Files
project_6/cccl_upstream/libcudacxx/include/cuda/std/numbers
EngineX CI 56fd68e7dd [INFRA] Import NVIDIA/CCCL upstream as optimization reference library
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
2026-07-30 09:35:51 +00:00

287 lines
13 KiB
Plaintext

//===----------------------------------------------------------------------===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
// SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES.
//
//===----------------------------------------------------------------------===//
#ifndef _CUDA_STD_NUMBERS
#define _CUDA_STD_NUMBERS
#include <cuda/std/detail/__config>
#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 <cuda/std/__floating_point/cuda_fp_types.h>
#include <cuda/std/__floating_point/storage.h>
#include <cuda/std/__type_traits/always_false.h>
#include <cuda/std/version>
#include <cuda/std/__cccl/prologue.h>
_CCCL_DIAG_PUSH
_CCCL_DIAG_SUPPRESS_NVHPC(inexact_fp_conversion)
_CCCL_BEGIN_NAMESPACE_CUDA_STD
template <class _Tp>
struct __numbers
{
static_assert(__always_false_v<_Tp>,
"[math.constants] A program that instantiates a primary template of a mathematical constant variable "
"template is ill-formed.");
};
#define _CCCL_STD_NUMBERS_SPECIALATION_IMPL(_TYPE, _SUFFIX) \
template <> \
struct __numbers<_TYPE> \
{ \
[[nodiscard]] _CCCL_API static constexpr _TYPE __e() noexcept \
{ \
return 0x1.5bf0a8b1457695355fb8ac404e7ap+1##_SUFFIX; \
} \
[[nodiscard]] _CCCL_API static constexpr _TYPE __log2e() noexcept \
{ \
return 0x1.71547652b82fe1777d0ffda0d23ap+0##_SUFFIX; \
} \
[[nodiscard]] _CCCL_API static constexpr _TYPE __log10e() noexcept \
{ \
return 0x1.bcb7b1526e50e32a6ab7555f5a68p-2##_SUFFIX; \
} \
[[nodiscard]] _CCCL_API static constexpr _TYPE __pi() noexcept \
{ \
return 0x1.921fb54442d18469898cc51701b8p+1##_SUFFIX; \
} \
[[nodiscard]] _CCCL_API static constexpr _TYPE __inv_pi() noexcept \
{ \
return 0x1.45f306dc9c882a53f84eafa3ea6ap-2##_SUFFIX; \
} \
[[nodiscard]] _CCCL_API static constexpr _TYPE __inv_sqrtpi() noexcept \
{ \
return 0x1.20dd750429b6d11ae3a914fed7fep-1##_SUFFIX; \
} \
[[nodiscard]] _CCCL_API static constexpr _TYPE __ln2() noexcept \
{ \
return 0x1.62e42fefa39ef35793c7673007e6p-1##_SUFFIX; \
} \
[[nodiscard]] _CCCL_API static constexpr _TYPE __ln10() noexcept \
{ \
return 0x1.26bb1bbb5551582dd4adac5705a6p+1##_SUFFIX; \
} \
[[nodiscard]] _CCCL_API static constexpr _TYPE __sqrt2() noexcept \
{ \
return 0x1.6a09e667f3bcc908b2fb1366ea95p+0##_SUFFIX; \
} \
[[nodiscard]] _CCCL_API static constexpr _TYPE __sqrt3() noexcept \
{ \
return 0x1.bb67ae8584caa73b25742d7078b8p+0##_SUFFIX; \
} \
[[nodiscard]] _CCCL_API static constexpr _TYPE __inv_sqrt3() noexcept \
{ \
return 0x1.279a74590331c4d218f81e4afb25p-1##_SUFFIX; \
} \
[[nodiscard]] _CCCL_API static constexpr _TYPE __egamma() noexcept \
{ \
return 0x1.2788cfc6fb618f49a37c7f0202a6p-1##_SUFFIX; \
} \
[[nodiscard]] _CCCL_API static constexpr _TYPE __phi() noexcept \
{ \
return 0x1.9e3779b97f4a7c15f39cc0605ceep+0##_SUFFIX; \
} \
};
_CCCL_BEGIN_NV_DIAG_SUPPRESS(1046) // floating-point value cannot be represented exactly
_CCCL_STD_NUMBERS_SPECIALATION_IMPL(float, f);
_CCCL_STD_NUMBERS_SPECIALATION_IMPL(double, );
#if _CCCL_HAS_LONG_DOUBLE()
_CCCL_STD_NUMBERS_SPECIALATION_IMPL(long double, l);
#endif // _CCCL_HAS_LONG_DOUBLE()
#if _CCCL_HAS_FLOAT128()
_CCCL_STD_NUMBERS_SPECIALATION_IMPL(__float128, q);
#endif // _CCCL_HAS_FLOAT128()
_CCCL_END_NV_DIAG_SUPPRESS()
#undef _CCCL_STD_NUMBERS_SPECIALATION_IMPL
#if _CCCL_HAS_NVFP16()
template <>
struct __numbers<::__half>
{
[[nodiscard]] _CCCL_API static constexpr ::__half __e() noexcept
{
return ::cuda::std::__fp_from_storage<::__half>(__fp_storage_of_t<::__half>{0x4170u});
}
[[nodiscard]] _CCCL_API static constexpr ::__half __log2e() noexcept
{
return ::cuda::std::__fp_from_storage<::__half>(__fp_storage_of_t<::__half>{0x3dc5u});
}
[[nodiscard]] _CCCL_API static constexpr ::__half __log10e() noexcept
{
return ::cuda::std::__fp_from_storage<::__half>(__fp_storage_of_t<::__half>{0x36f3u});
}
[[nodiscard]] _CCCL_API static constexpr ::__half __pi() noexcept
{
return ::cuda::std::__fp_from_storage<::__half>(__fp_storage_of_t<::__half>{0x4248u});
}
[[nodiscard]] _CCCL_API static constexpr ::__half __inv_pi() noexcept
{
return ::cuda::std::__fp_from_storage<::__half>(__fp_storage_of_t<::__half>{0x3518u});
}
[[nodiscard]] _CCCL_API static constexpr ::__half __inv_sqrtpi() noexcept
{
return ::cuda::std::__fp_from_storage<::__half>(__fp_storage_of_t<::__half>{0x3883u});
}
[[nodiscard]] _CCCL_API static constexpr ::__half __ln2() noexcept
{
return ::cuda::std::__fp_from_storage<::__half>(__fp_storage_of_t<::__half>{0x398cu});
}
[[nodiscard]] _CCCL_API static constexpr ::__half __ln10() noexcept
{
return ::cuda::std::__fp_from_storage<::__half>(__fp_storage_of_t<::__half>{0x409bu});
}
[[nodiscard]] _CCCL_API static constexpr ::__half __sqrt2() noexcept
{
return ::cuda::std::__fp_from_storage<::__half>(__fp_storage_of_t<::__half>{0x3da8u});
}
[[nodiscard]] _CCCL_API static constexpr ::__half __sqrt3() noexcept
{
return ::cuda::std::__fp_from_storage<::__half>(__fp_storage_of_t<::__half>{0x3eeeu});
}
[[nodiscard]] _CCCL_API static constexpr ::__half __inv_sqrt3() noexcept
{
return ::cuda::std::__fp_from_storage<::__half>(__fp_storage_of_t<::__half>{0x389eu});
}
[[nodiscard]] _CCCL_API static constexpr ::__half __egamma() noexcept
{
return ::cuda::std::__fp_from_storage<::__half>(__fp_storage_of_t<::__half>{0x389eu});
}
[[nodiscard]] _CCCL_API static constexpr ::__half __phi() noexcept
{
return ::cuda::std::__fp_from_storage<::__half>(__fp_storage_of_t<::__half>{0x3e79u});
}
};
#endif // _CCCL_HAS_NVFP16()
#if _CCCL_HAS_NVBF16()
template <>
struct __numbers<__nv_bfloat16>
{
[[nodiscard]] _CCCL_API static constexpr ::__nv_bfloat16 __e() noexcept
{
return ::cuda::std::__fp_from_storage<::__nv_bfloat16>(__fp_storage_of_t<::__nv_bfloat16>{0x402eu});
}
[[nodiscard]] _CCCL_API static constexpr ::__nv_bfloat16 __log2e() noexcept
{
return ::cuda::std::__fp_from_storage<::__nv_bfloat16>(__fp_storage_of_t<::__nv_bfloat16>{0x3fb9u});
}
[[nodiscard]] _CCCL_API static constexpr ::__nv_bfloat16 __log10e() noexcept
{
return ::cuda::std::__fp_from_storage<::__nv_bfloat16>(__fp_storage_of_t<::__nv_bfloat16>{0x3edeu});
}
[[nodiscard]] _CCCL_API static constexpr ::__nv_bfloat16 __pi() noexcept
{
return ::cuda::std::__fp_from_storage<::__nv_bfloat16>(__fp_storage_of_t<::__nv_bfloat16>{0x4049u});
}
[[nodiscard]] _CCCL_API static constexpr ::__nv_bfloat16 __inv_pi() noexcept
{
return ::cuda::std::__fp_from_storage<::__nv_bfloat16>(__fp_storage_of_t<::__nv_bfloat16>{0x3ea3u});
}
[[nodiscard]] _CCCL_API static constexpr ::__nv_bfloat16 __inv_sqrtpi() noexcept
{
return ::cuda::std::__fp_from_storage<::__nv_bfloat16>(__fp_storage_of_t<::__nv_bfloat16>{0x3f10u});
}
[[nodiscard]] _CCCL_API static constexpr ::__nv_bfloat16 __ln2() noexcept
{
return ::cuda::std::__fp_from_storage<::__nv_bfloat16>(__fp_storage_of_t<::__nv_bfloat16>{0x3f31u});
}
[[nodiscard]] _CCCL_API static constexpr ::__nv_bfloat16 __ln10() noexcept
{
return ::cuda::std::__fp_from_storage<::__nv_bfloat16>(__fp_storage_of_t<::__nv_bfloat16>{0x4013u});
}
[[nodiscard]] _CCCL_API static constexpr ::__nv_bfloat16 __sqrt2() noexcept
{
return ::cuda::std::__fp_from_storage<::__nv_bfloat16>(__fp_storage_of_t<::__nv_bfloat16>{0x3fb5u});
}
[[nodiscard]] _CCCL_API static constexpr ::__nv_bfloat16 __sqrt3() noexcept
{
return ::cuda::std::__fp_from_storage<::__nv_bfloat16>(__fp_storage_of_t<::__nv_bfloat16>{0x3fdeu});
}
[[nodiscard]] _CCCL_API static constexpr ::__nv_bfloat16 __inv_sqrt3() noexcept
{
return ::cuda::std::__fp_from_storage<::__nv_bfloat16>(__fp_storage_of_t<::__nv_bfloat16>{0x3f14u});
}
[[nodiscard]] _CCCL_API static constexpr ::__nv_bfloat16 __egamma() noexcept
{
return ::cuda::std::__fp_from_storage<::__nv_bfloat16>(__fp_storage_of_t<::__nv_bfloat16>{0x3f14u});
}
[[nodiscard]] _CCCL_API static constexpr ::__nv_bfloat16 __phi() noexcept
{
return ::cuda::std::__fp_from_storage<::__nv_bfloat16>(__fp_storage_of_t<::__nv_bfloat16>{0x3fcfu});
}
};
#endif // _CCCL_HAS_NVBF16()
namespace numbers
{
#if !_CCCL_OS(WINDOWS)
template <class _Tp>
_CCCL_GLOBAL_CONSTANT auto e_v = __numbers<_Tp>::__e();
template <class _Tp>
_CCCL_GLOBAL_CONSTANT auto log2e_v = __numbers<_Tp>::__log2e();
template <class _Tp>
_CCCL_GLOBAL_CONSTANT auto log10e_v = __numbers<_Tp>::__log10e();
template <class _Tp>
_CCCL_GLOBAL_CONSTANT auto pi_v = __numbers<_Tp>::__pi();
template <class _Tp>
_CCCL_GLOBAL_CONSTANT auto inv_pi_v = __numbers<_Tp>::__inv_pi();
template <class _Tp>
_CCCL_GLOBAL_CONSTANT auto inv_sqrtpi_v = __numbers<_Tp>::__inv_sqrtpi();
template <class _Tp>
_CCCL_GLOBAL_CONSTANT auto ln2_v = __numbers<_Tp>::__ln2();
template <class _Tp>
_CCCL_GLOBAL_CONSTANT auto ln10_v = __numbers<_Tp>::__ln10();
template <class _Tp>
_CCCL_GLOBAL_CONSTANT auto sqrt2_v = __numbers<_Tp>::__sqrt2();
template <class _Tp>
_CCCL_GLOBAL_CONSTANT auto sqrt3_v = __numbers<_Tp>::__sqrt3();
template <class _Tp>
_CCCL_GLOBAL_CONSTANT auto inv_sqrt3_v = __numbers<_Tp>::__inv_sqrt3();
template <class _Tp>
_CCCL_GLOBAL_CONSTANT auto egamma_v = __numbers<_Tp>::__egamma();
template <class _Tp>
_CCCL_GLOBAL_CONSTANT auto phi_v = __numbers<_Tp>::__phi();
_CCCL_GLOBAL_CONSTANT auto e = __numbers<double>::__e();
_CCCL_GLOBAL_CONSTANT auto log2e = __numbers<double>::__log2e();
_CCCL_GLOBAL_CONSTANT auto log10e = __numbers<double>::__log10e();
_CCCL_GLOBAL_CONSTANT auto pi = __numbers<double>::__pi();
_CCCL_GLOBAL_CONSTANT auto inv_pi = __numbers<double>::__inv_pi();
_CCCL_GLOBAL_CONSTANT auto inv_sqrtpi = __numbers<double>::__inv_sqrtpi();
_CCCL_GLOBAL_CONSTANT auto ln2 = __numbers<double>::__ln2();
_CCCL_GLOBAL_CONSTANT auto ln10 = __numbers<double>::__ln10();
_CCCL_GLOBAL_CONSTANT auto sqrt2 = __numbers<double>::__sqrt2();
_CCCL_GLOBAL_CONSTANT auto sqrt3 = __numbers<double>::__sqrt3();
_CCCL_GLOBAL_CONSTANT auto inv_sqrt3 = __numbers<double>::__inv_sqrt3();
_CCCL_GLOBAL_CONSTANT auto egamma = __numbers<double>::__egamma();
_CCCL_GLOBAL_CONSTANT auto phi = __numbers<double>::__phi();
#endif // !_CCCL_OS(WINDOWS)
} // namespace numbers
_CCCL_END_NAMESPACE_CUDA_STD
_CCCL_DIAG_POP
#include <cuda/std/__cccl/epilogue.h>
#endif // _CUDA_STD_NUMBERS