Using znode instead of znodes inside kernels

[navidcy]

OceanBioME.jl/src/Light/2band.jl

Lines 16 to 29 in 888a107

	zᶜ = znodes(grid, Center(), Center(), Center())
	zᶠ = znodes(grid, Center(), Center(), Face())
	∫chlʳ = @inbounds (zᶠ[grid.Nz + 1] - zᶜ[grid.Nz]) * (P[i, j, grid.Nz] * Rᶜₚ / r) ^ eʳ
	∫chlᵇ = @inbounds (zᶠ[grid.Nz + 1] - zᶜ[grid.Nz]) * (P[i, j, grid.Nz] * Rᶜₚ / r) ^ eᵇ
	# first point below surface
	@inbounds PAR[i, j, grid.Nz] = PAR⁰ * (exp(kʳ * zᶜ[grid.Nz] - χʳ * ∫chlʳ) + exp(kᵇ * zᶜ[grid.Nz] - χᵇ * ∫chlᵇ)) / 2
	@inbounds for k in grid.Nz-1:-1:1
	∫chlʳ += (zᶜ[k + 1] - zᶠ[k + 1]) * (P[i, j, k+1] * Rᶜₚ / r) ^ eʳ + (zᶠ[k + 1] - zᶜ[k]) * (P[i, j, k] * Rᶜₚ / r) ^ eʳ
	∫chlᵇ += (zᶜ[k + 1] - zᶠ[k + 1]) * (P[i, j, k+1] * Rᶜₚ / r) ^ eᵇ + (zᶠ[k + 1] - zᶜ[k]) * (P[i, j, k] * Rᶜₚ / r) ^ eᵇ
	PAR[i, j, k] = PAR⁰ * (exp(kʳ * zᶜ[k] - χʳ * ∫chlʳ) + exp(kᵇ * zᶜ[k] - χᵇ * ∫chlᵇ)) / 2
	end

Here, I think it'd be much more efficient if instead of, e.g.,

zᶜ = znodes(grid, Center(), Center(), Center()) 
zᶜ[k + 1]

using Oceananigans.Grids: znode
znode(i, j, k+1, grid, Center(), Center(), Center())

The former create a whole array and selects a value from it. The latter just gets the value needed from the grid.

[navidcy]

Actually this might not be an issue. But definitely some bounds checking are missing here.