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 40 kB

7 years ago
4 years ago
13 years ago
4 years ago
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942943944945946947948949950951952953954955956957958959960961962963964965966967968969970971972973974975976977978979980981982983984985986987988989990991992993994995996997998999100010011002100310041005100610071008100910101011101210131014101510161017101810191020102110221023102410251026102710281029103010311032103310341035103610371038103910401041104210431044104510461047104810491050105110521053105410551056105710581059106010611062106310641065106610671068106910701071107210731074107510761077107810791080108110821083108410851086108710881089109010911092109310941095109610971098109911001101110211031104110511061107110811091110111111121113111411151116111711181119112011211122112311241125112611271128112911301131113211331134113511361137113811391140114111421143114411451146114711481149115011511152115311541155115611571158115911601161116211631164116511661167116811691170117111721173117411751176117711781179118011811182118311841185118611871188118911901191119211931194119511961197119811991200120112021203120412051206120712081209121012111212121312141215121612171218121912201221122212231224122512261227122812291230123112321233123412351236123712381239124012411242124312441245124612471248124912501251125212531254125512561257125812591260126112621263126412651266126712681269127012711272127312741275127612771278127912801281128212831284128512861287128812891290129112921293129412951296129712981299130013011302130313041305130613071308130913101311131213131314131513161317131813191320132113221323132413251326132713281329133013311332133313341335133613371338133913401341134213431344134513461347134813491350135113521353135413551356135713581359136013611362136313641365136613671368136913701371137213731374137513761377137813791380138113821383138413851386138713881389139013911392139313941395139613971398139914001401140214031404140514061407140814091410141114121413141414151416141714181419142014211422142314241425142614271428142914301431143214331434143514361437143814391440144114421443144414451446144714481449145014511452145314541455145614571458145914601461146214631464146514661467146814691470147114721473147414751476147714781479148014811482148314841485148614871488148914901491149214931494149514961497149814991500150115021503150415051506150715081509151015111512151315141515151615171518151915201521152215231524152515261527152815291530153115321533153415351536153715381539154015411542154315441545154615471548154915501551155215531554155515561557155815591560156115621563156415651566156715681569157015711572157315741575157615771578157915801581158215831584158515861587158815891590159115921593159415951596159715981599160016011602160316041605160616071608160916101611161216131614161516161617161816191620162116221623162416251626162716281629163016311632163316341635163616371638163916401641164216431644164516461647164816491650165116521653165416551656165716581659166016611662166316641665166616671668166916701671167216731674167516761677167816791680168116821683168416851686168716881689169016911692169316941695169616971698169917001701170217031704170517061707170817091710171117121713171417151716171717181719172017211722172317241725172617271728172917301731173217331734173517361737173817391740174117421743174417451746174717481749175017511752175317541755175617571758175917601761176217631764176517661767176817691770177117721773177417751776177717781779178017811782178317841785178617871788178917901791179217931794179517961797179817991800180118021803180418051806180718081809181018111812181318141815181618171818181918201821182218231824182518261827182818291830183118321833183418351836183718381839184018411842184318441845184618471848184918501851185218531854185518561857185818591860186118621863186418651866186718681869187018711872187318741875187618771878187918801881188218831884188518861887188818891890189118921893189418951896189718981899190019011902
  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 ($(ARCH), loongarch64)
  239. SMALL_MATRIX_OPT = 1
  240. endif
  241. ifeq ($(SMALL_MATRIX_OPT), 1)
  242. CCOMMON_OPT += -DSMALL_MATRIX_OPT
  243. endif
  244. # This operation is expensive, so execution should be once.
  245. ifndef GOTOBLAS_MAKEFILE
  246. export GOTOBLAS_MAKEFILE = 1
  247. # Generating Makefile.conf and config.h
  248. 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)
  249. endif
  250. ifndef TARGET_CORE
  251. -include $(TOPDIR)/Makefile.conf
  252. else
  253. HAVE_NEON=
  254. HAVE_VFP=
  255. HAVE_VFPV3=
  256. HAVE_VFPV4=
  257. HAVE_MMX=
  258. HAVE_SSE=
  259. HAVE_SSE2=
  260. HAVE_SSE3=
  261. HAVE_SSSE3=
  262. HAVE_SSE4_1=
  263. HAVE_SSE4_2=
  264. HAVE_SSE4A=
  265. HAVE_SSE5=
  266. HAVE_AVX=
  267. HAVE_AVX2=
  268. HAVE_FMA3=
  269. include $(TOPDIR)/Makefile_kernel.conf
  270. endif
  271. ifndef NUM_PARALLEL
  272. NUM_PARALLEL = 1
  273. endif
  274. ifndef NUM_THREADS
  275. NUM_THREADS = $(NUM_CORES)
  276. endif
  277. ifeq ($(NUM_THREADS), 1)
  278. override USE_THREAD = 0
  279. override USE_OPENMP = 0
  280. endif
  281. ifdef USE_THREAD
  282. ifeq ($(USE_THREAD), 0)
  283. SMP =
  284. else
  285. SMP = 1
  286. endif
  287. else
  288. ifeq ($(NUM_THREADS), 1)
  289. SMP =
  290. else
  291. SMP = 1
  292. endif
  293. endif
  294. ifeq ($(SMP), 1)
  295. USE_LOCKING =
  296. endif
  297. ifndef NEED_PIC
  298. NEED_PIC = 1
  299. endif
  300. ARFLAGS =
  301. CPP = $(COMPILER) -E
  302. AR ?= $(CROSS_SUFFIX)ar
  303. AS ?= $(CROSS_SUFFIX)as
  304. LD ?= $(CROSS_SUFFIX)ld
  305. RANLIB ?= $(CROSS_SUFFIX)ranlib
  306. NM = $(CROSS_SUFFIX)nm
  307. DLLWRAP = $(CROSS_SUFFIX)dllwrap
  308. OBJCOPY = $(CROSS_SUFFIX)objcopy
  309. OBJCONV = $(CROSS_SUFFIX)objconv
  310. # When fortran support was either not detected or actively deselected, only build BLAS.
  311. ifeq ($(NOFORTRAN), 1)
  312. C_LAPACK = 1
  313. override FEXTRALIB =
  314. ifeq ($(C_COMPILER), GCC)
  315. CCOMMON_OPT += -Wno-error=incompatible-pointer-types
  316. endif
  317. endif
  318. ifeq ($(C_COMPILER), GCC)
  319. GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
  320. GCCVERSIONGT4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 4)
  321. GCCVERSIONGT5 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 5)
  322. GCCVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 7)
  323. GCCVERSIONGTEQ8 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 8)
  324. GCCVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  325. GCCVERSIONGTEQ10 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 10)
  326. GCCVERSIONGTEQ11 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 11)
  327. GCCVERSIONGTEQ12 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 12)
  328. # Note that the behavior of -dumpversion is compile-time-configurable for
  329. # gcc-7.x and newer. Use -dumpfullversion there
  330. ifeq ($(GCCVERSIONGTEQ7),1)
  331. GCCDUMPVERSION_PARAM := -dumpfullversion
  332. else
  333. GCCDUMPVERSION_PARAM := -dumpversion
  334. endif
  335. GCCMINORVERSIONGTEQ1 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 1)
  336. GCCMINORVERSIONGTEQ2 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 2)
  337. GCCMINORVERSIONGTEQ4 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 4)
  338. GCCMINORVERSIONGTEQ7 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 7)
  339. endif
  340. ifeq ($(C_COMPILER), CLANG)
  341. CLANGVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  342. CLANGVERSIONGTEQ12 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 12)
  343. endif
  344. #
  345. # OS dependent settings
  346. #
  347. ifeq ($(OSNAME), Darwin)
  348. ifndef MACOSX_DEPLOYMENT_TARGET
  349. ifeq ($(ARCH), arm64)
  350. export MACOSX_DEPLOYMENT_TARGET=11.0
  351. ifeq ($(C_COMPILER), GCC)
  352. export NO_SVE = 1
  353. endif
  354. else
  355. export MACOSX_DEPLOYMENT_TARGET=10.8
  356. endif
  357. endif
  358. MD5SUM = md5 -r
  359. XCVER = $(shell pkgutil --pkg-info=com.apple.pkg.Xcode |awk '/version:/ {print $2}'|cut -d: -f2|cut -f1 -d.)
  360. ifeq (x$(XCVER)x,xx)
  361. XCVER = $(shell pkgutil --pkg-info=com.apple.pkg.CLTools_Executables |awk '/version:/ {print $2}'|cut -d: -f2|cut -f1 -d.)
  362. endif
  363. ifeq (x$(XCVER), x 15)
  364. CCOMMON_OPT += -Wl,-ld_classic
  365. FCOMMON_OPT += -Wl,-ld_classic
  366. endif
  367. endif
  368. ifneq (,$(findstring $(OSNAME), FreeBSD OpenBSD DragonFly))
  369. MD5SUM = md5 -r
  370. endif
  371. ifeq ($(OSNAME), NetBSD)
  372. MD5SUM = md5 -n
  373. endif
  374. ifeq ($(OSNAME), Linux)
  375. EXTRALIB += -lm
  376. NO_EXPRECISION = 1
  377. endif
  378. ifeq ($(OSNAME), Android)
  379. EXTRALIB += -lm
  380. endif
  381. ifeq ($(OSNAME), AIX)
  382. EXTRALIB += -lm
  383. endif
  384. ifeq ($(OSNAME), $(filter $(OSNAME),FreeBSD OpenBSD NetBSD DragonFly))
  385. ifeq ($(ARCH), $(filter $(ARCH),arm arm64))
  386. EXTRALIB += -lm
  387. endif
  388. endif
  389. ifeq ($(OSNAME), WINNT)
  390. NEED_PIC = 0
  391. NO_EXPRECISION = 1
  392. EXTRALIB += -defaultlib:advapi32
  393. SUFFIX = obj
  394. PSUFFIX = pobj
  395. LIBSUFFIX = a
  396. ifeq ($(C_COMPILER), CLANG)
  397. CCOMMON_OPT += -DMS_ABI
  398. endif
  399. #Version tests for supporting specific features (MS_ABI, POWER9 intrinsics)
  400. ifeq ($(GCCVERSIONGT4), 1)
  401. # GCC Major version > 4
  402. # It is compatible with MSVC ABI.
  403. CCOMMON_OPT += -DMS_ABI
  404. endif
  405. ifeq ($(GCCVERSIONGTEQ4), 1)
  406. ifeq ($(GCCMINORVERSIONGTEQ7), 1)
  407. # GCC Version >=4.7
  408. # It is compatible with MSVC ABI.
  409. CCOMMON_OPT += -DMS_ABI
  410. endif
  411. endif
  412. # Ensure the correct stack alignment on Win32
  413. # http://permalink.gmane.org/gmane.comp.lib.openblas.general/97
  414. ifeq ($(ARCH), x86)
  415. CCOMMON_OPT += -mincoming-stack-boundary=2
  416. FCOMMON_OPT += -mincoming-stack-boundary=2
  417. endif
  418. endif
  419. ifeq ($(OSNAME), Interix)
  420. NEED_PIC = 0
  421. NO_EXPRECISION = 1
  422. INTERIX_TOOL_DIR = /opt/gcc.3.3/i586-pc-interix3/bin
  423. endif
  424. ifeq ($(OSNAME), CYGWIN_NT)
  425. NEED_PIC = 0
  426. NO_EXPRECISION = 1
  427. OS_CYGWIN_NT = 1
  428. endif
  429. ifneq ($(OSNAME), WINNT)
  430. ifneq ($(OSNAME), CYGWIN_NT)
  431. ifneq ($(OSNAME), Interix)
  432. ifneq ($(OSNAME), Android)
  433. ifdef SMP
  434. EXTRALIB += -lpthread
  435. endif
  436. endif
  437. endif
  438. endif
  439. endif
  440. # ifeq logical or
  441. ifeq ($(OSNAME), $(filter $(OSNAME),WINNT CYGWIN_NT Interix))
  442. OS_WINDOWS=1
  443. endif
  444. ifdef QUAD_PRECISION
  445. CCOMMON_OPT += -DQUAD_PRECISION
  446. NO_EXPRECISION = 1
  447. endif
  448. ifneq ($(ARCH), x86)
  449. ifneq ($(ARCH), x86_64)
  450. NO_EXPRECISION = 1
  451. endif
  452. endif
  453. ifdef UTEST_CHECK
  454. CCOMMON_OPT += -DUTEST_CHECK
  455. SANITY_CHECK = 1
  456. endif
  457. ifdef SANITY_CHECK
  458. CCOMMON_OPT += -DSANITY_CHECK -DREFNAME=$(*F)f$(BU)
  459. endif
  460. MAX_STACK_ALLOC ?= 2048
  461. ifneq ($(MAX_STACK_ALLOC), 0)
  462. CCOMMON_OPT += -DMAX_STACK_ALLOC=$(MAX_STACK_ALLOC)
  463. endif
  464. ifdef USE_LOCKING
  465. ifneq ($(USE_LOCKING), 0)
  466. CCOMMON_OPT += -DUSE_LOCKING
  467. endif
  468. endif
  469. #
  470. # Architecture dependent settings
  471. #
  472. ifeq ($(ARCH), x86)
  473. ifndef BINARY
  474. NO_BINARY_MODE = 1
  475. endif
  476. ifeq ($(CORE), generic)
  477. NO_EXPRECISION = 1
  478. endif
  479. ifndef NO_EXPRECISION
  480. ifeq ($(F_COMPILER), GFORTRAN)
  481. # ifeq logical or. GCC or LSB
  482. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  483. EXPRECISION = 1
  484. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  485. FCOMMON_OPT += -m128bit-long-double
  486. endif
  487. ifeq ($(C_COMPILER), CLANG)
  488. EXPRECISION = 1
  489. CCOMMON_OPT += -DEXPRECISION
  490. FCOMMON_OPT += -m128bit-long-double
  491. endif
  492. endif
  493. endif
  494. endif
  495. ifeq ($(ARCH), x86_64)
  496. ifeq ($(CORE), generic)
  497. NO_EXPRECISION = 1
  498. endif
  499. ifndef NO_EXPRECISION
  500. ifeq ($(F_COMPILER), GFORTRAN)
  501. # ifeq logical or. GCC or LSB
  502. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  503. EXPRECISION = 1
  504. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  505. FCOMMON_OPT += -m128bit-long-double
  506. endif
  507. ifeq ($(C_COMPILER), CLANG)
  508. EXPRECISION = 1
  509. CCOMMON_OPT += -DEXPRECISION
  510. FCOMMON_OPT += -m128bit-long-double
  511. endif
  512. endif
  513. endif
  514. endif
  515. ifeq ($(C_COMPILER), INTEL)
  516. CCOMMON_OPT += -wd981
  517. endif
  518. ifeq ($(USE_OPENMP), 1)
  519. #check
  520. ifeq ($(USE_THREAD), 0)
  521. $(error OpenBLAS: Cannot set both USE_OPENMP=1 and USE_THREAD=0. The USE_THREAD=0 is only for building single thread version.)
  522. endif
  523. # ifeq logical or. GCC or LSB
  524. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  525. CCOMMON_OPT += -fopenmp
  526. endif
  527. ifeq ($(C_COMPILER), CLANG)
  528. CCOMMON_OPT += -fopenmp
  529. ifeq ($(F_COMPILER), GFORTRAN)
  530. FEXTRALIB := $(subst -lgomp,-lomp,$(FEXTRALIB))
  531. endif
  532. endif
  533. ifeq ($(C_COMPILER), INTEL)
  534. CCOMMON_OPT += -fopenmp
  535. endif
  536. ifeq ($(C_COMPILER), PGI)
  537. CCOMMON_OPT += -mp
  538. endif
  539. ifeq ($(C_COMPILER), OPEN64)
  540. CCOMMON_OPT += -mp
  541. CEXTRALIB += -lstdc++
  542. endif
  543. ifeq ($(C_COMPILER), PATHSCALE)
  544. CCOMMON_OPT += -mp
  545. endif
  546. endif
  547. ifeq ($(DYNAMIC_ARCH), 1)
  548. ifeq ($(ARCH), x86)
  549. DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
  550. CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
  551. endif
  552. ifeq ($(ARCH), x86_64)
  553. DYNAMIC_CORE = PRESCOTT CORE2
  554. ifeq ($(DYNAMIC_OLDER), 1)
  555. DYNAMIC_CORE += PENRYN DUNNINGTON
  556. endif
  557. DYNAMIC_CORE += NEHALEM
  558. ifeq ($(DYNAMIC_OLDER), 1)
  559. DYNAMIC_CORE += OPTERON OPTERON_SSE3
  560. endif
  561. DYNAMIC_CORE += BARCELONA
  562. ifeq ($(DYNAMIC_OLDER), 1)
  563. DYNAMIC_CORE += BOBCAT ATOM NANO
  564. endif
  565. ifneq ($(NO_AVX), 1)
  566. DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER STEAMROLLER EXCAVATOR
  567. endif
  568. ifneq ($(NO_AVX2), 1)
  569. DYNAMIC_CORE += HASWELL ZEN
  570. endif
  571. ifneq ($(NO_AVX512), 1)
  572. ifneq ($(NO_AVX2), 1)
  573. DYNAMIC_CORE += SKYLAKEX COOPERLAKE SAPPHIRERAPIDS
  574. endif
  575. endif
  576. endif
  577. ifdef DYNAMIC_LIST
  578. override DYNAMIC_CORE = PRESCOTT $(DYNAMIC_LIST)
  579. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_PRESCOTT
  580. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  581. CCOMMON_OPT += $(XCCOMMON_OPT)
  582. #CCOMMON_OPT += -DDYNAMIC_LIST='$(DYNAMIC_LIST)'
  583. endif
  584. ifeq ($(ARCH), arm64)
  585. DYNAMIC_CORE = ARMV8
  586. DYNAMIC_CORE += CORTEXA53
  587. DYNAMIC_CORE += CORTEXA57
  588. DYNAMIC_CORE += NEOVERSEN1
  589. ifneq ($(NO_SVE), 1)
  590. DYNAMIC_CORE += NEOVERSEV1
  591. DYNAMIC_CORE += NEOVERSEN2
  592. DYNAMIC_CORE += ARMV8SVE
  593. endif
  594. DYNAMIC_CORE += THUNDERX
  595. DYNAMIC_CORE += THUNDERX2T99
  596. DYNAMIC_CORE += TSV110
  597. DYNAMIC_CORE += EMAG8180
  598. DYNAMIC_CORE += THUNDERX3T110
  599. ifdef DYNAMIC_LIST
  600. override DYNAMIC_CORE = ARMV8 $(DYNAMIC_LIST)
  601. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_ARMV8
  602. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  603. endif
  604. endif
  605. ifeq ($(ARCH), mips64)
  606. DYNAMIC_CORE = LOONGSON3R3 LOONGSON3R4 MIPS64_GENERIC
  607. ifdef DYNAMIC_LIST
  608. override DYNAMIC_CORE = MIPS64_GENERIC $(DYNAMIC_LIST)
  609. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_MIPS64_GENERIC
  610. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  611. endif
  612. endif
  613. ifeq ($(ARCH), loongarch64)
  614. DYNAMIC_CORE = LOONGSON3R5 LOONGSON2K1000 LOONGSONGENERIC
  615. endif
  616. ifeq ($(ARCH), zarch)
  617. DYNAMIC_CORE = ZARCH_GENERIC
  618. # if the compiler accepts -march=arch11 or -march=z13 and can compile a file
  619. # with z13-specific inline assembly, then we can include support for Z13.
  620. # note: -march=z13 is equivalent to -march=arch11 yet some compiler releases
  621. # only support one or the other.
  622. # note: LLVM version 6.x supported -march=z13 yet could not handle vector
  623. # registers in inline assembly, so the check for supporting the -march flag is
  624. # not enough.
  625. ZARCH_TEST_COMPILE=-c $(TOPDIR)/kernel/zarch/damin_z13.c -I$(TOPDIR) -o /dev/null > /dev/null 2> /dev/null
  626. ZARCH_CC_SUPPORTS_ARCH11=$(shell $(CC) -march=arch11 $(ZARCH_TEST_COMPILE) && echo 1)
  627. ZARCH_CC_SUPPORTS_Z13=$(shell $(CC) -march=z13 $(ZARCH_TEST_COMPILE) && echo 1)
  628. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH11), $(ZARCH_CC_SUPPORTS_Z13)), 1)
  629. DYNAMIC_CORE += Z13
  630. CCOMMON_OPT += -DDYN_Z13
  631. else
  632. $(info OpenBLAS: Not building Z13 kernels because the compiler $(CC) does not support it)
  633. endif
  634. # as above for z13, check for -march=arch12 and z14 support in the compiler.
  635. ZARCH_CC_SUPPORTS_ARCH12=$(shell $(CC) -march=arch12 $(ZARCH_TEST_COMPILE) && echo 1)
  636. ZARCH_CC_SUPPORTS_Z14=$(shell $(CC) -march=z14 $(ZARCH_TEST_COMPILE) && echo 1)
  637. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH12), $(ZARCH_CC_SUPPORTS_Z14)), 1)
  638. DYNAMIC_CORE += Z14
  639. CCOMMON_OPT += -DDYN_Z14
  640. else
  641. $(info OpenBLAS: Not building Z14 kernels because the compiler $(CC) does not support it)
  642. endif
  643. endif # ARCH zarch
  644. ifeq ($(ARCH), power)
  645. ifneq ($(C_COMPILER), PGI)
  646. DYNAMIC_CORE = POWER6
  647. DYNAMIC_CORE += POWER8
  648. ifneq ($(C_COMPILER), GCC)
  649. DYNAMIC_CORE += POWER9
  650. DYNAMIC_CORE += POWER10
  651. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  652. endif
  653. ifeq ($(C_COMPILER), GCC)
  654. ifeq ($(GCCVERSIONGT5), 1)
  655. DYNAMIC_CORE += POWER9
  656. else
  657. $(info, OpenBLAS: Your gcc version is too old to build the POWER9 kernels.)
  658. endif
  659. ifeq ($(OSNAME), AIX)
  660. LDVERSIONGTEQ35 := 1
  661. else
  662. LDVERSIONGTEQ35 := $(shell expr `$(CC) -Wl,--version 2> /dev/null | head -1 | cut -f2 -d "." | cut -f1 -d "-"` \>= 35)
  663. endif
  664. ifeq ($(GCCVERSIONGTEQ11)$(LDVERSIONGTEQ35), 11)
  665. DYNAMIC_CORE += POWER10
  666. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  667. else ifeq ($(GCCVERSIONGTEQ10), 1)
  668. ifeq ($(GCCMINORVERSIONGTEQ2)$(LDVERSIONGTEQ35), 11)
  669. DYNAMIC_CORE += POWER10
  670. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  671. endif
  672. else
  673. $(info, OpenBLAS: Your gcc version is too old to build the POWER10 kernels.)
  674. endif
  675. endif
  676. else
  677. DYNAMIC_CORE = POWER8
  678. DYNAMIC_CORE += POWER9
  679. endif
  680. endif
  681. # If DYNAMIC_CORE is not set, DYNAMIC_ARCH cannot do anything, so force it to empty
  682. ifndef DYNAMIC_CORE
  683. override DYNAMIC_ARCH=
  684. endif
  685. endif
  686. ifeq ($(ARCH), ia64)
  687. NO_BINARY_MODE = 1
  688. BINARY_DEFINED = 1
  689. ifeq ($(F_COMPILER), GFORTRAN)
  690. ifeq ($(C_COMPILER), GCC)
  691. # EXPRECISION = 1
  692. # CCOMMON_OPT += -DEXPRECISION
  693. endif
  694. endif
  695. endif
  696. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  697. NO_BINARY_MODE = 1
  698. endif
  699. ifeq ($(ARCH), alpha)
  700. NO_BINARY_MODE = 1
  701. BINARY_DEFINED = 1
  702. endif
  703. ifeq ($(ARCH), arm)
  704. NO_BINARY_MODE = 1
  705. BINARY_DEFINED = 1
  706. ifneq ($(EMBEDDED), 1)
  707. CCOMMON_OPT += -marm
  708. FCOMMON_OPT += -marm
  709. else
  710. CCOMMON_OPT += -DOS_EMBEDDED -mthumb -mcpu=cortex-m4 -mfloat-abi=hard -mfpu=fpv4-sp-d16
  711. endif
  712. # If softfp abi is mentioned on the command line, force it.
  713. ifeq ($(ARM_SOFTFP_ABI), 1)
  714. CCOMMON_OPT += -mfloat-abi=softfp
  715. FCOMMON_OPT += -mfloat-abi=softfp
  716. endif
  717. ifeq ($(OSNAME), Android)
  718. ifeq ($(ARM_SOFTFP_ABI), 1)
  719. EXTRALIB += -lm
  720. else
  721. EXTRALIB += -Wl,-lm_hard
  722. endif
  723. endif
  724. endif
  725. ifeq ($(ARCH), arm64)
  726. NO_BINARY_MODE = 1
  727. BINARY_DEFINED = 1
  728. ifdef INTERFACE64
  729. ifneq ($(INTERFACE64), 0)
  730. ifeq ($(F_COMPILER), GFORTRAN)
  731. FCOMMON_OPT += -fdefault-integer-8
  732. endif
  733. ifeq ($(F_COMPILER), FLANG)
  734. FCOMMON_OPT += -i8
  735. endif
  736. endif
  737. endif
  738. endif
  739. ifeq ($(ARCH), riscv64)
  740. NO_BINARY_MODE = 1
  741. BINARY_DEFINED = 1
  742. ifdef INTERFACE64
  743. ifneq ($(INTERFACE64), 0)
  744. ifeq ($(F_COMPILER), GFORTRAN)
  745. FCOMMON_OPT += -fdefault-integer-8
  746. endif
  747. ifeq ($(F_COMPILER), FLANG)
  748. FCOMMON_OPT += -i8
  749. endif
  750. endif
  751. endif
  752. endif
  753. ifeq ($(ARCH), loongarch64)
  754. NO_BINARY_MODE = 1
  755. BINARY_DEFINED = 1
  756. ifdef INTERFACE64
  757. ifneq ($(INTERFACE64), 0)
  758. ifeq ($(F_COMPILER), GFORTRAN)
  759. FCOMMON_OPT += -fdefault-integer-8
  760. endif
  761. ifeq ($(F_COMPILER), FLANG)
  762. FCOMMON_OPT += -i8
  763. endif
  764. endif
  765. endif
  766. endif
  767. ifeq ($(ARCH), csky)
  768. NO_BINARY_MODE = 1
  769. BINARY_DEFINED = 1
  770. endif
  771. #
  772. # C Compiler dependent settings
  773. #
  774. # ifeq logical or. GCC or CLANG or LSB
  775. # http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
  776. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG LSB))
  777. CCOMMON_OPT += -Wall
  778. COMMON_PROF += -fno-inline
  779. NO_UNINITIALIZED_WARN = -Wno-uninitialized
  780. ifeq ($(QUIET_MAKE), 1)
  781. CCOMMON_OPT += $(NO_UNINITIALIZED_WARN) -Wno-unused
  782. endif
  783. ifdef NO_BINARY_MODE
  784. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  785. ifdef BINARY64
  786. CCOMMON_OPT += -mabi=64
  787. else
  788. CCOMMON_OPT += -mabi=n32
  789. endif
  790. BINARY_DEFINED = 1
  791. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  792. CCOMMON_OPT += -mabi=32
  793. BINARY_DEFINED = 1
  794. endif
  795. ifneq (, $(filter $(CORE), MIPS64_GENERIC))
  796. CCOMMON_OPT += -DNO_MSA
  797. FCOMMON_OPT += -DNO_MSA
  798. endif
  799. ifneq (, $(filter $(CORE),LOONGSON3R3 LOONGSON3R4))
  800. CCOMMON_OPT += -march=loongson3a
  801. FCOMMON_OPT += -march=loongson3a
  802. endif
  803. ifeq ($(CORE), MIPS24K)
  804. CCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  805. FCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  806. endif
  807. ifeq ($(CORE), MIPS1004K)
  808. CCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  809. FCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  810. endif
  811. ifeq ($(CORE), P5600)
  812. CCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  813. FCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  814. endif
  815. ifeq ($(CORE), I6400)
  816. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  817. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  818. endif
  819. ifeq ($(CORE), P6600)
  820. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  821. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  822. endif
  823. ifeq ($(CORE), I6500)
  824. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  825. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  826. endif
  827. ifeq ($(OSNAME), AIX)
  828. BINARY_DEFINED = 1
  829. endif
  830. ifeq ($(ARCH), loongarch64)
  831. LA64_ABI=$(shell $(CC) -mabi=lp64d -c $(TOPDIR)/cpuid_loongarch64.c -o /dev/null > /dev/null 2> /dev/null && echo lp64d)
  832. LA64_ARCH=$(shell $(CC) -march=loongarch64 -c $(TOPDIR)/cpuid_loongarch64.c -o /dev/null > /dev/null 2> /dev/null && echo loongarch64)
  833. ifneq ($(LA64_ABI), lp64d)
  834. LA64_ABI=lp64
  835. endif
  836. ifneq ($(LA64_ARCH), loongarch64)
  837. CCOMMON_OPT += -mabi=$(LA64_ABI)
  838. FCOMMON_OPT += -mabi=$(LA64_ABI)
  839. else
  840. CCOMMON_OPT += -march=loongarch64 -mabi=$(LA64_ABI)
  841. FCOMMON_OPT += -march=loongarch64 -mabi=$(LA64_ABI)
  842. endif
  843. endif
  844. endif
  845. ifndef BINARY_DEFINED
  846. ifneq ($(OSNAME), AIX)
  847. ifdef BINARY64
  848. ifneq ($(ARCH), riscv64)
  849. CCOMMON_OPT += -m64
  850. endif
  851. else
  852. CCOMMON_OPT += -m32
  853. endif
  854. endif
  855. endif
  856. endif
  857. ifeq ($(C_COMPILER), PGI)
  858. PGCVERSIONGT20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \> 20)
  859. PGCVERSIONEQ20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` == 20)
  860. PGCMINORVERSIONGE11 := $(shell expr `$(CC) --version|sed -n "2p" |cut -d "-" -f 1 |sed -e "s/[^0-9.]//g" |cut -c 4-5` \>= 11)
  861. PGCVERSIONCHECK := $(PGCVERSIONGT20)$(PGCVERSIONEQ20)$(PGCMINORVERSIONGE11)
  862. ifeq ($(PGCVERSIONCHECK), $(filter $(PGCVERSIONCHECK), 100 101 011))
  863. NEWPGI := 1
  864. PGCVERSIONGT21 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \> 21)
  865. PGCVERSIONEQ21 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` == 21)
  866. PGCVERSIONCHECK2 := $(PGCVERSIONGT21)$(PGCVERSIONEQ21)$(PGCMINORVERSIONGE11)
  867. ifeq ($(PGCVERSIONCHECK2), $(filter $(PGCVERSIONCHECK2), 100 101 011))
  868. NEWPGI2 := 1
  869. endif
  870. endif
  871. ifdef BINARY64
  872. ifeq ($(ARCH), x86_64)
  873. ifeq (,$(findstring tp,$(CFLAGS)))
  874. ifneq ($(NEWPGI2),1)
  875. CCOMMON_OPT += -tp p7-64
  876. else
  877. CCOMMON_OPT += -tp px
  878. endif
  879. endif
  880. ifneq ($(NEWPGI),1)
  881. CCOMMON_OPT += -D__MMX__ -Mnollvm
  882. endif
  883. else
  884. ifeq ($(ARCH), power)
  885. ifeq (,$(findstring tp,$(CFLAGS)))
  886. ifeq ($(CORE), POWER8)
  887. CCOMMON_OPT += -tp pwr8
  888. endif
  889. ifeq ($(CORE), POWER9)
  890. CCOMMON_OPT += -tp pwr9
  891. endif
  892. endif
  893. endif
  894. endif
  895. else
  896. ifneq ($(NEWPGI2),1)
  897. ifeq (,$(findstring tp,$(CFLAGS)))
  898. CCOMMON_OPT += -tp p7
  899. else
  900. CCOMMON_OPT += -tp px
  901. endif
  902. endif
  903. endif
  904. endif
  905. ifeq ($(C_COMPILER), PATHSCALE)
  906. ifdef BINARY64
  907. CCOMMON_OPT += -m64
  908. else
  909. CCOMMON_OPT += -m32
  910. endif
  911. endif
  912. #
  913. # Fortran Compiler dependent settings
  914. #
  915. ifeq ($(F_COMPILER), NAG)
  916. FCOMMON_OPT += -dcfuns -recursive -ieee=full -w=obs -thread_safe
  917. ifdef INTERFACE64
  918. ifneq ($(INTERFACE64), 0)
  919. FCOMMON_OPT += -i8
  920. endif
  921. endif
  922. ifeq ($(USE_OPENMP), 1)
  923. FCOMMON_OPT += -openmp
  924. endif
  925. endif
  926. ifeq ($(F_COMPILER), FLANG)
  927. CCOMMON_OPT += -DF_INTERFACE_FLANG
  928. FCOMMON_OPT += -Mrecursive -Kieee
  929. ifeq ($(OSNAME), Linux)
  930. ifeq ($(ARCH), x86_64)
  931. FLANG_VENDOR := $(shell $(FC) --version|head -1 |cut -f 1 -d " ")
  932. ifeq ($(FLANG_VENDOR), AMD)
  933. FCOMMON_OPT += -fno-unroll-loops
  934. endif
  935. endif
  936. endif
  937. ifdef BINARY64
  938. ifdef INTERFACE64
  939. ifneq ($(INTERFACE64), 0)
  940. FCOMMON_OPT += -i8
  941. endif
  942. endif
  943. FCOMMON_OPT += -Wall
  944. else
  945. FCOMMON_OPT += -Wall
  946. endif
  947. ifeq ($(USE_OPENMP), 1)
  948. FCOMMON_OPT += -fopenmp
  949. endif
  950. endif
  951. ifeq ($(F_COMPILER), G77)
  952. CCOMMON_OPT += -DF_INTERFACE_G77
  953. FCOMMON_OPT += -Wall
  954. ifndef NO_BINARY_MODE
  955. ifneq ($(OSNAME), AIX)
  956. ifdef BINARY64
  957. FCOMMON_OPT += -m64
  958. else
  959. FCOMMON_OPT += -m32
  960. endif
  961. endif
  962. endif
  963. endif
  964. ifeq ($(F_COMPILER), G95)
  965. CCOMMON_OPT += -DF_INTERFACE_G95
  966. FCOMMON_OPT += -Wall
  967. ifneq ($(OSNAME), AIX)
  968. ifndef NO_BINARY_MODE
  969. ifdef BINARY64
  970. FCOMMON_OPT += -m64
  971. else
  972. FCOMMON_OPT += -m32
  973. endif
  974. endif
  975. ifneq ($(NO_LAPACKE), 1)
  976. FCOMMON_OPT += -fno-second-underscore
  977. endif
  978. endif
  979. endif
  980. ifeq ($(F_COMPILER), $(filter $(F_COMPILER),GFORTRAN FLANGNEW))
  981. CCOMMON_OPT += -DF_INTERFACE_GFORT
  982. ifeq ($(F_COMPILER), GFORTRAN)
  983. FCOMMON_OPT += -Wall
  984. # make single-threaded LAPACK calls thread-safe #1847
  985. FCOMMON_OPT += -frecursive
  986. # work around ABI problem with passing single-character arguments
  987. FCOMMON_OPT += -fno-optimize-sibling-calls
  988. #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
  989. ifneq ($(NOFORTRAN), 1)
  990. ifneq ($(NOFORTRAN), 2)
  991. ifneq ($(NO_LAPACK), 1)
  992. EXTRALIB += -lgfortran
  993. endif
  994. endif
  995. endif
  996. endif
  997. ifdef NO_BINARY_MODE
  998. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  999. ifdef BINARY64
  1000. FCOMMON_OPT += -mabi=64
  1001. else
  1002. FCOMMON_OPT += -mabi=n32
  1003. endif
  1004. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  1005. FCOMMON_OPT += -mabi=32
  1006. endif
  1007. else
  1008. ifdef BINARY64
  1009. ifneq ($(OSNAME), AIX)
  1010. ifneq ($(ARCH), riscv64)
  1011. FCOMMON_OPT += -m64
  1012. endif
  1013. endif
  1014. ifdef INTERFACE64
  1015. ifneq ($(INTERFACE64), 0)
  1016. FCOMMON_OPT += -fdefault-integer-8
  1017. endif
  1018. endif
  1019. else
  1020. ifneq ($(OSNAME), AIX)
  1021. FCOMMON_OPT += -m32
  1022. endif
  1023. endif
  1024. endif
  1025. ifeq ($(USE_OPENMP), 1)
  1026. FCOMMON_OPT += -fopenmp
  1027. endif
  1028. endif
  1029. ifeq ($(F_COMPILER), INTEL)
  1030. CCOMMON_OPT += -DF_INTERFACE_INTEL
  1031. ifdef INTERFACE64
  1032. ifneq ($(INTERFACE64), 0)
  1033. FCOMMON_OPT += -i8
  1034. endif
  1035. endif
  1036. FCOMMON_OPT += -recursive -fp-model strict -assume protect-parens
  1037. ifeq ($(USE_OPENMP), 1)
  1038. FCOMMON_OPT += -fopenmp
  1039. endif
  1040. endif
  1041. ifeq ($(F_COMPILER), FUJITSU)
  1042. CCOMMON_OPT += -DF_INTERFACE_FUJITSU
  1043. ifeq ($(USE_OPENMP), 1)
  1044. FCOMMON_OPT += -openmp
  1045. endif
  1046. endif
  1047. ifeq ($(F_COMPILER), IBM)
  1048. CCOMMON_OPT += -DF_INTERFACE_IBM
  1049. FEXTRALIB += -lxlf90
  1050. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG))
  1051. FCOMMON_OPT += -qextname -qzerosize
  1052. endif
  1053. # FCOMMON_OPT += -qarch=440
  1054. ifdef BINARY64
  1055. FCOMMON_OPT += -q64
  1056. ifdef INTERFACE64
  1057. ifneq ($(INTERFACE64), 0)
  1058. FCOMMON_OPT += -qintsize=8
  1059. endif
  1060. endif
  1061. else
  1062. FCOMMON_OPT += -q32
  1063. endif
  1064. endif
  1065. ifeq ($(F_COMPILER), PGI)
  1066. CCOMMON_OPT += -DF_INTERFACE_PGI
  1067. COMMON_PROF += -DPGICOMPILER
  1068. ifdef BINARY64
  1069. ifdef INTERFACE64
  1070. ifneq ($(INTERFACE64), 0)
  1071. FCOMMON_OPT += -i8
  1072. endif
  1073. endif
  1074. ifeq ($(ARCH), x86_64)
  1075. ifneq ($(NEWPGI2),1)
  1076. FCOMMON_OPT += -tp p7-64
  1077. else
  1078. FCOMMON_OPT += -tp px
  1079. endif
  1080. else
  1081. ifeq ($(ARCH), power)
  1082. ifeq ($(CORE), POWER6)
  1083. $(warning NVIDIA HPC compilers do not support POWER6.)
  1084. endif
  1085. ifeq ($(CORE), POWER8)
  1086. FCOMMON_OPT += -tp pwr8
  1087. endif
  1088. ifeq ($(CORE), POWER9)
  1089. FCOMMON_OPT += -tp pwr9
  1090. endif
  1091. ifeq ($(CORE), POWER10)
  1092. $(warning NVIDIA HPC compilers do not support POWER10.)
  1093. endif
  1094. endif
  1095. endif
  1096. else
  1097. FCOMMON_OPT += -tp p7
  1098. endif
  1099. FCOMMON_OPT += -Mrecursive -Kieee
  1100. ifeq ($(USE_OPENMP), 1)
  1101. FCOMMON_OPT += -mp
  1102. endif
  1103. endif
  1104. ifeq ($(F_COMPILER), PATHSCALE)
  1105. CCOMMON_OPT += -DF_INTERFACE_PATHSCALE
  1106. ifdef BINARY64
  1107. ifdef INTERFACE64
  1108. ifneq ($(INTERFACE64), 0)
  1109. FCOMMON_OPT += -i8
  1110. endif
  1111. endif
  1112. endif
  1113. ifeq ($(USE_OPENMP), 1)
  1114. FCOMMON_OPT += -mp
  1115. endif
  1116. endif
  1117. ifeq ($(F_COMPILER), OPEN64)
  1118. CCOMMON_OPT += -DF_INTERFACE_OPEN64
  1119. ifdef BINARY64
  1120. ifdef INTERFACE64
  1121. ifneq ($(INTERFACE64), 0)
  1122. FCOMMON_OPT += -i8
  1123. endif
  1124. endif
  1125. endif
  1126. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  1127. ifndef BINARY64
  1128. FCOMMON_OPT += -n32
  1129. else
  1130. FCOMMON_OPT += -n64
  1131. endif
  1132. ifeq ($(CORE), LOONGSON3R3)
  1133. FCOMMON_OPT += -loongson3 -static
  1134. endif
  1135. ifeq ($(CORE), LOONGSON3R4)
  1136. FCOMMON_OPT += -loongson3 -static
  1137. endif
  1138. else
  1139. ifndef BINARY64
  1140. FCOMMON_OPT += -m32
  1141. else
  1142. FCOMMON_OPT += -m64
  1143. endif
  1144. endif
  1145. ifeq ($(USE_OPENMP), 1)
  1146. FEXTRALIB += -lstdc++
  1147. FCOMMON_OPT += -mp
  1148. endif
  1149. endif
  1150. ifeq ($(C_COMPILER), OPEN64)
  1151. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  1152. ifndef BINARY64
  1153. CCOMMON_OPT += -n32
  1154. else
  1155. CCOMMON_OPT += -n64
  1156. endif
  1157. ifeq ($(CORE), LOONGSON3R3)
  1158. CCOMMON_OPT += -loongson3 -static
  1159. endif
  1160. ifeq ($(CORE), LOONGSON3R4)
  1161. CCOMMON_OPT += -loongson3 -static
  1162. endif
  1163. else
  1164. ifndef BINARY64
  1165. CCOMMON_OPT += -m32
  1166. else
  1167. CCOMMON_OPT += -m64
  1168. endif
  1169. endif
  1170. endif
  1171. ifeq ($(C_COMPILER), SUN)
  1172. CCOMMON_OPT += -w
  1173. ifeq ($(ARCH), x86)
  1174. CCOMMON_OPT += -m32
  1175. else
  1176. ifdef BINARY64
  1177. CCOMMON_OPT += -m64
  1178. else
  1179. CCOMMON_OPT += -m32
  1180. endif
  1181. endif
  1182. endif
  1183. ifeq ($(F_COMPILER), SUN)
  1184. CCOMMON_OPT += -DF_INTERFACE_SUN
  1185. FCOMMON_OPT += -ftrap=%none -xrecursive
  1186. ifeq ($(ARCH), x86)
  1187. FCOMMON_OPT += -m32
  1188. else
  1189. ifdef BINARY64
  1190. FCOMMON_OPT += -m64
  1191. else
  1192. FCOMMON_OPT += -m32
  1193. endif
  1194. endif
  1195. ifeq ($(USE_OPENMP), 1)
  1196. FCOMMON_OPT += -xopenmp=parallel
  1197. endif
  1198. endif
  1199. ifeq ($(F_COMPILER), COMPAQ)
  1200. CCOMMON_OPT += -DF_INTERFACE_COMPAQ
  1201. ifeq ($(USE_OPENMP), 1)
  1202. FCOMMON_OPT += -openmp
  1203. endif
  1204. endif
  1205. ifeq ($(F_COMPILER), CRAY)
  1206. CCOMMON_OPT += -DF_INTERFACE_INTEL
  1207. FCOMMON_OPT += -hnopattern
  1208. ifdef INTERFACE64
  1209. ifneq ($(INTERFACE64), 0)
  1210. FCOMMON_OPT += -s integer64
  1211. endif
  1212. endif
  1213. ifneq ($(USE_OPENMP), 1)
  1214. FCOMMON_OPT += -O noomp
  1215. endif
  1216. endif
  1217. ifdef BINARY64
  1218. ifdef INTERFACE64
  1219. ifneq ($(INTERFACE64), 0)
  1220. CCOMMON_OPT +=
  1221. #-DUSE64BITINT
  1222. endif
  1223. endif
  1224. endif
  1225. ifeq ($(NEED_PIC), 1)
  1226. ifeq ($(C_COMPILER), IBM)
  1227. CCOMMON_OPT += -qpic=large
  1228. else
  1229. CCOMMON_OPT += -fPIC
  1230. endif
  1231. ifeq ($(F_COMPILER), SUN)
  1232. FCOMMON_OPT += -pic
  1233. else ifeq ($(F_COMPILER), NAG)
  1234. FCOMMON_OPT += -PIC
  1235. else ifeq ($(F_COMPILER), IBM)
  1236. FCOMMON_OPT += -qpic=large
  1237. else
  1238. FCOMMON_OPT += -fPIC
  1239. endif
  1240. endif
  1241. ifeq ($(DYNAMIC_ARCH), 1)
  1242. CCOMMON_OPT += -DDYNAMIC_ARCH
  1243. endif
  1244. ifeq ($(DYNAMIC_OLDER), 1)
  1245. CCOMMON_OPT += -DDYNAMIC_OLDER
  1246. endif
  1247. ifeq ($(C_LAPACK), 1)
  1248. CCOMMON_OPT += -DC_LAPACK
  1249. endif
  1250. ifeq ($(NO_LAPACK), 1)
  1251. CCOMMON_OPT += -DNO_LAPACK
  1252. #Disable LAPACK C interface
  1253. NO_LAPACKE = 1
  1254. endif
  1255. ifeq ($(NO_LAPACKE), 1)
  1256. CCOMMON_OPT += -DNO_LAPACKE
  1257. endif
  1258. ifeq ($(NO_AVX), 1)
  1259. CCOMMON_OPT += -DNO_AVX
  1260. endif
  1261. ifeq ($(ARCH), x86)
  1262. CCOMMON_OPT += -DNO_AVX
  1263. endif
  1264. ifeq ($(NO_AVX2), 1)
  1265. CCOMMON_OPT += -DNO_AVX2
  1266. endif
  1267. ifeq ($(NO_AVX512), 1)
  1268. CCOMMON_OPT += -DNO_AVX512
  1269. endif
  1270. ifeq ($(NO_SVE), 1)
  1271. CCOMMON_OPT += -DNO_SVE
  1272. endif
  1273. ifdef SMP
  1274. CCOMMON_OPT += -DSMP_SERVER
  1275. ifeq ($(ARCH), mips64)
  1276. USE_SIMPLE_THREADED_LEVEL3 = 1
  1277. endif
  1278. ifeq ($(USE_OPENMP), 1)
  1279. # USE_SIMPLE_THREADED_LEVEL3 = 1
  1280. # NO_AFFINITY = 1
  1281. CCOMMON_OPT += -DUSE_OPENMP
  1282. endif
  1283. ifeq ($(BIGNUMA), 1)
  1284. CCOMMON_OPT += -DBIGNUMA
  1285. endif
  1286. endif
  1287. ifeq ($(NO_WARMUP), 1)
  1288. CCOMMON_OPT += -DNO_WARMUP
  1289. endif
  1290. ifeq ($(CONSISTENT_FPCSR), 1)
  1291. CCOMMON_OPT += -DCONSISTENT_FPCSR
  1292. endif
  1293. # Only for development
  1294. # CCOMMON_OPT += -DPARAMTEST
  1295. # CCOMMON_OPT += -DPREFETCHTEST
  1296. # CCOMMON_OPT += -DNO_SWITCHING
  1297. # USE_PAPI = 1
  1298. ifdef USE_PAPI
  1299. CCOMMON_OPT += -DUSE_PAPI
  1300. EXTRALIB += -lpapi -lperfctr
  1301. endif
  1302. ifdef BUFFERSIZE
  1303. CCOMMON_OPT += -DBUFFERSIZE=$(BUFFERSIZE)
  1304. endif
  1305. ifdef DYNAMIC_THREADS
  1306. CCOMMON_OPT += -DDYNAMIC_THREADS
  1307. endif
  1308. CCOMMON_OPT += -DMAX_CPU_NUMBER=$(NUM_THREADS)
  1309. CCOMMON_OPT += -DMAX_PARALLEL_NUMBER=$(NUM_PARALLEL)
  1310. ifdef USE_SIMPLE_THREADED_LEVEL3
  1311. CCOMMON_OPT += -DUSE_SIMPLE_THREADED_LEVEL3
  1312. endif
  1313. ifeq ($(USE_TLS), 1)
  1314. CCOMMON_OPT += -DUSE_TLS
  1315. endif
  1316. ifeq ($(BUILD_BFLOAT16), 1)
  1317. CCOMMON_OPT += -DBUILD_BFLOAT16
  1318. endif
  1319. ifeq ($(BUILD_SINGLE), 1)
  1320. CCOMMON_OPT += -DBUILD_SINGLE=1
  1321. endif
  1322. ifeq ($(BUILD_DOUBLE), 1)
  1323. CCOMMON_OPT += -DBUILD_DOUBLE=1
  1324. endif
  1325. ifeq ($(BUILD_COMPLEX), 1)
  1326. CCOMMON_OPT += -DBUILD_COMPLEX=1
  1327. endif
  1328. ifeq ($(BUILD_COMPLEX16), 1)
  1329. CCOMMON_OPT += -DBUILD_COMPLEX16=1
  1330. endif
  1331. CCOMMON_OPT += -DVERSION=\"$(VERSION)\"
  1332. ifndef SYMBOLPREFIX
  1333. SYMBOLPREFIX =
  1334. endif
  1335. ifndef SYMBOLSUFFIX
  1336. SYMBOLSUFFIX =
  1337. endif
  1338. ifndef LIBSONAMEBASE
  1339. LIBSONAMEBASE = openblas
  1340. endif
  1341. ifndef LIBNAMEPREFIX
  1342. LIBNAMEPREFIX =
  1343. endif
  1344. SYMPREFIX=$(SYMBOLPREFIX)
  1345. ifeq ($(SYMBOLPREFIX),$(LIBNAMEPREFIX))
  1346. SYMPREFIX=
  1347. endif
  1348. SYMSUFFIX=$(SYMBOLSUFFIX)
  1349. ifeq ($(SYMBOLSUFFIX),$(LIBNAMESUFFIX))
  1350. SYMSUFFIX=
  1351. endif
  1352. ifndef LIBNAMESUFFIX
  1353. LIBNAMEBASE = $(SYMPREFIX)$(LIBSONAMEBASE)$(SYMSUFFIX)
  1354. else
  1355. LIBNAMEBASE = $(SYMPREFIX)$(LIBSONAMEBASE)$(SYMSUFFIX)$(LIBNAMESUFFIX)
  1356. endif
  1357. ifeq ($(OSNAME), CYGWIN_NT)
  1358. LIBPREFIX = cyg$(LIBNAMEPREFIX)$(LIBNAMEBASE)
  1359. else
  1360. LIBPREFIX = lib$(LIBNAMEPREFIX)$(LIBNAMEBASE)
  1361. endif
  1362. KERNELDIR = $(TOPDIR)/kernel/$(ARCH)
  1363. include $(TOPDIR)/Makefile.$(ARCH)
  1364. ifneq ($(C_COMPILER), PGI)
  1365. ifneq ($(C_COMPILER), SUN)
  1366. CCOMMON_OPT += -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME
  1367. endif
  1368. endif
  1369. CCOMMON_OPT += -DASMNAME=$(FU)$(*F) -DASMFNAME=$(FU)$(*F)$(BU) -DNAME=$(*F)$(BU) -DCNAME=$(*F) -DCHAR_NAME=\"$(*F)$(BU)\" -DCHAR_CNAME=\"$(*F)\"
  1370. ifeq ($(CORE), PPC440)
  1371. CCOMMON_OPT += -DALLOC_QALLOC
  1372. endif
  1373. ifeq ($(CORE), PPC440FP2)
  1374. STATIC_ALLOCATION = 1
  1375. endif
  1376. ifneq ($(OSNAME), Linux)
  1377. NO_AFFINITY = 1
  1378. endif
  1379. ifneq ($(ARCH), x86_64)
  1380. ifneq ($(ARCH), x86)
  1381. NO_AFFINITY = 1
  1382. endif
  1383. endif
  1384. ifdef NO_AFFINITY
  1385. ifeq ($(NO_AFFINITY), 0)
  1386. override undefine NO_AFFINITY
  1387. else
  1388. CCOMMON_OPT += -DNO_AFFINITY
  1389. endif
  1390. endif
  1391. ifdef FUNCTION_PROFILE
  1392. CCOMMON_OPT += -DFUNCTION_PROFILE
  1393. endif
  1394. ifdef SHMEM_ALLOCATION
  1395. ifneq ($(SHMEM_ALLOCATION), 0)
  1396. CCOMMON_OPT += -DALLOC_SHM
  1397. endif
  1398. endif
  1399. ifdef HUGETLB_ALLOCATION
  1400. ifneq ($(HUGETLB_ALLOCATION), 0)
  1401. CCOMMON_OPT += -DALLOC_HUGETLB
  1402. endif
  1403. endif
  1404. ifdef HUGETLBFILE_ALLOCATION
  1405. ifneq ($(HUGETLBFILE_ALLOCATION), 0)
  1406. CCOMMON_OPT += -DALLOC_HUGETLBFILE -DHUGETLB_FILE_NAME=$(HUGETLBFILE_ALLOCATION)
  1407. endif
  1408. endif
  1409. ifdef STATIC_ALLOCATION
  1410. CCOMMON_OPT += -DALLOC_STATIC
  1411. endif
  1412. ifdef DEVICEDRIVER_ALLOCATION
  1413. CCOMMON_OPT += -DALLOC_DEVICEDRIVER -DDEVICEDRIVER_NAME=\"/dev/mapper\"
  1414. endif
  1415. ifdef MIXED_MEMORY_ALLOCATION
  1416. CCOMMON_OPT += -DMIXED_MEMORY_ALLOCATION
  1417. endif
  1418. ifeq ($(OSNAME), SunOS)
  1419. TAR = gtar
  1420. PATCH = gpatch
  1421. GREP = ggrep
  1422. AWK = nawk
  1423. else
  1424. TAR = tar
  1425. PATCH = patch
  1426. GREP = grep
  1427. AWK = awk
  1428. endif
  1429. ifndef MD5SUM
  1430. MD5SUM = md5sum
  1431. endif
  1432. REVISION = -r$(VERSION)
  1433. MAJOR_VERSION = $(word 1,$(subst ., ,$(VERSION)))
  1434. ifeq ($(DEBUG), 1)
  1435. COMMON_OPT += -g
  1436. endif
  1437. ifeq ($(DEBUG), 1)
  1438. FCOMMON_OPT += -g
  1439. endif
  1440. ifndef COMMON_OPT
  1441. COMMON_OPT = -O2
  1442. endif
  1443. ifndef FCOMMON_OPT
  1444. FCOMMON_OPT = -O2 -frecursive
  1445. endif
  1446. override CFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
  1447. override PFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
  1448. override FFLAGS += $(COMMON_OPT) $(FCOMMON_OPT)
  1449. override FPFLAGS += $(FCOMMON_OPT) $(COMMON_PROF)
  1450. #MAKEOVERRIDES =
  1451. ifeq ($(NEED_PIC), 1)
  1452. ifeq (,$(findstring PIC,$(FFLAGS)))
  1453. ifneq ($(F_COMPILER),IBM)
  1454. override FFLAGS += -fPIC
  1455. endif
  1456. endif
  1457. endif
  1458. #For LAPACK Fortran codes.
  1459. #Disable -fopenmp for LAPACK Fortran codes on Windows.
  1460. ifdef OS_WINDOWS
  1461. LAPACK_FFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FFLAGS))
  1462. LAPACK_FPFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FPFLAGS))
  1463. else
  1464. LAPACK_FFLAGS := $(FFLAGS)
  1465. LAPACK_FPFLAGS := $(FPFLAGS)
  1466. endif
  1467. ifeq ($(F_COMPILER),NAG)
  1468. LAPACK_FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1469. override FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1470. endif
  1471. ifeq ($(F_COMPILER),CRAY)
  1472. LAPACK_FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1473. override FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1474. endif
  1475. ifeq ($(F_COMPILER),FLANGNEW)
  1476. LAPACK_FFLAGS := $(filter-out -m32 -m64 -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1477. override FFLAGS := $(filter-out -m32 -m64 -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1478. endif
  1479. LAPACK_CFLAGS = $(CFLAGS)
  1480. LAPACK_CFLAGS += -DHAVE_LAPACK_CONFIG_H
  1481. ifdef INTERFACE64
  1482. ifneq ($(INTERFACE64), 0)
  1483. LAPACK_CFLAGS += -DLAPACK_ILP64
  1484. endif
  1485. endif
  1486. ifdef OS_WINDOWS
  1487. LAPACK_CFLAGS += -DOPENBLAS_OS_WINDOWS
  1488. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1489. endif
  1490. ifeq ($(C_COMPILER), LSB)
  1491. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1492. endif
  1493. ifndef SUFFIX
  1494. SUFFIX = o
  1495. endif
  1496. ifndef PSUFFIX
  1497. PSUFFIX = po
  1498. endif
  1499. ifndef LIBSUFFIX
  1500. LIBSUFFIX = a
  1501. endif
  1502. ifneq ($(DYNAMIC_ARCH), 1)
  1503. ifndef SMP
  1504. LIBNAME = $(LIBPREFIX)_$(LIBCORE)$(REVISION).$(LIBSUFFIX)
  1505. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)$(REVISION)_p.$(LIBSUFFIX)
  1506. else
  1507. LIBNAME = $(LIBPREFIX)_$(LIBCORE)p$(REVISION).$(LIBSUFFIX)
  1508. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)p$(REVISION)_p.$(LIBSUFFIX)
  1509. endif
  1510. else
  1511. ifndef SMP
  1512. LIBNAME = $(LIBPREFIX)$(REVISION).$(LIBSUFFIX)
  1513. LIBNAME_P = $(LIBPREFIX)$(REVISION)_p.$(LIBSUFFIX)
  1514. else
  1515. LIBNAME = $(LIBPREFIX)p$(REVISION).$(LIBSUFFIX)
  1516. LIBNAME_P = $(LIBPREFIX)p$(REVISION)_p.$(LIBSUFFIX)
  1517. endif
  1518. endif
  1519. ifeq ($(FIXED_LIBNAME),1)
  1520. LIBNAME = lib$(LIBNAMEPREFIX)$(LIBSONAMEBASE)$(LIBNAMESUFFIX).$(LIBSUFFIX)
  1521. LIBNAME_P = lib$(LIBNAMEPREFIX)$(LISOBNAMEBASE)$(LIBNAMESUFFIX)_p.$(LIBSUFFIX)
  1522. endif
  1523. LIBDLLNAME = $(LIBPREFIX).dll
  1524. IMPLIBNAME = lib$(LIBNAMEBASE).dll.a
  1525. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.so)
  1526. LIBDYNNAME = $(LIBNAME:.$(LIBSUFFIX)=.dylib)
  1527. LIBDEFNAME = $(LIBNAME:.$(LIBSUFFIX)=.def)
  1528. LIBEXPNAME = $(LIBNAME:.$(LIBSUFFIX)=.exp)
  1529. LIBZIPNAME = $(LIBNAME:.$(LIBSUFFIX)=.zip)
  1530. LIBS = $(TOPDIR)/$(LIBNAME)
  1531. LIBS_P = $(TOPDIR)/$(LIBNAME_P)
  1532. LIB_COMPONENTS = BLAS
  1533. ifneq ($(NO_CBLAS), 1)
  1534. LIB_COMPONENTS += CBLAS
  1535. endif
  1536. ifneq ($(NO_LAPACK), 1)
  1537. LIB_COMPONENTS += LAPACK
  1538. ifneq ($(NO_LAPACKE), 1)
  1539. LIB_COMPONENTS += LAPACKE
  1540. endif
  1541. ifeq ($(BUILD_RELAPACK), 1)
  1542. LIB_COMPONENTS += ReLAPACK
  1543. endif
  1544. endif
  1545. ifeq ($(ONLY_CBLAS), 1)
  1546. LIB_COMPONENTS = CBLAS
  1547. endif
  1548. export OSNAME
  1549. export ARCH
  1550. export CORE
  1551. export LIBCORE
  1552. export __BYTE_ORDER__
  1553. export ELF_VERSION
  1554. export PGCPATH
  1555. export CONFIG
  1556. export CC
  1557. export FC
  1558. export BU
  1559. export FU
  1560. export NEED2UNDERSCORES
  1561. export USE_THREAD
  1562. export NUM_THREADS
  1563. export NUM_CORES
  1564. export SMP
  1565. export MAKEFILE_RULE
  1566. export NEED_PIC
  1567. export BINARY
  1568. export BINARY32
  1569. export BINARY64
  1570. export F_COMPILER
  1571. export C_COMPILER
  1572. export USE_OPENMP
  1573. export CROSS
  1574. export CROSS_SUFFIX
  1575. export NOFORTRAN
  1576. export C_LAPACK
  1577. export NO_FBLAS
  1578. export EXTRALIB
  1579. export CEXTRALIB
  1580. export FEXTRALIB
  1581. export HAVE_SSE
  1582. export HAVE_SSE2
  1583. export HAVE_SSE3
  1584. export HAVE_SSSE3
  1585. export HAVE_SSE4_1
  1586. export HAVE_SSE4_2
  1587. export HAVE_SSE4A
  1588. export HAVE_SSE5
  1589. export HAVE_AVX
  1590. export HAVE_AVX2
  1591. export HAVE_FMA3
  1592. export HAVE_VFP
  1593. export HAVE_VFPV3
  1594. export HAVE_VFPV4
  1595. export HAVE_NEON
  1596. ifndef NO_MSA
  1597. export HAVE_MSA
  1598. export MSA_FLAGS
  1599. endif
  1600. export KERNELDIR
  1601. export FUNCTION_PROFILE
  1602. export TARGET_CORE
  1603. export NO_AVX512
  1604. export NO_AVX2
  1605. export BUILD_BFLOAT16
  1606. export NO_LSX
  1607. export NO_LASX
  1608. export SBGEMM_UNROLL_M
  1609. export SBGEMM_UNROLL_N
  1610. export SGEMM_UNROLL_M
  1611. export SGEMM_UNROLL_N
  1612. export DGEMM_UNROLL_M
  1613. export DGEMM_UNROLL_N
  1614. export QGEMM_UNROLL_M
  1615. export QGEMM_UNROLL_N
  1616. export CGEMM_UNROLL_M
  1617. export CGEMM_UNROLL_N
  1618. export ZGEMM_UNROLL_M
  1619. export ZGEMM_UNROLL_N
  1620. export XGEMM_UNROLL_M
  1621. export XGEMM_UNROLL_N
  1622. export CGEMM3M_UNROLL_M
  1623. export CGEMM3M_UNROLL_N
  1624. export ZGEMM3M_UNROLL_M
  1625. export ZGEMM3M_UNROLL_N
  1626. export XGEMM3M_UNROLL_M
  1627. export XGEMM3M_UNROLL_N
  1628. ifdef USE_CUDA
  1629. export CUDADIR
  1630. export CUCC
  1631. export CUFLAGS
  1632. export CULIB
  1633. endif
  1634. .SUFFIXES: .$(PSUFFIX) .$(SUFFIX) .f
  1635. .f.$(SUFFIX):
  1636. $(FC) $(FFLAGS) -c $< -o $(@F)
  1637. .f.$(PSUFFIX):
  1638. $(FC) $(FPFLAGS) -pg -c $< -o $(@F)
  1639. ifdef BINARY64
  1640. PATHSCALEPATH = /opt/pathscale/lib/3.1
  1641. PGIPATH = /opt/pgi/linux86-64/7.1-5/lib
  1642. else
  1643. PATHSCALEPATH = /opt/pathscale/lib/3.1/32
  1644. PGIPATH = /opt/pgi/linux86/7.1-5/lib
  1645. endif
  1646. ACMLPATH = /opt/acml/4.3.0
  1647. ifneq ($(OSNAME), Darwin)
  1648. MKLPATH = /opt/intel/mkl/10.2.2.025/lib
  1649. else
  1650. MKLPATH = /Library/Frameworks/Intel_MKL.framework/Versions/10.0.1.014/lib
  1651. endif
  1652. ATLASPATH = /opt/atlas/3.9.17/opteron
  1653. FLAMEPATH = $(HOME)/flame/lib
  1654. ifneq ($(OSNAME), SunOS)
  1655. SUNPATH = /opt/sunstudio12.1
  1656. else
  1657. SUNPATH = /opt/SUNWspro
  1658. endif