[FEA] Python bindings for `rmm::mr::pinned_host_memory_resource`

[harrism]

Is your feature request related to a problem? Please describe.

Add Python bindings for pinned_host_memory_resource added in #1392 .

[willtryagain]

Hi, I would like to work on this issue. But I don't fully understand how to solve this. I am going through the cython documentation. I have also located the files I need to change - https://github.com/rapidsai/rmm/blob/branch-24.02/python/rmm/_lib/memory_resource.pyx

Would you mind sharing the steps to take to do so?
Here is my (maybe incorrect) understanding so far -

1. Create a class definition like https://github.com/rapidsai/rmm/blob/branch-24.02/python/rmm/_lib/memory_resource.pyx#L126
2. Create python class PinnedHostMemoryResource and define the functions inside
3. Add entry to python/rmm/mr.py
4. Add tests to python/rmm/tests/test_rmm.py

[harrism]

@willtryagain that sounds about right! Thanks and we look forward to your contribution. Please make sure you also look at the RAPIDS contributing docs. Also this page is useful for dates; we are just starting 24.04 so make sure you target branch-24.04 with your PR.

[willtryagain]

Hey @harrism ,

Can you direct me on how to debug this error please?

I got this when I ran

cmake .. -DCMAKE_INSTALL_PREFIX=$CONDA_PREFIX.

I was busy with other deadlines so sorry for the late response.

[wence-]

Hey @harrism ,

Can you direct me on how to debug this error please?

I got this when I ran

cmake .. -DCMAKE_INSTALL_PREFIX=$CONDA_PREFIX.

I was busy with other deadlines so sorry for the late response.

Something looks very strange with your environment from that error message. What did you do to set up the development environment for RMM?

[willtryagain]

Hello @wence- ,

I have followed the README instructions

1. git clone the latest version
2. create conda env using conda env create --name rmm_dev --file conda/environments/all_cuda-122_arch-x86_64.yaml. I have nvcc 12.3 so this may be a problem.
3. follow build instructions
   $ mkdir build # make a build directory
   $ cd build # enter the build directory
   $ cmake .. -DCMAKE_INSTALL_PREFIX=/install/path # configure cmake ... use $CONDA_PREFIX if you're using Anaconda

Thanks.

[wence-]

OK, thanks. I have a CTK 12.3 install here, so maybe I can reproduce.

[wence-]

Hmm, I could not reproduce, I did (I used mamba not conda, but everything else is the same). On ce3af2c:

git clean -fdx # there's no magic behind the curtain
mamba env create --name rmm-dev --file conda/environments/all_cuda-122_arch-x86_64.yaml
mamba activate rmm-dev
mkdir build-test
cd build-test
cmake .. -DCMAKE_INSTALL_PREFIX=$CONDA_PREFIX
make -j

Without any issues.

Can you show the output (in the activated rmm_dev environment) of conda list?

An alternative to trying to get the development environment set up manually is (easiest if you use vscode, but possible without) to use the devcontainer setup we have in the repository (needs some setup in advance setting up the nvidia container toolkit locally, https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html).

[willtryagain]

Hey @wence- ,

Thanks.
I also wanted to inform that while running cmake it failed CMAKE_HAVE_LIBC_PTHREAD.

Here is the conda list output.