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 31 kB

6 years ago
13 years ago
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942943944945946947948949950951952953954955956957958959960961962963964965966967968969970971972973974975976977978979980981982983984985986987988989990991992993994995996997998999100010011002100310041005100610071008100910101011101210131014101510161017101810191020102110221023102410251026102710281029103010311032103310341035103610371038103910401041104210431044104510461047104810491050105110521053105410551056105710581059106010611062106310641065106610671068106910701071107210731074107510761077107810791080108110821083108410851086108710881089109010911092109310941095109610971098109911001101110211031104110511061107110811091110111111121113111411151116111711181119112011211122112311241125112611271128112911301131113211331134113511361137113811391140114111421143114411451146114711481149115011511152115311541155115611571158115911601161116211631164116511661167116811691170117111721173117411751176117711781179118011811182118311841185118611871188118911901191119211931194119511961197119811991200120112021203120412051206120712081209121012111212121312141215121612171218121912201221122212231224122512261227122812291230123112321233123412351236123712381239124012411242124312441245124612471248124912501251125212531254125512561257125812591260126112621263126412651266126712681269127012711272127312741275127612771278127912801281128212831284128512861287128812891290129112921293129412951296129712981299130013011302130313041305130613071308130913101311131213131314131513161317131813191320132113221323132413251326132713281329133013311332133313341335133613371338133913401341134213431344134513461347134813491350135113521353135413551356135713581359136013611362136313641365136613671368136913701371137213731374137513761377137813791380138113821383138413851386138713881389139013911392139313941395139613971398139914001401140214031404140514061407140814091410141114121413141414151416141714181419142014211422142314241425142614271428142914301431143214331434143514361437143814391440144114421443144414451446144714481449145014511452145314541455145614571458145914601461146214631464146514661467146814691470147114721473147414751476147714781479148014811482148314841485148614871488148914901491149214931494149514961497149814991500150115021503150415051506150715081509151015111512151315141515151615171518151915201521152215231524152515261527152815291530153115321533153415351536153715381539154015411542154315441545154615471548154915501551155215531554155515561557155815591560156115621563156415651566156715681569157015711572157315741575157615771578157915801581158215831584158515861587
  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), powerpc)
  21. override ARCH=power
  22. else ifeq ($(ARCH), i386)
  23. override ARCH=x86
  24. else ifeq ($(ARCH), armv6)
  25. override ARCH=arm
  26. else ifeq ($(ARCH), armv7)
  27. override ARCH=arm
  28. else ifeq ($(ARCH), aarch64)
  29. override ARCH=arm64
  30. else ifeq ($(ARCH), zarch)
  31. override ARCH=zarch
  32. endif
  33. NETLIB_LAPACK_DIR = $(TOPDIR)/lapack-netlib
  34. # Default C compiler
  35. # - Only set if not specified on the command line or inherited from the environment.
  36. # - CC is an implicit variable so neither '?=' or 'ifndef' can be used.
  37. # http://stackoverflow.com/questions/4029274/mingw-and-make-variables
  38. # - Default value is 'cc' which is not always a valid command (e.g. MinGW).
  39. ifeq ($(origin CC),default)
  40. # Check if $(CC) refers to a valid command and set the value to gcc if not
  41. ifneq ($(findstring cmd.exe,$(SHELL)),)
  42. ifeq ($(shell where $(CC) 2>NUL),)
  43. CC = gcc
  44. endif
  45. else # POSIX-ish
  46. ifeq ($(shell command -v $(CC) 2>/dev/null),)
  47. ifeq ($(shell uname -s),Darwin)
  48. CC = clang
  49. # EXTRALIB += -Wl,-no_compact_unwind
  50. else
  51. CC = gcc
  52. endif # Darwin
  53. endif # CC exists
  54. endif # Shell is sane
  55. endif # CC is set to default
  56. # Default Fortran compiler (FC) is selected by f_check.
  57. ifndef MAKEFILE_RULE
  58. include $(TOPDIR)/Makefile.rule
  59. else
  60. include $(TOPDIR)/$(MAKEFILE_RULE)
  61. endif
  62. #
  63. # Beginning of system configuration
  64. #
  65. ifneq ($(BUILD_SINGLE),1)
  66. ifneq ($(BUILD_DOUBLE),1)
  67. ifneq ($(BUILD_COMPLEX),1)
  68. ifneq ($(BUILD_COMPLEX16),1)
  69. override BUILD_SINGLE=1
  70. override BUILD_DOUBLE=1
  71. override BUILD_COMPLEX=1
  72. override BUILD_COMPLEX16=1
  73. endif
  74. endif
  75. endif
  76. endif
  77. ifndef HOSTCC
  78. HOSTCC = $(CC)
  79. endif
  80. ifdef TARGET
  81. GETARCH_FLAGS := -DFORCE_$(TARGET)
  82. GETARCH_FLAGS += -DUSER_TARGET
  83. endif
  84. # Force fallbacks for 32bit
  85. ifeq ($(BINARY), 32)
  86. ifeq ($(TARGET), HASWELL)
  87. GETARCH_FLAGS := -DFORCE_NEHALEM
  88. endif
  89. ifeq ($(TARGET), SKYLAKEX)
  90. GETARCH_FLAGS := -DFORCE_NEHALEM
  91. endif
  92. ifeq ($(TARGET), COOPERLAKE)
  93. GETARCH_FLAGS := -DFORCE_NEHALEM
  94. endif
  95. ifeq ($(TARGET), SANDYBRIDGE)
  96. GETARCH_FLAGS := -DFORCE_NEHALEM
  97. endif
  98. ifeq ($(TARGET), BULLDOZER)
  99. GETARCH_FLAGS := -DFORCE_BARCELONA
  100. endif
  101. ifeq ($(TARGET), PILEDRIVER)
  102. GETARCH_FLAGS := -DFORCE_BARCELONA
  103. endif
  104. ifeq ($(TARGET), STEAMROLLER)
  105. GETARCH_FLAGS := -DFORCE_BARCELONA
  106. endif
  107. ifeq ($(TARGET), EXCAVATOR)
  108. GETARCH_FLAGS := -DFORCE_BARCELONA
  109. endif
  110. ifeq ($(TARGET), ZEN)
  111. GETARCH_FLAGS := -DFORCE_BARCELONA
  112. endif
  113. ifeq ($(TARGET), ARMV8)
  114. GETARCH_FLAGS := -DFORCE_ARMV7
  115. endif
  116. ifeq ($(TARGET), POWER8)
  117. GETARCH_FLAGS := -DFORCE_POWER6
  118. endif
  119. endif
  120. #TARGET_CORE will override TARGET which is used in DYNAMIC_ARCH=1.
  121. #
  122. ifdef TARGET_CORE
  123. GETARCH_FLAGS := -DFORCE_$(TARGET_CORE)
  124. endif
  125. # Force fallbacks for 32bit
  126. ifeq ($(BINARY), 32)
  127. ifeq ($(TARGET_CORE), HASWELL)
  128. GETARCH_FLAGS := -DFORCE_NEHALEM
  129. endif
  130. ifeq ($(TARGET_CORE), SKYLAKEX)
  131. GETARCH_FLAGS := -DFORCE_NEHALEM
  132. endif
  133. ifeq ($(TARGET_CORE), COOPERLAKE)
  134. GETARCH_FLAGS := -DFORCE_NEHALEM
  135. endif
  136. ifeq ($(TARGET_CORE), SANDYBRIDGE)
  137. GETARCH_FLAGS := -DFORCE_NEHALEM
  138. endif
  139. ifeq ($(TARGET_CORE), BULLDOZER)
  140. GETARCH_FLAGS := -DFORCE_BARCELONA
  141. endif
  142. ifeq ($(TARGET_CORE), PILEDRIVER)
  143. GETARCH_FLAGS := -DFORCE_BARCELONA
  144. endif
  145. ifeq ($(TARGET_CORE), STEAMROLLER)
  146. GETARCH_FLAGS := -DFORCE_BARCELONA
  147. endif
  148. ifeq ($(TARGET_CORE), EXCAVATOR)
  149. GETARCH_FLAGS := -DFORCE_BARCELONA
  150. endif
  151. ifeq ($(TARGET_CORE), ZEN)
  152. GETARCH_FLAGS := -DFORCE_BARCELONA
  153. endif
  154. endif
  155. # On x86_64 build getarch with march=native unless the compiler is PGI. This is required to detect AVX512 support in getarch.
  156. ifeq ($(HOSTARCH), x86_64)
  157. ifeq ($(findstring pgcc,$(HOSTCC)),)
  158. GETARCH_FLAGS += -march=native
  159. endif
  160. endif
  161. ifdef INTERFACE64
  162. ifneq ($(INTERFACE64), 0)
  163. GETARCH_FLAGS += -DUSE64BITINT
  164. endif
  165. endif
  166. ifndef GEMM_MULTITHREAD_THRESHOLD
  167. GEMM_MULTITHREAD_THRESHOLD=4
  168. endif
  169. GETARCH_FLAGS += -DGEMM_MULTITHREAD_THRESHOLD=$(GEMM_MULTITHREAD_THRESHOLD)
  170. ifeq ($(NO_AVX), 1)
  171. GETARCH_FLAGS += -DNO_AVX
  172. endif
  173. ifeq ($(BINARY), 32)
  174. GETARCH_FLAGS += -DNO_AVX -DNO_AVX2 -DNO_AVX512
  175. NO_AVX512 = 1
  176. endif
  177. ifeq ($(NO_AVX2), 1)
  178. GETARCH_FLAGS += -DNO_AVX2
  179. endif
  180. ifeq ($(NO_AVX512), 1)
  181. GETARCH_FLAGS += -DNO_AVX512
  182. endif
  183. ifeq ($(DEBUG), 1)
  184. GETARCH_FLAGS += -g
  185. endif
  186. ifeq ($(QUIET_MAKE), 1)
  187. MAKE += -s
  188. endif
  189. ifndef NO_PARALLEL_MAKE
  190. NO_PARALLEL_MAKE=0
  191. endif
  192. GETARCH_FLAGS += -DNO_PARALLEL_MAKE=$(NO_PARALLEL_MAKE)
  193. ifdef MAKE_NB_JOBS
  194. GETARCH_FLAGS += -DMAKE_NB_JOBS=$(MAKE_NB_JOBS)
  195. endif
  196. ifeq ($(HOSTCC), loongcc)
  197. GETARCH_FLAGS += -static
  198. endif
  199. #if don't use Fortran, it will only compile CBLAS.
  200. ifeq ($(ONLY_CBLAS), 1)
  201. NO_LAPACK = 1
  202. else
  203. ONLY_CBLAS = 0
  204. endif
  205. # This operation is expensive, so execution should be once.
  206. ifndef GOTOBLAS_MAKEFILE
  207. export GOTOBLAS_MAKEFILE = 1
  208. # Generating Makefile.conf and config.h
  209. 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)
  210. ifndef TARGET_CORE
  211. include $(TOPDIR)/Makefile.conf
  212. else
  213. include $(TOPDIR)/Makefile_kernel.conf
  214. endif
  215. endif
  216. ifndef NUM_PARALLEL
  217. NUM_PARALLEL = 1
  218. endif
  219. ifndef NUM_THREADS
  220. NUM_THREADS = $(NUM_CORES)
  221. endif
  222. ifeq ($(NUM_THREADS), 1)
  223. override USE_THREAD = 0
  224. override USE_OPENMP = 0
  225. endif
  226. ifdef USE_THREAD
  227. ifeq ($(USE_THREAD), 0)
  228. SMP =
  229. else
  230. SMP = 1
  231. endif
  232. else
  233. ifeq ($(NUM_THREAD), 1)
  234. SMP =
  235. else
  236. SMP = 1
  237. endif
  238. endif
  239. ifeq ($(SMP), 1)
  240. USE_LOCKING =
  241. endif
  242. ifndef NEED_PIC
  243. NEED_PIC = 1
  244. endif
  245. ARFLAGS =
  246. CPP = $(COMPILER) -E
  247. AR ?= $(CROSS_SUFFIX)ar
  248. AS ?= $(CROSS_SUFFIX)as
  249. LD ?= $(CROSS_SUFFIX)ld
  250. RANLIB ?= $(CROSS_SUFFIX)ranlib
  251. NM = $(CROSS_SUFFIX)nm
  252. DLLWRAP = $(CROSS_SUFFIX)dllwrap
  253. OBJCOPY = $(CROSS_SUFFIX)objcopy
  254. OBJCONV = $(CROSS_SUFFIX)objconv
  255. # When fortran support was either not detected or actively deselected, only build BLAS.
  256. ifeq ($(NOFORTRAN), 1)
  257. NO_LAPACK = 1
  258. override FEXTRALIB =
  259. endif
  260. ifeq ($(C_COMPILER), GCC)
  261. GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
  262. GCCVERSIONGT4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 4)
  263. GCCVERSIONGT5 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 5)
  264. GCCVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 7)
  265. GCCVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  266. GCCVERSIONGTEQ11 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 11)
  267. GCCVERSIONGTEQ10 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 10)
  268. # Note that the behavior of -dumpversion is compile-time-configurable for
  269. # gcc-7.x and newer. Use -dumpfullversion there
  270. ifeq ($(GCCVERSIONGTEQ7),1)
  271. GCCDUMPVERSION_PARAM := -dumpfullversion
  272. else
  273. GCCDUMPVERSION_PARAM := -dumpversion
  274. endif
  275. GCCMINORVERSIONGTEQ2 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 2)
  276. GCCMINORVERSIONGTEQ7 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 7)
  277. endif
  278. #
  279. # OS dependent settings
  280. #
  281. ifeq ($(OSNAME), Darwin)
  282. ifndef MACOSX_DEPLOYMENT_TARGET
  283. export MACOSX_DEPLOYMENT_TARGET=10.8
  284. endif
  285. MD5SUM = md5 -r
  286. endif
  287. ifneq (,$(findstring $(OSNAME), FreeBSD OpenBSD DragonFly))
  288. MD5SUM = md5 -r
  289. endif
  290. ifeq ($(OSNAME), NetBSD)
  291. MD5SUM = md5 -n
  292. endif
  293. ifeq ($(OSNAME), Linux)
  294. EXTRALIB += -lm
  295. NO_EXPRECISION = 1
  296. endif
  297. ifeq ($(OSNAME), Android)
  298. EXTRALIB += -lm
  299. endif
  300. ifeq ($(OSNAME), AIX)
  301. EXTRALIB += -lm
  302. endif
  303. ifeq ($(OSNAME), WINNT)
  304. NEED_PIC = 0
  305. NO_EXPRECISION = 1
  306. EXTRALIB += -defaultlib:advapi32
  307. SUFFIX = obj
  308. PSUFFIX = pobj
  309. LIBSUFFIX = a
  310. ifeq ($(C_COMPILER), CLANG)
  311. CCOMMON_OPT += -DMS_ABI
  312. endif
  313. #Version tests for supporting specific features (MS_ABI, POWER9 intrinsics)
  314. ifeq ($(GCCVERSIONGT4), 1)
  315. # GCC Major version > 4
  316. # It is compatible with MSVC ABI.
  317. CCOMMON_OPT += -DMS_ABI
  318. endif
  319. ifeq ($(GCCVERSIONGTEQ4), 1)
  320. ifeq ($(GCCMINORVERSIONGTEQ7), 1)
  321. # GCC Version >=4.7
  322. # It is compatible with MSVC ABI.
  323. CCOMMON_OPT += -DMS_ABI
  324. endif
  325. endif
  326. # Ensure the correct stack alignment on Win32
  327. # http://permalink.gmane.org/gmane.comp.lib.openblas.general/97
  328. ifeq ($(ARCH), x86)
  329. CCOMMON_OPT += -mincoming-stack-boundary=2
  330. FCOMMON_OPT += -mincoming-stack-boundary=2
  331. endif
  332. endif
  333. ifeq ($(OSNAME), Interix)
  334. NEED_PIC = 0
  335. NO_EXPRECISION = 1
  336. INTERIX_TOOL_DIR = /opt/gcc.3.3/i586-pc-interix3/bin
  337. endif
  338. ifeq ($(OSNAME), CYGWIN_NT)
  339. NEED_PIC = 0
  340. NO_EXPRECISION = 1
  341. OS_CYGWIN_NT = 1
  342. endif
  343. ifneq ($(OSNAME), WINNT)
  344. ifneq ($(OSNAME), CYGWIN_NT)
  345. ifneq ($(OSNAME), Interix)
  346. ifneq ($(OSNAME), Android)
  347. ifdef SMP
  348. EXTRALIB += -lpthread
  349. endif
  350. endif
  351. endif
  352. endif
  353. endif
  354. # ifeq logical or
  355. ifeq ($(OSNAME), $(filter $(OSNAME),WINNT CYGWIN_NT Interix))
  356. OS_WINDOWS=1
  357. endif
  358. ifdef QUAD_PRECISION
  359. CCOMMON_OPT += -DQUAD_PRECISION
  360. NO_EXPRECISION = 1
  361. endif
  362. ifneq ($(ARCH), x86)
  363. ifneq ($(ARCH), x86_64)
  364. NO_EXPRECISION = 1
  365. endif
  366. endif
  367. ifdef UTEST_CHECK
  368. CCOMMON_OPT += -DUTEST_CHECK
  369. SANITY_CHECK = 1
  370. endif
  371. ifdef SANITY_CHECK
  372. CCOMMON_OPT += -DSANITY_CHECK -DREFNAME=$(*F)f$(BU)
  373. endif
  374. MAX_STACK_ALLOC ?= 2048
  375. ifneq ($(MAX_STACK_ALLOC), 0)
  376. CCOMMON_OPT += -DMAX_STACK_ALLOC=$(MAX_STACK_ALLOC)
  377. endif
  378. ifdef USE_LOCKING
  379. ifneq ($(USE_LOCKING), 0)
  380. CCOMMON_OPT += -DUSE_LOCKING
  381. endif
  382. endif
  383. #
  384. # Architecture dependent settings
  385. #
  386. ifeq ($(ARCH), x86)
  387. ifndef BINARY
  388. NO_BINARY_MODE = 1
  389. endif
  390. ifeq ($(CORE), generic)
  391. NO_EXPRECISION = 1
  392. endif
  393. ifndef NO_EXPRECISION
  394. ifeq ($(F_COMPILER), GFORTRAN)
  395. # ifeq logical or. GCC or LSB
  396. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  397. EXPRECISION = 1
  398. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  399. FCOMMON_OPT += -m128bit-long-double
  400. endif
  401. ifeq ($(C_COMPILER), CLANG)
  402. EXPRECISION = 1
  403. CCOMMON_OPT += -DEXPRECISION
  404. FCOMMON_OPT += -m128bit-long-double
  405. endif
  406. endif
  407. endif
  408. endif
  409. ifeq ($(ARCH), x86_64)
  410. ifeq ($(CORE), generic)
  411. NO_EXPRECISION = 1
  412. endif
  413. ifndef NO_EXPRECISION
  414. ifeq ($(F_COMPILER), GFORTRAN)
  415. # ifeq logical or. GCC or LSB
  416. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  417. EXPRECISION = 1
  418. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  419. FCOMMON_OPT += -m128bit-long-double
  420. endif
  421. ifeq ($(C_COMPILER), CLANG)
  422. EXPRECISION = 1
  423. CCOMMON_OPT += -DEXPRECISION
  424. FCOMMON_OPT += -m128bit-long-double
  425. endif
  426. endif
  427. endif
  428. endif
  429. ifeq ($(C_COMPILER), INTEL)
  430. CCOMMON_OPT += -wd981
  431. endif
  432. ifeq ($(USE_OPENMP), 1)
  433. #check
  434. ifeq ($(USE_THREAD), 0)
  435. $(error OpenBLAS: Cannot set both USE_OPENMP=1 and USE_THREAD=0. The USE_THREAD=0 is only for building single thread version.)
  436. endif
  437. # ifeq logical or. GCC or LSB
  438. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  439. CCOMMON_OPT += -fopenmp
  440. endif
  441. ifeq ($(C_COMPILER), CLANG)
  442. CCOMMON_OPT += -fopenmp
  443. endif
  444. ifeq ($(C_COMPILER), INTEL)
  445. CCOMMON_OPT += -fopenmp
  446. endif
  447. ifeq ($(C_COMPILER), PGI)
  448. CCOMMON_OPT += -mp
  449. endif
  450. ifeq ($(C_COMPILER), OPEN64)
  451. CCOMMON_OPT += -mp
  452. CEXTRALIB += -lstdc++
  453. endif
  454. ifeq ($(C_COMPILER), PATHSCALE)
  455. CCOMMON_OPT += -mp
  456. endif
  457. endif
  458. ifeq ($(DYNAMIC_ARCH), 1)
  459. ifeq ($(ARCH), x86)
  460. DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
  461. CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
  462. endif
  463. ifeq ($(ARCH), x86_64)
  464. DYNAMIC_CORE = PRESCOTT CORE2
  465. ifeq ($(DYNAMIC_OLDER), 1)
  466. DYNAMIC_CORE += PENRYN DUNNINGTON
  467. endif
  468. DYNAMIC_CORE += NEHALEM
  469. ifeq ($(DYNAMIC_OLDER), 1)
  470. DYNAMIC_CORE += OPTERON OPTERON_SSE3
  471. endif
  472. DYNAMIC_CORE += BARCELONA
  473. ifeq ($(DYNAMIC_OLDER), 1)
  474. DYNAMIC_CORE += BOBCAT ATOM NANO
  475. endif
  476. ifneq ($(NO_AVX), 1)
  477. DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER STEAMROLLER EXCAVATOR
  478. endif
  479. ifneq ($(NO_AVX2), 1)
  480. DYNAMIC_CORE += HASWELL ZEN
  481. endif
  482. ifneq ($(NO_AVX512), 1)
  483. ifneq ($(NO_AVX2), 1)
  484. DYNAMIC_CORE += SKYLAKEX COOPERLAKE
  485. endif
  486. endif
  487. endif
  488. ifdef DYNAMIC_LIST
  489. override DYNAMIC_CORE = PRESCOTT $(DYNAMIC_LIST)
  490. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_PRESCOTT
  491. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  492. CCOMMON_OPT += $(XCCOMMON_OPT)
  493. #CCOMMON_OPT += -DDYNAMIC_LIST='$(DYNAMIC_LIST)'
  494. endif
  495. ifeq ($(ARCH), arm64)
  496. DYNAMIC_CORE = ARMV8
  497. DYNAMIC_CORE += CORTEXA53
  498. DYNAMIC_CORE += CORTEXA57
  499. DYNAMIC_CORE += CORTEXA72
  500. DYNAMIC_CORE += CORTEXA73
  501. DYNAMIC_CORE += NEOVERSEN1
  502. DYNAMIC_CORE += FALKOR
  503. DYNAMIC_CORE += THUNDERX
  504. DYNAMIC_CORE += THUNDERX2T99
  505. DYNAMIC_CORE += TSV110
  506. DYNAMIC_CORE += EMAG8180
  507. DYNAMIC_CORE += THUNDERX3T110
  508. endif
  509. ifeq ($(ARCH), zarch)
  510. DYNAMIC_CORE = ZARCH_GENERIC
  511. # if the compiler accepts -march=arch11 or -march=z13 and can compile a file
  512. # with z13-specific inline assembly, then we can include support for Z13.
  513. # note: -march=z13 is equivalent to -march=arch11 yet some compiler releases
  514. # only support one or the other.
  515. # note: LLVM version 6.x supported -march=z13 yet could not handle vector
  516. # registers in inline assembly, so the check for supporting the -march flag is
  517. # not enough.
  518. ZARCH_TEST_COMPILE=-c $(TOPDIR)/kernel/zarch/damin_z13.c -I$(TOPDIR) -o /dev/null > /dev/null 2> /dev/null
  519. ZARCH_CC_SUPPORTS_ARCH11=$(shell $(CC) -march=arch11 $(ZARCH_TEST_COMPILE) && echo 1)
  520. ZARCH_CC_SUPPORTS_Z13=$(shell $(CC) -march=z13 $(ZARCH_TEST_COMPILE) && echo 1)
  521. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH11), $(ZARCH_CC_SUPPORTS_Z13)), 1)
  522. DYNAMIC_CORE += Z13
  523. CCOMMON_OPT += -DDYN_Z13
  524. else
  525. $(info OpenBLAS: Not building Z13 kernels because the compiler $(CC) does not support it)
  526. endif
  527. # as above for z13, check for -march=arch12 and z14 support in the compiler.
  528. ZARCH_CC_SUPPORTS_ARCH12=$(shell $(CC) -march=arch12 $(ZARCH_TEST_COMPILE) && echo 1)
  529. ZARCH_CC_SUPPORTS_Z14=$(shell $(CC) -march=z14 $(ZARCH_TEST_COMPILE) && echo 1)
  530. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH12), $(ZARCH_CC_SUPPORTS_Z14)), 1)
  531. DYNAMIC_CORE += Z14
  532. CCOMMON_OPT += -DDYN_Z14
  533. else
  534. $(info OpenBLAS: Not building Z14 kernels because the compiler $(CC) does not support it)
  535. endif
  536. endif # ARCH zarch
  537. ifeq ($(ARCH), power)
  538. DYNAMIC_CORE = POWER6
  539. DYNAMIC_CORE += POWER8
  540. ifneq ($(C_COMPILER), GCC)
  541. DYNAMIC_CORE += POWER9
  542. DYNAMIC_CORE += POWER10
  543. endif
  544. ifeq ($(C_COMPILER), GCC)
  545. ifeq ($(GCCVERSIONGT5), 1)
  546. DYNAMIC_CORE += POWER9
  547. else
  548. $(info, OpenBLAS: Your gcc version is too old to build the POWER9 kernels.)
  549. endif
  550. ifeq ($(GCCVERSIONGTEQ11), 1)
  551. DYNAMIC_CORE += POWER10
  552. else ifeq ($(GCCVERSIONGTEQ10), 1)
  553. ifeq ($(GCCMINORVERSIONGTEQ2), 1)
  554. DYNAMIC_CORE += POWER10
  555. endif
  556. else
  557. $(info, OpenBLAS: Your gcc version is too old to build the POWER10 kernels.)
  558. endif
  559. endif
  560. endif
  561. # If DYNAMIC_CORE is not set, DYNAMIC_ARCH cannot do anything, so force it to empty
  562. ifndef DYNAMIC_CORE
  563. override DYNAMIC_ARCH=
  564. endif
  565. endif
  566. ifeq ($(ARCH), ia64)
  567. NO_BINARY_MODE = 1
  568. BINARY_DEFINED = 1
  569. ifeq ($(F_COMPILER), GFORTRAN)
  570. ifeq ($(C_COMPILER), GCC)
  571. # EXPRECISION = 1
  572. # CCOMMON_OPT += -DEXPRECISION
  573. endif
  574. endif
  575. endif
  576. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  577. NO_BINARY_MODE = 1
  578. endif
  579. ifeq ($(ARCH), alpha)
  580. NO_BINARY_MODE = 1
  581. BINARY_DEFINED = 1
  582. endif
  583. ifeq ($(ARCH), arm)
  584. NO_BINARY_MODE = 1
  585. BINARY_DEFINED = 1
  586. CCOMMON_OPT += -marm
  587. FCOMMON_OPT += -marm
  588. # If softfp abi is mentioned on the command line, force it.
  589. ifeq ($(ARM_SOFTFP_ABI), 1)
  590. CCOMMON_OPT += -mfloat-abi=softfp
  591. FCOMMON_OPT += -mfloat-abi=softfp
  592. endif
  593. ifeq ($(OSNAME), Android)
  594. ifeq ($(ARM_SOFTFP_ABI), 1)
  595. EXTRALIB += -lm
  596. else
  597. EXTRALIB += -Wl,-lm_hard
  598. endif
  599. endif
  600. endif
  601. ifeq ($(ARCH), arm64)
  602. NO_BINARY_MODE = 1
  603. BINARY_DEFINED = 1
  604. ifdef INTERFACE64
  605. ifneq ($(INTERFACE64), 0)
  606. ifeq ($(F_COMPILER), GFORTRAN)
  607. FCOMMON_OPT += -fdefault-integer-8
  608. endif
  609. ifeq ($(F_COMPILER), FLANG)
  610. FCOMMON_OPT += -i8
  611. endif
  612. endif
  613. endif
  614. endif
  615. ifeq ($(ARCH), riscv64)
  616. NO_BINARY_MODE = 1
  617. BINARY_DEFINED = 1
  618. endif
  619. #
  620. # C Compiler dependent settings
  621. #
  622. # ifeq logical or. GCC or CLANG or LSB
  623. # http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
  624. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG LSB))
  625. CCOMMON_OPT += -Wall
  626. COMMON_PROF += -fno-inline
  627. NO_UNINITIALIZED_WARN = -Wno-uninitialized
  628. ifeq ($(QUIET_MAKE), 1)
  629. CCOMMON_OPT += $(NO_UNINITIALIZED_WARN) -Wno-unused
  630. endif
  631. ifdef NO_BINARY_MODE
  632. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  633. ifdef BINARY64
  634. CCOMMON_OPT += -mabi=64
  635. else
  636. CCOMMON_OPT += -mabi=n32
  637. endif
  638. BINARY_DEFINED = 1
  639. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  640. CCOMMON_OPT += -mabi=32
  641. BINARY_DEFINED = 1
  642. endif
  643. ifeq ($(CORE), LOONGSON3A)
  644. CCOMMON_OPT += -march=mips64
  645. FCOMMON_OPT += -march=mips64
  646. endif
  647. ifeq ($(CORE), LOONGSON3B)
  648. CCOMMON_OPT += -march=mips64
  649. FCOMMON_OPT += -march=mips64
  650. endif
  651. ifeq ($(CORE), MIPS24K)
  652. CCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  653. FCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  654. endif
  655. ifeq ($(CORE), MIPS1004K)
  656. CCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  657. FCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  658. endif
  659. ifeq ($(CORE), P5600)
  660. CCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  661. FCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  662. endif
  663. ifeq ($(CORE), I6400)
  664. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  665. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  666. endif
  667. ifeq ($(CORE), P6600)
  668. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  669. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  670. endif
  671. ifeq ($(CORE), I6500)
  672. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  673. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  674. endif
  675. ifeq ($(OSNAME), AIX)
  676. BINARY_DEFINED = 1
  677. endif
  678. endif
  679. ifndef BINARY_DEFINED
  680. ifneq ($(OSNAME), AIX)
  681. ifdef BINARY64
  682. CCOMMON_OPT += -m64
  683. else
  684. CCOMMON_OPT += -m32
  685. endif
  686. endif
  687. endif
  688. endif
  689. ifeq ($(C_COMPILER), PGI)
  690. ifdef BINARY64
  691. ifeq ($(ARCH), x86_64)
  692. CCOMMON_OPT += -tp p7-64 -D__MMX__ -Mnollvm
  693. else
  694. ifeq ($(ARCH), power)
  695. ifeq ($(CORE), POWER8)
  696. CCOMMON_OPT += -tp pwr8
  697. endif
  698. ifeq ($(CORE), POWER9)
  699. CCOMMON_OPT += -tp pwr9
  700. endif
  701. endif
  702. endif
  703. else
  704. CCOMMON_OPT += -tp p7
  705. endif
  706. endif
  707. ifeq ($(C_COMPILER), PATHSCALE)
  708. ifdef BINARY64
  709. CCOMMON_OPT += -m64
  710. else
  711. CCOMMON_OPT += -m32
  712. endif
  713. endif
  714. #
  715. # Fortran Compiler dependent settings
  716. #
  717. ifeq ($(F_COMPILER), FLANG)
  718. CCOMMON_OPT += -DF_INTERFACE_FLANG
  719. FCOMMON_OPT += -Mrecursive -Kieee
  720. ifeq ($(OSNAME), Linux)
  721. ifeq ($(ARCH), x86_64)
  722. FLANG_VENDOR := $(shell expr `$(FC) --version|cut -f 1 -d "."|head -1`)
  723. ifeq ($(FLANG_VENDOR),AOCC)
  724. FCOMMON_OPT += -fno-unroll-loops
  725. endif
  726. endif
  727. endif
  728. ifdef BINARY64
  729. ifdef INTERFACE64
  730. ifneq ($(INTERFACE64), 0)
  731. FCOMMON_OPT += -i8
  732. endif
  733. endif
  734. FCOMMON_OPT += -Wall
  735. else
  736. FCOMMON_OPT += -Wall
  737. endif
  738. ifeq ($(USE_OPENMP), 1)
  739. FCOMMON_OPT += -fopenmp
  740. endif
  741. endif
  742. ifeq ($(F_COMPILER), G77)
  743. CCOMMON_OPT += -DF_INTERFACE_G77
  744. FCOMMON_OPT += -Wall
  745. ifndef NO_BINARY_MODE
  746. ifneq ($(OSNAME), AIX)
  747. ifdef BINARY64
  748. FCOMMON_OPT += -m64
  749. else
  750. FCOMMON_OPT += -m32
  751. endif
  752. endif
  753. endif
  754. endif
  755. ifeq ($(F_COMPILER), G95)
  756. CCOMMON_OPT += -DF_INTERFACE_G95
  757. FCOMMON_OPT += -Wall
  758. ifneq ($(OSNAME), AIX)
  759. ifndef NO_BINARY_MODE
  760. ifdef BINARY64
  761. FCOMMON_OPT += -m64
  762. else
  763. FCOMMON_OPT += -m32
  764. endif
  765. endif
  766. ifneq ($(NO_LAPACKE), 1)
  767. FCOMMON_OPT += -fno-second-underscore
  768. endif
  769. endif
  770. endif
  771. ifeq ($(F_COMPILER), GFORTRAN)
  772. CCOMMON_OPT += -DF_INTERFACE_GFORT
  773. FCOMMON_OPT += -Wall
  774. # make single-threaded LAPACK calls thread-safe #1847
  775. FCOMMON_OPT += -frecursive
  776. # work around ABI problem with passing single-character arguments
  777. FCOMMON_OPT += -fno-optimize-sibling-calls
  778. #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
  779. ifneq ($(NO_LAPACK), 1)
  780. EXTRALIB += -lgfortran
  781. endif
  782. ifdef NO_BINARY_MODE
  783. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  784. ifdef BINARY64
  785. FCOMMON_OPT += -mabi=64
  786. else
  787. FCOMMON_OPT += -mabi=n32
  788. endif
  789. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  790. FCOMMON_OPT += -mabi=32
  791. endif
  792. else
  793. ifdef BINARY64
  794. ifneq ($(OSNAME), AIX)
  795. FCOMMON_OPT += -m64
  796. endif
  797. ifdef INTERFACE64
  798. ifneq ($(INTERFACE64), 0)
  799. FCOMMON_OPT += -fdefault-integer-8
  800. endif
  801. endif
  802. else
  803. ifneq ($(OSNAME), AIX)
  804. FCOMMON_OPT += -m32
  805. endif
  806. endif
  807. endif
  808. ifeq ($(USE_OPENMP), 1)
  809. FCOMMON_OPT += -fopenmp
  810. endif
  811. endif
  812. ifeq ($(F_COMPILER), INTEL)
  813. CCOMMON_OPT += -DF_INTERFACE_INTEL
  814. ifdef INTERFACE64
  815. ifneq ($(INTERFACE64), 0)
  816. FCOMMON_OPT += -i8
  817. endif
  818. endif
  819. FCOMMON_OPT += -recursive -fp-model strict -assume protect-parens
  820. ifeq ($(USE_OPENMP), 1)
  821. FCOMMON_OPT += -fopenmp
  822. endif
  823. endif
  824. ifeq ($(F_COMPILER), FUJITSU)
  825. CCOMMON_OPT += -DF_INTERFACE_FUJITSU
  826. ifeq ($(USE_OPENMP), 1)
  827. FCOMMON_OPT += -openmp
  828. endif
  829. endif
  830. ifeq ($(F_COMPILER), IBM)
  831. CCOMMON_OPT += -DF_INTERFACE_IBM
  832. # FCOMMON_OPT += -qarch=440
  833. ifdef BINARY64
  834. FCOMMON_OPT += -q64
  835. ifdef INTERFACE64
  836. ifneq ($(INTERFACE64), 0)
  837. FCOMMON_OPT += -qintsize=8
  838. endif
  839. endif
  840. else
  841. FCOMMON_OPT += -q32
  842. endif
  843. ifeq ($(USE_OPENMP), 1)
  844. FCOMMON_OPT += -openmp
  845. endif
  846. endif
  847. ifeq ($(F_COMPILER), PGI)
  848. CCOMMON_OPT += -DF_INTERFACE_PGI
  849. COMMON_PROF += -DPGICOMPILER
  850. ifdef BINARY64
  851. ifdef INTERFACE64
  852. ifneq ($(INTERFACE64), 0)
  853. FCOMMON_OPT += -i8
  854. endif
  855. endif
  856. ifeq ($(ARCH), x86_64)
  857. FCOMMON_OPT += -tp p7-64
  858. else
  859. ifeq ($(ARCH), power)
  860. ifeq ($(CORE), POWER8)
  861. FCOMMON_OPT += -tp pwr8
  862. endif
  863. ifeq ($(CORE), POWER9)
  864. FCOMMON_OPT += -tp pwr9
  865. endif
  866. endif
  867. endif
  868. else
  869. FCOMMON_OPT += -tp p7
  870. endif
  871. FCOMMON_OPT += -Mrecursive
  872. ifeq ($(USE_OPENMP), 1)
  873. FCOMMON_OPT += -mp
  874. endif
  875. endif
  876. ifeq ($(F_COMPILER), PATHSCALE)
  877. CCOMMON_OPT += -DF_INTERFACE_PATHSCALE
  878. ifdef BINARY64
  879. ifdef INTERFACE64
  880. ifneq ($(INTERFACE64), 0)
  881. FCOMMON_OPT += -i8
  882. endif
  883. endif
  884. endif
  885. ifeq ($(USE_OPENMP), 1)
  886. FCOMMON_OPT += -mp
  887. endif
  888. endif
  889. ifeq ($(F_COMPILER), OPEN64)
  890. CCOMMON_OPT += -DF_INTERFACE_OPEN64
  891. ifdef BINARY64
  892. ifdef INTERFACE64
  893. ifneq ($(INTERFACE64), 0)
  894. FCOMMON_OPT += -i8
  895. endif
  896. endif
  897. endif
  898. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  899. ifndef BINARY64
  900. FCOMMON_OPT += -n32
  901. else
  902. FCOMMON_OPT += -n64
  903. endif
  904. ifeq ($(CORE), LOONGSON3A)
  905. FCOMMON_OPT += -loongson3 -static
  906. endif
  907. ifeq ($(CORE), LOONGSON3B)
  908. FCOMMON_OPT += -loongson3 -static
  909. endif
  910. else
  911. ifndef BINARY64
  912. FCOMMON_OPT += -m32
  913. else
  914. FCOMMON_OPT += -m64
  915. endif
  916. endif
  917. ifeq ($(USE_OPENMP), 1)
  918. FEXTRALIB += -lstdc++
  919. FCOMMON_OPT += -mp
  920. endif
  921. endif
  922. ifeq ($(C_COMPILER), OPEN64)
  923. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  924. ifndef BINARY64
  925. CCOMMON_OPT += -n32
  926. else
  927. CCOMMON_OPT += -n64
  928. endif
  929. ifeq ($(CORE), LOONGSON3A)
  930. CCOMMON_OPT += -loongson3 -static
  931. endif
  932. ifeq ($(CORE), LOONGSON3B)
  933. CCOMMON_OPT += -loongson3 -static
  934. endif
  935. else
  936. ifndef BINARY64
  937. CCOMMON_OPT += -m32
  938. else
  939. CCOMMON_OPT += -m64
  940. endif
  941. endif
  942. endif
  943. ifeq ($(C_COMPILER), SUN)
  944. CCOMMON_OPT += -w
  945. ifeq ($(ARCH), x86)
  946. CCOMMON_OPT += -m32
  947. else
  948. FCOMMON_OPT += -m64
  949. endif
  950. endif
  951. ifeq ($(F_COMPILER), SUN)
  952. CCOMMON_OPT += -DF_INTERFACE_SUN
  953. ifeq ($(ARCH), x86)
  954. FCOMMON_OPT += -m32
  955. else
  956. FCOMMON_OPT += -m64
  957. endif
  958. ifeq ($(USE_OPENMP), 1)
  959. FCOMMON_OPT += -xopenmp=parallel
  960. endif
  961. endif
  962. ifeq ($(F_COMPILER), COMPAQ)
  963. CCOMMON_OPT += -DF_INTERFACE_COMPAQ
  964. ifeq ($(USE_OPENMP), 1)
  965. FCOMMON_OPT += -openmp
  966. endif
  967. endif
  968. ifdef BINARY64
  969. ifdef INTERFACE64
  970. ifneq ($(INTERFACE64), 0)
  971. CCOMMON_OPT +=
  972. #-DUSE64BITINT
  973. endif
  974. endif
  975. endif
  976. ifeq ($(NEED_PIC), 1)
  977. ifeq ($(C_COMPILER), IBM)
  978. CCOMMON_OPT += -qpic=large
  979. else
  980. CCOMMON_OPT += -fPIC
  981. endif
  982. ifeq ($(F_COMPILER), SUN)
  983. FCOMMON_OPT += -pic
  984. else
  985. FCOMMON_OPT += -fPIC
  986. endif
  987. endif
  988. ifeq ($(DYNAMIC_ARCH), 1)
  989. CCOMMON_OPT += -DDYNAMIC_ARCH
  990. endif
  991. ifeq ($(DYNAMIC_OLDER), 1)
  992. CCOMMON_OPT += -DDYNAMIC_OLDER
  993. endif
  994. ifeq ($(NO_LAPACK), 1)
  995. CCOMMON_OPT += -DNO_LAPACK
  996. #Disable LAPACK C interface
  997. NO_LAPACKE = 1
  998. endif
  999. ifeq ($(NO_LAPACKE), 1)
  1000. CCOMMON_OPT += -DNO_LAPACKE
  1001. endif
  1002. ifeq ($(NO_AVX), 1)
  1003. CCOMMON_OPT += -DNO_AVX
  1004. endif
  1005. ifeq ($(ARCH), x86)
  1006. CCOMMON_OPT += -DNO_AVX
  1007. endif
  1008. ifeq ($(NO_AVX2), 1)
  1009. CCOMMON_OPT += -DNO_AVX2
  1010. endif
  1011. ifeq ($(NO_AVX512), 1)
  1012. CCOMMON_OPT += -DNO_AVX512
  1013. endif
  1014. ifdef SMP
  1015. CCOMMON_OPT += -DSMP_SERVER
  1016. ifeq ($(ARCH), mips64)
  1017. ifneq ($(CORE), LOONGSON3B)
  1018. USE_SIMPLE_THREADED_LEVEL3 = 1
  1019. endif
  1020. endif
  1021. ifeq ($(USE_OPENMP), 1)
  1022. # USE_SIMPLE_THREADED_LEVEL3 = 1
  1023. # NO_AFFINITY = 1
  1024. CCOMMON_OPT += -DUSE_OPENMP
  1025. endif
  1026. ifeq ($(BIGNUMA), 1)
  1027. CCOMMON_OPT += -DBIGNUMA
  1028. endif
  1029. endif
  1030. ifeq ($(NO_WARMUP), 1)
  1031. CCOMMON_OPT += -DNO_WARMUP
  1032. endif
  1033. ifeq ($(CONSISTENT_FPCSR), 1)
  1034. CCOMMON_OPT += -DCONSISTENT_FPCSR
  1035. endif
  1036. # Only for development
  1037. # CCOMMON_OPT += -DPARAMTEST
  1038. # CCOMMON_OPT += -DPREFETCHTEST
  1039. # CCOMMON_OPT += -DNO_SWITCHING
  1040. # USE_PAPI = 1
  1041. ifdef USE_PAPI
  1042. CCOMMON_OPT += -DUSE_PAPI
  1043. EXTRALIB += -lpapi -lperfctr
  1044. endif
  1045. ifdef DYNAMIC_THREADS
  1046. CCOMMON_OPT += -DDYNAMIC_THREADS
  1047. endif
  1048. CCOMMON_OPT += -DMAX_CPU_NUMBER=$(NUM_THREADS)
  1049. CCOMMON_OPT += -DMAX_PARALLEL_NUMBER=$(NUM_PARALLEL)
  1050. ifdef USE_SIMPLE_THREADED_LEVEL3
  1051. CCOMMON_OPT += -DUSE_SIMPLE_THREADED_LEVEL3
  1052. endif
  1053. ifeq ($(USE_TLS), 1)
  1054. CCOMMON_OPT += -DUSE_TLS
  1055. endif
  1056. ifeq ($(BUILD_BFLOAT16), 1)
  1057. CCOMMON_OPT += -DBUILD_BFLOAT16
  1058. endif
  1059. ifeq ($(BUILD_SINGLE), 1)
  1060. CCOMMON_OPT += -DBUILD_SINGLE=1
  1061. endif
  1062. ifeq ($(BUILD_DOUBLE), 1)
  1063. CCOMMON_OPT += -DBUILD_DOUBLE=1
  1064. endif
  1065. ifeq ($(BUILD_COMPLEX), 1)
  1066. CCOMMON_OPT += -DBUILD_COMPLEX=1
  1067. endif
  1068. ifeq ($(BUILD_COMPLEX16), 1)
  1069. CCOMMON_OPT += -DBUILD_COMPLEX16=1
  1070. endif
  1071. CCOMMON_OPT += -DVERSION=\"$(VERSION)\"
  1072. ifndef SYMBOLPREFIX
  1073. SYMBOLPREFIX =
  1074. endif
  1075. ifndef SYMBOLSUFFIX
  1076. SYMBOLSUFFIX =
  1077. endif
  1078. ifndef LIBNAMESUFFIX
  1079. LIBNAMEBASE = $(SYMBOLPREFIX)openblas$(SYMBOLSUFFIX)
  1080. else
  1081. LIBNAMEBASE = $(SYMBOLPREFIX)openblas$(SYMBOLSUFFIX)_$(LIBNAMESUFFIX)
  1082. endif
  1083. ifeq ($(OSNAME), CYGWIN_NT)
  1084. LIBPREFIX = cyg$(LIBNAMEBASE)
  1085. else
  1086. LIBPREFIX = lib$(LIBNAMEBASE)
  1087. endif
  1088. KERNELDIR = $(TOPDIR)/kernel/$(ARCH)
  1089. include $(TOPDIR)/Makefile.$(ARCH)
  1090. ifneq ($(C_COMPILER), PGI)
  1091. CCOMMON_OPT += -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME
  1092. endif
  1093. CCOMMON_OPT += -DASMNAME=$(FU)$(*F) -DASMFNAME=$(FU)$(*F)$(BU) -DNAME=$(*F)$(BU) -DCNAME=$(*F) -DCHAR_NAME=\"$(*F)$(BU)\" -DCHAR_CNAME=\"$(*F)\"
  1094. ifeq ($(CORE), PPC440)
  1095. CCOMMON_OPT += -DALLOC_QALLOC
  1096. endif
  1097. ifeq ($(CORE), PPC440FP2)
  1098. STATIC_ALLOCATION = 1
  1099. endif
  1100. ifneq ($(OSNAME), Linux)
  1101. NO_AFFINITY = 1
  1102. endif
  1103. ifneq ($(ARCH), x86_64)
  1104. ifneq ($(ARCH), x86)
  1105. ifneq ($(CORE), LOONGSON3B)
  1106. NO_AFFINITY = 1
  1107. endif
  1108. endif
  1109. endif
  1110. ifdef NO_AFFINITY
  1111. ifeq ($(NO_AFFINITY), 0)
  1112. override undefine NO_AFFINITY
  1113. else
  1114. CCOMMON_OPT += -DNO_AFFINITY
  1115. endif
  1116. endif
  1117. ifdef FUNCTION_PROFILE
  1118. CCOMMON_OPT += -DFUNCTION_PROFILE
  1119. endif
  1120. ifdef HUGETLB_ALLOCATION
  1121. CCOMMON_OPT += -DALLOC_HUGETLB
  1122. endif
  1123. ifdef HUGETLBFILE_ALLOCATION
  1124. CCOMMON_OPT += -DALLOC_HUGETLBFILE -DHUGETLB_FILE_NAME=$(HUGETLBFILE_ALLOCATION)
  1125. endif
  1126. ifdef STATIC_ALLOCATION
  1127. CCOMMON_OPT += -DALLOC_STATIC
  1128. endif
  1129. ifdef DEVICEDRIVER_ALLOCATION
  1130. CCOMMON_OPT += -DALLOC_DEVICEDRIVER -DDEVICEDRIVER_NAME=\"/dev/mapper\"
  1131. endif
  1132. ifdef MIXED_MEMORY_ALLOCATION
  1133. CCOMMON_OPT += -DMIXED_MEMORY_ALLOCATION
  1134. endif
  1135. ifeq ($(OSNAME), SunOS)
  1136. TAR = gtar
  1137. PATCH = gpatch
  1138. GREP = ggrep
  1139. AWK = nawk
  1140. else
  1141. TAR = tar
  1142. PATCH = patch
  1143. GREP = grep
  1144. AWK = awk
  1145. endif
  1146. ifndef MD5SUM
  1147. MD5SUM = md5sum
  1148. endif
  1149. REVISION = -r$(VERSION)
  1150. MAJOR_VERSION = $(word 1,$(subst ., ,$(VERSION)))
  1151. ifeq ($(DEBUG), 1)
  1152. COMMON_OPT += -g
  1153. endif
  1154. ifeq ($(DEBUG), 1)
  1155. FCOMMON_OPT += -g
  1156. endif
  1157. ifndef COMMON_OPT
  1158. COMMON_OPT = -O2
  1159. endif
  1160. ifndef FCOMMON_OPT
  1161. FCOMMON_OPT = -O2 -frecursive
  1162. endif
  1163. override CFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
  1164. override PFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
  1165. override FFLAGS += $(COMMON_OPT) $(FCOMMON_OPT)
  1166. override FPFLAGS += $(FCOMMON_OPT) $(COMMON_PROF)
  1167. #MAKEOVERRIDES =
  1168. ifdef NEED_PIC
  1169. ifeq (,$(findstring PIC,$(FFLAGS)))
  1170. override FFLAGS += -fPIC
  1171. endif
  1172. endif
  1173. #For LAPACK Fortran codes.
  1174. #Disable -fopenmp for LAPACK Fortran codes on Windows.
  1175. ifdef OS_WINDOWS
  1176. LAPACK_FFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FFLAGS))
  1177. LAPACK_FPFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FPFLAGS))
  1178. else
  1179. LAPACK_FFLAGS := $(FFLAGS)
  1180. LAPACK_FPFLAGS := $(FPFLAGS)
  1181. endif
  1182. LAPACK_CFLAGS = $(CFLAGS)
  1183. LAPACK_CFLAGS += -DHAVE_LAPACK_CONFIG_H
  1184. ifdef INTERFACE64
  1185. ifneq ($(INTERFACE64), 0)
  1186. LAPACK_CFLAGS += -DLAPACK_ILP64
  1187. endif
  1188. endif
  1189. ifdef OS_WINDOWS
  1190. LAPACK_CFLAGS += -DOPENBLAS_OS_WINDOWS
  1191. endif
  1192. ifeq ($(C_COMPILER), LSB)
  1193. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1194. endif
  1195. ifndef SUFFIX
  1196. SUFFIX = o
  1197. endif
  1198. ifndef PSUFFIX
  1199. PSUFFIX = po
  1200. endif
  1201. ifndef LIBSUFFIX
  1202. LIBSUFFIX = a
  1203. endif
  1204. ifneq ($(DYNAMIC_ARCH), 1)
  1205. ifndef SMP
  1206. LIBNAME = $(LIBPREFIX)_$(LIBCORE)$(REVISION).$(LIBSUFFIX)
  1207. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)$(REVISION)_p.$(LIBSUFFIX)
  1208. else
  1209. LIBNAME = $(LIBPREFIX)_$(LIBCORE)p$(REVISION).$(LIBSUFFIX)
  1210. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)p$(REVISION)_p.$(LIBSUFFIX)
  1211. endif
  1212. else
  1213. ifndef SMP
  1214. LIBNAME = $(LIBPREFIX)$(REVISION).$(LIBSUFFIX)
  1215. LIBNAME_P = $(LIBPREFIX)$(REVISION)_p.$(LIBSUFFIX)
  1216. else
  1217. LIBNAME = $(LIBPREFIX)p$(REVISION).$(LIBSUFFIX)
  1218. LIBNAME_P = $(LIBPREFIX)p$(REVISION)_p.$(LIBSUFFIX)
  1219. endif
  1220. endif
  1221. LIBDLLNAME = $(LIBPREFIX).dll
  1222. IMPLIBNAME = lib$(LIBNAMEBASE).dll.a
  1223. ifneq ($(OSNAME), AIX)
  1224. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.so)
  1225. else
  1226. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.a)
  1227. endif
  1228. LIBDYNNAME = $(LIBNAME:.$(LIBSUFFIX)=.dylib)
  1229. LIBDEFNAME = $(LIBNAME:.$(LIBSUFFIX)=.def)
  1230. LIBEXPNAME = $(LIBNAME:.$(LIBSUFFIX)=.exp)
  1231. LIBZIPNAME = $(LIBNAME:.$(LIBSUFFIX)=.zip)
  1232. LIBS = $(TOPDIR)/$(LIBNAME)
  1233. LIBS_P = $(TOPDIR)/$(LIBNAME_P)
  1234. LIB_COMPONENTS = BLAS
  1235. ifneq ($(NO_CBLAS), 1)
  1236. LIB_COMPONENTS += CBLAS
  1237. endif
  1238. ifneq ($(NO_LAPACK), 1)
  1239. LIB_COMPONENTS += LAPACK
  1240. ifneq ($(NO_LAPACKE), 1)
  1241. LIB_COMPONENTS += LAPACKE
  1242. endif
  1243. ifeq ($(BUILD_RELAPACK), 1)
  1244. LIB_COMPONENTS += ReLAPACK
  1245. endif
  1246. endif
  1247. ifeq ($(ONLY_CBLAS), 1)
  1248. LIB_COMPONENTS = CBLAS
  1249. endif
  1250. export OSNAME
  1251. export ARCH
  1252. export CORE
  1253. export LIBCORE
  1254. export __BYTE_ORDER__
  1255. export ELF_VERSION
  1256. export PGCPATH
  1257. export CONFIG
  1258. export CC
  1259. export FC
  1260. export BU
  1261. export FU
  1262. export NEED2UNDERSCORES
  1263. export USE_THREAD
  1264. export NUM_THREADS
  1265. export NUM_CORES
  1266. export SMP
  1267. export MAKEFILE_RULE
  1268. export NEED_PIC
  1269. export BINARY
  1270. export BINARY32
  1271. export BINARY64
  1272. export F_COMPILER
  1273. export C_COMPILER
  1274. export USE_OPENMP
  1275. export CROSS
  1276. export CROSS_SUFFIX
  1277. export NOFORTRAN
  1278. export NO_FBLAS
  1279. export EXTRALIB
  1280. export CEXTRALIB
  1281. export FEXTRALIB
  1282. export HAVE_SSE
  1283. export HAVE_SSE2
  1284. export HAVE_SSE3
  1285. export HAVE_SSSE3
  1286. export HAVE_SSE4_1
  1287. export HAVE_SSE4_2
  1288. export HAVE_SSE4A
  1289. export HAVE_SSE5
  1290. export HAVE_AVX
  1291. export HAVE_VFP
  1292. export HAVE_VFPV3
  1293. export HAVE_VFPV4
  1294. export HAVE_NEON
  1295. export HAVE_MSA
  1296. export MSA_FLAGS
  1297. export KERNELDIR
  1298. export FUNCTION_PROFILE
  1299. export TARGET_CORE
  1300. export NO_AVX512
  1301. export BUILD_BFLOAT16
  1302. export SBGEMM_UNROLL_M
  1303. export SBGEMM_UNROLL_N
  1304. export SGEMM_UNROLL_M
  1305. export SGEMM_UNROLL_N
  1306. export DGEMM_UNROLL_M
  1307. export DGEMM_UNROLL_N
  1308. export QGEMM_UNROLL_M
  1309. export QGEMM_UNROLL_N
  1310. export CGEMM_UNROLL_M
  1311. export CGEMM_UNROLL_N
  1312. export ZGEMM_UNROLL_M
  1313. export ZGEMM_UNROLL_N
  1314. export XGEMM_UNROLL_M
  1315. export XGEMM_UNROLL_N
  1316. export CGEMM3M_UNROLL_M
  1317. export CGEMM3M_UNROLL_N
  1318. export ZGEMM3M_UNROLL_M
  1319. export ZGEMM3M_UNROLL_N
  1320. export XGEMM3M_UNROLL_M
  1321. export XGEMM3M_UNROLL_N
  1322. ifdef USE_CUDA
  1323. export CUDADIR
  1324. export CUCC
  1325. export CUFLAGS
  1326. export CULIB
  1327. endif
  1328. .SUFFIXES: .$(PSUFFIX) .$(SUFFIX) .f
  1329. .f.$(SUFFIX):
  1330. $(FC) $(FFLAGS) -c $< -o $(@F)
  1331. .f.$(PSUFFIX):
  1332. $(FC) $(FPFLAGS) -pg -c $< -o $(@F)
  1333. ifdef BINARY64
  1334. PATHSCALEPATH = /opt/pathscale/lib/3.1
  1335. PGIPATH = /opt/pgi/linux86-64/7.1-5/lib
  1336. else
  1337. PATHSCALEPATH = /opt/pathscale/lib/3.1/32
  1338. PGIPATH = /opt/pgi/linux86/7.1-5/lib
  1339. endif
  1340. ACMLPATH = /opt/acml/4.3.0
  1341. ifneq ($(OSNAME), Darwin)
  1342. MKLPATH = /opt/intel/mkl/10.2.2.025/lib
  1343. else
  1344. MKLPATH = /Library/Frameworks/Intel_MKL.framework/Versions/10.0.1.014/lib
  1345. endif
  1346. ATLASPATH = /opt/atlas/3.9.17/opteron
  1347. FLAMEPATH = $(HOME)/flame/lib
  1348. ifneq ($(OSNAME), SunOS)
  1349. SUNPATH = /opt/sunstudio12.1
  1350. else
  1351. SUNPATH = /opt/SUNWspro
  1352. endif