Wed, 31 Dec 2014 06:09:35 +0100
Cloned upstream origin tor-browser at tor-browser-31.3.0esr-4.5-1-build1
revision ID fc1c9ff7c1b2defdbc039f12214767608f46423f for hacking purpose.
michael@0 | 1 | #define COMPONENT_SIZE 8 |
michael@0 | 2 | #define MASK 0xff |
michael@0 | 3 | #define ONE_HALF 0x80 |
michael@0 | 4 | |
michael@0 | 5 | #define A_SHIFT 8 * 3 |
michael@0 | 6 | #define R_SHIFT 8 * 2 |
michael@0 | 7 | #define G_SHIFT 8 |
michael@0 | 8 | #define A_MASK 0xff000000 |
michael@0 | 9 | #define R_MASK 0xff0000 |
michael@0 | 10 | #define G_MASK 0xff00 |
michael@0 | 11 | |
michael@0 | 12 | #define RB_MASK 0xff00ff |
michael@0 | 13 | #define AG_MASK 0xff00ff00 |
michael@0 | 14 | #define RB_ONE_HALF 0x800080 |
michael@0 | 15 | #define RB_MASK_PLUS_ONE 0x10000100 |
michael@0 | 16 | |
michael@0 | 17 | #define ALPHA_8(x) ((x) >> A_SHIFT) |
michael@0 | 18 | #define RED_8(x) (((x) >> R_SHIFT) & MASK) |
michael@0 | 19 | #define GREEN_8(x) (((x) >> G_SHIFT) & MASK) |
michael@0 | 20 | #define BLUE_8(x) ((x) & MASK) |
michael@0 | 21 | |
michael@0 | 22 | /* |
michael@0 | 23 | * ARMv6 has UQADD8 instruction, which implements unsigned saturated |
michael@0 | 24 | * addition for 8-bit values packed in 32-bit registers. It is very useful |
michael@0 | 25 | * for UN8x4_ADD_UN8x4, UN8_rb_ADD_UN8_rb and ADD_UN8 macros (which would |
michael@0 | 26 | * otherwise need a lot of arithmetic operations to simulate this operation). |
michael@0 | 27 | * Since most of the major ARM linux distros are built for ARMv7, we are |
michael@0 | 28 | * much less dependent on runtime CPU detection and can get practical |
michael@0 | 29 | * benefits from conditional compilation here for a lot of users. |
michael@0 | 30 | */ |
michael@0 | 31 | |
michael@0 | 32 | #if defined(USE_GCC_INLINE_ASM) && defined(__arm__) && \ |
michael@0 | 33 | !defined(__aarch64__) && (!defined(__thumb__) || defined(__thumb2__)) |
michael@0 | 34 | #if defined(__ARM_ARCH_6__) || defined(__ARM_ARCH_6J__) || \ |
michael@0 | 35 | defined(__ARM_ARCH_6K__) || defined(__ARM_ARCH_6Z__) || \ |
michael@0 | 36 | defined(__ARM_ARCH_6ZK__) || defined(__ARM_ARCH_6T2__) || \ |
michael@0 | 37 | defined(__ARM_ARCH_6M__) || defined(__ARM_ARCH_7__) || \ |
michael@0 | 38 | defined(__ARM_ARCH_7A__) || defined(__ARM_ARCH_7R__) || \ |
michael@0 | 39 | defined(__ARM_ARCH_7M__) || defined(__ARM_ARCH_7EM__) |
michael@0 | 40 | |
michael@0 | 41 | static force_inline uint32_t |
michael@0 | 42 | un8x4_add_un8x4 (uint32_t x, uint32_t y) |
michael@0 | 43 | { |
michael@0 | 44 | uint32_t t; |
michael@0 | 45 | asm ("uqadd8 %0, %1, %2" : "=r" (t) : "%r" (x), "r" (y)); |
michael@0 | 46 | return t; |
michael@0 | 47 | } |
michael@0 | 48 | |
michael@0 | 49 | #define UN8x4_ADD_UN8x4(x, y) \ |
michael@0 | 50 | ((x) = un8x4_add_un8x4 ((x), (y))) |
michael@0 | 51 | |
michael@0 | 52 | #define UN8_rb_ADD_UN8_rb(x, y, t) \ |
michael@0 | 53 | ((t) = un8x4_add_un8x4 ((x), (y)), (x) = (t)) |
michael@0 | 54 | |
michael@0 | 55 | #define ADD_UN8(x, y, t) \ |
michael@0 | 56 | ((t) = (x), un8x4_add_un8x4 ((t), (y))) |
michael@0 | 57 | |
michael@0 | 58 | #endif |
michael@0 | 59 | #endif |
michael@0 | 60 | |
michael@0 | 61 | /*****************************************************************************/ |
michael@0 | 62 | |
michael@0 | 63 | /* |
michael@0 | 64 | * Helper macros. |
michael@0 | 65 | */ |
michael@0 | 66 | |
michael@0 | 67 | #define MUL_UN8(a, b, t) \ |
michael@0 | 68 | ((t) = (a) * (uint16_t)(b) + ONE_HALF, ((((t) >> G_SHIFT ) + (t) ) >> G_SHIFT )) |
michael@0 | 69 | |
michael@0 | 70 | #define DIV_UN8(a, b) \ |
michael@0 | 71 | (((uint16_t) (a) * MASK + ((b) / 2)) / (b)) |
michael@0 | 72 | |
michael@0 | 73 | #ifndef ADD_UN8 |
michael@0 | 74 | #define ADD_UN8(x, y, t) \ |
michael@0 | 75 | ((t) = (x) + (y), \ |
michael@0 | 76 | (uint32_t) (uint8_t) ((t) | (0 - ((t) >> G_SHIFT)))) |
michael@0 | 77 | #endif |
michael@0 | 78 | |
michael@0 | 79 | #define DIV_ONE_UN8(x) \ |
michael@0 | 80 | (((x) + ONE_HALF + (((x) + ONE_HALF) >> G_SHIFT)) >> G_SHIFT) |
michael@0 | 81 | |
michael@0 | 82 | /* |
michael@0 | 83 | * The methods below use some tricks to be able to do two color |
michael@0 | 84 | * components at the same time. |
michael@0 | 85 | */ |
michael@0 | 86 | |
michael@0 | 87 | /* |
michael@0 | 88 | * x_rb = (x_rb * a) / 255 |
michael@0 | 89 | */ |
michael@0 | 90 | #define UN8_rb_MUL_UN8(x, a, t) \ |
michael@0 | 91 | do \ |
michael@0 | 92 | { \ |
michael@0 | 93 | t = ((x) & RB_MASK) * (a); \ |
michael@0 | 94 | t += RB_ONE_HALF; \ |
michael@0 | 95 | x = (t + ((t >> G_SHIFT) & RB_MASK)) >> G_SHIFT; \ |
michael@0 | 96 | x &= RB_MASK; \ |
michael@0 | 97 | } while (0) |
michael@0 | 98 | |
michael@0 | 99 | /* |
michael@0 | 100 | * x_rb = min (x_rb + y_rb, 255) |
michael@0 | 101 | */ |
michael@0 | 102 | #ifndef UN8_rb_ADD_UN8_rb |
michael@0 | 103 | #define UN8_rb_ADD_UN8_rb(x, y, t) \ |
michael@0 | 104 | do \ |
michael@0 | 105 | { \ |
michael@0 | 106 | t = ((x) + (y)); \ |
michael@0 | 107 | t |= RB_MASK_PLUS_ONE - ((t >> G_SHIFT) & RB_MASK); \ |
michael@0 | 108 | x = (t & RB_MASK); \ |
michael@0 | 109 | } while (0) |
michael@0 | 110 | #endif |
michael@0 | 111 | |
michael@0 | 112 | /* |
michael@0 | 113 | * x_rb = (x_rb * a_rb) / 255 |
michael@0 | 114 | */ |
michael@0 | 115 | #define UN8_rb_MUL_UN8_rb(x, a, t) \ |
michael@0 | 116 | do \ |
michael@0 | 117 | { \ |
michael@0 | 118 | t = (x & MASK) * (a & MASK); \ |
michael@0 | 119 | t |= (x & R_MASK) * ((a >> R_SHIFT) & MASK); \ |
michael@0 | 120 | t += RB_ONE_HALF; \ |
michael@0 | 121 | t = (t + ((t >> G_SHIFT) & RB_MASK)) >> G_SHIFT; \ |
michael@0 | 122 | x = t & RB_MASK; \ |
michael@0 | 123 | } while (0) |
michael@0 | 124 | |
michael@0 | 125 | /* |
michael@0 | 126 | * x_c = (x_c * a) / 255 |
michael@0 | 127 | */ |
michael@0 | 128 | #define UN8x4_MUL_UN8(x, a) \ |
michael@0 | 129 | do \ |
michael@0 | 130 | { \ |
michael@0 | 131 | uint32_t r1__, r2__, t__; \ |
michael@0 | 132 | \ |
michael@0 | 133 | r1__ = (x); \ |
michael@0 | 134 | UN8_rb_MUL_UN8 (r1__, (a), t__); \ |
michael@0 | 135 | \ |
michael@0 | 136 | r2__ = (x) >> G_SHIFT; \ |
michael@0 | 137 | UN8_rb_MUL_UN8 (r2__, (a), t__); \ |
michael@0 | 138 | \ |
michael@0 | 139 | (x) = r1__ | (r2__ << G_SHIFT); \ |
michael@0 | 140 | } while (0) |
michael@0 | 141 | |
michael@0 | 142 | /* |
michael@0 | 143 | * x_c = (x_c * a) / 255 + y_c |
michael@0 | 144 | */ |
michael@0 | 145 | #define UN8x4_MUL_UN8_ADD_UN8x4(x, a, y) \ |
michael@0 | 146 | do \ |
michael@0 | 147 | { \ |
michael@0 | 148 | uint32_t r1__, r2__, r3__, t__; \ |
michael@0 | 149 | \ |
michael@0 | 150 | r1__ = (x); \ |
michael@0 | 151 | r2__ = (y) & RB_MASK; \ |
michael@0 | 152 | UN8_rb_MUL_UN8 (r1__, (a), t__); \ |
michael@0 | 153 | UN8_rb_ADD_UN8_rb (r1__, r2__, t__); \ |
michael@0 | 154 | \ |
michael@0 | 155 | r2__ = (x) >> G_SHIFT; \ |
michael@0 | 156 | r3__ = ((y) >> G_SHIFT) & RB_MASK; \ |
michael@0 | 157 | UN8_rb_MUL_UN8 (r2__, (a), t__); \ |
michael@0 | 158 | UN8_rb_ADD_UN8_rb (r2__, r3__, t__); \ |
michael@0 | 159 | \ |
michael@0 | 160 | (x) = r1__ | (r2__ << G_SHIFT); \ |
michael@0 | 161 | } while (0) |
michael@0 | 162 | |
michael@0 | 163 | /* |
michael@0 | 164 | * x_c = (x_c * a + y_c * b) / 255 |
michael@0 | 165 | */ |
michael@0 | 166 | #define UN8x4_MUL_UN8_ADD_UN8x4_MUL_UN8(x, a, y, b) \ |
michael@0 | 167 | do \ |
michael@0 | 168 | { \ |
michael@0 | 169 | uint32_t r1__, r2__, r3__, t__; \ |
michael@0 | 170 | \ |
michael@0 | 171 | r1__ = (x); \ |
michael@0 | 172 | r2__ = (y); \ |
michael@0 | 173 | UN8_rb_MUL_UN8 (r1__, (a), t__); \ |
michael@0 | 174 | UN8_rb_MUL_UN8 (r2__, (b), t__); \ |
michael@0 | 175 | UN8_rb_ADD_UN8_rb (r1__, r2__, t__); \ |
michael@0 | 176 | \ |
michael@0 | 177 | r2__ = ((x) >> G_SHIFT); \ |
michael@0 | 178 | r3__ = ((y) >> G_SHIFT); \ |
michael@0 | 179 | UN8_rb_MUL_UN8 (r2__, (a), t__); \ |
michael@0 | 180 | UN8_rb_MUL_UN8 (r3__, (b), t__); \ |
michael@0 | 181 | UN8_rb_ADD_UN8_rb (r2__, r3__, t__); \ |
michael@0 | 182 | \ |
michael@0 | 183 | (x) = r1__ | (r2__ << G_SHIFT); \ |
michael@0 | 184 | } while (0) |
michael@0 | 185 | |
michael@0 | 186 | /* |
michael@0 | 187 | * x_c = (x_c * a_c) / 255 |
michael@0 | 188 | */ |
michael@0 | 189 | #define UN8x4_MUL_UN8x4(x, a) \ |
michael@0 | 190 | do \ |
michael@0 | 191 | { \ |
michael@0 | 192 | uint32_t r1__, r2__, r3__, t__; \ |
michael@0 | 193 | \ |
michael@0 | 194 | r1__ = (x); \ |
michael@0 | 195 | r2__ = (a); \ |
michael@0 | 196 | UN8_rb_MUL_UN8_rb (r1__, r2__, t__); \ |
michael@0 | 197 | \ |
michael@0 | 198 | r2__ = (x) >> G_SHIFT; \ |
michael@0 | 199 | r3__ = (a) >> G_SHIFT; \ |
michael@0 | 200 | UN8_rb_MUL_UN8_rb (r2__, r3__, t__); \ |
michael@0 | 201 | \ |
michael@0 | 202 | (x) = r1__ | (r2__ << G_SHIFT); \ |
michael@0 | 203 | } while (0) |
michael@0 | 204 | |
michael@0 | 205 | /* |
michael@0 | 206 | * x_c = (x_c * a_c) / 255 + y_c |
michael@0 | 207 | */ |
michael@0 | 208 | #define UN8x4_MUL_UN8x4_ADD_UN8x4(x, a, y) \ |
michael@0 | 209 | do \ |
michael@0 | 210 | { \ |
michael@0 | 211 | uint32_t r1__, r2__, r3__, t__; \ |
michael@0 | 212 | \ |
michael@0 | 213 | r1__ = (x); \ |
michael@0 | 214 | r2__ = (a); \ |
michael@0 | 215 | UN8_rb_MUL_UN8_rb (r1__, r2__, t__); \ |
michael@0 | 216 | r2__ = (y) & RB_MASK; \ |
michael@0 | 217 | UN8_rb_ADD_UN8_rb (r1__, r2__, t__); \ |
michael@0 | 218 | \ |
michael@0 | 219 | r2__ = ((x) >> G_SHIFT); \ |
michael@0 | 220 | r3__ = ((a) >> G_SHIFT); \ |
michael@0 | 221 | UN8_rb_MUL_UN8_rb (r2__, r3__, t__); \ |
michael@0 | 222 | r3__ = ((y) >> G_SHIFT) & RB_MASK; \ |
michael@0 | 223 | UN8_rb_ADD_UN8_rb (r2__, r3__, t__); \ |
michael@0 | 224 | \ |
michael@0 | 225 | (x) = r1__ | (r2__ << G_SHIFT); \ |
michael@0 | 226 | } while (0) |
michael@0 | 227 | |
michael@0 | 228 | /* |
michael@0 | 229 | * x_c = (x_c * a_c + y_c * b) / 255 |
michael@0 | 230 | */ |
michael@0 | 231 | #define UN8x4_MUL_UN8x4_ADD_UN8x4_MUL_UN8(x, a, y, b) \ |
michael@0 | 232 | do \ |
michael@0 | 233 | { \ |
michael@0 | 234 | uint32_t r1__, r2__, r3__, t__; \ |
michael@0 | 235 | \ |
michael@0 | 236 | r1__ = (x); \ |
michael@0 | 237 | r2__ = (a); \ |
michael@0 | 238 | UN8_rb_MUL_UN8_rb (r1__, r2__, t__); \ |
michael@0 | 239 | r2__ = (y); \ |
michael@0 | 240 | UN8_rb_MUL_UN8 (r2__, (b), t__); \ |
michael@0 | 241 | UN8_rb_ADD_UN8_rb (r1__, r2__, t__); \ |
michael@0 | 242 | \ |
michael@0 | 243 | r2__ = (x) >> G_SHIFT; \ |
michael@0 | 244 | r3__ = (a) >> G_SHIFT; \ |
michael@0 | 245 | UN8_rb_MUL_UN8_rb (r2__, r3__, t__); \ |
michael@0 | 246 | r3__ = (y) >> G_SHIFT; \ |
michael@0 | 247 | UN8_rb_MUL_UN8 (r3__, (b), t__); \ |
michael@0 | 248 | UN8_rb_ADD_UN8_rb (r2__, r3__, t__); \ |
michael@0 | 249 | \ |
michael@0 | 250 | x = r1__ | (r2__ << G_SHIFT); \ |
michael@0 | 251 | } while (0) |
michael@0 | 252 | |
michael@0 | 253 | /* |
michael@0 | 254 | x_c = min(x_c + y_c, 255) |
michael@0 | 255 | */ |
michael@0 | 256 | #ifndef UN8x4_ADD_UN8x4 |
michael@0 | 257 | #define UN8x4_ADD_UN8x4(x, y) \ |
michael@0 | 258 | do \ |
michael@0 | 259 | { \ |
michael@0 | 260 | uint32_t r1__, r2__, r3__, t__; \ |
michael@0 | 261 | \ |
michael@0 | 262 | r1__ = (x) & RB_MASK; \ |
michael@0 | 263 | r2__ = (y) & RB_MASK; \ |
michael@0 | 264 | UN8_rb_ADD_UN8_rb (r1__, r2__, t__); \ |
michael@0 | 265 | \ |
michael@0 | 266 | r2__ = ((x) >> G_SHIFT) & RB_MASK; \ |
michael@0 | 267 | r3__ = ((y) >> G_SHIFT) & RB_MASK; \ |
michael@0 | 268 | UN8_rb_ADD_UN8_rb (r2__, r3__, t__); \ |
michael@0 | 269 | \ |
michael@0 | 270 | x = r1__ | (r2__ << G_SHIFT); \ |
michael@0 | 271 | } while (0) |
michael@0 | 272 | #endif |