libclc/generic/lib/geometric/length.cl

   1 /*
   2  * Copyright (c) 2014 Advanced Micro Devices, Inc.
   3  *
   4  * Permission is hereby granted, free of charge, to any person obtaining a copy
   5  * of this software and associated documentation files (the "Software"), to deal
   6  * in the Software without restriction, including without limitation the rights
   7  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
   8  * copies of the Software, and to permit persons to whom the Software is
   9  * furnished to do so, subject to the following conditions:
  10  *
  11  * The above copyright notice and this permission notice shall be included in
  12  * all copies or substantial portions of the Software.
  13  *
  14  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  15  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  16  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
  17  * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  18  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
  19  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
  20  * THE SOFTWARE.
  21  */
  22
  23 #include <clc/clc.h>
  24
  25 _CLC_OVERLOAD _CLC_DEF float length(float p) {
  26   return fabs(p);
  27 }
  28
  29 #define V_FLENGTH(p)                     \
  30   float l2 = dot(p, p);                  \
  31                                          \
  32   if (l2 < FLT_MIN) {                    \
  33     p *= 0x1.0p+86F;                     \
  34     return sqrt(dot(p, p)) * 0x1.0p-86F; \
  35   } else if (l2 == INFINITY) {           \
  36     p *= 0x1.0p-65F;                     \
  37     return sqrt(dot(p, p)) * 0x1.0p+65F; \
  38   }                                      \
  39                                          \
  40   return sqrt(l2);
  41
  42 _CLC_OVERLOAD _CLC_DEF float length(float2 p) {
  43   V_FLENGTH(p);
  44 }
  45
  46 _CLC_OVERLOAD _CLC_DEF float length(float3 p) {
  47   V_FLENGTH(p);
  48 }
  49
  50 _CLC_OVERLOAD _CLC_DEF float length(float4 p) {
  51   V_FLENGTH(p);
  52 }
  53
  54 #ifdef cl_khr_fp64
  55 #pragma OPENCL EXTENSION cl_khr_fp64 : enable
  56
  57 _CLC_OVERLOAD _CLC_DEF double length(double p){
  58   return fabs(p);
  59 }
  60
  61 #define V_DLENGTH(p)                       \
  62   double l2 = dot(p, p);                   \
  63                                            \
  64   if (l2 < DBL_MIN) {                      \
  65       p *= 0x1.0p+563;                     \
  66       return sqrt(dot(p, p)) * 0x1.0p-563; \
  67   } else if (l2 == INFINITY) {             \
  68       p *= 0x1.0p-513;                     \
  69       return sqrt(dot(p, p)) * 0x1.0p+513; \
  70   }                                        \
  71                                            \
  72   return sqrt(l2);
  73
  74 _CLC_OVERLOAD _CLC_DEF double length(double2 p) {
  75   V_DLENGTH(p);
  76 }
  77
  78 _CLC_OVERLOAD _CLC_DEF double length(double3 p) {
  79   V_DLENGTH(p);
  80 }
  81
  82 _CLC_OVERLOAD _CLC_DEF double length(double4 p) {
  83   V_DLENGTH(p);
  84 }
  85
  86 #endif
  87
  88 #ifdef cl_khr_fp16
  89 #pragma OPENCL EXTENSION cl_khr_fp16 : enable
  90
  91 _CLC_OVERLOAD _CLC_DEF half length(half p){
  92   return fabs(p);
  93 }
  94
  95 // Only available in CLC1.2
  96 #ifndef HALF_MIN
  97 #define HALF_MIN   0x1.0p-14h
  98 #endif
  99
 100 #define V_HLENGTH(p)                       \
 101   half l2 = dot(p, p);                     \
 102                                            \
 103   if (l2 < HALF_MIN) {                     \
 104       p *= 0x1.0p+12h;                     \
 105       return sqrt(dot(p, p)) * 0x1.0p-12h; \
 106   } else if (l2 == INFINITY) {             \
 107       p *= 0x1.0p-7h;                      \
 108       return sqrt(dot(p, p)) * 0x1.0p+7h;  \
 109   }                                        \
 110                                            \
 111   return sqrt(l2);
 112
 113 _CLC_OVERLOAD _CLC_DEF half length(half2 p) {
 114   V_HLENGTH(p);
 115 }
 116
 117 _CLC_OVERLOAD _CLC_DEF half length(half3 p) {
 118   V_HLENGTH(p);
 119 }
 120
 121 _CLC_OVERLOAD _CLC_DEF half length(half4 p) {
 122   V_HLENGTH(p);
 123 }
 124
 125 #endif