From 189f4e712e0275df14d5b1dbd397a722d407331c Mon Sep 17 00:00:00 2001 From: galabovaa Date: Wed, 5 Feb 2025 21:07:56 +0000 Subject: [PATCH] Update README.md --- src/pdlp/README.md | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-) diff --git a/src/pdlp/README.md b/src/pdlp/README.md index 150eb73df7..f331706c30 100644 --- a/src/pdlp/README.md +++ b/src/pdlp/README.md @@ -12,9 +12,9 @@ To utilize the maximum potential of any hardware, relevant drivers for that hard * Before installing anything, run the following in a terminal to check if nvidia-smi is present and can see the GPUs -`` +``` nvidia-smi -`` +``` This should print a table of GPU-related stats. If it doesn't don't worry: installing the Cuda Toolkit from https://developer.nvidia.com/cuda-toolkit should help! @@ -23,24 +23,24 @@ This should print a table of GPU-related stats. If it doesn't don't worry: insta Download Cuda from https://developer.nvidia.com/cuda-toolkit for the appropriate OS and architecture. Note, that installing the `nvidia-cuda-toolkit` package from apt will work on many Linux installations, *but* if the drivers are missing and you attempt to install them with -`` +``` apt install nvidia-cuda-toolkit apt install nvidia-utils-515 -`` +``` or `nvidia-utils-some-other-number`, this may lead to errors with versions of cuda libraries on versions of Lunux that are resolved by cleaning up the cuda drivers and toolkit from the machine and installing the toolkit from the website, which should give you the drivers as well. If you run into any trouble, please get in touch with Ivet right away so we can look at it together. Ivet did have some trouble getting it all to work on her laptop and is still learning about supporting different versions of Cuda on different operating systems. Check again that the drivers are present by running -`` +``` nvidia-smi -`` +``` * Run the following in a terminal -`` +``` nvcc --version -`` +``` If you see a version, this means that the cuda compiler nvcc is up and running. @@ -51,7 +51,7 @@ When both work, you should check that `/usr/local/cuda` exists. * Checkout the HiGHS codebase from Github, and use the branch `cuda-updates` * From your HiGHS build directory, run -`` +``` cmake -DCUPDLP_GPU=ON -DALL_TESTS=ON ..; make -`` +```