# L-21 MCS 572 Mon 14 Oct 2024 : organization.jl

# Illustrates the launching of a kernel
# with a three dimensional grid of three dimensional blocks.

using CUDA

"""
    function matFill!(A)

fills the array using the blockIdx and threadIdx.
"""
function matFill!(A)
    bx = blockIdx().x - 1
    by = blockIdx().y - 1
    tx = threadIdx().x - 1
    ty = threadIdx().y - 1
    row = by*blockDim().y + ty 
    col = bx*blockDim().x + tx
    dim = gridDim().x*blockDim().x 
    idx = 1 + row*dim + col 
    A[idx] = idx
    return nothing
end

xb = 2 # gridDim.x
yb = 2 # gridDim.y
zb = 1 # gridDim.z
xt = 2 # blockDim.x
yt = 2 # blockDim.y
zt = 1 # blockDim.z

dim = xb*yb*zb*xt*yt*zt
A_h = zeros(dim)
A_d = CuArray(A_h)

@cuda threads=(xt, yt, zt) blocks=(xb, yb, zb) matFill!(A_d) 

A_h = Array(A_d)
println(A_d)
println(A_h)