numba-cuda 0.12.1__py3-none-any.whl → 0.13.0__py3-none-any.whl

numba_cuda/VERSION

@@ -1 +1 @@
-0.12.1
+0.13.0

numba_cuda/numba/cuda/codegen.py

@@ -5,7 +5,7 @@ from numba.core.codegen import Codegen, CodeLibrary
 from .cudadrv import devices, driver, nvvm, runtime
 from numba.cuda.cudadrv.libs import get_cudalib
 from numba.cuda.cudadrv.linkable_code import LinkableCode
-from numba.cuda.runtime.nrt import NRT_LIBRARY
+from numba.cuda.memory_management.nrt import NRT_LIBRARY
 
 import os
 import subprocess

numba_cuda/numba/cuda/cudadrv/nvrtc.py

@@ -397,7 +397,7 @@ def compile(src, name, cc, ltoir=False):
     else:
         extra_includes = []
 
-    nrt_path = os.path.join(numba_cuda_path, "runtime")
+    nrt_path = os.path.join(numba_cuda_path, "memory_management")
     nrt_include = f"-I{nrt_path}"
 
     options = [

numba_cuda/numba/cuda/dispatcher.py

@@ -27,8 +27,8 @@ from numba.cuda.errors import (
     normalize_kernel_dimensions,
 )
 from numba.cuda import types as cuda_types
-from numba.cuda.runtime.nrt import rtsys, NRT_LIBRARY
 from numba.cuda.locks import module_init_lock
+from numba.cuda.memory_management.nrt import rtsys, NRT_LIBRARY
 
 from numba import cuda
 from numba import _dispatcher

numba_cuda/numba/cuda/memory_management/__init__.py

@@ -0,0 +1 @@
+from numba.cuda.memory_management.nrt import rtsys  # noqa: F401

numba_cuda/numba/cuda/simulator/__init__.py

@@ -38,11 +38,20 @@ if config.ENABLE_CUDASIM:
     sys.modules["numba.cuda.cudadrv.devicearray"] = cudadrv.devicearray
     sys.modules["numba.cuda.cudadrv.devices"] = cudadrv.devices
     sys.modules["numba.cuda.cudadrv.driver"] = cudadrv.driver
+    sys.modules["numba.cuda.cudadrv.linkable_code"] = cudadrv.linkable_code
     sys.modules["numba.cuda.cudadrv.runtime"] = cudadrv.runtime
     sys.modules["numba.cuda.cudadrv.drvapi"] = cudadrv.drvapi
     sys.modules["numba.cuda.cudadrv.error"] = cudadrv.error
     sys.modules["numba.cuda.cudadrv.nvvm"] = cudadrv.nvvm
 
-    from . import compiler
+    from . import bf16, compiler, _internal
 
+    sys.modules["numba.cuda.bf16"] = bf16
     sys.modules["numba.cuda.compiler"] = compiler
+    sys.modules["numba.cuda._internal"] = _internal
+    sys.modules["numba.cuda._internal.cuda_bf16"] = _internal.cuda_bf16
+
+    from numba.cuda.simulator import memory_management
+
+    sys.modules["numba.cuda.memory_management"] = memory_management
+    sys.modules["numba.cuda.memory_management.nrt"] = memory_management.nrt

numba_cuda/numba/cuda/simulator/_internal/__init__.py

@@ -0,0 +1 @@
+from numba.cuda.simulator._internal import cuda_bf16  # noqa: F401

numba_cuda/numba/cuda/simulator/api.py

@@ -7,6 +7,15 @@ Contains CUDA API functions
 from contextlib import contextmanager
 
 from .cudadrv.devices import require_context, reset, gpus  # noqa: F401
+from .cudadrv.linkable_code import (
+    PTXSource,  # noqa: F401
+    CUSource,  # noqa: F401
+    Cubin,  # noqa: F401
+    Fatbin,  # noqa: F401
+    Archive,  # noqa: F401
+    Object,  # noqa: F401
+    LTOIR,  # noqa: F401
+)  # noqa: F401
 from .kernel import FakeCUDAKernel
 from numba.core.sigutils import is_signature
 from numba.core import config
@@ -22,6 +31,10 @@ def is_float16_supported():
     return True
 
 
+def is_bfloat16_supported():
+    return False
+
+
 class stream(object):
     """
     The stream API is supported in the simulator - however, all execution
@@ -72,6 +85,10 @@ def list_devices():
     return gpus
 
 
+def get_current_device():
+    return gpus[0].device
+
+
 # Events
 
 

numba_cuda/numba/cuda/simulator/bf16.py

@@ -0,0 +1 @@
+bfloat16 = None

numba_cuda/numba/cuda/simulator/compiler.py

@@ -7,3 +7,4 @@ compile = None
 compile_for_current_device = None
 compile_ptx = None
 compile_ptx_for_current_device = None
+declare_device_function = None

numba_cuda/numba/cuda/simulator/cudadrv/driver.py

@@ -3,6 +3,8 @@ Most of the driver API is unsupported in the simulator, but some stubs are
 provided to allow tests to import correctly.
 """
 
+from numba import config
+
 
 def device_memset(dst, val, size, stream=0):
     dst.view("u1")[:size].fill(bytes([val])[0])
@@ -60,3 +62,8 @@ def launch_kernel(*args, **kwargs):
 
 
 USE_NV_BINDING = False
+
+PyNvJitLinker = None
+
+if config.ENABLE_CUDASIM:
+    config.CUDA_ENABLE_PYNVJITLINK = False

numba_cuda/numba/cuda/simulator/cudadrv/libs.py

@@ -1,2 +1,6 @@
 def check_static_lib(lib):
     raise FileNotFoundError("Linking libraries not supported by cudasim")
+
+
+def get_cuda_include_dir():
+    raise FileNotFoundError("CUDA includes not supported by cudasim")

numba_cuda/numba/cuda/simulator/cudadrv/linkable_code.py

@@ -0,0 +1,57 @@
+class LinkableCode:
+    """An object that holds code to be linked from memory.
+
+    :param data: A buffer containing the data to link.
+    :param name: The name of the file to be referenced in any compilation or
+                 linking errors that may be produced.
+    """
+
+    def __init__(self, data, name=None):
+        self.data = data
+        self._name = name
+
+    @property
+    def name(self):
+        return self._name or self.default_name
+
+
+class PTXSource(LinkableCode):
+    """PTX source code in memory."""
+
+    default_name = "<unnamed-ptx>"
+
+
+class CUSource(LinkableCode):
+    """CUDA C/C++ source code in memory."""
+
+    default_name = "<unnamed-cu>"
+
+
+class Fatbin(LinkableCode):
+    """An ELF Fatbin in memory."""
+
+    default_name = "<unnamed-fatbin>"
+
+
+class Cubin(LinkableCode):
+    """An ELF Cubin in memory."""
+
+    default_name = "<unnamed-cubin>"
+
+
+class Archive(LinkableCode):
+    """An archive of objects in memory."""
+
+    default_name = "<unnamed-archive>"
+
+
+class Object(LinkableCode):
+    """An object file in memory."""
+
+    default_name = "<unnamed-object>"
+
+
+class LTOIR(LinkableCode):
+    """An LTOIR file in memory."""
+
+    default_name = "<unnamed-ltoir>"

numba_cuda/numba/cuda/simulator/cudadrv/nvrtc.py

@@ -0,0 +1,8 @@
+"""
+NVVM is not supported in the simulator, but stubs are provided to allow tests
+to import correctly.
+"""
+
+
+def compile(src, name, cc, ltoir=False):
+    raise RuntimeError("NVRTC is not supported in the simulator")

numba_cuda/numba/cuda/simulator/kernel.py

@@ -78,7 +78,7 @@ class FakeCUDAKernel(object):
         functools.update_wrapper(self, fn)
 
     def __call__(self, *args):
-        if self._device:
+        if self._device or _kernel_context:
             with swapped_cuda_module(self.fn, _get_kernel_context()):
                 return self.fn(*args)
 

numba_cuda/numba/cuda/simulator/kernelapi.py

@@ -63,7 +63,10 @@ class FakeCUDALocal(object):
     CUDA Local arrays
     """
 
-    def array(self, shape, dtype):
+    def array(self, shape, dtype, alignment=None):
+        if alignment is not None:
+            raise RuntimeError("Array alignment is not supported in cudasim")
+
         if isinstance(dtype, types.Type):
             dtype = numpy_support.as_dtype(dtype)
         return np.empty(shape, dtype)
@@ -102,7 +105,10 @@ class FakeCUDAShared(object):
         self._dynshared_size = dynshared_size
         self._dynshared = np.zeros(dynshared_size, dtype=np.byte)
 
-    def array(self, shape, dtype):
+    def array(self, shape, dtype, alignment=None):
+        if alignment is not None:
+            raise RuntimeError("Array alignment is not supported in cudasim")
+
         if isinstance(dtype, types.Type):
             dtype = numpy_support.as_dtype(dtype)
         # Dynamic shared memory is requested with size 0 - this all shares the

numba_cuda/numba/cuda/simulator/memory_management/__init__.py

@@ -0,0 +1 @@
+from .nrt import rtsys  # noqa: F401

numba_cuda/numba/cuda/simulator/memory_management/nrt.py

@@ -0,0 +1,6 @@
+from numba import config
+
+rtsys = None
+
+config.CUDA_NRT_STATS = False
+config.CUDA_ENABLE_NRT = False

numba_cuda/numba/cuda/testing.py

@@ -116,20 +116,26 @@ def skip_on_arm(reason):
 def skip_if_cuda_includes_missing(fn):
     # Skip when cuda.h is not available - generally this should indicate
     # whether the CUDA includes are available or not
-    cuda_include_path = libs.get_cuda_include_dir()
+    reason = "CUDA include dir not available on this system"
+    try:
+        cuda_include_path = libs.get_cuda_include_dir()
+    except FileNotFoundError:
+        return unittest.skip(reason)(fn)
     cuda_h = os.path.join(cuda_include_path, "cuda.h")
     cuda_h_file = os.path.exists(cuda_h) and os.path.isfile(cuda_h)
-    reason = "CUDA include dir not available on this system"
     return unittest.skipUnless(cuda_h_file, reason)(fn)
 
 
 def skip_if_curand_kernel_missing(fn):
-    cuda_include_path = libs.get_cuda_include_dir()
+    reason = "curand_kernel.h not available on this system"
+    try:
+        cuda_include_path = libs.get_cuda_include_dir()
+    except FileNotFoundError:
+        return unittest.skip(reason)(fn)
     curand_kernel_h = os.path.join(cuda_include_path, "curand_kernel.h")
     curand_kernel_h_file = os.path.exists(curand_kernel_h) and os.path.isfile(
         curand_kernel_h
     )
-    reason = "curand_kernel.h not available on this system"
     return unittest.skipUnless(curand_kernel_h_file, reason)(fn)
 
 

numba_cuda/numba/cuda/tests/cudadrv/test_cuda_ndarray.py

@@ -476,12 +476,14 @@ class TestArrayMethod(CUDATestCase):
             host_array, dev_array.copy_to_host().astype(dtype)
         )
 
+    @skip_on_cudasim("Simulator does not use __array__()")
     @unittest.skipUnless(IS_NUMPY_2, "NumPy 1.x does not pass copy kwarg")
     def test_np_array_copy_false(self):
         dev_array = cuda.to_device(np.asarray([1.0, 2.0, 3.0]))
         with self.assertRaisesRegex(ValueError, "`copy=False` is not"):
             np.array(dev_array, copy=False)
 
+    @skip_on_cudasim("Simulator does not use __array__()")
     @unittest.skipUnless(IS_NUMPY_2, "NumPy 1.x does not pass copy kwarg")
     def test_np_array_copy_true(self):
         dev_array = cuda.to_device(np.asarray([1.0, 2.0, 3.0]))

numba_cuda/numba/cuda/tests/cudadrv/test_module_callbacks.py

@@ -5,14 +5,19 @@ import numpy as np
 
 from numba import cuda, config
 from numba.cuda.cudadrv.linkable_code import CUSource
-from numba.cuda.testing import CUDATestCase, ContextResettingTestCase
+from numba.cuda.testing import (
+    CUDATestCase,
+    ContextResettingTestCase,
+    skip_on_cudasim,
+)
 
-from cuda.bindings.driver import cuModuleGetGlobal, cuMemcpyHtoD
+if not config.ENABLE_CUDASIM:
+    from cuda.bindings.driver import cuModuleGetGlobal, cuMemcpyHtoD
 
-if config.CUDA_USE_NVIDIA_BINDING:
-    from cuda.cuda import CUmodule as cu_module_type
-else:
-    from numba.cuda.cudadrv.drvapi import cu_module as cu_module_type
+    if config.CUDA_USE_NVIDIA_BINDING:
+        from cuda.cuda import CUmodule as cu_module_type
+    else:
+        from numba.cuda.cudadrv.drvapi import cu_module as cu_module_type
 
 
 def wipe_all_modules_in_context():
@@ -32,6 +37,7 @@ def get_hashable_handle_value(handle):
     return handle
 
 
+@skip_on_cudasim("Module loading not implemented in the simulator")
 class TestModuleCallbacksBasic(ContextResettingTestCase):
     def test_basic(self):
         counter = 0
@@ -136,6 +142,7 @@ class TestModuleCallbacksBasic(ContextResettingTestCase):
         self.assertEqual(len(teardown_seen), 2)
 
 
+@skip_on_cudasim("Module loading not implemented in the simulator")
 class TestModuleCallbacksAPICompleteness(CUDATestCase):
     def test_api(self):
         def setup(handle):
@@ -164,6 +171,7 @@ class TestModuleCallbacksAPICompleteness(CUDATestCase):
                 kernel[1, 1]()
 
 
+@skip_on_cudasim("Module loading not implemented in the simulator")
 class TestModuleCallbacks(CUDATestCase):
     def setUp(self):
         super().setUp()
@@ -213,6 +221,7 @@ __device__ int get_num(int &retval) {
         self.assertEqual(arr[0], 42)
 
 
+@skip_on_cudasim("Module loading not implemented in the simulator")
 class TestMultithreadedCallbacks(CUDATestCase):
     def test_concurrent_initialization(self):
         seen_mods = set()

numba_cuda/numba/cuda/tests/cudadrv/test_nvjitlink.py

@@ -267,6 +267,7 @@ class TestLinker(CUDATestCase):
     not PYNVJITLINK_INSTALLED or not TEST_BIN_DIR,
     reason="pynvjitlink not enabled",
 )
+@skip_on_cudasim("Linking unsupported in the simulator")
 class TestLinkerUsage(CUDATestCase):
     """Test that whether pynvjitlink can be enabled by both environment variable
     and modification of config at runtime.

numba_cuda/numba/cuda/tests/cudapy/test_array.py

@@ -310,9 +310,6 @@ class TestCudaArray(CUDATestCase):
             check(array_reshape, array_reshape1d, arr, 0)
             check(array_reshape, array_reshape1d, arr, (0,))
             check(array_reshape, array_reshape3d, arr, (1, 0, 2))
-            check_only_shape(array_reshape2d, arr, (0, -1), (0, 0))
-            check_only_shape(array_reshape2d, arr, (4, -1), (4, 0))
-            check_only_shape(array_reshape3d, arr, (-1, 0, 4), (0, 0, 4))
 
         # C-contiguous
         arr = np.arange(24)

numba_cuda/numba/cuda/tests/cudapy/test_array_alignment.py

@@ -3,7 +3,11 @@ import itertools
 import numpy as np
 from numba import cuda
 from numba.core.errors import TypingError
-from numba.cuda.testing import CUDATestCase
+from numba.cuda.testing import (
+    CUDATestCase,
+    skip_on_cudasim,
+    skip_unless_cudasim,
+)
 import unittest
 
 
@@ -65,6 +69,7 @@ for align in (True, False):
 #      with the test_alignment.TestArrayAlignment class.
 
 
+@skip_on_cudasim("Array alignment not supported on cudasim")
 class TestArrayAddressAlignment(CUDATestCase):
     """
     Test cuda.local.array and cuda.shared.array support for an alignment
@@ -232,5 +237,24 @@ class TestArrayAddressAlignment(CUDATestCase):
                     print(".", end="", flush=True)
 
 
+@skip_unless_cudasim("Only check for alignment unsupported in the simulator")
+class TestCudasimUnsupportedAlignment(CUDATestCase):
+    def test_local_unsupported(self):
+        @cuda.jit
+        def f():
+            cuda.local.array(1, dtype=np.uint8, alignment=16)
+
+        with self.assertRaisesRegex(RuntimeError, "not supported in cudasim"):
+            f[1, 1]()
+
+    def test_shared_unsupported(self):
+        @cuda.jit
+        def f():
+            cuda.shared.array(1, dtype=np.uint8, alignment=16)
+
+        with self.assertRaisesRegex(RuntimeError, "not supported in cudasim"):
+            f[1, 1]()
+
+
 if __name__ == "__main__":
     unittest.main()

numba_cuda/numba/cuda/tests/cudapy/test_bfloat16.py

@@ -23,6 +23,15 @@ class TestBfloat16HighLevelBindings(CUDATestCase):
 
     def test_math_bindings(self):
         self.skip_unsupported()
+
+        exp_functions = [math.exp]
+        try:
+            from math import exp2
+
+            exp_functions += [exp2]
+        except ImportError:
+            pass
+
         functions = [
             math.trunc,
             math.ceil,
@@ -33,9 +42,7 @@ class TestBfloat16HighLevelBindings(CUDATestCase):
             math.cos,
             math.sin,
             math.tanh,
-            math.exp,
-            math.exp2,
-        ]
+        ] + exp_functions
 
         for f in functions:
             with self.subTest(func=f):
@@ -49,7 +56,7 @@ class TestBfloat16HighLevelBindings(CUDATestCase):
                 arr = cuda.device_array((1,), dtype="float32")
                 kernel[1, 1](arr)
 
-                if f in (math.exp, math.exp2):
+                if f in exp_functions:
                     self.assertAlmostEqual(arr[0], f(3.14), delta=1e-1)
                 else:
                     self.assertAlmostEqual(arr[0], f(3.14), delta=1e-2)

numba_cuda/numba/cuda/tests/cudapy/test_bfloat16_bindings.py

@@ -2,29 +2,40 @@ import numba.cuda as cuda
 from numba.cuda.testing import unittest, CUDATestCase
 import numpy as np
 
-from numba import int16, int32, int64, uint16, uint32, uint64, float32, float64
+from numba import (
+    config,
+    int16,
+    int32,
+    int64,
+    uint16,
+    uint32,
+    uint64,
+    float32,
+    float64,
+)
 from numba.types import float16
 
-from numba.cuda._internal.cuda_bf16 import (
-    nv_bfloat16,
-    htrunc,
-    hceil,
-    hfloor,
-    hrint,
-    hsqrt,
-    hrsqrt,
-    hrcp,
-    hlog,
-    hlog2,
-    hlog10,
-    hcos,
-    hsin,
-    hexp,
-    hexp2,
-    hexp10,
-    htanh,
-    htanh_approx,
-)
+if not config.ENABLE_CUDASIM:
+    from numba.cuda._internal.cuda_bf16 import (
+        nv_bfloat16,
+        htrunc,
+        hceil,
+        hfloor,
+        hrint,
+        hsqrt,
+        hrsqrt,
+        hrcp,
+        hlog,
+        hlog2,
+        hlog10,
+        hcos,
+        hsin,
+        hexp,
+        hexp2,
+        hexp10,
+        htanh,
+        htanh_approx,
+    )
 
 dtypes = [int16, int32, int64, uint16, uint32, uint64, float32]
 
@@ -263,6 +274,8 @@ class Bfloat16Test(CUDATestCase):
         np.testing.assert_allclose(arr, [8], atol=1e-2)
 
     def test_use_binding_inside_dfunc(self):
+        self.skip_unsupported()
+
         @cuda.jit(device=True)
         def f(arr):
             pi = nv_bfloat16(3.14)

numba_cuda/numba/cuda/tests/cudapy/test_cooperative_groups.py

@@ -157,6 +157,7 @@ class TestCudaCooperativeGroups(CUDATestCase):
         self.assertEqual(blocks1d, blocks2d)
         self.assertEqual(blocks1d, blocks3d)
 
+    @skip_on_cudasim("External code unsupported on cudasim")
     @skip_unless_cc_60
     def test_external_cooperative_func(self):
         cudapy_test_path = os.path.dirname(__file__)
@@ -171,12 +172,13 @@ class TestCudaCooperativeGroups(CUDATestCase):
             "cta_barrier", sig=sig, link=[src], use_cooperative=True
         )
 
-        @cuda.jit
+        @cuda.jit("void()")
         def kernel():
             cta_barrier()
 
+        overload = kernel.overloads[()]
         block_size = 32
-        grid_size = 1024
+        grid_size = overload.max_cooperative_grid_blocks(block_size)
 
         kernel[grid_size, block_size]()
 

numba_cuda/numba/cuda/tests/cudapy/test_enums.py

@@ -116,6 +116,7 @@ class EnumTest(CUDATestCase):
         got = cuda_func(arr)
         self.assertPreciseEqual(expected, got)
 
+    @skip_on_cudasim("No typing context in CUDA simulator")
     def test_int_enum_no_conversion(self):
         # Ported from Numba PR #10047: "Fix IntEnumMember.can_convert_to() when
         # no conversions found", https://github.com/numba/numba/pull/10047.

numba_cuda/numba/cuda/tests/cudapy/test_extending.py

@@ -191,7 +191,9 @@ if TEST_BIN_DIR:
     )
 
 
+@skip_on_cudasim("Extensions not supported in the simulator")
 class TestExtendingLinkage(CUDATestCase):
+    @unittest.skipUnless(TEST_BIN_DIR, "Necessary binaries are not available")
     def test_extension_adds_linkable_code(self):
         cuda_major_version = cuda.runtime.get_version()[0]
 

numba_cuda/numba/cuda/tests/nrt/test_nrt.py

@@ -3,16 +3,13 @@ import os
 
 import numpy as np
 import unittest
-from numba.cuda.testing import CUDATestCase
+from numba.cuda.testing import CUDATestCase, skip_on_cudasim
 from numba.tests.support import run_in_subprocess, override_config
 from numba.cuda import get_current_device
 from numba.cuda.cudadrv.nvrtc import compile
-from numba import types
-from numba.cuda.cudadecl import registry as cuda_decl_registry
+from numba import config, types
 from numba.core.typing import signature
-from numba.cuda.cudaimpl import lower as cuda_lower
 from numba import cuda
-from numba.cuda.runtime.nrt import rtsys, get_include
 from numba.core.typing.templates import AbstractTemplate
 from numba.cuda.cudadrv.linkable_code import (
     CUSource,
@@ -23,67 +20,68 @@ from numba.cuda.cudadrv.linkable_code import (
     Object,
 )
 
-
 TEST_BIN_DIR = os.getenv("NUMBA_CUDA_TEST_BIN_DIR")
 
-if TEST_BIN_DIR:
-
-    def make_linkable_code(name, kind, mode):
-        path = os.path.join(TEST_BIN_DIR, name)
-        with open(path, mode) as f:
-            contents = f.read()
-        return kind(contents, nrt=True)
-
-    nrt_extern_a = make_linkable_code("nrt_extern.a", Archive, "rb")
-    nrt_extern_cubin = make_linkable_code("nrt_extern.cubin", Cubin, "rb")
-    nrt_extern_cu = make_linkable_code(
-        "nrt_extern.cu",
-        CUSource,
-        "rb",
-    )
-    nrt_extern_fatbin = make_linkable_code("nrt_extern.fatbin", Fatbin, "rb")
-    nrt_extern_fatbin_multi = make_linkable_code(
-        "nrt_extern_multi.fatbin", Fatbin, "rb"
-    )
-    nrt_extern_o = make_linkable_code("nrt_extern.o", Object, "rb")
-    nrt_extern_ptx = make_linkable_code("nrt_extern.ptx", PTXSource, "rb")
+if not config.ENABLE_CUDASIM:
+    from numba.cuda.memory_management.nrt import rtsys, get_include
+    from numba.cuda.cudadecl import registry as cuda_decl_registry
+    from numba.cuda.cudaimpl import lower as cuda_lower
 
+    def allocate_deallocate_handle():
+        """
+        Handle to call NRT_Allocate and NRT_Free
+        """
+        pass
 
-def allocate_deallocate_handle():
-    """
-    Handle to call NRT_Allocate and NRT_Free
-    """
-    pass
-
-
-@cuda_decl_registry.register_global(allocate_deallocate_handle)
-class AllocateShimImpl(AbstractTemplate):
-    def generic(self, args, kws):
-        return signature(types.void)
-
+    @cuda_decl_registry.register_global(allocate_deallocate_handle)
+    class AllocateShimImpl(AbstractTemplate):
+        def generic(self, args, kws):
+            return signature(types.void)
 
-device_fun_shim = cuda.declare_device(
-    "device_allocate_deallocate", types.int32()
-)
+    device_fun_shim = cuda.declare_device(
+        "device_allocate_deallocate", types.int32()
+    )
 
+    # wrapper to turn the above into a python callable
+    def call_device_fun_shim():
+        return device_fun_shim()
+
+    @cuda_lower(allocate_deallocate_handle)
+    def allocate_deallocate_impl(context, builder, sig, args):
+        sig_ = types.int32()
+        # call the external function, passing the pointer
+        result = context.compile_internal(
+            builder,
+            call_device_fun_shim,
+            sig_,
+            (),
+        )
 
-# wrapper to turn the above into a python callable
-def call_device_fun_shim():
-    return device_fun_shim()
+        return result
 
+    if TEST_BIN_DIR:
 
-@cuda_lower(allocate_deallocate_handle)
-def allocate_deallocate_impl(context, builder, sig, args):
-    sig_ = types.int32()
-    # call the external function, passing the pointer
-    result = context.compile_internal(
-        builder,
-        call_device_fun_shim,
-        sig_,
-        (),
-    )
+        def make_linkable_code(name, kind, mode):
+            path = os.path.join(TEST_BIN_DIR, name)
+            with open(path, mode) as f:
+                contents = f.read()
+            return kind(contents, nrt=True)
 
-    return result
+        nrt_extern_a = make_linkable_code("nrt_extern.a", Archive, "rb")
+        nrt_extern_cubin = make_linkable_code("nrt_extern.cubin", Cubin, "rb")
+        nrt_extern_cu = make_linkable_code(
+            "nrt_extern.cu",
+            CUSource,
+            "rb",
+        )
+        nrt_extern_fatbin = make_linkable_code(
+            "nrt_extern.fatbin", Fatbin, "rb"
+        )
+        nrt_extern_fatbin_multi = make_linkable_code(
+            "nrt_extern_multi.fatbin", Fatbin, "rb"
+        )
+        nrt_extern_o = make_linkable_code("nrt_extern.o", Object, "rb")
+        nrt_extern_ptx = make_linkable_code("nrt_extern.ptx", PTXSource, "rb")
 
 
 class TestNrtBasic(CUDATestCase):
@@ -104,6 +102,7 @@ class TestNrtBasic(CUDATestCase):
         g[1, 1]()
         cuda.synchronize()
 
+    @skip_on_cudasim("CUDA Simulator does not produce PTX")
     def test_nrt_ptx_contains_refcount(self):
         @cuda.jit
         def f(x):
@@ -157,6 +156,7 @@ class TestNrtLinking(CUDATestCase):
         with override_config("CUDA_ENABLE_NRT", True):
             super(TestNrtLinking, self).run(result)
 
+    @skip_on_cudasim("CUDA Simulator does not link PTX")
     def test_nrt_detect_linked_ptx_file(self):
         src = f"#include <{get_include()}/nrt.cuh>"
         src += """
@@ -176,6 +176,7 @@ class TestNrtLinking(CUDATestCase):
         kernel[1, 1]()
 
     @unittest.skipIf(not TEST_BIN_DIR, "necessary binaries not generated.")
+    @skip_on_cudasim("CUDA Simulator does not link code")
     def test_nrt_detect_linkable_code(self):
         codes = (
             nrt_extern_a,
@@ -196,6 +197,7 @@ class TestNrtLinking(CUDATestCase):
                 kernel[1, 1]()
 
 
+@skip_on_cudasim("CUDASIM does not have NRT statistics")
 class TestNrtStatistics(CUDATestCase):
     def setUp(self):
         self._stream = cuda.default_stream()
@@ -213,7 +215,7 @@ class TestNrtStatistics(CUDATestCase):
         # Checks that explicitly turning the stats on via the env var works.
         src = """if 1:
         from numba import cuda
-        from numba.cuda.runtime import rtsys
+        from numba.cuda.memory_management import rtsys
         import numpy as np
 
         @cuda.jit
@@ -252,7 +254,7 @@ class TestNrtStatistics(CUDATestCase):
         src = """if 1:
         from numba import cuda
         import numpy as np
-        from numba.cuda.runtime import rtsys
+        from numba.cuda.memory_management import rtsys
 
         @cuda.jit
         def foo():

numba_cuda/numba/cuda/tests/nrt/test_nrt_refct.py

@@ -1,13 +1,14 @@
 import numpy as np
 import unittest
 from numba.tests.support import override_config
-from numba.cuda.runtime import rtsys
+from numba.cuda.memory_management import rtsys
 from numba.cuda.tests.support import EnableNRTStatsMixin
-from numba.cuda.testing import CUDATestCase
+from numba.cuda.testing import CUDATestCase, skip_on_cudasim
 
 from numba import cuda
 
 
+@skip_on_cudasim("No refcounting in the simulator")
 class TestNrtRefCt(EnableNRTStatsMixin, CUDATestCase):
     def setUp(self):
         super(TestNrtRefCt, self).setUp()

numba_cuda/numba/cuda/tests/support.py

@@ -1,4 +1,4 @@
-from numba.cuda.runtime.nrt import rtsys
+from numba.cuda.memory_management.nrt import rtsys
 
 
 class EnableNRTStatsMixin(object):

numba_cuda/numba/cuda/tests/test_binary_generation/Makefile

@@ -40,7 +40,7 @@ LTOIR_FLAGS := $(LTOIR_GENCODE) -dc
 
 OUTPUT_DIR := ./
 
-NRT_INCLUDE_DIR := $(shell python -c "from numba.cuda.runtime.nrt import get_include; print(get_include())")
+NRT_INCLUDE_DIR := $(shell python -c "from numba.cuda.memory_management.nrt import get_include; print(get_include())")
 
 all:
 	@echo "GPU CC: $(GPU_CC)"

numba_cuda/numba/cuda/tests/test_binary_generation/generate_raw_ltoir.py

@@ -7,7 +7,7 @@ import subprocess
 import sys
 
 from cuda import nvrtc
-from numba.cuda.runtime.nrt import get_include
+from numba.cuda.memory_management.nrt import get_include
 
 # Magic number found at the start of an LTO-IR file
 LTOIR_MAGIC = 0x7F4E43ED

{numba_cuda-0.12.1.dist-info → numba_cuda-0.13.0.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: numba-cuda
-Version: 0.12.1
+Version: 0.13.0
 Summary: CUDA target for Numba
 Author: Anaconda Inc., NVIDIA Corporation
 License: BSD 2-clause

{numba_cuda-0.12.1.dist-info → numba_cuda-0.13.0.dist-info}/RECORD

@@ -1,6 +1,6 @@
 _numba_cuda_redirector.pth,sha256=cmfMMmV0JPh3yEpl4bGeM9AuXiVVMSo6Z_b7RaQL3XE,30
 _numba_cuda_redirector.py,sha256=n_r8MYbu5-vcXMnLJW147k8DnFXXvgb7nPIXnlXwTyQ,2659
-numba_cuda/VERSION,sha256=9u5pvxxLJ6JCJmzLWutKqMgwY0W56-T_czW4yUBFK4E,7
+numba_cuda/VERSION,sha256=2EyeWWx9apTl90V5742JEqgHsNKFgkdJAK0137Pt_PQ,7
 numba_cuda/__init__.py,sha256=atXeUvJKR3JHcAiCFbXCVOJQUHgB1TulmsqSL_9RT3Q,114
 numba_cuda/_version.py,sha256=nzrrJXi85d18m6SPdsPsetJNClDETkmF1MrEhGLYDBs,734
 numba_cuda/numba/cuda/__init__.py,sha256=3siqMXEKqa9ezQ8RxPC3KMdebUjgJt-EKxxV4CX9818,607
@@ -9,7 +9,7 @@ numba_cuda/numba/cuda/api_util.py,sha256=jK8oUD3zf_D5IX7vbjc3uY_5kmOxwgEqO2m_lDH
 numba_cuda/numba/cuda/args.py,sha256=UlTHTJpwPeCtnW0Bb-Wetm5UO9TPR-PCgIt5ys8b8tQ,1894
 numba_cuda/numba/cuda/bf16.py,sha256=PXuitxHhPMjnti3g9IOSoL90ofGgVRcDfqFg7AqCXpU,1778
 numba_cuda/numba/cuda/cg.py,sha256=n-sBj05ut6U_GgFIq-PTCjPad4nXWAc0GVg_J9xD_Pc,1602
-numba_cuda/numba/cuda/codegen.py,sha256=vZtLahHSLYzRpQ3GSbmMm5qYp4FS5mAlzGgSgJbaoz0,16709
+numba_cuda/numba/cuda/codegen.py,sha256=u2J0mRRDBiPceB1G5WR4KQ0KUFGGawaDaaoUf9zLQzE,16719
 numba_cuda/numba/cuda/compiler.py,sha256=aZwEVP8KXCIyccSw4vJyG6Qaai9oXsFuBAo_Ghwwai4,25607
 numba_cuda/numba/cuda/cpp_function_wrappers.cu,sha256=8lUPmU6FURxphzEqkPLZRPYBCEK_wmDtHq2voPkckfs,950
 numba_cuda/numba/cuda/cuda_paths.py,sha256=kMIJ_1yV2qtcKEM5rCgSDJ3Gz7bgxbfAWh54E5cDndg,15872
@@ -21,7 +21,7 @@ numba_cuda/numba/cuda/decorators.py,sha256=bR8yOAIC68lhm8mSMU-DUt1qFrEogbmSAtzAI
 numba_cuda/numba/cuda/descriptor.py,sha256=t1rSVJSCAlVACC5_Un3FQ7iubdTTBe-euqz88cvs2tI,985
 numba_cuda/numba/cuda/device_init.py,sha256=Rtwd6hQMHMLMkj6MXtndbWYFJfkIaRe0MwOIJF2nzhU,3449
 numba_cuda/numba/cuda/deviceufunc.py,sha256=zj9BbLiZD-dPttHew4olw8ANgR2nXnXEE9qjCeGLrQI,30731
-numba_cuda/numba/cuda/dispatcher.py,sha256=cLXD2pnsU7k-bN5clfjuWqifFCr7LfECKtK7YeeHwis,43162
+numba_cuda/numba/cuda/dispatcher.py,sha256=m8kXKk08ldcW7Cl3KpFxsKMTxVgZeRJke9bKzO6_JjE,43172
 numba_cuda/numba/cuda/errors.py,sha256=WRso1Q_jCoWP5yrDBMhihRhhVtVo1-7KdN8QVE9j46o,1712
 numba_cuda/numba/cuda/extending.py,sha256=VwuU5F0AQFlJsqaiwoWk-6Itihew1FsjVT_BVjhY8Us,2278
 numba_cuda/numba/cuda/initialize.py,sha256=0SnpjccQEYiWITIyfAJx833H1yhYFFDY42EpnwYyMn8,487
@@ -42,7 +42,7 @@ numba_cuda/numba/cuda/reshape_funcs.cu,sha256=frw1uoeMSYlkPC38LiKE8Tz2P70X2e4UZG
 numba_cuda/numba/cuda/simulator_init.py,sha256=Hvzty6NJp1SeKspyb-b887xpeNLMMI0x9aPmV--X77E,450
 numba_cuda/numba/cuda/stubs.py,sha256=JMs4Xg8IHlAq5L6SBYWcYNzXfJGM6v0lZCQaOb5x9CQ,23014
 numba_cuda/numba/cuda/target.py,sha256=ymYBdkt7iNK_PJCfyqupKpcSj7j-UQzkWIq3KjoLBD8,12963
-numba_cuda/numba/cuda/testing.py,sha256=OR37AuDdzg7vLG4G_4s2uRAkNTScZc-BzHmTMJYuxhQ,6827
+numba_cuda/numba/cuda/testing.py,sha256=a5Ay9z7x9vPiFfbfuwIfb3uyA_OSAchQZtye4u71UKE,6995
 numba_cuda/numba/cuda/types.py,sha256=hC1MUvgUwy-SLgbzFzXwssJzPR8BxQwqUcjwGJFzVac,1317
 numba_cuda/numba/cuda/ufuncs.py,sha256=AJifQgapyv62fdJeMm939R1I5TvIRmaA8dJ83Jy8DCw,23559
 numba_cuda/numba/cuda/utils.py,sha256=VRphC0PLr8Klq3D1FMONu4aRdVO23HOCBg4bxnsqmfc,785
@@ -61,7 +61,7 @@ numba_cuda/numba/cuda/cudadrv/libs.py,sha256=qjknQxYXd2ucwDLQqzhWC_srNg6FnwvcVHI
 numba_cuda/numba/cuda/cudadrv/linkable_code.py,sha256=IZ13laEG_altDQyi9HkdMcwW-YYEIn2erqz6AnYsqHg,2808
 numba_cuda/numba/cuda/cudadrv/mappings.py,sha256=9uEs1KepeVGRbEpVhLjtxSsvZpZsbrHnPywmx--y88A,804
 numba_cuda/numba/cuda/cudadrv/ndarray.py,sha256=HtULWWFyDlgqvrH5459yyPTvU4UbUo2DSdtcNfvbH00,473
-numba_cuda/numba/cuda/cudadrv/nvrtc.py,sha256=pDc5YsxOMdMbLnUKm1st2FVmFPRU-Mhlpd9mau9KZ-0,14976
+numba_cuda/numba/cuda/cudadrv/nvrtc.py,sha256=UD8kASyGUU896tNWAtVxmbzDTP5jDbiOAZjCsELOg6U,14986
 numba_cuda/numba/cuda/cudadrv/nvvm.py,sha256=7tTy6-VEbMBpDUmuSMnUwqPFfBndTh3aPq_n7nxhEA0,26344
 numba_cuda/numba/cuda/cudadrv/rtapi.py,sha256=J6PRGGK07XSLRzgCw5xs8VU5xVoqavvhojk1mxiQsi4,226
 numba_cuda/numba/cuda/cudadrv/runtime.py,sha256=CFumwg4iblWap_E7l7GM_hMYz1PsbH81-N0tZwFFooA,4372
@@ -76,31 +76,38 @@ numba_cuda/numba/cuda/include/12/cuda_fp16.hpp,sha256=o1ITDmuN67N8YUGUcvTpV3IdpS
 numba_cuda/numba/cuda/kernels/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 numba_cuda/numba/cuda/kernels/reduction.py,sha256=RsVubg8uNumxNxo9HBlFVCDicA-KZKsksKId0ktgQyY,9101
 numba_cuda/numba/cuda/kernels/transpose.py,sha256=FbtFmOqaj_e7ARR_kkiTpSvj4BJyqBta5ci1CWtJ690,2033
-numba_cuda/numba/cuda/runtime/__init__.py,sha256=H-KOPDk6wMO_ADWzlZbA9U9hroX79WjenRcNpCSICiM,55
-numba_cuda/numba/cuda/runtime/memsys.cu,sha256=gMBM9_Hnv3EO3Gw_GKvII8y2hGoNtwrlZ43AUjTcsVo,2387
-numba_cuda/numba/cuda/runtime/memsys.cuh,sha256=hPGBQgKyOfYY25ntoBXlhYyeXzxJyz0ByeTszkaKJUM,504
-numba_cuda/numba/cuda/runtime/nrt.cu,sha256=1hzbAKyqh9783UVdVT67ZxfvJyl_Ojt8e0AbHUC86ss,4818
-numba_cuda/numba/cuda/runtime/nrt.cuh,sha256=p2GQ-l-EfCoO0sBTyKXhIY3hxGWbPhEJcR-mLLT_V3M,2173
-numba_cuda/numba/cuda/runtime/nrt.py,sha256=6yXKBUvjIw_9BJ48iDIuckREaQVskzQAXm7uIRGFVuc,10039
-numba_cuda/numba/cuda/simulator/__init__.py,sha256=OByuGEDv4JDPx5LtO_NFApGIlaOEuhhtjrST2nthqeI,1580
-numba_cuda/numba/cuda/simulator/api.py,sha256=bzvwedFWG-EZWp6iWhRdqQisHu-Ik1TkyqW6Q5pGyrM,3210
-numba_cuda/numba/cuda/simulator/compiler.py,sha256=MnGuww-h5uzcNuLsXfkpBRSDNhy-8PlHAWwIvtt3PI4,232
-numba_cuda/numba/cuda/simulator/kernel.py,sha256=DtQz7cEpqH43M3r1Ic3xNmjT_1W6-96BpEtre6ephWU,10429
-numba_cuda/numba/cuda/simulator/kernelapi.py,sha256=Lp-x6hTd6ByAkQKv9bHEbVxUUgUba2NdBE25zAXDS_g,12404
+numba_cuda/numba/cuda/memory_management/__init__.py,sha256=r4RWOgWft81ChDdRaJdABHF0y_HxrTIv4mgXpo0imYA,65
+numba_cuda/numba/cuda/memory_management/memsys.cu,sha256=gMBM9_Hnv3EO3Gw_GKvII8y2hGoNtwrlZ43AUjTcsVo,2387
+numba_cuda/numba/cuda/memory_management/memsys.cuh,sha256=hPGBQgKyOfYY25ntoBXlhYyeXzxJyz0ByeTszkaKJUM,504
+numba_cuda/numba/cuda/memory_management/nrt.cu,sha256=1hzbAKyqh9783UVdVT67ZxfvJyl_Ojt8e0AbHUC86ss,4818
+numba_cuda/numba/cuda/memory_management/nrt.cuh,sha256=p2GQ-l-EfCoO0sBTyKXhIY3hxGWbPhEJcR-mLLT_V3M,2173
+numba_cuda/numba/cuda/memory_management/nrt.py,sha256=6yXKBUvjIw_9BJ48iDIuckREaQVskzQAXm7uIRGFVuc,10039
+numba_cuda/numba/cuda/simulator/__init__.py,sha256=ONoWJ3SwE53di0p-lFRH7NOZea2jEUWyn9sDpkOVjCw,2040
+numba_cuda/numba/cuda/simulator/api.py,sha256=hFSFPIrg-aUd-MHg2GCSosFJiL8x2XRiQaqolfTGA3A,3551
+numba_cuda/numba/cuda/simulator/bf16.py,sha256=1ZWkY4Adv8dY46YyorGKGQj3KEBqeet6rsyM8jwfAb4,16
+numba_cuda/numba/cuda/simulator/compiler.py,sha256=bK3qZhZ5ZszjGEtR-a5q-0wqKk_mjZ8Z5yXFsRVu4kE,263
+numba_cuda/numba/cuda/simulator/kernel.py,sha256=9SWZwD0kLQ3phRUIxZJdzxKiOgLQ6FCbyxVGC52niwM,10448
+numba_cuda/numba/cuda/simulator/kernelapi.py,sha256=wAKQih_0KRVgtMgOW2SpfmKQhZxU-D2-9D6SZc78CXM,12662
 numba_cuda/numba/cuda/simulator/reduction.py,sha256=_l5xzoKfoDoaSWW5uFOkXz0YKA6XBqL84wUFNLLNCgM,301
 numba_cuda/numba/cuda/simulator/vector_types.py,sha256=bN347hH6w2khtYD42d_VT8n1lAmWCee1QUxtGvpdzBg,1692
+numba_cuda/numba/cuda/simulator/_internal/__init__.py,sha256=arjWguQQhHaD0pAQe9XaC8gI6eye0lboHAxd6TgCXHQ,67
+numba_cuda/numba/cuda/simulator/_internal/cuda_bf16.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 numba_cuda/numba/cuda/simulator/cudadrv/__init__.py,sha256=P8VK4Q8w7H_-sagOYWNLeYjWsmoI6_KcCAceBmSyHx0,120
 numba_cuda/numba/cuda/simulator/cudadrv/devicearray.py,sha256=On6Vx4WQC79yWr8ayUOts00sOADsCITVxLMnuiG-4LI,13503
 numba_cuda/numba/cuda/simulator/cudadrv/devices.py,sha256=7kq0XuzlgFeZPlYiFl1smEEmR0BCigalkt-CODsfPCo,2691
-numba_cuda/numba/cuda/simulator/cudadrv/driver.py,sha256=0S3NT64iIZMrSxxZK7ppecD0orYWkIJgtZAyMAHLgFk,1124
+numba_cuda/numba/cuda/simulator/cudadrv/driver.py,sha256=-o16cYfkuta7BgltIBcprvQL81yQwlzZ4xpy9DD6V9Q,1242
 numba_cuda/numba/cuda/simulator/cudadrv/drvapi.py,sha256=MfYr-UGYhMaysqxsEstyTvmV5Gee6mFZb3PCvcLywos,111
 numba_cuda/numba/cuda/simulator/cudadrv/dummyarray.py,sha256=DYIpIehz3cZSane651UYdJP5fehDuJkxyCg_90A1heU,163
 numba_cuda/numba/cuda/simulator/cudadrv/error.py,sha256=ACSQ7ZvhuCHnvV4GmvRuKWZ5bBLVzq7ncZ75oiWyLdM,87
-numba_cuda/numba/cuda/simulator/cudadrv/libs.py,sha256=P_59f2bM-KbXJQ_ccnLrxLidCNFWJuAoyDRPIiEWnTQ,101
+numba_cuda/numba/cuda/simulator/cudadrv/libs.py,sha256=pmAZxP40Md1B8sEBlkQ0ckYraSiuxH7tt3TmjvXDmr0,201
+numba_cuda/numba/cuda/simulator/cudadrv/linkable_code.py,sha256=u8I2keYsmAcWaSvQhTLiBxX_O4qJhkMrK15esx0m1PM,1192
+numba_cuda/numba/cuda/simulator/cudadrv/nvrtc.py,sha256=OSE_gPUqqxdIMs01yacVT-eGwmirQAY4DIve15wODRI,216
 numba_cuda/numba/cuda/simulator/cudadrv/nvvm.py,sha256=AFJg67iVcXROuj2ncHcmpkBsooEbEK8CQVOvxA_RUrc,474
 numba_cuda/numba/cuda/simulator/cudadrv/runtime.py,sha256=OtbGKMObt5eTPbVV8G5bKviT2sKiGLQ6z9crC1wfiCQ,360
+numba_cuda/numba/cuda/simulator/memory_management/__init__.py,sha256=-_jZfXlheXqZDSGW-sHYykVl0dDvljpEg4Pp_f0EZC4,37
+numba_cuda/numba/cuda/simulator/memory_management/nrt.py,sha256=1wcgPJtKUjqzP3KDDE1F_fwVR0S2emBeEPfHBdfnKkU,101
 numba_cuda/numba/cuda/tests/__init__.py,sha256=qrIMTeP56g1SJlyekm4vQ3fcsWGR9qGTnMfE_uq0eeQ,2387
-numba_cuda/numba/cuda/tests/support.py,sha256=1og4VLrK2x2LF5m5ARrrHVe-JhYx9Gv9ODKt6-8r6Aw,253
+numba_cuda/numba/cuda/tests/support.py,sha256=IpWXM2pELCeoqdQIUsvy9Rsm460omp15HMMpJsxTt9U,263
 numba_cuda/numba/cuda/tests/cudadrv/__init__.py,sha256=43EXdiXXRBd6yIcVGMrU9F_EJCD9Uw3mzOP3SB53AEE,260
 numba_cuda/numba/cuda/tests/cudadrv/test_array_attr.py,sha256=np2UccpkNqLFLcbjJNHGkCJF6TZIyFljB6UXD7U71nM,5299
 numba_cuda/numba/cuda/tests/cudadrv/test_context_stack.py,sha256=m7q1bEsH3rJD2gngB9RAjQuvjr0FB-GrOjLCWryI1lQ,4495
@@ -110,7 +117,7 @@ numba_cuda/numba/cuda/tests/cudadrv/test_cuda_devicerecord.py,sha256=JkMbKFa3CBS
 numba_cuda/numba/cuda/tests/cudadrv/test_cuda_driver.py,sha256=bn9OBNmNq5WTgv5LXQTyi-3V3auKbIBNoC-vNfzeX9I,7536
 numba_cuda/numba/cuda/tests/cudadrv/test_cuda_libraries.py,sha256=KWGON5OSb5Vp74QFDIiupK6ytMwwwDfbYqpENAB4lGE,801
 numba_cuda/numba/cuda/tests/cudadrv/test_cuda_memory.py,sha256=nN1pk7CEm4j8A6XYlDpIWkpFOSO7IGz-7rwa0fFnerY,6485
-numba_cuda/numba/cuda/tests/cudadrv/test_cuda_ndarray.py,sha256=YpeBvh3tPugd0qcWfCAR2Qk9Ac9A0t6mE03jbDS4uzU,21581
+numba_cuda/numba/cuda/tests/cudadrv/test_cuda_ndarray.py,sha256=MzKXO2RLCHA_0XU29JfjaLrmKTwwu7PA3cspTfQgCPM,21699
 numba_cuda/numba/cuda/tests/cudadrv/test_deallocations.py,sha256=VKYRuIOPdEWkI-6E6-pRCNC1U4-Qxi0d-jX_q_7x1dI,8420
 numba_cuda/numba/cuda/tests/cudadrv/test_detect.py,sha256=DUYZeNlDgL1mQN1xHDYzTRfc-zetuikcULyULcRaC1A,2657
 numba_cuda/numba/cuda/tests/cudadrv/test_emm_plugins.py,sha256=xom2V_KImeDX5AN9XKcWln_MFhIvPrwJDxJWvE6Nspw,7070
@@ -121,9 +128,9 @@ numba_cuda/numba/cuda/tests/cudadrv/test_inline_ptx.py,sha256=B_fYsBUpd9SxYSOmuW
 numba_cuda/numba/cuda/tests/cudadrv/test_is_fp16.py,sha256=0KPe4E9wOZsSV_0QI0LmjUeMTjWpYT8BXExUUsmUCDI,394
 numba_cuda/numba/cuda/tests/cudadrv/test_linker.py,sha256=ymv2ujRLLIIURikNEdC0SshJFwXhIx9j462va_QvPTw,10133
 numba_cuda/numba/cuda/tests/cudadrv/test_managed_alloc.py,sha256=2tkf766GjIta_wL5NGlMIqmrDMFN2rZmnP_c9A8cWA8,5084
-numba_cuda/numba/cuda/tests/cudadrv/test_module_callbacks.py,sha256=eR4868v9davyqdCel3Qia4fNBwKHGIMOD868uHm0GEM,8057
+numba_cuda/numba/cuda/tests/cudadrv/test_module_callbacks.py,sha256=176Ma2ZVLnc4w4bfYwbF1eeRq3x3rbOvDieRJLSuNpI,8413
 numba_cuda/numba/cuda/tests/cudadrv/test_mvc.py,sha256=9MLFEXn7DnLkuuXK_qjilA1jxQwC-AeSBOcRYzZogRY,1513
-numba_cuda/numba/cuda/tests/cudadrv/test_nvjitlink.py,sha256=9RkSpQ-8selsJtr7lrwx6brRbJOz6wxNY7xydl6G2uU,11494
+numba_cuda/numba/cuda/tests/cudadrv/test_nvjitlink.py,sha256=8SSSAotk8rhGClwxQCnwL_JhoD9NbvXxEa7KfjaZO3M,11551
 numba_cuda/numba/cuda/tests/cudadrv/test_nvvm_driver.py,sha256=1r817QeIrIEs8BcK0XKBR9g_mkO3e7WI-oW-sNsO_Ho,7353
 numba_cuda/numba/cuda/tests/cudadrv/test_pinned.py,sha256=PGuv4bt9qiIGlkLhyQCOXFIf1SK5Nj-RjcpWqeO1TMM,943
 numba_cuda/numba/cuda/tests/cudadrv/test_profiler.py,sha256=xbSFmvqOIcWY-TI9p1MDcGwE-24iaK4j-_UenMvTnR4,508
@@ -140,13 +147,13 @@ numba_cuda/numba/cuda/tests/cudapy/extensions_usecases.py,sha256=2i_xq4B1t1tctr6
 numba_cuda/numba/cuda/tests/cudapy/jitlink.ptx,sha256=PKVafUhDH1SKRWXkt4N3v8SDMh4RyDFiJM-CMksa5uc,519
 numba_cuda/numba/cuda/tests/cudapy/recursion_usecases.py,sha256=wrWx8AeRhBHM74iYPKKrZqiyWrYCtQU3J-g3Zv7JmoY,1782
 numba_cuda/numba/cuda/tests/cudapy/test_alignment.py,sha256=RkhAcVkGtze8JpZTlYYvqTesDYE7xfKQZd1izgxDQpU,1219
-numba_cuda/numba/cuda/tests/cudapy/test_array.py,sha256=lT7XWXl0_lqtXyyXN-w0cd0wH7EBklRElYDnHUM5G1I,13215
-numba_cuda/numba/cuda/tests/cudapy/test_array_alignment.py,sha256=JDKbbRieNE0C3wDsA83B5fTcRMu1OvSaBrs1uV0XtSM,8333
+numba_cuda/numba/cuda/tests/cudapy/test_array.py,sha256=tqALZEr39aM8ZzbnEuOpKTOWb6Diz6Eti1i0a6WyI7k,13005
+numba_cuda/numba/cuda/tests/cudapy/test_array_alignment.py,sha256=Bkz6z-GrSbwAjI9YQMcTKnFMNIQv5zyALYJZtaet9w0,9059
 numba_cuda/numba/cuda/tests/cudapy/test_array_args.py,sha256=iiFrt5Yn7gfheAGOYG2VBeWeuW3JlBhRLXNfSz4cHAA,4982
 numba_cuda/numba/cuda/tests/cudapy/test_array_methods.py,sha256=SWa1MvpwG07yBkrFIUeM9pm3BIwUbhttMNBdUW-CpSM,969
 numba_cuda/numba/cuda/tests/cudapy/test_atomics.py,sha256=agsfUN3WOoh6ICAECtuMuxZNcKq5ivK30Ew3h_m76m0,57689
-numba_cuda/numba/cuda/tests/cudapy/test_bfloat16.py,sha256=DdP7WlHev8R5DdY6DEEgOF45ljh8LwKeqmkvGLjNC7E,1444
-numba_cuda/numba/cuda/tests/cudapy/test_bfloat16_bindings.py,sha256=wNP0NNtqVgaekY9fXp_H4LpPNLX-rDu9gp-_-e965Lg,7420
+numba_cuda/numba/cuda/tests/cudapy/test_bfloat16.py,sha256=TpHRxyIwvwv6KUKTS15ukYVDTcT5iYsNp8fCcocY7M0,1573
+numba_cuda/numba/cuda/tests/cudapy/test_bfloat16_bindings.py,sha256=wvEbMdlmqAZvlZoUgFG-KC0w13CeMfMdR-gf2lN0Sj0,7612
 numba_cuda/numba/cuda/tests/cudapy/test_blackscholes.py,sha256=0_wr6MSeHh0QVzPeH8SB7j0Nv_RrPAK01hNoQ_dGT5I,4417
 numba_cuda/numba/cuda/tests/cudapy/test_boolean.py,sha256=j4mIOv4rJTLjJzpKk1O9UFLT41_iOQRtwsmteXdKZ-M,547
 numba_cuda/numba/cuda/tests/cudapy/test_caching.py,sha256=obUSTJSP2Lh-YNElq8PZpVnRJOeq-uqV_VyLHtsXwAw,18427
@@ -157,7 +164,7 @@ numba_cuda/numba/cuda/tests/cudapy/test_complex.py,sha256=hmAcyZim46yueXZDqDSJYq
 numba_cuda/numba/cuda/tests/cudapy/test_complex_kernel.py,sha256=KIuXQ0ihgQQXM-eH7s3xAxhKe35YL1qDTHCVTWA4ut8,497
 numba_cuda/numba/cuda/tests/cudapy/test_const_string.py,sha256=li1UsV5vc2M01cJ7k6_526VPtuAOAKr8e7kb1CDUXi4,4323
 numba_cuda/numba/cuda/tests/cudapy/test_constmem.py,sha256=ZWmyKvFokRMjqyXjVpZVOnR6LR694GWcbUn2jVEQV14,5170
-numba_cuda/numba/cuda/tests/cudapy/test_cooperative_groups.py,sha256=kkrK5Mo9E8nNH3PYfQAEel0hY7CXZNsn88BAo7heX9g,5818
+numba_cuda/numba/cuda/tests/cudapy/test_cooperative_groups.py,sha256=3OkjhcjPp_P3Pnc1zbteGpAGpoN07cG8Xtdnunx5yWA,5973
 numba_cuda/numba/cuda/tests/cudapy/test_cuda_array_interface.py,sha256=RXCNHAZM35sbUf3Gi-x2E8-a6BmhFb2rhQkBOeiS_fo,15757
 numba_cuda/numba/cuda/tests/cudapy/test_cuda_jit_no_types.py,sha256=8prL2FTiaajW-UHSL9al-nBniygOfpdAOT_Dkej4PWI,2138
 numba_cuda/numba/cuda/tests/cudapy/test_datetime.py,sha256=MnOeDWMz-rL3-07FsswM06Laxmm0KjTmTwhrP3rmchQ,3526
@@ -165,10 +172,10 @@ numba_cuda/numba/cuda/tests/cudapy/test_debug.py,sha256=1P369s02AvGu7fSIEe_YxSgh
 numba_cuda/numba/cuda/tests/cudapy/test_debuginfo.py,sha256=AE8D4U4dAv4nYP9oatDwROW6knpJ0-iggP4BaHymo6g,13170
 numba_cuda/numba/cuda/tests/cudapy/test_device_func.py,sha256=LNGBZfqFGUtVVQeC6FcHo8T3DbG-j6AjeBwJmwp9HH4,13157
 numba_cuda/numba/cuda/tests/cudapy/test_dispatcher.py,sha256=Oc6CdI1j9Ad_wklHdIYSMytrzUpzK6oXD0BGe45sTwg,26636
-numba_cuda/numba/cuda/tests/cudapy/test_enums.py,sha256=Yxac6S5P6C8GN0kMwieL3dQb1uogOVZQEx969B0AMpM,4533
+numba_cuda/numba/cuda/tests/cudapy/test_enums.py,sha256=VQGPLcTbT1nhS1BE4VALK-TaQEsPec5zu-XVlWV0sHA,4593
 numba_cuda/numba/cuda/tests/cudapy/test_errors.py,sha256=w6ipW9UIvUD_ZIt_6fQ-uJsHyKLyHVqv2bym-9vyGyY,2757
 numba_cuda/numba/cuda/tests/cudapy/test_exception.py,sha256=W5NF022DOOTaEjFmhfr8BnfhRXvYyXHiGwznQrm_9T4,5507
-numba_cuda/numba/cuda/tests/cudapy/test_extending.py,sha256=OEEuJ87D-ZcYtWrcOfcSsKPSDccgZVreUo61dxDv7DE,8537
+numba_cuda/numba/cuda/tests/cudapy/test_extending.py,sha256=G6KcFAiJnDEfa5f7HW72Ocqxrv6xRvGMRTbwttTsuec,8678
 numba_cuda/numba/cuda/tests/cudapy/test_fastmath.py,sha256=fiUoOiwWjctZNFN-DGw1A8eGfHLqNulo2OQ7v1DFS9o,8552
 numba_cuda/numba/cuda/tests/cudapy/test_forall.py,sha256=Ory5s-_9MauSCP2RuWUEmcGFvP0kS7ytV-3iYPFYR6o,1470
 numba_cuda/numba/cuda/tests/cudapy/test_freevar.py,sha256=JvWn7Lw137HI61mouKnPvDxZIqLppiCF_351osxQQYE,753
@@ -259,15 +266,15 @@ numba_cuda/numba/cuda/tests/nocuda/test_import.py,sha256=avrMV0jlve3KmDOrYtCeQ4r
 numba_cuda/numba/cuda/tests/nocuda/test_library_lookup.py,sha256=O_S_HG59Ak071b43BQ2s-xJDw9D8Iy_H1-CechHOZnc,7948
 numba_cuda/numba/cuda/tests/nocuda/test_nvvm.py,sha256=E_OdYlxgfRTFKONCMlgjvwvwHQkPS1ne1KTwzBDH9GE,1968
 numba_cuda/numba/cuda/tests/nrt/__init__.py,sha256=43EXdiXXRBd6yIcVGMrU9F_EJCD9Uw3mzOP3SB53AEE,260
-numba_cuda/numba/cuda/tests/nrt/test_nrt.py,sha256=12IJTEgY2gFIgKagYXNvLksQHIj7MrbRYlX9DU4SGoc,12194
-numba_cuda/numba/cuda/tests/nrt/test_nrt_refct.py,sha256=kbkPw8F7rTiY7RBbf-2Uiw19jRN0a9Y4YchpMep1HQ4,3133
-numba_cuda/numba/cuda/tests/test_binary_generation/Makefile,sha256=gFLhuz0vmU8ZrUXUcy78M-KTRs9vf0yOGnv28Ds7fuA,3631
-numba_cuda/numba/cuda/tests/test_binary_generation/generate_raw_ltoir.py,sha256=UBkdU7Hl5QTnIdZdMu1EJ_1fhtwliBL49EpGTt9fL_A,5172
+numba_cuda/numba/cuda/tests/nrt/test_nrt.py,sha256=2yzKcfpaAYuxO2xG9BO2BySyMXkQFki3yEEZobuoqPA,12720
+numba_cuda/numba/cuda/tests/nrt/test_nrt_refct.py,sha256=vishrSKwg0TRQ-FAEhRh7-Xbdg6Q8xVRy4C4-T1s5Pg,3212
+numba_cuda/numba/cuda/tests/test_binary_generation/Makefile,sha256=Gf5Q06VWY2DhUS6hZrLundTfUzWHXhr8LsbAddzdTE8,3641
+numba_cuda/numba/cuda/tests/test_binary_generation/generate_raw_ltoir.py,sha256=QRFDUQFsIk6zQ7U3sCK2a0F40pOmJdUzdP3deUjCOwk,5182
 numba_cuda/numba/cuda/tests/test_binary_generation/nrt_extern.cu,sha256=T9ubst3fFUK7EXyXXMi73wAban3VFFQ986cY5OcKfvI,157
 numba_cuda/numba/cuda/tests/test_binary_generation/test_device_functions.cu,sha256=IB5t-dVhrKVoue3AbUx3yVMxPG0hBF_yZbzb4642sf0,538
 numba_cuda/numba/cuda/tests/test_binary_generation/undefined_extern.cu,sha256=q3oxZziT8KDodeNcEBiWULH6vMrHCWucmJmtrg8C0d0,128
-numba_cuda-0.12.1.dist-info/licenses/LICENSE,sha256=eHeYE-XjASmwbxfsP5AImgfzRwZurZGqH1f6OFwJ4io,1326
-numba_cuda-0.12.1.dist-info/METADATA,sha256=H6JW6cSrhykHqICS50fIbGkrZ6SRgh_cTC3hTC2-XvQ,1859
-numba_cuda-0.12.1.dist-info/WHEEL,sha256=zaaOINJESkSfm_4HQVc5ssNzHCPXhJm0kEUakpsEHaU,91
-numba_cuda-0.12.1.dist-info/top_level.txt,sha256=C50SsH-8tXDmt7I0Y3nlJYhS5s6pqWflCPdobe9vx2M,11
-numba_cuda-0.12.1.dist-info/RECORD,,
+numba_cuda-0.13.0.dist-info/licenses/LICENSE,sha256=eHeYE-XjASmwbxfsP5AImgfzRwZurZGqH1f6OFwJ4io,1326
+numba_cuda-0.13.0.dist-info/METADATA,sha256=clEe3q5Jb4S4sixwT6RAgkGqLieoRYtWoyWEqBvSyZk,1859
+numba_cuda-0.13.0.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+numba_cuda-0.13.0.dist-info/top_level.txt,sha256=C50SsH-8tXDmt7I0Y3nlJYhS5s6pqWflCPdobe9vx2M,11
+numba_cuda-0.13.0.dist-info/RECORD,,

{numba_cuda-0.12.1.dist-info → numba_cuda-0.13.0.dist-info}/WHEEL

@@ -1,5 +1,5 @@
 Wheel-Version: 1.0
-Generator: setuptools (80.8.0)
+Generator: setuptools (80.9.0)
 Root-Is-Purelib: true
 Tag: py3-none-any
 

numba_cuda/numba/cuda/runtime/__init__.py

@@ -1 +0,0 @@
-from numba.cuda.runtime.nrt import rtsys  # noqa: F401