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
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942943944945946947948949950951952953954955956957958959960961962963964965966967968969970971972973974975976977978979980981982983984985986987988989990991992993994995996997998999100010011002100310041005100610071008100910101011101210131014101510161017101810191020102110221023102410251026102710281029103010311032103310341035103610371038103910401041104210431044104510461047104810491050105110521053105410551056105710581059106010611062106310641065106610671068106910701071107210731074107510761077107810791080108110821083108410851086108710881089109010911092109310941095109610971098109911001101110211031104110511061107110811091110111111121113111411151116111711181119112011211122112311241125112611271128112911301131113211331134113511361137113811391140114111421143114411451146114711481149115011511152115311541155115611571158115911601161116211631164116511661167116811691170117111721173117411751176117711781179118011811182118311841185118611871188118911901191119211931194119511961197119811991200120112021203120412051206120712081209121012111212121312141215121612171218121912201221122212231224122512261227122812291230123112321233123412351236123712381239124012411242124312441245124612471248124912501251125212531254125512561257125812591260126112621263126412651266126712681269127012711272127312741275127612771278127912801281128212831284128512861287128812891290129112921293129412951296129712981299130013011302130313041305130613071308130913101311131213131314131513161317131813191320132113221323132413251326132713281329133013311332133313341335133613371338133913401341134213431344134513461347134813491350135113521353135413551356135713581359136013611362136313641365136613671368136913701371137213731374137513761377137813791380138113821383138413851386138713881389139013911392139313941395139613971398139914001401140214031404140514061407140814091410141114121413141414151416141714181419142014211422142314241425142614271428142914301431143214331434143514361437143814391440144114421443144414451446144714481449145014511452145314541455145614571458145914601461146214631464146514661467146814691470147114721473147414751476147714781479148014811482148314841485148614871488148914901491149214931494149514961497149814991500150115021503150415051506150715081509151015111512151315141515151615171518151915201521152215231524152515261527152815291530153115321533153415351536153715381539154015411542154315441545154615471548154915501551155215531554155515561557155815591560156115621563156415651566156715681569157015711572157315741575157615771578157915801581158215831584158515861587158815891590159115921593159415951596159715981599160016011602160316041605160616071608160916101611161216131614161516161617161816191620162116221623162416251626162716281629163016311632163316341635163616371638163916401641164216431644164516461647164816491650165116521653165416551656165716581659166016611662166316641665166616671668166916701671167216731674167516761677167816791680168116821683168416851686168716881689169016911692169316941695169616971698169917001701170217031704170517061707170817091710171117121713171417151716171717181719172017211722172317241725172617271728172917301731173217331734173517361737173817391740174117421743174417451746174717481749175017511752175317541755175617571758175917601761176217631764176517661767176817691770177117721773177417751776177717781779178017811782178317841785178617871788178917901791179217931794179517961797179817991800180118021803180418051806180718081809181018111812181318141815181618171818181918201821182218231824182518261827182818291830183118321833183418351836183718381839184018411842184318441845184618471848184918501851185218531854185518561857185818591860186118621863186418651866186718681869187018711872187318741875187618771878187918801881188218831884188518861887188818891890189118921893189418951896189718981899190019011902190319041905190619071908190919101911191219131914191519161917191819191920192119221923192419251926192719281929193019311932193319341935193619371938193919401941194219431944194519461947
  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 ($(ARCH), arm64)
  244. GEMM_GEMV_FORWARD = 1
  245. endif
  246. ifeq ($(ARCH), riscv)
  247. GEMM_GEMV_FORWARD = 1
  248. endif
  249. ifeq ($(ARCH), power)
  250. GEMM_GEMV_FORWARD = 1
  251. GEMM_GEMV_FORWARD_BF16 = 1
  252. endif
  253. ifeq ($(SMALL_MATRIX_OPT), 1)
  254. CCOMMON_OPT += -DSMALL_MATRIX_OPT
  255. endif
  256. ifneq ($(ONLY_CBLAS), 1)
  257. ifeq ($(GEMM_GEMV_FORWARD), 1)
  258. CCOMMON_OPT += -DGEMM_GEMV_FORWARD
  259. endif
  260. ifeq ($(GEMM_GEMV_FORWARD_BF16), 1)
  261. CCOMMON_OPT += -DGEMM_GEMV_FORWARD_BF16
  262. endif
  263. endif
  264. # This operation is expensive, so execution should be once.
  265. ifndef GOTOBLAS_MAKEFILE
  266. export GOTOBLAS_MAKEFILE = 1
  267. # Generating Makefile.conf and config.h
  268. 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)
  269. endif
  270. ifndef TARGET_CORE
  271. -include $(TOPDIR)/Makefile.conf
  272. else
  273. HAVE_NEON=
  274. HAVE_VFP=
  275. HAVE_VFPV3=
  276. HAVE_VFPV4=
  277. HAVE_MMX=
  278. HAVE_SSE=
  279. HAVE_SSE2=
  280. HAVE_SSE3=
  281. HAVE_SSSE3=
  282. HAVE_SSE4_1=
  283. HAVE_SSE4_2=
  284. HAVE_SSE4A=
  285. HAVE_SSE5=
  286. HAVE_AVX=
  287. HAVE_AVX2=
  288. HAVE_FMA3=
  289. include $(TOPDIR)/Makefile_kernel.conf
  290. endif
  291. ifndef NUM_PARALLEL
  292. NUM_PARALLEL = 1
  293. endif
  294. ifndef NUM_THREADS
  295. NUM_THREADS = $(NUM_CORES)
  296. endif
  297. ifeq ($(NUM_THREADS), 1)
  298. override USE_THREAD = 0
  299. override USE_OPENMP = 0
  300. endif
  301. ifdef USE_THREAD
  302. ifeq ($(USE_THREAD), 0)
  303. SMP =
  304. else
  305. SMP = 1
  306. endif
  307. else
  308. ifeq ($(NUM_THREADS), 1)
  309. SMP =
  310. else
  311. SMP = 1
  312. endif
  313. endif
  314. ifeq ($(SMP), 1)
  315. USE_LOCKING =
  316. endif
  317. ifndef NEED_PIC
  318. NEED_PIC = 1
  319. endif
  320. ARFLAGS =
  321. CPP = $(COMPILER) -E
  322. AR ?= $(CROSS_SUFFIX)ar
  323. AS ?= $(CROSS_SUFFIX)as
  324. LD ?= $(CROSS_SUFFIX)ld
  325. RANLIB ?= $(CROSS_SUFFIX)ranlib
  326. NM = $(CROSS_SUFFIX)nm
  327. DLLWRAP = $(CROSS_SUFFIX)dllwrap
  328. OBJCOPY = $(CROSS_SUFFIX)objcopy
  329. OBJCONV = $(CROSS_SUFFIX)objconv
  330. # When fortran support was either not detected or actively deselected, only build BLAS.
  331. ifeq ($(NOFORTRAN), 1)
  332. C_LAPACK = 1
  333. override FEXTRALIB =
  334. endif
  335. ifeq ($(C_COMPILER), GCC)
  336. GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
  337. GCCVERSIONGT4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 4)
  338. GCCVERSIONGT5 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 5)
  339. GCCVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 7)
  340. GCCVERSIONGTEQ8 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 8)
  341. GCCVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  342. GCCVERSIONGTEQ10 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 10)
  343. GCCVERSIONGTEQ11 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 11)
  344. GCCVERSIONGTEQ12 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 12)
  345. # Note that the behavior of -dumpversion is compile-time-configurable for
  346. # gcc-7.x and newer. Use -dumpfullversion there
  347. ifeq ($(GCCVERSIONGTEQ7),1)
  348. GCCDUMPVERSION_PARAM := -dumpfullversion
  349. else
  350. GCCDUMPVERSION_PARAM := -dumpversion
  351. endif
  352. GCCMINORVERSIONGTEQ1 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 1)
  353. GCCMINORVERSIONGTEQ2 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 2)
  354. GCCMINORVERSIONGTEQ4 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 4)
  355. GCCMINORVERSIONGTEQ7 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 7)
  356. endif
  357. ifeq ($(C_COMPILER), CLANG)
  358. CLANGVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  359. CLANGVERSIONGTEQ12 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 12)
  360. endif
  361. #
  362. # OS dependent settings
  363. #
  364. ifeq ($(OSNAME), Darwin)
  365. ifndef MACOSX_DEPLOYMENT_TARGET
  366. ifeq ($(ARCH), arm64)
  367. export MACOSX_DEPLOYMENT_TARGET=11.0
  368. ifeq ($(C_COMPILER), GCC)
  369. export NO_SVE = 1
  370. export NO_SME = 1
  371. endif
  372. else
  373. export MACOSX_DEPLOYMENT_TARGET=10.8
  374. endif
  375. endif
  376. MD5SUM = md5 -r
  377. XCVER = $(shell pkgutil --pkg-info=com.apple.pkg.Xcode |awk '/version:/ {print $2}'|cut -d: -f2|cut -f1 -d.)
  378. ifeq (x$(XCVER)x,xx)
  379. XCVER = $(shell pkgutil --pkg-info=com.apple.pkg.CLTools_Executables |awk '/version:/ {print $2}'|cut -d: -f2|cut -f1 -d.)
  380. endif
  381. ifeq (x$(XCVER), x 15)
  382. CCOMMON_OPT += -Wl,-ld_classic
  383. FCOMMON_OPT += -Wl,-ld_classic
  384. endif
  385. endif
  386. ifneq (,$(findstring $(OSNAME), FreeBSD OpenBSD DragonFly))
  387. MD5SUM = md5 -r
  388. endif
  389. ifeq ($(OSNAME), NetBSD)
  390. MD5SUM = md5 -n
  391. endif
  392. ifeq ($(OSNAME), Linux)
  393. EXTRALIB += -lm
  394. #NO_EXPRECISION = 1
  395. endif
  396. ifeq ($(OSNAME), Android)
  397. EXTRALIB += -lm
  398. endif
  399. ifeq ($(OSNAME), AIX)
  400. EXTRALIB += -lm
  401. endif
  402. ifeq ($(OSNAME), $(filter $(OSNAME),FreeBSD OpenBSD NetBSD DragonFly))
  403. ifeq ($(ARCH), $(filter $(ARCH),arm arm64))
  404. EXTRALIB += -lm
  405. endif
  406. endif
  407. ifeq ($(OSNAME), WINNT)
  408. NEED_PIC = 0
  409. NO_EXPRECISION = 1
  410. EXTRALIB += -defaultlib:advapi32
  411. SUFFIX = obj
  412. PSUFFIX = pobj
  413. LIBSUFFIX = a
  414. ifeq ($(C_COMPILER), CLANG)
  415. CCOMMON_OPT += -DMS_ABI
  416. endif
  417. #Version tests for supporting specific features (MS_ABI, POWER9 intrinsics)
  418. ifeq ($(GCCVERSIONGT4), 1)
  419. # GCC Major version > 4
  420. # It is compatible with MSVC ABI.
  421. CCOMMON_OPT += -DMS_ABI
  422. endif
  423. ifeq ($(GCCVERSIONGTEQ4), 1)
  424. ifeq ($(GCCMINORVERSIONGTEQ7), 1)
  425. # GCC Version >=4.7
  426. # It is compatible with MSVC ABI.
  427. CCOMMON_OPT += -DMS_ABI
  428. endif
  429. endif
  430. # Ensure the correct stack alignment on Win32
  431. # http://permalink.gmane.org/gmane.comp.lib.openblas.general/97
  432. ifeq ($(ARCH), x86)
  433. CCOMMON_OPT += -mincoming-stack-boundary=2
  434. FCOMMON_OPT += -mincoming-stack-boundary=2
  435. endif
  436. endif
  437. ifeq ($(OSNAME), Interix)
  438. NEED_PIC = 0
  439. NO_EXPRECISION = 1
  440. INTERIX_TOOL_DIR = /opt/gcc.3.3/i586-pc-interix3/bin
  441. endif
  442. ifeq ($(OSNAME), CYGWIN_NT)
  443. NEED_PIC = 0
  444. NO_EXPRECISION = 1
  445. OS_CYGWIN_NT = 1
  446. endif
  447. ifneq ($(OSNAME), WINNT)
  448. ifneq ($(OSNAME), CYGWIN_NT)
  449. ifneq ($(OSNAME), Interix)
  450. ifneq ($(OSNAME), Android)
  451. ifdef SMP
  452. EXTRALIB += -lpthread
  453. endif
  454. endif
  455. endif
  456. endif
  457. endif
  458. # ifeq logical or
  459. ifeq ($(OSNAME), $(filter $(OSNAME),WINNT CYGWIN_NT Interix))
  460. OS_WINDOWS=1
  461. endif
  462. ifdef QUAD_PRECISION
  463. CCOMMON_OPT += -DQUAD_PRECISION
  464. NO_EXPRECISION = 1
  465. endif
  466. ifneq ($(ARCH), x86)
  467. ifneq ($(ARCH), x86_64)
  468. NO_EXPRECISION = 1
  469. endif
  470. endif
  471. ifdef UTEST_CHECK
  472. CCOMMON_OPT += -DUTEST_CHECK
  473. SANITY_CHECK = 1
  474. endif
  475. ifdef SANITY_CHECK
  476. CCOMMON_OPT += -DSANITY_CHECK -DREFNAME=$(*F)f$(BU)
  477. endif
  478. MAX_STACK_ALLOC ?= 2048
  479. ifneq ($(MAX_STACK_ALLOC), 0)
  480. CCOMMON_OPT += -DMAX_STACK_ALLOC=$(MAX_STACK_ALLOC)
  481. endif
  482. ifdef USE_LOCKING
  483. ifneq ($(USE_LOCKING), 0)
  484. CCOMMON_OPT += -DUSE_LOCKING
  485. endif
  486. endif
  487. #
  488. # Architecture dependent settings
  489. #
  490. ifeq ($(ARCH), x86)
  491. ifndef BINARY
  492. NO_BINARY_MODE = 1
  493. endif
  494. ifeq ($(CORE), generic)
  495. #NO_EXPRECISION = 1
  496. endif
  497. ifndef NO_EXPRECISION
  498. ifeq ($(F_COMPILER), GFORTRAN)
  499. # ifeq logical or. GCC or LSB
  500. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  501. EXPRECISION = 1
  502. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  503. FCOMMON_OPT += -m128bit-long-double
  504. endif
  505. ifeq ($(C_COMPILER), CLANG)
  506. EXPRECISION = 1
  507. CCOMMON_OPT += -DEXPRECISION
  508. FCOMMON_OPT += -m128bit-long-double
  509. endif
  510. endif
  511. endif
  512. endif
  513. ifeq ($(ARCH), x86_64)
  514. ifeq ($(CORE), generic)
  515. #NO_EXPRECISION = 1
  516. endif
  517. ifndef NO_EXPRECISION
  518. ifeq ($(F_COMPILER), GFORTRAN)
  519. # ifeq logical or. GCC or LSB
  520. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  521. EXPRECISION = 1
  522. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  523. FCOMMON_OPT += -m128bit-long-double
  524. endif
  525. ifeq ($(C_COMPILER), CLANG)
  526. EXPRECISION = 1
  527. CCOMMON_OPT += -DEXPRECISION
  528. FCOMMON_OPT += -m128bit-long-double
  529. endif
  530. endif
  531. endif
  532. endif
  533. ifeq ($(C_COMPILER), INTEL)
  534. CCOMMON_OPT += -wd981
  535. endif
  536. ifeq ($(USE_OPENMP), 1)
  537. #check
  538. ifeq ($(USE_THREAD), 0)
  539. $(error OpenBLAS: Cannot set both USE_OPENMP=1 and USE_THREAD=0. The USE_THREAD=0 is only for building single thread version.)
  540. endif
  541. # ifeq logical or. GCC or LSB
  542. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  543. CCOMMON_OPT += -fopenmp
  544. endif
  545. ifeq ($(C_COMPILER), CLANG)
  546. CCOMMON_OPT += -fopenmp
  547. ifeq ($(F_COMPILER), GFORTRAN)
  548. FEXTRALIB := $(subst -lgomp,-lomp,$(FEXTRALIB))
  549. endif
  550. endif
  551. ifeq ($(C_COMPILER), INTEL)
  552. CCOMMON_OPT += -fopenmp
  553. endif
  554. ifeq ($(C_COMPILER), PGI)
  555. CCOMMON_OPT += -mp
  556. endif
  557. ifeq ($(C_COMPILER), OPEN64)
  558. CCOMMON_OPT += -mp
  559. CEXTRALIB += -lstdc++
  560. endif
  561. ifeq ($(C_COMPILER), PATHSCALE)
  562. CCOMMON_OPT += -mp
  563. endif
  564. endif
  565. ifeq ($(DYNAMIC_ARCH), 1)
  566. ifeq ($(ARCH), x86)
  567. DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
  568. CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
  569. endif
  570. ifeq ($(ARCH), x86_64)
  571. DYNAMIC_CORE = PRESCOTT CORE2
  572. ifeq ($(DYNAMIC_OLDER), 1)
  573. DYNAMIC_CORE += PENRYN DUNNINGTON
  574. endif
  575. DYNAMIC_CORE += NEHALEM
  576. ifeq ($(DYNAMIC_OLDER), 1)
  577. DYNAMIC_CORE += OPTERON OPTERON_SSE3
  578. endif
  579. DYNAMIC_CORE += BARCELONA
  580. ifeq ($(DYNAMIC_OLDER), 1)
  581. DYNAMIC_CORE += BOBCAT ATOM NANO
  582. endif
  583. ifneq ($(NO_AVX), 1)
  584. DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER STEAMROLLER EXCAVATOR
  585. endif
  586. ifneq ($(NO_AVX2), 1)
  587. DYNAMIC_CORE += HASWELL ZEN
  588. endif
  589. ifneq ($(NO_AVX512), 1)
  590. ifneq ($(NO_AVX2), 1)
  591. DYNAMIC_CORE += SKYLAKEX COOPERLAKE SAPPHIRERAPIDS
  592. endif
  593. endif
  594. endif
  595. ifdef DYNAMIC_LIST
  596. override DYNAMIC_CORE = PRESCOTT $(DYNAMIC_LIST)
  597. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_PRESCOTT
  598. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  599. CCOMMON_OPT += $(XCCOMMON_OPT)
  600. #CCOMMON_OPT += -DDYNAMIC_LIST='$(DYNAMIC_LIST)'
  601. endif
  602. ifeq ($(ARCH), arm64)
  603. DYNAMIC_CORE = ARMV8
  604. DYNAMIC_CORE += CORTEXA53
  605. DYNAMIC_CORE += CORTEXA57
  606. DYNAMIC_CORE += NEOVERSEN1
  607. ifneq ($(NO_SVE), 1)
  608. DYNAMIC_CORE += NEOVERSEV1
  609. DYNAMIC_CORE += NEOVERSEN2
  610. DYNAMIC_CORE += ARMV8SVE
  611. DYNAMIC_CORE += A64FX
  612. endif
  613. ifneq ($(NO_SME), 1)
  614. DYNAMIC_CORE += ARMV9SME
  615. endif
  616. DYNAMIC_CORE += THUNDERX
  617. DYNAMIC_CORE += THUNDERX2T99
  618. DYNAMIC_CORE += TSV110
  619. DYNAMIC_CORE += EMAG8180
  620. DYNAMIC_CORE += THUNDERX3T110
  621. ifdef DYNAMIC_LIST
  622. override DYNAMIC_CORE = ARMV8 $(DYNAMIC_LIST)
  623. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_ARMV8
  624. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  625. endif
  626. endif
  627. ifeq ($(ARCH), mips64)
  628. DYNAMIC_CORE = LOONGSON3R3 LOONGSON3R4 MIPS64_GENERIC
  629. ifdef DYNAMIC_LIST
  630. override DYNAMIC_CORE = MIPS64_GENERIC $(DYNAMIC_LIST)
  631. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_MIPS64_GENERIC
  632. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  633. endif
  634. endif
  635. ifeq ($(ARCH), loongarch64)
  636. DYNAMIC_CORE = LA64_GENERIC LA264 LA464
  637. endif
  638. ifeq ($(ARCH), riscv64)
  639. DYNAMIC_CORE = RISCV64_GENERIC
  640. DYNAMIC_CORE += RISCV64_ZVL128B
  641. DYNAMIC_CORE += RISCV64_ZVL256B
  642. ifdef DYNAMIC_LIST
  643. override DYNAMIC_CORE = RISCV64_GENERIC $(DYNAMIC_LIST)
  644. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_RISCV64_GENERIC
  645. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  646. endif
  647. endif
  648. ifeq ($(ARCH), zarch)
  649. DYNAMIC_CORE = ZARCH_GENERIC
  650. # if the compiler accepts -march=arch11 or -march=z13 and can compile a file
  651. # with z13-specific inline assembly, then we can include support for Z13.
  652. # note: -march=z13 is equivalent to -march=arch11 yet some compiler releases
  653. # only support one or the other.
  654. # note: LLVM version 6.x supported -march=z13 yet could not handle vector
  655. # registers in inline assembly, so the check for supporting the -march flag is
  656. # not enough.
  657. ZARCH_TEST_COMPILE=-c $(TOPDIR)/kernel/zarch/damin_z13.c -I$(TOPDIR) -o /dev/null > /dev/null 2> /dev/null
  658. ZARCH_CC_SUPPORTS_ARCH11=$(shell $(CC) -march=arch11 $(ZARCH_TEST_COMPILE) && echo 1)
  659. ZARCH_CC_SUPPORTS_Z13=$(shell $(CC) -march=z13 $(ZARCH_TEST_COMPILE) && echo 1)
  660. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH11), $(ZARCH_CC_SUPPORTS_Z13)), 1)
  661. DYNAMIC_CORE += Z13
  662. CCOMMON_OPT += -DDYN_Z13
  663. else
  664. $(info OpenBLAS: Not building Z13 kernels because the compiler $(CC) does not support it)
  665. endif
  666. # as above for z13, check for -march=arch12 and z14 support in the compiler.
  667. ZARCH_CC_SUPPORTS_ARCH12=$(shell $(CC) -march=arch12 $(ZARCH_TEST_COMPILE) && echo 1)
  668. ZARCH_CC_SUPPORTS_Z14=$(shell $(CC) -march=z14 $(ZARCH_TEST_COMPILE) && echo 1)
  669. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH12), $(ZARCH_CC_SUPPORTS_Z14)), 1)
  670. DYNAMIC_CORE += Z14
  671. CCOMMON_OPT += -DDYN_Z14
  672. else
  673. $(info OpenBLAS: Not building Z14 kernels because the compiler $(CC) does not support it)
  674. endif
  675. endif # ARCH zarch
  676. ifeq ($(ARCH), power)
  677. ifneq ($(C_COMPILER), PGI)
  678. DYNAMIC_CORE = POWER6
  679. DYNAMIC_CORE += POWER8
  680. ifneq ($(C_COMPILER), GCC)
  681. DYNAMIC_CORE += POWER9
  682. DYNAMIC_CORE += POWER10
  683. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  684. endif
  685. ifeq ($(C_COMPILER), GCC)
  686. ifeq ($(GCCVERSIONGT5), 1)
  687. DYNAMIC_CORE += POWER9
  688. else
  689. $(info, OpenBLAS: Your gcc version is too old to build the POWER9 kernels.)
  690. endif
  691. ifeq ($(OSNAME), AIX)
  692. LDVERSIONGTEQ35 := 1
  693. else
  694. LDVERSIONGTEQ35 := $(shell expr `$(CC) -Wl,--version 2> /dev/null | head -1 | cut -f2 -d "." | cut -f1 -d "-"` \>= 35)
  695. endif
  696. ifeq ($(GCCVERSIONGTEQ11)$(LDVERSIONGTEQ35), 11)
  697. DYNAMIC_CORE += POWER10
  698. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  699. else ifeq ($(GCCVERSIONGTEQ10), 1)
  700. ifeq ($(GCCMINORVERSIONGTEQ2)$(LDVERSIONGTEQ35), 11)
  701. DYNAMIC_CORE += POWER10
  702. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  703. endif
  704. else
  705. $(info, OpenBLAS: Your gcc version is too old to build the POWER10 kernels.)
  706. endif
  707. endif
  708. else
  709. DYNAMIC_CORE = POWER8
  710. DYNAMIC_CORE += POWER9
  711. endif
  712. endif
  713. # If DYNAMIC_CORE is not set, DYNAMIC_ARCH cannot do anything, so force it to empty
  714. ifndef DYNAMIC_CORE
  715. override DYNAMIC_ARCH=
  716. endif
  717. endif
  718. ifeq ($(ARCH), ia64)
  719. NO_BINARY_MODE = 1
  720. BINARY_DEFINED = 1
  721. ifeq ($(F_COMPILER), GFORTRAN)
  722. ifeq ($(C_COMPILER), GCC)
  723. EXPRECISION = 1
  724. CCOMMON_OPT += -DEXPRECISION
  725. endif
  726. endif
  727. endif
  728. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  729. NO_BINARY_MODE = 1
  730. endif
  731. ifeq ($(ARCH), alpha)
  732. NO_BINARY_MODE = 1
  733. BINARY_DEFINED = 1
  734. endif
  735. ifeq ($(ARCH), arm)
  736. NO_BINARY_MODE = 1
  737. BINARY_DEFINED = 1
  738. ifneq ($(EMBEDDED), 1)
  739. CCOMMON_OPT += -marm
  740. FCOMMON_OPT += -marm
  741. else
  742. CCOMMON_OPT += -DOS_EMBEDDED -mthumb -mcpu=cortex-m4 -mfloat-abi=hard -mfpu=fpv4-sp-d16
  743. endif
  744. # If softfp abi is mentioned on the command line, force it.
  745. ifeq ($(ARM_SOFTFP_ABI), 1)
  746. CCOMMON_OPT += -mfloat-abi=softfp
  747. FCOMMON_OPT += -mfloat-abi=softfp
  748. endif
  749. ifeq ($(OSNAME), Android)
  750. ifeq ($(ARM_SOFTFP_ABI), 1)
  751. EXTRALIB += -lm
  752. else
  753. EXTRALIB += -Wl,-lm_hard
  754. endif
  755. endif
  756. endif
  757. ifeq ($(ARCH), arm64)
  758. NO_BINARY_MODE = 1
  759. BINARY_DEFINED = 1
  760. ifdef INTERFACE64
  761. ifneq ($(INTERFACE64), 0)
  762. ifeq ($(F_COMPILER), GFORTRAN)
  763. FCOMMON_OPT += -fdefault-integer-8
  764. endif
  765. ifeq ($(F_COMPILER), FLANG)
  766. FCOMMON_OPT += -i8
  767. endif
  768. endif
  769. endif
  770. endif
  771. ifeq ($(ARCH), riscv64)
  772. NO_BINARY_MODE = 1
  773. BINARY_DEFINED = 1
  774. ifdef INTERFACE64
  775. ifneq ($(INTERFACE64), 0)
  776. ifeq ($(F_COMPILER), GFORTRAN)
  777. FCOMMON_OPT += -fdefault-integer-8
  778. endif
  779. ifeq ($(F_COMPILER), FLANG)
  780. FCOMMON_OPT += -i8
  781. endif
  782. endif
  783. endif
  784. endif
  785. ifeq ($(ARCH), loongarch64)
  786. NO_BINARY_MODE = 1
  787. BINARY_DEFINED = 1
  788. ifdef INTERFACE64
  789. ifneq ($(INTERFACE64), 0)
  790. ifeq ($(F_COMPILER), GFORTRAN)
  791. FCOMMON_OPT += -fdefault-integer-8
  792. endif
  793. ifeq ($(F_COMPILER), FLANG)
  794. FCOMMON_OPT += -i8
  795. endif
  796. endif
  797. endif
  798. endif
  799. ifeq ($(ARCH), csky)
  800. NO_BINARY_MODE = 1
  801. BINARY_DEFINED = 1
  802. endif
  803. #
  804. # C Compiler dependent settings
  805. #
  806. # ifeq logical or. GCC or CLANG or LSB
  807. # http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
  808. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG LSB))
  809. CCOMMON_OPT += -Wall
  810. COMMON_PROF += -fno-inline
  811. NO_UNINITIALIZED_WARN = -Wno-uninitialized
  812. ifeq ($(QUIET_MAKE), 1)
  813. CCOMMON_OPT += $(NO_UNINITIALIZED_WARN) -Wno-unused
  814. endif
  815. ifdef NO_BINARY_MODE
  816. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  817. ifdef BINARY64
  818. CCOMMON_OPT += -mabi=64
  819. else
  820. CCOMMON_OPT += -mabi=n32
  821. endif
  822. BINARY_DEFINED = 1
  823. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  824. CCOMMON_OPT += -mabi=32
  825. BINARY_DEFINED = 1
  826. endif
  827. ifneq (, $(filter $(CORE), MIPS64_GENERIC))
  828. CCOMMON_OPT += -DNO_MSA
  829. FCOMMON_OPT += -DNO_MSA
  830. endif
  831. ifneq (, $(filter $(CORE),LOONGSON3R3 LOONGSON3R4))
  832. CCOMMON_OPT += -march=loongson3a
  833. FCOMMON_OPT += -march=loongson3a
  834. endif
  835. ifeq ($(CORE), MIPS24K)
  836. CCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  837. FCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  838. endif
  839. ifeq ($(CORE), MIPS1004K)
  840. CCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  841. FCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  842. endif
  843. ifeq ($(CORE), P5600)
  844. CCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  845. FCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  846. endif
  847. ifeq ($(CORE), I6400)
  848. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  849. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  850. endif
  851. ifeq ($(CORE), P6600)
  852. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  853. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  854. endif
  855. ifeq ($(CORE), I6500)
  856. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  857. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  858. endif
  859. ifeq ($(OSNAME), AIX)
  860. BINARY_DEFINED = 1
  861. endif
  862. ifeq ($(ARCH), loongarch64)
  863. LA64_ABI=$(shell $(CC) -mabi=lp64d -c $(TOPDIR)/cpuid_loongarch64.c -o /dev/null > /dev/null 2> /dev/null && echo lp64d)
  864. LA64_ARCH=$(shell $(CC) -march=loongarch64 -c $(TOPDIR)/cpuid_loongarch64.c -o /dev/null > /dev/null 2> /dev/null && echo loongarch64)
  865. ifneq ($(LA64_ABI), lp64d)
  866. LA64_ABI=lp64
  867. endif
  868. ifneq ($(LA64_ARCH), loongarch64)
  869. CCOMMON_OPT += -mabi=$(LA64_ABI)
  870. FCOMMON_OPT += -mabi=$(LA64_ABI)
  871. else
  872. CCOMMON_OPT += -march=loongarch64 -mabi=$(LA64_ABI)
  873. FCOMMON_OPT += -march=loongarch64 -mabi=$(LA64_ABI)
  874. endif
  875. endif
  876. endif
  877. ifndef BINARY_DEFINED
  878. ifneq ($(OSNAME), AIX)
  879. ifdef BINARY64
  880. ifneq ($(ARCH), riscv64)
  881. CCOMMON_OPT += -m64
  882. endif
  883. else
  884. CCOMMON_OPT += -m32
  885. endif
  886. endif
  887. endif
  888. endif
  889. ifeq ($(C_COMPILER), PGI)
  890. PGCVERSIONGT20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \> 20)
  891. PGCVERSIONEQ20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` == 20)
  892. PGCMINORVERSIONGE11 := $(shell expr `$(CC) --version|sed -n "2p" |cut -d "-" -f 1 |sed -e "s/[^0-9.]//g" |cut -c 4-5` \>= 11)
  893. PGCVERSIONCHECK := $(PGCVERSIONGT20)$(PGCVERSIONEQ20)$(PGCMINORVERSIONGE11)
  894. ifeq ($(PGCVERSIONCHECK), $(filter $(PGCVERSIONCHECK), 100 101 011))
  895. NEWPGI := 1
  896. PGCVERSIONGT21 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \> 21)
  897. PGCVERSIONEQ21 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` == 21)
  898. PGCVERSIONCHECK2 := $(PGCVERSIONGT21)$(PGCVERSIONEQ21)$(PGCMINORVERSIONGE11)
  899. ifeq ($(PGCVERSIONCHECK2), $(filter $(PGCVERSIONCHECK2), 100 101 011))
  900. NEWPGI2 := 1
  901. endif
  902. endif
  903. ifdef BINARY64
  904. ifeq ($(ARCH), x86_64)
  905. ifeq (,$(findstring tp,$(CFLAGS)))
  906. ifneq ($(NEWPGI2),1)
  907. CCOMMON_OPT += -tp p7-64
  908. else
  909. CCOMMON_OPT += -tp px
  910. endif
  911. endif
  912. ifneq ($(NEWPGI),1)
  913. CCOMMON_OPT += -D__MMX__ -Mnollvm
  914. endif
  915. else
  916. ifeq ($(ARCH), power)
  917. ifeq (,$(findstring tp,$(CFLAGS)))
  918. ifeq ($(CORE), POWER8)
  919. CCOMMON_OPT += -tp pwr8
  920. endif
  921. ifeq ($(CORE), POWER9)
  922. CCOMMON_OPT += -tp pwr9
  923. endif
  924. endif
  925. endif
  926. endif
  927. else
  928. ifneq ($(NEWPGI2),1)
  929. ifeq (,$(findstring tp,$(CFLAGS)))
  930. CCOMMON_OPT += -tp p7
  931. else
  932. CCOMMON_OPT += -tp px
  933. endif
  934. endif
  935. endif
  936. endif
  937. ifeq ($(C_COMPILER), PATHSCALE)
  938. ifdef BINARY64
  939. CCOMMON_OPT += -m64
  940. else
  941. CCOMMON_OPT += -m32
  942. endif
  943. endif
  944. #
  945. # Fortran Compiler dependent settings
  946. #
  947. ifeq ($(F_COMPILER), NAG)
  948. FCOMMON_OPT += -dcfuns -recursive -ieee=full -w=obs -thread_safe
  949. ifdef INTERFACE64
  950. ifneq ($(INTERFACE64), 0)
  951. FCOMMON_OPT += -i8
  952. endif
  953. endif
  954. ifeq ($(USE_OPENMP), 1)
  955. FCOMMON_OPT += -openmp
  956. endif
  957. endif
  958. ifeq ($(F_COMPILER), FLANG)
  959. CCOMMON_OPT += -DF_INTERFACE_FLANG
  960. FCOMMON_OPT += -Mrecursive -Kieee
  961. ifeq ($(OSNAME), Linux)
  962. ifeq ($(ARCH), x86_64)
  963. FLANG_VENDOR := $(shell $(FC) --version|head -1 |cut -f 1 -d " ")
  964. ifeq ($(FLANG_VENDOR), AMD)
  965. FCOMMON_OPT += -fno-unroll-loops
  966. endif
  967. endif
  968. endif
  969. ifdef BINARY64
  970. ifdef INTERFACE64
  971. ifneq ($(INTERFACE64), 0)
  972. FCOMMON_OPT += -i8
  973. endif
  974. endif
  975. FCOMMON_OPT += -Wall
  976. else
  977. FCOMMON_OPT += -Wall
  978. endif
  979. ifeq ($(USE_OPENMP), 1)
  980. FCOMMON_OPT += -fopenmp
  981. endif
  982. endif
  983. ifeq ($(F_COMPILER), G77)
  984. CCOMMON_OPT += -DF_INTERFACE_G77
  985. FCOMMON_OPT += -Wall
  986. ifndef NO_BINARY_MODE
  987. ifneq ($(OSNAME), AIX)
  988. ifdef BINARY64
  989. FCOMMON_OPT += -m64
  990. else
  991. FCOMMON_OPT += -m32
  992. endif
  993. endif
  994. endif
  995. endif
  996. ifeq ($(F_COMPILER), G95)
  997. CCOMMON_OPT += -DF_INTERFACE_G95
  998. FCOMMON_OPT += -Wall
  999. ifneq ($(OSNAME), AIX)
  1000. ifndef NO_BINARY_MODE
  1001. ifdef BINARY64
  1002. FCOMMON_OPT += -m64
  1003. else
  1004. FCOMMON_OPT += -m32
  1005. endif
  1006. endif
  1007. ifneq ($(NO_LAPACKE), 1)
  1008. FCOMMON_OPT += -fno-second-underscore
  1009. endif
  1010. endif
  1011. endif
  1012. ifeq ($(F_COMPILER), $(filter $(F_COMPILER),GFORTRAN FLANGNEW))
  1013. CCOMMON_OPT += -DF_INTERFACE_GFORT
  1014. ifeq ($(F_COMPILER), GFORTRAN)
  1015. FCOMMON_OPT += -Wall
  1016. # make single-threaded LAPACK calls thread-safe #1847
  1017. FCOMMON_OPT += -frecursive
  1018. # work around ABI problem with passing single-character arguments
  1019. FCOMMON_OPT += -fno-optimize-sibling-calls
  1020. #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
  1021. ifneq ($(NOFORTRAN), 1)
  1022. ifneq ($(NOFORTRAN), 2)
  1023. ifneq ($(NO_LAPACK), 1)
  1024. EXTRALIB += -lgfortran
  1025. endif
  1026. endif
  1027. endif
  1028. endif
  1029. ifdef NO_BINARY_MODE
  1030. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  1031. ifdef BINARY64
  1032. FCOMMON_OPT += -mabi=64
  1033. else
  1034. FCOMMON_OPT += -mabi=n32
  1035. endif
  1036. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  1037. FCOMMON_OPT += -mabi=32
  1038. endif
  1039. else
  1040. ifdef BINARY64
  1041. ifneq ($(OSNAME), AIX)
  1042. ifneq ($(ARCH), riscv64)
  1043. FCOMMON_OPT += -m64
  1044. endif
  1045. endif
  1046. ifdef INTERFACE64
  1047. ifneq ($(INTERFACE64), 0)
  1048. FCOMMON_OPT += -fdefault-integer-8
  1049. endif
  1050. endif
  1051. else
  1052. ifneq ($(OSNAME), AIX)
  1053. FCOMMON_OPT += -m32
  1054. endif
  1055. endif
  1056. endif
  1057. ifeq ($(USE_OPENMP), 1)
  1058. FCOMMON_OPT += -fopenmp
  1059. endif
  1060. endif
  1061. ifeq ($(F_COMPILER), INTEL)
  1062. CCOMMON_OPT += -DF_INTERFACE_INTEL
  1063. ifdef INTERFACE64
  1064. ifneq ($(INTERFACE64), 0)
  1065. FCOMMON_OPT += -i8
  1066. endif
  1067. endif
  1068. FCOMMON_OPT += -recursive -fp-model strict -assume protect-parens
  1069. ifeq ($(USE_OPENMP), 1)
  1070. FCOMMON_OPT += -fopenmp
  1071. endif
  1072. endif
  1073. ifeq ($(F_COMPILER), FUJITSU)
  1074. CCOMMON_OPT += -DF_INTERFACE_FUJITSU
  1075. ifeq ($(USE_OPENMP), 1)
  1076. FCOMMON_OPT += -openmp
  1077. endif
  1078. endif
  1079. ifeq ($(F_COMPILER), IBM)
  1080. CCOMMON_OPT += -DF_INTERFACE_IBM
  1081. FEXTRALIB += -lxlf90
  1082. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG))
  1083. FCOMMON_OPT += -qextname -qzerosize
  1084. endif
  1085. # FCOMMON_OPT += -qarch=440
  1086. ifdef BINARY64
  1087. FCOMMON_OPT += -q64
  1088. ifdef INTERFACE64
  1089. ifneq ($(INTERFACE64), 0)
  1090. FCOMMON_OPT += -qintsize=8
  1091. endif
  1092. endif
  1093. else
  1094. FCOMMON_OPT += -q32
  1095. endif
  1096. endif
  1097. ifeq ($(F_COMPILER), PGI)
  1098. CCOMMON_OPT += -DF_INTERFACE_PGI
  1099. COMMON_PROF += -DPGICOMPILER
  1100. ifdef BINARY64
  1101. ifdef INTERFACE64
  1102. ifneq ($(INTERFACE64), 0)
  1103. FCOMMON_OPT += -i8
  1104. endif
  1105. endif
  1106. ifeq ($(ARCH), x86_64)
  1107. ifneq ($(NEWPGI2),1)
  1108. FCOMMON_OPT += -tp p7-64
  1109. else
  1110. FCOMMON_OPT += -tp px
  1111. endif
  1112. else
  1113. ifeq ($(ARCH), power)
  1114. ifeq ($(CORE), POWER6)
  1115. $(warning NVIDIA HPC compilers do not support POWER6.)
  1116. endif
  1117. ifeq ($(CORE), POWER8)
  1118. FCOMMON_OPT += -tp pwr8
  1119. endif
  1120. ifeq ($(CORE), POWER9)
  1121. FCOMMON_OPT += -tp pwr9
  1122. endif
  1123. ifeq ($(CORE), POWER10)
  1124. $(warning NVIDIA HPC compilers do not support POWER10.)
  1125. endif
  1126. endif
  1127. endif
  1128. else
  1129. FCOMMON_OPT += -tp p7
  1130. endif
  1131. FCOMMON_OPT += -Mrecursive -Kieee
  1132. ifeq ($(USE_OPENMP), 1)
  1133. FCOMMON_OPT += -mp
  1134. endif
  1135. endif
  1136. ifeq ($(F_COMPILER), PATHSCALE)
  1137. CCOMMON_OPT += -DF_INTERFACE_PATHSCALE
  1138. ifdef BINARY64
  1139. ifdef INTERFACE64
  1140. ifneq ($(INTERFACE64), 0)
  1141. FCOMMON_OPT += -i8
  1142. endif
  1143. endif
  1144. endif
  1145. ifeq ($(USE_OPENMP), 1)
  1146. FCOMMON_OPT += -mp
  1147. endif
  1148. endif
  1149. ifeq ($(F_COMPILER), OPEN64)
  1150. CCOMMON_OPT += -DF_INTERFACE_OPEN64
  1151. ifdef BINARY64
  1152. ifdef INTERFACE64
  1153. ifneq ($(INTERFACE64), 0)
  1154. FCOMMON_OPT += -i8
  1155. endif
  1156. endif
  1157. endif
  1158. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  1159. ifndef BINARY64
  1160. FCOMMON_OPT += -n32
  1161. else
  1162. FCOMMON_OPT += -n64
  1163. endif
  1164. ifeq ($(CORE), LOONGSON3R3)
  1165. FCOMMON_OPT += -loongson3 -static
  1166. endif
  1167. ifeq ($(CORE), LOONGSON3R4)
  1168. FCOMMON_OPT += -loongson3 -static
  1169. endif
  1170. else
  1171. ifndef BINARY64
  1172. FCOMMON_OPT += -m32
  1173. else
  1174. FCOMMON_OPT += -m64
  1175. endif
  1176. endif
  1177. ifeq ($(USE_OPENMP), 1)
  1178. FEXTRALIB += -lstdc++
  1179. FCOMMON_OPT += -mp
  1180. endif
  1181. endif
  1182. ifeq ($(C_COMPILER), OPEN64)
  1183. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  1184. ifndef BINARY64
  1185. CCOMMON_OPT += -n32
  1186. else
  1187. CCOMMON_OPT += -n64
  1188. endif
  1189. ifeq ($(CORE), LOONGSON3R3)
  1190. CCOMMON_OPT += -loongson3 -static
  1191. endif
  1192. ifeq ($(CORE), LOONGSON3R4)
  1193. CCOMMON_OPT += -loongson3 -static
  1194. endif
  1195. else
  1196. ifndef BINARY64
  1197. CCOMMON_OPT += -m32
  1198. else
  1199. CCOMMON_OPT += -m64
  1200. endif
  1201. endif
  1202. endif
  1203. ifeq ($(C_COMPILER), SUN)
  1204. CCOMMON_OPT += -w
  1205. ifeq ($(ARCH), x86)
  1206. CCOMMON_OPT += -m32
  1207. else
  1208. ifdef BINARY64
  1209. CCOMMON_OPT += -m64
  1210. else
  1211. CCOMMON_OPT += -m32
  1212. endif
  1213. endif
  1214. endif
  1215. ifeq ($(F_COMPILER), SUN)
  1216. CCOMMON_OPT += -DF_INTERFACE_SUN
  1217. FCOMMON_OPT += -ftrap=%none -xrecursive
  1218. ifeq ($(ARCH), x86)
  1219. FCOMMON_OPT += -m32
  1220. else
  1221. ifdef BINARY64
  1222. FCOMMON_OPT += -m64
  1223. else
  1224. FCOMMON_OPT += -m32
  1225. endif
  1226. endif
  1227. ifeq ($(USE_OPENMP), 1)
  1228. FCOMMON_OPT += -xopenmp=parallel
  1229. endif
  1230. endif
  1231. ifeq ($(F_COMPILER), COMPAQ)
  1232. CCOMMON_OPT += -DF_INTERFACE_COMPAQ
  1233. ifeq ($(USE_OPENMP), 1)
  1234. FCOMMON_OPT += -openmp
  1235. endif
  1236. endif
  1237. ifeq ($(F_COMPILER), CRAY)
  1238. CCOMMON_OPT += -DF_INTERFACE_INTEL
  1239. FCOMMON_OPT += -hnopattern
  1240. ifdef INTERFACE64
  1241. ifneq ($(INTERFACE64), 0)
  1242. FCOMMON_OPT += -s integer64
  1243. endif
  1244. endif
  1245. ifneq ($(USE_OPENMP), 1)
  1246. FCOMMON_OPT += -O noomp
  1247. endif
  1248. endif
  1249. ifdef BINARY64
  1250. ifdef INTERFACE64
  1251. ifneq ($(INTERFACE64), 0)
  1252. CCOMMON_OPT +=
  1253. #-DUSE64BITINT
  1254. endif
  1255. endif
  1256. endif
  1257. ifeq ($(NEED_PIC), 1)
  1258. ifeq ($(C_COMPILER), IBM)
  1259. CCOMMON_OPT += -qpic=large
  1260. else
  1261. CCOMMON_OPT += -fPIC
  1262. endif
  1263. ifeq ($(F_COMPILER), SUN)
  1264. FCOMMON_OPT += -pic
  1265. else ifeq ($(F_COMPILER), NAG)
  1266. FCOMMON_OPT += -PIC
  1267. else ifeq ($(F_COMPILER), IBM)
  1268. FCOMMON_OPT += -qpic=large
  1269. else
  1270. FCOMMON_OPT += -fPIC
  1271. endif
  1272. endif
  1273. ifeq ($(DYNAMIC_ARCH), 1)
  1274. CCOMMON_OPT += -DDYNAMIC_ARCH
  1275. endif
  1276. ifeq ($(DYNAMIC_OLDER), 1)
  1277. CCOMMON_OPT += -DDYNAMIC_OLDER
  1278. endif
  1279. ifeq ($(C_LAPACK), 1)
  1280. CCOMMON_OPT += -DC_LAPACK
  1281. endif
  1282. ifeq ($(NO_LAPACK), 1)
  1283. CCOMMON_OPT += -DNO_LAPACK
  1284. #Disable LAPACK C interface
  1285. NO_LAPACKE = 1
  1286. endif
  1287. ifeq ($(NO_LAPACKE), 1)
  1288. CCOMMON_OPT += -DNO_LAPACKE
  1289. endif
  1290. ifeq ($(NO_AVX), 1)
  1291. CCOMMON_OPT += -DNO_AVX
  1292. endif
  1293. ifeq ($(ARCH), x86)
  1294. CCOMMON_OPT += -DNO_AVX
  1295. endif
  1296. ifeq ($(NO_AVX2), 1)
  1297. CCOMMON_OPT += -DNO_AVX2
  1298. endif
  1299. ifeq ($(NO_AVX512), 1)
  1300. CCOMMON_OPT += -DNO_AVX512
  1301. endif
  1302. ifeq ($(NO_SVE), 1)
  1303. CCOMMON_OPT += -DNO_SVE
  1304. endif
  1305. ifeq ($(NO_SME), 1)
  1306. CCOMMON_OPT += -DNO_SME
  1307. endif
  1308. ifdef SMP
  1309. CCOMMON_OPT += -DSMP_SERVER
  1310. ifeq ($(ARCH), mips64)
  1311. USE_SIMPLE_THREADED_LEVEL3 = 1
  1312. endif
  1313. ifeq ($(USE_OPENMP), 1)
  1314. # USE_SIMPLE_THREADED_LEVEL3 = 1
  1315. # NO_AFFINITY = 1
  1316. CCOMMON_OPT += -DUSE_OPENMP
  1317. endif
  1318. ifeq ($(BIGNUMA), 1)
  1319. CCOMMON_OPT += -DBIGNUMA
  1320. endif
  1321. endif
  1322. ifeq ($(NO_WARMUP), 1)
  1323. CCOMMON_OPT += -DNO_WARMUP
  1324. endif
  1325. ifeq ($(CONSISTENT_FPCSR), 1)
  1326. CCOMMON_OPT += -DCONSISTENT_FPCSR
  1327. endif
  1328. # Only for development
  1329. # CCOMMON_OPT += -DPARAMTEST
  1330. # CCOMMON_OPT += -DPREFETCHTEST
  1331. # CCOMMON_OPT += -DNO_SWITCHING
  1332. # USE_PAPI = 1
  1333. ifdef USE_PAPI
  1334. CCOMMON_OPT += -DUSE_PAPI
  1335. EXTRALIB += -lpapi -lperfctr
  1336. endif
  1337. ifdef BUFFERSIZE
  1338. CCOMMON_OPT += -DBUFFERSIZE=$(BUFFERSIZE)
  1339. endif
  1340. ifdef DYNAMIC_THREADS
  1341. CCOMMON_OPT += -DDYNAMIC_THREADS
  1342. endif
  1343. CCOMMON_OPT += -DMAX_CPU_NUMBER=$(NUM_THREADS)
  1344. CCOMMON_OPT += -DMAX_PARALLEL_NUMBER=$(NUM_PARALLEL)
  1345. ifdef USE_SIMPLE_THREADED_LEVEL3
  1346. CCOMMON_OPT += -DUSE_SIMPLE_THREADED_LEVEL3
  1347. endif
  1348. ifeq ($(USE_TLS), 1)
  1349. CCOMMON_OPT += -DUSE_TLS
  1350. endif
  1351. ifeq ($(BUILD_BFLOAT16), 1)
  1352. CCOMMON_OPT += -DBUILD_BFLOAT16
  1353. endif
  1354. ifeq ($(BUILD_SINGLE), 1)
  1355. CCOMMON_OPT += -DBUILD_SINGLE=1
  1356. endif
  1357. ifeq ($(BUILD_DOUBLE), 1)
  1358. CCOMMON_OPT += -DBUILD_DOUBLE=1
  1359. endif
  1360. ifeq ($(BUILD_COMPLEX), 1)
  1361. CCOMMON_OPT += -DBUILD_COMPLEX=1
  1362. endif
  1363. ifeq ($(BUILD_COMPLEX16), 1)
  1364. CCOMMON_OPT += -DBUILD_COMPLEX16=1
  1365. endif
  1366. CCOMMON_OPT += -DVERSION=\"$(VERSION)\"
  1367. ifndef SYMBOLPREFIX
  1368. SYMBOLPREFIX =
  1369. endif
  1370. ifndef SYMBOLSUFFIX
  1371. SYMBOLSUFFIX =
  1372. endif
  1373. ifndef LIBSONAMEBASE
  1374. LIBSONAMEBASE = openblas
  1375. endif
  1376. ifndef LIBNAMEPREFIX
  1377. LIBNAMEPREFIX =
  1378. endif
  1379. SYMPREFIX=$(SYMBOLPREFIX)
  1380. ifeq ($(SYMBOLPREFIX),$(LIBNAMEPREFIX))
  1381. SYMPREFIX=
  1382. endif
  1383. SYMSUFFIX=$(SYMBOLSUFFIX)
  1384. ifeq ($(SYMBOLSUFFIX),$(LIBNAMESUFFIX))
  1385. SYMSUFFIX=
  1386. endif
  1387. ifndef LIBNAMESUFFIX
  1388. LIBNAMEBASE = $(SYMPREFIX)$(LIBSONAMEBASE)$(SYMSUFFIX)
  1389. else
  1390. LIBNAMEBASE = $(SYMPREFIX)$(LIBSONAMEBASE)$(SYMSUFFIX)$(LIBNAMESUFFIX)
  1391. endif
  1392. ifeq ($(OSNAME), CYGWIN_NT)
  1393. LIBPREFIX = cyg$(LIBNAMEPREFIX)$(LIBNAMEBASE)
  1394. else
  1395. LIBPREFIX = lib$(LIBNAMEPREFIX)$(LIBNAMEBASE)
  1396. endif
  1397. KERNELDIR = $(TOPDIR)/kernel/$(ARCH)
  1398. include $(TOPDIR)/Makefile.$(ARCH)
  1399. ifneq ($(C_COMPILER), PGI)
  1400. ifneq ($(C_COMPILER), SUN)
  1401. CCOMMON_OPT += -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME
  1402. endif
  1403. endif
  1404. CCOMMON_OPT += -DASMNAME=$(FU)$(*F) -DASMFNAME=$(FU)$(*F)$(BU) -DNAME=$(*F)$(BU) -DCNAME=$(*F) -DCHAR_NAME=\"$(*F)$(BU)\" -DCHAR_CNAME=\"$(*F)\"
  1405. ifeq ($(CORE), PPC440)
  1406. CCOMMON_OPT += -DALLOC_QALLOC
  1407. endif
  1408. ifeq ($(CORE), PPC440FP2)
  1409. STATIC_ALLOCATION = 1
  1410. endif
  1411. ifneq ($(OSNAME), Linux)
  1412. NO_AFFINITY = 1
  1413. endif
  1414. ifneq ($(ARCH), x86_64)
  1415. ifneq ($(ARCH), x86)
  1416. NO_AFFINITY = 1
  1417. endif
  1418. endif
  1419. ifeq ($(ARCH), POWER)
  1420. ifeq ($(DEBUG), 1)
  1421. CCOMMON_OPT := $(filter-out -O%, $(CCOMMON_OPT)) -O0
  1422. FCOMMON_OPT := $(filter-out -O%, $(FCOMMON_OPT)) -O0
  1423. endif
  1424. endif
  1425. ifdef NO_AFFINITY
  1426. ifeq ($(NO_AFFINITY), 0)
  1427. override undefine NO_AFFINITY
  1428. else
  1429. CCOMMON_OPT += -DNO_AFFINITY
  1430. endif
  1431. endif
  1432. ifdef FUNCTION_PROFILE
  1433. CCOMMON_OPT += -DFUNCTION_PROFILE
  1434. endif
  1435. ifdef SHMEM_ALLOCATION
  1436. ifneq ($(SHMEM_ALLOCATION), 0)
  1437. CCOMMON_OPT += -DALLOC_SHM
  1438. endif
  1439. endif
  1440. ifdef HUGETLB_ALLOCATION
  1441. ifneq ($(HUGETLB_ALLOCATION), 0)
  1442. CCOMMON_OPT += -DALLOC_HUGETLB
  1443. endif
  1444. endif
  1445. ifdef HUGETLBFILE_ALLOCATION
  1446. ifneq ($(HUGETLBFILE_ALLOCATION), 0)
  1447. CCOMMON_OPT += -DALLOC_HUGETLBFILE -DHUGETLB_FILE_NAME=$(HUGETLBFILE_ALLOCATION)
  1448. endif
  1449. endif
  1450. ifdef STATIC_ALLOCATION
  1451. CCOMMON_OPT += -DALLOC_STATIC
  1452. endif
  1453. ifdef DEVICEDRIVER_ALLOCATION
  1454. CCOMMON_OPT += -DALLOC_DEVICEDRIVER -DDEVICEDRIVER_NAME=\"/dev/mapper\"
  1455. endif
  1456. ifdef MIXED_MEMORY_ALLOCATION
  1457. CCOMMON_OPT += -DMIXED_MEMORY_ALLOCATION
  1458. endif
  1459. ifeq ($(OSNAME), SunOS)
  1460. TAR = gtar
  1461. PATCH = gpatch
  1462. GREP = ggrep
  1463. AWK = nawk
  1464. else
  1465. TAR = tar
  1466. PATCH = patch
  1467. GREP = grep
  1468. AWK = awk
  1469. endif
  1470. ifndef MD5SUM
  1471. MD5SUM = md5sum
  1472. endif
  1473. REVISION = -r$(VERSION)
  1474. MAJOR_VERSION = $(word 1,$(subst ., ,$(VERSION)))
  1475. ifeq ($(DEBUG), 1)
  1476. COMMON_OPT += -g
  1477. endif
  1478. ifeq ($(DEBUG), 1)
  1479. FCOMMON_OPT += -g
  1480. endif
  1481. ifndef COMMON_OPT
  1482. COMMON_OPT = -O2
  1483. endif
  1484. ifndef FCOMMON_OPT
  1485. FCOMMON_OPT = -O2 -frecursive
  1486. endif
  1487. override CFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
  1488. override PFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
  1489. override FFLAGS += $(COMMON_OPT) $(FCOMMON_OPT)
  1490. override FPFLAGS += $(FCOMMON_OPT) $(COMMON_PROF)
  1491. #MAKEOVERRIDES =
  1492. ifeq ($(NEED_PIC), 1)
  1493. ifeq (,$(findstring PIC,$(FFLAGS)))
  1494. ifneq ($(F_COMPILER),IBM)
  1495. override FFLAGS += -fPIC
  1496. endif
  1497. endif
  1498. endif
  1499. #For LAPACK Fortran codes.
  1500. #Disable -fopenmp for LAPACK Fortran codes on Windows.
  1501. ifdef OS_WINDOWS
  1502. LAPACK_FFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FFLAGS))
  1503. LAPACK_FPFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FPFLAGS))
  1504. else
  1505. LAPACK_FFLAGS := $(FFLAGS)
  1506. LAPACK_FPFLAGS := $(FPFLAGS)
  1507. endif
  1508. ifeq ($(F_COMPILER),NAG)
  1509. LAPACK_FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1510. override FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1511. endif
  1512. ifeq ($(F_COMPILER),CRAY)
  1513. LAPACK_FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1514. override FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1515. endif
  1516. ifeq ($(F_COMPILER),FLANGNEW)
  1517. LAPACK_FFLAGS := $(filter-out -m32 -m64 -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 -mtune=% -mabi=% ,$(FFLAGS))
  1518. override FFLAGS := $(filter-out -m32 -m64 -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 -mtune=% -mabi=% ,$(FFLAGS))
  1519. endif
  1520. LAPACK_CFLAGS = $(CFLAGS)
  1521. LAPACK_CFLAGS += -DHAVE_LAPACK_CONFIG_H
  1522. ifdef INTERFACE64
  1523. ifneq ($(INTERFACE64), 0)
  1524. LAPACK_CFLAGS += -DLAPACK_ILP64
  1525. endif
  1526. endif
  1527. ifdef OS_WINDOWS
  1528. LAPACK_CFLAGS += -DOPENBLAS_OS_WINDOWS
  1529. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1530. endif
  1531. ifeq ($(C_COMPILER), LSB)
  1532. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1533. endif
  1534. ifndef SUFFIX
  1535. SUFFIX = o
  1536. endif
  1537. ifndef PSUFFIX
  1538. PSUFFIX = po
  1539. endif
  1540. ifndef LIBSUFFIX
  1541. LIBSUFFIX = a
  1542. endif
  1543. ifneq ($(DYNAMIC_ARCH), 1)
  1544. ifndef SMP
  1545. LIBNAME = $(LIBPREFIX)_$(LIBCORE)$(REVISION).$(LIBSUFFIX)
  1546. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)$(REVISION)_p.$(LIBSUFFIX)
  1547. else
  1548. LIBNAME = $(LIBPREFIX)_$(LIBCORE)p$(REVISION).$(LIBSUFFIX)
  1549. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)p$(REVISION)_p.$(LIBSUFFIX)
  1550. endif
  1551. else
  1552. ifndef SMP
  1553. LIBNAME = $(LIBPREFIX)$(REVISION).$(LIBSUFFIX)
  1554. LIBNAME_P = $(LIBPREFIX)$(REVISION)_p.$(LIBSUFFIX)
  1555. else
  1556. LIBNAME = $(LIBPREFIX)p$(REVISION).$(LIBSUFFIX)
  1557. LIBNAME_P = $(LIBPREFIX)p$(REVISION)_p.$(LIBSUFFIX)
  1558. endif
  1559. endif
  1560. ifeq ($(FIXED_LIBNAME),1)
  1561. LIBNAME = lib$(LIBNAMEPREFIX)$(LIBSONAMEBASE)$(LIBNAMESUFFIX).$(LIBSUFFIX)
  1562. LIBNAME_P = lib$(LIBNAMEPREFIX)$(LISOBNAMEBASE)$(LIBNAMESUFFIX)_p.$(LIBSUFFIX)
  1563. endif
  1564. LIBDLLNAME = $(LIBPREFIX).dll
  1565. IMPLIBNAME = lib$(LIBNAMEBASE).dll.a
  1566. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.so)
  1567. LIBDYNNAME = $(LIBNAME:.$(LIBSUFFIX)=.dylib)
  1568. LIBDEFNAME = $(LIBNAME:.$(LIBSUFFIX)=.def)
  1569. LIBEXPNAME = $(LIBNAME:.$(LIBSUFFIX)=.exp)
  1570. LIBZIPNAME = $(LIBNAME:.$(LIBSUFFIX)=.zip)
  1571. LIBS = $(TOPDIR)/$(LIBNAME)
  1572. LIBS_P = $(TOPDIR)/$(LIBNAME_P)
  1573. LIB_COMPONENTS = BLAS
  1574. ifneq ($(NO_CBLAS), 1)
  1575. LIB_COMPONENTS += CBLAS
  1576. endif
  1577. ifneq ($(NO_LAPACK), 1)
  1578. LIB_COMPONENTS += LAPACK
  1579. ifneq ($(NO_LAPACKE), 1)
  1580. LIB_COMPONENTS += LAPACKE
  1581. endif
  1582. ifeq ($(BUILD_RELAPACK), 1)
  1583. LIB_COMPONENTS += ReLAPACK
  1584. endif
  1585. endif
  1586. ifeq ($(ONLY_CBLAS), 1)
  1587. LIB_COMPONENTS = CBLAS
  1588. endif
  1589. export OSNAME
  1590. export ARCH
  1591. export CORE
  1592. export LIBCORE
  1593. export __BYTE_ORDER__
  1594. export ELF_VERSION
  1595. export PGCPATH
  1596. export CONFIG
  1597. export CC
  1598. export FC
  1599. export BU
  1600. export FU
  1601. export NEED2UNDERSCORES
  1602. export USE_THREAD
  1603. export NUM_THREADS
  1604. export NUM_CORES
  1605. export SMP
  1606. export MAKEFILE_RULE
  1607. export NEED_PIC
  1608. export BINARY
  1609. export BINARY32
  1610. export BINARY64
  1611. export F_COMPILER
  1612. export C_COMPILER
  1613. export USE_OPENMP
  1614. export CROSS
  1615. export CROSS_SUFFIX
  1616. export NOFORTRAN
  1617. export C_LAPACK
  1618. export NO_FBLAS
  1619. export EXTRALIB
  1620. export CEXTRALIB
  1621. export FEXTRALIB
  1622. export HAVE_SSE
  1623. export HAVE_SSE2
  1624. export HAVE_SSE3
  1625. export HAVE_SSSE3
  1626. export HAVE_SSE4_1
  1627. export HAVE_SSE4_2
  1628. export HAVE_SSE4A
  1629. export HAVE_SSE5
  1630. export HAVE_AVX
  1631. export HAVE_AVX2
  1632. export HAVE_FMA3
  1633. export HAVE_VFP
  1634. export HAVE_VFPV3
  1635. export HAVE_VFPV4
  1636. export HAVE_NEON
  1637. ifndef NO_MSA
  1638. export HAVE_MSA
  1639. export MSA_FLAGS
  1640. endif
  1641. export KERNELDIR
  1642. export FUNCTION_PROFILE
  1643. export TARGET_CORE
  1644. export NO_AVX512
  1645. export NO_AVX2
  1646. export BUILD_BFLOAT16
  1647. export NO_LSX
  1648. export NO_LASX
  1649. export SBGEMM_UNROLL_M
  1650. export SBGEMM_UNROLL_N
  1651. export SGEMM_UNROLL_M
  1652. export SGEMM_UNROLL_N
  1653. export DGEMM_UNROLL_M
  1654. export DGEMM_UNROLL_N
  1655. export QGEMM_UNROLL_M
  1656. export QGEMM_UNROLL_N
  1657. export CGEMM_UNROLL_M
  1658. export CGEMM_UNROLL_N
  1659. export ZGEMM_UNROLL_M
  1660. export ZGEMM_UNROLL_N
  1661. export XGEMM_UNROLL_M
  1662. export XGEMM_UNROLL_N
  1663. export CGEMM3M_UNROLL_M
  1664. export CGEMM3M_UNROLL_N
  1665. export ZGEMM3M_UNROLL_M
  1666. export ZGEMM3M_UNROLL_N
  1667. export XGEMM3M_UNROLL_M
  1668. export XGEMM3M_UNROLL_N
  1669. ifdef USE_CUDA
  1670. export CUDADIR
  1671. export CUCC
  1672. export CUFLAGS
  1673. export CULIB
  1674. endif
  1675. .SUFFIXES: .$(PSUFFIX) .$(SUFFIX) .f
  1676. .f.$(SUFFIX):
  1677. $(FC) $(FFLAGS) -c $< -o $(@F)
  1678. .f.$(PSUFFIX):
  1679. $(FC) $(FPFLAGS) -pg -c $< -o $(@F)
  1680. ifdef BINARY64
  1681. PATHSCALEPATH = /opt/pathscale/lib/3.1
  1682. PGIPATH = /opt/pgi/linux86-64/7.1-5/lib
  1683. else
  1684. PATHSCALEPATH = /opt/pathscale/lib/3.1/32
  1685. PGIPATH = /opt/pgi/linux86/7.1-5/lib
  1686. endif
  1687. ACMLPATH = /opt/acml/4.3.0
  1688. ifneq ($(OSNAME), Darwin)
  1689. MKLPATH = /opt/intel/mkl/10.2.2.025/lib
  1690. else
  1691. MKLPATH = /Library/Frameworks/Intel_MKL.framework/Versions/10.0.1.014/lib
  1692. endif
  1693. ATLASPATH = /opt/atlas/3.9.17/opteron
  1694. FLAMEPATH = $(HOME)/flame/lib
  1695. ifneq ($(OSNAME), SunOS)
  1696. SUNPATH = /opt/sunstudio12.1
  1697. else
  1698. SUNPATH = /opt/SUNWspro
  1699. endif