Question About 'ComputationSliceState' Struct in Affine Dialect (original) (raw)

Hi,

I am having trouble understanding ComputationSliceState Struct (mlir/include/mlir/Dialect/Affine/Analysis/Utils.h):

/// ComputationSliceState aggregates loop IVs, loop bound AffineMaps and their
/// associated operands for a set of loops within a loop nest (typically the
/// set of loops surrounding a store operation). Loop bound AffineMaps which
/// are non-null represent slices of that loop's iteration space.
struct ComputationSliceState {
  // List of sliced loop IVs (ordered from outermost to innermost).
  // EX: 'ivs[i]' has lower bound 'lbs[i]' and upper bound 'ubs[i]'.
  SmallVector<Value, 4> ivs;
  // List of lower bound AffineMaps.
  SmallVector<AffineMap, 4> lbs;
  // List of upper bound AffineMaps.
  SmallVector<AffineMap, 4> ubs;
  // List of lower bound operands (lbOperands[i] are used by 'lbs[i]').
  std::vector<SmallVector<Value, 4>> lbOperands;
  // List of upper bound operands (ubOperands[i] are used by 'ubs[i]').
  std::vector<SmallVector<Value, 4>> ubOperands;
  // Slice loop nest insertion point in target loop nest.
  Block::iterator insertPoint;
  ...
}

And maybe could someone explain this date structure based on this example I extracted from the comment in the same file:

//  Backward slice example:
//
//    affine.for %i0 = 0 to 10 {
//      affine.store %cst, %0[%i0] : memref<100xf32>  // 'depSourceAccess'
//    }
//    affine.for %i1 = 0 to 10 {
//      %v = affine.load %0[%i1] : memref<100xf32>    // 'depSinkAccess'
//    }
//
//    // Backward computation slice of loop nest '%i0'.
//    affine.for %i0 = (d0) -> (d0)(%i1) to (d0) -> (d0 + 1)(%i1) {
//      affine.store %cst, %0[%i0] : memref<100xf32>  // 'depSourceAccess'
//    }
//
//  Forward slice example:
//
//    affine.for %i0 = 0 to 10 {
//      affine.store %cst, %0[%i0] : memref<100xf32>  // 'depSourceAccess'
//    }
//    affine.for %i1 = 0 to 10 {
//      %v = affine.load %0[%i1] : memref<100xf32>    // 'depSinkAccess'
//    }
//
//    // Forward computation slice of loop nest '%i1'.
//    affine.for %i1 = (d0) -> (d0)(%i0) to (d0) -> (d0 + 1)(%i0) {
//      %v = affine.load %0[%i1] : memref<100xf32>    // 'depSinkAccess'
//    }
//

Specifically, what is a computation slice? What is a ‘slice of the loop’s iteration space’ ? and what is the lower bound and upper bound of the computation slice? And how is the loop bound operands related to loop bounds?

Thanks!