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

7 years ago
1 year ago
4 years ago
13 years ago
4 years ago
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942943944945946947948949950951952953954955956957958959960961962963964965966967968969970971972973974975976977978979980981982983984985986987988989990991992993994995996997998999100010011002100310041005100610071008100910101011101210131014101510161017101810191020102110221023102410251026102710281029103010311032103310341035103610371038103910401041104210431044104510461047104810491050105110521053105410551056105710581059106010611062106310641065106610671068106910701071107210731074107510761077107810791080108110821083108410851086108710881089109010911092109310941095109610971098109911001101110211031104110511061107110811091110111111121113111411151116111711181119112011211122112311241125112611271128112911301131113211331134113511361137113811391140114111421143114411451146114711481149115011511152115311541155115611571158115911601161116211631164116511661167116811691170117111721173117411751176117711781179118011811182118311841185118611871188118911901191119211931194119511961197119811991200120112021203120412051206120712081209121012111212121312141215121612171218121912201221122212231224122512261227122812291230123112321233123412351236123712381239124012411242124312441245124612471248124912501251125212531254125512561257125812591260126112621263126412651266126712681269127012711272127312741275127612771278127912801281128212831284128512861287128812891290129112921293129412951296129712981299130013011302130313041305130613071308130913101311131213131314131513161317131813191320132113221323132413251326132713281329133013311332133313341335133613371338133913401341134213431344134513461347134813491350135113521353135413551356135713581359136013611362136313641365136613671368136913701371137213731374137513761377137813791380138113821383138413851386138713881389139013911392139313941395139613971398139914001401140214031404140514061407140814091410141114121413141414151416141714181419142014211422142314241425142614271428142914301431143214331434143514361437143814391440144114421443144414451446144714481449145014511452145314541455145614571458145914601461146214631464146514661467146814691470147114721473147414751476147714781479148014811482148314841485148614871488148914901491149214931494149514961497149814991500150115021503150415051506150715081509151015111512151315141515151615171518151915201521152215231524152515261527152815291530153115321533153415351536153715381539154015411542154315441545154615471548154915501551155215531554155515561557155815591560156115621563156415651566156715681569157015711572157315741575157615771578157915801581158215831584158515861587158815891590159115921593159415951596159715981599160016011602160316041605160616071608160916101611161216131614161516161617161816191620162116221623162416251626162716281629163016311632163316341635163616371638163916401641164216431644164516461647164816491650165116521653165416551656165716581659166016611662166316641665166616671668166916701671167216731674167516761677167816791680168116821683168416851686168716881689169016911692169316941695169616971698169917001701170217031704170517061707170817091710171117121713171417151716171717181719172017211722172317241725172617271728172917301731173217331734173517361737173817391740174117421743174417451746174717481749175017511752175317541755175617571758175917601761176217631764176517661767176817691770177117721773177417751776177717781779178017811782178317841785178617871788178917901791179217931794179517961797179817991800180118021803180418051806180718081809181018111812181318141815181618171818181918201821182218231824182518261827182818291830183118321833183418351836183718381839184018411842184318441845184618471848184918501851185218531854185518561857185818591860186118621863186418651866186718681869187018711872187318741875187618771878187918801881188218831884188518861887188818891890189118921893189418951896189718981899190019011902190319041905190619071908190919101911191219131914191519161917191819191920192119221923192419251926192719281929193019311932193319341935193619371938193919401941194219431944194519461947194819491950195119521953
  1. #
  2. # Include user definition
  3. #
  4. # TO suppress recursive includes
  5. INCLUDED = 1
  6. ifndef TOPDIR
  7. TOPDIR = .
  8. endif
  9. ifndef RELAPACK_REPLACE
  10. RELAPACK_REPLACE=0
  11. endif
  12. # we need to use the host system's architecture for getarch compile options even especially when cross-compiling
  13. HOSTARCH := $(shell uname -m)
  14. ifeq ($(HOSTARCH), amd64)
  15. HOSTARCH=x86_64
  16. endif
  17. # Catch conflicting usage of ARCH in some BSD environments
  18. ifeq ($(ARCH), amd64)
  19. override ARCH=x86_64
  20. else ifeq ($(ARCH), powerpc64)
  21. override ARCH=power
  22. else ifeq ($(ARCH), powerpc64le)
  23. override ARCH=power
  24. else ifeq ($(ARCH), powerpc)
  25. override ARCH=power
  26. else ifeq ($(ARCH), i386)
  27. override ARCH=x86
  28. else ifeq ($(ARCH), armv6)
  29. override ARCH=arm
  30. else ifeq ($(ARCH), armv7)
  31. override ARCH=arm
  32. else ifeq ($(ARCH), aarch64)
  33. override ARCH=arm64
  34. else ifeq ($(ARCH), mipsel)
  35. override ARCH=mips
  36. else ifeq ($(ARCH), mips64el)
  37. override ARCH=mips64
  38. else ifeq ($(ARCH), zarch)
  39. override ARCH=zarch
  40. endif
  41. NETLIB_LAPACK_DIR = $(TOPDIR)/lapack-netlib
  42. # Default C compiler
  43. # - Only set if not specified on the command line or inherited from the environment.
  44. # - CC is an implicit variable so neither '?=' or 'ifndef' can be used.
  45. # http://stackoverflow.com/questions/4029274/mingw-and-make-variables
  46. # - Default value is 'cc' which is not always a valid command (e.g. MinGW).
  47. ifeq ($(origin CC),default)
  48. # Check if $(CC) refers to a valid command and set the value to gcc if not
  49. ifneq ($(findstring cmd.exe,$(SHELL)),)
  50. ifeq ($(shell where $(CC) 2>NUL),)
  51. CC = gcc
  52. endif
  53. else # POSIX-ish
  54. ifeq ($(shell command -v $(CC) 2>/dev/null),)
  55. ifeq ($(shell uname -s),Darwin)
  56. CC = clang
  57. # EXTRALIB += -Wl,-no_compact_unwind
  58. else
  59. CC = gcc
  60. endif # Darwin
  61. endif # CC exists
  62. endif # Shell is sane
  63. endif # CC is set to default
  64. # Default Fortran compiler (FC) is selected by f_check.
  65. ifndef MAKEFILE_RULE
  66. include $(TOPDIR)/Makefile.rule
  67. else
  68. include $(TOPDIR)/$(MAKEFILE_RULE)
  69. endif
  70. #
  71. # Beginning of system configuration
  72. #
  73. ifneq ($(BUILD_SINGLE),1)
  74. ifneq ($(BUILD_DOUBLE),1)
  75. ifneq ($(BUILD_COMPLEX),1)
  76. ifneq ($(BUILD_COMPLEX16),1)
  77. override BUILD_SINGLE=1
  78. override BUILD_DOUBLE=1
  79. override BUILD_COMPLEX=1
  80. override BUILD_COMPLEX16=1
  81. endif
  82. endif
  83. endif
  84. endif
  85. ifndef HOSTCC
  86. HOSTCC = $(CC)
  87. endif
  88. ifdef TARGET
  89. GETARCH_FLAGS := -DFORCE_$(TARGET)
  90. GETARCH_FLAGS += -DUSER_TARGET
  91. ifeq ($(TARGET), GENERIC)
  92. ifeq ($(DYNAMIC_ARCH), 1)
  93. override NO_EXPRECISION=1
  94. export NO_EXPRECISION
  95. endif
  96. endif
  97. endif
  98. # Force fallbacks for 32bit
  99. ifeq ($(BINARY), 32)
  100. ifeq ($(TARGET), HASWELL)
  101. GETARCH_FLAGS := -DFORCE_NEHALEM
  102. endif
  103. ifeq ($(TARGET), SKYLAKEX)
  104. GETARCH_FLAGS := -DFORCE_NEHALEM
  105. endif
  106. ifeq ($(TARGET), COOPERLAKE)
  107. GETARCH_FLAGS := -DFORCE_NEHALEM
  108. endif
  109. ifeq ($(TARGET), SAPPHIRERAPIDS)
  110. GETARCH_FLAGS := -DFORCE_NEHALEM
  111. endif
  112. ifeq ($(TARGET), SANDYBRIDGE)
  113. GETARCH_FLAGS := -DFORCE_NEHALEM
  114. endif
  115. ifeq ($(TARGET), BULLDOZER)
  116. GETARCH_FLAGS := -DFORCE_BARCELONA
  117. endif
  118. ifeq ($(TARGET), PILEDRIVER)
  119. GETARCH_FLAGS := -DFORCE_BARCELONA
  120. endif
  121. ifeq ($(TARGET), STEAMROLLER)
  122. GETARCH_FLAGS := -DFORCE_BARCELONA
  123. endif
  124. ifeq ($(TARGET), EXCAVATOR)
  125. GETARCH_FLAGS := -DFORCE_BARCELONA
  126. endif
  127. ifeq ($(TARGET), ZEN)
  128. GETARCH_FLAGS := -DFORCE_BARCELONA
  129. endif
  130. ifeq ($(TARGET), ARMV8)
  131. GETARCH_FLAGS := -DFORCE_ARMV7
  132. endif
  133. ifeq ($(TARGET), POWER8)
  134. GETARCH_FLAGS := -DFORCE_POWER6
  135. endif
  136. ifeq ($(TARGET), POWER9)
  137. GETARCH_FLAGS := -DFORCE_POWER6
  138. endif
  139. ifeq ($(TARGET), POWER10)
  140. GETARCH_FLAGS := -DFORCE_POWER6
  141. endif
  142. endif
  143. #TARGET_CORE will override TARGET which is used in DYNAMIC_ARCH=1.
  144. #
  145. ifdef TARGET_CORE
  146. GETARCH_FLAGS := -DFORCE_$(TARGET_CORE)
  147. endif
  148. # Force fallbacks for 32bit
  149. ifeq ($(BINARY), 32)
  150. ifeq ($(TARGET_CORE), HASWELL)
  151. GETARCH_FLAGS := -DFORCE_NEHALEM
  152. endif
  153. ifeq ($(TARGET_CORE), SKYLAKEX)
  154. GETARCH_FLAGS := -DFORCE_NEHALEM
  155. endif
  156. ifeq ($(TARGET_CORE), COOPERLAKE)
  157. GETARCH_FLAGS := -DFORCE_NEHALEM
  158. endif
  159. ifeq ($(TARGET_CORE), SAPPHIRERAPIDS)
  160. GETARCH_FLAGS := -DFORCE_NEHALEM
  161. endif
  162. ifeq ($(TARGET_CORE), SANDYBRIDGE)
  163. GETARCH_FLAGS := -DFORCE_NEHALEM
  164. endif
  165. ifeq ($(TARGET_CORE), BULLDOZER)
  166. GETARCH_FLAGS := -DFORCE_BARCELONA
  167. endif
  168. ifeq ($(TARGET_CORE), PILEDRIVER)
  169. GETARCH_FLAGS := -DFORCE_BARCELONA
  170. endif
  171. ifeq ($(TARGET_CORE), STEAMROLLER)
  172. GETARCH_FLAGS := -DFORCE_BARCELONA
  173. endif
  174. ifeq ($(TARGET_CORE), EXCAVATOR)
  175. GETARCH_FLAGS := -DFORCE_BARCELONA
  176. endif
  177. ifeq ($(TARGET_CORE), ZEN)
  178. GETARCH_FLAGS := -DFORCE_BARCELONA
  179. endif
  180. endif
  181. # On x86_64 build getarch with march=native unless the compiler is PGI. This is required to detect AVX512 support in getarch.
  182. ifeq ($(HOSTARCH), x86_64)
  183. ifeq ($(findstring pgcc,$(HOSTCC))$(findstring nvc,$(HOSTCC)),)
  184. GETARCH_FLAGS += -march=native
  185. endif
  186. endif
  187. ifdef INTERFACE64
  188. ifneq ($(INTERFACE64), 0)
  189. GETARCH_FLAGS += -DUSE64BITINT
  190. endif
  191. endif
  192. ifndef GEMM_MULTITHREAD_THRESHOLD
  193. GEMM_MULTITHREAD_THRESHOLD=4
  194. endif
  195. GETARCH_FLAGS += -DGEMM_MULTITHREAD_THRESHOLD=$(GEMM_MULTITHREAD_THRESHOLD)
  196. ifeq ($(NO_AVX), 1)
  197. GETARCH_FLAGS += -DNO_AVX
  198. endif
  199. ifeq ($(BINARY), 32)
  200. GETARCH_FLAGS += -DNO_AVX -DNO_AVX2 -DNO_AVX512
  201. NO_AVX512 = 1
  202. endif
  203. ifeq ($(NO_AVX2), 1)
  204. GETARCH_FLAGS += -DNO_AVX2
  205. endif
  206. ifeq ($(NO_AVX512), 1)
  207. GETARCH_FLAGS += -DNO_AVX512
  208. endif
  209. ifeq ($(DEBUG), 1)
  210. GETARCH_FLAGS += -g
  211. endif
  212. ifeq ($(QUIET_MAKE), 1)
  213. MAKE += -s
  214. endif
  215. ifndef NO_PARALLEL_MAKE
  216. NO_PARALLEL_MAKE=0
  217. endif
  218. GETARCH_FLAGS += -DNO_PARALLEL_MAKE=$(NO_PARALLEL_MAKE)
  219. ifdef MAKE_NB_JOBS
  220. GETARCH_FLAGS += -DMAKE_NB_JOBS=$(MAKE_NB_JOBS)
  221. endif
  222. ifeq ($(HOSTCC), loongcc)
  223. GETARCH_FLAGS += -static
  224. endif
  225. #if don't use Fortran, it will only compile CBLAS.
  226. ifeq ($(ONLY_CBLAS), 1)
  227. NO_LAPACK = 1
  228. else
  229. ONLY_CBLAS = 0
  230. endif
  231. #For small matrix optimization
  232. ifeq ($(ARCH), x86_64)
  233. SMALL_MATRIX_OPT = 1
  234. else ifeq ($(ARCH), power)
  235. SMALL_MATRIX_OPT = 1
  236. BUILD_BFLOAT16 = 1
  237. else ifeq ($(ARCH), arm64)
  238. SMALL_MATRIX_OPT = 1
  239. endif
  240. ifeq ($(ARCH), loongarch64)
  241. SMALL_MATRIX_OPT = 1
  242. endif
  243. ifeq ($(ARCH), arm64)
  244. GEMM_GEMV_FORWARD = 1
  245. GEMM_GEMV_FORWARD_BF16 = 1
  246. endif
  247. ifeq ($(ARCH), riscv)
  248. GEMM_GEMV_FORWARD = 1
  249. endif
  250. ifeq ($(ARCH), power)
  251. GEMM_GEMV_FORWARD = 1
  252. GEMM_GEMV_FORWARD_BF16 = 1
  253. endif
  254. ifeq ($(SMALL_MATRIX_OPT), 1)
  255. CCOMMON_OPT += -DSMALL_MATRIX_OPT
  256. endif
  257. ifneq ($(ONLY_CBLAS), 1)
  258. ifeq ($(GEMM_GEMV_FORWARD), 1)
  259. CCOMMON_OPT += -DGEMM_GEMV_FORWARD
  260. endif
  261. ifeq ($(GEMM_GEMV_FORWARD_BF16), 1)
  262. CCOMMON_OPT += -DGEMM_GEMV_FORWARD_BF16
  263. endif
  264. endif
  265. # This operation is expensive, so execution should be once.
  266. ifndef GOTOBLAS_MAKEFILE
  267. export GOTOBLAS_MAKEFILE = 1
  268. # Generating Makefile.conf and config.h
  269. DUMMY := $(shell $(MAKE) -C $(TOPDIR) -f Makefile.prebuild CC="$(CC)" FC="$(FC)" HOSTCC="$(HOSTCC)" HOST_CFLAGS="$(GETARCH_FLAGS)" CFLAGS="$(CFLAGS)" BINARY=$(BINARY) USE_OPENMP=$(USE_OPENMP) DYNAMIC_ARCH=$(DYNAMIC_ARCH) TARGET_CORE=$(TARGET_CORE) ONLY_CBLAS=$(ONLY_CBLAS) TARGET=$(TARGET) all)
  270. endif
  271. ifndef TARGET_CORE
  272. -include $(TOPDIR)/Makefile.conf
  273. else
  274. HAVE_NEON=
  275. HAVE_VFP=
  276. HAVE_VFPV3=
  277. HAVE_VFPV4=
  278. HAVE_MMX=
  279. HAVE_SSE=
  280. HAVE_SSE2=
  281. HAVE_SSE3=
  282. HAVE_SSSE3=
  283. HAVE_SSE4_1=
  284. HAVE_SSE4_2=
  285. HAVE_SSE4A=
  286. HAVE_SSE5=
  287. HAVE_AVX=
  288. HAVE_AVX2=
  289. HAVE_FMA3=
  290. include $(TOPDIR)/Makefile_kernel.conf
  291. endif
  292. ifndef NUM_PARALLEL
  293. NUM_PARALLEL = 1
  294. endif
  295. ifndef NUM_THREADS
  296. NUM_THREADS = $(NUM_CORES)
  297. endif
  298. ifeq ($(NUM_THREADS), 1)
  299. override USE_THREAD = 0
  300. override USE_OPENMP = 0
  301. endif
  302. ifdef USE_THREAD
  303. ifeq ($(USE_THREAD), 0)
  304. SMP =
  305. else
  306. SMP = 1
  307. endif
  308. else
  309. ifeq ($(NUM_THREADS), 1)
  310. SMP =
  311. else
  312. SMP = 1
  313. endif
  314. endif
  315. ifeq ($(SMP), 1)
  316. USE_LOCKING =
  317. endif
  318. ifndef NEED_PIC
  319. NEED_PIC = 1
  320. endif
  321. ARFLAGS =
  322. CPP = $(COMPILER) -E
  323. AR ?= $(CROSS_SUFFIX)ar
  324. AS ?= $(CROSS_SUFFIX)as
  325. LD ?= $(CROSS_SUFFIX)ld
  326. RANLIB ?= $(CROSS_SUFFIX)ranlib
  327. NM = $(CROSS_SUFFIX)nm
  328. DLLWRAP = $(CROSS_SUFFIX)dllwrap
  329. OBJCOPY = $(CROSS_SUFFIX)objcopy
  330. OBJCONV = $(CROSS_SUFFIX)objconv
  331. # When fortran support was either not detected or actively deselected, only build BLAS.
  332. ifeq ($(NOFORTRAN), 1)
  333. C_LAPACK = 1
  334. override FEXTRALIB =
  335. endif
  336. ifeq ($(C_COMPILER), GCC)
  337. GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
  338. GCCVERSIONGT4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 4)
  339. GCCVERSIONGT5 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 5)
  340. GCCVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 7)
  341. GCCVERSIONGTEQ8 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 8)
  342. GCCVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  343. GCCVERSIONGTEQ10 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 10)
  344. GCCVERSIONGTEQ11 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 11)
  345. GCCVERSIONGTEQ12 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 12)
  346. # Note that the behavior of -dumpversion is compile-time-configurable for
  347. # gcc-7.x and newer. Use -dumpfullversion there
  348. ifeq ($(GCCVERSIONGTEQ7),1)
  349. GCCDUMPVERSION_PARAM := -dumpfullversion
  350. else
  351. GCCDUMPVERSION_PARAM := -dumpversion
  352. endif
  353. GCCMINORVERSIONGTEQ1 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 1)
  354. GCCMINORVERSIONGTEQ2 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 2)
  355. GCCMINORVERSIONGTEQ4 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 4)
  356. GCCMINORVERSIONGTEQ7 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 7)
  357. endif
  358. ifeq ($(C_COMPILER), CLANG)
  359. CLANGVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  360. CLANGVERSIONGTEQ12 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 12)
  361. endif
  362. #
  363. # OS dependent settings
  364. #
  365. ifeq ($(OSNAME), Darwin)
  366. ifndef MACOSX_DEPLOYMENT_TARGET
  367. ifeq ($(ARCH), arm64)
  368. export MACOSX_DEPLOYMENT_TARGET=11.0
  369. ifeq ($(C_COMPILER), GCC)
  370. export NO_SVE = 1
  371. export NO_SME = 1
  372. endif
  373. else
  374. export MACOSX_DEPLOYMENT_TARGET=10.8
  375. endif
  376. endif
  377. MD5SUM = md5 -r
  378. XCVER = $(shell pkgutil --pkg-info=com.apple.pkg.Xcode |awk '/version:/ {print $2}'|cut -d: -f2|cut -f1 -d.)
  379. ifeq (x$(XCVER)x,xx)
  380. XCVER = $(shell pkgutil --pkg-info=com.apple.pkg.CLTools_Executables |awk '/version:/ {print $2}'|cut -d: -f2|cut -f1 -d.)
  381. endif
  382. ifeq (x$(XCVER), x 15)
  383. CCOMMON_OPT += -Wl,-ld_classic
  384. FCOMMON_OPT += -Wl,-ld_classic
  385. endif
  386. ifeq (x$(XCVER), x 16)
  387. ifeq ($(F_COMPILER), GFORTRAN)
  388. override CEXTRALIB := $(filter-out(-lto_library, $(CEXTRALIB)))
  389. endif
  390. endif
  391. endif
  392. ifneq (,$(findstring $(OSNAME), FreeBSD OpenBSD DragonFly))
  393. MD5SUM = md5 -r
  394. endif
  395. ifeq ($(OSNAME), NetBSD)
  396. MD5SUM = md5 -n
  397. endif
  398. ifeq ($(OSNAME), Linux)
  399. EXTRALIB += -lm
  400. #NO_EXPRECISION = 1
  401. endif
  402. ifeq ($(OSNAME), Android)
  403. EXTRALIB += -lm
  404. endif
  405. ifeq ($(OSNAME), AIX)
  406. EXTRALIB += -lm
  407. endif
  408. ifeq ($(OSNAME), $(filter $(OSNAME),FreeBSD OpenBSD NetBSD DragonFly))
  409. ifeq ($(ARCH), $(filter $(ARCH),arm arm64))
  410. EXTRALIB += -lm
  411. endif
  412. endif
  413. ifeq ($(OSNAME), WINNT)
  414. NEED_PIC = 0
  415. NO_EXPRECISION = 1
  416. EXTRALIB += -defaultlib:advapi32
  417. SUFFIX = obj
  418. PSUFFIX = pobj
  419. LIBSUFFIX = a
  420. ifeq ($(C_COMPILER), CLANG)
  421. CCOMMON_OPT += -DMS_ABI
  422. endif
  423. #Version tests for supporting specific features (MS_ABI, POWER9 intrinsics)
  424. ifeq ($(GCCVERSIONGT4), 1)
  425. # GCC Major version > 4
  426. # It is compatible with MSVC ABI.
  427. CCOMMON_OPT += -DMS_ABI
  428. endif
  429. ifeq ($(GCCVERSIONGTEQ4), 1)
  430. ifeq ($(GCCMINORVERSIONGTEQ7), 1)
  431. # GCC Version >=4.7
  432. # It is compatible with MSVC ABI.
  433. CCOMMON_OPT += -DMS_ABI
  434. endif
  435. endif
  436. # Ensure the correct stack alignment on Win32
  437. # http://permalink.gmane.org/gmane.comp.lib.openblas.general/97
  438. ifeq ($(ARCH), x86)
  439. CCOMMON_OPT += -mincoming-stack-boundary=2
  440. FCOMMON_OPT += -mincoming-stack-boundary=2
  441. endif
  442. endif
  443. ifeq ($(OSNAME), Interix)
  444. NEED_PIC = 0
  445. NO_EXPRECISION = 1
  446. INTERIX_TOOL_DIR = /opt/gcc.3.3/i586-pc-interix3/bin
  447. endif
  448. ifeq ($(OSNAME), CYGWIN_NT)
  449. NEED_PIC = 0
  450. NO_EXPRECISION = 1
  451. OS_CYGWIN_NT = 1
  452. endif
  453. ifneq ($(OSNAME), WINNT)
  454. ifneq ($(OSNAME), CYGWIN_NT)
  455. ifneq ($(OSNAME), Interix)
  456. ifneq ($(OSNAME), Android)
  457. ifdef SMP
  458. EXTRALIB += -lpthread
  459. endif
  460. endif
  461. endif
  462. endif
  463. endif
  464. # ifeq logical or
  465. ifeq ($(OSNAME), $(filter $(OSNAME),WINNT CYGWIN_NT Interix))
  466. OS_WINDOWS=1
  467. endif
  468. ifdef QUAD_PRECISION
  469. CCOMMON_OPT += -DQUAD_PRECISION
  470. NO_EXPRECISION = 1
  471. endif
  472. ifneq ($(ARCH), x86)
  473. ifneq ($(ARCH), x86_64)
  474. NO_EXPRECISION = 1
  475. endif
  476. endif
  477. ifdef UTEST_CHECK
  478. CCOMMON_OPT += -DUTEST_CHECK
  479. SANITY_CHECK = 1
  480. endif
  481. ifdef SANITY_CHECK
  482. CCOMMON_OPT += -DSANITY_CHECK -DREFNAME=$(*F)f$(BU)
  483. endif
  484. MAX_STACK_ALLOC ?= 2048
  485. ifneq ($(MAX_STACK_ALLOC), 0)
  486. CCOMMON_OPT += -DMAX_STACK_ALLOC=$(MAX_STACK_ALLOC)
  487. endif
  488. ifdef USE_LOCKING
  489. ifneq ($(USE_LOCKING), 0)
  490. CCOMMON_OPT += -DUSE_LOCKING
  491. endif
  492. endif
  493. #
  494. # Architecture dependent settings
  495. #
  496. ifeq ($(ARCH), x86)
  497. ifndef BINARY
  498. NO_BINARY_MODE = 1
  499. endif
  500. ifeq ($(CORE), generic)
  501. #NO_EXPRECISION = 1
  502. endif
  503. ifndef NO_EXPRECISION
  504. ifeq ($(F_COMPILER), GFORTRAN)
  505. # ifeq logical or. GCC or LSB
  506. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  507. EXPRECISION = 1
  508. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  509. FCOMMON_OPT += -m128bit-long-double
  510. endif
  511. ifeq ($(C_COMPILER), CLANG)
  512. EXPRECISION = 1
  513. CCOMMON_OPT += -DEXPRECISION
  514. FCOMMON_OPT += -m128bit-long-double
  515. endif
  516. endif
  517. endif
  518. endif
  519. ifeq ($(ARCH), x86_64)
  520. ifeq ($(CORE), generic)
  521. #NO_EXPRECISION = 1
  522. endif
  523. ifndef NO_EXPRECISION
  524. ifeq ($(F_COMPILER), GFORTRAN)
  525. # ifeq logical or. GCC or LSB
  526. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  527. EXPRECISION = 1
  528. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  529. FCOMMON_OPT += -m128bit-long-double
  530. endif
  531. ifeq ($(C_COMPILER), CLANG)
  532. EXPRECISION = 1
  533. CCOMMON_OPT += -DEXPRECISION
  534. FCOMMON_OPT += -m128bit-long-double
  535. endif
  536. endif
  537. endif
  538. endif
  539. ifeq ($(C_COMPILER), INTEL)
  540. CCOMMON_OPT += -wd981
  541. endif
  542. ifeq ($(USE_OPENMP), 1)
  543. #check
  544. ifeq ($(USE_THREAD), 0)
  545. $(error OpenBLAS: Cannot set both USE_OPENMP=1 and USE_THREAD=0. The USE_THREAD=0 is only for building single thread version.)
  546. endif
  547. # ifeq logical or. GCC or LSB
  548. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  549. CCOMMON_OPT += -fopenmp
  550. endif
  551. ifeq ($(C_COMPILER), CLANG)
  552. CCOMMON_OPT += -fopenmp
  553. ifeq ($(F_COMPILER), GFORTRAN)
  554. FEXTRALIB := $(subst -lgomp,-lomp,$(FEXTRALIB))
  555. endif
  556. endif
  557. ifeq ($(C_COMPILER), INTEL)
  558. CCOMMON_OPT += -fopenmp
  559. endif
  560. ifeq ($(C_COMPILER), PGI)
  561. CCOMMON_OPT += -mp
  562. endif
  563. ifeq ($(C_COMPILER), OPEN64)
  564. CCOMMON_OPT += -mp
  565. CEXTRALIB += -lstdc++
  566. endif
  567. ifeq ($(C_COMPILER), PATHSCALE)
  568. CCOMMON_OPT += -mp
  569. endif
  570. endif
  571. ifeq ($(DYNAMIC_ARCH), 1)
  572. ifeq ($(ARCH), x86)
  573. DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
  574. CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
  575. endif
  576. ifeq ($(ARCH), x86_64)
  577. DYNAMIC_CORE = PRESCOTT CORE2
  578. ifeq ($(DYNAMIC_OLDER), 1)
  579. DYNAMIC_CORE += PENRYN DUNNINGTON
  580. endif
  581. DYNAMIC_CORE += NEHALEM
  582. ifeq ($(DYNAMIC_OLDER), 1)
  583. DYNAMIC_CORE += OPTERON OPTERON_SSE3
  584. endif
  585. DYNAMIC_CORE += BARCELONA
  586. ifeq ($(DYNAMIC_OLDER), 1)
  587. DYNAMIC_CORE += BOBCAT ATOM NANO
  588. endif
  589. ifneq ($(NO_AVX), 1)
  590. DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER STEAMROLLER EXCAVATOR
  591. endif
  592. ifneq ($(NO_AVX2), 1)
  593. DYNAMIC_CORE += HASWELL ZEN
  594. endif
  595. ifneq ($(NO_AVX512), 1)
  596. ifneq ($(NO_AVX2), 1)
  597. DYNAMIC_CORE += SKYLAKEX COOPERLAKE SAPPHIRERAPIDS
  598. endif
  599. endif
  600. endif
  601. ifdef DYNAMIC_LIST
  602. override DYNAMIC_CORE = PRESCOTT $(DYNAMIC_LIST)
  603. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_PRESCOTT
  604. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  605. CCOMMON_OPT += $(XCCOMMON_OPT)
  606. #CCOMMON_OPT += -DDYNAMIC_LIST='$(DYNAMIC_LIST)'
  607. endif
  608. ifeq ($(ARCH), arm64)
  609. DYNAMIC_CORE = ARMV8
  610. DYNAMIC_CORE += CORTEXA53
  611. DYNAMIC_CORE += CORTEXA57
  612. DYNAMIC_CORE += NEOVERSEN1
  613. ifneq ($(NO_SVE), 1)
  614. DYNAMIC_CORE += NEOVERSEV1
  615. DYNAMIC_CORE += NEOVERSEN2
  616. DYNAMIC_CORE += ARMV8SVE
  617. DYNAMIC_CORE += A64FX
  618. endif
  619. ifneq ($(NO_SME), 1)
  620. DYNAMIC_CORE += ARMV9SME
  621. endif
  622. DYNAMIC_CORE += THUNDERX
  623. DYNAMIC_CORE += THUNDERX2T99
  624. DYNAMIC_CORE += TSV110
  625. DYNAMIC_CORE += EMAG8180
  626. DYNAMIC_CORE += THUNDERX3T110
  627. ifdef DYNAMIC_LIST
  628. override DYNAMIC_CORE = ARMV8 $(DYNAMIC_LIST)
  629. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_ARMV8
  630. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  631. endif
  632. endif
  633. ifeq ($(ARCH), mips64)
  634. DYNAMIC_CORE = LOONGSON3R3 LOONGSON3R4 MIPS64_GENERIC
  635. ifdef DYNAMIC_LIST
  636. override DYNAMIC_CORE = MIPS64_GENERIC $(DYNAMIC_LIST)
  637. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_MIPS64_GENERIC
  638. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  639. endif
  640. endif
  641. ifeq ($(ARCH), loongarch64)
  642. DYNAMIC_CORE = LA64_GENERIC LA264 LA464
  643. endif
  644. ifeq ($(ARCH), riscv64)
  645. DYNAMIC_CORE = RISCV64_GENERIC
  646. DYNAMIC_CORE += RISCV64_ZVL128B
  647. DYNAMIC_CORE += RISCV64_ZVL256B
  648. ifdef DYNAMIC_LIST
  649. override DYNAMIC_CORE = RISCV64_GENERIC $(DYNAMIC_LIST)
  650. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_RISCV64_GENERIC
  651. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  652. endif
  653. endif
  654. ifeq ($(ARCH), zarch)
  655. DYNAMIC_CORE = ZARCH_GENERIC
  656. # if the compiler accepts -march=arch11 or -march=z13 and can compile a file
  657. # with z13-specific inline assembly, then we can include support for Z13.
  658. # note: -march=z13 is equivalent to -march=arch11 yet some compiler releases
  659. # only support one or the other.
  660. # note: LLVM version 6.x supported -march=z13 yet could not handle vector
  661. # registers in inline assembly, so the check for supporting the -march flag is
  662. # not enough.
  663. ZARCH_TEST_COMPILE=-c $(TOPDIR)/kernel/zarch/damin_z13.c -I$(TOPDIR) -o /dev/null > /dev/null 2> /dev/null
  664. ZARCH_CC_SUPPORTS_ARCH11=$(shell $(CC) -march=arch11 $(ZARCH_TEST_COMPILE) && echo 1)
  665. ZARCH_CC_SUPPORTS_Z13=$(shell $(CC) -march=z13 $(ZARCH_TEST_COMPILE) && echo 1)
  666. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH11), $(ZARCH_CC_SUPPORTS_Z13)), 1)
  667. DYNAMIC_CORE += Z13
  668. CCOMMON_OPT += -DDYN_Z13
  669. else
  670. $(info OpenBLAS: Not building Z13 kernels because the compiler $(CC) does not support it)
  671. endif
  672. # as above for z13, check for -march=arch12 and z14 support in the compiler.
  673. ZARCH_CC_SUPPORTS_ARCH12=$(shell $(CC) -march=arch12 $(ZARCH_TEST_COMPILE) && echo 1)
  674. ZARCH_CC_SUPPORTS_Z14=$(shell $(CC) -march=z14 $(ZARCH_TEST_COMPILE) && echo 1)
  675. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH12), $(ZARCH_CC_SUPPORTS_Z14)), 1)
  676. DYNAMIC_CORE += Z14
  677. CCOMMON_OPT += -DDYN_Z14
  678. else
  679. $(info OpenBLAS: Not building Z14 kernels because the compiler $(CC) does not support it)
  680. endif
  681. endif # ARCH zarch
  682. ifeq ($(ARCH), power)
  683. ifneq ($(C_COMPILER), PGI)
  684. DYNAMIC_CORE = POWER6
  685. DYNAMIC_CORE += POWER8
  686. ifneq ($(C_COMPILER), GCC)
  687. DYNAMIC_CORE += POWER9
  688. DYNAMIC_CORE += POWER10
  689. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  690. endif
  691. ifeq ($(C_COMPILER), GCC)
  692. ifeq ($(GCCVERSIONGT5), 1)
  693. DYNAMIC_CORE += POWER9
  694. else
  695. $(info, OpenBLAS: Your gcc version is too old to build the POWER9 kernels.)
  696. endif
  697. ifeq ($(OSNAME), AIX)
  698. LDVERSIONGTEQ35 := 1
  699. else
  700. LDVERSIONGTEQ35 := $(shell expr `$(CC) -Wl,--version 2> /dev/null | head -1 | cut -f2 -d "." | cut -f1 -d "-"` \>= 35)
  701. endif
  702. ifeq ($(GCCVERSIONGTEQ11)$(LDVERSIONGTEQ35), 11)
  703. DYNAMIC_CORE += POWER10
  704. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  705. else ifeq ($(GCCVERSIONGTEQ10), 1)
  706. ifeq ($(GCCMINORVERSIONGTEQ2)$(LDVERSIONGTEQ35), 11)
  707. DYNAMIC_CORE += POWER10
  708. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  709. endif
  710. else
  711. $(info, OpenBLAS: Your gcc version is too old to build the POWER10 kernels.)
  712. endif
  713. endif
  714. else
  715. DYNAMIC_CORE = POWER8
  716. DYNAMIC_CORE += POWER9
  717. endif
  718. endif
  719. # If DYNAMIC_CORE is not set, DYNAMIC_ARCH cannot do anything, so force it to empty
  720. ifndef DYNAMIC_CORE
  721. override DYNAMIC_ARCH=
  722. endif
  723. endif
  724. ifeq ($(ARCH), ia64)
  725. NO_BINARY_MODE = 1
  726. BINARY_DEFINED = 1
  727. ifeq ($(F_COMPILER), GFORTRAN)
  728. ifeq ($(C_COMPILER), GCC)
  729. EXPRECISION = 1
  730. CCOMMON_OPT += -DEXPRECISION
  731. endif
  732. endif
  733. endif
  734. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  735. NO_BINARY_MODE = 1
  736. endif
  737. ifeq ($(ARCH), alpha)
  738. NO_BINARY_MODE = 1
  739. BINARY_DEFINED = 1
  740. endif
  741. ifeq ($(ARCH), arm)
  742. NO_BINARY_MODE = 1
  743. BINARY_DEFINED = 1
  744. ifneq ($(EMBEDDED), 1)
  745. CCOMMON_OPT += -marm
  746. FCOMMON_OPT += -marm
  747. else
  748. CCOMMON_OPT += -DOS_EMBEDDED -mthumb -mcpu=cortex-m4 -mfloat-abi=hard -mfpu=fpv4-sp-d16
  749. endif
  750. # If softfp abi is mentioned on the command line, force it.
  751. ifeq ($(ARM_SOFTFP_ABI), 1)
  752. CCOMMON_OPT += -mfloat-abi=softfp
  753. FCOMMON_OPT += -mfloat-abi=softfp
  754. endif
  755. ifeq ($(OSNAME), Android)
  756. ifeq ($(ARM_SOFTFP_ABI), 1)
  757. EXTRALIB += -lm
  758. else
  759. EXTRALIB += -Wl,-lm_hard
  760. endif
  761. endif
  762. endif
  763. ifeq ($(ARCH), arm64)
  764. NO_BINARY_MODE = 1
  765. BINARY_DEFINED = 1
  766. ifdef INTERFACE64
  767. ifneq ($(INTERFACE64), 0)
  768. ifeq ($(F_COMPILER), GFORTRAN)
  769. FCOMMON_OPT += -fdefault-integer-8
  770. endif
  771. ifeq ($(F_COMPILER), FLANG)
  772. FCOMMON_OPT += -i8
  773. endif
  774. endif
  775. endif
  776. endif
  777. ifeq ($(ARCH), riscv64)
  778. NO_BINARY_MODE = 1
  779. BINARY_DEFINED = 1
  780. ifdef INTERFACE64
  781. ifneq ($(INTERFACE64), 0)
  782. ifeq ($(F_COMPILER), GFORTRAN)
  783. FCOMMON_OPT += -fdefault-integer-8
  784. endif
  785. ifeq ($(F_COMPILER), FLANG)
  786. FCOMMON_OPT += -i8
  787. endif
  788. endif
  789. endif
  790. endif
  791. ifeq ($(ARCH), loongarch64)
  792. NO_BINARY_MODE = 1
  793. BINARY_DEFINED = 1
  794. ifdef INTERFACE64
  795. ifneq ($(INTERFACE64), 0)
  796. ifeq ($(F_COMPILER), GFORTRAN)
  797. FCOMMON_OPT += -fdefault-integer-8
  798. endif
  799. ifeq ($(F_COMPILER), FLANG)
  800. FCOMMON_OPT += -i8
  801. endif
  802. endif
  803. endif
  804. endif
  805. ifeq ($(ARCH), csky)
  806. NO_BINARY_MODE = 1
  807. BINARY_DEFINED = 1
  808. endif
  809. #
  810. # C Compiler dependent settings
  811. #
  812. # ifeq logical or. GCC or CLANG or LSB
  813. # http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
  814. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG LSB))
  815. CCOMMON_OPT += -Wall
  816. COMMON_PROF += -fno-inline
  817. NO_UNINITIALIZED_WARN = -Wno-uninitialized
  818. ifeq ($(QUIET_MAKE), 1)
  819. CCOMMON_OPT += $(NO_UNINITIALIZED_WARN) -Wno-unused
  820. endif
  821. ifdef NO_BINARY_MODE
  822. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  823. ifdef BINARY64
  824. CCOMMON_OPT += -mabi=64
  825. else
  826. CCOMMON_OPT += -mabi=n32
  827. endif
  828. BINARY_DEFINED = 1
  829. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  830. CCOMMON_OPT += -mabi=32
  831. BINARY_DEFINED = 1
  832. endif
  833. ifneq (, $(filter $(CORE), MIPS64_GENERIC))
  834. CCOMMON_OPT += -DNO_MSA
  835. FCOMMON_OPT += -DNO_MSA
  836. endif
  837. ifneq (, $(filter $(CORE),LOONGSON3R3 LOONGSON3R4))
  838. CCOMMON_OPT += -march=loongson3a
  839. FCOMMON_OPT += -march=loongson3a
  840. endif
  841. ifeq ($(CORE), MIPS24K)
  842. CCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  843. FCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  844. endif
  845. ifeq ($(CORE), MIPS1004K)
  846. CCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  847. FCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  848. endif
  849. ifeq ($(CORE), P5600)
  850. CCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  851. FCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  852. endif
  853. ifeq ($(CORE), I6400)
  854. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  855. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  856. endif
  857. ifeq ($(CORE), P6600)
  858. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  859. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  860. endif
  861. ifeq ($(CORE), I6500)
  862. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  863. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  864. endif
  865. ifeq ($(OSNAME), AIX)
  866. BINARY_DEFINED = 1
  867. endif
  868. ifeq ($(ARCH), loongarch64)
  869. LA64_ABI=$(shell $(CC) -mabi=lp64d -c $(TOPDIR)/cpuid_loongarch64.c -o /dev/null > /dev/null 2> /dev/null && echo lp64d)
  870. LA64_ARCH=$(shell $(CC) -march=loongarch64 -c $(TOPDIR)/cpuid_loongarch64.c -o /dev/null > /dev/null 2> /dev/null && echo loongarch64)
  871. ifneq ($(LA64_ABI), lp64d)
  872. LA64_ABI=lp64
  873. endif
  874. ifneq ($(LA64_ARCH), loongarch64)
  875. CCOMMON_OPT += -mabi=$(LA64_ABI)
  876. FCOMMON_OPT += -mabi=$(LA64_ABI)
  877. else
  878. CCOMMON_OPT += -march=loongarch64 -mabi=$(LA64_ABI)
  879. FCOMMON_OPT += -march=loongarch64 -mabi=$(LA64_ABI)
  880. endif
  881. endif
  882. endif
  883. ifndef BINARY_DEFINED
  884. ifneq ($(OSNAME), AIX)
  885. ifdef BINARY64
  886. ifneq ($(ARCH), riscv64)
  887. CCOMMON_OPT += -m64
  888. endif
  889. else
  890. CCOMMON_OPT += -m32
  891. endif
  892. endif
  893. endif
  894. endif
  895. ifeq ($(C_COMPILER), PGI)
  896. PGCVERSIONGT20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \> 20)
  897. PGCVERSIONEQ20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` == 20)
  898. PGCMINORVERSIONGE11 := $(shell expr `$(CC) --version|sed -n "2p" |cut -d "-" -f 1 |sed -e "s/[^0-9.]//g" |cut -c 4-5` \>= 11)
  899. PGCVERSIONCHECK := $(PGCVERSIONGT20)$(PGCVERSIONEQ20)$(PGCMINORVERSIONGE11)
  900. ifeq ($(PGCVERSIONCHECK), $(filter $(PGCVERSIONCHECK), 100 101 011))
  901. NEWPGI := 1
  902. PGCVERSIONGT21 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \> 21)
  903. PGCVERSIONEQ21 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` == 21)
  904. PGCVERSIONCHECK2 := $(PGCVERSIONGT21)$(PGCVERSIONEQ21)$(PGCMINORVERSIONGE11)
  905. ifeq ($(PGCVERSIONCHECK2), $(filter $(PGCVERSIONCHECK2), 100 101 011))
  906. NEWPGI2 := 1
  907. endif
  908. endif
  909. ifdef BINARY64
  910. ifeq ($(ARCH), x86_64)
  911. ifeq (,$(findstring tp,$(CFLAGS)))
  912. ifneq ($(NEWPGI2),1)
  913. CCOMMON_OPT += -tp p7-64
  914. else
  915. CCOMMON_OPT += -tp px
  916. endif
  917. endif
  918. ifneq ($(NEWPGI),1)
  919. CCOMMON_OPT += -D__MMX__ -Mnollvm
  920. endif
  921. else
  922. ifeq ($(ARCH), power)
  923. ifeq (,$(findstring tp,$(CFLAGS)))
  924. ifeq ($(CORE), POWER8)
  925. CCOMMON_OPT += -tp pwr8
  926. endif
  927. ifeq ($(CORE), POWER9)
  928. CCOMMON_OPT += -tp pwr9
  929. endif
  930. endif
  931. endif
  932. endif
  933. else
  934. ifneq ($(NEWPGI2),1)
  935. ifeq (,$(findstring tp,$(CFLAGS)))
  936. CCOMMON_OPT += -tp p7
  937. else
  938. CCOMMON_OPT += -tp px
  939. endif
  940. endif
  941. endif
  942. endif
  943. ifeq ($(C_COMPILER), PATHSCALE)
  944. ifdef BINARY64
  945. CCOMMON_OPT += -m64
  946. else
  947. CCOMMON_OPT += -m32
  948. endif
  949. endif
  950. #
  951. # Fortran Compiler dependent settings
  952. #
  953. ifeq ($(F_COMPILER), NAG)
  954. FCOMMON_OPT += -dcfuns -recursive -ieee=full -w=obs -thread_safe
  955. ifdef INTERFACE64
  956. ifneq ($(INTERFACE64), 0)
  957. FCOMMON_OPT += -i8
  958. endif
  959. endif
  960. ifeq ($(USE_OPENMP), 1)
  961. FCOMMON_OPT += -openmp
  962. endif
  963. endif
  964. ifeq ($(F_COMPILER), FLANG)
  965. CCOMMON_OPT += -DF_INTERFACE_FLANG
  966. FCOMMON_OPT += -Mrecursive -Kieee
  967. ifeq ($(OSNAME), Linux)
  968. ifeq ($(ARCH), x86_64)
  969. FLANG_VENDOR := $(shell $(FC) --version|head -1 |cut -f 1 -d " ")
  970. ifeq ($(FLANG_VENDOR), AMD)
  971. FCOMMON_OPT += -fno-unroll-loops
  972. endif
  973. endif
  974. endif
  975. ifdef BINARY64
  976. ifdef INTERFACE64
  977. ifneq ($(INTERFACE64), 0)
  978. FCOMMON_OPT += -i8
  979. endif
  980. endif
  981. FCOMMON_OPT += -Wall
  982. else
  983. FCOMMON_OPT += -Wall
  984. endif
  985. ifeq ($(USE_OPENMP), 1)
  986. FCOMMON_OPT += -fopenmp
  987. endif
  988. endif
  989. ifeq ($(F_COMPILER), G77)
  990. CCOMMON_OPT += -DF_INTERFACE_G77
  991. FCOMMON_OPT += -Wall
  992. ifndef NO_BINARY_MODE
  993. ifneq ($(OSNAME), AIX)
  994. ifdef BINARY64
  995. FCOMMON_OPT += -m64
  996. else
  997. FCOMMON_OPT += -m32
  998. endif
  999. endif
  1000. endif
  1001. endif
  1002. ifeq ($(F_COMPILER), G95)
  1003. CCOMMON_OPT += -DF_INTERFACE_G95
  1004. FCOMMON_OPT += -Wall
  1005. ifneq ($(OSNAME), AIX)
  1006. ifndef NO_BINARY_MODE
  1007. ifdef BINARY64
  1008. FCOMMON_OPT += -m64
  1009. else
  1010. FCOMMON_OPT += -m32
  1011. endif
  1012. endif
  1013. ifneq ($(NO_LAPACKE), 1)
  1014. FCOMMON_OPT += -fno-second-underscore
  1015. endif
  1016. endif
  1017. endif
  1018. ifeq ($(F_COMPILER), $(filter $(F_COMPILER),GFORTRAN FLANGNEW))
  1019. CCOMMON_OPT += -DF_INTERFACE_GFORT
  1020. ifeq ($(F_COMPILER), GFORTRAN)
  1021. FCOMMON_OPT += -Wall
  1022. # make single-threaded LAPACK calls thread-safe #1847
  1023. FCOMMON_OPT += -frecursive
  1024. # work around ABI problem with passing single-character arguments
  1025. FCOMMON_OPT += -fno-optimize-sibling-calls
  1026. #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
  1027. ifneq ($(NOFORTRAN), 1)
  1028. ifneq ($(NOFORTRAN), 2)
  1029. ifneq ($(NO_LAPACK), 1)
  1030. EXTRALIB += -lgfortran
  1031. endif
  1032. endif
  1033. endif
  1034. endif
  1035. ifdef NO_BINARY_MODE
  1036. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  1037. ifdef BINARY64
  1038. FCOMMON_OPT += -mabi=64
  1039. else
  1040. FCOMMON_OPT += -mabi=n32
  1041. endif
  1042. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  1043. FCOMMON_OPT += -mabi=32
  1044. endif
  1045. else
  1046. ifdef BINARY64
  1047. ifneq ($(OSNAME), AIX)
  1048. ifneq ($(ARCH), riscv64)
  1049. FCOMMON_OPT += -m64
  1050. endif
  1051. endif
  1052. ifdef INTERFACE64
  1053. ifneq ($(INTERFACE64), 0)
  1054. FCOMMON_OPT += -fdefault-integer-8
  1055. endif
  1056. endif
  1057. else
  1058. ifneq ($(OSNAME), AIX)
  1059. FCOMMON_OPT += -m32
  1060. endif
  1061. endif
  1062. endif
  1063. ifeq ($(USE_OPENMP), 1)
  1064. FCOMMON_OPT += -fopenmp
  1065. endif
  1066. endif
  1067. ifeq ($(F_COMPILER), INTEL)
  1068. CCOMMON_OPT += -DF_INTERFACE_INTEL
  1069. ifdef INTERFACE64
  1070. ifneq ($(INTERFACE64), 0)
  1071. FCOMMON_OPT += -i8
  1072. endif
  1073. endif
  1074. FCOMMON_OPT += -recursive -fp-model strict -assume protect-parens
  1075. ifeq ($(USE_OPENMP), 1)
  1076. FCOMMON_OPT += -fopenmp
  1077. endif
  1078. endif
  1079. ifeq ($(F_COMPILER), FUJITSU)
  1080. CCOMMON_OPT += -DF_INTERFACE_FUJITSU
  1081. ifeq ($(USE_OPENMP), 1)
  1082. FCOMMON_OPT += -openmp
  1083. endif
  1084. endif
  1085. ifeq ($(F_COMPILER), IBM)
  1086. CCOMMON_OPT += -DF_INTERFACE_IBM
  1087. FEXTRALIB += -lxlf90
  1088. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG))
  1089. FCOMMON_OPT += -qextname -qzerosize
  1090. endif
  1091. # FCOMMON_OPT += -qarch=440
  1092. ifdef BINARY64
  1093. FCOMMON_OPT += -q64
  1094. ifdef INTERFACE64
  1095. ifneq ($(INTERFACE64), 0)
  1096. FCOMMON_OPT += -qintsize=8
  1097. endif
  1098. endif
  1099. else
  1100. FCOMMON_OPT += -q32
  1101. endif
  1102. endif
  1103. ifeq ($(F_COMPILER), PGI)
  1104. CCOMMON_OPT += -DF_INTERFACE_PGI
  1105. COMMON_PROF += -DPGICOMPILER
  1106. ifdef BINARY64
  1107. ifdef INTERFACE64
  1108. ifneq ($(INTERFACE64), 0)
  1109. FCOMMON_OPT += -i8
  1110. endif
  1111. endif
  1112. ifeq ($(ARCH), x86_64)
  1113. ifneq ($(NEWPGI2),1)
  1114. FCOMMON_OPT += -tp p7-64
  1115. else
  1116. FCOMMON_OPT += -tp px
  1117. endif
  1118. else
  1119. ifeq ($(ARCH), power)
  1120. ifeq ($(CORE), POWER6)
  1121. $(warning NVIDIA HPC compilers do not support POWER6.)
  1122. endif
  1123. ifeq ($(CORE), POWER8)
  1124. FCOMMON_OPT += -tp pwr8
  1125. endif
  1126. ifeq ($(CORE), POWER9)
  1127. FCOMMON_OPT += -tp pwr9
  1128. endif
  1129. ifeq ($(CORE), POWER10)
  1130. $(warning NVIDIA HPC compilers do not support POWER10.)
  1131. endif
  1132. endif
  1133. endif
  1134. else
  1135. FCOMMON_OPT += -tp p7
  1136. endif
  1137. FCOMMON_OPT += -Mrecursive -Kieee
  1138. ifeq ($(USE_OPENMP), 1)
  1139. FCOMMON_OPT += -mp
  1140. endif
  1141. endif
  1142. ifeq ($(F_COMPILER), PATHSCALE)
  1143. CCOMMON_OPT += -DF_INTERFACE_PATHSCALE
  1144. ifdef BINARY64
  1145. ifdef INTERFACE64
  1146. ifneq ($(INTERFACE64), 0)
  1147. FCOMMON_OPT += -i8
  1148. endif
  1149. endif
  1150. endif
  1151. ifeq ($(USE_OPENMP), 1)
  1152. FCOMMON_OPT += -mp
  1153. endif
  1154. endif
  1155. ifeq ($(F_COMPILER), OPEN64)
  1156. CCOMMON_OPT += -DF_INTERFACE_OPEN64
  1157. ifdef BINARY64
  1158. ifdef INTERFACE64
  1159. ifneq ($(INTERFACE64), 0)
  1160. FCOMMON_OPT += -i8
  1161. endif
  1162. endif
  1163. endif
  1164. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  1165. ifndef BINARY64
  1166. FCOMMON_OPT += -n32
  1167. else
  1168. FCOMMON_OPT += -n64
  1169. endif
  1170. ifeq ($(CORE), LOONGSON3R3)
  1171. FCOMMON_OPT += -loongson3 -static
  1172. endif
  1173. ifeq ($(CORE), LOONGSON3R4)
  1174. FCOMMON_OPT += -loongson3 -static
  1175. endif
  1176. else
  1177. ifndef BINARY64
  1178. FCOMMON_OPT += -m32
  1179. else
  1180. FCOMMON_OPT += -m64
  1181. endif
  1182. endif
  1183. ifeq ($(USE_OPENMP), 1)
  1184. FEXTRALIB += -lstdc++
  1185. FCOMMON_OPT += -mp
  1186. endif
  1187. endif
  1188. ifeq ($(C_COMPILER), OPEN64)
  1189. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  1190. ifndef BINARY64
  1191. CCOMMON_OPT += -n32
  1192. else
  1193. CCOMMON_OPT += -n64
  1194. endif
  1195. ifeq ($(CORE), LOONGSON3R3)
  1196. CCOMMON_OPT += -loongson3 -static
  1197. endif
  1198. ifeq ($(CORE), LOONGSON3R4)
  1199. CCOMMON_OPT += -loongson3 -static
  1200. endif
  1201. else
  1202. ifndef BINARY64
  1203. CCOMMON_OPT += -m32
  1204. else
  1205. CCOMMON_OPT += -m64
  1206. endif
  1207. endif
  1208. endif
  1209. ifeq ($(C_COMPILER), SUN)
  1210. CCOMMON_OPT += -w
  1211. ifeq ($(ARCH), x86)
  1212. CCOMMON_OPT += -m32
  1213. else
  1214. ifdef BINARY64
  1215. CCOMMON_OPT += -m64
  1216. else
  1217. CCOMMON_OPT += -m32
  1218. endif
  1219. endif
  1220. endif
  1221. ifeq ($(F_COMPILER), SUN)
  1222. CCOMMON_OPT += -DF_INTERFACE_SUN
  1223. FCOMMON_OPT += -ftrap=%none -xrecursive
  1224. ifeq ($(ARCH), x86)
  1225. FCOMMON_OPT += -m32
  1226. else
  1227. ifdef BINARY64
  1228. FCOMMON_OPT += -m64
  1229. else
  1230. FCOMMON_OPT += -m32
  1231. endif
  1232. endif
  1233. ifeq ($(USE_OPENMP), 1)
  1234. FCOMMON_OPT += -xopenmp=parallel
  1235. endif
  1236. endif
  1237. ifeq ($(F_COMPILER), COMPAQ)
  1238. CCOMMON_OPT += -DF_INTERFACE_COMPAQ
  1239. ifeq ($(USE_OPENMP), 1)
  1240. FCOMMON_OPT += -openmp
  1241. endif
  1242. endif
  1243. ifeq ($(F_COMPILER), CRAY)
  1244. CCOMMON_OPT += -DF_INTERFACE_INTEL
  1245. FCOMMON_OPT += -hnopattern
  1246. ifdef INTERFACE64
  1247. ifneq ($(INTERFACE64), 0)
  1248. FCOMMON_OPT += -s integer64
  1249. endif
  1250. endif
  1251. ifneq ($(USE_OPENMP), 1)
  1252. FCOMMON_OPT += -O noomp
  1253. endif
  1254. endif
  1255. ifdef BINARY64
  1256. ifdef INTERFACE64
  1257. ifneq ($(INTERFACE64), 0)
  1258. CCOMMON_OPT +=
  1259. #-DUSE64BITINT
  1260. endif
  1261. endif
  1262. endif
  1263. ifeq ($(NEED_PIC), 1)
  1264. ifeq ($(C_COMPILER), IBM)
  1265. CCOMMON_OPT += -qpic=large
  1266. else
  1267. CCOMMON_OPT += -fPIC
  1268. endif
  1269. ifeq ($(F_COMPILER), SUN)
  1270. FCOMMON_OPT += -pic
  1271. else ifeq ($(F_COMPILER), NAG)
  1272. FCOMMON_OPT += -PIC
  1273. else ifeq ($(F_COMPILER), IBM)
  1274. FCOMMON_OPT += -qpic=large
  1275. else
  1276. FCOMMON_OPT += -fPIC
  1277. endif
  1278. endif
  1279. ifeq ($(DYNAMIC_ARCH), 1)
  1280. CCOMMON_OPT += -DDYNAMIC_ARCH
  1281. endif
  1282. ifeq ($(DYNAMIC_OLDER), 1)
  1283. CCOMMON_OPT += -DDYNAMIC_OLDER
  1284. endif
  1285. ifeq ($(C_LAPACK), 1)
  1286. CCOMMON_OPT += -DC_LAPACK
  1287. endif
  1288. ifeq ($(NO_LAPACK), 1)
  1289. CCOMMON_OPT += -DNO_LAPACK
  1290. #Disable LAPACK C interface
  1291. NO_LAPACKE = 1
  1292. endif
  1293. ifeq ($(NO_LAPACKE), 1)
  1294. CCOMMON_OPT += -DNO_LAPACKE
  1295. endif
  1296. ifeq ($(NO_AVX), 1)
  1297. CCOMMON_OPT += -DNO_AVX
  1298. endif
  1299. ifeq ($(ARCH), x86)
  1300. CCOMMON_OPT += -DNO_AVX
  1301. endif
  1302. ifeq ($(NO_AVX2), 1)
  1303. CCOMMON_OPT += -DNO_AVX2
  1304. endif
  1305. ifeq ($(NO_AVX512), 1)
  1306. CCOMMON_OPT += -DNO_AVX512
  1307. endif
  1308. ifeq ($(NO_SVE), 1)
  1309. CCOMMON_OPT += -DNO_SVE
  1310. endif
  1311. ifeq ($(NO_SME), 1)
  1312. CCOMMON_OPT += -DNO_SME
  1313. endif
  1314. ifdef SMP
  1315. CCOMMON_OPT += -DSMP_SERVER
  1316. ifeq ($(ARCH), mips64)
  1317. USE_SIMPLE_THREADED_LEVEL3 = 1
  1318. endif
  1319. ifeq ($(USE_OPENMP), 1)
  1320. # USE_SIMPLE_THREADED_LEVEL3 = 1
  1321. # NO_AFFINITY = 1
  1322. CCOMMON_OPT += -DUSE_OPENMP
  1323. endif
  1324. ifeq ($(BIGNUMA), 1)
  1325. CCOMMON_OPT += -DBIGNUMA
  1326. endif
  1327. endif
  1328. ifeq ($(NO_WARMUP), 1)
  1329. CCOMMON_OPT += -DNO_WARMUP
  1330. endif
  1331. ifeq ($(CONSISTENT_FPCSR), 1)
  1332. CCOMMON_OPT += -DCONSISTENT_FPCSR
  1333. endif
  1334. # Only for development
  1335. # CCOMMON_OPT += -DPARAMTEST
  1336. # CCOMMON_OPT += -DPREFETCHTEST
  1337. # CCOMMON_OPT += -DNO_SWITCHING
  1338. # USE_PAPI = 1
  1339. ifdef USE_PAPI
  1340. CCOMMON_OPT += -DUSE_PAPI
  1341. EXTRALIB += -lpapi -lperfctr
  1342. endif
  1343. ifdef BUFFERSIZE
  1344. CCOMMON_OPT += -DBUFFERSIZE=$(BUFFERSIZE)
  1345. endif
  1346. ifdef DYNAMIC_THREADS
  1347. CCOMMON_OPT += -DDYNAMIC_THREADS
  1348. endif
  1349. CCOMMON_OPT += -DMAX_CPU_NUMBER=$(NUM_THREADS)
  1350. CCOMMON_OPT += -DMAX_PARALLEL_NUMBER=$(NUM_PARALLEL)
  1351. ifdef USE_SIMPLE_THREADED_LEVEL3
  1352. CCOMMON_OPT += -DUSE_SIMPLE_THREADED_LEVEL3
  1353. endif
  1354. ifeq ($(USE_TLS), 1)
  1355. CCOMMON_OPT += -DUSE_TLS
  1356. endif
  1357. ifeq ($(BUILD_BFLOAT16), 1)
  1358. CCOMMON_OPT += -DBUILD_BFLOAT16
  1359. endif
  1360. ifeq ($(BUILD_SINGLE), 1)
  1361. CCOMMON_OPT += -DBUILD_SINGLE=1
  1362. endif
  1363. ifeq ($(BUILD_DOUBLE), 1)
  1364. CCOMMON_OPT += -DBUILD_DOUBLE=1
  1365. endif
  1366. ifeq ($(BUILD_COMPLEX), 1)
  1367. CCOMMON_OPT += -DBUILD_COMPLEX=1
  1368. endif
  1369. ifeq ($(BUILD_COMPLEX16), 1)
  1370. CCOMMON_OPT += -DBUILD_COMPLEX16=1
  1371. endif
  1372. CCOMMON_OPT += -DVERSION=\"$(VERSION)\"
  1373. ifndef SYMBOLPREFIX
  1374. SYMBOLPREFIX =
  1375. endif
  1376. ifndef SYMBOLSUFFIX
  1377. SYMBOLSUFFIX =
  1378. endif
  1379. ifndef LIBSONAMEBASE
  1380. LIBSONAMEBASE = openblas
  1381. endif
  1382. ifndef LIBNAMEPREFIX
  1383. LIBNAMEPREFIX =
  1384. endif
  1385. SYMPREFIX=$(SYMBOLPREFIX)
  1386. ifeq ($(SYMBOLPREFIX),$(LIBNAMEPREFIX))
  1387. SYMPREFIX=
  1388. endif
  1389. SYMSUFFIX=$(SYMBOLSUFFIX)
  1390. ifeq ($(SYMBOLSUFFIX),$(LIBNAMESUFFIX))
  1391. SYMSUFFIX=
  1392. endif
  1393. ifndef LIBNAMESUFFIX
  1394. LIBNAMEBASE = $(SYMPREFIX)$(LIBSONAMEBASE)$(SYMSUFFIX)
  1395. else
  1396. LIBNAMEBASE = $(SYMPREFIX)$(LIBSONAMEBASE)$(SYMSUFFIX)$(LIBNAMESUFFIX)
  1397. endif
  1398. ifeq ($(OSNAME), CYGWIN_NT)
  1399. LIBPREFIX = cyg$(LIBNAMEPREFIX)$(LIBNAMEBASE)
  1400. else
  1401. LIBPREFIX = lib$(LIBNAMEPREFIX)$(LIBNAMEBASE)
  1402. endif
  1403. KERNELDIR = $(TOPDIR)/kernel/$(ARCH)
  1404. include $(TOPDIR)/Makefile.$(ARCH)
  1405. ifneq ($(C_COMPILER), PGI)
  1406. ifneq ($(C_COMPILER), SUN)
  1407. CCOMMON_OPT += -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME
  1408. endif
  1409. endif
  1410. CCOMMON_OPT += -DASMNAME=$(FU)$(*F) -DASMFNAME=$(FU)$(*F)$(BU) -DNAME=$(*F)$(BU) -DCNAME=$(*F) -DCHAR_NAME=\"$(*F)$(BU)\" -DCHAR_CNAME=\"$(*F)\"
  1411. ifeq ($(CORE), PPC440)
  1412. CCOMMON_OPT += -DALLOC_QALLOC
  1413. endif
  1414. ifeq ($(CORE), PPC440FP2)
  1415. STATIC_ALLOCATION = 1
  1416. endif
  1417. ifneq ($(OSNAME), Linux)
  1418. NO_AFFINITY = 1
  1419. endif
  1420. ifneq ($(ARCH), x86_64)
  1421. ifneq ($(ARCH), x86)
  1422. NO_AFFINITY = 1
  1423. endif
  1424. endif
  1425. ifeq ($(ARCH), POWER)
  1426. ifeq ($(DEBUG), 1)
  1427. CCOMMON_OPT := $(filter-out -O%, $(CCOMMON_OPT)) -O0
  1428. FCOMMON_OPT := $(filter-out -O%, $(FCOMMON_OPT)) -O0
  1429. endif
  1430. endif
  1431. ifdef NO_AFFINITY
  1432. ifeq ($(NO_AFFINITY), 0)
  1433. override undefine NO_AFFINITY
  1434. else
  1435. CCOMMON_OPT += -DNO_AFFINITY
  1436. endif
  1437. endif
  1438. ifdef FUNCTION_PROFILE
  1439. CCOMMON_OPT += -DFUNCTION_PROFILE
  1440. endif
  1441. ifdef SHMEM_ALLOCATION
  1442. ifneq ($(SHMEM_ALLOCATION), 0)
  1443. CCOMMON_OPT += -DALLOC_SHM
  1444. endif
  1445. endif
  1446. ifdef HUGETLB_ALLOCATION
  1447. ifneq ($(HUGETLB_ALLOCATION), 0)
  1448. CCOMMON_OPT += -DALLOC_HUGETLB
  1449. endif
  1450. endif
  1451. ifdef HUGETLBFILE_ALLOCATION
  1452. ifneq ($(HUGETLBFILE_ALLOCATION), 0)
  1453. CCOMMON_OPT += -DALLOC_HUGETLBFILE -DHUGETLB_FILE_NAME=$(HUGETLBFILE_ALLOCATION)
  1454. endif
  1455. endif
  1456. ifdef STATIC_ALLOCATION
  1457. CCOMMON_OPT += -DALLOC_STATIC
  1458. endif
  1459. ifdef DEVICEDRIVER_ALLOCATION
  1460. CCOMMON_OPT += -DALLOC_DEVICEDRIVER -DDEVICEDRIVER_NAME=\"/dev/mapper\"
  1461. endif
  1462. ifdef MIXED_MEMORY_ALLOCATION
  1463. CCOMMON_OPT += -DMIXED_MEMORY_ALLOCATION
  1464. endif
  1465. ifeq ($(OSNAME), SunOS)
  1466. TAR = gtar
  1467. PATCH = gpatch
  1468. GREP = ggrep
  1469. AWK = nawk
  1470. else
  1471. TAR = tar
  1472. PATCH = patch
  1473. GREP = grep
  1474. AWK = awk
  1475. endif
  1476. ifndef MD5SUM
  1477. MD5SUM = md5sum
  1478. endif
  1479. REVISION = -r$(VERSION)
  1480. MAJOR_VERSION = $(word 1,$(subst ., ,$(VERSION)))
  1481. ifeq ($(DEBUG), 1)
  1482. COMMON_OPT += -g
  1483. endif
  1484. ifeq ($(DEBUG), 1)
  1485. FCOMMON_OPT += -g
  1486. endif
  1487. ifndef COMMON_OPT
  1488. COMMON_OPT = -O2
  1489. endif
  1490. ifndef FCOMMON_OPT
  1491. FCOMMON_OPT = -O2 -frecursive
  1492. endif
  1493. override CFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
  1494. override PFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
  1495. override FFLAGS += $(COMMON_OPT) $(FCOMMON_OPT)
  1496. override FPFLAGS += $(FCOMMON_OPT) $(COMMON_PROF)
  1497. #MAKEOVERRIDES =
  1498. ifeq ($(NEED_PIC), 1)
  1499. ifeq (,$(findstring PIC,$(FFLAGS)))
  1500. ifneq ($(F_COMPILER),IBM)
  1501. override FFLAGS += -fPIC
  1502. endif
  1503. endif
  1504. endif
  1505. #For LAPACK Fortran codes.
  1506. #Disable -fopenmp for LAPACK Fortran codes on Windows.
  1507. ifdef OS_WINDOWS
  1508. LAPACK_FFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FFLAGS))
  1509. LAPACK_FPFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FPFLAGS))
  1510. else
  1511. LAPACK_FFLAGS := $(FFLAGS)
  1512. LAPACK_FPFLAGS := $(FPFLAGS)
  1513. endif
  1514. ifeq ($(F_COMPILER),NAG)
  1515. LAPACK_FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1516. override FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1517. endif
  1518. ifeq ($(F_COMPILER),CRAY)
  1519. LAPACK_FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1520. override FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1521. endif
  1522. ifeq ($(F_COMPILER),FLANGNEW)
  1523. LAPACK_FFLAGS := $(filter-out -m32 -m64 -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 -mtune=% -mabi=% ,$(FFLAGS))
  1524. override FFLAGS := $(filter-out -m32 -m64 -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 -mtune=% -mabi=% ,$(FFLAGS))
  1525. endif
  1526. LAPACK_CFLAGS = $(CFLAGS)
  1527. LAPACK_CFLAGS += -DHAVE_LAPACK_CONFIG_H
  1528. ifdef INTERFACE64
  1529. ifneq ($(INTERFACE64), 0)
  1530. LAPACK_CFLAGS += -DLAPACK_ILP64
  1531. endif
  1532. endif
  1533. ifdef OS_WINDOWS
  1534. LAPACK_CFLAGS += -DOPENBLAS_OS_WINDOWS
  1535. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1536. endif
  1537. ifeq ($(C_COMPILER), LSB)
  1538. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1539. endif
  1540. ifndef SUFFIX
  1541. SUFFIX = o
  1542. endif
  1543. ifndef PSUFFIX
  1544. PSUFFIX = po
  1545. endif
  1546. ifndef LIBSUFFIX
  1547. LIBSUFFIX = a
  1548. endif
  1549. ifneq ($(DYNAMIC_ARCH), 1)
  1550. ifndef SMP
  1551. LIBNAME = $(LIBPREFIX)_$(LIBCORE)$(REVISION).$(LIBSUFFIX)
  1552. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)$(REVISION)_p.$(LIBSUFFIX)
  1553. else
  1554. LIBNAME = $(LIBPREFIX)_$(LIBCORE)p$(REVISION).$(LIBSUFFIX)
  1555. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)p$(REVISION)_p.$(LIBSUFFIX)
  1556. endif
  1557. else
  1558. ifndef SMP
  1559. LIBNAME = $(LIBPREFIX)$(REVISION).$(LIBSUFFIX)
  1560. LIBNAME_P = $(LIBPREFIX)$(REVISION)_p.$(LIBSUFFIX)
  1561. else
  1562. LIBNAME = $(LIBPREFIX)p$(REVISION).$(LIBSUFFIX)
  1563. LIBNAME_P = $(LIBPREFIX)p$(REVISION)_p.$(LIBSUFFIX)
  1564. endif
  1565. endif
  1566. ifeq ($(FIXED_LIBNAME),1)
  1567. LIBNAME = lib$(LIBNAMEPREFIX)$(LIBSONAMEBASE)$(LIBNAMESUFFIX).$(LIBSUFFIX)
  1568. LIBNAME_P = lib$(LIBNAMEPREFIX)$(LISOBNAMEBASE)$(LIBNAMESUFFIX)_p.$(LIBSUFFIX)
  1569. endif
  1570. LIBDLLNAME = $(LIBPREFIX).dll
  1571. IMPLIBNAME = lib$(LIBNAMEBASE).dll.a
  1572. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.so)
  1573. LIBDYNNAME = $(LIBNAME:.$(LIBSUFFIX)=.dylib)
  1574. LIBDEFNAME = $(LIBNAME:.$(LIBSUFFIX)=.def)
  1575. LIBEXPNAME = $(LIBNAME:.$(LIBSUFFIX)=.exp)
  1576. LIBZIPNAME = $(LIBNAME:.$(LIBSUFFIX)=.zip)
  1577. LIBS = $(TOPDIR)/$(LIBNAME)
  1578. LIBS_P = $(TOPDIR)/$(LIBNAME_P)
  1579. LIB_COMPONENTS = BLAS
  1580. ifneq ($(NO_CBLAS), 1)
  1581. LIB_COMPONENTS += CBLAS
  1582. endif
  1583. ifneq ($(NO_LAPACK), 1)
  1584. LIB_COMPONENTS += LAPACK
  1585. ifneq ($(NO_LAPACKE), 1)
  1586. LIB_COMPONENTS += LAPACKE
  1587. endif
  1588. ifeq ($(BUILD_RELAPACK), 1)
  1589. LIB_COMPONENTS += ReLAPACK
  1590. endif
  1591. endif
  1592. ifeq ($(ONLY_CBLAS), 1)
  1593. LIB_COMPONENTS = CBLAS
  1594. endif
  1595. export OSNAME
  1596. export ARCH
  1597. export CORE
  1598. export LIBCORE
  1599. export __BYTE_ORDER__
  1600. export ELF_VERSION
  1601. export PGCPATH
  1602. export CONFIG
  1603. export CC
  1604. export FC
  1605. export BU
  1606. export FU
  1607. export NEED2UNDERSCORES
  1608. export USE_THREAD
  1609. export NUM_THREADS
  1610. export NUM_CORES
  1611. export SMP
  1612. export MAKEFILE_RULE
  1613. export NEED_PIC
  1614. export BINARY
  1615. export BINARY32
  1616. export BINARY64
  1617. export F_COMPILER
  1618. export C_COMPILER
  1619. export USE_OPENMP
  1620. export CROSS
  1621. export CROSS_SUFFIX
  1622. export NOFORTRAN
  1623. export C_LAPACK
  1624. export NO_FBLAS
  1625. export EXTRALIB
  1626. export CEXTRALIB
  1627. export FEXTRALIB
  1628. export HAVE_SSE
  1629. export HAVE_SSE2
  1630. export HAVE_SSE3
  1631. export HAVE_SSSE3
  1632. export HAVE_SSE4_1
  1633. export HAVE_SSE4_2
  1634. export HAVE_SSE4A
  1635. export HAVE_SSE5
  1636. export HAVE_AVX
  1637. export HAVE_AVX2
  1638. export HAVE_FMA3
  1639. export HAVE_VFP
  1640. export HAVE_VFPV3
  1641. export HAVE_VFPV4
  1642. export HAVE_NEON
  1643. ifndef NO_MSA
  1644. export HAVE_MSA
  1645. export MSA_FLAGS
  1646. endif
  1647. export KERNELDIR
  1648. export FUNCTION_PROFILE
  1649. export TARGET_CORE
  1650. export NO_AVX512
  1651. export NO_AVX2
  1652. export BUILD_BFLOAT16
  1653. export NO_LSX
  1654. export NO_LASX
  1655. export SBGEMM_UNROLL_M
  1656. export SBGEMM_UNROLL_N
  1657. export SGEMM_UNROLL_M
  1658. export SGEMM_UNROLL_N
  1659. export DGEMM_UNROLL_M
  1660. export DGEMM_UNROLL_N
  1661. export QGEMM_UNROLL_M
  1662. export QGEMM_UNROLL_N
  1663. export CGEMM_UNROLL_M
  1664. export CGEMM_UNROLL_N
  1665. export ZGEMM_UNROLL_M
  1666. export ZGEMM_UNROLL_N
  1667. export XGEMM_UNROLL_M
  1668. export XGEMM_UNROLL_N
  1669. export CGEMM3M_UNROLL_M
  1670. export CGEMM3M_UNROLL_N
  1671. export ZGEMM3M_UNROLL_M
  1672. export ZGEMM3M_UNROLL_N
  1673. export XGEMM3M_UNROLL_M
  1674. export XGEMM3M_UNROLL_N
  1675. ifdef USE_CUDA
  1676. export CUDADIR
  1677. export CUCC
  1678. export CUFLAGS
  1679. export CULIB
  1680. endif
  1681. .SUFFIXES: .$(PSUFFIX) .$(SUFFIX) .f
  1682. .f.$(SUFFIX):
  1683. $(FC) $(FFLAGS) -c $< -o $(@F)
  1684. .f.$(PSUFFIX):
  1685. $(FC) $(FPFLAGS) -pg -c $< -o $(@F)
  1686. ifdef BINARY64
  1687. PATHSCALEPATH = /opt/pathscale/lib/3.1
  1688. PGIPATH = /opt/pgi/linux86-64/7.1-5/lib
  1689. else
  1690. PATHSCALEPATH = /opt/pathscale/lib/3.1/32
  1691. PGIPATH = /opt/pgi/linux86/7.1-5/lib
  1692. endif
  1693. ACMLPATH = /opt/acml/4.3.0
  1694. ifneq ($(OSNAME), Darwin)
  1695. MKLPATH = /opt/intel/mkl/10.2.2.025/lib
  1696. else
  1697. MKLPATH = /Library/Frameworks/Intel_MKL.framework/Versions/10.0.1.014/lib
  1698. endif
  1699. ATLASPATH = /opt/atlas/3.9.17/opteron
  1700. FLAMEPATH = $(HOME)/flame/lib
  1701. ifneq ($(OSNAME), SunOS)
  1702. SUNPATH = /opt/sunstudio12.1
  1703. else
  1704. SUNPATH = /opt/SUNWspro
  1705. endif