Scalar support

[raikonenfnu]

So in GPU/Wave programming, you'd typically distribute dimensions across certain threads. But for scalar, we'd kind of want to broadcast/ let all threads to have access to that instead of trying to distribute it