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 36 kB

7 years ago
4 years ago
13 years ago
4 years ago
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942943944945946947948949950951952953954955956957958959960961962963964965966967968969970971972973974975976977978979980981982983984985986987988989990991992993994995996997998999100010011002100310041005100610071008100910101011101210131014101510161017101810191020102110221023102410251026102710281029103010311032103310341035103610371038103910401041104210431044104510461047104810491050105110521053105410551056105710581059106010611062106310641065106610671068106910701071107210731074107510761077107810791080108110821083108410851086108710881089109010911092109310941095109610971098109911001101110211031104110511061107110811091110111111121113111411151116111711181119112011211122112311241125112611271128112911301131113211331134113511361137113811391140114111421143114411451146114711481149115011511152115311541155115611571158115911601161116211631164116511661167116811691170117111721173117411751176117711781179118011811182118311841185118611871188118911901191119211931194119511961197119811991200120112021203120412051206120712081209121012111212121312141215121612171218121912201221122212231224122512261227122812291230123112321233123412351236123712381239124012411242124312441245124612471248124912501251125212531254125512561257125812591260126112621263126412651266126712681269127012711272127312741275127612771278127912801281128212831284128512861287128812891290129112921293129412951296129712981299130013011302130313041305130613071308130913101311131213131314131513161317131813191320132113221323132413251326132713281329133013311332133313341335133613371338133913401341134213431344134513461347134813491350135113521353135413551356135713581359136013611362136313641365136613671368136913701371137213731374137513761377137813791380138113821383138413851386138713881389139013911392139313941395139613971398139914001401140214031404140514061407140814091410141114121413141414151416141714181419142014211422142314241425142614271428142914301431143214331434143514361437143814391440144114421443144414451446144714481449145014511452145314541455145614571458145914601461146214631464146514661467146814691470147114721473147414751476147714781479148014811482148314841485148614871488148914901491149214931494149514961497149814991500150115021503150415051506150715081509151015111512151315141515151615171518151915201521152215231524152515261527152815291530153115321533153415351536153715381539154015411542154315441545154615471548154915501551155215531554155515561557155815591560156115621563156415651566156715681569157015711572157315741575157615771578157915801581158215831584158515861587158815891590159115921593159415951596159715981599160016011602160316041605160616071608160916101611161216131614161516161617161816191620162116221623162416251626162716281629163016311632163316341635163616371638163916401641164216431644164516461647164816491650165116521653165416551656165716581659166016611662166316641665166616671668166916701671167216731674167516761677167816791680168116821683168416851686168716881689169016911692169316941695169616971698169917001701170217031704170517061707170817091710171117121713171417151716171717181719172017211722172317241725172617271728172917301731173217331734173517361737173817391740174117421743174417451746174717481749175017511752
  1. #
  2. # Include user definition
  3. #
  4. # TO suppress recursive includes
  5. INCLUDED = 1
  6. ifndef TOPDIR
  7. TOPDIR = .
  8. endif
  9. # we need to use the host system's architecture for getarch compile options even especially when cross-compiling
  10. HOSTARCH := $(shell uname -m)
  11. ifeq ($(HOSTARCH), amd64)
  12. HOSTARCH=x86_64
  13. endif
  14. # Catch conflicting usage of ARCH in some BSD environments
  15. ifeq ($(ARCH), amd64)
  16. override ARCH=x86_64
  17. else ifeq ($(ARCH), powerpc64)
  18. override ARCH=power
  19. else ifeq ($(ARCH), powerpc64le)
  20. override ARCH=power
  21. else ifeq ($(ARCH), powerpc)
  22. override ARCH=power
  23. else ifeq ($(ARCH), i386)
  24. override ARCH=x86
  25. else ifeq ($(ARCH), armv6)
  26. override ARCH=arm
  27. else ifeq ($(ARCH), armv7)
  28. override ARCH=arm
  29. else ifeq ($(ARCH), aarch64)
  30. override ARCH=arm64
  31. else ifeq ($(ARCH), mipsel)
  32. override ARCH=mips
  33. else ifeq ($(ARCH), mips64el)
  34. override ARCH=mips64
  35. else ifeq ($(ARCH), zarch)
  36. override ARCH=zarch
  37. endif
  38. NETLIB_LAPACK_DIR = $(TOPDIR)/lapack-netlib
  39. # Default C compiler
  40. # - Only set if not specified on the command line or inherited from the environment.
  41. # - CC is an implicit variable so neither '?=' or 'ifndef' can be used.
  42. # http://stackoverflow.com/questions/4029274/mingw-and-make-variables
  43. # - Default value is 'cc' which is not always a valid command (e.g. MinGW).
  44. ifeq ($(origin CC),default)
  45. # Check if $(CC) refers to a valid command and set the value to gcc if not
  46. ifneq ($(findstring cmd.exe,$(SHELL)),)
  47. ifeq ($(shell where $(CC) 2>NUL),)
  48. CC = gcc
  49. endif
  50. else # POSIX-ish
  51. ifeq ($(shell command -v $(CC) 2>/dev/null),)
  52. ifeq ($(shell uname -s),Darwin)
  53. CC = clang
  54. # EXTRALIB += -Wl,-no_compact_unwind
  55. else
  56. CC = gcc
  57. endif # Darwin
  58. endif # CC exists
  59. endif # Shell is sane
  60. endif # CC is set to default
  61. # Default Fortran compiler (FC) is selected by f_check.
  62. ifndef MAKEFILE_RULE
  63. include $(TOPDIR)/Makefile.rule
  64. else
  65. include $(TOPDIR)/$(MAKEFILE_RULE)
  66. endif
  67. #
  68. # Beginning of system configuration
  69. #
  70. ifneq ($(BUILD_SINGLE),1)
  71. ifneq ($(BUILD_DOUBLE),1)
  72. ifneq ($(BUILD_COMPLEX),1)
  73. ifneq ($(BUILD_COMPLEX16),1)
  74. override BUILD_SINGLE=1
  75. override BUILD_DOUBLE=1
  76. override BUILD_COMPLEX=1
  77. override BUILD_COMPLEX16=1
  78. endif
  79. endif
  80. endif
  81. endif
  82. ifndef HOSTCC
  83. HOSTCC = $(CC)
  84. endif
  85. ifdef TARGET
  86. GETARCH_FLAGS := -DFORCE_$(TARGET)
  87. GETARCH_FLAGS += -DUSER_TARGET
  88. ifeq ($(TARGET), GENERIC)
  89. ifeq ($(DYNAMIC_ARCH), 1)
  90. override NO_EXPRECISION=1
  91. export NO_EXPRECISION
  92. endif
  93. endif
  94. endif
  95. # Force fallbacks for 32bit
  96. ifeq ($(BINARY), 32)
  97. ifeq ($(TARGET), HASWELL)
  98. GETARCH_FLAGS := -DFORCE_NEHALEM
  99. endif
  100. ifeq ($(TARGET), SKYLAKEX)
  101. GETARCH_FLAGS := -DFORCE_NEHALEM
  102. endif
  103. ifeq ($(TARGET), COOPERLAKE)
  104. GETARCH_FLAGS := -DFORCE_NEHALEM
  105. endif
  106. ifeq ($(TARGET), SAPPHIRERAPIDS)
  107. GETARCH_FLAGS := -DFORCE_NEHALEM
  108. endif
  109. ifeq ($(TARGET), SANDYBRIDGE)
  110. GETARCH_FLAGS := -DFORCE_NEHALEM
  111. endif
  112. ifeq ($(TARGET), BULLDOZER)
  113. GETARCH_FLAGS := -DFORCE_BARCELONA
  114. endif
  115. ifeq ($(TARGET), PILEDRIVER)
  116. GETARCH_FLAGS := -DFORCE_BARCELONA
  117. endif
  118. ifeq ($(TARGET), STEAMROLLER)
  119. GETARCH_FLAGS := -DFORCE_BARCELONA
  120. endif
  121. ifeq ($(TARGET), EXCAVATOR)
  122. GETARCH_FLAGS := -DFORCE_BARCELONA
  123. endif
  124. ifeq ($(TARGET), ZEN)
  125. GETARCH_FLAGS := -DFORCE_BARCELONA
  126. endif
  127. ifeq ($(TARGET), ARMV8)
  128. GETARCH_FLAGS := -DFORCE_ARMV7
  129. endif
  130. ifeq ($(TARGET), POWER8)
  131. GETARCH_FLAGS := -DFORCE_POWER6
  132. endif
  133. ifeq ($(TARGET), POWER9)
  134. GETARCH_FLAGS := -DFORCE_POWER6
  135. endif
  136. ifeq ($(TARGET), POWER10)
  137. GETARCH_FLAGS := -DFORCE_POWER6
  138. endif
  139. endif
  140. #TARGET_CORE will override TARGET which is used in DYNAMIC_ARCH=1.
  141. #
  142. ifdef TARGET_CORE
  143. GETARCH_FLAGS := -DFORCE_$(TARGET_CORE)
  144. endif
  145. # Force fallbacks for 32bit
  146. ifeq ($(BINARY), 32)
  147. ifeq ($(TARGET_CORE), HASWELL)
  148. GETARCH_FLAGS := -DFORCE_NEHALEM
  149. endif
  150. ifeq ($(TARGET_CORE), SKYLAKEX)
  151. GETARCH_FLAGS := -DFORCE_NEHALEM
  152. endif
  153. ifeq ($(TARGET_CORE), COOPERLAKE)
  154. GETARCH_FLAGS := -DFORCE_NEHALEM
  155. endif
  156. ifeq ($(TARGET_CORE), SAPPHIRERAPIDS)
  157. GETARCH_FLAGS := -DFORCE_NEHALEM
  158. endif
  159. ifeq ($(TARGET_CORE), SANDYBRIDGE)
  160. GETARCH_FLAGS := -DFORCE_NEHALEM
  161. endif
  162. ifeq ($(TARGET_CORE), BULLDOZER)
  163. GETARCH_FLAGS := -DFORCE_BARCELONA
  164. endif
  165. ifeq ($(TARGET_CORE), PILEDRIVER)
  166. GETARCH_FLAGS := -DFORCE_BARCELONA
  167. endif
  168. ifeq ($(TARGET_CORE), STEAMROLLER)
  169. GETARCH_FLAGS := -DFORCE_BARCELONA
  170. endif
  171. ifeq ($(TARGET_CORE), EXCAVATOR)
  172. GETARCH_FLAGS := -DFORCE_BARCELONA
  173. endif
  174. ifeq ($(TARGET_CORE), ZEN)
  175. GETARCH_FLAGS := -DFORCE_BARCELONA
  176. endif
  177. endif
  178. # On x86_64 build getarch with march=native unless the compiler is PGI. This is required to detect AVX512 support in getarch.
  179. ifeq ($(HOSTARCH), x86_64)
  180. ifeq ($(findstring pgcc,$(HOSTCC))$(findstring nvc,$(HOSTCC)),)
  181. GETARCH_FLAGS += -march=native
  182. endif
  183. endif
  184. ifdef INTERFACE64
  185. ifneq ($(INTERFACE64), 0)
  186. GETARCH_FLAGS += -DUSE64BITINT
  187. endif
  188. endif
  189. ifndef GEMM_MULTITHREAD_THRESHOLD
  190. GEMM_MULTITHREAD_THRESHOLD=4
  191. endif
  192. GETARCH_FLAGS += -DGEMM_MULTITHREAD_THRESHOLD=$(GEMM_MULTITHREAD_THRESHOLD)
  193. ifeq ($(NO_AVX), 1)
  194. GETARCH_FLAGS += -DNO_AVX
  195. endif
  196. ifeq ($(BINARY), 32)
  197. GETARCH_FLAGS += -DNO_AVX -DNO_AVX2 -DNO_AVX512
  198. NO_AVX512 = 1
  199. endif
  200. ifeq ($(NO_AVX2), 1)
  201. GETARCH_FLAGS += -DNO_AVX2
  202. endif
  203. ifeq ($(NO_AVX512), 1)
  204. GETARCH_FLAGS += -DNO_AVX512
  205. endif
  206. ifeq ($(DEBUG), 1)
  207. GETARCH_FLAGS += -g
  208. endif
  209. ifeq ($(QUIET_MAKE), 1)
  210. MAKE += -s
  211. endif
  212. ifndef NO_PARALLEL_MAKE
  213. NO_PARALLEL_MAKE=0
  214. endif
  215. GETARCH_FLAGS += -DNO_PARALLEL_MAKE=$(NO_PARALLEL_MAKE)
  216. ifdef MAKE_NB_JOBS
  217. GETARCH_FLAGS += -DMAKE_NB_JOBS=$(MAKE_NB_JOBS)
  218. endif
  219. ifeq ($(HOSTCC), loongcc)
  220. GETARCH_FLAGS += -static
  221. endif
  222. #if don't use Fortran, it will only compile CBLAS.
  223. ifeq ($(ONLY_CBLAS), 1)
  224. NO_LAPACK = 1
  225. else
  226. ONLY_CBLAS = 0
  227. endif
  228. #For small matrix optimization
  229. ifeq ($(ARCH), x86_64)
  230. SMALL_MATRIX_OPT = 1
  231. else ifeq ($(CORE), POWER10)
  232. SMALL_MATRIX_OPT = 1
  233. endif
  234. ifeq ($(SMALL_MATRIX_OPT), 1)
  235. CCOMMON_OPT += -DSMALL_MATRIX_OPT
  236. endif
  237. # This operation is expensive, so execution should be once.
  238. ifndef GOTOBLAS_MAKEFILE
  239. export GOTOBLAS_MAKEFILE = 1
  240. # Determine if the assembler is GNU Assembler
  241. HAVE_GAS := $(shell $(AS) -v < /dev/null 2>&1 | grep GNU 2>&1 >/dev/null ; echo $$?)
  242. GETARCH_FLAGS += -DHAVE_GAS=$(HAVE_GAS)
  243. # Generating Makefile.conf and config.h
  244. 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)
  245. ifndef TARGET_CORE
  246. include $(TOPDIR)/Makefile.conf
  247. else
  248. HAVE_NEON=
  249. HAVE_VFP=
  250. HAVE_VFPV3=
  251. HAVE_VFPV4=
  252. HAVE_MMX=
  253. HAVE_SSE=
  254. HAVE_SSE2=
  255. HAVE_SSE3=
  256. HAVE_SSSE3=
  257. HAVE_SSE4_1=
  258. HAVE_SSE4_2=
  259. HAVE_SSE4A=
  260. HAVE_SSE5=
  261. HAVE_AVX=
  262. HAVE_AVX2=
  263. HAVE_FMA3=
  264. include $(TOPDIR)/Makefile_kernel.conf
  265. endif
  266. endif
  267. ifndef NUM_PARALLEL
  268. NUM_PARALLEL = 1
  269. endif
  270. ifndef NUM_THREADS
  271. NUM_THREADS = $(NUM_CORES)
  272. endif
  273. ifeq ($(NUM_THREADS), 1)
  274. override USE_THREAD = 0
  275. override USE_OPENMP = 0
  276. endif
  277. ifdef USE_THREAD
  278. ifeq ($(USE_THREAD), 0)
  279. SMP =
  280. else
  281. SMP = 1
  282. endif
  283. else
  284. ifeq ($(NUM_THREADS), 1)
  285. SMP =
  286. else
  287. SMP = 1
  288. endif
  289. endif
  290. ifeq ($(SMP), 1)
  291. USE_LOCKING =
  292. endif
  293. ifndef NEED_PIC
  294. NEED_PIC = 1
  295. endif
  296. ARFLAGS =
  297. CPP = $(COMPILER) -E
  298. AR ?= $(CROSS_SUFFIX)ar
  299. AS ?= $(CROSS_SUFFIX)as
  300. LD ?= $(CROSS_SUFFIX)ld
  301. RANLIB ?= $(CROSS_SUFFIX)ranlib
  302. NM = $(CROSS_SUFFIX)nm
  303. DLLWRAP = $(CROSS_SUFFIX)dllwrap
  304. OBJCOPY = $(CROSS_SUFFIX)objcopy
  305. OBJCONV = $(CROSS_SUFFIX)objconv
  306. # When fortran support was either not detected or actively deselected, only build BLAS.
  307. ifeq ($(NOFORTRAN), 1)
  308. NO_LAPACK = 1
  309. override FEXTRALIB =
  310. endif
  311. ifeq ($(C_COMPILER), GCC)
  312. GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
  313. GCCVERSIONGT4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 4)
  314. GCCVERSIONGT5 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 5)
  315. GCCVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 7)
  316. GCCVERSIONGTEQ8 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 8)
  317. GCCVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  318. GCCVERSIONGTEQ11 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 11)
  319. GCCVERSIONGTEQ10 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 10)
  320. # Note that the behavior of -dumpversion is compile-time-configurable for
  321. # gcc-7.x and newer. Use -dumpfullversion there
  322. ifeq ($(GCCVERSIONGTEQ7),1)
  323. GCCDUMPVERSION_PARAM := -dumpfullversion
  324. else
  325. GCCDUMPVERSION_PARAM := -dumpversion
  326. endif
  327. GCCMINORVERSIONGTEQ1 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 1)
  328. GCCMINORVERSIONGTEQ2 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 2)
  329. GCCMINORVERSIONGTEQ4 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 4)
  330. GCCMINORVERSIONGTEQ7 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 7)
  331. endif
  332. #
  333. # OS dependent settings
  334. #
  335. ifeq ($(OSNAME), Darwin)
  336. ifndef MACOSX_DEPLOYMENT_TARGET
  337. export MACOSX_DEPLOYMENT_TARGET=10.8
  338. endif
  339. MD5SUM = md5 -r
  340. endif
  341. ifneq (,$(findstring $(OSNAME), FreeBSD OpenBSD DragonFly))
  342. MD5SUM = md5 -r
  343. endif
  344. ifeq ($(OSNAME), NetBSD)
  345. MD5SUM = md5 -n
  346. endif
  347. ifeq ($(OSNAME), Linux)
  348. EXTRALIB += -lm
  349. NO_EXPRECISION = 1
  350. endif
  351. ifeq ($(OSNAME), Android)
  352. EXTRALIB += -lm
  353. endif
  354. ifeq ($(OSNAME), AIX)
  355. EXTRALIB += -lm
  356. endif
  357. ifeq ($(OSNAME), FreeBSD)
  358. ifeq ($(ARCH), $(filter $(ARCH),arm arm64))
  359. EXTRALIB += -lm
  360. endif
  361. endif
  362. ifeq ($(OSNAME), WINNT)
  363. NEED_PIC = 0
  364. NO_EXPRECISION = 1
  365. EXTRALIB += -defaultlib:advapi32
  366. SUFFIX = obj
  367. PSUFFIX = pobj
  368. LIBSUFFIX = a
  369. ifeq ($(C_COMPILER), CLANG)
  370. CCOMMON_OPT += -DMS_ABI
  371. endif
  372. #Version tests for supporting specific features (MS_ABI, POWER9 intrinsics)
  373. ifeq ($(GCCVERSIONGT4), 1)
  374. # GCC Major version > 4
  375. # It is compatible with MSVC ABI.
  376. CCOMMON_OPT += -DMS_ABI
  377. endif
  378. ifeq ($(GCCVERSIONGTEQ4), 1)
  379. ifeq ($(GCCMINORVERSIONGTEQ7), 1)
  380. # GCC Version >=4.7
  381. # It is compatible with MSVC ABI.
  382. CCOMMON_OPT += -DMS_ABI
  383. endif
  384. endif
  385. # Ensure the correct stack alignment on Win32
  386. # http://permalink.gmane.org/gmane.comp.lib.openblas.general/97
  387. ifeq ($(ARCH), x86)
  388. CCOMMON_OPT += -mincoming-stack-boundary=2
  389. FCOMMON_OPT += -mincoming-stack-boundary=2
  390. endif
  391. endif
  392. ifeq ($(OSNAME), Interix)
  393. NEED_PIC = 0
  394. NO_EXPRECISION = 1
  395. INTERIX_TOOL_DIR = /opt/gcc.3.3/i586-pc-interix3/bin
  396. endif
  397. ifeq ($(OSNAME), CYGWIN_NT)
  398. NEED_PIC = 0
  399. NO_EXPRECISION = 1
  400. OS_CYGWIN_NT = 1
  401. endif
  402. ifneq ($(OSNAME), WINNT)
  403. ifneq ($(OSNAME), CYGWIN_NT)
  404. ifneq ($(OSNAME), Interix)
  405. ifneq ($(OSNAME), Android)
  406. ifdef SMP
  407. EXTRALIB += -lpthread
  408. endif
  409. endif
  410. endif
  411. endif
  412. endif
  413. # ifeq logical or
  414. ifeq ($(OSNAME), $(filter $(OSNAME),WINNT CYGWIN_NT Interix))
  415. OS_WINDOWS=1
  416. endif
  417. ifdef QUAD_PRECISION
  418. CCOMMON_OPT += -DQUAD_PRECISION
  419. NO_EXPRECISION = 1
  420. endif
  421. ifneq ($(ARCH), x86)
  422. ifneq ($(ARCH), x86_64)
  423. NO_EXPRECISION = 1
  424. endif
  425. endif
  426. ifdef UTEST_CHECK
  427. CCOMMON_OPT += -DUTEST_CHECK
  428. SANITY_CHECK = 1
  429. endif
  430. ifdef SANITY_CHECK
  431. CCOMMON_OPT += -DSANITY_CHECK -DREFNAME=$(*F)f$(BU)
  432. endif
  433. MAX_STACK_ALLOC ?= 2048
  434. ifneq ($(MAX_STACK_ALLOC), 0)
  435. CCOMMON_OPT += -DMAX_STACK_ALLOC=$(MAX_STACK_ALLOC)
  436. endif
  437. ifdef USE_LOCKING
  438. ifneq ($(USE_LOCKING), 0)
  439. CCOMMON_OPT += -DUSE_LOCKING
  440. endif
  441. endif
  442. #
  443. # Architecture dependent settings
  444. #
  445. ifeq ($(ARCH), x86)
  446. ifndef BINARY
  447. NO_BINARY_MODE = 1
  448. endif
  449. ifeq ($(CORE), generic)
  450. NO_EXPRECISION = 1
  451. endif
  452. ifndef NO_EXPRECISION
  453. ifeq ($(F_COMPILER), GFORTRAN)
  454. # ifeq logical or. GCC or LSB
  455. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  456. EXPRECISION = 1
  457. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  458. FCOMMON_OPT += -m128bit-long-double
  459. endif
  460. ifeq ($(C_COMPILER), CLANG)
  461. EXPRECISION = 1
  462. CCOMMON_OPT += -DEXPRECISION
  463. FCOMMON_OPT += -m128bit-long-double
  464. endif
  465. endif
  466. endif
  467. endif
  468. ifeq ($(ARCH), x86_64)
  469. ifeq ($(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 ($(C_COMPILER), INTEL)
  489. CCOMMON_OPT += -wd981
  490. endif
  491. ifeq ($(USE_OPENMP), 1)
  492. #check
  493. ifeq ($(USE_THREAD), 0)
  494. $(error OpenBLAS: Cannot set both USE_OPENMP=1 and USE_THREAD=0. The USE_THREAD=0 is only for building single thread version.)
  495. endif
  496. # ifeq logical or. GCC or LSB
  497. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  498. CCOMMON_OPT += -fopenmp
  499. endif
  500. ifeq ($(C_COMPILER), CLANG)
  501. CCOMMON_OPT += -fopenmp
  502. endif
  503. ifeq ($(C_COMPILER), INTEL)
  504. CCOMMON_OPT += -fopenmp
  505. endif
  506. ifeq ($(C_COMPILER), PGI)
  507. CCOMMON_OPT += -mp
  508. endif
  509. ifeq ($(C_COMPILER), OPEN64)
  510. CCOMMON_OPT += -mp
  511. CEXTRALIB += -lstdc++
  512. endif
  513. ifeq ($(C_COMPILER), PATHSCALE)
  514. CCOMMON_OPT += -mp
  515. endif
  516. endif
  517. ifeq ($(DYNAMIC_ARCH), 1)
  518. ifeq ($(ARCH), x86)
  519. DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
  520. CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
  521. endif
  522. ifeq ($(ARCH), x86_64)
  523. DYNAMIC_CORE = PRESCOTT CORE2
  524. ifeq ($(DYNAMIC_OLDER), 1)
  525. DYNAMIC_CORE += PENRYN DUNNINGTON
  526. endif
  527. DYNAMIC_CORE += NEHALEM
  528. ifeq ($(DYNAMIC_OLDER), 1)
  529. DYNAMIC_CORE += OPTERON OPTERON_SSE3
  530. endif
  531. DYNAMIC_CORE += BARCELONA
  532. ifeq ($(DYNAMIC_OLDER), 1)
  533. DYNAMIC_CORE += BOBCAT ATOM NANO
  534. endif
  535. ifneq ($(NO_AVX), 1)
  536. DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER STEAMROLLER EXCAVATOR
  537. endif
  538. ifneq ($(NO_AVX2), 1)
  539. DYNAMIC_CORE += HASWELL ZEN
  540. endif
  541. ifneq ($(NO_AVX512), 1)
  542. ifneq ($(NO_AVX2), 1)
  543. DYNAMIC_CORE += SKYLAKEX COOPERLAKE
  544. endif
  545. endif
  546. endif
  547. ifdef DYNAMIC_LIST
  548. override DYNAMIC_CORE = PRESCOTT $(DYNAMIC_LIST)
  549. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_PRESCOTT
  550. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  551. CCOMMON_OPT += $(XCCOMMON_OPT)
  552. #CCOMMON_OPT += -DDYNAMIC_LIST='$(DYNAMIC_LIST)'
  553. endif
  554. ifeq ($(ARCH), arm64)
  555. DYNAMIC_CORE = ARMV8
  556. DYNAMIC_CORE += CORTEXA53
  557. DYNAMIC_CORE += CORTEXA57
  558. DYNAMIC_CORE += CORTEXA72
  559. DYNAMIC_CORE += CORTEXA73
  560. DYNAMIC_CORE += NEOVERSEN1
  561. DYNAMIC_CORE += NEOVERSEV1
  562. DYNAMIC_CORE += NEOVERSEN2
  563. DYNAMIC_CORE += CORTEXA55
  564. DYNAMIC_CORE += FALKOR
  565. DYNAMIC_CORE += THUNDERX
  566. DYNAMIC_CORE += THUNDERX2T99
  567. DYNAMIC_CORE += TSV110
  568. DYNAMIC_CORE += EMAG8180
  569. DYNAMIC_CORE += THUNDERX3T110
  570. ifdef DYNAMIC_LIST
  571. override DYNAMIC_CORE = ARMV8 $(DYNAMIC_LIST)
  572. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_ARMV8
  573. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  574. endif
  575. endif
  576. ifeq ($(ARCH), mips64)
  577. DYNAMIC_CORE = LOONGSON3R3 LOONGSON3R4
  578. endif
  579. ifeq ($(ARCH), zarch)
  580. DYNAMIC_CORE = ZARCH_GENERIC
  581. # if the compiler accepts -march=arch11 or -march=z13 and can compile a file
  582. # with z13-specific inline assembly, then we can include support for Z13.
  583. # note: -march=z13 is equivalent to -march=arch11 yet some compiler releases
  584. # only support one or the other.
  585. # note: LLVM version 6.x supported -march=z13 yet could not handle vector
  586. # registers in inline assembly, so the check for supporting the -march flag is
  587. # not enough.
  588. ZARCH_TEST_COMPILE=-c $(TOPDIR)/kernel/zarch/damin_z13.c -I$(TOPDIR) -o /dev/null > /dev/null 2> /dev/null
  589. ZARCH_CC_SUPPORTS_ARCH11=$(shell $(CC) -march=arch11 $(ZARCH_TEST_COMPILE) && echo 1)
  590. ZARCH_CC_SUPPORTS_Z13=$(shell $(CC) -march=z13 $(ZARCH_TEST_COMPILE) && echo 1)
  591. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH11), $(ZARCH_CC_SUPPORTS_Z13)), 1)
  592. DYNAMIC_CORE += Z13
  593. CCOMMON_OPT += -DDYN_Z13
  594. else
  595. $(info OpenBLAS: Not building Z13 kernels because the compiler $(CC) does not support it)
  596. endif
  597. # as above for z13, check for -march=arch12 and z14 support in the compiler.
  598. ZARCH_CC_SUPPORTS_ARCH12=$(shell $(CC) -march=arch12 $(ZARCH_TEST_COMPILE) && echo 1)
  599. ZARCH_CC_SUPPORTS_Z14=$(shell $(CC) -march=z14 $(ZARCH_TEST_COMPILE) && echo 1)
  600. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH12), $(ZARCH_CC_SUPPORTS_Z14)), 1)
  601. DYNAMIC_CORE += Z14
  602. CCOMMON_OPT += -DDYN_Z14
  603. else
  604. $(info OpenBLAS: Not building Z14 kernels because the compiler $(CC) does not support it)
  605. endif
  606. endif # ARCH zarch
  607. ifeq ($(ARCH), power)
  608. ifneq ($(C_COMPILER), PGI)
  609. DYNAMIC_CORE = POWER6
  610. DYNAMIC_CORE += POWER8
  611. ifneq ($(C_COMPILER), GCC)
  612. DYNAMIC_CORE += POWER9
  613. DYNAMIC_CORE += POWER10
  614. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  615. endif
  616. ifeq ($(C_COMPILER), GCC)
  617. ifeq ($(GCCVERSIONGT5), 1)
  618. DYNAMIC_CORE += POWER9
  619. else
  620. $(info, OpenBLAS: Your gcc version is too old to build the POWER9 kernels.)
  621. endif
  622. LDVERSIONGTEQ35 := $(shell expr `$(CC) -Wl,--version 2> /dev/null | head -1 | cut -f2 -d "." | cut -f1 -d "-"` \>= 35)
  623. ifeq ($(GCCVERSIONGTEQ11)$(LDVERSIONGTEQ35), 11)
  624. DYNAMIC_CORE += POWER10
  625. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  626. else ifeq ($(GCCVERSIONGTEQ10), 1)
  627. ifeq ($(GCCMINORVERSIONGTEQ2)$(LDVERSIONGTEQ35), 11)
  628. DYNAMIC_CORE += POWER10
  629. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  630. endif
  631. else
  632. $(info, OpenBLAS: Your gcc version is too old to build the POWER10 kernels.)
  633. endif
  634. endif
  635. else
  636. DYNAMIC_CORE = POWER8
  637. DYNAMIC_CORE += POWER9
  638. endif
  639. endif
  640. # If DYNAMIC_CORE is not set, DYNAMIC_ARCH cannot do anything, so force it to empty
  641. ifndef DYNAMIC_CORE
  642. override DYNAMIC_ARCH=
  643. endif
  644. endif
  645. ifeq ($(ARCH), ia64)
  646. NO_BINARY_MODE = 1
  647. BINARY_DEFINED = 1
  648. ifeq ($(F_COMPILER), GFORTRAN)
  649. ifeq ($(C_COMPILER), GCC)
  650. # EXPRECISION = 1
  651. # CCOMMON_OPT += -DEXPRECISION
  652. endif
  653. endif
  654. endif
  655. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  656. NO_BINARY_MODE = 1
  657. endif
  658. ifeq ($(ARCH), alpha)
  659. NO_BINARY_MODE = 1
  660. BINARY_DEFINED = 1
  661. endif
  662. ifeq ($(ARCH), arm)
  663. NO_BINARY_MODE = 1
  664. BINARY_DEFINED = 1
  665. CCOMMON_OPT += -marm
  666. FCOMMON_OPT += -marm
  667. # If softfp abi is mentioned on the command line, force it.
  668. ifeq ($(ARM_SOFTFP_ABI), 1)
  669. CCOMMON_OPT += -mfloat-abi=softfp
  670. FCOMMON_OPT += -mfloat-abi=softfp
  671. endif
  672. ifeq ($(OSNAME), Android)
  673. ifeq ($(ARM_SOFTFP_ABI), 1)
  674. EXTRALIB += -lm
  675. else
  676. EXTRALIB += -Wl,-lm_hard
  677. endif
  678. endif
  679. endif
  680. ifeq ($(ARCH), arm64)
  681. NO_BINARY_MODE = 1
  682. BINARY_DEFINED = 1
  683. ifdef INTERFACE64
  684. ifneq ($(INTERFACE64), 0)
  685. ifeq ($(F_COMPILER), GFORTRAN)
  686. FCOMMON_OPT += -fdefault-integer-8
  687. endif
  688. ifeq ($(F_COMPILER), FLANG)
  689. FCOMMON_OPT += -i8
  690. endif
  691. endif
  692. endif
  693. endif
  694. ifeq ($(ARCH), riscv64)
  695. NO_BINARY_MODE = 1
  696. BINARY_DEFINED = 1
  697. endif
  698. ifeq ($(ARCH), loongarch64)
  699. NO_BINARY_MODE = 1
  700. BINARY_DEFINED = 1
  701. endif
  702. #
  703. # C Compiler dependent settings
  704. #
  705. # ifeq logical or. GCC or CLANG or LSB
  706. # http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
  707. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG LSB))
  708. CCOMMON_OPT += -Wall
  709. COMMON_PROF += -fno-inline
  710. NO_UNINITIALIZED_WARN = -Wno-uninitialized
  711. ifeq ($(QUIET_MAKE), 1)
  712. CCOMMON_OPT += $(NO_UNINITIALIZED_WARN) -Wno-unused
  713. endif
  714. ifdef NO_BINARY_MODE
  715. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  716. ifdef BINARY64
  717. CCOMMON_OPT += -mabi=64
  718. else
  719. CCOMMON_OPT += -mabi=n32
  720. endif
  721. BINARY_DEFINED = 1
  722. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  723. CCOMMON_OPT += -mabi=32
  724. BINARY_DEFINED = 1
  725. endif
  726. ifeq ($(CORE), $(filter $(CORE),LOONGSON3R3 LOONGSON3R4))
  727. CCOMMON_OPT += -march=loongson3a
  728. FCOMMON_OPT += -march=loongson3a
  729. endif
  730. ifeq ($(CORE), MIPS24K)
  731. CCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  732. FCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  733. endif
  734. ifeq ($(CORE), MIPS1004K)
  735. CCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  736. FCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  737. endif
  738. ifeq ($(CORE), P5600)
  739. CCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  740. FCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  741. endif
  742. ifeq ($(CORE), I6400)
  743. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  744. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  745. endif
  746. ifeq ($(CORE), P6600)
  747. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  748. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  749. endif
  750. ifeq ($(CORE), I6500)
  751. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  752. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  753. endif
  754. ifeq ($(OSNAME), AIX)
  755. BINARY_DEFINED = 1
  756. endif
  757. ifeq ($(ARCH), loongarch64)
  758. ifeq ($(CORE), LOONGSON3R5)
  759. CCOMMON_OPT += -march=loongarch64 -mabi=lp64
  760. FCOMMON_OPT += -march=loongarch64 -mabi=lp64
  761. endif
  762. endif
  763. endif
  764. ifndef BINARY_DEFINED
  765. ifneq ($(OSNAME), AIX)
  766. ifdef BINARY64
  767. ifneq ($(ARCH), riscv64)
  768. CCOMMON_OPT += -m64
  769. endif
  770. else
  771. CCOMMON_OPT += -m32
  772. endif
  773. endif
  774. endif
  775. endif
  776. ifeq ($(C_COMPILER), PGI)
  777. PGCVERSIONGT20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \> 20)
  778. PGCVERSIONEQ20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` == 20)
  779. PGCMINORVERSIONGE11 := $(shell expr `$(CC) --version|sed -n "2p" |cut -d "-" -f 1 |sed -e "s/[^0-9.]//g" |cut -c 4-5` \>= 11)
  780. PGCVERSIONCHECK := $(PGCVERSIONGT20)$(PGCVERSIONEQ20)$(PGCMINORVERSIONGE11)
  781. ifeq ($(PGCVERSIONCHECK), $(filter $(PGCVERSIONCHECK), 100 101 011))
  782. NEWPGI := 1
  783. PGCVERSIONGT21 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \> 21)
  784. PGCVERSIONEQ21 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` == 21)
  785. PGCVERSIONCHECK2 := $(PGCVERSIONGT21)$(PGCVERSIONEQ21)$(PGCMINORVERSIONGE11)
  786. ifeq ($(PGCVERSIONCHECK2), $(filter $(PGCVERSIONCHECK2), 100 101 011))
  787. NEWPGI2 := 1
  788. endif
  789. endif
  790. ifdef BINARY64
  791. ifeq ($(ARCH), x86_64)
  792. ifneq ($(NEWPGI2),1)
  793. CCOMMON_OPT += -tp p7-64
  794. else
  795. CCOMMON_OPT += -tp px
  796. endif
  797. ifneq ($(NEWPGI),1)
  798. CCOMMON_OPT += -D__MMX__ -Mnollvm
  799. endif
  800. else
  801. ifeq ($(ARCH), power)
  802. ifeq ($(CORE), POWER8)
  803. CCOMMON_OPT += -tp pwr8
  804. endif
  805. ifeq ($(CORE), POWER9)
  806. CCOMMON_OPT += -tp pwr9
  807. endif
  808. endif
  809. endif
  810. else
  811. ifneq ($(NEWPGI2),1)
  812. CCOMMON_OPT += -tp p7
  813. else
  814. CCOMMON_OPT += -tp px
  815. endif
  816. endif
  817. endif
  818. ifeq ($(C_COMPILER), PATHSCALE)
  819. ifdef BINARY64
  820. CCOMMON_OPT += -m64
  821. else
  822. CCOMMON_OPT += -m32
  823. endif
  824. endif
  825. #
  826. # Fortran Compiler dependent settings
  827. #
  828. ifeq ($(F_COMPILER), NAG)
  829. FCOMMON_OPT += -dcfuns -recursive -ieee=full -w=obs -thread_safe
  830. ifdef INTERFACE64
  831. ifneq ($(INTERFACE64), 0)
  832. FCOMMON_OPT += -i8
  833. endif
  834. endif
  835. ifeq ($(USE_OPENMP), 1)
  836. FCOMMON_OPT += -openmp
  837. endif
  838. endif
  839. ifeq ($(F_COMPILER), FLANG)
  840. CCOMMON_OPT += -DF_INTERFACE_FLANG
  841. FCOMMON_OPT += -Mrecursive -Kieee
  842. ifeq ($(OSNAME), Linux)
  843. ifeq ($(ARCH), x86_64)
  844. FLANG_VENDOR := $(shell $(FC) --version|head -1 |cut -f 1 -d " ")
  845. ifeq ($(FLANG_VENDOR), AMD)
  846. FCOMMON_OPT += -fno-unroll-loops
  847. endif
  848. endif
  849. endif
  850. ifdef BINARY64
  851. ifdef INTERFACE64
  852. ifneq ($(INTERFACE64), 0)
  853. FCOMMON_OPT += -i8
  854. endif
  855. endif
  856. FCOMMON_OPT += -Wall
  857. else
  858. FCOMMON_OPT += -Wall
  859. endif
  860. ifeq ($(USE_OPENMP), 1)
  861. FCOMMON_OPT += -fopenmp
  862. endif
  863. endif
  864. ifeq ($(F_COMPILER), G77)
  865. CCOMMON_OPT += -DF_INTERFACE_G77
  866. FCOMMON_OPT += -Wall
  867. ifndef NO_BINARY_MODE
  868. ifneq ($(OSNAME), AIX)
  869. ifdef BINARY64
  870. FCOMMON_OPT += -m64
  871. else
  872. FCOMMON_OPT += -m32
  873. endif
  874. endif
  875. endif
  876. endif
  877. ifeq ($(F_COMPILER), G95)
  878. CCOMMON_OPT += -DF_INTERFACE_G95
  879. FCOMMON_OPT += -Wall
  880. ifneq ($(OSNAME), AIX)
  881. ifndef NO_BINARY_MODE
  882. ifdef BINARY64
  883. FCOMMON_OPT += -m64
  884. else
  885. FCOMMON_OPT += -m32
  886. endif
  887. endif
  888. ifneq ($(NO_LAPACKE), 1)
  889. FCOMMON_OPT += -fno-second-underscore
  890. endif
  891. endif
  892. endif
  893. ifeq ($(F_COMPILER), GFORTRAN)
  894. CCOMMON_OPT += -DF_INTERFACE_GFORT
  895. FCOMMON_OPT += -Wall
  896. # make single-threaded LAPACK calls thread-safe #1847
  897. FCOMMON_OPT += -frecursive
  898. # work around ABI problem with passing single-character arguments
  899. FCOMMON_OPT += -fno-optimize-sibling-calls
  900. #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
  901. ifneq ($(NO_LAPACK), 1)
  902. EXTRALIB += -lgfortran
  903. endif
  904. ifdef NO_BINARY_MODE
  905. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  906. ifdef BINARY64
  907. FCOMMON_OPT += -mabi=64
  908. else
  909. FCOMMON_OPT += -mabi=n32
  910. endif
  911. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  912. FCOMMON_OPT += -mabi=32
  913. endif
  914. else
  915. ifdef BINARY64
  916. ifneq ($(OSNAME), AIX)
  917. ifneq ($(ARCH), riscv64)
  918. FCOMMON_OPT += -m64
  919. endif
  920. endif
  921. ifdef INTERFACE64
  922. ifneq ($(INTERFACE64), 0)
  923. FCOMMON_OPT += -fdefault-integer-8
  924. endif
  925. endif
  926. else
  927. ifneq ($(OSNAME), AIX)
  928. FCOMMON_OPT += -m32
  929. endif
  930. endif
  931. endif
  932. ifeq ($(USE_OPENMP), 1)
  933. FCOMMON_OPT += -fopenmp
  934. endif
  935. endif
  936. ifeq ($(F_COMPILER), INTEL)
  937. CCOMMON_OPT += -DF_INTERFACE_INTEL
  938. ifdef INTERFACE64
  939. ifneq ($(INTERFACE64), 0)
  940. FCOMMON_OPT += -i8
  941. endif
  942. endif
  943. FCOMMON_OPT += -recursive -fp-model strict -assume protect-parens
  944. ifeq ($(USE_OPENMP), 1)
  945. FCOMMON_OPT += -fopenmp
  946. endif
  947. endif
  948. ifeq ($(F_COMPILER), FUJITSU)
  949. CCOMMON_OPT += -DF_INTERFACE_FUJITSU
  950. ifeq ($(USE_OPENMP), 1)
  951. FCOMMON_OPT += -openmp
  952. endif
  953. endif
  954. ifeq ($(F_COMPILER), IBM)
  955. CCOMMON_OPT += -DF_INTERFACE_IBM
  956. # FCOMMON_OPT += -qarch=440
  957. ifdef BINARY64
  958. FCOMMON_OPT += -q64
  959. ifdef INTERFACE64
  960. ifneq ($(INTERFACE64), 0)
  961. FCOMMON_OPT += -qintsize=8
  962. endif
  963. endif
  964. else
  965. FCOMMON_OPT += -q32
  966. endif
  967. ifeq ($(USE_OPENMP), 1)
  968. FCOMMON_OPT += -openmp
  969. endif
  970. endif
  971. ifeq ($(F_COMPILER), PGI)
  972. CCOMMON_OPT += -DF_INTERFACE_PGI
  973. COMMON_PROF += -DPGICOMPILER
  974. ifdef BINARY64
  975. ifdef INTERFACE64
  976. ifneq ($(INTERFACE64), 0)
  977. FCOMMON_OPT += -i8
  978. endif
  979. endif
  980. ifeq ($(ARCH), x86_64)
  981. ifneq ($(NEWPGI2),1)
  982. FCOMMON_OPT += -tp p7-64
  983. else
  984. FCOMMON_OPT += -tp px
  985. endif
  986. else
  987. ifeq ($(ARCH), power)
  988. ifeq ($(CORE), POWER6)
  989. $(warning NVIDIA HPC compilers do not support POWER6.)
  990. endif
  991. ifeq ($(CORE), POWER8)
  992. FCOMMON_OPT += -tp pwr8
  993. endif
  994. ifeq ($(CORE), POWER9)
  995. FCOMMON_OPT += -tp pwr9
  996. endif
  997. ifeq ($(CORE), POWER10)
  998. $(warning NVIDIA HPC compilers do not support POWER10.)
  999. endif
  1000. endif
  1001. endif
  1002. else
  1003. FCOMMON_OPT += -tp p7
  1004. endif
  1005. FCOMMON_OPT += -Mrecursive -Kieee
  1006. ifeq ($(USE_OPENMP), 1)
  1007. FCOMMON_OPT += -mp
  1008. endif
  1009. endif
  1010. ifeq ($(F_COMPILER), PATHSCALE)
  1011. CCOMMON_OPT += -DF_INTERFACE_PATHSCALE
  1012. ifdef BINARY64
  1013. ifdef INTERFACE64
  1014. ifneq ($(INTERFACE64), 0)
  1015. FCOMMON_OPT += -i8
  1016. endif
  1017. endif
  1018. endif
  1019. ifeq ($(USE_OPENMP), 1)
  1020. FCOMMON_OPT += -mp
  1021. endif
  1022. endif
  1023. ifeq ($(F_COMPILER), OPEN64)
  1024. CCOMMON_OPT += -DF_INTERFACE_OPEN64
  1025. ifdef BINARY64
  1026. ifdef INTERFACE64
  1027. ifneq ($(INTERFACE64), 0)
  1028. FCOMMON_OPT += -i8
  1029. endif
  1030. endif
  1031. endif
  1032. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  1033. ifndef BINARY64
  1034. FCOMMON_OPT += -n32
  1035. else
  1036. FCOMMON_OPT += -n64
  1037. endif
  1038. ifeq ($(CORE), LOONGSON3R3)
  1039. FCOMMON_OPT += -loongson3 -static
  1040. endif
  1041. ifeq ($(CORE), LOONGSON3R4)
  1042. FCOMMON_OPT += -loongson3 -static
  1043. endif
  1044. else
  1045. ifndef BINARY64
  1046. FCOMMON_OPT += -m32
  1047. else
  1048. FCOMMON_OPT += -m64
  1049. endif
  1050. endif
  1051. ifeq ($(USE_OPENMP), 1)
  1052. FEXTRALIB += -lstdc++
  1053. FCOMMON_OPT += -mp
  1054. endif
  1055. endif
  1056. ifeq ($(C_COMPILER), OPEN64)
  1057. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  1058. ifndef BINARY64
  1059. CCOMMON_OPT += -n32
  1060. else
  1061. CCOMMON_OPT += -n64
  1062. endif
  1063. ifeq ($(CORE), LOONGSON3R3)
  1064. CCOMMON_OPT += -loongson3 -static
  1065. endif
  1066. ifeq ($(CORE), LOONGSON3R4)
  1067. CCOMMON_OPT += -loongson3 -static
  1068. endif
  1069. else
  1070. ifndef BINARY64
  1071. CCOMMON_OPT += -m32
  1072. else
  1073. CCOMMON_OPT += -m64
  1074. endif
  1075. endif
  1076. endif
  1077. ifeq ($(C_COMPILER), SUN)
  1078. CCOMMON_OPT += -w
  1079. ifeq ($(ARCH), x86)
  1080. CCOMMON_OPT += -m32
  1081. else
  1082. ifdef BINARY64
  1083. CCOMMON_OPT += -m64
  1084. else
  1085. CCOMMON_OPT += -m32
  1086. endif
  1087. endif
  1088. endif
  1089. ifeq ($(F_COMPILER), SUN)
  1090. CCOMMON_OPT += -DF_INTERFACE_SUN
  1091. FCOMMON_OPT += -ftrap=%none -xrecursive
  1092. ifeq ($(ARCH), x86)
  1093. FCOMMON_OPT += -m32
  1094. else
  1095. ifdef BINARY64
  1096. FCOMMON_OPT += -m64
  1097. else
  1098. FCOMMON_OPT += -m32
  1099. endif
  1100. endif
  1101. ifeq ($(USE_OPENMP), 1)
  1102. FCOMMON_OPT += -xopenmp=parallel
  1103. endif
  1104. endif
  1105. ifeq ($(F_COMPILER), COMPAQ)
  1106. CCOMMON_OPT += -DF_INTERFACE_COMPAQ
  1107. ifeq ($(USE_OPENMP), 1)
  1108. FCOMMON_OPT += -openmp
  1109. endif
  1110. endif
  1111. ifdef BINARY64
  1112. ifdef INTERFACE64
  1113. ifneq ($(INTERFACE64), 0)
  1114. CCOMMON_OPT +=
  1115. #-DUSE64BITINT
  1116. endif
  1117. endif
  1118. endif
  1119. ifeq ($(NEED_PIC), 1)
  1120. ifeq ($(C_COMPILER), IBM)
  1121. CCOMMON_OPT += -qpic=large
  1122. else
  1123. CCOMMON_OPT += -fPIC
  1124. endif
  1125. ifeq ($(F_COMPILER), SUN)
  1126. FCOMMON_OPT += -pic
  1127. else ifeq ($(F_COMPILER), NAG)
  1128. FCOMMON_OPT += -PIC
  1129. else
  1130. FCOMMON_OPT += -fPIC
  1131. endif
  1132. endif
  1133. ifeq ($(DYNAMIC_ARCH), 1)
  1134. CCOMMON_OPT += -DDYNAMIC_ARCH
  1135. endif
  1136. ifeq ($(DYNAMIC_OLDER), 1)
  1137. CCOMMON_OPT += -DDYNAMIC_OLDER
  1138. endif
  1139. ifeq ($(NO_LAPACK), 1)
  1140. CCOMMON_OPT += -DNO_LAPACK
  1141. #Disable LAPACK C interface
  1142. NO_LAPACKE = 1
  1143. endif
  1144. ifeq ($(NO_LAPACKE), 1)
  1145. CCOMMON_OPT += -DNO_LAPACKE
  1146. endif
  1147. ifeq ($(NO_AVX), 1)
  1148. CCOMMON_OPT += -DNO_AVX
  1149. endif
  1150. ifeq ($(ARCH), x86)
  1151. CCOMMON_OPT += -DNO_AVX
  1152. endif
  1153. ifeq ($(NO_AVX2), 1)
  1154. CCOMMON_OPT += -DNO_AVX2
  1155. endif
  1156. ifeq ($(NO_AVX512), 1)
  1157. CCOMMON_OPT += -DNO_AVX512
  1158. endif
  1159. ifdef SMP
  1160. CCOMMON_OPT += -DSMP_SERVER
  1161. ifeq ($(ARCH), mips64)
  1162. USE_SIMPLE_THREADED_LEVEL3 = 1
  1163. endif
  1164. ifeq ($(USE_OPENMP), 1)
  1165. # USE_SIMPLE_THREADED_LEVEL3 = 1
  1166. # NO_AFFINITY = 1
  1167. CCOMMON_OPT += -DUSE_OPENMP
  1168. endif
  1169. ifeq ($(BIGNUMA), 1)
  1170. CCOMMON_OPT += -DBIGNUMA
  1171. endif
  1172. endif
  1173. ifeq ($(NO_WARMUP), 1)
  1174. CCOMMON_OPT += -DNO_WARMUP
  1175. endif
  1176. ifeq ($(CONSISTENT_FPCSR), 1)
  1177. CCOMMON_OPT += -DCONSISTENT_FPCSR
  1178. endif
  1179. # Only for development
  1180. # CCOMMON_OPT += -DPARAMTEST
  1181. # CCOMMON_OPT += -DPREFETCHTEST
  1182. # CCOMMON_OPT += -DNO_SWITCHING
  1183. # USE_PAPI = 1
  1184. ifdef USE_PAPI
  1185. CCOMMON_OPT += -DUSE_PAPI
  1186. EXTRALIB += -lpapi -lperfctr
  1187. endif
  1188. ifdef BUFFERSIZE
  1189. CCOMMON_OPT += -DBUFFERSIZE=$(BUFFERSIZE)
  1190. endif
  1191. ifdef DYNAMIC_THREADS
  1192. CCOMMON_OPT += -DDYNAMIC_THREADS
  1193. endif
  1194. CCOMMON_OPT += -DMAX_CPU_NUMBER=$(NUM_THREADS)
  1195. CCOMMON_OPT += -DMAX_PARALLEL_NUMBER=$(NUM_PARALLEL)
  1196. ifdef USE_SIMPLE_THREADED_LEVEL3
  1197. CCOMMON_OPT += -DUSE_SIMPLE_THREADED_LEVEL3
  1198. endif
  1199. ifeq ($(USE_TLS), 1)
  1200. CCOMMON_OPT += -DUSE_TLS
  1201. endif
  1202. ifeq ($(BUILD_BFLOAT16), 1)
  1203. CCOMMON_OPT += -DBUILD_BFLOAT16
  1204. endif
  1205. ifeq ($(BUILD_SINGLE), 1)
  1206. CCOMMON_OPT += -DBUILD_SINGLE=1
  1207. endif
  1208. ifeq ($(BUILD_DOUBLE), 1)
  1209. CCOMMON_OPT += -DBUILD_DOUBLE=1
  1210. endif
  1211. ifeq ($(BUILD_COMPLEX), 1)
  1212. CCOMMON_OPT += -DBUILD_COMPLEX=1
  1213. endif
  1214. ifeq ($(BUILD_COMPLEX16), 1)
  1215. CCOMMON_OPT += -DBUILD_COMPLEX16=1
  1216. endif
  1217. CCOMMON_OPT += -DVERSION=\"$(VERSION)\"
  1218. ifndef SYMBOLPREFIX
  1219. SYMBOLPREFIX =
  1220. endif
  1221. ifndef SYMBOLSUFFIX
  1222. SYMBOLSUFFIX =
  1223. endif
  1224. ifndef LIBSONAMEBASE
  1225. LIBSONAMEBASE = openblas
  1226. endif
  1227. ifndef LIBNAMESUFFIX
  1228. LIBNAMEBASE = $(SYMBOLPREFIX)$(LIBSONAMEBASE)$(SYMBOLSUFFIX)
  1229. else
  1230. LIBNAMEBASE = $(SYMBOLPREFIX)$(LIBSONAMEBASE)$(SYMBOLSUFFIX)_$(LIBNAMESUFFIX)
  1231. endif
  1232. ifeq ($(OSNAME), CYGWIN_NT)
  1233. LIBPREFIX = cyg$(LIBNAMEBASE)
  1234. else
  1235. LIBPREFIX = lib$(LIBNAMEBASE)
  1236. endif
  1237. KERNELDIR = $(TOPDIR)/kernel/$(ARCH)
  1238. include $(TOPDIR)/Makefile.$(ARCH)
  1239. ifneq ($(C_COMPILER), PGI)
  1240. ifneq ($(C_COMPILER), SUN)
  1241. CCOMMON_OPT += -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME
  1242. endif
  1243. endif
  1244. CCOMMON_OPT += -DASMNAME=$(FU)$(*F) -DASMFNAME=$(FU)$(*F)$(BU) -DNAME=$(*F)$(BU) -DCNAME=$(*F) -DCHAR_NAME=\"$(*F)$(BU)\" -DCHAR_CNAME=\"$(*F)\"
  1245. ifeq ($(CORE), PPC440)
  1246. CCOMMON_OPT += -DALLOC_QALLOC
  1247. endif
  1248. ifeq ($(CORE), PPC440FP2)
  1249. STATIC_ALLOCATION = 1
  1250. endif
  1251. ifneq ($(OSNAME), Linux)
  1252. NO_AFFINITY = 1
  1253. endif
  1254. ifneq ($(ARCH), x86_64)
  1255. ifneq ($(ARCH), x86)
  1256. NO_AFFINITY = 1
  1257. endif
  1258. endif
  1259. ifdef NO_AFFINITY
  1260. ifeq ($(NO_AFFINITY), 0)
  1261. override undefine NO_AFFINITY
  1262. else
  1263. CCOMMON_OPT += -DNO_AFFINITY
  1264. endif
  1265. endif
  1266. ifdef FUNCTION_PROFILE
  1267. CCOMMON_OPT += -DFUNCTION_PROFILE
  1268. endif
  1269. ifdef HUGETLB_ALLOCATION
  1270. CCOMMON_OPT += -DALLOC_HUGETLB
  1271. endif
  1272. ifdef HUGETLBFILE_ALLOCATION
  1273. CCOMMON_OPT += -DALLOC_HUGETLBFILE -DHUGETLB_FILE_NAME=$(HUGETLBFILE_ALLOCATION)
  1274. endif
  1275. ifdef STATIC_ALLOCATION
  1276. CCOMMON_OPT += -DALLOC_STATIC
  1277. endif
  1278. ifdef DEVICEDRIVER_ALLOCATION
  1279. CCOMMON_OPT += -DALLOC_DEVICEDRIVER -DDEVICEDRIVER_NAME=\"/dev/mapper\"
  1280. endif
  1281. ifdef MIXED_MEMORY_ALLOCATION
  1282. CCOMMON_OPT += -DMIXED_MEMORY_ALLOCATION
  1283. endif
  1284. ifeq ($(OSNAME), SunOS)
  1285. TAR = gtar
  1286. PATCH = gpatch
  1287. GREP = ggrep
  1288. AWK = nawk
  1289. else
  1290. TAR = tar
  1291. PATCH = patch
  1292. GREP = grep
  1293. AWK = awk
  1294. endif
  1295. ifndef MD5SUM
  1296. MD5SUM = md5sum
  1297. endif
  1298. REVISION = -r$(VERSION)
  1299. MAJOR_VERSION = $(word 1,$(subst ., ,$(VERSION)))
  1300. ifeq ($(DEBUG), 1)
  1301. COMMON_OPT += -g
  1302. endif
  1303. ifeq ($(DEBUG), 1)
  1304. FCOMMON_OPT += -g
  1305. endif
  1306. ifndef COMMON_OPT
  1307. COMMON_OPT = -O2
  1308. endif
  1309. ifndef FCOMMON_OPT
  1310. FCOMMON_OPT = -O2 -frecursive
  1311. endif
  1312. override CFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
  1313. override PFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
  1314. override FFLAGS += $(COMMON_OPT) $(FCOMMON_OPT)
  1315. override FPFLAGS += $(FCOMMON_OPT) $(COMMON_PROF)
  1316. #MAKEOVERRIDES =
  1317. ifdef NEED_PIC
  1318. ifeq (,$(findstring PIC,$(FFLAGS)))
  1319. override FFLAGS += -fPIC
  1320. endif
  1321. endif
  1322. #For LAPACK Fortran codes.
  1323. #Disable -fopenmp for LAPACK Fortran codes on Windows.
  1324. ifdef OS_WINDOWS
  1325. LAPACK_FFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FFLAGS))
  1326. LAPACK_FPFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FPFLAGS))
  1327. else
  1328. LAPACK_FFLAGS := $(FFLAGS)
  1329. LAPACK_FPFLAGS := $(FPFLAGS)
  1330. endif
  1331. ifeq ($(F_COMPILER),NAG)
  1332. LAPACK_FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1333. endif
  1334. LAPACK_CFLAGS = $(CFLAGS)
  1335. LAPACK_CFLAGS += -DHAVE_LAPACK_CONFIG_H
  1336. ifdef INTERFACE64
  1337. ifneq ($(INTERFACE64), 0)
  1338. LAPACK_CFLAGS += -DLAPACK_ILP64
  1339. endif
  1340. endif
  1341. ifdef OS_WINDOWS
  1342. LAPACK_CFLAGS += -DOPENBLAS_OS_WINDOWS
  1343. endif
  1344. ifeq ($(C_COMPILER), LSB)
  1345. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1346. endif
  1347. ifndef SUFFIX
  1348. SUFFIX = o
  1349. endif
  1350. ifndef PSUFFIX
  1351. PSUFFIX = po
  1352. endif
  1353. ifndef LIBSUFFIX
  1354. LIBSUFFIX = a
  1355. endif
  1356. ifneq ($(DYNAMIC_ARCH), 1)
  1357. ifndef SMP
  1358. LIBNAME = $(LIBPREFIX)_$(LIBCORE)$(REVISION).$(LIBSUFFIX)
  1359. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)$(REVISION)_p.$(LIBSUFFIX)
  1360. else
  1361. LIBNAME = $(LIBPREFIX)_$(LIBCORE)p$(REVISION).$(LIBSUFFIX)
  1362. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)p$(REVISION)_p.$(LIBSUFFIX)
  1363. endif
  1364. else
  1365. ifndef SMP
  1366. LIBNAME = $(LIBPREFIX)$(REVISION).$(LIBSUFFIX)
  1367. LIBNAME_P = $(LIBPREFIX)$(REVISION)_p.$(LIBSUFFIX)
  1368. else
  1369. LIBNAME = $(LIBPREFIX)p$(REVISION).$(LIBSUFFIX)
  1370. LIBNAME_P = $(LIBPREFIX)p$(REVISION)_p.$(LIBSUFFIX)
  1371. endif
  1372. endif
  1373. LIBDLLNAME = $(LIBPREFIX).dll
  1374. IMPLIBNAME = lib$(LIBNAMEBASE).dll.a
  1375. ifneq ($(OSNAME), AIX)
  1376. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.so)
  1377. else
  1378. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.a)
  1379. endif
  1380. LIBDYNNAME = $(LIBNAME:.$(LIBSUFFIX)=.dylib)
  1381. LIBDEFNAME = $(LIBNAME:.$(LIBSUFFIX)=.def)
  1382. LIBEXPNAME = $(LIBNAME:.$(LIBSUFFIX)=.exp)
  1383. LIBZIPNAME = $(LIBNAME:.$(LIBSUFFIX)=.zip)
  1384. LIBS = $(TOPDIR)/$(LIBNAME)
  1385. LIBS_P = $(TOPDIR)/$(LIBNAME_P)
  1386. LIB_COMPONENTS = BLAS
  1387. ifneq ($(NO_CBLAS), 1)
  1388. LIB_COMPONENTS += CBLAS
  1389. endif
  1390. ifneq ($(NO_LAPACK), 1)
  1391. LIB_COMPONENTS += LAPACK
  1392. ifneq ($(NO_LAPACKE), 1)
  1393. LIB_COMPONENTS += LAPACKE
  1394. endif
  1395. ifeq ($(BUILD_RELAPACK), 1)
  1396. LIB_COMPONENTS += ReLAPACK
  1397. endif
  1398. endif
  1399. ifeq ($(ONLY_CBLAS), 1)
  1400. LIB_COMPONENTS = CBLAS
  1401. endif
  1402. export OSNAME
  1403. export ARCH
  1404. export CORE
  1405. export LIBCORE
  1406. export __BYTE_ORDER__
  1407. export ELF_VERSION
  1408. export PGCPATH
  1409. export CONFIG
  1410. export CC
  1411. export FC
  1412. export BU
  1413. export FU
  1414. export NEED2UNDERSCORES
  1415. export USE_THREAD
  1416. export NUM_THREADS
  1417. export NUM_CORES
  1418. export SMP
  1419. export MAKEFILE_RULE
  1420. export NEED_PIC
  1421. export BINARY
  1422. export BINARY32
  1423. export BINARY64
  1424. export F_COMPILER
  1425. export C_COMPILER
  1426. export USE_OPENMP
  1427. export CROSS
  1428. export CROSS_SUFFIX
  1429. export NOFORTRAN
  1430. export NO_FBLAS
  1431. export EXTRALIB
  1432. export CEXTRALIB
  1433. export FEXTRALIB
  1434. export HAVE_SSE
  1435. export HAVE_SSE2
  1436. export HAVE_SSE3
  1437. export HAVE_SSSE3
  1438. export HAVE_SSE4_1
  1439. export HAVE_SSE4_2
  1440. export HAVE_SSE4A
  1441. export HAVE_SSE5
  1442. export HAVE_AVX
  1443. export HAVE_AVX2
  1444. export HAVE_FMA3
  1445. export HAVE_VFP
  1446. export HAVE_VFPV3
  1447. export HAVE_VFPV4
  1448. export HAVE_NEON
  1449. ifndef NO_MSA
  1450. export HAVE_MSA
  1451. export MSA_FLAGS
  1452. endif
  1453. export KERNELDIR
  1454. export FUNCTION_PROFILE
  1455. export TARGET_CORE
  1456. export NO_AVX512
  1457. export NO_AVX2
  1458. export BUILD_BFLOAT16
  1459. export SBGEMM_UNROLL_M
  1460. export SBGEMM_UNROLL_N
  1461. export SGEMM_UNROLL_M
  1462. export SGEMM_UNROLL_N
  1463. export DGEMM_UNROLL_M
  1464. export DGEMM_UNROLL_N
  1465. export QGEMM_UNROLL_M
  1466. export QGEMM_UNROLL_N
  1467. export CGEMM_UNROLL_M
  1468. export CGEMM_UNROLL_N
  1469. export ZGEMM_UNROLL_M
  1470. export ZGEMM_UNROLL_N
  1471. export XGEMM_UNROLL_M
  1472. export XGEMM_UNROLL_N
  1473. export CGEMM3M_UNROLL_M
  1474. export CGEMM3M_UNROLL_N
  1475. export ZGEMM3M_UNROLL_M
  1476. export ZGEMM3M_UNROLL_N
  1477. export XGEMM3M_UNROLL_M
  1478. export XGEMM3M_UNROLL_N
  1479. ifdef USE_CUDA
  1480. export CUDADIR
  1481. export CUCC
  1482. export CUFLAGS
  1483. export CULIB
  1484. endif
  1485. .SUFFIXES: .$(PSUFFIX) .$(SUFFIX) .f
  1486. .f.$(SUFFIX):
  1487. $(FC) $(FFLAGS) -c $< -o $(@F)
  1488. .f.$(PSUFFIX):
  1489. $(FC) $(FPFLAGS) -pg -c $< -o $(@F)
  1490. ifdef BINARY64
  1491. PATHSCALEPATH = /opt/pathscale/lib/3.1
  1492. PGIPATH = /opt/pgi/linux86-64/7.1-5/lib
  1493. else
  1494. PATHSCALEPATH = /opt/pathscale/lib/3.1/32
  1495. PGIPATH = /opt/pgi/linux86/7.1-5/lib
  1496. endif
  1497. ACMLPATH = /opt/acml/4.3.0
  1498. ifneq ($(OSNAME), Darwin)
  1499. MKLPATH = /opt/intel/mkl/10.2.2.025/lib
  1500. else
  1501. MKLPATH = /Library/Frameworks/Intel_MKL.framework/Versions/10.0.1.014/lib
  1502. endif
  1503. ATLASPATH = /opt/atlas/3.9.17/opteron
  1504. FLAMEPATH = $(HOME)/flame/lib
  1505. ifneq ($(OSNAME), SunOS)
  1506. SUNPATH = /opt/sunstudio12.1
  1507. else
  1508. SUNPATH = /opt/SUNWspro
  1509. endif