[Mlir-commits] [mlir] [mlir][memref] Add runtime verification for `memref.atomic_rmw` (PR #130414)

Matthias Springer llvmlistbot at llvm.org
Thu Mar 20 00:31:22 PDT 2025 

https://github.com/matthias-springer updated https://github.com/llvm/llvm-project/pull/130414

>From 18f917b8c68bf7164b893d274c484d19769ea712 Mon Sep 17 00:00:00 2001
From: Matthias Springer <mspringer at nvidia.com>
Date: Sat, 8 Mar 2025 14:34:54 +0100
Subject: [PATCH] [mlir][memref] Add runtime verification for
 `memref.atomic_rmw`

---
 .../Transforms/RuntimeOpVerification.cpp      | 45 +++++++++++--------
 .../atomic-rmw-runtime-verification.mlir      | 45 +++++++++++++++++++
 .../MemRef/store-runtime-verification.mlir    | 45 +++++++++++++++++++
 3 files changed, 116 insertions(+), 19 deletions(-)
 create mode 100644 mlir/test/Integration/Dialect/MemRef/atomic-rmw-runtime-verification.mlir
 create mode 100644 mlir/test/Integration/Dialect/MemRef/store-runtime-verification.mlir

diff --git a/mlir/lib/Dialect/MemRef/Transforms/RuntimeOpVerification.cpp b/mlir/lib/Dialect/MemRef/Transforms/RuntimeOpVerification.cpp
index 134e8b5efcfdf..be0f4724ea63a 100644
--- a/mlir/lib/Dialect/MemRef/Transforms/RuntimeOpVerification.cpp
+++ b/mlir/lib/Dialect/MemRef/Transforms/RuntimeOpVerification.cpp
@@ -35,6 +35,26 @@ Value generateInBoundsCheck(OpBuilder &builder, Location loc, Value value,
   return inBounds;
 }
 
+/// Generate a runtime check to see if the given indices are in-bounds with
+/// respect to the given ranked memref.
+Value generateIndicesInBoundsCheck(OpBuilder &builder, Location loc,
+                                   Value memref, ValueRange indices) {
+  auto memrefType = cast<MemRefType>(memref.getType());
+  assert(memrefType.getRank() == static_cast<int64_t>(indices.size()) &&
+         "rank mismatch");
+  Value cond = builder.create<arith::ConstantOp>(
+      loc, builder.getIntegerAttr(builder.getI1Type(), 1));
+
+  auto zero = builder.create<arith::ConstantIndexOp>(loc, 0);
+  for (auto [dim, idx] : llvm::enumerate(indices)) {
+    Value dimOp = builder.createOrFold<memref::DimOp>(loc, memref, dim);
+    Value inBounds = generateInBoundsCheck(builder, loc, idx, zero, dimOp);
+    cond = builder.createOrFold<arith::AndIOp>(loc, cond, inBounds);
+  }
+
+  return cond;
+}
+
 struct AssumeAlignmentOpInterface
     : public RuntimeVerifiableOpInterface::ExternalModel<
           AssumeAlignmentOpInterface, AssumeAlignmentOp> {
@@ -230,26 +250,10 @@ struct LoadStoreOpInterface
   void generateRuntimeVerification(Operation *op, OpBuilder &builder,
                                    Location loc) const {
     auto loadStoreOp = cast<LoadStoreOp>(op);
-
-    auto memref = loadStoreOp.getMemref();
-    auto rank = memref.getType().getRank();
-    if (rank == 0) {
-      return;
-    }
-    auto indices = loadStoreOp.getIndices();
-
-    auto zero = builder.create<arith::ConstantIndexOp>(loc, 0);
-    Value assertCond;
-    for (auto i : llvm::seq<int64_t>(0, rank)) {
-      Value dimOp = builder.createOrFold<memref::DimOp>(loc, memref, i);
-      Value inBounds =
-          generateInBoundsCheck(builder, loc, indices[i], zero, dimOp);
-      assertCond =
-          i > 0 ? builder.createOrFold<arith::AndIOp>(loc, assertCond, inBounds)
-                : inBounds;
-    }
+    Value cond = generateIndicesInBoundsCheck(
+        builder, loc, loadStoreOp.getMemref(), loadStoreOp.getIndices());
     builder.create<cf::AssertOp>(
-        loc, assertCond,
+        loc, cond,
         RuntimeVerifiableOpInterface::generateErrorMessage(
             op, "out-of-bounds access"));
   }
@@ -421,10 +425,13 @@ void mlir::memref::registerRuntimeVerifiableOpInterfaceExternalModels(
     DialectRegistry &registry) {
   registry.addExtension(+[](MLIRContext *ctx, memref::MemRefDialect *dialect) {
     AssumeAlignmentOp::attachInterface<AssumeAlignmentOpInterface>(*ctx);
+    AtomicRMWOp::attachInterface<LoadStoreOpInterface<AtomicRMWOp>>(*ctx);
     CastOp::attachInterface<CastOpInterface>(*ctx);
     CopyOp::attachInterface<CopyOpInterface>(*ctx);
     DimOp::attachInterface<DimOpInterface>(*ctx);
     ExpandShapeOp::attachInterface<ExpandShapeOpInterface>(*ctx);
+    GenericAtomicRMWOp::attachInterface<
+        LoadStoreOpInterface<GenericAtomicRMWOp>>(*ctx);
     LoadOp::attachInterface<LoadStoreOpInterface<LoadOp>>(*ctx);
     ReinterpretCastOp::attachInterface<ReinterpretCastOpInterface>(*ctx);
     StoreOp::attachInterface<LoadStoreOpInterface<StoreOp>>(*ctx);
diff --git a/mlir/test/Integration/Dialect/MemRef/atomic-rmw-runtime-verification.mlir b/mlir/test/Integration/Dialect/MemRef/atomic-rmw-runtime-verification.mlir
new file mode 100644
index 0000000000000..9f70c5ca66f65
--- /dev/null
+++ b/mlir/test/Integration/Dialect/MemRef/atomic-rmw-runtime-verification.mlir
@@ -0,0 +1,45 @@
+// RUN: mlir-opt %s -generate-runtime-verification \
+// RUN:     -test-cf-assert \
+// RUN:     -expand-strided-metadata \
+// RUN:     -lower-affine \
+// RUN:     -convert-to-llvm | \
+// RUN: mlir-runner -e main -entry-point-result=void \
+// RUN:     -shared-libs=%mlir_runner_utils 2>&1 | \
+// RUN: FileCheck %s
+
+func.func @store_dynamic(%memref: memref<?xf32>, %index: index) {
+  %cst = arith.constant 1.0 : f32
+  memref.atomic_rmw addf %cst, %memref[%index] : (f32, memref<?xf32>) -> f32
+  return
+}
+
+func.func @main() {
+  // Allocate a memref<10xf32>, but disguise it as a memref<5xf32>. This is
+  // necessary because "-test-cf-assert" does not abort the program and we do
+  // not want to segfault when running the test case.
+  %alloc = memref.alloca() : memref<10xf32>
+  %ptr = memref.extract_aligned_pointer_as_index %alloc : memref<10xf32> -> index
+  %ptr_i64 = arith.index_cast %ptr : index to i64
+  %ptr_llvm = llvm.inttoptr %ptr_i64 : i64 to !llvm.ptr
+  %c0 = llvm.mlir.constant(0 : index) : i64
+  %c1 = llvm.mlir.constant(1 : index) : i64
+  %c5 = llvm.mlir.constant(5 : index) : i64
+  %4 = llvm.mlir.poison : !llvm.struct<(ptr, ptr, i64, array<1 x i64>, array<1 x i64>)>
+  %5 = llvm.insertvalue %ptr_llvm, %4[0] : !llvm.struct<(ptr, ptr, i64, array<1 x i64>, array<1 x i64>)>
+  %6 = llvm.insertvalue %ptr_llvm, %5[1] : !llvm.struct<(ptr, ptr, i64, array<1 x i64>, array<1 x i64>)>
+  %8 = llvm.insertvalue %c0, %6[2] : !llvm.struct<(ptr, ptr, i64, array<1 x i64>, array<1 x i64>)>
+  %9 = llvm.insertvalue %c5, %8[3, 0] : !llvm.struct<(ptr, ptr, i64, array<1 x i64>, array<1 x i64>)>
+  %10 = llvm.insertvalue %c1, %9[4, 0] : !llvm.struct<(ptr, ptr, i64, array<1 x i64>, array<1 x i64>)>
+  %buffer = builtin.unrealized_conversion_cast %10 : !llvm.struct<(ptr, ptr, i64, array<1 x i64>, array<1 x i64>)> to memref<5xf32>
+  %cast = memref.cast %buffer : memref<5xf32> to memref<?xf32>
+
+  //      CHECK: ERROR: Runtime op verification failed
+  // CHECK-NEXT: "memref.atomic_rmw"(%{{.*}}, %{{.*}}, %{{.*}}) <{kind = 0 : i64}> : (f32, memref<?xf32>, index) -> f32
+  // CHECK-NEXT: ^ out-of-bounds access
+  // CHECK-NEXT: Location: loc({{.*}})
+  %c9 = arith.constant 9 : index
+  func.call @store_dynamic(%cast, %c9) : (memref<?xf32>, index) -> ()
+
+  return
+}
+
diff --git a/mlir/test/Integration/Dialect/MemRef/store-runtime-verification.mlir b/mlir/test/Integration/Dialect/MemRef/store-runtime-verification.mlir
new file mode 100644
index 0000000000000..58961ba31d93a
--- /dev/null
+++ b/mlir/test/Integration/Dialect/MemRef/store-runtime-verification.mlir
@@ -0,0 +1,45 @@
+// RUN: mlir-opt %s -generate-runtime-verification \
+// RUN:     -test-cf-assert \
+// RUN:     -expand-strided-metadata \
+// RUN:     -lower-affine \
+// RUN:     -convert-to-llvm | \
+// RUN: mlir-runner -e main -entry-point-result=void \
+// RUN:     -shared-libs=%mlir_runner_utils 2>&1 | \
+// RUN: FileCheck %s
+
+func.func @store_dynamic(%memref: memref<?xf32>, %index: index) {
+  %cst = arith.constant 1.0 : f32
+  memref.store %cst, %memref[%index] :  memref<?xf32>
+  return
+}
+
+func.func @main() {
+  // Allocate a memref<10xf32>, but disguise it as a memref<5xf32>. This is
+  // necessary because "-test-cf-assert" does not abort the program and we do
+  // not want to segfault when running the test case.
+  %alloc = memref.alloca() : memref<10xf32>
+  %ptr = memref.extract_aligned_pointer_as_index %alloc : memref<10xf32> -> index
+  %ptr_i64 = arith.index_cast %ptr : index to i64
+  %ptr_llvm = llvm.inttoptr %ptr_i64 : i64 to !llvm.ptr
+  %c0 = llvm.mlir.constant(0 : index) : i64
+  %c1 = llvm.mlir.constant(1 : index) : i64
+  %c5 = llvm.mlir.constant(5 : index) : i64
+  %4 = llvm.mlir.poison : !llvm.struct<(ptr, ptr, i64, array<1 x i64>, array<1 x i64>)>
+  %5 = llvm.insertvalue %ptr_llvm, %4[0] : !llvm.struct<(ptr, ptr, i64, array<1 x i64>, array<1 x i64>)>
+  %6 = llvm.insertvalue %ptr_llvm, %5[1] : !llvm.struct<(ptr, ptr, i64, array<1 x i64>, array<1 x i64>)>
+  %8 = llvm.insertvalue %c0, %6[2] : !llvm.struct<(ptr, ptr, i64, array<1 x i64>, array<1 x i64>)>
+  %9 = llvm.insertvalue %c5, %8[3, 0] : !llvm.struct<(ptr, ptr, i64, array<1 x i64>, array<1 x i64>)>
+  %10 = llvm.insertvalue %c1, %9[4, 0] : !llvm.struct<(ptr, ptr, i64, array<1 x i64>, array<1 x i64>)>
+  %buffer = builtin.unrealized_conversion_cast %10 : !llvm.struct<(ptr, ptr, i64, array<1 x i64>, array<1 x i64>)> to memref<5xf32>
+  %cast = memref.cast %buffer : memref<5xf32> to memref<?xf32>
+
+  //      CHECK: ERROR: Runtime op verification failed
+  // CHECK-NEXT: "memref.store"(%{{.*}}, %{{.*}}, %{{.*}}) : (f32, memref<?xf32>, index) -> ()
+  // CHECK-NEXT: ^ out-of-bounds access
+  // CHECK-NEXT: Location: loc({{.*}})
+  %c9 = arith.constant 9 : index
+  func.call @store_dynamic(%cast, %c9) : (memref<?xf32>, index) -> ()
+
+  return
+}
+

More information about the Mlir-commits mailing list