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 42 kB

7 years ago
1 year ago
4 years ago
13 years ago
4 years ago
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942943944945946947948949950951952953954955956957958959960961962963964965966967968969970971972973974975976977978979980981982983984985986987988989990991992993994995996997998999100010011002100310041005100610071008100910101011101210131014101510161017101810191020102110221023102410251026102710281029103010311032103310341035103610371038103910401041104210431044104510461047104810491050105110521053105410551056105710581059106010611062106310641065106610671068106910701071107210731074107510761077107810791080108110821083108410851086108710881089109010911092109310941095109610971098109911001101110211031104110511061107110811091110111111121113111411151116111711181119112011211122112311241125112611271128112911301131113211331134113511361137113811391140114111421143114411451146114711481149115011511152115311541155115611571158115911601161116211631164116511661167116811691170117111721173117411751176117711781179118011811182118311841185118611871188118911901191119211931194119511961197119811991200120112021203120412051206120712081209121012111212121312141215121612171218121912201221122212231224122512261227122812291230123112321233123412351236123712381239124012411242124312441245124612471248124912501251125212531254125512561257125812591260126112621263126412651266126712681269127012711272127312741275127612771278127912801281128212831284128512861287128812891290129112921293129412951296129712981299130013011302130313041305130613071308130913101311131213131314131513161317131813191320132113221323132413251326132713281329133013311332133313341335133613371338133913401341134213431344134513461347134813491350135113521353135413551356135713581359136013611362136313641365136613671368136913701371137213731374137513761377137813791380138113821383138413851386138713881389139013911392139313941395139613971398139914001401140214031404140514061407140814091410141114121413141414151416141714181419142014211422142314241425142614271428142914301431143214331434143514361437143814391440144114421443144414451446144714481449145014511452145314541455145614571458145914601461146214631464146514661467146814691470147114721473147414751476147714781479148014811482148314841485148614871488148914901491149214931494149514961497149814991500150115021503150415051506150715081509151015111512151315141515151615171518151915201521152215231524152515261527152815291530153115321533153415351536153715381539154015411542154315441545154615471548154915501551155215531554155515561557155815591560156115621563156415651566156715681569157015711572157315741575157615771578157915801581158215831584158515861587158815891590159115921593159415951596159715981599160016011602160316041605160616071608160916101611161216131614161516161617161816191620162116221623162416251626162716281629163016311632163316341635163616371638163916401641164216431644164516461647164816491650165116521653165416551656165716581659166016611662166316641665166616671668166916701671167216731674167516761677167816791680168116821683168416851686168716881689169016911692169316941695169616971698169917001701170217031704170517061707170817091710171117121713171417151716171717181719172017211722172317241725172617271728172917301731173217331734173517361737173817391740174117421743174417451746174717481749175017511752175317541755175617571758175917601761176217631764176517661767176817691770177117721773177417751776177717781779178017811782178317841785178617871788178917901791179217931794179517961797179817991800180118021803180418051806180718081809181018111812181318141815181618171818181918201821182218231824182518261827182818291830183118321833183418351836183718381839184018411842184318441845184618471848184918501851185218531854185518561857185818591860186118621863186418651866186718681869187018711872187318741875187618771878187918801881188218831884188518861887188818891890189118921893189418951896189718981899190019011902190319041905190619071908190919101911191219131914191519161917191819191920192119221923192419251926192719281929193019311932193319341935193619371938193919401941194219431944194519461947194819491950195119521953195419551956195719581959196019611962196319641965196619671968196919701971197219731974
  1. #
  2. # Include user definition
  3. #
  4. # TO suppress recursive includes
  5. INCLUDED = 1
  6. ifndef TOPDIR
  7. TOPDIR = .
  8. endif
  9. ifndef RELAPACK_REPLACE
  10. RELAPACK_REPLACE=0
  11. endif
  12. # we need to use the host system's architecture for getarch compile options even especially when cross-compiling
  13. HOSTARCH := $(shell uname -m)
  14. ifeq ($(HOSTARCH), amd64)
  15. HOSTARCH=x86_64
  16. endif
  17. # Catch conflicting usage of ARCH in some BSD environments
  18. ifeq ($(ARCH), amd64)
  19. override ARCH=x86_64
  20. else ifeq ($(ARCH), powerpc64)
  21. override ARCH=power
  22. else ifeq ($(ARCH), powerpc64le)
  23. override ARCH=power
  24. else ifeq ($(ARCH), powerpc)
  25. override ARCH=power
  26. else ifeq ($(ARCH), i386)
  27. override ARCH=x86
  28. else ifeq ($(ARCH), armv6)
  29. override ARCH=arm
  30. else ifeq ($(ARCH), armv7)
  31. override ARCH=arm
  32. else ifeq ($(ARCH), aarch64)
  33. override ARCH=arm64
  34. else ifeq ($(ARCH), mipsel)
  35. override ARCH=mips
  36. else ifeq ($(ARCH), mips64el)
  37. override ARCH=mips64
  38. else ifeq ($(ARCH), zarch)
  39. override ARCH=zarch
  40. endif
  41. NETLIB_LAPACK_DIR = $(TOPDIR)/lapack-netlib
  42. # Default C compiler
  43. # - Only set if not specified on the command line or inherited from the environment.
  44. # - CC is an implicit variable so neither '?=' or 'ifndef' can be used.
  45. # http://stackoverflow.com/questions/4029274/mingw-and-make-variables
  46. # - Default value is 'cc' which is not always a valid command (e.g. MinGW).
  47. ifeq ($(origin CC),default)
  48. # Check if $(CC) refers to a valid command and set the value to gcc if not
  49. ifneq ($(findstring cmd.exe,$(SHELL)),)
  50. ifeq ($(shell where $(CC) 2>NUL),)
  51. CC = gcc
  52. endif
  53. else # POSIX-ish
  54. ifeq ($(shell command -v $(CC) 2>/dev/null),)
  55. ifeq ($(shell uname -s),Darwin)
  56. CC = clang
  57. # EXTRALIB += -Wl,-no_compact_unwind
  58. else
  59. CC = gcc
  60. endif # Darwin
  61. endif # CC exists
  62. endif # Shell is sane
  63. endif # CC is set to default
  64. # Default Fortran compiler (FC) is selected by f_check.
  65. ifndef MAKEFILE_RULE
  66. include $(TOPDIR)/Makefile.rule
  67. else
  68. include $(TOPDIR)/$(MAKEFILE_RULE)
  69. endif
  70. #
  71. # Beginning of system configuration
  72. #
  73. ifneq ($(BUILD_SINGLE),1)
  74. ifneq ($(BUILD_DOUBLE),1)
  75. ifneq ($(BUILD_COMPLEX),1)
  76. ifneq ($(BUILD_COMPLEX16),1)
  77. override BUILD_SINGLE=1
  78. override BUILD_DOUBLE=1
  79. override BUILD_COMPLEX=1
  80. override BUILD_COMPLEX16=1
  81. endif
  82. endif
  83. endif
  84. endif
  85. ifndef HOSTCC
  86. HOSTCC = $(CC)
  87. endif
  88. ifdef TARGET
  89. GETARCH_FLAGS := -DFORCE_$(TARGET)
  90. GETARCH_FLAGS += -DUSER_TARGET
  91. ifeq ($(TARGET), GENERIC)
  92. ifeq ($(DYNAMIC_ARCH), 1)
  93. override NO_EXPRECISION=1
  94. export NO_EXPRECISION
  95. endif
  96. endif
  97. endif
  98. # Force fallbacks for 32bit
  99. ifeq ($(BINARY), 32)
  100. ifeq ($(TARGET), HASWELL)
  101. GETARCH_FLAGS := -DFORCE_NEHALEM
  102. endif
  103. ifeq ($(TARGET), SKYLAKEX)
  104. GETARCH_FLAGS := -DFORCE_NEHALEM
  105. endif
  106. ifeq ($(TARGET), COOPERLAKE)
  107. GETARCH_FLAGS := -DFORCE_NEHALEM
  108. endif
  109. ifeq ($(TARGET), SAPPHIRERAPIDS)
  110. GETARCH_FLAGS := -DFORCE_NEHALEM
  111. endif
  112. ifeq ($(TARGET), SANDYBRIDGE)
  113. GETARCH_FLAGS := -DFORCE_NEHALEM
  114. endif
  115. ifeq ($(TARGET), BULLDOZER)
  116. GETARCH_FLAGS := -DFORCE_BARCELONA
  117. endif
  118. ifeq ($(TARGET), PILEDRIVER)
  119. GETARCH_FLAGS := -DFORCE_BARCELONA
  120. endif
  121. ifeq ($(TARGET), STEAMROLLER)
  122. GETARCH_FLAGS := -DFORCE_BARCELONA
  123. endif
  124. ifeq ($(TARGET), EXCAVATOR)
  125. GETARCH_FLAGS := -DFORCE_BARCELONA
  126. endif
  127. ifeq ($(TARGET), ZEN)
  128. GETARCH_FLAGS := -DFORCE_BARCELONA
  129. endif
  130. ifeq ($(TARGET), ARMV8)
  131. GETARCH_FLAGS := -DFORCE_ARMV7
  132. endif
  133. ifeq ($(TARGET), POWER8)
  134. GETARCH_FLAGS := -DFORCE_POWER6
  135. endif
  136. ifeq ($(TARGET), POWER9)
  137. GETARCH_FLAGS := -DFORCE_POWER6
  138. endif
  139. ifeq ($(TARGET), POWER10)
  140. GETARCH_FLAGS := -DFORCE_POWER6
  141. endif
  142. endif
  143. #TARGET_CORE will override TARGET which is used in DYNAMIC_ARCH=1.
  144. #
  145. ifdef TARGET_CORE
  146. GETARCH_FLAGS := -DFORCE_$(TARGET_CORE)
  147. endif
  148. # Force fallbacks for 32bit
  149. ifeq ($(BINARY), 32)
  150. ifeq ($(TARGET_CORE), HASWELL)
  151. GETARCH_FLAGS := -DFORCE_NEHALEM
  152. endif
  153. ifeq ($(TARGET_CORE), SKYLAKEX)
  154. GETARCH_FLAGS := -DFORCE_NEHALEM
  155. endif
  156. ifeq ($(TARGET_CORE), COOPERLAKE)
  157. GETARCH_FLAGS := -DFORCE_NEHALEM
  158. endif
  159. ifeq ($(TARGET_CORE), SAPPHIRERAPIDS)
  160. GETARCH_FLAGS := -DFORCE_NEHALEM
  161. endif
  162. ifeq ($(TARGET_CORE), SANDYBRIDGE)
  163. GETARCH_FLAGS := -DFORCE_NEHALEM
  164. endif
  165. ifeq ($(TARGET_CORE), BULLDOZER)
  166. GETARCH_FLAGS := -DFORCE_BARCELONA
  167. endif
  168. ifeq ($(TARGET_CORE), PILEDRIVER)
  169. GETARCH_FLAGS := -DFORCE_BARCELONA
  170. endif
  171. ifeq ($(TARGET_CORE), STEAMROLLER)
  172. GETARCH_FLAGS := -DFORCE_BARCELONA
  173. endif
  174. ifeq ($(TARGET_CORE), EXCAVATOR)
  175. GETARCH_FLAGS := -DFORCE_BARCELONA
  176. endif
  177. ifeq ($(TARGET_CORE), ZEN)
  178. GETARCH_FLAGS := -DFORCE_BARCELONA
  179. endif
  180. endif
  181. # On x86_64 build getarch with march=native unless the compiler is PGI. This is required to detect AVX512 support in getarch.
  182. ifeq ($(HOSTARCH), x86_64)
  183. ifeq ($(findstring pgcc,$(HOSTCC))$(findstring nvc,$(HOSTCC)),)
  184. GETARCH_FLAGS += -march=native
  185. endif
  186. endif
  187. ifdef INTERFACE64
  188. ifneq ($(INTERFACE64), 0)
  189. GETARCH_FLAGS += -DUSE64BITINT
  190. endif
  191. endif
  192. ifndef GEMM_MULTITHREAD_THRESHOLD
  193. GEMM_MULTITHREAD_THRESHOLD=4
  194. endif
  195. GETARCH_FLAGS += -DGEMM_MULTITHREAD_THRESHOLD=$(GEMM_MULTITHREAD_THRESHOLD)
  196. ifeq ($(NO_AVX), 1)
  197. GETARCH_FLAGS += -DNO_AVX
  198. endif
  199. ifeq ($(BINARY), 32)
  200. GETARCH_FLAGS += -DNO_AVX -DNO_AVX2 -DNO_AVX512
  201. NO_AVX512 = 1
  202. endif
  203. ifeq ($(NO_AVX2), 1)
  204. GETARCH_FLAGS += -DNO_AVX2
  205. endif
  206. ifeq ($(NO_AVX512), 1)
  207. GETARCH_FLAGS += -DNO_AVX512
  208. endif
  209. ifeq ($(DEBUG), 1)
  210. GETARCH_FLAGS += -g
  211. endif
  212. ifeq ($(QUIET_MAKE), 1)
  213. MAKE += -s
  214. endif
  215. ifndef NO_PARALLEL_MAKE
  216. NO_PARALLEL_MAKE=0
  217. endif
  218. GETARCH_FLAGS += -DNO_PARALLEL_MAKE=$(NO_PARALLEL_MAKE)
  219. ifdef MAKE_NB_JOBS
  220. GETARCH_FLAGS += -DMAKE_NB_JOBS=$(MAKE_NB_JOBS)
  221. endif
  222. ifeq ($(HOSTCC), loongcc)
  223. GETARCH_FLAGS += -static
  224. endif
  225. #if don't use Fortran, it will only compile CBLAS.
  226. ifeq ($(ONLY_CBLAS), 1)
  227. NO_LAPACK = 1
  228. else
  229. ONLY_CBLAS = 0
  230. endif
  231. #For small matrix optimization
  232. ifeq ($(ARCH), x86_64)
  233. SMALL_MATRIX_OPT = 1
  234. else ifeq ($(ARCH), power)
  235. SMALL_MATRIX_OPT = 1
  236. BUILD_BFLOAT16 = 1
  237. else ifeq ($(ARCH), arm64)
  238. SMALL_MATRIX_OPT = 1
  239. BUILD_BFLOAT16 = 1
  240. endif
  241. ifeq ($(ARCH), loongarch64)
  242. SMALL_MATRIX_OPT = 1
  243. endif
  244. ifeq ($(ARCH), arm64)
  245. GEMM_GEMV_FORWARD = 1
  246. SBGEMM_GEMV_FORWARD = 1
  247. BGEMM_GEMV_FORWARD = 1
  248. endif
  249. ifeq ($(ARCH), riscv)
  250. GEMM_GEMV_FORWARD = 1
  251. endif
  252. ifeq ($(ARCH), power)
  253. GEMM_GEMV_FORWARD = 1
  254. SBGEMM_GEMV_FORWARD = 1
  255. endif
  256. ifeq ($(SMALL_MATRIX_OPT), 1)
  257. CCOMMON_OPT += -DSMALL_MATRIX_OPT
  258. endif
  259. ifneq ($(ONLY_CBLAS), 1)
  260. ifeq ($(GEMM_GEMV_FORWARD), 1)
  261. CCOMMON_OPT += -DGEMM_GEMV_FORWARD
  262. endif
  263. ifeq ($(SBGEMM_GEMV_FORWARD), 1)
  264. CCOMMON_OPT += -DSBGEMM_GEMV_FORWARD
  265. endif
  266. ifeq ($(BGEMM_GEMV_FORWARD), 1)
  267. CCOMMON_OPT += -DBGEMM_GEMV_FORWARD
  268. endif
  269. endif
  270. # This operation is expensive, so execution should be once.
  271. ifndef GOTOBLAS_MAKEFILE
  272. export GOTOBLAS_MAKEFILE = 1
  273. # Generating Makefile.conf and config.h
  274. DUMMY := $(shell $(MAKE) -C $(TOPDIR) -f Makefile.prebuild CC="$(CC)" FC="$(FC)" HOSTCC="$(HOSTCC)" HOST_CFLAGS="$(GETARCH_FLAGS)" CFLAGS="$(CFLAGS)" BINARY=$(BINARY) USE_OPENMP=$(USE_OPENMP) DYNAMIC_ARCH=$(DYNAMIC_ARCH) TARGET_CORE=$(TARGET_CORE) ONLY_CBLAS=$(ONLY_CBLAS) TARGET=$(TARGET) all)
  275. endif
  276. ifndef TARGET_CORE
  277. -include $(TOPDIR)/Makefile.conf
  278. else
  279. HAVE_NEON=
  280. HAVE_VFP=
  281. HAVE_VFPV3=
  282. HAVE_VFPV4=
  283. HAVE_MMX=
  284. HAVE_SSE=
  285. HAVE_SSE2=
  286. HAVE_SSE3=
  287. HAVE_SSSE3=
  288. HAVE_SSE4_1=
  289. HAVE_SSE4_2=
  290. HAVE_SSE4A=
  291. HAVE_SSE5=
  292. HAVE_AVX=
  293. HAVE_AVX2=
  294. HAVE_FMA3=
  295. include $(TOPDIR)/Makefile_kernel.conf
  296. endif
  297. ifndef NUM_PARALLEL
  298. NUM_PARALLEL = 1
  299. endif
  300. ifndef NUM_THREADS
  301. NUM_THREADS = $(NUM_CORES)
  302. endif
  303. ifeq ($(NUM_THREADS), 1)
  304. override USE_THREAD = 0
  305. override USE_OPENMP = 0
  306. endif
  307. ifdef USE_THREAD
  308. ifeq ($(USE_THREAD), 0)
  309. SMP =
  310. else
  311. SMP = 1
  312. endif
  313. else
  314. ifeq ($(NUM_THREADS), 1)
  315. SMP =
  316. else
  317. SMP = 1
  318. endif
  319. endif
  320. ifeq ($(SMP), 1)
  321. USE_LOCKING =
  322. endif
  323. ifndef NEED_PIC
  324. NEED_PIC = 1
  325. endif
  326. ARFLAGS =
  327. CPP = $(COMPILER) -E
  328. AR ?= $(CROSS_SUFFIX)ar
  329. AS ?= $(CROSS_SUFFIX)as
  330. LD ?= $(CROSS_SUFFIX)ld
  331. RANLIB ?= $(CROSS_SUFFIX)ranlib
  332. NM = $(CROSS_SUFFIX)nm
  333. DLLWRAP = $(CROSS_SUFFIX)dllwrap
  334. OBJCOPY = $(CROSS_SUFFIX)objcopy
  335. OBJCONV = $(CROSS_SUFFIX)objconv
  336. # When fortran support was either not detected or actively deselected, only build BLAS.
  337. ifeq ($(NOFORTRAN), 1)
  338. C_LAPACK = 1
  339. override FEXTRALIB =
  340. endif
  341. ifeq ($(C_COMPILER), GCC)
  342. GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
  343. GCCVERSIONGT4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 4)
  344. GCCVERSIONGT5 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 5)
  345. GCCVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 7)
  346. GCCVERSIONGTEQ8 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 8)
  347. GCCVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  348. GCCVERSIONGTEQ10 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 10)
  349. GCCVERSIONGTEQ11 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 11)
  350. GCCVERSIONGTEQ12 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 12)
  351. GCCVERSIONGTEQ13 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 13)
  352. GCCVERSIONGTEQ14 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 14)
  353. # Note that the behavior of -dumpversion is compile-time-configurable for
  354. # gcc-7.x and newer. Use -dumpfullversion there
  355. ifeq ($(GCCVERSIONGTEQ7),1)
  356. GCCDUMPVERSION_PARAM := -dumpfullversion
  357. else
  358. GCCDUMPVERSION_PARAM := -dumpversion
  359. endif
  360. GCCMINORVERSIONGTEQ1 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 1)
  361. GCCMINORVERSIONGTEQ2 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 2)
  362. GCCMINORVERSIONGTEQ4 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 4)
  363. GCCMINORVERSIONGTEQ7 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 7)
  364. endif
  365. ifeq ($(C_COMPILER), CLANG)
  366. CLANGVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  367. CLANGVERSIONGTEQ12 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 12)
  368. endif
  369. #
  370. # OS dependent settings
  371. #
  372. ifeq ($(OSNAME), Darwin)
  373. ifndef MACOSX_DEPLOYMENT_TARGET
  374. ifeq ($(ARCH), arm64)
  375. export MACOSX_DEPLOYMENT_TARGET=11.0
  376. export NO_SVE = 1
  377. # export NO_SME = 1
  378. else
  379. export MACOSX_DEPLOYMENT_TARGET=10.8
  380. endif
  381. endif
  382. MD5SUM = md5 -r
  383. XCVER = $(shell pkgutil --pkg-info=com.apple.pkg.Xcode |awk '/version:/ {print $2}'|cut -d: -f2|cut -f1 -d.)
  384. ifeq (x$(XCVER)x,xx)
  385. XCVER = $(shell pkgutil --pkg-info=com.apple.pkg.CLTools_Executables |awk '/version:/ {print $2}'|cut -d: -f2|cut -f1 -d.)
  386. endif
  387. ifeq (x$(XCVER), x 15)
  388. CCOMMON_OPT += -Wl,-ld_classic
  389. FCOMMON_OPT += -Wl,-ld_classic
  390. endif
  391. ifeq (x$(XCVER), x 16)
  392. ifeq ($(F_COMPILER), GFORTRAN)
  393. override CEXTRALIB := $(filter-out(-lto_library, $(CEXTRALIB)))
  394. endif
  395. endif
  396. endif
  397. ifneq (,$(findstring $(OSNAME), FreeBSD OpenBSD DragonFly))
  398. MD5SUM = md5 -r
  399. endif
  400. ifeq ($(OSNAME), NetBSD)
  401. MD5SUM = md5 -n
  402. endif
  403. ifeq ($(OSNAME), Linux)
  404. EXTRALIB += -lm
  405. #NO_EXPRECISION = 1
  406. endif
  407. ifeq ($(OSNAME), Android)
  408. EXTRALIB += -lm
  409. endif
  410. ifeq ($(OSNAME), AIX)
  411. EXTRALIB += -lm
  412. endif
  413. ifeq ($(OSNAME), $(filter $(OSNAME),FreeBSD OpenBSD NetBSD DragonFly))
  414. ifeq ($(ARCH), $(filter $(ARCH),arm arm64))
  415. EXTRALIB += -lm
  416. endif
  417. endif
  418. ifeq ($(OSNAME), WINNT)
  419. NEED_PIC = 0
  420. NO_EXPRECISION = 1
  421. EXTRALIB += -defaultlib:advapi32
  422. SUFFIX = obj
  423. PSUFFIX = pobj
  424. LIBSUFFIX = a
  425. ifeq ($(C_COMPILER), CLANG)
  426. CCOMMON_OPT += -DMS_ABI
  427. endif
  428. #Version tests for supporting specific features (MS_ABI, POWER9 intrinsics)
  429. ifeq ($(GCCVERSIONGT4), 1)
  430. # GCC Major version > 4
  431. # It is compatible with MSVC ABI.
  432. CCOMMON_OPT += -DMS_ABI
  433. endif
  434. ifeq ($(GCCVERSIONGTEQ4), 1)
  435. ifeq ($(GCCMINORVERSIONGTEQ7), 1)
  436. # GCC Version >=4.7
  437. # It is compatible with MSVC ABI.
  438. CCOMMON_OPT += -DMS_ABI
  439. endif
  440. endif
  441. # Ensure the correct stack alignment on Win32
  442. # http://permalink.gmane.org/gmane.comp.lib.openblas.general/97
  443. ifeq ($(ARCH), x86)
  444. CCOMMON_OPT += -mincoming-stack-boundary=2
  445. FCOMMON_OPT += -mincoming-stack-boundary=2
  446. endif
  447. endif
  448. ifeq ($(OSNAME), Interix)
  449. NEED_PIC = 0
  450. NO_EXPRECISION = 1
  451. INTERIX_TOOL_DIR = /opt/gcc.3.3/i586-pc-interix3/bin
  452. endif
  453. ifeq ($(OSNAME), CYGWIN_NT)
  454. NEED_PIC = 0
  455. NO_EXPRECISION = 1
  456. OS_CYGWIN_NT = 1
  457. endif
  458. ifneq ($(OSNAME), WINNT)
  459. ifneq ($(OSNAME), CYGWIN_NT)
  460. ifneq ($(OSNAME), Interix)
  461. ifneq ($(OSNAME), Android)
  462. ifdef SMP
  463. EXTRALIB += -lpthread
  464. endif
  465. endif
  466. endif
  467. endif
  468. endif
  469. # ifeq logical or
  470. ifeq ($(OSNAME), $(filter $(OSNAME),WINNT CYGWIN_NT Interix))
  471. OS_WINDOWS=1
  472. endif
  473. ifdef QUAD_PRECISION
  474. CCOMMON_OPT += -DQUAD_PRECISION
  475. NO_EXPRECISION = 1
  476. endif
  477. ifneq ($(ARCH), x86)
  478. ifneq ($(ARCH), x86_64)
  479. NO_EXPRECISION = 1
  480. endif
  481. endif
  482. ifdef UTEST_CHECK
  483. CCOMMON_OPT += -DUTEST_CHECK
  484. SANITY_CHECK = 1
  485. endif
  486. ifdef SANITY_CHECK
  487. CCOMMON_OPT += -DSANITY_CHECK -DREFNAME=$(*F)f$(BU)
  488. endif
  489. MAX_STACK_ALLOC ?= 2048
  490. ifneq ($(MAX_STACK_ALLOC), 0)
  491. CCOMMON_OPT += -DMAX_STACK_ALLOC=$(MAX_STACK_ALLOC)
  492. endif
  493. ifdef USE_LOCKING
  494. ifneq ($(USE_LOCKING), 0)
  495. CCOMMON_OPT += -DUSE_LOCKING
  496. endif
  497. endif
  498. #
  499. # Architecture dependent settings
  500. #
  501. ifeq ($(ARCH), x86)
  502. ifndef BINARY
  503. NO_BINARY_MODE = 1
  504. endif
  505. ifeq ($(CORE), generic)
  506. #NO_EXPRECISION = 1
  507. endif
  508. ifndef NO_EXPRECISION
  509. ifeq ($(F_COMPILER), GFORTRAN)
  510. # ifeq logical or. GCC or LSB
  511. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  512. EXPRECISION = 1
  513. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  514. FCOMMON_OPT += -m128bit-long-double
  515. endif
  516. ifeq ($(C_COMPILER), CLANG)
  517. EXPRECISION = 1
  518. CCOMMON_OPT += -DEXPRECISION
  519. FCOMMON_OPT += -m128bit-long-double
  520. endif
  521. endif
  522. endif
  523. endif
  524. ifeq ($(ARCH), x86_64)
  525. ifeq ($(CORE), generic)
  526. #NO_EXPRECISION = 1
  527. endif
  528. ifndef NO_EXPRECISION
  529. ifeq ($(F_COMPILER), GFORTRAN)
  530. # ifeq logical or. GCC or LSB
  531. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  532. EXPRECISION = 1
  533. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  534. FCOMMON_OPT += -m128bit-long-double
  535. endif
  536. ifeq ($(C_COMPILER), CLANG)
  537. EXPRECISION = 1
  538. CCOMMON_OPT += -DEXPRECISION
  539. FCOMMON_OPT += -m128bit-long-double
  540. endif
  541. endif
  542. endif
  543. endif
  544. ifeq ($(C_COMPILER), INTEL)
  545. CCOMMON_OPT += -wd981
  546. endif
  547. ifeq ($(USE_OPENMP), 1)
  548. #check
  549. ifeq ($(USE_THREAD), 0)
  550. $(error OpenBLAS: Cannot set both USE_OPENMP=1 and USE_THREAD=0. The USE_THREAD=0 is only for building single thread version.)
  551. endif
  552. # ifeq logical or. GCC or LSB
  553. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  554. CCOMMON_OPT += -fopenmp
  555. endif
  556. ifeq ($(C_COMPILER), CLANG)
  557. CCOMMON_OPT += -fopenmp
  558. ifeq ($(F_COMPILER), GFORTRAN)
  559. FEXTRALIB := $(subst -lgomp,-lomp,$(FEXTRALIB))
  560. endif
  561. endif
  562. ifeq ($(C_COMPILER), INTEL)
  563. CCOMMON_OPT += -fopenmp
  564. endif
  565. ifeq ($(C_COMPILER), PGI)
  566. CCOMMON_OPT += -mp
  567. endif
  568. ifeq ($(C_COMPILER), OPEN64)
  569. CCOMMON_OPT += -mp
  570. CEXTRALIB += -lstdc++
  571. endif
  572. ifeq ($(C_COMPILER), PATHSCALE)
  573. CCOMMON_OPT += -mp
  574. endif
  575. endif
  576. ifeq ($(DYNAMIC_ARCH), 1)
  577. ifeq ($(ARCH), x86)
  578. DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
  579. CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
  580. endif
  581. ifeq ($(ARCH), x86_64)
  582. DYNAMIC_CORE = PRESCOTT CORE2
  583. ifeq ($(DYNAMIC_OLDER), 1)
  584. DYNAMIC_CORE += PENRYN DUNNINGTON
  585. endif
  586. DYNAMIC_CORE += NEHALEM
  587. ifeq ($(DYNAMIC_OLDER), 1)
  588. DYNAMIC_CORE += OPTERON OPTERON_SSE3
  589. endif
  590. DYNAMIC_CORE += BARCELONA
  591. ifeq ($(DYNAMIC_OLDER), 1)
  592. DYNAMIC_CORE += BOBCAT ATOM NANO
  593. endif
  594. ifneq ($(NO_AVX), 1)
  595. DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER STEAMROLLER EXCAVATOR
  596. endif
  597. ifneq ($(NO_AVX2), 1)
  598. DYNAMIC_CORE += HASWELL ZEN
  599. endif
  600. ifneq ($(NO_AVX512), 1)
  601. ifneq ($(NO_AVX2), 1)
  602. DYNAMIC_CORE += SKYLAKEX COOPERLAKE SAPPHIRERAPIDS
  603. endif
  604. endif
  605. endif
  606. ifdef DYNAMIC_LIST
  607. override DYNAMIC_CORE = PRESCOTT $(DYNAMIC_LIST)
  608. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_PRESCOTT
  609. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  610. CCOMMON_OPT += $(XCCOMMON_OPT)
  611. #CCOMMON_OPT += -DDYNAMIC_LIST='$(DYNAMIC_LIST)'
  612. endif
  613. ifeq ($(ARCH), arm64)
  614. DYNAMIC_CORE = ARMV8
  615. DYNAMIC_CORE += CORTEXA53
  616. DYNAMIC_CORE += CORTEXA57
  617. DYNAMIC_CORE += NEOVERSEN1
  618. ifneq ($(NO_SVE), 1)
  619. DYNAMIC_CORE += NEOVERSEV1
  620. DYNAMIC_CORE += NEOVERSEN2
  621. DYNAMIC_CORE += ARMV8SVE
  622. DYNAMIC_CORE += A64FX
  623. endif
  624. ifneq ($(NO_SME), 1)
  625. DYNAMIC_CORE += ARMV9SME
  626. DYNAMIC_CORE += VORTEXM4
  627. endif
  628. DYNAMIC_CORE += THUNDERX
  629. DYNAMIC_CORE += THUNDERX2T99
  630. DYNAMIC_CORE += TSV110
  631. DYNAMIC_CORE += EMAG8180
  632. DYNAMIC_CORE += THUNDERX3T110
  633. ifdef DYNAMIC_LIST
  634. override DYNAMIC_CORE = ARMV8 $(DYNAMIC_LIST)
  635. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_ARMV8
  636. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  637. endif
  638. endif
  639. ifeq ($(ARCH), mips64)
  640. DYNAMIC_CORE = LOONGSON3R3 LOONGSON3R4 MIPS64_GENERIC
  641. ifdef DYNAMIC_LIST
  642. override DYNAMIC_CORE = MIPS64_GENERIC $(DYNAMIC_LIST)
  643. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_MIPS64_GENERIC
  644. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  645. endif
  646. endif
  647. ifeq ($(ARCH), loongarch64)
  648. DYNAMIC_CORE = LA64_GENERIC LA264 LA464
  649. endif
  650. ifeq ($(ARCH), riscv64)
  651. DYNAMIC_CORE = RISCV64_GENERIC
  652. DYNAMIC_CORE += RISCV64_ZVL128B
  653. DYNAMIC_CORE += RISCV64_ZVL256B
  654. ifdef DYNAMIC_LIST
  655. override DYNAMIC_CORE = RISCV64_GENERIC $(DYNAMIC_LIST)
  656. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_RISCV64_GENERIC
  657. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  658. endif
  659. endif
  660. ifeq ($(ARCH), zarch)
  661. DYNAMIC_CORE = ZARCH_GENERIC
  662. # if the compiler accepts -march=arch11 or -march=z13 and can compile a file
  663. # with z13-specific inline assembly, then we can include support for Z13.
  664. # note: -march=z13 is equivalent to -march=arch11 yet some compiler releases
  665. # only support one or the other.
  666. # note: LLVM version 6.x supported -march=z13 yet could not handle vector
  667. # registers in inline assembly, so the check for supporting the -march flag is
  668. # not enough.
  669. ZARCH_TEST_COMPILE=-c $(TOPDIR)/kernel/zarch/damin_z13.c -I$(TOPDIR) -o /dev/null > /dev/null 2> /dev/null
  670. ZARCH_CC_SUPPORTS_ARCH11=$(shell $(CC) -march=arch11 $(ZARCH_TEST_COMPILE) && echo 1)
  671. ZARCH_CC_SUPPORTS_Z13=$(shell $(CC) -march=z13 $(ZARCH_TEST_COMPILE) && echo 1)
  672. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH11), $(ZARCH_CC_SUPPORTS_Z13)), 1)
  673. DYNAMIC_CORE += Z13
  674. CCOMMON_OPT += -DDYN_Z13
  675. else
  676. $(info OpenBLAS: Not building Z13 kernels because the compiler $(CC) does not support it)
  677. endif
  678. # as above for z13, check for -march=arch12 and z14 support in the compiler.
  679. ZARCH_CC_SUPPORTS_ARCH12=$(shell $(CC) -march=arch12 $(ZARCH_TEST_COMPILE) && echo 1)
  680. ZARCH_CC_SUPPORTS_Z14=$(shell $(CC) -march=z14 $(ZARCH_TEST_COMPILE) && echo 1)
  681. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH12), $(ZARCH_CC_SUPPORTS_Z14)), 1)
  682. DYNAMIC_CORE += Z14
  683. CCOMMON_OPT += -DDYN_Z14
  684. else
  685. $(info OpenBLAS: Not building Z14 kernels because the compiler $(CC) does not support it)
  686. endif
  687. endif # ARCH zarch
  688. ifeq ($(ARCH), power)
  689. ifneq ($(C_COMPILER), PGI)
  690. DYNAMIC_CORE = POWER6
  691. DYNAMIC_CORE += POWER8
  692. ifneq ($(C_COMPILER), GCC)
  693. DYNAMIC_CORE += POWER9
  694. DYNAMIC_CORE += POWER10
  695. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  696. endif
  697. ifeq ($(C_COMPILER), GCC)
  698. ifeq ($(GCCVERSIONGT5), 1)
  699. DYNAMIC_CORE += POWER9
  700. else
  701. $(info, OpenBLAS: Your gcc version is too old to build the POWER9 kernels.)
  702. endif
  703. ifeq ($(OSNAME), AIX)
  704. LDVERSIONGTEQ35 := 1
  705. else
  706. LDVERSIONGTEQ35 := $(shell expr `$(CC) -Wl,--version 2> /dev/null | head -1 | cut -f2 -d "." | cut -f1 -d "-"` \>= 35)
  707. endif
  708. ifeq ($(GCCVERSIONGTEQ11)$(LDVERSIONGTEQ35), 11)
  709. DYNAMIC_CORE += POWER10
  710. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  711. else ifeq ($(GCCVERSIONGTEQ10), 1)
  712. ifeq ($(GCCMINORVERSIONGTEQ2)$(LDVERSIONGTEQ35), 11)
  713. DYNAMIC_CORE += POWER10
  714. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  715. endif
  716. else
  717. $(info, OpenBLAS: Your gcc version is too old to build the POWER10 kernels.)
  718. endif
  719. endif
  720. else
  721. DYNAMIC_CORE = POWER8
  722. DYNAMIC_CORE += POWER9
  723. endif
  724. endif
  725. # If DYNAMIC_CORE is not set, DYNAMIC_ARCH cannot do anything, so force it to empty
  726. ifndef DYNAMIC_CORE
  727. override DYNAMIC_ARCH=
  728. endif
  729. endif
  730. ifeq ($(ARCH), ia64)
  731. NO_BINARY_MODE = 1
  732. BINARY_DEFINED = 1
  733. ifeq ($(F_COMPILER), GFORTRAN)
  734. ifeq ($(C_COMPILER), GCC)
  735. EXPRECISION = 1
  736. CCOMMON_OPT += -DEXPRECISION
  737. endif
  738. endif
  739. endif
  740. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  741. NO_BINARY_MODE = 1
  742. endif
  743. ifeq ($(ARCH), alpha)
  744. NO_BINARY_MODE = 1
  745. BINARY_DEFINED = 1
  746. endif
  747. ifeq ($(ARCH), arm)
  748. NO_BINARY_MODE = 1
  749. BINARY_DEFINED = 1
  750. ifneq ($(EMBEDDED), 1)
  751. CCOMMON_OPT += -marm
  752. FCOMMON_OPT += -marm
  753. else
  754. CCOMMON_OPT += -DOS_EMBEDDED -mthumb -mcpu=cortex-m4 -mfloat-abi=hard -mfpu=fpv4-sp-d16
  755. endif
  756. # If softfp abi is mentioned on the command line, force it.
  757. ifeq ($(ARM_SOFTFP_ABI), 1)
  758. CCOMMON_OPT += -mfloat-abi=softfp
  759. FCOMMON_OPT += -mfloat-abi=softfp
  760. endif
  761. ifeq ($(OSNAME), Android)
  762. ifeq ($(ARM_SOFTFP_ABI), 1)
  763. EXTRALIB += -lm
  764. else
  765. EXTRALIB += -Wl,-lm_hard
  766. endif
  767. endif
  768. endif
  769. ifeq ($(ARCH), arm64)
  770. NO_BINARY_MODE = 1
  771. BINARY_DEFINED = 1
  772. ifdef INTERFACE64
  773. ifneq ($(INTERFACE64), 0)
  774. ifeq ($(F_COMPILER), GFORTRAN)
  775. FCOMMON_OPT += -fdefault-integer-8
  776. endif
  777. ifeq ($(F_COMPILER), FLANG)
  778. FCOMMON_OPT += -i8
  779. endif
  780. endif
  781. endif
  782. endif
  783. ifeq ($(ARCH), riscv64)
  784. NO_BINARY_MODE = 1
  785. BINARY_DEFINED = 1
  786. ifdef INTERFACE64
  787. ifneq ($(INTERFACE64), 0)
  788. ifeq ($(F_COMPILER), GFORTRAN)
  789. FCOMMON_OPT += -fdefault-integer-8
  790. endif
  791. ifeq ($(F_COMPILER), FLANG)
  792. FCOMMON_OPT += -i8
  793. endif
  794. endif
  795. endif
  796. endif
  797. ifeq ($(ARCH), loongarch64)
  798. NO_BINARY_MODE = 1
  799. BINARY_DEFINED = 1
  800. ifdef INTERFACE64
  801. ifneq ($(INTERFACE64), 0)
  802. ifeq ($(F_COMPILER), GFORTRAN)
  803. FCOMMON_OPT += -fdefault-integer-8
  804. endif
  805. ifeq ($(F_COMPILER), FLANG)
  806. FCOMMON_OPT += -i8
  807. endif
  808. endif
  809. endif
  810. endif
  811. ifeq ($(ARCH), csky)
  812. NO_BINARY_MODE = 1
  813. BINARY_DEFINED = 1
  814. endif
  815. #
  816. # C Compiler dependent settings
  817. #
  818. # ifeq logical or. GCC or CLANG or LSB
  819. # http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
  820. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG LSB))
  821. CCOMMON_OPT += -Wall
  822. COMMON_PROF += -fno-inline
  823. NO_UNINITIALIZED_WARN = -Wno-uninitialized
  824. ifeq ($(QUIET_MAKE), 1)
  825. CCOMMON_OPT += $(NO_UNINITIALIZED_WARN) -Wno-unused
  826. endif
  827. ifdef NO_BINARY_MODE
  828. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  829. ifdef BINARY64
  830. CCOMMON_OPT += -mabi=64
  831. else
  832. CCOMMON_OPT += -mabi=n32
  833. endif
  834. BINARY_DEFINED = 1
  835. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  836. CCOMMON_OPT += -mabi=32
  837. BINARY_DEFINED = 1
  838. endif
  839. ifneq (, $(filter $(CORE), MIPS64_GENERIC))
  840. CCOMMON_OPT += -DNO_MSA
  841. FCOMMON_OPT += -DNO_MSA
  842. endif
  843. ifneq (, $(filter $(CORE),LOONGSON3R3 LOONGSON3R4))
  844. CCOMMON_OPT += -march=loongson3a
  845. FCOMMON_OPT += -march=loongson3a
  846. endif
  847. ifeq ($(CORE), MIPS24K)
  848. CCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  849. FCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  850. endif
  851. ifeq ($(CORE), MIPS1004K)
  852. CCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  853. FCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  854. endif
  855. ifeq ($(CORE), P5600)
  856. CCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  857. FCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  858. endif
  859. ifeq ($(CORE), I6400)
  860. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  861. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  862. endif
  863. ifeq ($(CORE), P6600)
  864. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  865. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  866. endif
  867. ifeq ($(CORE), I6500)
  868. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  869. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  870. endif
  871. ifeq ($(OSNAME), AIX)
  872. BINARY_DEFINED = 1
  873. endif
  874. ifeq ($(ARCH), loongarch64)
  875. LA64_ABI=$(shell $(CC) -mabi=lp64d -c $(TOPDIR)/cpuid_loongarch64.c -o /dev/null > /dev/null 2> /dev/null && echo lp64d)
  876. LA64_ARCH=$(shell $(CC) -march=loongarch64 -c $(TOPDIR)/cpuid_loongarch64.c -o /dev/null > /dev/null 2> /dev/null && echo loongarch64)
  877. ifneq ($(LA64_ABI), lp64d)
  878. LA64_ABI=lp64
  879. endif
  880. ifneq ($(LA64_ARCH), loongarch64)
  881. CCOMMON_OPT += -mabi=$(LA64_ABI)
  882. FCOMMON_OPT += -mabi=$(LA64_ABI)
  883. else
  884. CCOMMON_OPT += -march=loongarch64 -mabi=$(LA64_ABI)
  885. FCOMMON_OPT += -march=loongarch64 -mabi=$(LA64_ABI)
  886. endif
  887. endif
  888. endif
  889. ifndef BINARY_DEFINED
  890. ifneq ($(OSNAME), AIX)
  891. ifdef BINARY64
  892. ifneq ($(ARCH), riscv64)
  893. CCOMMON_OPT += -m64
  894. endif
  895. else
  896. CCOMMON_OPT += -m32
  897. endif
  898. endif
  899. endif
  900. endif
  901. ifeq ($(C_COMPILER), PGI)
  902. PGCVERSIONGT20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \> 20)
  903. PGCVERSIONEQ20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` == 20)
  904. PGCMINORVERSIONGE11 := $(shell expr `$(CC) --version|sed -n "2p" |cut -d "-" -f 1 |sed -e "s/[^0-9.]//g" |cut -c 4-5` \>= 11)
  905. PGCVERSIONCHECK := $(PGCVERSIONGT20)$(PGCVERSIONEQ20)$(PGCMINORVERSIONGE11)
  906. ifeq ($(PGCVERSIONCHECK), $(filter $(PGCVERSIONCHECK), 100 101 011))
  907. NEWPGI := 1
  908. PGCVERSIONGT21 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \> 21)
  909. PGCVERSIONEQ21 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` == 21)
  910. PGCVERSIONCHECK2 := $(PGCVERSIONGT21)$(PGCVERSIONEQ21)$(PGCMINORVERSIONGE11)
  911. ifeq ($(PGCVERSIONCHECK2), $(filter $(PGCVERSIONCHECK2), 100 101 011))
  912. NEWPGI2 := 1
  913. endif
  914. endif
  915. ifdef BINARY64
  916. ifeq ($(ARCH), x86_64)
  917. ifeq (,$(findstring tp,$(CFLAGS)))
  918. ifneq ($(NEWPGI2),1)
  919. CCOMMON_OPT += -tp p7-64
  920. else
  921. CCOMMON_OPT += -tp px
  922. endif
  923. endif
  924. ifneq ($(NEWPGI),1)
  925. CCOMMON_OPT += -D__MMX__ -Mnollvm
  926. endif
  927. else
  928. ifeq ($(ARCH), power)
  929. ifeq (,$(findstring tp,$(CFLAGS)))
  930. ifeq ($(CORE), POWER8)
  931. CCOMMON_OPT += -tp pwr8
  932. endif
  933. ifeq ($(CORE), POWER9)
  934. CCOMMON_OPT += -tp pwr9
  935. endif
  936. endif
  937. endif
  938. endif
  939. else
  940. ifneq ($(NEWPGI2),1)
  941. ifeq (,$(findstring tp,$(CFLAGS)))
  942. CCOMMON_OPT += -tp p7
  943. else
  944. CCOMMON_OPT += -tp px
  945. endif
  946. endif
  947. endif
  948. endif
  949. ifeq ($(C_COMPILER), PATHSCALE)
  950. ifdef BINARY64
  951. CCOMMON_OPT += -m64
  952. else
  953. CCOMMON_OPT += -m32
  954. endif
  955. endif
  956. #
  957. # Fortran Compiler dependent settings
  958. #
  959. ifeq ($(F_COMPILER), NAG)
  960. FCOMMON_OPT += -dcfuns -recursive -ieee=full -w=obs -thread_safe
  961. ifdef INTERFACE64
  962. ifneq ($(INTERFACE64), 0)
  963. FCOMMON_OPT += -i8
  964. endif
  965. endif
  966. ifeq ($(USE_OPENMP), 1)
  967. FCOMMON_OPT += -openmp
  968. endif
  969. endif
  970. ifeq ($(F_COMPILER), FLANG)
  971. CCOMMON_OPT += -DF_INTERFACE_FLANG
  972. FCOMMON_OPT += -Mrecursive -Kieee
  973. ifeq ($(OSNAME), Linux)
  974. ifeq ($(ARCH), x86_64)
  975. FLANG_VENDOR := $(shell $(FC) --version|head -1 |cut -f 1 -d " ")
  976. ifeq ($(FLANG_VENDOR), AMD)
  977. FCOMMON_OPT += -fno-unroll-loops
  978. endif
  979. endif
  980. endif
  981. ifdef BINARY64
  982. ifdef INTERFACE64
  983. ifneq ($(INTERFACE64), 0)
  984. FCOMMON_OPT += -i8
  985. endif
  986. endif
  987. FCOMMON_OPT += -Wall
  988. else
  989. FCOMMON_OPT += -Wall
  990. endif
  991. ifeq ($(USE_OPENMP), 1)
  992. FCOMMON_OPT += -fopenmp
  993. endif
  994. endif
  995. ifeq ($(F_COMPILER), G77)
  996. CCOMMON_OPT += -DF_INTERFACE_G77
  997. FCOMMON_OPT += -Wall
  998. ifndef NO_BINARY_MODE
  999. ifneq ($(OSNAME), AIX)
  1000. ifdef BINARY64
  1001. FCOMMON_OPT += -m64
  1002. else
  1003. FCOMMON_OPT += -m32
  1004. endif
  1005. endif
  1006. endif
  1007. endif
  1008. ifeq ($(F_COMPILER), G95)
  1009. CCOMMON_OPT += -DF_INTERFACE_G95
  1010. FCOMMON_OPT += -Wall
  1011. ifneq ($(OSNAME), AIX)
  1012. ifndef NO_BINARY_MODE
  1013. ifdef BINARY64
  1014. FCOMMON_OPT += -m64
  1015. else
  1016. FCOMMON_OPT += -m32
  1017. endif
  1018. endif
  1019. ifneq ($(NO_LAPACKE), 1)
  1020. FCOMMON_OPT += -fno-second-underscore
  1021. endif
  1022. endif
  1023. endif
  1024. ifeq ($(F_COMPILER), $(filter $(F_COMPILER),GFORTRAN FLANGNEW))
  1025. CCOMMON_OPT += -DF_INTERFACE_GFORT
  1026. ifeq ($(F_COMPILER), GFORTRAN)
  1027. FCOMMON_OPT += -Wall
  1028. # make single-threaded LAPACK calls thread-safe #1847
  1029. FCOMMON_OPT += -frecursive
  1030. # work around ABI problem with passing single-character arguments
  1031. FCOMMON_OPT += -fno-optimize-sibling-calls
  1032. #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
  1033. ifneq ($(NOFORTRAN), 1)
  1034. ifneq ($(NOFORTRAN), 2)
  1035. ifneq ($(NO_LAPACK), 1)
  1036. EXTRALIB += -lgfortran
  1037. endif
  1038. endif
  1039. endif
  1040. endif
  1041. ifdef NO_BINARY_MODE
  1042. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  1043. ifdef BINARY64
  1044. FCOMMON_OPT += -mabi=64
  1045. else
  1046. FCOMMON_OPT += -mabi=n32
  1047. endif
  1048. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  1049. FCOMMON_OPT += -mabi=32
  1050. endif
  1051. ifeq ($(ARCH), $(filter $(ARCH),loongarch64))
  1052. ifdef INTERFACE64
  1053. ifneq ($(INTERFACE64), 0)
  1054. FCOMMON_OPT += -fdefault-integer-8
  1055. endif
  1056. endif
  1057. endif
  1058. else
  1059. ifdef BINARY64
  1060. ifneq ($(OSNAME), AIX)
  1061. ifneq ($(ARCH), riscv64)
  1062. FCOMMON_OPT += -m64
  1063. endif
  1064. endif
  1065. ifdef INTERFACE64
  1066. ifneq ($(INTERFACE64), 0)
  1067. FCOMMON_OPT += -fdefault-integer-8
  1068. endif
  1069. endif
  1070. else
  1071. ifneq ($(OSNAME), AIX)
  1072. FCOMMON_OPT += -m32
  1073. endif
  1074. endif
  1075. endif
  1076. ifeq ($(USE_OPENMP), 1)
  1077. FCOMMON_OPT += -fopenmp
  1078. endif
  1079. endif
  1080. ifeq ($(F_COMPILER), INTEL)
  1081. CCOMMON_OPT += -DF_INTERFACE_INTEL
  1082. ifdef INTERFACE64
  1083. ifneq ($(INTERFACE64), 0)
  1084. FCOMMON_OPT += -i8
  1085. endif
  1086. endif
  1087. FCOMMON_OPT += -recursive -fp-model strict -assume protect-parens
  1088. ifeq ($(USE_OPENMP), 1)
  1089. FCOMMON_OPT += -fopenmp
  1090. endif
  1091. endif
  1092. ifeq ($(F_COMPILER), FUJITSU)
  1093. CCOMMON_OPT += -DF_INTERFACE_FUJITSU
  1094. ifeq ($(USE_OPENMP), 1)
  1095. FCOMMON_OPT += -openmp
  1096. endif
  1097. endif
  1098. ifeq ($(F_COMPILER), IBM)
  1099. CCOMMON_OPT += -DF_INTERFACE_IBM
  1100. FEXTRALIB += -lxlf90
  1101. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG))
  1102. FCOMMON_OPT += -qextname -qzerosize
  1103. endif
  1104. # FCOMMON_OPT += -qarch=440
  1105. ifdef BINARY64
  1106. FCOMMON_OPT += -q64
  1107. ifdef INTERFACE64
  1108. ifneq ($(INTERFACE64), 0)
  1109. FCOMMON_OPT += -qintsize=8
  1110. endif
  1111. endif
  1112. else
  1113. FCOMMON_OPT += -q32
  1114. endif
  1115. endif
  1116. ifeq ($(F_COMPILER), PGI)
  1117. CCOMMON_OPT += -DF_INTERFACE_PGI
  1118. COMMON_PROF += -DPGICOMPILER
  1119. ifdef BINARY64
  1120. ifdef INTERFACE64
  1121. ifneq ($(INTERFACE64), 0)
  1122. FCOMMON_OPT += -i8
  1123. endif
  1124. endif
  1125. ifeq ($(ARCH), x86_64)
  1126. ifneq ($(NEWPGI2),1)
  1127. FCOMMON_OPT += -tp p7-64
  1128. else
  1129. FCOMMON_OPT += -tp px
  1130. endif
  1131. else
  1132. ifeq ($(ARCH), power)
  1133. ifeq ($(CORE), POWER6)
  1134. $(warning NVIDIA HPC compilers do not support POWER6.)
  1135. endif
  1136. ifeq ($(CORE), POWER8)
  1137. FCOMMON_OPT += -tp pwr8
  1138. endif
  1139. ifeq ($(CORE), POWER9)
  1140. FCOMMON_OPT += -tp pwr9
  1141. endif
  1142. ifeq ($(CORE), POWER10)
  1143. $(warning NVIDIA HPC compilers do not support POWER10.)
  1144. endif
  1145. endif
  1146. endif
  1147. else
  1148. FCOMMON_OPT += -tp p7
  1149. endif
  1150. FCOMMON_OPT += -Mrecursive -Kieee
  1151. ifeq ($(USE_OPENMP), 1)
  1152. FCOMMON_OPT += -mp
  1153. endif
  1154. endif
  1155. ifeq ($(F_COMPILER), PATHSCALE)
  1156. CCOMMON_OPT += -DF_INTERFACE_PATHSCALE
  1157. ifdef BINARY64
  1158. ifdef INTERFACE64
  1159. ifneq ($(INTERFACE64), 0)
  1160. FCOMMON_OPT += -i8
  1161. endif
  1162. endif
  1163. endif
  1164. ifeq ($(USE_OPENMP), 1)
  1165. FCOMMON_OPT += -mp
  1166. endif
  1167. endif
  1168. ifeq ($(F_COMPILER), OPEN64)
  1169. CCOMMON_OPT += -DF_INTERFACE_OPEN64
  1170. ifdef BINARY64
  1171. ifdef INTERFACE64
  1172. ifneq ($(INTERFACE64), 0)
  1173. FCOMMON_OPT += -i8
  1174. endif
  1175. endif
  1176. endif
  1177. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  1178. ifndef BINARY64
  1179. FCOMMON_OPT += -n32
  1180. else
  1181. FCOMMON_OPT += -n64
  1182. endif
  1183. ifeq ($(CORE), LOONGSON3R3)
  1184. FCOMMON_OPT += -loongson3 -static
  1185. endif
  1186. ifeq ($(CORE), LOONGSON3R4)
  1187. FCOMMON_OPT += -loongson3 -static
  1188. endif
  1189. else
  1190. ifndef BINARY64
  1191. FCOMMON_OPT += -m32
  1192. else
  1193. FCOMMON_OPT += -m64
  1194. endif
  1195. endif
  1196. ifeq ($(USE_OPENMP), 1)
  1197. FEXTRALIB += -lstdc++
  1198. FCOMMON_OPT += -mp
  1199. endif
  1200. endif
  1201. ifeq ($(C_COMPILER), OPEN64)
  1202. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  1203. ifndef BINARY64
  1204. CCOMMON_OPT += -n32
  1205. else
  1206. CCOMMON_OPT += -n64
  1207. endif
  1208. ifeq ($(CORE), LOONGSON3R3)
  1209. CCOMMON_OPT += -loongson3 -static
  1210. endif
  1211. ifeq ($(CORE), LOONGSON3R4)
  1212. CCOMMON_OPT += -loongson3 -static
  1213. endif
  1214. else
  1215. ifndef BINARY64
  1216. CCOMMON_OPT += -m32
  1217. else
  1218. CCOMMON_OPT += -m64
  1219. endif
  1220. endif
  1221. endif
  1222. ifeq ($(C_COMPILER), SUN)
  1223. CCOMMON_OPT += -w
  1224. ifeq ($(ARCH), x86)
  1225. CCOMMON_OPT += -m32
  1226. else
  1227. ifdef BINARY64
  1228. CCOMMON_OPT += -m64
  1229. else
  1230. CCOMMON_OPT += -m32
  1231. endif
  1232. endif
  1233. endif
  1234. ifeq ($(F_COMPILER), SUN)
  1235. CCOMMON_OPT += -DF_INTERFACE_SUN
  1236. FCOMMON_OPT += -ftrap=%none -xrecursive
  1237. ifeq ($(ARCH), x86)
  1238. FCOMMON_OPT += -m32
  1239. else
  1240. ifdef BINARY64
  1241. FCOMMON_OPT += -m64
  1242. else
  1243. FCOMMON_OPT += -m32
  1244. endif
  1245. endif
  1246. ifeq ($(USE_OPENMP), 1)
  1247. FCOMMON_OPT += -xopenmp=parallel
  1248. endif
  1249. endif
  1250. ifeq ($(F_COMPILER), COMPAQ)
  1251. CCOMMON_OPT += -DF_INTERFACE_COMPAQ
  1252. ifeq ($(USE_OPENMP), 1)
  1253. FCOMMON_OPT += -openmp
  1254. endif
  1255. endif
  1256. ifeq ($(F_COMPILER), CRAY)
  1257. CCOMMON_OPT += -DF_INTERFACE_INTEL
  1258. FCOMMON_OPT += -hnopattern
  1259. ifdef INTERFACE64
  1260. ifneq ($(INTERFACE64), 0)
  1261. FCOMMON_OPT += -s integer64
  1262. endif
  1263. endif
  1264. ifneq ($(USE_OPENMP), 1)
  1265. FCOMMON_OPT += -O noomp
  1266. endif
  1267. endif
  1268. ifdef BINARY64
  1269. ifdef INTERFACE64
  1270. ifneq ($(INTERFACE64), 0)
  1271. CCOMMON_OPT +=
  1272. #-DUSE64BITINT
  1273. endif
  1274. endif
  1275. endif
  1276. ifeq ($(NEED_PIC), 1)
  1277. ifeq ($(C_COMPILER), IBM)
  1278. CCOMMON_OPT += -qpic=large
  1279. else
  1280. CCOMMON_OPT += -fPIC
  1281. endif
  1282. ifeq ($(F_COMPILER), SUN)
  1283. FCOMMON_OPT += -pic
  1284. else ifeq ($(F_COMPILER), NAG)
  1285. FCOMMON_OPT += -PIC
  1286. else ifeq ($(F_COMPILER), IBM)
  1287. FCOMMON_OPT += -qpic=large
  1288. else
  1289. FCOMMON_OPT += -fPIC
  1290. endif
  1291. endif
  1292. ifeq ($(DYNAMIC_ARCH), 1)
  1293. CCOMMON_OPT += -DDYNAMIC_ARCH
  1294. endif
  1295. ifeq ($(DYNAMIC_OLDER), 1)
  1296. CCOMMON_OPT += -DDYNAMIC_OLDER
  1297. endif
  1298. ifeq ($(C_LAPACK), 1)
  1299. CCOMMON_OPT += -DC_LAPACK
  1300. endif
  1301. ifeq ($(NO_LAPACK), 1)
  1302. CCOMMON_OPT += -DNO_LAPACK
  1303. #Disable LAPACK C interface
  1304. NO_LAPACKE = 1
  1305. endif
  1306. ifeq ($(NO_LAPACKE), 1)
  1307. CCOMMON_OPT += -DNO_LAPACKE
  1308. endif
  1309. ifeq ($(NO_AVX), 1)
  1310. CCOMMON_OPT += -DNO_AVX
  1311. endif
  1312. ifeq ($(ARCH), x86)
  1313. CCOMMON_OPT += -DNO_AVX
  1314. endif
  1315. ifeq ($(NO_AVX2), 1)
  1316. CCOMMON_OPT += -DNO_AVX2
  1317. endif
  1318. ifeq ($(NO_AVX512), 1)
  1319. CCOMMON_OPT += -DNO_AVX512
  1320. endif
  1321. ifeq ($(NO_SVE), 1)
  1322. CCOMMON_OPT += -DNO_SVE
  1323. endif
  1324. ifeq ($(NO_SME), 1)
  1325. CCOMMON_OPT += -DNO_SME
  1326. endif
  1327. ifdef SMP
  1328. CCOMMON_OPT += -DSMP_SERVER
  1329. ifeq ($(ARCH), mips64)
  1330. USE_SIMPLE_THREADED_LEVEL3 = 1
  1331. endif
  1332. ifeq ($(USE_OPENMP), 1)
  1333. # USE_SIMPLE_THREADED_LEVEL3 = 1
  1334. # NO_AFFINITY = 1
  1335. CCOMMON_OPT += -DUSE_OPENMP
  1336. endif
  1337. ifeq ($(BIGNUMA), 1)
  1338. CCOMMON_OPT += -DBIGNUMA
  1339. endif
  1340. endif
  1341. ifeq ($(NO_WARMUP), 1)
  1342. CCOMMON_OPT += -DNO_WARMUP
  1343. endif
  1344. ifeq ($(CONSISTENT_FPCSR), 1)
  1345. CCOMMON_OPT += -DCONSISTENT_FPCSR
  1346. endif
  1347. # Only for development
  1348. # CCOMMON_OPT += -DPARAMTEST
  1349. # CCOMMON_OPT += -DPREFETCHTEST
  1350. # CCOMMON_OPT += -DNO_SWITCHING
  1351. # USE_PAPI = 1
  1352. ifdef USE_PAPI
  1353. CCOMMON_OPT += -DUSE_PAPI
  1354. EXTRALIB += -lpapi -lperfctr
  1355. endif
  1356. ifdef BUFFERSIZE
  1357. CCOMMON_OPT += -DBUFFERSIZE=$(BUFFERSIZE)
  1358. endif
  1359. ifdef DYNAMIC_THREADS
  1360. CCOMMON_OPT += -DDYNAMIC_THREADS
  1361. endif
  1362. CCOMMON_OPT += -DMAX_CPU_NUMBER=$(NUM_THREADS)
  1363. CCOMMON_OPT += -DMAX_PARALLEL_NUMBER=$(NUM_PARALLEL)
  1364. ifdef USE_SIMPLE_THREADED_LEVEL3
  1365. CCOMMON_OPT += -DUSE_SIMPLE_THREADED_LEVEL3
  1366. endif
  1367. ifeq ($(USE_TLS), 1)
  1368. CCOMMON_OPT += -DUSE_TLS
  1369. endif
  1370. ifeq ($(BUILD_BFLOAT16), 1)
  1371. CCOMMON_OPT += -DBUILD_BFLOAT16
  1372. endif
  1373. ifeq ($(BUILD_HFLOAT16), 1)
  1374. CCOMMON_OPT += -DBUILD_HFLOAT16
  1375. endif
  1376. ifeq ($(BUILD_SINGLE), 1)
  1377. CCOMMON_OPT += -DBUILD_SINGLE=1
  1378. endif
  1379. ifeq ($(BUILD_DOUBLE), 1)
  1380. CCOMMON_OPT += -DBUILD_DOUBLE=1
  1381. endif
  1382. ifeq ($(BUILD_COMPLEX), 1)
  1383. CCOMMON_OPT += -DBUILD_COMPLEX=1
  1384. endif
  1385. ifeq ($(BUILD_COMPLEX16), 1)
  1386. CCOMMON_OPT += -DBUILD_COMPLEX16=1
  1387. endif
  1388. CCOMMON_OPT += -DVERSION=\"$(VERSION)\"
  1389. ifndef SYMBOLPREFIX
  1390. SYMBOLPREFIX =
  1391. endif
  1392. ifndef SYMBOLSUFFIX
  1393. SYMBOLSUFFIX =
  1394. endif
  1395. ifndef LIBSONAMEBASE
  1396. LIBSONAMEBASE = openblas
  1397. endif
  1398. ifndef LIBNAMEPREFIX
  1399. LIBNAMEPREFIX =
  1400. endif
  1401. SYMPREFIX=$(SYMBOLPREFIX)
  1402. ifeq ($(SYMBOLPREFIX),$(LIBNAMEPREFIX))
  1403. SYMPREFIX=
  1404. endif
  1405. SYMSUFFIX=$(SYMBOLSUFFIX)
  1406. ifeq ($(SYMBOLSUFFIX),$(LIBNAMESUFFIX))
  1407. SYMSUFFIX=
  1408. endif
  1409. ifndef LIBNAMESUFFIX
  1410. LIBNAMEBASE = $(SYMPREFIX)$(LIBSONAMEBASE)$(SYMSUFFIX)
  1411. else
  1412. LIBNAMEBASE = $(SYMPREFIX)$(LIBSONAMEBASE)$(SYMSUFFIX)$(LIBNAMESUFFIX)
  1413. endif
  1414. ifeq ($(OSNAME), CYGWIN_NT)
  1415. LIBPREFIX = cyg$(LIBNAMEPREFIX)$(LIBNAMEBASE)
  1416. else
  1417. LIBPREFIX = lib$(LIBNAMEPREFIX)$(LIBNAMEBASE)
  1418. endif
  1419. KERNELDIR = $(TOPDIR)/kernel/$(ARCH)
  1420. include $(TOPDIR)/Makefile.$(ARCH)
  1421. ifneq ($(C_COMPILER), PGI)
  1422. ifneq ($(C_COMPILER), SUN)
  1423. CCOMMON_OPT += -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME
  1424. endif
  1425. endif
  1426. CCOMMON_OPT += -DASMNAME=$(FU)$(*F) -DASMFNAME=$(FU)$(*F)$(BU) -DNAME=$(*F)$(BU) -DCNAME=$(*F) -DCHAR_NAME=\"$(*F)$(BU)\" -DCHAR_CNAME=\"$(*F)\"
  1427. ifeq ($(CORE), PPC440)
  1428. CCOMMON_OPT += -DALLOC_QALLOC
  1429. endif
  1430. ifeq ($(CORE), PPC440FP2)
  1431. STATIC_ALLOCATION = 1
  1432. endif
  1433. ifneq ($(OSNAME), Linux)
  1434. NO_AFFINITY = 1
  1435. endif
  1436. ifneq ($(ARCH), x86_64)
  1437. ifneq ($(ARCH), x86)
  1438. NO_AFFINITY = 1
  1439. endif
  1440. endif
  1441. ifeq ($(ARCH), POWER)
  1442. ifeq ($(DEBUG), 1)
  1443. CCOMMON_OPT := $(filter-out -O%, $(CCOMMON_OPT)) -O0
  1444. FCOMMON_OPT := $(filter-out -O%, $(FCOMMON_OPT)) -O0
  1445. endif
  1446. endif
  1447. ifdef NO_AFFINITY
  1448. ifeq ($(NO_AFFINITY), 0)
  1449. override undefine NO_AFFINITY
  1450. else
  1451. CCOMMON_OPT += -DNO_AFFINITY
  1452. endif
  1453. endif
  1454. ifdef FUNCTION_PROFILE
  1455. CCOMMON_OPT += -DFUNCTION_PROFILE
  1456. endif
  1457. ifdef SHMEM_ALLOCATION
  1458. ifneq ($(SHMEM_ALLOCATION), 0)
  1459. CCOMMON_OPT += -DALLOC_SHM
  1460. endif
  1461. endif
  1462. ifdef HUGETLB_ALLOCATION
  1463. ifneq ($(HUGETLB_ALLOCATION), 0)
  1464. CCOMMON_OPT += -DALLOC_HUGETLB
  1465. endif
  1466. endif
  1467. ifdef HUGETLBFILE_ALLOCATION
  1468. ifneq ($(HUGETLBFILE_ALLOCATION), 0)
  1469. CCOMMON_OPT += -DALLOC_HUGETLBFILE -DHUGETLB_FILE_NAME=$(HUGETLBFILE_ALLOCATION)
  1470. endif
  1471. endif
  1472. ifdef STATIC_ALLOCATION
  1473. CCOMMON_OPT += -DALLOC_STATIC
  1474. endif
  1475. ifdef DEVICEDRIVER_ALLOCATION
  1476. CCOMMON_OPT += -DALLOC_DEVICEDRIVER -DDEVICEDRIVER_NAME=\"/dev/mapper\"
  1477. endif
  1478. ifdef MIXED_MEMORY_ALLOCATION
  1479. CCOMMON_OPT += -DMIXED_MEMORY_ALLOCATION
  1480. endif
  1481. ifeq ($(OSNAME), SunOS)
  1482. TAR = gtar
  1483. PATCH = gpatch
  1484. GREP = ggrep
  1485. AWK = nawk
  1486. else
  1487. TAR = tar
  1488. PATCH = patch
  1489. GREP = grep
  1490. AWK = awk
  1491. endif
  1492. ifndef MD5SUM
  1493. MD5SUM = md5sum
  1494. endif
  1495. REVISION = -r$(VERSION)
  1496. MAJOR_VERSION = $(word 1,$(subst ., ,$(VERSION)))
  1497. ifeq ($(DEBUG), 1)
  1498. COMMON_OPT += -g
  1499. endif
  1500. ifeq ($(DEBUG), 1)
  1501. FCOMMON_OPT += -g
  1502. endif
  1503. ifndef COMMON_OPT
  1504. COMMON_OPT = -O2
  1505. endif
  1506. ifndef FCOMMON_OPT
  1507. FCOMMON_OPT = -O2 -frecursive
  1508. endif
  1509. override CFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
  1510. override PFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
  1511. override FFLAGS += $(COMMON_OPT) $(FCOMMON_OPT)
  1512. override FPFLAGS += $(FCOMMON_OPT) $(COMMON_PROF)
  1513. #MAKEOVERRIDES =
  1514. ifeq ($(NEED_PIC), 1)
  1515. ifeq (,$(findstring PIC,$(FFLAGS)))
  1516. ifneq ($(F_COMPILER),IBM)
  1517. override FFLAGS += -fPIC
  1518. endif
  1519. endif
  1520. endif
  1521. #For LAPACK Fortran codes.
  1522. #Disable -fopenmp for LAPACK Fortran codes on Windows.
  1523. ifdef OS_WINDOWS
  1524. LAPACK_FFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FFLAGS))
  1525. LAPACK_FPFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FPFLAGS))
  1526. else
  1527. LAPACK_FFLAGS := $(FFLAGS)
  1528. LAPACK_FPFLAGS := $(FPFLAGS)
  1529. endif
  1530. ifeq ($(F_COMPILER),NAG)
  1531. LAPACK_FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1532. override FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1533. endif
  1534. ifeq ($(F_COMPILER),CRAY)
  1535. LAPACK_FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1536. override FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1537. endif
  1538. ifeq ($(F_COMPILER),FLANGNEW)
  1539. LAPACK_FFLAGS := $(filter-out -m32 -m64 -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 -mtune=% -mabi=% ,$(FFLAGS))
  1540. override FFLAGS := $(filter-out -m32 -m64 -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 -mtune=% -mabi=% ,$(FFLAGS))
  1541. endif
  1542. LAPACK_CFLAGS = $(CFLAGS)
  1543. LAPACK_CFLAGS += -DHAVE_LAPACK_CONFIG_H
  1544. ifdef INTERFACE64
  1545. ifneq ($(INTERFACE64), 0)
  1546. LAPACK_CFLAGS += -DLAPACK_ILP64
  1547. endif
  1548. endif
  1549. ifdef OS_WINDOWS
  1550. LAPACK_CFLAGS += -DOPENBLAS_OS_WINDOWS
  1551. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1552. endif
  1553. ifeq ($(C_COMPILER), LSB)
  1554. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1555. endif
  1556. ifndef SUFFIX
  1557. SUFFIX = o
  1558. endif
  1559. ifndef PSUFFIX
  1560. PSUFFIX = po
  1561. endif
  1562. ifndef LIBSUFFIX
  1563. LIBSUFFIX = a
  1564. endif
  1565. ifneq ($(DYNAMIC_ARCH), 1)
  1566. ifndef SMP
  1567. LIBNAME = $(LIBPREFIX)_$(LIBCORE)$(REVISION).$(LIBSUFFIX)
  1568. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)$(REVISION)_p.$(LIBSUFFIX)
  1569. else
  1570. LIBNAME = $(LIBPREFIX)_$(LIBCORE)p$(REVISION).$(LIBSUFFIX)
  1571. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)p$(REVISION)_p.$(LIBSUFFIX)
  1572. endif
  1573. else
  1574. ifndef SMP
  1575. LIBNAME = $(LIBPREFIX)$(REVISION).$(LIBSUFFIX)
  1576. LIBNAME_P = $(LIBPREFIX)$(REVISION)_p.$(LIBSUFFIX)
  1577. else
  1578. LIBNAME = $(LIBPREFIX)p$(REVISION).$(LIBSUFFIX)
  1579. LIBNAME_P = $(LIBPREFIX)p$(REVISION)_p.$(LIBSUFFIX)
  1580. endif
  1581. endif
  1582. ifeq ($(FIXED_LIBNAME),1)
  1583. LIBNAME = lib$(LIBNAMEPREFIX)$(LIBSONAMEBASE)$(LIBNAMESUFFIX).$(LIBSUFFIX)
  1584. LIBNAME_P = lib$(LIBNAMEPREFIX)$(LISOBNAMEBASE)$(LIBNAMESUFFIX)_p.$(LIBSUFFIX)
  1585. endif
  1586. LIBDLLNAME = $(LIBPREFIX).dll
  1587. IMPLIBNAME = lib$(LIBNAMEBASE).dll.a
  1588. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.so)
  1589. LIBDYNNAME = $(LIBNAME:.$(LIBSUFFIX)=.dylib)
  1590. LIBDEFNAME = $(LIBNAME:.$(LIBSUFFIX)=.def)
  1591. LIBEXPNAME = $(LIBNAME:.$(LIBSUFFIX)=.exp)
  1592. LIBZIPNAME = $(LIBNAME:.$(LIBSUFFIX)=.zip)
  1593. LIBS = $(TOPDIR)/$(LIBNAME)
  1594. LIBS_P = $(TOPDIR)/$(LIBNAME_P)
  1595. LIB_COMPONENTS = BLAS
  1596. ifneq ($(NO_CBLAS), 1)
  1597. LIB_COMPONENTS += CBLAS
  1598. endif
  1599. ifneq ($(NO_LAPACK), 1)
  1600. LIB_COMPONENTS += LAPACK
  1601. ifneq ($(NO_LAPACKE), 1)
  1602. LIB_COMPONENTS += LAPACKE
  1603. endif
  1604. ifeq ($(BUILD_RELAPACK), 1)
  1605. LIB_COMPONENTS += ReLAPACK
  1606. endif
  1607. endif
  1608. ifeq ($(ONLY_CBLAS), 1)
  1609. LIB_COMPONENTS = CBLAS
  1610. endif
  1611. export OSNAME
  1612. export ARCH
  1613. export CORE
  1614. export LIBCORE
  1615. export __BYTE_ORDER__
  1616. export ELF_VERSION
  1617. export PGCPATH
  1618. export CONFIG
  1619. export CC
  1620. export FC
  1621. export BU
  1622. export FU
  1623. export NEED2UNDERSCORES
  1624. export USE_THREAD
  1625. export NUM_THREADS
  1626. export NUM_CORES
  1627. export SMP
  1628. export MAKEFILE_RULE
  1629. export NEED_PIC
  1630. export BINARY
  1631. export BINARY32
  1632. export BINARY64
  1633. export F_COMPILER
  1634. export C_COMPILER
  1635. export USE_OPENMP
  1636. export CROSS
  1637. export CROSS_SUFFIX
  1638. export NOFORTRAN
  1639. export C_LAPACK
  1640. export NO_FBLAS
  1641. export EXTRALIB
  1642. export CEXTRALIB
  1643. export FEXTRALIB
  1644. export HAVE_SSE
  1645. export HAVE_SSE2
  1646. export HAVE_SSE3
  1647. export HAVE_SSSE3
  1648. export HAVE_SSE4_1
  1649. export HAVE_SSE4_2
  1650. export HAVE_SSE4A
  1651. export HAVE_SSE5
  1652. export HAVE_AVX
  1653. export HAVE_AVX2
  1654. export HAVE_FMA3
  1655. export HAVE_VFP
  1656. export HAVE_VFPV3
  1657. export HAVE_VFPV4
  1658. export HAVE_NEON
  1659. ifndef NO_MSA
  1660. export HAVE_MSA
  1661. export MSA_FLAGS
  1662. endif
  1663. export KERNELDIR
  1664. export FUNCTION_PROFILE
  1665. export TARGET_CORE
  1666. export NO_AVX512
  1667. export NO_AVX2
  1668. export BUILD_BFLOAT16
  1669. export BUILD_HFLOAT16
  1670. export NO_LSX
  1671. export NO_LASX
  1672. export BGEMM_UNROLL_M
  1673. export BGEMM_UNROLL_N
  1674. export SBGEMM_UNROLL_M
  1675. export SBGEMM_UNROLL_N
  1676. export SHGEMM_UNROLL_M
  1677. export SHGEMM_UNROLL_N
  1678. export SGEMM_UNROLL_M
  1679. export SGEMM_UNROLL_N
  1680. export DGEMM_UNROLL_M
  1681. export DGEMM_UNROLL_N
  1682. export QGEMM_UNROLL_M
  1683. export QGEMM_UNROLL_N
  1684. export CGEMM_UNROLL_M
  1685. export CGEMM_UNROLL_N
  1686. export ZGEMM_UNROLL_M
  1687. export ZGEMM_UNROLL_N
  1688. export XGEMM_UNROLL_M
  1689. export XGEMM_UNROLL_N
  1690. export CGEMM3M_UNROLL_M
  1691. export CGEMM3M_UNROLL_N
  1692. export ZGEMM3M_UNROLL_M
  1693. export ZGEMM3M_UNROLL_N
  1694. export XGEMM3M_UNROLL_M
  1695. export XGEMM3M_UNROLL_N
  1696. ifdef USE_CUDA
  1697. export CUDADIR
  1698. export CUCC
  1699. export CUFLAGS
  1700. export CULIB
  1701. endif
  1702. .SUFFIXES: .$(PSUFFIX) .$(SUFFIX) .f
  1703. .f.$(SUFFIX):
  1704. $(FC) $(FFLAGS) -c $< -o $(@F)
  1705. .f.$(PSUFFIX):
  1706. $(FC) $(FPFLAGS) -pg -c $< -o $(@F)
  1707. ifdef BINARY64
  1708. PATHSCALEPATH = /opt/pathscale/lib/3.1
  1709. PGIPATH = /opt/pgi/linux86-64/7.1-5/lib
  1710. else
  1711. PATHSCALEPATH = /opt/pathscale/lib/3.1/32
  1712. PGIPATH = /opt/pgi/linux86/7.1-5/lib
  1713. endif
  1714. ACMLPATH = /opt/acml/4.3.0
  1715. ifneq ($(OSNAME), Darwin)
  1716. MKLPATH = /opt/intel/mkl/10.2.2.025/lib
  1717. else
  1718. MKLPATH = /Library/Frameworks/Intel_MKL.framework/Versions/10.0.1.014/lib
  1719. endif
  1720. ATLASPATH = /opt/atlas/3.9.17/opteron
  1721. FLAMEPATH = $(HOME)/flame/lib
  1722. ifneq ($(OSNAME), SunOS)
  1723. SUNPATH = /opt/sunstudio12.1
  1724. else
  1725. SUNPATH = /opt/SUNWspro
  1726. endif