1.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
1.2 +++ b/gfx/cairo/libpixman/src/pixman-combine32.h Wed Dec 31 06:09:35 2014 +0100
1.3 @@ -0,0 +1,272 @@
1.4 +#define COMPONENT_SIZE 8
1.5 +#define MASK 0xff
1.6 +#define ONE_HALF 0x80
1.7 +
1.8 +#define A_SHIFT 8 * 3
1.9 +#define R_SHIFT 8 * 2
1.10 +#define G_SHIFT 8
1.11 +#define A_MASK 0xff000000
1.12 +#define R_MASK 0xff0000
1.13 +#define G_MASK 0xff00
1.14 +
1.15 +#define RB_MASK 0xff00ff
1.16 +#define AG_MASK 0xff00ff00
1.17 +#define RB_ONE_HALF 0x800080
1.18 +#define RB_MASK_PLUS_ONE 0x10000100
1.19 +
1.20 +#define ALPHA_8(x) ((x) >> A_SHIFT)
1.21 +#define RED_8(x) (((x) >> R_SHIFT) & MASK)
1.22 +#define GREEN_8(x) (((x) >> G_SHIFT) & MASK)
1.23 +#define BLUE_8(x) ((x) & MASK)
1.24 +
1.25 +/*
1.26 + * ARMv6 has UQADD8 instruction, which implements unsigned saturated
1.27 + * addition for 8-bit values packed in 32-bit registers. It is very useful
1.28 + * for UN8x4_ADD_UN8x4, UN8_rb_ADD_UN8_rb and ADD_UN8 macros (which would
1.29 + * otherwise need a lot of arithmetic operations to simulate this operation).
1.30 + * Since most of the major ARM linux distros are built for ARMv7, we are
1.31 + * much less dependent on runtime CPU detection and can get practical
1.32 + * benefits from conditional compilation here for a lot of users.
1.33 + */
1.34 +
1.35 +#if defined(USE_GCC_INLINE_ASM) && defined(__arm__) && \
1.36 + !defined(__aarch64__) && (!defined(__thumb__) || defined(__thumb2__))
1.37 +#if defined(__ARM_ARCH_6__) || defined(__ARM_ARCH_6J__) || \
1.38 + defined(__ARM_ARCH_6K__) || defined(__ARM_ARCH_6Z__) || \
1.39 + defined(__ARM_ARCH_6ZK__) || defined(__ARM_ARCH_6T2__) || \
1.40 + defined(__ARM_ARCH_6M__) || defined(__ARM_ARCH_7__) || \
1.41 + defined(__ARM_ARCH_7A__) || defined(__ARM_ARCH_7R__) || \
1.42 + defined(__ARM_ARCH_7M__) || defined(__ARM_ARCH_7EM__)
1.43 +
1.44 +static force_inline uint32_t
1.45 +un8x4_add_un8x4 (uint32_t x, uint32_t y)
1.46 +{
1.47 + uint32_t t;
1.48 + asm ("uqadd8 %0, %1, %2" : "=r" (t) : "%r" (x), "r" (y));
1.49 + return t;
1.50 +}
1.51 +
1.52 +#define UN8x4_ADD_UN8x4(x, y) \
1.53 + ((x) = un8x4_add_un8x4 ((x), (y)))
1.54 +
1.55 +#define UN8_rb_ADD_UN8_rb(x, y, t) \
1.56 + ((t) = un8x4_add_un8x4 ((x), (y)), (x) = (t))
1.57 +
1.58 +#define ADD_UN8(x, y, t) \
1.59 + ((t) = (x), un8x4_add_un8x4 ((t), (y)))
1.60 +
1.61 +#endif
1.62 +#endif
1.63 +
1.64 +/*****************************************************************************/
1.65 +
1.66 +/*
1.67 + * Helper macros.
1.68 + */
1.69 +
1.70 +#define MUL_UN8(a, b, t) \
1.71 + ((t) = (a) * (uint16_t)(b) + ONE_HALF, ((((t) >> G_SHIFT ) + (t) ) >> G_SHIFT ))
1.72 +
1.73 +#define DIV_UN8(a, b) \
1.74 + (((uint16_t) (a) * MASK + ((b) / 2)) / (b))
1.75 +
1.76 +#ifndef ADD_UN8
1.77 +#define ADD_UN8(x, y, t) \
1.78 + ((t) = (x) + (y), \
1.79 + (uint32_t) (uint8_t) ((t) | (0 - ((t) >> G_SHIFT))))
1.80 +#endif
1.81 +
1.82 +#define DIV_ONE_UN8(x) \
1.83 + (((x) + ONE_HALF + (((x) + ONE_HALF) >> G_SHIFT)) >> G_SHIFT)
1.84 +
1.85 +/*
1.86 + * The methods below use some tricks to be able to do two color
1.87 + * components at the same time.
1.88 + */
1.89 +
1.90 +/*
1.91 + * x_rb = (x_rb * a) / 255
1.92 + */
1.93 +#define UN8_rb_MUL_UN8(x, a, t) \
1.94 + do \
1.95 + { \
1.96 + t = ((x) & RB_MASK) * (a); \
1.97 + t += RB_ONE_HALF; \
1.98 + x = (t + ((t >> G_SHIFT) & RB_MASK)) >> G_SHIFT; \
1.99 + x &= RB_MASK; \
1.100 + } while (0)
1.101 +
1.102 +/*
1.103 + * x_rb = min (x_rb + y_rb, 255)
1.104 + */
1.105 +#ifndef UN8_rb_ADD_UN8_rb
1.106 +#define UN8_rb_ADD_UN8_rb(x, y, t) \
1.107 + do \
1.108 + { \
1.109 + t = ((x) + (y)); \
1.110 + t |= RB_MASK_PLUS_ONE - ((t >> G_SHIFT) & RB_MASK); \
1.111 + x = (t & RB_MASK); \
1.112 + } while (0)
1.113 +#endif
1.114 +
1.115 +/*
1.116 + * x_rb = (x_rb * a_rb) / 255
1.117 + */
1.118 +#define UN8_rb_MUL_UN8_rb(x, a, t) \
1.119 + do \
1.120 + { \
1.121 + t = (x & MASK) * (a & MASK); \
1.122 + t |= (x & R_MASK) * ((a >> R_SHIFT) & MASK); \
1.123 + t += RB_ONE_HALF; \
1.124 + t = (t + ((t >> G_SHIFT) & RB_MASK)) >> G_SHIFT; \
1.125 + x = t & RB_MASK; \
1.126 + } while (0)
1.127 +
1.128 +/*
1.129 + * x_c = (x_c * a) / 255
1.130 + */
1.131 +#define UN8x4_MUL_UN8(x, a) \
1.132 + do \
1.133 + { \
1.134 + uint32_t r1__, r2__, t__; \
1.135 + \
1.136 + r1__ = (x); \
1.137 + UN8_rb_MUL_UN8 (r1__, (a), t__); \
1.138 + \
1.139 + r2__ = (x) >> G_SHIFT; \
1.140 + UN8_rb_MUL_UN8 (r2__, (a), t__); \
1.141 + \
1.142 + (x) = r1__ | (r2__ << G_SHIFT); \
1.143 + } while (0)
1.144 +
1.145 +/*
1.146 + * x_c = (x_c * a) / 255 + y_c
1.147 + */
1.148 +#define UN8x4_MUL_UN8_ADD_UN8x4(x, a, y) \
1.149 + do \
1.150 + { \
1.151 + uint32_t r1__, r2__, r3__, t__; \
1.152 + \
1.153 + r1__ = (x); \
1.154 + r2__ = (y) & RB_MASK; \
1.155 + UN8_rb_MUL_UN8 (r1__, (a), t__); \
1.156 + UN8_rb_ADD_UN8_rb (r1__, r2__, t__); \
1.157 + \
1.158 + r2__ = (x) >> G_SHIFT; \
1.159 + r3__ = ((y) >> G_SHIFT) & RB_MASK; \
1.160 + UN8_rb_MUL_UN8 (r2__, (a), t__); \
1.161 + UN8_rb_ADD_UN8_rb (r2__, r3__, t__); \
1.162 + \
1.163 + (x) = r1__ | (r2__ << G_SHIFT); \
1.164 + } while (0)
1.165 +
1.166 +/*
1.167 + * x_c = (x_c * a + y_c * b) / 255
1.168 + */
1.169 +#define UN8x4_MUL_UN8_ADD_UN8x4_MUL_UN8(x, a, y, b) \
1.170 + do \
1.171 + { \
1.172 + uint32_t r1__, r2__, r3__, t__; \
1.173 + \
1.174 + r1__ = (x); \
1.175 + r2__ = (y); \
1.176 + UN8_rb_MUL_UN8 (r1__, (a), t__); \
1.177 + UN8_rb_MUL_UN8 (r2__, (b), t__); \
1.178 + UN8_rb_ADD_UN8_rb (r1__, r2__, t__); \
1.179 + \
1.180 + r2__ = ((x) >> G_SHIFT); \
1.181 + r3__ = ((y) >> G_SHIFT); \
1.182 + UN8_rb_MUL_UN8 (r2__, (a), t__); \
1.183 + UN8_rb_MUL_UN8 (r3__, (b), t__); \
1.184 + UN8_rb_ADD_UN8_rb (r2__, r3__, t__); \
1.185 + \
1.186 + (x) = r1__ | (r2__ << G_SHIFT); \
1.187 + } while (0)
1.188 +
1.189 +/*
1.190 + * x_c = (x_c * a_c) / 255
1.191 + */
1.192 +#define UN8x4_MUL_UN8x4(x, a) \
1.193 + do \
1.194 + { \
1.195 + uint32_t r1__, r2__, r3__, t__; \
1.196 + \
1.197 + r1__ = (x); \
1.198 + r2__ = (a); \
1.199 + UN8_rb_MUL_UN8_rb (r1__, r2__, t__); \
1.200 + \
1.201 + r2__ = (x) >> G_SHIFT; \
1.202 + r3__ = (a) >> G_SHIFT; \
1.203 + UN8_rb_MUL_UN8_rb (r2__, r3__, t__); \
1.204 + \
1.205 + (x) = r1__ | (r2__ << G_SHIFT); \
1.206 + } while (0)
1.207 +
1.208 +/*
1.209 + * x_c = (x_c * a_c) / 255 + y_c
1.210 + */
1.211 +#define UN8x4_MUL_UN8x4_ADD_UN8x4(x, a, y) \
1.212 + do \
1.213 + { \
1.214 + uint32_t r1__, r2__, r3__, t__; \
1.215 + \
1.216 + r1__ = (x); \
1.217 + r2__ = (a); \
1.218 + UN8_rb_MUL_UN8_rb (r1__, r2__, t__); \
1.219 + r2__ = (y) & RB_MASK; \
1.220 + UN8_rb_ADD_UN8_rb (r1__, r2__, t__); \
1.221 + \
1.222 + r2__ = ((x) >> G_SHIFT); \
1.223 + r3__ = ((a) >> G_SHIFT); \
1.224 + UN8_rb_MUL_UN8_rb (r2__, r3__, t__); \
1.225 + r3__ = ((y) >> G_SHIFT) & RB_MASK; \
1.226 + UN8_rb_ADD_UN8_rb (r2__, r3__, t__); \
1.227 + \
1.228 + (x) = r1__ | (r2__ << G_SHIFT); \
1.229 + } while (0)
1.230 +
1.231 +/*
1.232 + * x_c = (x_c * a_c + y_c * b) / 255
1.233 + */
1.234 +#define UN8x4_MUL_UN8x4_ADD_UN8x4_MUL_UN8(x, a, y, b) \
1.235 + do \
1.236 + { \
1.237 + uint32_t r1__, r2__, r3__, t__; \
1.238 + \
1.239 + r1__ = (x); \
1.240 + r2__ = (a); \
1.241 + UN8_rb_MUL_UN8_rb (r1__, r2__, t__); \
1.242 + r2__ = (y); \
1.243 + UN8_rb_MUL_UN8 (r2__, (b), t__); \
1.244 + UN8_rb_ADD_UN8_rb (r1__, r2__, t__); \
1.245 + \
1.246 + r2__ = (x) >> G_SHIFT; \
1.247 + r3__ = (a) >> G_SHIFT; \
1.248 + UN8_rb_MUL_UN8_rb (r2__, r3__, t__); \
1.249 + r3__ = (y) >> G_SHIFT; \
1.250 + UN8_rb_MUL_UN8 (r3__, (b), t__); \
1.251 + UN8_rb_ADD_UN8_rb (r2__, r3__, t__); \
1.252 + \
1.253 + x = r1__ | (r2__ << G_SHIFT); \
1.254 + } while (0)
1.255 +
1.256 +/*
1.257 + x_c = min(x_c + y_c, 255)
1.258 +*/
1.259 +#ifndef UN8x4_ADD_UN8x4
1.260 +#define UN8x4_ADD_UN8x4(x, y) \
1.261 + do \
1.262 + { \
1.263 + uint32_t r1__, r2__, r3__, t__; \
1.264 + \
1.265 + r1__ = (x) & RB_MASK; \
1.266 + r2__ = (y) & RB_MASK; \
1.267 + UN8_rb_ADD_UN8_rb (r1__, r2__, t__); \
1.268 + \
1.269 + r2__ = ((x) >> G_SHIFT) & RB_MASK; \
1.270 + r3__ = ((y) >> G_SHIFT) & RB_MASK; \
1.271 + UN8_rb_ADD_UN8_rb (r2__, r3__, t__); \
1.272 + \
1.273 + x = r1__ | (r2__ << G_SHIFT); \
1.274 + } while (0)
1.275 +#endif
