You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.

Makefile.system 34 kB

7 years ago
13 years ago
4 years ago
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942943944945946947948949950951952953954955956957958959960961962963964965966967968969970971972973974975976977978979980981982983984985986987988989990991992993994995996997998999100010011002100310041005100610071008100910101011101210131014101510161017101810191020102110221023102410251026102710281029103010311032103310341035103610371038103910401041104210431044104510461047104810491050105110521053105410551056105710581059106010611062106310641065106610671068106910701071107210731074107510761077107810791080108110821083108410851086108710881089109010911092109310941095109610971098109911001101110211031104110511061107110811091110111111121113111411151116111711181119112011211122112311241125112611271128112911301131113211331134113511361137113811391140114111421143114411451146114711481149115011511152115311541155115611571158115911601161116211631164116511661167116811691170117111721173117411751176117711781179118011811182118311841185118611871188118911901191119211931194119511961197119811991200120112021203120412051206120712081209121012111212121312141215121612171218121912201221122212231224122512261227122812291230123112321233123412351236123712381239124012411242124312441245124612471248124912501251125212531254125512561257125812591260126112621263126412651266126712681269127012711272127312741275127612771278127912801281128212831284128512861287128812891290129112921293129412951296129712981299130013011302130313041305130613071308130913101311131213131314131513161317131813191320132113221323132413251326132713281329133013311332133313341335133613371338133913401341134213431344134513461347134813491350135113521353135413551356135713581359136013611362136313641365136613671368136913701371137213731374137513761377137813791380138113821383138413851386138713881389139013911392139313941395139613971398139914001401140214031404140514061407140814091410141114121413141414151416141714181419142014211422142314241425142614271428142914301431143214331434143514361437143814391440144114421443144414451446144714481449145014511452145314541455145614571458145914601461146214631464146514661467146814691470147114721473147414751476147714781479148014811482148314841485148614871488148914901491149214931494149514961497149814991500150115021503150415051506150715081509151015111512151315141515151615171518151915201521152215231524152515261527152815291530153115321533153415351536153715381539154015411542154315441545154615471548154915501551155215531554155515561557155815591560156115621563156415651566156715681569157015711572157315741575157615771578157915801581158215831584158515861587158815891590159115921593159415951596159715981599160016011602160316041605160616071608160916101611161216131614161516161617161816191620162116221623162416251626162716281629163016311632163316341635163616371638163916401641164216431644164516461647164816491650165116521653165416551656165716581659166016611662166316641665166616671668166916701671167216731674167516761677167816791680168116821683168416851686168716881689
  1. #
  2. # Include user definition
  3. #
  4. # TO suppress recursive includes
  5. INCLUDED = 1
  6. ifndef TOPDIR
  7. TOPDIR = .
  8. endif
  9. # If ARCH is not set, we use the host system's architecture for getarch compile options.
  10. ifndef ARCH
  11. HOSTARCH := $(shell uname -m)
  12. else
  13. HOSTARCH = $(ARCH)
  14. endif
  15. # Catch conflicting usage of ARCH in some BSD environments
  16. ifeq ($(ARCH), amd64)
  17. override ARCH=x86_64
  18. else ifeq ($(ARCH), powerpc64)
  19. override ARCH=power
  20. else ifeq ($(ARCH), powerpc64le)
  21. override ARCH=power
  22. else ifeq ($(ARCH), powerpc)
  23. override ARCH=power
  24. else ifeq ($(ARCH), i386)
  25. override ARCH=x86
  26. else ifeq ($(ARCH), armv6)
  27. override ARCH=arm
  28. else ifeq ($(ARCH), armv7)
  29. override ARCH=arm
  30. else ifeq ($(ARCH), aarch64)
  31. override ARCH=arm64
  32. else ifeq ($(ARCH), zarch)
  33. override ARCH=zarch
  34. endif
  35. NETLIB_LAPACK_DIR = $(TOPDIR)/lapack-netlib
  36. # Default C compiler
  37. # - Only set if not specified on the command line or inherited from the environment.
  38. # - CC is an implicit variable so neither '?=' or 'ifndef' can be used.
  39. # http://stackoverflow.com/questions/4029274/mingw-and-make-variables
  40. # - Default value is 'cc' which is not always a valid command (e.g. MinGW).
  41. ifeq ($(origin CC),default)
  42. # Check if $(CC) refers to a valid command and set the value to gcc if not
  43. ifneq ($(findstring cmd.exe,$(SHELL)),)
  44. ifeq ($(shell where $(CC) 2>NUL),)
  45. CC = gcc
  46. endif
  47. else # POSIX-ish
  48. ifeq ($(shell command -v $(CC) 2>/dev/null),)
  49. ifeq ($(shell uname -s),Darwin)
  50. CC = clang
  51. # EXTRALIB += -Wl,-no_compact_unwind
  52. else
  53. CC = gcc
  54. endif # Darwin
  55. endif # CC exists
  56. endif # Shell is sane
  57. endif # CC is set to default
  58. # Default Fortran compiler (FC) is selected by f_check.
  59. ifndef MAKEFILE_RULE
  60. include $(TOPDIR)/Makefile.rule
  61. else
  62. include $(TOPDIR)/$(MAKEFILE_RULE)
  63. endif
  64. #
  65. # Beginning of system configuration
  66. #
  67. ifneq ($(BUILD_SINGLE),1)
  68. ifneq ($(BUILD_DOUBLE),1)
  69. ifneq ($(BUILD_COMPLEX),1)
  70. ifneq ($(BUILD_COMPLEX16),1)
  71. override BUILD_SINGLE=1
  72. override BUILD_DOUBLE=1
  73. override BUILD_COMPLEX=1
  74. override BUILD_COMPLEX16=1
  75. endif
  76. endif
  77. endif
  78. endif
  79. ifndef HOSTCC
  80. HOSTCC = $(CC)
  81. endif
  82. ifdef TARGET
  83. GETARCH_FLAGS := -DFORCE_$(TARGET)
  84. GETARCH_FLAGS += -DUSER_TARGET
  85. ifeq ($(TARGET), GENERIC)
  86. ifeq ($(DYNAMIC_ARCH), 1)
  87. override NO_EXPRECISION=1
  88. export NO_EXPRECiSION
  89. endif
  90. endif
  91. endif
  92. # Force fallbacks for 32bit
  93. ifeq ($(BINARY), 32)
  94. ifeq ($(TARGET), HASWELL)
  95. GETARCH_FLAGS := -DFORCE_NEHALEM
  96. endif
  97. ifeq ($(TARGET), SKYLAKEX)
  98. GETARCH_FLAGS := -DFORCE_NEHALEM
  99. endif
  100. ifeq ($(TARGET), COOPERLAKE)
  101. GETARCH_FLAGS := -DFORCE_NEHALEM
  102. endif
  103. ifeq ($(TARGET), SANDYBRIDGE)
  104. GETARCH_FLAGS := -DFORCE_NEHALEM
  105. endif
  106. ifeq ($(TARGET), BULLDOZER)
  107. GETARCH_FLAGS := -DFORCE_BARCELONA
  108. endif
  109. ifeq ($(TARGET), PILEDRIVER)
  110. GETARCH_FLAGS := -DFORCE_BARCELONA
  111. endif
  112. ifeq ($(TARGET), STEAMROLLER)
  113. GETARCH_FLAGS := -DFORCE_BARCELONA
  114. endif
  115. ifeq ($(TARGET), EXCAVATOR)
  116. GETARCH_FLAGS := -DFORCE_BARCELONA
  117. endif
  118. ifeq ($(TARGET), ZEN)
  119. GETARCH_FLAGS := -DFORCE_BARCELONA
  120. endif
  121. ifeq ($(TARGET), ARMV8)
  122. GETARCH_FLAGS := -DFORCE_ARMV7
  123. endif
  124. ifeq ($(TARGET), POWER8)
  125. GETARCH_FLAGS := -DFORCE_POWER6
  126. endif
  127. endif
  128. #TARGET_CORE will override TARGET which is used in DYNAMIC_ARCH=1.
  129. #
  130. ifdef TARGET_CORE
  131. GETARCH_FLAGS := -DFORCE_$(TARGET_CORE)
  132. endif
  133. # Force fallbacks for 32bit
  134. ifeq ($(BINARY), 32)
  135. ifeq ($(TARGET_CORE), HASWELL)
  136. GETARCH_FLAGS := -DFORCE_NEHALEM
  137. endif
  138. ifeq ($(TARGET_CORE), SKYLAKEX)
  139. GETARCH_FLAGS := -DFORCE_NEHALEM
  140. endif
  141. ifeq ($(TARGET_CORE), COOPERLAKE)
  142. GETARCH_FLAGS := -DFORCE_NEHALEM
  143. endif
  144. ifeq ($(TARGET_CORE), SANDYBRIDGE)
  145. GETARCH_FLAGS := -DFORCE_NEHALEM
  146. endif
  147. ifeq ($(TARGET_CORE), BULLDOZER)
  148. GETARCH_FLAGS := -DFORCE_BARCELONA
  149. endif
  150. ifeq ($(TARGET_CORE), PILEDRIVER)
  151. GETARCH_FLAGS := -DFORCE_BARCELONA
  152. endif
  153. ifeq ($(TARGET_CORE), STEAMROLLER)
  154. GETARCH_FLAGS := -DFORCE_BARCELONA
  155. endif
  156. ifeq ($(TARGET_CORE), EXCAVATOR)
  157. GETARCH_FLAGS := -DFORCE_BARCELONA
  158. endif
  159. ifeq ($(TARGET_CORE), ZEN)
  160. GETARCH_FLAGS := -DFORCE_BARCELONA
  161. endif
  162. endif
  163. # On x86_64 build getarch with march=native unless the compiler is PGI. This is required to detect AVX512 support in getarch.
  164. ifeq ($(HOSTARCH), x86_64)
  165. ifeq ($(findstring pgcc,$(HOSTCC))$(findstring nvc,$(HOSTCC)),)
  166. GETARCH_FLAGS += -march=native
  167. endif
  168. endif
  169. ifdef INTERFACE64
  170. ifneq ($(INTERFACE64), 0)
  171. GETARCH_FLAGS += -DUSE64BITINT
  172. endif
  173. endif
  174. ifndef GEMM_MULTITHREAD_THRESHOLD
  175. GEMM_MULTITHREAD_THRESHOLD=4
  176. endif
  177. GETARCH_FLAGS += -DGEMM_MULTITHREAD_THRESHOLD=$(GEMM_MULTITHREAD_THRESHOLD)
  178. ifeq ($(NO_AVX), 1)
  179. GETARCH_FLAGS += -DNO_AVX
  180. endif
  181. ifeq ($(BINARY), 32)
  182. GETARCH_FLAGS += -DNO_AVX -DNO_AVX2 -DNO_AVX512
  183. NO_AVX512 = 1
  184. endif
  185. ifeq ($(NO_AVX2), 1)
  186. GETARCH_FLAGS += -DNO_AVX2
  187. endif
  188. ifeq ($(NO_AVX512), 1)
  189. GETARCH_FLAGS += -DNO_AVX512
  190. endif
  191. ifeq ($(DEBUG), 1)
  192. GETARCH_FLAGS += -g
  193. endif
  194. ifeq ($(QUIET_MAKE), 1)
  195. MAKE += -s
  196. endif
  197. ifndef NO_PARALLEL_MAKE
  198. NO_PARALLEL_MAKE=0
  199. endif
  200. GETARCH_FLAGS += -DNO_PARALLEL_MAKE=$(NO_PARALLEL_MAKE)
  201. ifdef MAKE_NB_JOBS
  202. GETARCH_FLAGS += -DMAKE_NB_JOBS=$(MAKE_NB_JOBS)
  203. endif
  204. ifeq ($(HOSTCC), loongcc)
  205. GETARCH_FLAGS += -static
  206. endif
  207. #if don't use Fortran, it will only compile CBLAS.
  208. ifeq ($(ONLY_CBLAS), 1)
  209. NO_LAPACK = 1
  210. else
  211. ONLY_CBLAS = 0
  212. endif
  213. # This operation is expensive, so execution should be once.
  214. ifndef GOTOBLAS_MAKEFILE
  215. export GOTOBLAS_MAKEFILE = 1
  216. # Generating Makefile.conf and config.h
  217. DUMMY := $(shell $(MAKE) -C $(TOPDIR) -f Makefile.prebuild CC="$(CC)" FC="$(FC)" HOSTCC="$(HOSTCC)" HOST_CFLAGS="$(GETARCH_FLAGS)" CFLAGS="$(CFLAGS)" BINARY=$(BINARY) USE_OPENMP=$(USE_OPENMP) TARGET_CORE=$(TARGET_CORE) ONLY_CBLAS=$(ONLY_CBLAS) TARGET=$(TARGET) all)
  218. ifndef TARGET_CORE
  219. include $(TOPDIR)/Makefile.conf
  220. else
  221. HAVE_NEON=
  222. HAVE_VFP=
  223. HAVE_VFPV3=
  224. HAVE_VFPV4=
  225. HAVE_MMX=
  226. HAVE_SSE=
  227. HAVE_SSE2=
  228. HAVE_SSE3=
  229. HAVE_SSSE3=
  230. HAVE_SSE4_1=
  231. HAVE_SSE4_2=
  232. HAVE_SSE4A=
  233. HAVE_SSE5=
  234. HAVE_AVX=
  235. HAVE_AVX2=
  236. HAVE_FMA3=
  237. include $(TOPDIR)/Makefile_kernel.conf
  238. endif
  239. endif
  240. ifndef NUM_PARALLEL
  241. NUM_PARALLEL = 1
  242. endif
  243. ifndef NUM_THREADS
  244. NUM_THREADS = $(NUM_CORES)
  245. endif
  246. ifeq ($(NUM_THREADS), 1)
  247. override USE_THREAD = 0
  248. override USE_OPENMP = 0
  249. endif
  250. ifdef USE_THREAD
  251. ifeq ($(USE_THREAD), 0)
  252. SMP =
  253. else
  254. SMP = 1
  255. endif
  256. else
  257. ifeq ($(NUM_THREAD), 1)
  258. SMP =
  259. else
  260. SMP = 1
  261. endif
  262. endif
  263. ifeq ($(SMP), 1)
  264. USE_LOCKING =
  265. endif
  266. ifndef NEED_PIC
  267. NEED_PIC = 1
  268. endif
  269. ARFLAGS =
  270. CPP = $(COMPILER) -E
  271. AR ?= $(CROSS_SUFFIX)ar
  272. AS ?= $(CROSS_SUFFIX)as
  273. LD ?= $(CROSS_SUFFIX)ld
  274. RANLIB ?= $(CROSS_SUFFIX)ranlib
  275. NM = $(CROSS_SUFFIX)nm
  276. DLLWRAP = $(CROSS_SUFFIX)dllwrap
  277. OBJCOPY = $(CROSS_SUFFIX)objcopy
  278. OBJCONV = $(CROSS_SUFFIX)objconv
  279. # When fortran support was either not detected or actively deselected, only build BLAS.
  280. ifeq ($(NOFORTRAN), 1)
  281. NO_LAPACK = 1
  282. override FEXTRALIB =
  283. endif
  284. ifeq ($(C_COMPILER), GCC)
  285. GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
  286. GCCVERSIONGT4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 4)
  287. GCCVERSIONGT5 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 5)
  288. GCCVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 7)
  289. GCCVERSIONGTEQ8 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 8)
  290. GCCVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  291. GCCVERSIONGTEQ11 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 11)
  292. GCCVERSIONGTEQ10 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 10)
  293. # Note that the behavior of -dumpversion is compile-time-configurable for
  294. # gcc-7.x and newer. Use -dumpfullversion there
  295. ifeq ($(GCCVERSIONGTEQ7),1)
  296. GCCDUMPVERSION_PARAM := -dumpfullversion
  297. else
  298. GCCDUMPVERSION_PARAM := -dumpversion
  299. endif
  300. GCCMINORVERSIONGTEQ1 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 1)
  301. GCCMINORVERSIONGTEQ2 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 2)
  302. GCCMINORVERSIONGTEQ7 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 7)
  303. endif
  304. #
  305. # OS dependent settings
  306. #
  307. ifeq ($(OSNAME), Darwin)
  308. ifndef MACOSX_DEPLOYMENT_TARGET
  309. export MACOSX_DEPLOYMENT_TARGET=10.8
  310. endif
  311. MD5SUM = md5 -r
  312. endif
  313. ifneq (,$(findstring $(OSNAME), FreeBSD OpenBSD DragonFly))
  314. MD5SUM = md5 -r
  315. endif
  316. ifeq ($(OSNAME), NetBSD)
  317. MD5SUM = md5 -n
  318. endif
  319. ifeq ($(OSNAME), Linux)
  320. EXTRALIB += -lm
  321. NO_EXPRECISION = 1
  322. endif
  323. ifeq ($(OSNAME), Android)
  324. EXTRALIB += -lm
  325. endif
  326. ifeq ($(OSNAME), AIX)
  327. EXTRALIB += -lm
  328. endif
  329. ifeq ($(OSNAME), FreeBSD)
  330. ifeq ($(ARCH), $(filter $(ARCH),arm arm64))
  331. EXTRALIB += -lm
  332. endif
  333. endif
  334. ifeq ($(OSNAME), WINNT)
  335. NEED_PIC = 0
  336. NO_EXPRECISION = 1
  337. EXTRALIB += -defaultlib:advapi32
  338. SUFFIX = obj
  339. PSUFFIX = pobj
  340. LIBSUFFIX = a
  341. ifeq ($(C_COMPILER), CLANG)
  342. CCOMMON_OPT += -DMS_ABI
  343. endif
  344. #Version tests for supporting specific features (MS_ABI, POWER9 intrinsics)
  345. ifeq ($(GCCVERSIONGT4), 1)
  346. # GCC Major version > 4
  347. # It is compatible with MSVC ABI.
  348. CCOMMON_OPT += -DMS_ABI
  349. endif
  350. ifeq ($(GCCVERSIONGTEQ4), 1)
  351. ifeq ($(GCCMINORVERSIONGTEQ7), 1)
  352. # GCC Version >=4.7
  353. # It is compatible with MSVC ABI.
  354. CCOMMON_OPT += -DMS_ABI
  355. endif
  356. endif
  357. # Ensure the correct stack alignment on Win32
  358. # http://permalink.gmane.org/gmane.comp.lib.openblas.general/97
  359. ifeq ($(ARCH), x86)
  360. CCOMMON_OPT += -mincoming-stack-boundary=2
  361. FCOMMON_OPT += -mincoming-stack-boundary=2
  362. endif
  363. endif
  364. ifeq ($(OSNAME), Interix)
  365. NEED_PIC = 0
  366. NO_EXPRECISION = 1
  367. INTERIX_TOOL_DIR = /opt/gcc.3.3/i586-pc-interix3/bin
  368. endif
  369. ifeq ($(OSNAME), CYGWIN_NT)
  370. NEED_PIC = 0
  371. NO_EXPRECISION = 1
  372. OS_CYGWIN_NT = 1
  373. endif
  374. ifneq ($(OSNAME), WINNT)
  375. ifneq ($(OSNAME), CYGWIN_NT)
  376. ifneq ($(OSNAME), Interix)
  377. ifneq ($(OSNAME), Android)
  378. ifdef SMP
  379. EXTRALIB += -lpthread
  380. endif
  381. endif
  382. endif
  383. endif
  384. endif
  385. # ifeq logical or
  386. ifeq ($(OSNAME), $(filter $(OSNAME),WINNT CYGWIN_NT Interix))
  387. OS_WINDOWS=1
  388. endif
  389. ifdef QUAD_PRECISION
  390. CCOMMON_OPT += -DQUAD_PRECISION
  391. NO_EXPRECISION = 1
  392. endif
  393. ifneq ($(ARCH), x86)
  394. ifneq ($(ARCH), x86_64)
  395. NO_EXPRECISION = 1
  396. endif
  397. endif
  398. ifdef UTEST_CHECK
  399. CCOMMON_OPT += -DUTEST_CHECK
  400. SANITY_CHECK = 1
  401. endif
  402. ifdef SANITY_CHECK
  403. CCOMMON_OPT += -DSANITY_CHECK -DREFNAME=$(*F)f$(BU)
  404. endif
  405. MAX_STACK_ALLOC ?= 2048
  406. ifneq ($(MAX_STACK_ALLOC), 0)
  407. CCOMMON_OPT += -DMAX_STACK_ALLOC=$(MAX_STACK_ALLOC)
  408. endif
  409. ifdef USE_LOCKING
  410. ifneq ($(USE_LOCKING), 0)
  411. CCOMMON_OPT += -DUSE_LOCKING
  412. endif
  413. endif
  414. #
  415. # Architecture dependent settings
  416. #
  417. ifeq ($(ARCH), x86)
  418. ifndef BINARY
  419. NO_BINARY_MODE = 1
  420. endif
  421. ifeq ($(CORE), generic)
  422. NO_EXPRECISION = 1
  423. endif
  424. ifndef NO_EXPRECISION
  425. ifeq ($(F_COMPILER), GFORTRAN)
  426. # ifeq logical or. GCC or LSB
  427. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  428. EXPRECISION = 1
  429. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  430. FCOMMON_OPT += -m128bit-long-double
  431. endif
  432. ifeq ($(C_COMPILER), CLANG)
  433. EXPRECISION = 1
  434. CCOMMON_OPT += -DEXPRECISION
  435. FCOMMON_OPT += -m128bit-long-double
  436. endif
  437. endif
  438. endif
  439. endif
  440. ifeq ($(ARCH), x86_64)
  441. ifeq ($(CORE), generic)
  442. NO_EXPRECISION = 1
  443. endif
  444. ifndef NO_EXPRECISION
  445. ifeq ($(F_COMPILER), GFORTRAN)
  446. # ifeq logical or. GCC or LSB
  447. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  448. EXPRECISION = 1
  449. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  450. FCOMMON_OPT += -m128bit-long-double
  451. endif
  452. ifeq ($(C_COMPILER), CLANG)
  453. EXPRECISION = 1
  454. CCOMMON_OPT += -DEXPRECISION
  455. FCOMMON_OPT += -m128bit-long-double
  456. endif
  457. endif
  458. endif
  459. endif
  460. ifeq ($(C_COMPILER), INTEL)
  461. CCOMMON_OPT += -wd981
  462. endif
  463. ifeq ($(USE_OPENMP), 1)
  464. #check
  465. ifeq ($(USE_THREAD), 0)
  466. $(error OpenBLAS: Cannot set both USE_OPENMP=1 and USE_THREAD=0. The USE_THREAD=0 is only for building single thread version.)
  467. endif
  468. # ifeq logical or. GCC or LSB
  469. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  470. CCOMMON_OPT += -fopenmp
  471. endif
  472. ifeq ($(C_COMPILER), CLANG)
  473. CCOMMON_OPT += -fopenmp
  474. endif
  475. ifeq ($(C_COMPILER), INTEL)
  476. CCOMMON_OPT += -fopenmp
  477. endif
  478. ifeq ($(C_COMPILER), PGI)
  479. CCOMMON_OPT += -mp
  480. endif
  481. ifeq ($(C_COMPILER), OPEN64)
  482. CCOMMON_OPT += -mp
  483. CEXTRALIB += -lstdc++
  484. endif
  485. ifeq ($(C_COMPILER), PATHSCALE)
  486. CCOMMON_OPT += -mp
  487. endif
  488. endif
  489. ifeq ($(DYNAMIC_ARCH), 1)
  490. ifeq ($(ARCH), x86)
  491. DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
  492. CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
  493. endif
  494. ifeq ($(ARCH), x86_64)
  495. DYNAMIC_CORE = PRESCOTT CORE2
  496. ifeq ($(DYNAMIC_OLDER), 1)
  497. DYNAMIC_CORE += PENRYN DUNNINGTON
  498. endif
  499. DYNAMIC_CORE += NEHALEM
  500. ifeq ($(DYNAMIC_OLDER), 1)
  501. DYNAMIC_CORE += OPTERON OPTERON_SSE3
  502. endif
  503. DYNAMIC_CORE += BARCELONA
  504. ifeq ($(DYNAMIC_OLDER), 1)
  505. DYNAMIC_CORE += BOBCAT ATOM NANO
  506. endif
  507. ifneq ($(NO_AVX), 1)
  508. DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER STEAMROLLER EXCAVATOR
  509. endif
  510. ifneq ($(NO_AVX2), 1)
  511. DYNAMIC_CORE += HASWELL ZEN
  512. endif
  513. ifneq ($(NO_AVX512), 1)
  514. ifneq ($(NO_AVX2), 1)
  515. DYNAMIC_CORE += SKYLAKEX COOPERLAKE
  516. endif
  517. endif
  518. endif
  519. ifdef DYNAMIC_LIST
  520. override DYNAMIC_CORE = PRESCOTT $(DYNAMIC_LIST)
  521. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_PRESCOTT
  522. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  523. CCOMMON_OPT += $(XCCOMMON_OPT)
  524. #CCOMMON_OPT += -DDYNAMIC_LIST='$(DYNAMIC_LIST)'
  525. endif
  526. ifeq ($(ARCH), arm64)
  527. DYNAMIC_CORE = ARMV8
  528. DYNAMIC_CORE += CORTEXA53
  529. DYNAMIC_CORE += CORTEXA57
  530. DYNAMIC_CORE += CORTEXA72
  531. DYNAMIC_CORE += CORTEXA73
  532. DYNAMIC_CORE += NEOVERSEN1
  533. DYNAMIC_CORE += CORTEXA55
  534. DYNAMIC_CORE += FALKOR
  535. DYNAMIC_CORE += THUNDERX
  536. DYNAMIC_CORE += THUNDERX2T99
  537. DYNAMIC_CORE += TSV110
  538. DYNAMIC_CORE += EMAG8180
  539. DYNAMIC_CORE += THUNDERX3T110
  540. ifdef DYNAMIC_LIST
  541. override DYNAMIC_CORE = ARMV8 $(DYNAMIC_LIST)
  542. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_ARMV8
  543. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  544. endif
  545. endif
  546. ifeq ($(ARCH), mips64)
  547. DYNAMIC_CORE = LOONGSON3R3 LOONGSON3R4
  548. endif
  549. ifeq ($(ARCH), zarch)
  550. DYNAMIC_CORE = ZARCH_GENERIC
  551. # if the compiler accepts -march=arch11 or -march=z13 and can compile a file
  552. # with z13-specific inline assembly, then we can include support for Z13.
  553. # note: -march=z13 is equivalent to -march=arch11 yet some compiler releases
  554. # only support one or the other.
  555. # note: LLVM version 6.x supported -march=z13 yet could not handle vector
  556. # registers in inline assembly, so the check for supporting the -march flag is
  557. # not enough.
  558. ZARCH_TEST_COMPILE=-c $(TOPDIR)/kernel/zarch/damin_z13.c -I$(TOPDIR) -o /dev/null > /dev/null 2> /dev/null
  559. ZARCH_CC_SUPPORTS_ARCH11=$(shell $(CC) -march=arch11 $(ZARCH_TEST_COMPILE) && echo 1)
  560. ZARCH_CC_SUPPORTS_Z13=$(shell $(CC) -march=z13 $(ZARCH_TEST_COMPILE) && echo 1)
  561. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH11), $(ZARCH_CC_SUPPORTS_Z13)), 1)
  562. DYNAMIC_CORE += Z13
  563. CCOMMON_OPT += -DDYN_Z13
  564. else
  565. $(info OpenBLAS: Not building Z13 kernels because the compiler $(CC) does not support it)
  566. endif
  567. # as above for z13, check for -march=arch12 and z14 support in the compiler.
  568. ZARCH_CC_SUPPORTS_ARCH12=$(shell $(CC) -march=arch12 $(ZARCH_TEST_COMPILE) && echo 1)
  569. ZARCH_CC_SUPPORTS_Z14=$(shell $(CC) -march=z14 $(ZARCH_TEST_COMPILE) && echo 1)
  570. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH12), $(ZARCH_CC_SUPPORTS_Z14)), 1)
  571. DYNAMIC_CORE += Z14
  572. CCOMMON_OPT += -DDYN_Z14
  573. else
  574. $(info OpenBLAS: Not building Z14 kernels because the compiler $(CC) does not support it)
  575. endif
  576. endif # ARCH zarch
  577. ifeq ($(ARCH), power)
  578. ifneq ($(C_COMPILER), PGI)
  579. DYNAMIC_CORE = POWER6
  580. DYNAMIC_CORE += POWER8
  581. ifneq ($(C_COMPILER), GCC)
  582. DYNAMIC_CORE += POWER9
  583. DYNAMIC_CORE += POWER10
  584. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  585. endif
  586. ifeq ($(C_COMPILER), GCC)
  587. ifeq ($(GCCVERSIONGT5), 1)
  588. DYNAMIC_CORE += POWER9
  589. else
  590. $(info, OpenBLAS: Your gcc version is too old to build the POWER9 kernels.)
  591. endif
  592. LDVERSIONGTEQ35 := $(shell expr `$(CC) -Wl,--version 2> /dev/null | head -1 | cut -f2 -d "." | cut -f1 -d "-"` \>= 35)
  593. ifeq ($(GCCVERSIONGTEQ11)$(LDVERSIONGTEQ35), 11)
  594. DYNAMIC_CORE += POWER10
  595. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  596. else ifeq ($(GCCVERSIONGTEQ10), 1)
  597. ifeq ($(GCCMINORVERSIONGTEQ2)$(LDVERSIONGTEQ35), 11)
  598. DYNAMIC_CORE += POWER10
  599. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  600. endif
  601. else
  602. $(info, OpenBLAS: Your gcc version is too old to build the POWER10 kernels.)
  603. endif
  604. endif
  605. else
  606. DYNAMIC_CORE = POWER8
  607. DYNAMIC_CORE += POWER9
  608. endif
  609. endif
  610. # If DYNAMIC_CORE is not set, DYNAMIC_ARCH cannot do anything, so force it to empty
  611. ifndef DYNAMIC_CORE
  612. override DYNAMIC_ARCH=
  613. endif
  614. endif
  615. ifeq ($(ARCH), ia64)
  616. NO_BINARY_MODE = 1
  617. BINARY_DEFINED = 1
  618. ifeq ($(F_COMPILER), GFORTRAN)
  619. ifeq ($(C_COMPILER), GCC)
  620. # EXPRECISION = 1
  621. # CCOMMON_OPT += -DEXPRECISION
  622. endif
  623. endif
  624. endif
  625. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  626. NO_BINARY_MODE = 1
  627. endif
  628. ifeq ($(ARCH), alpha)
  629. NO_BINARY_MODE = 1
  630. BINARY_DEFINED = 1
  631. endif
  632. ifeq ($(ARCH), arm)
  633. NO_BINARY_MODE = 1
  634. BINARY_DEFINED = 1
  635. CCOMMON_OPT += -marm
  636. FCOMMON_OPT += -marm
  637. # If softfp abi is mentioned on the command line, force it.
  638. ifeq ($(ARM_SOFTFP_ABI), 1)
  639. CCOMMON_OPT += -mfloat-abi=softfp
  640. FCOMMON_OPT += -mfloat-abi=softfp
  641. endif
  642. ifeq ($(OSNAME), Android)
  643. ifeq ($(ARM_SOFTFP_ABI), 1)
  644. EXTRALIB += -lm
  645. else
  646. EXTRALIB += -Wl,-lm_hard
  647. endif
  648. endif
  649. endif
  650. ifeq ($(ARCH), arm64)
  651. NO_BINARY_MODE = 1
  652. BINARY_DEFINED = 1
  653. ifdef INTERFACE64
  654. ifneq ($(INTERFACE64), 0)
  655. ifeq ($(F_COMPILER), GFORTRAN)
  656. FCOMMON_OPT += -fdefault-integer-8
  657. endif
  658. ifeq ($(F_COMPILER), FLANG)
  659. FCOMMON_OPT += -i8
  660. endif
  661. endif
  662. endif
  663. endif
  664. ifeq ($(ARCH), riscv64)
  665. NO_BINARY_MODE = 1
  666. BINARY_DEFINED = 1
  667. endif
  668. #
  669. # C Compiler dependent settings
  670. #
  671. # ifeq logical or. GCC or CLANG or LSB
  672. # http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
  673. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG LSB))
  674. CCOMMON_OPT += -Wall
  675. COMMON_PROF += -fno-inline
  676. NO_UNINITIALIZED_WARN = -Wno-uninitialized
  677. ifeq ($(QUIET_MAKE), 1)
  678. CCOMMON_OPT += $(NO_UNINITIALIZED_WARN) -Wno-unused
  679. endif
  680. ifdef NO_BINARY_MODE
  681. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  682. ifdef BINARY64
  683. CCOMMON_OPT += -mabi=64
  684. else
  685. CCOMMON_OPT += -mabi=n32
  686. endif
  687. BINARY_DEFINED = 1
  688. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  689. CCOMMON_OPT += -mabi=32
  690. BINARY_DEFINED = 1
  691. endif
  692. ifeq ($(CORE), $(filter $(CORE),LOONGSON3R3 LOONGSON3R4))
  693. CCOMMON_OPT += -march=loongson3a
  694. FCOMMON_OPT += -march=loongson3a
  695. endif
  696. ifeq ($(CORE), MIPS24K)
  697. CCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  698. FCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  699. endif
  700. ifeq ($(CORE), MIPS1004K)
  701. CCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  702. FCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  703. endif
  704. ifeq ($(CORE), P5600)
  705. CCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  706. FCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  707. endif
  708. ifeq ($(CORE), I6400)
  709. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  710. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  711. endif
  712. ifeq ($(CORE), P6600)
  713. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  714. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  715. endif
  716. ifeq ($(CORE), I6500)
  717. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  718. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  719. endif
  720. ifeq ($(OSNAME), AIX)
  721. BINARY_DEFINED = 1
  722. endif
  723. endif
  724. ifndef BINARY_DEFINED
  725. ifneq ($(OSNAME), AIX)
  726. ifdef BINARY64
  727. ifneq ($(ARCH), riscv64)
  728. CCOMMON_OPT += -m64
  729. endif
  730. else
  731. CCOMMON_OPT += -m32
  732. endif
  733. endif
  734. endif
  735. endif
  736. ifeq ($(C_COMPILER), PGI)
  737. PGCVERSIONGT20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \> 20)
  738. PGCVERSIONGTEQ20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \>= 20)
  739. PGCMINORVERSIONGE11 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -c 4-5` == 11)
  740. PGCVERSIONCHECK := $(PGCVERSIONGT20)$(PGCVERSIONEQ20)$(PGCMINORVERSIONGE11)
  741. ifeq ($(PGCVERSIONCHECK), $(filter $(PGCVERSIONCHECK), 110 111 011))
  742. NEWPGI := 1
  743. endif
  744. ifdef BINARY64
  745. ifeq ($(ARCH), x86_64)
  746. CCOMMON_OPT += -tp p7-64
  747. ifneq ($(NEWPGI),1)
  748. CCOMMON_OPT += -D__MMX__ -Mnollvm
  749. endif
  750. else
  751. ifeq ($(ARCH), power)
  752. ifeq ($(CORE), POWER8)
  753. CCOMMON_OPT += -tp pwr8
  754. endif
  755. ifeq ($(CORE), POWER9)
  756. CCOMMON_OPT += -tp pwr9
  757. endif
  758. endif
  759. endif
  760. else
  761. CCOMMON_OPT += -tp p7
  762. endif
  763. endif
  764. ifeq ($(C_COMPILER), PATHSCALE)
  765. ifdef BINARY64
  766. CCOMMON_OPT += -m64
  767. else
  768. CCOMMON_OPT += -m32
  769. endif
  770. endif
  771. #
  772. # Fortran Compiler dependent settings
  773. #
  774. ifeq ($(F_COMPILER), NAG)
  775. FCOMMON_OPT += -dcfuns -recursive -ieee=full -w=obs -thread_safe
  776. ifdef INTERFACE64
  777. ifneq ($(INTERFACE64), 0)
  778. FCOMMON_OPT += -i8
  779. endif
  780. endif
  781. ifeq ($(USE_OPENMP), 1)
  782. FCOMMON_OPT += -openmp
  783. endif
  784. endif
  785. ifeq ($(F_COMPILER), FLANG)
  786. CCOMMON_OPT += -DF_INTERFACE_FLANG
  787. FCOMMON_OPT += -Mrecursive -Kieee
  788. ifeq ($(OSNAME), Linux)
  789. ifeq ($(ARCH), x86_64)
  790. FLANG_VENDOR := $(shell $(FC) --version|head -1 |cut -f 1 -d " ")
  791. ifeq ($(FLANG_VENDOR), AMD)
  792. FCOMMON_OPT += -fno-unroll-loops
  793. endif
  794. endif
  795. endif
  796. ifdef BINARY64
  797. ifdef INTERFACE64
  798. ifneq ($(INTERFACE64), 0)
  799. FCOMMON_OPT += -i8
  800. endif
  801. endif
  802. FCOMMON_OPT += -Wall
  803. else
  804. FCOMMON_OPT += -Wall
  805. endif
  806. ifeq ($(USE_OPENMP), 1)
  807. FCOMMON_OPT += -fopenmp
  808. endif
  809. endif
  810. ifeq ($(F_COMPILER), G77)
  811. CCOMMON_OPT += -DF_INTERFACE_G77
  812. FCOMMON_OPT += -Wall
  813. ifndef NO_BINARY_MODE
  814. ifneq ($(OSNAME), AIX)
  815. ifdef BINARY64
  816. FCOMMON_OPT += -m64
  817. else
  818. FCOMMON_OPT += -m32
  819. endif
  820. endif
  821. endif
  822. endif
  823. ifeq ($(F_COMPILER), G95)
  824. CCOMMON_OPT += -DF_INTERFACE_G95
  825. FCOMMON_OPT += -Wall
  826. ifneq ($(OSNAME), AIX)
  827. ifndef NO_BINARY_MODE
  828. ifdef BINARY64
  829. FCOMMON_OPT += -m64
  830. else
  831. FCOMMON_OPT += -m32
  832. endif
  833. endif
  834. ifneq ($(NO_LAPACKE), 1)
  835. FCOMMON_OPT += -fno-second-underscore
  836. endif
  837. endif
  838. endif
  839. ifeq ($(F_COMPILER), GFORTRAN)
  840. CCOMMON_OPT += -DF_INTERFACE_GFORT
  841. FCOMMON_OPT += -Wall
  842. # make single-threaded LAPACK calls thread-safe #1847
  843. FCOMMON_OPT += -frecursive
  844. # work around ABI problem with passing single-character arguments
  845. FCOMMON_OPT += -fno-optimize-sibling-calls
  846. #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
  847. ifneq ($(NO_LAPACK), 1)
  848. EXTRALIB += -lgfortran
  849. endif
  850. ifdef NO_BINARY_MODE
  851. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  852. ifdef BINARY64
  853. FCOMMON_OPT += -mabi=64
  854. else
  855. FCOMMON_OPT += -mabi=n32
  856. endif
  857. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  858. FCOMMON_OPT += -mabi=32
  859. endif
  860. else
  861. ifdef BINARY64
  862. ifneq ($(OSNAME), AIX)
  863. ifneq ($(ARCH), riscv64)
  864. FCOMMON_OPT += -m64
  865. endif
  866. endif
  867. ifdef INTERFACE64
  868. ifneq ($(INTERFACE64), 0)
  869. FCOMMON_OPT += -fdefault-integer-8
  870. endif
  871. endif
  872. else
  873. ifneq ($(OSNAME), AIX)
  874. FCOMMON_OPT += -m32
  875. endif
  876. endif
  877. endif
  878. ifeq ($(USE_OPENMP), 1)
  879. FCOMMON_OPT += -fopenmp
  880. endif
  881. endif
  882. ifeq ($(F_COMPILER), INTEL)
  883. CCOMMON_OPT += -DF_INTERFACE_INTEL
  884. ifdef INTERFACE64
  885. ifneq ($(INTERFACE64), 0)
  886. FCOMMON_OPT += -i8
  887. endif
  888. endif
  889. FCOMMON_OPT += -recursive -fp-model strict -assume protect-parens
  890. ifeq ($(USE_OPENMP), 1)
  891. FCOMMON_OPT += -fopenmp
  892. endif
  893. endif
  894. ifeq ($(F_COMPILER), FUJITSU)
  895. CCOMMON_OPT += -DF_INTERFACE_FUJITSU
  896. ifeq ($(USE_OPENMP), 1)
  897. FCOMMON_OPT += -openmp
  898. endif
  899. endif
  900. ifeq ($(F_COMPILER), IBM)
  901. CCOMMON_OPT += -DF_INTERFACE_IBM
  902. # FCOMMON_OPT += -qarch=440
  903. ifdef BINARY64
  904. FCOMMON_OPT += -q64
  905. ifdef INTERFACE64
  906. ifneq ($(INTERFACE64), 0)
  907. FCOMMON_OPT += -qintsize=8
  908. endif
  909. endif
  910. else
  911. FCOMMON_OPT += -q32
  912. endif
  913. ifeq ($(USE_OPENMP), 1)
  914. FCOMMON_OPT += -openmp
  915. endif
  916. endif
  917. ifeq ($(F_COMPILER), PGI)
  918. CCOMMON_OPT += -DF_INTERFACE_PGI
  919. COMMON_PROF += -DPGICOMPILER
  920. ifdef BINARY64
  921. ifdef INTERFACE64
  922. ifneq ($(INTERFACE64), 0)
  923. FCOMMON_OPT += -i8
  924. endif
  925. endif
  926. ifeq ($(ARCH), x86_64)
  927. FCOMMON_OPT += -tp p7-64
  928. else
  929. ifeq ($(ARCH), power)
  930. ifeq ($(CORE), POWER6)
  931. $(warning NVIDIA HPC compilers do not support POWER6.)
  932. endif
  933. ifeq ($(CORE), POWER8)
  934. FCOMMON_OPT += -tp pwr8
  935. endif
  936. ifeq ($(CORE), POWER9)
  937. FCOMMON_OPT += -tp pwr9
  938. endif
  939. ifeq ($(CORE), POWER10)
  940. $(warning NVIDIA HPC compilers do not support POWER10.)
  941. endif
  942. endif
  943. endif
  944. else
  945. FCOMMON_OPT += -tp p7
  946. endif
  947. FCOMMON_OPT += -Mrecursive -Kieee
  948. ifeq ($(USE_OPENMP), 1)
  949. FCOMMON_OPT += -mp
  950. endif
  951. endif
  952. ifeq ($(F_COMPILER), PATHSCALE)
  953. CCOMMON_OPT += -DF_INTERFACE_PATHSCALE
  954. ifdef BINARY64
  955. ifdef INTERFACE64
  956. ifneq ($(INTERFACE64), 0)
  957. FCOMMON_OPT += -i8
  958. endif
  959. endif
  960. endif
  961. ifeq ($(USE_OPENMP), 1)
  962. FCOMMON_OPT += -mp
  963. endif
  964. endif
  965. ifeq ($(F_COMPILER), OPEN64)
  966. CCOMMON_OPT += -DF_INTERFACE_OPEN64
  967. ifdef BINARY64
  968. ifdef INTERFACE64
  969. ifneq ($(INTERFACE64), 0)
  970. FCOMMON_OPT += -i8
  971. endif
  972. endif
  973. endif
  974. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  975. ifndef BINARY64
  976. FCOMMON_OPT += -n32
  977. else
  978. FCOMMON_OPT += -n64
  979. endif
  980. ifeq ($(CORE), LOONGSON3R3)
  981. FCOMMON_OPT += -loongson3 -static
  982. endif
  983. ifeq ($(CORE), LOONGSON3R4)
  984. FCOMMON_OPT += -loongson3 -static
  985. endif
  986. else
  987. ifndef BINARY64
  988. FCOMMON_OPT += -m32
  989. else
  990. FCOMMON_OPT += -m64
  991. endif
  992. endif
  993. ifeq ($(USE_OPENMP), 1)
  994. FEXTRALIB += -lstdc++
  995. FCOMMON_OPT += -mp
  996. endif
  997. endif
  998. ifeq ($(C_COMPILER), OPEN64)
  999. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  1000. ifndef BINARY64
  1001. CCOMMON_OPT += -n32
  1002. else
  1003. CCOMMON_OPT += -n64
  1004. endif
  1005. ifeq ($(CORE), LOONGSON3R3)
  1006. CCOMMON_OPT += -loongson3 -static
  1007. endif
  1008. ifeq ($(CORE), LOONGSON3R4)
  1009. CCOMMON_OPT += -loongson3 -static
  1010. endif
  1011. else
  1012. ifndef BINARY64
  1013. CCOMMON_OPT += -m32
  1014. else
  1015. CCOMMON_OPT += -m64
  1016. endif
  1017. endif
  1018. endif
  1019. ifeq ($(C_COMPILER), SUN)
  1020. CCOMMON_OPT += -w
  1021. ifeq ($(ARCH), x86)
  1022. CCOMMON_OPT += -m32
  1023. else
  1024. ifdef BINARY64
  1025. CCOMMON_OPT += -m64
  1026. else
  1027. CCOMMON_OPT += -m32
  1028. endif
  1029. endif
  1030. endif
  1031. ifeq ($(F_COMPILER), SUN)
  1032. CCOMMON_OPT += -DF_INTERFACE_SUN
  1033. FCOMMON_OPT += -ftrap=%none -xrecursive
  1034. ifeq ($(ARCH), x86)
  1035. FCOMMON_OPT += -m32
  1036. else
  1037. ifdef BINARY64
  1038. FCOMMON_OPT += -m64
  1039. else
  1040. FCOMMON_OPT += -m32
  1041. endif
  1042. endif
  1043. ifeq ($(USE_OPENMP), 1)
  1044. FCOMMON_OPT += -xopenmp=parallel
  1045. endif
  1046. endif
  1047. ifeq ($(F_COMPILER), COMPAQ)
  1048. CCOMMON_OPT += -DF_INTERFACE_COMPAQ
  1049. ifeq ($(USE_OPENMP), 1)
  1050. FCOMMON_OPT += -openmp
  1051. endif
  1052. endif
  1053. ifdef BINARY64
  1054. ifdef INTERFACE64
  1055. ifneq ($(INTERFACE64), 0)
  1056. CCOMMON_OPT +=
  1057. #-DUSE64BITINT
  1058. endif
  1059. endif
  1060. endif
  1061. ifeq ($(NEED_PIC), 1)
  1062. ifeq ($(C_COMPILER), IBM)
  1063. CCOMMON_OPT += -qpic=large
  1064. else
  1065. CCOMMON_OPT += -fPIC
  1066. endif
  1067. ifeq ($(F_COMPILER), SUN)
  1068. FCOMMON_OPT += -pic
  1069. else ifeq ($(F_COMPILER), NAG)
  1070. FCOMMON_OPT += -PIC
  1071. else
  1072. FCOMMON_OPT += -fPIC
  1073. endif
  1074. endif
  1075. ifeq ($(DYNAMIC_ARCH), 1)
  1076. CCOMMON_OPT += -DDYNAMIC_ARCH
  1077. endif
  1078. ifeq ($(DYNAMIC_OLDER), 1)
  1079. CCOMMON_OPT += -DDYNAMIC_OLDER
  1080. endif
  1081. ifeq ($(NO_LAPACK), 1)
  1082. CCOMMON_OPT += -DNO_LAPACK
  1083. #Disable LAPACK C interface
  1084. NO_LAPACKE = 1
  1085. endif
  1086. ifeq ($(NO_LAPACKE), 1)
  1087. CCOMMON_OPT += -DNO_LAPACKE
  1088. endif
  1089. ifeq ($(NO_AVX), 1)
  1090. CCOMMON_OPT += -DNO_AVX
  1091. endif
  1092. ifeq ($(ARCH), x86)
  1093. CCOMMON_OPT += -DNO_AVX
  1094. endif
  1095. ifeq ($(NO_AVX2), 1)
  1096. CCOMMON_OPT += -DNO_AVX2
  1097. endif
  1098. ifeq ($(NO_AVX512), 1)
  1099. CCOMMON_OPT += -DNO_AVX512
  1100. endif
  1101. ifdef SMP
  1102. CCOMMON_OPT += -DSMP_SERVER
  1103. ifeq ($(ARCH), mips64)
  1104. USE_SIMPLE_THREADED_LEVEL3 = 1
  1105. endif
  1106. ifeq ($(USE_OPENMP), 1)
  1107. # USE_SIMPLE_THREADED_LEVEL3 = 1
  1108. # NO_AFFINITY = 1
  1109. CCOMMON_OPT += -DUSE_OPENMP
  1110. endif
  1111. ifeq ($(BIGNUMA), 1)
  1112. CCOMMON_OPT += -DBIGNUMA
  1113. endif
  1114. endif
  1115. ifeq ($(NO_WARMUP), 1)
  1116. CCOMMON_OPT += -DNO_WARMUP
  1117. endif
  1118. ifeq ($(CONSISTENT_FPCSR), 1)
  1119. CCOMMON_OPT += -DCONSISTENT_FPCSR
  1120. endif
  1121. # Only for development
  1122. # CCOMMON_OPT += -DPARAMTEST
  1123. # CCOMMON_OPT += -DPREFETCHTEST
  1124. # CCOMMON_OPT += -DNO_SWITCHING
  1125. # USE_PAPI = 1
  1126. ifdef USE_PAPI
  1127. CCOMMON_OPT += -DUSE_PAPI
  1128. EXTRALIB += -lpapi -lperfctr
  1129. endif
  1130. ifdef BUFFERSIZE
  1131. CCOMMON_OPT += -DBUFFERSIZE=$(BUFFERSIZE)
  1132. endif
  1133. ifdef DYNAMIC_THREADS
  1134. CCOMMON_OPT += -DDYNAMIC_THREADS
  1135. endif
  1136. CCOMMON_OPT += -DMAX_CPU_NUMBER=$(NUM_THREADS)
  1137. CCOMMON_OPT += -DMAX_PARALLEL_NUMBER=$(NUM_PARALLEL)
  1138. ifdef USE_SIMPLE_THREADED_LEVEL3
  1139. CCOMMON_OPT += -DUSE_SIMPLE_THREADED_LEVEL3
  1140. endif
  1141. ifeq ($(USE_TLS), 1)
  1142. CCOMMON_OPT += -DUSE_TLS
  1143. endif
  1144. ifeq ($(BUILD_BFLOAT16), 1)
  1145. CCOMMON_OPT += -DBUILD_BFLOAT16
  1146. endif
  1147. ifeq ($(BUILD_SINGLE), 1)
  1148. CCOMMON_OPT += -DBUILD_SINGLE=1
  1149. endif
  1150. ifeq ($(BUILD_DOUBLE), 1)
  1151. CCOMMON_OPT += -DBUILD_DOUBLE=1
  1152. endif
  1153. ifeq ($(BUILD_COMPLEX), 1)
  1154. CCOMMON_OPT += -DBUILD_COMPLEX=1
  1155. endif
  1156. ifeq ($(BUILD_COMPLEX16), 1)
  1157. CCOMMON_OPT += -DBUILD_COMPLEX16=1
  1158. endif
  1159. CCOMMON_OPT += -DVERSION=\"$(VERSION)\"
  1160. ifndef SYMBOLPREFIX
  1161. SYMBOLPREFIX =
  1162. endif
  1163. ifndef SYMBOLSUFFIX
  1164. SYMBOLSUFFIX =
  1165. endif
  1166. ifndef LIBSONAMEBASE
  1167. LIBSONAMEBASE = openblas
  1168. endif
  1169. ifndef LIBNAMESUFFIX
  1170. LIBNAMEBASE = $(SYMBOLPREFIX)$(LIBSONAMEBASE)$(SYMBOLSUFFIX)
  1171. else
  1172. LIBNAMEBASE = $(SYMBOLPREFIX)$(LIBSONAMEBASE)$(SYMBOLSUFFIX)_$(LIBNAMESUFFIX)
  1173. endif
  1174. ifeq ($(OSNAME), CYGWIN_NT)
  1175. LIBPREFIX = cyg$(LIBNAMEBASE)
  1176. else
  1177. LIBPREFIX = lib$(LIBNAMEBASE)
  1178. endif
  1179. KERNELDIR = $(TOPDIR)/kernel/$(ARCH)
  1180. include $(TOPDIR)/Makefile.$(ARCH)
  1181. ifneq ($(C_COMPILER), PGI)
  1182. ifneq ($(C_COMPILER), SUN)
  1183. CCOMMON_OPT += -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME
  1184. endif
  1185. endif
  1186. CCOMMON_OPT += -DASMNAME=$(FU)$(*F) -DASMFNAME=$(FU)$(*F)$(BU) -DNAME=$(*F)$(BU) -DCNAME=$(*F) -DCHAR_NAME=\"$(*F)$(BU)\" -DCHAR_CNAME=\"$(*F)\"
  1187. ifeq ($(CORE), PPC440)
  1188. CCOMMON_OPT += -DALLOC_QALLOC
  1189. endif
  1190. ifeq ($(CORE), PPC440FP2)
  1191. STATIC_ALLOCATION = 1
  1192. endif
  1193. ifneq ($(OSNAME), Linux)
  1194. NO_AFFINITY = 1
  1195. endif
  1196. ifneq ($(ARCH), x86_64)
  1197. ifneq ($(ARCH), x86)
  1198. NO_AFFINITY = 1
  1199. endif
  1200. endif
  1201. ifdef NO_AFFINITY
  1202. ifeq ($(NO_AFFINITY), 0)
  1203. override undefine NO_AFFINITY
  1204. else
  1205. CCOMMON_OPT += -DNO_AFFINITY
  1206. endif
  1207. endif
  1208. ifdef FUNCTION_PROFILE
  1209. CCOMMON_OPT += -DFUNCTION_PROFILE
  1210. endif
  1211. ifdef HUGETLB_ALLOCATION
  1212. CCOMMON_OPT += -DALLOC_HUGETLB
  1213. endif
  1214. ifdef HUGETLBFILE_ALLOCATION
  1215. CCOMMON_OPT += -DALLOC_HUGETLBFILE -DHUGETLB_FILE_NAME=$(HUGETLBFILE_ALLOCATION)
  1216. endif
  1217. ifdef STATIC_ALLOCATION
  1218. CCOMMON_OPT += -DALLOC_STATIC
  1219. endif
  1220. ifdef DEVICEDRIVER_ALLOCATION
  1221. CCOMMON_OPT += -DALLOC_DEVICEDRIVER -DDEVICEDRIVER_NAME=\"/dev/mapper\"
  1222. endif
  1223. ifdef MIXED_MEMORY_ALLOCATION
  1224. CCOMMON_OPT += -DMIXED_MEMORY_ALLOCATION
  1225. endif
  1226. ifeq ($(OSNAME), SunOS)
  1227. TAR = gtar
  1228. PATCH = gpatch
  1229. GREP = ggrep
  1230. AWK = nawk
  1231. else
  1232. TAR = tar
  1233. PATCH = patch
  1234. GREP = grep
  1235. AWK = awk
  1236. endif
  1237. ifndef MD5SUM
  1238. MD5SUM = md5sum
  1239. endif
  1240. REVISION = -r$(VERSION)
  1241. MAJOR_VERSION = $(word 1,$(subst ., ,$(VERSION)))
  1242. ifeq ($(DEBUG), 1)
  1243. COMMON_OPT += -g
  1244. endif
  1245. ifeq ($(DEBUG), 1)
  1246. FCOMMON_OPT += -g
  1247. endif
  1248. ifndef COMMON_OPT
  1249. COMMON_OPT = -O2
  1250. endif
  1251. ifndef FCOMMON_OPT
  1252. FCOMMON_OPT = -O2 -frecursive
  1253. endif
  1254. override CFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
  1255. override PFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
  1256. override FFLAGS += $(COMMON_OPT) $(FCOMMON_OPT)
  1257. override FPFLAGS += $(FCOMMON_OPT) $(COMMON_PROF)
  1258. #MAKEOVERRIDES =
  1259. ifdef NEED_PIC
  1260. ifeq (,$(findstring PIC,$(FFLAGS)))
  1261. override FFLAGS += -fPIC
  1262. endif
  1263. endif
  1264. #For LAPACK Fortran codes.
  1265. #Disable -fopenmp for LAPACK Fortran codes on Windows.
  1266. ifdef OS_WINDOWS
  1267. LAPACK_FFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FFLAGS))
  1268. LAPACK_FPFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FPFLAGS))
  1269. else
  1270. LAPACK_FFLAGS := $(FFLAGS)
  1271. LAPACK_FPFLAGS := $(FPFLAGS)
  1272. endif
  1273. ifeq ($(F_COMPILER),NAG)
  1274. LAPACK_FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1275. endif
  1276. LAPACK_CFLAGS = $(CFLAGS)
  1277. LAPACK_CFLAGS += -DHAVE_LAPACK_CONFIG_H
  1278. ifdef INTERFACE64
  1279. ifneq ($(INTERFACE64), 0)
  1280. LAPACK_CFLAGS += -DLAPACK_ILP64
  1281. endif
  1282. endif
  1283. ifdef OS_WINDOWS
  1284. LAPACK_CFLAGS += -DOPENBLAS_OS_WINDOWS
  1285. endif
  1286. ifeq ($(C_COMPILER), LSB)
  1287. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1288. endif
  1289. ifndef SUFFIX
  1290. SUFFIX = o
  1291. endif
  1292. ifndef PSUFFIX
  1293. PSUFFIX = po
  1294. endif
  1295. ifndef LIBSUFFIX
  1296. LIBSUFFIX = a
  1297. endif
  1298. ifneq ($(DYNAMIC_ARCH), 1)
  1299. ifndef SMP
  1300. LIBNAME = $(LIBPREFIX)_$(LIBCORE)$(REVISION).$(LIBSUFFIX)
  1301. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)$(REVISION)_p.$(LIBSUFFIX)
  1302. else
  1303. LIBNAME = $(LIBPREFIX)_$(LIBCORE)p$(REVISION).$(LIBSUFFIX)
  1304. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)p$(REVISION)_p.$(LIBSUFFIX)
  1305. endif
  1306. else
  1307. ifndef SMP
  1308. LIBNAME = $(LIBPREFIX)$(REVISION).$(LIBSUFFIX)
  1309. LIBNAME_P = $(LIBPREFIX)$(REVISION)_p.$(LIBSUFFIX)
  1310. else
  1311. LIBNAME = $(LIBPREFIX)p$(REVISION).$(LIBSUFFIX)
  1312. LIBNAME_P = $(LIBPREFIX)p$(REVISION)_p.$(LIBSUFFIX)
  1313. endif
  1314. endif
  1315. LIBDLLNAME = $(LIBPREFIX).dll
  1316. IMPLIBNAME = lib$(LIBNAMEBASE).dll.a
  1317. ifneq ($(OSNAME), AIX)
  1318. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.so)
  1319. else
  1320. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.a)
  1321. endif
  1322. LIBDYNNAME = $(LIBNAME:.$(LIBSUFFIX)=.dylib)
  1323. LIBDEFNAME = $(LIBNAME:.$(LIBSUFFIX)=.def)
  1324. LIBEXPNAME = $(LIBNAME:.$(LIBSUFFIX)=.exp)
  1325. LIBZIPNAME = $(LIBNAME:.$(LIBSUFFIX)=.zip)
  1326. LIBS = $(TOPDIR)/$(LIBNAME)
  1327. LIBS_P = $(TOPDIR)/$(LIBNAME_P)
  1328. LIB_COMPONENTS = BLAS
  1329. ifneq ($(NO_CBLAS), 1)
  1330. LIB_COMPONENTS += CBLAS
  1331. endif
  1332. ifneq ($(NO_LAPACK), 1)
  1333. LIB_COMPONENTS += LAPACK
  1334. ifneq ($(NO_LAPACKE), 1)
  1335. LIB_COMPONENTS += LAPACKE
  1336. endif
  1337. ifeq ($(BUILD_RELAPACK), 1)
  1338. LIB_COMPONENTS += ReLAPACK
  1339. endif
  1340. endif
  1341. ifeq ($(ONLY_CBLAS), 1)
  1342. LIB_COMPONENTS = CBLAS
  1343. endif
  1344. export OSNAME
  1345. export ARCH
  1346. export CORE
  1347. export LIBCORE
  1348. export __BYTE_ORDER__
  1349. export ELF_VERSION
  1350. export PGCPATH
  1351. export CONFIG
  1352. export CC
  1353. export FC
  1354. export BU
  1355. export FU
  1356. export NEED2UNDERSCORES
  1357. export USE_THREAD
  1358. export NUM_THREADS
  1359. export NUM_CORES
  1360. export SMP
  1361. export MAKEFILE_RULE
  1362. export NEED_PIC
  1363. export BINARY
  1364. export BINARY32
  1365. export BINARY64
  1366. export F_COMPILER
  1367. export C_COMPILER
  1368. export USE_OPENMP
  1369. export CROSS
  1370. export CROSS_SUFFIX
  1371. export NOFORTRAN
  1372. export NO_FBLAS
  1373. export EXTRALIB
  1374. export CEXTRALIB
  1375. export FEXTRALIB
  1376. export HAVE_SSE
  1377. export HAVE_SSE2
  1378. export HAVE_SSE3
  1379. export HAVE_SSSE3
  1380. export HAVE_SSE4_1
  1381. export HAVE_SSE4_2
  1382. export HAVE_SSE4A
  1383. export HAVE_SSE5
  1384. export HAVE_AVX
  1385. export HAVE_AVX2
  1386. export HAVE_FMA3
  1387. export HAVE_VFP
  1388. export HAVE_VFPV3
  1389. export HAVE_VFPV4
  1390. export HAVE_NEON
  1391. export HAVE_MSA
  1392. export MSA_FLAGS
  1393. export KERNELDIR
  1394. export FUNCTION_PROFILE
  1395. export TARGET_CORE
  1396. export NO_AVX512
  1397. export NO_AVX2
  1398. export BUILD_BFLOAT16
  1399. export SBGEMM_UNROLL_M
  1400. export SBGEMM_UNROLL_N
  1401. export SGEMM_UNROLL_M
  1402. export SGEMM_UNROLL_N
  1403. export DGEMM_UNROLL_M
  1404. export DGEMM_UNROLL_N
  1405. export QGEMM_UNROLL_M
  1406. export QGEMM_UNROLL_N
  1407. export CGEMM_UNROLL_M
  1408. export CGEMM_UNROLL_N
  1409. export ZGEMM_UNROLL_M
  1410. export ZGEMM_UNROLL_N
  1411. export XGEMM_UNROLL_M
  1412. export XGEMM_UNROLL_N
  1413. export CGEMM3M_UNROLL_M
  1414. export CGEMM3M_UNROLL_N
  1415. export ZGEMM3M_UNROLL_M
  1416. export ZGEMM3M_UNROLL_N
  1417. export XGEMM3M_UNROLL_M
  1418. export XGEMM3M_UNROLL_N
  1419. ifdef USE_CUDA
  1420. export CUDADIR
  1421. export CUCC
  1422. export CUFLAGS
  1423. export CULIB
  1424. endif
  1425. .SUFFIXES: .$(PSUFFIX) .$(SUFFIX) .f
  1426. .f.$(SUFFIX):
  1427. $(FC) $(FFLAGS) -c $< -o $(@F)
  1428. .f.$(PSUFFIX):
  1429. $(FC) $(FPFLAGS) -pg -c $< -o $(@F)
  1430. ifdef BINARY64
  1431. PATHSCALEPATH = /opt/pathscale/lib/3.1
  1432. PGIPATH = /opt/pgi/linux86-64/7.1-5/lib
  1433. else
  1434. PATHSCALEPATH = /opt/pathscale/lib/3.1/32
  1435. PGIPATH = /opt/pgi/linux86/7.1-5/lib
  1436. endif
  1437. ACMLPATH = /opt/acml/4.3.0
  1438. ifneq ($(OSNAME), Darwin)
  1439. MKLPATH = /opt/intel/mkl/10.2.2.025/lib
  1440. else
  1441. MKLPATH = /Library/Frameworks/Intel_MKL.framework/Versions/10.0.1.014/lib
  1442. endif
  1443. ATLASPATH = /opt/atlas/3.9.17/opteron
  1444. FLAMEPATH = $(HOME)/flame/lib
  1445. ifneq ($(OSNAME), SunOS)
  1446. SUNPATH = /opt/sunstudio12.1
  1447. else
  1448. SUNPATH = /opt/SUNWspro
  1449. endif