[mlir][loops] Reland Refactor LoopFuseSiblingOp and support parallel fusion #94391

mlir/include/mlir/Dialect/SCF/IR/SCFOps.td

@@ -303,7 +303,8 @@ def ForallOp : SCF_Op<"forall", [
        DeclareOpInterfaceMethods<LoopLikeOpInterface,
           ["getInitsMutable", "getRegionIterArgs", "getLoopInductionVars", 
            "getLoopLowerBounds", "getLoopUpperBounds", "getLoopSteps",
-           "promoteIfSingleIteration", "yieldTiledValuesAndReplace"]>,
+           "replaceWithAdditionalYields", "promoteIfSingleIteration",
+           "yieldTiledValuesAndReplace"]>,
        RecursiveMemoryEffects,
        SingleBlockImplicitTerminator<"scf::InParallelOp">,
        DeclareOpInterfaceMethods<RegionBranchOpInterface>,

mlir/include/mlir/Dialect/SCF/Utils/Utils.h

@@ -182,6 +182,16 @@ Loops tilePerfectlyNested(scf::ForOp rootForOp, ArrayRef<Value> sizes);
 void getPerfectlyNestedLoops(SmallVectorImpl<scf::ForOp> &nestedLoops,
                              scf::ForOp root);
 
+//===----------------------------------------------------------------------===//
+// Fusion related helpers
+//===----------------------------------------------------------------------===//
+
+/// Check structural compatibility between two loops such as iteration space
+/// and dominance.
+bool checkFusionStructuralLegality(LoopLikeOpInterface target,
+                                   LoopLikeOpInterface source,
+                                   Diagnostic &diag);
+
 /// Given two scf.forall loops, `target` and `source`, fuses `target` into
 /// `source`. Assumes that the given loops are siblings and are independent of
 /// each other.
@@ -203,6 +213,16 @@ scf::ForallOp fuseIndependentSiblingForallLoops(scf::ForallOp target,
 scf::ForOp fuseIndependentSiblingForLoops(scf::ForOp target, scf::ForOp source,
                                           RewriterBase &rewriter);
 
+/// Given two scf.parallel loops, `target` and `source`, fuses `target` into
+/// `source`. Assumes that the given loops are siblings and are independent of
+/// each other.
+///
+/// This function does not perform any legality checks and simply fuses the
+/// loops. The caller is responsible for ensuring that the loops are legal to
+/// fuse.
+scf::ParallelOp fuseIndependentSiblingParallelLoops(scf::ParallelOp target,
+                                                    scf::ParallelOp source,
+                                                    RewriterBase &rewriter);
 } // namespace mlir
 
 #endif // MLIR_DIALECT_SCF_UTILS_UTILS_H_

mlir/include/mlir/Interfaces/LoopLikeInterface.h

@@ -90,4 +90,24 @@ struct JamBlockGatherer {
 /// Include the generated interface declarations.
 #include "mlir/Interfaces/LoopLikeInterface.h.inc"
 
+namespace mlir {
+/// A function that rewrites `target`'s terminator as a teminator obtained by
+/// fusing `source` into `target`.
+using FuseTerminatorFn =
+    function_ref<void(RewriterBase &rewriter, LoopLikeOpInterface source,
+                      LoopLikeOpInterface &target, IRMapping mapping)>;
+
+/// Returns a fused `LoopLikeOpInterface` created by fusing `source` to
+/// `target`.  The `NewYieldValuesFn` callback is used to pass to the
+/// `replaceWithAdditionalYields` interface method to replace the loop with a
+/// new loop with (possibly) additional yields, while the `FuseTerminatorFn`
+/// callback is repsonsible for updating the fused loop terminator.
+LoopLikeOpInterface createFused(LoopLikeOpInterface target,
+                                LoopLikeOpInterface source,
+                                RewriterBase &rewriter,
+                                NewYieldValuesFn newYieldValuesFn,
+                                FuseTerminatorFn fuseTerminatorFn);
+
+} // namespace mlir
+
 #endif // MLIR_INTERFACES_LOOPLIKEINTERFACE_H_

mlir/lib/Dialect/SCF/IR/SCF.cpp

@@ -618,6 +618,44 @@ void ForOp::getSuccessorRegions(RegionBranchPoint point,
 
 SmallVector<Region *> ForallOp::getLoopRegions() { return {&getRegion()}; }
 
+FailureOr<LoopLikeOpInterface> ForallOp::replaceWithAdditionalYields(
+    RewriterBase &rewriter, ValueRange newInitOperands,
+    bool replaceInitOperandUsesInLoop,
+    const NewYieldValuesFn &newYieldValuesFn) {
+  // Create a new loop before the existing one, with the extra operands.
+  OpBuilder::InsertionGuard g(rewriter);
+  rewriter.setInsertionPoint(getOperation());
+  SmallVector<Value> inits(getOutputs());
+  llvm::append_range(inits, newInitOperands);
+  scf::ForallOp newLoop = rewriter.create<scf::ForallOp>(
+      getLoc(), getMixedLowerBound(), getMixedUpperBound(), getMixedStep(),
+      inits, getMapping(),
+      /*bodyBuilderFn =*/[](OpBuilder &, Location, ValueRange) {});
+
+  // Move the loop body to the new op.
+  rewriter.mergeBlocks(getBody(), newLoop.getBody(),
+                       newLoop.getBody()->getArguments().take_front(
+                           getBody()->getNumArguments()));
+
+  if (replaceInitOperandUsesInLoop) {
+    // Replace all uses of `newInitOperands` with the corresponding basic block
+    // arguments.
+    for (auto &&[newOperand, oldOperand] :
+         llvm::zip(newInitOperands, newLoop.getBody()->getArguments().take_back(
+                                        newInitOperands.size()))) {
+      rewriter.replaceUsesWithIf(newOperand, oldOperand, [&](OpOperand &use) {
+        Operation *user = use.getOwner();
+        return newLoop->isProperAncestor(user);
+      });
+    }
+  }
+
+  // Replace the old loop.
+  rewriter.replaceOp(getOperation(),
+                     newLoop->getResults().take_front(getNumResults()));
+  return cast<LoopLikeOpInterface>(newLoop.getOperation());
+}
+
 /// Promotes the loop body of a forallOp to its containing block if it can be
 /// determined that the loop has a single iteration.
 LogicalResult scf::ForallOp::promoteIfSingleIteration(RewriterBase &rewriter) {

mlir/lib/Dialect/SCF/TransformOps/SCFTransformOps.cpp

@@ -261,8 +261,10 @@ loopScheduling(scf::ForOp forOp,
     return 1;
   };
 
-  std::optional<int64_t> ubConstant = getConstantIntValue(forOp.getUpperBound());
-  std::optional<int64_t> lbConstant = getConstantIntValue(forOp.getLowerBound());
+  std::optional<int64_t> ubConstant =
+      getConstantIntValue(forOp.getUpperBound());
+  std::optional<int64_t> lbConstant =
+      getConstantIntValue(forOp.getLowerBound());
   DenseMap<Operation *, unsigned> opCycles;
   std::map<unsigned, std::vector<Operation *>> wrappedSchedule;
   for (Operation &op : forOp.getBody()->getOperations()) {
@@ -447,113 +449,6 @@ void transform::TakeAssumedBranchOp::getEffects(
 // LoopFuseSiblingOp
 //===----------------------------------------------------------------------===//
 
-/// Check if `target` and `source` are siblings, in the context that `target`
-/// is being fused into `source`.
-///
-/// This is a simple check that just checks if both operations are in the same
-/// block and some checks to ensure that the fused IR does not violate
-/// dominance.
-static DiagnosedSilenceableFailure isOpSibling(Operation *target,
-                                               Operation *source) {
-  // Check if both operations are same.
-  if (target == source)
-    return emitSilenceableFailure(source)
-           << "target and source need to be different loops";
-
-  // Check if both operations are in the same block.
-  if (target->getBlock() != source->getBlock())
-    return emitSilenceableFailure(source)
-           << "target and source are not in the same block";
-
-  // Check if fusion will violate dominance.
-  DominanceInfo domInfo(source);
-  if (target->isBeforeInBlock(source)) {
-    // Since `target` is before `source`, all users of results of `target`
-    // need to be dominated by `source`.
-    for (Operation *user : target->getUsers()) {
-      if (!domInfo.properlyDominates(source, user, /*enclosingOpOk=*/false)) {
-        return emitSilenceableFailure(target)
-               << "user of results of target should be properly dominated by "
-                  "source";
-      }
-    }
-  } else {
-    // Since `target` is after `source`, all values used by `target` need
-    // to dominate `source`.
-
-    // Check if operands of `target` are dominated by `source`.
-    for (Value operand : target->getOperands()) {
-      Operation *operandOp = operand.getDefiningOp();
-      // Operands without defining operations are block arguments. When `target`
-      // and `source` occur in the same block, these operands dominate `source`.
-      if (!operandOp)
-        continue;
-
-      // Operand's defining operation should properly dominate `source`.
-      if (!domInfo.properlyDominates(operandOp, source,
-                                     /*enclosingOpOk=*/false))
-        return emitSilenceableFailure(target)
-               << "operands of target should be properly dominated by source";
-    }
-
-    // Check if values used by `target` are dominated by `source`.
-    bool failed = false;
-    OpOperand *failedValue = nullptr;
-    visitUsedValuesDefinedAbove(target->getRegions(), [&](OpOperand *operand) {
-      Operation *operandOp = operand->get().getDefiningOp();
-      if (operandOp && !domInfo.properlyDominates(operandOp, source,
-                                                  /*enclosingOpOk=*/false)) {
-        // `operand` is not an argument of an enclosing block and the defining
-        // op of `operand` is outside `target` but does not dominate `source`.
-        failed = true;
-        failedValue = operand;
-      }
-    });
-
-    if (failed)
-      return emitSilenceableFailure(failedValue->getOwner())
-             << "values used inside regions of target should be properly "
-                "dominated by source";
-  }
-
-  return DiagnosedSilenceableFailure::success();
-}
-
-/// Check if `target` scf.forall can be fused into `source` scf.forall.
-///
-/// This simply checks if both loops have the same bounds, steps and mapping.
-/// No attempt is made at checking that the side effects of `target` and
-/// `source` are independent of each other.
-static bool isForallWithIdenticalConfiguration(Operation *target,
-                                               Operation *source) {
-  auto targetOp = dyn_cast<scf::ForallOp>(target);
-  auto sourceOp = dyn_cast<scf::ForallOp>(source);
-  if (!targetOp || !sourceOp)
-    return false;
-
-  return targetOp.getMixedLowerBound() == sourceOp.getMixedLowerBound() &&
-         targetOp.getMixedUpperBound() == sourceOp.getMixedUpperBound() &&
-         targetOp.getMixedStep() == sourceOp.getMixedStep() &&
-         targetOp.getMapping() == sourceOp.getMapping();
-}
-
-/// Check if `target` scf.for can be fused into `source` scf.for.
-///
-/// This simply checks if both loops have the same bounds and steps. No attempt
-/// is made at checking that the side effects of `target` and `source` are
-/// independent of each other.
-static bool isForWithIdenticalConfiguration(Operation *target,
-                                            Operation *source) {
-  auto targetOp = dyn_cast<scf::ForOp>(target);
-  auto sourceOp = dyn_cast<scf::ForOp>(source);
-  if (!targetOp || !sourceOp)
-    return false;
-
-  return targetOp.getLowerBound() == sourceOp.getLowerBound() &&
-         targetOp.getUpperBound() == sourceOp.getUpperBound() &&
-         targetOp.getStep() == sourceOp.getStep();
-}
-
 DiagnosedSilenceableFailure
 transform::LoopFuseSiblingOp::apply(transform::TransformRewriter &rewriter,
                                     transform::TransformResults &results,
@@ -569,25 +464,32 @@ transform::LoopFuseSiblingOp::apply(transform::TransformRewriter &rewriter,
            << "source handle (got " << llvm::range_size(sourceOps) << ")";
   }
 
-  Operation *target = *targetOps.begin();
-  Operation *source = *sourceOps.begin();
+  auto target = dyn_cast<LoopLikeOpInterface>(*targetOps.begin());
+  auto source = dyn_cast<LoopLikeOpInterface>(*sourceOps.begin());
+  if (!target || !source)
+    return emitSilenceableFailure(target->getLoc())
+           << "target or source is not a loop op";
 
-  // Check if the target and source are siblings.
-  DiagnosedSilenceableFailure diag = isOpSibling(target, source);
-  if (!diag.succeeded())
-    return diag;
+  // Check if loops can be fused
+  Diagnostic diag(target.getLoc(), DiagnosticSeverity::Error);
+  if (!mlir::checkFusionStructuralLegality(target, source, diag))
+    return DiagnosedSilenceableFailure::silenceableFailure(std::move(diag));
 
   Operation *fusedLoop;
-  /// TODO: Support fusion for loop-like ops besides scf.for and scf.forall.
-  if (isForWithIdenticalConfiguration(target, source)) {
+  // TODO: Support fusion for loop-like ops besides scf.for, scf.forall
+  // and scf.parallel.
+  if (isa<scf::ForOp>(target) && isa<scf::ForOp>(source)) {
     fusedLoop = fuseIndependentSiblingForLoops(
         cast<scf::ForOp>(target), cast<scf::ForOp>(source), rewriter);
-  } else if (isForallWithIdenticalConfiguration(target, source)) {
+  } else if (isa<scf::ForallOp>(target) && isa<scf::ForallOp>(source)) {
     fusedLoop = fuseIndependentSiblingForallLoops(
         cast<scf::ForallOp>(target), cast<scf::ForallOp>(source), rewriter);
+  } else if (isa<scf::ParallelOp>(target) && isa<scf::ParallelOp>(source)) {
+    fusedLoop = fuseIndependentSiblingParallelLoops(
+        cast<scf::ParallelOp>(target), cast<scf::ParallelOp>(source), rewriter);
   } else
     return emitSilenceableFailure(target->getLoc())
-           << "operations cannot be fused";
+           << "unsupported loop type for fusion";
 
   assert(fusedLoop && "failed to fuse operations");
 

mlir/lib/Dialect/SCF/Transforms/ParallelLoopFusion.cpp

@@ -16,6 +16,7 @@
 #include "mlir/Dialect/MemRef/IR/MemRef.h"
 #include "mlir/Dialect/SCF/IR/SCF.h"
 #include "mlir/Dialect/SCF/Transforms/Transforms.h"
+#include "mlir/Dialect/SCF/Utils/Utils.h"
 #include "mlir/IR/Builders.h"
 #include "mlir/IR/IRMapping.h"
 #include "mlir/IR/OpDefinition.h"
@@ -37,24 +38,6 @@ static bool hasNestedParallelOp(ParallelOp ploop) {
   return walkResult.wasInterrupted();
 }
 
-/// Verify equal iteration spaces.
-static bool equalIterationSpaces(ParallelOp firstPloop,
-                                 ParallelOp secondPloop) {
-  if (firstPloop.getNumLoops() != secondPloop.getNumLoops())
-    return false;
-
-  auto matchOperands = [&](const OperandRange &lhs,
-                           const OperandRange &rhs) -> bool {
-    // TODO: Extend this to support aliases and equal constants.
-    return std::equal(lhs.begin(), lhs.end(), rhs.begin());
-  };
-  return matchOperands(firstPloop.getLowerBound(),
-                       secondPloop.getLowerBound()) &&
-         matchOperands(firstPloop.getUpperBound(),
-                       secondPloop.getUpperBound()) &&
-         matchOperands(firstPloop.getStep(), secondPloop.getStep());
-}
-
 /// Checks if the parallel loops have mixed access to the same buffers. Returns
 /// `true` if the first parallel loop writes to the same indices that the second
 /// loop reads.
@@ -153,9 +136,10 @@ verifyDependencies(ParallelOp firstPloop, ParallelOp secondPloop,
 static bool isFusionLegal(ParallelOp firstPloop, ParallelOp secondPloop,
                           const IRMapping &firstToSecondPloopIndices,
                           llvm::function_ref<bool(Value, Value)> mayAlias) {
+  Diagnostic diag(firstPloop.getLoc(), DiagnosticSeverity::Remark);
   return !hasNestedParallelOp(firstPloop) &&
          !hasNestedParallelOp(secondPloop) &&
-         equalIterationSpaces(firstPloop, secondPloop) &&
+         checkFusionStructuralLegality(firstPloop, secondPloop, diag) &&
          succeeded(verifyDependencies(firstPloop, secondPloop,
                                       firstToSecondPloopIndices, mayAlias));
 }
@@ -174,61 +158,9 @@ static void fuseIfLegal(ParallelOp firstPloop, ParallelOp &secondPloop,
                      mayAlias))
     return;
 
-  DominanceInfo dom;
-  // We are fusing first loop into second, make sure there are no users of the
-  // first loop results between loops.
-  for (Operation *user : firstPloop->getUsers())
-    if (!dom.properlyDominates(secondPloop, user, /*enclosingOpOk*/ false))
-      return;
-
-  ValueRange inits1 = firstPloop.getInitVals();
-  ValueRange inits2 = secondPloop.getInitVals();
-
-  SmallVector<Value> newInitVars(inits1.begin(), inits1.end());
-  newInitVars.append(inits2.begin(), inits2.end());
-
-  IRRewriter b(builder);
-  b.setInsertionPoint(secondPloop);
-  auto newSecondPloop = b.create<ParallelOp>(
-      secondPloop.getLoc(), secondPloop.getLowerBound(),
-      secondPloop.getUpperBound(), secondPloop.getStep(), newInitVars);
-
-  Block *newBlock = newSecondPloop.getBody();
-  auto term1 = cast<ReduceOp>(block1->getTerminator());
-  auto term2 = cast<ReduceOp>(block2->getTerminator());
-
-  b.inlineBlockBefore(block2, newBlock, newBlock->begin(),
-                      newBlock->getArguments());
-  b.inlineBlockBefore(block1, newBlock, newBlock->begin(),
-                      newBlock->getArguments());
-
-  ValueRange results = newSecondPloop.getResults();
-  if (!results.empty()) {
-    b.setInsertionPointToEnd(newBlock);
-
-    ValueRange reduceArgs1 = term1.getOperands();
-    ValueRange reduceArgs2 = term2.getOperands();
-    SmallVector<Value> newReduceArgs(reduceArgs1.begin(), reduceArgs1.end());
-    newReduceArgs.append(reduceArgs2.begin(), reduceArgs2.end());
-
-    auto newReduceOp = b.create<scf::ReduceOp>(term2.getLoc(), newReduceArgs);
-
-    for (auto &&[i, reg] : llvm::enumerate(llvm::concat<Region>(
-             term1.getReductions(), term2.getReductions()))) {
-      Block &oldRedBlock = reg.front();
-      Block &newRedBlock = newReduceOp.getReductions()[i].front();
-      b.inlineBlockBefore(&oldRedBlock, &newRedBlock, newRedBlock.begin(),
-                          newRedBlock.getArguments());
-    }
-
-    firstPloop.replaceAllUsesWith(results.take_front(inits1.size()));
-    secondPloop.replaceAllUsesWith(results.take_back(inits2.size()));
-  }
-  term1->erase();
-  term2->erase();
-  firstPloop.erase();
-  secondPloop.erase();
-  secondPloop = newSecondPloop;
+  IRRewriter rewriter(builder);
+  secondPloop = mlir::fuseIndependentSiblingParallelLoops(
+      firstPloop, secondPloop, rewriter);
 }
 
 void mlir::scf::naivelyFuseParallelOps(