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 32 kB

6 years ago
13 years ago
12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448144914501451145214531454145514561457145814591460146114621463146414651466146714681469147014711472147314741475147614771478147914801481148214831484148514861487148814891490149114921493149414951496149714981499150015011502150315041505150615071508150915101511151215131514151515161517151815191520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559156015611562156315641565156615671568156915701571157215731574157515761577157815791580158115821583158415851586158715881589159015911592159315941595159615971598159916001601160216031604160516061607160816091610161116121613161416151616161716181619162016211622162316241625
  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. ifeq ($(TARGET), GENERIC)
  84. ifeq ($(DYNAMIC_ARCH), 1)
  85. override NO_EXPRECISION=1
  86. export NO_EXPRECiSION
  87. endif
  88. endif
  89. endif
  90. # Force fallbacks for 32bit
  91. ifeq ($(BINARY), 32)
  92. ifeq ($(TARGET), HASWELL)
  93. GETARCH_FLAGS := -DFORCE_NEHALEM
  94. endif
  95. ifeq ($(TARGET), SKYLAKEX)
  96. GETARCH_FLAGS := -DFORCE_NEHALEM
  97. endif
  98. ifeq ($(TARGET), COOPERLAKE)
  99. GETARCH_FLAGS := -DFORCE_NEHALEM
  100. endif
  101. ifeq ($(TARGET), SANDYBRIDGE)
  102. GETARCH_FLAGS := -DFORCE_NEHALEM
  103. endif
  104. ifeq ($(TARGET), BULLDOZER)
  105. GETARCH_FLAGS := -DFORCE_BARCELONA
  106. endif
  107. ifeq ($(TARGET), PILEDRIVER)
  108. GETARCH_FLAGS := -DFORCE_BARCELONA
  109. endif
  110. ifeq ($(TARGET), STEAMROLLER)
  111. GETARCH_FLAGS := -DFORCE_BARCELONA
  112. endif
  113. ifeq ($(TARGET), EXCAVATOR)
  114. GETARCH_FLAGS := -DFORCE_BARCELONA
  115. endif
  116. ifeq ($(TARGET), ZEN)
  117. GETARCH_FLAGS := -DFORCE_BARCELONA
  118. endif
  119. ifeq ($(TARGET), ARMV8)
  120. GETARCH_FLAGS := -DFORCE_ARMV7
  121. endif
  122. ifeq ($(TARGET), POWER8)
  123. GETARCH_FLAGS := -DFORCE_POWER6
  124. endif
  125. endif
  126. #TARGET_CORE will override TARGET which is used in DYNAMIC_ARCH=1.
  127. #
  128. ifdef TARGET_CORE
  129. GETARCH_FLAGS := -DFORCE_$(TARGET_CORE)
  130. endif
  131. # Force fallbacks for 32bit
  132. ifeq ($(BINARY), 32)
  133. ifeq ($(TARGET_CORE), HASWELL)
  134. GETARCH_FLAGS := -DFORCE_NEHALEM
  135. endif
  136. ifeq ($(TARGET_CORE), SKYLAKEX)
  137. GETARCH_FLAGS := -DFORCE_NEHALEM
  138. endif
  139. ifeq ($(TARGET_CORE), COOPERLAKE)
  140. GETARCH_FLAGS := -DFORCE_NEHALEM
  141. endif
  142. ifeq ($(TARGET_CORE), SANDYBRIDGE)
  143. GETARCH_FLAGS := -DFORCE_NEHALEM
  144. endif
  145. ifeq ($(TARGET_CORE), BULLDOZER)
  146. GETARCH_FLAGS := -DFORCE_BARCELONA
  147. endif
  148. ifeq ($(TARGET_CORE), PILEDRIVER)
  149. GETARCH_FLAGS := -DFORCE_BARCELONA
  150. endif
  151. ifeq ($(TARGET_CORE), STEAMROLLER)
  152. GETARCH_FLAGS := -DFORCE_BARCELONA
  153. endif
  154. ifeq ($(TARGET_CORE), EXCAVATOR)
  155. GETARCH_FLAGS := -DFORCE_BARCELONA
  156. endif
  157. ifeq ($(TARGET_CORE), ZEN)
  158. GETARCH_FLAGS := -DFORCE_BARCELONA
  159. endif
  160. endif
  161. # On x86_64 build getarch with march=native unless the compiler is PGI. This is required to detect AVX512 support in getarch.
  162. ifeq ($(HOSTARCH), x86_64)
  163. ifeq ($(findstring pgcc,$(HOSTCC)),)
  164. GETARCH_FLAGS += -march=native
  165. endif
  166. endif
  167. ifdef INTERFACE64
  168. ifneq ($(INTERFACE64), 0)
  169. GETARCH_FLAGS += -DUSE64BITINT
  170. endif
  171. endif
  172. ifndef GEMM_MULTITHREAD_THRESHOLD
  173. GEMM_MULTITHREAD_THRESHOLD=4
  174. endif
  175. GETARCH_FLAGS += -DGEMM_MULTITHREAD_THRESHOLD=$(GEMM_MULTITHREAD_THRESHOLD)
  176. ifeq ($(NO_AVX), 1)
  177. GETARCH_FLAGS += -DNO_AVX
  178. endif
  179. ifeq ($(BINARY), 32)
  180. GETARCH_FLAGS += -DNO_AVX -DNO_AVX2 -DNO_AVX512
  181. NO_AVX512 = 1
  182. endif
  183. ifeq ($(NO_AVX2), 1)
  184. GETARCH_FLAGS += -DNO_AVX2
  185. endif
  186. ifeq ($(NO_AVX512), 1)
  187. GETARCH_FLAGS += -DNO_AVX512
  188. endif
  189. ifeq ($(DEBUG), 1)
  190. GETARCH_FLAGS += -g
  191. endif
  192. ifeq ($(QUIET_MAKE), 1)
  193. MAKE += -s
  194. endif
  195. ifndef NO_PARALLEL_MAKE
  196. NO_PARALLEL_MAKE=0
  197. endif
  198. GETARCH_FLAGS += -DNO_PARALLEL_MAKE=$(NO_PARALLEL_MAKE)
  199. ifdef MAKE_NB_JOBS
  200. GETARCH_FLAGS += -DMAKE_NB_JOBS=$(MAKE_NB_JOBS)
  201. endif
  202. ifeq ($(HOSTCC), loongcc)
  203. GETARCH_FLAGS += -static
  204. endif
  205. #if don't use Fortran, it will only compile CBLAS.
  206. ifeq ($(ONLY_CBLAS), 1)
  207. NO_LAPACK = 1
  208. else
  209. ONLY_CBLAS = 0
  210. endif
  211. # This operation is expensive, so execution should be once.
  212. ifndef GOTOBLAS_MAKEFILE
  213. export GOTOBLAS_MAKEFILE = 1
  214. # Generating Makefile.conf and config.h
  215. 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)
  216. ifndef TARGET_CORE
  217. include $(TOPDIR)/Makefile.conf
  218. else
  219. HAVE_NEON=
  220. HAVE_VFP=
  221. HAVE_VFPV3=
  222. HAVE_VFPV4=
  223. HAVE_MMX=
  224. HAVE_SSE=
  225. HAVE_SSE2=
  226. HAVE_SSE3=
  227. HAVE_SSSE3=
  228. HAVE_SSE4_1=
  229. HAVE_SSE4_2=
  230. HAVE_SSE4A=
  231. HAVE_SSE5=
  232. HAVE_AVX=
  233. HAVE_AVX2=
  234. HAVE_FMA3=
  235. include $(TOPDIR)/Makefile_kernel.conf
  236. endif
  237. endif
  238. ifndef NUM_PARALLEL
  239. NUM_PARALLEL = 1
  240. endif
  241. ifndef NUM_THREADS
  242. NUM_THREADS = $(NUM_CORES)
  243. endif
  244. ifeq ($(NUM_THREADS), 1)
  245. override USE_THREAD = 0
  246. override USE_OPENMP = 0
  247. endif
  248. ifdef USE_THREAD
  249. ifeq ($(USE_THREAD), 0)
  250. SMP =
  251. else
  252. SMP = 1
  253. endif
  254. else
  255. ifeq ($(NUM_THREAD), 1)
  256. SMP =
  257. else
  258. SMP = 1
  259. endif
  260. endif
  261. ifeq ($(SMP), 1)
  262. USE_LOCKING =
  263. endif
  264. ifndef NEED_PIC
  265. NEED_PIC = 1
  266. endif
  267. ARFLAGS =
  268. CPP = $(COMPILER) -E
  269. AR ?= $(CROSS_SUFFIX)ar
  270. AS ?= $(CROSS_SUFFIX)as
  271. LD ?= $(CROSS_SUFFIX)ld
  272. RANLIB ?= $(CROSS_SUFFIX)ranlib
  273. NM = $(CROSS_SUFFIX)nm
  274. DLLWRAP = $(CROSS_SUFFIX)dllwrap
  275. OBJCOPY = $(CROSS_SUFFIX)objcopy
  276. OBJCONV = $(CROSS_SUFFIX)objconv
  277. # When fortran support was either not detected or actively deselected, only build BLAS.
  278. ifeq ($(NOFORTRAN), 1)
  279. NO_LAPACK = 1
  280. override FEXTRALIB =
  281. endif
  282. ifeq ($(C_COMPILER), GCC)
  283. GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
  284. GCCVERSIONGT4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 4)
  285. GCCVERSIONGT5 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 5)
  286. GCCVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 7)
  287. GCCVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  288. GCCVERSIONGTEQ11 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 11)
  289. GCCVERSIONGTEQ10 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 10)
  290. # Note that the behavior of -dumpversion is compile-time-configurable for
  291. # gcc-7.x and newer. Use -dumpfullversion there
  292. ifeq ($(GCCVERSIONGTEQ7),1)
  293. GCCDUMPVERSION_PARAM := -dumpfullversion
  294. else
  295. GCCDUMPVERSION_PARAM := -dumpversion
  296. endif
  297. GCCMINORVERSIONGTEQ1 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 1)
  298. GCCMINORVERSIONGTEQ2 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 2)
  299. GCCMINORVERSIONGTEQ7 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 7)
  300. endif
  301. #
  302. # OS dependent settings
  303. #
  304. ifeq ($(OSNAME), Darwin)
  305. ifndef MACOSX_DEPLOYMENT_TARGET
  306. export MACOSX_DEPLOYMENT_TARGET=10.8
  307. endif
  308. MD5SUM = md5 -r
  309. endif
  310. ifneq (,$(findstring $(OSNAME), FreeBSD OpenBSD DragonFly))
  311. MD5SUM = md5 -r
  312. endif
  313. ifeq ($(OSNAME), NetBSD)
  314. MD5SUM = md5 -n
  315. endif
  316. ifeq ($(OSNAME), Linux)
  317. EXTRALIB += -lm
  318. NO_EXPRECISION = 1
  319. endif
  320. ifeq ($(OSNAME), Android)
  321. EXTRALIB += -lm
  322. endif
  323. ifeq ($(OSNAME), AIX)
  324. EXTRALIB += -lm
  325. endif
  326. ifeq ($(OSNAME), WINNT)
  327. NEED_PIC = 0
  328. NO_EXPRECISION = 1
  329. EXTRALIB += -defaultlib:advapi32
  330. SUFFIX = obj
  331. PSUFFIX = pobj
  332. LIBSUFFIX = a
  333. ifeq ($(C_COMPILER), CLANG)
  334. CCOMMON_OPT += -DMS_ABI
  335. endif
  336. #Version tests for supporting specific features (MS_ABI, POWER9 intrinsics)
  337. ifeq ($(GCCVERSIONGT4), 1)
  338. # GCC Major version > 4
  339. # It is compatible with MSVC ABI.
  340. CCOMMON_OPT += -DMS_ABI
  341. endif
  342. ifeq ($(GCCVERSIONGTEQ4), 1)
  343. ifeq ($(GCCMINORVERSIONGTEQ7), 1)
  344. # GCC Version >=4.7
  345. # It is compatible with MSVC ABI.
  346. CCOMMON_OPT += -DMS_ABI
  347. endif
  348. endif
  349. # Ensure the correct stack alignment on Win32
  350. # http://permalink.gmane.org/gmane.comp.lib.openblas.general/97
  351. ifeq ($(ARCH), x86)
  352. CCOMMON_OPT += -mincoming-stack-boundary=2
  353. FCOMMON_OPT += -mincoming-stack-boundary=2
  354. endif
  355. endif
  356. ifeq ($(OSNAME), Interix)
  357. NEED_PIC = 0
  358. NO_EXPRECISION = 1
  359. INTERIX_TOOL_DIR = /opt/gcc.3.3/i586-pc-interix3/bin
  360. endif
  361. ifeq ($(OSNAME), CYGWIN_NT)
  362. NEED_PIC = 0
  363. NO_EXPRECISION = 1
  364. OS_CYGWIN_NT = 1
  365. endif
  366. ifneq ($(OSNAME), WINNT)
  367. ifneq ($(OSNAME), CYGWIN_NT)
  368. ifneq ($(OSNAME), Interix)
  369. ifneq ($(OSNAME), Android)
  370. ifdef SMP
  371. EXTRALIB += -lpthread
  372. endif
  373. endif
  374. endif
  375. endif
  376. endif
  377. # ifeq logical or
  378. ifeq ($(OSNAME), $(filter $(OSNAME),WINNT CYGWIN_NT Interix))
  379. OS_WINDOWS=1
  380. endif
  381. ifdef QUAD_PRECISION
  382. CCOMMON_OPT += -DQUAD_PRECISION
  383. NO_EXPRECISION = 1
  384. endif
  385. ifneq ($(ARCH), x86)
  386. ifneq ($(ARCH), x86_64)
  387. NO_EXPRECISION = 1
  388. endif
  389. endif
  390. ifdef UTEST_CHECK
  391. CCOMMON_OPT += -DUTEST_CHECK
  392. SANITY_CHECK = 1
  393. endif
  394. ifdef SANITY_CHECK
  395. CCOMMON_OPT += -DSANITY_CHECK -DREFNAME=$(*F)f$(BU)
  396. endif
  397. MAX_STACK_ALLOC ?= 2048
  398. ifneq ($(MAX_STACK_ALLOC), 0)
  399. CCOMMON_OPT += -DMAX_STACK_ALLOC=$(MAX_STACK_ALLOC)
  400. endif
  401. ifdef USE_LOCKING
  402. ifneq ($(USE_LOCKING), 0)
  403. CCOMMON_OPT += -DUSE_LOCKING
  404. endif
  405. endif
  406. #
  407. # Architecture dependent settings
  408. #
  409. ifeq ($(ARCH), x86)
  410. ifndef BINARY
  411. NO_BINARY_MODE = 1
  412. endif
  413. ifeq ($(CORE), generic)
  414. NO_EXPRECISION = 1
  415. endif
  416. ifndef NO_EXPRECISION
  417. ifeq ($(F_COMPILER), GFORTRAN)
  418. # ifeq logical or. GCC or LSB
  419. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  420. EXPRECISION = 1
  421. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  422. FCOMMON_OPT += -m128bit-long-double
  423. endif
  424. ifeq ($(C_COMPILER), CLANG)
  425. EXPRECISION = 1
  426. CCOMMON_OPT += -DEXPRECISION
  427. FCOMMON_OPT += -m128bit-long-double
  428. endif
  429. endif
  430. endif
  431. endif
  432. ifeq ($(ARCH), x86_64)
  433. ifeq ($(CORE), generic)
  434. NO_EXPRECISION = 1
  435. endif
  436. ifndef NO_EXPRECISION
  437. ifeq ($(F_COMPILER), GFORTRAN)
  438. # ifeq logical or. GCC or LSB
  439. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  440. EXPRECISION = 1
  441. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  442. FCOMMON_OPT += -m128bit-long-double
  443. endif
  444. ifeq ($(C_COMPILER), CLANG)
  445. EXPRECISION = 1
  446. CCOMMON_OPT += -DEXPRECISION
  447. FCOMMON_OPT += -m128bit-long-double
  448. endif
  449. endif
  450. endif
  451. endif
  452. ifeq ($(C_COMPILER), INTEL)
  453. CCOMMON_OPT += -wd981
  454. endif
  455. ifeq ($(USE_OPENMP), 1)
  456. #check
  457. ifeq ($(USE_THREAD), 0)
  458. $(error OpenBLAS: Cannot set both USE_OPENMP=1 and USE_THREAD=0. The USE_THREAD=0 is only for building single thread version.)
  459. endif
  460. # ifeq logical or. GCC or LSB
  461. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  462. CCOMMON_OPT += -fopenmp
  463. endif
  464. ifeq ($(C_COMPILER), CLANG)
  465. CCOMMON_OPT += -fopenmp
  466. endif
  467. ifeq ($(C_COMPILER), INTEL)
  468. CCOMMON_OPT += -fopenmp
  469. endif
  470. ifeq ($(C_COMPILER), PGI)
  471. CCOMMON_OPT += -mp
  472. endif
  473. ifeq ($(C_COMPILER), OPEN64)
  474. CCOMMON_OPT += -mp
  475. CEXTRALIB += -lstdc++
  476. endif
  477. ifeq ($(C_COMPILER), PATHSCALE)
  478. CCOMMON_OPT += -mp
  479. endif
  480. endif
  481. ifeq ($(DYNAMIC_ARCH), 1)
  482. ifeq ($(ARCH), x86)
  483. DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
  484. CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
  485. endif
  486. ifeq ($(ARCH), x86_64)
  487. DYNAMIC_CORE = PRESCOTT CORE2
  488. ifeq ($(DYNAMIC_OLDER), 1)
  489. DYNAMIC_CORE += PENRYN DUNNINGTON
  490. endif
  491. DYNAMIC_CORE += NEHALEM
  492. ifeq ($(DYNAMIC_OLDER), 1)
  493. DYNAMIC_CORE += OPTERON OPTERON_SSE3
  494. endif
  495. DYNAMIC_CORE += BARCELONA
  496. ifeq ($(DYNAMIC_OLDER), 1)
  497. DYNAMIC_CORE += BOBCAT ATOM NANO
  498. endif
  499. ifneq ($(NO_AVX), 1)
  500. DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER STEAMROLLER EXCAVATOR
  501. endif
  502. ifneq ($(NO_AVX2), 1)
  503. DYNAMIC_CORE += HASWELL ZEN
  504. endif
  505. ifneq ($(NO_AVX512), 1)
  506. ifneq ($(NO_AVX2), 1)
  507. DYNAMIC_CORE += SKYLAKEX COOPERLAKE
  508. endif
  509. endif
  510. endif
  511. ifdef DYNAMIC_LIST
  512. override DYNAMIC_CORE = PRESCOTT $(DYNAMIC_LIST)
  513. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_PRESCOTT
  514. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  515. CCOMMON_OPT += $(XCCOMMON_OPT)
  516. #CCOMMON_OPT += -DDYNAMIC_LIST='$(DYNAMIC_LIST)'
  517. endif
  518. ifeq ($(ARCH), arm64)
  519. DYNAMIC_CORE = ARMV8
  520. DYNAMIC_CORE += CORTEXA53
  521. DYNAMIC_CORE += CORTEXA57
  522. DYNAMIC_CORE += CORTEXA72
  523. DYNAMIC_CORE += CORTEXA73
  524. DYNAMIC_CORE += NEOVERSEN1
  525. DYNAMIC_CORE += FALKOR
  526. DYNAMIC_CORE += THUNDERX
  527. DYNAMIC_CORE += THUNDERX2T99
  528. DYNAMIC_CORE += TSV110
  529. DYNAMIC_CORE += EMAG8180
  530. DYNAMIC_CORE += THUNDERX3T110
  531. endif
  532. ifeq ($(ARCH), zarch)
  533. DYNAMIC_CORE = ZARCH_GENERIC
  534. # if the compiler accepts -march=arch11 or -march=z13 and can compile a file
  535. # with z13-specific inline assembly, then we can include support for Z13.
  536. # note: -march=z13 is equivalent to -march=arch11 yet some compiler releases
  537. # only support one or the other.
  538. # note: LLVM version 6.x supported -march=z13 yet could not handle vector
  539. # registers in inline assembly, so the check for supporting the -march flag is
  540. # not enough.
  541. ZARCH_TEST_COMPILE=-c $(TOPDIR)/kernel/zarch/damin_z13.c -I$(TOPDIR) -o /dev/null > /dev/null 2> /dev/null
  542. ZARCH_CC_SUPPORTS_ARCH11=$(shell $(CC) -march=arch11 $(ZARCH_TEST_COMPILE) && echo 1)
  543. ZARCH_CC_SUPPORTS_Z13=$(shell $(CC) -march=z13 $(ZARCH_TEST_COMPILE) && echo 1)
  544. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH11), $(ZARCH_CC_SUPPORTS_Z13)), 1)
  545. DYNAMIC_CORE += Z13
  546. CCOMMON_OPT += -DDYN_Z13
  547. else
  548. $(info OpenBLAS: Not building Z13 kernels because the compiler $(CC) does not support it)
  549. endif
  550. # as above for z13, check for -march=arch12 and z14 support in the compiler.
  551. ZARCH_CC_SUPPORTS_ARCH12=$(shell $(CC) -march=arch12 $(ZARCH_TEST_COMPILE) && echo 1)
  552. ZARCH_CC_SUPPORTS_Z14=$(shell $(CC) -march=z14 $(ZARCH_TEST_COMPILE) && echo 1)
  553. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH12), $(ZARCH_CC_SUPPORTS_Z14)), 1)
  554. DYNAMIC_CORE += Z14
  555. CCOMMON_OPT += -DDYN_Z14
  556. else
  557. $(info OpenBLAS: Not building Z14 kernels because the compiler $(CC) does not support it)
  558. endif
  559. endif # ARCH zarch
  560. ifeq ($(ARCH), power)
  561. DYNAMIC_CORE = POWER6
  562. DYNAMIC_CORE += POWER8
  563. ifneq ($(C_COMPILER), GCC)
  564. DYNAMIC_CORE += POWER9
  565. DYNAMIC_CORE += POWER10
  566. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  567. endif
  568. ifeq ($(C_COMPILER), GCC)
  569. ifeq ($(GCCVERSIONGT5), 1)
  570. DYNAMIC_CORE += POWER9
  571. else
  572. $(info, OpenBLAS: Your gcc version is too old to build the POWER9 kernels.)
  573. endif
  574. LDVERSIONGTEQ35 := $(shell expr `ld --version | head -1 | cut -f2 -d "." | cut -f1 -d "-"` >= 35)
  575. ifeq ($(GCCVERSIONGTEQ11)$(LDVERSIONGTEQ35), 11)
  576. DYNAMIC_CORE += POWER10
  577. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  578. else ifeq ($(GCCVERSIONGTEQ10), 1)
  579. ifeq ($(GCCMINORVERSIONGTEQ2)$(LDVERSIONGTEQ35), 11)
  580. DYNAMIC_CORE += POWER10
  581. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  582. endif
  583. else
  584. $(info, OpenBLAS: Your gcc version is too old to build the POWER10 kernels.)
  585. endif
  586. endif
  587. endif
  588. # If DYNAMIC_CORE is not set, DYNAMIC_ARCH cannot do anything, so force it to empty
  589. ifndef DYNAMIC_CORE
  590. override DYNAMIC_ARCH=
  591. endif
  592. endif
  593. ifeq ($(ARCH), ia64)
  594. NO_BINARY_MODE = 1
  595. BINARY_DEFINED = 1
  596. ifeq ($(F_COMPILER), GFORTRAN)
  597. ifeq ($(C_COMPILER), GCC)
  598. # EXPRECISION = 1
  599. # CCOMMON_OPT += -DEXPRECISION
  600. endif
  601. endif
  602. endif
  603. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  604. NO_BINARY_MODE = 1
  605. endif
  606. ifeq ($(ARCH), alpha)
  607. NO_BINARY_MODE = 1
  608. BINARY_DEFINED = 1
  609. endif
  610. ifeq ($(ARCH), arm)
  611. NO_BINARY_MODE = 1
  612. BINARY_DEFINED = 1
  613. CCOMMON_OPT += -marm
  614. FCOMMON_OPT += -marm
  615. # If softfp abi is mentioned on the command line, force it.
  616. ifeq ($(ARM_SOFTFP_ABI), 1)
  617. CCOMMON_OPT += -mfloat-abi=softfp
  618. FCOMMON_OPT += -mfloat-abi=softfp
  619. endif
  620. ifeq ($(OSNAME), Android)
  621. ifeq ($(ARM_SOFTFP_ABI), 1)
  622. EXTRALIB += -lm
  623. else
  624. EXTRALIB += -Wl,-lm_hard
  625. endif
  626. endif
  627. endif
  628. ifeq ($(ARCH), arm64)
  629. NO_BINARY_MODE = 1
  630. BINARY_DEFINED = 1
  631. ifdef INTERFACE64
  632. ifneq ($(INTERFACE64), 0)
  633. ifeq ($(F_COMPILER), GFORTRAN)
  634. FCOMMON_OPT += -fdefault-integer-8
  635. endif
  636. ifeq ($(F_COMPILER), FLANG)
  637. FCOMMON_OPT += -i8
  638. endif
  639. endif
  640. endif
  641. endif
  642. ifeq ($(ARCH), riscv64)
  643. NO_BINARY_MODE = 1
  644. BINARY_DEFINED = 1
  645. endif
  646. #
  647. # C Compiler dependent settings
  648. #
  649. # ifeq logical or. GCC or CLANG or LSB
  650. # http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
  651. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG LSB))
  652. CCOMMON_OPT += -Wall
  653. COMMON_PROF += -fno-inline
  654. NO_UNINITIALIZED_WARN = -Wno-uninitialized
  655. ifeq ($(QUIET_MAKE), 1)
  656. CCOMMON_OPT += $(NO_UNINITIALIZED_WARN) -Wno-unused
  657. endif
  658. ifdef NO_BINARY_MODE
  659. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  660. ifdef BINARY64
  661. CCOMMON_OPT += -mabi=64
  662. else
  663. CCOMMON_OPT += -mabi=n32
  664. endif
  665. BINARY_DEFINED = 1
  666. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  667. CCOMMON_OPT += -mabi=32
  668. BINARY_DEFINED = 1
  669. endif
  670. ifeq ($(CORE), LOONGSON3A)
  671. CCOMMON_OPT += -march=mips64
  672. FCOMMON_OPT += -march=mips64
  673. endif
  674. ifeq ($(CORE), LOONGSON3B)
  675. CCOMMON_OPT += -march=mips64
  676. FCOMMON_OPT += -march=mips64
  677. endif
  678. ifeq ($(CORE), MIPS24K)
  679. CCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  680. FCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  681. endif
  682. ifeq ($(CORE), MIPS1004K)
  683. CCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  684. FCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  685. endif
  686. ifeq ($(CORE), P5600)
  687. CCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  688. FCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  689. endif
  690. ifeq ($(CORE), I6400)
  691. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  692. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  693. endif
  694. ifeq ($(CORE), P6600)
  695. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  696. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  697. endif
  698. ifeq ($(CORE), I6500)
  699. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  700. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  701. endif
  702. ifeq ($(OSNAME), AIX)
  703. BINARY_DEFINED = 1
  704. endif
  705. endif
  706. ifndef BINARY_DEFINED
  707. ifneq ($(OSNAME), AIX)
  708. ifdef BINARY64
  709. ifneq ($(ARCH), riscv64)
  710. CCOMMON_OPT += -m64
  711. endif
  712. else
  713. CCOMMON_OPT += -m32
  714. endif
  715. endif
  716. endif
  717. endif
  718. ifeq ($(C_COMPILER), PGI)
  719. ifdef BINARY64
  720. ifeq ($(ARCH), x86_64)
  721. CCOMMON_OPT += -tp p7-64 -D__MMX__ -Mnollvm
  722. else
  723. ifeq ($(ARCH), power)
  724. ifeq ($(CORE), POWER8)
  725. CCOMMON_OPT += -tp pwr8
  726. endif
  727. ifeq ($(CORE), POWER9)
  728. CCOMMON_OPT += -tp pwr9
  729. endif
  730. endif
  731. endif
  732. else
  733. CCOMMON_OPT += -tp p7
  734. endif
  735. endif
  736. ifeq ($(C_COMPILER), PATHSCALE)
  737. ifdef BINARY64
  738. CCOMMON_OPT += -m64
  739. else
  740. CCOMMON_OPT += -m32
  741. endif
  742. endif
  743. #
  744. # Fortran Compiler dependent settings
  745. #
  746. ifeq ($(F_COMPILER), FLANG)
  747. CCOMMON_OPT += -DF_INTERFACE_FLANG
  748. FCOMMON_OPT += -Mrecursive -Kieee
  749. ifeq ($(OSNAME), Linux)
  750. ifeq ($(ARCH), x86_64)
  751. FLANG_VENDOR := $(shell `$(FC) --version|cut -f 1 -d "."|head -1`)
  752. ifeq ($(FLANG_VENDOR),AOCC)
  753. FCOMMON_OPT += -fno-unroll-loops
  754. endif
  755. endif
  756. endif
  757. ifdef BINARY64
  758. ifdef INTERFACE64
  759. ifneq ($(INTERFACE64), 0)
  760. FCOMMON_OPT += -i8
  761. endif
  762. endif
  763. FCOMMON_OPT += -Wall
  764. else
  765. FCOMMON_OPT += -Wall
  766. endif
  767. ifeq ($(USE_OPENMP), 1)
  768. FCOMMON_OPT += -fopenmp
  769. endif
  770. endif
  771. ifeq ($(F_COMPILER), G77)
  772. CCOMMON_OPT += -DF_INTERFACE_G77
  773. FCOMMON_OPT += -Wall
  774. ifndef NO_BINARY_MODE
  775. ifneq ($(OSNAME), AIX)
  776. ifdef BINARY64
  777. FCOMMON_OPT += -m64
  778. else
  779. FCOMMON_OPT += -m32
  780. endif
  781. endif
  782. endif
  783. endif
  784. ifeq ($(F_COMPILER), G95)
  785. CCOMMON_OPT += -DF_INTERFACE_G95
  786. FCOMMON_OPT += -Wall
  787. ifneq ($(OSNAME), AIX)
  788. ifndef NO_BINARY_MODE
  789. ifdef BINARY64
  790. FCOMMON_OPT += -m64
  791. else
  792. FCOMMON_OPT += -m32
  793. endif
  794. endif
  795. ifneq ($(NO_LAPACKE), 1)
  796. FCOMMON_OPT += -fno-second-underscore
  797. endif
  798. endif
  799. endif
  800. ifeq ($(F_COMPILER), GFORTRAN)
  801. CCOMMON_OPT += -DF_INTERFACE_GFORT
  802. FCOMMON_OPT += -Wall
  803. # make single-threaded LAPACK calls thread-safe #1847
  804. FCOMMON_OPT += -frecursive
  805. # work around ABI problem with passing single-character arguments
  806. FCOMMON_OPT += -fno-optimize-sibling-calls
  807. #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
  808. ifneq ($(NO_LAPACK), 1)
  809. EXTRALIB += -lgfortran
  810. endif
  811. ifdef NO_BINARY_MODE
  812. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  813. ifdef BINARY64
  814. FCOMMON_OPT += -mabi=64
  815. else
  816. FCOMMON_OPT += -mabi=n32
  817. endif
  818. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  819. FCOMMON_OPT += -mabi=32
  820. endif
  821. else
  822. ifdef BINARY64
  823. ifneq ($(OSNAME), AIX)
  824. ifneq ($(ARCH), riscv64)
  825. FCOMMON_OPT += -m64
  826. endif
  827. endif
  828. ifdef INTERFACE64
  829. ifneq ($(INTERFACE64), 0)
  830. FCOMMON_OPT += -fdefault-integer-8
  831. endif
  832. endif
  833. else
  834. ifneq ($(OSNAME), AIX)
  835. FCOMMON_OPT += -m32
  836. endif
  837. endif
  838. endif
  839. ifeq ($(USE_OPENMP), 1)
  840. FCOMMON_OPT += -fopenmp
  841. endif
  842. endif
  843. ifeq ($(F_COMPILER), INTEL)
  844. CCOMMON_OPT += -DF_INTERFACE_INTEL
  845. ifdef INTERFACE64
  846. ifneq ($(INTERFACE64), 0)
  847. FCOMMON_OPT += -i8
  848. endif
  849. endif
  850. FCOMMON_OPT += -recursive -fp-model strict -assume protect-parens
  851. ifeq ($(USE_OPENMP), 1)
  852. FCOMMON_OPT += -fopenmp
  853. endif
  854. endif
  855. ifeq ($(F_COMPILER), FUJITSU)
  856. CCOMMON_OPT += -DF_INTERFACE_FUJITSU
  857. ifeq ($(USE_OPENMP), 1)
  858. FCOMMON_OPT += -openmp
  859. endif
  860. endif
  861. ifeq ($(F_COMPILER), IBM)
  862. CCOMMON_OPT += -DF_INTERFACE_IBM
  863. # FCOMMON_OPT += -qarch=440
  864. ifdef BINARY64
  865. FCOMMON_OPT += -q64
  866. ifdef INTERFACE64
  867. ifneq ($(INTERFACE64), 0)
  868. FCOMMON_OPT += -qintsize=8
  869. endif
  870. endif
  871. else
  872. FCOMMON_OPT += -q32
  873. endif
  874. ifeq ($(USE_OPENMP), 1)
  875. FCOMMON_OPT += -openmp
  876. endif
  877. endif
  878. ifeq ($(F_COMPILER), PGI)
  879. CCOMMON_OPT += -DF_INTERFACE_PGI
  880. COMMON_PROF += -DPGICOMPILER
  881. ifdef BINARY64
  882. ifdef INTERFACE64
  883. ifneq ($(INTERFACE64), 0)
  884. FCOMMON_OPT += -i8
  885. endif
  886. endif
  887. ifeq ($(ARCH), x86_64)
  888. FCOMMON_OPT += -tp p7-64
  889. else
  890. ifeq ($(ARCH), power)
  891. ifeq ($(CORE), POWER8)
  892. FCOMMON_OPT += -tp pwr8
  893. endif
  894. ifeq ($(CORE), POWER9)
  895. FCOMMON_OPT += -tp pwr9
  896. endif
  897. endif
  898. endif
  899. else
  900. FCOMMON_OPT += -tp p7
  901. endif
  902. FCOMMON_OPT += -Mrecursive
  903. ifeq ($(USE_OPENMP), 1)
  904. FCOMMON_OPT += -mp
  905. endif
  906. endif
  907. ifeq ($(F_COMPILER), PATHSCALE)
  908. CCOMMON_OPT += -DF_INTERFACE_PATHSCALE
  909. ifdef BINARY64
  910. ifdef INTERFACE64
  911. ifneq ($(INTERFACE64), 0)
  912. FCOMMON_OPT += -i8
  913. endif
  914. endif
  915. endif
  916. ifeq ($(USE_OPENMP), 1)
  917. FCOMMON_OPT += -mp
  918. endif
  919. endif
  920. ifeq ($(F_COMPILER), OPEN64)
  921. CCOMMON_OPT += -DF_INTERFACE_OPEN64
  922. ifdef BINARY64
  923. ifdef INTERFACE64
  924. ifneq ($(INTERFACE64), 0)
  925. FCOMMON_OPT += -i8
  926. endif
  927. endif
  928. endif
  929. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  930. ifndef BINARY64
  931. FCOMMON_OPT += -n32
  932. else
  933. FCOMMON_OPT += -n64
  934. endif
  935. ifeq ($(CORE), LOONGSON3A)
  936. FCOMMON_OPT += -loongson3 -static
  937. endif
  938. ifeq ($(CORE), LOONGSON3B)
  939. FCOMMON_OPT += -loongson3 -static
  940. endif
  941. else
  942. ifndef BINARY64
  943. FCOMMON_OPT += -m32
  944. else
  945. FCOMMON_OPT += -m64
  946. endif
  947. endif
  948. ifeq ($(USE_OPENMP), 1)
  949. FEXTRALIB += -lstdc++
  950. FCOMMON_OPT += -mp
  951. endif
  952. endif
  953. ifeq ($(C_COMPILER), OPEN64)
  954. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  955. ifndef BINARY64
  956. CCOMMON_OPT += -n32
  957. else
  958. CCOMMON_OPT += -n64
  959. endif
  960. ifeq ($(CORE), LOONGSON3A)
  961. CCOMMON_OPT += -loongson3 -static
  962. endif
  963. ifeq ($(CORE), LOONGSON3B)
  964. CCOMMON_OPT += -loongson3 -static
  965. endif
  966. else
  967. ifndef BINARY64
  968. CCOMMON_OPT += -m32
  969. else
  970. CCOMMON_OPT += -m64
  971. endif
  972. endif
  973. endif
  974. ifeq ($(C_COMPILER), SUN)
  975. CCOMMON_OPT += -w
  976. ifeq ($(ARCH), x86)
  977. CCOMMON_OPT += -m32
  978. else
  979. FCOMMON_OPT += -m64
  980. endif
  981. endif
  982. ifeq ($(F_COMPILER), SUN)
  983. CCOMMON_OPT += -DF_INTERFACE_SUN
  984. ifeq ($(ARCH), x86)
  985. FCOMMON_OPT += -m32
  986. else
  987. FCOMMON_OPT += -m64
  988. endif
  989. ifeq ($(USE_OPENMP), 1)
  990. FCOMMON_OPT += -xopenmp=parallel
  991. endif
  992. endif
  993. ifeq ($(F_COMPILER), COMPAQ)
  994. CCOMMON_OPT += -DF_INTERFACE_COMPAQ
  995. ifeq ($(USE_OPENMP), 1)
  996. FCOMMON_OPT += -openmp
  997. endif
  998. endif
  999. ifdef BINARY64
  1000. ifdef INTERFACE64
  1001. ifneq ($(INTERFACE64), 0)
  1002. CCOMMON_OPT +=
  1003. #-DUSE64BITINT
  1004. endif
  1005. endif
  1006. endif
  1007. ifeq ($(NEED_PIC), 1)
  1008. ifeq ($(C_COMPILER), IBM)
  1009. CCOMMON_OPT += -qpic=large
  1010. else
  1011. CCOMMON_OPT += -fPIC
  1012. endif
  1013. ifeq ($(F_COMPILER), SUN)
  1014. FCOMMON_OPT += -pic
  1015. else
  1016. FCOMMON_OPT += -fPIC
  1017. endif
  1018. endif
  1019. ifeq ($(DYNAMIC_ARCH), 1)
  1020. CCOMMON_OPT += -DDYNAMIC_ARCH
  1021. endif
  1022. ifeq ($(DYNAMIC_OLDER), 1)
  1023. CCOMMON_OPT += -DDYNAMIC_OLDER
  1024. endif
  1025. ifeq ($(NO_LAPACK), 1)
  1026. CCOMMON_OPT += -DNO_LAPACK
  1027. #Disable LAPACK C interface
  1028. NO_LAPACKE = 1
  1029. endif
  1030. ifeq ($(NO_LAPACKE), 1)
  1031. CCOMMON_OPT += -DNO_LAPACKE
  1032. endif
  1033. ifeq ($(NO_AVX), 1)
  1034. CCOMMON_OPT += -DNO_AVX
  1035. endif
  1036. ifeq ($(ARCH), x86)
  1037. CCOMMON_OPT += -DNO_AVX
  1038. endif
  1039. ifeq ($(NO_AVX2), 1)
  1040. CCOMMON_OPT += -DNO_AVX2
  1041. endif
  1042. ifeq ($(NO_AVX512), 1)
  1043. CCOMMON_OPT += -DNO_AVX512
  1044. endif
  1045. ifdef SMP
  1046. CCOMMON_OPT += -DSMP_SERVER
  1047. ifeq ($(ARCH), mips64)
  1048. ifneq ($(CORE), LOONGSON3B)
  1049. USE_SIMPLE_THREADED_LEVEL3 = 1
  1050. endif
  1051. endif
  1052. ifeq ($(USE_OPENMP), 1)
  1053. # USE_SIMPLE_THREADED_LEVEL3 = 1
  1054. # NO_AFFINITY = 1
  1055. CCOMMON_OPT += -DUSE_OPENMP
  1056. endif
  1057. ifeq ($(BIGNUMA), 1)
  1058. CCOMMON_OPT += -DBIGNUMA
  1059. endif
  1060. endif
  1061. ifeq ($(NO_WARMUP), 1)
  1062. CCOMMON_OPT += -DNO_WARMUP
  1063. endif
  1064. ifeq ($(CONSISTENT_FPCSR), 1)
  1065. CCOMMON_OPT += -DCONSISTENT_FPCSR
  1066. endif
  1067. # Only for development
  1068. # CCOMMON_OPT += -DPARAMTEST
  1069. # CCOMMON_OPT += -DPREFETCHTEST
  1070. # CCOMMON_OPT += -DNO_SWITCHING
  1071. # USE_PAPI = 1
  1072. ifdef USE_PAPI
  1073. CCOMMON_OPT += -DUSE_PAPI
  1074. EXTRALIB += -lpapi -lperfctr
  1075. endif
  1076. ifdef DYNAMIC_THREADS
  1077. CCOMMON_OPT += -DDYNAMIC_THREADS
  1078. endif
  1079. CCOMMON_OPT += -DMAX_CPU_NUMBER=$(NUM_THREADS)
  1080. CCOMMON_OPT += -DMAX_PARALLEL_NUMBER=$(NUM_PARALLEL)
  1081. ifdef USE_SIMPLE_THREADED_LEVEL3
  1082. CCOMMON_OPT += -DUSE_SIMPLE_THREADED_LEVEL3
  1083. endif
  1084. ifeq ($(USE_TLS), 1)
  1085. CCOMMON_OPT += -DUSE_TLS
  1086. endif
  1087. ifeq ($(BUILD_BFLOAT16), 1)
  1088. CCOMMON_OPT += -DBUILD_BFLOAT16
  1089. endif
  1090. ifeq ($(BUILD_SINGLE), 1)
  1091. CCOMMON_OPT += -DBUILD_SINGLE=1
  1092. endif
  1093. ifeq ($(BUILD_DOUBLE), 1)
  1094. CCOMMON_OPT += -DBUILD_DOUBLE=1
  1095. endif
  1096. ifeq ($(BUILD_COMPLEX), 1)
  1097. CCOMMON_OPT += -DBUILD_COMPLEX=1
  1098. endif
  1099. ifeq ($(BUILD_COMPLEX16), 1)
  1100. CCOMMON_OPT += -DBUILD_COMPLEX16=1
  1101. endif
  1102. CCOMMON_OPT += -DVERSION=\"$(VERSION)\"
  1103. ifndef SYMBOLPREFIX
  1104. SYMBOLPREFIX =
  1105. endif
  1106. ifndef SYMBOLSUFFIX
  1107. SYMBOLSUFFIX =
  1108. endif
  1109. ifndef LIBSONAMEBASE
  1110. LIBSONAMEBASE = openblas
  1111. endif
  1112. ifndef LIBNAMESUFFIX
  1113. LIBNAMEBASE = $(SYMBOLPREFIX)$(LIBSONAMEBASE)$(SYMBOLSUFFIX)
  1114. else
  1115. LIBNAMEBASE = $(SYMBOLPREFIX)$(LIBSONAMEBASE)$(SYMBOLSUFFIX)_$(LIBNAMESUFFIX)
  1116. endif
  1117. ifeq ($(OSNAME), CYGWIN_NT)
  1118. LIBPREFIX = cyg$(LIBNAMEBASE)
  1119. else
  1120. LIBPREFIX = lib$(LIBNAMEBASE)
  1121. endif
  1122. KERNELDIR = $(TOPDIR)/kernel/$(ARCH)
  1123. include $(TOPDIR)/Makefile.$(ARCH)
  1124. ifneq ($(C_COMPILER), PGI)
  1125. CCOMMON_OPT += -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME
  1126. endif
  1127. CCOMMON_OPT += -DASMNAME=$(FU)$(*F) -DASMFNAME=$(FU)$(*F)$(BU) -DNAME=$(*F)$(BU) -DCNAME=$(*F) -DCHAR_NAME=\"$(*F)$(BU)\" -DCHAR_CNAME=\"$(*F)\"
  1128. ifeq ($(CORE), PPC440)
  1129. CCOMMON_OPT += -DALLOC_QALLOC
  1130. endif
  1131. ifeq ($(CORE), PPC440FP2)
  1132. STATIC_ALLOCATION = 1
  1133. endif
  1134. ifneq ($(OSNAME), Linux)
  1135. NO_AFFINITY = 1
  1136. endif
  1137. ifneq ($(ARCH), x86_64)
  1138. ifneq ($(ARCH), x86)
  1139. ifneq ($(CORE), LOONGSON3B)
  1140. NO_AFFINITY = 1
  1141. endif
  1142. endif
  1143. endif
  1144. ifdef NO_AFFINITY
  1145. ifeq ($(NO_AFFINITY), 0)
  1146. override undefine NO_AFFINITY
  1147. else
  1148. CCOMMON_OPT += -DNO_AFFINITY
  1149. endif
  1150. endif
  1151. ifdef FUNCTION_PROFILE
  1152. CCOMMON_OPT += -DFUNCTION_PROFILE
  1153. endif
  1154. ifdef HUGETLB_ALLOCATION
  1155. CCOMMON_OPT += -DALLOC_HUGETLB
  1156. endif
  1157. ifdef HUGETLBFILE_ALLOCATION
  1158. CCOMMON_OPT += -DALLOC_HUGETLBFILE -DHUGETLB_FILE_NAME=$(HUGETLBFILE_ALLOCATION)
  1159. endif
  1160. ifdef STATIC_ALLOCATION
  1161. CCOMMON_OPT += -DALLOC_STATIC
  1162. endif
  1163. ifdef DEVICEDRIVER_ALLOCATION
  1164. CCOMMON_OPT += -DALLOC_DEVICEDRIVER -DDEVICEDRIVER_NAME=\"/dev/mapper\"
  1165. endif
  1166. ifdef MIXED_MEMORY_ALLOCATION
  1167. CCOMMON_OPT += -DMIXED_MEMORY_ALLOCATION
  1168. endif
  1169. ifeq ($(OSNAME), SunOS)
  1170. TAR = gtar
  1171. PATCH = gpatch
  1172. GREP = ggrep
  1173. AWK = nawk
  1174. else
  1175. TAR = tar
  1176. PATCH = patch
  1177. GREP = grep
  1178. AWK = awk
  1179. endif
  1180. ifndef MD5SUM
  1181. MD5SUM = md5sum
  1182. endif
  1183. REVISION = -r$(VERSION)
  1184. MAJOR_VERSION = $(word 1,$(subst ., ,$(VERSION)))
  1185. ifeq ($(DEBUG), 1)
  1186. COMMON_OPT += -g
  1187. endif
  1188. ifeq ($(DEBUG), 1)
  1189. FCOMMON_OPT += -g
  1190. endif
  1191. ifndef COMMON_OPT
  1192. COMMON_OPT = -O2
  1193. endif
  1194. ifndef FCOMMON_OPT
  1195. FCOMMON_OPT = -O2 -frecursive
  1196. endif
  1197. override CFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
  1198. override PFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
  1199. override FFLAGS += $(COMMON_OPT) $(FCOMMON_OPT)
  1200. override FPFLAGS += $(FCOMMON_OPT) $(COMMON_PROF)
  1201. #MAKEOVERRIDES =
  1202. ifdef NEED_PIC
  1203. ifeq (,$(findstring PIC,$(FFLAGS)))
  1204. override FFLAGS += -fPIC
  1205. endif
  1206. endif
  1207. #For LAPACK Fortran codes.
  1208. #Disable -fopenmp for LAPACK Fortran codes on Windows.
  1209. ifdef OS_WINDOWS
  1210. LAPACK_FFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FFLAGS))
  1211. LAPACK_FPFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FPFLAGS))
  1212. else
  1213. LAPACK_FFLAGS := $(FFLAGS)
  1214. LAPACK_FPFLAGS := $(FPFLAGS)
  1215. endif
  1216. LAPACK_CFLAGS = $(CFLAGS)
  1217. LAPACK_CFLAGS += -DHAVE_LAPACK_CONFIG_H
  1218. ifdef INTERFACE64
  1219. ifneq ($(INTERFACE64), 0)
  1220. LAPACK_CFLAGS += -DLAPACK_ILP64
  1221. endif
  1222. endif
  1223. ifdef OS_WINDOWS
  1224. LAPACK_CFLAGS += -DOPENBLAS_OS_WINDOWS
  1225. endif
  1226. ifeq ($(C_COMPILER), LSB)
  1227. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1228. endif
  1229. ifndef SUFFIX
  1230. SUFFIX = o
  1231. endif
  1232. ifndef PSUFFIX
  1233. PSUFFIX = po
  1234. endif
  1235. ifndef LIBSUFFIX
  1236. LIBSUFFIX = a
  1237. endif
  1238. ifneq ($(DYNAMIC_ARCH), 1)
  1239. ifndef SMP
  1240. LIBNAME = $(LIBPREFIX)_$(LIBCORE)$(REVISION).$(LIBSUFFIX)
  1241. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)$(REVISION)_p.$(LIBSUFFIX)
  1242. else
  1243. LIBNAME = $(LIBPREFIX)_$(LIBCORE)p$(REVISION).$(LIBSUFFIX)
  1244. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)p$(REVISION)_p.$(LIBSUFFIX)
  1245. endif
  1246. else
  1247. ifndef SMP
  1248. LIBNAME = $(LIBPREFIX)$(REVISION).$(LIBSUFFIX)
  1249. LIBNAME_P = $(LIBPREFIX)$(REVISION)_p.$(LIBSUFFIX)
  1250. else
  1251. LIBNAME = $(LIBPREFIX)p$(REVISION).$(LIBSUFFIX)
  1252. LIBNAME_P = $(LIBPREFIX)p$(REVISION)_p.$(LIBSUFFIX)
  1253. endif
  1254. endif
  1255. LIBDLLNAME = $(LIBPREFIX).dll
  1256. IMPLIBNAME = lib$(LIBNAMEBASE).dll.a
  1257. ifneq ($(OSNAME), AIX)
  1258. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.so)
  1259. else
  1260. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.a)
  1261. endif
  1262. LIBDYNNAME = $(LIBNAME:.$(LIBSUFFIX)=.dylib)
  1263. LIBDEFNAME = $(LIBNAME:.$(LIBSUFFIX)=.def)
  1264. LIBEXPNAME = $(LIBNAME:.$(LIBSUFFIX)=.exp)
  1265. LIBZIPNAME = $(LIBNAME:.$(LIBSUFFIX)=.zip)
  1266. LIBS = $(TOPDIR)/$(LIBNAME)
  1267. LIBS_P = $(TOPDIR)/$(LIBNAME_P)
  1268. LIB_COMPONENTS = BLAS
  1269. ifneq ($(NO_CBLAS), 1)
  1270. LIB_COMPONENTS += CBLAS
  1271. endif
  1272. ifneq ($(NO_LAPACK), 1)
  1273. LIB_COMPONENTS += LAPACK
  1274. ifneq ($(NO_LAPACKE), 1)
  1275. LIB_COMPONENTS += LAPACKE
  1276. endif
  1277. ifeq ($(BUILD_RELAPACK), 1)
  1278. LIB_COMPONENTS += ReLAPACK
  1279. endif
  1280. endif
  1281. ifeq ($(ONLY_CBLAS), 1)
  1282. LIB_COMPONENTS = CBLAS
  1283. endif
  1284. export OSNAME
  1285. export ARCH
  1286. export CORE
  1287. export LIBCORE
  1288. export __BYTE_ORDER__
  1289. export ELF_VERSION
  1290. export PGCPATH
  1291. export CONFIG
  1292. export CC
  1293. export FC
  1294. export BU
  1295. export FU
  1296. export NEED2UNDERSCORES
  1297. export USE_THREAD
  1298. export NUM_THREADS
  1299. export NUM_CORES
  1300. export SMP
  1301. export MAKEFILE_RULE
  1302. export NEED_PIC
  1303. export BINARY
  1304. export BINARY32
  1305. export BINARY64
  1306. export F_COMPILER
  1307. export C_COMPILER
  1308. export USE_OPENMP
  1309. export CROSS
  1310. export CROSS_SUFFIX
  1311. export NOFORTRAN
  1312. export NO_FBLAS
  1313. export EXTRALIB
  1314. export CEXTRALIB
  1315. export FEXTRALIB
  1316. export HAVE_SSE
  1317. export HAVE_SSE2
  1318. export HAVE_SSE3
  1319. export HAVE_SSSE3
  1320. export HAVE_SSE4_1
  1321. export HAVE_SSE4_2
  1322. export HAVE_SSE4A
  1323. export HAVE_SSE5
  1324. export HAVE_AVX
  1325. export HAVE_AVX2
  1326. export HAVE_FMA3
  1327. export HAVE_VFP
  1328. export HAVE_VFPV3
  1329. export HAVE_VFPV4
  1330. export HAVE_NEON
  1331. export HAVE_MSA
  1332. export MSA_FLAGS
  1333. export KERNELDIR
  1334. export FUNCTION_PROFILE
  1335. export TARGET_CORE
  1336. export NO_AVX512
  1337. export NO_AVX2
  1338. export BUILD_BFLOAT16
  1339. export SBGEMM_UNROLL_M
  1340. export SBGEMM_UNROLL_N
  1341. export SGEMM_UNROLL_M
  1342. export SGEMM_UNROLL_N
  1343. export DGEMM_UNROLL_M
  1344. export DGEMM_UNROLL_N
  1345. export QGEMM_UNROLL_M
  1346. export QGEMM_UNROLL_N
  1347. export CGEMM_UNROLL_M
  1348. export CGEMM_UNROLL_N
  1349. export ZGEMM_UNROLL_M
  1350. export ZGEMM_UNROLL_N
  1351. export XGEMM_UNROLL_M
  1352. export XGEMM_UNROLL_N
  1353. export CGEMM3M_UNROLL_M
  1354. export CGEMM3M_UNROLL_N
  1355. export ZGEMM3M_UNROLL_M
  1356. export ZGEMM3M_UNROLL_N
  1357. export XGEMM3M_UNROLL_M
  1358. export XGEMM3M_UNROLL_N
  1359. ifdef USE_CUDA
  1360. export CUDADIR
  1361. export CUCC
  1362. export CUFLAGS
  1363. export CULIB
  1364. endif
  1365. .SUFFIXES: .$(PSUFFIX) .$(SUFFIX) .f
  1366. .f.$(SUFFIX):
  1367. $(FC) $(FFLAGS) -c $< -o $(@F)
  1368. .f.$(PSUFFIX):
  1369. $(FC) $(FPFLAGS) -pg -c $< -o $(@F)
  1370. ifdef BINARY64
  1371. PATHSCALEPATH = /opt/pathscale/lib/3.1
  1372. PGIPATH = /opt/pgi/linux86-64/7.1-5/lib
  1373. else
  1374. PATHSCALEPATH = /opt/pathscale/lib/3.1/32
  1375. PGIPATH = /opt/pgi/linux86/7.1-5/lib
  1376. endif
  1377. ACMLPATH = /opt/acml/4.3.0
  1378. ifneq ($(OSNAME), Darwin)
  1379. MKLPATH = /opt/intel/mkl/10.2.2.025/lib
  1380. else
  1381. MKLPATH = /Library/Frameworks/Intel_MKL.framework/Versions/10.0.1.014/lib
  1382. endif
  1383. ATLASPATH = /opt/atlas/3.9.17/opteron
  1384. FLAMEPATH = $(HOME)/flame/lib
  1385. ifneq ($(OSNAME), SunOS)
  1386. SUNPATH = /opt/sunstudio12.1
  1387. else
  1388. SUNPATH = /opt/SUNWspro
  1389. endif