[PATCH] D107540: [OMPIRBuilder] Clarify CanonicalLoopInfo. NFC.

[Alex Zinenko via Phabricator via llvm-commits]

ftynse accepted this revision.
ftynse added a comment.
This revision is now accepted and ready to land.

Thanks!



================
Comment at: llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h:364
+  /// over all chunks that are executed on the same thread. Returning
+  /// CanonicalLoopInfo objects representing then may eventually be useful for
+  /// the apply clause planned in OpenMP 6.0, but currently whether these are
----------------



================
Comment at: llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h:393
+  /// \param DL       Debug location for instructions added for the
+  ///                workshare-loop construct itself.
   /// \param CLI      A descriptor of the canonical loop to workshare.
----------------



================
Comment at: llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h:420
+  /// \param DL       Debug location for instructions added for the
+  ///                workshare-loop construct itself.
   /// \param CLI      A descriptor of the canonical loop to workshare.
----------------



================
Comment at: llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h:1287
+/// The loop is thought to start at PreheaderIP (at the Preheader's terminator,
+/// including) and at AfterIP (at the After's first instruction, excluding).
+/// That is, instructions in the Preheader and After blocks (except the
----------------



================
Comment at: llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h:1315
+/// basic blocks. After invalidation, the CanonicalLoopInfo must not be used
+/// anymore as its underlaying control flow does not exist anymore.
+/// Loop-transformation methods such as tileLoops, collapseLoops and unrollLoop
----------------
I suppose it may still exist in some cases.


================
Comment at: llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h:1323
+/// modified loop. What is done is an implementation detail of
+/// transformation-implementing method and callers should always assume the the
+/// CanonicalLoopInfo passed to it is invalidated and a new object is returned.
----------------



================
Comment at: llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h:1330
+/// Generally, methods consuming CanonicalLoopInfo do not need an
+/// OpenMPIRBuilder::InsertPointTy as argument, but uses the locations of the
+/// CanonicalLoopInfo to insert new or modify existing instructions. Unless
----------------



Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D107540/new/

https://reviews.llvm.org/D107540