diff --git a/scripts_future_API/profileme.sh b/scripts_future_API/profileme.sh
deleted file mode 100755
index 59b405fe..00000000
--- a/scripts_future_API/profileme.sh
+++ /dev/null
@@ -1,16 +0,0 @@
-#!/bin/bash
-
-source /users/lurass/scratch/setenv_lumi.sh
-
-# basic
-# srun --cpu-bind=map_cpu:49,57,17,25,1,9,33,41 -N1 -n8 --gpus-per-node=8 profileme.sh
-
-# optimal using only single GCD per MI250x Module
-# srun --cpu-bind=map_cpu:49,17,1,33 -N1 -n1 --gpus-per-node=8 profileme.sh
-# srun --cpu-bind=map_cpu:49,17,1,33 -N4 -n16 --gpus-per-node=8 profileme.sh
-export ROCR_VISIBLE_DEVICES=0,2,4,6
-
-# julia --project benchmark_diffusion_3D.jl
-julia --project --color=yes tm_stokes_mpi_wip.jl
-
-# ENABLE_JITPROFILING=1 rocprof --hip-trace --hsa-trace -d ./prof_out${SLURM_PROCID} -o ./prof_out${SLURM_PROCID}/results${SLURM_PROCID}.csv julia --project bench3d.jl
diff --git a/scripts_future_API/runme.sh b/scripts_future_API/runme.sh
index 45ed56ab..f7c803f1 100755
--- a/scripts_future_API/runme.sh
+++ b/scripts_future_API/runme.sh
@@ -1,14 +1,23 @@
 #!/bin/bash
 
-module load LUMI/22.08
-module load partition/G
-module load rocm/5.3.3
+source /users/lurass/scratch/setenv_lumi.sh
+# module load LUMI/22.08
+# module load partition/G
+# module load rocm/5.3.3
 
 # ROCm-aware MPI set to 1, else 0
 export MPICH_GPU_SUPPORT_ENABLED=1
-export IGG_ROCMAWARE_MPI=1
 
-# Needs to know about location of GTL lib
-export LD_PRELOAD=${CRAY_MPICH_ROOTDIR}/gtl/lib/libmpi_gtl_hsa.so
+## basic
+# srun --cpu-bind=map_cpu:49,57,17,25,1,9,33,41 -N1 -n8 --gpus-per-node=8 profileme.sh
 
-julia --project rocmaware.jl
\ No newline at end of file
+## optimal using only single GCD per MI250x Module
+# srun --cpu-bind=map_cpu:49,17,1,33 -N1 -n1 --gpus-per-node=8 profileme.sh
+# srun --cpu-bind=map_cpu:49,17,1,33 -N4 -n16 --gpus-per-node=8 profileme.sh
+export ROCR_VISIBLE_DEVICES=0,2,4,6
+
+# julia --project benchmark_diffusion_3D.jl
+julia --project --color=yes tm_stokes_mpi_wip.jl
+
+# Profiling
+# ENABLE_JITPROFILING=1 rocprof --hip-trace --hsa-trace -d ./prof_out${SLURM_PROCID} -o ./prof_out${SLURM_PROCID}/results${SLURM_PROCID}.csv julia --project bench3d.jl
diff --git a/scripts_future_API/submit.sh b/scripts_future_API/submit.sh
index 1014ee59..77ff13fe 100644
--- a/scripts_future_API/submit.sh
+++ b/scripts_future_API/submit.sh
@@ -2,14 +2,17 @@
 #SBATCH --job-name="FastIce3D"
 #SBATCH --output=FastIce3D.%j.o
 #SBATCH --error=FastIce3D.%j.e
-#SBATCH --time=00:05:00
-#SBATCH --nodes=16
-#SBATCH --ntasks=64
-# #SBATCH --ntasks-per-node=8
+#SBATCH --time=00:10:00
+#SBATCH --nodes=4
+#SBATCH --ntasks=16
+# #SBATCH --ntasks-per-node=8 # this somehow fails...
 #SBATCH --gpus-per-node=8
 #SBATCH --partition=standard-g
 #SBATCH --account project_465000557
 
-# export ROCR_VISIBLE_DEVICES=0,2,4,6
+# CPU_BIND="map_cpu:49,57,17,25,1,9,33,41"
 
-srun --cpu-bind=map_cpu:49,17,1,33 ./profileme.sh
+# export ROCR_VISIBLE_DEVICES=0,2,4,6 # -> done in runme.sh
+CPU_BIND="map_cpu:49,17,1,33"
+
+srun --cpu-bind=${CPU_BIND} ./runme.sh
diff --git a/scripts_future_API/tm_stokes_mpi_wip.jl b/scripts_future_API/tm_stokes_mpi_wip.jl
index 7f1e8ca1..347fe608 100644
--- a/scripts_future_API/tm_stokes_mpi_wip.jl
+++ b/scripts_future_API/tm_stokes_mpi_wip.jl
@@ -36,7 +36,7 @@ function main(; do_visu=false, do_save=false)
     MPI.Init()
 
     backend = ROCBackend()
-    dims = (2, 1, 1)
+    dims = (4, 2, 2)
     # dims = (4, 2, 2)
     # dims = (2, 1, 1)
     topo = CartesianTopology(dims)
diff --git a/scripts_future_API/tm_stokes_wip.jl b/scripts_future_API/tm_stokes_wip.jl
index 97cc14ff..ca12b24f 100644
--- a/scripts_future_API/tm_stokes_wip.jl
+++ b/scripts_future_API/tm_stokes_wip.jl
@@ -15,6 +15,7 @@ const SBC = BoundaryCondition{Slip}
 using LinearAlgebra, Printf
 using KernelAbstractions
 # using CUDA
+# using AMDGPU
 
 using CairoMakie
 # using GLMakie