The MaterializeTrivialGoto pass, tries to remove superfluous `goto` from
the `ScopeGraph`, by verifying that the resulting `ScopeGraph` does not
become cyclic or undecided.
In that case, it rolls back the changes.
Introduce the `inline-divergent-scopes` pass. This pass in is charge of
reducing the exit nodes in a `ScopeGraph`, applying the _Inline
Divergent Scopes_ transformation, which is able, in presence of
divergent exits nodes, wrt. to a conditional, to remove the exit node
and to transform the `ScopeGraph` in a form where semantics is
preserved.
Introduce the `select-scope` pass. This pass is in charge of enforcing
the relative decidedness property for all the nodes in the `ScopeGraph`,
wrt. to all the conditional nodes. It does this by applying the
relative decidedness definition, electing one scope for each conditional
node in the graph, and by transforming all the edges that would violate
the property into goto edges.
Implement the DAGify pass. This pass, using the results exposed by the
`GenericRegionInfo` analysis, transforms all the retreating edges of
each identified `GenericRegion`, processed in a bottom-up fashion, into
a `goto` edge on the `ScopeGraph`.
Introduce the Enforce Single Exit pass, whose task is to normalize a
generic `ScopeGraph`, which may have multiple exit blocks (and/or
infinite loop regions), in order to have a single `sink_block` as exit
block.
This is done by adding a new entry block, a `sink_block`, and some
`scope_closer` edges (which are visible only on the `ScopeGraph`) that
enforce the property.
This is done taking inspiration from how the internally the
`PostDominatorTree` pass construct the temporary graph on which the post
dominance information is computed on.
Some unit tests are added in order to verify that the pass works as
expected.
Introduce the `ScopeCloser` and `GotoTarget` annotations in the IR, and
the relative necessary machinery, needed to handle scope closer and goto
edges for the new backend.
A specialization of the `llvm::GraphTraits`, called `ScopeGraph`, that
is able to handle both the above mentioned annotations is provided.
For the `llvm::GraphTraits` implementation, we introduce the
`GeneratorIterator` class, which uses a coroutine to store the status of
the iteration.
A debug logger pass is added, so that we are able to test the
functionality with `FileCheck`.
We implement the `GenericRegionInfo` analysis.
The analysis accepts a template parameter which enables to run it on
every graph which exposes `llvm::GraphTraits`.
The `GenericRegionPass` is responsible for instantiating and running the
analysis on a `llvm::Function`.
The `GenericRegionInfo` analysis uses, and takes insipiration from the
`GenericCycleInfo` LLVM analysis.
The analysis exposes a tree of well nested `GenericRegion`s, which are
constructed starting from the well nested tree of `GenericCycle`s.
In addition, we perform the election of the `Head` of each
`GenericRegion`, and the election of the retreating edges.
We introduce the `CycleEquivalence` analysis.
This is an analysis which implements the _Cycle Equivalence_ computation
algorithm, and provides as result the _cycle equivalence classes_
Specifically, we introduce:
- The `CycleEquivalenceClass`, which is the unitary element computed by
the analysis.
- The `CycleEquivalenceAnalysis`, which contains the algorithm to
compute the cycle equivalence.
- The `CycleEquivalencePass`, a `FunctionPass` that can be used to
perform the analysis on a `llvm::Function`.
The algorithm is composed by various stages:
- We construct a new `GenericGraph` object, replicating the input CFG,
with the addition of the `exit`->`entry` edge.
- Taking advantage of `llvm::GraphTraits<Undirected<>>`, we can now
implement the algorithm working on an undirected version of the input.
- We perform the `CycleEquivalence` computation, returning a
`llvm::SmallVector` of `CycleEquivalenceClass` objects.
In addition to using the `llvm::GraphTraits<Undirected<>>` traits to
walk on the equivalent undirected graph, we also need to:
- Perform a preliminary DFS, in order to:
- Assign the DFS number to each node in the graph.
- Compute the spanning tree, and use this information to distinguish
tree edges and back edges when running the algorithm.
The internal graph used by the analysis also normalizes the graph in
order to have a sinle exit node (called sink), which is a requirement
for the `CycleEquivalence` algorithm.
We also implement the `llvm::DOTGraphTraits` for the
`CycleEquivalenceAnalysis<llvm::Function *>` specialization. In this
way, we can have a graphical representation of the undirected graph used
internally in the `CycleEquivalenceAnalysis` core implementation.
We add some `FileCheck` tests on some well-known graph topologies.
We introduce a "dead code" optimization like pass on the AST.
The goal is to simplify away, in a `SequenceNode`, everything that
follows a node with sports a `nofallthrough behavior.
We introduce the `InlineDispatcherSwitch` beautify pass. Its goal is to
try and inline the body of some of the `case`s of a exit dispatcher, in
place of the `SetNode` corresponding to that `case`, if this doesn't
introduce duplication in the code (i.e., a single `SetNode` for that
specific case value is present).
Additionally, if the inlining procedure is able to completely remove the
necessity of an exit dispatcher altogether, the pass removes it.
The pass is able to handle chains of weaved dispatcher `switch`es
referring to the same original dispatcher `switch`, by handling the
inline operation and the possible simplification level-wise.
The inlining procedure, cannot take place if a `SetNode` is contained in
the body of the case we are trying to inline, since this can possibly
break the semantics of the state variable of a loop, by placing a
`SetNode` in a more internal loop.
We introduce the PromoteCallNoReturn beautification pass. Its goal is to
restructure sequence of statements, in order to have `call`s to
`noreturn` functions as _inlined_ in the middle of the statement
sequence, and leave _non local control flow statements_ at the end of
that scope. E.g., we prefer:
```
if (cond)
call noreturnfunc();
return;
```
to
```
if (!cond)
return;
call noreturnfunc();
```
In order to do this, contextually, we restructure the routine computing
the `fallthrough` property, in order to be able to differentiate between
the _non local control flow statements_, a call to a `noreturn`
function, or a generic mix of the two (useful when combining results
from the two situations above).
The new analysis is also used in the `promoteNoFallThrough` promotion
pass.
We introduce a simplification step, which looks for `switch`es that can
be reduced to simpler `if` statements.
Specifically, the logic is the following:
1) When we identify a `switch` statement composed by a single `case` and
a possible default, we transform it into an `if` with the `case` now
corresponding to the `then`, and the `default` corresponding to the
`else`, if present.
2) When we identify a `switch` statement composed by two `case`s, and no
`default` is present, we can promote it to an `if` with `then` `else`
branches.
Other key details:
- The promotion happens only if we can identify at least one of the
`case`s that have a single element in the `case` label. If this is not
the case, we do not promote one to RHS of the `if` condition.
- A new `CompareNode` class, inheriting from `ExprNode`, is created, in
order to represent the equality or inequality condition of an `IfNode`
instance that is the result of the promotion. This `CompareNode` can
represent for the LHS both an `llvm::Value` or the `loop_state_var`,
while it embeds the RHS constant which completes the comparison.
- We remove `SwitchBreak` AST nodes that may now appear as children of
an `if` node promoted from a `switch`.
- We introduce in the `CompareNode` the `weaved` concept. Indeed, if a
promotion of a weaved `switch` happens, we should avoid the
serialization of the instructions leading to the computation of the
condition of the original `switch`, because they have been already
emitted by the main related dominating `switch`.
We also introduce an additional simplification step, which takes care
of:
- Promoting `!(==)` to `(!=)` and `!(!=)` to `(==)`, if the inner
equal/not equal is represented via a `CompareNode`.
- Promoting `x == 0` to `!x` and `x != 0` to `x`.
To be able to correctly emit (or not) the instructions computing a
condition of an `IfNode`, we need to add the `EmittBB` flag, an
additional parameter to the `buildGHASTCondition` function, which
controls the emission of the statements of a basic block computing a
condition.
Consequently, the `IfNode` acquires a `IsWeaved` field, which is used to
mirror the property having the same name on `SwitchNode`. Being now
possible a promotion from a dual `SwitchNode` to an `IfNode`, we need to
represent this property on the `IfNode` too.
Introduce implicit statements simplification phase, specifically:
- A implicit `return` simplification: `return` statements in `void`
type functions, which are not followed by any other scope, can be
omitted.
- A implicit `continue` simplification: `continue` statements whose
fallthrough leads directly to the end of the cycle scope (i.e., to
execute another iteration of the enclosing loop), can be omitted.
In order to avoid the printing of the implicit `return`, we need an
additional `emitReturn` parameter in the `emitBasicBlock` method of the
`CCodeGenerator` class.
Introduce the hybrid beautifier aiming to simplify double `not`s (one on
the IR side (e.g., a `icmp ne`), and one on the `ExprNode` side on the
`GHAST` in the form of a `NotNode`.
To do this hybrid beautifier, we basically need to do the following:
- Implement a preliminary collection phase which computes which
`ExprNode`s are affected by which `BasicBlock`.
- Compute the so-called consesus, i.e., verify that for all the
`ExprNode`s affected by a certain `BasicBlock`, the transformation
would be beneficial in terms of output. Basically:
1) That the number of `!(!=)` transformed into `(==)` outscores the
number of correlated `(!=)` which are transformed into `!(==)`.
2) That the number of `!(==)` transformed into `(!=)` outscores the
number of correlated `(==)` which are transformed into `!(!=)`.
- Actually perform the transformation for those situations where the
consensus agrees.
The beautification library was split into a separate library for
historical reason, having to do with LLVM passes, their dependencies,
and dirty hacks we had around.
Now the LLVM passes, and the hacks are gone, so we can merge back
everything together.
Now that the late stage of the decompilation pipeline is managed with
revng-pipeline, and we have a standalone tool for decompilation, we
have no need to use LLVM passes anymore for those stages.
This commit also renames the directories to avoid the confusing Pass
suffix, not what LLVM passes are gone.
Andrea Gussoni