[Mlir-commits] [mlir] [mlir][mesh, MPI] Mesh2mpi (PR #104566)

[Frank Schlimbach]

================
@@ -0,0 +1,240 @@
+//===- MeshToMPI.cpp - Mesh to MPI  dialect conversion -----------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// This file implements a translation of Mesh communication ops tp MPI ops.
+//
+//===----------------------------------------------------------------------===//
+
+#include "mlir/Conversion/MeshToMPI/MeshToMPI.h"
+
+#include "mlir/Dialect/Arith/IR/Arith.h"
+#include "mlir/Dialect/MPI/IR/MPI.h"
+#include "mlir/Dialect/MemRef/IR/MemRef.h"
+#include "mlir/Dialect/Mesh/IR/MeshOps.h"
+#include "mlir/Dialect/SCF/IR/SCF.h"
+#include "mlir/Dialect/Utils/StaticValueUtils.h"
+#include "mlir/IR/Builders.h"
+#include "mlir/IR/BuiltinAttributes.h"
+#include "mlir/IR/BuiltinTypes.h"
+#include "mlir/IR/PatternMatch.h"
+#include "mlir/Transforms/GreedyPatternRewriteDriver.h"
+
+#define DEBUG_TYPE "mesh-to-mpi"
+#define DBGS() (llvm::dbgs() << "[" DEBUG_TYPE "]: ")
+
+namespace mlir {
+#define GEN_PASS_DEF_CONVERTMESHTOMPIPASS
+#include "mlir/Conversion/Passes.h.inc"
+} // namespace mlir
+
+using namespace mlir;
+using namespace mlir::mesh;
+
+namespace {
+
+// This pattern converts the mesh.update_halo operation to MPI calls
+struct ConvertUpdateHaloOp
+    : public mlir::OpRewritePattern<mlir::mesh::UpdateHaloOp> {
+  using OpRewritePattern::OpRewritePattern;
+
+  mlir::LogicalResult
+  matchAndRewrite(mlir::mesh::UpdateHaloOp op,
----------------
fschlimb wrote:

Agree, let's add a tensor-based `updateHalo` once needed.

We need the memref-based version no matter what. Array/numpy semantics are partially reference-based. For `subview` and `insert_slice` they have `memref` semantics, copies are disallowed. This cannot really be expressed on the tensor-level.

My question was not so much about optimizations in general (which of course are simpler on the tensor level). I was wondering if early, implicit bufferization - when converting to MPI - would do any harm, like when done right before bufferization.

We need the memref-based version no matter what. Array/numpy semantics are partially not value-based. For `subview` and `insert_slice` they have `memref` semantics, copies are disallowed.

https://github.com/llvm/llvm-project/pull/104566