[Mlir-commits] [mlir] [mlir][SCF] Deprecate `linalg::tileToForallOp` and `linalg::tileToForallOpUsingTileSizes` (PR #91878)

[Nicolas Vasilache]

================
@@ -530,27 +764,25 @@ static LogicalResult addInitOperandsToLoopNest(
 FailureOr<scf::SCFTilingResult>
 mlir::scf::tileUsingSCF(RewriterBase &rewriter, TilingInterface op,
                         const scf::SCFTilingOptions &options) {
+  if (failed(verifyTileSizeOptions(rewriter, op.getLoc(), options))) {
+    return failure();
+  }
+
   OpBuilder::InsertionGuard guard(rewriter);
   rewriter.setInsertionPointAfter(op);
 
-  if (!options.tileSizeComputationFunction) {
-    return rewriter.notifyMatchFailure(
-        op, "missing tile size computation function");
-  }
-
   // 1. Get the range of the loops that are represented by the operation.
   SmallVector<Range> iterationDomain = op.getIterationDomain(rewriter);
-  size_t numLoops = iterationDomain.size();
 
-  // 2. Materialize the tile sizes. Enforce the convention that "tiling by zero"
-  // skips tiling a particular dimension. This convention is significantly
-  // simpler to handle instead of adjusting affine maps to account for missing
-  // dimensions.
-  SmallVector<OpFoldResult> tileSizes =
-      options.tileSizeComputationFunction(rewriter, op);
-  if (tileSizes.size() < iterationDomain.size()) {
-    auto zero = rewriter.getIndexAttr(0);
-    tileSizes.append(numLoops - tileSizes.size(), zero);
+  // 2. Materialize the tile sizes and/or number of threads;
+  SmallVector<OpFoldResult> tileSizes, numThreads;
+  std::tie(tileSizes, numThreads) =
+      getTileSizes(rewriter, op, iterationDomain, options);
----------------
nicolasvasilache wrote:

is this `getTileSizes` precomputation needed because of `checkSafeToTileToForall` or something else ?

https://github.com/llvm/llvm-project/pull/91878