[flang] Schedule InlineHLFIRAssign after BufferizeHLFIR. (#121863)

This helps to get rid of *some* calls to AssignTemporary runtime
that are appearing due to temporary_lhs hlfir.assign produced
in BufferizeHLFIR. I only tested it on `tonto`, and did not see
any performance changes. I will run more performance testing
before merging this.

Author: vzakhari

flang/lib/Optimizer/Passes/Pipelines.cpp

@@ -240,6 +240,16 @@ void createHLFIRToFIRPassPipeline(mlir::PassManager &pm, bool enableOpenMP,
   pm.addPass(hlfir::createLowerHLFIROrderedAssignments());
   pm.addPass(hlfir::createLowerHLFIRIntrinsics());
   pm.addPass(hlfir::createBufferizeHLFIR());
+  // Run hlfir.assign inlining again after BufferizeHLFIR,
+  // because the latter may introduce new hlfir.assign operations,
+  // e.g. for copying an array into a temporary due to
+  // hlfir.associate.
+  // TODO: we can remove the previous InlineHLFIRAssign, when
+  // FIR AliasAnalysis is good enough to say that a temporary
+  // array does not alias with any user object.
+  if (optLevel.isOptimizingForSpeed())
+    addNestedPassToAllTopLevelOperations<PassConstructor>(
+        pm, hlfir::createInlineHLFIRAssign);
   pm.addPass(hlfir::createConvertHLFIRtoFIR());
   if (enableOpenMP)
     pm.addPass(flangomp::createLowerWorkshare());

flang/test/Driver/mlir-pass-pipeline.f90

@@ -49,6 +49,15 @@
 ! ALL: LowerHLFIROrderedAssignments
 ! ALL-NEXT: LowerHLFIRIntrinsics
 ! ALL-NEXT: BufferizeHLFIR
+! O2-NEXT: Pipeline Collection : ['fir.global', 'func.func', 'omp.declare_reduction', 'omp.private']
+! O2-NEXT:   'fir.global' Pipeline
+! O2-NEXT:     InlineHLFIRAssign
+! O2-NEXT:   'func.func' Pipeline
+! O2-NEXT:     InlineHLFIRAssign
+! O2-NEXT:   'omp.declare_reduction' Pipeline
+! O2-NEXT:     InlineHLFIRAssign
+! O2-NEXT:   'omp.private' Pipeline
+! O2-NEXT:     InlineHLFIRAssign
 ! ALL-NEXT: ConvertHLFIRtoFIR
 ! ALL-NEXT: CSE
 ! Ideally, we need an output with only the pass names, but

flang/test/Fir/basic-program.fir

@@ -50,6 +50,15 @@ func.func @_QQmain() {
 // PASSES-NEXT:   LowerHLFIROrderedAssignments
 // PASSES-NEXT:   LowerHLFIRIntrinsics
 // PASSES-NEXT:   BufferizeHLFIR
+// PASSES-NEXT: Pipeline Collection : ['fir.global', 'func.func', 'omp.declare_reduction', 'omp.private']
+// PASSES-NEXT:   'fir.global' Pipeline
+// PASSES-NEXT:     InlineHLFIRAssign
+// PASSES-NEXT:   'func.func' Pipeline
+// PASSES-NEXT:     InlineHLFIRAssign
+// PASSES-NEXT:   'omp.declare_reduction' Pipeline
+// PASSES-NEXT:     InlineHLFIRAssign
+// PASSES-NEXT:   'omp.private' Pipeline
+// PASSES-NEXT:     InlineHLFIRAssign
 // PASSES-NEXT:   ConvertHLFIRtoFIR
 // PASSES-NEXT:   LowerWorkshare
 // PASSES-NEXT:   CSE