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 29 kB

7 years ago
13 years ago
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942943944945946947948949950951952953954955956957958959960961962963964965966967968969970971972973974975976977978979980981982983984985986987988989990991992993994995996997998999100010011002100310041005100610071008100910101011101210131014101510161017101810191020102110221023102410251026102710281029103010311032103310341035103610371038103910401041104210431044104510461047104810491050105110521053105410551056105710581059106010611062106310641065106610671068106910701071107210731074107510761077107810791080108110821083108410851086108710881089109010911092109310941095109610971098109911001101110211031104110511061107110811091110111111121113111411151116111711181119112011211122112311241125112611271128112911301131113211331134113511361137113811391140114111421143114411451146114711481149115011511152115311541155115611571158115911601161116211631164116511661167116811691170117111721173117411751176117711781179118011811182118311841185118611871188118911901191119211931194119511961197119811991200120112021203120412051206120712081209121012111212121312141215121612171218121912201221122212231224122512261227122812291230123112321233123412351236123712381239124012411242124312441245124612471248124912501251125212531254125512561257125812591260126112621263126412651266126712681269127012711272127312741275127612771278127912801281128212831284128512861287128812891290129112921293129412951296129712981299130013011302130313041305130613071308130913101311131213131314131513161317131813191320132113221323132413251326132713281329133013311332133313341335133613371338133913401341134213431344134513461347134813491350135113521353135413551356135713581359136013611362136313641365136613671368136913701371137213731374137513761377137813791380138113821383138413851386138713881389139013911392139313941395139613971398139914001401140214031404140514061407140814091410141114121413141414151416141714181419142014211422142314241425142614271428142914301431143214331434143514361437143814391440144114421443144414451446144714481449145014511452145314541455145614571458145914601461146214631464146514661467146814691470147114721473147414751476147714781479148014811482148314841485148614871488148914901491149214931494149514961497149814991500150115021503150415051506150715081509
  1. #
  2. # Include user definition
  3. #
  4. # TO suppress recursive includes
  5. INCLUDED = 1
  6. ifndef TOPDIR
  7. TOPDIR = .
  8. endif
  9. # If ARCH is not set, we use the host system's architecture for getarch compile options.
  10. ifndef ARCH
  11. HOSTARCH := $(shell uname -m)
  12. else
  13. HOSTARCH = $(ARCH)
  14. endif
  15. # Catch conflicting usage of ARCH in some BSD environments
  16. ifeq ($(ARCH), amd64)
  17. override ARCH=x86_64
  18. else ifeq ($(ARCH), powerpc64)
  19. override ARCH=power
  20. else ifeq ($(ARCH), powerpc)
  21. override ARCH=power
  22. else ifeq ($(ARCH), i386)
  23. override ARCH=x86
  24. else ifeq ($(ARCH), aarch64)
  25. override ARCH=arm64
  26. else ifeq ($(ARCH), zarch)
  27. override ARCH=zarch
  28. endif
  29. NETLIB_LAPACK_DIR = $(TOPDIR)/lapack-netlib
  30. # Default C compiler
  31. # - Only set if not specified on the command line or inherited from the environment.
  32. # - CC is an implicit variable so neither '?=' or 'ifndef' can be used.
  33. # http://stackoverflow.com/questions/4029274/mingw-and-make-variables
  34. # - Default value is 'cc' which is not always a valid command (e.g. MinGW).
  35. ifeq ($(origin CC),default)
  36. # Check if $(CC) refers to a valid command and set the value to gcc if not
  37. ifneq ($(findstring cmd.exe,$(SHELL)),)
  38. ifeq ($(shell where $(CC) 2>NUL),)
  39. CC = gcc
  40. endif
  41. else # POSIX-ish
  42. ifeq ($(shell command -v $(CC) 2>/dev/null),)
  43. ifeq ($(shell uname -s),Darwin)
  44. CC = clang
  45. # EXTRALIB += -Wl,-no_compact_unwind
  46. else
  47. CC = gcc
  48. endif # Darwin
  49. endif # CC exists
  50. endif # Shell is sane
  51. endif # CC is set to default
  52. # Default Fortran compiler (FC) is selected by f_check.
  53. ifndef MAKEFILE_RULE
  54. include $(TOPDIR)/Makefile.rule
  55. else
  56. include $(TOPDIR)/$(MAKEFILE_RULE)
  57. endif
  58. #
  59. # Beginning of system configuration
  60. #
  61. ifndef HOSTCC
  62. HOSTCC = $(CC)
  63. endif
  64. ifdef TARGET
  65. GETARCH_FLAGS := -DFORCE_$(TARGET)
  66. GETARCH_FLAGS += -DUSER_TARGET
  67. endif
  68. # Force fallbacks for 32bit
  69. ifeq ($(BINARY), 32)
  70. ifeq ($(TARGET), HASWELL)
  71. GETARCH_FLAGS := -DFORCE_NEHALEM
  72. endif
  73. ifeq ($(TARGET), SKYLAKEX)
  74. GETARCH_FLAGS := -DFORCE_NEHALEM
  75. endif
  76. ifeq ($(TARGET), SANDYBRIDGE)
  77. GETARCH_FLAGS := -DFORCE_NEHALEM
  78. endif
  79. ifeq ($(TARGET), BULLDOZER)
  80. GETARCH_FLAGS := -DFORCE_BARCELONA
  81. endif
  82. ifeq ($(TARGET), PILEDRIVER)
  83. GETARCH_FLAGS := -DFORCE_BARCELONA
  84. endif
  85. ifeq ($(TARGET), STEAMROLLER)
  86. GETARCH_FLAGS := -DFORCE_BARCELONA
  87. endif
  88. ifeq ($(TARGET), EXCAVATOR)
  89. GETARCH_FLAGS := -DFORCE_BARCELONA
  90. endif
  91. ifeq ($(TARGET), ZEN)
  92. GETARCH_FLAGS := -DFORCE_BARCELONA
  93. endif
  94. ifeq ($(TARGET), ARMV8)
  95. GETARCH_FLAGS := -DFORCE_ARMV7
  96. endif
  97. endif
  98. #TARGET_CORE will override TARGET which is used in DYNAMIC_ARCH=1.
  99. #
  100. ifdef TARGET_CORE
  101. GETARCH_FLAGS := -DFORCE_$(TARGET_CORE)
  102. endif
  103. # Force fallbacks for 32bit
  104. ifeq ($(BINARY), 32)
  105. ifeq ($(TARGET_CORE), HASWELL)
  106. GETARCH_FLAGS := -DFORCE_NEHALEM
  107. endif
  108. ifeq ($(TARGET_CORE), SKYLAKEX)
  109. GETARCH_FLAGS := -DFORCE_NEHALEM
  110. endif
  111. ifeq ($(TARGET_CORE), SANDYBRIDGE)
  112. GETARCH_FLAGS := -DFORCE_NEHALEM
  113. endif
  114. ifeq ($(TARGET_CORE), BULLDOZER)
  115. GETARCH_FLAGS := -DFORCE_BARCELONA
  116. endif
  117. ifeq ($(TARGET_CORE), PILEDRIVER)
  118. GETARCH_FLAGS := -DFORCE_BARCELONA
  119. endif
  120. ifeq ($(TARGET_CORE), STEAMROLLER)
  121. GETARCH_FLAGS := -DFORCE_BARCELONA
  122. endif
  123. ifeq ($(TARGET_CORE), EXCAVATOR)
  124. GETARCH_FLAGS := -DFORCE_BARCELONA
  125. endif
  126. ifeq ($(TARGET_CORE), ZEN)
  127. GETARCH_FLAGS := -DFORCE_BARCELONA
  128. endif
  129. endif
  130. # On x86_64 build getarch with march=native unless the compiler is PGI. This is required to detect AVX512 support in getarch.
  131. ifeq ($(HOSTARCH), x86_64)
  132. ifeq ($(findstring pgcc,$(HOSTCC)),)
  133. GETARCH_FLAGS += -march=native
  134. endif
  135. endif
  136. ifdef INTERFACE64
  137. ifneq ($(INTERFACE64), 0)
  138. GETARCH_FLAGS += -DUSE64BITINT
  139. endif
  140. endif
  141. ifndef GEMM_MULTITHREAD_THRESHOLD
  142. GEMM_MULTITHREAD_THRESHOLD=4
  143. endif
  144. GETARCH_FLAGS += -DGEMM_MULTITHREAD_THRESHOLD=$(GEMM_MULTITHREAD_THRESHOLD)
  145. ifeq ($(NO_AVX), 1)
  146. GETARCH_FLAGS += -DNO_AVX
  147. endif
  148. ifeq ($(BINARY), 32)
  149. GETARCH_FLAGS += -DNO_AVX -DNO_AVX2 -DNO_AVX512
  150. NO_AVX512 = 1
  151. endif
  152. ifeq ($(NO_AVX2), 1)
  153. GETARCH_FLAGS += -DNO_AVX2
  154. endif
  155. ifeq ($(NO_AVX512), 1)
  156. GETARCH_FLAGS += -DNO_AVX512
  157. endif
  158. ifeq ($(DEBUG), 1)
  159. GETARCH_FLAGS += -g
  160. endif
  161. ifeq ($(QUIET_MAKE), 1)
  162. MAKE += -s
  163. endif
  164. ifndef NO_PARALLEL_MAKE
  165. NO_PARALLEL_MAKE=0
  166. endif
  167. GETARCH_FLAGS += -DNO_PARALLEL_MAKE=$(NO_PARALLEL_MAKE)
  168. ifdef MAKE_NB_JOBS
  169. GETARCH_FLAGS += -DMAKE_NB_JOBS=$(MAKE_NB_JOBS)
  170. endif
  171. ifeq ($(HOSTCC), loongcc)
  172. GETARCH_FLAGS += -static
  173. endif
  174. #if don't use Fortran, it will only compile CBLAS.
  175. ifeq ($(ONLY_CBLAS), 1)
  176. NO_LAPACK = 1
  177. else
  178. ONLY_CBLAS = 0
  179. endif
  180. # This operation is expensive, so execution should be once.
  181. ifndef GOTOBLAS_MAKEFILE
  182. export GOTOBLAS_MAKEFILE = 1
  183. # Generating Makefile.conf and config.h
  184. DUMMY := $(shell $(MAKE) -C $(TOPDIR) -f Makefile.prebuild CC="$(CC)" FC="$(FC)" HOSTCC="$(HOSTCC)" HOST_CFLAGS="$(GETARCH_FLAGS)" CFLAGS="$(CFLAGS)" BINARY=$(BINARY) USE_OPENMP=$(USE_OPENMP) TARGET_CORE=$(TARGET_CORE) ONLY_CBLAS=$(ONLY_CBLAS) TARGET=$(TARGET) all)
  185. ifndef TARGET_CORE
  186. include $(TOPDIR)/Makefile.conf
  187. else
  188. include $(TOPDIR)/Makefile_kernel.conf
  189. endif
  190. endif
  191. ifndef NUM_PARALLEL
  192. NUM_PARALLEL = 1
  193. endif
  194. ifndef NUM_THREADS
  195. NUM_THREADS = $(NUM_CORES)
  196. endif
  197. ifeq ($(NUM_THREADS), 1)
  198. override USE_THREAD = 0
  199. override USE_OPENMP = 0
  200. endif
  201. ifdef USE_THREAD
  202. ifeq ($(USE_THREAD), 0)
  203. SMP =
  204. else
  205. SMP = 1
  206. endif
  207. else
  208. ifeq ($(NUM_THREAD), 1)
  209. SMP =
  210. else
  211. SMP = 1
  212. endif
  213. endif
  214. ifeq ($(SMP), 1)
  215. USE_LOCKING =
  216. endif
  217. ifndef NEED_PIC
  218. NEED_PIC = 1
  219. endif
  220. ARFLAGS =
  221. CPP = $(COMPILER) -E
  222. AR ?= $(CROSS_SUFFIX)ar
  223. AS ?= $(CROSS_SUFFIX)as
  224. LD ?= $(CROSS_SUFFIX)ld
  225. RANLIB ?= $(CROSS_SUFFIX)ranlib
  226. NM = $(CROSS_SUFFIX)nm
  227. DLLWRAP = $(CROSS_SUFFIX)dllwrap
  228. OBJCOPY = $(CROSS_SUFFIX)objcopy
  229. OBJCONV = $(CROSS_SUFFIX)objconv
  230. # When fortran support was either not detected or actively deselected, only build BLAS.
  231. ifeq ($(NOFORTRAN), 1)
  232. NO_LAPACK = 1
  233. override FEXTRALIB =
  234. endif
  235. ifeq ($(C_COMPILER), GCC)
  236. GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
  237. GCCVERSIONGT4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 4)
  238. GCCVERSIONEQ5 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` = 5)
  239. GCCVERSIONGT5 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 5)
  240. GCCVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 7)
  241. GCCVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  242. GCCMINORVERSIONGTEQ2 := $(shell expr `$(CC) -dumpversion | cut -f2 -d.` \>= 2)
  243. GCCMINORVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f2 -d.` \>= 7)
  244. endif
  245. #
  246. # OS dependent settings
  247. #
  248. ifeq ($(OSNAME), Darwin)
  249. ifndef MACOSX_DEPLOYMENT_TARGET
  250. export MACOSX_DEPLOYMENT_TARGET=10.8
  251. endif
  252. MD5SUM = md5 -r
  253. endif
  254. ifneq (,$(findstring $(OSNAME), FreeBSD OpenBSD DragonFly))
  255. MD5SUM = md5 -r
  256. endif
  257. ifeq ($(OSNAME), NetBSD)
  258. MD5SUM = md5 -n
  259. endif
  260. ifeq ($(OSNAME), Linux)
  261. EXTRALIB += -lm
  262. NO_EXPRECISION = 1
  263. endif
  264. ifeq ($(OSNAME), Android)
  265. EXTRALIB += -lm
  266. endif
  267. ifeq ($(OSNAME), AIX)
  268. EXTRALIB += -lm
  269. endif
  270. ifeq ($(OSNAME), WINNT)
  271. NEED_PIC = 0
  272. NO_EXPRECISION = 1
  273. EXTRALIB += -defaultlib:advapi32
  274. SUFFIX = obj
  275. PSUFFIX = pobj
  276. LIBSUFFIX = a
  277. ifeq ($(C_COMPILER), CLANG)
  278. CCOMMON_OPT += -DMS_ABI
  279. endif
  280. #Version tests for supporting specific features (MS_ABI, POWER9 intrinsics)
  281. ifeq ($(GCCVERSIONGT4), 1)
  282. # GCC Major version > 4
  283. # It is compatible with MSVC ABI.
  284. CCOMMON_OPT += -DMS_ABI
  285. endif
  286. ifeq ($(GCCVERSIONGTEQ4), 1)
  287. ifeq ($(GCCMINORVERSIONGTEQ7), 1)
  288. # GCC Version >=4.7
  289. # It is compatible with MSVC ABI.
  290. CCOMMON_OPT += -DMS_ABI
  291. endif
  292. endif
  293. # Ensure the correct stack alignment on Win32
  294. # http://permalink.gmane.org/gmane.comp.lib.openblas.general/97
  295. ifeq ($(ARCH), x86)
  296. CCOMMON_OPT += -mincoming-stack-boundary=2
  297. FCOMMON_OPT += -mincoming-stack-boundary=2
  298. endif
  299. endif
  300. ifeq ($(OSNAME), Interix)
  301. NEED_PIC = 0
  302. NO_EXPRECISION = 1
  303. INTERIX_TOOL_DIR = /opt/gcc.3.3/i586-pc-interix3/bin
  304. endif
  305. ifeq ($(OSNAME), CYGWIN_NT)
  306. NEED_PIC = 0
  307. NO_EXPRECISION = 1
  308. OS_CYGWIN_NT = 1
  309. endif
  310. ifneq ($(OSNAME), WINNT)
  311. ifneq ($(OSNAME), CYGWIN_NT)
  312. ifneq ($(OSNAME), Interix)
  313. ifneq ($(OSNAME), Android)
  314. ifdef SMP
  315. EXTRALIB += -lpthread
  316. endif
  317. endif
  318. endif
  319. endif
  320. endif
  321. # ifeq logical or
  322. ifeq ($(OSNAME), $(filter $(OSNAME),WINNT CYGWIN_NT Interix))
  323. OS_WINDOWS=1
  324. endif
  325. ifdef QUAD_PRECISION
  326. CCOMMON_OPT += -DQUAD_PRECISION
  327. NO_EXPRECISION = 1
  328. endif
  329. ifneq ($(ARCH), x86)
  330. ifneq ($(ARCH), x86_64)
  331. NO_EXPRECISION = 1
  332. endif
  333. endif
  334. ifdef UTEST_CHECK
  335. CCOMMON_OPT += -DUTEST_CHECK
  336. SANITY_CHECK = 1
  337. endif
  338. ifdef SANITY_CHECK
  339. CCOMMON_OPT += -DSANITY_CHECK -DREFNAME=$(*F)f$(BU)
  340. endif
  341. MAX_STACK_ALLOC ?= 2048
  342. ifneq ($(MAX_STACK_ALLOC), 0)
  343. CCOMMON_OPT += -DMAX_STACK_ALLOC=$(MAX_STACK_ALLOC)
  344. endif
  345. ifdef USE_LOCKING
  346. ifneq ($(USE_LOCKING), 0)
  347. CCOMMON_OPT += -DUSE_LOCKING
  348. endif
  349. endif
  350. #
  351. # Architecture dependent settings
  352. #
  353. ifeq ($(ARCH), x86)
  354. ifndef BINARY
  355. NO_BINARY_MODE = 1
  356. endif
  357. ifeq ($(CORE), generic)
  358. NO_EXPRECISION = 1
  359. endif
  360. ifndef NO_EXPRECISION
  361. ifeq ($(F_COMPILER), GFORTRAN)
  362. # ifeq logical or. GCC or LSB
  363. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  364. EXPRECISION = 1
  365. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  366. FCOMMON_OPT += -m128bit-long-double
  367. endif
  368. ifeq ($(C_COMPILER), CLANG)
  369. EXPRECISION = 1
  370. CCOMMON_OPT += -DEXPRECISION
  371. FCOMMON_OPT += -m128bit-long-double
  372. endif
  373. endif
  374. endif
  375. endif
  376. ifeq ($(ARCH), x86_64)
  377. ifeq ($(CORE), generic)
  378. NO_EXPRECISION = 1
  379. endif
  380. ifndef NO_EXPRECISION
  381. ifeq ($(F_COMPILER), GFORTRAN)
  382. # ifeq logical or. GCC or LSB
  383. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  384. EXPRECISION = 1
  385. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  386. FCOMMON_OPT += -m128bit-long-double
  387. endif
  388. ifeq ($(C_COMPILER), CLANG)
  389. EXPRECISION = 1
  390. CCOMMON_OPT += -DEXPRECISION
  391. FCOMMON_OPT += -m128bit-long-double
  392. endif
  393. endif
  394. endif
  395. endif
  396. ifeq ($(C_COMPILER), INTEL)
  397. CCOMMON_OPT += -wd981
  398. endif
  399. ifeq ($(USE_OPENMP), 1)
  400. #check
  401. ifeq ($(USE_THREAD), 0)
  402. $(error OpenBLAS: Cannot set both USE_OPENMP=1 and USE_THREAD=0. The USE_THREAD=0 is only for building single thread version.)
  403. endif
  404. # ifeq logical or. GCC or LSB
  405. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  406. CCOMMON_OPT += -fopenmp
  407. endif
  408. ifeq ($(C_COMPILER), CLANG)
  409. CCOMMON_OPT += -fopenmp
  410. endif
  411. ifeq ($(C_COMPILER), INTEL)
  412. CCOMMON_OPT += -fopenmp
  413. endif
  414. ifeq ($(C_COMPILER), PGI)
  415. CCOMMON_OPT += -mp
  416. endif
  417. ifeq ($(C_COMPILER), OPEN64)
  418. CCOMMON_OPT += -mp
  419. CEXTRALIB += -lstdc++
  420. endif
  421. ifeq ($(C_COMPILER), PATHSCALE)
  422. CCOMMON_OPT += -mp
  423. endif
  424. endif
  425. ifeq ($(DYNAMIC_ARCH), 1)
  426. ifeq ($(ARCH), x86)
  427. DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
  428. CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
  429. endif
  430. ifeq ($(ARCH), x86_64)
  431. DYNAMIC_CORE = PRESCOTT CORE2
  432. ifeq ($(DYNAMIC_OLDER), 1)
  433. DYNAMIC_CORE += PENRYN DUNNINGTON
  434. endif
  435. DYNAMIC_CORE += NEHALEM
  436. ifeq ($(DYNAMIC_OLDER), 1)
  437. DYNAMIC_CORE += OPTERON OPTERON_SSE3
  438. endif
  439. DYNAMIC_CORE += BARCELONA
  440. ifeq ($(DYNAMIC_OLDER), 1)
  441. DYNAMIC_CORE += BOBCAT ATOM NANO
  442. endif
  443. ifneq ($(NO_AVX), 1)
  444. DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER STEAMROLLER EXCAVATOR
  445. endif
  446. ifneq ($(NO_AVX2), 1)
  447. DYNAMIC_CORE += HASWELL ZEN
  448. endif
  449. ifneq ($(NO_AVX512), 1)
  450. ifneq ($(NO_AVX2), 1)
  451. DYNAMIC_CORE += SKYLAKEX
  452. endif
  453. endif
  454. endif
  455. ifdef DYNAMIC_LIST
  456. override DYNAMIC_CORE = PRESCOTT $(DYNAMIC_LIST)
  457. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_PRESCOTT
  458. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  459. CCOMMON_OPT += $(XCCOMMON_OPT)
  460. #CCOMMON_OPT += -DDYNAMIC_LIST='$(DYNAMIC_LIST)'
  461. endif
  462. ifeq ($(ARCH), arm64)
  463. DYNAMIC_CORE = ARMV8
  464. DYNAMIC_CORE += CORTEXA53
  465. DYNAMIC_CORE += CORTEXA57
  466. DYNAMIC_CORE += CORTEXA72
  467. DYNAMIC_CORE += CORTEXA73
  468. DYNAMIC_CORE += NEOVERSEN1
  469. DYNAMIC_CORE += FALKOR
  470. DYNAMIC_CORE += THUNDERX
  471. DYNAMIC_CORE += THUNDERX2T99
  472. DYNAMIC_CORE += TSV110
  473. DYNAMIC_CORE += EMAG8180
  474. endif
  475. ifeq ($(ARCH), zarch)
  476. DYNAMIC_CORE = ZARCH_GENERIC
  477. # Z13 is supported since gcc-5.2, gcc-6, and in RHEL 7.3 and newer
  478. ifeq ($(GCCVERSIONGT5), 1)
  479. ZARCH_SUPPORT_Z13 := 1
  480. else ifeq ($(GCCVERSIONEQ5), 1)
  481. ifeq ($(GCCMINORVERSIONGTEQ2), 1)
  482. ZARCH_SUPPORT_Z13 := 1
  483. endif
  484. endif
  485. ifeq ($(wildcard /etc/redhat-release), /etc/redhat-release)
  486. ifeq ($(shell source /etc/os-release ; expr $$VERSION_ID \>= "7.3"), 1)
  487. ZARCH_SUPPORT_Z13 := 1
  488. endif
  489. endif
  490. ifeq ($(ZARCH_SUPPORT_Z13), 1)
  491. DYNAMIC_CORE += Z13
  492. else
  493. $(info OpenBLAS: Not building Z13 kernels because gcc is older than 5.2 or 6.x)
  494. endif
  495. ifeq ($(GCCVERSIONGTEQ7), 1)
  496. DYNAMIC_CORE += Z14
  497. else
  498. $(info OpenBLAS: Not building Z14 kernels because gcc is older than 7.x)
  499. endif
  500. endif
  501. ifeq ($(ARCH), power)
  502. DYNAMIC_CORE = POWER6
  503. DYNAMIC_CORE += POWER8
  504. ifneq ($(C_COMPILER), GCC)
  505. DYNAMIC_CORE += POWER9
  506. DYNAMIC_CORE += POWER10
  507. endif
  508. ifeq ($(C_COMPILER), GCC)
  509. ifeq ($(GCCVERSIONGT5), 1)
  510. DYNAMIC_CORE += POWER9
  511. else
  512. $(info, OpenBLAS: Your gcc version is too old to build the POWER9 kernels.)
  513. endif
  514. GCCVERSIONGTEQ11 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 11)
  515. ifeq ($(GCCVERSIONGTEQ11), 1)
  516. DYNAMIC_CORE += POWER10
  517. else
  518. $(info, OpenBLAS: Your gcc version is too old to build the POWER10 kernels.)
  519. endif
  520. endif
  521. endif
  522. # If DYNAMIC_CORE is not set, DYNAMIC_ARCH cannot do anything, so force it to empty
  523. ifndef DYNAMIC_CORE
  524. override DYNAMIC_ARCH=
  525. endif
  526. endif
  527. ifeq ($(ARCH), ia64)
  528. NO_BINARY_MODE = 1
  529. BINARY_DEFINED = 1
  530. ifeq ($(F_COMPILER), GFORTRAN)
  531. ifeq ($(C_COMPILER), GCC)
  532. # EXPRECISION = 1
  533. # CCOMMON_OPT += -DEXPRECISION
  534. endif
  535. endif
  536. endif
  537. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  538. NO_BINARY_MODE = 1
  539. endif
  540. ifeq ($(ARCH), alpha)
  541. NO_BINARY_MODE = 1
  542. BINARY_DEFINED = 1
  543. endif
  544. ifeq ($(ARCH), arm)
  545. NO_BINARY_MODE = 1
  546. BINARY_DEFINED = 1
  547. CCOMMON_OPT += -marm
  548. FCOMMON_OPT += -marm
  549. # If softfp abi is mentioned on the command line, force it.
  550. ifeq ($(ARM_SOFTFP_ABI), 1)
  551. CCOMMON_OPT += -mfloat-abi=softfp
  552. FCOMMON_OPT += -mfloat-abi=softfp
  553. endif
  554. ifeq ($(OSNAME), Android)
  555. ifeq ($(ARM_SOFTFP_ABI), 1)
  556. EXTRALIB += -lm
  557. else
  558. EXTRALIB += -Wl,-lm_hard
  559. endif
  560. endif
  561. endif
  562. ifeq ($(ARCH), arm64)
  563. NO_BINARY_MODE = 1
  564. BINARY_DEFINED = 1
  565. ifdef INTERFACE64
  566. ifneq ($(INTERFACE64), 0)
  567. ifeq ($(F_COMPILER), GFORTRAN)
  568. FCOMMON_OPT += -fdefault-integer-8
  569. endif
  570. ifeq ($(F_COMPILER), FLANG)
  571. FCOMMON_OPT += -i8
  572. endif
  573. endif
  574. endif
  575. endif
  576. #
  577. # C Compiler dependent settings
  578. #
  579. # ifeq logical or. GCC or CLANG or LSB
  580. # http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
  581. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG LSB))
  582. CCOMMON_OPT += -Wall
  583. COMMON_PROF += -fno-inline
  584. NO_UNINITIALIZED_WARN = -Wno-uninitialized
  585. ifeq ($(QUIET_MAKE), 1)
  586. CCOMMON_OPT += $(NO_UNINITIALIZED_WARN) -Wno-unused
  587. endif
  588. ifdef NO_BINARY_MODE
  589. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  590. ifdef BINARY64
  591. CCOMMON_OPT += -mabi=64
  592. else
  593. CCOMMON_OPT += -mabi=n32
  594. endif
  595. BINARY_DEFINED = 1
  596. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  597. CCOMMON_OPT += -mabi=32
  598. BINARY_DEFINED = 1
  599. endif
  600. ifeq ($(CORE), LOONGSON3A)
  601. CCOMMON_OPT += -march=mips64
  602. FCOMMON_OPT += -march=mips64
  603. endif
  604. ifeq ($(CORE), LOONGSON3B)
  605. CCOMMON_OPT += -march=mips64
  606. FCOMMON_OPT += -march=mips64
  607. endif
  608. ifeq ($(CORE), MIPS24K)
  609. CCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  610. FCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  611. endif
  612. ifeq ($(CORE), MIPS1004K)
  613. CCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  614. FCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  615. endif
  616. ifeq ($(CORE), P5600)
  617. CCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  618. FCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  619. endif
  620. ifeq ($(CORE), I6400)
  621. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  622. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  623. endif
  624. ifeq ($(CORE), P6600)
  625. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  626. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  627. endif
  628. ifeq ($(CORE), I6500)
  629. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  630. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  631. endif
  632. ifeq ($(OSNAME), AIX)
  633. BINARY_DEFINED = 1
  634. endif
  635. endif
  636. ifndef BINARY_DEFINED
  637. ifneq ($(OSNAME), AIX)
  638. ifdef BINARY64
  639. CCOMMON_OPT += -m64
  640. else
  641. CCOMMON_OPT += -m32
  642. endif
  643. endif
  644. endif
  645. endif
  646. ifeq ($(C_COMPILER), PGI)
  647. ifdef BINARY64
  648. CCOMMON_OPT += -tp p7-64 -D__MMX__ -Mnollvm
  649. else
  650. CCOMMON_OPT += -tp p7
  651. endif
  652. endif
  653. ifeq ($(C_COMPILER), PATHSCALE)
  654. ifdef BINARY64
  655. CCOMMON_OPT += -m64
  656. else
  657. CCOMMON_OPT += -m32
  658. endif
  659. endif
  660. #
  661. # Fortran Compiler dependent settings
  662. #
  663. ifeq ($(F_COMPILER), FLANG)
  664. CCOMMON_OPT += -DF_INTERFACE_FLANG
  665. FCOMMON_OPT += -Mrecursive -Kieee
  666. ifeq ($(OSNAME), Linux)
  667. ifeq ($(ARCH), x86_64)
  668. FLANG_VENDOR := $(shell expr `$(FC) --version|cut -f 1 -d "."|head -1`)
  669. ifeq ($(FLANG_VENDOR),AOCC)
  670. FCOMMON_OPT += -fno-unroll-loops
  671. endif
  672. endif
  673. endif
  674. ifdef BINARY64
  675. ifdef INTERFACE64
  676. ifneq ($(INTERFACE64), 0)
  677. FCOMMON_OPT += -i8
  678. endif
  679. endif
  680. FCOMMON_OPT += -Wall
  681. else
  682. FCOMMON_OPT += -Wall
  683. endif
  684. ifeq ($(USE_OPENMP), 1)
  685. FCOMMON_OPT += -fopenmp
  686. endif
  687. endif
  688. ifeq ($(F_COMPILER), G77)
  689. CCOMMON_OPT += -DF_INTERFACE_G77
  690. FCOMMON_OPT += -Wall
  691. ifndef NO_BINARY_MODE
  692. ifneq ($(OSNAME), AIX)
  693. ifdef BINARY64
  694. FCOMMON_OPT += -m64
  695. else
  696. FCOMMON_OPT += -m32
  697. endif
  698. endif
  699. endif
  700. endif
  701. ifeq ($(F_COMPILER), G95)
  702. CCOMMON_OPT += -DF_INTERFACE_G95
  703. FCOMMON_OPT += -Wall
  704. ifneq ($(OSNAME), AIX)
  705. ifndef NO_BINARY_MODE
  706. ifdef BINARY64
  707. FCOMMON_OPT += -m64
  708. else
  709. FCOMMON_OPT += -m32
  710. endif
  711. endif
  712. ifneq ($(NO_LAPACKE), 1)
  713. FCOMMON_OPT += -fno-second-underscore
  714. endif
  715. endif
  716. endif
  717. ifeq ($(F_COMPILER), GFORTRAN)
  718. CCOMMON_OPT += -DF_INTERFACE_GFORT
  719. FCOMMON_OPT += -Wall
  720. # make single-threaded LAPACK calls thread-safe #1847
  721. FCOMMON_OPT += -frecursive
  722. # work around ABI problem with passing single-character arguments
  723. FCOMMON_OPT += -fno-optimize-sibling-calls
  724. #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
  725. ifneq ($(NO_LAPACK), 1)
  726. EXTRALIB += -lgfortran
  727. endif
  728. ifdef NO_BINARY_MODE
  729. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  730. ifdef BINARY64
  731. FCOMMON_OPT += -mabi=64
  732. else
  733. FCOMMON_OPT += -mabi=n32
  734. endif
  735. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  736. FCOMMON_OPT += -mabi=32
  737. endif
  738. else
  739. ifdef BINARY64
  740. ifneq ($(OSNAME), AIX)
  741. FCOMMON_OPT += -m64
  742. endif
  743. ifdef INTERFACE64
  744. ifneq ($(INTERFACE64), 0)
  745. FCOMMON_OPT += -fdefault-integer-8
  746. endif
  747. endif
  748. else
  749. ifneq ($(OSNAME), AIX)
  750. FCOMMON_OPT += -m32
  751. endif
  752. endif
  753. endif
  754. ifeq ($(USE_OPENMP), 1)
  755. FCOMMON_OPT += -fopenmp
  756. endif
  757. endif
  758. ifeq ($(F_COMPILER), INTEL)
  759. CCOMMON_OPT += -DF_INTERFACE_INTEL
  760. ifdef INTERFACE64
  761. ifneq ($(INTERFACE64), 0)
  762. FCOMMON_OPT += -i8
  763. endif
  764. endif
  765. FCOMMON_OPT += -recursive -fp-model strict -assume protect-parens
  766. ifeq ($(USE_OPENMP), 1)
  767. FCOMMON_OPT += -fopenmp
  768. endif
  769. endif
  770. ifeq ($(F_COMPILER), FUJITSU)
  771. CCOMMON_OPT += -DF_INTERFACE_FUJITSU
  772. ifeq ($(USE_OPENMP), 1)
  773. FCOMMON_OPT += -openmp
  774. endif
  775. endif
  776. ifeq ($(F_COMPILER), IBM)
  777. CCOMMON_OPT += -DF_INTERFACE_IBM
  778. # FCOMMON_OPT += -qarch=440
  779. ifdef BINARY64
  780. FCOMMON_OPT += -q64
  781. ifdef INTERFACE64
  782. ifneq ($(INTERFACE64), 0)
  783. FCOMMON_OPT += -qintsize=8
  784. endif
  785. endif
  786. else
  787. FCOMMON_OPT += -q32
  788. endif
  789. ifeq ($(USE_OPENMP), 1)
  790. FCOMMON_OPT += -openmp
  791. endif
  792. endif
  793. ifeq ($(F_COMPILER), PGI)
  794. CCOMMON_OPT += -DF_INTERFACE_PGI
  795. COMMON_PROF += -DPGICOMPILER
  796. ifdef BINARY64
  797. ifdef INTERFACE64
  798. ifneq ($(INTERFACE64), 0)
  799. FCOMMON_OPT += -i8
  800. endif
  801. endif
  802. FCOMMON_OPT += -tp p7-64
  803. else
  804. FCOMMON_OPT += -tp p7
  805. endif
  806. FCOMMON_OPT += -Mrecursive
  807. ifeq ($(USE_OPENMP), 1)
  808. FCOMMON_OPT += -mp
  809. endif
  810. endif
  811. ifeq ($(F_COMPILER), PATHSCALE)
  812. CCOMMON_OPT += -DF_INTERFACE_PATHSCALE
  813. ifdef BINARY64
  814. ifdef INTERFACE64
  815. ifneq ($(INTERFACE64), 0)
  816. FCOMMON_OPT += -i8
  817. endif
  818. endif
  819. endif
  820. ifeq ($(USE_OPENMP), 1)
  821. FCOMMON_OPT += -mp
  822. endif
  823. endif
  824. ifeq ($(F_COMPILER), OPEN64)
  825. CCOMMON_OPT += -DF_INTERFACE_OPEN64
  826. ifdef BINARY64
  827. ifdef INTERFACE64
  828. ifneq ($(INTERFACE64), 0)
  829. FCOMMON_OPT += -i8
  830. endif
  831. endif
  832. endif
  833. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  834. ifndef BINARY64
  835. FCOMMON_OPT += -n32
  836. else
  837. FCOMMON_OPT += -n64
  838. endif
  839. ifeq ($(CORE), LOONGSON3A)
  840. FCOMMON_OPT += -loongson3 -static
  841. endif
  842. ifeq ($(CORE), LOONGSON3B)
  843. FCOMMON_OPT += -loongson3 -static
  844. endif
  845. else
  846. ifndef BINARY64
  847. FCOMMON_OPT += -m32
  848. else
  849. FCOMMON_OPT += -m64
  850. endif
  851. endif
  852. ifeq ($(USE_OPENMP), 1)
  853. FEXTRALIB += -lstdc++
  854. FCOMMON_OPT += -mp
  855. endif
  856. endif
  857. ifeq ($(C_COMPILER), OPEN64)
  858. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  859. ifndef BINARY64
  860. CCOMMON_OPT += -n32
  861. else
  862. CCOMMON_OPT += -n64
  863. endif
  864. ifeq ($(CORE), LOONGSON3A)
  865. CCOMMON_OPT += -loongson3 -static
  866. endif
  867. ifeq ($(CORE), LOONGSON3B)
  868. CCOMMON_OPT += -loongson3 -static
  869. endif
  870. else
  871. ifndef BINARY64
  872. CCOMMON_OPT += -m32
  873. else
  874. CCOMMON_OPT += -m64
  875. endif
  876. endif
  877. endif
  878. ifeq ($(C_COMPILER), SUN)
  879. CCOMMON_OPT += -w
  880. ifeq ($(ARCH), x86)
  881. CCOMMON_OPT += -m32
  882. else
  883. FCOMMON_OPT += -m64
  884. endif
  885. endif
  886. ifeq ($(F_COMPILER), SUN)
  887. CCOMMON_OPT += -DF_INTERFACE_SUN
  888. ifeq ($(ARCH), x86)
  889. FCOMMON_OPT += -m32
  890. else
  891. FCOMMON_OPT += -m64
  892. endif
  893. ifeq ($(USE_OPENMP), 1)
  894. FCOMMON_OPT += -xopenmp=parallel
  895. endif
  896. endif
  897. ifeq ($(F_COMPILER), COMPAQ)
  898. CCOMMON_OPT += -DF_INTERFACE_COMPAQ
  899. ifeq ($(USE_OPENMP), 1)
  900. FCOMMON_OPT += -openmp
  901. endif
  902. endif
  903. ifdef BINARY64
  904. ifdef INTERFACE64
  905. ifneq ($(INTERFACE64), 0)
  906. CCOMMON_OPT +=
  907. #-DUSE64BITINT
  908. endif
  909. endif
  910. endif
  911. ifeq ($(NEED_PIC), 1)
  912. ifeq ($(C_COMPILER), IBM)
  913. CCOMMON_OPT += -qpic=large
  914. else
  915. CCOMMON_OPT += -fPIC
  916. endif
  917. ifeq ($(F_COMPILER), SUN)
  918. FCOMMON_OPT += -pic
  919. else
  920. FCOMMON_OPT += -fPIC
  921. endif
  922. endif
  923. ifeq ($(DYNAMIC_ARCH), 1)
  924. CCOMMON_OPT += -DDYNAMIC_ARCH
  925. endif
  926. ifeq ($(DYNAMIC_OLDER), 1)
  927. CCOMMON_OPT += -DDYNAMIC_OLDER
  928. endif
  929. ifeq ($(NO_LAPACK), 1)
  930. CCOMMON_OPT += -DNO_LAPACK
  931. #Disable LAPACK C interface
  932. NO_LAPACKE = 1
  933. endif
  934. ifeq ($(NO_LAPACKE), 1)
  935. CCOMMON_OPT += -DNO_LAPACKE
  936. endif
  937. ifeq ($(NO_AVX), 1)
  938. CCOMMON_OPT += -DNO_AVX
  939. endif
  940. ifeq ($(ARCH), x86)
  941. CCOMMON_OPT += -DNO_AVX
  942. endif
  943. ifeq ($(NO_AVX2), 1)
  944. CCOMMON_OPT += -DNO_AVX2
  945. endif
  946. ifeq ($(NO_AVX512), 1)
  947. CCOMMON_OPT += -DNO_AVX512
  948. endif
  949. ifdef SMP
  950. CCOMMON_OPT += -DSMP_SERVER
  951. ifeq ($(ARCH), mips64)
  952. ifneq ($(CORE), LOONGSON3B)
  953. USE_SIMPLE_THREADED_LEVEL3 = 1
  954. endif
  955. endif
  956. ifeq ($(USE_OPENMP), 1)
  957. # USE_SIMPLE_THREADED_LEVEL3 = 1
  958. # NO_AFFINITY = 1
  959. CCOMMON_OPT += -DUSE_OPENMP
  960. endif
  961. ifeq ($(BIGNUMA), 1)
  962. CCOMMON_OPT += -DBIGNUMA
  963. endif
  964. endif
  965. ifeq ($(NO_WARMUP), 1)
  966. CCOMMON_OPT += -DNO_WARMUP
  967. endif
  968. ifeq ($(CONSISTENT_FPCSR), 1)
  969. CCOMMON_OPT += -DCONSISTENT_FPCSR
  970. endif
  971. # Only for development
  972. # CCOMMON_OPT += -DPARAMTEST
  973. # CCOMMON_OPT += -DPREFETCHTEST
  974. # CCOMMON_OPT += -DNO_SWITCHING
  975. # USE_PAPI = 1
  976. ifdef USE_PAPI
  977. CCOMMON_OPT += -DUSE_PAPI
  978. EXTRALIB += -lpapi -lperfctr
  979. endif
  980. ifdef DYNAMIC_THREADS
  981. CCOMMON_OPT += -DDYNAMIC_THREADS
  982. endif
  983. CCOMMON_OPT += -DMAX_CPU_NUMBER=$(NUM_THREADS)
  984. CCOMMON_OPT += -DMAX_PARALLEL_NUMBER=$(NUM_PARALLEL)
  985. ifdef USE_SIMPLE_THREADED_LEVEL3
  986. CCOMMON_OPT += -DUSE_SIMPLE_THREADED_LEVEL3
  987. endif
  988. ifeq ($(USE_TLS), 1)
  989. CCOMMON_OPT += -DUSE_TLS
  990. endif
  991. ifeq ($(BUILD_HALF), 1)
  992. CCOMMON_OPT += -DBUILD_HALF
  993. endif
  994. CCOMMON_OPT += -DVERSION=\"$(VERSION)\"
  995. ifndef SYMBOLPREFIX
  996. SYMBOLPREFIX =
  997. endif
  998. ifndef SYMBOLSUFFIX
  999. SYMBOLSUFFIX =
  1000. endif
  1001. ifndef LIBNAMESUFFIX
  1002. LIBNAMEBASE = $(SYMBOLPREFIX)openblas$(SYMBOLSUFFIX)
  1003. else
  1004. LIBNAMEBASE = $(SYMBOLPREFIX)openblas$(SYMBOLSUFFIX)_$(LIBNAMESUFFIX)
  1005. endif
  1006. ifeq ($(OSNAME), CYGWIN_NT)
  1007. LIBPREFIX = cyg$(LIBNAMEBASE)
  1008. else
  1009. LIBPREFIX = lib$(LIBNAMEBASE)
  1010. endif
  1011. KERNELDIR = $(TOPDIR)/kernel/$(ARCH)
  1012. include $(TOPDIR)/Makefile.$(ARCH)
  1013. CCOMMON_OPT += -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME
  1014. CCOMMON_OPT += -DASMNAME=$(FU)$(*F) -DASMFNAME=$(FU)$(*F)$(BU) -DNAME=$(*F)$(BU) -DCNAME=$(*F) -DCHAR_NAME=\"$(*F)$(BU)\" -DCHAR_CNAME=\"$(*F)\"
  1015. ifeq ($(CORE), PPC440)
  1016. CCOMMON_OPT += -DALLOC_QALLOC
  1017. endif
  1018. ifeq ($(CORE), PPC440FP2)
  1019. STATIC_ALLOCATION = 1
  1020. endif
  1021. ifneq ($(OSNAME), Linux)
  1022. NO_AFFINITY = 1
  1023. endif
  1024. ifneq ($(ARCH), x86_64)
  1025. ifneq ($(ARCH), x86)
  1026. ifneq ($(CORE), LOONGSON3B)
  1027. NO_AFFINITY = 1
  1028. endif
  1029. endif
  1030. endif
  1031. ifdef NO_AFFINITY
  1032. ifeq ($(NO_AFFINITY), 0)
  1033. override undefine NO_AFFINITY
  1034. else
  1035. CCOMMON_OPT += -DNO_AFFINITY
  1036. endif
  1037. endif
  1038. ifdef FUNCTION_PROFILE
  1039. CCOMMON_OPT += -DFUNCTION_PROFILE
  1040. endif
  1041. ifdef HUGETLB_ALLOCATION
  1042. CCOMMON_OPT += -DALLOC_HUGETLB
  1043. endif
  1044. ifdef HUGETLBFILE_ALLOCATION
  1045. CCOMMON_OPT += -DALLOC_HUGETLBFILE -DHUGETLB_FILE_NAME=$(HUGETLBFILE_ALLOCATION)
  1046. endif
  1047. ifdef STATIC_ALLOCATION
  1048. CCOMMON_OPT += -DALLOC_STATIC
  1049. endif
  1050. ifdef DEVICEDRIVER_ALLOCATION
  1051. CCOMMON_OPT += -DALLOC_DEVICEDRIVER -DDEVICEDRIVER_NAME=\"/dev/mapper\"
  1052. endif
  1053. ifdef MIXED_MEMORY_ALLOCATION
  1054. CCOMMON_OPT += -DMIXED_MEMORY_ALLOCATION
  1055. endif
  1056. ifeq ($(OSNAME), SunOS)
  1057. TAR = gtar
  1058. PATCH = gpatch
  1059. GREP = ggrep
  1060. AWK = nawk
  1061. else
  1062. TAR = tar
  1063. PATCH = patch
  1064. GREP = grep
  1065. AWK = awk
  1066. endif
  1067. ifndef MD5SUM
  1068. MD5SUM = md5sum
  1069. endif
  1070. REVISION = -r$(VERSION)
  1071. MAJOR_VERSION = $(word 1,$(subst ., ,$(VERSION)))
  1072. ifeq ($(DEBUG), 1)
  1073. COMMON_OPT += -g
  1074. endif
  1075. ifeq ($(DEBUG), 1)
  1076. FCOMMON_OPT += -g
  1077. endif
  1078. ifndef COMMON_OPT
  1079. COMMON_OPT = -O2
  1080. endif
  1081. ifndef FCOMMON_OPT
  1082. FCOMMON_OPT = -O2 -frecursive
  1083. endif
  1084. override CFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
  1085. override PFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
  1086. override FFLAGS += $(COMMON_OPT) $(FCOMMON_OPT)
  1087. override FPFLAGS += $(FCOMMON_OPT) $(COMMON_PROF)
  1088. #MAKEOVERRIDES =
  1089. ifdef NEED_PIC
  1090. ifeq (,$(findstring PIC,$(FFLAGS)))
  1091. override FFLAGS += -fPIC
  1092. endif
  1093. endif
  1094. #For LAPACK Fortran codes.
  1095. #Disable -fopenmp for LAPACK Fortran codes on Windows.
  1096. ifdef OS_WINDOWS
  1097. LAPACK_FFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FFLAGS))
  1098. LAPACK_FPFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FPFLAGS))
  1099. else
  1100. LAPACK_FFLAGS := $(FFLAGS)
  1101. LAPACK_FPFLAGS := $(FPFLAGS)
  1102. endif
  1103. LAPACK_CFLAGS = $(CFLAGS)
  1104. LAPACK_CFLAGS += -DHAVE_LAPACK_CONFIG_H
  1105. ifdef INTERFACE64
  1106. ifneq ($(INTERFACE64), 0)
  1107. LAPACK_CFLAGS += -DLAPACK_ILP64
  1108. endif
  1109. endif
  1110. ifdef OS_WINDOWS
  1111. LAPACK_CFLAGS += -DOPENBLAS_OS_WINDOWS
  1112. endif
  1113. ifeq ($(C_COMPILER), LSB)
  1114. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1115. endif
  1116. ifndef SUFFIX
  1117. SUFFIX = o
  1118. endif
  1119. ifndef PSUFFIX
  1120. PSUFFIX = po
  1121. endif
  1122. ifndef LIBSUFFIX
  1123. LIBSUFFIX = a
  1124. endif
  1125. ifneq ($(DYNAMIC_ARCH), 1)
  1126. ifndef SMP
  1127. LIBNAME = $(LIBPREFIX)_$(LIBCORE)$(REVISION).$(LIBSUFFIX)
  1128. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)$(REVISION)_p.$(LIBSUFFIX)
  1129. else
  1130. LIBNAME = $(LIBPREFIX)_$(LIBCORE)p$(REVISION).$(LIBSUFFIX)
  1131. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)p$(REVISION)_p.$(LIBSUFFIX)
  1132. endif
  1133. else
  1134. ifndef SMP
  1135. LIBNAME = $(LIBPREFIX)$(REVISION).$(LIBSUFFIX)
  1136. LIBNAME_P = $(LIBPREFIX)$(REVISION)_p.$(LIBSUFFIX)
  1137. else
  1138. LIBNAME = $(LIBPREFIX)p$(REVISION).$(LIBSUFFIX)
  1139. LIBNAME_P = $(LIBPREFIX)p$(REVISION)_p.$(LIBSUFFIX)
  1140. endif
  1141. endif
  1142. LIBDLLNAME = $(LIBPREFIX).dll
  1143. IMPLIBNAME = lib$(LIBNAMEBASE).dll.a
  1144. ifneq ($(OSNAME), AIX)
  1145. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.so)
  1146. else
  1147. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.a)
  1148. endif
  1149. LIBDYNNAME = $(LIBNAME:.$(LIBSUFFIX)=.dylib)
  1150. LIBDEFNAME = $(LIBNAME:.$(LIBSUFFIX)=.def)
  1151. LIBEXPNAME = $(LIBNAME:.$(LIBSUFFIX)=.exp)
  1152. LIBZIPNAME = $(LIBNAME:.$(LIBSUFFIX)=.zip)
  1153. LIBS = $(TOPDIR)/$(LIBNAME)
  1154. LIBS_P = $(TOPDIR)/$(LIBNAME_P)
  1155. LIB_COMPONENTS = BLAS
  1156. ifneq ($(NO_CBLAS), 1)
  1157. LIB_COMPONENTS += CBLAS
  1158. endif
  1159. ifneq ($(NO_LAPACK), 1)
  1160. LIB_COMPONENTS += LAPACK
  1161. ifneq ($(NO_LAPACKE), 1)
  1162. LIB_COMPONENTS += LAPACKE
  1163. endif
  1164. ifeq ($(BUILD_RELAPACK), 1)
  1165. LIB_COMPONENTS += ReLAPACK
  1166. endif
  1167. endif
  1168. ifeq ($(ONLY_CBLAS), 1)
  1169. LIB_COMPONENTS = CBLAS
  1170. endif
  1171. export OSNAME
  1172. export ARCH
  1173. export CORE
  1174. export LIBCORE
  1175. export __BYTE_ORDER__
  1176. export ELF_VERSION
  1177. export PGCPATH
  1178. export CONFIG
  1179. export CC
  1180. export FC
  1181. export BU
  1182. export FU
  1183. export NEED2UNDERSCORES
  1184. export USE_THREAD
  1185. export NUM_THREADS
  1186. export NUM_CORES
  1187. export SMP
  1188. export MAKEFILE_RULE
  1189. export NEED_PIC
  1190. export BINARY
  1191. export BINARY32
  1192. export BINARY64
  1193. export F_COMPILER
  1194. export C_COMPILER
  1195. export USE_OPENMP
  1196. export CROSS
  1197. export CROSS_SUFFIX
  1198. export NOFORTRAN
  1199. export NO_FBLAS
  1200. export EXTRALIB
  1201. export CEXTRALIB
  1202. export FEXTRALIB
  1203. export HAVE_SSE
  1204. export HAVE_SSE2
  1205. export HAVE_SSE3
  1206. export HAVE_SSSE3
  1207. export HAVE_SSE4_1
  1208. export HAVE_SSE4_2
  1209. export HAVE_SSE4A
  1210. export HAVE_SSE5
  1211. export HAVE_AVX
  1212. export HAVE_VFP
  1213. export HAVE_VFPV3
  1214. export HAVE_VFPV4
  1215. export HAVE_NEON
  1216. export HAVE_MSA
  1217. export MSA_FLAGS
  1218. export KERNELDIR
  1219. export FUNCTION_PROFILE
  1220. export TARGET_CORE
  1221. export NO_AVX512
  1222. export BUILD_HALF
  1223. export SHGEMM_UNROLL_M
  1224. export SHGEMM_UNROLL_N
  1225. export SGEMM_UNROLL_M
  1226. export SGEMM_UNROLL_N
  1227. export DGEMM_UNROLL_M
  1228. export DGEMM_UNROLL_N
  1229. export QGEMM_UNROLL_M
  1230. export QGEMM_UNROLL_N
  1231. export CGEMM_UNROLL_M
  1232. export CGEMM_UNROLL_N
  1233. export ZGEMM_UNROLL_M
  1234. export ZGEMM_UNROLL_N
  1235. export XGEMM_UNROLL_M
  1236. export XGEMM_UNROLL_N
  1237. export CGEMM3M_UNROLL_M
  1238. export CGEMM3M_UNROLL_N
  1239. export ZGEMM3M_UNROLL_M
  1240. export ZGEMM3M_UNROLL_N
  1241. export XGEMM3M_UNROLL_M
  1242. export XGEMM3M_UNROLL_N
  1243. ifdef USE_CUDA
  1244. export CUDADIR
  1245. export CUCC
  1246. export CUFLAGS
  1247. export CULIB
  1248. endif
  1249. .SUFFIXES: .$(PSUFFIX) .$(SUFFIX) .f
  1250. .f.$(SUFFIX):
  1251. $(FC) $(FFLAGS) -c $< -o $(@F)
  1252. .f.$(PSUFFIX):
  1253. $(FC) $(FPFLAGS) -pg -c $< -o $(@F)
  1254. ifdef BINARY64
  1255. PATHSCALEPATH = /opt/pathscale/lib/3.1
  1256. PGIPATH = /opt/pgi/linux86-64/7.1-5/lib
  1257. else
  1258. PATHSCALEPATH = /opt/pathscale/lib/3.1/32
  1259. PGIPATH = /opt/pgi/linux86/7.1-5/lib
  1260. endif
  1261. ACMLPATH = /opt/acml/4.3.0
  1262. ifneq ($(OSNAME), Darwin)
  1263. MKLPATH = /opt/intel/mkl/10.2.2.025/lib
  1264. else
  1265. MKLPATH = /Library/Frameworks/Intel_MKL.framework/Versions/10.0.1.014/lib
  1266. endif
  1267. ATLASPATH = /opt/atlas/3.9.17/opteron
  1268. FLAMEPATH = $(HOME)/flame/lib
  1269. ifneq ($(OSNAME), SunOS)
  1270. SUNPATH = /opt/sunstudio12.1
  1271. else
  1272. SUNPATH = /opt/SUNWspro
  1273. endif