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
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942943944945946947948949950951952953954955956957958959960961962963964965966967968969970971972973974975976977978979980981982983984985986987988989990991992993994995996997998999100010011002100310041005100610071008100910101011101210131014101510161017101810191020102110221023102410251026102710281029103010311032103310341035103610371038103910401041104210431044104510461047104810491050105110521053105410551056105710581059106010611062106310641065106610671068106910701071107210731074107510761077107810791080108110821083108410851086108710881089109010911092109310941095109610971098109911001101110211031104110511061107110811091110111111121113111411151116111711181119112011211122112311241125112611271128112911301131113211331134113511361137113811391140114111421143114411451146114711481149115011511152115311541155115611571158115911601161116211631164116511661167116811691170117111721173117411751176117711781179118011811182118311841185118611871188118911901191119211931194119511961197119811991200120112021203120412051206120712081209121012111212121312141215121612171218121912201221122212231224122512261227122812291230123112321233123412351236123712381239124012411242124312441245124612471248124912501251125212531254125512561257125812591260126112621263126412651266126712681269127012711272127312741275127612771278127912801281128212831284128512861287128812891290129112921293129412951296129712981299130013011302130313041305130613071308130913101311131213131314131513161317131813191320132113221323132413251326132713281329133013311332133313341335133613371338133913401341134213431344134513461347134813491350135113521353135413551356135713581359136013611362136313641365136613671368136913701371137213731374137513761377137813791380138113821383138413851386138713881389139013911392139313941395139613971398139914001401140214031404140514061407140814091410141114121413141414151416141714181419142014211422142314241425142614271428142914301431143214331434143514361437143814391440144114421443144414451446144714481449145014511452145314541455145614571458145914601461146214631464146514661467146814691470147114721473147414751476147714781479148014811482148314841485148614871488148914901491149214931494149514961497149814991500150115021503150415051506150715081509151015111512151315141515151615171518151915201521152215231524152515261527152815291530153115321533153415351536153715381539154015411542154315441545154615471548154915501551155215531554155515561557155815591560156115621563156415651566156715681569157015711572157315741575157615771578157915801581158215831584158515861587158815891590159115921593159415951596159715981599160016011602160316041605160616071608160916101611161216131614161516161617161816191620162116221623162416251626162716281629163016311632163316341635163616371638163916401641164216431644164516461647164816491650165116521653165416551656165716581659166016611662166316641665166616671668166916701671167216731674167516761677167816791680168116821683168416851686168716881689169016911692169316941695169616971698169917001701170217031704170517061707170817091710171117121713171417151716171717181719172017211722172317241725172617271728172917301731173217331734173517361737173817391740174117421743174417451746174717481749175017511752175317541755175617571758175917601761176217631764176517661767176817691770177117721773177417751776177717781779178017811782178317841785178617871788178917901791179217931794179517961797179817991800180118021803180418051806180718081809181018111812181318141815181618171818181918201821182218231824182518261827182818291830183118321833183418351836183718381839184018411842184318441845184618471848184918501851185218531854185518561857185818591860186118621863186418651866186718681869187018711872187318741875187618771878187918801881188218831884188518861887188818891890189118921893189418951896189718981899190019011902190319041905190619071908190919101911191219131914191519161917191819191920192119221923192419251926192719281929193019311932193319341935193619371938193919401941194219431944194519461947194819491950195119521953195419551956195719581959196019611962196319641965196619671968196919701971
  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. override CEXTRALIB := $(filter-out(-lto_library, $(CEXTRALIB)))
  393. endif
  394. endif
  395. ifneq (,$(findstring $(OSNAME), FreeBSD OpenBSD DragonFly))
  396. MD5SUM = md5 -r
  397. endif
  398. ifeq ($(OSNAME), NetBSD)
  399. MD5SUM = md5 -n
  400. endif
  401. ifeq ($(OSNAME), Linux)
  402. EXTRALIB += -lm
  403. #NO_EXPRECISION = 1
  404. endif
  405. ifeq ($(OSNAME), Android)
  406. EXTRALIB += -lm
  407. endif
  408. ifeq ($(OSNAME), AIX)
  409. EXTRALIB += -lm
  410. endif
  411. ifeq ($(OSNAME), $(filter $(OSNAME),FreeBSD OpenBSD NetBSD DragonFly))
  412. ifeq ($(ARCH), $(filter $(ARCH),arm arm64))
  413. EXTRALIB += -lm
  414. endif
  415. endif
  416. ifeq ($(OSNAME), WINNT)
  417. NEED_PIC = 0
  418. NO_EXPRECISION = 1
  419. EXTRALIB += -defaultlib:advapi32
  420. SUFFIX = obj
  421. PSUFFIX = pobj
  422. LIBSUFFIX = a
  423. ifeq ($(C_COMPILER), CLANG)
  424. CCOMMON_OPT += -DMS_ABI
  425. endif
  426. #Version tests for supporting specific features (MS_ABI, POWER9 intrinsics)
  427. ifeq ($(GCCVERSIONGT4), 1)
  428. # GCC Major version > 4
  429. # It is compatible with MSVC ABI.
  430. CCOMMON_OPT += -DMS_ABI
  431. endif
  432. ifeq ($(GCCVERSIONGTEQ4), 1)
  433. ifeq ($(GCCMINORVERSIONGTEQ7), 1)
  434. # GCC Version >=4.7
  435. # It is compatible with MSVC ABI.
  436. CCOMMON_OPT += -DMS_ABI
  437. endif
  438. endif
  439. # Ensure the correct stack alignment on Win32
  440. # http://permalink.gmane.org/gmane.comp.lib.openblas.general/97
  441. ifeq ($(ARCH), x86)
  442. CCOMMON_OPT += -mincoming-stack-boundary=2
  443. FCOMMON_OPT += -mincoming-stack-boundary=2
  444. endif
  445. endif
  446. ifeq ($(OSNAME), Interix)
  447. NEED_PIC = 0
  448. NO_EXPRECISION = 1
  449. INTERIX_TOOL_DIR = /opt/gcc.3.3/i586-pc-interix3/bin
  450. endif
  451. ifeq ($(OSNAME), CYGWIN_NT)
  452. NEED_PIC = 0
  453. NO_EXPRECISION = 1
  454. OS_CYGWIN_NT = 1
  455. endif
  456. ifneq ($(OSNAME), WINNT)
  457. ifneq ($(OSNAME), CYGWIN_NT)
  458. ifneq ($(OSNAME), Interix)
  459. ifneq ($(OSNAME), Android)
  460. ifdef SMP
  461. EXTRALIB += -lpthread
  462. endif
  463. endif
  464. endif
  465. endif
  466. endif
  467. # ifeq logical or
  468. ifeq ($(OSNAME), $(filter $(OSNAME),WINNT CYGWIN_NT Interix))
  469. OS_WINDOWS=1
  470. endif
  471. ifdef QUAD_PRECISION
  472. CCOMMON_OPT += -DQUAD_PRECISION
  473. NO_EXPRECISION = 1
  474. endif
  475. ifneq ($(ARCH), x86)
  476. ifneq ($(ARCH), x86_64)
  477. NO_EXPRECISION = 1
  478. endif
  479. endif
  480. ifdef UTEST_CHECK
  481. CCOMMON_OPT += -DUTEST_CHECK
  482. SANITY_CHECK = 1
  483. endif
  484. ifdef SANITY_CHECK
  485. CCOMMON_OPT += -DSANITY_CHECK -DREFNAME=$(*F)f$(BU)
  486. endif
  487. MAX_STACK_ALLOC ?= 2048
  488. ifneq ($(MAX_STACK_ALLOC), 0)
  489. CCOMMON_OPT += -DMAX_STACK_ALLOC=$(MAX_STACK_ALLOC)
  490. endif
  491. ifdef USE_LOCKING
  492. ifneq ($(USE_LOCKING), 0)
  493. CCOMMON_OPT += -DUSE_LOCKING
  494. endif
  495. endif
  496. #
  497. # Architecture dependent settings
  498. #
  499. ifeq ($(ARCH), x86)
  500. ifndef BINARY
  501. NO_BINARY_MODE = 1
  502. endif
  503. ifeq ($(CORE), generic)
  504. #NO_EXPRECISION = 1
  505. endif
  506. ifndef NO_EXPRECISION
  507. ifeq ($(F_COMPILER), GFORTRAN)
  508. # ifeq logical or. GCC or LSB
  509. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  510. EXPRECISION = 1
  511. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  512. FCOMMON_OPT += -m128bit-long-double
  513. endif
  514. ifeq ($(C_COMPILER), CLANG)
  515. EXPRECISION = 1
  516. CCOMMON_OPT += -DEXPRECISION
  517. FCOMMON_OPT += -m128bit-long-double
  518. endif
  519. endif
  520. endif
  521. endif
  522. ifeq ($(ARCH), x86_64)
  523. ifeq ($(CORE), generic)
  524. #NO_EXPRECISION = 1
  525. endif
  526. ifndef NO_EXPRECISION
  527. ifeq ($(F_COMPILER), GFORTRAN)
  528. # ifeq logical or. GCC or LSB
  529. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  530. EXPRECISION = 1
  531. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  532. FCOMMON_OPT += -m128bit-long-double
  533. endif
  534. ifeq ($(C_COMPILER), CLANG)
  535. EXPRECISION = 1
  536. CCOMMON_OPT += -DEXPRECISION
  537. FCOMMON_OPT += -m128bit-long-double
  538. endif
  539. endif
  540. endif
  541. endif
  542. ifeq ($(C_COMPILER), INTEL)
  543. CCOMMON_OPT += -wd981
  544. endif
  545. ifeq ($(USE_OPENMP), 1)
  546. #check
  547. ifeq ($(USE_THREAD), 0)
  548. $(error OpenBLAS: Cannot set both USE_OPENMP=1 and USE_THREAD=0. The USE_THREAD=0 is only for building single thread version.)
  549. endif
  550. # ifeq logical or. GCC or LSB
  551. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  552. CCOMMON_OPT += -fopenmp
  553. endif
  554. ifeq ($(C_COMPILER), CLANG)
  555. CCOMMON_OPT += -fopenmp
  556. ifeq ($(F_COMPILER), GFORTRAN)
  557. FEXTRALIB := $(subst -lgomp,-lomp,$(FEXTRALIB))
  558. endif
  559. endif
  560. ifeq ($(C_COMPILER), INTEL)
  561. CCOMMON_OPT += -fopenmp
  562. endif
  563. ifeq ($(C_COMPILER), PGI)
  564. CCOMMON_OPT += -mp
  565. endif
  566. ifeq ($(C_COMPILER), OPEN64)
  567. CCOMMON_OPT += -mp
  568. CEXTRALIB += -lstdc++
  569. endif
  570. ifeq ($(C_COMPILER), PATHSCALE)
  571. CCOMMON_OPT += -mp
  572. endif
  573. endif
  574. ifeq ($(DYNAMIC_ARCH), 1)
  575. ifeq ($(ARCH), x86)
  576. DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
  577. CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
  578. endif
  579. ifeq ($(ARCH), x86_64)
  580. DYNAMIC_CORE = PRESCOTT CORE2
  581. ifeq ($(DYNAMIC_OLDER), 1)
  582. DYNAMIC_CORE += PENRYN DUNNINGTON
  583. endif
  584. DYNAMIC_CORE += NEHALEM
  585. ifeq ($(DYNAMIC_OLDER), 1)
  586. DYNAMIC_CORE += OPTERON OPTERON_SSE3
  587. endif
  588. DYNAMIC_CORE += BARCELONA
  589. ifeq ($(DYNAMIC_OLDER), 1)
  590. DYNAMIC_CORE += BOBCAT ATOM NANO
  591. endif
  592. ifneq ($(NO_AVX), 1)
  593. DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER STEAMROLLER EXCAVATOR
  594. endif
  595. ifneq ($(NO_AVX2), 1)
  596. DYNAMIC_CORE += HASWELL ZEN
  597. endif
  598. ifneq ($(NO_AVX512), 1)
  599. ifneq ($(NO_AVX2), 1)
  600. DYNAMIC_CORE += SKYLAKEX COOPERLAKE SAPPHIRERAPIDS
  601. endif
  602. endif
  603. endif
  604. ifdef DYNAMIC_LIST
  605. override DYNAMIC_CORE = PRESCOTT $(DYNAMIC_LIST)
  606. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_PRESCOTT
  607. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  608. CCOMMON_OPT += $(XCCOMMON_OPT)
  609. #CCOMMON_OPT += -DDYNAMIC_LIST='$(DYNAMIC_LIST)'
  610. endif
  611. ifeq ($(ARCH), arm64)
  612. DYNAMIC_CORE = ARMV8
  613. DYNAMIC_CORE += CORTEXA53
  614. DYNAMIC_CORE += CORTEXA57
  615. DYNAMIC_CORE += NEOVERSEN1
  616. ifneq ($(NO_SVE), 1)
  617. DYNAMIC_CORE += NEOVERSEV1
  618. DYNAMIC_CORE += NEOVERSEN2
  619. DYNAMIC_CORE += ARMV8SVE
  620. DYNAMIC_CORE += A64FX
  621. endif
  622. ifneq ($(NO_SME), 1)
  623. DYNAMIC_CORE += ARMV9SME
  624. endif
  625. DYNAMIC_CORE += THUNDERX
  626. DYNAMIC_CORE += THUNDERX2T99
  627. DYNAMIC_CORE += TSV110
  628. DYNAMIC_CORE += EMAG8180
  629. DYNAMIC_CORE += THUNDERX3T110
  630. ifdef DYNAMIC_LIST
  631. override DYNAMIC_CORE = ARMV8 $(DYNAMIC_LIST)
  632. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_ARMV8
  633. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  634. endif
  635. endif
  636. ifeq ($(ARCH), mips64)
  637. DYNAMIC_CORE = LOONGSON3R3 LOONGSON3R4 MIPS64_GENERIC
  638. ifdef DYNAMIC_LIST
  639. override DYNAMIC_CORE = MIPS64_GENERIC $(DYNAMIC_LIST)
  640. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_MIPS64_GENERIC
  641. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  642. endif
  643. endif
  644. ifeq ($(ARCH), loongarch64)
  645. DYNAMIC_CORE = LA64_GENERIC LA264 LA464
  646. endif
  647. ifeq ($(ARCH), riscv64)
  648. DYNAMIC_CORE = RISCV64_GENERIC
  649. DYNAMIC_CORE += RISCV64_ZVL128B
  650. DYNAMIC_CORE += RISCV64_ZVL256B
  651. ifdef DYNAMIC_LIST
  652. override DYNAMIC_CORE = RISCV64_GENERIC $(DYNAMIC_LIST)
  653. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_RISCV64_GENERIC
  654. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  655. endif
  656. endif
  657. ifeq ($(ARCH), zarch)
  658. DYNAMIC_CORE = ZARCH_GENERIC
  659. # if the compiler accepts -march=arch11 or -march=z13 and can compile a file
  660. # with z13-specific inline assembly, then we can include support for Z13.
  661. # note: -march=z13 is equivalent to -march=arch11 yet some compiler releases
  662. # only support one or the other.
  663. # note: LLVM version 6.x supported -march=z13 yet could not handle vector
  664. # registers in inline assembly, so the check for supporting the -march flag is
  665. # not enough.
  666. ZARCH_TEST_COMPILE=-c $(TOPDIR)/kernel/zarch/damin_z13.c -I$(TOPDIR) -o /dev/null > /dev/null 2> /dev/null
  667. ZARCH_CC_SUPPORTS_ARCH11=$(shell $(CC) -march=arch11 $(ZARCH_TEST_COMPILE) && echo 1)
  668. ZARCH_CC_SUPPORTS_Z13=$(shell $(CC) -march=z13 $(ZARCH_TEST_COMPILE) && echo 1)
  669. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH11), $(ZARCH_CC_SUPPORTS_Z13)), 1)
  670. DYNAMIC_CORE += Z13
  671. CCOMMON_OPT += -DDYN_Z13
  672. else
  673. $(info OpenBLAS: Not building Z13 kernels because the compiler $(CC) does not support it)
  674. endif
  675. # as above for z13, check for -march=arch12 and z14 support in the compiler.
  676. ZARCH_CC_SUPPORTS_ARCH12=$(shell $(CC) -march=arch12 $(ZARCH_TEST_COMPILE) && echo 1)
  677. ZARCH_CC_SUPPORTS_Z14=$(shell $(CC) -march=z14 $(ZARCH_TEST_COMPILE) && echo 1)
  678. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH12), $(ZARCH_CC_SUPPORTS_Z14)), 1)
  679. DYNAMIC_CORE += Z14
  680. CCOMMON_OPT += -DDYN_Z14
  681. else
  682. $(info OpenBLAS: Not building Z14 kernels because the compiler $(CC) does not support it)
  683. endif
  684. endif # ARCH zarch
  685. ifeq ($(ARCH), power)
  686. ifneq ($(C_COMPILER), PGI)
  687. DYNAMIC_CORE = POWER6
  688. DYNAMIC_CORE += POWER8
  689. ifneq ($(C_COMPILER), GCC)
  690. DYNAMIC_CORE += POWER9
  691. DYNAMIC_CORE += POWER10
  692. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  693. endif
  694. ifeq ($(C_COMPILER), GCC)
  695. ifeq ($(GCCVERSIONGT5), 1)
  696. DYNAMIC_CORE += POWER9
  697. else
  698. $(info, OpenBLAS: Your gcc version is too old to build the POWER9 kernels.)
  699. endif
  700. ifeq ($(OSNAME), AIX)
  701. LDVERSIONGTEQ35 := 1
  702. else
  703. LDVERSIONGTEQ35 := $(shell expr `$(CC) -Wl,--version 2> /dev/null | head -1 | cut -f2 -d "." | cut -f1 -d "-"` \>= 35)
  704. endif
  705. ifeq ($(GCCVERSIONGTEQ11)$(LDVERSIONGTEQ35), 11)
  706. DYNAMIC_CORE += POWER10
  707. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  708. else ifeq ($(GCCVERSIONGTEQ10), 1)
  709. ifeq ($(GCCMINORVERSIONGTEQ2)$(LDVERSIONGTEQ35), 11)
  710. DYNAMIC_CORE += POWER10
  711. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  712. endif
  713. else
  714. $(info, OpenBLAS: Your gcc version is too old to build the POWER10 kernels.)
  715. endif
  716. endif
  717. else
  718. DYNAMIC_CORE = POWER8
  719. DYNAMIC_CORE += POWER9
  720. endif
  721. endif
  722. # If DYNAMIC_CORE is not set, DYNAMIC_ARCH cannot do anything, so force it to empty
  723. ifndef DYNAMIC_CORE
  724. override DYNAMIC_ARCH=
  725. endif
  726. endif
  727. ifeq ($(ARCH), ia64)
  728. NO_BINARY_MODE = 1
  729. BINARY_DEFINED = 1
  730. ifeq ($(F_COMPILER), GFORTRAN)
  731. ifeq ($(C_COMPILER), GCC)
  732. EXPRECISION = 1
  733. CCOMMON_OPT += -DEXPRECISION
  734. endif
  735. endif
  736. endif
  737. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  738. NO_BINARY_MODE = 1
  739. endif
  740. ifeq ($(ARCH), alpha)
  741. NO_BINARY_MODE = 1
  742. BINARY_DEFINED = 1
  743. endif
  744. ifeq ($(ARCH), arm)
  745. NO_BINARY_MODE = 1
  746. BINARY_DEFINED = 1
  747. ifneq ($(EMBEDDED), 1)
  748. CCOMMON_OPT += -marm
  749. FCOMMON_OPT += -marm
  750. else
  751. CCOMMON_OPT += -DOS_EMBEDDED -mthumb -mcpu=cortex-m4 -mfloat-abi=hard -mfpu=fpv4-sp-d16
  752. endif
  753. # If softfp abi is mentioned on the command line, force it.
  754. ifeq ($(ARM_SOFTFP_ABI), 1)
  755. CCOMMON_OPT += -mfloat-abi=softfp
  756. FCOMMON_OPT += -mfloat-abi=softfp
  757. endif
  758. ifeq ($(OSNAME), Android)
  759. ifeq ($(ARM_SOFTFP_ABI), 1)
  760. EXTRALIB += -lm
  761. else
  762. EXTRALIB += -Wl,-lm_hard
  763. endif
  764. endif
  765. endif
  766. ifeq ($(ARCH), arm64)
  767. NO_BINARY_MODE = 1
  768. BINARY_DEFINED = 1
  769. ifdef INTERFACE64
  770. ifneq ($(INTERFACE64), 0)
  771. ifeq ($(F_COMPILER), GFORTRAN)
  772. FCOMMON_OPT += -fdefault-integer-8
  773. endif
  774. ifeq ($(F_COMPILER), FLANG)
  775. FCOMMON_OPT += -i8
  776. endif
  777. endif
  778. endif
  779. endif
  780. ifeq ($(ARCH), riscv64)
  781. NO_BINARY_MODE = 1
  782. BINARY_DEFINED = 1
  783. ifdef INTERFACE64
  784. ifneq ($(INTERFACE64), 0)
  785. ifeq ($(F_COMPILER), GFORTRAN)
  786. FCOMMON_OPT += -fdefault-integer-8
  787. endif
  788. ifeq ($(F_COMPILER), FLANG)
  789. FCOMMON_OPT += -i8
  790. endif
  791. endif
  792. endif
  793. endif
  794. ifeq ($(ARCH), loongarch64)
  795. NO_BINARY_MODE = 1
  796. BINARY_DEFINED = 1
  797. ifdef INTERFACE64
  798. ifneq ($(INTERFACE64), 0)
  799. ifeq ($(F_COMPILER), GFORTRAN)
  800. FCOMMON_OPT += -fdefault-integer-8
  801. endif
  802. ifeq ($(F_COMPILER), FLANG)
  803. FCOMMON_OPT += -i8
  804. endif
  805. endif
  806. endif
  807. endif
  808. ifeq ($(ARCH), csky)
  809. NO_BINARY_MODE = 1
  810. BINARY_DEFINED = 1
  811. endif
  812. #
  813. # C Compiler dependent settings
  814. #
  815. # ifeq logical or. GCC or CLANG or LSB
  816. # http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
  817. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG LSB))
  818. CCOMMON_OPT += -Wall
  819. COMMON_PROF += -fno-inline
  820. NO_UNINITIALIZED_WARN = -Wno-uninitialized
  821. ifeq ($(QUIET_MAKE), 1)
  822. CCOMMON_OPT += $(NO_UNINITIALIZED_WARN) -Wno-unused
  823. endif
  824. ifdef NO_BINARY_MODE
  825. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  826. ifdef BINARY64
  827. CCOMMON_OPT += -mabi=64
  828. else
  829. CCOMMON_OPT += -mabi=n32
  830. endif
  831. BINARY_DEFINED = 1
  832. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  833. CCOMMON_OPT += -mabi=32
  834. BINARY_DEFINED = 1
  835. endif
  836. ifneq (, $(filter $(CORE), MIPS64_GENERIC))
  837. CCOMMON_OPT += -DNO_MSA
  838. FCOMMON_OPT += -DNO_MSA
  839. endif
  840. ifneq (, $(filter $(CORE),LOONGSON3R3 LOONGSON3R4))
  841. CCOMMON_OPT += -march=loongson3a
  842. FCOMMON_OPT += -march=loongson3a
  843. endif
  844. ifeq ($(CORE), MIPS24K)
  845. CCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  846. FCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  847. endif
  848. ifeq ($(CORE), MIPS1004K)
  849. CCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  850. FCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  851. endif
  852. ifeq ($(CORE), P5600)
  853. CCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  854. FCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  855. endif
  856. ifeq ($(CORE), I6400)
  857. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  858. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  859. endif
  860. ifeq ($(CORE), P6600)
  861. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  862. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  863. endif
  864. ifeq ($(CORE), I6500)
  865. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  866. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  867. endif
  868. ifeq ($(OSNAME), AIX)
  869. BINARY_DEFINED = 1
  870. endif
  871. ifeq ($(ARCH), loongarch64)
  872. LA64_ABI=$(shell $(CC) -mabi=lp64d -c $(TOPDIR)/cpuid_loongarch64.c -o /dev/null > /dev/null 2> /dev/null && echo lp64d)
  873. LA64_ARCH=$(shell $(CC) -march=loongarch64 -c $(TOPDIR)/cpuid_loongarch64.c -o /dev/null > /dev/null 2> /dev/null && echo loongarch64)
  874. ifneq ($(LA64_ABI), lp64d)
  875. LA64_ABI=lp64
  876. endif
  877. ifneq ($(LA64_ARCH), loongarch64)
  878. CCOMMON_OPT += -mabi=$(LA64_ABI)
  879. FCOMMON_OPT += -mabi=$(LA64_ABI)
  880. else
  881. CCOMMON_OPT += -march=loongarch64 -mabi=$(LA64_ABI)
  882. FCOMMON_OPT += -march=loongarch64 -mabi=$(LA64_ABI)
  883. endif
  884. endif
  885. endif
  886. ifndef BINARY_DEFINED
  887. ifneq ($(OSNAME), AIX)
  888. ifdef BINARY64
  889. ifneq ($(ARCH), riscv64)
  890. CCOMMON_OPT += -m64
  891. endif
  892. else
  893. CCOMMON_OPT += -m32
  894. endif
  895. endif
  896. endif
  897. endif
  898. ifeq ($(C_COMPILER), PGI)
  899. PGCVERSIONGT20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \> 20)
  900. PGCVERSIONEQ20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` == 20)
  901. PGCMINORVERSIONGE11 := $(shell expr `$(CC) --version|sed -n "2p" |cut -d "-" -f 1 |sed -e "s/[^0-9.]//g" |cut -c 4-5` \>= 11)
  902. PGCVERSIONCHECK := $(PGCVERSIONGT20)$(PGCVERSIONEQ20)$(PGCMINORVERSIONGE11)
  903. ifeq ($(PGCVERSIONCHECK), $(filter $(PGCVERSIONCHECK), 100 101 011))
  904. NEWPGI := 1
  905. PGCVERSIONGT21 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \> 21)
  906. PGCVERSIONEQ21 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` == 21)
  907. PGCVERSIONCHECK2 := $(PGCVERSIONGT21)$(PGCVERSIONEQ21)$(PGCMINORVERSIONGE11)
  908. ifeq ($(PGCVERSIONCHECK2), $(filter $(PGCVERSIONCHECK2), 100 101 011))
  909. NEWPGI2 := 1
  910. endif
  911. endif
  912. ifdef BINARY64
  913. ifeq ($(ARCH), x86_64)
  914. ifeq (,$(findstring tp,$(CFLAGS)))
  915. ifneq ($(NEWPGI2),1)
  916. CCOMMON_OPT += -tp p7-64
  917. else
  918. CCOMMON_OPT += -tp px
  919. endif
  920. endif
  921. ifneq ($(NEWPGI),1)
  922. CCOMMON_OPT += -D__MMX__ -Mnollvm
  923. endif
  924. else
  925. ifeq ($(ARCH), power)
  926. ifeq (,$(findstring tp,$(CFLAGS)))
  927. ifeq ($(CORE), POWER8)
  928. CCOMMON_OPT += -tp pwr8
  929. endif
  930. ifeq ($(CORE), POWER9)
  931. CCOMMON_OPT += -tp pwr9
  932. endif
  933. endif
  934. endif
  935. endif
  936. else
  937. ifneq ($(NEWPGI2),1)
  938. ifeq (,$(findstring tp,$(CFLAGS)))
  939. CCOMMON_OPT += -tp p7
  940. else
  941. CCOMMON_OPT += -tp px
  942. endif
  943. endif
  944. endif
  945. endif
  946. ifeq ($(C_COMPILER), PATHSCALE)
  947. ifdef BINARY64
  948. CCOMMON_OPT += -m64
  949. else
  950. CCOMMON_OPT += -m32
  951. endif
  952. endif
  953. #
  954. # Fortran Compiler dependent settings
  955. #
  956. ifeq ($(F_COMPILER), NAG)
  957. FCOMMON_OPT += -dcfuns -recursive -ieee=full -w=obs -thread_safe
  958. ifdef INTERFACE64
  959. ifneq ($(INTERFACE64), 0)
  960. FCOMMON_OPT += -i8
  961. endif
  962. endif
  963. ifeq ($(USE_OPENMP), 1)
  964. FCOMMON_OPT += -openmp
  965. endif
  966. endif
  967. ifeq ($(F_COMPILER), FLANG)
  968. CCOMMON_OPT += -DF_INTERFACE_FLANG
  969. FCOMMON_OPT += -Mrecursive -Kieee
  970. ifeq ($(OSNAME), Linux)
  971. ifeq ($(ARCH), x86_64)
  972. FLANG_VENDOR := $(shell $(FC) --version|head -1 |cut -f 1 -d " ")
  973. ifeq ($(FLANG_VENDOR), AMD)
  974. FCOMMON_OPT += -fno-unroll-loops
  975. endif
  976. endif
  977. endif
  978. ifdef BINARY64
  979. ifdef INTERFACE64
  980. ifneq ($(INTERFACE64), 0)
  981. FCOMMON_OPT += -i8
  982. endif
  983. endif
  984. FCOMMON_OPT += -Wall
  985. else
  986. FCOMMON_OPT += -Wall
  987. endif
  988. ifeq ($(USE_OPENMP), 1)
  989. FCOMMON_OPT += -fopenmp
  990. endif
  991. endif
  992. ifeq ($(F_COMPILER), G77)
  993. CCOMMON_OPT += -DF_INTERFACE_G77
  994. FCOMMON_OPT += -Wall
  995. ifndef NO_BINARY_MODE
  996. ifneq ($(OSNAME), AIX)
  997. ifdef BINARY64
  998. FCOMMON_OPT += -m64
  999. else
  1000. FCOMMON_OPT += -m32
  1001. endif
  1002. endif
  1003. endif
  1004. endif
  1005. ifeq ($(F_COMPILER), G95)
  1006. CCOMMON_OPT += -DF_INTERFACE_G95
  1007. FCOMMON_OPT += -Wall
  1008. ifneq ($(OSNAME), AIX)
  1009. ifndef NO_BINARY_MODE
  1010. ifdef BINARY64
  1011. FCOMMON_OPT += -m64
  1012. else
  1013. FCOMMON_OPT += -m32
  1014. endif
  1015. endif
  1016. ifneq ($(NO_LAPACKE), 1)
  1017. FCOMMON_OPT += -fno-second-underscore
  1018. endif
  1019. endif
  1020. endif
  1021. ifeq ($(F_COMPILER), $(filter $(F_COMPILER),GFORTRAN FLANGNEW))
  1022. CCOMMON_OPT += -DF_INTERFACE_GFORT
  1023. ifeq ($(F_COMPILER), GFORTRAN)
  1024. FCOMMON_OPT += -Wall
  1025. # make single-threaded LAPACK calls thread-safe #1847
  1026. FCOMMON_OPT += -frecursive
  1027. # work around ABI problem with passing single-character arguments
  1028. FCOMMON_OPT += -fno-optimize-sibling-calls
  1029. #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
  1030. ifneq ($(NOFORTRAN), 1)
  1031. ifneq ($(NOFORTRAN), 2)
  1032. ifneq ($(NO_LAPACK), 1)
  1033. EXTRALIB += -lgfortran
  1034. endif
  1035. endif
  1036. endif
  1037. endif
  1038. ifdef NO_BINARY_MODE
  1039. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  1040. ifdef BINARY64
  1041. FCOMMON_OPT += -mabi=64
  1042. else
  1043. FCOMMON_OPT += -mabi=n32
  1044. endif
  1045. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  1046. FCOMMON_OPT += -mabi=32
  1047. endif
  1048. ifeq ($(ARCH), $(filter $(ARCH),loongarch64))
  1049. ifdef INTERFACE64
  1050. ifneq ($(INTERFACE64), 0)
  1051. FCOMMON_OPT += -fdefault-integer-8
  1052. endif
  1053. endif
  1054. endif
  1055. else
  1056. ifdef BINARY64
  1057. ifneq ($(OSNAME), AIX)
  1058. ifneq ($(ARCH), riscv64)
  1059. FCOMMON_OPT += -m64
  1060. endif
  1061. endif
  1062. ifdef INTERFACE64
  1063. ifneq ($(INTERFACE64), 0)
  1064. FCOMMON_OPT += -fdefault-integer-8
  1065. endif
  1066. endif
  1067. else
  1068. ifneq ($(OSNAME), AIX)
  1069. FCOMMON_OPT += -m32
  1070. endif
  1071. endif
  1072. endif
  1073. ifeq ($(USE_OPENMP), 1)
  1074. FCOMMON_OPT += -fopenmp
  1075. endif
  1076. endif
  1077. ifeq ($(F_COMPILER), INTEL)
  1078. CCOMMON_OPT += -DF_INTERFACE_INTEL
  1079. ifdef INTERFACE64
  1080. ifneq ($(INTERFACE64), 0)
  1081. FCOMMON_OPT += -i8
  1082. endif
  1083. endif
  1084. FCOMMON_OPT += -recursive -fp-model strict -assume protect-parens
  1085. ifeq ($(USE_OPENMP), 1)
  1086. FCOMMON_OPT += -fopenmp
  1087. endif
  1088. endif
  1089. ifeq ($(F_COMPILER), FUJITSU)
  1090. CCOMMON_OPT += -DF_INTERFACE_FUJITSU
  1091. ifeq ($(USE_OPENMP), 1)
  1092. FCOMMON_OPT += -openmp
  1093. endif
  1094. endif
  1095. ifeq ($(F_COMPILER), IBM)
  1096. CCOMMON_OPT += -DF_INTERFACE_IBM
  1097. FEXTRALIB += -lxlf90
  1098. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG))
  1099. FCOMMON_OPT += -qextname -qzerosize
  1100. endif
  1101. # FCOMMON_OPT += -qarch=440
  1102. ifdef BINARY64
  1103. FCOMMON_OPT += -q64
  1104. ifdef INTERFACE64
  1105. ifneq ($(INTERFACE64), 0)
  1106. FCOMMON_OPT += -qintsize=8
  1107. endif
  1108. endif
  1109. else
  1110. FCOMMON_OPT += -q32
  1111. endif
  1112. endif
  1113. ifeq ($(F_COMPILER), PGI)
  1114. CCOMMON_OPT += -DF_INTERFACE_PGI
  1115. COMMON_PROF += -DPGICOMPILER
  1116. ifdef BINARY64
  1117. ifdef INTERFACE64
  1118. ifneq ($(INTERFACE64), 0)
  1119. FCOMMON_OPT += -i8
  1120. endif
  1121. endif
  1122. ifeq ($(ARCH), x86_64)
  1123. ifneq ($(NEWPGI2),1)
  1124. FCOMMON_OPT += -tp p7-64
  1125. else
  1126. FCOMMON_OPT += -tp px
  1127. endif
  1128. else
  1129. ifeq ($(ARCH), power)
  1130. ifeq ($(CORE), POWER6)
  1131. $(warning NVIDIA HPC compilers do not support POWER6.)
  1132. endif
  1133. ifeq ($(CORE), POWER8)
  1134. FCOMMON_OPT += -tp pwr8
  1135. endif
  1136. ifeq ($(CORE), POWER9)
  1137. FCOMMON_OPT += -tp pwr9
  1138. endif
  1139. ifeq ($(CORE), POWER10)
  1140. $(warning NVIDIA HPC compilers do not support POWER10.)
  1141. endif
  1142. endif
  1143. endif
  1144. else
  1145. FCOMMON_OPT += -tp p7
  1146. endif
  1147. FCOMMON_OPT += -Mrecursive -Kieee
  1148. ifeq ($(USE_OPENMP), 1)
  1149. FCOMMON_OPT += -mp
  1150. endif
  1151. endif
  1152. ifeq ($(F_COMPILER), PATHSCALE)
  1153. CCOMMON_OPT += -DF_INTERFACE_PATHSCALE
  1154. ifdef BINARY64
  1155. ifdef INTERFACE64
  1156. ifneq ($(INTERFACE64), 0)
  1157. FCOMMON_OPT += -i8
  1158. endif
  1159. endif
  1160. endif
  1161. ifeq ($(USE_OPENMP), 1)
  1162. FCOMMON_OPT += -mp
  1163. endif
  1164. endif
  1165. ifeq ($(F_COMPILER), OPEN64)
  1166. CCOMMON_OPT += -DF_INTERFACE_OPEN64
  1167. ifdef BINARY64
  1168. ifdef INTERFACE64
  1169. ifneq ($(INTERFACE64), 0)
  1170. FCOMMON_OPT += -i8
  1171. endif
  1172. endif
  1173. endif
  1174. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  1175. ifndef BINARY64
  1176. FCOMMON_OPT += -n32
  1177. else
  1178. FCOMMON_OPT += -n64
  1179. endif
  1180. ifeq ($(CORE), LOONGSON3R3)
  1181. FCOMMON_OPT += -loongson3 -static
  1182. endif
  1183. ifeq ($(CORE), LOONGSON3R4)
  1184. FCOMMON_OPT += -loongson3 -static
  1185. endif
  1186. else
  1187. ifndef BINARY64
  1188. FCOMMON_OPT += -m32
  1189. else
  1190. FCOMMON_OPT += -m64
  1191. endif
  1192. endif
  1193. ifeq ($(USE_OPENMP), 1)
  1194. FEXTRALIB += -lstdc++
  1195. FCOMMON_OPT += -mp
  1196. endif
  1197. endif
  1198. ifeq ($(C_COMPILER), OPEN64)
  1199. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  1200. ifndef BINARY64
  1201. CCOMMON_OPT += -n32
  1202. else
  1203. CCOMMON_OPT += -n64
  1204. endif
  1205. ifeq ($(CORE), LOONGSON3R3)
  1206. CCOMMON_OPT += -loongson3 -static
  1207. endif
  1208. ifeq ($(CORE), LOONGSON3R4)
  1209. CCOMMON_OPT += -loongson3 -static
  1210. endif
  1211. else
  1212. ifndef BINARY64
  1213. CCOMMON_OPT += -m32
  1214. else
  1215. CCOMMON_OPT += -m64
  1216. endif
  1217. endif
  1218. endif
  1219. ifeq ($(C_COMPILER), SUN)
  1220. CCOMMON_OPT += -w
  1221. ifeq ($(ARCH), x86)
  1222. CCOMMON_OPT += -m32
  1223. else
  1224. ifdef BINARY64
  1225. CCOMMON_OPT += -m64
  1226. else
  1227. CCOMMON_OPT += -m32
  1228. endif
  1229. endif
  1230. endif
  1231. ifeq ($(F_COMPILER), SUN)
  1232. CCOMMON_OPT += -DF_INTERFACE_SUN
  1233. FCOMMON_OPT += -ftrap=%none -xrecursive
  1234. ifeq ($(ARCH), x86)
  1235. FCOMMON_OPT += -m32
  1236. else
  1237. ifdef BINARY64
  1238. FCOMMON_OPT += -m64
  1239. else
  1240. FCOMMON_OPT += -m32
  1241. endif
  1242. endif
  1243. ifeq ($(USE_OPENMP), 1)
  1244. FCOMMON_OPT += -xopenmp=parallel
  1245. endif
  1246. endif
  1247. ifeq ($(F_COMPILER), COMPAQ)
  1248. CCOMMON_OPT += -DF_INTERFACE_COMPAQ
  1249. ifeq ($(USE_OPENMP), 1)
  1250. FCOMMON_OPT += -openmp
  1251. endif
  1252. endif
  1253. ifeq ($(F_COMPILER), CRAY)
  1254. CCOMMON_OPT += -DF_INTERFACE_INTEL
  1255. FCOMMON_OPT += -hnopattern
  1256. ifdef INTERFACE64
  1257. ifneq ($(INTERFACE64), 0)
  1258. FCOMMON_OPT += -s integer64
  1259. endif
  1260. endif
  1261. ifneq ($(USE_OPENMP), 1)
  1262. FCOMMON_OPT += -O noomp
  1263. endif
  1264. endif
  1265. ifdef BINARY64
  1266. ifdef INTERFACE64
  1267. ifneq ($(INTERFACE64), 0)
  1268. CCOMMON_OPT +=
  1269. #-DUSE64BITINT
  1270. endif
  1271. endif
  1272. endif
  1273. ifeq ($(NEED_PIC), 1)
  1274. ifeq ($(C_COMPILER), IBM)
  1275. CCOMMON_OPT += -qpic=large
  1276. else
  1277. CCOMMON_OPT += -fPIC
  1278. endif
  1279. ifeq ($(F_COMPILER), SUN)
  1280. FCOMMON_OPT += -pic
  1281. else ifeq ($(F_COMPILER), NAG)
  1282. FCOMMON_OPT += -PIC
  1283. else ifeq ($(F_COMPILER), IBM)
  1284. FCOMMON_OPT += -qpic=large
  1285. else
  1286. FCOMMON_OPT += -fPIC
  1287. endif
  1288. endif
  1289. ifeq ($(DYNAMIC_ARCH), 1)
  1290. CCOMMON_OPT += -DDYNAMIC_ARCH
  1291. endif
  1292. ifeq ($(DYNAMIC_OLDER), 1)
  1293. CCOMMON_OPT += -DDYNAMIC_OLDER
  1294. endif
  1295. ifeq ($(C_LAPACK), 1)
  1296. CCOMMON_OPT += -DC_LAPACK
  1297. endif
  1298. ifeq ($(NO_LAPACK), 1)
  1299. CCOMMON_OPT += -DNO_LAPACK
  1300. #Disable LAPACK C interface
  1301. NO_LAPACKE = 1
  1302. endif
  1303. ifeq ($(NO_LAPACKE), 1)
  1304. CCOMMON_OPT += -DNO_LAPACKE
  1305. endif
  1306. ifeq ($(NO_AVX), 1)
  1307. CCOMMON_OPT += -DNO_AVX
  1308. endif
  1309. ifeq ($(ARCH), x86)
  1310. CCOMMON_OPT += -DNO_AVX
  1311. endif
  1312. ifeq ($(NO_AVX2), 1)
  1313. CCOMMON_OPT += -DNO_AVX2
  1314. endif
  1315. ifeq ($(NO_AVX512), 1)
  1316. CCOMMON_OPT += -DNO_AVX512
  1317. endif
  1318. ifeq ($(NO_SVE), 1)
  1319. CCOMMON_OPT += -DNO_SVE
  1320. endif
  1321. ifeq ($(NO_SME), 1)
  1322. CCOMMON_OPT += -DNO_SME
  1323. endif
  1324. ifdef SMP
  1325. CCOMMON_OPT += -DSMP_SERVER
  1326. ifeq ($(ARCH), mips64)
  1327. USE_SIMPLE_THREADED_LEVEL3 = 1
  1328. endif
  1329. ifeq ($(USE_OPENMP), 1)
  1330. # USE_SIMPLE_THREADED_LEVEL3 = 1
  1331. # NO_AFFINITY = 1
  1332. CCOMMON_OPT += -DUSE_OPENMP
  1333. endif
  1334. ifeq ($(BIGNUMA), 1)
  1335. CCOMMON_OPT += -DBIGNUMA
  1336. endif
  1337. endif
  1338. ifeq ($(NO_WARMUP), 1)
  1339. CCOMMON_OPT += -DNO_WARMUP
  1340. endif
  1341. ifeq ($(CONSISTENT_FPCSR), 1)
  1342. CCOMMON_OPT += -DCONSISTENT_FPCSR
  1343. endif
  1344. # Only for development
  1345. # CCOMMON_OPT += -DPARAMTEST
  1346. # CCOMMON_OPT += -DPREFETCHTEST
  1347. # CCOMMON_OPT += -DNO_SWITCHING
  1348. # USE_PAPI = 1
  1349. ifdef USE_PAPI
  1350. CCOMMON_OPT += -DUSE_PAPI
  1351. EXTRALIB += -lpapi -lperfctr
  1352. endif
  1353. ifdef BUFFERSIZE
  1354. CCOMMON_OPT += -DBUFFERSIZE=$(BUFFERSIZE)
  1355. endif
  1356. ifdef DYNAMIC_THREADS
  1357. CCOMMON_OPT += -DDYNAMIC_THREADS
  1358. endif
  1359. CCOMMON_OPT += -DMAX_CPU_NUMBER=$(NUM_THREADS)
  1360. CCOMMON_OPT += -DMAX_PARALLEL_NUMBER=$(NUM_PARALLEL)
  1361. ifdef USE_SIMPLE_THREADED_LEVEL3
  1362. CCOMMON_OPT += -DUSE_SIMPLE_THREADED_LEVEL3
  1363. endif
  1364. ifeq ($(USE_TLS), 1)
  1365. CCOMMON_OPT += -DUSE_TLS
  1366. endif
  1367. ifeq ($(BUILD_BFLOAT16), 1)
  1368. CCOMMON_OPT += -DBUILD_BFLOAT16
  1369. endif
  1370. ifeq ($(BUILD_HFLOAT16), 1)
  1371. CCOMMON_OPT += -DBUILD_HFLOAT16
  1372. endif
  1373. ifeq ($(BUILD_SINGLE), 1)
  1374. CCOMMON_OPT += -DBUILD_SINGLE=1
  1375. endif
  1376. ifeq ($(BUILD_DOUBLE), 1)
  1377. CCOMMON_OPT += -DBUILD_DOUBLE=1
  1378. endif
  1379. ifeq ($(BUILD_COMPLEX), 1)
  1380. CCOMMON_OPT += -DBUILD_COMPLEX=1
  1381. endif
  1382. ifeq ($(BUILD_COMPLEX16), 1)
  1383. CCOMMON_OPT += -DBUILD_COMPLEX16=1
  1384. endif
  1385. CCOMMON_OPT += -DVERSION=\"$(VERSION)\"
  1386. ifndef SYMBOLPREFIX
  1387. SYMBOLPREFIX =
  1388. endif
  1389. ifndef SYMBOLSUFFIX
  1390. SYMBOLSUFFIX =
  1391. endif
  1392. ifndef LIBSONAMEBASE
  1393. LIBSONAMEBASE = openblas
  1394. endif
  1395. ifndef LIBNAMEPREFIX
  1396. LIBNAMEPREFIX =
  1397. endif
  1398. SYMPREFIX=$(SYMBOLPREFIX)
  1399. ifeq ($(SYMBOLPREFIX),$(LIBNAMEPREFIX))
  1400. SYMPREFIX=
  1401. endif
  1402. SYMSUFFIX=$(SYMBOLSUFFIX)
  1403. ifeq ($(SYMBOLSUFFIX),$(LIBNAMESUFFIX))
  1404. SYMSUFFIX=
  1405. endif
  1406. ifndef LIBNAMESUFFIX
  1407. LIBNAMEBASE = $(SYMPREFIX)$(LIBSONAMEBASE)$(SYMSUFFIX)
  1408. else
  1409. LIBNAMEBASE = $(SYMPREFIX)$(LIBSONAMEBASE)$(SYMSUFFIX)$(LIBNAMESUFFIX)
  1410. endif
  1411. ifeq ($(OSNAME), CYGWIN_NT)
  1412. LIBPREFIX = cyg$(LIBNAMEPREFIX)$(LIBNAMEBASE)
  1413. else
  1414. LIBPREFIX = lib$(LIBNAMEPREFIX)$(LIBNAMEBASE)
  1415. endif
  1416. KERNELDIR = $(TOPDIR)/kernel/$(ARCH)
  1417. include $(TOPDIR)/Makefile.$(ARCH)
  1418. ifneq ($(C_COMPILER), PGI)
  1419. ifneq ($(C_COMPILER), SUN)
  1420. CCOMMON_OPT += -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME
  1421. endif
  1422. endif
  1423. CCOMMON_OPT += -DASMNAME=$(FU)$(*F) -DASMFNAME=$(FU)$(*F)$(BU) -DNAME=$(*F)$(BU) -DCNAME=$(*F) -DCHAR_NAME=\"$(*F)$(BU)\" -DCHAR_CNAME=\"$(*F)\"
  1424. ifeq ($(CORE), PPC440)
  1425. CCOMMON_OPT += -DALLOC_QALLOC
  1426. endif
  1427. ifeq ($(CORE), PPC440FP2)
  1428. STATIC_ALLOCATION = 1
  1429. endif
  1430. ifneq ($(OSNAME), Linux)
  1431. NO_AFFINITY = 1
  1432. endif
  1433. ifneq ($(ARCH), x86_64)
  1434. ifneq ($(ARCH), x86)
  1435. NO_AFFINITY = 1
  1436. endif
  1437. endif
  1438. ifeq ($(ARCH), POWER)
  1439. ifeq ($(DEBUG), 1)
  1440. CCOMMON_OPT := $(filter-out -O%, $(CCOMMON_OPT)) -O0
  1441. FCOMMON_OPT := $(filter-out -O%, $(FCOMMON_OPT)) -O0
  1442. endif
  1443. endif
  1444. ifdef NO_AFFINITY
  1445. ifeq ($(NO_AFFINITY), 0)
  1446. override undefine NO_AFFINITY
  1447. else
  1448. CCOMMON_OPT += -DNO_AFFINITY
  1449. endif
  1450. endif
  1451. ifdef FUNCTION_PROFILE
  1452. CCOMMON_OPT += -DFUNCTION_PROFILE
  1453. endif
  1454. ifdef SHMEM_ALLOCATION
  1455. ifneq ($(SHMEM_ALLOCATION), 0)
  1456. CCOMMON_OPT += -DALLOC_SHM
  1457. endif
  1458. endif
  1459. ifdef HUGETLB_ALLOCATION
  1460. ifneq ($(HUGETLB_ALLOCATION), 0)
  1461. CCOMMON_OPT += -DALLOC_HUGETLB
  1462. endif
  1463. endif
  1464. ifdef HUGETLBFILE_ALLOCATION
  1465. ifneq ($(HUGETLBFILE_ALLOCATION), 0)
  1466. CCOMMON_OPT += -DALLOC_HUGETLBFILE -DHUGETLB_FILE_NAME=$(HUGETLBFILE_ALLOCATION)
  1467. endif
  1468. endif
  1469. ifdef STATIC_ALLOCATION
  1470. CCOMMON_OPT += -DALLOC_STATIC
  1471. endif
  1472. ifdef DEVICEDRIVER_ALLOCATION
  1473. CCOMMON_OPT += -DALLOC_DEVICEDRIVER -DDEVICEDRIVER_NAME=\"/dev/mapper\"
  1474. endif
  1475. ifdef MIXED_MEMORY_ALLOCATION
  1476. CCOMMON_OPT += -DMIXED_MEMORY_ALLOCATION
  1477. endif
  1478. ifeq ($(OSNAME), SunOS)
  1479. TAR = gtar
  1480. PATCH = gpatch
  1481. GREP = ggrep
  1482. AWK = nawk
  1483. else
  1484. TAR = tar
  1485. PATCH = patch
  1486. GREP = grep
  1487. AWK = awk
  1488. endif
  1489. ifndef MD5SUM
  1490. MD5SUM = md5sum
  1491. endif
  1492. REVISION = -r$(VERSION)
  1493. MAJOR_VERSION = $(word 1,$(subst ., ,$(VERSION)))
  1494. ifeq ($(DEBUG), 1)
  1495. COMMON_OPT += -g
  1496. endif
  1497. ifeq ($(DEBUG), 1)
  1498. FCOMMON_OPT += -g
  1499. endif
  1500. ifndef COMMON_OPT
  1501. COMMON_OPT = -O2
  1502. endif
  1503. ifndef FCOMMON_OPT
  1504. FCOMMON_OPT = -O2 -frecursive
  1505. endif
  1506. override CFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
  1507. override PFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
  1508. override FFLAGS += $(COMMON_OPT) $(FCOMMON_OPT)
  1509. override FPFLAGS += $(FCOMMON_OPT) $(COMMON_PROF)
  1510. #MAKEOVERRIDES =
  1511. ifeq ($(NEED_PIC), 1)
  1512. ifeq (,$(findstring PIC,$(FFLAGS)))
  1513. ifneq ($(F_COMPILER),IBM)
  1514. override FFLAGS += -fPIC
  1515. endif
  1516. endif
  1517. endif
  1518. #For LAPACK Fortran codes.
  1519. #Disable -fopenmp for LAPACK Fortran codes on Windows.
  1520. ifdef OS_WINDOWS
  1521. LAPACK_FFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FFLAGS))
  1522. LAPACK_FPFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FPFLAGS))
  1523. else
  1524. LAPACK_FFLAGS := $(FFLAGS)
  1525. LAPACK_FPFLAGS := $(FPFLAGS)
  1526. endif
  1527. ifeq ($(F_COMPILER),NAG)
  1528. LAPACK_FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1529. override FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1530. endif
  1531. ifeq ($(F_COMPILER),CRAY)
  1532. LAPACK_FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1533. override FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1534. endif
  1535. ifeq ($(F_COMPILER),FLANGNEW)
  1536. LAPACK_FFLAGS := $(filter-out -m32 -m64 -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 -mtune=% -mabi=% ,$(FFLAGS))
  1537. override FFLAGS := $(filter-out -m32 -m64 -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 -mtune=% -mabi=% ,$(FFLAGS))
  1538. endif
  1539. LAPACK_CFLAGS = $(CFLAGS)
  1540. LAPACK_CFLAGS += -DHAVE_LAPACK_CONFIG_H
  1541. ifdef INTERFACE64
  1542. ifneq ($(INTERFACE64), 0)
  1543. LAPACK_CFLAGS += -DLAPACK_ILP64
  1544. endif
  1545. endif
  1546. ifdef OS_WINDOWS
  1547. LAPACK_CFLAGS += -DOPENBLAS_OS_WINDOWS
  1548. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1549. endif
  1550. ifeq ($(C_COMPILER), LSB)
  1551. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1552. endif
  1553. ifndef SUFFIX
  1554. SUFFIX = o
  1555. endif
  1556. ifndef PSUFFIX
  1557. PSUFFIX = po
  1558. endif
  1559. ifndef LIBSUFFIX
  1560. LIBSUFFIX = a
  1561. endif
  1562. ifneq ($(DYNAMIC_ARCH), 1)
  1563. ifndef SMP
  1564. LIBNAME = $(LIBPREFIX)_$(LIBCORE)$(REVISION).$(LIBSUFFIX)
  1565. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)$(REVISION)_p.$(LIBSUFFIX)
  1566. else
  1567. LIBNAME = $(LIBPREFIX)_$(LIBCORE)p$(REVISION).$(LIBSUFFIX)
  1568. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)p$(REVISION)_p.$(LIBSUFFIX)
  1569. endif
  1570. else
  1571. ifndef SMP
  1572. LIBNAME = $(LIBPREFIX)$(REVISION).$(LIBSUFFIX)
  1573. LIBNAME_P = $(LIBPREFIX)$(REVISION)_p.$(LIBSUFFIX)
  1574. else
  1575. LIBNAME = $(LIBPREFIX)p$(REVISION).$(LIBSUFFIX)
  1576. LIBNAME_P = $(LIBPREFIX)p$(REVISION)_p.$(LIBSUFFIX)
  1577. endif
  1578. endif
  1579. ifeq ($(FIXED_LIBNAME),1)
  1580. LIBNAME = lib$(LIBNAMEPREFIX)$(LIBSONAMEBASE)$(LIBNAMESUFFIX).$(LIBSUFFIX)
  1581. LIBNAME_P = lib$(LIBNAMEPREFIX)$(LISOBNAMEBASE)$(LIBNAMESUFFIX)_p.$(LIBSUFFIX)
  1582. endif
  1583. LIBDLLNAME = $(LIBPREFIX).dll
  1584. IMPLIBNAME = lib$(LIBNAMEBASE).dll.a
  1585. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.so)
  1586. LIBDYNNAME = $(LIBNAME:.$(LIBSUFFIX)=.dylib)
  1587. LIBDEFNAME = $(LIBNAME:.$(LIBSUFFIX)=.def)
  1588. LIBEXPNAME = $(LIBNAME:.$(LIBSUFFIX)=.exp)
  1589. LIBZIPNAME = $(LIBNAME:.$(LIBSUFFIX)=.zip)
  1590. LIBS = $(TOPDIR)/$(LIBNAME)
  1591. LIBS_P = $(TOPDIR)/$(LIBNAME_P)
  1592. LIB_COMPONENTS = BLAS
  1593. ifneq ($(NO_CBLAS), 1)
  1594. LIB_COMPONENTS += CBLAS
  1595. endif
  1596. ifneq ($(NO_LAPACK), 1)
  1597. LIB_COMPONENTS += LAPACK
  1598. ifneq ($(NO_LAPACKE), 1)
  1599. LIB_COMPONENTS += LAPACKE
  1600. endif
  1601. ifeq ($(BUILD_RELAPACK), 1)
  1602. LIB_COMPONENTS += ReLAPACK
  1603. endif
  1604. endif
  1605. ifeq ($(ONLY_CBLAS), 1)
  1606. LIB_COMPONENTS = CBLAS
  1607. endif
  1608. export OSNAME
  1609. export ARCH
  1610. export CORE
  1611. export LIBCORE
  1612. export __BYTE_ORDER__
  1613. export ELF_VERSION
  1614. export PGCPATH
  1615. export CONFIG
  1616. export CC
  1617. export FC
  1618. export BU
  1619. export FU
  1620. export NEED2UNDERSCORES
  1621. export USE_THREAD
  1622. export NUM_THREADS
  1623. export NUM_CORES
  1624. export SMP
  1625. export MAKEFILE_RULE
  1626. export NEED_PIC
  1627. export BINARY
  1628. export BINARY32
  1629. export BINARY64
  1630. export F_COMPILER
  1631. export C_COMPILER
  1632. export USE_OPENMP
  1633. export CROSS
  1634. export CROSS_SUFFIX
  1635. export NOFORTRAN
  1636. export C_LAPACK
  1637. export NO_FBLAS
  1638. export EXTRALIB
  1639. export CEXTRALIB
  1640. export FEXTRALIB
  1641. export HAVE_SSE
  1642. export HAVE_SSE2
  1643. export HAVE_SSE3
  1644. export HAVE_SSSE3
  1645. export HAVE_SSE4_1
  1646. export HAVE_SSE4_2
  1647. export HAVE_SSE4A
  1648. export HAVE_SSE5
  1649. export HAVE_AVX
  1650. export HAVE_AVX2
  1651. export HAVE_FMA3
  1652. export HAVE_VFP
  1653. export HAVE_VFPV3
  1654. export HAVE_VFPV4
  1655. export HAVE_NEON
  1656. ifndef NO_MSA
  1657. export HAVE_MSA
  1658. export MSA_FLAGS
  1659. endif
  1660. export KERNELDIR
  1661. export FUNCTION_PROFILE
  1662. export TARGET_CORE
  1663. export NO_AVX512
  1664. export NO_AVX2
  1665. export BUILD_BFLOAT16
  1666. export BUILD_HFLOAT16
  1667. export NO_LSX
  1668. export NO_LASX
  1669. export BGEMM_UNROLL_M
  1670. export BGEMM_UNROLL_N
  1671. export SBGEMM_UNROLL_M
  1672. export SBGEMM_UNROLL_N
  1673. export SHGEMM_UNROLL_M
  1674. export SHGEMM_UNROLL_N
  1675. export SGEMM_UNROLL_M
  1676. export SGEMM_UNROLL_N
  1677. export DGEMM_UNROLL_M
  1678. export DGEMM_UNROLL_N
  1679. export QGEMM_UNROLL_M
  1680. export QGEMM_UNROLL_N
  1681. export CGEMM_UNROLL_M
  1682. export CGEMM_UNROLL_N
  1683. export ZGEMM_UNROLL_M
  1684. export ZGEMM_UNROLL_N
  1685. export XGEMM_UNROLL_M
  1686. export XGEMM_UNROLL_N
  1687. export CGEMM3M_UNROLL_M
  1688. export CGEMM3M_UNROLL_N
  1689. export ZGEMM3M_UNROLL_M
  1690. export ZGEMM3M_UNROLL_N
  1691. export XGEMM3M_UNROLL_M
  1692. export XGEMM3M_UNROLL_N
  1693. ifdef USE_CUDA
  1694. export CUDADIR
  1695. export CUCC
  1696. export CUFLAGS
  1697. export CULIB
  1698. endif
  1699. .SUFFIXES: .$(PSUFFIX) .$(SUFFIX) .f
  1700. .f.$(SUFFIX):
  1701. $(FC) $(FFLAGS) -c $< -o $(@F)
  1702. .f.$(PSUFFIX):
  1703. $(FC) $(FPFLAGS) -pg -c $< -o $(@F)
  1704. ifdef BINARY64
  1705. PATHSCALEPATH = /opt/pathscale/lib/3.1
  1706. PGIPATH = /opt/pgi/linux86-64/7.1-5/lib
  1707. else
  1708. PATHSCALEPATH = /opt/pathscale/lib/3.1/32
  1709. PGIPATH = /opt/pgi/linux86/7.1-5/lib
  1710. endif
  1711. ACMLPATH = /opt/acml/4.3.0
  1712. ifneq ($(OSNAME), Darwin)
  1713. MKLPATH = /opt/intel/mkl/10.2.2.025/lib
  1714. else
  1715. MKLPATH = /Library/Frameworks/Intel_MKL.framework/Versions/10.0.1.014/lib
  1716. endif
  1717. ATLASPATH = /opt/atlas/3.9.17/opteron
  1718. FLAMEPATH = $(HOME)/flame/lib
  1719. ifneq ($(OSNAME), SunOS)
  1720. SUNPATH = /opt/sunstudio12.1
  1721. else
  1722. SUNPATH = /opt/SUNWspro
  1723. endif