================
@@ -158,24 +170,27 @@ mlir::LogicalResult 
CIRGenFunction::emitOpenACCOpAssociatedStmt(
 
   llvm::SmallVector<mlir::Type> retTy;
   llvm::SmallVector<mlir::Value> operands;
-
-  // Clause-emitter must be here because it might modify operands.
-  OpenACCClauseCIREmitter clauseEmitter(getCIRGenModule(), dirKind, dirLoc);
-  clauseEmitter.VisitClauseList(clauses);
-
   auto op = builder.create<Op>(start, retTy, operands);
 
-  // Apply the attributes derived from the clauses.
-  clauseEmitter.applyAttributes(builder, op);
+  {
+    mlir::OpBuilder::InsertionGuard guardCase(builder);
+    // Sets insertion point before the 'op', since every new expression needs 
to
+    // be before the operation.
+    builder.setInsertionPoint(op);
----------------
erichkeane wrote:

See here for the magic around getting the 'insertion' point right.  We now only 
view the clauses information 1x in the emitter (rather than collect data, then 
emit data for attributes), but all of the ops they 'insert' need to happen 
before the OpenACC operation.

https://github.com/llvm/llvm-project/pull/135851
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
  • [clang] [... Erich Keane via cfe-commits
    • [cla... via cfe-commits
    • [cla... via cfe-commits
    • [cla... Erich Keane via cfe-commits
    • [cla... Erich Keane via cfe-commits
    • [cla... Erich Keane via cfe-commits
    • [cla... Erich Keane via cfe-commits
    • [cla... Erich Keane via cfe-commits
    • [cla... Erich Keane via cfe-commits
    • [cla... Erich Keane via cfe-commits
    • [cla... Valentin Clement バレンタイン クレメン via cfe-commits
    • [cla... Valentin Clement バレンタイン クレメン via cfe-commits
    • [cla... Erich Keane via cfe-commits
    • [cla... Erich Keane via cfe-commits
    • [cla... Andy Kaylor via cfe-commits
    • [cla... Andy Kaylor via cfe-commits
    • [cla... Andy Kaylor via cfe-commits
    • [cla... Andy Kaylor via cfe-commits
    • [cla... Erich Keane via cfe-commits

Reply via email to