Loading src/TNL/Containers/Algorithms/ArrayOperationsCuda_impl.h +4 −0 Original line number Diff line number Diff line Loading @@ -15,6 +15,7 @@ #include <TNL/tnlConfig.h> #include <TNL/Math.h> #include <TNL/Exceptions/CudaSupportMissing.h> #include <TNL/Exceptions/CudaBadAlloc.h> #include <TNL/Containers/Algorithms/ArrayOperations.h> #include <TNL/Containers/Algorithms/Reduction.h> #include <TNL/Containers/Algorithms/reduction-operations.h> Loading @@ -33,7 +34,10 @@ allocateMemory( Element*& data, checkCudaDevice; if( cudaMalloc( ( void** ) &data, ( size_t ) size * sizeof( Element ) ) != cudaSuccess ) { data = 0; throw Exceptions::CudaBadAlloc(); } return checkCudaDevice; #else throw Exceptions::CudaSupportMissing(); Loading src/TNL/Containers/Algorithms/CudaReductionBuffer.h +3 −1 Original line number Diff line number Diff line Loading @@ -11,7 +11,9 @@ #pragma once #include <stdlib.h> #include <TNL/Devices/Cuda.h> #include <TNL/Exceptions/CudaBadAlloc.h> namespace TNL { namespace Containers { Loading @@ -35,8 +37,8 @@ class CudaReductionBuffer this->size = size; if( cudaMalloc( ( void** ) &this->data, size ) != cudaSuccess ) { std::cerr << "I am not able to allocate reduction buffer on the GPU." << std::endl; this->data = 0; throw Exceptions::CudaBadAlloc(); } return checkCudaDevice; } Loading src/TNL/Containers/Algorithms/cuda-prefix-sum_impl.h +3 −6 Original line number Diff line number Diff line Loading @@ -11,7 +11,9 @@ #pragma once #include <iostream> #include <TNL/Devices/Cuda.h> #include <TNL/Exceptions/CudaBadAlloc.h> #include <TNL/Containers/Algorithms/reduction-operations.h> #ifdef HAVE_CUDA Loading Loading @@ -194,12 +196,7 @@ bool cudaRecursivePrefixSum( const enumPrefixSumType prefixSumType, if( cudaMalloc( ( void** ) &auxArray1, auxArraySize ) != cudaSuccess || cudaMalloc( ( void** ) &auxArray2, auxArraySize ) != cudaSuccess ) { { std::cerr << "Not enough memory on device to allocate auxilliary arrays." << std::endl; return false; } } throw Exceptions::CudaBadAlloc(); /**** * Setup block and grid size. Loading src/TNL/Devices/Cuda.cu +2 −376 Original line number Diff line number Diff line Loading @@ -9,6 +9,7 @@ /* See Copyright Notice in tnl/Copyright */ #include <TNL/Devices/Cuda.h> #include <TNL/Exceptions/CudaRuntimeError.h> #include <TNL/Config/ConfigDescription.h> #include <TNL/Config/ParameterContainer.h> Loading Loading @@ -106,382 +107,7 @@ bool Cuda::checkDevice( const char* file_name, int line, cudaError error ) { if( error == cudaSuccess ) return true; std::cerr << "CUDA ERROR(" << error << ") at line " << line << " in " << file_name << ":" << std::endl; std::cerr << cudaGetErrorString( error ) << std::endl; std::cerr << "Detailed description is: " << std::endl; switch( error ) { // 1 case cudaErrorMissingConfiguration: std::cerr << "The device function being invoked (usually via ::cudaLaunch()) was not " << std::endl << "previously configured via the ::cudaConfigureCall() function. " << std::endl; break; // 2 case cudaErrorMemoryAllocation: std::cerr << "The API call failed because it was unable to allocate enough memory to " << std::endl << "perform the requested operation. " << std::endl; break; // 3 case cudaErrorInitializationError: std::cerr << "The API call failed because the CUDA driver and runtime could not be " << std::endl << "initialized. " << std::endl; break; // 4 case cudaErrorLaunchFailure: std::cerr << "An exception occurred on the device while executing a kernel. Common " << std::endl << "causes include dereferencing an invalid device pointer and accessing " << std::endl << "out of bounds shared memory. The device cannot be used until " << std::endl << "::cudaThreadExit() is called. All existing device memory allocations " << std::endl << "are invalid and must be reconstructed if the program is to continue " << std::endl << "using CUDA. " << std::endl; break; // 5 case cudaErrorPriorLaunchFailure: std::cerr << "This indicated that a previous kernel launch failed. This was previously " << std::endl << "used for device emulation of kernel launches. " << std::endl << "This error return is deprecated as of CUDA 3.1. Device emulation mode was " << std::endl << "removed with the CUDA 3.1 release. " << std::endl; break; // 6 case cudaErrorLaunchTimeout: std::cerr << "This indicates that the device kernel took too long to execute. This can " << std::endl << "only occur if timeouts are enabled - see the device property " << std::endl << "ref ::cudaDeviceProp::kernelExecTimeoutEnabled \"kernelExecTimeoutEnabled\" " << std::endl << "for more information. The device cannot be used until ::cudaThreadExit() " << std::endl << "is called. All existing device memory allocations are invalid and must be " << std::endl << "reconstructed if the program is to continue using CUDA. " << std::endl; break; // 7 case cudaErrorLaunchOutOfResources: std::cerr << "This indicates that a launch did not occur because it did not have " << std::endl << "appropriate resources. Although this error is similar to " << std::endl << "::cudaErrorInvalidConfiguration, this error usually indicates that the " << std::endl << "user has attempted to pass too many arguments to the device kernel, or the " << std::endl << "kernel launch specifies too many threads for the kernel's register count. " << std::endl; break; // 8 case cudaErrorInvalidDeviceFunction: std::cerr << "The requested device function does not exist or is not compiled for the " << std::endl << "proper device architecture. " << std::endl; break; // 9 case cudaErrorInvalidConfiguration: std::cerr << "This indicates that a kernel launch is requesting resources that can " << std::endl << "never be satisfied by the current device. Requesting more shared memory " << std::endl << "per block than the device supports will trigger this error, as will " << std::endl << "requesting too many threads or blocks. See ::cudaDeviceProp for more " << std::endl << "device limitations. " << std::endl; break; // 10 case cudaErrorInvalidDevice: std::cerr << "This indicates that the device ordinal supplied by the user does not " << std::endl << "correspond to a valid CUDA device. " << std::endl; break; // 11 case cudaErrorInvalidValue: std::cerr << "This indicates that one or more of the parameters passed to the API call " << std::endl << "is not within an acceptable range of values. " << std::endl; break; // 12 case cudaErrorInvalidPitchValue: std::cerr << "This indicates that one or more of the pitch-related parameters passed " << std::endl << "to the API call is not within the acceptable range for pitch. " << std::endl; break; // 13 case cudaErrorInvalidSymbol: std::cerr << "This indicates that the symbol name/identifier passed to the API call " << std::endl << "is not a valid name or identifier. " << std::endl; break; // 14 case cudaErrorMapBufferObjectFailed: std::cerr << "This indicates that the buffer object could not be mapped. " << std::endl; break; // 15 case cudaErrorUnmapBufferObjectFailed: std::cerr << "This indicates that the buffer object could not be unmapped. " << std::endl; break; // 16 case cudaErrorInvalidHostPointer: std::cerr << "This indicates that at least one host pointer passed to the API call is " << std::endl << "not a valid host pointer. " << std::endl; break; // 17 case cudaErrorInvalidDevicePointer: std::cerr << "This indicates that at least one device pointer passed to the API call is " << std::endl << "not a valid device pointer. " << std::endl; break; case cudaErrorInvalidTexture: std::cerr << "This indicates that the texture passed to the API call is not a valid " << std::endl << "texture. " << std::endl; break; case cudaErrorInvalidTextureBinding: std::cerr << "This indicates that the texture binding is not valid. This occurs if you " << std::endl << "call ::cudaGetTextureAlignmentOffset() with an unbound texture. " << std::endl; break; case cudaErrorInvalidChannelDescriptor: std::cerr << "This indicates that the channel descriptor passed to the API call is not " << std::endl << "valid. This occurs if the format is not one of the formats specified by " << std::endl << "::cudaChannelFormatKind, or if one of the dimensions is invalid. " << std::endl; break; case cudaErrorInvalidMemcpyDirection: std::cerr << "This indicates that the direction of the memcpy passed to the API call is " << std::endl << "not one of the types specified by ::cudaMemcpyKind. " << std::endl; break; case cudaErrorAddressOfConstant: std::cerr << "This indicated that the user has taken the address of a constant variable, " << std::endl << "which was forbidden up until the CUDA 3.1 release. " << std::endl << "This error return is deprecated as of CUDA 3.1. Variables in constant " << std::endl << "memory may now have their address taken by the runtime via " << std::endl << "::cudaGetSymbolAddress(). " << std::endl; break; case cudaErrorTextureFetchFailed: std::cerr << "This indicated that a texture fetch was not able to be performed. " << std::endl << "This was previously used for device emulation of texture operations. " << std::endl << "This error return is deprecated as of CUDA 3.1. Device emulation mode was " << std::endl << "removed with the CUDA 3.1 release. " << std::endl; break; case cudaErrorTextureNotBound: std::cerr << "This indicated that a texture was not bound for access. " << std::endl << "This was previously used for device emulation of texture operations. " << std::endl << "This error return is deprecated as of CUDA 3.1. Device emulation mode was " << std::endl << "removed with the CUDA 3.1 release. " << std::endl; break; case cudaErrorSynchronizationError: std::cerr << "This indicated that a synchronization operation had failed. " << std::endl << "This was previously used for some device emulation functions. " << std::endl << "This error return is deprecated as of CUDA 3.1. Device emulation mode was " << std::endl << "removed with the CUDA 3.1 release. " << std::endl; break; case cudaErrorInvalidFilterSetting: std::cerr << "This indicates that a non-float texture was being accessed with linear " << std::endl << "filtering. This is not supported by CUDA. " << std::endl; break; case cudaErrorInvalidNormSetting: std::cerr << "This indicates that an attempt was made to read a non-float texture as a " << std::endl << "normalized float. This is not supported by CUDA. " << std::endl; break; case cudaErrorMixedDeviceExecution: std::cerr << "Mixing of device and device emulation code was not allowed. " << std::endl << "This error return is deprecated as of CUDA 3.1. Device emulation mode was " << std::endl << "removed with the CUDA 3.1 release. " << std::endl; break; case cudaErrorCudartUnloading: std::cerr << "This indicated an issue with calling API functions during the unload " << std::endl << "process of the CUDA runtime in prior releases. " << std::endl << "This error return is deprecated as of CUDA 3.2. " << std::endl; break; case cudaErrorUnknown: std::cerr << "This indicates that an unknown internal error has occurred. " << std::endl; break; case cudaErrorNotYetImplemented: std::cerr << "This indicates that the API call is not yet implemented. Production " << std::endl << "releases of CUDA will never return this error. " << std::endl; break; case cudaErrorMemoryValueTooLarge: std::cerr << "This indicated that an emulated device pointer exceeded the 32-bit address " << std::endl << "range. " << std::endl << "This error return is deprecated as of CUDA 3.1. Device emulation mode was " << std::endl << "removed with the CUDA 3.1 release. " << std::endl; break; case cudaErrorInvalidResourceHandle: std::cerr << "This indicates that a resource handle passed to the API call was not " << std::endl << "valid. Resource handles are opaque types like ::cudaStream_t and " << std::endl << "::cudaEvent_t. " << std::endl; break; case cudaErrorNotReady: std::cerr << "This indicates that asynchronous operations issued previously have not " << std::endl << "completed yet. This result is not actually an error, but must be indicated " << std::endl << "differently than ::cudaSuccess (which indicates completion). Calls that " << std::endl << "may return this value include ::cudaEventQuery() and ::cudaStreamQuery(). " << std::endl; break; case cudaErrorInsufficientDriver: std::cerr << "This indicates that the installed NVIDIA CUDA driver is older than the " << std::endl << "CUDA runtime library. This is not a supported configuration. Users should " << std::endl << "install an updated NVIDIA display driver to allow the application to run. " << std::endl; break; case cudaErrorSetOnActiveProcess: std::cerr << "This indicates that the user has called ::cudaSetDevice(), " << std::endl << "::cudaSetValidDevices(), ::cudaSetDeviceFlags(), " << std::endl << "::cudaD3D9SetDirect3DDevice(), ::cudaD3D10SetDirect3DDevice, " << std::endl << "::cudaD3D11SetDirect3DDevice(), * or ::cudaVDPAUSetVDPAUDevice() after " << std::endl << "initializing the CUDA runtime by calling non-device management operations " << std::endl << "(allocating memory and launching kernels are examples of non-device " << std::endl << "management operations). This error can also be returned if using " << std::endl << "runtime/driver interoperability and there is an existing ::CUcontext " << std::endl << "active on the host thread. " << std::endl; break; case cudaErrorInvalidSurface: std::cerr << "This indicates that the surface passed to the API call is not a valid " << std::endl << "surface. " << std::endl; break; case cudaErrorNoDevice: std::cerr << "This indicates that no CUDA-capable devices were detected by the installed " << std::endl << "CUDA driver. " << std::endl; break; case cudaErrorECCUncorrectable: std::cerr << "This indicates that an uncorrectable ECC error was detected during " << std::endl << "execution. " << std::endl; break; case cudaErrorSharedObjectSymbolNotFound: std::cerr << "This indicates that a link to a shared object failed to resolve. " << std::endl; break; case cudaErrorSharedObjectInitFailed: std::cerr << "This indicates that initialization of a shared object failed. " << std::endl; break; case cudaErrorUnsupportedLimit: std::cerr << "This indicates that the ::cudaLimit passed to the API call is not " << std::endl << "supported by the active device. " << std::endl; break; case cudaErrorDuplicateVariableName: std::cerr << "This indicates that multiple global or constant variables (across separate " << std::endl << "CUDA source files in the application) share the same string name. " << std::endl; break; case cudaErrorDuplicateTextureName: std::cerr << "This indicates that multiple textures (across separate CUDA source " << std::endl << "files in the application) share the same string name. " << std::endl; break; case cudaErrorDuplicateSurfaceName: std::cerr << "This indicates that multiple surfaces (across separate CUDA source " << std::endl << "files in the application) share the same string name. " << std::endl; break; case cudaErrorDevicesUnavailable: std::cerr << "This indicates that all CUDA devices are busy or unavailable at the current " << std::endl << "time. Devices are often busy/unavailable due to use of " << std::endl << "::cudaComputeModeExclusive or ::cudaComputeModeProhibited. They can also " << std::endl << "be unavailable due to memory constraints on a device that already has " << std::endl << "active CUDA work being performed. " << std::endl; break; case cudaErrorInvalidKernelImage: std::cerr << "This indicates that the device kernel image is invalid. " << std::endl; break; case cudaErrorNoKernelImageForDevice: std::cerr << "This indicates that there is no kernel image available that is suitable " << std::endl << "for the device. This can occur when a user specifies code generation " << std::endl << "options for a particular CUDA source file that do not include the " << std::endl << "corresponding device configuration. " << std::endl; break; case cudaErrorIncompatibleDriverContext: std::cerr << "This indicates that the current context is not compatible with this " << std::endl << "version of the CUDA Runtime. This can only occur if you are using CUDA " << std::endl << "Runtime/Driver interoperability and have created an existing Driver " << std::endl << "context using an older API. Please see \\ref CUDART_DRIVER " << std::endl << "\"Interactions with the CUDA Driver API\" for more information. " << std::endl; break; case cudaErrorStartupFailure: std::cerr << "This indicates an internal startup failure in the CUDA runtime. " << std::endl; break; case cudaErrorApiFailureBase: std::cerr << "Any unhandled CUDA driver error is added to this value and returned via " << std::endl << "the runtime. Production releases of CUDA should not return such errors. " << std::endl; break; default: std::cerr << "(detailed description is not available)" << std::endl; break; } throw EXIT_FAILURE; return false; throw Exceptions::CudaRuntimeError( error, file_name, line ); } std::ostream& operator << ( std::ostream& str, const dim3& d ) Loading src/TNL/Devices/Cuda.h +4 −1 Original line number Diff line number Diff line Loading @@ -185,11 +185,14 @@ class Cuda #ifdef HAVE_CUDA #define checkCudaDevice ::TNL::Devices::Cuda::checkDevice( __FILE__, __LINE__, cudaGetLastError() ) std::ostream& operator << ( std::ostream& str, const dim3& d ); #else #define checkCudaDevice ::TNL::Devices::Cuda::checkDevice() #endif #ifdef HAVE_CUDA std::ostream& operator << ( std::ostream& str, const dim3& d ); #endif } // namespace Devices } // namespace TNL Loading Loading
src/TNL/Containers/Algorithms/ArrayOperationsCuda_impl.h +4 −0 Original line number Diff line number Diff line Loading @@ -15,6 +15,7 @@ #include <TNL/tnlConfig.h> #include <TNL/Math.h> #include <TNL/Exceptions/CudaSupportMissing.h> #include <TNL/Exceptions/CudaBadAlloc.h> #include <TNL/Containers/Algorithms/ArrayOperations.h> #include <TNL/Containers/Algorithms/Reduction.h> #include <TNL/Containers/Algorithms/reduction-operations.h> Loading @@ -33,7 +34,10 @@ allocateMemory( Element*& data, checkCudaDevice; if( cudaMalloc( ( void** ) &data, ( size_t ) size * sizeof( Element ) ) != cudaSuccess ) { data = 0; throw Exceptions::CudaBadAlloc(); } return checkCudaDevice; #else throw Exceptions::CudaSupportMissing(); Loading
src/TNL/Containers/Algorithms/CudaReductionBuffer.h +3 −1 Original line number Diff line number Diff line Loading @@ -11,7 +11,9 @@ #pragma once #include <stdlib.h> #include <TNL/Devices/Cuda.h> #include <TNL/Exceptions/CudaBadAlloc.h> namespace TNL { namespace Containers { Loading @@ -35,8 +37,8 @@ class CudaReductionBuffer this->size = size; if( cudaMalloc( ( void** ) &this->data, size ) != cudaSuccess ) { std::cerr << "I am not able to allocate reduction buffer on the GPU." << std::endl; this->data = 0; throw Exceptions::CudaBadAlloc(); } return checkCudaDevice; } Loading
src/TNL/Containers/Algorithms/cuda-prefix-sum_impl.h +3 −6 Original line number Diff line number Diff line Loading @@ -11,7 +11,9 @@ #pragma once #include <iostream> #include <TNL/Devices/Cuda.h> #include <TNL/Exceptions/CudaBadAlloc.h> #include <TNL/Containers/Algorithms/reduction-operations.h> #ifdef HAVE_CUDA Loading Loading @@ -194,12 +196,7 @@ bool cudaRecursivePrefixSum( const enumPrefixSumType prefixSumType, if( cudaMalloc( ( void** ) &auxArray1, auxArraySize ) != cudaSuccess || cudaMalloc( ( void** ) &auxArray2, auxArraySize ) != cudaSuccess ) { { std::cerr << "Not enough memory on device to allocate auxilliary arrays." << std::endl; return false; } } throw Exceptions::CudaBadAlloc(); /**** * Setup block and grid size. Loading
src/TNL/Devices/Cuda.cu +2 −376 Original line number Diff line number Diff line Loading @@ -9,6 +9,7 @@ /* See Copyright Notice in tnl/Copyright */ #include <TNL/Devices/Cuda.h> #include <TNL/Exceptions/CudaRuntimeError.h> #include <TNL/Config/ConfigDescription.h> #include <TNL/Config/ParameterContainer.h> Loading Loading @@ -106,382 +107,7 @@ bool Cuda::checkDevice( const char* file_name, int line, cudaError error ) { if( error == cudaSuccess ) return true; std::cerr << "CUDA ERROR(" << error << ") at line " << line << " in " << file_name << ":" << std::endl; std::cerr << cudaGetErrorString( error ) << std::endl; std::cerr << "Detailed description is: " << std::endl; switch( error ) { // 1 case cudaErrorMissingConfiguration: std::cerr << "The device function being invoked (usually via ::cudaLaunch()) was not " << std::endl << "previously configured via the ::cudaConfigureCall() function. " << std::endl; break; // 2 case cudaErrorMemoryAllocation: std::cerr << "The API call failed because it was unable to allocate enough memory to " << std::endl << "perform the requested operation. " << std::endl; break; // 3 case cudaErrorInitializationError: std::cerr << "The API call failed because the CUDA driver and runtime could not be " << std::endl << "initialized. " << std::endl; break; // 4 case cudaErrorLaunchFailure: std::cerr << "An exception occurred on the device while executing a kernel. Common " << std::endl << "causes include dereferencing an invalid device pointer and accessing " << std::endl << "out of bounds shared memory. The device cannot be used until " << std::endl << "::cudaThreadExit() is called. All existing device memory allocations " << std::endl << "are invalid and must be reconstructed if the program is to continue " << std::endl << "using CUDA. " << std::endl; break; // 5 case cudaErrorPriorLaunchFailure: std::cerr << "This indicated that a previous kernel launch failed. This was previously " << std::endl << "used for device emulation of kernel launches. " << std::endl << "This error return is deprecated as of CUDA 3.1. Device emulation mode was " << std::endl << "removed with the CUDA 3.1 release. " << std::endl; break; // 6 case cudaErrorLaunchTimeout: std::cerr << "This indicates that the device kernel took too long to execute. This can " << std::endl << "only occur if timeouts are enabled - see the device property " << std::endl << "ref ::cudaDeviceProp::kernelExecTimeoutEnabled \"kernelExecTimeoutEnabled\" " << std::endl << "for more information. The device cannot be used until ::cudaThreadExit() " << std::endl << "is called. All existing device memory allocations are invalid and must be " << std::endl << "reconstructed if the program is to continue using CUDA. " << std::endl; break; // 7 case cudaErrorLaunchOutOfResources: std::cerr << "This indicates that a launch did not occur because it did not have " << std::endl << "appropriate resources. Although this error is similar to " << std::endl << "::cudaErrorInvalidConfiguration, this error usually indicates that the " << std::endl << "user has attempted to pass too many arguments to the device kernel, or the " << std::endl << "kernel launch specifies too many threads for the kernel's register count. " << std::endl; break; // 8 case cudaErrorInvalidDeviceFunction: std::cerr << "The requested device function does not exist or is not compiled for the " << std::endl << "proper device architecture. " << std::endl; break; // 9 case cudaErrorInvalidConfiguration: std::cerr << "This indicates that a kernel launch is requesting resources that can " << std::endl << "never be satisfied by the current device. Requesting more shared memory " << std::endl << "per block than the device supports will trigger this error, as will " << std::endl << "requesting too many threads or blocks. See ::cudaDeviceProp for more " << std::endl << "device limitations. " << std::endl; break; // 10 case cudaErrorInvalidDevice: std::cerr << "This indicates that the device ordinal supplied by the user does not " << std::endl << "correspond to a valid CUDA device. " << std::endl; break; // 11 case cudaErrorInvalidValue: std::cerr << "This indicates that one or more of the parameters passed to the API call " << std::endl << "is not within an acceptable range of values. " << std::endl; break; // 12 case cudaErrorInvalidPitchValue: std::cerr << "This indicates that one or more of the pitch-related parameters passed " << std::endl << "to the API call is not within the acceptable range for pitch. " << std::endl; break; // 13 case cudaErrorInvalidSymbol: std::cerr << "This indicates that the symbol name/identifier passed to the API call " << std::endl << "is not a valid name or identifier. " << std::endl; break; // 14 case cudaErrorMapBufferObjectFailed: std::cerr << "This indicates that the buffer object could not be mapped. " << std::endl; break; // 15 case cudaErrorUnmapBufferObjectFailed: std::cerr << "This indicates that the buffer object could not be unmapped. " << std::endl; break; // 16 case cudaErrorInvalidHostPointer: std::cerr << "This indicates that at least one host pointer passed to the API call is " << std::endl << "not a valid host pointer. " << std::endl; break; // 17 case cudaErrorInvalidDevicePointer: std::cerr << "This indicates that at least one device pointer passed to the API call is " << std::endl << "not a valid device pointer. " << std::endl; break; case cudaErrorInvalidTexture: std::cerr << "This indicates that the texture passed to the API call is not a valid " << std::endl << "texture. " << std::endl; break; case cudaErrorInvalidTextureBinding: std::cerr << "This indicates that the texture binding is not valid. This occurs if you " << std::endl << "call ::cudaGetTextureAlignmentOffset() with an unbound texture. " << std::endl; break; case cudaErrorInvalidChannelDescriptor: std::cerr << "This indicates that the channel descriptor passed to the API call is not " << std::endl << "valid. This occurs if the format is not one of the formats specified by " << std::endl << "::cudaChannelFormatKind, or if one of the dimensions is invalid. " << std::endl; break; case cudaErrorInvalidMemcpyDirection: std::cerr << "This indicates that the direction of the memcpy passed to the API call is " << std::endl << "not one of the types specified by ::cudaMemcpyKind. " << std::endl; break; case cudaErrorAddressOfConstant: std::cerr << "This indicated that the user has taken the address of a constant variable, " << std::endl << "which was forbidden up until the CUDA 3.1 release. " << std::endl << "This error return is deprecated as of CUDA 3.1. Variables in constant " << std::endl << "memory may now have their address taken by the runtime via " << std::endl << "::cudaGetSymbolAddress(). " << std::endl; break; case cudaErrorTextureFetchFailed: std::cerr << "This indicated that a texture fetch was not able to be performed. " << std::endl << "This was previously used for device emulation of texture operations. " << std::endl << "This error return is deprecated as of CUDA 3.1. Device emulation mode was " << std::endl << "removed with the CUDA 3.1 release. " << std::endl; break; case cudaErrorTextureNotBound: std::cerr << "This indicated that a texture was not bound for access. " << std::endl << "This was previously used for device emulation of texture operations. " << std::endl << "This error return is deprecated as of CUDA 3.1. Device emulation mode was " << std::endl << "removed with the CUDA 3.1 release. " << std::endl; break; case cudaErrorSynchronizationError: std::cerr << "This indicated that a synchronization operation had failed. " << std::endl << "This was previously used for some device emulation functions. " << std::endl << "This error return is deprecated as of CUDA 3.1. Device emulation mode was " << std::endl << "removed with the CUDA 3.1 release. " << std::endl; break; case cudaErrorInvalidFilterSetting: std::cerr << "This indicates that a non-float texture was being accessed with linear " << std::endl << "filtering. This is not supported by CUDA. " << std::endl; break; case cudaErrorInvalidNormSetting: std::cerr << "This indicates that an attempt was made to read a non-float texture as a " << std::endl << "normalized float. This is not supported by CUDA. " << std::endl; break; case cudaErrorMixedDeviceExecution: std::cerr << "Mixing of device and device emulation code was not allowed. " << std::endl << "This error return is deprecated as of CUDA 3.1. Device emulation mode was " << std::endl << "removed with the CUDA 3.1 release. " << std::endl; break; case cudaErrorCudartUnloading: std::cerr << "This indicated an issue with calling API functions during the unload " << std::endl << "process of the CUDA runtime in prior releases. " << std::endl << "This error return is deprecated as of CUDA 3.2. " << std::endl; break; case cudaErrorUnknown: std::cerr << "This indicates that an unknown internal error has occurred. " << std::endl; break; case cudaErrorNotYetImplemented: std::cerr << "This indicates that the API call is not yet implemented. Production " << std::endl << "releases of CUDA will never return this error. " << std::endl; break; case cudaErrorMemoryValueTooLarge: std::cerr << "This indicated that an emulated device pointer exceeded the 32-bit address " << std::endl << "range. " << std::endl << "This error return is deprecated as of CUDA 3.1. Device emulation mode was " << std::endl << "removed with the CUDA 3.1 release. " << std::endl; break; case cudaErrorInvalidResourceHandle: std::cerr << "This indicates that a resource handle passed to the API call was not " << std::endl << "valid. Resource handles are opaque types like ::cudaStream_t and " << std::endl << "::cudaEvent_t. " << std::endl; break; case cudaErrorNotReady: std::cerr << "This indicates that asynchronous operations issued previously have not " << std::endl << "completed yet. This result is not actually an error, but must be indicated " << std::endl << "differently than ::cudaSuccess (which indicates completion). Calls that " << std::endl << "may return this value include ::cudaEventQuery() and ::cudaStreamQuery(). " << std::endl; break; case cudaErrorInsufficientDriver: std::cerr << "This indicates that the installed NVIDIA CUDA driver is older than the " << std::endl << "CUDA runtime library. This is not a supported configuration. Users should " << std::endl << "install an updated NVIDIA display driver to allow the application to run. " << std::endl; break; case cudaErrorSetOnActiveProcess: std::cerr << "This indicates that the user has called ::cudaSetDevice(), " << std::endl << "::cudaSetValidDevices(), ::cudaSetDeviceFlags(), " << std::endl << "::cudaD3D9SetDirect3DDevice(), ::cudaD3D10SetDirect3DDevice, " << std::endl << "::cudaD3D11SetDirect3DDevice(), * or ::cudaVDPAUSetVDPAUDevice() after " << std::endl << "initializing the CUDA runtime by calling non-device management operations " << std::endl << "(allocating memory and launching kernels are examples of non-device " << std::endl << "management operations). This error can also be returned if using " << std::endl << "runtime/driver interoperability and there is an existing ::CUcontext " << std::endl << "active on the host thread. " << std::endl; break; case cudaErrorInvalidSurface: std::cerr << "This indicates that the surface passed to the API call is not a valid " << std::endl << "surface. " << std::endl; break; case cudaErrorNoDevice: std::cerr << "This indicates that no CUDA-capable devices were detected by the installed " << std::endl << "CUDA driver. " << std::endl; break; case cudaErrorECCUncorrectable: std::cerr << "This indicates that an uncorrectable ECC error was detected during " << std::endl << "execution. " << std::endl; break; case cudaErrorSharedObjectSymbolNotFound: std::cerr << "This indicates that a link to a shared object failed to resolve. " << std::endl; break; case cudaErrorSharedObjectInitFailed: std::cerr << "This indicates that initialization of a shared object failed. " << std::endl; break; case cudaErrorUnsupportedLimit: std::cerr << "This indicates that the ::cudaLimit passed to the API call is not " << std::endl << "supported by the active device. " << std::endl; break; case cudaErrorDuplicateVariableName: std::cerr << "This indicates that multiple global or constant variables (across separate " << std::endl << "CUDA source files in the application) share the same string name. " << std::endl; break; case cudaErrorDuplicateTextureName: std::cerr << "This indicates that multiple textures (across separate CUDA source " << std::endl << "files in the application) share the same string name. " << std::endl; break; case cudaErrorDuplicateSurfaceName: std::cerr << "This indicates that multiple surfaces (across separate CUDA source " << std::endl << "files in the application) share the same string name. " << std::endl; break; case cudaErrorDevicesUnavailable: std::cerr << "This indicates that all CUDA devices are busy or unavailable at the current " << std::endl << "time. Devices are often busy/unavailable due to use of " << std::endl << "::cudaComputeModeExclusive or ::cudaComputeModeProhibited. They can also " << std::endl << "be unavailable due to memory constraints on a device that already has " << std::endl << "active CUDA work being performed. " << std::endl; break; case cudaErrorInvalidKernelImage: std::cerr << "This indicates that the device kernel image is invalid. " << std::endl; break; case cudaErrorNoKernelImageForDevice: std::cerr << "This indicates that there is no kernel image available that is suitable " << std::endl << "for the device. This can occur when a user specifies code generation " << std::endl << "options for a particular CUDA source file that do not include the " << std::endl << "corresponding device configuration. " << std::endl; break; case cudaErrorIncompatibleDriverContext: std::cerr << "This indicates that the current context is not compatible with this " << std::endl << "version of the CUDA Runtime. This can only occur if you are using CUDA " << std::endl << "Runtime/Driver interoperability and have created an existing Driver " << std::endl << "context using an older API. Please see \\ref CUDART_DRIVER " << std::endl << "\"Interactions with the CUDA Driver API\" for more information. " << std::endl; break; case cudaErrorStartupFailure: std::cerr << "This indicates an internal startup failure in the CUDA runtime. " << std::endl; break; case cudaErrorApiFailureBase: std::cerr << "Any unhandled CUDA driver error is added to this value and returned via " << std::endl << "the runtime. Production releases of CUDA should not return such errors. " << std::endl; break; default: std::cerr << "(detailed description is not available)" << std::endl; break; } throw EXIT_FAILURE; return false; throw Exceptions::CudaRuntimeError( error, file_name, line ); } std::ostream& operator << ( std::ostream& str, const dim3& d ) Loading
src/TNL/Devices/Cuda.h +4 −1 Original line number Diff line number Diff line Loading @@ -185,11 +185,14 @@ class Cuda #ifdef HAVE_CUDA #define checkCudaDevice ::TNL::Devices::Cuda::checkDevice( __FILE__, __LINE__, cudaGetLastError() ) std::ostream& operator << ( std::ostream& str, const dim3& d ); #else #define checkCudaDevice ::TNL::Devices::Cuda::checkDevice() #endif #ifdef HAVE_CUDA std::ostream& operator << ( std::ostream& str, const dim3& d ); #endif } // namespace Devices } // namespace TNL Loading
