15 #ifndef MLIR_DIALECT_AFFINE_LOOPUTILS_H 16 #define MLIR_DIALECT_AFFINE_LOOPUTILS_H 25 class LoopLikeOpInterface;
51 AffineForOp forOp, uint64_t unrollFactor,
53 bool cleanUpUnroll =
false);
75 uint64_t unrollJamFactor);
80 uint64_t unrollJamFactor);
96 bool unrollPrologueEpilogue =
false);
297 #endif // MLIR_DIALECT_AFFINE_LOOPUTILS_H Include the generated interface declarations.
LogicalResult affineForOpBodySkew(AffineForOp forOp, ArrayRef< uint64_t > shifts, bool unrollPrologueEpilogue=false)
Skew the operations in an affine.for's body with the specified operation-wise shifts.
Operation is a basic unit of execution within MLIR.
bool LLVM_ATTRIBUTE_UNUSED isPerfectlyNested(ArrayRef< AffineForOp > loops)
Returns true if loops is a perfectly nested loop nest, where loops appear in it from outermost to inn...
unsigned permuteLoops(MutableArrayRef< AffineForOp > inputNest, ArrayRef< unsigned > permMap)
Performs a loop permutation on a perfectly nested loop nest inputNest (where the contained loops appe...
LogicalResult coalesceLoops(MutableArrayRef< AffineForOp > loops)
Replace a perfect nest of "for" loops with a single linearized loop.
LogicalResult promoteIfSingleIteration(AffineForOp forOp)
Promotes the loop body of a AffineForOp to its containing block if the loop was known to have a singl...
LogicalResult loopUnrollByFactor(AffineForOp forOp, uint64_t unrollFactor, function_ref< void(unsigned, Operation *, OpBuilder)> annotateFn=nullptr, bool cleanUpUnroll=false)
Unrolls this for operation by the specified unroll factor.
void gatherLoops(func::FuncOp func, std::vector< SmallVector< AffineForOp, 2 >> &depthToLoops)
Gathers all AffineForOps in 'func.func' grouped by loop depth.
This class defines the main interface for locations in MLIR and acts as a non-nullable wrapper around...
LogicalResult tilePerfectlyNested(MutableArrayRef< AffineForOp > input, ArrayRef< unsigned > tileSizes, SmallVectorImpl< AffineForOp > *tiledNest=nullptr)
Tiles the specified band of perfectly nested loops creating tile-space loops and intra-tile loops...
Result for calling generateCopyForMemRegion.
LogicalResult loopUnrollJamUpToFactor(AffineForOp forOp, uint64_t unrollJamFactor)
Unrolls and jams this loop by the specified factor or by the trip count (if constant), whichever is lower.
This class represents an efficient way to signal success or failure.
LogicalResult loopUnrollFull(AffineForOp forOp)
Unrolls this for operation completely if the trip count is known to be constant.
OpListType::iterator iterator
LogicalResult tilePerfectlyNestedParametric(MutableArrayRef< AffineForOp > input, ArrayRef< Value > tileSizes, SmallVectorImpl< AffineForOp > *tiledNest=nullptr)
Tiles the specified band of perfectly nested loops creating tile-space loops and intra-tile loops...
SmallVector< SmallVector< AffineForOp, 8 >, 8 > tile(ArrayRef< AffineForOp > forOps, ArrayRef< uint64_t > sizes, ArrayRef< AffineForOp > targets)
Performs tiling fo imperfectly nested loops (with interchange) by strip-mining the forOps by sizes an...
void getPerfectlyNestedLoops(SmallVectorImpl< AffineForOp > &nestedLoops, AffineForOp root)
Get perfectly nested sequence of loops starting at root of loop nest (the first op being another Affi...
AffineForOp sinkSequentialLoops(AffineForOp forOp)
LogicalResult loopUnrollUpToFactor(AffineForOp forOp, uint64_t unrollFactor)
Unrolls this loop by the specified unroll factor or its trip count, whichever is lower.
uint64_t fastMemCapacityBytes
A multi-dimensional affine map Affine map's are immutable like Type's, and they are uniqued...
LogicalResult separateFullTiles(MutableArrayRef< AffineForOp > nest, SmallVectorImpl< AffineForOp > *fullTileNest=nullptr)
Separates full tiles from partial tiles for a perfect nest nest by generating a conditional guard tha...
Explicit copy / DMA generation options for mlir::affineDataCopyGenerate.
LogicalResult affineDataCopyGenerate(Block::iterator begin, Block::iterator end, const AffineCopyOptions ©Options, Optional< Value > filterMemRef, DenseSet< Operation *> ©Nests)
Performs explicit copying for the contiguous sequence of operations in the block iterator range [`beg...
LogicalResult generateCopyForMemRegion(const MemRefRegion &memrefRegion, Operation *analyzedOp, const AffineCopyOptions ©Options, CopyGenerateResult &result)
generateCopyForMemRegion is similar to affineDataCopyGenerate, but works with a single memref region...
void getTileableBands(func::FuncOp f, std::vector< SmallVector< AffineForOp, 6 >> *bands)
Identify valid and profitable bands of loops to tile.
void promoteSingleIterationLoops(func::FuncOp f)
Promotes all single iteration AffineForOp's in the Function, i.e., moves their body into the containi...
void interchangeLoops(AffineForOp forOpA, AffineForOp forOpB)
Performs loop interchange on 'forOpA' and 'forOpB'.
LogicalResult loopUnrollJamByFactor(AffineForOp forOp, uint64_t unrollJamFactor)
Unrolls and jams this loop by the specified factor.
bool isValidLoopInterchangePermutation(ArrayRef< AffineForOp > loops, ArrayRef< unsigned > loopPermMap)
Checks if the loop interchange permutation 'loopPermMap', of the perfectly nested sequence of loops i...
A region of a memref's data space; this is typically constructed by analyzing load/store op's on this...
void mapLoopToProcessorIds(scf::ForOp forOp, ArrayRef< Value > processorId, ArrayRef< Value > numProcessors)
Maps forOp for execution on a parallel grid of virtual processorIds of size given by numProcessors...
This class helps build Operations.
This class provides an abstraction over the different types of ranges over Values.
AffineForOp createCanonicalizedAffineForOp(OpBuilder b, Location loc, ValueRange lbOperands, AffineMap lbMap, ValueRange ubOperands, AffineMap ubMap, int64_t step=1)
Creates an AffineForOp while ensuring that the lower and upper bounds are canonicalized, i.e., unused and duplicate operands are removed, any constant operands propagated/folded in, and duplicate bound maps dropped.