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 34 kB

7 years ago
4 years ago
13 years ago
4 years ago
12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448144914501451145214531454145514561457145814591460146114621463146414651466146714681469147014711472147314741475147614771478147914801481148214831484148514861487148814891490149114921493149414951496149714981499150015011502150315041505150615071508150915101511151215131514151515161517151815191520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559156015611562156315641565156615671568156915701571157215731574157515761577157815791580158115821583158415851586158715881589159015911592159315941595159615971598159916001601160216031604160516061607160816091610161116121613161416151616161716181619162016211622162316241625162616271628162916301631163216331634163516361637163816391640164116421643164416451646164716481649165016511652165316541655165616571658165916601661166216631664166516661667166816691670167116721673167416751676167716781679168016811682168316841685168616871688168916901691169216931694169516961697169816991700170117021703170417051706170717081709171017111712171317141715171617171718171917201721
  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. HAVE_GAS := $(shell as -v < /dev/null 2>&1 | grep GNU 2>&1 >/dev/null)
  16. # Catch conflicting usage of ARCH in some BSD environments
  17. ifeq ($(ARCH), amd64)
  18. override ARCH=x86_64
  19. else ifeq ($(ARCH), powerpc64)
  20. override ARCH=power
  21. else ifeq ($(ARCH), powerpc64le)
  22. override ARCH=power
  23. else ifeq ($(ARCH), powerpc)
  24. override ARCH=power
  25. else ifeq ($(ARCH), i386)
  26. override ARCH=x86
  27. else ifeq ($(ARCH), armv6)
  28. override ARCH=arm
  29. else ifeq ($(ARCH), armv7)
  30. override ARCH=arm
  31. else ifeq ($(ARCH), aarch64)
  32. override ARCH=arm64
  33. else ifeq ($(ARCH), mipsel)
  34. override ARCH=mips
  35. else ifeq ($(ARCH), mips64el)
  36. override ARCH=mips64
  37. else ifeq ($(ARCH), zarch)
  38. override ARCH=zarch
  39. endif
  40. NETLIB_LAPACK_DIR = $(TOPDIR)/lapack-netlib
  41. # Default C compiler
  42. # - Only set if not specified on the command line or inherited from the environment.
  43. # - CC is an implicit variable so neither '?=' or 'ifndef' can be used.
  44. # http://stackoverflow.com/questions/4029274/mingw-and-make-variables
  45. # - Default value is 'cc' which is not always a valid command (e.g. MinGW).
  46. ifeq ($(origin CC),default)
  47. # Check if $(CC) refers to a valid command and set the value to gcc if not
  48. ifneq ($(findstring cmd.exe,$(SHELL)),)
  49. ifeq ($(shell where $(CC) 2>NUL),)
  50. CC = gcc
  51. endif
  52. else # POSIX-ish
  53. ifeq ($(shell command -v $(CC) 2>/dev/null),)
  54. ifeq ($(shell uname -s),Darwin)
  55. CC = clang
  56. # EXTRALIB += -Wl,-no_compact_unwind
  57. else
  58. CC = gcc
  59. endif # Darwin
  60. endif # CC exists
  61. endif # Shell is sane
  62. endif # CC is set to default
  63. # Default Fortran compiler (FC) is selected by f_check.
  64. ifndef MAKEFILE_RULE
  65. include $(TOPDIR)/Makefile.rule
  66. else
  67. include $(TOPDIR)/$(MAKEFILE_RULE)
  68. endif
  69. #
  70. # Beginning of system configuration
  71. #
  72. ifneq ($(BUILD_SINGLE),1)
  73. ifneq ($(BUILD_DOUBLE),1)
  74. ifneq ($(BUILD_COMPLEX),1)
  75. ifneq ($(BUILD_COMPLEX16),1)
  76. override BUILD_SINGLE=1
  77. override BUILD_DOUBLE=1
  78. override BUILD_COMPLEX=1
  79. override BUILD_COMPLEX16=1
  80. endif
  81. endif
  82. endif
  83. endif
  84. ifndef HOSTCC
  85. HOSTCC = $(CC)
  86. endif
  87. ifdef TARGET
  88. GETARCH_FLAGS := -DFORCE_$(TARGET)
  89. GETARCH_FLAGS += -DUSER_TARGET
  90. ifeq ($(TARGET), GENERIC)
  91. ifeq ($(DYNAMIC_ARCH), 1)
  92. override NO_EXPRECISION=1
  93. export NO_EXPRECiSION
  94. endif
  95. endif
  96. endif
  97. # Force fallbacks for 32bit
  98. ifeq ($(BINARY), 32)
  99. ifeq ($(TARGET), HASWELL)
  100. GETARCH_FLAGS := -DFORCE_NEHALEM
  101. endif
  102. ifeq ($(TARGET), SKYLAKEX)
  103. GETARCH_FLAGS := -DFORCE_NEHALEM
  104. endif
  105. ifeq ($(TARGET), COOPERLAKE)
  106. GETARCH_FLAGS := -DFORCE_NEHALEM
  107. endif
  108. ifeq ($(TARGET), SAPPHIRERAPIDS)
  109. GETARCH_FLAGS := -DFORCE_NEHALEM
  110. endif
  111. ifeq ($(TARGET), SANDYBRIDGE)
  112. GETARCH_FLAGS := -DFORCE_NEHALEM
  113. endif
  114. ifeq ($(TARGET), BULLDOZER)
  115. GETARCH_FLAGS := -DFORCE_BARCELONA
  116. endif
  117. ifeq ($(TARGET), PILEDRIVER)
  118. GETARCH_FLAGS := -DFORCE_BARCELONA
  119. endif
  120. ifeq ($(TARGET), STEAMROLLER)
  121. GETARCH_FLAGS := -DFORCE_BARCELONA
  122. endif
  123. ifeq ($(TARGET), EXCAVATOR)
  124. GETARCH_FLAGS := -DFORCE_BARCELONA
  125. endif
  126. ifeq ($(TARGET), ZEN)
  127. GETARCH_FLAGS := -DFORCE_BARCELONA
  128. endif
  129. ifeq ($(TARGET), ARMV8)
  130. GETARCH_FLAGS := -DFORCE_ARMV7
  131. endif
  132. ifeq ($(TARGET), POWER8)
  133. GETARCH_FLAGS := -DFORCE_POWER6
  134. endif
  135. endif
  136. #TARGET_CORE will override TARGET which is used in DYNAMIC_ARCH=1.
  137. #
  138. ifdef TARGET_CORE
  139. GETARCH_FLAGS := -DFORCE_$(TARGET_CORE)
  140. endif
  141. # Force fallbacks for 32bit
  142. ifeq ($(BINARY), 32)
  143. ifeq ($(TARGET_CORE), HASWELL)
  144. GETARCH_FLAGS := -DFORCE_NEHALEM
  145. endif
  146. ifeq ($(TARGET_CORE), SKYLAKEX)
  147. GETARCH_FLAGS := -DFORCE_NEHALEM
  148. endif
  149. ifeq ($(TARGET_CORE), COOPERLAKE)
  150. GETARCH_FLAGS := -DFORCE_NEHALEM
  151. endif
  152. ifeq ($(TARGET_CORE), SAPPHIRERAPIDS)
  153. GETARCH_FLAGS := -DFORCE_NEHALEM
  154. endif
  155. ifeq ($(TARGET_CORE), SANDYBRIDGE)
  156. GETARCH_FLAGS := -DFORCE_NEHALEM
  157. endif
  158. ifeq ($(TARGET_CORE), BULLDOZER)
  159. GETARCH_FLAGS := -DFORCE_BARCELONA
  160. endif
  161. ifeq ($(TARGET_CORE), PILEDRIVER)
  162. GETARCH_FLAGS := -DFORCE_BARCELONA
  163. endif
  164. ifeq ($(TARGET_CORE), STEAMROLLER)
  165. GETARCH_FLAGS := -DFORCE_BARCELONA
  166. endif
  167. ifeq ($(TARGET_CORE), EXCAVATOR)
  168. GETARCH_FLAGS := -DFORCE_BARCELONA
  169. endif
  170. ifeq ($(TARGET_CORE), ZEN)
  171. GETARCH_FLAGS := -DFORCE_BARCELONA
  172. endif
  173. endif
  174. # On x86_64 build getarch with march=native unless the compiler is PGI. This is required to detect AVX512 support in getarch.
  175. ifeq ($(HOSTARCH), x86_64)
  176. ifeq ($(findstring pgcc,$(HOSTCC))$(findstring nvc,$(HOSTCC)),)
  177. GETARCH_FLAGS += -march=native
  178. endif
  179. endif
  180. ifdef INTERFACE64
  181. ifneq ($(INTERFACE64), 0)
  182. GETARCH_FLAGS += -DUSE64BITINT
  183. endif
  184. endif
  185. ifndef GEMM_MULTITHREAD_THRESHOLD
  186. GEMM_MULTITHREAD_THRESHOLD=4
  187. endif
  188. GETARCH_FLAGS += -DGEMM_MULTITHREAD_THRESHOLD=$(GEMM_MULTITHREAD_THRESHOLD)
  189. ifeq ($(NO_AVX), 1)
  190. GETARCH_FLAGS += -DNO_AVX
  191. endif
  192. ifeq ($(BINARY), 32)
  193. GETARCH_FLAGS += -DNO_AVX -DNO_AVX2 -DNO_AVX512
  194. NO_AVX512 = 1
  195. endif
  196. ifeq ($(NO_AVX2), 1)
  197. GETARCH_FLAGS += -DNO_AVX2
  198. endif
  199. ifeq ($(NO_AVX512), 1)
  200. GETARCH_FLAGS += -DNO_AVX512
  201. endif
  202. ifeq ($(DEBUG), 1)
  203. GETARCH_FLAGS += -g
  204. endif
  205. ifeq ($(QUIET_MAKE), 1)
  206. MAKE += -s
  207. endif
  208. ifndef NO_PARALLEL_MAKE
  209. NO_PARALLEL_MAKE=0
  210. endif
  211. GETARCH_FLAGS += -DNO_PARALLEL_MAKE=$(NO_PARALLEL_MAKE)
  212. ifdef MAKE_NB_JOBS
  213. GETARCH_FLAGS += -DMAKE_NB_JOBS=$(MAKE_NB_JOBS)
  214. endif
  215. ifeq ($(HOSTCC), loongcc)
  216. GETARCH_FLAGS += -static
  217. endif
  218. #if don't use Fortran, it will only compile CBLAS.
  219. ifeq ($(ONLY_CBLAS), 1)
  220. NO_LAPACK = 1
  221. else
  222. ONLY_CBLAS = 0
  223. endif
  224. #For small matrix optimization
  225. ifeq ($(ARCH), x86_64)
  226. SMALL_MATRIX_OPT = 1
  227. endif
  228. ifeq ($(SMALL_MATRIX_OPT), 1)
  229. CCOMMON_OPT += -DSMALL_MATRIX_OPT
  230. endif
  231. # This operation is expensive, so execution should be once.
  232. ifndef GOTOBLAS_MAKEFILE
  233. export GOTOBLAS_MAKEFILE = 1
  234. # Generating Makefile.conf and config.h
  235. 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)
  236. ifndef TARGET_CORE
  237. include $(TOPDIR)/Makefile.conf
  238. else
  239. HAVE_NEON=
  240. HAVE_VFP=
  241. HAVE_VFPV3=
  242. HAVE_VFPV4=
  243. HAVE_MMX=
  244. HAVE_SSE=
  245. HAVE_SSE2=
  246. HAVE_SSE3=
  247. HAVE_SSSE3=
  248. HAVE_SSE4_1=
  249. HAVE_SSE4_2=
  250. HAVE_SSE4A=
  251. HAVE_SSE5=
  252. HAVE_AVX=
  253. HAVE_AVX2=
  254. HAVE_FMA3=
  255. include $(TOPDIR)/Makefile_kernel.conf
  256. endif
  257. endif
  258. ifndef NUM_PARALLEL
  259. NUM_PARALLEL = 1
  260. endif
  261. ifndef NUM_THREADS
  262. NUM_THREADS = $(NUM_CORES)
  263. endif
  264. ifeq ($(NUM_THREADS), 1)
  265. override USE_THREAD = 0
  266. override USE_OPENMP = 0
  267. endif
  268. ifdef USE_THREAD
  269. ifeq ($(USE_THREAD), 0)
  270. SMP =
  271. else
  272. SMP = 1
  273. endif
  274. else
  275. ifeq ($(NUM_THREADS), 1)
  276. SMP =
  277. else
  278. SMP = 1
  279. endif
  280. endif
  281. ifeq ($(SMP), 1)
  282. USE_LOCKING =
  283. endif
  284. ifndef NEED_PIC
  285. NEED_PIC = 1
  286. endif
  287. ARFLAGS =
  288. CPP = $(COMPILER) -E
  289. AR ?= $(CROSS_SUFFIX)ar
  290. AS ?= $(CROSS_SUFFIX)as
  291. LD ?= $(CROSS_SUFFIX)ld
  292. RANLIB ?= $(CROSS_SUFFIX)ranlib
  293. NM = $(CROSS_SUFFIX)nm
  294. DLLWRAP = $(CROSS_SUFFIX)dllwrap
  295. OBJCOPY = $(CROSS_SUFFIX)objcopy
  296. OBJCONV = $(CROSS_SUFFIX)objconv
  297. # When fortran support was either not detected or actively deselected, only build BLAS.
  298. ifeq ($(NOFORTRAN), 1)
  299. NO_LAPACK = 1
  300. override FEXTRALIB =
  301. endif
  302. ifeq ($(C_COMPILER), GCC)
  303. GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
  304. GCCVERSIONGT4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 4)
  305. GCCVERSIONGT5 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 5)
  306. GCCVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 7)
  307. GCCVERSIONGTEQ8 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 8)
  308. GCCVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  309. GCCVERSIONGTEQ11 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 11)
  310. GCCVERSIONGTEQ10 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 10)
  311. # Note that the behavior of -dumpversion is compile-time-configurable for
  312. # gcc-7.x and newer. Use -dumpfullversion there
  313. ifeq ($(GCCVERSIONGTEQ7),1)
  314. GCCDUMPVERSION_PARAM := -dumpfullversion
  315. else
  316. GCCDUMPVERSION_PARAM := -dumpversion
  317. endif
  318. GCCMINORVERSIONGTEQ1 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 1)
  319. GCCMINORVERSIONGTEQ2 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 2)
  320. GCCMINORVERSIONGTEQ7 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 7)
  321. endif
  322. #
  323. # OS dependent settings
  324. #
  325. ifeq ($(OSNAME), Darwin)
  326. ifndef MACOSX_DEPLOYMENT_TARGET
  327. export MACOSX_DEPLOYMENT_TARGET=10.8
  328. endif
  329. MD5SUM = md5 -r
  330. endif
  331. ifneq (,$(findstring $(OSNAME), FreeBSD OpenBSD DragonFly))
  332. MD5SUM = md5 -r
  333. endif
  334. ifeq ($(OSNAME), NetBSD)
  335. MD5SUM = md5 -n
  336. endif
  337. ifeq ($(OSNAME), Linux)
  338. EXTRALIB += -lm
  339. NO_EXPRECISION = 1
  340. endif
  341. ifeq ($(OSNAME), Android)
  342. EXTRALIB += -lm
  343. endif
  344. ifeq ($(OSNAME), AIX)
  345. EXTRALIB += -lm
  346. endif
  347. ifeq ($(OSNAME), FreeBSD)
  348. ifeq ($(ARCH), $(filter $(ARCH),arm arm64))
  349. EXTRALIB += -lm
  350. endif
  351. endif
  352. ifeq ($(OSNAME), WINNT)
  353. NEED_PIC = 0
  354. NO_EXPRECISION = 1
  355. EXTRALIB += -defaultlib:advapi32
  356. SUFFIX = obj
  357. PSUFFIX = pobj
  358. LIBSUFFIX = a
  359. ifeq ($(C_COMPILER), CLANG)
  360. CCOMMON_OPT += -DMS_ABI
  361. endif
  362. #Version tests for supporting specific features (MS_ABI, POWER9 intrinsics)
  363. ifeq ($(GCCVERSIONGT4), 1)
  364. # GCC Major version > 4
  365. # It is compatible with MSVC ABI.
  366. CCOMMON_OPT += -DMS_ABI
  367. endif
  368. ifeq ($(GCCVERSIONGTEQ4), 1)
  369. ifeq ($(GCCMINORVERSIONGTEQ7), 1)
  370. # GCC Version >=4.7
  371. # It is compatible with MSVC ABI.
  372. CCOMMON_OPT += -DMS_ABI
  373. endif
  374. endif
  375. # Ensure the correct stack alignment on Win32
  376. # http://permalink.gmane.org/gmane.comp.lib.openblas.general/97
  377. ifeq ($(ARCH), x86)
  378. CCOMMON_OPT += -mincoming-stack-boundary=2
  379. FCOMMON_OPT += -mincoming-stack-boundary=2
  380. endif
  381. endif
  382. ifeq ($(OSNAME), Interix)
  383. NEED_PIC = 0
  384. NO_EXPRECISION = 1
  385. INTERIX_TOOL_DIR = /opt/gcc.3.3/i586-pc-interix3/bin
  386. endif
  387. ifeq ($(OSNAME), CYGWIN_NT)
  388. NEED_PIC = 0
  389. NO_EXPRECISION = 1
  390. OS_CYGWIN_NT = 1
  391. endif
  392. ifneq ($(OSNAME), WINNT)
  393. ifneq ($(OSNAME), CYGWIN_NT)
  394. ifneq ($(OSNAME), Interix)
  395. ifneq ($(OSNAME), Android)
  396. ifdef SMP
  397. EXTRALIB += -lpthread
  398. endif
  399. endif
  400. endif
  401. endif
  402. endif
  403. # ifeq logical or
  404. ifeq ($(OSNAME), $(filter $(OSNAME),WINNT CYGWIN_NT Interix))
  405. OS_WINDOWS=1
  406. endif
  407. ifdef QUAD_PRECISION
  408. CCOMMON_OPT += -DQUAD_PRECISION
  409. NO_EXPRECISION = 1
  410. endif
  411. ifneq ($(ARCH), x86)
  412. ifneq ($(ARCH), x86_64)
  413. NO_EXPRECISION = 1
  414. endif
  415. endif
  416. ifdef UTEST_CHECK
  417. CCOMMON_OPT += -DUTEST_CHECK
  418. SANITY_CHECK = 1
  419. endif
  420. ifdef SANITY_CHECK
  421. CCOMMON_OPT += -DSANITY_CHECK -DREFNAME=$(*F)f$(BU)
  422. endif
  423. MAX_STACK_ALLOC ?= 2048
  424. ifneq ($(MAX_STACK_ALLOC), 0)
  425. CCOMMON_OPT += -DMAX_STACK_ALLOC=$(MAX_STACK_ALLOC)
  426. endif
  427. ifdef USE_LOCKING
  428. ifneq ($(USE_LOCKING), 0)
  429. CCOMMON_OPT += -DUSE_LOCKING
  430. endif
  431. endif
  432. #
  433. # Architecture dependent settings
  434. #
  435. ifeq ($(ARCH), x86)
  436. ifndef BINARY
  437. NO_BINARY_MODE = 1
  438. endif
  439. ifeq ($(CORE), generic)
  440. NO_EXPRECISION = 1
  441. endif
  442. ifndef NO_EXPRECISION
  443. ifeq ($(F_COMPILER), GFORTRAN)
  444. # ifeq logical or. GCC or LSB
  445. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  446. EXPRECISION = 1
  447. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  448. FCOMMON_OPT += -m128bit-long-double
  449. endif
  450. ifeq ($(C_COMPILER), CLANG)
  451. EXPRECISION = 1
  452. CCOMMON_OPT += -DEXPRECISION
  453. FCOMMON_OPT += -m128bit-long-double
  454. endif
  455. endif
  456. endif
  457. endif
  458. ifeq ($(ARCH), x86_64)
  459. ifeq ($(CORE), generic)
  460. NO_EXPRECISION = 1
  461. endif
  462. ifndef NO_EXPRECISION
  463. ifeq ($(F_COMPILER), GFORTRAN)
  464. # ifeq logical or. GCC or LSB
  465. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  466. EXPRECISION = 1
  467. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  468. FCOMMON_OPT += -m128bit-long-double
  469. endif
  470. ifeq ($(C_COMPILER), CLANG)
  471. EXPRECISION = 1
  472. CCOMMON_OPT += -DEXPRECISION
  473. FCOMMON_OPT += -m128bit-long-double
  474. endif
  475. endif
  476. endif
  477. endif
  478. ifeq ($(C_COMPILER), INTEL)
  479. CCOMMON_OPT += -wd981
  480. endif
  481. ifeq ($(USE_OPENMP), 1)
  482. #check
  483. ifeq ($(USE_THREAD), 0)
  484. $(error OpenBLAS: Cannot set both USE_OPENMP=1 and USE_THREAD=0. The USE_THREAD=0 is only for building single thread version.)
  485. endif
  486. # ifeq logical or. GCC or LSB
  487. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  488. CCOMMON_OPT += -fopenmp
  489. endif
  490. ifeq ($(C_COMPILER), CLANG)
  491. CCOMMON_OPT += -fopenmp
  492. endif
  493. ifeq ($(C_COMPILER), INTEL)
  494. CCOMMON_OPT += -fopenmp
  495. endif
  496. ifeq ($(C_COMPILER), PGI)
  497. CCOMMON_OPT += -mp
  498. endif
  499. ifeq ($(C_COMPILER), OPEN64)
  500. CCOMMON_OPT += -mp
  501. CEXTRALIB += -lstdc++
  502. endif
  503. ifeq ($(C_COMPILER), PATHSCALE)
  504. CCOMMON_OPT += -mp
  505. endif
  506. endif
  507. ifeq ($(DYNAMIC_ARCH), 1)
  508. ifeq ($(ARCH), x86)
  509. DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
  510. CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
  511. endif
  512. ifeq ($(ARCH), x86_64)
  513. DYNAMIC_CORE = PRESCOTT CORE2
  514. ifeq ($(DYNAMIC_OLDER), 1)
  515. DYNAMIC_CORE += PENRYN DUNNINGTON
  516. endif
  517. DYNAMIC_CORE += NEHALEM
  518. ifeq ($(DYNAMIC_OLDER), 1)
  519. DYNAMIC_CORE += OPTERON OPTERON_SSE3
  520. endif
  521. DYNAMIC_CORE += BARCELONA
  522. ifeq ($(DYNAMIC_OLDER), 1)
  523. DYNAMIC_CORE += BOBCAT ATOM NANO
  524. endif
  525. ifneq ($(NO_AVX), 1)
  526. DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER STEAMROLLER EXCAVATOR
  527. endif
  528. ifneq ($(NO_AVX2), 1)
  529. DYNAMIC_CORE += HASWELL ZEN
  530. endif
  531. ifneq ($(NO_AVX512), 1)
  532. ifneq ($(NO_AVX2), 1)
  533. DYNAMIC_CORE += SKYLAKEX COOPERLAKE
  534. endif
  535. endif
  536. endif
  537. ifdef DYNAMIC_LIST
  538. override DYNAMIC_CORE = PRESCOTT $(DYNAMIC_LIST)
  539. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_PRESCOTT
  540. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  541. CCOMMON_OPT += $(XCCOMMON_OPT)
  542. #CCOMMON_OPT += -DDYNAMIC_LIST='$(DYNAMIC_LIST)'
  543. endif
  544. ifeq ($(ARCH), arm64)
  545. DYNAMIC_CORE = ARMV8
  546. DYNAMIC_CORE += CORTEXA53
  547. DYNAMIC_CORE += CORTEXA57
  548. DYNAMIC_CORE += CORTEXA72
  549. DYNAMIC_CORE += CORTEXA73
  550. DYNAMIC_CORE += NEOVERSEN1
  551. DYNAMIC_CORE += CORTEXA55
  552. DYNAMIC_CORE += FALKOR
  553. DYNAMIC_CORE += THUNDERX
  554. DYNAMIC_CORE += THUNDERX2T99
  555. DYNAMIC_CORE += TSV110
  556. DYNAMIC_CORE += EMAG8180
  557. DYNAMIC_CORE += THUNDERX3T110
  558. ifdef DYNAMIC_LIST
  559. override DYNAMIC_CORE = ARMV8 $(DYNAMIC_LIST)
  560. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_ARMV8
  561. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  562. endif
  563. endif
  564. ifeq ($(ARCH), mips64)
  565. DYNAMIC_CORE = LOONGSON3R3 LOONGSON3R4
  566. endif
  567. ifeq ($(ARCH), zarch)
  568. DYNAMIC_CORE = ZARCH_GENERIC
  569. # if the compiler accepts -march=arch11 or -march=z13 and can compile a file
  570. # with z13-specific inline assembly, then we can include support for Z13.
  571. # note: -march=z13 is equivalent to -march=arch11 yet some compiler releases
  572. # only support one or the other.
  573. # note: LLVM version 6.x supported -march=z13 yet could not handle vector
  574. # registers in inline assembly, so the check for supporting the -march flag is
  575. # not enough.
  576. ZARCH_TEST_COMPILE=-c $(TOPDIR)/kernel/zarch/damin_z13.c -I$(TOPDIR) -o /dev/null > /dev/null 2> /dev/null
  577. ZARCH_CC_SUPPORTS_ARCH11=$(shell $(CC) -march=arch11 $(ZARCH_TEST_COMPILE) && echo 1)
  578. ZARCH_CC_SUPPORTS_Z13=$(shell $(CC) -march=z13 $(ZARCH_TEST_COMPILE) && echo 1)
  579. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH11), $(ZARCH_CC_SUPPORTS_Z13)), 1)
  580. DYNAMIC_CORE += Z13
  581. CCOMMON_OPT += -DDYN_Z13
  582. else
  583. $(info OpenBLAS: Not building Z13 kernels because the compiler $(CC) does not support it)
  584. endif
  585. # as above for z13, check for -march=arch12 and z14 support in the compiler.
  586. ZARCH_CC_SUPPORTS_ARCH12=$(shell $(CC) -march=arch12 $(ZARCH_TEST_COMPILE) && echo 1)
  587. ZARCH_CC_SUPPORTS_Z14=$(shell $(CC) -march=z14 $(ZARCH_TEST_COMPILE) && echo 1)
  588. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH12), $(ZARCH_CC_SUPPORTS_Z14)), 1)
  589. DYNAMIC_CORE += Z14
  590. CCOMMON_OPT += -DDYN_Z14
  591. else
  592. $(info OpenBLAS: Not building Z14 kernels because the compiler $(CC) does not support it)
  593. endif
  594. endif # ARCH zarch
  595. ifeq ($(ARCH), power)
  596. ifneq ($(C_COMPILER), PGI)
  597. DYNAMIC_CORE = POWER6
  598. DYNAMIC_CORE += POWER8
  599. ifneq ($(C_COMPILER), GCC)
  600. DYNAMIC_CORE += POWER9
  601. DYNAMIC_CORE += POWER10
  602. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  603. endif
  604. ifeq ($(C_COMPILER), GCC)
  605. ifeq ($(GCCVERSIONGT5), 1)
  606. DYNAMIC_CORE += POWER9
  607. else
  608. $(info, OpenBLAS: Your gcc version is too old to build the POWER9 kernels.)
  609. endif
  610. LDVERSIONGTEQ35 := $(shell expr `$(CC) -Wl,--version 2> /dev/null | head -1 | cut -f2 -d "." | cut -f1 -d "-"` \>= 35)
  611. ifeq ($(GCCVERSIONGTEQ11)$(LDVERSIONGTEQ35), 11)
  612. DYNAMIC_CORE += POWER10
  613. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  614. else ifeq ($(GCCVERSIONGTEQ10), 1)
  615. ifeq ($(GCCMINORVERSIONGTEQ2)$(LDVERSIONGTEQ35), 11)
  616. DYNAMIC_CORE += POWER10
  617. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  618. endif
  619. else
  620. $(info, OpenBLAS: Your gcc version is too old to build the POWER10 kernels.)
  621. endif
  622. endif
  623. else
  624. DYNAMIC_CORE = POWER8
  625. DYNAMIC_CORE += POWER9
  626. endif
  627. endif
  628. # If DYNAMIC_CORE is not set, DYNAMIC_ARCH cannot do anything, so force it to empty
  629. ifndef DYNAMIC_CORE
  630. override DYNAMIC_ARCH=
  631. endif
  632. endif
  633. ifeq ($(ARCH), ia64)
  634. NO_BINARY_MODE = 1
  635. BINARY_DEFINED = 1
  636. ifeq ($(F_COMPILER), GFORTRAN)
  637. ifeq ($(C_COMPILER), GCC)
  638. # EXPRECISION = 1
  639. # CCOMMON_OPT += -DEXPRECISION
  640. endif
  641. endif
  642. endif
  643. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  644. NO_BINARY_MODE = 1
  645. endif
  646. ifeq ($(ARCH), alpha)
  647. NO_BINARY_MODE = 1
  648. BINARY_DEFINED = 1
  649. endif
  650. ifeq ($(ARCH), arm)
  651. NO_BINARY_MODE = 1
  652. BINARY_DEFINED = 1
  653. CCOMMON_OPT += -marm
  654. FCOMMON_OPT += -marm
  655. # If softfp abi is mentioned on the command line, force it.
  656. ifeq ($(ARM_SOFTFP_ABI), 1)
  657. CCOMMON_OPT += -mfloat-abi=softfp
  658. FCOMMON_OPT += -mfloat-abi=softfp
  659. endif
  660. ifeq ($(OSNAME), Android)
  661. ifeq ($(ARM_SOFTFP_ABI), 1)
  662. EXTRALIB += -lm
  663. else
  664. EXTRALIB += -Wl,-lm_hard
  665. endif
  666. endif
  667. endif
  668. ifeq ($(ARCH), arm64)
  669. NO_BINARY_MODE = 1
  670. BINARY_DEFINED = 1
  671. ifdef INTERFACE64
  672. ifneq ($(INTERFACE64), 0)
  673. ifeq ($(F_COMPILER), GFORTRAN)
  674. FCOMMON_OPT += -fdefault-integer-8
  675. endif
  676. ifeq ($(F_COMPILER), FLANG)
  677. FCOMMON_OPT += -i8
  678. endif
  679. endif
  680. endif
  681. endif
  682. ifeq ($(ARCH), riscv64)
  683. NO_BINARY_MODE = 1
  684. BINARY_DEFINED = 1
  685. endif
  686. ifeq ($(ARCH), loongarch64)
  687. NO_BINARY_MODE = 1
  688. BINARY_DEFINED = 1
  689. endif
  690. #
  691. # C Compiler dependent settings
  692. #
  693. # ifeq logical or. GCC or CLANG or LSB
  694. # http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
  695. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG LSB))
  696. CCOMMON_OPT += -Wall
  697. COMMON_PROF += -fno-inline
  698. NO_UNINITIALIZED_WARN = -Wno-uninitialized
  699. ifeq ($(QUIET_MAKE), 1)
  700. CCOMMON_OPT += $(NO_UNINITIALIZED_WARN) -Wno-unused
  701. endif
  702. ifdef NO_BINARY_MODE
  703. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  704. ifdef BINARY64
  705. CCOMMON_OPT += -mabi=64
  706. else
  707. CCOMMON_OPT += -mabi=n32
  708. endif
  709. BINARY_DEFINED = 1
  710. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  711. CCOMMON_OPT += -mabi=32
  712. BINARY_DEFINED = 1
  713. endif
  714. ifeq ($(CORE), $(filter $(CORE),LOONGSON3R3 LOONGSON3R4))
  715. CCOMMON_OPT += -march=loongson3a
  716. FCOMMON_OPT += -march=loongson3a
  717. endif
  718. ifeq ($(CORE), MIPS24K)
  719. CCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  720. FCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  721. endif
  722. ifeq ($(CORE), MIPS1004K)
  723. CCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  724. FCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  725. endif
  726. ifeq ($(CORE), P5600)
  727. CCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  728. FCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  729. endif
  730. ifeq ($(CORE), I6400)
  731. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  732. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  733. endif
  734. ifeq ($(CORE), P6600)
  735. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  736. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  737. endif
  738. ifeq ($(CORE), I6500)
  739. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  740. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  741. endif
  742. ifeq ($(OSNAME), AIX)
  743. BINARY_DEFINED = 1
  744. endif
  745. ifeq ($(ARCH), loongarch64)
  746. ifeq ($(CORE), LOONGSON3R5)
  747. CCOMMON_OPT += -march=loongarch64 -mabi=lp64
  748. FCOMMON_OPT += -march=loongarch64 -mabi=lp64
  749. endif
  750. endif
  751. endif
  752. ifndef BINARY_DEFINED
  753. ifneq ($(OSNAME), AIX)
  754. ifdef BINARY64
  755. ifneq ($(ARCH), riscv64)
  756. CCOMMON_OPT += -m64
  757. endif
  758. else
  759. CCOMMON_OPT += -m32
  760. endif
  761. endif
  762. endif
  763. endif
  764. ifeq ($(C_COMPILER), PGI)
  765. PGCVERSIONGT20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \> 20)
  766. PGCVERSIONGTEQ20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \>= 20)
  767. PGCMINORVERSIONGE11 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -c 4-5` == 11)
  768. PGCVERSIONCHECK := $(PGCVERSIONGT20)$(PGCVERSIONEQ20)$(PGCMINORVERSIONGE11)
  769. ifeq ($(PGCVERSIONCHECK), $(filter $(PGCVERSIONCHECK), 110 111 011))
  770. NEWPGI := 1
  771. endif
  772. ifdef BINARY64
  773. ifeq ($(ARCH), x86_64)
  774. CCOMMON_OPT += -tp p7-64
  775. ifneq ($(NEWPGI),1)
  776. CCOMMON_OPT += -D__MMX__ -Mnollvm
  777. endif
  778. else
  779. ifeq ($(ARCH), power)
  780. ifeq ($(CORE), POWER8)
  781. CCOMMON_OPT += -tp pwr8
  782. endif
  783. ifeq ($(CORE), POWER9)
  784. CCOMMON_OPT += -tp pwr9
  785. endif
  786. endif
  787. endif
  788. else
  789. CCOMMON_OPT += -tp p7
  790. endif
  791. endif
  792. ifeq ($(C_COMPILER), PATHSCALE)
  793. ifdef BINARY64
  794. CCOMMON_OPT += -m64
  795. else
  796. CCOMMON_OPT += -m32
  797. endif
  798. endif
  799. #
  800. # Fortran Compiler dependent settings
  801. #
  802. ifeq ($(F_COMPILER), NAG)
  803. FCOMMON_OPT += -dcfuns -recursive -ieee=full -w=obs -thread_safe
  804. ifdef INTERFACE64
  805. ifneq ($(INTERFACE64), 0)
  806. FCOMMON_OPT += -i8
  807. endif
  808. endif
  809. ifeq ($(USE_OPENMP), 1)
  810. FCOMMON_OPT += -openmp
  811. endif
  812. endif
  813. ifeq ($(F_COMPILER), FLANG)
  814. CCOMMON_OPT += -DF_INTERFACE_FLANG
  815. FCOMMON_OPT += -Mrecursive -Kieee
  816. ifeq ($(OSNAME), Linux)
  817. ifeq ($(ARCH), x86_64)
  818. FLANG_VENDOR := $(shell $(FC) --version|head -1 |cut -f 1 -d " ")
  819. ifeq ($(FLANG_VENDOR), AMD)
  820. FCOMMON_OPT += -fno-unroll-loops
  821. endif
  822. endif
  823. endif
  824. ifdef BINARY64
  825. ifdef INTERFACE64
  826. ifneq ($(INTERFACE64), 0)
  827. FCOMMON_OPT += -i8
  828. endif
  829. endif
  830. FCOMMON_OPT += -Wall
  831. else
  832. FCOMMON_OPT += -Wall
  833. endif
  834. ifeq ($(USE_OPENMP), 1)
  835. FCOMMON_OPT += -fopenmp
  836. endif
  837. endif
  838. ifeq ($(F_COMPILER), G77)
  839. CCOMMON_OPT += -DF_INTERFACE_G77
  840. FCOMMON_OPT += -Wall
  841. ifndef NO_BINARY_MODE
  842. ifneq ($(OSNAME), AIX)
  843. ifdef BINARY64
  844. FCOMMON_OPT += -m64
  845. else
  846. FCOMMON_OPT += -m32
  847. endif
  848. endif
  849. endif
  850. endif
  851. ifeq ($(F_COMPILER), G95)
  852. CCOMMON_OPT += -DF_INTERFACE_G95
  853. FCOMMON_OPT += -Wall
  854. ifneq ($(OSNAME), AIX)
  855. ifndef NO_BINARY_MODE
  856. ifdef BINARY64
  857. FCOMMON_OPT += -m64
  858. else
  859. FCOMMON_OPT += -m32
  860. endif
  861. endif
  862. ifneq ($(NO_LAPACKE), 1)
  863. FCOMMON_OPT += -fno-second-underscore
  864. endif
  865. endif
  866. endif
  867. ifeq ($(F_COMPILER), GFORTRAN)
  868. CCOMMON_OPT += -DF_INTERFACE_GFORT
  869. FCOMMON_OPT += -Wall
  870. # make single-threaded LAPACK calls thread-safe #1847
  871. FCOMMON_OPT += -frecursive
  872. # work around ABI problem with passing single-character arguments
  873. FCOMMON_OPT += -fno-optimize-sibling-calls
  874. #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
  875. ifneq ($(NO_LAPACK), 1)
  876. EXTRALIB += -lgfortran
  877. endif
  878. ifdef NO_BINARY_MODE
  879. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  880. ifdef BINARY64
  881. FCOMMON_OPT += -mabi=64
  882. else
  883. FCOMMON_OPT += -mabi=n32
  884. endif
  885. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  886. FCOMMON_OPT += -mabi=32
  887. endif
  888. else
  889. ifdef BINARY64
  890. ifneq ($(OSNAME), AIX)
  891. ifneq ($(ARCH), riscv64)
  892. FCOMMON_OPT += -m64
  893. endif
  894. endif
  895. ifdef INTERFACE64
  896. ifneq ($(INTERFACE64), 0)
  897. FCOMMON_OPT += -fdefault-integer-8
  898. endif
  899. endif
  900. else
  901. ifneq ($(OSNAME), AIX)
  902. FCOMMON_OPT += -m32
  903. endif
  904. endif
  905. endif
  906. ifeq ($(USE_OPENMP), 1)
  907. FCOMMON_OPT += -fopenmp
  908. endif
  909. endif
  910. ifeq ($(F_COMPILER), INTEL)
  911. CCOMMON_OPT += -DF_INTERFACE_INTEL
  912. ifdef INTERFACE64
  913. ifneq ($(INTERFACE64), 0)
  914. FCOMMON_OPT += -i8
  915. endif
  916. endif
  917. FCOMMON_OPT += -recursive -fp-model strict -assume protect-parens
  918. ifeq ($(USE_OPENMP), 1)
  919. FCOMMON_OPT += -fopenmp
  920. endif
  921. endif
  922. ifeq ($(F_COMPILER), FUJITSU)
  923. CCOMMON_OPT += -DF_INTERFACE_FUJITSU
  924. ifeq ($(USE_OPENMP), 1)
  925. FCOMMON_OPT += -openmp
  926. endif
  927. endif
  928. ifeq ($(F_COMPILER), IBM)
  929. CCOMMON_OPT += -DF_INTERFACE_IBM
  930. # FCOMMON_OPT += -qarch=440
  931. ifdef BINARY64
  932. FCOMMON_OPT += -q64
  933. ifdef INTERFACE64
  934. ifneq ($(INTERFACE64), 0)
  935. FCOMMON_OPT += -qintsize=8
  936. endif
  937. endif
  938. else
  939. FCOMMON_OPT += -q32
  940. endif
  941. ifeq ($(USE_OPENMP), 1)
  942. FCOMMON_OPT += -openmp
  943. endif
  944. endif
  945. ifeq ($(F_COMPILER), PGI)
  946. CCOMMON_OPT += -DF_INTERFACE_PGI
  947. COMMON_PROF += -DPGICOMPILER
  948. ifdef BINARY64
  949. ifdef INTERFACE64
  950. ifneq ($(INTERFACE64), 0)
  951. FCOMMON_OPT += -i8
  952. endif
  953. endif
  954. ifeq ($(ARCH), x86_64)
  955. FCOMMON_OPT += -tp p7-64
  956. else
  957. ifeq ($(ARCH), power)
  958. ifeq ($(CORE), POWER6)
  959. $(warning NVIDIA HPC compilers do not support POWER6.)
  960. endif
  961. ifeq ($(CORE), POWER8)
  962. FCOMMON_OPT += -tp pwr8
  963. endif
  964. ifeq ($(CORE), POWER9)
  965. FCOMMON_OPT += -tp pwr9
  966. endif
  967. ifeq ($(CORE), POWER10)
  968. $(warning NVIDIA HPC compilers do not support POWER10.)
  969. endif
  970. endif
  971. endif
  972. else
  973. FCOMMON_OPT += -tp p7
  974. endif
  975. FCOMMON_OPT += -Mrecursive -Kieee
  976. ifeq ($(USE_OPENMP), 1)
  977. FCOMMON_OPT += -mp
  978. endif
  979. endif
  980. ifeq ($(F_COMPILER), PATHSCALE)
  981. CCOMMON_OPT += -DF_INTERFACE_PATHSCALE
  982. ifdef BINARY64
  983. ifdef INTERFACE64
  984. ifneq ($(INTERFACE64), 0)
  985. FCOMMON_OPT += -i8
  986. endif
  987. endif
  988. endif
  989. ifeq ($(USE_OPENMP), 1)
  990. FCOMMON_OPT += -mp
  991. endif
  992. endif
  993. ifeq ($(F_COMPILER), OPEN64)
  994. CCOMMON_OPT += -DF_INTERFACE_OPEN64
  995. ifdef BINARY64
  996. ifdef INTERFACE64
  997. ifneq ($(INTERFACE64), 0)
  998. FCOMMON_OPT += -i8
  999. endif
  1000. endif
  1001. endif
  1002. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  1003. ifndef BINARY64
  1004. FCOMMON_OPT += -n32
  1005. else
  1006. FCOMMON_OPT += -n64
  1007. endif
  1008. ifeq ($(CORE), LOONGSON3R3)
  1009. FCOMMON_OPT += -loongson3 -static
  1010. endif
  1011. ifeq ($(CORE), LOONGSON3R4)
  1012. FCOMMON_OPT += -loongson3 -static
  1013. endif
  1014. else
  1015. ifndef BINARY64
  1016. FCOMMON_OPT += -m32
  1017. else
  1018. FCOMMON_OPT += -m64
  1019. endif
  1020. endif
  1021. ifeq ($(USE_OPENMP), 1)
  1022. FEXTRALIB += -lstdc++
  1023. FCOMMON_OPT += -mp
  1024. endif
  1025. endif
  1026. ifeq ($(C_COMPILER), OPEN64)
  1027. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  1028. ifndef BINARY64
  1029. CCOMMON_OPT += -n32
  1030. else
  1031. CCOMMON_OPT += -n64
  1032. endif
  1033. ifeq ($(CORE), LOONGSON3R3)
  1034. CCOMMON_OPT += -loongson3 -static
  1035. endif
  1036. ifeq ($(CORE), LOONGSON3R4)
  1037. CCOMMON_OPT += -loongson3 -static
  1038. endif
  1039. else
  1040. ifndef BINARY64
  1041. CCOMMON_OPT += -m32
  1042. else
  1043. CCOMMON_OPT += -m64
  1044. endif
  1045. endif
  1046. endif
  1047. ifeq ($(C_COMPILER), SUN)
  1048. CCOMMON_OPT += -w
  1049. ifeq ($(ARCH), x86)
  1050. CCOMMON_OPT += -m32
  1051. else
  1052. ifdef BINARY64
  1053. CCOMMON_OPT += -m64
  1054. else
  1055. CCOMMON_OPT += -m32
  1056. endif
  1057. endif
  1058. endif
  1059. ifeq ($(F_COMPILER), SUN)
  1060. CCOMMON_OPT += -DF_INTERFACE_SUN
  1061. FCOMMON_OPT += -ftrap=%none -xrecursive
  1062. ifeq ($(ARCH), x86)
  1063. FCOMMON_OPT += -m32
  1064. else
  1065. ifdef BINARY64
  1066. FCOMMON_OPT += -m64
  1067. else
  1068. FCOMMON_OPT += -m32
  1069. endif
  1070. endif
  1071. ifeq ($(USE_OPENMP), 1)
  1072. FCOMMON_OPT += -xopenmp=parallel
  1073. endif
  1074. endif
  1075. ifeq ($(F_COMPILER), COMPAQ)
  1076. CCOMMON_OPT += -DF_INTERFACE_COMPAQ
  1077. ifeq ($(USE_OPENMP), 1)
  1078. FCOMMON_OPT += -openmp
  1079. endif
  1080. endif
  1081. ifdef BINARY64
  1082. ifdef INTERFACE64
  1083. ifneq ($(INTERFACE64), 0)
  1084. CCOMMON_OPT +=
  1085. #-DUSE64BITINT
  1086. endif
  1087. endif
  1088. endif
  1089. ifeq ($(NEED_PIC), 1)
  1090. ifeq ($(C_COMPILER), IBM)
  1091. CCOMMON_OPT += -qpic=large
  1092. else
  1093. CCOMMON_OPT += -fPIC
  1094. endif
  1095. ifeq ($(F_COMPILER), SUN)
  1096. FCOMMON_OPT += -pic
  1097. else ifeq ($(F_COMPILER), NAG)
  1098. FCOMMON_OPT += -PIC
  1099. else
  1100. FCOMMON_OPT += -fPIC
  1101. endif
  1102. endif
  1103. ifeq ($(DYNAMIC_ARCH), 1)
  1104. CCOMMON_OPT += -DDYNAMIC_ARCH
  1105. endif
  1106. ifeq ($(DYNAMIC_OLDER), 1)
  1107. CCOMMON_OPT += -DDYNAMIC_OLDER
  1108. endif
  1109. ifeq ($(NO_LAPACK), 1)
  1110. CCOMMON_OPT += -DNO_LAPACK
  1111. #Disable LAPACK C interface
  1112. NO_LAPACKE = 1
  1113. endif
  1114. ifeq ($(NO_LAPACKE), 1)
  1115. CCOMMON_OPT += -DNO_LAPACKE
  1116. endif
  1117. ifeq ($(NO_AVX), 1)
  1118. CCOMMON_OPT += -DNO_AVX
  1119. endif
  1120. ifeq ($(ARCH), x86)
  1121. CCOMMON_OPT += -DNO_AVX
  1122. endif
  1123. ifeq ($(NO_AVX2), 1)
  1124. CCOMMON_OPT += -DNO_AVX2
  1125. endif
  1126. ifeq ($(NO_AVX512), 1)
  1127. CCOMMON_OPT += -DNO_AVX512
  1128. endif
  1129. ifdef SMP
  1130. CCOMMON_OPT += -DSMP_SERVER
  1131. ifeq ($(ARCH), mips64)
  1132. USE_SIMPLE_THREADED_LEVEL3 = 1
  1133. endif
  1134. ifeq ($(USE_OPENMP), 1)
  1135. # USE_SIMPLE_THREADED_LEVEL3 = 1
  1136. # NO_AFFINITY = 1
  1137. CCOMMON_OPT += -DUSE_OPENMP
  1138. endif
  1139. ifeq ($(BIGNUMA), 1)
  1140. CCOMMON_OPT += -DBIGNUMA
  1141. endif
  1142. endif
  1143. ifeq ($(NO_WARMUP), 1)
  1144. CCOMMON_OPT += -DNO_WARMUP
  1145. endif
  1146. ifeq ($(CONSISTENT_FPCSR), 1)
  1147. CCOMMON_OPT += -DCONSISTENT_FPCSR
  1148. endif
  1149. # Only for development
  1150. # CCOMMON_OPT += -DPARAMTEST
  1151. # CCOMMON_OPT += -DPREFETCHTEST
  1152. # CCOMMON_OPT += -DNO_SWITCHING
  1153. # USE_PAPI = 1
  1154. ifdef USE_PAPI
  1155. CCOMMON_OPT += -DUSE_PAPI
  1156. EXTRALIB += -lpapi -lperfctr
  1157. endif
  1158. ifdef BUFFERSIZE
  1159. CCOMMON_OPT += -DBUFFERSIZE=$(BUFFERSIZE)
  1160. endif
  1161. ifdef DYNAMIC_THREADS
  1162. CCOMMON_OPT += -DDYNAMIC_THREADS
  1163. endif
  1164. CCOMMON_OPT += -DMAX_CPU_NUMBER=$(NUM_THREADS)
  1165. CCOMMON_OPT += -DMAX_PARALLEL_NUMBER=$(NUM_PARALLEL)
  1166. ifdef USE_SIMPLE_THREADED_LEVEL3
  1167. CCOMMON_OPT += -DUSE_SIMPLE_THREADED_LEVEL3
  1168. endif
  1169. ifeq ($(USE_TLS), 1)
  1170. CCOMMON_OPT += -DUSE_TLS
  1171. endif
  1172. ifeq ($(BUILD_BFLOAT16), 1)
  1173. CCOMMON_OPT += -DBUILD_BFLOAT16
  1174. endif
  1175. ifeq ($(BUILD_SINGLE), 1)
  1176. CCOMMON_OPT += -DBUILD_SINGLE=1
  1177. endif
  1178. ifeq ($(BUILD_DOUBLE), 1)
  1179. CCOMMON_OPT += -DBUILD_DOUBLE=1
  1180. endif
  1181. ifeq ($(BUILD_COMPLEX), 1)
  1182. CCOMMON_OPT += -DBUILD_COMPLEX=1
  1183. endif
  1184. ifeq ($(BUILD_COMPLEX16), 1)
  1185. CCOMMON_OPT += -DBUILD_COMPLEX16=1
  1186. endif
  1187. CCOMMON_OPT += -DVERSION=\"$(VERSION)\"
  1188. ifndef SYMBOLPREFIX
  1189. SYMBOLPREFIX =
  1190. endif
  1191. ifndef SYMBOLSUFFIX
  1192. SYMBOLSUFFIX =
  1193. endif
  1194. ifndef LIBSONAMEBASE
  1195. LIBSONAMEBASE = openblas
  1196. endif
  1197. ifndef LIBNAMESUFFIX
  1198. LIBNAMEBASE = $(SYMBOLPREFIX)$(LIBSONAMEBASE)$(SYMBOLSUFFIX)
  1199. else
  1200. LIBNAMEBASE = $(SYMBOLPREFIX)$(LIBSONAMEBASE)$(SYMBOLSUFFIX)_$(LIBNAMESUFFIX)
  1201. endif
  1202. ifeq ($(OSNAME), CYGWIN_NT)
  1203. LIBPREFIX = cyg$(LIBNAMEBASE)
  1204. else
  1205. LIBPREFIX = lib$(LIBNAMEBASE)
  1206. endif
  1207. KERNELDIR = $(TOPDIR)/kernel/$(ARCH)
  1208. include $(TOPDIR)/Makefile.$(ARCH)
  1209. ifneq ($(C_COMPILER), PGI)
  1210. ifneq ($(C_COMPILER), SUN)
  1211. CCOMMON_OPT += -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME
  1212. endif
  1213. endif
  1214. CCOMMON_OPT += -DASMNAME=$(FU)$(*F) -DASMFNAME=$(FU)$(*F)$(BU) -DNAME=$(*F)$(BU) -DCNAME=$(*F) -DCHAR_NAME=\"$(*F)$(BU)\" -DCHAR_CNAME=\"$(*F)\"
  1215. ifeq ($(CORE), PPC440)
  1216. CCOMMON_OPT += -DALLOC_QALLOC
  1217. endif
  1218. ifeq ($(CORE), PPC440FP2)
  1219. STATIC_ALLOCATION = 1
  1220. endif
  1221. ifneq ($(OSNAME), Linux)
  1222. NO_AFFINITY = 1
  1223. endif
  1224. ifneq ($(ARCH), x86_64)
  1225. ifneq ($(ARCH), x86)
  1226. NO_AFFINITY = 1
  1227. endif
  1228. endif
  1229. ifdef NO_AFFINITY
  1230. ifeq ($(NO_AFFINITY), 0)
  1231. override undefine NO_AFFINITY
  1232. else
  1233. CCOMMON_OPT += -DNO_AFFINITY
  1234. endif
  1235. endif
  1236. ifdef FUNCTION_PROFILE
  1237. CCOMMON_OPT += -DFUNCTION_PROFILE
  1238. endif
  1239. ifdef HUGETLB_ALLOCATION
  1240. CCOMMON_OPT += -DALLOC_HUGETLB
  1241. endif
  1242. ifdef HUGETLBFILE_ALLOCATION
  1243. CCOMMON_OPT += -DALLOC_HUGETLBFILE -DHUGETLB_FILE_NAME=$(HUGETLBFILE_ALLOCATION)
  1244. endif
  1245. ifdef STATIC_ALLOCATION
  1246. CCOMMON_OPT += -DALLOC_STATIC
  1247. endif
  1248. ifdef DEVICEDRIVER_ALLOCATION
  1249. CCOMMON_OPT += -DALLOC_DEVICEDRIVER -DDEVICEDRIVER_NAME=\"/dev/mapper\"
  1250. endif
  1251. ifdef MIXED_MEMORY_ALLOCATION
  1252. CCOMMON_OPT += -DMIXED_MEMORY_ALLOCATION
  1253. endif
  1254. ifeq ($(OSNAME), SunOS)
  1255. TAR = gtar
  1256. PATCH = gpatch
  1257. GREP = ggrep
  1258. AWK = nawk
  1259. else
  1260. TAR = tar
  1261. PATCH = patch
  1262. GREP = grep
  1263. AWK = awk
  1264. endif
  1265. ifndef MD5SUM
  1266. MD5SUM = md5sum
  1267. endif
  1268. REVISION = -r$(VERSION)
  1269. MAJOR_VERSION = $(word 1,$(subst ., ,$(VERSION)))
  1270. ifeq ($(DEBUG), 1)
  1271. COMMON_OPT += -g
  1272. endif
  1273. ifeq ($(DEBUG), 1)
  1274. FCOMMON_OPT += -g
  1275. endif
  1276. ifndef COMMON_OPT
  1277. COMMON_OPT = -O2
  1278. endif
  1279. ifndef FCOMMON_OPT
  1280. FCOMMON_OPT = -O2 -frecursive
  1281. endif
  1282. override CFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
  1283. override PFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
  1284. override FFLAGS += $(COMMON_OPT) $(FCOMMON_OPT)
  1285. override FPFLAGS += $(FCOMMON_OPT) $(COMMON_PROF)
  1286. #MAKEOVERRIDES =
  1287. ifdef NEED_PIC
  1288. ifeq (,$(findstring PIC,$(FFLAGS)))
  1289. override FFLAGS += -fPIC
  1290. endif
  1291. endif
  1292. #For LAPACK Fortran codes.
  1293. #Disable -fopenmp for LAPACK Fortran codes on Windows.
  1294. ifdef OS_WINDOWS
  1295. LAPACK_FFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FFLAGS))
  1296. LAPACK_FPFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FPFLAGS))
  1297. else
  1298. LAPACK_FFLAGS := $(FFLAGS)
  1299. LAPACK_FPFLAGS := $(FPFLAGS)
  1300. endif
  1301. ifeq ($(F_COMPILER),NAG)
  1302. LAPACK_FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1303. endif
  1304. LAPACK_CFLAGS = $(CFLAGS)
  1305. LAPACK_CFLAGS += -DHAVE_LAPACK_CONFIG_H
  1306. ifdef INTERFACE64
  1307. ifneq ($(INTERFACE64), 0)
  1308. LAPACK_CFLAGS += -DLAPACK_ILP64
  1309. endif
  1310. endif
  1311. ifdef OS_WINDOWS
  1312. LAPACK_CFLAGS += -DOPENBLAS_OS_WINDOWS
  1313. endif
  1314. ifeq ($(C_COMPILER), LSB)
  1315. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1316. endif
  1317. ifndef SUFFIX
  1318. SUFFIX = o
  1319. endif
  1320. ifndef PSUFFIX
  1321. PSUFFIX = po
  1322. endif
  1323. ifndef LIBSUFFIX
  1324. LIBSUFFIX = a
  1325. endif
  1326. ifneq ($(DYNAMIC_ARCH), 1)
  1327. ifndef SMP
  1328. LIBNAME = $(LIBPREFIX)_$(LIBCORE)$(REVISION).$(LIBSUFFIX)
  1329. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)$(REVISION)_p.$(LIBSUFFIX)
  1330. else
  1331. LIBNAME = $(LIBPREFIX)_$(LIBCORE)p$(REVISION).$(LIBSUFFIX)
  1332. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)p$(REVISION)_p.$(LIBSUFFIX)
  1333. endif
  1334. else
  1335. ifndef SMP
  1336. LIBNAME = $(LIBPREFIX)$(REVISION).$(LIBSUFFIX)
  1337. LIBNAME_P = $(LIBPREFIX)$(REVISION)_p.$(LIBSUFFIX)
  1338. else
  1339. LIBNAME = $(LIBPREFIX)p$(REVISION).$(LIBSUFFIX)
  1340. LIBNAME_P = $(LIBPREFIX)p$(REVISION)_p.$(LIBSUFFIX)
  1341. endif
  1342. endif
  1343. LIBDLLNAME = $(LIBPREFIX).dll
  1344. IMPLIBNAME = lib$(LIBNAMEBASE).dll.a
  1345. ifneq ($(OSNAME), AIX)
  1346. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.so)
  1347. else
  1348. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.a)
  1349. endif
  1350. LIBDYNNAME = $(LIBNAME:.$(LIBSUFFIX)=.dylib)
  1351. LIBDEFNAME = $(LIBNAME:.$(LIBSUFFIX)=.def)
  1352. LIBEXPNAME = $(LIBNAME:.$(LIBSUFFIX)=.exp)
  1353. LIBZIPNAME = $(LIBNAME:.$(LIBSUFFIX)=.zip)
  1354. LIBS = $(TOPDIR)/$(LIBNAME)
  1355. LIBS_P = $(TOPDIR)/$(LIBNAME_P)
  1356. LIB_COMPONENTS = BLAS
  1357. ifneq ($(NO_CBLAS), 1)
  1358. LIB_COMPONENTS += CBLAS
  1359. endif
  1360. ifneq ($(NO_LAPACK), 1)
  1361. LIB_COMPONENTS += LAPACK
  1362. ifneq ($(NO_LAPACKE), 1)
  1363. LIB_COMPONENTS += LAPACKE
  1364. endif
  1365. ifeq ($(BUILD_RELAPACK), 1)
  1366. LIB_COMPONENTS += ReLAPACK
  1367. endif
  1368. endif
  1369. ifeq ($(ONLY_CBLAS), 1)
  1370. LIB_COMPONENTS = CBLAS
  1371. endif
  1372. export OSNAME
  1373. export ARCH
  1374. export CORE
  1375. export LIBCORE
  1376. export __BYTE_ORDER__
  1377. export ELF_VERSION
  1378. export PGCPATH
  1379. export CONFIG
  1380. export CC
  1381. export FC
  1382. export BU
  1383. export FU
  1384. export NEED2UNDERSCORES
  1385. export USE_THREAD
  1386. export NUM_THREADS
  1387. export NUM_CORES
  1388. export SMP
  1389. export MAKEFILE_RULE
  1390. export NEED_PIC
  1391. export BINARY
  1392. export BINARY32
  1393. export BINARY64
  1394. export F_COMPILER
  1395. export C_COMPILER
  1396. export USE_OPENMP
  1397. export CROSS
  1398. export CROSS_SUFFIX
  1399. export NOFORTRAN
  1400. export NO_FBLAS
  1401. export EXTRALIB
  1402. export CEXTRALIB
  1403. export FEXTRALIB
  1404. export HAVE_SSE
  1405. export HAVE_SSE2
  1406. export HAVE_SSE3
  1407. export HAVE_SSSE3
  1408. export HAVE_SSE4_1
  1409. export HAVE_SSE4_2
  1410. export HAVE_SSE4A
  1411. export HAVE_SSE5
  1412. export HAVE_AVX
  1413. export HAVE_AVX2
  1414. export HAVE_FMA3
  1415. export HAVE_VFP
  1416. export HAVE_VFPV3
  1417. export HAVE_VFPV4
  1418. export HAVE_NEON
  1419. export HAVE_MSA
  1420. export MSA_FLAGS
  1421. export KERNELDIR
  1422. export FUNCTION_PROFILE
  1423. export TARGET_CORE
  1424. export NO_AVX512
  1425. export NO_AVX2
  1426. export BUILD_BFLOAT16
  1427. export SBGEMM_UNROLL_M
  1428. export SBGEMM_UNROLL_N
  1429. export SGEMM_UNROLL_M
  1430. export SGEMM_UNROLL_N
  1431. export DGEMM_UNROLL_M
  1432. export DGEMM_UNROLL_N
  1433. export QGEMM_UNROLL_M
  1434. export QGEMM_UNROLL_N
  1435. export CGEMM_UNROLL_M
  1436. export CGEMM_UNROLL_N
  1437. export ZGEMM_UNROLL_M
  1438. export ZGEMM_UNROLL_N
  1439. export XGEMM_UNROLL_M
  1440. export XGEMM_UNROLL_N
  1441. export CGEMM3M_UNROLL_M
  1442. export CGEMM3M_UNROLL_N
  1443. export ZGEMM3M_UNROLL_M
  1444. export ZGEMM3M_UNROLL_N
  1445. export XGEMM3M_UNROLL_M
  1446. export XGEMM3M_UNROLL_N
  1447. ifdef USE_CUDA
  1448. export CUDADIR
  1449. export CUCC
  1450. export CUFLAGS
  1451. export CULIB
  1452. endif
  1453. .SUFFIXES: .$(PSUFFIX) .$(SUFFIX) .f
  1454. .f.$(SUFFIX):
  1455. $(FC) $(FFLAGS) -c $< -o $(@F)
  1456. .f.$(PSUFFIX):
  1457. $(FC) $(FPFLAGS) -pg -c $< -o $(@F)
  1458. ifdef BINARY64
  1459. PATHSCALEPATH = /opt/pathscale/lib/3.1
  1460. PGIPATH = /opt/pgi/linux86-64/7.1-5/lib
  1461. else
  1462. PATHSCALEPATH = /opt/pathscale/lib/3.1/32
  1463. PGIPATH = /opt/pgi/linux86/7.1-5/lib
  1464. endif
  1465. ACMLPATH = /opt/acml/4.3.0
  1466. ifneq ($(OSNAME), Darwin)
  1467. MKLPATH = /opt/intel/mkl/10.2.2.025/lib
  1468. else
  1469. MKLPATH = /Library/Frameworks/Intel_MKL.framework/Versions/10.0.1.014/lib
  1470. endif
  1471. ATLASPATH = /opt/atlas/3.9.17/opteron
  1472. FLAMEPATH = $(HOME)/flame/lib
  1473. ifneq ($(OSNAME), SunOS)
  1474. SUNPATH = /opt/sunstudio12.1
  1475. else
  1476. SUNPATH = /opt/SUNWspro
  1477. endif