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
|
\documentclass[11pt]{article}
%\usepackage{psfig}
\usepackage{epsfig}
\newcommand{\Wskip}[1]{ }
\newcommand{\Wceil}[1]{\lceil #1 \rceil}
\newcommand{\Wfloor}[1]{\lfloor #1 \rfloor}
\newenvironment{routdef}[1]
{
\begin{list}{BLANK}
{
\setlength{\parsep}{0in}
\setlength{\itemsep}{.01in}
\setlength{\partopsep}{0in}
\setlength{\topsep}{0.1in}
\setlength{\labelsep}{0in}
\setlength{\labelwidth}{#1in}
\setlength{\leftmargin}{#1in}
}
} {\end{list}}
\newcommand{\rditem}[2]{\item[#1\hfill(~]#2 )}
\textwidth=6in
\textheight=8.7in
\hoffset = -.6in
\voffset = -.6in
\usepackage{fancyhdr}
\pagestyle{fancy}
\lfoot{}
\cfoot{}
\rfoot{}
\begin{document}
\headheight=25.3pt
\begin{titlepage}
\title{ATLAS Installation Guide
\footnote{This work was supported in part by National Science
Foundation CRI grant SNS-0551504}}
\vspace{.4in}
\author
{
R. Clint Whaley \thanks { {\tt rwhaley@users.sourceforge.net},
{\tt www.cs.utsa.edu/$\sim$whaley}}
}
\end{titlepage}
\maketitle
\fancypagestyle{plain}{}
%\lhead{\footnotesize \bf ATLAS Installation Guide}}
%\chead{\footnotesize \bf ATLAS Installation Guide}
%\rhead{\footnotesize CONTENTS}
%\fancypagestyle{plain}{\lhead{\bf University of Texas at San Antonio,
% Department of Computer Science, Technical Report CS-TR-2008-002}\chead{}\rhead{}}
\begin{abstract}
This note provides a brief overview of ATLAS, and describes how to install
it. It includes extensive discussion of common configure options, and
describes why they might be employed on various platforms. In addition
to discussing how to configure and build the ATLAS package,
this note also describes how an installer can confirm that the resulting
libraries are producing correct answers and running efficiently.
Extensive examples are provided, including a full-length example showing
the installation of both ATLAS and LAPACK on an example architecture.
\end{abstract}
\newpage
\headheight=12.5pt
%\lhead{\footnotesize \bf UTSA/CS Technical Report}
%\chead{\footnotesize \bf CS-TR-2008-002}
\rhead{\footnotesize CONTENTS}
\begin{small}
\tableofcontents
\end{small}
%\listoftables
%\listoffigures
\newpage
\setcounter{page}{1}
%\lhead{\footnotesize \bf UTSA/CS Technical Report}
%\chead{\footnotesize \bf CS-TR-2008-002}
\rhead{{\footnotesize\bf Whaley}~~~~~~\thepage}
\section{Introduction}
This note provides a quick reference to installing and using
ATLAS~\cite{atlas-hp,atlas_wn97,atlas_sc98,atlas_siam,WN147,whaley04}. ATLAS
(Automatically Tuned Linear Algebra Software), is an empirical tuning
system that produces a BLAS~\cite{blas3,blas2a,blas2b,blas1a,blas1b}
(Basic Linear Algebra Subprograms) library which has been specifically optimized
for the platform you install ATLAS on. The BLAS are a set of building block
routines which, when tuned well, allow more complicated Linear Algebra
operations such as solving linear equations or finding eigenvalues to run
extremely efficiently (this is important, since these operations are
computationally intensive). For a list of the BLAS routines, see the FORTRAN77
and C API quick references guides available in the ATLAS tarfile at:
\begin{verbatim}
ATLAS/doc/cblasqref.pdf
ATLAS/doc/f77blasqref.pdf
\end{verbatim}
ATLAS also natively provides a few routines from the LAPACK~\cite{lug}
(Linear Algebra PACKage). LAPACK is an extremely comprehensive
FORTRAN77 package for solving the most commonly occurring problems in numerical
linear algebra. LAPACK is available as an open source FORTRAN77 package from
netlib~\cite{lapack-hp}, and its size and complexity effectively rule out the
idea of ATLAS providing a full implementation. Therefore, we add support
for particular LAPACK routines only when we believe that the potential
performance win we can offer make the extra development and maintenance
costs worthwhile. Presently, ATLAS provides roughly 40 routines, all of
which derive from our improved LU and Cholesky factorizations, which use
recursive blocking. The standard LAPACK routines use statically blocked
routines, which typically run slower than recursively blocked for all
problem sizes. ATLAS's LU and Cholesky factorizations are based on
the work of~\cite{Toledo-LU,RecFred,gustavson98a,WN146,RecQR}.
In addition to providing the standard FORTRAN77 interface to LAPACK,
ATLAS also provides its own C interface, modeled after the official C interface
to the BLAS~\cite{Blast,blast-toms}, which includes support for row-major
storage in addition to the standard column-major implementations. Note that
there is no official C interface to LAPACK, and so there is no general C
API that allows users to easily substitute one C-interface LAPACK for
another, as there is when one uses the standard FORTRAN77 API.
For a list of the LAPACK routines that ATLAS natively supplies, see the
FORTRAN77 and C API quick references guide available in the ATLAS tarfile at:
\begin{verbatim}
ATLAS/doc/lapackqref.pdf
\end{verbatim}
Note that although ATLAS provides only a handful of LAPACK routines, it is
designed so that it can easily be combined with netlib LAPACK in order
to provide the complete library. See Section~\ref{sec-lapack} for details.
\section{Overview of an ATLAS Installation}
\label{sec-install}
\subsection{Downloading the software and checking for known errors}
The main ATLAS homepage is at:
\vspace*{-0.1in}
\begin{verbatim}
http://math-atlas.sourceforge.net/
\end{verbatim}
The software link off of this page allows for downloading the tarfile.
The explicit download link is:
\vspace*{-0.1in}
\begin{verbatim}
https://sourceforge.net/project/showfiles.php?group_id=23725
\end{verbatim}
Once you have obtained the tarfile, you untar it in the directory where
you want to keep the ATLAS source directory. The tarfile will create
a subdirectory called {\tt ATLAS}, which you may want to rename to make
less generic. For instance, assuming I have saved the tarfile to
{\tt /home/whaley/dload}, and want to put the source in
{\tt /home/whaley/numerics}, I could create ATLAS's source directory
({\tt SRCdir}) with the following commands:
\begin{verbatim}
cd ~/numerics
bunzip2 -c ~/dload/atlas3.8.0.tar.bz2 | tar xfm -
mv ATLAS ATLAS3.8.0
\end{verbatim}
Before doing anything else, scope the ATLAS errata file for known
errors/problems that you should fix/be aware of before installation:
\vspace*{-0.1in}
\begin{verbatim}
http://math-atlas.sourceforge.net/errata.html
\end{verbatim}
This file contains not only all bugs found, but also all kinds
of platform-specific installation and tuning help.
\subsection{Turn off CPU throttling when installing ATLAS}
Most OSes (including Linux) now turn on CPU throttling for power management\\
\underline{even if you are using a desktop machine}. CPU throttling makes
pretty much all timings completely random, and so any ATLAS install will be
junk. Therefore, before installing ATLAS, turn off CPU throttling. For most
PCs, you can switch it off in the BIOS (eg., on my Athlon-64 machine, I can say
"No" to "Cool and Quiet" under "Power Management"). Most OSes also provide a
way to switch off CPU throttling, but that varies from OS to OS. Under
Fedora, at any rate, the following command seemed to work:
\begin{verbatim}
/usr/bin/cpufreq-selector -g performance
\end{verbatim}
On my Core2Duo, cpufreq-selector only
changes the parameters of the first CPU, regardless of which cpu you specify.
I suspect this is a bug, because on earlier systems, the remaining CPUs were
controlled via a logical link to /sys/devices/system/cpu/cpu0/. In this case,
the only way I found to force the second processor to also run at its peak
frequency was to issue the following as root after setting CPU0 to performance:
\begin{verbatim}
cp /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor \
/sys/devices/system/cpu/cpu1/cpufreq/scaling_governor
\end{verbatim}
Under MacOS or Windows,
you may be able to change this under the power settings.
ATLAS config tries to detect if CPU throttling is enabled, but it may not
always detect it, and sometimes may detect it after you have disabled it.
In the latter case, to force the {\tt configure} to continue regardless of the
results of the CPU throttling probe, pass this flag to {\tt configure}:
\begin{verbatim}
-Si cputhrchk 0
\end{verbatim}
\subsection{Basic Steps of an ATLAS install}
An ATLAS install is performed in 5 steps, only the first two of which
are mandatory.
This install process is very similar to other free software installs,
particularly gnu, though the fact that ATLAS does an extremely complex
empirical tuning step can make the build step particularly long running.
There are two directories that we will refer to constantly in this note,
which indicate both the ATLAS source and build directories:
\begin{description}
\item[{\tt SRCdir}]: This handle should be replaced by the path to your
ATLAS source directory (eg, {\tt /home/whaley/ATLAS3.8.0}).
\item[{\tt BLDdir}]: This handle should be replaced by the path to your
ATLAS build directory (eg, {\tt /home/whaley/ATLAS3.8.0/Linux\_P4E64SSE3}).
\end{description}
Note that these two directories cannot be the same (i.e. you cannot build
the libraries directly in the source directory). The examples in this
note show the {\tt BLDdir} being a subdirectory of the {\tt SRCdir}, but
this is not required (in fact, any directory to which the installer has
read/write permission other than {\tt SRCdir} can be used).
\vspace*{0.1in}
\noindent
The ATLAS install steps are:
\begin{enumerate}
\item {\bf configure} (\S\ref{sec-configure}):
Tell the ATLAS build harness where your {\tt SRCdir}
and {\tt BLDdir} directories are, and allow ATLAS to probe the
platform to create ATLAS's {\tt Make.inc} and {\tt BLDdir} directory
tree.
\item {\bf build} (\S\ref{sec-build}):
Tune ATLAS for your platform, and build the libraries.
\item {\bf check}\footnotemark[1] (\S\ref{sec-check}):
Run sanity tests to ensure your libraries are producing
correct answers.
\item {\bf time}\footnotemark[1] (\S\ref{sec-time}):
Run basic timing on various ATLAS kernels in order to
make sure the tuning done in the {\bf build} step
has resulted in efficient implementations.
\item {\bf install}\footnotemark[1] (\S\ref{sec-install-step}):
Copy ATLAS's libraries from the BLDdir to some
standard location.
\end{enumerate}
\footnotetext[1]{Optional step}
It is extremely important that you read Section~\ref{sec-configure} in
particular, as most users will want to throw at least one flag during
the {\tt configure} step. In particular, most installers will
want to set whether to build 32 or 64-bit libraries
(Section~\ref{sec-bitwidth}), and fine-tune the timer used, as discussed
in Section~\ref{sec-time-flags}.
However, for the impatient, here is the way a
typical install might look (see \S\ref{sec-configure} for an explanation
of the {\tt configure} flags, since they will not work on all systems);
note that the characters after the \verb+#+ character are comments,
and not meant to be typed in:
\begin{verbatim}
bunzip2 -c atlas3.9.x.tar.bz2 | tar xfm - # create SRCdir
mv ATLAS ATLAS3.9.x # get unique dir name
cd ATLAS3.9.x # enter SRCdir
mkdir Linux_C2D64SSE3 # create BLDdir
cd Linux_C2D64SSE3 # enter BLDdir
../configure -b 64 -D c -DPentiumCPS=2400 \ # configure command
--prefix=/home/whaley/lib/atlas \ # install dir
--with-netlib-lapack-tarfile=/home/whaley/dload/lapack.tgz
make build # tune & build lib
make check # sanity check correct answer
make time # check if lib is fast
make install # copy libs to install dir
\end{verbatim}
\section{The ATLAS configure step}
\label{sec-configure}
In this step, ATLAS builds all the subdirectories of the {\tt BLDdir}, and
creates the {\tt make} include file used in all ATLAS's Makefiles
({\tt Make.inc}).
In order to do this successfully, you inform ATLAS where your
{\tt SRCdir} and {\tt BLDdir} are located, and pass flags which tell
{\tt configure}
what type of install you want to do. The basic way to do a configure step is:
\begin{verbatim}
cd BLDdir ; SRCdir/configure [flags]
\end{verbatim}
A complete list of flags is beyond the
scope of this paper, but you can get a list of them by passing {\tt --help} to
{\tt configure}. In this note, we will discuss some of the more important
flags only. ATLAS takes two types of flags: flags that are consumed by the
initial {\tt configure} script itself begin with {\tt --}, and flags that are
passed by {\tt configure} to a later config step begin with only a
single~{\tt -}.
We first discuss flags and steps for building a full netlib library
using netlib's LAPACK~(\S\ref{sec-lapack}),
building a shared library~(\S\ref{sec-shared}),
changing the compilers~(\S\ref{sec-comps}), and
a flag~(\S\ref{sec-nof77}) to indicate that you have no FORTRAN compiler
(and thus don't need any FORTRAN APIs), and changing the way ATLAS
does timings~(\S\ref{sec-time-flags}).
%and manually overriding various
%things configure detects~(\S\ref{sec-OS-override})
%such as OS type and assembly dialect.
Finally, we consider a few
miscellaneous flags~(\S\ref{sec-flag-other}), including the flag telling
ATLAS whether the resulting libraries should assume a 64 or 32 bit
address space~(\S\ref{sec-bitwidth}).
\subsection{Building a full LAPACK library using ATLAS and netlib's LAPACK}
\label{sec-lapack}
ATLAS natively provides only a relative handful of the routines which comprise
LAPACK. However, ATLAS is designed so that its routines can easily be
added to netlib's standard LAPACK in order to get a full LAPACK library.
If you want your final libraries to have all the LAPACK routines,
then you just need to pass the {\tt --with-netlib-lapack-tarfile} flag
to configure, along with the netlib tarfile that you have previously
downloaded. For instance, assuming you have previously downloaded
the lapack tarfile to {\tt /home/whaley/dload/lapack.tgz}, you
would add the following to your {\tt configure} flags:
\begin{verbatim}
--with-netlib-lapack-tarfile=/home/whaley/dload/lapack.tgz
\end{verbatim}
Configure then auto-builds a {\tt make.inc} for LAPACK to use,
and builds netlib LAPACK as part of the ATLAS install process.
Section~\ref{sec-lpdetails} provides details on how this works
in case it breaks down for future releases of LAPACK (it has been
tested to work with netlib LAPACK 3.1.1 and 3.2.1).
\subsubsection{LAPACK APIs}
Note that there is no standard C API to LAPACK. Therefore, when
you build the netlib LAPACK, you get only the Fortran77 API on
all platforms. Various vendor libraries provide various C APIs.
ATLAS provides two types of LAPACK APIs for C.
\paragraph{ATLAS's clapack API:}
ATLAS's original C interface to the LU and Cholesky-related routines
is built from the {\tt ATLAS/interfaces/lapack/C/src/} directory,
and is documented in {\tt ATLAS/doc/cblasqref.pdf}. This API
is like that of the {\tt cblas}, in that all routines take a
new argument that allows matrices to be either row- or column-major.
This API is difficult to extend to all of LAPACK, since the F77
LAPACK provided by netlib only handles column-major. This API
uses the CBLAS enum types for F77's string arguments, and the
appropriate pass-by-value or pass-by-address. This API prefixes
{\tt clapack\_} to the native lapack routine name.
\paragraph{ATLAS's C2F API:}
For routines where we depend on the Fortran LAPACK for implementation,
we have created a new C/LAPACK API, which can be found in \\
{\tt ATLAS/interfaces/lapack/C2F/src}. This API is presently
undocumented. It prefixes {\tt ATL\_C2F} to the native LAPACK
routine name.
This interface is designed to be quick to extend, and it's main
goal is to provide access to the F77 LAPACK without requiring the
caller to know the arcana of calling a Fortran77 routine from C
(which varies by compiler and platform). Therefore, you get a
nice C interface, using enums, with correct pass-by-address/value.
For every routine that takes a work argument, this API provides
two routines: one with the normal name which does \underline{not} take
a work, where the API does the {\tt malloc} for the user, and a second
routine with {\tt \_wrk} suffixed to the name that works exactly
like the F77 routine (user must pass in appropriate workspace).
So, for instance the lapack routine {\tt geqrf} takes work, and
so for double precision this API provides a routine {\tt ATL\_C2Fdgeqrf}
which takes no workspace-related parameters, and a
{\tt ATL\_C2Fdgeqrf\_wrk} which takes workspace parameters.
The C2F API is built by simple wrappers around the F77 LAPACK,
and presently we have wrappers only for the QR-related routines.
It is extremely straightfoward to expand this API to cover all
of LAPACK, but I won't do so unless users ask for given routines,
or my own research requires them (no use producing code that
one using), so let me know if there are routines you'd like to
see ATLAS provide via this API.
\subsubsection{Details on ATLAS's LAPACK autobuild}
\label{sec-lpdetails}
The main thing that you need to install the netlib LAPACK is a correct
{\tt make.inc} which provides the correct compilers, flags, etc. When
mixing ATLAS and netlib lapack, you need to be sure to use compatible
compiler flags. ATLAS's lapack autobuild creates a {\tt make.inc} for
LAPACK to use by substituting the ATLAS options in {\tt make.inc.example}
that netlib LAPACK provides. It auto-sets the macros
{\tt FORTRAN}, {\tt OPTS}, {\tt NOOPT}, {\tt LOADER}, {\tt LOADOPTS},
and {\tt TIMER}. If in future releases of LAPACK they remove the
{\tt make.inc.example} file, rename the above macros, or provide
additional architecture-specific macros, then this autobuild process
may have to be modified. In this case, you may be able to get things
working by creating a compatible {\tt make.inc} yourself after
the ATLAS configure step, but before the build step. You can
see the Makefile targets for autobuilding netlib's ATLAS in
{\tt BLDdir/interfaces/lapack/C2F/Makefile}.
\Wskip{
Note that there is no standard C API to LAPACK, so this will get you
only the FORTRAN API. As previously mentioned, ATLAS provides its own
C/F77 API only for those LAPACK routines it natively provides. Therefore,
C users wishing to call an LAPACK routine not natively provided by ATLAS
will need to build the full LAPACK library as outlined in this section,
and then call the FORTRAN API. Note that we are currently working on an
add-on package that provides a uniform C API for any FORTRAN77 API LAPACK,
but this work may not be available for some time (or indeed, ever).
Here are the steps to get a full FORTRAN77 API LAPACK which uses ATLAS's
improved routines when possible, and the standard netlib routines when not:
\begin{enumerate}
\item Download and install netlib's LAPACK (see~\S\ref{sec-netlib-lp} for
help with this). Assume this is installed in {\tt NLAPACKdir}
(eg., my {\tt NLAPACKdir} might be {\tt /home/whaley/numerical/lapack-3.1.1}).
\item During configure of ATLAS, pass {\tt configure} the flag so that it will
add your previously install netlib LAPACK to its own LAPACK routines
by passing the following flags to your ATLAS configure line:
\vspace*{-0.1in}
\begin{verbatim}
--with-netlib-lapack=NLAPACKdir/<your lapack library name>
\end{verbatim}
(eg.,
{\tt --with-netlib-lapack=/home/whaley/numerical/lapack-3.1.1/lapack\_linux.a}).
\item Do rest of ATLAS install as normal, full FORTRAN77 LAPACK API will
be available in {\tt BLDdir/lib/liblapack.a}.
\end{enumerate}
These directions allow you to produce a full LAPACK when doing an ATLAS
install. Section~\ref{sec-lapack-post} describes how to easily add netlib's
LAPACK to an already existing ATLAS build.
}
\subsubsection{Obtaining netlib's LAPACK}
\label{sec-netlib-lp}
You can download the LAPACK reference implementation from
{\tt www.netlib.org/lapack/}. As of this writing, the newest LAPACK
tarfile was {\tt http://www.netlib.org/lapack/lapack.tgz}.
For more standard information on LAPACK, please scope the following URLs:
\vspace*{-0.1in}
\begin{itemize}
\item \verb|http://www.netlib.org/lapack/|
\vspace*{-0.1in}
\item \verb|http://www.netlib.org/lapack/lawn81/index.html|
\vspace*{-0.1in}
\item \verb|http://www.netlib.org/lapack/lawn41/index.html|
\vspace*{-0.1in}
\item \verb|http://www.netlib.org/lapack/release_notes.html|
\vspace*{-0.1in}
\item \verb|http://www.netlib.org/lapack/lug/index.html|
\end{itemize}
\subsection{Changing the compilers and flags that ATLAS uses for the build}
\label{sec-comps}
ATLAS defines eight different compilers and associated flag macros in
its {\tt Make.inc} which are used to compile various files during the
install process. ATLAS's {\tt configure} provides flags for changing both the
compiler and flags for each of these macros. In the following list,
the macro name is given first, and the configure flag abbreviation is in
parentheses:
\begin{enumerate}
\item {\tt XCC} ({\tt xc}): C compiler used to compile ATLAS's build harness routines
(these never appear in any user-callable library)
\item {\tt GOODGCC} ({\tt gc}): gcc with any required architectural flags
(eg. {\tt -m64}), which will be used to assemble cpp-enabled assembly
and to compile certain multiple implementation routines that specifically
request {\tt gcc}
\item {\tt F77} ({\tt if}): FORTRAN compiler used to compile ATLAS's FORTRAN77
API interface routines.
\item {\tt ICC} ({\tt ic}): C compiler used to compile ATLAS's C API interface
routines.
\item {\tt DMC} ({\tt dm}): C compiler used to compile ATLAS's generated
double precision (real and complex) matmul kernels
\item {\tt SMC} ({\tt sm}): C compiler used to compile ATLAS's generated
single precision (real and complex) matmul kernels
\item {\tt DKC} ({\tt dk}): C compiler used to compile all other double
precision routines (mainly used for other kernels, thus the
{\tt K})
\item {\tt SKC} ({\tt sk}): C compiler used to compile all other single
precision routines (mainly used for other kernels,
thus the {\tt K})
\end{enumerate}
It is almost \underline{never} a good idea to change {\tt DMC} or {\tt SMC},
and it is only very rarely a good idea to change {\tt DKC} or {\tt SKC}.
For ATLAS~3.8.0, all architectural defaults are set using gcc 4.2 only
(the one exception is MIPS/IRIX, where SGI's compiler is used). In most
cases, switching these compilers will get you worse performance and accuracy,
even when you are \underline{\em absolutely} sure it is a better compiler
and flag combination! In particular we tried the Intel compiler {\tt icc}
(called {\tt icl} on Windows) on Intel x86 platforms,
and overall performance was lower than {\tt gcc}. Even worse,
from the documentation
{\tt icc} does not seem to have any firm IEEE floating point compliance unless
you want to run so slow that you could compute it by hand faster. This means
that whenever {\tt icc} achieves reasonable performance, I
have no idea if the error will be bounded or not. I could not obtain
access to {\tt icc} on the Itaniums, where {\tt icc} has historically been
much faster than {\tt gcc}, but I note that the performance of
{\tt gcc4.2} is much better than {\tt gcc3} for most routines, so {\tt gcc}
may be the best compiler there now as well.
There is almost never a need to change {\tt XCC}, since it doesn't affect
the output libraries in any way, and we have seen that changing the kernel
compilers is a bad idea. However, what if you yourself use a non-gnu
compiler, like Intel's {\tt icc} or {\tt ifort}, then what you need to
do is tell ATLAS to compile its interface routines with your compilers,
which is discussed in Section~\ref{sec-cc-iface}. Another common problem
is that your OS has been built with an older {\tt gcc} whose libraries
are incompatible with {\tt gcc 4.2}. In this case, creating an executable
with {\tt gcc4.2} can cause problems, and so what you want to do is keep
{\tt gcc3} as you default compiler (compiling ATLAS interface routines
with it, as well as using it for all linking) but
compile the ATLAS kernel routines with {\tt gcc4}. This
case is discussed in Section~\ref{sec-cc-4v3}.
For those who insist on monkeying with
other compilers, Section~\ref{sec-cc-override} gives some guidance.
Finally installing ATLAS without a FORTRAN compiler
is discussed in Section~\ref{sec-nof77}.
\subsubsection{Changing ATLAS interface compilers to match your usage}
\label{sec-cc-iface}
As mentioned, ATLAS typically gets its best performance when compiled
with {\tt gcc} using the flags that ATLAS automatically picks for
your platform (this assumes you are installing on a system that ATLAS
provides architectural defaults for). However, you can vary the interface
(API) compilers without affecting ATLAS's performance. Since most compilers
are interoperable with {\tt gcc} this is what we recommend you do if you
are using a non-default compiler. Note that almost all compilers can
interoperate with {\tt gcc}, though you may have to throw some special
flags (eg., {\tt /iface:cref} for MSVC++).
The configure flags to override the C interface compiler and flags are:
\vspace*{-0.1in}
\begin{verbatim}
-C ic <C compiler> -F ic '<compiler flags>'
\end{verbatim}
The configure flags to override the FORTRAN interface compiler and flags are:
\vspace*{-0.1in}
\begin{verbatim}
-C if <FORTRAN compiler> -F if '<compiler flags>'
\end{verbatim}
A few example will help here. If I wanted to use Intel's FORTRAN and C
compilers under windows on a P4, I could issue:
\vspace*{-0.1in}
\begin{verbatim}
-C if ifort -F if '-O2 -fltconsistency -nologo' \
-C ic icl -F ic '-QxN -O3 -Qprec -fp:extended -fp:except -nologo -Oy'
\end{verbatim}
On the same system, if I wanted to use Intel for FORTRAN and MSVC++ for C:
\vspace*{-0.1in}
\begin{verbatim}
-C if ifort -F if '-O2 -fltconsistency -nologo' \
-C ic icl -F ic '-Oy -Ox -arch:SSE2 -nologo'
\end{verbatim}
For Windows, we can note a couple of things. First, while these flags are
straight from the Windows compiler documentation, we have replaced the
Windows `{\tt /}' flag character with the Unix `{\tt -}' flag character.
This is because ATLAS doesn't call native Windows compilers directly, but
rather calls a wrapper routine that makes these compilers work with make
like a standard Unix compiler. The second thing to notice is that we don't
have to say to use the {\tt /iface:cref} flag, because this same wrapper
always throws this flag (ATLAS does not work with the other rather bizarre
naming strategies).
For a non-Windows example, assume you use the Sun Workshop compilers available
under Solaris. You can instruct {\tt configure} to use them for building the
APIs rather than the gnu compilers with something like:
\vspace*{-0.1in}
\begin{verbatim}
-C if f77 -F if '-dalign -native -xO5' \
-C ic cc -F ic '-dalign -fsingle -xO5 -native'
\end{verbatim}
\subsubsection{Compiling ATLAS with gcc 4.2 when your OS uses an
incompatible gcc}
\label{sec-cc-4v3}
As previously mentioned, {\tt gcc4.2} is what the architectural defaults
are built for, and previous versions are likely to hurt your performance.
For systems with {\tt gcc4.1} (the worst-performing {\tt gcc} for x86
machines), you can usually just install {\tt gcc4.2}, and change your
path so that {\tt gcc4.2} is your default compiler. However, between
major releases the {\tt gcc} system libraries change too much for this
to work right. Therefore, if your OS was built with {\tt gcc3}, for example,
what will often happen is that executables built with {\tt gcc4} will
not be able to run, unless you fiddle with your {\tt LD\_LIBRARY\_PATH} so
that the {\tt gcc4} libraries are found before those of {\tt gcc3}.
However, if you do this, then often {\tt gcc3}-built objects, which include
the majority of things you use every day (eg., editors), won't run because
they find the {\tt gcc4} libraries instead of the expected libs from {\tt gcc3}!
Therefore, you don't want to make {\tt gcc4.2} your default compiler, but
you want to have ATLAS use it to compile all the kernel routines, while
compiling interface routines and doing any linking with {\tt gcc3}. To
do this, leave the system {\tt gcc} as the default one in your path, but
pass the following flag to {\tt configure}:
\vspace*{-0.1in}
\begin{verbatim}
-Ss kern <path to gcc4.2>
\end{verbatim}
This tells ATLAS to use all non-kernel compilers as normal, but to change
all kernel compilers to the given compiler. Therefore, if I have installed
{\tt gcc4.2} on my {\tt gcc3}-built OS in my own home area at
{\tt /home/whaley/local/gcc42}, I would add something like:
\vspace*{-0.1in}
\begin{verbatim}
-Ss kern /home/whaley/local/gcc42/bin/gcc
\end{verbatim}
\subsubsection{Rough guide to overriding ATLAS's compiler choice/changing flags}
\label{sec-cc-override}
Previous sections have discussed the more useful cases of overriding
ATLAS's compiler and flags, which typically leave ATLAS's kernel compilers
alone. Users often wish to add flags or change arbitrary compilers, however.
This is rarely a good idea, and almost always provides reduced performance.
However, you can do it. You can find more details by passing
{\tt --help} to {\tt configure}.
As previously mentioned (\S\ref{sec-cc-iface}), you can specify what compiler
(flag setting) to
override by passing the appropriate abbreviation to the {\tt -C} ({\tt -F})
configure flags in order to change the compiler (compiler flags). For
example, you would pass {\tt -C if} to override interface FORTRAN compiler.
{\tt configure} also supports appending certain compiler flags, so that user
flags are simply added to the defaults that ATLAS uses. This is done:
\vspace*{-0.1in}
\begin{verbatim}
-Fa <abbr> '<comp flags to append>'
\end{verbatim}
where \verb|<abbr>| is one of:
\begin{itemize}
\item One of the already discussed compiler abbreviations
(eg, {\tt xc}, {\tt ic}, {\tt if}, {\tt sk}, {\tt dc}, {\tt sm}
or {\tt dm})
\item {\tt al}: all compilers (including FORTRAN) except {\tt GOODGCC}
\item {\tt alg} all compilers (including FORTRAN) including {\tt GOODGCC}
\item {\tt ac}: all C compilers except {\tt GOODGCC}
\item {\tt acg}: all C compilers including {\tt GOODGCC}
\end{itemize}
Therefore, by passing the following to {\tt configure}:
\vspace*{-0.1in}
\begin{verbatim}
-Fa acg '-DUsingDynamic -fPIC'
\end{verbatim}
We would have all C routines compiled with {\tt -fPIC}, and also have the
macro {\tt UsingDynamic} defined (ATLAS does not use this macro, this is
for example only).
The compiler overriding flag {-C} can also take the abbreviation {\tt ac}
which will override all C compilers except {\tt GOODGCC} with the given
C compiler. There is currently no flag to override {\tt GOODGCC} on
the command line, so if you need to do this, you will need to edit the
output {\tt Make.inc} after configure.
As an example, if I want to use SunOS's {\tt f77} rather than {\tt gfortran},
I could pass the following compiler and flag override:
\vspace*{-0.1in}
\begin{verbatim}
-C if f77 -F if 'dalign -native -xO5'
\end{verbatim}
{\bf IMPORTANT NOTE}: If you change the default flags in any way for
the kernel compilers
(even just appending flags), you may reduce performance. Therefore once
your build is finished, you should make sure to compare your achieved
performance against what ATLAS's architectural defaults achieved. See
Section~\ref{sec-time-nondefault} for details on how to do this.
If your compiler is a different version of {\tt gcc}, you may also
want to tell ATLAS not to use the architectural defaults, as described
in Section~\ref{sec-nodefaults}.
\subsubsection{Installing ATLAS when you don't have access to a FORTRAN
compiler}
\label{sec-nof77}
By default, ATLAS expects to find a FORTRAN compiler on your system. If
you cannot install a FORTRAN compiler, you can still install ATLAS, but
ATLAS will be unable to build the FORTRAN77 APIs for both BLAS and LAPACK.
Further, certain tests will not be able to even compile, as their testers
are at least partially implemented in FORTRAN. To tell ATLAS you wish
to install w/o a FORTRAN compiler, simply add the flag:
\vspace*{-0.1in}
\begin{verbatim}
--nof77
\end{verbatim}
to your {\tt configure} command.
{\tt IMPORTANT NOTE}: When you install ATLAS w/o a FORTRAN compiler,
your build step will end with a bunch of {\tt make} errors about being
unable to compile some FORTRAN routines. This is because the {\tt Makefile}s
always attempt to compile the FORTRAN APIs: they simply continue the install
if they don't succeed in building them. So, just because you get a lot
of {\tt make} messages about FORTRAN, don't assume your library is messed
up. As long as {\tt make check} and {\tt make time} say your {\tt -nof77}
install is OK, you should be fine.
\subsection{Building dynamic/shared libraries}
\label{sec-shared}
ATLAS natively builds static libraries (i.e. libs that usually end in
`\texttt{.a}' under Unix and `\texttt{.lib}' under windows).
ATLAS always builds such a library,
but it can also optionally be requested to build a dynamic/shared library
(typically ending in .so for Unix or .dll windows) as well. In order to do so,
you must tell ATLAS up front to compile with the proper flags (the same is
true when building netlib's LAPACK, see \S\ref{sec-lapack} for more details).
Assuming
you are using the gnu C and FORTRAN compilers, you can add the following
commands to your {\tt configure} command:
\vspace*{-0.1in}
\begin{verbatim}
-Fa alg -fPIC
\end{verbatim}
to force ATLAS to be built using position independent code (required for a
dynamic lib). If you use non-gnu compilers, you'll need to use -Fa to
pass the correct flag(s) to append to force position independent code for
each compiler (don't forget the gcc compiler used in the index files).
After your build is complete, you can {\tt cd} to your {\tt OBJdir/lib}
directory, and
ask ATLAS to build the {\tt .so} you want. If you want all libraries, including
the FORTRAN77 routines, the target choices are:
\begin{description}
\item [shared] : create shared versions of ATLAS's sequential libs
\item [ptshared] : create shared versions of ATLAS's threaded libs
\end{description}
If you want only C routines (eg., you don't have a FORTRAN compiler):
\begin{description}
\item [cshared] : create shared versions of ATLAS's sequential libs
\item [cptshared] : create shared versions of ATLAS's threaded libs
\end{description}
Note that this support for building dynamic libraries is new in this release,
and not well debugged or supported, and is much less likely to work for
non-gnu compilers.
\textbf{IMPORTANT NOTE}: Since gcc uses one less integer register when
compiling with this flag, this
could potentially impact performance of the architectural defaults,
but we have not seen it so far. Therefore, do not throw this flag
unless you want dynamic libraries. If you want both static and dynamic
libs, the safest thing is probably to build ATLAS twice, once static
and once dynamic, rather than getting both from a dynamic install.
\subsection{Changing the way ATLAS does timings}
\label{sec-time-flags}
By default ATLAS does all timings with a CPU timer, so that the install
can be done on a machine that is experiencing relatively heavy load. However,
CPU time has very poor resolution, and so this makes the timings less repeatable
and provides for only a rough idea of overall performance. Therefore, if
you are installing ATLAS on a machine which is not heavily loaded, you will
want to improve your install by instructing ATLAS to use one of its higher
resolution wall timers.
For x86 machines, ATLAS has access to a cycle accurate wall timer, assuming
you are using {\tt gcc} as your interface compiler (we use {\tt gcc}'s inline
assembly to enable this timer -- under Linux, Intel's {\tt icc} also supports
this form of inline
assembly). ATLAS needs to be able to translate the cycle count returned by
this function into seconds, so you must pass your machine's clock rate to
ATLAS. In order to do this, you add the following flags to your
configure flags:
\vspace*{-0.1in}
\begin{verbatim}
-D c -DPentiumCPS=<your Mhz>
\end{verbatim}
So, for my 2.4Ghz Core2Duo, I would pass:
\vspace*{-0.1in}
\begin{verbatim}
-D c -DPentiumCPS=2400
\end{verbatim}
If you are not on an x86 machine, or if your interface compiler is not {\tt gcc}
(or {\tt icc} if on Linux), then you cannot use the above cycle-accurate
wall timer. However, wall time is still much more accurate than CPU time,
so you can indicate ATLAS should use its wall timer for the install by passing
the flag:
\vspace*{-0.1in}
\begin{verbatim}
-D c -DWALL
\end{verbatim}
Note that on Windows XP/NT/2000, this should still get you a cycle-accurate
walltime, since it calls some undocumented Windows APIs that purport to do so.
For Solaris, the high resolution timer {\tt gethrtime} will be used. For
all other OSes, this will call a standard wall timer such as {\tt gettimeofday},
which is still usually much more accurate than the CPU timer.
\Wskip{
\subsection{Manually setting OS, architecture, assembly dialect}
\label{sec-OS-override}
Configure's selection of operating system, architecture, assembly dialect
and SIMD vectorization type are all controlled by enumerated types.
Occasionally, configure will misdetect one of these values and so
configure provides
flags for overriding configures detecting of these features.
\subsubsection{Manually overriding the assembly dialect}
\subsubsection{Manually overriding SIMD vectorization type}
\label{sec-SIMD-override}
Also, the installer
may occasionally want to refrain from using a feature that this machine
possesses, but other machines that plan to use the same library don't.
For instance, the installation machine might have SSE3, but the installer
wants to use the same library on machines that possess only SSE2, and so
the installer wants to use only SSE2, even though this will make the library
sub-optimal on the machine possessing SSE3.
}
\subsection{Various other flags}
\label{sec-flag-other}
\subsubsection{Changing pointer bitwidth (64 or 32 bits)}
\label{sec-bitwidth}
Most modern platforms allow for compiling libraries to handle either 32 or
64 bit address spaces. On the x86, this selection strongly affects the ISA
used (eg., whether to use IA32 or x86-64). The x86-64 ISA, with 16 rather
than 8 registers, is more
amenable to optimization than the IA32, so if the user has no preference,
64-bit pointers are recommended.
If ATLAS's guess is not correct,
you can tell configure what address space to build for. In order to force
32-bit pointer width, pass the flag:
\vspace*{-0.1in}
\begin{verbatim}
-b 32
\end{verbatim}
and in order to force 64 bit pointers, pass:
\vspace*{-0.1in}
\begin{verbatim}
-b 64
\end{verbatim}
(the {\tt b} stands for bitwidth).
This tells ATLAS to throw the appropriate compiler flags for compilers it
knows about, as well as effecting various configure probes. Therefore, if
you override ATLAS's compiler choices, be sure that you give the correct
flags to match this setting.
\subsubsection{Changing configure verbosity}
\label{sec-verb}
{\tt configure} does a series of architectural probes to figure out how to do
an install on your system. Many of the probes that are run don't produce
output during the configure step. You can tell {\tt configure} that you want
to see more output by cranking up the verbosity. Presently, maximum
verbosity is enabled by adding the flag:
\vspace*{-0.1in}
\begin{verbatim}
-v 2
\end{verbatim}
\subsubsection{Controlling where ATLAS will move files to during install step}
ATLAS supplies some flags to control where ATLAS will move files to when
you do the {\tt make install} step (\S\ref{sec-install}). These flags
are taken from gnu {\tt configure}, and they are:
\begin{itemize}
\item \verb|--prefix=<dirname>| : Top level installation directory.
include files will be moved to \verb|<dirname>/include| and
libraries will be moved to \verb|<dirname>/lib|.
Default: {\tt /usr/local/atlas}
\item \verb|--incdir=<dirname>| : Installation directory for ATLAS's
include files. Default: {\tt /usr/local/atlas/include}.
\item \verb|--incdir=<dirname>| : Installation directory for ATLAS's
libraries. \\ Default: {\tt /usr/local/atlas/lib}.
\end{itemize}
\subsubsection{Telling ATLAS to ignore architectural defaults}
\label{sec-nodefaults}
Architectural defaults are partial results of past searches when
the compiler and architecture are known. They allow you skip the
full ATLAS search, which makes install time much quicker. They also
ensure that you have good results, since they typically represent
several searches and/or user intervention into the usual search so
that maximum performance is found. This doesn't typically mean a huge
performance difference, since the empirical search usually does an adequate
job, but it often provides a few extra percentage points of performance. Also,
occasionally the empirical search will, due to machine load or other timing
problems, produce inadequate code, and using the architectural defaults
prevents this from happening.
By default, ATLAS automatically uses the architectural defaults anytime
it has results for the given architecture and compiler. However, the
compiler detection is based on the compiler name, not version, and so
ATLAS's architectural defaults for gnu {\tt gcc4.2} might not be best
for {\tt gcc3} or apple's {\tt gcc}, etc, even though {\tt configure} would
use the architectural defaults in such cases.
So, there are times when you want to tell ATLAS to ignore any architectural
defaults it might have. Common reasons include the fact that you have
overridden the compiler flags ATLAS uses, or are using an earlier version
of the supported compiler. In these cases, the best idea is often to
install both with and without the architectural defaults, and compare timings.
If both your installs (homegrown-compiler/flags+archdef,
homegrown-compiler/flags+search) are slower than the architectural defaults
using the default compiler, you should probably install the default
compiler. However, if your results are largely the same, you know your
changes haven't depressed performance and so it is OK to use the generated
libraries (see Section~\ref{sec-time} for details on timing an ATLAS
install). If your timing results are substantially better, and you haven't
enabled IEEE-destroying flags, you should send your improved compiler and
flags to the ATLAS team!
To force ATLAS to ignore the architectural defaults (and thus to perform
a full ATLAS search), pass the following flags to {\tt configure}:
\begin{verbatim}
-Si archdef 0
\end{verbatim}
\section{The ATLAS build step}
\label{sec-build}
This is the step where ATLAS performs all its empirical tuning, and then
uses the discovered kernels to build all required libraries. It uses
the {\tt BLDdir} created by the configure step, and is invoked from
the {\tt BLDdir} with the {\tt make build} command, or simply by {\tt make}.
This step can be quite long, depending on your platform and whether or
not you use architectural defaults. For a system like the Core2Duo with
architectural defaults, the build step may take 10 or 20 minutes, while in
order to complete a full ATLAS search on a slower platform (eg. MIPS) could
take anywhere between a couple of hours and a full day.
\section{The ATLAS check step}
\label{sec-check}
In this optional step, ATLAS runs various testers in order to make sure
that the generated library is not producing completely bogus results.
For each precision, ATLAS runs the standard BLAS testers (both C and F77
interface), and then various of ATLAS's homegrown testers that appear
in {\tt ATLAS/bin}. If you have installed without a FORTRAN compiler,
then the standard BLAS testers cannot be run (the standard BLAS testers,
downloadable from netlib, require FORTRAN even to test the C interface),
and so your testing will be less comprehensive.
There are two possible targets, {\tt check} which tests ATLAS's serial
routines, and {\tt ptcheck} which check the parallel routines. You cannot
run {\tt ptcheck} if you haven't installed the parallel libraries.
This step is invoked from {\tt BLDdir} by typing:
\begin{verbatim}
make check # test serial routines
make ptcheck # check parallel routines
\end{verbatim}
Both of these commands will first do a lot of compilation, and then they
will finish with results such as:
\vspace*{-0.1in}
\begin{verbatim}
core2.home.net. make check
...................................................
..... A WHOLE LOT OF COMPILATION AND RUNNING ......
...................................................
DONE BUILDING TESTERS, RUNNING:
SCOPING FOR FAILURES IN BIN TESTS:
fgrep -e fault -e FAULT -e error -e ERROR -e fail -e FAIL \
bin/sanity.out
8 cases: 8 passed, 0 skipped, 0 failed
4 cases: 4 passed, 0 skipped, 0 failed
8 cases: 8 passed, 0 skipped, 0 failed
4 cases: 4 passed, 0 skipped, 0 failed
8 cases: 8 passed, 0 skipped, 0 failed
4 cases: 4 passed, 0 skipped, 0 failed
8 cases: 8 passed, 0 skipped, 0 failed
4 cases: 4 passed, 0 skipped, 0 failed
DONE
SCOPING FOR FAILURES IN CBLAS TESTS:
fgrep -e fault -e FAULT -e error -e ERROR -e fail -e FAIL \
interfaces/blas/C/testing/sanity.out | \
fgrep -v PASSED
make[1]: [sanity_test] Error 1 (ignored)
DONE
SCOPING FOR FAILURES IN F77BLAS TESTS:
fgrep -e fault -e FAULT -e error -e ERROR -e fail -e FAIL \
interfaces/blas/F77/testing/sanity.out | \
fgrep -v PASSED
make[1]: [sanity_test] Error 1 (ignored)
DONE
make[1]: Leaving directory `/home/whaley/TEST/ATLAS3.7.36.0/obj64'
\end{verbatim}
Notice that the {\tt Error 1 (ignored)} commands come from {\tt make}, and they
indicate that {\tt fgrep} is not finding any errors in the output files
(thus this {\tt make} output does not represent the finding of an error).
When true errors occur, the lines of the form
\vspace*{-0.1in}
\begin{verbatim}
8 cases: 8 passed, 0 skipped, 0 failed
\end{verbatim}
will have non-zero numbers for {\tt failed}, or you will see other tester
output discussing errors, such as the printing of large residuals.
As mentioned, this is really sanity checking, and it runs only a few tests
on a handful of problem sizes. This is usually adequate to catch most
blatant problems (eg., compiler producing incorrect output). More subtle
or rarely-occurring bugs may require running the LAPACK and/or full ATLAS
testers. The ATLAS developer guide~\cite{atlas-devel} provides instructions
on how to use the full ATLAS tester, as well as help in diagnosing problems.
The developer guide is provided in the ATLAS tarfile as
{\tt ATLAS/doc/atlas\_devel.pdf}
\section{The ATLAS time step}
\label{sec-time}
In this optional step, ATLAS times certain kernel routines and reports their
performance as a percentage of clock rate. Its purpose is to provide a
quick way to ensure that your install has resulted in a library that obtains
adequate performance. If you are installing using architectural
defaults, this step will print a timing comparison against the performance
that the ATLAS maintainer got when creating the architectural defaults.
To invoke this step, issue the following command in your {\tt BLDdir}:
\vspace*{-0.1in}
\begin{verbatim}
make time
\end{verbatim}
\begin{figure}[p]
\begin{footnotesize}
\begin{verbatim}
NAMING ABBREVIATIONS:
kSelMM : selected matmul kernel (may be hand-tuned)
kGenMM : generated matmul kernel
kMM_NT : worst no-copy kernel
kMM_TN : best no-copy kernel
BIG_MM : large GEMM timing (usually N=1600); estimate of asymptotic peak
kMV_N : NoTranspose matvec kernel
kMV_T : Transpose matvec kernel
kGER : GER (rank-1 update) kernel
Kernel routines are not called by the user directly, and their
performance is often somewhat different than the total
algorithm (eg, dGER perf may differ from dkGER)
Reference clock rate=2394Mhz, new rate=2394Mhz
Refrenc : % of clock rate achieved by reference install
Present : % of clock rate achieved by present ATLAS install
single precision double precision
******************************** *******************************
real complex real complex
--------------- --------------- --------------- ---------------
Benchmark Refrenc Present Refrenc Present Refrenc Present Refrenc Present
========= ======= ======= ======= ======= ======= ======= ======= =======
kSelMM 535.0 551.4 525.4 509.6 311.5 312.7 298.0 296.5
kGenMM 175.5 174.0 175.5 173.6 160.5 159.7 165.4 166.9
kMM_NT 145.2 143.7 149.3 150.7 135.3 131.0 132.3 134.3
kMM_TN 163.2 158.0 161.1 164.6 148.7 144.8 146.0 155.4
BIG_MM 510.1 544.5 504.0 545.9 307.7 301.5 293.0 304.9
kMV_N 113.5 109.1 216.9 208.3 58.9 56.2 97.4 88.8
kMV_T 89.9 85.9 94.6 96.4 47.2 44.4 74.1 77.1
kGER 154.2 154.1 119.4 116.9 29.1 26.0 46.8 45.6
\end{verbatim}
\end{footnotesize}
\caption{Normal results for {\tt make time} on Core2Duo64SSE3}
\label{fig-tim}
\end{figure}
\begin{figure}[p]
\begin{footnotesize}
\begin{verbatim}
Reference clock rate=2200Mhz, new rate=1597Mhz
....
single precision double precision
******************************** *******************************
real complex real complex
--------------- --------------- --------------- ---------------
Benchmark Refrenc Present Refrenc Present Refrenc Present Refrenc Present
========= ======= ======= ======= ======= ======= ======= ======= =======
kSelMM 335.5 338.8 329.4 331.6 178.9 180.8 180.3 178.7
kGenMM 175.4 100.4 174.2 100.3 163.7 92.6 141.4 94.9
kMM_NT 142.0 86.8 141.2 92.0 125.3 85.2 138.1 88.8
kMM_TN 143.0 92.7 141.1 95.2 139.4 87.8 137.4 90.1
BIG_MM 327.1 325.2 318.6 320.0 169.8 171.3 171.0 172.0
kMV_N 61.4 35.5 139.3 98.9 47.2 30.7 71.9 74.2
kMV_T 73.6 53.6 75.3 62.5 31.6 20.2 52.7 36.6
kGER 43.6 28.8 91.8 65.1 23.7 18.3 46.8 40.3
\end{verbatim}
\end{footnotesize}
\caption{Timings results when architectural defaults are compiled with
substandard gcc4.1, rather than gcc4.2}
\label{fig-tim-gcc4}
\end{figure}
In Figure~\ref{fig-tim} we see a typical printout of a successful install,
in this case ran on my 2.4Ghz Core2Duo.
The {\tt Refrenc}
columns provide the performance achieved by the architectural defaults when
they were
originally created, while the {\tt Present} columns provide the results
obtained using the new ATLAS install we have just completed. We see that
the {\tt Present} columns wins occasionally (eg. single precision real
{\tt kSelMM}), and loses sometimes (eg. single precision complex {\tt kSelMM}),
but that the timings are relatively similar across the board. This tells
us that the install is OK from a performance angle.
As a general rule, performance for both data types of a particular precision
should be roughly comparable, but may vary dramatically between precisions
(due mainly to differing vector lengths in SIMD instructions).
The timings are normalized to the clock rate, which is why the clock rate
of both the reference and present install are printed. It is expected that
as clock rates rise, performance as a percent of it may fall slightly
(since memory bus speeds do not usually rise in exact lockstep). Therefore,
if I installed on a 3.2Ghz Core2Duo, I would not be surprised if the
{\tt Present} install lost by a few percentage points in most cases.
True problems typically display a significant loss that occurs in a pattern.
The most common problem is from installing with a poor compiler, which will
lower the performance of most compiled kernels, without affecting the speed
of assembly kernels. Figure~\ref{fig-tim-gcc4} shows
such an example, where {\tt gcc 4.1} (a terrible
compiler for floating point arithmetic on x86 machines) has been used to
install ATLAS on an Opteron, rather than {\tt gcc 4.2}, which was the compiler
that was used to create the architectural defaults.
Here, we see that the present machine is actually slower than the machine
that was used to create the defaults, so if anything, we expect it to
achieve a greater percentage of clock rate. Indeed, this is more or less
true of the first line, {\tt kSelMM}. On this platform, {\tt kSelMM} is
written totally in assembly, and {\tt BIG\_MM} calls these kernels,
and so the {\tt Present} results are good for these rows. All the other
rows show kernels that are written in C, and so we see that the use of
a bad compiler has markedly depressed performance across the board.
Anytime you see a pattern such as this, the first thing you should check
is if you are using a recommended compiler, and if not, install and use
that compiler.
On the other hand, if only your {\tt BIG\_MM} column is depressed, it is
likely you have a bad setting for the {\tt CacheEdge} or the complex-to-real
crossover point (if the performance is depressed only for both complex types).
\subsection{Contrasting non-default install performance}
\label{sec-time-nondefault}
If you do not install using the architectural defaults, {\tt make time} will
only print out the {\tt Present} columns. This gives you a good summary of
ATLAS's library performance, but it can be hard to tell what is good and
bad if you are not familiar with ATLAS on this hardware. Sometimes, ATLAS
has architectural defaults for your platform, but your install doesn't use
them. This is usually because the installer has specified the use of a
non-default compiler, or has explicitly asked that the architectural defaults
not be used, or has overridden the detection of the architecture, etc. In
this case, {\tt make time} does not do the comparison
against the architectural defaults, and so only the {\tt Present} columns
are printed.
However, if you wish to ensure that your library is as good as one that
uses the architectural defaults, then you can manually tell the program
called by {\tt make time} ({\tt xatlbench} to do the comparison. The most
common example would
be you have switched to an unsupported compiler (eg., the Intel compiler),
and now you want to see if the library you built using it is as fast or faster
than the one using the default {\tt gcc 4.2} compiler. Another example would
be that you want to compare the performance of two closely related
architectures. This is what we will do here, where we contrast the performance
of the 32 and 64 bit versions of the library on my Core2Duo.
In order to manually do a comparison between a present install and any of
the results stored in ATLAS's architectural defaults you'll need to
perform the following steps:
\begin{enumerate}
\item {\tt make time} issued in the {\tt BLDdir} of your non-default install.
This does the timings of the present build, and stores the results
in {\tt BLDdir/bin/INSTALL\_LOG}.
\item {\tt cd SRCdir/CONFIG/ARCHS}, and find the tarfile containing the
results you wish to compare against. In our case, we choose
{\tt Core2Duo32SSE3.tgz} to compare against our own Core2Duo64SSE
results.
\item {\tt gunzip -c Core2Duo32SSE3.tgz | tar xvf -} untars the
selected architectural results (replace {\tt Core2Duo32SSE3.tgz} with the
tarfile you have selected in step\#2).
\item {\tt cd BLDdir}
\item \verb|./xatlbench -dp SRCdir/CONFIG/ARCHS/<ARCH> -dc BLDdir/bin/INSTALL_LOG| \\
{\tt xatlbench} is the program that compares two sets of results, with
the {\tt -dp} pointing to the previous ({\tt Refrenc}) install result
directory and {\tt -dc} pointing to the current ({\tt Present})
install result directory.
\end{enumerate}
Figure~\ref{fig-tim-32v64} shows me doing this on my Core2Duo, with
{\tt SRCdir} = {\tt /home/whaley/TEST/ATLAS3.7.36.0} and
{\tt BLDdir} = {\tt /home/whaley/TEST/ATLAS3.7.36.0/obj64}, where we compare
the present 64-bit install to the stored 32-bit install.
We see that the 64-bit install, which gets to use 16 rather than 8 registers,
is slightly faster for almost all kernels and precisions, as one might expect.
\begin{figure}[htbp]
\begin{footnotesize}
\begin{verbatim}
core2.home.net. cd /home/whaley/TEST/ATLAS3.7.36.0/obj64
core2.home.net. make time
..... lots of output .....
core2.home.net. pushd ~/TEST/ATLAS3.7.36.0/CONFIG/ARCHS/
core2.home.net. ls
BOZOL1.tgz CreateTar.sh MIPSICE964.tgz POWER564.tgz
Core2Duo32SSE3/ HAMMER64SSE2.tgz MIPSR1xK64.tgz PPCG532AltiVec.tgz
Core2Duo32SSE3.tgz HAMMER64SSE3.tgz negflt.c PPCG564AltiVec.tgz
Core2Duo64SSE3/ IA64Itan264.tgz P432SSE2.tgz USIV32.tgz
Core2Duo64SSE3.tgz KillDirs.sh P4E32SSE3.tgz USIV64.tgz
CoreDuo32SSE3.tgz Make.ext P4E64SSE3.tgz
CreateDef.sh Makefile POWER432.tgz
CreateDirs.sh MIPSICE932.tgz POWER464.tgz
core2.home.net. gunzip -c Core2Duo32SSE3.tgz | tar xvf -
..... lots of output .....
core2.home.net. pushd
core2.home.net. ./xatlbench \
-dp /home/whaley/TEST/ATLAS3.7.36.0/CONFIG/ARCHS/Core2Duo32SSE3 \
-dc /home/whaley/TEST/ATLAS3.7.36.0/obj64/bin/INSTALL_LOG/
.....
Reference clock rate=2394Mhz, new rate=2394Mhz
.....
single precision double precision
******************************** *******************************
real complex real complex
--------------- --------------- --------------- ---------------
Benchmark Refrenc Present Refrenc Present Refrenc Present Refrenc Present
========= ======= ======= ======= ======= ======= ======= ======= =======
kSelMM 539.0 551.4 496.5 509.6 299.4 312.7 289.0 296.5
kGenMM 165.1 174.0 165.1 173.6 156.1 159.7 153.8 166.9
kMM_NT 137.6 143.7 134.7 150.7 115.7 131.0 123.5 134.3
kMM_TN 116.3 158.0 112.3 164.6 101.3 144.8 110.9 155.4
BIG_MM 521.3 544.5 476.5 545.9 282.6 301.5 282.8 304.9
kMV_N 69.0 109.1 206.9 208.3 56.3 56.2 69.4 88.8
kMV_T 84.8 85.9 117.3 96.4 48.0 44.4 87.9 77.1
kGER 90.1 154.1 114.2 116.9 27.9 26.0 41.5 45.6
\end{verbatim}
\end{footnotesize}
\caption{Comparing 32 and 64 bit libraries on a 2.4 Ghz Core2Duo}
\label{fig-tim-32v64}
\end{figure}
\subsection{Discussion of timing targets}
Presently, ATLAS times mostly kernel routines, which are used to build
higher level routines that then appear in the BLAS or LAPACK.
{\tt kSelMM} is the matrix multiply kernel that is being used for large
GEMM calls, which will be the best kernel found in the generator and
multiple implementation searches. Therefore this kernel may be written
in assembly on some platforms.
{\tt kGenMM} is the fastest generated kernel that matches {\tt kSelMM},
and it may be used for some types of cleanup. All generated kernels
are written in ANSI~C, and thus their peak performance will strongly
depend on the compiler being used.
{\tt kMM\_NT} and {\tt kMM\_TN} are two of the four generated kernels that
will be used for small-case GEMM when we cannot afford to copy the input
matrices. The last two characters indicate the transpose settings.
The other two kernels' performance lies between these extremes:
{\tt NT} is typically the slowest kernel (all non-contiguous access),
and {\tt TN} is typically the fastest (all contiguous access).
{\tt BIG\_MM} is the only non-kernel timing we presently report, and it is
the speed found when doing a large GEMM call. ``Large'' can vary by platform:
it is typically $M=N=K=1600$, except where we were unable to allocate that
much memory, where it will be less. On many machines, this line gives you
a rough asymptotic bound on BLAS performance.
The next three lines report Level 2 BLAS kernel performance (the Level~2
BLAS' performance will follow these kernels in roughly the same way that
the Level~3 follow the GEMM kernels).
We should eventually supply an expanded timing comparison that would include
higher level timings, such as LAPACK routines and threaded performance,
but do not currently do so.
\section{The ATLAS install step}
\label{sec-install-step}
This final optional step instructs ATLAS to copy the created libraries
and include files into the appropriate directories, as specified in
the configure step. This functionality is new, and so far
is not bullet-proof (for instance, it copies only static libraries,
and so presently fails to copy any dynamic libraries the user has built).
From your {\tt BLDdir}, it may be invoked by:
\vspace*{-0.1in}
\begin{verbatim}
make install
\end{verbatim}
By default, this command will copy all the static libraries to
{\tt /usr/local/atlas/lib} and all the user-includable header files to
{\tt /usr/local/atlas/include}. You may override this default directory
during the configure step using the gnu-like flags {\tt --prefix},
{\tt --incdir} and/or {\tt --libdir}. Assuming you didn't issue
{\tt --incdir} or {\tt --libdir}, you can also override the prefix
directory at install time with the command:
\vspace*{-0.1in}
\begin{verbatim}
make install DESTDIR=<prefix directory to install atlas in>
\end{verbatim}
\section{Example: Installing ATLAS with full LAPACK on Linux/AMD64}
\label{sec-example}
In this section, I show a complete ATLAS install, including installing
LAPACK. We assume I have already downloaded the tarfiles
{\tt atlas3.9.12.tar.bz2} and {\tt lapack.tgz} into the
{\tt /home/whaley/dload} directory.
%We will assume we are installing
%both libraries in {\tt /home/whaley/numerics}.
\subsection{Figuring out configure flags}
\label{sec-ex-configure}
The system is a Fedora Core 8 system, which unfortunately uses the broken
{\tt gcc 4.1.2}, which would cripple ATLAS performance. Therefore, prior to
installing ATLAS, I have installed gcc~4.2.1, with
{\tt --prefix=/home/whaley/local/gcc-4.2.1}
I therefore add the following lines to my {\tt .cshrc} so that ATLAS
will use this gcc (it is put first in the path), and will be able
to find the {\tt gcc 4.2} libraries:
\begin{footnotesize}
\vspace*{-0.1in}
\begin{verbatim}
set path = (/home/whaley/local/gcc-4.2.1/bin $path)
setenv LD_LIBRARY_PATH /home/whaley/local/gcc-4.2.1/lib64:/home/whaley/local/gcc-4.2.1/lib
\end{verbatim}
\end{footnotesize}
\noindent
I source the C shell startup file, and then check that I'm now getting the
correct compiler:
\vspace*{-0.1in}
\begin{verbatim}
etl-opt8>source ~/.cshrc
etl-opt8>gcc -v
Using built-in specs.
Target: x86_64-unknown-linux-gnu
Configured with: ../configure --prefix=/home/whaley/local/gcc-4.2.1 --enable-languages=c
Thread model: posix
gcc version 4.2.1
\end{verbatim}
Now, I don't need to pass a lot of flags to set what compiler to use,
since ATLAS will find {\tt gcc 4.2} as the first compiler, and it will
have the libraries it needs to work. However, I want to build dynamic
libraries for this install, so I know I'll need to add the
{\tt --shared} configure flag; config will automatically add the required
{\tt -fPIC} flag to all gnu compilers so they can build shared object code.
Now, I do a top on {\tt etl-opt8} (the machine name) and see that
I'm alone on the machine. Therefore, I will want to
use the cycle-accurate x86-specific wall timer in order to improve the
accuracy of my install. This requires me to figure out what the Mhz of
my machine is.
Under Linux, I can discover this with {\tt cat /proc/cpuinfo}, which tells
me {\tt cpu~MHz~:~2100.000}. Therefore, I will throw
{\tt -D c -DPentiumCPS=2100}.
I want ATLAS to install the resulting libraries and header files in
the directory \\
{\tt /home/whaley/local/atlas}, so I'll pass
{\tt --prefix=/home/whaley/local/atlas} as well.
I want a 64 bit install, and to build a full LAPACK library, so I will also
want to throw {\tt -b 64} and \\
\verb|--with-netlib-lapack-tarfile=/home/whaley/dload/lapack.tgz|.
\Wskip{
\subsection{Creating source directories and building LAPACK}
\label{sec-ex-lapack}
Now that I know what ATLAS {\tt configure} flags I want to throw, I am ready to
begin installing LAPACK. The first step is to create the {\tt numerics}
directory that will hold the libraries, and untar them:
\vspace*{-0.1in}
\begin{verbatim}
animal>cd ~/
animal>mkdir numerics
animal>cd numerics/
animal>bunzip2 -c ~/atlas3.8.0.tar.bz2 | tar xfm -
animal>mv ATLAS ATLAS3.8.0
animal>gunzip -c ~/dload/lapack-3.1.1.tgz | tar xfm -
animal>ls
ATLAS3.7.38/ lapack-3.1.1/
\end{verbatim}
Now, we are needing to set the LAPACK Make.inc appropriately. First,
I go into the LAPACK directory, and copy the platform-specific {\tt make.inc}
to {\tt make.inc}. In my case this is:
\vspace*{-0.1in}
\begin{verbatim}
animal>cd lapack-3.1.1/
animal>cp INSTALL/make.inc.LINUX make.inc
\end{verbatim}
I now edit the created {\tt make.inc} ({\tt vi make.inc}), and here are
the {\tt make} macros that I change:
\vspace*{-0.1in}
\begin{verbatim}
FORTRAN = <want to set to ATLAS's F77 macro>
OPTS = <want to set to ATLAS F77FLAGS macro>
DRVOPTS = $(OPTS)
NOOPT = <F77FLAGS w/o optimization>
LOADER = $(FORTRAN)
LOADOPTS = $(OPTS)
TIMER = <need to know what compiler I'm using to set>
\end{verbatim}
So far, I have only been able to fill in {\tt DRVOPTS}, {\tt LOADER} and
{\tt LOADOPTS},
which are defined in terms of the macros I've yet to fill in! The reason
is that I want to use the same compiler and flags as ATLAS, so that
I'm sure my LAPACK library can interoperate with my ATLAS-tuned library.
I will set the {\tt FORTRAN} macro to the compiler indicated by ATLAS's
{\tt F77} macro, and {\tt OPTS} will be the same as {\tt F77FLAGS}.
So, I change to the ATLAS source directory, and produce a dry-run {\tt BLDdir}
in order to get this information by:
\vspace*{-0.1in}
\begin{verbatim}
animal>cd ../ATLAS3.7.38/
animal>mkdir bogus
animal>cd bogus/
animal>../configure -b 64 -D c -DPentiumCPS=2200 -Fa alg -fPIC
...................................................
............<A WHOLE LOT OF OUTPUT>................
...................................................
animal>fgrep "F77 =" Make.inc
F77 = gfortran
animal>fgrep "F77FLAGS =" Make.inc
F77FLAGS = -fomit-frame-pointer -mfpmath=387 -O2 -falign-loops=4 -fPIC -m64
\end{verbatim}
With this info in hand, I am ready to delete this
bogus directory, and go back and edit the LAPACK make.inc:
\vspace*{-0.1in}
\begin{verbatim}
animal>cd ..
animal>rm -rf bogus/
animal>cd ../lapack-3.1.1/
animal>vi make.inc
\end{verbatim}
I now fill in my {\tt make.inc} macros as:
\vspace*{-0.1in}
\begin{verbatim}
FORTRAN = gfortran
OPTS = -fomit-frame-pointer -mfpmath=387 -O2 -falign-loops=4 -fPIC -m64
DRVOPTS = $(OPTS)
NOOPT = -fomit-frame-pointer -mfpmath=387 -m64
LOADER = $(FORTRAN)
LOADOPTS = $(OPTS)
TIMER = INT_ETIME
\end{verbatim}
I chose the setting of {\tt TIMER} based on the fact that the example file's
comments said it is the correct setting when the compiler is {\tt gfortran}.
Now I perform the LAPACK install:
\vspace*{-0.1in}
\begin{footnotesize}
\begin{verbatim}
animal>make lib
./testdlamch; ./testsecond; ./testdsecnd; ./testversion )
make[1]: Entering directory `/home/whaley/numerics/lapack-3.1.1/INSTALL'
gfortran -fomit-frame-pointer -mfpmath=387 -O2 -falign-loops=4 -m64 -c lsame.f -o lsame.o
........................................................
.............<WHOLE LOT OF COMPILATION>.................
........................................................
ar cr ../../tmglib_LINUX.a slatms.o slatme.o slatmr.o slagge.o slagsy.o
slakf2.o slarge.o slaror.o slarot.o slatm2.o slatm3.o slatm5.o slatm6.o
clatms.o clatme.o clatmr.o clagge.o claghe.o clagsy.o clakf2.o clarge.o
claror.o clarot.o clatm1.o clarnd.o clatm2.o clatm3.o clatm5.o clatm6.o
slatm1.o slaran.o slarnd.o dlatms.o dlatme.o dlatmr.o dlagge.o dlagsy.o
dlakf2.o dlarge.o dlaror.o dlarot.o dlatm2.o dlatm3.o dlatm5.o dlatm6.o
zlatms.o zlatme.o zlatmr.o zlagge.o zlaghe.o zlagsy.o zlakf2.o zlarge.o
zlaror.o zlarot.o zlatm1.o zlarnd.o zlatm2.o zlatm3.o zlatm5.o zlatm6.o
dlatm1.o dlaran.o dlarnd.o
ranlib ../../tmglib_LINUX.a
make[1]: Leaving directory `/home/whaley/numerics/lapack-3.1.1/TESTING/MATGEN'
227.482u 20.093s 4:09.81 99.1% 0+0k 0+0io 12pf+0w
animal>
animal>ls
BLAS/ INSTALL/ make.inc README tmglib_LINUX.a
COPYING lapack_LINUX.a make.inc.example SRC/
html/ Makefile manpages/ TESTING/
\end{verbatim}
\end{footnotesize}
So, we have succesfully created the LAPACK library, and now we need to
install ATLAS and a complete LAPACK using it.
}
\subsection{Creating {\tt BLDdir} and installing ATLAS}
\label{sec-ex-install}
I'm ready to install ATLAS and LAPACK. I just need to untar the
ATLAS tarfile, issue, create my {\tt BLDdir}, and issue the
previously selected flags to configure:
\vspace*{-0.1in}
\begin{footnotesize}
\begin{verbatim}
etl-opt8>bunzip2 -c ~/dload/atlas3.9.12.tar.bz2 | tar xfm -
etl-opt8>mv ATLAS ATLAS3.9.12.1
etl-opt8>cd ATLAS3.9.12.1/
etl-opt8>mkdir obj64
etl-opt8>cd obj64/
etl-opt8>../configure -b 64 -D c -DPentiumCPS=2100 -Fa alg --shared \
--prefix=/home/whaley/local/atlas \
--with-netlib-lapack-tarfile=/home/whaley/dload/lapack.tgz
...................................................
............<A WHOLE LOT OF OUTPUT>................
...................................................
etl-opt8>ls
ARCHS/ Makefile xconfig* xprobe_3dnow* xprobe_OS*
atlcomp.txt Make.inc xctest* xprobe_arch* xprobe_pmake*
atlconf.txt Make.top xf2cint* xprobe_asm* xprobe_sse1*
bin/ src/ xf2cname* xprobe_comp* xprobe_sse2*
include/ tune/ xf2cstr* xprobe_f2c* xprobe_sse3*
interfaces/ xarchinfo_linux* xf77test* xprobe_gas_x8632* xprobe_vec*
lib/ xarchinfo_x86* xflibchk* xprobe_gas_x8664* xspew*
etl-opt8>make
.........................................................
............<A WHOLE WHOLE LOT OF OUTPUT>................
.........................................................
ATLAS install complete. Examine
ATLAS/bin/<arch>/INSTALL_LOG/SUMMARY.LOG for details.
make[1]: Leaving directory `/home/whaley/TEST/ATLAS3.9.12.1/obj64'
make clean
make[1]: Entering directory `/home/whaley/TEST/ATLAS3.9.12.1/obj64'
rm -f *.o x* config?.out *core*
make[1]: Leaving directory `/home/whaley/TEST/ATLAS3.9.12.1/obj64'
1628.011u 153.212s 23:05.34 128.5% 0+0k 32+3325928io 0pf+0w
\end{verbatim}
\end{footnotesize}
OK, in a little over 20 minutes, we've got ATLAS and LAPACK built. Now, we need
to see if it passes the sanity tests, which we do by:
\vspace*{-0.1in}
\begin{footnotesize}
\begin{verbatim}
etl-opt8>make check
........................................................
............<A WHOLE LOT OF COMPILATION>................
........................................................
DONE BUILDING TESTERS, RUNNING:
SCOPING FOR FAILURES IN BIN TESTS:
fgrep -e fault -e FAULT -e error -e ERROR -e fail -e FAIL \
bin/sanity.out
8 cases: 8 passed, 0 skipped, 0 failed
4 cases: 4 passed, 0 skipped, 0 failed
8 cases: 8 passed, 0 skipped, 0 failed
4 cases: 4 passed, 0 skipped, 0 failed
8 cases: 8 passed, 0 skipped, 0 failed
4 cases: 4 passed, 0 skipped, 0 failed
8 cases: 8 passed, 0 skipped, 0 failed
4 cases: 4 passed, 0 skipped, 0 failed
DONE
SCOPING FOR FAILURES IN CBLAS TESTS:
fgrep -e fault -e FAULT -e error -e ERROR -e fail -e FAIL \
interfaces/blas/C/testing/sanity.out | \
fgrep -v PASSED
make[1]: [sanity_test] Error 1 (ignored)
DONE
SCOPING FOR FAILURES IN F77BLAS TESTS:
fgrep -e fault -e FAULT -e error -e ERROR -e fail -e FAIL \
interfaces/blas/F77/testing/sanity.out | \
fgrep -v PASSED
make[1]: [sanity_test] Error 1 (ignored)
DONE
make[1]: Leaving directory `/home/whaley/TEST/ATLAS3.9.12.1/obj64'
61.684u 6.485s 1:08.66 99.2% 0+0k 0+163768io 0pf+0w
\end{verbatim}
\end{footnotesize}
So, since we see no failures, we passed. I get essentially the same
output when I check the parallel interfaces (my machine has eight processors)
via {\tt make ptcheck}.
Now, I am ready to make sure my libraries are getting the expected performance,
so I do:
\vspace*{-0.1in}
\begin{footnotesize}
{\samepage
\begin{verbatim}
etl-opt8>make time
........................................................
............<A WHOLE LOT OF COMPILATION>................
........................................................
single precision double precision
******************************** *******************************
real complex real complex
--------------- --------------- --------------- ---------------
Benchmark Refrenc Present Refrenc Present Refrenc Present Refrenc Present
========= ======= ======= ======= ======= ======= ======= ======= =======
kSelMM 643.4 642.9 622.0 621.8 323.8 343.5 320.5 316.9
kGenMM 191.4 192.1 161.8 174.1 178.3 164.3 172.9 172.4
kMM_NT 140.0 138.5 127.4 129.3 137.4 136.1 126.4 131.8
kMM_TN 165.2 165.3 159.8 157.0 163.0 161.6 158.0 155.2
BIG_MM 604.1 617.0 601.8 599.8 311.3 332.3 309.2 292.1
kMV_N 74.3 70.2 211.2 197.5 51.9 48.4 107.3 99.7
kMV_T 82.2 79.8 97.2 95.3 46.4 43.9 77.6 73.3
kGER 60.1 56.9 153.5 130.3 38.8 32.0 77.5 64.8
\end{verbatim}
}
\end{footnotesize}
We see that load and timer issues have made it so there is not an exact
match, but that neither install is worse overall, and so this install
looks good! Now we are finally ready to install the libraries. We can
do so, and then check what got installed by:
\begin{footnotesize}
\begin{verbatim}
etl-opt8>make install
...............................................
..............<A LOT OF OUTPUT>................
...............................................
etl-opt8>cd ~/local/atlas/
etl-opt8>ls
include/ lib/
etl-opt8>ls include/
atlas/ cblas.h clapack.h
etl-opt8>ls include/atlas/
atlas_buildinfo.h atlas_dr1kernels.h atlas_strsmXover.h
atlas_cacheedge.h atlas_dr1_L1.h atlas_tcacheedge.h
atlas_cGetNB_gelqf.h atlas_dr1_L2.h atlas_trsmNB.h
atlas_cGetNB_geqlf.h atlas_dsyr2.h atlas_type.h
atlas_cGetNB_geqrf.h atlas_dsyr.h atlas_zdNKB.h
atlas_cGetNB_gerqf.h atlas_dsyr_L1.h atlas_zGetNB_gelqf.h
atlas_cmv.h atlas_dsyr_L2.h atlas_zGetNB_geqlf.h
atlas_cmvN.h atlas_dsysinfo.h atlas_zGetNB_geqrf.h
atlas_cmvS.h atlas_dtGetNB_gelqf.h atlas_zGetNB_gerqf.h
atlas_cmvT.h atlas_dtGetNB_geqlf.h atlas_zmv.h
atlas_cNCmm.h atlas_dtGetNB_geqrf.h atlas_zmvN.h
atlas_cr1.h atlas_dtGetNB_gerqf.h atlas_zmvS.h
atlas_cr1kernels.h atlas_dtrsmXover.h atlas_zmvT.h
atlas_cr1_L1.h atlas_pthreads.h atlas_zNCmm.h
atlas_cr1_L2.h atlas_sGetNB_gelqf.h atlas_zr1.h
atlas_csNKB.h atlas_sGetNB_geqlf.h atlas_zr1kernels.h
atlas_csyr2.h atlas_sGetNB_geqrf.h atlas_zr1_L1.h
atlas_csyr.h atlas_sGetNB_gerqf.h atlas_zr1_L2.h
atlas_csyr_L1.h atlas_smv.h atlas_zsyr2.h
atlas_csyr_L2.h atlas_smvN.h atlas_zsyr.h
atlas_csysinfo.h atlas_smvS.h atlas_zsyr_L1.h
atlas_ctGetNB_gelqf.h atlas_smvT.h atlas_zsyr_L2.h
atlas_ctGetNB_geqlf.h atlas_sNCmm.h atlas_zsysinfo.h
atlas_ctGetNB_geqrf.h atlas_sr1.h atlas_ztGetNB_gelqf.h
atlas_ctGetNB_gerqf.h atlas_sr1kernels.h atlas_ztGetNB_geqlf.h
atlas_ctrsmXover.h atlas_sr1_L1.h atlas_ztGetNB_geqrf.h
atlas_dGetNB_gelqf.h atlas_sr1_L2.h atlas_ztGetNB_gerqf.h
atlas_dGetNB_geqlf.h atlas_ssyr2.h atlas_ztrsmXover.h
atlas_dGetNB_geqrf.h atlas_ssyr.h cmm.h
atlas_dGetNB_gerqf.h atlas_ssyr_L1.h cXover.h
atlas_dmv.h atlas_ssyr_L2.h dmm.h
atlas_dmvN.h atlas_ssysinfo.h dXover.h
atlas_dmvS.h atlas_stGetNB_gelqf.h smm.h
atlas_dmvT.h atlas_stGetNB_geqlf.h sXover.h
atlas_dNCmm.h atlas_stGetNB_geqrf.h zmm.h
atlas_dr1.h atlas_stGetNB_gerqf.h zXover.h
etl-opt8>ls lib/
libatlas.a libcblas.so liblapack.a libptcblas.so
libatlas.so libf77blas.a liblapack.so libptf77blas.a
libcblas.a libf77blas.so libptcblas.a libptf77blas.so
\end{verbatim}
\end{footnotesize}
The shared object support in ATLAS is still experimental, so we
can get some idea if our shared objects work by running an undocumented
tester. To try a dynamically linked LU factorization, we:
\begin{footnotesize}
\begin{verbatim}
animal>cd ../bin
animal>make xdlutst_dyn
...............................................................
............<A WHOLE LOT OF UP-TO-DATE CHECKING>...............
...............................................................
make[1]: Leaving directory `/home/whaley/numerics/ATLAS3.7.38/animal64/bin'
gfortran -O -fPIC -m64 -o xdlutst_dyn dlutst.o \
/home/whaley/numerics/ATLAS3.7.38/animal64/lib/libtstatlas.a \
/home/whaley/numerics/ATLAS3.7.38/animal64/lib/liblapack.so \
/home/whaley/numerics/ATLAS3.7.38/animal64/lib/libf77blas.so \
/home/whaley/numerics/ATLAS3.7.38/animal64/lib/libcblas.so \
/home/whaley/numerics/ATLAS3.7.38/animal64/lib/libatlas.so \
-Wl,--rpath /home/whaley/numerics/ATLAS3.7.38/animal64/lib
animal>./xdlutst_dyn
NREPS Major M N lda NPVTS TIME MFLOP RESID
===== ===== ===== ===== ===== ===== ======== ======== ========
0 Col 100 100 100 95 0.001 1273.153 1.416e-02
0 Col 200 200 200 194 0.002 2453.930 1.087e-02
0 Col 300 300 300 295 0.007 2574.077 8.561e-03
0 Col 400 400 400 394 0.017 2531.312 8.480e-03
0 Col 500 500 500 490 0.031 2701.090 7.610e-03
0 Col 600 600 600 594 0.051 2796.150 8.332e-03
0 Col 700 700 700 693 0.081 2832.877 7.681e-03
0 Col 800 800 800 793 0.116 2938.840 7.091e-03
0 Col 900 900 900 893 0.161 3014.142 6.856e-03
0 Col 1000 1000 1000 995 0.221 3019.330 7.097e-03
10 cases ran, 10 cases passed
\end{verbatim}
\end{footnotesize}
So, we appear to be good, and the install is complete! Now we point our
users to the installed libs, and wait for the error reports to roll in.
\section{Special Instructions for some platforms}
\subsection{Special Instructions for Windows users}
\label{sec-windows}
ATLAS presently requires cygwin in order to install under Windows. Cygwin
provides a Unix-style shell environment (including standard utilities such as
{\tt gcc} and {\tt make}) for Windows.
Cygwin is free, and can be downloaded from {\tt www.cygwin.com}.
We presently do
not support Interix (AKA Windows Services for Unix, etc.) as provided
by Microsoft, but a user has submitted code to help with this, and so
we hope to add support in the future. We have had requests
to support MinGW (\verb|http://www.mingw.org/|), but no one has submitted
suggested code to help, and I have never successfully figured out how to
install and use it, so this is probably not coming soon unless something
changes.
Once cygwin is installed, you are ready to install ATLAS. If you want
to call ATLAS from code using {\tt gcc} and {\tt gfortran}, then you
can just install as usual.
If you want to call ATLAS from code compiled by native compilers such
as the Intel or Microsoft compilers, you must set up some environment
variables so that these compilers can be called from cygwin's shell.
Details on how do do this are available in the ATLAS errata file:
\vspace*{-0.1in}
\begin{verbatim}
http://math-atlas.sourceforge.net/errata.html#WinComp
\end{verbatim}
If you want multithreaded (eg., shared-memory parallel) ATLAS libraries,
you must use {\tt gcc} to compile the main library, and if you use a native
compiler for interface compilation, manually link to the cygwin library.
This is because
ATLAS uses the POSIX threading standard, which of course Microsoft
does not support, and so you need the cygwin emulation layer to use
a decade-old standard.
\subsection{Special instructions for AIX}
Under AIX, it is critical that you define an envirnment variable
indicating whether you are building 64 or 32 bit libraries, and
this definition must match what you pass to {\tt configure} via the
{\tt -b} flag. You need to define the environment variable
{\tt OBJECT\_MODE} to either 64 or 32, depending on which of
these you pass to {\tt configure} using the {\tt -b} flag. So, if you
are building 64-bit libraries and you use a {\tt bash} derivative
shell, you would issue {\tt export OBJECT\_MODE=64} before starting
the ATLAS configure step. On the other hand, if you use a {\tt csh}
derivative shell and want to build 32 but libaries, you would need
to issue {\tt setenv OBJECT\_MODE 32} before the build step.
\subsection{Special instructions for SunOS}
Solaris has its own version of the Unix utilities, which differ sharply
from the more common gnu tools. In particular, SunOS offers two fgreps,
one of which works correctly for ATLAS's {\tt make check} step, and
one of which does not. On my SunOS machine, I had to make sure
{\tt /usr/xpg4/bin} was in my path before {\tt /bin} in order
to get an {\tt fgrep} that can take multiple expression arguments
(as {\tt make check} requires).
Also, if gcc isn't compiled with with the correct gnu utilities, ATLAS
may fail to autodetect the assembly dialect of your machine. This
will cause the build to fail since it can't assemble the UltraSPARC
assembly kernels, and you can see if it happened by examining your
{\tt Make.inc}'s ARCHDEF macro. If this macro does not include
the definition {\tt -DATL\_GAS\_SPARC}, then this has happened to you.
On some systems, you can get the install to work by adding the flag
{\tt -s 3} to your {\tt configure} invocation. If this still doesn't
fix the problem, you'll need to get a better gcc install. Note that
this error causes linking to assembled files to die with messages like:
\vspace{-0.05in}
\begin{verbatim}
ld: fatal: relocation error: R_SPARC_32: file /var/tmp//ccccPppx.o:
symbol <unknown>: offset 0xff061776 is non-aligned
\end{verbatim}
\section{Troubleshooting}
The first thing you need to do is scope the errata file to see if your
problem is already covered:
\vspace*{-0.1in}
\begin{verbatim}
http://math-atlas.sourceforge.net/errata.html
\end{verbatim}
Probably the most common error is when ATLAS dies because its timings are
varying widely. This can often be fixed with a simple restart, as described:
\vspace*{-0.1in}
\begin{verbatim}
http://math-atlas.sourceforge.net/errata.html#tol
\end{verbatim}
If you are unable to find anything relevant in the errata file, you can
submit a support request to the ATLAS support tracker ({\bf not} the
bug tracker, which is for developer-confirmed bugs only):
\vspace*{-0.1in}
\begin{verbatim}
https://sourceforge.net/tracker/?atid=379483&group_id=23725&func=browse
\end{verbatim}
When you create the support request, be sure to attach the error report.
It should appear as \verb|BLDdir/error_<arch>.tgz|. If this file doesn't
exist, you can create it by typing \verb|make error_report| in your
{\tt BLDdir}. More details on submitting support requests can be found
in the ATLAS FAQ at:
\vspace*{-0.1in}
\begin{verbatim}
http://math-atlas.sourceforge.net/faq.html#help
\end{verbatim}
\bibliographystyle{plain}
\bibliography{atlas}
\end{document}
|