Lesson 3 — Elementwise map

An elementwise map applies a function independently to every element of an array. Because there are no dependencies between elements, this is the ideal GPU workload — every thread does the same work on a different datum.

Available math functions

Sarek exposes the standard GPU math library. Some useful functions in the Float32 module:

• Float32.sin x, Float32.cos x
• Float32.sqrt x, Float32.exp x, Float32.log x
• Float32.abs x, Float32.ceil x, Float32.floor x

Inline arithmetic: x *. x is faster than calling a power function for squaring.

Your task

Write a kernel that squares every element of a into b: b[i] = a[i] * a[i].

Kernel source fun (a : float32 vector) (b : float32 vector) -> let i = global_thread_id in b.(i) <- (* TODO *)

Click "Run on my GPU" to execute your kernel.

Hint

a.(i) *. a.(i) — that is all. You can also try Float32.sqrt (a.(i) *. a.(i)) to verify that sqrt(x²) = |x| for non-negative inputs.