diff --git a/cuda_bindings/regenerate_abi.sh b/cuda_bindings/regenerate_abi.sh
new file mode 100755
index 000000000..60cc888ba
--- /dev/null
+++ b/cuda_bindings/regenerate_abi.sh
@@ -0,0 +1,8 @@
+#!/usr/bin/env bash
+
+# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# SPDX-License-Identifier: Apache-2.0
+
+SCRIPT_DIR=$( cd -- "$( dirname -- "${BASH_SOURCE[0]}" )" &> /dev/null && pwd )
+python ${SCRIPT_DIR}/tests/test_cython_abi.py regenerate
diff --git a/cuda_bindings/tests/abi/cuda/bindings/_bindings/cydriver.abi.json b/cuda_bindings/tests/abi/cuda/bindings/_bindings/cydriver.abi.json
new file mode 100644
index 000000000..b374c1acf
--- /dev/null
+++ b/cuda_bindings/tests/abi/cuda/bindings/_bindings/cydriver.abi.json
@@ -0,0 +1,472 @@
+{
+  "_cuArray3DCreate_v2": "CUresult (CUarray *, CUDA_ARRAY3D_DESCRIPTOR const *)",
+  "_cuArray3DGetDescriptor_v2": "CUresult (CUDA_ARRAY3D_DESCRIPTOR *, CUarray)",
+  "_cuArrayCreate_v2": "CUresult (CUarray *, CUDA_ARRAY_DESCRIPTOR const *)",
+  "_cuArrayDestroy": "CUresult (CUarray)",
+  "_cuArrayGetDescriptor_v2": "CUresult (CUDA_ARRAY_DESCRIPTOR *, CUarray)",
+  "_cuArrayGetMemoryRequirements": "CUresult (CUDA_ARRAY_MEMORY_REQUIREMENTS *, CUarray, CUdevice)",
+  "_cuArrayGetPlane": "CUresult (CUarray *, CUarray, unsigned int)",
+  "_cuArrayGetSparseProperties": "CUresult (CUDA_ARRAY_SPARSE_PROPERTIES *, CUarray)",
+  "_cuCheckpointProcessCheckpoint": "CUresult (int, CUcheckpointCheckpointArgs *)",
+  "_cuCheckpointProcessGetRestoreThreadId": "CUresult (int, int *)",
+  "_cuCheckpointProcessGetState": "CUresult (int, CUprocessState *)",
+  "_cuCheckpointProcessLock": "CUresult (int, CUcheckpointLockArgs *)",
+  "_cuCheckpointProcessRestore": "CUresult (int, CUcheckpointRestoreArgs *)",
+  "_cuCheckpointProcessUnlock": "CUresult (int, CUcheckpointUnlockArgs *)",
+  "_cuCoredumpGetAttribute": "CUresult (CUcoredumpSettings, void *, size_t *)",
+  "_cuCoredumpGetAttributeGlobal": "CUresult (CUcoredumpSettings, void *, size_t *)",
+  "_cuCoredumpSetAttribute": "CUresult (CUcoredumpSettings, void *, size_t *)",
+  "_cuCoredumpSetAttributeGlobal": "CUresult (CUcoredumpSettings, void *, size_t *)",
+  "_cuCtxAttach": "CUresult (CUcontext *, unsigned int)",
+  "_cuCtxCreate_v4": "CUresult (CUcontext *, CUctxCreateParams *, unsigned int, CUdevice)",
+  "_cuCtxDestroy_v2": "CUresult (CUcontext)",
+  "_cuCtxDetach": "CUresult (CUcontext)",
+  "_cuCtxDisablePeerAccess": "CUresult (CUcontext)",
+  "_cuCtxEnablePeerAccess": "CUresult (CUcontext, unsigned int)",
+  "_cuCtxFromGreenCtx": "CUresult (CUcontext *, CUgreenCtx)",
+  "_cuCtxGetApiVersion": "CUresult (CUcontext, unsigned int *)",
+  "_cuCtxGetCacheConfig": "CUresult (CUfunc_cache *)",
+  "_cuCtxGetCurrent": "CUresult (CUcontext *)",
+  "_cuCtxGetDevResource": "CUresult (CUcontext, CUdevResource *, CUdevResourceType)",
+  "_cuCtxGetDevice": "CUresult (CUdevice *)",
+  "_cuCtxGetDevice_v2": "CUresult (CUdevice *, CUcontext)",
+  "_cuCtxGetExecAffinity": "CUresult (CUexecAffinityParam *, CUexecAffinityType)",
+  "_cuCtxGetFlags": "CUresult (unsigned int *)",
+  "_cuCtxGetId": "CUresult (CUcontext, unsigned PY_LONG_LONG *)",
+  "_cuCtxGetLimit": "CUresult (size_t *, CUlimit)",
+  "_cuCtxGetSharedMemConfig": "CUresult (CUsharedconfig *)",
+  "_cuCtxGetStreamPriorityRange": "CUresult (int *, int *)",
+  "_cuCtxPopCurrent_v2": "CUresult (CUcontext *)",
+  "_cuCtxPushCurrent_v2": "CUresult (CUcontext)",
+  "_cuCtxRecordEvent": "CUresult (CUcontext, CUevent)",
+  "_cuCtxResetPersistingL2Cache": "CUresult (void)",
+  "_cuCtxSetCacheConfig": "CUresult (CUfunc_cache)",
+  "_cuCtxSetCurrent": "CUresult (CUcontext)",
+  "_cuCtxSetFlags": "CUresult (unsigned int)",
+  "_cuCtxSetLimit": "CUresult (CUlimit, size_t)",
+  "_cuCtxSetSharedMemConfig": "CUresult (CUsharedconfig)",
+  "_cuCtxSynchronize": "CUresult (void)",
+  "_cuCtxSynchronize_v2": "CUresult (CUcontext)",
+  "_cuCtxWaitEvent": "CUresult (CUcontext, CUevent)",
+  "_cuDestroyExternalMemory": "CUresult (CUexternalMemory)",
+  "_cuDestroyExternalSemaphore": "CUresult (CUexternalSemaphore)",
+  "_cuDevResourceGenerateDesc": "CUresult (CUdevResourceDesc *, CUdevResource *, unsigned int)",
+  "_cuDevSmResourceSplitByCount": "CUresult (CUdevResource *, unsigned int *, CUdevResource const *, CUdevResource *, unsigned int, unsigned int)",
+  "_cuDeviceCanAccessPeer": "CUresult (int *, CUdevice, CUdevice)",
+  "_cuDeviceComputeCapability": "CUresult (int *, int *, CUdevice)",
+  "_cuDeviceGet": "CUresult (CUdevice *, int)",
+  "_cuDeviceGetAttribute": "CUresult (int *, CUdevice_attribute, CUdevice)",
+  "_cuDeviceGetByPCIBusId": "CUresult (CUdevice *, char const *)",
+  "_cuDeviceGetCount": "CUresult (int *)",
+  "_cuDeviceGetDefaultMemPool": "CUresult (CUmemoryPool *, CUdevice)",
+  "_cuDeviceGetDevResource": "CUresult (CUdevice, CUdevResource *, CUdevResourceType)",
+  "_cuDeviceGetExecAffinitySupport": "CUresult (int *, CUexecAffinityType, CUdevice)",
+  "_cuDeviceGetGraphMemAttribute": "CUresult (CUdevice, CUgraphMem_attribute, void *)",
+  "_cuDeviceGetHostAtomicCapabilities": "CUresult (unsigned int *, CUatomicOperation const *, unsigned int, CUdevice)",
+  "_cuDeviceGetLuid": "CUresult (char *, unsigned int *, CUdevice)",
+  "_cuDeviceGetMemPool": "CUresult (CUmemoryPool *, CUdevice)",
+  "_cuDeviceGetName": "CUresult (char *, int, CUdevice)",
+  "_cuDeviceGetNvSciSyncAttributes": "CUresult (void *, CUdevice, int)",
+  "_cuDeviceGetP2PAtomicCapabilities": "CUresult (unsigned int *, CUatomicOperation const *, unsigned int, CUdevice, CUdevice)",
+  "_cuDeviceGetP2PAttribute": "CUresult (int *, CUdevice_P2PAttribute, CUdevice, CUdevice)",
+  "_cuDeviceGetPCIBusId": "CUresult (char *, int, CUdevice)",
+  "_cuDeviceGetProperties": "CUresult (CUdevprop *, CUdevice)",
+  "_cuDeviceGetTexture1DLinearMaxWidth": "CUresult (size_t *, CUarray_format, unsigned int, CUdevice)",
+  "_cuDeviceGetUuid_v2": "CUresult (CUuuid *, CUdevice)",
+  "_cuDeviceGraphMemTrim": "CUresult (CUdevice)",
+  "_cuDevicePrimaryCtxGetState": "CUresult (CUdevice, unsigned int *, int *)",
+  "_cuDevicePrimaryCtxRelease_v2": "CUresult (CUdevice)",
+  "_cuDevicePrimaryCtxReset_v2": "CUresult (CUdevice)",
+  "_cuDevicePrimaryCtxRetain": "CUresult (CUcontext *, CUdevice)",
+  "_cuDevicePrimaryCtxSetFlags_v2": "CUresult (CUdevice, unsigned int)",
+  "_cuDeviceRegisterAsyncNotification": "CUresult (CUdevice, CUasyncCallback, void *, CUasyncCallbackHandle *)",
+  "_cuDeviceSetGraphMemAttribute": "CUresult (CUdevice, CUgraphMem_attribute, void *)",
+  "_cuDeviceSetMemPool": "CUresult (CUdevice, CUmemoryPool)",
+  "_cuDeviceTotalMem_v2": "CUresult (size_t *, CUdevice)",
+  "_cuDeviceUnregisterAsyncNotification": "CUresult (CUdevice, CUasyncCallbackHandle)",
+  "_cuDriverGetVersion": "CUresult (int *)",
+  "_cuEGLStreamConsumerAcquireFrame": "CUresult (__pyx_t_4cuda_8bindings_8cydriver_CUeglStreamConnection *, CUgraphicsResource *, CUstream *, unsigned int)",
+  "_cuEGLStreamConsumerConnect": "CUresult (__pyx_t_4cuda_8bindings_8cydriver_CUeglStreamConnection *, __pyx_t_4cuda_8bindings_8cydriver_EGLStreamKHR)",
+  "_cuEGLStreamConsumerConnectWithFlags": "CUresult (__pyx_t_4cuda_8bindings_8cydriver_CUeglStreamConnection *, __pyx_t_4cuda_8bindings_8cydriver_EGLStreamKHR, unsigned int)",
+  "_cuEGLStreamConsumerDisconnect": "CUresult (__pyx_t_4cuda_8bindings_8cydriver_CUeglStreamConnection *)",
+  "_cuEGLStreamConsumerReleaseFrame": "CUresult (__pyx_t_4cuda_8bindings_8cydriver_CUeglStreamConnection *, CUgraphicsResource, CUstream *)",
+  "_cuEGLStreamProducerConnect": "CUresult (__pyx_t_4cuda_8bindings_8cydriver_CUeglStreamConnection *, __pyx_t_4cuda_8bindings_8cydriver_EGLStreamKHR, __pyx_t_4cuda_8bindings_8cydriver_EGLint, __pyx_t_4cuda_8bindings_8cydriver_EGLint)",
+  "_cuEGLStreamProducerDisconnect": "CUresult (__pyx_t_4cuda_8bindings_8cydriver_CUeglStreamConnection *)",
+  "_cuEGLStreamProducerPresentFrame": "CUresult (__pyx_t_4cuda_8bindings_8cydriver_CUeglStreamConnection *, __pyx_t_4cuda_8bindings_8cydriver_CUeglFrame, CUstream *)",
+  "_cuEGLStreamProducerReturnFrame": "CUresult (__pyx_t_4cuda_8bindings_8cydriver_CUeglStreamConnection *, __pyx_t_4cuda_8bindings_8cydriver_CUeglFrame *, CUstream *)",
+  "_cuEventCreate": "CUresult (CUevent *, unsigned int)",
+  "_cuEventCreateFromEGLSync": "CUresult (CUevent *, __pyx_t_4cuda_8bindings_8cydriver_EGLSyncKHR, unsigned int)",
+  "_cuEventDestroy_v2": "CUresult (CUevent)",
+  "_cuEventElapsedTime_v2": "CUresult (float *, CUevent, CUevent)",
+  "_cuEventQuery": "CUresult (CUevent)",
+  "_cuEventRecord": "CUresult (CUevent, CUstream)",
+  "_cuEventRecordWithFlags": "CUresult (CUevent, CUstream, unsigned int)",
+  "_cuEventSynchronize": "CUresult (CUevent)",
+  "_cuExternalMemoryGetMappedBuffer": "CUresult (CUdeviceptr *, CUexternalMemory, CUDA_EXTERNAL_MEMORY_BUFFER_DESC const *)",
+  "_cuExternalMemoryGetMappedMipmappedArray": "CUresult (CUmipmappedArray *, CUexternalMemory, CUDA_EXTERNAL_MEMORY_MIPMAPPED_ARRAY_DESC const *)",
+  "_cuFlushGPUDirectRDMAWrites": "CUresult (CUflushGPUDirectRDMAWritesTarget, CUflushGPUDirectRDMAWritesScope)",
+  "_cuFuncGetAttribute": "CUresult (int *, CUfunction_attribute, CUfunction)",
+  "_cuFuncGetModule": "CUresult (CUmodule *, CUfunction)",
+  "_cuFuncGetName": "CUresult (char const **, CUfunction)",
+  "_cuFuncGetParamInfo": "CUresult (CUfunction, size_t, size_t *, size_t *)",
+  "_cuFuncIsLoaded": "CUresult (CUfunctionLoadingState *, CUfunction)",
+  "_cuFuncLoad": "CUresult (CUfunction)",
+  "_cuFuncSetAttribute": "CUresult (CUfunction, CUfunction_attribute, int)",
+  "_cuFuncSetBlockShape": "CUresult (CUfunction, int, int, int)",
+  "_cuFuncSetCacheConfig": "CUresult (CUfunction, CUfunc_cache)",
+  "_cuFuncSetSharedMemConfig": "CUresult (CUfunction, CUsharedconfig)",
+  "_cuFuncSetSharedSize": "CUresult (CUfunction, unsigned int)",
+  "_cuGLGetDevices_v2": "CUresult (unsigned int *, CUdevice *, unsigned int, __pyx_t_4cuda_8bindings_8cydriver_CUGLDeviceList)",
+  "_cuGetErrorName": "CUresult (CUresult, char const **)",
+  "_cuGetErrorString": "CUresult (CUresult, char const **)",
+  "_cuGetExportTable": "CUresult (void const **, CUuuid const *)",
+  "_cuGetProcAddress_v2": "CUresult (char const *, void **, int, cuuint64_t, CUdriverProcAddressQueryResult *)",
+  "_cuGraphAddBatchMemOpNode": "CUresult (CUgraphNode *, CUgraph, CUgraphNode const *, size_t, CUDA_BATCH_MEM_OP_NODE_PARAMS const *)",
+  "_cuGraphAddChildGraphNode": "CUresult (CUgraphNode *, CUgraph, CUgraphNode const *, size_t, CUgraph)",
+  "_cuGraphAddDependencies_v2": "CUresult (CUgraph, CUgraphNode const *, CUgraphNode const *, CUgraphEdgeData const *, size_t)",
+  "_cuGraphAddEmptyNode": "CUresult (CUgraphNode *, CUgraph, CUgraphNode const *, size_t)",
+  "_cuGraphAddEventRecordNode": "CUresult (CUgraphNode *, CUgraph, CUgraphNode const *, size_t, CUevent)",
+  "_cuGraphAddEventWaitNode": "CUresult (CUgraphNode *, CUgraph, CUgraphNode const *, size_t, CUevent)",
+  "_cuGraphAddExternalSemaphoresSignalNode": "CUresult (CUgraphNode *, CUgraph, CUgraphNode const *, size_t, CUDA_EXT_SEM_SIGNAL_NODE_PARAMS const *)",
+  "_cuGraphAddExternalSemaphoresWaitNode": "CUresult (CUgraphNode *, CUgraph, CUgraphNode const *, size_t, CUDA_EXT_SEM_WAIT_NODE_PARAMS const *)",
+  "_cuGraphAddHostNode": "CUresult (CUgraphNode *, CUgraph, CUgraphNode const *, size_t, CUDA_HOST_NODE_PARAMS const *)",
+  "_cuGraphAddKernelNode_v2": "CUresult (CUgraphNode *, CUgraph, CUgraphNode const *, size_t, CUDA_KERNEL_NODE_PARAMS const *)",
+  "_cuGraphAddMemAllocNode": "CUresult (CUgraphNode *, CUgraph, CUgraphNode const *, size_t, CUDA_MEM_ALLOC_NODE_PARAMS *)",
+  "_cuGraphAddMemFreeNode": "CUresult (CUgraphNode *, CUgraph, CUgraphNode const *, size_t, CUdeviceptr)",
+  "_cuGraphAddMemcpyNode": "CUresult (CUgraphNode *, CUgraph, CUgraphNode const *, size_t, CUDA_MEMCPY3D const *, CUcontext)",
+  "_cuGraphAddMemsetNode": "CUresult (CUgraphNode *, CUgraph, CUgraphNode const *, size_t, CUDA_MEMSET_NODE_PARAMS const *, CUcontext)",
+  "_cuGraphAddNode_v2": "CUresult (CUgraphNode *, CUgraph, CUgraphNode const *, CUgraphEdgeData const *, size_t, CUgraphNodeParams *)",
+  "_cuGraphBatchMemOpNodeGetParams": "CUresult (CUgraphNode, CUDA_BATCH_MEM_OP_NODE_PARAMS *)",
+  "_cuGraphBatchMemOpNodeSetParams": "CUresult (CUgraphNode, CUDA_BATCH_MEM_OP_NODE_PARAMS const *)",
+  "_cuGraphChildGraphNodeGetGraph": "CUresult (CUgraphNode, CUgraph *)",
+  "_cuGraphClone": "CUresult (CUgraph *, CUgraph)",
+  "_cuGraphConditionalHandleCreate": "CUresult (CUgraphConditionalHandle *, CUgraph, CUcontext, unsigned int, unsigned int)",
+  "_cuGraphCreate": "CUresult (CUgraph *, unsigned int)",
+  "_cuGraphDebugDotPrint": "CUresult (CUgraph, char const *, unsigned int)",
+  "_cuGraphDestroy": "CUresult (CUgraph)",
+  "_cuGraphDestroyNode": "CUresult (CUgraphNode)",
+  "_cuGraphEventRecordNodeGetEvent": "CUresult (CUgraphNode, CUevent *)",
+  "_cuGraphEventRecordNodeSetEvent": "CUresult (CUgraphNode, CUevent)",
+  "_cuGraphEventWaitNodeGetEvent": "CUresult (CUgraphNode, CUevent *)",
+  "_cuGraphEventWaitNodeSetEvent": "CUresult (CUgraphNode, CUevent)",
+  "_cuGraphExecBatchMemOpNodeSetParams": "CUresult (CUgraphExec, CUgraphNode, CUDA_BATCH_MEM_OP_NODE_PARAMS const *)",
+  "_cuGraphExecChildGraphNodeSetParams": "CUresult (CUgraphExec, CUgraphNode, CUgraph)",
+  "_cuGraphExecDestroy": "CUresult (CUgraphExec)",
+  "_cuGraphExecEventRecordNodeSetEvent": "CUresult (CUgraphExec, CUgraphNode, CUevent)",
+  "_cuGraphExecEventWaitNodeSetEvent": "CUresult (CUgraphExec, CUgraphNode, CUevent)",
+  "_cuGraphExecExternalSemaphoresSignalNodeSetParams": "CUresult (CUgraphExec, CUgraphNode, CUDA_EXT_SEM_SIGNAL_NODE_PARAMS const *)",
+  "_cuGraphExecExternalSemaphoresWaitNodeSetParams": "CUresult (CUgraphExec, CUgraphNode, CUDA_EXT_SEM_WAIT_NODE_PARAMS const *)",
+  "_cuGraphExecGetFlags": "CUresult (CUgraphExec, cuuint64_t *)",
+  "_cuGraphExecHostNodeSetParams": "CUresult (CUgraphExec, CUgraphNode, CUDA_HOST_NODE_PARAMS const *)",
+  "_cuGraphExecKernelNodeSetParams_v2": "CUresult (CUgraphExec, CUgraphNode, CUDA_KERNEL_NODE_PARAMS const *)",
+  "_cuGraphExecMemcpyNodeSetParams": "CUresult (CUgraphExec, CUgraphNode, CUDA_MEMCPY3D const *, CUcontext)",
+  "_cuGraphExecMemsetNodeSetParams": "CUresult (CUgraphExec, CUgraphNode, CUDA_MEMSET_NODE_PARAMS const *, CUcontext)",
+  "_cuGraphExecNodeSetParams": "CUresult (CUgraphExec, CUgraphNode, CUgraphNodeParams *)",
+  "_cuGraphExecUpdate_v2": "CUresult (CUgraphExec, CUgraph, CUgraphExecUpdateResultInfo *)",
+  "_cuGraphExternalSemaphoresSignalNodeGetParams": "CUresult (CUgraphNode, CUDA_EXT_SEM_SIGNAL_NODE_PARAMS *)",
+  "_cuGraphExternalSemaphoresSignalNodeSetParams": "CUresult (CUgraphNode, CUDA_EXT_SEM_SIGNAL_NODE_PARAMS const *)",
+  "_cuGraphExternalSemaphoresWaitNodeGetParams": "CUresult (CUgraphNode, CUDA_EXT_SEM_WAIT_NODE_PARAMS *)",
+  "_cuGraphExternalSemaphoresWaitNodeSetParams": "CUresult (CUgraphNode, CUDA_EXT_SEM_WAIT_NODE_PARAMS const *)",
+  "_cuGraphGetEdges_v2": "CUresult (CUgraph, CUgraphNode *, CUgraphNode *, CUgraphEdgeData *, size_t *)",
+  "_cuGraphGetNodes": "CUresult (CUgraph, CUgraphNode *, size_t *)",
+  "_cuGraphGetRootNodes": "CUresult (CUgraph, CUgraphNode *, size_t *)",
+  "_cuGraphHostNodeGetParams": "CUresult (CUgraphNode, CUDA_HOST_NODE_PARAMS *)",
+  "_cuGraphHostNodeSetParams": "CUresult (CUgraphNode, CUDA_HOST_NODE_PARAMS const *)",
+  "_cuGraphInstantiateWithFlags": "CUresult (CUgraphExec *, CUgraph, unsigned PY_LONG_LONG)",
+  "_cuGraphInstantiateWithParams": "CUresult (CUgraphExec *, CUgraph, CUDA_GRAPH_INSTANTIATE_PARAMS *)",
+  "_cuGraphKernelNodeCopyAttributes": "CUresult (CUgraphNode, CUgraphNode)",
+  "_cuGraphKernelNodeGetAttribute": "CUresult (CUgraphNode, CUkernelNodeAttrID, CUkernelNodeAttrValue *)",
+  "_cuGraphKernelNodeGetParams_v2": "CUresult (CUgraphNode, CUDA_KERNEL_NODE_PARAMS *)",
+  "_cuGraphKernelNodeSetAttribute": "CUresult (CUgraphNode, CUkernelNodeAttrID, CUkernelNodeAttrValue const *)",
+  "_cuGraphKernelNodeSetParams_v2": "CUresult (CUgraphNode, CUDA_KERNEL_NODE_PARAMS const *)",
+  "_cuGraphLaunch": "CUresult (CUgraphExec, CUstream)",
+  "_cuGraphMemAllocNodeGetParams": "CUresult (CUgraphNode, CUDA_MEM_ALLOC_NODE_PARAMS *)",
+  "_cuGraphMemFreeNodeGetParams": "CUresult (CUgraphNode, CUdeviceptr *)",
+  "_cuGraphMemcpyNodeGetParams": "CUresult (CUgraphNode, CUDA_MEMCPY3D *)",
+  "_cuGraphMemcpyNodeSetParams": "CUresult (CUgraphNode, CUDA_MEMCPY3D const *)",
+  "_cuGraphMemsetNodeGetParams": "CUresult (CUgraphNode, CUDA_MEMSET_NODE_PARAMS *)",
+  "_cuGraphMemsetNodeSetParams": "CUresult (CUgraphNode, CUDA_MEMSET_NODE_PARAMS const *)",
+  "_cuGraphNodeFindInClone": "CUresult (CUgraphNode *, CUgraphNode, CUgraph)",
+  "_cuGraphNodeGetDependencies_v2": "CUresult (CUgraphNode, CUgraphNode *, CUgraphEdgeData *, size_t *)",
+  "_cuGraphNodeGetDependentNodes_v2": "CUresult (CUgraphNode, CUgraphNode *, CUgraphEdgeData *, size_t *)",
+  "_cuGraphNodeGetEnabled": "CUresult (CUgraphExec, CUgraphNode, unsigned int *)",
+  "_cuGraphNodeGetType": "CUresult (CUgraphNode, CUgraphNodeType *)",
+  "_cuGraphNodeSetEnabled": "CUresult (CUgraphExec, CUgraphNode, unsigned int)",
+  "_cuGraphNodeSetParams": "CUresult (CUgraphNode, CUgraphNodeParams *)",
+  "_cuGraphReleaseUserObject": "CUresult (CUgraph, CUuserObject, unsigned int)",
+  "_cuGraphRemoveDependencies_v2": "CUresult (CUgraph, CUgraphNode const *, CUgraphNode const *, CUgraphEdgeData const *, size_t)",
+  "_cuGraphRetainUserObject": "CUresult (CUgraph, CUuserObject, unsigned int, unsigned int)",
+  "_cuGraphUpload": "CUresult (CUgraphExec, CUstream)",
+  "_cuGraphicsEGLRegisterImage": "CUresult (CUgraphicsResource *, __pyx_t_4cuda_8bindings_8cydriver_EGLImageKHR, unsigned int)",
+  "_cuGraphicsGLRegisterBuffer": "CUresult (CUgraphicsResource *, __pyx_t_4cuda_8bindings_8cydriver_GLuint, unsigned int)",
+  "_cuGraphicsGLRegisterImage": "CUresult (CUgraphicsResource *, __pyx_t_4cuda_8bindings_8cydriver_GLuint, __pyx_t_4cuda_8bindings_8cydriver_GLenum, unsigned int)",
+  "_cuGraphicsMapResources": "CUresult (unsigned int, CUgraphicsResource *, CUstream)",
+  "_cuGraphicsResourceGetMappedEglFrame": "CUresult (__pyx_t_4cuda_8bindings_8cydriver_CUeglFrame *, CUgraphicsResource, unsigned int, unsigned int)",
+  "_cuGraphicsResourceGetMappedMipmappedArray": "CUresult (CUmipmappedArray *, CUgraphicsResource)",
+  "_cuGraphicsResourceGetMappedPointer_v2": "CUresult (CUdeviceptr *, size_t *, CUgraphicsResource)",
+  "_cuGraphicsResourceSetMapFlags_v2": "CUresult (CUgraphicsResource, unsigned int)",
+  "_cuGraphicsSubResourceGetMappedArray": "CUresult (CUarray *, CUgraphicsResource, unsigned int, unsigned int)",
+  "_cuGraphicsUnmapResources": "CUresult (unsigned int, CUgraphicsResource *, CUstream)",
+  "_cuGraphicsUnregisterResource": "CUresult (CUgraphicsResource)",
+  "_cuGraphicsVDPAURegisterOutputSurface": "CUresult (CUgraphicsResource *, __pyx_t_4cuda_8bindings_8cydriver_VdpOutputSurface, unsigned int)",
+  "_cuGraphicsVDPAURegisterVideoSurface": "CUresult (CUgraphicsResource *, __pyx_t_4cuda_8bindings_8cydriver_VdpVideoSurface, unsigned int)",
+  "_cuGreenCtxCreate": "CUresult (CUgreenCtx *, CUdevResourceDesc, CUdevice, unsigned int)",
+  "_cuGreenCtxDestroy": "CUresult (CUgreenCtx)",
+  "_cuGreenCtxGetDevResource": "CUresult (CUgreenCtx, CUdevResource *, CUdevResourceType)",
+  "_cuGreenCtxGetId": "CUresult (CUgreenCtx, unsigned PY_LONG_LONG *)",
+  "_cuGreenCtxRecordEvent": "CUresult (CUgreenCtx, CUevent)",
+  "_cuGreenCtxStreamCreate": "CUresult (CUstream *, CUgreenCtx, unsigned int, int)",
+  "_cuGreenCtxWaitEvent": "CUresult (CUgreenCtx, CUevent)",
+  "_cuImportExternalMemory": "CUresult (CUexternalMemory *, CUDA_EXTERNAL_MEMORY_HANDLE_DESC const *)",
+  "_cuImportExternalSemaphore": "CUresult (CUexternalSemaphore *, CUDA_EXTERNAL_SEMAPHORE_HANDLE_DESC const *)",
+  "_cuInit": "CUresult (unsigned int)",
+  "_cuIpcCloseMemHandle": "CUresult (CUdeviceptr)",
+  "_cuIpcGetEventHandle": "CUresult (CUipcEventHandle *, CUevent)",
+  "_cuIpcGetMemHandle": "CUresult (CUipcMemHandle *, CUdeviceptr)",
+  "_cuIpcOpenEventHandle": "CUresult (CUevent *, CUipcEventHandle)",
+  "_cuIpcOpenMemHandle_v2": "CUresult (CUdeviceptr *, CUipcMemHandle, unsigned int)",
+  "_cuKernelGetAttribute": "CUresult (int *, CUfunction_attribute, CUkernel, CUdevice)",
+  "_cuKernelGetFunction": "CUresult (CUfunction *, CUkernel)",
+  "_cuKernelGetLibrary": "CUresult (CUlibrary *, CUkernel)",
+  "_cuKernelGetName": "CUresult (char const **, CUkernel)",
+  "_cuKernelGetParamInfo": "CUresult (CUkernel, size_t, size_t *, size_t *)",
+  "_cuKernelSetAttribute": "CUresult (CUfunction_attribute, int, CUkernel, CUdevice)",
+  "_cuKernelSetCacheConfig": "CUresult (CUkernel, CUfunc_cache, CUdevice)",
+  "_cuLaunch": "CUresult (CUfunction)",
+  "_cuLaunchCooperativeKernel": "CUresult (CUfunction, unsigned int, unsigned int, unsigned int, unsigned int, unsigned int, unsigned int, unsigned int, CUstream, void **)",
+  "_cuLaunchCooperativeKernelMultiDevice": "CUresult (CUDA_LAUNCH_PARAMS *, unsigned int, unsigned int)",
+  "_cuLaunchGrid": "CUresult (CUfunction, int, int)",
+  "_cuLaunchGridAsync": "CUresult (CUfunction, int, int, CUstream)",
+  "_cuLaunchHostFunc": "CUresult (CUstream, CUhostFn, void *)",
+  "_cuLaunchKernel": "CUresult (CUfunction, unsigned int, unsigned int, unsigned int, unsigned int, unsigned int, unsigned int, unsigned int, CUstream, void **, void **)",
+  "_cuLaunchKernelEx": "CUresult (CUlaunchConfig const *, CUfunction, void **, void **)",
+  "_cuLibraryEnumerateKernels": "CUresult (CUkernel *, unsigned int, CUlibrary)",
+  "_cuLibraryGetGlobal": "CUresult (CUdeviceptr *, size_t *, CUlibrary, char const *)",
+  "_cuLibraryGetKernel": "CUresult (CUkernel *, CUlibrary, char const *)",
+  "_cuLibraryGetKernelCount": "CUresult (unsigned int *, CUlibrary)",
+  "_cuLibraryGetManaged": "CUresult (CUdeviceptr *, size_t *, CUlibrary, char const *)",
+  "_cuLibraryGetModule": "CUresult (CUmodule *, CUlibrary)",
+  "_cuLibraryGetUnifiedFunction": "CUresult (void **, CUlibrary, char const *)",
+  "_cuLibraryLoadData": "CUresult (CUlibrary *, void const *, CUjit_option *, void **, unsigned int, CUlibraryOption *, void **, unsigned int)",
+  "_cuLibraryLoadFromFile": "CUresult (CUlibrary *, char const *, CUjit_option *, void **, unsigned int, CUlibraryOption *, void **, unsigned int)",
+  "_cuLibraryUnload": "CUresult (CUlibrary)",
+  "_cuLinkAddData_v2": "CUresult (CUlinkState, CUjitInputType, void *, size_t, char const *, unsigned int, CUjit_option *, void **)",
+  "_cuLinkAddFile_v2": "CUresult (CUlinkState, CUjitInputType, char const *, unsigned int, CUjit_option *, void **)",
+  "_cuLinkComplete": "CUresult (CUlinkState, void **, size_t *)",
+  "_cuLinkCreate_v2": "CUresult (unsigned int, CUjit_option *, void **, CUlinkState *)",
+  "_cuLinkDestroy": "CUresult (CUlinkState)",
+  "_cuLogsCurrent": "CUresult (CUlogIterator *, unsigned int)",
+  "_cuLogsDumpToFile": "CUresult (CUlogIterator *, char const *, unsigned int)",
+  "_cuLogsDumpToMemory": "CUresult (CUlogIterator *, char *, size_t *, unsigned int)",
+  "_cuLogsRegisterCallback": "CUresult (CUlogsCallback, void *, CUlogsCallbackHandle *)",
+  "_cuLogsUnregisterCallback": "CUresult (CUlogsCallbackHandle)",
+  "_cuMemAddressFree": "CUresult (CUdeviceptr, size_t)",
+  "_cuMemAddressReserve": "CUresult (CUdeviceptr *, size_t, size_t, CUdeviceptr, unsigned PY_LONG_LONG)",
+  "_cuMemAdvise_v2": "CUresult (CUdeviceptr, size_t, CUmem_advise, CUmemLocation)",
+  "_cuMemAllocAsync": "CUresult (CUdeviceptr *, size_t, CUstream)",
+  "_cuMemAllocFromPoolAsync": "CUresult (CUdeviceptr *, size_t, CUmemoryPool, CUstream)",
+  "_cuMemAllocHost_v2": "CUresult (void **, size_t)",
+  "_cuMemAllocManaged": "CUresult (CUdeviceptr *, size_t, unsigned int)",
+  "_cuMemAllocPitch_v2": "CUresult (CUdeviceptr *, size_t *, size_t, size_t, unsigned int)",
+  "_cuMemAlloc_v2": "CUresult (CUdeviceptr *, size_t)",
+  "_cuMemBatchDecompressAsync": "CUresult (CUmemDecompressParams *, size_t, unsigned int, size_t *, CUstream)",
+  "_cuMemCreate": "CUresult (CUmemGenericAllocationHandle *, size_t, CUmemAllocationProp const *, unsigned PY_LONG_LONG)",
+  "_cuMemDiscardAndPrefetchBatchAsync": "CUresult (CUdeviceptr *, size_t *, size_t, CUmemLocation *, size_t *, size_t, unsigned PY_LONG_LONG, CUstream)",
+  "_cuMemDiscardBatchAsync": "CUresult (CUdeviceptr *, size_t *, size_t, unsigned PY_LONG_LONG, CUstream)",
+  "_cuMemExportToShareableHandle": "CUresult (void *, CUmemGenericAllocationHandle, CUmemAllocationHandleType, unsigned PY_LONG_LONG)",
+  "_cuMemFreeAsync": "CUresult (CUdeviceptr, CUstream)",
+  "_cuMemFreeHost": "CUresult (void *)",
+  "_cuMemFree_v2": "CUresult (CUdeviceptr)",
+  "_cuMemGetAccess": "CUresult (unsigned PY_LONG_LONG *, CUmemLocation const *, CUdeviceptr)",
+  "_cuMemGetAddressRange_v2": "CUresult (CUdeviceptr *, size_t *, CUdeviceptr)",
+  "_cuMemGetAllocationGranularity": "CUresult (size_t *, CUmemAllocationProp const *, CUmemAllocationGranularity_flags)",
+  "_cuMemGetAllocationPropertiesFromHandle": "CUresult (CUmemAllocationProp *, CUmemGenericAllocationHandle)",
+  "_cuMemGetDefaultMemPool": "CUresult (CUmemoryPool *, CUmemLocation *, CUmemAllocationType)",
+  "_cuMemGetHandleForAddressRange": "CUresult (void *, CUdeviceptr, size_t, CUmemRangeHandleType, unsigned PY_LONG_LONG)",
+  "_cuMemGetInfo_v2": "CUresult (size_t *, size_t *)",
+  "_cuMemGetMemPool": "CUresult (CUmemoryPool *, CUmemLocation *, CUmemAllocationType)",
+  "_cuMemHostAlloc": "CUresult (void **, size_t, unsigned int)",
+  "_cuMemHostGetDevicePointer_v2": "CUresult (CUdeviceptr *, void *, unsigned int)",
+  "_cuMemHostGetFlags": "CUresult (unsigned int *, void *)",
+  "_cuMemHostRegister_v2": "CUresult (void *, size_t, unsigned int)",
+  "_cuMemHostUnregister": "CUresult (void *)",
+  "_cuMemImportFromShareableHandle": "CUresult (CUmemGenericAllocationHandle *, void *, CUmemAllocationHandleType)",
+  "_cuMemMap": "CUresult (CUdeviceptr, size_t, size_t, CUmemGenericAllocationHandle, unsigned PY_LONG_LONG)",
+  "_cuMemMapArrayAsync": "CUresult (CUarrayMapInfo *, unsigned int, CUstream)",
+  "_cuMemPoolCreate": "CUresult (CUmemoryPool *, CUmemPoolProps const *)",
+  "_cuMemPoolDestroy": "CUresult (CUmemoryPool)",
+  "_cuMemPoolExportPointer": "CUresult (CUmemPoolPtrExportData *, CUdeviceptr)",
+  "_cuMemPoolExportToShareableHandle": "CUresult (void *, CUmemoryPool, CUmemAllocationHandleType, unsigned PY_LONG_LONG)",
+  "_cuMemPoolGetAccess": "CUresult (CUmemAccess_flags *, CUmemoryPool, CUmemLocation *)",
+  "_cuMemPoolGetAttribute": "CUresult (CUmemoryPool, CUmemPool_attribute, void *)",
+  "_cuMemPoolImportFromShareableHandle": "CUresult (CUmemoryPool *, void *, CUmemAllocationHandleType, unsigned PY_LONG_LONG)",
+  "_cuMemPoolImportPointer": "CUresult (CUdeviceptr *, CUmemoryPool, CUmemPoolPtrExportData *)",
+  "_cuMemPoolSetAccess": "CUresult (CUmemoryPool, CUmemAccessDesc const *, size_t)",
+  "_cuMemPoolSetAttribute": "CUresult (CUmemoryPool, CUmemPool_attribute, void *)",
+  "_cuMemPoolTrimTo": "CUresult (CUmemoryPool, size_t)",
+  "_cuMemPrefetchAsync_v2": "CUresult (CUdeviceptr, size_t, CUmemLocation, unsigned int, CUstream)",
+  "_cuMemPrefetchBatchAsync": "CUresult (CUdeviceptr *, size_t *, size_t, CUmemLocation *, size_t *, size_t, unsigned PY_LONG_LONG, CUstream)",
+  "_cuMemRangeGetAttribute": "CUresult (void *, size_t, CUmem_range_attribute, CUdeviceptr, size_t)",
+  "_cuMemRangeGetAttributes": "CUresult (void **, size_t *, CUmem_range_attribute *, size_t, CUdeviceptr, size_t)",
+  "_cuMemRelease": "CUresult (CUmemGenericAllocationHandle)",
+  "_cuMemRetainAllocationHandle": "CUresult (CUmemGenericAllocationHandle *, void *)",
+  "_cuMemSetAccess": "CUresult (CUdeviceptr, size_t, CUmemAccessDesc const *, size_t)",
+  "_cuMemSetMemPool": "CUresult (CUmemLocation *, CUmemAllocationType, CUmemoryPool)",
+  "_cuMemUnmap": "CUresult (CUdeviceptr, size_t)",
+  "_cuMemcpy": "CUresult (CUdeviceptr, CUdeviceptr, size_t)",
+  "_cuMemcpy2DAsync_v2": "CUresult (CUDA_MEMCPY2D const *, CUstream)",
+  "_cuMemcpy2DUnaligned_v2": "CUresult (CUDA_MEMCPY2D const *)",
+  "_cuMemcpy2D_v2": "CUresult (CUDA_MEMCPY2D const *)",
+  "_cuMemcpy3DAsync_v2": "CUresult (CUDA_MEMCPY3D const *, CUstream)",
+  "_cuMemcpy3DBatchAsync_v2": "CUresult (size_t, CUDA_MEMCPY3D_BATCH_OP *, unsigned PY_LONG_LONG, CUstream)",
+  "_cuMemcpy3DPeer": "CUresult (CUDA_MEMCPY3D_PEER const *)",
+  "_cuMemcpy3DPeerAsync": "CUresult (CUDA_MEMCPY3D_PEER const *, CUstream)",
+  "_cuMemcpy3D_v2": "CUresult (CUDA_MEMCPY3D const *)",
+  "_cuMemcpyAsync": "CUresult (CUdeviceptr, CUdeviceptr, size_t, CUstream)",
+  "_cuMemcpyAtoA_v2": "CUresult (CUarray, size_t, CUarray, size_t, size_t)",
+  "_cuMemcpyAtoD_v2": "CUresult (CUdeviceptr, CUarray, size_t, size_t)",
+  "_cuMemcpyAtoHAsync_v2": "CUresult (void *, CUarray, size_t, size_t, CUstream)",
+  "_cuMemcpyAtoH_v2": "CUresult (void *, CUarray, size_t, size_t)",
+  "_cuMemcpyBatchAsync_v2": "CUresult (CUdeviceptr *, CUdeviceptr *, size_t *, size_t, CUmemcpyAttributes *, size_t *, size_t, CUstream)",
+  "_cuMemcpyDtoA_v2": "CUresult (CUarray, size_t, CUdeviceptr, size_t)",
+  "_cuMemcpyDtoDAsync_v2": "CUresult (CUdeviceptr, CUdeviceptr, size_t, CUstream)",
+  "_cuMemcpyDtoD_v2": "CUresult (CUdeviceptr, CUdeviceptr, size_t)",
+  "_cuMemcpyDtoHAsync_v2": "CUresult (void *, CUdeviceptr, size_t, CUstream)",
+  "_cuMemcpyDtoH_v2": "CUresult (void *, CUdeviceptr, size_t)",
+  "_cuMemcpyHtoAAsync_v2": "CUresult (CUarray, size_t, void const *, size_t, CUstream)",
+  "_cuMemcpyHtoA_v2": "CUresult (CUarray, size_t, void const *, size_t)",
+  "_cuMemcpyHtoDAsync_v2": "CUresult (CUdeviceptr, void const *, size_t, CUstream)",
+  "_cuMemcpyHtoD_v2": "CUresult (CUdeviceptr, void const *, size_t)",
+  "_cuMemcpyPeer": "CUresult (CUdeviceptr, CUcontext, CUdeviceptr, CUcontext, size_t)",
+  "_cuMemcpyPeerAsync": "CUresult (CUdeviceptr, CUcontext, CUdeviceptr, CUcontext, size_t, CUstream)",
+  "_cuMemsetD16Async": "CUresult (CUdeviceptr, unsigned short, size_t, CUstream)",
+  "_cuMemsetD16_v2": "CUresult (CUdeviceptr, unsigned short, size_t)",
+  "_cuMemsetD2D16Async": "CUresult (CUdeviceptr, size_t, unsigned short, size_t, size_t, CUstream)",
+  "_cuMemsetD2D16_v2": "CUresult (CUdeviceptr, size_t, unsigned short, size_t, size_t)",
+  "_cuMemsetD2D32Async": "CUresult (CUdeviceptr, size_t, unsigned int, size_t, size_t, CUstream)",
+  "_cuMemsetD2D32_v2": "CUresult (CUdeviceptr, size_t, unsigned int, size_t, size_t)",
+  "_cuMemsetD2D8Async": "CUresult (CUdeviceptr, size_t, unsigned char, size_t, size_t, CUstream)",
+  "_cuMemsetD2D8_v2": "CUresult (CUdeviceptr, size_t, unsigned char, size_t, size_t)",
+  "_cuMemsetD32Async": "CUresult (CUdeviceptr, unsigned int, size_t, CUstream)",
+  "_cuMemsetD32_v2": "CUresult (CUdeviceptr, unsigned int, size_t)",
+  "_cuMemsetD8Async": "CUresult (CUdeviceptr, unsigned char, size_t, CUstream)",
+  "_cuMemsetD8_v2": "CUresult (CUdeviceptr, unsigned char, size_t)",
+  "_cuMipmappedArrayCreate": "CUresult (CUmipmappedArray *, CUDA_ARRAY3D_DESCRIPTOR const *, unsigned int)",
+  "_cuMipmappedArrayDestroy": "CUresult (CUmipmappedArray)",
+  "_cuMipmappedArrayGetLevel": "CUresult (CUarray *, CUmipmappedArray, unsigned int)",
+  "_cuMipmappedArrayGetMemoryRequirements": "CUresult (CUDA_ARRAY_MEMORY_REQUIREMENTS *, CUmipmappedArray, CUdevice)",
+  "_cuMipmappedArrayGetSparseProperties": "CUresult (CUDA_ARRAY_SPARSE_PROPERTIES *, CUmipmappedArray)",
+  "_cuModuleEnumerateFunctions": "CUresult (CUfunction *, unsigned int, CUmodule)",
+  "_cuModuleGetFunction": "CUresult (CUfunction *, CUmodule, char const *)",
+  "_cuModuleGetFunctionCount": "CUresult (unsigned int *, CUmodule)",
+  "_cuModuleGetGlobal_v2": "CUresult (CUdeviceptr *, size_t *, CUmodule, char const *)",
+  "_cuModuleGetLoadingMode": "CUresult (CUmoduleLoadingMode *)",
+  "_cuModuleGetSurfRef": "CUresult (CUsurfref *, CUmodule, char const *)",
+  "_cuModuleGetTexRef": "CUresult (CUtexref *, CUmodule, char const *)",
+  "_cuModuleLoad": "CUresult (CUmodule *, char const *)",
+  "_cuModuleLoadData": "CUresult (CUmodule *, void const *)",
+  "_cuModuleLoadDataEx": "CUresult (CUmodule *, void const *, unsigned int, CUjit_option *, void **)",
+  "_cuModuleLoadFatBinary": "CUresult (CUmodule *, void const *)",
+  "_cuModuleUnload": "CUresult (CUmodule)",
+  "_cuMulticastAddDevice": "CUresult (CUmemGenericAllocationHandle, CUdevice)",
+  "_cuMulticastBindAddr": "CUresult (CUmemGenericAllocationHandle, size_t, CUdeviceptr, size_t, unsigned PY_LONG_LONG)",
+  "_cuMulticastBindMem": "CUresult (CUmemGenericAllocationHandle, size_t, CUmemGenericAllocationHandle, size_t, size_t, unsigned PY_LONG_LONG)",
+  "_cuMulticastCreate": "CUresult (CUmemGenericAllocationHandle *, CUmulticastObjectProp const *)",
+  "_cuMulticastGetGranularity": "CUresult (size_t *, CUmulticastObjectProp const *, CUmulticastGranularity_flags)",
+  "_cuMulticastUnbind": "CUresult (CUmemGenericAllocationHandle, CUdevice, size_t, size_t)",
+  "_cuOccupancyAvailableDynamicSMemPerBlock": "CUresult (size_t *, CUfunction, int, int)",
+  "_cuOccupancyMaxActiveBlocksPerMultiprocessor": "CUresult (int *, CUfunction, int, size_t)",
+  "_cuOccupancyMaxActiveBlocksPerMultiprocessorWithFlags": "CUresult (int *, CUfunction, int, size_t, unsigned int)",
+  "_cuOccupancyMaxActiveClusters": "CUresult (int *, CUfunction, CUlaunchConfig const *)",
+  "_cuOccupancyMaxPotentialBlockSize": "CUresult (int *, int *, CUfunction, CUoccupancyB2DSize, size_t, int)",
+  "_cuOccupancyMaxPotentialBlockSizeWithFlags": "CUresult (int *, int *, CUfunction, CUoccupancyB2DSize, size_t, int, unsigned int)",
+  "_cuOccupancyMaxPotentialClusterSize": "CUresult (int *, CUfunction, CUlaunchConfig const *)",
+  "_cuParamSetSize": "CUresult (CUfunction, unsigned int)",
+  "_cuParamSetTexRef": "CUresult (CUfunction, int, CUtexref)",
+  "_cuParamSetf": "CUresult (CUfunction, int, float)",
+  "_cuParamSeti": "CUresult (CUfunction, int, unsigned int)",
+  "_cuParamSetv": "CUresult (CUfunction, int, void *, unsigned int)",
+  "_cuPointerGetAttribute": "CUresult (void *, CUpointer_attribute, CUdeviceptr)",
+  "_cuPointerGetAttributes": "CUresult (unsigned int, CUpointer_attribute *, void **, CUdeviceptr)",
+  "_cuPointerSetAttribute": "CUresult (void const *, CUpointer_attribute, CUdeviceptr)",
+  "_cuProfilerStart": "CUresult (void)",
+  "_cuProfilerStop": "CUresult (void)",
+  "_cuSignalExternalSemaphoresAsync": "CUresult (CUexternalSemaphore const *, CUDA_EXTERNAL_SEMAPHORE_SIGNAL_PARAMS const *, unsigned int, CUstream)",
+  "_cuStreamAddCallback": "CUresult (CUstream, CUstreamCallback, void *, unsigned int)",
+  "_cuStreamAttachMemAsync": "CUresult (CUstream, CUdeviceptr, size_t, unsigned int)",
+  "_cuStreamBatchMemOp_v2": "CUresult (CUstream, unsigned int, CUstreamBatchMemOpParams *, unsigned int)",
+  "_cuStreamBeginCaptureToGraph": "CUresult (CUstream, CUgraph, CUgraphNode const *, CUgraphEdgeData const *, size_t, CUstreamCaptureMode)",
+  "_cuStreamBeginCapture_v2": "CUresult (CUstream, CUstreamCaptureMode)",
+  "_cuStreamCopyAttributes": "CUresult (CUstream, CUstream)",
+  "_cuStreamCreate": "CUresult (CUstream *, unsigned int)",
+  "_cuStreamCreateWithPriority": "CUresult (CUstream *, unsigned int, int)",
+  "_cuStreamDestroy_v2": "CUresult (CUstream)",
+  "_cuStreamEndCapture": "CUresult (CUstream, CUgraph *)",
+  "_cuStreamGetAttribute": "CUresult (CUstream, CUstreamAttrID, CUstreamAttrValue *)",
+  "_cuStreamGetCaptureInfo_v3": "CUresult (CUstream, CUstreamCaptureStatus *, cuuint64_t *, CUgraph *, CUgraphNode const **, CUgraphEdgeData const **, size_t *)",
+  "_cuStreamGetCtx": "CUresult (CUstream, CUcontext *)",
+  "_cuStreamGetCtx_v2": "CUresult (CUstream, CUcontext *, CUgreenCtx *)",
+  "_cuStreamGetDevice": "CUresult (CUstream, CUdevice *)",
+  "_cuStreamGetFlags": "CUresult (CUstream, unsigned int *)",
+  "_cuStreamGetGreenCtx": "CUresult (CUstream, CUgreenCtx *)",
+  "_cuStreamGetId": "CUresult (CUstream, unsigned PY_LONG_LONG *)",
+  "_cuStreamGetPriority": "CUresult (CUstream, int *)",
+  "_cuStreamIsCapturing": "CUresult (CUstream, CUstreamCaptureStatus *)",
+  "_cuStreamQuery": "CUresult (CUstream)",
+  "_cuStreamSetAttribute": "CUresult (CUstream, CUstreamAttrID, CUstreamAttrValue const *)",
+  "_cuStreamSynchronize": "CUresult (CUstream)",
+  "_cuStreamUpdateCaptureDependencies_v2": "CUresult (CUstream, CUgraphNode *, CUgraphEdgeData const *, size_t, unsigned int)",
+  "_cuStreamWaitEvent": "CUresult (CUstream, CUevent, unsigned int)",
+  "_cuStreamWaitValue32_v2": "CUresult (CUstream, CUdeviceptr, cuuint32_t, unsigned int)",
+  "_cuStreamWaitValue64_v2": "CUresult (CUstream, CUdeviceptr, cuuint64_t, unsigned int)",
+  "_cuStreamWriteValue32_v2": "CUresult (CUstream, CUdeviceptr, cuuint32_t, unsigned int)",
+  "_cuStreamWriteValue64_v2": "CUresult (CUstream, CUdeviceptr, cuuint64_t, unsigned int)",
+  "_cuSurfObjectCreate": "CUresult (CUsurfObject *, CUDA_RESOURCE_DESC const *)",
+  "_cuSurfObjectDestroy": "CUresult (CUsurfObject)",
+  "_cuSurfObjectGetResourceDesc": "CUresult (CUDA_RESOURCE_DESC *, CUsurfObject)",
+  "_cuSurfRefGetArray": "CUresult (CUarray *, CUsurfref)",
+  "_cuSurfRefSetArray": "CUresult (CUsurfref, CUarray, unsigned int)",
+  "_cuTensorMapEncodeIm2col": "CUresult (CUtensorMap *, CUtensorMapDataType, cuuint32_t, void *, cuuint64_t const *, cuuint64_t const *, int const *, int const *, cuuint32_t, cuuint32_t, cuuint32_t const *, CUtensorMapInterleave, CUtensorMapSwizzle, CUtensorMapL2promotion, CUtensorMapFloatOOBfill)",
+  "_cuTensorMapEncodeIm2colWide": "CUresult (CUtensorMap *, CUtensorMapDataType, cuuint32_t, void *, cuuint64_t const *, cuuint64_t const *, int, int, cuuint32_t, cuuint32_t, cuuint32_t const *, CUtensorMapInterleave, CUtensorMapIm2ColWideMode, CUtensorMapSwizzle, CUtensorMapL2promotion, CUtensorMapFloatOOBfill)",
+  "_cuTensorMapEncodeTiled": "CUresult (CUtensorMap *, CUtensorMapDataType, cuuint32_t, void *, cuuint64_t const *, cuuint64_t const *, cuuint32_t const *, cuuint32_t const *, CUtensorMapInterleave, CUtensorMapSwizzle, CUtensorMapL2promotion, CUtensorMapFloatOOBfill)",
+  "_cuTensorMapReplaceAddress": "CUresult (CUtensorMap *, void *)",
+  "_cuTexObjectCreate": "CUresult (CUtexObject *, CUDA_RESOURCE_DESC const *, CUDA_TEXTURE_DESC const *, CUDA_RESOURCE_VIEW_DESC const *)",
+  "_cuTexObjectDestroy": "CUresult (CUtexObject)",
+  "_cuTexObjectGetResourceDesc": "CUresult (CUDA_RESOURCE_DESC *, CUtexObject)",
+  "_cuTexObjectGetResourceViewDesc": "CUresult (CUDA_RESOURCE_VIEW_DESC *, CUtexObject)",
+  "_cuTexObjectGetTextureDesc": "CUresult (CUDA_TEXTURE_DESC *, CUtexObject)",
+  "_cuTexRefCreate": "CUresult (CUtexref *)",
+  "_cuTexRefDestroy": "CUresult (CUtexref)",
+  "_cuTexRefGetAddressMode": "CUresult (CUaddress_mode *, CUtexref, int)",
+  "_cuTexRefGetAddress_v2": "CUresult (CUdeviceptr *, CUtexref)",
+  "_cuTexRefGetArray": "CUresult (CUarray *, CUtexref)",
+  "_cuTexRefGetBorderColor": "CUresult (float *, CUtexref)",
+  "_cuTexRefGetFilterMode": "CUresult (CUfilter_mode *, CUtexref)",
+  "_cuTexRefGetFlags": "CUresult (unsigned int *, CUtexref)",
+  "_cuTexRefGetFormat": "CUresult (CUarray_format *, int *, CUtexref)",
+  "_cuTexRefGetMaxAnisotropy": "CUresult (int *, CUtexref)",
+  "_cuTexRefGetMipmapFilterMode": "CUresult (CUfilter_mode *, CUtexref)",
+  "_cuTexRefGetMipmapLevelBias": "CUresult (float *, CUtexref)",
+  "_cuTexRefGetMipmapLevelClamp": "CUresult (float *, float *, CUtexref)",
+  "_cuTexRefGetMipmappedArray": "CUresult (CUmipmappedArray *, CUtexref)",
+  "_cuTexRefSetAddress2D_v3": "CUresult (CUtexref, CUDA_ARRAY_DESCRIPTOR const *, CUdeviceptr, size_t)",
+  "_cuTexRefSetAddressMode": "CUresult (CUtexref, int, CUaddress_mode)",
+  "_cuTexRefSetAddress_v2": "CUresult (size_t *, CUtexref, CUdeviceptr, size_t)",
+  "_cuTexRefSetArray": "CUresult (CUtexref, CUarray, unsigned int)",
+  "_cuTexRefSetBorderColor": "CUresult (CUtexref, float *)",
+  "_cuTexRefSetFilterMode": "CUresult (CUtexref, CUfilter_mode)",
+  "_cuTexRefSetFlags": "CUresult (CUtexref, unsigned int)",
+  "_cuTexRefSetFormat": "CUresult (CUtexref, CUarray_format, int)",
+  "_cuTexRefSetMaxAnisotropy": "CUresult (CUtexref, unsigned int)",
+  "_cuTexRefSetMipmapFilterMode": "CUresult (CUtexref, CUfilter_mode)",
+  "_cuTexRefSetMipmapLevelBias": "CUresult (CUtexref, float)",
+  "_cuTexRefSetMipmapLevelClamp": "CUresult (CUtexref, float, float)",
+  "_cuTexRefSetMipmappedArray": "CUresult (CUtexref, CUmipmappedArray, unsigned int)",
+  "_cuThreadExchangeStreamCaptureMode": "CUresult (CUstreamCaptureMode *)",
+  "_cuUserObjectCreate": "CUresult (CUuserObject *, void *, CUhostFn, unsigned int, unsigned int)",
+  "_cuUserObjectRelease": "CUresult (CUuserObject, unsigned int)",
+  "_cuUserObjectRetain": "CUresult (CUuserObject, unsigned int)",
+  "_cuVDPAUCtxCreate_v2": "CUresult (CUcontext *, unsigned int, CUdevice, __pyx_t_4cuda_8bindings_8cydriver_VdpDevice, __pyx_t_4cuda_8bindings_8cydriver_VdpGetProcAddress *)",
+  "_cuVDPAUGetDevice": "CUresult (CUdevice *, __pyx_t_4cuda_8bindings_8cydriver_VdpDevice, __pyx_t_4cuda_8bindings_8cydriver_VdpGetProcAddress *)",
+  "_cuWaitExternalSemaphoresAsync": "CUresult (CUexternalSemaphore const *, CUDA_EXTERNAL_SEMAPHORE_WAIT_PARAMS const *, unsigned int, CUstream)"
+}
diff --git a/cuda_bindings/tests/abi/cuda/bindings/_bindings/cynvrtc.abi.json b/cuda_bindings/tests/abi/cuda/bindings/_bindings/cynvrtc.abi.json
new file mode 100644
index 000000000..801f41ddb
--- /dev/null
+++ b/cuda_bindings/tests/abi/cuda/bindings/_bindings/cynvrtc.abi.json
@@ -0,0 +1,26 @@
+{
+  "_nvrtcAddNameExpression": "nvrtcResult (nvrtcProgram, char const *)",
+  "_nvrtcCompileProgram": "nvrtcResult (nvrtcProgram, int, char const **)",
+  "_nvrtcCreateProgram": "nvrtcResult (nvrtcProgram *, char const *, char const *, int, char const **, char const **)",
+  "_nvrtcDestroyProgram": "nvrtcResult (nvrtcProgram *)",
+  "_nvrtcGetCUBIN": "nvrtcResult (nvrtcProgram, char *)",
+  "_nvrtcGetCUBINSize": "nvrtcResult (nvrtcProgram, size_t *)",
+  "_nvrtcGetErrorString": "char const *(nvrtcResult)",
+  "_nvrtcGetLTOIR": "nvrtcResult (nvrtcProgram, char *)",
+  "_nvrtcGetLTOIRSize": "nvrtcResult (nvrtcProgram, size_t *)",
+  "_nvrtcGetLoweredName": "nvrtcResult (nvrtcProgram, char const *, char const **)",
+  "_nvrtcGetNumSupportedArchs": "nvrtcResult (int *)",
+  "_nvrtcGetOptiXIR": "nvrtcResult (nvrtcProgram, char *)",
+  "_nvrtcGetOptiXIRSize": "nvrtcResult (nvrtcProgram, size_t *)",
+  "_nvrtcGetPCHCreateStatus": "nvrtcResult (nvrtcProgram)",
+  "_nvrtcGetPCHHeapSize": "nvrtcResult (size_t *)",
+  "_nvrtcGetPCHHeapSizeRequired": "nvrtcResult (nvrtcProgram, size_t *)",
+  "_nvrtcGetPTX": "nvrtcResult (nvrtcProgram, char *)",
+  "_nvrtcGetPTXSize": "nvrtcResult (nvrtcProgram, size_t *)",
+  "_nvrtcGetProgramLog": "nvrtcResult (nvrtcProgram, char *)",
+  "_nvrtcGetProgramLogSize": "nvrtcResult (nvrtcProgram, size_t *)",
+  "_nvrtcGetSupportedArchs": "nvrtcResult (int *)",
+  "_nvrtcSetFlowCallback": "nvrtcResult (nvrtcProgram, void *, void *)",
+  "_nvrtcSetPCHHeapSize": "nvrtcResult (size_t)",
+  "_nvrtcVersion": "nvrtcResult (int *, int *)"
+}
diff --git a/cuda_bindings/tests/abi/cuda/bindings/_bindings/cyruntime.abi.json b/cuda_bindings/tests/abi/cuda/bindings/_bindings/cyruntime.abi.json
new file mode 100644
index 000000000..39aa55db1
--- /dev/null
+++ b/cuda_bindings/tests/abi/cuda/bindings/_bindings/cyruntime.abi.json
@@ -0,0 +1,317 @@
+{
+  "_cudaArrayGetInfo": "cudaError_t (struct cudaChannelFormatDesc *, struct cudaExtent *, unsigned int *, cudaArray_t)",
+  "_cudaArrayGetMemoryRequirements": "cudaError_t (struct cudaArrayMemoryRequirements *, cudaArray_t, int)",
+  "_cudaArrayGetPlane": "cudaError_t (cudaArray_t *, cudaArray_t, unsigned int)",
+  "_cudaArrayGetSparseProperties": "cudaError_t (struct cudaArraySparseProperties *, cudaArray_t)",
+  "_cudaChooseDevice": "cudaError_t (int *, struct cudaDeviceProp const *)",
+  "_cudaCreateChannelDesc": "struct cudaChannelFormatDesc (int, int, int, int, enum cudaChannelFormatKind)",
+  "_cudaCreateSurfaceObject": "cudaError_t (cudaSurfaceObject_t *, struct cudaResourceDesc const *)",
+  "_cudaCreateTextureObject": "cudaError_t (cudaTextureObject_t *, struct cudaResourceDesc const *, struct cudaTextureDesc const *, struct cudaResourceViewDesc const *)",
+  "_cudaCtxResetPersistingL2Cache": "cudaError_t (void)",
+  "_cudaDestroyExternalMemory": "cudaError_t (cudaExternalMemory_t)",
+  "_cudaDestroyExternalSemaphore": "cudaError_t (cudaExternalSemaphore_t)",
+  "_cudaDestroySurfaceObject": "cudaError_t (cudaSurfaceObject_t)",
+  "_cudaDestroyTextureObject": "cudaError_t (cudaTextureObject_t)",
+  "_cudaDeviceCanAccessPeer": "cudaError_t (int *, int, int)",
+  "_cudaDeviceDisablePeerAccess": "cudaError_t (int)",
+  "_cudaDeviceEnablePeerAccess": "cudaError_t (int, unsigned int)",
+  "_cudaDeviceFlushGPUDirectRDMAWrites": "cudaError_t (enum cudaFlushGPUDirectRDMAWritesTarget, enum cudaFlushGPUDirectRDMAWritesScope)",
+  "_cudaDeviceGetAttribute": "cudaError_t (int *, enum cudaDeviceAttr, int)",
+  "_cudaDeviceGetByPCIBusId": "cudaError_t (int *, char const *)",
+  "_cudaDeviceGetCacheConfig": "cudaError_t (enum cudaFuncCache *)",
+  "_cudaDeviceGetDefaultMemPool": "cudaError_t (cudaMemPool_t *, int)",
+  "_cudaDeviceGetGraphMemAttribute": "cudaError_t (int, enum cudaGraphMemAttributeType, void *)",
+  "_cudaDeviceGetHostAtomicCapabilities": "cudaError_t (unsigned int *, enum cudaAtomicOperation const *, unsigned int, int)",
+  "_cudaDeviceGetLimit": "cudaError_t (size_t *, enum cudaLimit)",
+  "_cudaDeviceGetMemPool": "cudaError_t (cudaMemPool_t *, int)",
+  "_cudaDeviceGetNvSciSyncAttributes": "cudaError_t (void *, int, int)",
+  "_cudaDeviceGetP2PAtomicCapabilities": "cudaError_t (unsigned int *, enum cudaAtomicOperation const *, unsigned int, int, int)",
+  "_cudaDeviceGetP2PAttribute": "cudaError_t (int *, enum cudaDeviceP2PAttr, int, int)",
+  "_cudaDeviceGetPCIBusId": "cudaError_t (char *, int, int)",
+  "_cudaDeviceGetSharedMemConfig": "cudaError_t (enum cudaSharedMemConfig *)",
+  "_cudaDeviceGetStreamPriorityRange": "cudaError_t (int *, int *)",
+  "_cudaDeviceGetTexture1DLinearMaxWidth": "cudaError_t (size_t *, struct cudaChannelFormatDesc const *, int)",
+  "_cudaDeviceGraphMemTrim": "cudaError_t (int)",
+  "_cudaDeviceRegisterAsyncNotification": "cudaError_t (int, cudaAsyncCallback, void *, cudaAsyncCallbackHandle_t *)",
+  "_cudaDeviceReset": "cudaError_t (void)",
+  "_cudaDeviceSetCacheConfig": "cudaError_t (enum cudaFuncCache)",
+  "_cudaDeviceSetGraphMemAttribute": "cudaError_t (int, enum cudaGraphMemAttributeType, void *)",
+  "_cudaDeviceSetLimit": "cudaError_t (enum cudaLimit, size_t)",
+  "_cudaDeviceSetMemPool": "cudaError_t (int, cudaMemPool_t)",
+  "_cudaDeviceSetSharedMemConfig": "cudaError_t (enum cudaSharedMemConfig)",
+  "_cudaDeviceSynchronize": "cudaError_t (void)",
+  "_cudaDeviceUnregisterAsyncNotification": "cudaError_t (int, cudaAsyncCallbackHandle_t)",
+  "_cudaDriverGetVersion": "cudaError_t (int *)",
+  "_cudaEGLStreamConsumerAcquireFrame": "cudaError_t (__pyx_t_4cuda_8bindings_9cyruntime_cudaEglStreamConnection *, cudaGraphicsResource_t *, cudaStream_t *, unsigned int)",
+  "_cudaEGLStreamConsumerConnect": "cudaError_t (__pyx_t_4cuda_8bindings_9cyruntime_cudaEglStreamConnection *, __pyx_t_4cuda_8bindings_9cyruntime_EGLStreamKHR)",
+  "_cudaEGLStreamConsumerConnectWithFlags": "cudaError_t (__pyx_t_4cuda_8bindings_9cyruntime_cudaEglStreamConnection *, __pyx_t_4cuda_8bindings_9cyruntime_EGLStreamKHR, unsigned int)",
+  "_cudaEGLStreamConsumerDisconnect": "cudaError_t (__pyx_t_4cuda_8bindings_9cyruntime_cudaEglStreamConnection *)",
+  "_cudaEGLStreamConsumerReleaseFrame": "cudaError_t (__pyx_t_4cuda_8bindings_9cyruntime_cudaEglStreamConnection *, cudaGraphicsResource_t, cudaStream_t *)",
+  "_cudaEGLStreamProducerConnect": "cudaError_t (__pyx_t_4cuda_8bindings_9cyruntime_cudaEglStreamConnection *, __pyx_t_4cuda_8bindings_9cyruntime_EGLStreamKHR, __pyx_t_4cuda_8bindings_9cyruntime_EGLint, __pyx_t_4cuda_8bindings_9cyruntime_EGLint)",
+  "_cudaEGLStreamProducerDisconnect": "cudaError_t (__pyx_t_4cuda_8bindings_9cyruntime_cudaEglStreamConnection *)",
+  "_cudaEGLStreamProducerPresentFrame": "cudaError_t (__pyx_t_4cuda_8bindings_9cyruntime_cudaEglStreamConnection *, __pyx_t_4cuda_8bindings_9cyruntime_cudaEglFrame, cudaStream_t *)",
+  "_cudaEGLStreamProducerReturnFrame": "cudaError_t (__pyx_t_4cuda_8bindings_9cyruntime_cudaEglStreamConnection *, __pyx_t_4cuda_8bindings_9cyruntime_cudaEglFrame *, cudaStream_t *)",
+  "_cudaEventCreate": "cudaError_t (cudaEvent_t *)",
+  "_cudaEventCreateFromEGLSync": "cudaError_t (cudaEvent_t *, __pyx_t_4cuda_8bindings_9cyruntime_EGLSyncKHR, unsigned int)",
+  "_cudaEventCreateWithFlags": "cudaError_t (cudaEvent_t *, unsigned int)",
+  "_cudaEventDestroy": "cudaError_t (cudaEvent_t)",
+  "_cudaEventElapsedTime": "cudaError_t (float *, cudaEvent_t, cudaEvent_t)",
+  "_cudaEventQuery": "cudaError_t (cudaEvent_t)",
+  "_cudaEventRecord": "cudaError_t (cudaEvent_t, cudaStream_t)",
+  "_cudaEventRecordWithFlags": "cudaError_t (cudaEvent_t, cudaStream_t, unsigned int)",
+  "_cudaEventSynchronize": "cudaError_t (cudaEvent_t)",
+  "_cudaExternalMemoryGetMappedBuffer": "cudaError_t (void **, cudaExternalMemory_t, struct cudaExternalMemoryBufferDesc const *)",
+  "_cudaExternalMemoryGetMappedMipmappedArray": "cudaError_t (cudaMipmappedArray_t *, cudaExternalMemory_t, struct cudaExternalMemoryMipmappedArrayDesc const *)",
+  "_cudaFree": "cudaError_t (void *)",
+  "_cudaFreeArray": "cudaError_t (cudaArray_t)",
+  "_cudaFreeAsync": "cudaError_t (void *, cudaStream_t)",
+  "_cudaFreeHost": "cudaError_t (void *)",
+  "_cudaFreeMipmappedArray": "cudaError_t (cudaMipmappedArray_t)",
+  "_cudaFuncGetAttributes": "cudaError_t (struct cudaFuncAttributes *, void const *)",
+  "_cudaFuncSetAttribute": "cudaError_t (void const *, enum cudaFuncAttribute, int)",
+  "_cudaFuncSetCacheConfig": "cudaError_t (void const *, enum cudaFuncCache)",
+  "_cudaFuncSetSharedMemConfig": "cudaError_t (void const *, enum cudaSharedMemConfig)",
+  "_cudaGLGetDevices": "cudaError_t (unsigned int *, int *, unsigned int, enum __pyx_t_4cuda_8bindings_9cyruntime_cudaGLDeviceList)",
+  "_cudaGetChannelDesc": "cudaError_t (struct cudaChannelFormatDesc *, cudaArray_const_t)",
+  "_cudaGetDevice": "cudaError_t (int *)",
+  "_cudaGetDeviceCount": "cudaError_t (int *)",
+  "_cudaGetDeviceFlags": "cudaError_t (unsigned int *)",
+  "_cudaGetDeviceProperties": "cudaError_t (struct cudaDeviceProp *, int)",
+  "_cudaGetDriverEntryPoint": "cudaError_t (char const *, void **, unsigned PY_LONG_LONG, enum cudaDriverEntryPointQueryResult *)",
+  "_cudaGetDriverEntryPointByVersion": "cudaError_t (char const *, void **, unsigned int, unsigned PY_LONG_LONG, enum cudaDriverEntryPointQueryResult *)",
+  "_cudaGetErrorName": "char const *(cudaError_t)",
+  "_cudaGetErrorString": "char const *(cudaError_t)",
+  "_cudaGetExportTable": "cudaError_t (void const **, cudaUUID_t const *)",
+  "_cudaGetKernel": "cudaError_t (cudaKernel_t *, void const *)",
+  "_cudaGetLastError": "cudaError_t (void)",
+  "_cudaGetMipmappedArrayLevel": "cudaError_t (cudaArray_t *, cudaMipmappedArray_const_t, unsigned int)",
+  "_cudaGetSurfaceObjectResourceDesc": "cudaError_t (struct cudaResourceDesc *, cudaSurfaceObject_t)",
+  "_cudaGetTextureObjectResourceDesc": "cudaError_t (struct cudaResourceDesc *, cudaTextureObject_t)",
+  "_cudaGetTextureObjectResourceViewDesc": "cudaError_t (struct cudaResourceViewDesc *, cudaTextureObject_t)",
+  "_cudaGetTextureObjectTextureDesc": "cudaError_t (struct cudaTextureDesc *, cudaTextureObject_t)",
+  "_cudaGraphAddChildGraphNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, cudaGraph_t)",
+  "_cudaGraphAddDependencies": "cudaError_t (cudaGraph_t, cudaGraphNode_t const *, cudaGraphNode_t const *, cudaGraphEdgeData const *, size_t)",
+  "_cudaGraphAddEmptyNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t)",
+  "_cudaGraphAddEventRecordNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, cudaEvent_t)",
+  "_cudaGraphAddEventWaitNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, cudaEvent_t)",
+  "_cudaGraphAddExternalSemaphoresSignalNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, struct cudaExternalSemaphoreSignalNodeParams const *)",
+  "_cudaGraphAddExternalSemaphoresWaitNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, struct cudaExternalSemaphoreWaitNodeParams const *)",
+  "_cudaGraphAddHostNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, struct cudaHostNodeParams const *)",
+  "_cudaGraphAddKernelNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, struct cudaKernelNodeParams const *)",
+  "_cudaGraphAddMemAllocNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, struct cudaMemAllocNodeParams *)",
+  "_cudaGraphAddMemFreeNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, void *)",
+  "_cudaGraphAddMemcpyNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, struct cudaMemcpy3DParms const *)",
+  "_cudaGraphAddMemcpyNode1D": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, void *, void const *, size_t, enum cudaMemcpyKind)",
+  "_cudaGraphAddMemsetNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, struct cudaMemsetParams const *)",
+  "_cudaGraphAddNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, cudaGraphEdgeData const *, size_t, struct cudaGraphNodeParams *)",
+  "_cudaGraphChildGraphNodeGetGraph": "cudaError_t (cudaGraphNode_t, cudaGraph_t *)",
+  "_cudaGraphClone": "cudaError_t (cudaGraph_t *, cudaGraph_t)",
+  "_cudaGraphConditionalHandleCreate": "cudaError_t (cudaGraphConditionalHandle *, cudaGraph_t, unsigned int, unsigned int)",
+  "_cudaGraphCreate": "cudaError_t (cudaGraph_t *, unsigned int)",
+  "_cudaGraphDebugDotPrint": "cudaError_t (cudaGraph_t, char const *, unsigned int)",
+  "_cudaGraphDestroy": "cudaError_t (cudaGraph_t)",
+  "_cudaGraphDestroyNode": "cudaError_t (cudaGraphNode_t)",
+  "_cudaGraphEventRecordNodeGetEvent": "cudaError_t (cudaGraphNode_t, cudaEvent_t *)",
+  "_cudaGraphEventRecordNodeSetEvent": "cudaError_t (cudaGraphNode_t, cudaEvent_t)",
+  "_cudaGraphEventWaitNodeGetEvent": "cudaError_t (cudaGraphNode_t, cudaEvent_t *)",
+  "_cudaGraphEventWaitNodeSetEvent": "cudaError_t (cudaGraphNode_t, cudaEvent_t)",
+  "_cudaGraphExecChildGraphNodeSetParams": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, cudaGraph_t)",
+  "_cudaGraphExecDestroy": "cudaError_t (cudaGraphExec_t)",
+  "_cudaGraphExecEventRecordNodeSetEvent": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, cudaEvent_t)",
+  "_cudaGraphExecEventWaitNodeSetEvent": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, cudaEvent_t)",
+  "_cudaGraphExecExternalSemaphoresSignalNodeSetParams": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, struct cudaExternalSemaphoreSignalNodeParams const *)",
+  "_cudaGraphExecExternalSemaphoresWaitNodeSetParams": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, struct cudaExternalSemaphoreWaitNodeParams const *)",
+  "_cudaGraphExecGetFlags": "cudaError_t (cudaGraphExec_t, unsigned PY_LONG_LONG *)",
+  "_cudaGraphExecHostNodeSetParams": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, struct cudaHostNodeParams const *)",
+  "_cudaGraphExecKernelNodeSetParams": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, struct cudaKernelNodeParams const *)",
+  "_cudaGraphExecMemcpyNodeSetParams": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, struct cudaMemcpy3DParms const *)",
+  "_cudaGraphExecMemcpyNodeSetParams1D": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, void *, void const *, size_t, enum cudaMemcpyKind)",
+  "_cudaGraphExecMemsetNodeSetParams": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, struct cudaMemsetParams const *)",
+  "_cudaGraphExecNodeSetParams": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, struct cudaGraphNodeParams *)",
+  "_cudaGraphExecUpdate": "cudaError_t (cudaGraphExec_t, cudaGraph_t, cudaGraphExecUpdateResultInfo *)",
+  "_cudaGraphExternalSemaphoresSignalNodeGetParams": "cudaError_t (cudaGraphNode_t, struct cudaExternalSemaphoreSignalNodeParams *)",
+  "_cudaGraphExternalSemaphoresSignalNodeSetParams": "cudaError_t (cudaGraphNode_t, struct cudaExternalSemaphoreSignalNodeParams const *)",
+  "_cudaGraphExternalSemaphoresWaitNodeGetParams": "cudaError_t (cudaGraphNode_t, struct cudaExternalSemaphoreWaitNodeParams *)",
+  "_cudaGraphExternalSemaphoresWaitNodeSetParams": "cudaError_t (cudaGraphNode_t, struct cudaExternalSemaphoreWaitNodeParams const *)",
+  "_cudaGraphGetEdges": "cudaError_t (cudaGraph_t, cudaGraphNode_t *, cudaGraphNode_t *, cudaGraphEdgeData *, size_t *)",
+  "_cudaGraphGetNodes": "cudaError_t (cudaGraph_t, cudaGraphNode_t *, size_t *)",
+  "_cudaGraphGetRootNodes": "cudaError_t (cudaGraph_t, cudaGraphNode_t *, size_t *)",
+  "_cudaGraphHostNodeGetParams": "cudaError_t (cudaGraphNode_t, struct cudaHostNodeParams *)",
+  "_cudaGraphHostNodeSetParams": "cudaError_t (cudaGraphNode_t, struct cudaHostNodeParams const *)",
+  "_cudaGraphInstantiate": "cudaError_t (cudaGraphExec_t *, cudaGraph_t, unsigned PY_LONG_LONG)",
+  "_cudaGraphInstantiateWithFlags": "cudaError_t (cudaGraphExec_t *, cudaGraph_t, unsigned PY_LONG_LONG)",
+  "_cudaGraphInstantiateWithParams": "cudaError_t (cudaGraphExec_t *, cudaGraph_t, cudaGraphInstantiateParams *)",
+  "_cudaGraphKernelNodeCopyAttributes": "cudaError_t (cudaGraphNode_t, cudaGraphNode_t)",
+  "_cudaGraphKernelNodeGetAttribute": "cudaError_t (cudaGraphNode_t, __pyx_t_4cuda_8bindings_9_bindings_9cyruntime_cudaKernelNodeAttrID, __pyx_t_4cuda_8bindings_9_bindings_9cyruntime_cudaKernelNodeAttrValue *)",
+  "_cudaGraphKernelNodeGetParams": "cudaError_t (cudaGraphNode_t, struct cudaKernelNodeParams *)",
+  "_cudaGraphKernelNodeSetAttribute": "cudaError_t (cudaGraphNode_t, __pyx_t_4cuda_8bindings_9_bindings_9cyruntime_cudaKernelNodeAttrID, __pyx_t_4cuda_8bindings_9_bindings_9cyruntime_cudaKernelNodeAttrValue const *)",
+  "_cudaGraphKernelNodeSetParams": "cudaError_t (cudaGraphNode_t, struct cudaKernelNodeParams const *)",
+  "_cudaGraphLaunch": "cudaError_t (cudaGraphExec_t, cudaStream_t)",
+  "_cudaGraphMemAllocNodeGetParams": "cudaError_t (cudaGraphNode_t, struct cudaMemAllocNodeParams *)",
+  "_cudaGraphMemFreeNodeGetParams": "cudaError_t (cudaGraphNode_t, void *)",
+  "_cudaGraphMemcpyNodeGetParams": "cudaError_t (cudaGraphNode_t, struct cudaMemcpy3DParms *)",
+  "_cudaGraphMemcpyNodeSetParams": "cudaError_t (cudaGraphNode_t, struct cudaMemcpy3DParms const *)",
+  "_cudaGraphMemcpyNodeSetParams1D": "cudaError_t (cudaGraphNode_t, void *, void const *, size_t, enum cudaMemcpyKind)",
+  "_cudaGraphMemsetNodeGetParams": "cudaError_t (cudaGraphNode_t, struct cudaMemsetParams *)",
+  "_cudaGraphMemsetNodeSetParams": "cudaError_t (cudaGraphNode_t, struct cudaMemsetParams const *)",
+  "_cudaGraphNodeFindInClone": "cudaError_t (cudaGraphNode_t *, cudaGraphNode_t, cudaGraph_t)",
+  "_cudaGraphNodeGetDependencies": "cudaError_t (cudaGraphNode_t, cudaGraphNode_t *, cudaGraphEdgeData *, size_t *)",
+  "_cudaGraphNodeGetDependentNodes": "cudaError_t (cudaGraphNode_t, cudaGraphNode_t *, cudaGraphEdgeData *, size_t *)",
+  "_cudaGraphNodeGetEnabled": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, unsigned int *)",
+  "_cudaGraphNodeGetType": "cudaError_t (cudaGraphNode_t, enum cudaGraphNodeType *)",
+  "_cudaGraphNodeSetEnabled": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, unsigned int)",
+  "_cudaGraphNodeSetParams": "cudaError_t (cudaGraphNode_t, struct cudaGraphNodeParams *)",
+  "_cudaGraphReleaseUserObject": "cudaError_t (cudaGraph_t, cudaUserObject_t, unsigned int)",
+  "_cudaGraphRemoveDependencies": "cudaError_t (cudaGraph_t, cudaGraphNode_t const *, cudaGraphNode_t const *, cudaGraphEdgeData const *, size_t)",
+  "_cudaGraphRetainUserObject": "cudaError_t (cudaGraph_t, cudaUserObject_t, unsigned int, unsigned int)",
+  "_cudaGraphUpload": "cudaError_t (cudaGraphExec_t, cudaStream_t)",
+  "_cudaGraphicsEGLRegisterImage": "cudaError_t (cudaGraphicsResource_t *, __pyx_t_4cuda_8bindings_9cyruntime_EGLImageKHR, unsigned int)",
+  "_cudaGraphicsGLRegisterBuffer": "cudaError_t (struct cudaGraphicsResource **, __pyx_t_4cuda_8bindings_9cyruntime_GLuint, unsigned int)",
+  "_cudaGraphicsGLRegisterImage": "cudaError_t (struct cudaGraphicsResource **, __pyx_t_4cuda_8bindings_9cyruntime_GLuint, __pyx_t_4cuda_8bindings_9cyruntime_GLenum, unsigned int)",
+  "_cudaGraphicsMapResources": "cudaError_t (int, cudaGraphicsResource_t *, cudaStream_t)",
+  "_cudaGraphicsResourceGetMappedEglFrame": "cudaError_t (__pyx_t_4cuda_8bindings_9cyruntime_cudaEglFrame *, cudaGraphicsResource_t, unsigned int, unsigned int)",
+  "_cudaGraphicsResourceGetMappedMipmappedArray": "cudaError_t (cudaMipmappedArray_t *, cudaGraphicsResource_t)",
+  "_cudaGraphicsResourceGetMappedPointer": "cudaError_t (void **, size_t *, cudaGraphicsResource_t)",
+  "_cudaGraphicsResourceSetMapFlags": "cudaError_t (cudaGraphicsResource_t, unsigned int)",
+  "_cudaGraphicsSubResourceGetMappedArray": "cudaError_t (cudaArray_t *, cudaGraphicsResource_t, unsigned int, unsigned int)",
+  "_cudaGraphicsUnmapResources": "cudaError_t (int, cudaGraphicsResource_t *, cudaStream_t)",
+  "_cudaGraphicsUnregisterResource": "cudaError_t (cudaGraphicsResource_t)",
+  "_cudaGraphicsVDPAURegisterOutputSurface": "cudaError_t (struct cudaGraphicsResource **, __pyx_t_4cuda_8bindings_9cyruntime_VdpOutputSurface, unsigned int)",
+  "_cudaGraphicsVDPAURegisterVideoSurface": "cudaError_t (struct cudaGraphicsResource **, __pyx_t_4cuda_8bindings_9cyruntime_VdpVideoSurface, unsigned int)",
+  "_cudaHostAlloc": "cudaError_t (void **, size_t, unsigned int)",
+  "_cudaHostGetDevicePointer": "cudaError_t (void **, void *, unsigned int)",
+  "_cudaHostGetFlags": "cudaError_t (unsigned int *, void *)",
+  "_cudaHostRegister": "cudaError_t (void *, size_t, unsigned int)",
+  "_cudaHostUnregister": "cudaError_t (void *)",
+  "_cudaImportExternalMemory": "cudaError_t (cudaExternalMemory_t *, struct cudaExternalMemoryHandleDesc const *)",
+  "_cudaImportExternalSemaphore": "cudaError_t (cudaExternalSemaphore_t *, struct cudaExternalSemaphoreHandleDesc const *)",
+  "_cudaInitDevice": "cudaError_t (int, unsigned int, unsigned int)",
+  "_cudaIpcCloseMemHandle": "cudaError_t (void *)",
+  "_cudaIpcGetEventHandle": "cudaError_t (cudaIpcEventHandle_t *, cudaEvent_t)",
+  "_cudaIpcGetMemHandle": "cudaError_t (cudaIpcMemHandle_t *, void *)",
+  "_cudaIpcOpenEventHandle": "cudaError_t (cudaEvent_t *, cudaIpcEventHandle_t)",
+  "_cudaIpcOpenMemHandle": "cudaError_t (void **, cudaIpcMemHandle_t, unsigned int)",
+  "_cudaKernelSetAttributeForDevice": "cudaError_t (cudaKernel_t, enum cudaFuncAttribute, int, int)",
+  "_cudaLaunchHostFunc": "cudaError_t (cudaStream_t, cudaHostFn_t, void *)",
+  "_cudaLibraryEnumerateKernels": "cudaError_t (cudaKernel_t *, unsigned int, cudaLibrary_t)",
+  "_cudaLibraryGetGlobal": "cudaError_t (void **, size_t *, cudaLibrary_t, char const *)",
+  "_cudaLibraryGetKernel": "cudaError_t (cudaKernel_t *, cudaLibrary_t, char const *)",
+  "_cudaLibraryGetKernelCount": "cudaError_t (unsigned int *, cudaLibrary_t)",
+  "_cudaLibraryGetManaged": "cudaError_t (void **, size_t *, cudaLibrary_t, char const *)",
+  "_cudaLibraryGetUnifiedFunction": "cudaError_t (void **, cudaLibrary_t, char const *)",
+  "_cudaLibraryLoadData": "cudaError_t (cudaLibrary_t *, void const *, enum cudaJitOption *, void **, unsigned int, enum cudaLibraryOption *, void **, unsigned int)",
+  "_cudaLibraryLoadFromFile": "cudaError_t (cudaLibrary_t *, char const *, enum cudaJitOption *, void **, unsigned int, enum cudaLibraryOption *, void **, unsigned int)",
+  "_cudaLibraryUnload": "cudaError_t (cudaLibrary_t)",
+  "_cudaLogsCurrent": "cudaError_t (cudaLogIterator *, unsigned int)",
+  "_cudaLogsDumpToFile": "cudaError_t (cudaLogIterator *, char const *, unsigned int)",
+  "_cudaLogsDumpToMemory": "cudaError_t (cudaLogIterator *, char *, size_t *, unsigned int)",
+  "_cudaLogsRegisterCallback": "cudaError_t (cudaLogsCallback_t, void *, cudaLogsCallbackHandle *)",
+  "_cudaLogsUnregisterCallback": "cudaError_t (cudaLogsCallbackHandle)",
+  "_cudaMalloc": "cudaError_t (void **, size_t)",
+  "_cudaMalloc3D": "cudaError_t (struct cudaPitchedPtr *, struct cudaExtent)",
+  "_cudaMalloc3DArray": "cudaError_t (cudaArray_t *, struct cudaChannelFormatDesc const *, struct cudaExtent, unsigned int)",
+  "_cudaMallocArray": "cudaError_t (cudaArray_t *, struct cudaChannelFormatDesc const *, size_t, size_t, unsigned int)",
+  "_cudaMallocAsync": "cudaError_t (void **, size_t, cudaStream_t)",
+  "_cudaMallocFromPoolAsync": "cudaError_t (void **, size_t, cudaMemPool_t, cudaStream_t)",
+  "_cudaMallocHost": "cudaError_t (void **, size_t)",
+  "_cudaMallocManaged": "cudaError_t (void **, size_t, unsigned int)",
+  "_cudaMallocMipmappedArray": "cudaError_t (cudaMipmappedArray_t *, struct cudaChannelFormatDesc const *, struct cudaExtent, unsigned int, unsigned int)",
+  "_cudaMallocPitch": "cudaError_t (void **, size_t *, size_t, size_t)",
+  "_cudaMemAdvise": "cudaError_t (void const *, size_t, enum cudaMemoryAdvise, struct cudaMemLocation)",
+  "_cudaMemDiscardAndPrefetchBatchAsync": "cudaError_t (void **, size_t *, size_t, struct cudaMemLocation *, size_t *, size_t, unsigned PY_LONG_LONG, cudaStream_t)",
+  "_cudaMemDiscardBatchAsync": "cudaError_t (void **, size_t *, size_t, unsigned PY_LONG_LONG, cudaStream_t)",
+  "_cudaMemGetDefaultMemPool": "cudaError_t (cudaMemPool_t *, struct cudaMemLocation *, enum cudaMemAllocationType)",
+  "_cudaMemGetInfo": "cudaError_t (size_t *, size_t *)",
+  "_cudaMemGetMemPool": "cudaError_t (cudaMemPool_t *, struct cudaMemLocation *, enum cudaMemAllocationType)",
+  "_cudaMemPoolCreate": "cudaError_t (cudaMemPool_t *, struct cudaMemPoolProps const *)",
+  "_cudaMemPoolDestroy": "cudaError_t (cudaMemPool_t)",
+  "_cudaMemPoolExportPointer": "cudaError_t (struct cudaMemPoolPtrExportData *, void *)",
+  "_cudaMemPoolExportToShareableHandle": "cudaError_t (void *, cudaMemPool_t, enum cudaMemAllocationHandleType, unsigned int)",
+  "_cudaMemPoolGetAccess": "cudaError_t (enum cudaMemAccessFlags *, cudaMemPool_t, struct cudaMemLocation *)",
+  "_cudaMemPoolGetAttribute": "cudaError_t (cudaMemPool_t, enum cudaMemPoolAttr, void *)",
+  "_cudaMemPoolImportFromShareableHandle": "cudaError_t (cudaMemPool_t *, void *, enum cudaMemAllocationHandleType, unsigned int)",
+  "_cudaMemPoolImportPointer": "cudaError_t (void **, cudaMemPool_t, struct cudaMemPoolPtrExportData *)",
+  "_cudaMemPoolSetAccess": "cudaError_t (cudaMemPool_t, struct cudaMemAccessDesc const *, size_t)",
+  "_cudaMemPoolSetAttribute": "cudaError_t (cudaMemPool_t, enum cudaMemPoolAttr, void *)",
+  "_cudaMemPoolTrimTo": "cudaError_t (cudaMemPool_t, size_t)",
+  "_cudaMemPrefetchAsync": "cudaError_t (void const *, size_t, struct cudaMemLocation, unsigned int, cudaStream_t)",
+  "_cudaMemPrefetchBatchAsync": "cudaError_t (void **, size_t *, size_t, struct cudaMemLocation *, size_t *, size_t, unsigned PY_LONG_LONG, cudaStream_t)",
+  "_cudaMemRangeGetAttribute": "cudaError_t (void *, size_t, enum cudaMemRangeAttribute, void const *, size_t)",
+  "_cudaMemRangeGetAttributes": "cudaError_t (void **, size_t *, enum cudaMemRangeAttribute *, size_t, void const *, size_t)",
+  "_cudaMemSetMemPool": "cudaError_t (struct cudaMemLocation *, enum cudaMemAllocationType, cudaMemPool_t)",
+  "_cudaMemcpy": "cudaError_t (void *, void const *, size_t, enum cudaMemcpyKind)",
+  "_cudaMemcpy2D": "cudaError_t (void *, size_t, void const *, size_t, size_t, size_t, enum cudaMemcpyKind)",
+  "_cudaMemcpy2DArrayToArray": "cudaError_t (cudaArray_t, size_t, size_t, cudaArray_const_t, size_t, size_t, size_t, size_t, enum cudaMemcpyKind)",
+  "_cudaMemcpy2DAsync": "cudaError_t (void *, size_t, void const *, size_t, size_t, size_t, enum cudaMemcpyKind, cudaStream_t)",
+  "_cudaMemcpy2DFromArray": "cudaError_t (void *, size_t, cudaArray_const_t, size_t, size_t, size_t, size_t, enum cudaMemcpyKind)",
+  "_cudaMemcpy2DFromArrayAsync": "cudaError_t (void *, size_t, cudaArray_const_t, size_t, size_t, size_t, size_t, enum cudaMemcpyKind, cudaStream_t)",
+  "_cudaMemcpy2DToArray": "cudaError_t (cudaArray_t, size_t, size_t, void const *, size_t, size_t, size_t, enum cudaMemcpyKind)",
+  "_cudaMemcpy2DToArrayAsync": "cudaError_t (cudaArray_t, size_t, size_t, void const *, size_t, size_t, size_t, enum cudaMemcpyKind, cudaStream_t)",
+  "_cudaMemcpy3D": "cudaError_t (struct cudaMemcpy3DParms const *)",
+  "_cudaMemcpy3DAsync": "cudaError_t (struct cudaMemcpy3DParms const *, cudaStream_t)",
+  "_cudaMemcpy3DBatchAsync": "cudaError_t (size_t, struct cudaMemcpy3DBatchOp *, unsigned PY_LONG_LONG, cudaStream_t)",
+  "_cudaMemcpy3DPeer": "cudaError_t (struct cudaMemcpy3DPeerParms const *)",
+  "_cudaMemcpy3DPeerAsync": "cudaError_t (struct cudaMemcpy3DPeerParms const *, cudaStream_t)",
+  "_cudaMemcpyArrayToArray": "cudaError_t (cudaArray_t, size_t, size_t, cudaArray_const_t, size_t, size_t, size_t, enum cudaMemcpyKind)",
+  "_cudaMemcpyAsync": "cudaError_t (void *, void const *, size_t, enum cudaMemcpyKind, cudaStream_t)",
+  "_cudaMemcpyBatchAsync": "cudaError_t (void const **, void const **, size_t const *, size_t, struct cudaMemcpyAttributes *, size_t *, size_t, cudaStream_t)",
+  "_cudaMemcpyFromArray": "cudaError_t (void *, cudaArray_const_t, size_t, size_t, size_t, enum cudaMemcpyKind)",
+  "_cudaMemcpyFromArrayAsync": "cudaError_t (void *, cudaArray_const_t, size_t, size_t, size_t, enum cudaMemcpyKind, cudaStream_t)",
+  "_cudaMemcpyPeer": "cudaError_t (void *, int, void const *, int, size_t)",
+  "_cudaMemcpyPeerAsync": "cudaError_t (void *, int, void const *, int, size_t, cudaStream_t)",
+  "_cudaMemcpyToArray": "cudaError_t (cudaArray_t, size_t, size_t, void const *, size_t, enum cudaMemcpyKind)",
+  "_cudaMemcpyToArrayAsync": "cudaError_t (cudaArray_t, size_t, size_t, void const *, size_t, enum cudaMemcpyKind, cudaStream_t)",
+  "_cudaMemset": "cudaError_t (void *, int, size_t)",
+  "_cudaMemset2D": "cudaError_t (void *, size_t, int, size_t, size_t)",
+  "_cudaMemset2DAsync": "cudaError_t (void *, size_t, int, size_t, size_t, cudaStream_t)",
+  "_cudaMemset3D": "cudaError_t (struct cudaPitchedPtr, int, struct cudaExtent)",
+  "_cudaMemset3DAsync": "cudaError_t (struct cudaPitchedPtr, int, struct cudaExtent, cudaStream_t)",
+  "_cudaMemsetAsync": "cudaError_t (void *, int, size_t, cudaStream_t)",
+  "_cudaMipmappedArrayGetMemoryRequirements": "cudaError_t (struct cudaArrayMemoryRequirements *, cudaMipmappedArray_t, int)",
+  "_cudaMipmappedArrayGetSparseProperties": "cudaError_t (struct cudaArraySparseProperties *, cudaMipmappedArray_t)",
+  "_cudaOccupancyAvailableDynamicSMemPerBlock": "cudaError_t (size_t *, void const *, int, int)",
+  "_cudaOccupancyMaxActiveBlocksPerMultiprocessor": "cudaError_t (int *, void const *, int, size_t)",
+  "_cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags": "cudaError_t (int *, void const *, int, size_t, unsigned int)",
+  "_cudaPeekAtLastError": "cudaError_t (void)",
+  "_cudaPointerGetAttributes": "cudaError_t (struct cudaPointerAttributes *, void const *)",
+  "_cudaProfilerStart": "cudaError_t (void)",
+  "_cudaProfilerStop": "cudaError_t (void)",
+  "_cudaRuntimeGetVersion": "cudaError_t (int *)",
+  "_cudaSetDevice": "cudaError_t (int)",
+  "_cudaSetDeviceFlags": "cudaError_t (unsigned int)",
+  "_cudaSignalExternalSemaphoresAsync": "cudaError_t (cudaExternalSemaphore_t const *, struct cudaExternalSemaphoreSignalParams const *, unsigned int, cudaStream_t)",
+  "_cudaStreamAddCallback": "cudaError_t (cudaStream_t, cudaStreamCallback_t, void *, unsigned int)",
+  "_cudaStreamAttachMemAsync": "cudaError_t (cudaStream_t, void *, size_t, unsigned int)",
+  "_cudaStreamBeginCapture": "cudaError_t (cudaStream_t, enum cudaStreamCaptureMode)",
+  "_cudaStreamBeginCaptureToGraph": "cudaError_t (cudaStream_t, cudaGraph_t, cudaGraphNode_t const *, cudaGraphEdgeData const *, size_t, enum cudaStreamCaptureMode)",
+  "_cudaStreamCopyAttributes": "cudaError_t (cudaStream_t, cudaStream_t)",
+  "_cudaStreamCreate": "cudaError_t (cudaStream_t *)",
+  "_cudaStreamCreateWithFlags": "cudaError_t (cudaStream_t *, unsigned int)",
+  "_cudaStreamCreateWithPriority": "cudaError_t (cudaStream_t *, unsigned int, int)",
+  "_cudaStreamDestroy": "cudaError_t (cudaStream_t)",
+  "_cudaStreamEndCapture": "cudaError_t (cudaStream_t, cudaGraph_t *)",
+  "_cudaStreamGetAttribute": "cudaError_t (cudaStream_t, __pyx_t_4cuda_8bindings_9_bindings_9cyruntime_cudaStreamAttrID, __pyx_t_4cuda_8bindings_9_bindings_9cyruntime_cudaStreamAttrValue *)",
+  "_cudaStreamGetCaptureInfo": "cudaError_t (cudaStream_t, enum cudaStreamCaptureStatus *, unsigned PY_LONG_LONG *, cudaGraph_t *, cudaGraphNode_t const **, cudaGraphEdgeData const **, size_t *)",
+  "_cudaStreamGetDevice": "cudaError_t (cudaStream_t, int *)",
+  "_cudaStreamGetFlags": "cudaError_t (cudaStream_t, unsigned int *)",
+  "_cudaStreamGetId": "cudaError_t (cudaStream_t, unsigned PY_LONG_LONG *)",
+  "_cudaStreamGetPriority": "cudaError_t (cudaStream_t, int *)",
+  "_cudaStreamIsCapturing": "cudaError_t (cudaStream_t, enum cudaStreamCaptureStatus *)",
+  "_cudaStreamQuery": "cudaError_t (cudaStream_t)",
+  "_cudaStreamSetAttribute": "cudaError_t (cudaStream_t, __pyx_t_4cuda_8bindings_9_bindings_9cyruntime_cudaStreamAttrID, __pyx_t_4cuda_8bindings_9_bindings_9cyruntime_cudaStreamAttrValue const *)",
+  "_cudaStreamSynchronize": "cudaError_t (cudaStream_t)",
+  "_cudaStreamUpdateCaptureDependencies": "cudaError_t (cudaStream_t, cudaGraphNode_t *, cudaGraphEdgeData const *, size_t, unsigned int)",
+  "_cudaStreamWaitEvent": "cudaError_t (cudaStream_t, cudaEvent_t, unsigned int)",
+  "_cudaThreadExchangeStreamCaptureMode": "cudaError_t (enum cudaStreamCaptureMode *)",
+  "_cudaUserObjectCreate": "cudaError_t (cudaUserObject_t *, void *, cudaHostFn_t, unsigned int, unsigned int)",
+  "_cudaUserObjectRelease": "cudaError_t (cudaUserObject_t, unsigned int)",
+  "_cudaUserObjectRetain": "cudaError_t (cudaUserObject_t, unsigned int)",
+  "_cudaVDPAUGetDevice": "cudaError_t (int *, __pyx_t_4cuda_8bindings_9cyruntime_VdpDevice, __pyx_t_4cuda_8bindings_9cyruntime_VdpGetProcAddress *)",
+  "_cudaVDPAUSetVDPAUDevice": "cudaError_t (int, __pyx_t_4cuda_8bindings_9cyruntime_VdpDevice, __pyx_t_4cuda_8bindings_9cyruntime_VdpGetProcAddress *)",
+  "_cudaWaitExternalSemaphoresAsync": "cudaError_t (cudaExternalSemaphore_t const *, struct cudaExternalSemaphoreWaitParams const *, unsigned int, cudaStream_t)",
+  "_make_cudaExtent": "struct cudaExtent (size_t, size_t, size_t)",
+  "_make_cudaPitchedPtr": "struct cudaPitchedPtr (void *, size_t, size_t, size_t)",
+  "_make_cudaPos": "struct cudaPos (size_t, size_t, size_t)",
+  "getDriverEglFrame": "cudaError_t (__pyx_t_4cuda_8bindings_8cydriver_CUeglFrame *, __pyx_t_4cuda_8bindings_9cyruntime_cudaEglFrame)",
+  "getRuntimeEglFrame": "cudaError_t (__pyx_t_4cuda_8bindings_9cyruntime_cudaEglFrame *, __pyx_t_4cuda_8bindings_8cydriver_CUeglFrame)"
+}
diff --git a/cuda_bindings/tests/abi/cuda/bindings/_bindings/cyruntime_ptds.abi.json b/cuda_bindings/tests/abi/cuda/bindings/_bindings/cyruntime_ptds.abi.json
new file mode 100644
index 000000000..a35b2ca80
--- /dev/null
+++ b/cuda_bindings/tests/abi/cuda/bindings/_bindings/cyruntime_ptds.abi.json
@@ -0,0 +1,296 @@
+{
+  "_cudaArrayGetInfo": "cudaError_t (struct cudaChannelFormatDesc *, struct cudaExtent *, unsigned int *, cudaArray_t)",
+  "_cudaArrayGetMemoryRequirements": "cudaError_t (struct cudaArrayMemoryRequirements *, cudaArray_t, int)",
+  "_cudaArrayGetPlane": "cudaError_t (cudaArray_t *, cudaArray_t, unsigned int)",
+  "_cudaArrayGetSparseProperties": "cudaError_t (struct cudaArraySparseProperties *, cudaArray_t)",
+  "_cudaChooseDevice": "cudaError_t (int *, struct cudaDeviceProp const *)",
+  "_cudaCreateChannelDesc": "struct cudaChannelFormatDesc (int, int, int, int, enum cudaChannelFormatKind)",
+  "_cudaCreateSurfaceObject": "cudaError_t (cudaSurfaceObject_t *, struct cudaResourceDesc const *)",
+  "_cudaCreateTextureObject": "cudaError_t (cudaTextureObject_t *, struct cudaResourceDesc const *, struct cudaTextureDesc const *, struct cudaResourceViewDesc const *)",
+  "_cudaCtxResetPersistingL2Cache": "cudaError_t (void)",
+  "_cudaDestroyExternalMemory": "cudaError_t (cudaExternalMemory_t)",
+  "_cudaDestroyExternalSemaphore": "cudaError_t (cudaExternalSemaphore_t)",
+  "_cudaDestroySurfaceObject": "cudaError_t (cudaSurfaceObject_t)",
+  "_cudaDestroyTextureObject": "cudaError_t (cudaTextureObject_t)",
+  "_cudaDeviceCanAccessPeer": "cudaError_t (int *, int, int)",
+  "_cudaDeviceDisablePeerAccess": "cudaError_t (int)",
+  "_cudaDeviceEnablePeerAccess": "cudaError_t (int, unsigned int)",
+  "_cudaDeviceFlushGPUDirectRDMAWrites": "cudaError_t (enum cudaFlushGPUDirectRDMAWritesTarget, enum cudaFlushGPUDirectRDMAWritesScope)",
+  "_cudaDeviceGetAttribute": "cudaError_t (int *, enum cudaDeviceAttr, int)",
+  "_cudaDeviceGetByPCIBusId": "cudaError_t (int *, char const *)",
+  "_cudaDeviceGetCacheConfig": "cudaError_t (enum cudaFuncCache *)",
+  "_cudaDeviceGetDefaultMemPool": "cudaError_t (cudaMemPool_t *, int)",
+  "_cudaDeviceGetGraphMemAttribute": "cudaError_t (int, enum cudaGraphMemAttributeType, void *)",
+  "_cudaDeviceGetHostAtomicCapabilities": "cudaError_t (unsigned int *, enum cudaAtomicOperation const *, unsigned int, int)",
+  "_cudaDeviceGetLimit": "cudaError_t (size_t *, enum cudaLimit)",
+  "_cudaDeviceGetMemPool": "cudaError_t (cudaMemPool_t *, int)",
+  "_cudaDeviceGetNvSciSyncAttributes": "cudaError_t (void *, int, int)",
+  "_cudaDeviceGetP2PAtomicCapabilities": "cudaError_t (unsigned int *, enum cudaAtomicOperation const *, unsigned int, int, int)",
+  "_cudaDeviceGetP2PAttribute": "cudaError_t (int *, enum cudaDeviceP2PAttr, int, int)",
+  "_cudaDeviceGetPCIBusId": "cudaError_t (char *, int, int)",
+  "_cudaDeviceGetSharedMemConfig": "cudaError_t (enum cudaSharedMemConfig *)",
+  "_cudaDeviceGetStreamPriorityRange": "cudaError_t (int *, int *)",
+  "_cudaDeviceGetTexture1DLinearMaxWidth": "cudaError_t (size_t *, struct cudaChannelFormatDesc const *, int)",
+  "_cudaDeviceGraphMemTrim": "cudaError_t (int)",
+  "_cudaDeviceRegisterAsyncNotification": "cudaError_t (int, cudaAsyncCallback, void *, cudaAsyncCallbackHandle_t *)",
+  "_cudaDeviceReset": "cudaError_t (void)",
+  "_cudaDeviceSetCacheConfig": "cudaError_t (enum cudaFuncCache)",
+  "_cudaDeviceSetGraphMemAttribute": "cudaError_t (int, enum cudaGraphMemAttributeType, void *)",
+  "_cudaDeviceSetLimit": "cudaError_t (enum cudaLimit, size_t)",
+  "_cudaDeviceSetMemPool": "cudaError_t (int, cudaMemPool_t)",
+  "_cudaDeviceSetSharedMemConfig": "cudaError_t (enum cudaSharedMemConfig)",
+  "_cudaDeviceSynchronize": "cudaError_t (void)",
+  "_cudaDeviceUnregisterAsyncNotification": "cudaError_t (int, cudaAsyncCallbackHandle_t)",
+  "_cudaDriverGetVersion": "cudaError_t (int *)",
+  "_cudaEventCreate": "cudaError_t (cudaEvent_t *)",
+  "_cudaEventCreateWithFlags": "cudaError_t (cudaEvent_t *, unsigned int)",
+  "_cudaEventDestroy": "cudaError_t (cudaEvent_t)",
+  "_cudaEventElapsedTime": "cudaError_t (float *, cudaEvent_t, cudaEvent_t)",
+  "_cudaEventQuery": "cudaError_t (cudaEvent_t)",
+  "_cudaEventRecord": "cudaError_t (cudaEvent_t, cudaStream_t)",
+  "_cudaEventRecordWithFlags": "cudaError_t (cudaEvent_t, cudaStream_t, unsigned int)",
+  "_cudaEventSynchronize": "cudaError_t (cudaEvent_t)",
+  "_cudaExternalMemoryGetMappedBuffer": "cudaError_t (void **, cudaExternalMemory_t, struct cudaExternalMemoryBufferDesc const *)",
+  "_cudaExternalMemoryGetMappedMipmappedArray": "cudaError_t (cudaMipmappedArray_t *, cudaExternalMemory_t, struct cudaExternalMemoryMipmappedArrayDesc const *)",
+  "_cudaFree": "cudaError_t (void *)",
+  "_cudaFreeArray": "cudaError_t (cudaArray_t)",
+  "_cudaFreeAsync": "cudaError_t (void *, cudaStream_t)",
+  "_cudaFreeHost": "cudaError_t (void *)",
+  "_cudaFreeMipmappedArray": "cudaError_t (cudaMipmappedArray_t)",
+  "_cudaFuncGetAttributes": "cudaError_t (struct cudaFuncAttributes *, void const *)",
+  "_cudaFuncSetAttribute": "cudaError_t (void const *, enum cudaFuncAttribute, int)",
+  "_cudaFuncSetCacheConfig": "cudaError_t (void const *, enum cudaFuncCache)",
+  "_cudaFuncSetSharedMemConfig": "cudaError_t (void const *, enum cudaSharedMemConfig)",
+  "_cudaGetChannelDesc": "cudaError_t (struct cudaChannelFormatDesc *, cudaArray_const_t)",
+  "_cudaGetDevice": "cudaError_t (int *)",
+  "_cudaGetDeviceCount": "cudaError_t (int *)",
+  "_cudaGetDeviceFlags": "cudaError_t (unsigned int *)",
+  "_cudaGetDeviceProperties": "cudaError_t (struct cudaDeviceProp *, int)",
+  "_cudaGetDriverEntryPoint": "cudaError_t (char const *, void **, unsigned PY_LONG_LONG, enum cudaDriverEntryPointQueryResult *)",
+  "_cudaGetDriverEntryPointByVersion": "cudaError_t (char const *, void **, unsigned int, unsigned PY_LONG_LONG, enum cudaDriverEntryPointQueryResult *)",
+  "_cudaGetErrorName": "char const *(cudaError_t)",
+  "_cudaGetErrorString": "char const *(cudaError_t)",
+  "_cudaGetExportTable": "cudaError_t (void const **, cudaUUID_t const *)",
+  "_cudaGetKernel": "cudaError_t (cudaKernel_t *, void const *)",
+  "_cudaGetLastError": "cudaError_t (void)",
+  "_cudaGetMipmappedArrayLevel": "cudaError_t (cudaArray_t *, cudaMipmappedArray_const_t, unsigned int)",
+  "_cudaGetSurfaceObjectResourceDesc": "cudaError_t (struct cudaResourceDesc *, cudaSurfaceObject_t)",
+  "_cudaGetTextureObjectResourceDesc": "cudaError_t (struct cudaResourceDesc *, cudaTextureObject_t)",
+  "_cudaGetTextureObjectResourceViewDesc": "cudaError_t (struct cudaResourceViewDesc *, cudaTextureObject_t)",
+  "_cudaGetTextureObjectTextureDesc": "cudaError_t (struct cudaTextureDesc *, cudaTextureObject_t)",
+  "_cudaGraphAddChildGraphNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, cudaGraph_t)",
+  "_cudaGraphAddDependencies": "cudaError_t (cudaGraph_t, cudaGraphNode_t const *, cudaGraphNode_t const *, cudaGraphEdgeData const *, size_t)",
+  "_cudaGraphAddEmptyNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t)",
+  "_cudaGraphAddEventRecordNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, cudaEvent_t)",
+  "_cudaGraphAddEventWaitNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, cudaEvent_t)",
+  "_cudaGraphAddExternalSemaphoresSignalNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, struct cudaExternalSemaphoreSignalNodeParams const *)",
+  "_cudaGraphAddExternalSemaphoresWaitNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, struct cudaExternalSemaphoreWaitNodeParams const *)",
+  "_cudaGraphAddHostNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, struct cudaHostNodeParams const *)",
+  "_cudaGraphAddKernelNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, struct cudaKernelNodeParams const *)",
+  "_cudaGraphAddMemAllocNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, struct cudaMemAllocNodeParams *)",
+  "_cudaGraphAddMemFreeNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, void *)",
+  "_cudaGraphAddMemcpyNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, struct cudaMemcpy3DParms const *)",
+  "_cudaGraphAddMemcpyNode1D": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, void *, void const *, size_t, enum cudaMemcpyKind)",
+  "_cudaGraphAddMemsetNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, struct cudaMemsetParams const *)",
+  "_cudaGraphAddNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, cudaGraphEdgeData const *, size_t, struct cudaGraphNodeParams *)",
+  "_cudaGraphChildGraphNodeGetGraph": "cudaError_t (cudaGraphNode_t, cudaGraph_t *)",
+  "_cudaGraphClone": "cudaError_t (cudaGraph_t *, cudaGraph_t)",
+  "_cudaGraphConditionalHandleCreate": "cudaError_t (cudaGraphConditionalHandle *, cudaGraph_t, unsigned int, unsigned int)",
+  "_cudaGraphCreate": "cudaError_t (cudaGraph_t *, unsigned int)",
+  "_cudaGraphDebugDotPrint": "cudaError_t (cudaGraph_t, char const *, unsigned int)",
+  "_cudaGraphDestroy": "cudaError_t (cudaGraph_t)",
+  "_cudaGraphDestroyNode": "cudaError_t (cudaGraphNode_t)",
+  "_cudaGraphEventRecordNodeGetEvent": "cudaError_t (cudaGraphNode_t, cudaEvent_t *)",
+  "_cudaGraphEventRecordNodeSetEvent": "cudaError_t (cudaGraphNode_t, cudaEvent_t)",
+  "_cudaGraphEventWaitNodeGetEvent": "cudaError_t (cudaGraphNode_t, cudaEvent_t *)",
+  "_cudaGraphEventWaitNodeSetEvent": "cudaError_t (cudaGraphNode_t, cudaEvent_t)",
+  "_cudaGraphExecChildGraphNodeSetParams": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, cudaGraph_t)",
+  "_cudaGraphExecDestroy": "cudaError_t (cudaGraphExec_t)",
+  "_cudaGraphExecEventRecordNodeSetEvent": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, cudaEvent_t)",
+  "_cudaGraphExecEventWaitNodeSetEvent": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, cudaEvent_t)",
+  "_cudaGraphExecExternalSemaphoresSignalNodeSetParams": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, struct cudaExternalSemaphoreSignalNodeParams const *)",
+  "_cudaGraphExecExternalSemaphoresWaitNodeSetParams": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, struct cudaExternalSemaphoreWaitNodeParams const *)",
+  "_cudaGraphExecGetFlags": "cudaError_t (cudaGraphExec_t, unsigned PY_LONG_LONG *)",
+  "_cudaGraphExecHostNodeSetParams": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, struct cudaHostNodeParams const *)",
+  "_cudaGraphExecKernelNodeSetParams": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, struct cudaKernelNodeParams const *)",
+  "_cudaGraphExecMemcpyNodeSetParams": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, struct cudaMemcpy3DParms const *)",
+  "_cudaGraphExecMemcpyNodeSetParams1D": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, void *, void const *, size_t, enum cudaMemcpyKind)",
+  "_cudaGraphExecMemsetNodeSetParams": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, struct cudaMemsetParams const *)",
+  "_cudaGraphExecNodeSetParams": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, struct cudaGraphNodeParams *)",
+  "_cudaGraphExecUpdate": "cudaError_t (cudaGraphExec_t, cudaGraph_t, cudaGraphExecUpdateResultInfo *)",
+  "_cudaGraphExternalSemaphoresSignalNodeGetParams": "cudaError_t (cudaGraphNode_t, struct cudaExternalSemaphoreSignalNodeParams *)",
+  "_cudaGraphExternalSemaphoresSignalNodeSetParams": "cudaError_t (cudaGraphNode_t, struct cudaExternalSemaphoreSignalNodeParams const *)",
+  "_cudaGraphExternalSemaphoresWaitNodeGetParams": "cudaError_t (cudaGraphNode_t, struct cudaExternalSemaphoreWaitNodeParams *)",
+  "_cudaGraphExternalSemaphoresWaitNodeSetParams": "cudaError_t (cudaGraphNode_t, struct cudaExternalSemaphoreWaitNodeParams const *)",
+  "_cudaGraphGetEdges": "cudaError_t (cudaGraph_t, cudaGraphNode_t *, cudaGraphNode_t *, cudaGraphEdgeData *, size_t *)",
+  "_cudaGraphGetNodes": "cudaError_t (cudaGraph_t, cudaGraphNode_t *, size_t *)",
+  "_cudaGraphGetRootNodes": "cudaError_t (cudaGraph_t, cudaGraphNode_t *, size_t *)",
+  "_cudaGraphHostNodeGetParams": "cudaError_t (cudaGraphNode_t, struct cudaHostNodeParams *)",
+  "_cudaGraphHostNodeSetParams": "cudaError_t (cudaGraphNode_t, struct cudaHostNodeParams const *)",
+  "_cudaGraphInstantiate": "cudaError_t (cudaGraphExec_t *, cudaGraph_t, unsigned PY_LONG_LONG)",
+  "_cudaGraphInstantiateWithFlags": "cudaError_t (cudaGraphExec_t *, cudaGraph_t, unsigned PY_LONG_LONG)",
+  "_cudaGraphInstantiateWithParams": "cudaError_t (cudaGraphExec_t *, cudaGraph_t, cudaGraphInstantiateParams *)",
+  "_cudaGraphKernelNodeCopyAttributes": "cudaError_t (cudaGraphNode_t, cudaGraphNode_t)",
+  "_cudaGraphKernelNodeGetAttribute": "cudaError_t (cudaGraphNode_t, __pyx_t_4cuda_8bindings_9_bindings_14cyruntime_ptds_cudaKernelNodeAttrID, __pyx_t_4cuda_8bindings_9_bindings_14cyruntime_ptds_cudaKernelNodeAttrValue *)",
+  "_cudaGraphKernelNodeGetParams": "cudaError_t (cudaGraphNode_t, struct cudaKernelNodeParams *)",
+  "_cudaGraphKernelNodeSetAttribute": "cudaError_t (cudaGraphNode_t, __pyx_t_4cuda_8bindings_9_bindings_14cyruntime_ptds_cudaKernelNodeAttrID, __pyx_t_4cuda_8bindings_9_bindings_14cyruntime_ptds_cudaKernelNodeAttrValue const *)",
+  "_cudaGraphKernelNodeSetParams": "cudaError_t (cudaGraphNode_t, struct cudaKernelNodeParams const *)",
+  "_cudaGraphLaunch": "cudaError_t (cudaGraphExec_t, cudaStream_t)",
+  "_cudaGraphMemAllocNodeGetParams": "cudaError_t (cudaGraphNode_t, struct cudaMemAllocNodeParams *)",
+  "_cudaGraphMemFreeNodeGetParams": "cudaError_t (cudaGraphNode_t, void *)",
+  "_cudaGraphMemcpyNodeGetParams": "cudaError_t (cudaGraphNode_t, struct cudaMemcpy3DParms *)",
+  "_cudaGraphMemcpyNodeSetParams": "cudaError_t (cudaGraphNode_t, struct cudaMemcpy3DParms const *)",
+  "_cudaGraphMemcpyNodeSetParams1D": "cudaError_t (cudaGraphNode_t, void *, void const *, size_t, enum cudaMemcpyKind)",
+  "_cudaGraphMemsetNodeGetParams": "cudaError_t (cudaGraphNode_t, struct cudaMemsetParams *)",
+  "_cudaGraphMemsetNodeSetParams": "cudaError_t (cudaGraphNode_t, struct cudaMemsetParams const *)",
+  "_cudaGraphNodeFindInClone": "cudaError_t (cudaGraphNode_t *, cudaGraphNode_t, cudaGraph_t)",
+  "_cudaGraphNodeGetDependencies": "cudaError_t (cudaGraphNode_t, cudaGraphNode_t *, cudaGraphEdgeData *, size_t *)",
+  "_cudaGraphNodeGetDependentNodes": "cudaError_t (cudaGraphNode_t, cudaGraphNode_t *, cudaGraphEdgeData *, size_t *)",
+  "_cudaGraphNodeGetEnabled": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, unsigned int *)",
+  "_cudaGraphNodeGetType": "cudaError_t (cudaGraphNode_t, enum cudaGraphNodeType *)",
+  "_cudaGraphNodeSetEnabled": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, unsigned int)",
+  "_cudaGraphNodeSetParams": "cudaError_t (cudaGraphNode_t, struct cudaGraphNodeParams *)",
+  "_cudaGraphReleaseUserObject": "cudaError_t (cudaGraph_t, cudaUserObject_t, unsigned int)",
+  "_cudaGraphRemoveDependencies": "cudaError_t (cudaGraph_t, cudaGraphNode_t const *, cudaGraphNode_t const *, cudaGraphEdgeData const *, size_t)",
+  "_cudaGraphRetainUserObject": "cudaError_t (cudaGraph_t, cudaUserObject_t, unsigned int, unsigned int)",
+  "_cudaGraphUpload": "cudaError_t (cudaGraphExec_t, cudaStream_t)",
+  "_cudaGraphicsMapResources": "cudaError_t (int, cudaGraphicsResource_t *, cudaStream_t)",
+  "_cudaGraphicsResourceGetMappedMipmappedArray": "cudaError_t (cudaMipmappedArray_t *, cudaGraphicsResource_t)",
+  "_cudaGraphicsResourceGetMappedPointer": "cudaError_t (void **, size_t *, cudaGraphicsResource_t)",
+  "_cudaGraphicsResourceSetMapFlags": "cudaError_t (cudaGraphicsResource_t, unsigned int)",
+  "_cudaGraphicsSubResourceGetMappedArray": "cudaError_t (cudaArray_t *, cudaGraphicsResource_t, unsigned int, unsigned int)",
+  "_cudaGraphicsUnmapResources": "cudaError_t (int, cudaGraphicsResource_t *, cudaStream_t)",
+  "_cudaGraphicsUnregisterResource": "cudaError_t (cudaGraphicsResource_t)",
+  "_cudaHostAlloc": "cudaError_t (void **, size_t, unsigned int)",
+  "_cudaHostGetDevicePointer": "cudaError_t (void **, void *, unsigned int)",
+  "_cudaHostGetFlags": "cudaError_t (unsigned int *, void *)",
+  "_cudaHostRegister": "cudaError_t (void *, size_t, unsigned int)",
+  "_cudaHostUnregister": "cudaError_t (void *)",
+  "_cudaImportExternalMemory": "cudaError_t (cudaExternalMemory_t *, struct cudaExternalMemoryHandleDesc const *)",
+  "_cudaImportExternalSemaphore": "cudaError_t (cudaExternalSemaphore_t *, struct cudaExternalSemaphoreHandleDesc const *)",
+  "_cudaInitDevice": "cudaError_t (int, unsigned int, unsigned int)",
+  "_cudaIpcCloseMemHandle": "cudaError_t (void *)",
+  "_cudaIpcGetEventHandle": "cudaError_t (cudaIpcEventHandle_t *, cudaEvent_t)",
+  "_cudaIpcGetMemHandle": "cudaError_t (cudaIpcMemHandle_t *, void *)",
+  "_cudaIpcOpenEventHandle": "cudaError_t (cudaEvent_t *, cudaIpcEventHandle_t)",
+  "_cudaIpcOpenMemHandle": "cudaError_t (void **, cudaIpcMemHandle_t, unsigned int)",
+  "_cudaKernelSetAttributeForDevice": "cudaError_t (cudaKernel_t, enum cudaFuncAttribute, int, int)",
+  "_cudaLaunchHostFunc": "cudaError_t (cudaStream_t, cudaHostFn_t, void *)",
+  "_cudaLibraryEnumerateKernels": "cudaError_t (cudaKernel_t *, unsigned int, cudaLibrary_t)",
+  "_cudaLibraryGetGlobal": "cudaError_t (void **, size_t *, cudaLibrary_t, char const *)",
+  "_cudaLibraryGetKernel": "cudaError_t (cudaKernel_t *, cudaLibrary_t, char const *)",
+  "_cudaLibraryGetKernelCount": "cudaError_t (unsigned int *, cudaLibrary_t)",
+  "_cudaLibraryGetManaged": "cudaError_t (void **, size_t *, cudaLibrary_t, char const *)",
+  "_cudaLibraryGetUnifiedFunction": "cudaError_t (void **, cudaLibrary_t, char const *)",
+  "_cudaLibraryLoadData": "cudaError_t (cudaLibrary_t *, void const *, enum cudaJitOption *, void **, unsigned int, enum cudaLibraryOption *, void **, unsigned int)",
+  "_cudaLibraryLoadFromFile": "cudaError_t (cudaLibrary_t *, char const *, enum cudaJitOption *, void **, unsigned int, enum cudaLibraryOption *, void **, unsigned int)",
+  "_cudaLibraryUnload": "cudaError_t (cudaLibrary_t)",
+  "_cudaLogsCurrent": "cudaError_t (cudaLogIterator *, unsigned int)",
+  "_cudaLogsDumpToFile": "cudaError_t (cudaLogIterator *, char const *, unsigned int)",
+  "_cudaLogsDumpToMemory": "cudaError_t (cudaLogIterator *, char *, size_t *, unsigned int)",
+  "_cudaLogsRegisterCallback": "cudaError_t (cudaLogsCallback_t, void *, cudaLogsCallbackHandle *)",
+  "_cudaLogsUnregisterCallback": "cudaError_t (cudaLogsCallbackHandle)",
+  "_cudaMalloc": "cudaError_t (void **, size_t)",
+  "_cudaMalloc3D": "cudaError_t (struct cudaPitchedPtr *, struct cudaExtent)",
+  "_cudaMalloc3DArray": "cudaError_t (cudaArray_t *, struct cudaChannelFormatDesc const *, struct cudaExtent, unsigned int)",
+  "_cudaMallocArray": "cudaError_t (cudaArray_t *, struct cudaChannelFormatDesc const *, size_t, size_t, unsigned int)",
+  "_cudaMallocAsync": "cudaError_t (void **, size_t, cudaStream_t)",
+  "_cudaMallocFromPoolAsync": "cudaError_t (void **, size_t, cudaMemPool_t, cudaStream_t)",
+  "_cudaMallocHost": "cudaError_t (void **, size_t)",
+  "_cudaMallocManaged": "cudaError_t (void **, size_t, unsigned int)",
+  "_cudaMallocMipmappedArray": "cudaError_t (cudaMipmappedArray_t *, struct cudaChannelFormatDesc const *, struct cudaExtent, unsigned int, unsigned int)",
+  "_cudaMallocPitch": "cudaError_t (void **, size_t *, size_t, size_t)",
+  "_cudaMemAdvise": "cudaError_t (void const *, size_t, enum cudaMemoryAdvise, struct cudaMemLocation)",
+  "_cudaMemDiscardAndPrefetchBatchAsync": "cudaError_t (void **, size_t *, size_t, struct cudaMemLocation *, size_t *, size_t, unsigned PY_LONG_LONG, cudaStream_t)",
+  "_cudaMemDiscardBatchAsync": "cudaError_t (void **, size_t *, size_t, unsigned PY_LONG_LONG, cudaStream_t)",
+  "_cudaMemGetDefaultMemPool": "cudaError_t (cudaMemPool_t *, struct cudaMemLocation *, enum cudaMemAllocationType)",
+  "_cudaMemGetInfo": "cudaError_t (size_t *, size_t *)",
+  "_cudaMemGetMemPool": "cudaError_t (cudaMemPool_t *, struct cudaMemLocation *, enum cudaMemAllocationType)",
+  "_cudaMemPoolCreate": "cudaError_t (cudaMemPool_t *, struct cudaMemPoolProps const *)",
+  "_cudaMemPoolDestroy": "cudaError_t (cudaMemPool_t)",
+  "_cudaMemPoolExportPointer": "cudaError_t (struct cudaMemPoolPtrExportData *, void *)",
+  "_cudaMemPoolExportToShareableHandle": "cudaError_t (void *, cudaMemPool_t, enum cudaMemAllocationHandleType, unsigned int)",
+  "_cudaMemPoolGetAccess": "cudaError_t (enum cudaMemAccessFlags *, cudaMemPool_t, struct cudaMemLocation *)",
+  "_cudaMemPoolGetAttribute": "cudaError_t (cudaMemPool_t, enum cudaMemPoolAttr, void *)",
+  "_cudaMemPoolImportFromShareableHandle": "cudaError_t (cudaMemPool_t *, void *, enum cudaMemAllocationHandleType, unsigned int)",
+  "_cudaMemPoolImportPointer": "cudaError_t (void **, cudaMemPool_t, struct cudaMemPoolPtrExportData *)",
+  "_cudaMemPoolSetAccess": "cudaError_t (cudaMemPool_t, struct cudaMemAccessDesc const *, size_t)",
+  "_cudaMemPoolSetAttribute": "cudaError_t (cudaMemPool_t, enum cudaMemPoolAttr, void *)",
+  "_cudaMemPoolTrimTo": "cudaError_t (cudaMemPool_t, size_t)",
+  "_cudaMemPrefetchAsync": "cudaError_t (void const *, size_t, struct cudaMemLocation, unsigned int, cudaStream_t)",
+  "_cudaMemPrefetchBatchAsync": "cudaError_t (void **, size_t *, size_t, struct cudaMemLocation *, size_t *, size_t, unsigned PY_LONG_LONG, cudaStream_t)",
+  "_cudaMemRangeGetAttribute": "cudaError_t (void *, size_t, enum cudaMemRangeAttribute, void const *, size_t)",
+  "_cudaMemRangeGetAttributes": "cudaError_t (void **, size_t *, enum cudaMemRangeAttribute *, size_t, void const *, size_t)",
+  "_cudaMemSetMemPool": "cudaError_t (struct cudaMemLocation *, enum cudaMemAllocationType, cudaMemPool_t)",
+  "_cudaMemcpy": "cudaError_t (void *, void const *, size_t, enum cudaMemcpyKind)",
+  "_cudaMemcpy2D": "cudaError_t (void *, size_t, void const *, size_t, size_t, size_t, enum cudaMemcpyKind)",
+  "_cudaMemcpy2DArrayToArray": "cudaError_t (cudaArray_t, size_t, size_t, cudaArray_const_t, size_t, size_t, size_t, size_t, enum cudaMemcpyKind)",
+  "_cudaMemcpy2DAsync": "cudaError_t (void *, size_t, void const *, size_t, size_t, size_t, enum cudaMemcpyKind, cudaStream_t)",
+  "_cudaMemcpy2DFromArray": "cudaError_t (void *, size_t, cudaArray_const_t, size_t, size_t, size_t, size_t, enum cudaMemcpyKind)",
+  "_cudaMemcpy2DFromArrayAsync": "cudaError_t (void *, size_t, cudaArray_const_t, size_t, size_t, size_t, size_t, enum cudaMemcpyKind, cudaStream_t)",
+  "_cudaMemcpy2DToArray": "cudaError_t (cudaArray_t, size_t, size_t, void const *, size_t, size_t, size_t, enum cudaMemcpyKind)",
+  "_cudaMemcpy2DToArrayAsync": "cudaError_t (cudaArray_t, size_t, size_t, void const *, size_t, size_t, size_t, enum cudaMemcpyKind, cudaStream_t)",
+  "_cudaMemcpy3D": "cudaError_t (struct cudaMemcpy3DParms const *)",
+  "_cudaMemcpy3DAsync": "cudaError_t (struct cudaMemcpy3DParms const *, cudaStream_t)",
+  "_cudaMemcpy3DBatchAsync": "cudaError_t (size_t, struct cudaMemcpy3DBatchOp *, unsigned PY_LONG_LONG, cudaStream_t)",
+  "_cudaMemcpy3DPeer": "cudaError_t (struct cudaMemcpy3DPeerParms const *)",
+  "_cudaMemcpy3DPeerAsync": "cudaError_t (struct cudaMemcpy3DPeerParms const *, cudaStream_t)",
+  "_cudaMemcpyArrayToArray": "cudaError_t (cudaArray_t, size_t, size_t, cudaArray_const_t, size_t, size_t, size_t, enum cudaMemcpyKind)",
+  "_cudaMemcpyAsync": "cudaError_t (void *, void const *, size_t, enum cudaMemcpyKind, cudaStream_t)",
+  "_cudaMemcpyBatchAsync": "cudaError_t (void const **, void const **, size_t const *, size_t, struct cudaMemcpyAttributes *, size_t *, size_t, cudaStream_t)",
+  "_cudaMemcpyFromArray": "cudaError_t (void *, cudaArray_const_t, size_t, size_t, size_t, enum cudaMemcpyKind)",
+  "_cudaMemcpyFromArrayAsync": "cudaError_t (void *, cudaArray_const_t, size_t, size_t, size_t, enum cudaMemcpyKind, cudaStream_t)",
+  "_cudaMemcpyPeer": "cudaError_t (void *, int, void const *, int, size_t)",
+  "_cudaMemcpyPeerAsync": "cudaError_t (void *, int, void const *, int, size_t, cudaStream_t)",
+  "_cudaMemcpyToArray": "cudaError_t (cudaArray_t, size_t, size_t, void const *, size_t, enum cudaMemcpyKind)",
+  "_cudaMemcpyToArrayAsync": "cudaError_t (cudaArray_t, size_t, size_t, void const *, size_t, enum cudaMemcpyKind, cudaStream_t)",
+  "_cudaMemset": "cudaError_t (void *, int, size_t)",
+  "_cudaMemset2D": "cudaError_t (void *, size_t, int, size_t, size_t)",
+  "_cudaMemset2DAsync": "cudaError_t (void *, size_t, int, size_t, size_t, cudaStream_t)",
+  "_cudaMemset3D": "cudaError_t (struct cudaPitchedPtr, int, struct cudaExtent)",
+  "_cudaMemset3DAsync": "cudaError_t (struct cudaPitchedPtr, int, struct cudaExtent, cudaStream_t)",
+  "_cudaMemsetAsync": "cudaError_t (void *, int, size_t, cudaStream_t)",
+  "_cudaMipmappedArrayGetMemoryRequirements": "cudaError_t (struct cudaArrayMemoryRequirements *, cudaMipmappedArray_t, int)",
+  "_cudaMipmappedArrayGetSparseProperties": "cudaError_t (struct cudaArraySparseProperties *, cudaMipmappedArray_t)",
+  "_cudaOccupancyAvailableDynamicSMemPerBlock": "cudaError_t (size_t *, void const *, int, int)",
+  "_cudaOccupancyMaxActiveBlocksPerMultiprocessor": "cudaError_t (int *, void const *, int, size_t)",
+  "_cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags": "cudaError_t (int *, void const *, int, size_t, unsigned int)",
+  "_cudaPeekAtLastError": "cudaError_t (void)",
+  "_cudaPointerGetAttributes": "cudaError_t (struct cudaPointerAttributes *, void const *)",
+  "_cudaProfilerStart": "cudaError_t (void)",
+  "_cudaProfilerStop": "cudaError_t (void)",
+  "_cudaRuntimeGetVersion": "cudaError_t (int *)",
+  "_cudaSetDevice": "cudaError_t (int)",
+  "_cudaSetDeviceFlags": "cudaError_t (unsigned int)",
+  "_cudaSignalExternalSemaphoresAsync": "cudaError_t (cudaExternalSemaphore_t const *, struct cudaExternalSemaphoreSignalParams const *, unsigned int, cudaStream_t)",
+  "_cudaStreamAddCallback": "cudaError_t (cudaStream_t, cudaStreamCallback_t, void *, unsigned int)",
+  "_cudaStreamAttachMemAsync": "cudaError_t (cudaStream_t, void *, size_t, unsigned int)",
+  "_cudaStreamBeginCapture": "cudaError_t (cudaStream_t, enum cudaStreamCaptureMode)",
+  "_cudaStreamBeginCaptureToGraph": "cudaError_t (cudaStream_t, cudaGraph_t, cudaGraphNode_t const *, cudaGraphEdgeData const *, size_t, enum cudaStreamCaptureMode)",
+  "_cudaStreamCopyAttributes": "cudaError_t (cudaStream_t, cudaStream_t)",
+  "_cudaStreamCreate": "cudaError_t (cudaStream_t *)",
+  "_cudaStreamCreateWithFlags": "cudaError_t (cudaStream_t *, unsigned int)",
+  "_cudaStreamCreateWithPriority": "cudaError_t (cudaStream_t *, unsigned int, int)",
+  "_cudaStreamDestroy": "cudaError_t (cudaStream_t)",
+  "_cudaStreamEndCapture": "cudaError_t (cudaStream_t, cudaGraph_t *)",
+  "_cudaStreamGetAttribute": "cudaError_t (cudaStream_t, __pyx_t_4cuda_8bindings_9_bindings_14cyruntime_ptds_cudaStreamAttrID, __pyx_t_4cuda_8bindings_9_bindings_14cyruntime_ptds_cudaStreamAttrValue *)",
+  "_cudaStreamGetCaptureInfo": "cudaError_t (cudaStream_t, enum cudaStreamCaptureStatus *, unsigned PY_LONG_LONG *, cudaGraph_t *, cudaGraphNode_t const **, cudaGraphEdgeData const **, size_t *)",
+  "_cudaStreamGetDevice": "cudaError_t (cudaStream_t, int *)",
+  "_cudaStreamGetFlags": "cudaError_t (cudaStream_t, unsigned int *)",
+  "_cudaStreamGetId": "cudaError_t (cudaStream_t, unsigned PY_LONG_LONG *)",
+  "_cudaStreamGetPriority": "cudaError_t (cudaStream_t, int *)",
+  "_cudaStreamIsCapturing": "cudaError_t (cudaStream_t, enum cudaStreamCaptureStatus *)",
+  "_cudaStreamQuery": "cudaError_t (cudaStream_t)",
+  "_cudaStreamSetAttribute": "cudaError_t (cudaStream_t, __pyx_t_4cuda_8bindings_9_bindings_14cyruntime_ptds_cudaStreamAttrID, __pyx_t_4cuda_8bindings_9_bindings_14cyruntime_ptds_cudaStreamAttrValue const *)",
+  "_cudaStreamSynchronize": "cudaError_t (cudaStream_t)",
+  "_cudaStreamUpdateCaptureDependencies": "cudaError_t (cudaStream_t, cudaGraphNode_t *, cudaGraphEdgeData const *, size_t, unsigned int)",
+  "_cudaStreamWaitEvent": "cudaError_t (cudaStream_t, cudaEvent_t, unsigned int)",
+  "_cudaThreadExchangeStreamCaptureMode": "cudaError_t (enum cudaStreamCaptureMode *)",
+  "_cudaUserObjectCreate": "cudaError_t (cudaUserObject_t *, void *, cudaHostFn_t, unsigned int, unsigned int)",
+  "_cudaUserObjectRelease": "cudaError_t (cudaUserObject_t, unsigned int)",
+  "_cudaUserObjectRetain": "cudaError_t (cudaUserObject_t, unsigned int)",
+  "_cudaWaitExternalSemaphoresAsync": "cudaError_t (cudaExternalSemaphore_t const *, struct cudaExternalSemaphoreWaitParams const *, unsigned int, cudaStream_t)",
+  "_make_cudaExtent": "struct cudaExtent (size_t, size_t, size_t)",
+  "_make_cudaPitchedPtr": "struct cudaPitchedPtr (void *, size_t, size_t, size_t)",
+  "_make_cudaPos": "struct cudaPos (size_t, size_t, size_t)"
+}
diff --git a/cuda_bindings/tests/abi/cuda/bindings/_internal/cufile.abi.json b/cuda_bindings/tests/abi/cuda/bindings/_internal/cufile.abi.json
new file mode 100644
index 000000000..00327707c
--- /dev/null
+++ b/cuda_bindings/tests/abi/cuda/bindings/_internal/cufile.abi.json
@@ -0,0 +1,45 @@
+{
+  "_cuFileBatchIOCancel": "CUfileError_t (CUfileBatchHandle_t)",
+  "_cuFileBatchIODestroy": "void (CUfileBatchHandle_t)",
+  "_cuFileBatchIOGetStatus": "CUfileError_t (CUfileBatchHandle_t, unsigned int, unsigned int *, CUfileIOEvents_t *, struct timespec *)",
+  "_cuFileBatchIOSetUp": "CUfileError_t (CUfileBatchHandle_t *, unsigned int)",
+  "_cuFileBatchIOSubmit": "CUfileError_t (CUfileBatchHandle_t, unsigned int, CUfileIOParams_t *, unsigned int)",
+  "_cuFileBufDeregister": "CUfileError_t (void const *)",
+  "_cuFileBufRegister": "CUfileError_t (void const *, size_t, int)",
+  "_cuFileDriverClose": "CUfileError_t (void)",
+  "_cuFileDriverClose_v2": "CUfileError_t (void)",
+  "_cuFileDriverGetProperties": "CUfileError_t (CUfileDrvProps_t *)",
+  "_cuFileDriverOpen": "CUfileError_t (void)",
+  "_cuFileDriverSetMaxCacheSize": "CUfileError_t (size_t)",
+  "_cuFileDriverSetMaxDirectIOSize": "CUfileError_t (size_t)",
+  "_cuFileDriverSetMaxPinnedMemSize": "CUfileError_t (size_t)",
+  "_cuFileDriverSetPollMode": "CUfileError_t (bool, size_t)",
+  "_cuFileGetBARSizeInKB": "CUfileError_t (int, size_t *)",
+  "_cuFileGetParameterBool": "CUfileError_t (CUFileBoolConfigParameter_t, bool *)",
+  "_cuFileGetParameterMinMaxValue": "CUfileError_t (CUFileSizeTConfigParameter_t, size_t *, size_t *)",
+  "_cuFileGetParameterPosixPoolSlabArray": "CUfileError_t (size_t *, size_t *, int)",
+  "_cuFileGetParameterSizeT": "CUfileError_t (CUFileSizeTConfigParameter_t, size_t *)",
+  "_cuFileGetParameterString": "CUfileError_t (CUFileStringConfigParameter_t, char *, int)",
+  "_cuFileGetStatsL1": "CUfileError_t (CUfileStatsLevel1_t *)",
+  "_cuFileGetStatsL2": "CUfileError_t (CUfileStatsLevel2_t *)",
+  "_cuFileGetStatsL3": "CUfileError_t (CUfileStatsLevel3_t *)",
+  "_cuFileGetStatsLevel": "CUfileError_t (int *)",
+  "_cuFileGetVersion": "CUfileError_t (int *)",
+  "_cuFileHandleDeregister": "void (CUfileHandle_t)",
+  "_cuFileHandleRegister": "CUfileError_t (CUfileHandle_t *, CUfileDescr_t *)",
+  "_cuFileRead": "Py_ssize_t (CUfileHandle_t, void *, size_t, off_t, off_t)",
+  "_cuFileReadAsync": "CUfileError_t (CUfileHandle_t, void *, size_t *, off_t *, off_t *, Py_ssize_t *, CUstream)",
+  "_cuFileSetParameterBool": "CUfileError_t (CUFileBoolConfigParameter_t, bool)",
+  "_cuFileSetParameterPosixPoolSlabArray": "CUfileError_t (size_t const *, size_t const *, int)",
+  "_cuFileSetParameterSizeT": "CUfileError_t (CUFileSizeTConfigParameter_t, size_t)",
+  "_cuFileSetParameterString": "CUfileError_t (CUFileStringConfigParameter_t, char const *)",
+  "_cuFileSetStatsLevel": "CUfileError_t (int)",
+  "_cuFileStatsReset": "CUfileError_t (void)",
+  "_cuFileStatsStart": "CUfileError_t (void)",
+  "_cuFileStatsStop": "CUfileError_t (void)",
+  "_cuFileStreamDeregister": "CUfileError_t (CUstream)",
+  "_cuFileStreamRegister": "CUfileError_t (CUstream, unsigned int)",
+  "_cuFileUseCount": "long (void)",
+  "_cuFileWrite": "Py_ssize_t (CUfileHandle_t, void const *, size_t, off_t, off_t)",
+  "_cuFileWriteAsync": "CUfileError_t (CUfileHandle_t, void *, size_t *, off_t *, off_t *, Py_ssize_t *, CUstream)"
+}
diff --git a/cuda_bindings/tests/abi/cuda/bindings/_internal/nvjitlink.abi.json b/cuda_bindings/tests/abi/cuda/bindings/_internal/nvjitlink.abi.json
new file mode 100644
index 000000000..4f8ddd0f4
--- /dev/null
+++ b/cuda_bindings/tests/abi/cuda/bindings/_internal/nvjitlink.abi.json
@@ -0,0 +1,16 @@
+{
+  "_nvJitLinkAddData": "nvJitLinkResult (nvJitLinkHandle, nvJitLinkInputType, void const *, size_t, char const *)",
+  "_nvJitLinkAddFile": "nvJitLinkResult (nvJitLinkHandle, nvJitLinkInputType, char const *)",
+  "_nvJitLinkComplete": "nvJitLinkResult (nvJitLinkHandle)",
+  "_nvJitLinkCreate": "nvJitLinkResult (nvJitLinkHandle *, uint32_t, char const **)",
+  "_nvJitLinkDestroy": "nvJitLinkResult (nvJitLinkHandle *)",
+  "_nvJitLinkGetErrorLog": "nvJitLinkResult (nvJitLinkHandle, char *)",
+  "_nvJitLinkGetErrorLogSize": "nvJitLinkResult (nvJitLinkHandle, size_t *)",
+  "_nvJitLinkGetInfoLog": "nvJitLinkResult (nvJitLinkHandle, char *)",
+  "_nvJitLinkGetInfoLogSize": "nvJitLinkResult (nvJitLinkHandle, size_t *)",
+  "_nvJitLinkGetLinkedCubin": "nvJitLinkResult (nvJitLinkHandle, void *)",
+  "_nvJitLinkGetLinkedCubinSize": "nvJitLinkResult (nvJitLinkHandle, size_t *)",
+  "_nvJitLinkGetLinkedPtx": "nvJitLinkResult (nvJitLinkHandle, char *)",
+  "_nvJitLinkGetLinkedPtxSize": "nvJitLinkResult (nvJitLinkHandle, size_t *)",
+  "_nvJitLinkVersion": "nvJitLinkResult (unsigned int *, unsigned int *)"
+}
diff --git a/cuda_bindings/tests/abi/cuda/bindings/_internal/nvvm.abi.json b/cuda_bindings/tests/abi/cuda/bindings/_internal/nvvm.abi.json
new file mode 100644
index 000000000..8fca1d807
--- /dev/null
+++ b/cuda_bindings/tests/abi/cuda/bindings/_internal/nvvm.abi.json
@@ -0,0 +1,15 @@
+{
+  "_nvvmAddModuleToProgram": "nvvmResult (nvvmProgram, char const *, size_t, char const *)",
+  "_nvvmCompileProgram": "nvvmResult (nvvmProgram, int, char const **)",
+  "_nvvmCreateProgram": "nvvmResult (nvvmProgram *)",
+  "_nvvmDestroyProgram": "nvvmResult (nvvmProgram *)",
+  "_nvvmGetCompiledResult": "nvvmResult (nvvmProgram, char *)",
+  "_nvvmGetCompiledResultSize": "nvvmResult (nvvmProgram, size_t *)",
+  "_nvvmGetErrorString": "char const *(nvvmResult)",
+  "_nvvmGetProgramLog": "nvvmResult (nvvmProgram, char *)",
+  "_nvvmGetProgramLogSize": "nvvmResult (nvvmProgram, size_t *)",
+  "_nvvmIRVersion": "nvvmResult (int *, int *, int *, int *)",
+  "_nvvmLazyAddModuleToProgram": "nvvmResult (nvvmProgram, char const *, size_t, char const *)",
+  "_nvvmVerifyProgram": "nvvmResult (nvvmProgram, int, char const **)",
+  "_nvvmVersion": "nvvmResult (int *, int *)"
+}
diff --git a/cuda_bindings/tests/abi/cuda/bindings/_internal/utils.abi.json b/cuda_bindings/tests/abi/cuda/bindings/_internal/utils.abi.json
new file mode 100644
index 000000000..4e42aeb28
--- /dev/null
+++ b/cuda_bindings/tests/abi/cuda/bindings/_internal/utils.abi.json
@@ -0,0 +1,17 @@
+{
+  "__pyx_fuse_0get_nested_resource_ptr": "int (__pyx_t_4cuda_8bindings_9_internal_5utils_nested_resource<int>  &, PyObject *, int *)",
+  "__pyx_fuse_0get_resource_ptr": "int (nullable_unique_ptr<std::vector<int> >  &, PyObject *, int *)",
+  "__pyx_fuse_0get_resource_ptrs": "int (nullable_unique_ptr<std::vector<void *> >  &, PyObject *, void *)",
+  "__pyx_fuse_1get_nested_resource_ptr": "int (__pyx_t_4cuda_8bindings_9_internal_5utils_nested_resource<int32_t>  &, PyObject *, int32_t *)",
+  "__pyx_fuse_1get_resource_ptr": "int (nullable_unique_ptr<std::vector<int32_t> >  &, PyObject *, int32_t *)",
+  "__pyx_fuse_2get_nested_resource_ptr": "int (__pyx_t_4cuda_8bindings_9_internal_5utils_nested_resource<int64_t>  &, PyObject *, int64_t *)",
+  "__pyx_fuse_2get_resource_ptr": "int (nullable_unique_ptr<std::vector<int64_t> >  &, PyObject *, int64_t *)",
+  "__pyx_fuse_3get_nested_resource_ptr": "int (__pyx_t_4cuda_8bindings_9_internal_5utils_nested_resource<char>  &, PyObject *, char *)",
+  "__pyx_fuse_3get_resource_ptr": "int (nullable_unique_ptr<std::vector<char> >  &, PyObject *, char *)",
+  "__pyx_fuse_4get_nested_resource_ptr": "int (__pyx_t_4cuda_8bindings_9_internal_5utils_nested_resource<float>  &, PyObject *, float *)",
+  "__pyx_fuse_4get_resource_ptr": "int (nullable_unique_ptr<std::vector<float> >  &, PyObject *, float *)",
+  "__pyx_fuse_5get_nested_resource_ptr": "int (__pyx_t_4cuda_8bindings_9_internal_5utils_nested_resource<double>  &, PyObject *, double *)",
+  "__pyx_fuse_5get_resource_ptr": "int (nullable_unique_ptr<std::vector<double> >  &, PyObject *, double *)",
+  "get_buffer_pointer": "void *(PyObject *, Py_ssize_t, struct __pyx_opt_args_4cuda_8bindings_9_internal_5utils_get_buffer_pointer *__pyx_optional_args)",
+  "is_nested_sequence": "int (PyObject *)"
+}
diff --git a/cuda_bindings/tests/abi/cuda/bindings/cufile.abi.json b/cuda_bindings/tests/abi/cuda/bindings/cufile.abi.json
new file mode 100644
index 000000000..3d4c06193
--- /dev/null
+++ b/cuda_bindings/tests/abi/cuda/bindings/cufile.abi.json
@@ -0,0 +1,31 @@
+{
+  "batch_io_cancel": "PyObject *(intptr_t, int __pyx_skip_dispatch)",
+  "batch_io_destroy": "void (intptr_t, int __pyx_skip_dispatch)",
+  "batch_io_get_status": "PyObject *(intptr_t, unsigned int, intptr_t, intptr_t, intptr_t, int __pyx_skip_dispatch)",
+  "batch_io_set_up": "intptr_t (unsigned int, int __pyx_skip_dispatch)",
+  "batch_io_submit": "PyObject *(intptr_t, unsigned int, intptr_t, unsigned int, int __pyx_skip_dispatch)",
+  "buf_deregister": "PyObject *(intptr_t, int __pyx_skip_dispatch)",
+  "buf_register": "PyObject *(intptr_t, size_t, int, int __pyx_skip_dispatch)",
+  "driver_get_properties": "PyObject *(intptr_t, int __pyx_skip_dispatch)",
+  "driver_open": "PyObject *(int __pyx_skip_dispatch)",
+  "driver_set_max_cache_size": "PyObject *(size_t, int __pyx_skip_dispatch)",
+  "driver_set_max_direct_io_size": "PyObject *(size_t, int __pyx_skip_dispatch)",
+  "driver_set_max_pinned_mem_size": "PyObject *(size_t, int __pyx_skip_dispatch)",
+  "driver_set_poll_mode": "PyObject *(int, size_t, int __pyx_skip_dispatch)",
+  "get_parameter_bool": "int (int, int __pyx_skip_dispatch)",
+  "get_parameter_size_t": "size_t (int, int __pyx_skip_dispatch)",
+  "get_parameter_string": "PyObject *(int, int, int __pyx_skip_dispatch)",
+  "get_version": "int (int __pyx_skip_dispatch)",
+  "handle_deregister": "void (intptr_t, int __pyx_skip_dispatch)",
+  "handle_register": "intptr_t (intptr_t, int __pyx_skip_dispatch)",
+  "read": "PyObject *(intptr_t, intptr_t, size_t, off_t, off_t, int __pyx_skip_dispatch)",
+  "read_async": "PyObject *(intptr_t, intptr_t, intptr_t, intptr_t, intptr_t, intptr_t, intptr_t, int __pyx_skip_dispatch)",
+  "set_parameter_bool": "PyObject *(int, int, int __pyx_skip_dispatch)",
+  "set_parameter_size_t": "PyObject *(int, size_t, int __pyx_skip_dispatch)",
+  "set_parameter_string": "PyObject *(int, intptr_t, int __pyx_skip_dispatch)",
+  "stream_deregister": "PyObject *(intptr_t, int __pyx_skip_dispatch)",
+  "stream_register": "PyObject *(intptr_t, unsigned int, int __pyx_skip_dispatch)",
+  "use_count": "PyObject *(int __pyx_skip_dispatch)",
+  "write": "PyObject *(intptr_t, intptr_t, size_t, off_t, off_t, int __pyx_skip_dispatch)",
+  "write_async": "PyObject *(intptr_t, intptr_t, intptr_t, intptr_t, intptr_t, intptr_t, intptr_t, int __pyx_skip_dispatch)"
+}
diff --git a/cuda_bindings/tests/abi/cuda/bindings/cycufile.abi.json b/cuda_bindings/tests/abi/cuda/bindings/cycufile.abi.json
new file mode 100644
index 000000000..6790befeb
--- /dev/null
+++ b/cuda_bindings/tests/abi/cuda/bindings/cycufile.abi.json
@@ -0,0 +1,45 @@
+{
+  "cuFileBatchIOCancel": "CUfileError_t (CUfileBatchHandle_t)",
+  "cuFileBatchIODestroy": "void (CUfileBatchHandle_t)",
+  "cuFileBatchIOGetStatus": "CUfileError_t (CUfileBatchHandle_t, unsigned int, unsigned int *, CUfileIOEvents_t *, struct timespec *)",
+  "cuFileBatchIOSetUp": "CUfileError_t (CUfileBatchHandle_t *, unsigned int)",
+  "cuFileBatchIOSubmit": "CUfileError_t (CUfileBatchHandle_t, unsigned int, CUfileIOParams_t *, unsigned int)",
+  "cuFileBufDeregister": "CUfileError_t (void const *)",
+  "cuFileBufRegister": "CUfileError_t (void const *, size_t, int)",
+  "cuFileDriverClose": "CUfileError_t (void)",
+  "cuFileDriverClose_v2": "CUfileError_t (void)",
+  "cuFileDriverGetProperties": "CUfileError_t (CUfileDrvProps_t *)",
+  "cuFileDriverOpen": "CUfileError_t (void)",
+  "cuFileDriverSetMaxCacheSize": "CUfileError_t (size_t)",
+  "cuFileDriverSetMaxDirectIOSize": "CUfileError_t (size_t)",
+  "cuFileDriverSetMaxPinnedMemSize": "CUfileError_t (size_t)",
+  "cuFileDriverSetPollMode": "CUfileError_t (bool, size_t)",
+  "cuFileGetBARSizeInKB": "CUfileError_t (int, size_t *)",
+  "cuFileGetParameterBool": "CUfileError_t (CUFileBoolConfigParameter_t, bool *)",
+  "cuFileGetParameterMinMaxValue": "CUfileError_t (CUFileSizeTConfigParameter_t, size_t *, size_t *)",
+  "cuFileGetParameterPosixPoolSlabArray": "CUfileError_t (size_t *, size_t *, int)",
+  "cuFileGetParameterSizeT": "CUfileError_t (CUFileSizeTConfigParameter_t, size_t *)",
+  "cuFileGetParameterString": "CUfileError_t (CUFileStringConfigParameter_t, char *, int)",
+  "cuFileGetStatsL1": "CUfileError_t (CUfileStatsLevel1_t *)",
+  "cuFileGetStatsL2": "CUfileError_t (CUfileStatsLevel2_t *)",
+  "cuFileGetStatsL3": "CUfileError_t (CUfileStatsLevel3_t *)",
+  "cuFileGetStatsLevel": "CUfileError_t (int *)",
+  "cuFileGetVersion": "CUfileError_t (int *)",
+  "cuFileHandleDeregister": "void (CUfileHandle_t)",
+  "cuFileHandleRegister": "CUfileError_t (CUfileHandle_t *, CUfileDescr_t *)",
+  "cuFileRead": "Py_ssize_t (CUfileHandle_t, void *, size_t, off_t, off_t)",
+  "cuFileReadAsync": "CUfileError_t (CUfileHandle_t, void *, size_t *, off_t *, off_t *, Py_ssize_t *, CUstream)",
+  "cuFileSetParameterBool": "CUfileError_t (CUFileBoolConfigParameter_t, bool)",
+  "cuFileSetParameterPosixPoolSlabArray": "CUfileError_t (size_t const *, size_t const *, int)",
+  "cuFileSetParameterSizeT": "CUfileError_t (CUFileSizeTConfigParameter_t, size_t)",
+  "cuFileSetParameterString": "CUfileError_t (CUFileStringConfigParameter_t, char const *)",
+  "cuFileSetStatsLevel": "CUfileError_t (int)",
+  "cuFileStatsReset": "CUfileError_t (void)",
+  "cuFileStatsStart": "CUfileError_t (void)",
+  "cuFileStatsStop": "CUfileError_t (void)",
+  "cuFileStreamDeregister": "CUfileError_t (CUstream)",
+  "cuFileStreamRegister": "CUfileError_t (CUstream, unsigned int)",
+  "cuFileUseCount": "long (void)",
+  "cuFileWrite": "Py_ssize_t (CUfileHandle_t, void const *, size_t, off_t, off_t)",
+  "cuFileWriteAsync": "CUfileError_t (CUfileHandle_t, void *, size_t *, off_t *, off_t *, Py_ssize_t *, CUstream)"
+}
diff --git a/cuda_bindings/tests/abi/cuda/bindings/cydriver.abi.json b/cuda_bindings/tests/abi/cuda/bindings/cydriver.abi.json
new file mode 100644
index 000000000..93e96d72d
--- /dev/null
+++ b/cuda_bindings/tests/abi/cuda/bindings/cydriver.abi.json
@@ -0,0 +1,472 @@
+{
+  "cuArray3DCreate": "CUresult (CUarray *, CUDA_ARRAY3D_DESCRIPTOR const *)",
+  "cuArray3DGetDescriptor": "CUresult (CUDA_ARRAY3D_DESCRIPTOR *, CUarray)",
+  "cuArrayCreate": "CUresult (CUarray *, CUDA_ARRAY_DESCRIPTOR const *)",
+  "cuArrayDestroy": "CUresult (CUarray)",
+  "cuArrayGetDescriptor": "CUresult (CUDA_ARRAY_DESCRIPTOR *, CUarray)",
+  "cuArrayGetMemoryRequirements": "CUresult (CUDA_ARRAY_MEMORY_REQUIREMENTS *, CUarray, CUdevice)",
+  "cuArrayGetPlane": "CUresult (CUarray *, CUarray, unsigned int)",
+  "cuArrayGetSparseProperties": "CUresult (CUDA_ARRAY_SPARSE_PROPERTIES *, CUarray)",
+  "cuCheckpointProcessCheckpoint": "CUresult (int, CUcheckpointCheckpointArgs *)",
+  "cuCheckpointProcessGetRestoreThreadId": "CUresult (int, int *)",
+  "cuCheckpointProcessGetState": "CUresult (int, CUprocessState *)",
+  "cuCheckpointProcessLock": "CUresult (int, CUcheckpointLockArgs *)",
+  "cuCheckpointProcessRestore": "CUresult (int, CUcheckpointRestoreArgs *)",
+  "cuCheckpointProcessUnlock": "CUresult (int, CUcheckpointUnlockArgs *)",
+  "cuCoredumpGetAttribute": "CUresult (CUcoredumpSettings, void *, size_t *)",
+  "cuCoredumpGetAttributeGlobal": "CUresult (CUcoredumpSettings, void *, size_t *)",
+  "cuCoredumpSetAttribute": "CUresult (CUcoredumpSettings, void *, size_t *)",
+  "cuCoredumpSetAttributeGlobal": "CUresult (CUcoredumpSettings, void *, size_t *)",
+  "cuCtxAttach": "CUresult (CUcontext *, unsigned int)",
+  "cuCtxCreate": "CUresult (CUcontext *, CUctxCreateParams *, unsigned int, CUdevice)",
+  "cuCtxDestroy": "CUresult (CUcontext)",
+  "cuCtxDetach": "CUresult (CUcontext)",
+  "cuCtxDisablePeerAccess": "CUresult (CUcontext)",
+  "cuCtxEnablePeerAccess": "CUresult (CUcontext, unsigned int)",
+  "cuCtxFromGreenCtx": "CUresult (CUcontext *, CUgreenCtx)",
+  "cuCtxGetApiVersion": "CUresult (CUcontext, unsigned int *)",
+  "cuCtxGetCacheConfig": "CUresult (CUfunc_cache *)",
+  "cuCtxGetCurrent": "CUresult (CUcontext *)",
+  "cuCtxGetDevResource": "CUresult (CUcontext, CUdevResource *, CUdevResourceType)",
+  "cuCtxGetDevice": "CUresult (CUdevice *)",
+  "cuCtxGetDevice_v2": "CUresult (CUdevice *, CUcontext)",
+  "cuCtxGetExecAffinity": "CUresult (CUexecAffinityParam *, CUexecAffinityType)",
+  "cuCtxGetFlags": "CUresult (unsigned int *)",
+  "cuCtxGetId": "CUresult (CUcontext, unsigned PY_LONG_LONG *)",
+  "cuCtxGetLimit": "CUresult (size_t *, CUlimit)",
+  "cuCtxGetSharedMemConfig": "CUresult (CUsharedconfig *)",
+  "cuCtxGetStreamPriorityRange": "CUresult (int *, int *)",
+  "cuCtxPopCurrent": "CUresult (CUcontext *)",
+  "cuCtxPushCurrent": "CUresult (CUcontext)",
+  "cuCtxRecordEvent": "CUresult (CUcontext, CUevent)",
+  "cuCtxResetPersistingL2Cache": "CUresult (void)",
+  "cuCtxSetCacheConfig": "CUresult (CUfunc_cache)",
+  "cuCtxSetCurrent": "CUresult (CUcontext)",
+  "cuCtxSetFlags": "CUresult (unsigned int)",
+  "cuCtxSetLimit": "CUresult (CUlimit, size_t)",
+  "cuCtxSetSharedMemConfig": "CUresult (CUsharedconfig)",
+  "cuCtxSynchronize": "CUresult (void)",
+  "cuCtxSynchronize_v2": "CUresult (CUcontext)",
+  "cuCtxWaitEvent": "CUresult (CUcontext, CUevent)",
+  "cuDestroyExternalMemory": "CUresult (CUexternalMemory)",
+  "cuDestroyExternalSemaphore": "CUresult (CUexternalSemaphore)",
+  "cuDevResourceGenerateDesc": "CUresult (CUdevResourceDesc *, CUdevResource *, unsigned int)",
+  "cuDevSmResourceSplitByCount": "CUresult (CUdevResource *, unsigned int *, CUdevResource const *, CUdevResource *, unsigned int, unsigned int)",
+  "cuDeviceCanAccessPeer": "CUresult (int *, CUdevice, CUdevice)",
+  "cuDeviceComputeCapability": "CUresult (int *, int *, CUdevice)",
+  "cuDeviceGet": "CUresult (CUdevice *, int)",
+  "cuDeviceGetAttribute": "CUresult (int *, CUdevice_attribute, CUdevice)",
+  "cuDeviceGetByPCIBusId": "CUresult (CUdevice *, char const *)",
+  "cuDeviceGetCount": "CUresult (int *)",
+  "cuDeviceGetDefaultMemPool": "CUresult (CUmemoryPool *, CUdevice)",
+  "cuDeviceGetDevResource": "CUresult (CUdevice, CUdevResource *, CUdevResourceType)",
+  "cuDeviceGetExecAffinitySupport": "CUresult (int *, CUexecAffinityType, CUdevice)",
+  "cuDeviceGetGraphMemAttribute": "CUresult (CUdevice, CUgraphMem_attribute, void *)",
+  "cuDeviceGetHostAtomicCapabilities": "CUresult (unsigned int *, CUatomicOperation const *, unsigned int, CUdevice)",
+  "cuDeviceGetLuid": "CUresult (char *, unsigned int *, CUdevice)",
+  "cuDeviceGetMemPool": "CUresult (CUmemoryPool *, CUdevice)",
+  "cuDeviceGetName": "CUresult (char *, int, CUdevice)",
+  "cuDeviceGetNvSciSyncAttributes": "CUresult (void *, CUdevice, int)",
+  "cuDeviceGetP2PAtomicCapabilities": "CUresult (unsigned int *, CUatomicOperation const *, unsigned int, CUdevice, CUdevice)",
+  "cuDeviceGetP2PAttribute": "CUresult (int *, CUdevice_P2PAttribute, CUdevice, CUdevice)",
+  "cuDeviceGetPCIBusId": "CUresult (char *, int, CUdevice)",
+  "cuDeviceGetProperties": "CUresult (CUdevprop *, CUdevice)",
+  "cuDeviceGetTexture1DLinearMaxWidth": "CUresult (size_t *, CUarray_format, unsigned int, CUdevice)",
+  "cuDeviceGetUuid": "CUresult (CUuuid *, CUdevice)",
+  "cuDeviceGraphMemTrim": "CUresult (CUdevice)",
+  "cuDevicePrimaryCtxGetState": "CUresult (CUdevice, unsigned int *, int *)",
+  "cuDevicePrimaryCtxRelease": "CUresult (CUdevice)",
+  "cuDevicePrimaryCtxReset": "CUresult (CUdevice)",
+  "cuDevicePrimaryCtxRetain": "CUresult (CUcontext *, CUdevice)",
+  "cuDevicePrimaryCtxSetFlags": "CUresult (CUdevice, unsigned int)",
+  "cuDeviceRegisterAsyncNotification": "CUresult (CUdevice, CUasyncCallback, void *, CUasyncCallbackHandle *)",
+  "cuDeviceSetGraphMemAttribute": "CUresult (CUdevice, CUgraphMem_attribute, void *)",
+  "cuDeviceSetMemPool": "CUresult (CUdevice, CUmemoryPool)",
+  "cuDeviceTotalMem": "CUresult (size_t *, CUdevice)",
+  "cuDeviceUnregisterAsyncNotification": "CUresult (CUdevice, CUasyncCallbackHandle)",
+  "cuDriverGetVersion": "CUresult (int *)",
+  "cuEGLStreamConsumerAcquireFrame": "CUresult (__pyx_t_4cuda_8bindings_8cydriver_CUeglStreamConnection *, CUgraphicsResource *, CUstream *, unsigned int)",
+  "cuEGLStreamConsumerConnect": "CUresult (__pyx_t_4cuda_8bindings_8cydriver_CUeglStreamConnection *, __pyx_t_4cuda_8bindings_8cydriver_EGLStreamKHR)",
+  "cuEGLStreamConsumerConnectWithFlags": "CUresult (__pyx_t_4cuda_8bindings_8cydriver_CUeglStreamConnection *, __pyx_t_4cuda_8bindings_8cydriver_EGLStreamKHR, unsigned int)",
+  "cuEGLStreamConsumerDisconnect": "CUresult (__pyx_t_4cuda_8bindings_8cydriver_CUeglStreamConnection *)",
+  "cuEGLStreamConsumerReleaseFrame": "CUresult (__pyx_t_4cuda_8bindings_8cydriver_CUeglStreamConnection *, CUgraphicsResource, CUstream *)",
+  "cuEGLStreamProducerConnect": "CUresult (__pyx_t_4cuda_8bindings_8cydriver_CUeglStreamConnection *, __pyx_t_4cuda_8bindings_8cydriver_EGLStreamKHR, __pyx_t_4cuda_8bindings_8cydriver_EGLint, __pyx_t_4cuda_8bindings_8cydriver_EGLint)",
+  "cuEGLStreamProducerDisconnect": "CUresult (__pyx_t_4cuda_8bindings_8cydriver_CUeglStreamConnection *)",
+  "cuEGLStreamProducerPresentFrame": "CUresult (__pyx_t_4cuda_8bindings_8cydriver_CUeglStreamConnection *, __pyx_t_4cuda_8bindings_8cydriver_CUeglFrame, CUstream *)",
+  "cuEGLStreamProducerReturnFrame": "CUresult (__pyx_t_4cuda_8bindings_8cydriver_CUeglStreamConnection *, __pyx_t_4cuda_8bindings_8cydriver_CUeglFrame *, CUstream *)",
+  "cuEventCreate": "CUresult (CUevent *, unsigned int)",
+  "cuEventCreateFromEGLSync": "CUresult (CUevent *, __pyx_t_4cuda_8bindings_8cydriver_EGLSyncKHR, unsigned int)",
+  "cuEventDestroy": "CUresult (CUevent)",
+  "cuEventElapsedTime": "CUresult (float *, CUevent, CUevent)",
+  "cuEventQuery": "CUresult (CUevent)",
+  "cuEventRecord": "CUresult (CUevent, CUstream)",
+  "cuEventRecordWithFlags": "CUresult (CUevent, CUstream, unsigned int)",
+  "cuEventSynchronize": "CUresult (CUevent)",
+  "cuExternalMemoryGetMappedBuffer": "CUresult (CUdeviceptr *, CUexternalMemory, CUDA_EXTERNAL_MEMORY_BUFFER_DESC const *)",
+  "cuExternalMemoryGetMappedMipmappedArray": "CUresult (CUmipmappedArray *, CUexternalMemory, CUDA_EXTERNAL_MEMORY_MIPMAPPED_ARRAY_DESC const *)",
+  "cuFlushGPUDirectRDMAWrites": "CUresult (CUflushGPUDirectRDMAWritesTarget, CUflushGPUDirectRDMAWritesScope)",
+  "cuFuncGetAttribute": "CUresult (int *, CUfunction_attribute, CUfunction)",
+  "cuFuncGetModule": "CUresult (CUmodule *, CUfunction)",
+  "cuFuncGetName": "CUresult (char const **, CUfunction)",
+  "cuFuncGetParamInfo": "CUresult (CUfunction, size_t, size_t *, size_t *)",
+  "cuFuncIsLoaded": "CUresult (CUfunctionLoadingState *, CUfunction)",
+  "cuFuncLoad": "CUresult (CUfunction)",
+  "cuFuncSetAttribute": "CUresult (CUfunction, CUfunction_attribute, int)",
+  "cuFuncSetBlockShape": "CUresult (CUfunction, int, int, int)",
+  "cuFuncSetCacheConfig": "CUresult (CUfunction, CUfunc_cache)",
+  "cuFuncSetSharedMemConfig": "CUresult (CUfunction, CUsharedconfig)",
+  "cuFuncSetSharedSize": "CUresult (CUfunction, unsigned int)",
+  "cuGLGetDevices": "CUresult (unsigned int *, CUdevice *, unsigned int, __pyx_t_4cuda_8bindings_8cydriver_CUGLDeviceList)",
+  "cuGetErrorName": "CUresult (CUresult, char const **)",
+  "cuGetErrorString": "CUresult (CUresult, char const **)",
+  "cuGetExportTable": "CUresult (void const **, CUuuid const *)",
+  "cuGetProcAddress": "CUresult (char const *, void **, int, cuuint64_t, CUdriverProcAddressQueryResult *)",
+  "cuGraphAddBatchMemOpNode": "CUresult (CUgraphNode *, CUgraph, CUgraphNode const *, size_t, CUDA_BATCH_MEM_OP_NODE_PARAMS const *)",
+  "cuGraphAddChildGraphNode": "CUresult (CUgraphNode *, CUgraph, CUgraphNode const *, size_t, CUgraph)",
+  "cuGraphAddDependencies": "CUresult (CUgraph, CUgraphNode const *, CUgraphNode const *, CUgraphEdgeData const *, size_t)",
+  "cuGraphAddEmptyNode": "CUresult (CUgraphNode *, CUgraph, CUgraphNode const *, size_t)",
+  "cuGraphAddEventRecordNode": "CUresult (CUgraphNode *, CUgraph, CUgraphNode const *, size_t, CUevent)",
+  "cuGraphAddEventWaitNode": "CUresult (CUgraphNode *, CUgraph, CUgraphNode const *, size_t, CUevent)",
+  "cuGraphAddExternalSemaphoresSignalNode": "CUresult (CUgraphNode *, CUgraph, CUgraphNode const *, size_t, CUDA_EXT_SEM_SIGNAL_NODE_PARAMS const *)",
+  "cuGraphAddExternalSemaphoresWaitNode": "CUresult (CUgraphNode *, CUgraph, CUgraphNode const *, size_t, CUDA_EXT_SEM_WAIT_NODE_PARAMS const *)",
+  "cuGraphAddHostNode": "CUresult (CUgraphNode *, CUgraph, CUgraphNode const *, size_t, CUDA_HOST_NODE_PARAMS const *)",
+  "cuGraphAddKernelNode": "CUresult (CUgraphNode *, CUgraph, CUgraphNode const *, size_t, CUDA_KERNEL_NODE_PARAMS const *)",
+  "cuGraphAddMemAllocNode": "CUresult (CUgraphNode *, CUgraph, CUgraphNode const *, size_t, CUDA_MEM_ALLOC_NODE_PARAMS *)",
+  "cuGraphAddMemFreeNode": "CUresult (CUgraphNode *, CUgraph, CUgraphNode const *, size_t, CUdeviceptr)",
+  "cuGraphAddMemcpyNode": "CUresult (CUgraphNode *, CUgraph, CUgraphNode const *, size_t, CUDA_MEMCPY3D const *, CUcontext)",
+  "cuGraphAddMemsetNode": "CUresult (CUgraphNode *, CUgraph, CUgraphNode const *, size_t, CUDA_MEMSET_NODE_PARAMS const *, CUcontext)",
+  "cuGraphAddNode": "CUresult (CUgraphNode *, CUgraph, CUgraphNode const *, CUgraphEdgeData const *, size_t, CUgraphNodeParams *)",
+  "cuGraphBatchMemOpNodeGetParams": "CUresult (CUgraphNode, CUDA_BATCH_MEM_OP_NODE_PARAMS *)",
+  "cuGraphBatchMemOpNodeSetParams": "CUresult (CUgraphNode, CUDA_BATCH_MEM_OP_NODE_PARAMS const *)",
+  "cuGraphChildGraphNodeGetGraph": "CUresult (CUgraphNode, CUgraph *)",
+  "cuGraphClone": "CUresult (CUgraph *, CUgraph)",
+  "cuGraphConditionalHandleCreate": "CUresult (CUgraphConditionalHandle *, CUgraph, CUcontext, unsigned int, unsigned int)",
+  "cuGraphCreate": "CUresult (CUgraph *, unsigned int)",
+  "cuGraphDebugDotPrint": "CUresult (CUgraph, char const *, unsigned int)",
+  "cuGraphDestroy": "CUresult (CUgraph)",
+  "cuGraphDestroyNode": "CUresult (CUgraphNode)",
+  "cuGraphEventRecordNodeGetEvent": "CUresult (CUgraphNode, CUevent *)",
+  "cuGraphEventRecordNodeSetEvent": "CUresult (CUgraphNode, CUevent)",
+  "cuGraphEventWaitNodeGetEvent": "CUresult (CUgraphNode, CUevent *)",
+  "cuGraphEventWaitNodeSetEvent": "CUresult (CUgraphNode, CUevent)",
+  "cuGraphExecBatchMemOpNodeSetParams": "CUresult (CUgraphExec, CUgraphNode, CUDA_BATCH_MEM_OP_NODE_PARAMS const *)",
+  "cuGraphExecChildGraphNodeSetParams": "CUresult (CUgraphExec, CUgraphNode, CUgraph)",
+  "cuGraphExecDestroy": "CUresult (CUgraphExec)",
+  "cuGraphExecEventRecordNodeSetEvent": "CUresult (CUgraphExec, CUgraphNode, CUevent)",
+  "cuGraphExecEventWaitNodeSetEvent": "CUresult (CUgraphExec, CUgraphNode, CUevent)",
+  "cuGraphExecExternalSemaphoresSignalNodeSetParams": "CUresult (CUgraphExec, CUgraphNode, CUDA_EXT_SEM_SIGNAL_NODE_PARAMS const *)",
+  "cuGraphExecExternalSemaphoresWaitNodeSetParams": "CUresult (CUgraphExec, CUgraphNode, CUDA_EXT_SEM_WAIT_NODE_PARAMS const *)",
+  "cuGraphExecGetFlags": "CUresult (CUgraphExec, cuuint64_t *)",
+  "cuGraphExecHostNodeSetParams": "CUresult (CUgraphExec, CUgraphNode, CUDA_HOST_NODE_PARAMS const *)",
+  "cuGraphExecKernelNodeSetParams": "CUresult (CUgraphExec, CUgraphNode, CUDA_KERNEL_NODE_PARAMS const *)",
+  "cuGraphExecMemcpyNodeSetParams": "CUresult (CUgraphExec, CUgraphNode, CUDA_MEMCPY3D const *, CUcontext)",
+  "cuGraphExecMemsetNodeSetParams": "CUresult (CUgraphExec, CUgraphNode, CUDA_MEMSET_NODE_PARAMS const *, CUcontext)",
+  "cuGraphExecNodeSetParams": "CUresult (CUgraphExec, CUgraphNode, CUgraphNodeParams *)",
+  "cuGraphExecUpdate": "CUresult (CUgraphExec, CUgraph, CUgraphExecUpdateResultInfo *)",
+  "cuGraphExternalSemaphoresSignalNodeGetParams": "CUresult (CUgraphNode, CUDA_EXT_SEM_SIGNAL_NODE_PARAMS *)",
+  "cuGraphExternalSemaphoresSignalNodeSetParams": "CUresult (CUgraphNode, CUDA_EXT_SEM_SIGNAL_NODE_PARAMS const *)",
+  "cuGraphExternalSemaphoresWaitNodeGetParams": "CUresult (CUgraphNode, CUDA_EXT_SEM_WAIT_NODE_PARAMS *)",
+  "cuGraphExternalSemaphoresWaitNodeSetParams": "CUresult (CUgraphNode, CUDA_EXT_SEM_WAIT_NODE_PARAMS const *)",
+  "cuGraphGetEdges": "CUresult (CUgraph, CUgraphNode *, CUgraphNode *, CUgraphEdgeData *, size_t *)",
+  "cuGraphGetNodes": "CUresult (CUgraph, CUgraphNode *, size_t *)",
+  "cuGraphGetRootNodes": "CUresult (CUgraph, CUgraphNode *, size_t *)",
+  "cuGraphHostNodeGetParams": "CUresult (CUgraphNode, CUDA_HOST_NODE_PARAMS *)",
+  "cuGraphHostNodeSetParams": "CUresult (CUgraphNode, CUDA_HOST_NODE_PARAMS const *)",
+  "cuGraphInstantiate": "CUresult (CUgraphExec *, CUgraph, unsigned PY_LONG_LONG)",
+  "cuGraphInstantiateWithParams": "CUresult (CUgraphExec *, CUgraph, CUDA_GRAPH_INSTANTIATE_PARAMS *)",
+  "cuGraphKernelNodeCopyAttributes": "CUresult (CUgraphNode, CUgraphNode)",
+  "cuGraphKernelNodeGetAttribute": "CUresult (CUgraphNode, CUkernelNodeAttrID, CUkernelNodeAttrValue *)",
+  "cuGraphKernelNodeGetParams": "CUresult (CUgraphNode, CUDA_KERNEL_NODE_PARAMS *)",
+  "cuGraphKernelNodeSetAttribute": "CUresult (CUgraphNode, CUkernelNodeAttrID, CUkernelNodeAttrValue const *)",
+  "cuGraphKernelNodeSetParams": "CUresult (CUgraphNode, CUDA_KERNEL_NODE_PARAMS const *)",
+  "cuGraphLaunch": "CUresult (CUgraphExec, CUstream)",
+  "cuGraphMemAllocNodeGetParams": "CUresult (CUgraphNode, CUDA_MEM_ALLOC_NODE_PARAMS *)",
+  "cuGraphMemFreeNodeGetParams": "CUresult (CUgraphNode, CUdeviceptr *)",
+  "cuGraphMemcpyNodeGetParams": "CUresult (CUgraphNode, CUDA_MEMCPY3D *)",
+  "cuGraphMemcpyNodeSetParams": "CUresult (CUgraphNode, CUDA_MEMCPY3D const *)",
+  "cuGraphMemsetNodeGetParams": "CUresult (CUgraphNode, CUDA_MEMSET_NODE_PARAMS *)",
+  "cuGraphMemsetNodeSetParams": "CUresult (CUgraphNode, CUDA_MEMSET_NODE_PARAMS const *)",
+  "cuGraphNodeFindInClone": "CUresult (CUgraphNode *, CUgraphNode, CUgraph)",
+  "cuGraphNodeGetDependencies": "CUresult (CUgraphNode, CUgraphNode *, CUgraphEdgeData *, size_t *)",
+  "cuGraphNodeGetDependentNodes": "CUresult (CUgraphNode, CUgraphNode *, CUgraphEdgeData *, size_t *)",
+  "cuGraphNodeGetEnabled": "CUresult (CUgraphExec, CUgraphNode, unsigned int *)",
+  "cuGraphNodeGetType": "CUresult (CUgraphNode, CUgraphNodeType *)",
+  "cuGraphNodeSetEnabled": "CUresult (CUgraphExec, CUgraphNode, unsigned int)",
+  "cuGraphNodeSetParams": "CUresult (CUgraphNode, CUgraphNodeParams *)",
+  "cuGraphReleaseUserObject": "CUresult (CUgraph, CUuserObject, unsigned int)",
+  "cuGraphRemoveDependencies": "CUresult (CUgraph, CUgraphNode const *, CUgraphNode const *, CUgraphEdgeData const *, size_t)",
+  "cuGraphRetainUserObject": "CUresult (CUgraph, CUuserObject, unsigned int, unsigned int)",
+  "cuGraphUpload": "CUresult (CUgraphExec, CUstream)",
+  "cuGraphicsEGLRegisterImage": "CUresult (CUgraphicsResource *, __pyx_t_4cuda_8bindings_8cydriver_EGLImageKHR, unsigned int)",
+  "cuGraphicsGLRegisterBuffer": "CUresult (CUgraphicsResource *, __pyx_t_4cuda_8bindings_8cydriver_GLuint, unsigned int)",
+  "cuGraphicsGLRegisterImage": "CUresult (CUgraphicsResource *, __pyx_t_4cuda_8bindings_8cydriver_GLuint, __pyx_t_4cuda_8bindings_8cydriver_GLenum, unsigned int)",
+  "cuGraphicsMapResources": "CUresult (unsigned int, CUgraphicsResource *, CUstream)",
+  "cuGraphicsResourceGetMappedEglFrame": "CUresult (__pyx_t_4cuda_8bindings_8cydriver_CUeglFrame *, CUgraphicsResource, unsigned int, unsigned int)",
+  "cuGraphicsResourceGetMappedMipmappedArray": "CUresult (CUmipmappedArray *, CUgraphicsResource)",
+  "cuGraphicsResourceGetMappedPointer": "CUresult (CUdeviceptr *, size_t *, CUgraphicsResource)",
+  "cuGraphicsResourceSetMapFlags": "CUresult (CUgraphicsResource, unsigned int)",
+  "cuGraphicsSubResourceGetMappedArray": "CUresult (CUarray *, CUgraphicsResource, unsigned int, unsigned int)",
+  "cuGraphicsUnmapResources": "CUresult (unsigned int, CUgraphicsResource *, CUstream)",
+  "cuGraphicsUnregisterResource": "CUresult (CUgraphicsResource)",
+  "cuGraphicsVDPAURegisterOutputSurface": "CUresult (CUgraphicsResource *, __pyx_t_4cuda_8bindings_8cydriver_VdpOutputSurface, unsigned int)",
+  "cuGraphicsVDPAURegisterVideoSurface": "CUresult (CUgraphicsResource *, __pyx_t_4cuda_8bindings_8cydriver_VdpVideoSurface, unsigned int)",
+  "cuGreenCtxCreate": "CUresult (CUgreenCtx *, CUdevResourceDesc, CUdevice, unsigned int)",
+  "cuGreenCtxDestroy": "CUresult (CUgreenCtx)",
+  "cuGreenCtxGetDevResource": "CUresult (CUgreenCtx, CUdevResource *, CUdevResourceType)",
+  "cuGreenCtxGetId": "CUresult (CUgreenCtx, unsigned PY_LONG_LONG *)",
+  "cuGreenCtxRecordEvent": "CUresult (CUgreenCtx, CUevent)",
+  "cuGreenCtxStreamCreate": "CUresult (CUstream *, CUgreenCtx, unsigned int, int)",
+  "cuGreenCtxWaitEvent": "CUresult (CUgreenCtx, CUevent)",
+  "cuImportExternalMemory": "CUresult (CUexternalMemory *, CUDA_EXTERNAL_MEMORY_HANDLE_DESC const *)",
+  "cuImportExternalSemaphore": "CUresult (CUexternalSemaphore *, CUDA_EXTERNAL_SEMAPHORE_HANDLE_DESC const *)",
+  "cuInit": "CUresult (unsigned int)",
+  "cuIpcCloseMemHandle": "CUresult (CUdeviceptr)",
+  "cuIpcGetEventHandle": "CUresult (CUipcEventHandle *, CUevent)",
+  "cuIpcGetMemHandle": "CUresult (CUipcMemHandle *, CUdeviceptr)",
+  "cuIpcOpenEventHandle": "CUresult (CUevent *, CUipcEventHandle)",
+  "cuIpcOpenMemHandle": "CUresult (CUdeviceptr *, CUipcMemHandle, unsigned int)",
+  "cuKernelGetAttribute": "CUresult (int *, CUfunction_attribute, CUkernel, CUdevice)",
+  "cuKernelGetFunction": "CUresult (CUfunction *, CUkernel)",
+  "cuKernelGetLibrary": "CUresult (CUlibrary *, CUkernel)",
+  "cuKernelGetName": "CUresult (char const **, CUkernel)",
+  "cuKernelGetParamInfo": "CUresult (CUkernel, size_t, size_t *, size_t *)",
+  "cuKernelSetAttribute": "CUresult (CUfunction_attribute, int, CUkernel, CUdevice)",
+  "cuKernelSetCacheConfig": "CUresult (CUkernel, CUfunc_cache, CUdevice)",
+  "cuLaunch": "CUresult (CUfunction)",
+  "cuLaunchCooperativeKernel": "CUresult (CUfunction, unsigned int, unsigned int, unsigned int, unsigned int, unsigned int, unsigned int, unsigned int, CUstream, void **)",
+  "cuLaunchCooperativeKernelMultiDevice": "CUresult (CUDA_LAUNCH_PARAMS *, unsigned int, unsigned int)",
+  "cuLaunchGrid": "CUresult (CUfunction, int, int)",
+  "cuLaunchGridAsync": "CUresult (CUfunction, int, int, CUstream)",
+  "cuLaunchHostFunc": "CUresult (CUstream, CUhostFn, void *)",
+  "cuLaunchKernel": "CUresult (CUfunction, unsigned int, unsigned int, unsigned int, unsigned int, unsigned int, unsigned int, unsigned int, CUstream, void **, void **)",
+  "cuLaunchKernelEx": "CUresult (CUlaunchConfig const *, CUfunction, void **, void **)",
+  "cuLibraryEnumerateKernels": "CUresult (CUkernel *, unsigned int, CUlibrary)",
+  "cuLibraryGetGlobal": "CUresult (CUdeviceptr *, size_t *, CUlibrary, char const *)",
+  "cuLibraryGetKernel": "CUresult (CUkernel *, CUlibrary, char const *)",
+  "cuLibraryGetKernelCount": "CUresult (unsigned int *, CUlibrary)",
+  "cuLibraryGetManaged": "CUresult (CUdeviceptr *, size_t *, CUlibrary, char const *)",
+  "cuLibraryGetModule": "CUresult (CUmodule *, CUlibrary)",
+  "cuLibraryGetUnifiedFunction": "CUresult (void **, CUlibrary, char const *)",
+  "cuLibraryLoadData": "CUresult (CUlibrary *, void const *, CUjit_option *, void **, unsigned int, CUlibraryOption *, void **, unsigned int)",
+  "cuLibraryLoadFromFile": "CUresult (CUlibrary *, char const *, CUjit_option *, void **, unsigned int, CUlibraryOption *, void **, unsigned int)",
+  "cuLibraryUnload": "CUresult (CUlibrary)",
+  "cuLinkAddData": "CUresult (CUlinkState, CUjitInputType, void *, size_t, char const *, unsigned int, CUjit_option *, void **)",
+  "cuLinkAddFile": "CUresult (CUlinkState, CUjitInputType, char const *, unsigned int, CUjit_option *, void **)",
+  "cuLinkComplete": "CUresult (CUlinkState, void **, size_t *)",
+  "cuLinkCreate": "CUresult (unsigned int, CUjit_option *, void **, CUlinkState *)",
+  "cuLinkDestroy": "CUresult (CUlinkState)",
+  "cuLogsCurrent": "CUresult (CUlogIterator *, unsigned int)",
+  "cuLogsDumpToFile": "CUresult (CUlogIterator *, char const *, unsigned int)",
+  "cuLogsDumpToMemory": "CUresult (CUlogIterator *, char *, size_t *, unsigned int)",
+  "cuLogsRegisterCallback": "CUresult (CUlogsCallback, void *, CUlogsCallbackHandle *)",
+  "cuLogsUnregisterCallback": "CUresult (CUlogsCallbackHandle)",
+  "cuMemAddressFree": "CUresult (CUdeviceptr, size_t)",
+  "cuMemAddressReserve": "CUresult (CUdeviceptr *, size_t, size_t, CUdeviceptr, unsigned PY_LONG_LONG)",
+  "cuMemAdvise": "CUresult (CUdeviceptr, size_t, CUmem_advise, CUmemLocation)",
+  "cuMemAlloc": "CUresult (CUdeviceptr *, size_t)",
+  "cuMemAllocAsync": "CUresult (CUdeviceptr *, size_t, CUstream)",
+  "cuMemAllocFromPoolAsync": "CUresult (CUdeviceptr *, size_t, CUmemoryPool, CUstream)",
+  "cuMemAllocHost": "CUresult (void **, size_t)",
+  "cuMemAllocManaged": "CUresult (CUdeviceptr *, size_t, unsigned int)",
+  "cuMemAllocPitch": "CUresult (CUdeviceptr *, size_t *, size_t, size_t, unsigned int)",
+  "cuMemBatchDecompressAsync": "CUresult (CUmemDecompressParams *, size_t, unsigned int, size_t *, CUstream)",
+  "cuMemCreate": "CUresult (CUmemGenericAllocationHandle *, size_t, CUmemAllocationProp const *, unsigned PY_LONG_LONG)",
+  "cuMemDiscardAndPrefetchBatchAsync": "CUresult (CUdeviceptr *, size_t *, size_t, CUmemLocation *, size_t *, size_t, unsigned PY_LONG_LONG, CUstream)",
+  "cuMemDiscardBatchAsync": "CUresult (CUdeviceptr *, size_t *, size_t, unsigned PY_LONG_LONG, CUstream)",
+  "cuMemExportToShareableHandle": "CUresult (void *, CUmemGenericAllocationHandle, CUmemAllocationHandleType, unsigned PY_LONG_LONG)",
+  "cuMemFree": "CUresult (CUdeviceptr)",
+  "cuMemFreeAsync": "CUresult (CUdeviceptr, CUstream)",
+  "cuMemFreeHost": "CUresult (void *)",
+  "cuMemGetAccess": "CUresult (unsigned PY_LONG_LONG *, CUmemLocation const *, CUdeviceptr)",
+  "cuMemGetAddressRange": "CUresult (CUdeviceptr *, size_t *, CUdeviceptr)",
+  "cuMemGetAllocationGranularity": "CUresult (size_t *, CUmemAllocationProp const *, CUmemAllocationGranularity_flags)",
+  "cuMemGetAllocationPropertiesFromHandle": "CUresult (CUmemAllocationProp *, CUmemGenericAllocationHandle)",
+  "cuMemGetDefaultMemPool": "CUresult (CUmemoryPool *, CUmemLocation *, CUmemAllocationType)",
+  "cuMemGetHandleForAddressRange": "CUresult (void *, CUdeviceptr, size_t, CUmemRangeHandleType, unsigned PY_LONG_LONG)",
+  "cuMemGetInfo": "CUresult (size_t *, size_t *)",
+  "cuMemGetMemPool": "CUresult (CUmemoryPool *, CUmemLocation *, CUmemAllocationType)",
+  "cuMemHostAlloc": "CUresult (void **, size_t, unsigned int)",
+  "cuMemHostGetDevicePointer": "CUresult (CUdeviceptr *, void *, unsigned int)",
+  "cuMemHostGetFlags": "CUresult (unsigned int *, void *)",
+  "cuMemHostRegister": "CUresult (void *, size_t, unsigned int)",
+  "cuMemHostUnregister": "CUresult (void *)",
+  "cuMemImportFromShareableHandle": "CUresult (CUmemGenericAllocationHandle *, void *, CUmemAllocationHandleType)",
+  "cuMemMap": "CUresult (CUdeviceptr, size_t, size_t, CUmemGenericAllocationHandle, unsigned PY_LONG_LONG)",
+  "cuMemMapArrayAsync": "CUresult (CUarrayMapInfo *, unsigned int, CUstream)",
+  "cuMemPoolCreate": "CUresult (CUmemoryPool *, CUmemPoolProps const *)",
+  "cuMemPoolDestroy": "CUresult (CUmemoryPool)",
+  "cuMemPoolExportPointer": "CUresult (CUmemPoolPtrExportData *, CUdeviceptr)",
+  "cuMemPoolExportToShareableHandle": "CUresult (void *, CUmemoryPool, CUmemAllocationHandleType, unsigned PY_LONG_LONG)",
+  "cuMemPoolGetAccess": "CUresult (CUmemAccess_flags *, CUmemoryPool, CUmemLocation *)",
+  "cuMemPoolGetAttribute": "CUresult (CUmemoryPool, CUmemPool_attribute, void *)",
+  "cuMemPoolImportFromShareableHandle": "CUresult (CUmemoryPool *, void *, CUmemAllocationHandleType, unsigned PY_LONG_LONG)",
+  "cuMemPoolImportPointer": "CUresult (CUdeviceptr *, CUmemoryPool, CUmemPoolPtrExportData *)",
+  "cuMemPoolSetAccess": "CUresult (CUmemoryPool, CUmemAccessDesc const *, size_t)",
+  "cuMemPoolSetAttribute": "CUresult (CUmemoryPool, CUmemPool_attribute, void *)",
+  "cuMemPoolTrimTo": "CUresult (CUmemoryPool, size_t)",
+  "cuMemPrefetchAsync": "CUresult (CUdeviceptr, size_t, CUmemLocation, unsigned int, CUstream)",
+  "cuMemPrefetchBatchAsync": "CUresult (CUdeviceptr *, size_t *, size_t, CUmemLocation *, size_t *, size_t, unsigned PY_LONG_LONG, CUstream)",
+  "cuMemRangeGetAttribute": "CUresult (void *, size_t, CUmem_range_attribute, CUdeviceptr, size_t)",
+  "cuMemRangeGetAttributes": "CUresult (void **, size_t *, CUmem_range_attribute *, size_t, CUdeviceptr, size_t)",
+  "cuMemRelease": "CUresult (CUmemGenericAllocationHandle)",
+  "cuMemRetainAllocationHandle": "CUresult (CUmemGenericAllocationHandle *, void *)",
+  "cuMemSetAccess": "CUresult (CUdeviceptr, size_t, CUmemAccessDesc const *, size_t)",
+  "cuMemSetMemPool": "CUresult (CUmemLocation *, CUmemAllocationType, CUmemoryPool)",
+  "cuMemUnmap": "CUresult (CUdeviceptr, size_t)",
+  "cuMemcpy": "CUresult (CUdeviceptr, CUdeviceptr, size_t)",
+  "cuMemcpy2D": "CUresult (CUDA_MEMCPY2D const *)",
+  "cuMemcpy2DAsync": "CUresult (CUDA_MEMCPY2D const *, CUstream)",
+  "cuMemcpy2DUnaligned": "CUresult (CUDA_MEMCPY2D const *)",
+  "cuMemcpy3D": "CUresult (CUDA_MEMCPY3D const *)",
+  "cuMemcpy3DAsync": "CUresult (CUDA_MEMCPY3D const *, CUstream)",
+  "cuMemcpy3DBatchAsync": "CUresult (size_t, CUDA_MEMCPY3D_BATCH_OP *, unsigned PY_LONG_LONG, CUstream)",
+  "cuMemcpy3DPeer": "CUresult (CUDA_MEMCPY3D_PEER const *)",
+  "cuMemcpy3DPeerAsync": "CUresult (CUDA_MEMCPY3D_PEER const *, CUstream)",
+  "cuMemcpyAsync": "CUresult (CUdeviceptr, CUdeviceptr, size_t, CUstream)",
+  "cuMemcpyAtoA": "CUresult (CUarray, size_t, CUarray, size_t, size_t)",
+  "cuMemcpyAtoD": "CUresult (CUdeviceptr, CUarray, size_t, size_t)",
+  "cuMemcpyAtoH": "CUresult (void *, CUarray, size_t, size_t)",
+  "cuMemcpyAtoHAsync": "CUresult (void *, CUarray, size_t, size_t, CUstream)",
+  "cuMemcpyBatchAsync": "CUresult (CUdeviceptr *, CUdeviceptr *, size_t *, size_t, CUmemcpyAttributes *, size_t *, size_t, CUstream)",
+  "cuMemcpyDtoA": "CUresult (CUarray, size_t, CUdeviceptr, size_t)",
+  "cuMemcpyDtoD": "CUresult (CUdeviceptr, CUdeviceptr, size_t)",
+  "cuMemcpyDtoDAsync": "CUresult (CUdeviceptr, CUdeviceptr, size_t, CUstream)",
+  "cuMemcpyDtoH": "CUresult (void *, CUdeviceptr, size_t)",
+  "cuMemcpyDtoHAsync": "CUresult (void *, CUdeviceptr, size_t, CUstream)",
+  "cuMemcpyHtoA": "CUresult (CUarray, size_t, void const *, size_t)",
+  "cuMemcpyHtoAAsync": "CUresult (CUarray, size_t, void const *, size_t, CUstream)",
+  "cuMemcpyHtoD": "CUresult (CUdeviceptr, void const *, size_t)",
+  "cuMemcpyHtoDAsync": "CUresult (CUdeviceptr, void const *, size_t, CUstream)",
+  "cuMemcpyPeer": "CUresult (CUdeviceptr, CUcontext, CUdeviceptr, CUcontext, size_t)",
+  "cuMemcpyPeerAsync": "CUresult (CUdeviceptr, CUcontext, CUdeviceptr, CUcontext, size_t, CUstream)",
+  "cuMemsetD16": "CUresult (CUdeviceptr, unsigned short, size_t)",
+  "cuMemsetD16Async": "CUresult (CUdeviceptr, unsigned short, size_t, CUstream)",
+  "cuMemsetD2D16": "CUresult (CUdeviceptr, size_t, unsigned short, size_t, size_t)",
+  "cuMemsetD2D16Async": "CUresult (CUdeviceptr, size_t, unsigned short, size_t, size_t, CUstream)",
+  "cuMemsetD2D32": "CUresult (CUdeviceptr, size_t, unsigned int, size_t, size_t)",
+  "cuMemsetD2D32Async": "CUresult (CUdeviceptr, size_t, unsigned int, size_t, size_t, CUstream)",
+  "cuMemsetD2D8": "CUresult (CUdeviceptr, size_t, unsigned char, size_t, size_t)",
+  "cuMemsetD2D8Async": "CUresult (CUdeviceptr, size_t, unsigned char, size_t, size_t, CUstream)",
+  "cuMemsetD32": "CUresult (CUdeviceptr, unsigned int, size_t)",
+  "cuMemsetD32Async": "CUresult (CUdeviceptr, unsigned int, size_t, CUstream)",
+  "cuMemsetD8": "CUresult (CUdeviceptr, unsigned char, size_t)",
+  "cuMemsetD8Async": "CUresult (CUdeviceptr, unsigned char, size_t, CUstream)",
+  "cuMipmappedArrayCreate": "CUresult (CUmipmappedArray *, CUDA_ARRAY3D_DESCRIPTOR const *, unsigned int)",
+  "cuMipmappedArrayDestroy": "CUresult (CUmipmappedArray)",
+  "cuMipmappedArrayGetLevel": "CUresult (CUarray *, CUmipmappedArray, unsigned int)",
+  "cuMipmappedArrayGetMemoryRequirements": "CUresult (CUDA_ARRAY_MEMORY_REQUIREMENTS *, CUmipmappedArray, CUdevice)",
+  "cuMipmappedArrayGetSparseProperties": "CUresult (CUDA_ARRAY_SPARSE_PROPERTIES *, CUmipmappedArray)",
+  "cuModuleEnumerateFunctions": "CUresult (CUfunction *, unsigned int, CUmodule)",
+  "cuModuleGetFunction": "CUresult (CUfunction *, CUmodule, char const *)",
+  "cuModuleGetFunctionCount": "CUresult (unsigned int *, CUmodule)",
+  "cuModuleGetGlobal": "CUresult (CUdeviceptr *, size_t *, CUmodule, char const *)",
+  "cuModuleGetLoadingMode": "CUresult (CUmoduleLoadingMode *)",
+  "cuModuleGetSurfRef": "CUresult (CUsurfref *, CUmodule, char const *)",
+  "cuModuleGetTexRef": "CUresult (CUtexref *, CUmodule, char const *)",
+  "cuModuleLoad": "CUresult (CUmodule *, char const *)",
+  "cuModuleLoadData": "CUresult (CUmodule *, void const *)",
+  "cuModuleLoadDataEx": "CUresult (CUmodule *, void const *, unsigned int, CUjit_option *, void **)",
+  "cuModuleLoadFatBinary": "CUresult (CUmodule *, void const *)",
+  "cuModuleUnload": "CUresult (CUmodule)",
+  "cuMulticastAddDevice": "CUresult (CUmemGenericAllocationHandle, CUdevice)",
+  "cuMulticastBindAddr": "CUresult (CUmemGenericAllocationHandle, size_t, CUdeviceptr, size_t, unsigned PY_LONG_LONG)",
+  "cuMulticastBindMem": "CUresult (CUmemGenericAllocationHandle, size_t, CUmemGenericAllocationHandle, size_t, size_t, unsigned PY_LONG_LONG)",
+  "cuMulticastCreate": "CUresult (CUmemGenericAllocationHandle *, CUmulticastObjectProp const *)",
+  "cuMulticastGetGranularity": "CUresult (size_t *, CUmulticastObjectProp const *, CUmulticastGranularity_flags)",
+  "cuMulticastUnbind": "CUresult (CUmemGenericAllocationHandle, CUdevice, size_t, size_t)",
+  "cuOccupancyAvailableDynamicSMemPerBlock": "CUresult (size_t *, CUfunction, int, int)",
+  "cuOccupancyMaxActiveBlocksPerMultiprocessor": "CUresult (int *, CUfunction, int, size_t)",
+  "cuOccupancyMaxActiveBlocksPerMultiprocessorWithFlags": "CUresult (int *, CUfunction, int, size_t, unsigned int)",
+  "cuOccupancyMaxActiveClusters": "CUresult (int *, CUfunction, CUlaunchConfig const *)",
+  "cuOccupancyMaxPotentialBlockSize": "CUresult (int *, int *, CUfunction, CUoccupancyB2DSize, size_t, int)",
+  "cuOccupancyMaxPotentialBlockSizeWithFlags": "CUresult (int *, int *, CUfunction, CUoccupancyB2DSize, size_t, int, unsigned int)",
+  "cuOccupancyMaxPotentialClusterSize": "CUresult (int *, CUfunction, CUlaunchConfig const *)",
+  "cuParamSetSize": "CUresult (CUfunction, unsigned int)",
+  "cuParamSetTexRef": "CUresult (CUfunction, int, CUtexref)",
+  "cuParamSetf": "CUresult (CUfunction, int, float)",
+  "cuParamSeti": "CUresult (CUfunction, int, unsigned int)",
+  "cuParamSetv": "CUresult (CUfunction, int, void *, unsigned int)",
+  "cuPointerGetAttribute": "CUresult (void *, CUpointer_attribute, CUdeviceptr)",
+  "cuPointerGetAttributes": "CUresult (unsigned int, CUpointer_attribute *, void **, CUdeviceptr)",
+  "cuPointerSetAttribute": "CUresult (void const *, CUpointer_attribute, CUdeviceptr)",
+  "cuProfilerStart": "CUresult (void)",
+  "cuProfilerStop": "CUresult (void)",
+  "cuSignalExternalSemaphoresAsync": "CUresult (CUexternalSemaphore const *, CUDA_EXTERNAL_SEMAPHORE_SIGNAL_PARAMS const *, unsigned int, CUstream)",
+  "cuStreamAddCallback": "CUresult (CUstream, CUstreamCallback, void *, unsigned int)",
+  "cuStreamAttachMemAsync": "CUresult (CUstream, CUdeviceptr, size_t, unsigned int)",
+  "cuStreamBatchMemOp": "CUresult (CUstream, unsigned int, CUstreamBatchMemOpParams *, unsigned int)",
+  "cuStreamBeginCapture": "CUresult (CUstream, CUstreamCaptureMode)",
+  "cuStreamBeginCaptureToGraph": "CUresult (CUstream, CUgraph, CUgraphNode const *, CUgraphEdgeData const *, size_t, CUstreamCaptureMode)",
+  "cuStreamCopyAttributes": "CUresult (CUstream, CUstream)",
+  "cuStreamCreate": "CUresult (CUstream *, unsigned int)",
+  "cuStreamCreateWithPriority": "CUresult (CUstream *, unsigned int, int)",
+  "cuStreamDestroy": "CUresult (CUstream)",
+  "cuStreamEndCapture": "CUresult (CUstream, CUgraph *)",
+  "cuStreamGetAttribute": "CUresult (CUstream, CUstreamAttrID, CUstreamAttrValue *)",
+  "cuStreamGetCaptureInfo": "CUresult (CUstream, CUstreamCaptureStatus *, cuuint64_t *, CUgraph *, CUgraphNode const **, CUgraphEdgeData const **, size_t *)",
+  "cuStreamGetCtx": "CUresult (CUstream, CUcontext *)",
+  "cuStreamGetCtx_v2": "CUresult (CUstream, CUcontext *, CUgreenCtx *)",
+  "cuStreamGetDevice": "CUresult (CUstream, CUdevice *)",
+  "cuStreamGetFlags": "CUresult (CUstream, unsigned int *)",
+  "cuStreamGetGreenCtx": "CUresult (CUstream, CUgreenCtx *)",
+  "cuStreamGetId": "CUresult (CUstream, unsigned PY_LONG_LONG *)",
+  "cuStreamGetPriority": "CUresult (CUstream, int *)",
+  "cuStreamIsCapturing": "CUresult (CUstream, CUstreamCaptureStatus *)",
+  "cuStreamQuery": "CUresult (CUstream)",
+  "cuStreamSetAttribute": "CUresult (CUstream, CUstreamAttrID, CUstreamAttrValue const *)",
+  "cuStreamSynchronize": "CUresult (CUstream)",
+  "cuStreamUpdateCaptureDependencies": "CUresult (CUstream, CUgraphNode *, CUgraphEdgeData const *, size_t, unsigned int)",
+  "cuStreamWaitEvent": "CUresult (CUstream, CUevent, unsigned int)",
+  "cuStreamWaitValue32": "CUresult (CUstream, CUdeviceptr, cuuint32_t, unsigned int)",
+  "cuStreamWaitValue64": "CUresult (CUstream, CUdeviceptr, cuuint64_t, unsigned int)",
+  "cuStreamWriteValue32": "CUresult (CUstream, CUdeviceptr, cuuint32_t, unsigned int)",
+  "cuStreamWriteValue64": "CUresult (CUstream, CUdeviceptr, cuuint64_t, unsigned int)",
+  "cuSurfObjectCreate": "CUresult (CUsurfObject *, CUDA_RESOURCE_DESC const *)",
+  "cuSurfObjectDestroy": "CUresult (CUsurfObject)",
+  "cuSurfObjectGetResourceDesc": "CUresult (CUDA_RESOURCE_DESC *, CUsurfObject)",
+  "cuSurfRefGetArray": "CUresult (CUarray *, CUsurfref)",
+  "cuSurfRefSetArray": "CUresult (CUsurfref, CUarray, unsigned int)",
+  "cuTensorMapEncodeIm2col": "CUresult (CUtensorMap *, CUtensorMapDataType, cuuint32_t, void *, cuuint64_t const *, cuuint64_t const *, int const *, int const *, cuuint32_t, cuuint32_t, cuuint32_t const *, CUtensorMapInterleave, CUtensorMapSwizzle, CUtensorMapL2promotion, CUtensorMapFloatOOBfill)",
+  "cuTensorMapEncodeIm2colWide": "CUresult (CUtensorMap *, CUtensorMapDataType, cuuint32_t, void *, cuuint64_t const *, cuuint64_t const *, int, int, cuuint32_t, cuuint32_t, cuuint32_t const *, CUtensorMapInterleave, CUtensorMapIm2ColWideMode, CUtensorMapSwizzle, CUtensorMapL2promotion, CUtensorMapFloatOOBfill)",
+  "cuTensorMapEncodeTiled": "CUresult (CUtensorMap *, CUtensorMapDataType, cuuint32_t, void *, cuuint64_t const *, cuuint64_t const *, cuuint32_t const *, cuuint32_t const *, CUtensorMapInterleave, CUtensorMapSwizzle, CUtensorMapL2promotion, CUtensorMapFloatOOBfill)",
+  "cuTensorMapReplaceAddress": "CUresult (CUtensorMap *, void *)",
+  "cuTexObjectCreate": "CUresult (CUtexObject *, CUDA_RESOURCE_DESC const *, CUDA_TEXTURE_DESC const *, CUDA_RESOURCE_VIEW_DESC const *)",
+  "cuTexObjectDestroy": "CUresult (CUtexObject)",
+  "cuTexObjectGetResourceDesc": "CUresult (CUDA_RESOURCE_DESC *, CUtexObject)",
+  "cuTexObjectGetResourceViewDesc": "CUresult (CUDA_RESOURCE_VIEW_DESC *, CUtexObject)",
+  "cuTexObjectGetTextureDesc": "CUresult (CUDA_TEXTURE_DESC *, CUtexObject)",
+  "cuTexRefCreate": "CUresult (CUtexref *)",
+  "cuTexRefDestroy": "CUresult (CUtexref)",
+  "cuTexRefGetAddress": "CUresult (CUdeviceptr *, CUtexref)",
+  "cuTexRefGetAddressMode": "CUresult (CUaddress_mode *, CUtexref, int)",
+  "cuTexRefGetArray": "CUresult (CUarray *, CUtexref)",
+  "cuTexRefGetBorderColor": "CUresult (float *, CUtexref)",
+  "cuTexRefGetFilterMode": "CUresult (CUfilter_mode *, CUtexref)",
+  "cuTexRefGetFlags": "CUresult (unsigned int *, CUtexref)",
+  "cuTexRefGetFormat": "CUresult (CUarray_format *, int *, CUtexref)",
+  "cuTexRefGetMaxAnisotropy": "CUresult (int *, CUtexref)",
+  "cuTexRefGetMipmapFilterMode": "CUresult (CUfilter_mode *, CUtexref)",
+  "cuTexRefGetMipmapLevelBias": "CUresult (float *, CUtexref)",
+  "cuTexRefGetMipmapLevelClamp": "CUresult (float *, float *, CUtexref)",
+  "cuTexRefGetMipmappedArray": "CUresult (CUmipmappedArray *, CUtexref)",
+  "cuTexRefSetAddress": "CUresult (size_t *, CUtexref, CUdeviceptr, size_t)",
+  "cuTexRefSetAddress2D": "CUresult (CUtexref, CUDA_ARRAY_DESCRIPTOR const *, CUdeviceptr, size_t)",
+  "cuTexRefSetAddressMode": "CUresult (CUtexref, int, CUaddress_mode)",
+  "cuTexRefSetArray": "CUresult (CUtexref, CUarray, unsigned int)",
+  "cuTexRefSetBorderColor": "CUresult (CUtexref, float *)",
+  "cuTexRefSetFilterMode": "CUresult (CUtexref, CUfilter_mode)",
+  "cuTexRefSetFlags": "CUresult (CUtexref, unsigned int)",
+  "cuTexRefSetFormat": "CUresult (CUtexref, CUarray_format, int)",
+  "cuTexRefSetMaxAnisotropy": "CUresult (CUtexref, unsigned int)",
+  "cuTexRefSetMipmapFilterMode": "CUresult (CUtexref, CUfilter_mode)",
+  "cuTexRefSetMipmapLevelBias": "CUresult (CUtexref, float)",
+  "cuTexRefSetMipmapLevelClamp": "CUresult (CUtexref, float, float)",
+  "cuTexRefSetMipmappedArray": "CUresult (CUtexref, CUmipmappedArray, unsigned int)",
+  "cuThreadExchangeStreamCaptureMode": "CUresult (CUstreamCaptureMode *)",
+  "cuUserObjectCreate": "CUresult (CUuserObject *, void *, CUhostFn, unsigned int, unsigned int)",
+  "cuUserObjectRelease": "CUresult (CUuserObject, unsigned int)",
+  "cuUserObjectRetain": "CUresult (CUuserObject, unsigned int)",
+  "cuVDPAUCtxCreate": "CUresult (CUcontext *, unsigned int, CUdevice, __pyx_t_4cuda_8bindings_8cydriver_VdpDevice, __pyx_t_4cuda_8bindings_8cydriver_VdpGetProcAddress *)",
+  "cuVDPAUGetDevice": "CUresult (CUdevice *, __pyx_t_4cuda_8bindings_8cydriver_VdpDevice, __pyx_t_4cuda_8bindings_8cydriver_VdpGetProcAddress *)",
+  "cuWaitExternalSemaphoresAsync": "CUresult (CUexternalSemaphore const *, CUDA_EXTERNAL_SEMAPHORE_WAIT_PARAMS const *, unsigned int, CUstream)"
+}
diff --git a/cuda_bindings/tests/abi/cuda/bindings/cynvjitlink.abi.json b/cuda_bindings/tests/abi/cuda/bindings/cynvjitlink.abi.json
new file mode 100644
index 000000000..6add6d965
--- /dev/null
+++ b/cuda_bindings/tests/abi/cuda/bindings/cynvjitlink.abi.json
@@ -0,0 +1,16 @@
+{
+  "nvJitLinkAddData": "nvJitLinkResult (nvJitLinkHandle, nvJitLinkInputType, void const *, size_t, char const *)",
+  "nvJitLinkAddFile": "nvJitLinkResult (nvJitLinkHandle, nvJitLinkInputType, char const *)",
+  "nvJitLinkComplete": "nvJitLinkResult (nvJitLinkHandle)",
+  "nvJitLinkCreate": "nvJitLinkResult (nvJitLinkHandle *, uint32_t, char const **)",
+  "nvJitLinkDestroy": "nvJitLinkResult (nvJitLinkHandle *)",
+  "nvJitLinkGetErrorLog": "nvJitLinkResult (nvJitLinkHandle, char *)",
+  "nvJitLinkGetErrorLogSize": "nvJitLinkResult (nvJitLinkHandle, size_t *)",
+  "nvJitLinkGetInfoLog": "nvJitLinkResult (nvJitLinkHandle, char *)",
+  "nvJitLinkGetInfoLogSize": "nvJitLinkResult (nvJitLinkHandle, size_t *)",
+  "nvJitLinkGetLinkedCubin": "nvJitLinkResult (nvJitLinkHandle, void *)",
+  "nvJitLinkGetLinkedCubinSize": "nvJitLinkResult (nvJitLinkHandle, size_t *)",
+  "nvJitLinkGetLinkedPtx": "nvJitLinkResult (nvJitLinkHandle, char *)",
+  "nvJitLinkGetLinkedPtxSize": "nvJitLinkResult (nvJitLinkHandle, size_t *)",
+  "nvJitLinkVersion": "nvJitLinkResult (unsigned int *, unsigned int *)"
+}
diff --git a/cuda_bindings/tests/abi/cuda/bindings/cynvrtc.abi.json b/cuda_bindings/tests/abi/cuda/bindings/cynvrtc.abi.json
new file mode 100644
index 000000000..73448d03a
--- /dev/null
+++ b/cuda_bindings/tests/abi/cuda/bindings/cynvrtc.abi.json
@@ -0,0 +1,26 @@
+{
+  "nvrtcAddNameExpression": "nvrtcResult (nvrtcProgram, char const *)",
+  "nvrtcCompileProgram": "nvrtcResult (nvrtcProgram, int, char const **)",
+  "nvrtcCreateProgram": "nvrtcResult (nvrtcProgram *, char const *, char const *, int, char const **, char const **)",
+  "nvrtcDestroyProgram": "nvrtcResult (nvrtcProgram *)",
+  "nvrtcGetCUBIN": "nvrtcResult (nvrtcProgram, char *)",
+  "nvrtcGetCUBINSize": "nvrtcResult (nvrtcProgram, size_t *)",
+  "nvrtcGetErrorString": "char const *(nvrtcResult)",
+  "nvrtcGetLTOIR": "nvrtcResult (nvrtcProgram, char *)",
+  "nvrtcGetLTOIRSize": "nvrtcResult (nvrtcProgram, size_t *)",
+  "nvrtcGetLoweredName": "nvrtcResult (nvrtcProgram, char const *, char const **)",
+  "nvrtcGetNumSupportedArchs": "nvrtcResult (int *)",
+  "nvrtcGetOptiXIR": "nvrtcResult (nvrtcProgram, char *)",
+  "nvrtcGetOptiXIRSize": "nvrtcResult (nvrtcProgram, size_t *)",
+  "nvrtcGetPCHCreateStatus": "nvrtcResult (nvrtcProgram)",
+  "nvrtcGetPCHHeapSize": "nvrtcResult (size_t *)",
+  "nvrtcGetPCHHeapSizeRequired": "nvrtcResult (nvrtcProgram, size_t *)",
+  "nvrtcGetPTX": "nvrtcResult (nvrtcProgram, char *)",
+  "nvrtcGetPTXSize": "nvrtcResult (nvrtcProgram, size_t *)",
+  "nvrtcGetProgramLog": "nvrtcResult (nvrtcProgram, char *)",
+  "nvrtcGetProgramLogSize": "nvrtcResult (nvrtcProgram, size_t *)",
+  "nvrtcGetSupportedArchs": "nvrtcResult (int *)",
+  "nvrtcSetFlowCallback": "nvrtcResult (nvrtcProgram, void *, void *)",
+  "nvrtcSetPCHHeapSize": "nvrtcResult (size_t)",
+  "nvrtcVersion": "nvrtcResult (int *, int *)"
+}
diff --git a/cuda_bindings/tests/abi/cuda/bindings/cynvvm.abi.json b/cuda_bindings/tests/abi/cuda/bindings/cynvvm.abi.json
new file mode 100644
index 000000000..ed6c3e61c
--- /dev/null
+++ b/cuda_bindings/tests/abi/cuda/bindings/cynvvm.abi.json
@@ -0,0 +1,15 @@
+{
+  "nvvmAddModuleToProgram": "nvvmResult (nvvmProgram, char const *, size_t, char const *)",
+  "nvvmCompileProgram": "nvvmResult (nvvmProgram, int, char const **)",
+  "nvvmCreateProgram": "nvvmResult (nvvmProgram *)",
+  "nvvmDestroyProgram": "nvvmResult (nvvmProgram *)",
+  "nvvmGetCompiledResult": "nvvmResult (nvvmProgram, char *)",
+  "nvvmGetCompiledResultSize": "nvvmResult (nvvmProgram, size_t *)",
+  "nvvmGetErrorString": "char const *(nvvmResult)",
+  "nvvmGetProgramLog": "nvvmResult (nvvmProgram, char *)",
+  "nvvmGetProgramLogSize": "nvvmResult (nvvmProgram, size_t *)",
+  "nvvmIRVersion": "nvvmResult (int *, int *, int *, int *)",
+  "nvvmLazyAddModuleToProgram": "nvvmResult (nvvmProgram, char const *, size_t, char const *)",
+  "nvvmVerifyProgram": "nvvmResult (nvvmProgram, int, char const **)",
+  "nvvmVersion": "nvvmResult (int *, int *)"
+}
diff --git a/cuda_bindings/tests/abi/cuda/bindings/cyruntime.abi.json b/cuda_bindings/tests/abi/cuda/bindings/cyruntime.abi.json
new file mode 100644
index 000000000..93bce2e45
--- /dev/null
+++ b/cuda_bindings/tests/abi/cuda/bindings/cyruntime.abi.json
@@ -0,0 +1,316 @@
+{
+  "cudaArrayGetInfo": "cudaError_t (struct cudaChannelFormatDesc *, struct cudaExtent *, unsigned int *, cudaArray_t)",
+  "cudaArrayGetMemoryRequirements": "cudaError_t (struct cudaArrayMemoryRequirements *, cudaArray_t, int)",
+  "cudaArrayGetPlane": "cudaError_t (cudaArray_t *, cudaArray_t, unsigned int)",
+  "cudaArrayGetSparseProperties": "cudaError_t (struct cudaArraySparseProperties *, cudaArray_t)",
+  "cudaChooseDevice": "cudaError_t (int *, struct cudaDeviceProp const *)",
+  "cudaCreateChannelDesc": "struct cudaChannelFormatDesc (int, int, int, int, enum cudaChannelFormatKind)",
+  "cudaCreateSurfaceObject": "cudaError_t (cudaSurfaceObject_t *, struct cudaResourceDesc const *)",
+  "cudaCreateTextureObject": "cudaError_t (cudaTextureObject_t *, struct cudaResourceDesc const *, struct cudaTextureDesc const *, struct cudaResourceViewDesc const *)",
+  "cudaCtxResetPersistingL2Cache": "cudaError_t (void)",
+  "cudaDestroyExternalMemory": "cudaError_t (cudaExternalMemory_t)",
+  "cudaDestroyExternalSemaphore": "cudaError_t (cudaExternalSemaphore_t)",
+  "cudaDestroySurfaceObject": "cudaError_t (cudaSurfaceObject_t)",
+  "cudaDestroyTextureObject": "cudaError_t (cudaTextureObject_t)",
+  "cudaDeviceCanAccessPeer": "cudaError_t (int *, int, int)",
+  "cudaDeviceDisablePeerAccess": "cudaError_t (int)",
+  "cudaDeviceEnablePeerAccess": "cudaError_t (int, unsigned int)",
+  "cudaDeviceFlushGPUDirectRDMAWrites": "cudaError_t (enum cudaFlushGPUDirectRDMAWritesTarget, enum cudaFlushGPUDirectRDMAWritesScope)",
+  "cudaDeviceGetAttribute": "cudaError_t (int *, enum cudaDeviceAttr, int)",
+  "cudaDeviceGetByPCIBusId": "cudaError_t (int *, char const *)",
+  "cudaDeviceGetCacheConfig": "cudaError_t (enum cudaFuncCache *)",
+  "cudaDeviceGetDefaultMemPool": "cudaError_t (cudaMemPool_t *, int)",
+  "cudaDeviceGetGraphMemAttribute": "cudaError_t (int, enum cudaGraphMemAttributeType, void *)",
+  "cudaDeviceGetHostAtomicCapabilities": "cudaError_t (unsigned int *, enum cudaAtomicOperation const *, unsigned int, int)",
+  "cudaDeviceGetLimit": "cudaError_t (size_t *, enum cudaLimit)",
+  "cudaDeviceGetMemPool": "cudaError_t (cudaMemPool_t *, int)",
+  "cudaDeviceGetNvSciSyncAttributes": "cudaError_t (void *, int, int)",
+  "cudaDeviceGetP2PAtomicCapabilities": "cudaError_t (unsigned int *, enum cudaAtomicOperation const *, unsigned int, int, int)",
+  "cudaDeviceGetP2PAttribute": "cudaError_t (int *, enum cudaDeviceP2PAttr, int, int)",
+  "cudaDeviceGetPCIBusId": "cudaError_t (char *, int, int)",
+  "cudaDeviceGetSharedMemConfig": "cudaError_t (enum cudaSharedMemConfig *)",
+  "cudaDeviceGetStreamPriorityRange": "cudaError_t (int *, int *)",
+  "cudaDeviceGetTexture1DLinearMaxWidth": "cudaError_t (size_t *, struct cudaChannelFormatDesc const *, int)",
+  "cudaDeviceGraphMemTrim": "cudaError_t (int)",
+  "cudaDeviceRegisterAsyncNotification": "cudaError_t (int, cudaAsyncCallback, void *, cudaAsyncCallbackHandle_t *)",
+  "cudaDeviceReset": "cudaError_t (void)",
+  "cudaDeviceSetCacheConfig": "cudaError_t (enum cudaFuncCache)",
+  "cudaDeviceSetGraphMemAttribute": "cudaError_t (int, enum cudaGraphMemAttributeType, void *)",
+  "cudaDeviceSetLimit": "cudaError_t (enum cudaLimit, size_t)",
+  "cudaDeviceSetMemPool": "cudaError_t (int, cudaMemPool_t)",
+  "cudaDeviceSetSharedMemConfig": "cudaError_t (enum cudaSharedMemConfig)",
+  "cudaDeviceSynchronize": "cudaError_t (void)",
+  "cudaDeviceUnregisterAsyncNotification": "cudaError_t (int, cudaAsyncCallbackHandle_t)",
+  "cudaDriverGetVersion": "cudaError_t (int *)",
+  "cudaEGLStreamConsumerAcquireFrame": "cudaError_t (__pyx_t_4cuda_8bindings_9cyruntime_cudaEglStreamConnection *, cudaGraphicsResource_t *, cudaStream_t *, unsigned int)",
+  "cudaEGLStreamConsumerConnect": "cudaError_t (__pyx_t_4cuda_8bindings_9cyruntime_cudaEglStreamConnection *, __pyx_t_4cuda_8bindings_9cyruntime_EGLStreamKHR)",
+  "cudaEGLStreamConsumerConnectWithFlags": "cudaError_t (__pyx_t_4cuda_8bindings_9cyruntime_cudaEglStreamConnection *, __pyx_t_4cuda_8bindings_9cyruntime_EGLStreamKHR, unsigned int)",
+  "cudaEGLStreamConsumerDisconnect": "cudaError_t (__pyx_t_4cuda_8bindings_9cyruntime_cudaEglStreamConnection *)",
+  "cudaEGLStreamConsumerReleaseFrame": "cudaError_t (__pyx_t_4cuda_8bindings_9cyruntime_cudaEglStreamConnection *, cudaGraphicsResource_t, cudaStream_t *)",
+  "cudaEGLStreamProducerConnect": "cudaError_t (__pyx_t_4cuda_8bindings_9cyruntime_cudaEglStreamConnection *, __pyx_t_4cuda_8bindings_9cyruntime_EGLStreamKHR, __pyx_t_4cuda_8bindings_9cyruntime_EGLint, __pyx_t_4cuda_8bindings_9cyruntime_EGLint)",
+  "cudaEGLStreamProducerDisconnect": "cudaError_t (__pyx_t_4cuda_8bindings_9cyruntime_cudaEglStreamConnection *)",
+  "cudaEGLStreamProducerPresentFrame": "cudaError_t (__pyx_t_4cuda_8bindings_9cyruntime_cudaEglStreamConnection *, __pyx_t_4cuda_8bindings_9cyruntime_cudaEglFrame, cudaStream_t *)",
+  "cudaEGLStreamProducerReturnFrame": "cudaError_t (__pyx_t_4cuda_8bindings_9cyruntime_cudaEglStreamConnection *, __pyx_t_4cuda_8bindings_9cyruntime_cudaEglFrame *, cudaStream_t *)",
+  "cudaEventCreate": "cudaError_t (cudaEvent_t *)",
+  "cudaEventCreateFromEGLSync": "cudaError_t (cudaEvent_t *, __pyx_t_4cuda_8bindings_9cyruntime_EGLSyncKHR, unsigned int)",
+  "cudaEventCreateWithFlags": "cudaError_t (cudaEvent_t *, unsigned int)",
+  "cudaEventDestroy": "cudaError_t (cudaEvent_t)",
+  "cudaEventElapsedTime": "cudaError_t (float *, cudaEvent_t, cudaEvent_t)",
+  "cudaEventQuery": "cudaError_t (cudaEvent_t)",
+  "cudaEventRecord": "cudaError_t (cudaEvent_t, cudaStream_t)",
+  "cudaEventRecordWithFlags": "cudaError_t (cudaEvent_t, cudaStream_t, unsigned int)",
+  "cudaEventSynchronize": "cudaError_t (cudaEvent_t)",
+  "cudaExternalMemoryGetMappedBuffer": "cudaError_t (void **, cudaExternalMemory_t, struct cudaExternalMemoryBufferDesc const *)",
+  "cudaExternalMemoryGetMappedMipmappedArray": "cudaError_t (cudaMipmappedArray_t *, cudaExternalMemory_t, struct cudaExternalMemoryMipmappedArrayDesc const *)",
+  "cudaFree": "cudaError_t (void *)",
+  "cudaFreeArray": "cudaError_t (cudaArray_t)",
+  "cudaFreeAsync": "cudaError_t (void *, cudaStream_t)",
+  "cudaFreeHost": "cudaError_t (void *)",
+  "cudaFreeMipmappedArray": "cudaError_t (cudaMipmappedArray_t)",
+  "cudaFuncGetAttributes": "cudaError_t (struct cudaFuncAttributes *, void const *)",
+  "cudaFuncSetAttribute": "cudaError_t (void const *, enum cudaFuncAttribute, int)",
+  "cudaFuncSetCacheConfig": "cudaError_t (void const *, enum cudaFuncCache)",
+  "cudaFuncSetSharedMemConfig": "cudaError_t (void const *, enum cudaSharedMemConfig)",
+  "cudaGLGetDevices": "cudaError_t (unsigned int *, int *, unsigned int, enum __pyx_t_4cuda_8bindings_9cyruntime_cudaGLDeviceList)",
+  "cudaGetChannelDesc": "cudaError_t (struct cudaChannelFormatDesc *, cudaArray_const_t)",
+  "cudaGetDevice": "cudaError_t (int *)",
+  "cudaGetDeviceCount": "cudaError_t (int *)",
+  "cudaGetDeviceFlags": "cudaError_t (unsigned int *)",
+  "cudaGetDeviceProperties": "cudaError_t (struct cudaDeviceProp *, int)",
+  "cudaGetDriverEntryPoint": "cudaError_t (char const *, void **, unsigned PY_LONG_LONG, enum cudaDriverEntryPointQueryResult *)",
+  "cudaGetDriverEntryPointByVersion": "cudaError_t (char const *, void **, unsigned int, unsigned PY_LONG_LONG, enum cudaDriverEntryPointQueryResult *)",
+  "cudaGetErrorName": "char const *(cudaError_t)",
+  "cudaGetErrorString": "char const *(cudaError_t)",
+  "cudaGetExportTable": "cudaError_t (void const **, cudaUUID_t const *)",
+  "cudaGetKernel": "cudaError_t (cudaKernel_t *, void const *)",
+  "cudaGetLastError": "cudaError_t (void)",
+  "cudaGetMipmappedArrayLevel": "cudaError_t (cudaArray_t *, cudaMipmappedArray_const_t, unsigned int)",
+  "cudaGetSurfaceObjectResourceDesc": "cudaError_t (struct cudaResourceDesc *, cudaSurfaceObject_t)",
+  "cudaGetTextureObjectResourceDesc": "cudaError_t (struct cudaResourceDesc *, cudaTextureObject_t)",
+  "cudaGetTextureObjectResourceViewDesc": "cudaError_t (struct cudaResourceViewDesc *, cudaTextureObject_t)",
+  "cudaGetTextureObjectTextureDesc": "cudaError_t (struct cudaTextureDesc *, cudaTextureObject_t)",
+  "cudaGraphAddChildGraphNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, cudaGraph_t)",
+  "cudaGraphAddDependencies": "cudaError_t (cudaGraph_t, cudaGraphNode_t const *, cudaGraphNode_t const *, cudaGraphEdgeData const *, size_t)",
+  "cudaGraphAddEmptyNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t)",
+  "cudaGraphAddEventRecordNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, cudaEvent_t)",
+  "cudaGraphAddEventWaitNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, cudaEvent_t)",
+  "cudaGraphAddExternalSemaphoresSignalNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, struct cudaExternalSemaphoreSignalNodeParams const *)",
+  "cudaGraphAddExternalSemaphoresWaitNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, struct cudaExternalSemaphoreWaitNodeParams const *)",
+  "cudaGraphAddHostNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, struct cudaHostNodeParams const *)",
+  "cudaGraphAddKernelNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, struct cudaKernelNodeParams const *)",
+  "cudaGraphAddMemAllocNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, struct cudaMemAllocNodeParams *)",
+  "cudaGraphAddMemFreeNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, void *)",
+  "cudaGraphAddMemcpyNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, struct cudaMemcpy3DParms const *)",
+  "cudaGraphAddMemcpyNode1D": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, void *, void const *, size_t, enum cudaMemcpyKind)",
+  "cudaGraphAddMemsetNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, size_t, struct cudaMemsetParams const *)",
+  "cudaGraphAddNode": "cudaError_t (cudaGraphNode_t *, cudaGraph_t, cudaGraphNode_t const *, cudaGraphEdgeData const *, size_t, struct cudaGraphNodeParams *)",
+  "cudaGraphChildGraphNodeGetGraph": "cudaError_t (cudaGraphNode_t, cudaGraph_t *)",
+  "cudaGraphClone": "cudaError_t (cudaGraph_t *, cudaGraph_t)",
+  "cudaGraphConditionalHandleCreate": "cudaError_t (cudaGraphConditionalHandle *, cudaGraph_t, unsigned int, unsigned int)",
+  "cudaGraphCreate": "cudaError_t (cudaGraph_t *, unsigned int)",
+  "cudaGraphDebugDotPrint": "cudaError_t (cudaGraph_t, char const *, unsigned int)",
+  "cudaGraphDestroy": "cudaError_t (cudaGraph_t)",
+  "cudaGraphDestroyNode": "cudaError_t (cudaGraphNode_t)",
+  "cudaGraphEventRecordNodeGetEvent": "cudaError_t (cudaGraphNode_t, cudaEvent_t *)",
+  "cudaGraphEventRecordNodeSetEvent": "cudaError_t (cudaGraphNode_t, cudaEvent_t)",
+  "cudaGraphEventWaitNodeGetEvent": "cudaError_t (cudaGraphNode_t, cudaEvent_t *)",
+  "cudaGraphEventWaitNodeSetEvent": "cudaError_t (cudaGraphNode_t, cudaEvent_t)",
+  "cudaGraphExecChildGraphNodeSetParams": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, cudaGraph_t)",
+  "cudaGraphExecDestroy": "cudaError_t (cudaGraphExec_t)",
+  "cudaGraphExecEventRecordNodeSetEvent": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, cudaEvent_t)",
+  "cudaGraphExecEventWaitNodeSetEvent": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, cudaEvent_t)",
+  "cudaGraphExecExternalSemaphoresSignalNodeSetParams": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, struct cudaExternalSemaphoreSignalNodeParams const *)",
+  "cudaGraphExecExternalSemaphoresWaitNodeSetParams": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, struct cudaExternalSemaphoreWaitNodeParams const *)",
+  "cudaGraphExecGetFlags": "cudaError_t (cudaGraphExec_t, unsigned PY_LONG_LONG *)",
+  "cudaGraphExecHostNodeSetParams": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, struct cudaHostNodeParams const *)",
+  "cudaGraphExecKernelNodeSetParams": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, struct cudaKernelNodeParams const *)",
+  "cudaGraphExecMemcpyNodeSetParams": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, struct cudaMemcpy3DParms const *)",
+  "cudaGraphExecMemcpyNodeSetParams1D": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, void *, void const *, size_t, enum cudaMemcpyKind)",
+  "cudaGraphExecMemsetNodeSetParams": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, struct cudaMemsetParams const *)",
+  "cudaGraphExecNodeSetParams": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, struct cudaGraphNodeParams *)",
+  "cudaGraphExecUpdate": "cudaError_t (cudaGraphExec_t, cudaGraph_t, cudaGraphExecUpdateResultInfo *)",
+  "cudaGraphExternalSemaphoresSignalNodeGetParams": "cudaError_t (cudaGraphNode_t, struct cudaExternalSemaphoreSignalNodeParams *)",
+  "cudaGraphExternalSemaphoresSignalNodeSetParams": "cudaError_t (cudaGraphNode_t, struct cudaExternalSemaphoreSignalNodeParams const *)",
+  "cudaGraphExternalSemaphoresWaitNodeGetParams": "cudaError_t (cudaGraphNode_t, struct cudaExternalSemaphoreWaitNodeParams *)",
+  "cudaGraphExternalSemaphoresWaitNodeSetParams": "cudaError_t (cudaGraphNode_t, struct cudaExternalSemaphoreWaitNodeParams const *)",
+  "cudaGraphGetEdges": "cudaError_t (cudaGraph_t, cudaGraphNode_t *, cudaGraphNode_t *, cudaGraphEdgeData *, size_t *)",
+  "cudaGraphGetNodes": "cudaError_t (cudaGraph_t, cudaGraphNode_t *, size_t *)",
+  "cudaGraphGetRootNodes": "cudaError_t (cudaGraph_t, cudaGraphNode_t *, size_t *)",
+  "cudaGraphHostNodeGetParams": "cudaError_t (cudaGraphNode_t, struct cudaHostNodeParams *)",
+  "cudaGraphHostNodeSetParams": "cudaError_t (cudaGraphNode_t, struct cudaHostNodeParams const *)",
+  "cudaGraphInstantiate": "cudaError_t (cudaGraphExec_t *, cudaGraph_t, unsigned PY_LONG_LONG)",
+  "cudaGraphInstantiateWithFlags": "cudaError_t (cudaGraphExec_t *, cudaGraph_t, unsigned PY_LONG_LONG)",
+  "cudaGraphInstantiateWithParams": "cudaError_t (cudaGraphExec_t *, cudaGraph_t, cudaGraphInstantiateParams *)",
+  "cudaGraphKernelNodeCopyAttributes": "cudaError_t (cudaGraphNode_t, cudaGraphNode_t)",
+  "cudaGraphKernelNodeGetAttribute": "cudaError_t (cudaGraphNode_t, __pyx_t_4cuda_8bindings_9cyruntime_cudaKernelNodeAttrID, __pyx_t_4cuda_8bindings_9cyruntime_cudaKernelNodeAttrValue *)",
+  "cudaGraphKernelNodeGetParams": "cudaError_t (cudaGraphNode_t, struct cudaKernelNodeParams *)",
+  "cudaGraphKernelNodeSetAttribute": "cudaError_t (cudaGraphNode_t, __pyx_t_4cuda_8bindings_9cyruntime_cudaKernelNodeAttrID, __pyx_t_4cuda_8bindings_9cyruntime_cudaKernelNodeAttrValue const *)",
+  "cudaGraphKernelNodeSetParams": "cudaError_t (cudaGraphNode_t, struct cudaKernelNodeParams const *)",
+  "cudaGraphLaunch": "cudaError_t (cudaGraphExec_t, cudaStream_t)",
+  "cudaGraphMemAllocNodeGetParams": "cudaError_t (cudaGraphNode_t, struct cudaMemAllocNodeParams *)",
+  "cudaGraphMemFreeNodeGetParams": "cudaError_t (cudaGraphNode_t, void *)",
+  "cudaGraphMemcpyNodeGetParams": "cudaError_t (cudaGraphNode_t, struct cudaMemcpy3DParms *)",
+  "cudaGraphMemcpyNodeSetParams": "cudaError_t (cudaGraphNode_t, struct cudaMemcpy3DParms const *)",
+  "cudaGraphMemcpyNodeSetParams1D": "cudaError_t (cudaGraphNode_t, void *, void const *, size_t, enum cudaMemcpyKind)",
+  "cudaGraphMemsetNodeGetParams": "cudaError_t (cudaGraphNode_t, struct cudaMemsetParams *)",
+  "cudaGraphMemsetNodeSetParams": "cudaError_t (cudaGraphNode_t, struct cudaMemsetParams const *)",
+  "cudaGraphNodeFindInClone": "cudaError_t (cudaGraphNode_t *, cudaGraphNode_t, cudaGraph_t)",
+  "cudaGraphNodeGetDependencies": "cudaError_t (cudaGraphNode_t, cudaGraphNode_t *, cudaGraphEdgeData *, size_t *)",
+  "cudaGraphNodeGetDependentNodes": "cudaError_t (cudaGraphNode_t, cudaGraphNode_t *, cudaGraphEdgeData *, size_t *)",
+  "cudaGraphNodeGetEnabled": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, unsigned int *)",
+  "cudaGraphNodeGetType": "cudaError_t (cudaGraphNode_t, enum cudaGraphNodeType *)",
+  "cudaGraphNodeSetEnabled": "cudaError_t (cudaGraphExec_t, cudaGraphNode_t, unsigned int)",
+  "cudaGraphNodeSetParams": "cudaError_t (cudaGraphNode_t, struct cudaGraphNodeParams *)",
+  "cudaGraphReleaseUserObject": "cudaError_t (cudaGraph_t, cudaUserObject_t, unsigned int)",
+  "cudaGraphRemoveDependencies": "cudaError_t (cudaGraph_t, cudaGraphNode_t const *, cudaGraphNode_t const *, cudaGraphEdgeData const *, size_t)",
+  "cudaGraphRetainUserObject": "cudaError_t (cudaGraph_t, cudaUserObject_t, unsigned int, unsigned int)",
+  "cudaGraphUpload": "cudaError_t (cudaGraphExec_t, cudaStream_t)",
+  "cudaGraphicsEGLRegisterImage": "cudaError_t (struct cudaGraphicsResource **, __pyx_t_4cuda_8bindings_9cyruntime_EGLImageKHR, unsigned int)",
+  "cudaGraphicsGLRegisterBuffer": "cudaError_t (struct cudaGraphicsResource **, __pyx_t_4cuda_8bindings_9cyruntime_GLuint, unsigned int)",
+  "cudaGraphicsGLRegisterImage": "cudaError_t (struct cudaGraphicsResource **, __pyx_t_4cuda_8bindings_9cyruntime_GLuint, __pyx_t_4cuda_8bindings_9cyruntime_GLenum, unsigned int)",
+  "cudaGraphicsMapResources": "cudaError_t (int, cudaGraphicsResource_t *, cudaStream_t)",
+  "cudaGraphicsResourceGetMappedEglFrame": "cudaError_t (__pyx_t_4cuda_8bindings_9cyruntime_cudaEglFrame *, cudaGraphicsResource_t, unsigned int, unsigned int)",
+  "cudaGraphicsResourceGetMappedMipmappedArray": "cudaError_t (cudaMipmappedArray_t *, cudaGraphicsResource_t)",
+  "cudaGraphicsResourceGetMappedPointer": "cudaError_t (void **, size_t *, cudaGraphicsResource_t)",
+  "cudaGraphicsResourceSetMapFlags": "cudaError_t (cudaGraphicsResource_t, unsigned int)",
+  "cudaGraphicsSubResourceGetMappedArray": "cudaError_t (cudaArray_t *, cudaGraphicsResource_t, unsigned int, unsigned int)",
+  "cudaGraphicsUnmapResources": "cudaError_t (int, cudaGraphicsResource_t *, cudaStream_t)",
+  "cudaGraphicsUnregisterResource": "cudaError_t (cudaGraphicsResource_t)",
+  "cudaGraphicsVDPAURegisterOutputSurface": "cudaError_t (struct cudaGraphicsResource **, __pyx_t_4cuda_8bindings_9cyruntime_VdpOutputSurface, unsigned int)",
+  "cudaGraphicsVDPAURegisterVideoSurface": "cudaError_t (struct cudaGraphicsResource **, __pyx_t_4cuda_8bindings_9cyruntime_VdpVideoSurface, unsigned int)",
+  "cudaHostAlloc": "cudaError_t (void **, size_t, unsigned int)",
+  "cudaHostGetDevicePointer": "cudaError_t (void **, void *, unsigned int)",
+  "cudaHostGetFlags": "cudaError_t (unsigned int *, void *)",
+  "cudaHostRegister": "cudaError_t (void *, size_t, unsigned int)",
+  "cudaHostUnregister": "cudaError_t (void *)",
+  "cudaImportExternalMemory": "cudaError_t (cudaExternalMemory_t *, struct cudaExternalMemoryHandleDesc const *)",
+  "cudaImportExternalSemaphore": "cudaError_t (cudaExternalSemaphore_t *, struct cudaExternalSemaphoreHandleDesc const *)",
+  "cudaInitDevice": "cudaError_t (int, unsigned int, unsigned int)",
+  "cudaIpcCloseMemHandle": "cudaError_t (void *)",
+  "cudaIpcGetEventHandle": "cudaError_t (cudaIpcEventHandle_t *, cudaEvent_t)",
+  "cudaIpcGetMemHandle": "cudaError_t (cudaIpcMemHandle_t *, void *)",
+  "cudaIpcOpenEventHandle": "cudaError_t (cudaEvent_t *, cudaIpcEventHandle_t)",
+  "cudaIpcOpenMemHandle": "cudaError_t (void **, cudaIpcMemHandle_t, unsigned int)",
+  "cudaKernelSetAttributeForDevice": "cudaError_t (cudaKernel_t, enum cudaFuncAttribute, int, int)",
+  "cudaLaunchHostFunc": "cudaError_t (cudaStream_t, cudaHostFn_t, void *)",
+  "cudaLibraryEnumerateKernels": "cudaError_t (cudaKernel_t *, unsigned int, cudaLibrary_t)",
+  "cudaLibraryGetGlobal": "cudaError_t (void **, size_t *, cudaLibrary_t, char const *)",
+  "cudaLibraryGetKernel": "cudaError_t (cudaKernel_t *, cudaLibrary_t, char const *)",
+  "cudaLibraryGetKernelCount": "cudaError_t (unsigned int *, cudaLibrary_t)",
+  "cudaLibraryGetManaged": "cudaError_t (void **, size_t *, cudaLibrary_t, char const *)",
+  "cudaLibraryGetUnifiedFunction": "cudaError_t (void **, cudaLibrary_t, char const *)",
+  "cudaLibraryLoadData": "cudaError_t (cudaLibrary_t *, void const *, enum cudaJitOption *, void **, unsigned int, enum cudaLibraryOption *, void **, unsigned int)",
+  "cudaLibraryLoadFromFile": "cudaError_t (cudaLibrary_t *, char const *, enum cudaJitOption *, void **, unsigned int, enum cudaLibraryOption *, void **, unsigned int)",
+  "cudaLibraryUnload": "cudaError_t (cudaLibrary_t)",
+  "cudaLogsCurrent": "cudaError_t (cudaLogIterator *, unsigned int)",
+  "cudaLogsDumpToFile": "cudaError_t (cudaLogIterator *, char const *, unsigned int)",
+  "cudaLogsDumpToMemory": "cudaError_t (cudaLogIterator *, char *, size_t *, unsigned int)",
+  "cudaLogsRegisterCallback": "cudaError_t (cudaLogsCallback_t, void *, cudaLogsCallbackHandle *)",
+  "cudaLogsUnregisterCallback": "cudaError_t (cudaLogsCallbackHandle)",
+  "cudaMalloc": "cudaError_t (void **, size_t)",
+  "cudaMalloc3D": "cudaError_t (struct cudaPitchedPtr *, struct cudaExtent)",
+  "cudaMalloc3DArray": "cudaError_t (cudaArray_t *, struct cudaChannelFormatDesc const *, struct cudaExtent, unsigned int)",
+  "cudaMallocArray": "cudaError_t (cudaArray_t *, struct cudaChannelFormatDesc const *, size_t, size_t, unsigned int)",
+  "cudaMallocAsync": "cudaError_t (void **, size_t, cudaStream_t)",
+  "cudaMallocFromPoolAsync": "cudaError_t (void **, size_t, cudaMemPool_t, cudaStream_t)",
+  "cudaMallocHost": "cudaError_t (void **, size_t)",
+  "cudaMallocManaged": "cudaError_t (void **, size_t, unsigned int)",
+  "cudaMallocMipmappedArray": "cudaError_t (cudaMipmappedArray_t *, struct cudaChannelFormatDesc const *, struct cudaExtent, unsigned int, unsigned int)",
+  "cudaMallocPitch": "cudaError_t (void **, size_t *, size_t, size_t)",
+  "cudaMemAdvise": "cudaError_t (void const *, size_t, enum cudaMemoryAdvise, struct cudaMemLocation)",
+  "cudaMemDiscardAndPrefetchBatchAsync": "cudaError_t (void **, size_t *, size_t, struct cudaMemLocation *, size_t *, size_t, unsigned PY_LONG_LONG, cudaStream_t)",
+  "cudaMemDiscardBatchAsync": "cudaError_t (void **, size_t *, size_t, unsigned PY_LONG_LONG, cudaStream_t)",
+  "cudaMemGetDefaultMemPool": "cudaError_t (cudaMemPool_t *, struct cudaMemLocation *, enum cudaMemAllocationType)",
+  "cudaMemGetInfo": "cudaError_t (size_t *, size_t *)",
+  "cudaMemGetMemPool": "cudaError_t (cudaMemPool_t *, struct cudaMemLocation *, enum cudaMemAllocationType)",
+  "cudaMemPoolCreate": "cudaError_t (cudaMemPool_t *, struct cudaMemPoolProps const *)",
+  "cudaMemPoolDestroy": "cudaError_t (cudaMemPool_t)",
+  "cudaMemPoolExportPointer": "cudaError_t (struct cudaMemPoolPtrExportData *, void *)",
+  "cudaMemPoolExportToShareableHandle": "cudaError_t (void *, cudaMemPool_t, enum cudaMemAllocationHandleType, unsigned int)",
+  "cudaMemPoolGetAccess": "cudaError_t (enum cudaMemAccessFlags *, cudaMemPool_t, struct cudaMemLocation *)",
+  "cudaMemPoolGetAttribute": "cudaError_t (cudaMemPool_t, enum cudaMemPoolAttr, void *)",
+  "cudaMemPoolImportFromShareableHandle": "cudaError_t (cudaMemPool_t *, void *, enum cudaMemAllocationHandleType, unsigned int)",
+  "cudaMemPoolImportPointer": "cudaError_t (void **, cudaMemPool_t, struct cudaMemPoolPtrExportData *)",
+  "cudaMemPoolSetAccess": "cudaError_t (cudaMemPool_t, struct cudaMemAccessDesc const *, size_t)",
+  "cudaMemPoolSetAttribute": "cudaError_t (cudaMemPool_t, enum cudaMemPoolAttr, void *)",
+  "cudaMemPoolTrimTo": "cudaError_t (cudaMemPool_t, size_t)",
+  "cudaMemPrefetchAsync": "cudaError_t (void const *, size_t, struct cudaMemLocation, unsigned int, cudaStream_t)",
+  "cudaMemPrefetchBatchAsync": "cudaError_t (void **, size_t *, size_t, struct cudaMemLocation *, size_t *, size_t, unsigned PY_LONG_LONG, cudaStream_t)",
+  "cudaMemRangeGetAttribute": "cudaError_t (void *, size_t, enum cudaMemRangeAttribute, void const *, size_t)",
+  "cudaMemRangeGetAttributes": "cudaError_t (void **, size_t *, enum cudaMemRangeAttribute *, size_t, void const *, size_t)",
+  "cudaMemSetMemPool": "cudaError_t (struct cudaMemLocation *, enum cudaMemAllocationType, cudaMemPool_t)",
+  "cudaMemcpy": "cudaError_t (void *, void const *, size_t, enum cudaMemcpyKind)",
+  "cudaMemcpy2D": "cudaError_t (void *, size_t, void const *, size_t, size_t, size_t, enum cudaMemcpyKind)",
+  "cudaMemcpy2DArrayToArray": "cudaError_t (cudaArray_t, size_t, size_t, cudaArray_const_t, size_t, size_t, size_t, size_t, enum cudaMemcpyKind)",
+  "cudaMemcpy2DAsync": "cudaError_t (void *, size_t, void const *, size_t, size_t, size_t, enum cudaMemcpyKind, cudaStream_t)",
+  "cudaMemcpy2DFromArray": "cudaError_t (void *, size_t, cudaArray_const_t, size_t, size_t, size_t, size_t, enum cudaMemcpyKind)",
+  "cudaMemcpy2DFromArrayAsync": "cudaError_t (void *, size_t, cudaArray_const_t, size_t, size_t, size_t, size_t, enum cudaMemcpyKind, cudaStream_t)",
+  "cudaMemcpy2DToArray": "cudaError_t (cudaArray_t, size_t, size_t, void const *, size_t, size_t, size_t, enum cudaMemcpyKind)",
+  "cudaMemcpy2DToArrayAsync": "cudaError_t (cudaArray_t, size_t, size_t, void const *, size_t, size_t, size_t, enum cudaMemcpyKind, cudaStream_t)",
+  "cudaMemcpy3D": "cudaError_t (struct cudaMemcpy3DParms const *)",
+  "cudaMemcpy3DAsync": "cudaError_t (struct cudaMemcpy3DParms const *, cudaStream_t)",
+  "cudaMemcpy3DBatchAsync": "cudaError_t (size_t, struct cudaMemcpy3DBatchOp *, unsigned PY_LONG_LONG, cudaStream_t)",
+  "cudaMemcpy3DPeer": "cudaError_t (struct cudaMemcpy3DPeerParms const *)",
+  "cudaMemcpy3DPeerAsync": "cudaError_t (struct cudaMemcpy3DPeerParms const *, cudaStream_t)",
+  "cudaMemcpyArrayToArray": "cudaError_t (cudaArray_t, size_t, size_t, cudaArray_const_t, size_t, size_t, size_t, enum cudaMemcpyKind)",
+  "cudaMemcpyAsync": "cudaError_t (void *, void const *, size_t, enum cudaMemcpyKind, cudaStream_t)",
+  "cudaMemcpyBatchAsync": "cudaError_t (void const **, void const **, size_t const *, size_t, struct cudaMemcpyAttributes *, size_t *, size_t, cudaStream_t)",
+  "cudaMemcpyFromArray": "cudaError_t (void *, cudaArray_const_t, size_t, size_t, size_t, enum cudaMemcpyKind)",
+  "cudaMemcpyFromArrayAsync": "cudaError_t (void *, cudaArray_const_t, size_t, size_t, size_t, enum cudaMemcpyKind, cudaStream_t)",
+  "cudaMemcpyPeer": "cudaError_t (void *, int, void const *, int, size_t)",
+  "cudaMemcpyPeerAsync": "cudaError_t (void *, int, void const *, int, size_t, cudaStream_t)",
+  "cudaMemcpyToArray": "cudaError_t (cudaArray_t, size_t, size_t, void const *, size_t, enum cudaMemcpyKind)",
+  "cudaMemcpyToArrayAsync": "cudaError_t (cudaArray_t, size_t, size_t, void const *, size_t, enum cudaMemcpyKind, cudaStream_t)",
+  "cudaMemset": "cudaError_t (void *, int, size_t)",
+  "cudaMemset2D": "cudaError_t (void *, size_t, int, size_t, size_t)",
+  "cudaMemset2DAsync": "cudaError_t (void *, size_t, int, size_t, size_t, cudaStream_t)",
+  "cudaMemset3D": "cudaError_t (struct cudaPitchedPtr, int, struct cudaExtent)",
+  "cudaMemset3DAsync": "cudaError_t (struct cudaPitchedPtr, int, struct cudaExtent, cudaStream_t)",
+  "cudaMemsetAsync": "cudaError_t (void *, int, size_t, cudaStream_t)",
+  "cudaMipmappedArrayGetMemoryRequirements": "cudaError_t (struct cudaArrayMemoryRequirements *, cudaMipmappedArray_t, int)",
+  "cudaMipmappedArrayGetSparseProperties": "cudaError_t (struct cudaArraySparseProperties *, cudaMipmappedArray_t)",
+  "cudaOccupancyAvailableDynamicSMemPerBlock": "cudaError_t (size_t *, void const *, int, int)",
+  "cudaOccupancyMaxActiveBlocksPerMultiprocessor": "cudaError_t (int *, void const *, int, size_t)",
+  "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags": "cudaError_t (int *, void const *, int, size_t, unsigned int)",
+  "cudaPeekAtLastError": "cudaError_t (void)",
+  "cudaPointerGetAttributes": "cudaError_t (struct cudaPointerAttributes *, void const *)",
+  "cudaProfilerStart": "cudaError_t (void)",
+  "cudaProfilerStop": "cudaError_t (void)",
+  "cudaRuntimeGetVersion": "cudaError_t (int *)",
+  "cudaSetDevice": "cudaError_t (int)",
+  "cudaSetDeviceFlags": "cudaError_t (unsigned int)",
+  "cudaSignalExternalSemaphoresAsync": "cudaError_t (cudaExternalSemaphore_t const *, struct cudaExternalSemaphoreSignalParams const *, unsigned int, cudaStream_t)",
+  "cudaStreamAddCallback": "cudaError_t (cudaStream_t, cudaStreamCallback_t, void *, unsigned int)",
+  "cudaStreamAttachMemAsync": "cudaError_t (cudaStream_t, void *, size_t, unsigned int)",
+  "cudaStreamBeginCapture": "cudaError_t (cudaStream_t, enum cudaStreamCaptureMode)",
+  "cudaStreamBeginCaptureToGraph": "cudaError_t (cudaStream_t, cudaGraph_t, cudaGraphNode_t const *, cudaGraphEdgeData const *, size_t, enum cudaStreamCaptureMode)",
+  "cudaStreamCopyAttributes": "cudaError_t (cudaStream_t, cudaStream_t)",
+  "cudaStreamCreate": "cudaError_t (cudaStream_t *)",
+  "cudaStreamCreateWithFlags": "cudaError_t (cudaStream_t *, unsigned int)",
+  "cudaStreamCreateWithPriority": "cudaError_t (cudaStream_t *, unsigned int, int)",
+  "cudaStreamDestroy": "cudaError_t (cudaStream_t)",
+  "cudaStreamEndCapture": "cudaError_t (cudaStream_t, cudaGraph_t *)",
+  "cudaStreamGetAttribute": "cudaError_t (cudaStream_t, __pyx_t_4cuda_8bindings_9cyruntime_cudaStreamAttrID, __pyx_t_4cuda_8bindings_9cyruntime_cudaStreamAttrValue *)",
+  "cudaStreamGetCaptureInfo": "cudaError_t (cudaStream_t, enum cudaStreamCaptureStatus *, unsigned PY_LONG_LONG *, cudaGraph_t *, cudaGraphNode_t const **, cudaGraphEdgeData const **, size_t *)",
+  "cudaStreamGetDevice": "cudaError_t (cudaStream_t, int *)",
+  "cudaStreamGetFlags": "cudaError_t (cudaStream_t, unsigned int *)",
+  "cudaStreamGetId": "cudaError_t (cudaStream_t, unsigned PY_LONG_LONG *)",
+  "cudaStreamGetPriority": "cudaError_t (cudaStream_t, int *)",
+  "cudaStreamIsCapturing": "cudaError_t (cudaStream_t, enum cudaStreamCaptureStatus *)",
+  "cudaStreamQuery": "cudaError_t (cudaStream_t)",
+  "cudaStreamSetAttribute": "cudaError_t (cudaStream_t, __pyx_t_4cuda_8bindings_9cyruntime_cudaStreamAttrID, __pyx_t_4cuda_8bindings_9cyruntime_cudaStreamAttrValue const *)",
+  "cudaStreamSynchronize": "cudaError_t (cudaStream_t)",
+  "cudaStreamUpdateCaptureDependencies": "cudaError_t (cudaStream_t, cudaGraphNode_t *, cudaGraphEdgeData const *, size_t, unsigned int)",
+  "cudaStreamWaitEvent": "cudaError_t (cudaStream_t, cudaEvent_t, unsigned int)",
+  "cudaThreadExchangeStreamCaptureMode": "cudaError_t (enum cudaStreamCaptureMode *)",
+  "cudaUserObjectCreate": "cudaError_t (cudaUserObject_t *, void *, cudaHostFn_t, unsigned int, unsigned int)",
+  "cudaUserObjectRelease": "cudaError_t (cudaUserObject_t, unsigned int)",
+  "cudaUserObjectRetain": "cudaError_t (cudaUserObject_t, unsigned int)",
+  "cudaVDPAUGetDevice": "cudaError_t (int *, __pyx_t_4cuda_8bindings_9cyruntime_VdpDevice, __pyx_t_4cuda_8bindings_9cyruntime_VdpGetProcAddress *)",
+  "cudaVDPAUSetVDPAUDevice": "cudaError_t (int, __pyx_t_4cuda_8bindings_9cyruntime_VdpDevice, __pyx_t_4cuda_8bindings_9cyruntime_VdpGetProcAddress *)",
+  "cudaWaitExternalSemaphoresAsync": "cudaError_t (cudaExternalSemaphore_t const *, struct cudaExternalSemaphoreWaitParams const *, unsigned int, cudaStream_t)",
+  "getLocalRuntimeVersion": "cudaError_t (int *)",
+  "make_cudaExtent": "struct cudaExtent (size_t, size_t, size_t)",
+  "make_cudaPitchedPtr": "struct cudaPitchedPtr (void *, size_t, size_t, size_t)",
+  "make_cudaPos": "struct cudaPos (size_t, size_t, size_t)"
+}
diff --git a/cuda_bindings/tests/abi/cuda/bindings/nvjitlink.abi.json b/cuda_bindings/tests/abi/cuda/bindings/nvjitlink.abi.json
new file mode 100644
index 000000000..040b8aa34
--- /dev/null
+++ b/cuda_bindings/tests/abi/cuda/bindings/nvjitlink.abi.json
@@ -0,0 +1,15 @@
+{
+  "add_data": "PyObject *(intptr_t, int, PyObject *, size_t, PyObject *, int __pyx_skip_dispatch)",
+  "add_file": "PyObject *(intptr_t, int, PyObject *, int __pyx_skip_dispatch)",
+  "complete": "PyObject *(intptr_t, int __pyx_skip_dispatch)",
+  "create": "intptr_t (uint32_t, PyObject *, int __pyx_skip_dispatch)",
+  "get_error_log": "PyObject *(intptr_t, PyObject *, int __pyx_skip_dispatch)",
+  "get_error_log_size": "size_t (intptr_t, int __pyx_skip_dispatch)",
+  "get_info_log": "PyObject *(intptr_t, PyObject *, int __pyx_skip_dispatch)",
+  "get_info_log_size": "size_t (intptr_t, int __pyx_skip_dispatch)",
+  "get_linked_cubin": "PyObject *(intptr_t, PyObject *, int __pyx_skip_dispatch)",
+  "get_linked_cubin_size": "size_t (intptr_t, int __pyx_skip_dispatch)",
+  "get_linked_ptx": "PyObject *(intptr_t, PyObject *, int __pyx_skip_dispatch)",
+  "get_linked_ptx_size": "size_t (intptr_t, int __pyx_skip_dispatch)",
+  "version": "PyObject *(int __pyx_skip_dispatch)"
+}
diff --git a/cuda_bindings/tests/abi/cuda/bindings/nvvm.abi.json b/cuda_bindings/tests/abi/cuda/bindings/nvvm.abi.json
new file mode 100644
index 000000000..41a9416f1
--- /dev/null
+++ b/cuda_bindings/tests/abi/cuda/bindings/nvvm.abi.json
@@ -0,0 +1,14 @@
+{
+  "add_module_to_program": "PyObject *(intptr_t, PyObject *, size_t, PyObject *, int __pyx_skip_dispatch)",
+  "compile_program": "PyObject *(intptr_t, int, PyObject *, int __pyx_skip_dispatch)",
+  "create_program": "intptr_t (int __pyx_skip_dispatch)",
+  "get_compiled_result": "PyObject *(intptr_t, PyObject *, int __pyx_skip_dispatch)",
+  "get_compiled_result_size": "size_t (intptr_t, int __pyx_skip_dispatch)",
+  "get_error_string": "PyObject *(int, int __pyx_skip_dispatch)",
+  "get_program_log": "PyObject *(intptr_t, PyObject *, int __pyx_skip_dispatch)",
+  "get_program_log_size": "size_t (intptr_t, int __pyx_skip_dispatch)",
+  "ir_version": "PyObject *(int __pyx_skip_dispatch)",
+  "lazy_add_module_to_program": "PyObject *(intptr_t, PyObject *, size_t, PyObject *, int __pyx_skip_dispatch)",
+  "verify_program": "PyObject *(intptr_t, int, PyObject *, int __pyx_skip_dispatch)",
+  "version": "PyObject *(int __pyx_skip_dispatch)"
+}
diff --git a/cuda_bindings/tests/test_cython_abi.py b/cuda_bindings/tests/test_cython_abi.py
new file mode 100644
index 000000000..73fe58e2e
--- /dev/null
+++ b/cuda_bindings/tests/test_cython_abi.py
@@ -0,0 +1,149 @@
+# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# SPDX-License-Identifier: Apache-2.0
+
+
+import importlib
+import json
+import re
+import sysconfig
+from pathlib import Path
+
+EXT_SUFFIX = sysconfig.get_config_var("EXT_SUFFIX")
+ABI_SUFFIX = ".abi.json"
+PACKAGE = "cuda.bindings"
+ABI_PATH = Path(__file__).parent / Path("abi/cuda/bindings")
+
+
+def short_stem(name: str) -> str:
+    return name[: name.find(".")]
+
+
+def get_package_path(package_name: str) -> Path:
+    package = importlib.import_module(package_name)
+    return Path(package.__file__).parent
+
+
+def import_from_path(root_package: str, root_dir: Path, path: Path) -> object:
+    path = path.relative_to(root_dir)
+    parts = [root_package] + list(path.parts[:-1]) + [short_stem(path.name)]
+    return importlib.import_module(".".join(parts))
+
+
+def so_path_to_abi_path(so_path: Path, build_dir: Path, abi_dir: Path) -> Path:
+    abi_name = short_stem(so_path.name) + ABI_SUFFIX
+    return abi_dir / so_path.parent.relative_to(build_dir) / abi_name
+
+
+def abi_path_to_so_path(abi_path: Path, build_dir: Path, abi_dir: Path) -> Path:
+    so_name = short_stem(abi_path.name) + EXT_SUFFIX
+    return build_dir / abi_path.parent.relative_to(abi_dir) / so_name
+
+
+def pyx_capi_to_json(d: dict[str, object]) -> dict[str, str]:
+    """
+    Converts the __pyx_capi__ dictionary to a JSON-serializable dictionary,
+    removing any memory addresses that are irrelevant for comparison.
+    """
+
+    def extract_name(v: object) -> str:
+        v = str(v)
+        match = re.match(r'<capsule object "([^\"]+)" at 0x[0-9a-fA-F]+>', v)
+        assert match, f"Could not parse __pyx_capi__ entry: {v}"
+        return match.group(1)
+
+    # Sort the dictionary by keys to make diffs in the JSON files smaller
+    return {k: extract_name(d[k]) for k in sorted(d.keys())}
+
+
+def check_abi(expected: dict[str, str], found: dict[str, str]) -> tuple[bool, bool]:
+    has_errors = False
+    has_allowed_changes = False
+    for k, v in expected.items():
+        if k not in found:
+            print(f"  Missing symbol: {k}")
+            has_errors = True
+        elif found[k] != v:
+            print(f"  Changed symbol: {k}: expected {v}, got {found[k]}")
+            has_errors = True
+    for k, v in found.items():
+        if k not in expected:
+            print(f"  Added symbol: {k}")
+            has_allowed_changes = True
+    return has_errors, has_allowed_changes
+
+
+def check(package: str, abi_dir: Path) -> tuple[bool, bool]:
+    build_dir = get_package_path(package)
+
+    has_errors = False
+    has_allowed_changes = False
+    for abi_path in Path(abi_dir).glob(f"**/*{ABI_SUFFIX}"):
+        so_path = abi_path_to_so_path(abi_path, build_dir, abi_dir)
+        if so_path.is_file():
+            module = import_from_path(package, build_dir, so_path)
+            if hasattr(module, "__pyx_capi__"):
+                found_json = pyx_capi_to_json(module.__pyx_capi__)
+                with open(abi_path, encoding="utf-8") as f:
+                    expected_json = json.load(f)
+                print(f"Checking module: {so_path.relative_to(build_dir)}")
+                check_errors, check_allowed_changes = check_abi(expected_json, found_json)
+                has_errors |= check_errors
+                has_allowed_changes |= check_allowed_changes
+            else:
+                print(f"Module no longer has an exposed ABI: {so_path.relative_to(build_dir)}")
+                has_errors = True
+        else:
+            print(f"No module found for {abi_path.relative_to(abi_dir)}")
+            has_errors = True
+
+    for so_path in Path(build_dir).glob(f"**/*{EXT_SUFFIX}"):
+        module = import_from_path(package, build_dir, so_path)
+        if hasattr(module, "__pyx_capi__"):
+            abi_path = so_path_to_abi_path(so_path, build_dir, abi_dir)
+            if not abi_path.is_file():
+                print(f"New module added {so_path.relative_to(build_dir)}")
+                has_allowed_changes = True
+
+    return has_errors, has_allowed_changes
+
+
+def regenerate(package: str, abi_dir: Path) -> None:
+    build_dir = get_package_path(package)
+    for so_path in Path(build_dir).glob("**/*{EXT_SUFFIX}"):
+        print(f"Generating ABI from {so_path.relative_to(build_dir)}")
+        module = import_from_path(package, build_dir, so_path)
+        if hasattr(module, "__pyx_capi__"):
+            abi_path = so_path_to_abi_path(so_path, build_dir, abi_dir)
+            abi_path.parent.mkdir(parents=True, exist_ok=True)
+            with open(abi_path, "w", encoding="utf-8") as f:
+                json.dump(pyx_capi_to_json(module.__pyx_capi__), f, indent=2)
+
+
+def test_cython_abi_changes():
+    has_errors, has_allowed_changes = check(PACKAGE, ABI_PATH)
+
+    if has_errors:
+        print("Cython ABI check failed.")
+        print("We generally do not allow these changes during a major release,")
+        print("but if they are expected, you can regenerate the ABI files by")
+        print("running:")
+        print("")
+        print("  cuda_bindings/regenerate_abi.sh")
+        print("")
+        print("and committing the result.")
+        raise AssertionError("Cython ABI check failed")
+
+    if has_allowed_changes:
+        print("New elements were added to the Cython ABI.")
+        print("This is allowed at any time, but the ABI files must be regenerated.")
+        print("You can regenerate the ABI files by running:")
+        print("")
+        print("  cuda_bindings/regenerate_abi.sh")
+        print("")
+        print("and committing the result.")
+        raise AssertionError("Cython ABI updated")
+
+
+if __name__ == "__main__":
+    regenerate(PACKAGE, ABI_PATH)