NVIDIA CUDA

Reference Manual

Version 2.3
July 2009
Contents

1 Module Index 1
1.1 Modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

2 Data Structure Index 3

2.1 Data Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

3 Module Documentation 5
3.1 CUDA Runtime API . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.1.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.2 Thread Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.2.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.2.2 Function Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.2.2.1 cudaThreadExit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.2.2.2 cudaThreadSynchronize . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.3 Error Handling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
3.3.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
3.3.2 Function Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
3.3.2.1 cudaGetErrorString . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
3.3.2.2 cudaGetLastError . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
3.4 Device Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.4.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.4.2 Function Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.4.2.1 cudaChooseDevice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.4.2.2 cudaGetDevice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.4.2.3 cudaGetDeviceCount . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.4.2.4 cudaGetDeviceProperties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.4.2.5 cudaSetDevice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.4.2.6 cudaSetDeviceFlags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.4.2.7 cudaSetValidDevices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
ii CONTENTS

3.5 Stream Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

3.5.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.5.2 Function Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.5.2.1 cudaStreamCreate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.5.2.2 cudaStreamDestroy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.5.2.3 cudaStreamQuery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.5.2.4 cudaStreamSynchronize . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.6 Event Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.6.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.6.2 Function Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.6.2.1 cudaEventCreate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.6.2.2 cudaEventCreateWithFlags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.6.2.3 cudaEventDestroy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.6.2.4 cudaEventElapsedTime . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.6.2.5 cudaEventQuery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.6.2.6 cudaEventRecord . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.6.2.7 cudaEventSynchronize . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.7 Execution Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.7.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.7.2 Function Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.7.2.1 cudaConfigureCall . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.7.2.2 cudaFuncGetAttributes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.7.2.3 cudaLaunch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.7.2.4 cudaSetDoubleForDevice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.7.2.5 cudaSetDoubleForHost . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.7.2.6 cudaSetupArgument . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.8 Memory Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.8.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.8.2 Function Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.8.2.1 cudaFree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.8.2.2 cudaFreeArray . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.8.2.3 cudaFreeHost . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.8.2.4 cudaGetSymbolAddress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.8.2.5 cudaGetSymbolSize . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.8.2.6 cudaHostAlloc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.8.2.7 cudaHostGetDevicePointer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.8.2.8 cudaHostGetFlags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

Generated for NVIDIA CUDA Library by Doxygen

CONTENTS iii

3.8.2.9 cudaMalloc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.8.2.10 cudaMalloc3D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.8.2.11 cudaMalloc3DArray . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.8.2.12 cudaMallocArray . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.8.2.13 cudaMallocHost . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.8.2.14 cudaMallocPitch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.8.2.15 cudaMemcpy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.8.2.16 cudaMemcpy2D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.8.2.17 cudaMemcpy2DArrayToArray . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.8.2.18 cudaMemcpy2DAsync . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.8.2.19 cudaMemcpy2DFromArray . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.8.2.20 cudaMemcpy2DFromArrayAsync . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.8.2.21 cudaMemcpy2DToArray . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.8.2.22 cudaMemcpy2DToArrayAsync . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.8.2.23 cudaMemcpy3D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.8.2.24 cudaMemcpy3DAsync . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.8.2.25 cudaMemcpyArrayToArray . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.8.2.26 cudaMemcpyAsync . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.8.2.27 cudaMemcpyFromArray . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.8.2.28 cudaMemcpyFromArrayAsync . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.8.2.29 cudaMemcpyFromSymbol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.8.2.30 cudaMemcpyFromSymbolAsync . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.8.2.31 cudaMemcpyToArray . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.8.2.32 cudaMemcpyToArrayAsync . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.8.2.33 cudaMemcpyToSymbol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.8.2.34 cudaMemcpyToSymbolAsync . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.8.2.35 cudaMemset . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.8.2.36 cudaMemset2D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.8.2.37 cudaMemset3D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.9 OpenGL Interoperability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.9.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.9.2 Function Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.9.2.1 cudaGLMapBufferObject . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.9.2.2 cudaGLMapBufferObjectAsync . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.9.2.3 cudaGLRegisterBufferObject . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.9.2.4 cudaGLSetBufferObjectMapFlags . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.9.2.5 cudaGLSetGLDevice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

Generated for NVIDIA CUDA Library by Doxygen

iv CONTENTS

3.9.2.6 cudaGLUnmapBufferObject . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.9.2.7 cudaGLUnmapBufferObjectAsync . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.9.2.8 cudaGLUnregisterBufferObject . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.9.2.9 cudaWGLGetDevice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.10 Direct3D 9 Interoperability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.10.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.10.2 Enumeration Type Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.10.2.1 cudaD3D9MapFlags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.10.2.2 cudaD3D9RegisterFlags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.10.3 Function Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.10.3.1 cudaD3D9GetDevice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.10.3.2 cudaD3D9GetDirect3DDevice . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.10.3.3 cudaD3D9MapResources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.10.3.4 cudaD3D9RegisterResource . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3.10.3.5 cudaD3D9ResourceGetMappedArray . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.10.3.6 cudaD3D9ResourceGetMappedPitch . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.10.3.7 cudaD3D9ResourceGetMappedPointer . . . . . . . . . . . . . . . . . . . . . . . . 63
3.10.3.8 cudaD3D9ResourceGetMappedSize . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.10.3.9 cudaD3D9ResourceGetSurfaceDimensions . . . . . . . . . . . . . . . . . . . . . . 64
3.10.3.10 cudaD3D9ResourceSetMapFlags . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.10.3.11 cudaD3D9SetDirect3DDevice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.10.3.12 cudaD3D9UnmapResources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.10.3.13 cudaD3D9UnregisterResource . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.11 Direct3D 10 Interoperability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.11.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
3.11.2 Enumeration Type Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
3.11.2.1 cudaD3D10MapFlags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
3.11.2.2 cudaD3D10RegisterFlags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
3.11.3 Function Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
3.11.3.1 cudaD3D10GetDevice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
3.11.3.2 cudaD3D10MapResources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.11.3.3 cudaD3D10RegisterResource . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.11.3.4 cudaD3D10ResourceGetMappedArray . . . . . . . . . . . . . . . . . . . . . . . . 72
3.11.3.5 cudaD3D10ResourceGetMappedPitch . . . . . . . . . . . . . . . . . . . . . . . . 72
3.11.3.6 cudaD3D10ResourceGetMappedPointer . . . . . . . . . . . . . . . . . . . . . . . 73
3.11.3.7 cudaD3D10ResourceGetMappedSize . . . . . . . . . . . . . . . . . . . . . . . . . 74
3.11.3.8 cudaD3D10ResourceGetSurfaceDimensions . . . . . . . . . . . . . . . . . . . . . 74

Generated for NVIDIA CUDA Library by Doxygen

CONTENTS v

3.11.3.9 cudaD3D10ResourceSetMapFlags . . . . . . . . . . . . . . . . . . . . . . . . . . 75
3.11.3.10 cudaD3D10SetDirect3DDevice . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
3.11.3.11 cudaD3D10UnmapResources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
3.11.3.12 cudaD3D10UnregisterResource . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
3.12 Texture Reference Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
3.12.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
3.12.2 Function Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
3.12.2.1 cudaBindTexture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
3.12.2.2 cudaBindTexture2D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
3.12.2.3 cudaBindTextureToArray . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
3.12.2.4 cudaCreateChannelDesc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
3.12.2.5 cudaGetChannelDesc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
3.12.2.6 cudaGetTextureAlignmentOffset . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
3.12.2.7 cudaGetTextureReference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
3.12.2.8 cudaUnbindTexture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
3.13 Version Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
3.13.1 Function Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
3.13.1.1 cudaDriverGetVersion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
3.13.1.2 cudaRuntimeGetVersion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
3.14 C++ API Routines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
3.14.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
3.14.2 Function Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
3.14.2.1 cudaBindTexture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
3.14.2.2 cudaBindTexture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
3.14.2.3 cudaBindTexture2D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
3.14.2.4 cudaBindTextureToArray . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
3.14.2.5 cudaBindTextureToArray . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
3.14.2.6 cudaCreateChannelDesc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
3.14.2.7 cudaFuncGetAttributes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
3.14.2.8 cudaGetSymbolAddress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
3.14.2.9 cudaGetSymbolSize . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
3.14.2.10 cudaGetTextureAlignmentOffset . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
3.14.2.11 cudaLaunch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
3.14.2.12 cudaSetupArgument . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
3.14.2.13 cudaUnbindTexture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
3.15 Data types used by CUDA Runtime . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
3.15.1 Typedef Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94

Generated for NVIDIA CUDA Library by Doxygen

vi CONTENTS

3.15.1.1 cudaError_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
3.15.1.2 cudaEvent_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
3.15.1.3 cudaStream_t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
3.15.2 Enumeration Type Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
3.15.2.1 cudaChannelFormatKind . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
3.15.2.2 cudaComputeMode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
3.15.2.3 cudaError . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
3.15.2.4 cudaMemcpyKind . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
3.16 CUDA Driver API . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
3.16.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
3.17 Initialization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
3.17.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
3.17.2 Function Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
3.17.2.1 cuInit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
3.18 Device Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
3.18.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
3.18.2 Function Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
3.18.2.1 cuDeviceComputeCapability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
3.18.2.2 cuDeviceGet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
3.18.2.3 cuDeviceGetAttribute . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
3.18.2.4 cuDeviceGetCount . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
3.18.2.5 cuDeviceGetName . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
3.18.2.6 cuDeviceGetProperties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
3.18.2.7 cuDeviceTotalMem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
3.19 Version Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
3.19.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
3.19.2 Function Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
3.19.2.1 cuDriverGetVersion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
3.20 Context Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
3.20.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
3.20.2 Function Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
3.20.2.1 cuCtxAttach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
3.20.2.2 cuCtxCreate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
3.20.2.3 cuCtxDestroy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
3.20.2.4 cuCtxDetach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
3.20.2.5 cuCtxGetDevice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
3.20.2.6 cuCtxPopCurrent . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

Generated for NVIDIA CUDA Library by Doxygen

CONTENTS vii

3.20.2.7 cuCtxPushCurrent . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110

3.20.2.8 cuCtxSynchronize . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
3.21 Module Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
3.21.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
3.21.2 Function Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
3.21.2.1 cuModuleGetFunction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
3.21.2.2 cuModuleGetGlobal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
3.21.2.3 cuModuleGetTexRef . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
3.21.2.4 cuModuleLoad . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
3.21.2.5 cuModuleLoadData . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
3.21.2.6 cuModuleLoadDataEx . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
3.21.2.7 cuModuleLoadFatBinary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
3.21.2.8 cuModuleUnload . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
3.22 Stream Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
3.22.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
3.22.2 Function Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
3.22.2.1 cuStreamCreate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
3.22.2.2 cuStreamDestroy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
3.22.2.3 cuStreamQuery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
3.22.2.4 cuStreamSynchronize . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
3.23 Event Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
3.23.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
3.23.2 Function Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
3.23.2.1 cuEventCreate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
3.23.2.2 cuEventDestroy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
3.23.2.3 cuEventElapsedTime . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
3.23.2.4 cuEventQuery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
3.23.2.5 cuEventRecord . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
3.23.2.6 cuEventSynchronize . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
3.24 Execution Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
3.24.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
3.24.2 Function Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
3.24.2.1 cuFuncGetAttribute . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
3.24.2.2 cuFuncSetBlockShape . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
3.24.2.3 cuFuncSetSharedSize . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
3.24.2.4 cuLaunch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
3.24.2.5 cuLaunchGrid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125

Generated for NVIDIA CUDA Library by Doxygen

viii CONTENTS

3.24.2.6 cuLaunchGridAsync . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126

3.24.2.7 cuParamSetf . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
3.24.2.8 cuParamSeti . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
3.24.2.9 cuParamSetSize . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
3.24.2.10 cuParamSetTexRef . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
3.24.2.11 cuParamSetv . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
3.25 Memory Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
3.25.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
3.25.2 Function Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
3.25.2.1 cuArray3DCreate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
3.25.2.2 cuArray3DGetDescriptor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
3.25.2.3 cuArrayCreate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
3.25.2.4 cuArrayDestroy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
3.25.2.5 cuArrayGetDescriptor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
3.25.2.6 cuMemAlloc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
3.25.2.7 cuMemAllocHost . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
3.25.2.8 cuMemAllocPitch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
3.25.2.9 cuMemcpy2D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
3.25.2.10 cuMemcpy2DAsync . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
3.25.2.11 cuMemcpy2DUnaligned . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
3.25.2.12 cuMemcpy3D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
3.25.2.13 cuMemcpy3DAsync . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
3.25.2.14 cuMemcpyAtoA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
3.25.2.15 cuMemcpyAtoD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
3.25.2.16 cuMemcpyAtoH . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
3.25.2.17 cuMemcpyAtoHAsync . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
3.25.2.18 cuMemcpyDtoA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
3.25.2.19 cuMemcpyDtoD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
3.25.2.20 cuMemcpyDtoH . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
3.25.2.21 cuMemcpyDtoHAsync . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
3.25.2.22 cuMemcpyHtoA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
3.25.2.23 cuMemcpyHtoAAsync . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
3.25.2.24 cuMemcpyHtoD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
3.25.2.25 cuMemcpyHtoDAsync . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
3.25.2.26 cuMemFree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
3.25.2.27 cuMemFreeHost . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
3.25.2.28 cuMemGetAddressRange . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158

Generated for NVIDIA CUDA Library by Doxygen

CONTENTS ix

3.25.2.29 cuMemGetInfo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159

3.25.2.30 cuMemHostAlloc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
3.25.2.31 cuMemHostGetDevicePointer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
3.25.2.32 cuMemHostGetFlags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
3.25.2.33 cuMemsetD16 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
3.25.2.34 cuMemsetD2D16 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
3.25.2.35 cuMemsetD2D32 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
3.25.2.36 cuMemsetD2D8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
3.25.2.37 cuMemsetD32 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
3.25.2.38 cuMemsetD8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
3.26 Texture Reference Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
3.26.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
3.26.2 Function Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
3.26.2.1 cuTexRefCreate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
3.26.2.2 cuTexRefDestroy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
3.26.2.3 cuTexRefGetAddress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
3.26.2.4 cuTexRefGetAddressMode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
3.26.2.5 cuTexRefGetArray . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
3.26.2.6 cuTexRefGetFilterMode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
3.26.2.7 cuTexRefGetFlags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
3.26.2.8 cuTexRefGetFormat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
3.26.2.9 cuTexRefSetAddress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
3.26.2.10 cuTexRefSetAddress2D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
3.26.2.11 cuTexRefSetAddressMode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
3.26.2.12 cuTexRefSetArray . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
3.26.2.13 cuTexRefSetFilterMode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
3.26.2.14 cuTexRefSetFlags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
3.26.2.15 cuTexRefSetFormat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
3.27 OpenGL Interoperability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
3.27.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
3.27.2 Function Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
3.27.2.1 cuGLCtxCreate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
3.27.2.2 cuGLInit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
3.27.2.3 cuGLMapBufferObject . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
3.27.2.4 cuGLMapBufferObjectAsync . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
3.27.2.5 cuGLRegisterBufferObject . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
3.27.2.6 cuGLSetBufferObjectMapFlags . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177

Generated for NVIDIA CUDA Library by Doxygen

x CONTENTS

3.27.2.7 cuGLUnmapBufferObject . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178

3.27.2.8 cuGLUnmapBufferObjectAsync . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
3.27.2.9 cuGLUnregisterBufferObject . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
3.27.2.10 cuWGLGetDevice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
3.28 Direct3D 9 Interoperability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
3.28.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
3.28.2 Function Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
3.28.2.1 cuD3D9CtxCreate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
3.28.2.2 cuD3D9GetDevice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
3.28.2.3 cuD3D9GetDirect3DDevice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
3.28.2.4 cuD3D9MapResources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
3.28.2.5 cuD3D9RegisterResource . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
3.28.2.6 cuD3D9ResourceGetMappedArray . . . . . . . . . . . . . . . . . . . . . . . . . . 184
3.28.2.7 cuD3D9ResourceGetMappedPitch . . . . . . . . . . . . . . . . . . . . . . . . . . 185
3.28.2.8 cuD3D9ResourceGetMappedPointer . . . . . . . . . . . . . . . . . . . . . . . . . 186
3.28.2.9 cuD3D9ResourceGetMappedSize . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
3.28.2.10 cuD3D9ResourceGetSurfaceDimensions . . . . . . . . . . . . . . . . . . . . . . . 187
3.28.2.11 cuD3D9ResourceSetMapFlags . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
3.28.2.12 cuD3D9UnmapResources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
3.28.2.13 cuD3D9UnregisterResource . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
3.29 Direct3D 10 Interoperability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
3.29.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
3.29.2 Function Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
3.29.2.1 cuD3D10CtxCreate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
3.29.2.2 cuD3D10GetDevice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
3.29.2.3 cuD3D10MapResources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
3.29.2.4 cuD3D10RegisterResource . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
3.29.2.5 cuD3D10ResourceGetMappedArray . . . . . . . . . . . . . . . . . . . . . . . . . 195
3.29.2.6 cuD3D10ResourceGetMappedPitch . . . . . . . . . . . . . . . . . . . . . . . . . . 195
3.29.2.7 cuD3D10ResourceGetMappedPointer . . . . . . . . . . . . . . . . . . . . . . . . 196
3.29.2.8 cuD3D10ResourceGetMappedSize . . . . . . . . . . . . . . . . . . . . . . . . . . 197
3.29.2.9 cuD3D10ResourceGetSurfaceDimensions . . . . . . . . . . . . . . . . . . . . . . 197
3.29.2.10 cuD3D10ResourceSetMapFlags . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
3.29.2.11 cuD3D10UnmapResources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
3.29.2.12 cuD3D10UnregisterResource . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
3.30 Data types used by CUDA driver . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
3.30.1 Define Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205

Generated for NVIDIA CUDA Library by Doxygen

CONTENTS xi

3.30.1.1 CU_MEMHOSTALLOC_DEVICEMAP . . . . . . . . . . . . . . . . . . . . . . . 205

3.30.1.2 CU_MEMHOSTALLOC_PORTABLE . . . . . . . . . . . . . . . . . . . . . . . . 205
3.30.1.3 CU_MEMHOSTALLOC_WRITECOMBINED . . . . . . . . . . . . . . . . . . . 205
3.30.1.4 CU_PARAM_TR_DEFAULT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
3.30.1.5 CU_TRSA_OVERRIDE_FORMAT . . . . . . . . . . . . . . . . . . . . . . . . . 205
3.30.1.6 CU_TRSF_NORMALIZED_COORDINATES . . . . . . . . . . . . . . . . . . . . 205
3.30.1.7 CU_TRSF_READ_AS_INTEGER . . . . . . . . . . . . . . . . . . . . . . . . . . 205
3.30.1.8 CUDA_VERSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
3.30.2 Typedef Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
3.30.2.1 CUaddress_mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
3.30.2.2 CUarray_format . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
3.30.2.3 CUcomputemode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
3.30.2.4 CUctx_flags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
3.30.2.5 CUDA_MEMCPY2D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
3.30.2.6 CUDA_MEMCPY3D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
3.30.2.7 CUdevice_attribute . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
3.30.2.8 CUdevprop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
3.30.2.9 CUevent_flags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
3.30.2.10 CUfilter_mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
3.30.2.11 CUfunction_attribute . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
3.30.2.12 CUjit_fallback . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
3.30.2.13 CUjit_option . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
3.30.2.14 CUjit_target . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
3.30.2.15 CUmemorytype . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
3.30.2.16 CUresult . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
3.30.3 Enumeration Type Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
3.30.3.1 CUaddress_mode_enum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
3.30.3.2 CUarray_format_enum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
3.30.3.3 CUcomputemode_enum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
3.30.3.4 CUctx_flags_enum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
3.30.3.5 cudaError_enum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
3.30.3.6 CUdevice_attribute_enum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
3.30.3.7 CUevent_flags_enum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
3.30.3.8 CUfilter_mode_enum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
3.30.3.9 CUfunction_attribute_enum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
3.30.3.10 CUjit_fallback_enum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
3.30.3.11 CUjit_option_enum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210

Generated for NVIDIA CUDA Library by Doxygen

xii CONTENTS

3.30.3.12 CUjit_target_enum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211

3.30.3.13 CUmemorytype_enum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211

4 Data Structure Documentation 213

4.1 CUDA_ARRAY3D_DESCRIPTOR Struct Reference . . . . . . . . . . . . . . . . . . . . . . . . . . 213
4.1.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
4.2 CUDA_ARRAY_DESCRIPTOR Struct Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
4.2.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
4.3 CUDA_MEMCPY2D_st Struct Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
4.3.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
4.4 CUDA_MEMCPY3D_st Struct Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
4.4.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
4.5 cudaChannelFormatDesc Struct Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
4.5.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
4.6 cudaDeviceProp Struct Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
4.6.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
4.7 cudaExtent Struct Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
4.7.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
4.8 cudaFuncAttributes Struct Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
4.8.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
4.9 cudaMemcpy3DParms Struct Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
4.9.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
4.10 cudaPitchedPtr Struct Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
4.10.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
4.11 cudaPos Struct Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
4.11.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
4.12 CUdevprop_st Struct Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
4.12.1 Detailed Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227

Generated for NVIDIA CUDA Library by Doxygen

Chapter 1

Module Index

1.1 Modules
Here is a list of all modules:
CUDA Runtime API . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Thread Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Error Handling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Device Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Stream Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
Event Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
Execution Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
Memory Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
OpenGL Interoperability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
Direct3D 9 Interoperability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
Direct3D 10 Interoperability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
Texture Reference Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
Version Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
C++ API Routines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
Data types used by CUDA Runtime . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
CUDA Driver API . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
Initialization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
Device Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
Version Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
Context Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
Module Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
Stream Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
Event Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
Execution Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
Memory Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
Texture Reference Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
OpenGL Interoperability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
Direct3D 9 Interoperability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
Direct3D 10 Interoperability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
Data types used by CUDA driver . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
2 Module Index

Generated for NVIDIA CUDA Library by Doxygen

Chapter 2

Data Structure Index

2.1 Data Structures

Here are the data structures with brief descriptions:
CUDA_ARRAY3D_DESCRIPTOR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
CUDA_ARRAY_DESCRIPTOR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
CUDA_MEMCPY2D_st . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
CUDA_MEMCPY3D_st . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
cudaChannelFormatDesc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
cudaDeviceProp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
cudaExtent . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
cudaFuncAttributes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
cudaMemcpy3DParms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
cudaPitchedPtr . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
cudaPos . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
CUdevprop_st . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
4 Data Structure Index

Generated for NVIDIA CUDA Library by Doxygen

Chapter 3

Module Documentation

3.1 CUDA Runtime API

Modules
• Thread Management
• Error Handling
• Device Management
• Stream Management
• Event Management
• Execution Control
• Memory Management
• OpenGL Interoperability
• Direct3D 9 Interoperability
• Direct3D 10 Interoperability
• Texture Reference Management
• Version Management
• C++ API Routines
C++-style interface built on top of CUDA runtime API.

• Data types used by CUDA Runtime

3.1.1 Detailed Description

There are two levels for the runtime API.

The C API (cuda_runtime_api.h) is a C-style interface that does not require compiling with nvcc.
The C++ API (cuda_runtime.h) is a C++-style interface built on top of the C API. It wraps some of the C API routines,
using overloading, references and default arguments. These wrappers can be used from C++ code and can be compiled
with any C++ compiler. The C++ API also has some CUDA-specific wrappers that wrap C API routines that deal with
symbols, textures, and device functions. These wrappers require the use of nvcc because they depend on code being
generated by the compiler. For example, the execution configuration syntax to invoke kernels is only available in
source code compiled with nvcc.
6 Module Documentation

3.2 Thread Management

Functions
• cudaError_t cudaThreadExit (void)
Exit and clean up from CUDA launches.

• cudaError_t cudaThreadSynchronize (void)

Wait for compute device to finish.

3.2.1 Detailed Description

This section describes the thread management functions of the CUDA runtime application programming interface.

3.2.2 Function Documentation

3.2.2.1 cudaError_t cudaThreadExit (void)

Explicitly cleans up all runtime-related resources associated with the calling host thread. Any subsequent API call
reinitializes the runtime. cudaThreadExit() is implicitly called on host thread exit.

Returns:
cudaSuccess

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaThreadSynchronize

3.2.2.2 cudaError_t cudaThreadSynchronize (void)

Blocks until the device has completed all preceding requested tasks. cudaThreadSynchronize() returns an error if one
of the preceding tasks has failed.

Returns:
cudaSuccess

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaThreadExit

Generated for NVIDIA CUDA Library by Doxygen

3.3 Error Handling 7

3.3 Error Handling

Functions
• const char ∗ cudaGetErrorString (cudaError_t error)
Returns the message string from an error code.

• cudaError_t cudaGetLastError (void)

Returns the last error from a runtime call.

3.3.1 Detailed Description

This section describes the error handling functions of the CUDA runtime application programming interface.

3.3.2 Function Documentation

3.3.2.1 const char∗ cudaGetErrorString (cudaError_t error)

Returns the message string from an error code.

Parameters:
error - Error code to convert to string

Returns:
char∗ pointer to a NULL-terminated string

See also:
cudaGetLastError, cudaError

3.3.2.2 cudaError_t cudaGetLastError (void)

Returns the last error that has been produced by any of the runtime calls in the same host thread and resets it to
cudaSuccess.

Returns:
cudaSuccess, cudaErrorMissingConfiguration, cudaErrorMemoryAllocation, cudaErrorInitializationError, cu-
daErrorLaunchFailure, cudaErrorPriorLaunchFailure, cudaErrorLaunchTimeout, cudaErrorLaunchOutOfRe-
sources, cudaErrorInvalidDeviceFunction, cudaErrorInvalidConfiguration, cudaErrorInvalidDevice cudaError-
InvalidValue cudaErrorInvalidPitchValue cudaErrorInvalidSymbol cudaErrorMapBufferObjectFailed cudaEr-
rorUnmapBufferObjectFailed cudaErrorInvalidHostPointer cudaErrorInvalidDevicePointer cudaErrorInvalidTex-
ture cudaErrorInvalidTextureBinding cudaErrorInvalidChannelDescriptor cudaErrorInvalidMemcpyDirection cu-
daErrorAddressOfConstant cudaErrorTextureFetchFailed cudaErrorTextureNotBound cudaErrorSynchroniza-
tionError cudaErrorInvalidFilterSetting cudaErrorInvalidNormSetting cudaErrorMixedDeviceExecution cudaEr-
rorCudartUnloading cudaErrorUnknown cudaErrorNotYetImplemented cudaErrorMemoryValueTooLarge cu-
daErrorInvalidResourceHandle cudaErrorNotReady cudaErrorInsufficientDriver cudaErrorSetOnActiveProcess
cudaErrorStartupFailure cudaErrorApiFailureBase

Generated for NVIDIA CUDA Library by Doxygen

8 Module Documentation

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaGetErrorString, cudaError

Generated for NVIDIA CUDA Library by Doxygen

3.4 Device Management 9

3.4 Device Management

Functions
• cudaError_t cudaChooseDevice (int ∗device, const struct cudaDeviceProp ∗prop)
Select compute-device which best matches criteria.

• cudaError_t cudaGetDevice (int ∗device)

Returns which device is currently being used.

• cudaError_t cudaGetDeviceCount (int ∗count)

Returns the number of compute-capable devices.

• cudaError_t cudaGetDeviceProperties (struct cudaDeviceProp ∗prop, int device)

Returns information about the compute-device.

• cudaError_t cudaSetDevice (int device)

Set device to be used for GPU executions.

• cudaError_t cudaSetDeviceFlags (int flags)

Sets flags to be used for device executions.

• cudaError_t cudaSetValidDevices (int ∗device_arr, int len)

Set a list of devices that can be used for CUDA.

3.4.1 Detailed Description

This section describes the device management functions of the CUDA runtime application programming interface.

3.4.2 Function Documentation

3.4.2.1 cudaError_t cudaChooseDevice (int ∗ device, const struct cudaDeviceProp ∗ prop)

Returns in ∗device the device which has properties that best match ∗prop.

Parameters:
device - Device with best match
prop - Desired device properties

Returns:
cudaSuccess, cudaErrorInvalidValue

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaGetDeviceCount, cudaGetDevice, cudaSetDevice, cudaGetDeviceProperties

Generated for NVIDIA CUDA Library by Doxygen

10 Module Documentation

3.4.2.2 cudaError_t cudaGetDevice (int ∗ device)

Returns in ∗device the device on which the active host thread executes the device code.

Parameters:
device - Returns the device on which the active host thread executes the device code.

Returns:
cudaSuccess

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaGetDeviceCount, cudaSetDevice, cudaGetDeviceProperties, cudaChooseDevice

3.4.2.3 cudaError_t cudaGetDeviceCount (int ∗ count)

Returns in ∗count the number of devices with compute capability greater or equal to 1.0 that are available for
execution. If there is no such device, cudaGetDeviceCount() returns 1 and device 0 only supports device
emulation mode. Since this device will be able to emulate all hardware features, this device will report major and
minor compute capability versions of 9999.

Parameters:
count - Returns the number of devices with compute capability greater or equal to 1.0

Returns:
cudaSuccess

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaGetDevice, cudaSetDevice, cudaGetDeviceProperties, cudaChooseDevice

3.4.2.4 cudaError_t cudaGetDeviceProperties (struct cudaDeviceProp ∗ prop, int device)

Returns in ∗prop the properties of device dev. The cudaDeviceProp structure is defined as:

struct cudaDeviceProp {
char name[256];
size_t totalGlobalMem;
size_t sharedMemPerBlock;
int regsPerBlock;
int warpSize;
size_t memPitch;
int maxThreadsPerBlock;
int maxThreadsDim[3];
int maxGridSize[3];

Generated for NVIDIA CUDA Library by Doxygen

3.4 Device Management 11

size_t totalConstMem;
int major;
int minor;
int clockRate;
size_t textureAlignment;
int deviceOverlap;
int multiProcessorCount;
int kernelExecTimeoutEnabled;
int integrated;
int canMapHostMemory;
int computeMode;
}

where:

• name is an ASCII string identifying the device;

• totalGlobalMem is the total amount of global memory available on the device in bytes;

• sharedMemPerBlock is the maximum amount of shared memory available to a thread block in bytes; this
amount is shared by all thread blocks simultaneously resident on a multiprocessor;

• regsPerBlock is the maximum number of 32-bit registers available to a thread block; this number is shared
by all thread blocks simultaneously resident on a multiprocessor;

• warpSize is the warp size in threads;

• memPitch is the maximum pitch in bytes allowed by the memory copy functions that involve memory regions
allocated through cudaMallocPitch();

• maxThreadsPerBlock is the maximum number of threads per block;

• maxThreadsDim[3] contains the maximum size of each dimension of a block;

• maxGridSize[3] contains the maximum size of each dimension of a grid;

• clockRate is the clock frequency in kilohertz;

• totalConstMem is the total amount of constant memory available on the device in bytes;

• major, minor are the major and minor revision numbers defining the device’s compute capability;

• textureAlignment is the alignment requirement; texture base addresses that are aligned to
textureAlignment bytes do not need an offset applied to texture fetches;

• deviceOverlap is 1 if the device can concurrently copy memory between host and device while executing a
kernel, or 0 if not;

• multiProcessorCount is the number of multiprocessors on the device;

• kernelExecTimeoutEnabled is 1 if there is a run time limit for kernels executed on the device, or 0 if
not.

• integrated is 1 if the device is an integrated (motherboard) GPU and 0 if it is a discrete (card) component

• canMapHostMemory is 1 if the device can map host memory into the CUDA address space for use with
cudaHostAlloc()/cudaHostGetDevicePointer(), or 0 if not;

• computeMode is the compute mode that the device is currently in. Available modes are as follows:

– cudaComputeModeDefault: Default mode - Device is not restricted and multiple threads can use cudaSet-
Device() with this device.

Generated for NVIDIA CUDA Library by Doxygen

12 Module Documentation

– cudaComputeModeExclusive: Compute-exclusive mode - Only one thread will be able to use cudaSetDe-
vice() with this device.
– cudaComputeModeProhibited: Compute-prohibited mode - No threads can use cudaSetDevice() with this
device. Any errors from calling cudaSetDevice() with an exclusive (and occupied) or prohibited device will
only show up after a non-device management runtime function is called. At that time, cudaErrorNoDevice
will be returned.

Parameters:
prop - Properties for the specified device
device - Device number to get properties for

Returns:
cudaSuccess, cudaErrorInvalidDevice

See also:
cudaGetDeviceCount, cudaGetDevice, cudaSetDevice, cudaChooseDevice

3.4.2.5 cudaError_t cudaSetDevice (int device)

Records device as the device on which the active host thread executes the device code. If the host thread has already
initialized the CUDA runtime by calling non-device management runtime functions, this call returns cudaErrorSetOn-
ActiveProcess.

Parameters:
device - Device on which the active host thread should execute the device code.

Returns:
cudaSuccess, cudaErrorInvalidDevice, cudaErrorSetOnActiveProcess

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaGetDeviceCount, cudaGetDevice, cudaGetDeviceProperties, cudaChooseDevice

3.4.2.6 cudaError_t cudaSetDeviceFlags (int flags)

Records flags as the flags to use when the active host thread executes device code. If the host thread has already
initialized the CUDA runtime by calling non-device management runtime functions, this call returns cudaErrorSetOn-
ActiveProcess.
The two LSBs of the flags parameter can be used to control how the CPU thread interacts with the OS scheduler
when waiting for results from the device.

• cudaDeviceScheduleAuto: The default value if the flags parameter is zero, uses a heuristic based on the
number of active CUDA contexts in the process C and the number of logical processors in the system P. If C >
P, then CUDA will yield to other OS threads when waiting for the device, otherwise CUDA will not yield while
waiting for results and actively spin on the processor.

Generated for NVIDIA CUDA Library by Doxygen

3.4 Device Management 13

• cudaDeviceScheduleSpin: Instruct CUDA to actively spin when waiting for results from the device. This can
decrease latency when waiting for the device, but may lower the performance of CPU threads if they are per-
forming work in parallel with the CUDA thread.
• cudaDeviceScheduleYield: Instruct CUDA to yield its thread when waiting for results from the device. This
can increase latency when waiting for the device, but can increase the performance of CPU threads performing
work in parallel with the device.
• cudaDeviceBlockingSync: Instruct CUDA to block the CPU thread on a synchronization primitive when waiting
for the device to finish work.
• cudaDeviceMapHost: This flag must be set in order to allocate pinned host memory that is accessible to the
device. If this flag is not set, cudaHostGetDevicePointer() will always return a failure code.

Parameters:
flags - Parameters for device operation

Returns:
cudaSuccess, cudaErrorInvalidDevice, cudaErrorSetOnActiveProcess

See also:
cudaGetDeviceCount, cudaGetDevice, cudaGetDeviceProperties, cudaSetDevice, cudaSetValidDevices, cuda-
ChooseDevice

3.4.2.7 cudaError_t cudaSetValidDevices (int ∗ device_arr, int len)

Sets a list of devices for CUDA execution in priority order using device_arr. The parameter len specifies the
number of elements in the list. CUDA will try devices from the list sequentially until it finds one that works. If this
function is not called, or if it is called with a len of 0, then CUDA will go back to its default behavior of trying devices
sequentially from a default list containing all of the available CUDA devices in the system. If a specified device ID in
the list does not exist, this function will return cudaErrorInvalidDevice. If len is not 0 and device_arr is NULL
or if len is greater than the number of devices in the system, then cudaErrorInvalidValue is returned.

Parameters:
device_arr - List of devices to try
len - Number of devices in specified list

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidDevice

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaGetDeviceCount, cudaSetDevice, cudaGetDeviceProperties, cudaSetDeviceFlags, cudaChooseDevice

Generated for NVIDIA CUDA Library by Doxygen

14 Module Documentation

3.5 Stream Management

Functions
• cudaError_t cudaStreamCreate (cudaStream_t ∗pStream)
Create an asynchronous stream.

• cudaError_t cudaStreamDestroy (cudaStream_t stream)

Destroys and cleans up an asynchronous stream.

• cudaError_t cudaStreamQuery (cudaStream_t stream)

Queries an asynchronous stream for completion status.

• cudaError_t cudaStreamSynchronize (cudaStream_t stream)

Waits for stream tasks to complete.

3.5.1 Detailed Description

This section describes the stream management functions of the CUDA runtime application programming interface.

3.5.2 Function Documentation

3.5.2.1 cudaError_t cudaStreamCreate (cudaStream_t ∗ pStream)

Creates a new asynchronous stream.

Parameters:
pStream - Pointer to new stream identifier

Returns:
cudaSuccess, cudaErrorInvalidValue

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaStreamQuery, cudaStreamSynchronize, cudaStreamDestroy

3.5.2.2 cudaError_t cudaStreamDestroy (cudaStream_t stream)

Destroys and cleans up the asynchronous stream specified by stream.

Parameters:
stream - Stream identifier

Generated for NVIDIA CUDA Library by Doxygen

3.5 Stream Management 15

Returns:
cudaSuccess, cudaErrorInvalidResourceHandle

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaStreamCreate, cudaStreamQuery, cudaStreamSynchronize

3.5.2.3 cudaError_t cudaStreamQuery (cudaStream_t stream)

Returns cudaSuccess if all operations in stream have completed, or cudaErrorNotReady if not.

Parameters:
stream - Stream identifier

Returns:
cudaSuccess, cudaErrorNotReady cudaErrorInvalidResourceHandle

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaStreamCreate, cudaStreamSynchronize, cudaStreamDestroy

3.5.2.4 cudaError_t cudaStreamSynchronize (cudaStream_t stream)

Blocks until stream has completed all operations.

Parameters:
stream - Stream identifier

Returns:
cudaSuccess, cudaErrorInvalidResourceHandle

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaStreamCreate, cudaStreamQuery, cudaStreamDestroy

Generated for NVIDIA CUDA Library by Doxygen

16 Module Documentation

3.6 Event Management

Functions
• cudaError_t cudaEventCreate (cudaEvent_t ∗event)
Creates an event object.

• cudaError_t cudaEventCreateWithFlags (cudaEvent_t ∗event, int flags)

Creates an event object with the specified flags.

• cudaError_t cudaEventDestroy (cudaEvent_t event)

Destroys an event object.

• cudaError_t cudaEventElapsedTime (float ∗ms, cudaEvent_t start, cudaEvent_t end)

Computes the elapsed time between events.

• cudaError_t cudaEventQuery (cudaEvent_t event)

Query if an event has been recorded.

• cudaError_t cudaEventRecord (cudaEvent_t event, cudaStream_t stream)

Records an event.

• cudaError_t cudaEventSynchronize (cudaEvent_t event)

Wait for an event to be recorded.

3.6.1 Detailed Description

This section describes the event management functions of the CUDA runtime application programming interface.

3.6.2 Function Documentation

3.6.2.1 cudaError_t cudaEventCreate (cudaEvent_t ∗ event)

Creates an event object using cudaEventDefault.

Parameters:
event - Newly created event

Returns:
cudaSuccess, cudaErrorInitializationError, cudaErrorPriorLaunchFailure, cudaErrorInvalidValue, cudaEr-
rorMemoryAllocation

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaEventCreateWithFlags, cudaEventRecord, cudaEventQuery, cudaEventSynchronize, cudaEventDestroy, cu-
daEventElapsedTime

Generated for NVIDIA CUDA Library by Doxygen

3.6 Event Management 17

3.6.2.2 cudaError_t cudaEventCreateWithFlags (cudaEvent_t ∗ event, int flags)

Creates an event object with the specified flags.

Valid flags include: cudaEventDefault, cudaEventBlockingSync

Parameters:
event - Newly created event
flags - Flags for new event

Returns:
cudaSuccess, cudaErrorInitializationError, cudaErrorPriorLaunchFailure, cudaErrorInvalidValue, cudaEr-
rorMemoryAllocation

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaEventCreate, cudaEventRecord, cudaEventQuery, cudaEventSynchronize, cudaEventDestroy, cudaEventE-
lapsedTime

3.6.2.3 cudaError_t cudaEventDestroy (cudaEvent_t event)

Destroys the specified event object.

Parameters:
event - Event to destroy

Returns:
cudaSuccess, cudaErrorInitializationError, cudaErrorPriorLaunchFailure, cudaErrorInvalidValue

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaEventCreate, cudaEventCreateWithFlags, cudaEventQuery, cudaEventSynchronize, cudaEventRecord, cud-
aEventElapsedTime

3.6.2.4 cudaError_t cudaEventElapsedTime (float ∗ ms, cudaEvent_t start, cudaEvent_t end)

Computes the elapsed time between two events (in milliseconds with a resolution of around 0.5 microseconds). If
either event has not been recorded yet, this function returns cudaErrorInvalidValue. If either event has been recorded
with a non-zero stream, the result is undefined.

Parameters:
ms - Time between start and stop in ms

Generated for NVIDIA CUDA Library by Doxygen

18 Module Documentation

start - Starting event

end - Stopping event

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInitializationError, cudaErrorPriorLaunchFailure, cudaErrorIn-
validResourceHandle

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaEventCreate, cudaEventCreateWithFlags, cudaEventQuery, cudaEventSynchronize, cudaEventDestroy, cu-
daEventRecord

3.6.2.5 cudaError_t cudaEventQuery (cudaEvent_t event)

Returns cudaSuccess if the event has actually been recorded, or cudaErrorNotReady if not. If cudaEventRecord() has
not been called on this event, the function returns cudaErrorInvalidValue.

Parameters:
event - Event to query

Returns:
cudaSuccess, cudaErrorNotReady, cudaErrorInitializationError, cudaErrorPriorLaunchFailure, cudaErrorInvalid-
Value cudaErrorInvalidResourceHandle

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaEventCreate, cudaEventCreateWithFlags, cudaEventRecord, cudaEventSynchronize, cudaEventDestroy, cu-
daEventElapsedTime

3.6.2.6 cudaError_t cudaEventRecord (cudaEvent_t event, cudaStream_t stream)

Records an event. If stream is non-zero, the event is recorded after all preceding operations in the stream have been
completed; otherwise, it is recorded after all preceding operations in the CUDA context have been completed. Since
this operation is asynchronous, cudaEventQuery() and/or cudaEventSynchronize() must be used to determine when
the event has actually been recorded.
If cudaEventRecord() has previously been called and the event has not been recorded yet, this function returns cud-
aErrorInvalidValue.

Parameters:
event - Event to record
stream - Stream in which to record event

Generated for NVIDIA CUDA Library by Doxygen

3.6 Event Management 19

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInitializationError, cudaErrorPriorLaunchFailure, cudaErrorIn-
validResourceHandle

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaEventCreate, cudaEventCreateWithFlags, cudaEventQuery, cudaEventSynchronize, cudaEventDestroy, cu-
daEventElapsedTime

3.6.2.7 cudaError_t cudaEventSynchronize (cudaEvent_t event)

Blocks until the event has actually been recorded. If cudaEventRecord() has not been called on this event, the function
returns cudaErrorInvalidValue.

Parameters:
event - Event to wait for

Returns:
cudaSuccess, cudaErrorInitializationError, cudaErrorPriorLaunchFailure, cudaErrorInvalidValue cudaErrorIn-
validResourceHandle

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaEventCreate, cudaEventCreateWithFlags, cudaEventRecord, cudaEventQuery, cudaEventDestroy, cud-
aEventElapsedTime

Generated for NVIDIA CUDA Library by Doxygen

20 Module Documentation

3.7 Execution Control

Functions
• cudaError_t cudaConfigureCall (dim3 gridDim, dim3 blockDim, size_t sharedMem, cudaStream_t stream)
Configure a device-launch.

• cudaError_t cudaFuncGetAttributes (struct cudaFuncAttributes ∗attr, const char ∗func)

Find out attributes for a given function.

• cudaError_t cudaLaunch (const char ∗entry)

Launches a device function.

• cudaError_t cudaSetDoubleForDevice (double ∗d)

Converts a double argument to be executed on a device.

• cudaError_t cudaSetDoubleForHost (double ∗d)

Converts a double argument after execution on a device.

• cudaError_t cudaSetupArgument (const void ∗arg, size_t size, size_t offset)

Configure a device launch.

3.7.1 Detailed Description

This section describes the execution control functions of the CUDA runtime application programming interface.

3.7.2 Function Documentation

3.7.2.1 cudaError_t cudaConfigureCall (dim3 gridDim, dim3 blockDim, size_t sharedMem, cudaStream_t
stream)

Specifies the grid and block dimensions for the device call to be executed similar to the execution configuration
syntax. cudaConfigureCall() is stack based. Each call pushes data on top of an execution stack. This data contains the
dimension for the grid and thread blocks, together with any arguments for the call.

Parameters:
gridDim - Grid dimensions
blockDim - Block dimensions
sharedMem - Shared memory
stream - Stream identifier

Returns:
cudaSuccess, cudaErrorInvalidConfiguration

Note:
Note that this function may also return error codes from previous, asynchronous launches.

Generated for NVIDIA CUDA Library by Doxygen

3.7 Execution Control 21

See also:
cudaFuncGetAttributes (C API), cudaLaunch (C API), cudaSetDoubleForDevice, cudaSetDoubleForHost, cud-
aSetupArgument (C API),

3.7.2.2 cudaError_t cudaFuncGetAttributes (struct cudaFuncAttributes ∗ attr, const char ∗ func)

This function obtains the attributes of a function specified via func. The fetched attributes are placed in attr. If the
specified function does not exist, then cudaErrorInvalidDeviceFunction is returned.

Parameters:
attr - Return pointer to function’s attributes
func - Function to get attributes of

Returns:
cudaSuccess, cudaErrorInitializationError, cudaErrorInvalidDeviceFunction

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaConfigureCall, cudaFuncGetAttributes (C++ API), cudaLaunch (C API), cudaSetDoubleForDevice, cudaSet-
DoubleForHost, cudaSetupArgument (C API)

3.7.2.3 cudaError_t cudaLaunch (const char ∗ entry)

Launches the function entry on the device. The parameter entry must be a character string naming a function
that executes on the device. The parameter specified by entry must be declared as a __global__ function. cud-
aLaunch() must be preceded by a call to cudaConfigureCall() since it pops the data that was pushed by cudaConfig-
ureCall() from the execution stack.

Parameters:
entry - Device char string naming device function to execute

Returns:
cudaSuccess, cudaErrorInvalidDeviceFunction, cudaErrorInvalidConfiguration, cudaErrorLaunchFailure, cud-
aErrorPriorLaunchFailure, cudaErrorLaunchTimeout, cudaErrorLaunchOutOfResources

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaConfigureCall, cudaFuncGetAttributes (C API), cudaLaunch (C++ API), cudaSetDoubleForDevice, cudaSet-
DoubleForHost, cudaSetupArgument (C API)

Generated for NVIDIA CUDA Library by Doxygen

22 Module Documentation

3.7.2.4 cudaError_t cudaSetDoubleForDevice (double ∗ d)

Parameters:
d - Double to convert

Converts the double value of d to an internal float representation if the device does not support double arithmetic. If
the device does natively support doubles, then this function does nothing.

Returns:
cudaSuccess

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaConfigureCall, cudaFuncGetAttributes (C API), cudaLaunch (C API), cudaSetDoubleForHost, cudaSetupAr-
gument (C API)

3.7.2.5 cudaError_t cudaSetDoubleForHost (double ∗ d)

Converts the double value of d from a potentially internal float representation if the device does not support double
arithmetic. If the device does natively support doubles, then this function does nothing.

Parameters:
d - Double to convert

Returns:
cudaSuccess

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaConfigureCall, cudaFuncGetAttributes (C API), cudaLaunch (C API), cudaSetDoubleForDevice, cudaSetu-
pArgument (C API)

3.7.2.6 cudaError_t cudaSetupArgument (const void ∗ arg, size_t size, size_t offset)

Pushes size bytes of the argument pointed to by arg at offset bytes from the start of the parameter passing area,
which starts at offset 0. The arguments are stored in the top of the execution stack. cudaSetupArgument() must be
preceded by a call to cudaConfigureCall().

Parameters:
arg - Argument to push for a kernel launch
size - Size of argument
offset - Offset in argument stack to push new arg

Generated for NVIDIA CUDA Library by Doxygen

3.7 Execution Control 23

Returns:
cudaSuccess

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaConfigureCall, cudaFuncGetAttributes (C API), cudaLaunch (C API), cudaSetDoubleForDevice, cudaSet-
DoubleForHost, cudaSetupArgument (C++ API),

Generated for NVIDIA CUDA Library by Doxygen

24 Module Documentation

3.8 Memory Management

Functions
• cudaError_t cudaFree (void ∗devPtr)
Frees memory on the device.

• cudaError_t cudaFreeArray (struct cudaArray ∗array)

Frees an array on the device.

• cudaError_t cudaFreeHost (void ∗ptr)

Frees page-locked memory.

• cudaError_t cudaGetSymbolAddress (void ∗∗devPtr, const char ∗symbol)

Finds the address associated with a CUDA symbol.

• cudaError_t cudaGetSymbolSize (size_t ∗size, const char ∗symbol)

Finds the size of the object associated with a CUDA symbol.

• cudaError_t cudaHostAlloc (void ∗∗ptr, size_t size, unsigned int flags)

Allocates page-locked memory on the host.

• cudaError_t cudaHostGetDevicePointer (void ∗∗pDevice, void ∗pHost, unsigned int flags)

Passes back device pointer of mapped host memory allocated by cudaHostAlloc().

• cudaError_t cudaHostGetFlags (unsigned int ∗pFlags, void ∗pHost)

Passes back flags used to allocate pinned host memory allocated by cudaHostAlloc().

• cudaError_t cudaMalloc (void ∗∗devPtr, size_t size)

Allocate memory on the device.

• cudaError_t cudaMalloc3D (struct cudaPitchedPtr ∗pitchedDevPtr, struct cudaExtent extent)

Allocates logical 1D, 2D, or 3D memory objects on the device.

• cudaError_t cudaMalloc3DArray (struct cudaArray ∗∗arrayPtr, const struct cudaChannelFormatDesc ∗desc,

struct cudaExtent extent)
Allocate an array on the device.

• cudaError_t cudaMallocArray (struct cudaArray ∗∗arrayPtr, const struct cudaChannelFormatDesc ∗desc, size_t
width, size_t height)
Allocate an array on the device.

• cudaError_t cudaMallocHost (void ∗∗ptr, size_t size)

Allocates page-locked memory on the host.

• cudaError_t cudaMallocPitch (void ∗∗devPtr, size_t ∗pitch, size_t width, size_t height)
Allocates pitched memory on the device.

• cudaError_t cudaMemcpy (void ∗dst, const void ∗src, size_t count, enum cudaMemcpyKind kind)
Copies data between host and device.

Generated for NVIDIA CUDA Library by Doxygen

3.8 Memory Management 25

• cudaError_t cudaMemcpy2D (void ∗dst, size_t dpitch, const void ∗src, size_t spitch, size_t width, size_t height,
enum cudaMemcpyKind kind)
Copies data between host and device.

• cudaError_t cudaMemcpy2DArrayToArray (struct cudaArray ∗dst, size_t wOffsetDst, size_t hOffsetDst, const
struct cudaArray ∗src, size_t wOffsetSrc, size_t hOffsetSrc, size_t width, size_t height, enum cudaMemcpyKind
kind)
Copies data between host and device.

• cudaError_t cudaMemcpy2DAsync (void ∗dst, size_t dpitch, const void ∗src, size_t spitch, size_t width, size_t
height, enum cudaMemcpyKind kind, cudaStream_t stream)
Copies data between host and device.

• cudaError_t cudaMemcpy2DFromArray (void ∗dst, size_t dpitch, const struct cudaArray ∗src, size_t wOffset,
size_t hOffset, size_t width, size_t height, enum cudaMemcpyKind kind)
Copies data between host and device.

• cudaError_t cudaMemcpy2DFromArrayAsync (void ∗dst, size_t dpitch, const struct cudaArray ∗src, size_t
wOffset, size_t hOffset, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream)
Copies data between host and device.

• cudaError_t cudaMemcpy2DToArray (struct cudaArray ∗dst, size_t wOffset, size_t hOffset, const void ∗src,
size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind)
Copies data between host and device.

• cudaError_t cudaMemcpy2DToArrayAsync (struct cudaArray ∗dst, size_t wOffset, size_t hOffset, const void
∗src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream)
Copies data between host and device.

• cudaError_t cudaMemcpy3D (const struct cudaMemcpy3DParms ∗p)

Copies data between 3D objects.

• cudaError_t cudaMemcpy3DAsync (const struct cudaMemcpy3DParms ∗p, cudaStream_t stream)

Copies data between 3D objects.

• cudaError_t cudaMemcpyArrayToArray (struct cudaArray ∗dst, size_t wOffsetDst, size_t hOffsetDst, const
struct cudaArray ∗src, size_t wOffsetSrc, size_t hOffsetSrc, size_t count, enum cudaMemcpyKind kind)
Copies data between host and device.

• cudaError_t cudaMemcpyAsync (void ∗dst, const void ∗src, size_t count, enum cudaMemcpyKind kind,
cudaStream_t stream)
Copies data between host and device.

• cudaError_t cudaMemcpyFromArray (void ∗dst, const struct cudaArray ∗src, size_t wOffset, size_t hOffset,
size_t count, enum cudaMemcpyKind kind)
Copies data between host and device.

• cudaError_t cudaMemcpyFromArrayAsync (void ∗dst, const struct cudaArray ∗src, size_t wOffset, size_t hOff-
set, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream)
Copies data between host and device.

Generated for NVIDIA CUDA Library by Doxygen

26 Module Documentation

• cudaError_t cudaMemcpyFromSymbol (void ∗dst, const char ∗symbol, size_t count, size_t offset, enum cud-
aMemcpyKind kind)
Copies data from the given symbol on the device.

• cudaError_t cudaMemcpyFromSymbolAsync (void ∗dst, const char ∗symbol, size_t count, size_t offset, enum
cudaMemcpyKind kind, cudaStream_t stream)
Copies data from the given symbol on the device.

• cudaError_t cudaMemcpyToArray (struct cudaArray ∗dst, size_t wOffset, size_t hOffset, const void ∗src, size_t
count, enum cudaMemcpyKind kind)
Copies data between host and device.

• cudaError_t cudaMemcpyToArrayAsync (struct cudaArray ∗dst, size_t wOffset, size_t hOffset, const void ∗src,
size_t count, enum cudaMemcpyKind kind, cudaStream_t stream)
Copies data between host and device.

• cudaError_t cudaMemcpyToSymbol (const char ∗symbol, const void ∗src, size_t count, size_t offset, enum
cudaMemcpyKind kind)
Copies data to the given symbol on the device.

• cudaError_t cudaMemcpyToSymbolAsync (const char ∗symbol, const void ∗src, size_t count, size_t offset,
enum cudaMemcpyKind kind, cudaStream_t stream)
Copies data to the given symbol on the device.

• cudaError_t cudaMemset (void ∗devPtr, int value, size_t count)

Initializes or sets device memory to a value.

• cudaError_t cudaMemset2D (void ∗devPtr, size_t pitch, int value, size_t width, size_t height)
Initializes or sets device memory to a value.

• cudaError_t cudaMemset3D (struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent)
Initializes or sets device memory to a value.

3.8.1 Detailed Description

This section describes the memory management functions of the CUDA runtime application programming interface.

3.8.2 Function Documentation

3.8.2.1 cudaError_t cudaFree (void ∗ devPtr)

Frees the memory space pointed to by devPtr, which must have been returned by a previous call to cudaMalloc()
or cudaMallocPitch(). Otherwise, or if cudaFree(devPtr) has already been called before, an error is returned. If
devPtr is 0, no operation is performed. cudaFree() returns cudaErrorInvalidDevicePointer in case of failure.

Parameters:
devPtr - Device pointer to memory to free

Generated for NVIDIA CUDA Library by Doxygen

3.8 Memory Management 27

Returns:
cudaSuccess, cudaErrorInvalidDevicePointer, cudaErrorInitializationError

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaMalloc, cudaMallocPitch, cudaMallocArray, cudaFreeArray, cudaMallocHost, cudaFreeHost, cudaMal-
loc3D, cudaMalloc3DArray, cudaHostAlloc

3.8.2.2 cudaError_t cudaFreeArray (struct cudaArray ∗ array)

Frees the CUDA array array, which must have been ∗ returned by a previous call to cudaMallocArray(). If cud-
aFreeArray(array) has already been called before, cudaErrorInvalidValue is returned. If devPtr is 0, no operation
is performed.

Parameters:
array - Pointer to array to free

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInitializationError

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaMalloc, cudaMallocPitch, cudaFree, cudaMallocArray, cudaMallocHost, cudaFreeHost, cudaHostAlloc

3.8.2.3 cudaError_t cudaFreeHost (void ∗ ptr)

Frees the memory space pointed to by hostPtr, which must have been returned by a previous call to cudaMallo-
cHost() or cudaHostAlloc().

Parameters:
ptr - Pointer to memory to free

Returns:
cudaSuccess, cudaErrorInitializationError

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaMalloc, cudaMallocPitch, cudaFree, cudaMallocArray, cudaFreeArray, cudaMallocHost, cudaMalloc3D, cu-
daMalloc3DArray, cudaHostAlloc

Generated for NVIDIA CUDA Library by Doxygen

28 Module Documentation

3.8.2.4 cudaError_t cudaGetSymbolAddress (void ∗∗ devPtr, const char ∗ symbol)

Returns in ∗devPtr the address of symbol symbol on the device. symbol can either be a variable that resides
in global memory space, or it can be a character string, naming a variable that resides in global memory space. If
symbol cannot be found, or if symbol is not declared in the global memory space, ∗devPtr is unchanged and the
error cudaErrorInvalidSymbol is returned.

Parameters:
devPtr - Return device pointer associated with symbol
symbol - Global variable or string symbol to search for

Returns:
cudaSuccess, cudaErrorInvalidSymbol, cudaErrorAddressOfConstant

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaGetSymbolAddress (C++ API) cudaGetSymbolSize (C API)

3.8.2.5 cudaError_t cudaGetSymbolSize (size_t ∗ size, const char ∗ symbol)

Returns in ∗size the size of symbol symbol. symbol can either be a variable that resides in global or constant
memory space, or it can be a character string, naming a variable that resides in global or constant memory space. If
symbol cannot be found, or if symbol is not declared in global or constant memory space, ∗size is unchanged
and the error cudaErrorInvalidSymbol is returned.

Parameters:
size - Size of object associated with symbol
symbol - Global variable or string symbol to find size of

Returns:
cudaSuccess, cudaErrorInvalidSymbol

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaGetSymbolAddress (C API) cudaGetSymbolSize (C++ API)

3.8.2.6 cudaError_t cudaHostAlloc (void ∗∗ ptr, size_t size, unsigned int flags)

Allocates count bytes of host memory that is page-locked and accessible to the device. The driver tracks the virtual
memory ranges allocated with this function and automatically accelerates calls to functions such as cudaMemcpy().
Since the memory can be accessed directly by the device, it can be read or written with much higher bandwidth than
pageable memory obtained with functions such as malloc(). Allocating excessive amounts of pinned memory may

Generated for NVIDIA CUDA Library by Doxygen

3.8 Memory Management 29

degrade system performance, since it reduces the amount of memory available to the system for paging. As a result,
this function is best used sparingly to allocate staging areas for data exchange between host and device.
The flags parameter enables different options to be specified that affect the allocation, as follows.

• cudaHostAllocDefault: This flag’s value is defined to be 0 and causes cudaHostAlloc() to emulate cudaMallo-
cHost().

• cudaHostAllocPortable: The memory returned by this call will be considered as pinned memory by all CUDA
contexts, not just the one that performed the allocation.

• cudaHostAllocMapped: Maps the allocation into the CUDA address space. The device pointer to the memory
may be obtained by calling cudaHostGetDevicePointer().

• cudaHostAllocWriteCombined: Allocates the memory as write-combined (WC). WC memory can be trans-

ferred across the PCI Express bus more quickly on some system configurations, but cannot be read efficiently
by most CPUs. WC memory is a good option for buffers that will be written by the CPU and read by the device
via mapped pinned memory or host->device transfers.

All of these flags are orthogonal to one another: a developer may allocate memory that is portable, mapped and/or
write-combined with no restrictions.
cudaSetDeviceFlags() must have been called with the cudaDeviceMapHost flag in order for the cudaHostAllocMapped
flag to have any effect.
The cudaHostAllocMapped flag may be specified on CUDA contexts for devices that do not support mapped pinned
memory. The failure is deferred to cudaHostGetDevicePointer() because the memory may be mapped into other
CUDA contexts via the cudaHostAllocPortable flag.
Memory allocated by this function must be freed with cudaFreeHost().

Parameters:
ptr - Device pointer to allocated memory
size - Requested allocation size in bytes
flags - Requested properties of allocated memory

Returns:
cudaSuccess, cudaErrorMemoryAllocation

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaSetDeviceFlags, cudaMallocHost, cudaFreeHost

3.8.2.7 cudaError_t cudaHostGetDevicePointer (void ∗∗ pDevice, void ∗ pHost, unsigned int flags)

Passes back the device pointer corresponding to the mapped, pinned host buffer allocated by cudaHostAlloc().
cudaHostGetDevicePointer() will fail if the cudaDeviceMapHost flag was not specified before deferred context cre-
ation occurred, or if called on a device that does not support mapped, pinned memory.
flags provides for future releases. For now, it must be set to 0.

Generated for NVIDIA CUDA Library by Doxygen

30 Module Documentation

Parameters:
pDevice - Returned device pointer for mapped memory
pHost - Requested host pointer mapping
flags - Flags for extensions (must be 0 for now)

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorMemoryAllocation

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaSetDeviceFlags, cudaHostAlloc

3.8.2.8 cudaError_t cudaHostGetFlags (unsigned int ∗ pFlags, void ∗ pHost)

cudaHostGetFlags() will fail if the input pointer does not reside in an address range allocated by cudaHostAlloc().

Parameters:
pFlags - Returned flags word
pHost - Host pointer

Returns:
cudaSuccess, cudaErrorInvalidValue

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaHostAlloc

3.8.2.9 cudaError_t cudaMalloc (void ∗∗ devPtr, size_t size)

Allocates size bytes of linear memory on the device and returns in ∗devPtr a pointer to the allocated memory.
The allocated memory is suitably aligned for any kind of variable. The memory is not cleared. cudaMalloc() returns
cudaErrorMemoryAllocation in case of failure.

Parameters:
devPtr - Pointer to allocated device memory
size - Requested allocation size in bytes

Returns:
cudaSuccess, cudaErrorMemoryAllocation

See also:
cudaMallocPitch, cudaFree, cudaMallocArray, cudaFreeArray, cudaMalloc3D, cudaMalloc3DArray, cudaMallo-
cHost, cudaFreeHost, cudaHostAlloc

Generated for NVIDIA CUDA Library by Doxygen

3.8 Memory Management 31

3.8.2.10 cudaError_t cudaMalloc3D (struct cudaPitchedPtr ∗ pitchedDevPtr, struct cudaExtent extent)

Allocates at least width ∗ height ∗ depth bytes of linear memory on the device and returns a cudaPitchedPtr in
which ptr is a pointer to the allocated memory. The function may pad the allocation to ensure hardware alignment
requirements are met. The pitch returned in the pitch field of pitchedDevPtr is the width in bytes of the
allocation.
The returned cudaPitchedPtr contains additional fields xsize and ysize, the logical width and height of the allo-
cation, which are equivalent to the width and height extent parameters provided by the programmer during
allocation.
For allocations of 2D and 3D objects, it is highly recommended that programmers perform allocations using cudaMal-
loc3D() or cudaMallocPitch(). Due to alignment restrictions in the hardware, this is especially true if the application
will be performing memory copies involving 2D or 3D objects (whether linear memory or CUDA arrays).

Parameters:
pitchedDevPtr - Pointer to allocated pitched device memory
extent - Requested allocation size

Returns:
cudaSuccess, cudaErrorMemoryAllocation

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaMallocPitch, cudaFree, cudaMemcpy3D, cudaMemset3D, cudaMalloc3DArray, cudaMallocArray, cud-
aFreeArray, cudaMallocHost, cudaFreeHost, cudaHostAlloc, make_cudaPitchedPtr, make_cudaExtent

3.8.2.11 cudaError_t cudaMalloc3DArray (struct cudaArray ∗∗ arrayPtr, const struct

cudaChannelFormatDesc ∗ desc, struct cudaExtent extent)

Allocates a CUDA array according to the cudaChannelFormatDesc structure desc and returns a handle to the new
CUDA array in ∗arrayPtr.
The cudaChannelFormatDesc is defined as:

struct cudaChannelFormatDesc {
int x, y, z, w;
enum cudaChannelFormatKind f;
};

where cudaChannelFormatKind is one of cudaChannelFormatKindSigned, cudaChannelFormatKindUnsigned, or cu-

daChannelFormatKindFloat.
cudaMalloc3DArray() is able to allocate 1D, 2D, or 3D arrays.

• A 1D array is allocated if the height and depth extent are both zero. For 1D arrays valid extent ranges are {(1,
8192), 0, 0}.

• A 2D array is allocated if only the depth extent is zero. For 2D arrays valid extent ranges are {(1, 65536), (1,
32768), 0}.

Generated for NVIDIA CUDA Library by Doxygen

32 Module Documentation

• A 3D array is allocated if all three extents are non-zero. For 3D arrays valid extent ranges are {(1, 2048), (1,
2048), (1, 2048)}.

Note:
Due to the differing extent limits, it may be advantageous to use a degenerate array (with unused dimensions set
to one) of higher dimensionality. For instance, a degenerate 2D array allows for significantly more linear storage
than a 1D array.

Parameters:
arrayPtr - Pointer to allocated array in device memory
desc - Requested channel format
extent - Requested allocation size

Returns:
cudaSuccess, cudaErrorMemoryAllocation

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaMalloc3D, cudaMalloc, cudaMallocPitch, cudaFree, cudaFreeArray, cudaMallocHost, cudaFreeHost, cuda-
HostAlloc, make_cudaExtent

3.8.2.12 cudaError_t cudaMallocArray (struct cudaArray ∗∗ arrayPtr, const struct cudaChannelFormatDesc

∗ desc, size_t width, size_t height)

Allocates a CUDA array according to the cudaChannelFormatDesc structure desc and returns a handle to the new
CUDA array in ∗array.
The cudaChannelFormatDesc is defined as:

struct cudaChannelFormatDesc {
int x, y, z, w;
enum cudaChannelFormatKind f;
};

where cudaChannelFormatKind is one of cudaChannelFormatKindSigned, cudaChannelFormatKindUnsigned, or cu-

daChannelFormatKindFloat.

Parameters:
arrayPtr - Pointer to allocated array in device memory
desc - Requested channel format
width - Requested array allocation width
height - Requested array allocation height

Returns:
cudaSuccess, cudaErrorMemoryAllocation

Generated for NVIDIA CUDA Library by Doxygen

3.8 Memory Management 33

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaMalloc, cudaMallocPitch, cudaFree, cudaFreeArray, cudaMallocHost, cudaFreeHost, cudaMalloc3D, cud-
aMalloc3DArray, cudaHostAlloc

3.8.2.13 cudaError_t cudaMallocHost (void ∗∗ ptr, size_t size)

Allocates size bytes of host memory that is page-locked and accessible to the device. The driver tracks the virtual
memory ranges allocated with this function and automatically accelerates calls to functions such as cudaMemcpy∗().
Since the memory can be accessed directly by the device, it can be read or written with much higher bandwidth
than pageable memory obtained with functions such as malloc(). Allocating excessive amounts of memory with
cudaMallocHost() may degrade system performance, since it reduces the amount of memory available to the system
for paging. As a result, this function is best used sparingly to allocate staging areas for data exchange between host
and device.

Parameters:
ptr - Pointer to allocated host memory
size - Requested allocation size in bytes

Returns:
cudaSuccess, cudaErrorMemoryAllocation

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaMalloc, cudaMallocPitch, cudaMallocArray, cudaMalloc3D, cudaMalloc3DArray, cudaHostAlloc, cud-
aFree, cudaFreeArray, cudaFreeHost, cudaHostAlloc

3.8.2.14 cudaError_t cudaMallocPitch (void ∗∗ devPtr, size_t ∗ pitch, size_t width, size_t height)

Allocates at least widthInBytes ∗ height bytes of linear memory on the device and returns in ∗devPtr a pointer
to the allocated memory. The function may pad the allocation to ensure that corresponding pointers in any given row
will continue to meet the alignment requirements for coalescing as the address is updated from row to row. The pitch
returned in ∗pitch by cudaMallocPitch() is the width in bytes of the allocation. The intended usage of pitch is as
a separate parameter of the allocation, used to compute addresses within the 2D array. Given the row and column of
an array element of type T, the address is computed as:

T* pElement = (T*)((char*)BaseAddress + Row * pitch) + Column;

For allocations of 2D arrays, it is recommended that programmers consider performing pitch allocations using cud-
aMallocPitch(). Due to pitch alignment restrictions in the hardware, this is especially true if the application will be
performing 2D memory copies between different regions of device memory (whether linear memory or CUDA arrays).

Parameters:
devPtr - Pointer to allocated pitched device memory

Generated for NVIDIA CUDA Library by Doxygen

34 Module Documentation

pitch - Pitch for allocation

width - Requested pitched allocation width
height - Requested pitched allocation height

Returns:
cudaSuccess, cudaErrorMemoryAllocation

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaMalloc, cudaFree, cudaMallocArray, cudaFreeArray, cudaMallocHost, cudaFreeHost, cudaMalloc3D, cud-
aMalloc3DArray, cudaHostAlloc

3.8.2.15 cudaError_t cudaMemcpy (void ∗ dst, const void ∗ src, size_t count, enum cudaMemcpyKind kind)

Copies count bytes from the memory area pointed to by src to the memory area pointed to by dst, where kind is
one of cudaMemcpyHostToHost, cudaMemcpyHostToDevice, cudaMemcpyDeviceToHost, or cudaMemcpyDevice-
ToDevice, and specifies the direction of the copy. The memory areas may not overlap. Calling cudaMemcpy() with
dst and src pointers that do not match the direction of the copy results in an undefined behavior.

Parameters:
dst - Destination memory address
src - Source memory address
count - Size in bytes to copy
kind - Type of transfer

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidDevicePointer, cudaErrorInvalidMemcpyDirection

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaMemcpy2D, cudaMemcpyToArray, cudaMemcpy2DToArray, cudaMemcpyFromArray, cudaMem-
cpy2DFromArray, cudaMemcpyArrayToArray, cudaMemcpy2DArrayToArray, cudaMemcpyToSymbol,
cudaMemcpyFromSymbol, cudaMemcpyAsync, cudaMemcpy2DAsync, cudaMemcpyToArrayAsync, cud-
aMemcpy2DToArrayAsync, cudaMemcpyFromArrayAsync, cudaMemcpy2DFromArrayAsync, cudaMemcpy-
ToSymbolAsync, cudaMemcpyFromSymbolAsync

3.8.2.16 cudaError_t cudaMemcpy2D (void ∗ dst, size_t dpitch, const void ∗ src, size_t spitch, size_t width,
size_t height, enum cudaMemcpyKind kind)

Copies a matrix (height rows of width bytes each) from the memory area pointed to by src to the memory area
pointed to by dst, where kind is one of cudaMemcpyHostToHost, cudaMemcpyHostToDevice, cudaMemcpyDe-
viceToHost, or cudaMemcpyDeviceToDevice, and specifies the direction of the copy. dpitch and spitch are the

Generated for NVIDIA CUDA Library by Doxygen

3.8 Memory Management 35

widths in memory in bytes of the 2D arrays pointed to by dst and src, including any padding added to the end of
each row. The memory areas may not overlap. Calling cudaMemcpy2D() with dst and src pointers that do not
match the direction of the copy results in an undefined behavior. cudaMemcpy2D() returns an error if dpitch or
spitch is greater than the maximum allowed.

Parameters:
dst - Destination memory address
dpitch - Pitch of destination memory
src - Source memory address
spitch - Pitch of source memory
width - Width of matrix transfer (columns in bytes)
height - Height of matrix transfer (rows)
kind - Type of transfer

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidPitchValue, cudaErrorInvalidDevicePointer, cudaErrorIn-
validMemcpyDirection

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaMemcpy, cudaMemcpyToArray, cudaMemcpy2DToArray, cudaMemcpyFromArray, cudaMem-
cpy2DFromArray, cudaMemcpyArrayToArray, cudaMemcpy2DArrayToArray, cudaMemcpyToSymbol,
cudaMemcpyFromSymbol, cudaMemcpyAsync, cudaMemcpy2DAsync, cudaMemcpyToArrayAsync, cud-
aMemcpy2DToArrayAsync, cudaMemcpyFromArrayAsync, cudaMemcpy2DFromArrayAsync, cudaMemcpy-
ToSymbolAsync, cudaMemcpyFromSymbolAsync

3.8.2.17 cudaError_t cudaMemcpy2DArrayToArray (struct cudaArray ∗ dst, size_t wOffsetDst, size_t

hOffsetDst, const struct cudaArray ∗ src, size_t wOffsetSrc, size_t hOffsetSrc, size_t width, size_t
height, enum cudaMemcpyKind kind)

Copies a matrix (height rows of width bytes each) from the CUDA array srcArray starting at the upper left
corner (wOffsetSrc, hOffsetSrc) to the CUDA array dst starting at the upper left corner (wOffsetDst,
hOffsetDst), where kind is one of cudaMemcpyHostToHost, cudaMemcpyHostToDevice, cudaMemcpyDevice-
ToHost, or cudaMemcpyDeviceToDevice, and specifies the direction of the copy.

Parameters:
dst - Destination memory address
wOffsetDst - Destination starting X offset
hOffsetDst - Destination starting Y offset
src - Source memory address
wOffsetSrc - Source starting X offset
hOffsetSrc - Source starting Y offset
width - Width of matrix transfer (columns in bytes)
height - Height of matrix transfer (rows)

Generated for NVIDIA CUDA Library by Doxygen

36 Module Documentation

kind - Type of transfer

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidMemcpyDirection

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaMemcpy, cudaMemcpy2D, cudaMemcpyToArray, cudaMemcpy2DToArray, cudaMemcpyFromArray, cu-
daMemcpy2DFromArray, cudaMemcpyArrayToArray, cudaMemcpyToSymbol, cudaMemcpyFromSymbol, cu-
daMemcpyAsync, cudaMemcpy2DAsync, cudaMemcpyToArrayAsync, cudaMemcpy2DToArrayAsync, cu-
daMemcpyFromArrayAsync, cudaMemcpy2DFromArrayAsync, cudaMemcpyToSymbolAsync, cudaMem-
cpyFromSymbolAsync

3.8.2.18 cudaError_t cudaMemcpy2DAsync (void ∗ dst, size_t dpitch, const void ∗ src, size_t spitch, size_t
width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream)

Copies a matrix (height rows of width bytes each) from the memory area pointed to by src to the memory area
pointed to by dst, where kind is one of cudaMemcpyHostToHost, cudaMemcpyHostToDevice, cudaMemcpyDe-
viceToHost, or cudaMemcpyDeviceToDevice, and specifies the direction of the copy. dpitch and spitch are the
widths in memory in bytes of the 2D arrays pointed to by dst and src, including any padding added to the end
of each row. The memory areas may not overlap. Calling cudaMemcpy2DAsync() with dst and src pointers that
do not match the direction of the copy results in an undefined behavior. cudaMemcpy2DAsync() returns an error if
dpitch or spitch is greater than the maximum allowed.
cudaMemcpy2DAsync() is asynchronous with respect to the host, so the call may return before the copy is complete.
It only works on page-locked host memory and returns an error if a pointer to pageable memory is passed as input.
The copy can optionally be associated to a stream by passing a non-zero stream argument. If kind is cudaMem-
cpyHostToDevice or cudaMemcpyDeviceToHost and stream is non-zero, the copy may overlap with operations in
other streams.
IMPORTANT NOTE: Copies with kind == cudaMemcpyDeviceToDevice are asynchronous with respect to the
host, but never overlap with kernel execution.

Parameters:
dst - Destination memory address
dpitch - Pitch of destination memory
src - Source memory address
spitch - Pitch of source memory
width - Width of matrix transfer (columns in bytes)
height - Height of matrix transfer (rows)
kind - Type of transfer
stream - Stream identifier

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidPitchValue, cudaErrorInvalidDevicePointer, cudaErrorIn-
validMemcpyDirection

Generated for NVIDIA CUDA Library by Doxygen

3.8 Memory Management 37

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaMemcpy, cudaMemcpy2D, cudaMemcpyToArray, cudaMemcpy2DToArray, cudaMemcpyFromArray, cu-
daMemcpy2DFromArray, cudaMemcpyArrayToArray, cudaMemcpy2DArrayToArray, cudaMemcpyToSymbol,
cudaMemcpyFromSymbol, cudaMemcpyAsync, cudaMemcpyToArrayAsync, cudaMemcpy2DToArrayAsync,
cudaMemcpyFromArrayAsync, cudaMemcpy2DFromArrayAsync, cudaMemcpyToSymbolAsync, cudaMem-
cpyFromSymbolAsync

3.8.2.19 cudaError_t cudaMemcpy2DFromArray (void ∗ dst, size_t dpitch, const struct cudaArray ∗ src,
size_t wOffset, size_t hOffset, size_t width, size_t height, enum cudaMemcpyKind kind)

Copies a matrix (height rows of width bytes each) from the CUDA array srcArray starting at the upper left
corner (wOffset, hOffset) to the memory area pointed to by dst, where kind is one of cudaMemcpyHost-
ToHost, cudaMemcpyHostToDevice, cudaMemcpyDeviceToHost, or cudaMemcpyDeviceToDevice, and specifies the
direction of the copy. dpitch is the width in memory in bytes of the 2D array pointed to by dst, including any
padding added to the end of each row. cudaMemcpy2DFromArray() returns an error if dpitch is greater than the
maximum allowed.

Parameters:
dst - Destination memory address
dpitch - Pitch of destination memory
src - Source memory address
wOffset - Source starting X offset
hOffset - Source starting Y offset
width - Width of matrix transfer (columns in bytes)
height - Height of matrix transfer (rows)
kind - Type of transfer

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidDevicePointer, cudaErrorInvalidPitchValue, cudaErrorIn-
validMemcpyDirection

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaMemcpy, cudaMemcpy2D, cudaMemcpyToArray, cudaMemcpy2DToArray, cudaMemcpyFromArray,
cudaMemcpyArrayToArray, cudaMemcpy2DArrayToArray, cudaMemcpyToSymbol, cudaMemcpyFromSym-
bol, cudaMemcpyAsync, cudaMemcpy2DAsync, cudaMemcpyToArrayAsync, cudaMemcpy2DToArrayAsync,
cudaMemcpyFromArrayAsync, cudaMemcpy2DFromArrayAsync, cudaMemcpyToSymbolAsync, cudaMem-
cpyFromSymbolAsync

Generated for NVIDIA CUDA Library by Doxygen

38 Module Documentation

3.8.2.20 cudaError_t cudaMemcpy2DFromArrayAsync (void ∗ dst, size_t dpitch, const struct cudaArray
∗ src, size_t wOffset, size_t hOffset, size_t width, size_t height, enum cudaMemcpyKind kind,
cudaStream_t stream)

Copies a matrix (height rows of width bytes each) from the CUDA array srcArray starting at the upper left
corner (wOffset, hOffset) to the memory area pointed to by dst, where kind is one of cudaMemcpyHost-
ToHost, cudaMemcpyHostToDevice, cudaMemcpyDeviceToHost, or cudaMemcpyDeviceToDevice, and specifies the
direction of the copy. dpitch is the width in memory in bytes of the 2D array pointed to by dst, including any
padding added to the end of each row. cudaMemcpy2DFromArrayAsync() returns an error if dpitch is greater than
the maximum allowed.
cudaMemcpy2DFromArrayAsync() is asynchronous with respect to the host, so the call may return before the copy is
complete. It only works on page-locked host memory and returns an error if a pointer to pageable memory is passed
as input. The copy can optionally be associated to a stream by passing a non-zero stream argument. If kind
is cudaMemcpyHostToDevice or cudaMemcpyDeviceToHost and stream is non-zero, the copy may overlap with
operations in other streams.
IMPORTANT NOTE: Copies with kind == cudaMemcpyDeviceToDevice are asynchronous with respect to the
host, but never overlap with kernel execution.

Parameters:
dst - Destination memory address
dpitch - Pitch of destination memory
src - Source memory address
wOffset - Source starting X offset
hOffset - Source starting Y offset
width - Width of matrix transfer (columns in bytes)
height - Height of matrix transfer (rows)
kind - Type of transfer
stream - Stream identifier

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidDevicePointer, cudaErrorInvalidPitchValue, cudaErrorIn-
validMemcpyDirection

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaMemcpy, cudaMemcpy2D, cudaMemcpyToArray, cudaMemcpy2DToArray, cudaMemcpyFromArray, cu-
daMemcpy2DFromArray, cudaMemcpyArrayToArray, cudaMemcpy2DArrayToArray, cudaMemcpyToSym-
bol, cudaMemcpyFromSymbol, cudaMemcpyAsync, cudaMemcpy2DAsync, cudaMemcpyToArrayAsync, cud-
aMemcpy2DToArrayAsync, cudaMemcpyFromArrayAsync, cudaMemcpyToSymbolAsync, cudaMemcpyFrom-
SymbolAsync

3.8.2.21 cudaError_t cudaMemcpy2DToArray (struct cudaArray ∗ dst, size_t wOffset, size_t hOffset, const
void ∗ src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind)

Copies a matrix (height rows of width bytes each) from the memory area pointed to by src to the CUDA array
dst starting at the upper left corner (wOffset, hOffset) where kind is one of cudaMemcpyHostToHost, cud-
aMemcpyHostToDevice, cudaMemcpyDeviceToHost, or cudaMemcpyDeviceToDevice, and specifies the direction of

Generated for NVIDIA CUDA Library by Doxygen

3.8 Memory Management 39

the copy. spitch is the width in memory in bytes of the 2D array pointed to by src, including any padding added
to the end of each row. cudaMemcpy2DToArray() returns an error if spitch is greater than the maximum allowed.

Parameters:
dst - Destination memory address
wOffset - Destination starting X offset
hOffset - Destination starting Y offset
src - Source memory address
spitch - Pitch of source memory
width - Width of matrix transfer (columns in bytes)
height - Height of matrix transfer (rows)
kind - Type of transfer

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidDevicePointer, cudaErrorInvalidPitchValue, cudaErrorIn-
validMemcpyDirection

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaMemcpy, cudaMemcpy2D, cudaMemcpyToArray, cudaMemcpyFromArray, cudaMemcpy2DFromArray,
cudaMemcpyArrayToArray, cudaMemcpy2DArrayToArray, cudaMemcpyToSymbol, cudaMemcpyFromSym-
bol, cudaMemcpyAsync, cudaMemcpy2DAsync, cudaMemcpyToArrayAsync, cudaMemcpy2DToArrayAsync,
cudaMemcpyFromArrayAsync, cudaMemcpy2DFromArrayAsync, cudaMemcpyToSymbolAsync, cudaMem-
cpyFromSymbolAsync

3.8.2.22 cudaError_t cudaMemcpy2DToArrayAsync (struct cudaArray ∗ dst, size_t wOffset, size_t

hOffset, const void ∗ src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind,
cudaStream_t stream)

Copies a matrix (height rows of width bytes each) from the memory area pointed to by src to the CUDA array
dst starting at the upper left corner (wOffset, hOffset) where kind is one of cudaMemcpyHostToHost, cud-
aMemcpyHostToDevice, cudaMemcpyDeviceToHost, or cudaMemcpyDeviceToDevice, and specifies the direction of
the copy. spitch is the width in memory in bytes of the 2D array pointed to by src, including any padding added
to the end of each row. cudaMemcpy2DToArrayAsync() returns an error if spitch is greater than the maximum
allowed.
cudaMemcpy2DToArrayAsync() is asynchronous with respect to the host, so the call may return before the copy is
complete. It only works on page-locked host memory and returns an error if a pointer to pageable memory is passed
as input. The copy can optionally be associated to a stream by passing a non-zero stream argument. If kind
is cudaMemcpyHostToDevice or cudaMemcpyDeviceToHost and stream is non-zero, the copy may overlap with
operations in other streams.
IMPORTANT NOTE: Copies with kind == cudaMemcpyDeviceToDevice are asynchronous with respect to the
host, but never overlap with kernel execution.

Parameters:
dst - Destination memory address

Generated for NVIDIA CUDA Library by Doxygen

40 Module Documentation

wOffset - Destination starting X offset

hOffset - Destination starting Y offset
src - Source memory address
spitch - Pitch of source memory
width - Width of matrix transfer (columns in bytes)
height - Height of matrix transfer (rows)
kind - Type of transfer
stream - Stream identifier

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidDevicePointer, cudaErrorInvalidPitchValue, cudaErrorIn-
validMemcpyDirection

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaMemcpy, cudaMemcpy2D, cudaMemcpyToArray, cudaMemcpy2DToArray, cudaMemcpyFromArray, cu-
daMemcpy2DFromArray, cudaMemcpyArrayToArray, cudaMemcpy2DArrayToArray, cudaMemcpyToSym-
bol, cudaMemcpyFromSymbol, cudaMemcpyAsync, cudaMemcpy2DAsync, cudaMemcpyToArrayAsync,
cudaMemcpyFromArrayAsync, cudaMemcpy2DFromArrayAsync, cudaMemcpyToSymbolAsync, cudaMem-
cpyFromSymbolAsync

3.8.2.23 cudaError_t cudaMemcpy3D (const struct cudaMemcpy3DParms ∗ p)

struct cudaExtent {
size_t width;
size_t height;
size_t depth;
};
struct cudaExtent make_cudaExtent(size_t w, size_t h, size_t d);

struct cudaPos {
size_t x;
size_t y;
size_t z;
};
struct cudaPos make_cudaPos(size_t x, size_t y, size_t z);

struct cudaMemcpy3DParms {
struct cudaArray *srcArray;
struct cudaPos srcPos;
struct cudaPitchedPtr srcPtr;
struct cudaArray *dstArray;
struct cudaPos dstPos;
struct cudaPitchedPtr dstPtr;
struct cudaExtent extent;
enum cudaMemcpyKind kind;
};

cudaMemcpy3D() copies data betwen two 3D objects. The source and destination objects may be in either host
memory, device memory, or a CUDA array. The source, destination, extent, and kind of copy performed is specified
by the cudaMemcpy3DParms struct which should be initialized to zero before use:

Generated for NVIDIA CUDA Library by Doxygen

3.8 Memory Management 41

cudaMemcpy3DParms myParms = {0};

The struct passed to cudaMemcpy3D() must specify one of srcArray or srcPtr and one of dstArray or
dstPtr. Passing more than one non-zero source or destination will cause cudaMemcpy3D() to return an error.
The srcPos and dstPos fields are optional offsets into the source and destination objects and are defined in units of
each object’s elements. The element for a host or device pointer is assumed to be unsigned char. For CUDA arrays,
positions must be in the range [0, 2048) for any dimension.
The extent field defines the dimensions of the transferred area in elements. If a CUDA array is participating in the
copy, the extent is defined in terms of that array’s elements. If no CUDA array is participating in the copy then the
extents are defined in elements of unsigned char.
The kind field defines the direction of the copy. It must be one of cudaMemcpyHostToHost, cudaMemcpyHostToDe-
vice, cudaMemcpyDeviceToHost, or cudaMemcpyDeviceToDevice.
If the source and destination are both arrays, cudaMemcpy3D() will return an error if they do not have the same
element size.
The source and destination object may not overlap. If overlapping source and destination objects are specified, unde-
fined behavior will result.
cudaMemcpy3D() returns an error if the pitch of srcPtr or dstPtr is greater than the maximum allowed. The
pitch of a cudaPitchedPtr allocated with cudaMalloc3D() will always be valid.

Parameters:
p - 3D memory copy parameters

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidDevicePointer, cudaErrorInvalidPitchValue, cudaErrorIn-
validMemcpyDirection

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaMalloc3D, cudaMalloc3DArray, cudaMemset3D, cudaMemcpy3DAsync, cudaMemcpy, cudaMemcpy2D,
cudaMemcpyToArray, cudaMemcpy2DToArray, cudaMemcpyFromArray, cudaMemcpy2DFromArray, cu-
daMemcpyArrayToArray, cudaMemcpy2DArrayToArray, cudaMemcpyToSymbol, cudaMemcpyFromSym-
bol, cudaMemcpyAsync, cudaMemcpy2DAsync, cudaMemcpyToArrayAsync, cudaMemcpy2DToArrayAsync,
cudaMemcpyFromArrayAsync, cudaMemcpy2DFromArrayAsync, cudaMemcpyToSymbolAsync, cudaMem-
cpyFromSymbolAsync, make_cudaExtent, make_cudaPos

3.8.2.24 cudaError_t cudaMemcpy3DAsync (const struct cudaMemcpy3DParms ∗ p, cudaStream_t stream)

struct cudaExtent {
size_t width;
size_t height;
size_t depth;
};
struct cudaExtent make_cudaExtent(size_t w, size_t h, size_t d);

struct cudaPos {
size_t x;
size_t y;

Generated for NVIDIA CUDA Library by Doxygen

42 Module Documentation

size_t z;
};
struct cudaPos make_cudaPos(size_t x, size_t y, size_t z);

struct cudaMemcpy3DParms {
struct cudaArray *srcArray;
struct cudaPos srcPos;
struct cudaPitchedPtr srcPtr;
struct cudaArray *dstArray;
struct cudaPos dstPos;
struct cudaPitchedPtr dstPtr;
struct cudaExtent extent;
enum cudaMemcpyKind kind;
};

cudaMemcpy3DAsync() copies data betwen two 3D objects. The source and destination objects may be in either host
memory, device memory, or a CUDA array. The source, destination, extent, and kind of copy performed is specified
by the cudaMemcpy3DParms struct which should be initialized to zero before use:

cudaMemcpy3DParms myParms = {0};

The struct passed to cudaMemcpy3DAsync() must specify one of srcArray or srcPtr and one of dstArray or
dstPtr. Passing more than one non-zero source or destination will cause cudaMemcpy3DAsync() to return an error.
The srcPos and dstPos fields are optional offsets into the source and destination objects and are defined in units of
each object’s elements. The element for a host or device pointer is assumed to be unsigned char. For CUDA arrays,
positions must be in the range [0, 2048) for any dimension.
The extent field defines the dimensions of the transferred area in elements. If a CUDA array is participating in the
copy, the extent is defined in terms of that array’s elements. If no CUDA array is participating in the copy then the
extents are defined in elements of unsigned char.
The kind field defines the direction of the copy. It must be one of cudaMemcpyHostToHost, cudaMemcpyHostToDe-
vice, cudaMemcpyDeviceToHost, or cudaMemcpyDeviceToDevice.
If the source and destination are both arrays, cudaMemcpy3DAsync() will return an error if they do not have the same
element size.
The source and destination object may not overlap. If overlapping source and destination objects are specified, unde-
fined behavior will result.
cudaMemcpy3DAsync() returns an error if the pitch of srcPtr or dstPtr is greater than the maximum allowed.
The pitch of a cudaPitchedPtr allocated with cudaMalloc3D() will always be valid.
cudaMemcpy3DAsync() is asynchronous with respect to the host, so the call may return before the copy is complete.
It only works on page-locked host memory and returns an error if a pointer to pageable memory is passed as input.
The copy can optionally be associated to a stream by passing a non-zero stream argument. If kind is cudaMem-
cpyHostToDevice or cudaMemcpyDeviceToHost and stream is non-zero, the copy may overlap with operations in
other streams.
IMPORTANT NOTE: Copies with kind == cudaMemcpyDeviceToDevice are asynchronous with respect to the
host, but never overlap with kernel execution.

Parameters:
p - 3D memory copy parameters
stream - Stream identifier

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidDevicePointer, cudaErrorInvalidPitchValue, cudaErrorIn-
validMemcpyDirection

Generated for NVIDIA CUDA Library by Doxygen

3.8 Memory Management 43

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaMalloc3D, cudaMalloc3DArray, cudaMemset3D, cudaMemcpy3D, cudaMemcpy, cudaMemcpy2D, cud-
aMemcpyToArray, cudaMemcpy2DToArray, cudaMemcpyFromArray, cudaMemcpy2DFromArray, cudaMem-
cpyArrayToArray, cudaMemcpy2DArrayToArray, cudaMemcpyToSymbol, cudaMemcpyFromSymbol, cu-
daMemcpyAsync, cudaMemcpy2DAsync, cudaMemcpyToArrayAsync, cudaMemcpy2DToArrayAsync, cu-
daMemcpyFromArrayAsync, cudaMemcpy2DFromArrayAsync, cudaMemcpyToSymbolAsync, cudaMem-
cpyFromSymbolAsync, make_cudaExtent, make_cudaPos

3.8.2.25 cudaError_t cudaMemcpyArrayToArray (struct cudaArray ∗ dst, size_t wOffsetDst, size_t

hOffsetDst, const struct cudaArray ∗ src, size_t wOffsetSrc, size_t hOffsetSrc, size_t count, enum
cudaMemcpyKind kind)

Copies count bytes from the CUDA array src starting at the upper left corner (wOffsetSrc, hOffsetSrc)
to the CUDA array dst starting at the upper left corner (wOffsetDst, hOffsetDst) where kind is one of cu-
daMemcpyHostToHost, cudaMemcpyHostToDevice, cudaMemcpyDeviceToHost, or cudaMemcpyDeviceToDevice,
and specifies the direction of the copy.

Parameters:
dst - Destination memory address
wOffsetDst - Destination starting X offset
hOffsetDst - Destination starting Y offset
src - Source memory address
wOffsetSrc - Source starting X offset
hOffsetSrc - Source starting Y offset
count - Size in bytes to copy
kind - Type of transfer

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidMemcpyDirection

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaMemcpy, cudaMemcpy2D, cudaMemcpyToArray, cudaMemcpy2DToArray, cudaMemcpyFromArray,
cudaMemcpy2DFromArray, cudaMemcpy2DArrayToArray, cudaMemcpyToSymbol, cudaMemcpyFromSym-
bol, cudaMemcpyAsync, cudaMemcpy2DAsync, cudaMemcpyToArrayAsync, cudaMemcpy2DToArrayAsync,
cudaMemcpyFromArrayAsync, cudaMemcpy2DFromArrayAsync, cudaMemcpyToSymbolAsync, cudaMem-
cpyFromSymbolAsync

Generated for NVIDIA CUDA Library by Doxygen

44 Module Documentation

3.8.2.26 cudaError_t cudaMemcpyAsync (void ∗ dst, const void ∗ src, size_t count, enum cudaMemcpyKind
kind, cudaStream_t stream)

Copies count bytes from the memory area pointed to by src to the memory area pointed to by dst, where kind is
one of cudaMemcpyHostToHost, cudaMemcpyHostToDevice, cudaMemcpyDeviceToHost, or cudaMemcpyDevice-
ToDevice, and specifies the direction of the copy. The memory areas may not overlap. Calling cudaMemcpyAsync()
with dst and src pointers that do not match the direction of the copy results in an undefined behavior.
cudaMemcpyAsync() is asynchronous with respect to the host, so the call may return before the copy is complete. It
only works on page-locked host memory and returns an error if a pointer to pageable memory is passed as input. The
copy can optionally be associated to a stream by passing a non-zero stream argument. If kind is cudaMemcpy-
HostToDevice or cudaMemcpyDeviceToHost and the stream is non-zero, the copy may overlap with operations in
other streams.
IMPORTANT NOTE: Copies with kind == cudaMemcpyDeviceToDevice are asynchronous with respect to the
host, but never overlap with kernel execution.

Parameters:
dst - Destination memory address
src - Source memory address
count - Size in bytes to copy
kind - Type of transfer
stream - Stream identifier

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidDevicePointer, cudaErrorInvalidMemcpyDirection

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaMemcpy, cudaMemcpy2D, cudaMemcpyToArray, cudaMemcpy2DToArray, cudaMemcpyFromArray,
cudaMemcpy2DFromArray, cudaMemcpyArrayToArray, cudaMemcpy2DArrayToArray, cudaMemcpy-
ToSymbol, cudaMemcpyFromSymbol, cudaMemcpy2DAsync, cudaMemcpyToArrayAsync, cudaMem-
cpy2DToArrayAsync, cudaMemcpyFromArrayAsync, cudaMemcpy2DFromArrayAsync, cudaMemcpyToSym-
bolAsync, cudaMemcpyFromSymbolAsync

3.8.2.27 cudaError_t cudaMemcpyFromArray (void ∗ dst, const struct cudaArray ∗ src, size_t wOffset,
size_t hOffset, size_t count, enum cudaMemcpyKind kind)

Copies count bytes from the CUDA array src starting at the upper left corner (wOffset, hOffset) to the memory
area pointed to by dst, where kind is one of cudaMemcpyHostToHost, cudaMemcpyHostToDevice, cudaMemcpy-
DeviceToHost, or cudaMemcpyDeviceToDevice, and specifies the direction of the copy.

Parameters:
dst - Destination memory address
src - Source memory address
wOffset - Source starting X offset
hOffset - Source starting Y offset

Generated for NVIDIA CUDA Library by Doxygen

3.8 Memory Management 45

count - Size in bytes to copy

kind - Type of transfer

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidDevicePointer, cudaErrorInvalidMemcpyDirection

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaMemcpy, cudaMemcpy2D, cudaMemcpyToArray, cudaMemcpy2DToArray, cudaMemcpy2DFromArray,
cudaMemcpyArrayToArray, cudaMemcpy2DArrayToArray, cudaMemcpyToSymbol, cudaMemcpyFromSym-
bol, cudaMemcpyAsync, cudaMemcpy2DAsync, cudaMemcpyToArrayAsync, cudaMemcpy2DToArrayAsync,
cudaMemcpyFromArrayAsync, cudaMemcpy2DFromArrayAsync, cudaMemcpyToSymbolAsync, cudaMem-
cpyFromSymbolAsync

3.8.2.28 cudaError_t cudaMemcpyFromArrayAsync (void ∗ dst, const struct cudaArray ∗ src, size_t
wOffset, size_t hOffset, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream)

Copies count bytes from the CUDA array src starting at the upper left corner (wOffset, hOffset) to the memory
area pointed to by dst, where kind is one of cudaMemcpyHostToHost, cudaMemcpyHostToDevice, cudaMemcpy-
DeviceToHost, or cudaMemcpyDeviceToDevice, and specifies the direction of the copy.
cudaMemcpyFromArrayAsync() is asynchronous with respect to the host, so the call may return before the copy is
complete. It only works on page-locked host memory and returns an error if a pointer to pageable memory is passed
as input. The copy can optionally be associated to a stream by passing a non-zero stream argument. If kind
is cudaMemcpyHostToDevice or cudaMemcpyDeviceToHost and stream is non-zero, the copy may overlap with
operations in other streams.
IMPORTANT NOTE: Copies with kind == cudaMemcpyDeviceToDevice are asynchronous with respect to the
host, but never overlap with kernel execution.

Parameters:
dst - Destination memory address
src - Source memory address
wOffset - Source starting X offset
hOffset - Source starting Y offset
count - Size in bytes to copy
kind - Type of transfer
stream - Stream identifier

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidDevicePointer, cudaErrorInvalidMemcpyDirection

Note:
Note that this function may also return error codes from previous, asynchronous launches.

Generated for NVIDIA CUDA Library by Doxygen

46 Module Documentation

See also:
cudaMemcpy, cudaMemcpy2D, cudaMemcpyToArray, cudaMemcpy2DToArray, cudaMemcpyFromArray, cu-
daMemcpy2DFromArray, cudaMemcpyArrayToArray, cudaMemcpy2DArrayToArray, cudaMemcpyToSym-
bol, cudaMemcpyFromSymbol, cudaMemcpyAsync, cudaMemcpy2DAsync, cudaMemcpyToArrayAsync,
cudaMemcpy2DToArrayAsync, cudaMemcpy2DFromArrayAsync, cudaMemcpyToSymbolAsync, cudaMem-
cpyFromSymbolAsync

3.8.2.29 cudaError_t cudaMemcpyFromSymbol (void ∗ dst, const char ∗ symbol, size_t count, size_t offset,
enum cudaMemcpyKind kind)

Copies count bytes from the memory area pointed to by offset bytes from the start of symbol symbol to the
memory area pointed to by dst. The memory areas may not overlap. symbol can either be a variable that resides in
global or constant memory space, or it can be a character string, naming a variable that resides in global or constant
memory space. kind can be either cudaMemcpyDeviceToHost or cudaMemcpyDeviceToDevice.

Parameters:
dst - Destination memory address
symbol - Symbol source from device
count - Size in bytes to copy
offset - Offset from start of symbol in bytes
kind - Type of transfer

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidSymbol, cudaErrorInvalidDevicePointer, cudaErrorIn-
validMemcpyDirection

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaMemcpy, cudaMemcpy2D, cudaMemcpyToArray, cudaMemcpy2DToArray, cudaMemcpyFromArray, cu-
daMemcpy2DFromArray, cudaMemcpyArrayToArray, cudaMemcpy2DArrayToArray, cudaMemcpyToSym-
bol, cudaMemcpyAsync, cudaMemcpy2DAsync, cudaMemcpyToArrayAsync, cudaMemcpy2DToArrayAsync,
cudaMemcpyFromArrayAsync, cudaMemcpy2DFromArrayAsync, cudaMemcpyToSymbolAsync, cudaMem-
cpyFromSymbolAsync

3.8.2.30 cudaError_t cudaMemcpyFromSymbolAsync (void ∗ dst, const char ∗ symbol, size_t count, size_t
offset, enum cudaMemcpyKind kind, cudaStream_t stream)

Copies count bytes from the memory area pointed to by offset bytes from the start of symbol symbol to the
memory area pointed to by dst. The memory areas may not overlap. symbol can either be a variable that resides in
global or constant memory space, or it can be a character string, naming a variable that resides in global or constant
memory space. kind can be either cudaMemcpyDeviceToHost or cudaMemcpyDeviceToDevice.
cudaMemcpyFromSymbolAsync() is asynchronous with respect to the host, so the call may return before the copy is
complete. It only works on page-locked host memory and returns an error if a pointer to pageable memory is passed
as input. The copy can optionally be associated to a stream by passing a non-zero stream argument. If kind is
cudaMemcpyDeviceToHost and stream is non-zero, the copy may overlap with operations in other streams.

Generated for NVIDIA CUDA Library by Doxygen

3.8 Memory Management 47

IMPORTANT NOTE: Copies with kind == cudaMemcpyDeviceToDevice are asynchronous with respect to the
host, but never overlap with kernel execution.

Parameters:
dst - Destination memory address
symbol - Symbol source from device
count - Size in bytes to copy
offset - Offset from start of symbol in bytes
kind - Type of transfer
stream - Stream identifier

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidSymbol, cudaErrorInvalidDevicePointer, cudaErrorIn-
validMemcpyDirection

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaMemcpy, cudaMemcpy2D, cudaMemcpyToArray, cudaMemcpy2DToArray, cudaMemcpyFromArray, cu-
daMemcpy2DFromArray, cudaMemcpyArrayToArray, cudaMemcpy2DArrayToArray, cudaMemcpyToSym-
bol, cudaMemcpyFromSymbol, cudaMemcpyAsync, cudaMemcpy2DAsync, cudaMemcpyToArrayAsync, cu-
daMemcpy2DToArrayAsync, cudaMemcpyFromArrayAsync, cudaMemcpy2DFromArrayAsync, cudaMemcpy-
ToSymbolAsync

3.8.2.31 cudaError_t cudaMemcpyToArray (struct cudaArray ∗ dst, size_t wOffset, size_t hOffset, const
void ∗ src, size_t count, enum cudaMemcpyKind kind)

Copies count bytes from the memory area pointed to by src to the CUDA array dst starting at the upper left corner
(wOffset, hOffset), where kind is one of cudaMemcpyHostToHost, cudaMemcpyHostToDevice, cudaMemcpy-
DeviceToHost, or cudaMemcpyDeviceToDevice, and specifies the direction of the copy.

Parameters:
dst - Destination memory address
wOffset - Destination starting X offset
hOffset - Destination starting Y offset
src - Source memory address
count - Size in bytes to copy
kind - Type of transfer

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidDevicePointer, cudaErrorInvalidMemcpyDirection

Note:
Note that this function may also return error codes from previous, asynchronous launches.

Generated for NVIDIA CUDA Library by Doxygen

48 Module Documentation

See also:
cudaMemcpy, cudaMemcpy2D, cudaMemcpy2DToArray, cudaMemcpyFromArray, cudaMem-
cpy2DFromArray, cudaMemcpyArrayToArray, cudaMemcpy2DArrayToArray, cudaMemcpyToSymbol,
cudaMemcpyFromSymbol, cudaMemcpyAsync, cudaMemcpy2DAsync, cudaMemcpyToArrayAsync, cud-
aMemcpy2DToArrayAsync, cudaMemcpyFromArrayAsync, cudaMemcpy2DFromArrayAsync, cudaMemcpy-
ToSymbolAsync, cudaMemcpyFromSymbolAsync

3.8.2.32 cudaError_t cudaMemcpyToArrayAsync (struct cudaArray ∗ dst, size_t wOffset, size_t hOffset,
const void ∗ src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream)

Copies count bytes from the memory area pointed to by src to the CUDA array dst starting at the upper left corner
(wOffset, hOffset), where kind is one of cudaMemcpyHostToHost, cudaMemcpyHostToDevice, cudaMemcpy-
DeviceToHost, or cudaMemcpyDeviceToDevice, and specifies the direction of the copy.
cudaMemcpyToArrayAsync() is asynchronous with respect to the host, so the call may return before the copy is
complete. It only works on page-locked host memory and returns an error if a pointer to pageable memory is passed
as input. The copy can optionally be associated to a stream by passing a non-zero stream argument. If kind
is cudaMemcpyHostToDevice or cudaMemcpyDeviceToHost and stream is non-zero, the copy may overlap with
operations in other streams.
IMPORTANT NOTE: Copies with kind == cudaMemcpyDeviceToDevice are asynchronous with respect to the
host, but never overlap with kernel execution.

Parameters:
dst - Destination memory address
wOffset - Destination starting X offset
hOffset - Destination starting Y offset
src - Source memory address
count - Size in bytes to copy
kind - Type of transfer
stream - Stream identifier

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidDevicePointer, cudaErrorInvalidMemcpyDirection

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaMemcpy, cudaMemcpy2D, cudaMemcpyToArray, cudaMemcpy2DToArray, cudaMemcpyFromArray, cu-
daMemcpy2DFromArray, cudaMemcpyArrayToArray, cudaMemcpy2DArrayToArray, cudaMemcpyToSym-
bol, cudaMemcpyFromSymbol, cudaMemcpyAsync, cudaMemcpy2DAsync, cudaMemcpy2DToArrayAsync,
cudaMemcpyFromArrayAsync, cudaMemcpy2DFromArrayAsync, cudaMemcpyToSymbolAsync, cudaMem-
cpyFromSymbolAsync

Generated for NVIDIA CUDA Library by Doxygen

3.8 Memory Management 49

3.8.2.33 cudaError_t cudaMemcpyToSymbol (const char ∗ symbol, const void ∗ src, size_t count, size_t
offset, enum cudaMemcpyKind kind)

Copies count bytes from the memory area pointed to by src to the memory area pointed to by offset bytes from
the start of symbol symbol. The memory areas may not overlap. symbol can either be a variable that resides in
global or constant memory space, or it can be a character string, naming a variable that resides in global or constant
memory space. kind can be either cudaMemcpyHostToDevice or cudaMemcpyDeviceToDevice.

Parameters:
symbol - Symbol destination on device
src - Source memory address
count - Size in bytes to copy
offset - Offset from start of symbol in bytes
kind - Type of transfer

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidSymbol, cudaErrorInvalidDevicePointer, cudaErrorIn-
validMemcpyDirection

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaMemcpy, cudaMemcpy2D, cudaMemcpyToArray, cudaMemcpy2DToArray, cudaMemcpyFromArray, cu-
daMemcpy2DFromArray, cudaMemcpyArrayToArray, cudaMemcpy2DArrayToArray, cudaMemcpyFromSym-
bol, cudaMemcpyAsync, cudaMemcpy2DAsync, cudaMemcpyToArrayAsync, cudaMemcpy2DToArrayAsync,
cudaMemcpyFromArrayAsync, cudaMemcpy2DFromArrayAsync, cudaMemcpyToSymbolAsync, cudaMem-
cpyFromSymbolAsync

3.8.2.34 cudaError_t cudaMemcpyToSymbolAsync (const char ∗ symbol, const void ∗ src, size_t count,
size_t offset, enum cudaMemcpyKind kind, cudaStream_t stream)

Copies count bytes from the memory area pointed to by src to the memory area pointed to by offset bytes from
the start of symbol symbol. The memory areas may not overlap. symbol can either be a variable that resides in
global or constant memory space, or it can be a character string, naming a variable that resides in global or constant
memory space. kind can be either cudaMemcpyHostToDevice or cudaMemcpyDeviceToDevice.
cudaMemcpyToSymbolAsync() is asynchronous with respect to the host, so the call may return before the copy is
complete. It only works on page-locked host memory and returns an error if a pointer to pageable memory is passed
as input. The copy can optionally be associated to a stream by passing a non-zero stream argument. If kind is
cudaMemcpyHostToDevice and stream is non-zero, the copy may overlap with operations in other streams.
IMPORTANT NOTE: Copies with kind == cudaMemcpyDeviceToDevice are asynchronous with respect to the
host, but never overlap with kernel execution.

Parameters:
symbol - Symbol destination on device
src - Source memory address
count - Size in bytes to copy

Generated for NVIDIA CUDA Library by Doxygen

50 Module Documentation

offset - Offset from start of symbol in bytes

kind - Type of transfer
stream - Stream identifier

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidSymbol, cudaErrorInvalidDevicePointer, cudaErrorIn-
validMemcpyDirection

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaMemcpy, cudaMemcpy2D, cudaMemcpyToArray, cudaMemcpy2DToArray, cudaMemcpyFromArray, cu-
daMemcpy2DFromArray, cudaMemcpyArrayToArray, cudaMemcpy2DArrayToArray, cudaMemcpyToSym-
bol, cudaMemcpyFromSymbol, cudaMemcpyAsync, cudaMemcpy2DAsync, cudaMemcpyToArrayAsync, cu-
daMemcpy2DToArrayAsync, cudaMemcpyFromArrayAsync, cudaMemcpy2DFromArrayAsync, cudaMem-
cpyFromSymbolAsync

3.8.2.35 cudaError_t cudaMemset (void ∗ devPtr, int value, size_t count)

Fills the first count bytes of the memory area pointed to by devPtr with the constant byte value value.

Parameters:
devPtr - Pointer to device memory
value - Value to set for each byte of specified memory
count - Size in bytes to set

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidDevicePointer

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaMemset2D, cudaMemset3D

3.8.2.36 cudaError_t cudaMemset2D (void ∗ devPtr, size_t pitch, int value, size_t width, size_t height)

Sets to the specified value value a matrix (height rows of width bytes each) pointed to by dstPtr. pitch is
the width in bytes of the 2D array pointed to by dstPtr, including any padding added to the end of each row. This
function performs fastest when the pitch is one that has been passed back by cudaMallocPitch().

Parameters:
devPtr - Pointer to 2D device memory
pitch - Pitch in bytes of 2D device memory

Generated for NVIDIA CUDA Library by Doxygen

3.8 Memory Management 51

value - Value to set for each byte of specified memory

width - Width of matrix set (columns in bytes)
height - Height of matrix set (rows)

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidDevicePointer

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaMemset, cudaMemset3D

3.8.2.37 cudaError_t cudaMemset3D (struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent
extent)

Initializes each element of a 3D array to the specified value value. The object to initialize is defined by
pitchedDevPtr. The pitch field of pitchedDevPtr is the width in memory in bytes of the 3D array pointed
to by pitchedDevPtr, including any padding added to the end of each row. The xsize field specifies the logical
width of each row in bytes, while the ysize field specifies the height of each 2D slice in rows.
The extents of the initialized region are specified as a width in bytes, a height in rows, and a depth in slices.
Extents with width greater than or equal to the xsize of pitchedDevPtr may perform significantly faster than
extents narrower than the xsize. Secondarily, extents with height equal to the ysize of pitchedDevPtr will
perform faster than when the height is shorter than the ysize.
This function performs fastest when the pitchedDevPtr has been allocated by cudaMalloc3D().

Parameters:
pitchedDevPtr - Pointer to pitched device memory
value - Value to set for each byte of specified memory
extent - Size parameters for where to set device memory

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidDevicePointer

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaMemset, cudaMemset2D, cudaMalloc3D, make_cudaPitchedPtr, make_cudaExtent

Generated for NVIDIA CUDA Library by Doxygen

52 Module Documentation

3.9 OpenGL Interoperability

Functions
• cudaError_t cudaGLMapBufferObject (void ∗∗devPtr, GLuint bufObj)
Maps a buffer object for access by CUDA.

• cudaError_t cudaGLMapBufferObjectAsync (void ∗∗devPtr, GLuint bufObj, cudaStream_t stream)

Maps a buffer object for access by CUDA.

• cudaError_t cudaGLRegisterBufferObject (GLuint bufObj)

Registers a buffer object for access by CUDA.

• cudaError_t cudaGLSetBufferObjectMapFlags (GLuint bufObj, unsigned int flags)

Set usage flags for mapping an OpenGL buffer.

• cudaError_t cudaGLSetGLDevice (int device)

Sets the CUDA device for use with OpenGL interoperability.

• cudaError_t cudaGLUnmapBufferObject (GLuint bufObj)

Unmaps a buffer object for access by CUDA.

• cudaError_t cudaGLUnmapBufferObjectAsync (GLuint bufObj, cudaStream_t stream)

Unmaps a buffer object for access by CUDA.

• cudaError_t cudaGLUnregisterBufferObject (GLuint bufObj)

Unregisters a buffer object for access by CUDA.

• cudaError_t cudaWGLGetDevice (int ∗device, HGPUNV hGpu)

Gets the CUDA device associated with hGpu.

3.9.1 Detailed Description

This section describes the OpenGL interoperability functions of the CUDA runtime application programming inter-
face.

3.9.2 Function Documentation

3.9.2.1 cudaError_t cudaGLMapBufferObject (void ∗∗ devPtr, GLuint bufObj)

Maps the buffer object of ID bufObj into the address space of CUDA and returns in ∗devPtr the base pointer of the
resulting mapping. The buffer must have previously been registered by calling cudaGLRegisterBufferObject(). While
a buffer is mapped by CUDA, any OpenGL operation which references the buffer will result in undefined behavior.
The OpenGL context used to create the buffer, or another context from the same share group, must be bound to the
current thread when this is called.
All streams in the current thread are synchronized with the current GL context.

Parameters:
devPtr - Returned device pointer to CUDA object

Generated for NVIDIA CUDA Library by Doxygen

3.9 OpenGL Interoperability 53

bufObj - Buffer object ID to map

Returns:
cudaSuccess, cudaErrorMapBufferObjectFailed

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaGLSetGLDevice, cudaGLRegisterBufferObject, cudaGLUnmapBufferObject, cudaGLUnregisterBufferOb-
ject, cudaGLMapBufferObjectAsync, cudaGLUnmapBufferObjectAsync

3.9.2.2 cudaError_t cudaGLMapBufferObjectAsync (void ∗∗ devPtr, GLuint bufObj, cudaStream_t stream)

Maps the buffer object of ID bufObj into the address space of CUDA and returns in ∗devPtr the base pointer of the
resulting mapping. The buffer must have previously been registered by calling cudaGLRegisterBufferObject(). While
a buffer is mapped by CUDA, any OpenGL operation which references the buffer will result in undefined behavior.
The OpenGL context used to create the buffer, or another context from the same share group, must be bound to the
current thread when this is called.
Stream /p stream is synchronized with the current GL context.

Parameters:
devPtr - Returned device pointer to CUDA object
bufObj - Buffer object ID to map
stream - Stream to synchronize

Returns:
cudaSuccess, cudaErrorMapBufferObjectFailed

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaGLSetGLDevice, cudaGLRegisterBufferObject, cudaGLUnmapBufferObject, cudaGLUnregisterBufferOb-
ject, cudaGLUnmapBufferObjectAsync

3.9.2.3 cudaError_t cudaGLRegisterBufferObject (GLuint bufObj)

Registers the buffer object of ID bufObj for access by CUDA. This function must be called before CUDA can map
the buffer object. The OpenGL context used to create the buffer, or another context from the same share group, must
be bound to the current thread when this is called.

Parameters:
bufObj - Buffer object ID to register

Generated for NVIDIA CUDA Library by Doxygen

54 Module Documentation

Returns:
cudaSuccess, cudaErrorInitializationError

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaGLSetGLDevice, cudaGLMapBufferObject, cudaGLUnmapBufferObject, cudaGLUnregisterBufferObject,
cudaGLMapBufferObjectAsync, cudaGLUnmapBufferObjectAsync

3.9.2.4 cudaError_t cudaGLSetBufferObjectMapFlags (GLuint bufObj, unsigned int flags)

Set flags for mapping the OpenGL buffer bufObj

Changes to flags will take effect the next time bufObj is mapped. The flags argument may be any of the following:

• cudaGLMapFlagsNone: Specifies no hints about how this buffer will be used. It is therefore assumed that this
buffer will be read from and written to by CUDA kernels. This is the default value.

• cudaGLMapFlagsReadOnly: Specifies that CUDA kernels which access this buffer will not write to the buffer.

• cudaGLMapFlagsWriteDiscard: Specifies that CUDA kernels which access this buffer will not read from the
buffer and will write over the entire contents of the buffer, so none of the data previously stored in the buffer
will be preserved.

If bufObj has not been registered for use with CUDA, then cudaErrorInvalidResourceHandle is returned. If bufObj
is presently mapped for access by CUDA, then cudaErrorUnknown is returned.

Parameters:
bufObj - Registered buffer object to set flags for
flags - Parameters for buffer mapping

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidResourceHandle, cudaErrorUnknown

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaGLSetGLDevice, cudaGLRegisterBufferObject, cudaGLUnregisterBufferObject, cudaGLMapBufferObject,
cudaGLUnmapBufferObject, cudaGLMapBufferObjectAsync, cudaGLUnmapBufferObjectAsync

3.9.2.5 cudaError_t cudaGLSetGLDevice (int device)

Records device as the device on which the active host thread executes the device code. Records the thread as using
OpenGL interopability. If the host thread has already initialized the CUDA runtime by calling non-device management
runtime functions, this call returns cudaErrorSetOnActiveProcess.

Generated for NVIDIA CUDA Library by Doxygen

3.9 OpenGL Interoperability 55

Parameters:
device - Device to use for OpenGL interoperability

Returns:
cudaSuccess, cudaErrorInvalidDevice, cudaErrorSetOnActiveProcess

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaGLRegisterBufferObject, cudaGLMapBufferObject, cudaGLUnmapBufferObject, cudaGLUnregisterBuffer-
Object, cudaGLMapBufferObjectAsync, cudaGLUnmapBufferObjectAsync

3.9.2.6 cudaError_t cudaGLUnmapBufferObject (GLuint bufObj)

Unmaps the buffer object of ID bufObj for access by CUDA. When a buffer is unmapped, the base address returned
by cudaGLMapBufferObject() is invalid and subsequent references to the address result in undefined behavior. The
OpenGL context used to create the buffer, or another context from the same share group, must be bound to the current
thread when this is called.
All streams in the current thread are synchronized with the current GL context.

Parameters:
bufObj - Buffer object to unmap

Returns:
cudaSuccess, cudaErrorInvalidDevicePointer, cudaErrorUnmapBufferObjectFailed

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaGLSetGLDevice, cudaGLRegisterBufferObject, cudaGLMapBufferObject, cudaGLUnregisterBufferObject,
cudaGLMapBufferObjectAsync, cudaGLUnmapBufferObjectAsync

3.9.2.7 cudaError_t cudaGLUnmapBufferObjectAsync (GLuint bufObj, cudaStream_t stream)

Unmaps the buffer object of ID bufObj for access by CUDA. When a buffer is unmapped, the base address returned
by cudaGLMapBufferObject() is invalid and subsequent references to the address result in undefined behavior. The
OpenGL context used to create the buffer, or another context from the same share group, must be bound to the current
thread when this is called.
Stream /p stream is synchronized with the current GL context.

Parameters:
bufObj - Buffer object to unmap
stream - Stream to synchronize

Generated for NVIDIA CUDA Library by Doxygen

56 Module Documentation

Returns:
cudaSuccess, cudaErrorInvalidDevicePointer, cudaErrorUnmapBufferObjectFailed

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaGLSetGLDevice, cudaGLRegisterBufferObject, cudaGLMapBufferObject, cudaGLUnregisterBufferObject,
cudaGLMapBufferObjectAsync

3.9.2.8 cudaError_t cudaGLUnregisterBufferObject (GLuint bufObj)

Unregisters the buffer object of ID bufObj for access by CUDA and releases any CUDA resources associated with
the buffer. Once a buffer is unregistered, it may no longer be mapped by CUDA. The GL context used to create the
buffer, or another context from the same share group, must be bound to the current thread when this is called.

Parameters:
bufObj - Buffer object to unregister

Returns:
cudaSuccess

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaGLSetGLDevice, cudaGLRegisterBufferObject, cudaGLMapBufferObject, cudaGLUnmapBufferObject,
cudaGLMapBufferObjectAsync, cudaGLUnmapBufferObjectAsync

3.9.2.9 cudaError_t cudaWGLGetDevice (int ∗ device, HGPUNV hGpu)

Returns the CUDA device associated with a hGpu, if applicable.

Parameters:
device - Returns the device associated with hGpu, or -1 if hGpu is not a compute device.
hGpu - Handle to a GPU, as queried via WGL_NV_gpu_affinity()

Returns:
cudaSuccess

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
WGL_NV_gpu_affinity, cudaGLSetGLDevice

Generated for NVIDIA CUDA Library by Doxygen

3.10 Direct3D 9 Interoperability 57

3.10 Direct3D 9 Interoperability

Direct3D 9 Interoperability with CUDA Runtime

Direct3D 9 Interoperability with CUDA Runtime

Author:
NVIDIA Corporation

• enum cudaD3D9MapFlags {
cudaD3D9MapFlagsNone,
cudaD3D9MapFlagsReadOnly,
cudaD3D9MapFlagsWriteDiscard }
• enum cudaD3D9RegisterFlags {
cudaD3D9RegisterFlagsNone,
cudaD3D9RegisterFlagsArray }
• cudaError_t cudaD3D9GetDevice (int ∗device, const char ∗pszAdapterName)
Gets the device number for an adapter.

• cudaError_t cudaD3D9GetDirect3DDevice (IDirect3DDevice9 ∗∗ppDxDevice)

Gets the Direct3D device against which the current CUDA context was created.

• cudaError_t cudaD3D9MapResources (int count, IDirect3DResource9 ∗∗ppResources)

Map Direct3D resources for access by CUDA.

• cudaError_t cudaD3D9RegisterResource (IDirect3DResource9 ∗pResource, unsigned int flags)

Registers a Direct3D resource for access by CUDA.

• cudaError_t cudaD3D9ResourceGetMappedArray (cudaArray ∗∗ppArray, IDirect3DResource9 ∗pResource,

unsigned int face, unsigned int level)
Get an array through which to access a subresource of a Direct3D resource which has been mapped for access by CUDA.

• cudaError_t cudaD3D9ResourceGetMappedPitch (size_t ∗pPitch, size_t ∗pPitchSlice, IDirect3DResource9

∗pResource, unsigned int face, unsigned int level)
Get the pitch of a subresource of a Direct3D resource which has been mapped for access by CUDA.

• cudaError_t cudaD3D9ResourceGetMappedPointer (void ∗∗pPointer, IDirect3DResource9 ∗pResource, un-

signed int face, unsigned int level)
Get a pointer through which to access a subresource of a Direct3D resource which has been mapped for access by
CUDA.

• cudaError_t cudaD3D9ResourceGetMappedSize (size_t ∗pSize, IDirect3DResource9 ∗pResource, unsigned int

face, unsigned int level)
Get the size of a subresource of a Direct3D resource which has been mapped for access by CUDA.

• cudaError_t cudaD3D9ResourceGetSurfaceDimensions (size_t ∗pWidth, size_t ∗pHeight, size_t ∗pDepth, IDi-

rect3DResource9 ∗pResource, unsigned int face, unsigned int level)
Get the dimensions of a registered Direct3D surface.

Generated for NVIDIA CUDA Library by Doxygen

58 Module Documentation

• cudaError_t cudaD3D9ResourceSetMapFlags (IDirect3DResource9 ∗pResource, unsigned int flags)

Set usage flags for mapping a Direct3D resource.

• cudaError_t cudaD3D9SetDirect3DDevice (IDirect3DDevice9 ∗pDxDevice)

Sets the Direct3D device to use for interoperability in this thread.

• cudaError_t cudaD3D9UnmapResources (int count, IDirect3DResource9 ∗∗ppResources)

Unmap Direct3D resources for access by CUDA.

• cudaError_t cudaD3D9UnregisterResource (IDirect3DResource9 ∗pResource)

Unregisters a Direct3D resource for access by CUDA.

3.10.1 Detailed Description

This section describes the Direct3D 9 interoperability functions of the CUDA runtime application programming inter-
face.

3.10.2 Enumeration Type Documentation

3.10.2.1 enum cudaD3D9MapFlags

CUDA D3D9 Map Flags

Enumerator:
cudaD3D9MapFlagsNone Default; Assume resource can be read/written.
cudaD3D9MapFlagsReadOnly CUDA kernels will not write to this resource.
cudaD3D9MapFlagsWriteDiscard CUDA kernels will only write to and will not read from this resource.

3.10.2.2 enum cudaD3D9RegisterFlags

CUDA D3D9 Register Flags

Enumerator:
cudaD3D9RegisterFlagsNone Default; Resource can be accessed througa void∗.
cudaD3D9RegisterFlagsArray Resource can be accessed through a CUarray∗.

3.10.3 Function Documentation

3.10.3.1 cudaError_t cudaD3D9GetDevice (int ∗ device, const char ∗ pszAdapterName)

Returns in ∗device the CUDA-compatible device corresponding to the adapter name pszAdapterName ob-
tained from EnumDisplayDevices or IDirect3D9::GetAdapterIdentifier(). If no device on the adapter with name
pszAdapterName is CUDA-compatible then the call will fail.

Parameters:
device - Returns the device corresponding to pszAdapterName

Generated for NVIDIA CUDA Library by Doxygen

3.10 Direct3D 9 Interoperability 59

pszAdapterName - D3D9 adapter to get device for

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorUnknown

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaD3D9SetDirect3DDevice, cudaD3D9GetDirect3DDevice, cudaD3D9RegisterResource,
cudaD3D9UnregisterResource, cudaD3D9MapResources, cudaD3D9UnmapResources,
cudaD3D9ResourceGetSurfaceDimensions, cudaD3D9ResourceSetMapFlags, cu-
daD3D9ResourceGetMappedArray, cudaD3D9ResourceGetMappedPointer, cu-
daD3D9ResourceGetMappedSize, cudaD3D9ResourceGetMappedPitch

3.10.3.2 cudaError_t cudaD3D9GetDirect3DDevice (IDirect3DDevice9 ∗∗ ppDxDevice)

Returns in ∗ppDxDevice the Direct3D device against which this CUDA context was created in cu-
daD3D9SetDirect3DDevice().

Parameters:
ppDxDevice - Returns the Direct3D device for this thread

Returns:
cudaSuccess, cudaErrorUnknown

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaD3D9GetDevice, cudaD3D9SetDirect3DDevice, cudaD3D9RegisterResource, cu-
daD3D9UnregisterResource, cudaD3D9MapResources, cudaD3D9UnmapResources,
cudaD3D9ResourceGetSurfaceDimensions, cudaD3D9ResourceSetMapFlags, cu-
daD3D9ResourceGetMappedArray, cudaD3D9ResourceGetMappedPointer, cu-
daD3D9ResourceGetMappedSize, cudaD3D9ResourceGetMappedPitch

3.10.3.3 cudaError_t cudaD3D9MapResources (int count, IDirect3DResource9 ∗∗ ppResources)

Maps the count Direct3D resources in ppResources for access by CUDA.

The resources in ppResources may be accessed in CUDA kernels until they are unmapped. Direct3D should not
access any resources while they are mapped by CUDA. If an application does so, the results are undefined.
This function provides the synchronization guarantee that any Direct3D calls issued before cudaD3D9MapResources()
will complete before any CUDA kernels issued after cudaD3D9MapResources() begin.
If any of ppResources have not been registered for use with CUDA or if ppResources contains any duplicate
entries then cudaErrorInvalidResourceHandle is returned. If any of ppResources are presently mapped for access
by CUDA then cudaErrorUnknown is returned.

Generated for NVIDIA CUDA Library by Doxygen

60 Module Documentation

Parameters:
count - Number of resources to map for CUDA
ppResources - Resources to map for CUDA

Returns:
cudaSuccess, cudaErrorInvalidResourceHandle, cudaErrorUnknown

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaD3D9GetDevice, cudaD3D9SetDirect3DDevice, cudaD3D9GetDirect3DDevice, cu-
daD3D9RegisterResource, cudaD3D9UnregisterResource, cudaD3D9UnmapResources,
cudaD3D9ResourceGetSurfaceDimensions, cudaD3D9ResourceSetMapFlags, cu-
daD3D9ResourceGetMappedArray, cudaD3D9ResourceGetMappedPointer, cu-
daD3D9ResourceGetMappedSize, cudaD3D9ResourceGetMappedPitch

3.10.3.4 cudaError_t cudaD3D9RegisterResource (IDirect3DResource9 ∗ pResource, unsigned int flags)

Registers the Direct3D resource pResource for access by CUDA.

If this call is successful, then the application will be able to map and unmap this resource until it is unregistered through
cudaD3D9UnregisterResource(). Also on success, this call will increase the internal reference count on pResource.
This reference count will be decremented when this resource is unregistered through cudaD3D9UnregisterResource().
This call is potentially high-overhead and should not be called every frame in interactive applications.
The type of pResource must be one of the following.

• IDirect3DVertexBuffer9: No notes.
• IDirect3DIndexBuffer9: No notes.
• IDirect3DSurface9: Only stand-alone objects of type IDirect3DSurface9 may be explicitly shared. In particular,
individual mipmap levels and faces of cube maps may not be registered directly. To access individual surfaces
associated with a texture, one must register the base texture object.
• IDirect3DBaseTexture9: When a texture is registered, all surfaces associated with all mipmap levels of all faces
of the texture will be accessible to CUDA.

The flags argument specifies the mechanism through which CUDA will access the Direct3D resource. The following
value is allowed:

• cudaD3D9RegisterFlagsNone: Specifies that CUDA will access this resource through a

void∗. The pointer, size, and pitch for each subresource of this resource may be queried
through cudaD3D9ResourceGetMappedPointer(), cudaD3D9ResourceGetMappedSize(), and cu-
daD3D9ResourceGetMappedPitch() respectively. This option is valid for all resource types.

Not all Direct3D resources of the above types may be used for interoperability with CUDA. The following are some
limitations:

• The primary rendertarget may not be registered with CUDA.

Generated for NVIDIA CUDA Library by Doxygen

3.10 Direct3D 9 Interoperability 61

• Resources allocated as shared may not be registered with CUDA.

• Any resources allocated in D3DPOOL_SYSTEMMEM or D3DPOOL_MANAGED may not be registered with

CUDA.

• Textures which are not of a format which is 1, 2, or 4 channels of 8, 16, or 32-bit integer or floating-point data
cannot be shared.

• Surfaces of depth or stencil formats cannot be shared.

If Direct3D interoperability is not initialized on this context, then cudaErrorInvalidDevice is returned. If pResource
is of incorrect type (e.g, is a non-stand-alone IDirect3DSurface9) or is already registered, then cudaErrorIn-
validResourceHandle is returned. If pResource cannot be registered then cudaErrorUnknown is returned.

Parameters:
pResource - Resource to register
flags - Parameters for resource registration

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidResourceHandle, cudaErrorUnknown

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaD3D9GetDevice, cudaD3D9SetDirect3DDevice, cudaD3D9GetDirect3DDevice, cu-
daD3D9UnregisterResource, cudaD3D9MapResources, cudaD3D9UnmapResources,
cudaD3D9ResourceGetSurfaceDimensions, cudaD3D9ResourceSetMapFlags, cu-
daD3D9ResourceGetMappedArray, cudaD3D9ResourceGetMappedPointer, cu-
daD3D9ResourceGetMappedSize, cudaD3D9ResourceGetMappedPitch

3.10.3.5 cudaError_t cudaD3D9ResourceGetMappedArray (cudaArray ∗∗ ppArray, IDirect3DResource9 ∗

pResource, unsigned int face, unsigned int level)

Returns in ∗pArray an array through which the subresource of the mapped Direct3D resource pResource,
which corresponds to face and level may be accessed. The value set in pArray may change every time that
pResource is mapped.
If pResource is not registered then cudaErrorInvalidResourceHandle is returned. If pResource was not registered
with usage flags cudaD3D9RegisterFlagsArray, then cudaErrorInvalidResourceHandle is returned. If pResource is
not mapped, then cudaErrorUnknown is returned.
For usage requirements of face and level parameters, see cudaD3D9ResourceGetMappedPointer().

Parameters:
ppArray - Returned array corresponding to subresource
pResource - Mapped resource to access
face - Face of resource to access
level - Level of resource to access

Generated for NVIDIA CUDA Library by Doxygen

62 Module Documentation

Returns:
cudaSuccess, cudaErrorInvalidResourceHandle, cudaErrorUnknown

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaD3D9GetDevice, cudaD3D9SetDirect3DDevice, cudaD3D9GetDirect3DDevice, cu-
daD3D9RegisterResource, cudaD3D9UnregisterResource, cudaD3D9MapResources, cu-
daD3D9UnmapResources, cudaD3D9ResourceGetSurfaceDimensions, cudaD3D9ResourceSetMapFlags, cu-
daD3D9ResourceGetMappedPointer, cudaD3D9ResourceGetMappedSize, cudaD3D9ResourceGetMappedPitch

3.10.3.6 cudaError_t cudaD3D9ResourceGetMappedPitch (size_t ∗ pPitch, size_t ∗ pPitchSlice,

IDirect3DResource9 ∗ pResource, unsigned int face, unsigned int level)

Returns in ∗pPitch and ∗pPitchSlice the pitch and Z-slice pitch of the subresource of the mapped Direct3D
resource pResource, which corresponds to face and level. The values set in pPitch and pPitchSlice may
change every time that pResource is mapped.
The pitch and Z-slice pitch values may be used to compute the location of a sample on a surface as follows.
For a 2D surface, the byte offset of the sample at position x, y from the base pointer of the surface is:
y ∗ pitch + (bytes per pixel) ∗ x
For a 3D surface, the byte offset of the sample at position x, y, z from the base pointer of the surface is:
z∗ slicePitch + y ∗ pitch + (bytes per pixel) ∗ x
Both parameters pPitch and pPitchSlice are optional and may be set to NULL.
If pResource is not of type IDirect3DBaseTexture9 or one of its sub-types or if pResource has not been registered
for use with CUDA, then cudaErrorInvalidResourceHandle is returned. If pResource was not registered with usage
flags cudaD3D9RegisterFlagsNone, then cudaErrorInvalidResourceHandle is returned. If pResource is not mapped
for access by CUDA then cudaErrorUnknown is returned.
For usage requirements of face and level parameters, see cudaD3D9ResourceGetMappedPointer().

Parameters:
pPitch - Returned pitch of subresource
pPitchSlice - Returned Z-slice pitch of subresource
pResource - Mapped resource to access
face - Face of resource to access
level - Level of resource to access

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidResourceHandle, cudaErrorUnknown

Note:
Note that this function may also return error codes from previous, asynchronous launches.

Generated for NVIDIA CUDA Library by Doxygen

3.10 Direct3D 9 Interoperability 63

See also:
cudaD3D9GetDevice, cudaD3D9SetDirect3DDevice, cudaD3D9GetDirect3DDevice, cu-
daD3D9RegisterResource, cudaD3D9UnregisterResource, cudaD3D9MapResources, cu-
daD3D9UnmapResources, cudaD3D9ResourceGetSurfaceDimensions, cudaD3D9ResourceSetMapFlags,
cudaD3D9ResourceGetMappedArray, cudaD3D9ResourceGetMappedPointer, cu-
daD3D9ResourceGetMappedSize

3.10.3.7 cudaError_t cudaD3D9ResourceGetMappedPointer (void ∗∗ pPointer, IDirect3DResource9 ∗

pResource, unsigned int face, unsigned int level)

Returns in ∗pPointer the base pointer of the subresource of the mapped Direct3D resource pResource, which
corresponds to face and level. The value set in pPointer may change every time that pResource is mapped.
If pResource is not registered, then cudaErrorInvalidResourceHandle is returned. If pResource was not registered
with usage flags cudaD3D9RegisterFlagsNone, then cudaErrorInvalidResourceHandle is returned. If pResource is
not mapped, then cudaErrorUnknown is returned.
If pResource is of type IDirect3DCubeTexture9, then face must one of the values enumerated by type
D3DCUBEMAP_FACES. For all other types, face must be 0. If face is invalid, then cudaErrorInvalidValue is
returned.
If pResource is of type IDirect3DBaseTexture9, then level must correspond to a valid mipmap level. Only
mipmap level 0 is supported for now. For all other types level must be 0. If level is invalid, then cudaErrorIn-
validValue is returned.

Parameters:
pPointer - Returned pointer corresponding to subresource
pResource - Mapped resource to access
face - Face of resource to access
level - Level of resource to access

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidResourceHandle, cudaErrorUnknown

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaD3D9GetDevice, cudaD3D9SetDirect3DDevice, cudaD3D9GetDirect3DDevice, cu-
daD3D9RegisterResource, cudaD3D9UnregisterResource, cudaD3D9MapResources, cu-
daD3D9UnmapResources, cudaD3D9ResourceGetSurfaceDimensions, cudaD3D9ResourceSetMapFlags, cu-
daD3D9ResourceGetMappedArray, cudaD3D9ResourceGetMappedSize, cudaD3D9ResourceGetMappedPitch

3.10.3.8 cudaError_t cudaD3D9ResourceGetMappedSize (size_t ∗ pSize, IDirect3DResource9 ∗ pResource,

unsigned int face, unsigned int level)

Returns in ∗pSize the size of the subresource of the mapped Direct3D resource pResource, which corresponds to
face and level. The value set in pSize may change every time that pResource is mapped.

Generated for NVIDIA CUDA Library by Doxygen

64 Module Documentation

If pResource has not been registered for use with CUDA then cudaErrorInvalidResourceHandle is returned. If
pResource was not registered with usage flags cudaD3D9RegisterFlagsNone, then cudaErrorInvalidResourceHan-
dle is returned. If pResource is not mapped for access by CUDA then cudaErrorUnknown is returned.
For usage requirements of face and level parameters, see cudaD3D9ResourceGetMappedPointer().

Parameters:
pSize - Returned size of subresource
pResource - Mapped resource to access
face - Face of resource to access
level - Level of resource to access

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidResourceHandle, cudaErrorUnknown

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaD3D9GetDevice, cudaD3D9SetDirect3DDevice, cudaD3D9GetDirect3DDevice, cu-
daD3D9RegisterResource, cudaD3D9UnregisterResource, cudaD3D9MapResources, cu-
daD3D9UnmapResources, cudaD3D9ResourceGetSurfaceDimensions, cudaD3D9ResourceSetMapFlags,
cudaD3D9ResourceGetMappedArray, cudaD3D9ResourceGetMappedPointer, cu-
daD3D9ResourceGetMappedPitch

3.10.3.9 cudaError_t cudaD3D9ResourceGetSurfaceDimensions (size_t ∗ pWidth, size_t ∗ pHeight, size_t ∗

pDepth, IDirect3DResource9 ∗ pResource, unsigned int face, unsigned int level)

Returns in ∗pWidth, ∗pHeight, and ∗pDepth the dimensions of the subresource of the mapped Direct3D resource
pResource which corresponds to face and level.
Because anti-aliased surfaces may have multiple samples per pixel, it is possible that the dimensions of a resource will
be an integer factor larger than the dimensions reported by the Direct3D runtime.
The parameters pWidth, pHeight, and pDepth are optional. For 2D surfaces, the value returned in ∗pDepth will
be 0.
If pResource is not of type IDirect3DBaseTexture9 or IDirect3DSurface9 or if pResource has not been registered
for use with CUDA, then cudaErrorInvalidResourceHandle is returned.
For usage requirements of face and level parameters, see cudaD3D9ResourceGetMappedPointer.

Parameters:
pWidth - Returned width of surface
pHeight - Returned height of surface
pDepth - Returned depth of surface
pResource - Registered resource to access
face - Face of resource to access
level - Level of resource to access

Generated for NVIDIA CUDA Library by Doxygen

3.10 Direct3D 9 Interoperability 65

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidResourceHandle,

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaD3D9GetDevice, cudaD3D9SetDirect3DDevice, cudaD3D9GetDirect3DDevice, cu-
daD3D9RegisterResource, cudaD3D9UnregisterResource, cudaD3D9MapResources, cu-
daD3D9UnmapResources, cudaD3D9ResourceSetMapFlags, cudaD3D9ResourceGetMappedArray, cu-
daD3D9ResourceGetMappedPointer, cudaD3D9ResourceGetMappedSize, cudaD3D9ResourceGetMappedPitch

3.10.3.10 cudaError_t cudaD3D9ResourceSetMapFlags (IDirect3DResource9 ∗ pResource, unsigned int

flags)

Set flags for mapping the Direct3D resource pResource.

Changes to flags will take effect the next time pResource is mapped. The flags argument may be any of the
following:

• cudaD3D9MapFlagsNone: Specifies no hints about how this resource will be used. It is therefore assumed that
this resource will be read from and written to by CUDA kernels. This is the default value.
• cudaD3D9MapFlagsReadOnly: Specifies that CUDA kernels which access this resource will not write to this
resource.
• cudaD3D9MapFlagsWriteDiscard: Specifies that CUDA kernels which access this resource will not read from
this resource and will write over the entire contents of the resource, so none of the data previously stored in the
resource will be preserved.

If pResource has not been registered for use with CUDA, then cudaErrorInvalidResourceHandle is returned. If
pResource is presently mapped for access by CUDA, then cudaErrorUnknown is returned.

Parameters:
pResource - Registered resource to set flags for
flags - Parameters for resource mapping

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidResourceHandle, cudaErrorUnknown

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaD3D9GetDevice, cudaD3D9SetDirect3DDevice, cudaD3D9GetDirect3DDevice, cu-
daD3D9RegisterResource, cudaD3D9UnregisterResource, cudaD3D9MapResources, cu-
daD3D9UnmapResources, cudaD3D9ResourceGetSurfaceDimensions, cudaD3D9ResourceGetMappedArray,
cudaD3D9ResourceGetMappedPointer, cudaD3D9ResourceGetMappedSize, cu-
daD3D9ResourceGetMappedPitch

Generated for NVIDIA CUDA Library by Doxygen

66 Module Documentation

3.10.3.11 cudaError_t cudaD3D9SetDirect3DDevice (IDirect3DDevice9 ∗ pDxDevice)

Records pDxDevice as the Direct3D device to use for Direct3D interoperability on this host thread. In order to
use Direct3D interoperability, this call must be made before any non-device management CUDA runtime calls on this
thread. In that case, this call will return cudaErrorSetOnActiveProcess.
Successful context creation on pDxDevice will increase the internal reference count on pDxDevice. This reference
count will be decremented upon destruction of this context through cudaThreadExit().

Parameters:
pDxDevice - Direct3D device to use for this thread

Returns:
cudaSuccess, cudaErrorInitializationError, cudaErrorInvalidValue, cudaErrorSetOnActiveProcess

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaD3D9GetDevice, cudaD3D9GetDirect3DDevice, cudaD3D9RegisterResource, cu-
daD3D9UnregisterResource, cudaD3D9MapResources, cudaD3D9UnmapResources,
cudaD3D9ResourceGetSurfaceDimensions, cudaD3D9ResourceSetMapFlags, cu-
daD3D9ResourceGetMappedArray, cudaD3D9ResourceGetMappedPointer, cu-
daD3D9ResourceGetMappedSize, cudaD3D9ResourceGetMappedPitch

3.10.3.12 cudaError_t cudaD3D9UnmapResources (int count, IDirect3DResource9 ∗∗ ppResources)

Unmaps the count Direct3D resources in ppResources.

This function provides the synchronization guarantee that any CUDA kernels issued before cu-
daD3D9UnmapResources() will complete before any Direct3D calls issued after cudaD3D9UnmapResources()
begin.
If any of ppResources have not been registered for use with CUDA or if ppResources contains any duplicate
entries, then cudaErrorInvalidResourceHandle is returned. If any of ppResources are not presently mapped for
access by CUDA then cudaErrorUnknown is returned.

Parameters:
count - Number of resources to unmap for CUDA
ppResources - Resources to unmap for CUDA

Returns:
cudaSuccess, cudaErrorInvalidResourceHandle, cudaErrorUnknown

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaD3D9GetDevice, cudaD3D9SetDirect3DDevice, cudaD3D9GetDirect3DDevice, cu-
daD3D9RegisterResource, cudaD3D9UnregisterResource, cudaD3D9MapResources,

Generated for NVIDIA CUDA Library by Doxygen

3.10 Direct3D 9 Interoperability 67

cudaD3D9ResourceGetSurfaceDimensions, cudaD3D9ResourceSetMapFlags, cu-

daD3D9ResourceGetMappedArray, cudaD3D9ResourceGetMappedPointer, cu-
daD3D9ResourceGetMappedSize, cudaD3D9ResourceGetMappedPitch

3.10.3.13 cudaError_t cudaD3D9UnregisterResource (IDirect3DResource9 ∗ pResource)

Unregisters the Direct3D resource pResource so it is not accessible by CUDA unless registered again.
If pResource is not registered, then cudaErrorInvalidResourceHandle is returned.

Parameters:
pResource - Resource to unregister

Returns:
cudaSuccess, cudaErrorInvalidResourceHandle, cudaErrorUnknown

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaD3D9GetDevice, cudaD3D9SetDirect3DDevice, cudaD3D9GetDirect3DDevice, cu-
daD3D9RegisterResource, cudaD3D9MapResources, cudaD3D9UnmapResources,
cudaD3D9ResourceGetSurfaceDimensions, cudaD3D9ResourceSetMapFlags, cu-
daD3D9ResourceGetMappedArray, cudaD3D9ResourceGetMappedPointer, cu-
daD3D9ResourceGetMappedSize, cudaD3D9ResourceGetMappedPitch

Generated for NVIDIA CUDA Library by Doxygen

68 Module Documentation

3.11 Direct3D 10 Interoperability

Direct3D 10 Interoperability with CUDA Runtime

Direct3D 10 Interoperability with CUDA Runtime

Author:
NVIDIA Corporation

• enum cudaD3D10MapFlags {
cudaD3D10MapFlagsNone,
cudaD3D10MapFlagsReadOnly,
cudaD3D10MapFlagsWriteDiscard }
• enum cudaD3D10RegisterFlags {
cudaD3D10RegisterFlagsNone,
cudaD3D10RegisterFlagsArray }
• cudaError_t cudaD3D10GetDevice (int ∗device, IDXGIAdapter ∗pAdapter)
Gets the device number for an adapter.

• cudaError_t cudaD3D10MapResources (int count, ID3D10Resource ∗∗ppResources)

Map Direct3D Resources for access by CUDA.

• cudaError_t cudaD3D10RegisterResource (ID3D10Resource ∗pResource, unsigned int flags)

Register a Direct3D resource for access by CUDA.

• cudaError_t cudaD3D10ResourceGetMappedArray (cudaArray ∗∗ppArray, ID3D10Resource ∗pResource, un-

signed int subResource)
Get an array through which to access a subresource of a Direct3D resource which has been mapped for access by CUDA.

• cudaError_t cudaD3D10ResourceGetMappedPitch (size_t ∗pPitch, size_t ∗pPitchSlice, ID3D10Resource

∗pResource, unsigned int subResource)
Get the pitch of a subresource of a Direct3D resource which has been mapped for access by CUDA.

• cudaError_t cudaD3D10ResourceGetMappedPointer (void ∗∗pPointer, ID3D10Resource ∗pResource, unsigned

int subResource)
Get a pointer through which to access a subresource of a Direct3D resource which has been mapped for access by
CUDA.

• cudaError_t cudaD3D10ResourceGetMappedSize (size_t ∗pSize, ID3D10Resource ∗pResource, unsigned int

subResource)
Get the size of a subresource of a Direct3D resource which has been mapped for access by CUDA.

• cudaError_t cudaD3D10ResourceGetSurfaceDimensions (size_t ∗pWidth, size_t ∗pHeight, size_t ∗pDepth,

ID3D10Resource ∗pResource, unsigned int subResource)
Get the dimensions of a registered Direct3D surface.

• cudaError_t cudaD3D10ResourceSetMapFlags (ID3D10Resource ∗pResource, unsigned int flags)

Set usage flags for mapping a Direct3D resource.

Generated for NVIDIA CUDA Library by Doxygen

3.11 Direct3D 10 Interoperability 69

• cudaError_t cudaD3D10SetDirect3DDevice (ID3D10Device ∗pDxDevice)

Sets the Direct3D device to use for interoperability in this thread.

• cudaError_t cudaD3D10UnmapResources (int count, ID3D10Resource ∗∗ppResources)

Unmaps Direct3D resources.

• cudaError_t cudaD3D10UnregisterResource (ID3D10Resource ∗pResource)

Unregisters a Direct3D resource.

3.11.1 Detailed Description

This section describes the Direct3D 10 interoperability functions of the CUDA runtime application programming
interface.

3.11.2 Enumeration Type Documentation

3.11.2.1 enum cudaD3D10MapFlags

CUDA D3D10 Map Flags

Enumerator:
cudaD3D10MapFlagsNone Default; Assume resource can be read/written.
cudaD3D10MapFlagsReadOnly CUDA kernels will not write to this resource.
cudaD3D10MapFlagsWriteDiscard CUDA kernels will only write to and will not read from this resource.

3.11.2.2 enum cudaD3D10RegisterFlags

CUDA D3D10 Register Flags

Enumerator:
cudaD3D10RegisterFlagsNone Default; Resource can be accessed through a void∗.
cudaD3D10RegisterFlagsArray Resource can be accessed through a CUarray∗.

3.11.3 Function Documentation

3.11.3.1 cudaError_t cudaD3D10GetDevice (int ∗ device, IDXGIAdapter ∗ pAdapter)

Returns in ∗device the CUDA-compatible device corresponding to the adapter pAdapter obtained from IDXGI-
Factory::EnumAdapters. This call will succeed only if a device on adapter pAdapter is Cuda-compatible.

Parameters:
device - Returns the device corresponding to pAdapter
pAdapter - D3D10 adapter to get device for

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorUnknown

Generated for NVIDIA CUDA Library by Doxygen

70 Module Documentation

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaD3D10SetDirect3DDevice, cudaD3D10RegisterResource, cudaD3D10UnregisterResource, cu-
daD3D10MapResources, cudaD3D10UnmapResources, cudaD3D10ResourceGetSurfaceDimensions,
cudaD3D10ResourceSetMapFlags, cudaD3D10ResourceGetMappedArray, cu-
daD3D10ResourceGetMappedPointer, cudaD3D10ResourceGetMappedSize, cu-
daD3D10ResourceGetMappedPitch

3.11.3.2 cudaError_t cudaD3D10MapResources (int count, ID3D10Resource ∗∗ ppResources)

Maps the count Direct3D resources in ppResources for access by CUDA.

The resources in ppResources may be accessed in CUDA kernels until they are unmapped. Direct3D should not
access any resources while they are mapped by CUDA. If an application does so, the results are undefined.
This function provides the synchronization guarantee that any Direct3D calls issued before cu-
daD3D10MapResources() will complete before any CUDA kernels issued after cudaD3D10MapResources()
begin.
If any of ppResources have not been registered for use with CUDA or if ppResources contains any duplicate
entries then cudaErrorInvalidResourceHandle is returned. If any of ppResources are presently mapped for access
by CUDA then cudaErrorUnknown is returned.

Parameters:
count - Number of resources to map for CUDA
ppResources - Resources to map for CUDA

Returns:
cudaSuccess, cudaErrorInvalidResourceHandle, cudaErrorUnknown

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaD3D10GetDevice, cudaD3D10SetDirect3DDevice, cudaD3D10RegisterResource, cu-
daD3D10UnregisterResource, cudaD3D10UnmapResources, cudaD3D10ResourceGetSurfaceDimensions,
cudaD3D10ResourceSetMapFlags, cudaD3D10ResourceGetMappedArray, cu-
daD3D10ResourceGetMappedPointer, cudaD3D10ResourceGetMappedSize, cu-
daD3D10ResourceGetMappedPitch

3.11.3.3 cudaError_t cudaD3D10RegisterResource (ID3D10Resource ∗ pResource, unsigned int flags)

Registers the Direct3D resource pResource for access by CUDA.

If this call is successful, then the application will be able to map and unmap this resource until it is unreg-
istered through cudaD3D10UnregisterResource(). Also on success, this call will increase the internal reference
count on pResource. This reference count will be decremented when this resource is unregistered through cu-
daD3D10UnregisterResource().

Generated for NVIDIA CUDA Library by Doxygen

3.11 Direct3D 10 Interoperability 71

This call is potentially high-overhead and should not be called every frame in interactive applications.
The type of pResource must be one of the following:

• ID3D10Buffer: Cannot be used with flags set to cudaD3D10RegisterFlagsArray.

• ID3D10Texture1D: No restrictions.
• ID3D10Texture2D: No restrictions.
• ID3D10Texture3D: No restrictions.

The flags argument specifies the mechanism through which CUDA will access the Direct3D resource. The following
values are allowed.

• cudaD3D10RegisterFlagsNone: Specifies that CUDA will access this resource through a

void∗. The pointer, size, and pitch for each subresource of this resource may be queried
through cudaD3D10ResourceGetMappedPointer(), cudaD3D10ResourceGetMappedSize(), and cu-
daD3D10ResourceGetMappedPitch() respectively. This option is valid for all resource types.
• cudaD3D10RegisterFlagsArray: Specifies that CUDA will access this resource through a CUarray queried on
a sub-resource basis through cuD3D10ResourceGetMappedArray(). This option is only valid for resources of
type ID3D10Texture1D, ID3D10Texture2D, and ID3D10Texture3D.

Not all Direct3D resources of the above types may be used for interoperability with CUDA. The following are some
limitations.

• The primary rendertarget may not be registered with CUDA.

• Resources allocated as shared may not be registered with CUDA.
• Textures which are not of a format which is 1, 2, or 4 channels of 8, 16, or 32-bit integer or floating-point data
cannot be shared.
• Surfaces of depth or stencil formats cannot be shared.

If Direct3D interoperability is not initialized on this context then cudaErrorInvalidDevice is returned. If pResource
is of incorrect type or is already registered then cudaErrorInvalidResourceHandle is returned. If pResource cannot
be registered then cudaErrorUnknown is returned.

Parameters:
pResource - Resource to register
flags - Parameters for resource registration

Returns:
cudaSuccess, cudaErrorInvalidDevice, cudaErrorInvalidValue, cudaErrorInvalidResourceHandle, cudaErrorUnk-
nown
Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaD3D10GetDevice, cudaD3D10SetDirect3DDevice, cudaD3D10UnregisterResource, cu-
daD3D10MapResources, cudaD3D10UnmapResources, cudaD3D10ResourceGetSurfaceDimensions,
cudaD3D10ResourceSetMapFlags, cudaD3D10ResourceGetMappedArray, cu-
daD3D10ResourceGetMappedPointer, cudaD3D10ResourceGetMappedSize, cu-
daD3D10ResourceGetMappedPitch

Generated for NVIDIA CUDA Library by Doxygen

72 Module Documentation

3.11.3.4 cudaError_t cudaD3D10ResourceGetMappedArray (cudaArray ∗∗ ppArray, ID3D10Resource ∗

pResource, unsigned int subResource)

Returns in ∗ppArray an array through which the subresource of the mapped Direct3D resource pResource which
corresponds to subResource may be accessed. The value set in ppArray may change every time that pResource
is mapped.
If pResource is not registered, then cudaErrorInvalidResourceHandle is returned. If pResource was not registered
with usage flags cudaD3D10RegisterFlagsArray, then cudaErrorInvalidResourceHandle is returned. If pResource
is not mapped then cudaErrorUnknown is returned.
For usage requirements of the subResource parameter, see cudaD3D10ResourceGetMappedPointer().

Parameters:
ppArray - Returned array corresponding to subresource
pResource - Mapped resource to access
subResource - Subresource of pResource to access

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidResourceHandle, cudaErrorUnknown

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaD3D10GetDevice, cudaD3D10SetDirect3DDevice, cudaD3D10RegisterResource, cu-
daD3D10UnregisterResource, cudaD3D10MapResources, cudaD3D10UnmapResources,
cudaD3D10ResourceGetSurfaceDimensions, cudaD3D10ResourceSetMapFlags, cu-
daD3D10ResourceGetMappedPointer, cudaD3D10ResourceGetMappedSize, cu-
daD3D10ResourceGetMappedPitch

3.11.3.5 cudaError_t cudaD3D10ResourceGetMappedPitch (size_t ∗ pPitch, size_t ∗ pPitchSlice,

ID3D10Resource ∗ pResource, unsigned int subResource)

Returns in ∗pPitch and ∗pPitchSlice the pitch and Z-slice pitch of the subresource of the mapped Direct3D
resource pResource, which corresponds to subResource. The values set in pPitch and pPitchSlice may
change every time that pResource is mapped.
The pitch and Z-slice pitch values may be used to compute the location of a sample on a surface as follows.
For a 2D surface, the byte offset of the sample at position x, y from the base pointer of the surface is:
y ∗ pitch + (bytes per pixel) ∗ x
For a 3D surface, the byte offset of the sample at position x, y, z from the base pointer of the surface is:
z∗ slicePitch + y ∗ pitch + (bytes per pixel) ∗ x
Both parameters pPitch and pPitchSlice are optional and may be set to NULL.
If pResource is not of type ID3D10Texture1D, ID3D10Texture2D, or ID3D10Texture3D, or if pResource has
not been registered for use with CUDA, then cudaErrorInvalidResourceHandle is returned. If pResource was
not registered with usage flags cudaD3D10RegisterFlagsNone, then cudaErrorInvalidResourceHandle is returned. If
pResource is not mapped for access by CUDA then cudaErrorUnknown is returned.
For usage requirements of the subResource parameter see cudaD3D10ResourceGetMappedPointer().

Generated for NVIDIA CUDA Library by Doxygen

3.11 Direct3D 10 Interoperability 73

Parameters:
pPitch - Returned pitch of subresource
pPitchSlice - Returned Z-slice pitch of subresource
pResource - Mapped resource to access
subResource - Subresource of pResource to access

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidResourceHandle, cudaErrorUnknown

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaD3D10GetDevice, cudaD3D10SetDirect3DDevice, cudaD3D10RegisterResource, cu-
daD3D10UnregisterResource, cudaD3D10MapResources, cudaD3D10UnmapResources,
cudaD3D10ResourceGetSurfaceDimensions, cudaD3D10ResourceSetMapFlags, cu-
daD3D10ResourceGetMappedArray, cudaD3D10ResourceGetMappedPointer, cu-
daD3D10ResourceGetMappedSize

3.11.3.6 cudaError_t cudaD3D10ResourceGetMappedPointer (void ∗∗ pPointer, ID3D10Resource ∗

pResource, unsigned int subResource)

Returns in ∗pPointer the base pointer of the subresource of the mapped Direct3D resource pResource which
corresponds to subResource. The value set in pPointer may change every time that pResource is mapped.
If pResource is not registered, then cudaErrorInvalidResourceHandle is returned. If pResource was not registered
with usage flags cudaD3D9RegisterFlagsNone, then cudaErrorInvalidResourceHandle is returned. If pResource is
not mapped then cudaErrorUnknown is returned.
If pResource is of type ID3D10Buffer then subResource must be 0. If pResource is of any other type, then
the value of subResource must come from the subresource calculation in D3D10CalcSubResource().

Parameters:
pPointer - Returned pointer corresponding to subresource
pResource - Mapped resource to access
subResource - Subresource of pResource to access

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidResourceHandle, cudaErrorUnknown

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaD3D10GetDevice, cudaD3D10SetDirect3DDevice, cudaD3D10RegisterResource, cu-
daD3D10UnregisterResource, cudaD3D10MapResources, cudaD3D10UnmapResources,
cudaD3D10ResourceGetSurfaceDimensions, cudaD3D10ResourceSetMapFlags, cu-
daD3D10ResourceGetMappedArray, cudaD3D10ResourceGetMappedSize, cu-
daD3D10ResourceGetMappedPitch

Generated for NVIDIA CUDA Library by Doxygen

74 Module Documentation

3.11.3.7 cudaError_t cudaD3D10ResourceGetMappedSize (size_t ∗ pSize, ID3D10Resource ∗ pResource,

unsigned int subResource)

Returns in ∗pSize the size of the subresource of the mapped Direct3D resource pResource which corresponds to
subResource. The value set in pSize may change every time that pResource is mapped.
If pResource has not been registered for use with CUDA then cudaErrorInvalidHandle is returned. If pResource
was not registered with usage flags cudaD3D10RegisterFlagsNone, then cudaErrorInvalidHandle is returned. If
pResource is not mapped for access by CUDA then cudaErrorUnknown is returned.
For usage requirements of the subResource parameter see cudaD3D10ResourceGetMappedPointer().

Parameters:
pSize - Returned size of subresource
pResource - Mapped resource to access
subResource - Subresource of pResource to access

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidResourceHandle, cudaErrorUnknown

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaD3D10GetDevice, cudaD3D10SetDirect3DDevice, cudaD3D10RegisterResource, cu-
daD3D10UnregisterResource, cudaD3D10MapResources, cudaD3D10UnmapResources,
cudaD3D10ResourceGetSurfaceDimensions, cudaD3D10ResourceSetMapFlags, cu-
daD3D10ResourceGetMappedPointer, cudaD3D10ResourceGetMappedArray, cu-
daD3D10ResourceGetMappedPitch

3.11.3.8 cudaError_t cudaD3D10ResourceGetSurfaceDimensions (size_t ∗ pWidth, size_t ∗ pHeight, size_t ∗

pDepth, ID3D10Resource ∗ pResource, unsigned int subResource)

Returns in ∗pWidth, ∗pHeight, and ∗pDepth the dimensions of the subresource of the mapped Direct3D resource
pResource which corresponds to subResource.
Because anti-aliased surfaces may have multiple samples per pixel, it is possible that the dimensions of a resource will
be an integer factor larger than the dimensions reported by the Direct3D runtime.
The parameters pWidth, pHeight, and pDepth are optional. For 2D surfaces, the value returned in ∗pDepth will
be 0.
If pResource is not of type ID3D10Texture1D, ID3D10Texture2D, or ID3D10Texture3D, or if pResource has
not been registered for use with CUDA, then cudaErrorInvalidHandle is returned.
For usage requirements of subResource parameters see cudaD3D10ResourceGetMappedPointer().

Parameters:
pWidth - Returned width of surface
pHeight - Returned height of surface
pDepth - Returned depth of surface

Generated for NVIDIA CUDA Library by Doxygen

3.11 Direct3D 10 Interoperability 75

pResource - Registered resource to access

subResource - Subresource of pResource to access

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidResourceHandle,

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaD3D10GetDevice, cudaD3D10SetDirect3DDevice, cudaD3D10RegisterResource, cu-
daD3D10UnregisterResource, cudaD3D10MapResources, cudaD3D10UnmapResources,
cudaD3D10ResourceSetMapFlags, cudaD3D10ResourceGetMappedArray, cu-
daD3D10ResourceGetMappedPointer, cudaD3D10ResourceGetMappedSize, cu-
daD3D10ResourceGetMappedPitch

3.11.3.9 cudaError_t cudaD3D10ResourceSetMapFlags (ID3D10Resource ∗ pResource, unsigned int flags)

Set usage flags for mapping the Direct3D resource pResource.

Changes to flags will take effect the next time pResource is mapped. The flags argument may be any of the
following:

• cudaD3D10MapFlagsNone: Specifies no hints about how this resource will be used. It is therefore assumed that
this resource will be read from and written to by CUDA kernels. This is the default value.
• cudaD3D10MapFlagsReadOnly: Specifies that CUDA kernels which access this resource will not write to this
resource.
• cudaD3D10MapFlagsWriteDiscard: Specifies that CUDA kernels which access this resource will not read from
this resource and will write over the entire contents of the resource, so none of the data previously stored in the
resource will be preserved.

If pResource has not been registered for use with CUDA then cudaErrorInvalidHandle is returned. If pResource
is presently mapped for access by CUDA then cudaErrorUnknown is returned.

Parameters:
pResource - Registered resource to set flags for
flags - Parameters for resource mapping

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidResourceHandle, cudaErrorUnknown,

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaD3D10GetDevice, cudaD3D10SetDirect3DDevice, cudaD3D10RegisterResource, cu-
daD3D10UnregisterResource, cudaD3D10MapResources, cudaD3D10UnmapResources,
cudaD3D10ResourceGetSurfaceDimensions, cudaD3D10ResourceGetMappedArray, cu-
daD3D10ResourceGetMappedPointer, cudaD3D10ResourceGetMappedSize, cu-
daD3D10ResourceGetMappedPitch

Generated for NVIDIA CUDA Library by Doxygen

76 Module Documentation

3.11.3.10 cudaError_t cudaD3D10SetDirect3DDevice (ID3D10Device ∗ pDxDevice)

Records pDxDevice as the Direct3D device to use for Direct3D interoperability on this host thread. In order to
use Direct3D interoperability, this call must be made before any non-device management CUDA runtime calls on this
thread. In that case, this call will return cudaErrorSetOnActiveProcess.
Successful context creation on pDxDevice will increase the internal reference count on pDxDevice. This reference
count will be decremented upon destruction of this context through cudaThreadExit().

Parameters:
pDxDevice - Direct3D device to use for interoperability

Returns:
cudaSuccess, cudaErrorInitializationError, cudaErrorInvalidValue, cudaErrorSetOnActiveProcess

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaD3D10GetDevice, cudaD3D10RegisterResource, cudaD3D10UnregisterResource, cu-
daD3D10MapResources, cudaD3D10UnmapResources, cudaD3D10ResourceGetSurfaceDimensions,
cudaD3D10ResourceSetMapFlags, cudaD3D10ResourceGetMappedArray, cu-
daD3D10ResourceGetMappedPointer, cudaD3D10ResourceGetMappedSize, cu-
daD3D10ResourceGetMappedPitch

3.11.3.11 cudaError_t cudaD3D10UnmapResources (int count, ID3D10Resource ∗∗ ppResources)

Unmaps the count Direct3D resource in ppResources.

This function provides the synchronization guarantee that any CUDA kernels issued before cu-
daD3D10UnmapResources() will complete before any Direct3D calls issued after cudaD3D10UnmapResources()
begin.
If any of ppResources have not been registered for use with CUDA or if ppResources contains any duplicate
entries, then cudaErrorInvalidResourceHandle is returned. If any of ppResources are not presently mapped for
access by CUDA then cudaErrorUnknown is returned.

Parameters:
count - Number of resources to unmap for CUDA
ppResources - Resources to unmap for CUDA

Returns:
cudaSuccess, cudaErrorInvalidResourceHandle, cudaErrorUnknown

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaD3D10GetDevice, cudaD3D10SetDirect3DDevice, cudaD3D10RegisterResource, cu-
daD3D10UnregisterResource, cudaD3D10MapResources, cudaD3D10ResourceGetSurfaceDimensions,

Generated for NVIDIA CUDA Library by Doxygen

3.11 Direct3D 10 Interoperability 77

cudaD3D10ResourceSetMapFlags, cudaD3D10ResourceGetMappedArray, cu-

daD3D10ResourceGetMappedPointer, cudaD3D10ResourceGetMappedSize, cu-
daD3D10ResourceGetMappedPitch

3.11.3.12 cudaError_t cudaD3D10UnregisterResource (ID3D10Resource ∗ pResource)

Unregisters the Direct3D resource resource so it is not accessible by CUDA unless registered again.
If pResource is not registered, then cudaErrorInvalidResourceHandle is returned.

Parameters:
pResource - Resource to unregister

Returns:
cudaSuccess, cudaErrorInvalidResourceHandle, cudaErrorUnknown

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaD3D10GetDevice, cudaD3D10SetDirect3DDevice, cudaD3D10RegisterResource, cu-
daD3D10MapResources, cudaD3D10UnmapResources, cudaD3D10ResourceGetSurfaceDimensions,
cudaD3D10ResourceSetMapFlags, cudaD3D10ResourceGetMappedArray, cu-
daD3D10ResourceGetMappedPointer, cudaD3D10ResourceGetMappedSize, cu-
daD3D10ResourceGetMappedPitch

Generated for NVIDIA CUDA Library by Doxygen

78 Module Documentation

3.12 Texture Reference Management

Functions
• cudaError_t cudaBindTexture (size_t ∗offset, const struct textureReference ∗texref, const void ∗devPtr, const
struct cudaChannelFormatDesc ∗desc, size_t size)
Binds a memory area to a texture.

• cudaError_t cudaBindTexture2D (size_t ∗offset, const struct textureReference ∗texref, const void ∗devPtr, const
struct cudaChannelFormatDesc ∗desc, size_t width, size_t height, size_t pitch)
Binds a 2D memory area to a texture.

• cudaError_t cudaBindTextureToArray (const struct textureReference ∗texref, const struct cudaArray ∗array,
const struct cudaChannelFormatDesc ∗desc)
Binds an array to a texture.

• struct cudaChannelFormatDesc cudaCreateChannelDesc (int x, int y, int z, int w, enum cudaChannelFormatKind

f)
Returns a channel descriptor using the specified format.

• cudaError_t cudaGetChannelDesc (struct cudaChannelFormatDesc ∗desc, const struct cudaArray ∗array)

Get the channel descriptor of an array.

• cudaError_t cudaGetTextureAlignmentOffset (size_t ∗offset, const struct textureReference ∗texref)

Get the alignment offset of a texture.

• cudaError_t cudaGetTextureReference (const struct textureReference ∗∗texref, const char ∗symbol)

Get the texture reference associated with a symbol.

• cudaError_t cudaUnbindTexture (const struct textureReference ∗texref)

Unbinds a texture.

3.12.1 Detailed Description

This section describes the low level texture reference management functions of the CUDA runtime application pro-
gramming interface.

3.12.2 Function Documentation

3.12.2.1 cudaError_t cudaBindTexture (size_t ∗ offset, const struct textureReference ∗ texref, const void ∗
devPtr, const struct cudaChannelFormatDesc ∗ desc, size_t size)

Binds size bytes of the memory area pointed to by devPtr to the texture reference texref. desc describes
how the memory is interpreted when fetching values from the texture. Any memory previously bound to texref is
unbound.
Since the hardware enforces an alignment requirement on texture base addresses, cudaBindTexture() returns in
∗offset a byte offset that must be applied to texture fetches in order to read from the desired memory. This off-
set must be divided by the texel size and passed to kernels that read from the texture so they can be applied to the

Generated for NVIDIA CUDA Library by Doxygen

3.12 Texture Reference Management 79

tex1Dfetch() function. If the device memory pointer was returned from cudaMalloc(), the offset is guaranteed to be 0
and NULL may be passed as the offset parameter.

Parameters:
offset - Offset in bytes
texref - Texture to bind
devPtr - Memory area on device
desc - Channel format
size - Size of the memory area pointed to by devPtr

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidDevicePointer, cudaErrorInvalidTexture

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaCreateChannelDesc (C API), cudaGetChannelDesc, cudaGetTextureReference, cudaBindTexture (C++ API),
cudaBindTexture2D (C API), cudaBindTextureToArray (C API), cudaUnbindTexture (C API), cudaGetTexture-
AlignmentOffset (C API)

3.12.2.2 cudaError_t cudaBindTexture2D (size_t ∗ offset, const struct textureReference ∗ texref, const void
∗ devPtr, const struct cudaChannelFormatDesc ∗ desc, size_t width, size_t height, size_t pitch)

Binds the 2D memory area pointed to by devPtr to the texture reference texref. The size of the area is constrained
by width in texel units, height in texel units, and pitch in byte units. desc describes how the memory is
interpreted when fetching values from the texture. Any memory previously bound to texref is unbound.
Since the hardware enforces an alignment requirement on texture base addresses, cudaBindTexture2D() returns in
∗offset a byte offset that must be applied to texture fetches in order to read from the desired memory. This offset
must be divided by the texel size and passed to kernels that read from the texture so they can be applied to the tex2D()
function. If the device memory pointer was returned from cudaMalloc(), the offset is guaranteed to be 0 and NULL
may be passed as the offset parameter.

Parameters:
offset - Offset in bytes
texref - Texture reference to bind
devPtr - 2D memory area on device
desc - Channel format
width - Width in texel units
height - Height in texel units
pitch - Pitch in bytes

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidDevicePointer, cudaErrorInvalidTexture

Generated for NVIDIA CUDA Library by Doxygen

80 Module Documentation

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaCreateChannelDesc (C API), cudaGetChannelDesc, cudaGetTextureReference, cudaBindTexture (C API),
cudaBindTexture2D (C++ API), cudaBindTextureToArray (C API), cudaBindTextureToArray (C API), cudaGet-
TextureAlignmentOffset (C API)

3.12.2.3 cudaError_t cudaBindTextureToArray (const struct textureReference ∗ texref, const struct

cudaArray ∗ array, const struct cudaChannelFormatDesc ∗ desc)

Binds the CUDA array array to the texture reference texref. desc describes how the memory is interpreted when
fetching values from the texture. Any CUDA array previously bound to texref is unbound.

Parameters:
texref - Texture to bind
array - Memory array on device
desc - Channel format

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidDevicePointer, cudaErrorInvalidTexture

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaCreateChannelDesc (C API), cudaGetChannelDesc, cudaGetTextureReference, cudaBindTexture (C API),
cudaBindTexture2D (C API), cudaBindTextureToArray (C++ API), cudaUnbindTexture (C API), cudaGetTex-
tureAlignmentOffset (C API)

3.12.2.4 struct cudaChannelFormatDesc cudaCreateChannelDesc (int x, int y, int z, int w, enum

cudaChannelFormatKind f) [read]

Returns a channel descriptor with format f and number of bits of each component x, y, z, and w. The cudaChan-
nelFormatDesc is defined as:

struct cudaChannelFormatDesc {
int x, y, z, w;
enum cudaChannelFormatKind f;
};

where cudaChannelFormatKind is one of cudaChannelFormatKindSigned, cudaChannelFormatKindUnsigned, or cu-

daChannelFormatKindFloat.

Parameters:
x - X component
y - Y component

Generated for NVIDIA CUDA Library by Doxygen

3.12 Texture Reference Management 81

z - Z component
w - W component
f - Channel format

Returns:
Channel descriptor with format f

See also:
cudaCreateChannelDesc (C++ API), cudaGetChannelDesc, cudaGetTextureReference, cudaBindTexture (C API),
cudaBindTexture2D (C API), cudaBindTextureToArray (C API), cudaUnbindTexture (C API), cudaGetTexture-
AlignmentOffset (C API)

3.12.2.5 cudaError_t cudaGetChannelDesc (struct cudaChannelFormatDesc ∗ desc, const struct cudaArray

∗ array)

Returns in ∗desc the channel descriptor of the CUDA array array.

Parameters:
desc - Channel format
array - Memory array on device

Returns:
cudaSuccess, cudaErrorInvalidValue

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaCreateChannelDesc (C API), cudaGetTextureReference, cudaBindTexture (C API), cudaBindTexture2D (C
API), cudaBindTextureToArray (C API), cudaUnbindTexture (C API), cudaGetTextureAlignmentOffset (C API)

3.12.2.6 cudaError_t cudaGetTextureAlignmentOffset (size_t ∗ offset, const struct textureReference ∗ texref)

Returns in ∗offset the offset that was returned when texture reference texref was bound.

Parameters:
offset - Offset of texture reference in bytes
texref - Texture to get offset of

Returns:
cudaSuccess, cudaErrorInvalidTexture, cudaErrorInvalidTextureBinding

Note:
Note that this function may also return error codes from previous, asynchronous launches.

Generated for NVIDIA CUDA Library by Doxygen

82 Module Documentation

See also:
cudaCreateChannelDesc (C API), cudaGetChannelDesc, cudaGetTextureReference, cudaBindTexture (C API),
cudaBindTexture2D (C API), cudaBindTextureToArray (C API), cudaUnbindTexture (C API), cudaGetTexture-
AlignmentOffset (C++ API)

3.12.2.7 cudaError_t cudaGetTextureReference (const struct textureReference ∗∗ texref, const char ∗

symbol)

Returns in ∗texref the structure associated to the texture reference defined by symbol symbol.

Parameters:
texref - Texture associated with symbol
symbol - Symbol to find texture reference for

Returns:
cudaSuccess, cudaErrorInvalidTexture,

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaCreateChannelDesc (C API), cudaGetChannelDesc, cudaGetTextureAlignmentOffset (C API), cudaBind-
Texture (C API), cudaBindTexture2D (C API), cudaBindTextureToArray (C API), cudaUnbindTexture (C API)

3.12.2.8 cudaError_t cudaUnbindTexture (const struct textureReference ∗ texref)

Unbinds the texture bound to texref.

Parameters:
texref - Texture to unbind

Returns:
cudaSuccess

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaCreateChannelDesc (C API), cudaGetChannelDesc, cudaGetTextureReference, cudaBindTexture (C API),
cudaBindTexture2D (C API), cudaBindTextureToArray (C API), cudaUnbindTexture (C++ API), cudaGetTex-
tureAlignmentOffset (C API)

Generated for NVIDIA CUDA Library by Doxygen

3.13 Version Management 83

3.13 Version Management

Functions
• cudaError_t cudaDriverGetVersion (int ∗driverVersion)
Returns the CUDA driver version.

• cudaError_t cudaRuntimeGetVersion (int ∗runtimeVersion)

Returns the CUDA Runtime version.

3.13.1 Function Documentation

3.13.1.1 cudaError_t cudaDriverGetVersion (int ∗ driverVersion)

Returns in ∗driverVersion the version number of the installed CUDA driver. If no driver is installed, then 0 is
returned as the driver version (via driverVersion). This function automatically returns cudaErrorInvalidValue if
the driverVersion argument is NULL.

Parameters:
driverVersion - Returns the CUDA driver version.

Returns:
cudaSuccess, cudaErrorInvalidValue

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaRuntimeGetVersion

3.13.1.2 cudaError_t cudaRuntimeGetVersion (int ∗ runtimeVersion)

Returns in ∗runtimeVersion the version number of the installed CUDA Runtime. This function automatically
returns cudaErrorInvalidValue if the runtimeVersion argument is NULL.

Parameters:
runtimeVersion - Returns the CUDA Runtime version.

Returns:
cudaSuccess, cudaErrorInvalidValue

See also:
cudaDriverGetVersion

Generated for NVIDIA CUDA Library by Doxygen

84 Module Documentation

3.14 C++ API Routines

C++-style interface built on top of CUDA runtime API.

Functions
• template<class T , int dim, enum cudaTextureReadMode readMode>
cudaError_t cudaBindTexture (size_t ∗offset, const struct texture< T, dim, readMode > &tex, const void
∗devPtr, size_t size=UINT_MAX)
[C++ API] Binds a memory area to a texture

• template<class T , int dim, enum cudaTextureReadMode readMode>

cudaError_t cudaBindTexture (size_t ∗offset, const struct texture< T, dim, readMode > &tex, const void
∗devPtr, const struct cudaChannelFormatDesc &desc, size_t size=UINT_MAX)
[C++ API] Binds a memory area to a texture

• template<class T , int dim, enum cudaTextureReadMode readMode>

cudaError_t cudaBindTexture2D (size_t ∗offset, const struct texture< T, dim, readMode > &tex, const void
∗devPtr, const struct cudaChannelFormatDesc &desc, size_t width, size_t height, size_t pitch)
[C++ API] Binds a 2D memory area to a texture

• template<class T , int dim, enum cudaTextureReadMode readMode>

cudaError_t cudaBindTextureToArray (const struct texture< T, dim, readMode > &tex, const struct cudaArray
∗array)
[C++ API] Binds an array to a texture

• template<class T , int dim, enum cudaTextureReadMode readMode>

cudaError_t cudaBindTextureToArray (const struct texture< T, dim, readMode > &tex, const struct cudaArray
∗array, const struct cudaChannelFormatDesc &desc)
[C++ API] Binds an array to a texture

• template<class T >
cudaChannelFormatDesc cudaCreateChannelDesc (void)
[C++ API] Returns a channel descriptor using the specified format

• template<class T >
cudaError_t cudaFuncGetAttributes (struct cudaFuncAttributes ∗attr, T ∗entry)
[C++ API] Find out attributes for a given function

• template<class T >
cudaError_t cudaGetSymbolAddress (void ∗∗devPtr, const T &symbol)
[C++ API] Finds the address associated with a CUDA symbol

• template<class T >
cudaError_t cudaGetSymbolSize (size_t ∗size, const T &symbol)
[C++ API] Finds the size of the object associated with a CUDA symbol

• template<class T , int dim, enum cudaTextureReadMode readMode>

cudaError_t cudaGetTextureAlignmentOffset (size_t ∗offset, const struct texture< T, dim, readMode > &tex)
[C++ API] Get the alignment offset of a texture

Generated for NVIDIA CUDA Library by Doxygen

3.14 C++ API Routines 85

• template<class T >
cudaError_t cudaLaunch (T ∗entry)
[C++ API] Launches a device function

• template<class T >
cudaError_t cudaSetupArgument (T arg, size_t offset)
[C++ API] Configure a device launch

• template<class T , int dim, enum cudaTextureReadMode readMode>

cudaError_t cudaUnbindTexture (const struct texture< T, dim, readMode > &tex)
[C++ API] Unbinds a texture

3.14.1 Detailed Description

This section describes the C++ high level API functions of the CUDA runtime application programming interface. To
use these functions, your application needs to be compiled with the nvcc compiler.

3.14.2 Function Documentation

3.14.2.1 template<class T , int dim, enum cudaTextureReadMode readMode> cudaError_t

cudaBindTexture (size_t ∗ offset, const struct texture< T, dim, readMode > & tex, const void ∗
devPtr, size_t size = UINT_MAX)

Binds size bytes of the memory area pointed to by devPtr to texture reference tex. The channel descriptor is
inherited from the texture reference type. The offset parameter is an optional byte offset as with the low-level
cudaBindTexture(size_t∗, const struct textureReference∗, const void∗, const struct cudaChannelFormatDesc∗, size_t)
function. Any memory previously bound to tex is unbound.

Parameters:
offset - Offset in bytes
tex - Texture to bind
devPtr - Memory area on device
size - Size of the memory area pointed to by devPtr

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidDevicePointer, cudaErrorInvalidTexture

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaCreateChannelDesc (C++ API), cudaGetChannelDesc, cudaGetTextureReference, cudaBindTexture (C API),
cudaBindTexture (C++ API, inherited channel descriptor), cudaBindTexture2D (C++ API), cudaBindTexture-
ToArray (C++ API), cudaBindTextureToArray (C++ API, inherited channel descriptor), cudaUnbindTexture (C++
API), cudaGetTextureAlignmentOffset (C++ API)

Generated for NVIDIA CUDA Library by Doxygen

86 Module Documentation

3.14.2.2 template<class T , int dim, enum cudaTextureReadMode readMode> cudaError_t

cudaBindTexture (size_t ∗ offset, const struct texture< T, dim, readMode > & tex, const void ∗
devPtr, const struct cudaChannelFormatDesc & desc, size_t size = UINT_MAX)

Binds size bytes of the memory area pointed to by devPtr to texture reference tex. desc describes how the
memory is interpreted when fetching values from the texture. The offset parameter is an optional byte offset as
with the low-level cudaBindTexture() function. Any memory previously bound to tex is unbound.

Parameters:
offset - Offset in bytes
tex - Texture to bind
devPtr - Memory area on device
desc - Channel format
size - Size of the memory area pointed to by devPtr

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidDevicePointer, cudaErrorInvalidTexture

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaCreateChannelDesc (C++ API), cudaGetChannelDesc, cudaGetTextureReference, cudaBindTexture (C API),
cudaBindTexture (C++ API, inherited channel descriptor), cudaBindTexture2D (C++ API), cudaBindTexture-
ToArray (C++ API), cudaBindTextureToArray (C++ API, inherited channel descriptor), cudaUnbindTexture (C++
API), cudaGetTextureAlignmentOffset (C++ API)

3.14.2.3 template<class T , int dim, enum cudaTextureReadMode readMode> cudaError_t

cudaBindTexture2D (size_t ∗ offset, const struct texture< T, dim, readMode > & tex, const void ∗
devPtr, const struct cudaChannelFormatDesc & desc, size_t width, size_t height, size_t pitch)

Binds the 2D memory area pointed to by devPtr to the texture reference tex. The size of the area is constrained by
width in texel units, height in texel units, and pitch in byte units. desc describes how the memory is interpreted
when fetching values from the texture. Any memory previously bound to tex is unbound.
Since the hardware enforces an alignment requirement on texture base addresses, cudaBindTexture2D() returns in
∗offset a byte offset that must be applied to texture fetches in order to read from the desired memory. This offset
must be divided by the texel size and passed to kernels that read from the texture so they can be applied to the tex2D()
function. If the device memory pointer was returned from cudaMalloc(), the offset is guaranteed to be 0 and NULL
may be passed as the offset parameter.

Parameters:
offset - Offset in bytes
tex - Texture reference to bind
devPtr - 2D memory area on device
desc - Channel format
width - Width in texel units

Generated for NVIDIA CUDA Library by Doxygen

3.14 C++ API Routines 87

height - Height in texel units

pitch - Pitch in bytes

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidDevicePointer, cudaErrorInvalidTexture

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaCreateChannelDesc (C++ API), cudaGetChannelDesc, cudaGetTextureReference, cudaBindTexture (C++
API), cudaBindTexture (C++ API, inherited channel descriptor), cudaBindTexture2D (C API), cudaBindTex-
tureToArray (C++ API), cudaBindTextureToArray (C++ API, inherited channel descriptor), cudaUnbindTexture
(C++ API), cudaGetTextureAlignmentOffset (C++ API)

3.14.2.4 template<class T , int dim, enum cudaTextureReadMode readMode> cudaError_t

cudaBindTextureToArray (const struct texture< T, dim, readMode > & tex, const struct cudaArray
∗ array)

Binds the CUDA array array to the texture reference tex. The channel descriptor is inherited from the CUDA array.
Any CUDA array previously bound to tex is unbound.

Parameters:
tex - Texture to bind
array - Memory array on device

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidDevicePointer, cudaErrorInvalidTexture

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaCreateChannelDesc (C++ API), cudaGetChannelDesc, cudaGetTextureReference, cudaBindTexture (C++
API), cudaBindTexture (C++ API, inherited channel descriptor), cudaBindTexture2D (C++ API), cudaBind-
TextureToArray (C API), cudaBindTextureToArray (C++ API), cudaUnbindTexture (C++ API), cudaGetTexture-
AlignmentOffset (C++ API)

3.14.2.5 template<class T , int dim, enum cudaTextureReadMode readMode> cudaError_t

cudaBindTextureToArray (const struct texture< T, dim, readMode > & tex, const struct cudaArray
∗ array, const struct cudaChannelFormatDesc & desc)

Binds the CUDA array array to the texture reference tex. desc describes how the memory is interpreted when
fetching values from the texture. Any CUDA array previously bound to tex is unbound.

Parameters:
tex - Texture to bind

Generated for NVIDIA CUDA Library by Doxygen

88 Module Documentation

array - Memory array on device

desc - Channel format

Returns:
cudaSuccess, cudaErrorInvalidValue, cudaErrorInvalidDevicePointer, cudaErrorInvalidTexture

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaCreateChannelDesc (C++ API), cudaGetChannelDesc, cudaGetTextureReference, cudaBindTexture (C++
API), cudaBindTexture (C++ API, inherited channel descriptor), cudaBindTexture2D (C++ API), cudaBind-
TextureToArray (C API), cudaBindTextureToArray (C++ API, inherited channel descriptor), cudaUnbindTexture
(C++ API), cudaGetTextureAlignmentOffset (C++ API)

3.14.2.6 template<class T > cudaChannelFormatDesc cudaCreateChannelDesc (void)

Returns a channel descriptor with format f and number of bits of each component x, y, z, and w. The cudaChan-
nelFormatDesc is defined as:

struct cudaChannelFormatDesc {
int x, y, z, w;
enum cudaChannelFormatKind f;
};

where cudaChannelFormatKind is one of cudaChannelFormatKindSigned, cudaChannelFormatKindUnsigned, or cu-

daChannelFormatKindFloat.

Returns:
Channel descriptor with format f

See also:
cudaCreateChannelDesc (Low level), cudaGetChannelDesc, cudaGetTextureReference, cudaBindTexture (High
level), cudaBindTexture (High level, inherited channel descriptor), cudaBindTexture2D (High level), cudaBind-
TextureToArray (High level), cudaBindTextureToArray (High level, inherited channel descriptor), cudaUnbind-
Texture (High level), cudaGetTextureAlignmentOffset (High level)

3.14.2.7 template<class T > cudaError_t cudaFuncGetAttributes (struct cudaFuncAttributes ∗ attr, T ∗

entry)

This function obtains the attributes of a function specified via entry. The parameter entry can either be a function
that executes on the device, or it can be a character string, naming a function that executes on the device. The parameter
specified by entry must be declared as a __global__ function. The fetched attributes are placed in attr. If the
specified function does not exist, then cudaErrorInvalidDeviceFunction is returned.

Parameters:
attr - Return pointer to function’s attributes
entry - Function to get attributes of

Generated for NVIDIA CUDA Library by Doxygen

3.14 C++ API Routines 89

Returns:
cudaSuccess, cudaErrorInitializationError, cudaErrorInvalidDeviceFunction

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaConfigureCall, cudaFuncGetAttributes (C API), cudaLaunch (C++ API), cudaSetDoubleForDevice, cudaSet-
DoubleForHost, cudaSetupArgument (C++ API)

3.14.2.8 template<class T > cudaError_t cudaGetSymbolAddress (void ∗∗ devPtr, const T & symbol)

Returns in ∗devPtr the address of symbol symbol on the device. symbol can either be a variable that resides
in global memory space, or it can be a character string, naming a variable that resides in global memory space. If
symbol cannot be found, or if symbol is not declared in the global memory space, ∗devPtr is unchanged and the
error cudaErrorInvalidSymbol is returned.

Parameters:
devPtr - Return device pointer associated with symbol
symbol - Global variable or string symbol to search for

Returns:
cudaSuccess, cudaErrorInvalidSymbol, cudaErrorAddressOfConstant

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaGetSymbolAddress (C API) cudaGetSymbolSize (C++ API)

3.14.2.9 template<class T > cudaError_t cudaGetSymbolSize (size_t ∗ size, const T & symbol)

Returns in ∗size the size of symbol symbol. symbol can either be a variable that resides in global or constant
memory space, or it can be a character string, naming a variable that resides in global or constant memory space. If
symbol cannot be found, or if symbol is not declared in global or constant memory space, ∗size is unchanged
and the error cudaErrorInvalidSymbol is returned.

Parameters:
size - Size of object associated with symbol
symbol - Global variable or string symbol to find size of

Returns:
cudaSuccess, cudaErrorInvalidSymbol

Note:
Note that this function may also return error codes from previous, asynchronous launches.

Generated for NVIDIA CUDA Library by Doxygen

90 Module Documentation

See also:
cudaGetSymbolAddress (C++ API) cudaGetSymbolSize (C API)

3.14.2.10 template<class T , int dim, enum cudaTextureReadMode readMode> cudaError_t

cudaGetTextureAlignmentOffset (size_t ∗ offset, const struct texture< T, dim, readMode > & tex)

Returns in ∗offset the offset that was returned when texture reference tex was bound.

Parameters:
offset - Offset of texture reference in bytes
tex - Texture to get offset of

Returns:
cudaSuccess, cudaErrorInvalidTexture, cudaErrorInvalidTextureBinding

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaCreateChannelDesc (C++ API), cudaGetChannelDesc, cudaGetTextureReference, cudaBindTexture (C++
API), cudaBindTexture (C++ API, inherited channel descriptor), cudaBindTexture2D (C++ API), cudaBindTex-
tureToArray (C++ API), cudaBindTextureToArray (C++ API, inherited channel descriptor), cudaUnbindTexture
(C++ API), cudaGetTextureAlignmentOffset (C API)

3.14.2.11 template<class T > cudaError_t cudaLaunch (T ∗ entry)

Launches the function entry on the device. The parameter entry can either be a function that executes on the
device, or it can be a character string, naming a function that executes on the device. The parameter specified by entry
must be declared as a __global__ function. cudaLaunch() must be preceded by a call to cudaConfigureCall() since
it pops the data that was pushed by cudaConfigureCall() from the execution stack.

Parameters:
entry - Device function pointer or char string naming device function to execute

Returns:
cudaSuccess, cudaErrorInvalidDeviceFunction, cudaErrorInvalidConfiguration, cudaErrorLaunchFailure, cud-
aErrorPriorLaunchFailure, cudaErrorLaunchTimeout, cudaErrorLaunchOutOfResources

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaConfigureCall, cudaFuncGetAttributes (C++ API), cudaLaunch (C API), cudaSetDoubleForDevice, cudaSet-
DoubleForHost, cudaSetupArgument (C++ API)

Generated for NVIDIA CUDA Library by Doxygen

3.14 C++ API Routines 91

3.14.2.12 template<class T > cudaError_t cudaSetupArgument (T arg, size_t offset)

Pushes size bytes of the argument pointed to by arg at offset bytes from the start of the parameter passing area,
which starts at offset 0. The arguments are stored in the top of the execution stack. cudaSetupArgument() must be
preceded by a call to cudaConfigureCall().

Parameters:
arg - Argument to push for a kernel launch
offset - Offset in argument stack to push new arg

Returns:
cudaSuccess

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaConfigureCall, cudaFuncGetAttributes (C++ API), cudaLaunch (C++ API), cudaSetDoubleForDevice, cud-
aSetDoubleForHost, cudaSetupArgument (C API)

3.14.2.13 template<class T , int dim, enum cudaTextureReadMode readMode> cudaError_t

cudaUnbindTexture (const struct texture< T, dim, readMode > & tex)

Unbinds the texture bound to tex.

Parameters:
tex - Texture to unbind

Returns:
cudaSuccess

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cudaCreateChannelDesc (C++ API), cudaGetChannelDesc, cudaGetTextureReference, cudaBindTexture (C++
API), cudaBindTexture (C++ API, inherited channel descriptor), cudaBindTexture2D (C++ API), cudaBindTex-
tureToArray (C++ API), cudaBindTextureToArray (C++ API, inherited channel descriptor), cudaUnbindTexture
(C API), cudaGetTextureAlignmentOffset (C++ API)

Generated for NVIDIA CUDA Library by Doxygen

92 Module Documentation

3.15 Data types used by CUDA Runtime

Data Structures
• struct cudaChannelFormatDesc
• struct cudaDeviceProp
• struct cudaExtent
• struct cudaFuncAttributes
• struct cudaMemcpy3DParms
• struct cudaPitchedPtr
• struct cudaPos

Data types used by CUDA Runtime

Data types used by CUDA Runtime

Author:
NVIDIA Corporation

• enum cudaChannelFormatKind {
cudaChannelFormatKindSigned,
cudaChannelFormatKindUnsigned,
cudaChannelFormatKindFloat,
cudaChannelFormatKindNone }
• enum cudaComputeMode {
cudaComputeModeDefault,
cudaComputeModeExclusive,
cudaComputeModeProhibited }
• enum cudaError {
cudaSuccess,
cudaErrorMissingConfiguration,
cudaErrorMemoryAllocation,
cudaErrorInitializationError,
cudaErrorLaunchFailure,
cudaErrorPriorLaunchFailure,
cudaErrorLaunchTimeout,
cudaErrorLaunchOutOfResources,
cudaErrorInvalidDeviceFunction,
cudaErrorInvalidConfiguration,
cudaErrorInvalidDevice,
cudaErrorInvalidValue,
cudaErrorInvalidPitchValue,
cudaErrorInvalidSymbol,
cudaErrorMapBufferObjectFailed,

Generated for NVIDIA CUDA Library by Doxygen

3.15 Data types used by CUDA Runtime 93

cudaErrorUnmapBufferObjectFailed,
cudaErrorInvalidHostPointer,
cudaErrorInvalidDevicePointer,
cudaErrorInvalidTexture,
cudaErrorInvalidTextureBinding,
cudaErrorInvalidChannelDescriptor,
cudaErrorInvalidMemcpyDirection,
cudaErrorAddressOfConstant,
cudaErrorTextureFetchFailed,
cudaErrorTextureNotBound,
cudaErrorSynchronizationError,
cudaErrorInvalidFilterSetting,
cudaErrorInvalidNormSetting,
cudaErrorMixedDeviceExecution,
cudaErrorCudartUnloading,
cudaErrorUnknown,
cudaErrorNotYetImplemented,
cudaErrorMemoryValueTooLarge,
cudaErrorInvalidResourceHandle,
cudaErrorNotReady,
cudaErrorInsufficientDriver,
cudaErrorSetOnActiveProcess,
cudaErrorNoDevice,
cudaErrorStartupFailure,
cudaErrorApiFailureBase }
• enum cudaMemcpyKind {
cudaMemcpyHostToHost,
cudaMemcpyHostToDevice,
cudaMemcpyDeviceToHost,
cudaMemcpyDeviceToDevice }
• typedef enum cudaError cudaError_t
• typedef int cudaEvent_t
• typedef int cudaStream_t
• #define cudaDeviceBlockingSync
Device flag - Use blocking synchronization.

• #define cudaDeviceMapHost
Device flag - Support mapped pinned allocations.

• #define cudaDeviceMask
Device flags mask.

• #define cudaDevicePropDontCare
Empty device properties.

Generated for NVIDIA CUDA Library by Doxygen

94 Module Documentation

• #define cudaDeviceScheduleAuto
Device flag - Automatic scheduling.

• #define cudaDeviceScheduleSpin
Device flag - Spin default scheduling.

• #define cudaDeviceScheduleYield
Device flag - Yield default scheduling.

• #define cudaEventBlockingSync
Event uses blocking synchronization.

• #define cudaEventDefault
Default event flag.

• #define cudaHostAllocDefault
Default page-locked allocation flag.

• #define cudaHostAllocMapped
Map allocation into device space.

• #define cudaHostAllocPortable
Pinned memory accessible by all CUDA contexts.

• #define cudaHostAllocWriteCombined
Write-combined memory.

3.15.1 Typedef Documentation

3.15.1.1 typedef enum cudaError cudaError_t

CUDA Error types

3.15.1.2 typedef int cudaEvent_t

CUDA event types

3.15.1.3 typedef int cudaStream_t

CUDA stream

3.15.2 Enumeration Type Documentation

3.15.2.1 enum cudaChannelFormatKind

Channel format kind

Generated for NVIDIA CUDA Library by Doxygen

3.15 Data types used by CUDA Runtime 95

Enumerator:
cudaChannelFormatKindSigned Signed channel format.
cudaChannelFormatKindUnsigned Unsigned channel format.
cudaChannelFormatKindFloat Float channel format.
cudaChannelFormatKindNone No channel format.

3.15.2.2 enum cudaComputeMode

CUDA device compute modes

Enumerator:
cudaComputeModeDefault Default compute mode (Multiple threads can use cudaSetDevice() with this device).

cudaComputeModeExclusive Compute-exclusive mode (Only one thread will be able to use cudaSetDevice()
with this device).
cudaComputeModeProhibited Compute-prohibited mode (No threads can use cudaSetDevice() with this de-
vice).

3.15.2.3 enum cudaError

CUDA error types

Enumerator:
cudaSuccess No errors.
cudaErrorMissingConfiguration Missing configuration error.
cudaErrorMemoryAllocation Memory allocation error.
cudaErrorInitializationError Initialization error.
cudaErrorLaunchFailure Launch failure.
cudaErrorPriorLaunchFailure Prior launch failure.
cudaErrorLaunchTimeout Launch timeout error.
cudaErrorLaunchOutOfResources Launch out of resources error.
cudaErrorInvalidDeviceFunction Invalid device function.
cudaErrorInvalidConfiguration Invalid configuration.
cudaErrorInvalidDevice Invalid device.
cudaErrorInvalidValue Invalid value.
cudaErrorInvalidPitchValue Invalid pitch value.
cudaErrorInvalidSymbol Invalid symbol.
cudaErrorMapBufferObjectFailed Map buffer object failed.
cudaErrorUnmapBufferObjectFailed Unmap buffer object failed.
cudaErrorInvalidHostPointer Invalid host pointer.
cudaErrorInvalidDevicePointer Invalid device pointer.
cudaErrorInvalidTexture Invalid texture.
cudaErrorInvalidTextureBinding Invalid texture binding.

Generated for NVIDIA CUDA Library by Doxygen

96 Module Documentation

cudaErrorInvalidChannelDescriptor Invalid channel descriptor.

cudaErrorInvalidMemcpyDirection Invalid memcpy direction.
cudaErrorAddressOfConstant Address of constant error.
cudaErrorTextureFetchFailed Texture fetch failed.
cudaErrorTextureNotBound Texture not bound error.
cudaErrorSynchronizationError Synchronization error.
cudaErrorInvalidFilterSetting Invalid filter setting.
cudaErrorInvalidNormSetting Invalid norm setting.
cudaErrorMixedDeviceExecution Mixed device execution.
cudaErrorCudartUnloading CUDA runtime unloading.
cudaErrorUnknown Unknown error condition.
cudaErrorNotYetImplemented Function not yet implemented.
cudaErrorMemoryValueTooLarge Memory value too large.
cudaErrorInvalidResourceHandle Invalid resource handle.
cudaErrorNotReady Not ready error.
cudaErrorInsufficientDriver CUDA runtime is newer than driver.
cudaErrorSetOnActiveProcess Set on active process error.
cudaErrorNoDevice No available CUDA device.
cudaErrorStartupFailure Startup failure.
cudaErrorApiFailureBase API failure base.

3.15.2.4 enum cudaMemcpyKind

CUDA memory copy types

Enumerator:
cudaMemcpyHostToHost Host -> Host.
cudaMemcpyHostToDevice Host -> Device.
cudaMemcpyDeviceToHost Device -> Host.
cudaMemcpyDeviceToDevice Device -> Device.

Generated for NVIDIA CUDA Library by Doxygen

3.16 CUDA Driver API 97

3.16 CUDA Driver API

Modules
• Initialization
• Device Management
• Version Management
• Context Management
• Module Management
• Stream Management
• Event Management
• Execution Control
• Memory Management
• Texture Reference Management
• OpenGL Interoperability
• Direct3D 9 Interoperability
• Direct3D 10 Interoperability
• Data types used by CUDA driver

3.16.1 Detailed Description

This section describes the low-level CUDA driver application programming interface.

Generated for NVIDIA CUDA Library by Doxygen

98 Module Documentation

3.17 Initialization

Functions
• CUresult cuInit (unsigned int Flags)
Initialize the CUDA driver API.

3.17.1 Detailed Description

This section describes the initialization functions of the low-level CUDA driver application programming interface.

3.17.2 Function Documentation

3.17.2.1 cuInit (unsigned int Flags)

Initializes the driver API and must be called before any other function from the driver API. Currently, the Flags
parameter must be 0. If cuInit() has not been called, any function from the driver API will return CUDA_ERROR_-
NOT_INITIALIZED.

Parameters:
Flags - Initialization flag for CUDA.

Returns:
CUDA_SUCCESS, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_INVALID_DEVICE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

Generated for NVIDIA CUDA Library by Doxygen

3.18 Device Management 99

3.18 Device Management

Functions
• CUresult cuDeviceComputeCapability (int ∗major, int ∗minor, CUdevice dev)
Returns the compute capability of the device.

• CUresult cuDeviceGet (CUdevice ∗device, int ordinal)

Returns a handle to a compute device.

• CUresult cuDeviceGetAttribute (int ∗pi, CUdevice_attribute attrib, CUdevice dev)

Returns information about the device.

• CUresult cuDeviceGetCount (int ∗count)

Returns the number of compute-capable devices.

• CUresult cuDeviceGetName (char ∗name, int len, CUdevice dev)

Returns an identifer string for the device.

• CUresult cuDeviceGetProperties (CUdevprop ∗prop, CUdevice dev)

Returns properties for a selected device.

• CUresult cuDeviceTotalMem (unsigned int ∗bytes, CUdevice dev)

Returns the total amount of memory on the device.

3.18.1 Detailed Description

This section describes the device management functions of the low-level CUDA driver application programming in-
terface.

3.18.2 Function Documentation

3.18.2.1 cuDeviceComputeCapability (int ∗ major, int ∗ minor, CUdevice dev)

Returns in ∗major and ∗minor the major and minor revision numbers that define the compute capability of the
device dev.

Parameters:
major - Major revision number
minor - Minor revision number
dev - Device handle

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_INVALID_DEVICE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

Generated for NVIDIA CUDA Library by Doxygen

100 Module Documentation

See also:
cuDeviceGetAttribute, cuDeviceGetCount, cuDeviceGetName, cuDeviceGet, cuDeviceGetProperties, cuDevice-
TotalMem

3.18.2.2 cuDeviceGet (CUdevice ∗ device, int ordinal)

Returns in ∗device a device handle given an ordinal in the range [0, cuDeviceGetCount()-1].

Parameters:
device - Returned device handle
ordinal - Device number to get handle for

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_INVALID_DEVICE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuDeviceComputeCapability, cuDeviceGetAttribute, cuDeviceGetCount, cuDeviceGetName, cuDeviceGetProp-
erties, cuDeviceTotalMem

3.18.2.3 cuDeviceGetAttribute (int ∗ pi, CUdevice_attribute attrib, CUdevice dev)

Returns in ∗pi the integer value of the attribute attrib on device dev. The supported attributes are:

• CU_DEVICE_ATTRIBUTE_MAX_THREADS_PER_BLOCK: Maximum number of threads per block;

• CU_DEVICE_ATTRIBUTE_MAX_BLOCK_DIM_X: Maximum x-dimension of a block;
• CU_DEVICE_ATTRIBUTE_MAX_BLOCK_DIM_Y: Maximum y-dimension of a block;
• CU_DEVICE_ATTRIBUTE_MAX_BLOCK_DIM_Z: Maximum z-dimension of a block;
• CU_DEVICE_ATTRIBUTE_MAX_GRID_DIM_X: Maximum x-dimension of a grid;
• CU_DEVICE_ATTRIBUTE_MAX_GRID_DIM_Y: Maximum y-dimension of a grid;
• CU_DEVICE_ATTRIBUTE_MAX_GRID_DIM_Z: Maximum z-dimension of a grid;
• CU_DEVICE_ATTRIBUTE_MAX_SHARED_MEMORY_PER_BLOCK: Maximum amount of shared mem-
ory available to a thread block in bytes; this amount is shared by all thread blocks simultaneously resident on a
multiprocessor;
• CU_DEVICE_ATTRIBUTE_TOTAL_CONSTANT_MEMORY: Memory available on device for __constant_-
_ variables in a CUDA C kernel in bytes;
• CU_DEVICE_ATTRIBUTE_WARP_SIZE: Warp size in threads;
• CU_DEVICE_ATTRIBUTE_MAX_PITCH: Maximum pitch in bytes allowed by the memory copy functions
that involve memory regions allocated through cuMemAllocPitch();

Generated for NVIDIA CUDA Library by Doxygen

3.18 Device Management 101

• CU_DEVICE_ATTRIBUTE_MAX_REGISTERS_PER_BLOCK: Maximum number of 32-bit registers avail-

able to a thread block; this number is shared by all thread blocks simultaneously resident on a multiprocessor;

• CU_DEVICE_ATTRIBUTE_CLOCK_RATE: Peak clock frequency in kilohertz;

• CU_DEVICE_ATTRIBUTE_TEXTURE_ALIGNMENT: Alignment requirement; texture base addresses

aligned to textureAlign bytes do not need an offset applied to texture fetches;

• CU_DEVICE_ATTRIBUTE_GPU_OVERLAP: 1 if the device can concurrently copy memory between host

and device while executing a kernel, or 0 if not;

• CU_DEVICE_ATTRIBUTE_MULTIPROCESSOR_COUNT: Number of multiprocessors on the device;

• CU_DEVICE_ATTRIBUTE_KERNEL_EXEC_TIMEOUT: 1 if there is a run time limit for kernels executed

on the device, or 0 if not;

• CU_DEVICE_ATTRIBUTE_INTEGRATED: 1 if the device is integrated with the memory subsystem, or 0 if

not;

• CU_DEVICE_ATTRIBUTE_CAN_MAP_HOST_MEMORY: 1 if the device can map host memory into the

CUDA address space, or 0 if not;

• CU_DEVICE_ATTRIBUTE_COMPUTE_MODE: Compute mode that device is currently in. Available modes

are as follows:

– CU_COMPUTEMODE_DEFAULT: Default mode - Device is not restricted and can have multiple CUDA
contexts present at a single time.
– CU_COMPUTEMODE_EXCLUSIVE: Compute-exclusive mode - Device can have only one CUDA con-
text present on it at a time.
– CU_COMPUTEMODE_PROHIBITED: Compute-prohibited mode - Device is prohibited from creating
new CUDA contexts.

Parameters:
pi - Returned device attribute value
attrib - Device attribute to query
dev - Device handle

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_INVALID_DEVICE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuDeviceComputeCapability, cuDeviceGetCount, cuDeviceGetName, cuDeviceGet, cuDeviceGetProperties,
cuDeviceTotalMem

Generated for NVIDIA CUDA Library by Doxygen

102 Module Documentation

3.18.2.4 cuDeviceGetCount (int ∗ count)

Returns in ∗count the number of devices with compute capability greater than or equal to 1.0 that are available for
execution. If there is no such device, cuDeviceGetCount() returns 0.

Parameters:
count - Returned number of compute-capable devices

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuDeviceComputeCapability, cuDeviceGetAttribute, cuDeviceGetName, cuDeviceGet, cuDeviceGetProperties,
cuDeviceTotalMem

3.18.2.5 cuDeviceGetName (char ∗ name, int len, CUdevice dev)

Returns an ASCII string identifying the device dev in the NULL-terminated string pointed to by name. len specifies
the maximum length of the string that may be returned.

Parameters:
name - Returned identifier string for the device
len - Maximum length of string to store in name
dev - Device to get identifier string for

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_INVALID_DEVICE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuDeviceComputeCapability, cuDeviceGetAttribute, cuDeviceGetCount, cuDeviceGet, cuDeviceGetProperties,
cuDeviceTotalMem

3.18.2.6 cuDeviceGetProperties (CUdevprop ∗ prop, CUdevice dev)

Returns in ∗prop the properties of device dev. The CUdevprop structure is defined as:

Generated for NVIDIA CUDA Library by Doxygen

3.18 Device Management 103

typedef struct CUdevprop_st {

int maxThreadsPerBlock;
int maxThreadsDim[3];
int maxGridSize[3];
int sharedMemPerBlock;
int totalConstantMemory;
int SIMDWidth;
int memPitch;
int regsPerBlock;
int clockRate;
int textureAlign
} CUdevprop;

where:

• maxThreadsPerBlock is the maximum number of threads per block;

• maxThreadsDim[3] is the maximum sizes of each dimension of a block;

• maxGridSize[3] is the maximum sizes of each dimension of a grid;

• sharedMemPerBlock is the total amount of shared memory available per block in bytes;

• totalConstantMemory is the total amount of constant memory available on the device in bytes;

• SIMDWidth is the warp size;

• memPitch is the maximum pitch allowed by the memory copy functions that involve memory regions allocated
through cuMemAllocPitch();

• regsPerBlock is the total number of registers available per block;

• clockRate is the clock frequency in kilohertz;

• textureAlign is the alignment requirement; texture base addresses that are aligned to textureAlign bytes do not
need an offset applied to texture fetches.

Parameters:
prop - Returned properties of device
dev - Device to get properties for

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_INVALID_DEVICE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuDeviceComputeCapability, cuDeviceGetAttribute, cuDeviceGetCount, cuDeviceGetName, cuDeviceGet, cuD-
eviceTotalMem

Generated for NVIDIA CUDA Library by Doxygen

104 Module Documentation

3.18.2.7 cuDeviceTotalMem (unsigned int ∗ bytes, CUdevice dev)

Returns in ∗bytes the total amount of memory available on the device dev in bytes.

Parameters:
bytes - Returned memory available on device in bytes
dev - Device handle

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_INVALID_DEVICE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuDeviceComputeCapability, cuDeviceGetAttribute, cuDeviceGetCount, cuDeviceGetName, cuDeviceGet, cuD-
eviceGetProperties,

Generated for NVIDIA CUDA Library by Doxygen

3.19 Version Management 105

3.19 Version Management

Functions
• CUresult cuDriverGetVersion (int ∗driverVersion)
Returns the CUDA driver version.

3.19.1 Detailed Description

This section describes the version management functions of the low-level CUDA driver application programming
interface.

3.19.2 Function Documentation

3.19.2.1 cuDriverGetVersion (int ∗ driverVersion)

Returns in ∗driverVersion the version number of the installed CUDA driver. This function automatically returns
CUDA_ERROR_INVALID_VALUE if the driverVersion argument is NULL.

Parameters:
driverVersion - Returns the CUDA driver version

Returns:
CUDA_SUCCESS, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

Generated for NVIDIA CUDA Library by Doxygen

106 Module Documentation

3.20 Context Management

Functions

• CUresult cuCtxAttach (CUcontext ∗pctx, unsigned int flags)

Increment a context’s usage-count.

• CUresult cuCtxCreate (CUcontext ∗pctx, unsigned int flags, CUdevice dev)

Create a CUDA context.

• CUresult cuCtxDestroy (CUcontext ctx)

Destroy the current context or a floating CUDA context.

• CUresult cuCtxDetach (CUcontext ctx)

Decrement a context’s usage-count.

• CUresult cuCtxGetDevice (CUdevice ∗device)

Returns the device ID for the current context.

• CUresult cuCtxPopCurrent (CUcontext ∗pctx)

Pops the current CUDA context from the current CPU thread.

• CUresult cuCtxPushCurrent (CUcontext ctx)

Pushes a floating context on the current CPU thread.

• CUresult cuCtxSynchronize (void)

Block for a context’s tasks to complete.

3.20.1 Detailed Description

This section describes the context management functions of the low-level CUDA driver application programming
interface.

3.20.2 Function Documentation

3.20.2.1 cuCtxAttach (CUcontext ∗ pctx, unsigned int flags)

Increments the usage count of the context and passes back a context handle in ∗pctx that must be passed to cuC-
txDetach() when the application is done with the context. cuCtxAttach() fails if there is no context current to the
thread.
Currently, the flags parameter must be 0.

Parameters:
pctx - Returned context handle of the current context
flags - Context attach flags (must be 0)

Generated for NVIDIA CUDA Library by Doxygen

3.20 Context Management 107

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuCtxCreate, cuCtxDestroy, cuCtxDetach, cuCtxGetDevice, cuCtxPopCurrent, cuCtxPushCurrent, cuCtxSyn-
chronize

3.20.2.2 cuCtxCreate (CUcontext ∗ pctx, unsigned int flags, CUdevice dev)

Creates a new CUDA context and associates it with the calling thread. The flags parameter is described below. The
context is created with a usage count of 1 and the caller of cuCtxCreate() must call cuCtxDestroy() or cuCtxDetach()
when done using the context. If a context is already current to the thread, it is supplanted by the newly created context
and may be restored by a subsequent call to cuCtxPopCurrent().
The two LSBs of the flags parameter can be used to control how the OS thread, which owns the CUDA context at
the time of an API call, interacts with the OS scheduler when waiting for results from the GPU.

• CU_CTX_SCHED_AUTO: The default value if the flags parameter is zero, uses a heuristic based on the
number of active CUDA contexts in the process C and the number of logical processors in the system P. If C >
P, then CUDA will yield to other OS threads when waiting for the GPU, otherwise CUDA will not yield while
waiting for results and actively spin on the processor.

• CU_CTX_SCHED_SPIN: Instruct CUDA to actively spin when waiting for results from the GPU. This can de-
crease latency when waiting for the GPU, but may lower the performance of CPU threads if they are performing
work in parallel with the CUDA thread.

• CU_CTX_SCHED_YIELD: Instruct CUDA to yield its thread when waiting for results from the GPU. This can
increase latency when waiting for the GPU, but can increase the performance of CPU threads performing work
in parallel with the GPU.

• CU_CTX_BLOCKING_SYNC: Instruct CUDA to block the CPU thread on a synchronization primitive when
waiting for the GPU to finish work.

• CU_CTX_MAP_HOST: Instruct CUDA to support mapped pinned allocations. This flag must be set in order
to allocate pinned host memory that is accessible to the GPU.

Note to Linux users:

Context creation will fail with CUDA_ERROR_UNKNOWN if the compute mode of the device is CU_-
COMPUTEMODE_PROHIBITED. Similarly, context creation will also fail with CUDA_ERROR_UNKNOWN if
the compute mode for the device is set to CU_COMPUTEMODE_EXCLUSIVE and there is already an active context
on the device. The function cuDeviceGetAttribute() can be used with CU_DEVICE_ATTRIBUTE_COMPUTE_-
MODE to determine the compute mode of the device. The nvidia-smi tool can be used to set the compute mode for
devices. Documentation for nvidia-smi can be obtained by passing a -h option to it.

Parameters:
pctx - Returned context handle of the new context

Generated for NVIDIA CUDA Library by Doxygen

108 Module Documentation

flags - Context creation flags

dev - Device to create context on

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_DEVICE, CUDA_ERROR_INVALID_VALUE,
CUDA_ERROR_OUT_OF_MEMORY, CUDA_ERROR_UNKNOWN

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuCtxAttach, cuCtxDestroy, cuCtxDetach, cuCtxGetDevice, cuCtxPopCurrent, cuCtxPushCurrent, cuCtxSyn-
chronize

3.20.2.3 cuCtxDestroy (CUcontext ctx)

Destroys the CUDA context specified by ctx. If the context usage count is not equal to 1, or the context is current
to any CPU thread other than the current one, this function fails. Floating contexts (detached from a CPU thread via
cuCtxPopCurrent()) may be destroyed by this function.

Parameters:
ctx - Context to destroy

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuCtxAttach, cuCtxCreate, cuCtxDetach, cuCtxGetDevice, cuCtxPopCurrent, cuCtxPushCurrent, cuCtxSynchro-
nize

3.20.2.4 cuCtxDetach (CUcontext ctx)

Decrements the usage count of the context ctx, and destroys the context if the usage count goes to 0. The context
must be a handle that was passed back by cuCtxCreate() or cuCtxAttach(), and must be current to the calling thread.

Parameters:
ctx - Context to destroy

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT

Generated for NVIDIA CUDA Library by Doxygen

3.20 Context Management 109

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuCtxAttach, cuCtxCreate, cuCtxDestroy, cuCtxGetDevice, cuCtxPopCurrent, cuCtxPushCurrent, cuCtxSyn-
chronize

3.20.2.5 cuCtxGetDevice (CUdevice ∗ device)

Returns in ∗device the ordinal of the current context’s device.

Parameters:
device - Returned device ID for the current context
Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE,

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuCtxAttach, cuCtxCreate, cuCtxDestroy, cuCtxDetach, cuCtxPopCurrent, cuCtxPushCurrent, cuCtxSynchro-
nize

3.20.2.6 cuCtxPopCurrent (CUcontext ∗ pctx)

Pops the current CUDA context from the CPU thread. The CUDA context must have a usage count of 1. CUDA con-
texts have a usage count of 1 upon creation; the usage count may be incremented with cuCtxAttach() and decremented
with cuCtxDetach().
If successful, cuCtxPopCurrent() passes back the new context handle in ∗pctx. The old context may then be made
current to a different CPU thread by calling cuCtxPushCurrent().
Floating contexts may be destroyed by calling cuCtxDestroy().
If a context was current to the CPU thread before cuCtxCreate() or cuCtxPushCurrent() was called, this function makes
that context current to the CPU thread again.

Parameters:
pctx - Returned new context handle

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuCtxAttach, cuCtxCreate, cuCtxDestroy, cuCtxDetach, cuCtxGetDevice, cuCtxPushCurrent, cuCtxSynchronize

Generated for NVIDIA CUDA Library by Doxygen

110 Module Documentation

3.20.2.7 cuCtxPushCurrent (CUcontext ctx)

Pushes the given context ctx onto the CPU thread’s stack of current contexts. The specified context becomes the
CPU thread’s current context, so all CUDA functions that operate on the current context are affected.
The previous current context may be made current again by calling cuCtxDestroy() or cuCtxPopCurrent().
The context must be "floating," i.e. not attached to any thread. Contexts are made to float by calling cuCtxPopCur-
rent().

Parameters:
ctx - Floating context to attach

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuCtxAttach, cuCtxCreate, cuCtxDestroy, cuCtxDetach, cuCtxGetDevice, cuCtxPopCurrent, cuCtxSynchronize

3.20.2.8 cuCtxSynchronize (void)

Blocks until the device has completed all preceding requested tasks. cuCtxSynchronize() returns an error if one of the
preceding tasks failed.

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuCtxAttach, cuCtxCreate, cuCtxDestroy, cuCtxDetach, cuCtxGetDevice, cuCtxPopCurrent, cuCtxSynchronize

Generated for NVIDIA CUDA Library by Doxygen

3.21 Module Management 111

3.21 Module Management

Functions
• CUresult cuModuleGetFunction (CUfunction ∗hfunc, CUmodule hmod, const char ∗name)
Returns a function handle.

• CUresult cuModuleGetGlobal (CUdeviceptr ∗dptr, unsigned int ∗bytes, CUmodule hmod, const char ∗name)
Returns a global pointer from a module.

• CUresult cuModuleGetTexRef (CUtexref ∗pTexRef, CUmodule hmod, const char ∗name)

Returns a handle to a texture-reference.

• CUresult cuModuleLoad (CUmodule ∗module, const char ∗fname)

Loads a compute module.

• CUresult cuModuleLoadData (CUmodule ∗module, const void ∗image)

Load a module’s data.

• CUresult cuModuleLoadDataEx (CUmodule ∗module, const void ∗image, unsigned int numOptions, CUjit_-
option ∗options, void ∗∗optionValues)
Load a module’s data with options.

• CUresult cuModuleLoadFatBinary (CUmodule ∗module, const void ∗fatCubin)

Load a module’s data.

• CUresult cuModuleUnload (CUmodule hmod)

Unloads a module.

3.21.1 Detailed Description

This section describes the module management functions of the low-level CUDA driver application programming
interface.

3.21.2 Function Documentation

3.21.2.1 cuModuleGetFunction (CUfunction ∗ hfunc, CUmodule hmod, const char ∗ name)

Returns in ∗hfunc the handle of the function of name name located in module hmod. If no function of that name
exists, cuModuleGetFunction() returns CUDA_ERROR_NOT_FOUND.

Parameters:
hfunc - Returned function handle
hmod - Module to retrieve function from
name - Name of function to retrieve

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_NOT_FOUND

Generated for NVIDIA CUDA Library by Doxygen

112 Module Documentation

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuModuleGetGlobal, cuModuleGetTexRef, cuModuleLoad, cuModuleLoadData, cuModuleLoadDataEx, cu-
ModuleLoadFatBinary, cuModuleUnload

3.21.2.2 cuModuleGetGlobal (CUdeviceptr ∗ dptr, unsigned int ∗ bytes, CUmodule hmod, const char ∗
name)

Returns in ∗dptr and ∗bytes the base pointer and size of the global of name name located in module hmod. If no
variable of that name exists, cuModuleGetGlobal() returns CUDA_ERROR_NOT_FOUND. Both parameters dptr
and bytes are optional. If one of them is NULL, it is ignored.

Parameters:
dptr - Returned global device pointer
bytes - Returned global size in bytes
hmod - Module to retrieve function from
name - Name of global to retrieve

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_NOT_FOUND

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuModuleGetFunction, cuModuleGetTexRef, cuModuleLoad, cuModuleLoadData, cuModuleLoadDataEx, cu-
ModuleLoadFatBinary, cuModuleUnload

3.21.2.3 cuModuleGetTexRef (CUtexref ∗ pTexRef, CUmodule hmod, const char ∗ name)

Returns in ∗pTexRef the handle of the texture reference of name name in the module hmod. If no texture reference
of that name exists, cuModuleGetTexRef() returns CUDA_ERROR_NOT_FOUND. This texture reference handle
should not be destroyed, since it will be destroyed when the module is unloaded.

Parameters:
pTexRef - Returned global device pointer
hmod - Module to retrieve texture-reference from
name - Name of texture-reference to retrieve

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_NOT_FOUND

Generated for NVIDIA CUDA Library by Doxygen

3.21 Module Management 113

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuModuleGetFunction, cuModuleGetGlobal, cuModuleLoad, cuModuleLoadData, cuModuleLoadDataEx, cu-
ModuleLoadFatBinary, cuModuleUnload

3.21.2.4 cuModuleLoad (CUmodule ∗ module, const char ∗ fname)

Takes a filename fname and loads the corresponding module module into the current context. The CUDA driver API
does not attempt to lazily allocate the resources needed by a module; if the memory for functions and data (constant
and global) needed by the module cannot be allocated, cuModuleLoad() fails. The file should be a cubin file as output
by nvcc or a PTX file, either as output by nvcc or handwrtten.

Parameters:
module - Returned module
fname - Filename of module to load

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_NOT_FOUND,
CUDA_ERROR_OUT_OF_MEMORY, CUDA_ERROR_FILE_NOT_FOUND

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuModuleGetFunction, cuModuleGetGlobal, cuModuleGetTexRef, cuModuleLoadData, cuModuleLoadDataEx,
cuModuleLoadFatBinary, cuModuleUnload

3.21.2.5 cuModuleLoadData (CUmodule ∗ module, const void ∗ image)

Takes a pointer image and loads the corresponding module module into the current context. The pointer may be ob-
tained by mapping a cubin or PTX file, passing a cubin or PTX file as a NULL-terminated text string, or incorporating
a cubin object into the executable resources and using operating system calls such as Windows FindResource()
to obtain the pointer.

Parameters:
module - Returned module
image - Module data to load

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_OUT_OF_MEMORY

Note:
Note that this function may also return error codes from previous, asynchronous launches.

Generated for NVIDIA CUDA Library by Doxygen

114 Module Documentation

See also:
cuModuleGetFunction, cuModuleGetGlobal, cuModuleGetTexRef, cuModuleLoad, cuModuleLoadDataEx, cu-
ModuleLoadFatBinary, cuModuleUnload

3.21.2.6 cuModuleLoadDataEx (CUmodule ∗ module, const void ∗ image, unsigned int numOptions,
CUjit_option ∗ options, void ∗∗ optionValues)

Takes a pointer image and loads the corresponding module module into the current context. The pointer may be ob-
tained by mapping a cubin or PTX file, passing a cubin or PTX file as a NULL-terminated text string, or incorporating
a cubin object into the executable resources and using operating system calls such as Windows FindResource()
to obtain the pointer. Options are passed as an array via options and any corresponding parameters are passed
in optionValues. The number of total options is supplied via numOptions. Any outputs will be returned via
optionValues. Supported options are:

• CU_JIT_MAX_REGISTERS: input specifies the maximum number of registers per thread;

• CU_JIT_THREADS_PER_BLOCK: input specifies number of threads per block to target compilation for; out-
put returns the number of threads the compiler actually targeted;
• CU_JIT_WALL_TIME: output returns the float value of wall clock time, in milliseconds, spent compiling the
PTX code;
• CU_JIT_INFO_LOG_BUFFER: input is a pointer to a buffer in which to print any informational log messages
from PTX assembly;
• CU_JIT_INFO_LOG_BUFFER_SIZE_BYTES: input is the size in bytes of the buffer; output is the number of
bytes filled with messages;
• CU_JIT_ERROR_LOG_BUFFER: input is a pointer to a buffer in which to print any error log messages from
PTX assembly;
• CU_JIT_ERROR_LOG_BUFFER_SIZE_BYTES: input is the size in bytes of the buffer; output is the number
of bytes filled with messages;
• CU_JIT_OPTIMIZATION_LEVEL: input is the level of optimization to apply to generated code (0 - 4), with 4
being the default and highest level;
• CU_JIT_TARGET_FROM_CUCONTEXT: causes compilation target to be determined based on current at-
tached context (default);
• CU_JIT_TARGET: input is the compilation target based on supplied CUjit_target_enum; possible values are:
– CU_TARGET_COMPUTE_10
– CU_TARGET_COMPUTE_11
– CU_TARGET_COMPUTE_12
– CU_TARGET_COMPUTE_13

Parameters:
module - Returned module
image - Module data to load
numOptions - Number of options
options - Options for JIT
optionValues - Option values for JIT

Generated for NVIDIA CUDA Library by Doxygen

3.21 Module Management 115

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_OUT_OF_MEMORY,
CUDA_ERROR_NO_BINARY_FOR_GPU

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuModuleGetFunction, cuModuleGetGlobal, cuModuleGetTexRef, cuModuleLoad, cuModuleLoadData, cu-
ModuleLoadFatBinary, cuModuleUnload

3.21.2.7 cuModuleLoadFatBinary (CUmodule ∗ module, const void ∗ fatCubin)

Takes a pointer fatCubin and loads the corresponding module module into the current context. The pointer
represents a fat binary object, which is a collection of different cubin files, all representing the same device code, but
compiled and optimized for different architectures. There is currently no documented API for constructing and using
fat binary objects by programmers, and therefore this function is an internal function in this version of CUDA. More
information can be found in the nvcc document.

Parameters:
module - Returned module
fatCubin - Fat binary to load

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_NOT_FOUND,
CUDA_ERROR_OUT_OF_MEMORY, CUDA_ERROR_NO_BINARY_FOR_GPU

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuModuleGetFunction, cuModuleGetGlobal, cuModuleGetTexRef, cuModuleLoad, cuModuleLoadData, cu-
ModuleLoadDataEx, cuModuleUnload

3.21.2.8 cuModuleUnload (CUmodule hmod)

Unloads a module hmod from the current context.

Parameters:
hmod - Module to unload

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Generated for NVIDIA CUDA Library by Doxygen

116 Module Documentation

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuModuleGetFunction, cuModuleGetGlobal, cuModuleGetTexRef, cuModuleLoad, cuModuleLoadData, cu-
ModuleLoadDataEx, cuModuleLoadFatBinary

Generated for NVIDIA CUDA Library by Doxygen

3.22 Stream Management 117

3.22 Stream Management

Functions
• CUresult cuStreamCreate (CUstream ∗phStream, unsigned int Flags)
Create a stream.

• CUresult cuStreamDestroy (CUstream hStream)

Destroys a stream.

• CUresult cuStreamQuery (CUstream hStream)

Determine status of a compute stream.

• CUresult cuStreamSynchronize (CUstream hStream)

Wait until a stream’s tasks are completed.

3.22.1 Detailed Description

This section describes the stream management functions of the low-level CUDA driver application programming
interface.

3.22.2 Function Documentation

3.22.2.1 cuStreamCreate (CUstream ∗ phStream, unsigned int Flags)

Creates a stream and returns a handle in phStream. Flags is required to be 0.

Parameters:
phStream - Returned newly created stream
Flags - Parameters for stream creation (must be 0)

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_OUT_OF_MEMORY

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuStreamDestroy, cuStreamQuery, cuStreamSynchronize

3.22.2.2 cuStreamDestroy (CUstream hStream)

Destroys the stream specified by hStream.

Generated for NVIDIA CUDA Library by Doxygen

118 Module Documentation

Parameters:
hStream - Stream to destroy

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuStreamCreate, cuStreamQuery, cuStreamSynchronize

3.22.2.3 cuStreamQuery (CUstream hStream)

Returns CUDA_SUCCESS if all operations in the stream specified by hStream have completed, or CUDA_-
ERROR_NOT_READY if not.

Parameters:
hStream - Stream to query status of

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_HANDLE, CUDA_ERROR_NOT_READY

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuStreamCreate, cuStreamDestroy, cuStreamSynchronize

3.22.2.4 cuStreamSynchronize (CUstream hStream)

Waits until the device has completed all operations in the stream specified by hStream.

Parameters:
hStream - Stream to wait for

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_HANDLE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuStreamCreate, cuStreamDestroy, cuStreamQuery

Generated for NVIDIA CUDA Library by Doxygen

3.23 Event Management 119

3.23 Event Management

Functions
• CUresult cuEventCreate (CUevent ∗phEvent, unsigned int Flags)
Creates an event.

• CUresult cuEventDestroy (CUevent hEvent)

Destroys an event.

• CUresult cuEventElapsedTime (float ∗pMilliseconds, CUevent hStart, CUevent hEnd)

Computes the elapsed time between two events.

• CUresult cuEventQuery (CUevent hEvent)

Queries an event’s status.

• CUresult cuEventRecord (CUevent hEvent, CUstream hStream)

Records an event.

• CUresult cuEventSynchronize (CUevent hEvent)

Waits for an event to complete.

3.23.1 Detailed Description

This section describes the event management functions of the low-level CUDA driver application programming inter-
face.

3.23.2 Function Documentation

3.23.2.1 cuEventCreate (CUevent ∗ phEvent, unsigned int Flags)

Creates an event ∗phEvent with the flags specified via Flags. Valid flags include:

• CU_EVENT_DEFAULT: Default event creation flag

• CU_EVENT_BLOCKING_SYNC: Specifies that event should use blocking synchronization

Parameters:
phEvent - Returns newly created event
Flags - Event creation flags

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_OUT_OF_MEMORY

Note:
Note that this function may also return error codes from previous, asynchronous launches.

Generated for NVIDIA CUDA Library by Doxygen

120 Module Documentation

See also:
cuEventRecord, cuEventQuery, cuEventSynchronize, cuEventDestroy, cuEventElapsedTime

3.23.2.2 cuEventDestroy (CUevent hEvent)

Destroys the event specified by event.

Parameters:
hEvent - Event to destroy

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_HANDLE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuEventCreate, cuEventRecord, cuEventQuery, cuEventSynchronize, cuEventElapsedTime

3.23.2.3 cuEventElapsedTime (float ∗ pMilliseconds, CUevent hStart, CUevent hEnd)

Computes the elapsed time between two events (in milliseconds with a resolution of around 0.5 microseconds). If
either event has not been recorded yet, this function returns CUDA_ERROR_NOT_READY. If either event has been
recorded with a non-zero stream, the result is undefined.

Parameters:
pMilliseconds - Returned elapsed time in milliseconds
hStart - Starting event
hEnd - Ending event

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_HANDLE, CUDA_ERROR_NOT_READY

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuEventCreate, cuEventRecord, cuEventQuery, cuEventSynchronize, cuEventDestroy

Generated for NVIDIA CUDA Library by Doxygen

3.23 Event Management 121

3.23.2.4 cuEventQuery (CUevent hEvent)

Returns CUDA_SUCCESS if the event has actually been recorded, or CUDA_ERROR_NOT_READY if not. If
cuEventRecord() has not been called on this event, the function returns CUDA_ERROR_INVALID_VALUE.

Parameters:
hEvent - Event to query

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_HANDLE, CUDA_ERROR_INVALID_VALUE,
CUDA_ERROR_NOT_READY

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuEventCreate, cuEventRecord, cuEventSynchronize, cuEventDestroy, cuEventElapsedTime

3.23.2.5 cuEventRecord (CUevent hEvent, CUstream hStream)

Records an event. If stream is non-zero, the event is recorded after all preceding operations in the stream have been
completed; otherwise, it is recorded after all preceding operations in the CUDA context have been completed. Since
operation is asynchronous, cuEventQuery() and/or cuEventSynchronize() must be used to determine when the event
has actually been recorded.
If cuEventRecord() has previously been called and the event has not been recorded yet, this function returns CUDA_-
ERROR_INVALID_VALUE.

Parameters:
hEvent - Event to record
hStream - Stream to record event for

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_HANDLE, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuEventCreate, cuEventQuery, cuEventSynchronize, cuEventDestroy, cuEventElapsedTime

3.23.2.6 cuEventSynchronize (CUevent hEvent)

Waits until the event has actually been recorded. If cuEventRecord() has been called on this event, the function returns
CUDA_ERROR_INVALID_VALUE.
If cuEventRecord() has previously been called and the event has not been recorded yet, this function returns CUDA_-
ERROR_INVALID_VALUE.

Generated for NVIDIA CUDA Library by Doxygen

122 Module Documentation

Parameters:
hEvent - Event to wait for

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_HANDLE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuEventCreate, cuEventRecord, cuEventQuery, cuEventDestroy, cuEventElapsedTime

Generated for NVIDIA CUDA Library by Doxygen

3.24 Execution Control 123

3.24 Execution Control

Functions
• CUresult cuFuncGetAttribute (int ∗pi, CUfunction_attribute attrib, CUfunction hfunc)
Returns information about a function.

• CUresult cuFuncSetBlockShape (CUfunction hfunc, int x, int y, int z)

Sets the block-dimensions for the function.

• CUresult cuFuncSetSharedSize (CUfunction hfunc, unsigned int bytes)

Sets the dynamic shared-memory size for the function.

• CUresult cuLaunch (CUfunction f)

Launches a CUDA function.

• CUresult cuLaunchGrid (CUfunction f, int grid_width, int grid_height)

Launches a CUDA function.

• CUresult cuLaunchGridAsync (CUfunction f, int grid_width, int grid_height, CUstream hStream)

Launches a CUDA function.

• CUresult cuParamSetf (CUfunction hfunc, int offset, float value)

Adds a floating-point parameter to the function’s argument list.

• CUresult cuParamSeti (CUfunction hfunc, int offset, unsigned int value)

Adds an integer parameter to the function’s argument list.

• CUresult cuParamSetSize (CUfunction hfunc, unsigned int numbytes)

Sets the parameter size for the function.

• CUresult cuParamSetTexRef (CUfunction hfunc, int texunit, CUtexref hTexRef)

Adds a texture-reference to the function’s argument list.

• CUresult cuParamSetv (CUfunction hfunc, int offset, void ∗ptr, unsigned int numbytes)
Adds arbitrary data to the function’s argument list.

3.24.1 Detailed Description

This section describes the execution control functions of the low-level CUDA driver application programming inter-
face.

3.24.2 Function Documentation

3.24.2.1 cuFuncGetAttribute (int ∗ pi, CUfunction_attribute attrib, CUfunction hfunc)

Returns in ∗pi the integer value of the attribute attrib on the kernel given by hfunc. The supported attributes are:

Generated for NVIDIA CUDA Library by Doxygen

124 Module Documentation

• CU_FUNC_ATTRIBUTE_MAX_THREADS_PER_BLOCK: The number of threads beyond which a launch

of the function would fail. This number depends on both the function and the device on which the function is
currently loaded.
• CU_FUNC_ATTRIBUTE_SHARED_SIZE_BYTES: The size in bytes of statically-allocated shared memory
required by this function. This does not include dynamically-allocated shared memory requested by the user at
runtime.
• CU_FUNC_ATTRIBUTE_CONST_SIZE_BYTES: The size in bytes of user-allocated constant memory re-
quired by this function.
• CU_FUNC_ATTRIBUTE_LOCAL_SIZE_BYTES: The size in bytes of thread local memory used by this func-
tion.
• CU_FUNC_ATTRIBUTE_NUM_REGS: The number of registers used by each thread of this function.

Parameters:
pi - Returned attribute value
attrib - Attribute requested
hfunc - Function to query attribute of

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_HANDLE, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuFuncSetBlockShape, cuFuncSetSharedSize, cuParamSetSize, cuParamSeti, cuParamSetf, cuParamSetv, cu-
ParamSetTexRef, cuLaunch, cuLaunchGrid, cuLaunchGridAsync

3.24.2.2 cuFuncSetBlockShape (CUfunction hfunc, int x, int y, int z)

Specifies the x, y, and z dimensions of the thread blocks that are created when the kernel given by hfunc is launched.

Parameters:
hfunc - Kernel to specify dimensions of
x - X dimension
y - Y dimension
z - Z dimension
Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_HANDLE, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuFuncSetSharedSize, cuFuncGetAttribute, cuParamSetSize, cuParamSeti, cuParamSetf, cuParamSetv, cuParam-
SetTexRef, cuLaunch, cuLaunchGrid, cuLaunchGridAsync

Generated for NVIDIA CUDA Library by Doxygen

3.24 Execution Control 125

3.24.2.3 cuFuncSetSharedSize (CUfunction hfunc, unsigned int bytes)

Sets through bytes the amount of dynamic shared memory that will be available to each thread block when the kernel
given by hfunc is launched.

Parameters:
hfunc - Kernel to specify dynamic shared-memory size for
bytes - Dynamic shared-memory size per thread in bytes

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_HANDLE, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuFuncSetBlockShape, cuFuncGetAttribute, cuParamSetSize, cuParamSeti, cuParamSetf, cuParamSetv, cu-
ParamSetTexRef, cuLaunch, cuLaunchGrid, cuLaunchGridAsync

3.24.2.4 cuLaunch (CUfunction f)

Invokes the kernel f on a 1 x 1 x 1 grid of blocks. The block contains the number of threads specified by a previous
call to cuFuncSetBlockShape().

Parameters:
f - Kernel to launch

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_LAUNCH_FAILED,
CUDA_ERROR_LAUNCH_OUT_OF_RESOURCES, CUDA_ERROR_LAUNCH_TIMEOUT, CUDA_-
ERROR_LAUNCH_INCOMPATIBLE_TEXTURING

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuFuncSetBlockShape, cuFuncSetSharedSize, cuFuncGetAttribute, cuParamSetSize, cuParamSetf, cuParamSeti,
cuParamSetv, cuParamSetTexRef, cuLaunchGrid, cuLaunchGridAsync

3.24.2.5 cuLaunchGrid (CUfunction f, int grid_width, int grid_height)

Invokes the kernel f on a grid_width x grid_height grid of blocks. Each block contains the number of threads
specified by a previous call to cuFuncSetBlockShape().

Generated for NVIDIA CUDA Library by Doxygen

126 Module Documentation

Parameters:
f - Kernel to launch
grid_width - Width of grid in blocks
grid_height - Height of grid in blocks

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_LAUNCH_FAILED,
CUDA_ERROR_LAUNCH_OUT_OF_RESOURCES, CUDA_ERROR_LAUNCH_TIMEOUT, CUDA_-
ERROR_LAUNCH_INCOMPATIBLE_TEXTURING

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuFuncSetBlockShape, cuFuncSetSharedSize, cuFuncGetAttribute, cuParamSetSize, cuParamSetf, cuParamSeti,
cuParamSetv, cuParamSetTexRef, cuLaunch, cuLaunchGridAsync

3.24.2.6 cuLaunchGridAsync (CUfunction f, int grid_width, int grid_height, CUstream hStream)

Invokes the kernel f on a grid_width x grid_height grid of blocks. Each block contains the number of threads
specified by a previous call to cuFuncSetBlockShape().
cuLaunchGridAsync() can optionally be associated to a stream by passing a non-zero hStream argument.

Parameters:
f - Kernel to launch
grid_width - Width of grid in blocks
grid_height - Height of grid in blocks
hStream - Stream identifier

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_LAUNCH_FAILED,
CUDA_ERROR_LAUNCH_OUT_OF_RESOURCES, CUDA_ERROR_LAUNCH_TIMEOUT, CUDA_-
ERROR_LAUNCH_INCOMPATIBLE_TEXTURING

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuFuncSetBlockShape, cuFuncSetSharedSize, cuFuncGetAttribute, cuParamSetSize, cuParamSetf, cuParamSeti,
cuParamSetv, cuParamSetTexRef, cuLaunch, cuLaunchGrid

Generated for NVIDIA CUDA Library by Doxygen

3.24 Execution Control 127

3.24.2.7 cuParamSetf (CUfunction hfunc, int offset, float value)

Sets a floating-point parameter that will be specified the next time the kernel corresponding to hfunc will be invoked.
offset is a byte offset.

Parameters:
hfunc - Kernel to add parameter to
offset - Offset to add parameter to argument list
value - Value of parameter

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuFuncSetBlockShape, cuFuncSetSharedSize, cuFuncGetAttribute, cuParamSetSize, cuParamSeti, cuParamSetv,
cuParamSetTexRef, cuLaunch, cuLaunchGrid, cuLaunchGridAsync

3.24.2.8 cuParamSeti (CUfunction hfunc, int offset, unsigned int value)

Sets an integer parameter that will be specified the next time the kernel corresponding to hfunc will be invoked.
offset is a byte offset.

Parameters:
hfunc - Kernel to add parameter to
offset - Offset to add parameter to argument list
value - Value of parameter

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuFuncSetBlockShape, cuFuncSetSharedSize, cuFuncGetAttribute, cuParamSetSize, cuParamSetf, cuParamSetv,
cuParamSetTexRef, cuLaunch, cuLaunchGrid, cuLaunchGridAsync

Generated for NVIDIA CUDA Library by Doxygen

128 Module Documentation

3.24.2.9 cuParamSetSize (CUfunction hfunc, unsigned int numbytes)

Sets through numbytes the total size in bytes needed by the function parameters of the kernel corresponding to
hfunc.

Parameters:
hfunc - Kernel to set parameter size for
numbytes - Size of parameter list in bytes

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuFuncSetBlockShape, cuFuncSetSharedSize, cuFuncGetAttribute, cuParamSetf, cuParamSeti, cuParamSetv, cu-
ParamSetTexRef, cuLaunch, cuLaunchGrid, cuLaunchGridAsync

3.24.2.10 cuParamSetTexRef (CUfunction hfunc, int texunit, CUtexref hTexRef)

Makes the CUDA array or linear memory bound to the texture reference hTexRef available to a device program as a
texture. In this version of CUDA, the texture-reference must be obtained via cuModuleGetTexRef() and the texunit
parameter must be set to CU_PARAM_TR_DEFAULT.

Parameters:
hfunc - Kernel to add texture-reference to
texunit - Texture unit (must be CU_PARAM_TR_DEFAULT)
hTexRef - Texture-reference to add to argument list

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuFuncSetBlockShape, cuFuncSetSharedSize, cuFuncGetAttribute, cuParamSetSize, cuParamSetf, cuParamSeti,
cuParamSetv, cuLaunch, cuLaunchGrid, cuLaunchGridAsync

3.24.2.11 cuParamSetv (CUfunction hfunc, int offset, void ∗ ptr, unsigned int numbytes)

Copies an arbitrary amount of data (specified in numbytes) from ptr into the parameter space of the kernel corre-
sponding to hfunc. offset is a byte offset.

Generated for NVIDIA CUDA Library by Doxygen

3.24 Execution Control 129

Parameters:
hfunc - Kernel to add data to
offset - Offset to add data to argument list
ptr - Pointer to arbitrary data
numbytes - Size of data to copy in bytes

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuFuncSetBlockShape, cuFuncSetSharedSize, cuFuncGetAttribute, cuParamSetSize, cuParamSetf, cuParamSeti,
cuParamSetTexRef, cuLaunch, cuLaunchGrid, cuLaunchGridAsync

Generated for NVIDIA CUDA Library by Doxygen

130 Module Documentation

3.25 Memory Management

Functions
• CUresult cuArray3DCreate (CUarray ∗pHandle, const CUDA_ARRAY3D_DESCRIPTOR ∗pAllocateArray)
Creates a 3D CUDA array.

• CUresult cuArray3DGetDescriptor (CUDA_ARRAY3D_DESCRIPTOR ∗pArrayDescriptor, CUarray hAr-

ray)
Get a 3D CUDA array descriptor.

• CUresult cuArrayCreate (CUarray ∗pHandle, const CUDA_ARRAY_DESCRIPTOR ∗pAllocateArray)

Creates a 1D or 2D CUDA array.

• CUresult cuArrayDestroy (CUarray hArray)

Destroys a CUDA array.

• CUresult cuArrayGetDescriptor (CUDA_ARRAY_DESCRIPTOR ∗pArrayDescriptor, CUarray hArray)

Get a 1D or 2D CUDA array descriptor.

• CUresult cuMemAlloc (CUdeviceptr ∗dptr, unsigned int bytesize)

Allocates device memory.

• CUresult cuMemAllocHost (void ∗∗pp, unsigned int bytesize)

Allocates page-locked host memory.

• CUresult cuMemAllocPitch (CUdeviceptr ∗dptr, unsigned int ∗pPitch, unsigned int WidthInBytes, unsigned int
Height, unsigned int ElementSizeBytes)
Allocates pitched device memory.

• CUresult cuMemcpy2D (const CUDA_MEMCPY2D ∗pCopy)

Copies memory for 2D arrays.

• CUresult cuMemcpy2DAsync (const CUDA_MEMCPY2D ∗pCopy, CUstream hStream)

Copies memory for 2D arrays.

• CUresult cuMemcpy2DUnaligned (const CUDA_MEMCPY2D ∗pCopy)

Copies memory for 2D arrays.

• CUresult cuMemcpy3D (const CUDA_MEMCPY3D ∗pCopy)

Copies memory for 3D arrays.

• CUresult cuMemcpy3DAsync (const CUDA_MEMCPY3D ∗pCopy, CUstream hStream)

Copies memory for 3D arrays.

• CUresult cuMemcpyAtoA (CUarray dstArray, unsigned int dstIndex, CUarray srcArray, unsigned int srcIndex,
unsigned int ByteCount)
Copies memory from Array to Array.

• CUresult cuMemcpyAtoD (CUdeviceptr dstDevice, CUarray hSrc, unsigned int SrcIndex, unsigned int Byte-
Count)

Generated for NVIDIA CUDA Library by Doxygen

3.25 Memory Management 131

Copies memory from Array to Device.

• CUresult cuMemcpyAtoH (void ∗dstHost, CUarray srcArray, unsigned int srcIndex, unsigned int ByteCount)
Copies memory from Array to Host.

• CUresult cuMemcpyAtoHAsync (void ∗dstHost, CUarray srcArray, unsigned int srcIndex, unsigned int Byte-
Count, CUstream hStream)
Copies memory from Array to Host.

• CUresult cuMemcpyDtoA (CUarray dstArray, unsigned int dstIndex, CUdeviceptr srcDevice, unsigned int Byte-
Count)
Copies memory from Device to Array.

• CUresult cuMemcpyDtoD (CUdeviceptr dstDevice, CUdeviceptr srcDevice, unsigned int ByteCount)

Copies memory from Device to Device.

• CUresult cuMemcpyDtoH (void ∗dstHost, CUdeviceptr srcDevice, unsigned int ByteCount)

Copies memory from Device to Host.

• CUresult cuMemcpyDtoHAsync (void ∗dstHost, CUdeviceptr srcDevice, unsigned int ByteCount, CUstream
hStream)
Copies memory from Device to Host.

• CUresult cuMemcpyHtoA (CUarray dstArray, unsigned int dstIndex, const void ∗pSrc, unsigned int Byte-
Count)
Copies memory from Host to Array.

• CUresult cuMemcpyHtoAAsync (CUarray dstArray, unsigned int dstIndex, const void ∗pSrc, unsigned int Byte-
Count, CUstream hStream)
Copies memory from Host to Array.

• CUresult cuMemcpyHtoD (CUdeviceptr dstDevice, const void ∗srcHost, unsigned int ByteCount)
Copies memory from Host to Device.

• CUresult cuMemcpyHtoDAsync (CUdeviceptr dstDevice, const void ∗srcHost, unsigned int ByteCount, CUs-
tream hStream)
Copies memory from Host to Device.

• CUresult cuMemFree (CUdeviceptr dptr)

Frees device memory.

• CUresult cuMemFreeHost (void ∗p)

Frees page-locked host memory.

• CUresult cuMemGetAddressRange (CUdeviceptr ∗pbase, unsigned int ∗psize, CUdeviceptr dptr)

Get information on memory allocations.

• CUresult cuMemGetInfo (unsigned int ∗free, unsigned int ∗total)

Gets free and total memory.

• CUresult cuMemHostAlloc (void ∗∗pp, size_t bytesize, unsigned int Flags)

Generated for NVIDIA CUDA Library by Doxygen

132 Module Documentation

Allocates page-locked host memory.

• CUresult cuMemHostGetDevicePointer (CUdeviceptr ∗pdptr, void ∗p, unsigned int Flags)

Passes back device pointer of mapped pinned memory.

• CUresult cuMemHostGetFlags (unsigned int ∗pFlags, void ∗p)

Passes back flags that were used for a pinned allocation.

• CUresult cuMemsetD16 (CUdeviceptr dstDevice, unsigned short us, unsigned int N)

Initializes device memory.

• CUresult cuMemsetD2D16 (CUdeviceptr dstDevice, unsigned int dstPitch, unsigned short us, unsigned int
Width, unsigned int Height)
Initializes device memory.

• CUresult cuMemsetD2D32 (CUdeviceptr dstDevice, unsigned int dstPitch, unsigned int ui, unsigned int Width,
unsigned int Height)
Initializes device memory.

• CUresult cuMemsetD2D8 (CUdeviceptr dstDevice, unsigned int dstPitch, unsigned char uc, unsigned int Width,
unsigned int Height)
Initializes device memory.

• CUresult cuMemsetD32 (CUdeviceptr dstDevice, unsigned int ui, unsigned int N)

Initializes device memory.

• CUresult cuMemsetD8 (CUdeviceptr dstDevice, unsigned char uc, unsigned int N)

Initializes device memory.

3.25.1 Detailed Description

This section describes the memory management functions of the low-level CUDA driver application programming
interface.

3.25.2 Function Documentation

3.25.2.1 cuArray3DCreate (CUarray ∗ pHandle, const CUDA_ARRAY3D_DESCRIPTOR ∗ pAllocateArray)

Creates a CUDA array according to the CUDA_ARRAY3D_DESCRIPTOR structure pAllocateArray and returns
a handle to the new CUDA array in ∗pHandle. The CUDA_ARRAY3D_DESCRIPTOR is defined as:

typedef struct {
unsigned int Width;
unsigned int Height;
unsigned int Depth;
CUarray_format Format;
unsigned int NumChannels;
unsigned int Flags;
} CUDA_ARRAY3D_DESCRIPTOR;

where:

Generated for NVIDIA CUDA Library by Doxygen

3.25 Memory Management 133

• Width, Height, and Depth are the width, height, and depth of the CUDA array (in elements); the CUDA
array is one-dimensional if height and depth are 0, two-dimensional if depth is 0, and three-dimensional other-
wise;
• Format specifies the format of the elements; CUarray_format is defined as:

typedef enum CUarray_format_enum {

CU_AD_FORMAT_UNSIGNED_INT8 = 0x01,
CU_AD_FORMAT_UNSIGNED_INT16 = 0x02,
CU_AD_FORMAT_UNSIGNED_INT32 = 0x03,
CU_AD_FORMAT_SIGNED_INT8 = 0x08,
CU_AD_FORMAT_SIGNED_INT16 = 0x09,
CU_AD_FORMAT_SIGNED_INT32 = 0x0a,
CU_AD_FORMAT_HALF = 0x10,
CU_AD_FORMAT_FLOAT = 0x20
} CUarray_format;

• NumChannels specifies the number of packed components per CUDA array element; it may be 1, 2, or 4;
• Flags provides for future features. For now, it must be set to 0.

Here are examples of CUDA array descriptions:

Description for a CUDA array of 2048 floats:

CUDA_ARRAY3D_DESCRIPTOR desc;
desc.Format = CU_AD_FORMAT_FLOAT;
desc.NumChannels = 1;
desc.Width = 2048;
desc.Height = 0;
desc.Depth = 0;

Description for a 64 x 64 CUDA array of floats:

CUDA_ARRAY3D_DESCRIPTOR desc;
desc.Format = CU_AD_FORMAT_FLOAT;
desc.NumChannels = 1;
desc.Width = 64;
desc.Height = 64;
desc.Depth = 0;

Description for a width x height x depth CUDA array of 64-bit, 4x16-bit float16’s:

CUDA_ARRAY3D_DESCRIPTOR desc;
desc.FormatFlags = CU_AD_FORMAT_HALF;
desc.NumChannels = 4;
desc.Width = width;
desc.Height = height;
desc.Depth = depth;

Parameters:
pHandle - Returned array
pAllocateArray - 3D array descriptor

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_OUT_OF_MEMORY,
CUDA_ERROR_UNKNOWN

Generated for NVIDIA CUDA Library by Doxygen

134 Module Documentation

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuArray3DGetDescriptor, cuArrayCreate, cuArrayDestroy, cuArrayGetDescriptor, cuMemAlloc, cuMemAl-
locHost, cuMemAllocPitch, cuMemcpy2D, cuMemcpy2DAsync, cuMemcpy2DUnaligned, cuMemcpy3D,
cuMemcpy3DAsync, cuMemcpyAtoA, cuMemcpyAtoD, cuMemcpyAtoH, cuMemcpyAtoHAsync, cuMemcpy-
DtoA, cuMemcpyDtoD, cuMemcpyDtoH, cuMemcpyDtoHAsync, cuMemcpyHtoA, cuMemcpyHtoAAsync,
cuMemcpyHtoD, cuMemcpyHtoDAsync, cuMemFree, cuMemFreeHost, cuMemGetAddressRange, cuMemGet-
Info, cuMemHostAlloc, cuMemHostGetDevicePointer, cuMemsetD2D8, cuMemsetD2D16, cuMemsetD2D32,
cuMemsetD8, cuMemsetD16, cuMemsetD32

3.25.2.2 cuArray3DGetDescriptor (CUDA_ARRAY3D_DESCRIPTOR ∗ pArrayDescriptor, CUarray

hArray)

Returns in ∗pArrayDescriptor a descriptor containing information on the format and dimensions of the CUDA
array hArray. It is useful for subroutines that have been passed a CUDA array, but need to know the CUDA array
parameters for validation or other purposes.
This function may be called on 1D and 2D arrays, in which case the Height and/or Depth members of the descriptor
struct will be set to 0.

Parameters:
pArrayDescriptor - Returned 3D array descriptor
hArray - 3D array to get descriptor of

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_INVALID_HANDLE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuArray3DCreate, cuArrayCreate, cuArrayDestroy, cuArrayGetDescriptor, cuMemAlloc, cuMemAllocHost,
cuMemAllocPitch, cuMemcpy2D, cuMemcpy2DAsync, cuMemcpy2DUnaligned, cuMemcpy3D, cuMem-
cpy3DAsync, cuMemcpyAtoA, cuMemcpyAtoD, cuMemcpyAtoH, cuMemcpyAtoHAsync, cuMemcpyD-
toA, cuMemcpyDtoD, cuMemcpyDtoH, cuMemcpyDtoHAsync, cuMemcpyHtoA, cuMemcpyHtoAAsync,
cuMemcpyHtoD, cuMemcpyHtoDAsync, cuMemFree, cuMemFreeHost, cuMemGetAddressRange, cuMemGet-
Info, cuMemHostAlloc, cuMemHostGetDevicePointer, cuMemsetD2D8, cuMemsetD2D16, cuMemsetD2D32,
cuMemsetD8, cuMemsetD16, cuMemsetD32

3.25.2.3 cuArrayCreate (CUarray ∗ pHandle, const CUDA_ARRAY_DESCRIPTOR ∗ pAllocateArray)

Creates a CUDA array according to the CUDA_ARRAY_DESCRIPTOR structure pAllocateArray and returns a
handle to the new CUDA array in ∗pHandle. The CUDA_ARRAY_DESCRIPTOR is defined as:

Generated for NVIDIA CUDA Library by Doxygen

3.25 Memory Management 135

typedef struct {
unsigned int Width;
unsigned int Height;
CUarray_format Format;
unsigned int NumChannels;
} CUDA_ARRAY_DESCRIPTOR;

where:

• Width, and Height are the width, and height of the CUDA array (in elements); the CUDA array is one-
dimensional if height is 0, two-dimensional otherwise;
• Format specifies the format of the elements; CUarray_format is defined as:

typedef enum CUarray_format_enum {

CU_AD_FORMAT_UNSIGNED_INT8 = 0x01,
CU_AD_FORMAT_UNSIGNED_INT16 = 0x02,
CU_AD_FORMAT_UNSIGNED_INT32 = 0x03,
CU_AD_FORMAT_SIGNED_INT8 = 0x08,
CU_AD_FORMAT_SIGNED_INT16 = 0x09,
CU_AD_FORMAT_SIGNED_INT32 = 0x0a,
CU_AD_FORMAT_HALF = 0x10,
CU_AD_FORMAT_FLOAT = 0x20
} CUarray_format;

• NumChannels specifies the number of packed components per CUDA array element; it may be 1, 2, or 4;

Here are examples of CUDA array descriptions:

Description for a CUDA array of 2048 floats:

CUDA_ARRAY_DESCRIPTOR desc;
desc.Format = CU_AD_FORMAT_FLOAT;
desc.NumChannels = 1;
desc.Width = 2048;
desc.Height = 1;

Description for a 64 x 64 CUDA array of floats:

CUDA_ARRAY_DESCRIPTOR desc;
desc.Format = CU_AD_FORMAT_FLOAT;
desc.NumChannels = 1;
desc.Width = 64;
desc.Height = 64;

Description for a width x height CUDA array of 64-bit, 4x16-bit float16’s:

CUDA_ARRAY_DESCRIPTOR desc;
desc.FormatFlags = CU_AD_FORMAT_HALF;
desc.NumChannels = 4;
desc.Width = width;
desc.Height = height;

Description for a width x height CUDA array of 16-bit elements, each of which is two 8-bit unsigned chars:

CUDA_ARRAY_DESCRIPTOR arrayDesc;
desc.FormatFlags = CU_AD_FORMAT_UNSIGNED_INT8;
desc.NumChannels = 2;
desc.Width = width;
desc.Height = height;

Generated for NVIDIA CUDA Library by Doxygen

136 Module Documentation

Parameters:
pHandle - Returned array
pAllocateArray - Array descriptor

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_OUT_OF_MEMORY,
CUDA_ERROR_UNKNOWN

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuArray3DCreate, cuArray3DGetDescriptor, cuArrayDestroy, cuArrayGetDescriptor, cuMemAlloc, cuMemAl-
locHost, cuMemAllocPitch, cuMemcpy2D, cuMemcpy2DAsync, cuMemcpy2DUnaligned, cuMemcpy3D,
cuMemcpy3DAsync, cuMemcpyAtoA, cuMemcpyAtoD, cuMemcpyAtoH, cuMemcpyAtoHAsync, cuMemcpy-
DtoA, cuMemcpyDtoD, cuMemcpyDtoH, cuMemcpyDtoHAsync, cuMemcpyHtoA, cuMemcpyHtoAAsync,
cuMemcpyHtoD, cuMemcpyHtoDAsync, cuMemFree, cuMemFreeHost, cuMemGetAddressRange, cuMemGet-
Info, cuMemHostAlloc, cuMemHostGetDevicePointer, cuMemsetD2D8, cuMemsetD2D16, cuMemsetD2D32,
cuMemsetD8, cuMemsetD16, cuMemsetD32

3.25.2.4 cuArrayDestroy (CUarray hArray)

Destroys the CUDA array hArray.

Parameters:
hArray - Array to destroy

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_HANDLE, CUDA_ERROR_ARRAY_IS_-
MAPPED

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuArray3DCreate, cuArray3DGetDescriptor, cuArrayCreate, cuArrayGetDescriptor, cuMemAlloc, cuMemAl-
locHost, cuMemAllocPitch, cuMemcpy2D, cuMemcpy2DAsync, cuMemcpy2DUnaligned, cuMemcpy3D,
cuMemcpy3DAsync, cuMemcpyAtoA, cuMemcpyAtoD, cuMemcpyAtoH, cuMemcpyAtoHAsync, cuMemcpy-
DtoA, cuMemcpyDtoD, cuMemcpyDtoH, cuMemcpyDtoHAsync, cuMemcpyHtoA, cuMemcpyHtoAAsync,
cuMemcpyHtoD, cuMemcpyHtoDAsync, cuMemFree, cuMemFreeHost, cuMemGetAddressRange, cuMemGet-
Info, cuMemHostAlloc, cuMemHostGetDevicePointer, cuMemsetD2D8, cuMemsetD2D16, cuMemsetD2D32,
cuMemsetD8, cuMemsetD16, cuMemsetD32

Generated for NVIDIA CUDA Library by Doxygen

3.25 Memory Management 137

3.25.2.5 cuArrayGetDescriptor (CUDA_ARRAY_DESCRIPTOR ∗ pArrayDescriptor, CUarray hArray)

Returns in ∗pArrayDescriptor a descriptor containing information on the format and dimensions of the CUDA
array hArray. It is useful for subroutines that have been passed a CUDA array, but need to know the CUDA array
parameters for validation or other purposes.

Parameters:
pArrayDescriptor - Returned array descriptor
hArray - Array to get descriptor of

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_INVALID_HANDLE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuArray3DCreate, cuArray3DGetDescriptor, cuArrayCreate, cuArrayDestroy, cuMemAlloc, cuMemAllocHost,
cuMemAllocPitch, cuMemcpy2D, cuMemcpy2DAsync, cuMemcpy2DUnaligned, cuMemcpy3D, cuMem-
cpy3DAsync, cuMemcpyAtoA, cuMemcpyAtoD, cuMemcpyAtoH, cuMemcpyAtoHAsync, cuMemcpyD-
toA, cuMemcpyDtoD, cuMemcpyDtoH, cuMemcpyDtoHAsync, cuMemcpyHtoA, cuMemcpyHtoAAsync,
cuMemcpyHtoD, cuMemcpyHtoDAsync, cuMemFree, cuMemFreeHost, cuMemGetAddressRange, cuMemGet-
Info, cuMemHostAlloc, cuMemHostGetDevicePointer, cuMemsetD2D8, cuMemsetD2D16, cuMemsetD2D32,
cuMemsetD8, cuMemsetD16, cuMemsetD32

3.25.2.6 cuMemAlloc (CUdeviceptr ∗ dptr, unsigned int bytesize)

Allocates bytesize bytes of linear memory on the device and returns in ∗dptr a pointer to the allocated memory.
The allocated memory is suitably aligned for any kind of variable. The memory is not cleared. If bytesize is 0,
cuMemAlloc() returns CUDA_ERROR_INVALID_VALUE.

Parameters:
dptr - Returned device pointer
bytesize - Requested allocation size in bytes

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_OUT_OF_MEMORY

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuArray3DCreate, cuArray3DGetDescriptor, cuArrayCreate, cuArrayDestroy, cuArrayGetDescriptor, cuMemAl-
locHost, cuMemAllocPitch, cuMemcpy2D, cuMemcpy2DAsync, cuMemcpy2DUnaligned, cuMemcpy3D,
cuMemcpy3DAsync, cuMemcpyAtoA, cuMemcpyAtoD, cuMemcpyAtoH, cuMemcpyAtoHAsync, cuMemcpy-
DtoA, cuMemcpyDtoD, cuMemcpyDtoH, cuMemcpyDtoHAsync, cuMemcpyHtoA, cuMemcpyHtoAAsync,

Generated for NVIDIA CUDA Library by Doxygen

138 Module Documentation

cuMemcpyHtoD, cuMemcpyHtoDAsync, cuMemFree, cuMemFreeHost, cuMemGetAddressRange, cuMemGet-

Info, cuMemHostAlloc, cuMemHostGetDevicePointer, cuMemsetD2D8, cuMemsetD2D16, cuMemsetD2D32,
cuMemsetD8, cuMemsetD16, cuMemsetD32

3.25.2.7 cuMemAllocHost (void ∗∗ pp, unsigned int bytesize)

Allocates bytesize bytes of host memory that is page-locked and accessible to the device. The driver tracks the vir-
tual memory ranges allocated with this function and automatically accelerates calls to functions such as cuMemcpy().
Since the memory can be accessed directly by the device, it can be read or written with much higher bandwidth than
pageable memory obtained with functions such as malloc(). Allocating excessive amounts of memory with cuMemAl-
locHost() may degrade system performance, since it reduces the amount of memory available to the system for paging.
As a result, this function is best used sparingly to allocate staging areas for data exchange between host and device.

Parameters:
pp - Returned host pointer to page-locked memory
bytesize - Requested allocation size in bytes

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_OUT_OF_MEMORY

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuArray3DCreate, cuArray3DGetDescriptor, cuArrayCreate, cuArrayDestroy, cuArrayGetDescriptor, cuMemAl-
loc, cuMemAllocPitch, cuMemcpy2D, cuMemcpy2DAsync, cuMemcpy2DUnaligned, cuMemcpy3D, cuMem-
cpy3DAsync, cuMemcpyAtoA, cuMemcpyAtoD, cuMemcpyAtoH, cuMemcpyAtoHAsync, cuMemcpyD-
toA, cuMemcpyDtoD, cuMemcpyDtoH, cuMemcpyDtoHAsync, cuMemcpyHtoA, cuMemcpyHtoAAsync,
cuMemcpyHtoD, cuMemcpyHtoDAsync, cuMemFree, cuMemFreeHost, cuMemGetAddressRange, cuMemGet-
Info, cuMemHostAlloc, cuMemHostGetDevicePointer, cuMemsetD2D8, cuMemsetD2D16, cuMemsetD2D32,
cuMemsetD8, cuMemsetD16, cuMemsetD32

3.25.2.8 cuMemAllocPitch (CUdeviceptr ∗ dptr, unsigned int ∗ pPitch, unsigned int WidthInBytes, unsigned
int Height, unsigned int ElementSizeBytes)

Allocates at least WidthInBytes ∗ Height bytes of linear memory on the device and returns in ∗dptr a pointer
to the allocated memory. The function may pad the allocation to ensure that corresponding pointers in any given
row will continue to meet the alignment requirements for coalescing as the address is updated from row to row.
ElementSizeBytes specifies the size of the largest reads and writes that will be performed on the memory range.
ElementSizeBytes may be 4, 8 or 16 (since coalesced memory transactions are not possible on other data sizes). If
ElementSizeBytes is smaller than the actual read/write size of a kernel, the kernel will run correctly, but possibly
at reduced speed. The pitch returned in ∗pPitch by cuMemAllocPitch() is the width in bytes of the allocation. The
intended usage of pitch is as a separate parameter of the allocation, used to compute addresses within the 2D array.
Given the row and column of an array element of type T, the address is computed as:

T* pElement = (T*)((char*)BaseAddress + Row * Pitch) + Column;

Generated for NVIDIA CUDA Library by Doxygen

3.25 Memory Management 139

The pitch returned by cuMemAllocPitch() is guaranteed to work with cuMemcpy2D() under all circumstances. For
allocations of 2D arrays, it is recommended that programmers consider performing pitch allocations using cuMemAl-
locPitch(). Due to alignment restrictions in the hardware, this is especially true if the application will be performing
2D memory copies between different regions of device memory (whether linear memory or CUDA arrays).

Parameters:
dptr - Returned device pointer
pPitch - Returned pitch of allocation in bytes
WidthInBytes - Requested allocation width in bytes
Height - Requested allocation height in rows
ElementSizeBytes - Size of largest reads/writes for range

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_OUT_OF_MEMORY

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuArray3DCreate, cuArray3DGetDescriptor, cuArrayCreate, cuArrayDestroy, cuArrayGetDescriptor, cuMemAl-
loc, cuMemAllocHost, cuMemcpy2D, cuMemcpy2DAsync, cuMemcpy2DUnaligned, cuMemcpy3D, cuMem-
cpy3DAsync, cuMemcpyAtoA, cuMemcpyAtoD, cuMemcpyAtoH, cuMemcpyAtoHAsync, cuMemcpyD-
toA, cuMemcpyDtoD, cuMemcpyDtoH, cuMemcpyDtoHAsync, cuMemcpyHtoA, cuMemcpyHtoAAsync,
cuMemcpyHtoD, cuMemcpyHtoDAsync, cuMemFree, cuMemFreeHost, cuMemGetAddressRange, cuMemGet-
Info, cuMemHostAlloc, cuMemHostGetDevicePointer, cuMemsetD2D8, cuMemsetD2D16, cuMemsetD2D32,
cuMemsetD8, cuMemsetD16, cuMemsetD32

3.25.2.9 cuMemcpy2D (const CUDA_MEMCPY2D ∗ pCopy)

Perform a 2D memory copy according to the parameters specified in pCopy. The CUDA_MEMCPY2D structure is
defined as:

typedef struct CUDA_MEMCPY2D_st {

unsigned int srcXInBytes, srcY;
CUmemorytype srcMemoryType;
const void *srcHost;
CUdeviceptr srcDevice;
CUarray srcArray;
unsigned int srcPitch;

unsigned int dstXInBytes, dstY;

CUmemorytype dstMemoryType;
void *dstHost;
CUdeviceptr dstDevice;
CUarray dstArray;
unsigned int dstPitch;

unsigned int WidthInBytes;

unsigned int Height;
} CUDA_MEMCPY2D;

where:

Generated for NVIDIA CUDA Library by Doxygen

140 Module Documentation

• srcMemoryType and dstMemoryType specify the type of memory of the source and destination, respectively;
CUmemorytype_enum is defined as:

typedef enum CUmemorytype_enum {

CU_MEMORYTYPE_HOST = 0x01,
CU_MEMORYTYPE_DEVICE = 0x02,
CU_MEMORYTYPE_ARRAY = 0x03
} CUmemorytype;

If srcMemoryType is CU_MEMORYTYPE_HOST, srcHost and srcPitch specify the (host) base address of the
source data and the bytes per row to apply. srcArray is ignored.

If srcMemoryType is CU_MEMORYTYPE_DEVICE, srcDevice and srcPitch specify the (device) base address
of the source data and the bytes per row to apply. srcArray is ignored.

If srcMemoryType is CU_MEMORYTYPE_ARRAY, srcArray specifies the handle of the source data. srcHost,
srcDevice and srcPitch are ignored.

If dstMemoryType is CU_MEMORYTYPE_HOST, dstHost and dstPitch specify the (host) base address of the
destination data and the bytes per row to apply. dstArray is ignored.

If dstMemoryType is CU_MEMORYTYPE_DEVICE, dstDevice and dstPitch specify the (device) base address
of the destination data and the bytes per row to apply. dstArray is ignored.

If dstMemoryType is CU_MEMORYTYPE_ARRAY, dstArray specifies the handle of the destination data.

dstHost, dstDevice and dstPitch are ignored.

• srcXInBytes and srcY specify the base address of the source data for the copy.

For host pointers, the starting address is

void* Start = (void*)((char*)srcHost+srcY*srcPitch + srcXInBytes);

For device pointers, the starting address is

CUdeviceptr Start = srcDevice+srcY*srcPitch+srcXInBytes;

For CUDA arrays, srcXInBytes must be evenly divisible by the array element size.

• dstXInBytes and dstY specify the base address of the destination data for the copy.

For host pointers, the base address is

void* dstStart = (void*)((char*)dstHost+dstY*dstPitch + dstXInBytes);

For device pointers, the starting address is

Generated for NVIDIA CUDA Library by Doxygen

3.25 Memory Management 141

CUdeviceptr dstStart = dstDevice+dstY*dstPitch+dstXInBytes;

For CUDA arrays, dstXInBytes must be evenly divisible by the array element size.

• WidthInBytes and Height specify the width (in bytes) and height of the 2D copy being performed. Any pitches
must be greater than or equal to WidthInBytes.

cuMemcpy2D() returns an error if any pitch is greater than the maximum allowed (CU_DEVICE_ATTRIBUTE_-
MAX_PITCH). cuMemAllocPitch() passes back pitches that always work with cuMemcpy2D(). On intra-device
memory copies (device ? device, CUDA array ? device, CUDA array ? CUDA array), cuMemcpy2D() may fail
for pitches not computed by cuMemAllocPitch(). cuMemcpy2DUnaligned() does not have this restriction, but
may run significantly slower in the cases where cuMemcpy2D() would have returned an error code.

Parameters:
pCopy - Parameters for the memory copy

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuArray3DCreate, cuArray3DGetDescriptor, cuArrayCreate, cuArrayDestroy, cuArrayGetDescriptor, cuMemAl-
loc, cuMemAllocHost, cuMemAllocPitch, cuMemcpy2DAsync, cuMemcpy2DUnaligned, cuMemcpy3D,
cuMemcpy3DAsync, cuMemcpyAtoA, cuMemcpyAtoD, cuMemcpyAtoH, cuMemcpyAtoHAsync, cuMemcpy-
DtoA, cuMemcpyDtoD, cuMemcpyDtoH, cuMemcpyDtoHAsync, cuMemcpyHtoA, cuMemcpyHtoAAsync,
cuMemcpyHtoD, cuMemcpyHtoDAsync, cuMemFree, cuMemFreeHost, cuMemGetAddressRange, cuMemGet-
Info, cuMemHostAlloc, cuMemHostGetDevicePointer, cuMemsetD2D8, cuMemsetD2D16, cuMemsetD2D32,
cuMemsetD8, cuMemsetD16, cuMemsetD32

3.25.2.10 cuMemcpy2DAsync (const CUDA_MEMCPY2D ∗ pCopy, CUstream hStream)

Perform a 2D memory copy according to the parameters specified in pCopy. The CUDA_MEMCPY2D structure is
defined as:

typedef struct CUDA_MEMCPY2D_st {

unsigned int srcXInBytes, srcY;
CUmemorytype srcMemoryType;
const void *srcHost;
CUdeviceptr srcDevice;
CUarray srcArray;
unsigned int srcPitch;
unsigned int dstXInBytes, dstY;
CUmemorytype dstMemoryType;
void *dstHost;
CUdeviceptr dstDevice;
CUarray dstArray;
unsigned int dstPitch;
unsigned int WidthInBytes;
unsigned int Height;
} CUDA_MEMCPY2D;

Generated for NVIDIA CUDA Library by Doxygen

142 Module Documentation

where:

• srcMemoryType and dstMemoryType specify the type of memory of the source and destination, respectively;
CUmemorytype_enum is defined as:

typedef enum CUmemorytype_enum {

CU_MEMORYTYPE_HOST = 0x01,
CU_MEMORYTYPE_DEVICE = 0x02,
CU_MEMORYTYPE_ARRAY = 0x03
} CUmemorytype;

If srcMemoryType is CU_MEMORYTYPE_HOST, srcHost and srcPitch specify the (host) base address of the
source data and the bytes per row to apply. srcArray is ignored.

If srcMemoryType is CU_MEMORYTYPE_DEVICE, srcDevice and srcPitch specify the (device) base address
of the source data and the bytes per row to apply. srcArray is ignored.

If srcMemoryType is CU_MEMORYTYPE_ARRAY, srcArray specifies the handle of the source data. srcHost,
srcDevice and srcPitch are ignored.

If dstMemoryType is CU_MEMORYTYPE_HOST, dstHost and dstPitch specify the (host) base address of the
destination data and the bytes per row to apply. dstArray is ignored.

If dstMemoryType is CU_MEMORYTYPE_DEVICE, dstDevice and dstPitch specify the (device) base address
of the destination data and the bytes per row to apply. dstArray is ignored.

If dstMemoryType is CU_MEMORYTYPE_ARRAY, dstArray specifies the handle of the destination data.

dstHost, dstDevice and dstPitch are ignored.

• srcXInBytes and srcY specify the base address of the source data for the copy.

For host pointers, the starting address is

void* Start = (void*)((char*)srcHost+srcY*srcPitch + srcXInBytes);

For device pointers, the starting address is

CUdeviceptr Start = srcDevice+srcY*srcPitch+srcXInBytes;

For CUDA arrays, srcXInBytes must be evenly divisible by the array element size.

• dstXInBytes and dstY specify the base address of the destination data for the copy.

For host pointers, the base address is

void* dstStart = (void*)((char*)dstHost+dstY*dstPitch + dstXInBytes);

Generated for NVIDIA CUDA Library by Doxygen

3.25 Memory Management 143

For device pointers, the starting address is

CUdeviceptr dstStart = dstDevice+dstY*dstPitch+dstXInBytes;

For CUDA arrays, dstXInBytes must be evenly divisible by the array element size.

• WidthInBytes and Height specify the width (in bytes) and height of the 2D copy being performed. Any pitches
must be greater than or equal to WidthInBytes.

cuMemcpy2D() returns an error if any pitch is greater than the maximum allowed (CU_DEVICE_ATTRIBUTE_-
MAX_PITCH). cuMemAllocPitch() passes back pitches that always work with cuMemcpy2D(). On intra-device
memory copies (device ? device, CUDA array ? device, CUDA array ? CUDA array), cuMemcpy2D() may fail
for pitches not computed by cuMemAllocPitch(). cuMemcpy2DUnaligned() does not have this restriction, but
may run significantly slower in the cases where cuMemcpy2D() would have returned an error code.

cuMemcpy2DAsync() is asynchronous and can optionally be associated to a stream by passing a non-zero hStream
argument. It only works on page-locked host memory and returns an error if a pointer to pageable memory is passed
as input.

Parameters:
pCopy - Parameters for the memory copy
hStream - Stream identifier

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuArray3DCreate, cuArray3DGetDescriptor, cuArrayCreate, cuArrayDestroy, cuArrayGetDescriptor, cuMemAl-
loc, cuMemAllocHost, cuMemAllocPitch, cuMemcpy2D, cuMemcpy2DUnaligned, cuMemcpy3D, cuMem-
cpy3DAsync, cuMemcpyAtoA, cuMemcpyAtoD, cuMemcpyAtoH, cuMemcpyAtoHAsync, cuMemcpyD-
toA, cuMemcpyDtoD, cuMemcpyDtoH, cuMemcpyDtoHAsync, cuMemcpyHtoA, cuMemcpyHtoAAsync,
cuMemcpyHtoD, cuMemcpyHtoDAsync, cuMemFree, cuMemFreeHost, cuMemGetAddressRange, cuMemGet-
Info, cuMemHostAlloc, cuMemHostGetDevicePointer, cuMemsetD2D8, cuMemsetD2D16, cuMemsetD2D32,
cuMemsetD8, cuMemsetD16, cuMemsetD32

3.25.2.11 cuMemcpy2DUnaligned (const CUDA_MEMCPY2D ∗ pCopy)

Perform a 2D memory copy according to the parameters specified in pCopy. The CUDA_MEMCPY2D structure is
defined as:

typedef struct CUDA_MEMCPY2D_st {

unsigned int srcXInBytes, srcY;
CUmemorytype srcMemoryType;
const void *srcHost;

Generated for NVIDIA CUDA Library by Doxygen

144 Module Documentation

CUdeviceptr srcDevice;
CUarray srcArray;
unsigned int srcPitch;
unsigned int dstXInBytes, dstY;
CUmemorytype dstMemoryType;
void *dstHost;
CUdeviceptr dstDevice;
CUarray dstArray;
unsigned int dstPitch;
unsigned int WidthInBytes;
unsigned int Height;
} CUDA_MEMCPY2D;

where:

• srcMemoryType and dstMemoryType specify the type of memory of the source and destination, respectively;
CUmemorytype_enum is defined as:

typedef enum CUmemorytype_enum {

CU_MEMORYTYPE_HOST = 0x01,
CU_MEMORYTYPE_DEVICE = 0x02,
CU_MEMORYTYPE_ARRAY = 0x03
} CUmemorytype;

If srcMemoryType is CU_MEMORYTYPE_HOST, srcHost and srcPitch specify the (host) base address of the
source data and the bytes per row to apply. srcArray is ignored.

If srcMemoryType is CU_MEMORYTYPE_DEVICE, srcDevice and srcPitch specify the (device) base address
of the source data and the bytes per row to apply. srcArray is ignored.

If srcMemoryType is CU_MEMORYTYPE_ARRAY, srcArray specifies the handle of the source data. srcHost,
srcDevice and srcPitch are ignored.

If dstMemoryType is CU_MEMORYTYPE_HOST, dstHost and dstPitch specify the (host) base address of the
destination data and the bytes per row to apply. dstArray is ignored.

If dstMemoryType is CU_MEMORYTYPE_DEVICE, dstDevice and dstPitch specify the (device) base address
of the destination data and the bytes per row to apply. dstArray is ignored.

If dstMemoryType is CU_MEMORYTYPE_ARRAY, dstArray specifies the handle of the destination data.

dstHost, dstDevice and dstPitch are ignored.

• srcXInBytes and srcY specify the base address of the source data for the copy.

For host pointers, the starting address is

void* Start = (void*)((char*)srcHost+srcY*srcPitch + srcXInBytes);

For device pointers, the starting address is

Generated for NVIDIA CUDA Library by Doxygen

3.25 Memory Management 145

CUdeviceptr Start = srcDevice+srcY*srcPitch+srcXInBytes;

For CUDA arrays, srcXInBytes must be evenly divisible by the array element size.

• dstXInBytes and dstY specify the base address of the destination data for the copy.

For host pointers, the base address is

void* dstStart = (void*)((char*)dstHost+dstY*dstPitch + dstXInBytes);

For device pointers, the starting address is

CUdeviceptr dstStart = dstDevice+dstY*dstPitch+dstXInBytes;

For CUDA arrays, dstXInBytes must be evenly divisible by the array element size.

• WidthInBytes and Height specify the width (in bytes) and height of the 2D copy being performed. Any pitches
must be greater than or equal to WidthInBytes.

cuMemcpy2D() returns an error if any pitch is greater than the maximum allowed (CU_DEVICE_ATTRIBUTE_-
MAX_PITCH). cuMemAllocPitch() passes back pitches that always work with cuMemcpy2D(). On intra-device
memory copies (device ? device, CUDA array ? device, CUDA array ? CUDA array), cuMemcpy2D() may fail
for pitches not computed by cuMemAllocPitch(). cuMemcpy2DUnaligned() does not have this restriction, but
may run significantly slower in the cases where cuMemcpy2D() would have returned an error code.

Parameters:
pCopy - Parameters for the memory copy

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuArray3DCreate, cuArray3DGetDescriptor, cuArrayCreate, cuArrayDestroy, cuArrayGetDescriptor, cuMemAl-
loc, cuMemAllocHost, cuMemAllocPitch, cuMemcpy2D, cuMemcpy2DAsync, cuMemcpy3D, cuMem-
cpy3DAsync, cuMemcpyAtoA, cuMemcpyAtoD, cuMemcpyAtoH, cuMemcpyAtoHAsync, cuMemcpyD-
toA, cuMemcpyDtoD, cuMemcpyDtoH, cuMemcpyDtoHAsync, cuMemcpyHtoA, cuMemcpyHtoAAsync,
cuMemcpyHtoD, cuMemcpyHtoDAsync, cuMemFree, cuMemFreeHost, cuMemGetAddressRange, cuMemGet-
Info, cuMemHostAlloc, cuMemHostGetDevicePointer, cuMemsetD2D8, cuMemsetD2D16, cuMemsetD2D32,
cuMemsetD8, cuMemsetD16, cuMemsetD32

Generated for NVIDIA CUDA Library by Doxygen

146 Module Documentation

3.25.2.12 cuMemcpy3D (const CUDA_MEMCPY3D ∗ pCopy)

Perform a 3D memory copy according to the parameters specified in pCopy. The CUDA_MEMCPY3D structure is
defined as:

typedef struct CUDA_MEMCPY3D_st {

unsigned int srcXInBytes, srcY, srcZ;

unsigned int srcLOD;
CUmemorytype srcMemoryType;
const void *srcHost;
CUdeviceptr srcDevice;
CUarray srcArray;
unsigned int srcPitch; // ignored when src is array
unsigned int srcHeight; // ignored when src is array; may be 0 if Depth==1

unsigned int dstXInBytes, dstY, dstZ;

unsigned int dstLOD;
CUmemorytype dstMemoryType;
void *dstHost;
CUdeviceptr dstDevice;
CUarray dstArray;
unsigned int dstPitch; // ignored when dst is array
unsigned int dstHeight; // ignored when dst is array; may be 0 if Depth==1

unsigned int WidthInBytes;

unsigned int Height;
unsigned int Depth;
} CUDA_MEMCPY3D;

where:

• srcMemoryType and dstMemoryType specify the type of memory of the source and destination, respectively;
CUmemorytype_enum is defined as:

typedef enum CUmemorytype_enum {

CU_MEMORYTYPE_HOST = 0x01,
CU_MEMORYTYPE_DEVICE = 0x02,
CU_MEMORYTYPE_ARRAY = 0x03
} CUmemorytype;

If srcMemoryType is CU_MEMORYTYPE_HOST, srcHost, srcPitch and srcHeight specify the (host) base ad-
dress of the source data, the bytes per row, and the height of each 2D slice of the 3D array. srcArray is ignored.

If srcMemoryType is CU_MEMORYTYPE_DEVICE, srcDevice, srcPitch and srcHeight specify the (device)

base address of the source data, the bytes per row, and the height of each 2D slice of the 3D array. srcArray is
ignored.

If srcMemoryType is CU_MEMORYTYPE_ARRAY, srcArray specifies the handle of the source data. srcHost,
srcDevice, srcPitch and srcHeight are ignored.

If dstMemoryType is CU_MEMORYTYPE_HOST, dstHost and dstPitch specify the (host) base address of the
destination data, the bytes per row, and the height of each 2D slice of the 3D array. dstArray is ignored.

If dstMemoryType is CU_MEMORYTYPE_DEVICE, dstDevice and dstPitch specify the (device) base address
of the destination data, the bytes per row, and the height of each 2D slice of the 3D array. dstArray is ignored.

Generated for NVIDIA CUDA Library by Doxygen

3.25 Memory Management 147

If dstMemoryType is CU_MEMORYTYPE_ARRAY, dstArray specifies the handle of the destination data.

dstHost, dstDevice, dstPitch and dstHeight are ignored.

• srcXInBytes, srcY and srcZ specify the base address of the source data for the copy.

For host pointers, the starting address is

void* Start = (void*)((char*)srcHost+(srcZ*srcHeight+srcY)*srcPitch + srcXInBytes);

For device pointers, the starting address is

CUdeviceptr Start = srcDevice+(srcZ*srcHeight+srcY)*srcPitch+srcXInBytes;

For CUDA arrays, srcXInBytes must be evenly divisible by the array element size.

• dstXInBytes, dstY and dstZ specify the base address of the destination data for the copy.

For host pointers, the base address is

void* dstStart = (void*)((char*)dstHost+(dstZ*dstHeight+dstY)*dstPitch + dstXInBytes);

For device pointers, the starting address is

CUdeviceptr dstStart = dstDevice+(dstZ*dstHeight+dstY)*dstPitch+dstXInBytes;

For CUDA arrays, dstXInBytes must be evenly divisible by the array element size.

• WidthInBytes, Height and Depth specify the width (in bytes), height and depth of the 3D copy being performed.
Any pitches must be greater than or equal to WidthInBytes.

cuMemcpy3D() returns an error if any pitch is greater than the maximum allowed (CU_DEVICE_ATTRIBUTE_-
MAX_PITCH).

The srcLOD and dstLOD members of the CUDA_MEMCPY3D structure must be set to 0.

Parameters:
pCopy - Parameters for the memory copy

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

Generated for NVIDIA CUDA Library by Doxygen

148 Module Documentation

See also:
cuArray3DCreate, cuArray3DGetDescriptor, cuArrayCreate, cuArrayDestroy, cuArrayGetDescriptor, cuMemAl-
loc, cuMemAllocHost, cuMemAllocPitch, cuMemcpy2D, cuMemcpy2DAsync, cuMemcpy2DUnaligned,
cuMemcpy3DAsync, cuMemcpyAtoA, cuMemcpyAtoD, cuMemcpyAtoH, cuMemcpyAtoHAsync, cuMemcpy-
DtoA, cuMemcpyDtoD, cuMemcpyDtoH, cuMemcpyDtoHAsync, cuMemcpyHtoA, cuMemcpyHtoAAsync,
cuMemcpyHtoD, cuMemcpyHtoDAsync, cuMemFree, cuMemFreeHost, cuMemGetAddressRange, cuMemGet-
Info, cuMemHostAlloc, cuMemHostGetDevicePointer, cuMemsetD2D8, cuMemsetD2D16, cuMemsetD2D32,
cuMemsetD8, cuMemsetD16, cuMemsetD32

3.25.2.13 cuMemcpy3DAsync (const CUDA_MEMCPY3D ∗ pCopy, CUstream hStream)

Perform a 3D memory copy according to the parameters specified in pCopy. The CUDA_MEMCPY3D structure is
defined as:

typedef struct CUDA_MEMCPY3D_st {

unsigned int srcXInBytes, srcY, srcZ;

unsigned int srcLOD;
CUmemorytype srcMemoryType;
const void *srcHost;
CUdeviceptr srcDevice;
CUarray srcArray;
unsigned int srcPitch; // ignored when src is array
unsigned int srcHeight; // ignored when src is array; may be 0 if Depth==1

unsigned int dstXInBytes, dstY, dstZ;

unsigned int dstLOD;
CUmemorytype dstMemoryType;
void *dstHost;
CUdeviceptr dstDevice;
CUarray dstArray;
unsigned int dstPitch; // ignored when dst is array
unsigned int dstHeight; // ignored when dst is array; may be 0 if Depth==1

unsigned int WidthInBytes;

unsigned int Height;
unsigned int Depth;
} CUDA_MEMCPY3D;

where:

• srcMemoryType and dstMemoryType specify the type of memory of the source and destination, respectively;
CUmemorytype_enum is defined as:

typedef enum CUmemorytype_enum {

CU_MEMORYTYPE_HOST = 0x01,
CU_MEMORYTYPE_DEVICE = 0x02,
CU_MEMORYTYPE_ARRAY = 0x03
} CUmemorytype;

If srcMemoryType is CU_MEMORYTYPE_HOST, srcHost, srcPitch and srcHeight specify the (host) base ad-
dress of the source data, the bytes per row, and the height of each 2D slice of the 3D array. srcArray is ignored.

If srcMemoryType is CU_MEMORYTYPE_DEVICE, srcDevice, srcPitch and srcHeight specify the (device)

base address of the source data, the bytes per row, and the height of each 2D slice of the 3D array. srcArray is
ignored.

Generated for NVIDIA CUDA Library by Doxygen

3.25 Memory Management 149

If srcMemoryType is CU_MEMORYTYPE_ARRAY, srcArray specifies the handle of the source data. srcHost,
srcDevice, srcPitch and srcHeight are ignored.

If dstMemoryType is CU_MEMORYTYPE_HOST, dstHost and dstPitch specify the (host) base address of the
destination data, the bytes per row, and the height of each 2D slice of the 3D array. dstArray is ignored.

If dstMemoryType is CU_MEMORYTYPE_DEVICE, dstDevice and dstPitch specify the (device) base address
of the destination data, the bytes per row, and the height of each 2D slice of the 3D array. dstArray is ignored.

If dstMemoryType is CU_MEMORYTYPE_ARRAY, dstArray specifies the handle of the destination data.

dstHost, dstDevice, dstPitch and dstHeight are ignored.

• srcXInBytes, srcY and srcZ specify the base address of the source data for the copy.

For host pointers, the starting address is

void* Start = (void*)((char*)srcHost+(srcZ*srcHeight+srcY)*srcPitch + srcXInBytes);

For device pointers, the starting address is

CUdeviceptr Start = srcDevice+(srcZ*srcHeight+srcY)*srcPitch+srcXInBytes;

For CUDA arrays, srcXInBytes must be evenly divisible by the array element size.

• dstXInBytes, dstY and dstZ specify the base address of the destination data for the copy.

For host pointers, the base address is

void* dstStart = (void*)((char*)dstHost+(dstZ*dstHeight+dstY)*dstPitch + dstXInBytes);

For device pointers, the starting address is

CUdeviceptr dstStart = dstDevice+(dstZ*dstHeight+dstY)*dstPitch+dstXInBytes;

For CUDA arrays, dstXInBytes must be evenly divisible by the array element size.

• WidthInBytes, Height and Depth specify the width (in bytes), height and depth of the 3D copy being performed.
Any pitches must be greater than or equal to WidthInBytes.

cuMemcpy3D() returns an error if any pitch is greater than the maximum allowed (CU_DEVICE_ATTRIBUTE_-
MAX_PITCH).

cuMemcpy3DAsync() is asynchronous and can optionally be associated to a stream by passing a non-zero hStream
argument. It only works on page-locked host memory and returns an error if a pointer to pageable memory is passed
as input.
The srcLOD and dstLOD members of the CUDA_MEMCPY3D structure must be set to 0.

Generated for NVIDIA CUDA Library by Doxygen

150 Module Documentation

Parameters:
pCopy - Parameters for the memory copy
hStream - Stream identifier

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuArray3DCreate, cuArray3DGetDescriptor, cuArrayCreate, cuArrayDestroy, cuArrayGetDescriptor, cuMemAl-
loc, cuMemAllocHost, cuMemAllocPitch, cuMemcpy2D, cuMemcpy2DAsync, cuMemcpy2DUnaligned,
cuMemcpy3D, cuMemcpyAtoA, cuMemcpyAtoD, cuMemcpyAtoH, cuMemcpyAtoHAsync, cuMemcpyD-
toA, cuMemcpyDtoD, cuMemcpyDtoH, cuMemcpyDtoHAsync, cuMemcpyHtoA, cuMemcpyHtoAAsync,
cuMemcpyHtoD, cuMemcpyHtoDAsync, cuMemFree, cuMemFreeHost, cuMemGetAddressRange, cuMemGet-
Info, cuMemHostAlloc, cuMemHostGetDevicePointer, cuMemsetD2D8, cuMemsetD2D16, cuMemsetD2D32,
cuMemsetD8, cuMemsetD16, cuMemsetD32

3.25.2.14 cuMemcpyAtoA (CUarray dstArray, unsigned int dstIndex, CUarray srcArray, unsigned int
srcIndex, unsigned int ByteCount)

Copies from one 1D CUDA array to another. dstArray and srcArray specify the handles of the destination
and source CUDA arrays for the copy, respectively. dstIndex and srcIndex specify the destination and source
indices into the CUDA array. These values are in the range [0, Width-1] for the CUDA array; they are not byte offsets.
ByteCount is the number of bytes to be copied. The size of the elements in the CUDA arrays need not be the same
format, but the elements must be the same size; and count must be evenly divisible by that size.

Parameters:
dstArray - Destination array
dstIndex - Offset of destination array
srcArray - Source array
srcIndex - Offset of source array
ByteCount - Size of memory copy in bytes

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuArray3DCreate, cuArray3DGetDescriptor, cuArrayCreate, cuArrayDestroy, cuArrayGetDescriptor, cuMemAl-
loc, cuMemAllocHost, cuMemAllocPitch, cuMemcpy2D, cuMemcpy2DAsync, cuMemcpy2DUnaligned,

Generated for NVIDIA CUDA Library by Doxygen

3.25 Memory Management 151

cuMemcpy3D, cuMemcpy3DAsync, cuMemcpyAtoD, cuMemcpyAtoH, cuMemcpyAtoHAsync, cuMemcpy-

DtoA, cuMemcpyDtoD, cuMemcpyDtoH, cuMemcpyDtoHAsync, cuMemcpyHtoA, cuMemcpyHtoAAsync,
cuMemcpyHtoD, cuMemcpyHtoDAsync, cuMemFree, cuMemFreeHost, cuMemGetAddressRange, cuMemGet-
Info, cuMemHostAlloc, cuMemHostGetDevicePointer, cuMemsetD2D8, cuMemsetD2D16, cuMemsetD2D32,
cuMemsetD8, cuMemsetD16, cuMemsetD32

3.25.2.15 cuMemcpyAtoD (CUdeviceptr dstDevice, CUarray hSrc, unsigned int SrcIndex, unsigned int
ByteCount)

Copies from one 1D CUDA array to device memory. dstDevice specifies the base pointer of the destination and
must be naturally aligned with the CUDA array elements. hSrc and SrcIndex specify the CUDA array handle and
the index (in array elements) of the array element where the copy is to begin. ByteCount specifies the number of
bytes to copy and must be evenly divisible by the array element size.

Parameters:
dstDevice - Destination device pointer
hSrc - Source array
SrcIndex - Offset of source array
ByteCount - Size of memory copy in bytes

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuArray3DCreate, cuArray3DGetDescriptor, cuArrayCreate, cuArrayDestroy, cuArrayGetDescriptor, cuMemAl-
loc, cuMemAllocHost, cuMemAllocPitch, cuMemcpy2D, cuMemcpy2DAsync, cuMemcpy2DUnaligned,
cuMemcpy3D, cuMemcpy3DAsync, cuMemcpyAtoA, cuMemcpyAtoH, cuMemcpyAtoHAsync, cuMemcpy-
DtoA, cuMemcpyDtoD, cuMemcpyDtoH, cuMemcpyDtoHAsync, cuMemcpyHtoA, cuMemcpyHtoAAsync,
cuMemcpyHtoD, cuMemcpyHtoDAsync, cuMemFree, cuMemFreeHost, cuMemGetAddressRange, cuMemGet-
Info, cuMemHostAlloc, cuMemHostGetDevicePointer, cuMemsetD2D8, cuMemsetD2D16, cuMemsetD2D32,
cuMemsetD8, cuMemsetD16, cuMemsetD32

3.25.2.16 cuMemcpyAtoH (void ∗ dstHost, CUarray srcArray, unsigned int srcIndex, unsigned int
ByteCount)

Copies from one 1D CUDA array to host memory. dstHost specifies the base pointer of the destination. srcArray
and srcIndex specify the CUDA array handle and starting index of the source data. ByteCount specifies the
number of bytes to copy.

Parameters:
dstHost - Destination device pointer
srcArray - Source array
srcIndex - Offset of source array

Generated for NVIDIA CUDA Library by Doxygen

152 Module Documentation

ByteCount - Size of memory copy in bytes

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuArray3DCreate, cuArray3DGetDescriptor, cuArrayCreate, cuArrayDestroy, cuArrayGetDescriptor, cuMemAl-
loc, cuMemAllocHost, cuMemAllocPitch, cuMemcpy2D, cuMemcpy2DAsync, cuMemcpy2DUnaligned,
cuMemcpy3D, cuMemcpy3DAsync, cuMemcpyAtoA, cuMemcpyAtoD, cuMemcpyAtoHAsync, cuMemcpy-
DtoA, cuMemcpyDtoD, cuMemcpyDtoH, cuMemcpyDtoHAsync, cuMemcpyHtoA, cuMemcpyHtoAAsync,
cuMemcpyHtoD, cuMemcpyHtoDAsync, cuMemFree, cuMemFreeHost, cuMemGetAddressRange, cuMemGet-
Info, cuMemHostAlloc, cuMemHostGetDevicePointer, cuMemsetD2D8, cuMemsetD2D16, cuMemsetD2D32,
cuMemsetD8, cuMemsetD16, cuMemsetD32

3.25.2.17 cuMemcpyAtoHAsync (void ∗ dstHost, CUarray srcArray, unsigned int srcIndex, unsigned int
ByteCount, CUstream hStream)

Copies from one 1D CUDA array to host memory. dstHost specifies the base pointer of the destination. srcArray
and srcIndex specify the CUDA array handle and starting index of the source data. ByteCount specifies the
number of bytes to copy.
cuMemcpyAtoHAsync() is asynchronous and can optionally be associated to a stream by passing a non-zero stream
argument. It only works on page-locked host memory and returns an error if a pointer to pageable memory is passed
as input.

Parameters:
dstHost - Destination pointer
srcArray - Source array
srcIndex - Offset of source array
ByteCount - Size of memory copy in bytes
hStream - Stream identifier
Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuArray3DCreate, cuArray3DGetDescriptor, cuArrayCreate, cuArrayDestroy, cuArrayGetDescriptor, cuMemAl-
loc, cuMemAllocHost, cuMemAllocPitch, cuMemcpy2D, cuMemcpy2DAsync, cuMemcpy2DUnaligned,
cuMemcpy3D, cuMemcpy3DAsync, cuMemcpyAtoA, cuMemcpyAtoD, cuMemcpyAtoH, cuMemcpyDtoA,
cuMemcpyDtoD, cuMemcpyDtoH, cuMemcpyDtoHAsync, cuMemcpyHtoA, cuMemcpyHtoAAsync, cuMem-
cpyHtoD, cuMemcpyHtoDAsync, cuMemFree, cuMemFreeHost, cuMemGetAddressRange, cuMemGet-
Info, cuMemHostAlloc, cuMemHostGetDevicePointer, cuMemsetD2D8, cuMemsetD2D16, cuMemsetD2D32,
cuMemsetD8, cuMemsetD16, cuMemsetD32

Generated for NVIDIA CUDA Library by Doxygen

3.25 Memory Management 153

3.25.2.18 cuMemcpyDtoA (CUarray dstArray, unsigned int dstIndex, CUdeviceptr srcDevice, unsigned int
ByteCount)

Copies from device memory to a 1D CUDA array. dstArray and dstIndex specify the CUDA array handle and
starting index of the destination data. srcDevice specifies the base pointer of the source. ByteCount specifies
the number of bytes to copy.

Parameters:
dstArray - Destination array
dstIndex - Offset of destination array
srcDevice - Source device pointer
ByteCount - Size of memory copy in bytes

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuArray3DCreate, cuArray3DGetDescriptor, cuArrayCreate, cuArrayDestroy, cuArrayGetDescriptor, cuMemAl-
loc, cuMemAllocHost, cuMemAllocPitch, cuMemcpy2D, cuMemcpy2DAsync, cuMemcpy2DUnaligned,
cuMemcpy3D, cuMemcpy3DAsync, cuMemcpyAtoA, cuMemcpyAtoD, cuMemcpyAtoH, cuMemcpyAto-
HAsync, cuMemcpyDtoD, cuMemcpyDtoH, cuMemcpyDtoHAsync, cuMemcpyHtoA, cuMemcpyHtoAAsync,
cuMemcpyHtoD, cuMemcpyHtoDAsync, cuMemFree, cuMemFreeHost, cuMemGetAddressRange, cuMemGet-
Info, cuMemHostAlloc, cuMemHostGetDevicePointer, cuMemsetD2D8, cuMemsetD2D16, cuMemsetD2D32,
cuMemsetD8, cuMemsetD16, cuMemsetD32

3.25.2.19 cuMemcpyDtoD (CUdeviceptr dstDevice, CUdeviceptr srcDevice, unsigned int ByteCount)

Copies from device memory to device memory. dstDevice and srcDevice are the base pointers of the destination
and source, respectively. ByteCount specifies the number of bytes to copy. Note that this function is asynchronous.

Parameters:
dstDevice - Destination device pointer
srcDevice - Source device pointer
ByteCount - Size of memory copy in bytes

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

Generated for NVIDIA CUDA Library by Doxygen

154 Module Documentation

See also:
cuArray3DCreate, cuArray3DGetDescriptor, cuArrayCreate, cuArrayDestroy, cuArrayGetDescriptor, cuMemAl-
loc, cuMemAllocHost, cuMemAllocPitch, cuMemcpy2D, cuMemcpy2DAsync, cuMemcpy2DUnaligned,
cuMemcpy3D, cuMemcpy3DAsync, cuMemcpyAtoA, cuMemcpyAtoD, cuMemcpyAtoH, cuMemcpyAto-
HAsync, cuMemcpyDtoA, cuMemcpyDtoH, cuMemcpyDtoHAsync, cuMemcpyHtoA, cuMemcpyHtoAAsync,
cuMemcpyHtoD, cuMemcpyHtoDAsync, cuMemFree, cuMemFreeHost, cuMemGetAddressRange, cuMemGet-
Info, cuMemHostAlloc, cuMemHostGetDevicePointer, cuMemsetD2D8, cuMemsetD2D16, cuMemsetD2D32,
cuMemsetD8, cuMemsetD16, cuMemsetD32

3.25.2.20 cuMemcpyDtoH (void ∗ dstHost, CUdeviceptr srcDevice, unsigned int ByteCount)

Copies from device to host memory. dstHost and srcDevice specify the base pointers of the destination and
source, respectively. ByteCount specifies the number of bytes to copy. Note that this function is synchronous.

Parameters:
dstHost - Destination host pointer
srcDevice - Source device pointer
ByteCount - Size of memory copy in bytes

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuArray3DCreate, cuArray3DGetDescriptor, cuArrayCreate, cuArrayDestroy, cuArrayGetDescriptor, cuMemAl-
loc, cuMemAllocHost, cuMemAllocPitch, cuMemcpy2D, cuMemcpy2DAsync, cuMemcpy2DUnaligned,
cuMemcpy3D, cuMemcpy3DAsync, cuMemcpyAtoA, cuMemcpyAtoD, cuMemcpyAtoH, cuMemcpyAto-
HAsync, cuMemcpyDtoA, cuMemcpyDtoD, cuMemcpyDtoHAsync, cuMemcpyHtoA, cuMemcpyHtoAAsync,
cuMemcpyHtoD, cuMemcpyHtoDAsync, cuMemFree, cuMemFreeHost, cuMemGetAddressRange, cuMemGet-
Info, cuMemHostAlloc, cuMemHostGetDevicePointer, cuMemsetD2D8, cuMemsetD2D16, cuMemsetD2D32,
cuMemsetD8, cuMemsetD16, cuMemsetD32

3.25.2.21 cuMemcpyDtoHAsync (void ∗ dstHost, CUdeviceptr srcDevice, unsigned int ByteCount, CUstream
hStream)

Copies from device to host memory. dstHost and srcDevice specify the base pointers of the destination and
source, respectively. ByteCount specifies the number of bytes to copy.
cuMemcpyDtoHAsync() is asynchronous and can optionally be associated to a stream by passing a non-zero
hStream argument. It only works on page-locked memory and returns an error if a pointer to pageable memory
is passed as input.

Parameters:
dstHost - Destination host pointer
srcDevice - Source device pointer

Generated for NVIDIA CUDA Library by Doxygen

3.25 Memory Management 155

ByteCount - Size of memory copy in bytes

hStream - Stream identifier

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuArray3DCreate, cuArray3DGetDescriptor, cuArrayCreate, cuArrayDestroy, cuArrayGetDescriptor, cuMemAl-
loc, cuMemAllocHost, cuMemAllocPitch, cuMemcpy2D, cuMemcpy2DAsync, cuMemcpy2DUnaligned,
cuMemcpy3D, cuMemcpy3DAsync, cuMemcpyAtoA, cuMemcpyAtoD, cuMemcpyAtoH, cuMemcpyAto-
HAsync, cuMemcpyDtoA, cuMemcpyDtoD, cuMemcpyDtoH, cuMemcpyHtoA, cuMemcpyHtoAAsync,
cuMemcpyHtoD, cuMemcpyHtoDAsync, cuMemFree, cuMemFreeHost, cuMemGetAddressRange, cuMemGet-
Info, cuMemHostAlloc, cuMemHostGetDevicePointer, cuMemsetD2D8, cuMemsetD2D16, cuMemsetD2D32,
cuMemsetD8, cuMemsetD16, cuMemsetD32

3.25.2.22 cuMemcpyHtoA (CUarray dstArray, unsigned int dstIndex, const void ∗ pSrc, unsigned int
ByteCount)

Copies from host memory to a 1D CUDA array. dstArray and dstIndex specify the CUDA array handle and
starting index of the destination data. pSrc specifies the base address of the source. ByteCount specifies the
number of bytes to copy.

Parameters:
dstArray - Destination array
dstIndex - Offset of destination array
pSrc - Source host pointer
ByteCount - Size of memory copy in bytes

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuArray3DCreate, cuArray3DGetDescriptor, cuArrayCreate, cuArrayDestroy, cuArrayGetDescriptor, cuMemAl-
loc, cuMemAllocHost, cuMemAllocPitch, cuMemcpy2D, cuMemcpy2DAsync, cuMemcpy2DUnaligned,
cuMemcpy3D, cuMemcpy3DAsync, cuMemcpyAtoA, cuMemcpyAtoD, cuMemcpyAtoH, cuMemcpyAto-
HAsync, cuMemcpyDtoA, cuMemcpyDtoD, cuMemcpyDtoH, cuMemcpyDtoHAsync, cuMemcpyHtoAAsync,
cuMemcpyHtoD, cuMemcpyHtoDAsync, cuMemFree, cuMemFreeHost, cuMemGetAddressRange, cuMemGet-
Info, cuMemHostAlloc, cuMemHostGetDevicePointer, cuMemsetD2D8, cuMemsetD2D16, cuMemsetD2D32,
cuMemsetD8, cuMemsetD16, cuMemsetD32

Generated for NVIDIA CUDA Library by Doxygen

156 Module Documentation

3.25.2.23 cuMemcpyHtoAAsync (CUarray dstArray, unsigned int dstIndex, const void ∗ pSrc, unsigned int
ByteCount, CUstream hStream)

Copies from host memory to a 1D CUDA array. dstArray and dstIndex specify the CUDA array handle and
starting index of the destination data. pSrc specifies the base address of the source. ByteCount specifies the
number of bytes to copy.
cuMemcpyHtoAAsync() is asynchronous and can optionally be associated to a stream by passing a non-zero
hStream argument. It only works on page-locked memory and returns an error if a pointer to pageable memory
is passed as input.

Parameters:
dstArray - Destination array
dstIndex - Offset of destination array
pSrc - Source host pointer
ByteCount - Size of memory copy in bytes
hStream - Stream identifier

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuArray3DCreate, cuArray3DGetDescriptor, cuArrayCreate, cuArrayDestroy, cuArrayGetDescriptor, cuMemAl-
loc, cuMemAllocHost, cuMemAllocPitch, cuMemcpy2D, cuMemcpy2DAsync, cuMemcpy2DUnaligned,
cuMemcpy3D, cuMemcpy3DAsync, cuMemcpyAtoA, cuMemcpyAtoD, cuMemcpyAtoH, cuMemcpyAto-
HAsync, cuMemcpyDtoA, cuMemcpyDtoD, cuMemcpyDtoH, cuMemcpyDtoHAsync, cuMemcpyHtoA,
cuMemcpyHtoD, cuMemcpyHtoDAsync, cuMemFree, cuMemFreeHost, cuMemGetAddressRange, cuMemGet-
Info, cuMemHostAlloc, cuMemHostGetDevicePointer, cuMemsetD2D8, cuMemsetD2D16, cuMemsetD2D32,
cuMemsetD8, cuMemsetD16, cuMemsetD32

3.25.2.24 cuMemcpyHtoD (CUdeviceptr dstDevice, const void ∗ srcHost, unsigned int ByteCount)

Copies from host memory to device memory. dstDevice and srcHost are the base addresses of the destination
and source, respectively. ByteCount specifies the number of bytes to copy. Note that this function is synchronous.

Parameters:
dstDevice - Destination device pointer
srcHost - Source host pointer
ByteCount - Size of memory copy in bytes

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Generated for NVIDIA CUDA Library by Doxygen

3.25 Memory Management 157

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuArray3DCreate, cuArray3DGetDescriptor, cuArrayCreate, cuArrayDestroy, cuArrayGetDescriptor, cuMemAl-
loc, cuMemAllocHost, cuMemAllocPitch, cuMemcpy2D, cuMemcpy2DAsync, cuMemcpy2DUnaligned,
cuMemcpy3D, cuMemcpy3DAsync, cuMemcpyAtoA, cuMemcpyAtoD, cuMemcpyAtoH, cuMemcpyAto-
HAsync, cuMemcpyDtoA, cuMemcpyDtoD, cuMemcpyDtoH, cuMemcpyDtoHAsync, cuMemcpyHtoA,
cuMemcpyHtoAAsync, cuMemcpyHtoDAsync, cuMemFree, cuMemFreeHost, cuMemGetAddressRange,
cuMemGetInfo, cuMemHostAlloc, cuMemHostGetDevicePointer, cuMemsetD2D8, cuMemsetD2D16, cuMem-
setD2D32, cuMemsetD8, cuMemsetD16, cuMemsetD32

3.25.2.25 cuMemcpyHtoDAsync (CUdeviceptr dstDevice, const void ∗ srcHost, unsigned int ByteCount,
CUstream hStream)

Copies from host memory to device memory. dstDevice and srcHost are the base addresses of the destination
and source, respectively. ByteCount specifies the number of bytes to copy.
cuMemcpyHtoDAsync() is asynchronous and can optionally be associated to a stream by passing a non-zero hStream
argument. It only works on page-locked memory and returns an error if a pointer to pageable memory is passed as
input.

Parameters:
dstDevice - Destination device pointer
srcHost - Source host pointer
ByteCount - Size of memory copy in bytes
hStream - Stream identifier

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuArray3DCreate, cuArray3DGetDescriptor, cuArrayCreate, cuArrayDestroy, cuArrayGetDescriptor, cuMemAl-
loc, cuMemAllocHost, cuMemAllocPitch, cuMemcpy2D, cuMemcpy2DAsync, cuMemcpy2DUnaligned,
cuMemcpy3D, cuMemcpy3DAsync, cuMemcpyAtoA, cuMemcpyAtoD, cuMemcpyAtoH, cuMemcpyAto-
HAsync, cuMemcpyDtoA, cuMemcpyDtoD, cuMemcpyDtoH, cuMemcpyDtoHAsync, cuMemcpyHtoA,
cuMemcpyHtoAAsync, cuMemcpyHtoD, cuMemFree, cuMemFreeHost, cuMemGetAddressRange, cuMemGet-
Info, cuMemHostAlloc, cuMemHostGetDevicePointer, cuMemsetD2D8, cuMemsetD2D16, cuMemsetD2D32,
cuMemsetD8, cuMemsetD16, cuMemsetD32

3.25.2.26 cuMemFree (CUdeviceptr dptr)

Frees the memory space pointed to by dptr, which must have been returned by a previous call to cuMemAlloc() or
cuMemAllocPitch().

Generated for NVIDIA CUDA Library by Doxygen

158 Module Documentation

Parameters:
dptr - Pointer to memory to free

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuArray3DCreate, cuArray3DGetDescriptor, cuArrayCreate, cuArrayDestroy, cuArrayGetDescriptor, cuMemAl-
loc, cuMemAllocHost, cuMemAllocPitch, cuMemcpy2D, cuMemcpy2DAsync, cuMemcpy2DUnaligned,
cuMemcpy3D, cuMemcpy3DAsync, cuMemcpyAtoA, cuMemcpyAtoD, cuMemcpyAtoH, cuMemcpyAto-
HAsync, cuMemcpyDtoA, cuMemcpyDtoD, cuMemcpyDtoH, cuMemcpyDtoHAsync, cuMemcpyHtoA,
cuMemcpyHtoAAsync, cuMemcpyHtoD, cuMemcpyHtoDAsync, cuMemFreeHost, cuMemGetAddressRange,
cuMemGetInfo, cuMemHostAlloc, cuMemHostGetDevicePointer, cuMemsetD2D8, cuMemsetD2D16, cuMem-
setD2D32, cuMemsetD8, cuMemsetD16, cuMemsetD32

3.25.2.27 cuMemFreeHost (void ∗ p)

Frees the memory space pointed to by p, which must have been returned by a previous call to cuMemAllocHost().

Parameters:
p - Pointer to memory to free

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuArray3DCreate, cuArray3DGetDescriptor, cuArrayCreate, cuArrayDestroy, cuArrayGetDescriptor, cuMemAl-
loc, cuMemAllocHost, cuMemAllocPitch, cuMemcpy2D, cuMemcpy2DAsync, cuMemcpy2DUnaligned,
cuMemcpy3D, cuMemcpy3DAsync, cuMemcpyAtoA, cuMemcpyAtoD, cuMemcpyAtoH, cuMemcpyAto-
HAsync, cuMemcpyDtoA, cuMemcpyDtoD, cuMemcpyDtoH, cuMemcpyDtoHAsync, cuMemcpyHtoA,
cuMemcpyHtoAAsync, cuMemcpyHtoD, cuMemcpyHtoDAsync, cuMemFree, cuMemGetAddressRange,
cuMemGetInfo, cuMemHostAlloc, cuMemHostGetDevicePointer, cuMemsetD2D8, cuMemsetD2D16, cuMem-
setD2D32, cuMemsetD8, cuMemsetD16, cuMemsetD32

3.25.2.28 cuMemGetAddressRange (CUdeviceptr ∗ pbase, unsigned int ∗ psize, CUdeviceptr dptr)

Returns the base address in ∗pbase and size in ∗psize of the allocation by cuMemAlloc() or cuMemAllocPitch()
that contains the input pointer dptr. Both parameters pbase and psize are optional. If one of them is NULL, it is
ignored.

Generated for NVIDIA CUDA Library by Doxygen

3.25 Memory Management 159

Parameters:
pbase - Returned base address
psize - Returned size of device memory allocation
dptr - Device pointer to query

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuArray3DCreate, cuArray3DGetDescriptor, cuArrayCreate, cuArrayDestroy, cuArrayGetDescriptor, cuMemAl-
loc, cuMemAllocHost, cuMemAllocPitch, cuMemcpy2D, cuMemcpy2DAsync, cuMemcpy2DUnaligned,
cuMemcpy3D, cuMemcpy3DAsync, cuMemcpyAtoA, cuMemcpyAtoD, cuMemcpyAtoH, cuMemcpyAto-
HAsync, cuMemcpyDtoA, cuMemcpyDtoD, cuMemcpyDtoH, cuMemcpyDtoHAsync, cuMemcpyHtoA,
cuMemcpyHtoAAsync, cuMemcpyHtoD, cuMemcpyHtoDAsync, cuMemFree, cuMemFreeHost, cuMemGet-
Info, cuMemHostAlloc, cuMemHostGetDevicePointer, cuMemsetD2D8, cuMemsetD2D16, cuMemsetD2D32,
cuMemsetD8, cuMemsetD16, cuMemsetD32

3.25.2.29 cuMemGetInfo (unsigned int ∗ free, unsigned int ∗ total)

Returns in ∗free and ∗total respectively, the free and total amount of memory available for allocation by the
CUDA context, in bytes.

Parameters:
free - Returned free memory in bytes
total - Returned total memory in bytes

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuArray3DCreate, cuArray3DGetDescriptor, cuArrayCreate, cuArrayDestroy, cuArrayGetDescriptor, cuMemAl-
loc, cuMemAllocHost, cuMemAllocPitch, cuMemcpy2D, cuMemcpy2DAsync, cuMemcpy2DUnaligned,
cuMemcpy3D, cuMemcpy3DAsync, cuMemcpyAtoA, cuMemcpyAtoD, cuMemcpyAtoH, cuMemcpyAto-
HAsync, cuMemcpyDtoA, cuMemcpyDtoD, cuMemcpyDtoH, cuMemcpyDtoHAsync, cuMemcpyHtoA,
cuMemcpyHtoAAsync, cuMemcpyHtoD, cuMemcpyHtoDAsync, cuMemFree, cuMemFreeHost, cuMemGe-
tAddressRange, cuMemHostAlloc, cuMemHostGetDevicePointer, cuMemsetD2D8, cuMemsetD2D16, cuMem-
setD2D32, cuMemsetD8, cuMemsetD16, cuMemsetD32

Generated for NVIDIA CUDA Library by Doxygen

160 Module Documentation

3.25.2.30 cuMemHostAlloc (void ∗∗ pp, size_t bytesize, unsigned int Flags)

Allocates bytesize bytes of host memory that is page-locked and accessible to the device. The driver tracks the
virtual memory ranges allocated with this function and automatically accelerates calls to functions such as cuMem-
cpyHtoD(). Since the memory can be accessed directly by the device, it can be read or written with much higher
bandwidth than pageable memory obtained with functions such as malloc(). Allocating excessive amounts of pinned
memory may degrade system performance, since it reduces the amount of memory available to the system for paging.
As a result, this function is best used sparingly to allocate staging areas for data exchange between host and device.
The Flags parameter enables different options to be specified that affect the allocation, as follows.

• CU_MEMHOSTALLOC_PORTABLE: The memory returned by this call will be considered as pinned memory
by all CUDA contexts, not just the one that performed the allocation.

• CU_MEMHOSTALLOC_DEVICEMAP: Maps the allocation into the CUDA address space. The device pointer
to the memory may be obtained by calling cuMemHostGetDevicePointer(). This feature is available only on
GPUs with compute capability greater than or equal to 1.1.

• CU_MEMHOSTALLOC_WRITECOMBINED: Allocates the memory as write-combined (WC). WC memory

can be transferred across the PCI Express bus more quickly on some system configurations, but cannot be read
efficiently by most CPUs. WC memory is a good option for buffers that will be written by the CPU and read by
the GPU via mapped pinned memory or host->device transfers.

All of these flags are orthogonal to one another: a developer may allocate memory that is portable, mapped and/or
write-combined with no restrictions.
The CUDA context must have been created with the CU_CTX_MAP_HOST flag in order for the CU_-
MEMHOSTALLOC_MAPPED flag to have any effect.
The CU_MEMHOSTALLOC_MAPPED flag may be specified on CUDA contexts for devices that do not support
mapped pinned memory. The failure is deferred to cuMemHostGetDevicePointer() because the memory may be
mapped into other CUDA contexts via the CU_MEMHOSTALLOC_PORTABLE flag.
The memory allocated by this function must be freed with cuMemFreeHost().

Parameters:
pp - Returned host pointer to page-locked memory
bytesize - Requested allocation size in bytes
Flags - Flags for allocation request

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_OUT_OF_MEMORY

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuArray3DCreate, cuArray3DGetDescriptor, cuArrayCreate, cuArrayDestroy, cuArrayGetDescriptor, cuMemAl-
loc, cuMemAllocHost, cuMemAllocPitch, cuMemcpy2D, cuMemcpy2DAsync, cuMemcpy2DUnaligned,
cuMemcpy3D, cuMemcpy3DAsync, cuMemcpyAtoA, cuMemcpyAtoD, cuMemcpyAtoH, cuMemcpyAto-
HAsync, cuMemcpyDtoA, cuMemcpyDtoD, cuMemcpyDtoH, cuMemcpyDtoHAsync, cuMemcpyHtoA,
cuMemcpyHtoAAsync, cuMemcpyHtoD, cuMemcpyHtoDAsync, cuMemFree, cuMemFreeHost, cuMemGe-
tAddressRange, cuMemGetInfo, cuMemHostGetDevicePointer, cuMemsetD2D8, cuMemsetD2D16, cuMem-
setD2D32, cuMemsetD8, cuMemsetD16, cuMemsetD32

Generated for NVIDIA CUDA Library by Doxygen

3.25 Memory Management 161

3.25.2.31 cuMemHostGetDevicePointer (CUdeviceptr ∗ pdptr, void ∗ p, unsigned int Flags)

Passes back the device pointer pdptr corresponding to the mapped, pinned host buffer p allocated by cuMemHostAl-
loc.
cuMemHostGetDevicePointer() will fail if the CU_MEMALLOCHOST_DEVICEMAP flag was not specified at the
time the memory was allocated, or if the function is called on a GPU that does not support mapped pinned memory.
Flags provides for future releases. For now, it must be set to 0.

Parameters:
pdptr - Returned device pointer
p - Host pointer
Flags - Options (must be 0)

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuArray3DCreate, cuArray3DGetDescriptor, cuArrayCreate, cuArrayDestroy, cuArrayGetDescriptor, cuMemAl-
loc, cuMemAllocHost, cuMemAllocPitch, cuMemcpy2D, cuMemcpy2DAsync, cuMemcpy2DUnaligned,
cuMemcpy3D, cuMemcpy3DAsync, cuMemcpyAtoA, cuMemcpyAtoD, cuMemcpyAtoH, cuMemcpyAto-
HAsync, cuMemcpyDtoA, cuMemcpyDtoD, cuMemcpyDtoH, cuMemcpyDtoHAsync, cuMemcpyHtoA,
cuMemcpyHtoAAsync, cuMemcpyHtoD, cuMemcpyHtoDAsync, cuMemFree, cuMemFreeHost, cuMemGetAd-
dressRange, cuMemGetInfo, cuMemHostAlloc, cuMemsetD2D8, cuMemsetD2D16, cuMemsetD2D32, cuMem-
setD8, cuMemsetD16, cuMemsetD32

3.25.2.32 cuMemHostGetFlags (unsigned int ∗ pFlags, void ∗ p)

Passes back the flags pFlags that were specified when allocating the pinned host buffer p allocated by
cuMemHostAlloc.
cuMemHostGetFlags() will fail if the pointer does not reside in an allocation performed by cuMemAllocHost() or
cuMemHostAlloc().

Parameters:
pFlags - Returned flags word
p - Host pointer

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuMemAllocHost, cuMemHostAlloc

Generated for NVIDIA CUDA Library by Doxygen

162 Module Documentation

3.25.2.33 cuMemsetD16 (CUdeviceptr dstDevice, unsigned short us, unsigned int N)

Sets the memory range of N 16-bit values to the specified value us.

Parameters:
dstDevice - Destination device pointer
us - Value to set
N - Number of elements

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuArray3DCreate, cuArray3DGetDescriptor, cuArrayCreate, cuArrayDestroy, cuArrayGetDescriptor, cuMemAl-
loc, cuMemAllocHost, cuMemAllocPitch, cuMemcpy2D, cuMemcpy2DAsync, cuMemcpy2DUnaligned,
cuMemcpy3D, cuMemcpy3DAsync, cuMemcpyAtoA, cuMemcpyAtoD, cuMemcpyAtoH, cuMemcpyAto-
HAsync, cuMemcpyDtoA, cuMemcpyDtoD, cuMemcpyDtoH, cuMemcpyDtoHAsync, cuMemcpyHtoA,
cuMemcpyHtoAAsync, cuMemcpyHtoD, cuMemcpyHtoDAsync, cuMemFree, cuMemFreeHost, cuMemGe-
tAddressRange, cuMemGetInfo, cuMemHostAlloc, cuMemHostGetDevicePointer, cuMemsetD2D8, cuMem-
setD2D16, cuMemsetD2D32, cuMemsetD8, cuMemsetD32

3.25.2.34 cuMemsetD2D16 (CUdeviceptr dstDevice, unsigned int dstPitch, unsigned short us, unsigned int
Width, unsigned int Height)

Sets the 2D memory range of Width 16-bit values to the specified value us. Height specifies the number of rows
to set, and dstPitch specifies the number of bytes between each row. This function performs fastest when the pitch
is one that has been passed back by cuMemAllocPitch().

Parameters:
dstDevice - Destination device pointer
dstPitch - Pitch of destination device pointer
us - Value to set
Width - Width of row
Height - Number of rows

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

Generated for NVIDIA CUDA Library by Doxygen

3.25 Memory Management 163

See also:
cuArray3DCreate, cuArray3DGetDescriptor, cuArrayCreate, cuArrayDestroy, cuArrayGetDescriptor, cuMemAl-
loc, cuMemAllocHost, cuMemAllocPitch, cuMemcpy2D, cuMemcpy2DAsync, cuMemcpy2DUnaligned,
cuMemcpy3D, cuMemcpy3DAsync, cuMemcpyAtoA, cuMemcpyAtoD, cuMemcpyAtoH, cuMemcpyAto-
HAsync, cuMemcpyDtoA, cuMemcpyDtoD, cuMemcpyDtoH, cuMemcpyDtoHAsync, cuMemcpyHtoA,
cuMemcpyHtoAAsync, cuMemcpyHtoD, cuMemcpyHtoDAsync, cuMemFree, cuMemFreeHost, cuMemGe-
tAddressRange, cuMemGetInfo, cuMemHostAlloc, cuMemHostGetDevicePointer, cuMemsetD2D8, cuMem-
setD2D32, cuMemsetD8, cuMemsetD16, cuMemsetD32

3.25.2.35 cuMemsetD2D32 (CUdeviceptr dstDevice, unsigned int dstPitch, unsigned int ui, unsigned int
Width, unsigned int Height)

Sets the 2D memory range of Width 32-bit values to the specified value ui. Height specifies the number of rows
to set, and dstPitch specifies the number of bytes between each row. This function performs fastest when the pitch
is one that has been passed back by cuMemAllocPitch().

Parameters:
dstDevice - Destination device pointer
dstPitch - Pitch of destination device pointer
ui - Value to set
Width - Width of row
Height - Number of rows

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuArray3DCreate, cuArray3DGetDescriptor, cuArrayCreate, cuArrayDestroy, cuArrayGetDescriptor, cuMemAl-
loc, cuMemAllocHost, cuMemAllocPitch, cuMemcpy2D, cuMemcpy2DAsync, cuMemcpy2DUnaligned,
cuMemcpy3D, cuMemcpy3DAsync, cuMemcpyAtoA, cuMemcpyAtoD, cuMemcpyAtoH, cuMemcpyAto-
HAsync, cuMemcpyDtoA, cuMemcpyDtoD, cuMemcpyDtoH, cuMemcpyDtoHAsync, cuMemcpyHtoA,
cuMemcpyHtoAAsync, cuMemcpyHtoD, cuMemcpyHtoDAsync, cuMemFree, cuMemFreeHost, cuMemGe-
tAddressRange, cuMemGetInfo, cuMemHostAlloc, cuMemHostGetDevicePointer, cuMemsetD2D8, cuMem-
setD2D16, cuMemsetD8, cuMemsetD16, cuMemsetD32

3.25.2.36 cuMemsetD2D8 (CUdeviceptr dstDevice, unsigned int dstPitch, unsigned char uc, unsigned int
Width, unsigned int Height)

Sets the 2D memory range of Width 8-bit values to the specified value uc. Height specifies the number of rows to
set, and dstPitch specifies the number of bytes between each row. This function performs fastest when the pitch is
one that has been passed back by cuMemAllocPitch().

Parameters:
dstDevice - Destination device pointer

Generated for NVIDIA CUDA Library by Doxygen

164 Module Documentation

dstPitch - Pitch of destination device pointer

uc - Value to set
Width - Width of row
Height - Number of rows

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuArray3DCreate, cuArray3DGetDescriptor, cuArrayCreate, cuArrayDestroy, cuArrayGetDescriptor, cuMemAl-
loc, cuMemAllocHost, cuMemAllocPitch, cuMemcpy2D, cuMemcpy2DAsync, cuMemcpy2DUnaligned,
cuMemcpy3D, cuMemcpy3DAsync, cuMemcpyAtoA, cuMemcpyAtoD, cuMemcpyAtoH, cuMemcpyAto-
HAsync, cuMemcpyDtoA, cuMemcpyDtoD, cuMemcpyDtoH, cuMemcpyDtoHAsync, cuMemcpyHtoA,
cuMemcpyHtoAAsync, cuMemcpyHtoD, cuMemcpyHtoDAsync, cuMemFree, cuMemFreeHost, cuMemGe-
tAddressRange, cuMemGetInfo, cuMemHostAlloc, cuMemHostGetDevicePointer, cuMemsetD2D16, cuMem-
setD2D32, cuMemsetD8, cuMemsetD16, cuMemsetD32

3.25.2.37 cuMemsetD32 (CUdeviceptr dstDevice, unsigned int ui, unsigned int N)

Sets the memory range of N 32-bit values to the specified value ui.

Parameters:
dstDevice - Destination device pointer
ui - Value to set
N - Number of elements

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuArray3DCreate, cuArray3DGetDescriptor, cuArrayCreate, cuArrayDestroy, cuArrayGetDescriptor, cuMemAl-
loc, cuMemAllocHost, cuMemAllocPitch, cuMemcpy2D, cuMemcpy2DAsync, cuMemcpy2DUnaligned,
cuMemcpy3D, cuMemcpy3DAsync, cuMemcpyAtoA, cuMemcpyAtoD, cuMemcpyAtoH, cuMemcpyAto-
HAsync, cuMemcpyDtoA, cuMemcpyDtoD, cuMemcpyDtoH, cuMemcpyDtoHAsync, cuMemcpyHtoA,
cuMemcpyHtoAAsync, cuMemcpyHtoD, cuMemcpyHtoDAsync, cuMemFree, cuMemFreeHost, cuMemGe-
tAddressRange, cuMemGetInfo, cuMemHostAlloc, cuMemHostGetDevicePointer, cuMemsetD2D8, cuMem-
setD2D16, cuMemsetD2D32, cuMemsetD8, cuMemsetD16

Generated for NVIDIA CUDA Library by Doxygen

3.25 Memory Management 165

3.25.2.38 cuMemsetD8 (CUdeviceptr dstDevice, unsigned char uc, unsigned int N)

Sets the memory range of N 8-bit values to the specified value uc.

Parameters:
dstDevice - Destination device pointer
uc - Value to set
N - Number of elements

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuArray3DCreate, cuArray3DGetDescriptor, cuArrayCreate, cuArrayDestroy, cuArrayGetDescriptor, cuMemAl-
loc, cuMemAllocHost, cuMemAllocPitch, cuMemcpy2D, cuMemcpy2DAsync, cuMemcpy2DUnaligned,
cuMemcpy3D, cuMemcpy3DAsync, cuMemcpyAtoA, cuMemcpyAtoD, cuMemcpyAtoH, cuMemcpyAto-
HAsync, cuMemcpyDtoA, cuMemcpyDtoD, cuMemcpyDtoH, cuMemcpyDtoHAsync, cuMemcpyHtoA,
cuMemcpyHtoAAsync, cuMemcpyHtoD, cuMemcpyHtoDAsync, cuMemFree, cuMemFreeHost, cuMemGe-
tAddressRange, cuMemGetInfo, cuMemHostAlloc, cuMemHostGetDevicePointer, cuMemsetD2D8, cuMem-
setD2D16, cuMemsetD2D32, cuMemsetD16, cuMemsetD32

Generated for NVIDIA CUDA Library by Doxygen

166 Module Documentation

3.26 Texture Reference Management

Functions
• CUresult cuTexRefCreate (CUtexref ∗pTexRef)
Creates a texture reference.

• CUresult cuTexRefDestroy (CUtexref hTexRef)

Destroys a texture reference.

• CUresult cuTexRefGetAddress (CUdeviceptr ∗pdptr, CUtexref hTexRef)

Gets the address associated with a texture reference.

• CUresult cuTexRefGetAddressMode (CUaddress_mode ∗pam, CUtexref hTexRef, int dim)

Gets the addressing mode used by a texture reference.

• CUresult cuTexRefGetArray (CUarray ∗phArray, CUtexref hTexRef)

Gets the array bound to a texture reference.

• CUresult cuTexRefGetFilterMode (CUfilter_mode ∗pfm, CUtexref hTexRef)

Gets the filter-mode used by a texture reference.

• CUresult cuTexRefGetFlags (unsigned int ∗pFlags, CUtexref hTexRef)

Gets the flags used by a texture reference.

• CUresult cuTexRefGetFormat (CUarray_format ∗pFormat, int ∗pNumChannels, CUtexref hTexRef)

Gets the format used by a texture reference.

• CUresult cuTexRefSetAddress (unsigned int ∗ByteOffset, CUtexref hTexRef, CUdeviceptr dptr, unsigned int
bytes)
Binds an address as a texture reference.

• CUresult cuTexRefSetAddress2D (CUtexref hTexRef, const CUDA_ARRAY_DESCRIPTOR ∗desc, CUdevi-

ceptr dptr, unsigned int Pitch)
Binds an address as a 2D texture reference.

• CUresult cuTexRefSetAddressMode (CUtexref hTexRef, int dim, CUaddress_mode am)

Sets the addressing mode for a texture reference.

• CUresult cuTexRefSetArray (CUtexref hTexRef, CUarray hArray, unsigned int Flags)

Binds an address as a texture reference.

• CUresult cuTexRefSetFilterMode (CUtexref hTexRef, CUfilter_mode fm)

Sets the filtering mode for a texture reference.

• CUresult cuTexRefSetFlags (CUtexref hTexRef, unsigned int Flags)

Sets the flags for a texture reference.

• CUresult cuTexRefSetFormat (CUtexref hTexRef, CUarray_format fmt, int NumPackedComponents)

Sets the format for a texture reference.

Generated for NVIDIA CUDA Library by Doxygen

3.26 Texture Reference Management 167

3.26.1 Detailed Description

This section describes the texture reference management functions of the low-level CUDA driver application program-
ming interface.

3.26.2 Function Documentation

3.26.2.1 cuTexRefCreate (CUtexref ∗ pTexRef)

Creates a texture reference and returns its handle in ∗pTexRef. Once created, the application must call cuTexRefSe-
tArray() or cuTexRefSetAddress() to associate the reference with allocated memory. Other texture reference functions
are used to specify the format and interpretation (addressing, filtering, etc.) to be used when the memory is read
through this texture reference. To associate the texture reference with a texture ordinal for a given function, the
application should call cuParamSetTexRef().

Parameters:
pTexRef - Returned texture reference

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

See also:
cuTexRefDestroy, cuTexRefSetAddress, cuTexRefSetAddress2D, cuTexRefSetAddressMode, cuTexRefSetAr-
ray, cuTexRefSetFilterMode, cuTexRefSetFlags, cuTexRefSetFormat, cuTexRefGetAddress, cuTexRefGetAd-
dressMode, cuTexRefGetArray, cuTexRefGetFilterMode, cuTexRefGetFlags, cuTexRefGetFormat

3.26.2.2 cuTexRefDestroy (CUtexref hTexRef)

Destroys the texture reference specified by hTexRef.

Parameters:
hTexRef - Texture reference to destroy

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

See also:
cuTexRefCreate, cuTexRefSetAddress, cuTexRefSetAddress2D, cuTexRefSetAddressMode, cuTexRefSetArray,
cuTexRefSetFilterMode, cuTexRefSetFlags, cuTexRefSetFormat, cuTexRefGetAddress, cuTexRefGetAddress-
Mode, cuTexRefGetArray, cuTexRefGetFilterMode, cuTexRefGetFlags, cuTexRefGetFormat

3.26.2.3 cuTexRefGetAddress (CUdeviceptr ∗ pdptr, CUtexref hTexRef)

Returns in ∗pdptr the base address bound to the texture reference hTexRef, or returns CUDA_ERROR_-
INVALID_VALUE if the texture reference is not bound to any device memory range.

Generated for NVIDIA CUDA Library by Doxygen

168 Module Documentation

Parameters:
pdptr - Returned device address
hTexRef - Texture reference

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

See also:
cuTexRefCreate, cuTexRefDestroy, cuTexRefSetAddress, cuTexRefSetAddress2D, cuTexRefSetAddressMode,
cuTexRefSetArray, cuTexRefSetFilterMode, cuTexRefSetFlags, cuTexRefSetFormat, cuTexRefGetAddress-
Mode, cuTexRefGetArray, cuTexRefGetFilterMode, cuTexRefGetFlags, cuTexRefGetFormat

3.26.2.4 cuTexRefGetAddressMode (CUaddress_mode ∗ pam, CUtexref hTexRef, int dim)

Returns in ∗pam the addressing mode corresponding to the dimension dim of the texture reference hTexRef. Cur-
rently, the only valid value for dim are 0 and 1.

Parameters:
pam - Returned addressing mode
hTexRef - Texture reference
dim - Dimension
Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

See also:
cuTexRefCreate, cuTexRefDestroy, cuTexRefSetAddress, cuTexRefSetAddress2D, cuTexRefSetAddressMode,
cuTexRefSetArray, cuTexRefSetFilterMode, cuTexRefSetFlags, cuTexRefSetFormat, cuTexRefGetAddress, cu-
TexRefGetArray, cuTexRefGetFilterMode, cuTexRefGetFlags, cuTexRefGetFormat

3.26.2.5 cuTexRefGetArray (CUarray ∗ phArray, CUtexref hTexRef)

Returns in ∗phArray the CUDA array bound to the texture reference hTexRef, or returns CUDA_ERROR_-
INVALID_VALUE if the texture reference is not bound to any CUDA array.

Parameters:
phArray - Returned array
hTexRef - Texture reference

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

See also:
cuTexRefCreate, cuTexRefDestroy, cuTexRefSetAddress, cuTexRefSetAddress2D, cuTexRefSetAddressMode,
cuTexRefSetArray, cuTexRefSetFilterMode, cuTexRefSetFlags, cuTexRefSetFormat, cuTexRefGetAddress, cu-
TexRefGetAddressMode, cuTexRefGetFilterMode, cuTexRefGetFlags, cuTexRefGetFormat

Generated for NVIDIA CUDA Library by Doxygen

3.26 Texture Reference Management 169

3.26.2.6 cuTexRefGetFilterMode (CUfilter_mode ∗ pfm, CUtexref hTexRef)

Returns in ∗pfm the filtering mode of the texture reference hTexRef.

Parameters:
pfm - Returned filtering mode
hTexRef - Texture reference

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

See also:
cuTexRefCreate, cuTexRefDestroy, cuTexRefSetAddress, cuTexRefSetAddress2D, cuTexRefSetAddressMode,
cuTexRefSetArray, cuTexRefSetFilterMode, cuTexRefSetFlags, cuTexRefSetFormat, cuTexRefGetAddress, cu-
TexRefGetAddressMode, cuTexRefGetArray, cuTexRefGetFlags, cuTexRefGetFormat

3.26.2.7 cuTexRefGetFlags (unsigned int ∗ pFlags, CUtexref hTexRef)

Returns in ∗pFlags the flags of the texture reference hTexRef.

Parameters:
pFlags - Returned flags
hTexRef - Texture reference

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

See also:
cuTexRefCreate, cuTexRefDestroy, cuTexRefSetAddress, cuTexRefSetAddress2D, cuTexRefSetAddressMode,
cuTexRefSetArray, cuTexRefSetFilterMode, cuTexRefSetFlags, cuTexRefSetFormat, cuTexRefGetAddress, cu-
TexRefGetAddressMode, cuTexRefGetArray, cuTexRefGetFilterMode, cuTexRefGetFormat

3.26.2.8 cuTexRefGetFormat (CUarray_format ∗ pFormat, int ∗ pNumChannels, CUtexref hTexRef)

Returns in ∗pFormat and ∗pNumChannels the format and number of components of the CUDA array bound to
the texture reference hTexRef. If pFormat or pNumChannels is NULL, it will be ignored.

Parameters:
pFormat - Returned format
pNumChannels - Returned number of components
hTexRef - Texture reference

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Generated for NVIDIA CUDA Library by Doxygen

170 Module Documentation

See also:
cuTexRefCreate, cuTexRefDestroy, cuTexRefSetAddress, cuTexRefSetAddress2D, cuTexRefSetAddressMode,
cuTexRefSetArray, cuTexRefSetFilterMode, cuTexRefSetFlags, cuTexRefSetFormat, cuTexRefGetAddress, cu-
TexRefGetAddressMode, cuTexRefGetArray, cuTexRefGetFilterMode, cuTexRefGetFlags

3.26.2.9 cuTexRefSetAddress (unsigned int ∗ ByteOffset, CUtexref hTexRef, CUdeviceptr dptr, unsigned int
bytes)

Binds a linear address range to the texture reference hTexRef. Any previous address or CUDA array state associated
with the texture reference is superseded by this function. Any memory previously bound to hTexRef is unbound.
Since the hardware enforces an alignment requirement on texture base addresses, cuTexRefSetAddress() passes back
a byte offset in ∗ByteOffset that must be applied to texture fetches in order to read from the desired memory. This
offset must be divided by the texel size and passed to kernels that read from the texture so they can be applied to the
tex1Dfetch() function.
If the device memory pointer was returned from cuMemAlloc(), the offset is guaranteed to be 0 and NULL may be
passed as the ByteOffset parameter.

Parameters:
ByteOffset - Returned byte offset
hTexRef - Texture reference to bind
dptr - Device pointer to bind
bytes - Size of memory to bind in bytes

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

See also:
cuTexRefCreate, cuTexRefDestroy, cuTexRefSetAddress2D, cuTexRefSetAddressMode, cuTexRefSetArray, cu-
TexRefSetFilterMode, cuTexRefSetFlags, cuTexRefSetFormat, cuTexRefGetAddress, cuTexRefGetAddress-
Mode, cuTexRefGetArray, cuTexRefGetFilterMode, cuTexRefGetFlags, cuTexRefGetFormat

3.26.2.10 cuTexRefSetAddress2D (CUtexref hTexRef, const CUDA_ARRAY_DESCRIPTOR ∗ desc,

CUdeviceptr dptr, unsigned int Pitch)

Binds a linear address range to the texture reference hTexRef. Any previous address or CUDA array state associated
with the texture reference is superseded by this function. Any memory previously bound to hTexRef is unbound.
When using a tex2D() function inside a kernel, we have to either call cuTexRefSetArray() to bind the corresponding
texture reference to an array, or cuTexRefSetAddress2D() to bind the texture reference to linear memory.
Function calls to cuTexRefSetFormat() cannot follow calls to cuTexRefSetAddress2D() for the same texture reference.

Parameters:
hTexRef - Texture reference to bind
desc - Descriptor of CUDA array
dptr - Device pointer to bind

Generated for NVIDIA CUDA Library by Doxygen

3.26 Texture Reference Management 171

Pitch - Line pitch in bytes

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

See also:
cuTexRefCreate, cuTexRefDestroy, cuTexRefSetAddress, cuTexRefSetAddressMode, cuTexRefSetArray, cu-
TexRefSetFilterMode, cuTexRefSetFlags, cuTexRefSetFormat, cuTexRefGetAddress, cuTexRefGetAddress-
Mode, cuTexRefGetArray, cuTexRefGetFilterMode, cuTexRefGetFlags, cuTexRefGetFormat

3.26.2.11 cuTexRefSetAddressMode (CUtexref hTexRef, int dim, CUaddress_mode am)

Specifies the addressing mode am for the given dimension dim of the texture reference hTexRef. If dim is zero,
the addressing mode is applied to the first parameter of the functions used to fetch from the texture; if dim is 1, the
second, and so on. CUaddress_mode is defined as:

typedef enum CUaddress_mode_enum {

CU_TR_ADDRESS_MODE_WRAP = 0,
CU_TR_ADDRESS_MODE_CLAMP = 1,
CU_TR_ADDRESS_MODE_MIRROR = 2,
} CUaddress_mode;

Note that this call has no effect if hTexRef is bound to linear memory.

Parameters:
hTexRef - Texture reference
dim - Dimension
am - Addressing mode to set

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

See also:
cuTexRefCreate, cuTexRefDestroy, cuTexRefSetAddress, cuTexRefSetAddress2D, cuTexRefSetArray, cu-
TexRefSetFilterMode, cuTexRefSetFlags, cuTexRefSetFormat, cuTexRefGetAddress, cuTexRefGetAddress-
Mode, cuTexRefGetArray, cuTexRefGetFilterMode, cuTexRefGetFlags, cuTexRefGetFormat

3.26.2.12 cuTexRefSetArray (CUtexref hTexRef, CUarray hArray, unsigned int Flags)

Binds the CUDA array hArray to the texture reference hTexRef. Any previous address or CUDA array state
associated with the texture reference is superseded by this function. Flags must be set to CU_TRSA_OVERRIDE_-
FORMAT. Any CUDA array previously bound to hTexRef is unbound.

Parameters:
hTexRef - Texture reference to bind

Generated for NVIDIA CUDA Library by Doxygen

172 Module Documentation

hArray - Array to bind

Flags - Options (must be CU_TRSA_OVERRIDE_FORMAT)

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

See also:
cuTexRefCreate, cuTexRefDestroy, cuTexRefSetAddress, cuTexRefSetAddress2D, cuTexRefSetAddressMode,
cuTexRefSetFilterMode, cuTexRefSetFlags, cuTexRefSetFormat, cuTexRefGetAddress, cuTexRefGetAddress-
Mode, cuTexRefGetArray, cuTexRefGetFilterMode, cuTexRefGetFlags, cuTexRefGetFormat

3.26.2.13 cuTexRefSetFilterMode (CUtexref hTexRef, CUfilter_mode fm)

Specifies the filtering mode fm to be used when reading memory through the texture reference hTexRef. CUfilter_-
mode_enum is defined as:

typedef enum CUfilter_mode_enum {

CU_TR_FILTER_MODE_POINT = 0,
CU_TR_FILTER_MODE_LINEAR = 1
} CUfilter_mode;

Note that this call has no effect if hTexRef is bound to linear memory.

Parameters:
hTexRef - Texture reference
fm - Filtering mode to set

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

See also:
cuTexRefCreate, cuTexRefDestroy, cuTexRefSetAddress, cuTexRefSetAddress2D, cuTexRefSetAddressMode,
cuTexRefSetArray, cuTexRefSetFlags, cuTexRefSetFormat, cuTexRefGetAddress, cuTexRefGetAddressMode,
cuTexRefGetArray, cuTexRefGetFilterMode, cuTexRefGetFlags, cuTexRefGetFormat

3.26.2.14 cuTexRefSetFlags (CUtexref hTexRef, unsigned int Flags)

Specifies optional flags via Flags to specify the behavior of data returned through the texture reference hTexRef.
The valid flags are:

• CU_TRSF_READ_AS_INTEGER, which suppresses the default behavior of having the texture promote integer
data to floating point data in the range [0, 1];
• CU_TRSF_NORMALIZED_COORDINATES, which suppresses the default behavior of having the texture co-
ordinates range from [0, Dim) where Dim is the width or height of the CUDA array. Instead, the texture
coordinates [0, 1.0) reference the entire breadth of the array dimension;

Generated for NVIDIA CUDA Library by Doxygen

3.26 Texture Reference Management 173

Parameters:
hTexRef - Texture reference
Flags - Optional flags to set

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

See also:
cuTexRefCreate, cuTexRefDestroy, cuTexRefSetAddress, cuTexRefSetAddress2D, cuTexRefSetAddressMode,
cuTexRefSetArray, cuTexRefSetFilterMode, cuTexRefSetFormat, cuTexRefGetAddress, cuTexRefGetAddress-
Mode, cuTexRefGetArray, cuTexRefGetFilterMode, cuTexRefGetFlags, cuTexRefGetFormat

3.26.2.15 cuTexRefSetFormat (CUtexref hTexRef, CUarray_format fmt, int NumPackedComponents)

Specifies the format of the data to be read by the texture reference hTexRef. fmt and NumPackedComponents
are exactly analogous to the Format and NumChannels members of the CUDA_ARRAY_DESCRIPTOR structure:
They specify the format of each component and the number of components per array element.

Parameters:
hTexRef - Texture reference
fmt - Format to set
NumPackedComponents - Number of components per array element

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

See also:
cuTexRefCreate, cuTexRefDestroy, cuTexRefSetAddress, cuTexRefSetAddress2D, cuTexRefSetAddressMode,
cuTexRefSetArray, cuTexRefSetFilterMode, cuTexRefSetFlags, cuTexRefGetAddress, cuTexRefGetAddress-
Mode, cuTexRefGetArray, cuTexRefGetFilterMode, cuTexRefGetFlags, cuTexRefGetFormat

Generated for NVIDIA CUDA Library by Doxygen

174 Module Documentation

3.27 OpenGL Interoperability

Functions
• CUresult cuGLCtxCreate (CUcontext ∗pCtx, unsigned int Flags, CUdevice device)
Create a CUDA context for interoperability with OpenGL.

• CUresult cuGLInit (void)

Initializes OpenGL interoperability.

• CUresult cuGLMapBufferObject (CUdeviceptr ∗dptr, unsigned int ∗size, GLuint bufferobj)

Maps an OpenGL buffer object.

• CUresult cuGLMapBufferObjectAsync (CUdeviceptr ∗dptr, unsigned int ∗size, GLuint bufferobj, CUstream
hStream)
Maps an OpenGL buffer object.

• CUresult cuGLRegisterBufferObject (GLuint bufferobj)

Registers an OpenGL buffer object.

• CUresult cuGLSetBufferObjectMapFlags (GLuint bufferobj, unsigned int Flags)

Set the map flags for an OpenGL buffer object.

• CUresult cuGLUnmapBufferObject (GLuint bufferobj)

Unmaps an OpenGL buffer object.

• CUresult cuGLUnmapBufferObjectAsync (GLuint bufferobj, CUstream hStream)

Unmaps an OpenGL buffer object.

• CUresult cuGLUnregisterBufferObject (GLuint bufferobj)

Unregister an OpenGL buffer object.

• CUresult cuWGLGetDevice (CUdevice ∗pDevice, HGPUNV hGpu)

Gets the CUDA device associated with hGpu.

3.27.1 Detailed Description

This section describes the OpenGL interoperability functions of the low-level CUDA driver application programming
interface.

3.27.2 Function Documentation

3.27.2.1 cuGLCtxCreate (CUcontext ∗ pCtx, unsigned int Flags, CUdevice device)

Creates a new CUDA context, initializes OpenGL interoperability, and associates the CUDA context with the calling
thread. It must be called before performing any other OpenGL interoperability operations. It may fail if the needed
OpenGL driver facilities are not available. For usage of the Flags parameter, see cuCtxCreate().

Generated for NVIDIA CUDA Library by Doxygen

3.27 OpenGL Interoperability 175

Parameters:
pCtx - Returned CUDA context
Flags - Options for CUDA context creation
device - Device on which to create the context

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_OUT_OF_MEMORY

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuCtxCreate, cuGLInit, cuGLMapBufferObject, cuGLRegisterBufferObject, cuGLUnmapBufferObject,
cuGLUnregisterBufferObject, cuGLMapBufferObjectAsync, cuGLUnmapBufferObjectAsync, cuGLSetBuffer-
ObjectMapFlags, cuWGLGetDevice

3.27.2.2 cuGLInit (void)

Initializes OpenGL interoperability. It must be called before performing any other OpenGL interoperability operations.
It may fail if the needed OpenGL driver facilities are not available.

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_UNKNOWN

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuGLCtxCreate, cuGLMapBufferObject, cuGLRegisterBufferObject, cuGLUnmapBufferObject, cuGLUn-
registerBufferObject, cuGLMapBufferObjectAsync, cuGLUnmapBufferObjectAsync, cuGLSetBufferOb-
jectMapFlags, cuWGLGetDevice

3.27.2.3 cuGLMapBufferObject (CUdeviceptr ∗ dptr, unsigned int ∗ size, GLuint bufferobj)

Maps the buffer object of ID bufferobj into the address space of the current CUDA context and returns in ∗dptr
and ∗size the base pointer and size of the resulting mapping.
All streams in the current CUDA context are synchronized with the current GL context.

Parameters:
dptr - Returned mapped base pointer
size - Returned size of mapping
bufferobj - Buffer object to map

Generated for NVIDIA CUDA Library by Doxygen

176 Module Documentation

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_MAP_FAILED

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuGLCtxCreate, cuGLInit, cuGLRegisterBufferObject, cuGLUnmapBufferObject, cuGLUnregisterBufferObject,
cuGLMapBufferObjectAsync, cuGLUnmapBufferObjectAsync, cuGLSetBufferObjectMapFlags, cuWGLGet-
Device

3.27.2.4 cuGLMapBufferObjectAsync (CUdeviceptr ∗ dptr, unsigned int ∗ size, GLuint bufferobj,

CUstream hStream)

Maps the buffer object of ID bufferobj into the address space of the current CUDA context and returns in ∗dptr
and ∗size the base pointer and size of the resulting mapping.
Stream hStream in the current CUDA context is synchronized with the current GL context.

Parameters:
dptr - Returned mapped base pointer
size - Returned size of mapping
bufferobj - Buffer object to map
hStream - Stream to synchronize

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_MAP_FAILED

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuGLCtxCreate, cuGLInit, cuGLRegisterBufferObject, cuGLUnmapBufferObject, cuGLUnregisterBufferObject,
cuGLUnmapBufferObjectAsync, cuGLSetBufferObjectMapFlags,cuWGLGetDevice

3.27.2.5 cuGLRegisterBufferObject (GLuint bufferobj)

Registers the buffer object of ID bufferobj for access by CUDA. This function must be called before CUDA can
map the buffer object. While it is registered, the buffer object cannot be used by any OpenGL commands, except as a
data source for OpenGL drawing commands.

Parameters:
bufferobj - Buffer object to register

Generated for NVIDIA CUDA Library by Doxygen

3.27 OpenGL Interoperability 177

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_ALREADY_MAPPED

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuGLCtxCreate, cuGLInit, cuGLMapBufferObject, cuGLUnmapBufferObject, cuGLMapBufferObjectAsync,
cuGLUnmapBufferObjectAsync, cuGLSetBufferObjectMapFlags, cuGLUnregisterBufferObject, cuWGLGetDe-
vice

3.27.2.6 cuGLSetBufferObjectMapFlags (GLuint bufferobj, unsigned int Flags)

Sets the map flags for the registered buffer object of ID bufferobj.
Changes to Flags will take effect the next time bufferobj is mapped. The Flags argument may be any of the
following:

• CU_GL_MAP_RESOURCE_FLAGS_NONE: Specifies no hints about how this resource will be used. It is

therefore assumed that this resource will be read from and written to by CUDA kernels. This is the default
value.

• CU_GL_MAP_RESOURCE_FLAGS_READ_ONLY: Specifies that CUDA kernels which access this resource

will not write to this resource.

• CU_GL_MAP_RESOURCE_FLAGS_WRITE_DISCARD: Specifies that CUDA kernels which access this re-

source will not read from this resource and will write over the entire contents of the resource, so none of the
data previously stored in the resource will be preserved.

If bufferobj has not been registered for use with CUDA, then CUDA_ERROR_INVALID_HANDLE is returned.
If bufferobj is presently mapped for access by CUDA, then CUDA_ERROR_ALREADY_MAPPED is returned.

Parameters:
bufferobj - Buffer object to unmap
Flags - Map flags

Returns:
CUDA_SUCCESS, CUDA_ERROR_NOT_INITIALIZED, CUDA_ERROR_INVALID_HANDLE, CUDA_-
ERROR_ALREADY_MAPPED, CUDA_ERROR_INVALID_CONTEXT,

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuGLCtxCreate, cuGLInit, cuGLMapBufferObject, cuGLRegisterBufferObject, cuGLUnregisterBufferObject,
cuGLUnmapBufferObject, cuGLMapBufferObjectAsync, cuGLUnmapBufferObjectAsync, cuWGLGetDevice

Generated for NVIDIA CUDA Library by Doxygen

178 Module Documentation

3.27.2.7 cuGLUnmapBufferObject (GLuint bufferobj)

Unmaps the buffer object of ID bufferobj for access by CUDA.

All streams in the current CUDA context are synchronized with the current GL context.

Parameters:
bufferobj - Buffer object to unmap

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuGLCtxCreate, cuGLInit, cuGLMapBufferObject, cuGLRegisterBufferObject, cuGLUnregisterBufferObject,
cuGLMapBufferObjectAsync, cuGLUnmapBufferObjectAsync, cuGLSetBufferObjectMapFlags, cuWGLGet-
Device

3.27.2.8 cuGLUnmapBufferObjectAsync (GLuint bufferobj, CUstream hStream)

Unmaps the buffer object of ID bufferobj for access by CUDA.

Stream hStream in the current CUDA context is synchronized with the current GL context.

Parameters:
bufferobj - Buffer object to unmap
hStream - Stream to synchronize

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuGLCtxCreate, cuGLInit, cuGLMapBufferObject, cuGLRegisterBufferObject, cuGLUnregisterBufferObject,
cuGLMapBufferObjectAsync, cuGLSetBufferObjectMapFlags, cuWGLGetDevice

3.27.2.9 cuGLUnregisterBufferObject (GLuint bufferobj)

Unregisters the buffer object of ID bufferobj for access by CUDA.

Parameters:
bufferobj - Buffer object to unmap

Generated for NVIDIA CUDA Library by Doxygen

3.27 OpenGL Interoperability 179

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuGLCtxCreate, cuGLInit, cuGLMapBufferObject, cuGLRegisterBufferObject, cuGLUnmapBufferObject,
cuGLMapBufferObjectAsync, cuGLUnmapBufferObjectAsync, cuGLSetBufferObjectMapFlags, cuWGLGet-
Device

3.27.2.10 cuWGLGetDevice (CUdevice ∗ pDevice, HGPUNV hGpu)

Returns in ∗pDevice the CUDA device associated with a hGpu, if applicable.

Parameters:
pDevice - Device associated with hGpu
hGpu - Handle to a GPU, as queried via WGL_NV_gpu_affinity()

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuGLCtxCreate, cuGLInit, cuGLMapBufferObject, cuGLRegisterBufferObject, cuGLUnmapBufferObject,
cuGLUnregisterBufferObject, cuGLUnmapBufferObjectAsync, cuGLSetBufferObjectMapFlags

Generated for NVIDIA CUDA Library by Doxygen

180 Module Documentation

3.28 Direct3D 9 Interoperability

Functions
• CUresult cuD3D9CtxCreate (CUcontext ∗pCtx, CUdevice ∗pCuDevice, unsigned int Flags, IDirect3DDevice9
∗pDxDevice)
Create a CUDA context for interoperability with Direct3D.

• CUresult cuD3D9GetDevice (CUdevice ∗pDevice, const char ∗pszAdapterName)

Gets the device number for an adapter.

• CUresult cuD3D9GetDirect3DDevice (IDirect3DDevice9 ∗∗ppDxDevice)

Get the Direct3D device against which the current CUDA context was created.

• CUresult cuD3D9MapResources (unsigned int count, IDirect3DResource9 ∗∗ppResource)

Map Direct3D resources for access by CUDA.

• CUresult cuD3D9RegisterResource (IDirect3DResource9 ∗pResource, unsigned int Flags)

Register a Direct3D resource for access by CUDA.

• CUresult cuD3D9ResourceGetMappedArray (CUarray ∗pArray, IDirect3DResource9 ∗pResource, unsigned int

Face, unsigned int Level)
Get an array through which to access a subresource of a Direct3D resource which has been mapped for access by CUDA.

• CUresult cuD3D9ResourceGetMappedPitch (unsigned int ∗pPitch, unsigned int ∗pPitchSlice, IDi-

rect3DResource9 ∗pResource, unsigned int Face, unsigned int Level)
Get the pitch of a subresource of a Direct3D resource which has been mapped for access by CUDA.

• CUresult cuD3D9ResourceGetMappedPointer (CUdeviceptr ∗pDevPtr, IDirect3DResource9 ∗pResource, un-

signed int Face, unsigned int Level)
Get the pointer through which to access a subresource of a Direct3D resource which has been mapped for access by
CUDA.

• CUresult cuD3D9ResourceGetMappedSize (unsigned int ∗pSize, IDirect3DResource9 ∗pResource, unsigned

int Face, unsigned int Level)
Get the size of a subresource of a Direct3D resource which has been mapped for access by CUDA.

• CUresult cuD3D9ResourceGetSurfaceDimensions (unsigned int ∗pWidth, unsigned int ∗pHeight, unsigned int
∗pDepth, IDirect3DResource9 ∗pResource, unsigned int Face, unsigned int Level)
Get the dimensions of a registered surface.

• CUresult cuD3D9ResourceSetMapFlags (IDirect3DResource9 ∗pResource, unsigned int Flags)

Set usage flags for mapping a Direct3D resource.

• CUresult cuD3D9UnmapResources (unsigned int count, IDirect3DResource9 ∗∗ppResource)

Unmaps Direct3D resources.

• CUresult cuD3D9UnregisterResource (IDirect3DResource9 ∗pResource)

Unregister a Direct3D resource.

Generated for NVIDIA CUDA Library by Doxygen

3.28 Direct3D 9 Interoperability 181

3.28.1 Detailed Description

This section describes the Direct3D 9 interoperability functions of the low-level CUDA driver application program-
ming interface.

3.28.2 Function Documentation

3.28.2.1 cuD3D9CtxCreate (CUcontext ∗ pCtx, CUdevice ∗ pCuDevice, unsigned int Flags,

IDirect3DDevice9 ∗ pDxDevice)

Creates a new CUDA context, enables interoperability for that context with the Direct3D device pDxDevice, and as-
sociates the created CUDA context with the calling thread. The CUcontext will be returned in ∗pCtx. If pCuDevice
is non-NULL, then the CUdevice on which this CUDA context was created will be returned in ∗pCuDevice. For
usage of the Flags parameter, see cuCtxCreate(). Direct3D resources from this device may be registered and mapped
through the lifetime of this CUDA context.
This context will function only until its Direct3D device is destroyed. On success, this call will increase the internal
reference count on pDxDevice. This reference count will be decremented upon destruction of this context through
cuCtxDestroy().

Parameters:
pCtx - Returned newly created CUDA context
pCuDevice - Returned pointer to device on which context was created
Flags - Context creation flags (see cuCtxCreate() for details)
pDxDevice - Direct3D device to create interoperability context with

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_VALUE, CUDA_ERROR_OUT_OF_MEMORY, CUDA_ERROR_UNKNOWN

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuD3D9GetDevice, cuD3D9GetDirect3DDevice, cuD3D9MapResources, cuD3D9RegisterResource,
cuD3D9ResourceGetMappedArray, cuD3D9ResourceGetMappedPitch, cuD3D9ResourceGetMappedPointer,
cuD3D9ResourceGetMappedSize, cuD3D9ResourceGetSurfaceDimensions, cuD3D9ResourceSetMapFlags,
cuD3D9UnmapResources, cuD3D9UnregisterResource

3.28.2.2 cuD3D9GetDevice (CUdevice ∗ pDevice, const char ∗ pszAdapterName)

Returns in ∗pDevice the CUDA-compatible device corresponding to the adapter name pszAdapterName ob-
tained from EnumDisplayDevices() or IDirect3D9::GetAdapterIdentifier(). If no device on the adapter with name
pszAdapterName is CUDA-compatible, then the call will fail.

Parameters:
pDevice - Returned CUDA device corresponding to pszAdapterName
pszAdapterName - Adapter name to query for device

Generated for NVIDIA CUDA Library by Doxygen

182 Module Documentation

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_VALUE, CUDA_ERROR_UNKNOWN

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuD3D9CtxCreate, cuD3D9GetDirect3DDevice, cuD3D9MapResources, cuD3D9RegisterResource,
cuD3D9ResourceGetMappedArray, cuD3D9ResourceGetMappedPitch, cuD3D9ResourceGetMappedPointer,
cuD3D9ResourceGetMappedSize, cuD3D9ResourceGetSurfaceDimensions, cuD3D9ResourceSetMapFlags,
cuD3D9UnmapResources, cuD3D9UnregisterResource

3.28.2.3 cuD3D9GetDirect3DDevice (IDirect3DDevice9 ∗∗ ppDxDevice)

Returns in ∗ppDxDevice the Direct3D device against which this CUDA context was created in cuD3D9CtxCreate().

Parameters:
ppDxDevice - Returned Direct3D device corresponding to CUDA context

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuD3D9CtxCreate, cuD3D9GetDevice, cuD3D9MapResources, cuD3D9RegisterResource,
cuD3D9ResourceGetMappedArray, cuD3D9ResourceGetMappedPitch, cuD3D9ResourceGetMappedPointer,
cuD3D9ResourceGetMappedSize, cuD3D9ResourceGetSurfaceDimensions, cuD3D9ResourceSetMapFlags,
cuD3D9UnmapResources, cuD3D9UnregisterResource

3.28.2.4 cuD3D9MapResources (unsigned int count, IDirect3DResource9 ∗∗ ppResource)

Maps the count Direct3D resources in ppResource for access by CUDA.

The resources in ppResource may be accessed in CUDA kernels until they are unmapped. Direct3D should not
access any resources while they are mapped by CUDA. If an application does so the results are undefined.
This function provides the synchronization guarantee that any Direct3D calls issued before cuD3D9MapResources()
will complete before any CUDA kernels issued after cuD3D9MapResources() begin.
If any of ppResource have not been registered for use with CUDA or if ppResource contains any duplicate
entries, then CUDA_ERROR_INVALID_HANDLE is returned. If any of ppResource are presently mapped for
access by CUDA, then CUDA_ERROR_ALREADY_MAPPED is returned.

Parameters:
count - Number of resources in ppResource

Generated for NVIDIA CUDA Library by Doxygen

3.28 Direct3D 9 Interoperability 183

ppResource - Resources to map for CUDA usage

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_HANDLE, CUDA_ERROR_ALREADY_-
MAPPED, CUDA_ERROR_UNKNOWN

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuD3D9CtxCreate, cuD3D9GetDevice, cuD3D9GetDirect3DDevice, cuD3D9RegisterResource,
cuD3D9ResourceGetMappedArray, cuD3D9ResourceGetMappedPitch, cuD3D9ResourceGetMappedPointer,
cuD3D9ResourceGetMappedSize, cuD3D9ResourceGetSurfaceDimensions, cuD3D9ResourceSetMapFlags,
cuD3D9UnmapResources, cuD3D9UnregisterResource

3.28.2.5 cuD3D9RegisterResource (IDirect3DResource9 ∗ pResource, unsigned int Flags)

Registers the Direct3D resource pResource for access by CUDA.

If this call is successful, then the application will be able to map and unmap this resource until it is unregistered through
cuD3D9UnregisterResource(). Also on success, this call will increase the internal reference count on pResource.
This reference count will be decremented when this resource is unregistered through cuD3D9UnregisterResource().
This call is potentially high-overhead and should not be called every frame in interactive applications.
The type of pResource must be one of the following.

• IDirect3DVertexBuffer9: Cannot be used with Flags set to CU_D3D9_REGISTER_FLAGS_ARRAY.

• IDirect3DIndexBuffer9: Cannot be used with Flags set to CU_D3D9_REGISTER_FLAGS_ARRAY.
• IDirect3DSurface9: Only stand-alone objects of type IDirect3DSurface9 may be explicitly shared. In particular,
individual mipmap levels and faces of cube maps may not be registered directly. To access individual surfaces
associated with a texture, one must register the base texture object. For restrictions on the Flags parameter,
see type IDirect3DBaseTexture9.
• IDirect3DBaseTexture9: When a texture is registered, all surfaces associated with the all mipmap levels of all
faces of the texture will be accessible to CUDA.

The Flags argument specifies the mechanism through which CUDA will access the Direct3D resource. The following
values are allowed.

• CU_D3D9_REGISTER_FLAGS_NONE: Specifies that CUDA will access this resource through a CUd-
eviceptr. The pointer, size, and (for textures), pitch for each subresource of this allocation
may be queried through cuD3D9ResourceGetMappedPointer(), cuD3D9ResourceGetMappedSize(), and
cuD3D9ResourceGetMappedPitch() respectively. This option is valid for all resource types.
• CU_D3D9_REGISTER_FLAGS_ARRAY: Specifies that CUDA will access this resource through a CUarray
queried on a sub-resource basis through cuD3D9ResourceGetMappedArray(). This option is only valid for
resources of type IDirect3DSurface9 and subtypes of IDirect3DBaseTexture9.

Not all Direct3D resources of the above types may be used for interoperability with CUDA. The following are some
limitations.

Generated for NVIDIA CUDA Library by Doxygen

184 Module Documentation

• The primary rendertarget may not be registered with CUDA.

• Resources allocated as shared may not be registered with CUDA.

• Any resources allocated in D3DPOOL_SYSTEMMEM or D3DPOOL_MANAGED may not be registered with

CUDA.

• Textures which are not of a format which is 1, 2, or 4 channels of 8, 16, or 32-bit integer or floating-point data
cannot be shared.

• Surfaces of depth or stencil formats cannot be shared.

If Direct3D interoperability is not initialized on this context, then CUDA_ERROR_INVALID_CONTEXT is re-

turned. If pResource is of incorrect type (e.g. is a non-stand-alone IDirect3DSurface9) or is already registered,
then CUDA_ERROR_INVALID_HANDLE is returned. If pResource cannot be registered then CUDA_ERROR_-
UNKNOWN is returned.

Parameters:
pResource - Resource to register for CUDA access
Flags - Flags for resource registration

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_INVALID_HANDLE,
CUDA_ERROR_OUT_OF_MEMORY, CUDA_ERROR_UNKNOWN

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuD3D9CtxCreate, cuD3D9GetDevice, cuD3D9GetDirect3DDevice, cuD3D9MapResources,
cuD3D9ResourceGetMappedArray, cuD3D9ResourceGetMappedPitch, cuD3D9ResourceGetMappedPointer,
cuD3D9ResourceGetMappedSize, cuD3D9ResourceGetSurfaceDimensions, cuD3D9ResourceSetMapFlags,
cuD3D9UnmapResources, cuD3D9UnregisterResource

3.28.2.6 cuD3D9ResourceGetMappedArray (CUarray ∗ pArray, IDirect3DResource9 ∗ pResource, unsigned

int Face, unsigned int Level)

Returns in ∗pArray an array through which the subresource of the mapped Direct3D resource pResource
which corresponds to Face and Level may be accessed. The value set in pArray may change every time that
pResource is mapped.
If pResource is not registered then CUDA_ERROR_INVALID_HANDLE is returned. If pResource was not
registered with usage flags CU_D3D9_REGISTER_FLAGS_ARRAY then CUDA_ERROR_INVALID_HANDLE is
returned. If pResource is not mapped then CUDA_ERROR_NOT_MAPPED is returned.
For usage requirements of Face and Level parameters, see cuD3D9ResourceGetMappedPointer().

Parameters:
pArray - Returned array corresponding to subresource
pResource - Mapped resource to access

Generated for NVIDIA CUDA Library by Doxygen

3.28 Direct3D 9 Interoperability 185

Face - Face of resource to access

Level - Level of resource to access

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_INVALID_HANDLE,
CUDA_ERROR_NOT_MAPPED

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuD3D9CtxCreate, cuD3D9GetDevice, cuD3D9GetDirect3DDevice, cuD3D9MapResources,
cuD3D9RegisterResource, cuD3D9ResourceGetMappedPitch, cuD3D9ResourceGetMappedPointer,
cuD3D9ResourceGetMappedSize, cuD3D9ResourceGetSurfaceDimensions, cuD3D9ResourceSetMapFlags,
cuD3D9UnmapResources, cuD3D9UnregisterResource

3.28.2.7 cuD3D9ResourceGetMappedPitch (unsigned int ∗ pPitch, unsigned int ∗ pPitchSlice,

IDirect3DResource9 ∗ pResource, unsigned int Face, unsigned int Level)

Returns in ∗pPitch and ∗pPitchSlice the pitch and Z-slice pitch of the subresource of the mapped Direct3D
resource pResource, which corresponds to Face and Level. The values set in pPitch and pPitchSlice may
change every time that pResource is mapped.
The pitch and Z-slice pitch values may be used to compute the location of a sample on a surface as follows.
For a 2D surface, the byte offset of the sample at position x, y from the base pointer of the surface is:
y ∗ pitch + (bytes per pixel) ∗ x
For a 3D surface, the byte offset of the sample at position x, y, z from the base pointer of the surface is:
z∗ slicePitch + y ∗ pitch + (bytes per pixel) ∗ x
Both parameters pPitch and pPitchSlice are optional and may be set to NULL.
If pResource is not of type IDirect3DBaseTexture9 or one of its sub-types or if pResource has not been reg-
istered for use with CUDA, then cudaErrorInvalidResourceHandle is returned. If pResource was not registered
with usage flags CU_D3D9_REGISTER_FLAGS_NONE, then CUDA_ERROR_INVALID_HANDLE is returned. If
pResource is not mapped for access by CUDA then CUDA_ERROR_NOT_MAPPED is returned.
For usage requirements of Face and Level parameters, see cuD3D9ResourceGetMappedPointer().

Parameters:
pPitch - Returned pitch of subresource
pPitchSlice - Returned Z-slice pitch of subresource
pResource - Mapped resource to access
Face - Face of resource to access
Level - Level of resource to access

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_INVALID_HANDLE,
CUDA_ERROR_NOT_MAPPED

Generated for NVIDIA CUDA Library by Doxygen

186 Module Documentation

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuD3D9CtxCreate, cuD3D9GetDevice, cuD3D9GetDirect3DDevice, cuD3D9MapResources,
cuD3D9RegisterResource, cuD3D9ResourceGetMappedArray, cuD3D9ResourceGetMappedPointer,
cuD3D9ResourceGetMappedSize, cuD3D9ResourceGetSurfaceDimensions, cuD3D9ResourceSetMapFlags,
cuD3D9UnmapResources, cuD3D9UnregisterResource

3.28.2.8 cuD3D9ResourceGetMappedPointer (CUdeviceptr ∗ pDevPtr, IDirect3DResource9 ∗ pResource,

unsigned int Face, unsigned int Level)

Returns in ∗pDevPtr the base pointer of the subresource of the mapped Direct3D resource pResource, which
corresponds to Face and Level. The value set in pDevPtr may change every time that pResource is mapped.
If pResource is not registered, then CUDA_ERROR_INVALID_HANDLE is returned. If pResource was not
registered with usage flags CU_D3D9_REGISTER_FLAGS_NONE, then CUDA_ERROR_INVALID_HANDLE is
returned. If pResource is not mapped, then CUDA_ERROR_NOT_MAPPED is returned.
If pResource is of type IDirect3DCubeTexture9, then Face must one of the values enumerated by type
D3DCUBEMAP_FACES. For all other types Face must be 0. If Face is invalid, then CUDA_ERROR_INVALID_-
VALUE is returned.
If pResource is of type IDirect3DBaseTexture9, then Level must correspond to a valid mipmap level. At present
only mipmap level 0 is supported. For all other types Level must be 0. If Level is invalid, then CUDA_ERROR_-
INVALID_VALUE is returned.

Parameters:
pDevPtr - Returned pointer corresponding to subresource
pResource - Mapped resource to access
Face - Face of resource to access
Level - Level of resource to access

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_INVALID_HANDLE,
CUDA_ERROR_NOT_MAPPED

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuD3D9CtxCreate, cuD3D9GetDevice, cuD3D9GetDirect3DDevice, cuD3D9MapResources,
cuD3D9RegisterResource, cuD3D9ResourceGetMappedArray, cuD3D9ResourceGetMappedPitch,
cuD3D9ResourceGetMappedSize, cuD3D9ResourceGetSurfaceDimensions, cuD3D9ResourceSetMapFlags,
cuD3D9UnmapResources, cuD3D9UnregisterResource

Generated for NVIDIA CUDA Library by Doxygen

3.28 Direct3D 9 Interoperability 187

3.28.2.9 cuD3D9ResourceGetMappedSize (unsigned int ∗ pSize, IDirect3DResource9 ∗ pResource, unsigned

int Face, unsigned int Level)

Returns in ∗pSize the size of the subresource of the mapped Direct3D resource pResource, which corresponds to
Face and Level. The value set in pSize may change every time that pResource is mapped.
If pResource has not been registered for use with CUDA, then CUDA_ERROR_INVALID_HANDLE is returned.
If pResource was not registered with usage flags CU_D3D9_REGISTER_FLAGS_NONE, then CUDA_ERROR_-
INVALID_HANDLE is returned. If pResource is not mapped for access by CUDA, then CUDA_ERROR_NOT_-
MAPPED is returned.
For usage requirements of Face and Level parameters, see cuD3D9ResourceGetMappedPointer.

Parameters:
pSize - Returned size of subresource
pResource - Mapped resource to access
Face - Face of resource to access
Level - Level of resource to access

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_INVALID_HANDLE,
CUDA_ERROR_NOT_MAPPED

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuD3D9CtxCreate, cuD3D9GetDevice, cuD3D9GetDirect3DDevice, cuD3D9MapResources,
cuD3D9RegisterResource, cuD3D9ResourceGetMappedArray, cuD3D9ResourceGetMappedPitch,
cuD3D9ResourceGetMappedPointer, cuD3D9ResourceGetSurfaceDimensions, cuD3D9ResourceSetMapFlags,
cuD3D9UnmapResources, cuD3D9UnregisterResource

3.28.2.10 cuD3D9ResourceGetSurfaceDimensions (unsigned int ∗ pWidth, unsigned int ∗ pHeight, unsigned

int ∗ pDepth, IDirect3DResource9 ∗ pResource, unsigned int Face, unsigned int Level)

Returns in ∗pWidth, ∗pHeight, and ∗pDepth the dimensions of the subresource of the mapped Direct3D resource
pResource, which corresponds to Face and Level.
Because anti-aliased surfaces may have multiple samples per pixel, it is possible that the dimensions of a resource will
be an integer factor larger than the dimensions reported by the Direct3D runtime.
The parameters pWidth, pHeight, and pDepth are optional. For 2D surfaces, the value returned in ∗pDepth will
be 0.
If pResource is not of type IDirect3DBaseTexture9 or IDirect3DSurface9 or if pResource has not been registered
for use with CUDA, then CUDA_ERROR_INVALID_HANDLE is returned.
For usage requirements of Face and Level parameters, see cuD3D9ResourceGetMappedPointer().

Parameters:
pWidth - Returned width of surface

Generated for NVIDIA CUDA Library by Doxygen

188 Module Documentation

pHeight - Returned height of surface

pDepth - Returned depth of surface
pResource - Registered resource to access
Face - Face of resource to access
Level - Level of resource to access

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_INVALID_HANDLE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuD3D9CtxCreate, cuD3D9GetDevice, cuD3D9GetDirect3DDevice, cuD3D9MapResources,
cuD3D9RegisterResource, cuD3D9ResourceGetMappedArray, cuD3D9ResourceGetMappedPitch,
cuD3D9ResourceGetMappedPointer, cuD3D9ResourceGetMappedSize, cuD3D9ResourceSetMapFlags,
cuD3D9UnmapResources, cuD3D9UnregisterResource

3.28.2.11 cuD3D9ResourceSetMapFlags (IDirect3DResource9 ∗ pResource, unsigned int Flags)

Set Flags for mapping the Direct3D resource pResource.

Changes to Flags will take effect the next time pResource is mapped. The Flags argument may be any of the
following:

• CU_D3D9_MAPRESOURCE_FLAGS_NONE: Specifies no hints about how this resource will be used. It is

therefore assumed that this resource will be read from and written to by CUDA kernels. This is the default value.
• CU_D3D9_MAPRESOURCE_FLAGS_READONLY: Specifies that CUDA kernels which access this resource
will not write to this resource.
• CU_D3D9_MAPRESOURCE_FLAGS_WRITEDISCARD: Specifies that CUDA kernels which access this re-
source will not read from this resource and will write over the entire contents of the resource, so none of the
data previously stored in the resource will be preserved.

If pResource has not been registered for use with CUDA, then CUDA_ERROR_INVALID_HANDLE is returned.
If pResource is presently mapped for access by CUDA, then CUDA_ERROR_ALREADY_MAPPED is returned.

Parameters:
pResource - Registered resource to set flags for
Flags - Parameters for resource mapping

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_INVALID_HANDLE,
CUDA_ERROR_ALREADY_MAPPED

Note:
Note that this function may also return error codes from previous, asynchronous launches.

Generated for NVIDIA CUDA Library by Doxygen

3.28 Direct3D 9 Interoperability 189

See also:
cuD3D9CtxCreate, cuD3D9GetDevice, cuD3D9GetDirect3DDevice, cuD3D9MapResources,
cuD3D9RegisterResource, cuD3D9ResourceGetMappedArray, cuD3D9ResourceGetMappedPitch,
cuD3D9ResourceGetMappedPointer, cuD3D9ResourceGetMappedSize, cuD3D9ResourceGetSurfaceDimensions,
cuD3D9UnmapResources, cuD3D9UnregisterResource

3.28.2.12 cuD3D9UnmapResources (unsigned int count, IDirect3DResource9 ∗∗ ppResource)

Unmaps the count Direct3D resources in ppResource.

This function provides the synchronization guarantee that any CUDA kernels issued before
cuD3D9UnmapResources() will complete before any Direct3D calls issued after cuD3D9UnmapResources()
begin.
If any of ppResource have not been registered for use with CUDA or if ppResource contains any duplicate
entries, then CUDA_ERROR_INVALID_HANDLE is returned. If any of ppResource are not presently mapped
for access by CUDA, then CUDA_ERROR_NOT_MAPPED is returned.

Parameters:
count - Number of resources to unmap for CUDA
ppResource - Resources to unmap for CUDA

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_HANDLE, CUDA_ERROR_NOT_MAPPED,
CUDA_ERROR_UNKNOWN

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuD3D9CtxCreate, cuD3D9GetDevice, cuD3D9GetDirect3DDevice, cuD3D9MapResources,
cuD3D9RegisterResource, cuD3D9ResourceGetMappedArray, cuD3D9ResourceGetMappedPitch,
cuD3D9ResourceGetMappedPointer, cuD3D9ResourceGetMappedSize, cuD3D9ResourceGetSurfaceDimensions,
cuD3D9ResourceSetMapFlags, cuD3D9UnregisterResource

3.28.2.13 cuD3D9UnregisterResource (IDirect3DResource9 ∗ pResource)

Unregisters the Direct3D resource pResource so it is not accessible by CUDA unless registered again.
If pResource is not registered, then CUDA_ERROR_INVALID_HANDLE is returned.

Parameters:
pResource - Resource to unregister

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_HANDLE, CUDA_ERROR_UNKNOWN

Generated for NVIDIA CUDA Library by Doxygen

190 Module Documentation

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuD3D9CtxCreate, cuD3D9GetDevice, cuD3D9GetDirect3DDevice, cuD3D9MapResources,
cuD3D9RegisterResource, cuD3D9ResourceGetMappedArray, cuD3D9ResourceGetMappedPitch,
cuD3D9ResourceGetMappedPointer, cuD3D9ResourceGetMappedSize, cuD3D9ResourceGetSurfaceDimensions,
cuD3D9ResourceSetMapFlags, cuD3D9UnmapResources

Generated for NVIDIA CUDA Library by Doxygen

3.29 Direct3D 10 Interoperability 191

3.29 Direct3D 10 Interoperability

Functions
• CUresult cuD3D10CtxCreate (CUcontext ∗pCtx, CUdevice ∗pCuDevice, unsigned int Flags, ID3D10Device
∗pDxDevice)
Create a CUDA context for interoperability with Direct3D.

• CUresult cuD3D10GetDevice (CUdevice ∗pDevice, IDXGIAdapter ∗pAdapter)

Gets the device number for an adapter.

• CUresult cuD3D10MapResources (unsigned int count, ID3D10Resource ∗∗ppResources)

Map Direct3D resources for access by CUDA.

• CUresult cuD3D10RegisterResource (ID3D10Resource ∗pResource, unsigned int Flags)

Register a Direct3D resource for access by CUDA.

• CUresult cuD3D10ResourceGetMappedArray (CUarray ∗pArray, ID3D10Resource ∗pResource, unsigned int

SubResource)
Get an array through which to access a subresource of a Direct3D resource which has been mapped for access by CUDA.

• CUresult cuD3D10ResourceGetMappedPitch (unsigned int ∗pPitch, unsigned int ∗pPitchSlice,

ID3D10Resource ∗pResource, unsigned int SubResource)
Get the pitch of a subresource of a Direct3D resource which has been mapped for access by CUDA.

• CUresult cuD3D10ResourceGetMappedPointer (CUdeviceptr ∗pDevPtr, ID3D10Resource ∗pResource, un-

signed int SubResource)
Get a pointer through which to access a subresource of a Direct3D resource which has been mapped for access by
CUDA.

• CUresult cuD3D10ResourceGetMappedSize (unsigned int ∗pSize, ID3D10Resource ∗pResource, unsigned int

SubResource)
Get the size of a subresource of a Direct3D resource which has been mapped for access by CUDA.

• CUresult cuD3D10ResourceGetSurfaceDimensions (unsigned int ∗pWidth, unsigned int ∗pHeight, unsigned int
∗pDepth, ID3D10Resource ∗pResource, unsigned int SubResource)
Get the dimensions of a registered surface.

• CUresult cuD3D10ResourceSetMapFlags (ID3D10Resource ∗pResource, unsigned int Flags)

Set usage flags for mapping a Direct3D resource.

• CUresult cuD3D10UnmapResources (unsigned int count, ID3D10Resource ∗∗ppResources)

Unmap Direct3D resources.

• CUresult cuD3D10UnregisterResource (ID3D10Resource ∗pResource)

Unregister a Direct3D resource.

3.29.1 Detailed Description

This section describes the Direct3D 10 interoperability functions of the low-level CUDA driver application program-
ming interface.

Generated for NVIDIA CUDA Library by Doxygen

192 Module Documentation

3.29.2 Function Documentation

3.29.2.1 cuD3D10CtxCreate (CUcontext ∗ pCtx, CUdevice ∗ pCuDevice, unsigned int Flags, ID3D10Device
∗ pDxDevice)

Creates a new CUDA context, enables interoperability for that context with the Direct3D device pDxDevice, and as-
sociates the created CUDA context with the calling thread. The CUcontext will be returned in ∗pCtx. If pCuDevice
is non-NULL, then the CUdevice on which this CUDA context was created will be returned in ∗pCuDevice. For
usage of the Flags parameter, see cuCtxCreate(). Direct3D resources from this device may be registered and mapped
through the lifetime of this CUDA context.
This context will function only until its Direct3D device is destroyed. On success, this call will increase the internal
reference count on pDxDevice. This reference count will be decremented upon destruction of this context through
cuCtxDestroy().

Parameters:
pCtx - Returned newly created CUDA context
pCuDevice - Returned pointer to device on which context was created
Flags - Context creation flags (see cuCtxCreate() for details)
pDxDevice - Direct3D device to create interoperability context with

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_VALUE, CUDA_ERROR_OUT_OF_MEMORY, CUDA_ERROR_UNKNOWN

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuD3D10GetDevice, cuD3D10MapResources, cuD3D10RegisterResource, cuD3D10ResourceGetMappedArray,
cuD3D10ResourceGetMappedPitch, cuD3D10ResourceGetMappedPointer, cuD3D10ResourceGetMappedSize,
cuD3D10ResourceGetSurfaceDimensions, cuD3D10ResourceSetMapFlags, cuD3D10UnmapResources,
cuD3D10UnregisterResource

3.29.2.2 cuD3D10GetDevice (CUdevice ∗ pDevice, IDXGIAdapter ∗ pAdapter)

Returns in ∗pDevice the Cuda-compatible device corresponding to the adapter pAdapter obtained from IDXGI-
Factory::EnumAdapters. This call will succeed only if a device on adapter pAdapter is Cuda-compatible.

Parameters:
pDevice - Returns the device corresponding to pAdapter
pAdapter - D3D10 adapter to get device for

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_VALUE, CUDA_ERROR_UNKNOWN

Note:
Note that this function may also return error codes from previous, asynchronous launches.

Generated for NVIDIA CUDA Library by Doxygen

3.29 Direct3D 10 Interoperability 193

See also:
cuD3D10CtxCreate, cuD3D10MapResources, cuD3D10RegisterResource, cuD3D10ResourceGetMappedArray,
cuD3D10ResourceGetMappedPitch, cuD3D10ResourceGetMappedPointer, cuD3D10ResourceGetMappedSize,
cuD3D10ResourceGetSurfaceDimensions, cuD3D10ResourceSetMapFlags, cuD3D10UnmapResources,
cuD3D10UnregisterResource

3.29.2.3 cuD3D10MapResources (unsigned int count, ID3D10Resource ∗∗ ppResources)

Maps the count Direct3D resources in ppResources for access by CUDA.

The resources in ppResources may be accessed in CUDA kernels until they are unmapped. Direct3D should not
access any resources while they are mapped by CUDA. If an application does so, the results are undefined.
This function provides the synchronization guarantee that any Direct3D calls issued before cuD3D10MapResources()
will complete before any CUDA kernels issued after cuD3D10MapResources() begin.
If any of ppResources have not been registered for use with CUDA or if ppResources contains any duplicate
entries, then CUDA_ERROR_INVALID_HANDLE is returned. If any of ppResources are presently mapped for
access by CUDA, then CUDA_ERROR_ALREADY_MAPPED is returned.

Parameters:
count - Number of resources to map for CUDA
ppResources - Resources to map for CUDA

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_HANDLE, CUDA_ERROR_ALREADY_-
MAPPED, CUDA_ERROR_UNKNOWN

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuD3D10CtxCreate, cuD3D10GetDevice, cuD3D10RegisterResource, cuD3D10ResourceGetMappedArray,
cuD3D10ResourceGetMappedPitch, cuD3D10ResourceGetMappedPointer, cuD3D10ResourceGetMappedSize,
cuD3D10ResourceGetSurfaceDimensions, cuD3D10ResourceSetMapFlags, cuD3D10UnmapResources,
cuD3D10UnregisterResource

3.29.2.4 cuD3D10RegisterResource (ID3D10Resource ∗ pResource, unsigned int Flags)

Registers the Direct3D resource pResource for access by CUDA.

If this call is successful, then the application will be able to map and unmap this resource until it is unregistered through
cuD3D10UnregisterResource(). Also on success, this call will increase the internal reference count on pResource.
This reference count will be decremented when this resource is unregistered through cuD3D10UnregisterResource().
This call is potentially high-overhead and should not be called every frame in interactive applications.
The type of pResource must be one of the following.

• ID3D10Buffer: Cannot be used with Flags set to CU_D3D10_REGISTER_FLAGS_ARRAY.

Generated for NVIDIA CUDA Library by Doxygen

194 Module Documentation

• ID3D10Texture1D: No restrictions.
• ID3D10Texture2D: No restrictions.
• ID3D10Texture3D: No restrictions.

The Flags argument specifies the mechanism through which CUDA will access the Direct3D resource. The following
values are allowed.

• CU_D3D10_REGISTER_FLAGS_NONE: Specifies that CUDA will access this resource through a

CUdeviceptr. The pointer, size, and (for textures), pitch for each subresource of this allocation
may be queried through cuD3D10ResourceGetMappedPointer(), cuD3D10ResourceGetMappedSize(), and
cuD3D10ResourceGetMappedPitch() respectively. This option is valid for all resource types.
• CU_D3D10_REGISTER_FLAGS_ARRAY: Specifies that CUDA will access this resource through a CUarray
queried on a sub-resource basis through cuD3D10ResourceGetMappedArray(). This option is only valid for
resources of type ID3D10Texture1D, ID3D10Texture2D, and ID3D10Texture3D.

Not all Direct3D resources of the above types may be used for interoperability with CUDA. The following are some
limitations.

• The primary rendertarget may not be registered with CUDA.

• Resources allocated as shared may not be registered with CUDA.
• Textures which are not of a format which is 1, 2, or 4 channels of 8, 16, or 32-bit integer or floating-point data
cannot be shared.
• Surfaces of depth or stencil formats cannot be shared.

If Direct3D interoperability is not initialized on this context then CUDA_ERROR_INVALID_CONTEXT is returned.

If pResource is of incorrect type or is already registered, then CUDA_ERROR_INVALID_HANDLE is returned.
If pResource cannot be registered, then CUDA_ERROR_UNKNOWN is returned.

Parameters:
pResource - Resource to register
Flags - Parameters for resource registration

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_INVALID_HANDLE,
CUDA_ERROR_OUT_OF_MEMORY, CUDA_ERROR_UNKNOWN

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuD3D10CtxCreate, cuD3D10GetDevice, cuD3D10MapResources, cuD3D10ResourceGetMappedArray,
cuD3D10ResourceGetMappedPitch, cuD3D10ResourceGetMappedPointer, cuD3D10ResourceGetMappedSize,
cuD3D10ResourceGetSurfaceDimensions, cuD3D10ResourceSetMapFlags, cuD3D10UnmapResources,
cuD3D10UnregisterResource

Generated for NVIDIA CUDA Library by Doxygen

3.29 Direct3D 10 Interoperability 195

3.29.2.5 cuD3D10ResourceGetMappedArray (CUarray ∗ pArray, ID3D10Resource ∗ pResource, unsigned

int SubResource)

Returns in ∗pArray an array through which the subresource of the mapped Direct3D resource pResource, which
corresponds to SubResource may be accessed. The value set in pArray may change every time that pResource
is mapped.
If pResource is not registered, then CUDA_ERROR_INVALID_HANDLE is returned. If pResource was not
registered with usage flags CU_D3D10_REGISTER_FLAGS_ARRAY, then CUDA_ERROR_INVALID_HANDLE
is returned. If pResource is not mapped, then CUDA_ERROR_NOT_MAPPED is returned.
For usage requirements of the SubResource parameter, see cuD3D10ResourceGetMappedPointer().

Parameters:
pArray - Returned array corresponding to subresource
pResource - Mapped resource to access
SubResource - Subresource of pResource to access

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_INVALID_HANDLE,
CUDA_ERROR_NOT_MAPPED

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuD3D10CtxCreate, cuD3D10GetDevice, cuD3D10MapResources, cuD3D10RegisterResource,
cuD3D10ResourceGetMappedPitch, cuD3D10ResourceGetMappedPointer, cuD3D10ResourceGetMappedSize,
cuD3D10ResourceGetSurfaceDimensions, cuD3D10ResourceSetMapFlags, cuD3D10UnmapResources,
cuD3D10UnregisterResource

3.29.2.6 cuD3D10ResourceGetMappedPitch (unsigned int ∗ pPitch, unsigned int ∗ pPitchSlice,

ID3D10Resource ∗ pResource, unsigned int SubResource)

Returns in ∗pPitch and ∗pPitchSlice the pitch and Z-slice pitch of the subresource of the mapped Direct3D
resource pResource, which corresponds to SubResource. The values set in pPitch and pPitchSlice may
change every time that pResource is mapped.
The pitch and Z-slice pitch values may be used to compute the location of a sample on a surface as follows.
For a 2D surface, the byte offset of the sample at position x, y from the base pointer of the surface is:
y ∗ pitch + (bytes per pixel) ∗ x
For a 3D surface, the byte offset of the sample at position x, y, z from the base pointer of the surface is:
z∗ slicePitch + y ∗ pitch + (bytes per pixel) ∗ x
Both parameters pPitch and pPitchSlice are optional and may be set to NULL.
If pResource is not of type IDirect3DBaseTexture10 or one of its sub-types or if pResource has not been regis-
tered for use with CUDA, then CUDA_ERROR_INVALID_HANDLE is returned. If pResource was not registered
with usage flags CU_D3D10_REGISTER_FLAGS_NONE, then CUDA_ERROR_INVALID_HANDLE is returned.
If pResource is not mapped for access by CUDA, then CUDA_ERROR_NOT_MAPPED is returned.

Generated for NVIDIA CUDA Library by Doxygen

196 Module Documentation

For usage requirements of the SubResource parameter, see cuD3D10ResourceGetMappedPointer().

Parameters:
pPitch - Returned pitch of subresource
pPitchSlice - Returned Z-slice pitch of subresource
pResource - Mapped resource to access
SubResource - Subresource of pResource to access

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_INVALID_HANDLE,
CUDA_ERROR_NOT_MAPPED

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuD3D10CtxCreate, cuD3D10GetDevice, cuD3D10MapResources, cuD3D10RegisterResource,
cuD3D10ResourceGetMappedArray, cuD3D10ResourceGetMappedPointer, cuD3D10ResourceGetMappedSize,
cuD3D10ResourceGetSurfaceDimensions, cuD3D10ResourceSetMapFlags, cuD3D10UnmapResources,
cuD3D10UnregisterResource

3.29.2.7 cuD3D10ResourceGetMappedPointer (CUdeviceptr ∗ pDevPtr, ID3D10Resource ∗ pResource,

unsigned int SubResource)

Returns in ∗pDevPtr the base pointer of the subresource of the mapped Direct3D resource pResource, which
corresponds to SubResource. The value set in pDevPtr may change every time that pResource is mapped.
If pResource is not registered, then CUDA_ERROR_INVALID_HANDLE is returned. If pResource was not
registered with usage flags CU_D3D10_REGISTER_FLAGS_NONE, then CUDA_ERROR_INVALID_HANDLE is
returned. If pResource is not mapped, then CUDA_ERROR_NOT_MAPPED is returned.
If pResource is of type ID3D10Buffer, then SubResource must be 0. If pResource is of any other type, then
the value of SubResource must come from the subresource calculation in D3D10CalcSubResource().

Parameters:
pDevPtr - Returned pointer corresponding to subresource
pResource - Mapped resource to access
SubResource - Subresource of pResource to access

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_INVALID_HANDLE,
CUDA_ERROR_NOT_MAPPED

Note:
Note that this function may also return error codes from previous, asynchronous launches.

Generated for NVIDIA CUDA Library by Doxygen

3.29 Direct3D 10 Interoperability 197

See also:
cuD3D10CtxCreate, cuD3D10GetDevice, cuD3D10MapResources, cuD3D10RegisterResource,
cuD3D10ResourceGetMappedArray, cuD3D10ResourceGetMappedPitch, cuD3D10ResourceGetMappedSize,
cuD3D10ResourceGetSurfaceDimensions, cuD3D10ResourceSetMapFlags, cuD3D10UnmapResources,
cuD3D10UnregisterResource

3.29.2.8 cuD3D10ResourceGetMappedSize (unsigned int ∗ pSize, ID3D10Resource ∗ pResource, unsigned

int SubResource)

Returns in ∗pSize the size of the subresource of the mapped Direct3D resource pResource, which corresponds to
SubResource. The value set in pSize may change every time that pResource is mapped.
If pResource has not been registered for use with CUDA, then CUDA_ERROR_INVALID_HANDLE is returned. If
pResource was not registered with usage flags CU_D3D10_REGISTER_FLAGS_NONE, then CUDA_ERROR_-
INVALID_HANDLE is returned. If pResource is not mapped for access by CUDA, then CUDA_ERROR_NOT_-
MAPPED is returned.
For usage requirements of the SubResource parameter, see cuD3D10ResourceGetMappedPointer().

Parameters:
pSize - Returned size of subresource
pResource - Mapped resource to access
SubResource - Subresource of pResource to access

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_INVALID_HANDLE,
CUDA_ERROR_NOT_MAPPED

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuD3D10CtxCreate, cuD3D10GetDevice, cuD3D10MapResources, cuD3D10RegisterResource,
cuD3D10ResourceGetMappedArray, cuD3D10ResourceGetMappedPitch, cuD3D10ResourceGetMappedPointer,
cuD3D10ResourceGetSurfaceDimensions, cuD3D10ResourceSetMapFlags, cuD3D10UnmapResources,
cuD3D10UnregisterResource

3.29.2.9 cuD3D10ResourceGetSurfaceDimensions (unsigned int ∗ pWidth, unsigned int ∗ pHeight, unsigned

int ∗ pDepth, ID3D10Resource ∗ pResource, unsigned int SubResource)

Returns in ∗pWidth, ∗pHeight, and ∗pDepth the dimensions of the subresource of the mapped Direct3D resource
pResource, which corresponds to SubResource.
Because anti-aliased surfaces may have multiple samples per pixel, it is possible that the dimensions of a resource will
be an integer factor larger than the dimensions reported by the Direct3D runtime.
The parameters pWidth, pHeight, and pDepth are optional. For 2D surfaces, the value returned in ∗pDepth will
be 0.

Generated for NVIDIA CUDA Library by Doxygen

198 Module Documentation

If pResource is not of type IDirect3DBaseTexture10 or IDirect3DSurface10 or if pResource has not been regis-
tered for use with CUDA, then CUDA_ERROR_INVALID_HANDLE is returned.
For usage requirements of the SubResource parameter, see cuD3D10ResourceGetMappedPointer().

Parameters:
pWidth - Returned width of surface
pHeight - Returned height of surface
pDepth - Returned depth of surface
pResource - Registered resource to access
SubResource - Subresource of pResource to access

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_INVALID_HANDLE

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuD3D10CtxCreate, cuD3D10GetDevice, cuD3D10MapResources, cuD3D10RegisterResource,
cuD3D10ResourceGetMappedArray, cuD3D10ResourceGetMappedPitch, cuD3D10ResourceGetMappedPointer,
cuD3D10ResourceGetMappedSize, cuD3D10ResourceSetMapFlags, cuD3D10UnmapResources,
cuD3D10UnregisterResource

3.29.2.10 cuD3D10ResourceSetMapFlags (ID3D10Resource ∗ pResource, unsigned int Flags)

Set flags for mapping the Direct3D resource pResource.

Changes to flags will take effect the next time pResource is mapped. The Flags argument may be any of the
following.

• CU_D3D10_MAPRESOURCE_FLAGS_NONE: Specifies no hints about how this resource will be used. It is

therefore assumed that this resource will be read from and written to by CUDA kernels. This is the default value.

• CU_D3D10_MAPRESOURCE_FLAGS_READONLY: Specifies that CUDA kernels which access this re-

source will not write to this resource.

• CU_D3D10_MAPRESOURCE_FLAGS_WRITEDISCARD: Specifies that CUDA kernels which access this

resource will not read from this resource and will write over the entire contents of the resource, so none of the
data previously stored in the resource will be preserved.

If pResource has not been registered for use with CUDA, then CUDA_ERROR_INVALID_HANDLE is returned.
If pResource is presently mapped for access by CUDA then CUDA_ERROR_ALREADY_MAPPED is returned.

Parameters:
pResource - Registered resource to set flags for
Flags - Parameters for resource mapping

Generated for NVIDIA CUDA Library by Doxygen

3.29 Direct3D 10 Interoperability 199

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_INVALID_HANDLE,
CUDA_ERROR_ALREADY_MAPPED

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuD3D10CtxCreate, cuD3D10GetDevice, cuD3D10MapResources, cuD3D10RegisterResource,
cuD3D10ResourceGetMappedArray, cuD3D10ResourceGetMappedPitch, cuD3D10ResourceGetMappedPointer,
cuD3D10ResourceGetMappedSize, cuD3D10ResourceGetSurfaceDimensions, cuD3D10UnmapResources,
cuD3D10UnregisterResource

3.29.2.11 cuD3D10UnmapResources (unsigned int count, ID3D10Resource ∗∗ ppResources)

Unmaps the count Direct3D resources in ppResources.

This function provides the synchronization guarantee that any CUDA kernels issued before
cuD3D10UnmapResources() will complete before any Direct3D calls issued after cuD3D10UnmapResources()
begin.
If any of ppResources have not been registered for use with CUDA or if ppResources contains any duplicate
entries, then CUDA_ERROR_INVALID_HANDLE is returned. If any of ppResources are not presently mapped
for access by CUDA, then CUDA_ERROR_NOT_MAPPED is returned.

Parameters:
count - Number of resources to unmap for CUDA
ppResources - Resources to unmap for CUDA

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_VALUE, CUDA_ERROR_INVALID_HANDLE,
CUDA_ERROR_NOT_MAPPED, CUDA_ERROR_UNKNOWN

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuD3D10CtxCreate, cuD3D10GetDevice, cuD3D10MapResources, cuD3D10RegisterResource,
cuD3D10ResourceGetMappedArray, cuD3D10ResourceGetMappedPitch, cuD3D10ResourceGetMappedPointer,
cuD3D10ResourceGetMappedSize, cuD3D10ResourceGetSurfaceDimensions, cuD3D10ResourceSetMapFlags,
cuD3D10UnregisterResource

3.29.2.12 cuD3D10UnregisterResource (ID3D10Resource ∗ pResource)

Unregisters the Direct3D resource pResource so it is not accessible by CUDA unless registered again.
If pResource is not registered, then CUDA_ERROR_INVALID_HANDLE is returned.

Generated for NVIDIA CUDA Library by Doxygen

200 Module Documentation

Parameters:
pResource - Resources to unregister

Returns:
CUDA_SUCCESS, CUDA_ERROR_DEINITIALIZED, CUDA_ERROR_NOT_INITIALIZED, CUDA_-
ERROR_INVALID_CONTEXT, CUDA_ERROR_INVALID_HANDLE, CUDA_ERROR_UNKNOWN

Note:
Note that this function may also return error codes from previous, asynchronous launches.

See also:
cuD3D10CtxCreate, cuD3D10GetDevice, cuD3D10MapResources, cuD3D10RegisterResource,
cuD3D10ResourceGetMappedArray, cuD3D10ResourceGetMappedPitch, cuD3D10ResourceGetMappedPointer,
cuD3D10ResourceGetMappedSize, cuD3D10ResourceGetSurfaceDimensions, cuD3D10ResourceSetMapFlags,
cuD3D10UnmapResources

Generated for NVIDIA CUDA Library by Doxygen

3.30 Data types used by CUDA driver 201

3.30 Data types used by CUDA driver

Data Structures

• struct CUDA_ARRAY3D_DESCRIPTOR
• struct CUDA_ARRAY_DESCRIPTOR
• struct CUDA_MEMCPY2D_st
• struct CUDA_MEMCPY3D_st
• struct CUdevprop_st

Data types used by CUDA driver

Data types used by CUDA driver

Author:
NVIDIA Corporation

• enum CUaddress_mode_enum {
CU_TR_ADDRESS_MODE_WRAP,
CU_TR_ADDRESS_MODE_CLAMP,
CU_TR_ADDRESS_MODE_MIRROR }
• enum CUarray_format_enum {
CU_AD_FORMAT_UNSIGNED_INT8,
CU_AD_FORMAT_UNSIGNED_INT16,
CU_AD_FORMAT_UNSIGNED_INT32,
CU_AD_FORMAT_SIGNED_INT8,
CU_AD_FORMAT_SIGNED_INT16,
CU_AD_FORMAT_SIGNED_INT32,
CU_AD_FORMAT_HALF,
CU_AD_FORMAT_FLOAT }
• enum CUcomputemode_enum {
CU_COMPUTEMODE_DEFAULT,
CU_COMPUTEMODE_EXCLUSIVE,
CU_COMPUTEMODE_PROHIBITED }
• enum CUctx_flags_enum {
CU_CTX_SCHED_AUTO,
CU_CTX_SCHED_SPIN,
CU_CTX_SCHED_YIELD ,
CU_CTX_BLOCKING_SYNC,
CU_CTX_MAP_HOST,
CU_CTX_LMEM_RESIZE_TO_MAX }

Generated for NVIDIA CUDA Library by Doxygen

202 Module Documentation

• enum cudaError_enum {
CUDA_SUCCESS,
CUDA_ERROR_INVALID_VALUE,
CUDA_ERROR_OUT_OF_MEMORY,
CUDA_ERROR_NOT_INITIALIZED,
CUDA_ERROR_DEINITIALIZED,
CUDA_ERROR_NO_DEVICE,
CUDA_ERROR_INVALID_DEVICE,
CUDA_ERROR_INVALID_IMAGE,
CUDA_ERROR_INVALID_CONTEXT,
CUDA_ERROR_CONTEXT_ALREADY_CURRENT,
CUDA_ERROR_MAP_FAILED,
CUDA_ERROR_UNMAP_FAILED,
CUDA_ERROR_ARRAY_IS_MAPPED,
CUDA_ERROR_ALREADY_MAPPED,
CUDA_ERROR_NO_BINARY_FOR_GPU,
CUDA_ERROR_ALREADY_ACQUIRED,
CUDA_ERROR_NOT_MAPPED,
CUDA_ERROR_INVALID_SOURCE,
CUDA_ERROR_FILE_NOT_FOUND,
CUDA_ERROR_INVALID_HANDLE,
CUDA_ERROR_NOT_FOUND,
CUDA_ERROR_NOT_READY,
CUDA_ERROR_LAUNCH_FAILED,
CUDA_ERROR_LAUNCH_OUT_OF_RESOURCES,
CUDA_ERROR_LAUNCH_TIMEOUT,
CUDA_ERROR_LAUNCH_INCOMPATIBLE_TEXTURING,
CUDA_ERROR_UNKNOWN }
• enum CUdevice_attribute_enum {
CU_DEVICE_ATTRIBUTE_MAX_THREADS_PER_BLOCK,
CU_DEVICE_ATTRIBUTE_MAX_BLOCK_DIM_X,
CU_DEVICE_ATTRIBUTE_MAX_BLOCK_DIM_Y,
CU_DEVICE_ATTRIBUTE_MAX_BLOCK_DIM_Z,
CU_DEVICE_ATTRIBUTE_MAX_GRID_DIM_X,
CU_DEVICE_ATTRIBUTE_MAX_GRID_DIM_Y,
CU_DEVICE_ATTRIBUTE_MAX_GRID_DIM_Z,
CU_DEVICE_ATTRIBUTE_MAX_SHARED_MEMORY_PER_BLOCK,
CU_DEVICE_ATTRIBUTE_SHARED_MEMORY_PER_BLOCK,
CU_DEVICE_ATTRIBUTE_TOTAL_CONSTANT_MEMORY,
CU_DEVICE_ATTRIBUTE_WARP_SIZE,
CU_DEVICE_ATTRIBUTE_MAX_PITCH,

Generated for NVIDIA CUDA Library by Doxygen

3.30 Data types used by CUDA driver 203

CU_DEVICE_ATTRIBUTE_MAX_REGISTERS_PER_BLOCK,
CU_DEVICE_ATTRIBUTE_REGISTERS_PER_BLOCK,
CU_DEVICE_ATTRIBUTE_CLOCK_RATE,
CU_DEVICE_ATTRIBUTE_TEXTURE_ALIGNMENT,
CU_DEVICE_ATTRIBUTE_GPU_OVERLAP,
CU_DEVICE_ATTRIBUTE_MULTIPROCESSOR_COUNT,
CU_DEVICE_ATTRIBUTE_KERNEL_EXEC_TIMEOUT,
CU_DEVICE_ATTRIBUTE_INTEGRATED,
CU_DEVICE_ATTRIBUTE_CAN_MAP_HOST_MEMORY,
CU_DEVICE_ATTRIBUTE_COMPUTE_MODE }
• enum CUevent_flags_enum {
CU_EVENT_DEFAULT,
CU_EVENT_BLOCKING_SYNC }
• enum CUfilter_mode_enum {
CU_TR_FILTER_MODE_POINT,
CU_TR_FILTER_MODE_LINEAR }
• enum CUfunction_attribute_enum {
CU_FUNC_ATTRIBUTE_MAX_THREADS_PER_BLOCK,
CU_FUNC_ATTRIBUTE_SHARED_SIZE_BYTES,
CU_FUNC_ATTRIBUTE_CONST_SIZE_BYTES,
CU_FUNC_ATTRIBUTE_LOCAL_SIZE_BYTES,
CU_FUNC_ATTRIBUTE_NUM_REGS }
• enum CUjit_fallback_enum {
CU_PREFER_PTX,
CU_PREFER_BINARY }
• enum CUjit_option_enum {
CU_JIT_MAX_REGISTERS,
CU_JIT_THREADS_PER_BLOCK,
CU_JIT_WALL_TIME,
CU_JIT_INFO_LOG_BUFFER,
CU_JIT_INFO_LOG_BUFFER_SIZE_BYTES,
CU_JIT_ERROR_LOG_BUFFER,
CU_JIT_ERROR_LOG_BUFFER_SIZE_BYTES,
CU_JIT_OPTIMIZATION_LEVEL,
CU_JIT_TARGET_FROM_CUCONTEXT,
CU_JIT_TARGET,
CU_JIT_FALLBACK_STRATEGY }
• enum CUjit_target_enum {
CU_TARGET_COMPUTE_10,
CU_TARGET_COMPUTE_11,
CU_TARGET_COMPUTE_12,
CU_TARGET_COMPUTE_13 }

Generated for NVIDIA CUDA Library by Doxygen

204 Module Documentation

• enum CUmemorytype_enum {
CU_MEMORYTYPE_HOST,
CU_MEMORYTYPE_DEVICE,
CU_MEMORYTYPE_ARRAY }
• typedef enum CUaddress_mode_enum CUaddress_mode
• typedef struct CUarray_st ∗ CUarray
CUDA array.

• typedef enum CUarray_format_enum CUarray_format

• typedef enum CUcomputemode_enum CUcomputemode
• typedef struct CUctx_st ∗ CUcontext
CUDA context.

• typedef enum CUctx_flags_enum CUctx_flags

• typedef struct CUDA_MEMCPY2D_st CUDA_MEMCPY2D
• typedef struct CUDA_MEMCPY3D_st CUDA_MEMCPY3D
• typedef int CUdevice
CUDA device.

• typedef enum CUdevice_attribute_enum CUdevice_attribute

• typedef unsigned int CUdeviceptr
CUDA device pointer.

• typedef struct CUdevprop_st CUdevprop

• typedef struct CUevent_st ∗ CUevent
CUDA event.

• typedef enum CUevent_flags_enum CUevent_flags

• typedef enum CUfilter_mode_enum CUfilter_mode
• typedef struct CUfunc_st ∗ CUfunction
CUDA function.

• typedef enum CUfunction_attribute_enum CUfunction_attribute

• typedef enum CUjit_fallback_enum CUjit_fallback
• typedef enum CUjit_option_enum CUjit_option
• typedef enum CUjit_target_enum CUjit_target
• typedef enum CUmemorytype_enum CUmemorytype
• typedef struct CUmod_st ∗ CUmodule
CUDA module.

• typedef enum cudaError_enum CUresult

• typedef struct CUstream_st ∗ CUstream
CUDA stream.

• typedef struct CUtexref_st ∗ CUtexref

CUDA texture reference.

• #define CU_MEMHOSTALLOC_DEVICEMAP
• #define CU_MEMHOSTALLOC_PORTABLE
• #define CU_MEMHOSTALLOC_WRITECOMBINED

Generated for NVIDIA CUDA Library by Doxygen

3.30 Data types used by CUDA driver 205

• #define CU_PARAM_TR_DEFAULT
• #define CU_TRSA_OVERRIDE_FORMAT
• #define CU_TRSF_NORMALIZED_COORDINATES
• #define CU_TRSF_READ_AS_INTEGER
• #define CUDA_VERSION

3.30.1 Define Documentation

3.30.1.1 #define CU_MEMHOSTALLOC_DEVICEMAP

If set, host memory is mapped into CUDA address space and cuMemHostGetDevicePointer() may be called on the
host pointer. Flag for cuMemHostAlloc()

3.30.1.2 #define CU_MEMHOSTALLOC_PORTABLE

If set, host memory is portable between CUDA contexts. Flag for cuMemHostAlloc()

3.30.1.3 #define CU_MEMHOSTALLOC_WRITECOMBINED

If set, host memory is allocated as write-combined - fast to write, faster to DMA, slow to read except via SSE4
streaming load instruction (MOVNTDQA). Flag for cuMemHostAlloc()

3.30.1.4 #define CU_PARAM_TR_DEFAULT

For texture references loaded into the module, use default texunit from texture reference.

3.30.1.5 #define CU_TRSA_OVERRIDE_FORMAT

Override the texref format with a format inferred from the array. Flag for cuTexRefSetArray()

3.30.1.6 #define CU_TRSF_NORMALIZED_COORDINATES

Use normalized texture coordinates in the range [0,1) instead of [0,dim). Flag for cuTexRefSetFlags()

3.30.1.7 #define CU_TRSF_READ_AS_INTEGER

Read the texture as integers rather than promoting the values to floats in the range [0,1]. Flag for cuTexRefSetFlags()

3.30.1.8 #define CUDA_VERSION

CUDA API version number

3.30.2 Typedef Documentation

3.30.2.1 typedef enum CUaddress_mode_enum CUaddress_mode

Texture reference addressing modes

Generated for NVIDIA CUDA Library by Doxygen

206 Module Documentation

3.30.2.2 typedef enum CUarray_format_enum CUarray_format

Array formats

3.30.2.3 typedef enum CUcomputemode_enum CUcomputemode

Compute Modes

3.30.2.4 typedef enum CUctx_flags_enum CUctx_flags

Context creation flags

3.30.2.5 typedef struct CUDA_MEMCPY2D_st CUDA_MEMCPY2D

2D memory copy parameters

3.30.2.6 typedef struct CUDA_MEMCPY3D_st CUDA_MEMCPY3D

3D memory copy parameters

3.30.2.7 typedef enum CUdevice_attribute_enum CUdevice_attribute

Device properties

3.30.2.8 typedef struct CUdevprop_st CUdevprop

Legacy device properties

3.30.2.9 typedef enum CUevent_flags_enum CUevent_flags

Event creation flags

3.30.2.10 typedef enum CUfilter_mode_enum CUfilter_mode

Texture reference filtering modes

3.30.2.11 typedef enum CUfunction_attribute_enum CUfunction_attribute

Function properties

3.30.2.12 typedef enum CUjit_fallback_enum CUjit_fallback

Cubin matching fallback strategies

3.30.2.13 typedef enum CUjit_option_enum CUjit_option

Online compiler options

Generated for NVIDIA CUDA Library by Doxygen

3.30 Data types used by CUDA driver 207

3.30.2.14 typedef enum CUjit_target_enum CUjit_target

Online compilation targets

3.30.2.15 typedef enum CUmemorytype_enum CUmemorytype

Memory types

3.30.2.16 typedef enum cudaError_enum CUresult

Error codes

3.30.3 Enumeration Type Documentation

3.30.3.1 enum CUaddress_mode_enum

Texture reference addressing modes

Enumerator:
CU_TR_ADDRESS_MODE_WRAP Wrapping address mode.
CU_TR_ADDRESS_MODE_CLAMP Clamp to edge address mode.
CU_TR_ADDRESS_MODE_MIRROR Mirror address mode.

3.30.3.2 enum CUarray_format_enum

Array formats

Enumerator:
CU_AD_FORMAT_UNSIGNED_INT8 Unsigned 8-bit integers.
CU_AD_FORMAT_UNSIGNED_INT16 Unsigned 16-bit integers.
CU_AD_FORMAT_UNSIGNED_INT32 Unsigned 32-bit integers.
CU_AD_FORMAT_SIGNED_INT8 Signed 8-bit integers.
CU_AD_FORMAT_SIGNED_INT16 Signed 16-bit integers.
CU_AD_FORMAT_SIGNED_INT32 Signed 32-bit integers.
CU_AD_FORMAT_HALF 16-bit floating point
CU_AD_FORMAT_FLOAT 32-bit floating point

3.30.3.3 enum CUcomputemode_enum

Compute Modes

Enumerator:
CU_COMPUTEMODE_DEFAULT Default compute mode (Multiple contexts allowed per device).
CU_COMPUTEMODE_EXCLUSIVE Compute-exclusive mode (Only one context can be present on this de-
vice at a time).
CU_COMPUTEMODE_PROHIBITED Compute-prohibited mode (No contexts can be created on this device
at this time).

Generated for NVIDIA CUDA Library by Doxygen

208 Module Documentation

3.30.3.4 enum CUctx_flags_enum

Context creation flags

Enumerator:
CU_CTX_SCHED_AUTO Automatic scheduling.
CU_CTX_SCHED_SPIN Set spin as default scheduling.
CU_CTX_SCHED_YIELD Set yield as default scheduling.
CU_CTX_BLOCKING_SYNC Use blocking synchronization.
CU_CTX_MAP_HOST Support mapped pinned allocations.
CU_CTX_LMEM_RESIZE_TO_MAX Keep local memory allocation after launch.

3.30.3.5 enum cudaError_enum

Error codes
Enumerator:
CUDA_SUCCESS No errors.
CUDA_ERROR_INVALID_VALUE Invalid value.
CUDA_ERROR_OUT_OF_MEMORY Out of memory.
CUDA_ERROR_NOT_INITIALIZED Driver not initialized.
CUDA_ERROR_DEINITIALIZED Driver deinitialized.
CUDA_ERROR_NO_DEVICE No CUDA-capable device available.
CUDA_ERROR_INVALID_DEVICE Invalid device.
CUDA_ERROR_INVALID_IMAGE Invalid kernel image.
CUDA_ERROR_INVALID_CONTEXT Invalid context.
CUDA_ERROR_CONTEXT_ALREADY_CURRENT Context already current.
CUDA_ERROR_MAP_FAILED Map failed.
CUDA_ERROR_UNMAP_FAILED Unmap failed.
CUDA_ERROR_ARRAY_IS_MAPPED Array is mapped.
CUDA_ERROR_ALREADY_MAPPED Already mapped.
CUDA_ERROR_NO_BINARY_FOR_GPU No binary for GPU.
CUDA_ERROR_ALREADY_ACQUIRED Already acquired.
CUDA_ERROR_NOT_MAPPED Not mapped.
CUDA_ERROR_INVALID_SOURCE Invalid source.
CUDA_ERROR_FILE_NOT_FOUND File not found.
CUDA_ERROR_INVALID_HANDLE Invalid handle.
CUDA_ERROR_NOT_FOUND Not found.
CUDA_ERROR_NOT_READY CUDA not ready.
CUDA_ERROR_LAUNCH_FAILED Launch failed.
CUDA_ERROR_LAUNCH_OUT_OF_RESOURCES Launch exceeded resources.
CUDA_ERROR_LAUNCH_TIMEOUT Launch exceeded timeout.
CUDA_ERROR_LAUNCH_INCOMPATIBLE_TEXTURING Launch with incompatible texturing.
CUDA_ERROR_UNKNOWN Unknown error.

Generated for NVIDIA CUDA Library by Doxygen

3.30 Data types used by CUDA driver 209

3.30.3.6 enum CUdevice_attribute_enum

Device properties

Enumerator:
CU_DEVICE_ATTRIBUTE_MAX_THREADS_PER_BLOCK Maximum number of threads per block.
CU_DEVICE_ATTRIBUTE_MAX_BLOCK_DIM_X Maximum block dimension X.
CU_DEVICE_ATTRIBUTE_MAX_BLOCK_DIM_Y Maximum block dimension Y.
CU_DEVICE_ATTRIBUTE_MAX_BLOCK_DIM_Z Maximum block dimension Z.
CU_DEVICE_ATTRIBUTE_MAX_GRID_DIM_X Maximum grid dimension X.
CU_DEVICE_ATTRIBUTE_MAX_GRID_DIM_Y Maximum grid dimension Y.
CU_DEVICE_ATTRIBUTE_MAX_GRID_DIM_Z Maximum grid dimension Z.
CU_DEVICE_ATTRIBUTE_MAX_SHARED_MEMORY_PER_BLOCK Maximum shared memory avail-
able per block in bytes.
CU_DEVICE_ATTRIBUTE_SHARED_MEMORY_PER_BLOCK Deprecated, use CU_DEVICE_-
ATTRIBUTE_MAX_SHARED_MEMORY_PER_BLOCK.
CU_DEVICE_ATTRIBUTE_TOTAL_CONSTANT_MEMORY Memory available on device for __constant_-
_ variables in a CUDA C kernel in bytes.
CU_DEVICE_ATTRIBUTE_WARP_SIZE Warp size in threads.
CU_DEVICE_ATTRIBUTE_MAX_PITCH Maximum pitch in bytes allowed by memory copies.
CU_DEVICE_ATTRIBUTE_MAX_REGISTERS_PER_BLOCK Maximum number of 32-bit registers avail-
able per block.
CU_DEVICE_ATTRIBUTE_REGISTERS_PER_BLOCK Deprecated, use CU_DEVICE_ATTRIBUTE_-
MAX_REGISTERS_PER_BLOCK.
CU_DEVICE_ATTRIBUTE_CLOCK_RATE Peak clock frequency in kilohertz.
CU_DEVICE_ATTRIBUTE_TEXTURE_ALIGNMENT Alignment requirement for textures.
CU_DEVICE_ATTRIBUTE_GPU_OVERLAP Device can possibly copy memory and execute a kernel con-
currently.
CU_DEVICE_ATTRIBUTE_MULTIPROCESSOR_COUNT Number of multiprocessors on device.
CU_DEVICE_ATTRIBUTE_KERNEL_EXEC_TIMEOUT Specifies whether there is a run time limit on ker-
nels.
CU_DEVICE_ATTRIBUTE_INTEGRATED Device is integrated with host memory.
CU_DEVICE_ATTRIBUTE_CAN_MAP_HOST_MEMORY Device can map host memory into CUDA ad-
dress space.
CU_DEVICE_ATTRIBUTE_COMPUTE_MODE Compute mode (See CUcomputemode for details).

3.30.3.7 enum CUevent_flags_enum

Event creation flags

Enumerator:
CU_EVENT_DEFAULT Default event flag.
CU_EVENT_BLOCKING_SYNC Event uses blocking synchronization.

Generated for NVIDIA CUDA Library by Doxygen

210 Module Documentation

3.30.3.8 enum CUfilter_mode_enum

Texture reference filtering modes

Enumerator:
CU_TR_FILTER_MODE_POINT Point filter mode.
CU_TR_FILTER_MODE_LINEAR Linear filter mode.

3.30.3.9 enum CUfunction_attribute_enum

Function properties

Enumerator:
CU_FUNC_ATTRIBUTE_MAX_THREADS_PER_BLOCK The number of threads beyond which a launch of
the function would fail. This number depends on both the function and the device on which the function is
currently loaded.
CU_FUNC_ATTRIBUTE_SHARED_SIZE_BYTES The size in bytes of statically-allocated shared memory
required by this function. This does not include dynamically-allocated shared memory requested by the user
at runtime.
CU_FUNC_ATTRIBUTE_CONST_SIZE_BYTES The size in bytes of user-allocated constant memory re-
quired by this function.
CU_FUNC_ATTRIBUTE_LOCAL_SIZE_BYTES The size in bytes of thread local memory used by this func-
tion.
CU_FUNC_ATTRIBUTE_NUM_REGS The number of registers used by each thread of this function.

3.30.3.10 enum CUjit_fallback_enum

Cubin matching fallback strategies

Enumerator:
CU_PREFER_PTX Prefer to compile ptx
CU_PREFER_BINARY Prefer to fall back to compatible binary code

3.30.3.11 enum CUjit_option_enum

Online compiler options

Enumerator:
CU_JIT_MAX_REGISTERS Max number of registers that a thread may use.
CU_JIT_THREADS_PER_BLOCK IN: Specifies minimum number of threads per block to target compilation
for
OUT: Returns the number of threads the compiler actually targeted. This restricts the resource utilization
fo the compiler (e.g. max registers) such that a block with the given number of threads should be able to
launch based on register limitations. Note, this option does not currently take into account any other resource
limitations, such as shared memory utilization.

Generated for NVIDIA CUDA Library by Doxygen

3.30 Data types used by CUDA driver 211

CU_JIT_WALL_TIME Returns a float value in the option of the wall clock time, in milliseconds, spent creating
the cubin
CU_JIT_INFO_LOG_BUFFER Pointer to a buffer in which to print any log messsages from PTXAS that are
informational in nature
CU_JIT_INFO_LOG_BUFFER_SIZE_BYTES IN: Log buffer size in bytes. Log messages will be capped at
this size (including null terminator)
OUT: Amount of log buffer filled with messages
CU_JIT_ERROR_LOG_BUFFER Pointer to a buffer in which to print any log messages from PTXAS that
reflect errors
CU_JIT_ERROR_LOG_BUFFER_SIZE_BYTES IN: Log buffer size in bytes. Log messages will be capped
at this size (including null terminator)
OUT: Amount of log buffer filled with messages
CU_JIT_OPTIMIZATION_LEVEL Level of optimizations to apply to generated code (0 - 4), with 4 being the
default and highest level of optimizations.
CU_JIT_TARGET_FROM_CUCONTEXT No option value required. Determines the target based on the cur-
rent attached context (default)
CU_JIT_TARGET Target is chosen based on supplied CUjit_target_enum.
CU_JIT_FALLBACK_STRATEGY Specifies choice of fallback strategy if matching cubin is not found. Choice
is based on supplied CUjit_fallback_enum.

3.30.3.12 enum CUjit_target_enum

Online compilation targets

Enumerator:
CU_TARGET_COMPUTE_10 Compute device class 1.0.
CU_TARGET_COMPUTE_11 Compute device class 1.1.
CU_TARGET_COMPUTE_12 Compute device class 1.2.
CU_TARGET_COMPUTE_13 Compute device class 1.3.

3.30.3.13 enum CUmemorytype_enum

Memory types

Enumerator:
CU_MEMORYTYPE_HOST Host memory.
CU_MEMORYTYPE_DEVICE Device memory.
CU_MEMORYTYPE_ARRAY Array memory.

Generated for NVIDIA CUDA Library by Doxygen

212 Module Documentation

Generated for NVIDIA CUDA Library by Doxygen

Chapter 4

Data Structure Documentation

4.1 CUDA_ARRAY3D_DESCRIPTOR Struct Reference

Data Fields
• unsigned int Depth
Depth of 3D array.

• unsigned int Flags

Flags.

• CUarray_format Format
Array format.

• unsigned int Height

Height of 3D array.

• unsigned int NumChannels

Channels per array element.

• unsigned int Width

Width of 3D array.

4.1.1 Detailed Description

3D array descriptor
214 Data Structure Documentation

4.2 CUDA_ARRAY_DESCRIPTOR Struct Reference

Data Fields
• CUarray_format Format
Array format.

• unsigned int Height

Height of array.

• unsigned int NumChannels

Channels per array element.

• unsigned int Width

Width of array.

4.2.1 Detailed Description

Array descriptor

Generated for NVIDIA CUDA Library by Doxygen

4.3 CUDA_MEMCPY2D_st Struct Reference 215

4.3 CUDA_MEMCPY2D_st Struct Reference

Data Fields
• CUarray dstArray
Destination array reference.

• CUdeviceptr dstDevice
Destination device pointer.

• void ∗ dstHost
Destination host pointer.

• CUmemorytype dstMemoryType
Destination memory type (host, device, array).

• unsigned int dstPitch

Destination pitch (ignored when dst is array).

• unsigned int dstXInBytes

Destination X in bytes.

• unsigned int dstY

Destination Y.

• unsigned int Height

Height of 2D memory copy.

• CUarray srcArray
Source array reference.

• CUdeviceptr srcDevice
Source device pointer.

• const void ∗ srcHost

Source host pointer.

• CUmemorytype srcMemoryType
Source memory type (host, device, array).

• unsigned int srcPitch

Source pitch (ignored when src is array).

• unsigned int srcXInBytes

Source X in bytes.

• unsigned int srcY

Source Y.

• unsigned int WidthInBytes

Width of 2D memory copy in bytes.

Generated for NVIDIA CUDA Library by Doxygen

216 Data Structure Documentation

4.3.1 Detailed Description

2D memory copy parameters

Generated for NVIDIA CUDA Library by Doxygen

4.4 CUDA_MEMCPY3D_st Struct Reference 217

4.4 CUDA_MEMCPY3D_st Struct Reference

Data Fields
• unsigned int Depth
Depth of 3D memory copy.

• CUarray dstArray
Destination array reference.

• CUdeviceptr dstDevice
Destination device pointer.

• unsigned int dstHeight

Destination height (ignored when dst is array; may be 0 if Depth==1).

• void ∗ dstHost
Destination host pointer.

• unsigned int dstLOD

Destination LOD.

• CUmemorytype dstMemoryType
Destination memory type (host, device, array).

• unsigned int dstPitch

Destination pitch (ignored when dst is array).

• unsigned int dstXInBytes

Destination X in bytes.

• unsigned int dstY

Destination Y.

• unsigned int dstZ

Destination Z.

• unsigned int Height

Height of 3D memory copy.

• void ∗ reserved0
Must be NULL.

• void ∗ reserved1
Must be NULL.

• CUarray srcArray
Source array reference.

• CUdeviceptr srcDevice

Generated for NVIDIA CUDA Library by Doxygen

218 Data Structure Documentation

Source device pointer.

• unsigned int srcHeight

Source height (ignored when src is array; may be 0 if Depth==1).

• const void ∗ srcHost

Source host pointer.

• unsigned int srcLOD

Source LOD.

• CUmemorytype srcMemoryType
Source memory type (host, device, array).

• unsigned int srcPitch

Source pitch (ignored when src is array).

• unsigned int srcXInBytes

Source X in bytes.

• unsigned int srcY

Source Y.

• unsigned int srcZ

Source Z.

• unsigned int WidthInBytes

Width of 3D memory copy in bytes.

4.4.1 Detailed Description

3D memory copy parameters

Generated for NVIDIA CUDA Library by Doxygen

4.5 cudaChannelFormatDesc Struct Reference 219

4.5 cudaChannelFormatDesc Struct Reference

Data Fields
• enum cudaChannelFormatKind f
Channel format kind.

• int w
w

• int x
x

• int y
y

• int z
z

4.5.1 Detailed Description

CUDA Channel format descriptor

Generated for NVIDIA CUDA Library by Doxygen

220 Data Structure Documentation

4.6 cudaDeviceProp Struct Reference

Data Fields
• int canMapHostMemory
Device can map host memory with cudaHostAlloc/cudaHostGetDevicePointer.

• int clockRate
Clock frequency in kilohertz.

• int computeMode
Compute mode (See cudaComputeMode).

• int deviceOverlap
Device can concurrently copy memory and execute a kernel.

• int integrated
Device is integrated as opposed to discrete.

• int kernelExecTimeoutEnabled
Specified whether there is a run time limit on kernels.

• int major
Major compute capability.

• int maxGridSize [3]

Maximum size of each dimension of a grid.

• int maxThreadsDim [3]

Maximum size of each dimension of a block.

• int maxThreadsPerBlock
Maximum number of threads per block.

• size_t memPitch
Maximum pitch in bytes allowed by memory copies.

• int minor
Minor compute capability.

• int multiProcessorCount
Number of multiprocessors on device.

• char name [256]

ASCII string identifying device.

• int regsPerBlock
32-bit registers available per block

• size_t sharedMemPerBlock

Generated for NVIDIA CUDA Library by Doxygen

4.6 cudaDeviceProp Struct Reference 221

Shared memory available per block in bytes.

• size_t textureAlignment
Alignment requirement for textures.

• size_t totalConstMem
Constant memory available on device in bytes.

• size_t totalGlobalMem
Global memory available on device in bytes.

• int warpSize
Warp size in threads.

4.6.1 Detailed Description

CUDA device properties

Generated for NVIDIA CUDA Library by Doxygen

222 Data Structure Documentation

4.7 cudaExtent Struct Reference

Data Fields
• size_t depth
Depth in bytes.

• size_t height
Height in bytes.

• size_t width
Width in bytes.

4.7.1 Detailed Description

CUDA extent

Generated for NVIDIA CUDA Library by Doxygen

4.8 cudaFuncAttributes Struct Reference 223

4.8 cudaFuncAttributes Struct Reference

Data Fields
• size_t constSizeBytes
Size of constant memory in bytes.

• size_t localSizeBytes
Size of local memory in bytes.

• int maxThreadsPerBlock
Maximum number of threads per block.

• int numRegs
Number of registers used.

• size_t sharedSizeBytes
Size of shared memory in bytes.

4.8.1 Detailed Description

CUDA function attributes

Generated for NVIDIA CUDA Library by Doxygen

224 Data Structure Documentation

4.9 cudaMemcpy3DParms Struct Reference

Data Fields
• struct cudaArray ∗ dstArray
Destination memory address.

• struct cudaPos dstPos

Destination position offset.

• struct cudaPitchedPtr dstPtr

Pitched destination memory address.

• struct cudaExtent extent

Requested memory copy size.

• enum cudaMemcpyKind kind

Type of transfer.

• struct cudaArray ∗ srcArray

Source memory address.

• struct cudaPos srcPos

Source position offset.

• struct cudaPitchedPtr srcPtr

Pitched source memory address.

4.9.1 Detailed Description

CUDA 3D memory copying parameters

Generated for NVIDIA CUDA Library by Doxygen

4.10 cudaPitchedPtr Struct Reference 225

4.10 cudaPitchedPtr Struct Reference

Data Fields
• size_t pitch
Pitch of allocated memory in bytes.

• void ∗ ptr
Pointer to allocated memory.

• size_t xsize
Logical width of allocation in elements.

• size_t ysize
Logical height of allocation in elements.

4.10.1 Detailed Description

CUDA Pitched memory pointer

Generated for NVIDIA CUDA Library by Doxygen

226 Data Structure Documentation

4.11 cudaPos Struct Reference

Data Fields
• size_t x
x

• size_t y
y

• size_t z
z

4.11.1 Detailed Description

CUDA 3D position

Generated for NVIDIA CUDA Library by Doxygen

4.12 CUdevprop_st Struct Reference 227

4.12 CUdevprop_st Struct Reference

Data Fields
• int clockRate
Clock frequency in kilohertz.

• int maxGridSize [3]

Maximum size of each dimension of a grid.

• int maxThreadsDim [3]

Maximum size of each dimension of a block.

• int maxThreadsPerBlock
Maximum number of threads per block.

• int memPitch
Maximum pitch in bytes allowed by memory copies.

• int regsPerBlock
32-bit registers available per block

• int sharedMemPerBlock
Shared memory available per block in bytes.

• int SIMDWidth
Warp size in threads.

• int textureAlign
Alignment requirement for textures.

• int totalConstantMemory
Constant memory available on device in bytes.

4.12.1 Detailed Description

Legacy device properties

Generated for NVIDIA CUDA Library by Doxygen

Index

C++ API Routines, 84 CU_DEVICE_ATTRIBUTE_INTEGRATED

Context Management, 106 CUDA_TYPES, 209
CU_AD_FORMAT_FLOAT CU_DEVICE_ATTRIBUTE_KERNEL_EXEC_-
CUDA_TYPES, 207 TIMEOUT
CU_AD_FORMAT_HALF CUDA_TYPES, 209
CUDA_TYPES, 207 CU_DEVICE_ATTRIBUTE_MAX_BLOCK_DIM_X
CU_AD_FORMAT_SIGNED_INT16 CUDA_TYPES, 209
CUDA_TYPES, 207 CU_DEVICE_ATTRIBUTE_MAX_BLOCK_DIM_Y
CU_AD_FORMAT_SIGNED_INT32 CUDA_TYPES, 209
CUDA_TYPES, 207 CU_DEVICE_ATTRIBUTE_MAX_BLOCK_DIM_Z
CU_AD_FORMAT_SIGNED_INT8 CUDA_TYPES, 209
CUDA_TYPES, 207 CU_DEVICE_ATTRIBUTE_MAX_GRID_DIM_X
CU_AD_FORMAT_UNSIGNED_INT16 CUDA_TYPES, 209
CUDA_TYPES, 207 CU_DEVICE_ATTRIBUTE_MAX_GRID_DIM_Y
CU_AD_FORMAT_UNSIGNED_INT32 CUDA_TYPES, 209
CUDA_TYPES, 207 CU_DEVICE_ATTRIBUTE_MAX_GRID_DIM_Z
CU_AD_FORMAT_UNSIGNED_INT8 CUDA_TYPES, 209
CUDA_TYPES, 207 CU_DEVICE_ATTRIBUTE_MAX_PITCH
CU_COMPUTEMODE_DEFAULT CUDA_TYPES, 209
CUDA_TYPES, 207 CU_DEVICE_ATTRIBUTE_MAX_REGISTERS_-
CU_COMPUTEMODE_EXCLUSIVE PER_BLOCK
CUDA_TYPES, 207 CUDA_TYPES, 209
CU_COMPUTEMODE_PROHIBITED CU_DEVICE_ATTRIBUTE_MAX_SHARED_-
CUDA_TYPES, 207 MEMORY_PER_BLOCK
CU_CTX_BLOCKING_SYNC CUDA_TYPES, 209
CUDA_TYPES, 208 CU_DEVICE_ATTRIBUTE_MAX_THREADS_PER_-
CU_CTX_LMEM_RESIZE_TO_MAX BLOCK
CUDA_TYPES, 208 CUDA_TYPES, 209
CU_CTX_MAP_HOST CU_DEVICE_ATTRIBUTE_MULTIPROCESSOR_-
CUDA_TYPES, 208 COUNT
CU_CTX_SCHED_AUTO CUDA_TYPES, 209
CUDA_TYPES, 208 CU_DEVICE_ATTRIBUTE_REGISTERS_PER_-
CU_CTX_SCHED_SPIN BLOCK
CUDA_TYPES, 208 CUDA_TYPES, 209
CU_CTX_SCHED_YIELD CU_DEVICE_ATTRIBUTE_SHARED_MEMORY_-
CUDA_TYPES, 208 PER_BLOCK
CU_DEVICE_ATTRIBUTE_CAN_MAP_HOST_- CUDA_TYPES, 209
MEMORY CU_DEVICE_ATTRIBUTE_TEXTURE_ALIGNMENT
CUDA_TYPES, 209 CUDA_TYPES, 209
CU_DEVICE_ATTRIBUTE_CLOCK_RATE CU_DEVICE_ATTRIBUTE_TOTAL_CONSTANT_-
CUDA_TYPES, 209 MEMORY
CU_DEVICE_ATTRIBUTE_COMPUTE_MODE CUDA_TYPES, 209
CUDA_TYPES, 209 CU_DEVICE_ATTRIBUTE_WARP_SIZE
CU_DEVICE_ATTRIBUTE_GPU_OVERLAP CUDA_TYPES, 209
CUDA_TYPES, 209 CU_EVENT_BLOCKING_SYNC
INDEX 229

CUDA_TYPES, 209 CU_TR_ADDRESS_MODE_CLAMP

CU_EVENT_DEFAULT CUDA_TYPES, 207
CUDA_TYPES, 209 CU_TR_ADDRESS_MODE_MIRROR
CU_FUNC_ATTRIBUTE_CONST_SIZE_BYTES CUDA_TYPES, 207
CUDA_TYPES, 210 CU_TR_ADDRESS_MODE_WRAP
CU_FUNC_ATTRIBUTE_LOCAL_SIZE_BYTES CUDA_TYPES, 207
CUDA_TYPES, 210 CU_TR_FILTER_MODE_LINEAR
CU_FUNC_ATTRIBUTE_MAX_THREADS_PER_- CUDA_TYPES, 210
BLOCK CU_TR_FILTER_MODE_POINT
CUDA_TYPES, 210 CUDA_TYPES, 210
CU_FUNC_ATTRIBUTE_NUM_REGS CU_MEMHOSTALLOC_DEVICEMAP
CUDA_TYPES, 210 CUDA_TYPES, 205
CU_FUNC_ATTRIBUTE_SHARED_SIZE_BYTES CU_MEMHOSTALLOC_PORTABLE
CUDA_TYPES, 210 CUDA_TYPES, 205
CU_JIT_ERROR_LOG_BUFFER CU_MEMHOSTALLOC_WRITECOMBINED
CUDA_TYPES, 211 CUDA_TYPES, 205
CU_JIT_ERROR_LOG_BUFFER_SIZE_BYTES CU_PARAM_TR_DEFAULT
CUDA_TYPES, 211 CUDA_TYPES, 205
CU_JIT_FALLBACK_STRATEGY CU_TRSA_OVERRIDE_FORMAT
CUDA_TYPES, 211 CUDA_TYPES, 205
CU_JIT_INFO_LOG_BUFFER CU_TRSF_NORMALIZED_COORDINATES
CUDA_TYPES, 211 CUDA_TYPES, 205
CU_JIT_INFO_LOG_BUFFER_SIZE_BYTES CU_TRSF_READ_AS_INTEGER
CUDA_TYPES, 211 CUDA_TYPES, 205
CU_JIT_MAX_REGISTERS CUaddress_mode
CUDA_TYPES, 210 CUDA_TYPES, 205
CU_JIT_OPTIMIZATION_LEVEL CUaddress_mode_enum
CUDA_TYPES, 211 CUDA_TYPES, 207
CU_JIT_TARGET cuArray3DCreate
CUDA_TYPES, 211 CUMEM, 132
CU_JIT_TARGET_FROM_CUCONTEXT cuArray3DGetDescriptor
CUDA_TYPES, 211 CUMEM, 134
CU_JIT_THREADS_PER_BLOCK CUarray_format
CUDA_TYPES, 210 CUDA_TYPES, 205
CU_JIT_WALL_TIME CUarray_format_enum
CUDA_TYPES, 210 CUDA_TYPES, 207
CU_MEMORYTYPE_ARRAY cuArrayCreate
CUDA_TYPES, 211 CUMEM, 134
CU_MEMORYTYPE_DEVICE cuArrayDestroy
CUDA_TYPES, 211 CUMEM, 136
CU_MEMORYTYPE_HOST cuArrayGetDescriptor
CUDA_TYPES, 211 CUMEM, 136
CU_PREFER_BINARY CUcomputemode
CUDA_TYPES, 210 CUDA_TYPES, 206
CU_PREFER_PTX CUcomputemode_enum
CUDA_TYPES, 210 CUDA_TYPES, 207
CU_TARGET_COMPUTE_10 CUCTX
CUDA_TYPES, 211 cuCtxAttach, 106
CU_TARGET_COMPUTE_11 cuCtxCreate, 107
CUDA_TYPES, 211 cuCtxDestroy, 108
CU_TARGET_COMPUTE_12 cuCtxDetach, 108
CUDA_TYPES, 211 cuCtxGetDevice, 109
CU_TARGET_COMPUTE_13 cuCtxPopCurrent, 109
CUDA_TYPES, 211 cuCtxPushCurrent, 109

Generated for NVIDIA CUDA Library by Doxygen

230 INDEX

cuCtxSynchronize, 110 cuD3D10UnmapResources

CUctx_flags CUD3D10, 199
CUDA_TYPES, 206 cuD3D10UnregisterResource
CUctx_flags_enum CUD3D10, 199
CUDA_TYPES, 207 CUD3D9
cuCtxAttach cuD3D9CtxCreate, 181
CUCTX, 106 cuD3D9GetDevice, 181
cuCtxCreate cuD3D9GetDirect3DDevice, 182
CUCTX, 107 cuD3D9MapResources, 182
cuCtxDestroy cuD3D9RegisterResource, 183
CUCTX, 108 cuD3D9ResourceGetMappedArray, 184
cuCtxDetach cuD3D9ResourceGetMappedPitch, 185
CUCTX, 108 cuD3D9ResourceGetMappedPointer, 186
cuCtxGetDevice cuD3D9ResourceGetMappedSize, 186
CUCTX, 109 cuD3D9ResourceGetSurfaceDimensions, 187
cuCtxPopCurrent cuD3D9ResourceSetMapFlags, 188
CUCTX, 109 cuD3D9UnmapResources, 189
cuCtxPushCurrent cuD3D9UnregisterResource, 189
CUCTX, 109 cuD3D9CtxCreate
cuCtxSynchronize CUD3D9, 181
CUCTX, 110 cuD3D9GetDevice
CUD3D10 CUD3D9, 181
cuD3D10CtxCreate, 192 cuD3D9GetDirect3DDevice
cuD3D10GetDevice, 192 CUD3D9, 182
cuD3D10MapResources, 193 cuD3D9MapResources
cuD3D10RegisterResource, 193 CUD3D9, 182
cuD3D10ResourceGetMappedArray, 194 cuD3D9RegisterResource
cuD3D10ResourceGetMappedPitch, 195 CUD3D9, 183
cuD3D10ResourceGetMappedPointer, 196 cuD3D9ResourceGetMappedArray
cuD3D10ResourceGetMappedSize, 197 CUD3D9, 184
cuD3D10ResourceGetSurfaceDimensions, 197 cuD3D9ResourceGetMappedPitch
cuD3D10ResourceSetMapFlags, 198 CUD3D9, 185
cuD3D10UnmapResources, 199 cuD3D9ResourceGetMappedPointer
cuD3D10UnregisterResource, 199 CUD3D9, 186
cuD3D10CtxCreate cuD3D9ResourceGetMappedSize
CUD3D10, 192 CUD3D9, 186
cuD3D10GetDevice cuD3D9ResourceGetSurfaceDimensions
CUD3D10, 192 CUD3D9, 187
cuD3D10MapResources cuD3D9ResourceSetMapFlags
CUD3D10, 193 CUD3D9, 188
cuD3D10RegisterResource cuD3D9UnmapResources
CUD3D10, 193 CUD3D9, 189
cuD3D10ResourceGetMappedArray cuD3D9UnregisterResource
CUD3D10, 194 CUD3D9, 189
cuD3D10ResourceGetMappedPitch CUDA Driver API, 97
CUD3D10, 195 CUDA Runtime API, 5
cuD3D10ResourceGetMappedPointer CUDA_ERROR_ALREADY_ACQUIRED
CUD3D10, 196 CUDA_TYPES, 208
cuD3D10ResourceGetMappedSize CUDA_ERROR_ALREADY_MAPPED
CUD3D10, 197 CUDA_TYPES, 208
cuD3D10ResourceGetSurfaceDimensions CUDA_ERROR_ARRAY_IS_MAPPED
CUD3D10, 197 CUDA_TYPES, 208
cuD3D10ResourceSetMapFlags CUDA_ERROR_CONTEXT_ALREADY_CURRENT
CUD3D10, 198 CUDA_TYPES, 208

Generated for NVIDIA CUDA Library by Doxygen

INDEX 231

CUDA_ERROR_DEINITIALIZED CU_AD_FORMAT_UNSIGNED_INT32, 207

CUDA_TYPES, 208 CU_AD_FORMAT_UNSIGNED_INT8, 207
CUDA_ERROR_FILE_NOT_FOUND CU_COMPUTEMODE_DEFAULT, 207
CUDA_TYPES, 208 CU_COMPUTEMODE_EXCLUSIVE, 207
CUDA_ERROR_INVALID_CONTEXT CU_COMPUTEMODE_PROHIBITED, 207
CUDA_TYPES, 208 CU_CTX_BLOCKING_SYNC, 208
CUDA_ERROR_INVALID_DEVICE CU_CTX_LMEM_RESIZE_TO_MAX, 208
CUDA_TYPES, 208 CU_CTX_MAP_HOST, 208
CUDA_ERROR_INVALID_HANDLE CU_CTX_SCHED_AUTO, 208
CUDA_TYPES, 208 CU_CTX_SCHED_SPIN, 208
CUDA_ERROR_INVALID_IMAGE CU_CTX_SCHED_YIELD, 208
CUDA_TYPES, 208 CU_DEVICE_ATTRIBUTE_CAN_MAP_HOST_-
CUDA_ERROR_INVALID_SOURCE MEMORY, 209
CUDA_TYPES, 208 CU_DEVICE_ATTRIBUTE_CLOCK_RATE, 209
CUDA_ERROR_INVALID_VALUE CU_DEVICE_ATTRIBUTE_COMPUTE_MODE,
CUDA_TYPES, 208 209
CUDA_ERROR_LAUNCH_FAILED CU_DEVICE_ATTRIBUTE_GPU_OVERLAP,
CUDA_TYPES, 208 209
CUDA_ERROR_LAUNCH_INCOMPATIBLE_- CU_DEVICE_ATTRIBUTE_INTEGRATED, 209
TEXTURING CU_DEVICE_ATTRIBUTE_KERNEL_EXEC_-
CUDA_TYPES, 208 TIMEOUT, 209
CUDA_ERROR_LAUNCH_OUT_OF_RESOURCES CU_DEVICE_ATTRIBUTE_MAX_BLOCK_-
CUDA_TYPES, 208 DIM_X, 209
CUDA_ERROR_LAUNCH_TIMEOUT CU_DEVICE_ATTRIBUTE_MAX_BLOCK_-
CUDA_TYPES, 208 DIM_Y, 209
CUDA_ERROR_MAP_FAILED CU_DEVICE_ATTRIBUTE_MAX_BLOCK_-
CUDA_TYPES, 208 DIM_Z, 209
CUDA_ERROR_NO_BINARY_FOR_GPU CU_DEVICE_ATTRIBUTE_MAX_GRID_DIM_-
CUDA_TYPES, 208 X, 209
CUDA_ERROR_NO_DEVICE CU_DEVICE_ATTRIBUTE_MAX_GRID_DIM_-
CUDA_TYPES, 208 Y, 209
CUDA_ERROR_NOT_FOUND CU_DEVICE_ATTRIBUTE_MAX_GRID_DIM_-
CUDA_TYPES, 208 Z, 209
CUDA_ERROR_NOT_INITIALIZED CU_DEVICE_ATTRIBUTE_MAX_PITCH, 209
CUDA_TYPES, 208 CU_DEVICE_ATTRIBUTE_MAX_-
CUDA_ERROR_NOT_MAPPED REGISTERS_PER_BLOCK, 209
CUDA_TYPES, 208 CU_DEVICE_ATTRIBUTE_MAX_SHARED_-
CUDA_ERROR_NOT_READY MEMORY_PER_BLOCK, 209
CUDA_TYPES, 208 CU_DEVICE_ATTRIBUTE_MAX_THREADS_-
CUDA_ERROR_OUT_OF_MEMORY PER_BLOCK, 209
CUDA_TYPES, 208 CU_DEVICE_ATTRIBUTE_-
CUDA_ERROR_UNKNOWN MULTIPROCESSOR_COUNT, 209
CUDA_TYPES, 208 CU_DEVICE_ATTRIBUTE_REGISTERS_PER_-
CUDA_ERROR_UNMAP_FAILED BLOCK, 209
CUDA_TYPES, 208 CU_DEVICE_ATTRIBUTE_SHARED_-
CUDA_SUCCESS MEMORY_PER_BLOCK, 209
CUDA_TYPES, 208 CU_DEVICE_ATTRIBUTE_TEXTURE_-
CUDA_TYPES ALIGNMENT, 209
CU_AD_FORMAT_FLOAT, 207 CU_DEVICE_ATTRIBUTE_TOTAL_-
CU_AD_FORMAT_HALF, 207 CONSTANT_MEMORY, 209
CU_AD_FORMAT_SIGNED_INT16, 207 CU_DEVICE_ATTRIBUTE_WARP_SIZE, 209
CU_AD_FORMAT_SIGNED_INT32, 207 CU_EVENT_BLOCKING_SYNC, 209
CU_AD_FORMAT_SIGNED_INT8, 207 CU_EVENT_DEFAULT, 209
CU_AD_FORMAT_UNSIGNED_INT16, 207 CU_FUNC_ATTRIBUTE_CONST_SIZE_BYTES,

Generated for NVIDIA CUDA Library by Doxygen

232 INDEX

210 CUDA_ERROR_MAP_FAILED, 208

CU_FUNC_ATTRIBUTE_LOCAL_SIZE_BYTES, CUDA_ERROR_NO_BINARY_FOR_GPU, 208
210 CUDA_ERROR_NO_DEVICE, 208
CU_FUNC_ATTRIBUTE_MAX_THREADS_- CUDA_ERROR_NOT_FOUND, 208
PER_BLOCK, 210 CUDA_ERROR_NOT_INITIALIZED, 208
CU_FUNC_ATTRIBUTE_NUM_REGS, 210 CUDA_ERROR_NOT_MAPPED, 208
CU_FUNC_ATTRIBUTE_SHARED_SIZE_- CUDA_ERROR_NOT_READY, 208
BYTES, 210 CUDA_ERROR_OUT_OF_MEMORY, 208
CU_JIT_ERROR_LOG_BUFFER, 211 CUDA_ERROR_UNKNOWN, 208
CU_JIT_ERROR_LOG_BUFFER_SIZE_BYTES, CUDA_ERROR_UNMAP_FAILED, 208
211 CUDA_SUCCESS, 208
CU_JIT_FALLBACK_STRATEGY, 211 CUDA_ARRAY3D_DESCRIPTOR, 213
CU_JIT_INFO_LOG_BUFFER, 211 CUDA_ARRAY_DESCRIPTOR, 214
CU_JIT_INFO_LOG_BUFFER_SIZE_BYTES, CUDA_MEMCPY2D
211 CUDA_TYPES, 206
CU_JIT_MAX_REGISTERS, 210 CUDA_MEMCPY2D_st, 215
CU_JIT_OPTIMIZATION_LEVEL, 211 CUDA_MEMCPY3D
CU_JIT_TARGET, 211 CUDA_TYPES, 206
CU_JIT_TARGET_FROM_CUCONTEXT, 211 CUDA_MEMCPY3D_st, 217
CU_JIT_THREADS_PER_BLOCK, 210 CUDA_TYPES
CU_JIT_WALL_TIME, 210 CU_MEMHOSTALLOC_DEVICEMAP, 205
CU_MEMORYTYPE_ARRAY, 211 CU_MEMHOSTALLOC_PORTABLE, 205
CU_MEMORYTYPE_DEVICE, 211 CU_MEMHOSTALLOC_WRITECOMBINED,
CU_MEMORYTYPE_HOST, 211 205
CU_PREFER_BINARY, 210 CU_PARAM_TR_DEFAULT, 205
CU_PREFER_PTX, 210 CU_TRSA_OVERRIDE_FORMAT, 205
CU_TARGET_COMPUTE_10, 211 CU_TRSF_NORMALIZED_COORDINATES, 205
CU_TARGET_COMPUTE_11, 211 CU_TRSF_READ_AS_INTEGER, 205
CU_TARGET_COMPUTE_12, 211 CUaddress_mode, 205
CU_TARGET_COMPUTE_13, 211 CUaddress_mode_enum, 207
CU_TR_ADDRESS_MODE_CLAMP, 207 CUarray_format, 205
CU_TR_ADDRESS_MODE_MIRROR, 207 CUarray_format_enum, 207
CU_TR_ADDRESS_MODE_WRAP, 207 CUcomputemode, 206
CU_TR_FILTER_MODE_LINEAR, 210 CUcomputemode_enum, 207
CU_TR_FILTER_MODE_POINT, 210 CUctx_flags, 206
CUDA_ERROR_ALREADY_ACQUIRED, 208 CUctx_flags_enum, 207
CUDA_ERROR_ALREADY_MAPPED, 208 CUDA_MEMCPY2D, 206
CUDA_ERROR_ARRAY_IS_MAPPED, 208 CUDA_MEMCPY3D, 206
CUDA_ERROR_CONTEXT_ALREADY_- CUDA_VERSION, 205
CURRENT, 208 cudaError_enum, 208
CUDA_ERROR_DEINITIALIZED, 208 CUdevice_attribute, 206
CUDA_ERROR_FILE_NOT_FOUND, 208 CUdevice_attribute_enum, 208
CUDA_ERROR_INVALID_CONTEXT, 208 CUdevprop, 206
CUDA_ERROR_INVALID_DEVICE, 208 CUevent_flags, 206
CUDA_ERROR_INVALID_HANDLE, 208 CUevent_flags_enum, 209
CUDA_ERROR_INVALID_IMAGE, 208 CUfilter_mode, 206
CUDA_ERROR_INVALID_SOURCE, 208 CUfilter_mode_enum, 209
CUDA_ERROR_INVALID_VALUE, 208 CUfunction_attribute, 206
CUDA_ERROR_LAUNCH_FAILED, 208 CUfunction_attribute_enum, 210
CUDA_ERROR_LAUNCH_INCOMPATIBLE_- CUjit_fallback, 206
TEXTURING, 208 CUjit_fallback_enum, 210
CUDA_ERROR_LAUNCH_OUT_OF_- CUjit_option, 206
RESOURCES, 208 CUjit_option_enum, 210
CUDA_ERROR_LAUNCH_TIMEOUT, 208 CUjit_target, 206

Generated for NVIDIA CUDA Library by Doxygen

INDEX 233

CUjit_target_enum, 211 CUDART_D3D10, 69

CUmemorytype, 207 cudaD3D10RegisterFlagsArray
CUmemorytype_enum, 211 CUDART_D3D10, 69
CUresult, 207 cudaD3D10RegisterFlagsNone
CUDA_VERSION CUDART_D3D10, 69
CUDA_TYPES, 205 cudaD3D10RegisterResource
cudaBindTexture CUDART_D3D10, 70
CUDART_HIGHLEVEL, 85 cudaD3D10ResourceGetMappedArray
CUDART_TEXTURE, 78 CUDART_D3D10, 71
cudaBindTexture2D cudaD3D10ResourceGetMappedPitch
CUDART_HIGHLEVEL, 86 CUDART_D3D10, 72
CUDART_TEXTURE, 79 cudaD3D10ResourceGetMappedPointer
cudaBindTextureToArray CUDART_D3D10, 73
CUDART_HIGHLEVEL, 87 cudaD3D10ResourceGetMappedSize
CUDART_TEXTURE, 80 CUDART_D3D10, 73
cudaChannelFormatDesc, 219 cudaD3D10ResourceGetSurfaceDimensions
cudaChannelFormatKind CUDART_D3D10, 74
CUDART_TYPES, 94 cudaD3D10ResourceSetMapFlags
cudaChannelFormatKindFloat CUDART_D3D10, 75
CUDART_TYPES, 95 cudaD3D10SetDirect3DDevice
cudaChannelFormatKindNone CUDART_D3D10, 75
CUDART_TYPES, 95 cudaD3D10UnmapResources
cudaChannelFormatKindSigned CUDART_D3D10, 76
CUDART_TYPES, 95 cudaD3D10UnregisterResource
cudaChannelFormatKindUnsigned CUDART_D3D10, 77
CUDART_TYPES, 95 cudaD3D9GetDevice
cudaChooseDevice CUDART_D3D9, 58
CUDART_DEVICE, 9 cudaD3D9GetDirect3DDevice
cudaComputeMode CUDART_D3D9, 59
CUDART_TYPES, 95 cudaD3D9MapFlags
cudaComputeModeDefault CUDART_D3D9, 58
CUDART_TYPES, 95 cudaD3D9MapFlagsNone
cudaComputeModeExclusive CUDART_D3D9, 58
CUDART_TYPES, 95 cudaD3D9MapFlagsReadOnly
cudaComputeModeProhibited CUDART_D3D9, 58
CUDART_TYPES, 95 cudaD3D9MapFlagsWriteDiscard
cudaConfigureCall CUDART_D3D9, 58
CUDART_EXECUTION, 20 cudaD3D9MapResources
cudaCreateChannelDesc CUDART_D3D9, 59
CUDART_HIGHLEVEL, 88 cudaD3D9RegisterFlags
CUDART_TEXTURE, 80 CUDART_D3D9, 58
cudaD3D10GetDevice cudaD3D9RegisterFlagsArray
CUDART_D3D10, 69 CUDART_D3D9, 58
cudaD3D10MapFlags cudaD3D9RegisterFlagsNone
CUDART_D3D10, 69 CUDART_D3D9, 58
cudaD3D10MapFlagsNone cudaD3D9RegisterResource
CUDART_D3D10, 69 CUDART_D3D9, 60
cudaD3D10MapFlagsReadOnly cudaD3D9ResourceGetMappedArray
CUDART_D3D10, 69 CUDART_D3D9, 61
cudaD3D10MapFlagsWriteDiscard cudaD3D9ResourceGetMappedPitch
CUDART_D3D10, 69 CUDART_D3D9, 62
cudaD3D10MapResources cudaD3D9ResourceGetMappedPointer
CUDART_D3D10, 70 CUDART_D3D9, 63
cudaD3D10RegisterFlags cudaD3D9ResourceGetMappedSize

Generated for NVIDIA CUDA Library by Doxygen

234 INDEX

CUDART_D3D9, 63 cudaErrorInvalidTexture
cudaD3D9ResourceGetSurfaceDimensions CUDART_TYPES, 95
CUDART_D3D9, 64 cudaErrorInvalidTextureBinding
cudaD3D9ResourceSetMapFlags CUDART_TYPES, 95
CUDART_D3D9, 65 cudaErrorInvalidValue
cudaD3D9SetDirect3DDevice CUDART_TYPES, 95
CUDART_D3D9, 65 cudaErrorLaunchFailure
cudaD3D9UnmapResources CUDART_TYPES, 95
CUDART_D3D9, 66 cudaErrorLaunchOutOfResources
cudaD3D9UnregisterResource CUDART_TYPES, 95
CUDART_D3D9, 67 cudaErrorLaunchTimeout
cudaDeviceProp, 220 CUDART_TYPES, 95
cudaDriverGetVersion cudaErrorMapBufferObjectFailed
CUDART_VERSION, 83 CUDART_TYPES, 95
cudaError cudaErrorMemoryAllocation
CUDART_TYPES, 95 CUDART_TYPES, 95
cudaError_enum cudaErrorMemoryValueTooLarge
CUDA_TYPES, 208 CUDART_TYPES, 96
cudaError_t cudaErrorMissingConfiguration
CUDART_TYPES, 94 CUDART_TYPES, 95
cudaErrorAddressOfConstant cudaErrorMixedDeviceExecution
CUDART_TYPES, 96 CUDART_TYPES, 96
cudaErrorApiFailureBase cudaErrorNoDevice
CUDART_TYPES, 96 CUDART_TYPES, 96
cudaErrorCudartUnloading cudaErrorNotReady
CUDART_TYPES, 96 CUDART_TYPES, 96
cudaErrorInitializationError cudaErrorNotYetImplemented
CUDART_TYPES, 95 CUDART_TYPES, 96
cudaErrorInsufficientDriver cudaErrorPriorLaunchFailure
CUDART_TYPES, 96 CUDART_TYPES, 95
cudaErrorInvalidChannelDescriptor cudaErrorSetOnActiveProcess
CUDART_TYPES, 95 CUDART_TYPES, 96
cudaErrorInvalidConfiguration cudaErrorStartupFailure
CUDART_TYPES, 95 CUDART_TYPES, 96
cudaErrorInvalidDevice cudaErrorSynchronizationError
CUDART_TYPES, 95 CUDART_TYPES, 96
cudaErrorInvalidDeviceFunction cudaErrorTextureFetchFailed
CUDART_TYPES, 95 CUDART_TYPES, 96
cudaErrorInvalidDevicePointer cudaErrorTextureNotBound
CUDART_TYPES, 95 CUDART_TYPES, 96
cudaErrorInvalidFilterSetting cudaErrorUnknown
CUDART_TYPES, 96 CUDART_TYPES, 96
cudaErrorInvalidHostPointer cudaErrorUnmapBufferObjectFailed
CUDART_TYPES, 95 CUDART_TYPES, 95
cudaErrorInvalidMemcpyDirection cudaEvent_t
CUDART_TYPES, 96 CUDART_TYPES, 94
cudaErrorInvalidNormSetting cudaEventCreate
CUDART_TYPES, 96 CUDART_EVENT, 16
cudaErrorInvalidPitchValue cudaEventCreateWithFlags
CUDART_TYPES, 95 CUDART_EVENT, 16
cudaErrorInvalidResourceHandle cudaEventDestroy
CUDART_TYPES, 96 CUDART_EVENT, 17
cudaErrorInvalidSymbol cudaEventElapsedTime
CUDART_TYPES, 95 CUDART_EVENT, 17

Generated for NVIDIA CUDA Library by Doxygen

INDEX 235

cudaEventQuery cudaGLUnregisterBufferObject
CUDART_EVENT, 18 CUDART_OPENGL, 56
cudaEventRecord cudaHostAlloc
CUDART_EVENT, 18 CUDART_MEMORY, 28
cudaEventSynchronize cudaHostGetDevicePointer
CUDART_EVENT, 19 CUDART_MEMORY, 29
cudaExtent, 222 cudaHostGetFlags
cudaFree CUDART_MEMORY, 30
CUDART_MEMORY, 26 cudaLaunch
cudaFreeArray CUDART_EXECUTION, 21
CUDART_MEMORY, 27 CUDART_HIGHLEVEL, 90
cudaFreeHost cudaMalloc
CUDART_MEMORY, 27 CUDART_MEMORY, 30
cudaFuncAttributes, 223 cudaMalloc3D
cudaFuncGetAttributes CUDART_MEMORY, 30
CUDART_EXECUTION, 21 cudaMalloc3DArray
CUDART_HIGHLEVEL, 88 CUDART_MEMORY, 31
cudaGetChannelDesc cudaMallocArray
CUDART_TEXTURE, 81 CUDART_MEMORY, 32
cudaGetDevice cudaMallocHost
CUDART_DEVICE, 9 CUDART_MEMORY, 33
cudaGetDeviceCount cudaMallocPitch
CUDART_DEVICE, 10 CUDART_MEMORY, 33
cudaGetDeviceProperties cudaMemcpy
CUDART_DEVICE, 10 CUDART_MEMORY, 34
cudaGetErrorString cudaMemcpy2D
CUDART_ERROR, 7 CUDART_MEMORY, 34
cudaGetLastError cudaMemcpy2DArrayToArray
CUDART_ERROR, 7 CUDART_MEMORY, 35
cudaGetSymbolAddress cudaMemcpy2DAsync
CUDART_HIGHLEVEL, 89 CUDART_MEMORY, 36
CUDART_MEMORY, 27 cudaMemcpy2DFromArray
cudaGetSymbolSize CUDART_MEMORY, 37
CUDART_HIGHLEVEL, 89 cudaMemcpy2DFromArrayAsync
CUDART_MEMORY, 28 CUDART_MEMORY, 37
cudaGetTextureAlignmentOffset cudaMemcpy2DToArray
CUDART_HIGHLEVEL, 90 CUDART_MEMORY, 38
CUDART_TEXTURE, 81 cudaMemcpy2DToArrayAsync
cudaGetTextureReference CUDART_MEMORY, 39
CUDART_TEXTURE, 82 cudaMemcpy3D
cudaGLMapBufferObject CUDART_MEMORY, 40
CUDART_OPENGL, 52 cudaMemcpy3DAsync
cudaGLMapBufferObjectAsync CUDART_MEMORY, 41
CUDART_OPENGL, 53 cudaMemcpy3DParms, 224
cudaGLRegisterBufferObject cudaMemcpyArrayToArray
CUDART_OPENGL, 53 CUDART_MEMORY, 43
cudaGLSetBufferObjectMapFlags cudaMemcpyAsync
CUDART_OPENGL, 54 CUDART_MEMORY, 43
cudaGLSetGLDevice cudaMemcpyDeviceToDevice
CUDART_OPENGL, 54 CUDART_TYPES, 96
cudaGLUnmapBufferObject cudaMemcpyDeviceToHost
CUDART_OPENGL, 55 CUDART_TYPES, 96
cudaGLUnmapBufferObjectAsync cudaMemcpyFromArray
CUDART_OPENGL, 55 CUDART_MEMORY, 44

Generated for NVIDIA CUDA Library by Doxygen

236 INDEX

cudaMemcpyFromArrayAsync cudaErrorInvalidConfiguration, 95
CUDART_MEMORY, 45 cudaErrorInvalidDevice, 95
cudaMemcpyFromSymbol cudaErrorInvalidDeviceFunction, 95
CUDART_MEMORY, 46 cudaErrorInvalidDevicePointer, 95
cudaMemcpyFromSymbolAsync cudaErrorInvalidFilterSetting, 96
CUDART_MEMORY, 46 cudaErrorInvalidHostPointer, 95
cudaMemcpyHostToDevice cudaErrorInvalidMemcpyDirection, 96
CUDART_TYPES, 96 cudaErrorInvalidNormSetting, 96
cudaMemcpyHostToHost cudaErrorInvalidPitchValue, 95
CUDART_TYPES, 96 cudaErrorInvalidResourceHandle, 96
cudaMemcpyKind cudaErrorInvalidSymbol, 95
CUDART_TYPES, 96 cudaErrorInvalidTexture, 95
cudaMemcpyToArray cudaErrorInvalidTextureBinding, 95
CUDART_MEMORY, 47 cudaErrorInvalidValue, 95
cudaMemcpyToArrayAsync cudaErrorLaunchFailure, 95
CUDART_MEMORY, 48 cudaErrorLaunchOutOfResources, 95
cudaMemcpyToSymbol cudaErrorLaunchTimeout, 95
CUDART_MEMORY, 48 cudaErrorMapBufferObjectFailed, 95
cudaMemcpyToSymbolAsync cudaErrorMemoryAllocation, 95
CUDART_MEMORY, 49 cudaErrorMemoryValueTooLarge, 96
cudaMemset cudaErrorMissingConfiguration, 95
CUDART_MEMORY, 50 cudaErrorMixedDeviceExecution, 96
cudaMemset2D cudaErrorNoDevice, 96
CUDART_MEMORY, 50 cudaErrorNotReady, 96
cudaMemset3D cudaErrorNotYetImplemented, 96
CUDART_MEMORY, 51 cudaErrorPriorLaunchFailure, 95
cudaPitchedPtr, 225 cudaErrorSetOnActiveProcess, 96
cudaPos, 226 cudaErrorStartupFailure, 96
CUDART_D3D10 cudaErrorSynchronizationError, 96
cudaD3D10MapFlagsNone, 69 cudaErrorTextureFetchFailed, 96
cudaD3D10MapFlagsReadOnly, 69 cudaErrorTextureNotBound, 96
cudaD3D10MapFlagsWriteDiscard, 69 cudaErrorUnknown, 96
cudaD3D10RegisterFlagsArray, 69 cudaErrorUnmapBufferObjectFailed, 95
cudaD3D10RegisterFlagsNone, 69 cudaMemcpyDeviceToDevice, 96
CUDART_D3D9 cudaMemcpyDeviceToHost, 96
cudaD3D9MapFlagsNone, 58 cudaMemcpyHostToDevice, 96
cudaD3D9MapFlagsReadOnly, 58 cudaMemcpyHostToHost, 96
cudaD3D9MapFlagsWriteDiscard, 58 cudaSuccess, 95
cudaD3D9RegisterFlagsArray, 58 CUDART_D3D10
cudaD3D9RegisterFlagsNone, 58 cudaD3D10GetDevice, 69
CUDART_TYPES cudaD3D10MapFlags, 69
cudaChannelFormatKindFloat, 95 cudaD3D10MapResources, 70
cudaChannelFormatKindNone, 95 cudaD3D10RegisterFlags, 69
cudaChannelFormatKindSigned, 95 cudaD3D10RegisterResource, 70
cudaChannelFormatKindUnsigned, 95 cudaD3D10ResourceGetMappedArray, 71
cudaComputeModeDefault, 95 cudaD3D10ResourceGetMappedPitch, 72
cudaComputeModeExclusive, 95 cudaD3D10ResourceGetMappedPointer, 73
cudaComputeModeProhibited, 95 cudaD3D10ResourceGetMappedSize, 73
cudaErrorAddressOfConstant, 96 cudaD3D10ResourceGetSurfaceDimensions, 74
cudaErrorApiFailureBase, 96 cudaD3D10ResourceSetMapFlags, 75
cudaErrorCudartUnloading, 96 cudaD3D10SetDirect3DDevice, 75
cudaErrorInitializationError, 95 cudaD3D10UnmapResources, 76
cudaErrorInsufficientDriver, 96 cudaD3D10UnregisterResource, 77
cudaErrorInvalidChannelDescriptor, 95 CUDART_D3D9

Generated for NVIDIA CUDA Library by Doxygen

INDEX 237

cudaD3D9GetDevice, 58 cudaFree, 26
cudaD3D9GetDirect3DDevice, 59 cudaFreeArray, 27
cudaD3D9MapFlags, 58 cudaFreeHost, 27
cudaD3D9MapResources, 59 cudaGetSymbolAddress, 27
cudaD3D9RegisterFlags, 58 cudaGetSymbolSize, 28
cudaD3D9RegisterResource, 60 cudaHostAlloc, 28
cudaD3D9ResourceGetMappedArray, 61 cudaHostGetDevicePointer, 29
cudaD3D9ResourceGetMappedPitch, 62 cudaHostGetFlags, 30
cudaD3D9ResourceGetMappedPointer, 63 cudaMalloc, 30
cudaD3D9ResourceGetMappedSize, 63 cudaMalloc3D, 30
cudaD3D9ResourceGetSurfaceDimensions, 64 cudaMalloc3DArray, 31
cudaD3D9ResourceSetMapFlags, 65 cudaMallocArray, 32
cudaD3D9SetDirect3DDevice, 65 cudaMallocHost, 33
cudaD3D9UnmapResources, 66 cudaMallocPitch, 33
cudaD3D9UnregisterResource, 67 cudaMemcpy, 34
CUDART_DEVICE cudaMemcpy2D, 34
cudaChooseDevice, 9 cudaMemcpy2DArrayToArray, 35
cudaGetDevice, 9 cudaMemcpy2DAsync, 36
cudaGetDeviceCount, 10 cudaMemcpy2DFromArray, 37
cudaGetDeviceProperties, 10 cudaMemcpy2DFromArrayAsync, 37
cudaSetDevice, 12 cudaMemcpy2DToArray, 38
cudaSetDeviceFlags, 12 cudaMemcpy2DToArrayAsync, 39
cudaSetValidDevices, 13 cudaMemcpy3D, 40
CUDART_ERROR cudaMemcpy3DAsync, 41
cudaGetErrorString, 7 cudaMemcpyArrayToArray, 43
cudaGetLastError, 7 cudaMemcpyAsync, 43
CUDART_EVENT cudaMemcpyFromArray, 44
cudaEventCreate, 16 cudaMemcpyFromArrayAsync, 45
cudaEventCreateWithFlags, 16 cudaMemcpyFromSymbol, 46
cudaEventDestroy, 17 cudaMemcpyFromSymbolAsync, 46
cudaEventElapsedTime, 17 cudaMemcpyToArray, 47
cudaEventQuery, 18 cudaMemcpyToArrayAsync, 48
cudaEventRecord, 18 cudaMemcpyToSymbol, 48
cudaEventSynchronize, 19 cudaMemcpyToSymbolAsync, 49
CUDART_EXECUTION cudaMemset, 50
cudaConfigureCall, 20 cudaMemset2D, 50
cudaFuncGetAttributes, 21 cudaMemset3D, 51
cudaLaunch, 21 CUDART_OPENGL
cudaSetDoubleForDevice, 21 cudaGLMapBufferObject, 52
cudaSetDoubleForHost, 22 cudaGLMapBufferObjectAsync, 53
cudaSetupArgument, 22 cudaGLRegisterBufferObject, 53
CUDART_HIGHLEVEL cudaGLSetBufferObjectMapFlags, 54
cudaBindTexture, 85 cudaGLSetGLDevice, 54
cudaBindTexture2D, 86 cudaGLUnmapBufferObject, 55
cudaBindTextureToArray, 87 cudaGLUnmapBufferObjectAsync, 55
cudaCreateChannelDesc, 88 cudaGLUnregisterBufferObject, 56
cudaFuncGetAttributes, 88 cudaWGLGetDevice, 56
cudaGetSymbolAddress, 89 CUDART_STREAM
cudaGetSymbolSize, 89 cudaStreamCreate, 14
cudaGetTextureAlignmentOffset, 90 cudaStreamDestroy, 14
cudaLaunch, 90 cudaStreamQuery, 15
cudaSetupArgument, 90 cudaStreamSynchronize, 15
cudaUnbindTexture, 91 CUDART_TEXTURE
CUDART_MEMORY cudaBindTexture, 78

Generated for NVIDIA CUDA Library by Doxygen

238 INDEX

cudaBindTexture2D, 79 CUDART_TEXTURE, 82
cudaBindTextureToArray, 80 cudaWGLGetDevice
cudaCreateChannelDesc, 80 CUDART_OPENGL, 56
cudaGetChannelDesc, 81 CUDEVICE
cudaGetTextureAlignmentOffset, 81 cuDeviceComputeCapability, 99
cudaGetTextureReference, 82 cuDeviceGet, 100
cudaUnbindTexture, 82 cuDeviceGetAttribute, 100
CUDART_THREAD cuDeviceGetCount, 101
cudaThreadExit, 6 cuDeviceGetName, 102
cudaThreadSynchronize, 6 cuDeviceGetProperties, 102
CUDART_TYPES cuDeviceTotalMem, 103
cudaChannelFormatKind, 94 CUdevice_attribute
cudaComputeMode, 95 CUDA_TYPES, 206
cudaError, 95 CUdevice_attribute_enum
cudaError_t, 94 CUDA_TYPES, 208
cudaEvent_t, 94 cuDeviceComputeCapability
cudaMemcpyKind, 96 CUDEVICE, 99
cudaStream_t, 94 cuDeviceGet
CUDART_VERSION CUDEVICE, 100
cudaDriverGetVersion, 83 cuDeviceGetAttribute
cudaRuntimeGetVersion, 83 CUDEVICE, 100
cudaRuntimeGetVersion cuDeviceGetCount
CUDART_VERSION, 83 CUDEVICE, 101
cudaSetDevice cuDeviceGetName
CUDART_DEVICE, 12 CUDEVICE, 102
cudaSetDeviceFlags cuDeviceGetProperties
CUDART_DEVICE, 12 CUDEVICE, 102
cudaSetDoubleForDevice cuDeviceTotalMem
CUDART_EXECUTION, 21 CUDEVICE, 103
cudaSetDoubleForHost CUdevprop
CUDART_EXECUTION, 22 CUDA_TYPES, 206
cudaSetupArgument CUdevprop_st, 227
CUDART_EXECUTION, 22 cuDriverGetVersion
CUDART_HIGHLEVEL, 90 CUVERSION, 105
cudaSetValidDevices CUEVENT
CUDART_DEVICE, 13 cuEventCreate, 119
cudaStream_t cuEventDestroy, 120
CUDART_TYPES, 94 cuEventElapsedTime, 120
cudaStreamCreate cuEventQuery, 120
CUDART_STREAM, 14 cuEventRecord, 121
cudaStreamDestroy cuEventSynchronize, 121
CUDART_STREAM, 14 CUevent_flags
cudaStreamQuery CUDA_TYPES, 206
CUDART_STREAM, 15 CUevent_flags_enum
cudaStreamSynchronize CUDA_TYPES, 209
CUDART_STREAM, 15 cuEventCreate
cudaSuccess CUEVENT, 119
CUDART_TYPES, 95 cuEventDestroy
cudaThreadExit CUEVENT, 120
CUDART_THREAD, 6 cuEventElapsedTime
cudaThreadSynchronize CUEVENT, 120
CUDART_THREAD, 6 cuEventQuery
cudaUnbindTexture CUEVENT, 120
CUDART_HIGHLEVEL, 91 cuEventRecord

Generated for NVIDIA CUDA Library by Doxygen

INDEX 239

CUEVENT, 121 cuGLUnmapBufferObjectAsync

cuEventSynchronize CUGL, 178
CUEVENT, 121 cuGLUnregisterBufferObject
CUEXEC CUGL, 178
cuFuncGetAttribute, 123 CUINIT
cuFuncSetBlockShape, 124 cuInit, 98
cuFuncSetSharedSize, 124 cuInit
cuLaunch, 125 CUINIT, 98
cuLaunchGrid, 125 CUjit_fallback
cuLaunchGridAsync, 126 CUDA_TYPES, 206
cuParamSetf, 126 CUjit_fallback_enum
cuParamSeti, 127 CUDA_TYPES, 210
cuParamSetSize, 127 CUjit_option
cuParamSetTexRef, 128 CUDA_TYPES, 206
cuParamSetv, 128 CUjit_option_enum
CUfilter_mode CUDA_TYPES, 210
CUDA_TYPES, 206 CUjit_target
CUfilter_mode_enum CUDA_TYPES, 206
CUDA_TYPES, 209 CUjit_target_enum
cuFuncGetAttribute CUDA_TYPES, 211
CUEXEC, 123 cuLaunch
cuFuncSetBlockShape CUEXEC, 125
CUEXEC, 124 cuLaunchGrid
cuFuncSetSharedSize CUEXEC, 125
CUEXEC, 124 cuLaunchGridAsync
CUfunction_attribute CUEXEC, 126
CUDA_TYPES, 206 CUMEM
CUfunction_attribute_enum cuArray3DCreate, 132
CUDA_TYPES, 210 cuArray3DGetDescriptor, 134
CUGL cuArrayCreate, 134
cuGLCtxCreate, 174 cuArrayDestroy, 136
cuGLInit, 175 cuArrayGetDescriptor, 136
cuGLMapBufferObject, 175 cuMemAlloc, 137
cuGLMapBufferObjectAsync, 176 cuMemAllocHost, 138
cuGLRegisterBufferObject, 176 cuMemAllocPitch, 138
cuGLSetBufferObjectMapFlags, 177 cuMemcpy2D, 139
cuGLUnmapBufferObject, 177 cuMemcpy2DAsync, 141
cuGLUnmapBufferObjectAsync, 178 cuMemcpy2DUnaligned, 143
cuGLUnregisterBufferObject, 178 cuMemcpy3D, 145
cuWGLGetDevice, 179 cuMemcpy3DAsync, 148
cuGLCtxCreate cuMemcpyAtoA, 150
CUGL, 174 cuMemcpyAtoD, 151
cuGLInit cuMemcpyAtoH, 151
CUGL, 175 cuMemcpyAtoHAsync, 152
cuGLMapBufferObject cuMemcpyDtoA, 152
CUGL, 175 cuMemcpyDtoD, 153
cuGLMapBufferObjectAsync cuMemcpyDtoH, 154
CUGL, 176 cuMemcpyDtoHAsync, 154
cuGLRegisterBufferObject cuMemcpyHtoA, 155
CUGL, 176 cuMemcpyHtoAAsync, 155
cuGLSetBufferObjectMapFlags cuMemcpyHtoD, 156
CUGL, 177 cuMemcpyHtoDAsync, 157
cuGLUnmapBufferObject cuMemFree, 157
CUGL, 177 cuMemFreeHost, 158

Generated for NVIDIA CUDA Library by Doxygen

240 INDEX

cuMemGetAddressRange, 158 CUMEM, 158

cuMemGetInfo, 159 cuMemGetAddressRange
cuMemHostAlloc, 159 CUMEM, 158
cuMemHostGetDevicePointer, 160 cuMemGetInfo
cuMemHostGetFlags, 161 CUMEM, 159
cuMemsetD16, 161 cuMemHostAlloc
cuMemsetD2D16, 162 CUMEM, 159
cuMemsetD2D32, 163 cuMemHostGetDevicePointer
cuMemsetD2D8, 163 CUMEM, 160
cuMemsetD32, 164 cuMemHostGetFlags
cuMemsetD8, 164 CUMEM, 161
cuMemAlloc CUmemorytype
CUMEM, 137 CUDA_TYPES, 207
cuMemAllocHost CUmemorytype_enum
CUMEM, 138 CUDA_TYPES, 211
cuMemAllocPitch cuMemsetD16
CUMEM, 138 CUMEM, 161
cuMemcpy2D cuMemsetD2D16
CUMEM, 139 CUMEM, 162
cuMemcpy2DAsync cuMemsetD2D32
CUMEM, 141 CUMEM, 163
cuMemcpy2DUnaligned cuMemsetD2D8
CUMEM, 143 CUMEM, 163
cuMemcpy3D cuMemsetD32
CUMEM, 145 CUMEM, 164
cuMemcpy3DAsync cuMemsetD8
CUMEM, 148 CUMEM, 164
cuMemcpyAtoA CUMODULE
CUMEM, 150 cuModuleGetFunction, 111
cuMemcpyAtoD cuModuleGetGlobal, 112
CUMEM, 151 cuModuleGetTexRef, 112
cuMemcpyAtoH cuModuleLoad, 113
CUMEM, 151 cuModuleLoadData, 113
cuMemcpyAtoHAsync cuModuleLoadDataEx, 114
CUMEM, 152 cuModuleLoadFatBinary, 115
cuMemcpyDtoA cuModuleUnload, 115
CUMEM, 152 cuModuleGetFunction
cuMemcpyDtoD CUMODULE, 111
CUMEM, 153 cuModuleGetGlobal
cuMemcpyDtoH CUMODULE, 112
CUMEM, 154 cuModuleGetTexRef
cuMemcpyDtoHAsync CUMODULE, 112
CUMEM, 154 cuModuleLoad
cuMemcpyHtoA CUMODULE, 113
CUMEM, 155 cuModuleLoadData
cuMemcpyHtoAAsync CUMODULE, 113
CUMEM, 155 cuModuleLoadDataEx
cuMemcpyHtoD CUMODULE, 114
CUMEM, 156 cuModuleLoadFatBinary
cuMemcpyHtoDAsync CUMODULE, 115
CUMEM, 157 cuModuleUnload
cuMemFree CUMODULE, 115
CUMEM, 157 cuParamSetf
cuMemFreeHost CUEXEC, 126

Generated for NVIDIA CUDA Library by Doxygen

INDEX 241

cuParamSeti CUTEXREF, 169

CUEXEC, 127 cuTexRefSetAddress
cuParamSetSize CUTEXREF, 170
CUEXEC, 127 cuTexRefSetAddress2D
cuParamSetTexRef CUTEXREF, 170
CUEXEC, 128 cuTexRefSetAddressMode
cuParamSetv CUTEXREF, 171
CUEXEC, 128 cuTexRefSetArray
CUresult CUTEXREF, 171
CUDA_TYPES, 207 cuTexRefSetFilterMode
CUSTREAM CUTEXREF, 172
cuStreamCreate, 117 cuTexRefSetFlags
cuStreamDestroy, 117 CUTEXREF, 172
cuStreamQuery, 118 cuTexRefSetFormat
cuStreamSynchronize, 118 CUTEXREF, 173
cuStreamCreate CUVERSION
CUSTREAM, 117 cuDriverGetVersion, 105
cuStreamDestroy cuWGLGetDevice
CUSTREAM, 117 CUGL, 179
cuStreamQuery
CUSTREAM, 118 Data types used by CUDA driver, 201
cuStreamSynchronize Data types used by CUDA Runtime, 92
CUSTREAM, 118 Device Management, 9, 99
CUTEXREF Direct3D 10 Interoperability, 68, 191
cuTexRefCreate, 167 Direct3D 9 Interoperability, 57, 180
cuTexRefDestroy, 167
Error Handling, 7
cuTexRefGetAddress, 167
Event Management, 16, 119
cuTexRefGetAddressMode, 168
Execution Control, 20, 123
cuTexRefGetArray, 168
cuTexRefGetFilterMode, 168 Initialization, 98
cuTexRefGetFlags, 169
cuTexRefGetFormat, 169 Memory Management, 24, 130
cuTexRefSetAddress, 170 Module Management, 111
cuTexRefSetAddress2D, 170
cuTexRefSetAddressMode, 171 OpenGL Interoperability, 52, 174
cuTexRefSetArray, 171
cuTexRefSetFilterMode, 172 Stream Management, 14, 117
cuTexRefSetFlags, 172
Texture Reference Management, 78, 166
cuTexRefSetFormat, 173
Thread Management, 6
cuTexRefCreate
CUTEXREF, 167 Version Management, 83, 105
cuTexRefDestroy
CUTEXREF, 167
cuTexRefGetAddress
CUTEXREF, 167
cuTexRefGetAddressMode
CUTEXREF, 168
cuTexRefGetArray
CUTEXREF, 168
cuTexRefGetFilterMode
CUTEXREF, 168
cuTexRefGetFlags
CUTEXREF, 169
cuTexRefGetFormat

Generated for NVIDIA CUDA Library by Doxygen

Notice
ALL NVIDIA DESIGN SPECIFICATIONS, REFERENCE BOARDS, FILES, DRAWINGS, DIAGNOSTICS, LISTS, AND
OTHER DOCUMENTS (TOGETHER AND SEPARATELY, “MATERIALS”) ARE BEING PROVIDED “AS IS.” NVIDIA
MAKES NO WARRANTIES, EXPRESSED, IMPLIED, STATUTORY, OR OTHERWISE WITH RESPECT TO THE
MATERIALS, AND EXPRESSLY DISCLAIMS ALL IMPLIED WARRANTIES OF NONINFRINGEMENT,
MERCHANTABILITY, AND FITNESS FOR A PARTICULAR PURPOSE.
Information furnished is believed to be accurate and reliable. However, NVIDIA Corporation assumes no
responsibility for the consequences of use of such information or for any infringement of patents or other
rights of third parties that may result from its use. No license is granted by implication or otherwise under any
patent or patent rights of NVIDIA Corporation. Specifications mentioned in this publication are subject to
change without notice. This publication supersedes and replaces all information previously supplied. NVIDIA
Corporation products are not authorized for use as critical components in life support devices or systems
without express written approval of NVIDIA Corporation.
Trademarks
NVIDIA, the NVIDIA logo, GeForce, Tesla, and Quadro are trademarks or registered trademarks of NVIDIA
Corporation. Other company and product names may be trademarks of the respective companies with which
they are associated.
Copyright
© 2007-2009 NVIDIA Corporation. All rights reserved.

NVIDIA Corporation
2701 San Tomas Expressway
Santa Clara, CA 95050
www.nvidia.com