GHC.Cmm.Dataflow.Graph

Documentation

type Body (n :: Extensibility -> Extensibility -> Type) = LabelMap (Block n C C)

A (possibly empty) collection of closed/closed blocks

type Graph = Graph' Block

A control-flow graph, which may take any of four shapes (O/O, OC, CO, C/C). A graph open at the entry has a single, distinguished, anonymous entry point; if a graph is closed at the entry, its entry point(s) are supplied by a context.

data Graph' (block :: (Extensibility -> Extensibility -> Type) -> Extensibility -> Extensibility -> Type) (n :: Extensibility -> Extensibility -> Type) (e :: Extensibility) (x :: Extensibility) where

Graph' is abstracted over the block type, so that we can build graphs of annotated blocks for example (Compiler.Hoopl.Dataflow needs this).

Constructors

GNil :: forall (block :: (Extensibility -> Extensibility -> Type) -> Extensibility -> Extensibility -> Type) (n :: Extensibility -> Extensibility -> Type). Graph' block n 'Open 'Open

GUnit :: forall (block :: (Extensibility -> Extensibility -> Type) -> Extensibility -> Extensibility -> Type) (n :: Extensibility -> Extensibility -> Type). block n O O -> Graph' block n 'Open 'Open

GMany :: forall (e :: Extensibility) (block :: (Extensibility -> Extensibility -> Type) -> Extensibility -> Extensibility -> Type) (n :: Extensibility -> Extensibility -> Type) (x :: Extensibility). MaybeO e (block n O C) -> Body' block n -> MaybeO x (block n C O) -> Graph' block n e x

Instances

OutputableP Platform (Graph CmmNode e x)

Defined in GHC.Cmm.Node

Methods

pdoc :: Platform -> Graph CmmNode e x -> SDoc

class NonLocal (thing :: Extensibility -> Extensibility -> Type) where

Gives access to the anchor points for nonlocal edges as well as the edges themselves

Methods

entryLabel

Arguments

:: forall (x :: Extensibility). thing C x

-> Label

The label of a first node or block

successors

Arguments

:: forall (e :: Extensibility). thing e C

-> [Label]

Gives control-flow successors

Instances

NonLocal CmmNode

Defined in GHC.Cmm.Node

Methods

entryLabel :: forall (x :: Extensibility). CmmNode C x -> Label

successors :: forall (e :: Extensibility). CmmNode e C -> [Label]

NonLocal n => NonLocal (Block n)

Defined in GHC.Cmm.Dataflow.Graph

Methods

entryLabel :: forall (x :: Extensibility). Block n C x -> Label

successors :: forall (e :: Extensibility). Block n e C -> [Label]

addBlock :: (NonLocal block, HasDebugCallStack) => block C C -> LabelMap (block C C) -> LabelMap (block C C)

bodyList :: forall block (n :: Extensibility -> Extensibility -> Type). Body' block n -> [(Label, block n C C)]

bodyToBlockList :: forall (n :: Extensibility -> Extensibility -> Type). Body n -> [Block n C C]

emptyBody :: forall block (n :: Extensibility -> Extensibility -> Type). Body' block n

labelsDefined :: forall (block :: (Extensibility -> Extensibility -> Type) -> Extensibility -> Extensibility -> Type) (n :: Extensibility -> Extensibility -> Type) (e :: Extensibility) (x :: Extensibility). NonLocal (block n) => Graph' block n e x -> LabelSet

mapGraph :: forall n n' (e :: Extensibility) (x :: Extensibility). (forall (e1 :: Extensibility) (x1 :: Extensibility). n e1 x1 -> n' e1 x1) -> Graph n e x -> Graph n' e x

Maps over all nodes in a graph.

mapGraphBlocks :: forall block (n :: Extensibility -> Extensibility -> Type) block' (n' :: Extensibility -> Extensibility -> Type) (e :: Extensibility) (x :: Extensibility). (forall (e1 :: Extensibility) (x1 :: Extensibility). block n e1 x1 -> block' n' e1 x1) -> Graph' block n e x -> Graph' block' n' e x

Function mapGraphBlocks enables a change of representation of blocks, nodes, or both. It lifts a polymorphic block transform into a polymorphic graph transform. When the block representation stabilizes, a similar function should be provided for blocks.

revPostorderFrom :: NonLocal block => LabelMap (block C C) -> Label -> [block C C]

Returns a list of blocks reachable from the provided Labels in the reverse postorder.

This is the most important traversal over this data structure. It drops unreachable code and puts blocks in an order that is good for solving forward dataflow problems quickly. The reverse order is good for solving backward dataflow problems quickly. The forward order is also reasonably good for emitting instructions, except that it will not usually exploit Forrest Baskett's trick of eliminating the unconditional branch from a loop. For that you would need a more serious analysis, probably based on dominators, to identify loop headers.

For forward analyses we want reverse postorder visitation, consider: A -> [B,C] B -> D C -> D Postorder: [D, C, B, A] (or [D, B, C, A]) Reverse postorder: [A, B, C, D] (or [A, C, B, D]) This matters for, e.g., forward analysis, because we want to analyze *both* B and C before we analyze D.