Merge pull request #129 from JanJereczek/fix-parallelstencil

Fix parallelstencil initialisation
JanJereczek · Aug 17, 2024 · 8986aa0 · 8986aa0
2 parents c6a7d7e + fd00762
commit 8986aa0
Show file tree

Hide file tree

Showing 3 changed files with 10 additions and 6 deletions.
diff --git a/Project.toml b/Project.toml
@@ -40,7 +40,7 @@ FFTW = "1.8"
 FastGaussQuadrature = "1.0"
 Interpolations = "0.14, 0.15"
 NLsolve = "4.5"
-NetCDF = "0.11"
+NetCDF = "0.11, 0.12"
 OrdinaryDiffEq = "6.58"
 ParallelStencil = "0.13"
 Reexport = "1"

diff --git a/src/FastIsostasy.jl b/src/FastIsostasy.jl
@@ -17,6 +17,10 @@ using ParallelStencil: ParallelStencil, @init_parallel_stencil, @parallel,
 using Statistics: mean, cov
 using SpecialFunctions: besselj0, besselj1
 
+# Init stencil on GPU. Will only be used if specified in ComputationDomain.
+allowscalar(false)
+@init_parallel_stencil(CUDA, Float64, 3);
+
 using Reexport: Reexport, @reexport
 @reexport using Interpolations
 @reexport using OrdinaryDiffEq: Euler, Midpoint, Heun, Ralston, BS3, BS5, RK4,

diff --git a/src/utils.jl b/src/utils.jl
@@ -520,11 +520,11 @@ function zeropad_extension(M::Matrix{T}, Nx::Int, Ny::Int) where {T<:AbstractFlo
     return M_zeropadded
 end
 
-function init()
-    println("Initializing CUDA Stencil")
-    @init_parallel_stencil(CUDA, Float64, 2)
-    CUDA.allowscalar(false)
-end
+# function init()
+#     println("Initializing CUDA Stencil")
+#     @init_parallel_stencil(CUDA, Float64, 2)
+#     CUDA.allowscalar(false)
+# end
 
 #####################################################
 # Example utils