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 30 kB

7 years ago
13 years ago
12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448144914501451145214531454145514561457145814591460146114621463146414651466146714681469147014711472147314741475147614771478147914801481148214831484148514861487148814891490149114921493149414951496149714981499150015011502150315041505150615071508150915101511151215131514151515161517151815191520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559
  1. #
  2. # Include user definition
  3. #
  4. # TO suppress recursive includes
  5. INCLUDED = 1
  6. ifndef TOPDIR
  7. TOPDIR = .
  8. endif
  9. # If ARCH is not set, we use the host system's architecture for getarch compile options.
  10. ifndef ARCH
  11. HOSTARCH := $(shell uname -m)
  12. else
  13. HOSTARCH = $(ARCH)
  14. endif
  15. # Catch conflicting usage of ARCH in some BSD environments
  16. ifeq ($(ARCH), amd64)
  17. override ARCH=x86_64
  18. else ifeq ($(ARCH), powerpc64)
  19. override ARCH=power
  20. else ifeq ($(ARCH), powerpc)
  21. override ARCH=power
  22. else ifeq ($(ARCH), i386)
  23. override ARCH=x86
  24. else ifeq ($(ARCH), armv6)
  25. override ARCH=arm
  26. else ifeq ($(ARCH), armv7)
  27. override ARCH=arm
  28. else ifeq ($(ARCH), aarch64)
  29. override ARCH=arm64
  30. else ifeq ($(ARCH), zarch)
  31. override ARCH=zarch
  32. endif
  33. NETLIB_LAPACK_DIR = $(TOPDIR)/lapack-netlib
  34. # Default C compiler
  35. # - Only set if not specified on the command line or inherited from the environment.
  36. # - CC is an implicit variable so neither '?=' or 'ifndef' can be used.
  37. # http://stackoverflow.com/questions/4029274/mingw-and-make-variables
  38. # - Default value is 'cc' which is not always a valid command (e.g. MinGW).
  39. ifeq ($(origin CC),default)
  40. # Check if $(CC) refers to a valid command and set the value to gcc if not
  41. ifneq ($(findstring cmd.exe,$(SHELL)),)
  42. ifeq ($(shell where $(CC) 2>NUL),)
  43. CC = gcc
  44. endif
  45. else # POSIX-ish
  46. ifeq ($(shell command -v $(CC) 2>/dev/null),)
  47. ifeq ($(shell uname -s),Darwin)
  48. CC = clang
  49. # EXTRALIB += -Wl,-no_compact_unwind
  50. else
  51. CC = gcc
  52. endif # Darwin
  53. endif # CC exists
  54. endif # Shell is sane
  55. endif # CC is set to default
  56. # Default Fortran compiler (FC) is selected by f_check.
  57. ifndef MAKEFILE_RULE
  58. include $(TOPDIR)/Makefile.rule
  59. else
  60. include $(TOPDIR)/$(MAKEFILE_RULE)
  61. endif
  62. #
  63. # Beginning of system configuration
  64. #
  65. ifndef HOSTCC
  66. HOSTCC = $(CC)
  67. endif
  68. ifdef TARGET
  69. GETARCH_FLAGS := -DFORCE_$(TARGET)
  70. GETARCH_FLAGS += -DUSER_TARGET
  71. endif
  72. # Force fallbacks for 32bit
  73. ifeq ($(BINARY), 32)
  74. ifeq ($(TARGET), HASWELL)
  75. GETARCH_FLAGS := -DFORCE_NEHALEM
  76. endif
  77. ifeq ($(TARGET), SKYLAKEX)
  78. GETARCH_FLAGS := -DFORCE_NEHALEM
  79. endif
  80. ifeq ($(TARGET), COOPERLAKE)
  81. GETARCH_FLAGS := -DFORCE_NEHALEM
  82. endif
  83. ifeq ($(TARGET), SANDYBRIDGE)
  84. GETARCH_FLAGS := -DFORCE_NEHALEM
  85. endif
  86. ifeq ($(TARGET), BULLDOZER)
  87. GETARCH_FLAGS := -DFORCE_BARCELONA
  88. endif
  89. ifeq ($(TARGET), PILEDRIVER)
  90. GETARCH_FLAGS := -DFORCE_BARCELONA
  91. endif
  92. ifeq ($(TARGET), STEAMROLLER)
  93. GETARCH_FLAGS := -DFORCE_BARCELONA
  94. endif
  95. ifeq ($(TARGET), EXCAVATOR)
  96. GETARCH_FLAGS := -DFORCE_BARCELONA
  97. endif
  98. ifeq ($(TARGET), ZEN)
  99. GETARCH_FLAGS := -DFORCE_BARCELONA
  100. endif
  101. ifeq ($(TARGET), ARMV8)
  102. GETARCH_FLAGS := -DFORCE_ARMV7
  103. endif
  104. ifeq ($(TARGET), POWER8)
  105. GETARCH_FLAGS := -DFORCE_POWER6
  106. endif
  107. endif
  108. #TARGET_CORE will override TARGET which is used in DYNAMIC_ARCH=1.
  109. #
  110. ifdef TARGET_CORE
  111. GETARCH_FLAGS := -DFORCE_$(TARGET_CORE)
  112. endif
  113. # Force fallbacks for 32bit
  114. ifeq ($(BINARY), 32)
  115. ifeq ($(TARGET_CORE), HASWELL)
  116. GETARCH_FLAGS := -DFORCE_NEHALEM
  117. endif
  118. ifeq ($(TARGET_CORE), SKYLAKEX)
  119. GETARCH_FLAGS := -DFORCE_NEHALEM
  120. endif
  121. ifeq ($(TARGET_CORE), COOPERLAKE)
  122. GETARCH_FLAGS := -DFORCE_NEHALEM
  123. endif
  124. ifeq ($(TARGET_CORE), SANDYBRIDGE)
  125. GETARCH_FLAGS := -DFORCE_NEHALEM
  126. endif
  127. ifeq ($(TARGET_CORE), BULLDOZER)
  128. GETARCH_FLAGS := -DFORCE_BARCELONA
  129. endif
  130. ifeq ($(TARGET_CORE), PILEDRIVER)
  131. GETARCH_FLAGS := -DFORCE_BARCELONA
  132. endif
  133. ifeq ($(TARGET_CORE), STEAMROLLER)
  134. GETARCH_FLAGS := -DFORCE_BARCELONA
  135. endif
  136. ifeq ($(TARGET_CORE), EXCAVATOR)
  137. GETARCH_FLAGS := -DFORCE_BARCELONA
  138. endif
  139. ifeq ($(TARGET_CORE), ZEN)
  140. GETARCH_FLAGS := -DFORCE_BARCELONA
  141. endif
  142. endif
  143. # On x86_64 build getarch with march=native unless the compiler is PGI. This is required to detect AVX512 support in getarch.
  144. ifeq ($(HOSTARCH), x86_64)
  145. ifeq ($(findstring pgcc,$(HOSTCC)),)
  146. GETARCH_FLAGS += -march=native
  147. endif
  148. endif
  149. ifdef INTERFACE64
  150. ifneq ($(INTERFACE64), 0)
  151. GETARCH_FLAGS += -DUSE64BITINT
  152. endif
  153. endif
  154. ifndef GEMM_MULTITHREAD_THRESHOLD
  155. GEMM_MULTITHREAD_THRESHOLD=4
  156. endif
  157. GETARCH_FLAGS += -DGEMM_MULTITHREAD_THRESHOLD=$(GEMM_MULTITHREAD_THRESHOLD)
  158. ifeq ($(NO_AVX), 1)
  159. GETARCH_FLAGS += -DNO_AVX
  160. endif
  161. ifeq ($(BINARY), 32)
  162. GETARCH_FLAGS += -DNO_AVX -DNO_AVX2 -DNO_AVX512
  163. NO_AVX512 = 1
  164. endif
  165. ifeq ($(NO_AVX2), 1)
  166. GETARCH_FLAGS += -DNO_AVX2
  167. endif
  168. ifeq ($(NO_AVX512), 1)
  169. GETARCH_FLAGS += -DNO_AVX512
  170. endif
  171. ifeq ($(DEBUG), 1)
  172. GETARCH_FLAGS += -g
  173. endif
  174. ifeq ($(QUIET_MAKE), 1)
  175. MAKE += -s
  176. endif
  177. ifndef NO_PARALLEL_MAKE
  178. NO_PARALLEL_MAKE=0
  179. endif
  180. GETARCH_FLAGS += -DNO_PARALLEL_MAKE=$(NO_PARALLEL_MAKE)
  181. ifdef MAKE_NB_JOBS
  182. GETARCH_FLAGS += -DMAKE_NB_JOBS=$(MAKE_NB_JOBS)
  183. endif
  184. ifeq ($(HOSTCC), loongcc)
  185. GETARCH_FLAGS += -static
  186. endif
  187. #if don't use Fortran, it will only compile CBLAS.
  188. ifeq ($(ONLY_CBLAS), 1)
  189. NO_LAPACK = 1
  190. else
  191. ONLY_CBLAS = 0
  192. endif
  193. # This operation is expensive, so execution should be once.
  194. ifndef GOTOBLAS_MAKEFILE
  195. export GOTOBLAS_MAKEFILE = 1
  196. # Generating Makefile.conf and config.h
  197. 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)
  198. ifndef TARGET_CORE
  199. include $(TOPDIR)/Makefile.conf
  200. else
  201. include $(TOPDIR)/Makefile_kernel.conf
  202. endif
  203. endif
  204. ifndef NUM_PARALLEL
  205. NUM_PARALLEL = 1
  206. endif
  207. ifndef NUM_THREADS
  208. NUM_THREADS = $(NUM_CORES)
  209. endif
  210. ifeq ($(NUM_THREADS), 1)
  211. override USE_THREAD = 0
  212. override USE_OPENMP = 0
  213. endif
  214. ifdef USE_THREAD
  215. ifeq ($(USE_THREAD), 0)
  216. SMP =
  217. else
  218. SMP = 1
  219. endif
  220. else
  221. ifeq ($(NUM_THREAD), 1)
  222. SMP =
  223. else
  224. SMP = 1
  225. endif
  226. endif
  227. ifeq ($(SMP), 1)
  228. USE_LOCKING =
  229. endif
  230. ifndef NEED_PIC
  231. NEED_PIC = 1
  232. endif
  233. ARFLAGS =
  234. CPP = $(COMPILER) -E
  235. AR ?= $(CROSS_SUFFIX)ar
  236. AS ?= $(CROSS_SUFFIX)as
  237. LD ?= $(CROSS_SUFFIX)ld
  238. RANLIB ?= $(CROSS_SUFFIX)ranlib
  239. NM = $(CROSS_SUFFIX)nm
  240. DLLWRAP = $(CROSS_SUFFIX)dllwrap
  241. OBJCOPY = $(CROSS_SUFFIX)objcopy
  242. OBJCONV = $(CROSS_SUFFIX)objconv
  243. # When fortran support was either not detected or actively deselected, only build BLAS.
  244. ifeq ($(NOFORTRAN), 1)
  245. NO_LAPACK = 1
  246. override FEXTRALIB =
  247. endif
  248. ifeq ($(C_COMPILER), GCC)
  249. GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
  250. GCCVERSIONGT4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 4)
  251. GCCVERSIONEQ5 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` = 5)
  252. GCCVERSIONGT5 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 5)
  253. GCCVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 7)
  254. GCCVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  255. GCCVERSIONGTEQ11 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 11)
  256. GCCVERSIONGTEQ10 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 10)
  257. # Note that the behavior of -dumpversion is compile-time-configurable for
  258. # gcc-7.x and newer. Use -dumpfullversion there
  259. ifeq ($(GCCVERSIONGTEQ7),1)
  260. GCCDUMPVERSION_PARAM := -dumpfullversion
  261. else
  262. GCCDUMPVERSION_PARAM := -dumpversion
  263. endif
  264. GCCMINORVERSIONGTEQ2 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 2)
  265. GCCMINORVERSIONGTEQ7 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 7)
  266. endif
  267. #
  268. # OS dependent settings
  269. #
  270. ifeq ($(OSNAME), Darwin)
  271. ifndef MACOSX_DEPLOYMENT_TARGET
  272. export MACOSX_DEPLOYMENT_TARGET=10.8
  273. endif
  274. MD5SUM = md5 -r
  275. endif
  276. ifneq (,$(findstring $(OSNAME), FreeBSD OpenBSD DragonFly))
  277. MD5SUM = md5 -r
  278. endif
  279. ifeq ($(OSNAME), NetBSD)
  280. MD5SUM = md5 -n
  281. endif
  282. ifeq ($(OSNAME), Linux)
  283. EXTRALIB += -lm
  284. NO_EXPRECISION = 1
  285. endif
  286. ifeq ($(OSNAME), Android)
  287. EXTRALIB += -lm
  288. endif
  289. ifeq ($(OSNAME), AIX)
  290. EXTRALIB += -lm
  291. endif
  292. ifeq ($(OSNAME), WINNT)
  293. NEED_PIC = 0
  294. NO_EXPRECISION = 1
  295. EXTRALIB += -defaultlib:advapi32
  296. SUFFIX = obj
  297. PSUFFIX = pobj
  298. LIBSUFFIX = a
  299. ifeq ($(C_COMPILER), CLANG)
  300. CCOMMON_OPT += -DMS_ABI
  301. endif
  302. #Version tests for supporting specific features (MS_ABI, POWER9 intrinsics)
  303. ifeq ($(GCCVERSIONGT4), 1)
  304. # GCC Major version > 4
  305. # It is compatible with MSVC ABI.
  306. CCOMMON_OPT += -DMS_ABI
  307. endif
  308. ifeq ($(GCCVERSIONGTEQ4), 1)
  309. ifeq ($(GCCMINORVERSIONGTEQ7), 1)
  310. # GCC Version >=4.7
  311. # It is compatible with MSVC ABI.
  312. CCOMMON_OPT += -DMS_ABI
  313. endif
  314. endif
  315. # Ensure the correct stack alignment on Win32
  316. # http://permalink.gmane.org/gmane.comp.lib.openblas.general/97
  317. ifeq ($(ARCH), x86)
  318. CCOMMON_OPT += -mincoming-stack-boundary=2
  319. FCOMMON_OPT += -mincoming-stack-boundary=2
  320. endif
  321. endif
  322. ifeq ($(OSNAME), Interix)
  323. NEED_PIC = 0
  324. NO_EXPRECISION = 1
  325. INTERIX_TOOL_DIR = /opt/gcc.3.3/i586-pc-interix3/bin
  326. endif
  327. ifeq ($(OSNAME), CYGWIN_NT)
  328. NEED_PIC = 0
  329. NO_EXPRECISION = 1
  330. OS_CYGWIN_NT = 1
  331. endif
  332. ifneq ($(OSNAME), WINNT)
  333. ifneq ($(OSNAME), CYGWIN_NT)
  334. ifneq ($(OSNAME), Interix)
  335. ifneq ($(OSNAME), Android)
  336. ifdef SMP
  337. EXTRALIB += -lpthread
  338. endif
  339. endif
  340. endif
  341. endif
  342. endif
  343. # ifeq logical or
  344. ifeq ($(OSNAME), $(filter $(OSNAME),WINNT CYGWIN_NT Interix))
  345. OS_WINDOWS=1
  346. endif
  347. ifdef QUAD_PRECISION
  348. CCOMMON_OPT += -DQUAD_PRECISION
  349. NO_EXPRECISION = 1
  350. endif
  351. ifneq ($(ARCH), x86)
  352. ifneq ($(ARCH), x86_64)
  353. NO_EXPRECISION = 1
  354. endif
  355. endif
  356. ifdef UTEST_CHECK
  357. CCOMMON_OPT += -DUTEST_CHECK
  358. SANITY_CHECK = 1
  359. endif
  360. ifdef SANITY_CHECK
  361. CCOMMON_OPT += -DSANITY_CHECK -DREFNAME=$(*F)f$(BU)
  362. endif
  363. MAX_STACK_ALLOC ?= 2048
  364. ifneq ($(MAX_STACK_ALLOC), 0)
  365. CCOMMON_OPT += -DMAX_STACK_ALLOC=$(MAX_STACK_ALLOC)
  366. endif
  367. ifdef USE_LOCKING
  368. ifneq ($(USE_LOCKING), 0)
  369. CCOMMON_OPT += -DUSE_LOCKING
  370. endif
  371. endif
  372. #
  373. # Architecture dependent settings
  374. #
  375. ifeq ($(ARCH), x86)
  376. ifndef BINARY
  377. NO_BINARY_MODE = 1
  378. endif
  379. ifeq ($(CORE), generic)
  380. NO_EXPRECISION = 1
  381. endif
  382. ifndef NO_EXPRECISION
  383. ifeq ($(F_COMPILER), GFORTRAN)
  384. # ifeq logical or. GCC or LSB
  385. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  386. EXPRECISION = 1
  387. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  388. FCOMMON_OPT += -m128bit-long-double
  389. endif
  390. ifeq ($(C_COMPILER), CLANG)
  391. EXPRECISION = 1
  392. CCOMMON_OPT += -DEXPRECISION
  393. FCOMMON_OPT += -m128bit-long-double
  394. endif
  395. endif
  396. endif
  397. endif
  398. ifeq ($(ARCH), x86_64)
  399. ifeq ($(CORE), generic)
  400. NO_EXPRECISION = 1
  401. endif
  402. ifndef NO_EXPRECISION
  403. ifeq ($(F_COMPILER), GFORTRAN)
  404. # ifeq logical or. GCC or LSB
  405. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  406. EXPRECISION = 1
  407. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  408. FCOMMON_OPT += -m128bit-long-double
  409. endif
  410. ifeq ($(C_COMPILER), CLANG)
  411. EXPRECISION = 1
  412. CCOMMON_OPT += -DEXPRECISION
  413. FCOMMON_OPT += -m128bit-long-double
  414. endif
  415. endif
  416. endif
  417. endif
  418. ifeq ($(C_COMPILER), INTEL)
  419. CCOMMON_OPT += -wd981
  420. endif
  421. ifeq ($(USE_OPENMP), 1)
  422. #check
  423. ifeq ($(USE_THREAD), 0)
  424. $(error OpenBLAS: Cannot set both USE_OPENMP=1 and USE_THREAD=0. The USE_THREAD=0 is only for building single thread version.)
  425. endif
  426. # ifeq logical or. GCC or LSB
  427. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  428. CCOMMON_OPT += -fopenmp
  429. endif
  430. ifeq ($(C_COMPILER), CLANG)
  431. CCOMMON_OPT += -fopenmp
  432. endif
  433. ifeq ($(C_COMPILER), INTEL)
  434. CCOMMON_OPT += -fopenmp
  435. endif
  436. ifeq ($(C_COMPILER), PGI)
  437. CCOMMON_OPT += -mp
  438. endif
  439. ifeq ($(C_COMPILER), OPEN64)
  440. CCOMMON_OPT += -mp
  441. CEXTRALIB += -lstdc++
  442. endif
  443. ifeq ($(C_COMPILER), PATHSCALE)
  444. CCOMMON_OPT += -mp
  445. endif
  446. endif
  447. ifeq ($(DYNAMIC_ARCH), 1)
  448. ifeq ($(ARCH), x86)
  449. DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
  450. CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
  451. endif
  452. ifeq ($(ARCH), x86_64)
  453. DYNAMIC_CORE = PRESCOTT CORE2
  454. ifeq ($(DYNAMIC_OLDER), 1)
  455. DYNAMIC_CORE += PENRYN DUNNINGTON
  456. endif
  457. DYNAMIC_CORE += NEHALEM
  458. ifeq ($(DYNAMIC_OLDER), 1)
  459. DYNAMIC_CORE += OPTERON OPTERON_SSE3
  460. endif
  461. DYNAMIC_CORE += BARCELONA
  462. ifeq ($(DYNAMIC_OLDER), 1)
  463. DYNAMIC_CORE += BOBCAT ATOM NANO
  464. endif
  465. ifneq ($(NO_AVX), 1)
  466. DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER STEAMROLLER EXCAVATOR
  467. endif
  468. ifneq ($(NO_AVX2), 1)
  469. DYNAMIC_CORE += HASWELL ZEN
  470. endif
  471. ifneq ($(NO_AVX512), 1)
  472. ifneq ($(NO_AVX2), 1)
  473. DYNAMIC_CORE += SKYLAKEX COOPERLAKE
  474. endif
  475. endif
  476. endif
  477. ifdef DYNAMIC_LIST
  478. override DYNAMIC_CORE = PRESCOTT $(DYNAMIC_LIST)
  479. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_PRESCOTT
  480. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  481. CCOMMON_OPT += $(XCCOMMON_OPT)
  482. #CCOMMON_OPT += -DDYNAMIC_LIST='$(DYNAMIC_LIST)'
  483. endif
  484. ifeq ($(ARCH), arm64)
  485. DYNAMIC_CORE = ARMV8
  486. DYNAMIC_CORE += CORTEXA53
  487. DYNAMIC_CORE += CORTEXA57
  488. DYNAMIC_CORE += CORTEXA72
  489. DYNAMIC_CORE += CORTEXA73
  490. DYNAMIC_CORE += NEOVERSEN1
  491. DYNAMIC_CORE += FALKOR
  492. DYNAMIC_CORE += THUNDERX
  493. DYNAMIC_CORE += THUNDERX2T99
  494. DYNAMIC_CORE += TSV110
  495. DYNAMIC_CORE += EMAG8180
  496. DYNAMIC_CORE += THUNDERX3T110
  497. endif
  498. ifeq ($(ARCH), zarch)
  499. DYNAMIC_CORE = ZARCH_GENERIC
  500. # Z13 is supported since gcc-5.2, gcc-6, and in RHEL 7.3 and newer
  501. ifeq ($(GCCVERSIONGT5), 1)
  502. ZARCH_SUPPORT_Z13 := 1
  503. else ifeq ($(GCCVERSIONEQ5), 1)
  504. ifeq ($(GCCMINORVERSIONGTEQ2), 1)
  505. ZARCH_SUPPORT_Z13 := 1
  506. endif
  507. endif
  508. ifeq ($(wildcard /etc/redhat-release), /etc/redhat-release)
  509. ifeq ($(shell source /etc/os-release ; expr $$VERSION_ID \>= "7.3"), 1)
  510. ZARCH_SUPPORT_Z13 := 1
  511. endif
  512. endif
  513. ifeq ($(ZARCH_SUPPORT_Z13), 1)
  514. DYNAMIC_CORE += Z13
  515. else
  516. $(info OpenBLAS: Not building Z13 kernels because gcc is older than 5.2 or 6.x)
  517. endif
  518. ifeq ($(GCCVERSIONGTEQ7), 1)
  519. DYNAMIC_CORE += Z14
  520. else
  521. $(info OpenBLAS: Not building Z14 kernels because gcc is older than 7.x)
  522. endif
  523. endif
  524. ifeq ($(ARCH), power)
  525. DYNAMIC_CORE = POWER6
  526. DYNAMIC_CORE += POWER8
  527. ifneq ($(C_COMPILER), GCC)
  528. DYNAMIC_CORE += POWER9
  529. DYNAMIC_CORE += POWER10
  530. endif
  531. ifeq ($(C_COMPILER), GCC)
  532. ifeq ($(GCCVERSIONGT5), 1)
  533. DYNAMIC_CORE += POWER9
  534. else
  535. $(info, OpenBLAS: Your gcc version is too old to build the POWER9 kernels.)
  536. endif
  537. ifeq ($(GCCVERSIONGTEQ11), 1)
  538. DYNAMIC_CORE += POWER10
  539. else ifeq ($(GCCVERSIONGTEQ10), 1)
  540. ifeq ($(GCCMINORVERSIONGTEQ2), 1)
  541. DYNAMIC_CORE += POWER10
  542. endif
  543. else
  544. $(info, OpenBLAS: Your gcc version is too old to build the POWER10 kernels.)
  545. endif
  546. endif
  547. endif
  548. # If DYNAMIC_CORE is not set, DYNAMIC_ARCH cannot do anything, so force it to empty
  549. ifndef DYNAMIC_CORE
  550. override DYNAMIC_ARCH=
  551. endif
  552. endif
  553. ifeq ($(ARCH), ia64)
  554. NO_BINARY_MODE = 1
  555. BINARY_DEFINED = 1
  556. ifeq ($(F_COMPILER), GFORTRAN)
  557. ifeq ($(C_COMPILER), GCC)
  558. # EXPRECISION = 1
  559. # CCOMMON_OPT += -DEXPRECISION
  560. endif
  561. endif
  562. endif
  563. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  564. NO_BINARY_MODE = 1
  565. endif
  566. ifeq ($(ARCH), alpha)
  567. NO_BINARY_MODE = 1
  568. BINARY_DEFINED = 1
  569. endif
  570. ifeq ($(ARCH), arm)
  571. NO_BINARY_MODE = 1
  572. BINARY_DEFINED = 1
  573. CCOMMON_OPT += -marm
  574. FCOMMON_OPT += -marm
  575. # If softfp abi is mentioned on the command line, force it.
  576. ifeq ($(ARM_SOFTFP_ABI), 1)
  577. CCOMMON_OPT += -mfloat-abi=softfp
  578. FCOMMON_OPT += -mfloat-abi=softfp
  579. endif
  580. ifeq ($(OSNAME), Android)
  581. ifeq ($(ARM_SOFTFP_ABI), 1)
  582. EXTRALIB += -lm
  583. else
  584. EXTRALIB += -Wl,-lm_hard
  585. endif
  586. endif
  587. endif
  588. ifeq ($(ARCH), arm64)
  589. NO_BINARY_MODE = 1
  590. BINARY_DEFINED = 1
  591. ifdef INTERFACE64
  592. ifneq ($(INTERFACE64), 0)
  593. ifeq ($(F_COMPILER), GFORTRAN)
  594. FCOMMON_OPT += -fdefault-integer-8
  595. endif
  596. ifeq ($(F_COMPILER), FLANG)
  597. FCOMMON_OPT += -i8
  598. endif
  599. endif
  600. endif
  601. endif
  602. #
  603. # C Compiler dependent settings
  604. #
  605. # ifeq logical or. GCC or CLANG or LSB
  606. # http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
  607. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG LSB))
  608. CCOMMON_OPT += -Wall
  609. COMMON_PROF += -fno-inline
  610. NO_UNINITIALIZED_WARN = -Wno-uninitialized
  611. ifeq ($(QUIET_MAKE), 1)
  612. CCOMMON_OPT += $(NO_UNINITIALIZED_WARN) -Wno-unused
  613. endif
  614. ifdef NO_BINARY_MODE
  615. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  616. ifdef BINARY64
  617. CCOMMON_OPT += -mabi=64
  618. else
  619. CCOMMON_OPT += -mabi=n32
  620. endif
  621. BINARY_DEFINED = 1
  622. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  623. CCOMMON_OPT += -mabi=32
  624. BINARY_DEFINED = 1
  625. endif
  626. ifeq ($(CORE), LOONGSON3A)
  627. CCOMMON_OPT += -march=mips64
  628. FCOMMON_OPT += -march=mips64
  629. endif
  630. ifeq ($(CORE), LOONGSON3B)
  631. CCOMMON_OPT += -march=mips64
  632. FCOMMON_OPT += -march=mips64
  633. endif
  634. ifeq ($(CORE), MIPS24K)
  635. CCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  636. FCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  637. endif
  638. ifeq ($(CORE), MIPS1004K)
  639. CCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  640. FCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  641. endif
  642. ifeq ($(CORE), P5600)
  643. CCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  644. FCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  645. endif
  646. ifeq ($(CORE), I6400)
  647. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  648. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  649. endif
  650. ifeq ($(CORE), P6600)
  651. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  652. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  653. endif
  654. ifeq ($(CORE), I6500)
  655. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  656. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  657. endif
  658. ifeq ($(OSNAME), AIX)
  659. BINARY_DEFINED = 1
  660. endif
  661. endif
  662. ifndef BINARY_DEFINED
  663. ifneq ($(OSNAME), AIX)
  664. ifdef BINARY64
  665. CCOMMON_OPT += -m64
  666. else
  667. CCOMMON_OPT += -m32
  668. endif
  669. endif
  670. endif
  671. endif
  672. ifeq ($(C_COMPILER), PGI)
  673. ifdef BINARY64
  674. ifeq ($(ARCH), x86_64)
  675. CCOMMON_OPT += -tp p7-64 -D__MMX__ -Mnollvm
  676. else
  677. ifeq ($(ARCH), power)
  678. ifeq ($(CORE), POWER8)
  679. CCOMMON_OPT += -tp pwr8
  680. endif
  681. ifeq ($(CORE), POWER9)
  682. CCOMMON_OPT += -tp pwr9
  683. endif
  684. endif
  685. endif
  686. else
  687. CCOMMON_OPT += -tp p7
  688. endif
  689. endif
  690. ifeq ($(C_COMPILER), PATHSCALE)
  691. ifdef BINARY64
  692. CCOMMON_OPT += -m64
  693. else
  694. CCOMMON_OPT += -m32
  695. endif
  696. endif
  697. #
  698. # Fortran Compiler dependent settings
  699. #
  700. ifeq ($(F_COMPILER), FLANG)
  701. CCOMMON_OPT += -DF_INTERFACE_FLANG
  702. FCOMMON_OPT += -Mrecursive -Kieee
  703. ifeq ($(OSNAME), Linux)
  704. ifeq ($(ARCH), x86_64)
  705. FLANG_VENDOR := $(shell expr `$(FC) --version|cut -f 1 -d "."|head -1`)
  706. ifeq ($(FLANG_VENDOR),AOCC)
  707. FCOMMON_OPT += -fno-unroll-loops
  708. endif
  709. endif
  710. endif
  711. ifdef BINARY64
  712. ifdef INTERFACE64
  713. ifneq ($(INTERFACE64), 0)
  714. FCOMMON_OPT += -i8
  715. endif
  716. endif
  717. FCOMMON_OPT += -Wall
  718. else
  719. FCOMMON_OPT += -Wall
  720. endif
  721. ifeq ($(USE_OPENMP), 1)
  722. FCOMMON_OPT += -fopenmp
  723. endif
  724. endif
  725. ifeq ($(F_COMPILER), G77)
  726. CCOMMON_OPT += -DF_INTERFACE_G77
  727. FCOMMON_OPT += -Wall
  728. ifndef NO_BINARY_MODE
  729. ifneq ($(OSNAME), AIX)
  730. ifdef BINARY64
  731. FCOMMON_OPT += -m64
  732. else
  733. FCOMMON_OPT += -m32
  734. endif
  735. endif
  736. endif
  737. endif
  738. ifeq ($(F_COMPILER), G95)
  739. CCOMMON_OPT += -DF_INTERFACE_G95
  740. FCOMMON_OPT += -Wall
  741. ifneq ($(OSNAME), AIX)
  742. ifndef NO_BINARY_MODE
  743. ifdef BINARY64
  744. FCOMMON_OPT += -m64
  745. else
  746. FCOMMON_OPT += -m32
  747. endif
  748. endif
  749. ifneq ($(NO_LAPACKE), 1)
  750. FCOMMON_OPT += -fno-second-underscore
  751. endif
  752. endif
  753. endif
  754. ifeq ($(F_COMPILER), GFORTRAN)
  755. CCOMMON_OPT += -DF_INTERFACE_GFORT
  756. FCOMMON_OPT += -Wall
  757. # make single-threaded LAPACK calls thread-safe #1847
  758. FCOMMON_OPT += -frecursive
  759. # work around ABI problem with passing single-character arguments
  760. FCOMMON_OPT += -fno-optimize-sibling-calls
  761. #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
  762. ifneq ($(NO_LAPACK), 1)
  763. EXTRALIB += -lgfortran
  764. endif
  765. ifdef NO_BINARY_MODE
  766. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  767. ifdef BINARY64
  768. FCOMMON_OPT += -mabi=64
  769. else
  770. FCOMMON_OPT += -mabi=n32
  771. endif
  772. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  773. FCOMMON_OPT += -mabi=32
  774. endif
  775. else
  776. ifdef BINARY64
  777. ifneq ($(OSNAME), AIX)
  778. FCOMMON_OPT += -m64
  779. endif
  780. ifdef INTERFACE64
  781. ifneq ($(INTERFACE64), 0)
  782. FCOMMON_OPT += -fdefault-integer-8
  783. endif
  784. endif
  785. else
  786. ifneq ($(OSNAME), AIX)
  787. FCOMMON_OPT += -m32
  788. endif
  789. endif
  790. endif
  791. ifeq ($(USE_OPENMP), 1)
  792. FCOMMON_OPT += -fopenmp
  793. endif
  794. endif
  795. ifeq ($(F_COMPILER), INTEL)
  796. CCOMMON_OPT += -DF_INTERFACE_INTEL
  797. ifdef INTERFACE64
  798. ifneq ($(INTERFACE64), 0)
  799. FCOMMON_OPT += -i8
  800. endif
  801. endif
  802. FCOMMON_OPT += -recursive -fp-model strict -assume protect-parens
  803. ifeq ($(USE_OPENMP), 1)
  804. FCOMMON_OPT += -fopenmp
  805. endif
  806. endif
  807. ifeq ($(F_COMPILER), FUJITSU)
  808. CCOMMON_OPT += -DF_INTERFACE_FUJITSU
  809. ifeq ($(USE_OPENMP), 1)
  810. FCOMMON_OPT += -openmp
  811. endif
  812. endif
  813. ifeq ($(F_COMPILER), IBM)
  814. CCOMMON_OPT += -DF_INTERFACE_IBM
  815. # FCOMMON_OPT += -qarch=440
  816. ifdef BINARY64
  817. FCOMMON_OPT += -q64
  818. ifdef INTERFACE64
  819. ifneq ($(INTERFACE64), 0)
  820. FCOMMON_OPT += -qintsize=8
  821. endif
  822. endif
  823. else
  824. FCOMMON_OPT += -q32
  825. endif
  826. ifeq ($(USE_OPENMP), 1)
  827. FCOMMON_OPT += -openmp
  828. endif
  829. endif
  830. ifeq ($(F_COMPILER), PGI)
  831. CCOMMON_OPT += -DF_INTERFACE_PGI
  832. COMMON_PROF += -DPGICOMPILER
  833. ifdef BINARY64
  834. ifdef INTERFACE64
  835. ifneq ($(INTERFACE64), 0)
  836. FCOMMON_OPT += -i8
  837. endif
  838. endif
  839. ifeq ($(ARCH), x86_64)
  840. FCOMMON_OPT += -tp p7-64
  841. else
  842. ifeq ($(ARCH), power)
  843. ifeq ($(CORE), POWER8)
  844. FCOMMON_OPT += -tp pwr8
  845. endif
  846. ifeq ($(CORE), POWER9)
  847. FCOMMON_OPT += -tp pwr9
  848. endif
  849. endif
  850. endif
  851. else
  852. FCOMMON_OPT += -tp p7
  853. endif
  854. FCOMMON_OPT += -Mrecursive
  855. ifeq ($(USE_OPENMP), 1)
  856. FCOMMON_OPT += -mp
  857. endif
  858. endif
  859. ifeq ($(F_COMPILER), PATHSCALE)
  860. CCOMMON_OPT += -DF_INTERFACE_PATHSCALE
  861. ifdef BINARY64
  862. ifdef INTERFACE64
  863. ifneq ($(INTERFACE64), 0)
  864. FCOMMON_OPT += -i8
  865. endif
  866. endif
  867. endif
  868. ifeq ($(USE_OPENMP), 1)
  869. FCOMMON_OPT += -mp
  870. endif
  871. endif
  872. ifeq ($(F_COMPILER), OPEN64)
  873. CCOMMON_OPT += -DF_INTERFACE_OPEN64
  874. ifdef BINARY64
  875. ifdef INTERFACE64
  876. ifneq ($(INTERFACE64), 0)
  877. FCOMMON_OPT += -i8
  878. endif
  879. endif
  880. endif
  881. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  882. ifndef BINARY64
  883. FCOMMON_OPT += -n32
  884. else
  885. FCOMMON_OPT += -n64
  886. endif
  887. ifeq ($(CORE), LOONGSON3A)
  888. FCOMMON_OPT += -loongson3 -static
  889. endif
  890. ifeq ($(CORE), LOONGSON3B)
  891. FCOMMON_OPT += -loongson3 -static
  892. endif
  893. else
  894. ifndef BINARY64
  895. FCOMMON_OPT += -m32
  896. else
  897. FCOMMON_OPT += -m64
  898. endif
  899. endif
  900. ifeq ($(USE_OPENMP), 1)
  901. FEXTRALIB += -lstdc++
  902. FCOMMON_OPT += -mp
  903. endif
  904. endif
  905. ifeq ($(C_COMPILER), OPEN64)
  906. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  907. ifndef BINARY64
  908. CCOMMON_OPT += -n32
  909. else
  910. CCOMMON_OPT += -n64
  911. endif
  912. ifeq ($(CORE), LOONGSON3A)
  913. CCOMMON_OPT += -loongson3 -static
  914. endif
  915. ifeq ($(CORE), LOONGSON3B)
  916. CCOMMON_OPT += -loongson3 -static
  917. endif
  918. else
  919. ifndef BINARY64
  920. CCOMMON_OPT += -m32
  921. else
  922. CCOMMON_OPT += -m64
  923. endif
  924. endif
  925. endif
  926. ifeq ($(C_COMPILER), SUN)
  927. CCOMMON_OPT += -w
  928. ifeq ($(ARCH), x86)
  929. CCOMMON_OPT += -m32
  930. else
  931. FCOMMON_OPT += -m64
  932. endif
  933. endif
  934. ifeq ($(F_COMPILER), SUN)
  935. CCOMMON_OPT += -DF_INTERFACE_SUN
  936. ifeq ($(ARCH), x86)
  937. FCOMMON_OPT += -m32
  938. else
  939. FCOMMON_OPT += -m64
  940. endif
  941. ifeq ($(USE_OPENMP), 1)
  942. FCOMMON_OPT += -xopenmp=parallel
  943. endif
  944. endif
  945. ifeq ($(F_COMPILER), COMPAQ)
  946. CCOMMON_OPT += -DF_INTERFACE_COMPAQ
  947. ifeq ($(USE_OPENMP), 1)
  948. FCOMMON_OPT += -openmp
  949. endif
  950. endif
  951. ifdef BINARY64
  952. ifdef INTERFACE64
  953. ifneq ($(INTERFACE64), 0)
  954. CCOMMON_OPT +=
  955. #-DUSE64BITINT
  956. endif
  957. endif
  958. endif
  959. ifeq ($(NEED_PIC), 1)
  960. ifeq ($(C_COMPILER), IBM)
  961. CCOMMON_OPT += -qpic=large
  962. else
  963. CCOMMON_OPT += -fPIC
  964. endif
  965. ifeq ($(F_COMPILER), SUN)
  966. FCOMMON_OPT += -pic
  967. else
  968. FCOMMON_OPT += -fPIC
  969. endif
  970. endif
  971. ifeq ($(DYNAMIC_ARCH), 1)
  972. CCOMMON_OPT += -DDYNAMIC_ARCH
  973. endif
  974. ifeq ($(DYNAMIC_OLDER), 1)
  975. CCOMMON_OPT += -DDYNAMIC_OLDER
  976. endif
  977. ifeq ($(NO_LAPACK), 1)
  978. CCOMMON_OPT += -DNO_LAPACK
  979. #Disable LAPACK C interface
  980. NO_LAPACKE = 1
  981. endif
  982. ifeq ($(NO_LAPACKE), 1)
  983. CCOMMON_OPT += -DNO_LAPACKE
  984. endif
  985. ifeq ($(NO_AVX), 1)
  986. CCOMMON_OPT += -DNO_AVX
  987. endif
  988. ifeq ($(ARCH), x86)
  989. CCOMMON_OPT += -DNO_AVX
  990. endif
  991. ifeq ($(NO_AVX2), 1)
  992. CCOMMON_OPT += -DNO_AVX2
  993. endif
  994. ifeq ($(NO_AVX512), 1)
  995. CCOMMON_OPT += -DNO_AVX512
  996. endif
  997. ifdef SMP
  998. CCOMMON_OPT += -DSMP_SERVER
  999. ifeq ($(ARCH), mips64)
  1000. ifneq ($(CORE), LOONGSON3B)
  1001. USE_SIMPLE_THREADED_LEVEL3 = 1
  1002. endif
  1003. endif
  1004. ifeq ($(USE_OPENMP), 1)
  1005. # USE_SIMPLE_THREADED_LEVEL3 = 1
  1006. # NO_AFFINITY = 1
  1007. CCOMMON_OPT += -DUSE_OPENMP
  1008. endif
  1009. ifeq ($(BIGNUMA), 1)
  1010. CCOMMON_OPT += -DBIGNUMA
  1011. endif
  1012. endif
  1013. ifeq ($(NO_WARMUP), 1)
  1014. CCOMMON_OPT += -DNO_WARMUP
  1015. endif
  1016. ifeq ($(CONSISTENT_FPCSR), 1)
  1017. CCOMMON_OPT += -DCONSISTENT_FPCSR
  1018. endif
  1019. # Only for development
  1020. # CCOMMON_OPT += -DPARAMTEST
  1021. # CCOMMON_OPT += -DPREFETCHTEST
  1022. # CCOMMON_OPT += -DNO_SWITCHING
  1023. # USE_PAPI = 1
  1024. ifdef USE_PAPI
  1025. CCOMMON_OPT += -DUSE_PAPI
  1026. EXTRALIB += -lpapi -lperfctr
  1027. endif
  1028. ifdef DYNAMIC_THREADS
  1029. CCOMMON_OPT += -DDYNAMIC_THREADS
  1030. endif
  1031. CCOMMON_OPT += -DMAX_CPU_NUMBER=$(NUM_THREADS)
  1032. CCOMMON_OPT += -DMAX_PARALLEL_NUMBER=$(NUM_PARALLEL)
  1033. ifdef USE_SIMPLE_THREADED_LEVEL3
  1034. CCOMMON_OPT += -DUSE_SIMPLE_THREADED_LEVEL3
  1035. endif
  1036. ifeq ($(USE_TLS), 1)
  1037. CCOMMON_OPT += -DUSE_TLS
  1038. endif
  1039. ifeq ($(BUILD_HALF), 1)
  1040. CCOMMON_OPT += -DBUILD_HALF
  1041. endif
  1042. CCOMMON_OPT += -DVERSION=\"$(VERSION)\"
  1043. ifndef SYMBOLPREFIX
  1044. SYMBOLPREFIX =
  1045. endif
  1046. ifndef SYMBOLSUFFIX
  1047. SYMBOLSUFFIX =
  1048. endif
  1049. ifndef LIBNAMESUFFIX
  1050. LIBNAMEBASE = $(SYMBOLPREFIX)openblas$(SYMBOLSUFFIX)
  1051. else
  1052. LIBNAMEBASE = $(SYMBOLPREFIX)openblas$(SYMBOLSUFFIX)_$(LIBNAMESUFFIX)
  1053. endif
  1054. ifeq ($(OSNAME), CYGWIN_NT)
  1055. LIBPREFIX = cyg$(LIBNAMEBASE)
  1056. else
  1057. LIBPREFIX = lib$(LIBNAMEBASE)
  1058. endif
  1059. KERNELDIR = $(TOPDIR)/kernel/$(ARCH)
  1060. include $(TOPDIR)/Makefile.$(ARCH)
  1061. ifneq ($(C_COMPILER), PGI)
  1062. CCOMMON_OPT += -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME
  1063. endif
  1064. CCOMMON_OPT += -DASMNAME=$(FU)$(*F) -DASMFNAME=$(FU)$(*F)$(BU) -DNAME=$(*F)$(BU) -DCNAME=$(*F) -DCHAR_NAME=\"$(*F)$(BU)\" -DCHAR_CNAME=\"$(*F)\"
  1065. ifeq ($(CORE), PPC440)
  1066. CCOMMON_OPT += -DALLOC_QALLOC
  1067. endif
  1068. ifeq ($(CORE), PPC440FP2)
  1069. STATIC_ALLOCATION = 1
  1070. endif
  1071. ifneq ($(OSNAME), Linux)
  1072. NO_AFFINITY = 1
  1073. endif
  1074. ifneq ($(ARCH), x86_64)
  1075. ifneq ($(ARCH), x86)
  1076. ifneq ($(CORE), LOONGSON3B)
  1077. NO_AFFINITY = 1
  1078. endif
  1079. endif
  1080. endif
  1081. ifdef NO_AFFINITY
  1082. ifeq ($(NO_AFFINITY), 0)
  1083. override undefine NO_AFFINITY
  1084. else
  1085. CCOMMON_OPT += -DNO_AFFINITY
  1086. endif
  1087. endif
  1088. ifdef FUNCTION_PROFILE
  1089. CCOMMON_OPT += -DFUNCTION_PROFILE
  1090. endif
  1091. ifdef HUGETLB_ALLOCATION
  1092. CCOMMON_OPT += -DALLOC_HUGETLB
  1093. endif
  1094. ifdef HUGETLBFILE_ALLOCATION
  1095. CCOMMON_OPT += -DALLOC_HUGETLBFILE -DHUGETLB_FILE_NAME=$(HUGETLBFILE_ALLOCATION)
  1096. endif
  1097. ifdef STATIC_ALLOCATION
  1098. CCOMMON_OPT += -DALLOC_STATIC
  1099. endif
  1100. ifdef DEVICEDRIVER_ALLOCATION
  1101. CCOMMON_OPT += -DALLOC_DEVICEDRIVER -DDEVICEDRIVER_NAME=\"/dev/mapper\"
  1102. endif
  1103. ifdef MIXED_MEMORY_ALLOCATION
  1104. CCOMMON_OPT += -DMIXED_MEMORY_ALLOCATION
  1105. endif
  1106. ifeq ($(OSNAME), SunOS)
  1107. TAR = gtar
  1108. PATCH = gpatch
  1109. GREP = ggrep
  1110. AWK = nawk
  1111. else
  1112. TAR = tar
  1113. PATCH = patch
  1114. GREP = grep
  1115. AWK = awk
  1116. endif
  1117. ifndef MD5SUM
  1118. MD5SUM = md5sum
  1119. endif
  1120. REVISION = -r$(VERSION)
  1121. MAJOR_VERSION = $(word 1,$(subst ., ,$(VERSION)))
  1122. ifeq ($(DEBUG), 1)
  1123. COMMON_OPT += -g
  1124. endif
  1125. ifeq ($(DEBUG), 1)
  1126. FCOMMON_OPT += -g
  1127. endif
  1128. ifndef COMMON_OPT
  1129. COMMON_OPT = -O2
  1130. endif
  1131. ifndef FCOMMON_OPT
  1132. FCOMMON_OPT = -O2 -frecursive
  1133. endif
  1134. override CFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
  1135. override PFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
  1136. override FFLAGS += $(COMMON_OPT) $(FCOMMON_OPT)
  1137. override FPFLAGS += $(FCOMMON_OPT) $(COMMON_PROF)
  1138. #MAKEOVERRIDES =
  1139. ifdef NEED_PIC
  1140. ifeq (,$(findstring PIC,$(FFLAGS)))
  1141. override FFLAGS += -fPIC
  1142. endif
  1143. endif
  1144. #For LAPACK Fortran codes.
  1145. #Disable -fopenmp for LAPACK Fortran codes on Windows.
  1146. ifdef OS_WINDOWS
  1147. LAPACK_FFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FFLAGS))
  1148. LAPACK_FPFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FPFLAGS))
  1149. else
  1150. LAPACK_FFLAGS := $(FFLAGS)
  1151. LAPACK_FPFLAGS := $(FPFLAGS)
  1152. endif
  1153. LAPACK_CFLAGS = $(CFLAGS)
  1154. LAPACK_CFLAGS += -DHAVE_LAPACK_CONFIG_H
  1155. ifdef INTERFACE64
  1156. ifneq ($(INTERFACE64), 0)
  1157. LAPACK_CFLAGS += -DLAPACK_ILP64
  1158. endif
  1159. endif
  1160. ifdef OS_WINDOWS
  1161. LAPACK_CFLAGS += -DOPENBLAS_OS_WINDOWS
  1162. endif
  1163. ifeq ($(C_COMPILER), LSB)
  1164. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1165. endif
  1166. ifndef SUFFIX
  1167. SUFFIX = o
  1168. endif
  1169. ifndef PSUFFIX
  1170. PSUFFIX = po
  1171. endif
  1172. ifndef LIBSUFFIX
  1173. LIBSUFFIX = a
  1174. endif
  1175. ifneq ($(DYNAMIC_ARCH), 1)
  1176. ifndef SMP
  1177. LIBNAME = $(LIBPREFIX)_$(LIBCORE)$(REVISION).$(LIBSUFFIX)
  1178. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)$(REVISION)_p.$(LIBSUFFIX)
  1179. else
  1180. LIBNAME = $(LIBPREFIX)_$(LIBCORE)p$(REVISION).$(LIBSUFFIX)
  1181. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)p$(REVISION)_p.$(LIBSUFFIX)
  1182. endif
  1183. else
  1184. ifndef SMP
  1185. LIBNAME = $(LIBPREFIX)$(REVISION).$(LIBSUFFIX)
  1186. LIBNAME_P = $(LIBPREFIX)$(REVISION)_p.$(LIBSUFFIX)
  1187. else
  1188. LIBNAME = $(LIBPREFIX)p$(REVISION).$(LIBSUFFIX)
  1189. LIBNAME_P = $(LIBPREFIX)p$(REVISION)_p.$(LIBSUFFIX)
  1190. endif
  1191. endif
  1192. LIBDLLNAME = $(LIBPREFIX).dll
  1193. IMPLIBNAME = lib$(LIBNAMEBASE).dll.a
  1194. ifneq ($(OSNAME), AIX)
  1195. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.so)
  1196. else
  1197. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.a)
  1198. endif
  1199. LIBDYNNAME = $(LIBNAME:.$(LIBSUFFIX)=.dylib)
  1200. LIBDEFNAME = $(LIBNAME:.$(LIBSUFFIX)=.def)
  1201. LIBEXPNAME = $(LIBNAME:.$(LIBSUFFIX)=.exp)
  1202. LIBZIPNAME = $(LIBNAME:.$(LIBSUFFIX)=.zip)
  1203. LIBS = $(TOPDIR)/$(LIBNAME)
  1204. LIBS_P = $(TOPDIR)/$(LIBNAME_P)
  1205. LIB_COMPONENTS = BLAS
  1206. ifneq ($(NO_CBLAS), 1)
  1207. LIB_COMPONENTS += CBLAS
  1208. endif
  1209. ifneq ($(NO_LAPACK), 1)
  1210. LIB_COMPONENTS += LAPACK
  1211. ifneq ($(NO_LAPACKE), 1)
  1212. LIB_COMPONENTS += LAPACKE
  1213. endif
  1214. ifeq ($(BUILD_RELAPACK), 1)
  1215. LIB_COMPONENTS += ReLAPACK
  1216. endif
  1217. endif
  1218. ifeq ($(ONLY_CBLAS), 1)
  1219. LIB_COMPONENTS = CBLAS
  1220. endif
  1221. export OSNAME
  1222. export ARCH
  1223. export CORE
  1224. export LIBCORE
  1225. export __BYTE_ORDER__
  1226. export ELF_VERSION
  1227. export PGCPATH
  1228. export CONFIG
  1229. export CC
  1230. export FC
  1231. export BU
  1232. export FU
  1233. export NEED2UNDERSCORES
  1234. export USE_THREAD
  1235. export NUM_THREADS
  1236. export NUM_CORES
  1237. export SMP
  1238. export MAKEFILE_RULE
  1239. export NEED_PIC
  1240. export BINARY
  1241. export BINARY32
  1242. export BINARY64
  1243. export F_COMPILER
  1244. export C_COMPILER
  1245. export USE_OPENMP
  1246. export CROSS
  1247. export CROSS_SUFFIX
  1248. export NOFORTRAN
  1249. export NO_FBLAS
  1250. export EXTRALIB
  1251. export CEXTRALIB
  1252. export FEXTRALIB
  1253. export HAVE_SSE
  1254. export HAVE_SSE2
  1255. export HAVE_SSE3
  1256. export HAVE_SSSE3
  1257. export HAVE_SSE4_1
  1258. export HAVE_SSE4_2
  1259. export HAVE_SSE4A
  1260. export HAVE_SSE5
  1261. export HAVE_AVX
  1262. export HAVE_VFP
  1263. export HAVE_VFPV3
  1264. export HAVE_VFPV4
  1265. export HAVE_NEON
  1266. export HAVE_MSA
  1267. export MSA_FLAGS
  1268. export KERNELDIR
  1269. export FUNCTION_PROFILE
  1270. export TARGET_CORE
  1271. export NO_AVX512
  1272. export BUILD_HALF
  1273. export SHGEMM_UNROLL_M
  1274. export SHGEMM_UNROLL_N
  1275. export SGEMM_UNROLL_M
  1276. export SGEMM_UNROLL_N
  1277. export DGEMM_UNROLL_M
  1278. export DGEMM_UNROLL_N
  1279. export QGEMM_UNROLL_M
  1280. export QGEMM_UNROLL_N
  1281. export CGEMM_UNROLL_M
  1282. export CGEMM_UNROLL_N
  1283. export ZGEMM_UNROLL_M
  1284. export ZGEMM_UNROLL_N
  1285. export XGEMM_UNROLL_M
  1286. export XGEMM_UNROLL_N
  1287. export CGEMM3M_UNROLL_M
  1288. export CGEMM3M_UNROLL_N
  1289. export ZGEMM3M_UNROLL_M
  1290. export ZGEMM3M_UNROLL_N
  1291. export XGEMM3M_UNROLL_M
  1292. export XGEMM3M_UNROLL_N
  1293. ifdef USE_CUDA
  1294. export CUDADIR
  1295. export CUCC
  1296. export CUFLAGS
  1297. export CULIB
  1298. endif
  1299. .SUFFIXES: .$(PSUFFIX) .$(SUFFIX) .f
  1300. .f.$(SUFFIX):
  1301. $(FC) $(FFLAGS) -c $< -o $(@F)
  1302. .f.$(PSUFFIX):
  1303. $(FC) $(FPFLAGS) -pg -c $< -o $(@F)
  1304. ifdef BINARY64
  1305. PATHSCALEPATH = /opt/pathscale/lib/3.1
  1306. PGIPATH = /opt/pgi/linux86-64/7.1-5/lib
  1307. else
  1308. PATHSCALEPATH = /opt/pathscale/lib/3.1/32
  1309. PGIPATH = /opt/pgi/linux86/7.1-5/lib
  1310. endif
  1311. ACMLPATH = /opt/acml/4.3.0
  1312. ifneq ($(OSNAME), Darwin)
  1313. MKLPATH = /opt/intel/mkl/10.2.2.025/lib
  1314. else
  1315. MKLPATH = /Library/Frameworks/Intel_MKL.framework/Versions/10.0.1.014/lib
  1316. endif
  1317. ATLASPATH = /opt/atlas/3.9.17/opteron
  1318. FLAMEPATH = $(HOME)/flame/lib
  1319. ifneq ($(OSNAME), SunOS)
  1320. SUNPATH = /opt/sunstudio12.1
  1321. else
  1322. SUNPATH = /opt/SUNWspro
  1323. endif