1 /* 2 * Copyright 2009 Red Hat Inc. 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 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 17 * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR 18 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, 19 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR 20 * OTHER DEALINGS IN THE SOFTWARE. 21 * 22 * Authors: Dave Airlie 23 * Christian König 24 */ 25 #ifndef DRM_FIXED_H 26 #define DRM_FIXED_H 27 28 #include <linux/math64.h> 29 #include <linux/types.h> 30 #include <linux/wordpart.h> 31 32 typedef union dfixed { 33 u32 full; 34 } fixed20_12; 35 36 37 #define dfixed_const(A) (u32)(((A) << 12))/* + ((B + 0.000122)*4096)) */ 38 #define dfixed_const_half(A) (u32)(((A) << 12) + 2048) 39 #define dfixed_const_666(A) (u32)(((A) << 12) + 2731) 40 #define dfixed_const_8(A) (u32)(((A) << 12) + 3277) 41 #define dfixed_mul(A, B) ((u64)((u64)(A).full * (B).full + 2048) >> 12) 42 #define dfixed_init(A) { .full = dfixed_const((A)) } 43 #define dfixed_init_half(A) { .full = dfixed_const_half((A)) } 44 #define dfixed_trunc(A) ((A).full >> 12) 45 #define dfixed_frac(A) ((A).full & ((1 << 12) - 1)) 46 47 static inline u32 dfixed_floor(fixed20_12 A) 48 { 49 u32 non_frac = dfixed_trunc(A); 50 51 return dfixed_const(non_frac); 52 } 53 54 static inline u32 dfixed_ceil(fixed20_12 A) 55 { 56 u32 non_frac = dfixed_trunc(A); 57 58 if (A.full > dfixed_const(non_frac)) 59 return dfixed_const(non_frac + 1); 60 else 61 return dfixed_const(non_frac); 62 } 63 64 static inline u32 dfixed_div(fixed20_12 A, fixed20_12 B) 65 { 66 u64 tmp = ((u64)A.full << 13); 67 68 do_div(tmp, B.full); 69 tmp += 1; 70 tmp /= 2; 71 return lower_32_bits(tmp); 72 } 73 74 #define DRM_FIXED_POINT 32 75 #define DRM_FIXED_ONE (1ULL << DRM_FIXED_POINT) 76 #define DRM_FIXED_DECIMAL_MASK (DRM_FIXED_ONE - 1) 77 #define DRM_FIXED_DIGITS_MASK (~DRM_FIXED_DECIMAL_MASK) 78 #define DRM_FIXED_EPSILON 1LL 79 #define DRM_FIXED_ALMOST_ONE (DRM_FIXED_ONE - DRM_FIXED_EPSILON) 80 81 /** 82 * drm_sm2fixp() - convert signed-magnitude to fixed point 83 * @a: 1.31.32 signed-magnitude fixed point 84 * 85 * Convert a 1.31.32 signed-magnitude fixed point to 32.32 86 * 2s-complement fixed point 87 * 88 * @return s64 2s-complement fixed point 89 */ 90 static inline s64 drm_sm2fixp(__u64 a) 91 { 92 if ((a & (1LL << 63))) { 93 return -(a & 0x7fffffffffffffffll); 94 } else { 95 return a; 96 } 97 } 98 99 static inline s64 drm_int2fixp(int a) 100 { 101 return ((s64)a) << DRM_FIXED_POINT; 102 } 103 104 static inline int drm_fixp2int(s64 a) 105 { 106 return ((s64)a) >> DRM_FIXED_POINT; 107 } 108 109 static inline int drm_fixp2int_round(s64 a) 110 { 111 return drm_fixp2int(a + DRM_FIXED_ONE / 2); 112 } 113 114 static inline int drm_fixp2int_ceil(s64 a) 115 { 116 if (a >= 0) 117 return drm_fixp2int(a + DRM_FIXED_ALMOST_ONE); 118 else 119 return drm_fixp2int(a - DRM_FIXED_ALMOST_ONE); 120 } 121 122 static inline unsigned drm_fixp_msbset(s64 a) 123 { 124 unsigned shift, sign = (a >> 63) & 1; 125 126 for (shift = 62; shift > 0; --shift) 127 if (((a >> shift) & 1) != sign) 128 return shift; 129 130 return 0; 131 } 132 133 static inline s64 drm_fixp_mul(s64 a, s64 b) 134 { 135 unsigned shift = drm_fixp_msbset(a) + drm_fixp_msbset(b); 136 s64 result; 137 138 if (shift > 61) { 139 shift = shift - 61; 140 a >>= (shift >> 1) + (shift & 1); 141 b >>= shift >> 1; 142 } else 143 shift = 0; 144 145 result = a * b; 146 147 if (shift > DRM_FIXED_POINT) 148 return result << (shift - DRM_FIXED_POINT); 149 150 if (shift < DRM_FIXED_POINT) 151 return result >> (DRM_FIXED_POINT - shift); 152 153 return result; 154 } 155 156 static inline s64 drm_fixp_div(s64 a, s64 b) 157 { 158 unsigned shift = 62 - drm_fixp_msbset(a); 159 s64 result; 160 161 a <<= shift; 162 163 if (shift < DRM_FIXED_POINT) 164 b >>= (DRM_FIXED_POINT - shift); 165 166 result = div64_s64(a, b); 167 168 if (shift > DRM_FIXED_POINT) 169 return result >> (shift - DRM_FIXED_POINT); 170 171 return result; 172 } 173 174 static inline s64 drm_fixp_from_fraction(s64 a, s64 b) 175 { 176 s64 res; 177 bool a_neg = a < 0; 178 bool b_neg = b < 0; 179 u64 a_abs = a_neg ? -a : a; 180 u64 b_abs = b_neg ? -b : b; 181 u64 rem; 182 183 /* determine integer part */ 184 u64 res_abs = div64_u64_rem(a_abs, b_abs, &rem); 185 186 /* determine fractional part */ 187 { 188 u32 i = DRM_FIXED_POINT; 189 190 do { 191 rem <<= 1; 192 res_abs <<= 1; 193 if (rem >= b_abs) { 194 res_abs |= 1; 195 rem -= b_abs; 196 } 197 } while (--i != 0); 198 } 199 200 /* round up LSB */ 201 { 202 u64 summand = (rem << 1) >= b_abs; 203 204 res_abs += summand; 205 } 206 207 res = (s64) res_abs; 208 if (a_neg ^ b_neg) 209 res = -res; 210 return res; 211 } 212 213 static inline s64 drm_fixp_exp(s64 x) 214 { 215 s64 tolerance = div64_s64(DRM_FIXED_ONE, 1000000); 216 s64 sum = DRM_FIXED_ONE, term, y = x; 217 u64 count = 1; 218 219 if (x < 0) 220 y = -1 * x; 221 222 term = y; 223 224 while (term >= tolerance) { 225 sum = sum + term; 226 count = count + 1; 227 term = drm_fixp_mul(term, div64_s64(y, count)); 228 } 229 230 if (x < 0) 231 sum = drm_fixp_div(DRM_FIXED_ONE, sum); 232 233 return sum; 234 } 235 236 static inline int fxp_q4_from_int(int val_int) 237 { 238 return val_int << 4; 239 } 240 241 static inline int fxp_q4_to_int(int val_q4) 242 { 243 return val_q4 >> 4; 244 } 245 246 static inline int fxp_q4_to_int_roundup(int val_q4) 247 { 248 return (val_q4 + 0xf) >> 4; 249 } 250 251 static inline int fxp_q4_to_frac(int val_q4) 252 { 253 return val_q4 & 0xf; 254 } 255 256 #define FXP_Q4_FMT "%d.%04d" 257 #define FXP_Q4_ARGS(val_q4) fxp_q4_to_int(val_q4), (fxp_q4_to_frac(val_q4) * 625) 258 259 #endif 260