[MLIR] Expose `getAsValues` in `StaticValueUtils.h` (NFC)

[chelini]

Summary

The utility function should live in StaticValueUtils.h as it provides
a convenient way to convert a vector of OpFoldResults into a vector of
Values.

[ftynse]

mlir/include/mlir/Dialect/Utils/StaticValueUtils.h

83–84

Nit: please document that this expects constant values to be IntegerAttributes and that it will create Index-typed constants.

[chelini]

Update comments.

[chelini]

Fix shared libs build.

[cota]

This change introduces a circular build dependence: DialectUtils <- ArithmeticUtils <- ArithDialect <- DialectUtils.

chelini: can you either revert or break the cycle? I don't see a trivial way of doing the latter.

[chelini]

In D134451#3815941, @cota wrote:

This change introduces a circular build dependence: DialectUtils <- ArithmeticUtils <- ArithDialect <- DialectUtils.

chelini: can you either revert or break the cycle? I don't see a trivial way of doing the latter.

Reverted. Thanks for catching this.

[akuegel]

In D134451#3815941, @cota wrote:

This change introduces a circular build dependence: DialectUtils <- ArithmeticUtils <- ArithDialect <- DialectUtils.

chelini: can you either revert or break the cycle? I don't see a trivial way of doing the latter.

I think you talk about the Bazel build, right? Actually I think the ArithDialect to DialectUtils dependency is unused. Yesterday when this change landed first, I had the Bazel build working by removing that unused dependency.
Please check https://github.com/llvm/llvm-project/commit/ec8f08ce08c26f6db61b142ba40590c3c1d158a0 and https://github.com/llvm/llvm-project/commit/b4cc363e86e7506e41560af333b12beefd9f4bbd

@chelini if you roll forward, please consider updating the Bazel BUILD files like I did in the two revisions mentioned above

[rengolin]

In D134451#3817114, @akuegel wrote:

@chelini if you roll forward, please consider updating the Bazel BUILD files like I did in the two revisions mentioned above

The Bazel build isn't officially supported by the LLVM community and should not be considered a strong motive for reverting patches, nor it should be a requirement that people build with Bazel *before* landing their patches on generic LLVM code.

If the Bazel community wants to have that kind of support in LLVM, then they'll need a new round of RFCs and convince everyone that this is a good idea. IMO, this will not happen any time soon.

@chelini, if this is the last problem with your patch, feel free to re-land it and let the Bazel folks fix this upstream on their end.

cheers,
--renato

[akuegel]

In D134451#3817453, @rengolin wrote:

In D134451#3817114, @akuegel wrote:

@chelini if you roll forward, please consider updating the Bazel BUILD files like I did in the two revisions mentioned above

The Bazel build isn't officially supported by the LLVM community and should not be considered a strong motive for reverting patches, nor it should be a requirement that people build with Bazel *before* landing their patches on generic LLVM code.

If the Bazel community wants to have that kind of support in LLVM, then they'll need a new round of RFCs and convince everyone that this is a good idea. IMO, this will not happen any time soon.

@chelini, if this is the last problem with your patch, feel free to re-land it and let the Bazel folks fix this upstream on their end.

cheers,
--renato

Sorry, I didn't intend to say that it is required to fix the Bazel build when relanding the change. My intent was just to say it would be nice if that is done, given that it is just applying the two patches I pointed to. We can certainly also do this ourselves.

[cota]

@akuegel is right, my apologies. I wrongly assumed that in Cmake, ArithDialect depends on DialectUtils, just like it (unnecessarily) does in Bazel. I confirm that removing that superfluous dependence fixes the cycle in Bazel.

@chelini: please feel free to reland this and I will update the Bazel files.

[cota]

@rengolin: AFAIK we do not want dependency cycles in the build system (Cmake or otherwise). I wrongly thought I was reporting one such case.

[chelini]

Update Bazel build as suggested.

[chelini]

@cota and @akuegel please have a look a the latest changes. If you are happy with them please land it. Thanks!

mlir/include/mlir/Dialect/Utils/StaticValueUtils.h

 bool isConstantIntValue(OpFoldResult ofr, int64_t value);
 
 /// Return true if ofr1 and ofr2 are the same integer constant attribute values
 /// or the same SSA value.
 /// Ignore integer bitwitdh and type mismatch that come from the fact there is
 /// no IndexAttr and that IndexType have no bitwidth.
 bool isEqualConstantIntOrValue(OpFoldResult ofr1, OpFoldResult ofr2);
 
+/// Helper function to convert a vector of `OpFoldResult`s into a vector of
+/// `Value`s. For each `OpFoldResult` in `valueOrAttrVec` return the fold result
+/// if it casts to  a `Value` or create an index-type constant if it casts to
+/// `IntegerAttr`. No other attribute types are supported.
+SmallVector<Value> getAsValues(OpBuilder &b, Location loc,
+                               ArrayRef<OpFoldResult> valueOrAttrVec);
+
 } // namespace mlir
 
 #endif // MLIR_DIALECT_UTILS_STATICVALUEUTILS_H

mlir/lib/Dialect/Tensor/IR/TensorInferTypeOpInterfaceImpl.cpp

 //===- InferTypeOpImpl.cpp - InferType Interface external models *- 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
 //
 //===----------------------------------------------------------------------===//
 
 #include "mlir/Dialect/Tensor/IR/TensorInferTypeOpInterfaceImpl.h"
 #include "mlir/Dialect/Affine/IR/AffineOps.h"
-#include "mlir/Dialect/Arithmetic/Utils/Utils.h"
 #include "mlir/Dialect/Tensor/IR/Tensor.h"
+#include "mlir/Dialect/Utils/StaticValueUtils.h"
 #include "mlir/Interfaces/InferTypeOpInterface.h"
 
 using namespace mlir;
 using namespace mlir::tensor;
 
 /// Compute a map that for a given dimension of the expanded type gives the
 /// dimension in the collapsed type it maps to. Essentially its the inverse of
 /// the `reassocation` maps.
getReshapeOutputShapeFromInputShape(OpBuilder &builder, Location loc, Value src,
   return dstStaticShape.size() >
                  static_cast<size_t>(src.getType().cast<ShapedType>().getRank())
              ? getExpandedOutputShapeFromInputShape(
                    builder, loc, src, dstStaticShape, reassocation)
              : getCollapsedOutputShapeFromInputShape(
                    builder, loc, src, dstStaticShape, reassocation);
 }
 
-/// Helper function to convert a vector of `OpFoldResult`s into a vector of
-/// `Value`s.
-static SmallVector<Value> getAsValues(OpBuilder &b, Location loc,
-                                      ArrayRef<OpFoldResult> valueOrAttrVec) {
-  return llvm::to_vector<4>(
-      llvm::map_range(valueOrAttrVec, [&](OpFoldResult value) -> Value {
-        return getValueOrCreateConstantIndexOp(b, loc, value);
-      }));
-}
-
 template <typename OpTy>
 struct ReifyExpandOrCollapseShapeOp
     : public ReifyRankedShapedTypeOpInterface::ExternalModel<
           ReifyExpandOrCollapseShapeOp<OpTy>, OpTy> {
   LogicalResult
   reifyResultShapes(Operation *op, OpBuilder &b,
                     ReifiedRankedShapedTypeDims &reifiedReturnShapes) const {
     auto loc = op->getLoc();

mlir/lib/Dialect/Utils/CMakeLists.txt

 add_mlir_library(MLIRDialectUtils
   IndexingUtils.cpp
   ReshapeOpsUtils.cpp
   StructuredOpsUtils.cpp
   StaticValueUtils.cpp
 
   LINK_LIBS PUBLIC
+  MLIRArithmeticUtils
   MLIRIR
 )

mlir/lib/Dialect/Utils/StaticValueUtils.cpp

 //===- StaticValueUtils.cpp - Utilities for dealing with static values ----===//
 //
 // 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/Utils/StaticValueUtils.h"
+#include "mlir/Dialect/Arithmetic/Utils/Utils.h"
 #include "mlir/IR/Matchers.h"
 #include "mlir/Support/LLVM.h"
 #include "llvm/ADT/APSInt.h"
 
 namespace mlir {
 
 std::tuple<SmallVector<OpFoldResult>, SmallVector<OpFoldResult>,
            SmallVector<OpFoldResult>>

 /// no IndexAttr and that IndexType has no bitwidth.
 bool isEqualConstantIntOrValue(OpFoldResult ofr1, OpFoldResult ofr2) {
   auto cst1 = getConstantIntValue(ofr1), cst2 = getConstantIntValue(ofr2);
   if (cst1 && cst2 && *cst1 == *cst2)
     return true;
   auto v1 = ofr1.dyn_cast<Value>(), v2 = ofr2.dyn_cast<Value>();
   return v1 && v1 == v2;
 }
+
+/// Helper function to convert a vector of `OpFoldResult`s into a vector of
+/// `Value`s. For each `OpFoldResult` in `valueOrAttrVec` return the fold result
+/// if it casts to  a `Value` or create an index-type constant if it casts to
+/// `IntegerAttr`. No other attribute types are supported.
+SmallVector<Value> getAsValues(OpBuilder &b, Location loc,
+                               ArrayRef<OpFoldResult> valueOrAttrVec) {
+  return llvm::to_vector<4>(
+      llvm::map_range(valueOrAttrVec, [&](OpFoldResult value) -> Value {
+        return getValueOrCreateConstantIndexOp(b, loc, value);
+      }));
+}
 } // namespace mlir

utils/bazel/llvm-project-overlay/mlir/BUILD.bazel

srcs = glob([
         "lib/Dialect/Utils/*.cpp",
         "lib/Dialect/Utils/*.h",
     ]),
     hdrs = glob([
         "include/mlir/Dialect/Utils/*.h",
     ]),
     includes = ["include"],
     deps = [
+        ":ArithmeticUtils",
         ":DialectUtilsIncGen",
         ":IR",
         ":Support",
         "//llvm:Support",
     ],
 )
 
 cc_library(

 
 cc_library(
     name = "TensorInferTypeOpInterfaceImpl",
     srcs = ["lib/Dialect/Tensor/IR/TensorInferTypeOpInterfaceImpl.cpp"],
     hdrs = ["include/mlir/Dialect/Tensor/IR/TensorInferTypeOpInterfaceImpl.h"],
     includes = ["include"],
     deps = [
         ":AffineDialect",
-        ":ArithmeticUtils",
+        ":DialectUtils",
         ":IR",
         ":InferTypeOpInterface",
         ":TensorDialect",
         "//llvm:Support",
     ],
 )
 
 cc_library(
hdrs = [
         "include/mlir/Transforms/InliningUtils.h",
     ],
     includes = ["include"],
     deps = [
         ":ArithmeticBaseIncGen",
         ":ArithmeticCanonicalizationIncGen",
         ":ArithmeticOpsIncGen",
         ":CommonFolders",
-        ":DialectUtils",
         ":IR",
         ":InferIntRangeInterface",
         ":InferTypeOpInterface",
         ":SideEffectInterfaces",
         ":VectorInterfaces",
         "//llvm:Support",
     ],
 )