Hash :
703c5960
Author :
Date :
2024-07-24T17:02:28
Vulkan: PPO must warmup pipeline cache according to GPL support
This commit caused regression:
https://chromium-review.googlesource.com/c/angle/angle/+/5366173
`ProgramPipelineVk::link()` uses Complete subset for warmup regardless
of the GPL support. However, `ContextVk::createGraphicsPipeline()` will
respect the GPL support. Before the above change, PPO was respecting
GPL for warmup.
This change removed `subset` parameter from the
`getPipelineCacheWarmUpTasks()` not only to avoid copy-paste logic, but
also to ensure consistency with `waitForGraphicsPostLinkTasks()`.
Bug: angleproject:8601
Change-Id: Iab5df1b55921649a8f98a34bb07d8e6a145bfd4d
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5738153
Reviewed-by: mohan maiya <m.maiya@samsung.com>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
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
//
// Copyright 2020 The ANGLE Project Authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
//
// ProgramExecutableVk.cpp: Collects the information and interfaces common to both ProgramVks and
// ProgramPipelineVks in order to execute/draw with either.
#include "libANGLE/renderer/vulkan/ProgramExecutableVk.h"
#include "common/string_utils.h"
#include "libANGLE/renderer/vulkan/BufferVk.h"
#include "libANGLE/renderer/vulkan/DisplayVk.h"
#include "libANGLE/renderer/vulkan/FramebufferVk.h"
#include "libANGLE/renderer/vulkan/ProgramPipelineVk.h"
#include "libANGLE/renderer/vulkan/ProgramVk.h"
#include "libANGLE/renderer/vulkan/TextureVk.h"
#include "libANGLE/renderer/vulkan/TransformFeedbackVk.h"
#include "libANGLE/renderer/vulkan/vk_helpers.h"
#include "libANGLE/renderer/vulkan/vk_utils.h"
namespace rx
{
namespace
{
// Limit decompressed vulkan pipelines to 10MB per program.
static constexpr size_t kMaxLocalPipelineCacheSize = 10 * 1024 * 1024;
bool ValidateTransformedSpirV(vk::Context *context,
const gl::ShaderBitSet &linkedShaderStages,
const ShaderInterfaceVariableInfoMap &variableInfoMap,
const gl::ShaderMap<angle::spirv::Blob> &spirvBlobs)
{
gl::ShaderType lastPreFragmentStage = gl::GetLastPreFragmentStage(linkedShaderStages);
for (gl::ShaderType shaderType : linkedShaderStages)
{
SpvTransformOptions options;
options.shaderType = shaderType;
options.isLastPreFragmentStage =
shaderType == lastPreFragmentStage && shaderType != gl::ShaderType::TessControl;
options.isTransformFeedbackStage = options.isLastPreFragmentStage;
options.useSpirvVaryingPrecisionFixer =
context->getFeatures().varyingsRequireMatchingPrecisionInSpirv.enabled;
angle::spirv::Blob transformed;
if (SpvTransformSpirvCode(options, variableInfoMap, spirvBlobs[shaderType], &transformed) !=
angle::Result::Continue)
{
return false;
}
}
return true;
}
uint32_t GetInterfaceBlockArraySize(const std::vector<gl::InterfaceBlock> &blocks,
uint32_t bufferIndex)
{
const gl::InterfaceBlock &block = blocks[bufferIndex];
if (!block.pod.isArray)
{
return 1;
}
ASSERT(block.pod.arrayElement == 0);
// Search consecutively until all array indices of this block are visited.
uint32_t arraySize;
for (arraySize = 1; bufferIndex + arraySize < blocks.size(); ++arraySize)
{
const gl::InterfaceBlock &nextBlock = blocks[bufferIndex + arraySize];
if (nextBlock.pod.arrayElement != arraySize)
{
break;
}
// It's unexpected for an array to start at a non-zero array size, so we can always rely on
// the sequential `arrayElement`s to belong to the same block.
ASSERT(nextBlock.name == block.name);
ASSERT(nextBlock.pod.isArray);
}
return arraySize;
}
void SetupDefaultPipelineState(const vk::Context *context,
const gl::ProgramExecutable &glExecutable,
gl::PrimitiveMode mode,
vk::PipelineRobustness pipelineRobustness,
vk::PipelineProtectedAccess pipelineProtectedAccess,
vk::GraphicsPipelineSubset subset,
vk::GraphicsPipelineDesc *graphicsPipelineDescOut)
{
graphicsPipelineDescOut->initDefaults(context, vk::GraphicsPipelineSubset::Complete,
pipelineRobustness, pipelineProtectedAccess);
// Set render pass state, affecting both complete and shaders-only pipelines.
graphicsPipelineDescOut->setTopology(mode);
graphicsPipelineDescOut->setRenderPassSampleCount(1);
graphicsPipelineDescOut->setRenderPassFramebufferFetchMode(glExecutable.usesFramebufferFetch());
const std::vector<gl::ProgramOutput> &outputVariables = glExecutable.getOutputVariables();
const std::vector<gl::VariableLocation> &outputLocations = glExecutable.getOutputLocations();
gl::DrawBufferMask drawBuffers;
for (const gl::VariableLocation &outputLocation : outputLocations)
{
if (outputLocation.arrayIndex == 0 && outputLocation.used() && !outputLocation.ignored)
{
const gl::ProgramOutput &outputVar = outputVariables[outputLocation.index];
if (angle::BeginsWith(outputVar.name, "gl_") && outputVar.name != "gl_FragColor")
{
continue;
}
uint32_t location = 0;
if (outputVar.pod.location != -1)
{
location = outputVar.pod.location;
}
GLenum type = gl::VariableComponentType(outputVar.pod.type);
angle::FormatID format = angle::FormatID::R8G8B8A8_UNORM;
if (type == GL_INT)
{
format = angle::FormatID::R8G8B8A8_SINT;
}
else if (type == GL_UNSIGNED_INT)
{
format = angle::FormatID::R8G8B8A8_UINT;
}
const size_t arraySize = outputVar.isArray() ? outputVar.getOutermostArraySize() : 1;
for (size_t arrayIndex = 0; arrayIndex < arraySize; ++arrayIndex)
{
graphicsPipelineDescOut->setRenderPassColorAttachmentFormat(location + arrayIndex,
format);
drawBuffers.set(location + arrayIndex);
}
}
}
for (const gl::ProgramOutput &outputVar : outputVariables)
{
if (outputVar.name == "gl_FragColor" || outputVar.name == "gl_FragData")
{
const size_t arraySize = outputVar.isArray() ? outputVar.getOutermostArraySize() : 1;
for (size_t arrayIndex = 0; arrayIndex < arraySize; ++arrayIndex)
{
graphicsPipelineDescOut->setRenderPassColorAttachmentFormat(
arrayIndex, angle::FormatID::R8G8B8A8_UNORM);
drawBuffers.set(arrayIndex);
}
}
}
if (subset == vk::GraphicsPipelineSubset::Complete)
{
// Include vertex input state
graphicsPipelineDescOut->setVertexShaderComponentTypes(
glExecutable.getNonBuiltinAttribLocationsMask(), glExecutable.getAttributesTypeMask());
// Include fragment output state
gl::BlendStateExt::ColorMaskStorage::Type colorMask =
gl::BlendStateExt::ColorMaskStorage::GetReplicatedValue(
gl::BlendStateExt::PackColorMask(true, true, true, true),
gl::BlendStateExt::ColorMaskStorage::GetMask(gl::IMPLEMENTATION_MAX_DRAW_BUFFERS));
graphicsPipelineDescOut->setColorWriteMasks(colorMask, {}, drawBuffers);
}
}
void GetPipelineCacheData(ContextVk *contextVk,
const vk::PipelineCache &pipelineCache,
angle::MemoryBuffer *cacheDataOut)
{
ASSERT(pipelineCache.valid() || contextVk->getState().isGLES1() ||
!contextVk->getFeatures().warmUpPipelineCacheAtLink.enabled ||
!contextVk->getFeatures().hasEffectivePipelineCacheSerialization.enabled);
if (!pipelineCache.valid() ||
!contextVk->getFeatures().hasEffectivePipelineCacheSerialization.enabled)
{
return;
}
// Extract the pipeline data. If failed, or empty, it's simply not stored on disk.
size_t pipelineCacheSize = 0;
VkResult result =
pipelineCache.getCacheData(contextVk->getDevice(), &pipelineCacheSize, nullptr);
if (result != VK_SUCCESS || pipelineCacheSize == 0)
{
return;
}
if (contextVk->getFeatures().enablePipelineCacheDataCompression.enabled)
{
std::vector<uint8_t> pipelineCacheData(pipelineCacheSize);
result = pipelineCache.getCacheData(contextVk->getDevice(), &pipelineCacheSize,
pipelineCacheData.data());
if (result != VK_SUCCESS && result != VK_INCOMPLETE)
{
return;
}
// Compress it.
if (!angle::CompressBlob(pipelineCacheData.size(), pipelineCacheData.data(), cacheDataOut))
{
cacheDataOut->clear();
}
}
else
{
if (!cacheDataOut->resize(pipelineCacheSize))
{
ERR() << "Failed to allocate memory for pipeline cache data.";
return;
}
result = pipelineCache.getCacheData(contextVk->getDevice(), &pipelineCacheSize,
cacheDataOut->data());
if (result != VK_SUCCESS && result != VK_INCOMPLETE)
{
cacheDataOut->clear();
}
}
}
vk::SpecializationConstants MakeSpecConsts(ProgramTransformOptions transformOptions,
const vk::GraphicsPipelineDesc &desc)
{
vk::SpecializationConstants specConsts;
specConsts.surfaceRotation = transformOptions.surfaceRotation;
specConsts.dither = desc.getEmulatedDitherControl();
return specConsts;
}
template <typename T>
void UpdateDefaultUniformBlock(GLsizei count,
uint32_t arrayIndex,
int componentCount,
const T *v,
const sh::BlockMemberInfo &layoutInfo,
angle::MemoryBuffer *uniformData)
{
const int elementSize = sizeof(T) * componentCount;
uint8_t *dst = uniformData->data() + layoutInfo.offset;
if (layoutInfo.arrayStride == 0 || layoutInfo.arrayStride == elementSize)
{
uint32_t arrayOffset = arrayIndex * layoutInfo.arrayStride;
uint8_t *writePtr = dst + arrayOffset;
ASSERT(writePtr + (elementSize * count) <= uniformData->data() + uniformData->size());
memcpy(writePtr, v, elementSize * count);
}
else
{
// Have to respect the arrayStride between each element of the array.
int maxIndex = arrayIndex + count;
for (int writeIndex = arrayIndex, readIndex = 0; writeIndex < maxIndex;
writeIndex++, readIndex++)
{
const int arrayOffset = writeIndex * layoutInfo.arrayStride;
uint8_t *writePtr = dst + arrayOffset;
const T *readPtr = v + (readIndex * componentCount);
ASSERT(writePtr + elementSize <= uniformData->data() + uniformData->size());
memcpy(writePtr, readPtr, elementSize);
}
}
}
template <typename T>
void ReadFromDefaultUniformBlock(int componentCount,
uint32_t arrayIndex,
T *dst,
const sh::BlockMemberInfo &layoutInfo,
const angle::MemoryBuffer *uniformData)
{
ASSERT(layoutInfo.offset != -1);
const int elementSize = sizeof(T) * componentCount;
const uint8_t *source = uniformData->data() + layoutInfo.offset;
if (layoutInfo.arrayStride == 0 || layoutInfo.arrayStride == elementSize)
{
const uint8_t *readPtr = source + arrayIndex * layoutInfo.arrayStride;
memcpy(dst, readPtr, elementSize);
}
else
{
// Have to respect the arrayStride between each element of the array.
const int arrayOffset = arrayIndex * layoutInfo.arrayStride;
const uint8_t *readPtr = source + arrayOffset;
memcpy(dst, readPtr, elementSize);
}
}
template <typename T>
void SetUniformImpl(const gl::ProgramExecutable *executable,
GLint location,
GLsizei count,
const T *v,
GLenum entryPointType,
DefaultUniformBlockMap *defaultUniformBlocks,
gl::ShaderBitSet *defaultUniformBlocksDirty)
{
const gl::VariableLocation &locationInfo = executable->getUniformLocations()[location];
const gl::LinkedUniform &linkedUniform = executable->getUniforms()[locationInfo.index];
ASSERT(!linkedUniform.isSampler());
if (linkedUniform.getType() == entryPointType)
{
for (const gl::ShaderType shaderType : executable->getLinkedShaderStages())
{
DefaultUniformBlockVk &uniformBlock = *(*defaultUniformBlocks)[shaderType];
const sh::BlockMemberInfo &layoutInfo = uniformBlock.uniformLayout[location];
// Assume an offset of -1 means the block is unused.
if (layoutInfo.offset == -1)
{
continue;
}
const GLint componentCount = linkedUniform.getElementComponents();
UpdateDefaultUniformBlock(count, locationInfo.arrayIndex, componentCount, v, layoutInfo,
&uniformBlock.uniformData);
defaultUniformBlocksDirty->set(shaderType);
}
}
else
{
for (const gl::ShaderType shaderType : executable->getLinkedShaderStages())
{
DefaultUniformBlockVk &uniformBlock = *(*defaultUniformBlocks)[shaderType];
const sh::BlockMemberInfo &layoutInfo = uniformBlock.uniformLayout[location];
// Assume an offset of -1 means the block is unused.
if (layoutInfo.offset == -1)
{
continue;
}
const GLint componentCount = linkedUniform.getElementComponents();
ASSERT(linkedUniform.getType() == gl::VariableBoolVectorType(entryPointType));
GLint initialArrayOffset =
locationInfo.arrayIndex * layoutInfo.arrayStride + layoutInfo.offset;
for (GLint i = 0; i < count; i++)
{
GLint elementOffset = i * layoutInfo.arrayStride + initialArrayOffset;
GLint *dst =
reinterpret_cast<GLint *>(uniformBlock.uniformData.data() + elementOffset);
const T *source = v + i * componentCount;
for (int c = 0; c < componentCount; c++)
{
dst[c] = (source[c] == static_cast<T>(0)) ? GL_FALSE : GL_TRUE;
}
}
defaultUniformBlocksDirty->set(shaderType);
}
}
}
template <int cols, int rows>
void SetUniformMatrixfv(const gl::ProgramExecutable *executable,
GLint location,
GLsizei count,
GLboolean transpose,
const GLfloat *value,
DefaultUniformBlockMap *defaultUniformBlocks,
gl::ShaderBitSet *defaultUniformBlocksDirty)
{
const gl::VariableLocation &locationInfo = executable->getUniformLocations()[location];
const gl::LinkedUniform &linkedUniform = executable->getUniforms()[locationInfo.index];
for (const gl::ShaderType shaderType : executable->getLinkedShaderStages())
{
DefaultUniformBlockVk &uniformBlock = *(*defaultUniformBlocks)[shaderType];
const sh::BlockMemberInfo &layoutInfo = uniformBlock.uniformLayout[location];
// Assume an offset of -1 means the block is unused.
if (layoutInfo.offset == -1)
{
continue;
}
SetFloatUniformMatrixGLSL<cols, rows>::Run(
locationInfo.arrayIndex, linkedUniform.getBasicTypeElementCount(), count, transpose,
value, uniformBlock.uniformData.data() + layoutInfo.offset);
defaultUniformBlocksDirty->set(shaderType);
}
}
vk::GraphicsPipelineSubset GetWarmUpSubset(const angle::FeaturesVk &features)
{
// Only build the shaders subset of the pipeline if VK_EXT_graphics_pipeline_library is
// supported.
return features.supportsGraphicsPipelineLibrary.enabled ? vk::GraphicsPipelineSubset::Shaders
: vk::GraphicsPipelineSubset::Complete;
}
} // namespace
class ProgramExecutableVk::WarmUpTaskCommon : public vk::Context, public LinkSubTask
{
public:
WarmUpTaskCommon(vk::Renderer *renderer) : vk::Context(renderer) {}
WarmUpTaskCommon(vk::Renderer *renderer,
ProgramExecutableVk *executableVk,
vk::PipelineRobustness pipelineRobustness,
vk::PipelineProtectedAccess pipelineProtectedAccess)
: vk::Context(renderer),
mExecutableVk(executableVk),
mPipelineRobustness(pipelineRobustness),
mPipelineProtectedAccess(pipelineProtectedAccess)
{}
~WarmUpTaskCommon() override = default;
void handleError(VkResult result,
const char *file,
const char *function,
unsigned int line) override
{
mErrorCode = result;
mErrorFile = file;
mErrorFunction = function;
mErrorLine = line;
}
void operator()() override { UNREACHABLE(); }
angle::Result getResult(const gl::Context *context, gl::InfoLog &infoLog) override
{
ContextVk *contextVk = vk::GetImpl(context);
return getResultImpl(contextVk, infoLog);
}
angle::Result getResultImpl(ContextVk *contextVk, gl::InfoLog &infoLog)
{
// Forward any errors
if (mErrorCode != VK_SUCCESS)
{
contextVk->handleError(mErrorCode, mErrorFile, mErrorFunction, mErrorLine);
return angle::Result::Stop;
}
// Accumulate relevant perf counters
const angle::VulkanPerfCounters &from = getPerfCounters();
angle::VulkanPerfCounters &to = contextVk->getPerfCounters();
to.pipelineCreationCacheHits += from.pipelineCreationCacheHits;
to.pipelineCreationCacheMisses += from.pipelineCreationCacheMisses;
to.pipelineCreationTotalCacheHitsDurationNs +=
from.pipelineCreationTotalCacheHitsDurationNs;
to.pipelineCreationTotalCacheMissesDurationNs +=
from.pipelineCreationTotalCacheMissesDurationNs;
return angle::Result::Continue;
}
protected:
void mergeProgramExecutablePipelineCacheToRenderer()
{
angle::Result mergeResult = mExecutableVk->mergePipelineCacheToRenderer(this);
// Treat error during merge as non fatal, log it and move on
if (mergeResult != angle::Result::Continue)
{
INFO() << "Error while merging to Renderer's pipeline cache";
}
}
// The front-end ensures that the program is not modified while the subtask is running, so it is
// safe to directly access the executable from this parallel job. Note that this is the reason
// why the front-end does not let the parallel job continue when a relink happens or the first
// draw with this program.
ProgramExecutableVk *mExecutableVk = nullptr;
const vk::PipelineRobustness mPipelineRobustness = vk::PipelineRobustness::NonRobust;
const vk::PipelineProtectedAccess mPipelineProtectedAccess =
vk::PipelineProtectedAccess::Unprotected;
// Error handling
VkResult mErrorCode = VK_SUCCESS;
const char *mErrorFile = nullptr;
const char *mErrorFunction = nullptr;
unsigned int mErrorLine = 0;
};
class ProgramExecutableVk::WarmUpComputeTask : public WarmUpTaskCommon
{
public:
WarmUpComputeTask(vk::Renderer *renderer,
ProgramExecutableVk *executableVk,
vk::PipelineRobustness pipelineRobustness,
vk::PipelineProtectedAccess pipelineProtectedAccess)
: WarmUpTaskCommon(renderer, executableVk, pipelineRobustness, pipelineProtectedAccess)
{}
~WarmUpComputeTask() override = default;
void operator()() override
{
angle::Result result = mExecutableVk->warmUpComputePipelineCache(this, mPipelineRobustness,
mPipelineProtectedAccess);
ASSERT((result == angle::Result::Continue) == (mErrorCode == VK_SUCCESS));
mergeProgramExecutablePipelineCacheToRenderer();
}
};
using SharedRenderPass = vk::AtomicRefCounted<vk::RenderPass>;
class ProgramExecutableVk::WarmUpGraphicsTask : public WarmUpTaskCommon
{
public:
WarmUpGraphicsTask(vk::Renderer *renderer,
ProgramExecutableVk *executableVk,
vk::PipelineRobustness pipelineRobustness,
vk::PipelineProtectedAccess pipelineProtectedAccess,
vk::GraphicsPipelineSubset subset,
const bool isSurfaceRotated,
const vk::GraphicsPipelineDesc &graphicsPipelineDesc,
SharedRenderPass *compatibleRenderPass,
vk::PipelineHelper *placeholderPipelineHelper)
: WarmUpTaskCommon(renderer, executableVk, pipelineRobustness, pipelineProtectedAccess),
mPipelineSubset(subset),
mIsSurfaceRotated(isSurfaceRotated),
mGraphicsPipelineDesc(graphicsPipelineDesc),
mWarmUpPipelineHelper(placeholderPipelineHelper),
mCompatibleRenderPass(compatibleRenderPass)
{
ASSERT(mCompatibleRenderPass);
mCompatibleRenderPass->addRef();
}
~WarmUpGraphicsTask() override = default;
void operator()() override
{
angle::Result result = mExecutableVk->warmUpGraphicsPipelineCache(
this, mPipelineRobustness, mPipelineProtectedAccess, mPipelineSubset, mIsSurfaceRotated,
mGraphicsPipelineDesc, mCompatibleRenderPass->get(), mWarmUpPipelineHelper);
ASSERT((result == angle::Result::Continue) == (mErrorCode == VK_SUCCESS));
// Release reference to shared renderpass. If this is the last reference -
// 1. merge ProgramExecutableVk's pipeline cache into the Renderer's cache
// 2. cleanup temporary renderpass
//
// Note: with dynamic rendering, |mCompatibleRenderPass| holds a VK_NULL_HANDLE, and it's
// just used as a ref count for this purpose.
const bool isLastWarmUpTask = mCompatibleRenderPass->getAndReleaseRef() == 1;
if (isLastWarmUpTask)
{
mergeProgramExecutablePipelineCacheToRenderer();
mCompatibleRenderPass->get().destroy(getDevice());
SafeDelete(mCompatibleRenderPass);
}
}
private:
vk::GraphicsPipelineSubset mPipelineSubset;
bool mIsSurfaceRotated;
vk::GraphicsPipelineDesc mGraphicsPipelineDesc;
vk::PipelineHelper *mWarmUpPipelineHelper;
// Temporary objects to clean up at the end
SharedRenderPass *mCompatibleRenderPass;
};
DefaultUniformBlockVk::DefaultUniformBlockVk() = default;
DefaultUniformBlockVk::~DefaultUniformBlockVk() = default;
// ShaderInfo implementation.
ShaderInfo::ShaderInfo() {}
ShaderInfo::~ShaderInfo() = default;
angle::Result ShaderInfo::initShaders(vk::Context *context,
const gl::ShaderBitSet &linkedShaderStages,
const gl::ShaderMap<const angle::spirv::Blob *> &spirvBlobs,
const ShaderInterfaceVariableInfoMap &variableInfoMap,
bool isGLES1)
{
clear();
for (gl::ShaderType shaderType : gl::AllShaderTypes())
{
if (spirvBlobs[shaderType] != nullptr)
{
mSpirvBlobs[shaderType] = *spirvBlobs[shaderType];
}
}
// Assert that SPIR-V transformation is correct, even if the test never issues a draw call.
// Don't validate GLES1 programs because they are always created right before a draw, so they
// will naturally be validated. This improves GLES1 test run times.
if (!isGLES1)
{
ASSERT(ValidateTransformedSpirV(context, linkedShaderStages, variableInfoMap, mSpirvBlobs));
}
mIsInitialized = true;
return angle::Result::Continue;
}
void ShaderInfo::initShaderFromProgram(gl::ShaderType shaderType,
const ShaderInfo &programShaderInfo)
{
mSpirvBlobs[shaderType] = programShaderInfo.mSpirvBlobs[shaderType];
mIsInitialized = true;
}
void ShaderInfo::clear()
{
for (angle::spirv::Blob &spirvBlob : mSpirvBlobs)
{
spirvBlob.clear();
}
mIsInitialized = false;
}
void ShaderInfo::load(gl::BinaryInputStream *stream)
{
clear();
// Read in shader codes for all shader types
for (gl::ShaderType shaderType : gl::AllShaderTypes())
{
stream->readVector(&mSpirvBlobs[shaderType]);
}
mIsInitialized = true;
}
void ShaderInfo::save(gl::BinaryOutputStream *stream)
{
ASSERT(valid());
// Write out shader codes for all shader types
for (gl::ShaderType shaderType : gl::AllShaderTypes())
{
stream->writeVector(mSpirvBlobs[shaderType]);
}
}
// ProgramInfo implementation.
ProgramInfo::ProgramInfo() {}
ProgramInfo::~ProgramInfo() = default;
angle::Result ProgramInfo::initProgram(vk::Context *context,
gl::ShaderType shaderType,
bool isLastPreFragmentStage,
bool isTransformFeedbackProgram,
const ShaderInfo &shaderInfo,
ProgramTransformOptions optionBits,
const ShaderInterfaceVariableInfoMap &variableInfoMap)
{
const gl::ShaderMap<angle::spirv::Blob> &originalSpirvBlobs = shaderInfo.getSpirvBlobs();
const angle::spirv::Blob &originalSpirvBlob = originalSpirvBlobs[shaderType];
gl::ShaderMap<angle::spirv::Blob> transformedSpirvBlobs;
angle::spirv::Blob &transformedSpirvBlob = transformedSpirvBlobs[shaderType];
SpvTransformOptions options;
options.shaderType = shaderType;
options.isLastPreFragmentStage = isLastPreFragmentStage;
options.isTransformFeedbackStage = isLastPreFragmentStage && isTransformFeedbackProgram &&
!optionBits.removeTransformFeedbackEmulation;
options.isTransformFeedbackEmulated = context->getFeatures().emulateTransformFeedback.enabled;
options.isMultisampledFramebufferFetch =
optionBits.multiSampleFramebufferFetch && shaderType == gl::ShaderType::Fragment;
options.enableSampleShading = optionBits.enableSampleShading;
options.useSpirvVaryingPrecisionFixer =
context->getFeatures().varyingsRequireMatchingPrecisionInSpirv.enabled;
ANGLE_TRY(
SpvTransformSpirvCode(options, variableInfoMap, originalSpirvBlob, &transformedSpirvBlob));
ANGLE_TRY(vk::InitShaderModule(context, &mShaders[shaderType].get(),
transformedSpirvBlob.data(),
transformedSpirvBlob.size() * sizeof(uint32_t)));
mProgramHelper.setShader(shaderType, &mShaders[shaderType]);
return angle::Result::Continue;
}
void ProgramInfo::release(ContextVk *contextVk)
{
mProgramHelper.release(contextVk);
for (vk::RefCounted<vk::ShaderModule> &shader : mShaders)
{
shader.get().destroy(contextVk->getDevice());
}
}
ProgramExecutableVk::ProgramExecutableVk(const gl::ProgramExecutable *executable)
: ProgramExecutableImpl(executable),
mImmutableSamplersMaxDescriptorCount(1),
mUniformBufferDescriptorType(VK_DESCRIPTOR_TYPE_MAX_ENUM),
mDynamicUniformDescriptorOffsets{},
mValidGraphicsPermutations{},
mValidComputePermutations{}
{
mDescriptorSets.fill(VK_NULL_HANDLE);
for (std::shared_ptr<DefaultUniformBlockVk> &defaultBlock : mDefaultUniformBlocks)
{
defaultBlock = std::make_shared<DefaultUniformBlockVk>();
}
}
ProgramExecutableVk::~ProgramExecutableVk()
{
ASSERT(!mPipelineCache.valid());
}
void ProgramExecutableVk::destroy(const gl::Context *context)
{
reset(vk::GetImpl(context));
}
void ProgramExecutableVk::resetLayout(ContextVk *contextVk)
{
if (!mPipelineLayout.valid())
{
ASSERT(mValidGraphicsPermutations.none());
ASSERT(mValidComputePermutations.none());
return;
}
waitForPostLinkTasksImpl(contextVk);
for (auto &descriptorSetLayout : mDescriptorSetLayouts)
{
descriptorSetLayout.reset();
}
mImmutableSamplersMaxDescriptorCount = 1;
mImmutableSamplerIndexMap.clear();
mDescriptorSets.fill(VK_NULL_HANDLE);
for (vk::RefCountedDescriptorPoolBinding &binding : mDescriptorPoolBindings)
{
binding.reset();
}
for (vk::DescriptorPoolPointer &pool : mDescriptorPools)
{
pool.reset();
}
// Initialize with an invalid BufferSerial
mCurrentDefaultUniformBufferSerial = vk::BufferSerial();
for (size_t index : mValidGraphicsPermutations)
{
mCompleteGraphicsPipelines[index].release(contextVk);
mShadersGraphicsPipelines[index].release(contextVk);
// Program infos and pipeline layout must be released after pipelines are; they might be
// having pending jobs that are referencing them.
mGraphicsProgramInfos[index].release(contextVk);
}
mValidGraphicsPermutations.reset();
for (size_t index : mValidComputePermutations)
{
mComputePipelines[index].release(contextVk);
}
mComputeProgramInfo.release(contextVk);
mValidComputePermutations.reset();
mPipelineLayout.reset();
contextVk->onProgramExecutableReset(this);
}
void ProgramExecutableVk::reset(ContextVk *contextVk)
{
resetLayout(contextVk);
if (mPipelineCache.valid())
{
mPipelineCache.destroy(contextVk->getDevice());
}
}
angle::Result ProgramExecutableVk::initializePipelineCache(vk::Context *context,
bool compressed,
const std::vector<uint8_t> &pipelineData)
{
ASSERT(!mPipelineCache.valid());
size_t dataSize = pipelineData.size();
const uint8_t *dataPointer = pipelineData.data();
angle::MemoryBuffer uncompressedData;
if (compressed)
{
if (!angle::DecompressBlob(dataPointer, dataSize, kMaxLocalPipelineCacheSize,
&uncompressedData))
{
return angle::Result::Stop;
}
dataSize = uncompressedData.size();
dataPointer = uncompressedData.data();
}
VkPipelineCacheCreateInfo pipelineCacheCreateInfo = {};
pipelineCacheCreateInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_CACHE_CREATE_INFO;
pipelineCacheCreateInfo.initialDataSize = dataSize;
pipelineCacheCreateInfo.pInitialData = dataPointer;
ANGLE_VK_TRY(context, mPipelineCache.init(context->getDevice(), pipelineCacheCreateInfo));
// Merge the pipeline cache into Renderer's.
if (context->getFeatures().mergeProgramPipelineCachesToGlobalCache.enabled)
{
ANGLE_TRY(context->getRenderer()->mergeIntoPipelineCache(context, mPipelineCache));
}
return angle::Result::Continue;
}
angle::Result ProgramExecutableVk::ensurePipelineCacheInitialized(vk::Context *context)
{
if (!mPipelineCache.valid())
{
VkPipelineCacheCreateInfo pipelineCacheCreateInfo = {};
pipelineCacheCreateInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_CACHE_CREATE_INFO;
ANGLE_VK_TRY(context, mPipelineCache.init(context->getDevice(), pipelineCacheCreateInfo));
}
return angle::Result::Continue;
}
angle::Result ProgramExecutableVk::load(ContextVk *contextVk,
bool isSeparable,
gl::BinaryInputStream *stream,
egl::CacheGetResult *resultOut)
{
mVariableInfoMap.load(stream);
mOriginalShaderInfo.load(stream);
// Deserializes the uniformLayout data of mDefaultUniformBlocks
for (gl::ShaderType shaderType : gl::AllShaderTypes())
{
stream->readVector(&mDefaultUniformBlocks[shaderType]->uniformLayout);
}
// Deserializes required uniform block memory sizes
gl::ShaderMap<size_t> requiredBufferSize;
stream->readPackedEnumMap(&requiredBufferSize);
if (!isSeparable)
{
size_t compressedPipelineDataSize = 0;
stream->readInt<size_t>(&compressedPipelineDataSize);
std::vector<uint8_t> compressedPipelineData(compressedPipelineDataSize);
if (compressedPipelineDataSize > 0)
{
bool compressedData = false;
stream->readBool(&compressedData);
stream->readBytes(compressedPipelineData.data(), compressedPipelineDataSize);
// Initialize the pipeline cache based on cached data.
ANGLE_TRY(initializePipelineCache(contextVk, compressedData, compressedPipelineData));
}
}
// Initialize and resize the mDefaultUniformBlocks' memory
ANGLE_TRY(resizeUniformBlockMemory(contextVk, requiredBufferSize));
resetLayout(contextVk);
ANGLE_TRY(createPipelineLayout(contextVk, &contextVk->getPipelineLayoutCache(),
&contextVk->getDescriptorSetLayoutCache(), nullptr));
ANGLE_TRY(initializeDescriptorPools(contextVk, &contextVk->getDescriptorSetLayoutCache(),
&contextVk->getMetaDescriptorPools()));
*resultOut = egl::CacheGetResult::Success;
return angle::Result::Continue;
}
void ProgramExecutableVk::save(ContextVk *contextVk,
bool isSeparable,
gl::BinaryOutputStream *stream)
{
mVariableInfoMap.save(stream);
mOriginalShaderInfo.save(stream);
// Serializes the uniformLayout data of mDefaultUniformBlocks
for (gl::ShaderType shaderType : gl::AllShaderTypes())
{
stream->writeVector(mDefaultUniformBlocks[shaderType]->uniformLayout);
}
// Serializes required uniform block memory sizes
gl::ShaderMap<size_t> uniformDataSize;
for (gl::ShaderType shaderType : gl::AllShaderTypes())
{
uniformDataSize[shaderType] = mDefaultUniformBlocks[shaderType]->uniformData.size();
}
stream->writePackedEnumMap(uniformDataSize);
// Need to wait for warm up tasks to complete.
waitForPostLinkTasksImpl(contextVk);
// Compress and save mPipelineCache. Separable programs don't warm up the cache, while program
// pipelines do. However, currently ANGLE doesn't sync program pipelines to cache. ANGLE could
// potentially use VK_EXT_graphics_pipeline_library to create separate pipelines for
// pre-rasterization and fragment subsets, but currently those subsets are bundled together.
if (!isSeparable)
{
angle::MemoryBuffer cacheData;
GetPipelineCacheData(contextVk, mPipelineCache, &cacheData);
stream->writeInt(cacheData.size());
if (cacheData.size() > 0)
{
stream->writeBool(contextVk->getFeatures().enablePipelineCacheDataCompression.enabled);
stream->writeBytes(cacheData.data(), cacheData.size());
}
}
}
void ProgramExecutableVk::clearVariableInfoMap()
{
mVariableInfoMap.clear();
}
angle::Result ProgramExecutableVk::getPipelineCacheWarmUpTasks(
vk::Renderer *renderer,
vk::PipelineRobustness pipelineRobustness,
vk::PipelineProtectedAccess pipelineProtectedAccess,
std::vector<std::shared_ptr<LinkSubTask>> *postLinkSubTasksOut)
{
ASSERT(!postLinkSubTasksOut || postLinkSubTasksOut->empty());
const vk::GraphicsPipelineSubset subset = GetWarmUpSubset(renderer->getFeatures());
bool isCompute = false;
angle::FixedVector<bool, 2> surfaceRotationVariations = {false};
vk::GraphicsPipelineDesc *graphicsPipelineDesc = nullptr;
vk::RenderPass compatibleRenderPass;
WarmUpTaskCommon prepForWarmUpContext(renderer);
ANGLE_TRY(prepareForWarmUpPipelineCache(
&prepForWarmUpContext, pipelineRobustness, pipelineProtectedAccess, subset, &isCompute,
&surfaceRotationVariations, &graphicsPipelineDesc, &compatibleRenderPass));
std::vector<std::shared_ptr<rx::LinkSubTask>> warmUpSubTasks;
if (isCompute)
{
ASSERT(!compatibleRenderPass.valid());
warmUpSubTasks.push_back(std::make_shared<WarmUpComputeTask>(
renderer, this, pipelineRobustness, pipelineProtectedAccess));
}
else
{
ProgramTransformOptions transformOptions = {};
SharedRenderPass *sharedRenderPass = new SharedRenderPass(std::move(compatibleRenderPass));
for (bool surfaceRotation : surfaceRotationVariations)
{
// Add a placeholder entry in GraphicsPipelineCache
transformOptions.surfaceRotation = surfaceRotation;
const uint8_t programIndex = transformOptions.permutationIndex;
vk::PipelineHelper *pipelineHelper = nullptr;
if (subset == vk::GraphicsPipelineSubset::Complete)
{
CompleteGraphicsPipelineCache &pipelines = mCompleteGraphicsPipelines[programIndex];
pipelines.populate(mWarmUpGraphicsPipelineDesc, vk::Pipeline(), &pipelineHelper);
}
else
{
ASSERT(subset == vk::GraphicsPipelineSubset::Shaders);
ShadersGraphicsPipelineCache &pipelines = mShadersGraphicsPipelines[programIndex];
pipelines.populate(mWarmUpGraphicsPipelineDesc, vk::Pipeline(), &pipelineHelper);
}
warmUpSubTasks.push_back(std::make_shared<WarmUpGraphicsTask>(
renderer, this, pipelineRobustness, pipelineProtectedAccess, subset,
surfaceRotation, *graphicsPipelineDesc, sharedRenderPass, pipelineHelper));
}
}
// If the caller hasn't provided a valid async task container, inline the warmUp tasks.
if (postLinkSubTasksOut)
{
*postLinkSubTasksOut = std::move(warmUpSubTasks);
}
else
{
for (std::shared_ptr<rx::LinkSubTask> &task : warmUpSubTasks)
{
(*task)();
}
warmUpSubTasks.clear();
}
ASSERT(warmUpSubTasks.empty());
return angle::Result::Continue;
}
angle::Result ProgramExecutableVk::prepareForWarmUpPipelineCache(
vk::Context *context,
vk::PipelineRobustness pipelineRobustness,
vk::PipelineProtectedAccess pipelineProtectedAccess,
vk::GraphicsPipelineSubset subset,
bool *isComputeOut,
angle::FixedVector<bool, 2> *surfaceRotationVariationsOut,
vk::GraphicsPipelineDesc **graphicsPipelineDescOut,
vk::RenderPass *renderPassOut)
{
ASSERT(isComputeOut);
ASSERT(surfaceRotationVariationsOut);
ASSERT(graphicsPipelineDescOut);
ASSERT(renderPassOut);
ASSERT(context->getFeatures().warmUpPipelineCacheAtLink.enabled);
ANGLE_TRY(ensurePipelineCacheInitialized(context));
*isComputeOut = false;
const bool isCompute = mExecutable->hasLinkedShaderStage(gl::ShaderType::Compute);
if (isCompute)
{
// Initialize compute program.
vk::ComputePipelineOptions pipelineOptions =
vk::GetComputePipelineOptions(pipelineRobustness, pipelineProtectedAccess);
ANGLE_TRY(
initComputeProgram(context, &mComputeProgramInfo, mVariableInfoMap, pipelineOptions));
*isComputeOut = true;
return angle::Result::Continue;
}
// It is only at drawcall time that we will have complete information required to build the
// graphics pipeline descriptor. Use the most "commonly seen" state values and create the
// pipeline.
gl::PrimitiveMode mode = (mExecutable->hasLinkedShaderStage(gl::ShaderType::TessControl) ||
mExecutable->hasLinkedShaderStage(gl::ShaderType::TessEvaluation))
? gl::PrimitiveMode::Patches
: gl::PrimitiveMode::TriangleStrip;
SetupDefaultPipelineState(context, *mExecutable, mode, pipelineRobustness,
pipelineProtectedAccess, subset, &mWarmUpGraphicsPipelineDesc);
// Create a temporary compatible RenderPass. The render pass cache in ContextVk cannot be used
// because this function may be called from a worker thread.
vk::AttachmentOpsArray ops;
RenderPassCache::InitializeOpsForCompatibleRenderPass(
mWarmUpGraphicsPipelineDesc.getRenderPassDesc(), &ops);
if (!context->getFeatures().preferDynamicRendering.enabled)
{
ANGLE_TRY(RenderPassCache::MakeRenderPass(
context, mWarmUpGraphicsPipelineDesc.getRenderPassDesc(), ops, renderPassOut, nullptr));
}
*graphicsPipelineDescOut = &mWarmUpGraphicsPipelineDesc;
// Variations that definitely matter:
//
// - PreRotation: It's a boolean specialization constant
// - Depth correction: It's a SPIR-V transformation
//
// There are a number of states that are not currently dynamic (and may never be, such as sample
// shading), but pre-creating shaders for them is impractical. Most such state is likely unused
// by most applications, but variations can be added here for certain apps that are known to
// benefit from it.
*surfaceRotationVariationsOut = {false};
if (context->getFeatures().enablePreRotateSurfaces.enabled &&
!context->getFeatures().preferDriverUniformOverSpecConst.enabled)
{
surfaceRotationVariationsOut->push_back(true);
}
ProgramTransformOptions transformOptions = {};
for (bool rotation : *surfaceRotationVariationsOut)
{
// Initialize graphics programs.
transformOptions.surfaceRotation = rotation;
ANGLE_TRY(initGraphicsShaderPrograms(context, transformOptions));
}
return angle::Result::Continue;
}
angle::Result ProgramExecutableVk::warmUpComputePipelineCache(
vk::Context *context,
vk::PipelineRobustness pipelineRobustness,
vk::PipelineProtectedAccess pipelineProtectedAccess)
{
ANGLE_TRACE_EVENT0("gpu.angle", "ProgramExecutableVk::warmUpComputePipelineCache");
// This method assumes that all the state necessary to create a compute pipeline has already
// been setup by the caller. Assert that all required state is valid so all that is left will
// be the call to `vkCreateComputePipelines`
// Make sure the shader module for compute shader stage is valid.
ASSERT(mComputeProgramInfo.valid(gl::ShaderType::Compute));
// No synchronization necessary since mPipelineCache is internally synchronized.
vk::PipelineCacheAccess pipelineCache;
pipelineCache.init(&mPipelineCache, nullptr);
// There is no state associated with compute programs, so only one pipeline needs creation
// to warm up the cache.
vk::PipelineHelper *pipeline = nullptr;
ANGLE_TRY(getOrCreateComputePipeline(context, &pipelineCache, PipelineSource::WarmUp,
pipelineRobustness, pipelineProtectedAccess, &pipeline));
return angle::Result::Continue;
}
angle::Result ProgramExecutableVk::warmUpGraphicsPipelineCache(
vk::Context *context,
vk::PipelineRobustness pipelineRobustness,
vk::PipelineProtectedAccess pipelineProtectedAccess,
vk::GraphicsPipelineSubset subset,
const bool isSurfaceRotated,
const vk::GraphicsPipelineDesc &graphicsPipelineDesc,
const vk::RenderPass &renderPass,
vk::PipelineHelper *placeholderPipelineHelper)
{
ANGLE_TRACE_EVENT0("gpu.angle", "ProgramExecutableVk::warmUpGraphicsPipelineCache");
ASSERT(placeholderPipelineHelper && !placeholderPipelineHelper->valid());
// No synchronization necessary since mPipelineCache is internally synchronized.
vk::PipelineCacheAccess pipelineCache;
pipelineCache.init(&mPipelineCache, nullptr);
const vk::GraphicsPipelineDesc *descPtr = nullptr;
ProgramTransformOptions transformOptions = {};
transformOptions.surfaceRotation = isSurfaceRotated;
ANGLE_TRY(createGraphicsPipelineImpl(context, transformOptions, subset, &pipelineCache,
PipelineSource::WarmUp, graphicsPipelineDesc, renderPass,
&descPtr, &placeholderPipelineHelper));
ASSERT(placeholderPipelineHelper->valid());
return angle::Result::Continue;
}
void ProgramExecutableVk::waitForPostLinkTasksImpl(ContextVk *contextVk)
{
const std::vector<std::shared_ptr<rx::LinkSubTask>> &postLinkSubTasks =
mExecutable->getPostLinkSubTasks();
if (postLinkSubTasks.empty())
{
return;
}
// Wait for all post-link tasks to finish
angle::WaitableEvent::WaitMany(&mExecutable->getPostLinkSubTaskWaitableEvents());
// Get results and clean up
for (const std::shared_ptr<rx::LinkSubTask> &task : postLinkSubTasks)
{
WarmUpTaskCommon *warmUpTask = static_cast<WarmUpTaskCommon *>(task.get());
// As these tasks can be run post-link, their results are ignored. Failure is harmless, but
// more importantly the error (effectively due to a link event) may not be allowed through
// the entry point that results in this call.
gl::InfoLog infoLog;
angle::Result result = warmUpTask->getResultImpl(contextVk, infoLog);
if (result != angle::Result::Continue)
{
ANGLE_PERF_WARNING(contextVk->getDebug(), GL_DEBUG_SEVERITY_LOW,
"Post-link task unexpectedly failed. Performance may degrade, or "
"device may soon be lost");
}
}
mExecutable->onPostLinkTasksComplete();
}
void ProgramExecutableVk::waitForGraphicsPostLinkTasks(
ContextVk *contextVk,
const vk::GraphicsPipelineDesc ¤tGraphicsPipelineDesc)
{
ASSERT(mExecutable->hasLinkedShaderStage(gl::ShaderType::Vertex));
if (mExecutable->getPostLinkSubTasks().empty())
{
return;
}
const vk::GraphicsPipelineSubset subset = GetWarmUpSubset(contextVk->getFeatures());
if (!mWarmUpGraphicsPipelineDesc.keyEqual(currentGraphicsPipelineDesc, subset))
{
// The GraphicsPipelineDesc used for warmup differs from the one used by the draw call.
// There is no need to wait for the warmup tasks to complete.
ANGLE_PERF_WARNING(
contextVk->getDebug(), GL_DEBUG_SEVERITY_LOW,
"GraphicsPipelineDesc used for warmup differs from the one used by draw.");
// If the warm up tasks are finished anyway, let |waitForPostLinkTasksImpl| clean them up.
if (!angle::WaitableEvent::AllReady(&mExecutable->getPostLinkSubTaskWaitableEvents()))
{
return;
}
}
waitForPostLinkTasksImpl(contextVk);
}
angle::Result ProgramExecutableVk::mergePipelineCacheToRenderer(vk::Context *context) const
{
// Merge the cache with Renderer's
if (context->getFeatures().mergeProgramPipelineCachesToGlobalCache.enabled)
{
ANGLE_TRACE_EVENT0("gpu.angle", "ProgramExecutableVk::mergePipelineCacheToRenderer");
ANGLE_TRY(context->getRenderer()->mergeIntoPipelineCache(context, mPipelineCache));
}
return angle::Result::Continue;
}
void ProgramExecutableVk::addInterfaceBlockDescriptorSetDesc(
const std::vector<gl::InterfaceBlock> &blocks,
gl::ShaderBitSet shaderTypes,
VkDescriptorType descType,
vk::DescriptorSetLayoutDesc *descOut)
{
for (uint32_t bufferIndex = 0, arraySize = 0; bufferIndex < blocks.size();
bufferIndex += arraySize)
{
gl::InterfaceBlock block = blocks[bufferIndex];
arraySize = GetInterfaceBlockArraySize(blocks, bufferIndex);
if (block.activeShaders().none())
{
continue;
}
const gl::ShaderType firstShaderType = block.getFirstActiveShaderType();
const ShaderInterfaceVariableInfo &info =
mVariableInfoMap.getVariableById(firstShaderType, block.getId(firstShaderType));
const VkShaderStageFlags activeStages = gl_vk::GetShaderStageFlags(info.activeStages);
descOut->addBinding(info.binding, descType, arraySize, activeStages, nullptr);
}
}
void ProgramExecutableVk::addAtomicCounterBufferDescriptorSetDesc(
const std::vector<gl::AtomicCounterBuffer> &atomicCounterBuffers,
vk::DescriptorSetLayoutDesc *descOut)
{
if (atomicCounterBuffers.empty())
{
return;
}
const ShaderInterfaceVariableInfo &info =
mVariableInfoMap.getAtomicCounterInfo(atomicCounterBuffers[0].getFirstActiveShaderType());
VkShaderStageFlags activeStages = gl_vk::GetShaderStageFlags(info.activeStages);
// A single storage buffer array is used for all stages for simplicity.
descOut->addBinding(info.binding, vk::kStorageBufferDescriptorType,
gl::IMPLEMENTATION_MAX_ATOMIC_COUNTER_BUFFER_BINDINGS, activeStages,
nullptr);
}
void ProgramExecutableVk::addImageDescriptorSetDesc(vk::DescriptorSetLayoutDesc *descOut)
{
const std::vector<gl::ImageBinding> &imageBindings = mExecutable->getImageBindings();
const std::vector<gl::LinkedUniform> &uniforms = mExecutable->getUniforms();
for (uint32_t imageIndex = 0; imageIndex < imageBindings.size(); ++imageIndex)
{
uint32_t uniformIndex = mExecutable->getUniformIndexFromImageIndex(imageIndex);
const gl::LinkedUniform &imageUniform = uniforms[uniformIndex];
// 2D arrays are split into multiple 1D arrays when generating LinkedUniforms. Since they
// are flattened into one array, ignore the nonzero elements and expand the array to the
// total array size.
if (imageUniform.activeShaders().none() || imageUniform.getOuterArrayOffset() > 0)
{
ASSERT(gl::SamplerNameContainsNonZeroArrayElement(
mExecutable->getUniformNameByIndex(uniformIndex)));
continue;
}
ASSERT(!gl::SamplerNameContainsNonZeroArrayElement(
mExecutable->getUniformNameByIndex(uniformIndex)));
// The front-end always binds array image units sequentially.
const gl::ImageBinding &imageBinding = imageBindings[imageIndex];
uint32_t arraySize = static_cast<uint32_t>(imageBinding.boundImageUnits.size());
arraySize *= imageUniform.getOuterArraySizeProduct();
const gl::ShaderType firstShaderType = imageUniform.getFirstActiveShaderType();
const ShaderInterfaceVariableInfo &info =
mVariableInfoMap.getVariableById(firstShaderType, imageUniform.getId(firstShaderType));
const VkShaderStageFlags activeStages = gl_vk::GetShaderStageFlags(info.activeStages);
const VkDescriptorType descType = imageBinding.textureType == gl::TextureType::Buffer
? VK_DESCRIPTOR_TYPE_STORAGE_TEXEL_BUFFER
: VK_DESCRIPTOR_TYPE_STORAGE_IMAGE;
descOut->addBinding(info.binding, descType, arraySize, activeStages, nullptr);
}
}
void ProgramExecutableVk::addInputAttachmentDescriptorSetDesc(vk::DescriptorSetLayoutDesc *descOut)
{
if (!mExecutable->getLinkedShaderStages()[gl::ShaderType::Fragment])
{
return;
}
if (!mExecutable->usesFramebufferFetch())
{
return;
}
const uint32_t firstInputAttachment =
static_cast<uint32_t>(mExecutable->getFragmentInoutIndices().first());
const ShaderInterfaceVariableInfo &baseInfo = mVariableInfoMap.getVariableById(
gl::ShaderType::Fragment, sh::vk::spirv::kIdInputAttachment0 + firstInputAttachment);
uint32_t baseBinding = baseInfo.binding - firstInputAttachment;
for (uint32_t colorIndex = 0; colorIndex < gl::IMPLEMENTATION_MAX_DRAW_BUFFERS; ++colorIndex)
{
descOut->addBinding(baseBinding, VK_DESCRIPTOR_TYPE_INPUT_ATTACHMENT, 1,
VK_SHADER_STAGE_FRAGMENT_BIT, nullptr);
baseBinding++;
}
}
angle::Result ProgramExecutableVk::addTextureDescriptorSetDesc(
vk::Context *context,
const gl::ActiveTextureArray<TextureVk *> *activeTextures,
vk::DescriptorSetLayoutDesc *descOut)
{
const std::vector<gl::SamplerBinding> &samplerBindings = mExecutable->getSamplerBindings();
const std::vector<gl::LinkedUniform> &uniforms = mExecutable->getUniforms();
const std::vector<GLuint> &samplerBoundTextureUnits =
mExecutable->getSamplerBoundTextureUnits();
for (uint32_t samplerIndex = 0; samplerIndex < samplerBindings.size(); ++samplerIndex)
{
uint32_t uniformIndex = mExecutable->getUniformIndexFromSamplerIndex(samplerIndex);
const gl::LinkedUniform &samplerUniform = uniforms[uniformIndex];
// 2D arrays are split into multiple 1D arrays when generating LinkedUniforms. Since they
// are flattened into one array, ignore the nonzero elements and expand the array to the
// total array size.
if (samplerUniform.activeShaders().none() || samplerUniform.getOuterArrayOffset() > 0)
{
ASSERT(gl::SamplerNameContainsNonZeroArrayElement(
mExecutable->getUniformNameByIndex(uniformIndex)));
continue;
}
ASSERT(!gl::SamplerNameContainsNonZeroArrayElement(
mExecutable->getUniformNameByIndex(uniformIndex)));
// The front-end always binds array sampler units sequentially.
const gl::SamplerBinding &samplerBinding = samplerBindings[samplerIndex];
uint32_t arraySize = static_cast<uint32_t>(samplerBinding.textureUnitsCount);
arraySize *= samplerUniform.getOuterArraySizeProduct();
const gl::ShaderType firstShaderType = samplerUniform.getFirstActiveShaderType();
const ShaderInterfaceVariableInfo &info = mVariableInfoMap.getVariableById(
firstShaderType, samplerUniform.getId(firstShaderType));
const VkShaderStageFlags activeStages = gl_vk::GetShaderStageFlags(info.activeStages);
// TODO: https://issuetracker.google.com/issues/158215272: how do we handle array of
// immutable samplers?
GLuint textureUnit = samplerBinding.getTextureUnit(samplerBoundTextureUnits, 0);
if (activeTextures != nullptr &&
(*activeTextures)[textureUnit]->getImage().hasImmutableSampler())
{
ASSERT(samplerBinding.textureUnitsCount == 1);
// In the case of samplerExternal2DY2YEXT, we need
// samplerYcbcrConversion object with IDENTITY conversion model
bool isSamplerExternalY2Y =
samplerBinding.samplerType == GL_SAMPLER_EXTERNAL_2D_Y2Y_EXT;
// Always take the texture's sampler, that's only way to get to yuv conversion for
// externalFormat
const TextureVk *textureVk = (*activeTextures)[textureUnit];
const vk::Sampler &immutableSampler = textureVk->getSampler(isSamplerExternalY2Y).get();
descOut->addBinding(info.binding, VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER, arraySize,
activeStages, &immutableSampler);
const vk::ImageHelper &image = textureVk->getImage();
const vk::YcbcrConversionDesc ycbcrConversionDesc =
isSamplerExternalY2Y ? image.getY2YConversionDesc()
: image.getYcbcrConversionDesc();
mImmutableSamplerIndexMap[ycbcrConversionDesc] = samplerIndex;
// The Vulkan spec has the following note -
// All descriptors in a binding use the same maximum
// combinedImageSamplerDescriptorCount descriptors to allow implementations to use a
// uniform stride for dynamic indexing of the descriptors in the binding.
uint64_t externalFormat = image.getExternalFormat();
uint32_t formatDescriptorCount = 0;
vk::Renderer *renderer = context->getRenderer();
if (externalFormat != 0)
{
ANGLE_TRY(renderer->getFormatDescriptorCountForExternalFormat(
context, externalFormat, &formatDescriptorCount));
}
else
{
VkFormat vkFormat = image.getActualVkFormat();
ASSERT(vkFormat != 0);
ANGLE_TRY(renderer->getFormatDescriptorCountForVkFormat(context, vkFormat,
&formatDescriptorCount));
}
ASSERT(formatDescriptorCount > 0);
mImmutableSamplersMaxDescriptorCount =
std::max(mImmutableSamplersMaxDescriptorCount, formatDescriptorCount);
}
else
{
const VkDescriptorType descType = samplerBinding.textureType == gl::TextureType::Buffer
? VK_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER
: VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER;
descOut->addBinding(info.binding, descType, arraySize, activeStages, nullptr);
}
}
return angle::Result::Continue;
}
void ProgramExecutableVk::initializeWriteDescriptorDesc(vk::Context *context)
{
const gl::ShaderBitSet &linkedShaderStages = mExecutable->getLinkedShaderStages();
// Update mShaderResourceWriteDescriptorDescBuilder
mShaderResourceWriteDescriptorDescs.reset();
mShaderResourceWriteDescriptorDescs.updateShaderBuffers(
mVariableInfoMap, mExecutable->getUniformBlocks(), getUniformBufferDescriptorType());
mShaderResourceWriteDescriptorDescs.updateShaderBuffers(
mVariableInfoMap, mExecutable->getShaderStorageBlocks(), getStorageBufferDescriptorType());
mShaderResourceWriteDescriptorDescs.updateAtomicCounters(
mVariableInfoMap, mExecutable->getAtomicCounterBuffers());
mShaderResourceWriteDescriptorDescs.updateImages(*mExecutable, mVariableInfoMap);
mShaderResourceWriteDescriptorDescs.updateDynamicDescriptorsCount();
// Update mTextureWriteDescriptors
mTextureWriteDescriptorDescs.reset();
mTextureWriteDescriptorDescs.updateExecutableActiveTextures(mVariableInfoMap, *mExecutable);
mTextureWriteDescriptorDescs.updateDynamicDescriptorsCount();
// Update mDefaultUniformWriteDescriptors
mDefaultUniformWriteDescriptorDescs.reset();
mDefaultUniformWriteDescriptorDescs.updateDefaultUniform(linkedShaderStages, mVariableInfoMap,
*mExecutable);
mDefaultUniformWriteDescriptorDescs.updateDynamicDescriptorsCount();
mDefaultUniformAndXfbWriteDescriptorDescs.reset();
if (mExecutable->hasTransformFeedbackOutput() &&
context->getRenderer()->getFeatures().emulateTransformFeedback.enabled)
{
// Update mDefaultUniformAndXfbWriteDescriptorDescs for the emulation code path.
mDefaultUniformAndXfbWriteDescriptorDescs.updateDefaultUniform(
linkedShaderStages, mVariableInfoMap, *mExecutable);
if (linkedShaderStages[gl::ShaderType::Vertex])
{
mDefaultUniformAndXfbWriteDescriptorDescs.updateTransformFeedbackWrite(mVariableInfoMap,
*mExecutable);
}
mDefaultUniformAndXfbWriteDescriptorDescs.updateDynamicDescriptorsCount();
}
else
{
// Otherwise it will be the same as default uniform
mDefaultUniformAndXfbWriteDescriptorDescs = mDefaultUniformWriteDescriptorDescs;
}
}
ProgramTransformOptions ProgramExecutableVk::getTransformOptions(
ContextVk *contextVk,
const vk::GraphicsPipelineDesc &desc)
{
ProgramTransformOptions transformOptions = {};
transformOptions.surfaceRotation = desc.getSurfaceRotation();
transformOptions.removeTransformFeedbackEmulation =
contextVk->getFeatures().emulateTransformFeedback.enabled &&
!contextVk->getState().isTransformFeedbackActiveUnpaused();
FramebufferVk *drawFrameBuffer = vk::GetImpl(contextVk->getState().getDrawFramebuffer());
const bool hasFramebufferFetch = mExecutable->usesFramebufferFetch();
const bool isMultisampled = drawFrameBuffer->getSamples() > 1;
transformOptions.multiSampleFramebufferFetch = hasFramebufferFetch && isMultisampled;
transformOptions.enableSampleShading =
contextVk->getState().isSampleShadingEnabled() && isMultisampled;
return transformOptions;
}
angle::Result ProgramExecutableVk::initGraphicsShaderPrograms(
vk::Context *context,
ProgramTransformOptions transformOptions)
{
ASSERT(mExecutable->hasLinkedShaderStage(gl::ShaderType::Vertex));
const uint8_t programIndex = transformOptions.permutationIndex;
ProgramInfo &programInfo = mGraphicsProgramInfos[programIndex];
const gl::ShaderBitSet linkedShaderStages = mExecutable->getLinkedShaderStages();
gl::ShaderType lastPreFragmentStage = gl::GetLastPreFragmentStage(linkedShaderStages);
const bool isTransformFeedbackProgram =
!mExecutable->getLinkedTransformFeedbackVaryings().empty();
for (gl::ShaderType shaderType : linkedShaderStages)
{
ANGLE_TRY(initGraphicsShaderProgram(context, shaderType, shaderType == lastPreFragmentStage,
isTransformFeedbackProgram, transformOptions,
&programInfo, mVariableInfoMap));
}
return angle::Result::Continue;
}
angle::Result ProgramExecutableVk::initProgramThenCreateGraphicsPipeline(
vk::Context *context,
ProgramTransformOptions transformOptions,
vk::GraphicsPipelineSubset pipelineSubset,
vk::PipelineCacheAccess *pipelineCache,
PipelineSource source,
const vk::GraphicsPipelineDesc &desc,
const vk::RenderPass &compatibleRenderPass,
const vk::GraphicsPipelineDesc **descPtrOut,
vk::PipelineHelper **pipelineOut)
{
ANGLE_TRY(initGraphicsShaderPrograms(context, transformOptions));
return createGraphicsPipelineImpl(context, transformOptions, pipelineSubset, pipelineCache,
source, desc, compatibleRenderPass, descPtrOut, pipelineOut);
}
angle::Result ProgramExecutableVk::createGraphicsPipelineImpl(
vk::Context *context,
ProgramTransformOptions transformOptions,
vk::GraphicsPipelineSubset pipelineSubset,
vk::PipelineCacheAccess *pipelineCache,
PipelineSource source,
const vk::GraphicsPipelineDesc &desc,
const vk::RenderPass &compatibleRenderPass,
const vk::GraphicsPipelineDesc **descPtrOut,
vk::PipelineHelper **pipelineOut)
{
// This method assumes that all the state necessary to create a graphics pipeline has already
// been setup by the caller. Assert that all required state is valid so all that is left will
// be the call to `vkCreateGraphicsPipelines`
// Make sure program index is within range
const uint8_t programIndex = transformOptions.permutationIndex;
ASSERT(programIndex >= 0 && programIndex < ProgramTransformOptions::kPermutationCount);
// Make sure the shader modules for all linked shader stages are valid.
ProgramInfo &programInfo = mGraphicsProgramInfos[programIndex];
for (gl::ShaderType shaderType : mExecutable->getLinkedShaderStages())
{
ASSERT(programInfo.valid(shaderType));
}
// Generate spec consts, a change in which results in a new pipeline.
vk::SpecializationConstants specConsts = MakeSpecConsts(transformOptions, desc);
// Choose appropriate pipeline cache based on pipeline subset
if (pipelineSubset == vk::GraphicsPipelineSubset::Complete)
{
CompleteGraphicsPipelineCache &pipelines = mCompleteGraphicsPipelines[programIndex];
return programInfo.getShaderProgram().createGraphicsPipeline(
context, &pipelines, pipelineCache, compatibleRenderPass, getPipelineLayout(), source,
desc, specConsts, descPtrOut, pipelineOut);
}
else
{
// Vertex input and fragment output subsets are independent of shaders, and are not created
// through the program executable.
ASSERT(pipelineSubset == vk::GraphicsPipelineSubset::Shaders);
ShadersGraphicsPipelineCache &pipelines = mShadersGraphicsPipelines[programIndex];
return programInfo.getShaderProgram().createGraphicsPipeline(
context, &pipelines, pipelineCache, compatibleRenderPass, getPipelineLayout(), source,
desc, specConsts, descPtrOut, pipelineOut);
}
}
angle::Result ProgramExecutableVk::getGraphicsPipeline(ContextVk *contextVk,
vk::GraphicsPipelineSubset pipelineSubset,
const vk::GraphicsPipelineDesc &desc,
const vk::GraphicsPipelineDesc **descPtrOut,
vk::PipelineHelper **pipelineOut)
{
ProgramTransformOptions transformOptions = getTransformOptions(contextVk, desc);
ANGLE_TRY(initGraphicsShaderPrograms(contextVk, transformOptions));
const uint8_t programIndex = transformOptions.permutationIndex;
*descPtrOut = nullptr;
*pipelineOut = nullptr;
if (pipelineSubset == vk::GraphicsPipelineSubset::Complete)
{
mCompleteGraphicsPipelines[programIndex].getPipeline(desc, descPtrOut, pipelineOut);
}
else
{
// Vertex input and fragment output subsets are independent of shaders, and are not created
// through the program executable.
ASSERT(pipelineSubset == vk::GraphicsPipelineSubset::Shaders);
mShadersGraphicsPipelines[programIndex].getPipeline(desc, descPtrOut, pipelineOut);
}
return angle::Result::Continue;
}
angle::Result ProgramExecutableVk::createGraphicsPipeline(
ContextVk *contextVk,
vk::GraphicsPipelineSubset pipelineSubset,
vk::PipelineCacheAccess *pipelineCache,
PipelineSource source,
const vk::GraphicsPipelineDesc &desc,
const vk::GraphicsPipelineDesc **descPtrOut,
vk::PipelineHelper **pipelineOut)
{
ProgramTransformOptions transformOptions = getTransformOptions(contextVk, desc);
// When creating monolithic pipelines, the renderer's pipeline cache is used as passed in.
// When creating the shaders subset of pipelines, the program's own pipeline cache is used.
vk::PipelineCacheAccess perProgramPipelineCache;
const bool useProgramPipelineCache = pipelineSubset == vk::GraphicsPipelineSubset::Shaders;
if (useProgramPipelineCache)
{
ANGLE_TRY(ensurePipelineCacheInitialized(contextVk));
perProgramPipelineCache.init(&mPipelineCache, nullptr);
pipelineCache = &perProgramPipelineCache;
}
// Pull in a compatible RenderPass.
const vk::RenderPass *compatibleRenderPass = nullptr;
ANGLE_TRY(contextVk->getCompatibleRenderPass(desc.getRenderPassDesc(), &compatibleRenderPass));
ANGLE_TRY(initProgramThenCreateGraphicsPipeline(
contextVk, transformOptions, pipelineSubset, pipelineCache, source, desc,
*compatibleRenderPass, descPtrOut, pipelineOut));
if (useProgramPipelineCache &&
contextVk->getFeatures().mergeProgramPipelineCachesToGlobalCache.enabled)
{
ANGLE_TRY(contextVk->getRenderer()->mergeIntoPipelineCache(contextVk, mPipelineCache));
}
return angle::Result::Continue;
}
angle::Result ProgramExecutableVk::linkGraphicsPipelineLibraries(
ContextVk *contextVk,
vk::PipelineCacheAccess *pipelineCache,
const vk::GraphicsPipelineDesc &desc,
vk::PipelineHelper *vertexInputPipeline,
vk::PipelineHelper *shadersPipeline,
vk::PipelineHelper *fragmentOutputPipeline,
const vk::GraphicsPipelineDesc **descPtrOut,
vk::PipelineHelper **pipelineOut)
{
ProgramTransformOptions transformOptions = getTransformOptions(contextVk, desc);
const uint8_t programIndex = transformOptions.permutationIndex;
ANGLE_TRY(mCompleteGraphicsPipelines[programIndex].linkLibraries(
contextVk, pipelineCache, desc, getPipelineLayout(), vertexInputPipeline, shadersPipeline,
fragmentOutputPipeline, descPtrOut, pipelineOut));
// If monolithic pipelines are preferred over libraries, create a task so that it can be created
// asynchronously.
if (contextVk->getFeatures().preferMonolithicPipelinesOverLibraries.enabled)
{
vk::SpecializationConstants specConsts = MakeSpecConsts(transformOptions, desc);
mGraphicsProgramInfos[programIndex].getShaderProgram().createMonolithicPipelineCreationTask(
contextVk, pipelineCache, desc, getPipelineLayout(), specConsts, *pipelineOut);
}
return angle::Result::Continue;
}
angle::Result ProgramExecutableVk::getOrCreateComputePipeline(
vk::Context *context,
vk::PipelineCacheAccess *pipelineCache,
PipelineSource source,
vk::PipelineRobustness pipelineRobustness,
vk::PipelineProtectedAccess pipelineProtectedAccess,
vk::PipelineHelper **pipelineOut)
{
ASSERT(mExecutable->hasLinkedShaderStage(gl::ShaderType::Compute));
vk::ComputePipelineOptions pipelineOptions =
vk::GetComputePipelineOptions(pipelineRobustness, pipelineProtectedAccess);
ANGLE_TRY(initComputeProgram(context, &mComputeProgramInfo, mVariableInfoMap, pipelineOptions));
return mComputeProgramInfo.getShaderProgram().getOrCreateComputePipeline(
context, &mComputePipelines, pipelineCache, getPipelineLayout(), pipelineOptions, source,
pipelineOut, nullptr, nullptr);
}
angle::Result ProgramExecutableVk::createPipelineLayout(
vk::Context *context,
PipelineLayoutCache *pipelineLayoutCache,
DescriptorSetLayoutCache *descriptorSetLayoutCache,
gl::ActiveTextureArray<TextureVk *> *activeTextures)
{
const gl::ShaderBitSet &linkedShaderStages = mExecutable->getLinkedShaderStages();
// Store a reference to the pipeline and descriptor set layouts. This will create them if they
// don't already exist in the cache.
// Default uniforms and transform feedback:
mDefaultUniformAndXfbSetDesc = {};
uint32_t numDefaultUniformDescriptors = 0;
for (gl::ShaderType shaderType : linkedShaderStages)
{
const ShaderInterfaceVariableInfo &info =
mVariableInfoMap.getDefaultUniformInfo(shaderType);
// Note that currently the default uniform block is added unconditionally.
ASSERT(info.activeStages[shaderType]);
mDefaultUniformAndXfbSetDesc.addBinding(info.binding,
VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER_DYNAMIC, 1,
gl_vk::kShaderStageMap[shaderType], nullptr);
numDefaultUniformDescriptors++;
}
gl::ShaderType linkedTransformFeedbackStage = mExecutable->getLinkedTransformFeedbackStage();
bool hasXfbVaryings = linkedTransformFeedbackStage != gl::ShaderType::InvalidEnum &&
!mExecutable->getLinkedTransformFeedbackVaryings().empty();
if (context->getFeatures().emulateTransformFeedback.enabled && hasXfbVaryings)
{
size_t xfbBufferCount = mExecutable->getTransformFeedbackBufferCount();
for (uint32_t bufferIndex = 0; bufferIndex < xfbBufferCount; ++bufferIndex)
{
const uint32_t binding = mVariableInfoMap.getEmulatedXfbBufferBinding(bufferIndex);
ASSERT(binding != std::numeric_limits<uint32_t>::max());
mDefaultUniformAndXfbSetDesc.addBinding(binding, VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, 1,
VK_SHADER_STAGE_VERTEX_BIT, nullptr);
}
}
ANGLE_TRY(descriptorSetLayoutCache->getDescriptorSetLayout(
context, mDefaultUniformAndXfbSetDesc,
&mDescriptorSetLayouts[DescriptorSetIndex::UniformsAndXfb]));
// Uniform and storage buffers, atomic counter buffers and images:
mShaderResourceSetDesc = {};
// Count the number of active uniform buffer descriptors.
uint32_t numActiveUniformBufferDescriptors = 0;
const std::vector<gl::InterfaceBlock> &blocks = mExecutable->getUniformBlocks();
for (uint32_t bufferIndex = 0; bufferIndex < blocks.size();)
{
const gl::InterfaceBlock &block = blocks[bufferIndex];
const uint32_t arraySize = GetInterfaceBlockArraySize(blocks, bufferIndex);
bufferIndex += arraySize;
if (block.activeShaders().any())
{
numActiveUniformBufferDescriptors += arraySize;
}
}
// Decide if we should use dynamic or fixed descriptor types.
VkPhysicalDeviceLimits limits = context->getRenderer()->getPhysicalDeviceProperties().limits;
uint32_t totalDynamicUniformBufferCount =
numActiveUniformBufferDescriptors + numDefaultUniformDescriptors;
if (totalDynamicUniformBufferCount <= limits.maxDescriptorSetUniformBuffersDynamic)
{
mUniformBufferDescriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER_DYNAMIC;
}
else
{
mUniformBufferDescriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER;
}
addInterfaceBlockDescriptorSetDesc(mExecutable->getUniformBlocks(), linkedShaderStages,
mUniformBufferDescriptorType, &mShaderResourceSetDesc);
addInterfaceBlockDescriptorSetDesc(mExecutable->getShaderStorageBlocks(), linkedShaderStages,
vk::kStorageBufferDescriptorType, &mShaderResourceSetDesc);
addAtomicCounterBufferDescriptorSetDesc(mExecutable->getAtomicCounterBuffers(),
&mShaderResourceSetDesc);
addImageDescriptorSetDesc(&mShaderResourceSetDesc);
addInputAttachmentDescriptorSetDesc(&mShaderResourceSetDesc);
ANGLE_TRY(descriptorSetLayoutCache->getDescriptorSetLayout(
context, mShaderResourceSetDesc,
&mDescriptorSetLayouts[DescriptorSetIndex::ShaderResource]));
// Textures:
mTextureSetDesc = {};
ANGLE_TRY(addTextureDescriptorSetDesc(context, activeTextures, &mTextureSetDesc));
ANGLE_TRY(descriptorSetLayoutCache->getDescriptorSetLayout(
context, mTextureSetDesc, &mDescriptorSetLayouts[DescriptorSetIndex::Texture]));
// Create pipeline layout with these 3 descriptor sets.
vk::PipelineLayoutDesc pipelineLayoutDesc;
pipelineLayoutDesc.updateDescriptorSetLayout(DescriptorSetIndex::UniformsAndXfb,
mDefaultUniformAndXfbSetDesc);
pipelineLayoutDesc.updateDescriptorSetLayout(DescriptorSetIndex::ShaderResource,
mShaderResourceSetDesc);
pipelineLayoutDesc.updateDescriptorSetLayout(DescriptorSetIndex::Texture, mTextureSetDesc);
// Set up driver uniforms as push constants. The size is set for a graphics pipeline, as there
// are more driver uniforms for a graphics pipeline than there are for a compute pipeline. As
// for the shader stages, both graphics and compute stages are used.
VkShaderStageFlags pushConstantShaderStageFlags =
context->getRenderer()->getSupportedVulkanShaderStageMask();
uint32_t pushConstantSize = GetDriverUniformSize(context, PipelineType::Graphics);
pipelineLayoutDesc.updatePushConstantRange(pushConstantShaderStageFlags, 0, pushConstantSize);
ANGLE_TRY(pipelineLayoutCache->getPipelineLayout(context, pipelineLayoutDesc,
mDescriptorSetLayouts, &mPipelineLayout));
mDynamicUniformDescriptorOffsets.clear();
mDynamicUniformDescriptorOffsets.resize(mExecutable->getLinkedShaderStageCount(), 0);
initializeWriteDescriptorDesc(context);
return angle::Result::Continue;
}
angle::Result ProgramExecutableVk::initializeDescriptorPools(
vk::Context *context,
DescriptorSetLayoutCache *descriptorSetLayoutCache,
vk::DescriptorSetArray<vk::MetaDescriptorPool> *metaDescriptorPools)
{
ANGLE_TRY((*metaDescriptorPools)[DescriptorSetIndex::UniformsAndXfb].bindCachedDescriptorPool(
context, mDefaultUniformAndXfbSetDesc, 1, descriptorSetLayoutCache,
&mDescriptorPools[DescriptorSetIndex::UniformsAndXfb]));
ANGLE_TRY((*metaDescriptorPools)[DescriptorSetIndex::Texture].bindCachedDescriptorPool(
context, mTextureSetDesc, mImmutableSamplersMaxDescriptorCount, descriptorSetLayoutCache,
&mDescriptorPools[DescriptorSetIndex::Texture]));
return (*metaDescriptorPools)[DescriptorSetIndex::ShaderResource].bindCachedDescriptorPool(
context, mShaderResourceSetDesc, 1, descriptorSetLayoutCache,
&mDescriptorPools[DescriptorSetIndex::ShaderResource]);
}
void ProgramExecutableVk::resolvePrecisionMismatch(const gl::ProgramMergedVaryings &mergedVaryings)
{
for (const gl::ProgramVaryingRef &mergedVarying : mergedVaryings)
{
if (!mergedVarying.frontShader || !mergedVarying.backShader)
{
continue;
}
GLenum frontPrecision = mergedVarying.frontShader->precision;
GLenum backPrecision = mergedVarying.backShader->precision;
if (frontPrecision == backPrecision)
{
continue;
}
ASSERT(frontPrecision >= GL_LOW_FLOAT && frontPrecision <= GL_HIGH_INT);
ASSERT(backPrecision >= GL_LOW_FLOAT && backPrecision <= GL_HIGH_INT);
if (frontPrecision > backPrecision)
{
// The output is higher precision than the input
ShaderInterfaceVariableInfo &info = mVariableInfoMap.getMutable(
mergedVarying.frontShaderStage, mergedVarying.frontShader->id);
info.varyingIsOutput = true;
info.useRelaxedPrecision = true;
}
else
{
// The output is lower precision than the input, adjust the input
ASSERT(backPrecision > frontPrecision);
ShaderInterfaceVariableInfo &info = mVariableInfoMap.getMutable(
mergedVarying.backShaderStage, mergedVarying.backShader->id);
info.varyingIsInput = true;
info.useRelaxedPrecision = true;
}
}
}
angle::Result ProgramExecutableVk::getOrAllocateDescriptorSet(
vk::Context *context,
UpdateDescriptorSetsBuilder *updateBuilder,
vk::CommandBufferHelperCommon *commandBufferHelper,
const vk::DescriptorSetDescBuilder &descriptorSetDesc,
const vk::WriteDescriptorDescs &writeDescriptorDescs,
DescriptorSetIndex setIndex,
vk::SharedDescriptorSetCacheKey *newSharedCacheKeyOut)
{
ANGLE_TRY(mDescriptorPools[setIndex].get().getOrAllocateDescriptorSet(
context, commandBufferHelper, descriptorSetDesc.getDesc(),
mDescriptorSetLayouts[setIndex].get(), &mDescriptorPoolBindings[setIndex],
&mDescriptorSets[setIndex], newSharedCacheKeyOut));
ASSERT(mDescriptorSets[setIndex] != VK_NULL_HANDLE);
if (*newSharedCacheKeyOut != nullptr)
{
// Cache miss. A new cache entry has been created.
descriptorSetDesc.updateDescriptorSet(context->getRenderer(), writeDescriptorDescs,
updateBuilder, mDescriptorSets[setIndex]);
}
else
{
commandBufferHelper->retainResource(&mDescriptorPoolBindings[setIndex].get());
}
return angle::Result::Continue;
}
angle::Result ProgramExecutableVk::updateShaderResourcesDescriptorSet(
vk::Context *context,
UpdateDescriptorSetsBuilder *updateBuilder,
const vk::WriteDescriptorDescs &writeDescriptorDescs,
vk::CommandBufferHelperCommon *commandBufferHelper,
const vk::DescriptorSetDescBuilder &shaderResourcesDesc,
vk::SharedDescriptorSetCacheKey *newSharedCacheKeyOut)
{
if (!mDescriptorPools[DescriptorSetIndex::ShaderResource].get().valid())
{
*newSharedCacheKeyOut = nullptr;
return angle::Result::Continue;
}
ANGLE_TRY(getOrAllocateDescriptorSet(context, updateBuilder, commandBufferHelper,
shaderResourcesDesc, writeDescriptorDescs,
DescriptorSetIndex::ShaderResource, newSharedCacheKeyOut));
size_t numOffsets = writeDescriptorDescs.getDynamicDescriptorSetCount();
mDynamicShaderResourceDescriptorOffsets.resize(numOffsets);
if (numOffsets > 0)
{
memcpy(mDynamicShaderResourceDescriptorOffsets.data(),
shaderResourcesDesc.getDynamicOffsets(), numOffsets * sizeof(uint32_t));
}
return angle::Result::Continue;
}
angle::Result ProgramExecutableVk::updateUniformsAndXfbDescriptorSet(
vk::Context *context,
UpdateDescriptorSetsBuilder *updateBuilder,
const vk::WriteDescriptorDescs &writeDescriptorDescs,
vk::CommandBufferHelperCommon *commandBufferHelper,
vk::BufferHelper *defaultUniformBuffer,
vk::DescriptorSetDescBuilder *uniformsAndXfbDesc,
vk::SharedDescriptorSetCacheKey *sharedCacheKeyOut)
{
mCurrentDefaultUniformBufferSerial =
defaultUniformBuffer ? defaultUniformBuffer->getBufferSerial() : vk::kInvalidBufferSerial;
return getOrAllocateDescriptorSet(context, updateBuilder, commandBufferHelper,
*uniformsAndXfbDesc, writeDescriptorDescs,
DescriptorSetIndex::UniformsAndXfb, sharedCacheKeyOut);
}
angle::Result ProgramExecutableVk::updateTexturesDescriptorSet(
vk::Context *context,
const gl::ActiveTextureArray<TextureVk *> &textures,
const gl::SamplerBindingVector &samplers,
PipelineType pipelineType,
UpdateDescriptorSetsBuilder *updateBuilder,
vk::CommandBufferHelperCommon *commandBufferHelper,
const vk::DescriptorSetDesc &texturesDesc)
{
vk::SharedDescriptorSetCacheKey newSharedCacheKey;
ANGLE_TRY(mDescriptorPools[DescriptorSetIndex::Texture].get().getOrAllocateDescriptorSet(
context, commandBufferHelper, texturesDesc,
mDescriptorSetLayouts[DescriptorSetIndex::Texture].get(),
&mDescriptorPoolBindings[DescriptorSetIndex::Texture],
&mDescriptorSets[DescriptorSetIndex::Texture], &newSharedCacheKey));
ASSERT(mDescriptorSets[DescriptorSetIndex::Texture] != VK_NULL_HANDLE);
if (newSharedCacheKey != nullptr)
{
vk::DescriptorSetDescBuilder fullDesc(
mTextureWriteDescriptorDescs.getTotalDescriptorCount());
// Cache miss. A new cache entry has been created.
ANGLE_TRY(fullDesc.updateFullActiveTextures(
context, mVariableInfoMap, mTextureWriteDescriptorDescs, *mExecutable, textures,
samplers, pipelineType, newSharedCacheKey));
fullDesc.updateDescriptorSet(context->getRenderer(), mTextureWriteDescriptorDescs,
updateBuilder, mDescriptorSets[DescriptorSetIndex::Texture]);
}
else
{
commandBufferHelper->retainResource(
&mDescriptorPoolBindings[DescriptorSetIndex::Texture].get());
}
return angle::Result::Continue;
}
template <typename CommandBufferT>
angle::Result ProgramExecutableVk::bindDescriptorSets(
vk::Context *context,
vk::CommandBufferHelperCommon *commandBufferHelper,
CommandBufferT *commandBuffer,
PipelineType pipelineType)
{
// Can probably use better dirty bits here.
// Find the maximum non-null descriptor set. This is used in conjunction with a driver
// workaround to bind empty descriptor sets only for gaps in between 0 and max and avoid
// binding unnecessary empty descriptor sets for the sets beyond max.
DescriptorSetIndex lastNonNullDescriptorSetIndex = DescriptorSetIndex::InvalidEnum;
for (DescriptorSetIndex descriptorSetIndex : angle::AllEnums<DescriptorSetIndex>())
{
if (mDescriptorSets[descriptorSetIndex] != VK_NULL_HANDLE)
{
lastNonNullDescriptorSetIndex = descriptorSetIndex;
}
}
const VkPipelineBindPoint pipelineBindPoint = pipelineType == PipelineType::Compute
? VK_PIPELINE_BIND_POINT_COMPUTE
: VK_PIPELINE_BIND_POINT_GRAPHICS;
for (DescriptorSetIndex descriptorSetIndex : angle::AllEnums<DescriptorSetIndex>())
{
if (ToUnderlying(descriptorSetIndex) > ToUnderlying(lastNonNullDescriptorSetIndex))
{
continue;
}
VkDescriptorSet descSet = mDescriptorSets[descriptorSetIndex];
if (descSet == VK_NULL_HANDLE)
{
continue;
}
// Default uniforms are encompassed in a block per shader stage, and they are assigned
// through dynamic uniform buffers (requiring dynamic offsets). No other descriptor
// requires a dynamic offset.
if (descriptorSetIndex == DescriptorSetIndex::UniformsAndXfb)
{
commandBuffer->bindDescriptorSets(
getPipelineLayout(), pipelineBindPoint, descriptorSetIndex, 1, &descSet,
static_cast<uint32_t>(mDynamicUniformDescriptorOffsets.size()),
mDynamicUniformDescriptorOffsets.data());
}
else if (descriptorSetIndex == DescriptorSetIndex::ShaderResource)
{
commandBuffer->bindDescriptorSets(
getPipelineLayout(), pipelineBindPoint, descriptorSetIndex, 1, &descSet,
static_cast<uint32_t>(mDynamicShaderResourceDescriptorOffsets.size()),
mDynamicShaderResourceDescriptorOffsets.data());
}
else
{
commandBuffer->bindDescriptorSets(getPipelineLayout(), pipelineBindPoint,
descriptorSetIndex, 1, &descSet, 0, nullptr);
}
}
return angle::Result::Continue;
}
template angle::Result ProgramExecutableVk::bindDescriptorSets<vk::priv::SecondaryCommandBuffer>(
vk::Context *context,
vk::CommandBufferHelperCommon *commandBufferHelper,
vk::priv::SecondaryCommandBuffer *commandBuffer,
PipelineType pipelineType);
template angle::Result ProgramExecutableVk::bindDescriptorSets<vk::VulkanSecondaryCommandBuffer>(
vk::Context *context,
vk::CommandBufferHelperCommon *commandBufferHelper,
vk::VulkanSecondaryCommandBuffer *commandBuffer,
PipelineType pipelineType);
void ProgramExecutableVk::setAllDefaultUniformsDirty()
{
mDefaultUniformBlocksDirty.reset();
for (gl::ShaderType shaderType : mExecutable->getLinkedShaderStages())
{
if (!mDefaultUniformBlocks[shaderType]->uniformData.empty())
{
mDefaultUniformBlocksDirty.set(shaderType);
}
}
}
angle::Result ProgramExecutableVk::updateUniforms(
vk::Context *context,
UpdateDescriptorSetsBuilder *updateBuilder,
vk::CommandBufferHelperCommon *commandBufferHelper,
vk::BufferHelper *emptyBuffer,
vk::DynamicBuffer *defaultUniformStorage,
bool isTransformFeedbackActiveUnpaused,
TransformFeedbackVk *transformFeedbackVk)
{
ASSERT(mDefaultUniformBlocksDirty.any());
vk::BufferHelper *defaultUniformBuffer;
bool anyNewBufferAllocated = false;
gl::ShaderMap<VkDeviceSize> offsets = {}; // offset to the beginning of bufferData
uint32_t offsetIndex = 0;
size_t requiredSpace;
// We usually only update uniform data for shader stages that are actually dirty. But when the
// buffer for uniform data have switched, because all shader stages are using the same buffer,
// we then must update uniform data for all shader stages to keep all shader stages' uniform
// data in the same buffer.
requiredSpace = calcUniformUpdateRequiredSpace(context, &offsets);
ASSERT(requiredSpace > 0);
// Allocate space from dynamicBuffer. Always try to allocate from the current buffer first.
// If that failed, we deal with fall out and try again.
if (!defaultUniformStorage->allocateFromCurrentBuffer(requiredSpace, &defaultUniformBuffer))
{
setAllDefaultUniformsDirty();
requiredSpace = calcUniformUpdateRequiredSpace(context, &offsets);
ANGLE_TRY(defaultUniformStorage->allocate(context, requiredSpace, &defaultUniformBuffer,
&anyNewBufferAllocated));
}
ASSERT(defaultUniformBuffer);
uint8_t *bufferData = defaultUniformBuffer->getMappedMemory();
VkDeviceSize bufferOffset = defaultUniformBuffer->getOffset();
for (gl::ShaderType shaderType : mExecutable->getLinkedShaderStages())
{
if (mDefaultUniformBlocksDirty[shaderType])
{
const angle::MemoryBuffer &uniformData = mDefaultUniformBlocks[shaderType]->uniformData;
memcpy(&bufferData[offsets[shaderType]], uniformData.data(), uniformData.size());
mDynamicUniformDescriptorOffsets[offsetIndex] =
static_cast<uint32_t>(bufferOffset + offsets[shaderType]);
mDefaultUniformBlocksDirty.reset(shaderType);
}
++offsetIndex;
}
ANGLE_TRY(defaultUniformBuffer->flush(context->getRenderer()));
// Because the uniform buffers are per context, we can't rely on dynamicBuffer's allocate
// function to tell us if you have got a new buffer or not. Other program's use of the buffer
// might already pushed dynamicBuffer to a new buffer. We record which buffer (represented by
// the unique BufferSerial number) we were using with the current descriptor set and then we
// use that recorded BufferSerial compare to the current uniform buffer to quickly detect if
// there is a buffer switch or not. We need to retrieve from the descriptor set cache or
// allocate a new descriptor set whenever there is uniform buffer switch.
if (mCurrentDefaultUniformBufferSerial != defaultUniformBuffer->getBufferSerial())
{
// We need to reinitialize the descriptor sets if we newly allocated buffers since we can't
// modify the descriptor sets once initialized.
const vk::WriteDescriptorDescs &writeDescriptorDescs =
getDefaultUniformWriteDescriptorDescs(transformFeedbackVk);
vk::DescriptorSetDescBuilder uniformsAndXfbDesc(
writeDescriptorDescs.getTotalDescriptorCount());
uniformsAndXfbDesc.updateUniformsAndXfb(
context, *mExecutable, writeDescriptorDescs, defaultUniformBuffer, *emptyBuffer,
isTransformFeedbackActiveUnpaused,
mExecutable->hasTransformFeedbackOutput() ? transformFeedbackVk : nullptr);
vk::SharedDescriptorSetCacheKey newSharedCacheKey;
ANGLE_TRY(updateUniformsAndXfbDescriptorSet(context, updateBuilder, writeDescriptorDescs,
commandBufferHelper, defaultUniformBuffer,
&uniformsAndXfbDesc, &newSharedCacheKey));
if (newSharedCacheKey)
{
defaultUniformBuffer->getBufferBlock()->onNewDescriptorSet(newSharedCacheKey);
if (mExecutable->hasTransformFeedbackOutput() &&
context->getFeatures().emulateTransformFeedback.enabled)
{
transformFeedbackVk->onNewDescriptorSet(*mExecutable, newSharedCacheKey);
}
}
}
return angle::Result::Continue;
}
size_t ProgramExecutableVk::calcUniformUpdateRequiredSpace(
vk::Context *context,
gl::ShaderMap<VkDeviceSize> *uniformOffsets) const
{
size_t requiredSpace = 0;
for (gl::ShaderType shaderType : mExecutable->getLinkedShaderStages())
{
if (mDefaultUniformBlocksDirty[shaderType])
{
(*uniformOffsets)[shaderType] = requiredSpace;
requiredSpace += getDefaultUniformAlignedSize(context, shaderType);
}
}
return requiredSpace;
}
void ProgramExecutableVk::onProgramBind()
{
// Because all programs share default uniform buffers, when we switch programs, we have to
// re-update all uniform data. We could do more tracking to avoid update if the context's
// current uniform buffer is still the same buffer we last time used and buffer has not been
// recycled. But statistics gathered on gfxbench shows that app always update uniform data on
// program bind anyway, so not really worth it to add more tracking logic here.
//
// Note: if this is changed, PPO uniform checks need to be updated as well
setAllDefaultUniformsDirty();
}
angle::Result ProgramExecutableVk::resizeUniformBlockMemory(
vk::Context *context,
const gl::ShaderMap<size_t> &requiredBufferSize)
{
for (gl::ShaderType shaderType : mExecutable->getLinkedShaderStages())
{
if (requiredBufferSize[shaderType] > 0)
{
if (!mDefaultUniformBlocks[shaderType]->uniformData.resize(
requiredBufferSize[shaderType]))
{
ANGLE_VK_CHECK(context, false, VK_ERROR_OUT_OF_HOST_MEMORY);
}
// Initialize uniform buffer memory to zero by default.
mDefaultUniformBlocks[shaderType]->uniformData.fill(0);
mDefaultUniformBlocksDirty.set(shaderType);
}
}
return angle::Result::Continue;
}
template <typename T>
void ProgramExecutableVk::getUniformImpl(GLint location, T *v, GLenum entryPointType) const
{
const gl::VariableLocation &locationInfo = mExecutable->getUniformLocations()[location];
const gl::LinkedUniform &linkedUniform = mExecutable->getUniforms()[locationInfo.index];
ASSERT(!linkedUniform.isSampler() && !linkedUniform.isImage());
const gl::ShaderType shaderType = linkedUniform.getFirstActiveShaderType();
ASSERT(shaderType != gl::ShaderType::InvalidEnum);
const DefaultUniformBlockVk &uniformBlock = *mDefaultUniformBlocks[shaderType];
const sh::BlockMemberInfo &layoutInfo = uniformBlock.uniformLayout[location];
ASSERT(linkedUniform.getUniformTypeInfo().componentType == entryPointType ||
linkedUniform.getUniformTypeInfo().componentType ==
gl::VariableBoolVectorType(entryPointType));
if (gl::IsMatrixType(linkedUniform.getType()))
{
const uint8_t *ptrToElement = uniformBlock.uniformData.data() + layoutInfo.offset +
(locationInfo.arrayIndex * layoutInfo.arrayStride);
GetMatrixUniform(linkedUniform.getType(), v, reinterpret_cast<const T *>(ptrToElement),
false);
}
else
{
ReadFromDefaultUniformBlock(linkedUniform.getElementComponents(), locationInfo.arrayIndex,
v, layoutInfo, &uniformBlock.uniformData);
}
}
void ProgramExecutableVk::setUniform1fv(GLint location, GLsizei count, const GLfloat *v)
{
SetUniformImpl(mExecutable, location, count, v, GL_FLOAT, &mDefaultUniformBlocks,
&mDefaultUniformBlocksDirty);
}
void ProgramExecutableVk::setUniform2fv(GLint location, GLsizei count, const GLfloat *v)
{
SetUniformImpl(mExecutable, location, count, v, GL_FLOAT_VEC2, &mDefaultUniformBlocks,
&mDefaultUniformBlocksDirty);
}
void ProgramExecutableVk::setUniform3fv(GLint location, GLsizei count, const GLfloat *v)
{
SetUniformImpl(mExecutable, location, count, v, GL_FLOAT_VEC3, &mDefaultUniformBlocks,
&mDefaultUniformBlocksDirty);
}
void ProgramExecutableVk::setUniform4fv(GLint location, GLsizei count, const GLfloat *v)
{
SetUniformImpl(mExecutable, location, count, v, GL_FLOAT_VEC4, &mDefaultUniformBlocks,
&mDefaultUniformBlocksDirty);
}
void ProgramExecutableVk::setUniform1iv(GLint location, GLsizei count, const GLint *v)
{
const gl::VariableLocation &locationInfo = mExecutable->getUniformLocations()[location];
const gl::LinkedUniform &linkedUniform = mExecutable->getUniforms()[locationInfo.index];
if (linkedUniform.isSampler())
{
// We could potentially cache some indexing here. For now this is a no-op since the mapping
// is handled entirely in ContextVk.
return;
}
SetUniformImpl(mExecutable, location, count, v, GL_INT, &mDefaultUniformBlocks,
&mDefaultUniformBlocksDirty);
}
void ProgramExecutableVk::setUniform2iv(GLint location, GLsizei count, const GLint *v)
{
SetUniformImpl(mExecutable, location, count, v, GL_INT_VEC2, &mDefaultUniformBlocks,
&mDefaultUniformBlocksDirty);
}
void ProgramExecutableVk::setUniform3iv(GLint location, GLsizei count, const GLint *v)
{
SetUniformImpl(mExecutable, location, count, v, GL_INT_VEC3, &mDefaultUniformBlocks,
&mDefaultUniformBlocksDirty);
}
void ProgramExecutableVk::setUniform4iv(GLint location, GLsizei count, const GLint *v)
{
SetUniformImpl(mExecutable, location, count, v, GL_INT_VEC4, &mDefaultUniformBlocks,
&mDefaultUniformBlocksDirty);
}
void ProgramExecutableVk::setUniform1uiv(GLint location, GLsizei count, const GLuint *v)
{
SetUniformImpl(mExecutable, location, count, v, GL_UNSIGNED_INT, &mDefaultUniformBlocks,
&mDefaultUniformBlocksDirty);
}
void ProgramExecutableVk::setUniform2uiv(GLint location, GLsizei count, const GLuint *v)
{
SetUniformImpl(mExecutable, location, count, v, GL_UNSIGNED_INT_VEC2, &mDefaultUniformBlocks,
&mDefaultUniformBlocksDirty);
}
void ProgramExecutableVk::setUniform3uiv(GLint location, GLsizei count, const GLuint *v)
{
SetUniformImpl(mExecutable, location, count, v, GL_UNSIGNED_INT_VEC3, &mDefaultUniformBlocks,
&mDefaultUniformBlocksDirty);
}
void ProgramExecutableVk::setUniform4uiv(GLint location, GLsizei count, const GLuint *v)
{
SetUniformImpl(mExecutable, location, count, v, GL_UNSIGNED_INT_VEC4, &mDefaultUniformBlocks,
&mDefaultUniformBlocksDirty);
}
void ProgramExecutableVk::setUniformMatrix2fv(GLint location,
GLsizei count,
GLboolean transpose,
const GLfloat *value)
{
SetUniformMatrixfv<2, 2>(mExecutable, location, count, transpose, value, &mDefaultUniformBlocks,
&mDefaultUniformBlocksDirty);
}
void ProgramExecutableVk::setUniformMatrix3fv(GLint location,
GLsizei count,
GLboolean transpose,
const GLfloat *value)
{
SetUniformMatrixfv<3, 3>(mExecutable, location, count, transpose, value, &mDefaultUniformBlocks,
&mDefaultUniformBlocksDirty);
}
void ProgramExecutableVk::setUniformMatrix4fv(GLint location,
GLsizei count,
GLboolean transpose,
const GLfloat *value)
{
SetUniformMatrixfv<4, 4>(mExecutable, location, count, transpose, value, &mDefaultUniformBlocks,
&mDefaultUniformBlocksDirty);
}
void ProgramExecutableVk::setUniformMatrix2x3fv(GLint location,
GLsizei count,
GLboolean transpose,
const GLfloat *value)
{
SetUniformMatrixfv<2, 3>(mExecutable, location, count, transpose, value, &mDefaultUniformBlocks,
&mDefaultUniformBlocksDirty);
}
void ProgramExecutableVk::setUniformMatrix3x2fv(GLint location,
GLsizei count,
GLboolean transpose,
const GLfloat *value)
{
SetUniformMatrixfv<3, 2>(mExecutable, location, count, transpose, value, &mDefaultUniformBlocks,
&mDefaultUniformBlocksDirty);
}
void ProgramExecutableVk::setUniformMatrix2x4fv(GLint location,
GLsizei count,
GLboolean transpose,
const GLfloat *value)
{
SetUniformMatrixfv<2, 4>(mExecutable, location, count, transpose, value, &mDefaultUniformBlocks,
&mDefaultUniformBlocksDirty);
}
void ProgramExecutableVk::setUniformMatrix4x2fv(GLint location,
GLsizei count,
GLboolean transpose,
const GLfloat *value)
{
SetUniformMatrixfv<4, 2>(mExecutable, location, count, transpose, value, &mDefaultUniformBlocks,
&mDefaultUniformBlocksDirty);
}
void ProgramExecutableVk::setUniformMatrix3x4fv(GLint location,
GLsizei count,
GLboolean transpose,
const GLfloat *value)
{
SetUniformMatrixfv<3, 4>(mExecutable, location, count, transpose, value, &mDefaultUniformBlocks,
&mDefaultUniformBlocksDirty);
}
void ProgramExecutableVk::setUniformMatrix4x3fv(GLint location,
GLsizei count,
GLboolean transpose,
const GLfloat *value)
{
SetUniformMatrixfv<4, 3>(mExecutable, location, count, transpose, value, &mDefaultUniformBlocks,
&mDefaultUniformBlocksDirty);
}
void ProgramExecutableVk::getUniformfv(const gl::Context *context,
GLint location,
GLfloat *params) const
{
getUniformImpl(location, params, GL_FLOAT);
}
void ProgramExecutableVk::getUniformiv(const gl::Context *context,
GLint location,
GLint *params) const
{
getUniformImpl(location, params, GL_INT);
}
void ProgramExecutableVk::getUniformuiv(const gl::Context *context,
GLint location,
GLuint *params) const
{
getUniformImpl(location, params, GL_UNSIGNED_INT);
}
} // namespace rx