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 > 7050
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
116 #undef __cxa_vec_new2
117 #undef __cxa_vec_new3
118 #undef __cxa_vec_delete2
119 #undef __cxa_vec_delete
120 #undef __cxa_vec_delete3
121 #undef __cxa_pure_virtual
128 #pragma push_macro("__host__")
130 #define __CUDACC_RTC__
131 #include "device_functions_decls.h"
132 #undef __CUDACC_RTC__
136 #define __host__ UNEXPECTED_HOST_ATTRIBUTE
142 #pragma push_macro("__forceinline__")
143 #define __forceinline__ __device__ __inline__ __attribute__((always_inline))
144 #include "device_functions.hpp"
152 #pragma push_macro("__USE_FAST_MATH__")
153 #if defined(__CLANG_CUDA_APPROX_TRANSCENDENTALS__)
154 #define __USE_FAST_MATH__
156 #include "math_functions.hpp"
157 #pragma pop_macro("__USE_FAST_MATH__")
159 #include "math_functions_dbl_ptx3.hpp"
160 #pragma pop_macro("__forceinline__")
164 #undef __MATH_FUNCTIONS_HPP__
166 #include "math_functions.hpp"
170 static inline float rsqrt(
float __a) {
return rsqrtf(__a); }
171 static inline float rcbrt(
float __a) {
return rcbrtf(__a); }
172 static inline float sinpi(
float __a) {
return sinpif(__a); }
173 static inline float cospi(
float __a) {
return cospif(__a); }
174 static inline void sincospi(
float __a,
float *
__b,
float *
__c) {
175 return sincospif(__a, __b, __c);
177 static inline float erfcinv(
float __a) {
return erfcinvf(__a); }
178 static inline float normcdfinv(
float __a) {
return normcdfinvf(__a); }
179 static inline float normcdf(
float __a) {
return normcdff(__a); }
180 static inline float erfcx(
float __a) {
return erfcxf(__a); }
184 static inline __device__
void __brkpt(
int __c) { __brkpt(); }
195 #undef __DEVICE_FUNCTIONS_HPP__
196 #include "device_atomic_functions.hpp"
197 #include "device_functions.hpp"
198 #include "sm_20_atomic_functions.hpp"
199 #include "sm_20_intrinsics.hpp"
200 #include "sm_32_atomic_functions.hpp"
210 #undef __MATH_FUNCTIONS_HPP__
216 #pragma push_macro("signbit")
217 #pragma push_macro("__GNUC__")
219 #define signbit __ignored_cuda_signbit
220 #include "math_functions.hpp"
221 #pragma pop_macro("__GNUC__")
222 #pragma pop_macro("signbit")
224 #pragma pop_macro("__host__")
226 #include "texture_indirect_functions.h"
229 #pragma pop_macro("__CUDA_ARCH__")
230 #pragma pop_macro("__THROW")
242 __device__
int vprintf(
const char *,
const char *);
243 __device__
void free(
void *) __attribute((nothrow));
244 __device__
void *malloc(
size_t) __attribute((nothrow))
__attribute__((malloc));
245 __device__
void __assertfail(const
char *__message, const
char *__file,
246 unsigned __line, const
char *__function,
251 __device__ static inline
void __assert_fail(const
char *__message,
252 const
char *__file,
unsigned __line,
253 const
char *__function) {
254 __assertfail(__message, __file, __line, __function,
sizeof(
char));
259 __device__
int printf(
const char *, ...);
264 __device__
static inline void free(
void *__ptr) { ::free(__ptr); }
265 __device__
static inline void *malloc(
size_t __size) {
266 return ::malloc(__size);
273 __device__
inline __cuda_builtin_threadIdx_t::operator uint3()
const {
281 __device__
inline __cuda_builtin_blockIdx_t::operator uint3()
const {
289 __device__
inline __cuda_builtin_blockDim_t::operator dim3()
const {
290 return dim3(x, y, z);
293 __device__
inline __cuda_builtin_gridDim_t::operator dim3()
const {
294 return dim3(x, y, z);
306 #pragma push_macro("dim3")
307 #pragma push_macro("uint3")
308 #define dim3 __cuda_builtin_blockDim_t
309 #define uint3 __cuda_builtin_threadIdx_t
310 #include "curand_mtgp32_kernel.h"
311 #pragma pop_macro("dim3")
312 #pragma pop_macro("uint3")
313 #pragma pop_macro("__USE_FAST_MATH__")
316 #endif // __CLANG_CUDA_RUNTIME_WRAPPER_H__
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).