smaller in shape and ensures that the offsets in the nested SDFG start with zero. This helps with subsequent transformations on the outer SDFGs.

For example, in the following program:

@dace.program
def func_a(y):
    return y[1:5] + 1

@dace.program
def main(x: dace.float32[N]):
    return func_a(x)

The memlet pointing to func_a will contain all of x (x[0:N]), and it is offset to y[1:5] in the function, with y’s size being N. After the transformation, the memlet connected to the nested SDFG of func_a would contain x[1:5] directly and the internal y array would have a size of 4, accessed as y[0:4].

static annotates_memlets()

Indicates whether the transformation annotates the edges it creates or modifies with the appropriate memlets. This determines whether to apply memlet propagation after the transformation.

apply(*args, **kwargs)

Applies this transformation instance on the matched pattern graph.

Parameters:

sdfg – The SDFG to apply the transformation to.

Returns:

A transformation-defined return value, which could be used to pass analysis data out, or nothing.

apply_pass(*args, **kwargs)

Applies the pass to the given SDFG.

Parameters:

• sdfg – The SDFG to apply the pass to.

• pipeline_results – If in the context of a Pipeline, a dictionary that is populated with prior Pass results as {Pass subclass name: returned object from pass}. If not run in a pipeline, an empty dictionary is expected.

Returns:

Some object if pass was applied, or None if nothing changed.

apply_pattern(*args, **kwargs)

Applies this transformation on the given SDFG, using the transformation instance to find the right control flow graph object (based on control flow graph ID), and applying memlet propagation as necessary.

Parameters:

• append – If True, appends the transformation to the SDFG transformation history.

• annotate – If True, applies memlet propagation as necessary.

Returns:

A transformation-defined return value, which could be used to pass analysis data out, or nothing.

apply_to(*args, **kwargs)

Applies this transformation to a given subgraph, defined by a set of nodes. Raises an error if arguments are invalid or transformation is not applicable.

The subgraph is defined by the where dictionary, where each key is taken from the PatternNode fields of the transformation. For example, applying MapCollapse on two maps can pe performed as follows:

` MapCollapse.apply_to(sdfg, outer_map_entry=map_a, inner_map_entry=map_b) `

Parameters:

• sdfg – The SDFG to apply the transformation to.

• options – A set of parameters to use for applying the transformation.

• expr_index – The pattern expression index to try to match with.

• verify – Check that can_be_applied returns True before applying.

• annotate – Run memlet propagation after application if necessary.

• permissive – Apply transformation in permissive mode.

• save – Save transformation as part of the SDFG file. Set to False if composing transformations.

• where – A dictionary of node names (from the transformation) to nodes in the SDFG or a single state.

can_be_applied(*args, **kwargs)

Returns True if this transformation can be applied on the candidate matched subgraph.

Parameters:

• graph – SDFGState object in which the match was found.

• candidate – A mapping between node IDs returned from PatternTransformation.expressions and the nodes in graph.

• expr_index – The list index from PatternTransformation.expressions that was matched.

• sdfg – The parent SDFG of the matched state.

• permissive – Whether transformation should run in permissive mode.

Returns:

True if the transformation can be applied.

classmethod expressions()

Returns a list of SDFG state subgraphs that will be matched in the subgraph isomorphism phase. Used as a pre-pass before calling can_be_applied.

nsdfg

Static field wrapper of a node or an SDFG state that designates it as part of a subgraph pattern. These objects are used in subclasses of PatternTransformation to represent the subgraph patterns.

Example use:

class MyTransformation(SingleStateTransformation):
    some_map_node = PatternNode(nodes.MapEntry)
    array = PatternNode(nodes.AccessNode)

The two nodes can then be used in the transformation static methods (e.g., expressions, can_be_applied) to represent the nodes, and in the instance methods to point to the nodes in the parent SDFG.

properties()

setup_match(*args, **kwargs)

Sets the transformation to a given subgraph pattern.

Parameters:

• cfg_id – A unique ID of the SDFG.

• state_id – The node ID of the SDFG state, if applicable. If transformation does not operate on a single state, the value should be -1.

• subgraph – A mapping between node IDs returned from PatternTransformation.expressions and the nodes in graph.

• expr_index – The list index from PatternTransformation.expressions that was matched.

• override – If True, accepts the subgraph dictionary as-is (mostly for internal use).

• options – An optional dictionary of transformation properties

Raises:

• TypeError – When transformation is not subclass of PatternTransformation.

• TypeError – When state_id is not instance of int.

• TypeError – When subgraph is not a dict of {PatternNode: int}.

dace.transformation.interstate.state_elimination module

State elimination transformations

class dace.transformation.interstate.state_elimination.EndStateElimination(*args, **kwargs)

Bases: MultiStateTransformation

End-state elimination removes a redundant state that has one incoming edge and no contents.

apply(graph, sdfg)

Applies this transformation instance on the matched pattern graph.

Parameters:

sdfg – The SDFG to apply the transformation to.

Returns:

A transformation-defined return value, which could be used to pass analysis data out, or nothing.

can_be_applied(graph, expr_index, sdfg, permissive=False)

Returns True if this transformation can be applied on the candidate matched subgraph.

Parameters:

• graph – SDFG object in which the match was found.

• candidate – A mapping between node IDs returned from PatternTransformation.expressions and the nodes in graph.

• expr_index – The list index from PatternTransformation.expressions that was matched.

• sdfg – The SDFG in which the match was found (equal to graph).

• permissive – Whether transformation should run in permissive mode.

Returns:

True if the transformation can be applied.

end_state

Static field wrapper of a node or an SDFG state that designates it as part of a subgraph pattern. These objects are used in subclasses of PatternTransformation to represent the subgraph patterns.

Example use:

class MyTransformation(SingleStateTransformation):
    some_map_node = PatternNode(nodes.MapEntry)
    array = PatternNode(nodes.AccessNode)

The two nodes can then be used in the transformation static methods (e.g., expressions, can_be_applied) to represent the nodes, and in the instance methods to point to the nodes in the parent SDFG.

classmethod expressions()

Returns a list of SDFG subgraphs that will be matched in the subgraph isomorphism phase. Used as a pre-pass before calling can_be_applied.

class dace.transformation.interstate.state_elimination.FalseConditionElimination(*args, **kwargs)

Bases: MultiStateTransformation

If a state transition condition is always false, removes edge.

apply(graph, sdfg)

Applies this transformation instance on the matched pattern graph.

Parameters:

sdfg (SDFG) – The SDFG to apply the transformation to.

Returns:

A transformation-defined return value, which could be used to pass analysis data out, or nothing.

can_be_applied(graph, expr_index, sdfg, permissive=False)

Returns True if this transformation can be applied on the candidate matched subgraph.

Parameters:

• graph (SDFG) – SDFG object in which the match was found.

• candidate – A mapping between node IDs returned from PatternTransformation.expressions and the nodes in graph.

• expr_index – The list index from PatternTransformation.expressions that was matched.

• sdfg (SDFG) – The SDFG in which the match was found (equal to graph).

• permissive – Whether transformation should run in permissive mode.

Returns:

True if the transformation can be applied.

classmethod expressions()

Returns a list of SDFG subgraphs that will be matched in the subgraph isomorphism phase. Used as a pre-pass before calling can_be_applied.

state_a

Static field wrapper of a node or an SDFG state that designates it as part of a subgraph pattern. These objects are used in subclasses of PatternTransformation to represent the subgraph patterns.

Example use:

class MyTransformation(SingleStateTransformation):
    some_map_node = PatternNode(nodes.MapEntry)
    array = PatternNode(nodes.AccessNode)

The two nodes can then be used in the transformation static methods (e.g., expressions, can_be_applied) to represent the nodes, and in the instance methods to point to the nodes in the parent SDFG.

state_b

Static field wrapper of a node or an SDFG state that designates it as part of a subgraph pattern. These objects are used in subclasses of PatternTransformation to represent the subgraph patterns.

Example use:

class MyTransformation(SingleStateTransformation):
    some_map_node = PatternNode(nodes.MapEntry)
    array = PatternNode(nodes.AccessNode)

The two nodes can then be used in the transformation static methods (e.g., expressions, can_be_applied) to represent the nodes, and in the instance methods to point to the nodes in the parent SDFG.

class dace.transformation.interstate.state_elimination.HoistState(*args, **kwargs)

Bases: SingleStateTransformation

Move a state out of a nested SDFG

apply(*args, **kwargs)

Applies this transformation instance on the matched pattern graph.

Parameters:

sdfg – The SDFG to apply the transformation to.

Returns:

A transformation-defined return value, which could be used to pass analysis data out, or nothing.

apply_pass(*args, **kwargs)

Applies the pass to the given SDFG.

Parameters:

• sdfg – The SDFG to apply the pass to.

• pipeline_results – If in the context of a Pipeline, a dictionary that is populated with prior Pass results as {Pass subclass name: returned object from pass}. If not run in a pipeline, an empty dictionary is expected.

Returns:

Some object if pass was applied, or None if nothing changed.

apply_pattern(*args, **kwargs)

Applies this transformation on the given SDFG, using the transformation instance to find the right control flow graph object (based on control flow graph ID), and applying memlet propagation as necessary.

Parameters:

• append – If True, appends the transformation to the SDFG transformation history.

• annotate – If True, applies memlet propagation as necessary.

Returns:

A transformation-defined return value, which could be used to pass analysis data out, or nothing.

apply_to(*args, **kwargs)

Applies this transformation to a given subgraph, defined by a set of nodes. Raises an error if arguments are invalid or transformation is not applicable.

The subgraph is defined by the where dictionary, where each key is taken from the PatternNode fields of the transformation. For example, applying MapCollapse on two maps can pe performed as follows:

` MapCollapse.apply_to(sdfg, outer_map_entry=map_a, inner_map_entry=map_b) `

Parameters:

• sdfg – The SDFG to apply the transformation to.

• options – A set of parameters to use for applying the transformation.

• expr_index – The pattern expression index to try to match with.

• verify – Check that can_be_applied returns True before applying.

• annotate – Run memlet propagation after application if necessary.

• permissive – Apply transformation in permissive mode.

• save – Save transformation as part of the SDFG file. Set to False if composing transformations.

• where – A dictionary of node names (from the transformation) to nodes in the SDFG or a single state.

can_be_applied(*args, **kwargs)

Returns True if this transformation can be applied on the candidate matched subgraph.

Parameters:

• graph – SDFGState object in which the match was found.

• candidate – A mapping between node IDs returned from PatternTransformation.expressions and the nodes in graph.

• expr_index – The list index from PatternTransformation.expressions that was matched.

• sdfg – The parent SDFG of the matched state.

• permissive – Whether transformation should run in permissive mode.

Returns:

True if the transformation can be applied.

classmethod expressions()

Returns a list of SDFG state subgraphs that will be matched in the subgraph isomorphism phase. Used as a pre-pass before calling can_be_applied.

nsdfg

Static field wrapper of a node or an SDFG state that designates it as part of a subgraph pattern. These objects are used in subclasses of PatternTransformation to represent the subgraph patterns.

Example use:

class MyTransformation(SingleStateTransformation):
    some_map_node = PatternNode(nodes.MapEntry)
    array = PatternNode(nodes.AccessNode)

The two nodes can then be used in the transformation static methods (e.g., expressions, can_be_applied) to represent the nodes, and in the instance methods to point to the nodes in the parent SDFG.

setup_match(*args, **kwargs)

Sets the transformation to a given subgraph pattern.

Parameters:

• cfg_id – A unique ID of the SDFG.

• state_id – The node ID of the SDFG state, if applicable. If transformation does not operate on a single state, the value should be -1.

• subgraph – A mapping between node IDs returned from PatternTransformation.expressions and the nodes in graph.

• expr_index – The list index from PatternTransformation.expressions that was matched.

• override – If True, accepts the subgraph dictionary as-is (mostly for internal use).

• options – An optional dictionary of transformation properties

Raises:

• TypeError – When transformation is not subclass of PatternTransformation.

• TypeError – When state_id is not instance of int.

• TypeError – When subgraph is not a dict of {PatternNode: int}.

class dace.transformation.interstate.state_elimination.StartStateElimination(*args, **kwargs)

Bases: MultiStateTransformation

Start-state elimination removes a redundant state that has one outgoing edge and no contents. This transformation applies only to nested SDFGs.

apply(graph, sdfg)

Applies this transformation instance on the matched pattern graph.

Parameters:

sdfg – The SDFG to apply the transformation to.

Returns:

A transformation-defined return value, which could be used to pass analysis data out, or nothing.

can_be_applied(graph, expr_index, sdfg, permissive=False)

Returns True if this transformation can be applied on the candidate matched subgraph.

Parameters:

• graph – SDFG object in which the match was found.

• candidate – A mapping between node IDs returned from PatternTransformation.expressions and the nodes in graph.

• expr_index – The list index from PatternTransformation.expressions that was matched.

• sdfg – The SDFG in which the match was found (equal to graph).

• permissive – Whether transformation should run in permissive mode.

Returns:

True if the transformation can be applied.

classmethod expressions()

Returns a list of SDFG subgraphs that will be matched in the subgraph isomorphism phase. Used as a pre-pass before calling can_be_applied.

start_state

Static field wrapper of a node or an SDFG state that designates it as part of a subgraph pattern. These objects are used in subclasses of PatternTransformation to represent the subgraph patterns.

Example use:

class MyTransformation(SingleStateTransformation):
    some_map_node = PatternNode(nodes.MapEntry)
    array = PatternNode(nodes.AccessNode)

The two nodes can then be used in the transformation static methods (e.g., expressions, can_be_applied) to represent the nodes, and in the instance methods to point to the nodes in the parent SDFG.

class dace.transformation.interstate.state_elimination.StateAssignElimination(*args, **kwargs)

Bases: MultiStateTransformation

State assign elimination removes all assignments into the final state and subsumes the assigned value into its contents.

apply(graph, sdfg)

Applies this transformation instance on the matched pattern graph.

Parameters:

sdfg – The SDFG to apply the transformation to.

Returns:

A transformation-defined return value, which could be used to pass analysis data out, or nothing.

can_be_applied(graph, expr_index, sdfg, permissive=False)

Returns True if this transformation can be applied on the candidate matched subgraph.

Parameters:

• graph – SDFG object in which the match was found.

• candidate – A mapping between node IDs returned from PatternTransformation.expressions and the nodes in graph.

• expr_index – The list index from PatternTransformation.expressions that was matched.

• sdfg – The SDFG in which the match was found (equal to graph).

• permissive – Whether transformation should run in permissive mode.

Returns:

True if the transformation can be applied.

end_state

Static field wrapper of a node or an SDFG state that designates it as part of a subgraph pattern. These objects are used in subclasses of PatternTransformation to represent the subgraph patterns.

Example use:

class MyTransformation(SingleStateTransformation):
    some_map_node = PatternNode(nodes.MapEntry)
    array = PatternNode(nodes.AccessNode)

The two nodes can then be used in the transformation static methods (e.g., expressions, can_be_applied) to represent the nodes, and in the instance methods to point to the nodes in the parent SDFG.

classmethod expressions()

Returns a list of SDFG subgraphs that will be matched in the subgraph isomorphism phase. Used as a pre-pass before calling can_be_applied.

class dace.transformation.interstate.state_elimination.SymbolAliasPromotion(*args, **kwargs)

Bases: MultiStateTransformation

SymbolAliasPromotion moves inter-state assignments that create symbolic aliases to the previous inter-state edge according to the topological order. The purpose of this transformation is to iteratively move symbolic aliases together, so that true duplicates can be easily removed.

apply(*args, **kwargs)

Applies this transformation instance on the matched pattern graph.

Parameters:

sdfg – The SDFG to apply the transformation to.

Returns:

A transformation-defined return value, which could be used to pass analysis data out, or nothing.

apply_pass(*args, **kwargs)

Applies the pass to the given SDFG.

Parameters:

• sdfg – The SDFG to apply the pass to.

• pipeline_results – If in the context of a Pipeline, a dictionary that is populated with prior Pass results as {Pass subclass name: returned object from pass}. If not run in a pipeline, an empty dictionary is expected.

Returns:

Some object if pass was applied, or None if nothing changed.

apply_pattern(*args, **kwargs)

Applies this transformation on the given SDFG, using the transformation instance to find the right control flow graph object (based on control flow graph ID), and applying memlet propagation as necessary.

Parameters:

• append – If True, appends the transformation to the SDFG transformation history.

• annotate – If True, applies memlet propagation as necessary.

Returns:

A transformation-defined return value, which could be used to pass analysis data out, or nothing.

apply_to(*args, **kwargs)

Applies this transformation to a given subgraph, defined by a set of nodes. Raises an error if arguments are invalid or transformation is not applicable.

The subgraph is defined by the where dictionary, where each key is taken from the PatternNode fields of the transformation. For example, applying MapCollapse on two maps can pe performed as follows:

` MapCollapse.apply_to(sdfg, outer_map_entry=map_a, inner_map_entry=map_b) `

Parameters:

• sdfg – The SDFG to apply the transformation to.

• options – A set of parameters to use for applying the transformation.

• expr_index – The pattern expression index to try to match with.

• verify – Check that can_be_applied returns True before applying.

• annotate – Run memlet propagation after application if necessary.

• permissive – Apply transformation in permissive mode.

• save – Save transformation as part of the SDFG file. Set to False if composing transformations.

• where – A dictionary of node names (from the transformation) to nodes in the SDFG or a single state.

can_be_applied(*args, **kwargs)

Returns True if this transformation can be applied on the candidate matched subgraph.

Parameters:

• graph – SDFG object in which the match was found.

• candidate – A mapping between node IDs returned from PatternTransformation.expressions and the nodes in graph.

• expr_index – The list index from PatternTransformation.expressions that was matched.

• sdfg – The SDFG in which the match was found (equal to graph).

• permissive – Whether transformation should run in permissive mode.

Returns:

True if the transformation can be applied.

classmethod expressions()

Returns a list of SDFG subgraphs that will be matched in the subgraph isomorphism phase. Used as a pre-pass before calling can_be_applied.

first_state

Static field wrapper of a node or an SDFG state that designates it as part of a subgraph pattern. These objects are used in subclasses of PatternTransformation to represent the subgraph patterns.

Example use:

class MyTransformation(SingleStateTransformation):
    some_map_node = PatternNode(nodes.MapEntry)
    array = PatternNode(nodes.AccessNode)

The two nodes can then be used in the transformation static methods (e.g., expressions, can_be_applied) to represent the nodes, and in the instance methods to point to the nodes in the parent SDFG.

second_state

Static field wrapper of a node or an SDFG state that designates it as part of a subgraph pattern. These objects are used in subclasses of PatternTransformation to represent the subgraph patterns.

Example use:

class MyTransformation(SingleStateTransformation):
    some_map_node = PatternNode(nodes.MapEntry)
    array = PatternNode(nodes.AccessNode)

The two nodes can then be used in the transformation static methods (e.g., expressions, can_be_applied) to represent the nodes, and in the instance methods to point to the nodes in the parent SDFG.

setup_match(*args, **kwargs)

Sets the transformation to a given subgraph pattern.

Parameters:

• cfg_id – A unique ID of the SDFG.

• state_id – The node ID of the SDFG state, if applicable. If transformation does not operate on a single state, the value should be -1.

• subgraph – A mapping between node IDs returned from PatternTransformation.expressions and the nodes in graph.

• expr_index – The list index from PatternTransformation.expressions that was matched.

• override – If True, accepts the subgraph dictionary as-is (mostly for internal use).

• options – An optional dictionary of transformation properties

Raises:

• TypeError – When transformation is not subclass of PatternTransformation.

• TypeError – When state_id is not instance of int.

• TypeError – When subgraph is not a dict of {PatternNode: int}.

class dace.transformation.interstate.state_elimination.TrueConditionElimination(*args, **kwargs)

Bases: MultiStateTransformation

If a state transition condition is always true, removes condition from edge.

apply(graph, sdfg)

Applies this transformation instance on the matched pattern graph.

Parameters:

sdfg (SDFG) – The SDFG to apply the transformation to.

Returns:

A transformation-defined return value, which could be used to pass analysis data out, or nothing.

can_be_applied(graph, expr_index, sdfg, permissive=False)

Returns True if this transformation can be applied on the candidate matched subgraph.

Parameters:

• graph (SDFG) – SDFG object in which the match was found.

• candidate – A mapping between node IDs returned from PatternTransformation.expressions and the nodes in graph.

• expr_index – The list index from PatternTransformation.expressions that was matched.

• sdfg (SDFG) – The SDFG in which the match was found (equal to graph).

• permissive – Whether transformation should run in permissive mode.

Returns:

True if the transformation can be applied.

classmethod expressions()

Returns a list of SDFG subgraphs that will be matched in the subgraph isomorphism phase. Used as a pre-pass before calling can_be_applied.

state_a

Static field wrapper of a node or an SDFG state that designates it as part of a subgraph pattern. These objects are used in subclasses of PatternTransformation to represent the subgraph patterns.

Example use:

class MyTransformation(SingleStateTransformation):
    some_map_node = PatternNode(nodes.MapEntry)
    array = PatternNode(nodes.AccessNode)

The two nodes can then be used in the transformation static methods (e.g., expressions, can_be_applied) to represent the nodes, and in the instance methods to point to the nodes in the parent SDFG.

state_b

Static field wrapper of a node or an SDFG state that designates it as part of a subgraph pattern. These objects are used in subclasses of PatternTransformation to represent the subgraph patterns.

Example use:

class MyTransformation(SingleStateTransformation):
    some_map_node = PatternNode(nodes.MapEntry)
    array = PatternNode(nodes.AccessNode)

The two nodes can then be used in the transformation static methods (e.g., expressions, can_be_applied) to represent the nodes, and in the instance methods to point to the nodes in the parent SDFG.

dace.transformation.interstate.state_fusion module

State fusion transformation

class dace.transformation.interstate.state_fusion.CCDesc(first_input_nodes, first_output_nodes, second_input_nodes, second_output_nodes)

Bases: object

class dace.transformation.interstate.state_fusion.StateFusion(*args, **kwargs)

Bases: MultiStateTransformation

Implements the state-fusion transformation.

State-fusion takes two states that are connected through a single edge, and fuses them into one state. If permissive, also applies if potential memory access hazards are created.

static annotates_memlets()

Indicates whether the transformation annotates the edges it creates or modifies with the appropriate memlets. This determines whether to apply memlet propagation after the transformation.

apply(_, sdfg)

Applies this transformation instance on the matched pattern graph.

Parameters:

sdfg – The SDFG to apply the transformation to.

Returns:

A transformation-defined return value, which could be used to pass analysis data out, or nothing.

can_be_applied(graph, expr_index, sdfg, permissive=False)

Returns True if this transformation can be applied on the candidate matched subgraph.

Parameters:

• graph – SDFG object in which the match was found.

• candidate – A mapping between node IDs returned from PatternTransformation.expressions and the nodes in graph.

• expr_index – The list index from PatternTransformation.expressions that was matched.

• sdfg – The SDFG in which the match was found (equal to graph).

• permissive – Whether transformation should run in permissive mode.

Returns:

True if the transformation can be applied.

classmethod expressions()

Returns a list of SDFG subgraphs that will be matched in the subgraph isomorphism phase. Used as a pre-pass before calling can_be_applied.

static find_fused_components(first_cc_input, first_cc_output, second_cc_input, second_cc_output)

Return type:

List[CCDesc]

first_state

Static field wrapper of a node or an SDFG state that designates it as part of a subgraph pattern. These objects are used in subclasses of PatternTransformation to represent the subgraph patterns.

Example use:

class MyTransformation(SingleStateTransformation):
    some_map_node = PatternNode(nodes.MapEntry)
    array = PatternNode(nodes.AccessNode)

The two nodes can then be used in the transformation static methods (e.g., expressions, can_be_applied) to represent the nodes, and in the instance methods to point to the nodes in the parent SDFG.

has_path(first_state, second_state, match_nodes, node_a, node_b)

Check for paths between the two states if they are fused.

Return type:

bool

static memlets_intersect(graph_a, group_a, inputs_a, graph_b, group_b, inputs_b)

Performs an all-pairs check for subset intersection on two groups of nodes. If group intersects or result is indeterminate, returns True as a precaution.

Parameters:

• graph_a (SDFGState) – The graph in which the first set of nodes reside.

• group_a (List[AccessNode]) – The first set of nodes to check.

• inputs_a (bool) – If True, checks inputs of the first group.

• graph_b (SDFGState) – The graph in which the second set of nodes reside.

• group_b (List[AccessNode]) – The second set of nodes to check.

• inputs_b (bool) – If True, checks inputs of the second group.

Return type:

bool

Returns:

True if subsets intersect or result is indeterminate.

second_state

Static field wrapper of a node or an SDFG state that designates it as part of a subgraph pattern. These objects are used in subclasses of PatternTransformation to represent the subgraph patterns.

Example use:

class MyTransformation(SingleStateTransformation):
    some_map_node = PatternNode(nodes.MapEntry)
    array = PatternNode(nodes.AccessNode)

The two nodes can then be used in the transformation static methods (e.g., expressions, can_be_applied) to represent the nodes, and in the instance methods to point to the nodes in the parent SDFG.

dace.transformation.interstate.state_fusion.top_level_nodes(state)

Module contents

This module initializes the inter-state transformations package.