[mlir][affine] Define affine.linearize_index (#114480)

krzysz00 · kuhar · web-flow · commit 847d50791c07 · 2024-11-05T10:44:02.000-06:00
`affine.linearize_index` is the inverse of `affine.delinearize_index` and general useful for representing computations (like those needed to move from N-D to 1-D memrefs) that put together indices. This commit introduces `affine.linearize_index` and one simple canonicalization for it. There are plans to add `affine.linearize_index` and `affine.delinearize_index` pair canonicalizations, but we are saving those for a followup PR (especially since having #113846 landed would make them nicer). Note while `affine` may not be the natural home for this operation, https://discourse.llvm.org/t/better-location-of-affine-delinearize-operation/80565/13 didn't come to any better consensus location. --------- Co-authored-by: Jakub Kuderski <kubakuderski@gmail.com>
diff --git a/mlir/include/mlir/Dialect/Affine/IR/AffineOps.td b/mlir/include/mlir/Dialect/Affine/IR/AffineOps.td
@@ -1113,4 +1113,73 @@ def AffineDelinearizeIndexOp : Affine_Op<"delinearize_index",
   let hasCanonicalizer = 1;
 }
 
+//===----------------------------------------------------------------------===//
+// AffineLinearizeIndexOp
+//===----------------------------------------------------------------------===//
+def AffineLinearizeIndexOp : Affine_Op<"linearize_index",
+    [Pure, AttrSizedOperandSegments]> {
+  let summary = "linearize an index";
+  let description = [{
+    The `affine.linearize_index` operation takes a sequence of index values and a
+    basis of the same length and linearizes the indices using that basis.
+
+    That is, for indices `%idx_1` through `%idx_N` and basis elements `b_1` through `b_N`,
+    it computes
+
+    ```
+    sum(i = 1 to N) %idx_i * product(j = i + 1 to N) B_j
+    ```
+
+    If the `disjoint` property is present, this is an optimization hint that,
+    for all `i`, `0 <= %idx_i < B_i` - that is, no index affects any other index,
+    except that `%idx_0` may be negative to make the index as a whole negative.
+
+    Note that the outputs of `affine.delinearize_index` are, by definition, `disjoint`.
+
+    Example:
+
+    ```mlir
+    %linear_index = affine.linearize_index [%index_0, %index_1, %index_2] (2, 3, 5) : index
+    ```
+
+    In the above example, `%linear_index` conceptually holds the following:
+
+    ```mlir
+    #map = affine_map<()[s0, s1, s2] -> (s0 * 15 + s1 * 5 + s2)>
+    %linear_index = affine.apply #map()[%index_0, %index_1, %index_2]
+    ```
+  }];
+
+  let arguments = (ins Variadic<Index>:$multi_index,
+    Variadic<Index>:$dynamic_basis,
+    DenseI64ArrayAttr:$static_basis,
+    UnitProperty:$disjoint);
+  let results = (outs Index:$linear_index);
+
+  let assemblyFormat = [{
+    (`disjoint` $disjoint^)? ` `
+    `[` $multi_index `]` `by` ` `
+    custom<DynamicIndexList>($dynamic_basis, $static_basis, "::mlir::AsmParser::Delimiter::Paren")
+    attr-dict `:` type($linear_index)
+  }];
+
+  let builders = [
+    OpBuilder<(ins "ValueRange":$multi_index, "ValueRange":$basis, CArg<"bool", "false">:$disjoint)>,
+    OpBuilder<(ins "ValueRange":$multi_index, "ArrayRef<OpFoldResult>":$basis, CArg<"bool", "false">:$disjoint)>,
+    OpBuilder<(ins "ValueRange":$multi_index, "ArrayRef<int64_t>":$basis, CArg<"bool", "false">:$disjoint)>
+  ];
+
+  let extraClassDeclaration = [{
+    /// Return a vector with all the static and dynamic basis values.
+    SmallVector<OpFoldResult> getMixedBasis() {
+      OpBuilder builder(getContext());
+      return ::mlir::getMixedValues(getStaticBasis(), getDynamicBasis(), builder);
+    }
+
+  }];
+
+  let hasVerifier = 1;
+  let hasCanonicalizer = 1;
+}
+
 #endif // AFFINE_OPS
diff --git a/mlir/include/mlir/Dialect/Affine/Utils.h b/mlir/include/mlir/Dialect/Affine/Utils.h
@@ -315,12 +315,17 @@ FailureOr<SmallVector<Value>> delinearizeIndex(OpBuilder &b, Location loc,
 FailureOr<SmallVector<Value>> delinearizeIndex(OpBuilder &b, Location loc,
                                                Value linearIndex,
                                                ArrayRef<OpFoldResult> basis);
+
 // Generate IR that extracts the linear index from a multi-index according to
 // a basis/shape.
 OpFoldResult linearizeIndex(ArrayRef<OpFoldResult> multiIndex,
                             ArrayRef<OpFoldResult> basis,
                             ImplicitLocOpBuilder &builder);
 
+OpFoldResult linearizeIndex(OpBuilder &builder, Location loc,
+                            ArrayRef<OpFoldResult> multiIndex,
+                            ArrayRef<OpFoldResult> basis);
+
 /// Ensure that all operations that could be executed after `start`
 /// (noninclusive) and prior to `memOp` (e.g. on a control flow/op path
 /// between the operations) do not have the potential memory effect
diff --git a/mlir/lib/Dialect/Affine/IR/AffineOps.cpp b/mlir/lib/Dialect/Affine/IR/AffineOps.cpp
@@ -4684,6 +4684,115 @@ void affine::AffineDelinearizeIndexOp::getCanonicalizationPatterns(
   patterns.insert<DropDelinearizeOfSingleLoop, DropUnitExtentBasis>(context);
 }
 
+//===----------------------------------------------------------------------===//
+// LinearizeIndexOp
+//===----------------------------------------------------------------------===//
+
+void AffineLinearizeIndexOp::build(OpBuilder &odsBuilder,
+                                   OperationState &odsState,
+                                   ValueRange multiIndex, ValueRange basis,
+                                   bool disjoint) {
+  SmallVector<Value> dynamicBasis;
+  SmallVector<int64_t> staticBasis;
+  dispatchIndexOpFoldResults(getAsOpFoldResult(basis), dynamicBasis,
+                             staticBasis);
+  build(odsBuilder, odsState, multiIndex, dynamicBasis, staticBasis, disjoint);
+}
+
+void AffineLinearizeIndexOp::build(OpBuilder &odsBuilder,
+                                   OperationState &odsState,
+                                   ValueRange multiIndex,
+                                   ArrayRef<OpFoldResult> basis,
+                                   bool disjoint) {
+  SmallVector<Value> dynamicBasis;
+  SmallVector<int64_t> staticBasis;
+  dispatchIndexOpFoldResults(basis, dynamicBasis, staticBasis);
+  build(odsBuilder, odsState, multiIndex, dynamicBasis, staticBasis, disjoint);
+}
+
+void AffineLinearizeIndexOp::build(OpBuilder &odsBuilder,
+                                   OperationState &odsState,
+                                   ValueRange multiIndex,
+                                   ArrayRef<int64_t> basis, bool disjoint) {
+  build(odsBuilder, odsState, multiIndex, ValueRange{}, basis, disjoint);
+}
+
+LogicalResult AffineLinearizeIndexOp::verify() {
+  if (getStaticBasis().empty())
+    return emitOpError("basis should not be empty");
+
+  if (getMultiIndex().size() != getStaticBasis().size())
+    return emitOpError("should be passed an index for each basis element");
+
+  auto dynamicMarkersCount =
+      llvm::count_if(getStaticBasis(), ShapedType::isDynamic);
+  if (static_cast<size_t>(dynamicMarkersCount) != getDynamicBasis().size())
+    return emitOpError(
+        "mismatch between dynamic and static basis (kDynamic marker but no "
+        "corresponding dynamic basis entry) -- this can only happen due to an "
+        "incorrect fold/rewrite");
+
+  return success();
+}
+
+namespace {
+/// Rewrite `affine.linearize_index disjoint [%...a, %x, %...b] by (%...c, 1,
+/// %...d)` to `affine.linearize_index disjoint [%...a, %...b] by (%...c,
+/// %...d)`.
+
+/// Note that `disjoint` is required here, because, without it, we could have
+/// `affine.linearize_index [%...a, %c64, %...b] by (%...c, 1, %...d)`
+/// is a valid operation where the `%c64` cannot be trivially dropped.
+///
+/// Alternatively, if `%x` in the above is a known constant 0, remove it even if
+/// the operation isn't asserted to be `disjoint`.
+struct DropLinearizeUnitComponentsIfDisjointOrZero final
+    : OpRewritePattern<affine::AffineLinearizeIndexOp> {
+  using OpRewritePattern::OpRewritePattern;
+
+  LogicalResult matchAndRewrite(affine::AffineLinearizeIndexOp op,
+                                PatternRewriter &rewriter) const override {
+    size_t numIndices = op.getMultiIndex().size();
+    SmallVector<Value> newIndices;
+    newIndices.reserve(numIndices);
+    SmallVector<OpFoldResult> newBasis;
+    newBasis.reserve(numIndices);
+
+    SmallVector<OpFoldResult> basis = op.getMixedBasis();
+    for (auto [index, basisElem] : llvm::zip_equal(op.getMultiIndex(), basis)) {
+      std::optional<int64_t> basisEntry = getConstantIntValue(basisElem);
+      if (!basisEntry || *basisEntry != 1) {
+        newIndices.push_back(index);
+        newBasis.push_back(basisElem);
+        continue;
+      }
+
+      std::optional<int64_t> indexValue = getConstantIntValue(index);
+      if (!op.getDisjoint() && (!indexValue || *indexValue != 0)) {
+        newIndices.push_back(index);
+        newBasis.push_back(basisElem);
+        continue;
+      }
+    }
+    if (newIndices.size() == numIndices)
+      return failure();
+
+    if (newIndices.size() == 0) {
+      rewriter.replaceOpWithNewOp<arith::ConstantIndexOp>(op, 0);
+      return success();
+    }
+    rewriter.replaceOpWithNewOp<affine::AffineLinearizeIndexOp>(
+        op, newIndices, newBasis, op.getDisjoint());
+    return success();
+  }
+};
+} // namespace
+
+void affine::AffineLinearizeIndexOp::getCanonicalizationPatterns(
+    RewritePatternSet &patterns, MLIRContext *context) {
+  patterns.add<DropLinearizeUnitComponentsIfDisjointOrZero>(context);
+}
+
 //===----------------------------------------------------------------------===//
 // TableGen'd op method definitions
 //===----------------------------------------------------------------------===//
diff --git a/mlir/lib/Dialect/Affine/Transforms/AffineExpandIndexOps.cpp b/mlir/lib/Dialect/Affine/Transforms/AffineExpandIndexOps.cpp
@@ -15,6 +15,7 @@
 #include "mlir/Dialect/Affine/IR/AffineOps.h"
 #include "mlir/Dialect/Affine/Transforms/Transforms.h"
 #include "mlir/Dialect/Affine/Utils.h"
+#include "mlir/Dialect/Arith/Utils/Utils.h"
 #include "mlir/Transforms/GreedyPatternRewriteDriver.h"
 
 namespace mlir {
@@ -44,6 +45,23 @@ struct LowerDelinearizeIndexOps
   }
 };
 
+/// Lowers `affine.linearize_index` into a sequence of multiplications and
+/// additions.
+struct LowerLinearizeIndexOps final : OpRewritePattern<AffineLinearizeIndexOp> {
+  using OpRewritePattern::OpRewritePattern;
+  LogicalResult matchAndRewrite(AffineLinearizeIndexOp op,
+                                PatternRewriter &rewriter) const override {
+    SmallVector<OpFoldResult> multiIndex =
+        getAsOpFoldResult(op.getMultiIndex());
+    OpFoldResult linearIndex =
+        linearizeIndex(rewriter, op.getLoc(), multiIndex, op.getMixedBasis());
+    Value linearIndexValue =
+        getValueOrCreateConstantIntOp(rewriter, op.getLoc(), linearIndex);
+    rewriter.replaceOp(op, linearIndexValue);
+    return success();
+  }
+};
+
 class ExpandAffineIndexOpsPass
     : public affine::impl::AffineExpandIndexOpsBase<ExpandAffineIndexOpsPass> {
 public:
@@ -63,7 +81,8 @@ class ExpandAffineIndexOpsPass
 
 void mlir::affine::populateAffineExpandIndexOpsPatterns(
     RewritePatternSet &patterns) {
-  patterns.insert<LowerDelinearizeIndexOps>(patterns.getContext());
+  patterns.insert<LowerDelinearizeIndexOps, LowerLinearizeIndexOps>(
+      patterns.getContext());
 }
 
 std::unique_ptr<Pass> mlir::affine::createAffineExpandIndexOpsPass() {
diff --git a/mlir/lib/Dialect/Affine/Utils/Utils.cpp b/mlir/lib/Dialect/Affine/Utils/Utils.cpp
@@ -1999,6 +1999,12 @@ mlir::affine::delinearizeIndex(OpBuilder &b, Location loc, Value linearIndex,
 OpFoldResult mlir::affine::linearizeIndex(ArrayRef<OpFoldResult> multiIndex,
                                           ArrayRef<OpFoldResult> basis,
                                           ImplicitLocOpBuilder &builder) {
+  return linearizeIndex(builder, builder.getLoc(), multiIndex, basis);
+}
+
+OpFoldResult mlir::affine::linearizeIndex(OpBuilder &builder, Location loc,
+                                          ArrayRef<OpFoldResult> multiIndex,
+                                          ArrayRef<OpFoldResult> basis) {
   assert(multiIndex.size() == basis.size());
   SmallVector<AffineExpr> basisAffine;
   for (size_t i = 0; i < basis.size(); ++i) {
@@ -2009,13 +2015,13 @@ OpFoldResult mlir::affine::linearizeIndex(ArrayRef<OpFoldResult> multiIndex,
   SmallVector<OpFoldResult> strides;
   strides.reserve(stridesAffine.size());
   llvm::transform(stridesAffine, std::back_inserter(strides),
-                  [&builder, &basis](AffineExpr strideExpr) {
+                  [&builder, &basis, loc](AffineExpr strideExpr) {
                     return affine::makeComposedFoldedAffineApply(
-                        builder, builder.getLoc(), strideExpr, basis);
+                        builder, loc, strideExpr, basis);
                   });
 
   auto &&[linearIndexExpr, multiIndexAndStrides] = computeLinearIndex(
       OpFoldResult(builder.getIndexAttr(0)), strides, multiIndex);
-  return affine::makeComposedFoldedAffineApply(
-      builder, builder.getLoc(), linearIndexExpr, multiIndexAndStrides);
+  return affine::makeComposedFoldedAffineApply(builder, loc, linearIndexExpr,
+                                               multiIndexAndStrides);
 }
diff --git a/mlir/test/Conversion/AffineToStandard/lower-affine.mlir b/mlir/test/Conversion/AffineToStandard/lower-affine.mlir
@@ -976,3 +976,20 @@ func.func @test_dilinearize_index(%linear_index: index) -> (index, index, index)
 // CHECK:           %[[VAL_38:.*]] = arith.select %[[VAL_36]], %[[VAL_37]], %[[VAL_34]] : index
 // CHECK:           return %[[VAL_11]], %[[VAL_32]], %[[VAL_38]] : index, index, index
 // CHECK:         }
+
+/////////////////////////////////////////////////////////////////////
+
+func.func @test_linearize_index(%arg0: index, %arg1: index, %arg2: index) -> index {
+  %ret = affine.linearize_index disjoint [%arg0, %arg1, %arg2] by (2, 3, 5) : index
+  return %ret : index
+}
+
+// CHECK-LABEL: @test_linearize_index
+// CHECK-SAME: (%[[arg0:.+]]: index, %[[arg1:.+]]: index, %[[arg2:.+]]: index)
+// CHECK: %[[c15:.+]] = arith.constant 15 : index
+// CHECK-NEXT: %[[tmp0:.+]] = arith.muli %[[arg0]], %[[c15]] : index
+// CHECK-NEXT: %[[c5:.+]] = arith.constant 5 : index
+// CHECK-NEXT: %[[tmp1:.+]] = arith.muli %[[arg1]], %[[c5]] : index
+// CHECK-NEXT: %[[tmp2:.+]] = arith.addi %[[tmp0]], %[[tmp1]] : index
+// CHECK-NEXT: %[[ret:.+]] = arith.addi %[[tmp2]], %[[arg2]] : index
+// CHECK-NEXT: return %[[ret]]
diff --git a/mlir/test/Dialect/Affine/affine-expand-index-ops.mlir b/mlir/test/Dialect/Affine/affine-expand-index-ops.mlir
@@ -41,3 +41,29 @@ func.func @dynamic_basis(%linear_index: index, %src: memref<?x?x?xf32>) -> (inde
   %1:3 = affine.delinearize_index %linear_index into (%b0, %b1, %b2) : index, index, index
   return %1#0, %1#1, %1#2 : index, index, index
 }
+
+// -----
+
+// CHECK-DAG: #[[$map0:.+]] = affine_map<()[s0, s1, s2] -> (s0 * 15 + s1 * 5 + s2)>
+
+// CHECK-LABEL: @linearize_static
+// CHECK-SAME: (%[[arg0:.+]]: index, %[[arg1:.+]]: index, %[[arg2:.+]]: index)
+// CHECK: %[[val_0:.+]] = affine.apply #[[$map0]]()[%[[arg0]], %[[arg1]], %[[arg2]]]
+// CHECK: return %[[val_0]]
+func.func @linearize_static(%arg0: index, %arg1: index, %arg2: index) -> index {
+  %0 = affine.linearize_index [%arg0, %arg1, %arg2] by (2, 3, 5) : index
+  func.return %0 : index
+}
+
+// -----
+
+// CHECK-DAG: #[[$map0:.+]] =  affine_map<()[s0, s1, s2, s3, s4] -> (s1 * s2 + s3 + s0 * (s2 * s4))>
+
+// CHECK-LABEL: @linearize_dynamic
+// CHECK-SAME: (%[[arg0:.+]]: index, %[[arg1:.+]]: index, %[[arg2:.+]]: index, %[[arg3:.+]]: index, %[[arg4:.+]]: index, %[[arg5:.+]]: index)
+// CHECK: %[[val_0:.+]] = affine.apply #[[$map0]]()[%[[arg0]], %[[arg1]], %[[arg5]], %[[arg2]], %[[arg4]]]
+// CHECK: return %[[val_0]]
+func.func @linearize_dynamic(%arg0: index, %arg1: index, %arg2: index, %arg3: index, %arg4: index, %arg5: index) -> index {
+  %0 = affine.linearize_index [%arg0, %arg1, %arg2] by (%arg3, %arg4, %arg5) : index
+  func.return %0 : index
+}
diff --git a/mlir/test/Dialect/Affine/canonicalize.mlir b/mlir/test/Dialect/Affine/canonicalize.mlir
@@ -1530,3 +1530,37 @@ func.func @delinearize_non_loop_like(%arg0: memref<?xi32>, %i : index) -> index
   %2 = affine.delinearize_index %i into (1024) : index
   return %2 : index
 }
+
+// -----
+
+// CHECK-LABEL: @linearize_unit_basis_disjoint
+// CHECK-SAME: (%[[arg0:.+]]: index, %[[arg1:.+]]: index, %[[arg2:.+]]: index, %[[arg3:.+]]: index)
+// CHECK: %[[ret:.+]] = affine.linearize_index disjoint [%[[arg0]], %[[arg2]]] by (3, %[[arg3]]) : index
+// CHECK: return %[[ret]]
+func.func @linearize_unit_basis_disjoint(%arg0: index, %arg1: index, %arg2: index, %arg3: index) -> index {
+  %ret = affine.linearize_index disjoint [%arg0, %arg1, %arg2] by (3, 1, %arg3) : index
+  return %ret : index
+}
+
+// -----
+
+// CHECK-LABEL: @linearize_unit_basis_zero
+// CHECK-SAME: (%[[arg0:.+]]: index, %[[arg1:.+]]: index, %[[arg2:.+]]: index)
+// CHECK: %[[ret:.+]] = affine.linearize_index [%[[arg0]], %[[arg1]]] by (3, %[[arg2]]) : index
+// CHECK: return %[[ret]]
+func.func @linearize_unit_basis_zero(%arg0: index, %arg1: index, %arg2: index) -> index {
+  %c0 = arith.constant 0 : index
+  %ret = affine.linearize_index [%arg0, %c0, %arg1] by (3, 1, %arg2) : index
+  return %ret : index
+}
+
+// -----
+
+// CHECK-LABEL: @linearize_all_zero_unit_basis
+// CHECK: arith.constant 0 : index
+// CHECK-NOT: affine.linearize_index
+func.func @linearize_all_zero_unit_basis() -> index {
+  %c0 = arith.constant 0 : index
+  %ret = affine.linearize_index [%c0, %c0] by (1, 1) : index
+  return %ret : index
+}
diff --git a/mlir/test/Dialect/Affine/invalid.mlir b/mlir/test/Dialect/Affine/invalid.mlir
@@ -548,6 +548,22 @@ func.func @delinearize(%idx: index, %basis0: index, %basis1 :index) {
 
 // -----
 
+func.func @linearize(%idx: index, %basis0: index, %basis1 :index) -> index {
+  // expected-error@+1 {{'affine.linearize_index' op should be passed an index for each basis element}}
+  %0 = affine.linearize_index [%idx] by (%basis0, %basis1) : index
+  return %0 : index
+}
+
+// -----
+
+func.func @linearize_empty() -> index {
+  // expected-error@+1 {{'affine.linearize_index' op basis should not be empty}}
+  %0 = affine.linearize_index [] by () : index
+  return %0 : index
+}
+
+// -----
+
 func.func @dynamic_dimension_index() {
   "unknown.region"() ({
     %idx = "unknown.test"() : () -> (index)
diff --git a/mlir/test/Dialect/Affine/ops.mlir b/mlir/test/Dialect/Affine/ops.mlir
