From a284f551d6b4b16a31dca1a14f3b760da8739458 Mon Sep 17 00:00:00 2001
From: Bradley Dice <bdice@bradleydice.com>
Date: Wed, 4 Dec 2024 18:20:22 -0600
Subject: [PATCH 1/4] Remove upper bounds on cuda-python to allow 12.6.2 and
 11.8.5 (#564)

Now that some upstream bugs have been fixed, we can allow cuda-python 12.6.2 and 11.8.5.

See https://github.com/NVIDIA/cuda-python/issues/226#issuecomment-2472355738 for more information.

Authors:
  - Bradley Dice (https://github.com/bdice)

Approvers:
  - James Lamb (https://github.com/jameslamb)

URL: https://github.com/rapidsai/kvikio/pull/564
---
 conda/environments/all_cuda-118_arch-aarch64.yaml | 2 +-
 conda/environments/all_cuda-118_arch-x86_64.yaml  | 2 +-
 conda/environments/all_cuda-125_arch-aarch64.yaml | 2 +-
 conda/environments/all_cuda-125_arch-x86_64.yaml  | 2 +-
 dependencies.yaml                                 | 4 ++--
 python/kvikio/pyproject.toml                      | 2 +-
 6 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/conda/environments/all_cuda-118_arch-aarch64.yaml b/conda/environments/all_cuda-118_arch-aarch64.yaml
index 2de4597657..54cbb12072 100644
--- a/conda/environments/all_cuda-118_arch-aarch64.yaml
+++ b/conda/environments/all_cuda-118_arch-aarch64.yaml
@@ -9,7 +9,7 @@ dependencies:
 - boto3>=1.21.21
 - c-compiler
 - cmake>=3.26.4,!=3.30.0
-- cuda-python>=11.7.1,<12.0a0,<=11.8.3
+- cuda-python>=11.7.1,<12.0a0
 - cuda-version=11.8
 - cudatoolkit
 - cupy>=12.0.0
diff --git a/conda/environments/all_cuda-118_arch-x86_64.yaml b/conda/environments/all_cuda-118_arch-x86_64.yaml
index 8b63a6726b..ff3da55951 100644
--- a/conda/environments/all_cuda-118_arch-x86_64.yaml
+++ b/conda/environments/all_cuda-118_arch-x86_64.yaml
@@ -9,7 +9,7 @@ dependencies:
 - boto3>=1.21.21
 - c-compiler
 - cmake>=3.26.4,!=3.30.0
-- cuda-python>=11.7.1,<12.0a0,<=11.8.3
+- cuda-python>=11.7.1,<12.0a0
 - cuda-version=11.8
 - cudatoolkit
 - cupy>=12.0.0
diff --git a/conda/environments/all_cuda-125_arch-aarch64.yaml b/conda/environments/all_cuda-125_arch-aarch64.yaml
index 87759faf53..41f6e6b3a3 100644
--- a/conda/environments/all_cuda-125_arch-aarch64.yaml
+++ b/conda/environments/all_cuda-125_arch-aarch64.yaml
@@ -10,7 +10,7 @@ dependencies:
 - c-compiler
 - cmake>=3.26.4,!=3.30.0
 - cuda-nvcc
-- cuda-python>=12.0,<13.0a0,<=12.6.0
+- cuda-python>=12.0,<13.0a0
 - cuda-version=12.5
 - cupy>=12.0.0
 - cxx-compiler
diff --git a/conda/environments/all_cuda-125_arch-x86_64.yaml b/conda/environments/all_cuda-125_arch-x86_64.yaml
index e0a10af655..cab86c6fdc 100644
--- a/conda/environments/all_cuda-125_arch-x86_64.yaml
+++ b/conda/environments/all_cuda-125_arch-x86_64.yaml
@@ -10,7 +10,7 @@ dependencies:
 - c-compiler
 - cmake>=3.26.4,!=3.30.0
 - cuda-nvcc
-- cuda-python>=12.0,<13.0a0,<=12.6.0
+- cuda-python>=12.0,<13.0a0
 - cuda-version=12.5
 - cupy>=12.0.0
 - cxx-compiler
diff --git a/dependencies.yaml b/dependencies.yaml
index 32d4f73c79..c541d05b6d 100644
--- a/dependencies.yaml
+++ b/dependencies.yaml
@@ -392,10 +392,10 @@ dependencies:
           - matrix:
               cuda: "12.*"
             packages:
-              - cuda-python>=12.0,<13.0a0,<=12.6.0
+              - cuda-python>=12.0,<13.0a0
           - matrix: # All CUDA 11 versions
             packages:
-              - cuda-python>=11.7.1,<12.0a0,<=11.8.3
+              - cuda-python>=11.7.1,<12.0a0
   test_java:
     common:
       - output_types: conda
diff --git a/python/kvikio/pyproject.toml b/python/kvikio/pyproject.toml
index 68d003c4d0..426ca726bd 100644
--- a/python/kvikio/pyproject.toml
+++ b/python/kvikio/pyproject.toml
@@ -41,7 +41,7 @@ classifiers = [
 [project.optional-dependencies]
 test = [
     "boto3>=1.21.21",
-    "cuda-python>=11.7.1,<12.0a0,<=11.8.3",
+    "cuda-python>=11.7.1,<12.0a0",
     "dask>=2022.05.2",
     "moto[server]>=4.0.8",
     "pytest",

From 3d2ab00d9dc279223f290cbbef07de36714087b9 Mon Sep 17 00:00:00 2001
From: "Mads R. B. Kristensen" <madsbk@gmail.com>
Date: Thu, 5 Dec 2024 10:06:00 +0100
Subject: [PATCH 2/4] cufile version (#565)

Fixes https://github.com/rapidsai/kvikio/issues/566

To clean up and make cufile capability detection more robust,  we now use `cuFileGetVersion()`.
Should fix the nightly failure: https://github.com/rapidsai/kvikio/actions/runs/12133477083/job/33877440755

NB:  `cuFileGetVersion()` first became available in cufile v1.8 (CTK v12.3) thus the stream and batch API detection will return false for  versions older than v1.8. I think this is acceptable for robustness.

Authors:
  - Mads R. B. Kristensen (https://github.com/madsbk)

Approvers:
  - Lawrence Mitchell (https://github.com/wence-)
  - Vyas Ramasubramani (https://github.com/vyasr)
  - Bradley Dice (https://github.com/bdice)

URL: https://github.com/rapidsai/kvikio/pull/565
---
 cpp/CMakeLists.txt                           | 17 +++-
 cpp/examples/basic_io.cpp                    |  2 +-
 cpp/include/kvikio/file_handle.hpp           | 29 +++---
 cpp/include/kvikio/remote_handle.hpp         |  2 +-
 cpp/include/kvikio/shim/cufile.hpp           | 99 +++++++++++++-------
 cpp/include/kvikio/shim/cufile_h_wrapper.hpp |  7 +-
 python/kvikio/kvikio/_lib/cufile_driver.pyx  |  5 +
 python/kvikio/kvikio/benchmarks/utils.py     |  6 ++
 python/kvikio/kvikio/cufile_driver.py        | 21 +++++
 python/kvikio/tests/test_cufile_driver.py    |  6 ++
 10 files changed, 140 insertions(+), 54 deletions(-)

diff --git a/cpp/CMakeLists.txt b/cpp/CMakeLists.txt
index 9c1450518b..a0639c5382 100644
--- a/cpp/CMakeLists.txt
+++ b/cpp/CMakeLists.txt
@@ -80,7 +80,7 @@ if(KvikIO_CUDA_SUPPORT)
   else()
     set(cuFile_FOUND 1)
 
-    # Check batch and stream API support (cuFile_BATCH_API_FOUND and cuFile_STREAM_API_FOUND)
+    # Check API support
     try_compile(
       cuFile_BATCH_API_FOUND SOURCE_FROM_CONTENT
       batch.cpp
@@ -109,6 +109,20 @@ if(KvikIO_CUDA_SUPPORT)
       OUTPUT_VARIABLE stream_output
     )
     message(STATUS "Found cuFile Stream API: ${cuFile_STREAM_API_FOUND}")
+    try_compile(
+      cuFile_VERSION_API_FOUND SOURCE_FROM_CONTENT
+      version.cpp
+      [[#include <cufile.h>
+      int main() {
+        int version;
+        cuFileGetVersion(&version);
+        return 0;
+      }
+      ]]
+      LINK_LIBRARIES CUDA::cuFile rt ${CMAKE_DL_LIBS}
+      OUTPUT_VARIABLE version_output
+    )
+    message(STATUS "Found cuFile Version API: ${cuFile_VERSION_API_FOUND}")
   endif()
 endif()
 
@@ -154,6 +168,7 @@ target_compile_definitions(
          $<$<BOOL:${cuFile_FOUND}>:KVIKIO_CUFILE_FOUND>
          $<$<BOOL:${cuFile_BATCH_API_FOUND}>:KVIKIO_CUFILE_BATCH_API_FOUND>
          $<$<BOOL:${cuFile_STREAM_API_FOUND}>:KVIKIO_CUFILE_STREAM_API_FOUND>
+         $<$<BOOL:${cuFile_VERSION_API_FOUND}>:KVIKIO_CUFILE_VERSION_API_FOUND>
 )
 
 set_target_properties(
diff --git a/cpp/examples/basic_io.cpp b/cpp/examples/basic_io.cpp
index 4d04391404..39bfc315cd 100644
--- a/cpp/examples/basic_io.cpp
+++ b/cpp/examples/basic_io.cpp
@@ -181,7 +181,7 @@ int main()
     cout << "Parallel POSIX read (" << kvikio::defaults::thread_pool_nthreads()
          << " threads): " << read << endl;
   }
-  if (kvikio::is_batch_and_stream_available() && !kvikio::defaults::is_compat_mode_preferred()) {
+  if (kvikio::is_batch_api_available() && !kvikio::defaults::is_compat_mode_preferred()) {
     std::cout << std::endl;
     Timer timer;
     // Here we use the batch API to read "/tmp/test-file" into `b_dev` by
diff --git a/cpp/include/kvikio/file_handle.hpp b/cpp/include/kvikio/file_handle.hpp
index abc6660de6..4880bd4f20 100644
--- a/cpp/include/kvikio/file_handle.hpp
+++ b/cpp/include/kvikio/file_handle.hpp
@@ -62,23 +62,20 @@ class FileHandle {
    */
   bool is_compat_mode_preferred_for_async(CompatMode requested_compat_mode)
   {
-    if (!defaults::is_compat_mode_preferred(requested_compat_mode)) {
-      if (!is_batch_and_stream_available()) {
-        if (requested_compat_mode == CompatMode::AUTO) { return true; }
-        throw std::runtime_error("Missing cuFile batch or stream library symbol.");
-      }
-
-      // When checking for availability, we also check if cuFile's config file exist. This is
-      // because even when the stream API is available, it doesn't work if no config file exist.
-      if (config_path().empty()) {
-        if (requested_compat_mode == CompatMode::AUTO) { return true; }
-        throw std::runtime_error("Missing cuFile configuration file.");
-      }
-
-      return false;
+    if (defaults::is_compat_mode_preferred(requested_compat_mode)) { return true; }
+
+    if (!is_stream_api_available()) {
+      if (requested_compat_mode == CompatMode::AUTO) { return true; }
+      throw std::runtime_error("Missing the cuFile stream api.");
     }
 
-    return true;
+    // When checking for availability, we also check if cuFile's config file exists. This is
+    // because even when the stream API is available, it doesn't work if no config file exists.
+    if (config_path().empty()) {
+      if (requested_compat_mode == CompatMode::AUTO) { return true; }
+      throw std::runtime_error("Missing cuFile configuration file.");
+    }
+    return false;
   }
 
  public:
@@ -670,7 +667,7 @@ class FileHandle {
    */
   [[nodiscard]] bool is_compat_mode_preferred_for_async() const noexcept
   {
-    static bool is_extra_symbol_available = is_batch_and_stream_available();
+    static bool is_extra_symbol_available = is_stream_api_available();
     static bool is_config_path_empty      = config_path().empty();
     return is_compat_mode_preferred() || !is_extra_symbol_available || is_config_path_empty;
   }
diff --git a/cpp/include/kvikio/remote_handle.hpp b/cpp/include/kvikio/remote_handle.hpp
index 8ac2798f31..e1b152b23c 100644
--- a/cpp/include/kvikio/remote_handle.hpp
+++ b/cpp/include/kvikio/remote_handle.hpp
@@ -41,7 +41,7 @@ namespace detail {
  * @note Is not thread-safe.
  */
 class BounceBufferH2D {
-  CUstream _stream;                 // The CUDA steam to use.
+  CUstream _stream;                 // The CUDA stream to use.
   CUdeviceptr _dev;                 // The output device buffer.
   AllocRetain::Alloc _host_buffer;  // The host buffer to bounce data on.
   std::ptrdiff_t _dev_offset{0};    // Number of bytes written to `_dev`.
diff --git a/cpp/include/kvikio/shim/cufile.hpp b/cpp/include/kvikio/shim/cufile.hpp
index 7f12c29c3d..5194d45e74 100644
--- a/cpp/include/kvikio/shim/cufile.hpp
+++ b/cpp/include/kvikio/shim/cufile.hpp
@@ -57,8 +57,11 @@ class cuFileAPI {
   decltype(cuFileDriverOpen)* DriverOpen{nullptr};
   decltype(cuFileDriverClose)* DriverClose{nullptr};
 
+  // Don't call `GetVersion` directly, use `cuFileAPI::instance().version`.
+  decltype(cuFileGetVersion)* GetVersion{nullptr};
+
  public:
-  bool stream_available = false;
+  int version{0};
 
  private:
 #ifdef KVIKIO_CUFILE_FOUND
@@ -88,33 +91,39 @@ class cuFileAPI {
     get_symbol(DriverSetMaxCacheSize, lib, KVIKIO_STRINGIFY(cuFileDriverSetMaxCacheSize));
     get_symbol(DriverSetMaxPinnedMemSize, lib, KVIKIO_STRINGIFY(cuFileDriverSetMaxPinnedMemSize));
 
-#ifdef KVIKIO_CUFILE_BATCH_API_FOUND
-    get_symbol(BatchIOSetUp, lib, KVIKIO_STRINGIFY(cuFileBatchIOSetUp));
-    get_symbol(BatchIOSubmit, lib, KVIKIO_STRINGIFY(cuFileBatchIOSubmit));
-    get_symbol(BatchIOGetStatus, lib, KVIKIO_STRINGIFY(cuFileBatchIOGetStatus));
-    get_symbol(BatchIOCancel, lib, KVIKIO_STRINGIFY(cuFileBatchIOCancel));
-    get_symbol(BatchIODestroy, lib, KVIKIO_STRINGIFY(cuFileBatchIODestroy));
-#endif
-
-#ifdef KVIKIO_CUFILE_STREAM_API_FOUND
-    get_symbol(ReadAsync, lib, KVIKIO_STRINGIFY(cuFileReadAsync));
-    get_symbol(WriteAsync, lib, KVIKIO_STRINGIFY(cuFileWriteAsync));
-    get_symbol(StreamRegister, lib, KVIKIO_STRINGIFY(cuFileStreamRegister));
-    get_symbol(StreamDeregister, lib, KVIKIO_STRINGIFY(cuFileStreamDeregister));
+#ifdef KVIKIO_CUFILE_VERSION_API_FOUND
     try {
-      void* s{};
-      get_symbol(s, lib, "cuFileReadAsync");
-      stream_available = true;
-    } catch (const std::runtime_error&) {
+      get_symbol(GetVersion, lib, KVIKIO_STRINGIFY(cuFileGetVersion));
+      int ver;
+      CUfileError_t const error = GetVersion(&ver);
+      if (error.err == CU_FILE_SUCCESS) { version = ver; }
+    } catch (std::runtime_error const&) {
     }
 #endif
 
+    // Some symbols were introduced in later versions, so version guards are required.
+    // Note: `version` is 0 for cuFile versions prior to v1.8 because `cuFileGetVersion`
+    // did not exist. As a result, the batch and stream APIs are not loaded in versions
+    // 1.6 and 1.7, respectively, even though they are available. This trade-off is made
+    // for improved robustness.
+    if (version >= 1060) {
+      get_symbol(BatchIOSetUp, lib, KVIKIO_STRINGIFY(cuFileBatchIOSetUp));
+      get_symbol(BatchIOSubmit, lib, KVIKIO_STRINGIFY(cuFileBatchIOSubmit));
+      get_symbol(BatchIOGetStatus, lib, KVIKIO_STRINGIFY(cuFileBatchIOGetStatus));
+      get_symbol(BatchIOCancel, lib, KVIKIO_STRINGIFY(cuFileBatchIOCancel));
+      get_symbol(BatchIODestroy, lib, KVIKIO_STRINGIFY(cuFileBatchIODestroy));
+    }
+    if (version >= 1070) {
+      get_symbol(ReadAsync, lib, KVIKIO_STRINGIFY(cuFileReadAsync));
+      get_symbol(WriteAsync, lib, KVIKIO_STRINGIFY(cuFileWriteAsync));
+      get_symbol(StreamRegister, lib, KVIKIO_STRINGIFY(cuFileStreamRegister));
+      get_symbol(StreamDeregister, lib, KVIKIO_STRINGIFY(cuFileStreamDeregister));
+    }
+
     // cuFile is supposed to open and close the driver automatically but
     // because of a bug in cuFile v1.4 (CUDA v11.8) it sometimes segfaults:
     // <https://github.com/rapidsai/kvikio/issues/159>.
-    // We use the stream API as a version indicator of cuFile since it was introduced
-    // in cuFile v1.7 (CUDA v12.2).
-    if (!stream_available) { driver_open(); }
+    if (version < 1050) { driver_open(); }
   }
 
   // Notice, we have to close the driver at program exit (if we opened it) even though we are
@@ -124,7 +133,7 @@ class cuFileAPI {
   // [1] <https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#initialization>
   ~cuFileAPI()
   {
-    if (!stream_available) { driver_close(); }
+    if (version < 1050) { driver_close(); }
   }
 #else
   cuFileAPI() { throw std::runtime_error("KvikIO not compiled with cuFile.h"); }
@@ -205,25 +214,49 @@ inline bool is_cufile_available()
 }
 
 /**
- * @brief Check if cuFile's batch and stream API is available
+ * @brief Get cufile version (or zero if older than v1.8).
  *
- * Technically, the batch API is available in CUDA 12.1 but since there is no good
- * way to check CUDA version using the driver API, we check for the existing of the
- * `cuFileReadAsync` symbol, which is defined in CUDA 12.2+.
+ * The version is returned as (1000*major + 10*minor). E.g., cufile v1.8.0 would
+ * be represented by 1080.
  *
- * @return The boolean answer
+ * Notice, this is not the version of the CUDA toolkit. cufile is part of the
+ * toolkit but follows its own version scheme.
+ *
+ * @return The version (1000*major + 10*minor) or zero if older than 1080.
  */
-#if defined(KVIKIO_CUFILE_STREAM_API_FOUND) && defined(KVIKIO_CUFILE_STREAM_API_FOUND)
-inline bool is_batch_and_stream_available() noexcept
+#ifdef KVIKIO_CUFILE_FOUND
+inline int cufile_version()
 {
   try {
-    return is_cufile_available() && cuFileAPI::instance().stream_available;
-  } catch (const std::runtime_error&) {
-    return false;
+    return cuFileAPI::instance().version;
+  } catch (std::runtime_error const&) {
+    return 0;
   }
 }
 #else
-constexpr bool is_batch_and_stream_available() { return false; }
+constexpr int cufile_version() { return 0; }
 #endif
 
+/**
+ * @brief Check if cuFile's batch API is available.
+ *
+ * Since `cuFileGetVersion()` first became available in cufile v1.8 (CTK v12.3),
+ * this function returns false for versions older than v1.8 even though the batch
+ * API became available in v1.6.
+ *
+ * @return The boolean answer
+ */
+inline bool is_batch_api_available() noexcept { return cufile_version() >= 1060; }
+
+/**
+ * @brief Check if cuFile's stream (async) API is available.
+ *
+ * Since `cuFileGetVersion()` first became available in cufile v1.8 (CTK v12.3),
+ * this function returns false for versions older than v1.8 even though the stream
+ * API became available in v1.7.
+ *
+ * @return The boolean answer
+ */
+inline bool is_stream_api_available() noexcept { return cufile_version() >= 1070; }
+
 }  // namespace kvikio
diff --git a/cpp/include/kvikio/shim/cufile_h_wrapper.hpp b/cpp/include/kvikio/shim/cufile_h_wrapper.hpp
index 33c3fee9a2..1c13d2d8a1 100644
--- a/cpp/include/kvikio/shim/cufile_h_wrapper.hpp
+++ b/cpp/include/kvikio/shim/cufile_h_wrapper.hpp
@@ -75,7 +75,7 @@ CUfileError_t cuFileDriverSetMaxPinnedMemSize(...);
 
 #endif
 
-// If the Batch API isn't defined, we define some of the data types here.
+// If some cufile APIs aren't defined, we define some of the data types here.
 // Notice, this doesn't need to be ABI compatible with the cufile definitions and
 // the lack of definitions is not a problem because the linker will never look for
 // these symbols because the "real" function calls are made through the shim instance.
@@ -105,10 +105,13 @@ CUfileError_t cuFileBatchIOCancel(...);
 CUfileError_t cuFileBatchIODestroy(...);
 #endif
 
-// If the Stream API isn't defined, we define some of the data types here.
 #ifndef KVIKIO_CUFILE_STREAM_API_FOUND
 CUfileError_t cuFileReadAsync(...);
 CUfileError_t cuFileWriteAsync(...);
 CUfileError_t cuFileStreamRegister(...);
 CUfileError_t cuFileStreamDeregister(...);
 #endif
+
+#ifndef KVIKIO_CUFILE_VERSION_API_FOUND
+CUfileError_t cuFileGetVersion(...);
+#endif
diff --git a/python/kvikio/kvikio/_lib/cufile_driver.pyx b/python/kvikio/kvikio/_lib/cufile_driver.pyx
index 29302a0104..0488eb3b20 100644
--- a/python/kvikio/kvikio/_lib/cufile_driver.pyx
+++ b/python/kvikio/kvikio/_lib/cufile_driver.pyx
@@ -9,10 +9,15 @@ from libcpp cimport bool
 
 
 cdef extern from "<kvikio/shim/cufile.hpp>" nogil:
+    cdef int cpp_libcufile_version "kvikio::cufile_version"() except +
     cdef void cpp_driver_open "kvikio::cuFileAPI::instance().driver_open"() except +
     cdef void cpp_driver_close "kvikio::cuFileAPI::instance().driver_close"() except +
 
 
+def libcufile_version() -> int:
+    return cpp_libcufile_version()
+
+
 def driver_open():
     cpp_driver_open()
 
diff --git a/python/kvikio/kvikio/benchmarks/utils.py b/python/kvikio/kvikio/benchmarks/utils.py
index 23c7731f24..fa25c361a4 100644
--- a/python/kvikio/kvikio/benchmarks/utils.py
+++ b/python/kvikio/kvikio/benchmarks/utils.py
@@ -27,6 +27,7 @@ def drop_vm_cache() -> None:
 def pprint_sys_info() -> None:
     """Pretty print system information"""
 
+    version = kvikio.cufile_driver.libcufile_version()
     props = kvikio.cufile_driver.DriverProperties()
     try:
         import pynvml
@@ -41,6 +42,10 @@ def pprint_sys_info() -> None:
         gpu_name = f"{pynvml.nvmlDeviceGetName(dev)} (dev #0)"
         mem_total = format_bytes(pynvml.nvmlDeviceGetMemoryInfo(dev).total)
         bar1_total = format_bytes(pynvml.nvmlDeviceGetBAR1MemoryInfo(dev).bar1Total)
+    if version == (0, 0):
+        libcufile_version = "unknown (earlier than cuFile 1.8)"
+    else:
+        libcufile_version = f"{version[0]}.{version[1]}"
     gds_version = "N/A (Compatibility Mode)"
     if props.is_gds_available:
         gds_version = f"v{props.major_version}.{props.minor_version}"
@@ -61,6 +66,7 @@ def pprint_sys_info() -> None:
     print(f"GPU               | {gpu_name}")
     print(f"GPU Memory Total  | {mem_total}")
     print(f"BAR1 Memory Total | {bar1_total}")
+    print(f"libcufile version | {libcufile_version}")
     print(f"GDS driver        | {gds_version}")
     print(f"GDS config.json   | {gds_config_json_path}")
 
diff --git a/python/kvikio/kvikio/cufile_driver.py b/python/kvikio/kvikio/cufile_driver.py
index e78242a514..fb32be347a 100644
--- a/python/kvikio/kvikio/cufile_driver.py
+++ b/python/kvikio/kvikio/cufile_driver.py
@@ -2,6 +2,7 @@
 # See file LICENSE for terms.
 
 import atexit
+from typing import Tuple
 
 from kvikio._lib import cufile_driver  # type: ignore
 
@@ -10,6 +11,26 @@
 DriverProperties = cufile_driver.DriverProperties
 
 
+def libcufile_version() -> Tuple[int, int]:
+    """Get the libcufile version.
+
+    Returns (0, 0) for cuFile versions prior to v1.8.
+
+    Notes
+    -----
+    This is not the version of the CUDA toolkit. cufile is part of the
+    toolkit but follows its own version scheme.
+
+    Returns
+    -------
+    The version as a tuple (MAJOR, MINOR).
+    """
+    v = cufile_driver.libcufile_version()
+    # Convert the integer version like 1080 to (1, 8).
+    major, minor = divmod(v, 1000)
+    return (major, minor // 10)
+
+
 def driver_open() -> None:
     """Open the cuFile driver
 
diff --git a/python/kvikio/tests/test_cufile_driver.py b/python/kvikio/tests/test_cufile_driver.py
index 0a64bf0952..a1dc3a6454 100644
--- a/python/kvikio/tests/test_cufile_driver.py
+++ b/python/kvikio/tests/test_cufile_driver.py
@@ -6,6 +6,12 @@
 import kvikio.cufile_driver
 
 
+def test_version():
+    major, minor = kvikio.cufile_driver.libcufile_version()
+    assert major >= 0
+    assert minor >= 0
+
+
 @pytest.mark.cufile
 def test_open_and_close():
     kvikio.cufile_driver.driver_open()

From dc6c0c1ffa22b61a0663ec266b4a23f1df173234 Mon Sep 17 00:00:00 2001
From: jakirkham <jakirkham@gmail.com>
Date: Fri, 6 Dec 2024 09:12:46 -0800
Subject: [PATCH 3/4] Add CUDA libs in Python Conda, Consolidate Conda CI
 installs & use `rapids-dask-dependency` (#513)

Explicitly include dependencies on `libcufile*` and `*cudart*` in the Python Conda package. These seem to be pulled in implicitly by the C++ package. However as they are dependencies of the Python built shared objects, they should also be listed here.

Also consolidate the Conda environment creation and installation of nightly dependencies into one step. By doing this in one step, we guarantee that all of our constraints are taken into account during environment creation.

Lastly switch from using `dask` directly to `rapids-dask-dependency` to align Dask installation and pinning with the rest of RAPIDS.

xref: https://github.com/rapidsai/build-planning/issues/22
xef: https://github.com/rapidsai/kvikio/issues/378

Authors:
  - https://github.com/jakirkham
  - Mads R. B. Kristensen (https://github.com/madsbk)

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

URL: https://github.com/rapidsai/kvikio/pull/513
---
 ci/build_docs.sh                              | 20 +++++++------------
 ci/release/update-version.sh                  |  3 +++
 ci/test_cpp.sh                                | 15 +++++---------
 ci/test_python.sh                             | 20 +++++++------------
 .../all_cuda-118_arch-aarch64.yaml            |  2 +-
 .../all_cuda-118_arch-x86_64.yaml             |  2 +-
 .../all_cuda-125_arch-aarch64.yaml            |  2 +-
 .../all_cuda-125_arch-x86_64.yaml             |  2 +-
 conda/recipes/kvikio/conda_build_config.yaml  | 17 ++++++++++++++--
 conda/recipes/kvikio/meta.yaml                | 12 ++++++++++-
 .../recipes/libkvikio/conda_build_config.yaml |  3 +++
 conda/recipes/libkvikio/meta.yaml             |  4 ++--
 dependencies.yaml                             | 17 +++++++++++++++-
 python/kvikio/pyproject.toml                  |  2 +-
 14 files changed, 74 insertions(+), 47 deletions(-)

diff --git a/ci/build_docs.sh b/ci/build_docs.sh
index 8c2ee8ae95..ea408b6940 100755
--- a/ci/build_docs.sh
+++ b/ci/build_docs.sh
@@ -6,28 +6,22 @@ set -euo pipefail
 rapids-logger "Create test conda environment"
 . /opt/conda/etc/profile.d/conda.sh
 
-export RAPIDS_VERSION="$(rapids-version)"
+rapids-logger "Downloading artifacts from previous jobs"
+CPP_CHANNEL=$(rapids-download-conda-from-s3 cpp)
+PYTHON_CHANNEL=$(rapids-download-conda-from-s3 python)
 
 rapids-dependency-file-generator \
   --output conda \
   --file-key docs \
-  --matrix "cuda=${RAPIDS_CUDA_VERSION%.*};arch=$(arch);py=${RAPIDS_PY_VERSION}" | tee env.yaml
+  --matrix "cuda=${RAPIDS_CUDA_VERSION%.*};arch=$(arch);py=${RAPIDS_PY_VERSION}" \
+  --prepend-channel "${CPP_CHANNEL}" --prepend-channel "${PYTHON_CHANNEL}" \
+  | tee env.yaml
 
-rapids-mamba-retry env create --yes -f env.yaml -n docs
+rapids-mamba-retry env create -yq -f env.yaml -n docs
 conda activate docs
 
 rapids-print-env
 
-rapids-logger "Downloading artifacts from previous jobs"
-CPP_CHANNEL=$(rapids-download-conda-from-s3 cpp)
-PYTHON_CHANNEL=$(rapids-download-conda-from-s3 python)
-
-rapids-mamba-retry install \
-  --channel "${CPP_CHANNEL}" \
-  --channel "${PYTHON_CHANNEL}" \
-  "kvikio=${RAPIDS_VERSION}" \
-  "libkvikio=${RAPIDS_VERSION}"
-
 export RAPIDS_DOCS_DIR="$(mktemp -d)"
 
 rapids-logger "Build CPP docs"
diff --git a/ci/release/update-version.sh b/ci/release/update-version.sh
index 6b9458b273..102beaa2ba 100755
--- a/ci/release/update-version.sh
+++ b/ci/release/update-version.sh
@@ -36,7 +36,10 @@ function sed_runner() {
 echo "${NEXT_FULL_TAG}" > VERSION
 
 DEPENDENCIES=(
+  kvikio
   libkvikio
+  libkvikio-tests
+  rapids-dask-dependency
 )
 for DEP in "${DEPENDENCIES[@]}"; do
   for FILE in dependencies.yaml conda/environments/*.yaml; do
diff --git a/ci/test_cpp.sh b/ci/test_cpp.sh
index 7dfde2187f..ef7933f150 100755
--- a/ci/test_cpp.sh
+++ b/ci/test_cpp.sh
@@ -5,33 +5,28 @@ set -euo pipefail
 
 . /opt/conda/etc/profile.d/conda.sh
 
-RAPIDS_VERSION="$(rapids-version)"
+CPP_CHANNEL=$(rapids-download-conda-from-s3 cpp)
 
 rapids-logger "Generate C++ testing dependencies"
 rapids-dependency-file-generator \
   --output conda \
   --file-key test_cpp \
-  --matrix "cuda=${RAPIDS_CUDA_VERSION%.*};arch=$(arch)" | tee env.yaml
+  --matrix "cuda=${RAPIDS_CUDA_VERSION%.*};arch=$(arch)" \
+  --prepend-channel "${CPP_CHANNEL}" \
+  | tee env.yaml
 
-rapids-mamba-retry env create --yes -f env.yaml -n test
+rapids-mamba-retry env create -qy -f env.yaml -n test
 
 # Temporarily allow unbound variables for conda activation.
 set +u
 conda activate test
 set -u
 
-CPP_CHANNEL=$(rapids-download-conda-from-s3 cpp)
 RAPIDS_TESTS_DIR=${RAPIDS_TESTS_DIR:-"${PWD}/test-results"}/
 mkdir -p "${RAPIDS_TESTS_DIR}"
-SUITEERROR=0
 
 rapids-print-env
 
-rapids-mamba-retry install \
-  --channel "${CPP_CHANNEL}" \
-  "libkvikio=${RAPIDS_VERSION}" \
-  "libkvikio-tests=${RAPIDS_VERSION}"
-
 rapids-logger "Check GPU usage"
 nvidia-smi
 
diff --git a/ci/test_python.sh b/ci/test_python.sh
index 1f4ff6d2f1..df16f20f9f 100755
--- a/ci/test_python.sh
+++ b/ci/test_python.sh
@@ -8,37 +8,31 @@ cd "$(dirname "$(realpath "${BASH_SOURCE[0]}")")"/../
 
 . /opt/conda/etc/profile.d/conda.sh
 
-RAPIDS_VERSION="$(rapids-version)"
+rapids-logger "Downloading artifacts from previous jobs"
+CPP_CHANNEL=$(rapids-download-conda-from-s3 cpp)
+PYTHON_CHANNEL=$(rapids-download-conda-from-s3 python)
 
 rapids-logger "Generate Python testing dependencies"
 rapids-dependency-file-generator \
   --output conda \
   --file-key test_python \
-  --matrix "cuda=${RAPIDS_CUDA_VERSION%.*};arch=$(arch);py=${RAPIDS_PY_VERSION}" | tee env.yaml
+  --matrix "cuda=${RAPIDS_CUDA_VERSION%.*};arch=$(arch);py=${RAPIDS_PY_VERSION}" \
+  --prepend-channel "${CPP_CHANNEL}" --prepend-channel "${PYTHON_CHANNEL}" \
+  | tee env.yaml
 
-rapids-mamba-retry env create --yes -f env.yaml -n test
+rapids-mamba-retry env create -qy -f env.yaml -n test
 
 # Temporarily allow unbound variables for conda activation.
 set +u
 conda activate test
 set -u
 
-rapids-logger "Downloading artifacts from previous jobs"
-CPP_CHANNEL=$(rapids-download-conda-from-s3 cpp)
-PYTHON_CHANNEL=$(rapids-download-conda-from-s3 python)
-
 RAPIDS_TESTS_DIR=${RAPIDS_TESTS_DIR:-"${PWD}/test-results"}
 RAPIDS_COVERAGE_DIR=${RAPIDS_COVERAGE_DIR:-"${PWD}/coverage-results"}
 mkdir -p "${RAPIDS_TESTS_DIR}" "${RAPIDS_COVERAGE_DIR}"
 
 rapids-print-env
 
-rapids-mamba-retry install \
-  --channel "${CPP_CHANNEL}" \
-  --channel "${PYTHON_CHANNEL}" \
-  "libkvikio=${RAPIDS_VERSION}" \
-  "kvikio=${RAPIDS_VERSION}"
-
 rapids-logger "Check GPU usage"
 nvidia-smi
 
diff --git a/conda/environments/all_cuda-118_arch-aarch64.yaml b/conda/environments/all_cuda-118_arch-aarch64.yaml
index 2de4597657..7fee941a1d 100644
--- a/conda/environments/all_cuda-118_arch-aarch64.yaml
+++ b/conda/environments/all_cuda-118_arch-aarch64.yaml
@@ -15,7 +15,6 @@ dependencies:
 - cupy>=12.0.0
 - cxx-compiler
 - cython>=3.0.0
-- dask>=2022.05.2
 - doxygen=1.9.1
 - gcc_linux-aarch64=11.*
 - libcurl>=7.87.0
@@ -33,6 +32,7 @@ dependencies:
 - python>=3.10,<3.13
 - rangehttpserver
 - rapids-build-backend>=0.3.0,<0.4.0.dev0
+- rapids-dask-dependency==24.12.*,>=0.0.0a0
 - scikit-build-core>=0.10.0
 - sphinx
 - sphinx-click
diff --git a/conda/environments/all_cuda-118_arch-x86_64.yaml b/conda/environments/all_cuda-118_arch-x86_64.yaml
index 8b63a6726b..b89468e062 100644
--- a/conda/environments/all_cuda-118_arch-x86_64.yaml
+++ b/conda/environments/all_cuda-118_arch-x86_64.yaml
@@ -15,7 +15,6 @@ dependencies:
 - cupy>=12.0.0
 - cxx-compiler
 - cython>=3.0.0
-- dask>=2022.05.2
 - doxygen=1.9.1
 - gcc_linux-64=11.*
 - libcufile-dev=1.4.0.31
@@ -35,6 +34,7 @@ dependencies:
 - python>=3.10,<3.13
 - rangehttpserver
 - rapids-build-backend>=0.3.0,<0.4.0.dev0
+- rapids-dask-dependency==24.12.*,>=0.0.0a0
 - scikit-build-core>=0.10.0
 - sphinx
 - sphinx-click
diff --git a/conda/environments/all_cuda-125_arch-aarch64.yaml b/conda/environments/all_cuda-125_arch-aarch64.yaml
index 87759faf53..e12ab092ef 100644
--- a/conda/environments/all_cuda-125_arch-aarch64.yaml
+++ b/conda/environments/all_cuda-125_arch-aarch64.yaml
@@ -15,7 +15,6 @@ dependencies:
 - cupy>=12.0.0
 - cxx-compiler
 - cython>=3.0.0
-- dask>=2022.05.2
 - doxygen=1.9.1
 - gcc_linux-aarch64=11.*
 - libcufile-dev
@@ -33,6 +32,7 @@ dependencies:
 - python>=3.10,<3.13
 - rangehttpserver
 - rapids-build-backend>=0.3.0,<0.4.0.dev0
+- rapids-dask-dependency==24.12.*,>=0.0.0a0
 - scikit-build-core>=0.10.0
 - sphinx
 - sphinx-click
diff --git a/conda/environments/all_cuda-125_arch-x86_64.yaml b/conda/environments/all_cuda-125_arch-x86_64.yaml
index e0a10af655..131efa85d5 100644
--- a/conda/environments/all_cuda-125_arch-x86_64.yaml
+++ b/conda/environments/all_cuda-125_arch-x86_64.yaml
@@ -15,7 +15,6 @@ dependencies:
 - cupy>=12.0.0
 - cxx-compiler
 - cython>=3.0.0
-- dask>=2022.05.2
 - doxygen=1.9.1
 - gcc_linux-64=11.*
 - libcufile-dev
@@ -33,6 +32,7 @@ dependencies:
 - python>=3.10,<3.13
 - rangehttpserver
 - rapids-build-backend>=0.3.0,<0.4.0.dev0
+- rapids-dask-dependency==24.12.*,>=0.0.0a0
 - scikit-build-core>=0.10.0
 - sphinx
 - sphinx-click
diff --git a/conda/recipes/kvikio/conda_build_config.yaml b/conda/recipes/kvikio/conda_build_config.yaml
index 8af5f6f9c2..639a56f509 100644
--- a/conda/recipes/kvikio/conda_build_config.yaml
+++ b/conda/recipes/kvikio/conda_build_config.yaml
@@ -4,6 +4,9 @@ c_compiler_version:
 cxx_compiler_version:
   - 11
 
+cmake_version:
+  - ">=3.26.4,!=3.30.0"
+
 cuda_compiler:
   - cuda-nvcc
 
@@ -16,8 +19,18 @@ c_stdlib:
 c_stdlib_version:
   - "2.17"
 
-cmake_version:
-  - ">=3.26.4,!=3.30.0"
+# The CTK libraries below are missing from the conda-forge::cudatoolkit package
+# for CUDA 11. The "*_host_*" version specifiers correspond to `11.8` packages
+# and the "*_run_*" version specifiers correspond to `11.x` packages.
+
+cuda11_libcufile_host_version:
+  - "1.4.0.31"
+
+cuda11_libcufile_run_version:
+  - ">=1.0.0.82,<=1.4.0.31"
+
+libcurl_version:
+  - "==7.87.0"
 
 nvcomp_version:
   - "=4.1.0.6"
diff --git a/conda/recipes/kvikio/meta.yaml b/conda/recipes/kvikio/meta.yaml
index 3c41af3310..271712ac51 100644
--- a/conda/recipes/kvikio/meta.yaml
+++ b/conda/recipes/kvikio/meta.yaml
@@ -37,6 +37,8 @@ build:
     - {{ compiler('cuda11') }}
     {% else %}
     - {{ compiler('cuda') }}
+    - cuda-cudart-dev
+    - libcufile-dev  # [linux]
     {% endif %}
 
 requirements:
@@ -58,13 +60,17 @@ requirements:
     - cython >=3.0.0
     {% if cuda_major == "11" %}
     - cudatoolkit
+    - libcufile {{ cuda11_libcufile_run_version }}  # [linux64]
+    {% else %}
+    - cuda-cudart-dev
+    - libcufile-dev  # [linux]
     {% endif %}
     - cuda-version ={{ cuda_version }}
     - nvcomp {{ nvcomp_version }}
     - rapids-build-backend >=0.3.0,<0.4.0.dev0
     - scikit-build-core >=0.10.0
     - libkvikio ={{ version }}
-    - libcurl==7.87.0
+    - libcurl {{ libcurl_version }}
   run:
     - python
     - numpy >=1.23,<3.0a0
@@ -76,6 +82,10 @@ requirements:
     - {{ pin_compatible('cuda-version', max_pin='x', min_pin='x') }}
     {% if cuda_major == "11" %}
     - cudatoolkit
+    - libcufile {{ cuda11_libcufile_run_version }}  # [linux64]
+    {% else %}
+    - cuda-cudart
+    - libcufile  # [linux]
     {% endif %}
 
 test:
diff --git a/conda/recipes/libkvikio/conda_build_config.yaml b/conda/recipes/libkvikio/conda_build_config.yaml
index 9cf2923599..b895b842f3 100644
--- a/conda/recipes/libkvikio/conda_build_config.yaml
+++ b/conda/recipes/libkvikio/conda_build_config.yaml
@@ -28,3 +28,6 @@ cuda11_libcufile_host_version:
 
 cuda11_libcufile_run_version:
   - ">=1.0.0.82,<=1.4.0.31"
+
+libcurl_version:
+  - "==7.87.0"
diff --git a/conda/recipes/libkvikio/meta.yaml b/conda/recipes/libkvikio/meta.yaml
index 999b9fc2c1..4019a55ec8 100644
--- a/conda/recipes/libkvikio/meta.yaml
+++ b/conda/recipes/libkvikio/meta.yaml
@@ -52,7 +52,7 @@ requirements:
     {% else %}
     - libcufile-dev  # [linux]
     {% endif %}
-    - libcurl==7.87.0
+    - libcurl {{ libcurl_version }}
 
 outputs:
   - name: libkvikio
@@ -75,7 +75,7 @@ outputs:
         - cmake {{ cmake_version }}
       host:
         - cuda-version ={{ cuda_version }}
-        - libcurl==7.87.0
+        - libcurl {{ libcurl_version }}
       run:
         - {{ pin_compatible('cuda-version', max_pin='x', min_pin='x') }}
         {% if cuda_major == "11" %}
diff --git a/dependencies.yaml b/dependencies.yaml
index fe85506ce5..5c35e5192c 100644
--- a/dependencies.yaml
+++ b/dependencies.yaml
@@ -24,12 +24,14 @@ files:
     output: none
     includes:
       - cuda_version
+      - test_libkvikio
       - test_cpp
   test_python:
     output: none
     includes:
       - cuda_version
       - py_version
+      - test_kvikio
       - test_python
   checks:
     output: none
@@ -42,6 +44,7 @@ files:
       - cuda_version
       - docs
       - py_version
+      - test_kvikio
   py_build_kvikio:
     output: pyproject
     pyproject_dir: python/kvikio
@@ -358,6 +361,18 @@ dependencies:
           # See https://github.com/zarr-developers/numcodecs/pull/475
           - numcodecs !=0.12.0
           - packaging
+  test_libkvikio:
+    common:
+      - output_types: conda
+        packages:
+          - libkvikio==24.12.*,>=0.0.0a0
+          - libkvikio-tests==24.12.*,>=0.0.0a0
+  test_kvikio:
+    common:
+      - output_types: conda
+        packages:
+          - libkvikio==24.12.*,>=0.0.0a0
+          - kvikio==24.12.*,>=0.0.0a0
   test_cpp:
     common:
       - output_types: conda
@@ -367,7 +382,7 @@ dependencies:
     common:
       - output_types: [conda, requirements, pyproject]
         packages:
-          - &dask dask>=2022.05.2
+          - rapids-dask-dependency==24.12.*,>=0.0.0a0
           - pytest
           - pytest-cov
           - rangehttpserver
diff --git a/python/kvikio/pyproject.toml b/python/kvikio/pyproject.toml
index cafde598bd..4869a63fab 100644
--- a/python/kvikio/pyproject.toml
+++ b/python/kvikio/pyproject.toml
@@ -42,11 +42,11 @@ classifiers = [
 test = [
     "boto3>=1.21.21",
     "cuda-python>=11.7.1,<12.0a0,<=11.8.3",
-    "dask>=2022.05.2",
     "moto[server]>=4.0.8",
     "pytest",
     "pytest-cov",
     "rangehttpserver",
+    "rapids-dask-dependency==24.12.*,>=0.0.0a0",
 ] # This list was generated by `rapids-dependency-file-generator`. To make changes, edit ../../dependencies.yaml and run `rapids-dependency-file-generator`.
 
 [project.urls]

From 924a8519a12a4d83e5153a153f14a42170ce90b4 Mon Sep 17 00:00:00 2001
From: Bradley Dice <bdice@bradleydice.com>
Date: Sat, 7 Dec 2024 00:35:16 -0600
Subject: [PATCH 4/4] Update cuda-python lower bounds to 12.6.2 / 11.8.5 (#573)

We require a newer cuda-python lower bound for new features and to use the new layout.
This will fix a number of errors observed when the runtime version of cuda-python is older than the version used to build packages using Cython features from cuda-python.

See https://github.com/rapidsai/build-planning/issues/117#issuecomment-2524250915 for details.

Authors:
  - Bradley Dice (https://github.com/bdice)

Approvers:
  - James Lamb (https://github.com/jameslamb)

URL: https://github.com/rapidsai/kvikio/pull/573
---
 conda/environments/all_cuda-118_arch-aarch64.yaml | 2 +-
 conda/environments/all_cuda-118_arch-x86_64.yaml  | 2 +-
 conda/environments/all_cuda-125_arch-aarch64.yaml | 2 +-
 conda/environments/all_cuda-125_arch-x86_64.yaml  | 2 +-
 dependencies.yaml                                 | 4 ++--
 python/kvikio/pyproject.toml                      | 2 +-
 6 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/conda/environments/all_cuda-118_arch-aarch64.yaml b/conda/environments/all_cuda-118_arch-aarch64.yaml
index 78f7dffb6b..328ae17a46 100644
--- a/conda/environments/all_cuda-118_arch-aarch64.yaml
+++ b/conda/environments/all_cuda-118_arch-aarch64.yaml
@@ -9,7 +9,7 @@ dependencies:
 - boto3>=1.21.21
 - c-compiler
 - cmake>=3.26.4,!=3.30.0
-- cuda-python>=11.7.1,<12.0a0
+- cuda-python>=11.8.5,<12.0a0
 - cuda-version=11.8
 - cudatoolkit
 - cupy>=12.0.0
diff --git a/conda/environments/all_cuda-118_arch-x86_64.yaml b/conda/environments/all_cuda-118_arch-x86_64.yaml
index 36b3745b52..d1dad00c50 100644
--- a/conda/environments/all_cuda-118_arch-x86_64.yaml
+++ b/conda/environments/all_cuda-118_arch-x86_64.yaml
@@ -9,7 +9,7 @@ dependencies:
 - boto3>=1.21.21
 - c-compiler
 - cmake>=3.26.4,!=3.30.0
-- cuda-python>=11.7.1,<12.0a0
+- cuda-python>=11.8.5,<12.0a0
 - cuda-version=11.8
 - cudatoolkit
 - cupy>=12.0.0
diff --git a/conda/environments/all_cuda-125_arch-aarch64.yaml b/conda/environments/all_cuda-125_arch-aarch64.yaml
index dde490d582..53ad0ab929 100644
--- a/conda/environments/all_cuda-125_arch-aarch64.yaml
+++ b/conda/environments/all_cuda-125_arch-aarch64.yaml
@@ -10,7 +10,7 @@ dependencies:
 - c-compiler
 - cmake>=3.26.4,!=3.30.0
 - cuda-nvcc
-- cuda-python>=12.0,<13.0a0
+- cuda-python>=12.6.2,<13.0a0
 - cuda-version=12.5
 - cupy>=12.0.0
 - cxx-compiler
diff --git a/conda/environments/all_cuda-125_arch-x86_64.yaml b/conda/environments/all_cuda-125_arch-x86_64.yaml
index 56a7b2b5df..68b9a72f6b 100644
--- a/conda/environments/all_cuda-125_arch-x86_64.yaml
+++ b/conda/environments/all_cuda-125_arch-x86_64.yaml
@@ -10,7 +10,7 @@ dependencies:
 - c-compiler
 - cmake>=3.26.4,!=3.30.0
 - cuda-nvcc
-- cuda-python>=12.0,<13.0a0
+- cuda-python>=12.6.2,<13.0a0
 - cuda-version=12.5
 - cupy>=12.0.0
 - cxx-compiler
diff --git a/dependencies.yaml b/dependencies.yaml
index aa3bc15016..6a715e08d7 100644
--- a/dependencies.yaml
+++ b/dependencies.yaml
@@ -407,10 +407,10 @@ dependencies:
           - matrix:
               cuda: "12.*"
             packages:
-              - cuda-python>=12.0,<13.0a0
+              - cuda-python>=12.6.2,<13.0a0
           - matrix: # All CUDA 11 versions
             packages:
-              - cuda-python>=11.7.1,<12.0a0
+              - cuda-python>=11.8.5,<12.0a0
   test_java:
     common:
       - output_types: conda
diff --git a/python/kvikio/pyproject.toml b/python/kvikio/pyproject.toml
index 661d9976bc..c9a9ab80d2 100644
--- a/python/kvikio/pyproject.toml
+++ b/python/kvikio/pyproject.toml
@@ -41,7 +41,7 @@ classifiers = [
 [project.optional-dependencies]
 test = [
     "boto3>=1.21.21",
-    "cuda-python>=11.7.1,<12.0a0",
+    "cuda-python>=11.8.5,<12.0a0",
     "moto[server]>=4.0.8",
     "pytest",
     "pytest-cov",