[Mlir-commits] [mlir] [mlir][dataflow] Improve DataFlowFramework debug output (PR #176632)

[Matthias Springer]

================
@@ -46,13 +46,27 @@ void AnalysisState::addDependency(ProgramPoint *dependent,
   DATAFLOW_DEBUG({
     if (inserted) {
       LDBG() << "Creating dependency between " << debugName << " of " << anchor
-             << "\nand " << debugName << " on " << *dependent;
+             << "\nand " << debugName << " on " << *dependent << "\nwith "
+             << analysis->debugName;
     }
   });
 }
 
 void AnalysisState::dump() const { print(llvm::errs()); }
 
+void AnalysisState::onUpdate(DataFlowSolver *solver) const {
+  for (const DataFlowSolver::WorkItem &item : dependents) {
+    DATAFLOW_DEBUG(LDBG() << "Enqueueing dependent work item: " << *item.first
----------------
matthias-springer wrote:

nit: I think the output would read slightly better if you prefix the analysis name: `[debugName] Enqueueing ...`. (Here and in other places.)

https://github.com/llvm/llvm-project/pull/176632