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
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942943944945946947948949950951952953954955956957958959960961962963964965966967968969970971972973974975976977978979980981982983984985986987988989990991992993994995996997998999100010011002100310041005100610071008100910101011101210131014101510161017101810191020102110221023102410251026102710281029103010311032103310341035103610371038103910401041104210431044104510461047104810491050105110521053105410551056105710581059106010611062106310641065106610671068106910701071107210731074107510761077107810791080108110821083108410851086108710881089109010911092109310941095109610971098109911001101110211031104110511061107110811091110111111121113111411151116111711181119112011211122112311241125112611271128112911301131113211331134113511361137113811391140114111421143114411451146114711481149115011511152115311541155115611571158115911601161116211631164116511661167116811691170117111721173117411751176117711781179118011811182118311841185118611871188118911901191119211931194119511961197119811991200120112021203120412051206120712081209121012111212121312141215121612171218121912201221122212231224122512261227122812291230123112321233123412351236123712381239124012411242124312441245124612471248124912501251125212531254125512561257125812591260126112621263126412651266126712681269127012711272127312741275127612771278127912801281128212831284128512861287128812891290129112921293129412951296129712981299130013011302130313041305130613071308130913101311131213131314131513161317131813191320132113221323132413251326132713281329133013311332133313341335133613371338133913401341134213431344134513461347134813491350135113521353135413551356135713581359136013611362136313641365136613671368136913701371137213731374137513761377137813791380138113821383138413851386138713881389139013911392139313941395139613971398139914001401140214031404140514061407140814091410141114121413141414151416141714181419142014211422142314241425142614271428142914301431143214331434143514361437143814391440144114421443144414451446144714481449145014511452145314541455145614571458145914601461146214631464146514661467146814691470147114721473147414751476147714781479148014811482148314841485148614871488148914901491149214931494149514961497149814991500150115021503150415051506150715081509151015111512151315141515151615171518151915201521152215231524152515261527152815291530153115321533153415351536153715381539154015411542154315441545154615471548154915501551155215531554155515561557155815591560156115621563156415651566156715681569157015711572157315741575157615771578157915801581158215831584158515861587158815891590159115921593159415951596159715981599160016011602160316041605160616071608160916101611161216131614161516161617161816191620162116221623162416251626162716281629163016311632163316341635163616371638163916401641164216431644164516461647164816491650165116521653165416551656165716581659166016611662166316641665166616671668166916701671167216731674167516761677167816791680168116821683168416851686168716881689169016911692169316941695169616971698169917001701170217031704170517061707170817091710171117121713171417151716171717181719172017211722172317241725172617271728172917301731173217331734173517361737173817391740174117421743174417451746174717481749175017511752175317541755175617571758175917601761176217631764176517661767176817691770177117721773177417751776177717781779178017811782178317841785178617871788178917901791179217931794179517961797179817991800180118021803180418051806180718081809181018111812181318141815181618171818181918201821182218231824182518261827182818291830183118321833183418351836183718381839184018411842184318441845184618471848184918501851185218531854185518561857185818591860186118621863186418651866186718681869187018711872187318741875187618771878187918801881188218831884188518861887188818891890189118921893189418951896189718981899190019011902190319041905190619071908190919101911191219131914191519161917191819191920192119221923192419251926192719281929193019311932193319341935193619371938193919401941194219431944194519461947194819491950195119521953195419551956195719581959196019611962196319641965196619671968
  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 ($(ARCH), arm64)
  244. GEMM_GEMV_FORWARD = 1
  245. GEMM_GEMV_FORWARD_BF16 = 1
  246. endif
  247. ifeq ($(ARCH), riscv)
  248. GEMM_GEMV_FORWARD = 1
  249. endif
  250. ifeq ($(ARCH), power)
  251. GEMM_GEMV_FORWARD = 1
  252. GEMM_GEMV_FORWARD_BF16 = 1
  253. endif
  254. ifeq ($(SMALL_MATRIX_OPT), 1)
  255. CCOMMON_OPT += -DSMALL_MATRIX_OPT
  256. endif
  257. ifneq ($(ONLY_CBLAS), 1)
  258. ifeq ($(GEMM_GEMV_FORWARD), 1)
  259. CCOMMON_OPT += -DGEMM_GEMV_FORWARD
  260. endif
  261. ifeq ($(GEMM_GEMV_FORWARD_BF16), 1)
  262. CCOMMON_OPT += -DGEMM_GEMV_FORWARD_BF16
  263. endif
  264. endif
  265. # This operation is expensive, so execution should be once.
  266. ifndef GOTOBLAS_MAKEFILE
  267. export GOTOBLAS_MAKEFILE = 1
  268. # Generating Makefile.conf and config.h
  269. 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)
  270. endif
  271. ifndef TARGET_CORE
  272. -include $(TOPDIR)/Makefile.conf
  273. else
  274. HAVE_NEON=
  275. HAVE_VFP=
  276. HAVE_VFPV3=
  277. HAVE_VFPV4=
  278. HAVE_MMX=
  279. HAVE_SSE=
  280. HAVE_SSE2=
  281. HAVE_SSE3=
  282. HAVE_SSSE3=
  283. HAVE_SSE4_1=
  284. HAVE_SSE4_2=
  285. HAVE_SSE4A=
  286. HAVE_SSE5=
  287. HAVE_AVX=
  288. HAVE_AVX2=
  289. HAVE_FMA3=
  290. include $(TOPDIR)/Makefile_kernel.conf
  291. endif
  292. ifndef NUM_PARALLEL
  293. NUM_PARALLEL = 1
  294. endif
  295. ifndef NUM_THREADS
  296. NUM_THREADS = $(NUM_CORES)
  297. endif
  298. ifeq ($(NUM_THREADS), 1)
  299. override USE_THREAD = 0
  300. override USE_OPENMP = 0
  301. endif
  302. ifdef USE_THREAD
  303. ifeq ($(USE_THREAD), 0)
  304. SMP =
  305. else
  306. SMP = 1
  307. endif
  308. else
  309. ifeq ($(NUM_THREADS), 1)
  310. SMP =
  311. else
  312. SMP = 1
  313. endif
  314. endif
  315. ifeq ($(SMP), 1)
  316. USE_LOCKING =
  317. endif
  318. ifndef NEED_PIC
  319. NEED_PIC = 1
  320. endif
  321. ARFLAGS =
  322. CPP = $(COMPILER) -E
  323. AR ?= $(CROSS_SUFFIX)ar
  324. AS ?= $(CROSS_SUFFIX)as
  325. LD ?= $(CROSS_SUFFIX)ld
  326. RANLIB ?= $(CROSS_SUFFIX)ranlib
  327. NM = $(CROSS_SUFFIX)nm
  328. DLLWRAP = $(CROSS_SUFFIX)dllwrap
  329. OBJCOPY = $(CROSS_SUFFIX)objcopy
  330. OBJCONV = $(CROSS_SUFFIX)objconv
  331. # When fortran support was either not detected or actively deselected, only build BLAS.
  332. ifeq ($(NOFORTRAN), 1)
  333. C_LAPACK = 1
  334. override FEXTRALIB =
  335. endif
  336. ifeq ($(C_COMPILER), GCC)
  337. GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
  338. GCCVERSIONGT4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 4)
  339. GCCVERSIONGT5 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 5)
  340. GCCVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 7)
  341. GCCVERSIONGTEQ8 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 8)
  342. GCCVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  343. GCCVERSIONGTEQ10 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 10)
  344. GCCVERSIONGTEQ11 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 11)
  345. GCCVERSIONGTEQ12 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 12)
  346. GCCVERSIONGTEQ13 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 13)
  347. GCCVERSIONGTEQ14 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 14)
  348. # Note that the behavior of -dumpversion is compile-time-configurable for
  349. # gcc-7.x and newer. Use -dumpfullversion there
  350. ifeq ($(GCCVERSIONGTEQ7),1)
  351. GCCDUMPVERSION_PARAM := -dumpfullversion
  352. else
  353. GCCDUMPVERSION_PARAM := -dumpversion
  354. endif
  355. GCCMINORVERSIONGTEQ1 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 1)
  356. GCCMINORVERSIONGTEQ2 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 2)
  357. GCCMINORVERSIONGTEQ4 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 4)
  358. GCCMINORVERSIONGTEQ7 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 7)
  359. endif
  360. ifeq ($(C_COMPILER), CLANG)
  361. CLANGVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  362. CLANGVERSIONGTEQ12 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 12)
  363. endif
  364. #
  365. # OS dependent settings
  366. #
  367. ifeq ($(OSNAME), Darwin)
  368. ifndef MACOSX_DEPLOYMENT_TARGET
  369. ifeq ($(ARCH), arm64)
  370. export MACOSX_DEPLOYMENT_TARGET=11.0
  371. ifeq ($(C_COMPILER), GCC)
  372. export NO_SVE = 1
  373. export NO_SME = 1
  374. endif
  375. else
  376. export MACOSX_DEPLOYMENT_TARGET=10.8
  377. endif
  378. endif
  379. MD5SUM = md5 -r
  380. XCVER = $(shell pkgutil --pkg-info=com.apple.pkg.Xcode |awk '/version:/ {print $2}'|cut -d: -f2|cut -f1 -d.)
  381. ifeq (x$(XCVER)x,xx)
  382. XCVER = $(shell pkgutil --pkg-info=com.apple.pkg.CLTools_Executables |awk '/version:/ {print $2}'|cut -d: -f2|cut -f1 -d.)
  383. endif
  384. ifeq (x$(XCVER), x 15)
  385. CCOMMON_OPT += -Wl,-ld_classic
  386. FCOMMON_OPT += -Wl,-ld_classic
  387. endif
  388. ifeq (x$(XCVER), x 16)
  389. ifeq ($(F_COMPILER), GFORTRAN)
  390. override CEXTRALIB := $(filter-out(-lto_library, $(CEXTRALIB)))
  391. endif
  392. endif
  393. endif
  394. ifneq (,$(findstring $(OSNAME), FreeBSD OpenBSD DragonFly))
  395. MD5SUM = md5 -r
  396. endif
  397. ifeq ($(OSNAME), NetBSD)
  398. MD5SUM = md5 -n
  399. endif
  400. ifeq ($(OSNAME), Linux)
  401. EXTRALIB += -lm
  402. #NO_EXPRECISION = 1
  403. endif
  404. ifeq ($(OSNAME), Android)
  405. EXTRALIB += -lm
  406. endif
  407. ifeq ($(OSNAME), AIX)
  408. EXTRALIB += -lm
  409. endif
  410. ifeq ($(OSNAME), $(filter $(OSNAME),FreeBSD OpenBSD NetBSD DragonFly))
  411. ifeq ($(ARCH), $(filter $(ARCH),arm arm64))
  412. EXTRALIB += -lm
  413. endif
  414. endif
  415. ifeq ($(OSNAME), WINNT)
  416. NEED_PIC = 0
  417. NO_EXPRECISION = 1
  418. EXTRALIB += -defaultlib:advapi32
  419. SUFFIX = obj
  420. PSUFFIX = pobj
  421. LIBSUFFIX = a
  422. ifeq ($(C_COMPILER), CLANG)
  423. CCOMMON_OPT += -DMS_ABI
  424. endif
  425. #Version tests for supporting specific features (MS_ABI, POWER9 intrinsics)
  426. ifeq ($(GCCVERSIONGT4), 1)
  427. # GCC Major version > 4
  428. # It is compatible with MSVC ABI.
  429. CCOMMON_OPT += -DMS_ABI
  430. endif
  431. ifeq ($(GCCVERSIONGTEQ4), 1)
  432. ifeq ($(GCCMINORVERSIONGTEQ7), 1)
  433. # GCC Version >=4.7
  434. # It is compatible with MSVC ABI.
  435. CCOMMON_OPT += -DMS_ABI
  436. endif
  437. endif
  438. # Ensure the correct stack alignment on Win32
  439. # http://permalink.gmane.org/gmane.comp.lib.openblas.general/97
  440. ifeq ($(ARCH), x86)
  441. CCOMMON_OPT += -mincoming-stack-boundary=2
  442. FCOMMON_OPT += -mincoming-stack-boundary=2
  443. endif
  444. endif
  445. ifeq ($(OSNAME), Interix)
  446. NEED_PIC = 0
  447. NO_EXPRECISION = 1
  448. INTERIX_TOOL_DIR = /opt/gcc.3.3/i586-pc-interix3/bin
  449. endif
  450. ifeq ($(OSNAME), CYGWIN_NT)
  451. NEED_PIC = 0
  452. NO_EXPRECISION = 1
  453. OS_CYGWIN_NT = 1
  454. endif
  455. ifneq ($(OSNAME), WINNT)
  456. ifneq ($(OSNAME), CYGWIN_NT)
  457. ifneq ($(OSNAME), Interix)
  458. ifneq ($(OSNAME), Android)
  459. ifdef SMP
  460. EXTRALIB += -lpthread
  461. endif
  462. endif
  463. endif
  464. endif
  465. endif
  466. # ifeq logical or
  467. ifeq ($(OSNAME), $(filter $(OSNAME),WINNT CYGWIN_NT Interix))
  468. OS_WINDOWS=1
  469. endif
  470. ifdef QUAD_PRECISION
  471. CCOMMON_OPT += -DQUAD_PRECISION
  472. NO_EXPRECISION = 1
  473. endif
  474. ifneq ($(ARCH), x86)
  475. ifneq ($(ARCH), x86_64)
  476. NO_EXPRECISION = 1
  477. endif
  478. endif
  479. ifdef UTEST_CHECK
  480. CCOMMON_OPT += -DUTEST_CHECK
  481. SANITY_CHECK = 1
  482. endif
  483. ifdef SANITY_CHECK
  484. CCOMMON_OPT += -DSANITY_CHECK -DREFNAME=$(*F)f$(BU)
  485. endif
  486. MAX_STACK_ALLOC ?= 2048
  487. ifneq ($(MAX_STACK_ALLOC), 0)
  488. CCOMMON_OPT += -DMAX_STACK_ALLOC=$(MAX_STACK_ALLOC)
  489. endif
  490. ifdef USE_LOCKING
  491. ifneq ($(USE_LOCKING), 0)
  492. CCOMMON_OPT += -DUSE_LOCKING
  493. endif
  494. endif
  495. #
  496. # Architecture dependent settings
  497. #
  498. ifeq ($(ARCH), x86)
  499. ifndef BINARY
  500. NO_BINARY_MODE = 1
  501. endif
  502. ifeq ($(CORE), generic)
  503. #NO_EXPRECISION = 1
  504. endif
  505. ifndef NO_EXPRECISION
  506. ifeq ($(F_COMPILER), GFORTRAN)
  507. # ifeq logical or. GCC or LSB
  508. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  509. EXPRECISION = 1
  510. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  511. FCOMMON_OPT += -m128bit-long-double
  512. endif
  513. ifeq ($(C_COMPILER), CLANG)
  514. EXPRECISION = 1
  515. CCOMMON_OPT += -DEXPRECISION
  516. FCOMMON_OPT += -m128bit-long-double
  517. endif
  518. endif
  519. endif
  520. endif
  521. ifeq ($(ARCH), x86_64)
  522. ifeq ($(CORE), generic)
  523. #NO_EXPRECISION = 1
  524. endif
  525. ifndef NO_EXPRECISION
  526. ifeq ($(F_COMPILER), GFORTRAN)
  527. # ifeq logical or. GCC or LSB
  528. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  529. EXPRECISION = 1
  530. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  531. FCOMMON_OPT += -m128bit-long-double
  532. endif
  533. ifeq ($(C_COMPILER), CLANG)
  534. EXPRECISION = 1
  535. CCOMMON_OPT += -DEXPRECISION
  536. FCOMMON_OPT += -m128bit-long-double
  537. endif
  538. endif
  539. endif
  540. endif
  541. ifeq ($(C_COMPILER), INTEL)
  542. CCOMMON_OPT += -wd981
  543. endif
  544. ifeq ($(USE_OPENMP), 1)
  545. #check
  546. ifeq ($(USE_THREAD), 0)
  547. $(error OpenBLAS: Cannot set both USE_OPENMP=1 and USE_THREAD=0. The USE_THREAD=0 is only for building single thread version.)
  548. endif
  549. # ifeq logical or. GCC or LSB
  550. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  551. CCOMMON_OPT += -fopenmp
  552. endif
  553. ifeq ($(C_COMPILER), CLANG)
  554. CCOMMON_OPT += -fopenmp
  555. ifeq ($(F_COMPILER), GFORTRAN)
  556. FEXTRALIB := $(subst -lgomp,-lomp,$(FEXTRALIB))
  557. endif
  558. endif
  559. ifeq ($(C_COMPILER), INTEL)
  560. CCOMMON_OPT += -fopenmp
  561. endif
  562. ifeq ($(C_COMPILER), PGI)
  563. CCOMMON_OPT += -mp
  564. endif
  565. ifeq ($(C_COMPILER), OPEN64)
  566. CCOMMON_OPT += -mp
  567. CEXTRALIB += -lstdc++
  568. endif
  569. ifeq ($(C_COMPILER), PATHSCALE)
  570. CCOMMON_OPT += -mp
  571. endif
  572. endif
  573. ifeq ($(DYNAMIC_ARCH), 1)
  574. ifeq ($(ARCH), x86)
  575. DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
  576. CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
  577. endif
  578. ifeq ($(ARCH), x86_64)
  579. DYNAMIC_CORE = PRESCOTT CORE2
  580. ifeq ($(DYNAMIC_OLDER), 1)
  581. DYNAMIC_CORE += PENRYN DUNNINGTON
  582. endif
  583. DYNAMIC_CORE += NEHALEM
  584. ifeq ($(DYNAMIC_OLDER), 1)
  585. DYNAMIC_CORE += OPTERON OPTERON_SSE3
  586. endif
  587. DYNAMIC_CORE += BARCELONA
  588. ifeq ($(DYNAMIC_OLDER), 1)
  589. DYNAMIC_CORE += BOBCAT ATOM NANO
  590. endif
  591. ifneq ($(NO_AVX), 1)
  592. DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER STEAMROLLER EXCAVATOR
  593. endif
  594. ifneq ($(NO_AVX2), 1)
  595. DYNAMIC_CORE += HASWELL ZEN
  596. endif
  597. ifneq ($(NO_AVX512), 1)
  598. ifneq ($(NO_AVX2), 1)
  599. DYNAMIC_CORE += SKYLAKEX COOPERLAKE SAPPHIRERAPIDS
  600. endif
  601. endif
  602. endif
  603. ifdef DYNAMIC_LIST
  604. override DYNAMIC_CORE = PRESCOTT $(DYNAMIC_LIST)
  605. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_PRESCOTT
  606. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  607. CCOMMON_OPT += $(XCCOMMON_OPT)
  608. #CCOMMON_OPT += -DDYNAMIC_LIST='$(DYNAMIC_LIST)'
  609. endif
  610. ifeq ($(ARCH), arm64)
  611. DYNAMIC_CORE = ARMV8
  612. DYNAMIC_CORE += CORTEXA53
  613. DYNAMIC_CORE += CORTEXA57
  614. DYNAMIC_CORE += NEOVERSEN1
  615. ifneq ($(NO_SVE), 1)
  616. DYNAMIC_CORE += NEOVERSEV1
  617. DYNAMIC_CORE += NEOVERSEN2
  618. DYNAMIC_CORE += ARMV8SVE
  619. DYNAMIC_CORE += A64FX
  620. endif
  621. ifneq ($(NO_SME), 1)
  622. DYNAMIC_CORE += ARMV9SME
  623. endif
  624. DYNAMIC_CORE += THUNDERX
  625. DYNAMIC_CORE += THUNDERX2T99
  626. DYNAMIC_CORE += TSV110
  627. DYNAMIC_CORE += EMAG8180
  628. DYNAMIC_CORE += THUNDERX3T110
  629. ifdef DYNAMIC_LIST
  630. override DYNAMIC_CORE = ARMV8 $(DYNAMIC_LIST)
  631. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_ARMV8
  632. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  633. endif
  634. endif
  635. ifeq ($(ARCH), mips64)
  636. DYNAMIC_CORE = LOONGSON3R3 LOONGSON3R4 MIPS64_GENERIC
  637. ifdef DYNAMIC_LIST
  638. override DYNAMIC_CORE = MIPS64_GENERIC $(DYNAMIC_LIST)
  639. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_MIPS64_GENERIC
  640. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  641. endif
  642. endif
  643. ifeq ($(ARCH), loongarch64)
  644. DYNAMIC_CORE = LA64_GENERIC LA264 LA464
  645. endif
  646. ifeq ($(ARCH), riscv64)
  647. DYNAMIC_CORE = RISCV64_GENERIC
  648. DYNAMIC_CORE += RISCV64_ZVL128B
  649. DYNAMIC_CORE += RISCV64_ZVL256B
  650. ifdef DYNAMIC_LIST
  651. override DYNAMIC_CORE = RISCV64_GENERIC $(DYNAMIC_LIST)
  652. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_RISCV64_GENERIC
  653. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  654. endif
  655. endif
  656. ifeq ($(ARCH), zarch)
  657. DYNAMIC_CORE = ZARCH_GENERIC
  658. # if the compiler accepts -march=arch11 or -march=z13 and can compile a file
  659. # with z13-specific inline assembly, then we can include support for Z13.
  660. # note: -march=z13 is equivalent to -march=arch11 yet some compiler releases
  661. # only support one or the other.
  662. # note: LLVM version 6.x supported -march=z13 yet could not handle vector
  663. # registers in inline assembly, so the check for supporting the -march flag is
  664. # not enough.
  665. ZARCH_TEST_COMPILE=-c $(TOPDIR)/kernel/zarch/damin_z13.c -I$(TOPDIR) -o /dev/null > /dev/null 2> /dev/null
  666. ZARCH_CC_SUPPORTS_ARCH11=$(shell $(CC) -march=arch11 $(ZARCH_TEST_COMPILE) && echo 1)
  667. ZARCH_CC_SUPPORTS_Z13=$(shell $(CC) -march=z13 $(ZARCH_TEST_COMPILE) && echo 1)
  668. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH11), $(ZARCH_CC_SUPPORTS_Z13)), 1)
  669. DYNAMIC_CORE += Z13
  670. CCOMMON_OPT += -DDYN_Z13
  671. else
  672. $(info OpenBLAS: Not building Z13 kernels because the compiler $(CC) does not support it)
  673. endif
  674. # as above for z13, check for -march=arch12 and z14 support in the compiler.
  675. ZARCH_CC_SUPPORTS_ARCH12=$(shell $(CC) -march=arch12 $(ZARCH_TEST_COMPILE) && echo 1)
  676. ZARCH_CC_SUPPORTS_Z14=$(shell $(CC) -march=z14 $(ZARCH_TEST_COMPILE) && echo 1)
  677. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH12), $(ZARCH_CC_SUPPORTS_Z14)), 1)
  678. DYNAMIC_CORE += Z14
  679. CCOMMON_OPT += -DDYN_Z14
  680. else
  681. $(info OpenBLAS: Not building Z14 kernels because the compiler $(CC) does not support it)
  682. endif
  683. endif # ARCH zarch
  684. ifeq ($(ARCH), power)
  685. ifneq ($(C_COMPILER), PGI)
  686. DYNAMIC_CORE = POWER6
  687. DYNAMIC_CORE += POWER8
  688. ifneq ($(C_COMPILER), GCC)
  689. DYNAMIC_CORE += POWER9
  690. DYNAMIC_CORE += POWER10
  691. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  692. endif
  693. ifeq ($(C_COMPILER), GCC)
  694. ifeq ($(GCCVERSIONGT5), 1)
  695. DYNAMIC_CORE += POWER9
  696. else
  697. $(info, OpenBLAS: Your gcc version is too old to build the POWER9 kernels.)
  698. endif
  699. ifeq ($(OSNAME), AIX)
  700. LDVERSIONGTEQ35 := 1
  701. else
  702. LDVERSIONGTEQ35 := $(shell expr `$(CC) -Wl,--version 2> /dev/null | head -1 | cut -f2 -d "." | cut -f1 -d "-"` \>= 35)
  703. endif
  704. ifeq ($(GCCVERSIONGTEQ11)$(LDVERSIONGTEQ35), 11)
  705. DYNAMIC_CORE += POWER10
  706. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  707. else ifeq ($(GCCVERSIONGTEQ10), 1)
  708. ifeq ($(GCCMINORVERSIONGTEQ2)$(LDVERSIONGTEQ35), 11)
  709. DYNAMIC_CORE += POWER10
  710. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  711. endif
  712. else
  713. $(info, OpenBLAS: Your gcc version is too old to build the POWER10 kernels.)
  714. endif
  715. endif
  716. else
  717. DYNAMIC_CORE = POWER8
  718. DYNAMIC_CORE += POWER9
  719. endif
  720. endif
  721. # If DYNAMIC_CORE is not set, DYNAMIC_ARCH cannot do anything, so force it to empty
  722. ifndef DYNAMIC_CORE
  723. override DYNAMIC_ARCH=
  724. endif
  725. endif
  726. ifeq ($(ARCH), ia64)
  727. NO_BINARY_MODE = 1
  728. BINARY_DEFINED = 1
  729. ifeq ($(F_COMPILER), GFORTRAN)
  730. ifeq ($(C_COMPILER), GCC)
  731. EXPRECISION = 1
  732. CCOMMON_OPT += -DEXPRECISION
  733. endif
  734. endif
  735. endif
  736. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  737. NO_BINARY_MODE = 1
  738. endif
  739. ifeq ($(ARCH), alpha)
  740. NO_BINARY_MODE = 1
  741. BINARY_DEFINED = 1
  742. endif
  743. ifeq ($(ARCH), arm)
  744. NO_BINARY_MODE = 1
  745. BINARY_DEFINED = 1
  746. ifneq ($(EMBEDDED), 1)
  747. CCOMMON_OPT += -marm
  748. FCOMMON_OPT += -marm
  749. else
  750. CCOMMON_OPT += -DOS_EMBEDDED -mthumb -mcpu=cortex-m4 -mfloat-abi=hard -mfpu=fpv4-sp-d16
  751. endif
  752. # If softfp abi is mentioned on the command line, force it.
  753. ifeq ($(ARM_SOFTFP_ABI), 1)
  754. CCOMMON_OPT += -mfloat-abi=softfp
  755. FCOMMON_OPT += -mfloat-abi=softfp
  756. endif
  757. ifeq ($(OSNAME), Android)
  758. ifeq ($(ARM_SOFTFP_ABI), 1)
  759. EXTRALIB += -lm
  760. else
  761. EXTRALIB += -Wl,-lm_hard
  762. endif
  763. endif
  764. endif
  765. ifeq ($(ARCH), arm64)
  766. NO_BINARY_MODE = 1
  767. BINARY_DEFINED = 1
  768. ifdef INTERFACE64
  769. ifneq ($(INTERFACE64), 0)
  770. ifeq ($(F_COMPILER), GFORTRAN)
  771. FCOMMON_OPT += -fdefault-integer-8
  772. endif
  773. ifeq ($(F_COMPILER), FLANG)
  774. FCOMMON_OPT += -i8
  775. endif
  776. endif
  777. endif
  778. endif
  779. ifeq ($(ARCH), riscv64)
  780. NO_BINARY_MODE = 1
  781. BINARY_DEFINED = 1
  782. ifdef INTERFACE64
  783. ifneq ($(INTERFACE64), 0)
  784. ifeq ($(F_COMPILER), GFORTRAN)
  785. FCOMMON_OPT += -fdefault-integer-8
  786. endif
  787. ifeq ($(F_COMPILER), FLANG)
  788. FCOMMON_OPT += -i8
  789. endif
  790. endif
  791. endif
  792. endif
  793. ifeq ($(ARCH), loongarch64)
  794. NO_BINARY_MODE = 1
  795. BINARY_DEFINED = 1
  796. ifdef INTERFACE64
  797. ifneq ($(INTERFACE64), 0)
  798. ifeq ($(F_COMPILER), GFORTRAN)
  799. FCOMMON_OPT += -fdefault-integer-8
  800. endif
  801. ifeq ($(F_COMPILER), FLANG)
  802. FCOMMON_OPT += -i8
  803. endif
  804. endif
  805. endif
  806. endif
  807. ifeq ($(ARCH), csky)
  808. NO_BINARY_MODE = 1
  809. BINARY_DEFINED = 1
  810. endif
  811. #
  812. # C Compiler dependent settings
  813. #
  814. # ifeq logical or. GCC or CLANG or LSB
  815. # http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
  816. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG LSB))
  817. CCOMMON_OPT += -Wall
  818. COMMON_PROF += -fno-inline
  819. NO_UNINITIALIZED_WARN = -Wno-uninitialized
  820. ifeq ($(QUIET_MAKE), 1)
  821. CCOMMON_OPT += $(NO_UNINITIALIZED_WARN) -Wno-unused
  822. endif
  823. ifdef NO_BINARY_MODE
  824. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  825. ifdef BINARY64
  826. CCOMMON_OPT += -mabi=64
  827. else
  828. CCOMMON_OPT += -mabi=n32
  829. endif
  830. BINARY_DEFINED = 1
  831. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  832. CCOMMON_OPT += -mabi=32
  833. BINARY_DEFINED = 1
  834. endif
  835. ifneq (, $(filter $(CORE), MIPS64_GENERIC))
  836. CCOMMON_OPT += -DNO_MSA
  837. FCOMMON_OPT += -DNO_MSA
  838. endif
  839. ifneq (, $(filter $(CORE),LOONGSON3R3 LOONGSON3R4))
  840. CCOMMON_OPT += -march=loongson3a
  841. FCOMMON_OPT += -march=loongson3a
  842. endif
  843. ifeq ($(CORE), MIPS24K)
  844. CCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  845. FCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  846. endif
  847. ifeq ($(CORE), MIPS1004K)
  848. CCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  849. FCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  850. endif
  851. ifeq ($(CORE), P5600)
  852. CCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  853. FCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  854. endif
  855. ifeq ($(CORE), I6400)
  856. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  857. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  858. endif
  859. ifeq ($(CORE), P6600)
  860. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  861. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  862. endif
  863. ifeq ($(CORE), I6500)
  864. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  865. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  866. endif
  867. ifeq ($(OSNAME), AIX)
  868. BINARY_DEFINED = 1
  869. endif
  870. ifeq ($(ARCH), loongarch64)
  871. LA64_ABI=$(shell $(CC) -mabi=lp64d -c $(TOPDIR)/cpuid_loongarch64.c -o /dev/null > /dev/null 2> /dev/null && echo lp64d)
  872. LA64_ARCH=$(shell $(CC) -march=loongarch64 -c $(TOPDIR)/cpuid_loongarch64.c -o /dev/null > /dev/null 2> /dev/null && echo loongarch64)
  873. ifneq ($(LA64_ABI), lp64d)
  874. LA64_ABI=lp64
  875. endif
  876. ifneq ($(LA64_ARCH), loongarch64)
  877. CCOMMON_OPT += -mabi=$(LA64_ABI)
  878. FCOMMON_OPT += -mabi=$(LA64_ABI)
  879. else
  880. CCOMMON_OPT += -march=loongarch64 -mabi=$(LA64_ABI)
  881. FCOMMON_OPT += -march=loongarch64 -mabi=$(LA64_ABI)
  882. endif
  883. endif
  884. endif
  885. ifndef BINARY_DEFINED
  886. ifneq ($(OSNAME), AIX)
  887. ifdef BINARY64
  888. ifneq ($(ARCH), riscv64)
  889. CCOMMON_OPT += -m64
  890. endif
  891. else
  892. CCOMMON_OPT += -m32
  893. endif
  894. endif
  895. endif
  896. endif
  897. ifeq ($(C_COMPILER), PGI)
  898. PGCVERSIONGT20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \> 20)
  899. PGCVERSIONEQ20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` == 20)
  900. PGCMINORVERSIONGE11 := $(shell expr `$(CC) --version|sed -n "2p" |cut -d "-" -f 1 |sed -e "s/[^0-9.]//g" |cut -c 4-5` \>= 11)
  901. PGCVERSIONCHECK := $(PGCVERSIONGT20)$(PGCVERSIONEQ20)$(PGCMINORVERSIONGE11)
  902. ifeq ($(PGCVERSIONCHECK), $(filter $(PGCVERSIONCHECK), 100 101 011))
  903. NEWPGI := 1
  904. PGCVERSIONGT21 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \> 21)
  905. PGCVERSIONEQ21 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` == 21)
  906. PGCVERSIONCHECK2 := $(PGCVERSIONGT21)$(PGCVERSIONEQ21)$(PGCMINORVERSIONGE11)
  907. ifeq ($(PGCVERSIONCHECK2), $(filter $(PGCVERSIONCHECK2), 100 101 011))
  908. NEWPGI2 := 1
  909. endif
  910. endif
  911. ifdef BINARY64
  912. ifeq ($(ARCH), x86_64)
  913. ifeq (,$(findstring tp,$(CFLAGS)))
  914. ifneq ($(NEWPGI2),1)
  915. CCOMMON_OPT += -tp p7-64
  916. else
  917. CCOMMON_OPT += -tp px
  918. endif
  919. endif
  920. ifneq ($(NEWPGI),1)
  921. CCOMMON_OPT += -D__MMX__ -Mnollvm
  922. endif
  923. else
  924. ifeq ($(ARCH), power)
  925. ifeq (,$(findstring tp,$(CFLAGS)))
  926. ifeq ($(CORE), POWER8)
  927. CCOMMON_OPT += -tp pwr8
  928. endif
  929. ifeq ($(CORE), POWER9)
  930. CCOMMON_OPT += -tp pwr9
  931. endif
  932. endif
  933. endif
  934. endif
  935. else
  936. ifneq ($(NEWPGI2),1)
  937. ifeq (,$(findstring tp,$(CFLAGS)))
  938. CCOMMON_OPT += -tp p7
  939. else
  940. CCOMMON_OPT += -tp px
  941. endif
  942. endif
  943. endif
  944. endif
  945. ifeq ($(C_COMPILER), PATHSCALE)
  946. ifdef BINARY64
  947. CCOMMON_OPT += -m64
  948. else
  949. CCOMMON_OPT += -m32
  950. endif
  951. endif
  952. #
  953. # Fortran Compiler dependent settings
  954. #
  955. ifeq ($(F_COMPILER), NAG)
  956. FCOMMON_OPT += -dcfuns -recursive -ieee=full -w=obs -thread_safe
  957. ifdef INTERFACE64
  958. ifneq ($(INTERFACE64), 0)
  959. FCOMMON_OPT += -i8
  960. endif
  961. endif
  962. ifeq ($(USE_OPENMP), 1)
  963. FCOMMON_OPT += -openmp
  964. endif
  965. endif
  966. ifeq ($(F_COMPILER), FLANG)
  967. CCOMMON_OPT += -DF_INTERFACE_FLANG
  968. FCOMMON_OPT += -Mrecursive -Kieee
  969. ifeq ($(OSNAME), Linux)
  970. ifeq ($(ARCH), x86_64)
  971. FLANG_VENDOR := $(shell $(FC) --version|head -1 |cut -f 1 -d " ")
  972. ifeq ($(FLANG_VENDOR), AMD)
  973. FCOMMON_OPT += -fno-unroll-loops
  974. endif
  975. endif
  976. endif
  977. ifdef BINARY64
  978. ifdef INTERFACE64
  979. ifneq ($(INTERFACE64), 0)
  980. FCOMMON_OPT += -i8
  981. endif
  982. endif
  983. FCOMMON_OPT += -Wall
  984. else
  985. FCOMMON_OPT += -Wall
  986. endif
  987. ifeq ($(USE_OPENMP), 1)
  988. FCOMMON_OPT += -fopenmp
  989. endif
  990. endif
  991. ifeq ($(F_COMPILER), G77)
  992. CCOMMON_OPT += -DF_INTERFACE_G77
  993. FCOMMON_OPT += -Wall
  994. ifndef NO_BINARY_MODE
  995. ifneq ($(OSNAME), AIX)
  996. ifdef BINARY64
  997. FCOMMON_OPT += -m64
  998. else
  999. FCOMMON_OPT += -m32
  1000. endif
  1001. endif
  1002. endif
  1003. endif
  1004. ifeq ($(F_COMPILER), G95)
  1005. CCOMMON_OPT += -DF_INTERFACE_G95
  1006. FCOMMON_OPT += -Wall
  1007. ifneq ($(OSNAME), AIX)
  1008. ifndef NO_BINARY_MODE
  1009. ifdef BINARY64
  1010. FCOMMON_OPT += -m64
  1011. else
  1012. FCOMMON_OPT += -m32
  1013. endif
  1014. endif
  1015. ifneq ($(NO_LAPACKE), 1)
  1016. FCOMMON_OPT += -fno-second-underscore
  1017. endif
  1018. endif
  1019. endif
  1020. ifeq ($(F_COMPILER), $(filter $(F_COMPILER),GFORTRAN FLANGNEW))
  1021. CCOMMON_OPT += -DF_INTERFACE_GFORT
  1022. ifeq ($(F_COMPILER), GFORTRAN)
  1023. FCOMMON_OPT += -Wall
  1024. # make single-threaded LAPACK calls thread-safe #1847
  1025. FCOMMON_OPT += -frecursive
  1026. # work around ABI problem with passing single-character arguments
  1027. FCOMMON_OPT += -fno-optimize-sibling-calls
  1028. #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
  1029. ifneq ($(NOFORTRAN), 1)
  1030. ifneq ($(NOFORTRAN), 2)
  1031. ifneq ($(NO_LAPACK), 1)
  1032. EXTRALIB += -lgfortran
  1033. endif
  1034. endif
  1035. endif
  1036. endif
  1037. ifdef NO_BINARY_MODE
  1038. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  1039. ifdef BINARY64
  1040. FCOMMON_OPT += -mabi=64
  1041. else
  1042. FCOMMON_OPT += -mabi=n32
  1043. endif
  1044. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  1045. FCOMMON_OPT += -mabi=32
  1046. endif
  1047. ifeq ($(ARCH), $(filter $(ARCH),loongarch64))
  1048. ifdef INTERFACE64
  1049. ifneq ($(INTERFACE64), 0)
  1050. FCOMMON_OPT += -fdefault-integer-8
  1051. endif
  1052. endif
  1053. endif
  1054. else
  1055. ifdef BINARY64
  1056. ifneq ($(OSNAME), AIX)
  1057. ifneq ($(ARCH), riscv64)
  1058. FCOMMON_OPT += -m64
  1059. endif
  1060. endif
  1061. ifdef INTERFACE64
  1062. ifneq ($(INTERFACE64), 0)
  1063. FCOMMON_OPT += -fdefault-integer-8
  1064. endif
  1065. endif
  1066. else
  1067. ifneq ($(OSNAME), AIX)
  1068. FCOMMON_OPT += -m32
  1069. endif
  1070. endif
  1071. endif
  1072. ifeq ($(USE_OPENMP), 1)
  1073. FCOMMON_OPT += -fopenmp
  1074. endif
  1075. endif
  1076. ifeq ($(F_COMPILER), INTEL)
  1077. CCOMMON_OPT += -DF_INTERFACE_INTEL
  1078. ifdef INTERFACE64
  1079. ifneq ($(INTERFACE64), 0)
  1080. FCOMMON_OPT += -i8
  1081. endif
  1082. endif
  1083. FCOMMON_OPT += -recursive -fp-model strict -assume protect-parens
  1084. ifeq ($(USE_OPENMP), 1)
  1085. FCOMMON_OPT += -fopenmp
  1086. endif
  1087. endif
  1088. ifeq ($(F_COMPILER), FUJITSU)
  1089. CCOMMON_OPT += -DF_INTERFACE_FUJITSU
  1090. ifeq ($(USE_OPENMP), 1)
  1091. FCOMMON_OPT += -openmp
  1092. endif
  1093. endif
  1094. ifeq ($(F_COMPILER), IBM)
  1095. CCOMMON_OPT += -DF_INTERFACE_IBM
  1096. FEXTRALIB += -lxlf90
  1097. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG))
  1098. FCOMMON_OPT += -qextname -qzerosize
  1099. endif
  1100. # FCOMMON_OPT += -qarch=440
  1101. ifdef BINARY64
  1102. FCOMMON_OPT += -q64
  1103. ifdef INTERFACE64
  1104. ifneq ($(INTERFACE64), 0)
  1105. FCOMMON_OPT += -qintsize=8
  1106. endif
  1107. endif
  1108. else
  1109. FCOMMON_OPT += -q32
  1110. endif
  1111. endif
  1112. ifeq ($(F_COMPILER), PGI)
  1113. CCOMMON_OPT += -DF_INTERFACE_PGI
  1114. COMMON_PROF += -DPGICOMPILER
  1115. ifdef BINARY64
  1116. ifdef INTERFACE64
  1117. ifneq ($(INTERFACE64), 0)
  1118. FCOMMON_OPT += -i8
  1119. endif
  1120. endif
  1121. ifeq ($(ARCH), x86_64)
  1122. ifneq ($(NEWPGI2),1)
  1123. FCOMMON_OPT += -tp p7-64
  1124. else
  1125. FCOMMON_OPT += -tp px
  1126. endif
  1127. else
  1128. ifeq ($(ARCH), power)
  1129. ifeq ($(CORE), POWER6)
  1130. $(warning NVIDIA HPC compilers do not support POWER6.)
  1131. endif
  1132. ifeq ($(CORE), POWER8)
  1133. FCOMMON_OPT += -tp pwr8
  1134. endif
  1135. ifeq ($(CORE), POWER9)
  1136. FCOMMON_OPT += -tp pwr9
  1137. endif
  1138. ifeq ($(CORE), POWER10)
  1139. $(warning NVIDIA HPC compilers do not support POWER10.)
  1140. endif
  1141. endif
  1142. endif
  1143. else
  1144. FCOMMON_OPT += -tp p7
  1145. endif
  1146. FCOMMON_OPT += -Mrecursive -Kieee
  1147. ifeq ($(USE_OPENMP), 1)
  1148. FCOMMON_OPT += -mp
  1149. endif
  1150. endif
  1151. ifeq ($(F_COMPILER), PATHSCALE)
  1152. CCOMMON_OPT += -DF_INTERFACE_PATHSCALE
  1153. ifdef BINARY64
  1154. ifdef INTERFACE64
  1155. ifneq ($(INTERFACE64), 0)
  1156. FCOMMON_OPT += -i8
  1157. endif
  1158. endif
  1159. endif
  1160. ifeq ($(USE_OPENMP), 1)
  1161. FCOMMON_OPT += -mp
  1162. endif
  1163. endif
  1164. ifeq ($(F_COMPILER), OPEN64)
  1165. CCOMMON_OPT += -DF_INTERFACE_OPEN64
  1166. ifdef BINARY64
  1167. ifdef INTERFACE64
  1168. ifneq ($(INTERFACE64), 0)
  1169. FCOMMON_OPT += -i8
  1170. endif
  1171. endif
  1172. endif
  1173. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  1174. ifndef BINARY64
  1175. FCOMMON_OPT += -n32
  1176. else
  1177. FCOMMON_OPT += -n64
  1178. endif
  1179. ifeq ($(CORE), LOONGSON3R3)
  1180. FCOMMON_OPT += -loongson3 -static
  1181. endif
  1182. ifeq ($(CORE), LOONGSON3R4)
  1183. FCOMMON_OPT += -loongson3 -static
  1184. endif
  1185. else
  1186. ifndef BINARY64
  1187. FCOMMON_OPT += -m32
  1188. else
  1189. FCOMMON_OPT += -m64
  1190. endif
  1191. endif
  1192. ifeq ($(USE_OPENMP), 1)
  1193. FEXTRALIB += -lstdc++
  1194. FCOMMON_OPT += -mp
  1195. endif
  1196. endif
  1197. ifeq ($(C_COMPILER), OPEN64)
  1198. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  1199. ifndef BINARY64
  1200. CCOMMON_OPT += -n32
  1201. else
  1202. CCOMMON_OPT += -n64
  1203. endif
  1204. ifeq ($(CORE), LOONGSON3R3)
  1205. CCOMMON_OPT += -loongson3 -static
  1206. endif
  1207. ifeq ($(CORE), LOONGSON3R4)
  1208. CCOMMON_OPT += -loongson3 -static
  1209. endif
  1210. else
  1211. ifndef BINARY64
  1212. CCOMMON_OPT += -m32
  1213. else
  1214. CCOMMON_OPT += -m64
  1215. endif
  1216. endif
  1217. endif
  1218. ifeq ($(C_COMPILER), SUN)
  1219. CCOMMON_OPT += -w
  1220. ifeq ($(ARCH), x86)
  1221. CCOMMON_OPT += -m32
  1222. else
  1223. ifdef BINARY64
  1224. CCOMMON_OPT += -m64
  1225. else
  1226. CCOMMON_OPT += -m32
  1227. endif
  1228. endif
  1229. endif
  1230. ifeq ($(F_COMPILER), SUN)
  1231. CCOMMON_OPT += -DF_INTERFACE_SUN
  1232. FCOMMON_OPT += -ftrap=%none -xrecursive
  1233. ifeq ($(ARCH), x86)
  1234. FCOMMON_OPT += -m32
  1235. else
  1236. ifdef BINARY64
  1237. FCOMMON_OPT += -m64
  1238. else
  1239. FCOMMON_OPT += -m32
  1240. endif
  1241. endif
  1242. ifeq ($(USE_OPENMP), 1)
  1243. FCOMMON_OPT += -xopenmp=parallel
  1244. endif
  1245. endif
  1246. ifeq ($(F_COMPILER), COMPAQ)
  1247. CCOMMON_OPT += -DF_INTERFACE_COMPAQ
  1248. ifeq ($(USE_OPENMP), 1)
  1249. FCOMMON_OPT += -openmp
  1250. endif
  1251. endif
  1252. ifeq ($(F_COMPILER), CRAY)
  1253. CCOMMON_OPT += -DF_INTERFACE_INTEL
  1254. FCOMMON_OPT += -hnopattern
  1255. ifdef INTERFACE64
  1256. ifneq ($(INTERFACE64), 0)
  1257. FCOMMON_OPT += -s integer64
  1258. endif
  1259. endif
  1260. ifneq ($(USE_OPENMP), 1)
  1261. FCOMMON_OPT += -O noomp
  1262. endif
  1263. endif
  1264. ifdef BINARY64
  1265. ifdef INTERFACE64
  1266. ifneq ($(INTERFACE64), 0)
  1267. CCOMMON_OPT +=
  1268. #-DUSE64BITINT
  1269. endif
  1270. endif
  1271. endif
  1272. ifeq ($(NEED_PIC), 1)
  1273. ifeq ($(C_COMPILER), IBM)
  1274. CCOMMON_OPT += -qpic=large
  1275. else
  1276. CCOMMON_OPT += -fPIC
  1277. endif
  1278. ifeq ($(F_COMPILER), SUN)
  1279. FCOMMON_OPT += -pic
  1280. else ifeq ($(F_COMPILER), NAG)
  1281. FCOMMON_OPT += -PIC
  1282. else ifeq ($(F_COMPILER), IBM)
  1283. FCOMMON_OPT += -qpic=large
  1284. else
  1285. FCOMMON_OPT += -fPIC
  1286. endif
  1287. endif
  1288. ifeq ($(DYNAMIC_ARCH), 1)
  1289. CCOMMON_OPT += -DDYNAMIC_ARCH
  1290. endif
  1291. ifeq ($(DYNAMIC_OLDER), 1)
  1292. CCOMMON_OPT += -DDYNAMIC_OLDER
  1293. endif
  1294. ifeq ($(C_LAPACK), 1)
  1295. CCOMMON_OPT += -DC_LAPACK
  1296. endif
  1297. ifeq ($(NO_LAPACK), 1)
  1298. CCOMMON_OPT += -DNO_LAPACK
  1299. #Disable LAPACK C interface
  1300. NO_LAPACKE = 1
  1301. endif
  1302. ifeq ($(NO_LAPACKE), 1)
  1303. CCOMMON_OPT += -DNO_LAPACKE
  1304. endif
  1305. ifeq ($(NO_AVX), 1)
  1306. CCOMMON_OPT += -DNO_AVX
  1307. endif
  1308. ifeq ($(ARCH), x86)
  1309. CCOMMON_OPT += -DNO_AVX
  1310. endif
  1311. ifeq ($(NO_AVX2), 1)
  1312. CCOMMON_OPT += -DNO_AVX2
  1313. endif
  1314. ifeq ($(NO_AVX512), 1)
  1315. CCOMMON_OPT += -DNO_AVX512
  1316. endif
  1317. ifeq ($(NO_SVE), 1)
  1318. CCOMMON_OPT += -DNO_SVE
  1319. endif
  1320. ifeq ($(NO_SME), 1)
  1321. CCOMMON_OPT += -DNO_SME
  1322. endif
  1323. ifdef SMP
  1324. CCOMMON_OPT += -DSMP_SERVER
  1325. ifeq ($(ARCH), mips64)
  1326. USE_SIMPLE_THREADED_LEVEL3 = 1
  1327. endif
  1328. ifeq ($(USE_OPENMP), 1)
  1329. # USE_SIMPLE_THREADED_LEVEL3 = 1
  1330. # NO_AFFINITY = 1
  1331. CCOMMON_OPT += -DUSE_OPENMP
  1332. endif
  1333. ifeq ($(BIGNUMA), 1)
  1334. CCOMMON_OPT += -DBIGNUMA
  1335. endif
  1336. endif
  1337. ifeq ($(NO_WARMUP), 1)
  1338. CCOMMON_OPT += -DNO_WARMUP
  1339. endif
  1340. ifeq ($(CONSISTENT_FPCSR), 1)
  1341. CCOMMON_OPT += -DCONSISTENT_FPCSR
  1342. endif
  1343. # Only for development
  1344. # CCOMMON_OPT += -DPARAMTEST
  1345. # CCOMMON_OPT += -DPREFETCHTEST
  1346. # CCOMMON_OPT += -DNO_SWITCHING
  1347. # USE_PAPI = 1
  1348. ifdef USE_PAPI
  1349. CCOMMON_OPT += -DUSE_PAPI
  1350. EXTRALIB += -lpapi -lperfctr
  1351. endif
  1352. ifdef BUFFERSIZE
  1353. CCOMMON_OPT += -DBUFFERSIZE=$(BUFFERSIZE)
  1354. endif
  1355. ifdef DYNAMIC_THREADS
  1356. CCOMMON_OPT += -DDYNAMIC_THREADS
  1357. endif
  1358. CCOMMON_OPT += -DMAX_CPU_NUMBER=$(NUM_THREADS)
  1359. CCOMMON_OPT += -DMAX_PARALLEL_NUMBER=$(NUM_PARALLEL)
  1360. ifdef USE_SIMPLE_THREADED_LEVEL3
  1361. CCOMMON_OPT += -DUSE_SIMPLE_THREADED_LEVEL3
  1362. endif
  1363. ifeq ($(USE_TLS), 1)
  1364. CCOMMON_OPT += -DUSE_TLS
  1365. endif
  1366. ifeq ($(BUILD_BFLOAT16), 1)
  1367. CCOMMON_OPT += -DBUILD_BFLOAT16
  1368. endif
  1369. ifeq ($(BUILD_HFLOAT16), 1)
  1370. CCOMMON_OPT += -DBUILD_HFLOAT16
  1371. endif
  1372. ifeq ($(BUILD_SINGLE), 1)
  1373. CCOMMON_OPT += -DBUILD_SINGLE=1
  1374. endif
  1375. ifeq ($(BUILD_DOUBLE), 1)
  1376. CCOMMON_OPT += -DBUILD_DOUBLE=1
  1377. endif
  1378. ifeq ($(BUILD_COMPLEX), 1)
  1379. CCOMMON_OPT += -DBUILD_COMPLEX=1
  1380. endif
  1381. ifeq ($(BUILD_COMPLEX16), 1)
  1382. CCOMMON_OPT += -DBUILD_COMPLEX16=1
  1383. endif
  1384. CCOMMON_OPT += -DVERSION=\"$(VERSION)\"
  1385. ifndef SYMBOLPREFIX
  1386. SYMBOLPREFIX =
  1387. endif
  1388. ifndef SYMBOLSUFFIX
  1389. SYMBOLSUFFIX =
  1390. endif
  1391. ifndef LIBSONAMEBASE
  1392. LIBSONAMEBASE = openblas
  1393. endif
  1394. ifndef LIBNAMEPREFIX
  1395. LIBNAMEPREFIX =
  1396. endif
  1397. SYMPREFIX=$(SYMBOLPREFIX)
  1398. ifeq ($(SYMBOLPREFIX),$(LIBNAMEPREFIX))
  1399. SYMPREFIX=
  1400. endif
  1401. SYMSUFFIX=$(SYMBOLSUFFIX)
  1402. ifeq ($(SYMBOLSUFFIX),$(LIBNAMESUFFIX))
  1403. SYMSUFFIX=
  1404. endif
  1405. ifndef LIBNAMESUFFIX
  1406. LIBNAMEBASE = $(SYMPREFIX)$(LIBSONAMEBASE)$(SYMSUFFIX)
  1407. else
  1408. LIBNAMEBASE = $(SYMPREFIX)$(LIBSONAMEBASE)$(SYMSUFFIX)$(LIBNAMESUFFIX)
  1409. endif
  1410. ifeq ($(OSNAME), CYGWIN_NT)
  1411. LIBPREFIX = cyg$(LIBNAMEPREFIX)$(LIBNAMEBASE)
  1412. else
  1413. LIBPREFIX = lib$(LIBNAMEPREFIX)$(LIBNAMEBASE)
  1414. endif
  1415. KERNELDIR = $(TOPDIR)/kernel/$(ARCH)
  1416. include $(TOPDIR)/Makefile.$(ARCH)
  1417. ifneq ($(C_COMPILER), PGI)
  1418. ifneq ($(C_COMPILER), SUN)
  1419. CCOMMON_OPT += -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME
  1420. endif
  1421. endif
  1422. CCOMMON_OPT += -DASMNAME=$(FU)$(*F) -DASMFNAME=$(FU)$(*F)$(BU) -DNAME=$(*F)$(BU) -DCNAME=$(*F) -DCHAR_NAME=\"$(*F)$(BU)\" -DCHAR_CNAME=\"$(*F)\"
  1423. ifeq ($(CORE), PPC440)
  1424. CCOMMON_OPT += -DALLOC_QALLOC
  1425. endif
  1426. ifeq ($(CORE), PPC440FP2)
  1427. STATIC_ALLOCATION = 1
  1428. endif
  1429. ifneq ($(OSNAME), Linux)
  1430. NO_AFFINITY = 1
  1431. endif
  1432. ifneq ($(ARCH), x86_64)
  1433. ifneq ($(ARCH), x86)
  1434. NO_AFFINITY = 1
  1435. endif
  1436. endif
  1437. ifeq ($(ARCH), POWER)
  1438. ifeq ($(DEBUG), 1)
  1439. CCOMMON_OPT := $(filter-out -O%, $(CCOMMON_OPT)) -O0
  1440. FCOMMON_OPT := $(filter-out -O%, $(FCOMMON_OPT)) -O0
  1441. endif
  1442. endif
  1443. ifdef NO_AFFINITY
  1444. ifeq ($(NO_AFFINITY), 0)
  1445. override undefine NO_AFFINITY
  1446. else
  1447. CCOMMON_OPT += -DNO_AFFINITY
  1448. endif
  1449. endif
  1450. ifdef FUNCTION_PROFILE
  1451. CCOMMON_OPT += -DFUNCTION_PROFILE
  1452. endif
  1453. ifdef SHMEM_ALLOCATION
  1454. ifneq ($(SHMEM_ALLOCATION), 0)
  1455. CCOMMON_OPT += -DALLOC_SHM
  1456. endif
  1457. endif
  1458. ifdef HUGETLB_ALLOCATION
  1459. ifneq ($(HUGETLB_ALLOCATION), 0)
  1460. CCOMMON_OPT += -DALLOC_HUGETLB
  1461. endif
  1462. endif
  1463. ifdef HUGETLBFILE_ALLOCATION
  1464. ifneq ($(HUGETLBFILE_ALLOCATION), 0)
  1465. CCOMMON_OPT += -DALLOC_HUGETLBFILE -DHUGETLB_FILE_NAME=$(HUGETLBFILE_ALLOCATION)
  1466. endif
  1467. endif
  1468. ifdef STATIC_ALLOCATION
  1469. CCOMMON_OPT += -DALLOC_STATIC
  1470. endif
  1471. ifdef DEVICEDRIVER_ALLOCATION
  1472. CCOMMON_OPT += -DALLOC_DEVICEDRIVER -DDEVICEDRIVER_NAME=\"/dev/mapper\"
  1473. endif
  1474. ifdef MIXED_MEMORY_ALLOCATION
  1475. CCOMMON_OPT += -DMIXED_MEMORY_ALLOCATION
  1476. endif
  1477. ifeq ($(OSNAME), SunOS)
  1478. TAR = gtar
  1479. PATCH = gpatch
  1480. GREP = ggrep
  1481. AWK = nawk
  1482. else
  1483. TAR = tar
  1484. PATCH = patch
  1485. GREP = grep
  1486. AWK = awk
  1487. endif
  1488. ifndef MD5SUM
  1489. MD5SUM = md5sum
  1490. endif
  1491. REVISION = -r$(VERSION)
  1492. MAJOR_VERSION = $(word 1,$(subst ., ,$(VERSION)))
  1493. ifeq ($(DEBUG), 1)
  1494. COMMON_OPT += -g
  1495. endif
  1496. ifeq ($(DEBUG), 1)
  1497. FCOMMON_OPT += -g
  1498. endif
  1499. ifndef COMMON_OPT
  1500. COMMON_OPT = -O2
  1501. endif
  1502. ifndef FCOMMON_OPT
  1503. FCOMMON_OPT = -O2 -frecursive
  1504. endif
  1505. override CFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
  1506. override PFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
  1507. override FFLAGS += $(COMMON_OPT) $(FCOMMON_OPT)
  1508. override FPFLAGS += $(FCOMMON_OPT) $(COMMON_PROF)
  1509. #MAKEOVERRIDES =
  1510. ifeq ($(NEED_PIC), 1)
  1511. ifeq (,$(findstring PIC,$(FFLAGS)))
  1512. ifneq ($(F_COMPILER),IBM)
  1513. override FFLAGS += -fPIC
  1514. endif
  1515. endif
  1516. endif
  1517. #For LAPACK Fortran codes.
  1518. #Disable -fopenmp for LAPACK Fortran codes on Windows.
  1519. ifdef OS_WINDOWS
  1520. LAPACK_FFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FFLAGS))
  1521. LAPACK_FPFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FPFLAGS))
  1522. else
  1523. LAPACK_FFLAGS := $(FFLAGS)
  1524. LAPACK_FPFLAGS := $(FPFLAGS)
  1525. endif
  1526. ifeq ($(F_COMPILER),NAG)
  1527. LAPACK_FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1528. override FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1529. endif
  1530. ifeq ($(F_COMPILER),CRAY)
  1531. LAPACK_FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1532. override FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1533. endif
  1534. ifeq ($(F_COMPILER),FLANGNEW)
  1535. LAPACK_FFLAGS := $(filter-out -m32 -m64 -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 -mtune=% -mabi=% ,$(FFLAGS))
  1536. override FFLAGS := $(filter-out -m32 -m64 -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 -mtune=% -mabi=% ,$(FFLAGS))
  1537. endif
  1538. LAPACK_CFLAGS = $(CFLAGS)
  1539. LAPACK_CFLAGS += -DHAVE_LAPACK_CONFIG_H
  1540. ifdef INTERFACE64
  1541. ifneq ($(INTERFACE64), 0)
  1542. LAPACK_CFLAGS += -DLAPACK_ILP64
  1543. endif
  1544. endif
  1545. ifdef OS_WINDOWS
  1546. LAPACK_CFLAGS += -DOPENBLAS_OS_WINDOWS
  1547. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1548. endif
  1549. ifeq ($(C_COMPILER), LSB)
  1550. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1551. endif
  1552. ifndef SUFFIX
  1553. SUFFIX = o
  1554. endif
  1555. ifndef PSUFFIX
  1556. PSUFFIX = po
  1557. endif
  1558. ifndef LIBSUFFIX
  1559. LIBSUFFIX = a
  1560. endif
  1561. ifneq ($(DYNAMIC_ARCH), 1)
  1562. ifndef SMP
  1563. LIBNAME = $(LIBPREFIX)_$(LIBCORE)$(REVISION).$(LIBSUFFIX)
  1564. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)$(REVISION)_p.$(LIBSUFFIX)
  1565. else
  1566. LIBNAME = $(LIBPREFIX)_$(LIBCORE)p$(REVISION).$(LIBSUFFIX)
  1567. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)p$(REVISION)_p.$(LIBSUFFIX)
  1568. endif
  1569. else
  1570. ifndef SMP
  1571. LIBNAME = $(LIBPREFIX)$(REVISION).$(LIBSUFFIX)
  1572. LIBNAME_P = $(LIBPREFIX)$(REVISION)_p.$(LIBSUFFIX)
  1573. else
  1574. LIBNAME = $(LIBPREFIX)p$(REVISION).$(LIBSUFFIX)
  1575. LIBNAME_P = $(LIBPREFIX)p$(REVISION)_p.$(LIBSUFFIX)
  1576. endif
  1577. endif
  1578. ifeq ($(FIXED_LIBNAME),1)
  1579. LIBNAME = lib$(LIBNAMEPREFIX)$(LIBSONAMEBASE)$(LIBNAMESUFFIX).$(LIBSUFFIX)
  1580. LIBNAME_P = lib$(LIBNAMEPREFIX)$(LISOBNAMEBASE)$(LIBNAMESUFFIX)_p.$(LIBSUFFIX)
  1581. endif
  1582. LIBDLLNAME = $(LIBPREFIX).dll
  1583. IMPLIBNAME = lib$(LIBNAMEBASE).dll.a
  1584. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.so)
  1585. LIBDYNNAME = $(LIBNAME:.$(LIBSUFFIX)=.dylib)
  1586. LIBDEFNAME = $(LIBNAME:.$(LIBSUFFIX)=.def)
  1587. LIBEXPNAME = $(LIBNAME:.$(LIBSUFFIX)=.exp)
  1588. LIBZIPNAME = $(LIBNAME:.$(LIBSUFFIX)=.zip)
  1589. LIBS = $(TOPDIR)/$(LIBNAME)
  1590. LIBS_P = $(TOPDIR)/$(LIBNAME_P)
  1591. LIB_COMPONENTS = BLAS
  1592. ifneq ($(NO_CBLAS), 1)
  1593. LIB_COMPONENTS += CBLAS
  1594. endif
  1595. ifneq ($(NO_LAPACK), 1)
  1596. LIB_COMPONENTS += LAPACK
  1597. ifneq ($(NO_LAPACKE), 1)
  1598. LIB_COMPONENTS += LAPACKE
  1599. endif
  1600. ifeq ($(BUILD_RELAPACK), 1)
  1601. LIB_COMPONENTS += ReLAPACK
  1602. endif
  1603. endif
  1604. ifeq ($(ONLY_CBLAS), 1)
  1605. LIB_COMPONENTS = CBLAS
  1606. endif
  1607. export OSNAME
  1608. export ARCH
  1609. export CORE
  1610. export LIBCORE
  1611. export __BYTE_ORDER__
  1612. export ELF_VERSION
  1613. export PGCPATH
  1614. export CONFIG
  1615. export CC
  1616. export FC
  1617. export BU
  1618. export FU
  1619. export NEED2UNDERSCORES
  1620. export USE_THREAD
  1621. export NUM_THREADS
  1622. export NUM_CORES
  1623. export SMP
  1624. export MAKEFILE_RULE
  1625. export NEED_PIC
  1626. export BINARY
  1627. export BINARY32
  1628. export BINARY64
  1629. export F_COMPILER
  1630. export C_COMPILER
  1631. export USE_OPENMP
  1632. export CROSS
  1633. export CROSS_SUFFIX
  1634. export NOFORTRAN
  1635. export C_LAPACK
  1636. export NO_FBLAS
  1637. export EXTRALIB
  1638. export CEXTRALIB
  1639. export FEXTRALIB
  1640. export HAVE_SSE
  1641. export HAVE_SSE2
  1642. export HAVE_SSE3
  1643. export HAVE_SSSE3
  1644. export HAVE_SSE4_1
  1645. export HAVE_SSE4_2
  1646. export HAVE_SSE4A
  1647. export HAVE_SSE5
  1648. export HAVE_AVX
  1649. export HAVE_AVX2
  1650. export HAVE_FMA3
  1651. export HAVE_VFP
  1652. export HAVE_VFPV3
  1653. export HAVE_VFPV4
  1654. export HAVE_NEON
  1655. ifndef NO_MSA
  1656. export HAVE_MSA
  1657. export MSA_FLAGS
  1658. endif
  1659. export KERNELDIR
  1660. export FUNCTION_PROFILE
  1661. export TARGET_CORE
  1662. export NO_AVX512
  1663. export NO_AVX2
  1664. export BUILD_BFLOAT16
  1665. export BUILD_HFLOAT16
  1666. export NO_LSX
  1667. export NO_LASX
  1668. export SBGEMM_UNROLL_M
  1669. export SBGEMM_UNROLL_N
  1670. export SHGEMM_UNROLL_M
  1671. export SHGEMM_UNROLL_N
  1672. export SGEMM_UNROLL_M
  1673. export SGEMM_UNROLL_N
  1674. export DGEMM_UNROLL_M
  1675. export DGEMM_UNROLL_N
  1676. export QGEMM_UNROLL_M
  1677. export QGEMM_UNROLL_N
  1678. export CGEMM_UNROLL_M
  1679. export CGEMM_UNROLL_N
  1680. export ZGEMM_UNROLL_M
  1681. export ZGEMM_UNROLL_N
  1682. export XGEMM_UNROLL_M
  1683. export XGEMM_UNROLL_N
  1684. export CGEMM3M_UNROLL_M
  1685. export CGEMM3M_UNROLL_N
  1686. export ZGEMM3M_UNROLL_M
  1687. export ZGEMM3M_UNROLL_N
  1688. export XGEMM3M_UNROLL_M
  1689. export XGEMM3M_UNROLL_N
  1690. ifdef USE_CUDA
  1691. export CUDADIR
  1692. export CUCC
  1693. export CUFLAGS
  1694. export CULIB
  1695. endif
  1696. .SUFFIXES: .$(PSUFFIX) .$(SUFFIX) .f
  1697. .f.$(SUFFIX):
  1698. $(FC) $(FFLAGS) -c $< -o $(@F)
  1699. .f.$(PSUFFIX):
  1700. $(FC) $(FPFLAGS) -pg -c $< -o $(@F)
  1701. ifdef BINARY64
  1702. PATHSCALEPATH = /opt/pathscale/lib/3.1
  1703. PGIPATH = /opt/pgi/linux86-64/7.1-5/lib
  1704. else
  1705. PATHSCALEPATH = /opt/pathscale/lib/3.1/32
  1706. PGIPATH = /opt/pgi/linux86/7.1-5/lib
  1707. endif
  1708. ACMLPATH = /opt/acml/4.3.0
  1709. ifneq ($(OSNAME), Darwin)
  1710. MKLPATH = /opt/intel/mkl/10.2.2.025/lib
  1711. else
  1712. MKLPATH = /Library/Frameworks/Intel_MKL.framework/Versions/10.0.1.014/lib
  1713. endif
  1714. ATLASPATH = /opt/atlas/3.9.17/opteron
  1715. FLAMEPATH = $(HOME)/flame/lib
  1716. ifneq ($(OSNAME), SunOS)
  1717. SUNPATH = /opt/sunstudio12.1
  1718. else
  1719. SUNPATH = /opt/SUNWspro
  1720. endif