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 38 kB

7 years ago
4 years ago
13 years ago
4 years ago
12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448144914501451145214531454145514561457145814591460146114621463146414651466146714681469147014711472147314741475147614771478147914801481148214831484148514861487148814891490149114921493149414951496149714981499150015011502150315041505150615071508150915101511151215131514151515161517151815191520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559156015611562156315641565156615671568156915701571157215731574157515761577157815791580158115821583158415851586158715881589159015911592159315941595159615971598159916001601160216031604160516061607160816091610161116121613161416151616161716181619162016211622162316241625162616271628162916301631163216331634163516361637163816391640164116421643164416451646164716481649165016511652165316541655165616571658165916601661166216631664166516661667166816691670167116721673167416751676167716781679168016811682168316841685168616871688168916901691169216931694169516961697169816991700170117021703170417051706170717081709171017111712171317141715171617171718171917201721172217231724172517261727172817291730173117321733173417351736173717381739174017411742174317441745174617471748174917501751175217531754175517561757175817591760176117621763176417651766176717681769177017711772177317741775177617771778177917801781178217831784178517861787178817891790179117921793179417951796179717981799180018011802180318041805180618071808180918101811181218131814181518161817181818191820182118221823182418251826182718281829183018311832183318341835183618371838183918401841184218431844184518461847184818491850185118521853
  1. #
  2. # Include user definition
  3. #
  4. # TO suppress recursive includes
  5. INCLUDED = 1
  6. ifndef TOPDIR
  7. TOPDIR = .
  8. endif
  9. ifndef RELAPACK_REPLACE
  10. RELAPACK_REPLACE=0
  11. endif
  12. # we need to use the host system's architecture for getarch compile options even especially when cross-compiling
  13. HOSTARCH := $(shell uname -m)
  14. ifeq ($(HOSTARCH), amd64)
  15. HOSTARCH=x86_64
  16. endif
  17. # Catch conflicting usage of ARCH in some BSD environments
  18. ifeq ($(ARCH), amd64)
  19. override ARCH=x86_64
  20. else ifeq ($(ARCH), powerpc64)
  21. override ARCH=power
  22. else ifeq ($(ARCH), powerpc64le)
  23. override ARCH=power
  24. else ifeq ($(ARCH), powerpc)
  25. override ARCH=power
  26. else ifeq ($(ARCH), i386)
  27. override ARCH=x86
  28. else ifeq ($(ARCH), armv6)
  29. override ARCH=arm
  30. else ifeq ($(ARCH), armv7)
  31. override ARCH=arm
  32. else ifeq ($(ARCH), aarch64)
  33. override ARCH=arm64
  34. else ifeq ($(ARCH), mipsel)
  35. override ARCH=mips
  36. else ifeq ($(ARCH), mips64el)
  37. override ARCH=mips64
  38. else ifeq ($(ARCH), zarch)
  39. override ARCH=zarch
  40. endif
  41. NETLIB_LAPACK_DIR = $(TOPDIR)/lapack-netlib
  42. # Default C compiler
  43. # - Only set if not specified on the command line or inherited from the environment.
  44. # - CC is an implicit variable so neither '?=' or 'ifndef' can be used.
  45. # http://stackoverflow.com/questions/4029274/mingw-and-make-variables
  46. # - Default value is 'cc' which is not always a valid command (e.g. MinGW).
  47. ifeq ($(origin CC),default)
  48. # Check if $(CC) refers to a valid command and set the value to gcc if not
  49. ifneq ($(findstring cmd.exe,$(SHELL)),)
  50. ifeq ($(shell where $(CC) 2>NUL),)
  51. CC = gcc
  52. endif
  53. else # POSIX-ish
  54. ifeq ($(shell command -v $(CC) 2>/dev/null),)
  55. ifeq ($(shell uname -s),Darwin)
  56. CC = clang
  57. # EXTRALIB += -Wl,-no_compact_unwind
  58. else
  59. CC = gcc
  60. endif # Darwin
  61. endif # CC exists
  62. endif # Shell is sane
  63. endif # CC is set to default
  64. # Default Fortran compiler (FC) is selected by f_check.
  65. ifndef MAKEFILE_RULE
  66. include $(TOPDIR)/Makefile.rule
  67. else
  68. include $(TOPDIR)/$(MAKEFILE_RULE)
  69. endif
  70. #
  71. # Beginning of system configuration
  72. #
  73. ifneq ($(BUILD_SINGLE),1)
  74. ifneq ($(BUILD_DOUBLE),1)
  75. ifneq ($(BUILD_COMPLEX),1)
  76. ifneq ($(BUILD_COMPLEX16),1)
  77. override BUILD_SINGLE=1
  78. override BUILD_DOUBLE=1
  79. override BUILD_COMPLEX=1
  80. override BUILD_COMPLEX16=1
  81. endif
  82. endif
  83. endif
  84. endif
  85. ifndef HOSTCC
  86. HOSTCC = $(CC)
  87. endif
  88. ifdef TARGET
  89. GETARCH_FLAGS := -DFORCE_$(TARGET)
  90. GETARCH_FLAGS += -DUSER_TARGET
  91. ifeq ($(TARGET), GENERIC)
  92. ifeq ($(DYNAMIC_ARCH), 1)
  93. override NO_EXPRECISION=1
  94. export NO_EXPRECISION
  95. endif
  96. endif
  97. endif
  98. # Force fallbacks for 32bit
  99. ifeq ($(BINARY), 32)
  100. ifeq ($(TARGET), HASWELL)
  101. GETARCH_FLAGS := -DFORCE_NEHALEM
  102. endif
  103. ifeq ($(TARGET), SKYLAKEX)
  104. GETARCH_FLAGS := -DFORCE_NEHALEM
  105. endif
  106. ifeq ($(TARGET), COOPERLAKE)
  107. GETARCH_FLAGS := -DFORCE_NEHALEM
  108. endif
  109. ifeq ($(TARGET), SAPPHIRERAPIDS)
  110. GETARCH_FLAGS := -DFORCE_NEHALEM
  111. endif
  112. ifeq ($(TARGET), SANDYBRIDGE)
  113. GETARCH_FLAGS := -DFORCE_NEHALEM
  114. endif
  115. ifeq ($(TARGET), BULLDOZER)
  116. GETARCH_FLAGS := -DFORCE_BARCELONA
  117. endif
  118. ifeq ($(TARGET), PILEDRIVER)
  119. GETARCH_FLAGS := -DFORCE_BARCELONA
  120. endif
  121. ifeq ($(TARGET), STEAMROLLER)
  122. GETARCH_FLAGS := -DFORCE_BARCELONA
  123. endif
  124. ifeq ($(TARGET), EXCAVATOR)
  125. GETARCH_FLAGS := -DFORCE_BARCELONA
  126. endif
  127. ifeq ($(TARGET), ZEN)
  128. GETARCH_FLAGS := -DFORCE_BARCELONA
  129. endif
  130. ifeq ($(TARGET), ARMV8)
  131. GETARCH_FLAGS := -DFORCE_ARMV7
  132. endif
  133. ifeq ($(TARGET), POWER8)
  134. GETARCH_FLAGS := -DFORCE_POWER6
  135. endif
  136. ifeq ($(TARGET), POWER9)
  137. GETARCH_FLAGS := -DFORCE_POWER6
  138. endif
  139. ifeq ($(TARGET), POWER10)
  140. GETARCH_FLAGS := -DFORCE_POWER6
  141. endif
  142. endif
  143. #TARGET_CORE will override TARGET which is used in DYNAMIC_ARCH=1.
  144. #
  145. ifdef TARGET_CORE
  146. GETARCH_FLAGS := -DFORCE_$(TARGET_CORE)
  147. endif
  148. # Force fallbacks for 32bit
  149. ifeq ($(BINARY), 32)
  150. ifeq ($(TARGET_CORE), HASWELL)
  151. GETARCH_FLAGS := -DFORCE_NEHALEM
  152. endif
  153. ifeq ($(TARGET_CORE), SKYLAKEX)
  154. GETARCH_FLAGS := -DFORCE_NEHALEM
  155. endif
  156. ifeq ($(TARGET_CORE), COOPERLAKE)
  157. GETARCH_FLAGS := -DFORCE_NEHALEM
  158. endif
  159. ifeq ($(TARGET_CORE), SAPPHIRERAPIDS)
  160. GETARCH_FLAGS := -DFORCE_NEHALEM
  161. endif
  162. ifeq ($(TARGET_CORE), SANDYBRIDGE)
  163. GETARCH_FLAGS := -DFORCE_NEHALEM
  164. endif
  165. ifeq ($(TARGET_CORE), BULLDOZER)
  166. GETARCH_FLAGS := -DFORCE_BARCELONA
  167. endif
  168. ifeq ($(TARGET_CORE), PILEDRIVER)
  169. GETARCH_FLAGS := -DFORCE_BARCELONA
  170. endif
  171. ifeq ($(TARGET_CORE), STEAMROLLER)
  172. GETARCH_FLAGS := -DFORCE_BARCELONA
  173. endif
  174. ifeq ($(TARGET_CORE), EXCAVATOR)
  175. GETARCH_FLAGS := -DFORCE_BARCELONA
  176. endif
  177. ifeq ($(TARGET_CORE), ZEN)
  178. GETARCH_FLAGS := -DFORCE_BARCELONA
  179. endif
  180. endif
  181. # On x86_64 build getarch with march=native unless the compiler is PGI. This is required to detect AVX512 support in getarch.
  182. ifeq ($(HOSTARCH), x86_64)
  183. ifeq ($(findstring pgcc,$(HOSTCC))$(findstring nvc,$(HOSTCC)),)
  184. GETARCH_FLAGS += -march=native
  185. endif
  186. endif
  187. ifdef INTERFACE64
  188. ifneq ($(INTERFACE64), 0)
  189. GETARCH_FLAGS += -DUSE64BITINT
  190. endif
  191. endif
  192. ifndef GEMM_MULTITHREAD_THRESHOLD
  193. GEMM_MULTITHREAD_THRESHOLD=4
  194. endif
  195. GETARCH_FLAGS += -DGEMM_MULTITHREAD_THRESHOLD=$(GEMM_MULTITHREAD_THRESHOLD)
  196. ifeq ($(NO_AVX), 1)
  197. GETARCH_FLAGS += -DNO_AVX
  198. endif
  199. ifeq ($(BINARY), 32)
  200. GETARCH_FLAGS += -DNO_AVX -DNO_AVX2 -DNO_AVX512
  201. NO_AVX512 = 1
  202. endif
  203. ifeq ($(NO_AVX2), 1)
  204. GETARCH_FLAGS += -DNO_AVX2
  205. endif
  206. ifeq ($(NO_AVX512), 1)
  207. GETARCH_FLAGS += -DNO_AVX512
  208. endif
  209. ifeq ($(DEBUG), 1)
  210. GETARCH_FLAGS += -g
  211. endif
  212. ifeq ($(QUIET_MAKE), 1)
  213. MAKE += -s
  214. endif
  215. ifndef NO_PARALLEL_MAKE
  216. NO_PARALLEL_MAKE=0
  217. endif
  218. GETARCH_FLAGS += -DNO_PARALLEL_MAKE=$(NO_PARALLEL_MAKE)
  219. ifdef MAKE_NB_JOBS
  220. GETARCH_FLAGS += -DMAKE_NB_JOBS=$(MAKE_NB_JOBS)
  221. endif
  222. ifeq ($(HOSTCC), loongcc)
  223. GETARCH_FLAGS += -static
  224. endif
  225. #if don't use Fortran, it will only compile CBLAS.
  226. ifeq ($(ONLY_CBLAS), 1)
  227. NO_LAPACK = 1
  228. else
  229. ONLY_CBLAS = 0
  230. endif
  231. #For small matrix optimization
  232. ifeq ($(ARCH), x86_64)
  233. SMALL_MATRIX_OPT = 1
  234. else ifeq ($(ARCH), power)
  235. SMALL_MATRIX_OPT = 1
  236. BUILD_BFLOAT16 = 1
  237. endif
  238. ifeq ($(SMALL_MATRIX_OPT), 1)
  239. CCOMMON_OPT += -DSMALL_MATRIX_OPT
  240. endif
  241. # This operation is expensive, so execution should be once.
  242. ifndef GOTOBLAS_MAKEFILE
  243. export GOTOBLAS_MAKEFILE = 1
  244. # Determine if the assembler is GNU Assembler
  245. HAVE_GAS := $(shell $(AS) -v < /dev/null 2>&1 | grep GNU 2>&1 >/dev/null ; echo $$?)
  246. GETARCH_FLAGS += -DHAVE_GAS=$(HAVE_GAS)
  247. # Generating Makefile.conf and config.h
  248. 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) DYNAMIC_ARCH=$(DYNAMIC_ARCH) TARGET_CORE=$(TARGET_CORE) ONLY_CBLAS=$(ONLY_CBLAS) TARGET=$(TARGET) all)
  249. endif
  250. ifndef TARGET_CORE
  251. -include $(TOPDIR)/Makefile.conf
  252. else
  253. HAVE_NEON=
  254. HAVE_VFP=
  255. HAVE_VFPV3=
  256. HAVE_VFPV4=
  257. HAVE_MMX=
  258. HAVE_SSE=
  259. HAVE_SSE2=
  260. HAVE_SSE3=
  261. HAVE_SSSE3=
  262. HAVE_SSE4_1=
  263. HAVE_SSE4_2=
  264. HAVE_SSE4A=
  265. HAVE_SSE5=
  266. HAVE_AVX=
  267. HAVE_AVX2=
  268. HAVE_FMA3=
  269. include $(TOPDIR)/Makefile_kernel.conf
  270. endif
  271. ifndef NUM_PARALLEL
  272. NUM_PARALLEL = 1
  273. endif
  274. ifndef NUM_THREADS
  275. NUM_THREADS = $(NUM_CORES)
  276. endif
  277. ifeq ($(NUM_THREADS), 1)
  278. override USE_THREAD = 0
  279. override USE_OPENMP = 0
  280. endif
  281. ifdef USE_THREAD
  282. ifeq ($(USE_THREAD), 0)
  283. SMP =
  284. else
  285. SMP = 1
  286. endif
  287. else
  288. ifeq ($(NUM_THREADS), 1)
  289. SMP =
  290. else
  291. SMP = 1
  292. endif
  293. endif
  294. ifeq ($(SMP), 1)
  295. USE_LOCKING =
  296. endif
  297. ifndef NEED_PIC
  298. NEED_PIC = 1
  299. endif
  300. ARFLAGS =
  301. CPP = $(COMPILER) -E
  302. AR ?= $(CROSS_SUFFIX)ar
  303. AS ?= $(CROSS_SUFFIX)as
  304. LD ?= $(CROSS_SUFFIX)ld
  305. RANLIB ?= $(CROSS_SUFFIX)ranlib
  306. NM = $(CROSS_SUFFIX)nm
  307. DLLWRAP = $(CROSS_SUFFIX)dllwrap
  308. OBJCOPY = $(CROSS_SUFFIX)objcopy
  309. OBJCONV = $(CROSS_SUFFIX)objconv
  310. # When fortran support was either not detected or actively deselected, only build BLAS.
  311. ifeq ($(NOFORTRAN), 1)
  312. C_LAPACK = 1
  313. override FEXTRALIB =
  314. endif
  315. ifeq ($(C_COMPILER), GCC)
  316. GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
  317. GCCVERSIONGT4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 4)
  318. GCCVERSIONGT5 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 5)
  319. GCCVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 7)
  320. GCCVERSIONGTEQ8 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 8)
  321. GCCVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  322. GCCVERSIONGTEQ11 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 11)
  323. GCCVERSIONGTEQ10 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 10)
  324. # Note that the behavior of -dumpversion is compile-time-configurable for
  325. # gcc-7.x and newer. Use -dumpfullversion there
  326. ifeq ($(GCCVERSIONGTEQ7),1)
  327. GCCDUMPVERSION_PARAM := -dumpfullversion
  328. else
  329. GCCDUMPVERSION_PARAM := -dumpversion
  330. endif
  331. GCCMINORVERSIONGTEQ1 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 1)
  332. GCCMINORVERSIONGTEQ2 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 2)
  333. GCCMINORVERSIONGTEQ4 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 4)
  334. GCCMINORVERSIONGTEQ7 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 7)
  335. endif
  336. ifeq ($(C_COMPILER), CLANG)
  337. CLANGVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  338. CLANGVERSIONGTEQ12 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 12)
  339. endif
  340. #
  341. # OS dependent settings
  342. #
  343. ifeq ($(OSNAME), Darwin)
  344. ifndef MACOSX_DEPLOYMENT_TARGET
  345. ifeq ($(ARCH), arm64)
  346. export MACOSX_DEPLOYMENT_TARGET=11.0
  347. ifeq ($(C_COMPILER), GCC)
  348. export NO_SVE = 1
  349. endif
  350. else
  351. export MACOSX_DEPLOYMENT_TARGET=10.8
  352. endif
  353. endif
  354. MD5SUM = md5 -r
  355. XCVER = $(shell pkgutil --pkg-info=com.apple.pkg.Xcode |awk '/version:/ {print $2}'|cut -d: -f2|cut -f1 -d.)
  356. ifeq (x$(XCVER)x,xx)
  357. XCVER = $(shell pkgutil --pkg-info=com.apple.pkg.CLTools_Executables |awk '/version:/ {print $2}'|cut -d: -f2|cut -f1 -d.)
  358. endif
  359. ifeq (x$(XCVER), x 15)
  360. CCOMMON_OPT += -Wl,-ld_classic
  361. endif
  362. endif
  363. ifneq (,$(findstring $(OSNAME), FreeBSD OpenBSD DragonFly))
  364. MD5SUM = md5 -r
  365. endif
  366. ifeq ($(OSNAME), NetBSD)
  367. MD5SUM = md5 -n
  368. endif
  369. ifeq ($(OSNAME), Linux)
  370. EXTRALIB += -lm
  371. NO_EXPRECISION = 1
  372. endif
  373. ifeq ($(OSNAME), Android)
  374. EXTRALIB += -lm
  375. endif
  376. ifeq ($(OSNAME), AIX)
  377. EXTRALIB += -lm
  378. endif
  379. ifeq ($(OSNAME), $(filter $(OSNAME),FreeBSD OpenBSD NetBSD DragonFly))
  380. ifeq ($(ARCH), $(filter $(ARCH),arm arm64))
  381. EXTRALIB += -lm
  382. endif
  383. endif
  384. ifeq ($(OSNAME), WINNT)
  385. NEED_PIC = 0
  386. NO_EXPRECISION = 1
  387. EXTRALIB += -defaultlib:advapi32
  388. SUFFIX = obj
  389. PSUFFIX = pobj
  390. LIBSUFFIX = a
  391. ifeq ($(C_COMPILER), CLANG)
  392. CCOMMON_OPT += -DMS_ABI
  393. endif
  394. #Version tests for supporting specific features (MS_ABI, POWER9 intrinsics)
  395. ifeq ($(GCCVERSIONGT4), 1)
  396. # GCC Major version > 4
  397. # It is compatible with MSVC ABI.
  398. CCOMMON_OPT += -DMS_ABI
  399. endif
  400. ifeq ($(GCCVERSIONGTEQ4), 1)
  401. ifeq ($(GCCMINORVERSIONGTEQ7), 1)
  402. # GCC Version >=4.7
  403. # It is compatible with MSVC ABI.
  404. CCOMMON_OPT += -DMS_ABI
  405. endif
  406. endif
  407. # Ensure the correct stack alignment on Win32
  408. # http://permalink.gmane.org/gmane.comp.lib.openblas.general/97
  409. ifeq ($(ARCH), x86)
  410. CCOMMON_OPT += -mincoming-stack-boundary=2
  411. FCOMMON_OPT += -mincoming-stack-boundary=2
  412. endif
  413. endif
  414. ifeq ($(OSNAME), Interix)
  415. NEED_PIC = 0
  416. NO_EXPRECISION = 1
  417. INTERIX_TOOL_DIR = /opt/gcc.3.3/i586-pc-interix3/bin
  418. endif
  419. ifeq ($(OSNAME), CYGWIN_NT)
  420. NEED_PIC = 0
  421. NO_EXPRECISION = 1
  422. OS_CYGWIN_NT = 1
  423. endif
  424. ifneq ($(OSNAME), WINNT)
  425. ifneq ($(OSNAME), CYGWIN_NT)
  426. ifneq ($(OSNAME), Interix)
  427. ifneq ($(OSNAME), Android)
  428. ifdef SMP
  429. EXTRALIB += -lpthread
  430. endif
  431. endif
  432. endif
  433. endif
  434. endif
  435. # ifeq logical or
  436. ifeq ($(OSNAME), $(filter $(OSNAME),WINNT CYGWIN_NT Interix))
  437. OS_WINDOWS=1
  438. endif
  439. ifdef QUAD_PRECISION
  440. CCOMMON_OPT += -DQUAD_PRECISION
  441. NO_EXPRECISION = 1
  442. endif
  443. ifneq ($(ARCH), x86)
  444. ifneq ($(ARCH), x86_64)
  445. NO_EXPRECISION = 1
  446. endif
  447. endif
  448. ifdef UTEST_CHECK
  449. CCOMMON_OPT += -DUTEST_CHECK
  450. SANITY_CHECK = 1
  451. endif
  452. ifdef SANITY_CHECK
  453. CCOMMON_OPT += -DSANITY_CHECK -DREFNAME=$(*F)f$(BU)
  454. endif
  455. MAX_STACK_ALLOC ?= 2048
  456. ifneq ($(MAX_STACK_ALLOC), 0)
  457. CCOMMON_OPT += -DMAX_STACK_ALLOC=$(MAX_STACK_ALLOC)
  458. endif
  459. ifdef USE_LOCKING
  460. ifneq ($(USE_LOCKING), 0)
  461. CCOMMON_OPT += -DUSE_LOCKING
  462. endif
  463. endif
  464. #
  465. # Architecture dependent settings
  466. #
  467. ifeq ($(ARCH), x86)
  468. ifndef BINARY
  469. NO_BINARY_MODE = 1
  470. endif
  471. ifeq ($(CORE), generic)
  472. NO_EXPRECISION = 1
  473. endif
  474. ifndef NO_EXPRECISION
  475. ifeq ($(F_COMPILER), GFORTRAN)
  476. # ifeq logical or. GCC or LSB
  477. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  478. EXPRECISION = 1
  479. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  480. FCOMMON_OPT += -m128bit-long-double
  481. endif
  482. ifeq ($(C_COMPILER), CLANG)
  483. EXPRECISION = 1
  484. CCOMMON_OPT += -DEXPRECISION
  485. FCOMMON_OPT += -m128bit-long-double
  486. endif
  487. endif
  488. endif
  489. endif
  490. ifeq ($(ARCH), x86_64)
  491. ifeq ($(CORE), generic)
  492. NO_EXPRECISION = 1
  493. endif
  494. ifndef NO_EXPRECISION
  495. ifeq ($(F_COMPILER), GFORTRAN)
  496. # ifeq logical or. GCC or LSB
  497. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  498. EXPRECISION = 1
  499. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  500. FCOMMON_OPT += -m128bit-long-double
  501. endif
  502. ifeq ($(C_COMPILER), CLANG)
  503. EXPRECISION = 1
  504. CCOMMON_OPT += -DEXPRECISION
  505. FCOMMON_OPT += -m128bit-long-double
  506. endif
  507. endif
  508. endif
  509. endif
  510. ifeq ($(C_COMPILER), INTEL)
  511. CCOMMON_OPT += -wd981
  512. endif
  513. ifeq ($(USE_OPENMP), 1)
  514. #check
  515. ifeq ($(USE_THREAD), 0)
  516. $(error OpenBLAS: Cannot set both USE_OPENMP=1 and USE_THREAD=0. The USE_THREAD=0 is only for building single thread version.)
  517. endif
  518. # ifeq logical or. GCC or LSB
  519. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  520. CCOMMON_OPT += -fopenmp
  521. endif
  522. ifeq ($(C_COMPILER), CLANG)
  523. CCOMMON_OPT += -fopenmp
  524. endif
  525. ifeq ($(C_COMPILER), INTEL)
  526. CCOMMON_OPT += -fopenmp
  527. endif
  528. ifeq ($(C_COMPILER), PGI)
  529. CCOMMON_OPT += -mp
  530. endif
  531. ifeq ($(C_COMPILER), OPEN64)
  532. CCOMMON_OPT += -mp
  533. CEXTRALIB += -lstdc++
  534. endif
  535. ifeq ($(C_COMPILER), PATHSCALE)
  536. CCOMMON_OPT += -mp
  537. endif
  538. endif
  539. ifeq ($(DYNAMIC_ARCH), 1)
  540. ifeq ($(ARCH), x86)
  541. DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
  542. CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
  543. endif
  544. ifeq ($(ARCH), x86_64)
  545. DYNAMIC_CORE = PRESCOTT CORE2
  546. ifeq ($(DYNAMIC_OLDER), 1)
  547. DYNAMIC_CORE += PENRYN DUNNINGTON
  548. endif
  549. DYNAMIC_CORE += NEHALEM
  550. ifeq ($(DYNAMIC_OLDER), 1)
  551. DYNAMIC_CORE += OPTERON OPTERON_SSE3
  552. endif
  553. DYNAMIC_CORE += BARCELONA
  554. ifeq ($(DYNAMIC_OLDER), 1)
  555. DYNAMIC_CORE += BOBCAT ATOM NANO
  556. endif
  557. ifneq ($(NO_AVX), 1)
  558. DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER STEAMROLLER EXCAVATOR
  559. endif
  560. ifneq ($(NO_AVX2), 1)
  561. DYNAMIC_CORE += HASWELL ZEN
  562. endif
  563. ifneq ($(NO_AVX512), 1)
  564. ifneq ($(NO_AVX2), 1)
  565. DYNAMIC_CORE += SKYLAKEX COOPERLAKE SAPPHIRERAPIDS
  566. endif
  567. endif
  568. endif
  569. ifdef DYNAMIC_LIST
  570. override DYNAMIC_CORE = PRESCOTT $(DYNAMIC_LIST)
  571. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_PRESCOTT
  572. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  573. CCOMMON_OPT += $(XCCOMMON_OPT)
  574. #CCOMMON_OPT += -DDYNAMIC_LIST='$(DYNAMIC_LIST)'
  575. endif
  576. ifeq ($(ARCH), arm64)
  577. DYNAMIC_CORE = ARMV8
  578. DYNAMIC_CORE += CORTEXA53
  579. DYNAMIC_CORE += CORTEXA57
  580. DYNAMIC_CORE += CORTEXA72
  581. DYNAMIC_CORE += CORTEXA73
  582. DYNAMIC_CORE += NEOVERSEN1
  583. ifneq ($(NO_SVE), 1)
  584. DYNAMIC_CORE += NEOVERSEV1
  585. DYNAMIC_CORE += NEOVERSEN2
  586. DYNAMIC_CORE += ARMV8SVE
  587. endif
  588. DYNAMIC_CORE += CORTEXA55
  589. DYNAMIC_CORE += FALKOR
  590. DYNAMIC_CORE += THUNDERX
  591. DYNAMIC_CORE += THUNDERX2T99
  592. DYNAMIC_CORE += TSV110
  593. DYNAMIC_CORE += EMAG8180
  594. DYNAMIC_CORE += THUNDERX3T110
  595. ifdef DYNAMIC_LIST
  596. override DYNAMIC_CORE = ARMV8 $(DYNAMIC_LIST)
  597. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_ARMV8
  598. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  599. endif
  600. endif
  601. ifeq ($(ARCH), mips64)
  602. DYNAMIC_CORE = LOONGSON3R3 LOONGSON3R4 MIPS64_GENERIC
  603. ifdef DYNAMIC_LIST
  604. override DYNAMIC_CORE = MIPS64_GENERIC $(DYNAMIC_LIST)
  605. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_MIPS64_GENERIC
  606. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  607. endif
  608. endif
  609. ifeq ($(ARCH), loongarch64)
  610. DYNAMIC_CORE = LOONGSON3R5 LOONGSON2K1000 LOONGSONGENERIC
  611. endif
  612. ifeq ($(ARCH), zarch)
  613. DYNAMIC_CORE = ZARCH_GENERIC
  614. # if the compiler accepts -march=arch11 or -march=z13 and can compile a file
  615. # with z13-specific inline assembly, then we can include support for Z13.
  616. # note: -march=z13 is equivalent to -march=arch11 yet some compiler releases
  617. # only support one or the other.
  618. # note: LLVM version 6.x supported -march=z13 yet could not handle vector
  619. # registers in inline assembly, so the check for supporting the -march flag is
  620. # not enough.
  621. ZARCH_TEST_COMPILE=-c $(TOPDIR)/kernel/zarch/damin_z13.c -I$(TOPDIR) -o /dev/null > /dev/null 2> /dev/null
  622. ZARCH_CC_SUPPORTS_ARCH11=$(shell $(CC) -march=arch11 $(ZARCH_TEST_COMPILE) && echo 1)
  623. ZARCH_CC_SUPPORTS_Z13=$(shell $(CC) -march=z13 $(ZARCH_TEST_COMPILE) && echo 1)
  624. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH11), $(ZARCH_CC_SUPPORTS_Z13)), 1)
  625. DYNAMIC_CORE += Z13
  626. CCOMMON_OPT += -DDYN_Z13
  627. else
  628. $(info OpenBLAS: Not building Z13 kernels because the compiler $(CC) does not support it)
  629. endif
  630. # as above for z13, check for -march=arch12 and z14 support in the compiler.
  631. ZARCH_CC_SUPPORTS_ARCH12=$(shell $(CC) -march=arch12 $(ZARCH_TEST_COMPILE) && echo 1)
  632. ZARCH_CC_SUPPORTS_Z14=$(shell $(CC) -march=z14 $(ZARCH_TEST_COMPILE) && echo 1)
  633. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH12), $(ZARCH_CC_SUPPORTS_Z14)), 1)
  634. DYNAMIC_CORE += Z14
  635. CCOMMON_OPT += -DDYN_Z14
  636. else
  637. $(info OpenBLAS: Not building Z14 kernels because the compiler $(CC) does not support it)
  638. endif
  639. endif # ARCH zarch
  640. ifeq ($(ARCH), power)
  641. ifneq ($(C_COMPILER), PGI)
  642. DYNAMIC_CORE = POWER6
  643. DYNAMIC_CORE += POWER8
  644. ifneq ($(C_COMPILER), GCC)
  645. DYNAMIC_CORE += POWER9
  646. DYNAMIC_CORE += POWER10
  647. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  648. endif
  649. ifeq ($(C_COMPILER), GCC)
  650. ifeq ($(GCCVERSIONGT5), 1)
  651. DYNAMIC_CORE += POWER9
  652. else
  653. $(info, OpenBLAS: Your gcc version is too old to build the POWER9 kernels.)
  654. endif
  655. LDVERSIONGTEQ35 := $(shell expr `$(CC) -Wl,--version 2> /dev/null | head -1 | cut -f2 -d "." | cut -f1 -d "-"` \>= 35)
  656. ifeq ($(GCCVERSIONGTEQ11)$(LDVERSIONGTEQ35), 11)
  657. DYNAMIC_CORE += POWER10
  658. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  659. else ifeq ($(GCCVERSIONGTEQ10), 1)
  660. ifeq ($(GCCMINORVERSIONGTEQ2)$(LDVERSIONGTEQ35), 11)
  661. DYNAMIC_CORE += POWER10
  662. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  663. endif
  664. else
  665. $(info, OpenBLAS: Your gcc version is too old to build the POWER10 kernels.)
  666. endif
  667. endif
  668. else
  669. DYNAMIC_CORE = POWER8
  670. DYNAMIC_CORE += POWER9
  671. endif
  672. endif
  673. # If DYNAMIC_CORE is not set, DYNAMIC_ARCH cannot do anything, so force it to empty
  674. ifndef DYNAMIC_CORE
  675. override DYNAMIC_ARCH=
  676. endif
  677. endif
  678. ifeq ($(ARCH), ia64)
  679. NO_BINARY_MODE = 1
  680. BINARY_DEFINED = 1
  681. ifeq ($(F_COMPILER), GFORTRAN)
  682. ifeq ($(C_COMPILER), GCC)
  683. # EXPRECISION = 1
  684. # CCOMMON_OPT += -DEXPRECISION
  685. endif
  686. endif
  687. endif
  688. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  689. NO_BINARY_MODE = 1
  690. endif
  691. ifeq ($(ARCH), alpha)
  692. NO_BINARY_MODE = 1
  693. BINARY_DEFINED = 1
  694. endif
  695. ifeq ($(ARCH), arm)
  696. NO_BINARY_MODE = 1
  697. BINARY_DEFINED = 1
  698. CCOMMON_OPT += -marm
  699. FCOMMON_OPT += -marm
  700. # If softfp abi is mentioned on the command line, force it.
  701. ifeq ($(ARM_SOFTFP_ABI), 1)
  702. CCOMMON_OPT += -mfloat-abi=softfp
  703. FCOMMON_OPT += -mfloat-abi=softfp
  704. endif
  705. ifeq ($(OSNAME), Android)
  706. ifeq ($(ARM_SOFTFP_ABI), 1)
  707. EXTRALIB += -lm
  708. else
  709. EXTRALIB += -Wl,-lm_hard
  710. endif
  711. endif
  712. endif
  713. ifeq ($(ARCH), arm64)
  714. NO_BINARY_MODE = 1
  715. BINARY_DEFINED = 1
  716. ifdef INTERFACE64
  717. ifneq ($(INTERFACE64), 0)
  718. ifeq ($(F_COMPILER), GFORTRAN)
  719. FCOMMON_OPT += -fdefault-integer-8
  720. endif
  721. ifeq ($(F_COMPILER), FLANG)
  722. FCOMMON_OPT += -i8
  723. endif
  724. endif
  725. endif
  726. endif
  727. ifeq ($(ARCH), riscv64)
  728. NO_BINARY_MODE = 1
  729. BINARY_DEFINED = 1
  730. ifdef INTERFACE64
  731. ifneq ($(INTERFACE64), 0)
  732. ifeq ($(F_COMPILER), GFORTRAN)
  733. FCOMMON_OPT += -fdefault-integer-8
  734. endif
  735. ifeq ($(F_COMPILER), FLANG)
  736. FCOMMON_OPT += -i8
  737. endif
  738. endif
  739. endif
  740. endif
  741. ifeq ($(ARCH), loongarch64)
  742. NO_BINARY_MODE = 1
  743. BINARY_DEFINED = 1
  744. ifdef INTERFACE64
  745. ifneq ($(INTERFACE64), 0)
  746. ifeq ($(F_COMPILER), GFORTRAN)
  747. FCOMMON_OPT += -fdefault-integer-8
  748. endif
  749. ifeq ($(F_COMPILER), FLANG)
  750. FCOMMON_OPT += -i8
  751. endif
  752. endif
  753. endif
  754. endif
  755. #
  756. # C Compiler dependent settings
  757. #
  758. # ifeq logical or. GCC or CLANG or LSB
  759. # http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
  760. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG LSB))
  761. CCOMMON_OPT += -Wall
  762. COMMON_PROF += -fno-inline
  763. NO_UNINITIALIZED_WARN = -Wno-uninitialized
  764. ifeq ($(QUIET_MAKE), 1)
  765. CCOMMON_OPT += $(NO_UNINITIALIZED_WARN) -Wno-unused
  766. endif
  767. ifdef NO_BINARY_MODE
  768. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  769. ifdef BINARY64
  770. CCOMMON_OPT += -mabi=64
  771. else
  772. CCOMMON_OPT += -mabi=n32
  773. endif
  774. BINARY_DEFINED = 1
  775. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  776. CCOMMON_OPT += -mabi=32
  777. BINARY_DEFINED = 1
  778. endif
  779. ifneq (, $(filter $(CORE), MIPS64_GENERIC))
  780. CCOMMON_OPT += -DNO_MSA
  781. FCOMMON_OPT += -DNO_MSA
  782. endif
  783. ifneq (, $(filter $(CORE),LOONGSON3R3 LOONGSON3R4))
  784. CCOMMON_OPT += -march=loongson3a
  785. FCOMMON_OPT += -march=loongson3a
  786. endif
  787. ifeq ($(CORE), MIPS24K)
  788. CCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  789. FCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  790. endif
  791. ifeq ($(CORE), MIPS1004K)
  792. CCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  793. FCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  794. endif
  795. ifeq ($(CORE), P5600)
  796. CCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  797. FCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  798. endif
  799. ifeq ($(CORE), I6400)
  800. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  801. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  802. endif
  803. ifeq ($(CORE), P6600)
  804. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  805. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  806. endif
  807. ifeq ($(CORE), I6500)
  808. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  809. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  810. endif
  811. ifeq ($(OSNAME), AIX)
  812. BINARY_DEFINED = 1
  813. endif
  814. ifeq ($(ARCH), loongarch64)
  815. LA64_ABI=$(shell $(CC) -mabi=lp64d -c $(TOPDIR)/cpuid_loongarch64.c -o /dev/null > /dev/null 2> /dev/null && echo lp64d)
  816. ifneq ($(LA64_ABI), lp64d)
  817. LA64_ABI=lp64
  818. endif
  819. CCOMMON_OPT += -march=loongarch64 -mabi=$(LA64_ABI)
  820. FCOMMON_OPT += -march=loongarch64 -mabi=$(LA64_ABI)
  821. endif
  822. endif
  823. ifndef BINARY_DEFINED
  824. ifneq ($(OSNAME), AIX)
  825. ifdef BINARY64
  826. ifneq ($(ARCH), riscv64)
  827. CCOMMON_OPT += -m64
  828. endif
  829. else
  830. CCOMMON_OPT += -m32
  831. endif
  832. endif
  833. endif
  834. endif
  835. ifeq ($(C_COMPILER), PGI)
  836. PGCVERSIONGT20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \> 20)
  837. PGCVERSIONEQ20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` == 20)
  838. PGCMINORVERSIONGE11 := $(shell expr `$(CC) --version|sed -n "2p" |cut -d "-" -f 1 |sed -e "s/[^0-9.]//g" |cut -c 4-5` \>= 11)
  839. PGCVERSIONCHECK := $(PGCVERSIONGT20)$(PGCVERSIONEQ20)$(PGCMINORVERSIONGE11)
  840. ifeq ($(PGCVERSIONCHECK), $(filter $(PGCVERSIONCHECK), 100 101 011))
  841. NEWPGI := 1
  842. PGCVERSIONGT21 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \> 21)
  843. PGCVERSIONEQ21 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` == 21)
  844. PGCVERSIONCHECK2 := $(PGCVERSIONGT21)$(PGCVERSIONEQ21)$(PGCMINORVERSIONGE11)
  845. ifeq ($(PGCVERSIONCHECK2), $(filter $(PGCVERSIONCHECK2), 100 101 011))
  846. NEWPGI2 := 1
  847. endif
  848. endif
  849. ifdef BINARY64
  850. ifeq ($(ARCH), x86_64)
  851. ifeq (,$(findstring tp,$(CFLAGS)))
  852. ifneq ($(NEWPGI2),1)
  853. CCOMMON_OPT += -tp p7-64
  854. else
  855. CCOMMON_OPT += -tp px
  856. endif
  857. endif
  858. ifneq ($(NEWPGI),1)
  859. CCOMMON_OPT += -D__MMX__ -Mnollvm
  860. endif
  861. else
  862. ifeq ($(ARCH), power)
  863. ifeq (,$(findstring tp,$(CFLAGS)))
  864. ifeq ($(CORE), POWER8)
  865. CCOMMON_OPT += -tp pwr8
  866. endif
  867. ifeq ($(CORE), POWER9)
  868. CCOMMON_OPT += -tp pwr9
  869. endif
  870. endif
  871. endif
  872. endif
  873. else
  874. ifneq ($(NEWPGI2),1)
  875. ifeq (,$(findstring tp,$(CFLAGS)))
  876. CCOMMON_OPT += -tp p7
  877. else
  878. CCOMMON_OPT += -tp px
  879. endif
  880. endif
  881. endif
  882. endif
  883. ifeq ($(C_COMPILER), PATHSCALE)
  884. ifdef BINARY64
  885. CCOMMON_OPT += -m64
  886. else
  887. CCOMMON_OPT += -m32
  888. endif
  889. endif
  890. #
  891. # Fortran Compiler dependent settings
  892. #
  893. ifeq ($(F_COMPILER), NAG)
  894. FCOMMON_OPT += -dcfuns -recursive -ieee=full -w=obs -thread_safe
  895. ifdef INTERFACE64
  896. ifneq ($(INTERFACE64), 0)
  897. FCOMMON_OPT += -i8
  898. endif
  899. endif
  900. ifeq ($(USE_OPENMP), 1)
  901. FCOMMON_OPT += -openmp
  902. endif
  903. endif
  904. ifeq ($(F_COMPILER), FLANG)
  905. CCOMMON_OPT += -DF_INTERFACE_FLANG
  906. FCOMMON_OPT += -Mrecursive -Kieee
  907. ifeq ($(OSNAME), Linux)
  908. ifeq ($(ARCH), x86_64)
  909. FLANG_VENDOR := $(shell $(FC) --version|head -1 |cut -f 1 -d " ")
  910. ifeq ($(FLANG_VENDOR), AMD)
  911. FCOMMON_OPT += -fno-unroll-loops
  912. endif
  913. endif
  914. endif
  915. ifdef BINARY64
  916. ifdef INTERFACE64
  917. ifneq ($(INTERFACE64), 0)
  918. FCOMMON_OPT += -i8
  919. endif
  920. endif
  921. FCOMMON_OPT += -Wall
  922. else
  923. FCOMMON_OPT += -Wall
  924. endif
  925. ifeq ($(USE_OPENMP), 1)
  926. FCOMMON_OPT += -fopenmp
  927. endif
  928. endif
  929. ifeq ($(F_COMPILER), G77)
  930. CCOMMON_OPT += -DF_INTERFACE_G77
  931. FCOMMON_OPT += -Wall
  932. ifndef NO_BINARY_MODE
  933. ifneq ($(OSNAME), AIX)
  934. ifdef BINARY64
  935. FCOMMON_OPT += -m64
  936. else
  937. FCOMMON_OPT += -m32
  938. endif
  939. endif
  940. endif
  941. endif
  942. ifeq ($(F_COMPILER), G95)
  943. CCOMMON_OPT += -DF_INTERFACE_G95
  944. FCOMMON_OPT += -Wall
  945. ifneq ($(OSNAME), AIX)
  946. ifndef NO_BINARY_MODE
  947. ifdef BINARY64
  948. FCOMMON_OPT += -m64
  949. else
  950. FCOMMON_OPT += -m32
  951. endif
  952. endif
  953. ifneq ($(NO_LAPACKE), 1)
  954. FCOMMON_OPT += -fno-second-underscore
  955. endif
  956. endif
  957. endif
  958. ifeq ($(F_COMPILER), $(filter $(F_COMPILER),GFORTRAN FLANGNEW))
  959. CCOMMON_OPT += -DF_INTERFACE_GFORT
  960. ifeq ($(F_COMPILER), GFORTRAN)
  961. FCOMMON_OPT += -Wall
  962. # make single-threaded LAPACK calls thread-safe #1847
  963. FCOMMON_OPT += -frecursive
  964. # work around ABI problem with passing single-character arguments
  965. FCOMMON_OPT += -fno-optimize-sibling-calls
  966. #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
  967. ifneq ($(NOFORTRAN), 1)
  968. ifneq ($(NOFORTRAN), 2)
  969. ifneq ($(NO_LAPACK), 1)
  970. EXTRALIB += -lgfortran
  971. endif
  972. endif
  973. endif
  974. endif
  975. ifdef NO_BINARY_MODE
  976. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  977. ifdef BINARY64
  978. FCOMMON_OPT += -mabi=64
  979. else
  980. FCOMMON_OPT += -mabi=n32
  981. endif
  982. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  983. FCOMMON_OPT += -mabi=32
  984. endif
  985. else
  986. ifdef BINARY64
  987. ifneq ($(OSNAME), AIX)
  988. ifneq ($(ARCH), riscv64)
  989. FCOMMON_OPT += -m64
  990. endif
  991. endif
  992. ifdef INTERFACE64
  993. ifneq ($(INTERFACE64), 0)
  994. FCOMMON_OPT += -fdefault-integer-8
  995. endif
  996. endif
  997. else
  998. ifneq ($(OSNAME), AIX)
  999. FCOMMON_OPT += -m32
  1000. endif
  1001. endif
  1002. endif
  1003. ifeq ($(USE_OPENMP), 1)
  1004. FCOMMON_OPT += -fopenmp
  1005. endif
  1006. endif
  1007. ifeq ($(F_COMPILER), INTEL)
  1008. CCOMMON_OPT += -DF_INTERFACE_INTEL
  1009. ifdef INTERFACE64
  1010. ifneq ($(INTERFACE64), 0)
  1011. FCOMMON_OPT += -i8
  1012. endif
  1013. endif
  1014. FCOMMON_OPT += -recursive -fp-model strict -assume protect-parens
  1015. ifeq ($(USE_OPENMP), 1)
  1016. FCOMMON_OPT += -fopenmp
  1017. endif
  1018. endif
  1019. ifeq ($(F_COMPILER), FUJITSU)
  1020. CCOMMON_OPT += -DF_INTERFACE_FUJITSU
  1021. ifeq ($(USE_OPENMP), 1)
  1022. FCOMMON_OPT += -openmp
  1023. endif
  1024. endif
  1025. ifeq ($(F_COMPILER), IBM)
  1026. CCOMMON_OPT += -DF_INTERFACE_IBM
  1027. FEXTRALIB += -lxlf90
  1028. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG))
  1029. FCOMMON_OPT += -qextname
  1030. endif
  1031. # FCOMMON_OPT += -qarch=440
  1032. ifdef BINARY64
  1033. FCOMMON_OPT += -q64
  1034. ifdef INTERFACE64
  1035. ifneq ($(INTERFACE64), 0)
  1036. FCOMMON_OPT += -qintsize=8
  1037. endif
  1038. endif
  1039. else
  1040. FCOMMON_OPT += -q32
  1041. endif
  1042. ifeq ($(USE_OPENMP), 1)
  1043. FCOMMON_OPT += -openmp
  1044. endif
  1045. endif
  1046. ifeq ($(F_COMPILER), PGI)
  1047. CCOMMON_OPT += -DF_INTERFACE_PGI
  1048. COMMON_PROF += -DPGICOMPILER
  1049. ifdef BINARY64
  1050. ifdef INTERFACE64
  1051. ifneq ($(INTERFACE64), 0)
  1052. FCOMMON_OPT += -i8
  1053. endif
  1054. endif
  1055. ifeq ($(ARCH), x86_64)
  1056. ifneq ($(NEWPGI2),1)
  1057. FCOMMON_OPT += -tp p7-64
  1058. else
  1059. FCOMMON_OPT += -tp px
  1060. endif
  1061. else
  1062. ifeq ($(ARCH), power)
  1063. ifeq ($(CORE), POWER6)
  1064. $(warning NVIDIA HPC compilers do not support POWER6.)
  1065. endif
  1066. ifeq ($(CORE), POWER8)
  1067. FCOMMON_OPT += -tp pwr8
  1068. endif
  1069. ifeq ($(CORE), POWER9)
  1070. FCOMMON_OPT += -tp pwr9
  1071. endif
  1072. ifeq ($(CORE), POWER10)
  1073. $(warning NVIDIA HPC compilers do not support POWER10.)
  1074. endif
  1075. endif
  1076. endif
  1077. else
  1078. FCOMMON_OPT += -tp p7
  1079. endif
  1080. FCOMMON_OPT += -Mrecursive -Kieee
  1081. ifeq ($(USE_OPENMP), 1)
  1082. FCOMMON_OPT += -mp
  1083. endif
  1084. endif
  1085. ifeq ($(F_COMPILER), PATHSCALE)
  1086. CCOMMON_OPT += -DF_INTERFACE_PATHSCALE
  1087. ifdef BINARY64
  1088. ifdef INTERFACE64
  1089. ifneq ($(INTERFACE64), 0)
  1090. FCOMMON_OPT += -i8
  1091. endif
  1092. endif
  1093. endif
  1094. ifeq ($(USE_OPENMP), 1)
  1095. FCOMMON_OPT += -mp
  1096. endif
  1097. endif
  1098. ifeq ($(F_COMPILER), OPEN64)
  1099. CCOMMON_OPT += -DF_INTERFACE_OPEN64
  1100. ifdef BINARY64
  1101. ifdef INTERFACE64
  1102. ifneq ($(INTERFACE64), 0)
  1103. FCOMMON_OPT += -i8
  1104. endif
  1105. endif
  1106. endif
  1107. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  1108. ifndef BINARY64
  1109. FCOMMON_OPT += -n32
  1110. else
  1111. FCOMMON_OPT += -n64
  1112. endif
  1113. ifeq ($(CORE), LOONGSON3R3)
  1114. FCOMMON_OPT += -loongson3 -static
  1115. endif
  1116. ifeq ($(CORE), LOONGSON3R4)
  1117. FCOMMON_OPT += -loongson3 -static
  1118. endif
  1119. else
  1120. ifndef BINARY64
  1121. FCOMMON_OPT += -m32
  1122. else
  1123. FCOMMON_OPT += -m64
  1124. endif
  1125. endif
  1126. ifeq ($(USE_OPENMP), 1)
  1127. FEXTRALIB += -lstdc++
  1128. FCOMMON_OPT += -mp
  1129. endif
  1130. endif
  1131. ifeq ($(C_COMPILER), OPEN64)
  1132. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  1133. ifndef BINARY64
  1134. CCOMMON_OPT += -n32
  1135. else
  1136. CCOMMON_OPT += -n64
  1137. endif
  1138. ifeq ($(CORE), LOONGSON3R3)
  1139. CCOMMON_OPT += -loongson3 -static
  1140. endif
  1141. ifeq ($(CORE), LOONGSON3R4)
  1142. CCOMMON_OPT += -loongson3 -static
  1143. endif
  1144. else
  1145. ifndef BINARY64
  1146. CCOMMON_OPT += -m32
  1147. else
  1148. CCOMMON_OPT += -m64
  1149. endif
  1150. endif
  1151. endif
  1152. ifeq ($(C_COMPILER), SUN)
  1153. CCOMMON_OPT += -w
  1154. ifeq ($(ARCH), x86)
  1155. CCOMMON_OPT += -m32
  1156. else
  1157. ifdef BINARY64
  1158. CCOMMON_OPT += -m64
  1159. else
  1160. CCOMMON_OPT += -m32
  1161. endif
  1162. endif
  1163. endif
  1164. ifeq ($(F_COMPILER), SUN)
  1165. CCOMMON_OPT += -DF_INTERFACE_SUN
  1166. FCOMMON_OPT += -ftrap=%none -xrecursive
  1167. ifeq ($(ARCH), x86)
  1168. FCOMMON_OPT += -m32
  1169. else
  1170. ifdef BINARY64
  1171. FCOMMON_OPT += -m64
  1172. else
  1173. FCOMMON_OPT += -m32
  1174. endif
  1175. endif
  1176. ifeq ($(USE_OPENMP), 1)
  1177. FCOMMON_OPT += -xopenmp=parallel
  1178. endif
  1179. endif
  1180. ifeq ($(F_COMPILER), COMPAQ)
  1181. CCOMMON_OPT += -DF_INTERFACE_COMPAQ
  1182. ifeq ($(USE_OPENMP), 1)
  1183. FCOMMON_OPT += -openmp
  1184. endif
  1185. endif
  1186. ifeq ($(F_COMPILER), CRAY)
  1187. CCOMMON_OPT += -DF_INTERFACE_INTEL
  1188. FCOMMON_OPT += -hnopattern
  1189. ifdef INTERFACE64
  1190. ifneq ($(INTERFACE64), 0)
  1191. FCOMMON_OPT += -s integer64
  1192. endif
  1193. endif
  1194. ifneq ($(USE_OPENMP), 1)
  1195. FCOMMON_OPT += -O noomp
  1196. endif
  1197. endif
  1198. ifdef BINARY64
  1199. ifdef INTERFACE64
  1200. ifneq ($(INTERFACE64), 0)
  1201. CCOMMON_OPT +=
  1202. #-DUSE64BITINT
  1203. endif
  1204. endif
  1205. endif
  1206. ifeq ($(NEED_PIC), 1)
  1207. ifeq ($(C_COMPILER), IBM)
  1208. CCOMMON_OPT += -qpic=large
  1209. else
  1210. CCOMMON_OPT += -fPIC
  1211. endif
  1212. ifeq ($(F_COMPILER), SUN)
  1213. FCOMMON_OPT += -pic
  1214. else ifeq ($(F_COMPILER), NAG)
  1215. FCOMMON_OPT += -PIC
  1216. else
  1217. FCOMMON_OPT += -fPIC
  1218. endif
  1219. endif
  1220. ifeq ($(DYNAMIC_ARCH), 1)
  1221. CCOMMON_OPT += -DDYNAMIC_ARCH
  1222. endif
  1223. ifeq ($(DYNAMIC_OLDER), 1)
  1224. CCOMMON_OPT += -DDYNAMIC_OLDER
  1225. endif
  1226. ifeq ($(C_LAPACK), 1)
  1227. CCOMMON_OPT += -DC_LAPACK
  1228. endif
  1229. ifeq ($(NO_LAPACK), 1)
  1230. CCOMMON_OPT += -DNO_LAPACK
  1231. #Disable LAPACK C interface
  1232. NO_LAPACKE = 1
  1233. endif
  1234. ifeq ($(NO_LAPACKE), 1)
  1235. CCOMMON_OPT += -DNO_LAPACKE
  1236. endif
  1237. ifeq ($(NO_AVX), 1)
  1238. CCOMMON_OPT += -DNO_AVX
  1239. endif
  1240. ifeq ($(ARCH), x86)
  1241. CCOMMON_OPT += -DNO_AVX
  1242. endif
  1243. ifeq ($(NO_AVX2), 1)
  1244. CCOMMON_OPT += -DNO_AVX2
  1245. endif
  1246. ifeq ($(NO_AVX512), 1)
  1247. CCOMMON_OPT += -DNO_AVX512
  1248. endif
  1249. ifeq ($(NO_SVE), 1)
  1250. CCOMMON_OPT += -DNO_SVE
  1251. endif
  1252. ifdef SMP
  1253. CCOMMON_OPT += -DSMP_SERVER
  1254. ifeq ($(ARCH), mips64)
  1255. USE_SIMPLE_THREADED_LEVEL3 = 1
  1256. endif
  1257. ifeq ($(USE_OPENMP), 1)
  1258. # USE_SIMPLE_THREADED_LEVEL3 = 1
  1259. # NO_AFFINITY = 1
  1260. CCOMMON_OPT += -DUSE_OPENMP
  1261. endif
  1262. ifeq ($(BIGNUMA), 1)
  1263. CCOMMON_OPT += -DBIGNUMA
  1264. endif
  1265. endif
  1266. ifeq ($(NO_WARMUP), 1)
  1267. CCOMMON_OPT += -DNO_WARMUP
  1268. endif
  1269. ifeq ($(CONSISTENT_FPCSR), 1)
  1270. CCOMMON_OPT += -DCONSISTENT_FPCSR
  1271. endif
  1272. # Only for development
  1273. # CCOMMON_OPT += -DPARAMTEST
  1274. # CCOMMON_OPT += -DPREFETCHTEST
  1275. # CCOMMON_OPT += -DNO_SWITCHING
  1276. # USE_PAPI = 1
  1277. ifdef USE_PAPI
  1278. CCOMMON_OPT += -DUSE_PAPI
  1279. EXTRALIB += -lpapi -lperfctr
  1280. endif
  1281. ifdef BUFFERSIZE
  1282. CCOMMON_OPT += -DBUFFERSIZE=$(BUFFERSIZE)
  1283. endif
  1284. ifdef DYNAMIC_THREADS
  1285. CCOMMON_OPT += -DDYNAMIC_THREADS
  1286. endif
  1287. CCOMMON_OPT += -DMAX_CPU_NUMBER=$(NUM_THREADS)
  1288. CCOMMON_OPT += -DMAX_PARALLEL_NUMBER=$(NUM_PARALLEL)
  1289. ifdef USE_SIMPLE_THREADED_LEVEL3
  1290. CCOMMON_OPT += -DUSE_SIMPLE_THREADED_LEVEL3
  1291. endif
  1292. ifeq ($(USE_TLS), 1)
  1293. CCOMMON_OPT += -DUSE_TLS
  1294. endif
  1295. ifeq ($(BUILD_BFLOAT16), 1)
  1296. CCOMMON_OPT += -DBUILD_BFLOAT16
  1297. endif
  1298. ifeq ($(BUILD_SINGLE), 1)
  1299. CCOMMON_OPT += -DBUILD_SINGLE=1
  1300. endif
  1301. ifeq ($(BUILD_DOUBLE), 1)
  1302. CCOMMON_OPT += -DBUILD_DOUBLE=1
  1303. endif
  1304. ifeq ($(BUILD_COMPLEX), 1)
  1305. CCOMMON_OPT += -DBUILD_COMPLEX=1
  1306. endif
  1307. ifeq ($(BUILD_COMPLEX16), 1)
  1308. CCOMMON_OPT += -DBUILD_COMPLEX16=1
  1309. endif
  1310. CCOMMON_OPT += -DVERSION=\"$(VERSION)\"
  1311. ifndef SYMBOLPREFIX
  1312. SYMBOLPREFIX =
  1313. endif
  1314. ifndef SYMBOLSUFFIX
  1315. SYMBOLSUFFIX =
  1316. endif
  1317. ifndef LIBSONAMEBASE
  1318. LIBSONAMEBASE = openblas
  1319. endif
  1320. ifndef LIBNAMESUFFIX
  1321. LIBNAMEBASE = $(SYMBOLPREFIX)$(LIBSONAMEBASE)$(SYMBOLSUFFIX)
  1322. else
  1323. LIBNAMEBASE = $(SYMBOLPREFIX)$(LIBSONAMEBASE)$(SYMBOLSUFFIX)_$(LIBNAMESUFFIX)
  1324. endif
  1325. ifeq ($(OSNAME), CYGWIN_NT)
  1326. LIBPREFIX = cyg$(LIBNAMEBASE)
  1327. else
  1328. LIBPREFIX = lib$(LIBNAMEBASE)
  1329. endif
  1330. KERNELDIR = $(TOPDIR)/kernel/$(ARCH)
  1331. include $(TOPDIR)/Makefile.$(ARCH)
  1332. ifneq ($(C_COMPILER), PGI)
  1333. ifneq ($(C_COMPILER), SUN)
  1334. CCOMMON_OPT += -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME
  1335. endif
  1336. endif
  1337. CCOMMON_OPT += -DASMNAME=$(FU)$(*F) -DASMFNAME=$(FU)$(*F)$(BU) -DNAME=$(*F)$(BU) -DCNAME=$(*F) -DCHAR_NAME=\"$(*F)$(BU)\" -DCHAR_CNAME=\"$(*F)\"
  1338. ifeq ($(CORE), PPC440)
  1339. CCOMMON_OPT += -DALLOC_QALLOC
  1340. endif
  1341. ifeq ($(CORE), PPC440FP2)
  1342. STATIC_ALLOCATION = 1
  1343. endif
  1344. ifneq ($(OSNAME), Linux)
  1345. NO_AFFINITY = 1
  1346. endif
  1347. ifneq ($(ARCH), x86_64)
  1348. ifneq ($(ARCH), x86)
  1349. NO_AFFINITY = 1
  1350. endif
  1351. endif
  1352. ifdef NO_AFFINITY
  1353. ifeq ($(NO_AFFINITY), 0)
  1354. override undefine NO_AFFINITY
  1355. else
  1356. CCOMMON_OPT += -DNO_AFFINITY
  1357. endif
  1358. endif
  1359. ifdef FUNCTION_PROFILE
  1360. CCOMMON_OPT += -DFUNCTION_PROFILE
  1361. endif
  1362. ifdef HUGETLB_ALLOCATION
  1363. CCOMMON_OPT += -DALLOC_HUGETLB
  1364. endif
  1365. ifdef HUGETLBFILE_ALLOCATION
  1366. CCOMMON_OPT += -DALLOC_HUGETLBFILE -DHUGETLB_FILE_NAME=$(HUGETLBFILE_ALLOCATION)
  1367. endif
  1368. ifdef STATIC_ALLOCATION
  1369. CCOMMON_OPT += -DALLOC_STATIC
  1370. endif
  1371. ifdef DEVICEDRIVER_ALLOCATION
  1372. CCOMMON_OPT += -DALLOC_DEVICEDRIVER -DDEVICEDRIVER_NAME=\"/dev/mapper\"
  1373. endif
  1374. ifdef MIXED_MEMORY_ALLOCATION
  1375. CCOMMON_OPT += -DMIXED_MEMORY_ALLOCATION
  1376. endif
  1377. ifeq ($(OSNAME), SunOS)
  1378. TAR = gtar
  1379. PATCH = gpatch
  1380. GREP = ggrep
  1381. AWK = nawk
  1382. else
  1383. TAR = tar
  1384. PATCH = patch
  1385. GREP = grep
  1386. AWK = awk
  1387. endif
  1388. ifndef MD5SUM
  1389. MD5SUM = md5sum
  1390. endif
  1391. REVISION = -r$(VERSION)
  1392. MAJOR_VERSION = $(word 1,$(subst ., ,$(VERSION)))
  1393. ifeq ($(DEBUG), 1)
  1394. COMMON_OPT += -g
  1395. endif
  1396. ifeq ($(DEBUG), 1)
  1397. FCOMMON_OPT += -g
  1398. endif
  1399. ifndef COMMON_OPT
  1400. COMMON_OPT = -O2
  1401. endif
  1402. ifndef FCOMMON_OPT
  1403. FCOMMON_OPT = -O2 -frecursive
  1404. endif
  1405. override CFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
  1406. override PFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
  1407. override FFLAGS += $(COMMON_OPT) $(FCOMMON_OPT)
  1408. override FPFLAGS += $(FCOMMON_OPT) $(COMMON_PROF)
  1409. #MAKEOVERRIDES =
  1410. ifeq ($(NEED_PIC), 1)
  1411. ifeq (,$(findstring PIC,$(FFLAGS)))
  1412. override FFLAGS += -fPIC
  1413. endif
  1414. endif
  1415. #For LAPACK Fortran codes.
  1416. #Disable -fopenmp for LAPACK Fortran codes on Windows.
  1417. ifdef OS_WINDOWS
  1418. LAPACK_FFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FFLAGS))
  1419. LAPACK_FPFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FPFLAGS))
  1420. else
  1421. LAPACK_FFLAGS := $(FFLAGS)
  1422. LAPACK_FPFLAGS := $(FPFLAGS)
  1423. endif
  1424. ifeq ($(F_COMPILER),NAG)
  1425. LAPACK_FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1426. FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1427. endif
  1428. ifeq ($(F_COMPILER),CRAY)
  1429. LAPACK_FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1430. FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1431. endif
  1432. LAPACK_CFLAGS = $(CFLAGS)
  1433. LAPACK_CFLAGS += -DHAVE_LAPACK_CONFIG_H
  1434. ifdef INTERFACE64
  1435. ifneq ($(INTERFACE64), 0)
  1436. LAPACK_CFLAGS += -DLAPACK_ILP64
  1437. endif
  1438. endif
  1439. ifdef OS_WINDOWS
  1440. LAPACK_CFLAGS += -DOPENBLAS_OS_WINDOWS
  1441. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1442. endif
  1443. ifeq ($(C_COMPILER), LSB)
  1444. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1445. endif
  1446. ifndef SUFFIX
  1447. SUFFIX = o
  1448. endif
  1449. ifndef PSUFFIX
  1450. PSUFFIX = po
  1451. endif
  1452. ifndef LIBSUFFIX
  1453. LIBSUFFIX = a
  1454. endif
  1455. ifneq ($(DYNAMIC_ARCH), 1)
  1456. ifndef SMP
  1457. LIBNAME = $(LIBPREFIX)_$(LIBCORE)$(REVISION).$(LIBSUFFIX)
  1458. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)$(REVISION)_p.$(LIBSUFFIX)
  1459. else
  1460. LIBNAME = $(LIBPREFIX)_$(LIBCORE)p$(REVISION).$(LIBSUFFIX)
  1461. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)p$(REVISION)_p.$(LIBSUFFIX)
  1462. endif
  1463. else
  1464. ifndef SMP
  1465. LIBNAME = $(LIBPREFIX)$(REVISION).$(LIBSUFFIX)
  1466. LIBNAME_P = $(LIBPREFIX)$(REVISION)_p.$(LIBSUFFIX)
  1467. else
  1468. LIBNAME = $(LIBPREFIX)p$(REVISION).$(LIBSUFFIX)
  1469. LIBNAME_P = $(LIBPREFIX)p$(REVISION)_p.$(LIBSUFFIX)
  1470. endif
  1471. endif
  1472. LIBDLLNAME = $(LIBPREFIX).dll
  1473. IMPLIBNAME = lib$(LIBNAMEBASE).dll.a
  1474. ifneq ($(OSNAME), AIX)
  1475. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.so)
  1476. else
  1477. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.a)
  1478. endif
  1479. LIBDYNNAME = $(LIBNAME:.$(LIBSUFFIX)=.dylib)
  1480. LIBDEFNAME = $(LIBNAME:.$(LIBSUFFIX)=.def)
  1481. LIBEXPNAME = $(LIBNAME:.$(LIBSUFFIX)=.exp)
  1482. LIBZIPNAME = $(LIBNAME:.$(LIBSUFFIX)=.zip)
  1483. LIBS = $(TOPDIR)/$(LIBNAME)
  1484. LIBS_P = $(TOPDIR)/$(LIBNAME_P)
  1485. LIB_COMPONENTS = BLAS
  1486. ifneq ($(NO_CBLAS), 1)
  1487. LIB_COMPONENTS += CBLAS
  1488. endif
  1489. ifneq ($(NO_LAPACK), 1)
  1490. LIB_COMPONENTS += LAPACK
  1491. ifneq ($(NO_LAPACKE), 1)
  1492. LIB_COMPONENTS += LAPACKE
  1493. endif
  1494. ifeq ($(BUILD_RELAPACK), 1)
  1495. LIB_COMPONENTS += ReLAPACK
  1496. endif
  1497. endif
  1498. ifeq ($(ONLY_CBLAS), 1)
  1499. LIB_COMPONENTS = CBLAS
  1500. endif
  1501. export OSNAME
  1502. export ARCH
  1503. export CORE
  1504. export LIBCORE
  1505. export __BYTE_ORDER__
  1506. export ELF_VERSION
  1507. export PGCPATH
  1508. export CONFIG
  1509. export CC
  1510. export FC
  1511. export BU
  1512. export FU
  1513. export NEED2UNDERSCORES
  1514. export USE_THREAD
  1515. export NUM_THREADS
  1516. export NUM_CORES
  1517. export SMP
  1518. export MAKEFILE_RULE
  1519. export NEED_PIC
  1520. export BINARY
  1521. export BINARY32
  1522. export BINARY64
  1523. export F_COMPILER
  1524. export C_COMPILER
  1525. export USE_OPENMP
  1526. export CROSS
  1527. export CROSS_SUFFIX
  1528. export NOFORTRAN
  1529. export C_LAPACK
  1530. export NO_FBLAS
  1531. export EXTRALIB
  1532. export CEXTRALIB
  1533. export FEXTRALIB
  1534. export HAVE_SSE
  1535. export HAVE_SSE2
  1536. export HAVE_SSE3
  1537. export HAVE_SSSE3
  1538. export HAVE_SSE4_1
  1539. export HAVE_SSE4_2
  1540. export HAVE_SSE4A
  1541. export HAVE_SSE5
  1542. export HAVE_AVX
  1543. export HAVE_AVX2
  1544. export HAVE_FMA3
  1545. export HAVE_VFP
  1546. export HAVE_VFPV3
  1547. export HAVE_VFPV4
  1548. export HAVE_NEON
  1549. ifndef NO_MSA
  1550. export HAVE_MSA
  1551. export MSA_FLAGS
  1552. endif
  1553. export KERNELDIR
  1554. export FUNCTION_PROFILE
  1555. export TARGET_CORE
  1556. export NO_AVX512
  1557. export NO_AVX2
  1558. export BUILD_BFLOAT16
  1559. export NO_LSX
  1560. export NO_LASX
  1561. export SBGEMM_UNROLL_M
  1562. export SBGEMM_UNROLL_N
  1563. export SGEMM_UNROLL_M
  1564. export SGEMM_UNROLL_N
  1565. export DGEMM_UNROLL_M
  1566. export DGEMM_UNROLL_N
  1567. export QGEMM_UNROLL_M
  1568. export QGEMM_UNROLL_N
  1569. export CGEMM_UNROLL_M
  1570. export CGEMM_UNROLL_N
  1571. export ZGEMM_UNROLL_M
  1572. export ZGEMM_UNROLL_N
  1573. export XGEMM_UNROLL_M
  1574. export XGEMM_UNROLL_N
  1575. export CGEMM3M_UNROLL_M
  1576. export CGEMM3M_UNROLL_N
  1577. export ZGEMM3M_UNROLL_M
  1578. export ZGEMM3M_UNROLL_N
  1579. export XGEMM3M_UNROLL_M
  1580. export XGEMM3M_UNROLL_N
  1581. ifdef USE_CUDA
  1582. export CUDADIR
  1583. export CUCC
  1584. export CUFLAGS
  1585. export CULIB
  1586. endif
  1587. .SUFFIXES: .$(PSUFFIX) .$(SUFFIX) .f
  1588. .f.$(SUFFIX):
  1589. $(FC) $(FFLAGS) -c $< -o $(@F)
  1590. .f.$(PSUFFIX):
  1591. $(FC) $(FPFLAGS) -pg -c $< -o $(@F)
  1592. ifdef BINARY64
  1593. PATHSCALEPATH = /opt/pathscale/lib/3.1
  1594. PGIPATH = /opt/pgi/linux86-64/7.1-5/lib
  1595. else
  1596. PATHSCALEPATH = /opt/pathscale/lib/3.1/32
  1597. PGIPATH = /opt/pgi/linux86/7.1-5/lib
  1598. endif
  1599. ACMLPATH = /opt/acml/4.3.0
  1600. ifneq ($(OSNAME), Darwin)
  1601. MKLPATH = /opt/intel/mkl/10.2.2.025/lib
  1602. else
  1603. MKLPATH = /Library/Frameworks/Intel_MKL.framework/Versions/10.0.1.014/lib
  1604. endif
  1605. ATLASPATH = /opt/atlas/3.9.17/opteron
  1606. FLAMEPATH = $(HOME)/flame/lib
  1607. ifneq ($(OSNAME), SunOS)
  1608. SUNPATH = /opt/sunstudio12.1
  1609. else
  1610. SUNPATH = /opt/SUNWspro
  1611. endif