/home/docs/checkouts/readthedocs.org/user_builds/advanced-micro-devices-composable-kernel/checkouts/develop/include/ck/utility/math_v2.hpp Source File

/home/docs/checkouts/readthedocs.org/user_builds/advanced-micro-devices-composable-kernel/checkouts/develop/include/ck/utility/math_v2.hpp Source File#

Composable Kernel: /home/docs/checkouts/readthedocs.org/user_builds/advanced-micro-devices-composable-kernel/checkouts/develop/include/ck/utility/math_v2.hpp Source File
Go to the documentation of this file.
 // SPDX-License-Identifier: MIT
 // Copyright (c) 2018-2025, Advanced Micro Devices, Inc. All rights reserved.
  
 #pragma once
  
 #ifndef __HIP_DEVICE_COMPILE__
 #include <cmath>
 #endif
  
 #include "ck/utility/data_type.hpp"
 #include "ck/utility/type.hpp"
 #include "ck/utility/type_convert.hpp"
  
 namespace ck {
 namespace math {
  
 #if CK_WORKAROUND_SWDEV_383542
 extern "C" __device__ float __ocml_native_recip_f32(float);
 #endif
  
 // math functions for the host,  some are implemented by calling C++ std functions
 #if !defined(__HIPCC_RTC__) || !defined(CK_CODE_GEN_RTC)
 static inline __host__ float abs(float x) { return std::abs(x); };
  
 static inline __host__ double abs(double x) { return std::abs(x); };
  
 static inline __host__ int8_t abs(int8_t x)
 {
     int8_t sgn = x >> (8 - 1);
  
     return (x ^ sgn) - sgn;
 };
  
 static inline __host__ int32_t abs(int32_t x)
 {
     int32_t sgn = x >> (32 - 1);
  
     return (x ^ sgn) - sgn;
 };
  
 static inline __host__ half_t abs(half_t x)
 {
     uint16_t xx = ck::bit_cast<uint16_t>(x);
  
     uint16_t abs_xx = xx & 0x7fff;
  
     half_t abs_x = ck::bit_cast<half_t>(abs_xx);
  
     return abs_x;
 };
  
 #ifdef CK_EXPERIMENTAL_BIT_INT_EXTENSION_INT4
 static inline __host__ int4_t abs(int4_t x)
 {
     int4_t sgn = x >> (4 - 1);
     return (x ^ sgn) - sgn;
 }
 #endif
  
 static inline __host__ bool isnan(float x) { return std::isnan(x); };
  
 static inline __host__ bool isnan(double x) { return std::isnan(x); };
  
 static inline __host__ bool isnan(int8_t x)
 {
     (void)x;
     return false;
 };
  
 static inline __host__ bool isnan(int32_t x)
 {
     (void)x;
     return false;
 };
  
 static inline __host__ bool isnan(half_t x)
 {
     uint16_t xx = ck::bit_cast<uint16_t>(x);
  
     return (xx & 0x7FFF) > 0x7C00;
 };
  
 static inline __host__ bool isnan(f8_t x) { return ck::fp8_is_nan(x); };
  
 #ifdef CK_EXPERIMENTAL_BIT_INT_EXTENSION_INT4
 static inline __host__ bool isnan(int4_t x)
 {
     (void)x;
     return false;
 };
 #endif
  
 static inline __host__ half_t sqrt(half_t x)
 {
     return static_cast<half_t>(std::sqrt(static_cast<float>(x)));
 };
  
 static inline __host__ float sqrt(float x) { return std::sqrt(x); };
  
 static inline __host__ double sqrt(double x) { return std::sqrt(x); };
  
 template <typename T>
 inline __host__ T tanh(T x)
 {
     return ck::type_convert<T>(std::tanhf(ck::type_convert<float>(x)));
 };
  
 template <>
 inline __host__ float tanh<float>(float x)
 {
     return std::tanhf(x);
 };
  
 template <>
 inline __host__ double tanh<double>(double x)
 {
     return std::tanh(x);
 };
  
 template <typename T>
 inline __host__ T acos(T x)
 {
     return ck::type_convert<T>(std::acosf(ck::type_convert<float>(x)));
 };
  
 template <>
 inline __host__ float acos<float>(float x)
 {
     return std::acosf(x);
 };
  
 template <>
 inline __host__ double acos<double>(double x)
 {
     return std::acos(x);
 };
  
 template <typename T>
 inline __host__ T neg(T x)
 {
     return ck::type_convert<T>(-(ck::type_convert<float>(x)));
 };
  
 template <>
 inline __host__ float neg<float>(float x)
 {
     return -x;
 };
  
 template <>
 inline __host__ double neg<double>(double x)
 {
     return -x;
 };
  
 template <>
 inline __host__ int32_t neg<int32_t>(int32_t x)
 {
     return -x;
 };
  
 template <>
 inline __host__ int8_t neg<int8_t>(int8_t x)
 {
     return -x;
 };
  
 template <typename T>
 inline __host__ T atan(T x)
 {
     return ck::type_convert<T>(std::atanf(ck::type_convert<float>(x)));
 };
  
 template <>
 inline __host__ float atan<float>(float x)
 {
     return std::atanf(x);
 };
  
 template <>
 inline __host__ double atan<double>(double x)
 {
     return std::atan(x);
 };
  
 template <typename T>
 inline __host__ T sin(T x)
 {
     return ck::type_convert<T>(std::sinf(ck::type_convert<float>(x)));
 };
  
 template <>
 inline __host__ float sin<float>(float x)
 {
     return std::sinf(x);
 };
  
 template <>
 inline __host__ double sin<double>(double x)
 {
     return std::sin(x);
 };
  
 template <typename T>
 inline __host__ T asin(T x)
 {
     return ck::type_convert<T>(std::asinf(ck::type_convert<float>(x)));
 };
  
 template <>
 inline __host__ float asin<float>(float x)
 {
     return std::asinf(x);
 };
  
 template <>
 inline __host__ double asin<double>(double x)
 {
     return std::asin(x);
 };
  
 template <typename T>
 inline __host__ T asinh(T x)
 {
     return ck::type_convert<T>(std::asinhf(ck::type_convert<float>(x)));
 };
  
 template <>
 inline __host__ float asinh<float>(float x)
 {
     return std::asinhf(x);
 };
  
 template <>
 inline __host__ double asinh<double>(double x)
 {
     return std::asinh(x);
 };
  
 template <typename T>
 inline __host__ T cos(T x)
 {
     return ck::type_convert<T>(std::cosf(ck::type_convert<float>(x)));
 };
  
 template <>
 inline __host__ float cos<float>(float x)
 {
     return std::cosf(x);
 };
  
 template <>
 inline __host__ double cos<double>(double x)
 {
     return std::cos(x);
 };
  
 template <typename T>
 inline __host__ T acosh(T x)
 {
     return ck::type_convert<T>(std::acoshf(ck::type_convert<float>(x)));
 };
  
 template <>
 inline __host__ float acosh<float>(float x)
 {
     return std::acoshf(x);
 };
  
 template <>
 inline __host__ double acosh<double>(double x)
 {
     return std::acosh(x);
 };
  
 template <typename T>
 inline __host__ T tan(T x)
 {
     return ck::type_convert<T>(std::tanf(ck::type_convert<float>(x)));
 };
  
 template <>
 inline __host__ float tan<float>(float x)
 {
     return std::tanf(x);
 };
  
 template <>
 inline __host__ double tan<double>(double x)
 {
     return std::tan(x);
 };
  
 template <typename T>
 inline __host__ T atanh(T x)
 {
     return ck::type_convert<T>(std::atanhf(ck::type_convert<float>(x)));
 };
  
 template <>
 inline __host__ float atanh<float>(float x)
 {
     return std::atanhf(x);
 };
  
 template <>
 inline __host__ double atanh<double>(double x)
 {
     return std::atanh(x);
 };
  
 template <typename T>
 inline __host__ T sinh(T x)
 {
     return ck::type_convert<T>(std::sinhf(ck::type_convert<float>(x)));
 };
  
 template <>
 inline __host__ float sinh<float>(float x)
 {
     return std::sinhf(x);
 };
  
 template <>
 inline __host__ double sinh<double>(double x)
 {
     return std::sinh(x);
 };
  
 template <typename T>
 inline __host__ T ceil(T x)
 {
     return ck::type_convert<T>(std::ceilf(ck::type_convert<float>(x)));
 };
  
 template <>
 inline __host__ float ceil<float>(float x)
 {
     return std::ceilf(x);
 };
  
 template <>
 inline __host__ double ceil<double>(double x)
 {
     return std::ceil(x);
 };
  
 template <typename T>
 inline __host__ T cosh(T x)
 {
     return ck::type_convert<T>(std::coshf(ck::type_convert<float>(x)));
 };
  
 template <>
 inline __host__ float cosh<float>(float x)
 {
     return std::coshf(x);
 };
  
 template <>
 inline __host__ double cosh<double>(double x)
 {
     return std::cosh(x);
 };
  
 template <typename T>
 inline __host__ T floor(T x)
 {
     return ck::type_convert<T>(std::floorf(ck::type_convert<float>(x)));
 };
  
 template <>
 inline __host__ float floor<float>(float x)
 {
     return std::floorf(x);
 };
  
 template <>
 inline __host__ double floor<double>(double x)
 {
     return std::floor(x);
 };
  
 template <typename T>
 inline __host__ T rcp(T x)
 {
     return ck::type_convert<T>(1.f / ck::type_convert<float>(x));
 };
  
 template <typename T>
 inline __host__ T exp(T x)
 {
     return ck::type_convert<T>(std::expf(ck::type_convert<float>(x)));
 }
  
 template <>
 inline __host__ float exp<float>(float x)
 {
     return std::expf(x);
 }
  
 template <>
 inline __host__ double exp<double>(double x)
 {
     return std::exp(x);
 }
  
 template <typename T>
 inline __host__ T log(T x)
 {
     return ck::type_convert<T>(std::logf(ck::type_convert<float>(x)));
 }
  
 template <>
 inline __host__ float log<float>(float x)
 {
     return std::logf(x);
 }
  
 template <>
 inline __host__ double log<double>(double x)
 {
     return std::log(x);
 }
  
 template <typename T>
 inline __host__ T pow(T x, T gamma)
 {
     return ck::type_convert<T>(
         std::powf(ck::type_convert<float>(x), ck::type_convert<float>(gamma)));
 }
  
 template <>
 inline __host__ float pow<float>(float x, float gamma)
 {
     return std::powf(x, gamma);
 }
  
 template <>
 inline __host__ double pow<double>(double x, double gamma)
 {
     return std::pow(x, gamma);
 }
  
 template <typename T>
 inline __host__ T expm1(T x)
 {
     return ck::type_convert<T>(std::expm1f(ck::type_convert<float>(x)));
 }
  
 template <>
 inline __host__ float expm1<float>(float x)
 {
     return std::expm1f(x);
 }
  
 template <>
 inline __host__ double expm1<double>(double x)
 {
     return std::expm1(x);
 }
 #endif
 // math functions for the HIP kernel,  some are implemented by calling hip builtin functions
  
 static inline __device__ float abs(float x) { return ::abs(x); };
  
 static inline __device__ double abs(double x) { return ::abs(x); };
  
 static inline __device__ int8_t abs(int8_t x)
 {
     int8_t sgn = x >> (8 - 1);
  
     return (x ^ sgn) - sgn;
 };
  
 static inline __device__ int32_t abs(int32_t x)
 {
     int32_t sgn = x >> (32 - 1);
  
     return (x ^ sgn) - sgn;
 };
  
 #ifdef CK_EXPERIMENTAL_BIT_INT_EXTENSION_INT4
 static inline __device__ int4_t abs(int4_t x)
 {
     int4_t sgn = x >> (4 - 1);
  
     return (x ^ sgn) - sgn;
 };
 #endif
  
 static inline __device__ half_t abs(half_t x)
 {
     uint16_t xx = ck::bit_cast<uint16_t>(x);
  
     uint16_t abs_xx = xx & 0x7fff;
  
     half_t abs_x = ck::bit_cast<half_t>(abs_xx);
  
     return abs_x;
 };
  
 static inline __device__ bool isnan(float x) { return ::isnan(x); };
  
 static inline __device__ bool isnan(double x) { return ::isnan(x); };
  
 static inline __device__ bool isnan(int8_t x)
 {
     (void)x;
     return false;
 };
  
 static inline __device__ bool isnan(int32_t x)
 {
     (void)x;
     return false;
 };
  
 #ifdef CK_EXPERIMENTAL_BIT_INT_EXTENSION_INT4
 static inline __device__ bool isnan(int4_t x)
 {
     (void)x;
     return false;
 };
 #endif
  
 static inline __device__ bool isnan(half_t x)
 {
     uint16_t xx = ck::bit_cast<uint16_t>(x);
  
     return (xx & 0x7FFF) > 0x7C00;
 };
  
 static inline __device__ bool isnan(f8_t x) { return ck::fp8_is_nan(x); };
  
 static inline __device__ half_t sqrt(half_t x)
 {
     return static_cast<half_t>(__builtin_amdgcn_sqrtf(static_cast<float>(x)));
 };
  
 static inline __device__ float sqrt(float x) { return __builtin_amdgcn_sqrtf(x); };
  
 static inline __device__ double sqrt(double x) { return __builtin_amdgcn_sqrt(x); };
  
 template <typename T>
 inline __device__ T tanh(T x)
 {
     return ck::type_convert<T>(::tanhf(ck::type_convert<float>(x)));
 };
  
 template <>
 inline __device__ float tanh<float>(float x)
 {
     return ::tanhf(x);
 };
  
 template <>
 inline __device__ double tanh<double>(double x)
 {
     return ::tanh(x);
 };
  
 template <typename T>
 inline __device__ T acos(T x)
 {
     return ck::type_convert<T>(::acosf(ck::type_convert<float>(x)));
 };
  
 template <>
 inline __device__ float acos<float>(float x)
 {
     return ::acosf(x);
 };
  
 template <>
 inline __device__ double acos<double>(double x)
 {
     return ::acos(x);
 };
  
 template <typename T>
 inline __device__ T neg(T x)
 {
     return ck::type_convert<T>(-(ck::type_convert<float>(x)));
 };
  
 template <>
 inline __device__ float neg<float>(float x)
 {
     return -x;
 };
  
 template <>
 inline __device__ double neg<double>(double x)
 {
     return -x;
 };
  
 template <>
 inline __device__ int32_t neg<int32_t>(int32_t x)
 {
     return -x;
 };
  
 template <>
 inline __device__ int8_t neg<int8_t>(int8_t x)
 {
     return -x;
 };
  
 template <>
 inline __device__ half_t neg<half_t>(half_t x)
 {
     return __hneg(static_cast<__half>(x));
 };
  
 template <typename T>
 inline __device__ T atan(T x)
 {
     return ck::type_convert<T>(::atanf(ck::type_convert<float>(x)));
 };
  
 template <>
 inline __device__ float atan<float>(float x)
 {
     return ::atanf(x);
 };
  
 template <>
 inline __device__ double atan<double>(double x)
 {
     return ::atan(x);
 };
  
 template <typename T>
 inline __device__ T sin(T x)
 {
     return ck::type_convert<T>(::sinf(ck::type_convert<float>(x)));
 };
  
 template <>
 inline __device__ float sin<float>(float x)
 {
     return ::sinf(x);
 };
  
 template <>
 inline __device__ double sin<double>(double x)
 {
     return ::sin(x);
 };
  
 template <>
 inline __device__ half_t sin<half_t>(half_t x)
 {
     return hsin(static_cast<__half>(x));
 };
  
 template <typename T>
 inline __device__ T asin(T x)
 {
     return ck::type_convert<T>(::asinf(ck::type_convert<float>(x)));
 };
  
 template <>
 inline __device__ float asin<float>(float x)
 {
     return ::asinf(x);
 };
  
 template <>
 inline __device__ double asin<double>(double x)
 {
     return ::asin(x);
 };
  
 template <typename T>
 inline __device__ T asinh(T x)
 {
     return ck::type_convert<T>(::asinhf(ck::type_convert<float>(x)));
 };
  
 template <>
 inline __device__ float asinh<float>(float x)
 {
     return ::asinhf(x);
 };
  
 template <>
 inline __device__ double asinh<double>(double x)
 {
     return ::asinh(x);
 };
  
 template <typename T>
 inline __device__ T acosh(T x)
 {
     return ck::type_convert<T>(::acoshf(ck::type_convert<float>(x)));
 };
  
 template <>
 inline __device__ float acosh<float>(float x)
 {
     return ::acoshf(x);
 };
  
 template <>
 inline __device__ double acosh<double>(double x)
 {
     return ::acosh(x);
 };
  
 template <typename T>
 inline __device__ T tan(T x)
 {
     return ck::type_convert<T>(::tanf(ck::type_convert<float>(x)));
 };
  
 template <>
 inline __device__ float tan<float>(float x)
 {
     return ::tanf(x);
 };
  
 template <>
 inline __device__ double tan<double>(double x)
 {
     return ::tan(x);
 };
  
 template <typename T>
 inline __device__ T atanh(T x)
 {
     return ck::type_convert<T>(::atanhf(ck::type_convert<float>(x)));
 };
  
 template <>
 inline __device__ float atanh<float>(float x)
 {
     return ::atanhf(x);
 };
  
 template <>
 inline __device__ double atanh<double>(double x)
 {
     return ::atanh(x);
 };
  
 template <typename T>
 inline __device__ T sinh(T x)
 {
     return ck::type_convert<T>(::sinhf(ck::type_convert<float>(x)));
 };
  
 template <>
 inline __device__ float sinh<float>(float x)
 {
     return ::sinhf(x);
 };
  
 template <>
 inline __device__ double sinh<double>(double x)
 {
     return ::sinh(x);
 };
  
 template <typename T>
 inline __device__ T ceil(T x)
 {
     return ck::type_convert<T>(::ceilf(ck::type_convert<float>(x)));
 };
  
 template <>
 inline __device__ float ceil<float>(float x)
 {
     return ::ceilf(x);
 };
  
 template <>
 inline __device__ double ceil<double>(double x)
 {
     return ::ceil(x);
 };
  
 template <>
 inline __device__ half_t ceil<half_t>(half_t x)
 {
     return hceil(static_cast<__half>(x));
 };
  
 template <typename T>
 inline __device__ T cosh(T x)
 {
     return ck::type_convert<T>(::coshf(ck::type_convert<float>(x)));
 };
  
 template <>
 inline __device__ float cosh<float>(float x)
 {
     return ::coshf(x);
 };
  
 template <>
 inline __device__ double cosh<double>(double x)
 {
     return ::cosh(x);
 };
  
 template <typename T>
 inline __device__ T floor(T x)
 {
     return ck::type_convert<T>(::floorf(ck::type_convert<float>(x)));
 };
  
 template <>
 inline __device__ float floor<float>(float x)
 {
     return ::floorf(x);
 };
  
 template <>
 inline __device__ double floor<double>(double x)
 {
     return ::floor(x);
 };
  
 template <>
 inline __device__ half_t floor<half_t>(half_t x)
 {
     return hfloor(static_cast<__half>(x));
 };
  
 template <typename T>
 inline __device__ T rcp(T x)
 {
 #if !CK_WORKAROUND_SWDEV_383542
     return __frcp_rn(x);
 #else
     return __ocml_native_recip_f32(x);
 #endif
 };
  
 template <typename T>
 inline __device__ T exp(T x)
 {
     return ck::type_convert<T>(__ocml_exp_f32(ck::type_convert<float>(x)));
 };
  
 template <>
 inline __device__ half_t exp<half_t>(half_t x)
 {
     return hexp(static_cast<__half>(x));
 };
  
 template <>
 inline __device__ float exp<float>(float x)
 {
     return __ocml_exp_f32(x);
 };
  
 template <>
 inline __device__ double exp<double>(double x)
 {
     return exp(x);
 };
  
 template <typename T>
 inline __device__ T log(T x)
 {
     return ck::type_convert<T>(__logf(ck::type_convert<float>(x)));
 };
  
 template <>
 inline __device__ half_t log<half_t>(half_t x)
 {
     return hlog(static_cast<__half>(x));
 };
  
 template <>
 inline __device__ float log<float>(float x)
 {
     return __logf(x);
 };
  
 template <>
 inline __device__ double log<double>(double x)
 {
     return log(x);
 };
  
 template <typename T>
 inline __device__ T pow(T x, T gamma)
 {
     return ck::type_convert<T>(powf(ck::type_convert<float>(x), ck::type_convert<float>(gamma)));
 };
  
 template <>
 inline __device__ float pow<float>(float x, float gamma)
 {
     return powf(x, gamma);
 };
  
 template <>
 inline __device__ double pow<double>(double x, double gamma)
 {
     return pow(x, gamma);
 };
  
 template <typename T>
 inline __device__ T expm1(T x)
 {
     return ck::type_convert<T>(expm1f(ck::type_convert<float>(x)));
 };
  
 template <>
 inline __device__ float expm1<float>(float x)
 {
     return expm1f(x);
 };
  
 template <>
 inline __device__ double expm1<double>(double x)
 {
     return expm1(x);
 };
  
 template <typename T>
 inline __device__ T cos(T x)
 {
     return ck::type_convert<T>(cosf(ck::type_convert<float>(x)));
 };
  
 template <>
 inline __device__ float cos<float>(float x)
 {
     return cosf(x);
 };
  
 template <>
 inline __device__ double cos<double>(double x)
 {
     return cos(x);
 };
  
 } // namespace math
 } // namespace ck