We now explicitly store the retreatings edges, that will eventually be
connected to the `continue` nodes, so that we can explicitly handle
them, without resorting to iterating over the predecessors of the
`EntryNode`, which although correct is a fragile method with respect to
defective situations.
In addition, we check that after the restructuring, all the retreating
edges that we connect to `continue` nodes, do point to the `EntryNode`
before the transformation.
Before this commit, the code was using an old arcane algorithm that was
adding and removing dummy frontier nodes, and relying on the
DominatorTree.
This commit drops the DominatorTree altogether, drops the use of
frontier nodes, and rewrites the algorithm only using successors and
predecessors.
Now a node outside a metaregion M is included in the metaregion only if
all its predecessors are part of the metaregion, and none of them is
already in another metaregion that is a transitive parent of M.
Before this commit, the whole body of each loop was outlined and
duplicated, only to be removed later if not necessary.
Now we only do this if there is more than one entry to the loop, which
is the simplest condition to detect when first iteration outlining is
needed.
- Add PrepareLLVMIRForMLIR Pass.
This Pass performs massaging of the LLVM IR
produced by revng so it can be parsed/accepted
for the LLVM MLIR Dialect.
- Introduce new pipeline branch: llvmir-to-mlir.yml.
- Add LLVMIRToMLIR step.
- Add ImportLLVMToMLIRPipe pipe.
This commit sets a name for types we use in OpaqueExtractValues.
These are StructTypes that can only be returned from isolated functions
with RawFunctionType prototype on the model, or from helpers that do not
represent isolated functions.
The name is required because when we will converto to MLIR LLVM Dialect,
there are checks in place that forbid unnamed types.
The code re-uses code that is used to print C code, in order to make
sure that the struct names we use for this are the same as those we emit
in C, to ease debugging.
This commit also cleans up the unittest to make it more concise, while
still testing the same things.
This commit drops some old broken code that was a leftover from when
MarkAssignments didn't inject calls to Assign, and that was poorly
migrated to Assign, causing it to be dead code.
Before this commit all the local variabels emitted by MarkAssignments
were using integer types obtained with llvmIntToModelType.
This commit enables using model::QualifiedTypes more often, resulting in
overall more readable C code.
This commit fixes a bug in the hasSideEffects function, used to evaluate
if an ExprNode has side effects.
Before this commit, when calling hasSideEffects on a NotNode, the
return value was true if the operand of NotNode **did not** have side
effects, which is wrong.
This commit changes the behaviour so that a NotNode has side effects if
and only if its only operand has side effects.
This commit fixes a bug in the hasSideEffects function, used to evaluate
if an ExprNode has side effects.
Before this commit, when calling hasSideEffects on an AndNode, the
return value was true if **both** the LHS and RHS of the AndNode had
side effects, which is wrong.
This commit changes the behaviour so that if **either** LHS **or** the
RHS of the And have any side effect, than hasSideEffects returns true
for the And.
Before this commit, the simplifyShortCircuit beautifier did not recur on
the then/else branch of an IfNode, basically stopping recursion very
early and potentially missing a lot of beautification opportunities.
Before this commit the DLA frontend was not ready to handle all the
possible combinations of integer- and pointer-typed arithmetic that was
used to compute SCEVs.
This could cause crashes when dealing with pointer-typed SCEVs.
This commit fixes the issues by converting all pointer-typed expressions
to integers, and resizing integers to the same size before adding them.
This commit adds caching for mutual reachability among children of a
node, in SimplifyInstanceAtOffset0.
This makes the DLAStep 10% to 50% faster on real-world benchmarks we
have measured, such as `updatedb.plocate` and `df` for Ubuntu 22.04
x86_64.
llvm::EquivalenceClasses is an efficient data structure from LLVM, to
compute equivalence classes among objects with Tarjan's union-find.
This commit uses that to avoid an hand-crafted very inefficient
algorithm.
The Decompiled container (which use FunctionStringMap) now uses
`.tar.gz` as the container format, as opposed to YAML. Change the
filename of the containers accordingly.
`revng` now requires all pipeline files to have a `Component` field
which tells the command line driver to what component the pipeline
definition belongs to.
Andrea Gussoni