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 34 kB

7 years ago
13 years ago
1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285128612871288128912901291129212931294129512961297129812991300130113021303130413051306130713081309131013111312131313141315131613171318131913201321132213231324132513261327132813291330133113321333133413351336133713381339134013411342134313441345134613471348134913501351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384138513861387138813891390139113921393139413951396139713981399140014011402140314041405140614071408140914101411141214131414141514161417141814191420142114221423142414251426142714281429143014311432143314341435143614371438143914401441144214431444144514461447144814491450145114521453145414551456145714581459146014611462146314641465146614671468146914701471147214731474147514761477147814791480148114821483148414851486148714881489149014911492149314941495149614971498149915001501150215031504150515061507150815091510151115121513151415151516151715181519152015211522152315241525152615271528152915301531153215331534153515361537153815391540154115421543154415451546154715481549155015511552155315541555155615571558155915601561156215631564156515661567156815691570157115721573157415751576157715781579158015811582158315841585158615871588158915901591159215931594159515961597159815991600160116021603160416051606160716081609161016111612161316141615161616171618161916201621162216231624162516261627162816291630163116321633163416351636163716381639164016411642164316441645164616471648164916501651165216531654165516561657165816591660166116621663166416651666166716681669167016711672167316741675167616771678167916801681168216831684168516861687
  1. #
  2. # Include user definition
  3. #
  4. # TO suppress recursive includes
  5. INCLUDED = 1
  6. ifndef TOPDIR
  7. TOPDIR = .
  8. endif
  9. # If ARCH is not set, we use the host system's architecture for getarch compile options.
  10. ifndef ARCH
  11. HOSTARCH := $(shell uname -m)
  12. else
  13. HOSTARCH = $(ARCH)
  14. endif
  15. # Catch conflicting usage of ARCH in some BSD environments
  16. ifeq ($(ARCH), amd64)
  17. override ARCH=x86_64
  18. else ifeq ($(ARCH), powerpc64)
  19. override ARCH=power
  20. else ifeq ($(ARCH), powerpc64le)
  21. override ARCH=power
  22. else ifeq ($(ARCH), powerpc)
  23. override ARCH=power
  24. else ifeq ($(ARCH), i386)
  25. override ARCH=x86
  26. else ifeq ($(ARCH), armv6)
  27. override ARCH=arm
  28. else ifeq ($(ARCH), armv7)
  29. override ARCH=arm
  30. else ifeq ($(ARCH), aarch64)
  31. override ARCH=arm64
  32. else ifeq ($(ARCH), zarch)
  33. override ARCH=zarch
  34. endif
  35. NETLIB_LAPACK_DIR = $(TOPDIR)/lapack-netlib
  36. # Default C compiler
  37. # - Only set if not specified on the command line or inherited from the environment.
  38. # - CC is an implicit variable so neither '?=' or 'ifndef' can be used.
  39. # http://stackoverflow.com/questions/4029274/mingw-and-make-variables
  40. # - Default value is 'cc' which is not always a valid command (e.g. MinGW).
  41. ifeq ($(origin CC),default)
  42. # Check if $(CC) refers to a valid command and set the value to gcc if not
  43. ifneq ($(findstring cmd.exe,$(SHELL)),)
  44. ifeq ($(shell where $(CC) 2>NUL),)
  45. CC = gcc
  46. endif
  47. else # POSIX-ish
  48. ifeq ($(shell command -v $(CC) 2>/dev/null),)
  49. ifeq ($(shell uname -s),Darwin)
  50. CC = clang
  51. # EXTRALIB += -Wl,-no_compact_unwind
  52. else
  53. CC = gcc
  54. endif # Darwin
  55. endif # CC exists
  56. endif # Shell is sane
  57. endif # CC is set to default
  58. # Default Fortran compiler (FC) is selected by f_check.
  59. ifndef MAKEFILE_RULE
  60. include $(TOPDIR)/Makefile.rule
  61. else
  62. include $(TOPDIR)/$(MAKEFILE_RULE)
  63. endif
  64. #
  65. # Beginning of system configuration
  66. #
  67. ifneq ($(BUILD_SINGLE),1)
  68. ifneq ($(BUILD_DOUBLE),1)
  69. ifneq ($(BUILD_COMPLEX),1)
  70. ifneq ($(BUILD_COMPLEX16),1)
  71. override BUILD_SINGLE=1
  72. override BUILD_DOUBLE=1
  73. override BUILD_COMPLEX=1
  74. override BUILD_COMPLEX16=1
  75. endif
  76. endif
  77. endif
  78. endif
  79. ifndef HOSTCC
  80. HOSTCC = $(CC)
  81. endif
  82. ifdef TARGET
  83. GETARCH_FLAGS := -DFORCE_$(TARGET)
  84. GETARCH_FLAGS += -DUSER_TARGET
  85. ifeq ($(TARGET), GENERIC)
  86. ifeq ($(DYNAMIC_ARCH), 1)
  87. override NO_EXPRECISION=1
  88. export NO_EXPRECiSION
  89. endif
  90. endif
  91. endif
  92. # Force fallbacks for 32bit
  93. ifeq ($(BINARY), 32)
  94. ifeq ($(TARGET), HASWELL)
  95. GETARCH_FLAGS := -DFORCE_NEHALEM
  96. endif
  97. ifeq ($(TARGET), SKYLAKEX)
  98. GETARCH_FLAGS := -DFORCE_NEHALEM
  99. endif
  100. ifeq ($(TARGET), COOPERLAKE)
  101. GETARCH_FLAGS := -DFORCE_NEHALEM
  102. endif
  103. ifeq ($(TARGET), SANDYBRIDGE)
  104. GETARCH_FLAGS := -DFORCE_NEHALEM
  105. endif
  106. ifeq ($(TARGET), BULLDOZER)
  107. GETARCH_FLAGS := -DFORCE_BARCELONA
  108. endif
  109. ifeq ($(TARGET), PILEDRIVER)
  110. GETARCH_FLAGS := -DFORCE_BARCELONA
  111. endif
  112. ifeq ($(TARGET), STEAMROLLER)
  113. GETARCH_FLAGS := -DFORCE_BARCELONA
  114. endif
  115. ifeq ($(TARGET), EXCAVATOR)
  116. GETARCH_FLAGS := -DFORCE_BARCELONA
  117. endif
  118. ifeq ($(TARGET), ZEN)
  119. GETARCH_FLAGS := -DFORCE_BARCELONA
  120. endif
  121. ifeq ($(TARGET), ARMV8)
  122. GETARCH_FLAGS := -DFORCE_ARMV7
  123. endif
  124. ifeq ($(TARGET), POWER8)
  125. GETARCH_FLAGS := -DFORCE_POWER6
  126. endif
  127. endif
  128. #TARGET_CORE will override TARGET which is used in DYNAMIC_ARCH=1.
  129. #
  130. ifdef TARGET_CORE
  131. GETARCH_FLAGS := -DFORCE_$(TARGET_CORE)
  132. endif
  133. # Force fallbacks for 32bit
  134. ifeq ($(BINARY), 32)
  135. ifeq ($(TARGET_CORE), HASWELL)
  136. GETARCH_FLAGS := -DFORCE_NEHALEM
  137. endif
  138. ifeq ($(TARGET_CORE), SKYLAKEX)
  139. GETARCH_FLAGS := -DFORCE_NEHALEM
  140. endif
  141. ifeq ($(TARGET_CORE), COOPERLAKE)
  142. GETARCH_FLAGS := -DFORCE_NEHALEM
  143. endif
  144. ifeq ($(TARGET_CORE), SANDYBRIDGE)
  145. GETARCH_FLAGS := -DFORCE_NEHALEM
  146. endif
  147. ifeq ($(TARGET_CORE), BULLDOZER)
  148. GETARCH_FLAGS := -DFORCE_BARCELONA
  149. endif
  150. ifeq ($(TARGET_CORE), PILEDRIVER)
  151. GETARCH_FLAGS := -DFORCE_BARCELONA
  152. endif
  153. ifeq ($(TARGET_CORE), STEAMROLLER)
  154. GETARCH_FLAGS := -DFORCE_BARCELONA
  155. endif
  156. ifeq ($(TARGET_CORE), EXCAVATOR)
  157. GETARCH_FLAGS := -DFORCE_BARCELONA
  158. endif
  159. ifeq ($(TARGET_CORE), ZEN)
  160. GETARCH_FLAGS := -DFORCE_BARCELONA
  161. endif
  162. endif
  163. # On x86_64 build getarch with march=native unless the compiler is PGI. This is required to detect AVX512 support in getarch.
  164. ifeq ($(HOSTARCH), x86_64)
  165. ifeq ($(findstring pgcc,$(HOSTCC))$(findstring nvc,$(HOSTCC)),)
  166. GETARCH_FLAGS += -march=native
  167. endif
  168. endif
  169. ifdef INTERFACE64
  170. ifneq ($(INTERFACE64), 0)
  171. GETARCH_FLAGS += -DUSE64BITINT
  172. endif
  173. endif
  174. ifndef GEMM_MULTITHREAD_THRESHOLD
  175. GEMM_MULTITHREAD_THRESHOLD=4
  176. endif
  177. GETARCH_FLAGS += -DGEMM_MULTITHREAD_THRESHOLD=$(GEMM_MULTITHREAD_THRESHOLD)
  178. ifeq ($(NO_AVX), 1)
  179. GETARCH_FLAGS += -DNO_AVX
  180. endif
  181. ifeq ($(BINARY), 32)
  182. GETARCH_FLAGS += -DNO_AVX -DNO_AVX2 -DNO_AVX512
  183. NO_AVX512 = 1
  184. endif
  185. ifeq ($(NO_AVX2), 1)
  186. GETARCH_FLAGS += -DNO_AVX2
  187. endif
  188. ifeq ($(NO_AVX512), 1)
  189. GETARCH_FLAGS += -DNO_AVX512
  190. endif
  191. ifeq ($(DEBUG), 1)
  192. GETARCH_FLAGS += -g
  193. endif
  194. ifeq ($(QUIET_MAKE), 1)
  195. MAKE += -s
  196. endif
  197. ifndef NO_PARALLEL_MAKE
  198. NO_PARALLEL_MAKE=0
  199. endif
  200. GETARCH_FLAGS += -DNO_PARALLEL_MAKE=$(NO_PARALLEL_MAKE)
  201. ifdef MAKE_NB_JOBS
  202. GETARCH_FLAGS += -DMAKE_NB_JOBS=$(MAKE_NB_JOBS)
  203. endif
  204. ifeq ($(HOSTCC), loongcc)
  205. GETARCH_FLAGS += -static
  206. endif
  207. #if don't use Fortran, it will only compile CBLAS.
  208. ifeq ($(ONLY_CBLAS), 1)
  209. NO_LAPACK = 1
  210. else
  211. ONLY_CBLAS = 0
  212. endif
  213. # This operation is expensive, so execution should be once.
  214. ifndef GOTOBLAS_MAKEFILE
  215. export GOTOBLAS_MAKEFILE = 1
  216. # Generating Makefile.conf and config.h
  217. 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) TARGET_CORE=$(TARGET_CORE) ONLY_CBLAS=$(ONLY_CBLAS) TARGET=$(TARGET) all)
  218. ifndef TARGET_CORE
  219. include $(TOPDIR)/Makefile.conf
  220. else
  221. HAVE_NEON=
  222. HAVE_VFP=
  223. HAVE_VFPV3=
  224. HAVE_VFPV4=
  225. HAVE_MMX=
  226. HAVE_SSE=
  227. HAVE_SSE2=
  228. HAVE_SSE3=
  229. HAVE_SSSE3=
  230. HAVE_SSE4_1=
  231. HAVE_SSE4_2=
  232. HAVE_SSE4A=
  233. HAVE_SSE5=
  234. HAVE_AVX=
  235. HAVE_AVX2=
  236. HAVE_FMA3=
  237. include $(TOPDIR)/Makefile_kernel.conf
  238. endif
  239. endif
  240. ifndef NUM_PARALLEL
  241. NUM_PARALLEL = 1
  242. endif
  243. ifndef NUM_THREADS
  244. NUM_THREADS = $(NUM_CORES)
  245. endif
  246. ifeq ($(NUM_THREADS), 1)
  247. override USE_THREAD = 0
  248. override USE_OPENMP = 0
  249. endif
  250. ifdef USE_THREAD
  251. ifeq ($(USE_THREAD), 0)
  252. SMP =
  253. else
  254. SMP = 1
  255. endif
  256. else
  257. ifeq ($(NUM_THREAD), 1)
  258. SMP =
  259. else
  260. SMP = 1
  261. endif
  262. endif
  263. ifeq ($(SMP), 1)
  264. USE_LOCKING =
  265. endif
  266. ifndef NEED_PIC
  267. NEED_PIC = 1
  268. endif
  269. ARFLAGS =
  270. CPP = $(COMPILER) -E
  271. AR ?= $(CROSS_SUFFIX)ar
  272. AS ?= $(CROSS_SUFFIX)as
  273. LD ?= $(CROSS_SUFFIX)ld
  274. RANLIB ?= $(CROSS_SUFFIX)ranlib
  275. NM = $(CROSS_SUFFIX)nm
  276. DLLWRAP = $(CROSS_SUFFIX)dllwrap
  277. OBJCOPY = $(CROSS_SUFFIX)objcopy
  278. OBJCONV = $(CROSS_SUFFIX)objconv
  279. # When fortran support was either not detected or actively deselected, only build BLAS.
  280. ifeq ($(NOFORTRAN), 1)
  281. NO_LAPACK = 1
  282. override FEXTRALIB =
  283. endif
  284. ifeq ($(C_COMPILER), GCC)
  285. GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
  286. GCCVERSIONGT4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 4)
  287. GCCVERSIONGT5 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 5)
  288. GCCVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 7)
  289. GCCVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  290. GCCVERSIONGTEQ11 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 11)
  291. GCCVERSIONGTEQ10 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 10)
  292. # Note that the behavior of -dumpversion is compile-time-configurable for
  293. # gcc-7.x and newer. Use -dumpfullversion there
  294. ifeq ($(GCCVERSIONGTEQ7),1)
  295. GCCDUMPVERSION_PARAM := -dumpfullversion
  296. else
  297. GCCDUMPVERSION_PARAM := -dumpversion
  298. endif
  299. GCCMINORVERSIONGTEQ1 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 1)
  300. GCCMINORVERSIONGTEQ2 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 2)
  301. GCCMINORVERSIONGTEQ7 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 7)
  302. endif
  303. #
  304. # OS dependent settings
  305. #
  306. ifeq ($(OSNAME), Darwin)
  307. ifndef MACOSX_DEPLOYMENT_TARGET
  308. export MACOSX_DEPLOYMENT_TARGET=10.8
  309. endif
  310. MD5SUM = md5 -r
  311. endif
  312. ifneq (,$(findstring $(OSNAME), FreeBSD OpenBSD DragonFly))
  313. MD5SUM = md5 -r
  314. endif
  315. ifeq ($(OSNAME), NetBSD)
  316. MD5SUM = md5 -n
  317. endif
  318. ifeq ($(OSNAME), Linux)
  319. EXTRALIB += -lm
  320. NO_EXPRECISION = 1
  321. endif
  322. ifeq ($(OSNAME), Android)
  323. EXTRALIB += -lm
  324. endif
  325. ifeq ($(OSNAME), AIX)
  326. EXTRALIB += -lm
  327. endif
  328. ifeq ($(OSNAME), FreeBSD)
  329. ifeq ($(ARCH), $(filter ($ARCH),arm arm64))
  330. EXTRALIB += -lm
  331. endif
  332. endif
  333. ifeq ($(OSNAME), WINNT)
  334. NEED_PIC = 0
  335. NO_EXPRECISION = 1
  336. EXTRALIB += -defaultlib:advapi32
  337. SUFFIX = obj
  338. PSUFFIX = pobj
  339. LIBSUFFIX = a
  340. ifeq ($(C_COMPILER), CLANG)
  341. CCOMMON_OPT += -DMS_ABI
  342. endif
  343. #Version tests for supporting specific features (MS_ABI, POWER9 intrinsics)
  344. ifeq ($(GCCVERSIONGT4), 1)
  345. # GCC Major version > 4
  346. # It is compatible with MSVC ABI.
  347. CCOMMON_OPT += -DMS_ABI
  348. endif
  349. ifeq ($(GCCVERSIONGTEQ4), 1)
  350. ifeq ($(GCCMINORVERSIONGTEQ7), 1)
  351. # GCC Version >=4.7
  352. # It is compatible with MSVC ABI.
  353. CCOMMON_OPT += -DMS_ABI
  354. endif
  355. endif
  356. # Ensure the correct stack alignment on Win32
  357. # http://permalink.gmane.org/gmane.comp.lib.openblas.general/97
  358. ifeq ($(ARCH), x86)
  359. CCOMMON_OPT += -mincoming-stack-boundary=2
  360. FCOMMON_OPT += -mincoming-stack-boundary=2
  361. endif
  362. endif
  363. ifeq ($(OSNAME), Interix)
  364. NEED_PIC = 0
  365. NO_EXPRECISION = 1
  366. INTERIX_TOOL_DIR = /opt/gcc.3.3/i586-pc-interix3/bin
  367. endif
  368. ifeq ($(OSNAME), CYGWIN_NT)
  369. NEED_PIC = 0
  370. NO_EXPRECISION = 1
  371. OS_CYGWIN_NT = 1
  372. endif
  373. ifneq ($(OSNAME), WINNT)
  374. ifneq ($(OSNAME), CYGWIN_NT)
  375. ifneq ($(OSNAME), Interix)
  376. ifneq ($(OSNAME), Android)
  377. ifdef SMP
  378. EXTRALIB += -lpthread
  379. endif
  380. endif
  381. endif
  382. endif
  383. endif
  384. # ifeq logical or
  385. ifeq ($(OSNAME), $(filter $(OSNAME),WINNT CYGWIN_NT Interix))
  386. OS_WINDOWS=1
  387. endif
  388. ifdef QUAD_PRECISION
  389. CCOMMON_OPT += -DQUAD_PRECISION
  390. NO_EXPRECISION = 1
  391. endif
  392. ifneq ($(ARCH), x86)
  393. ifneq ($(ARCH), x86_64)
  394. NO_EXPRECISION = 1
  395. endif
  396. endif
  397. ifdef UTEST_CHECK
  398. CCOMMON_OPT += -DUTEST_CHECK
  399. SANITY_CHECK = 1
  400. endif
  401. ifdef SANITY_CHECK
  402. CCOMMON_OPT += -DSANITY_CHECK -DREFNAME=$(*F)f$(BU)
  403. endif
  404. MAX_STACK_ALLOC ?= 2048
  405. ifneq ($(MAX_STACK_ALLOC), 0)
  406. CCOMMON_OPT += -DMAX_STACK_ALLOC=$(MAX_STACK_ALLOC)
  407. endif
  408. ifdef USE_LOCKING
  409. ifneq ($(USE_LOCKING), 0)
  410. CCOMMON_OPT += -DUSE_LOCKING
  411. endif
  412. endif
  413. #
  414. # Architecture dependent settings
  415. #
  416. ifeq ($(ARCH), x86)
  417. ifndef BINARY
  418. NO_BINARY_MODE = 1
  419. endif
  420. ifeq ($(CORE), generic)
  421. NO_EXPRECISION = 1
  422. endif
  423. ifndef NO_EXPRECISION
  424. ifeq ($(F_COMPILER), GFORTRAN)
  425. # ifeq logical or. GCC or LSB
  426. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  427. EXPRECISION = 1
  428. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  429. FCOMMON_OPT += -m128bit-long-double
  430. endif
  431. ifeq ($(C_COMPILER), CLANG)
  432. EXPRECISION = 1
  433. CCOMMON_OPT += -DEXPRECISION
  434. FCOMMON_OPT += -m128bit-long-double
  435. endif
  436. endif
  437. endif
  438. endif
  439. ifeq ($(ARCH), x86_64)
  440. ifeq ($(CORE), generic)
  441. NO_EXPRECISION = 1
  442. endif
  443. ifndef NO_EXPRECISION
  444. ifeq ($(F_COMPILER), GFORTRAN)
  445. # ifeq logical or. GCC or LSB
  446. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  447. EXPRECISION = 1
  448. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  449. FCOMMON_OPT += -m128bit-long-double
  450. endif
  451. ifeq ($(C_COMPILER), CLANG)
  452. EXPRECISION = 1
  453. CCOMMON_OPT += -DEXPRECISION
  454. FCOMMON_OPT += -m128bit-long-double
  455. endif
  456. endif
  457. endif
  458. endif
  459. ifeq ($(C_COMPILER), INTEL)
  460. CCOMMON_OPT += -wd981
  461. endif
  462. ifeq ($(USE_OPENMP), 1)
  463. #check
  464. ifeq ($(USE_THREAD), 0)
  465. $(error OpenBLAS: Cannot set both USE_OPENMP=1 and USE_THREAD=0. The USE_THREAD=0 is only for building single thread version.)
  466. endif
  467. # ifeq logical or. GCC or LSB
  468. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  469. CCOMMON_OPT += -fopenmp
  470. endif
  471. ifeq ($(C_COMPILER), CLANG)
  472. CCOMMON_OPT += -fopenmp
  473. endif
  474. ifeq ($(C_COMPILER), INTEL)
  475. CCOMMON_OPT += -fopenmp
  476. endif
  477. ifeq ($(C_COMPILER), PGI)
  478. CCOMMON_OPT += -mp
  479. endif
  480. ifeq ($(C_COMPILER), OPEN64)
  481. CCOMMON_OPT += -mp
  482. CEXTRALIB += -lstdc++
  483. endif
  484. ifeq ($(C_COMPILER), PATHSCALE)
  485. CCOMMON_OPT += -mp
  486. endif
  487. endif
  488. ifeq ($(DYNAMIC_ARCH), 1)
  489. ifeq ($(ARCH), x86)
  490. DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
  491. CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
  492. endif
  493. ifeq ($(ARCH), x86_64)
  494. DYNAMIC_CORE = PRESCOTT CORE2
  495. ifeq ($(DYNAMIC_OLDER), 1)
  496. DYNAMIC_CORE += PENRYN DUNNINGTON
  497. endif
  498. DYNAMIC_CORE += NEHALEM
  499. ifeq ($(DYNAMIC_OLDER), 1)
  500. DYNAMIC_CORE += OPTERON OPTERON_SSE3
  501. endif
  502. DYNAMIC_CORE += BARCELONA
  503. ifeq ($(DYNAMIC_OLDER), 1)
  504. DYNAMIC_CORE += BOBCAT ATOM NANO
  505. endif
  506. ifneq ($(NO_AVX), 1)
  507. DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER STEAMROLLER EXCAVATOR
  508. endif
  509. ifneq ($(NO_AVX2), 1)
  510. DYNAMIC_CORE += HASWELL ZEN
  511. endif
  512. ifneq ($(NO_AVX512), 1)
  513. ifneq ($(NO_AVX2), 1)
  514. DYNAMIC_CORE += SKYLAKEX COOPERLAKE
  515. endif
  516. endif
  517. endif
  518. ifdef DYNAMIC_LIST
  519. override DYNAMIC_CORE = PRESCOTT $(DYNAMIC_LIST)
  520. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_PRESCOTT
  521. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  522. CCOMMON_OPT += $(XCCOMMON_OPT)
  523. #CCOMMON_OPT += -DDYNAMIC_LIST='$(DYNAMIC_LIST)'
  524. endif
  525. ifeq ($(ARCH), arm64)
  526. DYNAMIC_CORE = ARMV8
  527. DYNAMIC_CORE += CORTEXA53
  528. DYNAMIC_CORE += CORTEXA57
  529. DYNAMIC_CORE += CORTEXA72
  530. DYNAMIC_CORE += CORTEXA73
  531. DYNAMIC_CORE += NEOVERSEN1
  532. DYNAMIC_CORE += FALKOR
  533. DYNAMIC_CORE += THUNDERX
  534. DYNAMIC_CORE += THUNDERX2T99
  535. DYNAMIC_CORE += TSV110
  536. DYNAMIC_CORE += EMAG8180
  537. DYNAMIC_CORE += THUNDERX3T110
  538. ifdef DYNAMIC_LIST
  539. override DYNAMIC_CORE = ARMV8 $(DYNAMIC_LIST)
  540. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_ARMV8
  541. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  542. endif
  543. endif
  544. ifeq ($(ARCH), mips64)
  545. DYNAMIC_CORE = LOONGSON3R3 LOONGSON3R4
  546. endif
  547. ifeq ($(ARCH), zarch)
  548. DYNAMIC_CORE = ZARCH_GENERIC
  549. # if the compiler accepts -march=arch11 or -march=z13 and can compile a file
  550. # with z13-specific inline assembly, then we can include support for Z13.
  551. # note: -march=z13 is equivalent to -march=arch11 yet some compiler releases
  552. # only support one or the other.
  553. # note: LLVM version 6.x supported -march=z13 yet could not handle vector
  554. # registers in inline assembly, so the check for supporting the -march flag is
  555. # not enough.
  556. ZARCH_TEST_COMPILE=-c $(TOPDIR)/kernel/zarch/damin_z13.c -I$(TOPDIR) -o /dev/null > /dev/null 2> /dev/null
  557. ZARCH_CC_SUPPORTS_ARCH11=$(shell $(CC) -march=arch11 $(ZARCH_TEST_COMPILE) && echo 1)
  558. ZARCH_CC_SUPPORTS_Z13=$(shell $(CC) -march=z13 $(ZARCH_TEST_COMPILE) && echo 1)
  559. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH11), $(ZARCH_CC_SUPPORTS_Z13)), 1)
  560. DYNAMIC_CORE += Z13
  561. CCOMMON_OPT += -DDYN_Z13
  562. else
  563. $(info OpenBLAS: Not building Z13 kernels because the compiler $(CC) does not support it)
  564. endif
  565. # as above for z13, check for -march=arch12 and z14 support in the compiler.
  566. ZARCH_CC_SUPPORTS_ARCH12=$(shell $(CC) -march=arch12 $(ZARCH_TEST_COMPILE) && echo 1)
  567. ZARCH_CC_SUPPORTS_Z14=$(shell $(CC) -march=z14 $(ZARCH_TEST_COMPILE) && echo 1)
  568. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH12), $(ZARCH_CC_SUPPORTS_Z14)), 1)
  569. DYNAMIC_CORE += Z14
  570. CCOMMON_OPT += -DDYN_Z14
  571. else
  572. $(info OpenBLAS: Not building Z14 kernels because the compiler $(CC) does not support it)
  573. endif
  574. endif # ARCH zarch
  575. ifeq ($(ARCH), power)
  576. ifneq ($(C_COMPILER), PGI)
  577. DYNAMIC_CORE = POWER6
  578. DYNAMIC_CORE += POWER8
  579. ifneq ($(C_COMPILER), GCC)
  580. DYNAMIC_CORE += POWER9
  581. DYNAMIC_CORE += POWER10
  582. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  583. endif
  584. ifeq ($(C_COMPILER), GCC)
  585. ifeq ($(GCCVERSIONGT5), 1)
  586. DYNAMIC_CORE += POWER9
  587. else
  588. $(info, OpenBLAS: Your gcc version is too old to build the POWER9 kernels.)
  589. endif
  590. LDVERSIONGTEQ35 := $(shell expr `$(CC) -Wl,--version 2> /dev/null | head -1 | cut -f2 -d "." | cut -f1 -d "-"` \>= 35)
  591. ifeq ($(GCCVERSIONGTEQ11)$(LDVERSIONGTEQ35), 11)
  592. DYNAMIC_CORE += POWER10
  593. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  594. else ifeq ($(GCCVERSIONGTEQ10), 1)
  595. ifeq ($(GCCMINORVERSIONGTEQ2)$(LDVERSIONGTEQ35), 11)
  596. DYNAMIC_CORE += POWER10
  597. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  598. endif
  599. else
  600. $(info, OpenBLAS: Your gcc version is too old to build the POWER10 kernels.)
  601. endif
  602. endif
  603. else
  604. DYNAMIC_CORE = POWER8
  605. DYNAMIC_CORE += POWER9
  606. endif
  607. endif
  608. # If DYNAMIC_CORE is not set, DYNAMIC_ARCH cannot do anything, so force it to empty
  609. ifndef DYNAMIC_CORE
  610. override DYNAMIC_ARCH=
  611. endif
  612. endif
  613. ifeq ($(ARCH), ia64)
  614. NO_BINARY_MODE = 1
  615. BINARY_DEFINED = 1
  616. ifeq ($(F_COMPILER), GFORTRAN)
  617. ifeq ($(C_COMPILER), GCC)
  618. # EXPRECISION = 1
  619. # CCOMMON_OPT += -DEXPRECISION
  620. endif
  621. endif
  622. endif
  623. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  624. NO_BINARY_MODE = 1
  625. endif
  626. ifeq ($(ARCH), alpha)
  627. NO_BINARY_MODE = 1
  628. BINARY_DEFINED = 1
  629. endif
  630. ifeq ($(ARCH), arm)
  631. NO_BINARY_MODE = 1
  632. BINARY_DEFINED = 1
  633. CCOMMON_OPT += -marm
  634. FCOMMON_OPT += -marm
  635. # If softfp abi is mentioned on the command line, force it.
  636. ifeq ($(ARM_SOFTFP_ABI), 1)
  637. CCOMMON_OPT += -mfloat-abi=softfp
  638. FCOMMON_OPT += -mfloat-abi=softfp
  639. endif
  640. ifeq ($(OSNAME), Android)
  641. ifeq ($(ARM_SOFTFP_ABI), 1)
  642. EXTRALIB += -lm
  643. else
  644. EXTRALIB += -Wl,-lm_hard
  645. endif
  646. endif
  647. endif
  648. ifeq ($(ARCH), arm64)
  649. NO_BINARY_MODE = 1
  650. BINARY_DEFINED = 1
  651. ifdef INTERFACE64
  652. ifneq ($(INTERFACE64), 0)
  653. ifeq ($(F_COMPILER), GFORTRAN)
  654. FCOMMON_OPT += -fdefault-integer-8
  655. endif
  656. ifeq ($(F_COMPILER), FLANG)
  657. FCOMMON_OPT += -i8
  658. endif
  659. endif
  660. endif
  661. endif
  662. ifeq ($(ARCH), riscv64)
  663. NO_BINARY_MODE = 1
  664. BINARY_DEFINED = 1
  665. endif
  666. #
  667. # C Compiler dependent settings
  668. #
  669. # ifeq logical or. GCC or CLANG or LSB
  670. # http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
  671. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG LSB))
  672. CCOMMON_OPT += -Wall
  673. COMMON_PROF += -fno-inline
  674. NO_UNINITIALIZED_WARN = -Wno-uninitialized
  675. ifeq ($(QUIET_MAKE), 1)
  676. CCOMMON_OPT += $(NO_UNINITIALIZED_WARN) -Wno-unused
  677. endif
  678. ifdef NO_BINARY_MODE
  679. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  680. ifdef BINARY64
  681. CCOMMON_OPT += -mabi=64
  682. else
  683. CCOMMON_OPT += -mabi=n32
  684. endif
  685. BINARY_DEFINED = 1
  686. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  687. CCOMMON_OPT += -mabi=32
  688. BINARY_DEFINED = 1
  689. endif
  690. ifeq ($(CORE), $(filter $(CORE),LOONGSON3R3 LOONGSON3R4))
  691. CCOMMON_OPT += -march=loongson3a
  692. FCOMMON_OPT += -march=loongson3a
  693. endif
  694. ifeq ($(CORE), MIPS24K)
  695. CCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  696. FCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  697. endif
  698. ifeq ($(CORE), MIPS1004K)
  699. CCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  700. FCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  701. endif
  702. ifeq ($(CORE), P5600)
  703. CCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  704. FCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  705. endif
  706. ifeq ($(CORE), I6400)
  707. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  708. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  709. endif
  710. ifeq ($(CORE), P6600)
  711. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  712. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  713. endif
  714. ifeq ($(CORE), I6500)
  715. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  716. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  717. endif
  718. ifeq ($(OSNAME), AIX)
  719. BINARY_DEFINED = 1
  720. endif
  721. endif
  722. ifndef BINARY_DEFINED
  723. ifneq ($(OSNAME), AIX)
  724. ifdef BINARY64
  725. ifneq ($(ARCH), riscv64)
  726. CCOMMON_OPT += -m64
  727. endif
  728. else
  729. CCOMMON_OPT += -m32
  730. endif
  731. endif
  732. endif
  733. endif
  734. ifeq ($(C_COMPILER), PGI)
  735. PGCVERSIONGT20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \> 20)
  736. PGCVERSIONGTEQ20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \>= 20)
  737. PGCMINORVERSIONGE11 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -c 4-5` == 11)
  738. PGCVERSIONCHECK := $(PGCVERSIONGT20)$(PGCVERSIONEQ20)$(PGCMINORVERSIONGE11)
  739. ifeq ($(PGCVERSIONCHECK), $(filter $(PGCVERSIONCHECK), 110 111 011))
  740. NEWPGI := 1
  741. endif
  742. ifdef BINARY64
  743. ifeq ($(ARCH), x86_64)
  744. CCOMMON_OPT += -tp p7-64
  745. ifneq ($(NEWPGI),1)
  746. CCOMMON_OPT += -D__MMX__ -Mnollvm
  747. endif
  748. else
  749. ifeq ($(ARCH), power)
  750. ifeq ($(CORE), POWER8)
  751. CCOMMON_OPT += -tp pwr8
  752. endif
  753. ifeq ($(CORE), POWER9)
  754. CCOMMON_OPT += -tp pwr9
  755. endif
  756. endif
  757. endif
  758. else
  759. CCOMMON_OPT += -tp p7
  760. endif
  761. endif
  762. ifeq ($(C_COMPILER), PATHSCALE)
  763. ifdef BINARY64
  764. CCOMMON_OPT += -m64
  765. else
  766. CCOMMON_OPT += -m32
  767. endif
  768. endif
  769. #
  770. # Fortran Compiler dependent settings
  771. #
  772. ifeq ($(F_COMPILER), NAG)
  773. FCOMMON_OPT += -dcfuns -recursive -ieee=full -w=obs -thread_safe
  774. ifdef INTERFACE64
  775. ifneq ($(INTERFACE64), 0)
  776. FCOMMON_OPT += -i8
  777. endif
  778. endif
  779. ifeq ($(USE_OPENMP), 1)
  780. FCOMMON_OPT += -openmp
  781. endif
  782. endif
  783. ifeq ($(F_COMPILER), FLANG)
  784. CCOMMON_OPT += -DF_INTERFACE_FLANG
  785. FCOMMON_OPT += -Mrecursive -Kieee
  786. ifeq ($(OSNAME), Linux)
  787. ifeq ($(ARCH), x86_64)
  788. FLANG_VENDOR := $(shell $(FC) --version|head -1 |cut -f 1 -d " ")
  789. ifeq ($(FLANG_VENDOR), AMD)
  790. FCOMMON_OPT += -fno-unroll-loops
  791. endif
  792. endif
  793. endif
  794. ifdef BINARY64
  795. ifdef INTERFACE64
  796. ifneq ($(INTERFACE64), 0)
  797. FCOMMON_OPT += -i8
  798. endif
  799. endif
  800. FCOMMON_OPT += -Wall
  801. else
  802. FCOMMON_OPT += -Wall
  803. endif
  804. ifeq ($(USE_OPENMP), 1)
  805. FCOMMON_OPT += -fopenmp
  806. endif
  807. endif
  808. ifeq ($(F_COMPILER), G77)
  809. CCOMMON_OPT += -DF_INTERFACE_G77
  810. FCOMMON_OPT += -Wall
  811. ifndef NO_BINARY_MODE
  812. ifneq ($(OSNAME), AIX)
  813. ifdef BINARY64
  814. FCOMMON_OPT += -m64
  815. else
  816. FCOMMON_OPT += -m32
  817. endif
  818. endif
  819. endif
  820. endif
  821. ifeq ($(F_COMPILER), G95)
  822. CCOMMON_OPT += -DF_INTERFACE_G95
  823. FCOMMON_OPT += -Wall
  824. ifneq ($(OSNAME), AIX)
  825. ifndef NO_BINARY_MODE
  826. ifdef BINARY64
  827. FCOMMON_OPT += -m64
  828. else
  829. FCOMMON_OPT += -m32
  830. endif
  831. endif
  832. ifneq ($(NO_LAPACKE), 1)
  833. FCOMMON_OPT += -fno-second-underscore
  834. endif
  835. endif
  836. endif
  837. ifeq ($(F_COMPILER), GFORTRAN)
  838. CCOMMON_OPT += -DF_INTERFACE_GFORT
  839. FCOMMON_OPT += -Wall
  840. # make single-threaded LAPACK calls thread-safe #1847
  841. FCOMMON_OPT += -frecursive
  842. # work around ABI problem with passing single-character arguments
  843. FCOMMON_OPT += -fno-optimize-sibling-calls
  844. #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
  845. ifneq ($(NO_LAPACK), 1)
  846. EXTRALIB += -lgfortran
  847. endif
  848. ifdef NO_BINARY_MODE
  849. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  850. ifdef BINARY64
  851. FCOMMON_OPT += -mabi=64
  852. else
  853. FCOMMON_OPT += -mabi=n32
  854. endif
  855. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  856. FCOMMON_OPT += -mabi=32
  857. endif
  858. else
  859. ifdef BINARY64
  860. ifneq ($(OSNAME), AIX)
  861. ifneq ($(ARCH), riscv64)
  862. FCOMMON_OPT += -m64
  863. endif
  864. endif
  865. ifdef INTERFACE64
  866. ifneq ($(INTERFACE64), 0)
  867. FCOMMON_OPT += -fdefault-integer-8
  868. endif
  869. endif
  870. else
  871. ifneq ($(OSNAME), AIX)
  872. FCOMMON_OPT += -m32
  873. endif
  874. endif
  875. endif
  876. ifeq ($(USE_OPENMP), 1)
  877. FCOMMON_OPT += -fopenmp
  878. endif
  879. endif
  880. ifeq ($(F_COMPILER), INTEL)
  881. CCOMMON_OPT += -DF_INTERFACE_INTEL
  882. ifdef INTERFACE64
  883. ifneq ($(INTERFACE64), 0)
  884. FCOMMON_OPT += -i8
  885. endif
  886. endif
  887. FCOMMON_OPT += -recursive -fp-model strict -assume protect-parens
  888. ifeq ($(USE_OPENMP), 1)
  889. FCOMMON_OPT += -fopenmp
  890. endif
  891. endif
  892. ifeq ($(F_COMPILER), FUJITSU)
  893. CCOMMON_OPT += -DF_INTERFACE_FUJITSU
  894. ifeq ($(USE_OPENMP), 1)
  895. FCOMMON_OPT += -openmp
  896. endif
  897. endif
  898. ifeq ($(F_COMPILER), IBM)
  899. CCOMMON_OPT += -DF_INTERFACE_IBM
  900. # FCOMMON_OPT += -qarch=440
  901. ifdef BINARY64
  902. FCOMMON_OPT += -q64
  903. ifdef INTERFACE64
  904. ifneq ($(INTERFACE64), 0)
  905. FCOMMON_OPT += -qintsize=8
  906. endif
  907. endif
  908. else
  909. FCOMMON_OPT += -q32
  910. endif
  911. ifeq ($(USE_OPENMP), 1)
  912. FCOMMON_OPT += -openmp
  913. endif
  914. endif
  915. ifeq ($(F_COMPILER), PGI)
  916. CCOMMON_OPT += -DF_INTERFACE_PGI
  917. COMMON_PROF += -DPGICOMPILER
  918. ifdef BINARY64
  919. ifdef INTERFACE64
  920. ifneq ($(INTERFACE64), 0)
  921. FCOMMON_OPT += -i8
  922. endif
  923. endif
  924. ifeq ($(ARCH), x86_64)
  925. FCOMMON_OPT += -tp p7-64
  926. else
  927. ifeq ($(ARCH), power)
  928. ifeq ($(CORE), POWER6)
  929. $(warning NVIDIA HPC compilers do not support POWER6.)
  930. endif
  931. ifeq ($(CORE), POWER8)
  932. FCOMMON_OPT += -tp pwr8
  933. endif
  934. ifeq ($(CORE), POWER9)
  935. FCOMMON_OPT += -tp pwr9
  936. endif
  937. ifeq ($(CORE), POWER10)
  938. $(warning NVIDIA HPC compilers do not support POWER10.)
  939. endif
  940. endif
  941. endif
  942. else
  943. FCOMMON_OPT += -tp p7
  944. endif
  945. FCOMMON_OPT += -Mrecursive -Kieee
  946. ifeq ($(USE_OPENMP), 1)
  947. FCOMMON_OPT += -mp
  948. endif
  949. endif
  950. ifeq ($(F_COMPILER), PATHSCALE)
  951. CCOMMON_OPT += -DF_INTERFACE_PATHSCALE
  952. ifdef BINARY64
  953. ifdef INTERFACE64
  954. ifneq ($(INTERFACE64), 0)
  955. FCOMMON_OPT += -i8
  956. endif
  957. endif
  958. endif
  959. ifeq ($(USE_OPENMP), 1)
  960. FCOMMON_OPT += -mp
  961. endif
  962. endif
  963. ifeq ($(F_COMPILER), OPEN64)
  964. CCOMMON_OPT += -DF_INTERFACE_OPEN64
  965. ifdef BINARY64
  966. ifdef INTERFACE64
  967. ifneq ($(INTERFACE64), 0)
  968. FCOMMON_OPT += -i8
  969. endif
  970. endif
  971. endif
  972. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  973. ifndef BINARY64
  974. FCOMMON_OPT += -n32
  975. else
  976. FCOMMON_OPT += -n64
  977. endif
  978. ifeq ($(CORE), LOONGSON3R3)
  979. FCOMMON_OPT += -loongson3 -static
  980. endif
  981. ifeq ($(CORE), LOONGSON3R4)
  982. FCOMMON_OPT += -loongson3 -static
  983. endif
  984. else
  985. ifndef BINARY64
  986. FCOMMON_OPT += -m32
  987. else
  988. FCOMMON_OPT += -m64
  989. endif
  990. endif
  991. ifeq ($(USE_OPENMP), 1)
  992. FEXTRALIB += -lstdc++
  993. FCOMMON_OPT += -mp
  994. endif
  995. endif
  996. ifeq ($(C_COMPILER), OPEN64)
  997. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  998. ifndef BINARY64
  999. CCOMMON_OPT += -n32
  1000. else
  1001. CCOMMON_OPT += -n64
  1002. endif
  1003. ifeq ($(CORE), LOONGSON3R3)
  1004. CCOMMON_OPT += -loongson3 -static
  1005. endif
  1006. ifeq ($(CORE), LOONGSON3R4)
  1007. CCOMMON_OPT += -loongson3 -static
  1008. endif
  1009. else
  1010. ifndef BINARY64
  1011. CCOMMON_OPT += -m32
  1012. else
  1013. CCOMMON_OPT += -m64
  1014. endif
  1015. endif
  1016. endif
  1017. ifeq ($(C_COMPILER), SUN)
  1018. CCOMMON_OPT += -w
  1019. ifeq ($(ARCH), x86)
  1020. CCOMMON_OPT += -m32
  1021. else
  1022. ifdef BINARY64
  1023. CCOMMON_OPT += -m64
  1024. else
  1025. CCOMMON_OPT += -m32
  1026. endif
  1027. endif
  1028. endif
  1029. ifeq ($(F_COMPILER), SUN)
  1030. CCOMMON_OPT += -DF_INTERFACE_SUN
  1031. FCOMMON_OPT += -ftrap=%none -xrecursive
  1032. ifeq ($(ARCH), x86)
  1033. FCOMMON_OPT += -m32
  1034. else
  1035. ifdef BINARY64
  1036. FCOMMON_OPT += -m64
  1037. else
  1038. FCOMMON_OPT += -m32
  1039. endif
  1040. endif
  1041. ifeq ($(USE_OPENMP), 1)
  1042. FCOMMON_OPT += -xopenmp=parallel
  1043. endif
  1044. endif
  1045. ifeq ($(F_COMPILER), COMPAQ)
  1046. CCOMMON_OPT += -DF_INTERFACE_COMPAQ
  1047. ifeq ($(USE_OPENMP), 1)
  1048. FCOMMON_OPT += -openmp
  1049. endif
  1050. endif
  1051. ifdef BINARY64
  1052. ifdef INTERFACE64
  1053. ifneq ($(INTERFACE64), 0)
  1054. CCOMMON_OPT +=
  1055. #-DUSE64BITINT
  1056. endif
  1057. endif
  1058. endif
  1059. ifeq ($(NEED_PIC), 1)
  1060. ifeq ($(C_COMPILER), IBM)
  1061. CCOMMON_OPT += -qpic=large
  1062. else
  1063. CCOMMON_OPT += -fPIC
  1064. endif
  1065. ifeq ($(F_COMPILER), SUN)
  1066. FCOMMON_OPT += -pic
  1067. else ifeq ($(F_COMPILER), NAG)
  1068. FCOMMON_OPT += -PIC
  1069. else
  1070. FCOMMON_OPT += -fPIC
  1071. endif
  1072. endif
  1073. ifeq ($(DYNAMIC_ARCH), 1)
  1074. CCOMMON_OPT += -DDYNAMIC_ARCH
  1075. endif
  1076. ifeq ($(DYNAMIC_OLDER), 1)
  1077. CCOMMON_OPT += -DDYNAMIC_OLDER
  1078. endif
  1079. ifeq ($(NO_LAPACK), 1)
  1080. CCOMMON_OPT += -DNO_LAPACK
  1081. #Disable LAPACK C interface
  1082. NO_LAPACKE = 1
  1083. endif
  1084. ifeq ($(NO_LAPACKE), 1)
  1085. CCOMMON_OPT += -DNO_LAPACKE
  1086. endif
  1087. ifeq ($(NO_AVX), 1)
  1088. CCOMMON_OPT += -DNO_AVX
  1089. endif
  1090. ifeq ($(ARCH), x86)
  1091. CCOMMON_OPT += -DNO_AVX
  1092. endif
  1093. ifeq ($(NO_AVX2), 1)
  1094. CCOMMON_OPT += -DNO_AVX2
  1095. endif
  1096. ifeq ($(NO_AVX512), 1)
  1097. CCOMMON_OPT += -DNO_AVX512
  1098. endif
  1099. ifdef SMP
  1100. CCOMMON_OPT += -DSMP_SERVER
  1101. ifeq ($(ARCH), mips64)
  1102. USE_SIMPLE_THREADED_LEVEL3 = 1
  1103. endif
  1104. ifeq ($(USE_OPENMP), 1)
  1105. # USE_SIMPLE_THREADED_LEVEL3 = 1
  1106. # NO_AFFINITY = 1
  1107. CCOMMON_OPT += -DUSE_OPENMP
  1108. endif
  1109. ifeq ($(BIGNUMA), 1)
  1110. CCOMMON_OPT += -DBIGNUMA
  1111. endif
  1112. endif
  1113. ifeq ($(NO_WARMUP), 1)
  1114. CCOMMON_OPT += -DNO_WARMUP
  1115. endif
  1116. ifeq ($(CONSISTENT_FPCSR), 1)
  1117. CCOMMON_OPT += -DCONSISTENT_FPCSR
  1118. endif
  1119. # Only for development
  1120. # CCOMMON_OPT += -DPARAMTEST
  1121. # CCOMMON_OPT += -DPREFETCHTEST
  1122. # CCOMMON_OPT += -DNO_SWITCHING
  1123. # USE_PAPI = 1
  1124. ifdef USE_PAPI
  1125. CCOMMON_OPT += -DUSE_PAPI
  1126. EXTRALIB += -lpapi -lperfctr
  1127. endif
  1128. ifdef BUFFERSIZE
  1129. CCOMMON_OPT += -DBUFFERSIZE=$(BUFFERSIZE)
  1130. endif
  1131. ifdef DYNAMIC_THREADS
  1132. CCOMMON_OPT += -DDYNAMIC_THREADS
  1133. endif
  1134. CCOMMON_OPT += -DMAX_CPU_NUMBER=$(NUM_THREADS)
  1135. CCOMMON_OPT += -DMAX_PARALLEL_NUMBER=$(NUM_PARALLEL)
  1136. ifdef USE_SIMPLE_THREADED_LEVEL3
  1137. CCOMMON_OPT += -DUSE_SIMPLE_THREADED_LEVEL3
  1138. endif
  1139. ifeq ($(USE_TLS), 1)
  1140. CCOMMON_OPT += -DUSE_TLS
  1141. endif
  1142. ifeq ($(BUILD_BFLOAT16), 1)
  1143. CCOMMON_OPT += -DBUILD_BFLOAT16
  1144. endif
  1145. ifeq ($(BUILD_SINGLE), 1)
  1146. CCOMMON_OPT += -DBUILD_SINGLE=1
  1147. endif
  1148. ifeq ($(BUILD_DOUBLE), 1)
  1149. CCOMMON_OPT += -DBUILD_DOUBLE=1
  1150. endif
  1151. ifeq ($(BUILD_COMPLEX), 1)
  1152. CCOMMON_OPT += -DBUILD_COMPLEX=1
  1153. endif
  1154. ifeq ($(BUILD_COMPLEX16), 1)
  1155. CCOMMON_OPT += -DBUILD_COMPLEX16=1
  1156. endif
  1157. CCOMMON_OPT += -DVERSION=\"$(VERSION)\"
  1158. ifndef SYMBOLPREFIX
  1159. SYMBOLPREFIX =
  1160. endif
  1161. ifndef SYMBOLSUFFIX
  1162. SYMBOLSUFFIX =
  1163. endif
  1164. ifndef LIBSONAMEBASE
  1165. LIBSONAMEBASE = openblas
  1166. endif
  1167. ifndef LIBNAMESUFFIX
  1168. LIBNAMEBASE = $(SYMBOLPREFIX)$(LIBSONAMEBASE)$(SYMBOLSUFFIX)
  1169. else
  1170. LIBNAMEBASE = $(SYMBOLPREFIX)$(LIBSONAMEBASE)$(SYMBOLSUFFIX)_$(LIBNAMESUFFIX)
  1171. endif
  1172. ifeq ($(OSNAME), CYGWIN_NT)
  1173. LIBPREFIX = cyg$(LIBNAMEBASE)
  1174. else
  1175. LIBPREFIX = lib$(LIBNAMEBASE)
  1176. endif
  1177. KERNELDIR = $(TOPDIR)/kernel/$(ARCH)
  1178. include $(TOPDIR)/Makefile.$(ARCH)
  1179. ifneq ($(C_COMPILER), PGI)
  1180. ifneq ($(C_COMPILER), SUN)
  1181. CCOMMON_OPT += -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME
  1182. endif
  1183. endif
  1184. CCOMMON_OPT += -DASMNAME=$(FU)$(*F) -DASMFNAME=$(FU)$(*F)$(BU) -DNAME=$(*F)$(BU) -DCNAME=$(*F) -DCHAR_NAME=\"$(*F)$(BU)\" -DCHAR_CNAME=\"$(*F)\"
  1185. ifeq ($(CORE), PPC440)
  1186. CCOMMON_OPT += -DALLOC_QALLOC
  1187. endif
  1188. ifeq ($(CORE), PPC440FP2)
  1189. STATIC_ALLOCATION = 1
  1190. endif
  1191. ifneq ($(OSNAME), Linux)
  1192. NO_AFFINITY = 1
  1193. endif
  1194. ifneq ($(ARCH), x86_64)
  1195. ifneq ($(ARCH), x86)
  1196. NO_AFFINITY = 1
  1197. endif
  1198. endif
  1199. ifdef NO_AFFINITY
  1200. ifeq ($(NO_AFFINITY), 0)
  1201. override undefine NO_AFFINITY
  1202. else
  1203. CCOMMON_OPT += -DNO_AFFINITY
  1204. endif
  1205. endif
  1206. ifdef FUNCTION_PROFILE
  1207. CCOMMON_OPT += -DFUNCTION_PROFILE
  1208. endif
  1209. ifdef HUGETLB_ALLOCATION
  1210. CCOMMON_OPT += -DALLOC_HUGETLB
  1211. endif
  1212. ifdef HUGETLBFILE_ALLOCATION
  1213. CCOMMON_OPT += -DALLOC_HUGETLBFILE -DHUGETLB_FILE_NAME=$(HUGETLBFILE_ALLOCATION)
  1214. endif
  1215. ifdef STATIC_ALLOCATION
  1216. CCOMMON_OPT += -DALLOC_STATIC
  1217. endif
  1218. ifdef DEVICEDRIVER_ALLOCATION
  1219. CCOMMON_OPT += -DALLOC_DEVICEDRIVER -DDEVICEDRIVER_NAME=\"/dev/mapper\"
  1220. endif
  1221. ifdef MIXED_MEMORY_ALLOCATION
  1222. CCOMMON_OPT += -DMIXED_MEMORY_ALLOCATION
  1223. endif
  1224. ifeq ($(OSNAME), SunOS)
  1225. TAR = gtar
  1226. PATCH = gpatch
  1227. GREP = ggrep
  1228. AWK = nawk
  1229. else
  1230. TAR = tar
  1231. PATCH = patch
  1232. GREP = grep
  1233. AWK = awk
  1234. endif
  1235. ifndef MD5SUM
  1236. MD5SUM = md5sum
  1237. endif
  1238. REVISION = -r$(VERSION)
  1239. MAJOR_VERSION = $(word 1,$(subst ., ,$(VERSION)))
  1240. ifeq ($(DEBUG), 1)
  1241. COMMON_OPT += -g
  1242. endif
  1243. ifeq ($(DEBUG), 1)
  1244. FCOMMON_OPT += -g
  1245. endif
  1246. ifndef COMMON_OPT
  1247. COMMON_OPT = -O2
  1248. endif
  1249. ifndef FCOMMON_OPT
  1250. FCOMMON_OPT = -O2 -frecursive
  1251. endif
  1252. override CFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
  1253. override PFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
  1254. override FFLAGS += $(COMMON_OPT) $(FCOMMON_OPT)
  1255. override FPFLAGS += $(FCOMMON_OPT) $(COMMON_PROF)
  1256. #MAKEOVERRIDES =
  1257. ifdef NEED_PIC
  1258. ifeq (,$(findstring PIC,$(FFLAGS)))
  1259. override FFLAGS += -fPIC
  1260. endif
  1261. endif
  1262. #For LAPACK Fortran codes.
  1263. #Disable -fopenmp for LAPACK Fortran codes on Windows.
  1264. ifdef OS_WINDOWS
  1265. LAPACK_FFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FFLAGS))
  1266. LAPACK_FPFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FPFLAGS))
  1267. else
  1268. LAPACK_FFLAGS := $(FFLAGS)
  1269. LAPACK_FPFLAGS := $(FPFLAGS)
  1270. endif
  1271. ifeq ($(F_COMPILER),NAG)
  1272. LAPACK_FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1273. endif
  1274. LAPACK_CFLAGS = $(CFLAGS)
  1275. LAPACK_CFLAGS += -DHAVE_LAPACK_CONFIG_H
  1276. ifdef INTERFACE64
  1277. ifneq ($(INTERFACE64), 0)
  1278. LAPACK_CFLAGS += -DLAPACK_ILP64
  1279. endif
  1280. endif
  1281. ifdef OS_WINDOWS
  1282. LAPACK_CFLAGS += -DOPENBLAS_OS_WINDOWS
  1283. endif
  1284. ifeq ($(C_COMPILER), LSB)
  1285. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1286. endif
  1287. ifndef SUFFIX
  1288. SUFFIX = o
  1289. endif
  1290. ifndef PSUFFIX
  1291. PSUFFIX = po
  1292. endif
  1293. ifndef LIBSUFFIX
  1294. LIBSUFFIX = a
  1295. endif
  1296. ifneq ($(DYNAMIC_ARCH), 1)
  1297. ifndef SMP
  1298. LIBNAME = $(LIBPREFIX)_$(LIBCORE)$(REVISION).$(LIBSUFFIX)
  1299. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)$(REVISION)_p.$(LIBSUFFIX)
  1300. else
  1301. LIBNAME = $(LIBPREFIX)_$(LIBCORE)p$(REVISION).$(LIBSUFFIX)
  1302. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)p$(REVISION)_p.$(LIBSUFFIX)
  1303. endif
  1304. else
  1305. ifndef SMP
  1306. LIBNAME = $(LIBPREFIX)$(REVISION).$(LIBSUFFIX)
  1307. LIBNAME_P = $(LIBPREFIX)$(REVISION)_p.$(LIBSUFFIX)
  1308. else
  1309. LIBNAME = $(LIBPREFIX)p$(REVISION).$(LIBSUFFIX)
  1310. LIBNAME_P = $(LIBPREFIX)p$(REVISION)_p.$(LIBSUFFIX)
  1311. endif
  1312. endif
  1313. LIBDLLNAME = $(LIBPREFIX).dll
  1314. IMPLIBNAME = lib$(LIBNAMEBASE).dll.a
  1315. ifneq ($(OSNAME), AIX)
  1316. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.so)
  1317. else
  1318. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.a)
  1319. endif
  1320. LIBDYNNAME = $(LIBNAME:.$(LIBSUFFIX)=.dylib)
  1321. LIBDEFNAME = $(LIBNAME:.$(LIBSUFFIX)=.def)
  1322. LIBEXPNAME = $(LIBNAME:.$(LIBSUFFIX)=.exp)
  1323. LIBZIPNAME = $(LIBNAME:.$(LIBSUFFIX)=.zip)
  1324. LIBS = $(TOPDIR)/$(LIBNAME)
  1325. LIBS_P = $(TOPDIR)/$(LIBNAME_P)
  1326. LIB_COMPONENTS = BLAS
  1327. ifneq ($(NO_CBLAS), 1)
  1328. LIB_COMPONENTS += CBLAS
  1329. endif
  1330. ifneq ($(NO_LAPACK), 1)
  1331. LIB_COMPONENTS += LAPACK
  1332. ifneq ($(NO_LAPACKE), 1)
  1333. LIB_COMPONENTS += LAPACKE
  1334. endif
  1335. ifeq ($(BUILD_RELAPACK), 1)
  1336. LIB_COMPONENTS += ReLAPACK
  1337. endif
  1338. endif
  1339. ifeq ($(ONLY_CBLAS), 1)
  1340. LIB_COMPONENTS = CBLAS
  1341. endif
  1342. export OSNAME
  1343. export ARCH
  1344. export CORE
  1345. export LIBCORE
  1346. export __BYTE_ORDER__
  1347. export ELF_VERSION
  1348. export PGCPATH
  1349. export CONFIG
  1350. export CC
  1351. export FC
  1352. export BU
  1353. export FU
  1354. export NEED2UNDERSCORES
  1355. export USE_THREAD
  1356. export NUM_THREADS
  1357. export NUM_CORES
  1358. export SMP
  1359. export MAKEFILE_RULE
  1360. export NEED_PIC
  1361. export BINARY
  1362. export BINARY32
  1363. export BINARY64
  1364. export F_COMPILER
  1365. export C_COMPILER
  1366. export USE_OPENMP
  1367. export CROSS
  1368. export CROSS_SUFFIX
  1369. export NOFORTRAN
  1370. export NO_FBLAS
  1371. export EXTRALIB
  1372. export CEXTRALIB
  1373. export FEXTRALIB
  1374. export HAVE_SSE
  1375. export HAVE_SSE2
  1376. export HAVE_SSE3
  1377. export HAVE_SSSE3
  1378. export HAVE_SSE4_1
  1379. export HAVE_SSE4_2
  1380. export HAVE_SSE4A
  1381. export HAVE_SSE5
  1382. export HAVE_AVX
  1383. export HAVE_AVX2
  1384. export HAVE_FMA3
  1385. export HAVE_VFP
  1386. export HAVE_VFPV3
  1387. export HAVE_VFPV4
  1388. export HAVE_NEON
  1389. export HAVE_MSA
  1390. export MSA_FLAGS
  1391. export KERNELDIR
  1392. export FUNCTION_PROFILE
  1393. export TARGET_CORE
  1394. export NO_AVX512
  1395. export NO_AVX2
  1396. export BUILD_BFLOAT16
  1397. export SBGEMM_UNROLL_M
  1398. export SBGEMM_UNROLL_N
  1399. export SGEMM_UNROLL_M
  1400. export SGEMM_UNROLL_N
  1401. export DGEMM_UNROLL_M
  1402. export DGEMM_UNROLL_N
  1403. export QGEMM_UNROLL_M
  1404. export QGEMM_UNROLL_N
  1405. export CGEMM_UNROLL_M
  1406. export CGEMM_UNROLL_N
  1407. export ZGEMM_UNROLL_M
  1408. export ZGEMM_UNROLL_N
  1409. export XGEMM_UNROLL_M
  1410. export XGEMM_UNROLL_N
  1411. export CGEMM3M_UNROLL_M
  1412. export CGEMM3M_UNROLL_N
  1413. export ZGEMM3M_UNROLL_M
  1414. export ZGEMM3M_UNROLL_N
  1415. export XGEMM3M_UNROLL_M
  1416. export XGEMM3M_UNROLL_N
  1417. ifdef USE_CUDA
  1418. export CUDADIR
  1419. export CUCC
  1420. export CUFLAGS
  1421. export CULIB
  1422. endif
  1423. .SUFFIXES: .$(PSUFFIX) .$(SUFFIX) .f
  1424. .f.$(SUFFIX):
  1425. $(FC) $(FFLAGS) -c $< -o $(@F)
  1426. .f.$(PSUFFIX):
  1427. $(FC) $(FPFLAGS) -pg -c $< -o $(@F)
  1428. ifdef BINARY64
  1429. PATHSCALEPATH = /opt/pathscale/lib/3.1
  1430. PGIPATH = /opt/pgi/linux86-64/7.1-5/lib
  1431. else
  1432. PATHSCALEPATH = /opt/pathscale/lib/3.1/32
  1433. PGIPATH = /opt/pgi/linux86/7.1-5/lib
  1434. endif
  1435. ACMLPATH = /opt/acml/4.3.0
  1436. ifneq ($(OSNAME), Darwin)
  1437. MKLPATH = /opt/intel/mkl/10.2.2.025/lib
  1438. else
  1439. MKLPATH = /Library/Frameworks/Intel_MKL.framework/Versions/10.0.1.014/lib
  1440. endif
  1441. ATLASPATH = /opt/atlas/3.9.17/opteron
  1442. FLAMEPATH = $(HOME)/flame/lib
  1443. ifneq ($(OSNAME), SunOS)
  1444. SUNPATH = /opt/sunstudio12.1
  1445. else
  1446. SUNPATH = /opt/SUNWspro
  1447. endif