[mlir][sparse] introduce sparse_tensor.iterate operation

Author: Peiming Liu

mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensor.h

@@ -17,9 +17,13 @@
 #include "mlir/IR/OpDefinition.h"
 #include "mlir/IR/OpImplementation.h"
 #include "mlir/IR/TensorEncoding.h"
+#include "mlir/Interfaces/ControlFlowInterfaces.h"
 #include "mlir/Interfaces/InferTypeOpInterface.h"
+#include "mlir/Interfaces/LoopLikeInterface.h"
 #include "mlir/Interfaces/SideEffectInterfaces.h"
 
+#include "llvm/ADT/bit.h"
+
 //===----------------------------------------------------------------------===//
 //
 // Type aliases to help code be more self-documenting. Unfortunately
@@ -41,6 +45,40 @@ using Level = uint64_t;
 /// including the value `ShapedType::kDynamic` (for shapes).
 using Size = int64_t;
 
+/// A simple wrapper to encode a bitset of defined  (at most 64) levels.
+class LevelSet {
+  uint64_t bits = 0;
+
+public:
+  LevelSet() = default;
+  explicit LevelSet(uint64_t bits) : bits(bits) {}
+  operator uint64_t() const { return bits; }
+
+  LevelSet &set(unsigned i) {
+    assert(i < 64);
+    bits |= 1 << i;
+    return *this;
+  }
+
+  LevelSet &operator|=(LevelSet lhs) {
+    bits |= static_cast<uint64_t>(lhs);
+    return *this;
+  }
+
+  LevelSet &lshift(unsigned offset) {
+    bits = bits << offset;
+    return *this;
+  }
+
+  bool operator[](unsigned i) const {
+    assert(i < 64);
+    return (bits & (1 << i)) != 0;
+  }
+
+  unsigned count() const { return llvm::popcount(bits); }
+  bool empty() const { return bits == 0; }
+};
+
 } // namespace sparse_tensor
 } // namespace mlir
 

mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorAttrDefs.td

@@ -19,6 +19,21 @@ class SparseTensor_Attr<string name,
                         list<Trait> traits = []>
     : AttrDef<SparseTensor_Dialect, name, traits>;
 
+//===----------------------------------------------------------------------===//
+// A simple bitset attribute wrapped over a single int64_t to encode a set of
+// sparse tensor levels.
+//===----------------------------------------------------------------------===//
+
+def LevelSetAttr :
+    TypedAttrBase<
+      I64, "IntegerAttr",
+      And<[CPred<"::llvm::isa<::mlir::IntegerAttr>($_self)">,
+           CPred<"::llvm::cast<::mlir::IntegerAttr>($_self).getType().isInteger(64)">]>,
+      "LevelSet attribute"> {
+  let returnType = [{::mlir::sparse_tensor::LevelSet}];
+  let convertFromStorage = [{::mlir::sparse_tensor::LevelSet($_self.getValue().getZExtValue())}];
+}
+
 //===----------------------------------------------------------------------===//
 // These attributes are just like `IndexAttr` except that they clarify whether
 // the index refers to a dimension (an axis of the semantic tensor) or a level

mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorOps.td

@@ -15,6 +15,8 @@ include "mlir/Dialect/SparseTensor/IR/SparseTensorTypes.td"
 include "mlir/Dialect/SparseTensor/IR/SparseTensorInterfaces.td"
 include "mlir/Interfaces/InferTypeOpInterface.td"
 include "mlir/Interfaces/SideEffectInterfaces.td"
+include "mlir/Interfaces/ControlFlowInterfaces.td"
+include "mlir/Interfaces/LoopLikeInterface.td"
 
 //===----------------------------------------------------------------------===//
 // Base class.
@@ -1277,7 +1279,7 @@ def SparseTensor_SelectOp : SparseTensor_Op<"select", [Pure, SameOperandsAndResu
 
 def SparseTensor_YieldOp : SparseTensor_Op<"yield", [Pure, Terminator,
     ParentOneOf<["BinaryOp", "UnaryOp", "ReduceOp", "SelectOp",
-                 "ForeachOp"]>]>,
+                 "ForeachOp", "IterateOp"]>]>,
     Arguments<(ins Variadic<AnyType>:$results)> {
   let summary = "Yield from sparse_tensor set-like operations";
   let description = [{
@@ -1490,6 +1492,103 @@ def ExtractIterSpaceOp : SparseTensor_Op<"extract_iteration_space",
                        " attr-dict `:` type($tensor) (`,` type($parentIter)^)?";
 }
 
+def IterateOp : SparseTensor_Op<"iterate",
+    [RecursiveMemoryEffects, RecursivelySpeculatable,
+     DeclareOpInterfaceMethods<LoopLikeOpInterface,
+      ["getInitsMutable", "getLoopResults", "getRegionIterArgs",
+       "getYieldedValuesMutable"]>,
+     DeclareOpInterfaceMethods<RegionBranchOpInterface,
+      ["getEntrySuccessorOperands"]>,
+     SingleBlockImplicitTerminator<"sparse_tensor::YieldOp">]> {
+
+  let arguments = (ins AnySparseIterSpace:$iterSpace,
+                       Variadic<AnyType>:$initArgs,
+                       LevelSetAttr:$crdUsedLvls);
+  let results = (outs Variadic<AnyType>:$results);
+  let regions = (region SizedRegion<1>:$region);
+
+  let summary = "Iterate over a sparse iteration space";
+  let description = [{
+      The `sparse_tensor.iterate` operations represents a loop over the
+      provided iteration space extracted from a specific sparse tensor.
+      The operation defines an SSA value for a sparse iterator that points
+      to the current stored element in the sparse tensor and SSA values
+      for coordinates of the stored element. The coordinates are always
+      converted to `index` type despite of the underlying sparse tensor
+      storage. When coordinates are not used, the SSA values can be skipped
+      by `_` symbols, which usually leads to simpler generated code after
+      sparsification. For example:
+
+      ```mlir
+      // The coordinate for level 0 is not used when iterating over a 2-D
+      // iteration space.
+      %sparse_tensor.iterate %iterator in %space at(_, %crd_1)
+        : !sparse_tensor.iter_space<#CSR, lvls = 0 to 2>
+      ```
+
+      `sparse_tensor.iterate` can also operate on loop-carried variables
+      and returns the final values after loop termination.
+      The initial values of the variables are passed as additional SSA operands
+      to the iterator SSA value and used coordinate SSA values mentioned
+      above. The operation region has an argument for the iterator, variadic
+      arguments for specified (used) coordiates and followed by one argument
+      for each loop-carried variable, representing the value of the variable
+      at the current iteration.
+      The body region must contain exactly one block that terminates with
+      `sparse_tensor.yield`.
+
+      `sparse_tensor.iterate` results hold the final values after the last
+      iteration. If the `sparse_tensor.iterate` defines any values, a yield
+      must be explicitly present.
+      The number and types of the `sparse_tensor.iterate` results must match
+      the initial values in the iter_args binding and the yield operands.
+
+
+      A nested `sparse_tensor.iterate` example that prints all the coordinates
+      stored in the sparse input:
+
+      ```mlir
+      func.func @nested_iterate(%sp : tensor<4x8xf32, #COO>) {
+        // Iterates over the first level of %sp
+        %l1 = sparse_tensor.extract_iteration_space %sp lvls = 0 : tensor<4x8xf32, #COO>
+        %r1 = sparse_tensor.iterate %it1 in %l1 at (%crd0)
+            : !sparse_tensor.iter_space<#COO, lvls = 0 to 1>  {
+          // Iterates over the second level of %sp
+          %l2 = sparse_tensor.extract_iteration_space %sp at %it1 lvls = 1
+              : tensor<4x8xf32, #COO>, !sparse_tensor.iterator<#COO, lvls = 0 to 1>
+          %r2 = sparse_tensor.iterate %it2 in %l2 at (crd1)
+              : !sparse_tensor.iter_space<#COO, lvls = 1 to 2>  {
+             vector.print %crd0 : index
+             vector.print %crd1 : index
+          }
+        }
+      }
+
+      ```
+  }];
+
+  let extraClassDeclaration = [{
+    unsigned getSpaceDim() {
+      return getIterSpace().getType().getSpaceDim();
+    }
+    BlockArgument getIterator() {
+      return getRegion().getArguments().front();
+    }
+    Block::BlockArgListType getCrds() {
+      // The first block argument is iterator, the remaining arguments are
+      // referenced coordinates.
+      return getRegion().getArguments().slice(1, getCrdUsedLvls().count());
+    }
+    unsigned getNumRegionIterArgs() {
+      return getRegion().getArguments().size() - 1 - getCrdUsedLvls().count();
+    }
+  }];
+
+  let hasVerifier = 1;
+  let hasRegionVerifier = 1;
+  let hasCustomAssemblyFormat = 1;
+}
+
 //===----------------------------------------------------------------------===//
 // Sparse Tensor Debugging and Test-Only Operations.
 //===----------------------------------------------------------------------===//