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 31 kB

7 years ago
13 years ago
12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448144914501451145214531454145514561457145814591460146114621463146414651466146714681469147014711472147314741475147614771478147914801481148214831484148514861487148814891490149114921493149414951496149714981499150015011502150315041505150615071508150915101511151215131514151515161517151815191520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559156015611562156315641565156615671568156915701571157215731574157515761577157815791580158115821583158415851586158715881589
  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. ifneq ($(BUILD_SINGLE),1)
  66. ifneq ($(BUILD_DOUBLE),1)
  67. ifneq ($(BUILD_COMPLEX),1)
  68. ifneq ($(BUILD_COMPLEX16),1)
  69. override BUILD_SINGLE=1
  70. override BUILD_DOUBLE=1
  71. override BUILD_COMPLEX=1
  72. override BUILD_COMPLEX16=1
  73. endif
  74. endif
  75. endif
  76. endif
  77. ifndef HOSTCC
  78. HOSTCC = $(CC)
  79. endif
  80. ifdef TARGET
  81. GETARCH_FLAGS := -DFORCE_$(TARGET)
  82. GETARCH_FLAGS += -DUSER_TARGET
  83. endif
  84. # Force fallbacks for 32bit
  85. ifeq ($(BINARY), 32)
  86. ifeq ($(TARGET), HASWELL)
  87. GETARCH_FLAGS := -DFORCE_NEHALEM
  88. endif
  89. ifeq ($(TARGET), SKYLAKEX)
  90. GETARCH_FLAGS := -DFORCE_NEHALEM
  91. endif
  92. ifeq ($(TARGET), COOPERLAKE)
  93. GETARCH_FLAGS := -DFORCE_NEHALEM
  94. endif
  95. ifeq ($(TARGET), SANDYBRIDGE)
  96. GETARCH_FLAGS := -DFORCE_NEHALEM
  97. endif
  98. ifeq ($(TARGET), BULLDOZER)
  99. GETARCH_FLAGS := -DFORCE_BARCELONA
  100. endif
  101. ifeq ($(TARGET), PILEDRIVER)
  102. GETARCH_FLAGS := -DFORCE_BARCELONA
  103. endif
  104. ifeq ($(TARGET), STEAMROLLER)
  105. GETARCH_FLAGS := -DFORCE_BARCELONA
  106. endif
  107. ifeq ($(TARGET), EXCAVATOR)
  108. GETARCH_FLAGS := -DFORCE_BARCELONA
  109. endif
  110. ifeq ($(TARGET), ZEN)
  111. GETARCH_FLAGS := -DFORCE_BARCELONA
  112. endif
  113. ifeq ($(TARGET), ARMV8)
  114. GETARCH_FLAGS := -DFORCE_ARMV7
  115. endif
  116. ifeq ($(TARGET), POWER8)
  117. GETARCH_FLAGS := -DFORCE_POWER6
  118. endif
  119. endif
  120. #TARGET_CORE will override TARGET which is used in DYNAMIC_ARCH=1.
  121. #
  122. ifdef TARGET_CORE
  123. GETARCH_FLAGS := -DFORCE_$(TARGET_CORE)
  124. endif
  125. # Force fallbacks for 32bit
  126. ifeq ($(BINARY), 32)
  127. ifeq ($(TARGET_CORE), HASWELL)
  128. GETARCH_FLAGS := -DFORCE_NEHALEM
  129. endif
  130. ifeq ($(TARGET_CORE), SKYLAKEX)
  131. GETARCH_FLAGS := -DFORCE_NEHALEM
  132. endif
  133. ifeq ($(TARGET_CORE), COOPERLAKE)
  134. GETARCH_FLAGS := -DFORCE_NEHALEM
  135. endif
  136. ifeq ($(TARGET_CORE), SANDYBRIDGE)
  137. GETARCH_FLAGS := -DFORCE_NEHALEM
  138. endif
  139. ifeq ($(TARGET_CORE), BULLDOZER)
  140. GETARCH_FLAGS := -DFORCE_BARCELONA
  141. endif
  142. ifeq ($(TARGET_CORE), PILEDRIVER)
  143. GETARCH_FLAGS := -DFORCE_BARCELONA
  144. endif
  145. ifeq ($(TARGET_CORE), STEAMROLLER)
  146. GETARCH_FLAGS := -DFORCE_BARCELONA
  147. endif
  148. ifeq ($(TARGET_CORE), EXCAVATOR)
  149. GETARCH_FLAGS := -DFORCE_BARCELONA
  150. endif
  151. ifeq ($(TARGET_CORE), ZEN)
  152. GETARCH_FLAGS := -DFORCE_BARCELONA
  153. endif
  154. endif
  155. # On x86_64 build getarch with march=native unless the compiler is PGI. This is required to detect AVX512 support in getarch.
  156. ifeq ($(HOSTARCH), x86_64)
  157. ifeq ($(findstring pgcc,$(HOSTCC)),)
  158. GETARCH_FLAGS += -march=native
  159. endif
  160. endif
  161. ifdef INTERFACE64
  162. ifneq ($(INTERFACE64), 0)
  163. GETARCH_FLAGS += -DUSE64BITINT
  164. endif
  165. endif
  166. ifndef GEMM_MULTITHREAD_THRESHOLD
  167. GEMM_MULTITHREAD_THRESHOLD=4
  168. endif
  169. GETARCH_FLAGS += -DGEMM_MULTITHREAD_THRESHOLD=$(GEMM_MULTITHREAD_THRESHOLD)
  170. ifeq ($(NO_AVX), 1)
  171. GETARCH_FLAGS += -DNO_AVX
  172. endif
  173. ifeq ($(BINARY), 32)
  174. GETARCH_FLAGS += -DNO_AVX -DNO_AVX2 -DNO_AVX512
  175. NO_AVX512 = 1
  176. endif
  177. ifeq ($(NO_AVX2), 1)
  178. GETARCH_FLAGS += -DNO_AVX2
  179. endif
  180. ifeq ($(NO_AVX512), 1)
  181. GETARCH_FLAGS += -DNO_AVX512
  182. endif
  183. ifeq ($(DEBUG), 1)
  184. GETARCH_FLAGS += -g
  185. endif
  186. ifeq ($(QUIET_MAKE), 1)
  187. MAKE += -s
  188. endif
  189. ifndef NO_PARALLEL_MAKE
  190. NO_PARALLEL_MAKE=0
  191. endif
  192. GETARCH_FLAGS += -DNO_PARALLEL_MAKE=$(NO_PARALLEL_MAKE)
  193. ifdef MAKE_NB_JOBS
  194. GETARCH_FLAGS += -DMAKE_NB_JOBS=$(MAKE_NB_JOBS)
  195. endif
  196. ifeq ($(HOSTCC), loongcc)
  197. GETARCH_FLAGS += -static
  198. endif
  199. #if don't use Fortran, it will only compile CBLAS.
  200. ifeq ($(ONLY_CBLAS), 1)
  201. NO_LAPACK = 1
  202. else
  203. ONLY_CBLAS = 0
  204. endif
  205. # This operation is expensive, so execution should be once.
  206. ifndef GOTOBLAS_MAKEFILE
  207. export GOTOBLAS_MAKEFILE = 1
  208. # Generating Makefile.conf and config.h
  209. 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)
  210. ifndef TARGET_CORE
  211. include $(TOPDIR)/Makefile.conf
  212. else
  213. include $(TOPDIR)/Makefile_kernel.conf
  214. endif
  215. endif
  216. ifndef NUM_PARALLEL
  217. NUM_PARALLEL = 1
  218. endif
  219. ifndef NUM_THREADS
  220. NUM_THREADS = $(NUM_CORES)
  221. endif
  222. ifeq ($(NUM_THREADS), 1)
  223. override USE_THREAD = 0
  224. override USE_OPENMP = 0
  225. endif
  226. ifdef USE_THREAD
  227. ifeq ($(USE_THREAD), 0)
  228. SMP =
  229. else
  230. SMP = 1
  231. endif
  232. else
  233. ifeq ($(NUM_THREAD), 1)
  234. SMP =
  235. else
  236. SMP = 1
  237. endif
  238. endif
  239. ifeq ($(SMP), 1)
  240. USE_LOCKING =
  241. endif
  242. ifndef NEED_PIC
  243. NEED_PIC = 1
  244. endif
  245. ARFLAGS =
  246. CPP = $(COMPILER) -E
  247. AR ?= $(CROSS_SUFFIX)ar
  248. AS ?= $(CROSS_SUFFIX)as
  249. LD ?= $(CROSS_SUFFIX)ld
  250. RANLIB ?= $(CROSS_SUFFIX)ranlib
  251. NM = $(CROSS_SUFFIX)nm
  252. DLLWRAP = $(CROSS_SUFFIX)dllwrap
  253. OBJCOPY = $(CROSS_SUFFIX)objcopy
  254. OBJCONV = $(CROSS_SUFFIX)objconv
  255. # When fortran support was either not detected or actively deselected, only build BLAS.
  256. ifeq ($(NOFORTRAN), 1)
  257. NO_LAPACK = 1
  258. override FEXTRALIB =
  259. endif
  260. ifeq ($(C_COMPILER), GCC)
  261. GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
  262. GCCVERSIONGT4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 4)
  263. GCCVERSIONGT5 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 5)
  264. GCCVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 7)
  265. GCCVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  266. GCCVERSIONGTEQ11 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 11)
  267. GCCVERSIONGTEQ10 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 10)
  268. # Note that the behavior of -dumpversion is compile-time-configurable for
  269. # gcc-7.x and newer. Use -dumpfullversion there
  270. ifeq ($(GCCVERSIONGTEQ7),1)
  271. GCCDUMPVERSION_PARAM := -dumpfullversion
  272. else
  273. GCCDUMPVERSION_PARAM := -dumpversion
  274. endif
  275. GCCMINORVERSIONGTEQ1 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 1)
  276. GCCMINORVERSIONGTEQ2 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 2)
  277. GCCMINORVERSIONGTEQ7 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 7)
  278. endif
  279. #
  280. # OS dependent settings
  281. #
  282. ifeq ($(OSNAME), Darwin)
  283. ifndef MACOSX_DEPLOYMENT_TARGET
  284. export MACOSX_DEPLOYMENT_TARGET=10.8
  285. endif
  286. MD5SUM = md5 -r
  287. endif
  288. ifneq (,$(findstring $(OSNAME), FreeBSD OpenBSD DragonFly))
  289. MD5SUM = md5 -r
  290. endif
  291. ifeq ($(OSNAME), NetBSD)
  292. MD5SUM = md5 -n
  293. endif
  294. ifeq ($(OSNAME), Linux)
  295. EXTRALIB += -lm
  296. NO_EXPRECISION = 1
  297. endif
  298. ifeq ($(OSNAME), Android)
  299. EXTRALIB += -lm
  300. endif
  301. ifeq ($(OSNAME), AIX)
  302. EXTRALIB += -lm
  303. endif
  304. ifeq ($(OSNAME), WINNT)
  305. NEED_PIC = 0
  306. NO_EXPRECISION = 1
  307. EXTRALIB += -defaultlib:advapi32
  308. SUFFIX = obj
  309. PSUFFIX = pobj
  310. LIBSUFFIX = a
  311. ifeq ($(C_COMPILER), CLANG)
  312. CCOMMON_OPT += -DMS_ABI
  313. endif
  314. #Version tests for supporting specific features (MS_ABI, POWER9 intrinsics)
  315. ifeq ($(GCCVERSIONGT4), 1)
  316. # GCC Major version > 4
  317. # It is compatible with MSVC ABI.
  318. CCOMMON_OPT += -DMS_ABI
  319. endif
  320. ifeq ($(GCCVERSIONGTEQ4), 1)
  321. ifeq ($(GCCMINORVERSIONGTEQ7), 1)
  322. # GCC Version >=4.7
  323. # It is compatible with MSVC ABI.
  324. CCOMMON_OPT += -DMS_ABI
  325. endif
  326. endif
  327. # Ensure the correct stack alignment on Win32
  328. # http://permalink.gmane.org/gmane.comp.lib.openblas.general/97
  329. ifeq ($(ARCH), x86)
  330. CCOMMON_OPT += -mincoming-stack-boundary=2
  331. FCOMMON_OPT += -mincoming-stack-boundary=2
  332. endif
  333. endif
  334. ifeq ($(OSNAME), Interix)
  335. NEED_PIC = 0
  336. NO_EXPRECISION = 1
  337. INTERIX_TOOL_DIR = /opt/gcc.3.3/i586-pc-interix3/bin
  338. endif
  339. ifeq ($(OSNAME), CYGWIN_NT)
  340. NEED_PIC = 0
  341. NO_EXPRECISION = 1
  342. OS_CYGWIN_NT = 1
  343. endif
  344. ifneq ($(OSNAME), WINNT)
  345. ifneq ($(OSNAME), CYGWIN_NT)
  346. ifneq ($(OSNAME), Interix)
  347. ifneq ($(OSNAME), Android)
  348. ifdef SMP
  349. EXTRALIB += -lpthread
  350. endif
  351. endif
  352. endif
  353. endif
  354. endif
  355. # ifeq logical or
  356. ifeq ($(OSNAME), $(filter $(OSNAME),WINNT CYGWIN_NT Interix))
  357. OS_WINDOWS=1
  358. endif
  359. ifdef QUAD_PRECISION
  360. CCOMMON_OPT += -DQUAD_PRECISION
  361. NO_EXPRECISION = 1
  362. endif
  363. ifneq ($(ARCH), x86)
  364. ifneq ($(ARCH), x86_64)
  365. NO_EXPRECISION = 1
  366. endif
  367. endif
  368. ifdef UTEST_CHECK
  369. CCOMMON_OPT += -DUTEST_CHECK
  370. SANITY_CHECK = 1
  371. endif
  372. ifdef SANITY_CHECK
  373. CCOMMON_OPT += -DSANITY_CHECK -DREFNAME=$(*F)f$(BU)
  374. endif
  375. MAX_STACK_ALLOC ?= 2048
  376. ifneq ($(MAX_STACK_ALLOC), 0)
  377. CCOMMON_OPT += -DMAX_STACK_ALLOC=$(MAX_STACK_ALLOC)
  378. endif
  379. ifdef USE_LOCKING
  380. ifneq ($(USE_LOCKING), 0)
  381. CCOMMON_OPT += -DUSE_LOCKING
  382. endif
  383. endif
  384. #
  385. # Architecture dependent settings
  386. #
  387. ifeq ($(ARCH), x86)
  388. ifndef BINARY
  389. NO_BINARY_MODE = 1
  390. endif
  391. ifeq ($(CORE), generic)
  392. NO_EXPRECISION = 1
  393. endif
  394. ifndef NO_EXPRECISION
  395. ifeq ($(F_COMPILER), GFORTRAN)
  396. # ifeq logical or. GCC or LSB
  397. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  398. EXPRECISION = 1
  399. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  400. FCOMMON_OPT += -m128bit-long-double
  401. endif
  402. ifeq ($(C_COMPILER), CLANG)
  403. EXPRECISION = 1
  404. CCOMMON_OPT += -DEXPRECISION
  405. FCOMMON_OPT += -m128bit-long-double
  406. endif
  407. endif
  408. endif
  409. endif
  410. ifeq ($(ARCH), x86_64)
  411. ifeq ($(CORE), generic)
  412. NO_EXPRECISION = 1
  413. endif
  414. ifndef NO_EXPRECISION
  415. ifeq ($(F_COMPILER), GFORTRAN)
  416. # ifeq logical or. GCC or LSB
  417. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  418. EXPRECISION = 1
  419. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  420. FCOMMON_OPT += -m128bit-long-double
  421. endif
  422. ifeq ($(C_COMPILER), CLANG)
  423. EXPRECISION = 1
  424. CCOMMON_OPT += -DEXPRECISION
  425. FCOMMON_OPT += -m128bit-long-double
  426. endif
  427. endif
  428. endif
  429. endif
  430. ifeq ($(C_COMPILER), INTEL)
  431. CCOMMON_OPT += -wd981
  432. endif
  433. ifeq ($(USE_OPENMP), 1)
  434. #check
  435. ifeq ($(USE_THREAD), 0)
  436. $(error OpenBLAS: Cannot set both USE_OPENMP=1 and USE_THREAD=0. The USE_THREAD=0 is only for building single thread version.)
  437. endif
  438. # ifeq logical or. GCC or LSB
  439. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  440. CCOMMON_OPT += -fopenmp
  441. endif
  442. ifeq ($(C_COMPILER), CLANG)
  443. CCOMMON_OPT += -fopenmp
  444. endif
  445. ifeq ($(C_COMPILER), INTEL)
  446. CCOMMON_OPT += -fopenmp
  447. endif
  448. ifeq ($(C_COMPILER), PGI)
  449. CCOMMON_OPT += -mp
  450. endif
  451. ifeq ($(C_COMPILER), OPEN64)
  452. CCOMMON_OPT += -mp
  453. CEXTRALIB += -lstdc++
  454. endif
  455. ifeq ($(C_COMPILER), PATHSCALE)
  456. CCOMMON_OPT += -mp
  457. endif
  458. endif
  459. ifeq ($(DYNAMIC_ARCH), 1)
  460. ifeq ($(ARCH), x86)
  461. DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
  462. CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
  463. endif
  464. ifeq ($(ARCH), x86_64)
  465. DYNAMIC_CORE = PRESCOTT CORE2
  466. ifeq ($(DYNAMIC_OLDER), 1)
  467. DYNAMIC_CORE += PENRYN DUNNINGTON
  468. endif
  469. DYNAMIC_CORE += NEHALEM
  470. ifeq ($(DYNAMIC_OLDER), 1)
  471. DYNAMIC_CORE += OPTERON OPTERON_SSE3
  472. endif
  473. DYNAMIC_CORE += BARCELONA
  474. ifeq ($(DYNAMIC_OLDER), 1)
  475. DYNAMIC_CORE += BOBCAT ATOM NANO
  476. endif
  477. ifneq ($(NO_AVX), 1)
  478. DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER STEAMROLLER EXCAVATOR
  479. endif
  480. ifneq ($(NO_AVX2), 1)
  481. DYNAMIC_CORE += HASWELL ZEN
  482. endif
  483. ifneq ($(NO_AVX512), 1)
  484. ifneq ($(NO_AVX2), 1)
  485. DYNAMIC_CORE += SKYLAKEX COOPERLAKE
  486. endif
  487. endif
  488. endif
  489. ifdef DYNAMIC_LIST
  490. override DYNAMIC_CORE = PRESCOTT $(DYNAMIC_LIST)
  491. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_PRESCOTT
  492. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  493. CCOMMON_OPT += $(XCCOMMON_OPT)
  494. #CCOMMON_OPT += -DDYNAMIC_LIST='$(DYNAMIC_LIST)'
  495. endif
  496. ifeq ($(ARCH), arm64)
  497. DYNAMIC_CORE = ARMV8
  498. DYNAMIC_CORE += CORTEXA53
  499. DYNAMIC_CORE += CORTEXA57
  500. DYNAMIC_CORE += CORTEXA72
  501. DYNAMIC_CORE += CORTEXA73
  502. DYNAMIC_CORE += NEOVERSEN1
  503. DYNAMIC_CORE += FALKOR
  504. DYNAMIC_CORE += THUNDERX
  505. DYNAMIC_CORE += THUNDERX2T99
  506. DYNAMIC_CORE += TSV110
  507. DYNAMIC_CORE += EMAG8180
  508. DYNAMIC_CORE += THUNDERX3T110
  509. endif
  510. ifeq ($(ARCH), zarch)
  511. DYNAMIC_CORE = ZARCH_GENERIC
  512. # if the compiler accepts -march=arch11 or -march=z13 and can compile a file
  513. # with z13-specific inline assembly, then we can include support for Z13.
  514. # note: -march=z13 is equivalent to -march=arch11 yet some compiler releases
  515. # only support one or the other.
  516. # note: LLVM version 6.x supported -march=z13 yet could not handle vector
  517. # registers in inline assembly, so the check for supporting the -march flag is
  518. # not enough.
  519. ZARCH_TEST_COMPILE=-c $(TOPDIR)/kernel/zarch/damin_z13.c -I$(TOPDIR) -o /dev/null > /dev/null 2> /dev/null
  520. ZARCH_CC_SUPPORTS_ARCH11=$(shell $(CC) -march=arch11 $(ZARCH_TEST_COMPILE) && echo 1)
  521. ZARCH_CC_SUPPORTS_Z13=$(shell $(CC) -march=z13 $(ZARCH_TEST_COMPILE) && echo 1)
  522. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH11), $(ZARCH_CC_SUPPORTS_Z13)), 1)
  523. DYNAMIC_CORE += Z13
  524. CCOMMON_OPT += -DDYN_Z13
  525. else
  526. $(info OpenBLAS: Not building Z13 kernels because the compiler $(CC) does not support it)
  527. endif
  528. # as above for z13, check for -march=arch12 and z14 support in the compiler.
  529. ZARCH_CC_SUPPORTS_ARCH12=$(shell $(CC) -march=arch12 $(ZARCH_TEST_COMPILE) && echo 1)
  530. ZARCH_CC_SUPPORTS_Z14=$(shell $(CC) -march=z14 $(ZARCH_TEST_COMPILE) && echo 1)
  531. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH12), $(ZARCH_CC_SUPPORTS_Z14)), 1)
  532. DYNAMIC_CORE += Z14
  533. CCOMMON_OPT += -DDYN_Z14
  534. else
  535. $(info OpenBLAS: Not building Z14 kernels because the compiler $(CC) does not support it)
  536. endif
  537. endif # ARCH zarch
  538. ifeq ($(ARCH), power)
  539. DYNAMIC_CORE = POWER6
  540. DYNAMIC_CORE += POWER8
  541. ifneq ($(C_COMPILER), GCC)
  542. DYNAMIC_CORE += POWER9
  543. DYNAMIC_CORE += POWER10
  544. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  545. endif
  546. ifeq ($(C_COMPILER), GCC)
  547. ifeq ($(GCCVERSIONGT5), 1)
  548. DYNAMIC_CORE += POWER9
  549. else
  550. $(info, OpenBLAS: Your gcc version is too old to build the POWER9 kernels.)
  551. endif
  552. LDVERSIONGTEQ35 := $(shell expr `ld --version | head -1 | cut -f2 -d "." | cut -f1 -d "-"` >= 35)
  553. ifeq ($(GCCVERSIONGTEQ11)$(LDVERSIONGTEQ35), 11)
  554. DYNAMIC_CORE += POWER10
  555. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  556. else ifeq ($(GCCVERSIONGTEQ10), 1)
  557. ifeq ($(GCCMINORVERSIONGTEQ2)$(LDVERSIONGTEQ35), 11)
  558. DYNAMIC_CORE += POWER10
  559. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  560. endif
  561. else
  562. $(info, OpenBLAS: Your gcc version is too old to build the POWER10 kernels.)
  563. endif
  564. endif
  565. endif
  566. # If DYNAMIC_CORE is not set, DYNAMIC_ARCH cannot do anything, so force it to empty
  567. ifndef DYNAMIC_CORE
  568. override DYNAMIC_ARCH=
  569. endif
  570. endif
  571. ifeq ($(ARCH), ia64)
  572. NO_BINARY_MODE = 1
  573. BINARY_DEFINED = 1
  574. ifeq ($(F_COMPILER), GFORTRAN)
  575. ifeq ($(C_COMPILER), GCC)
  576. # EXPRECISION = 1
  577. # CCOMMON_OPT += -DEXPRECISION
  578. endif
  579. endif
  580. endif
  581. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  582. NO_BINARY_MODE = 1
  583. endif
  584. ifeq ($(ARCH), alpha)
  585. NO_BINARY_MODE = 1
  586. BINARY_DEFINED = 1
  587. endif
  588. ifeq ($(ARCH), arm)
  589. NO_BINARY_MODE = 1
  590. BINARY_DEFINED = 1
  591. CCOMMON_OPT += -marm
  592. FCOMMON_OPT += -marm
  593. # If softfp abi is mentioned on the command line, force it.
  594. ifeq ($(ARM_SOFTFP_ABI), 1)
  595. CCOMMON_OPT += -mfloat-abi=softfp
  596. FCOMMON_OPT += -mfloat-abi=softfp
  597. endif
  598. ifeq ($(OSNAME), Android)
  599. ifeq ($(ARM_SOFTFP_ABI), 1)
  600. EXTRALIB += -lm
  601. else
  602. EXTRALIB += -Wl,-lm_hard
  603. endif
  604. endif
  605. endif
  606. ifeq ($(ARCH), arm64)
  607. NO_BINARY_MODE = 1
  608. BINARY_DEFINED = 1
  609. ifdef INTERFACE64
  610. ifneq ($(INTERFACE64), 0)
  611. ifeq ($(F_COMPILER), GFORTRAN)
  612. FCOMMON_OPT += -fdefault-integer-8
  613. endif
  614. ifeq ($(F_COMPILER), FLANG)
  615. FCOMMON_OPT += -i8
  616. endif
  617. endif
  618. endif
  619. endif
  620. #
  621. # C Compiler dependent settings
  622. #
  623. # ifeq logical or. GCC or CLANG or LSB
  624. # http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
  625. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG LSB))
  626. CCOMMON_OPT += -Wall
  627. COMMON_PROF += -fno-inline
  628. NO_UNINITIALIZED_WARN = -Wno-uninitialized
  629. ifeq ($(QUIET_MAKE), 1)
  630. CCOMMON_OPT += $(NO_UNINITIALIZED_WARN) -Wno-unused
  631. endif
  632. ifdef NO_BINARY_MODE
  633. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  634. ifdef BINARY64
  635. CCOMMON_OPT += -mabi=64
  636. else
  637. CCOMMON_OPT += -mabi=n32
  638. endif
  639. BINARY_DEFINED = 1
  640. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  641. CCOMMON_OPT += -mabi=32
  642. BINARY_DEFINED = 1
  643. endif
  644. ifeq ($(CORE), LOONGSON3A)
  645. CCOMMON_OPT += -march=mips64
  646. FCOMMON_OPT += -march=mips64
  647. endif
  648. ifeq ($(CORE), LOONGSON3B)
  649. CCOMMON_OPT += -march=mips64
  650. FCOMMON_OPT += -march=mips64
  651. endif
  652. ifeq ($(CORE), MIPS24K)
  653. CCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  654. FCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  655. endif
  656. ifeq ($(CORE), MIPS1004K)
  657. CCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  658. FCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  659. endif
  660. ifeq ($(CORE), P5600)
  661. CCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  662. FCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  663. endif
  664. ifeq ($(CORE), I6400)
  665. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  666. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  667. endif
  668. ifeq ($(CORE), P6600)
  669. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  670. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  671. endif
  672. ifeq ($(CORE), I6500)
  673. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  674. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  675. endif
  676. ifeq ($(OSNAME), AIX)
  677. BINARY_DEFINED = 1
  678. endif
  679. endif
  680. ifndef BINARY_DEFINED
  681. ifneq ($(OSNAME), AIX)
  682. ifdef BINARY64
  683. CCOMMON_OPT += -m64
  684. else
  685. CCOMMON_OPT += -m32
  686. endif
  687. endif
  688. endif
  689. endif
  690. ifeq ($(C_COMPILER), PGI)
  691. ifdef BINARY64
  692. ifeq ($(ARCH), x86_64)
  693. CCOMMON_OPT += -tp p7-64 -D__MMX__ -Mnollvm
  694. else
  695. ifeq ($(ARCH), power)
  696. ifeq ($(CORE), POWER8)
  697. CCOMMON_OPT += -tp pwr8
  698. endif
  699. ifeq ($(CORE), POWER9)
  700. CCOMMON_OPT += -tp pwr9
  701. endif
  702. endif
  703. endif
  704. else
  705. CCOMMON_OPT += -tp p7
  706. endif
  707. endif
  708. ifeq ($(C_COMPILER), PATHSCALE)
  709. ifdef BINARY64
  710. CCOMMON_OPT += -m64
  711. else
  712. CCOMMON_OPT += -m32
  713. endif
  714. endif
  715. #
  716. # Fortran Compiler dependent settings
  717. #
  718. ifeq ($(F_COMPILER), FLANG)
  719. CCOMMON_OPT += -DF_INTERFACE_FLANG
  720. FCOMMON_OPT += -Mrecursive -Kieee
  721. ifeq ($(OSNAME), Linux)
  722. ifeq ($(ARCH), x86_64)
  723. FLANG_VENDOR := $(shell `$(FC) --version|cut -f 1 -d "."|head -1`)
  724. ifeq ($(FLANG_VENDOR),AOCC)
  725. FCOMMON_OPT += -fno-unroll-loops
  726. endif
  727. endif
  728. endif
  729. ifdef BINARY64
  730. ifdef INTERFACE64
  731. ifneq ($(INTERFACE64), 0)
  732. FCOMMON_OPT += -i8
  733. endif
  734. endif
  735. FCOMMON_OPT += -Wall
  736. else
  737. FCOMMON_OPT += -Wall
  738. endif
  739. ifeq ($(USE_OPENMP), 1)
  740. FCOMMON_OPT += -fopenmp
  741. endif
  742. endif
  743. ifeq ($(F_COMPILER), G77)
  744. CCOMMON_OPT += -DF_INTERFACE_G77
  745. FCOMMON_OPT += -Wall
  746. ifndef NO_BINARY_MODE
  747. ifneq ($(OSNAME), AIX)
  748. ifdef BINARY64
  749. FCOMMON_OPT += -m64
  750. else
  751. FCOMMON_OPT += -m32
  752. endif
  753. endif
  754. endif
  755. endif
  756. ifeq ($(F_COMPILER), G95)
  757. CCOMMON_OPT += -DF_INTERFACE_G95
  758. FCOMMON_OPT += -Wall
  759. ifneq ($(OSNAME), AIX)
  760. ifndef NO_BINARY_MODE
  761. ifdef BINARY64
  762. FCOMMON_OPT += -m64
  763. else
  764. FCOMMON_OPT += -m32
  765. endif
  766. endif
  767. ifneq ($(NO_LAPACKE), 1)
  768. FCOMMON_OPT += -fno-second-underscore
  769. endif
  770. endif
  771. endif
  772. ifeq ($(F_COMPILER), GFORTRAN)
  773. CCOMMON_OPT += -DF_INTERFACE_GFORT
  774. FCOMMON_OPT += -Wall
  775. # make single-threaded LAPACK calls thread-safe #1847
  776. FCOMMON_OPT += -frecursive
  777. # work around ABI problem with passing single-character arguments
  778. FCOMMON_OPT += -fno-optimize-sibling-calls
  779. #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
  780. ifneq ($(NO_LAPACK), 1)
  781. EXTRALIB += -lgfortran
  782. endif
  783. ifdef NO_BINARY_MODE
  784. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  785. ifdef BINARY64
  786. FCOMMON_OPT += -mabi=64
  787. else
  788. FCOMMON_OPT += -mabi=n32
  789. endif
  790. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  791. FCOMMON_OPT += -mabi=32
  792. endif
  793. else
  794. ifdef BINARY64
  795. ifneq ($(OSNAME), AIX)
  796. FCOMMON_OPT += -m64
  797. endif
  798. ifdef INTERFACE64
  799. ifneq ($(INTERFACE64), 0)
  800. FCOMMON_OPT += -fdefault-integer-8
  801. endif
  802. endif
  803. else
  804. ifneq ($(OSNAME), AIX)
  805. FCOMMON_OPT += -m32
  806. endif
  807. endif
  808. endif
  809. ifeq ($(USE_OPENMP), 1)
  810. FCOMMON_OPT += -fopenmp
  811. endif
  812. endif
  813. ifeq ($(F_COMPILER), INTEL)
  814. CCOMMON_OPT += -DF_INTERFACE_INTEL
  815. ifdef INTERFACE64
  816. ifneq ($(INTERFACE64), 0)
  817. FCOMMON_OPT += -i8
  818. endif
  819. endif
  820. FCOMMON_OPT += -recursive -fp-model strict -assume protect-parens
  821. ifeq ($(USE_OPENMP), 1)
  822. FCOMMON_OPT += -fopenmp
  823. endif
  824. endif
  825. ifeq ($(F_COMPILER), FUJITSU)
  826. CCOMMON_OPT += -DF_INTERFACE_FUJITSU
  827. ifeq ($(USE_OPENMP), 1)
  828. FCOMMON_OPT += -openmp
  829. endif
  830. endif
  831. ifeq ($(F_COMPILER), IBM)
  832. CCOMMON_OPT += -DF_INTERFACE_IBM
  833. # FCOMMON_OPT += -qarch=440
  834. ifdef BINARY64
  835. FCOMMON_OPT += -q64
  836. ifdef INTERFACE64
  837. ifneq ($(INTERFACE64), 0)
  838. FCOMMON_OPT += -qintsize=8
  839. endif
  840. endif
  841. else
  842. FCOMMON_OPT += -q32
  843. endif
  844. ifeq ($(USE_OPENMP), 1)
  845. FCOMMON_OPT += -openmp
  846. endif
  847. endif
  848. ifeq ($(F_COMPILER), PGI)
  849. CCOMMON_OPT += -DF_INTERFACE_PGI
  850. COMMON_PROF += -DPGICOMPILER
  851. ifdef BINARY64
  852. ifdef INTERFACE64
  853. ifneq ($(INTERFACE64), 0)
  854. FCOMMON_OPT += -i8
  855. endif
  856. endif
  857. ifeq ($(ARCH), x86_64)
  858. FCOMMON_OPT += -tp p7-64
  859. else
  860. ifeq ($(ARCH), power)
  861. ifeq ($(CORE), POWER8)
  862. FCOMMON_OPT += -tp pwr8
  863. endif
  864. ifeq ($(CORE), POWER9)
  865. FCOMMON_OPT += -tp pwr9
  866. endif
  867. endif
  868. endif
  869. else
  870. FCOMMON_OPT += -tp p7
  871. endif
  872. FCOMMON_OPT += -Mrecursive
  873. ifeq ($(USE_OPENMP), 1)
  874. FCOMMON_OPT += -mp
  875. endif
  876. endif
  877. ifeq ($(F_COMPILER), PATHSCALE)
  878. CCOMMON_OPT += -DF_INTERFACE_PATHSCALE
  879. ifdef BINARY64
  880. ifdef INTERFACE64
  881. ifneq ($(INTERFACE64), 0)
  882. FCOMMON_OPT += -i8
  883. endif
  884. endif
  885. endif
  886. ifeq ($(USE_OPENMP), 1)
  887. FCOMMON_OPT += -mp
  888. endif
  889. endif
  890. ifeq ($(F_COMPILER), OPEN64)
  891. CCOMMON_OPT += -DF_INTERFACE_OPEN64
  892. ifdef BINARY64
  893. ifdef INTERFACE64
  894. ifneq ($(INTERFACE64), 0)
  895. FCOMMON_OPT += -i8
  896. endif
  897. endif
  898. endif
  899. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  900. ifndef BINARY64
  901. FCOMMON_OPT += -n32
  902. else
  903. FCOMMON_OPT += -n64
  904. endif
  905. ifeq ($(CORE), LOONGSON3A)
  906. FCOMMON_OPT += -loongson3 -static
  907. endif
  908. ifeq ($(CORE), LOONGSON3B)
  909. FCOMMON_OPT += -loongson3 -static
  910. endif
  911. else
  912. ifndef BINARY64
  913. FCOMMON_OPT += -m32
  914. else
  915. FCOMMON_OPT += -m64
  916. endif
  917. endif
  918. ifeq ($(USE_OPENMP), 1)
  919. FEXTRALIB += -lstdc++
  920. FCOMMON_OPT += -mp
  921. endif
  922. endif
  923. ifeq ($(C_COMPILER), OPEN64)
  924. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  925. ifndef BINARY64
  926. CCOMMON_OPT += -n32
  927. else
  928. CCOMMON_OPT += -n64
  929. endif
  930. ifeq ($(CORE), LOONGSON3A)
  931. CCOMMON_OPT += -loongson3 -static
  932. endif
  933. ifeq ($(CORE), LOONGSON3B)
  934. CCOMMON_OPT += -loongson3 -static
  935. endif
  936. else
  937. ifndef BINARY64
  938. CCOMMON_OPT += -m32
  939. else
  940. CCOMMON_OPT += -m64
  941. endif
  942. endif
  943. endif
  944. ifeq ($(C_COMPILER), SUN)
  945. CCOMMON_OPT += -w
  946. ifeq ($(ARCH), x86)
  947. CCOMMON_OPT += -m32
  948. else
  949. FCOMMON_OPT += -m64
  950. endif
  951. endif
  952. ifeq ($(F_COMPILER), SUN)
  953. CCOMMON_OPT += -DF_INTERFACE_SUN
  954. ifeq ($(ARCH), x86)
  955. FCOMMON_OPT += -m32
  956. else
  957. FCOMMON_OPT += -m64
  958. endif
  959. ifeq ($(USE_OPENMP), 1)
  960. FCOMMON_OPT += -xopenmp=parallel
  961. endif
  962. endif
  963. ifeq ($(F_COMPILER), COMPAQ)
  964. CCOMMON_OPT += -DF_INTERFACE_COMPAQ
  965. ifeq ($(USE_OPENMP), 1)
  966. FCOMMON_OPT += -openmp
  967. endif
  968. endif
  969. ifdef BINARY64
  970. ifdef INTERFACE64
  971. ifneq ($(INTERFACE64), 0)
  972. CCOMMON_OPT +=
  973. #-DUSE64BITINT
  974. endif
  975. endif
  976. endif
  977. ifeq ($(NEED_PIC), 1)
  978. ifeq ($(C_COMPILER), IBM)
  979. CCOMMON_OPT += -qpic=large
  980. else
  981. CCOMMON_OPT += -fPIC
  982. endif
  983. ifeq ($(F_COMPILER), SUN)
  984. FCOMMON_OPT += -pic
  985. else
  986. FCOMMON_OPT += -fPIC
  987. endif
  988. endif
  989. ifeq ($(DYNAMIC_ARCH), 1)
  990. CCOMMON_OPT += -DDYNAMIC_ARCH
  991. endif
  992. ifeq ($(DYNAMIC_OLDER), 1)
  993. CCOMMON_OPT += -DDYNAMIC_OLDER
  994. endif
  995. ifeq ($(NO_LAPACK), 1)
  996. CCOMMON_OPT += -DNO_LAPACK
  997. #Disable LAPACK C interface
  998. NO_LAPACKE = 1
  999. endif
  1000. ifeq ($(NO_LAPACKE), 1)
  1001. CCOMMON_OPT += -DNO_LAPACKE
  1002. endif
  1003. ifeq ($(NO_AVX), 1)
  1004. CCOMMON_OPT += -DNO_AVX
  1005. endif
  1006. ifeq ($(ARCH), x86)
  1007. CCOMMON_OPT += -DNO_AVX
  1008. endif
  1009. ifeq ($(NO_AVX2), 1)
  1010. CCOMMON_OPT += -DNO_AVX2
  1011. endif
  1012. ifeq ($(NO_AVX512), 1)
  1013. CCOMMON_OPT += -DNO_AVX512
  1014. endif
  1015. ifdef SMP
  1016. CCOMMON_OPT += -DSMP_SERVER
  1017. ifeq ($(ARCH), mips64)
  1018. ifneq ($(CORE), LOONGSON3B)
  1019. USE_SIMPLE_THREADED_LEVEL3 = 1
  1020. endif
  1021. endif
  1022. ifeq ($(USE_OPENMP), 1)
  1023. # USE_SIMPLE_THREADED_LEVEL3 = 1
  1024. # NO_AFFINITY = 1
  1025. CCOMMON_OPT += -DUSE_OPENMP
  1026. endif
  1027. ifeq ($(BIGNUMA), 1)
  1028. CCOMMON_OPT += -DBIGNUMA
  1029. endif
  1030. endif
  1031. ifeq ($(NO_WARMUP), 1)
  1032. CCOMMON_OPT += -DNO_WARMUP
  1033. endif
  1034. ifeq ($(CONSISTENT_FPCSR), 1)
  1035. CCOMMON_OPT += -DCONSISTENT_FPCSR
  1036. endif
  1037. # Only for development
  1038. # CCOMMON_OPT += -DPARAMTEST
  1039. # CCOMMON_OPT += -DPREFETCHTEST
  1040. # CCOMMON_OPT += -DNO_SWITCHING
  1041. # USE_PAPI = 1
  1042. ifdef USE_PAPI
  1043. CCOMMON_OPT += -DUSE_PAPI
  1044. EXTRALIB += -lpapi -lperfctr
  1045. endif
  1046. ifdef DYNAMIC_THREADS
  1047. CCOMMON_OPT += -DDYNAMIC_THREADS
  1048. endif
  1049. CCOMMON_OPT += -DMAX_CPU_NUMBER=$(NUM_THREADS)
  1050. CCOMMON_OPT += -DMAX_PARALLEL_NUMBER=$(NUM_PARALLEL)
  1051. ifdef USE_SIMPLE_THREADED_LEVEL3
  1052. CCOMMON_OPT += -DUSE_SIMPLE_THREADED_LEVEL3
  1053. endif
  1054. ifeq ($(USE_TLS), 1)
  1055. CCOMMON_OPT += -DUSE_TLS
  1056. endif
  1057. ifeq ($(BUILD_BFLOAT16), 1)
  1058. CCOMMON_OPT += -DBUILD_BFLOAT16
  1059. endif
  1060. ifeq ($(BUILD_SINGLE), 1)
  1061. CCOMMON_OPT += -DBUILD_SINGLE=1
  1062. endif
  1063. ifeq ($(BUILD_DOUBLE), 1)
  1064. CCOMMON_OPT += -DBUILD_DOUBLE=1
  1065. endif
  1066. ifeq ($(BUILD_COMPLEX), 1)
  1067. CCOMMON_OPT += -DBUILD_COMPLEX=1
  1068. endif
  1069. ifeq ($(BUILD_COMPLEX16), 1)
  1070. CCOMMON_OPT += -DBUILD_COMPLEX16=1
  1071. endif
  1072. CCOMMON_OPT += -DVERSION=\"$(VERSION)\"
  1073. ifndef SYMBOLPREFIX
  1074. SYMBOLPREFIX =
  1075. endif
  1076. ifndef SYMBOLSUFFIX
  1077. SYMBOLSUFFIX =
  1078. endif
  1079. ifndef LIBNAMESUFFIX
  1080. LIBNAMEBASE = $(SYMBOLPREFIX)openblas$(SYMBOLSUFFIX)
  1081. else
  1082. LIBNAMEBASE = $(SYMBOLPREFIX)openblas$(SYMBOLSUFFIX)_$(LIBNAMESUFFIX)
  1083. endif
  1084. ifeq ($(OSNAME), CYGWIN_NT)
  1085. LIBPREFIX = cyg$(LIBNAMEBASE)
  1086. else
  1087. LIBPREFIX = lib$(LIBNAMEBASE)
  1088. endif
  1089. KERNELDIR = $(TOPDIR)/kernel/$(ARCH)
  1090. include $(TOPDIR)/Makefile.$(ARCH)
  1091. ifneq ($(C_COMPILER), PGI)
  1092. CCOMMON_OPT += -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME
  1093. endif
  1094. CCOMMON_OPT += -DASMNAME=$(FU)$(*F) -DASMFNAME=$(FU)$(*F)$(BU) -DNAME=$(*F)$(BU) -DCNAME=$(*F) -DCHAR_NAME=\"$(*F)$(BU)\" -DCHAR_CNAME=\"$(*F)\"
  1095. ifeq ($(CORE), PPC440)
  1096. CCOMMON_OPT += -DALLOC_QALLOC
  1097. endif
  1098. ifeq ($(CORE), PPC440FP2)
  1099. STATIC_ALLOCATION = 1
  1100. endif
  1101. ifneq ($(OSNAME), Linux)
  1102. NO_AFFINITY = 1
  1103. endif
  1104. ifneq ($(ARCH), x86_64)
  1105. ifneq ($(ARCH), x86)
  1106. ifneq ($(CORE), LOONGSON3B)
  1107. NO_AFFINITY = 1
  1108. endif
  1109. endif
  1110. endif
  1111. ifdef NO_AFFINITY
  1112. ifeq ($(NO_AFFINITY), 0)
  1113. override undefine NO_AFFINITY
  1114. else
  1115. CCOMMON_OPT += -DNO_AFFINITY
  1116. endif
  1117. endif
  1118. ifdef FUNCTION_PROFILE
  1119. CCOMMON_OPT += -DFUNCTION_PROFILE
  1120. endif
  1121. ifdef HUGETLB_ALLOCATION
  1122. CCOMMON_OPT += -DALLOC_HUGETLB
  1123. endif
  1124. ifdef HUGETLBFILE_ALLOCATION
  1125. CCOMMON_OPT += -DALLOC_HUGETLBFILE -DHUGETLB_FILE_NAME=$(HUGETLBFILE_ALLOCATION)
  1126. endif
  1127. ifdef STATIC_ALLOCATION
  1128. CCOMMON_OPT += -DALLOC_STATIC
  1129. endif
  1130. ifdef DEVICEDRIVER_ALLOCATION
  1131. CCOMMON_OPT += -DALLOC_DEVICEDRIVER -DDEVICEDRIVER_NAME=\"/dev/mapper\"
  1132. endif
  1133. ifdef MIXED_MEMORY_ALLOCATION
  1134. CCOMMON_OPT += -DMIXED_MEMORY_ALLOCATION
  1135. endif
  1136. ifeq ($(OSNAME), SunOS)
  1137. TAR = gtar
  1138. PATCH = gpatch
  1139. GREP = ggrep
  1140. AWK = nawk
  1141. else
  1142. TAR = tar
  1143. PATCH = patch
  1144. GREP = grep
  1145. AWK = awk
  1146. endif
  1147. ifndef MD5SUM
  1148. MD5SUM = md5sum
  1149. endif
  1150. REVISION = -r$(VERSION)
  1151. MAJOR_VERSION = $(word 1,$(subst ., ,$(VERSION)))
  1152. ifeq ($(DEBUG), 1)
  1153. COMMON_OPT += -g
  1154. endif
  1155. ifeq ($(DEBUG), 1)
  1156. FCOMMON_OPT += -g
  1157. endif
  1158. ifndef COMMON_OPT
  1159. COMMON_OPT = -O2
  1160. endif
  1161. ifndef FCOMMON_OPT
  1162. FCOMMON_OPT = -O2 -frecursive
  1163. endif
  1164. override CFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
  1165. override PFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
  1166. override FFLAGS += $(COMMON_OPT) $(FCOMMON_OPT)
  1167. override FPFLAGS += $(FCOMMON_OPT) $(COMMON_PROF)
  1168. #MAKEOVERRIDES =
  1169. ifdef NEED_PIC
  1170. ifeq (,$(findstring PIC,$(FFLAGS)))
  1171. override FFLAGS += -fPIC
  1172. endif
  1173. endif
  1174. #For LAPACK Fortran codes.
  1175. #Disable -fopenmp for LAPACK Fortran codes on Windows.
  1176. ifdef OS_WINDOWS
  1177. LAPACK_FFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FFLAGS))
  1178. LAPACK_FPFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FPFLAGS))
  1179. else
  1180. LAPACK_FFLAGS := $(FFLAGS)
  1181. LAPACK_FPFLAGS := $(FPFLAGS)
  1182. endif
  1183. LAPACK_CFLAGS = $(CFLAGS)
  1184. LAPACK_CFLAGS += -DHAVE_LAPACK_CONFIG_H
  1185. ifdef INTERFACE64
  1186. ifneq ($(INTERFACE64), 0)
  1187. LAPACK_CFLAGS += -DLAPACK_ILP64
  1188. endif
  1189. endif
  1190. ifdef OS_WINDOWS
  1191. LAPACK_CFLAGS += -DOPENBLAS_OS_WINDOWS
  1192. endif
  1193. ifeq ($(C_COMPILER), LSB)
  1194. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1195. endif
  1196. ifndef SUFFIX
  1197. SUFFIX = o
  1198. endif
  1199. ifndef PSUFFIX
  1200. PSUFFIX = po
  1201. endif
  1202. ifndef LIBSUFFIX
  1203. LIBSUFFIX = a
  1204. endif
  1205. ifneq ($(DYNAMIC_ARCH), 1)
  1206. ifndef SMP
  1207. LIBNAME = $(LIBPREFIX)_$(LIBCORE)$(REVISION).$(LIBSUFFIX)
  1208. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)$(REVISION)_p.$(LIBSUFFIX)
  1209. else
  1210. LIBNAME = $(LIBPREFIX)_$(LIBCORE)p$(REVISION).$(LIBSUFFIX)
  1211. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)p$(REVISION)_p.$(LIBSUFFIX)
  1212. endif
  1213. else
  1214. ifndef SMP
  1215. LIBNAME = $(LIBPREFIX)$(REVISION).$(LIBSUFFIX)
  1216. LIBNAME_P = $(LIBPREFIX)$(REVISION)_p.$(LIBSUFFIX)
  1217. else
  1218. LIBNAME = $(LIBPREFIX)p$(REVISION).$(LIBSUFFIX)
  1219. LIBNAME_P = $(LIBPREFIX)p$(REVISION)_p.$(LIBSUFFIX)
  1220. endif
  1221. endif
  1222. LIBDLLNAME = $(LIBPREFIX).dll
  1223. IMPLIBNAME = lib$(LIBNAMEBASE).dll.a
  1224. ifneq ($(OSNAME), AIX)
  1225. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.so)
  1226. else
  1227. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.a)
  1228. endif
  1229. LIBDYNNAME = $(LIBNAME:.$(LIBSUFFIX)=.dylib)
  1230. LIBDEFNAME = $(LIBNAME:.$(LIBSUFFIX)=.def)
  1231. LIBEXPNAME = $(LIBNAME:.$(LIBSUFFIX)=.exp)
  1232. LIBZIPNAME = $(LIBNAME:.$(LIBSUFFIX)=.zip)
  1233. LIBS = $(TOPDIR)/$(LIBNAME)
  1234. LIBS_P = $(TOPDIR)/$(LIBNAME_P)
  1235. LIB_COMPONENTS = BLAS
  1236. ifneq ($(NO_CBLAS), 1)
  1237. LIB_COMPONENTS += CBLAS
  1238. endif
  1239. ifneq ($(NO_LAPACK), 1)
  1240. LIB_COMPONENTS += LAPACK
  1241. ifneq ($(NO_LAPACKE), 1)
  1242. LIB_COMPONENTS += LAPACKE
  1243. endif
  1244. ifeq ($(BUILD_RELAPACK), 1)
  1245. LIB_COMPONENTS += ReLAPACK
  1246. endif
  1247. endif
  1248. ifeq ($(ONLY_CBLAS), 1)
  1249. LIB_COMPONENTS = CBLAS
  1250. endif
  1251. export OSNAME
  1252. export ARCH
  1253. export CORE
  1254. export LIBCORE
  1255. export __BYTE_ORDER__
  1256. export ELF_VERSION
  1257. export PGCPATH
  1258. export CONFIG
  1259. export CC
  1260. export FC
  1261. export BU
  1262. export FU
  1263. export NEED2UNDERSCORES
  1264. export USE_THREAD
  1265. export NUM_THREADS
  1266. export NUM_CORES
  1267. export SMP
  1268. export MAKEFILE_RULE
  1269. export NEED_PIC
  1270. export BINARY
  1271. export BINARY32
  1272. export BINARY64
  1273. export F_COMPILER
  1274. export C_COMPILER
  1275. export USE_OPENMP
  1276. export CROSS
  1277. export CROSS_SUFFIX
  1278. export NOFORTRAN
  1279. export NO_FBLAS
  1280. export EXTRALIB
  1281. export CEXTRALIB
  1282. export FEXTRALIB
  1283. export HAVE_SSE
  1284. export HAVE_SSE2
  1285. export HAVE_SSE3
  1286. export HAVE_SSSE3
  1287. export HAVE_SSE4_1
  1288. export HAVE_SSE4_2
  1289. export HAVE_SSE4A
  1290. export HAVE_SSE5
  1291. export HAVE_AVX
  1292. export HAVE_VFP
  1293. export HAVE_VFPV3
  1294. export HAVE_VFPV4
  1295. export HAVE_NEON
  1296. export HAVE_MSA
  1297. export MSA_FLAGS
  1298. export KERNELDIR
  1299. export FUNCTION_PROFILE
  1300. export TARGET_CORE
  1301. export NO_AVX512
  1302. export BUILD_BFLOAT16
  1303. export SBGEMM_UNROLL_M
  1304. export SBGEMM_UNROLL_N
  1305. export SGEMM_UNROLL_M
  1306. export SGEMM_UNROLL_N
  1307. export DGEMM_UNROLL_M
  1308. export DGEMM_UNROLL_N
  1309. export QGEMM_UNROLL_M
  1310. export QGEMM_UNROLL_N
  1311. export CGEMM_UNROLL_M
  1312. export CGEMM_UNROLL_N
  1313. export ZGEMM_UNROLL_M
  1314. export ZGEMM_UNROLL_N
  1315. export XGEMM_UNROLL_M
  1316. export XGEMM_UNROLL_N
  1317. export CGEMM3M_UNROLL_M
  1318. export CGEMM3M_UNROLL_N
  1319. export ZGEMM3M_UNROLL_M
  1320. export ZGEMM3M_UNROLL_N
  1321. export XGEMM3M_UNROLL_M
  1322. export XGEMM3M_UNROLL_N
  1323. ifdef USE_CUDA
  1324. export CUDADIR
  1325. export CUCC
  1326. export CUFLAGS
  1327. export CULIB
  1328. endif
  1329. .SUFFIXES: .$(PSUFFIX) .$(SUFFIX) .f
  1330. .f.$(SUFFIX):
  1331. $(FC) $(FFLAGS) -c $< -o $(@F)
  1332. .f.$(PSUFFIX):
  1333. $(FC) $(FPFLAGS) -pg -c $< -o $(@F)
  1334. ifdef BINARY64
  1335. PATHSCALEPATH = /opt/pathscale/lib/3.1
  1336. PGIPATH = /opt/pgi/linux86-64/7.1-5/lib
  1337. else
  1338. PATHSCALEPATH = /opt/pathscale/lib/3.1/32
  1339. PGIPATH = /opt/pgi/linux86/7.1-5/lib
  1340. endif
  1341. ACMLPATH = /opt/acml/4.3.0
  1342. ifneq ($(OSNAME), Darwin)
  1343. MKLPATH = /opt/intel/mkl/10.2.2.025/lib
  1344. else
  1345. MKLPATH = /Library/Frameworks/Intel_MKL.framework/Versions/10.0.1.014/lib
  1346. endif
  1347. ATLASPATH = /opt/atlas/3.9.17/opteron
  1348. FLAMEPATH = $(HOME)/flame/lib
  1349. ifneq ($(OSNAME), SunOS)
  1350. SUNPATH = /opt/sunstudio12.1
  1351. else
  1352. SUNPATH = /opt/SUNWspro
  1353. endif