u01fixedpt.h

Go to the documentation of this file.

/*
Copyright 2011, D. E. Shaw Research.
All rights reserved.

Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are
met:

* Redistributions of source code must retain the above copyright
  notice, this list of conditions, and the following disclaimer.

* Redistributions in binary form must reproduce the above copyright
  notice, this list of conditions, and the following disclaimer in the
  documentation and/or other materials provided with the distribution.

* Neither the name of D. E. Shaw Research nor the names of its
  contributors may be used to endorse or promote products derived from
  this software without specific prior written permission.

THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#ifndef _random123_ufixed01_dot_h_
#define _random123_ufixed01_dot_h_

#include "features/compilerfeatures.h"

/* Hex floats were standardized by C in 1999, but weren't standardized
   by C++ until 2011.  So, we're obliged to write out our constants in
   decimal, even though they're most naturally expressed in binary.
   We cross our fingers and hope that the compiler does the compile-time
   constant arithmetic properly.
*/
#define R123_0x1p_31f (1.f/(1024.f*1024.f*1024.f*2.f))
#define R123_0x1p_24f (128.f*R123_0x1p_31f)
#define R123_0x1p_23f (256.f*R123_0x1p_31f)
#define R123_0x1p_32  (1./(1024.*1024.*1024.*4.))
#define R123_0x1p_63 (2.*R123_0x1p_32*R123_0x1p_32)
#define R123_0x1p_53 (1024.*R123_0x1p_63)
#define R123_0x1p_52 (2048.*R123_0x1p_63)

#ifndef R123_USE_U01_DOUBLE
#define R123_USE_U01_DOUBLE
#endif

#ifdef __cplusplus
extern "C"{
#endif

/* narrowing conversions:  uint32_t to float */
R123_CUDA_DEVICE R123_STATIC_INLINE float u01fixedpt_closed_closed_32_float(uint32_t i){
    /* N.B.  we ignore the high bit, so output is not monotonic */
    return ((i&0x7fffffc0) + (i&0x40))*R123_0x1p_31f; /* 0x1.p-31f */
}

R123_CUDA_DEVICE R123_STATIC_INLINE float u01fixedpt_closed_open_32_float(uint32_t i){
    return (i>>8)*R123_0x1p_24f; /* 0x1.0p-24f; */
}

R123_CUDA_DEVICE R123_STATIC_INLINE float u01fixedpt_open_closed_32_float(uint32_t i){
    return (1+(i>>8))*R123_0x1p_24f; /* *0x1.0p-24f; */
}

R123_CUDA_DEVICE R123_STATIC_INLINE float u01fixedpt_open_open_32_float(uint32_t i){
    return (0.5f+(i>>9))*R123_0x1p_23f; /* 0x1.p-23f; */
}

#if R123_USE_U01_DOUBLE
/* narrowing conversions:  uint64_t to double */
R123_CUDA_DEVICE R123_STATIC_INLINE double u01fixedpt_closed_closed_64_double(uint64_t i){
    /* N.B.  we ignore the high bit, so output is not monotonic */
    return ((i&R123_64BIT(0x7ffffffffffffe00)) + (i&0x200))*R123_0x1p_63; /* 0x1.p-63; */
}

R123_CUDA_DEVICE R123_STATIC_INLINE double u01fixedpt_closed_open_64_double(uint64_t i){
    return (i>>11)*R123_0x1p_53; /* 0x1.0p-53; */
}

R123_CUDA_DEVICE R123_STATIC_INLINE double u01fixedpt_open_closed_64_double(uint64_t i){
    return (1+(i>>11))*R123_0x1p_53; /* 0x1.0p-53; */
}

R123_CUDA_DEVICE R123_STATIC_INLINE double u01fixedpt_open_open_64_double(uint64_t i){
    return (0.5+(i>>12))*R123_0x1p_52; /* 0x1.0p-52; */
}

/* widening conversions:  u32 to double */
R123_CUDA_DEVICE R123_STATIC_INLINE double u01fixedpt_closed_closed_32_double(uint32_t i){
    /* j = i+(i&1) takes on 2^31+1 possible values with a 'trapezoid' distribution:
      p_j =  1 0 2 0 2 .... 2 0 2 0 1
      j   =  0 1 2 3 4 ....        2^32
      by converting to double *before* doing the add, we don't wrap the high bit.
    */
    return (((double)(i&1)) + i)*R123_0x1p_32; /* 0x1.p-32; */
}

R123_CUDA_DEVICE R123_STATIC_INLINE double u01fixedpt_closed_open_32_double(uint32_t i){
    return i*R123_0x1p_32; /* 0x1.p-32; */
}

R123_CUDA_DEVICE R123_STATIC_INLINE double u01fixedpt_open_closed_32_double(uint32_t i){
    return (1.+i)*R123_0x1p_32; /* 0x1.p-32; */
}

R123_CUDA_DEVICE R123_STATIC_INLINE double u01fixedpt_open_open_32_double(uint32_t i){
    return (0.5+i)*R123_0x1p_32; /* 0x1.p-32; */
}
#endif /* R123_USE_U01_DOUBLE */

#ifdef __cplusplus
}
#endif

#endif