Enable cuFile on ARM (#405)

Fixes #339

Builds CUDA 12.2+ ARM packages with cuFile and adds cuFile as a dependency. Continues to build CUDA 11 ARM packages without cuFile.

As cuFile already has the needed logic to `dlopen` cuFile at runtime (with fallbacks for when it is missing), there shouldn't be any additional handling needed for legacy CUDA versions with ARM support.

Authors:
  - https://github.com/jakirkham

Approvers:
  - Mads R. B. Kristensen (https://github.com/madsbk)
  - Mike Sarahan (https://github.com/msarahan)
  - Bradley Dice (https://github.com/bdice)

URL: #405

conda/environments/all_cuda-118_arch-aarch64.yaml

@@ -0,0 +1,38 @@
+# This file is generated by `rapids-dependency-file-generator`.
+# To make changes, edit ../../dependencies.yaml and run `rapids-dependency-file-generator`.
+channels:
+- rapidsai
+- rapidsai-nightly
+- conda-forge
+- nvidia
+dependencies:
+- c-compiler
+- cmake>=3.26.4,!=3.30.0
+- cuda-python>=11.7.1,<12.0a0
+- cuda-version=11.8
+- cudatoolkit
+- cupy>=12.0.0
+- cxx-compiler
+- cython>=3.0.0
+- dask>=2022.05.2
+- doxygen=1.9.1
+- gcc_linux-aarch64=11.*
+- ninja
+- numcodecs <0.12.0
+- numpy>=1.23,<2.0a0
+- numpydoc
+- nvcc_linux-aarch64=11.8
+- nvcomp==3.0.6
+- packaging
+- pre-commit
+- pytest
+- pytest-cov
+- python>=3.9,<3.12
+- rapids-build-backend>=0.3.0,<0.4.0.dev0
+- scikit-build-core>=0.7.0
+- sphinx
+- sphinx-click
+- sphinx_rtd_theme
+- sysroot_linux-aarch64=2.17
+- zarr
+name: all_cuda-118_arch-aarch64

conda/environments/all_cuda-125_arch-aarch64.yaml

@@ -0,0 +1,38 @@
+# This file is generated by `rapids-dependency-file-generator`.
+# To make changes, edit ../../dependencies.yaml and run `rapids-dependency-file-generator`.
+channels:
+- rapidsai
+- rapidsai-nightly
+- conda-forge
+- nvidia
+dependencies:
+- c-compiler
+- cmake>=3.26.4,!=3.30.0
+- cuda-nvcc
+- cuda-python>=12.0,<13.0a0
+- cuda-version=12.5
+- cupy>=12.0.0
+- cxx-compiler
+- cython>=3.0.0
+- dask>=2022.05.2
+- doxygen=1.9.1
+- gcc_linux-aarch64=11.*
+- libcufile-dev
+- ninja
+- numcodecs <0.12.0
+- numpy>=1.23,<2.0a0
+- numpydoc
+- nvcomp==3.0.6
+- packaging
+- pre-commit
+- pytest
+- pytest-cov
+- python>=3.9,<3.12
+- rapids-build-backend>=0.3.0,<0.4.0.dev0
+- scikit-build-core>=0.7.0
+- sphinx
+- sphinx-click
+- sphinx_rtd_theme
+- sysroot_linux-aarch64=2.17
+- zarr
+name: all_cuda-125_arch-aarch64

conda/recipes/libkvikio/meta.yaml

@@ -50,7 +50,7 @@ requirements:
     - libcufile {{ cuda11_libcufile_host_version }}      # [linux64]
     - libcufile-dev {{ cuda11_libcufile_host_version }}  # [linux64]
     {% else %}
-    - libcufile-dev  # [linux64]
+    - libcufile-dev  # [linux]
     {% endif %}
 
 outputs:
@@ -67,7 +67,7 @@ outputs:
         - {{ compiler('cuda11') }}
         {% else %}
         - {{ compiler('cuda') }}
-        - libcufile-dev  # [linux64]
+        - libcufile-dev  # [linux]
         {% endif %}
     requirements:
       build:
@@ -81,7 +81,7 @@ outputs:
         - libcufile {{ cuda11_libcufile_run_version }}      # [linux64]
         - libcufile-dev {{ cuda11_libcufile_run_version }}  # [linux64]
         {% else %}
-        - libcufile-dev  # [linux64]
+        - libcufile-dev  # [linux]
         {% endif %}
     test:
         commands:
@@ -104,7 +104,7 @@ outputs:
         {% else %}
         - {{ compiler('cuda') }}
         - cuda-cudart-dev
-        - libcufile-dev  # [linux64]
+        - libcufile-dev  # [linux]
         {% endif %}
     requirements:
       build:
@@ -116,7 +116,7 @@ outputs:
         - libcufile {{ cuda11_libcufile_run_version }}  # [linux64]
         {% else %}
         - cuda-cudart-dev
-        - libcufile-dev  # [linux64]
+        - libcufile-dev  # [linux]
         {% endif %}
       run:
         - {{ pin_compatible('cuda-version', max_pin='x', min_pin='x') }}
@@ -125,7 +125,7 @@ outputs:
         - libcufile {{ cuda11_libcufile_run_version }}  # [linux64]
         {% else %}
         - cuda-cudart
-        - libcufile  # [linux64]
+        - libcufile  # [linux]
         {% endif %}
     about:
       home: https://rapids.ai

dependencies.yaml

@@ -4,7 +4,7 @@ files:
     output: conda
     matrix:
       cuda: ["11.8", "12.5"]
-      arch: [x86_64]
+      arch: [aarch64, x86_64]
     includes:
       - build-universal
       - build-cpp
@@ -209,14 +209,14 @@ dependencies:
             packages:
       - output_types: conda
         matrices:
-          - matrix:
-              arch: aarch64
-            packages:
           - matrix:
               cuda: "12.*"
-              arch: x86_64
             packages:
               - libcufile-dev
+          - matrix:
+              arch: aarch64
+              cuda: "11.*"
+            packages:
           - matrix:
               cuda: "11.8"
               arch: x86_64