src/gallium/auxiliary/util/u_format_r11g11b10f.h

   1 /*
   2  * Copyright (C) 2011 Marek Olšák <maraeo@gmail.com>
   3  *
   4  * Permission is hereby granted, free of charge, to any person obtaining a
   5  * copy of this software and associated documentation files (the "Software"),
   6  * to deal in the Software without restriction, including without limitation
   7  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
   8  * and/or sell copies of the Software, and to permit persons to whom the
   9  * Software is furnished to do so, subject to the following conditions:
  10  *
  11  * The above copyright notice and this permission notice (including the next
  12  * paragraph) shall be included in all copies or substantial portions of the
  13  * Software.
  14  *
  15  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  16  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  17  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
  18  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  19  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
  20  * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
  21  * DEALINGS IN THE SOFTWARE.
  22  */
  23
  24 /* Based on code from The OpenGL Programming Guide / 7th Edition, Appendix J.
  25  * Available here: http://www.opengl-redbook.com/appendices/
  26  * The algorithm in the book contains a bug though, which is fixed in the code
  27  * below.
  28  */
  29
  30 #define UF11_EXPONENT_BIAS   15
  31 #define UF11_EXPONENT_BITS   0x1F
  32 #define UF11_EXPONENT_SHIFT  6
  33 #define UF11_MANTISSA_BITS   0x3F
  34 #define UF11_MANTISSA_SHIFT  (23 - UF11_EXPONENT_SHIFT)
  35 #define UF11_MAX_EXPONENT    (UF11_EXPONENT_BITS << UF11_EXPONENT_SHIFT)
  36
  37 #define UF10_EXPONENT_BIAS   15
  38 #define UF10_EXPONENT_BITS   0x1F
  39 #define UF10_EXPONENT_SHIFT  5
  40 #define UF10_MANTISSA_BITS   0x3F
  41 #define UF10_MANTISSA_SHIFT  (23 - UF10_EXPONENT_SHIFT)
  42 #define UF10_MAX_EXPONENT    (UF10_EXPONENT_BITS << UF10_EXPONENT_SHIFT)
  43
  44 #define F32_INFINITY         0x7f800000
  45
  46 static INLINE unsigned f32_to_uf11(float val)
  47 {
  48    uint32_t f32 = (*(uint32_t *) &val);
  49    uint16_t uf11 = 0;
  50
  51    /* Decode little-endian 32-bit floating-point value */
  52    int sign = (f32 >> 16) & 0x8000;
  53    /* Map exponent to the range [-127,128] */
  54    int exponent = ((f32 >> 23) & 0xff) - 127;
  55    int mantissa = f32 & 0x007fffff;
  56
  57    if (sign) return 0;
  58
  59    if (exponent == 128) { /* Infinity or NaN */
  60       uf11 = UF11_MAX_EXPONENT;
  61       if (mantissa) uf11 |= (mantissa & UF11_MANTISSA_BITS);
  62    }
  63    else if (exponent > 15) { /* Overflow - flush to Infinity */
  64       uf11 = UF11_MAX_EXPONENT;
  65    }
  66    else if (exponent > -15) { /* Representable value */
  67       exponent += UF11_EXPONENT_BIAS;
  68       mantissa >>= UF11_MANTISSA_SHIFT;
  69       uf11 = exponent << UF11_EXPONENT_SHIFT | mantissa;
  70    }
  71
  72    return uf11;
  73 }
  74
  75 static INLINE float uf11_to_f32(uint16_t val)
  76 {
  77    union {
  78       float f;
  79       uint32_t ui;
  80    } f32;
  81
  82    int exponent = (val & 0x07c0) >> UF11_EXPONENT_SHIFT;
  83    int mantissa = (val & 0x003f);
  84
  85    f32.f = 0.0;
  86
  87    if (exponent == 0) {
  88       if (mantissa != 0) {
  89          const float scale = 1.0 / (1 << 20);
  90          f32.f = scale * mantissa;
  91       }
  92    }
  93    else if (exponent == 31) {
  94       f32.ui = F32_INFINITY | mantissa;
  95    }
  96    else {
  97       float scale, decimal;
  98       exponent -= 15;
  99       if (exponent < 0) {
 100          scale = 1.0 / (1 << -exponent);
 101       }
 102       else {
 103          scale = 1 << exponent;
 104       }
 105       decimal = 1.0 + (float) mantissa / 64;
 106       f32.f = scale * decimal;
 107    }
 108
 109    return f32.f;
 110 }
 111
 112 static INLINE unsigned f32_to_uf10(float val)
 113 {
 114    uint32_t f32 = (*(uint32_t *) &val);
 115    uint16_t uf10 = 0;
 116
 117    /* Decode little-endian 32-bit floating-point value */
 118    int sign = (f32 >> 16) & 0x8000;
 119    /* Map exponent to the range [-127,128] */
 120    int exponent = ((f32 >> 23) & 0xff) - 127;
 121    int mantissa = f32 & 0x007fffff;
 122
 123    if (sign) return 0;
 124
 125    if (exponent == 128) { /* Infinity or NaN */
 126       uf10 = UF10_MAX_EXPONENT;
 127       if (mantissa) uf10 |= (mantissa & UF10_MANTISSA_BITS);
 128    }
 129    else if (exponent > 15) { /* Overflow - flush to Infinity */
 130       uf10 = UF10_MAX_EXPONENT;
 131    }
 132    else if (exponent > -15) { /* Representable value */
 133       exponent += UF10_EXPONENT_BIAS;
 134       mantissa >>= UF10_MANTISSA_SHIFT;
 135       uf10 = exponent << UF10_EXPONENT_SHIFT | mantissa;
 136    }
 137
 138    return uf10;
 139 }
 140
 141 static INLINE float uf10_to_f32(uint16_t val)
 142 {
 143    union {
 144       float f;
 145       uint32_t ui;
 146    } f32;
 147
 148    int exponent = (val & 0x07c0) >> UF10_EXPONENT_SHIFT;
 149    int mantissa = (val & 0x003f);
 150
 151    f32.f = 0.0;
 152
 153    if (exponent == 0) {
 154       if (mantissa != 0) {
 155          const float scale = 1.0 / (1 << 20);
 156          f32.f = scale * mantissa;
 157       }
 158    }
 159    else if (exponent == 31) {
 160       f32.ui = F32_INFINITY | mantissa;
 161    }
 162    else {
 163       float scale, decimal;
 164       exponent -= 15;
 165       if (exponent < 0) {
 166          scale = 1.0 / (1 << -exponent);
 167       }
 168       else {
 169          scale = 1 << exponent;
 170       }
 171       decimal = 1.0 + (float) mantissa / 32;
 172       f32.f = scale * decimal;
 173    }
 174
 175    return f32.f;
 176 }
 177
 178 static INLINE unsigned float3_to_r11g11b10f(const float rgb[3])
 179 {
 180    return ( f32_to_uf11(rgb[0]) & 0x7ff) |
 181           ((f32_to_uf11(rgb[1]) & 0x7ff) << 11) |
 182           ((f32_to_uf10(rgb[2]) & 0x3ff) << 22);
 183 }
 184
 185 static INLINE void r11g11b10f_to_float3(unsigned rgb, float retval[3])
 186 {
 187    retval[0] = uf11_to_f32( rgb        & 0x7ff);
 188    retval[1] = uf11_to_f32((rgb >> 11) & 0x7ff);
 189    retval[2] = uf10_to_f32((rgb >> 22) & 0x3ff);
 190 }