Skip to content
Snippets Groups Projects
Commit b29a9727 authored by CHAMONT David's avatar CHAMONT David
Browse files

General setup ready, but not specialized for MUST.

parent a33bec47
No related branches found
No related tags found
No related merge requests found
......@@ -25,7 +25,7 @@ Because the default compilation options on Code Reckons only generate the code f
Provided you have asked for an account ahead of time, on the [Intel Tiber Developer Cloud for oneAPI](https://devcloud.intel.com/oneapi/home/), and made the Intel recommended local ssh installation, that is how you may run the tests with `qsub`:
```
```sh
ssh devcloud
git clone https://gitlab.in2p3.fr/CodeursIntensifs/grayscott/GrayScottSyclSetup.git
......@@ -51,7 +51,7 @@ One can check the status of current jobs with `qstat`, and remove them with `qde
If you want to run within a container, a set of predefined images are usable:
- the [default oneAPI image](intel/oneapi-basekit) (4.26 GiB) which see your Intel CPU and iGPU.
- a ["smaller" oneAPI image](gitlab-registry.in2p3.fr/codeursintensifs/grayscott/grayscottsyclsetup/jfalcou:2024.1) (2.29 GiB) which only see your Intel CPU.
- a ["bigger" oneAPI image](gitlab-registry.in2p3.fr/codeursintensifs/grayscott/grayscottsyclsetup/dchamont:2024.1) (5.29 GiB) which add the NVidia codeplay plugin, and therefore also see your NVidia card.
- a ["bigger" oneAPI image](gitlab-registry.in2p3.fr/codeursintensifs/grayscott/grayscottsyclsetup/dchamont:2024.1) (5.29 GiB) which add the NVidia codeplay plugin, and therefore also see your NVidia card, if you have one.
We recommend you to start your container with the following command, which sets some useful options for Intel GPU (`--device=/dev/dri`) and|or NVidia card (`--gpus all`), and mount the current directory as `/work` in the container:
......@@ -61,7 +61,34 @@ docker pull ${IMG}
docker run --gpus all --device=/dev/dri --network host -it --rm -v ${PWD}:/work -w /work ${IMG}
```
Check that the installation is OK with `sycl-ls`, and `nvidia-smi` if you have a CUDA card.
First check that the installation is OK with `sycl-ls`, and `nvidia-smi` if you have a CUDA card.
Then you can go through the test program for CPU and|or CUDA:
```sh
# move to the top directory
cd GrayScottSyclSetup/
# check CPU nodes
IMG=gitlab-registry.in2p3.fr/codeursintensifs/grayscott/grayscottsyclsetup/jfalcou:2024.1
docker run --gpus all --device=/dev/dri --network host -it --rm -v ${PWD}:/work -w /work ${IMG}
cd CheckOneApiCpu
./sycl-ls.bash
./cmake.bash
./make.bash
./run.bash
exit
# Check GPU nodes
IMG=gitlab-registry.in2p3.fr/codeursintensifs/grayscott/grayscottsyclsetup/dchamont:2024.1
docker run --gpus all --device=/dev/dri --network host -it --rm -v ${PWD}:/work -w /work ${IMG}
cd CheckOneApiCuda
nvidia-smi
./sycl-ls.bash
./cmake.bash
./make.bash
./run.bash
exit
```
## Intel oneAPI Base Toolkit bare-metal installation
......@@ -76,20 +103,35 @@ For example:
Check that the installation seems OK with `sycl-ls`.
Note: when trying the recipe above, `sycl-ls` does not show the Intel iGPU. I wonder why.
Note: when trying myself the recipe above, `sycl-ls` does not show the Intel iGPU. I wonder why.
### Cuda
...
One way to proceed, for example for Ubuntu :
Check with `nvidia-smi` that your NVidia card is seen.
```sh
# Repo preparation: Driver Version: 555.42.02 CUDA Version: 12.5
wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/cuda-ubuntu2204.pin
mv -f cuda-ubuntu2204.pin /etc/apt/preferences.d/cuda-repository-pin-600
wget https://developer.download.nvidia.com/compute/cuda/12.5.0/local_installers/cuda-repo-ubuntu2204-12-5-local_12.5.0-555.42.02-1_amd64.deb
dpkg -i cuda-repo-ubuntu2204-12-5-local_12.5.0-555.42.02-1_amd64.deb
cp /var/cuda-repo-ubuntu2204-12-5-local/cuda-*-keyring.gpg /usr/share/keyrings/
# Installations apt-get
apt-get update
apt-get install -y cuda-toolkit-12-5
rm -rf /var/lib/apt/lists/*
# Check your card
nvidia-smi
```
### Codeplay plugin for NVidia
Get the installer from `https://developer.codeplay.com/products/oneapi/nvidia/download`,
and run it with `sh ...`.
Then check with `sycl-ls` that your NVidia card is seen.
Then check with `sycl-ls` that your NVidia card is now seen by SYCL.
......
File moved
File moved
File moved
File moved
0% Loading or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment