[mlir][tensor] Add runtime verification for cast/dim/extract/insert/extract_slice
#141332
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
matthias-springer
requested review from
hanhanW and
nicolasvasilache
as code owners
|
@llvm/pr-subscribers-mlir-tensor @llvm/pr-subscribers-mlir Author: Matthias Springer (matthias-springer) ChangesAdd
Patch is 24.25 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/141332.diff 8 Files Affected:
diff --git a/mlir/include/mlir/Dialect/Tensor/Transforms/RuntimeOpVerification.h b/mlir/include/mlir/Dialect/Tensor/Transforms/RuntimeOpVerification.h
new file mode 100644
index 0000000000000..812c4dd6d2cb7
--- /dev/null
+++ b/mlir/include/mlir/Dialect/Tensor/Transforms/RuntimeOpVerification.h
@@ -0,0 +1,21 @@
+//===- RuntimeOpVerification.h - Op Verification ----------------*- C++ -*-===//
+//
+// 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
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef MLIR_DIALECT_TENSOR_RUNTIMEOPVERIFICATION_H
+#define MLIR_DIALECT_TENSOR_RUNTIMEOPVERIFICATION_H
+
+namespace mlir {
+class DialectRegistry;
+
+namespace tensor {
+void registerRuntimeVerifiableOpInterfaceExternalModels(
+ DialectRegistry ®istry);
+} // namespace tensor
+} // namespace mlir
+
+#endif // MLIR_DIALECT_TENSOR_RUNTIMEOPVERIFICATION_H
diff --git a/mlir/include/mlir/InitAllDialects.h b/mlir/include/mlir/InitAllDialects.h
index ea285ac7f16e3..261b0e00bdf86 100644
--- a/mlir/include/mlir/InitAllDialects.h
+++ b/mlir/include/mlir/InitAllDialects.h
@@ -84,6 +84,7 @@
#include "mlir/Dialect/Tensor/IR/ValueBoundsOpInterfaceImpl.h"
#include "mlir/Dialect/Tensor/TransformOps/TensorTransformOps.h"
#include "mlir/Dialect/Tensor/Transforms/BufferizableOpInterfaceImpl.h"
+#include "mlir/Dialect/Tensor/Transforms/RuntimeOpVerification.h"
#include "mlir/Dialect/Tensor/Transforms/SubsetInsertionOpInterfaceImpl.h"
#include "mlir/Dialect/Tosa/IR/ShardingInterfaceImpl.h"
#include "mlir/Dialect/Tosa/IR/TosaOps.h"
@@ -186,6 +187,7 @@ inline void registerAllDialects(DialectRegistry ®istry) {
tensor::registerBufferizableOpInterfaceExternalModels(registry);
tensor::registerFindPayloadReplacementOpInterfaceExternalModels(registry);
tensor::registerInferTypeOpInterfaceExternalModels(registry);
+ tensor::registerRuntimeVerifiableOpInterfaceExternalModels(registry);
tensor::registerSubsetOpInterfaceExternalModels(registry);
tensor::registerTilingInterfaceExternalModels(registry);
tensor::registerValueBoundsOpInterfaceExternalModels(registry);
diff --git a/mlir/lib/Dialect/Tensor/Transforms/CMakeLists.txt b/mlir/lib/Dialect/Tensor/Transforms/CMakeLists.txt
index 7880d1c5a0c5d..ff84463cc507f 100644
--- a/mlir/lib/Dialect/Tensor/Transforms/CMakeLists.txt
+++ b/mlir/lib/Dialect/Tensor/Transforms/CMakeLists.txt
@@ -8,6 +8,7 @@ add_mlir_dialect_library(MLIRTensorTransforms
MergeConsecutiveInsertExtractSlicePatterns.cpp
ReshapePatterns.cpp
RewriteAsConstant.cpp
+ RuntimeOpVerification.cpp
SwapExtractSliceWithProducerPatterns.cpp
SubsetInsertionOpInterfaceImpl.cpp
diff --git a/mlir/lib/Dialect/Tensor/Transforms/RuntimeOpVerification.cpp b/mlir/lib/Dialect/Tensor/Transforms/RuntimeOpVerification.cpp
new file mode 100644
index 0000000000000..6138821ee8c61
--- /dev/null
+++ b/mlir/lib/Dialect/Tensor/Transforms/RuntimeOpVerification.cpp
@@ -0,0 +1,208 @@
+//===- RuntimeOpVerification.cpp - Op Verification ------------------------===//
+//
+// 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
+//
+//===----------------------------------------------------------------------===//
+
+#include "mlir/Dialect/Tensor/Transforms/RuntimeOpVerification.h"
+
+#include "mlir/Dialect/Arith/IR/Arith.h"
+#include "mlir/Dialect/Arith/Utils/Utils.h"
+#include "mlir/Dialect/ControlFlow/IR/ControlFlow.h"
+#include "mlir/Dialect/ControlFlow/IR/ControlFlowOps.h"
+#include "mlir/Dialect/Tensor/IR/Tensor.h"
+#include "mlir/Dialect/Utils/IndexingUtils.h"
+#include "mlir/Interfaces/RuntimeVerifiableOpInterface.h"
+
+using namespace mlir;
+
+namespace mlir {
+namespace tensor {
+namespace {
+/// Generate a runtime check for lb <= value < ub.
+Value generateInBoundsCheck(OpBuilder &builder, Location loc, Value value,
+ Value lb, Value ub) {
+ Value inBounds1 = builder.createOrFold<arith::CmpIOp>(
+ loc, arith::CmpIPredicate::sge, value, lb);
+ Value inBounds2 = builder.createOrFold<arith::CmpIOp>(
+ loc, arith::CmpIPredicate::slt, value, ub);
+ Value inBounds =
+ builder.createOrFold<arith::AndIOp>(loc, inBounds1, inBounds2);
+ return inBounds;
+}
+
+struct CastOpInterface
+ : public RuntimeVerifiableOpInterface::ExternalModel<CastOpInterface,
+ CastOp> {
+ void generateRuntimeVerification(Operation *op, OpBuilder &builder,
+ Location loc) const {
+ auto castOp = cast<CastOp>(op);
+ auto srcType = cast<TensorType>(castOp.getSource().getType());
+
+ // Nothing to check if the result is an unranked tensor.
+ auto resultType = dyn_cast<RankedTensorType>(castOp.getType());
+ if (!resultType)
+ return;
+
+ if (isa<UnrankedTensorType>(srcType)) {
+ // Check rank.
+ Value srcRank = builder.create<RankOp>(loc, castOp.getSource());
+ Value resultRank =
+ builder.create<arith::ConstantIndexOp>(loc, resultType.getRank());
+ Value isSameRank = builder.create<arith::CmpIOp>(
+ loc, arith::CmpIPredicate::eq, srcRank, resultRank);
+ builder.create<cf::AssertOp>(
+ loc, isSameRank,
+ RuntimeVerifiableOpInterface::generateErrorMessage(op,
+ "rank mismatch"));
+ }
+
+ // Check dimension sizes.
+ for (const auto &it : llvm::enumerate(resultType.getShape())) {
+ // Static dim size -> static/dynamic dim size does not need verification.
+ if (auto rankedSrcType = dyn_cast<RankedTensorType>(srcType))
+ if (!rankedSrcType.isDynamicDim(it.index()))
+ continue;
+
+ // Static/dynamic dim size -> dynamic dim size does not need verification.
+ if (resultType.isDynamicDim(it.index()))
+ continue;
+
+ Value srcDimSz =
+ builder.create<DimOp>(loc, castOp.getSource(), it.index());
+ Value resultDimSz =
+ builder.create<arith::ConstantIndexOp>(loc, it.value());
+ Value isSameSz = builder.create<arith::CmpIOp>(
+ loc, arith::CmpIPredicate::eq, srcDimSz, resultDimSz);
+ builder.create<cf::AssertOp>(
+ loc, isSameSz,
+ RuntimeVerifiableOpInterface::generateErrorMessage(
+ op, "size mismatch of dim " + std::to_string(it.index())));
+ }
+ }
+};
+
+struct DimOpInterface
+ : public RuntimeVerifiableOpInterface::ExternalModel<DimOpInterface,
+ DimOp> {
+ void generateRuntimeVerification(Operation *op, OpBuilder &builder,
+ Location loc) const {
+ auto dimOp = cast<DimOp>(op);
+ Value rank = builder.create<RankOp>(loc, dimOp.getSource());
+ Value zero = builder.create<arith::ConstantIndexOp>(loc, 0);
+ builder.create<cf::AssertOp>(
+ loc, generateInBoundsCheck(builder, loc, dimOp.getIndex(), zero, rank),
+ RuntimeVerifiableOpInterface::generateErrorMessage(
+ op, "index is out of bounds"));
+ }
+};
+
+/// Verifies that the indices on extract/insert ops are in-bounds of the
+/// tensor's index space: 0 <= index#i < dim#i
+template <typename OpTy>
+struct ExtractInsertOpInterface
+ : public RuntimeVerifiableOpInterface::ExternalModel<
+ ExtractInsertOpInterface<OpTy>, OpTy> {
+ void generateRuntimeVerification(Operation *op, OpBuilder &builder,
+ Location loc) const {
+ auto extractInsertOp = cast<OpTy>(op);
+
+ Value tensor;
+ if constexpr (std::is_same_v<OpTy, ExtractOp>) {
+ tensor = extractInsertOp.getTensor();
+ } else if constexpr (std::is_same_v<OpTy, InsertOp>) {
+ tensor = extractInsertOp.getDest();
+ } else {
+ llvm_unreachable("invalid op");
+ }
+ auto tensorType = cast<RankedTensorType>(tensor.getType());
+ auto rank = tensorType.getRank();
+ if (rank == 0) {
+ // Nothing to check for 0-d tensors.
+ return;
+ }
+
+ auto indices = extractInsertOp.getIndices();
+ auto zero = builder.create<arith::ConstantIndexOp>(loc, 0);
+ Value assertCond;
+ for (auto i : llvm::seq<int64_t>(0, rank)) {
+ Value dimOp = builder.createOrFold<tensor::DimOp>(loc, tensor, i);
+ Value inBounds =
+ generateInBoundsCheck(builder, loc, indices[i], zero, dimOp);
+ assertCond =
+ i > 0 ? builder.createOrFold<arith::AndIOp>(loc, assertCond, inBounds)
+ : inBounds;
+ }
+ builder.create<cf::AssertOp>(
+ loc, assertCond,
+ RuntimeVerifiableOpInterface::generateErrorMessage(
+ op, "out-of-bounds access"));
+ }
+};
+
+struct ExtractSliceOpInterface
+ : public RuntimeVerifiableOpInterface::ExternalModel<
+ ExtractSliceOpInterface, ExtractSliceOp> {
+ void generateRuntimeVerification(Operation *op, OpBuilder &builder,
+ Location loc) const {
+ auto extractSliceOp = cast<ExtractSliceOp>(op);
+ RankedTensorType sourceType = extractSliceOp.getSource().getType();
+
+ // For each dimension, assert that:
+ // 0 <= offset < dim_size
+ // 0 <= offset + (size - 1) * stride < dim_size
+ Value zero = builder.create<arith::ConstantIndexOp>(loc, 0);
+ Value one = builder.create<arith::ConstantIndexOp>(loc, 1);
+ for (int64_t i = 0, e = sourceType.getRank(); i < e; ++i) {
+ Value offset = getValueOrCreateConstantIndexOp(
+ builder, loc, extractSliceOp.getMixedOffsets()[i]);
+ Value size = getValueOrCreateConstantIndexOp(
+ builder, loc, extractSliceOp.getMixedSizes()[i]);
+ Value stride = getValueOrCreateConstantIndexOp(
+ builder, loc, extractSliceOp.getMixedStrides()[i]);
+
+ // Verify that offset is in-bounds.
+ Value dimSize = builder.createOrFold<tensor::DimOp>(
+ loc, extractSliceOp.getSource(), i);
+ Value offsetInBounds =
+ generateInBoundsCheck(builder, loc, offset, zero, dimSize);
+ builder.create<cf::AssertOp>(
+ loc, offsetInBounds,
+ RuntimeVerifiableOpInterface::generateErrorMessage(
+ op, "offset " + std::to_string(i) + " is out-of-bounds"));
+
+ // Verify that slice does not run out-of-bounds.
+ Value sizeMinusOne = builder.create<arith::SubIOp>(loc, size, one);
+ Value sizeMinusOneTimesStride =
+ builder.create<arith::MulIOp>(loc, sizeMinusOne, stride);
+ Value lastPos =
+ builder.create<arith::AddIOp>(loc, offset, sizeMinusOneTimesStride);
+ Value lastPosInBounds =
+ generateInBoundsCheck(builder, loc, lastPos, zero, dimSize);
+ builder.create<cf::AssertOp>(
+ loc, lastPosInBounds,
+ RuntimeVerifiableOpInterface::generateErrorMessage(
+ op, "extract_slice runs out-of-bounds along dimension " +
+ std::to_string(i)));
+ }
+ }
+};
+} // namespace
+} // namespace tensor
+} // namespace mlir
+
+void mlir::tensor::registerRuntimeVerifiableOpInterfaceExternalModels(
+ DialectRegistry ®istry) {
+ registry.addExtension(+[](MLIRContext *ctx, tensor::TensorDialect *dialect) {
+ CastOp::attachInterface<CastOpInterface>(*ctx);
+ DimOp::attachInterface<DimOpInterface>(*ctx);
+ ExtractOp::attachInterface<ExtractInsertOpInterface<ExtractOp>>(*ctx);
+ ExtractSliceOp::attachInterface<ExtractSliceOpInterface>(*ctx);
+ InsertOp::attachInterface<ExtractInsertOpInterface<InsertOp>>(*ctx);
+
+ // Load additional dialects of which ops may get created.
+ ctx->loadDialect<arith::ArithDialect, cf::ControlFlowDialect>();
+ });
+}
diff --git a/mlir/test/Integration/Dialect/Tensor/cast-runtime-verification.mlir b/mlir/test/Integration/Dialect/Tensor/cast-runtime-verification.mlir
new file mode 100644
index 0000000000000..e4aab32d4a390
--- /dev/null
+++ b/mlir/test/Integration/Dialect/Tensor/cast-runtime-verification.mlir
@@ -0,0 +1,50 @@
+// RUN: mlir-opt %s -generate-runtime-verification \
+// RUN: -one-shot-bufferize="bufferize-function-boundaries" \
+// RUN: -buffer-deallocation-pipeline=private-function-dynamic-ownership \
+// RUN: -test-cf-assert \
+// RUN: -convert-scf-to-cf \
+// RUN: -convert-to-llvm | \
+// RUN: mlir-runner -e main -entry-point-result=void \
+// RUN: -shared-libs=%tlir_runner_utils 2>&1 | \
+// RUN: FileCheck %s
+
+func.func private @cast_to_static_dim(%t: tensor<?xf32>) -> tensor<10xf32> {
+ %0 = tensor.cast %t : tensor<?xf32> to tensor<10xf32>
+ return %0 : tensor<10xf32>
+}
+
+func.func private @cast_to_ranked(%t: tensor<*xf32>) -> tensor<f32> {
+ %0 = tensor.cast %t : tensor<*xf32> to tensor<f32>
+ return %0 : tensor<f32>
+}
+
+func.func private @valid_cast(%t: tensor<*xf32>) -> tensor<?xf32> {
+ %0 = tensor.cast %t : tensor<*xf32> to tensor<?xf32>
+ return %0 : tensor<?xf32>
+}
+
+func.func @main() {
+ // All casts inside the called functions are invalid at runtime, except for
+ // the last one.
+ %alloc = tensor.empty() : tensor<5xf32>
+
+ // CHECK: ERROR: Runtime op verification failed
+ // CHECK-NEXT: "tensor.cast"(%{{.*}}) : (tensor<?xf32>) -> tensor<10xf32>
+ // CHECK-NEXT: ^ size mismatch of dim 0
+ // CHECK-NEXT: Location: loc({{.*}})
+ %1 = tensor.cast %alloc : tensor<5xf32> to tensor<?xf32>
+ func.call @cast_to_static_dim(%1) : (tensor<?xf32>) -> (tensor<10xf32>)
+
+ // CHECK-NEXT: ERROR: Runtime op verification failed
+ // CHECK-NEXT: "tensor.cast"(%{{.*}}) : (tensor<*xf32>) -> tensor<f32>
+ // CHECK-NEXT: ^ rank mismatch
+ // CHECK-NEXT: Location: loc({{.*}})
+ %3 = tensor.cast %alloc : tensor<5xf32> to tensor<*xf32>
+ func.call @cast_to_ranked(%3) : (tensor<*xf32>) -> (tensor<f32>)
+
+ // A last cast that actually succeeds.
+ // CHECK-NOT: ERROR: Runtime op verification failed
+ func.call @valid_cast(%3) : (tensor<*xf32>) -> (tensor<?xf32>)
+
+ return
+}
diff --git a/mlir/test/Integration/Dialect/Tensor/dim-runtime-verification.mlir b/mlir/test/Integration/Dialect/Tensor/dim-runtime-verification.mlir
new file mode 100644
index 0000000000000..c6d8f698b9433
--- /dev/null
+++ b/mlir/test/Integration/Dialect/Tensor/dim-runtime-verification.mlir
@@ -0,0 +1,21 @@
+// RUN: mlir-opt %s -generate-runtime-verification \
+// RUN: -one-shot-bufferize \
+// RUN: -buffer-deallocation-pipeline \
+// RUN: -test-cf-assert \
+// 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 @main() {
+ %c4 = arith.constant 4 : index
+ %tensor = tensor.empty() : tensor<1xf32>
+
+ // CHECK: ERROR: Runtime op verification failed
+ // CHECK-NEXT: "tensor.dim"(%{{.*}}, %{{.*}}) : (tensor<1xf32>, index) -> index
+ // CHECK-NEXT: ^ index is out of bounds
+ // CHECK-NEXT: Location: loc({{.*}})
+ %dim = tensor.dim %tensor, %c4 : tensor<1xf32>
+
+ return
+}
diff --git a/mlir/test/Integration/Dialect/Tensor/extract-runtime-verification.mlir b/mlir/test/Integration/Dialect/Tensor/extract-runtime-verification.mlir
new file mode 100644
index 0000000000000..8e3cab7be704d
--- /dev/null
+++ b/mlir/test/Integration/Dialect/Tensor/extract-runtime-verification.mlir
@@ -0,0 +1,64 @@
+// RUN: mlir-opt %s -generate-runtime-verification \
+// RUN: -one-shot-bufferize="bufferize-function-boundaries" \
+// RUN: -buffer-deallocation-pipeline=private-function-dynamic-ownership \
+// RUN: -test-cf-assert \
+// RUN: -convert-scf-to-cf \
+// RUN: -convert-to-llvm | \
+// RUN: mlir-runner -e main -entry-point-result=void \
+// RUN: -shared-libs=%tlir_runner_utils 2>&1 | \
+// RUN: FileCheck %s
+
+func.func @extract(%tensor: tensor<1xf32>, %index: index) {
+ tensor.extract %tensor[%index] : tensor<1xf32>
+ return
+}
+
+func.func @extract_dynamic(%tensor: tensor<?xf32>, %index: index) {
+ tensor.extract %tensor[%index] : tensor<?xf32>
+ return
+}
+
+func.func @extract_nd_dynamic(%tensor: tensor<?x?x?xf32>, %index0: index, %index1: index, %index2: index) {
+ tensor.extract %tensor[%index0, %index1, %index2] : tensor<?x?x?xf32>
+ return
+}
+
+func.func @main() {
+ %0 = arith.constant 0 : index
+ %1 = arith.constant 1 : index
+ %n1 = arith.constant -1 : index
+ %2 = arith.constant 2 : index
+ %alloca_1 = tensor.empty() : tensor<1xf32>
+ %alloc_1 = tensor.empty(%1) : tensor<?xf32>
+ %alloc_2x2x2 = tensor.empty(%2, %2, %2) : tensor<?x?x?xf32>
+
+ // CHECK: ERROR: Runtime op verification failed
+ // CHECK-NEXT: "tensor.extract"(%{{.*}}, %{{.*}}) : (tensor<1xf32>, index) -> f32
+ // CHECK-NEXT: ^ out-of-bounds access
+ // CHECK-NEXT: Location: loc({{.*}})
+ func.call @extract(%alloca_1, %1) : (tensor<1xf32>, index) -> ()
+
+ // CHECK: ERROR: Runtime op verification failed
+ // CHECK-NEXT: "tensor.extract"(%{{.*}}, %{{.*}}) : (tensor<?xf32>, index) -> f32
+ // CHECK-NEXT: ^ out-of-bounds access
+ // CHECK-NEXT: Location: loc({{.*}})
+ func.call @extract_dynamic(%alloc_1, %1) : (tensor<?xf32>, index) -> ()
+
+ // CHECK: ERROR: Runtime op verification failed
+ // CHECK-NEXT: "tensor.extract"(%{{.*}}, %{{.*}}) : (tensor<?x?x?xf32>, index, index, index) -> f32
+ // CHECK-NEXT: ^ out-of-bounds access
+ // CHECK-NEXT: Location: loc({{.*}})
+ func.call @extract_nd_dynamic(%alloc_2x2x2, %1, %n1, %0) : (tensor<?x?x?xf32>, index, index, index) -> ()
+
+ // CHECK-NOT: ERROR: Runtime op verification failed
+ func.call @extract(%alloca_1, %0) : (tensor<1xf32>, index) -> ()
+
+ // CHECK-NOT: ERROR: Runtime op verification failed
+ func.call @extract_dynamic(%alloc_1, %0) : (tensor<?xf32>, index) -> ()
+
+ // CHECK-NOT: ERROR: Runtime op verification failed
+ func.call @extract_nd_dynamic(%alloc_2x2x2, %1, %1, %0) : (tensor<?x?x?xf32>, index, index, index) -> ()
+
+ return
+}
+
diff --git a/mlir/test/Integration/Dialect/Tensor/extract_slice-runtime-verification.mlir b/mlir/test/Integration/Dialect/Tensor/extract_slice-runtime-verification.mlir
new file mode 100644
index 0000000000000..28f9be0fffe64
--- /dev/null
+++ b/mlir/test/Integration/Dialect/Tensor/extract_slice-runtime-verification.mlir
@@ -0,0 +1,95 @@
+// RUN: mlir-opt %s -generate-runtime-verification \
+// RUN: -one-shot-bufferize="bufferize-function-boundaries" \
+// RUN: -buffer-deallocation-pipeline=private-function-dynamic-ownership \
+// RUN: -test-cf-assert \
+// RUN: -convert-scf-to-cf \
+// RUN: -convert-to-llvm | \
+// RUN: mlir-runner -e main -entry-point-result=void \
+// RUN: -shared-libs=%tlir_runner_utils 2>&1 | \
+// RUN: FileCheck %s
+
+func.func @extract_slice(%tensor: tensor<1xf32>, %offset: index) {
+ tensor.extract_slice %tensor[%offset] [1] [1] : tensor<1xf32> to tensor<1xf32>
+ return
+}
+
+func.func @extract_slice_dynamic(%tensor: tensor<?x4xf32>, %offset: index, %size: index, %stride: index) {
+ tensor.extract_slice %tensor[%offset, 0] [%size, 4] [%stride, 1] : tensor<?x4xf32> to tensor<?x4xf32>
+ return
+}
+
+func.func @extract_slice_dynamic_rank_reduce(%tensor: tensor<?x4xf32>, %offset: index, %size: index, %stride: index) {
+ tensor.extract_slice %tensor[%offset, 0] [%size, 1] [%stride, 1] : tensor<?x4xf32> to tensor<?xf32>
+ return
+}
+
+func.func @main() {
+ %0 = arith.constant 0 : index
+ %1 = arith.constant 1 : index
+ %n1 = arith.constant -1 : index
+ %4 = arith.constant 4 : index
+ %5 = arith.constant 5 : index
+
+ %alloca = tensor.empty() : tensor<1xf32>
+ %alloca_4 = tensor.empty() : tensor<4x4xf32>
+ %alloca_4_dyn = tensor.cast %alloca_4 : tensor<4x4xf32> to tensor<?x4xf32>
+
+ // Offset is out-of-bounds and slice runs out-of-bounds
+ // CHECK: ERROR: Runtime op verification failed
+ // CHECK-NEXT: "tensor.extract_slice"(%arg0, %arg1, %arg2, %arg3) <{operandSegmentSizes = array<i32: 1, 1, 1, 1>, static_offsets = array<i64: -9223372036854775808, 0>, static_sizes = array<i64: -9223372036854775808, 1>, static_strides = array<i64: -9223372036854775808, 1>}> : (tensor<?x4xf32>, index, index, index) -> tensor<?xf32>
+ // CHECK-NEXT: ^ offset 0 is out-of-bounds
+ // CHECK-NEXT: Location: loc({{.*}})
+ // CH...
[truncated]
|
matthias-springer
force-pushed
the
users/matthias-springer/tensor_runtime_verification
branch
from
2045238 to
6367209
Compare
chelini
reviewed
May 26, 2025
chelini
approved these changes
May 28, 2025
…nsert`/`extract_slice`
matthias-springer
force-pushed
the
users/matthias-springer/tensor_runtime_verification
branch
from
6367209 to
9f7686b
Compare
matthias-springer
deleted the
users/matthias-springer/tensor_runtime_verification
branch
rupprecht
added a commit
to rupprecht/llvm-project
that referenced
this pull request
Jun 5, 2025
Commit e4c8ff9
rupprecht
added a commit
that referenced
this pull request
Jun 5, 2025
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/169/builds/12000 Here is the relevant piece of the build log for the reference |
rorth
pushed a commit
to rorth/llvm-project
that referenced
this pull request
Jun 11, 2025
…nsert`/`extract_slice` (llvm#141332) Add `RuntimeVerifiableOpInterface` implementations for the following ops. These were mostly copied from the respective memref implementations. Only the part that deals with offsets and strides was removed. * `tensor.cast`: `memref.cast` * `tensor.dim`: `memref.dim` * `tensor.extract`: `memref.load` * `tensor.insert`: `memref.store` * `tensor.extract_slice`: `memref.subview`
rorth
pushed a commit
to rorth/llvm-project
that referenced
this pull request
Jun 11, 2025
DhruvSrivastavaX
pushed a commit
to DhruvSrivastavaX/lldb-for-aix
that referenced
this pull request
Jun 12, 2025
…nsert`/`extract_slice` (llvm#141332) Add `RuntimeVerifiableOpInterface` implementations for the following ops. These were mostly copied from the respective memref implementations. Only the part that deals with offsets and strides was removed. * `tensor.cast`: `memref.cast` * `tensor.dim`: `memref.dim` * `tensor.extract`: `memref.load` * `tensor.insert`: `memref.store` * `tensor.extract_slice`: `memref.subview`
DhruvSrivastavaX
pushed a commit
to DhruvSrivastavaX/lldb-for-aix
that referenced
this pull request
Jun 12, 2025
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Add
RuntimeVerifiableOpInterfaceimplementations for the following ops. These were mostly copied from the respective memref implementations. Only the part that deals with offsets and strides was removed.tensor.cast:memref.casttensor.dim:memref.dimtensor.extract:memref.loadtensor.insert:memref.storetensor.extract_slice:memref.subview