flopscope.accounting.tensorsolve_cost

a_shape:tuple

Argument forwarded to the analytical linalg.tensorsolve cost formula.

:int

Weighted public cost estimate, floored to match runtime accounting.

This helper multiplies the analytical FLOP count by the active weight from flopscope._weights and then applies int(...) so public estimates match budget deductions.

flops.linalg.tensorsolve(a, b, axes=...) optionally transposes a to move axes to the trailing positions, then reshapes the result to a square (n, n) system before delegating to solve. A transpose only reorders axes, so it never changes a's total element count, and numpy requires a.size == n**2 for any valid call (it raises otherwise) -- both regardless of axes and of b's rank. So n = isqrt(prod(a_shape)) recovers the true solved dimension directly from a's original (untransposed) shape, without needing to locate the reshape split point.