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
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942943944945946947948949950951952953954955956957958959960961962963964965966967968969970971972973974975976977978979980981982983984985986987988989990991992993994995996997998999100010011002100310041005100610071008100910101011101210131014101510161017101810191020102110221023102410251026102710281029103010311032103310341035103610371038103910401041104210431044104510461047104810491050105110521053105410551056105710581059106010611062106310641065106610671068106910701071107210731074107510761077107810791080108110821083108410851086108710881089109010911092109310941095109610971098109911001101110211031104110511061107110811091110111111121113111411151116111711181119112011211122112311241125112611271128112911301131113211331134113511361137113811391140114111421143114411451146114711481149115011511152115311541155115611571158115911601161116211631164116511661167116811691170117111721173117411751176117711781179118011811182118311841185118611871188118911901191119211931194119511961197119811991200120112021203120412051206120712081209121012111212121312141215121612171218121912201221122212231224122512261227122812291230123112321233123412351236123712381239124012411242124312441245124612471248124912501251125212531254125512561257125812591260126112621263126412651266126712681269127012711272127312741275127612771278127912801281128212831284128512861287128812891290129112921293129412951296129712981299130013011302130313041305130613071308130913101311131213131314131513161317131813191320132113221323132413251326132713281329133013311332133313341335133613371338133913401341134213431344134513461347134813491350135113521353135413551356135713581359136013611362136313641365136613671368136913701371137213731374137513761377137813791380138113821383138413851386138713881389139013911392139313941395139613971398139914001401140214031404140514061407140814091410141114121413141414151416141714181419142014211422142314241425142614271428142914301431143214331434143514361437143814391440144114421443144414451446144714481449145014511452145314541455145614571458145914601461146214631464146514661467146814691470147114721473147414751476147714781479148014811482148314841485148614871488148914901491149214931494149514961497149814991500150115021503150415051506150715081509151015111512151315141515151615171518151915201521152215231524152515261527152815291530153115321533153415351536153715381539154015411542154315441545154615471548154915501551155215531554155515561557155815591560156115621563156415651566156715681569157015711572157315741575157615771578157915801581158215831584158515861587158815891590159115921593159415951596159715981599160016011602160316041605160616071608160916101611161216131614161516161617161816191620162116221623162416251626162716281629163016311632163316341635163616371638163916401641164216431644164516461647164816491650165116521653165416551656165716581659166016611662166316641665166616671668166916701671167216731674167516761677167816791680168116821683168416851686168716881689169016911692169316941695169616971698169917001701
  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 ($(ARCH), riscv64)
  665. NO_BINARY_MODE = 1
  666. BINARY_DEFINED = 1
  667. endif
  668. ifeq ($(ARCH), loongarch64)
  669. NO_BINARY_MODE = 1
  670. BINARY_DEFINED = 1
  671. endif
  672. #
  673. # C Compiler dependent settings
  674. #
  675. # ifeq logical or. GCC or CLANG or LSB
  676. # http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
  677. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG LSB))
  678. CCOMMON_OPT += -Wall
  679. COMMON_PROF += -fno-inline
  680. NO_UNINITIALIZED_WARN = -Wno-uninitialized
  681. ifeq ($(QUIET_MAKE), 1)
  682. CCOMMON_OPT += $(NO_UNINITIALIZED_WARN) -Wno-unused
  683. endif
  684. ifdef NO_BINARY_MODE
  685. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  686. ifdef BINARY64
  687. CCOMMON_OPT += -mabi=64
  688. else
  689. CCOMMON_OPT += -mabi=n32
  690. endif
  691. BINARY_DEFINED = 1
  692. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  693. CCOMMON_OPT += -mabi=32
  694. BINARY_DEFINED = 1
  695. endif
  696. ifeq ($(CORE), $(filter $(CORE),LOONGSON3R3 LOONGSON3R4))
  697. CCOMMON_OPT += -march=loongson3a
  698. FCOMMON_OPT += -march=loongson3a
  699. endif
  700. ifeq ($(CORE), MIPS24K)
  701. CCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  702. FCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  703. endif
  704. ifeq ($(CORE), MIPS1004K)
  705. CCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  706. FCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  707. endif
  708. ifeq ($(CORE), P5600)
  709. CCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  710. FCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  711. endif
  712. ifeq ($(CORE), I6400)
  713. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  714. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  715. endif
  716. ifeq ($(CORE), P6600)
  717. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  718. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  719. endif
  720. ifeq ($(CORE), I6500)
  721. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  722. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  723. endif
  724. ifeq ($(OSNAME), AIX)
  725. BINARY_DEFINED = 1
  726. endif
  727. ifeq ($(ARCH), loongarch64)
  728. ifeq ($(CORE), LOONGSONG3R5)
  729. CCOMMON_OPT += -march=loongarch64 -mabi=lp64
  730. FCOMMON_OPT += -march=loongarch64 -mabi=lp64
  731. endif
  732. endif
  733. endif
  734. ifndef BINARY_DEFINED
  735. ifneq ($(OSNAME), AIX)
  736. ifdef BINARY64
  737. ifneq ($(ARCH), riscv64)
  738. CCOMMON_OPT += -m64
  739. endif
  740. else
  741. CCOMMON_OPT += -m32
  742. endif
  743. endif
  744. endif
  745. endif
  746. ifeq ($(C_COMPILER), PGI)
  747. PGCVERSIONGT20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \> 20)
  748. PGCVERSIONGTEQ20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \>= 20)
  749. PGCMINORVERSIONGE11 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -c 4-5` == 11)
  750. PGCVERSIONCHECK := $(PGCVERSIONGT20)$(PGCVERSIONEQ20)$(PGCMINORVERSIONGE11)
  751. ifeq ($(PGCVERSIONCHECK), $(filter $(PGCVERSIONCHECK), 110 111 011))
  752. NEWPGI := 1
  753. endif
  754. ifdef BINARY64
  755. ifeq ($(ARCH), x86_64)
  756. CCOMMON_OPT += -tp p7-64
  757. ifneq ($(NEWPGI),1)
  758. CCOMMON_OPT += -D__MMX__ -Mnollvm
  759. endif
  760. else
  761. ifeq ($(ARCH), power)
  762. ifeq ($(CORE), POWER8)
  763. CCOMMON_OPT += -tp pwr8
  764. endif
  765. ifeq ($(CORE), POWER9)
  766. CCOMMON_OPT += -tp pwr9
  767. endif
  768. endif
  769. endif
  770. else
  771. CCOMMON_OPT += -tp p7
  772. endif
  773. endif
  774. ifeq ($(C_COMPILER), PATHSCALE)
  775. ifdef BINARY64
  776. CCOMMON_OPT += -m64
  777. else
  778. CCOMMON_OPT += -m32
  779. endif
  780. endif
  781. #
  782. # Fortran Compiler dependent settings
  783. #
  784. ifeq ($(F_COMPILER), NAG)
  785. FCOMMON_OPT += -dcfuns -recursive -ieee=full -w=obs -thread_safe
  786. ifdef INTERFACE64
  787. ifneq ($(INTERFACE64), 0)
  788. FCOMMON_OPT += -i8
  789. endif
  790. endif
  791. ifeq ($(USE_OPENMP), 1)
  792. FCOMMON_OPT += -openmp
  793. endif
  794. endif
  795. ifeq ($(F_COMPILER), FLANG)
  796. CCOMMON_OPT += -DF_INTERFACE_FLANG
  797. FCOMMON_OPT += -Mrecursive -Kieee
  798. ifeq ($(OSNAME), Linux)
  799. ifeq ($(ARCH), x86_64)
  800. FLANG_VENDOR := $(shell $(FC) --version|head -1 |cut -f 1 -d " ")
  801. ifeq ($(FLANG_VENDOR), AMD)
  802. FCOMMON_OPT += -fno-unroll-loops
  803. endif
  804. endif
  805. endif
  806. ifdef BINARY64
  807. ifdef INTERFACE64
  808. ifneq ($(INTERFACE64), 0)
  809. FCOMMON_OPT += -i8
  810. endif
  811. endif
  812. FCOMMON_OPT += -Wall
  813. else
  814. FCOMMON_OPT += -Wall
  815. endif
  816. ifeq ($(USE_OPENMP), 1)
  817. FCOMMON_OPT += -fopenmp
  818. endif
  819. endif
  820. ifeq ($(F_COMPILER), G77)
  821. CCOMMON_OPT += -DF_INTERFACE_G77
  822. FCOMMON_OPT += -Wall
  823. ifndef NO_BINARY_MODE
  824. ifneq ($(OSNAME), AIX)
  825. ifdef BINARY64
  826. FCOMMON_OPT += -m64
  827. else
  828. FCOMMON_OPT += -m32
  829. endif
  830. endif
  831. endif
  832. endif
  833. ifeq ($(F_COMPILER), G95)
  834. CCOMMON_OPT += -DF_INTERFACE_G95
  835. FCOMMON_OPT += -Wall
  836. ifneq ($(OSNAME), AIX)
  837. ifndef NO_BINARY_MODE
  838. ifdef BINARY64
  839. FCOMMON_OPT += -m64
  840. else
  841. FCOMMON_OPT += -m32
  842. endif
  843. endif
  844. ifneq ($(NO_LAPACKE), 1)
  845. FCOMMON_OPT += -fno-second-underscore
  846. endif
  847. endif
  848. endif
  849. ifeq ($(F_COMPILER), GFORTRAN)
  850. CCOMMON_OPT += -DF_INTERFACE_GFORT
  851. FCOMMON_OPT += -Wall
  852. # make single-threaded LAPACK calls thread-safe #1847
  853. FCOMMON_OPT += -frecursive
  854. # work around ABI problem with passing single-character arguments
  855. FCOMMON_OPT += -fno-optimize-sibling-calls
  856. #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
  857. ifneq ($(NO_LAPACK), 1)
  858. EXTRALIB += -lgfortran
  859. endif
  860. ifdef NO_BINARY_MODE
  861. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  862. ifdef BINARY64
  863. FCOMMON_OPT += -mabi=64
  864. else
  865. FCOMMON_OPT += -mabi=n32
  866. endif
  867. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  868. FCOMMON_OPT += -mabi=32
  869. endif
  870. else
  871. ifdef BINARY64
  872. ifneq ($(OSNAME), AIX)
  873. ifneq ($(ARCH), riscv64)
  874. FCOMMON_OPT += -m64
  875. endif
  876. endif
  877. ifdef INTERFACE64
  878. ifneq ($(INTERFACE64), 0)
  879. FCOMMON_OPT += -fdefault-integer-8
  880. endif
  881. endif
  882. else
  883. ifneq ($(OSNAME), AIX)
  884. FCOMMON_OPT += -m32
  885. endif
  886. endif
  887. endif
  888. ifeq ($(USE_OPENMP), 1)
  889. FCOMMON_OPT += -fopenmp
  890. endif
  891. endif
  892. ifeq ($(F_COMPILER), INTEL)
  893. CCOMMON_OPT += -DF_INTERFACE_INTEL
  894. ifdef INTERFACE64
  895. ifneq ($(INTERFACE64), 0)
  896. FCOMMON_OPT += -i8
  897. endif
  898. endif
  899. FCOMMON_OPT += -recursive -fp-model strict -assume protect-parens
  900. ifeq ($(USE_OPENMP), 1)
  901. FCOMMON_OPT += -fopenmp
  902. endif
  903. endif
  904. ifeq ($(F_COMPILER), FUJITSU)
  905. CCOMMON_OPT += -DF_INTERFACE_FUJITSU
  906. ifeq ($(USE_OPENMP), 1)
  907. FCOMMON_OPT += -openmp
  908. endif
  909. endif
  910. ifeq ($(F_COMPILER), IBM)
  911. CCOMMON_OPT += -DF_INTERFACE_IBM
  912. # FCOMMON_OPT += -qarch=440
  913. ifdef BINARY64
  914. FCOMMON_OPT += -q64
  915. ifdef INTERFACE64
  916. ifneq ($(INTERFACE64), 0)
  917. FCOMMON_OPT += -qintsize=8
  918. endif
  919. endif
  920. else
  921. FCOMMON_OPT += -q32
  922. endif
  923. ifeq ($(USE_OPENMP), 1)
  924. FCOMMON_OPT += -openmp
  925. endif
  926. endif
  927. ifeq ($(F_COMPILER), PGI)
  928. CCOMMON_OPT += -DF_INTERFACE_PGI
  929. COMMON_PROF += -DPGICOMPILER
  930. ifdef BINARY64
  931. ifdef INTERFACE64
  932. ifneq ($(INTERFACE64), 0)
  933. FCOMMON_OPT += -i8
  934. endif
  935. endif
  936. ifeq ($(ARCH), x86_64)
  937. FCOMMON_OPT += -tp p7-64
  938. else
  939. ifeq ($(ARCH), power)
  940. ifeq ($(CORE), POWER6)
  941. $(warning NVIDIA HPC compilers do not support POWER6.)
  942. endif
  943. ifeq ($(CORE), POWER8)
  944. FCOMMON_OPT += -tp pwr8
  945. endif
  946. ifeq ($(CORE), POWER9)
  947. FCOMMON_OPT += -tp pwr9
  948. endif
  949. ifeq ($(CORE), POWER10)
  950. $(warning NVIDIA HPC compilers do not support POWER10.)
  951. endif
  952. endif
  953. endif
  954. else
  955. FCOMMON_OPT += -tp p7
  956. endif
  957. FCOMMON_OPT += -Mrecursive -Kieee
  958. ifeq ($(USE_OPENMP), 1)
  959. FCOMMON_OPT += -mp
  960. endif
  961. endif
  962. ifeq ($(F_COMPILER), PATHSCALE)
  963. CCOMMON_OPT += -DF_INTERFACE_PATHSCALE
  964. ifdef BINARY64
  965. ifdef INTERFACE64
  966. ifneq ($(INTERFACE64), 0)
  967. FCOMMON_OPT += -i8
  968. endif
  969. endif
  970. endif
  971. ifeq ($(USE_OPENMP), 1)
  972. FCOMMON_OPT += -mp
  973. endif
  974. endif
  975. ifeq ($(F_COMPILER), OPEN64)
  976. CCOMMON_OPT += -DF_INTERFACE_OPEN64
  977. ifdef BINARY64
  978. ifdef INTERFACE64
  979. ifneq ($(INTERFACE64), 0)
  980. FCOMMON_OPT += -i8
  981. endif
  982. endif
  983. endif
  984. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  985. ifndef BINARY64
  986. FCOMMON_OPT += -n32
  987. else
  988. FCOMMON_OPT += -n64
  989. endif
  990. ifeq ($(CORE), LOONGSON3R3)
  991. FCOMMON_OPT += -loongson3 -static
  992. endif
  993. ifeq ($(CORE), LOONGSON3R4)
  994. FCOMMON_OPT += -loongson3 -static
  995. endif
  996. else
  997. ifndef BINARY64
  998. FCOMMON_OPT += -m32
  999. else
  1000. FCOMMON_OPT += -m64
  1001. endif
  1002. endif
  1003. ifeq ($(USE_OPENMP), 1)
  1004. FEXTRALIB += -lstdc++
  1005. FCOMMON_OPT += -mp
  1006. endif
  1007. endif
  1008. ifeq ($(C_COMPILER), OPEN64)
  1009. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  1010. ifndef BINARY64
  1011. CCOMMON_OPT += -n32
  1012. else
  1013. CCOMMON_OPT += -n64
  1014. endif
  1015. ifeq ($(CORE), LOONGSON3R3)
  1016. CCOMMON_OPT += -loongson3 -static
  1017. endif
  1018. ifeq ($(CORE), LOONGSON3R4)
  1019. CCOMMON_OPT += -loongson3 -static
  1020. endif
  1021. else
  1022. ifndef BINARY64
  1023. CCOMMON_OPT += -m32
  1024. else
  1025. CCOMMON_OPT += -m64
  1026. endif
  1027. endif
  1028. endif
  1029. ifeq ($(C_COMPILER), SUN)
  1030. CCOMMON_OPT += -w
  1031. ifeq ($(ARCH), x86)
  1032. CCOMMON_OPT += -m32
  1033. else
  1034. ifdef BINARY64
  1035. CCOMMON_OPT += -m64
  1036. else
  1037. CCOMMON_OPT += -m32
  1038. endif
  1039. endif
  1040. endif
  1041. ifeq ($(F_COMPILER), SUN)
  1042. CCOMMON_OPT += -DF_INTERFACE_SUN
  1043. FCOMMON_OPT += -ftrap=%none -xrecursive
  1044. ifeq ($(ARCH), x86)
  1045. FCOMMON_OPT += -m32
  1046. else
  1047. ifdef BINARY64
  1048. FCOMMON_OPT += -m64
  1049. else
  1050. FCOMMON_OPT += -m32
  1051. endif
  1052. endif
  1053. ifeq ($(USE_OPENMP), 1)
  1054. FCOMMON_OPT += -xopenmp=parallel
  1055. endif
  1056. endif
  1057. ifeq ($(F_COMPILER), COMPAQ)
  1058. CCOMMON_OPT += -DF_INTERFACE_COMPAQ
  1059. ifeq ($(USE_OPENMP), 1)
  1060. FCOMMON_OPT += -openmp
  1061. endif
  1062. endif
  1063. ifdef BINARY64
  1064. ifdef INTERFACE64
  1065. ifneq ($(INTERFACE64), 0)
  1066. CCOMMON_OPT +=
  1067. #-DUSE64BITINT
  1068. endif
  1069. endif
  1070. endif
  1071. ifeq ($(NEED_PIC), 1)
  1072. ifeq ($(C_COMPILER), IBM)
  1073. CCOMMON_OPT += -qpic=large
  1074. else
  1075. CCOMMON_OPT += -fPIC
  1076. endif
  1077. ifeq ($(F_COMPILER), SUN)
  1078. FCOMMON_OPT += -pic
  1079. else ifeq ($(F_COMPILER), NAG)
  1080. FCOMMON_OPT += -PIC
  1081. else
  1082. FCOMMON_OPT += -fPIC
  1083. endif
  1084. endif
  1085. ifeq ($(DYNAMIC_ARCH), 1)
  1086. CCOMMON_OPT += -DDYNAMIC_ARCH
  1087. endif
  1088. ifeq ($(DYNAMIC_OLDER), 1)
  1089. CCOMMON_OPT += -DDYNAMIC_OLDER
  1090. endif
  1091. ifeq ($(NO_LAPACK), 1)
  1092. CCOMMON_OPT += -DNO_LAPACK
  1093. #Disable LAPACK C interface
  1094. NO_LAPACKE = 1
  1095. endif
  1096. ifeq ($(NO_LAPACKE), 1)
  1097. CCOMMON_OPT += -DNO_LAPACKE
  1098. endif
  1099. ifeq ($(NO_AVX), 1)
  1100. CCOMMON_OPT += -DNO_AVX
  1101. endif
  1102. ifeq ($(ARCH), x86)
  1103. CCOMMON_OPT += -DNO_AVX
  1104. endif
  1105. ifeq ($(NO_AVX2), 1)
  1106. CCOMMON_OPT += -DNO_AVX2
  1107. endif
  1108. ifeq ($(NO_AVX512), 1)
  1109. CCOMMON_OPT += -DNO_AVX512
  1110. endif
  1111. ifdef SMP
  1112. CCOMMON_OPT += -DSMP_SERVER
  1113. ifeq ($(ARCH), mips64)
  1114. USE_SIMPLE_THREADED_LEVEL3 = 1
  1115. endif
  1116. ifeq ($(USE_OPENMP), 1)
  1117. # USE_SIMPLE_THREADED_LEVEL3 = 1
  1118. # NO_AFFINITY = 1
  1119. CCOMMON_OPT += -DUSE_OPENMP
  1120. endif
  1121. ifeq ($(BIGNUMA), 1)
  1122. CCOMMON_OPT += -DBIGNUMA
  1123. endif
  1124. endif
  1125. ifeq ($(NO_WARMUP), 1)
  1126. CCOMMON_OPT += -DNO_WARMUP
  1127. endif
  1128. ifeq ($(CONSISTENT_FPCSR), 1)
  1129. CCOMMON_OPT += -DCONSISTENT_FPCSR
  1130. endif
  1131. # Only for development
  1132. # CCOMMON_OPT += -DPARAMTEST
  1133. # CCOMMON_OPT += -DPREFETCHTEST
  1134. # CCOMMON_OPT += -DNO_SWITCHING
  1135. # USE_PAPI = 1
  1136. ifdef USE_PAPI
  1137. CCOMMON_OPT += -DUSE_PAPI
  1138. EXTRALIB += -lpapi -lperfctr
  1139. endif
  1140. ifdef BUFFERSIZE
  1141. CCOMMON_OPT += -DBUFFERSIZE=$(BUFFERSIZE)
  1142. endif
  1143. ifdef DYNAMIC_THREADS
  1144. CCOMMON_OPT += -DDYNAMIC_THREADS
  1145. endif
  1146. CCOMMON_OPT += -DMAX_CPU_NUMBER=$(NUM_THREADS)
  1147. CCOMMON_OPT += -DMAX_PARALLEL_NUMBER=$(NUM_PARALLEL)
  1148. ifdef USE_SIMPLE_THREADED_LEVEL3
  1149. CCOMMON_OPT += -DUSE_SIMPLE_THREADED_LEVEL3
  1150. endif
  1151. ifeq ($(USE_TLS), 1)
  1152. CCOMMON_OPT += -DUSE_TLS
  1153. endif
  1154. ifeq ($(BUILD_BFLOAT16), 1)
  1155. CCOMMON_OPT += -DBUILD_BFLOAT16
  1156. endif
  1157. ifeq ($(BUILD_SINGLE), 1)
  1158. CCOMMON_OPT += -DBUILD_SINGLE=1
  1159. endif
  1160. ifeq ($(BUILD_DOUBLE), 1)
  1161. CCOMMON_OPT += -DBUILD_DOUBLE=1
  1162. endif
  1163. ifeq ($(BUILD_COMPLEX), 1)
  1164. CCOMMON_OPT += -DBUILD_COMPLEX=1
  1165. endif
  1166. ifeq ($(BUILD_COMPLEX16), 1)
  1167. CCOMMON_OPT += -DBUILD_COMPLEX16=1
  1168. endif
  1169. CCOMMON_OPT += -DVERSION=\"$(VERSION)\"
  1170. ifndef SYMBOLPREFIX
  1171. SYMBOLPREFIX =
  1172. endif
  1173. ifndef SYMBOLSUFFIX
  1174. SYMBOLSUFFIX =
  1175. endif
  1176. ifndef LIBSONAMEBASE
  1177. LIBSONAMEBASE = openblas
  1178. endif
  1179. ifndef LIBNAMESUFFIX
  1180. LIBNAMEBASE = $(SYMBOLPREFIX)$(LIBSONAMEBASE)$(SYMBOLSUFFIX)
  1181. else
  1182. LIBNAMEBASE = $(SYMBOLPREFIX)$(LIBSONAMEBASE)$(SYMBOLSUFFIX)_$(LIBNAMESUFFIX)
  1183. endif
  1184. ifeq ($(OSNAME), CYGWIN_NT)
  1185. LIBPREFIX = cyg$(LIBNAMEBASE)
  1186. else
  1187. LIBPREFIX = lib$(LIBNAMEBASE)
  1188. endif
  1189. KERNELDIR = $(TOPDIR)/kernel/$(ARCH)
  1190. include $(TOPDIR)/Makefile.$(ARCH)
  1191. ifneq ($(C_COMPILER), PGI)
  1192. ifneq ($(C_COMPILER), SUN)
  1193. CCOMMON_OPT += -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME
  1194. endif
  1195. endif
  1196. CCOMMON_OPT += -DASMNAME=$(FU)$(*F) -DASMFNAME=$(FU)$(*F)$(BU) -DNAME=$(*F)$(BU) -DCNAME=$(*F) -DCHAR_NAME=\"$(*F)$(BU)\" -DCHAR_CNAME=\"$(*F)\"
  1197. ifeq ($(CORE), PPC440)
  1198. CCOMMON_OPT += -DALLOC_QALLOC
  1199. endif
  1200. ifeq ($(CORE), PPC440FP2)
  1201. STATIC_ALLOCATION = 1
  1202. endif
  1203. ifneq ($(OSNAME), Linux)
  1204. NO_AFFINITY = 1
  1205. endif
  1206. ifneq ($(ARCH), x86_64)
  1207. ifneq ($(ARCH), x86)
  1208. NO_AFFINITY = 1
  1209. endif
  1210. endif
  1211. ifdef NO_AFFINITY
  1212. ifeq ($(NO_AFFINITY), 0)
  1213. override undefine NO_AFFINITY
  1214. else
  1215. CCOMMON_OPT += -DNO_AFFINITY
  1216. endif
  1217. endif
  1218. ifdef FUNCTION_PROFILE
  1219. CCOMMON_OPT += -DFUNCTION_PROFILE
  1220. endif
  1221. ifdef HUGETLB_ALLOCATION
  1222. CCOMMON_OPT += -DALLOC_HUGETLB
  1223. endif
  1224. ifdef HUGETLBFILE_ALLOCATION
  1225. CCOMMON_OPT += -DALLOC_HUGETLBFILE -DHUGETLB_FILE_NAME=$(HUGETLBFILE_ALLOCATION)
  1226. endif
  1227. ifdef STATIC_ALLOCATION
  1228. CCOMMON_OPT += -DALLOC_STATIC
  1229. endif
  1230. ifdef DEVICEDRIVER_ALLOCATION
  1231. CCOMMON_OPT += -DALLOC_DEVICEDRIVER -DDEVICEDRIVER_NAME=\"/dev/mapper\"
  1232. endif
  1233. ifdef MIXED_MEMORY_ALLOCATION
  1234. CCOMMON_OPT += -DMIXED_MEMORY_ALLOCATION
  1235. endif
  1236. ifeq ($(OSNAME), SunOS)
  1237. TAR = gtar
  1238. PATCH = gpatch
  1239. GREP = ggrep
  1240. AWK = nawk
  1241. else
  1242. TAR = tar
  1243. PATCH = patch
  1244. GREP = grep
  1245. AWK = awk
  1246. endif
  1247. ifndef MD5SUM
  1248. MD5SUM = md5sum
  1249. endif
  1250. REVISION = -r$(VERSION)
  1251. MAJOR_VERSION = $(word 1,$(subst ., ,$(VERSION)))
  1252. ifeq ($(DEBUG), 1)
  1253. COMMON_OPT += -g
  1254. endif
  1255. ifeq ($(DEBUG), 1)
  1256. FCOMMON_OPT += -g
  1257. endif
  1258. ifndef COMMON_OPT
  1259. COMMON_OPT = -O2
  1260. endif
  1261. ifndef FCOMMON_OPT
  1262. FCOMMON_OPT = -O2 -frecursive
  1263. endif
  1264. override CFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
  1265. override PFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
  1266. override FFLAGS += $(COMMON_OPT) $(FCOMMON_OPT)
  1267. override FPFLAGS += $(FCOMMON_OPT) $(COMMON_PROF)
  1268. #MAKEOVERRIDES =
  1269. ifdef NEED_PIC
  1270. ifeq (,$(findstring PIC,$(FFLAGS)))
  1271. override FFLAGS += -fPIC
  1272. endif
  1273. endif
  1274. #For LAPACK Fortran codes.
  1275. #Disable -fopenmp for LAPACK Fortran codes on Windows.
  1276. ifdef OS_WINDOWS
  1277. LAPACK_FFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FFLAGS))
  1278. LAPACK_FPFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FPFLAGS))
  1279. else
  1280. LAPACK_FFLAGS := $(FFLAGS)
  1281. LAPACK_FPFLAGS := $(FPFLAGS)
  1282. endif
  1283. ifeq ($(F_COMPILER),NAG)
  1284. LAPACK_FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1285. endif
  1286. LAPACK_CFLAGS = $(CFLAGS)
  1287. LAPACK_CFLAGS += -DHAVE_LAPACK_CONFIG_H
  1288. ifdef INTERFACE64
  1289. ifneq ($(INTERFACE64), 0)
  1290. LAPACK_CFLAGS += -DLAPACK_ILP64
  1291. endif
  1292. endif
  1293. ifdef OS_WINDOWS
  1294. LAPACK_CFLAGS += -DOPENBLAS_OS_WINDOWS
  1295. endif
  1296. ifeq ($(C_COMPILER), LSB)
  1297. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1298. endif
  1299. ifndef SUFFIX
  1300. SUFFIX = o
  1301. endif
  1302. ifndef PSUFFIX
  1303. PSUFFIX = po
  1304. endif
  1305. ifndef LIBSUFFIX
  1306. LIBSUFFIX = a
  1307. endif
  1308. ifneq ($(DYNAMIC_ARCH), 1)
  1309. ifndef SMP
  1310. LIBNAME = $(LIBPREFIX)_$(LIBCORE)$(REVISION).$(LIBSUFFIX)
  1311. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)$(REVISION)_p.$(LIBSUFFIX)
  1312. else
  1313. LIBNAME = $(LIBPREFIX)_$(LIBCORE)p$(REVISION).$(LIBSUFFIX)
  1314. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)p$(REVISION)_p.$(LIBSUFFIX)
  1315. endif
  1316. else
  1317. ifndef SMP
  1318. LIBNAME = $(LIBPREFIX)$(REVISION).$(LIBSUFFIX)
  1319. LIBNAME_P = $(LIBPREFIX)$(REVISION)_p.$(LIBSUFFIX)
  1320. else
  1321. LIBNAME = $(LIBPREFIX)p$(REVISION).$(LIBSUFFIX)
  1322. LIBNAME_P = $(LIBPREFIX)p$(REVISION)_p.$(LIBSUFFIX)
  1323. endif
  1324. endif
  1325. LIBDLLNAME = $(LIBPREFIX).dll
  1326. IMPLIBNAME = lib$(LIBNAMEBASE).dll.a
  1327. ifneq ($(OSNAME), AIX)
  1328. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.so)
  1329. else
  1330. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.a)
  1331. endif
  1332. LIBDYNNAME = $(LIBNAME:.$(LIBSUFFIX)=.dylib)
  1333. LIBDEFNAME = $(LIBNAME:.$(LIBSUFFIX)=.def)
  1334. LIBEXPNAME = $(LIBNAME:.$(LIBSUFFIX)=.exp)
  1335. LIBZIPNAME = $(LIBNAME:.$(LIBSUFFIX)=.zip)
  1336. LIBS = $(TOPDIR)/$(LIBNAME)
  1337. LIBS_P = $(TOPDIR)/$(LIBNAME_P)
  1338. LIB_COMPONENTS = BLAS
  1339. ifneq ($(NO_CBLAS), 1)
  1340. LIB_COMPONENTS += CBLAS
  1341. endif
  1342. ifneq ($(NO_LAPACK), 1)
  1343. LIB_COMPONENTS += LAPACK
  1344. ifneq ($(NO_LAPACKE), 1)
  1345. LIB_COMPONENTS += LAPACKE
  1346. endif
  1347. ifeq ($(BUILD_RELAPACK), 1)
  1348. LIB_COMPONENTS += ReLAPACK
  1349. endif
  1350. endif
  1351. ifeq ($(ONLY_CBLAS), 1)
  1352. LIB_COMPONENTS = CBLAS
  1353. endif
  1354. export OSNAME
  1355. export ARCH
  1356. export CORE
  1357. export LIBCORE
  1358. export __BYTE_ORDER__
  1359. export ELF_VERSION
  1360. export PGCPATH
  1361. export CONFIG
  1362. export CC
  1363. export FC
  1364. export BU
  1365. export FU
  1366. export NEED2UNDERSCORES
  1367. export USE_THREAD
  1368. export NUM_THREADS
  1369. export NUM_CORES
  1370. export SMP
  1371. export MAKEFILE_RULE
  1372. export NEED_PIC
  1373. export BINARY
  1374. export BINARY32
  1375. export BINARY64
  1376. export F_COMPILER
  1377. export C_COMPILER
  1378. export USE_OPENMP
  1379. export CROSS
  1380. export CROSS_SUFFIX
  1381. export NOFORTRAN
  1382. export NO_FBLAS
  1383. export EXTRALIB
  1384. export CEXTRALIB
  1385. export FEXTRALIB
  1386. export HAVE_SSE
  1387. export HAVE_SSE2
  1388. export HAVE_SSE3
  1389. export HAVE_SSSE3
  1390. export HAVE_SSE4_1
  1391. export HAVE_SSE4_2
  1392. export HAVE_SSE4A
  1393. export HAVE_SSE5
  1394. export HAVE_AVX
  1395. export HAVE_AVX2
  1396. export HAVE_FMA3
  1397. export HAVE_VFP
  1398. export HAVE_VFPV3
  1399. export HAVE_VFPV4
  1400. export HAVE_NEON
  1401. export HAVE_MSA
  1402. export MSA_FLAGS
  1403. export KERNELDIR
  1404. export FUNCTION_PROFILE
  1405. export TARGET_CORE
  1406. export NO_AVX512
  1407. export NO_AVX2
  1408. export BUILD_BFLOAT16
  1409. export SBGEMM_UNROLL_M
  1410. export SBGEMM_UNROLL_N
  1411. export SGEMM_UNROLL_M
  1412. export SGEMM_UNROLL_N
  1413. export DGEMM_UNROLL_M
  1414. export DGEMM_UNROLL_N
  1415. export QGEMM_UNROLL_M
  1416. export QGEMM_UNROLL_N
  1417. export CGEMM_UNROLL_M
  1418. export CGEMM_UNROLL_N
  1419. export ZGEMM_UNROLL_M
  1420. export ZGEMM_UNROLL_N
  1421. export XGEMM_UNROLL_M
  1422. export XGEMM_UNROLL_N
  1423. export CGEMM3M_UNROLL_M
  1424. export CGEMM3M_UNROLL_N
  1425. export ZGEMM3M_UNROLL_M
  1426. export ZGEMM3M_UNROLL_N
  1427. export XGEMM3M_UNROLL_M
  1428. export XGEMM3M_UNROLL_N
  1429. ifdef USE_CUDA
  1430. export CUDADIR
  1431. export CUCC
  1432. export CUFLAGS
  1433. export CULIB
  1434. endif
  1435. .SUFFIXES: .$(PSUFFIX) .$(SUFFIX) .f
  1436. .f.$(SUFFIX):
  1437. $(FC) $(FFLAGS) -c $< -o $(@F)
  1438. .f.$(PSUFFIX):
  1439. $(FC) $(FPFLAGS) -pg -c $< -o $(@F)
  1440. ifdef BINARY64
  1441. PATHSCALEPATH = /opt/pathscale/lib/3.1
  1442. PGIPATH = /opt/pgi/linux86-64/7.1-5/lib
  1443. else
  1444. PATHSCALEPATH = /opt/pathscale/lib/3.1/32
  1445. PGIPATH = /opt/pgi/linux86/7.1-5/lib
  1446. endif
  1447. ACMLPATH = /opt/acml/4.3.0
  1448. ifneq ($(OSNAME), Darwin)
  1449. MKLPATH = /opt/intel/mkl/10.2.2.025/lib
  1450. else
  1451. MKLPATH = /Library/Frameworks/Intel_MKL.framework/Versions/10.0.1.014/lib
  1452. endif
  1453. ATLASPATH = /opt/atlas/3.9.17/opteron
  1454. FLAMEPATH = $(HOME)/flame/lib
  1455. ifneq ($(OSNAME), SunOS)
  1456. SUNPATH = /opt/sunstudio12.1
  1457. else
  1458. SUNPATH = /opt/SUNWspro
  1459. endif