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 38 kB

7 years ago
4 years ago
13 years ago
4 years ago
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942943944945946947948949950951952953954955956957958959960961962963964965966967968969970971972973974975976977978979980981982983984985986987988989990991992993994995996997998999100010011002100310041005100610071008100910101011101210131014101510161017101810191020102110221023102410251026102710281029103010311032103310341035103610371038103910401041104210431044104510461047104810491050105110521053105410551056105710581059106010611062106310641065106610671068106910701071107210731074107510761077107810791080108110821083108410851086108710881089109010911092109310941095109610971098109911001101110211031104110511061107110811091110111111121113111411151116111711181119112011211122112311241125112611271128112911301131113211331134113511361137113811391140114111421143114411451146114711481149115011511152115311541155115611571158115911601161116211631164116511661167116811691170117111721173117411751176117711781179118011811182118311841185118611871188118911901191119211931194119511961197119811991200120112021203120412051206120712081209121012111212121312141215121612171218121912201221122212231224122512261227122812291230123112321233123412351236123712381239124012411242124312441245124612471248124912501251125212531254125512561257125812591260126112621263126412651266126712681269127012711272127312741275127612771278127912801281128212831284128512861287128812891290129112921293129412951296129712981299130013011302130313041305130613071308130913101311131213131314131513161317131813191320132113221323132413251326132713281329133013311332133313341335133613371338133913401341134213431344134513461347134813491350135113521353135413551356135713581359136013611362136313641365136613671368136913701371137213731374137513761377137813791380138113821383138413851386138713881389139013911392139313941395139613971398139914001401140214031404140514061407140814091410141114121413141414151416141714181419142014211422142314241425142614271428142914301431143214331434143514361437143814391440144114421443144414451446144714481449145014511452145314541455145614571458145914601461146214631464146514661467146814691470147114721473147414751476147714781479148014811482148314841485148614871488148914901491149214931494149514961497149814991500150115021503150415051506150715081509151015111512151315141515151615171518151915201521152215231524152515261527152815291530153115321533153415351536153715381539154015411542154315441545154615471548154915501551155215531554155515561557155815591560156115621563156415651566156715681569157015711572157315741575157615771578157915801581158215831584158515861587158815891590159115921593159415951596159715981599160016011602160316041605160616071608160916101611161216131614161516161617161816191620162116221623162416251626162716281629163016311632163316341635163616371638163916401641164216431644164516461647164816491650165116521653165416551656165716581659166016611662166316641665166616671668166916701671167216731674167516761677167816791680168116821683168416851686168716881689169016911692169316941695169616971698169917001701170217031704170517061707170817091710171117121713171417151716171717181719172017211722172317241725172617271728172917301731173217331734173517361737173817391740174117421743174417451746174717481749175017511752175317541755175617571758175917601761176217631764176517661767176817691770177117721773177417751776177717781779178017811782178317841785178617871788178917901791179217931794179517961797179817991800180118021803180418051806180718081809181018111812181318141815181618171818181918201821182218231824182518261827182818291830183118321833183418351836183718381839184018411842184318441845184618471848184918501851185218531854185518561857
  1. #
  2. # Include user definition
  3. #
  4. # TO suppress recursive includes
  5. INCLUDED = 1
  6. ifndef TOPDIR
  7. TOPDIR = .
  8. endif
  9. ifndef RELAPACK_REPLACE
  10. RELAPACK_REPLACE=0
  11. endif
  12. # we need to use the host system's architecture for getarch compile options even especially when cross-compiling
  13. HOSTARCH := $(shell uname -m)
  14. ifeq ($(HOSTARCH), amd64)
  15. HOSTARCH=x86_64
  16. endif
  17. # Catch conflicting usage of ARCH in some BSD environments
  18. ifeq ($(ARCH), amd64)
  19. override ARCH=x86_64
  20. else ifeq ($(ARCH), powerpc64)
  21. override ARCH=power
  22. else ifeq ($(ARCH), powerpc64le)
  23. override ARCH=power
  24. else ifeq ($(ARCH), powerpc)
  25. override ARCH=power
  26. else ifeq ($(ARCH), i386)
  27. override ARCH=x86
  28. else ifeq ($(ARCH), armv6)
  29. override ARCH=arm
  30. else ifeq ($(ARCH), armv7)
  31. override ARCH=arm
  32. else ifeq ($(ARCH), aarch64)
  33. override ARCH=arm64
  34. else ifeq ($(ARCH), mipsel)
  35. override ARCH=mips
  36. else ifeq ($(ARCH), mips64el)
  37. override ARCH=mips64
  38. else ifeq ($(ARCH), zarch)
  39. override ARCH=zarch
  40. endif
  41. NETLIB_LAPACK_DIR = $(TOPDIR)/lapack-netlib
  42. # Default C compiler
  43. # - Only set if not specified on the command line or inherited from the environment.
  44. # - CC is an implicit variable so neither '?=' or 'ifndef' can be used.
  45. # http://stackoverflow.com/questions/4029274/mingw-and-make-variables
  46. # - Default value is 'cc' which is not always a valid command (e.g. MinGW).
  47. ifeq ($(origin CC),default)
  48. # Check if $(CC) refers to a valid command and set the value to gcc if not
  49. ifneq ($(findstring cmd.exe,$(SHELL)),)
  50. ifeq ($(shell where $(CC) 2>NUL),)
  51. CC = gcc
  52. endif
  53. else # POSIX-ish
  54. ifeq ($(shell command -v $(CC) 2>/dev/null),)
  55. ifeq ($(shell uname -s),Darwin)
  56. CC = clang
  57. # EXTRALIB += -Wl,-no_compact_unwind
  58. else
  59. CC = gcc
  60. endif # Darwin
  61. endif # CC exists
  62. endif # Shell is sane
  63. endif # CC is set to default
  64. # Default Fortran compiler (FC) is selected by f_check.
  65. ifndef MAKEFILE_RULE
  66. include $(TOPDIR)/Makefile.rule
  67. else
  68. include $(TOPDIR)/$(MAKEFILE_RULE)
  69. endif
  70. #
  71. # Beginning of system configuration
  72. #
  73. ifneq ($(BUILD_SINGLE),1)
  74. ifneq ($(BUILD_DOUBLE),1)
  75. ifneq ($(BUILD_COMPLEX),1)
  76. ifneq ($(BUILD_COMPLEX16),1)
  77. override BUILD_SINGLE=1
  78. override BUILD_DOUBLE=1
  79. override BUILD_COMPLEX=1
  80. override BUILD_COMPLEX16=1
  81. endif
  82. endif
  83. endif
  84. endif
  85. ifndef HOSTCC
  86. HOSTCC = $(CC)
  87. endif
  88. ifdef TARGET
  89. GETARCH_FLAGS := -DFORCE_$(TARGET)
  90. GETARCH_FLAGS += -DUSER_TARGET
  91. ifeq ($(TARGET), GENERIC)
  92. ifeq ($(DYNAMIC_ARCH), 1)
  93. override NO_EXPRECISION=1
  94. export NO_EXPRECISION
  95. endif
  96. endif
  97. endif
  98. # Force fallbacks for 32bit
  99. ifeq ($(BINARY), 32)
  100. ifeq ($(TARGET), HASWELL)
  101. GETARCH_FLAGS := -DFORCE_NEHALEM
  102. endif
  103. ifeq ($(TARGET), SKYLAKEX)
  104. GETARCH_FLAGS := -DFORCE_NEHALEM
  105. endif
  106. ifeq ($(TARGET), COOPERLAKE)
  107. GETARCH_FLAGS := -DFORCE_NEHALEM
  108. endif
  109. ifeq ($(TARGET), SAPPHIRERAPIDS)
  110. GETARCH_FLAGS := -DFORCE_NEHALEM
  111. endif
  112. ifeq ($(TARGET), SANDYBRIDGE)
  113. GETARCH_FLAGS := -DFORCE_NEHALEM
  114. endif
  115. ifeq ($(TARGET), BULLDOZER)
  116. GETARCH_FLAGS := -DFORCE_BARCELONA
  117. endif
  118. ifeq ($(TARGET), PILEDRIVER)
  119. GETARCH_FLAGS := -DFORCE_BARCELONA
  120. endif
  121. ifeq ($(TARGET), STEAMROLLER)
  122. GETARCH_FLAGS := -DFORCE_BARCELONA
  123. endif
  124. ifeq ($(TARGET), EXCAVATOR)
  125. GETARCH_FLAGS := -DFORCE_BARCELONA
  126. endif
  127. ifeq ($(TARGET), ZEN)
  128. GETARCH_FLAGS := -DFORCE_BARCELONA
  129. endif
  130. ifeq ($(TARGET), ARMV8)
  131. GETARCH_FLAGS := -DFORCE_ARMV7
  132. endif
  133. ifeq ($(TARGET), POWER8)
  134. GETARCH_FLAGS := -DFORCE_POWER6
  135. endif
  136. ifeq ($(TARGET), POWER9)
  137. GETARCH_FLAGS := -DFORCE_POWER6
  138. endif
  139. ifeq ($(TARGET), POWER10)
  140. GETARCH_FLAGS := -DFORCE_POWER6
  141. endif
  142. endif
  143. #TARGET_CORE will override TARGET which is used in DYNAMIC_ARCH=1.
  144. #
  145. ifdef TARGET_CORE
  146. GETARCH_FLAGS := -DFORCE_$(TARGET_CORE)
  147. endif
  148. # Force fallbacks for 32bit
  149. ifeq ($(BINARY), 32)
  150. ifeq ($(TARGET_CORE), HASWELL)
  151. GETARCH_FLAGS := -DFORCE_NEHALEM
  152. endif
  153. ifeq ($(TARGET_CORE), SKYLAKEX)
  154. GETARCH_FLAGS := -DFORCE_NEHALEM
  155. endif
  156. ifeq ($(TARGET_CORE), COOPERLAKE)
  157. GETARCH_FLAGS := -DFORCE_NEHALEM
  158. endif
  159. ifeq ($(TARGET_CORE), SAPPHIRERAPIDS)
  160. GETARCH_FLAGS := -DFORCE_NEHALEM
  161. endif
  162. ifeq ($(TARGET_CORE), SANDYBRIDGE)
  163. GETARCH_FLAGS := -DFORCE_NEHALEM
  164. endif
  165. ifeq ($(TARGET_CORE), BULLDOZER)
  166. GETARCH_FLAGS := -DFORCE_BARCELONA
  167. endif
  168. ifeq ($(TARGET_CORE), PILEDRIVER)
  169. GETARCH_FLAGS := -DFORCE_BARCELONA
  170. endif
  171. ifeq ($(TARGET_CORE), STEAMROLLER)
  172. GETARCH_FLAGS := -DFORCE_BARCELONA
  173. endif
  174. ifeq ($(TARGET_CORE), EXCAVATOR)
  175. GETARCH_FLAGS := -DFORCE_BARCELONA
  176. endif
  177. ifeq ($(TARGET_CORE), ZEN)
  178. GETARCH_FLAGS := -DFORCE_BARCELONA
  179. endif
  180. endif
  181. # On x86_64 build getarch with march=native unless the compiler is PGI. This is required to detect AVX512 support in getarch.
  182. ifeq ($(HOSTARCH), x86_64)
  183. ifeq ($(findstring pgcc,$(HOSTCC))$(findstring nvc,$(HOSTCC)),)
  184. GETARCH_FLAGS += -march=native
  185. endif
  186. endif
  187. ifdef INTERFACE64
  188. ifneq ($(INTERFACE64), 0)
  189. GETARCH_FLAGS += -DUSE64BITINT
  190. endif
  191. endif
  192. ifndef GEMM_MULTITHREAD_THRESHOLD
  193. GEMM_MULTITHREAD_THRESHOLD=4
  194. endif
  195. GETARCH_FLAGS += -DGEMM_MULTITHREAD_THRESHOLD=$(GEMM_MULTITHREAD_THRESHOLD)
  196. ifeq ($(NO_AVX), 1)
  197. GETARCH_FLAGS += -DNO_AVX
  198. endif
  199. ifeq ($(BINARY), 32)
  200. GETARCH_FLAGS += -DNO_AVX -DNO_AVX2 -DNO_AVX512
  201. NO_AVX512 = 1
  202. endif
  203. ifeq ($(NO_AVX2), 1)
  204. GETARCH_FLAGS += -DNO_AVX2
  205. endif
  206. ifeq ($(NO_AVX512), 1)
  207. GETARCH_FLAGS += -DNO_AVX512
  208. endif
  209. ifeq ($(DEBUG), 1)
  210. GETARCH_FLAGS += -g
  211. endif
  212. ifeq ($(QUIET_MAKE), 1)
  213. MAKE += -s
  214. endif
  215. ifndef NO_PARALLEL_MAKE
  216. NO_PARALLEL_MAKE=0
  217. endif
  218. GETARCH_FLAGS += -DNO_PARALLEL_MAKE=$(NO_PARALLEL_MAKE)
  219. ifdef MAKE_NB_JOBS
  220. GETARCH_FLAGS += -DMAKE_NB_JOBS=$(MAKE_NB_JOBS)
  221. endif
  222. ifeq ($(HOSTCC), loongcc)
  223. GETARCH_FLAGS += -static
  224. endif
  225. #if don't use Fortran, it will only compile CBLAS.
  226. ifeq ($(ONLY_CBLAS), 1)
  227. NO_LAPACK = 1
  228. else
  229. ONLY_CBLAS = 0
  230. endif
  231. #For small matrix optimization
  232. ifeq ($(ARCH), x86_64)
  233. SMALL_MATRIX_OPT = 1
  234. else ifeq ($(ARCH), power)
  235. SMALL_MATRIX_OPT = 1
  236. BUILD_BFLOAT16 = 1
  237. endif
  238. ifeq ($(SMALL_MATRIX_OPT), 1)
  239. CCOMMON_OPT += -DSMALL_MATRIX_OPT
  240. endif
  241. # This operation is expensive, so execution should be once.
  242. ifndef GOTOBLAS_MAKEFILE
  243. export GOTOBLAS_MAKEFILE = 1
  244. # Generating Makefile.conf and config.h
  245. 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) DYNAMIC_ARCH=$(DYNAMIC_ARCH) TARGET_CORE=$(TARGET_CORE) ONLY_CBLAS=$(ONLY_CBLAS) TARGET=$(TARGET) all)
  246. endif
  247. ifndef TARGET_CORE
  248. -include $(TOPDIR)/Makefile.conf
  249. else
  250. HAVE_NEON=
  251. HAVE_VFP=
  252. HAVE_VFPV3=
  253. HAVE_VFPV4=
  254. HAVE_MMX=
  255. HAVE_SSE=
  256. HAVE_SSE2=
  257. HAVE_SSE3=
  258. HAVE_SSSE3=
  259. HAVE_SSE4_1=
  260. HAVE_SSE4_2=
  261. HAVE_SSE4A=
  262. HAVE_SSE5=
  263. HAVE_AVX=
  264. HAVE_AVX2=
  265. HAVE_FMA3=
  266. include $(TOPDIR)/Makefile_kernel.conf
  267. endif
  268. ifndef NUM_PARALLEL
  269. NUM_PARALLEL = 1
  270. endif
  271. ifndef NUM_THREADS
  272. NUM_THREADS = $(NUM_CORES)
  273. endif
  274. ifeq ($(NUM_THREADS), 1)
  275. override USE_THREAD = 0
  276. override USE_OPENMP = 0
  277. endif
  278. ifdef USE_THREAD
  279. ifeq ($(USE_THREAD), 0)
  280. SMP =
  281. else
  282. SMP = 1
  283. endif
  284. else
  285. ifeq ($(NUM_THREADS), 1)
  286. SMP =
  287. else
  288. SMP = 1
  289. endif
  290. endif
  291. ifeq ($(SMP), 1)
  292. USE_LOCKING =
  293. endif
  294. ifndef NEED_PIC
  295. NEED_PIC = 1
  296. endif
  297. ARFLAGS =
  298. CPP = $(COMPILER) -E
  299. AR ?= $(CROSS_SUFFIX)ar
  300. AS ?= $(CROSS_SUFFIX)as
  301. LD ?= $(CROSS_SUFFIX)ld
  302. RANLIB ?= $(CROSS_SUFFIX)ranlib
  303. NM = $(CROSS_SUFFIX)nm
  304. DLLWRAP = $(CROSS_SUFFIX)dllwrap
  305. OBJCOPY = $(CROSS_SUFFIX)objcopy
  306. OBJCONV = $(CROSS_SUFFIX)objconv
  307. # When fortran support was either not detected or actively deselected, only build BLAS.
  308. ifeq ($(NOFORTRAN), 1)
  309. C_LAPACK = 1
  310. override FEXTRALIB =
  311. endif
  312. ifeq ($(C_COMPILER), GCC)
  313. GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
  314. GCCVERSIONGT4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 4)
  315. GCCVERSIONGT5 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 5)
  316. GCCVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 7)
  317. GCCVERSIONGTEQ8 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 8)
  318. GCCVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  319. GCCVERSIONGTEQ11 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 11)
  320. GCCVERSIONGTEQ10 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 10)
  321. # Note that the behavior of -dumpversion is compile-time-configurable for
  322. # gcc-7.x and newer. Use -dumpfullversion there
  323. ifeq ($(GCCVERSIONGTEQ7),1)
  324. GCCDUMPVERSION_PARAM := -dumpfullversion
  325. else
  326. GCCDUMPVERSION_PARAM := -dumpversion
  327. endif
  328. GCCMINORVERSIONGTEQ1 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 1)
  329. GCCMINORVERSIONGTEQ2 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 2)
  330. GCCMINORVERSIONGTEQ4 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 4)
  331. GCCMINORVERSIONGTEQ7 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 7)
  332. endif
  333. ifeq ($(C_COMPILER), CLANG)
  334. CLANGVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  335. CLANGVERSIONGTEQ12 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 12)
  336. endif
  337. #
  338. # OS dependent settings
  339. #
  340. ifeq ($(OSNAME), Darwin)
  341. ifndef MACOSX_DEPLOYMENT_TARGET
  342. ifeq ($(ARCH), arm64)
  343. export MACOSX_DEPLOYMENT_TARGET=11.0
  344. ifeq ($(C_COMPILER), GCC)
  345. export NO_SVE = 1
  346. endif
  347. else
  348. export MACOSX_DEPLOYMENT_TARGET=10.8
  349. endif
  350. endif
  351. MD5SUM = md5 -r
  352. XCVER = $(shell pkgutil --pkg-info=com.apple.pkg.Xcode |awk '/version:/ {print $2}'|cut -d: -f2|cut -f1 -d.)
  353. ifeq (x$(XCVER)x,xx)
  354. XCVER = $(shell pkgutil --pkg-info=com.apple.pkg.CLTools_Executables |awk '/version:/ {print $2}'|cut -d: -f2|cut -f1 -d.)
  355. endif
  356. ifeq (x$(XCVER), x 15)
  357. CCOMMON_OPT += -Wl,-ld_classic
  358. FCOMMON_OPT += -Wl,-ld_classic
  359. endif
  360. endif
  361. ifneq (,$(findstring $(OSNAME), FreeBSD OpenBSD DragonFly))
  362. MD5SUM = md5 -r
  363. endif
  364. ifeq ($(OSNAME), NetBSD)
  365. MD5SUM = md5 -n
  366. endif
  367. ifeq ($(OSNAME), Linux)
  368. EXTRALIB += -lm
  369. NO_EXPRECISION = 1
  370. endif
  371. ifeq ($(OSNAME), Android)
  372. EXTRALIB += -lm
  373. endif
  374. ifeq ($(OSNAME), AIX)
  375. EXTRALIB += -lm
  376. endif
  377. ifeq ($(OSNAME), $(filter $(OSNAME),FreeBSD OpenBSD NetBSD DragonFly))
  378. ifeq ($(ARCH), $(filter $(ARCH),arm arm64))
  379. EXTRALIB += -lm
  380. endif
  381. endif
  382. ifeq ($(OSNAME), WINNT)
  383. NEED_PIC = 0
  384. NO_EXPRECISION = 1
  385. EXTRALIB += -defaultlib:advapi32
  386. SUFFIX = obj
  387. PSUFFIX = pobj
  388. LIBSUFFIX = a
  389. ifeq ($(C_COMPILER), CLANG)
  390. CCOMMON_OPT += -DMS_ABI
  391. endif
  392. #Version tests for supporting specific features (MS_ABI, POWER9 intrinsics)
  393. ifeq ($(GCCVERSIONGT4), 1)
  394. # GCC Major version > 4
  395. # It is compatible with MSVC ABI.
  396. CCOMMON_OPT += -DMS_ABI
  397. endif
  398. ifeq ($(GCCVERSIONGTEQ4), 1)
  399. ifeq ($(GCCMINORVERSIONGTEQ7), 1)
  400. # GCC Version >=4.7
  401. # It is compatible with MSVC ABI.
  402. CCOMMON_OPT += -DMS_ABI
  403. endif
  404. endif
  405. # Ensure the correct stack alignment on Win32
  406. # http://permalink.gmane.org/gmane.comp.lib.openblas.general/97
  407. ifeq ($(ARCH), x86)
  408. CCOMMON_OPT += -mincoming-stack-boundary=2
  409. FCOMMON_OPT += -mincoming-stack-boundary=2
  410. endif
  411. endif
  412. ifeq ($(OSNAME), Interix)
  413. NEED_PIC = 0
  414. NO_EXPRECISION = 1
  415. INTERIX_TOOL_DIR = /opt/gcc.3.3/i586-pc-interix3/bin
  416. endif
  417. ifeq ($(OSNAME), CYGWIN_NT)
  418. NEED_PIC = 0
  419. NO_EXPRECISION = 1
  420. OS_CYGWIN_NT = 1
  421. endif
  422. ifneq ($(OSNAME), WINNT)
  423. ifneq ($(OSNAME), CYGWIN_NT)
  424. ifneq ($(OSNAME), Interix)
  425. ifneq ($(OSNAME), Android)
  426. ifdef SMP
  427. EXTRALIB += -lpthread
  428. endif
  429. endif
  430. endif
  431. endif
  432. endif
  433. # ifeq logical or
  434. ifeq ($(OSNAME), $(filter $(OSNAME),WINNT CYGWIN_NT Interix))
  435. OS_WINDOWS=1
  436. endif
  437. ifdef QUAD_PRECISION
  438. CCOMMON_OPT += -DQUAD_PRECISION
  439. NO_EXPRECISION = 1
  440. endif
  441. ifneq ($(ARCH), x86)
  442. ifneq ($(ARCH), x86_64)
  443. NO_EXPRECISION = 1
  444. endif
  445. endif
  446. ifdef UTEST_CHECK
  447. CCOMMON_OPT += -DUTEST_CHECK
  448. SANITY_CHECK = 1
  449. endif
  450. ifdef SANITY_CHECK
  451. CCOMMON_OPT += -DSANITY_CHECK -DREFNAME=$(*F)f$(BU)
  452. endif
  453. MAX_STACK_ALLOC ?= 2048
  454. ifneq ($(MAX_STACK_ALLOC), 0)
  455. CCOMMON_OPT += -DMAX_STACK_ALLOC=$(MAX_STACK_ALLOC)
  456. endif
  457. ifdef USE_LOCKING
  458. ifneq ($(USE_LOCKING), 0)
  459. CCOMMON_OPT += -DUSE_LOCKING
  460. endif
  461. endif
  462. #
  463. # Architecture dependent settings
  464. #
  465. ifeq ($(ARCH), x86)
  466. ifndef BINARY
  467. NO_BINARY_MODE = 1
  468. endif
  469. ifeq ($(CORE), generic)
  470. NO_EXPRECISION = 1
  471. endif
  472. ifndef NO_EXPRECISION
  473. ifeq ($(F_COMPILER), GFORTRAN)
  474. # ifeq logical or. GCC or LSB
  475. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  476. EXPRECISION = 1
  477. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  478. FCOMMON_OPT += -m128bit-long-double
  479. endif
  480. ifeq ($(C_COMPILER), CLANG)
  481. EXPRECISION = 1
  482. CCOMMON_OPT += -DEXPRECISION
  483. FCOMMON_OPT += -m128bit-long-double
  484. endif
  485. endif
  486. endif
  487. endif
  488. ifeq ($(ARCH), x86_64)
  489. ifeq ($(CORE), generic)
  490. NO_EXPRECISION = 1
  491. endif
  492. ifndef NO_EXPRECISION
  493. ifeq ($(F_COMPILER), GFORTRAN)
  494. # ifeq logical or. GCC or LSB
  495. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  496. EXPRECISION = 1
  497. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  498. FCOMMON_OPT += -m128bit-long-double
  499. endif
  500. ifeq ($(C_COMPILER), CLANG)
  501. EXPRECISION = 1
  502. CCOMMON_OPT += -DEXPRECISION
  503. FCOMMON_OPT += -m128bit-long-double
  504. endif
  505. endif
  506. endif
  507. endif
  508. ifeq ($(C_COMPILER), INTEL)
  509. CCOMMON_OPT += -wd981
  510. endif
  511. ifeq ($(USE_OPENMP), 1)
  512. #check
  513. ifeq ($(USE_THREAD), 0)
  514. $(error OpenBLAS: Cannot set both USE_OPENMP=1 and USE_THREAD=0. The USE_THREAD=0 is only for building single thread version.)
  515. endif
  516. # ifeq logical or. GCC or LSB
  517. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  518. CCOMMON_OPT += -fopenmp
  519. endif
  520. ifeq ($(C_COMPILER), CLANG)
  521. CCOMMON_OPT += -fopenmp
  522. ifeq ($(F_COMPILER), GFORTRAN)
  523. FEXTRALIB := $(subst -lgomp,-lomp,$(FEXTRALIB))
  524. endif
  525. endif
  526. ifeq ($(C_COMPILER), INTEL)
  527. CCOMMON_OPT += -fopenmp
  528. endif
  529. ifeq ($(C_COMPILER), PGI)
  530. CCOMMON_OPT += -mp
  531. endif
  532. ifeq ($(C_COMPILER), OPEN64)
  533. CCOMMON_OPT += -mp
  534. CEXTRALIB += -lstdc++
  535. endif
  536. ifeq ($(C_COMPILER), PATHSCALE)
  537. CCOMMON_OPT += -mp
  538. endif
  539. endif
  540. ifeq ($(DYNAMIC_ARCH), 1)
  541. ifeq ($(ARCH), x86)
  542. DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
  543. CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
  544. endif
  545. ifeq ($(ARCH), x86_64)
  546. DYNAMIC_CORE = PRESCOTT CORE2
  547. ifeq ($(DYNAMIC_OLDER), 1)
  548. DYNAMIC_CORE += PENRYN DUNNINGTON
  549. endif
  550. DYNAMIC_CORE += NEHALEM
  551. ifeq ($(DYNAMIC_OLDER), 1)
  552. DYNAMIC_CORE += OPTERON OPTERON_SSE3
  553. endif
  554. DYNAMIC_CORE += BARCELONA
  555. ifeq ($(DYNAMIC_OLDER), 1)
  556. DYNAMIC_CORE += BOBCAT ATOM NANO
  557. endif
  558. ifneq ($(NO_AVX), 1)
  559. DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER STEAMROLLER EXCAVATOR
  560. endif
  561. ifneq ($(NO_AVX2), 1)
  562. DYNAMIC_CORE += HASWELL ZEN
  563. endif
  564. ifneq ($(NO_AVX512), 1)
  565. ifneq ($(NO_AVX2), 1)
  566. DYNAMIC_CORE += SKYLAKEX COOPERLAKE SAPPHIRERAPIDS
  567. endif
  568. endif
  569. endif
  570. ifdef DYNAMIC_LIST
  571. override DYNAMIC_CORE = PRESCOTT $(DYNAMIC_LIST)
  572. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_PRESCOTT
  573. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  574. CCOMMON_OPT += $(XCCOMMON_OPT)
  575. #CCOMMON_OPT += -DDYNAMIC_LIST='$(DYNAMIC_LIST)'
  576. endif
  577. ifeq ($(ARCH), arm64)
  578. DYNAMIC_CORE = ARMV8
  579. DYNAMIC_CORE += CORTEXA53
  580. DYNAMIC_CORE += CORTEXA57
  581. DYNAMIC_CORE += NEOVERSEN1
  582. ifneq ($(NO_SVE), 1)
  583. DYNAMIC_CORE += NEOVERSEV1
  584. DYNAMIC_CORE += NEOVERSEN2
  585. DYNAMIC_CORE += ARMV8SVE
  586. endif
  587. DYNAMIC_CORE += THUNDERX
  588. DYNAMIC_CORE += THUNDERX2T99
  589. DYNAMIC_CORE += TSV110
  590. DYNAMIC_CORE += EMAG8180
  591. DYNAMIC_CORE += THUNDERX3T110
  592. ifdef DYNAMIC_LIST
  593. override DYNAMIC_CORE = ARMV8 $(DYNAMIC_LIST)
  594. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_ARMV8
  595. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  596. endif
  597. endif
  598. ifeq ($(ARCH), mips64)
  599. DYNAMIC_CORE = LOONGSON3R3 LOONGSON3R4 MIPS64_GENERIC
  600. ifdef DYNAMIC_LIST
  601. override DYNAMIC_CORE = MIPS64_GENERIC $(DYNAMIC_LIST)
  602. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_MIPS64_GENERIC
  603. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  604. endif
  605. endif
  606. ifeq ($(ARCH), loongarch64)
  607. DYNAMIC_CORE = LOONGSON3R5 LOONGSON2K1000 LOONGSONGENERIC
  608. endif
  609. ifeq ($(ARCH), zarch)
  610. DYNAMIC_CORE = ZARCH_GENERIC
  611. # if the compiler accepts -march=arch11 or -march=z13 and can compile a file
  612. # with z13-specific inline assembly, then we can include support for Z13.
  613. # note: -march=z13 is equivalent to -march=arch11 yet some compiler releases
  614. # only support one or the other.
  615. # note: LLVM version 6.x supported -march=z13 yet could not handle vector
  616. # registers in inline assembly, so the check for supporting the -march flag is
  617. # not enough.
  618. ZARCH_TEST_COMPILE=-c $(TOPDIR)/kernel/zarch/damin_z13.c -I$(TOPDIR) -o /dev/null > /dev/null 2> /dev/null
  619. ZARCH_CC_SUPPORTS_ARCH11=$(shell $(CC) -march=arch11 $(ZARCH_TEST_COMPILE) && echo 1)
  620. ZARCH_CC_SUPPORTS_Z13=$(shell $(CC) -march=z13 $(ZARCH_TEST_COMPILE) && echo 1)
  621. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH11), $(ZARCH_CC_SUPPORTS_Z13)), 1)
  622. DYNAMIC_CORE += Z13
  623. CCOMMON_OPT += -DDYN_Z13
  624. else
  625. $(info OpenBLAS: Not building Z13 kernels because the compiler $(CC) does not support it)
  626. endif
  627. # as above for z13, check for -march=arch12 and z14 support in the compiler.
  628. ZARCH_CC_SUPPORTS_ARCH12=$(shell $(CC) -march=arch12 $(ZARCH_TEST_COMPILE) && echo 1)
  629. ZARCH_CC_SUPPORTS_Z14=$(shell $(CC) -march=z14 $(ZARCH_TEST_COMPILE) && echo 1)
  630. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH12), $(ZARCH_CC_SUPPORTS_Z14)), 1)
  631. DYNAMIC_CORE += Z14
  632. CCOMMON_OPT += -DDYN_Z14
  633. else
  634. $(info OpenBLAS: Not building Z14 kernels because the compiler $(CC) does not support it)
  635. endif
  636. endif # ARCH zarch
  637. ifeq ($(ARCH), power)
  638. ifneq ($(C_COMPILER), PGI)
  639. DYNAMIC_CORE = POWER6
  640. DYNAMIC_CORE += POWER8
  641. ifneq ($(C_COMPILER), GCC)
  642. DYNAMIC_CORE += POWER9
  643. DYNAMIC_CORE += POWER10
  644. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  645. endif
  646. ifeq ($(C_COMPILER), GCC)
  647. ifeq ($(GCCVERSIONGT5), 1)
  648. DYNAMIC_CORE += POWER9
  649. else
  650. $(info, OpenBLAS: Your gcc version is too old to build the POWER9 kernels.)
  651. endif
  652. ifeq ($(OSNAME), AIX)
  653. LDVERSIONGTEQ35 := 1
  654. else
  655. LDVERSIONGTEQ35 := $(shell expr `$(CC) -Wl,--version 2> /dev/null | head -1 | cut -f2 -d "." | cut -f1 -d "-"` \>= 35)
  656. endif
  657. ifeq ($(GCCVERSIONGTEQ11)$(LDVERSIONGTEQ35), 11)
  658. DYNAMIC_CORE += POWER10
  659. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  660. else ifeq ($(GCCVERSIONGTEQ10), 1)
  661. ifeq ($(GCCMINORVERSIONGTEQ2)$(LDVERSIONGTEQ35), 11)
  662. DYNAMIC_CORE += POWER10
  663. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  664. endif
  665. else
  666. $(info, OpenBLAS: Your gcc version is too old to build the POWER10 kernels.)
  667. endif
  668. endif
  669. else
  670. DYNAMIC_CORE = POWER8
  671. DYNAMIC_CORE += POWER9
  672. endif
  673. endif
  674. # If DYNAMIC_CORE is not set, DYNAMIC_ARCH cannot do anything, so force it to empty
  675. ifndef DYNAMIC_CORE
  676. override DYNAMIC_ARCH=
  677. endif
  678. endif
  679. ifeq ($(ARCH), ia64)
  680. NO_BINARY_MODE = 1
  681. BINARY_DEFINED = 1
  682. ifeq ($(F_COMPILER), GFORTRAN)
  683. ifeq ($(C_COMPILER), GCC)
  684. # EXPRECISION = 1
  685. # CCOMMON_OPT += -DEXPRECISION
  686. endif
  687. endif
  688. endif
  689. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  690. NO_BINARY_MODE = 1
  691. endif
  692. ifeq ($(ARCH), alpha)
  693. NO_BINARY_MODE = 1
  694. BINARY_DEFINED = 1
  695. endif
  696. ifeq ($(ARCH), arm)
  697. NO_BINARY_MODE = 1
  698. BINARY_DEFINED = 1
  699. CCOMMON_OPT += -marm
  700. FCOMMON_OPT += -marm
  701. # If softfp abi is mentioned on the command line, force it.
  702. ifeq ($(ARM_SOFTFP_ABI), 1)
  703. CCOMMON_OPT += -mfloat-abi=softfp
  704. FCOMMON_OPT += -mfloat-abi=softfp
  705. endif
  706. ifeq ($(OSNAME), Android)
  707. ifeq ($(ARM_SOFTFP_ABI), 1)
  708. EXTRALIB += -lm
  709. else
  710. EXTRALIB += -Wl,-lm_hard
  711. endif
  712. endif
  713. endif
  714. ifeq ($(ARCH), arm64)
  715. NO_BINARY_MODE = 1
  716. BINARY_DEFINED = 1
  717. ifdef INTERFACE64
  718. ifneq ($(INTERFACE64), 0)
  719. ifeq ($(F_COMPILER), GFORTRAN)
  720. FCOMMON_OPT += -fdefault-integer-8
  721. endif
  722. ifeq ($(F_COMPILER), FLANG)
  723. FCOMMON_OPT += -i8
  724. endif
  725. endif
  726. endif
  727. endif
  728. ifeq ($(ARCH), riscv64)
  729. NO_BINARY_MODE = 1
  730. BINARY_DEFINED = 1
  731. ifdef INTERFACE64
  732. ifneq ($(INTERFACE64), 0)
  733. ifeq ($(F_COMPILER), GFORTRAN)
  734. FCOMMON_OPT += -fdefault-integer-8
  735. endif
  736. ifeq ($(F_COMPILER), FLANG)
  737. FCOMMON_OPT += -i8
  738. endif
  739. endif
  740. endif
  741. endif
  742. ifeq ($(ARCH), loongarch64)
  743. NO_BINARY_MODE = 1
  744. BINARY_DEFINED = 1
  745. ifdef INTERFACE64
  746. ifneq ($(INTERFACE64), 0)
  747. ifeq ($(F_COMPILER), GFORTRAN)
  748. FCOMMON_OPT += -fdefault-integer-8
  749. endif
  750. ifeq ($(F_COMPILER), FLANG)
  751. FCOMMON_OPT += -i8
  752. endif
  753. endif
  754. endif
  755. endif
  756. #
  757. # C Compiler dependent settings
  758. #
  759. # ifeq logical or. GCC or CLANG or LSB
  760. # http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
  761. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG LSB))
  762. CCOMMON_OPT += -Wall
  763. COMMON_PROF += -fno-inline
  764. NO_UNINITIALIZED_WARN = -Wno-uninitialized
  765. ifeq ($(QUIET_MAKE), 1)
  766. CCOMMON_OPT += $(NO_UNINITIALIZED_WARN) -Wno-unused
  767. endif
  768. ifdef NO_BINARY_MODE
  769. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  770. ifdef BINARY64
  771. CCOMMON_OPT += -mabi=64
  772. else
  773. CCOMMON_OPT += -mabi=n32
  774. endif
  775. BINARY_DEFINED = 1
  776. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  777. CCOMMON_OPT += -mabi=32
  778. BINARY_DEFINED = 1
  779. endif
  780. ifneq (, $(filter $(CORE), MIPS64_GENERIC))
  781. CCOMMON_OPT += -DNO_MSA
  782. FCOMMON_OPT += -DNO_MSA
  783. endif
  784. ifneq (, $(filter $(CORE),LOONGSON3R3 LOONGSON3R4))
  785. CCOMMON_OPT += -march=loongson3a
  786. FCOMMON_OPT += -march=loongson3a
  787. endif
  788. ifeq ($(CORE), MIPS24K)
  789. CCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  790. FCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  791. endif
  792. ifeq ($(CORE), MIPS1004K)
  793. CCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  794. FCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  795. endif
  796. ifeq ($(CORE), P5600)
  797. CCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  798. FCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  799. endif
  800. ifeq ($(CORE), I6400)
  801. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  802. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  803. endif
  804. ifeq ($(CORE), P6600)
  805. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  806. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  807. endif
  808. ifeq ($(CORE), I6500)
  809. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  810. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  811. endif
  812. ifeq ($(OSNAME), AIX)
  813. BINARY_DEFINED = 1
  814. endif
  815. ifeq ($(ARCH), loongarch64)
  816. LA64_ABI=$(shell $(CC) -mabi=lp64d -c $(TOPDIR)/cpuid_loongarch64.c -o /dev/null > /dev/null 2> /dev/null && echo lp64d)
  817. ifneq ($(LA64_ABI), lp64d)
  818. LA64_ABI=lp64
  819. endif
  820. CCOMMON_OPT += -march=loongarch64 -mabi=$(LA64_ABI)
  821. FCOMMON_OPT += -march=loongarch64 -mabi=$(LA64_ABI)
  822. endif
  823. endif
  824. ifndef BINARY_DEFINED
  825. ifneq ($(OSNAME), AIX)
  826. ifdef BINARY64
  827. ifneq ($(ARCH), riscv64)
  828. CCOMMON_OPT += -m64
  829. endif
  830. else
  831. CCOMMON_OPT += -m32
  832. endif
  833. endif
  834. endif
  835. endif
  836. ifeq ($(C_COMPILER), PGI)
  837. PGCVERSIONGT20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \> 20)
  838. PGCVERSIONEQ20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` == 20)
  839. PGCMINORVERSIONGE11 := $(shell expr `$(CC) --version|sed -n "2p" |cut -d "-" -f 1 |sed -e "s/[^0-9.]//g" |cut -c 4-5` \>= 11)
  840. PGCVERSIONCHECK := $(PGCVERSIONGT20)$(PGCVERSIONEQ20)$(PGCMINORVERSIONGE11)
  841. ifeq ($(PGCVERSIONCHECK), $(filter $(PGCVERSIONCHECK), 100 101 011))
  842. NEWPGI := 1
  843. PGCVERSIONGT21 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \> 21)
  844. PGCVERSIONEQ21 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` == 21)
  845. PGCVERSIONCHECK2 := $(PGCVERSIONGT21)$(PGCVERSIONEQ21)$(PGCMINORVERSIONGE11)
  846. ifeq ($(PGCVERSIONCHECK2), $(filter $(PGCVERSIONCHECK2), 100 101 011))
  847. NEWPGI2 := 1
  848. endif
  849. endif
  850. ifdef BINARY64
  851. ifeq ($(ARCH), x86_64)
  852. ifeq (,$(findstring tp,$(CFLAGS)))
  853. ifneq ($(NEWPGI2),1)
  854. CCOMMON_OPT += -tp p7-64
  855. else
  856. CCOMMON_OPT += -tp px
  857. endif
  858. endif
  859. ifneq ($(NEWPGI),1)
  860. CCOMMON_OPT += -D__MMX__ -Mnollvm
  861. endif
  862. else
  863. ifeq ($(ARCH), power)
  864. ifeq (,$(findstring tp,$(CFLAGS)))
  865. ifeq ($(CORE), POWER8)
  866. CCOMMON_OPT += -tp pwr8
  867. endif
  868. ifeq ($(CORE), POWER9)
  869. CCOMMON_OPT += -tp pwr9
  870. endif
  871. endif
  872. endif
  873. endif
  874. else
  875. ifneq ($(NEWPGI2),1)
  876. ifeq (,$(findstring tp,$(CFLAGS)))
  877. CCOMMON_OPT += -tp p7
  878. else
  879. CCOMMON_OPT += -tp px
  880. endif
  881. endif
  882. endif
  883. endif
  884. ifeq ($(C_COMPILER), PATHSCALE)
  885. ifdef BINARY64
  886. CCOMMON_OPT += -m64
  887. else
  888. CCOMMON_OPT += -m32
  889. endif
  890. endif
  891. #
  892. # Fortran Compiler dependent settings
  893. #
  894. ifeq ($(F_COMPILER), NAG)
  895. FCOMMON_OPT += -dcfuns -recursive -ieee=full -w=obs -thread_safe
  896. ifdef INTERFACE64
  897. ifneq ($(INTERFACE64), 0)
  898. FCOMMON_OPT += -i8
  899. endif
  900. endif
  901. ifeq ($(USE_OPENMP), 1)
  902. FCOMMON_OPT += -openmp
  903. endif
  904. endif
  905. ifeq ($(F_COMPILER), FLANG)
  906. CCOMMON_OPT += -DF_INTERFACE_FLANG
  907. FCOMMON_OPT += -Mrecursive -Kieee
  908. ifeq ($(OSNAME), Linux)
  909. ifeq ($(ARCH), x86_64)
  910. FLANG_VENDOR := $(shell $(FC) --version|head -1 |cut -f 1 -d " ")
  911. ifeq ($(FLANG_VENDOR), AMD)
  912. FCOMMON_OPT += -fno-unroll-loops
  913. endif
  914. endif
  915. endif
  916. ifdef BINARY64
  917. ifdef INTERFACE64
  918. ifneq ($(INTERFACE64), 0)
  919. FCOMMON_OPT += -i8
  920. endif
  921. endif
  922. FCOMMON_OPT += -Wall
  923. else
  924. FCOMMON_OPT += -Wall
  925. endif
  926. ifeq ($(USE_OPENMP), 1)
  927. FCOMMON_OPT += -fopenmp
  928. endif
  929. endif
  930. ifeq ($(F_COMPILER), G77)
  931. CCOMMON_OPT += -DF_INTERFACE_G77
  932. FCOMMON_OPT += -Wall
  933. ifndef NO_BINARY_MODE
  934. ifneq ($(OSNAME), AIX)
  935. ifdef BINARY64
  936. FCOMMON_OPT += -m64
  937. else
  938. FCOMMON_OPT += -m32
  939. endif
  940. endif
  941. endif
  942. endif
  943. ifeq ($(F_COMPILER), G95)
  944. CCOMMON_OPT += -DF_INTERFACE_G95
  945. FCOMMON_OPT += -Wall
  946. ifneq ($(OSNAME), AIX)
  947. ifndef NO_BINARY_MODE
  948. ifdef BINARY64
  949. FCOMMON_OPT += -m64
  950. else
  951. FCOMMON_OPT += -m32
  952. endif
  953. endif
  954. ifneq ($(NO_LAPACKE), 1)
  955. FCOMMON_OPT += -fno-second-underscore
  956. endif
  957. endif
  958. endif
  959. ifeq ($(F_COMPILER), $(filter $(F_COMPILER),GFORTRAN FLANGNEW))
  960. CCOMMON_OPT += -DF_INTERFACE_GFORT
  961. ifeq ($(F_COMPILER), GFORTRAN)
  962. FCOMMON_OPT += -Wall
  963. # make single-threaded LAPACK calls thread-safe #1847
  964. FCOMMON_OPT += -frecursive
  965. # work around ABI problem with passing single-character arguments
  966. FCOMMON_OPT += -fno-optimize-sibling-calls
  967. #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
  968. ifneq ($(NOFORTRAN), 1)
  969. ifneq ($(NOFORTRAN), 2)
  970. ifneq ($(NO_LAPACK), 1)
  971. EXTRALIB += -lgfortran
  972. endif
  973. endif
  974. endif
  975. endif
  976. ifdef NO_BINARY_MODE
  977. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  978. ifdef BINARY64
  979. FCOMMON_OPT += -mabi=64
  980. else
  981. FCOMMON_OPT += -mabi=n32
  982. endif
  983. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  984. FCOMMON_OPT += -mabi=32
  985. endif
  986. else
  987. ifdef BINARY64
  988. ifneq ($(OSNAME), AIX)
  989. ifneq ($(ARCH), riscv64)
  990. FCOMMON_OPT += -m64
  991. endif
  992. endif
  993. ifdef INTERFACE64
  994. ifneq ($(INTERFACE64), 0)
  995. FCOMMON_OPT += -fdefault-integer-8
  996. endif
  997. endif
  998. else
  999. ifneq ($(OSNAME), AIX)
  1000. FCOMMON_OPT += -m32
  1001. endif
  1002. endif
  1003. endif
  1004. ifeq ($(USE_OPENMP), 1)
  1005. FCOMMON_OPT += -fopenmp
  1006. endif
  1007. endif
  1008. ifeq ($(F_COMPILER), INTEL)
  1009. CCOMMON_OPT += -DF_INTERFACE_INTEL
  1010. ifdef INTERFACE64
  1011. ifneq ($(INTERFACE64), 0)
  1012. FCOMMON_OPT += -i8
  1013. endif
  1014. endif
  1015. FCOMMON_OPT += -recursive -fp-model strict -assume protect-parens
  1016. ifeq ($(USE_OPENMP), 1)
  1017. FCOMMON_OPT += -fopenmp
  1018. endif
  1019. endif
  1020. ifeq ($(F_COMPILER), FUJITSU)
  1021. CCOMMON_OPT += -DF_INTERFACE_FUJITSU
  1022. ifeq ($(USE_OPENMP), 1)
  1023. FCOMMON_OPT += -openmp
  1024. endif
  1025. endif
  1026. ifeq ($(F_COMPILER), IBM)
  1027. CCOMMON_OPT += -DF_INTERFACE_IBM
  1028. FEXTRALIB += -lxlf90
  1029. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG))
  1030. FCOMMON_OPT += -qextname
  1031. endif
  1032. # FCOMMON_OPT += -qarch=440
  1033. ifdef BINARY64
  1034. FCOMMON_OPT += -q64
  1035. ifdef INTERFACE64
  1036. ifneq ($(INTERFACE64), 0)
  1037. FCOMMON_OPT += -qintsize=8
  1038. endif
  1039. endif
  1040. else
  1041. FCOMMON_OPT += -q32
  1042. endif
  1043. ifeq ($(USE_OPENMP), 1)
  1044. FCOMMON_OPT += -openmp
  1045. endif
  1046. endif
  1047. ifeq ($(F_COMPILER), PGI)
  1048. CCOMMON_OPT += -DF_INTERFACE_PGI
  1049. COMMON_PROF += -DPGICOMPILER
  1050. ifdef BINARY64
  1051. ifdef INTERFACE64
  1052. ifneq ($(INTERFACE64), 0)
  1053. FCOMMON_OPT += -i8
  1054. endif
  1055. endif
  1056. ifeq ($(ARCH), x86_64)
  1057. ifneq ($(NEWPGI2),1)
  1058. FCOMMON_OPT += -tp p7-64
  1059. else
  1060. FCOMMON_OPT += -tp px
  1061. endif
  1062. else
  1063. ifeq ($(ARCH), power)
  1064. ifeq ($(CORE), POWER6)
  1065. $(warning NVIDIA HPC compilers do not support POWER6.)
  1066. endif
  1067. ifeq ($(CORE), POWER8)
  1068. FCOMMON_OPT += -tp pwr8
  1069. endif
  1070. ifeq ($(CORE), POWER9)
  1071. FCOMMON_OPT += -tp pwr9
  1072. endif
  1073. ifeq ($(CORE), POWER10)
  1074. $(warning NVIDIA HPC compilers do not support POWER10.)
  1075. endif
  1076. endif
  1077. endif
  1078. else
  1079. FCOMMON_OPT += -tp p7
  1080. endif
  1081. FCOMMON_OPT += -Mrecursive -Kieee
  1082. ifeq ($(USE_OPENMP), 1)
  1083. FCOMMON_OPT += -mp
  1084. endif
  1085. endif
  1086. ifeq ($(F_COMPILER), PATHSCALE)
  1087. CCOMMON_OPT += -DF_INTERFACE_PATHSCALE
  1088. ifdef BINARY64
  1089. ifdef INTERFACE64
  1090. ifneq ($(INTERFACE64), 0)
  1091. FCOMMON_OPT += -i8
  1092. endif
  1093. endif
  1094. endif
  1095. ifeq ($(USE_OPENMP), 1)
  1096. FCOMMON_OPT += -mp
  1097. endif
  1098. endif
  1099. ifeq ($(F_COMPILER), OPEN64)
  1100. CCOMMON_OPT += -DF_INTERFACE_OPEN64
  1101. ifdef BINARY64
  1102. ifdef INTERFACE64
  1103. ifneq ($(INTERFACE64), 0)
  1104. FCOMMON_OPT += -i8
  1105. endif
  1106. endif
  1107. endif
  1108. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  1109. ifndef BINARY64
  1110. FCOMMON_OPT += -n32
  1111. else
  1112. FCOMMON_OPT += -n64
  1113. endif
  1114. ifeq ($(CORE), LOONGSON3R3)
  1115. FCOMMON_OPT += -loongson3 -static
  1116. endif
  1117. ifeq ($(CORE), LOONGSON3R4)
  1118. FCOMMON_OPT += -loongson3 -static
  1119. endif
  1120. else
  1121. ifndef BINARY64
  1122. FCOMMON_OPT += -m32
  1123. else
  1124. FCOMMON_OPT += -m64
  1125. endif
  1126. endif
  1127. ifeq ($(USE_OPENMP), 1)
  1128. FEXTRALIB += -lstdc++
  1129. FCOMMON_OPT += -mp
  1130. endif
  1131. endif
  1132. ifeq ($(C_COMPILER), OPEN64)
  1133. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  1134. ifndef BINARY64
  1135. CCOMMON_OPT += -n32
  1136. else
  1137. CCOMMON_OPT += -n64
  1138. endif
  1139. ifeq ($(CORE), LOONGSON3R3)
  1140. CCOMMON_OPT += -loongson3 -static
  1141. endif
  1142. ifeq ($(CORE), LOONGSON3R4)
  1143. CCOMMON_OPT += -loongson3 -static
  1144. endif
  1145. else
  1146. ifndef BINARY64
  1147. CCOMMON_OPT += -m32
  1148. else
  1149. CCOMMON_OPT += -m64
  1150. endif
  1151. endif
  1152. endif
  1153. ifeq ($(C_COMPILER), SUN)
  1154. CCOMMON_OPT += -w
  1155. ifeq ($(ARCH), x86)
  1156. CCOMMON_OPT += -m32
  1157. else
  1158. ifdef BINARY64
  1159. CCOMMON_OPT += -m64
  1160. else
  1161. CCOMMON_OPT += -m32
  1162. endif
  1163. endif
  1164. endif
  1165. ifeq ($(F_COMPILER), SUN)
  1166. CCOMMON_OPT += -DF_INTERFACE_SUN
  1167. FCOMMON_OPT += -ftrap=%none -xrecursive
  1168. ifeq ($(ARCH), x86)
  1169. FCOMMON_OPT += -m32
  1170. else
  1171. ifdef BINARY64
  1172. FCOMMON_OPT += -m64
  1173. else
  1174. FCOMMON_OPT += -m32
  1175. endif
  1176. endif
  1177. ifeq ($(USE_OPENMP), 1)
  1178. FCOMMON_OPT += -xopenmp=parallel
  1179. endif
  1180. endif
  1181. ifeq ($(F_COMPILER), COMPAQ)
  1182. CCOMMON_OPT += -DF_INTERFACE_COMPAQ
  1183. ifeq ($(USE_OPENMP), 1)
  1184. FCOMMON_OPT += -openmp
  1185. endif
  1186. endif
  1187. ifeq ($(F_COMPILER), CRAY)
  1188. CCOMMON_OPT += -DF_INTERFACE_INTEL
  1189. FCOMMON_OPT += -hnopattern
  1190. ifdef INTERFACE64
  1191. ifneq ($(INTERFACE64), 0)
  1192. FCOMMON_OPT += -s integer64
  1193. endif
  1194. endif
  1195. ifneq ($(USE_OPENMP), 1)
  1196. FCOMMON_OPT += -O noomp
  1197. endif
  1198. endif
  1199. ifdef BINARY64
  1200. ifdef INTERFACE64
  1201. ifneq ($(INTERFACE64), 0)
  1202. CCOMMON_OPT +=
  1203. #-DUSE64BITINT
  1204. endif
  1205. endif
  1206. endif
  1207. ifeq ($(NEED_PIC), 1)
  1208. ifeq ($(C_COMPILER), IBM)
  1209. CCOMMON_OPT += -qpic=large
  1210. else
  1211. CCOMMON_OPT += -fPIC
  1212. endif
  1213. ifeq ($(F_COMPILER), SUN)
  1214. FCOMMON_OPT += -pic
  1215. else ifeq ($(F_COMPILER), NAG)
  1216. FCOMMON_OPT += -PIC
  1217. else ifeq ($(F_COMPILER), IBM)
  1218. FCOMMON_OPT += -qpic=large
  1219. else
  1220. FCOMMON_OPT += -fPIC
  1221. endif
  1222. endif
  1223. ifeq ($(DYNAMIC_ARCH), 1)
  1224. CCOMMON_OPT += -DDYNAMIC_ARCH
  1225. endif
  1226. ifeq ($(DYNAMIC_OLDER), 1)
  1227. CCOMMON_OPT += -DDYNAMIC_OLDER
  1228. endif
  1229. ifeq ($(C_LAPACK), 1)
  1230. CCOMMON_OPT += -DC_LAPACK
  1231. endif
  1232. ifeq ($(NO_LAPACK), 1)
  1233. CCOMMON_OPT += -DNO_LAPACK
  1234. #Disable LAPACK C interface
  1235. NO_LAPACKE = 1
  1236. endif
  1237. ifeq ($(NO_LAPACKE), 1)
  1238. CCOMMON_OPT += -DNO_LAPACKE
  1239. endif
  1240. ifeq ($(NO_AVX), 1)
  1241. CCOMMON_OPT += -DNO_AVX
  1242. endif
  1243. ifeq ($(ARCH), x86)
  1244. CCOMMON_OPT += -DNO_AVX
  1245. endif
  1246. ifeq ($(NO_AVX2), 1)
  1247. CCOMMON_OPT += -DNO_AVX2
  1248. endif
  1249. ifeq ($(NO_AVX512), 1)
  1250. CCOMMON_OPT += -DNO_AVX512
  1251. endif
  1252. ifeq ($(NO_SVE), 1)
  1253. CCOMMON_OPT += -DNO_SVE
  1254. endif
  1255. ifdef SMP
  1256. CCOMMON_OPT += -DSMP_SERVER
  1257. ifeq ($(ARCH), mips64)
  1258. USE_SIMPLE_THREADED_LEVEL3 = 1
  1259. endif
  1260. ifeq ($(USE_OPENMP), 1)
  1261. # USE_SIMPLE_THREADED_LEVEL3 = 1
  1262. # NO_AFFINITY = 1
  1263. CCOMMON_OPT += -DUSE_OPENMP
  1264. endif
  1265. ifeq ($(BIGNUMA), 1)
  1266. CCOMMON_OPT += -DBIGNUMA
  1267. endif
  1268. endif
  1269. ifeq ($(NO_WARMUP), 1)
  1270. CCOMMON_OPT += -DNO_WARMUP
  1271. endif
  1272. ifeq ($(CONSISTENT_FPCSR), 1)
  1273. CCOMMON_OPT += -DCONSISTENT_FPCSR
  1274. endif
  1275. # Only for development
  1276. # CCOMMON_OPT += -DPARAMTEST
  1277. # CCOMMON_OPT += -DPREFETCHTEST
  1278. # CCOMMON_OPT += -DNO_SWITCHING
  1279. # USE_PAPI = 1
  1280. ifdef USE_PAPI
  1281. CCOMMON_OPT += -DUSE_PAPI
  1282. EXTRALIB += -lpapi -lperfctr
  1283. endif
  1284. ifdef BUFFERSIZE
  1285. CCOMMON_OPT += -DBUFFERSIZE=$(BUFFERSIZE)
  1286. endif
  1287. ifdef DYNAMIC_THREADS
  1288. CCOMMON_OPT += -DDYNAMIC_THREADS
  1289. endif
  1290. CCOMMON_OPT += -DMAX_CPU_NUMBER=$(NUM_THREADS)
  1291. CCOMMON_OPT += -DMAX_PARALLEL_NUMBER=$(NUM_PARALLEL)
  1292. ifdef USE_SIMPLE_THREADED_LEVEL3
  1293. CCOMMON_OPT += -DUSE_SIMPLE_THREADED_LEVEL3
  1294. endif
  1295. ifeq ($(USE_TLS), 1)
  1296. CCOMMON_OPT += -DUSE_TLS
  1297. endif
  1298. ifeq ($(BUILD_BFLOAT16), 1)
  1299. CCOMMON_OPT += -DBUILD_BFLOAT16
  1300. endif
  1301. ifeq ($(BUILD_SINGLE), 1)
  1302. CCOMMON_OPT += -DBUILD_SINGLE=1
  1303. endif
  1304. ifeq ($(BUILD_DOUBLE), 1)
  1305. CCOMMON_OPT += -DBUILD_DOUBLE=1
  1306. endif
  1307. ifeq ($(BUILD_COMPLEX), 1)
  1308. CCOMMON_OPT += -DBUILD_COMPLEX=1
  1309. endif
  1310. ifeq ($(BUILD_COMPLEX16), 1)
  1311. CCOMMON_OPT += -DBUILD_COMPLEX16=1
  1312. endif
  1313. CCOMMON_OPT += -DVERSION=\"$(VERSION)\"
  1314. ifndef SYMBOLPREFIX
  1315. SYMBOLPREFIX =
  1316. endif
  1317. ifndef SYMBOLSUFFIX
  1318. SYMBOLSUFFIX =
  1319. endif
  1320. ifndef LIBSONAMEBASE
  1321. LIBSONAMEBASE = openblas
  1322. endif
  1323. ifndef LIBNAMESUFFIX
  1324. LIBNAMEBASE = $(SYMBOLPREFIX)$(LIBSONAMEBASE)$(SYMBOLSUFFIX)
  1325. else
  1326. LIBNAMEBASE = $(SYMBOLPREFIX)$(LIBSONAMEBASE)$(SYMBOLSUFFIX)_$(LIBNAMESUFFIX)
  1327. endif
  1328. ifeq ($(OSNAME), CYGWIN_NT)
  1329. LIBPREFIX = cyg$(LIBNAMEBASE)
  1330. else
  1331. LIBPREFIX = lib$(LIBNAMEBASE)
  1332. endif
  1333. KERNELDIR = $(TOPDIR)/kernel/$(ARCH)
  1334. include $(TOPDIR)/Makefile.$(ARCH)
  1335. ifneq ($(C_COMPILER), PGI)
  1336. ifneq ($(C_COMPILER), SUN)
  1337. CCOMMON_OPT += -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME
  1338. endif
  1339. endif
  1340. CCOMMON_OPT += -DASMNAME=$(FU)$(*F) -DASMFNAME=$(FU)$(*F)$(BU) -DNAME=$(*F)$(BU) -DCNAME=$(*F) -DCHAR_NAME=\"$(*F)$(BU)\" -DCHAR_CNAME=\"$(*F)\"
  1341. ifeq ($(CORE), PPC440)
  1342. CCOMMON_OPT += -DALLOC_QALLOC
  1343. endif
  1344. ifeq ($(CORE), PPC440FP2)
  1345. STATIC_ALLOCATION = 1
  1346. endif
  1347. ifneq ($(OSNAME), Linux)
  1348. NO_AFFINITY = 1
  1349. endif
  1350. ifneq ($(ARCH), x86_64)
  1351. ifneq ($(ARCH), x86)
  1352. NO_AFFINITY = 1
  1353. endif
  1354. endif
  1355. ifdef NO_AFFINITY
  1356. ifeq ($(NO_AFFINITY), 0)
  1357. override undefine NO_AFFINITY
  1358. else
  1359. CCOMMON_OPT += -DNO_AFFINITY
  1360. endif
  1361. endif
  1362. ifdef FUNCTION_PROFILE
  1363. CCOMMON_OPT += -DFUNCTION_PROFILE
  1364. endif
  1365. ifdef HUGETLB_ALLOCATION
  1366. CCOMMON_OPT += -DALLOC_HUGETLB
  1367. endif
  1368. ifdef HUGETLBFILE_ALLOCATION
  1369. CCOMMON_OPT += -DALLOC_HUGETLBFILE -DHUGETLB_FILE_NAME=$(HUGETLBFILE_ALLOCATION)
  1370. endif
  1371. ifdef STATIC_ALLOCATION
  1372. CCOMMON_OPT += -DALLOC_STATIC
  1373. endif
  1374. ifdef DEVICEDRIVER_ALLOCATION
  1375. CCOMMON_OPT += -DALLOC_DEVICEDRIVER -DDEVICEDRIVER_NAME=\"/dev/mapper\"
  1376. endif
  1377. ifdef MIXED_MEMORY_ALLOCATION
  1378. CCOMMON_OPT += -DMIXED_MEMORY_ALLOCATION
  1379. endif
  1380. ifeq ($(OSNAME), SunOS)
  1381. TAR = gtar
  1382. PATCH = gpatch
  1383. GREP = ggrep
  1384. AWK = nawk
  1385. else
  1386. TAR = tar
  1387. PATCH = patch
  1388. GREP = grep
  1389. AWK = awk
  1390. endif
  1391. ifndef MD5SUM
  1392. MD5SUM = md5sum
  1393. endif
  1394. REVISION = -r$(VERSION)
  1395. MAJOR_VERSION = $(word 1,$(subst ., ,$(VERSION)))
  1396. ifeq ($(DEBUG), 1)
  1397. COMMON_OPT += -g
  1398. endif
  1399. ifeq ($(DEBUG), 1)
  1400. FCOMMON_OPT += -g
  1401. endif
  1402. ifndef COMMON_OPT
  1403. COMMON_OPT = -O2
  1404. endif
  1405. ifndef FCOMMON_OPT
  1406. FCOMMON_OPT = -O2 -frecursive
  1407. endif
  1408. override CFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
  1409. override PFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
  1410. override FFLAGS += $(COMMON_OPT) $(FCOMMON_OPT)
  1411. override FPFLAGS += $(FCOMMON_OPT) $(COMMON_PROF)
  1412. #MAKEOVERRIDES =
  1413. ifeq ($(NEED_PIC), 1)
  1414. ifeq (,$(findstring PIC,$(FFLAGS)))
  1415. ifneq ($(F_COMPILER),IBM)
  1416. override FFLAGS += -fPIC
  1417. endif
  1418. endif
  1419. endif
  1420. #For LAPACK Fortran codes.
  1421. #Disable -fopenmp for LAPACK Fortran codes on Windows.
  1422. ifdef OS_WINDOWS
  1423. LAPACK_FFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FFLAGS))
  1424. LAPACK_FPFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FPFLAGS))
  1425. else
  1426. LAPACK_FFLAGS := $(FFLAGS)
  1427. LAPACK_FPFLAGS := $(FPFLAGS)
  1428. endif
  1429. ifeq ($(F_COMPILER),NAG)
  1430. LAPACK_FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1431. override FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1432. endif
  1433. ifeq ($(F_COMPILER),CRAY)
  1434. LAPACK_FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1435. override FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1436. endif
  1437. LAPACK_CFLAGS = $(CFLAGS)
  1438. LAPACK_CFLAGS += -DHAVE_LAPACK_CONFIG_H
  1439. ifdef INTERFACE64
  1440. ifneq ($(INTERFACE64), 0)
  1441. LAPACK_CFLAGS += -DLAPACK_ILP64
  1442. endif
  1443. endif
  1444. ifdef OS_WINDOWS
  1445. LAPACK_CFLAGS += -DOPENBLAS_OS_WINDOWS
  1446. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1447. endif
  1448. ifeq ($(C_COMPILER), LSB)
  1449. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1450. endif
  1451. ifndef SUFFIX
  1452. SUFFIX = o
  1453. endif
  1454. ifndef PSUFFIX
  1455. PSUFFIX = po
  1456. endif
  1457. ifndef LIBSUFFIX
  1458. LIBSUFFIX = a
  1459. endif
  1460. ifneq ($(DYNAMIC_ARCH), 1)
  1461. ifndef SMP
  1462. LIBNAME = $(LIBPREFIX)_$(LIBCORE)$(REVISION).$(LIBSUFFIX)
  1463. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)$(REVISION)_p.$(LIBSUFFIX)
  1464. else
  1465. LIBNAME = $(LIBPREFIX)_$(LIBCORE)p$(REVISION).$(LIBSUFFIX)
  1466. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)p$(REVISION)_p.$(LIBSUFFIX)
  1467. endif
  1468. else
  1469. ifndef SMP
  1470. LIBNAME = $(LIBPREFIX)$(REVISION).$(LIBSUFFIX)
  1471. LIBNAME_P = $(LIBPREFIX)$(REVISION)_p.$(LIBSUFFIX)
  1472. else
  1473. LIBNAME = $(LIBPREFIX)p$(REVISION).$(LIBSUFFIX)
  1474. LIBNAME_P = $(LIBPREFIX)p$(REVISION)_p.$(LIBSUFFIX)
  1475. endif
  1476. endif
  1477. LIBDLLNAME = $(LIBPREFIX).dll
  1478. IMPLIBNAME = lib$(LIBNAMEBASE).dll.a
  1479. ifneq ($(OSNAME), AIX)
  1480. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.so)
  1481. else
  1482. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.a)
  1483. endif
  1484. LIBDYNNAME = $(LIBNAME:.$(LIBSUFFIX)=.dylib)
  1485. LIBDEFNAME = $(LIBNAME:.$(LIBSUFFIX)=.def)
  1486. LIBEXPNAME = $(LIBNAME:.$(LIBSUFFIX)=.exp)
  1487. LIBZIPNAME = $(LIBNAME:.$(LIBSUFFIX)=.zip)
  1488. LIBS = $(TOPDIR)/$(LIBNAME)
  1489. LIBS_P = $(TOPDIR)/$(LIBNAME_P)
  1490. LIB_COMPONENTS = BLAS
  1491. ifneq ($(NO_CBLAS), 1)
  1492. LIB_COMPONENTS += CBLAS
  1493. endif
  1494. ifneq ($(NO_LAPACK), 1)
  1495. LIB_COMPONENTS += LAPACK
  1496. ifneq ($(NO_LAPACKE), 1)
  1497. LIB_COMPONENTS += LAPACKE
  1498. endif
  1499. ifeq ($(BUILD_RELAPACK), 1)
  1500. LIB_COMPONENTS += ReLAPACK
  1501. endif
  1502. endif
  1503. ifeq ($(ONLY_CBLAS), 1)
  1504. LIB_COMPONENTS = CBLAS
  1505. endif
  1506. export OSNAME
  1507. export ARCH
  1508. export CORE
  1509. export LIBCORE
  1510. export __BYTE_ORDER__
  1511. export ELF_VERSION
  1512. export PGCPATH
  1513. export CONFIG
  1514. export CC
  1515. export FC
  1516. export BU
  1517. export FU
  1518. export NEED2UNDERSCORES
  1519. export USE_THREAD
  1520. export NUM_THREADS
  1521. export NUM_CORES
  1522. export SMP
  1523. export MAKEFILE_RULE
  1524. export NEED_PIC
  1525. export BINARY
  1526. export BINARY32
  1527. export BINARY64
  1528. export F_COMPILER
  1529. export C_COMPILER
  1530. export USE_OPENMP
  1531. export CROSS
  1532. export CROSS_SUFFIX
  1533. export NOFORTRAN
  1534. export C_LAPACK
  1535. export NO_FBLAS
  1536. export EXTRALIB
  1537. export CEXTRALIB
  1538. export FEXTRALIB
  1539. export HAVE_SSE
  1540. export HAVE_SSE2
  1541. export HAVE_SSE3
  1542. export HAVE_SSSE3
  1543. export HAVE_SSE4_1
  1544. export HAVE_SSE4_2
  1545. export HAVE_SSE4A
  1546. export HAVE_SSE5
  1547. export HAVE_AVX
  1548. export HAVE_AVX2
  1549. export HAVE_FMA3
  1550. export HAVE_VFP
  1551. export HAVE_VFPV3
  1552. export HAVE_VFPV4
  1553. export HAVE_NEON
  1554. ifndef NO_MSA
  1555. export HAVE_MSA
  1556. export MSA_FLAGS
  1557. endif
  1558. export KERNELDIR
  1559. export FUNCTION_PROFILE
  1560. export TARGET_CORE
  1561. export NO_AVX512
  1562. export NO_AVX2
  1563. export BUILD_BFLOAT16
  1564. export NO_LSX
  1565. export NO_LASX
  1566. export SBGEMM_UNROLL_M
  1567. export SBGEMM_UNROLL_N
  1568. export SGEMM_UNROLL_M
  1569. export SGEMM_UNROLL_N
  1570. export DGEMM_UNROLL_M
  1571. export DGEMM_UNROLL_N
  1572. export QGEMM_UNROLL_M
  1573. export QGEMM_UNROLL_N
  1574. export CGEMM_UNROLL_M
  1575. export CGEMM_UNROLL_N
  1576. export ZGEMM_UNROLL_M
  1577. export ZGEMM_UNROLL_N
  1578. export XGEMM_UNROLL_M
  1579. export XGEMM_UNROLL_N
  1580. export CGEMM3M_UNROLL_M
  1581. export CGEMM3M_UNROLL_N
  1582. export ZGEMM3M_UNROLL_M
  1583. export ZGEMM3M_UNROLL_N
  1584. export XGEMM3M_UNROLL_M
  1585. export XGEMM3M_UNROLL_N
  1586. ifdef USE_CUDA
  1587. export CUDADIR
  1588. export CUCC
  1589. export CUFLAGS
  1590. export CULIB
  1591. endif
  1592. .SUFFIXES: .$(PSUFFIX) .$(SUFFIX) .f
  1593. .f.$(SUFFIX):
  1594. $(FC) $(FFLAGS) -c $< -o $(@F)
  1595. .f.$(PSUFFIX):
  1596. $(FC) $(FPFLAGS) -pg -c $< -o $(@F)
  1597. ifdef BINARY64
  1598. PATHSCALEPATH = /opt/pathscale/lib/3.1
  1599. PGIPATH = /opt/pgi/linux86-64/7.1-5/lib
  1600. else
  1601. PATHSCALEPATH = /opt/pathscale/lib/3.1/32
  1602. PGIPATH = /opt/pgi/linux86/7.1-5/lib
  1603. endif
  1604. ACMLPATH = /opt/acml/4.3.0
  1605. ifneq ($(OSNAME), Darwin)
  1606. MKLPATH = /opt/intel/mkl/10.2.2.025/lib
  1607. else
  1608. MKLPATH = /Library/Frameworks/Intel_MKL.framework/Versions/10.0.1.014/lib
  1609. endif
  1610. ATLASPATH = /opt/atlas/3.9.17/opteron
  1611. FLAMEPATH = $(HOME)/flame/lib
  1612. ifneq ($(OSNAME), SunOS)
  1613. SUNPATH = /opt/sunstudio12.1
  1614. else
  1615. SUNPATH = /opt/SUNWspro
  1616. endif