pub fn batched_outer_product_into<D, A, B, OpA, OpB>(
dest: &mut StridedViewMut<'_, D>,
lhs: &StridedView<'_, A, OpA>,
rhs: &StridedView<'_, B, OpB>,
lhs_free_ndim: usize,
rhs_free_ndim: usize,
exec: Exec<'_>,
) -> Result<()>Expand description
Compute dest[lhs_free..., rhs_free..., batch...] = lhs[lhs_free..., batch...] * rhs[rhs_free..., batch...].
This is a semantic convenience wrapper over broadcast_mul_into. The
broadcast/mul planner owns kernel selection, so explicit outer-product calls
and equivalent broadcasted multiplication use the same implementation path.
ยงErrors
Returns StridedError::RankMismatch when the free/batch split does not
describe the operand and destination ranks, StridedError::ShapeMismatch
when the batch or free extents disagree, or
StridedError::NonInjectiveOutputLayout when the destination layout
cannot be proven representable and injective. Validation completes before
the first write.