[clang] [clang][dataflow] Refactor processing of terminator element (PR #84499)
via cfe-commits
cfe-commits at lists.llvm.org
Tue Mar 12 04:13:39 PDT 2024
================
@@ -337,26 +274,33 @@ computeBlockInputState(const CFGBlock &Block, AnalysisContext &AC) {
AC.BlockStates[Pred->getBlockID()];
if (!MaybePredState)
continue;
-
- if (AC.Analysis.builtinOptions()) {
- if (const Stmt *PredTerminatorStmt = Pred->getTerminatorStmt()) {
- // We have a terminator: we need to mutate an environment to describe
- // when the terminator is taken. Copy now.
+ const TypeErasedDataflowAnalysisState &PredState = *MaybePredState;
+
+ if (const Stmt *PredTerminatorStmt = Pred->getTerminatorStmt()) {
----------------
martinboehme wrote:
To make this clearer, here's how the whole code block could be changed to make it (IMO) simpler (I can't do this as a suggested edit because I would have to include deleted lines in the block of code, and github doesn't allow me to do this):
```cxx
const Expr *TerminatorCond = getTerminatorCondition(Pred->getTerminatorStmt());
if (TerminatorCond == nullptr) {
Builder.addUnowned(PredState);
continue;
}
bool BranchVal = blockIndexInPredecessor(*Pred, Block) == 0;
// `transferBranch` may need to mutate the environment to describe the
// dynamic effect of the terminator for a given branch. Copy now.
TypeErasedDataflowAnalysisState Copy = PredState.fork();
if (AC.Analysis.builtinOptions()) {
auto *CondVal = Copy.Env.get<BoolValue>(*TerminatorCond);
// In transferCFGBlock(), we ensure that we always have a `Value`
// for the terminator condition, so assert this. We consciously
// assert ourselves instead of asserting via `cast()` so that we get
// a more meaningful line number if the assertion fails.
assert(CondVal != nullptr);
BoolValue *AssertedVal =
BranchVal ? CondVal : &Copy.Env.makeNot(*CondVal);
Copy.Env.assume(AssertedVal->formula());
}
AC.Analysis.transferBranchTypeErased(BranchVal, TerminatorCond, Copy.Lattice,
Copy.Env);
Builder.addOwned(std::move(Copy));
```
https://github.com/llvm/llvm-project/pull/84499
More information about the cfe-commits
mailing list