CSCfi · rkronberg · Jun 3, 2025 · Jun 4, 2025 · Jun 18, 2025 · Aug 13, 2025
diff --git a/csc-overrides/assets/snippets/graphical-connection.md b/csc-overrides/assets/snippets/graphical-connection.md
@@ -0,0 +1,4 @@
+!!! info "Note"
+    For performance reasons, we generally recommend using the
+    [HPC web interfaces](/computing/webinterface/index.md) to run applications which
+    require displaying graphics.
diff --git a/csc-overrides/assets/snippets/ssh-agent-forwarding.md b/csc-overrides/assets/snippets/ssh-agent-forwarding.md
@@ -0,0 +1,4 @@
+!!! warning "Note"
+    You should only forward your SSH agent to remote servers that you trust and
+    only when you really need it. Forwarding your SSH agent by default to any
+    server you connect to is considered insecure.
diff --git a/csc-overrides/assets/snippets/ssh-ca.md b/csc-overrides/assets/snippets/ssh-ca.md
@@ -0,0 +1,8 @@
+!!! warning "SSH certificates are required to connect to Roihu over SSH"
+
+    To connect to Roihu, users must sign their public key in MyCSC to obtain a
+    time-based SSH certificate. Each certificate is valid for 24 hours, and
+    once it expires, a new one must be generated by signing the public key
+    again.
+
+    [Read the detailed instructions on signing your public key](/computing/connecting/ssh-keys.md#signing-public-key).
diff --git a/csc-overrides/assets/snippets/using-ssh-keys.md b/csc-overrides/assets/snippets/using-ssh-keys.md
@@ -0,0 +1,8 @@
+!!! info "Using SSH keys"
+    See the page on [setting up SSH keys](/computing/connecting/ssh-keys.md)
+    for general information about using SSH keys and certificates for
+    authentication. Please note that it is mandatory to add your public key to
+    MyCSC – copying it directly to a CSC supercomputer does not work!
+
+    Supported key types are Ed25519 and RSA 4096 through 16384. **We strongly
+    recommend Ed25519**.
diff --git a/docs/apps/compute-san.md b/docs/apps/compute-san.md
@@ -10,27 +10,30 @@ catalog:
   available_on:
     - Puhti
     - Mahti
+    - Roihu
 ---
 
 # compute-sanitizer: functional correctness checking suite for CUDA programs
 
 ## Available
 
-- Puhti: 2022.2.0
-- Mahti: 2021.3.0
+- Puhti: version depends on the CUDA environment
+- Mahti: version depends on the CUDA environment
+- Roihu-GPU: version depends on the CUDA environment
 
 ## License
 
 Usage is possible for both academic and commercial purposes.
 
 ## Usage    
 
-[compute-sanitizer](https://docs.nvidia.com/cuda/compute-sanitizer/index.html) is a functional correctness checking suite included in the CUDA toolkit (starting from version 11). 
+[compute-sanitizer](https://docs.nvidia.com/cuda/compute-sanitizer/index.html) is a functional correctness checking suite included in the CUDA toolkit (`cuda` module needs to be loaded). 
 In order to use the tool, the CUDA code has to be compiled with the extra flags
 `-g` and `-G`.
 
-Debugging is started in an [interactive session](../computing/running/interactive-usage.md) 
-by running:
+Running the tool can be done either in an [interactive session](../computing/running/interactive-usage.md) or in a 
+[batch job](../computing/running/submitting-jobs.md). The application is started as (prepend `compute-sanitizer` by `srun`
+in a batch job): 
 
 ```bash
 compute-sanitizer  --tool <tool> ./cuda_program
@@ -44,3 +47,8 @@ where `<tool>` is one of the several sub-tools for different type of checks:
 * `initcheck`: can report cases where the GPU performs uninitialized accesses to global memory 
 
 * `synccheck`: can report cases where the application is attempting invalid usages of synchronization primitives
+
+!!! info 
+     Sometimes external libraries (e.g. MPI) generate a lot of false positives
+
+
diff --git a/docs/apps/cp2k.md b/docs/apps/cp2k.md
@@ -11,6 +11,7 @@ catalog:
     - LUMI
     - Puhti
     - Mahti
+    - Roihu
 ---
 
 # CP2K
@@ -22,6 +23,28 @@ parallel quantum chemistry calculations, in particular for AIMD.
 
 ## Available
 
+=== "Roihu-CPU"
+    | Version | Available modules | Notes       |
+    |:-------:|:------------------|:-----------:|
+    |2025.1   |`cp2k/2025.1`      | CPU version |
+    |2025.2   |`cp2k/2025.2`      | CPU version |
+    |2026.1   |`cp2k/2026.1`      | CPU version |
+
+=== "Roihu-GPU"
+    | Version | Available modules | Notes       |
+    |:-------:|:------------------|:-----------:|
+    |2025.1   |`cp2k/2025.1`      | GPU version |
+    |2025.2   |`cp2k/2025.2`      | GPU version |
+    |2026.1   |`cp2k/2026.1`      | GPU version |
+
+=== "LUMI"
+    | Version | Available modules                | Notes                 |
+    |:-------:|:---------------------------------|:---------------------:|
+    |2024.3   |`cp2k/2024.3`<br>`cp2k/2024.3-gpu`| GPU version available |
+    |2025.1   |`cp2k/2025.1`<br>`cp2k/2025.1-gpu`| GPU version available |
+    |2025.2   |`cp2k/2025.2`<br>`cp2k/2025.2-gpu`| GPU version available |
+    |2026.1   |`cp2k/2026.1`<br>`cp2k/2026.1-gpu`| GPU version available |
+
 === "Puhti"
     | Version | Available modules | Notes |
     |:-------:|:------------------|:-----:|
@@ -45,14 +68,6 @@ parallel quantum chemistry calculations, in particular for AIMD.
     |2024.2   |`cp2k/2024.2`      |       |
     |2025.1   |`cp2k/2025.1`      |       |
 
-=== "LUMI"
-    | Version | Available modules                | Notes                 |
-    |:-------:|:---------------------------------|:---------------------:|
-    |2024.3   |`cp2k/2024.3`<br>`cp2k/2024.3-gpu`| GPU version available |
-    |2025.1   |`cp2k/2025.1`<br>`cp2k/2025.1-gpu`| GPU version available |
-    |2025.2   |`cp2k/2025.2`<br>`cp2k/2025.2-gpu`| GPU version available |
-    |2026.1   |`cp2k/2026.1`<br>`cp2k/2026.1-gpu`| GPU version available |
-
 ## License
 
 CP2K is freely available under the GPL license.
@@ -91,43 +106,50 @@ double the number of cores the calculation should be at least 1.5 times faster.
 
 ### Example batch scripts
 
-=== "Puhti (MPI only)"
+=== "Roihu-CPU (hybrid MPI/OpenMP)"
 
     ```bash
     #!/bin/bash
-    #SBATCH --time=00:10:00
-    #SBATCH --ntasks-per-node=40
-    #SBATCH --nodes=2
-    #SBATCH --mem-per-cpu=2GB
-    #SBATCH --partition=large
+    #SBATCH --time=00:05:00
+    #SBATCH --nodes=1
+    #SBATCH --ntasks-per-node=96  # maximum 384
+    #SBATCH --cpus-per-task=4     # 384 / ntasks-per-node
+    #SBATCH --partition=medium
     #SBATCH --account=<project>
+    #SBATCH --hint=nomultithread
 
     module purge
-    module load gcc/14.2.0 openmpi/5.0.6
-    module load cp2k/2025.1
+    module load gcc/15.2.0 openmpi/5.0.10
+    module load cp2k/2026.1
 
-    srun cp2k.psmp H2O-64.inp > H2O-64.out
+    srun cp2k.psmp H2O-512.inp > H2O-512.out
     ```
 
-=== "Mahti (mixed MPI/OpenMP)"
+=== "Roihu-GPU (full GPU node)"
 
     ```bash
     #!/bin/bash
-    #SBATCH --time=00:05:00
-    #SBATCH --ntasks-per-node=32  # 2 - 128
-    #SBATCH --cpus-per-task=4     # 128 / ntasks-per-node
-    #SBATCH --nodes=2
-    #SBATCH --partition=test
+    #SBATCH --time=00:10:00
+    #SBATCH --nodes=1
+    #SBATCH --ntasks-per-node=64
+    #SBATCH --cpus-per-task=4
+    #SBATCH --gres=gpu:gh200:4
+    #SBATCH --partition=gpumedium
     #SBATCH --account=<project>
+    #SBATCH --hint=nomultithread
 
     module purge
-    module load gcc/14.2.0 openmpi/5.0.6
-    module load cp2k/2025.1
+    module load gcc/13.4.0 openmpi/5.0.10
+    module load cp2k/2026.1
 
-    export OMP_NUM_THREADS=$SLURM_CPUS_PER_TASK
-    export OMP_PLACES=cores
+    export OMP_NUM_THREADS=${SLURM_CPUS_PER_TASK}
 
-    srun cp2k.psmp H2O-64.inp > H2O-64.out
+    # Nvidia multi-process service (MPS) required to run multiple tasks per GPU
+    nvidia-cuda-mps-control -d
+
+    srun cp2k.psmp H2O-dft-ls.inp > H2O-dft-ls.out
+
+    echo quit | nvidia-cuda-mps-control
     ```
 
 === "LUMI-G (full GPU node)"
@@ -171,6 +193,45 @@ double the number of cores the calculation should be at least 1.5 times faster.
         separate GPU, you can reserve up to 8 "GPUs" per node. See more details
         in [LUMI Docs](https://docs.lumi-supercomputer.eu/hardware/lumig/).
 
+=== "Puhti (MPI only)"
+
+    ```bash
+    #!/bin/bash
+    #SBATCH --time=00:10:00
+    #SBATCH --ntasks-per-node=40
+    #SBATCH --nodes=2
+    #SBATCH --mem-per-cpu=2GB
+    #SBATCH --partition=large
+    #SBATCH --account=<project>
+
+    module purge
+    module load gcc/14.2.0 openmpi/5.0.6
+    module load cp2k/2025.1
+
+    srun cp2k.psmp H2O-64.inp > H2O-64.out
+    ```
+
+=== "Mahti (hybrid MPI/OpenMP)"
+
+    ```bash
+    #!/bin/bash
+    #SBATCH --time=00:05:00
+    #SBATCH --ntasks-per-node=32  # 2 - 128
+    #SBATCH --cpus-per-task=4     # 128 / ntasks-per-node
+    #SBATCH --nodes=2
+    #SBATCH --partition=test
+    #SBATCH --account=<project>
+
+    module purge
+    module load gcc/14.2.0 openmpi/5.0.6
+    module load cp2k/2025.1
+
+    export OMP_NUM_THREADS=$SLURM_CPUS_PER_TASK
+    export OMP_PLACES=cores
+
+    srun cp2k.psmp H2O-64.inp > H2O-64.out
+    ```
+
 ### Performance notes
 
 #### Mahti

diff --git a/docs/apps/cuda-gdb.md b/docs/apps/cuda-gdb.md
@@ -10,6 +10,7 @@ catalog:
   available_on:
     - Puhti
     - Mahti
+    - Roihu
 ---
 
 # cuda-gdb: CUDA debugger
@@ -18,6 +19,7 @@ catalog:
 
 - Puhti: 10.2
 - Mahti: 10.1
+- Roihu-GPU
 
 ## License
 
@@ -32,14 +34,14 @@ CUDA programs.
 In order to use tool the CUDA code has to be compiled with the extra flags
 `-g` and `-G`.
 
-Next in an [interactive session](../computing/running/interactive-usage.md) one needs to
+Next, in an [interactive session](../computing/running/interactive-usage.md) one needs to
 first load the CUDA module:
 
 ```bash
 module load cuda
 ```
 
-and then the debugging can be started by running:
+and then start the application under debugger by running:
 
 ```bash
 cuda-gdb ./cuda_program
@@ -52,6 +54,6 @@ specific to CUDA debugging:
 * Focus Commands: Commands to query or switch the focus of the debugger
 * Configuration Commands: Commands to configure the CUDA-specific commands
 
-Out of bonds accesses can be checked inside the debugger by activating
+Out of bounds accesses can be checked inside the debugger by activating
 the memory checker with `set cuda memcheck on`. Alternatively the `cuda-memcheck` or [`compute-sanitizer`](compute-san.md)
 tool can be used outside of the debugger (`cuda-memcheck ./cuda_program` or `compute-sanitizer ./cuda_program`).
diff --git a/docs/apps/gdb.md b/docs/apps/gdb.md
@@ -10,6 +10,7 @@ catalog:
   available_on:
     - Puhti
     - Mahti
+    - Roihu
 ---
 
 # gdb: GNU Debugger
@@ -18,6 +19,7 @@ catalog:
 
 - Puhti
 - Mahti
+- Roihu-CPU
 
 ## License
 
@@ -34,18 +36,21 @@ correct the effects of a bug.
 In order to use the debugger the program has to be compiled with `-g` flag to
 enable symbolic debugging.
 
-The debugger can either start a new process or attach to a running process.
-
-Example of starting a new process to be debugged:
+One can either start the application under the debugger, or attach the debugger 
+to a running application.
 
+In order to start the application under the debugger, launch first an
+[interactive session](../computing/running/interactive-usage.md) and execute then:
 ```
 gdb --tui ./myexecutable
 ```
 
-Example of attaching to an existing process (with process ID `pid`):
-
+In order to attach to a running application, 
+[connect first to a compute node](../computing/running/interactive-usage.md#connecting-to-a-compute-node-of-a-running-job).
+Next, you need to find the process ID `<pid>` e.g. by running the command `ps ux`,
+attach then debugger to that:
 ```
-gdb --tui ./myexecutable pid
+gdb --tui ./myexecutable <pid>
 ```
 
 If additional arguments are needed for the program, one can use the option

diff --git a/docs/apps/geoconda.md b/docs/apps/geoconda.md
@@ -147,6 +147,8 @@ The `geoconda` module is available:
 
 Version number is the same as the Python version.
 
+In Roihu geoconda is renamed to [python-geo](python-geo.md)
+
 ## Usage
 
 When using in LUMI, run this first: