pixel-a.asm 142 KB
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073 3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358 3359 3360 3361 3362 3363 3364 3365 3366 3367 3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378 3379 3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513 3514 3515 3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570 3571 3572 3573 3574 3575 3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588 3589 3590 3591 3592 3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604 3605 3606 3607 3608 3609 3610 3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624 3625 3626 3627 3628 3629 3630 3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675 3676 3677 3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688 3689 3690 3691 3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713 3714 3715 3716 3717 3718 3719 3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731 3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742 3743 3744 3745 3746 3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785 3786 3787 3788 3789 3790 3791 3792 3793 3794 3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805 3806 3807 3808 3809 3810 3811 3812 3813 3814 3815 3816 3817 3818 3819 3820 3821 3822 3823 3824 3825 3826 3827 3828 3829 3830 3831 3832 3833 3834 3835 3836 3837 3838 3839 3840 3841 3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853 3854 3855 3856 3857 3858 3859 3860 3861 3862 3863 3864 3865 3866 3867 3868 3869 3870 3871 3872 3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887 3888 3889 3890 3891 3892 3893 3894 3895 3896 3897 3898 3899 3900 3901 3902 3903 3904 3905 3906 3907 3908 3909 3910 3911 3912 3913 3914 3915 3916 3917 3918 3919 3920 3921 3922 3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 3933 3934 3935 3936 3937 3938 3939 3940 3941 3942 3943 3944 3945 3946 3947 3948 3949 3950 3951 3952 3953 3954 3955 3956 3957 3958 3959 3960 3961 3962 3963 3964 3965 3966 3967 3968 3969 3970 3971 3972 3973 3974 3975 3976 3977 3978 3979 3980 3981 3982 3983 3984 3985 3986 3987 3988 3989 3990 3991 3992 3993 3994 3995 3996 3997 3998 3999 4000 4001 4002 4003 4004 4005 4006 4007 4008 4009 4010 4011 4012 4013 4014 4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031 4032 4033 4034 4035 4036 4037 4038 4039 4040 4041 4042 4043 4044 4045 4046 4047 4048 4049 4050 4051 4052 4053 4054 4055 4056 4057 4058 4059 4060 4061 4062 4063 4064 4065 4066 4067 4068 4069 4070 4071 4072 4073 4074 4075 4076 4077 4078 4079 4080 4081 4082 4083 4084 4085 4086 4087 4088 4089 4090 4091 4092 4093 4094 4095 4096 4097 4098 4099 4100 4101 4102 4103 4104 4105 4106 4107 4108 4109 4110 4111 4112 4113 4114 4115 4116 4117 4118 4119 4120 4121 4122 4123 4124 4125 4126 4127 4128 4129 4130 4131 4132 4133 4134 4135 4136 4137 4138 4139 4140 4141 4142 4143 4144 4145 4146 4147 4148 4149 4150 4151 4152 4153 4154 4155 4156 4157 4158 4159 4160 4161 4162 4163 4164 4165 4166 4167 4168 4169 4170 4171 4172 4173 4174 4175 4176 4177 4178 4179 4180 4181 4182 4183 4184 4185 4186 4187 4188 4189 4190 4191 4192 4193 4194 4195 4196 4197 4198 4199 4200 4201 4202 4203 4204 4205 4206 4207 4208 4209 4210 4211 4212 4213 4214 4215 4216 4217 4218 4219 4220 4221 4222 4223 4224 4225 4226 4227 4228 4229 4230 4231 4232 4233 4234 4235 4236 4237 4238 4239 4240 4241 4242 4243 4244 4245 4246 4247 4248 4249 4250 4251 4252 4253 4254 4255 4256 4257 4258 4259 4260 4261 4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291 4292 4293 4294 4295 4296 4297 4298 4299 4300 4301 4302 4303 4304 4305 4306 4307 4308 4309 4310 4311 4312 4313 4314 4315 4316 4317 4318 4319 4320 4321 4322 4323 4324 4325 4326 4327 4328 4329 4330 4331 4332 4333 4334 4335 4336 4337 4338 4339 4340 4341 4342 4343 4344 4345 4346 4347 4348 4349 4350 4351 4352 4353 4354 4355 4356 4357 4358 4359 4360 4361 4362 4363 4364 4365 4366 4367 4368 4369 4370 4371 4372 4373 4374 4375 4376 4377 4378 4379 4380 4381 4382 4383 4384 4385 4386 4387 4388 4389 4390 4391 4392 4393 4394 4395 4396 4397 4398 4399 4400 4401 4402 4403 4404 4405 4406 4407 4408 4409 4410 4411 4412 4413 4414 4415 4416 4417 4418 4419 4420 4421 4422 4423 4424 4425 4426 4427 4428 4429 4430 4431 4432 4433 4434 4435 4436 4437 4438 4439 4440 4441 4442 4443 4444 4445 4446 4447 4448 4449 4450 4451 4452 4453 4454 4455 4456 4457 4458 4459 4460 4461 4462 4463 4464 4465 4466 4467 4468 4469 4470 4471 4472 4473 4474 4475 4476 4477 4478 4479 4480 4481 4482 4483 4484 4485 4486 4487 4488 4489 4490 4491 4492 4493 4494 4495 4496 4497 4498 4499 4500 4501 4502 4503 4504 4505 4506 4507 4508 4509 4510 4511 4512 4513 4514 4515 4516 4517 4518 4519 4520 4521 4522 4523 4524 4525 4526 4527 4528 4529 4530 4531 4532 4533 4534 4535 4536 4537 4538 4539 4540 4541 4542 4543 4544 4545 4546 4547 4548 4549 4550 4551 4552 4553 4554 4555 4556 4557 4558 4559 4560 4561 4562 4563 4564 4565 4566 4567 4568 4569 4570 4571 4572 4573 4574 4575 4576 4577 4578 4579 4580 4581 4582 4583 4584 4585 4586 4587 4588 4589 4590 4591 4592 4593 4594 4595 4596 4597 4598 4599 4600 4601 4602 4603 4604 4605 4606 4607 4608 4609 4610 4611 4612 4613 4614 4615 4616 4617 4618 4619 4620 4621 4622 4623 4624 4625 4626 4627 4628 4629 4630 4631 4632 4633 4634 4635 4636 4637 4638 4639 4640 4641 4642 4643 4644 4645 4646 4647 4648 4649 4650 4651 4652 4653 4654 4655 4656 4657 4658 4659 4660 4661 4662 4663 4664 4665 4666 4667 4668 4669 4670 4671 4672 4673 4674 4675 4676 4677 4678 4679 4680 4681 4682 4683 4684 4685 4686 4687 4688 4689 4690 4691 4692 4693 4694 4695 4696 4697 4698 4699 4700 4701 4702 4703 4704 4705 4706 4707 4708 4709 4710 4711 4712 4713 4714 4715 4716 4717 4718 4719 4720 4721 4722 4723 4724 4725 4726 4727 4728 4729 4730 4731 4732 4733 4734 4735 4736 4737 4738 4739 4740 4741 4742 4743 4744 4745 4746 4747 4748 4749 4750 4751 4752 4753 4754 4755 4756 4757 4758 4759 4760 4761 4762 4763 4764 4765 4766 4767 4768 4769 4770 4771 4772 4773 4774 4775 4776 4777 4778 4779 4780 4781 4782 4783 4784 4785 4786 4787 4788 4789 4790 4791 4792 4793 4794 4795 4796 4797 4798 4799 4800 4801 4802 4803 4804 4805 4806 4807 4808 4809 4810 4811 4812 4813 4814 4815 4816 4817 4818 4819 4820 4821 4822 4823 4824 4825 4826 4827 4828 4829 4830 4831 4832 4833 4834 4835 4836 4837 4838 4839 4840 4841 4842 4843 4844 4845 4846 4847 4848 4849 4850 4851 4852 4853 4854 4855 4856 4857 4858 4859 4860 4861 4862 4863 4864 4865 4866 4867 4868 4869 4870 4871 4872 4873 4874 4875 4876 4877 4878 4879 4880 4881 4882 4883 4884 4885 4886 4887 4888 4889 4890 4891 4892 4893 4894 4895 4896 4897 4898 4899 4900 4901 4902 4903 4904 4905 4906 4907 4908 4909 4910 4911 4912 4913 4914 4915 4916 4917 4918 4919 4920 4921 4922 4923 4924 4925 4926 4927 4928 4929 4930 4931 4932 4933 4934 4935 4936 4937 4938 4939 4940 4941 4942 4943 4944 4945 4946 4947 4948 4949 4950 4951 4952 4953 4954 4955 4956 4957 4958 4959 4960 4961 4962 4963 4964 4965 4966 4967 4968 4969 4970 4971 4972 4973 4974 4975 4976 4977 4978 4979 4980 4981 4982 4983 4984 4985 4986 4987 4988 4989 4990 4991 4992 4993 4994 4995 4996 4997 4998 4999 5000 5001 5002 5003 5004 5005 5006 5007 5008 5009 5010 5011 5012 5013 5014 5015 5016 5017 5018 5019 5020 5021 5022 5023 5024 5025 5026 5027 5028 5029 5030 5031 5032 5033 5034 5035 5036 5037 5038 5039 5040 5041 5042 5043 5044 5045 5046 5047 5048 5049 5050 5051 5052 5053 5054 5055 5056 5057 5058 5059 5060 5061 5062 5063 5064 5065 5066 5067 5068 5069 5070 5071 5072 5073 5074 5075 5076 5077 5078 5079 5080 5081 5082 5083 5084 5085 5086 5087 5088 5089 5090 5091 5092 5093 5094 5095 5096 5097 5098 5099 5100 5101 5102 5103 5104 5105 5106 5107 5108 5109 5110 5111 5112 5113 5114 5115 5116 5117 5118 5119 5120 5121 5122 5123 5124 5125 5126 5127 5128 5129 5130 5131 5132 5133 5134 5135 5136 5137 5138 5139 5140 5141 5142 5143 5144 5145 5146 5147 5148 5149 5150 5151 5152 5153 5154 5155 5156 5157 5158 5159 5160 5161 5162 5163 5164 5165 5166 5167 5168 5169 5170 5171 5172 5173 5174 5175 5176 5177 5178 5179 5180 5181 5182 5183 5184 5185 5186 5187 5188 5189 5190 5191 5192 5193 5194 5195 5196 5197 5198 5199 5200 5201 5202 5203 5204 5205 5206 5207 5208 5209 5210 5211 5212 5213 5214 5215 5216 5217 5218 5219 5220 5221 5222 5223 5224 5225 5226 5227 5228 5229 5230 5231 5232 5233 5234 5235 5236 5237 5238 5239 5240 5241 5242 5243 5244 5245 5246 5247 5248 5249 5250 5251 5252 5253 5254 5255 5256 5257 5258 5259 5260 5261 5262 5263 5264 5265 5266 5267 5268 5269 5270 5271 5272 5273 5274 5275 5276 5277 5278 5279 5280 5281 5282 5283 5284 5285 5286 5287 5288 5289 5290 5291 5292 5293 5294 5295 5296 5297 5298 5299 5300 5301 5302 5303 5304 5305 5306 5307 5308 5309 5310 5311 5312 5313 5314 5315 5316 5317 5318 5319 5320 5321 5322 5323 5324 5325 5326 5327 5328 5329 5330 5331 5332 5333 5334 5335 5336 5337 5338 5339 5340 5341 5342 5343 5344 5345 5346 5347 5348 5349 5350 5351 5352 5353 5354 5355 5356 5357 5358 5359 5360 5361 5362 5363 5364 5365 5366 5367 5368 5369 5370 5371 5372 5373 5374 5375 5376 5377 5378 5379 5380 5381 5382 5383 5384 5385 5386 5387 5388 5389 5390 5391 5392 5393 5394 5395 5396 5397 5398 5399 5400 5401 5402 5403 5404 5405 5406 5407 5408 5409 5410 5411 5412 5413 5414 5415 5416 5417 5418 5419 5420 5421 5422 5423 5424 5425 5426 5427 5428 5429 5430 5431 5432 5433 5434 5435 5436 5437 5438 5439 5440 5441 5442 5443 5444 5445 5446 5447 5448 5449 5450 5451 5452 5453 5454 5455 5456 5457 5458 5459 5460 5461 5462 5463 5464 5465 5466 5467 5468 5469 5470 5471 5472 5473 5474 5475 5476 5477 5478 5479 5480 5481 5482 5483 5484 5485 5486 5487 5488 5489 5490 5491 5492 5493 5494 5495 5496 5497 5498 5499 5500 5501 5502 5503 5504 5505 5506 5507 5508 5509 5510 5511 5512 5513 5514 5515 5516 5517 5518 5519 5520 5521 5522 5523 5524 5525 5526 5527 5528 5529 5530 5531 5532 5533 5534 5535 5536 5537 5538 5539 5540 5541 5542 5543 5544 5545 5546 5547 5548 5549 5550 5551 5552 5553 5554 5555 5556 5557 5558 5559 5560 5561 5562 5563 5564 5565 5566 5567 5568 5569 5570 5571 5572 5573 5574 5575 5576 5577 5578 5579 5580 5581 5582 5583 5584 5585 5586 5587 5588 5589 5590 5591 5592 5593 5594 5595 5596 5597 5598 5599 5600 5601 5602 5603 5604 5605 5606 5607 5608 5609 5610 5611 5612 5613 5614 5615 5616 5617 5618 5619 5620 5621 5622 5623 5624 5625 5626 5627 5628 5629 5630 5631 5632 5633 5634 5635 5636 5637 5638 5639 5640 5641 5642 5643 5644 5645 5646 5647 5648 5649 5650 5651 5652 5653 5654 5655 5656 5657 5658 5659 5660 5661 5662 5663 5664 5665 5666 5667 5668 5669 5670 5671 5672 5673 5674 5675 5676 5677 5678 5679 5680 5681 5682 5683 5684 5685 5686 5687 5688 5689 5690 5691 5692 5693 5694 5695 5696 5697 5698 5699 5700 5701 5702 5703 5704 5705 5706 5707 5708 5709 5710 5711 5712 5713 5714 5715 5716 5717 5718 5719 5720 5721 5722 5723 5724 5725 5726 5727 5728 5729 5730 5731 5732 5733 5734 5735 5736 5737 5738 5739 5740 5741 5742 5743 5744 5745 5746 5747 5748 5749 5750 5751 5752 5753 5754 5755 5756 5757 5758 5759 5760 5761 5762 5763 5764 5765 5766 5767 5768 5769 5770 5771 5772 5773 5774 5775 5776 5777 5778 5779 5780 5781 5782 5783 5784 5785 5786 5787 5788 5789 5790 5791 5792 5793 5794 5795 5796 5797 5798 5799 5800 5801 5802 5803 5804 5805 5806 5807 5808 5809 5810 5811 5812 5813 5814 5815 5816 5817 5818 5819 5820 5821 5822 5823 5824 5825 5826 5827 5828 5829 5830 5831 5832 5833 5834 5835 5836 5837 5838 5839 5840 5841 5842 5843 5844 5845 5846 5847 5848 5849 5850 5851
;*****************************************************************************
;* pixel.asm: x86 pixel metrics
;*****************************************************************************
;* Copyright (C) 2003-2024 x264 project
;*
;* Authors: Loren Merritt <lorenm@u.washington.edu>
;*          Holger Lubitz <holger@lubitz.org>
;*          Laurent Aimar <fenrir@via.ecp.fr>
;*          Alex Izvorski <aizvorksi@gmail.com>
;*          Fiona Glaser <fiona@x264.com>
;*          Oskar Arvidsson <oskar@irock.se>
;*
;* This program is free software; you can redistribute it and/or modify
;* it under the terms of the GNU General Public License as published by
;* the Free Software Foundation; either version 2 of the License, or
;* (at your option) any later version.
;*
;* This program is distributed in the hope that it will be useful,
;* but WITHOUT ANY WARRANTY; without even the implied warranty of
;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
;* GNU General Public License for more details.
;*
;* You should have received a copy of the GNU General Public License
;* along with this program; if not, write to the Free Software
;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02111, USA.
;*
;* This program is also available under a commercial proprietary license.
;* For more information, contact us at licensing@x264.com.
;*****************************************************************************

%include "x86inc.asm"
%include "x86util.asm"

SECTION_RODATA 32
var_shuf_avx512: db 0,-1, 1,-1, 2,-1, 3,-1, 4,-1, 5,-1, 6,-1, 7,-1
                 db 8,-1, 9,-1,10,-1,11,-1,12,-1,13,-1,14,-1,15,-1
hmul_16p:  times 16 db 1
           times 8 db 1, -1
hmul_8p:   times 8 db 1
           times 4 db 1, -1
           times 8 db 1
           times 4 db 1, -1
mask_ff:   times 16 db 0xff
           times 16 db 0
mask_ac4:  times 2 dw 0, -1, -1, -1, 0, -1, -1, -1
mask_ac4b: times 2 dw 0, -1, 0, -1, -1, -1, -1, -1
mask_ac8:  times 2 dw 0, -1, -1, -1, -1, -1, -1, -1
%if HIGH_BIT_DEPTH
ssd_nv12_shuf: db 0, 1, 4, 5, 2, 3, 6, 7, 8, 9, 12, 13, 10, 11, 14, 15
%endif
%if BIT_DEPTH == 10
ssim_c1:   times 4 dd 6697.7856    ; .01*.01*1023*1023*64
ssim_c2:   times 4 dd 3797644.4352 ; .03*.03*1023*1023*64*63
pf_64:     times 4 dd 64.0
pf_128:    times 4 dd 128.0
%elif BIT_DEPTH == 9
ssim_c1:   times 4 dd 1671         ; .01*.01*511*511*64
ssim_c2:   times 4 dd 947556       ; .03*.03*511*511*64*63
%else ; 8-bit
ssim_c1:   times 4 dd 416          ; .01*.01*255*255*64
ssim_c2:   times 4 dd 235963       ; .03*.03*255*255*64*63
%endif
hmul_4p:   times 2 db 1, 1, 1, 1, 1, -1, 1, -1
mask_10:   times 4 dw 0, -1
mask_1100: times 2 dd 0, -1
pb_pppm:   times 4 db 1,1,1,-1
deinterleave_shuf: db 0, 2, 4, 6, 8, 10, 12, 14, 1, 3, 5, 7, 9, 11, 13, 15
intrax3_shuf: db 7,6,7,6,5,4,5,4,3,2,3,2,1,0,1,0

intrax9a_ddlr1: db  6, 7, 8, 9, 7, 8, 9,10, 4, 5, 6, 7, 3, 4, 5, 6
intrax9a_ddlr2: db  8, 9,10,11, 9,10,11,12, 2, 3, 4, 5, 1, 2, 3, 4
intrax9a_hdu1:  db 15, 4, 5, 6,14, 3,15, 4,14, 2,13, 1,13, 1,12, 0
intrax9a_hdu2:  db 13, 2,14, 3,12, 1,13, 2,12, 0,11,11,11,11,11,11
intrax9a_vrl1:  db 10,11,12,13, 3, 4, 5, 6,11,12,13,14, 5, 6, 7, 8
intrax9a_vrl2:  db  2,10,11,12, 1, 3, 4, 5,12,13,14,15, 6, 7, 8, 9
intrax9a_vh1:   db  6, 7, 8, 9, 6, 7, 8, 9, 4, 4, 4, 4, 3, 3, 3, 3
intrax9a_vh2:   db  6, 7, 8, 9, 6, 7, 8, 9, 2, 2, 2, 2, 1, 1, 1, 1
intrax9a_dc:    db  1, 2, 3, 4, 6, 7, 8, 9,-1,-1,-1,-1,-1,-1,-1,-1
intrax9a_lut:   db 0x60,0x68,0x80,0x00,0x08,0x20,0x40,0x28,0x48,0,0,0,0,0,0,0
pw_s01234567:   dw 0x8000,0x8001,0x8002,0x8003,0x8004,0x8005,0x8006,0x8007
pw_s01234657:   dw 0x8000,0x8001,0x8002,0x8003,0x8004,0x8006,0x8005,0x8007
intrax9_edge:   db  0, 0, 1, 2, 3, 7, 8, 9,10,11,12,13,14,15,15,15

intrax9b_ddlr1: db  6, 7, 8, 9, 4, 5, 6, 7, 7, 8, 9,10, 3, 4, 5, 6
intrax9b_ddlr2: db  8, 9,10,11, 2, 3, 4, 5, 9,10,11,12, 1, 2, 3, 4
intrax9b_hdu1:  db 15, 4, 5, 6,14, 2,13, 1,14, 3,15, 4,13, 1,12, 0
intrax9b_hdu2:  db 13, 2,14, 3,12, 0,11,11,12, 1,13, 2,11,11,11,11
intrax9b_vrl1:  db 10,11,12,13,11,12,13,14, 3, 4, 5, 6, 5, 6, 7, 8
intrax9b_vrl2:  db  2,10,11,12,12,13,14,15, 1, 3, 4, 5, 6, 7, 8, 9
intrax9b_vh1:   db  6, 7, 8, 9, 4, 4, 4, 4, 6, 7, 8, 9, 3, 3, 3, 3
intrax9b_vh2:   db  6, 7, 8, 9, 2, 2, 2, 2, 6, 7, 8, 9, 1, 1, 1, 1
intrax9b_edge2: db  6, 7, 8, 9, 6, 7, 8, 9, 4, 3, 2, 1, 4, 3, 2, 1
intrax9b_v1:    db  0, 1,-1,-1,-1,-1,-1,-1, 4, 5,-1,-1,-1,-1,-1,-1
intrax9b_v2:    db  2, 3,-1,-1,-1,-1,-1,-1, 6, 7,-1,-1,-1,-1,-1,-1
intrax9b_lut:   db 0x60,0x64,0x80,0x00,0x04,0x20,0x40,0x24,0x44,0,0,0,0,0,0,0

ALIGN 32
intra8x9_h1:   db  7, 7, 7, 7, 7, 7, 7, 7, 5, 5, 5, 5, 5, 5, 5, 5
intra8x9_h2:   db  6, 6, 6, 6, 6, 6, 6, 6, 4, 4, 4, 4, 4, 4, 4, 4
intra8x9_h3:   db  3, 3, 3, 3, 3, 3, 3, 3, 1, 1, 1, 1, 1, 1, 1, 1
intra8x9_h4:   db  2, 2, 2, 2, 2, 2, 2, 2, 0, 0, 0, 0, 0, 0, 0, 0
intra8x9_ddl1: db  1, 2, 3, 4, 5, 6, 7, 8, 3, 4, 5, 6, 7, 8, 9,10
intra8x9_ddl2: db  2, 3, 4, 5, 6, 7, 8, 9, 4, 5, 6, 7, 8, 9,10,11
intra8x9_ddl3: db  5, 6, 7, 8, 9,10,11,12, 7, 8, 9,10,11,12,13,14
intra8x9_ddl4: db  6, 7, 8, 9,10,11,12,13, 8, 9,10,11,12,13,14,15
intra8x9_vl1:  db  0, 1, 2, 3, 4, 5, 6, 7, 1, 2, 3, 4, 5, 6, 7, 8
intra8x9_vl2:  db  1, 2, 3, 4, 5, 6, 7, 8, 2, 3, 4, 5, 6, 7, 8, 9
intra8x9_vl3:  db  2, 3, 4, 5, 6, 7, 8, 9, 3, 4, 5, 6, 7, 8, 9,10
intra8x9_vl4:  db  3, 4, 5, 6, 7, 8, 9,10, 4, 5, 6, 7, 8, 9,10,11
intra8x9_ddr1: db  8, 9,10,11,12,13,14,15, 6, 7, 8, 9,10,11,12,13
intra8x9_ddr2: db  7, 8, 9,10,11,12,13,14, 5, 6, 7, 8, 9,10,11,12
intra8x9_ddr3: db  4, 5, 6, 7, 8, 9,10,11, 2, 3, 4, 5, 6, 7, 8, 9
intra8x9_ddr4: db  3, 4, 5, 6, 7, 8, 9,10, 1, 2, 3, 4, 5, 6, 7, 8
intra8x9_vr1:  db  8, 9,10,11,12,13,14,15, 7, 8, 9,10,11,12,13,14
intra8x9_vr2:  db  8, 9,10,11,12,13,14,15, 6, 8, 9,10,11,12,13,14
intra8x9_vr3:  db  5, 7, 8, 9,10,11,12,13, 3, 5, 7, 8, 9,10,11,12
intra8x9_vr4:  db  4, 6, 8, 9,10,11,12,13, 2, 4, 6, 8, 9,10,11,12
intra8x9_hd1:  db  3, 8, 9,10,11,12,13,14, 1, 6, 2, 7, 3, 8, 9,10
intra8x9_hd2:  db  2, 7, 3, 8, 9,10,11,12, 0, 5, 1, 6, 2, 7, 3, 8
intra8x9_hd3:  db  7, 8, 9,10,11,12,13,14, 3, 4, 5, 6, 7, 8, 9,10
intra8x9_hd4:  db  5, 6, 7, 8, 9,10,11,12, 1, 2, 3, 4, 5, 6, 7, 8
intra8x9_hu1:  db 13,12,11,10, 9, 8, 7, 6, 9, 8, 7, 6, 5, 4, 3, 2
intra8x9_hu2:  db 11,10, 9, 8, 7, 6, 5, 4, 7, 6, 5, 4, 3, 2, 1, 0
intra8x9_hu3:  db  5, 4, 3, 2, 1, 0,15,15, 1, 0,15,15,15,15,15,15
intra8x9_hu4:  db  3, 2, 1, 0,15,15,15,15,15,15,15,15,15,15,15,15
pw_s00112233:  dw 0x8000,0x8000,0x8001,0x8001,0x8002,0x8002,0x8003,0x8003
pw_s00001111:  dw 0x8000,0x8000,0x8000,0x8000,0x8001,0x8001,0x8001,0x8001

transd_shuf1: SHUFFLE_MASK_W 0, 8, 2, 10, 4, 12, 6, 14
transd_shuf2: SHUFFLE_MASK_W 1, 9, 3, 11, 5, 13, 7, 15

sw_f0:     dq 0xfff0, 0
pd_f0:     times 4 dd 0xffff0000
pd_2:      times 4 dd 2

pw_76543210: dw 0, 1, 2, 3, 4, 5, 6, 7

ads_mvs_shuffle:
%macro ADS_MVS_SHUFFLE 8
    %assign y x
    %rep 8
        %rep 7
            %rotate (~y)&1
            %assign y y>>((~y)&1)
        %endrep
        db %1*2, %1*2+1
        %rotate 1
        %assign y y>>1
    %endrep
%endmacro
%assign x 0
%rep 256
    ADS_MVS_SHUFFLE 0, 1, 2, 3, 4, 5, 6, 7
%assign x x+1
%endrep

SECTION .text

cextern pb_0
cextern pb_1
cextern pw_1
cextern pw_8
cextern pw_16
cextern pw_32
cextern pw_00ff
cextern pw_ppppmmmm
cextern pw_ppmmppmm
cextern pw_pmpmpmpm
cextern pw_pmmpzzzz
cextern pd_1
cextern hsub_mul
cextern popcnt_table

;=============================================================================
; SSD
;=============================================================================

%if HIGH_BIT_DEPTH
;-----------------------------------------------------------------------------
; int pixel_ssd_WxH( uint16_t *, intptr_t, uint16_t *, intptr_t )
;-----------------------------------------------------------------------------
%macro SSD_ONE 2
cglobal pixel_ssd_%1x%2, 4,7,6
    FIX_STRIDES r1, r3
%if mmsize == %1*2
    %define offset0_1 r1
    %define offset0_2 r1*2
    %define offset0_3 r5
    %define offset1_1 r3
    %define offset1_2 r3*2
    %define offset1_3 r6
    lea     r5, [3*r1]
    lea     r6, [3*r3]
%elif mmsize == %1
    %define offset0_1 mmsize
    %define offset0_2 r1
    %define offset0_3 r1+mmsize
    %define offset1_1 mmsize
    %define offset1_2 r3
    %define offset1_3 r3+mmsize
%elif mmsize == %1/2
    %define offset0_1 mmsize
    %define offset0_2 mmsize*2
    %define offset0_3 mmsize*3
    %define offset1_1 mmsize
    %define offset1_2 mmsize*2
    %define offset1_3 mmsize*3
%endif
    %assign %%n %2/(2*mmsize/%1)
%if %%n > 1
    mov    r4d, %%n
%endif
    pxor    m0, m0
.loop:
    mova    m1, [r0]
    mova    m2, [r0+offset0_1]
    mova    m3, [r0+offset0_2]
    mova    m4, [r0+offset0_3]
    psubw   m1, [r2]
    psubw   m2, [r2+offset1_1]
    psubw   m3, [r2+offset1_2]
    psubw   m4, [r2+offset1_3]
%if %%n > 1
    lea     r0, [r0+r1*(%2/%%n)]
    lea     r2, [r2+r3*(%2/%%n)]
%endif
    pmaddwd m1, m1
    pmaddwd m2, m2
    pmaddwd m3, m3
    pmaddwd m4, m4
    paddd   m1, m2
    paddd   m3, m4
    paddd   m0, m1
    paddd   m0, m3
%if %%n > 1
    dec    r4d
    jg .loop
%endif
    HADDD   m0, m5
    movd   eax, xm0
    RET
%endmacro

INIT_MMX mmx2
SSD_ONE     4,  4
SSD_ONE     4,  8
SSD_ONE     4, 16
SSD_ONE     8,  4
SSD_ONE     8,  8
SSD_ONE     8, 16
SSD_ONE    16,  8
SSD_ONE    16, 16
INIT_XMM sse2
SSD_ONE     8,  4
SSD_ONE     8,  8
SSD_ONE     8, 16
SSD_ONE    16,  8
SSD_ONE    16, 16
INIT_YMM avx2
SSD_ONE    16,  8
SSD_ONE    16, 16
%endif ; HIGH_BIT_DEPTH

%if HIGH_BIT_DEPTH == 0
%macro SSD_LOAD_FULL 5
    mova      m1, [t0+%1]
    mova      m2, [t2+%2]
    mova      m3, [t0+%3]
    mova      m4, [t2+%4]
%if %5==1
    add       t0, t1
    add       t2, t3
%elif %5==2
    lea       t0, [t0+2*t1]
    lea       t2, [t2+2*t3]
%endif
%endmacro

%macro LOAD 5
    movh      m%1, %3
    movh      m%2, %4
%if %5
    lea       t0, [t0+2*t1]
%endif
%endmacro

%macro JOIN 7
    movh      m%3, %5
    movh      m%4, %6
%if %7
    lea       t2, [t2+2*t3]
%endif
    punpcklbw m%1, m7
    punpcklbw m%3, m7
    psubw     m%1, m%3
    punpcklbw m%2, m7
    punpcklbw m%4, m7
    psubw     m%2, m%4
%endmacro

%macro JOIN_SSE2 7
    movh      m%3, %5
    movh      m%4, %6
%if %7
    lea       t2, [t2+2*t3]
%endif
    punpcklqdq m%1, m%2
    punpcklqdq m%3, m%4
    DEINTB %2, %1, %4, %3, 7
    psubw m%2, m%4
    psubw m%1, m%3
%endmacro

%macro JOIN_SSSE3 7
    movh      m%3, %5
    movh      m%4, %6
%if %7
    lea       t2, [t2+2*t3]
%endif
    punpcklbw m%1, m%3
    punpcklbw m%2, m%4
%endmacro

%macro LOAD_AVX2 5
    mova     xm%1, %3
    vinserti128 m%1, m%1, %4, 1
%if %5
    lea       t0, [t0+2*t1]
%endif
%endmacro

%macro JOIN_AVX2 7
    mova     xm%2, %5
    vinserti128 m%2, m%2, %6, 1
%if %7
    lea       t2, [t2+2*t3]
%endif
    SBUTTERFLY bw, %1, %2, %3
%endmacro

%macro SSD_LOAD_HALF 5
    LOAD      1, 2, [t0+%1], [t0+%3], 1
    JOIN      1, 2, 3, 4, [t2+%2], [t2+%4], 1
    LOAD      3, 4, [t0+%1], [t0+%3], %5
    JOIN      3, 4, 5, 6, [t2+%2], [t2+%4], %5
%endmacro

%macro SSD_CORE 7-8
%ifidn %8, FULL
    mova      m%6, m%2
    mova      m%7, m%4
    psubusb   m%2, m%1
    psubusb   m%4, m%3
    psubusb   m%1, m%6
    psubusb   m%3, m%7
    por       m%1, m%2
    por       m%3, m%4
    punpcklbw m%2, m%1, m%5
    punpckhbw m%1, m%5
    punpcklbw m%4, m%3, m%5
    punpckhbw m%3, m%5
%endif
    pmaddwd   m%1, m%1
    pmaddwd   m%2, m%2
    pmaddwd   m%3, m%3
    pmaddwd   m%4, m%4
%endmacro

%macro SSD_CORE_SSE2 7-8
%ifidn %8, FULL
    DEINTB %6, %1, %7, %2, %5
    psubw m%6, m%7
    psubw m%1, m%2
    SWAP %6, %2, %1
    DEINTB %6, %3, %7, %4, %5
    psubw m%6, m%7
    psubw m%3, m%4
    SWAP %6, %4, %3
%endif
    pmaddwd   m%1, m%1
    pmaddwd   m%2, m%2
    pmaddwd   m%3, m%3
    pmaddwd   m%4, m%4
%endmacro

%macro SSD_CORE_SSSE3 7-8
%ifidn %8, FULL
    punpckhbw m%6, m%1, m%2
    punpckhbw m%7, m%3, m%4
    punpcklbw m%1, m%2
    punpcklbw m%3, m%4
    SWAP %6, %2, %3
    SWAP %7, %4
%endif
    pmaddubsw m%1, m%5
    pmaddubsw m%2, m%5
    pmaddubsw m%3, m%5
    pmaddubsw m%4, m%5
    pmaddwd   m%1, m%1
    pmaddwd   m%2, m%2
    pmaddwd   m%3, m%3
    pmaddwd   m%4, m%4
%endmacro

%macro SSD_ITER 6
    SSD_LOAD_%1 %2,%3,%4,%5,%6
    SSD_CORE  1, 2, 3, 4, 7, 5, 6, %1
    paddd     m1, m2
    paddd     m3, m4
    paddd     m0, m1
    paddd     m0, m3
%endmacro

;-----------------------------------------------------------------------------
; int pixel_ssd_16x16( uint8_t *, intptr_t, uint8_t *, intptr_t )
;-----------------------------------------------------------------------------
%macro SSD 2
%if %1 != %2
    %assign function_align 8
%else
    %assign function_align 16
%endif
cglobal pixel_ssd_%1x%2, 0,0,0
    mov     al, %1*%2/mmsize/2

%if %1 != %2
    jmp mangle(private_prefix %+ _pixel_ssd_%1x%1 %+ SUFFIX %+ .startloop)
%else

.startloop:
%if ARCH_X86_64
    DECLARE_REG_TMP 0,1,2,3
    PROLOGUE 0,0,8
%else
    PROLOGUE 0,5
    DECLARE_REG_TMP 1,2,3,4
    mov t0, r0m
    mov t1, r1m
    mov t2, r2m
    mov t3, r3m
%endif

%if cpuflag(ssse3)
    mova    m7, [hsub_mul]
%elifidn cpuname, sse2
    mova    m7, [pw_00ff]
%elif %1 >= mmsize
    pxor    m7, m7
%endif
    pxor    m0, m0

ALIGN 16
.loop:
%if %1 > mmsize
    SSD_ITER FULL, 0, 0, mmsize, mmsize, 1
%elif %1 == mmsize
    SSD_ITER FULL, 0, 0, t1, t3, 2
%else
    SSD_ITER HALF, 0, 0, t1, t3, 2
%endif
    dec     al
    jg .loop
%if mmsize==32
    vextracti128 xm1, m0, 1
    paddd  xm0, xm1
    HADDD  xm0, xm1
    movd   eax, xm0
%else
    HADDD   m0, m1
    movd   eax, m0
%endif
    RET
%endif
%endmacro

INIT_MMX mmx
SSD 16, 16
SSD 16,  8
SSD  8,  8
SSD  8, 16
SSD  4,  4
SSD  8,  4
SSD  4,  8
SSD  4, 16
INIT_XMM sse2slow
SSD 16, 16
SSD  8,  8
SSD 16,  8
SSD  8, 16
SSD  8,  4
INIT_XMM sse2
%define SSD_CORE SSD_CORE_SSE2
%define JOIN JOIN_SSE2
SSD 16, 16
SSD  8,  8
SSD 16,  8
SSD  8, 16
SSD  8,  4
INIT_XMM ssse3
%define SSD_CORE SSD_CORE_SSSE3
%define JOIN JOIN_SSSE3
SSD 16, 16
SSD  8,  8
SSD 16,  8
SSD  8, 16
SSD  8,  4
INIT_XMM avx
SSD 16, 16
SSD  8,  8
SSD 16,  8
SSD  8, 16
SSD  8,  4
INIT_MMX ssse3
SSD  4,  4
SSD  4,  8
SSD  4, 16
INIT_XMM xop
SSD 16, 16
SSD  8,  8
SSD 16,  8
SSD  8, 16
SSD  8,  4
%define LOAD LOAD_AVX2
%define JOIN JOIN_AVX2
INIT_YMM avx2
SSD 16, 16
SSD 16,  8
%assign function_align 16
%endif ; !HIGH_BIT_DEPTH

;-----------------------------------------------------------------------------
; void pixel_ssd_nv12_core( uint16_t *pixuv1, intptr_t stride1, uint16_t *pixuv2, intptr_t stride2,
;                           int width, int height, uint64_t *ssd_u, uint64_t *ssd_v )
;
; The maximum width this function can handle without risk of overflow is given
; in the following equation: (mmsize in bits)
;
;   2 * mmsize/32 * (2^32 - 1) / (2^BIT_DEPTH - 1)^2
;
; For 10-bit XMM this means width >= 32832. At sane distortion levels
; it will take much more than that though.
;-----------------------------------------------------------------------------
%if HIGH_BIT_DEPTH
%macro SSD_NV12 0
cglobal pixel_ssd_nv12_core, 6,7,7
    shl        r4d, 2
    FIX_STRIDES r1, r3
    add         r0, r4
    add         r2, r4
    neg         r4
    pxor        m4, m4
    pxor        m5, m5
%if mmsize == 32
    vbroadcasti128 m6, [ssd_nv12_shuf]
%endif
.loopy:
    mov         r6, r4
    pxor        m2, m2
    pxor        m3, m3
.loopx:
    mova        m0, [r0+r6]
    mova        m1, [r0+r6+mmsize]
    psubw       m0, [r2+r6]
    psubw       m1, [r2+r6+mmsize]
%if mmsize == 32
    pshufb      m0, m6
    pshufb      m1, m6
%else
    SBUTTERFLY wd, 0, 1, 6
%endif
%if cpuflag(xop)
    pmadcswd    m2, m0, m0, m2
    pmadcswd    m3, m1, m1, m3
%else
    pmaddwd     m0, m0
    pmaddwd     m1, m1
    paddd       m2, m0
    paddd       m3, m1
%endif
    add         r6, 2*mmsize
    jl .loopx
%if mmsize == 32 ; avx2 may overread by 32 bytes, that has to be handled
    jz .no_overread
    psubd       m3, m1
.no_overread:
%endif
    punpckhdq   m0, m2, m5 ; using HADDD would remove the mmsize/32 part from the
    punpckhdq   m1, m3, m5 ; equation above, putting the width limit at 8208
    punpckldq   m2, m5
    punpckldq   m3, m5
    paddq       m0, m1
    paddq       m2, m3
    paddq       m4, m0
    paddq       m4, m2
    add         r0, r1
    add         r2, r3
    dec        r5d
    jg .loopy
    mov         r0, r6m
    mov         r1, r7m
%if mmsize == 32
    vextracti128 xm0, m4, 1
    paddq      xm4, xm0
%endif
    movq      [r0], xm4
    movhps    [r1], xm4
    RET
%endmacro ; SSD_NV12

%else ; !HIGH_BIT_DEPTH
;-----------------------------------------------------------------------------
; void pixel_ssd_nv12_core( uint8_t *pixuv1, intptr_t stride1, uint8_t *pixuv2, intptr_t stride2,
;                           int width, int height, uint64_t *ssd_u, uint64_t *ssd_v )
;
; This implementation can potentially overflow on image widths >= 11008 (or
; 6604 if interlaced), since it is called on blocks of height up to 12 (resp
; 20). At sane distortion levels it will take much more than that though.
;-----------------------------------------------------------------------------
%macro SSD_NV12 0
cglobal pixel_ssd_nv12_core, 6,7
    add    r4d, r4d
    add     r0, r4
    add     r2, r4
    neg     r4
    pxor    m3, m3
    pxor    m4, m4
    mova    m5, [pw_00ff]
.loopy:
    mov     r6, r4
.loopx:
%if mmsize == 32 ; only 16-byte alignment is guaranteed
    movu    m2, [r0+r6]
    movu    m1, [r2+r6]
%else
    mova    m2, [r0+r6]
    mova    m1, [r2+r6]
%endif
    psubusb m0, m2, m1
    psubusb m1, m2
    por     m0, m1
    psrlw   m2, m0, 8
    pand    m0, m5
%if cpuflag(xop)
    pmadcswd m4, m2, m2, m4
    pmadcswd m3, m0, m0, m3
%else
    pmaddwd m2, m2
    pmaddwd m0, m0
    paddd   m4, m2
    paddd   m3, m0
%endif
    add     r6, mmsize
    jl .loopx
%if mmsize == 32 ; avx2 may overread by 16 bytes, that has to be handled
    jz .no_overread
    pcmpeqb xm1, xm1
    pandn   m0, m1, m0 ; zero the lower half
    pandn   m2, m1, m2
    psubd   m3, m0
    psubd   m4, m2
.no_overread:
%endif
    add     r0, r1
    add     r2, r3
    dec    r5d
    jg .loopy
    mov     r0, r6m
    mov     r1, r7m
%if cpuflag(ssse3)
    phaddd  m3, m4
%else
    SBUTTERFLY qdq, 3, 4, 0
    paddd   m3, m4
%endif
%if mmsize == 32
    vextracti128 xm4, m3, 1
    paddd  xm3, xm4
%endif
    psllq  xm4, xm3, 32
    paddd  xm3, xm4
    psrlq  xm3, 32
    movq  [r0], xm3
    movhps [r1], xm3
    RET
%endmacro ; SSD_NV12
%endif ; !HIGH_BIT_DEPTH

INIT_XMM sse2
SSD_NV12
INIT_XMM avx
SSD_NV12
INIT_XMM xop
SSD_NV12
INIT_YMM avx2
SSD_NV12

;=============================================================================
; variance
;=============================================================================

%macro VAR_START 1
    pxor  m5, m5    ; sum
    pxor  m6, m6    ; sum squared
%if HIGH_BIT_DEPTH == 0
%if %1
    mova  m7, [pw_00ff]
%elif mmsize == 16
    pxor  m7, m7    ; zero
%endif
%endif ; !HIGH_BIT_DEPTH
%endmacro

%macro VAR_END 0
    pmaddwd       m5, [pw_1]
    SBUTTERFLY    dq, 5, 6, 0
    paddd         m5, m6
%if mmsize == 32
    vextracti128 xm6, m5, 1
    paddd        xm5, xm6
%endif
    MOVHL        xm6, xm5
    paddd        xm5, xm6
%if ARCH_X86_64
    movq         rax, xm5
%else
    movd         eax, xm5
%if cpuflag(avx)
    pextrd       edx, xm5, 1
%else
    pshuflw      xm5, xm5, q1032
    movd         edx, xm5
%endif
%endif
    RET
%endmacro

%macro VAR_CORE 0
    paddw     m5, m0
    paddw     m5, m3
    paddw     m5, m1
    paddw     m5, m4
    pmaddwd   m0, m0
    pmaddwd   m3, m3
    pmaddwd   m1, m1
    pmaddwd   m4, m4
    paddd     m6, m0
    paddd     m6, m3
    paddd     m6, m1
    paddd     m6, m4
%endmacro

;-----------------------------------------------------------------------------
; int pixel_var_wxh( uint8_t *, intptr_t )
;-----------------------------------------------------------------------------
%if HIGH_BIT_DEPTH
%macro VAR 0
cglobal pixel_var_16x16, 2,3,8
    FIX_STRIDES r1
    VAR_START 0
    mov      r2d, 8
.loop:
    mova      m0, [r0]
    mova      m1, [r0+mmsize]
    mova      m3, [r0+r1]
    mova      m4, [r0+r1+mmsize]
    lea       r0, [r0+r1*2]
    VAR_CORE
    dec      r2d
    jg .loop
    VAR_END

cglobal pixel_var_8x8, 2,3,8
    lea       r2, [r1*3]
    VAR_START 0
    mova      m0, [r0]
    mova      m1, [r0+r1*2]
    mova      m3, [r0+r1*4]
    mova      m4, [r0+r2*2]
    lea       r0, [r0+r1*8]
    VAR_CORE
    mova      m0, [r0]
    mova      m1, [r0+r1*2]
    mova      m3, [r0+r1*4]
    mova      m4, [r0+r2*2]
    VAR_CORE
    VAR_END
%endmacro ; VAR

INIT_XMM sse2
VAR
INIT_XMM avx
VAR

%else ; HIGH_BIT_DEPTH == 0

%macro VAR 0
cglobal pixel_var_16x16, 2,3,8
    VAR_START 1
    mov      r2d, 8
.loop:
    mova      m0, [r0]
    mova      m3, [r0+r1]
    DEINTB    1, 0, 4, 3, 7
    lea       r0, [r0+r1*2]
    VAR_CORE
    dec r2d
    jg .loop
    VAR_END

cglobal pixel_var_8x8, 2,4,8
    VAR_START 1
    mov      r2d, 2
    lea       r3, [r1*3]
.loop:
    movh      m0, [r0]
    movh      m3, [r0+r1]
    movhps    m0, [r0+r1*2]
    movhps    m3, [r0+r3]
    DEINTB    1, 0, 4, 3, 7
    lea       r0, [r0+r1*4]
    VAR_CORE
    dec r2d
    jg .loop
    VAR_END

cglobal pixel_var_8x16, 2,4,8
    VAR_START 1
    mov      r2d, 4
    lea       r3, [r1*3]
.loop:
    movh      m0, [r0]
    movh      m3, [r0+r1]
    movhps    m0, [r0+r1*2]
    movhps    m3, [r0+r3]
    DEINTB    1, 0, 4, 3, 7
    lea       r0, [r0+r1*4]
    VAR_CORE
    dec r2d
    jg .loop
    VAR_END
%endmacro ; VAR

INIT_XMM sse2
VAR
INIT_XMM avx
VAR
%endif ; !HIGH_BIT_DEPTH

INIT_YMM avx2
cglobal pixel_var_16x16, 2,4,7
    FIX_STRIDES r1
    VAR_START 0
    mov      r2d, 4
    lea       r3, [r1*3]
.loop:
%if HIGH_BIT_DEPTH
    mova      m0, [r0]
    mova      m3, [r0+r1]
    mova      m1, [r0+r1*2]
    mova      m4, [r0+r3]
%else
    pmovzxbw  m0, [r0]
    pmovzxbw  m3, [r0+r1]
    pmovzxbw  m1, [r0+r1*2]
    pmovzxbw  m4, [r0+r3]
%endif
    lea       r0, [r0+r1*4]
    VAR_CORE
    dec r2d
    jg .loop
    VAR_END

%macro VAR_AVX512_CORE 1 ; accum
%if %1
    paddw    m0, m2
    pmaddwd  m2, m2
    paddw    m0, m3
    pmaddwd  m3, m3
    paddd    m1, m2
    paddd    m1, m3
%else
    paddw    m0, m2, m3
    pmaddwd  m2, m2
    pmaddwd  m3, m3
    paddd    m1, m2, m3
%endif
%endmacro

%macro VAR_AVX512_CORE_16x16 1 ; accum
%if HIGH_BIT_DEPTH
    mova            ym2, [r0]
    vinserti64x4     m2, [r0+r1], 1
    mova            ym3, [r0+2*r1]
    vinserti64x4     m3, [r0+r3], 1
%else
    vbroadcasti64x2 ym2, [r0]
    vbroadcasti64x2  m2 {k1}, [r0+r1]
    vbroadcasti64x2 ym3, [r0+2*r1]
    vbroadcasti64x2  m3 {k1}, [r0+r3]
    pshufb           m2, m4
    pshufb           m3, m4
%endif
    VAR_AVX512_CORE %1
%endmacro

%macro VAR_AVX512_CORE_8x8 1 ; accum
%if HIGH_BIT_DEPTH
    mova            xm2, [r0]
    mova            xm3, [r0+r1]
%else
    movq            xm2, [r0]
    movq            xm3, [r0+r1]
%endif
    vinserti128     ym2, [r0+2*r1], 1
    vinserti128     ym3, [r0+r2], 1
    lea              r0, [r0+4*r1]
    vinserti32x4     m2, [r0], 2
    vinserti32x4     m3, [r0+r1], 2
    vinserti32x4     m2, [r0+2*r1], 3
    vinserti32x4     m3, [r0+r2], 3
%if HIGH_BIT_DEPTH == 0
    punpcklbw        m2, m4
    punpcklbw        m3, m4
%endif
    VAR_AVX512_CORE %1
%endmacro

INIT_ZMM avx512
cglobal pixel_var_16x16, 2,4
    FIX_STRIDES     r1
    mov            r2d, 0xf0
    lea             r3, [3*r1]
%if HIGH_BIT_DEPTH == 0
    vbroadcasti64x4 m4, [var_shuf_avx512]
    kmovb           k1, r2d
%endif
    VAR_AVX512_CORE_16x16 0
.loop:
    lea             r0, [r0+4*r1]
    VAR_AVX512_CORE_16x16 1
    sub            r2d, 0x50
    jg .loop
%if ARCH_X86_64 == 0
    pop            r3d
    %assign regs_used 3
%endif
var_avx512_end:
    vbroadcasti32x4 m2, [pw_1]
    pmaddwd         m0, m2
    SBUTTERFLY      dq, 0, 1, 2
    paddd           m0, m1
    vextracti32x8  ym1, m0, 1
    paddd          ym0, ym1
    vextracti128   xm1, ym0, 1
    paddd         xmm0, xm0, xm1
    punpckhqdq    xmm1, xmm0, xmm0
    paddd         xmm0, xmm1
%if ARCH_X86_64
    movq           rax, xmm0
%else
    movd           eax, xmm0
    pextrd         edx, xmm0, 1
%endif
    RET

%if HIGH_BIT_DEPTH == 0 ; 8x8 doesn't benefit from AVX-512 in high bit-depth
cglobal pixel_var_8x8, 2,3
    lea             r2, [3*r1]
    pxor           xm4, xm4
    VAR_AVX512_CORE_8x8 0
    jmp var_avx512_end
%endif

cglobal pixel_var_8x16, 2,3
    FIX_STRIDES     r1
    lea             r2, [3*r1]
%if HIGH_BIT_DEPTH == 0
    pxor           xm4, xm4
%endif
    VAR_AVX512_CORE_8x8 0
    lea             r0, [r0+4*r1]
    VAR_AVX512_CORE_8x8 1
    jmp var_avx512_end

;-----------------------------------------------------------------------------
; int pixel_var2_8x8( pixel *fenc, pixel *fdec, int ssd[2] )
;-----------------------------------------------------------------------------

%if ARCH_X86_64
    DECLARE_REG_TMP 6
%else
    DECLARE_REG_TMP 2
%endif

%macro VAR2_END 3 ; src, tmp, shift
    movifnidn r2, r2mp
    pshufd    %2, %1, q3331
    pmuludq   %1, %1
    movq    [r2], %2 ; sqr_u sqr_v
    psrld     %1, %3
    psubd     %2, %1 ; sqr - (sum * sum >> shift)
    MOVHL     %1, %2
    paddd     %1, %2
    movd     eax, %1
    RET
%endmacro

%macro VAR2_8x8_SSE2 2
%if HIGH_BIT_DEPTH
cglobal pixel_var2_8x%1, 2,3,6
    pxor       m4, m4
    pxor       m5, m5
%define %%sum2 m4
%define %%sqr2 m5
%else
cglobal pixel_var2_8x%1, 2,3,7
    mova       m6, [pw_00ff]
%define %%sum2 m0
%define %%sqr2 m1
%endif
    pxor       m0, m0 ; sum
    pxor       m1, m1 ; sqr
    mov       t0d, (%1-1)*FENC_STRIDEB
.loop:
%if HIGH_BIT_DEPTH
    mova       m2, [r0+1*t0]
    psubw      m2, [r1+2*t0]
    mova       m3, [r0+1*t0+16]
    psubw      m3, [r1+2*t0+32]
%else
    mova       m3, [r0+1*t0]
    movq       m5, [r1+2*t0]
    punpcklqdq m5, [r1+2*t0+16]
    DEINTB      2, 3, 4, 5, 6
    psubw      m2, m4
    psubw      m3, m5
%endif
    paddw      m0, m2
    pmaddwd    m2, m2
    paddw  %%sum2, m3
    pmaddwd    m3, m3
    paddd      m1, m2
    paddd  %%sqr2, m3
    sub       t0d, FENC_STRIDEB
    jge .loop
%if HIGH_BIT_DEPTH
    SBUTTERFLY dq, 0, 4, 2
    paddw      m0, m4 ; sum_u sum_v
    pmaddwd    m0, [pw_1]
    SBUTTERFLY dq, 1, 5, 2
    paddd      m1, m5 ; sqr_u sqr_v
    SBUTTERFLY dq, 0, 1, 2
    paddd      m0, m1
%else
    pmaddwd    m0, [pw_1]
    shufps     m2, m0, m1, q2020
    shufps     m0, m1, q3131
    paddd      m0, m2
    pshufd     m0, m0, q3120 ; sum_u sqr_u sum_v sqr_v
%endif
    VAR2_END   m0, m1, %2
%endmacro

INIT_XMM sse2
VAR2_8x8_SSE2  8, 6
VAR2_8x8_SSE2 16, 7

%macro VAR2_CORE 3 ; src1, src2, accum
%if %3
    paddw    m0, %1
    pmaddwd  %1, %1
    paddw    m0, %2
    pmaddwd  %2, %2
    paddd    m1, %1
    paddd    m1, %2
%else
    paddw    m0, %1, %2
    pmaddwd  %1, %1
    pmaddwd  %2, %2
    paddd    m1, %1, %2
%endif
%endmacro

%if HIGH_BIT_DEPTH == 0
INIT_XMM ssse3
cglobal pixel_var2_internal
    pxor        m0, m0 ; sum
    pxor        m1, m1 ; sqr
.loop:
    movq        m2, [r0+1*t0]
    punpcklbw   m2, [r1+2*t0]
    movq        m3, [r0+1*t0-1*FENC_STRIDE]
    punpcklbw   m3, [r1+2*t0-1*FDEC_STRIDE]
    movq        m4, [r0+1*t0-2*FENC_STRIDE]
    punpcklbw   m4, [r1+2*t0-2*FDEC_STRIDE]
    movq        m5, [r0+1*t0-3*FENC_STRIDE]
    punpcklbw   m5, [r1+2*t0-3*FDEC_STRIDE]
    pmaddubsw   m2, m7
    pmaddubsw   m3, m7
    pmaddubsw   m4, m7
    pmaddubsw   m5, m7
    VAR2_CORE   m2, m3, 1
    VAR2_CORE   m4, m5, 1
    sub        t0d, 4*FENC_STRIDE
    jg .loop
    pmaddwd     m0, [pw_1]
    ret

%macro VAR2_8x8_SSSE3 2
cglobal pixel_var2_8x%1, 2,3,8
    mova        m7, [hsub_mul]
    mov        t0d, (%1-1)*FENC_STRIDE
    call pixel_var2_internal_ssse3 ; u
    add         r0, 8
    add         r1, 16
    SBUTTERFLY qdq, 0, 1, 6
    paddd       m1, m0
    mov        t0d, (%1-1)*FENC_STRIDE
    call pixel_var2_internal_ssse3 ; v
    SBUTTERFLY qdq, 0, 6, 2
    paddd       m0, m6
    phaddd      m1, m0 ; sum_u sqr_u sum_v sqr_v
    VAR2_END    m1, m0, %2
%endmacro

VAR2_8x8_SSSE3  8, 6
VAR2_8x8_SSSE3 16, 7
%endif ; !HIGH_BIT_DEPTH

%macro VAR2_AVX2_LOAD 3 ; offset_reg, row1_offset, row2_offset
%if HIGH_BIT_DEPTH
%if mmsize == 64
    mova        m2, [r1+2*%1+%2*FDEC_STRIDEB]
    vshufi32x4  m2, [r1+2*%1+%2*FDEC_STRIDEB+64], q2020
    mova        m3, [r1+2*%1+%3*FDEC_STRIDEB]
    vshufi32x4  m3, [r1+2*%1+%3*FDEC_STRIDEB+64], q2020
%else
    mova       xm2, [r1+2*%1+%2*FDEC_STRIDEB]
    vinserti128 m2, [r1+2*%1+%2*FDEC_STRIDEB+32], 1
    mova       xm3, [r1+2*%1+%3*FDEC_STRIDEB]
    vinserti128 m3, [r1+2*%1+%3*FDEC_STRIDEB+32], 1
%endif
    psubw       m2, [r0+1*%1+%2*FENC_STRIDEB]
    psubw       m3, [r0+1*%1+%3*FENC_STRIDEB]
%else
    pmovzxbw    m2, [r0+1*%1+%2*FENC_STRIDE]
    mova        m4, [r1+2*%1+%2*FDEC_STRIDE]
    pmovzxbw    m3, [r0+1*%1+%3*FENC_STRIDE]
    mova        m5, [r1+2*%1+%3*FDEC_STRIDE]
    punpcklbw   m4, m6
    punpcklbw   m5, m6
    psubw       m2, m4
    psubw       m3, m5
%endif
%endmacro

%macro VAR2_8x8_AVX2 2
%if HIGH_BIT_DEPTH
cglobal pixel_var2_8x%1, 2,3,4
%else
cglobal pixel_var2_8x%1, 2,3,7
    pxor           m6, m6
%endif
    mov           t0d, (%1-3)*FENC_STRIDEB
    VAR2_AVX2_LOAD t0, 2, 1
    VAR2_CORE      m2, m3, 0
.loop:
    VAR2_AVX2_LOAD t0, 0, -1
    VAR2_CORE      m2, m3, 1
    sub           t0d, 2*FENC_STRIDEB
    jg .loop

    pmaddwd        m0, [pw_1]
    SBUTTERFLY    qdq, 0, 1, 2
    paddd          m0, m1
    vextracti128  xm1, m0, 1
    phaddd        xm0, xm1
    VAR2_END      xm0, xm1, %2
%endmacro

INIT_YMM avx2
VAR2_8x8_AVX2  8, 6
VAR2_8x8_AVX2 16, 7

%macro VAR2_AVX512_END 1 ; shift
    vbroadcasti32x4 m2, [pw_1]
    pmaddwd         m0, m2
    SBUTTERFLY     qdq, 0, 1, 2
    paddd           m0, m1
    vextracti32x8  ym1, m0, 1
    paddd          ym0, ym1
    psrlq          ym1, ym0, 32
    paddd          ym0, ym1
    vpmovqd       xmm0, ym0 ; sum_u, sqr_u, sum_v, sqr_v
    VAR2_END      xmm0, xmm1, %1
%endmacro

INIT_ZMM avx512
cglobal pixel_var2_8x8, 2,3
%if HIGH_BIT_DEPTH == 0
    pxor          xm6, xm6
%endif
    VAR2_AVX2_LOAD  0, 0, 2
    VAR2_CORE      m2, m3, 0
    VAR2_AVX2_LOAD  0, 4, 6
    VAR2_CORE      m2, m3, 1
    VAR2_AVX512_END 6

cglobal pixel_var2_8x16, 2,3
%if HIGH_BIT_DEPTH == 0
    pxor          xm6, xm6
%endif
    mov           t0d, 10*FENC_STRIDEB
    VAR2_AVX2_LOAD  0, 14, 12
    VAR2_CORE      m2, m3, 0
.loop:
    VAR2_AVX2_LOAD t0, 0, -2
    VAR2_CORE      m2, m3, 1
    sub           t0d, 4*FENC_STRIDEB
    jg .loop
    VAR2_AVX512_END 7

;=============================================================================
; SATD
;=============================================================================

%macro JDUP 2
%if cpuflag(sse4)
    ; just use shufps on anything post conroe
    shufps %1, %2, 0
%elif cpuflag(ssse3) && notcpuflag(atom)
    ; join 2x 32 bit and duplicate them
    ; emulating shufps is faster on conroe
    punpcklqdq %1, %2
    movsldup %1, %1
%else
    ; doesn't need to dup. sse2 does things by zero extending to words and full h_2d
    punpckldq %1, %2
%endif
%endmacro

%macro HSUMSUB 5
    pmaddubsw m%2, m%5
    pmaddubsw m%1, m%5
    pmaddubsw m%4, m%5
    pmaddubsw m%3, m%5
%endmacro

%macro DIFF_UNPACK_SSE2 5
    punpcklbw m%1, m%5
    punpcklbw m%2, m%5
    punpcklbw m%3, m%5
    punpcklbw m%4, m%5
    psubw m%1, m%2
    psubw m%3, m%4
%endmacro

%macro DIFF_SUMSUB_SSSE3 5
    HSUMSUB %1, %2, %3, %4, %5
    psubw m%1, m%2
    psubw m%3, m%4
%endmacro

%macro LOAD_DUP_2x4P 4 ; dst, tmp, 2* pointer
    movd %1, %3
    movd %2, %4
    JDUP %1, %2
%endmacro

%macro LOAD_DUP_4x8P_CONROE 8 ; 4*dst, 4*pointer
    movddup m%3, %6
    movddup m%4, %8
    movddup m%1, %5
    movddup m%2, %7
%endmacro

%macro LOAD_DUP_4x8P_PENRYN 8
    ; penryn and nehalem run punpcklqdq and movddup in different units
    movh m%3, %6
    movh m%4, %8
    punpcklqdq m%3, m%3
    movddup m%1, %5
    punpcklqdq m%4, m%4
    movddup m%2, %7
%endmacro

%macro LOAD_SUMSUB_8x2P 9
    LOAD_DUP_4x8P %1, %2, %3, %4, %6, %7, %8, %9
    DIFF_SUMSUB_SSSE3 %1, %3, %2, %4, %5
%endmacro

%macro LOAD_SUMSUB_8x4P_SSSE3 7-11 r0, r2, 0, 0
; 4x dest, 2x tmp, 1x mul, [2* ptr], [increment?]
    LOAD_SUMSUB_8x2P %1, %2, %5, %6, %7, [%8], [%9], [%8+r1], [%9+r3]
    LOAD_SUMSUB_8x2P %3, %4, %5, %6, %7, [%8+2*r1], [%9+2*r3], [%8+r4], [%9+r5]
%if %10
    lea %8, [%8+4*r1]
    lea %9, [%9+4*r3]
%endif
%endmacro

%macro LOAD_SUMSUB_16P_SSSE3 7 ; 2*dst, 2*tmp, mul, 2*ptr
    movddup m%1, [%7]
    movddup m%2, [%7+8]
    mova m%4, [%6]
    movddup m%3, m%4
    punpckhqdq m%4, m%4
    DIFF_SUMSUB_SSSE3 %1, %3, %2, %4, %5
%endmacro

%macro LOAD_SUMSUB_16P_SSE2 7 ; 2*dst, 2*tmp, mask, 2*ptr
    movu  m%4, [%7]
    mova  m%2, [%6]
    DEINTB %1, %2, %3, %4, %5
    psubw m%1, m%3
    psubw m%2, m%4
    SUMSUB_BA w, %1, %2, %3
%endmacro

%macro LOAD_SUMSUB_16x4P 10-13 r0, r2, none
; 8x dest, 1x tmp, 1x mul, [2* ptr] [2nd tmp]
    LOAD_SUMSUB_16P %1, %5, %2, %3, %10, %11, %12
    LOAD_SUMSUB_16P %2, %6, %3, %4, %10, %11+r1, %12+r3
    LOAD_SUMSUB_16P %3, %7, %4, %9, %10, %11+2*r1, %12+2*r3
    LOAD_SUMSUB_16P %4, %8, %13, %9, %10, %11+r4, %12+r5
%endmacro

%macro LOAD_SUMSUB_16x2P_AVX2 9
; 2*dst, 2*tmp, mul, 4*ptr
    vbroadcasti128 m%1, [%6]
    vbroadcasti128 m%3, [%7]
    vbroadcasti128 m%2, [%8]
    vbroadcasti128 m%4, [%9]
    DIFF_SUMSUB_SSSE3 %1, %3, %2, %4, %5
%endmacro

%macro LOAD_SUMSUB_16x4P_AVX2 7-11 r0, r2, 0, 0
; 4x dest, 2x tmp, 1x mul, [2* ptr], [increment?]
    LOAD_SUMSUB_16x2P_AVX2 %1, %2, %5, %6, %7, %8, %9, %8+r1, %9+r3
    LOAD_SUMSUB_16x2P_AVX2 %3, %4, %5, %6, %7, %8+2*r1, %9+2*r3, %8+r4, %9+r5
%if %10
    lea  %8, [%8+4*r1]
    lea  %9, [%9+4*r3]
%endif
%endmacro

%macro LOAD_DUP_4x16P_AVX2 8 ; 4*dst, 4*pointer
    mova  xm%3, %6
    mova  xm%4, %8
    mova  xm%1, %5
    mova  xm%2, %7
    vpermq m%3, m%3, q0011
    vpermq m%4, m%4, q0011
    vpermq m%1, m%1, q0011
    vpermq m%2, m%2, q0011
%endmacro

%macro LOAD_SUMSUB8_16x2P_AVX2 9
; 2*dst, 2*tmp, mul, 4*ptr
    LOAD_DUP_4x16P_AVX2 %1, %2, %3, %4, %6, %7, %8, %9
    DIFF_SUMSUB_SSSE3 %1, %3, %2, %4, %5
%endmacro

%macro LOAD_SUMSUB8_16x4P_AVX2 7-11 r0, r2, 0, 0
; 4x dest, 2x tmp, 1x mul, [2* ptr], [increment?]
    LOAD_SUMSUB8_16x2P_AVX2 %1, %2, %5, %6, %7, [%8], [%9], [%8+r1], [%9+r3]
    LOAD_SUMSUB8_16x2P_AVX2 %3, %4, %5, %6, %7, [%8+2*r1], [%9+2*r3], [%8+r4], [%9+r5]
%if %10
    lea  %8, [%8+4*r1]
    lea  %9, [%9+4*r3]
%endif
%endmacro

; in: r4=3*stride1, r5=3*stride2
; in: %2 = horizontal offset
; in: %3 = whether we need to increment pix1 and pix2
; clobber: m3..m7
; out: %1 = satd
%macro SATD_4x4_MMX 3
    %xdefine %%n nn%1
    %assign offset %2*SIZEOF_PIXEL
    LOAD_DIFF m4, m3, none, [r0+     offset], [r2+     offset]
    LOAD_DIFF m5, m3, none, [r0+  r1+offset], [r2+  r3+offset]
    LOAD_DIFF m6, m3, none, [r0+2*r1+offset], [r2+2*r3+offset]
    LOAD_DIFF m7, m3, none, [r0+  r4+offset], [r2+  r5+offset]
%if %3
    lea  r0, [r0+4*r1]
    lea  r2, [r2+4*r3]
%endif
    HADAMARD4_2D 4, 5, 6, 7, 3, %%n
    paddw m4, m6
    SWAP %%n, 4
%endmacro

; in: %1 = horizontal if 0, vertical if 1
%macro SATD_8x4_SSE 8-9
%if %1
    HADAMARD4_2D_SSE %2, %3, %4, %5, %6, amax
%else
    HADAMARD4_V %2, %3, %4, %5, %6
    ; doing the abs first is a slight advantage
    ABSW2 m%2, m%4, m%2, m%4, m%6, m%7
    ABSW2 m%3, m%5, m%3, m%5, m%6, m%7
    HADAMARD 1, max, %2, %4, %6, %7
%endif
%ifnidn %9, swap
    paddw m%8, m%2
%else
    SWAP %8, %2
%endif
%if %1
    paddw m%8, m%4
%else
    HADAMARD 1, max, %3, %5, %6, %7
    paddw m%8, m%3
%endif
%endmacro

%macro SATD_START_MMX 0
    FIX_STRIDES r1, r3
    lea  r4, [3*r1] ; 3*stride1
    lea  r5, [3*r3] ; 3*stride2
%endmacro

%macro SATD_END_MMX 0
%if HIGH_BIT_DEPTH
    HADDUW      m0, m1
    movd       eax, m0
%else ; !HIGH_BIT_DEPTH
    pshufw      m1, m0, q1032
    paddw       m0, m1
    pshufw      m1, m0, q2301
    paddw       m0, m1
    movd       eax, m0
    and        eax, 0xffff
%endif ; HIGH_BIT_DEPTH
    RET
%endmacro

; FIXME avoid the spilling of regs to hold 3*stride.
; for small blocks on x86_32, modify pixel pointer instead.

;-----------------------------------------------------------------------------
; int pixel_satd_16x16( uint8_t *, intptr_t, uint8_t *, intptr_t )
;-----------------------------------------------------------------------------
INIT_MMX mmx2
cglobal pixel_satd_16x4_internal
    SATD_4x4_MMX m2,  0, 0
    SATD_4x4_MMX m1,  4, 0
    paddw        m0, m2
    SATD_4x4_MMX m2,  8, 0
    paddw        m0, m1
    SATD_4x4_MMX m1, 12, 0
    paddw        m0, m2
    paddw        m0, m1
    ret

cglobal pixel_satd_8x8_internal
    SATD_4x4_MMX m2,  0, 0
    SATD_4x4_MMX m1,  4, 1
    paddw        m0, m2
    paddw        m0, m1
pixel_satd_8x4_internal_mmx2:
    SATD_4x4_MMX m2,  0, 0
    SATD_4x4_MMX m1,  4, 0
    paddw        m0, m2
    paddw        m0, m1
    ret

%if HIGH_BIT_DEPTH
%macro SATD_MxN_MMX 3
cglobal pixel_satd_%1x%2, 4,7
    SATD_START_MMX
    pxor   m0, m0
    call pixel_satd_%1x%3_internal_mmx2
    HADDUW m0, m1
    movd  r6d, m0
%rep %2/%3-1
    pxor   m0, m0
    lea    r0, [r0+4*r1]
    lea    r2, [r2+4*r3]
    call pixel_satd_%1x%3_internal_mmx2
    movd   m2, r4
    HADDUW m0, m1
    movd   r4, m0
    add    r6, r4
    movd   r4, m2
%endrep
    movifnidn eax, r6d
    RET
%endmacro

SATD_MxN_MMX 16, 16, 4
SATD_MxN_MMX 16,  8, 4
SATD_MxN_MMX  8, 16, 8
%endif ; HIGH_BIT_DEPTH

%if HIGH_BIT_DEPTH == 0
cglobal pixel_satd_16x16, 4,6
    SATD_START_MMX
    pxor   m0, m0
%rep 3
    call pixel_satd_16x4_internal_mmx2
    lea  r0, [r0+4*r1]
    lea  r2, [r2+4*r3]
%endrep
    call pixel_satd_16x4_internal_mmx2
    HADDUW m0, m1
    movd  eax, m0
    RET

cglobal pixel_satd_16x8, 4,6
    SATD_START_MMX
    pxor   m0, m0
    call pixel_satd_16x4_internal_mmx2
    lea  r0, [r0+4*r1]
    lea  r2, [r2+4*r3]
    call pixel_satd_16x4_internal_mmx2
    SATD_END_MMX

cglobal pixel_satd_8x16, 4,6
    SATD_START_MMX
    pxor   m0, m0
    call pixel_satd_8x8_internal_mmx2
    lea  r0, [r0+4*r1]
    lea  r2, [r2+4*r3]
    call pixel_satd_8x8_internal_mmx2
    SATD_END_MMX
%endif ; !HIGH_BIT_DEPTH

cglobal pixel_satd_8x8, 4,6
    SATD_START_MMX
    pxor   m0, m0
    call pixel_satd_8x8_internal_mmx2
    SATD_END_MMX

cglobal pixel_satd_8x4, 4,6
    SATD_START_MMX
    pxor   m0, m0
    call pixel_satd_8x4_internal_mmx2
    SATD_END_MMX

cglobal pixel_satd_4x16, 4,6
    SATD_START_MMX
    SATD_4x4_MMX m0, 0, 1
    SATD_4x4_MMX m1, 0, 1
    paddw  m0, m1
    SATD_4x4_MMX m1, 0, 1
    paddw  m0, m1
    SATD_4x4_MMX m1, 0, 0
    paddw  m0, m1
    SATD_END_MMX

cglobal pixel_satd_4x8, 4,6
    SATD_START_MMX
    SATD_4x4_MMX m0, 0, 1
    SATD_4x4_MMX m1, 0, 0
    paddw  m0, m1
    SATD_END_MMX

cglobal pixel_satd_4x4, 4,6
    SATD_START_MMX
    SATD_4x4_MMX m0, 0, 0
    SATD_END_MMX

%macro SATD_START_SSE2 2-3 0
    FIX_STRIDES r1, r3
%if HIGH_BIT_DEPTH && %3
    pxor    %2, %2
%elif cpuflag(ssse3) && notcpuflag(atom)
%if mmsize==32
    mova    %2, [hmul_16p]
%else
    mova    %2, [hmul_8p]
%endif
%endif
    lea     r4, [3*r1]
    lea     r5, [3*r3]
    pxor    %1, %1
%endmacro

%macro SATD_END_SSE2 1-2
%if HIGH_BIT_DEPTH
    HADDUW  %1, xm0
%if %0 == 2
    paddd   %1, %2
%endif
%else
    HADDW   %1, xm7
%endif
    movd   eax, %1
    RET
%endmacro

%macro SATD_ACCUM 3
%if HIGH_BIT_DEPTH
    HADDUW %1, %2
    paddd  %3, %1
    pxor   %1, %1
%endif
%endmacro

%macro BACKUP_POINTERS 0
%if ARCH_X86_64
%if WIN64
    PUSH r7
%endif
    mov     r6, r0
    mov     r7, r2
%endif
%endmacro

%macro RESTORE_AND_INC_POINTERS 0
%if ARCH_X86_64
    lea     r0, [r6+8*SIZEOF_PIXEL]
    lea     r2, [r7+8*SIZEOF_PIXEL]
%if WIN64
    POP r7
%endif
%else
    mov     r0, r0mp
    mov     r2, r2mp
    add     r0, 8*SIZEOF_PIXEL
    add     r2, 8*SIZEOF_PIXEL
%endif
%endmacro

%macro SATD_4x8_SSE 3
%if HIGH_BIT_DEPTH
    movh    m0, [r0+0*r1]
    movh    m4, [r2+0*r3]
    movh    m1, [r0+1*r1]
    movh    m5, [r2+1*r3]
    movhps  m0, [r0+4*r1]
    movhps  m4, [r2+4*r3]
    movh    m2, [r0+2*r1]
    movh    m6, [r2+2*r3]
    psubw   m0, m4
    movh    m3, [r0+r4]
    movh    m4, [r2+r5]
    lea     r0, [r0+4*r1]
    lea     r2, [r2+4*r3]
    movhps  m1, [r0+1*r1]
    movhps  m5, [r2+1*r3]
    movhps  m2, [r0+2*r1]
    movhps  m6, [r2+2*r3]
    psubw   m1, m5
    movhps  m3, [r0+r4]
    movhps  m4, [r2+r5]
    psubw   m2, m6
    psubw   m3, m4
%else ; !HIGH_BIT_DEPTH
    movd m4, [r2]
    movd m5, [r2+r3]
    movd m6, [r2+2*r3]
    add r2, r5
    movd m0, [r0]
    movd m1, [r0+r1]
    movd m2, [r0+2*r1]
    add r0, r4
    movd m3, [r2+r3]
    JDUP m4, m3
    movd m3, [r0+r1]
    JDUP m0, m3
    movd m3, [r2+2*r3]
    JDUP m5, m3
    movd m3, [r0+2*r1]
    JDUP m1, m3
%if %1==0 && %2==1
    mova m3, [hmul_4p]
    DIFFOP 0, 4, 1, 5, 3
%else
    DIFFOP 0, 4, 1, 5, 7
%endif
    movd m5, [r2]
    add r2, r5
    movd m3, [r0]
    add r0, r4
    movd m4, [r2]
    JDUP m6, m4
    movd m4, [r0]
    JDUP m2, m4
    movd m4, [r2+r3]
    JDUP m5, m4
    movd m4, [r0+r1]
    JDUP m3, m4
%if %1==0 && %2==1
    mova m4, [hmul_4p]
    DIFFOP 2, 6, 3, 5, 4
%else
    DIFFOP 2, 6, 3, 5, 7
%endif
%endif ; HIGH_BIT_DEPTH
    SATD_8x4_SSE %1, 0, 1, 2, 3, 4, 5, 7, %3
%endmacro

;-----------------------------------------------------------------------------
; int pixel_satd_8x4( uint8_t *, intptr_t, uint8_t *, intptr_t )
;-----------------------------------------------------------------------------
%macro SATDS_SSE2 0
%define vertical ((notcpuflag(ssse3) || cpuflag(atom)) || HIGH_BIT_DEPTH)

%if cpuflag(ssse3) && (vertical==0 || HIGH_BIT_DEPTH)
cglobal pixel_satd_4x4, 4, 6, 6
    SATD_START_MMX
    mova m4, [hmul_4p]
    LOAD_DUP_2x4P m2, m5, [r2], [r2+r3]
    LOAD_DUP_2x4P m3, m5, [r2+2*r3], [r2+r5]
    LOAD_DUP_2x4P m0, m5, [r0], [r0+r1]
    LOAD_DUP_2x4P m1, m5, [r0+2*r1], [r0+r4]
    DIFF_SUMSUB_SSSE3 0, 2, 1, 3, 4
    HADAMARD 0, sumsub, 0, 1, 2, 3
    HADAMARD 4, sumsub, 0, 1, 2, 3
    HADAMARD 1, amax, 0, 1, 2, 3
    HADDW m0, m1
    movd eax, m0
    RET
%endif

cglobal pixel_satd_4x8, 4, 6, 8
    SATD_START_MMX
%if vertical==0
    mova m7, [hmul_4p]
%endif
    SATD_4x8_SSE vertical, 0, swap
    HADDW m7, m1
    movd eax, m7
    RET

cglobal pixel_satd_4x16, 4, 6, 8
    SATD_START_MMX
%if vertical==0
    mova m7, [hmul_4p]
%endif
    SATD_4x8_SSE vertical, 0, swap
    lea r0, [r0+r1*2*SIZEOF_PIXEL]
    lea r2, [r2+r3*2*SIZEOF_PIXEL]
    SATD_4x8_SSE vertical, 1, add
    HADDW m7, m1
    movd eax, m7
    RET

cglobal pixel_satd_8x8_internal
    LOAD_SUMSUB_8x4P 0, 1, 2, 3, 4, 5, 7, r0, r2, 1, 0
    SATD_8x4_SSE vertical, 0, 1, 2, 3, 4, 5, 6
%%pixel_satd_8x4_internal:
    LOAD_SUMSUB_8x4P 0, 1, 2, 3, 4, 5, 7, r0, r2, 1, 0
    SATD_8x4_SSE vertical, 0, 1, 2, 3, 4, 5, 6
    ret

; 16x8 regresses on phenom win64, 16x16 is almost the same (too many spilled registers)
; These aren't any faster on AVX systems with fast movddup (Bulldozer, Sandy Bridge)
%if HIGH_BIT_DEPTH == 0 && UNIX64 && notcpuflag(avx)
cglobal pixel_satd_16x4_internal
    LOAD_SUMSUB_16x4P 0, 1, 2, 3, 4, 8, 5, 9, 6, 7, r0, r2, 11
    lea  r2, [r2+4*r3]
    lea  r0, [r0+4*r1]
    ; always use horizontal mode here
    SATD_8x4_SSE 0, 0, 1, 2, 3, 6, 11, 10
    SATD_8x4_SSE 0, 4, 8, 5, 9, 6, 3, 10
    ret

cglobal pixel_satd_16x8, 4,6,12
    SATD_START_SSE2 m10, m7
%if vertical
    mova m7, [pw_00ff]
%endif
    jmp %%pixel_satd_16x8_internal

cglobal pixel_satd_16x16, 4,6,12
    SATD_START_SSE2 m10, m7
%if vertical
    mova m7, [pw_00ff]
%endif
    call pixel_satd_16x4_internal
    call pixel_satd_16x4_internal
%%pixel_satd_16x8_internal:
    call pixel_satd_16x4_internal
    call pixel_satd_16x4_internal
    SATD_END_SSE2 m10
%else
cglobal pixel_satd_16x8, 4,6,8
    SATD_START_SSE2 m6, m7
    BACKUP_POINTERS
    call pixel_satd_8x8_internal
    RESTORE_AND_INC_POINTERS
    call pixel_satd_8x8_internal
    SATD_END_SSE2 m6

cglobal pixel_satd_16x16, 4,6,8
    SATD_START_SSE2 m6, m7, 1
    BACKUP_POINTERS
    call pixel_satd_8x8_internal
    call pixel_satd_8x8_internal
    SATD_ACCUM m6, m0, m7
    RESTORE_AND_INC_POINTERS
    call pixel_satd_8x8_internal
    call pixel_satd_8x8_internal
    SATD_END_SSE2 m6, m7
%endif

cglobal pixel_satd_8x16, 4,6,8
    SATD_START_SSE2 m6, m7
    call pixel_satd_8x8_internal
    call pixel_satd_8x8_internal
    SATD_END_SSE2 m6

cglobal pixel_satd_8x8, 4,6,8
    SATD_START_SSE2 m6, m7
    call pixel_satd_8x8_internal
    SATD_END_SSE2 m6

cglobal pixel_satd_8x4, 4,6,8
    SATD_START_SSE2 m6, m7
    call %%pixel_satd_8x4_internal
    SATD_END_SSE2 m6
%endmacro ; SATDS_SSE2

%macro SA8D_INTER 0
%if ARCH_X86_64
    %define lh m10
    %define rh m0
%else
    %define lh m0
    %define rh [esp+48]
%endif
%if HIGH_BIT_DEPTH
    HADDUW  m0, m1
    paddd   lh, rh
%else
    paddusw lh, rh
%endif ; HIGH_BIT_DEPTH
%endmacro

%macro SA8D 0
; sse2 doesn't seem to like the horizontal way of doing things
%define vertical ((notcpuflag(ssse3) || cpuflag(atom)) || HIGH_BIT_DEPTH)

%if ARCH_X86_64
;-----------------------------------------------------------------------------
; int pixel_sa8d_8x8( uint8_t *, intptr_t, uint8_t *, intptr_t )
;-----------------------------------------------------------------------------
cglobal pixel_sa8d_8x8_internal
    lea  r6, [r0+4*r1]
    lea  r7, [r2+4*r3]
    LOAD_SUMSUB_8x4P 0, 1, 2, 8, 5, 6, 7, r0, r2
    LOAD_SUMSUB_8x4P 4, 5, 3, 9, 11, 6, 7, r6, r7
%if vertical
    HADAMARD8_2D 0, 1, 2, 8, 4, 5, 3, 9, 6, amax
%else ; non-sse2
    HADAMARD8_2D_HMUL 0, 1, 2, 8, 4, 5, 3, 9, 6, 11
%endif
    paddw m0, m1
    paddw m0, m2
    paddw m0, m8
    SAVE_MM_PERMUTATION
    ret

cglobal pixel_sa8d_8x8, 4,8,12
    FIX_STRIDES r1, r3
    lea  r4, [3*r1]
    lea  r5, [3*r3]
%if vertical == 0
    mova m7, [hmul_8p]
%endif
    call pixel_sa8d_8x8_internal
%if HIGH_BIT_DEPTH
    HADDUW m0, m1
%else
    HADDW m0, m1
%endif ; HIGH_BIT_DEPTH
    movd eax, m0
    add eax, 1
    shr eax, 1
    RET

cglobal pixel_sa8d_16x16, 4,8,12
    FIX_STRIDES r1, r3
    lea  r4, [3*r1]
    lea  r5, [3*r3]
%if vertical == 0
    mova m7, [hmul_8p]
%endif
    call pixel_sa8d_8x8_internal ; pix[0]
    add  r2, 8*SIZEOF_PIXEL
    add  r0, 8*SIZEOF_PIXEL
%if HIGH_BIT_DEPTH
    HADDUW m0, m1
%endif
    mova m10, m0
    call pixel_sa8d_8x8_internal ; pix[8]
    lea  r2, [r2+8*r3]
    lea  r0, [r0+8*r1]
    SA8D_INTER
    call pixel_sa8d_8x8_internal ; pix[8*stride+8]
    sub  r2, 8*SIZEOF_PIXEL
    sub  r0, 8*SIZEOF_PIXEL
    SA8D_INTER
    call pixel_sa8d_8x8_internal ; pix[8*stride]
    SA8D_INTER
    SWAP 0, 10
%if HIGH_BIT_DEPTH == 0
    HADDUW m0, m1
%endif
    movd eax, m0
    add  eax, 1
    shr  eax, 1
    RET

%else ; ARCH_X86_32
%if mmsize == 16
cglobal pixel_sa8d_8x8_internal
    %define spill0 [esp+4]
    %define spill1 [esp+20]
    %define spill2 [esp+36]
%if vertical
    LOAD_DIFF_8x4P 0, 1, 2, 3, 4, 5, 6, r0, r2, 1
    HADAMARD4_2D 0, 1, 2, 3, 4
    movdqa spill0, m3
    LOAD_DIFF_8x4P 4, 5, 6, 7, 3, 3, 2, r0, r2, 1
    HADAMARD4_2D 4, 5, 6, 7, 3
    HADAMARD2_2D 0, 4, 1, 5, 3, qdq, amax
    movdqa m3, spill0
    paddw m0, m1
    HADAMARD2_2D 2, 6, 3, 7, 5, qdq, amax
%else ; mmsize == 8
    mova m7, [hmul_8p]
    LOAD_SUMSUB_8x4P 0, 1, 2, 3, 5, 6, 7, r0, r2, 1
    ; could do first HADAMARD4_V here to save spilling later
    ; surprisingly, not a win on conroe or even p4
    mova spill0, m2
    mova spill1, m3
    mova spill2, m1
    SWAP 1, 7
    LOAD_SUMSUB_8x4P 4, 5, 6, 7, 2, 3, 1, r0, r2, 1
    HADAMARD4_V 4, 5, 6, 7, 3
    mova m1, spill2
    mova m2, spill0
    mova m3, spill1
    mova spill0, m6
    mova spill1, m7
    HADAMARD4_V 0, 1, 2, 3, 7
    SUMSUB_BADC w, 0, 4, 1, 5, 7
    HADAMARD 2, sumsub, 0, 4, 7, 6
    HADAMARD 2, sumsub, 1, 5, 7, 6
    HADAMARD 1, amax, 0, 4, 7, 6
    HADAMARD 1, amax, 1, 5, 7, 6
    mova m6, spill0
    mova m7, spill1
    paddw m0, m1
    SUMSUB_BADC w, 2, 6, 3, 7, 4
    HADAMARD 2, sumsub, 2, 6, 4, 5
    HADAMARD 2, sumsub, 3, 7, 4, 5
    HADAMARD 1, amax, 2, 6, 4, 5
    HADAMARD 1, amax, 3, 7, 4, 5
%endif ; sse2/non-sse2
    paddw m0, m2
    paddw m0, m3
    SAVE_MM_PERMUTATION
    ret
%endif ; ifndef mmx2

cglobal pixel_sa8d_8x8, 4,7
    FIX_STRIDES r1, r3
    mov    r6, esp
    and   esp, ~15
    sub   esp, 48
    lea    r4, [3*r1]
    lea    r5, [3*r3]
    call pixel_sa8d_8x8_internal
%if HIGH_BIT_DEPTH
    HADDUW m0, m1
%else
    HADDW  m0, m1
%endif ; HIGH_BIT_DEPTH
    movd  eax, m0
    add   eax, 1
    shr   eax, 1
    mov   esp, r6
    RET

cglobal pixel_sa8d_16x16, 4,7
    FIX_STRIDES r1, r3
    mov  r6, esp
    and  esp, ~15
    sub  esp, 64
    lea  r4, [3*r1]
    lea  r5, [3*r3]
    call pixel_sa8d_8x8_internal
%if mmsize == 8
    lea  r0, [r0+4*r1]
    lea  r2, [r2+4*r3]
%endif
%if HIGH_BIT_DEPTH
    HADDUW m0, m1
%endif
    mova [esp+48], m0
    call pixel_sa8d_8x8_internal
    mov  r0, [r6+20]
    mov  r2, [r6+28]
    add  r0, 8*SIZEOF_PIXEL
    add  r2, 8*SIZEOF_PIXEL
    SA8D_INTER
    mova [esp+48], m0
    call pixel_sa8d_8x8_internal
%if mmsize == 8
    lea  r0, [r0+4*r1]
    lea  r2, [r2+4*r3]
%else
    SA8D_INTER
%endif
    mova [esp+64-mmsize], m0
    call pixel_sa8d_8x8_internal
%if HIGH_BIT_DEPTH
    SA8D_INTER
%else ; !HIGH_BIT_DEPTH
    paddusw m0, [esp+64-mmsize]
%if mmsize == 16
    HADDUW m0, m1
%else
    mova m2, [esp+48]
    pxor m7, m7
    mova m1, m0
    mova m3, m2
    punpcklwd m0, m7
    punpckhwd m1, m7
    punpcklwd m2, m7
    punpckhwd m3, m7
    paddd m0, m1
    paddd m2, m3
    paddd m0, m2
    HADDD m0, m1
%endif
%endif ; HIGH_BIT_DEPTH
    movd eax, m0
    add  eax, 1
    shr  eax, 1
    mov  esp, r6
    RET
%endif ; !ARCH_X86_64
%endmacro ; SA8D

;=============================================================================
; SA8D_SATD
;=============================================================================

; %1: vertical/horizontal mode
; %2-%5: sa8d output regs (m0,m1,m2,m3,m4,m5,m8,m9)
; m10: satd result
; m6, m11-15: tmp regs
%macro SA8D_SATD_8x4 5
%if %1
    LOAD_DIFF_8x4P %2, %3, %4, %5, 6, 11, 7, r0, r2, 1
    HADAMARD   0, sumsub, %2, %3, 6
    HADAMARD   0, sumsub, %4, %5, 6
    SBUTTERFLY        wd, %2, %3, 6
    SBUTTERFLY        wd, %4, %5, 6
    HADAMARD2_2D  %2, %4, %3, %5, 6, dq

    mova   m12, m%2
    mova   m13, m%3
    mova   m14, m%4
    mova   m15, m%5
    HADAMARD 0, sumsub, %2, %3, 6
    HADAMARD 0, sumsub, %4, %5, 6
    SBUTTERFLY     qdq, 12, 13, 6
    HADAMARD   0, amax, 12, 13, 6
    SBUTTERFLY     qdq, 14, 15, 6
    paddw m10, m12
    HADAMARD   0, amax, 14, 15, 6
    paddw m10, m14
%else
    LOAD_SUMSUB_8x4P %2, %3, %4, %5, 6, 11, 7, r0, r2, 1
    HADAMARD4_V %2, %3, %4, %5, 6

    pabsw    m12, m%2 ; doing the abs first is a slight advantage
    pabsw    m14, m%4
    pabsw    m13, m%3
    pabsw    m15, m%5
    HADAMARD 1, max, 12, 14, 6, 11
    paddw    m10, m12
    HADAMARD 1, max, 13, 15, 6, 11
    paddw    m10, m13
%endif
%endmacro ; SA8D_SATD_8x4

; %1: add spilled regs?
; %2: spill regs?
%macro SA8D_SATD_ACCUM 2
%if HIGH_BIT_DEPTH
    pmaddwd m10, [pw_1]
    HADDUWD  m0, m1
%if %1
    paddd   m10, temp1
    paddd    m0, temp0
%endif
%if %2
    mova  temp1, m10
    pxor    m10, m10
%endif
%elif %1
    paddw    m0, temp0
%endif
%if %2
    mova  temp0, m0
%endif
%endmacro

%macro SA8D_SATD 0
%define vertical ((notcpuflag(ssse3) || cpuflag(atom)) || HIGH_BIT_DEPTH)
cglobal pixel_sa8d_satd_8x8_internal
    SA8D_SATD_8x4 vertical, 0, 1, 2, 3
    SA8D_SATD_8x4 vertical, 4, 5, 8, 9

%if vertical ; sse2-style
    HADAMARD2_2D 0, 4, 2, 8, 6, qdq, amax
    HADAMARD2_2D 1, 5, 3, 9, 6, qdq, amax
%else        ; complete sa8d
    SUMSUB_BADC w, 0, 4, 1, 5, 12
    HADAMARD 2, sumsub, 0, 4, 12, 11
    HADAMARD 2, sumsub, 1, 5, 12, 11
    SUMSUB_BADC w, 2, 8, 3, 9, 12
    HADAMARD 2, sumsub, 2, 8, 12, 11
    HADAMARD 2, sumsub, 3, 9, 12, 11
    HADAMARD 1, amax, 0, 4, 12, 11
    HADAMARD 1, amax, 1, 5, 12, 4
    HADAMARD 1, amax, 2, 8, 12, 4
    HADAMARD 1, amax, 3, 9, 12, 4
%endif

    ; create sa8d sub results
    paddw    m1, m2
    paddw    m0, m3
    paddw    m0, m1

    SAVE_MM_PERMUTATION
    ret

;-------------------------------------------------------------------------------
; uint64_t pixel_sa8d_satd_16x16( pixel *, intptr_t, pixel *, intptr_t )
;-------------------------------------------------------------------------------
cglobal pixel_sa8d_satd_16x16, 4,8-(mmsize/32),16,SIZEOF_PIXEL*mmsize
    %define temp0 [rsp+0*mmsize]
    %define temp1 [rsp+1*mmsize]
    FIX_STRIDES r1, r3
%if vertical==0
    mova     m7, [hmul_8p]
%endif
    lea      r4, [3*r1]
    lea      r5, [3*r3]
    pxor    m10, m10

%if mmsize==32
    call pixel_sa8d_satd_8x8_internal
    SA8D_SATD_ACCUM 0, 1
    call pixel_sa8d_satd_8x8_internal
    SA8D_SATD_ACCUM 1, 0
    vextracti128 xm1, m0, 1
    vextracti128 xm2, m10, 1
    paddw   xm0, xm1
    paddw  xm10, xm2
%else
    lea      r6, [r2+8*SIZEOF_PIXEL]
    lea      r7, [r0+8*SIZEOF_PIXEL]

    call pixel_sa8d_satd_8x8_internal
    SA8D_SATD_ACCUM 0, 1
    call pixel_sa8d_satd_8x8_internal
    SA8D_SATD_ACCUM 1, 1

    mov      r0, r7
    mov      r2, r6

    call pixel_sa8d_satd_8x8_internal
    SA8D_SATD_ACCUM 1, 1
    call pixel_sa8d_satd_8x8_internal
    SA8D_SATD_ACCUM 1, 0
%endif

; xop already has fast horizontal sums
%if cpuflag(sse4) && notcpuflag(xop) && HIGH_BIT_DEPTH==0
    pmaddwd xm10, [pw_1]
    HADDUWD xm0, xm1
    phaddd  xm0, xm10       ;  sa8d1  sa8d2  satd1  satd2
    pshufd  xm1, xm0, q2301 ;  sa8d2  sa8d1  satd2  satd1
    paddd   xm0, xm1        ;   sa8d   sa8d   satd   satd
    movd    r0d, xm0
    pextrd  eax, xm0, 2
%else
%if HIGH_BIT_DEPTH
    HADDD   xm0, xm1
    HADDD  xm10, xm2
%else
    HADDUW  xm0, xm1
    HADDW  xm10, xm2
%endif
    movd    r0d, xm0
    movd    eax, xm10
%endif
    add     r0d, 1
    shl     rax, 32
    shr     r0d, 1
    or      rax, r0
    RET
%endmacro ; SA8D_SATD

;=============================================================================
; INTRA SATD
;=============================================================================

%macro HSUMSUB2 8
    pshufd %4, %2, %7
    pshufd %5, %3, %7
    %1     %2, %8
    %1     %6, %8
    paddw  %2, %4
    paddw  %3, %5
%endmacro

; intra_sa8d_x3_8x8 and intra_satd_x3_4x4 are obsoleted by x9 on ssse3+,
; and are only retained for old cpus.
%macro INTRA_SA8D_SSE2 0
%if ARCH_X86_64
;-----------------------------------------------------------------------------
; void intra_sa8d_x3_8x8( uint8_t *fenc, uint8_t edge[36], int *res )
;-----------------------------------------------------------------------------
cglobal intra_sa8d_x3_8x8, 3,3,13
    ; 8x8 hadamard
    pxor        m8, m8
    movq        m0, [r0+0*FENC_STRIDE]
    movq        m1, [r0+1*FENC_STRIDE]
    movq        m2, [r0+2*FENC_STRIDE]
    movq        m3, [r0+3*FENC_STRIDE]
    movq        m4, [r0+4*FENC_STRIDE]
    movq        m5, [r0+5*FENC_STRIDE]
    movq        m6, [r0+6*FENC_STRIDE]
    movq        m7, [r0+7*FENC_STRIDE]
    punpcklbw   m0, m8
    punpcklbw   m1, m8
    punpcklbw   m2, m8
    punpcklbw   m3, m8
    punpcklbw   m4, m8
    punpcklbw   m5, m8
    punpcklbw   m6, m8
    punpcklbw   m7, m8

    HADAMARD8_2D 0, 1, 2, 3, 4, 5, 6, 7, 8

    ABSW2       m8, m9, m2, m3, m2, m3
    ABSW2      m10, m11, m4, m5, m4, m5
    paddw       m8, m10
    paddw       m9, m11
    ABSW2      m10, m11, m6, m7, m6, m7
    ABSW       m12, m1, m1
    paddw      m10, m11
    paddw       m8, m9
    paddw      m12, m10
    paddw      m12, m8

    ; 1D hadamard of edges
    movq        m8, [r1+7]
    movq        m9, [r1+16]
    pxor       m10, m10
    punpcklbw   m8, m10
    punpcklbw   m9, m10
    HSUMSUB2 pmullw, m8, m9, m10, m11, m11, q1032, [pw_ppppmmmm]
    HSUMSUB2 pmullw, m8, m9, m10, m11, m11, q2301, [pw_ppmmppmm]
    pshuflw    m10, m8, q2301
    pshuflw    m11, m9, q2301
    pshufhw    m10, m10, q2301
    pshufhw    m11, m11, q2301
    pmullw      m8, [pw_pmpmpmpm]
    pmullw     m11, [pw_pmpmpmpm]
    paddw       m8, m10
    paddw       m9, m11

    ; differences
    paddw      m10, m8, m9
    paddw      m10, [pw_8]
    pand       m10, [sw_f0]
    psllw       m8, 3 ; left edge
    psllw      m10, 2 ; dc
    psubw       m8, m0
    psubw      m10, m0
    punpcklwd   m0, m1
    punpcklwd   m2, m3
    punpcklwd   m4, m5
    punpcklwd   m6, m7
    ABSW       m10, m10, m1
    paddw      m10, m12
    punpckldq   m0, m2
    punpckldq   m4, m6
    punpcklqdq  m0, m4 ; transpose
    psllw       m9, 3 ; top edge
    psrldq      m2, m10, 2 ; 8x7 sum
    psubw       m0, m9  ; 8x1 sum
    ABSW2       m8, m0, m8, m0, m1, m3 ; 1x8 sum
    paddw       m8, m12
    paddusw     m2, m0

    ; 3x HADDW
    mova        m7, [pd_f0]
    pandn       m0, m7, m10
    psrld      m10, 16
    pandn       m1, m7, m8
    psrld       m8, 16
    pandn       m7, m2
    psrld       m2, 16
    paddd       m0, m10
    paddd       m1, m8
    paddd       m2, m7
    pshufd      m3, m0, q2301
    punpckhdq   m4, m2, m1
    punpckldq   m2, m1
    paddd       m3, m0
    paddd       m2, m4
    punpckhqdq  m0, m2, m3
    punpcklqdq  m2, m3
    paddd       m0, [pd_2]
    paddd       m0, m2
    psrld       m0, 2
    mova      [r2], m0
    RET
%endif ; ARCH_X86_64
%endmacro ; INTRA_SA8D_SSE2

; in: r0 = fenc
; out: m0..m3 = hadamard coefs
INIT_MMX
cglobal hadamard_load
; not really a global, but otherwise cycles get attributed to the wrong function in profiling
%if HIGH_BIT_DEPTH
    mova        m0, [r0+0*FENC_STRIDEB]
    mova        m1, [r0+1*FENC_STRIDEB]
    mova        m2, [r0+2*FENC_STRIDEB]
    mova        m3, [r0+3*FENC_STRIDEB]
%else
    pxor        m7, m7
    movd        m0, [r0+0*FENC_STRIDE]
    movd        m1, [r0+1*FENC_STRIDE]
    movd        m2, [r0+2*FENC_STRIDE]
    movd        m3, [r0+3*FENC_STRIDE]
    punpcklbw   m0, m7
    punpcklbw   m1, m7
    punpcklbw   m2, m7
    punpcklbw   m3, m7
%endif
    HADAMARD4_2D 0, 1, 2, 3, 4
    SAVE_MM_PERMUTATION
    ret

%macro SCALAR_HADAMARD 4-5 ; direction, offset, 3x tmp
%ifidn %1, top
%if HIGH_BIT_DEPTH
    mova        %3, [r1+%2*SIZEOF_PIXEL-FDEC_STRIDEB]
%else
    movd        %3, [r1+%2*SIZEOF_PIXEL-FDEC_STRIDEB]
    pxor        %5, %5
    punpcklbw   %3, %5
%endif
%else ; left
%ifnidn %2, 0
    shl         %2d, 5 ; log(FDEC_STRIDEB)
%endif
    movd        %3, [r1+%2*SIZEOF_PIXEL-4+1*FDEC_STRIDEB]
    pinsrw      %3, [r1+%2*SIZEOF_PIXEL-2+0*FDEC_STRIDEB], 0
    pinsrw      %3, [r1+%2*SIZEOF_PIXEL-2+2*FDEC_STRIDEB], 2
    pinsrw      %3, [r1+%2*SIZEOF_PIXEL-2+3*FDEC_STRIDEB], 3
%if HIGH_BIT_DEPTH == 0
    psrlw       %3, 8
%endif
%ifnidn %2, 0
    shr         %2d, 5
%endif
%endif ; direction
%if cpuflag(ssse3)
    %define %%sign psignw
%else
    %define %%sign pmullw
%endif
    pshufw      %4, %3, q1032
    %%sign      %4, [pw_ppmmppmm]
    paddw       %3, %4
    pshufw      %4, %3, q2301
    %%sign      %4, [pw_pmpmpmpm]
    paddw       %3, %4
    psllw       %3, 2
    mova        [%1_1d+2*%2], %3
%endmacro

%macro SUM_MM_X3 8 ; 3x sum, 4x tmp, op
    pxor        %7, %7
    pshufw      %4, %1, q1032
    pshufw      %5, %2, q1032
    pshufw      %6, %3, q1032
    paddw       %1, %4
    paddw       %2, %5
    paddw       %3, %6
    punpcklwd   %1, %7
    punpcklwd   %2, %7
    punpcklwd   %3, %7
    pshufw      %4, %1, q1032
    pshufw      %5, %2, q1032
    pshufw      %6, %3, q1032
    %8          %1, %4
    %8          %2, %5
    %8          %3, %6
%endmacro

; in: m1..m3
; out: m7
; clobber: m4..m6
%macro SUM3x4 0
    ABSW2       m4, m5, m1, m2, m1, m2
    ABSW        m7, m3, m3
    paddw       m4, m5
    paddw       m7, m4
%endmacro

; in: m0..m3 (4x4)
; out: m0 v, m4 h, m5 dc
; clobber: m1..m3
%macro SUM4x3 3 ; dc, left, top
    movq        m4, %2
%ifnum sizeof%1
    movq        m5, %1
%else
    movd        m5, %1
%endif
    psubw       m4, m0
    psubw       m5, m0
    punpcklwd   m0, m1
    punpcklwd   m2, m3
    punpckldq   m0, m2 ; transpose
    psubw       m0, %3
    ABSW2       m4, m5, m4, m5, m2, m3 ; 1x4 sum
    ABSW        m0, m0, m1 ; 4x1 sum
%endmacro

%macro INTRA_X3_MMX 0
;-----------------------------------------------------------------------------
; void intra_satd_x3_4x4( uint8_t *fenc, uint8_t *fdec, int *res )
;-----------------------------------------------------------------------------
cglobal intra_satd_x3_4x4, 3,3
%if UNIX64
    ; stack is 16 byte aligned because abi says so
    %define  top_1d  rsp-8  ; size 8
    %define  left_1d rsp-16 ; size 8
%else
    ; WIN64:  stack is 16 byte aligned because abi says so
    ; X86_32: stack is 16 byte aligned at least in gcc, and we've pushed 3 regs + return address, so it's still aligned
    SUB         rsp, 16
    %define  top_1d  rsp+8
    %define  left_1d rsp
%endif

    call hadamard_load
    SCALAR_HADAMARD left, 0, m4, m5
    SCALAR_HADAMARD top,  0, m6, m5, m7
    paddw       m6, m4
    pavgw       m6, [pw_16]
    pand        m6, [sw_f0] ; dc

    SUM3x4
    SUM4x3 m6, [left_1d], [top_1d]
    paddw       m4, m7
    paddw       m5, m7
    movq        m1, m5
    psrlq       m1, 16  ; 4x3 sum
    paddw       m0, m1

    SUM_MM_X3   m0, m4, m5, m1, m2, m3, m6, pavgw
    movd        [r2+0], m0 ; i4x4_v satd
    movd        [r2+4], m4 ; i4x4_h satd
    movd        [r2+8], m5 ; i4x4_dc satd
%if UNIX64 == 0
    ADD         rsp, 16
%endif
    RET

;-----------------------------------------------------------------------------
; void intra_satd_x3_16x16( uint8_t *fenc, uint8_t *fdec, int *res )
;-----------------------------------------------------------------------------
cglobal intra_satd_x3_16x16, 0,5
    %assign  stack_pad  120 + ((stack_offset+120+gprsize)&15)
    ; not really needed on x86_64, just shuts up valgrind about storing data below the stack across a function call
    SUB         rsp, stack_pad
%define sums    rsp+64 ; size 56
%define top_1d  rsp+32 ; size 32
%define left_1d rsp    ; size 32
    movifnidn   r1,  r1mp

    pxor        m7, m7
    mova [sums+ 0], m7
    mova [sums+ 8], m7
    mova [sums+16], m7
%if HIGH_BIT_DEPTH
    mova [sums+24], m7
    mova [sums+32], m7
    mova [sums+40], m7
    mova [sums+48], m7
%endif

    ; 1D hadamards
    mov        r3d, 12
    movd        m6, [pw_32]
.loop_edge:
    SCALAR_HADAMARD left, r3, m0, m1
    SCALAR_HADAMARD top,  r3, m1, m2, m3
    pavgw       m0, m1
    paddw       m6, m0
    sub        r3d, 4
    jge .loop_edge
    psrlw       m6, 2
    pand        m6, [sw_f0] ; dc

    ; 2D hadamards
    movifnidn   r0, r0mp
    mov         r3, -4
.loop_y:
    mov         r4, -4
.loop_x:
    call hadamard_load

    SUM3x4
    SUM4x3 m6, [left_1d+8*(r3+4)], [top_1d+8*(r4+4)]
    pavgw       m4, m7
    pavgw       m5, m7
    paddw       m0, [sums+ 0] ; i16x16_v satd
    paddw       m4, [sums+ 8] ; i16x16_h satd
    paddw       m5, [sums+16] ; i16x16_dc satd
    mova [sums+ 0], m0
    mova [sums+ 8], m4
    mova [sums+16], m5

    add         r0, 4*SIZEOF_PIXEL
    inc         r4
    jl  .loop_x
%if HIGH_BIT_DEPTH
    psrld       m7, m4, 16
    pslld       m4, 16
    psrld       m4, 16
    paddd       m4, m7
    psrld       m7, m0, 16
    pslld       m0, 16
    psrld       m0, 16
    paddd       m0, m7
    paddd       m4, [sums+32]
    paddd       m0, [sums+24]
    mova [sums+32], m4
    mova [sums+24], m0
    pxor        m7, m7
    punpckhwd   m3, m5, m7
    punpcklwd   m5, m7
    paddd       m3, [sums+48]
    paddd       m5, [sums+40]
    mova [sums+48], m3
    mova [sums+40], m5
    mova [sums+ 0], m7
    mova [sums+ 8], m7
    mova [sums+16], m7
%endif
    add         r0, 4*FENC_STRIDEB-16*SIZEOF_PIXEL
    inc         r3
    jl  .loop_y

; horizontal sum
    movifnidn   r2, r2mp
%if HIGH_BIT_DEPTH
    mova        m1, m5
    paddd       m5, m3
    HADDD       m5, m7 ; DC satd
    HADDD       m4, m7 ; H satd
    HADDD       m0, m7 ; the part of V satd that doesn't overlap with DC
    psrld       m0, 1
    psrlq       m1, 32 ; DC[1]
    paddd       m0, m3 ; DC[2]
    psrlq       m3, 32 ; DC[3]
    paddd       m0, m1
    paddd       m0, m3
%else
    mova        m7, m5
    SUM_MM_X3   m0, m4, m5, m3, m1, m2, m6, paddd
    psrld       m0, 1
    pslld       m7, 16
    psrld       m7, 16
    paddd       m0, m5
    psubd       m0, m7
%endif
    movd    [r2+8], m5 ; i16x16_dc satd
    movd    [r2+4], m4 ; i16x16_h satd
    movd    [r2+0], m0 ; i16x16_v satd
    ADD        rsp, stack_pad
    RET

%if ARCH_X86_64
    %define  t0 r6
%else
    %define  t0 r2
%endif

;-----------------------------------------------------------------------------
; void intra_satd_x3_8x8c( uint8_t *fenc, uint8_t *fdec, int *res )
;-----------------------------------------------------------------------------
cglobal intra_satd_x3_8x8c, 0,6
    ; not really needed on x86_64, just shuts up valgrind about storing data below the stack across a function call
    SUB          rsp, 72
%define  sums    rsp+48 ; size 24
%define  dc_1d   rsp+32 ; size 16
%define  top_1d  rsp+16 ; size 16
%define  left_1d rsp    ; size 16
    movifnidn   r1,  r1mp
    pxor        m7, m7
    mova [sums+ 0], m7
    mova [sums+ 8], m7
    mova [sums+16], m7

    ; 1D hadamards
    mov         r3d, 4
.loop_edge:
    SCALAR_HADAMARD left, r3, m0, m1
    SCALAR_HADAMARD top,  r3, m0, m1, m2
    sub         r3d, 4
    jge .loop_edge

    ; dc
    movzx       t0d, word [left_1d+0]
    movzx       r3d, word [top_1d+0]
    movzx       r4d, word [left_1d+8]
    movzx       r5d, word [top_1d+8]
    lea         t0d, [t0 + r3 + 16]
    lea         r3d, [r4 + r5 + 16]
    shr         t0d, 1
    shr         r3d, 1
    add         r4d, 8
    add         r5d, 8
    and         t0d, -16 ; tl
    and         r3d, -16 ; br
    and         r4d, -16 ; bl
    and         r5d, -16 ; tr
    mov         [dc_1d+ 0], t0d ; tl
    mov         [dc_1d+ 4], r5d ; tr
    mov         [dc_1d+ 8], r4d ; bl
    mov         [dc_1d+12], r3d ; br
    lea         r5, [dc_1d]

    ; 2D hadamards
    movifnidn   r0,  r0mp
    movifnidn   r2,  r2mp
    mov         r3,  -2
.loop_y:
    mov         r4,  -2
.loop_x:
    call hadamard_load

    SUM3x4
    SUM4x3 [r5+4*(r4+2)], [left_1d+8*(r3+2)], [top_1d+8*(r4+2)]
    pavgw       m4, m7
    pavgw       m5, m7
    paddw       m0, [sums+16] ; i4x4_v satd
    paddw       m4, [sums+8]  ; i4x4_h satd
    paddw       m5, [sums+0]  ; i4x4_dc satd
    movq        [sums+16], m0
    movq        [sums+8], m4
    movq        [sums+0], m5

    add         r0, 4*SIZEOF_PIXEL
    inc         r4
    jl  .loop_x
    add         r0, 4*FENC_STRIDEB-8*SIZEOF_PIXEL
    add         r5, 8
    inc         r3
    jl  .loop_y

; horizontal sum
    movq        m0, [sums+0]
    movq        m1, [sums+8]
    movq        m2, [sums+16]
    movq        m7, m0
%if HIGH_BIT_DEPTH
    psrlq       m7, 16
    HADDW       m7, m3
    SUM_MM_X3   m0, m1, m2, m3, m4, m5, m6, paddd
    psrld       m2, 1
    paddd       m2, m7
%else
    psrlq       m7, 15
    paddw       m2, m7
    SUM_MM_X3   m0, m1, m2, m3, m4, m5, m6, paddd
    psrld       m2, 1
%endif
    movd        [r2+0], m0 ; i8x8c_dc satd
    movd        [r2+4], m1 ; i8x8c_h satd
    movd        [r2+8], m2 ; i8x8c_v satd
    ADD         rsp, 72
    RET
%endmacro ; INTRA_X3_MMX



%macro PRED4x4_LOWPASS 5
%ifnum sizeof%5
    pavgb       %5, %2, %3
    pxor        %3, %2
    pand        %3, [pb_1]
    psubusb     %5, %3
    pavgb       %1, %4, %5
%else
    mova        %5, %2
    pavgb       %2, %3
    pxor        %3, %5
    pand        %3, [pb_1]
    psubusb     %2, %3
    pavgb       %1, %4, %2
%endif
%endmacro

%macro INTRA_X9_PRED 2
%if cpuflag(sse4)
    movu       m1, [r1-1*FDEC_STRIDE-8]
    pinsrb     m1, [r1+3*FDEC_STRIDE-1], 0
    pinsrb     m1, [r1+2*FDEC_STRIDE-1], 1
    pinsrb     m1, [r1+1*FDEC_STRIDE-1], 2
    pinsrb     m1, [r1+0*FDEC_STRIDE-1], 3
%else
    movd      mm0, [r1+3*FDEC_STRIDE-4]
    punpcklbw mm0, [r1+2*FDEC_STRIDE-4]
    movd      mm1, [r1+1*FDEC_STRIDE-4]
    punpcklbw mm1, [r1+0*FDEC_STRIDE-4]
    punpckhwd mm0, mm1
    psrlq     mm0, 32
    movq2dq    m0, mm0
    movu       m1, [r1-1*FDEC_STRIDE-8]
    movss      m1, m0                  ; l3 l2 l1 l0 __ __ __ lt t0 t1 t2 t3 t4 t5 t6 t7
%endif ; cpuflag
    pshufb     m1, [intrax9_edge]      ; l3 l3 l2 l1 l0 lt t0 t1 t2 t3 t4 t5 t6 t7 t7 __
    psrldq     m0, m1, 1               ; l3 l2 l1 l0 lt t0 t1 t2 t3 t4 t5 t6 t7 t7 __ __
    psrldq     m2, m1, 2               ; l2 l1 l0 lt t0 t1 t2 t3 t4 t5 t6 t7 t7 __ __ __
    pavgb      m5, m0, m1              ; Gl3 Gl2 Gl1 Gl0 Glt Gt0 Gt1 Gt2 Gt3 Gt4 Gt5  __  __ __ __ __
    mova       %2, m1
    PRED4x4_LOWPASS m0, m1, m2, m0, m4 ; Fl3 Fl2 Fl1 Fl0 Flt Ft0 Ft1 Ft2 Ft3 Ft4 Ft5 Ft6 Ft7 __ __ __
    ; ddl               ddr
    ; Ft1 Ft2 Ft3 Ft4   Flt Ft0 Ft1 Ft2
    ; Ft2 Ft3 Ft4 Ft5   Fl0 Flt Ft0 Ft1
    ; Ft3 Ft4 Ft5 Ft6   Fl1 Fl0 Flt Ft0
    ; Ft4 Ft5 Ft6 Ft7   Fl2 Fl1 Fl0 Flt
    pshufb     m2, m0, [%1_ddlr1] ; a: ddl row0, ddl row1, ddr row0, ddr row1 / b: ddl row0, ddr row0, ddl row1, ddr row1
    pshufb     m3, m0, [%1_ddlr2] ; rows 2,3
    ; hd                hu
    ; Glt Flt Ft0 Ft1   Gl0 Fl1 Gl1 Fl2
    ; Gl0 Fl0 Glt Flt   Gl1 Fl2 Gl2 Fl3
    ; Gl1 Fl1 Gl0 Fl0   Gl2 Fl3 Gl3 Gl3
    ; Gl2 Fl2 Gl1 Fl1   Gl3 Gl3 Gl3 Gl3
    pslldq     m0, 5                   ; ___ ___ ___ ___ ___ Fl3 Fl2 Fl1 Fl0 Flt Ft0 Ft1 Ft2 Ft3 Ft4 Ft5
    palignr    m7, m5, m0, 5           ; Fl3 Fl2 Fl1 Fl0 Flt Ft0 Ft1 Ft2 Ft3 Ft4 Ft5 Gl3 Gl2 Gl1 Gl0 Glt
    pshufb     m6, m7, [%1_hdu1]
    pshufb     m7, m7, [%1_hdu2]
    ; vr                vl
    ; Gt0 Gt1 Gt2 Gt3   Gt1 Gt2 Gt3 Gt4
    ; Flt Ft0 Ft1 Ft2   Ft1 Ft2 Ft3 Ft4
    ; Fl0 Gt0 Gt1 Gt2   Gt2 Gt3 Gt4 Gt5
    ; Fl1 Flt Ft0 Ft1   Ft2 Ft3 Ft4 Ft5
    psrldq     m5, 5                   ; Gt0 Gt1 Gt2 Gt3 Gt4 Gt5 ...
    palignr    m5, m0, 6               ; ___ Fl1 Fl0 Flt Ft0 Ft1 Ft2 Ft3 Ft4 Ft5 Gt0 Gt1 Gt2 Gt3 Gt4 Gt5
    pshufb     m4, m5, [%1_vrl1]
    pshufb     m5, m5, [%1_vrl2]
%endmacro ; INTRA_X9_PRED

%macro INTRA_X9_VHDC 5 ; edge, fenc01, fenc23, tmp, tmp
    pshufb     m2, m%1, [intrax9b_vh1]
    pshufb     m3, m%1, [intrax9b_vh2]
    mova      [pred_buf+0x60], m2
    mova      [pred_buf+0x70], m3
    pshufb    m%1, [intrax9b_edge2] ; t0 t1 t2 t3 t0 t1 t2 t3 l0 l1 l2 l3 l0 l1 l2 l3
    pmaddubsw m%1, [hmul_4p]
    pshufhw    m0, m%1, q2301
    pshuflw    m0, m0,  q2301
    psignw    m%1, [pw_pmpmpmpm]
    paddw      m0, m%1
    psllw      m0, 2 ; hadamard(top), hadamard(left)
    MOVHL      m3, m0
    pshufb     m1, m0, [intrax9b_v1]
    pshufb     m2, m0, [intrax9b_v2]
    paddw      m0, m3
    psignw     m3, [pw_pmmpzzzz] ; FIXME could this be eliminated?
    pavgw      m0, [pw_16]
    pand       m0, [sw_f0] ; dc
    ; This (as well as one of the steps in intra_satd_x9_4x4.satd_8x4) could be
    ; changed from a wd transpose to a qdq, with appropriate rearrangement of inputs.
    ; Which would be faster on conroe, but slower on penryn and sandybridge, and too invasive to ifdef.
    HADAMARD 0, sumsub, %2, %3, %4, %5
    HADAMARD 1, sumsub, %2, %3, %4, %5
    movd      r3d, m0
    shr       r3d, 4
    imul      r3d, 0x01010101
    mov       [pred_buf+0x80], r3d
    mov       [pred_buf+0x88], r3d
    mov       [pred_buf+0x90], r3d
    mov       [pred_buf+0x98], r3d
    psubw      m3, m%2
    psubw      m0, m%2
    psubw      m1, m%2
    psubw      m2, m%3
    pabsw     m%3, m%3
    pabsw      m3, m3
    pabsw      m0, m0
    pabsw      m1, m1
    pabsw      m2, m2
    pavgw      m3, m%3
    pavgw      m0, m%3
    pavgw      m1, m2
%if cpuflag(sse4)
    phaddw     m3, m0
%else
    SBUTTERFLY qdq, 3, 0, 2
    paddw      m3, m0
%endif
    MOVHL      m2, m1
    paddw      m1, m2
%if cpuflag(xop)
    vphaddwq   m3, m3
    vphaddwq   m1, m1
    packssdw   m1, m3
%else
    phaddw     m1, m3
    pmaddwd    m1, [pw_1] ; v, _, h, dc
%endif
%endmacro ; INTRA_X9_VHDC

%macro INTRA_X9_END 2
%if cpuflag(sse4)
    phminposuw m0, m0 ; h,dc,ddl,ddr,vr,hd,vl,hu
    movd      eax, m0
    add       eax, 1<<16
    cmp        ax, r3w
    cmovge    eax, r3d
%else
%if %1
    ; 4x4 sad is up to 12 bits; +bitcosts -> 13 bits; pack with 3 bit index
    psllw      m0, 3
    paddw      m0, [pw_s01234567] ; h,dc,ddl,ddr,vr,hd,vl,hu
%else
    ; 4x4 satd is up to 13 bits; +bitcosts and saturate -> 13 bits; pack with 3 bit index
    psllw      m0, 2
    paddusw    m0, m0
    paddw      m0, [pw_s01234657] ; h,dc,ddl,ddr,vr,vl,hd,hu
%endif
    movhlps    m1, m0
    pminsw     m0, m1
    pshuflw    m1, m0, q0032
    pminsw     m0, m1
    pshuflw    m1, m0, q0001
    pminsw     m0, m1
    movd      eax, m0
    movsx     r2d, ax
    and       eax, 7
    sar       r2d, 3
    shl       eax, 16
    ; 1<<16: increment index to match intra4x4_pred_e. couldn't do this before because it had to fit in 3 bits
    ; 1<<12: undo sign manipulation
    lea       eax, [rax+r2+(1<<16)+(1<<12)]
    cmp        ax, r3w
    cmovge    eax, r3d
%endif ; cpuflag

    ; output the predicted samples
    mov       r3d, eax
    shr       r3d, 16
%if ARCH_X86_64
    lea        r2, [%2_lut]
    movzx     r2d, byte [r2+r3]
%else
    movzx     r2d, byte [%2_lut+r3]
%endif
%if %1 ; sad
    movq      mm0, [pred_buf+r2]
    movq      mm1, [pred_buf+r2+16]
    movd     [r1+0*FDEC_STRIDE], mm0
    movd     [r1+2*FDEC_STRIDE], mm1
    psrlq     mm0, 32
    psrlq     mm1, 32
    movd     [r1+1*FDEC_STRIDE], mm0
    movd     [r1+3*FDEC_STRIDE], mm1
%else ; satd
%assign i 0
%rep 4
    mov       r3d, [pred_buf+r2+8*i]
    mov      [r1+i*FDEC_STRIDE], r3d
%assign i i+1
%endrep
%endif
%endmacro ; INTRA_X9_END

%macro INTRA_X9 0
;-----------------------------------------------------------------------------
; int intra_sad_x9_4x4( uint8_t *fenc, uint8_t *fdec, uint16_t *bitcosts )
;-----------------------------------------------------------------------------
%if notcpuflag(xop)
cglobal intra_sad_x9_4x4, 3,4,9
    %assign pad 0xc0-gprsize-(stack_offset&15)
    %define pred_buf rsp
    sub       rsp, pad
%if ARCH_X86_64
    INTRA_X9_PRED intrax9a, m8
%else
    INTRA_X9_PRED intrax9a, [rsp+0xa0]
%endif
    mova [rsp+0x00], m2
    mova [rsp+0x10], m3
    mova [rsp+0x20], m4
    mova [rsp+0x30], m5
    mova [rsp+0x40], m6
    mova [rsp+0x50], m7
%if cpuflag(sse4)
    movd       m0, [r0+0*FENC_STRIDE]
    pinsrd     m0, [r0+1*FENC_STRIDE], 1
    movd       m1, [r0+2*FENC_STRIDE]
    pinsrd     m1, [r0+3*FENC_STRIDE], 1
%else
    movd      mm0, [r0+0*FENC_STRIDE]
    punpckldq mm0, [r0+1*FENC_STRIDE]
    movd      mm1, [r0+2*FENC_STRIDE]
    punpckldq mm1, [r0+3*FENC_STRIDE]
    movq2dq    m0, mm0
    movq2dq    m1, mm1
%endif
    punpcklqdq m0, m0
    punpcklqdq m1, m1
    psadbw     m2, m0
    psadbw     m3, m1
    psadbw     m4, m0
    psadbw     m5, m1
    psadbw     m6, m0
    psadbw     m7, m1
    paddd      m2, m3
    paddd      m4, m5
    paddd      m6, m7
%if ARCH_X86_64
    SWAP        7, 8
    pxor       m8, m8
    %define %%zero m8
%else
    mova       m7, [rsp+0xa0]
    %define %%zero [pb_0]
%endif
    pshufb     m3, m7, [intrax9a_vh1]
    pshufb     m5, m7, [intrax9a_vh2]
    pshufb     m7, [intrax9a_dc]
    psadbw     m7, %%zero
    psrlw      m7, 2
    mova [rsp+0x60], m3
    mova [rsp+0x70], m5
    psadbw     m3, m0
    pavgw      m7, %%zero
    pshufb     m7, %%zero
    psadbw     m5, m1
    movq [rsp+0x80], m7
    movq [rsp+0x90], m7
    psadbw     m0, m7
    paddd      m3, m5
    psadbw     m1, m7
    paddd      m0, m1
    movzx     r3d, word [r2]
    movd      r0d, m3 ; v
    add       r3d, r0d
    punpckhqdq m3, m0 ; h, dc
    shufps     m3, m2, q2020
    psllq      m6, 32
    por        m4, m6
    movu       m0, [r2+2]
    packssdw   m3, m4
    paddw      m0, m3
    INTRA_X9_END 1, intrax9a
    add       rsp, pad
    RET
%endif ; cpuflag

%if ARCH_X86_64
;-----------------------------------------------------------------------------
; int intra_satd_x9_4x4( uint8_t *fenc, uint8_t *fdec, uint16_t *bitcosts )
;-----------------------------------------------------------------------------
cglobal intra_satd_x9_4x4, 3,4,16
    %assign pad 0xb0-gprsize-(stack_offset&15)
    %define pred_buf rsp
    sub       rsp, pad
    INTRA_X9_PRED intrax9b, m15
    mova [rsp+0x00], m2
    mova [rsp+0x10], m3
    mova [rsp+0x20], m4
    mova [rsp+0x30], m5
    mova [rsp+0x40], m6
    mova [rsp+0x50], m7
    movd       m8, [r0+0*FENC_STRIDE]
    movd       m9, [r0+1*FENC_STRIDE]
    movd      m10, [r0+2*FENC_STRIDE]
    movd      m11, [r0+3*FENC_STRIDE]
    mova      m12, [hmul_8p]
    pshufd     m8, m8, 0
    pshufd     m9, m9, 0
    pshufd    m10, m10, 0
    pshufd    m11, m11, 0
    pmaddubsw  m8, m12
    pmaddubsw  m9, m12
    pmaddubsw m10, m12
    pmaddubsw m11, m12
    movddup    m0, m2
    pshufd     m1, m2, q3232
    movddup    m2, m3
    punpckhqdq m3, m3
    call .satd_8x4 ; ddr, ddl
    movddup    m2, m5
    pshufd     m3, m5, q3232
    mova       m5, m0
    movddup    m0, m4
    pshufd     m1, m4, q3232
    call .satd_8x4 ; vr, vl
    movddup    m2, m7
    pshufd     m3, m7, q3232
    mova       m4, m0
    movddup    m0, m6
    pshufd     m1, m6, q3232
    call .satd_8x4 ; hd, hu
%if cpuflag(sse4)
    punpckldq  m4, m0
%else
    punpcklqdq m4, m0 ; conroe dislikes punpckldq, and ssse3 INTRA_X9_END can handle arbitrary orders whereas phminposuw can't
%endif
    mova       m1, [pw_ppmmppmm]
    psignw     m8, m1
    psignw    m10, m1
    paddw      m8, m9
    paddw     m10, m11
    INTRA_X9_VHDC 15, 8, 10, 6, 7
    ; find minimum
    movu       m0, [r2+2]
    movd      r3d, m1
    palignr    m5, m1, 8
%if notcpuflag(sse4)
    pshufhw    m0, m0, q3120 ; compensate for different order in unpack
%endif
    packssdw   m5, m4
    paddw      m0, m5
    movzx     r0d, word [r2]
    add       r3d, r0d
    INTRA_X9_END 0, intrax9b
    add       rsp, pad
    RET
RESET_MM_PERMUTATION
ALIGN 16
.satd_8x4:
    pmaddubsw  m0, m12
    pmaddubsw  m1, m12
    pmaddubsw  m2, m12
    pmaddubsw  m3, m12
    psubw      m0, m8
    psubw      m1, m9
    psubw      m2, m10
    psubw      m3, m11
    SATD_8x4_SSE 0, 0, 1, 2, 3, 13, 14, 0, swap
    pmaddwd    m0, [pw_1]
    MOVHL      m1, m0
    paddd    xmm0, m0, m1 ; consistent location of return value. only the avx version of hadamard permutes m0, so 3arg is free
    ret

%else ; !ARCH_X86_64
cglobal intra_satd_x9_4x4, 3,4,8
    %assign pad 0x120-gprsize-(stack_offset&15)
    %define fenc_buf rsp
    %define pred_buf rsp+0x40
    %define spill    rsp+0xe0
    sub       rsp, pad
    INTRA_X9_PRED intrax9b, [spill+0x20]
    mova [pred_buf+0x00], m2
    mova [pred_buf+0x10], m3
    mova [pred_buf+0x20], m4
    mova [pred_buf+0x30], m5
    mova [pred_buf+0x40], m6
    mova [pred_buf+0x50], m7
    movd       m4, [r0+0*FENC_STRIDE]
    movd       m5, [r0+1*FENC_STRIDE]
    movd       m6, [r0+2*FENC_STRIDE]
    movd       m0, [r0+3*FENC_STRIDE]
    mova       m7, [hmul_8p]
    pshufd     m4, m4, 0
    pshufd     m5, m5, 0
    pshufd     m6, m6, 0
    pshufd     m0, m0, 0
    pmaddubsw  m4, m7
    pmaddubsw  m5, m7
    pmaddubsw  m6, m7
    pmaddubsw  m0, m7
    mova [fenc_buf+0x00], m4
    mova [fenc_buf+0x10], m5
    mova [fenc_buf+0x20], m6
    mova [fenc_buf+0x30], m0
    movddup    m0, m2
    pshufd     m1, m2, q3232
    movddup    m2, m3
    punpckhqdq m3, m3
    pmaddubsw  m0, m7
    pmaddubsw  m1, m7
    pmaddubsw  m2, m7
    pmaddubsw  m3, m7
    psubw      m0, m4
    psubw      m1, m5
    psubw      m2, m6
    call .satd_8x4b ; ddr, ddl
    mova       m3, [pred_buf+0x30]
    mova       m1, [pred_buf+0x20]
    movddup    m2, m3
    punpckhqdq m3, m3
    movq [spill+0x08], m0
    movddup    m0, m1
    punpckhqdq m1, m1
    call .satd_8x4 ; vr, vl
    mova       m3, [pred_buf+0x50]
    mova       m1, [pred_buf+0x40]
    movddup    m2, m3
    punpckhqdq m3, m3
    movq [spill+0x10], m0
    movddup    m0, m1
    punpckhqdq m1, m1
    call .satd_8x4 ; hd, hu
    movq [spill+0x18], m0
    mova       m1, [spill+0x20]
    mova       m4, [fenc_buf+0x00]
    mova       m5, [fenc_buf+0x20]
    mova       m2, [pw_ppmmppmm]
    psignw     m4, m2
    psignw     m5, m2
    paddw      m4, [fenc_buf+0x10]
    paddw      m5, [fenc_buf+0x30]
    INTRA_X9_VHDC 1, 4, 5, 6, 7
    ; find minimum
    movu       m0, [r2+2]
    movd      r3d, m1
    punpckhqdq m1, [spill+0x00]
    packssdw   m1, [spill+0x10]
%if cpuflag(sse4)
    pshufhw    m1, m1, q3120
%else
    pshufhw    m0, m0, q3120
%endif
    paddw      m0, m1
    movzx     r0d, word [r2]
    add       r3d, r0d
    INTRA_X9_END 0, intrax9b
    add       rsp, pad
    RET
RESET_MM_PERMUTATION
ALIGN 16
.satd_8x4:
    pmaddubsw  m0, m7
    pmaddubsw  m1, m7
    pmaddubsw  m2, m7
    pmaddubsw  m3, m7
    %xdefine fenc_buf fenc_buf+gprsize
    psubw      m0, [fenc_buf+0x00]
    psubw      m1, [fenc_buf+0x10]
    psubw      m2, [fenc_buf+0x20]
.satd_8x4b:
    psubw      m3, [fenc_buf+0x30]
    SATD_8x4_SSE 0, 0, 1, 2, 3, 4, 5, 0, swap
    pmaddwd    m0, [pw_1]
    MOVHL      m1, m0
    paddd    xmm0, m0, m1
    ret
%endif ; ARCH
%endmacro ; INTRA_X9

%macro INTRA8_X9 0
;-----------------------------------------------------------------------------
; int intra_sad_x9_8x8( uint8_t *fenc, uint8_t *fdec, uint8_t edge[36], uint16_t *bitcosts, uint16_t *satds )
;-----------------------------------------------------------------------------
cglobal intra_sad_x9_8x8, 5,6,9
    %define fenc02 m4
    %define fenc13 m5
    %define fenc46 m6
    %define fenc57 m7
%if ARCH_X86_64
    %define tmp m8
    %assign padbase 0x0
%else
    %define tmp [rsp]
    %assign padbase 0x10
%endif
    %assign pad 0x240+0x10+padbase-gprsize-(stack_offset&15)
    %define pred(i,j) [rsp+i*0x40+j*0x10+padbase]

    SUB        rsp, pad
    movq    fenc02, [r0+FENC_STRIDE* 0]
    movq    fenc13, [r0+FENC_STRIDE* 1]
    movq    fenc46, [r0+FENC_STRIDE* 4]
    movq    fenc57, [r0+FENC_STRIDE* 5]
    movhps  fenc02, [r0+FENC_STRIDE* 2]
    movhps  fenc13, [r0+FENC_STRIDE* 3]
    movhps  fenc46, [r0+FENC_STRIDE* 6]
    movhps  fenc57, [r0+FENC_STRIDE* 7]

    ; save instruction size: avoid 4-byte memory offsets
    lea         r0, [intra8x9_h1+128]
    %define off(m) (r0+m-(intra8x9_h1+128))

; v
    movddup     m0, [r2+16]
    mova pred(0,0), m0
    psadbw      m1, m0, fenc02
    mova pred(0,1), m0
    psadbw      m2, m0, fenc13
    mova pred(0,2), m0
    psadbw      m3, m0, fenc46
    mova pred(0,3), m0
    psadbw      m0, m0, fenc57
    paddw       m1, m2
    paddw       m0, m3
    paddw       m0, m1
    MOVHL       m1, m0
    paddw       m0, m1
    movd    [r4+0], m0

; h
    movq        m0, [r2+7]
    pshufb      m1, m0, [off(intra8x9_h1)]
    pshufb      m2, m0, [off(intra8x9_h2)]
    mova pred(1,0), m1
    psadbw      m1, fenc02
    mova pred(1,1), m2
    psadbw      m2, fenc13
    paddw       m1, m2
    pshufb      m3, m0, [off(intra8x9_h3)]
    pshufb      m2, m0, [off(intra8x9_h4)]
    mova pred(1,2), m3
    psadbw      m3, fenc46
    mova pred(1,3), m2
    psadbw      m2, fenc57
    paddw       m1, m3
    paddw       m1, m2
    MOVHL       m2, m1
    paddw       m1, m2
    movd    [r4+2], m1

    lea         r5, [rsp+padbase+0x100]
    %define pred(i,j) [r5+i*0x40+j*0x10-0x100]

; dc
    movhps      m0, [r2+16]
    pxor        m2, m2
    psadbw      m0, m2
    MOVHL       m1, m0
    paddw       m0, m1
    psrlw       m0, 3
    pavgw       m0, m2
    pshufb      m0, m2
    mova pred(2,0), m0
    psadbw      m1, m0, fenc02
    mova pred(2,1), m0
    psadbw      m2, m0, fenc13
    mova pred(2,2), m0
    psadbw      m3, m0, fenc46
    mova pred(2,3), m0
    psadbw      m0, m0, fenc57
    paddw       m1, m2
    paddw       m0, m3
    paddw       m0, m1
    MOVHL       m1, m0
    paddw       m0, m1
    movd    [r4+4], m0

; ddl
; Ft1 Ft2 Ft3 Ft4 Ft5 Ft6 Ft7 Ft8
; Ft2 Ft3 Ft4 Ft5 Ft6 Ft7 Ft8 Ft9
; Ft3 Ft4 Ft5 Ft6 Ft7 Ft8 Ft9 FtA
; Ft4 Ft5 Ft6 Ft7 Ft8 Ft9 FtA FtB
; Ft5 Ft6 Ft7 Ft8 Ft9 FtA FtB FtC
; Ft6 Ft7 Ft8 Ft9 FtA FtB FtC FtD
; Ft7 Ft8 Ft9 FtA FtB FtC FtD FtE
; Ft8 Ft9 FtA FtB FtC FtD FtE FtF
    mova        m0, [r2+16]
    movu        m2, [r2+17]
    pslldq      m1, m0, 1
    pavgb       m3, m0, m2              ; Gt1 Gt2 Gt3 Gt4 Gt5 Gt6 Gt7 Gt8 Gt9 GtA GtB ___ ___ ___ ___ ___
    PRED4x4_LOWPASS m0, m1, m2, m0, tmp ; ___ Ft1 Ft2 Ft3 Ft4 Ft5 Ft6 Ft7 Ft8 Ft9 FtA FtB FtC FtD FtE FtF
    pshufb      m1, m0, [off(intra8x9_ddl1)]
    pshufb      m2, m0, [off(intra8x9_ddl2)]
    mova pred(3,0), m1
    psadbw      m1, fenc02
    mova pred(3,1), m2
    psadbw      m2, fenc13
    paddw       m1, m2
    pshufb      m2, m0, [off(intra8x9_ddl3)]
    mova pred(3,2), m2
    psadbw      m2, fenc46
    paddw       m1, m2
    pshufb      m2, m0, [off(intra8x9_ddl4)]
    mova pred(3,3), m2
    psadbw      m2, fenc57
    paddw       m1, m2
    MOVHL       m2, m1
    paddw       m1, m2
    movd    [r4+6], m1

; vl
; Gt1 Gt2 Gt3 Gt4 Gt5 Gt6 Gt7 Gt8
; Ft1 Ft2 Ft3 Ft4 Ft5 Ft6 Ft7 Ft8
; Gt2 Gt3 Gt4 Gt5 Gt6 Gt7 Gt8 Gt9
; Ft2 Ft3 Ft4 Ft5 Ft6 Ft7 Ft8 Ft9
; Gt3 Gt4 Gt5 Gt6 Gt7 Gt8 Gt9 GtA
; Ft3 Ft4 Ft5 Ft6 Ft7 Ft8 Ft9 FtA
; Gt4 Gt5 Gt6 Gt7 Gt8 Gt9 GtA GtB
; Ft4 Ft5 Ft6 Ft7 Ft8 Ft9 FtA FtB
    pshufb      m1, m3, [off(intra8x9_vl1)]
    pshufb      m2, m0, [off(intra8x9_vl2)]
    pshufb      m3, m3, [off(intra8x9_vl3)]
    pshufb      m0, m0, [off(intra8x9_vl4)]
    mova pred(7,0), m1
    psadbw      m1, fenc02
    mova pred(7,1), m2
    psadbw      m2, fenc13
    mova pred(7,2), m3
    psadbw      m3, fenc46
    mova pred(7,3), m0
    psadbw      m0, fenc57
    paddw       m1, m2
    paddw       m0, m3
    paddw       m0, m1
    MOVHL       m1, m0
    paddw       m0, m1
%if cpuflag(sse4)
    pextrw [r4+14], m0, 0
%else
    movd       r5d, m0
    mov    [r4+14], r5w
    lea         r5, [rsp+padbase+0x100]
%endif

; ddr
; Flt Ft0 Ft1 Ft2 Ft3 Ft4 Ft5 Ft6
; Fl0 Flt Ft0 Ft1 Ft2 Ft3 Ft4 Ft5
; Fl1 Fl0 Flt Ft0 Ft1 Ft2 Ft3 Ft4
; Fl2 Fl1 Fl0 Flt Ft0 Ft1 Ft2 Ft3
; Fl3 Fl2 Fl1 Fl0 Flt Ft0 Ft1 Ft2
; Fl4 Fl3 Fl2 Fl1 Fl0 Flt Ft0 Ft1
; Fl5 Fl4 Fl3 Fl2 Fl1 Fl0 Flt Ft0
; Fl6 Fl5 Fl4 Fl3 Fl2 Fl1 Fl0 Flt
    movu        m2, [r2+8]
    movu        m0, [r2+7]
    movu        m1, [r2+6]
    pavgb       m3, m2, m0              ; Gl6 Gl5 Gl4 Gl3 Gl2 Gl1 Gl0 Glt Gt0 Gt1 Gt2 Gt3 Gt4 Gt5 Gt6 Gt7
    PRED4x4_LOWPASS m0, m1, m2, m0, tmp ; Fl7 Fl6 Fl5 Fl4 Fl3 Fl2 Fl1 Fl0 Flt Ft0 Ft1 Ft2 Ft3 Ft4 Ft5 Ft6
    pshufb      m1, m0, [off(intra8x9_ddr1)]
    pshufb      m2, m0, [off(intra8x9_ddr2)]
    mova pred(4,0), m1
    psadbw      m1, fenc02
    mova pred(4,1), m2
    psadbw      m2, fenc13
    paddw       m1, m2
    pshufb      m2, m0, [off(intra8x9_ddr3)]
    mova pred(4,2), m2
    psadbw      m2, fenc46
    paddw       m1, m2
    pshufb      m2, m0, [off(intra8x9_ddr4)]
    mova pred(4,3), m2
    psadbw      m2, fenc57
    paddw       m1, m2
    MOVHL       m2, m1
    paddw       m1, m2
    movd    [r4+8], m1

    add         r0, 256
    add         r5, 0xC0
    %define off(m) (r0+m-(intra8x9_h1+256+128))
    %define pred(i,j) [r5+i*0x40+j*0x10-0x1C0]

; vr
; Gt0 Gt1 Gt2 Gt3 Gt4 Gt5 Gt6 Gt7
; Flt Ft0 Ft1 Ft2 Ft3 Ft4 Ft5 Ft6
; Fl0 Gt0 Gt1 Gt2 Gt3 Gt4 Gt5 Gt6
; Fl1 Flt Ft0 Ft1 Ft2 Ft3 Ft4 Ft5
; Fl2 Fl0 Gt0 Gt1 Gt2 Gt3 Gt4 Gt5
; Fl3 Fl1 Flt Ft0 Ft1 Ft2 Ft3 Ft4
; Fl4 Fl2 Fl0 Gt0 Gt1 Gt2 Gt3 Gt4
; Fl5 Fl3 Fl1 Flt Ft0 Ft1 Ft2 Ft3
    movsd       m2, m3, m0 ; Fl7 Fl6 Fl5 Fl4 Fl3 Fl2 Fl1 Fl0 Gt0 Gt1 Gt2 Gt3 Gt4 Gt5 Gt6 Gt7
    pshufb      m1, m2, [off(intra8x9_vr1)]
    pshufb      m2, m2, [off(intra8x9_vr3)]
    mova pred(5,0), m1
    psadbw      m1, fenc02
    mova pred(5,2), m2
    psadbw      m2, fenc46
    paddw       m1, m2
    pshufb      m2, m0, [off(intra8x9_vr2)]
    mova pred(5,1), m2
    psadbw      m2, fenc13
    paddw       m1, m2
    pshufb      m2, m0, [off(intra8x9_vr4)]
    mova pred(5,3), m2
    psadbw      m2, fenc57
    paddw       m1, m2
    MOVHL       m2, m1
    paddw       m1, m2
    movd   [r4+10], m1

; hd
; Glt Flt Ft0 Ft1 Ft2 Ft3 Ft4 Ft5
; Gl0 Fl0 Glt Flt Ft0 Ft1 Ft2 Ft3
; Gl1 Fl1 Gl0 Fl0 Glt Flt Ft0 Ft1
; Gl2 Fl2 Gl1 Fl1 Gl0 Fl0 Glt Flt
; Gl3 Fl3 Gl2 Fl2 Gl1 Fl1 Gl0 Fl0
; Gl4 Fl4 Gl3 Fl3 Gl2 Fl2 Gl1 Fl1
; Gl5 Fl5 Gl4 Fl4 Gl3 Fl3 Gl2 Fl2
; Gl6 Fl6 Gl5 Fl5 Gl4 Fl4 Gl3 Fl3
    pshufd      m2, m3, q0001
%if cpuflag(sse4)
    pblendw     m2, m0, q3330 ; Gl2 Gl1 Gl0 Glt ___ Fl2 Fl1 Fl0 Flt Ft0 Ft1 Ft2 Ft3 Ft4 Ft5 ___
%else
    movss       m1, m0, m2
    SWAP        1, 2
%endif
    punpcklbw   m0, m3        ; Fl7 Gl6 Fl6 Gl5 Fl5 Gl4 Fl4 Gl3 Fl3 Gl2 Fl2 Gl1 Fl1 Gl0 Fl0 ___
    pshufb      m1, m2, [off(intra8x9_hd1)]
    pshufb      m2, m2, [off(intra8x9_hd2)]
    mova pred(6,0), m1
    psadbw      m1, fenc02
    mova pred(6,1), m2
    psadbw      m2, fenc13
    paddw       m1, m2
    pshufb      m2, m0, [off(intra8x9_hd3)]
    pshufb      m3, m0, [off(intra8x9_hd4)]
    mova pred(6,2), m2
    psadbw      m2, fenc46
    mova pred(6,3), m3
    psadbw      m3, fenc57
    paddw       m1, m2
    paddw       m1, m3
    MOVHL       m2, m1
    paddw       m1, m2
    ; don't just store to [r4+12]. this is too close to the load of dqword [r4] and would cause a forwarding stall
    pslldq      m1, 12
    SWAP        3, 1

; hu
; Gl0 Fl1 Gl1 Fl2 Gl2 Fl3 Gl3 Fl4
; Gl1 Fl2 Gl2 Fl3 Gl3 Fl4 Gl4 Fl5
; Gl2 Fl3 Gl3 Gl3 Gl4 Fl5 Gl5 Fl6
; Gl3 Gl3 Gl4 Fl5 Gl5 Fl6 Gl6 Fl7
; Gl4 Fl5 Gl5 Fl6 Gl6 Fl7 Gl7 Gl7
; Gl5 Fl6 Gl6 Fl7 Gl7 Gl7 Gl7 Gl7
; Gl6 Fl7 Gl7 Gl7 Gl7 Gl7 Gl7 Gl7
; Gl7 Gl7 Gl7 Gl7 Gl7 Gl7 Gl7 Gl7
%if cpuflag(sse4)
    pinsrb      m0, [r2+7], 15 ; Gl7
%else
    movd        m1, [r2+7]
    pslldq      m0, 1
    palignr     m1, m0, 1
    SWAP        0, 1
%endif
    pshufb      m1, m0, [off(intra8x9_hu1)]
    pshufb      m2, m0, [off(intra8x9_hu2)]
    mova pred(8,0), m1
    psadbw      m1, fenc02
    mova pred(8,1), m2
    psadbw      m2, fenc13
    paddw       m1, m2
    pshufb      m2, m0, [off(intra8x9_hu3)]
    pshufb      m0, m0, [off(intra8x9_hu4)]
    mova pred(8,2), m2
    psadbw      m2, fenc46
    mova pred(8,3), m0
    psadbw      m0, fenc57
    paddw       m1, m2
    paddw       m1, m0
    MOVHL       m2, m1
    paddw       m1, m2
    movd       r2d, m1

    movu        m0, [r3]
    por         m3, [r4]
    paddw       m0, m3
    mova      [r4], m0
    movzx      r5d, word [r3+16]
    add        r2d, r5d
    mov    [r4+16], r2w

%if cpuflag(sse4)
    phminposuw m0, m0 ; v,h,dc,ddl,ddr,vr,hd,vl
    movd      eax, m0
%else
    ; 8x8 sad is up to 14 bits; +bitcosts and saturate -> 14 bits; pack with 2 bit index
    paddusw    m0, m0
    paddusw    m0, m0
    paddw      m0, [off(pw_s00112233)]
    MOVHL      m1, m0
    pminsw     m0, m1
    pshuflw    m1, m0, q0032
    pminsw     m0, m1
    movd      eax, m0
    ; repack with 3 bit index
    xor       eax, 0x80008000
    movzx     r3d, ax
    shr       eax, 15
    add       r3d, r3d
    or        eax, 1
    cmp       eax, r3d
    cmovg     eax, r3d
    ; reverse to phminposuw order
    mov       r3d, eax
    and       eax, 7
    shr       r3d, 3
    shl       eax, 16
    or        eax, r3d
%endif
    add       r2d, 8<<16
    cmp        ax, r2w
    cmovg     eax, r2d

    mov       r2d, eax
    shr       r2d, 16
    shl       r2d, 6
    add        r1, 4*FDEC_STRIDE
    mova       m0, [rsp+padbase+r2+0x00]
    mova       m1, [rsp+padbase+r2+0x10]
    mova       m2, [rsp+padbase+r2+0x20]
    mova       m3, [rsp+padbase+r2+0x30]
    movq   [r1+FDEC_STRIDE*-4], m0
    movhps [r1+FDEC_STRIDE*-2], m0
    movq   [r1+FDEC_STRIDE*-3], m1
    movhps [r1+FDEC_STRIDE*-1], m1
    movq   [r1+FDEC_STRIDE* 0], m2
    movhps [r1+FDEC_STRIDE* 2], m2
    movq   [r1+FDEC_STRIDE* 1], m3
    movhps [r1+FDEC_STRIDE* 3], m3
    ADD       rsp, pad
    RET

%if ARCH_X86_64
;-----------------------------------------------------------------------------
; int intra_sa8d_x9_8x8( uint8_t *fenc, uint8_t *fdec, uint8_t edge[36], uint16_t *bitcosts, uint16_t *satds )
;-----------------------------------------------------------------------------
cglobal intra_sa8d_x9_8x8, 5,6,16
    %assign pad 0x2c0+0x10-gprsize-(stack_offset&15)
    %define fenc_buf rsp
    %define pred_buf rsp+0x80
    SUB        rsp, pad
    mova       m15, [hmul_8p]
    pxor        m8, m8
%assign %%i 0
%rep 8
    movddup     m %+ %%i, [r0+%%i*FENC_STRIDE]
    pmaddubsw   m9, m %+ %%i, m15
    punpcklbw   m %+ %%i, m8
    mova [fenc_buf+%%i*0x10], m9
%assign %%i %%i+1
%endrep

    ; save instruction size: avoid 4-byte memory offsets
    lea         r0, [intra8x9_h1+0x80]
    %define off(m) (r0+m-(intra8x9_h1+0x80))
    lea         r5, [pred_buf+0x80]

; v, h, dc
    HADAMARD8_2D 0, 1, 2, 3, 4, 5, 6, 7, 8
    pabsw      m11, m1
%assign %%i 2
%rep 6
    pabsw       m8, m %+ %%i
    paddw      m11, m8
%assign %%i %%i+1
%endrep

    ; 1D hadamard of edges
    movq        m8, [r2+7]
    movddup     m9, [r2+16]
    mova [r5-0x80], m9
    mova [r5-0x70], m9
    mova [r5-0x60], m9
    mova [r5-0x50], m9
    punpcklwd   m8, m8
    pshufb      m9, [intrax3_shuf]
    pmaddubsw   m8, [pb_pppm]
    pmaddubsw   m9, [pb_pppm]
    HSUMSUB2 psignw, m8, m9, m12, m13, m9, q1032, [pw_ppppmmmm]
    HSUMSUB2 psignw, m8, m9, m12, m13, m9, q2301, [pw_ppmmppmm]

    ; dc
    paddw      m10, m8, m9
    paddw      m10, [pw_8]
    pand       m10, [sw_f0]
    psrlw      m12, m10, 4
    psllw      m10, 2
    pxor       m13, m13
    pshufb     m12, m13
    mova [r5+0x00], m12
    mova [r5+0x10], m12
    mova [r5+0x20], m12
    mova [r5+0x30], m12

    ; differences
    psllw       m8, 3 ; left edge
    psubw       m8, m0
    psubw      m10, m0
    pabsw       m8, m8 ; 1x8 sum
    pabsw      m10, m10
    paddw       m8, m11
    paddw      m11, m10
    punpcklwd   m0, m1
    punpcklwd   m2, m3
    punpcklwd   m4, m5
    punpcklwd   m6, m7
    punpckldq   m0, m2
    punpckldq   m4, m6
    punpcklqdq  m0, m4 ; transpose
    psllw       m9, 3  ; top edge
    psrldq     m10, m11, 2 ; 8x7 sum
    psubw       m0, m9 ; 8x1 sum
    pabsw       m0, m0
    paddw      m10, m0

    phaddd     m10, m8 ; logically phaddw, but this is faster and it won't overflow
    psrlw      m11, 1
    psrlw      m10, 1

; store h
    movq        m3, [r2+7]
    pshufb      m0, m3, [off(intra8x9_h1)]
    pshufb      m1, m3, [off(intra8x9_h2)]
    pshufb      m2, m3, [off(intra8x9_h3)]
    pshufb      m3, m3, [off(intra8x9_h4)]
    mova [r5-0x40], m0
    mova [r5-0x30], m1
    mova [r5-0x20], m2
    mova [r5-0x10], m3

; ddl
    mova        m8, [r2+16]
    movu        m2, [r2+17]
    pslldq      m1, m8, 1
    pavgb       m9, m8, m2
    PRED4x4_LOWPASS m8, m1, m2, m8, m3
    pshufb      m0, m8, [off(intra8x9_ddl1)]
    pshufb      m1, m8, [off(intra8x9_ddl2)]
    pshufb      m2, m8, [off(intra8x9_ddl3)]
    pshufb      m3, m8, [off(intra8x9_ddl4)]
    add         r5, 0x40
    call .sa8d
    phaddd     m11, m0

; vl
    pshufb      m0, m9, [off(intra8x9_vl1)]
    pshufb      m1, m8, [off(intra8x9_vl2)]
    pshufb      m2, m9, [off(intra8x9_vl3)]
    pshufb      m3, m8, [off(intra8x9_vl4)]
    add         r5, 0x100
    call .sa8d
    phaddd     m10, m11
    mova       m12, m0

; ddr
    movu        m2, [r2+8]
    movu        m8, [r2+7]
    movu        m1, [r2+6]
    pavgb       m9, m2, m8
    PRED4x4_LOWPASS m8, m1, m2, m8, m3
    pshufb      m0, m8, [off(intra8x9_ddr1)]
    pshufb      m1, m8, [off(intra8x9_ddr2)]
    pshufb      m2, m8, [off(intra8x9_ddr3)]
    pshufb      m3, m8, [off(intra8x9_ddr4)]
    sub         r5, 0xc0
    call .sa8d
    mova       m11, m0

    add         r0, 0x100
    %define off(m) (r0+m-(intra8x9_h1+0x180))

; vr
    movsd       m2, m9, m8
    pshufb      m0, m2, [off(intra8x9_vr1)]
    pshufb      m1, m8, [off(intra8x9_vr2)]
    pshufb      m2, m2, [off(intra8x9_vr3)]
    pshufb      m3, m8, [off(intra8x9_vr4)]
    add         r5, 0x40
    call .sa8d
    phaddd     m11, m0

; hd
%if cpuflag(sse4)
    pshufd      m1, m9, q0001
    pblendw     m1, m8, q3330
%else
    pshufd      m2, m9, q0001
    movss       m1, m8, m2
%endif
    punpcklbw   m8, m9
    pshufb      m0, m1, [off(intra8x9_hd1)]
    pshufb      m1, m1, [off(intra8x9_hd2)]
    pshufb      m2, m8, [off(intra8x9_hd3)]
    pshufb      m3, m8, [off(intra8x9_hd4)]
    add         r5, 0x40
    call .sa8d
    phaddd      m0, m12
    phaddd     m11, m0

; hu
%if cpuflag(sse4)
    pinsrb      m8, [r2+7], 15
%else
    movd        m9, [r2+7]
    pslldq      m8, 1
    palignr     m9, m8, 1
    SWAP        8, 9
%endif
    pshufb      m0, m8, [off(intra8x9_hu1)]
    pshufb      m1, m8, [off(intra8x9_hu2)]
    pshufb      m2, m8, [off(intra8x9_hu3)]
    pshufb      m3, m8, [off(intra8x9_hu4)]
    add         r5, 0x80
    call .sa8d

    pmaddwd     m0, [pw_1]
    phaddw     m10, m11
    MOVHL       m1, m0
    paddw       m0, m1
    pshuflw     m1, m0, q0032
    pavgw       m0, m1
    pxor        m2, m2
    pavgw      m10, m2
    movd       r2d, m0

    movu        m0, [r3]
    paddw       m0, m10
    mova      [r4], m0
    movzx      r5d, word [r3+16]
    add        r2d, r5d
    mov    [r4+16], r2w

%if cpuflag(sse4)
    phminposuw m0, m0
    movd      eax, m0
%else
    ; 8x8 sa8d is up to 15 bits; +bitcosts and saturate -> 15 bits; pack with 1 bit index
    paddusw    m0, m0
    paddw      m0, [off(pw_s00001111)]
    MOVHL      m1, m0
    pminsw     m0, m1
    pshuflw    m1, m0, q0032
    mova       m2, m0
    pminsw     m0, m1
    pcmpgtw    m2, m1 ; 2nd index bit
    movd      r3d, m0
    movd      r4d, m2
    ; repack with 3 bit index
    xor       r3d, 0x80008000
    and       r4d, 0x00020002
    movzx     eax, r3w
    movzx     r5d, r4w
    shr       r3d, 16
    shr       r4d, 16
    lea       eax, [rax*4+r5]
    lea       r3d, [ r3*4+r4+1]
    cmp       eax, r3d
    cmovg     eax, r3d
    ; reverse to phminposuw order
    mov       r3d, eax
    and       eax, 7
    shr       r3d, 3
    shl       eax, 16
    or        eax, r3d
%endif
    add       r2d, 8<<16
    cmp        ax, r2w
    cmovg     eax, r2d

    mov       r2d, eax
    shr       r2d, 16
    shl       r2d, 6
    add        r1, 4*FDEC_STRIDE
    mova       m0, [pred_buf+r2+0x00]
    mova       m1, [pred_buf+r2+0x10]
    mova       m2, [pred_buf+r2+0x20]
    mova       m3, [pred_buf+r2+0x30]
    movq   [r1+FDEC_STRIDE*-4], m0
    movhps [r1+FDEC_STRIDE*-2], m0
    movq   [r1+FDEC_STRIDE*-3], m1
    movhps [r1+FDEC_STRIDE*-1], m1
    movq   [r1+FDEC_STRIDE* 0], m2
    movhps [r1+FDEC_STRIDE* 2], m2
    movq   [r1+FDEC_STRIDE* 1], m3
    movhps [r1+FDEC_STRIDE* 3], m3
    ADD       rsp, pad
    RET

ALIGN 16
.sa8d:
    %xdefine mret m0
    %xdefine fenc_buf fenc_buf+gprsize
    mova [r5+0x00], m0
    mova [r5+0x10], m1
    mova [r5+0x20], m2
    mova [r5+0x30], m3
    movddup     m4, m0
    movddup     m5, m1
    movddup     m6, m2
    movddup     m7, m3
    punpckhqdq  m0, m0
    punpckhqdq  m1, m1
    punpckhqdq  m2, m2
    punpckhqdq  m3, m3
    PERMUTE 0,4, 1,5, 2,0, 3,1, 4,6, 5,7, 6,2, 7,3
    pmaddubsw   m0, m15
    pmaddubsw   m1, m15
    psubw       m0, [fenc_buf+0x00]
    psubw       m1, [fenc_buf+0x10]
    pmaddubsw   m2, m15
    pmaddubsw   m3, m15
    psubw       m2, [fenc_buf+0x20]
    psubw       m3, [fenc_buf+0x30]
    pmaddubsw   m4, m15
    pmaddubsw   m5, m15
    psubw       m4, [fenc_buf+0x40]
    psubw       m5, [fenc_buf+0x50]
    pmaddubsw   m6, m15
    pmaddubsw   m7, m15
    psubw       m6, [fenc_buf+0x60]
    psubw       m7, [fenc_buf+0x70]
    HADAMARD8_2D_HMUL 0, 1, 2, 3, 4, 5, 6, 7, 13, 14
    paddw       m0, m1
    paddw       m0, m2
    paddw mret, m0, m3
    ret
%endif ; ARCH_X86_64
%endmacro ; INTRA8_X9

; in:  r0=pix, r1=stride, r2=stride*3, r3=tmp, m6=mask_ac4, m7=0
; out: [tmp]=hadamard4, m0=satd
INIT_MMX mmx2
cglobal hadamard_ac_4x4
%if HIGH_BIT_DEPTH
    mova      m0, [r0]
    mova      m1, [r0+r1]
    mova      m2, [r0+r1*2]
    mova      m3, [r0+r2]
%else ; !HIGH_BIT_DEPTH
    movh      m0, [r0]
    movh      m1, [r0+r1]
    movh      m2, [r0+r1*2]
    movh      m3, [r0+r2]
    punpcklbw m0, m7
    punpcklbw m1, m7
    punpcklbw m2, m7
    punpcklbw m3, m7
%endif ; HIGH_BIT_DEPTH
    HADAMARD4_2D 0, 1, 2, 3, 4
    mova [r3],    m0
    mova [r3+8],  m1
    mova [r3+16], m2
    mova [r3+24], m3
    ABSW      m0, m0, m4
    ABSW      m1, m1, m4
    pand      m0, m6
    ABSW      m2, m2, m4
    ABSW      m3, m3, m4
    paddw     m0, m1
    paddw     m2, m3
    paddw     m0, m2
    SAVE_MM_PERMUTATION
    ret

cglobal hadamard_ac_2x2max
    mova      m0, [r3+0x00]
    mova      m1, [r3+0x20]
    mova      m2, [r3+0x40]
    mova      m3, [r3+0x60]
    sub       r3, 8
    SUMSUB_BADC w, 0, 1, 2, 3, 4
    ABSW2 m0, m2, m0, m2, m4, m5
    ABSW2 m1, m3, m1, m3, m4, m5
    HADAMARD 0, max, 0, 2, 4, 5
    HADAMARD 0, max, 1, 3, 4, 5
%if HIGH_BIT_DEPTH
    pmaddwd   m0, m7
    pmaddwd   m1, m7
    paddd     m6, m0
    paddd     m6, m1
%else ; !HIGH_BIT_DEPTH
    paddw     m7, m0
    paddw     m7, m1
%endif ; HIGH_BIT_DEPTH
    SAVE_MM_PERMUTATION
    ret

%macro AC_PREP 2
%if HIGH_BIT_DEPTH
    pmaddwd %1, %2
%endif
%endmacro

%macro AC_PADD 3
%if HIGH_BIT_DEPTH
    AC_PREP %2, %3
    paddd   %1, %2
%else
    paddw   %1, %2
%endif ; HIGH_BIT_DEPTH
%endmacro

cglobal hadamard_ac_8x8
    mova      m6, [mask_ac4]
%if HIGH_BIT_DEPTH
    mova      m7, [pw_1]
%else
    pxor      m7, m7
%endif ; HIGH_BIT_DEPTH
    call hadamard_ac_4x4_mmx2
    add       r0, 4*SIZEOF_PIXEL
    add       r3, 32
    mova      m5, m0
    AC_PREP   m5, m7
    call hadamard_ac_4x4_mmx2
    lea       r0, [r0+4*r1]
    add       r3, 64
    AC_PADD   m5, m0, m7
    call hadamard_ac_4x4_mmx2
    sub       r0, 4*SIZEOF_PIXEL
    sub       r3, 32
    AC_PADD   m5, m0, m7
    call hadamard_ac_4x4_mmx2
    AC_PADD   m5, m0, m7
    sub       r3, 40
    mova [rsp+gprsize+8], m5 ; save satd
%if HIGH_BIT_DEPTH
    pxor      m6, m6
%endif
%rep 3
    call hadamard_ac_2x2max_mmx2
%endrep
    mova      m0, [r3+0x00]
    mova      m1, [r3+0x20]
    mova      m2, [r3+0x40]
    mova      m3, [r3+0x60]
    SUMSUB_BADC w, 0, 1, 2, 3, 4
    HADAMARD 0, sumsub, 0, 2, 4, 5
    ABSW2 m1, m3, m1, m3, m4, m5
    ABSW2 m0, m2, m0, m2, m4, m5
    HADAMARD 0, max, 1, 3, 4, 5
%if HIGH_BIT_DEPTH
    pand      m0, [mask_ac4]
    pmaddwd   m1, m7
    pmaddwd   m0, m7
    pmaddwd   m2, m7
    paddd     m6, m1
    paddd     m0, m2
    paddd     m6, m6
    paddd     m0, m6
    SWAP       0,  6
%else ; !HIGH_BIT_DEPTH
    pand      m6, m0
    paddw     m7, m1
    paddw     m6, m2
    paddw     m7, m7
    paddw     m6, m7
%endif ; HIGH_BIT_DEPTH
    mova [rsp+gprsize], m6 ; save sa8d
    SWAP       0,  6
    SAVE_MM_PERMUTATION
    ret

%macro HADAMARD_AC_WXH_SUM_MMX 2
    mova    m1, [rsp+1*mmsize]
%if HIGH_BIT_DEPTH
%if %1*%2 >= 128
    paddd   m0, [rsp+2*mmsize]
    paddd   m1, [rsp+3*mmsize]
%endif
%if %1*%2 == 256
    mova    m2, [rsp+4*mmsize]
    paddd   m1, [rsp+5*mmsize]
    paddd   m2, [rsp+6*mmsize]
    mova    m3, m0
    paddd   m1, [rsp+7*mmsize]
    paddd   m0, m2
%endif
    psrld   m0, 1
    HADDD   m0, m2
    psrld   m1, 1
    HADDD   m1, m3
%else ; !HIGH_BIT_DEPTH
%if %1*%2 >= 128
    paddusw m0, [rsp+2*mmsize]
    paddusw m1, [rsp+3*mmsize]
%endif
%if %1*%2 == 256
    mova    m2, [rsp+4*mmsize]
    paddusw m1, [rsp+5*mmsize]
    paddusw m2, [rsp+6*mmsize]
    mova    m3, m0
    paddusw m1, [rsp+7*mmsize]
    pxor    m3, m2
    pand    m3, [pw_1]
    pavgw   m0, m2
    psubusw m0, m3
    HADDUW  m0, m2
%else
    psrlw   m0, 1
    HADDW   m0, m2
%endif
    psrlw   m1, 1
    HADDW   m1, m3
%endif ; HIGH_BIT_DEPTH
%endmacro

%macro HADAMARD_AC_WXH_MMX 2
cglobal pixel_hadamard_ac_%1x%2, 2,4
    %assign pad 16-gprsize-(stack_offset&15)
    %define ysub r1
    FIX_STRIDES r1
    sub  rsp, 16+128+pad
    lea  r2, [r1*3]
    lea  r3, [rsp+16]
    call hadamard_ac_8x8_mmx2
%if %2==16
    %define ysub r2
    lea  r0, [r0+r1*4]
    sub  rsp, 16
    call hadamard_ac_8x8_mmx2
%endif
%if %1==16
    neg  ysub
    sub  rsp, 16
    lea  r0, [r0+ysub*4+8*SIZEOF_PIXEL]
    neg  ysub
    call hadamard_ac_8x8_mmx2
%if %2==16
    lea  r0, [r0+r1*4]
    sub  rsp, 16
    call hadamard_ac_8x8_mmx2
%endif
%endif
    HADAMARD_AC_WXH_SUM_MMX %1, %2
    movd edx, m0
    movd eax, m1
    shr  edx, 1
%if ARCH_X86_64
    shl  rdx, 32
    add  rax, rdx
%endif
    add  rsp, 128+%1*%2/4+pad
    RET
%endmacro ; HADAMARD_AC_WXH_MMX

HADAMARD_AC_WXH_MMX 16, 16
HADAMARD_AC_WXH_MMX  8, 16
HADAMARD_AC_WXH_MMX 16,  8
HADAMARD_AC_WXH_MMX  8,  8

%macro LOAD_INC_8x4W_SSE2 5
%if HIGH_BIT_DEPTH
    movu      m%1, [r0]
    movu      m%2, [r0+r1]
    movu      m%3, [r0+r1*2]
    movu      m%4, [r0+r2]
%ifidn %1, 0
    lea       r0, [r0+r1*4]
%endif
%else ; !HIGH_BIT_DEPTH
    movh      m%1, [r0]
    movh      m%2, [r0+r1]
    movh      m%3, [r0+r1*2]
    movh      m%4, [r0+r2]
%ifidn %1, 0
    lea       r0, [r0+r1*4]
%endif
    punpcklbw m%1, m%5
    punpcklbw m%2, m%5
    punpcklbw m%3, m%5
    punpcklbw m%4, m%5
%endif ; HIGH_BIT_DEPTH
%endmacro

%macro LOAD_INC_8x4W_SSSE3 5
    LOAD_DUP_4x8P %3, %4, %1, %2, [r0+r1*2], [r0+r2], [r0], [r0+r1]
%ifidn %1, 0
    lea       r0, [r0+r1*4]
%endif
    HSUMSUB %1, %2, %3, %4, %5
%endmacro

%macro HADAMARD_AC_SSE2 0
; in:  r0=pix, r1=stride, r2=stride*3
; out: [esp+16]=sa8d, [esp+32]=satd, r0+=stride*4
cglobal hadamard_ac_8x8
%if ARCH_X86_64
    %define spill0 m8
    %define spill1 m9
    %define spill2 m10
%else
    %define spill0 [rsp+gprsize]
    %define spill1 [rsp+gprsize+mmsize]
    %define spill2 [rsp+gprsize+mmsize*2]
%endif
%if HIGH_BIT_DEPTH
    %define vertical 1
%elif cpuflag(ssse3) && notcpuflag(atom)
    %define vertical 0
    ;LOAD_INC loads sumsubs
    mova      m7, [hmul_8p]
%else
    %define vertical 1
    ;LOAD_INC only unpacks to words
    pxor      m7, m7
%endif
    LOAD_INC_8x4W 0, 1, 2, 3, 7
%if vertical
    HADAMARD4_2D_SSE 0, 1, 2, 3, 4
%else
    HADAMARD4_V 0, 1, 2, 3, 4
%endif
    mova  spill0, m1
    SWAP 1, 7
    LOAD_INC_8x4W 4, 5, 6, 7, 1
%if vertical
    HADAMARD4_2D_SSE 4, 5, 6, 7, 1
%else
    HADAMARD4_V 4, 5, 6, 7, 1
    ; FIXME SWAP
    mova      m1, spill0
    mova      spill0, m6
    mova      spill1, m7
    HADAMARD 1, sumsub, 0, 1, 6, 7
    HADAMARD 1, sumsub, 2, 3, 6, 7
    mova      m6, spill0
    mova      m7, spill1
    mova      spill0, m1
    mova      spill1, m0
    HADAMARD 1, sumsub, 4, 5, 1, 0
    HADAMARD 1, sumsub, 6, 7, 1, 0
    mova      m0, spill1
%endif
    mova  spill1, m2
    mova  spill2, m3
    ABSW      m1, m0, m0
    ABSW      m2, m4, m4
    ABSW      m3, m5, m5
    paddw     m1, m2
    SUMSUB_BA w, 0, 4
%if vertical
    pand      m1, [mask_ac4]
%else
    pand      m1, [mask_ac4b]
%endif
    AC_PREP   m1, [pw_1]
    ABSW      m2, spill0
    AC_PADD   m1, m3, [pw_1]
    ABSW      m3, spill1
    AC_PADD   m1, m2, [pw_1]
    ABSW      m2, spill2
    AC_PADD   m1, m3, [pw_1]
    ABSW      m3, m6, m6
    AC_PADD   m1, m2, [pw_1]
    ABSW      m2, m7, m7
    AC_PADD   m1, m3, [pw_1]
    AC_PADD   m1, m2, [pw_1]
    paddw     m3, m7, spill2
    psubw     m7, spill2
    mova  [rsp+gprsize+mmsize*2], m1 ; save satd
    paddw     m2, m6, spill1
    psubw     m6, spill1
    paddw     m1, m5, spill0
    psubw     m5, spill0
    %assign %%x 2
%if vertical
    %assign %%x 4
%endif
    mova  spill1, m4
    HADAMARD %%x, amax, 3, 7, 4
    HADAMARD %%x, amax, 2, 6, 7, 4
    mova      m4, spill1
    HADAMARD %%x, amax, 1, 5, 6, 7
    HADAMARD %%x, sumsub, 0, 4, 5, 6
    AC_PREP   m2, [pw_1]
    AC_PADD   m2, m3, [pw_1]
    AC_PADD   m2, m1, [pw_1]
%if HIGH_BIT_DEPTH
    paddd     m2, m2
%else
    paddw     m2, m2
%endif ; HIGH_BIT_DEPTH
    ABSW      m4, m4, m7
    pand      m0, [mask_ac8]
    ABSW      m0, m0, m7
    AC_PADD   m2, m4, [pw_1]
    AC_PADD   m2, m0, [pw_1]
    mova [rsp+gprsize+mmsize], m2 ; save sa8d
    SWAP       0, 2
    SAVE_MM_PERMUTATION
    ret

HADAMARD_AC_WXH_SSE2 16, 16
HADAMARD_AC_WXH_SSE2 16,  8
%if mmsize <= 16
HADAMARD_AC_WXH_SSE2  8, 16
HADAMARD_AC_WXH_SSE2  8,  8
%endif
%endmacro ; HADAMARD_AC_SSE2

%macro HADAMARD_AC_WXH_SUM_SSE2 2
    mova    m1, [rsp+2*mmsize]
%if HIGH_BIT_DEPTH
%if %1*%2 >= 128
    paddd   m0, [rsp+3*mmsize]
    paddd   m1, [rsp+4*mmsize]
%endif
%if %1*%2 == 256
    paddd   m0, [rsp+5*mmsize]
    paddd   m1, [rsp+6*mmsize]
    paddd   m0, [rsp+7*mmsize]
    paddd   m1, [rsp+8*mmsize]
    psrld   m0, 1
%endif
    HADDD  xm0, xm2
    HADDD  xm1, xm3
%else ; !HIGH_BIT_DEPTH
%if %1*%2*16/mmsize >= 128
    paddusw m0, [rsp+3*mmsize]
    paddusw m1, [rsp+4*mmsize]
%endif
%if %1*%2*16/mmsize == 256
    paddusw m0, [rsp+5*mmsize]
    paddusw m1, [rsp+6*mmsize]
    paddusw m0, [rsp+7*mmsize]
    paddusw m1, [rsp+8*mmsize]
    psrlw   m0, 1
%endif
%if mmsize==32
    vextracti128 xm2, m0, 1
    vextracti128 xm3, m1, 1
    paddusw xm0, xm2
    paddusw xm1, xm3
%endif
    HADDUW xm0, xm2
    HADDW  xm1, xm3
%endif ; HIGH_BIT_DEPTH
%endmacro

; struct { int satd, int sa8d; } pixel_hadamard_ac_16x16( uint8_t *pix, int stride )
%macro HADAMARD_AC_WXH_SSE2 2
cglobal pixel_hadamard_ac_%1x%2, 2,4,11
    %define ysub r1
    FIX_STRIDES r1
    mov   r3, rsp
    and  rsp, ~(mmsize-1)
    sub  rsp, mmsize*3
    lea   r2, [r1*3]
    call hadamard_ac_8x8
%if %2==16
    %define ysub r2
    lea   r0, [r0+r1*4]
    sub  rsp, mmsize*2
    call hadamard_ac_8x8
%endif
%if %1==16 && mmsize <= 16
    neg  ysub
    sub  rsp, mmsize*2
    lea   r0, [r0+ysub*4+8*SIZEOF_PIXEL]
    neg  ysub
    call hadamard_ac_8x8
%if %2==16
    lea   r0, [r0+r1*4]
    sub  rsp, mmsize*2
    call hadamard_ac_8x8
%endif
%endif
    HADAMARD_AC_WXH_SUM_SSE2 %1, %2
    movd edx, xm0
    movd eax, xm1
    shr  edx, 2 - (%1*%2*16/mmsize >> 8)
    shr  eax, 1
%if ARCH_X86_64
    shl  rdx, 32
    add  rax, rdx
%endif
    mov  rsp, r3
    RET
%endmacro ; HADAMARD_AC_WXH_SSE2

; instantiate satds

%if ARCH_X86_64 == 0 && HIGH_BIT_DEPTH == 0
cextern pixel_sa8d_8x8_internal_mmx2
INIT_MMX mmx2
SA8D
%endif

%define TRANS TRANS_SSE2
%define DIFFOP DIFF_UNPACK_SSE2
%define LOAD_INC_8x4W LOAD_INC_8x4W_SSE2
%define LOAD_SUMSUB_8x4P LOAD_DIFF_8x4P
%define LOAD_SUMSUB_16P  LOAD_SUMSUB_16P_SSE2
%define movdqa movaps ; doesn't hurt pre-nehalem, might as well save size
%define movdqu movups
%define punpcklqdq movlhps
INIT_XMM sse2
SA8D
SATDS_SSE2
%if ARCH_X86_64
SA8D_SATD
%endif
%if HIGH_BIT_DEPTH == 0
INTRA_SA8D_SSE2
%endif
INIT_MMX mmx2
INTRA_X3_MMX
INIT_XMM sse2
HADAMARD_AC_SSE2

%if HIGH_BIT_DEPTH == 0
INIT_XMM ssse3,atom
SATDS_SSE2
SA8D
HADAMARD_AC_SSE2
%if ARCH_X86_64
SA8D_SATD
%endif
%endif

%define DIFFOP DIFF_SUMSUB_SSSE3
%define LOAD_DUP_4x8P LOAD_DUP_4x8P_CONROE
%if HIGH_BIT_DEPTH == 0
%define LOAD_INC_8x4W LOAD_INC_8x4W_SSSE3
%define LOAD_SUMSUB_8x4P LOAD_SUMSUB_8x4P_SSSE3
%define LOAD_SUMSUB_16P  LOAD_SUMSUB_16P_SSSE3
%endif
INIT_XMM ssse3
SATDS_SSE2
SA8D
HADAMARD_AC_SSE2
%if ARCH_X86_64
SA8D_SATD
%endif
%if HIGH_BIT_DEPTH == 0
INTRA_X9
INTRA8_X9
%endif
%undef movdqa ; nehalem doesn't like movaps
%undef movdqu ; movups
%undef punpcklqdq ; or movlhps
%if HIGH_BIT_DEPTH == 0
INIT_MMX ssse3
INTRA_X3_MMX
%endif

%define TRANS TRANS_SSE4
%define LOAD_DUP_4x8P LOAD_DUP_4x8P_PENRYN
INIT_XMM sse4
SATDS_SSE2
SA8D
HADAMARD_AC_SSE2
%if ARCH_X86_64
SA8D_SATD
%endif
%if HIGH_BIT_DEPTH == 0
INTRA_X9
INTRA8_X9
%endif

; Sandy/Ivy Bridge and Bulldozer do movddup in the load unit, so
; it's effectively free.
%define LOAD_DUP_4x8P LOAD_DUP_4x8P_CONROE
INIT_XMM avx
SATDS_SSE2
SA8D
%if ARCH_X86_64
SA8D_SATD
%endif
%if HIGH_BIT_DEPTH == 0
INTRA_X9
INTRA8_X9
%endif
HADAMARD_AC_SSE2

%define TRANS TRANS_XOP
INIT_XMM xop
SATDS_SSE2
SA8D
%if ARCH_X86_64
SA8D_SATD
%endif
%if HIGH_BIT_DEPTH == 0
INTRA_X9
; no xop INTRA8_X9. it's slower than avx on bulldozer. dunno why.
%endif
HADAMARD_AC_SSE2


%if HIGH_BIT_DEPTH == 0
%define LOAD_SUMSUB_8x4P LOAD_SUMSUB8_16x4P_AVX2
%define LOAD_DUP_4x8P LOAD_DUP_4x16P_AVX2
%define TRANS TRANS_SSE4
INIT_YMM avx2
HADAMARD_AC_SSE2
%if ARCH_X86_64
SA8D_SATD
%endif

%macro LOAD_SUMSUB_8x8P_AVX2 7 ; 4*dst, 2*tmp, mul]
    movq   xm%1, [r0]
    movq   xm%3, [r2]
    movq   xm%2, [r0+r1]
    movq   xm%4, [r2+r3]
    vinserti128 m%1, m%1, [r0+4*r1], 1
    vinserti128 m%3, m%3, [r2+4*r3], 1
    vinserti128 m%2, m%2, [r0+r4], 1
    vinserti128 m%4, m%4, [r2+r5], 1
    punpcklqdq m%1, m%1
    punpcklqdq m%3, m%3
    punpcklqdq m%2, m%2
    punpcklqdq m%4, m%4
    DIFF_SUMSUB_SSSE3 %1, %3, %2, %4, %7
    lea      r0, [r0+2*r1]
    lea      r2, [r2+2*r3]

    movq   xm%3, [r0]
    movq   xm%5, [r2]
    movq   xm%4, [r0+r1]
    movq   xm%6, [r2+r3]
    vinserti128 m%3, m%3, [r0+4*r1], 1
    vinserti128 m%5, m%5, [r2+4*r3], 1
    vinserti128 m%4, m%4, [r0+r4], 1
    vinserti128 m%6, m%6, [r2+r5], 1
    punpcklqdq m%3, m%3
    punpcklqdq m%5, m%5
    punpcklqdq m%4, m%4
    punpcklqdq m%6, m%6
    DIFF_SUMSUB_SSSE3 %3, %5, %4, %6, %7
%endmacro

%macro SATD_START_AVX2 2-3 0
    FIX_STRIDES r1, r3
%if %3
    mova    %2, [hmul_8p]
    lea     r4, [5*r1]
    lea     r5, [5*r3]
%else
    mova    %2, [hmul_16p]
    lea     r4, [3*r1]
    lea     r5, [3*r3]
%endif
    pxor    %1, %1
%endmacro

%define TRANS TRANS_SSE4
INIT_YMM avx2
cglobal pixel_satd_16x8_internal
    LOAD_SUMSUB_16x4P_AVX2 0, 1, 2, 3, 4, 5, 7, r0, r2, 1
    SATD_8x4_SSE 0, 0, 1, 2, 3, 4, 5, 6
    LOAD_SUMSUB_16x4P_AVX2 0, 1, 2, 3, 4, 5, 7, r0, r2, 0
    SATD_8x4_SSE 0, 0, 1, 2, 3, 4, 5, 6
    ret

cglobal pixel_satd_16x16, 4,6,8
    SATD_START_AVX2 m6, m7
    call pixel_satd_16x8_internal
    lea  r0, [r0+4*r1]
    lea  r2, [r2+4*r3]
pixel_satd_16x8_internal:
    call pixel_satd_16x8_internal
    vextracti128 xm0, m6, 1
    paddw        xm0, xm6
    SATD_END_SSE2 xm0
    RET

cglobal pixel_satd_16x8, 4,6,8
    SATD_START_AVX2 m6, m7
    jmp pixel_satd_16x8_internal

cglobal pixel_satd_8x8_internal
    LOAD_SUMSUB_8x8P_AVX2 0, 1, 2, 3, 4, 5, 7
    SATD_8x4_SSE 0, 0, 1, 2, 3, 4, 5, 6
    ret

cglobal pixel_satd_8x16, 4,6,8
    SATD_START_AVX2 m6, m7, 1
    call pixel_satd_8x8_internal
    lea  r0, [r0+2*r1]
    lea  r2, [r2+2*r3]
    lea  r0, [r0+4*r1]
    lea  r2, [r2+4*r3]
    call pixel_satd_8x8_internal
    vextracti128 xm0, m6, 1
    paddw        xm0, xm6
    SATD_END_SSE2 xm0
    RET

cglobal pixel_satd_8x8, 4,6,8
    SATD_START_AVX2 m6, m7, 1
    call pixel_satd_8x8_internal
    vextracti128 xm0, m6, 1
    paddw        xm0, xm6
    SATD_END_SSE2 xm0
    RET

cglobal pixel_sa8d_8x8_internal
    LOAD_SUMSUB_8x8P_AVX2 0, 1, 2, 3, 4, 5, 7
    HADAMARD4_V 0, 1, 2, 3, 4
    HADAMARD 8, sumsub, 0, 1, 4, 5
    HADAMARD 8, sumsub, 2, 3, 4, 5
    HADAMARD 2, sumsub, 0, 1, 4, 5
    HADAMARD 2, sumsub, 2, 3, 4, 5
    HADAMARD 1, amax, 0, 1, 4, 5
    HADAMARD 1, amax, 2, 3, 4, 5
    paddw  m6, m0
    paddw  m6, m2
    ret

cglobal pixel_sa8d_8x8, 4,6,8
    SATD_START_AVX2 m6, m7, 1
    call pixel_sa8d_8x8_internal
    vextracti128 xm1, m6, 1
    paddw xm6, xm1
    HADDW xm6, xm1
    movd  eax, xm6
    add   eax, 1
    shr   eax, 1
    RET

cglobal intra_sad_x9_8x8, 5,7,8
    %define pred(i,j) [rsp+i*0x40+j*0x20]

    mov         r6, rsp
    and        rsp, ~31
    sub        rsp, 0x240
    movu        m5, [r0+0*FENC_STRIDE]
    movu        m6, [r0+4*FENC_STRIDE]
    punpcklqdq  m5, [r0+2*FENC_STRIDE]
    punpcklqdq  m6, [r0+6*FENC_STRIDE]

    ; save instruction size: avoid 4-byte memory offsets
    lea         r0, [intra8x9_h1+128]
    %define off(m) (r0+m-(intra8x9_h1+128))

    vpbroadcastq m0, [r2+16]
    psadbw      m4, m0, m5
    psadbw      m2, m0, m6
    mova pred(0,0), m0
    mova pred(0,1), m0
    paddw       m4, m2

    vpbroadcastq m1, [r2+7]
    pshufb      m3, m1, [off(intra8x9_h1)]
    pshufb      m2, m1, [off(intra8x9_h3)]
    mova pred(1,0), m3
    mova pred(1,1), m2
    psadbw      m3, m5
    psadbw      m2, m6
    paddw       m3, m2

    lea         r5, [rsp+0x100]
    %define pred(i,j) [r5+i*0x40+j*0x20-0x100]

    ; combine the first two
    pslldq      m3, 2
    por         m4, m3

    pxor        m2, m2
    psadbw      m0, m2
    psadbw      m1, m2
    paddw       m0, m1
    psrlw       m0, 3
    pavgw       m0, m2
    pshufb      m0, m2
    mova pred(2,0), m0
    mova pred(2,1), m0
    psadbw      m3, m0, m5
    psadbw      m2, m0, m6
    paddw       m3, m2

    pslldq      m3, 4
    por         m4, m3

    vbroadcasti128 m0, [r2+16]
    vbroadcasti128 m2, [r2+17]
    pslldq      m1, m0, 1
    pavgb       m3, m0, m2
    PRED4x4_LOWPASS m0, m1, m2, m0, m7
    pshufb      m1, m0, [off(intra8x9_ddl1)]
    pshufb      m2, m0, [off(intra8x9_ddl3)]
    mova pred(3,0), m1
    mova pred(3,1), m2
    psadbw      m1, m5
    psadbw      m2, m6
    paddw       m1, m2

    pslldq      m1, 6
    por         m4, m1
    vextracti128 xm1, m4, 1
    paddw      xm4, xm1
    mova      [r4], xm4

    ; for later
    vinserti128 m7, m3, xm0, 1

    vbroadcasti128 m2, [r2+8]
    vbroadcasti128 m0, [r2+7]
    vbroadcasti128 m1, [r2+6]
    pavgb       m3, m2, m0
    PRED4x4_LOWPASS m0, m1, m2, m0, m4
    pshufb      m1, m0, [off(intra8x9_ddr1)]
    pshufb      m2, m0, [off(intra8x9_ddr3)]
    mova pred(4,0), m1
    mova pred(4,1), m2
    psadbw      m4, m1, m5
    psadbw      m2, m6
    paddw       m4, m2

    add         r0, 256
    add         r5, 0xC0
    %define off(m) (r0+m-(intra8x9_h1+256+128))
    %define pred(i,j) [r5+i*0x40+j*0x20-0x1C0]

    vpblendd    m2, m3, m0, 11110011b
    pshufb      m1, m2, [off(intra8x9_vr1)]
    pshufb      m2, m2, [off(intra8x9_vr3)]
    mova pred(5,0), m1
    mova pred(5,1), m2
    psadbw      m1, m5
    psadbw      m2, m6
    paddw       m1, m2

    pslldq      m1, 2
    por         m4, m1

    psrldq      m2, m3, 4
    pblendw     m2, m0, q3330
    punpcklbw   m0, m3
    pshufb      m1, m2, [off(intra8x9_hd1)]
    pshufb      m2, m0, [off(intra8x9_hd3)]
    mova pred(6,0), m1
    mova pred(6,1), m2
    psadbw      m1, m5
    psadbw      m2, m6
    paddw       m1, m2

    pslldq      m1, 4
    por         m4, m1

    pshufb      m1, m7, [off(intra8x9_vl1)]
    pshufb      m2, m7, [off(intra8x9_vl3)]
    mova pred(7,0), m1
    mova pred(7,1), m2
    psadbw      m1, m5
    psadbw      m2, m6
    paddw       m1, m2

    pslldq      m1, 6
    por         m4, m1
    vextracti128 xm1, m4, 1
    paddw      xm4, xm1
    mova       xm3, [r4]
    SBUTTERFLY qdq, 3, 4, 7
    paddw      xm3, xm4

    pslldq      m1, m0, 1
    vpbroadcastd m0, [r2+7]
    palignr     m0, m1, 1
    pshufb      m1, m0, [off(intra8x9_hu1)]
    pshufb      m2, m0, [off(intra8x9_hu3)]
    mova pred(8,0), m1
    mova pred(8,1), m2
    psadbw      m1, m5
    psadbw      m2, m6
    paddw       m1, m2
    vextracti128 xm2, m1, 1
    paddw      xm1, xm2
    MOVHL      xm2, xm1
    paddw      xm1, xm2
    movd       r2d, xm1

    paddw      xm3, [r3]
    mova      [r4], xm3
    add        r2w, word [r3+16]
    mov    [r4+16], r2w

    phminposuw xm3, xm3
    movd       r3d, xm3
    add        r2d, 8<<16
    cmp        r3w, r2w
    cmovg      r3d, r2d

    mov        r2d, r3d
    shr         r3, 16
    shl         r3, 6
    add         r1, 4*FDEC_STRIDE
    mova       xm0, [rsp+r3+0x00]
    mova       xm1, [rsp+r3+0x10]
    mova       xm2, [rsp+r3+0x20]
    mova       xm3, [rsp+r3+0x30]
    movq   [r1+FDEC_STRIDE*-4], xm0
    movhps [r1+FDEC_STRIDE*-2], xm0
    movq   [r1+FDEC_STRIDE*-3], xm1
    movhps [r1+FDEC_STRIDE*-1], xm1
    movq   [r1+FDEC_STRIDE* 0], xm2
    movhps [r1+FDEC_STRIDE* 2], xm2
    movq   [r1+FDEC_STRIDE* 1], xm3
    movhps [r1+FDEC_STRIDE* 3], xm3
    mov        rsp, r6
    mov        eax, r2d
    RET

%macro SATD_AVX512_LOAD4 2 ; size, opmask
    vpbroadcast%1 m0, [r0]
    vpbroadcast%1 m0 {%2}, [r0+2*r1]
    vpbroadcast%1 m2, [r2]
    vpbroadcast%1 m2 {%2}, [r2+2*r3]
    add           r0, r1
    add           r2, r3
    vpbroadcast%1 m1, [r0]
    vpbroadcast%1 m1 {%2}, [r0+2*r1]
    vpbroadcast%1 m3, [r2]
    vpbroadcast%1 m3 {%2}, [r2+2*r3]
%endmacro

%macro SATD_AVX512_LOAD8 5 ; size, halfreg, opmask1, opmask2, opmask3
    vpbroadcast%1 %{2}0, [r0]
    vpbroadcast%1 %{2}0 {%3}, [r0+2*r1]
    vpbroadcast%1 %{2}2, [r2]
    vpbroadcast%1 %{2}2 {%3}, [r2+2*r3]
    vpbroadcast%1    m0 {%4}, [r0+4*r1]
    vpbroadcast%1    m2 {%4}, [r2+4*r3]
    vpbroadcast%1    m0 {%5}, [r0+2*r4]
    vpbroadcast%1    m2 {%5}, [r2+2*r5]
    vpbroadcast%1 %{2}1, [r0+r1]
    vpbroadcast%1 %{2}1 {%3}, [r0+r4]
    vpbroadcast%1 %{2}3, [r2+r3]
    vpbroadcast%1 %{2}3 {%3}, [r2+r5]
    lea              r0, [r0+4*r1]
    lea              r2, [r2+4*r3]
    vpbroadcast%1    m1 {%4}, [r0+r1]
    vpbroadcast%1    m3 {%4}, [r2+r3]
    vpbroadcast%1    m1 {%5}, [r0+r4]
    vpbroadcast%1    m3 {%5}, [r2+r5]
%endmacro

%macro SATD_AVX512_PACKED 0
    DIFF_SUMSUB_SSSE3 0, 2, 1, 3, 4
    SUMSUB_BA      w, 0, 1, 2
    SBUTTERFLY   qdq, 0, 1, 2
    SUMSUB_BA      w, 0, 1, 2
    HMAXABSW2         0, 1, 2, 3
%endmacro

%macro SATD_AVX512_END 0-1 0 ; sa8d
    vpaddw         m0 {k1}{z}, m1 ; zero-extend to dwords
%if ARCH_X86_64
%if mmsize == 64
    vextracti32x8 ym1, m0, 1
    paddd         ym0, ym1
%endif
%if mmsize >= 32
    vextracti128  xm1, ym0, 1
    paddd        xmm0, xm0, xm1
%endif
    punpckhqdq   xmm1, xmm0, xmm0
    paddd        xmm0, xmm1
    movq          rax, xmm0
    rorx          rdx, rax, 32
%if %1
    lea           eax, [rax+rdx+1]
    shr           eax, 1
%else
    add           eax, edx
%endif
%else
    HADDD          m0, m1
    movd          eax, xm0
%if %1
    inc           eax
    shr           eax, 1
%endif
%endif
    RET
%endmacro

%macro HMAXABSW2 4 ; a, b, tmp1, tmp2
    pabsw     m%1, m%1
    pabsw     m%2, m%2
    psrldq    m%3, m%1, 2
    psrld     m%4, m%2, 16
    pmaxsw    m%1, m%3
    pmaxsw    m%2, m%4
%endmacro

INIT_ZMM avx512
cglobal pixel_satd_16x8_internal
    vbroadcasti64x4 m6, [hmul_16p]
    kxnorb           k2, k2, k2
    mov             r4d, 0x55555555
    knotw            k2, k2
    kmovd            k1, r4d
    lea              r4, [3*r1]
    lea              r5, [3*r3]
satd_16x8_avx512:
    vbroadcasti128  ym0,      [r0]
    vbroadcasti32x4  m0 {k2}, [r0+4*r1] ; 0 0 4 4
    vbroadcasti128  ym4,      [r2]
    vbroadcasti32x4  m4 {k2}, [r2+4*r3]
    vbroadcasti128  ym2,      [r0+2*r1]
    vbroadcasti32x4  m2 {k2}, [r0+2*r4] ; 2 2 6 6
    vbroadcasti128  ym5,      [r2+2*r3]
    vbroadcasti32x4  m5 {k2}, [r2+2*r5]
    DIFF_SUMSUB_SSSE3 0, 4, 2, 5, 6
    vbroadcasti128  ym1,      [r0+r1]
    vbroadcasti128  ym4,      [r2+r3]
    vbroadcasti128  ym3,      [r0+r4]
    vbroadcasti128  ym5,      [r2+r5]
    lea              r0, [r0+4*r1]
    lea              r2, [r2+4*r3]
    vbroadcasti32x4  m1 {k2}, [r0+r1] ; 1 1 5 5
    vbroadcasti32x4  m4 {k2}, [r2+r3]
    vbroadcasti32x4  m3 {k2}, [r0+r4] ; 3 3 7 7
    vbroadcasti32x4  m5 {k2}, [r2+r5]
    DIFF_SUMSUB_SSSE3 1, 4, 3, 5, 6
    HADAMARD4_V       0, 1, 2, 3, 4
    HMAXABSW2         0, 2, 4, 5
    HMAXABSW2         1, 3, 4, 5
    paddw            m4, m0, m2 ; m1
    paddw            m2, m1, m3 ; m0
    ret

cglobal pixel_satd_8x8_internal
    vbroadcasti64x4 m4, [hmul_16p]
    mov     r4d, 0x55555555
    kmovd    k1, r4d   ; 01010101
    kshiftlb k2, k1, 5 ; 10100000
    kshiftlb k3, k1, 4 ; 01010000
    lea      r4, [3*r1]
    lea      r5, [3*r3]
satd_8x8_avx512:
    SATD_AVX512_LOAD8 q, ym, k1, k2, k3 ; 2 0 2 0 6 4 6 4
    SATD_AVX512_PACKED                  ; 3 1 3 1 7 5 7 5
    ret

cglobal pixel_satd_16x8, 4,6
    call pixel_satd_16x8_internal_avx512
    jmp satd_zmm_avx512_end

cglobal pixel_satd_16x16, 4,6
    call pixel_satd_16x8_internal_avx512
    lea      r0, [r0+4*r1]
    lea      r2, [r2+4*r3]
    paddw    m7, m0, m1
    call satd_16x8_avx512
    paddw    m1, m7
    jmp satd_zmm_avx512_end

cglobal pixel_satd_8x8, 4,6
    call pixel_satd_8x8_internal_avx512
satd_zmm_avx512_end:
    SATD_AVX512_END

cglobal pixel_satd_8x16, 4,6
    call pixel_satd_8x8_internal_avx512
    lea      r0, [r0+4*r1]
    lea      r2, [r2+4*r3]
    paddw    m5, m0, m1
    call satd_8x8_avx512
    paddw    m1, m5
    jmp satd_zmm_avx512_end

INIT_YMM avx512
cglobal pixel_satd_4x8_internal
    vbroadcasti128 m4, [hmul_4p]
    mov     r4d, 0x55550c
    kmovd    k2, r4d   ; 00001100
    kshiftlb k3, k2, 2 ; 00110000
    kshiftlb k4, k2, 4 ; 11000000
    kshiftrd k1, k2, 8 ; 01010101
    lea      r4, [3*r1]
    lea      r5, [3*r3]
satd_4x8_avx512:
    SATD_AVX512_LOAD8 d, xm, k2, k3, k4 ; 0 0 2 2 4 4 6 6
satd_ymm_avx512:                        ; 1 1 3 3 5 5 7 7
    SATD_AVX512_PACKED
    ret

cglobal pixel_satd_8x4, 4,5
    mova     m4, [hmul_16p]
    mov     r4d, 0x5555
    kmovw    k1, r4d
    SATD_AVX512_LOAD4 q, k1 ; 2 0 2 0
    call satd_ymm_avx512    ; 3 1 3 1
    jmp satd_ymm_avx512_end2

cglobal pixel_satd_4x8, 4,6
    call pixel_satd_4x8_internal_avx512
satd_ymm_avx512_end:
%if ARCH_X86_64 == 0
    pop     r5d
    %assign regs_used 5
%endif
satd_ymm_avx512_end2:
    SATD_AVX512_END

cglobal pixel_satd_4x16, 4,6
    call pixel_satd_4x8_internal_avx512
    lea      r0, [r0+4*r1]
    lea      r2, [r2+4*r3]
    paddw    m5, m0, m1
    call satd_4x8_avx512
    paddw    m1, m5
    jmp satd_ymm_avx512_end

INIT_XMM avx512
cglobal pixel_satd_4x4, 4,5
    mova     m4, [hmul_4p]
    mov     r4d, 0x550c
    kmovw    k2, r4d
    kshiftrw k1, k2, 8
    SATD_AVX512_LOAD4 d, k2 ; 0 0 2 2
    SATD_AVX512_PACKED      ; 1 1 3 3
    SWAP      0, 1
    SATD_AVX512_END

INIT_ZMM avx512
cglobal pixel_sa8d_8x8, 4,6
    vbroadcasti64x4 m4, [hmul_16p]
    mov     r4d, 0x55555555
    kmovd    k1, r4d   ; 01010101
    kshiftlb k2, k1, 5 ; 10100000
    kshiftlb k3, k1, 4 ; 01010000
    lea      r4, [3*r1]
    lea      r5, [3*r3]
    SATD_AVX512_LOAD8 q, ym, k1, k2, k3 ; 2 0 2 0 6 4 6 4
    DIFF_SUMSUB_SSSE3 0, 2, 1, 3, 4     ; 3 1 3 1 7 5 7 5
    SUMSUB_BA      w, 0, 1, 2
    SBUTTERFLY   qdq, 0, 1, 2
    SUMSUB_BA      w, 0, 1, 2
    shufps        m2, m0, m1, q2020
    shufps        m1, m0, m1, q3131
    SUMSUB_BA      w, 2, 1, 0
    vshufi32x4    m0, m2, m1, q1010
    vshufi32x4    m1, m2, m1, q3232
    SUMSUB_BA      w, 0, 1, 2
    HMAXABSW2      0, 1, 2, 3
    SATD_AVX512_END 1

%endif ; HIGH_BIT_DEPTH

;=============================================================================
; SSIM
;=============================================================================

;-----------------------------------------------------------------------------
; void pixel_ssim_4x4x2_core( const uint8_t *pix1, intptr_t stride1,
;                             const uint8_t *pix2, intptr_t stride2, int sums[2][4] )
;-----------------------------------------------------------------------------
%macro SSIM_ITER 1
%if HIGH_BIT_DEPTH
    movu      m4, [r0+(%1&1)*r1]
    movu      m5, [r2+(%1&1)*r3]
%elif cpuflag(avx)
    pmovzxbw  m4, [r0+(%1&1)*r1]
    pmovzxbw  m5, [r2+(%1&1)*r3]
%else
    movq      m4, [r0+(%1&1)*r1]
    movq      m5, [r2+(%1&1)*r3]
    punpcklbw m4, m7
    punpcklbw m5, m7
%endif
%if %1==1
    lea       r0, [r0+r1*2]
    lea       r2, [r2+r3*2]
%endif
%if %1 == 0 && cpuflag(avx)
    SWAP       0, 4
    SWAP       1, 5
    pmaddwd   m4, m0, m0
    pmaddwd   m5, m1, m1
    pmaddwd   m6, m0, m1
%else
%if %1 == 0
    mova      m0, m4
    mova      m1, m5
%else
    paddw     m0, m4
    paddw     m1, m5
%endif
    pmaddwd   m6, m4, m5
    pmaddwd   m4, m4
    pmaddwd   m5, m5
%endif
    ACCUM  paddd, 2, 4, %1
    ACCUM  paddd, 3, 6, %1
    paddd     m2, m5
%endmacro

%macro SSIM 0
%if HIGH_BIT_DEPTH
cglobal pixel_ssim_4x4x2_core, 4,4,7
    FIX_STRIDES r1, r3
%else
cglobal pixel_ssim_4x4x2_core, 4,4,7+notcpuflag(avx)
%if notcpuflag(avx)
    pxor      m7, m7
%endif
%endif
    SSIM_ITER 0
    SSIM_ITER 1
    SSIM_ITER 2
    SSIM_ITER 3
%if UNIX64
    DECLARE_REG_TMP 4
%else
    DECLARE_REG_TMP 0
    mov       t0, r4mp
%endif
%if cpuflag(ssse3)
    phaddw    m0, m1
    pmaddwd   m0, [pw_1]
    phaddd    m2, m3
%else
    mova      m4, [pw_1]
    pmaddwd   m0, m4
    pmaddwd   m1, m4
    packssdw  m0, m1
    shufps    m1, m2, m3, q2020
    shufps    m2, m3, q3131
    pmaddwd   m0, m4
    paddd     m2, m1
%endif
    shufps    m1, m0, m2, q2020
    shufps    m0, m2, q3131
    mova    [t0], m1
    mova [t0+16], m0
    RET

;-----------------------------------------------------------------------------
; float pixel_ssim_end( int sum0[5][4], int sum1[5][4], int width )
;-----------------------------------------------------------------------------
cglobal pixel_ssim_end4, 2,3
    mov      r2d, r2m
    mova      m0, [r0+ 0]
    mova      m1, [r0+16]
    mova      m2, [r0+32]
    mova      m3, [r0+48]
    mova      m4, [r0+64]
    paddd     m0, [r1+ 0]
    paddd     m1, [r1+16]
    paddd     m2, [r1+32]
    paddd     m3, [r1+48]
    paddd     m4, [r1+64]
    paddd     m0, m1
    paddd     m1, m2
    paddd     m2, m3
    paddd     m3, m4
    TRANSPOSE4x4D  0, 1, 2, 3, 4

;   s1=m0, s2=m1, ss=m2, s12=m3
%if BIT_DEPTH == 10
    cvtdq2ps  m0, m0
    cvtdq2ps  m1, m1
    cvtdq2ps  m2, m2
    cvtdq2ps  m3, m3
    mulps     m4, m0, m1  ; s1*s2
    mulps     m0, m0      ; s1*s1
    mulps     m1, m1      ; s2*s2
    mulps     m2, [pf_64] ; ss*64
    mulps     m3, [pf_128] ; s12*128
    addps     m4, m4      ; s1*s2*2
    addps     m0, m1      ; s1*s1 + s2*s2
    subps     m2, m0      ; vars
    subps     m3, m4      ; covar*2
    movaps    m1, [ssim_c1]
    addps     m4, m1      ; s1*s2*2 + ssim_c1
    addps     m0, m1      ; s1*s1 + s2*s2 + ssim_c1
    movaps    m1, [ssim_c2]
    addps     m2, m1      ; vars + ssim_c2
    addps     m3, m1      ; covar*2 + ssim_c2
%else
    pmaddwd   m4, m1, m0  ; s1*s2
    pslld     m1, 16
    por       m0, m1
    pmaddwd   m0, m0  ; s1*s1 + s2*s2
    pslld     m4, 1
    pslld     m3, 7
    pslld     m2, 6
    psubd     m3, m4  ; covar*2
    psubd     m2, m0  ; vars
    mova      m1, [ssim_c1]
    paddd     m0, m1
    paddd     m4, m1
    mova      m1, [ssim_c2]
    paddd     m3, m1
    paddd     m2, m1
    cvtdq2ps  m0, m0  ; (float)(s1*s1 + s2*s2 + ssim_c1)
    cvtdq2ps  m4, m4  ; (float)(s1*s2*2 + ssim_c1)
    cvtdq2ps  m3, m3  ; (float)(covar*2 + ssim_c2)
    cvtdq2ps  m2, m2  ; (float)(vars + ssim_c2)
%endif
    mulps     m4, m3
    mulps     m0, m2
    divps     m4, m0  ; ssim

    cmp       r2d, 4
    je .skip ; faster only if this is the common case; remove branch if we use ssim on a macroblock level
    neg       r2

%if ARCH_X86_64
    lea       r3, [mask_ff + 16]
    %xdefine %%mask r3
%else
    %xdefine %%mask mask_ff + 16
%endif
%if cpuflag(avx)
    andps     m4, [%%mask + r2*4]
%else
    movups    m0, [%%mask + r2*4]
    andps     m4, m0
%endif

.skip:
    movhlps   m0, m4
    addps     m0, m4
%if cpuflag(ssse3)
    movshdup  m4, m0
%else
    pshuflw   m4, m0, q0032
%endif
    addss     m0, m4
%if ARCH_X86_64 == 0
    movss    r0m, m0
    fld     dword r0m
%endif
    RET
%endmacro ; SSIM

INIT_XMM sse2
SSIM
INIT_XMM avx
SSIM

;-----------------------------------------------------------------------------
; int pixel_asd8( pixel *pix1, intptr_t stride1, pixel *pix2, intptr_t stride2, int height );
;-----------------------------------------------------------------------------
%macro ASD8 0
cglobal pixel_asd8, 5,5
    pxor     m0, m0
    pxor     m1, m1
.loop:
%if HIGH_BIT_DEPTH
    paddw    m0, [r0]
    paddw    m1, [r2]
    paddw    m0, [r0+2*r1]
    paddw    m1, [r2+2*r3]
    lea      r0, [r0+4*r1]
    paddw    m0, [r0]
    paddw    m1, [r2+4*r3]
    lea      r2, [r2+4*r3]
    paddw    m0, [r0+2*r1]
    paddw    m1, [r2+2*r3]
    lea      r0, [r0+4*r1]
    lea      r2, [r2+4*r3]
%else
    movq     m2, [r0]
    movq     m3, [r2]
    movhps   m2, [r0+r1]
    movhps   m3, [r2+r3]
    lea      r0, [r0+2*r1]
    psadbw   m2, m1
    psadbw   m3, m1
    movq     m4, [r0]
    movq     m5, [r2+2*r3]
    lea      r2, [r2+2*r3]
    movhps   m4, [r0+r1]
    movhps   m5, [r2+r3]
    lea      r0, [r0+2*r1]
    paddw    m0, m2
    psubw    m0, m3
    psadbw   m4, m1
    psadbw   m5, m1
    lea      r2, [r2+2*r3]
    paddw    m0, m4
    psubw    m0, m5
%endif
    sub     r4d, 4
    jg .loop
%if HIGH_BIT_DEPTH
    psubw    m0, m1
    HADDW    m0, m1
    ABSD     m1, m0
%else
    MOVHL    m1, m0
    paddw    m0, m1
    ABSW     m1, m0
%endif
    movd    eax, m1
    RET
%endmacro

INIT_XMM sse2
ASD8
INIT_XMM ssse3
ASD8
%if HIGH_BIT_DEPTH
INIT_XMM xop
ASD8
%endif

;=============================================================================
; Successive Elimination ADS
;=============================================================================

%macro ADS_START 0
%if UNIX64
    movsxd  r5, r5d
%else
    mov    r5d, r5m
%endif
    mov    r0d, r5d
    lea     r6, [r4+r5+(mmsize-1)]
    and     r6, ~(mmsize-1)
    shl     r2d,  1
%endmacro

%macro ADS_END 1-2 .loop ; unroll_size, loop_label
    add     r1, 2*%1
    add     r3, 2*%1
    add     r6, %1
    sub    r0d, %1
    jg %2
    WIN64_RESTORE_XMM_INTERNAL
%if mmsize==32
    vzeroupper
%endif
    lea     r6, [r4+r5+(mmsize-1)]
    and     r6, ~(mmsize-1)
%if cpuflag(ssse3)
    jmp ads_mvs_ssse3
%else
    jmp ads_mvs_mmx
%endif
%endmacro

;-----------------------------------------------------------------------------
; int pixel_ads4( int enc_dc[4], uint16_t *sums, int delta,
;                 uint16_t *cost_mvx, int16_t *mvs, int width, int thresh )
;-----------------------------------------------------------------------------
%if HIGH_BIT_DEPTH

%macro ADS_XMM 0
%if ARCH_X86_64
cglobal pixel_ads4, 5,7,9
%else
cglobal pixel_ads4, 5,7,8
%endif
%if mmsize >= 32
    vpbroadcastd m7, [r0+ 0]
    vpbroadcastd m6, [r0+ 4]
    vpbroadcastd m5, [r0+ 8]
    vpbroadcastd m4, [r0+12]
%else
    mova      m4, [r0]
    pshufd    m7, m4, 0
    pshufd    m6, m4, q1111
    pshufd    m5, m4, q2222
    pshufd    m4, m4, q3333
%endif
%if ARCH_X86_64
    SPLATD    m8, r6m
%endif
    ADS_START
.loop:
%if cpuflag(avx)
    pmovzxwd  m0, [r1]
    pmovzxwd  m1, [r1+16]
%else
    movh      m0, [r1]
    movh      m1, [r1+16]
    pxor      m3, m3
    punpcklwd m0, m3
    punpcklwd m1, m3
%endif
    psubd     m0, m7
    psubd     m1, m6
    ABSD      m0, m0, m2
    ABSD      m1, m1, m3
%if cpuflag(avx)
    pmovzxwd  m2, [r1+r2]
    pmovzxwd  m3, [r1+r2+16]
    paddd     m0, m1
%else
    movh      m2, [r1+r2]
    movh      m3, [r1+r2+16]
    paddd     m0, m1
    pxor      m1, m1
    punpcklwd m2, m1
    punpcklwd m3, m1
%endif
    psubd     m2, m5
    psubd     m3, m4
    ABSD      m2, m2, m1
    ABSD      m3, m3, m1
    paddd     m0, m2
    paddd     m0, m3
%if cpuflag(avx)
    pmovzxwd  m1, [r3]
%else
    movh      m1, [r3]
    pxor      m3, m3
    punpcklwd m1, m3
%endif
    paddd     m0, m1
%if ARCH_X86_64
    psubd     m1, m8, m0
%else
    SPLATD    m1, r6m
    psubd     m1, m0
%endif
    packssdw  m1, m1
%if mmsize == 32
    vpermq    m1, m1, q3120
    packuswb  m1, m1
    movq    [r6], xm1
%else
    packuswb  m1, m1
    movd    [r6], m1
%endif
    ADS_END mmsize/4

cglobal pixel_ads2, 5,7,8
%if mmsize >= 32
    vpbroadcastd m7, [r0+0]
    vpbroadcastd m6, [r0+4]
    vpbroadcastd m5, r6m
%else
    movq      m6, [r0]
    movd      m5, r6m
    pshufd    m7, m6, 0
    pshufd    m6, m6, q1111
    pshufd    m5, m5, 0
%endif
    pxor      m4, m4
    ADS_START
.loop:
%if cpuflag(avx)
    pmovzxwd  m0, [r1]
    pmovzxwd  m1, [r1+r2]
    pmovzxwd  m2, [r3]
%else
    movh      m0, [r1]
    movh      m1, [r1+r2]
    movh      m2, [r3]
    punpcklwd m0, m4
    punpcklwd m1, m4
    punpcklwd m2, m4
%endif
    psubd     m0, m7
    psubd     m1, m6
    ABSD      m0, m0, m3
    ABSD      m1, m1, m3
    paddd     m0, m1
    paddd     m0, m2
    psubd     m1, m5, m0
    packssdw  m1, m1
%if mmsize == 32
    vpermq    m1, m1, q3120
    packuswb  m1, m1
    movq    [r6], xm1
%else
    packuswb  m1, m1
    movd    [r6], m1
%endif
    ADS_END mmsize/4

cglobal pixel_ads1, 5,7,8
%if mmsize >= 32
    vpbroadcastd m7, [r0]
    vpbroadcastd m6, r6m
%else
    movd      m7, [r0]
    movd      m6, r6m
    pshufd    m7, m7, 0
    pshufd    m6, m6, 0
%endif
    pxor      m5, m5
    ADS_START
.loop:
    movu      m1, [r1]
    movu      m3, [r3]
    punpcklwd m0, m1, m5
    punpckhwd m1, m5
    punpcklwd m2, m3, m5
    punpckhwd m3, m5
    psubd     m0, m7
    psubd     m1, m7
    ABSD      m0, m0, m4
    ABSD      m1, m1, m4
    paddd     m0, m2
    paddd     m1, m3
    psubd     m2, m6, m0
    psubd     m3, m6, m1
    packssdw  m2, m3
    packuswb  m2, m2
%if mmsize == 32
    vpermq    m2, m2, q3120
    mova    [r6], xm2
%else
    movq    [r6], m2
%endif
    ADS_END mmsize/2
%endmacro

INIT_XMM sse2
ADS_XMM
INIT_XMM ssse3
ADS_XMM
INIT_XMM avx
ADS_XMM
INIT_YMM avx2
ADS_XMM

%else ; !HIGH_BIT_DEPTH

%macro ADS_XMM 0
%if ARCH_X86_64 && mmsize == 16
cglobal pixel_ads4, 5,7,12
%elif ARCH_X86_64 && mmsize != 8
cglobal pixel_ads4, 5,7,9
%else
cglobal pixel_ads4, 5,7,8
%endif
    test dword r6m, 0xffff0000
%if mmsize >= 32
    vpbroadcastw m7, [r0+ 0]
    vpbroadcastw m6, [r0+ 4]
    vpbroadcastw m5, [r0+ 8]
    vpbroadcastw m4, [r0+12]
%elif mmsize == 16
    mova       m4, [r0]
    pshuflw    m7, m4, 0
    pshuflw    m6, m4, q2222
    pshufhw    m5, m4, 0
    pshufhw    m4, m4, q2222
    punpcklqdq m7, m7
    punpcklqdq m6, m6
    punpckhqdq m5, m5
    punpckhqdq m4, m4
%else
    mova      m6, [r0]
    mova      m4, [r0+8]
    pshufw    m7, m6, 0
    pshufw    m6, m6, q2222
    pshufw    m5, m4, 0
    pshufw    m4, m4, q2222
%endif
    jnz .nz
    ADS_START
%if ARCH_X86_64 && mmsize == 16
    movu     m10, [r1]
    movu     m11, [r1+r2]
    SPLATW    m8, r6m
.loop:
    psubw     m0, m10, m7
    movu     m10, [r1+16]
    psubw     m1, m10, m6
    ABSW      m0, m0, m2
    ABSW      m1, m1, m3
    psubw     m2, m11, m5
    movu     m11, [r1+r2+16]
    paddw     m0, m1
    psubw     m3, m11, m4
    movu      m9, [r3]
    ABSW      m2, m2, m1
    ABSW      m3, m3, m1
    paddw     m0, m2
    paddw     m0, m3
    paddusw   m0, m9
    psubusw   m1, m8, m0
%else
%if ARCH_X86_64 && mmsize != 8
    SPLATW    m8, r6m
%endif
.loop:
    movu      m0, [r1]
    movu      m1, [r1+16]
    psubw     m0, m7
    psubw     m1, m6
    ABSW      m0, m0, m2
    ABSW      m1, m1, m3
    movu      m2, [r1+r2]
    movu      m3, [r1+r2+16]
    psubw     m2, m5
    psubw     m3, m4
    paddw     m0, m1
    ABSW      m2, m2, m1
    ABSW      m3, m3, m1
    paddw     m0, m2
    paddw     m0, m3
    movu      m2, [r3]
%if ARCH_X86_64 && mmsize != 8
    mova      m1, m8
%else
    SPLATW    m1, r6m
%endif
    paddusw   m0, m2
    psubusw   m1, m0
%endif ; ARCH
    packsswb  m1, m1
%if mmsize == 32
    vpermq    m1, m1, q3120
    mova    [r6], xm1
%else
    movh    [r6], m1
%endif
    ADS_END mmsize/2
.nz:
    ADS_START
%if ARCH_X86_64 && mmsize == 16
    movu     m10, [r1]
    movu     m11, [r1+r2]
    SPLATD    m8, r6m
.loop_nz:
    psubw     m0, m10, m7
    movu     m10, [r1+16]
    psubw     m1, m10, m6
    ABSW      m0, m0, m2
    ABSW      m1, m1, m3
    psubw     m2, m11, m5
    movu     m11, [r1+r2+16]
    paddw     m0, m1
    psubw     m3, m11, m4
    movu      m9, [r3]
    ABSW      m2, m2, m1
    ABSW      m3, m3, m1
    paddw     m0, m2
    paddw     m0, m3
    pxor      m3, m3
    mova      m2, m0
    mova      m1, m9
    punpcklwd m0, m3
    punpcklwd m9, m3
    punpckhwd m2, m3
    punpckhwd m1, m3
    paddd     m0, m9
    paddd     m2, m1
    psubd     m1, m8, m0
    psubd     m3, m8, m2
    packssdw  m1, m3
    packuswb  m1, m1
%else
%if ARCH_X86_64 && mmsize != 8
    SPLATD    m8, r6m
%endif
.loop_nz:
    movu      m0, [r1]
    movu      m1, [r1+16]
    psubw     m0, m7
    psubw     m1, m6
    ABSW      m0, m0, m2
    ABSW      m1, m1, m3
    movu      m2, [r1+r2]
    movu      m3, [r1+r2+16]
    psubw     m2, m5
    psubw     m3, m4
    paddw     m0, m1
    ABSW      m2, m2, m1
    ABSW      m3, m3, m1
    paddw     m0, m2
    paddw     m0, m3
%if mmsize == 32
    movu      m1, [r3]
%else
    movh      m1, [r3]
%endif
    pxor      m3, m3
    mova      m2, m0
    punpcklwd m0, m3
    punpcklwd m1, m3
    punpckhwd m2, m3
    paddd     m0, m1
%if mmsize == 32
    movu      m1, [r3]
    punpckhwd m1, m3
%else
    movh      m1, [r3+mmsize/2]
    punpcklwd m1, m3
%endif
    paddd     m2, m1
%if ARCH_X86_64 && mmsize != 8
    mova      m1, m8
%else
    SPLATD    m1, r6m
%endif
    mova      m3, m1
    psubd     m1, m0
    psubd     m3, m2
    packssdw  m1, m3
    packuswb  m1, m1
%endif ; ARCH
%if mmsize == 32
    vpermq    m1, m1, q3120
    mova    [r6], xm1
%else
    movh    [r6], m1
%endif
    ADS_END mmsize/2, .loop_nz

cglobal pixel_ads2, 5,7,8
    test dword r6m, 0xffff0000
%if mmsize >= 32
    vpbroadcastw m7, [r0+0]
    vpbroadcastw m6, [r0+4]
%elif mmsize == 16
    movq       m6, [r0]
    pshuflw    m7, m6, 0
    pshuflw    m6, m6, q2222
    punpcklqdq m7, m7
    punpcklqdq m6, m6
%else
    mova      m6, [r0]
    pshufw    m7, m6, 0
    pshufw    m6, m6, q2222
%endif
    jnz .nz
    ADS_START
    SPLATW    m5, r6m
.loop:
    movu      m0, [r1]
    movu      m1, [r1+r2]
    movu      m2, [r3]
    psubw     m0, m7
    psubw     m1, m6
    ABSW      m0, m0, m3
    ABSW      m1, m1, m4
    paddw     m0, m1
    paddusw   m0, m2
    psubusw   m1, m5, m0
    packsswb  m1, m1
%if mmsize == 32
    vpermq    m1, m1, q3120
    mova    [r6], xm1
%else
    movh    [r6], m1
%endif
    ADS_END mmsize/2
.nz:
    ADS_START
    SPLATD    m5, r6m
    pxor      m4, m4
.loop_nz:
    movu      m0, [r1]
    movu      m1, [r1+r2]
    movu      m2, [r3]
    psubw     m0, m7
    psubw     m1, m6
    ABSW      m0, m0, m3
    ABSW      m1, m1, m3
    paddw     m0, m1
    punpckhwd m3, m2, m4
    punpckhwd m1, m0, m4
    punpcklwd m2, m4
    punpcklwd m0, m4
    paddd     m1, m3
    paddd     m0, m2
    psubd     m3, m5, m1
    psubd     m2, m5, m0
    packssdw  m2, m3
    packuswb  m2, m2
%if mmsize == 32
    vpermq    m2, m2, q3120
    mova    [r6], xm2
%else
    movh    [r6], m2
%endif
    ADS_END mmsize/2, .loop_nz

cglobal pixel_ads1, 5,7,8
    test dword r6m, 0xffff0000
    SPLATW    m7, [r0]
    jnz .nz
    ADS_START
    SPLATW    m6, r6m
.loop:
    movu      m0, [r1]
    movu      m1, [r1+mmsize]
    movu      m2, [r3]
    movu      m3, [r3+mmsize]
    psubw     m0, m7
    psubw     m1, m7
    ABSW      m0, m0, m4
    ABSW      m1, m1, m5
    paddusw   m0, m2
    paddusw   m1, m3
    psubusw   m4, m6, m0
    psubusw   m5, m6, m1
    packsswb  m4, m5
%if mmsize == 32
    vpermq    m4, m4, q3120
%endif
    mova    [r6], m4
    ADS_END mmsize
.nz:
    ADS_START
    SPLATD    m6, r6m
    pxor      m5, m5
.loop_nz:
    movu      m0, [r1]
    movu      m1, [r1+mmsize]
    movu      m2, [r3]
    psubw     m0, m7
    psubw     m1, m7
    ABSW      m0, m0, m3
    ABSW      m1, m1, m4
    punpckhwd m3, m2, m5
    punpckhwd m4, m0, m5
    punpcklwd m2, m5
    punpcklwd m0, m5
    paddd     m4, m3
    paddd     m0, m2
    psubd     m3, m6, m4
    movu      m4, [r3+mmsize]
    psubd     m2, m6, m0
    packssdw  m2, m3
    punpckhwd m0, m1, m5
    punpckhwd m3, m4, m5
    punpcklwd m1, m5
    punpcklwd m4, m5
    paddd     m0, m3
    paddd     m1, m4
    psubd     m3, m6, m0
    psubd     m4, m6, m1
    packssdw  m4, m3
    packuswb  m2, m4
%if mmsize == 32
    vpermq    m2, m2, q3120
%endif
    mova    [r6], m2
    ADS_END mmsize, .loop_nz
%endmacro

INIT_MMX mmx2
ADS_XMM
INIT_XMM sse2
ADS_XMM
INIT_XMM ssse3
ADS_XMM
INIT_XMM avx
ADS_XMM
INIT_YMM avx2
ADS_XMM

%endif ; HIGH_BIT_DEPTH

; int pixel_ads_mvs( int16_t *mvs, uint8_t *masks, int width )
; {
;     int nmv=0, i, j;
;     *(uint32_t*)(masks+width) = 0;
;     for( i=0; i<width; i+=8 )
;     {
;         uint64_t mask = *(uint64_t*)(masks+i);
;         if( !mask ) continue;
;         for( j=0; j<8; j++ )
;             if( mask & (255<<j*8) )
;                 mvs[nmv++] = i+j;
;     }
;     return nmv;
; }

%macro TEST 1
    mov     [r4+r0*2], r1w
    test    r2d, 0xff<<(%1*8)
    setne   r3b
    add     r0d, r3d
    inc     r1d
%endmacro

INIT_MMX mmx
cglobal pixel_ads_mvs, 0,7,0
ads_mvs_mmx:
    ; mvs = r4
    ; masks = r6
    ; width = r5
    ; clear last block in case width isn't divisible by 8. (assume divisible by 4, so clearing 4 bytes is enough.)
    xor     r0d, r0d
    xor     r1d, r1d
    mov     [r6+r5], r0d
    jmp .loopi
ALIGN 16
.loopi0:
    add     r1d, 8
    cmp     r1d, r5d
    jge .end
.loopi:
    mov     r2,  [r6+r1]
%if ARCH_X86_64
    test    r2,  r2
%else
    mov     r3,  r2
    or     r3d, [r6+r1+4]
%endif
    jz .loopi0
    xor     r3d, r3d
    TEST 0
    TEST 1
    TEST 2
    TEST 3
%if ARCH_X86_64
    shr     r2,  32
%else
    mov     r2d, [r6+r1]
%endif
    TEST 0
    TEST 1
    TEST 2
    TEST 3
    cmp     r1d, r5d
    jl .loopi
.end:
    movifnidn eax, r0d
    RET

INIT_XMM ssse3
cglobal pixel_ads_mvs, 0,7,0
ads_mvs_ssse3:
    mova      m3, [pw_8]
    mova      m4, [pw_76543210]
    pxor      m5, m5
    add       r5, r6
    xor      r0d, r0d ; nmv
    mov     [r5], r0d
%if ARCH_X86_64
    lea       r1, [$$]
    %define GLOBAL +r1-$$
%else
    %define GLOBAL
%endif
.loop:
    movh      m0, [r6]
    pcmpeqb   m0, m5
    pmovmskb r2d, m0
    xor      r2d, 0xffff                         ; skipping if r2d is zero is slower (branch mispredictions)
    movzx    r3d, byte [r2+popcnt_table GLOBAL]  ; popcnt
    add      r2d, r2d
    ; shuffle counters based on mv mask
    pshufb    m2, m4, [r2*8+ads_mvs_shuffle GLOBAL]
    movu [r4+r0*2], m2
    add      r0d, r3d
    paddw     m4, m3                             ; {i*8+0, i*8+1, i*8+2, i*8+3, i*8+4, i*8+5, i*8+6, i*8+7}
    add       r6, 8
    cmp       r6, r5
    jl .loop
    movifnidn eax, r0d
    RET