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 41 kB

6 years ago
1 year ago
4 years ago
13 years ago
4 years ago
12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448144914501451145214531454145514561457145814591460146114621463146414651466146714681469147014711472147314741475147614771478147914801481148214831484148514861487148814891490149114921493149414951496149714981499150015011502150315041505150615071508150915101511151215131514151515161517151815191520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559156015611562156315641565156615671568156915701571157215731574157515761577157815791580158115821583158415851586158715881589159015911592159315941595159615971598159916001601160216031604160516061607160816091610161116121613161416151616161716181619162016211622162316241625162616271628162916301631163216331634163516361637163816391640164116421643164416451646164716481649165016511652165316541655165616571658165916601661166216631664166516661667166816691670167116721673167416751676167716781679168016811682168316841685168616871688168916901691169216931694169516961697169816991700170117021703170417051706170717081709171017111712171317141715171617171718171917201721172217231724172517261727172817291730173117321733173417351736173717381739174017411742174317441745174617471748174917501751175217531754175517561757175817591760176117621763176417651766176717681769177017711772177317741775177617771778177917801781178217831784178517861787178817891790179117921793179417951796179717981799180018011802180318041805180618071808180918101811181218131814181518161817181818191820182118221823182418251826182718281829183018311832183318341835183618371838183918401841184218431844184518461847184818491850185118521853185418551856185718581859186018611862186318641865186618671868186918701871187218731874187518761877187818791880188118821883188418851886188718881889189018911892189318941895189618971898189919001901190219031904190519061907190819091910191119121913191419151916191719181919192019211922192319241925192619271928192919301931193219331934193519361937193819391940194119421943194419451946194719481949195019511952195319541955195619571958195919601961196219631964196519661967196819691970197119721973
  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 ($(ARCH), arm64)
  238. SMALL_MATRIX_OPT = 1
  239. BUILD_BFLOAT16 = 1
  240. endif
  241. ifeq ($(ARCH), loongarch64)
  242. SMALL_MATRIX_OPT = 1
  243. endif
  244. ifeq ($(ARCH), arm64)
  245. GEMM_GEMV_FORWARD = 1
  246. SBGEMM_GEMV_FORWARD = 1
  247. BGEMM_GEMV_FORWARD = 1
  248. endif
  249. ifeq ($(ARCH), riscv)
  250. GEMM_GEMV_FORWARD = 1
  251. endif
  252. ifeq ($(ARCH), power)
  253. GEMM_GEMV_FORWARD = 1
  254. SBGEMM_GEMV_FORWARD = 1
  255. endif
  256. ifeq ($(SMALL_MATRIX_OPT), 1)
  257. CCOMMON_OPT += -DSMALL_MATRIX_OPT
  258. endif
  259. ifneq ($(ONLY_CBLAS), 1)
  260. ifeq ($(GEMM_GEMV_FORWARD), 1)
  261. CCOMMON_OPT += -DGEMM_GEMV_FORWARD
  262. endif
  263. ifeq ($(SBGEMM_GEMV_FORWARD), 1)
  264. CCOMMON_OPT += -DSBGEMM_GEMV_FORWARD
  265. endif
  266. ifeq ($(BGEMM_GEMV_FORWARD), 1)
  267. CCOMMON_OPT += -DBGEMM_GEMV_FORWARD
  268. endif
  269. endif
  270. # This operation is expensive, so execution should be once.
  271. ifndef GOTOBLAS_MAKEFILE
  272. export GOTOBLAS_MAKEFILE = 1
  273. # Generating Makefile.conf and config.h
  274. 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)
  275. endif
  276. ifndef TARGET_CORE
  277. -include $(TOPDIR)/Makefile.conf
  278. else
  279. HAVE_NEON=
  280. HAVE_VFP=
  281. HAVE_VFPV3=
  282. HAVE_VFPV4=
  283. HAVE_MMX=
  284. HAVE_SSE=
  285. HAVE_SSE2=
  286. HAVE_SSE3=
  287. HAVE_SSSE3=
  288. HAVE_SSE4_1=
  289. HAVE_SSE4_2=
  290. HAVE_SSE4A=
  291. HAVE_SSE5=
  292. HAVE_AVX=
  293. HAVE_AVX2=
  294. HAVE_FMA3=
  295. include $(TOPDIR)/Makefile_kernel.conf
  296. endif
  297. ifndef NUM_PARALLEL
  298. NUM_PARALLEL = 1
  299. endif
  300. ifndef NUM_THREADS
  301. NUM_THREADS = $(NUM_CORES)
  302. endif
  303. ifeq ($(NUM_THREADS), 1)
  304. override USE_THREAD = 0
  305. override USE_OPENMP = 0
  306. endif
  307. ifdef USE_THREAD
  308. ifeq ($(USE_THREAD), 0)
  309. SMP =
  310. else
  311. SMP = 1
  312. endif
  313. else
  314. ifeq ($(NUM_THREADS), 1)
  315. SMP =
  316. else
  317. SMP = 1
  318. endif
  319. endif
  320. ifeq ($(SMP), 1)
  321. USE_LOCKING =
  322. endif
  323. ifndef NEED_PIC
  324. NEED_PIC = 1
  325. endif
  326. ARFLAGS =
  327. CPP = $(COMPILER) -E
  328. AR ?= $(CROSS_SUFFIX)ar
  329. AS ?= $(CROSS_SUFFIX)as
  330. LD ?= $(CROSS_SUFFIX)ld
  331. RANLIB ?= $(CROSS_SUFFIX)ranlib
  332. NM = $(CROSS_SUFFIX)nm
  333. DLLWRAP = $(CROSS_SUFFIX)dllwrap
  334. OBJCOPY = $(CROSS_SUFFIX)objcopy
  335. OBJCONV = $(CROSS_SUFFIX)objconv
  336. # When fortran support was either not detected or actively deselected, only build BLAS.
  337. ifeq ($(NOFORTRAN), 1)
  338. C_LAPACK = 1
  339. override FEXTRALIB =
  340. endif
  341. ifeq ($(C_COMPILER), GCC)
  342. GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
  343. GCCVERSIONGT4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 4)
  344. GCCVERSIONGT5 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 5)
  345. GCCVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 7)
  346. GCCVERSIONGTEQ8 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 8)
  347. GCCVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  348. GCCVERSIONGTEQ10 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 10)
  349. GCCVERSIONGTEQ11 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 11)
  350. GCCVERSIONGTEQ12 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 12)
  351. GCCVERSIONGTEQ13 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 13)
  352. GCCVERSIONGTEQ14 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 14)
  353. # Note that the behavior of -dumpversion is compile-time-configurable for
  354. # gcc-7.x and newer. Use -dumpfullversion there
  355. ifeq ($(GCCVERSIONGTEQ7),1)
  356. GCCDUMPVERSION_PARAM := -dumpfullversion
  357. else
  358. GCCDUMPVERSION_PARAM := -dumpversion
  359. endif
  360. GCCMINORVERSIONGTEQ1 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 1)
  361. GCCMINORVERSIONGTEQ2 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 2)
  362. GCCMINORVERSIONGTEQ4 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 4)
  363. GCCMINORVERSIONGTEQ7 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 7)
  364. endif
  365. ifeq ($(C_COMPILER), CLANG)
  366. CLANGVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  367. CLANGVERSIONGTEQ12 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 12)
  368. endif
  369. #
  370. # OS dependent settings
  371. #
  372. ifeq ($(OSNAME), Darwin)
  373. ifndef MACOSX_DEPLOYMENT_TARGET
  374. ifeq ($(ARCH), arm64)
  375. export MACOSX_DEPLOYMENT_TARGET=11.0
  376. export NO_SVE = 1
  377. export NO_SME = 1
  378. else
  379. export MACOSX_DEPLOYMENT_TARGET=10.8
  380. endif
  381. endif
  382. MD5SUM = md5 -r
  383. XCVER = $(shell pkgutil --pkg-info=com.apple.pkg.Xcode |awk '/version:/ {print $2}'|cut -d: -f2|cut -f1 -d.)
  384. ifeq (x$(XCVER)x,xx)
  385. XCVER = $(shell pkgutil --pkg-info=com.apple.pkg.CLTools_Executables |awk '/version:/ {print $2}'|cut -d: -f2|cut -f1 -d.)
  386. endif
  387. ifeq (x$(XCVER), x 15)
  388. CCOMMON_OPT += -Wl,-ld_classic
  389. FCOMMON_OPT += -Wl,-ld_classic
  390. endif
  391. ifeq (x$(XCVER), x 16)
  392. ifeq ($(F_COMPILER), GFORTRAN)
  393. override CEXTRALIB := $(filter-out(-lto_library, $(CEXTRALIB)))
  394. endif
  395. endif
  396. endif
  397. ifneq (,$(findstring $(OSNAME), FreeBSD OpenBSD DragonFly))
  398. MD5SUM = md5 -r
  399. endif
  400. ifeq ($(OSNAME), NetBSD)
  401. MD5SUM = md5 -n
  402. endif
  403. ifeq ($(OSNAME), Linux)
  404. EXTRALIB += -lm
  405. #NO_EXPRECISION = 1
  406. endif
  407. ifeq ($(OSNAME), Android)
  408. EXTRALIB += -lm
  409. endif
  410. ifeq ($(OSNAME), AIX)
  411. EXTRALIB += -lm
  412. endif
  413. ifeq ($(OSNAME), $(filter $(OSNAME),FreeBSD OpenBSD NetBSD DragonFly))
  414. ifeq ($(ARCH), $(filter $(ARCH),arm arm64))
  415. EXTRALIB += -lm
  416. endif
  417. endif
  418. ifeq ($(OSNAME), WINNT)
  419. NEED_PIC = 0
  420. NO_EXPRECISION = 1
  421. EXTRALIB += -defaultlib:advapi32
  422. SUFFIX = obj
  423. PSUFFIX = pobj
  424. LIBSUFFIX = a
  425. ifeq ($(C_COMPILER), CLANG)
  426. CCOMMON_OPT += -DMS_ABI
  427. endif
  428. #Version tests for supporting specific features (MS_ABI, POWER9 intrinsics)
  429. ifeq ($(GCCVERSIONGT4), 1)
  430. # GCC Major version > 4
  431. # It is compatible with MSVC ABI.
  432. CCOMMON_OPT += -DMS_ABI
  433. endif
  434. ifeq ($(GCCVERSIONGTEQ4), 1)
  435. ifeq ($(GCCMINORVERSIONGTEQ7), 1)
  436. # GCC Version >=4.7
  437. # It is compatible with MSVC ABI.
  438. CCOMMON_OPT += -DMS_ABI
  439. endif
  440. endif
  441. # Ensure the correct stack alignment on Win32
  442. # http://permalink.gmane.org/gmane.comp.lib.openblas.general/97
  443. ifeq ($(ARCH), x86)
  444. CCOMMON_OPT += -mincoming-stack-boundary=2
  445. FCOMMON_OPT += -mincoming-stack-boundary=2
  446. endif
  447. endif
  448. ifeq ($(OSNAME), Interix)
  449. NEED_PIC = 0
  450. NO_EXPRECISION = 1
  451. INTERIX_TOOL_DIR = /opt/gcc.3.3/i586-pc-interix3/bin
  452. endif
  453. ifeq ($(OSNAME), CYGWIN_NT)
  454. NEED_PIC = 0
  455. NO_EXPRECISION = 1
  456. OS_CYGWIN_NT = 1
  457. endif
  458. ifneq ($(OSNAME), WINNT)
  459. ifneq ($(OSNAME), CYGWIN_NT)
  460. ifneq ($(OSNAME), Interix)
  461. ifneq ($(OSNAME), Android)
  462. ifdef SMP
  463. EXTRALIB += -lpthread
  464. endif
  465. endif
  466. endif
  467. endif
  468. endif
  469. # ifeq logical or
  470. ifeq ($(OSNAME), $(filter $(OSNAME),WINNT CYGWIN_NT Interix))
  471. OS_WINDOWS=1
  472. endif
  473. ifdef QUAD_PRECISION
  474. CCOMMON_OPT += -DQUAD_PRECISION
  475. NO_EXPRECISION = 1
  476. endif
  477. ifneq ($(ARCH), x86)
  478. ifneq ($(ARCH), x86_64)
  479. NO_EXPRECISION = 1
  480. endif
  481. endif
  482. ifdef UTEST_CHECK
  483. CCOMMON_OPT += -DUTEST_CHECK
  484. SANITY_CHECK = 1
  485. endif
  486. ifdef SANITY_CHECK
  487. CCOMMON_OPT += -DSANITY_CHECK -DREFNAME=$(*F)f$(BU)
  488. endif
  489. MAX_STACK_ALLOC ?= 2048
  490. ifneq ($(MAX_STACK_ALLOC), 0)
  491. CCOMMON_OPT += -DMAX_STACK_ALLOC=$(MAX_STACK_ALLOC)
  492. endif
  493. ifdef USE_LOCKING
  494. ifneq ($(USE_LOCKING), 0)
  495. CCOMMON_OPT += -DUSE_LOCKING
  496. endif
  497. endif
  498. #
  499. # Architecture dependent settings
  500. #
  501. ifeq ($(ARCH), x86)
  502. ifndef BINARY
  503. NO_BINARY_MODE = 1
  504. endif
  505. ifeq ($(CORE), generic)
  506. #NO_EXPRECISION = 1
  507. endif
  508. ifndef NO_EXPRECISION
  509. ifeq ($(F_COMPILER), GFORTRAN)
  510. # ifeq logical or. GCC or LSB
  511. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  512. EXPRECISION = 1
  513. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  514. FCOMMON_OPT += -m128bit-long-double
  515. endif
  516. ifeq ($(C_COMPILER), CLANG)
  517. EXPRECISION = 1
  518. CCOMMON_OPT += -DEXPRECISION
  519. FCOMMON_OPT += -m128bit-long-double
  520. endif
  521. endif
  522. endif
  523. endif
  524. ifeq ($(ARCH), x86_64)
  525. ifeq ($(CORE), generic)
  526. #NO_EXPRECISION = 1
  527. endif
  528. ifndef NO_EXPRECISION
  529. ifeq ($(F_COMPILER), GFORTRAN)
  530. # ifeq logical or. GCC or LSB
  531. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  532. EXPRECISION = 1
  533. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  534. FCOMMON_OPT += -m128bit-long-double
  535. endif
  536. ifeq ($(C_COMPILER), CLANG)
  537. EXPRECISION = 1
  538. CCOMMON_OPT += -DEXPRECISION
  539. FCOMMON_OPT += -m128bit-long-double
  540. endif
  541. endif
  542. endif
  543. endif
  544. ifeq ($(C_COMPILER), INTEL)
  545. CCOMMON_OPT += -wd981
  546. endif
  547. ifeq ($(USE_OPENMP), 1)
  548. #check
  549. ifeq ($(USE_THREAD), 0)
  550. $(error OpenBLAS: Cannot set both USE_OPENMP=1 and USE_THREAD=0. The USE_THREAD=0 is only for building single thread version.)
  551. endif
  552. # ifeq logical or. GCC or LSB
  553. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  554. CCOMMON_OPT += -fopenmp
  555. endif
  556. ifeq ($(C_COMPILER), CLANG)
  557. CCOMMON_OPT += -fopenmp
  558. ifeq ($(F_COMPILER), GFORTRAN)
  559. FEXTRALIB := $(subst -lgomp,-lomp,$(FEXTRALIB))
  560. endif
  561. endif
  562. ifeq ($(C_COMPILER), INTEL)
  563. CCOMMON_OPT += -fopenmp
  564. endif
  565. ifeq ($(C_COMPILER), PGI)
  566. CCOMMON_OPT += -mp
  567. endif
  568. ifeq ($(C_COMPILER), OPEN64)
  569. CCOMMON_OPT += -mp
  570. CEXTRALIB += -lstdc++
  571. endif
  572. ifeq ($(C_COMPILER), PATHSCALE)
  573. CCOMMON_OPT += -mp
  574. endif
  575. endif
  576. ifeq ($(DYNAMIC_ARCH), 1)
  577. ifeq ($(ARCH), x86)
  578. DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
  579. CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
  580. endif
  581. ifeq ($(ARCH), x86_64)
  582. DYNAMIC_CORE = PRESCOTT CORE2
  583. ifeq ($(DYNAMIC_OLDER), 1)
  584. DYNAMIC_CORE += PENRYN DUNNINGTON
  585. endif
  586. DYNAMIC_CORE += NEHALEM
  587. ifeq ($(DYNAMIC_OLDER), 1)
  588. DYNAMIC_CORE += OPTERON OPTERON_SSE3
  589. endif
  590. DYNAMIC_CORE += BARCELONA
  591. ifeq ($(DYNAMIC_OLDER), 1)
  592. DYNAMIC_CORE += BOBCAT ATOM NANO
  593. endif
  594. ifneq ($(NO_AVX), 1)
  595. DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER STEAMROLLER EXCAVATOR
  596. endif
  597. ifneq ($(NO_AVX2), 1)
  598. DYNAMIC_CORE += HASWELL ZEN
  599. endif
  600. ifneq ($(NO_AVX512), 1)
  601. ifneq ($(NO_AVX2), 1)
  602. DYNAMIC_CORE += SKYLAKEX COOPERLAKE SAPPHIRERAPIDS
  603. endif
  604. endif
  605. endif
  606. ifdef DYNAMIC_LIST
  607. override DYNAMIC_CORE = PRESCOTT $(DYNAMIC_LIST)
  608. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_PRESCOTT
  609. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  610. CCOMMON_OPT += $(XCCOMMON_OPT)
  611. #CCOMMON_OPT += -DDYNAMIC_LIST='$(DYNAMIC_LIST)'
  612. endif
  613. ifeq ($(ARCH), arm64)
  614. DYNAMIC_CORE = ARMV8
  615. DYNAMIC_CORE += CORTEXA53
  616. DYNAMIC_CORE += CORTEXA57
  617. DYNAMIC_CORE += NEOVERSEN1
  618. ifneq ($(NO_SVE), 1)
  619. DYNAMIC_CORE += NEOVERSEV1
  620. DYNAMIC_CORE += NEOVERSEN2
  621. DYNAMIC_CORE += ARMV8SVE
  622. DYNAMIC_CORE += A64FX
  623. endif
  624. ifneq ($(NO_SME), 1)
  625. DYNAMIC_CORE += ARMV9SME
  626. endif
  627. DYNAMIC_CORE += THUNDERX
  628. DYNAMIC_CORE += THUNDERX2T99
  629. DYNAMIC_CORE += TSV110
  630. DYNAMIC_CORE += EMAG8180
  631. DYNAMIC_CORE += THUNDERX3T110
  632. ifdef DYNAMIC_LIST
  633. override DYNAMIC_CORE = ARMV8 $(DYNAMIC_LIST)
  634. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_ARMV8
  635. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  636. endif
  637. endif
  638. ifeq ($(ARCH), mips64)
  639. DYNAMIC_CORE = LOONGSON3R3 LOONGSON3R4 MIPS64_GENERIC
  640. ifdef DYNAMIC_LIST
  641. override DYNAMIC_CORE = MIPS64_GENERIC $(DYNAMIC_LIST)
  642. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_MIPS64_GENERIC
  643. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  644. endif
  645. endif
  646. ifeq ($(ARCH), loongarch64)
  647. DYNAMIC_CORE = LA64_GENERIC LA264 LA464
  648. endif
  649. ifeq ($(ARCH), riscv64)
  650. DYNAMIC_CORE = RISCV64_GENERIC
  651. DYNAMIC_CORE += RISCV64_ZVL128B
  652. DYNAMIC_CORE += RISCV64_ZVL256B
  653. ifdef DYNAMIC_LIST
  654. override DYNAMIC_CORE = RISCV64_GENERIC $(DYNAMIC_LIST)
  655. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_RISCV64_GENERIC
  656. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  657. endif
  658. endif
  659. ifeq ($(ARCH), zarch)
  660. DYNAMIC_CORE = ZARCH_GENERIC
  661. # if the compiler accepts -march=arch11 or -march=z13 and can compile a file
  662. # with z13-specific inline assembly, then we can include support for Z13.
  663. # note: -march=z13 is equivalent to -march=arch11 yet some compiler releases
  664. # only support one or the other.
  665. # note: LLVM version 6.x supported -march=z13 yet could not handle vector
  666. # registers in inline assembly, so the check for supporting the -march flag is
  667. # not enough.
  668. ZARCH_TEST_COMPILE=-c $(TOPDIR)/kernel/zarch/damin_z13.c -I$(TOPDIR) -o /dev/null > /dev/null 2> /dev/null
  669. ZARCH_CC_SUPPORTS_ARCH11=$(shell $(CC) -march=arch11 $(ZARCH_TEST_COMPILE) && echo 1)
  670. ZARCH_CC_SUPPORTS_Z13=$(shell $(CC) -march=z13 $(ZARCH_TEST_COMPILE) && echo 1)
  671. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH11), $(ZARCH_CC_SUPPORTS_Z13)), 1)
  672. DYNAMIC_CORE += Z13
  673. CCOMMON_OPT += -DDYN_Z13
  674. else
  675. $(info OpenBLAS: Not building Z13 kernels because the compiler $(CC) does not support it)
  676. endif
  677. # as above for z13, check for -march=arch12 and z14 support in the compiler.
  678. ZARCH_CC_SUPPORTS_ARCH12=$(shell $(CC) -march=arch12 $(ZARCH_TEST_COMPILE) && echo 1)
  679. ZARCH_CC_SUPPORTS_Z14=$(shell $(CC) -march=z14 $(ZARCH_TEST_COMPILE) && echo 1)
  680. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH12), $(ZARCH_CC_SUPPORTS_Z14)), 1)
  681. DYNAMIC_CORE += Z14
  682. CCOMMON_OPT += -DDYN_Z14
  683. else
  684. $(info OpenBLAS: Not building Z14 kernels because the compiler $(CC) does not support it)
  685. endif
  686. endif # ARCH zarch
  687. ifeq ($(ARCH), power)
  688. ifneq ($(C_COMPILER), PGI)
  689. DYNAMIC_CORE = POWER6
  690. DYNAMIC_CORE += POWER8
  691. ifneq ($(C_COMPILER), GCC)
  692. DYNAMIC_CORE += POWER9
  693. DYNAMIC_CORE += POWER10
  694. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  695. endif
  696. ifeq ($(C_COMPILER), GCC)
  697. ifeq ($(GCCVERSIONGT5), 1)
  698. DYNAMIC_CORE += POWER9
  699. else
  700. $(info, OpenBLAS: Your gcc version is too old to build the POWER9 kernels.)
  701. endif
  702. ifeq ($(OSNAME), AIX)
  703. LDVERSIONGTEQ35 := 1
  704. else
  705. LDVERSIONGTEQ35 := $(shell expr `$(CC) -Wl,--version 2> /dev/null | head -1 | cut -f2 -d "." | cut -f1 -d "-"` \>= 35)
  706. endif
  707. ifeq ($(GCCVERSIONGTEQ11)$(LDVERSIONGTEQ35), 11)
  708. DYNAMIC_CORE += POWER10
  709. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  710. else ifeq ($(GCCVERSIONGTEQ10), 1)
  711. ifeq ($(GCCMINORVERSIONGTEQ2)$(LDVERSIONGTEQ35), 11)
  712. DYNAMIC_CORE += POWER10
  713. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  714. endif
  715. else
  716. $(info, OpenBLAS: Your gcc version is too old to build the POWER10 kernels.)
  717. endif
  718. endif
  719. else
  720. DYNAMIC_CORE = POWER8
  721. DYNAMIC_CORE += POWER9
  722. endif
  723. endif
  724. # If DYNAMIC_CORE is not set, DYNAMIC_ARCH cannot do anything, so force it to empty
  725. ifndef DYNAMIC_CORE
  726. override DYNAMIC_ARCH=
  727. endif
  728. endif
  729. ifeq ($(ARCH), ia64)
  730. NO_BINARY_MODE = 1
  731. BINARY_DEFINED = 1
  732. ifeq ($(F_COMPILER), GFORTRAN)
  733. ifeq ($(C_COMPILER), GCC)
  734. EXPRECISION = 1
  735. CCOMMON_OPT += -DEXPRECISION
  736. endif
  737. endif
  738. endif
  739. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  740. NO_BINARY_MODE = 1
  741. endif
  742. ifeq ($(ARCH), alpha)
  743. NO_BINARY_MODE = 1
  744. BINARY_DEFINED = 1
  745. endif
  746. ifeq ($(ARCH), arm)
  747. NO_BINARY_MODE = 1
  748. BINARY_DEFINED = 1
  749. ifneq ($(EMBEDDED), 1)
  750. CCOMMON_OPT += -marm
  751. FCOMMON_OPT += -marm
  752. else
  753. CCOMMON_OPT += -DOS_EMBEDDED -mthumb -mcpu=cortex-m4 -mfloat-abi=hard -mfpu=fpv4-sp-d16
  754. endif
  755. # If softfp abi is mentioned on the command line, force it.
  756. ifeq ($(ARM_SOFTFP_ABI), 1)
  757. CCOMMON_OPT += -mfloat-abi=softfp
  758. FCOMMON_OPT += -mfloat-abi=softfp
  759. endif
  760. ifeq ($(OSNAME), Android)
  761. ifeq ($(ARM_SOFTFP_ABI), 1)
  762. EXTRALIB += -lm
  763. else
  764. EXTRALIB += -Wl,-lm_hard
  765. endif
  766. endif
  767. endif
  768. ifeq ($(ARCH), arm64)
  769. NO_BINARY_MODE = 1
  770. BINARY_DEFINED = 1
  771. ifdef INTERFACE64
  772. ifneq ($(INTERFACE64), 0)
  773. ifeq ($(F_COMPILER), GFORTRAN)
  774. FCOMMON_OPT += -fdefault-integer-8
  775. endif
  776. ifeq ($(F_COMPILER), FLANG)
  777. FCOMMON_OPT += -i8
  778. endif
  779. endif
  780. endif
  781. endif
  782. ifeq ($(ARCH), riscv64)
  783. NO_BINARY_MODE = 1
  784. BINARY_DEFINED = 1
  785. ifdef INTERFACE64
  786. ifneq ($(INTERFACE64), 0)
  787. ifeq ($(F_COMPILER), GFORTRAN)
  788. FCOMMON_OPT += -fdefault-integer-8
  789. endif
  790. ifeq ($(F_COMPILER), FLANG)
  791. FCOMMON_OPT += -i8
  792. endif
  793. endif
  794. endif
  795. endif
  796. ifeq ($(ARCH), loongarch64)
  797. NO_BINARY_MODE = 1
  798. BINARY_DEFINED = 1
  799. ifdef INTERFACE64
  800. ifneq ($(INTERFACE64), 0)
  801. ifeq ($(F_COMPILER), GFORTRAN)
  802. FCOMMON_OPT += -fdefault-integer-8
  803. endif
  804. ifeq ($(F_COMPILER), FLANG)
  805. FCOMMON_OPT += -i8
  806. endif
  807. endif
  808. endif
  809. endif
  810. ifeq ($(ARCH), csky)
  811. NO_BINARY_MODE = 1
  812. BINARY_DEFINED = 1
  813. endif
  814. #
  815. # C Compiler dependent settings
  816. #
  817. # ifeq logical or. GCC or CLANG or LSB
  818. # http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
  819. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG LSB))
  820. CCOMMON_OPT += -Wall
  821. COMMON_PROF += -fno-inline
  822. NO_UNINITIALIZED_WARN = -Wno-uninitialized
  823. ifeq ($(QUIET_MAKE), 1)
  824. CCOMMON_OPT += $(NO_UNINITIALIZED_WARN) -Wno-unused
  825. endif
  826. ifdef NO_BINARY_MODE
  827. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  828. ifdef BINARY64
  829. CCOMMON_OPT += -mabi=64
  830. else
  831. CCOMMON_OPT += -mabi=n32
  832. endif
  833. BINARY_DEFINED = 1
  834. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  835. CCOMMON_OPT += -mabi=32
  836. BINARY_DEFINED = 1
  837. endif
  838. ifneq (, $(filter $(CORE), MIPS64_GENERIC))
  839. CCOMMON_OPT += -DNO_MSA
  840. FCOMMON_OPT += -DNO_MSA
  841. endif
  842. ifneq (, $(filter $(CORE),LOONGSON3R3 LOONGSON3R4))
  843. CCOMMON_OPT += -march=loongson3a
  844. FCOMMON_OPT += -march=loongson3a
  845. endif
  846. ifeq ($(CORE), MIPS24K)
  847. CCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  848. FCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  849. endif
  850. ifeq ($(CORE), MIPS1004K)
  851. CCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  852. FCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  853. endif
  854. ifeq ($(CORE), P5600)
  855. CCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  856. FCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  857. endif
  858. ifeq ($(CORE), I6400)
  859. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  860. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  861. endif
  862. ifeq ($(CORE), P6600)
  863. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  864. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  865. endif
  866. ifeq ($(CORE), I6500)
  867. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  868. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  869. endif
  870. ifeq ($(OSNAME), AIX)
  871. BINARY_DEFINED = 1
  872. endif
  873. ifeq ($(ARCH), loongarch64)
  874. LA64_ABI=$(shell $(CC) -mabi=lp64d -c $(TOPDIR)/cpuid_loongarch64.c -o /dev/null > /dev/null 2> /dev/null && echo lp64d)
  875. LA64_ARCH=$(shell $(CC) -march=loongarch64 -c $(TOPDIR)/cpuid_loongarch64.c -o /dev/null > /dev/null 2> /dev/null && echo loongarch64)
  876. ifneq ($(LA64_ABI), lp64d)
  877. LA64_ABI=lp64
  878. endif
  879. ifneq ($(LA64_ARCH), loongarch64)
  880. CCOMMON_OPT += -mabi=$(LA64_ABI)
  881. FCOMMON_OPT += -mabi=$(LA64_ABI)
  882. else
  883. CCOMMON_OPT += -march=loongarch64 -mabi=$(LA64_ABI)
  884. FCOMMON_OPT += -march=loongarch64 -mabi=$(LA64_ABI)
  885. endif
  886. endif
  887. endif
  888. ifndef BINARY_DEFINED
  889. ifneq ($(OSNAME), AIX)
  890. ifdef BINARY64
  891. ifneq ($(ARCH), riscv64)
  892. CCOMMON_OPT += -m64
  893. endif
  894. else
  895. CCOMMON_OPT += -m32
  896. endif
  897. endif
  898. endif
  899. endif
  900. ifeq ($(C_COMPILER), PGI)
  901. PGCVERSIONGT20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \> 20)
  902. PGCVERSIONEQ20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` == 20)
  903. PGCMINORVERSIONGE11 := $(shell expr `$(CC) --version|sed -n "2p" |cut -d "-" -f 1 |sed -e "s/[^0-9.]//g" |cut -c 4-5` \>= 11)
  904. PGCVERSIONCHECK := $(PGCVERSIONGT20)$(PGCVERSIONEQ20)$(PGCMINORVERSIONGE11)
  905. ifeq ($(PGCVERSIONCHECK), $(filter $(PGCVERSIONCHECK), 100 101 011))
  906. NEWPGI := 1
  907. PGCVERSIONGT21 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \> 21)
  908. PGCVERSIONEQ21 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` == 21)
  909. PGCVERSIONCHECK2 := $(PGCVERSIONGT21)$(PGCVERSIONEQ21)$(PGCMINORVERSIONGE11)
  910. ifeq ($(PGCVERSIONCHECK2), $(filter $(PGCVERSIONCHECK2), 100 101 011))
  911. NEWPGI2 := 1
  912. endif
  913. endif
  914. ifdef BINARY64
  915. ifeq ($(ARCH), x86_64)
  916. ifeq (,$(findstring tp,$(CFLAGS)))
  917. ifneq ($(NEWPGI2),1)
  918. CCOMMON_OPT += -tp p7-64
  919. else
  920. CCOMMON_OPT += -tp px
  921. endif
  922. endif
  923. ifneq ($(NEWPGI),1)
  924. CCOMMON_OPT += -D__MMX__ -Mnollvm
  925. endif
  926. else
  927. ifeq ($(ARCH), power)
  928. ifeq (,$(findstring tp,$(CFLAGS)))
  929. ifeq ($(CORE), POWER8)
  930. CCOMMON_OPT += -tp pwr8
  931. endif
  932. ifeq ($(CORE), POWER9)
  933. CCOMMON_OPT += -tp pwr9
  934. endif
  935. endif
  936. endif
  937. endif
  938. else
  939. ifneq ($(NEWPGI2),1)
  940. ifeq (,$(findstring tp,$(CFLAGS)))
  941. CCOMMON_OPT += -tp p7
  942. else
  943. CCOMMON_OPT += -tp px
  944. endif
  945. endif
  946. endif
  947. endif
  948. ifeq ($(C_COMPILER), PATHSCALE)
  949. ifdef BINARY64
  950. CCOMMON_OPT += -m64
  951. else
  952. CCOMMON_OPT += -m32
  953. endif
  954. endif
  955. #
  956. # Fortran Compiler dependent settings
  957. #
  958. ifeq ($(F_COMPILER), NAG)
  959. FCOMMON_OPT += -dcfuns -recursive -ieee=full -w=obs -thread_safe
  960. ifdef INTERFACE64
  961. ifneq ($(INTERFACE64), 0)
  962. FCOMMON_OPT += -i8
  963. endif
  964. endif
  965. ifeq ($(USE_OPENMP), 1)
  966. FCOMMON_OPT += -openmp
  967. endif
  968. endif
  969. ifeq ($(F_COMPILER), FLANG)
  970. CCOMMON_OPT += -DF_INTERFACE_FLANG
  971. FCOMMON_OPT += -Mrecursive -Kieee
  972. ifeq ($(OSNAME), Linux)
  973. ifeq ($(ARCH), x86_64)
  974. FLANG_VENDOR := $(shell $(FC) --version|head -1 |cut -f 1 -d " ")
  975. ifeq ($(FLANG_VENDOR), AMD)
  976. FCOMMON_OPT += -fno-unroll-loops
  977. endif
  978. endif
  979. endif
  980. ifdef BINARY64
  981. ifdef INTERFACE64
  982. ifneq ($(INTERFACE64), 0)
  983. FCOMMON_OPT += -i8
  984. endif
  985. endif
  986. FCOMMON_OPT += -Wall
  987. else
  988. FCOMMON_OPT += -Wall
  989. endif
  990. ifeq ($(USE_OPENMP), 1)
  991. FCOMMON_OPT += -fopenmp
  992. endif
  993. endif
  994. ifeq ($(F_COMPILER), G77)
  995. CCOMMON_OPT += -DF_INTERFACE_G77
  996. FCOMMON_OPT += -Wall
  997. ifndef NO_BINARY_MODE
  998. ifneq ($(OSNAME), AIX)
  999. ifdef BINARY64
  1000. FCOMMON_OPT += -m64
  1001. else
  1002. FCOMMON_OPT += -m32
  1003. endif
  1004. endif
  1005. endif
  1006. endif
  1007. ifeq ($(F_COMPILER), G95)
  1008. CCOMMON_OPT += -DF_INTERFACE_G95
  1009. FCOMMON_OPT += -Wall
  1010. ifneq ($(OSNAME), AIX)
  1011. ifndef NO_BINARY_MODE
  1012. ifdef BINARY64
  1013. FCOMMON_OPT += -m64
  1014. else
  1015. FCOMMON_OPT += -m32
  1016. endif
  1017. endif
  1018. ifneq ($(NO_LAPACKE), 1)
  1019. FCOMMON_OPT += -fno-second-underscore
  1020. endif
  1021. endif
  1022. endif
  1023. ifeq ($(F_COMPILER), $(filter $(F_COMPILER),GFORTRAN FLANGNEW))
  1024. CCOMMON_OPT += -DF_INTERFACE_GFORT
  1025. ifeq ($(F_COMPILER), GFORTRAN)
  1026. FCOMMON_OPT += -Wall
  1027. # make single-threaded LAPACK calls thread-safe #1847
  1028. FCOMMON_OPT += -frecursive
  1029. # work around ABI problem with passing single-character arguments
  1030. FCOMMON_OPT += -fno-optimize-sibling-calls
  1031. #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
  1032. ifneq ($(NOFORTRAN), 1)
  1033. ifneq ($(NOFORTRAN), 2)
  1034. ifneq ($(NO_LAPACK), 1)
  1035. EXTRALIB += -lgfortran
  1036. endif
  1037. endif
  1038. endif
  1039. endif
  1040. ifdef NO_BINARY_MODE
  1041. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  1042. ifdef BINARY64
  1043. FCOMMON_OPT += -mabi=64
  1044. else
  1045. FCOMMON_OPT += -mabi=n32
  1046. endif
  1047. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  1048. FCOMMON_OPT += -mabi=32
  1049. endif
  1050. ifeq ($(ARCH), $(filter $(ARCH),loongarch64))
  1051. ifdef INTERFACE64
  1052. ifneq ($(INTERFACE64), 0)
  1053. FCOMMON_OPT += -fdefault-integer-8
  1054. endif
  1055. endif
  1056. endif
  1057. else
  1058. ifdef BINARY64
  1059. ifneq ($(OSNAME), AIX)
  1060. ifneq ($(ARCH), riscv64)
  1061. FCOMMON_OPT += -m64
  1062. endif
  1063. endif
  1064. ifdef INTERFACE64
  1065. ifneq ($(INTERFACE64), 0)
  1066. FCOMMON_OPT += -fdefault-integer-8
  1067. endif
  1068. endif
  1069. else
  1070. ifneq ($(OSNAME), AIX)
  1071. FCOMMON_OPT += -m32
  1072. endif
  1073. endif
  1074. endif
  1075. ifeq ($(USE_OPENMP), 1)
  1076. FCOMMON_OPT += -fopenmp
  1077. endif
  1078. endif
  1079. ifeq ($(F_COMPILER), INTEL)
  1080. CCOMMON_OPT += -DF_INTERFACE_INTEL
  1081. ifdef INTERFACE64
  1082. ifneq ($(INTERFACE64), 0)
  1083. FCOMMON_OPT += -i8
  1084. endif
  1085. endif
  1086. FCOMMON_OPT += -recursive -fp-model strict -assume protect-parens
  1087. ifeq ($(USE_OPENMP), 1)
  1088. FCOMMON_OPT += -fopenmp
  1089. endif
  1090. endif
  1091. ifeq ($(F_COMPILER), FUJITSU)
  1092. CCOMMON_OPT += -DF_INTERFACE_FUJITSU
  1093. ifeq ($(USE_OPENMP), 1)
  1094. FCOMMON_OPT += -openmp
  1095. endif
  1096. endif
  1097. ifeq ($(F_COMPILER), IBM)
  1098. CCOMMON_OPT += -DF_INTERFACE_IBM
  1099. FEXTRALIB += -lxlf90
  1100. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG))
  1101. FCOMMON_OPT += -qextname -qzerosize
  1102. endif
  1103. # FCOMMON_OPT += -qarch=440
  1104. ifdef BINARY64
  1105. FCOMMON_OPT += -q64
  1106. ifdef INTERFACE64
  1107. ifneq ($(INTERFACE64), 0)
  1108. FCOMMON_OPT += -qintsize=8
  1109. endif
  1110. endif
  1111. else
  1112. FCOMMON_OPT += -q32
  1113. endif
  1114. endif
  1115. ifeq ($(F_COMPILER), PGI)
  1116. CCOMMON_OPT += -DF_INTERFACE_PGI
  1117. COMMON_PROF += -DPGICOMPILER
  1118. ifdef BINARY64
  1119. ifdef INTERFACE64
  1120. ifneq ($(INTERFACE64), 0)
  1121. FCOMMON_OPT += -i8
  1122. endif
  1123. endif
  1124. ifeq ($(ARCH), x86_64)
  1125. ifneq ($(NEWPGI2),1)
  1126. FCOMMON_OPT += -tp p7-64
  1127. else
  1128. FCOMMON_OPT += -tp px
  1129. endif
  1130. else
  1131. ifeq ($(ARCH), power)
  1132. ifeq ($(CORE), POWER6)
  1133. $(warning NVIDIA HPC compilers do not support POWER6.)
  1134. endif
  1135. ifeq ($(CORE), POWER8)
  1136. FCOMMON_OPT += -tp pwr8
  1137. endif
  1138. ifeq ($(CORE), POWER9)
  1139. FCOMMON_OPT += -tp pwr9
  1140. endif
  1141. ifeq ($(CORE), POWER10)
  1142. $(warning NVIDIA HPC compilers do not support POWER10.)
  1143. endif
  1144. endif
  1145. endif
  1146. else
  1147. FCOMMON_OPT += -tp p7
  1148. endif
  1149. FCOMMON_OPT += -Mrecursive -Kieee
  1150. ifeq ($(USE_OPENMP), 1)
  1151. FCOMMON_OPT += -mp
  1152. endif
  1153. endif
  1154. ifeq ($(F_COMPILER), PATHSCALE)
  1155. CCOMMON_OPT += -DF_INTERFACE_PATHSCALE
  1156. ifdef BINARY64
  1157. ifdef INTERFACE64
  1158. ifneq ($(INTERFACE64), 0)
  1159. FCOMMON_OPT += -i8
  1160. endif
  1161. endif
  1162. endif
  1163. ifeq ($(USE_OPENMP), 1)
  1164. FCOMMON_OPT += -mp
  1165. endif
  1166. endif
  1167. ifeq ($(F_COMPILER), OPEN64)
  1168. CCOMMON_OPT += -DF_INTERFACE_OPEN64
  1169. ifdef BINARY64
  1170. ifdef INTERFACE64
  1171. ifneq ($(INTERFACE64), 0)
  1172. FCOMMON_OPT += -i8
  1173. endif
  1174. endif
  1175. endif
  1176. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  1177. ifndef BINARY64
  1178. FCOMMON_OPT += -n32
  1179. else
  1180. FCOMMON_OPT += -n64
  1181. endif
  1182. ifeq ($(CORE), LOONGSON3R3)
  1183. FCOMMON_OPT += -loongson3 -static
  1184. endif
  1185. ifeq ($(CORE), LOONGSON3R4)
  1186. FCOMMON_OPT += -loongson3 -static
  1187. endif
  1188. else
  1189. ifndef BINARY64
  1190. FCOMMON_OPT += -m32
  1191. else
  1192. FCOMMON_OPT += -m64
  1193. endif
  1194. endif
  1195. ifeq ($(USE_OPENMP), 1)
  1196. FEXTRALIB += -lstdc++
  1197. FCOMMON_OPT += -mp
  1198. endif
  1199. endif
  1200. ifeq ($(C_COMPILER), OPEN64)
  1201. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  1202. ifndef BINARY64
  1203. CCOMMON_OPT += -n32
  1204. else
  1205. CCOMMON_OPT += -n64
  1206. endif
  1207. ifeq ($(CORE), LOONGSON3R3)
  1208. CCOMMON_OPT += -loongson3 -static
  1209. endif
  1210. ifeq ($(CORE), LOONGSON3R4)
  1211. CCOMMON_OPT += -loongson3 -static
  1212. endif
  1213. else
  1214. ifndef BINARY64
  1215. CCOMMON_OPT += -m32
  1216. else
  1217. CCOMMON_OPT += -m64
  1218. endif
  1219. endif
  1220. endif
  1221. ifeq ($(C_COMPILER), SUN)
  1222. CCOMMON_OPT += -w
  1223. ifeq ($(ARCH), x86)
  1224. CCOMMON_OPT += -m32
  1225. else
  1226. ifdef BINARY64
  1227. CCOMMON_OPT += -m64
  1228. else
  1229. CCOMMON_OPT += -m32
  1230. endif
  1231. endif
  1232. endif
  1233. ifeq ($(F_COMPILER), SUN)
  1234. CCOMMON_OPT += -DF_INTERFACE_SUN
  1235. FCOMMON_OPT += -ftrap=%none -xrecursive
  1236. ifeq ($(ARCH), x86)
  1237. FCOMMON_OPT += -m32
  1238. else
  1239. ifdef BINARY64
  1240. FCOMMON_OPT += -m64
  1241. else
  1242. FCOMMON_OPT += -m32
  1243. endif
  1244. endif
  1245. ifeq ($(USE_OPENMP), 1)
  1246. FCOMMON_OPT += -xopenmp=parallel
  1247. endif
  1248. endif
  1249. ifeq ($(F_COMPILER), COMPAQ)
  1250. CCOMMON_OPT += -DF_INTERFACE_COMPAQ
  1251. ifeq ($(USE_OPENMP), 1)
  1252. FCOMMON_OPT += -openmp
  1253. endif
  1254. endif
  1255. ifeq ($(F_COMPILER), CRAY)
  1256. CCOMMON_OPT += -DF_INTERFACE_INTEL
  1257. FCOMMON_OPT += -hnopattern
  1258. ifdef INTERFACE64
  1259. ifneq ($(INTERFACE64), 0)
  1260. FCOMMON_OPT += -s integer64
  1261. endif
  1262. endif
  1263. ifneq ($(USE_OPENMP), 1)
  1264. FCOMMON_OPT += -O noomp
  1265. endif
  1266. endif
  1267. ifdef BINARY64
  1268. ifdef INTERFACE64
  1269. ifneq ($(INTERFACE64), 0)
  1270. CCOMMON_OPT +=
  1271. #-DUSE64BITINT
  1272. endif
  1273. endif
  1274. endif
  1275. ifeq ($(NEED_PIC), 1)
  1276. ifeq ($(C_COMPILER), IBM)
  1277. CCOMMON_OPT += -qpic=large
  1278. else
  1279. CCOMMON_OPT += -fPIC
  1280. endif
  1281. ifeq ($(F_COMPILER), SUN)
  1282. FCOMMON_OPT += -pic
  1283. else ifeq ($(F_COMPILER), NAG)
  1284. FCOMMON_OPT += -PIC
  1285. else ifeq ($(F_COMPILER), IBM)
  1286. FCOMMON_OPT += -qpic=large
  1287. else
  1288. FCOMMON_OPT += -fPIC
  1289. endif
  1290. endif
  1291. ifeq ($(DYNAMIC_ARCH), 1)
  1292. CCOMMON_OPT += -DDYNAMIC_ARCH
  1293. endif
  1294. ifeq ($(DYNAMIC_OLDER), 1)
  1295. CCOMMON_OPT += -DDYNAMIC_OLDER
  1296. endif
  1297. ifeq ($(C_LAPACK), 1)
  1298. CCOMMON_OPT += -DC_LAPACK
  1299. endif
  1300. ifeq ($(NO_LAPACK), 1)
  1301. CCOMMON_OPT += -DNO_LAPACK
  1302. #Disable LAPACK C interface
  1303. NO_LAPACKE = 1
  1304. endif
  1305. ifeq ($(NO_LAPACKE), 1)
  1306. CCOMMON_OPT += -DNO_LAPACKE
  1307. endif
  1308. ifeq ($(NO_AVX), 1)
  1309. CCOMMON_OPT += -DNO_AVX
  1310. endif
  1311. ifeq ($(ARCH), x86)
  1312. CCOMMON_OPT += -DNO_AVX
  1313. endif
  1314. ifeq ($(NO_AVX2), 1)
  1315. CCOMMON_OPT += -DNO_AVX2
  1316. endif
  1317. ifeq ($(NO_AVX512), 1)
  1318. CCOMMON_OPT += -DNO_AVX512
  1319. endif
  1320. ifeq ($(NO_SVE), 1)
  1321. CCOMMON_OPT += -DNO_SVE
  1322. endif
  1323. ifeq ($(NO_SME), 1)
  1324. CCOMMON_OPT += -DNO_SME
  1325. endif
  1326. ifdef SMP
  1327. CCOMMON_OPT += -DSMP_SERVER
  1328. ifeq ($(ARCH), mips64)
  1329. USE_SIMPLE_THREADED_LEVEL3 = 1
  1330. endif
  1331. ifeq ($(USE_OPENMP), 1)
  1332. # USE_SIMPLE_THREADED_LEVEL3 = 1
  1333. # NO_AFFINITY = 1
  1334. CCOMMON_OPT += -DUSE_OPENMP
  1335. endif
  1336. ifeq ($(BIGNUMA), 1)
  1337. CCOMMON_OPT += -DBIGNUMA
  1338. endif
  1339. endif
  1340. ifeq ($(NO_WARMUP), 1)
  1341. CCOMMON_OPT += -DNO_WARMUP
  1342. endif
  1343. ifeq ($(CONSISTENT_FPCSR), 1)
  1344. CCOMMON_OPT += -DCONSISTENT_FPCSR
  1345. endif
  1346. # Only for development
  1347. # CCOMMON_OPT += -DPARAMTEST
  1348. # CCOMMON_OPT += -DPREFETCHTEST
  1349. # CCOMMON_OPT += -DNO_SWITCHING
  1350. # USE_PAPI = 1
  1351. ifdef USE_PAPI
  1352. CCOMMON_OPT += -DUSE_PAPI
  1353. EXTRALIB += -lpapi -lperfctr
  1354. endif
  1355. ifdef BUFFERSIZE
  1356. CCOMMON_OPT += -DBUFFERSIZE=$(BUFFERSIZE)
  1357. endif
  1358. ifdef DYNAMIC_THREADS
  1359. CCOMMON_OPT += -DDYNAMIC_THREADS
  1360. endif
  1361. CCOMMON_OPT += -DMAX_CPU_NUMBER=$(NUM_THREADS)
  1362. CCOMMON_OPT += -DMAX_PARALLEL_NUMBER=$(NUM_PARALLEL)
  1363. ifdef USE_SIMPLE_THREADED_LEVEL3
  1364. CCOMMON_OPT += -DUSE_SIMPLE_THREADED_LEVEL3
  1365. endif
  1366. ifeq ($(USE_TLS), 1)
  1367. CCOMMON_OPT += -DUSE_TLS
  1368. endif
  1369. ifeq ($(BUILD_BFLOAT16), 1)
  1370. CCOMMON_OPT += -DBUILD_BFLOAT16
  1371. endif
  1372. ifeq ($(BUILD_HFLOAT16), 1)
  1373. CCOMMON_OPT += -DBUILD_HFLOAT16
  1374. endif
  1375. ifeq ($(BUILD_SINGLE), 1)
  1376. CCOMMON_OPT += -DBUILD_SINGLE=1
  1377. endif
  1378. ifeq ($(BUILD_DOUBLE), 1)
  1379. CCOMMON_OPT += -DBUILD_DOUBLE=1
  1380. endif
  1381. ifeq ($(BUILD_COMPLEX), 1)
  1382. CCOMMON_OPT += -DBUILD_COMPLEX=1
  1383. endif
  1384. ifeq ($(BUILD_COMPLEX16), 1)
  1385. CCOMMON_OPT += -DBUILD_COMPLEX16=1
  1386. endif
  1387. CCOMMON_OPT += -DVERSION=\"$(VERSION)\"
  1388. ifndef SYMBOLPREFIX
  1389. SYMBOLPREFIX =
  1390. endif
  1391. ifndef SYMBOLSUFFIX
  1392. SYMBOLSUFFIX =
  1393. endif
  1394. ifndef LIBSONAMEBASE
  1395. LIBSONAMEBASE = openblas
  1396. endif
  1397. ifndef LIBNAMEPREFIX
  1398. LIBNAMEPREFIX =
  1399. endif
  1400. SYMPREFIX=$(SYMBOLPREFIX)
  1401. ifeq ($(SYMBOLPREFIX),$(LIBNAMEPREFIX))
  1402. SYMPREFIX=
  1403. endif
  1404. SYMSUFFIX=$(SYMBOLSUFFIX)
  1405. ifeq ($(SYMBOLSUFFIX),$(LIBNAMESUFFIX))
  1406. SYMSUFFIX=
  1407. endif
  1408. ifndef LIBNAMESUFFIX
  1409. LIBNAMEBASE = $(SYMPREFIX)$(LIBSONAMEBASE)$(SYMSUFFIX)
  1410. else
  1411. LIBNAMEBASE = $(SYMPREFIX)$(LIBSONAMEBASE)$(SYMSUFFIX)$(LIBNAMESUFFIX)
  1412. endif
  1413. ifeq ($(OSNAME), CYGWIN_NT)
  1414. LIBPREFIX = cyg$(LIBNAMEPREFIX)$(LIBNAMEBASE)
  1415. else
  1416. LIBPREFIX = lib$(LIBNAMEPREFIX)$(LIBNAMEBASE)
  1417. endif
  1418. KERNELDIR = $(TOPDIR)/kernel/$(ARCH)
  1419. include $(TOPDIR)/Makefile.$(ARCH)
  1420. ifneq ($(C_COMPILER), PGI)
  1421. ifneq ($(C_COMPILER), SUN)
  1422. CCOMMON_OPT += -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME
  1423. endif
  1424. endif
  1425. CCOMMON_OPT += -DASMNAME=$(FU)$(*F) -DASMFNAME=$(FU)$(*F)$(BU) -DNAME=$(*F)$(BU) -DCNAME=$(*F) -DCHAR_NAME=\"$(*F)$(BU)\" -DCHAR_CNAME=\"$(*F)\"
  1426. ifeq ($(CORE), PPC440)
  1427. CCOMMON_OPT += -DALLOC_QALLOC
  1428. endif
  1429. ifeq ($(CORE), PPC440FP2)
  1430. STATIC_ALLOCATION = 1
  1431. endif
  1432. ifneq ($(OSNAME), Linux)
  1433. NO_AFFINITY = 1
  1434. endif
  1435. ifneq ($(ARCH), x86_64)
  1436. ifneq ($(ARCH), x86)
  1437. NO_AFFINITY = 1
  1438. endif
  1439. endif
  1440. ifeq ($(ARCH), POWER)
  1441. ifeq ($(DEBUG), 1)
  1442. CCOMMON_OPT := $(filter-out -O%, $(CCOMMON_OPT)) -O0
  1443. FCOMMON_OPT := $(filter-out -O%, $(FCOMMON_OPT)) -O0
  1444. endif
  1445. endif
  1446. ifdef NO_AFFINITY
  1447. ifeq ($(NO_AFFINITY), 0)
  1448. override undefine NO_AFFINITY
  1449. else
  1450. CCOMMON_OPT += -DNO_AFFINITY
  1451. endif
  1452. endif
  1453. ifdef FUNCTION_PROFILE
  1454. CCOMMON_OPT += -DFUNCTION_PROFILE
  1455. endif
  1456. ifdef SHMEM_ALLOCATION
  1457. ifneq ($(SHMEM_ALLOCATION), 0)
  1458. CCOMMON_OPT += -DALLOC_SHM
  1459. endif
  1460. endif
  1461. ifdef HUGETLB_ALLOCATION
  1462. ifneq ($(HUGETLB_ALLOCATION), 0)
  1463. CCOMMON_OPT += -DALLOC_HUGETLB
  1464. endif
  1465. endif
  1466. ifdef HUGETLBFILE_ALLOCATION
  1467. ifneq ($(HUGETLBFILE_ALLOCATION), 0)
  1468. CCOMMON_OPT += -DALLOC_HUGETLBFILE -DHUGETLB_FILE_NAME=$(HUGETLBFILE_ALLOCATION)
  1469. endif
  1470. endif
  1471. ifdef STATIC_ALLOCATION
  1472. CCOMMON_OPT += -DALLOC_STATIC
  1473. endif
  1474. ifdef DEVICEDRIVER_ALLOCATION
  1475. CCOMMON_OPT += -DALLOC_DEVICEDRIVER -DDEVICEDRIVER_NAME=\"/dev/mapper\"
  1476. endif
  1477. ifdef MIXED_MEMORY_ALLOCATION
  1478. CCOMMON_OPT += -DMIXED_MEMORY_ALLOCATION
  1479. endif
  1480. ifeq ($(OSNAME), SunOS)
  1481. TAR = gtar
  1482. PATCH = gpatch
  1483. GREP = ggrep
  1484. AWK = nawk
  1485. else
  1486. TAR = tar
  1487. PATCH = patch
  1488. GREP = grep
  1489. AWK = awk
  1490. endif
  1491. ifndef MD5SUM
  1492. MD5SUM = md5sum
  1493. endif
  1494. REVISION = -r$(VERSION)
  1495. MAJOR_VERSION = $(word 1,$(subst ., ,$(VERSION)))
  1496. ifeq ($(DEBUG), 1)
  1497. COMMON_OPT += -g
  1498. endif
  1499. ifeq ($(DEBUG), 1)
  1500. FCOMMON_OPT += -g
  1501. endif
  1502. ifndef COMMON_OPT
  1503. COMMON_OPT = -O2
  1504. endif
  1505. ifndef FCOMMON_OPT
  1506. FCOMMON_OPT = -O2 -frecursive
  1507. endif
  1508. override CFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
  1509. override PFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
  1510. override FFLAGS += $(COMMON_OPT) $(FCOMMON_OPT)
  1511. override FPFLAGS += $(FCOMMON_OPT) $(COMMON_PROF)
  1512. #MAKEOVERRIDES =
  1513. ifeq ($(NEED_PIC), 1)
  1514. ifeq (,$(findstring PIC,$(FFLAGS)))
  1515. ifneq ($(F_COMPILER),IBM)
  1516. override FFLAGS += -fPIC
  1517. endif
  1518. endif
  1519. endif
  1520. #For LAPACK Fortran codes.
  1521. #Disable -fopenmp for LAPACK Fortran codes on Windows.
  1522. ifdef OS_WINDOWS
  1523. LAPACK_FFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FFLAGS))
  1524. LAPACK_FPFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FPFLAGS))
  1525. else
  1526. LAPACK_FFLAGS := $(FFLAGS)
  1527. LAPACK_FPFLAGS := $(FPFLAGS)
  1528. endif
  1529. ifeq ($(F_COMPILER),NAG)
  1530. LAPACK_FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1531. override FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1532. endif
  1533. ifeq ($(F_COMPILER),CRAY)
  1534. LAPACK_FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1535. override FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1536. endif
  1537. ifeq ($(F_COMPILER),FLANGNEW)
  1538. LAPACK_FFLAGS := $(filter-out -m32 -m64 -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 -mtune=% -mabi=% ,$(FFLAGS))
  1539. override FFLAGS := $(filter-out -m32 -m64 -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 -mtune=% -mabi=% ,$(FFLAGS))
  1540. endif
  1541. LAPACK_CFLAGS = $(CFLAGS)
  1542. LAPACK_CFLAGS += -DHAVE_LAPACK_CONFIG_H
  1543. ifdef INTERFACE64
  1544. ifneq ($(INTERFACE64), 0)
  1545. LAPACK_CFLAGS += -DLAPACK_ILP64
  1546. endif
  1547. endif
  1548. ifdef OS_WINDOWS
  1549. LAPACK_CFLAGS += -DOPENBLAS_OS_WINDOWS
  1550. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1551. endif
  1552. ifeq ($(C_COMPILER), LSB)
  1553. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1554. endif
  1555. ifndef SUFFIX
  1556. SUFFIX = o
  1557. endif
  1558. ifndef PSUFFIX
  1559. PSUFFIX = po
  1560. endif
  1561. ifndef LIBSUFFIX
  1562. LIBSUFFIX = a
  1563. endif
  1564. ifneq ($(DYNAMIC_ARCH), 1)
  1565. ifndef SMP
  1566. LIBNAME = $(LIBPREFIX)_$(LIBCORE)$(REVISION).$(LIBSUFFIX)
  1567. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)$(REVISION)_p.$(LIBSUFFIX)
  1568. else
  1569. LIBNAME = $(LIBPREFIX)_$(LIBCORE)p$(REVISION).$(LIBSUFFIX)
  1570. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)p$(REVISION)_p.$(LIBSUFFIX)
  1571. endif
  1572. else
  1573. ifndef SMP
  1574. LIBNAME = $(LIBPREFIX)$(REVISION).$(LIBSUFFIX)
  1575. LIBNAME_P = $(LIBPREFIX)$(REVISION)_p.$(LIBSUFFIX)
  1576. else
  1577. LIBNAME = $(LIBPREFIX)p$(REVISION).$(LIBSUFFIX)
  1578. LIBNAME_P = $(LIBPREFIX)p$(REVISION)_p.$(LIBSUFFIX)
  1579. endif
  1580. endif
  1581. ifeq ($(FIXED_LIBNAME),1)
  1582. LIBNAME = lib$(LIBNAMEPREFIX)$(LIBSONAMEBASE)$(LIBNAMESUFFIX).$(LIBSUFFIX)
  1583. LIBNAME_P = lib$(LIBNAMEPREFIX)$(LISOBNAMEBASE)$(LIBNAMESUFFIX)_p.$(LIBSUFFIX)
  1584. endif
  1585. LIBDLLNAME = $(LIBPREFIX).dll
  1586. IMPLIBNAME = lib$(LIBNAMEBASE).dll.a
  1587. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.so)
  1588. LIBDYNNAME = $(LIBNAME:.$(LIBSUFFIX)=.dylib)
  1589. LIBDEFNAME = $(LIBNAME:.$(LIBSUFFIX)=.def)
  1590. LIBEXPNAME = $(LIBNAME:.$(LIBSUFFIX)=.exp)
  1591. LIBZIPNAME = $(LIBNAME:.$(LIBSUFFIX)=.zip)
  1592. LIBS = $(TOPDIR)/$(LIBNAME)
  1593. LIBS_P = $(TOPDIR)/$(LIBNAME_P)
  1594. LIB_COMPONENTS = BLAS
  1595. ifneq ($(NO_CBLAS), 1)
  1596. LIB_COMPONENTS += CBLAS
  1597. endif
  1598. ifneq ($(NO_LAPACK), 1)
  1599. LIB_COMPONENTS += LAPACK
  1600. ifneq ($(NO_LAPACKE), 1)
  1601. LIB_COMPONENTS += LAPACKE
  1602. endif
  1603. ifeq ($(BUILD_RELAPACK), 1)
  1604. LIB_COMPONENTS += ReLAPACK
  1605. endif
  1606. endif
  1607. ifeq ($(ONLY_CBLAS), 1)
  1608. LIB_COMPONENTS = CBLAS
  1609. endif
  1610. export OSNAME
  1611. export ARCH
  1612. export CORE
  1613. export LIBCORE
  1614. export __BYTE_ORDER__
  1615. export ELF_VERSION
  1616. export PGCPATH
  1617. export CONFIG
  1618. export CC
  1619. export FC
  1620. export BU
  1621. export FU
  1622. export NEED2UNDERSCORES
  1623. export USE_THREAD
  1624. export NUM_THREADS
  1625. export NUM_CORES
  1626. export SMP
  1627. export MAKEFILE_RULE
  1628. export NEED_PIC
  1629. export BINARY
  1630. export BINARY32
  1631. export BINARY64
  1632. export F_COMPILER
  1633. export C_COMPILER
  1634. export USE_OPENMP
  1635. export CROSS
  1636. export CROSS_SUFFIX
  1637. export NOFORTRAN
  1638. export C_LAPACK
  1639. export NO_FBLAS
  1640. export EXTRALIB
  1641. export CEXTRALIB
  1642. export FEXTRALIB
  1643. export HAVE_SSE
  1644. export HAVE_SSE2
  1645. export HAVE_SSE3
  1646. export HAVE_SSSE3
  1647. export HAVE_SSE4_1
  1648. export HAVE_SSE4_2
  1649. export HAVE_SSE4A
  1650. export HAVE_SSE5
  1651. export HAVE_AVX
  1652. export HAVE_AVX2
  1653. export HAVE_FMA3
  1654. export HAVE_VFP
  1655. export HAVE_VFPV3
  1656. export HAVE_VFPV4
  1657. export HAVE_NEON
  1658. ifndef NO_MSA
  1659. export HAVE_MSA
  1660. export MSA_FLAGS
  1661. endif
  1662. export KERNELDIR
  1663. export FUNCTION_PROFILE
  1664. export TARGET_CORE
  1665. export NO_AVX512
  1666. export NO_AVX2
  1667. export BUILD_BFLOAT16
  1668. export BUILD_HFLOAT16
  1669. export NO_LSX
  1670. export NO_LASX
  1671. export BGEMM_UNROLL_M
  1672. export BGEMM_UNROLL_N
  1673. export SBGEMM_UNROLL_M
  1674. export SBGEMM_UNROLL_N
  1675. export SHGEMM_UNROLL_M
  1676. export SHGEMM_UNROLL_N
  1677. export SGEMM_UNROLL_M
  1678. export SGEMM_UNROLL_N
  1679. export DGEMM_UNROLL_M
  1680. export DGEMM_UNROLL_N
  1681. export QGEMM_UNROLL_M
  1682. export QGEMM_UNROLL_N
  1683. export CGEMM_UNROLL_M
  1684. export CGEMM_UNROLL_N
  1685. export ZGEMM_UNROLL_M
  1686. export ZGEMM_UNROLL_N
  1687. export XGEMM_UNROLL_M
  1688. export XGEMM_UNROLL_N
  1689. export CGEMM3M_UNROLL_M
  1690. export CGEMM3M_UNROLL_N
  1691. export ZGEMM3M_UNROLL_M
  1692. export ZGEMM3M_UNROLL_N
  1693. export XGEMM3M_UNROLL_M
  1694. export XGEMM3M_UNROLL_N
  1695. ifdef USE_CUDA
  1696. export CUDADIR
  1697. export CUCC
  1698. export CUFLAGS
  1699. export CULIB
  1700. endif
  1701. .SUFFIXES: .$(PSUFFIX) .$(SUFFIX) .f
  1702. .f.$(SUFFIX):
  1703. $(FC) $(FFLAGS) -c $< -o $(@F)
  1704. .f.$(PSUFFIX):
  1705. $(FC) $(FPFLAGS) -pg -c $< -o $(@F)
  1706. ifdef BINARY64
  1707. PATHSCALEPATH = /opt/pathscale/lib/3.1
  1708. PGIPATH = /opt/pgi/linux86-64/7.1-5/lib
  1709. else
  1710. PATHSCALEPATH = /opt/pathscale/lib/3.1/32
  1711. PGIPATH = /opt/pgi/linux86/7.1-5/lib
  1712. endif
  1713. ACMLPATH = /opt/acml/4.3.0
  1714. ifneq ($(OSNAME), Darwin)
  1715. MKLPATH = /opt/intel/mkl/10.2.2.025/lib
  1716. else
  1717. MKLPATH = /Library/Frameworks/Intel_MKL.framework/Versions/10.0.1.014/lib
  1718. endif
  1719. ATLASPATH = /opt/atlas/3.9.17/opteron
  1720. FLAMEPATH = $(HOME)/flame/lib
  1721. ifneq ($(OSNAME), SunOS)
  1722. SUNPATH = /opt/sunstudio12.1
  1723. else
  1724. SUNPATH = /opt/SUNWspro
  1725. endif