Simulation tips may be a little misleading

[glwagner]

The Simulation tips section of the docs probably needs updating. For example it says

For example, if can be difficult to just view a CuArray since Julia needs to access its elements to do that. Consider the example below:

This example is contrived though; the problem is that the show method for OffsetArrays has scalar indexing. But the show method for CuArray works fine, so

julia> using Oceananigans

julia> grid = RectilinearGrid(GPU(), size=(1, 1, 1), extent=(1, 1, 1), halo=(1, 1, 1))
1×1×1 RectilinearGrid{Float64, Periodic, Periodic, Bounded} on GPU with 1×1×1 halo
├── Periodic x ∈ [0.0, 1.0)  regularly spaced with Δx=1.0
├── Periodic y ∈ [0.0, 1.0)  regularly spaced with Δy=1.0
└── Bounded  z ∈ [-1.0, 0.0] regularly spaced with Δz=1.0

julia> c = CenterField(grid)
1×1×1 Field{Center, Center, Center} on RectilinearGrid on GPU
├── grid: 1×1×1 RectilinearGrid{Float64, Periodic, Periodic, Bounded} on GPU with 1×1×1 halo
├── boundary conditions: FieldBoundaryConditions
│   └── west: Periodic, east: Periodic, south: Periodic, north: Periodic, bottom: ZeroFlux, top: ZeroFlux, immersed: ZeroFlux
└── data: 3×3×3 OffsetArray(::CUDA.CuArray{Float64, 3, CUDA.Mem.DeviceBuffer}, 0:2, 0:2, 0:2) with eltype Float64 with indices 0:2×0:2×0:2
    └── max=0.0, min=0.0, mean=0.0

julia> c.data.parent
3×3×3 CUDA.CuArray{Float64, 3, CUDA.Mem.DeviceBuffer}:
[:, :, 1] =
 0.0  0.0  0.0
 0.0  0.0  0.0
 0.0  0.0  0.0

[:, :, 2] =
 0.0  0.0  0.0
 0.0  0.0  0.0
 0.0  0.0  0.0

[:, :, 3] =
 0.0  0.0  0.0
 0.0  0.0  0.0
 0.0  0.0  0.0

The example also creates a model which is probably overkill, a Field would do.

Also, the example uses

julia> adapt(Array, model.velocities.u.data)

To show an array on CPU. But we have a built-in method on_archiecture(CPU(), field) that would probably be better to use, since its more flexible and also generalizes to distributed architectures, etc...

[navidcy]

I also just noticed that this statement:

Oceananigans.jl/docs/src/simulation_tips.md

Lines 70 to 77 in 088f973

	```julia
	T₀ = 20 # ᵒC
	surface_temperature(x, y, t) = T₀ * sin(2π / 86400 * t)
	T_bcs = FieldBoundaryConditions(bottom = GradientBoundaryCondition(surface_temperature))
	```
	will throw an error if run on the GPU (and will run more slowly than it should on the CPU).

is wrong since the above code does not error. I believe the error will pop up down the line, when we try to pass this boundary condition to a model that uses a grid that lives on the GPU.

[glwagner]

right!

[glwagner]

Why don't we try to doctest / example this since we run docs on GPU