Add pynvjitlink as a dependency

ci/build_wheel.sh

@@ -48,7 +48,7 @@ fi
 if [[ $PACKAGE_CUDA_SUFFIX == "-cu12" ]]; then
     sed -i "s/cuda-python[<=>\.,0-9a]*/cuda-python>=12.0,<13.0a0/g" ${pyproject_file}
     sed -i "s/cupy-cuda11x/cupy-cuda12x/g" ${pyproject_file}
-    sed -i "/ptxcompiler/d" ${pyproject_file}
+    sed -i "s/ptxcompiler/pynvjitlink/g" ${pyproject_file}
     sed -i "/cubinlinker/d" ${pyproject_file}
 fi
 

conda/environments/all_cuda-120_arch-x86_64.yaml

@@ -70,6 +70,7 @@ dependencies:
 - protobuf>=4.21,<5
 - pyarrow==14.0.1.*
 - pydata-sphinx-theme!=0.14.2
+- pynvjitlink
 - pytest
 - pytest-benchmark
 - pytest-cases<3.8.2

conda/recipes/cudf/meta.yaml

@@ -1,4 +1,4 @@
-# Copyright (c) 2018-2023, NVIDIA CORPORATION.
+# Copyright (c) 2018-2024, NVIDIA CORPORATION.
 
 {% set version = environ['RAPIDS_PACKAGE_VERSION'].lstrip('v') %}
 {% set minor_version = version.split('.')[0] + '.' + version.split('.')[1] %}
@@ -98,6 +98,7 @@ requirements:
     # xref: https://github.com/rapidsai/cudf/issues/12822
     - cuda-nvrtc
     - cuda-python >=12.0,<13.0a0
+    - pynvjitlink
     {% endif %}
     - {{ pin_compatible('cuda-version', max_pin='x', min_pin='x') }}
     - nvtx >=0.2.1

dependencies.yaml

@@ -534,16 +534,19 @@ dependencies:
           - {matrix: null, packages: *run_cudf_packages_all_cu11}
       - output_types: conda
         matrices:
+          - matrix: {cuda: "12.*"}
+            packages:
+              - pynvjitlink
           - matrix: {cuda: "11.*"}
             packages:
               - cubinlinker
               - ptxcompiler
-          - {matrix: null, packages: null}
       - output_types: [requirements, pyproject]
         matrices:
           - matrix: {cuda: "12.*"}
             packages:
               - rmm-cu12==24.2.*
+              - pynvjitlink-cu12
           - matrix: {cuda: "11.*"}
             packages:
               - rmm-cu11==24.2.*

python/cudf/cudf/utils/_numba.py

@@ -3,23 +3,10 @@
 import glob
 import os
 import sys
-import warnings
 from functools import lru_cache
 
 from numba import config as numba_config
 
-try:
-    from pynvjitlink.patch import (
-        patch_numba_linker as patch_numba_linker_pynvjitlink,
-    )
-except ImportError:
-
-    def patch_numba_linker_pynvjitlink():
-        warnings.warn(
-            "CUDA Toolkit is newer than CUDA driver. "
-            "Numba features will not work in this configuration. "
-        )
-
 
 # Use an lru_cache with a single value to allow a delayed import of
 # strings_udf. This is the easiest way to break an otherwise circular import
@@ -117,11 +104,13 @@ def _setup_numba():
     version of the CUDA Toolkit used to build the PTX files shipped
     with the user cuDF package.
     """
-    # ptxcompiler is a requirement for cuda 11.x packages but not
-    # cuda 12.x packages. However its version checking machinery
-    # is still necessary. If a user happens to have ptxcompiler
-    # in a cuda 12 environment, it's use for the purposes of
-    # checking the driver and runtime versions is harmless
+
+    # Either ptxcompiler, or our vendored version (_ptxcompiler.py)
+    # is needed to determine the driver and runtime CUDA versions in
+    # the environment. In a CUDA 11.x environment, ptxcompiler is used
+    # to provide MVC directly, whereas for CUDA 12.x this is provided
+    # through pynvjitlink. The presence of either package does not
+    # perturb cuDF's operation in situations where they aren't used.
     try:
         from ptxcompiler.patch import NO_DRIVER, safe_get_versions
     except ModuleNotFoundError:
@@ -145,7 +134,9 @@ def _setup_numba():
             if driver_version < (12, 0):
                 patch_numba_linker_cuda_11()
             else:
-                patch_numba_linker_pynvjitlink()
+                from pynvjitlink.patch import patch_numba_linker
+
+                patch_numba_linker()
 
 
 def _get_cuda_version_from_ptx_file(path):