Type Alias MatmulOp<A, B>

MatmulOp: IsMatMulCompatible<A["__shape"], B["__shape"]> extends true
    ? MatMulShape<A["__shape"], B["__shape"]> extends never
        ? never
        : StorageTransformation<
            "matmul",
            TensorStorage<
                Promote<A["__dtype"], B["__dtype"]>,
                MatMulShape<A["__shape"], B["__shape"]>,
                ComputeStrides<MatMulShape<A["__shape"], B["__shape"]>>,
                MatmulOpLayout,
            >,
            readonly [A, B],
        >
    : never

Matrix multiplication operation with shape validation

This type performs compile-time validation that:

Both inputs are at least 1D
Inner dimensions match (A's last dim == B's second-to-last dim)
Batch dimensions are compatible

The output shape follows NumPy/PyTorch conventions:

1D × 1D → scalar
1D × 2D → 1D
2D × 1D → 1D
2D × 2D → 2D
ND × ND → ND with broadcasted batch dimensions

Type Parameters

A extends AnyTensorStorage
First tensor storage
B extends AnyTensorStorage
Second tensor storage

Type Alias MatmulOp<A, B>

Type Parameters

Settings