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
12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448144914501451145214531454145514561457145814591460146114621463146414651466146714681469147014711472147314741475147614771478147914801481148214831484148514861487148814891490149114921493149414951496149714981499150015011502150315041505150615071508150915101511151215131514151515161517151815191520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559156015611562156315641565156615671568156915701571157215731574157515761577157815791580158115821583158415851586158715881589159015911592159315941595159615971598159916001601160216031604160516061607160816091610161116121613161416151616161716181619162016211622162316241625162616271628162916301631163216331634163516361637163816391640164116421643164416451646164716481649165016511652165316541655165616571658165916601661166216631664166516661667166816691670167116721673167416751676167716781679168016811682168316841685168616871688168916901691169216931694169516961697169816991700170117021703170417051706170717081709171017111712171317141715171617171718171917201721172217231724172517261727172817291730173117321733173417351736173717381739174017411742174317441745174617471748174917501751175217531754175517561757175817591760176117621763176417651766176717681769177017711772177317741775177617771778177917801781178217831784178517861787178817891790179117921793179417951796179717981799180018011802180318041805180618071808180918101811181218131814181518161817181818191820182118221823182418251826182718281829183018311832183318341835183618371838183918401841184218431844184518461847184818491850185118521853185418551856185718581859186018611862186318641865186618671868186918701871187218731874187518761877187818791880188118821883188418851886188718881889189018911892189318941895189618971898189919001901190219031904190519061907190819091910191119121913191419151916
  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. else ifeq ($(ARCH), arm64)
  238. SMALL_MATRIX_OPT = 1
  239. endif
  240. ifeq ($(ARCH), loongarch64)
  241. SMALL_MATRIX_OPT = 1
  242. endif
  243. ifeq ($(SMALL_MATRIX_OPT), 1)
  244. CCOMMON_OPT += -DSMALL_MATRIX_OPT
  245. endif
  246. # This operation is expensive, so execution should be once.
  247. ifndef GOTOBLAS_MAKEFILE
  248. export GOTOBLAS_MAKEFILE = 1
  249. # Generating Makefile.conf and config.h
  250. 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)
  251. endif
  252. ifndef TARGET_CORE
  253. -include $(TOPDIR)/Makefile.conf
  254. else
  255. HAVE_NEON=
  256. HAVE_VFP=
  257. HAVE_VFPV3=
  258. HAVE_VFPV4=
  259. HAVE_MMX=
  260. HAVE_SSE=
  261. HAVE_SSE2=
  262. HAVE_SSE3=
  263. HAVE_SSSE3=
  264. HAVE_SSE4_1=
  265. HAVE_SSE4_2=
  266. HAVE_SSE4A=
  267. HAVE_SSE5=
  268. HAVE_AVX=
  269. HAVE_AVX2=
  270. HAVE_FMA3=
  271. include $(TOPDIR)/Makefile_kernel.conf
  272. endif
  273. ifndef NUM_PARALLEL
  274. NUM_PARALLEL = 1
  275. endif
  276. ifndef NUM_THREADS
  277. NUM_THREADS = $(NUM_CORES)
  278. endif
  279. ifeq ($(NUM_THREADS), 1)
  280. override USE_THREAD = 0
  281. override USE_OPENMP = 0
  282. endif
  283. ifdef USE_THREAD
  284. ifeq ($(USE_THREAD), 0)
  285. SMP =
  286. else
  287. SMP = 1
  288. endif
  289. else
  290. ifeq ($(NUM_THREADS), 1)
  291. SMP =
  292. else
  293. SMP = 1
  294. endif
  295. endif
  296. ifeq ($(SMP), 1)
  297. USE_LOCKING =
  298. endif
  299. ifndef NEED_PIC
  300. NEED_PIC = 1
  301. endif
  302. ARFLAGS =
  303. CPP = $(COMPILER) -E
  304. AR ?= $(CROSS_SUFFIX)ar
  305. AS ?= $(CROSS_SUFFIX)as
  306. LD ?= $(CROSS_SUFFIX)ld
  307. RANLIB ?= $(CROSS_SUFFIX)ranlib
  308. NM = $(CROSS_SUFFIX)nm
  309. DLLWRAP = $(CROSS_SUFFIX)dllwrap
  310. OBJCOPY = $(CROSS_SUFFIX)objcopy
  311. OBJCONV = $(CROSS_SUFFIX)objconv
  312. # When fortran support was either not detected or actively deselected, only build BLAS.
  313. ifeq ($(NOFORTRAN), 1)
  314. C_LAPACK = 1
  315. override FEXTRALIB =
  316. ifeq ($(C_COMPILER), GCC)
  317. CCOMMON_OPT += -Wno-error=incompatible-pointer-types
  318. endif
  319. endif
  320. ifeq ($(C_COMPILER), GCC)
  321. GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
  322. GCCVERSIONGT4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 4)
  323. GCCVERSIONGT5 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 5)
  324. GCCVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 7)
  325. GCCVERSIONGTEQ8 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 8)
  326. GCCVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  327. GCCVERSIONGTEQ10 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 10)
  328. GCCVERSIONGTEQ11 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 11)
  329. GCCVERSIONGTEQ12 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 12)
  330. # Note that the behavior of -dumpversion is compile-time-configurable for
  331. # gcc-7.x and newer. Use -dumpfullversion there
  332. ifeq ($(GCCVERSIONGTEQ7),1)
  333. GCCDUMPVERSION_PARAM := -dumpfullversion
  334. else
  335. GCCDUMPVERSION_PARAM := -dumpversion
  336. endif
  337. GCCMINORVERSIONGTEQ1 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 1)
  338. GCCMINORVERSIONGTEQ2 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 2)
  339. GCCMINORVERSIONGTEQ4 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 4)
  340. GCCMINORVERSIONGTEQ7 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 7)
  341. endif
  342. ifeq ($(C_COMPILER), CLANG)
  343. CLANGVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  344. CLANGVERSIONGTEQ12 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 12)
  345. endif
  346. #
  347. # OS dependent settings
  348. #
  349. ifeq ($(OSNAME), Darwin)
  350. ifndef MACOSX_DEPLOYMENT_TARGET
  351. ifeq ($(ARCH), arm64)
  352. export MACOSX_DEPLOYMENT_TARGET=11.0
  353. ifeq ($(C_COMPILER), GCC)
  354. export NO_SVE = 1
  355. endif
  356. else
  357. export MACOSX_DEPLOYMENT_TARGET=10.8
  358. endif
  359. endif
  360. MD5SUM = md5 -r
  361. XCVER = $(shell pkgutil --pkg-info=com.apple.pkg.Xcode |awk '/version:/ {print $2}'|cut -d: -f2|cut -f1 -d.)
  362. ifeq (x$(XCVER)x,xx)
  363. XCVER = $(shell pkgutil --pkg-info=com.apple.pkg.CLTools_Executables |awk '/version:/ {print $2}'|cut -d: -f2|cut -f1 -d.)
  364. endif
  365. ifeq (x$(XCVER), x 15)
  366. CCOMMON_OPT += -Wl,-ld_classic
  367. FCOMMON_OPT += -Wl,-ld_classic
  368. endif
  369. endif
  370. ifneq (,$(findstring $(OSNAME), FreeBSD OpenBSD DragonFly))
  371. MD5SUM = md5 -r
  372. endif
  373. ifeq ($(OSNAME), NetBSD)
  374. MD5SUM = md5 -n
  375. endif
  376. ifeq ($(OSNAME), Linux)
  377. EXTRALIB += -lm
  378. NO_EXPRECISION = 1
  379. endif
  380. ifeq ($(OSNAME), Android)
  381. EXTRALIB += -lm
  382. endif
  383. ifeq ($(OSNAME), AIX)
  384. EXTRALIB += -lm
  385. endif
  386. ifeq ($(OSNAME), $(filter $(OSNAME),FreeBSD OpenBSD NetBSD DragonFly))
  387. ifeq ($(ARCH), $(filter $(ARCH),arm arm64))
  388. EXTRALIB += -lm
  389. endif
  390. endif
  391. ifeq ($(OSNAME), WINNT)
  392. NEED_PIC = 0
  393. NO_EXPRECISION = 1
  394. EXTRALIB += -defaultlib:advapi32
  395. SUFFIX = obj
  396. PSUFFIX = pobj
  397. LIBSUFFIX = a
  398. ifeq ($(C_COMPILER), CLANG)
  399. CCOMMON_OPT += -DMS_ABI
  400. endif
  401. #Version tests for supporting specific features (MS_ABI, POWER9 intrinsics)
  402. ifeq ($(GCCVERSIONGT4), 1)
  403. # GCC Major version > 4
  404. # It is compatible with MSVC ABI.
  405. CCOMMON_OPT += -DMS_ABI
  406. endif
  407. ifeq ($(GCCVERSIONGTEQ4), 1)
  408. ifeq ($(GCCMINORVERSIONGTEQ7), 1)
  409. # GCC Version >=4.7
  410. # It is compatible with MSVC ABI.
  411. CCOMMON_OPT += -DMS_ABI
  412. endif
  413. endif
  414. # Ensure the correct stack alignment on Win32
  415. # http://permalink.gmane.org/gmane.comp.lib.openblas.general/97
  416. ifeq ($(ARCH), x86)
  417. CCOMMON_OPT += -mincoming-stack-boundary=2
  418. FCOMMON_OPT += -mincoming-stack-boundary=2
  419. endif
  420. endif
  421. ifeq ($(OSNAME), Interix)
  422. NEED_PIC = 0
  423. NO_EXPRECISION = 1
  424. INTERIX_TOOL_DIR = /opt/gcc.3.3/i586-pc-interix3/bin
  425. endif
  426. ifeq ($(OSNAME), CYGWIN_NT)
  427. NEED_PIC = 0
  428. NO_EXPRECISION = 1
  429. OS_CYGWIN_NT = 1
  430. endif
  431. ifneq ($(OSNAME), WINNT)
  432. ifneq ($(OSNAME), CYGWIN_NT)
  433. ifneq ($(OSNAME), Interix)
  434. ifneq ($(OSNAME), Android)
  435. ifdef SMP
  436. EXTRALIB += -lpthread
  437. endif
  438. endif
  439. endif
  440. endif
  441. endif
  442. # ifeq logical or
  443. ifeq ($(OSNAME), $(filter $(OSNAME),WINNT CYGWIN_NT Interix))
  444. OS_WINDOWS=1
  445. endif
  446. ifdef QUAD_PRECISION
  447. CCOMMON_OPT += -DQUAD_PRECISION
  448. NO_EXPRECISION = 1
  449. endif
  450. ifneq ($(ARCH), x86)
  451. ifneq ($(ARCH), x86_64)
  452. NO_EXPRECISION = 1
  453. endif
  454. endif
  455. ifdef UTEST_CHECK
  456. CCOMMON_OPT += -DUTEST_CHECK
  457. SANITY_CHECK = 1
  458. endif
  459. ifdef SANITY_CHECK
  460. CCOMMON_OPT += -DSANITY_CHECK -DREFNAME=$(*F)f$(BU)
  461. endif
  462. MAX_STACK_ALLOC ?= 2048
  463. ifneq ($(MAX_STACK_ALLOC), 0)
  464. CCOMMON_OPT += -DMAX_STACK_ALLOC=$(MAX_STACK_ALLOC)
  465. endif
  466. ifdef USE_LOCKING
  467. ifneq ($(USE_LOCKING), 0)
  468. CCOMMON_OPT += -DUSE_LOCKING
  469. endif
  470. endif
  471. #
  472. # Architecture dependent settings
  473. #
  474. ifeq ($(ARCH), x86)
  475. ifndef BINARY
  476. NO_BINARY_MODE = 1
  477. endif
  478. ifeq ($(CORE), generic)
  479. NO_EXPRECISION = 1
  480. endif
  481. ifndef NO_EXPRECISION
  482. ifeq ($(F_COMPILER), GFORTRAN)
  483. # ifeq logical or. GCC or LSB
  484. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  485. EXPRECISION = 1
  486. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  487. FCOMMON_OPT += -m128bit-long-double
  488. endif
  489. ifeq ($(C_COMPILER), CLANG)
  490. EXPRECISION = 1
  491. CCOMMON_OPT += -DEXPRECISION
  492. FCOMMON_OPT += -m128bit-long-double
  493. endif
  494. endif
  495. endif
  496. endif
  497. ifeq ($(ARCH), x86_64)
  498. ifeq ($(CORE), generic)
  499. NO_EXPRECISION = 1
  500. endif
  501. ifndef NO_EXPRECISION
  502. ifeq ($(F_COMPILER), GFORTRAN)
  503. # ifeq logical or. GCC or LSB
  504. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  505. EXPRECISION = 1
  506. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  507. FCOMMON_OPT += -m128bit-long-double
  508. endif
  509. ifeq ($(C_COMPILER), CLANG)
  510. EXPRECISION = 1
  511. CCOMMON_OPT += -DEXPRECISION
  512. FCOMMON_OPT += -m128bit-long-double
  513. endif
  514. endif
  515. endif
  516. endif
  517. ifeq ($(C_COMPILER), INTEL)
  518. CCOMMON_OPT += -wd981
  519. endif
  520. ifeq ($(USE_OPENMP), 1)
  521. #check
  522. ifeq ($(USE_THREAD), 0)
  523. $(error OpenBLAS: Cannot set both USE_OPENMP=1 and USE_THREAD=0. The USE_THREAD=0 is only for building single thread version.)
  524. endif
  525. # ifeq logical or. GCC or LSB
  526. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  527. CCOMMON_OPT += -fopenmp
  528. endif
  529. ifeq ($(C_COMPILER), CLANG)
  530. CCOMMON_OPT += -fopenmp
  531. ifeq ($(F_COMPILER), GFORTRAN)
  532. FEXTRALIB := $(subst -lgomp,-lomp,$(FEXTRALIB))
  533. endif
  534. endif
  535. ifeq ($(C_COMPILER), INTEL)
  536. CCOMMON_OPT += -fopenmp
  537. endif
  538. ifeq ($(C_COMPILER), PGI)
  539. CCOMMON_OPT += -mp
  540. endif
  541. ifeq ($(C_COMPILER), OPEN64)
  542. CCOMMON_OPT += -mp
  543. CEXTRALIB += -lstdc++
  544. endif
  545. ifeq ($(C_COMPILER), PATHSCALE)
  546. CCOMMON_OPT += -mp
  547. endif
  548. endif
  549. ifeq ($(DYNAMIC_ARCH), 1)
  550. ifeq ($(ARCH), x86)
  551. DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
  552. CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
  553. endif
  554. ifeq ($(ARCH), x86_64)
  555. DYNAMIC_CORE = PRESCOTT CORE2
  556. ifeq ($(DYNAMIC_OLDER), 1)
  557. DYNAMIC_CORE += PENRYN DUNNINGTON
  558. endif
  559. DYNAMIC_CORE += NEHALEM
  560. ifeq ($(DYNAMIC_OLDER), 1)
  561. DYNAMIC_CORE += OPTERON OPTERON_SSE3
  562. endif
  563. DYNAMIC_CORE += BARCELONA
  564. ifeq ($(DYNAMIC_OLDER), 1)
  565. DYNAMIC_CORE += BOBCAT ATOM NANO
  566. endif
  567. ifneq ($(NO_AVX), 1)
  568. DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER STEAMROLLER EXCAVATOR
  569. endif
  570. ifneq ($(NO_AVX2), 1)
  571. DYNAMIC_CORE += HASWELL ZEN
  572. endif
  573. ifneq ($(NO_AVX512), 1)
  574. ifneq ($(NO_AVX2), 1)
  575. DYNAMIC_CORE += SKYLAKEX COOPERLAKE SAPPHIRERAPIDS
  576. endif
  577. endif
  578. endif
  579. ifdef DYNAMIC_LIST
  580. override DYNAMIC_CORE = PRESCOTT $(DYNAMIC_LIST)
  581. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_PRESCOTT
  582. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  583. CCOMMON_OPT += $(XCCOMMON_OPT)
  584. #CCOMMON_OPT += -DDYNAMIC_LIST='$(DYNAMIC_LIST)'
  585. endif
  586. ifeq ($(ARCH), arm64)
  587. DYNAMIC_CORE = ARMV8
  588. DYNAMIC_CORE += CORTEXA53
  589. DYNAMIC_CORE += CORTEXA57
  590. DYNAMIC_CORE += NEOVERSEN1
  591. ifneq ($(NO_SVE), 1)
  592. DYNAMIC_CORE += NEOVERSEV1
  593. DYNAMIC_CORE += NEOVERSEN2
  594. DYNAMIC_CORE += ARMV8SVE
  595. DYNAMIC_CORE += A64FX
  596. endif
  597. DYNAMIC_CORE += THUNDERX
  598. DYNAMIC_CORE += THUNDERX2T99
  599. DYNAMIC_CORE += TSV110
  600. DYNAMIC_CORE += EMAG8180
  601. DYNAMIC_CORE += THUNDERX3T110
  602. ifdef DYNAMIC_LIST
  603. override DYNAMIC_CORE = ARMV8 $(DYNAMIC_LIST)
  604. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_ARMV8
  605. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  606. endif
  607. endif
  608. ifeq ($(ARCH), mips64)
  609. DYNAMIC_CORE = LOONGSON3R3 LOONGSON3R4 MIPS64_GENERIC
  610. ifdef DYNAMIC_LIST
  611. override DYNAMIC_CORE = MIPS64_GENERIC $(DYNAMIC_LIST)
  612. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_MIPS64_GENERIC
  613. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  614. endif
  615. endif
  616. ifeq ($(ARCH), loongarch64)
  617. DYNAMIC_CORE = LOONGSON3R5 LOONGSON2K1000 LOONGSONGENERIC
  618. endif
  619. ifeq ($(ARCH), riscv64)
  620. DYNAMIC_CORE = RISCV64_GENERIC
  621. DYNAMIC_CORE += RISCV64_ZVL128B
  622. DYNAMIC_CORE += RISCV64_ZVL256B
  623. ifdef DYNAMIC_LIST
  624. override DYNAMIC_CORE = RISCV64_GENERIC $(DYNAMIC_LIST)
  625. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_RISCV64_GENERIC
  626. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  627. endif
  628. endif
  629. ifeq ($(ARCH), zarch)
  630. DYNAMIC_CORE = ZARCH_GENERIC
  631. # if the compiler accepts -march=arch11 or -march=z13 and can compile a file
  632. # with z13-specific inline assembly, then we can include support for Z13.
  633. # note: -march=z13 is equivalent to -march=arch11 yet some compiler releases
  634. # only support one or the other.
  635. # note: LLVM version 6.x supported -march=z13 yet could not handle vector
  636. # registers in inline assembly, so the check for supporting the -march flag is
  637. # not enough.
  638. ZARCH_TEST_COMPILE=-c $(TOPDIR)/kernel/zarch/damin_z13.c -I$(TOPDIR) -o /dev/null > /dev/null 2> /dev/null
  639. ZARCH_CC_SUPPORTS_ARCH11=$(shell $(CC) -march=arch11 $(ZARCH_TEST_COMPILE) && echo 1)
  640. ZARCH_CC_SUPPORTS_Z13=$(shell $(CC) -march=z13 $(ZARCH_TEST_COMPILE) && echo 1)
  641. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH11), $(ZARCH_CC_SUPPORTS_Z13)), 1)
  642. DYNAMIC_CORE += Z13
  643. CCOMMON_OPT += -DDYN_Z13
  644. else
  645. $(info OpenBLAS: Not building Z13 kernels because the compiler $(CC) does not support it)
  646. endif
  647. # as above for z13, check for -march=arch12 and z14 support in the compiler.
  648. ZARCH_CC_SUPPORTS_ARCH12=$(shell $(CC) -march=arch12 $(ZARCH_TEST_COMPILE) && echo 1)
  649. ZARCH_CC_SUPPORTS_Z14=$(shell $(CC) -march=z14 $(ZARCH_TEST_COMPILE) && echo 1)
  650. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH12), $(ZARCH_CC_SUPPORTS_Z14)), 1)
  651. DYNAMIC_CORE += Z14
  652. CCOMMON_OPT += -DDYN_Z14
  653. else
  654. $(info OpenBLAS: Not building Z14 kernels because the compiler $(CC) does not support it)
  655. endif
  656. endif # ARCH zarch
  657. ifeq ($(ARCH), power)
  658. ifneq ($(C_COMPILER), PGI)
  659. DYNAMIC_CORE = POWER6
  660. DYNAMIC_CORE += POWER8
  661. ifneq ($(C_COMPILER), GCC)
  662. DYNAMIC_CORE += POWER9
  663. DYNAMIC_CORE += POWER10
  664. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  665. endif
  666. ifeq ($(C_COMPILER), GCC)
  667. ifeq ($(GCCVERSIONGT5), 1)
  668. DYNAMIC_CORE += POWER9
  669. else
  670. $(info, OpenBLAS: Your gcc version is too old to build the POWER9 kernels.)
  671. endif
  672. ifeq ($(OSNAME), AIX)
  673. LDVERSIONGTEQ35 := 1
  674. else
  675. LDVERSIONGTEQ35 := $(shell expr `$(CC) -Wl,--version 2> /dev/null | head -1 | cut -f2 -d "." | cut -f1 -d "-"` \>= 35)
  676. endif
  677. ifeq ($(GCCVERSIONGTEQ11)$(LDVERSIONGTEQ35), 11)
  678. DYNAMIC_CORE += POWER10
  679. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  680. else ifeq ($(GCCVERSIONGTEQ10), 1)
  681. ifeq ($(GCCMINORVERSIONGTEQ2)$(LDVERSIONGTEQ35), 11)
  682. DYNAMIC_CORE += POWER10
  683. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  684. endif
  685. else
  686. $(info, OpenBLAS: Your gcc version is too old to build the POWER10 kernels.)
  687. endif
  688. endif
  689. else
  690. DYNAMIC_CORE = POWER8
  691. DYNAMIC_CORE += POWER9
  692. endif
  693. endif
  694. # If DYNAMIC_CORE is not set, DYNAMIC_ARCH cannot do anything, so force it to empty
  695. ifndef DYNAMIC_CORE
  696. override DYNAMIC_ARCH=
  697. endif
  698. endif
  699. ifeq ($(ARCH), ia64)
  700. NO_BINARY_MODE = 1
  701. BINARY_DEFINED = 1
  702. ifeq ($(F_COMPILER), GFORTRAN)
  703. ifeq ($(C_COMPILER), GCC)
  704. # EXPRECISION = 1
  705. # CCOMMON_OPT += -DEXPRECISION
  706. endif
  707. endif
  708. endif
  709. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  710. NO_BINARY_MODE = 1
  711. endif
  712. ifeq ($(ARCH), alpha)
  713. NO_BINARY_MODE = 1
  714. BINARY_DEFINED = 1
  715. endif
  716. ifeq ($(ARCH), arm)
  717. NO_BINARY_MODE = 1
  718. BINARY_DEFINED = 1
  719. ifneq ($(EMBEDDED), 1)
  720. CCOMMON_OPT += -marm
  721. FCOMMON_OPT += -marm
  722. else
  723. CCOMMON_OPT += -DOS_EMBEDDED -mthumb -mcpu=cortex-m4 -mfloat-abi=hard -mfpu=fpv4-sp-d16
  724. endif
  725. # If softfp abi is mentioned on the command line, force it.
  726. ifeq ($(ARM_SOFTFP_ABI), 1)
  727. CCOMMON_OPT += -mfloat-abi=softfp
  728. FCOMMON_OPT += -mfloat-abi=softfp
  729. endif
  730. ifeq ($(OSNAME), Android)
  731. ifeq ($(ARM_SOFTFP_ABI), 1)
  732. EXTRALIB += -lm
  733. else
  734. EXTRALIB += -Wl,-lm_hard
  735. endif
  736. endif
  737. endif
  738. ifeq ($(ARCH), arm64)
  739. NO_BINARY_MODE = 1
  740. BINARY_DEFINED = 1
  741. ifdef INTERFACE64
  742. ifneq ($(INTERFACE64), 0)
  743. ifeq ($(F_COMPILER), GFORTRAN)
  744. FCOMMON_OPT += -fdefault-integer-8
  745. endif
  746. ifeq ($(F_COMPILER), FLANG)
  747. FCOMMON_OPT += -i8
  748. endif
  749. endif
  750. endif
  751. endif
  752. ifeq ($(ARCH), riscv64)
  753. NO_BINARY_MODE = 1
  754. BINARY_DEFINED = 1
  755. ifdef INTERFACE64
  756. ifneq ($(INTERFACE64), 0)
  757. ifeq ($(F_COMPILER), GFORTRAN)
  758. FCOMMON_OPT += -fdefault-integer-8
  759. endif
  760. ifeq ($(F_COMPILER), FLANG)
  761. FCOMMON_OPT += -i8
  762. endif
  763. endif
  764. endif
  765. endif
  766. ifeq ($(ARCH), loongarch64)
  767. NO_BINARY_MODE = 1
  768. BINARY_DEFINED = 1
  769. ifdef INTERFACE64
  770. ifneq ($(INTERFACE64), 0)
  771. ifeq ($(F_COMPILER), GFORTRAN)
  772. FCOMMON_OPT += -fdefault-integer-8
  773. endif
  774. ifeq ($(F_COMPILER), FLANG)
  775. FCOMMON_OPT += -i8
  776. endif
  777. endif
  778. endif
  779. endif
  780. ifeq ($(ARCH), csky)
  781. NO_BINARY_MODE = 1
  782. BINARY_DEFINED = 1
  783. endif
  784. #
  785. # C Compiler dependent settings
  786. #
  787. # ifeq logical or. GCC or CLANG or LSB
  788. # http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
  789. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG LSB))
  790. CCOMMON_OPT += -Wall
  791. COMMON_PROF += -fno-inline
  792. NO_UNINITIALIZED_WARN = -Wno-uninitialized
  793. ifeq ($(QUIET_MAKE), 1)
  794. CCOMMON_OPT += $(NO_UNINITIALIZED_WARN) -Wno-unused
  795. endif
  796. ifdef NO_BINARY_MODE
  797. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  798. ifdef BINARY64
  799. CCOMMON_OPT += -mabi=64
  800. else
  801. CCOMMON_OPT += -mabi=n32
  802. endif
  803. BINARY_DEFINED = 1
  804. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  805. CCOMMON_OPT += -mabi=32
  806. BINARY_DEFINED = 1
  807. endif
  808. ifneq (, $(filter $(CORE), MIPS64_GENERIC))
  809. CCOMMON_OPT += -DNO_MSA
  810. FCOMMON_OPT += -DNO_MSA
  811. endif
  812. ifneq (, $(filter $(CORE),LOONGSON3R3 LOONGSON3R4))
  813. CCOMMON_OPT += -march=loongson3a
  814. FCOMMON_OPT += -march=loongson3a
  815. endif
  816. ifeq ($(CORE), MIPS24K)
  817. CCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  818. FCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  819. endif
  820. ifeq ($(CORE), MIPS1004K)
  821. CCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  822. FCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  823. endif
  824. ifeq ($(CORE), P5600)
  825. CCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  826. FCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  827. endif
  828. ifeq ($(CORE), I6400)
  829. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  830. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  831. endif
  832. ifeq ($(CORE), P6600)
  833. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  834. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  835. endif
  836. ifeq ($(CORE), I6500)
  837. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  838. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  839. endif
  840. ifeq ($(OSNAME), AIX)
  841. BINARY_DEFINED = 1
  842. endif
  843. ifeq ($(ARCH), loongarch64)
  844. LA64_ABI=$(shell $(CC) -mabi=lp64d -c $(TOPDIR)/cpuid_loongarch64.c -o /dev/null > /dev/null 2> /dev/null && echo lp64d)
  845. LA64_ARCH=$(shell $(CC) -march=loongarch64 -c $(TOPDIR)/cpuid_loongarch64.c -o /dev/null > /dev/null 2> /dev/null && echo loongarch64)
  846. ifneq ($(LA64_ABI), lp64d)
  847. LA64_ABI=lp64
  848. endif
  849. ifneq ($(LA64_ARCH), loongarch64)
  850. CCOMMON_OPT += -mabi=$(LA64_ABI)
  851. FCOMMON_OPT += -mabi=$(LA64_ABI)
  852. else
  853. CCOMMON_OPT += -march=loongarch64 -mabi=$(LA64_ABI)
  854. FCOMMON_OPT += -march=loongarch64 -mabi=$(LA64_ABI)
  855. endif
  856. endif
  857. endif
  858. ifndef BINARY_DEFINED
  859. ifneq ($(OSNAME), AIX)
  860. ifdef BINARY64
  861. ifneq ($(ARCH), riscv64)
  862. CCOMMON_OPT += -m64
  863. endif
  864. else
  865. CCOMMON_OPT += -m32
  866. endif
  867. endif
  868. endif
  869. endif
  870. ifeq ($(C_COMPILER), PGI)
  871. PGCVERSIONGT20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \> 20)
  872. PGCVERSIONEQ20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` == 20)
  873. PGCMINORVERSIONGE11 := $(shell expr `$(CC) --version|sed -n "2p" |cut -d "-" -f 1 |sed -e "s/[^0-9.]//g" |cut -c 4-5` \>= 11)
  874. PGCVERSIONCHECK := $(PGCVERSIONGT20)$(PGCVERSIONEQ20)$(PGCMINORVERSIONGE11)
  875. ifeq ($(PGCVERSIONCHECK), $(filter $(PGCVERSIONCHECK), 100 101 011))
  876. NEWPGI := 1
  877. PGCVERSIONGT21 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \> 21)
  878. PGCVERSIONEQ21 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` == 21)
  879. PGCVERSIONCHECK2 := $(PGCVERSIONGT21)$(PGCVERSIONEQ21)$(PGCMINORVERSIONGE11)
  880. ifeq ($(PGCVERSIONCHECK2), $(filter $(PGCVERSIONCHECK2), 100 101 011))
  881. NEWPGI2 := 1
  882. endif
  883. endif
  884. ifdef BINARY64
  885. ifeq ($(ARCH), x86_64)
  886. ifeq (,$(findstring tp,$(CFLAGS)))
  887. ifneq ($(NEWPGI2),1)
  888. CCOMMON_OPT += -tp p7-64
  889. else
  890. CCOMMON_OPT += -tp px
  891. endif
  892. endif
  893. ifneq ($(NEWPGI),1)
  894. CCOMMON_OPT += -D__MMX__ -Mnollvm
  895. endif
  896. else
  897. ifeq ($(ARCH), power)
  898. ifeq (,$(findstring tp,$(CFLAGS)))
  899. ifeq ($(CORE), POWER8)
  900. CCOMMON_OPT += -tp pwr8
  901. endif
  902. ifeq ($(CORE), POWER9)
  903. CCOMMON_OPT += -tp pwr9
  904. endif
  905. endif
  906. endif
  907. endif
  908. else
  909. ifneq ($(NEWPGI2),1)
  910. ifeq (,$(findstring tp,$(CFLAGS)))
  911. CCOMMON_OPT += -tp p7
  912. else
  913. CCOMMON_OPT += -tp px
  914. endif
  915. endif
  916. endif
  917. endif
  918. ifeq ($(C_COMPILER), PATHSCALE)
  919. ifdef BINARY64
  920. CCOMMON_OPT += -m64
  921. else
  922. CCOMMON_OPT += -m32
  923. endif
  924. endif
  925. #
  926. # Fortran Compiler dependent settings
  927. #
  928. ifeq ($(F_COMPILER), NAG)
  929. FCOMMON_OPT += -dcfuns -recursive -ieee=full -w=obs -thread_safe
  930. ifdef INTERFACE64
  931. ifneq ($(INTERFACE64), 0)
  932. FCOMMON_OPT += -i8
  933. endif
  934. endif
  935. ifeq ($(USE_OPENMP), 1)
  936. FCOMMON_OPT += -openmp
  937. endif
  938. endif
  939. ifeq ($(F_COMPILER), FLANG)
  940. CCOMMON_OPT += -DF_INTERFACE_FLANG
  941. FCOMMON_OPT += -Mrecursive -Kieee
  942. ifeq ($(OSNAME), Linux)
  943. ifeq ($(ARCH), x86_64)
  944. FLANG_VENDOR := $(shell $(FC) --version|head -1 |cut -f 1 -d " ")
  945. ifeq ($(FLANG_VENDOR), AMD)
  946. FCOMMON_OPT += -fno-unroll-loops
  947. endif
  948. endif
  949. endif
  950. ifdef BINARY64
  951. ifdef INTERFACE64
  952. ifneq ($(INTERFACE64), 0)
  953. FCOMMON_OPT += -i8
  954. endif
  955. endif
  956. FCOMMON_OPT += -Wall
  957. else
  958. FCOMMON_OPT += -Wall
  959. endif
  960. ifeq ($(USE_OPENMP), 1)
  961. FCOMMON_OPT += -fopenmp
  962. endif
  963. endif
  964. ifeq ($(F_COMPILER), G77)
  965. CCOMMON_OPT += -DF_INTERFACE_G77
  966. FCOMMON_OPT += -Wall
  967. ifndef NO_BINARY_MODE
  968. ifneq ($(OSNAME), AIX)
  969. ifdef BINARY64
  970. FCOMMON_OPT += -m64
  971. else
  972. FCOMMON_OPT += -m32
  973. endif
  974. endif
  975. endif
  976. endif
  977. ifeq ($(F_COMPILER), G95)
  978. CCOMMON_OPT += -DF_INTERFACE_G95
  979. FCOMMON_OPT += -Wall
  980. ifneq ($(OSNAME), AIX)
  981. ifndef NO_BINARY_MODE
  982. ifdef BINARY64
  983. FCOMMON_OPT += -m64
  984. else
  985. FCOMMON_OPT += -m32
  986. endif
  987. endif
  988. ifneq ($(NO_LAPACKE), 1)
  989. FCOMMON_OPT += -fno-second-underscore
  990. endif
  991. endif
  992. endif
  993. ifeq ($(F_COMPILER), $(filter $(F_COMPILER),GFORTRAN FLANGNEW))
  994. CCOMMON_OPT += -DF_INTERFACE_GFORT
  995. ifeq ($(F_COMPILER), GFORTRAN)
  996. FCOMMON_OPT += -Wall
  997. # make single-threaded LAPACK calls thread-safe #1847
  998. FCOMMON_OPT += -frecursive
  999. # work around ABI problem with passing single-character arguments
  1000. FCOMMON_OPT += -fno-optimize-sibling-calls
  1001. #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
  1002. ifneq ($(NOFORTRAN), 1)
  1003. ifneq ($(NOFORTRAN), 2)
  1004. ifneq ($(NO_LAPACK), 1)
  1005. EXTRALIB += -lgfortran
  1006. endif
  1007. endif
  1008. endif
  1009. endif
  1010. ifdef NO_BINARY_MODE
  1011. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  1012. ifdef BINARY64
  1013. FCOMMON_OPT += -mabi=64
  1014. else
  1015. FCOMMON_OPT += -mabi=n32
  1016. endif
  1017. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  1018. FCOMMON_OPT += -mabi=32
  1019. endif
  1020. else
  1021. ifdef BINARY64
  1022. ifneq ($(OSNAME), AIX)
  1023. ifneq ($(ARCH), riscv64)
  1024. FCOMMON_OPT += -m64
  1025. endif
  1026. endif
  1027. ifdef INTERFACE64
  1028. ifneq ($(INTERFACE64), 0)
  1029. FCOMMON_OPT += -fdefault-integer-8
  1030. endif
  1031. endif
  1032. else
  1033. ifneq ($(OSNAME), AIX)
  1034. FCOMMON_OPT += -m32
  1035. endif
  1036. endif
  1037. endif
  1038. ifeq ($(USE_OPENMP), 1)
  1039. FCOMMON_OPT += -fopenmp
  1040. endif
  1041. endif
  1042. ifeq ($(F_COMPILER), INTEL)
  1043. CCOMMON_OPT += -DF_INTERFACE_INTEL
  1044. ifdef INTERFACE64
  1045. ifneq ($(INTERFACE64), 0)
  1046. FCOMMON_OPT += -i8
  1047. endif
  1048. endif
  1049. FCOMMON_OPT += -recursive -fp-model strict -assume protect-parens
  1050. ifeq ($(USE_OPENMP), 1)
  1051. FCOMMON_OPT += -fopenmp
  1052. endif
  1053. endif
  1054. ifeq ($(F_COMPILER), FUJITSU)
  1055. CCOMMON_OPT += -DF_INTERFACE_FUJITSU
  1056. ifeq ($(USE_OPENMP), 1)
  1057. FCOMMON_OPT += -openmp
  1058. endif
  1059. endif
  1060. ifeq ($(F_COMPILER), IBM)
  1061. CCOMMON_OPT += -DF_INTERFACE_IBM
  1062. FEXTRALIB += -lxlf90
  1063. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG))
  1064. FCOMMON_OPT += -qextname -qzerosize
  1065. endif
  1066. # FCOMMON_OPT += -qarch=440
  1067. ifdef BINARY64
  1068. FCOMMON_OPT += -q64
  1069. ifdef INTERFACE64
  1070. ifneq ($(INTERFACE64), 0)
  1071. FCOMMON_OPT += -qintsize=8
  1072. endif
  1073. endif
  1074. else
  1075. FCOMMON_OPT += -q32
  1076. endif
  1077. endif
  1078. ifeq ($(F_COMPILER), PGI)
  1079. CCOMMON_OPT += -DF_INTERFACE_PGI
  1080. COMMON_PROF += -DPGICOMPILER
  1081. ifdef BINARY64
  1082. ifdef INTERFACE64
  1083. ifneq ($(INTERFACE64), 0)
  1084. FCOMMON_OPT += -i8
  1085. endif
  1086. endif
  1087. ifeq ($(ARCH), x86_64)
  1088. ifneq ($(NEWPGI2),1)
  1089. FCOMMON_OPT += -tp p7-64
  1090. else
  1091. FCOMMON_OPT += -tp px
  1092. endif
  1093. else
  1094. ifeq ($(ARCH), power)
  1095. ifeq ($(CORE), POWER6)
  1096. $(warning NVIDIA HPC compilers do not support POWER6.)
  1097. endif
  1098. ifeq ($(CORE), POWER8)
  1099. FCOMMON_OPT += -tp pwr8
  1100. endif
  1101. ifeq ($(CORE), POWER9)
  1102. FCOMMON_OPT += -tp pwr9
  1103. endif
  1104. ifeq ($(CORE), POWER10)
  1105. $(warning NVIDIA HPC compilers do not support POWER10.)
  1106. endif
  1107. endif
  1108. endif
  1109. else
  1110. FCOMMON_OPT += -tp p7
  1111. endif
  1112. FCOMMON_OPT += -Mrecursive -Kieee
  1113. ifeq ($(USE_OPENMP), 1)
  1114. FCOMMON_OPT += -mp
  1115. endif
  1116. endif
  1117. ifeq ($(F_COMPILER), PATHSCALE)
  1118. CCOMMON_OPT += -DF_INTERFACE_PATHSCALE
  1119. ifdef BINARY64
  1120. ifdef INTERFACE64
  1121. ifneq ($(INTERFACE64), 0)
  1122. FCOMMON_OPT += -i8
  1123. endif
  1124. endif
  1125. endif
  1126. ifeq ($(USE_OPENMP), 1)
  1127. FCOMMON_OPT += -mp
  1128. endif
  1129. endif
  1130. ifeq ($(F_COMPILER), OPEN64)
  1131. CCOMMON_OPT += -DF_INTERFACE_OPEN64
  1132. ifdef BINARY64
  1133. ifdef INTERFACE64
  1134. ifneq ($(INTERFACE64), 0)
  1135. FCOMMON_OPT += -i8
  1136. endif
  1137. endif
  1138. endif
  1139. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  1140. ifndef BINARY64
  1141. FCOMMON_OPT += -n32
  1142. else
  1143. FCOMMON_OPT += -n64
  1144. endif
  1145. ifeq ($(CORE), LOONGSON3R3)
  1146. FCOMMON_OPT += -loongson3 -static
  1147. endif
  1148. ifeq ($(CORE), LOONGSON3R4)
  1149. FCOMMON_OPT += -loongson3 -static
  1150. endif
  1151. else
  1152. ifndef BINARY64
  1153. FCOMMON_OPT += -m32
  1154. else
  1155. FCOMMON_OPT += -m64
  1156. endif
  1157. endif
  1158. ifeq ($(USE_OPENMP), 1)
  1159. FEXTRALIB += -lstdc++
  1160. FCOMMON_OPT += -mp
  1161. endif
  1162. endif
  1163. ifeq ($(C_COMPILER), OPEN64)
  1164. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  1165. ifndef BINARY64
  1166. CCOMMON_OPT += -n32
  1167. else
  1168. CCOMMON_OPT += -n64
  1169. endif
  1170. ifeq ($(CORE), LOONGSON3R3)
  1171. CCOMMON_OPT += -loongson3 -static
  1172. endif
  1173. ifeq ($(CORE), LOONGSON3R4)
  1174. CCOMMON_OPT += -loongson3 -static
  1175. endif
  1176. else
  1177. ifndef BINARY64
  1178. CCOMMON_OPT += -m32
  1179. else
  1180. CCOMMON_OPT += -m64
  1181. endif
  1182. endif
  1183. endif
  1184. ifeq ($(C_COMPILER), SUN)
  1185. CCOMMON_OPT += -w
  1186. ifeq ($(ARCH), x86)
  1187. CCOMMON_OPT += -m32
  1188. else
  1189. ifdef BINARY64
  1190. CCOMMON_OPT += -m64
  1191. else
  1192. CCOMMON_OPT += -m32
  1193. endif
  1194. endif
  1195. endif
  1196. ifeq ($(F_COMPILER), SUN)
  1197. CCOMMON_OPT += -DF_INTERFACE_SUN
  1198. FCOMMON_OPT += -ftrap=%none -xrecursive
  1199. ifeq ($(ARCH), x86)
  1200. FCOMMON_OPT += -m32
  1201. else
  1202. ifdef BINARY64
  1203. FCOMMON_OPT += -m64
  1204. else
  1205. FCOMMON_OPT += -m32
  1206. endif
  1207. endif
  1208. ifeq ($(USE_OPENMP), 1)
  1209. FCOMMON_OPT += -xopenmp=parallel
  1210. endif
  1211. endif
  1212. ifeq ($(F_COMPILER), COMPAQ)
  1213. CCOMMON_OPT += -DF_INTERFACE_COMPAQ
  1214. ifeq ($(USE_OPENMP), 1)
  1215. FCOMMON_OPT += -openmp
  1216. endif
  1217. endif
  1218. ifeq ($(F_COMPILER), CRAY)
  1219. CCOMMON_OPT += -DF_INTERFACE_INTEL
  1220. FCOMMON_OPT += -hnopattern
  1221. ifdef INTERFACE64
  1222. ifneq ($(INTERFACE64), 0)
  1223. FCOMMON_OPT += -s integer64
  1224. endif
  1225. endif
  1226. ifneq ($(USE_OPENMP), 1)
  1227. FCOMMON_OPT += -O noomp
  1228. endif
  1229. endif
  1230. ifdef BINARY64
  1231. ifdef INTERFACE64
  1232. ifneq ($(INTERFACE64), 0)
  1233. CCOMMON_OPT +=
  1234. #-DUSE64BITINT
  1235. endif
  1236. endif
  1237. endif
  1238. ifeq ($(NEED_PIC), 1)
  1239. ifeq ($(C_COMPILER), IBM)
  1240. CCOMMON_OPT += -qpic=large
  1241. else
  1242. CCOMMON_OPT += -fPIC
  1243. endif
  1244. ifeq ($(F_COMPILER), SUN)
  1245. FCOMMON_OPT += -pic
  1246. else ifeq ($(F_COMPILER), NAG)
  1247. FCOMMON_OPT += -PIC
  1248. else ifeq ($(F_COMPILER), IBM)
  1249. FCOMMON_OPT += -qpic=large
  1250. else
  1251. FCOMMON_OPT += -fPIC
  1252. endif
  1253. endif
  1254. ifeq ($(DYNAMIC_ARCH), 1)
  1255. CCOMMON_OPT += -DDYNAMIC_ARCH
  1256. endif
  1257. ifeq ($(DYNAMIC_OLDER), 1)
  1258. CCOMMON_OPT += -DDYNAMIC_OLDER
  1259. endif
  1260. ifeq ($(C_LAPACK), 1)
  1261. CCOMMON_OPT += -DC_LAPACK
  1262. endif
  1263. ifeq ($(NO_LAPACK), 1)
  1264. CCOMMON_OPT += -DNO_LAPACK
  1265. #Disable LAPACK C interface
  1266. NO_LAPACKE = 1
  1267. endif
  1268. ifeq ($(NO_LAPACKE), 1)
  1269. CCOMMON_OPT += -DNO_LAPACKE
  1270. endif
  1271. ifeq ($(NO_AVX), 1)
  1272. CCOMMON_OPT += -DNO_AVX
  1273. endif
  1274. ifeq ($(ARCH), x86)
  1275. CCOMMON_OPT += -DNO_AVX
  1276. endif
  1277. ifeq ($(NO_AVX2), 1)
  1278. CCOMMON_OPT += -DNO_AVX2
  1279. endif
  1280. ifeq ($(NO_AVX512), 1)
  1281. CCOMMON_OPT += -DNO_AVX512
  1282. endif
  1283. ifeq ($(NO_SVE), 1)
  1284. CCOMMON_OPT += -DNO_SVE
  1285. endif
  1286. ifdef SMP
  1287. CCOMMON_OPT += -DSMP_SERVER
  1288. ifeq ($(ARCH), mips64)
  1289. USE_SIMPLE_THREADED_LEVEL3 = 1
  1290. endif
  1291. ifeq ($(USE_OPENMP), 1)
  1292. # USE_SIMPLE_THREADED_LEVEL3 = 1
  1293. # NO_AFFINITY = 1
  1294. CCOMMON_OPT += -DUSE_OPENMP
  1295. endif
  1296. ifeq ($(BIGNUMA), 1)
  1297. CCOMMON_OPT += -DBIGNUMA
  1298. endif
  1299. endif
  1300. ifeq ($(NO_WARMUP), 1)
  1301. CCOMMON_OPT += -DNO_WARMUP
  1302. endif
  1303. ifeq ($(CONSISTENT_FPCSR), 1)
  1304. CCOMMON_OPT += -DCONSISTENT_FPCSR
  1305. endif
  1306. # Only for development
  1307. # CCOMMON_OPT += -DPARAMTEST
  1308. # CCOMMON_OPT += -DPREFETCHTEST
  1309. # CCOMMON_OPT += -DNO_SWITCHING
  1310. # USE_PAPI = 1
  1311. ifdef USE_PAPI
  1312. CCOMMON_OPT += -DUSE_PAPI
  1313. EXTRALIB += -lpapi -lperfctr
  1314. endif
  1315. ifdef BUFFERSIZE
  1316. CCOMMON_OPT += -DBUFFERSIZE=$(BUFFERSIZE)
  1317. endif
  1318. ifdef DYNAMIC_THREADS
  1319. CCOMMON_OPT += -DDYNAMIC_THREADS
  1320. endif
  1321. CCOMMON_OPT += -DMAX_CPU_NUMBER=$(NUM_THREADS)
  1322. CCOMMON_OPT += -DMAX_PARALLEL_NUMBER=$(NUM_PARALLEL)
  1323. ifdef USE_SIMPLE_THREADED_LEVEL3
  1324. CCOMMON_OPT += -DUSE_SIMPLE_THREADED_LEVEL3
  1325. endif
  1326. ifeq ($(USE_TLS), 1)
  1327. CCOMMON_OPT += -DUSE_TLS
  1328. endif
  1329. ifeq ($(BUILD_BFLOAT16), 1)
  1330. CCOMMON_OPT += -DBUILD_BFLOAT16
  1331. endif
  1332. ifeq ($(BUILD_SINGLE), 1)
  1333. CCOMMON_OPT += -DBUILD_SINGLE=1
  1334. endif
  1335. ifeq ($(BUILD_DOUBLE), 1)
  1336. CCOMMON_OPT += -DBUILD_DOUBLE=1
  1337. endif
  1338. ifeq ($(BUILD_COMPLEX), 1)
  1339. CCOMMON_OPT += -DBUILD_COMPLEX=1
  1340. endif
  1341. ifeq ($(BUILD_COMPLEX16), 1)
  1342. CCOMMON_OPT += -DBUILD_COMPLEX16=1
  1343. endif
  1344. CCOMMON_OPT += -DVERSION=\"$(VERSION)\"
  1345. ifndef SYMBOLPREFIX
  1346. SYMBOLPREFIX =
  1347. endif
  1348. ifndef SYMBOLSUFFIX
  1349. SYMBOLSUFFIX =
  1350. endif
  1351. ifndef LIBSONAMEBASE
  1352. LIBSONAMEBASE = openblas
  1353. endif
  1354. ifndef LIBNAMEPREFIX
  1355. LIBNAMEPREFIX =
  1356. endif
  1357. SYMPREFIX=$(SYMBOLPREFIX)
  1358. ifeq ($(SYMBOLPREFIX),$(LIBNAMEPREFIX))
  1359. SYMPREFIX=
  1360. endif
  1361. SYMSUFFIX=$(SYMBOLSUFFIX)
  1362. ifeq ($(SYMBOLSUFFIX),$(LIBNAMESUFFIX))
  1363. SYMSUFFIX=
  1364. endif
  1365. ifndef LIBNAMESUFFIX
  1366. LIBNAMEBASE = $(SYMPREFIX)$(LIBSONAMEBASE)$(SYMSUFFIX)
  1367. else
  1368. LIBNAMEBASE = $(SYMPREFIX)$(LIBSONAMEBASE)$(SYMSUFFIX)$(LIBNAMESUFFIX)
  1369. endif
  1370. ifeq ($(OSNAME), CYGWIN_NT)
  1371. LIBPREFIX = cyg$(LIBNAMEPREFIX)$(LIBNAMEBASE)
  1372. else
  1373. LIBPREFIX = lib$(LIBNAMEPREFIX)$(LIBNAMEBASE)
  1374. endif
  1375. KERNELDIR = $(TOPDIR)/kernel/$(ARCH)
  1376. include $(TOPDIR)/Makefile.$(ARCH)
  1377. ifneq ($(C_COMPILER), PGI)
  1378. ifneq ($(C_COMPILER), SUN)
  1379. CCOMMON_OPT += -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME
  1380. endif
  1381. endif
  1382. CCOMMON_OPT += -DASMNAME=$(FU)$(*F) -DASMFNAME=$(FU)$(*F)$(BU) -DNAME=$(*F)$(BU) -DCNAME=$(*F) -DCHAR_NAME=\"$(*F)$(BU)\" -DCHAR_CNAME=\"$(*F)\"
  1383. ifeq ($(CORE), PPC440)
  1384. CCOMMON_OPT += -DALLOC_QALLOC
  1385. endif
  1386. ifeq ($(CORE), PPC440FP2)
  1387. STATIC_ALLOCATION = 1
  1388. endif
  1389. ifneq ($(OSNAME), Linux)
  1390. NO_AFFINITY = 1
  1391. endif
  1392. ifneq ($(ARCH), x86_64)
  1393. ifneq ($(ARCH), x86)
  1394. NO_AFFINITY = 1
  1395. endif
  1396. endif
  1397. ifdef NO_AFFINITY
  1398. ifeq ($(NO_AFFINITY), 0)
  1399. override undefine NO_AFFINITY
  1400. else
  1401. CCOMMON_OPT += -DNO_AFFINITY
  1402. endif
  1403. endif
  1404. ifdef FUNCTION_PROFILE
  1405. CCOMMON_OPT += -DFUNCTION_PROFILE
  1406. endif
  1407. ifdef SHMEM_ALLOCATION
  1408. ifneq ($(SHMEM_ALLOCATION), 0)
  1409. CCOMMON_OPT += -DALLOC_SHM
  1410. endif
  1411. endif
  1412. ifdef HUGETLB_ALLOCATION
  1413. ifneq ($(HUGETLB_ALLOCATION), 0)
  1414. CCOMMON_OPT += -DALLOC_HUGETLB
  1415. endif
  1416. endif
  1417. ifdef HUGETLBFILE_ALLOCATION
  1418. ifneq ($(HUGETLBFILE_ALLOCATION), 0)
  1419. CCOMMON_OPT += -DALLOC_HUGETLBFILE -DHUGETLB_FILE_NAME=$(HUGETLBFILE_ALLOCATION)
  1420. endif
  1421. endif
  1422. ifdef STATIC_ALLOCATION
  1423. CCOMMON_OPT += -DALLOC_STATIC
  1424. endif
  1425. ifdef DEVICEDRIVER_ALLOCATION
  1426. CCOMMON_OPT += -DALLOC_DEVICEDRIVER -DDEVICEDRIVER_NAME=\"/dev/mapper\"
  1427. endif
  1428. ifdef MIXED_MEMORY_ALLOCATION
  1429. CCOMMON_OPT += -DMIXED_MEMORY_ALLOCATION
  1430. endif
  1431. ifeq ($(OSNAME), SunOS)
  1432. TAR = gtar
  1433. PATCH = gpatch
  1434. GREP = ggrep
  1435. AWK = nawk
  1436. else
  1437. TAR = tar
  1438. PATCH = patch
  1439. GREP = grep
  1440. AWK = awk
  1441. endif
  1442. ifndef MD5SUM
  1443. MD5SUM = md5sum
  1444. endif
  1445. REVISION = -r$(VERSION)
  1446. MAJOR_VERSION = $(word 1,$(subst ., ,$(VERSION)))
  1447. ifeq ($(DEBUG), 1)
  1448. COMMON_OPT += -g
  1449. endif
  1450. ifeq ($(DEBUG), 1)
  1451. FCOMMON_OPT += -g
  1452. endif
  1453. ifndef COMMON_OPT
  1454. COMMON_OPT = -O2
  1455. endif
  1456. ifndef FCOMMON_OPT
  1457. FCOMMON_OPT = -O2 -frecursive
  1458. endif
  1459. override CFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
  1460. override PFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
  1461. override FFLAGS += $(COMMON_OPT) $(FCOMMON_OPT)
  1462. override FPFLAGS += $(FCOMMON_OPT) $(COMMON_PROF)
  1463. #MAKEOVERRIDES =
  1464. ifeq ($(NEED_PIC), 1)
  1465. ifeq (,$(findstring PIC,$(FFLAGS)))
  1466. ifneq ($(F_COMPILER),IBM)
  1467. override FFLAGS += -fPIC
  1468. endif
  1469. endif
  1470. endif
  1471. #For LAPACK Fortran codes.
  1472. #Disable -fopenmp for LAPACK Fortran codes on Windows.
  1473. ifdef OS_WINDOWS
  1474. LAPACK_FFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FFLAGS))
  1475. LAPACK_FPFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FPFLAGS))
  1476. else
  1477. LAPACK_FFLAGS := $(FFLAGS)
  1478. LAPACK_FPFLAGS := $(FPFLAGS)
  1479. endif
  1480. ifeq ($(F_COMPILER),NAG)
  1481. LAPACK_FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1482. override FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1483. endif
  1484. ifeq ($(F_COMPILER),CRAY)
  1485. LAPACK_FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1486. override FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1487. endif
  1488. ifeq ($(F_COMPILER),FLANGNEW)
  1489. LAPACK_FFLAGS := $(filter-out -m32 -m64 -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1490. override FFLAGS := $(filter-out -m32 -m64 -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1491. endif
  1492. LAPACK_CFLAGS = $(CFLAGS)
  1493. LAPACK_CFLAGS += -DHAVE_LAPACK_CONFIG_H
  1494. ifdef INTERFACE64
  1495. ifneq ($(INTERFACE64), 0)
  1496. LAPACK_CFLAGS += -DLAPACK_ILP64
  1497. endif
  1498. endif
  1499. ifdef OS_WINDOWS
  1500. LAPACK_CFLAGS += -DOPENBLAS_OS_WINDOWS
  1501. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1502. endif
  1503. ifeq ($(C_COMPILER), LSB)
  1504. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1505. endif
  1506. ifndef SUFFIX
  1507. SUFFIX = o
  1508. endif
  1509. ifndef PSUFFIX
  1510. PSUFFIX = po
  1511. endif
  1512. ifndef LIBSUFFIX
  1513. LIBSUFFIX = a
  1514. endif
  1515. ifneq ($(DYNAMIC_ARCH), 1)
  1516. ifndef SMP
  1517. LIBNAME = $(LIBPREFIX)_$(LIBCORE)$(REVISION).$(LIBSUFFIX)
  1518. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)$(REVISION)_p.$(LIBSUFFIX)
  1519. else
  1520. LIBNAME = $(LIBPREFIX)_$(LIBCORE)p$(REVISION).$(LIBSUFFIX)
  1521. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)p$(REVISION)_p.$(LIBSUFFIX)
  1522. endif
  1523. else
  1524. ifndef SMP
  1525. LIBNAME = $(LIBPREFIX)$(REVISION).$(LIBSUFFIX)
  1526. LIBNAME_P = $(LIBPREFIX)$(REVISION)_p.$(LIBSUFFIX)
  1527. else
  1528. LIBNAME = $(LIBPREFIX)p$(REVISION).$(LIBSUFFIX)
  1529. LIBNAME_P = $(LIBPREFIX)p$(REVISION)_p.$(LIBSUFFIX)
  1530. endif
  1531. endif
  1532. ifeq ($(FIXED_LIBNAME),1)
  1533. LIBNAME = lib$(LIBNAMEPREFIX)$(LIBSONAMEBASE)$(LIBNAMESUFFIX).$(LIBSUFFIX)
  1534. LIBNAME_P = lib$(LIBNAMEPREFIX)$(LISOBNAMEBASE)$(LIBNAMESUFFIX)_p.$(LIBSUFFIX)
  1535. endif
  1536. LIBDLLNAME = $(LIBPREFIX).dll
  1537. IMPLIBNAME = lib$(LIBNAMEBASE).dll.a
  1538. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.so)
  1539. LIBDYNNAME = $(LIBNAME:.$(LIBSUFFIX)=.dylib)
  1540. LIBDEFNAME = $(LIBNAME:.$(LIBSUFFIX)=.def)
  1541. LIBEXPNAME = $(LIBNAME:.$(LIBSUFFIX)=.exp)
  1542. LIBZIPNAME = $(LIBNAME:.$(LIBSUFFIX)=.zip)
  1543. LIBS = $(TOPDIR)/$(LIBNAME)
  1544. LIBS_P = $(TOPDIR)/$(LIBNAME_P)
  1545. LIB_COMPONENTS = BLAS
  1546. ifneq ($(NO_CBLAS), 1)
  1547. LIB_COMPONENTS += CBLAS
  1548. endif
  1549. ifneq ($(NO_LAPACK), 1)
  1550. LIB_COMPONENTS += LAPACK
  1551. ifneq ($(NO_LAPACKE), 1)
  1552. LIB_COMPONENTS += LAPACKE
  1553. endif
  1554. ifeq ($(BUILD_RELAPACK), 1)
  1555. LIB_COMPONENTS += ReLAPACK
  1556. endif
  1557. endif
  1558. ifeq ($(ONLY_CBLAS), 1)
  1559. LIB_COMPONENTS = CBLAS
  1560. endif
  1561. export OSNAME
  1562. export ARCH
  1563. export CORE
  1564. export LIBCORE
  1565. export __BYTE_ORDER__
  1566. export ELF_VERSION
  1567. export PGCPATH
  1568. export CONFIG
  1569. export CC
  1570. export FC
  1571. export BU
  1572. export FU
  1573. export NEED2UNDERSCORES
  1574. export USE_THREAD
  1575. export NUM_THREADS
  1576. export NUM_CORES
  1577. export SMP
  1578. export MAKEFILE_RULE
  1579. export NEED_PIC
  1580. export BINARY
  1581. export BINARY32
  1582. export BINARY64
  1583. export F_COMPILER
  1584. export C_COMPILER
  1585. export USE_OPENMP
  1586. export CROSS
  1587. export CROSS_SUFFIX
  1588. export NOFORTRAN
  1589. export C_LAPACK
  1590. export NO_FBLAS
  1591. export EXTRALIB
  1592. export CEXTRALIB
  1593. export FEXTRALIB
  1594. export HAVE_SSE
  1595. export HAVE_SSE2
  1596. export HAVE_SSE3
  1597. export HAVE_SSSE3
  1598. export HAVE_SSE4_1
  1599. export HAVE_SSE4_2
  1600. export HAVE_SSE4A
  1601. export HAVE_SSE5
  1602. export HAVE_AVX
  1603. export HAVE_AVX2
  1604. export HAVE_FMA3
  1605. export HAVE_VFP
  1606. export HAVE_VFPV3
  1607. export HAVE_VFPV4
  1608. export HAVE_NEON
  1609. ifndef NO_MSA
  1610. export HAVE_MSA
  1611. export MSA_FLAGS
  1612. endif
  1613. export KERNELDIR
  1614. export FUNCTION_PROFILE
  1615. export TARGET_CORE
  1616. export NO_AVX512
  1617. export NO_AVX2
  1618. export BUILD_BFLOAT16
  1619. export NO_LSX
  1620. export NO_LASX
  1621. export SBGEMM_UNROLL_M
  1622. export SBGEMM_UNROLL_N
  1623. export SGEMM_UNROLL_M
  1624. export SGEMM_UNROLL_N
  1625. export DGEMM_UNROLL_M
  1626. export DGEMM_UNROLL_N
  1627. export QGEMM_UNROLL_M
  1628. export QGEMM_UNROLL_N
  1629. export CGEMM_UNROLL_M
  1630. export CGEMM_UNROLL_N
  1631. export ZGEMM_UNROLL_M
  1632. export ZGEMM_UNROLL_N
  1633. export XGEMM_UNROLL_M
  1634. export XGEMM_UNROLL_N
  1635. export CGEMM3M_UNROLL_M
  1636. export CGEMM3M_UNROLL_N
  1637. export ZGEMM3M_UNROLL_M
  1638. export ZGEMM3M_UNROLL_N
  1639. export XGEMM3M_UNROLL_M
  1640. export XGEMM3M_UNROLL_N
  1641. ifdef USE_CUDA
  1642. export CUDADIR
  1643. export CUCC
  1644. export CUFLAGS
  1645. export CULIB
  1646. endif
  1647. .SUFFIXES: .$(PSUFFIX) .$(SUFFIX) .f
  1648. .f.$(SUFFIX):
  1649. $(FC) $(FFLAGS) -c $< -o $(@F)
  1650. .f.$(PSUFFIX):
  1651. $(FC) $(FPFLAGS) -pg -c $< -o $(@F)
  1652. ifdef BINARY64
  1653. PATHSCALEPATH = /opt/pathscale/lib/3.1
  1654. PGIPATH = /opt/pgi/linux86-64/7.1-5/lib
  1655. else
  1656. PATHSCALEPATH = /opt/pathscale/lib/3.1/32
  1657. PGIPATH = /opt/pgi/linux86/7.1-5/lib
  1658. endif
  1659. ACMLPATH = /opt/acml/4.3.0
  1660. ifneq ($(OSNAME), Darwin)
  1661. MKLPATH = /opt/intel/mkl/10.2.2.025/lib
  1662. else
  1663. MKLPATH = /Library/Frameworks/Intel_MKL.framework/Versions/10.0.1.014/lib
  1664. endif
  1665. ATLASPATH = /opt/atlas/3.9.17/opteron
  1666. FLAMEPATH = $(HOME)/flame/lib
  1667. ifneq ($(OSNAME), SunOS)
  1668. SUNPATH = /opt/sunstudio12.1
  1669. else
  1670. SUNPATH = /opt/SUNWspro
  1671. endif