memcopy3D.jl

# Memory copy 3D to return T_peak
const USE_GPU = true
using ParallelStencil
using ParallelStencil.FiniteDifferences3D
@static if USE_GPU
    @init_parallel_stencil(CUDA, Float64, 3)
else
    @init_parallel_stencil(Threads, Float64, 3)
end

@parallel function copy3D!(T2, T, Ci)
    @all(T2) = @all(T) + @all(Ci)
    return
end

function memcopy3D()
# Numerics
nx, ny, nz = 1024, 1024, 512                             # Number of grid points in dimensions x, y and z
nt = 100                                                 # Number of time steps

# Array initializations
T   = @zeros(nx, ny, nz)
T2  = @zeros(nx, ny, nz)
Ci  = @zeros(nx, ny, nz)

# Initial conditions
Ci .= 0.5
T  .= 1.7
T2 .= T

t_tic = 0.0
# Time loop
for it = 1:nt
    if (it == 11) t_tic=time() end  # Start measuring time.
    @parallel copy3D!(T2, T, Ci)
    T, T2 = T2, T
end
t_toc=time()-t_tic

# Performance
A_eff = (2*1+1)*1/1e9*nx*ny*nz*sizeof(Data.Number)      # Effective main memory access per iteration [GB] (Lower bound of required memory access: T has to be read and written: 2 whole-array memaccess; Ci has to be read: : 1 whole-array memaccess)
t_it  = t_toc/(nt-10)                                   # Execution time per iteration [s]
T_eff = A_eff/t_it                                      # Effective memory throughput [GB/s]
println("time_s=$t_toc T_eff=$T_eff")
end

memcopy3D()