libdevice.txt

LIBDEVICE USER'S GUIDE
Part 000 _v8.0 | February 2016TABLE OF CONTENTS
Chapter 1. Introduction.........................................................................................1
1.1. What Is libdevice?........................................................................................1
Chapter 2. Basic Usage..........................................................................................2
2.1. Linking with libdevice................................................................................... 2
2.2. Selecting Library Version................................................................................3
Chapter 3. Function Reference...............................................................................4
3.1. __nv_abs....................................................................................................4
3.2. __nv_acos..................................................................................................4
3.3. __nv_acosf.................................................................................................5
3.4. __nv_acosh.................................................................................................5
3.5. __nv_acoshf................................................................................................6
3.6. __nv_asin...................................................................................................7
3.7. __nv_asinf..................................................................................................7
3.8. __nv_asinh................................................................................................. 8
3.9. __nv_asinhf................................................................................................ 8
3.10. __nv_atan.................................................................................................9
3.11. __nv_atan2............................................................................................... 9
3.12. __nv_atan2f.............................................................................................10
3.13. __nv_atanf.............................................................................................. 10
3.14. __nv_atanh..............................................................................................11
3.15. __nv_atanhf.............................................................................................11
3.16. __nv_brev............................................................................................... 12
3.17. __nv_brevll..............................................................................................12
3.18. __nv_byte_perm........................................................................................13
3.19. __nv_cbrt................................................................................................14
3.20. __nv_cbrtf...............................................................................................14
3.21. __nv_ceil................................................................................................ 15
3.22. __nv_ceilf............................................................................................... 15
3.23. __nv_clz................................................................................................. 16
3.24. __nv_clzll................................................................................................16
3.25. __nv_copysign.......................................................................................... 17
3.26. __nv_copysignf......................................................................................... 17
3.27. __nv_cos.................................................................................................17
3.28. __nv_cosf................................................................................................18
3.29. __nv_cosh............................................................................................... 19
3.30. __nv_coshf.............................................................................................. 19
3.31. __nv_cospi...............................................................................................20
3.32. __nv_cospif..............................................................................................20
3.33. __nv_dadd_rd...........................................................................................21
3.34. __nv_dadd_rn...........................................................................................21
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | ii3.35. __nv_dadd_ru...........................................................................................22
3.36. __nv_dadd_rz...........................................................................................22
3.37. __nv_ddiv_rd............................................................................................23
3.38. __nv_ddiv_rn............................................................................................23
3.39. __nv_ddiv_ru............................................................................................24
3.40. __nv_ddiv_rz............................................................................................24
3.41. __nv_dmul_rd...........................................................................................25
3.42. __nv_dmul_rn...........................................................................................25
3.43. __nv_dmul_ru...........................................................................................26
3.44. __nv_dmul_rz...........................................................................................26
3.45. __nv_double2float_rd................................................................................. 27
3.46. __nv_double2float_rn................................................................................. 27
3.47. __nv_double2float_ru................................................................................. 28
3.48. __nv_double2float_rz..................................................................................28
3.49. __nv_double2hiint......................................................................................29
3.50. __nv_double2int_rd....................................................................................29
3.51. __nv_double2int_rn....................................................................................30
3.52. __nv_double2int_ru....................................................................................30
3.53. __nv_double2int_rz....................................................................................31
3.54. __nv_double2ll_rd......................................................................................31
3.55. __nv_double2ll_rn......................................................................................32
3.56. __nv_double2ll_ru......................................................................................32
3.57. __nv_double2ll_rz......................................................................................33
3.58. __nv_double2loint......................................................................................33
3.59. __nv_double2uint_rd..................................................................................34
3.60. __nv_double2uint_rn.................................................................................. 34
3.61. __nv_double2uint_ru.................................................................................. 35
3.62. __nv_double2uint_rz...................................................................................35
3.63. __nv_double2ull_rd....................................................................................36
3.64. __nv_double2ull_rn....................................................................................36
3.65. __nv_double2ull_ru....................................................................................37
3.66. __nv_double2ull_rz....................................................................................37
3.67. __nv_double_as_longlong.............................................................................38
3.68. __nv_drcp_rd........................................................................................... 38
3.69. __nv_drcp_rn........................................................................................... 39
3.70. __nv_drcp_ru........................................................................................... 39
3.71. __nv_drcp_rz............................................................................................40
3.72. __nv_dsqrt_rd...........................................................................................40
3.73. __nv_dsqrt_rn...........................................................................................41
3.74. __nv_dsqrt_ru...........................................................................................41
3.75. __nv_dsqrt_rz...........................................................................................42
3.76. __nv_erf.................................................................................................42
3.77. __nv_erfc................................................................................................43
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | iii3.78. __nv_erfcf...............................................................................................43
3.79. __nv_erfcinv............................................................................................44
3.80. __nv_erfcinvf...........................................................................................44
3.81. __nv_erfcx...............................................................................................45
3.82. __nv_erfcxf..............................................................................................46
3.83. __nv_erff................................................................................................46
3.84. __nv_erfinv..............................................................................................47
3.85. __nv_erfinvf.............................................................................................47
3.86. __nv_exp................................................................................................ 48
3.87. __nv_exp10..............................................................................................48
3.88. __nv_exp10f.............................................................................................49
3.89. __nv_exp2...............................................................................................49
3.90. __nv_exp2f..............................................................................................50
3.91. __nv_expf............................................................................................... 50
3.92. __nv_expm1.............................................................................................51
3.93. __nv_expm1f............................................................................................51
3.94. __nv_fabs................................................................................................52
3.95. __nv_fabsf...............................................................................................52
3.96. __nv_fadd_rd...........................................................................................53
3.97. __nv_fadd_rn........................................................................................... 53
3.98. __nv_fadd_ru........................................................................................... 54
3.99. __nv_fadd_rz............................................................................................54
3.100. __nv_fast_cosf........................................................................................ 55
3.101. __nv_fast_exp10f.....................................................................................56
3.102. __nv_fast_expf........................................................................................56
3.103. __nv_fast_fdividef....................................................................................57
3.104. __nv_fast_log10f......................................................................................57
3.105. __nv_fast_log2f.......................................................................................58
3.106. __nv_fast_logf.........................................................................................58
3.107. __nv_fast_powf....................................................................................... 59
3.108. __nv_fast_sincosf.....................................................................................59
3.109. __nv_fast_sinf.........................................................................................60
3.110. __nv_fast_tanf........................................................................................ 61
3.111. __nv_fdim..............................................................................................61
3.112. __nv_fdimf.............................................................................................62
3.113. __nv_fdiv_rd...........................................................................................62
3.114. __nv_fdiv_rn...........................................................................................63
3.115. __nv_fdiv_ru...........................................................................................63
3.116. __nv_fdiv_rz...........................................................................................64
3.117. __nv_ffs................................................................................................64
3.118. __nv_ffsll...............................................................................................65
3.119. __nv_finitef............................................................................................65
3.120. __nv_float2half_rn................................................................................... 66
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | iv3.121. __nv_float2int_rd.....................................................................................66
3.122. __nv_float2int_rn.....................................................................................67
3.123. __nv_float2int_ru.....................................................................................67
3.124. __nv_float2int_rz.....................................................................................68
3.125. __nv_float2ll_rd.......................................................................................68
3.126. __nv_float2ll_rn.......................................................................................69
3.127. __nv_float2ll_ru.......................................................................................69
3.128. __nv_float2ll_rz.......................................................................................70
3.129. __nv_float2uint_rd................................................................................... 70
3.130. __nv_float2uint_rn................................................................................... 71
3.131. __nv_float2uint_ru................................................................................... 71
3.132. __nv_float2uint_rz....................................................................................72
3.133. __nv_float2ull_rd.....................................................................................72
3.134. __nv_float2ull_rn.....................................................................................73
3.135. __nv_float2ull_ru.....................................................................................73
3.136. __nv_float2ull_rz.....................................................................................74
3.137. __nv_float_as_int.....................................................................................74
3.138. __nv_floor..............................................................................................74
3.139. __nv_floorf.............................................................................................75
3.140. __nv_fma...............................................................................................76
3.141. __nv_fma_rd...........................................................................................76
3.142. __nv_fma_rn...........................................................................................77
3.143. __nv_fma_ru...........................................................................................77
3.144. __nv_fma_rz...........................................................................................78
3.145. __nv_fmaf..............................................................................................79
3.146. __nv_fmaf_rd..........................................................................................79
3.147. __nv_fmaf_rn..........................................................................................80
3.148. __nv_fmaf_ru..........................................................................................81
3.149. __nv_fmaf_rz..........................................................................................81
3.150. __nv_fmax.............................................................................................82
3.151. __nv_fmaxf............................................................................................82
3.152. __nv_fmin..............................................................................................83
3.153. __nv_fminf.............................................................................................84
3.154. __nv_fmod.............................................................................................84
3.155. __nv_fmodf............................................................................................85
3.156. __nv_fmul_rd..........................................................................................86
3.157. __nv_fmul_rn..........................................................................................86
3.158. __nv_fmul_ru..........................................................................................87
3.159. __nv_fmul_rz..........................................................................................87
3.160. __nv_frcp_rd.......................................................................................... 88
3.161. __nv_frcp_rn.......................................................................................... 88
3.162. __nv_frcp_ru.......................................................................................... 89
3.163. __nv_frcp_rz...........................................................................................89
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | v3.164. __nv_frexp.............................................................................................90
3.165. __nv_frexpf............................................................................................90
3.166. __nv_frsqrt_rn.........................................................................................91
3.167. __nv_fsqrt_rd..........................................................................................92
3.168. __nv_fsqrt_rn..........................................................................................92
3.169. __nv_fsqrt_ru..........................................................................................93
3.170. __nv_fsqrt_rz..........................................................................................93
3.171. __nv_fsub_rd..........................................................................................94
3.172. __nv_fsub_rn..........................................................................................94
3.173. __nv_fsub_ru..........................................................................................95
3.174. __nv_fsub_rz.......................................................................................... 95
3.175. __nv_hadd............................................................................................. 96
3.176. __nv_half2float....................................................................................... 96
3.177. __nv_hiloint2double..................................................................................97
3.178. __nv_hypot............................................................................................ 97
3.179. __nv_hypotf........................................................................................... 98
3.180. __nv_ilogb............................................................................................. 98
3.181. __nv_ilogbf............................................................................................ 99
3.182. __nv_int2double_rn.................................................................................. 99
3.183. __nv_int2float_rd................................................................................... 100
3.184. __nv_int2float_rn................................................................................... 100
3.185. __nv_int2float_ru................................................................................... 101
3.186. __nv_int2float_rz....................................................................................101
3.187. __nv_int_as_float................................................................................... 102
3.188. __nv_isfinited........................................................................................102
3.189. __nv_isinfd...........................................................................................103
3.190. __nv_isinff............................................................................................103
3.191. __nv_isnand..........................................................................................103
3.192. __nv_isnanf...........................................................................................104
3.193. __nv_j0............................................................................................... 104
3.194. __nv_j0f.............................................................................................. 105
3.195. __nv_j1............................................................................................... 105
3.196. __nv_j1f.............................................................................................. 106
3.197. __nv_jn............................................................................................... 107
3.198. __nv_jnf.............................................................................................. 107
3.199. __nv_ldexp...........................................................................................108
3.200. __nv_ldexpf..........................................................................................108
3.201. __nv_lgamma........................................................................................ 109
3.202. __nv_lgammaf....................................................................................... 110
3.203. __nv_ll2double_rd...................................................................................110
3.204. __nv_ll2double_rn...................................................................................111
3.205. __nv_ll2double_ru...................................................................................111
3.206. __nv_ll2double_rz...................................................................................112
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | vi3.207. __nv_ll2float_rd.....................................................................................112
3.208. __nv_ll2float_rn.....................................................................................113
3.209. __nv_ll2float_ru.....................................................................................113
3.210. __nv_ll2float_rz..................................................................................... 114
3.211. __nv_llabs............................................................................................114
3.212. __nv_llmax...........................................................................................114
3.213. __nv_llmin............................................................................................115
3.214. __nv_llrint............................................................................................115
3.215. __nv_llrintf...........................................................................................116
3.216. __nv_llround......................................................................................... 116
3.217. __nv_llroundf........................................................................................ 117
3.218. __nv_log..............................................................................................117
3.219. __nv_log10........................................................................................... 118
3.220. __nv_log10f.......................................................................................... 118
3.221. __nv_log1p........................................................................................... 119
3.222. __nv_log1pf.......................................................................................... 119
3.223. __nv_log2.............................................................................................120
3.224. __nv_log2f............................................................................................121
3.225. __nv_logb.............................................................................................121
3.226. __nv_logbf............................................................................................122
3.227. __nv_logf.............................................................................................122
3.228. __nv_longlong_as_double..........................................................................123
3.229. __nv_max.............................................................................................123
3.230. __nv_min............................................................................................. 124
3.231. __nv_modf............................................................................................124
3.232. __nv_modff...........................................................................................125
3.233. __nv_mul24...........................................................................................125
3.234. __nv_mul64hi........................................................................................126
3.235. __nv_mulhi...........................................................................................126
3.236. __nv_nan..............................................................................................127
3.237. __nv_nanf.............................................................................................127
3.238. __nv_nearbyint......................................................................................128
3.239. __nv_nearbyintf.....................................................................................128
3.240. __nv_nextafter...................................................................................... 129
3.241. __nv_nextafterf..................................................................................... 129
3.242. __nv_normcdf........................................................................................130
3.243. __nv_normcdff.......................................................................................130
3.244. __nv_normcdfinv....................................................................................131
3.245. __nv_normcdfinvf...................................................................................131
3.246. __nv_popc............................................................................................132
3.247. __nv_popcll.......................................................................................... 132
3.248. __nv_pow.............................................................................................133
3.249. __nv_powf............................................................................................134
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | vii3.250. __nv_powi............................................................................................135
3.251. __nv_powif...........................................................................................136
3.252. __nv_rcbrt............................................................................................137
3.253. __nv_rcbrtf...........................................................................................137
3.254. __nv_remainder..................................................................................... 138
3.255. __nv_remainderf.................................................................................... 138
3.256. __nv_remquo.........................................................................................139
3.257. __nv_remquof........................................................................................140
3.258. __nv_rhadd...........................................................................................140
3.259. __nv_rint..............................................................................................141
3.260. __nv_rintf.............................................................................................141
3.261. __nv_round...........................................................................................142
3.262. __nv_roundf..........................................................................................142
3.263. __nv_rsqrt............................................................................................143
3.264. __nv_rsqrtf...........................................................................................143
3.265. __nv_sad..............................................................................................144
3.266. __nv_saturatef.......................................................................................144
3.267. __nv_scalbn..........................................................................................145
3.268. __nv_scalbnf.........................................................................................145
3.269. __nv_signbitd........................................................................................ 146
3.270. __nv_signbitf.........................................................................................146
3.271. __nv_sin...............................................................................................147
3.272. __nv_sincos...........................................................................................147
3.273. __nv_sincosf..........................................................................................148
3.274. __nv_sincospi........................................................................................ 148
3.275. __nv_sincospif....................................................................................... 149
3.276. __nv_sinf..............................................................................................150
3.277. __nv_sinh.............................................................................................150
3.278. __nv_sinhf............................................................................................151
3.279. __nv_sinpi............................................................................................151
3.280. __nv_sinpif...........................................................................................152
3.281. __nv_sqrt.............................................................................................152
3.282. __nv_sqrtf............................................................................................153
3.283. __nv_tan..............................................................................................153
3.284. __nv_tanf.............................................................................................154
3.285. __nv_tanh............................................................................................ 154
3.286. __nv_tanhf........................................................................................... 155
3.287. __nv_tgamma........................................................................................155
3.288. __nv_tgammaf.......................................................................................156
3.289. __nv_trunc........................................................................................... 157
3.290. __nv_truncf.......................................................................................... 157
3.291. __nv_uhadd.......................................................................................... 157
3.292. __nv_uint2double_rn............................................................................... 158
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | viii3.293. __nv_uint2float_rd..................................................................................158
3.294. __nv_uint2float_rn..................................................................................159
3.295. __nv_uint2float_ru..................................................................................159
3.296. __nv_uint2float_rz..................................................................................160
3.297. __nv_ull2double_rd.................................................................................160
3.298. __nv_ull2double_rn.................................................................................161
3.299. __nv_ull2double_ru.................................................................................161
3.300. __nv_ull2double_rz................................................................................. 162
3.301. __nv_ull2float_rd....................................................................................162
3.302. __nv_ull2float_rn....................................................................................163
3.303. __nv_ull2float_ru....................................................................................163
3.304. __nv_ull2float_rz....................................................................................164
3.305. __nv_ullmax..........................................................................................164
3.306. __nv_ullmin..........................................................................................164
3.307. __nv_umax........................................................................................... 165
3.308. __nv_umin............................................................................................165
3.309. __nv_umul24.........................................................................................166
3.310. __nv_umul64hi.......................................................................................166
3.311. __nv_umulhi..........................................................................................167
3.312. __nv_urhadd......................................................................................... 167
3.313. __nv_usad............................................................................................ 168
3.314. __nv_y0...............................................................................................168
3.315. __nv_y0f..............................................................................................169
3.316. __nv_y1...............................................................................................169
3.317. __nv_y1f..............................................................................................170
3.318. __nv_yn...............................................................................................171
3.319. __nv_ynf..............................................................................................171
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | ixLIST OF TABLES
Table 1  Supported Reflection Parameters....................................................................2
Table 2  Library version selection guidelines.................................................................3
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | xChapter 1.
INTRODUCTION
1.1. What Is libdevice?
The libdevice library is a collection of NVVM bitcode functions that implement common
functions for NVIDIA GPU devices, including math primitives and bit-manipulation
functions. These functions are optimized for particular GPU architectures, and are
intended to be linked with an NVVM IR module during compilation to PTX.
This guide documents both the functions available in libdevice and the basic usage of
the library from a compiler writer's perspective.
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 1Chapter 2.
BASIC USAGE
2.1. Linking with libdevice
The libdevice library ships as an LLVM bitcode library and is meant to be linked with
the target module early in the compilation process. The standard process for linking
with libdevice is to first link it with the target module, then run the standard LLVM
optimization and code generation passes. This allows the optimizers to inline and
perform analyses on the used library functions, and eliminate any used functions as
dead code.
Users of libnvvm can link with libdevice by adding the appropriate libdevice module
to the   object being compiled. In addition, the following options for
nvvmProgram
 affect the behavior of libdevice functions:
nvvmCompileProgram
Table 1  Supported Reflection Parameters
Parameter Values Description
preserve denormal values, when performing
0
(default) single-precision floating-point operations
-ftz
flush denormal values to zero, when performing
1 single-precision floating-point operations
use a faster approximation for single-
0 precision floating-point division and reciprocals
-prec-div
use IEEE round-to-nearest mode for single-
1
(default) precision floating-point division and reciprocals
use IEEE round-to-nearest mode for single-
0 precision floating-point square root
-prec-sqrt
1 use a faster approximation for single-precision floating-point square root
(default)
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 2Basic Usage
The following pseudo-code shows an example of linking an NVVM IR module with the
libdevice library using libnvvm:
nvvmProgram prog;
size_t libdeviceModSize;
const char *libdeviceMod = loadFile('/path/to/libdevice.*.bc',
                                    &libdeviceModSize);
const char *myIr = /* NVVM IR in text or binary format */;
size_t myIrSize = /* size of myIr in bytes */;
// Create NVVM program object
nvvmCreateProgram(&prog);
// Add libdevice module to program
nvvmAddModuleToProgram(prog, libdeviceMod, libdeviceModSize);
// Add custom IR to program
nvvmAddModuleToProgram(prog, myIr, myIrSize);
// Declare compile options
const char *options[] = { "-ftz=1" };
// Compile the program
nvvmCompileProgram(prog, 1, options);
It is the responsibility of the client program to locate and read the libdevice library
binary (represented by the   function in the example).
loadFile
2.2. Selecting Library Version
The libdevice library ships with several versions, each tuned for optimal performance on
a particular device architecture. The following table provides a guideline for choosing
the best libdevice version for the target architecture. All versions can be found in the
CUDA Toolkit under  .
nvvm/libdevice/<library-name>
Table 2  Library version selection guidelines
Compute Capability Library
2.0 ≤ Arch < 3.0
libdevice.compute_20.XX.bc
Arch = 3.0
libdevice.compute_30.XX.bc
3.1 ≤ Arch < 3.5
libdevice.compute_20.XX.bc
3.5 ≤ Arch ≤ 3.7
libdevice.compute_35.XX.bc
3.7 < Arch < 5.0
libdevice.compute_30.XX.bc
5.0 ≤ Arch ≤ 5.3
libdevice.compute_50.XX.bc
Arch > 5.3
libdevice.compute_30.XX.bc
The   in the library name corresponds to the libdevice library version number.
XX
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 3Chapter 3.
FUNCTION REFERENCE
This chapter describes all functions available in libdevice.
3.1. __nv_abs
Prototype:
i32 @__nv_abs(i32 %x) 
Description:
Determine the absolute value of the 32-bit signed integer  .
x
Returns:
Returns the absolute value of the 32-bit signed integer  .
x
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.2. __nv_acos
Prototype:
double @__nv_acos(double %x) 
Description:
Calculate the principal value of the arc cosine of the input argument  .
x
Returns:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 4Function Reference
Result will be in radians, in the interval [0,   ] for   inside [-1, +1].
x
‣ __nv_acos(1) returns +0.
‣ __nv_acos(x) returns NaN for x outside [-1, +1].
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.3. __nv_acosf
Prototype:
float @__nv_acosf(float %x) 
Description:
Calculate the principal value of the arc cosine of the input argument  .
x
Returns:
Result will be in radians, in the interval [0,   ] for   inside [-1, +1].
x
‣ __nv_acosf(1) returns +0.
‣ __nv_acosf(x) returns NaN for x outside [-1, +1].
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.4. __nv_acosh
Prototype:
double @__nv_acosh(double %x) 
Description:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 5Function Reference
Calculate the nonnegative arc hyperbolic cosine of the input argument  .
x
Returns:
Result will be in the interval [0,   ].
‣ __nv_acosh(1) returns 0.
‣ __nv_acosh(x) returns NaN for x in the interval [   , 1).
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.5. __nv_acoshf
Prototype:
float @__nv_acoshf(float %x) 
Description:
Calculate the nonnegative arc hyperbolic cosine of the input argument  .
x
Returns:
Result will be in the interval [0,   ].
‣ __nv_acoshf(1) returns 0.
‣ __nv_acoshf(x) returns NaN for x in the interval [   , 1).
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 6Function Reference
3.6. __nv_asin
Prototype:
double @__nv_asin(double %x) 
Description:
Calculate the principal value of the arc sine of the input argument  .
x
Returns:
Result will be in radians, in the interval [-   /2, +   /2] for   inside [-1, +1].
x
‣ __nv_asin(0) returns +0.
‣ __nv_asin(x) returns NaN for x outside [-1, +1].
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.7. __nv_asinf
Prototype:
float @__nv_asinf(float %x) 
Description:
Calculate the principal value of the arc sine of the input argument  .
x
Returns:
Result will be in radians, in the interval [-   /2, +   /2] for   inside [-1, +1].
x
‣ __nv_asinf(0) returns +0.
‣ __nv_asinf(x) returns NaN for x outside [-1, +1].
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 7Function Reference
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.8. __nv_asinh
Prototype:
double @__nv_asinh(double %x) 
Description:
Calculate the arc hyperbolic sine of the input argument  .
x
Returns:
‣ __nv_asinh(0) returns 1.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.9. __nv_asinhf
Prototype:
float @__nv_asinhf(float %x) 
Description:
Calculate the arc hyperbolic sine of the input argument  .
x
Returns:
‣ __nv_asinh(0) returns 1.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 8Function Reference
Compute 3.0: Yes
Compute 3.5: Yes
3.10. __nv_atan
Prototype:
double @__nv_atan(double %x) 
Description:
Calculate the principal value of the arc tangent of the input argument  .
x
Returns:
Result will be in radians, in the interval [-   /2, +   /2].
‣ __nv_atan(0) returns +0.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.11. __nv_atan2
Prototype:
double @__nv_atan2(double %x, double %y) 
Description:
Calculate the principal value of the arc tangent of the ratio of first and second input
arguments   /  . The quadrant of the result is determined by the signs of inputs   and  .
x y x y
Returns:
Result will be in radians, in the interval [-   /, +   ].
‣ __nv_atan2(0, 1) returns +0.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 9Function Reference
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.12. __nv_atan2f
Prototype:
float @__nv_atan2f(float %x, float %y) 
Description:
Calculate the principal value of the arc tangent of the ratio of first and second input
arguments   /  . The quadrant of the result is determined by the signs of inputs   and  .
x y x y
Returns:
Result will be in radians, in the interval [-   /, +   ].
‣ __nv_atan2f(0, 1) returns +0.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.13. __nv_atanf
Prototype:
float @__nv_atanf(float %x) 
Description:
Calculate the principal value of the arc tangent of the input argument  .
x
Returns:
Result will be in radians, in the interval [-   /2, +   /2].
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 10Function Reference
‣ __nv_atan(0) returns +0.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.14. __nv_atanh
Prototype:
double @__nv_atanh(double %x) 
Description:
Calculate the arc hyperbolic tangent of the input argument  .
x
Returns:
‣ __nv_atanh(   ) returns  .
‣ __nv_atanh(   ) returns  .
‣ __nv_atanh(x) returns NaN for x outside interval [-1, 1].
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.15. __nv_atanhf
Prototype:
float @__nv_atanhf(float %x) 
Description:
Calculate the arc hyperbolic tangent of the input argument  .
x
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 11Function Reference
Returns:
‣ __nv_atanhf(   ) returns  .
‣ __nv_atanhf(   ) returns  .
‣ __nv_atanhf(x) returns NaN for x outside interval [-1, 1].
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.16. __nv_brev
Prototype:
i32 @__nv_brev(i32 %x) 
Description:
Reverses the bit order of the 32 bit unsigned integer  .
x
Returns:
Returns the bit-reversed value of  . i.e. bit N of the return value corresponds to bit 31-N
x
of  .
x
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.17. __nv_brevll
Prototype:
i64 @__nv_brevll(i64 %x) 
Description:
Reverses the bit order of the 64 bit unsigned integer  .
x
Returns:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 12Function Reference
Returns the bit-reversed value of  . i.e. bit N of the return value corresponds to bit 63-N
x
of  .
x
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.18. __nv_byte_perm
Prototype:
i32 @__nv_byte_perm(i32 %x, i32 %y, i32 %z) 
Description:
__nv_byte_perm(x,y,s) returns a 32-bit integer consisting of four bytes from eight input
bytes provided in the two input integers   and  , as specified by a selector,  .
x y s
The input bytes are indexed as follows:
 input[0] = x<7:0>   input[1] = x<15:8>
 input[2] = x<23:16> input[3] = x<31:24>
 input[4] = y<7:0>   input[5] = y<15:8>
 input[6] = y<23:16> input[7] = y<31:24>
 
The selector indices are as follows (the upper 16-bits of the selector are not used):
 selector[0] = s<2:0>  selector[1] = s<6:4>
 selector[2] = s<10:8> selector[3] = s<14:12>
 
Returns:
The returned value r is computed to be: 
result[n] := input[selector[n]]
where   is the nth byte of r.
result[n]
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 13Function Reference
3.19. __nv_cbrt
Prototype:
double @__nv_cbrt(double %x) 
Description:
Calculate the cube root of  ,  .
x
Returns:
Returns  .
‣ __nv_cbrt(   ) returns  .
‣ __nv_cbrt(   ) returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.20. __nv_cbrtf
Prototype:
float @__nv_cbrtf(float %x) 
Description:
Calculate the cube root of  ,  .
x
Returns:
Returns  .
‣ __nv_cbrtf(   ) returns  .
‣ __nv_cbrtf(   ) returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 14Function Reference
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.21. __nv_ceil
Prototype:
double @__nv_ceil(double %x) 
Description:
Compute the smallest integer value not less than  .
x
Returns:
Returns   expressed as a floating-point number.
‣ __nv_ceil(   ) returns  .
‣ __nv_ceil(   ) returns  .
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.22. __nv_ceilf
Prototype:
float @__nv_ceilf(float %x) 
Description:
Compute the smallest integer value not less than  .
x
Returns:
Returns   expressed as a floating-point number.
‣ __nv_ceilf(   ) returns  .
‣ __nv_ceilf(   ) returns  .
Library Availability:
Compute 2.0: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 15Function Reference
Compute 3.0: Yes
Compute 3.5: Yes
3.23. __nv_clz
Prototype:
i32 @__nv_clz(i32 %x) 
Description:
Count the number of consecutive leading zero bits, starting at the most significant bit
(bit 31) of  .
x
Returns:
Returns a value between 0 and 32 inclusive representing the number of zero bits.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.24. __nv_clzll
Prototype:
i32 @__nv_clzll(i64 %x) 
Description:
Count the number of consecutive leading zero bits, starting at the most significant bit
(bit 63) of  .
x
Returns:
Returns a value between 0 and 64 inclusive representing the number of zero bits.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 16Function Reference
3.25. __nv_copysign
Prototype:
double @__nv_copysign(double %x, double %y) 
Description:
Create a floating-point value with the magnitude   and the sign of  .
x y
Returns:
Returns a value with the magnitude of   and the sign of  .
x y
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.26. __nv_copysignf
Prototype:
float @__nv_copysignf(float %x, float %y) 
Description:
Create a floating-point value with the magnitude   and the sign of  .
x y
Returns:
Returns a value with the magnitude of   and the sign of  .
x y
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.27. __nv_cos
Prototype:
double @__nv_cos(double %x) 
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 17Function Reference
Description:
Calculate the cosine of the input argument   (measured in radians).
x
Returns:
‣ __nv_cos(   ) returns 1.
‣ __nv_cos(   ) returns NaN.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.28. __nv_cosf
Prototype:
float @__nv_cosf(float %x) 
Description:
Calculate the cosine of the input argument   (measured in radians).
x
Returns:
‣ __nv_cosf(   ) returns 1.
‣ __nv_cosf(   ) returns NaN.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 18Function Reference
3.29. __nv_cosh
Prototype:
double @__nv_cosh(double %x) 
Description:
Calculate the hyperbolic cosine of the input argument  .
x
Returns:
‣ __nv_cosh(0) returns 1.
‣ __nv_cosh(   ) returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.30. __nv_coshf
Prototype:
float @__nv_coshf(float %x) 
Description:
Calculate the hyperbolic cosine of the input argument  .
x
Returns:
‣ __nv_coshf(0) returns 1.
‣ __nv_coshf(   ) returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 19Function Reference
Compute 3.5: Yes
3.31. __nv_cospi
Prototype:
double @__nv_cospi(double %x) 
Description:
Calculate the cosine of     (measured in radians), where   is the input argument.
x x
Returns:
‣ __nv_cospi(   ) returns 1.
‣ __nv_cospi(   ) returns NaN.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.32. __nv_cospif
Prototype:
float @__nv_cospif(float %x) 
Description:
Calculate the cosine of     (measured in radians), where   is the input argument.
x x
Returns:
‣ __nv_cospif(   ) returns 1.
‣ __nv_cospif(   ) returns NaN.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 20Function Reference
Compute 3.0: Yes
Compute 3.5: Yes
3.33. __nv_dadd_rd
Prototype:
double @__nv_dadd_rd(double %x, double %y) 
Description:
Adds two floating point values   and   in round-down (to negative infinity) mode.
x y
Returns:
Returns   +  .
x y
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
This operation will never be merged into a single multiply-add instruction.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.34. __nv_dadd_rn
Prototype:
double @__nv_dadd_rn(double %x, double %y) 
Description:
Adds two floating point values   and   in round-to-nearest-even mode.
x y
Returns:
Returns   +  .
x y
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
This operation will never be merged into a single multiply-add instruction.
Library Availability:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 21Function Reference
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.35. __nv_dadd_ru
Prototype:
double @__nv_dadd_ru(double %x, double %y) 
Description:
Adds two floating point values   and   in round-up (to positive infinity) mode.
x y
Returns:
Returns   +  .
x y
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
This operation will never be merged into a single multiply-add instruction.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.36. __nv_dadd_rz
Prototype:
double @__nv_dadd_rz(double %x, double %y) 
Description:
Adds two floating point values   and   in round-towards-zero mode.
x y
Returns:
Returns   +  .
x y
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
This operation will never be merged into a single multiply-add instruction.
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 22Function Reference
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.37. __nv_ddiv_rd
Prototype:
double @__nv_ddiv_rd(double %x, double %y) 
Description:
Divides two floating point values   by   in round-down (to negative infinity) mode.
x y
Returns:
Returns   /  .
x y
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Requires compute capability >= 2.0.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.38. __nv_ddiv_rn
Prototype:
double @__nv_ddiv_rn(double %x, double %y) 
Description:
Divides two floating point values   by   in round-to-nearest-even mode.
x y
Returns:
Returns   /  .
x y
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 23Function Reference
Requires compute capability >= 2.0.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.39. __nv_ddiv_ru
Prototype:
double @__nv_ddiv_ru(double %x, double %y) 
Description:
Divides two floating point values   by   in round-up (to positive infinity) mode.
x y
Returns:
Returns   /  .
x y
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Requires compute capability >= 2.0.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.40. __nv_ddiv_rz
Prototype:
double @__nv_ddiv_rz(double %x, double %y) 
Description:
Divides two floating point values   by   in round-towards-zero mode.
x y
Returns:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 24Function Reference
Returns   /  .
x y
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Requires compute capability >= 2.0.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.41. __nv_dmul_rd
Prototype:
double @__nv_dmul_rd(double %x, double %y) 
Description:
Multiplies two floating point values   and   in round-down (to negative infinity) mode.
x y
Returns:
Returns   *  .
x y
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
This operation will never be merged into a single multiply-add instruction.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.42. __nv_dmul_rn
Prototype:
double @__nv_dmul_rn(double %x, double %y) 
Description:
Multiplies two floating point values   and   in round-to-nearest-even mode.
x y
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 25Function Reference
Returns:
Returns   *  .
x y
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
This operation will never be merged into a single multiply-add instruction.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.43. __nv_dmul_ru
Prototype:
double @__nv_dmul_ru(double %x, double %y) 
Description:
Multiplies two floating point values   and   in round-up (to positive infinity) mode.
x y
Returns:
Returns   *  .
x y
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
This operation will never be merged into a single multiply-add instruction.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.44. __nv_dmul_rz
Prototype:
double @__nv_dmul_rz(double %x, double %y) 
Description:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 26Function Reference
Multiplies two floating point values   and   in round-towards-zero mode.
x y
Returns:
Returns   *  .
x y
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
This operation will never be merged into a single multiply-add instruction.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.45. __nv_double2float_rd
Prototype:
float @__nv_double2float_rd(double %d) 
Description:
Convert the double-precision floating point value   to a single-precision floating point
x
value in round-down (to negative infinity) mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.46. __nv_double2float_rn
Prototype:
float @__nv_double2float_rn(double %d) 
Description:
Convert the double-precision floating point value   to a single-precision floating point
x
value in round-to-nearest-even mode.
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 27Function Reference
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.47. __nv_double2float_ru
Prototype:
float @__nv_double2float_ru(double %d) 
Description:
Convert the double-precision floating point value   to a single-precision floating point
x
value in round-up (to positive infinity) mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.48. __nv_double2float_rz
Prototype:
float @__nv_double2float_rz(double %d) 
Description:
Convert the double-precision floating point value   to a single-precision floating point
x
value in round-towards-zero mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 28Function Reference
Compute 3.0: Yes
Compute 3.5: Yes
3.49. __nv_double2hiint
Prototype:
i32 @__nv_double2hiint(double %d) 
Description:
Reinterpret the high 32 bits in the double-precision floating point value   as a signed
x
integer.
Returns:
Returns reinterpreted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.50. __nv_double2int_rd
Prototype:
i32 @__nv_double2int_rd(double %d) 
Description:
Convert the double-precision floating point value   to a signed integer value in round-
x
down (to negative infinity) mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 29Function Reference
3.51. __nv_double2int_rn
Prototype:
i32 @__nv_double2int_rn(double %d) 
Description:
Convert the double-precision floating point value   to a signed integer value in round-
x
to-nearest-even mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.52. __nv_double2int_ru
Prototype:
i32 @__nv_double2int_ru(double %d) 
Description:
Convert the double-precision floating point value   to a signed integer value in round-
x
up (to positive infinity) mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 30Function Reference
3.53. __nv_double2int_rz
Prototype:
i32 @__nv_double2int_rz(double %d) 
Description:
Convert the double-precision floating point value   to a signed integer value in round-
x
towards-zero mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.54. __nv_double2ll_rd
Prototype:
i64 @__nv_double2ll_rd(double %f) 
Description:
Convert the double-precision floating point value   to a signed 64-bit integer value in
x
round-down (to negative infinity) mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 31Function Reference
3.55. __nv_double2ll_rn
Prototype:
i64 @__nv_double2ll_rn(double %f) 
Description:
Convert the double-precision floating point value   to a signed 64-bit integer value in
x
round-to-nearest-even mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.56. __nv_double2ll_ru
Prototype:
i64 @__nv_double2ll_ru(double %f) 
Description:
Convert the double-precision floating point value   to a signed 64-bit integer value in
x
round-up (to positive infinity) mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 32Function Reference
3.57. __nv_double2ll_rz
Prototype:
i64 @__nv_double2ll_rz(double %f) 
Description:
Convert the double-precision floating point value   to a signed 64-bit integer value in
x
round-towards-zero mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.58. __nv_double2loint
Prototype:
i32 @__nv_double2loint(double %d) 
Description:
Reinterpret the low 32 bits in the double-precision floating point value   as a signed
x
integer.
Returns:
Returns reinterpreted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 33Function Reference
3.59. __nv_double2uint_rd
Prototype:
i32 @__nv_double2uint_rd(double %d) 
Description:
Convert the double-precision floating point value   to an unsigned integer value in
x
round-down (to negative infinity) mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.60. __nv_double2uint_rn
Prototype:
i32 @__nv_double2uint_rn(double %d) 
Description:
Convert the double-precision floating point value   to an unsigned integer value in
x
round-to-nearest-even mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 34Function Reference
3.61. __nv_double2uint_ru
Prototype:
i32 @__nv_double2uint_ru(double %d) 
Description:
Convert the double-precision floating point value   to an unsigned integer value in
x
round-up (to positive infinity) mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.62. __nv_double2uint_rz
Prototype:
i32 @__nv_double2uint_rz(double %d) 
Description:
Convert the double-precision floating point value   to an unsigned integer value in
x
round-towards-zero mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 35Function Reference
3.63. __nv_double2ull_rd
Prototype:
i64 @__nv_double2ull_rd(double %f) 
Description:
Convert the double-precision floating point value   to an unsigned 64-bit integer value
x
in round-down (to negative infinity) mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.64. __nv_double2ull_rn
Prototype:
i64 @__nv_double2ull_rn(double %f) 
Description:
Convert the double-precision floating point value   to an unsigned 64-bit integer value
x
in round-to-nearest-even mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 36Function Reference
3.65. __nv_double2ull_ru
Prototype:
i64 @__nv_double2ull_ru(double %f) 
Description:
Convert the double-precision floating point value   to an unsigned 64-bit integer value
x
in round-up (to positive infinity) mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.66. __nv_double2ull_rz
Prototype:
i64 @__nv_double2ull_rz(double %f) 
Description:
Convert the double-precision floating point value   to an unsigned 64-bit integer value
x
in round-towards-zero mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 37Function Reference
3.67. __nv_double_as_longlong
Prototype:
i64 @__nv_double_as_longlong(double %x) 
Description:
Reinterpret the bits in the double-precision floating point value   as a signed 64-bit
x
integer.
Returns:
Returns reinterpreted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.68. __nv_drcp_rd
Prototype:
double @__nv_drcp_rd(double %x) 
Description:
Compute the reciprocal of   in round-down (to negative infinity) mode.
x
Returns:
Returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Requires compute capability >= 2.0.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 38Function Reference
3.69. __nv_drcp_rn
Prototype:
double @__nv_drcp_rn(double %x) 
Description:
Compute the reciprocal of   in round-to-nearest-even mode.
x
Returns:
Returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Requires compute capability >= 2.0.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.70. __nv_drcp_ru
Prototype:
double @__nv_drcp_ru(double %x) 
Description:
Compute the reciprocal of   in round-up (to positive infinity) mode.
x
Returns:
Returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Requires compute capability >= 2.0.
Library Availability:
Compute 2.0: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 39Function Reference
Compute 3.0: Yes
Compute 3.5: Yes
3.71. __nv_drcp_rz
Prototype:
double @__nv_drcp_rz(double %x) 
Description:
Compute the reciprocal of   in round-towards-zero mode.
x
Returns:
Returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Requires compute capability >= 2.0.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.72. __nv_dsqrt_rd
Prototype:
double @__nv_dsqrt_rd(double %x) 
Description:
Compute the square root of   in round-down (to negative infinity) mode.
x
Returns:
Returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Requires compute capability >= 2.0.
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 40Function Reference
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.73. __nv_dsqrt_rn
Prototype:
double @__nv_dsqrt_rn(double %x) 
Description:
Compute the square root of   in round-to-nearest-even mode.
x
Returns:
Returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Requires compute capability >= 2.0.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.74. __nv_dsqrt_ru
Prototype:
double @__nv_dsqrt_ru(double %x) 
Description:
Compute the square root of   in round-up (to positive infinity) mode.
x
Returns:
Returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 41Function Reference
Requires compute capability >= 2.0.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.75. __nv_dsqrt_rz
Prototype:
double @__nv_dsqrt_rz(double %x) 
Description:
Compute the square root of   in round-towards-zero mode.
x
Returns:
Returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Requires compute capability >= 2.0.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.76. __nv_erf
Prototype:
double @__nv_erf(double %x) 
Description:
Calculate the value of the error function for the input argument  ,  .
x
Returns:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 42Function Reference
‣ __nv_erf(   ) returns  .
‣ __nv_erf(   ) returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.77. __nv_erfc
Prototype:
double @__nv_erfc(double %x) 
Description:
Calculate the complementary error function of the input argument  , 1 - erf( ).
x x
Returns:
‣ __nv_erfc(   ) returns 2.
‣ __nv_erfc(   ) returns +0.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.78. __nv_erfcf
Prototype:
float @__nv_erfcf(float %x) 
Description:
Calculate the complementary error function of the input argument  , 1 - erf( ).
x x
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 43Function Reference
Returns:
‣ __nv_erfcf(   ) returns 2.
‣ __nv_erfcf(   ) returns +0.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.79. __nv_erfcinv
Prototype:
double @__nv_erfcinv(double %x) 
Description:
Calculate the inverse complementary error function of the input argument  , for   in the
y y
interval [0, 2]. The inverse complementary error function find the value   that satisfies
x
the equation   = erfc( ), for   , and  .
y x
Returns:
‣ __nv_erfcinv(0) returns  .
‣ __nv_erfcinv(2) returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.80. __nv_erfcinvf
Prototype:
float @__nv_erfcinvf(float %x) 
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 44Function Reference
Description:
Calculate the inverse complementary error function of the input argument  , for   in the
y y
interval [0, 2]. The inverse complementary error function find the value   that satisfies
x
the equation   = erfc( ), for   , and  .
y x
Returns:
‣ __nv_erfcinvf(0) returns  .
‣ __nv_erfcinvf(2) returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.81. __nv_erfcx
Prototype:
double @__nv_erfcx(double %x) 
Description:
Calculate the scaled complementary error function of the input argument  ,  .
x
Returns:
‣ __nv_erfcx(   ) returns 
‣ __nv_erfcx(   ) returns +0
‣ __nv_erfcx(x) returns   if the correctly calculated value is outside the double
floating point range.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 45Function Reference
3.82. __nv_erfcxf
Prototype:
float @__nv_erfcxf(float %x) 
Description:
Calculate the scaled complementary error function of the input argument  ,  .
x
Returns:
‣ __nv_erfcxf(   ) returns 
‣ __nv_erfcxf(   ) returns +0
‣ __nv_erfcxf(x) returns   if the correctly calculated value is outside the double
floating point range.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.83. __nv_erff
Prototype:
float @__nv_erff(float %x) 
Description:
Calculate the value of the error function for the input argument  ,  .
x
Returns:
‣ __nv_erff(   ) returns  .
‣ __nv_erff(   ) returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 46Function Reference
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.84. __nv_erfinv
Prototype:
double @__nv_erfinv(double %x) 
Description:
Calculate the inverse error function of the input argument  , for   in the interval [-1,
y y
1]. The inverse error function finds the value   that satisfies the equation   = erf( ), for
x y x
 , and  .
Returns:
‣ __nv_erfinv(1) returns  .
‣ __nv_erfinv(-1) returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.85. __nv_erfinvf
Prototype:
float @__nv_erfinvf(float %x) 
Description:
Calculate the inverse error function of the input argument  , for   in the interval [-1,
y y
1]. The inverse error function finds the value   that satisfies the equation   = erf( ), for
x y x
 , and  .
Returns:
‣ __nv_erfinvf(1) returns  .
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 47Function Reference
‣ __nv_erfinvf(-1) returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.86. __nv_exp
Prototype:
double @__nv_exp(double %x) 
Description:
Calculate the base   exponential of the input argument  .
x
Returns:
Returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.87. __nv_exp10
Prototype:
double @__nv_exp10(double %x) 
Description:
Calculate the base 10 exponential of the input argument  .
x
Returns:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 48Function Reference
Returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.88. __nv_exp10f
Prototype:
float @__nv_exp10f(float %x) 
Description:
Calculate the base 10 exponential of the input argument  .
x
Returns:
Returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.89. __nv_exp2
Prototype:
double @__nv_exp2(double %x) 
Description:
Calculate the base 2 exponential of the input argument  .
x
Returns:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 49Function Reference
Returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.90. __nv_exp2f
Prototype:
float @__nv_exp2f(float %x) 
Description:
Calculate the base 2 exponential of the input argument  .
x
Returns:
Returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.91. __nv_expf
Prototype:
float @__nv_expf(float %x) 
Description:
Calculate the base   exponential of the input argument  .
x
Returns:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 50Function Reference
Returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.92. __nv_expm1
Prototype:
double @__nv_expm1(double %x) 
Description:
Calculate the base   exponential of the input argument  , minus 1.
x
Returns:
Returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.93. __nv_expm1f
Prototype:
float @__nv_expm1f(float %x) 
Description:
Calculate the base   exponential of the input argument  , minus 1.
x
Returns:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 51Function Reference
Returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.94. __nv_fabs
Prototype:
double @__nv_fabs(double %f) 
Description:
Calculate the absolute value of the input argument  .
x
Returns:
Returns the absolute value of the input argument.
‣ __nv_fabs(   ) returns  .
‣ __nv_fabs(   ) returns 0.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.95. __nv_fabsf
Prototype:
float @__nv_fabsf(float %f) 
Description:
Calculate the absolute value of the input argument  .
x
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 52Function Reference
Returns:
Returns the absolute value of the input argument.
‣ __nv_fabsf(   ) returns  .
‣ __nv_fabsf(   ) returns 0.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.96. __nv_fadd_rd
Prototype:
float @__nv_fadd_rd(float %x, float %y) 
Description:
Compute the sum of   and   in round-down (to negative infinity) mode.
x y
Returns:
Returns   +  .
x y
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
This operation will never be merged into a single multiply-add instruction.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.97. __nv_fadd_rn
Prototype:
float @__nv_fadd_rn(float %x, float %y) 
Description:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 53Function Reference
Compute the sum of   and   in round-to-nearest-even rounding mode.
x y
Returns:
Returns   +  .
x y
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
This operation will never be merged into a single multiply-add instruction.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.98. __nv_fadd_ru
Prototype:
float @__nv_fadd_ru(float %x, float %y) 
Description:
Compute the sum of   and   in round-up (to positive infinity) mode.
x y
Returns:
Returns   +  .
x y
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
This operation will never be merged into a single multiply-add instruction.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.99. __nv_fadd_rz
Prototype:
float @__nv_fadd_rz(float %x, float %y) 
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 54Function Reference
Description:
Compute the sum of   and   in round-towards-zero mode.
x y
Returns:
Returns   +  .
x y
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
This operation will never be merged into a single multiply-add instruction.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.100. __nv_fast_cosf
Prototype:
float @__nv_fast_cosf(float %x) 
Description:
Calculate the fast approximate cosine of the input argument  , measured in radians.
x
Returns:
Returns the approximate cosine of  .
x
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.2, Table 9.
Input and output in the denormal range is flushed to sign preserving 0.0.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 55Function Reference
3.101. __nv_fast_exp10f
Prototype:
float @__nv_fast_exp10f(float %x) 
Description:
Calculate the fast approximate base 10 exponential of the input argument  ,  .
x
Returns:
Returns an approximation to  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.2, Table 9.
Most input and output values around denormal range are flushed to sign preserving
0.0.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.102. __nv_fast_expf
Prototype:
float @__nv_fast_expf(float %x) 
Description:
Calculate the fast approximate base   exponential of the input argument  ,  .
x
Returns:
Returns an approximation to  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.2, Table 9.
Most input and output values around denormal range are flushed to sign preserving
0.0.
Library Availability:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 56Function Reference
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.103. __nv_fast_fdividef
Prototype:
float @__nv_fast_fdividef(float %x, float %y) 
Description:
Calculate the fast approximate division of   by  .
x y
Returns:
Returns   /  .
x y
‣ __nv_fast_fdividef(   ,  ) returns NaN for  .
y
‣ __nv_fast_fdividef( ,  ) returns 0 for   and  .
x y
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.2, Table 9.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.104. __nv_fast_log10f
Prototype:
float @__nv_fast_log10f(float %x) 
Description:
Calculate the fast approximate base 10 logarithm of the input argument  .
x
Returns:
Returns an approximation to  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.2, Table 9.
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 57Function Reference
Most input and output values around denormal range are flushed to sign preserving
0.0.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.105. __nv_fast_log2f
Prototype:
float @__nv_fast_log2f(float %x) 
Description:
Calculate the fast approximate base 2 logarithm of the input argument  .
x
Returns:
Returns an approximation to  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.2, Table 9.
Input and output in the denormal range is flushed to sign preserving 0.0.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.106. __nv_fast_logf
Prototype:
float @__nv_fast_logf(float %x) 
Description:
Calculate the fast approximate base   logarithm of the input argument  .
x
Returns:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 58Function Reference
Returns an approximation to  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.2, Table 9.
Most input and output values around denormal range are flushed to sign preserving
0.0.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.107. __nv_fast_powf
Prototype:
float @__nv_fast_powf(float %x, float %y) 
Description:
Calculate the fast approximate of  , the first input argument, raised to the power of  ,
x y
the second input argument,  .
Returns:
Returns an approximation to  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.2, Table 9.
Most input and output values around denormal range are flushed to sign preserving
0.0.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.108. __nv_fast_sincosf
Prototype:
void @__nv_fast_sincosf(float %x, float* %sptr, float* %cptr) 
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 59Function Reference
Description:
Calculate the fast approximate of sine and cosine of the first input argument 
x
(measured in radians). The results for sine and cosine are written into the second
argument,  , and, respectively, third argument,  .
sptr zptr
Returns:
‣ none
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.2, Table 9.
Denorm input/output is flushed to sign preserving 0.0.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.109. __nv_fast_sinf
Prototype:
float @__nv_fast_sinf(float %x) 
Description:
Calculate the fast approximate sine of the input argument  , measured in radians.
x
Returns:
Returns the approximate sine of  .
x
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.2, Table 9.
Input and output in the denormal range is flushed to sign preserving 0.0.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 60Function Reference
3.110. __nv_fast_tanf
Prototype:
float @__nv_fast_tanf(float %x) 
Description:
Calculate the fast approximate tangent of the input argument  , measured in radians.
x
Returns:
Returns the approximate tangent of  .
x
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.2, Table 9.
The result is computed as the fast divide of __nv_sinf() by __nv_cosf(). Denormal
input and output are flushed to sign-preserving 0.0 at each step of the computation.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.111. __nv_fdim
Prototype:
double @__nv_fdim(double %x, double %y) 
Description:
Compute the positive difference between   and  . The positive difference is   -   when 
x y x y x
>   and +0 otherwise.
y
Returns:
Returns the positive difference between   and  .
x y
‣ __nv_fdim(x, y) returns x - y if x > y.
‣ __nv_fdim(x, y) returns +0 if x   y.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 61Function Reference
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.112. __nv_fdimf
Prototype:
float @__nv_fdimf(float %x, float %y) 
Description:
Compute the positive difference between   and  . The positive difference is   -   when 
x y x y x
>   and +0 otherwise.
y
Returns:
Returns the positive difference between   and  .
x y
‣ __nv_fdimf(x, y) returns x - y if x > y.
‣ __nv_fdimf(x, y) returns +0 if x   y.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.113. __nv_fdiv_rd
Prototype:
float @__nv_fdiv_rd(float %x, float %y) 
Description:
Divide two floating point values   by   in round-down (to negative infinity) mode.
x y
Returns:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 62Function Reference
Returns   /  .
x y
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.114. __nv_fdiv_rn
Prototype:
float @__nv_fdiv_rn(float %x, float %y) 
Description:
Divide two floating point values   by   in round-to-nearest-even mode.
x y
Returns:
Returns   /  .
x y
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.115. __nv_fdiv_ru
Prototype:
float @__nv_fdiv_ru(float %x, float %y) 
Description:
Divide two floating point values   by   in round-up (to positive infinity) mode.
x y
Returns:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 63Function Reference
Returns   /  .
x y
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.116. __nv_fdiv_rz
Prototype:
float @__nv_fdiv_rz(float %x, float %y) 
Description:
Divide two floating point values   by   in round-towards-zero mode.
x y
Returns:
Returns   /  .
x y
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.117. __nv_ffs
Prototype:
i32 @__nv_ffs(i32 %x) 
Description:
Find the position of the first (least significant) bit set to 1 in  , where the least significant
x
bit position is 1.
Returns:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 64Function Reference
Returns a value between 0 and 32 inclusive representing the position of the first bit set.
‣ __nv_ffs(0) returns 0.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.118. __nv_ffsll
Prototype:
i32 @__nv_ffsll(i64 %x) 
Description:
Find the position of the first (least significant) bit set to 1 in  , where the least significant
x
bit position is 1.
Returns:
Returns a value between 0 and 64 inclusive representing the position of the first bit set.
‣ __nv_ffsll(0) returns 0.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.119. __nv_finitef
Prototype:
i32 @__nv_finitef(float %x) 
Description:
Determine whether the floating-point value   is a finite value.
x
Returns:
Returns a non-zero value if and only if   is a finite value.
x
Library Availability:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 65Function Reference
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.120. __nv_float2half_rn
Prototype:
i16 @__nv_float2half_rn(float %f) 
Description:
Convert the single-precision float value   to a half-precision floating point value
x
represented in   format, in round-to-nearest-even mode.
unsigned short
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.121. __nv_float2int_rd
Prototype:
i32 @__nv_float2int_rd(float %in) 
Description:
Convert the single-precision floating point value   to a signed integer in round-down (to
x
negative infinity) mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 66Function Reference
3.122. __nv_float2int_rn
Prototype:
i32 @__nv_float2int_rn(float %in) 
Description:
Convert the single-precision floating point value   to a signed integer in round-to-
x
nearest-even mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.123. __nv_float2int_ru
Prototype:
i32 @__nv_float2int_ru(float %in) 
Description:
Convert the single-precision floating point value   to a signed integer in round-up (to
x
positive infinity) mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 67Function Reference
3.124. __nv_float2int_rz
Prototype:
i32 @__nv_float2int_rz(float %in) 
Description:
Convert the single-precision floating point value   to a signed integer in round-towards-
x
zero mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.125. __nv_float2ll_rd
Prototype:
i64 @__nv_float2ll_rd(float %f) 
Description:
Convert the single-precision floating point value   to a signed 64-bit integer in round-
x
down (to negative infinity) mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 68Function Reference
3.126. __nv_float2ll_rn
Prototype:
i64 @__nv_float2ll_rn(float %f) 
Description:
Convert the single-precision floating point value   to a signed 64-bit integer in round-to-
x
nearest-even mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.127. __nv_float2ll_ru
Prototype:
i64 @__nv_float2ll_ru(float %f) 
Description:
Convert the single-precision floating point value   to a signed 64-bit integer in round-up
x
(to positive infinity) mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 69Function Reference
3.128. __nv_float2ll_rz
Prototype:
i64 @__nv_float2ll_rz(float %f) 
Description:
Convert the single-precision floating point value   to a signed 64-bit integer in round-
x
towards-zero mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.129. __nv_float2uint_rd
Prototype:
i32 @__nv_float2uint_rd(float %in) 
Description:
Convert the single-precision floating point value   to an unsigned integer in round-
x
down (to negative infinity) mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 70Function Reference
3.130. __nv_float2uint_rn
Prototype:
i32 @__nv_float2uint_rn(float %in) 
Description:
Convert the single-precision floating point value   to an unsigned integer in round-to-
x
nearest-even mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.131. __nv_float2uint_ru
Prototype:
i32 @__nv_float2uint_ru(float %in) 
Description:
Convert the single-precision floating point value   to an unsigned integer in round-up
x
(to positive infinity) mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 71Function Reference
3.132. __nv_float2uint_rz
Prototype:
i32 @__nv_float2uint_rz(float %in) 
Description:
Convert the single-precision floating point value   to an unsigned integer in round-
x
towards-zero mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.133. __nv_float2ull_rd
Prototype:
i64 @__nv_float2ull_rd(float %f) 
Description:
Convert the single-precision floating point value   to an unsigned 64-bit integer in
x
round-down (to negative infinity) mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 72Function Reference
3.134. __nv_float2ull_rn
Prototype:
i64 @__nv_float2ull_rn(float %f) 
Description:
Convert the single-precision floating point value   to an unsigned 64-bit integer in
x
round-to-nearest-even mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.135. __nv_float2ull_ru
Prototype:
i64 @__nv_float2ull_ru(float %f) 
Description:
Convert the single-precision floating point value   to an unsigned 64-bit integer in
x
round-up (to positive infinity) mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 73Function Reference
3.136. __nv_float2ull_rz
Prototype:
i64 @__nv_float2ull_rz(float %f) 
Description:
Convert the single-precision floating point value   to an unsigned 64-bit integer in
x
round-towards_zero mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.137. __nv_float_as_int
Prototype:
i32 @__nv_float_as_int(float %x) 
Description:
Reinterpret the bits in the single-precision floating point value   as a signed integer.
x
Returns:
Returns reinterpreted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.138. __nv_floor
Prototype:
double @__nv_floor(double %f) 
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 74Function Reference
Description:
Calculates the largest integer value which is less than or equal to  .
x
Returns:
Returns the largest integer value which is less than or equal to x expressed as a floating-
point number.
‣ __nv_floor(   ) returns  .
‣ __nv_floor(   ) returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.139. __nv_floorf
Prototype:
float @__nv_floorf(float %f) 
Description:
Calculates the largest integer value which is less than or equal to  .
x
Returns:
Returns the largest integer value which is less than or equal to x expressed as a floating-
point number.
‣ __nv_floorf(   ) returns  .
‣ __nv_floorf(   ) returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 75Function Reference
3.140. __nv_fma
Prototype:
double @__nv_fma(double %x, double %y, double %z) 
Description:
Compute the value of   as a single ternary operation. After computing the value
to infinite precision, the value is rounded once.
Returns:
Returns the rounded value of   as a single operation.
‣ __nv_fma(   ,   , z) returns NaN.
‣ __nv_fma(   ,   , z) returns NaN.
‣ __nv_fma(x, y,   ) returns NaN if   is an exact  .
‣ __nv_fma(x, y,   ) returns NaN if   is an exact  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.141. __nv_fma_rd
Prototype:
double @__nv_fma_rd(double %x, double %y, double %z) 
Description:
Computes the value of   as a single ternary operation, rounding the result once
in round-down (to negative infinity) mode.
Returns:
Returns the rounded value of   as a single operation.
‣ __nv_fma_rd(   ,   , z) returns NaN.
‣ __nv_fma_rd(   ,   , z) returns NaN.
‣ __nv_fma_rd(x, y,   ) returns NaN if   is an exact 
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 76Function Reference
‣ __nv_fma_rd(x, y,   ) returns NaN if   is an exact 
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.142. __nv_fma_rn
Prototype:
double @__nv_fma_rn(double %x, double %y, double %z) 
Description:
Computes the value of   as a single ternary operation, rounding the result once
in round-to-nearest-even mode.
Returns:
Returns the rounded value of   as a single operation.
‣ __nv_fma_rn(   ,   , z) returns NaN.
‣ __nv_fma_rn(   ,   , z) returns NaN.
‣ __nv_fma_rn(x, y,   ) returns NaN if   is an exact 
‣ __nv_fma_rn(x, y,   ) returns NaN if   is an exact 
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.143. __nv_fma_ru
Prototype:
double @__nv_fma_ru(double %x, double %y, double %z) 
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 77Function Reference
Description:
Computes the value of   as a single ternary operation, rounding the result once
in round-up (to positive infinity) mode.
Returns:
Returns the rounded value of   as a single operation.
‣ __nv_fma_ru(   ,   , z) returns NaN.
‣ __nv_fma_ru(   ,   , z) returns NaN.
‣ __nv_fma_ru(x, y,   ) returns NaN if   is an exact 
‣ __nv_fma_ru(x, y,   ) returns NaN if   is an exact 
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.144. __nv_fma_rz
Prototype:
double @__nv_fma_rz(double %x, double %y, double %z) 
Description:
Computes the value of   as a single ternary operation, rounding the result once
in round-towards-zero mode.
Returns:
Returns the rounded value of   as a single operation.
‣ __nv_fma_rz(   ,   , z) returns NaN.
‣ __nv_fma_rz(   ,   , z) returns NaN.
‣ __nv_fma_rz(x, y,   ) returns NaN if   is an exact 
‣ __nv_fma_rz(x, y,   ) returns NaN if   is an exact 
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 78Function Reference
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.145. __nv_fmaf
Prototype:
float @__nv_fmaf(float %x, float %y, float %z) 
Description:
Compute the value of   as a single ternary operation. After computing the value
to infinite precision, the value is rounded once.
Returns:
Returns the rounded value of   as a single operation.
‣ __nv_fmaf(   ,   , z) returns NaN.
‣ __nv_fmaf(   ,   , z) returns NaN.
‣ __nv_fmaf(x, y,   ) returns NaN if   is an exact  .
‣ __nv_fmaf(x, y,   ) returns NaN if   is an exact  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.146. __nv_fmaf_rd
Prototype:
float @__nv_fmaf_rd(float %x, float %y, float %z) 
Description:
Computes the value of   as a single ternary operation, rounding the result once
in round-down (to negative infinity) mode.
Returns:
Returns the rounded value of   as a single operation.
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 79Function Reference
‣ __nv_fmaf_rd(   ,   , z) returns NaN.
‣ __nv_fmaf_rd(   ,   , z) returns NaN.
‣ __nv_fmaf_rd(x, y,   ) returns NaN if   is an exact  .
‣ __nv_fmaf_rd(x, y,   ) returns NaN if   is an exact  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.147. __nv_fmaf_rn
Prototype:
float @__nv_fmaf_rn(float %x, float %y, float %z) 
Description:
Computes the value of   as a single ternary operation, rounding the result once
in round-to-nearest-even mode.
Returns:
Returns the rounded value of   as a single operation.
‣ __nv_fmaf_rn(   ,   , z) returns NaN.
‣ __nv_fmaf_rn(   ,   , z) returns NaN.
‣ __nv_fmaf_rn(x, y,   ) returns NaN if   is an exact  .
‣ __nv_fmaf_rn(x, y,   ) returns NaN if   is an exact  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 80Function Reference
3.148. __nv_fmaf_ru
Prototype:
float @__nv_fmaf_ru(float %x, float %y, float %z) 
Description:
Computes the value of   as a single ternary operation, rounding the result once
in round-up (to positive infinity) mode.
Returns:
Returns the rounded value of   as a single operation.
‣ __nv_fmaf_ru(   ,   , z) returns NaN.
‣ __nv_fmaf_ru(   ,   , z) returns NaN.
‣ __nv_fmaf_ru(x, y,   ) returns NaN if   is an exact  .
‣ __nv_fmaf_ru(x, y,   ) returns NaN if   is an exact  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.149. __nv_fmaf_rz
Prototype:
float @__nv_fmaf_rz(float %x, float %y, float %z) 
Description:
Computes the value of   as a single ternary operation, rounding the result once
in round-towards-zero mode.
Returns:
Returns the rounded value of   as a single operation.
‣ __nv_fmaf_rz(   ,   , z) returns NaN.
‣ __nv_fmaf_rz(   ,   , z) returns NaN.
‣ __nv_fmaf_rz(x, y,   ) returns NaN if   is an exact  .
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 81Function Reference
‣ __nv_fmaf_rz(x, y,   ) returns NaN if   is an exact  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.150. __nv_fmax
Prototype:
double @__nv_fmax(double %x, double %y) 
Description:
Determines the maximum numeric value of the arguments   and  . Treats NaN
x y
arguments as missing data. If one argument is a NaN and the other is legitimate numeric
value, the numeric value is chosen.
Returns:
Returns the maximum numeric values of the arguments   and  .
x y
‣ If both arguments are NaN, returns NaN.
‣ If one argument is NaN, returns the numeric argument.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.151. __nv_fmaxf
Prototype:
float @__nv_fmaxf(float %x, float %y) 
Description:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 82Function Reference
Determines the maximum numeric value of the arguments   and  . Treats NaN
x y
arguments as missing data. If one argument is a NaN and the other is legitimate numeric
value, the numeric value is chosen.
Returns:
Returns the maximum numeric values of the arguments   and  .
x y
‣ If both arguments are NaN, returns NaN.
‣ If one argument is NaN, returns the numeric argument.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.152. __nv_fmin
Prototype:
double @__nv_fmin(double %x, double %y) 
Description:
Determines the minimum numeric value of the arguments   and  . Treats NaN
x y
arguments as missing data. If one argument is a NaN and the other is legitimate numeric
value, the numeric value is chosen.
Returns:
Returns the minimum numeric values of the arguments   and  .
x y
‣ If both arguments are NaN, returns NaN.
‣ If one argument is NaN, returns the numeric argument.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 83Function Reference
3.153. __nv_fminf
Prototype:
float @__nv_fminf(float %x, float %y) 
Description:
Determines the minimum numeric value of the arguments   and  . Treats NaN
x y
arguments as missing data. If one argument is a NaN and the other is legitimate numeric
value, the numeric value is chosen.
Returns:
Returns the minimum numeric values of the arguments   and  .
x y
‣ If both arguments are NaN, returns NaN.
‣ If one argument is NaN, returns the numeric argument.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.154. __nv_fmod
Prototype:
double @__nv_fmod(double %x, double %y) 
Description:
Calculate the floating-point remainder of   /  . The absolute value of the computed
x y
value is always less than   absolute value and will have the same sign as  .
y's x
Returns:
‣ Returns the floating point remainder of x / y.
‣ __nv_fmod(   , y) returns   if y is not zero.
‣ __nv_fmod(x, y) returns NaN and raised an invalid floating point exception if x is
 or   is zero.
y
‣ __nv_fmod(x, y) returns zero if y is zero or the result would overflow.
‣ __nv_fmod(x,   ) returns x if x is finite.
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 84Function Reference
‣ __nv_fmod(x, 0) returns NaN.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.155. __nv_fmodf
Prototype:
float @__nv_fmodf(float %x, float %y) 
Description:
Calculate the floating-point remainder of   /  . The absolute value of the computed
x y
value is always less than   absolute value and will have the same sign as  .
y's x
Returns:
‣ Returns the floating point remainder of x / y.
‣ __nv_fmodf(   , y) returns   if y is not zero.
‣ __nv_fmodf(x, y) returns NaN and raised an invalid floating point exception if x is
 or   is zero.
y
‣ __nv_fmodf(x, y) returns zero if y is zero or the result would overflow.
‣ __nv_fmodf(x,   ) returns x if x is finite.
‣ __nv_fmodf(x, 0) returns NaN.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 85Function Reference
3.156. __nv_fmul_rd
Prototype:
float @__nv_fmul_rd(float %x, float %y) 
Description:
Compute the product of   and   in round-down (to negative infinity) mode.
x y
Returns:
Returns   *  .
x y
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
This operation will never be merged into a single multiply-add instruction.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.157. __nv_fmul_rn
Prototype:
float @__nv_fmul_rn(float %x, float %y) 
Description:
Compute the product of   and   in round-to-nearest-even mode.
x y
Returns:
Returns   *  .
x y
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
This operation will never be merged into a single multiply-add instruction.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 86Function Reference
Compute 3.5: Yes
3.158. __nv_fmul_ru
Prototype:
float @__nv_fmul_ru(float %x, float %y) 
Description:
Compute the product of   and   in round-up (to positive infinity) mode.
x y
Returns:
Returns   *  .
x y
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
This operation will never be merged into a single multiply-add instruction.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.159. __nv_fmul_rz
Prototype:
float @__nv_fmul_rz(float %x, float %y) 
Description:
Compute the product of   and   in round-towards-zero mode.
x y
Returns:
Returns   *  .
x y
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
This operation will never be merged into a single multiply-add instruction.
Library Availability:
Compute 2.0: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 87Function Reference
Compute 3.0: Yes
Compute 3.5: Yes
3.160. __nv_frcp_rd
Prototype:
float @__nv_frcp_rd(float %x) 
Description:
Compute the reciprocal of   in round-down (to negative infinity) mode.
x
Returns:
Returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.161. __nv_frcp_rn
Prototype:
float @__nv_frcp_rn(float %x) 
Description:
Compute the reciprocal of   in round-to-nearest-even mode.
x
Returns:
Returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 88Function Reference
Compute 3.0: Yes
Compute 3.5: Yes
3.162. __nv_frcp_ru
Prototype:
float @__nv_frcp_ru(float %x) 
Description:
Compute the reciprocal of   in round-up (to positive infinity) mode.
x
Returns:
Returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.163. __nv_frcp_rz
Prototype:
float @__nv_frcp_rz(float %x) 
Description:
Compute the reciprocal of   in round-towards-zero mode.
x
Returns:
Returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 89Function Reference
Compute 3.0: Yes
Compute 3.5: Yes
3.164. __nv_frexp
Prototype:
double @__nv_frexp(double %x, i32* %b) 
Description:
Decompose the floating-point value   into a component   for the normalized fraction
x m
element and another term   for the exponent. The absolute value of   will be greater
n m
than or equal to 0.5 and less than 1.0 or it will be equal to 0;  . The integer
exponent   will be stored in the location to which   points.
n nptr
Returns:
Returns the fractional component  .
m
‣ __nv_frexp(0, nptr) returns 0 for the fractional component and zero for the integer
component.
‣ __nv_frexp(   , nptr) returns   and stores zero in the location pointed to by
.
nptr
‣ __nv_frexp(   , nptr) returns   and stores an unspecified value in the location
to which   points.
nptr
‣ __nv_frexp(NaN, y) returns a NaN and stores an unspecified value in the location to
which   points.
nptr
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.165. __nv_frexpf
Prototype:
float @__nv_frexpf(float %x, i32* %b) 
Description:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 90Function Reference
Decompose the floating-point value   into a component   for the normalized fraction
x m
element and another term   for the exponent. The absolute value of   will be greater
n m
than or equal to 0.5 and less than 1.0 or it will be equal to 0;  . The integer
exponent   will be stored in the location to which   points.
n nptr
Returns:
Returns the fractional component  .
m
‣ __nv_frexpf(0, nptr) returns 0 for the fractional component and zero for the integer
component.
‣ __nv_frexpf(   , nptr) returns   and stores zero in the location pointed to by
.
nptr
‣ __nv_frexpf(   , nptr) returns   and stores an unspecified value in the
location to which   points.
nptr
‣ __nv_frexpf(NaN, y) returns a NaN and stores an unspecified value in the location
to which   points.
nptr
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.166. __nv_frsqrt_rn
Prototype:
float @__nv_frsqrt_rn(float %x) 
Description:
Compute the reciprocal square root of   in round-to-nearest-even mode.
x
Returns:
Returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 91Function Reference
Compute 3.5: Yes
3.167. __nv_fsqrt_rd
Prototype:
float @__nv_fsqrt_rd(float %x) 
Description:
Compute the square root of   in round-down (to negative infinity) mode.
x
Returns:
Returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.168. __nv_fsqrt_rn
Prototype:
float @__nv_fsqrt_rn(float %x) 
Description:
Compute the square root of   in round-to-nearest-even mode.
x
Returns:
Returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 92Function Reference
Compute 3.5: Yes
3.169. __nv_fsqrt_ru
Prototype:
float @__nv_fsqrt_ru(float %x) 
Description:
Compute the square root of   in round-up (to positive infinity) mode.
x
Returns:
Returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.170. __nv_fsqrt_rz
Prototype:
float @__nv_fsqrt_rz(float %x) 
Description:
Compute the square root of   in round-towards-zero mode.
x
Returns:
Returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 93Function Reference
Compute 3.5: Yes
3.171. __nv_fsub_rd
Prototype:
float @__nv_fsub_rd(float %x, float %y) 
Description:
Compute the difference of   and   in round-down (to negative infinity) mode.
x y
Returns:
Returns   -  .
x y
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
This operation will never be merged into a single multiply-add instruction.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.172. __nv_fsub_rn
Prototype:
float @__nv_fsub_rn(float %x, float %y) 
Description:
Compute the difference of   and   in round-to-nearest-even rounding mode.
x y
Returns:
Returns   -  .
x y
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
This operation will never be merged into a single multiply-add instruction.
Library Availability:
Compute 2.0: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 94Function Reference
Compute 3.0: Yes
Compute 3.5: Yes
3.173. __nv_fsub_ru
Prototype:
float @__nv_fsub_ru(float %x, float %y) 
Description:
Compute the difference of   and   in round-up (to positive infinity) mode.
x y
Returns:
Returns   -  .
x y
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
This operation will never be merged into a single multiply-add instruction.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.174. __nv_fsub_rz
Prototype:
float @__nv_fsub_rz(float %x, float %y) 
Description:
Compute the difference of   and   in round-towards-zero mode.
x y
Returns:
Returns   -  .
x y
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
This operation will never be merged into a single multiply-add instruction.
Library Availability:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 95Function Reference
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.175. __nv_hadd
Prototype:
i32 @__nv_hadd(i32 %x, i32 %y) 
Description:
Compute average of signed input arguments   and   as (   +   ) >> 1, avoiding overflow
x y x y
in the intermediate sum.
Returns:
Returns a signed integer value representing the signed average value of the two inputs.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.176. __nv_half2float
Prototype:
float @__nv_half2float(i16 %h) 
Description:
Convert the half-precision floating point value   represented in 
x unsigned short
format to a single-precision floating point value.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 96Function Reference
3.177. __nv_hiloint2double
Prototype:
double @__nv_hiloint2double(i32 %x, i32 %y) 
Description:
Reinterpret the integer value of   as the high 32 bits of a double-precision floating
hi
point value and the integer value of   as the low 32 bits of the same double-precision
lo
floating point value.
Returns:
Returns reinterpreted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.178. __nv_hypot
Prototype:
double @__nv_hypot(double %x, double %y) 
Description:
Calculate the length of the hypotenuse of a right triangle whose two sides have lengths 
x
and   without undue overflow or underflow.
y
Returns:
Returns the length of the hypotenuse  . If the correct value would overflow,
returns  . If the correct value would underflow, returns 0. If one of the input
arguments is 0, returns the other argument
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 97Function Reference
Compute 3.5: Yes
3.179. __nv_hypotf
Prototype:
float @__nv_hypotf(float %x, float %y) 
Description:
Calculate the length of the hypotenuse of a right triangle whose two sides have lengths 
x
and   without undue overflow or underflow.
y
Returns:
Returns the length of the hypotenuse  . If the correct value would overflow,
returns  . If the correct value would underflow, returns 0. If one of the input
arguments is 0, returns the other argument
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.180. __nv_ilogb
Prototype:
i32 @__nv_ilogb(double %x) 
Description:
Calculates the unbiased integer exponent of the input argument  .
x
Returns:
‣ If successful, returns the unbiased exponent of the argument.
‣ __nv_ilogb(0) returns INT_MIN.
‣ __nv_ilogb(NaN) returns NaN.
‣ __nv_ilogb(x) returns INT_MAX if x is   or the correct value is greater than
.
INT_MAX
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 98Function Reference
‣ __nv_ilogb(x) return INT_MIN if the correct value is less than INT_MIN.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.181. __nv_ilogbf
Prototype:
i32 @__nv_ilogbf(float %x) 
Description:
Calculates the unbiased integer exponent of the input argument  .
x
Returns:
‣ If successful, returns the unbiased exponent of the argument.
‣ __nv_ilogbf(0) returns INT_MIN.
‣ __nv_ilogbf(NaN) returns NaN.
‣ __nv_ilogbf(x) returns INT_MAX if x is   or the correct value is greater than
.
INT_MAX
‣ __nv_ilogbf(x) return INT_MIN if the correct value is less than INT_MIN.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.182. __nv_int2double_rn
Prototype:
double @__nv_int2double_rn(i32 %i) 
Description:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 99Function Reference
Convert the signed integer value   to a double-precision floating point value.
x
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.183. __nv_int2float_rd
Prototype:
float @__nv_int2float_rd(i32 %in) 
Description:
Convert the signed integer value   to a single-precision floating point value in round-
x
down (to negative infinity) mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.184. __nv_int2float_rn
Prototype:
float @__nv_int2float_rn(i32 %in) 
Description:
Convert the signed integer value   to a single-precision floating point value in round-to-
x
nearest-even mode.
Returns:
Returns converted value.
Library Availability:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 100Function Reference
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.185. __nv_int2float_ru
Prototype:
float @__nv_int2float_ru(i32 %in) 
Description:
Convert the signed integer value   to a single-precision floating point value in round-up
x
(to positive infinity) mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.186. __nv_int2float_rz
Prototype:
float @__nv_int2float_rz(i32 %in) 
Description:
Convert the signed integer value   to a single-precision floating point value in round-
x
towards-zero mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 101Function Reference
3.187. __nv_int_as_float
Prototype:
float @__nv_int_as_float(i32 %x) 
Description:
Reinterpret the bits in the signed integer value   as a single-precision floating point
x
value.
Returns:
Returns reinterpreted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.188. __nv_isfinited
Prototype:
i32 @__nv_isfinited(double %x) 
Description:
Determine whether the floating-point value   is a finite value (zero, subnormal, or
x
normal and not infinity or NaN).
Returns:
Returns a nonzero value if and only if   is a finite value.
x
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 102Function Reference
3.189. __nv_isinfd
Prototype:
i32 @__nv_isinfd(double %x) 
Description:
Determine whether the floating-point value   is an infinite value (positive or negative).
x
Returns:
Returns a nonzero value if and only if   is a infinite value.
x
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.190. __nv_isinff
Prototype:
i32 @__nv_isinff(float %x) 
Description:
Determine whether the floating-point value   is an infinite value (positive or negative).
x
Returns:
Returns a nonzero value if and only if   is a infinite value.
x
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.191. __nv_isnand
Prototype:
i32 @__nv_isnand(double %x) 
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 103Function Reference
Description:
Determine whether the floating-point value   is a NaN.
x
Returns:
Returns a nonzero value if and only if   is a NaN value.
x
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.192. __nv_isnanf
Prototype:
i32 @__nv_isnanf(float %x) 
Description:
Determine whether the floating-point value   is a NaN.
x
Returns:
Returns a nonzero value if and only if   is a NaN value.
x
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.193. __nv_j0
Prototype:
double @__nv_j0(double %x) 
Description:
Calculate the value of the Bessel function of the first kind of order 0 for the input
argument  ,  .
x
Returns:
Returns the value of the Bessel function of the first kind of order 0.
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 104Function Reference
‣ __nv_j0(   ) returns +0.
‣ __nv_j0(NaN) returns NaN.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.194. __nv_j0f
Prototype:
float @__nv_j0f(float %x) 
Description:
Calculate the value of the Bessel function of the first kind of order 0 for the input
argument  ,  .
x
Returns:
Returns the value of the Bessel function of the first kind of order 0.
‣ __nv_j0f(   ) returns +0.
‣ __nv_j0f(NaN) returns NaN.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.195. __nv_j1
Prototype:
double @__nv_j1(double %x) 
Description:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 105Function Reference
Calculate the value of the Bessel function of the first kind of order 1 for the input
argument  ,  .
x
Returns:
Returns the value of the Bessel function of the first kind of order 1.
‣ __nv_j1(   ) returns  .
‣ __nv_j1(   ) returns +0.
‣ __nv_j1(NaN) returns NaN.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.196. __nv_j1f
Prototype:
float @__nv_j1f(float %x) 
Description:
Calculate the value of the Bessel function of the first kind of order 1 for the input
argument  ,  .
x
Returns:
Returns the value of the Bessel function of the first kind of order 1.
‣ __nv_j1f(   ) returns  .
‣ __nv_j1f(   ) returns +0.
‣ __nv_j1f(NaN) returns NaN.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 106Function Reference
3.197. __nv_jn
Prototype:
double @__nv_jn(i32 %n, double %x) 
Description:
Calculate the value of the Bessel function of the first kind of order   for the input
n
argument  ,  .
x
Returns:
Returns the value of the Bessel function of the first kind of order  .
n
‣ __nv_jn(n, NaN) returns NaN.
‣ __nv_jn(n, x) returns NaN for n < 0.
‣ __nv_jn(n,   ) returns +0.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.198. __nv_jnf
Prototype:
float @__nv_jnf(i32 %n, float %x) 
Description:
Calculate the value of the Bessel function of the first kind of order   for the input
n
argument  ,  .
x
Returns:
Returns the value of the Bessel function of the first kind of order  .
n
‣ __nv_jnf(n, NaN) returns NaN.
‣ __nv_jnf(n, x) returns NaN for n < 0.
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 107Function Reference
‣ __nv_jnf(n,   ) returns +0.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.199. __nv_ldexp
Prototype:
double @__nv_ldexp(double %x, i32 %y) 
Description:
Calculate the value of   of the input arguments   and  .
x exp
Returns:
‣ __nv_ldexp(x) returns   if the correctly calculated value is outside the double
floating point range.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.200. __nv_ldexpf
Prototype:
float @__nv_ldexpf(float %x, i32 %y) 
Description:
Calculate the value of   of the input arguments   and  .
x exp
Returns:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 108Function Reference
‣ __nv_ldexpf(x) returns   if the correctly calculated value is outside the double
floating point range.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.201. __nv_lgamma
Prototype:
double @__nv_lgamma(double %x) 
Description:
Calculate the natural logarithm of the absolute value of the gamma function of the input
argument  , namely the value of 
x
Returns:
‣ __nv_lgamma(1) returns +0.
‣ __nv_lgamma(2) returns +0.
‣ __nv_lgamma(x) returns   if the correctly calculated value is outside the double
floating point range.
‣ __nv_lgamma(x) returns   if x   0.
‣ __nv_lgamma(   ) returns  .
‣ __nv_lgamma(   ) returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 109Function Reference
3.202. __nv_lgammaf
Prototype:
float @__nv_lgammaf(float %x) 
Description:
Calculate the natural logarithm of the absolute value of the gamma function of the input
argument  , namely the value of 
x
Returns:
‣ __nv_lgammaf(1) returns +0.
‣ __nv_lgammaf(2) returns +0.
‣ __nv_lgammaf(x) returns   if the correctly calculated value is outside the double
floating point range.
‣ __nv_lgammaf(x) returns   if x   0.
‣ __nv_lgammaf(   ) returns  .
‣ __nv_lgammaf(   ) returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.203. __nv_ll2double_rd
Prototype:
double @__nv_ll2double_rd(i64 %l) 
Description:
Convert the signed 64-bit integer value   to a double-precision floating point value in
x
round-down (to negative infinity) mode.
Returns:
Returns converted value.
Library Availability:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 110Function Reference
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.204. __nv_ll2double_rn
Prototype:
double @__nv_ll2double_rn(i64 %l) 
Description:
Convert the signed 64-bit integer value   to a double-precision floating point value in
x
round-to-nearest-even mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.205. __nv_ll2double_ru
Prototype:
double @__nv_ll2double_ru(i64 %l) 
Description:
Convert the signed 64-bit integer value   to a double-precision floating point value in
x
round-up (to positive infinity) mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 111Function Reference
3.206. __nv_ll2double_rz
Prototype:
double @__nv_ll2double_rz(i64 %l) 
Description:
Convert the signed 64-bit integer value   to a double-precision floating point value in
x
round-towards-zero mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.207. __nv_ll2float_rd
Prototype:
float @__nv_ll2float_rd(i64 %l) 
Description:
Convert the signed integer value   to a single-precision floating point value in round-
x
down (to negative infinity) mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 112Function Reference
3.208. __nv_ll2float_rn
Prototype:
float @__nv_ll2float_rn(i64 %l) 
Description:
Convert the signed 64-bit integer value   to a single-precision floating point value in
x
round-to-nearest-even mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.209. __nv_ll2float_ru
Prototype:
float @__nv_ll2float_ru(i64 %l) 
Description:
Convert the signed integer value   to a single-precision floating point value in round-up
x
(to positive infinity) mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 113Function Reference
3.210. __nv_ll2float_rz
Prototype:
float @__nv_ll2float_rz(i64 %l) 
Description:
Convert the signed integer value   to a single-precision floating point value in round-
x
towards-zero mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.211. __nv_llabs
Prototype:
i64 @__nv_llabs(i64 %x) 
Description:
Determine the absolute value of the 64-bit signed integer  .
x
Returns:
Returns the absolute value of the 64-bit signed integer  .
x
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.212. __nv_llmax
Prototype:
i64 @__nv_llmax(i64 %x, i64 %y) 
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 114Function Reference
Description:
Determine the maximum value of the two 64-bit signed integers   and  .
x y
Returns:
Returns the maximum value of the two 64-bit signed integers   and  .
x y
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.213. __nv_llmin
Prototype:
i64 @__nv_llmin(i64 %x, i64 %y) 
Description:
Determine the minimum value of the two 64-bit signed integers   and  .
x y
Returns:
Returns the minimum value of the two 64-bit signed integers   and  .
x y
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.214. __nv_llrint
Prototype:
i64 @__nv_llrint(double %x) 
Description:
Round   to the nearest integer value, with halfway cases rounded towards zero. If the
x
result is outside the range of the return type, the result is undefined.
Returns:
Returns rounded integer value.
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 115Function Reference
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.215. __nv_llrintf
Prototype:
i64 @__nv_llrintf(float %x) 
Description:
Round   to the nearest integer value, with halfway cases rounded towards zero. If the
x
result is outside the range of the return type, the result is undefined.
Returns:
Returns rounded integer value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.216. __nv_llround
Prototype:
i64 @__nv_llround(double %x) 
Description:
Round   to the nearest integer value, with halfway cases rounded away from zero. If the
x
result is outside the range of the return type, the result is undefined.
Returns:
Returns rounded integer value.
This function may be slower than alternate rounding methods. See llrint().
Library Availability:
Compute 2.0: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 116Function Reference
Compute 3.0: Yes
Compute 3.5: Yes
3.217. __nv_llroundf
Prototype:
i64 @__nv_llroundf(float %x) 
Description:
Round   to the nearest integer value, with halfway cases rounded away from zero. If the
x
result is outside the range of the return type, the result is undefined.
Returns:
Returns rounded integer value.
This function may be slower than alternate rounding methods. See llrint().
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.218. __nv_log
Prototype:
double @__nv_log(double %x) 
Description:
Calculate the base   logarithm of the input argument  .
x
Returns:
‣ __nv_log(   ) returns  .
‣ __nv_log(1) returns +0.
‣ __nv_log(x) returns NaN for x < 0.
‣ __nv_log(   ) returns 
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 117Function Reference
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.219. __nv_log10
Prototype:
double @__nv_log10(double %x) 
Description:
Calculate the base 10 logarithm of the input argument  .
x
Returns:
‣ __nv_log10(   ) returns  .
‣ __nv_log10(1) returns +0.
‣ __nv_log10(x) returns NaN for x < 0.
‣ __nv_log10(   ) returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.220. __nv_log10f
Prototype:
float @__nv_log10f(float %x) 
Description:
Calculate the base 10 logarithm of the input argument  .
x
Returns:
‣ __nv_log10f(   ) returns  .
‣ __nv_log10f(1) returns +0.
‣ __nv_log10f(x) returns NaN for x < 0.
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 118Function Reference
‣ __nv_log10f(   ) returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.221. __nv_log1p
Prototype:
double @__nv_log1p(double %x) 
Description:
Calculate the value of   of the input argument  .
x
Returns:
‣ __nv_log1p(   ) returns  .
‣ __nv_log1p(-1) returns +0.
‣ __nv_log1p(x) returns NaN for x < -1.
‣ __nv_log1p(   ) returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.222. __nv_log1pf
Prototype:
float @__nv_log1pf(float %x) 
Description:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 119Function Reference
Calculate the value of   of the input argument  .
x
Returns:
‣ __nv_log1pf(   ) returns  .
‣ __nv_log1pf(-1) returns +0.
‣ __nv_log1pf(x) returns NaN for x < -1.
‣ __nv_log1pf(   ) returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.223. __nv_log2
Prototype:
double @__nv_log2(double %x) 
Description:
Calculate the base 2 logarithm of the input argument  .
x
Returns:
‣ __nv_log2(   ) returns  .
‣ __nv_log2(1) returns +0.
‣ __nv_log2(x) returns NaN for x < 0.
‣ __nv_log2(   ) returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 120Function Reference
3.224. __nv_log2f
Prototype:
float @__nv_log2f(float %x) 
Description:
Calculate the base 2 logarithm of the input argument  .
x
Returns:
‣ __nv_log2f(   ) returns  .
‣ __nv_log2f(1) returns +0.
‣ __nv_log2f(x) returns NaN for x < 0.
‣ __nv_log2f(   ) returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.225. __nv_logb
Prototype:
double @__nv_logb(double %x) 
Description:
Calculate the floating point representation of the exponent of the input argument  .
x
Returns:
‣ __nv_logb   returns 
‣ __nv_logb   returns 
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 121Function Reference
Compute 3.0: Yes
Compute 3.5: Yes
3.226. __nv_logbf
Prototype:
float @__nv_logbf(float %x) 
Description:
Calculate the floating point representation of the exponent of the input argument  .
x
Returns:
‣ __nv_logbf   returns 
‣ __nv_logbf   returns 
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.227. __nv_logf
Prototype:
float @__nv_logf(float %x) 
Description:
Calculate the base   logarithm of the input argument  .
x
Returns:
‣ __nv_logf(   ) returns  .
‣ __nv_logf(1) returns +0.
‣ __nv_logf(x) returns NaN for x < 0.
‣ __nv_logf(   ) returns 
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 122Function Reference
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.228. __nv_longlong_as_double
Prototype:
double @__nv_longlong_as_double(i64 %x) 
Description:
Reinterpret the bits in the 64-bit signed integer value   as a double-precision floating
x
point value.
Returns:
Returns reinterpreted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.229. __nv_max
Prototype:
i32 @__nv_max(i32 %x, i32 %y) 
Description:
Determine the maximum value of the two 32-bit signed integers   and  .
x y
Returns:
Returns the maximum value of the two 32-bit signed integers   and  .
x y
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 123Function Reference
3.230. __nv_min
Prototype:
i32 @__nv_min(i32 %x, i32 %y) 
Description:
Determine the minimum value of the two 32-bit signed integers   and  .
x y
Returns:
Returns the minimum value of the two 32-bit signed integers   and  .
x y
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.231. __nv_modf
Prototype:
double @__nv_modf(double %x, double* %b) 
Description:
Break down the argument   into fractional and integral parts. The integral part is stored
x
in the argument  . Fractional and integral parts are given the same sign as the
iptr
argument  .
x
Returns:
‣ __nv_modf(   , iptr) returns a result with the same sign as x.
‣ __nv_modf(   , iptr) returns   and stores   in the object pointed to by
.
iptr
‣ __nv_modf(NaN, iptr) stores a NaN in the object pointed to by iptr and returns a
NaN.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 124Function Reference
Compute 3.5: Yes
3.232. __nv_modff
Prototype:
float @__nv_modff(float %x, float* %b) 
Description:
Break down the argument   into fractional and integral parts. The integral part is stored
x
in the argument  . Fractional and integral parts are given the same sign as the
iptr
argument  .
x
Returns:
‣ __nv_modff(   , iptr) returns a result with the same sign as x.
‣ __nv_modff(   , iptr) returns   and stores   in the object pointed to by
.
iptr
‣ __nv_modff(NaN, iptr) stores a NaN in the object pointed to by iptr and returns
a NaN.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.233. __nv_mul24
Prototype:
i32 @__nv_mul24(i32 %x, i32 %y) 
Description:
Calculate the least significant 32 bits of the product of the least significant 24 bits of 
x
and  . The high order 8 bits of   and   are ignored.
y x y
Returns:
Returns the least significant 32 bits of the product   *  .
x y
Library Availability:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 125Function Reference
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.234. __nv_mul64hi
Prototype:
i64 @__nv_mul64hi(i64 %x, i64 %y) 
Description:
Calculate the most significant 64 bits of the 128-bit product   *  , where   and   are 64-
x y x y
bit integers.
Returns:
Returns the most significant 64 bits of the product   *  .
x y
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.235. __nv_mulhi
Prototype:
i32 @__nv_mulhi(i32 %x, i32 %y) 
Description:
Calculate the most significant 32 bits of the 64-bit product   *  , where   and   are 32-bit
x y x y
integers.
Returns:
Returns the most significant 32 bits of the product   *  .
x y
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 126Function Reference
3.236. __nv_nan
Prototype:
double @__nv_nan(i8* %tagp) 
Description:
Return a representation of a quiet NaN. Argument   selects one of the possible
tagp
representations.
Returns:
‣ __nv_nan(tagp) returns NaN.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.237. __nv_nanf
Prototype:
float @__nv_nanf(i8* %tagp) 
Description:
Return a representation of a quiet NaN. Argument   selects one of the possible
tagp
representations.
Returns:
‣ __nv_nanf(tagp) returns NaN.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 127Function Reference
Compute 3.5: Yes
3.238. __nv_nearbyint
Prototype:
double @__nv_nearbyint(double %x) 
Description:
Round argument   to an integer value in double precision floating-point format.
x
Returns:
‣ __nv_nearbyint(   ) returns  .
‣ __nv_nearbyint(   ) returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.239. __nv_nearbyintf
Prototype:
float @__nv_nearbyintf(float %x) 
Description:
Round argument   to an integer value in double precision floating-point format.
x
Returns:
‣ __nv_nearbyintf(   ) returns  .
‣ __nv_nearbyintf(   ) returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 128Function Reference
Compute 3.0: Yes
Compute 3.5: Yes
3.240. __nv_nextafter
Prototype:
double @__nv_nextafter(double %x, double %y) 
Description:
Calculate the next representable double-precision floating-point value following   in
x
the direction of  . For example, if   is greater than  , nextafter() returns the smallest
y y x
representable number greater than 
x
Returns:
‣ __nv_nextafter(   , y) returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.241. __nv_nextafterf
Prototype:
float @__nv_nextafterf(float %x, float %y) 
Description:
Calculate the next representable double-precision floating-point value following   in
x
the direction of  . For example, if   is greater than  , nextafter() returns the smallest
y y x
representable number greater than 
x
Returns:
‣ __nv_nextafterf(   , y) returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 129Function Reference
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.242. __nv_normcdf
Prototype:
double @__nv_normcdf(double %x) 
Description:
Calculate the cumulative distribution function of the standard normal distribution for
input argument  ,  .
y
Returns:
‣ __nv_normcdf(   ) returns 1
‣ __nv_normcdf(   ) returns +0
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.243. __nv_normcdff
Prototype:
float @__nv_normcdff(float %x) 
Description:
Calculate the cumulative distribution function of the standard normal distribution for
input argument  ,  .
y
Returns:
‣ __nv_normcdff(   ) returns 1
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 130Function Reference
‣ __nv_normcdff(   ) returns +0
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.244. __nv_normcdfinv
Prototype:
double @__nv_normcdfinv(double %x) 
Description:
Calculate the inverse of the standard normal cumulative distribution function for input
argument  ,  . The function is defined for input values in the interval  .
y
Returns:
‣ __nv_normcdfinv(0) returns  .
‣ __nv_normcdfinv(1) returns  .
‣ __nv_normcdfinv(x) returns NaN if x is not in the interval [0,1].
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.245. __nv_normcdfinvf
Prototype:
float @__nv_normcdfinvf(float %x) 
Description:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 131Function Reference
Calculate the inverse of the standard normal cumulative distribution function for input
argument  ,  . The function is defined for input values in the interval  .
y
Returns:
‣ __nv_normcdfinvf(0) returns  .
‣ __nv_normcdfinvf(1) returns  .
‣ __nv_normcdfinvf(x) returns NaN if x is not in the interval [0,1].
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.246. __nv_popc
Prototype:
i32 @__nv_popc(i32 %x) 
Description:
Count the number of bits that are set to 1 in  .
x
Returns:
Returns a value between 0 and 32 inclusive representing the number of set bits.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.247. __nv_popcll
Prototype:
i32 @__nv_popcll(i64 %x) 
Description:
Count the number of bits that are set to 1 in  .
x
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 132Function Reference
Returns:
Returns a value between 0 and 64 inclusive representing the number of set bits.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.248. __nv_pow
Prototype:
double @__nv_pow(double %x, double %y) 
Description:
Calculate the value of   to the power of 
x y
Returns:
‣ __nv_pow(   , y) returns   for y an integer less than 0.
‣ __nv_pow(   , y) returns   for y an odd integer greater than 0.
‣ __nv_pow(   , y) returns +0 for y > 0 and not and odd integer.
‣ __nv_pow(-1,   ) returns 1.
‣ __nv_pow(+1, y) returns 1 for any y, even a NaN.
‣ __nv_pow(x,   ) returns 1 for any x, even a NaN.
‣ __nv_pow(x, y) returns a NaN for finite x < 0 and finite non-integer y.
‣ __nv_pow( ,   ) returns   for  .
x
‣ __nv_pow( ,   ) returns +0 for  .
x
‣ __nv_pow( ,   ) returns +0 for  .
x
‣ __nv_pow( ,   ) returns   for  .
x
‣ __nv_pow(   , y) returns -0 for y an odd integer less than 0.
‣ __nv_pow(   , y) returns +0 for y < 0 and not an odd integer.
‣ __nv_pow(   , y) returns   for y an odd integer greater than 0.
‣ __nv_pow(   , y) returns   for y > 0 and not an odd integer.
‣ __nv_pow(   , y) returns +0 for y < 0.
‣ __nv_pow(   , y) returns   for y > 0.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 133Function Reference
Compute 3.0: Yes
Compute 3.5: Yes
3.249. __nv_powf
Prototype:
float @__nv_powf(float %x, float %y) 
Description:
Calculate the value of   to the power of 
x y
Returns:
‣ __nv_powf(   , y) returns   for y an integer less than 0.
‣ __nv_powf(   , y) returns   for y an odd integer greater than 0.
‣ __nv_powf(   , y) returns +0 for y > 0 and not and odd integer.
‣ __nv_powf(-1,   ) returns 1.
‣ __nv_powf(+1, y) returns 1 for any y, even a NaN.
‣ __nv_powf(x,   ) returns 1 for any x, even a NaN.
‣ __nv_powf(x, y) returns a NaN for finite x < 0 and finite non-integer y.
‣ __nv_powf( ,   ) returns   for  .
x
‣ __nv_powf( ,   ) returns +0 for  .
x
‣ __nv_powf( ,   ) returns +0 for  .
x
‣ __nv_powf( ,   ) returns   for  .
x
‣ __nv_powf(   , y) returns -0 for y an odd integer less than 0.
‣ __nv_powf(   , y) returns +0 for y < 0 and not an odd integer.
‣ __nv_powf(   , y) returns   for y an odd integer greater than 0.
‣ __nv_powf(   , y) returns   for y > 0 and not an odd integer.
‣ __nv_powf(   , y) returns +0 for y < 0.
‣ __nv_powf(   , y) returns   for y > 0.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 134Function Reference
3.250. __nv_powi
Prototype:
double @__nv_powi(double %x, i32 %y) 
Description:
Calculate the value of   to the power of 
x y
Returns:
‣ __nv_powi(   , y) returns   for y an integer less than 0.
‣ __nv_powi(   , y) returns   for y an odd integer greater than 0.
‣ __nv_powi(   , y) returns +0 for y > 0 and not and odd integer.
‣ __nv_powi(-1,   ) returns 1.
‣ __nv_powi(+1, y) returns 1 for any y, even a NaN.
‣ __nv_powi(x,   ) returns 1 for any x, even a NaN.
‣ __nv_powi(x, y) returns a NaN for finite x < 0 and finite non-integer y.
‣ __nv_powi( ,   ) returns   for  .
x
‣ __nv_powi( ,   ) returns +0 for  .
x
‣ __nv_powi( ,   ) returns +0 for  .
x
‣ __nv_powi( ,   ) returns   for  .
x
‣ __nv_powi(   , y) returns -0 for y an odd integer less than 0.
‣ __nv_powi(   , y) returns +0 for y < 0 and not an odd integer.
‣ __nv_powi(   , y) returns   for y an odd integer greater than 0.
‣ __nv_powi(   , y) returns   for y > 0 and not an odd integer.
‣ __nv_powi(   , y) returns +0 for y < 0.
‣ __nv_powi(   , y) returns   for y > 0.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 135Function Reference
3.251. __nv_powif
Prototype:
float @__nv_powif(float %x, i32 %y) 
Description:
Calculate the value of   to the power of  .
x y
Returns:
‣ __nv_powif(   , y) returns   for y an integer less than 0.
‣ __nv_powif(   , y) returns   for y an odd integer greater than 0.
‣ __nv_powif(   , y) returns +0 for y > 0 and not and odd integer.
‣ __nv_powif(-1,   ) returns 1.
‣ __nv_powif(+1, y) returns 1 for any y, even a NaN.
‣ __nv_powif(x,   ) returns 1 for any x, even a NaN.
‣ __nv_powif(x, y) returns a NaN for finite x < 0 and finite non-integer y.
‣ __nv_powif( ,   ) returns   for  .
x
‣ __nv_powif( ,   ) returns +0 for  .
x
‣ __nv_powif( ,   ) returns +0 for  .
x
‣ __nv_powif( ,   ) returns   for  .
x
‣ __nv_powif(   , y) returns -0 for y an odd integer less than 0.
‣ __nv_powif(   , y) returns +0 for y < 0 and not an odd integer.
‣ __nv_powif(   , y) returns   for y an odd integer greater than 0.
‣ __nv_powif(   , y) returns   for y > 0 and not an odd integer.
‣ __nv_powif(   , y) returns +0 for y < 0.
‣ __nv_powif(   , y) returns   for y > 0.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 136Function Reference
3.252. __nv_rcbrt
Prototype:
double @__nv_rcbrt(double %x) 
Description:
Calculate reciprocal cube root function of 
x
Returns:
‣ __nv_rcbrt(   ) returns  .
‣ __nv_rcbrt(   ) returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.253. __nv_rcbrtf
Prototype:
float @__nv_rcbrtf(float %x) 
Description:
Calculate reciprocal cube root function of 
x
Returns:
‣ __nv_rcbrtf(   ) returns  .
‣ __nv_rcbrtf(   ) returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 137Function Reference
Compute 3.5: Yes
3.254. __nv_remainder
Prototype:
double @__nv_remainder(double %x, double %y) 
Description:
Compute double-precision floating-point remainder   of dividing   by   for nonzero  .
r x y y
Thus  . The value   is the integer value nearest  . In the case when   ,
n
the even   value is chosen.
n
Returns:
‣ __nv_remainder(x, 0) returns NaN.
‣ __nv_remainder(   , y) returns NaN.
‣ __nv_remainder(x,   ) returns x for finite x.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.255. __nv_remainderf
Prototype:
float @__nv_remainderf(float %x, float %y) 
Description:
Compute double-precision floating-point remainder   of dividing   by   for nonzero  .
r x y y
Thus  . The value   is the integer value nearest  . In the case when   ,
n
the even   value is chosen.
n
Returns:
‣ __nv_remainderf(x, 0) returns NaN.
‣ __nv_remainderf(   , y) returns NaN.
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 138Function Reference
‣ __nv_remainderf(x,   ) returns x for finite x.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.256. __nv_remquo
Prototype:
double @__nv_remquo(double %x, double %y, i32* %c) 
Description:
Compute a double-precision floating-point remainder in the same way as the
remainder() function. Argument   returns part of quotient upon division of   by  .
quo x y
Value   has the same sign as   and may not be the exact quotient but agrees with the
quo
exact quotient in the low order 3 bits.
Returns:
Returns the remainder.
‣ __nv_remquo(x, 0, quo) returns NaN.
‣ __nv_remquo(   , y, quo) returns NaN.
‣ __nv_remquo(x,   , quo) returns x.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 139Function Reference
3.257. __nv_remquof
Prototype:
float @__nv_remquof(float %x, float %y, i32* %quo) 
Description:
Compute a double-precision floating-point remainder in the same way as the
remainder() function. Argument   returns part of quotient upon division of   by  .
quo x y
Value   has the same sign as   and may not be the exact quotient but agrees with the
quo
exact quotient in the low order 3 bits.
Returns:
Returns the remainder.
‣ __nv_remquof(x, 0, quo) returns NaN.
‣ __nv_remquof(   , y, quo) returns NaN.
‣ __nv_remquof(x,   , quo) returns x.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.258. __nv_rhadd
Prototype:
i32 @__nv_rhadd(i32 %x, i32 %y) 
Description:
Compute average of signed input arguments   and   as (   +   + 1 ) >> 1, avoiding
x y x y
overflow in the intermediate sum.
Returns:
Returns a signed integer value representing the signed rounded average value of the two
inputs.
Library Availability:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 140Function Reference
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.259. __nv_rint
Prototype:
double @__nv_rint(double %x) 
Description:
Round   to the nearest integer value in floating-point format, with halfway cases
x
rounded to the nearest even integer value.
Returns:
Returns rounded integer value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.260. __nv_rintf
Prototype:
float @__nv_rintf(float %x) 
Description:
Round   to the nearest integer value in floating-point format, with halfway cases
x
rounded to the nearest even integer value.
Returns:
Returns rounded integer value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 141Function Reference
3.261. __nv_round
Prototype:
double @__nv_round(double %x) 
Description:
Round   to the nearest integer value in floating-point format, with halfway cases
x
rounded away from zero.
Returns:
Returns rounded integer value.
This function may be slower than alternate rounding methods. See rint().
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.262. __nv_roundf
Prototype:
float @__nv_roundf(float %x) 
Description:
Round   to the nearest integer value in floating-point format, with halfway cases
x
rounded away from zero.
Returns:
Returns rounded integer value.
This function may be slower than alternate rounding methods. See rint().
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 142Function Reference
3.263. __nv_rsqrt
Prototype:
double @__nv_rsqrt(double %x) 
Description:
Calculate the reciprocal of the nonnegative square root of  ,  .
x
Returns:
Returns  .
‣ __nv_rsqrt(   ) returns +0.
‣ __nv_rsqrt(   ) returns  .
‣ __nv_rsqrt(x) returns NaN if x is less than 0.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.264. __nv_rsqrtf
Prototype:
float @__nv_rsqrtf(float %x) 
Description:
Calculate the reciprocal of the nonnegative square root of  ,  .
x
Returns:
Returns  .
‣ __nv_rsqrtf(   ) returns +0.
‣ __nv_rsqrtf(   ) returns  .
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 143Function Reference
‣ __nv_rsqrtf(x) returns NaN if x is less than 0.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.265. __nv_sad
Prototype:
i32 @__nv_sad(i32 %x, i32 %y, i32 %z) 
Description:
Calculate   , the 32-bit sum of the third argument   plus and the absolute value
z
of the difference between the first argument,  , and second argument,  .
x y
Inputs   and   are signed 32-bit integers, input   is a 32-bit unsigned integer.
x y z
Returns:
Returns  .
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.266. __nv_saturatef
Prototype:
float @__nv_saturatef(float %x) 
Description:
Clamp the input argument   to be within the interval [+0.0, 1.0].
x
Returns:
‣ __nv_saturatef(x) returns 0 if x < 0.
‣ __nv_saturatef(x) returns 1 if x > 1.
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 144Function Reference
‣ __nv_saturatef(x) returns x if  .
‣ __nv_saturatef(NaN) returns 0.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.267. __nv_scalbn
Prototype:
double @__nv_scalbn(double %x, i32 %y) 
Description:
Scale   by   by efficient manipulation of the floating-point exponent.
x
Returns:
Returns   *  .
x
‣ __nv_scalbn(   , n) returns  .
‣ __nv_scalbn(x, 0) returns x.
‣ __nv_scalbn(   , n) returns  .
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.268. __nv_scalbnf
Prototype:
float @__nv_scalbnf(float %x, i32 %y) 
Description:
Scale   by   by efficient manipulation of the floating-point exponent.
x
Returns:
Returns   *  .
x
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 145Function Reference
‣ __nv_scalbnf(   , n) returns  .
‣ __nv_scalbnf(x, 0) returns x.
‣ __nv_scalbnf(   , n) returns  .
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.269. __nv_signbitd
Prototype:
i32 @__nv_signbitd(double %x) 
Description:
Determine whether the floating-point value   is negative.
x
Returns:
Returns a nonzero value if and only if   is negative. Reports the sign bit of all values
x
including infinities, zeros, and NaNs.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.270. __nv_signbitf
Prototype:
i32 @__nv_signbitf(float %x) 
Description:
Determine whether the floating-point value   is negative.
x
Returns:
Returns a nonzero value if and only if   is negative. Reports the sign bit of all values
x
including infinities, zeros, and NaNs.
Library Availability:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 146Function Reference
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.271. __nv_sin
Prototype:
double @__nv_sin(double %x) 
Description:
Calculate the sine of the input argument   (measured in radians).
x
Returns:
‣ __nv_sin(   ) returns  .
‣ __nv_sin(   ) returns NaN.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.272. __nv_sincos
Prototype:
void @__nv_sincos(double %x, double* %sptr, double* %cptr) 
Description:
Calculate the sine and cosine of the first input argument   (measured in radians). The
x
results for sine and cosine are written into the second argument,  , and, respectively,
sptr
third argument,  .
zptr
Returns:
‣ none
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 147Function Reference
See __nv_sin() and __nv_cos().
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.273. __nv_sincosf
Prototype:
void @__nv_sincosf(float %x, float* %sptr, float* %cptr) 
Description:
Calculate the sine and cosine of the first input argument   (measured in radians). The
x
results for sine and cosine are written into the second argument,  , and, respectively,
sptr
third argument,  .
zptr
Returns:
‣ none
See __nv_sinf() and __nv_cosf().
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.274. __nv_sincospi
Prototype:
void @__nv_sincospi(double %x, double* %sptr, double* %cptr) 
Description:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 148Function Reference
Calculate the sine and cosine of the first input argument,   (measured in radians),
x
. The results for sine and cosine are written into the second argument,  , and,
sptr
respectively, third argument,  .
zptr
Returns:
‣ none
See __nv_sinpi() and __nv_cospi().
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.275. __nv_sincospif
Prototype:
void @__nv_sincospif(float %x, float* %sptr, float* %cptr) 
Description:
Calculate the sine and cosine of the first input argument,   (measured in radians),
x
. The results for sine and cosine are written into the second argument,  , and,
sptr
respectively, third argument,  .
zptr
Returns:
‣ none
See __nv_sinpif() and __nv_cospif().
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 149Function Reference
3.276. __nv_sinf
Prototype:
float @__nv_sinf(float %x) 
Description:
Calculate the sine of the input argument   (measured in radians).
x
Returns:
‣ __nv_sinf(   ) returns  .
‣ __nv_sinf(   ) returns NaN.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.277. __nv_sinh
Prototype:
double @__nv_sinh(double %x) 
Description:
Calculate the hyperbolic sine of the input argument  .
x
Returns:
‣ __nv_sinh(   ) returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 150Function Reference
3.278. __nv_sinhf
Prototype:
float @__nv_sinhf(float %x) 
Description:
Calculate the hyperbolic sine of the input argument  .
x
Returns:
‣ __nv_sinhf(   ) returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.279. __nv_sinpi
Prototype:
double @__nv_sinpi(double %x) 
Description:
Calculate the sine of     (measured in radians), where   is the input argument.
x x
Returns:
‣ __nv_sinpi(   ) returns  .
‣ __nv_sinpi(   ) returns NaN.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 151Function Reference
3.280. __nv_sinpif
Prototype:
float @__nv_sinpif(float %x) 
Description:
Calculate the sine of     (measured in radians), where   is the input argument.
x x
Returns:
‣ __nv_sinpif(   ) returns  .
‣ __nv_sinpif(   ) returns NaN.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.281. __nv_sqrt
Prototype:
double @__nv_sqrt(double %x) 
Description:
Calculate the nonnegative square root of  ,  .
x
Returns:
Returns  .
‣ __nv_sqrt(   ) returns  .
‣ __nv_sqrt(   ) returns  .
‣ __nv_sqrt(x) returns NaN if x is less than 0.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 152Function Reference
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.282. __nv_sqrtf
Prototype:
float @__nv_sqrtf(float %x) 
Description:
Calculate the nonnegative square root of  ,  .
x
Returns:
Returns  .
‣ __nv_sqrtf(   ) returns  .
‣ __nv_sqrtf(   ) returns  .
‣ __nv_sqrtf(x) returns NaN if x is less than 0.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.283. __nv_tan
Prototype:
double @__nv_tan(double %x) 
Description:
Calculate the tangent of the input argument   (measured in radians).
x
Returns:
‣ __nv_tan(   ) returns  .
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 153Function Reference
‣ __nv_tan(   ) returns NaN.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.284. __nv_tanf
Prototype:
float @__nv_tanf(float %x) 
Description:
Calculate the tangent of the input argument   (measured in radians).
x
Returns:
‣ __nv_tanf(   ) returns  .
‣ __nv_tanf(   ) returns NaN.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.285. __nv_tanh
Prototype:
double @__nv_tanh(double %x) 
Description:
Calculate the hyperbolic tangent of the input argument  .
x
Returns:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 154Function Reference
‣ __nv_tanh(   ) returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.286. __nv_tanhf
Prototype:
float @__nv_tanhf(float %x) 
Description:
Calculate the hyperbolic tangent of the input argument  .
x
Returns:
‣ __nv_tanhf(   ) returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.287. __nv_tgamma
Prototype:
double @__nv_tgamma(double %x) 
Description:
Calculate the gamma function of the input argument  , namely the value of  .
x
Returns:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 155Function Reference
‣ __nv_tgamma(   ) returns  .
‣ __nv_tgamma(2) returns +0.
‣ __nv_tgamma(x) returns   if the correctly calculated value is outside the double
floating point range.
‣ __nv_tgamma(x) returns NaN if x < 0.
‣ __nv_tgamma(   ) returns NaN.
‣ __nv_tgamma(   ) returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.288. __nv_tgammaf
Prototype:
float @__nv_tgammaf(float %x) 
Description:
Calculate the gamma function of the input argument  , namely the value of  .
x
Returns:
‣ __nv_tgammaf(   ) returns  .
‣ __nv_tgammaf(2) returns +0.
‣ __nv_tgammaf(x) returns   if the correctly calculated value is outside the double
floating point range.
‣ __nv_tgammaf(x) returns NaN if x < 0.
‣ __nv_tgammaf(   ) returns NaN.
‣ __nv_tgammaf(   ) returns  .
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 156Function Reference
3.289. __nv_trunc
Prototype:
double @__nv_trunc(double %x) 
Description:
Round   to the nearest integer value that does not exceed   in magnitude.
x x
Returns:
Returns truncated integer value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.290. __nv_truncf
Prototype:
float @__nv_truncf(float %x) 
Description:
Round   to the nearest integer value that does not exceed   in magnitude.
x x
Returns:
Returns truncated integer value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.291. __nv_uhadd
Prototype:
i32 @__nv_uhadd(i32 %x, i32 %y) 
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 157Function Reference
Description:
Compute average of unsigned input arguments   and   as (   +   ) >> 1, avoiding
x y x y
overflow in the intermediate sum.
Returns:
Returns an unsigned integer value representing the unsigned average value of the two
inputs.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.292. __nv_uint2double_rn
Prototype:
double @__nv_uint2double_rn(i32 %i) 
Description:
Convert the unsigned integer value   to a double-precision floating point value.
x
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.293. __nv_uint2float_rd
Prototype:
float @__nv_uint2float_rd(i32 %in) 
Description:
Convert the unsigned integer value   to a single-precision floating point value in round-
x
down (to negative infinity) mode.
Returns:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 158Function Reference
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.294. __nv_uint2float_rn
Prototype:
float @__nv_uint2float_rn(i32 %in) 
Description:
Convert the unsigned integer value   to a single-precision floating point value in round-
x
to-nearest-even mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.295. __nv_uint2float_ru
Prototype:
float @__nv_uint2float_ru(i32 %in) 
Description:
Convert the unsigned integer value   to a single-precision floating point value in round-
x
up (to positive infinity) mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 159Function Reference
Compute 3.5: Yes
3.296. __nv_uint2float_rz
Prototype:
float @__nv_uint2float_rz(i32 %in) 
Description:
Convert the unsigned integer value   to a single-precision floating point value in round-
x
towards-zero mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.297. __nv_ull2double_rd
Prototype:
double @__nv_ull2double_rd(i64 %l) 
Description:
Convert the unsigned 64-bit integer value   to a double-precision floating point value in
x
round-down (to negative infinity) mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 160Function Reference
3.298. __nv_ull2double_rn
Prototype:
double @__nv_ull2double_rn(i64 %l) 
Description:
Convert the unsigned 64-bit integer value   to a double-precision floating point value in
x
round-to-nearest-even mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.299. __nv_ull2double_ru
Prototype:
double @__nv_ull2double_ru(i64 %l) 
Description:
Convert the unsigned 64-bit integer value   to a double-precision floating point value in
x
round-up (to positive infinity) mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 161Function Reference
3.300. __nv_ull2double_rz
Prototype:
double @__nv_ull2double_rz(i64 %l) 
Description:
Convert the unsigned 64-bit integer value   to a double-precision floating point value in
x
round-towards-zero mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.301. __nv_ull2float_rd
Prototype:
float @__nv_ull2float_rd(i64 %l) 
Description:
Convert the unsigned integer value   to a single-precision floating point value in round-
x
down (to negative infinity) mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 162Function Reference
3.302. __nv_ull2float_rn
Prototype:
float @__nv_ull2float_rn(i64 %l) 
Description:
Convert the unsigned integer value   to a single-precision floating point value in round-
x
to-nearest-even mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.303. __nv_ull2float_ru
Prototype:
float @__nv_ull2float_ru(i64 %l) 
Description:
Convert the unsigned integer value   to a single-precision floating point value in round-
x
up (to positive infinity) mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 163Function Reference
3.304. __nv_ull2float_rz
Prototype:
float @__nv_ull2float_rz(i64 %l) 
Description:
Convert the unsigned integer value   to a single-precision floating point value in round-
x
towards-zero mode.
Returns:
Returns converted value.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.305. __nv_ullmax
Prototype:
i64 @__nv_ullmax(i64 %x, i64 %y) 
Description:
Determine the maximum value of the two 64-bit unsigned integers   and  .
x y
Returns:
Returns the maximum value of the two 64-bit unsigned integers   and  .
x y
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.306. __nv_ullmin
Prototype:
i64 @__nv_ullmin(i64 %x, i64 %y) 
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 164Function Reference
Description:
Determine the minimum value of the two 64-bit unsigned integers   and  .
x y
Returns:
Returns the minimum value of the two 64-bit unsigned integers   and  .
x y
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.307. __nv_umax
Prototype:
i32 @__nv_umax(i32 %x, i32 %y) 
Description:
Determine the maximum value of the two 32-bit unsigned integers   and  .
x y
Returns:
Returns the maximum value of the two 32-bit unsigned integers   and  .
x y
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.308. __nv_umin
Prototype:
i32 @__nv_umin(i32 %x, i32 %y) 
Description:
Determine the minimum value of the two 32-bit unsigned integers   and  .
x y
Returns:
Returns the minimum value of the two 32-bit unsigned integers   and  .
x y
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 165Function Reference
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.309. __nv_umul24
Prototype:
i32 @__nv_umul24(i32 %x, i32 %y) 
Description:
Calculate the least significant 32 bits of the product of the least significant 24 bits of 
x
and  . The high order 8 bits of   and   are ignored.
y x y
Returns:
Returns the least significant 32 bits of the product   *  .
x y
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.310. __nv_umul64hi
Prototype:
i64 @__nv_umul64hi(i64 %x, i64 %y) 
Description:
Calculate the most significant 64 bits of the 128-bit product   *  , where   and   are 64-
x y x y
bit unsigned integers.
Returns:
Returns the most significant 64 bits of the product   *  .
x y
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 166Function Reference
3.311. __nv_umulhi
Prototype:
i32 @__nv_umulhi(i32 %x, i32 %y) 
Description:
Calculate the most significant 32 bits of the 64-bit product   *  , where   and   are 32-bit
x y x y
unsigned integers.
Returns:
Returns the most significant 32 bits of the product   *  .
x y
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.312. __nv_urhadd
Prototype:
i32 @__nv_urhadd(i32 %x, i32 %y) 
Description:
Compute average of unsigned input arguments   and   as (   +   + 1 ) >> 1, avoiding
x y x y
overflow in the intermediate sum.
Returns:
Returns an unsigned integer value representing the unsigned rounded average value of
the two inputs.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 167Function Reference
3.313. __nv_usad
Prototype:
i32 @__nv_usad(i32 %x, i32 %y, i32 %z) 
Description:
Calculate   , the 32-bit sum of the third argument   plus and the absolute value
z
of the difference between the first argument,  , and second argument,  .
x y
Inputs  ,  , and   are unsigned 32-bit integers.
x y z
Returns:
Returns  .
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.314. __nv_y0
Prototype:
double @__nv_y0(double %x) 
Description:
Calculate the value of the Bessel function of the second kind of order 0 for the input
argument  ,  .
x
Returns:
Returns the value of the Bessel function of the second kind of order 0.
‣ __nv_y0(0) returns  .
‣ __nv_y0(x) returns NaN for x < 0.
‣ __nv_y0(   ) returns +0.
‣ __nv_y0(NaN) returns NaN.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 168Function Reference
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.315. __nv_y0f
Prototype:
float @__nv_y0f(float %x) 
Description:
Calculate the value of the Bessel function of the second kind of order 0 for the input
argument  ,  .
x
Returns:
Returns the value of the Bessel function of the second kind of order 0.
‣ __nv_y0f(0) returns  .
‣ __nv_y0f(x) returns NaN for x < 0.
‣ __nv_y0f(   ) returns +0.
‣ __nv_y0f(NaN) returns NaN.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.316. __nv_y1
Prototype:
double @__nv_y1(double %x) 
Description:
Calculate the value of the Bessel function of the second kind of order 1 for the input
argument  ,  .
x
Returns:
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 169Function Reference
Returns the value of the Bessel function of the second kind of order 1.
‣ __nv_y1(0) returns  .
‣ __nv_y1(x) returns NaN for x < 0.
‣ __nv_y1(   ) returns +0.
‣ __nv_y1(NaN) returns NaN.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.317. __nv_y1f
Prototype:
float @__nv_y1f(float %x) 
Description:
Calculate the value of the Bessel function of the second kind of order 1 for the input
argument  ,  .
x
Returns:
Returns the value of the Bessel function of the second kind of order 1.
‣ __nv_y1f(0) returns  .
‣ __nv_y1f(x) returns NaN for x < 0.
‣ __nv_y1f(   ) returns +0.
‣ __nv_y1f(NaN) returns NaN.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 170Function Reference
3.318. __nv_yn
Prototype:
double @__nv_yn(i32 %n, double %x) 
Description:
Calculate the value of the Bessel function of the second kind of order   for the input
n
argument  ,  .
x
Returns:
Returns the value of the Bessel function of the second kind of order  .
n
‣ __nv_yn(n, x) returns NaN for n < 0.
‣ __nv_yn(n, 0) returns  .
‣ __nv_yn(n, x) returns NaN for x < 0.
‣ __nv_yn(n,   ) returns +0.
‣ __nv_yn(n, NaN) returns NaN.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 7.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
3.319. __nv_ynf
Prototype:
float @__nv_ynf(i32 %n, float %x) 
Description:
Calculate the value of the Bessel function of the second kind of order   for the input
n
argument  ,  .
x
Returns:
Returns the value of the Bessel function of the second kind of order  .
n
‣ __nv_ynf(n, x) returns NaN for n < 0.
‣ __nv_ynf(n, 0) returns  .
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 171Function Reference
‣ __nv_ynf(n, x) returns NaN for x < 0.
‣ __nv_ynf(n,   ) returns +0.
‣ __nv_ynf(n, NaN) returns NaN.
For accuracy information for this function see the CUDA C Programming Guide,
Appendix D.1, Table 6.
Library Availability:
Compute 2.0: Yes
Compute 3.0: Yes
Compute 3.5: Yes
www.nvidia.com
Libdevice User's Guide Part 000 _v8.0 | 172Notice
ALL  NVIDIA  DESIGN  SPECIFICATIONS,  REFERENCE  BOARDS,  FILES,  DRAWINGS,
DIAGNOSTICS,  LISTS,  AND  OTHER  DOCUMENTS  (TOGETHER  AND  SEPARATELY,
"MATERIALS")  ARE  BEING  PROVIDED  "AS  IS."  NVIDIA  MAKES  NO  WARRANTIES,
EXPRESSED,  IMPLIED,  STATUTORY,  OR  OTHERWISE  WITH  RESPECT  TO  THE
MATERIALS,  AND  EXPRESSLY  DISCLAIMS  ALL  IMPLIED  WARRANTIES  OF
NONINFRINGEMENT,  MERCHANTABILITY,  AND  FITNESS  FOR  A  PARTICULAR
PURPOSE.
Information furnished is believed to be accurate and reliable. However, NVIDIA
Corporation  assumes  no  responsibility  for  the  consequences  of  use  of  such
information or for any infringement of patents or other rights of third parties
that may result from its use. No license is granted by implication of otherwise
under any patent rights of NVIDIA Corporation. Specifications mentioned in this
publication are subject to change without notice. This publication supersedes and
replaces all other information previously supplied. NVIDIA Corporation products
are not authorized as critical components in life support devices or systems
without express written approval of NVIDIA Corporation.
Trademarks
NVIDIA and the NVIDIA logo are trademarks or registered trademarks of NVIDIA
Corporation in the U.S. and other countries. Other company and product names
may be trademarks of the respective companies with which they are associated.
Copyright
© 2016 NVIDIA Corporation. All rights reserved.
www.nvidia.com