56fd68e7dd
CCCL (CUDA C++ Core Libraries) provides: - CUB: device/block/warp-level GPU primitives (reduce, scan, sort, topk) - Thrust: high-level parallel algorithms (transform_reduce, sort, scan) - libcudacxx: CUDA C++ standard library (atomics, barriers, memory) - cudax: experimental features (memory resources, allocators) - Tuning policies: per-SM hardware-specific algorithm parameters Competition optimization vectors mapped to CCCL: - Output TPS (83% weight): warp_reduce, block_reduce, device_topk - Input TPS (14% weight): device_scan, block_load, prefetch - Cache TPS (3% weight): prefix caching strategy patterns - Memory (0.9 util): pooled/cached/buddy allocators Source: https://github.com/NVIDIA/cccl (shallow clone, HEAD only) License: Apache-2.0
147 lines
6.3 KiB
C++
147 lines
6.3 KiB
C++
// SPDX-FileCopyrightText: Copyright (c) 2008-2013, NVIDIA Corporation. All rights reserved.
|
|
// SPDX-License-Identifier: Apache-2.0
|
|
|
|
/*! \file thrust/system/cuda/error.h
|
|
* \brief CUDA-specific error reporting
|
|
*/
|
|
|
|
#pragma once
|
|
|
|
#include <thrust/detail/config.h>
|
|
|
|
#if defined(_CCCL_IMPLICIT_SYSTEM_HEADER_GCC)
|
|
# pragma GCC system_header
|
|
#elif defined(_CCCL_IMPLICIT_SYSTEM_HEADER_CLANG)
|
|
# pragma clang system_header
|
|
#elif defined(_CCCL_IMPLICIT_SYSTEM_HEADER_MSVC)
|
|
# pragma system_header
|
|
#endif // no system header
|
|
#include <thrust/detail/type_traits.h>
|
|
#include <thrust/system/error_code.h>
|
|
|
|
THRUST_NAMESPACE_BEGIN
|
|
|
|
namespace system
|
|
{
|
|
// To construct an error_code after a CUDA Runtime error:
|
|
//
|
|
// error_code(::cudaGetLastError(), cuda_category())
|
|
|
|
// XXX N3000 prefers enum class errc { ... }
|
|
/*! Namespace for CUDA Runtime errors.
|
|
*/
|
|
namespace cuda::errc
|
|
{
|
|
/*! \p errc_t enumerates the kinds of CUDA Runtime errors.
|
|
*/
|
|
enum errc_t
|
|
{
|
|
// from cuda/include/driver_types.h
|
|
// mirror their order
|
|
success = cudaSuccess,
|
|
missing_configuration = cudaErrorMissingConfiguration,
|
|
memory_allocation = cudaErrorMemoryAllocation,
|
|
initialization_error = cudaErrorInitializationError,
|
|
launch_failure = cudaErrorLaunchFailure,
|
|
launch_timeout = cudaErrorLaunchTimeout,
|
|
launch_out_of_resources = cudaErrorLaunchOutOfResources,
|
|
invalid_device_function = cudaErrorInvalidDeviceFunction,
|
|
invalid_configuration = cudaErrorInvalidConfiguration,
|
|
invalid_device = cudaErrorInvalidDevice,
|
|
invalid_value = cudaErrorInvalidValue,
|
|
invalid_pitch_value = cudaErrorInvalidPitchValue,
|
|
invalid_symbol = cudaErrorInvalidSymbol,
|
|
map_buffer_object_failed = cudaErrorMapBufferObjectFailed,
|
|
unmap_buffer_object_failed = cudaErrorUnmapBufferObjectFailed,
|
|
invalid_texture = cudaErrorInvalidTexture,
|
|
invalid_texture_binding = cudaErrorInvalidTextureBinding,
|
|
invalid_channel_descriptor = cudaErrorInvalidChannelDescriptor,
|
|
invalid_memcpy_direction = cudaErrorInvalidMemcpyDirection,
|
|
invalid_filter_setting = cudaErrorInvalidFilterSetting,
|
|
invalid_norm_setting = cudaErrorInvalidNormSetting,
|
|
cuda_runtime_unloading = cudaErrorCudartUnloading,
|
|
unknown = cudaErrorUnknown,
|
|
invalid_resource_handle = cudaErrorInvalidResourceHandle,
|
|
not_ready = cudaErrorNotReady,
|
|
insufficient_driver = cudaErrorInsufficientDriver,
|
|
set_on_active_process_error = cudaErrorSetOnActiveProcess,
|
|
no_device = cudaErrorNoDevice,
|
|
ecc_uncorrectable = cudaErrorECCUncorrectable,
|
|
shared_object_symbol_not_found = cudaErrorSharedObjectSymbolNotFound,
|
|
shared_object_init_failed = cudaErrorSharedObjectInitFailed,
|
|
unsupported_limit = cudaErrorUnsupportedLimit,
|
|
duplicate_variable_name = cudaErrorDuplicateVariableName,
|
|
duplicate_texture_name = cudaErrorDuplicateTextureName,
|
|
duplicate_surface_name = cudaErrorDuplicateSurfaceName,
|
|
devices_unavailable = cudaErrorDevicesUnavailable,
|
|
invalid_kernel_image = cudaErrorInvalidKernelImage,
|
|
no_kernel_image_for_device = cudaErrorNoKernelImageForDevice,
|
|
incompatible_driver_context = cudaErrorIncompatibleDriverContext,
|
|
peer_access_already_enabled = cudaErrorPeerAccessAlreadyEnabled,
|
|
peer_access_not_enabled = cudaErrorPeerAccessNotEnabled,
|
|
device_already_in_use = cudaErrorDeviceAlreadyInUse,
|
|
profiler_disabled = cudaErrorProfilerDisabled,
|
|
assert_triggered = cudaErrorAssert,
|
|
too_many_peers = cudaErrorTooManyPeers,
|
|
host_memory_already_registered = cudaErrorHostMemoryAlreadyRegistered,
|
|
host_memory_not_registered = cudaErrorHostMemoryNotRegistered,
|
|
operating_system_error = cudaErrorOperatingSystem,
|
|
peer_access_unsupported = cudaErrorPeerAccessUnsupported,
|
|
launch_max_depth_exceeded = cudaErrorLaunchMaxDepthExceeded,
|
|
launch_file_scoped_texture_used = cudaErrorLaunchFileScopedTex,
|
|
launch_file_scoped_surface_used = cudaErrorLaunchFileScopedSurf,
|
|
sync_depth_exceeded = cudaErrorSyncDepthExceeded,
|
|
attempted_operation_not_permitted = cudaErrorNotPermitted,
|
|
attempted_operation_not_supported = cudaErrorNotSupported,
|
|
startup_failure = cudaErrorStartupFailure
|
|
}; // end errc_t
|
|
} // namespace cuda::errc
|
|
|
|
/*! \return A reference to an object of a type derived from class \p thrust::error_category.
|
|
* \note The object's \p equivalent virtual functions shall behave as specified
|
|
* for the class \p thrust::error_category. The object's \p name virtual function shall
|
|
* return a pointer to the string <tt>"cuda"</tt>. The object's
|
|
* \p default_error_condition virtual function shall behave as follows:
|
|
*
|
|
* If the argument <tt>ev</tt> corresponds to a CUDA error value, the function
|
|
* shall return <tt>error_condition(ev,cuda_category())</tt>.
|
|
* Otherwise, the function shall return <tt>system_category.default_error_condition(ev)</tt>.
|
|
*/
|
|
inline const error_category& cuda_category();
|
|
|
|
// XXX N3000 prefers is_error_code_enum<cuda::errc>
|
|
|
|
/*! Specialization of \p is_error_code_enum for \p cuda::errc::errc_t
|
|
*/
|
|
template <>
|
|
struct is_error_code_enum<cuda::errc::errc_t> : thrust::detail::true_type
|
|
{};
|
|
|
|
// XXX replace cuda::errc::errc_t with cuda::errc upon c++0x
|
|
/*! \return <tt>error_code(static_cast<int>(e), cuda::error_category())</tt>
|
|
*/
|
|
inline error_code make_error_code(cuda::errc::errc_t e);
|
|
|
|
// XXX replace cuda::errc::errc_t with cuda::errc upon c++0x
|
|
/*! \return <tt>error_condition(static_cast<int>(e), cuda::error_category())</tt>.
|
|
*/
|
|
inline error_condition make_error_condition(cuda::errc::errc_t e);
|
|
} // namespace system
|
|
|
|
namespace cuda_cub
|
|
{
|
|
namespace errc = system::cuda::errc;
|
|
} // namespace cuda_cub
|
|
|
|
namespace cuda
|
|
{
|
|
// XXX replace with using system::cuda_errc upon c++0x
|
|
namespace errc = system::cuda::errc;
|
|
} // namespace cuda
|
|
|
|
using system::cuda_category;
|
|
|
|
THRUST_NAMESPACE_END
|
|
|
|
#include <thrust/system/cuda/detail/error.inl>
|