40 #ifndef __CLANG_CUDA_RUNTIME_WRAPPER_H__
41 #define __CLANG_CUDA_RUNTIME_WRAPPER_H__
43 #if defined(__CUDA__) && defined(__clang__)
56 #pragma push_macro("__THROW")
57 #pragma push_macro("__CUDA_ARCH__")
63 #if !defined(CUDA_VERSION)
64 #error "cuda.h did not define CUDA_VERSION"
65 #elif CUDA_VERSION < 7000 || CUDA_VERSION > 8000
66 #error "Unsupported CUDA version!"
72 #define __CUDA_ARCH__ 350
79 #define __DEVICE_LAUNCH_PARAMETERS_H__
84 #define __DEVICE_FUNCTIONS_H__
85 #define __MATH_FUNCTIONS_H__
86 #define __COMMON_FUNCTIONS_H__
92 #include "driver_types.h"
93 #include "host_config.h"
94 #include "host_defines.h"
98 #include "cuda_runtime.h"
105 #define __nvvm_memcpy(s, d, n, a) __builtin_memcpy(s, d, n)
106 #define __nvvm_memset(d, c, n, a) __builtin_memset(d, c, n)
108 #include "crt/device_runtime.h"
109 #include "crt/host_runtime.h"
113 #undef __cxa_vec_ctor
114 #undef __cxa_vec_cctor
115 #undef __cxa_vec_dtor
117 #undef __cxa_vec_new2
118 #undef __cxa_vec_new3
119 #undef __cxa_vec_delete2
120 #undef __cxa_vec_delete
121 #undef __cxa_vec_delete3
122 #undef __cxa_pure_virtual
128 inline __host__
double __signbitd(
double x) {
138 #pragma push_macro("__host__")
140 #define __CUDACC_RTC__
141 #include "device_functions_decls.h"
142 #undef __CUDACC_RTC__
146 #define __host__ UNEXPECTED_HOST_ATTRIBUTE
153 #if defined(CU_DEVICE_INVALID)
154 #if !defined(__USE_FAST_MATH__)
155 #define __USE_FAST_MATH__ 0
158 #if !defined(__CUDA_PREC_DIV)
159 #define __CUDA_PREC_DIV 0
167 #pragma push_macro("__forceinline__")
168 #define __forceinline__ __device__ __inline__ __attribute__((always_inline))
169 #include "device_functions.hpp"
177 #pragma push_macro("__USE_FAST_MATH__")
178 #if defined(__CLANG_CUDA_APPROX_TRANSCENDENTALS__)
179 #define __USE_FAST_MATH__ 1
181 #include "math_functions.hpp"
182 #pragma pop_macro("__USE_FAST_MATH__")
184 #include "math_functions_dbl_ptx3.hpp"
185 #pragma pop_macro("__forceinline__")
189 #undef __MATH_FUNCTIONS_HPP__
191 #include "math_functions.hpp"
195 static inline float rsqrt(
float __a) {
return rsqrtf(__a); }
196 static inline float rcbrt(
float __a) {
return rcbrtf(__a); }
197 static inline float sinpi(
float __a) {
return sinpif(__a); }
198 static inline float cospi(
float __a) {
return cospif(__a); }
199 static inline void sincospi(
float __a,
float *
__b,
float *
__c) {
200 return sincospif(__a, __b, __c);
202 static inline float erfcinv(
float __a) {
return erfcinvf(__a); }
203 static inline float normcdfinv(
float __a) {
return normcdfinvf(__a); }
204 static inline float normcdf(
float __a) {
return normcdff(__a); }
205 static inline float erfcx(
float __a) {
return erfcxf(__a); }
209 static inline __device__
void __brkpt(
int __c) { __brkpt(); }
220 #undef __DEVICE_FUNCTIONS_HPP__
221 #include "device_atomic_functions.hpp"
222 #include "device_functions.hpp"
223 #include "sm_20_atomic_functions.hpp"
224 #include "sm_20_intrinsics.hpp"
225 #include "sm_32_atomic_functions.hpp"
235 #if CUDA_VERSION >= 8000
236 #include "sm_60_atomic_functions.hpp"
237 #include "sm_61_intrinsics.hpp"
240 #undef __MATH_FUNCTIONS_HPP__
246 #pragma push_macro("signbit")
247 #pragma push_macro("__GNUC__")
249 #define signbit __ignored_cuda_signbit
250 #include "math_functions.hpp"
251 #pragma pop_macro("__GNUC__")
252 #pragma pop_macro("signbit")
254 #pragma pop_macro("__host__")
256 #include "texture_indirect_functions.h"
259 #pragma pop_macro("__CUDA_ARCH__")
260 #pragma pop_macro("__THROW")
272 __device__
int vprintf(
const char *,
const char *);
273 __device__
void free(
void *) __attribute((nothrow));
274 __device__
void *malloc(
size_t) __attribute((nothrow))
__attribute__((malloc));
275 __device__
void __assertfail(const
char *__message, const
char *__file,
276 unsigned __line, const
char *__function,
281 __device__ static inline
void __assert_fail(const
char *__message,
282 const
char *__file,
unsigned __line,
283 const
char *__function) {
284 __assertfail(__message, __file, __line, __function,
sizeof(
char));
289 __device__
int printf(
const char *, ...);
294 __device__
static inline void free(
void *__ptr) { ::free(__ptr); }
295 __device__
static inline void *malloc(
size_t __size) {
296 return ::malloc(__size);
303 __device__
inline __cuda_builtin_threadIdx_t::operator uint3()
const {
311 __device__
inline __cuda_builtin_blockIdx_t::operator uint3()
const {
319 __device__
inline __cuda_builtin_blockDim_t::operator dim3()
const {
320 return dim3(x, y, z);
323 __device__
inline __cuda_builtin_gridDim_t::operator dim3()
const {
324 return dim3(x, y, z);
337 #pragma push_macro("dim3")
338 #pragma push_macro("uint3")
339 #define dim3 __cuda_builtin_blockDim_t
340 #define uint3 __cuda_builtin_threadIdx_t
341 #include "curand_mtgp32_kernel.h"
342 #pragma pop_macro("dim3")
343 #pragma pop_macro("uint3")
344 #pragma pop_macro("__USE_FAST_MATH__")
347 #endif // __CLANG_CUDA_RUNTIME_WRAPPER_H__
__DEVICE__ bool signbit(float __x)
Test for sign bit.
float __ovld __cnfn cospi(float x)
Compute cos (PI * x).
char __v64qi __attribute__((__vector_size__(64)))
float __ovld __cnfn rsqrt(float)
Compute inverse square root.
static __inline__ vector float vector float __b
int printf(__constant const char *st,...)
static __inline__ vector float vector float vector float __c
float __ovld __cnfn sinpi(float x)
Compute sin (PI * x).