[flang-commits] [flang] [llvm] [mlir] [flang][OpenMP] Enable tiling (PR #143715)

[Michael Kruse via flang-commits]

================
@@ -3449,13 +3500,9 @@ static void genOMPDispatch(lower::AbstractConverter &converter,
     newOp = genTeamsOp(converter, symTable, stmtCtx, semaCtx, eval, loc, queue,
                        item);
     break;
-  case llvm::omp::Directive::OMPD_tile: {
-    unsigned version = semaCtx.langOptions().OpenMPVersion;
-    if (!semaCtx.langOptions().OpenMPSimd)
-      TODO(loc, "Unhandled loop directive (" +
-                    llvm::omp::getOpenMPDirectiveName(dir, version) + ")");
+  case llvm::omp::Directive::OMPD_tile:
+    newOp = genLoopOp(converter, symTable, semaCtx, eval, loc, queue, item);
----------------
Meinersbur wrote:

> In the GenericLoopConversionPass there is a function rewriteStandaloneLoop that converts the loop to a simdloop. 

I think the semantics are different. Independent of how it is executed (parallel, simd, sequential, ...), the loop is duplicated with a copy nested in the other. This logically results in $n^2$ executions, where $n$ is the tripcount of the original loop. 

I was experimenting with it, and could observe those additional executions.


https://github.com/llvm/llvm-project/pull/143715