Rebase to CTK 12.1

Author: vzhurba01

README.md

@@ -7,8 +7,8 @@
 CUDA Python is supported on all platforms that CUDA is supported. Specific dependencies are as follows:
 
 * Driver: Linux (450.80.02 or later) Windows (456.38 or later)
-* CUDA Toolkit 12.0
-* Python 3.7 to 3.10
+* CUDA Toolkit 12.0 to 12.1
+* Python 3.8 to 3.11
 
 Only the NVRTC redistributable component is required from the CUDA Toolkit. [CUDA Toolkit Documentation](https://docs.nvidia.com/cuda/index.html) Installation Guides can be used for guidance. Note that the NVRTC component in the Toolkit can be obtained via PiPy, Conda or Local Installer.
 

cuda/_cuda/ccuda.pxd.in

@@ -1,4 +1,4 @@
-# Copyright 2021-2022 NVIDIA Corporation.  All rights reserved.
+# Copyright 2021-2023 NVIDIA Corporation.  All rights reserved.
 #
 # Please refer to the NVIDIA end user license agreement (EULA) associated
 # with this source code for terms and conditions that govern your use of
@@ -182,6 +182,11 @@ cdef CUresult _cuCtxGetDevice(CUdevice* device) nogil except ?CUDA_ERROR_NOT_FOU
 cdef CUresult _cuCtxGetFlags(unsigned int* flags) nogil except ?CUDA_ERROR_NOT_FOUND
 {{endif}}
 
+{{if 'cuCtxSetFlags' in found_functions}}
+
+cdef CUresult _cuCtxSetFlags(unsigned int flags) nogil except ?CUDA_ERROR_NOT_FOUND
+{{endif}}
+
 {{if 'cuCtxGetId' in found_functions}}
 
 cdef CUresult _cuCtxGetId(CUcontext ctx, unsigned long long* ctxId) nogil except ?CUDA_ERROR_NOT_FOUND
@@ -877,6 +882,36 @@ cdef CUresult _cuMemPoolExportPointer(CUmemPoolPtrExportData* shareData_out, CUd
 cdef CUresult _cuMemPoolImportPointer(CUdeviceptr* ptr_out, CUmemoryPool pool, CUmemPoolPtrExportData* shareData) nogil except ?CUDA_ERROR_NOT_FOUND
 {{endif}}
 
+{{if 'cuMulticastCreate' in found_functions}}
+
+cdef CUresult _cuMulticastCreate(CUmemGenericAllocationHandle* mcHandle, const CUmulticastObjectProp* prop) nogil except ?CUDA_ERROR_NOT_FOUND
+{{endif}}
+
+{{if 'cuMulticastAddDevice' in found_functions}}
+
+cdef CUresult _cuMulticastAddDevice(CUmemGenericAllocationHandle mcHandle, CUdevice dev) nogil except ?CUDA_ERROR_NOT_FOUND
+{{endif}}
+
+{{if 'cuMulticastBindMem' in found_functions}}
+
+cdef CUresult _cuMulticastBindMem(CUmemGenericAllocationHandle mcHandle, size_t mcOffset, CUmemGenericAllocationHandle memHandle, size_t memOffset, size_t size, unsigned long long flags) nogil except ?CUDA_ERROR_NOT_FOUND
+{{endif}}
+
+{{if 'cuMulticastBindAddr' in found_functions}}
+
+cdef CUresult _cuMulticastBindAddr(CUmemGenericAllocationHandle mcHandle, size_t mcOffset, CUdeviceptr memptr, size_t size, unsigned long long flags) nogil except ?CUDA_ERROR_NOT_FOUND
+{{endif}}
+
+{{if 'cuMulticastUnbind' in found_functions}}
+
+cdef CUresult _cuMulticastUnbind(CUmemGenericAllocationHandle mcHandle, CUdevice dev, size_t mcOffset, size_t size) nogil except ?CUDA_ERROR_NOT_FOUND
+{{endif}}
+
+{{if 'cuMulticastGetGranularity' in found_functions}}
+
+cdef CUresult _cuMulticastGetGranularity(size_t* granularity, const CUmulticastObjectProp* prop, CUmulticastGranularity_flags option) nogil except ?CUDA_ERROR_NOT_FOUND
+{{endif}}
+
 {{if 'cuPointerGetAttribute' in found_functions}}
 
 cdef CUresult _cuPointerGetAttribute(void* data, CUpointer_attribute attribute, CUdeviceptr ptr) nogil except ?CUDA_ERROR_NOT_FOUND
@@ -1902,6 +1937,26 @@ cdef CUresult _cuGraphicsUnmapResources(unsigned int count, CUgraphicsResource*
 cdef CUresult _cuGetProcAddress_v2(const char* symbol, void** pfn, int cudaVersion, cuuint64_t flags, CUdriverProcAddressQueryResult* symbolStatus) nogil except ?CUDA_ERROR_NOT_FOUND
 {{endif}}
 
+{{if 'cuCoredumpGetAttribute' in found_functions}}
+
+cdef CUresult _cuCoredumpGetAttribute(CUcoredumpSettings attrib, void* value, size_t* size) nogil except ?CUDA_ERROR_NOT_FOUND
+{{endif}}
+
+{{if 'cuCoredumpGetAttributeGlobal' in found_functions}}
+
+cdef CUresult _cuCoredumpGetAttributeGlobal(CUcoredumpSettings attrib, void* value, size_t* size) nogil except ?CUDA_ERROR_NOT_FOUND
+{{endif}}
+
+{{if 'cuCoredumpSetAttribute' in found_functions}}
+
+cdef CUresult _cuCoredumpSetAttribute(CUcoredumpSettings attrib, void* value, size_t* size) nogil except ?CUDA_ERROR_NOT_FOUND
+{{endif}}
+
+{{if 'cuCoredumpSetAttributeGlobal' in found_functions}}
+
+cdef CUresult _cuCoredumpSetAttributeGlobal(CUcoredumpSettings attrib, void* value, size_t* size) nogil except ?CUDA_ERROR_NOT_FOUND
+{{endif}}
+
 {{if 'cuGetExportTable' in found_functions}}
 
 cdef CUresult _cuGetExportTable(const void** ppExportTable, const CUuuid* pExportTableId) nogil except ?CUDA_ERROR_NOT_FOUND