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 40 kB

7 years ago
4 years ago
13 years ago
4 years ago
12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448144914501451145214531454145514561457145814591460146114621463146414651466146714681469147014711472147314741475147614771478147914801481148214831484148514861487148814891490149114921493149414951496149714981499150015011502150315041505150615071508150915101511151215131514151515161517151815191520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559156015611562156315641565156615671568156915701571157215731574157515761577157815791580158115821583158415851586158715881589159015911592159315941595159615971598159916001601160216031604160516061607160816091610161116121613161416151616161716181619162016211622162316241625162616271628162916301631163216331634163516361637163816391640164116421643164416451646164716481649165016511652165316541655165616571658165916601661166216631664166516661667166816691670167116721673167416751676167716781679168016811682168316841685168616871688168916901691169216931694169516961697169816991700170117021703170417051706170717081709171017111712171317141715171617171718171917201721172217231724172517261727172817291730173117321733173417351736173717381739174017411742174317441745174617471748174917501751175217531754175517561757175817591760176117621763176417651766176717681769177017711772177317741775177617771778177917801781178217831784178517861787178817891790179117921793179417951796179717981799180018011802180318041805180618071808180918101811181218131814181518161817181818191820182118221823182418251826182718281829183018311832183318341835183618371838183918401841184218431844184518461847184818491850185118521853185418551856185718581859186018611862186318641865186618671868186918701871187218731874187518761877187818791880188118821883188418851886188718881889
  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. # Generating Makefile.conf and config.h
  245. 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)
  246. endif
  247. ifndef TARGET_CORE
  248. -include $(TOPDIR)/Makefile.conf
  249. else
  250. HAVE_NEON=
  251. HAVE_VFP=
  252. HAVE_VFPV3=
  253. HAVE_VFPV4=
  254. HAVE_MMX=
  255. HAVE_SSE=
  256. HAVE_SSE2=
  257. HAVE_SSE3=
  258. HAVE_SSSE3=
  259. HAVE_SSE4_1=
  260. HAVE_SSE4_2=
  261. HAVE_SSE4A=
  262. HAVE_SSE5=
  263. HAVE_AVX=
  264. HAVE_AVX2=
  265. HAVE_FMA3=
  266. include $(TOPDIR)/Makefile_kernel.conf
  267. endif
  268. ifndef NUM_PARALLEL
  269. NUM_PARALLEL = 1
  270. endif
  271. ifndef NUM_THREADS
  272. NUM_THREADS = $(NUM_CORES)
  273. endif
  274. ifeq ($(NUM_THREADS), 1)
  275. override USE_THREAD = 0
  276. override USE_OPENMP = 0
  277. endif
  278. ifdef USE_THREAD
  279. ifeq ($(USE_THREAD), 0)
  280. SMP =
  281. else
  282. SMP = 1
  283. endif
  284. else
  285. ifeq ($(NUM_THREADS), 1)
  286. SMP =
  287. else
  288. SMP = 1
  289. endif
  290. endif
  291. ifeq ($(SMP), 1)
  292. USE_LOCKING =
  293. endif
  294. ifndef NEED_PIC
  295. NEED_PIC = 1
  296. endif
  297. ARFLAGS =
  298. CPP = $(COMPILER) -E
  299. AR ?= $(CROSS_SUFFIX)ar
  300. AS ?= $(CROSS_SUFFIX)as
  301. LD ?= $(CROSS_SUFFIX)ld
  302. RANLIB ?= $(CROSS_SUFFIX)ranlib
  303. NM = $(CROSS_SUFFIX)nm
  304. DLLWRAP = $(CROSS_SUFFIX)dllwrap
  305. OBJCOPY = $(CROSS_SUFFIX)objcopy
  306. OBJCONV = $(CROSS_SUFFIX)objconv
  307. # When fortran support was either not detected or actively deselected, only build BLAS.
  308. ifeq ($(NOFORTRAN), 1)
  309. C_LAPACK = 1
  310. override FEXTRALIB =
  311. ifeq ($(C_COMPILER), GCC)
  312. CCOMMON_OPT += -Wno-error=incompatible-pointer-types
  313. endif
  314. endif
  315. ifeq ($(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. GCCVERSIONGTEQ10 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 10)
  323. GCCVERSIONGTEQ11 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 11)
  324. GCCVERSIONGTEQ12 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 12)
  325. # Note that the behavior of -dumpversion is compile-time-configurable for
  326. # gcc-7.x and newer. Use -dumpfullversion there
  327. ifeq ($(GCCVERSIONGTEQ7),1)
  328. GCCDUMPVERSION_PARAM := -dumpfullversion
  329. else
  330. GCCDUMPVERSION_PARAM := -dumpversion
  331. endif
  332. GCCMINORVERSIONGTEQ1 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 1)
  333. GCCMINORVERSIONGTEQ2 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 2)
  334. GCCMINORVERSIONGTEQ4 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 4)
  335. GCCMINORVERSIONGTEQ7 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 7)
  336. endif
  337. ifeq ($(C_COMPILER), CLANG)
  338. CLANGVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  339. CLANGVERSIONGTEQ12 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 12)
  340. endif
  341. #
  342. # OS dependent settings
  343. #
  344. ifeq ($(OSNAME), Darwin)
  345. ifndef MACOSX_DEPLOYMENT_TARGET
  346. ifeq ($(ARCH), arm64)
  347. export MACOSX_DEPLOYMENT_TARGET=11.0
  348. ifeq ($(C_COMPILER), GCC)
  349. export NO_SVE = 1
  350. endif
  351. else
  352. export MACOSX_DEPLOYMENT_TARGET=10.8
  353. endif
  354. endif
  355. MD5SUM = md5 -r
  356. XCVER = $(shell pkgutil --pkg-info=com.apple.pkg.Xcode |awk '/version:/ {print $2}'|cut -d: -f2|cut -f1 -d.)
  357. ifeq (x$(XCVER)x,xx)
  358. XCVER = $(shell pkgutil --pkg-info=com.apple.pkg.CLTools_Executables |awk '/version:/ {print $2}'|cut -d: -f2|cut -f1 -d.)
  359. endif
  360. ifeq (x$(XCVER), x 15)
  361. CCOMMON_OPT += -Wl,-ld_classic
  362. FCOMMON_OPT += -Wl,-ld_classic
  363. endif
  364. endif
  365. ifneq (,$(findstring $(OSNAME), FreeBSD OpenBSD DragonFly))
  366. MD5SUM = md5 -r
  367. endif
  368. ifeq ($(OSNAME), NetBSD)
  369. MD5SUM = md5 -n
  370. endif
  371. ifeq ($(OSNAME), Linux)
  372. EXTRALIB += -lm
  373. NO_EXPRECISION = 1
  374. endif
  375. ifeq ($(OSNAME), Android)
  376. EXTRALIB += -lm
  377. endif
  378. ifeq ($(OSNAME), AIX)
  379. EXTRALIB += -lm
  380. endif
  381. ifeq ($(OSNAME), $(filter $(OSNAME),FreeBSD OpenBSD NetBSD DragonFly))
  382. ifeq ($(ARCH), $(filter $(ARCH),arm arm64))
  383. EXTRALIB += -lm
  384. endif
  385. endif
  386. ifeq ($(OSNAME), WINNT)
  387. NEED_PIC = 0
  388. NO_EXPRECISION = 1
  389. EXTRALIB += -defaultlib:advapi32
  390. SUFFIX = obj
  391. PSUFFIX = pobj
  392. LIBSUFFIX = a
  393. ifeq ($(C_COMPILER), CLANG)
  394. CCOMMON_OPT += -DMS_ABI
  395. endif
  396. #Version tests for supporting specific features (MS_ABI, POWER9 intrinsics)
  397. ifeq ($(GCCVERSIONGT4), 1)
  398. # GCC Major version > 4
  399. # It is compatible with MSVC ABI.
  400. CCOMMON_OPT += -DMS_ABI
  401. endif
  402. ifeq ($(GCCVERSIONGTEQ4), 1)
  403. ifeq ($(GCCMINORVERSIONGTEQ7), 1)
  404. # GCC Version >=4.7
  405. # It is compatible with MSVC ABI.
  406. CCOMMON_OPT += -DMS_ABI
  407. endif
  408. endif
  409. # Ensure the correct stack alignment on Win32
  410. # http://permalink.gmane.org/gmane.comp.lib.openblas.general/97
  411. ifeq ($(ARCH), x86)
  412. CCOMMON_OPT += -mincoming-stack-boundary=2
  413. FCOMMON_OPT += -mincoming-stack-boundary=2
  414. endif
  415. endif
  416. ifeq ($(OSNAME), Interix)
  417. NEED_PIC = 0
  418. NO_EXPRECISION = 1
  419. INTERIX_TOOL_DIR = /opt/gcc.3.3/i586-pc-interix3/bin
  420. endif
  421. ifeq ($(OSNAME), CYGWIN_NT)
  422. NEED_PIC = 0
  423. NO_EXPRECISION = 1
  424. OS_CYGWIN_NT = 1
  425. endif
  426. ifneq ($(OSNAME), WINNT)
  427. ifneq ($(OSNAME), CYGWIN_NT)
  428. ifneq ($(OSNAME), Interix)
  429. ifneq ($(OSNAME), Android)
  430. ifdef SMP
  431. EXTRALIB += -lpthread
  432. endif
  433. endif
  434. endif
  435. endif
  436. endif
  437. # ifeq logical or
  438. ifeq ($(OSNAME), $(filter $(OSNAME),WINNT CYGWIN_NT Interix))
  439. OS_WINDOWS=1
  440. endif
  441. ifdef QUAD_PRECISION
  442. CCOMMON_OPT += -DQUAD_PRECISION
  443. NO_EXPRECISION = 1
  444. endif
  445. ifneq ($(ARCH), x86)
  446. ifneq ($(ARCH), x86_64)
  447. NO_EXPRECISION = 1
  448. endif
  449. endif
  450. ifdef UTEST_CHECK
  451. CCOMMON_OPT += -DUTEST_CHECK
  452. SANITY_CHECK = 1
  453. endif
  454. ifdef SANITY_CHECK
  455. CCOMMON_OPT += -DSANITY_CHECK -DREFNAME=$(*F)f$(BU)
  456. endif
  457. MAX_STACK_ALLOC ?= 2048
  458. ifneq ($(MAX_STACK_ALLOC), 0)
  459. CCOMMON_OPT += -DMAX_STACK_ALLOC=$(MAX_STACK_ALLOC)
  460. endif
  461. ifdef USE_LOCKING
  462. ifneq ($(USE_LOCKING), 0)
  463. CCOMMON_OPT += -DUSE_LOCKING
  464. endif
  465. endif
  466. #
  467. # Architecture dependent settings
  468. #
  469. ifeq ($(ARCH), x86)
  470. ifndef BINARY
  471. NO_BINARY_MODE = 1
  472. endif
  473. ifeq ($(CORE), generic)
  474. NO_EXPRECISION = 1
  475. endif
  476. ifndef NO_EXPRECISION
  477. ifeq ($(F_COMPILER), GFORTRAN)
  478. # ifeq logical or. GCC or LSB
  479. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  480. EXPRECISION = 1
  481. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  482. FCOMMON_OPT += -m128bit-long-double
  483. endif
  484. ifeq ($(C_COMPILER), CLANG)
  485. EXPRECISION = 1
  486. CCOMMON_OPT += -DEXPRECISION
  487. FCOMMON_OPT += -m128bit-long-double
  488. endif
  489. endif
  490. endif
  491. endif
  492. ifeq ($(ARCH), x86_64)
  493. ifeq ($(CORE), generic)
  494. NO_EXPRECISION = 1
  495. endif
  496. ifndef NO_EXPRECISION
  497. ifeq ($(F_COMPILER), GFORTRAN)
  498. # ifeq logical or. GCC or LSB
  499. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  500. EXPRECISION = 1
  501. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  502. FCOMMON_OPT += -m128bit-long-double
  503. endif
  504. ifeq ($(C_COMPILER), CLANG)
  505. EXPRECISION = 1
  506. CCOMMON_OPT += -DEXPRECISION
  507. FCOMMON_OPT += -m128bit-long-double
  508. endif
  509. endif
  510. endif
  511. endif
  512. ifeq ($(C_COMPILER), INTEL)
  513. CCOMMON_OPT += -wd981
  514. endif
  515. ifeq ($(USE_OPENMP), 1)
  516. #check
  517. ifeq ($(USE_THREAD), 0)
  518. $(error OpenBLAS: Cannot set both USE_OPENMP=1 and USE_THREAD=0. The USE_THREAD=0 is only for building single thread version.)
  519. endif
  520. # ifeq logical or. GCC or LSB
  521. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  522. CCOMMON_OPT += -fopenmp
  523. endif
  524. ifeq ($(C_COMPILER), CLANG)
  525. CCOMMON_OPT += -fopenmp
  526. ifeq ($(F_COMPILER), GFORTRAN)
  527. FEXTRALIB := $(subst -lgomp,-lomp,$(FEXTRALIB))
  528. endif
  529. endif
  530. ifeq ($(C_COMPILER), INTEL)
  531. CCOMMON_OPT += -fopenmp
  532. endif
  533. ifeq ($(C_COMPILER), PGI)
  534. CCOMMON_OPT += -mp
  535. endif
  536. ifeq ($(C_COMPILER), OPEN64)
  537. CCOMMON_OPT += -mp
  538. CEXTRALIB += -lstdc++
  539. endif
  540. ifeq ($(C_COMPILER), PATHSCALE)
  541. CCOMMON_OPT += -mp
  542. endif
  543. endif
  544. ifeq ($(DYNAMIC_ARCH), 1)
  545. ifeq ($(ARCH), x86)
  546. DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
  547. CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
  548. endif
  549. ifeq ($(ARCH), x86_64)
  550. DYNAMIC_CORE = PRESCOTT CORE2
  551. ifeq ($(DYNAMIC_OLDER), 1)
  552. DYNAMIC_CORE += PENRYN DUNNINGTON
  553. endif
  554. DYNAMIC_CORE += NEHALEM
  555. ifeq ($(DYNAMIC_OLDER), 1)
  556. DYNAMIC_CORE += OPTERON OPTERON_SSE3
  557. endif
  558. DYNAMIC_CORE += BARCELONA
  559. ifeq ($(DYNAMIC_OLDER), 1)
  560. DYNAMIC_CORE += BOBCAT ATOM NANO
  561. endif
  562. ifneq ($(NO_AVX), 1)
  563. DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER STEAMROLLER EXCAVATOR
  564. endif
  565. ifneq ($(NO_AVX2), 1)
  566. DYNAMIC_CORE += HASWELL ZEN
  567. endif
  568. ifneq ($(NO_AVX512), 1)
  569. ifneq ($(NO_AVX2), 1)
  570. DYNAMIC_CORE += SKYLAKEX COOPERLAKE SAPPHIRERAPIDS
  571. endif
  572. endif
  573. endif
  574. ifdef DYNAMIC_LIST
  575. override DYNAMIC_CORE = PRESCOTT $(DYNAMIC_LIST)
  576. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_PRESCOTT
  577. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  578. CCOMMON_OPT += $(XCCOMMON_OPT)
  579. #CCOMMON_OPT += -DDYNAMIC_LIST='$(DYNAMIC_LIST)'
  580. endif
  581. ifeq ($(ARCH), arm64)
  582. DYNAMIC_CORE = ARMV8
  583. DYNAMIC_CORE += CORTEXA53
  584. DYNAMIC_CORE += CORTEXA57
  585. DYNAMIC_CORE += NEOVERSEN1
  586. ifneq ($(NO_SVE), 1)
  587. DYNAMIC_CORE += NEOVERSEV1
  588. DYNAMIC_CORE += NEOVERSEN2
  589. DYNAMIC_CORE += ARMV8SVE
  590. endif
  591. DYNAMIC_CORE += THUNDERX
  592. DYNAMIC_CORE += THUNDERX2T99
  593. DYNAMIC_CORE += TSV110
  594. DYNAMIC_CORE += EMAG8180
  595. DYNAMIC_CORE += THUNDERX3T110
  596. ifdef DYNAMIC_LIST
  597. override DYNAMIC_CORE = ARMV8 $(DYNAMIC_LIST)
  598. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_ARMV8
  599. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  600. endif
  601. endif
  602. ifeq ($(ARCH), mips64)
  603. DYNAMIC_CORE = LOONGSON3R3 LOONGSON3R4 MIPS64_GENERIC
  604. ifdef DYNAMIC_LIST
  605. override DYNAMIC_CORE = MIPS64_GENERIC $(DYNAMIC_LIST)
  606. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_MIPS64_GENERIC
  607. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  608. endif
  609. endif
  610. ifeq ($(ARCH), loongarch64)
  611. DYNAMIC_CORE = LOONGSON3R5 LOONGSON2K1000 LOONGSONGENERIC
  612. endif
  613. ifeq ($(ARCH), zarch)
  614. DYNAMIC_CORE = ZARCH_GENERIC
  615. # if the compiler accepts -march=arch11 or -march=z13 and can compile a file
  616. # with z13-specific inline assembly, then we can include support for Z13.
  617. # note: -march=z13 is equivalent to -march=arch11 yet some compiler releases
  618. # only support one or the other.
  619. # note: LLVM version 6.x supported -march=z13 yet could not handle vector
  620. # registers in inline assembly, so the check for supporting the -march flag is
  621. # not enough.
  622. ZARCH_TEST_COMPILE=-c $(TOPDIR)/kernel/zarch/damin_z13.c -I$(TOPDIR) -o /dev/null > /dev/null 2> /dev/null
  623. ZARCH_CC_SUPPORTS_ARCH11=$(shell $(CC) -march=arch11 $(ZARCH_TEST_COMPILE) && echo 1)
  624. ZARCH_CC_SUPPORTS_Z13=$(shell $(CC) -march=z13 $(ZARCH_TEST_COMPILE) && echo 1)
  625. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH11), $(ZARCH_CC_SUPPORTS_Z13)), 1)
  626. DYNAMIC_CORE += Z13
  627. CCOMMON_OPT += -DDYN_Z13
  628. else
  629. $(info OpenBLAS: Not building Z13 kernels because the compiler $(CC) does not support it)
  630. endif
  631. # as above for z13, check for -march=arch12 and z14 support in the compiler.
  632. ZARCH_CC_SUPPORTS_ARCH12=$(shell $(CC) -march=arch12 $(ZARCH_TEST_COMPILE) && echo 1)
  633. ZARCH_CC_SUPPORTS_Z14=$(shell $(CC) -march=z14 $(ZARCH_TEST_COMPILE) && echo 1)
  634. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH12), $(ZARCH_CC_SUPPORTS_Z14)), 1)
  635. DYNAMIC_CORE += Z14
  636. CCOMMON_OPT += -DDYN_Z14
  637. else
  638. $(info OpenBLAS: Not building Z14 kernels because the compiler $(CC) does not support it)
  639. endif
  640. endif # ARCH zarch
  641. ifeq ($(ARCH), power)
  642. ifneq ($(C_COMPILER), PGI)
  643. DYNAMIC_CORE = POWER6
  644. DYNAMIC_CORE += POWER8
  645. ifneq ($(C_COMPILER), GCC)
  646. DYNAMIC_CORE += POWER9
  647. DYNAMIC_CORE += POWER10
  648. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  649. endif
  650. ifeq ($(C_COMPILER), GCC)
  651. ifeq ($(GCCVERSIONGT5), 1)
  652. DYNAMIC_CORE += POWER9
  653. else
  654. $(info, OpenBLAS: Your gcc version is too old to build the POWER9 kernels.)
  655. endif
  656. ifeq ($(OSNAME), AIX)
  657. LDVERSIONGTEQ35 := 1
  658. else
  659. LDVERSIONGTEQ35 := $(shell expr `$(CC) -Wl,--version 2> /dev/null | head -1 | cut -f2 -d "." | cut -f1 -d "-"` \>= 35)
  660. endif
  661. ifeq ($(GCCVERSIONGTEQ11)$(LDVERSIONGTEQ35), 11)
  662. DYNAMIC_CORE += POWER10
  663. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  664. else ifeq ($(GCCVERSIONGTEQ10), 1)
  665. ifeq ($(GCCMINORVERSIONGTEQ2)$(LDVERSIONGTEQ35), 11)
  666. DYNAMIC_CORE += POWER10
  667. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  668. endif
  669. else
  670. $(info, OpenBLAS: Your gcc version is too old to build the POWER10 kernels.)
  671. endif
  672. endif
  673. else
  674. DYNAMIC_CORE = POWER8
  675. DYNAMIC_CORE += POWER9
  676. endif
  677. endif
  678. # If DYNAMIC_CORE is not set, DYNAMIC_ARCH cannot do anything, so force it to empty
  679. ifndef DYNAMIC_CORE
  680. override DYNAMIC_ARCH=
  681. endif
  682. endif
  683. ifeq ($(ARCH), ia64)
  684. NO_BINARY_MODE = 1
  685. BINARY_DEFINED = 1
  686. ifeq ($(F_COMPILER), GFORTRAN)
  687. ifeq ($(C_COMPILER), GCC)
  688. # EXPRECISION = 1
  689. # CCOMMON_OPT += -DEXPRECISION
  690. endif
  691. endif
  692. endif
  693. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  694. NO_BINARY_MODE = 1
  695. endif
  696. ifeq ($(ARCH), alpha)
  697. NO_BINARY_MODE = 1
  698. BINARY_DEFINED = 1
  699. endif
  700. ifeq ($(ARCH), arm)
  701. NO_BINARY_MODE = 1
  702. BINARY_DEFINED = 1
  703. ifneq ($(EMBEDDED), 1)
  704. CCOMMON_OPT += -marm
  705. FCOMMON_OPT += -marm
  706. else
  707. CCOMMON_OPT += -DOS_EMBEDDED -mthumb -mcpu=cortex-m4 -mfloat-abi=hard -mfpu=fpv4-sp-d16
  708. endif
  709. # If softfp abi is mentioned on the command line, force it.
  710. ifeq ($(ARM_SOFTFP_ABI), 1)
  711. CCOMMON_OPT += -mfloat-abi=softfp
  712. FCOMMON_OPT += -mfloat-abi=softfp
  713. endif
  714. ifeq ($(OSNAME), Android)
  715. ifeq ($(ARM_SOFTFP_ABI), 1)
  716. EXTRALIB += -lm
  717. else
  718. EXTRALIB += -Wl,-lm_hard
  719. endif
  720. endif
  721. endif
  722. ifeq ($(ARCH), arm64)
  723. NO_BINARY_MODE = 1
  724. BINARY_DEFINED = 1
  725. ifdef INTERFACE64
  726. ifneq ($(INTERFACE64), 0)
  727. ifeq ($(F_COMPILER), GFORTRAN)
  728. FCOMMON_OPT += -fdefault-integer-8
  729. endif
  730. ifeq ($(F_COMPILER), FLANG)
  731. FCOMMON_OPT += -i8
  732. endif
  733. endif
  734. endif
  735. endif
  736. ifeq ($(ARCH), riscv64)
  737. NO_BINARY_MODE = 1
  738. BINARY_DEFINED = 1
  739. ifdef INTERFACE64
  740. ifneq ($(INTERFACE64), 0)
  741. ifeq ($(F_COMPILER), GFORTRAN)
  742. FCOMMON_OPT += -fdefault-integer-8
  743. endif
  744. ifeq ($(F_COMPILER), FLANG)
  745. FCOMMON_OPT += -i8
  746. endif
  747. endif
  748. endif
  749. endif
  750. ifeq ($(ARCH), loongarch64)
  751. NO_BINARY_MODE = 1
  752. BINARY_DEFINED = 1
  753. ifdef INTERFACE64
  754. ifneq ($(INTERFACE64), 0)
  755. ifeq ($(F_COMPILER), GFORTRAN)
  756. FCOMMON_OPT += -fdefault-integer-8
  757. endif
  758. ifeq ($(F_COMPILER), FLANG)
  759. FCOMMON_OPT += -i8
  760. endif
  761. endif
  762. endif
  763. endif
  764. ifeq ($(ARCH), csky)
  765. NO_BINARY_MODE = 1
  766. BINARY_DEFINED = 1
  767. endif
  768. #
  769. # C Compiler dependent settings
  770. #
  771. # ifeq logical or. GCC or CLANG or LSB
  772. # http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
  773. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG LSB))
  774. CCOMMON_OPT += -Wall
  775. COMMON_PROF += -fno-inline
  776. NO_UNINITIALIZED_WARN = -Wno-uninitialized
  777. ifeq ($(QUIET_MAKE), 1)
  778. CCOMMON_OPT += $(NO_UNINITIALIZED_WARN) -Wno-unused
  779. endif
  780. ifdef NO_BINARY_MODE
  781. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  782. ifdef BINARY64
  783. CCOMMON_OPT += -mabi=64
  784. else
  785. CCOMMON_OPT += -mabi=n32
  786. endif
  787. BINARY_DEFINED = 1
  788. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  789. CCOMMON_OPT += -mabi=32
  790. BINARY_DEFINED = 1
  791. endif
  792. ifneq (, $(filter $(CORE), MIPS64_GENERIC))
  793. CCOMMON_OPT += -DNO_MSA
  794. FCOMMON_OPT += -DNO_MSA
  795. endif
  796. ifneq (, $(filter $(CORE),LOONGSON3R3 LOONGSON3R4))
  797. CCOMMON_OPT += -march=loongson3a
  798. FCOMMON_OPT += -march=loongson3a
  799. endif
  800. ifeq ($(CORE), MIPS24K)
  801. CCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  802. FCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  803. endif
  804. ifeq ($(CORE), MIPS1004K)
  805. CCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  806. FCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  807. endif
  808. ifeq ($(CORE), P5600)
  809. CCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  810. FCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  811. endif
  812. ifeq ($(CORE), I6400)
  813. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  814. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  815. endif
  816. ifeq ($(CORE), P6600)
  817. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  818. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  819. endif
  820. ifeq ($(CORE), I6500)
  821. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  822. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  823. endif
  824. ifeq ($(OSNAME), AIX)
  825. BINARY_DEFINED = 1
  826. endif
  827. ifeq ($(ARCH), loongarch64)
  828. LA64_ABI=$(shell $(CC) -mabi=lp64d -c $(TOPDIR)/cpuid_loongarch64.c -o /dev/null > /dev/null 2> /dev/null && echo lp64d)
  829. LA64_ARCH=$(shell $(CC) -march=loongarch64 -c $(TOPDIR)/cpuid_loongarch64.c -o /dev/null > /dev/null 2> /dev/null && echo loongarch64)
  830. ifneq ($(LA64_ABI), lp64d)
  831. LA64_ABI=lp64
  832. endif
  833. ifneq ($(LA64_ARCH), loongarch64)
  834. CCOMMON_OPT += -mabi=$(LA64_ABI)
  835. FCOMMON_OPT += -mabi=$(LA64_ABI)
  836. else
  837. CCOMMON_OPT += -march=loongarch64 -mabi=$(LA64_ABI)
  838. FCOMMON_OPT += -march=loongarch64 -mabi=$(LA64_ABI)
  839. endif
  840. endif
  841. endif
  842. ifndef BINARY_DEFINED
  843. ifneq ($(OSNAME), AIX)
  844. ifdef BINARY64
  845. ifneq ($(ARCH), riscv64)
  846. CCOMMON_OPT += -m64
  847. endif
  848. else
  849. CCOMMON_OPT += -m32
  850. endif
  851. endif
  852. endif
  853. endif
  854. ifeq ($(C_COMPILER), PGI)
  855. PGCVERSIONGT20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \> 20)
  856. PGCVERSIONEQ20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` == 20)
  857. PGCMINORVERSIONGE11 := $(shell expr `$(CC) --version|sed -n "2p" |cut -d "-" -f 1 |sed -e "s/[^0-9.]//g" |cut -c 4-5` \>= 11)
  858. PGCVERSIONCHECK := $(PGCVERSIONGT20)$(PGCVERSIONEQ20)$(PGCMINORVERSIONGE11)
  859. ifeq ($(PGCVERSIONCHECK), $(filter $(PGCVERSIONCHECK), 100 101 011))
  860. NEWPGI := 1
  861. PGCVERSIONGT21 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \> 21)
  862. PGCVERSIONEQ21 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` == 21)
  863. PGCVERSIONCHECK2 := $(PGCVERSIONGT21)$(PGCVERSIONEQ21)$(PGCMINORVERSIONGE11)
  864. ifeq ($(PGCVERSIONCHECK2), $(filter $(PGCVERSIONCHECK2), 100 101 011))
  865. NEWPGI2 := 1
  866. endif
  867. endif
  868. ifdef BINARY64
  869. ifeq ($(ARCH), x86_64)
  870. ifeq (,$(findstring tp,$(CFLAGS)))
  871. ifneq ($(NEWPGI2),1)
  872. CCOMMON_OPT += -tp p7-64
  873. else
  874. CCOMMON_OPT += -tp px
  875. endif
  876. endif
  877. ifneq ($(NEWPGI),1)
  878. CCOMMON_OPT += -D__MMX__ -Mnollvm
  879. endif
  880. else
  881. ifeq ($(ARCH), power)
  882. ifeq (,$(findstring tp,$(CFLAGS)))
  883. ifeq ($(CORE), POWER8)
  884. CCOMMON_OPT += -tp pwr8
  885. endif
  886. ifeq ($(CORE), POWER9)
  887. CCOMMON_OPT += -tp pwr9
  888. endif
  889. endif
  890. endif
  891. endif
  892. else
  893. ifneq ($(NEWPGI2),1)
  894. ifeq (,$(findstring tp,$(CFLAGS)))
  895. CCOMMON_OPT += -tp p7
  896. else
  897. CCOMMON_OPT += -tp px
  898. endif
  899. endif
  900. endif
  901. endif
  902. ifeq ($(C_COMPILER), PATHSCALE)
  903. ifdef BINARY64
  904. CCOMMON_OPT += -m64
  905. else
  906. CCOMMON_OPT += -m32
  907. endif
  908. endif
  909. #
  910. # Fortran Compiler dependent settings
  911. #
  912. ifeq ($(F_COMPILER), NAG)
  913. FCOMMON_OPT += -dcfuns -recursive -ieee=full -w=obs -thread_safe
  914. ifdef INTERFACE64
  915. ifneq ($(INTERFACE64), 0)
  916. FCOMMON_OPT += -i8
  917. endif
  918. endif
  919. ifeq ($(USE_OPENMP), 1)
  920. FCOMMON_OPT += -openmp
  921. endif
  922. endif
  923. ifeq ($(F_COMPILER), FLANG)
  924. CCOMMON_OPT += -DF_INTERFACE_FLANG
  925. FCOMMON_OPT += -Mrecursive -Kieee
  926. ifeq ($(OSNAME), Linux)
  927. ifeq ($(ARCH), x86_64)
  928. FLANG_VENDOR := $(shell $(FC) --version|head -1 |cut -f 1 -d " ")
  929. ifeq ($(FLANG_VENDOR), AMD)
  930. FCOMMON_OPT += -fno-unroll-loops
  931. endif
  932. endif
  933. endif
  934. ifdef BINARY64
  935. ifdef INTERFACE64
  936. ifneq ($(INTERFACE64), 0)
  937. FCOMMON_OPT += -i8
  938. endif
  939. endif
  940. FCOMMON_OPT += -Wall
  941. else
  942. FCOMMON_OPT += -Wall
  943. endif
  944. ifeq ($(USE_OPENMP), 1)
  945. FCOMMON_OPT += -fopenmp
  946. endif
  947. endif
  948. ifeq ($(F_COMPILER), G77)
  949. CCOMMON_OPT += -DF_INTERFACE_G77
  950. FCOMMON_OPT += -Wall
  951. ifndef NO_BINARY_MODE
  952. ifneq ($(OSNAME), AIX)
  953. ifdef BINARY64
  954. FCOMMON_OPT += -m64
  955. else
  956. FCOMMON_OPT += -m32
  957. endif
  958. endif
  959. endif
  960. endif
  961. ifeq ($(F_COMPILER), G95)
  962. CCOMMON_OPT += -DF_INTERFACE_G95
  963. FCOMMON_OPT += -Wall
  964. ifneq ($(OSNAME), AIX)
  965. ifndef NO_BINARY_MODE
  966. ifdef BINARY64
  967. FCOMMON_OPT += -m64
  968. else
  969. FCOMMON_OPT += -m32
  970. endif
  971. endif
  972. ifneq ($(NO_LAPACKE), 1)
  973. FCOMMON_OPT += -fno-second-underscore
  974. endif
  975. endif
  976. endif
  977. ifeq ($(F_COMPILER), $(filter $(F_COMPILER),GFORTRAN FLANGNEW))
  978. CCOMMON_OPT += -DF_INTERFACE_GFORT
  979. ifeq ($(F_COMPILER), GFORTRAN)
  980. FCOMMON_OPT += -Wall
  981. # make single-threaded LAPACK calls thread-safe #1847
  982. FCOMMON_OPT += -frecursive
  983. # work around ABI problem with passing single-character arguments
  984. FCOMMON_OPT += -fno-optimize-sibling-calls
  985. #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
  986. ifneq ($(NOFORTRAN), 1)
  987. ifneq ($(NOFORTRAN), 2)
  988. ifneq ($(NO_LAPACK), 1)
  989. EXTRALIB += -lgfortran
  990. endif
  991. endif
  992. endif
  993. endif
  994. ifdef NO_BINARY_MODE
  995. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  996. ifdef BINARY64
  997. FCOMMON_OPT += -mabi=64
  998. else
  999. FCOMMON_OPT += -mabi=n32
  1000. endif
  1001. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  1002. FCOMMON_OPT += -mabi=32
  1003. endif
  1004. else
  1005. ifdef BINARY64
  1006. ifneq ($(OSNAME), AIX)
  1007. ifneq ($(ARCH), riscv64)
  1008. FCOMMON_OPT += -m64
  1009. endif
  1010. endif
  1011. ifdef INTERFACE64
  1012. ifneq ($(INTERFACE64), 0)
  1013. FCOMMON_OPT += -fdefault-integer-8
  1014. endif
  1015. endif
  1016. else
  1017. ifneq ($(OSNAME), AIX)
  1018. FCOMMON_OPT += -m32
  1019. endif
  1020. endif
  1021. endif
  1022. ifeq ($(USE_OPENMP), 1)
  1023. FCOMMON_OPT += -fopenmp
  1024. endif
  1025. endif
  1026. ifeq ($(F_COMPILER), INTEL)
  1027. CCOMMON_OPT += -DF_INTERFACE_INTEL
  1028. ifdef INTERFACE64
  1029. ifneq ($(INTERFACE64), 0)
  1030. FCOMMON_OPT += -i8
  1031. endif
  1032. endif
  1033. FCOMMON_OPT += -recursive -fp-model strict -assume protect-parens
  1034. ifeq ($(USE_OPENMP), 1)
  1035. FCOMMON_OPT += -fopenmp
  1036. endif
  1037. endif
  1038. ifeq ($(F_COMPILER), FUJITSU)
  1039. CCOMMON_OPT += -DF_INTERFACE_FUJITSU
  1040. ifeq ($(USE_OPENMP), 1)
  1041. FCOMMON_OPT += -openmp
  1042. endif
  1043. endif
  1044. ifeq ($(F_COMPILER), IBM)
  1045. CCOMMON_OPT += -DF_INTERFACE_IBM
  1046. FEXTRALIB += -lxlf90
  1047. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG))
  1048. FCOMMON_OPT += -qextname -qzerosize
  1049. endif
  1050. # FCOMMON_OPT += -qarch=440
  1051. ifdef BINARY64
  1052. FCOMMON_OPT += -q64
  1053. ifdef INTERFACE64
  1054. ifneq ($(INTERFACE64), 0)
  1055. FCOMMON_OPT += -qintsize=8
  1056. endif
  1057. endif
  1058. else
  1059. FCOMMON_OPT += -q32
  1060. endif
  1061. endif
  1062. ifeq ($(F_COMPILER), PGI)
  1063. CCOMMON_OPT += -DF_INTERFACE_PGI
  1064. COMMON_PROF += -DPGICOMPILER
  1065. ifdef BINARY64
  1066. ifdef INTERFACE64
  1067. ifneq ($(INTERFACE64), 0)
  1068. FCOMMON_OPT += -i8
  1069. endif
  1070. endif
  1071. ifeq ($(ARCH), x86_64)
  1072. ifneq ($(NEWPGI2),1)
  1073. FCOMMON_OPT += -tp p7-64
  1074. else
  1075. FCOMMON_OPT += -tp px
  1076. endif
  1077. else
  1078. ifeq ($(ARCH), power)
  1079. ifeq ($(CORE), POWER6)
  1080. $(warning NVIDIA HPC compilers do not support POWER6.)
  1081. endif
  1082. ifeq ($(CORE), POWER8)
  1083. FCOMMON_OPT += -tp pwr8
  1084. endif
  1085. ifeq ($(CORE), POWER9)
  1086. FCOMMON_OPT += -tp pwr9
  1087. endif
  1088. ifeq ($(CORE), POWER10)
  1089. $(warning NVIDIA HPC compilers do not support POWER10.)
  1090. endif
  1091. endif
  1092. endif
  1093. else
  1094. FCOMMON_OPT += -tp p7
  1095. endif
  1096. FCOMMON_OPT += -Mrecursive -Kieee
  1097. ifeq ($(USE_OPENMP), 1)
  1098. FCOMMON_OPT += -mp
  1099. endif
  1100. endif
  1101. ifeq ($(F_COMPILER), PATHSCALE)
  1102. CCOMMON_OPT += -DF_INTERFACE_PATHSCALE
  1103. ifdef BINARY64
  1104. ifdef INTERFACE64
  1105. ifneq ($(INTERFACE64), 0)
  1106. FCOMMON_OPT += -i8
  1107. endif
  1108. endif
  1109. endif
  1110. ifeq ($(USE_OPENMP), 1)
  1111. FCOMMON_OPT += -mp
  1112. endif
  1113. endif
  1114. ifeq ($(F_COMPILER), OPEN64)
  1115. CCOMMON_OPT += -DF_INTERFACE_OPEN64
  1116. ifdef BINARY64
  1117. ifdef INTERFACE64
  1118. ifneq ($(INTERFACE64), 0)
  1119. FCOMMON_OPT += -i8
  1120. endif
  1121. endif
  1122. endif
  1123. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  1124. ifndef BINARY64
  1125. FCOMMON_OPT += -n32
  1126. else
  1127. FCOMMON_OPT += -n64
  1128. endif
  1129. ifeq ($(CORE), LOONGSON3R3)
  1130. FCOMMON_OPT += -loongson3 -static
  1131. endif
  1132. ifeq ($(CORE), LOONGSON3R4)
  1133. FCOMMON_OPT += -loongson3 -static
  1134. endif
  1135. else
  1136. ifndef BINARY64
  1137. FCOMMON_OPT += -m32
  1138. else
  1139. FCOMMON_OPT += -m64
  1140. endif
  1141. endif
  1142. ifeq ($(USE_OPENMP), 1)
  1143. FEXTRALIB += -lstdc++
  1144. FCOMMON_OPT += -mp
  1145. endif
  1146. endif
  1147. ifeq ($(C_COMPILER), OPEN64)
  1148. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  1149. ifndef BINARY64
  1150. CCOMMON_OPT += -n32
  1151. else
  1152. CCOMMON_OPT += -n64
  1153. endif
  1154. ifeq ($(CORE), LOONGSON3R3)
  1155. CCOMMON_OPT += -loongson3 -static
  1156. endif
  1157. ifeq ($(CORE), LOONGSON3R4)
  1158. CCOMMON_OPT += -loongson3 -static
  1159. endif
  1160. else
  1161. ifndef BINARY64
  1162. CCOMMON_OPT += -m32
  1163. else
  1164. CCOMMON_OPT += -m64
  1165. endif
  1166. endif
  1167. endif
  1168. ifeq ($(C_COMPILER), SUN)
  1169. CCOMMON_OPT += -w
  1170. ifeq ($(ARCH), x86)
  1171. CCOMMON_OPT += -m32
  1172. else
  1173. ifdef BINARY64
  1174. CCOMMON_OPT += -m64
  1175. else
  1176. CCOMMON_OPT += -m32
  1177. endif
  1178. endif
  1179. endif
  1180. ifeq ($(F_COMPILER), SUN)
  1181. CCOMMON_OPT += -DF_INTERFACE_SUN
  1182. FCOMMON_OPT += -ftrap=%none -xrecursive
  1183. ifeq ($(ARCH), x86)
  1184. FCOMMON_OPT += -m32
  1185. else
  1186. ifdef BINARY64
  1187. FCOMMON_OPT += -m64
  1188. else
  1189. FCOMMON_OPT += -m32
  1190. endif
  1191. endif
  1192. ifeq ($(USE_OPENMP), 1)
  1193. FCOMMON_OPT += -xopenmp=parallel
  1194. endif
  1195. endif
  1196. ifeq ($(F_COMPILER), COMPAQ)
  1197. CCOMMON_OPT += -DF_INTERFACE_COMPAQ
  1198. ifeq ($(USE_OPENMP), 1)
  1199. FCOMMON_OPT += -openmp
  1200. endif
  1201. endif
  1202. ifeq ($(F_COMPILER), CRAY)
  1203. CCOMMON_OPT += -DF_INTERFACE_INTEL
  1204. FCOMMON_OPT += -hnopattern
  1205. ifdef INTERFACE64
  1206. ifneq ($(INTERFACE64), 0)
  1207. FCOMMON_OPT += -s integer64
  1208. endif
  1209. endif
  1210. ifneq ($(USE_OPENMP), 1)
  1211. FCOMMON_OPT += -O noomp
  1212. endif
  1213. endif
  1214. ifdef BINARY64
  1215. ifdef INTERFACE64
  1216. ifneq ($(INTERFACE64), 0)
  1217. CCOMMON_OPT +=
  1218. #-DUSE64BITINT
  1219. endif
  1220. endif
  1221. endif
  1222. ifeq ($(NEED_PIC), 1)
  1223. ifeq ($(C_COMPILER), IBM)
  1224. CCOMMON_OPT += -qpic=large
  1225. else
  1226. CCOMMON_OPT += -fPIC
  1227. endif
  1228. ifeq ($(F_COMPILER), SUN)
  1229. FCOMMON_OPT += -pic
  1230. else ifeq ($(F_COMPILER), NAG)
  1231. FCOMMON_OPT += -PIC
  1232. else ifeq ($(F_COMPILER), IBM)
  1233. FCOMMON_OPT += -qpic=large
  1234. else
  1235. FCOMMON_OPT += -fPIC
  1236. endif
  1237. endif
  1238. ifeq ($(DYNAMIC_ARCH), 1)
  1239. CCOMMON_OPT += -DDYNAMIC_ARCH
  1240. endif
  1241. ifeq ($(DYNAMIC_OLDER), 1)
  1242. CCOMMON_OPT += -DDYNAMIC_OLDER
  1243. endif
  1244. ifeq ($(C_LAPACK), 1)
  1245. CCOMMON_OPT += -DC_LAPACK
  1246. endif
  1247. ifeq ($(NO_LAPACK), 1)
  1248. CCOMMON_OPT += -DNO_LAPACK
  1249. #Disable LAPACK C interface
  1250. NO_LAPACKE = 1
  1251. endif
  1252. ifeq ($(NO_LAPACKE), 1)
  1253. CCOMMON_OPT += -DNO_LAPACKE
  1254. endif
  1255. ifeq ($(NO_AVX), 1)
  1256. CCOMMON_OPT += -DNO_AVX
  1257. endif
  1258. ifeq ($(ARCH), x86)
  1259. CCOMMON_OPT += -DNO_AVX
  1260. endif
  1261. ifeq ($(NO_AVX2), 1)
  1262. CCOMMON_OPT += -DNO_AVX2
  1263. endif
  1264. ifeq ($(NO_AVX512), 1)
  1265. CCOMMON_OPT += -DNO_AVX512
  1266. endif
  1267. ifeq ($(NO_SVE), 1)
  1268. CCOMMON_OPT += -DNO_SVE
  1269. endif
  1270. ifdef SMP
  1271. CCOMMON_OPT += -DSMP_SERVER
  1272. ifeq ($(ARCH), mips64)
  1273. USE_SIMPLE_THREADED_LEVEL3 = 1
  1274. endif
  1275. ifeq ($(USE_OPENMP), 1)
  1276. # USE_SIMPLE_THREADED_LEVEL3 = 1
  1277. # NO_AFFINITY = 1
  1278. CCOMMON_OPT += -DUSE_OPENMP
  1279. endif
  1280. ifeq ($(BIGNUMA), 1)
  1281. CCOMMON_OPT += -DBIGNUMA
  1282. endif
  1283. endif
  1284. ifeq ($(NO_WARMUP), 1)
  1285. CCOMMON_OPT += -DNO_WARMUP
  1286. endif
  1287. ifeq ($(CONSISTENT_FPCSR), 1)
  1288. CCOMMON_OPT += -DCONSISTENT_FPCSR
  1289. endif
  1290. # Only for development
  1291. # CCOMMON_OPT += -DPARAMTEST
  1292. # CCOMMON_OPT += -DPREFETCHTEST
  1293. # CCOMMON_OPT += -DNO_SWITCHING
  1294. # USE_PAPI = 1
  1295. ifdef USE_PAPI
  1296. CCOMMON_OPT += -DUSE_PAPI
  1297. EXTRALIB += -lpapi -lperfctr
  1298. endif
  1299. ifdef BUFFERSIZE
  1300. CCOMMON_OPT += -DBUFFERSIZE=$(BUFFERSIZE)
  1301. endif
  1302. ifdef DYNAMIC_THREADS
  1303. CCOMMON_OPT += -DDYNAMIC_THREADS
  1304. endif
  1305. CCOMMON_OPT += -DMAX_CPU_NUMBER=$(NUM_THREADS)
  1306. CCOMMON_OPT += -DMAX_PARALLEL_NUMBER=$(NUM_PARALLEL)
  1307. ifdef USE_SIMPLE_THREADED_LEVEL3
  1308. CCOMMON_OPT += -DUSE_SIMPLE_THREADED_LEVEL3
  1309. endif
  1310. ifeq ($(USE_TLS), 1)
  1311. CCOMMON_OPT += -DUSE_TLS
  1312. endif
  1313. ifeq ($(BUILD_BFLOAT16), 1)
  1314. CCOMMON_OPT += -DBUILD_BFLOAT16
  1315. endif
  1316. ifeq ($(BUILD_SINGLE), 1)
  1317. CCOMMON_OPT += -DBUILD_SINGLE=1
  1318. endif
  1319. ifeq ($(BUILD_DOUBLE), 1)
  1320. CCOMMON_OPT += -DBUILD_DOUBLE=1
  1321. endif
  1322. ifeq ($(BUILD_COMPLEX), 1)
  1323. CCOMMON_OPT += -DBUILD_COMPLEX=1
  1324. endif
  1325. ifeq ($(BUILD_COMPLEX16), 1)
  1326. CCOMMON_OPT += -DBUILD_COMPLEX16=1
  1327. endif
  1328. CCOMMON_OPT += -DVERSION=\"$(VERSION)\"
  1329. ifndef SYMBOLPREFIX
  1330. SYMBOLPREFIX =
  1331. endif
  1332. ifndef SYMBOLSUFFIX
  1333. SYMBOLSUFFIX =
  1334. endif
  1335. ifndef LIBSONAMEBASE
  1336. LIBSONAMEBASE = openblas
  1337. endif
  1338. ifndef LIBNAMEPREFIX
  1339. LIBNAMEPREFIX =
  1340. endif
  1341. SYMPREFIX=$(SYMBOLPREFIX)
  1342. ifeq ($(SYMBOLPREFIX),$(LIBNAMEPREFIX))
  1343. SYMPREFIX=
  1344. endif
  1345. SYMSUFFIX=$(SYMBOLSUFFIX)
  1346. ifeq ($(SYMBOLSUFFIX),$(LIBNAMESUFFIX))
  1347. SYMSUFFIX=
  1348. endif
  1349. ifndef LIBNAMESUFFIX
  1350. LIBNAMEBASE = $(SYMPREFIX)$(LIBSONAMEBASE)$(SYMSUFFIX)
  1351. else
  1352. LIBNAMEBASE = $(SYMPREFIX)$(LIBSONAMEBASE)$(SYMSUFFIX)$(LIBNAMESUFFIX)
  1353. endif
  1354. ifeq ($(OSNAME), CYGWIN_NT)
  1355. LIBPREFIX = cyg$(LIBNAMEPREFIX)$(LIBNAMEBASE)
  1356. else
  1357. LIBPREFIX = lib$(LIBNAMEPREFIX)$(LIBNAMEBASE)
  1358. endif
  1359. KERNELDIR = $(TOPDIR)/kernel/$(ARCH)
  1360. include $(TOPDIR)/Makefile.$(ARCH)
  1361. ifneq ($(C_COMPILER), PGI)
  1362. ifneq ($(C_COMPILER), SUN)
  1363. CCOMMON_OPT += -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME
  1364. endif
  1365. endif
  1366. CCOMMON_OPT += -DASMNAME=$(FU)$(*F) -DASMFNAME=$(FU)$(*F)$(BU) -DNAME=$(*F)$(BU) -DCNAME=$(*F) -DCHAR_NAME=\"$(*F)$(BU)\" -DCHAR_CNAME=\"$(*F)\"
  1367. ifeq ($(CORE), PPC440)
  1368. CCOMMON_OPT += -DALLOC_QALLOC
  1369. endif
  1370. ifeq ($(CORE), PPC440FP2)
  1371. STATIC_ALLOCATION = 1
  1372. endif
  1373. ifneq ($(OSNAME), Linux)
  1374. NO_AFFINITY = 1
  1375. endif
  1376. ifneq ($(ARCH), x86_64)
  1377. ifneq ($(ARCH), x86)
  1378. NO_AFFINITY = 1
  1379. endif
  1380. endif
  1381. ifdef NO_AFFINITY
  1382. ifeq ($(NO_AFFINITY), 0)
  1383. override undefine NO_AFFINITY
  1384. else
  1385. CCOMMON_OPT += -DNO_AFFINITY
  1386. endif
  1387. endif
  1388. ifdef FUNCTION_PROFILE
  1389. CCOMMON_OPT += -DFUNCTION_PROFILE
  1390. endif
  1391. ifdef HUGETLB_ALLOCATION
  1392. CCOMMON_OPT += -DALLOC_HUGETLB
  1393. endif
  1394. ifdef HUGETLBFILE_ALLOCATION
  1395. CCOMMON_OPT += -DALLOC_HUGETLBFILE -DHUGETLB_FILE_NAME=$(HUGETLBFILE_ALLOCATION)
  1396. endif
  1397. ifdef STATIC_ALLOCATION
  1398. CCOMMON_OPT += -DALLOC_STATIC
  1399. endif
  1400. ifdef DEVICEDRIVER_ALLOCATION
  1401. CCOMMON_OPT += -DALLOC_DEVICEDRIVER -DDEVICEDRIVER_NAME=\"/dev/mapper\"
  1402. endif
  1403. ifdef MIXED_MEMORY_ALLOCATION
  1404. CCOMMON_OPT += -DMIXED_MEMORY_ALLOCATION
  1405. endif
  1406. ifeq ($(OSNAME), SunOS)
  1407. TAR = gtar
  1408. PATCH = gpatch
  1409. GREP = ggrep
  1410. AWK = nawk
  1411. else
  1412. TAR = tar
  1413. PATCH = patch
  1414. GREP = grep
  1415. AWK = awk
  1416. endif
  1417. ifndef MD5SUM
  1418. MD5SUM = md5sum
  1419. endif
  1420. REVISION = -r$(VERSION)
  1421. MAJOR_VERSION = $(word 1,$(subst ., ,$(VERSION)))
  1422. ifeq ($(DEBUG), 1)
  1423. COMMON_OPT += -g
  1424. endif
  1425. ifeq ($(DEBUG), 1)
  1426. FCOMMON_OPT += -g
  1427. endif
  1428. ifndef COMMON_OPT
  1429. COMMON_OPT = -O2
  1430. endif
  1431. ifndef FCOMMON_OPT
  1432. FCOMMON_OPT = -O2 -frecursive
  1433. endif
  1434. override CFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
  1435. override PFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
  1436. override FFLAGS += $(COMMON_OPT) $(FCOMMON_OPT)
  1437. override FPFLAGS += $(FCOMMON_OPT) $(COMMON_PROF)
  1438. #MAKEOVERRIDES =
  1439. ifeq ($(NEED_PIC), 1)
  1440. ifeq (,$(findstring PIC,$(FFLAGS)))
  1441. ifneq ($(F_COMPILER),IBM)
  1442. override FFLAGS += -fPIC
  1443. endif
  1444. endif
  1445. endif
  1446. #For LAPACK Fortran codes.
  1447. #Disable -fopenmp for LAPACK Fortran codes on Windows.
  1448. ifdef OS_WINDOWS
  1449. LAPACK_FFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FFLAGS))
  1450. LAPACK_FPFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FPFLAGS))
  1451. else
  1452. LAPACK_FFLAGS := $(FFLAGS)
  1453. LAPACK_FPFLAGS := $(FPFLAGS)
  1454. endif
  1455. ifeq ($(F_COMPILER),NAG)
  1456. LAPACK_FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1457. override FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1458. endif
  1459. ifeq ($(F_COMPILER),CRAY)
  1460. LAPACK_FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1461. override FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1462. endif
  1463. ifeq ($(F_COMPILER),FLANGNEW)
  1464. LAPACK_FFLAGS := $(filter-out -m32 -m64 -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1465. override FFLAGS := $(filter-out -m32 -m64 -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1466. endif
  1467. LAPACK_CFLAGS = $(CFLAGS)
  1468. LAPACK_CFLAGS += -DHAVE_LAPACK_CONFIG_H
  1469. ifdef INTERFACE64
  1470. ifneq ($(INTERFACE64), 0)
  1471. LAPACK_CFLAGS += -DLAPACK_ILP64
  1472. endif
  1473. endif
  1474. ifdef OS_WINDOWS
  1475. LAPACK_CFLAGS += -DOPENBLAS_OS_WINDOWS
  1476. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1477. endif
  1478. ifeq ($(C_COMPILER), LSB)
  1479. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1480. endif
  1481. ifndef SUFFIX
  1482. SUFFIX = o
  1483. endif
  1484. ifndef PSUFFIX
  1485. PSUFFIX = po
  1486. endif
  1487. ifndef LIBSUFFIX
  1488. LIBSUFFIX = a
  1489. endif
  1490. ifneq ($(DYNAMIC_ARCH), 1)
  1491. ifndef SMP
  1492. LIBNAME = $(LIBPREFIX)_$(LIBCORE)$(REVISION).$(LIBSUFFIX)
  1493. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)$(REVISION)_p.$(LIBSUFFIX)
  1494. else
  1495. LIBNAME = $(LIBPREFIX)_$(LIBCORE)p$(REVISION).$(LIBSUFFIX)
  1496. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)p$(REVISION)_p.$(LIBSUFFIX)
  1497. endif
  1498. else
  1499. ifndef SMP
  1500. LIBNAME = $(LIBPREFIX)$(REVISION).$(LIBSUFFIX)
  1501. LIBNAME_P = $(LIBPREFIX)$(REVISION)_p.$(LIBSUFFIX)
  1502. else
  1503. LIBNAME = $(LIBPREFIX)p$(REVISION).$(LIBSUFFIX)
  1504. LIBNAME_P = $(LIBPREFIX)p$(REVISION)_p.$(LIBSUFFIX)
  1505. endif
  1506. endif
  1507. ifeq ($(FIXED_LIBNAME),1)
  1508. LIBNAME = lib$(LIBNAMEPREFIX)$(LIBSONAMEBASE)$(LIBNAMESUFFIX).$(LIBSUFFIX)
  1509. LIBNAME_P = lib$(LIBNAMEPREFIX)$(LISOBNAMEBASE)$(LIBNAMESUFFIX)_p.$(LIBSUFFIX)
  1510. endif
  1511. LIBDLLNAME = $(LIBPREFIX).dll
  1512. IMPLIBNAME = lib$(LIBNAMEBASE).dll.a
  1513. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.so)
  1514. LIBDYNNAME = $(LIBNAME:.$(LIBSUFFIX)=.dylib)
  1515. LIBDEFNAME = $(LIBNAME:.$(LIBSUFFIX)=.def)
  1516. LIBEXPNAME = $(LIBNAME:.$(LIBSUFFIX)=.exp)
  1517. LIBZIPNAME = $(LIBNAME:.$(LIBSUFFIX)=.zip)
  1518. LIBS = $(TOPDIR)/$(LIBNAME)
  1519. LIBS_P = $(TOPDIR)/$(LIBNAME_P)
  1520. LIB_COMPONENTS = BLAS
  1521. ifneq ($(NO_CBLAS), 1)
  1522. LIB_COMPONENTS += CBLAS
  1523. endif
  1524. ifneq ($(NO_LAPACK), 1)
  1525. LIB_COMPONENTS += LAPACK
  1526. ifneq ($(NO_LAPACKE), 1)
  1527. LIB_COMPONENTS += LAPACKE
  1528. endif
  1529. ifeq ($(BUILD_RELAPACK), 1)
  1530. LIB_COMPONENTS += ReLAPACK
  1531. endif
  1532. endif
  1533. ifeq ($(ONLY_CBLAS), 1)
  1534. LIB_COMPONENTS = CBLAS
  1535. endif
  1536. export OSNAME
  1537. export ARCH
  1538. export CORE
  1539. export LIBCORE
  1540. export __BYTE_ORDER__
  1541. export ELF_VERSION
  1542. export PGCPATH
  1543. export CONFIG
  1544. export CC
  1545. export FC
  1546. export BU
  1547. export FU
  1548. export NEED2UNDERSCORES
  1549. export USE_THREAD
  1550. export NUM_THREADS
  1551. export NUM_CORES
  1552. export SMP
  1553. export MAKEFILE_RULE
  1554. export NEED_PIC
  1555. export BINARY
  1556. export BINARY32
  1557. export BINARY64
  1558. export F_COMPILER
  1559. export C_COMPILER
  1560. export USE_OPENMP
  1561. export CROSS
  1562. export CROSS_SUFFIX
  1563. export NOFORTRAN
  1564. export C_LAPACK
  1565. export NO_FBLAS
  1566. export EXTRALIB
  1567. export CEXTRALIB
  1568. export FEXTRALIB
  1569. export HAVE_SSE
  1570. export HAVE_SSE2
  1571. export HAVE_SSE3
  1572. export HAVE_SSSE3
  1573. export HAVE_SSE4_1
  1574. export HAVE_SSE4_2
  1575. export HAVE_SSE4A
  1576. export HAVE_SSE5
  1577. export HAVE_AVX
  1578. export HAVE_AVX2
  1579. export HAVE_FMA3
  1580. export HAVE_VFP
  1581. export HAVE_VFPV3
  1582. export HAVE_VFPV4
  1583. export HAVE_NEON
  1584. ifndef NO_MSA
  1585. export HAVE_MSA
  1586. export MSA_FLAGS
  1587. endif
  1588. export KERNELDIR
  1589. export FUNCTION_PROFILE
  1590. export TARGET_CORE
  1591. export NO_AVX512
  1592. export NO_AVX2
  1593. export BUILD_BFLOAT16
  1594. export NO_LSX
  1595. export NO_LASX
  1596. export SBGEMM_UNROLL_M
  1597. export SBGEMM_UNROLL_N
  1598. export SGEMM_UNROLL_M
  1599. export SGEMM_UNROLL_N
  1600. export DGEMM_UNROLL_M
  1601. export DGEMM_UNROLL_N
  1602. export QGEMM_UNROLL_M
  1603. export QGEMM_UNROLL_N
  1604. export CGEMM_UNROLL_M
  1605. export CGEMM_UNROLL_N
  1606. export ZGEMM_UNROLL_M
  1607. export ZGEMM_UNROLL_N
  1608. export XGEMM_UNROLL_M
  1609. export XGEMM_UNROLL_N
  1610. export CGEMM3M_UNROLL_M
  1611. export CGEMM3M_UNROLL_N
  1612. export ZGEMM3M_UNROLL_M
  1613. export ZGEMM3M_UNROLL_N
  1614. export XGEMM3M_UNROLL_M
  1615. export XGEMM3M_UNROLL_N
  1616. ifdef USE_CUDA
  1617. export CUDADIR
  1618. export CUCC
  1619. export CUFLAGS
  1620. export CULIB
  1621. endif
  1622. .SUFFIXES: .$(PSUFFIX) .$(SUFFIX) .f
  1623. .f.$(SUFFIX):
  1624. $(FC) $(FFLAGS) -c $< -o $(@F)
  1625. .f.$(PSUFFIX):
  1626. $(FC) $(FPFLAGS) -pg -c $< -o $(@F)
  1627. ifdef BINARY64
  1628. PATHSCALEPATH = /opt/pathscale/lib/3.1
  1629. PGIPATH = /opt/pgi/linux86-64/7.1-5/lib
  1630. else
  1631. PATHSCALEPATH = /opt/pathscale/lib/3.1/32
  1632. PGIPATH = /opt/pgi/linux86/7.1-5/lib
  1633. endif
  1634. ACMLPATH = /opt/acml/4.3.0
  1635. ifneq ($(OSNAME), Darwin)
  1636. MKLPATH = /opt/intel/mkl/10.2.2.025/lib
  1637. else
  1638. MKLPATH = /Library/Frameworks/Intel_MKL.framework/Versions/10.0.1.014/lib
  1639. endif
  1640. ATLASPATH = /opt/atlas/3.9.17/opteron
  1641. FLAMEPATH = $(HOME)/flame/lib
  1642. ifneq ($(OSNAME), SunOS)
  1643. SUNPATH = /opt/sunstudio12.1
  1644. else
  1645. SUNPATH = /opt/SUNWspro
  1646. endif