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 30 kB

7 years ago
13 years ago
1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285128612871288128912901291129212931294129512961297129812991300130113021303130413051306130713081309131013111312131313141315131613171318131913201321132213231324132513261327132813291330133113321333133413351336133713381339134013411342134313441345134613471348134913501351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384138513861387138813891390139113921393139413951396139713981399140014011402140314041405140614071408140914101411141214131414141514161417141814191420142114221423142414251426142714281429143014311432143314341435143614371438143914401441144214431444144514461447144814491450145114521453145414551456145714581459146014611462146314641465146614671468146914701471147214731474147514761477147814791480148114821483148414851486148714881489149014911492149314941495149614971498149915001501150215031504150515061507150815091510151115121513151415151516151715181519152015211522152315241525152615271528152915301531153215331534153515361537153815391540154115421543154415451546154715481549
  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), powerpc)
  21. override ARCH=power
  22. else ifeq ($(ARCH), i386)
  23. override ARCH=x86
  24. else ifeq ($(ARCH), aarch64)
  25. override ARCH=arm64
  26. else ifeq ($(ARCH), zarch)
  27. override ARCH=zarch
  28. endif
  29. NETLIB_LAPACK_DIR = $(TOPDIR)/lapack-netlib
  30. # Default C compiler
  31. # - Only set if not specified on the command line or inherited from the environment.
  32. # - CC is an implicit variable so neither '?=' or 'ifndef' can be used.
  33. # http://stackoverflow.com/questions/4029274/mingw-and-make-variables
  34. # - Default value is 'cc' which is not always a valid command (e.g. MinGW).
  35. ifeq ($(origin CC),default)
  36. # Check if $(CC) refers to a valid command and set the value to gcc if not
  37. ifneq ($(findstring cmd.exe,$(SHELL)),)
  38. ifeq ($(shell where $(CC) 2>NUL),)
  39. CC = gcc
  40. endif
  41. else # POSIX-ish
  42. ifeq ($(shell command -v $(CC) 2>/dev/null),)
  43. ifeq ($(shell uname -s),Darwin)
  44. CC = clang
  45. # EXTRALIB += -Wl,-no_compact_unwind
  46. else
  47. CC = gcc
  48. endif # Darwin
  49. endif # CC exists
  50. endif # Shell is sane
  51. endif # CC is set to default
  52. # Default Fortran compiler (FC) is selected by f_check.
  53. ifndef MAKEFILE_RULE
  54. include $(TOPDIR)/Makefile.rule
  55. else
  56. include $(TOPDIR)/$(MAKEFILE_RULE)
  57. endif
  58. #
  59. # Beginning of system configuration
  60. #
  61. ifndef HOSTCC
  62. HOSTCC = $(CC)
  63. endif
  64. ifdef TARGET
  65. GETARCH_FLAGS := -DFORCE_$(TARGET)
  66. GETARCH_FLAGS += -DUSER_TARGET
  67. endif
  68. # Force fallbacks for 32bit
  69. ifeq ($(BINARY), 32)
  70. ifeq ($(TARGET), HASWELL)
  71. GETARCH_FLAGS := -DFORCE_NEHALEM
  72. endif
  73. ifeq ($(TARGET), SKYLAKEX)
  74. GETARCH_FLAGS := -DFORCE_NEHALEM
  75. endif
  76. ifeq ($(TARGET), SANDYBRIDGE)
  77. GETARCH_FLAGS := -DFORCE_NEHALEM
  78. endif
  79. ifeq ($(TARGET), BULLDOZER)
  80. GETARCH_FLAGS := -DFORCE_BARCELONA
  81. endif
  82. ifeq ($(TARGET), PILEDRIVER)
  83. GETARCH_FLAGS := -DFORCE_BARCELONA
  84. endif
  85. ifeq ($(TARGET), STEAMROLLER)
  86. GETARCH_FLAGS := -DFORCE_BARCELONA
  87. endif
  88. ifeq ($(TARGET), EXCAVATOR)
  89. GETARCH_FLAGS := -DFORCE_BARCELONA
  90. endif
  91. ifeq ($(TARGET), ZEN)
  92. GETARCH_FLAGS := -DFORCE_BARCELONA
  93. endif
  94. ifeq ($(TARGET), ARMV8)
  95. GETARCH_FLAGS := -DFORCE_ARMV7
  96. endif
  97. ifeq ($(TARGET), POWER8)
  98. GETARCH_FLAGS := -DFORCE_POWER6
  99. endif
  100. endif
  101. #TARGET_CORE will override TARGET which is used in DYNAMIC_ARCH=1.
  102. #
  103. ifdef TARGET_CORE
  104. GETARCH_FLAGS := -DFORCE_$(TARGET_CORE)
  105. endif
  106. # Force fallbacks for 32bit
  107. ifeq ($(BINARY), 32)
  108. ifeq ($(TARGET_CORE), HASWELL)
  109. GETARCH_FLAGS := -DFORCE_NEHALEM
  110. endif
  111. ifeq ($(TARGET_CORE), SKYLAKEX)
  112. GETARCH_FLAGS := -DFORCE_NEHALEM
  113. endif
  114. ifeq ($(TARGET_CORE), SANDYBRIDGE)
  115. GETARCH_FLAGS := -DFORCE_NEHALEM
  116. endif
  117. ifeq ($(TARGET_CORE), BULLDOZER)
  118. GETARCH_FLAGS := -DFORCE_BARCELONA
  119. endif
  120. ifeq ($(TARGET_CORE), PILEDRIVER)
  121. GETARCH_FLAGS := -DFORCE_BARCELONA
  122. endif
  123. ifeq ($(TARGET_CORE), STEAMROLLER)
  124. GETARCH_FLAGS := -DFORCE_BARCELONA
  125. endif
  126. ifeq ($(TARGET_CORE), EXCAVATOR)
  127. GETARCH_FLAGS := -DFORCE_BARCELONA
  128. endif
  129. ifeq ($(TARGET_CORE), ZEN)
  130. GETARCH_FLAGS := -DFORCE_BARCELONA
  131. endif
  132. endif
  133. # On x86_64 build getarch with march=native unless the compiler is PGI. This is required to detect AVX512 support in getarch.
  134. ifeq ($(HOSTARCH), x86_64)
  135. ifeq ($(findstring pgcc,$(HOSTCC)),)
  136. GETARCH_FLAGS += -march=native
  137. endif
  138. endif
  139. ifdef INTERFACE64
  140. ifneq ($(INTERFACE64), 0)
  141. GETARCH_FLAGS += -DUSE64BITINT
  142. endif
  143. endif
  144. ifndef GEMM_MULTITHREAD_THRESHOLD
  145. GEMM_MULTITHREAD_THRESHOLD=4
  146. endif
  147. GETARCH_FLAGS += -DGEMM_MULTITHREAD_THRESHOLD=$(GEMM_MULTITHREAD_THRESHOLD)
  148. ifeq ($(NO_AVX), 1)
  149. GETARCH_FLAGS += -DNO_AVX
  150. endif
  151. ifeq ($(BINARY), 32)
  152. GETARCH_FLAGS += -DNO_AVX -DNO_AVX2 -DNO_AVX512
  153. NO_AVX512 = 1
  154. endif
  155. ifeq ($(NO_AVX2), 1)
  156. GETARCH_FLAGS += -DNO_AVX2
  157. endif
  158. ifeq ($(NO_AVX512), 1)
  159. GETARCH_FLAGS += -DNO_AVX512
  160. endif
  161. ifeq ($(DEBUG), 1)
  162. GETARCH_FLAGS += -g
  163. endif
  164. ifeq ($(QUIET_MAKE), 1)
  165. MAKE += -s
  166. endif
  167. ifndef NO_PARALLEL_MAKE
  168. NO_PARALLEL_MAKE=0
  169. endif
  170. GETARCH_FLAGS += -DNO_PARALLEL_MAKE=$(NO_PARALLEL_MAKE)
  171. ifdef MAKE_NB_JOBS
  172. GETARCH_FLAGS += -DMAKE_NB_JOBS=$(MAKE_NB_JOBS)
  173. endif
  174. ifeq ($(HOSTCC), loongcc)
  175. GETARCH_FLAGS += -static
  176. endif
  177. #if don't use Fortran, it will only compile CBLAS.
  178. ifeq ($(ONLY_CBLAS), 1)
  179. NO_LAPACK = 1
  180. else
  181. ONLY_CBLAS = 0
  182. endif
  183. # This operation is expensive, so execution should be once.
  184. ifndef GOTOBLAS_MAKEFILE
  185. export GOTOBLAS_MAKEFILE = 1
  186. # Generating Makefile.conf and config.h
  187. 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)
  188. ifndef TARGET_CORE
  189. include $(TOPDIR)/Makefile.conf
  190. else
  191. include $(TOPDIR)/Makefile_kernel.conf
  192. endif
  193. endif
  194. ifndef NUM_PARALLEL
  195. NUM_PARALLEL = 1
  196. endif
  197. ifndef NUM_THREADS
  198. NUM_THREADS = $(NUM_CORES)
  199. endif
  200. ifeq ($(NUM_THREADS), 1)
  201. override USE_THREAD = 0
  202. override USE_OPENMP = 0
  203. endif
  204. ifdef USE_THREAD
  205. ifeq ($(USE_THREAD), 0)
  206. SMP =
  207. else
  208. SMP = 1
  209. endif
  210. else
  211. ifeq ($(NUM_THREAD), 1)
  212. SMP =
  213. else
  214. SMP = 1
  215. endif
  216. endif
  217. ifeq ($(SMP), 1)
  218. USE_LOCKING =
  219. endif
  220. ifndef NEED_PIC
  221. NEED_PIC = 1
  222. endif
  223. ARFLAGS =
  224. CPP = $(COMPILER) -E
  225. AR ?= $(CROSS_SUFFIX)ar
  226. AS ?= $(CROSS_SUFFIX)as
  227. LD ?= $(CROSS_SUFFIX)ld
  228. RANLIB ?= $(CROSS_SUFFIX)ranlib
  229. NM = $(CROSS_SUFFIX)nm
  230. DLLWRAP = $(CROSS_SUFFIX)dllwrap
  231. OBJCOPY = $(CROSS_SUFFIX)objcopy
  232. OBJCONV = $(CROSS_SUFFIX)objconv
  233. # When fortran support was either not detected or actively deselected, only build BLAS.
  234. ifeq ($(NOFORTRAN), 1)
  235. NO_LAPACK = 1
  236. override FEXTRALIB =
  237. endif
  238. ifeq ($(C_COMPILER), GCC)
  239. GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
  240. GCCVERSIONGT4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 4)
  241. GCCVERSIONEQ5 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` = 5)
  242. GCCVERSIONGT5 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 5)
  243. GCCVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 7)
  244. GCCVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  245. GCCVERSIONGTEQ11 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 11)
  246. GCCVERSIONGTEQ10 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 10)
  247. # Note that the behavior of -dumpversion is compile-time-configurable for
  248. # gcc-7.x and newer. Use -dumpfullversion there
  249. ifeq ($(GCCVERSIONGTEQ7),1)
  250. GCCDUMPVERSION_PARAM := -dumpfullversion
  251. else
  252. GCCDUMPVERSION_PARAM := -dumpversion
  253. endif
  254. GCCMINORVERSIONGTEQ2 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 2)
  255. GCCMINORVERSIONGTEQ7 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 7)
  256. endif
  257. #
  258. # OS dependent settings
  259. #
  260. ifeq ($(OSNAME), Darwin)
  261. ifndef MACOSX_DEPLOYMENT_TARGET
  262. export MACOSX_DEPLOYMENT_TARGET=10.8
  263. endif
  264. MD5SUM = md5 -r
  265. endif
  266. ifneq (,$(findstring $(OSNAME), FreeBSD OpenBSD DragonFly))
  267. MD5SUM = md5 -r
  268. endif
  269. ifeq ($(OSNAME), NetBSD)
  270. MD5SUM = md5 -n
  271. endif
  272. ifeq ($(OSNAME), Linux)
  273. EXTRALIB += -lm
  274. NO_EXPRECISION = 1
  275. endif
  276. ifeq ($(OSNAME), Android)
  277. EXTRALIB += -lm
  278. endif
  279. ifeq ($(OSNAME), AIX)
  280. EXTRALIB += -lm
  281. endif
  282. ifeq ($(OSNAME), WINNT)
  283. NEED_PIC = 0
  284. NO_EXPRECISION = 1
  285. EXTRALIB += -defaultlib:advapi32
  286. SUFFIX = obj
  287. PSUFFIX = pobj
  288. LIBSUFFIX = a
  289. ifeq ($(C_COMPILER), CLANG)
  290. CCOMMON_OPT += -DMS_ABI
  291. endif
  292. #Version tests for supporting specific features (MS_ABI, POWER9 intrinsics)
  293. ifeq ($(GCCVERSIONGT4), 1)
  294. # GCC Major version > 4
  295. # It is compatible with MSVC ABI.
  296. CCOMMON_OPT += -DMS_ABI
  297. endif
  298. ifeq ($(GCCVERSIONGTEQ4), 1)
  299. ifeq ($(GCCMINORVERSIONGTEQ7), 1)
  300. # GCC Version >=4.7
  301. # It is compatible with MSVC ABI.
  302. CCOMMON_OPT += -DMS_ABI
  303. endif
  304. endif
  305. # Ensure the correct stack alignment on Win32
  306. # http://permalink.gmane.org/gmane.comp.lib.openblas.general/97
  307. ifeq ($(ARCH), x86)
  308. CCOMMON_OPT += -mincoming-stack-boundary=2
  309. FCOMMON_OPT += -mincoming-stack-boundary=2
  310. endif
  311. endif
  312. ifeq ($(OSNAME), Interix)
  313. NEED_PIC = 0
  314. NO_EXPRECISION = 1
  315. INTERIX_TOOL_DIR = /opt/gcc.3.3/i586-pc-interix3/bin
  316. endif
  317. ifeq ($(OSNAME), CYGWIN_NT)
  318. NEED_PIC = 0
  319. NO_EXPRECISION = 1
  320. OS_CYGWIN_NT = 1
  321. endif
  322. ifneq ($(OSNAME), WINNT)
  323. ifneq ($(OSNAME), CYGWIN_NT)
  324. ifneq ($(OSNAME), Interix)
  325. ifneq ($(OSNAME), Android)
  326. ifdef SMP
  327. EXTRALIB += -lpthread
  328. endif
  329. endif
  330. endif
  331. endif
  332. endif
  333. # ifeq logical or
  334. ifeq ($(OSNAME), $(filter $(OSNAME),WINNT CYGWIN_NT Interix))
  335. OS_WINDOWS=1
  336. endif
  337. ifdef QUAD_PRECISION
  338. CCOMMON_OPT += -DQUAD_PRECISION
  339. NO_EXPRECISION = 1
  340. endif
  341. ifneq ($(ARCH), x86)
  342. ifneq ($(ARCH), x86_64)
  343. NO_EXPRECISION = 1
  344. endif
  345. endif
  346. ifdef UTEST_CHECK
  347. CCOMMON_OPT += -DUTEST_CHECK
  348. SANITY_CHECK = 1
  349. endif
  350. ifdef SANITY_CHECK
  351. CCOMMON_OPT += -DSANITY_CHECK -DREFNAME=$(*F)f$(BU)
  352. endif
  353. MAX_STACK_ALLOC ?= 2048
  354. ifneq ($(MAX_STACK_ALLOC), 0)
  355. CCOMMON_OPT += -DMAX_STACK_ALLOC=$(MAX_STACK_ALLOC)
  356. endif
  357. ifdef USE_LOCKING
  358. ifneq ($(USE_LOCKING), 0)
  359. CCOMMON_OPT += -DUSE_LOCKING
  360. endif
  361. endif
  362. #
  363. # Architecture dependent settings
  364. #
  365. ifeq ($(ARCH), x86)
  366. ifndef BINARY
  367. NO_BINARY_MODE = 1
  368. endif
  369. ifeq ($(CORE), generic)
  370. NO_EXPRECISION = 1
  371. endif
  372. ifndef NO_EXPRECISION
  373. ifeq ($(F_COMPILER), GFORTRAN)
  374. # ifeq logical or. GCC or LSB
  375. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  376. EXPRECISION = 1
  377. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  378. FCOMMON_OPT += -m128bit-long-double
  379. endif
  380. ifeq ($(C_COMPILER), CLANG)
  381. EXPRECISION = 1
  382. CCOMMON_OPT += -DEXPRECISION
  383. FCOMMON_OPT += -m128bit-long-double
  384. endif
  385. endif
  386. endif
  387. endif
  388. ifeq ($(ARCH), x86_64)
  389. ifeq ($(CORE), generic)
  390. NO_EXPRECISION = 1
  391. endif
  392. ifndef NO_EXPRECISION
  393. ifeq ($(F_COMPILER), GFORTRAN)
  394. # ifeq logical or. GCC or LSB
  395. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  396. EXPRECISION = 1
  397. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  398. FCOMMON_OPT += -m128bit-long-double
  399. endif
  400. ifeq ($(C_COMPILER), CLANG)
  401. EXPRECISION = 1
  402. CCOMMON_OPT += -DEXPRECISION
  403. FCOMMON_OPT += -m128bit-long-double
  404. endif
  405. endif
  406. endif
  407. endif
  408. ifeq ($(C_COMPILER), INTEL)
  409. CCOMMON_OPT += -wd981
  410. endif
  411. ifeq ($(USE_OPENMP), 1)
  412. #check
  413. ifeq ($(USE_THREAD), 0)
  414. $(error OpenBLAS: Cannot set both USE_OPENMP=1 and USE_THREAD=0. The USE_THREAD=0 is only for building single thread version.)
  415. endif
  416. # ifeq logical or. GCC or LSB
  417. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  418. CCOMMON_OPT += -fopenmp
  419. endif
  420. ifeq ($(C_COMPILER), CLANG)
  421. CCOMMON_OPT += -fopenmp
  422. endif
  423. ifeq ($(C_COMPILER), INTEL)
  424. CCOMMON_OPT += -fopenmp
  425. endif
  426. ifeq ($(C_COMPILER), PGI)
  427. CCOMMON_OPT += -mp
  428. endif
  429. ifeq ($(C_COMPILER), OPEN64)
  430. CCOMMON_OPT += -mp
  431. CEXTRALIB += -lstdc++
  432. endif
  433. ifeq ($(C_COMPILER), PATHSCALE)
  434. CCOMMON_OPT += -mp
  435. endif
  436. endif
  437. ifeq ($(DYNAMIC_ARCH), 1)
  438. ifeq ($(ARCH), x86)
  439. DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
  440. CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
  441. endif
  442. ifeq ($(ARCH), x86_64)
  443. DYNAMIC_CORE = PRESCOTT CORE2
  444. ifeq ($(DYNAMIC_OLDER), 1)
  445. DYNAMIC_CORE += PENRYN DUNNINGTON
  446. endif
  447. DYNAMIC_CORE += NEHALEM
  448. ifeq ($(DYNAMIC_OLDER), 1)
  449. DYNAMIC_CORE += OPTERON OPTERON_SSE3
  450. endif
  451. DYNAMIC_CORE += BARCELONA
  452. ifeq ($(DYNAMIC_OLDER), 1)
  453. DYNAMIC_CORE += BOBCAT ATOM NANO
  454. endif
  455. ifneq ($(NO_AVX), 1)
  456. DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER STEAMROLLER EXCAVATOR
  457. endif
  458. ifneq ($(NO_AVX2), 1)
  459. DYNAMIC_CORE += HASWELL ZEN
  460. endif
  461. ifneq ($(NO_AVX512), 1)
  462. ifneq ($(NO_AVX2), 1)
  463. DYNAMIC_CORE += SKYLAKEX
  464. endif
  465. endif
  466. endif
  467. ifdef DYNAMIC_LIST
  468. override DYNAMIC_CORE = PRESCOTT $(DYNAMIC_LIST)
  469. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_PRESCOTT
  470. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  471. CCOMMON_OPT += $(XCCOMMON_OPT)
  472. #CCOMMON_OPT += -DDYNAMIC_LIST='$(DYNAMIC_LIST)'
  473. endif
  474. ifeq ($(ARCH), arm64)
  475. DYNAMIC_CORE = ARMV8
  476. DYNAMIC_CORE += CORTEXA53
  477. DYNAMIC_CORE += CORTEXA57
  478. DYNAMIC_CORE += CORTEXA72
  479. DYNAMIC_CORE += CORTEXA73
  480. DYNAMIC_CORE += NEOVERSEN1
  481. DYNAMIC_CORE += FALKOR
  482. DYNAMIC_CORE += THUNDERX
  483. DYNAMIC_CORE += THUNDERX2T99
  484. DYNAMIC_CORE += TSV110
  485. DYNAMIC_CORE += EMAG8180
  486. DYNAMIC_CORE += THUNDERX3T110
  487. endif
  488. ifeq ($(ARCH), zarch)
  489. DYNAMIC_CORE = ZARCH_GENERIC
  490. # Z13 is supported since gcc-5.2, gcc-6, and in RHEL 7.3 and newer
  491. ifeq ($(GCCVERSIONGT5), 1)
  492. ZARCH_SUPPORT_Z13 := 1
  493. else ifeq ($(GCCVERSIONEQ5), 1)
  494. ifeq ($(GCCMINORVERSIONGTEQ2), 1)
  495. ZARCH_SUPPORT_Z13 := 1
  496. endif
  497. endif
  498. ifeq ($(wildcard /etc/redhat-release), /etc/redhat-release)
  499. ifeq ($(shell source /etc/os-release ; expr $$VERSION_ID \>= "7.3"), 1)
  500. ZARCH_SUPPORT_Z13 := 1
  501. endif
  502. endif
  503. ifeq ($(ZARCH_SUPPORT_Z13), 1)
  504. DYNAMIC_CORE += Z13
  505. else
  506. $(info OpenBLAS: Not building Z13 kernels because gcc is older than 5.2 or 6.x)
  507. endif
  508. ifeq ($(GCCVERSIONGTEQ7), 1)
  509. DYNAMIC_CORE += Z14
  510. else
  511. $(info OpenBLAS: Not building Z14 kernels because gcc is older than 7.x)
  512. endif
  513. endif
  514. ifeq ($(ARCH), power)
  515. DYNAMIC_CORE = POWER6
  516. DYNAMIC_CORE += POWER8
  517. ifneq ($(C_COMPILER), GCC)
  518. DYNAMIC_CORE += POWER9
  519. DYNAMIC_CORE += POWER10
  520. endif
  521. ifeq ($(C_COMPILER), GCC)
  522. ifeq ($(GCCVERSIONGT5), 1)
  523. DYNAMIC_CORE += POWER9
  524. else
  525. $(info, OpenBLAS: Your gcc version is too old to build the POWER9 kernels.)
  526. endif
  527. ifeq ($(GCCVERSIONGTEQ11), 1)
  528. DYNAMIC_CORE += POWER10
  529. else ifeq ($(GCCVERSIONGTEQ10), 1)
  530. ifeq ($(GCCMINORVERSIONGTEQ2), 1)
  531. DYNAMIC_CORE += POWER10
  532. endif
  533. else
  534. $(info, OpenBLAS: Your gcc version is too old to build the POWER10 kernels.)
  535. endif
  536. endif
  537. endif
  538. # If DYNAMIC_CORE is not set, DYNAMIC_ARCH cannot do anything, so force it to empty
  539. ifndef DYNAMIC_CORE
  540. override DYNAMIC_ARCH=
  541. endif
  542. endif
  543. ifeq ($(ARCH), ia64)
  544. NO_BINARY_MODE = 1
  545. BINARY_DEFINED = 1
  546. ifeq ($(F_COMPILER), GFORTRAN)
  547. ifeq ($(C_COMPILER), GCC)
  548. # EXPRECISION = 1
  549. # CCOMMON_OPT += -DEXPRECISION
  550. endif
  551. endif
  552. endif
  553. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  554. NO_BINARY_MODE = 1
  555. endif
  556. ifeq ($(ARCH), alpha)
  557. NO_BINARY_MODE = 1
  558. BINARY_DEFINED = 1
  559. endif
  560. ifeq ($(ARCH), arm)
  561. NO_BINARY_MODE = 1
  562. BINARY_DEFINED = 1
  563. CCOMMON_OPT += -marm
  564. FCOMMON_OPT += -marm
  565. # If softfp abi is mentioned on the command line, force it.
  566. ifeq ($(ARM_SOFTFP_ABI), 1)
  567. CCOMMON_OPT += -mfloat-abi=softfp
  568. FCOMMON_OPT += -mfloat-abi=softfp
  569. endif
  570. ifeq ($(OSNAME), Android)
  571. ifeq ($(ARM_SOFTFP_ABI), 1)
  572. EXTRALIB += -lm
  573. else
  574. EXTRALIB += -Wl,-lm_hard
  575. endif
  576. endif
  577. endif
  578. ifeq ($(ARCH), arm64)
  579. NO_BINARY_MODE = 1
  580. BINARY_DEFINED = 1
  581. ifdef INTERFACE64
  582. ifneq ($(INTERFACE64), 0)
  583. ifeq ($(F_COMPILER), GFORTRAN)
  584. FCOMMON_OPT += -fdefault-integer-8
  585. endif
  586. ifeq ($(F_COMPILER), FLANG)
  587. FCOMMON_OPT += -i8
  588. endif
  589. endif
  590. endif
  591. endif
  592. #
  593. # C Compiler dependent settings
  594. #
  595. # ifeq logical or. GCC or CLANG or LSB
  596. # http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
  597. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG LSB))
  598. CCOMMON_OPT += -Wall
  599. COMMON_PROF += -fno-inline
  600. NO_UNINITIALIZED_WARN = -Wno-uninitialized
  601. ifeq ($(QUIET_MAKE), 1)
  602. CCOMMON_OPT += $(NO_UNINITIALIZED_WARN) -Wno-unused
  603. endif
  604. ifdef NO_BINARY_MODE
  605. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  606. ifdef BINARY64
  607. CCOMMON_OPT += -mabi=64
  608. else
  609. CCOMMON_OPT += -mabi=n32
  610. endif
  611. BINARY_DEFINED = 1
  612. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  613. CCOMMON_OPT += -mabi=32
  614. BINARY_DEFINED = 1
  615. endif
  616. ifeq ($(CORE), LOONGSON3A)
  617. CCOMMON_OPT += -march=mips64
  618. FCOMMON_OPT += -march=mips64
  619. endif
  620. ifeq ($(CORE), LOONGSON3B)
  621. CCOMMON_OPT += -march=mips64
  622. FCOMMON_OPT += -march=mips64
  623. endif
  624. ifeq ($(CORE), MIPS24K)
  625. CCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  626. FCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  627. endif
  628. ifeq ($(CORE), MIPS1004K)
  629. CCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  630. FCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  631. endif
  632. ifeq ($(CORE), P5600)
  633. CCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  634. FCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  635. endif
  636. ifeq ($(CORE), I6400)
  637. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  638. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  639. endif
  640. ifeq ($(CORE), P6600)
  641. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  642. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  643. endif
  644. ifeq ($(CORE), I6500)
  645. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  646. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  647. endif
  648. ifeq ($(OSNAME), AIX)
  649. BINARY_DEFINED = 1
  650. endif
  651. endif
  652. ifndef BINARY_DEFINED
  653. ifneq ($(OSNAME), AIX)
  654. ifdef BINARY64
  655. CCOMMON_OPT += -m64
  656. else
  657. CCOMMON_OPT += -m32
  658. endif
  659. endif
  660. endif
  661. endif
  662. ifeq ($(C_COMPILER), PGI)
  663. ifdef BINARY64
  664. ifeq ($(ARCH), x86_64)
  665. CCOMMON_OPT += -tp p7-64 -D__MMX__ -Mnollvm
  666. else
  667. ifeq ($(ARCH), power)
  668. ifeq ($(CORE), POWER8)
  669. CCOMMON_OPT += -tp pwr8
  670. endif
  671. ifeq ($(CORE), POWER9)
  672. CCOMMON_OPT += -tp pwr9
  673. endif
  674. endif
  675. endif
  676. else
  677. CCOMMON_OPT += -tp p7
  678. endif
  679. endif
  680. ifeq ($(C_COMPILER), PATHSCALE)
  681. ifdef BINARY64
  682. CCOMMON_OPT += -m64
  683. else
  684. CCOMMON_OPT += -m32
  685. endif
  686. endif
  687. #
  688. # Fortran Compiler dependent settings
  689. #
  690. ifeq ($(F_COMPILER), FLANG)
  691. CCOMMON_OPT += -DF_INTERFACE_FLANG
  692. FCOMMON_OPT += -Mrecursive -Kieee
  693. ifeq ($(OSNAME), Linux)
  694. ifeq ($(ARCH), x86_64)
  695. FLANG_VENDOR := $(shell expr `$(FC) --version|cut -f 1 -d "."|head -1`)
  696. ifeq ($(FLANG_VENDOR),AOCC)
  697. FCOMMON_OPT += -fno-unroll-loops
  698. endif
  699. endif
  700. endif
  701. ifdef BINARY64
  702. ifdef INTERFACE64
  703. ifneq ($(INTERFACE64), 0)
  704. FCOMMON_OPT += -i8
  705. endif
  706. endif
  707. FCOMMON_OPT += -Wall
  708. else
  709. FCOMMON_OPT += -Wall
  710. endif
  711. ifeq ($(USE_OPENMP), 1)
  712. FCOMMON_OPT += -fopenmp
  713. endif
  714. endif
  715. ifeq ($(F_COMPILER), G77)
  716. CCOMMON_OPT += -DF_INTERFACE_G77
  717. FCOMMON_OPT += -Wall
  718. ifndef NO_BINARY_MODE
  719. ifneq ($(OSNAME), AIX)
  720. ifdef BINARY64
  721. FCOMMON_OPT += -m64
  722. else
  723. FCOMMON_OPT += -m32
  724. endif
  725. endif
  726. endif
  727. endif
  728. ifeq ($(F_COMPILER), G95)
  729. CCOMMON_OPT += -DF_INTERFACE_G95
  730. FCOMMON_OPT += -Wall
  731. ifneq ($(OSNAME), AIX)
  732. ifndef NO_BINARY_MODE
  733. ifdef BINARY64
  734. FCOMMON_OPT += -m64
  735. else
  736. FCOMMON_OPT += -m32
  737. endif
  738. endif
  739. ifneq ($(NO_LAPACKE), 1)
  740. FCOMMON_OPT += -fno-second-underscore
  741. endif
  742. endif
  743. endif
  744. ifeq ($(F_COMPILER), GFORTRAN)
  745. CCOMMON_OPT += -DF_INTERFACE_GFORT
  746. FCOMMON_OPT += -Wall
  747. # make single-threaded LAPACK calls thread-safe #1847
  748. FCOMMON_OPT += -frecursive
  749. # work around ABI problem with passing single-character arguments
  750. FCOMMON_OPT += -fno-optimize-sibling-calls
  751. #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
  752. ifneq ($(NO_LAPACK), 1)
  753. EXTRALIB += -lgfortran
  754. endif
  755. ifdef NO_BINARY_MODE
  756. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  757. ifdef BINARY64
  758. FCOMMON_OPT += -mabi=64
  759. else
  760. FCOMMON_OPT += -mabi=n32
  761. endif
  762. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  763. FCOMMON_OPT += -mabi=32
  764. endif
  765. else
  766. ifdef BINARY64
  767. ifneq ($(OSNAME), AIX)
  768. FCOMMON_OPT += -m64
  769. endif
  770. ifdef INTERFACE64
  771. ifneq ($(INTERFACE64), 0)
  772. FCOMMON_OPT += -fdefault-integer-8
  773. endif
  774. endif
  775. else
  776. ifneq ($(OSNAME), AIX)
  777. FCOMMON_OPT += -m32
  778. endif
  779. endif
  780. endif
  781. ifeq ($(USE_OPENMP), 1)
  782. FCOMMON_OPT += -fopenmp
  783. endif
  784. endif
  785. ifeq ($(F_COMPILER), INTEL)
  786. CCOMMON_OPT += -DF_INTERFACE_INTEL
  787. ifdef INTERFACE64
  788. ifneq ($(INTERFACE64), 0)
  789. FCOMMON_OPT += -i8
  790. endif
  791. endif
  792. FCOMMON_OPT += -recursive -fp-model strict -assume protect-parens
  793. ifeq ($(USE_OPENMP), 1)
  794. FCOMMON_OPT += -fopenmp
  795. endif
  796. endif
  797. ifeq ($(F_COMPILER), FUJITSU)
  798. CCOMMON_OPT += -DF_INTERFACE_FUJITSU
  799. ifeq ($(USE_OPENMP), 1)
  800. FCOMMON_OPT += -openmp
  801. endif
  802. endif
  803. ifeq ($(F_COMPILER), IBM)
  804. CCOMMON_OPT += -DF_INTERFACE_IBM
  805. # FCOMMON_OPT += -qarch=440
  806. ifdef BINARY64
  807. FCOMMON_OPT += -q64
  808. ifdef INTERFACE64
  809. ifneq ($(INTERFACE64), 0)
  810. FCOMMON_OPT += -qintsize=8
  811. endif
  812. endif
  813. else
  814. FCOMMON_OPT += -q32
  815. endif
  816. ifeq ($(USE_OPENMP), 1)
  817. FCOMMON_OPT += -openmp
  818. endif
  819. endif
  820. ifeq ($(F_COMPILER), PGI)
  821. CCOMMON_OPT += -DF_INTERFACE_PGI
  822. COMMON_PROF += -DPGICOMPILER
  823. ifdef BINARY64
  824. ifdef INTERFACE64
  825. ifneq ($(INTERFACE64), 0)
  826. FCOMMON_OPT += -i8
  827. endif
  828. endif
  829. ifeq ($(ARCH), x86_64)
  830. FCOMMON_OPT += -tp p7-64
  831. else
  832. ifeq ($(ARCH), power)
  833. ifeq ($(CORE), POWER8)
  834. FCOMMON_OPT += -tp pwr8
  835. endif
  836. ifeq ($(CORE), POWER9)
  837. FCOMMON_OPT += -tp pwr9
  838. endif
  839. endif
  840. endif
  841. else
  842. FCOMMON_OPT += -tp p7
  843. endif
  844. FCOMMON_OPT += -Mrecursive
  845. ifeq ($(USE_OPENMP), 1)
  846. FCOMMON_OPT += -mp
  847. endif
  848. endif
  849. ifeq ($(F_COMPILER), PATHSCALE)
  850. CCOMMON_OPT += -DF_INTERFACE_PATHSCALE
  851. ifdef BINARY64
  852. ifdef INTERFACE64
  853. ifneq ($(INTERFACE64), 0)
  854. FCOMMON_OPT += -i8
  855. endif
  856. endif
  857. endif
  858. ifeq ($(USE_OPENMP), 1)
  859. FCOMMON_OPT += -mp
  860. endif
  861. endif
  862. ifeq ($(F_COMPILER), OPEN64)
  863. CCOMMON_OPT += -DF_INTERFACE_OPEN64
  864. ifdef BINARY64
  865. ifdef INTERFACE64
  866. ifneq ($(INTERFACE64), 0)
  867. FCOMMON_OPT += -i8
  868. endif
  869. endif
  870. endif
  871. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  872. ifndef BINARY64
  873. FCOMMON_OPT += -n32
  874. else
  875. FCOMMON_OPT += -n64
  876. endif
  877. ifeq ($(CORE), LOONGSON3A)
  878. FCOMMON_OPT += -loongson3 -static
  879. endif
  880. ifeq ($(CORE), LOONGSON3B)
  881. FCOMMON_OPT += -loongson3 -static
  882. endif
  883. else
  884. ifndef BINARY64
  885. FCOMMON_OPT += -m32
  886. else
  887. FCOMMON_OPT += -m64
  888. endif
  889. endif
  890. ifeq ($(USE_OPENMP), 1)
  891. FEXTRALIB += -lstdc++
  892. FCOMMON_OPT += -mp
  893. endif
  894. endif
  895. ifeq ($(C_COMPILER), OPEN64)
  896. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  897. ifndef BINARY64
  898. CCOMMON_OPT += -n32
  899. else
  900. CCOMMON_OPT += -n64
  901. endif
  902. ifeq ($(CORE), LOONGSON3A)
  903. CCOMMON_OPT += -loongson3 -static
  904. endif
  905. ifeq ($(CORE), LOONGSON3B)
  906. CCOMMON_OPT += -loongson3 -static
  907. endif
  908. else
  909. ifndef BINARY64
  910. CCOMMON_OPT += -m32
  911. else
  912. CCOMMON_OPT += -m64
  913. endif
  914. endif
  915. endif
  916. ifeq ($(C_COMPILER), SUN)
  917. CCOMMON_OPT += -w
  918. ifeq ($(ARCH), x86)
  919. CCOMMON_OPT += -m32
  920. else
  921. FCOMMON_OPT += -m64
  922. endif
  923. endif
  924. ifeq ($(F_COMPILER), SUN)
  925. CCOMMON_OPT += -DF_INTERFACE_SUN
  926. ifeq ($(ARCH), x86)
  927. FCOMMON_OPT += -m32
  928. else
  929. FCOMMON_OPT += -m64
  930. endif
  931. ifeq ($(USE_OPENMP), 1)
  932. FCOMMON_OPT += -xopenmp=parallel
  933. endif
  934. endif
  935. ifeq ($(F_COMPILER), COMPAQ)
  936. CCOMMON_OPT += -DF_INTERFACE_COMPAQ
  937. ifeq ($(USE_OPENMP), 1)
  938. FCOMMON_OPT += -openmp
  939. endif
  940. endif
  941. ifdef BINARY64
  942. ifdef INTERFACE64
  943. ifneq ($(INTERFACE64), 0)
  944. CCOMMON_OPT +=
  945. #-DUSE64BITINT
  946. endif
  947. endif
  948. endif
  949. ifeq ($(NEED_PIC), 1)
  950. ifeq ($(C_COMPILER), IBM)
  951. CCOMMON_OPT += -qpic=large
  952. else
  953. CCOMMON_OPT += -fPIC
  954. endif
  955. ifeq ($(F_COMPILER), SUN)
  956. FCOMMON_OPT += -pic
  957. else
  958. FCOMMON_OPT += -fPIC
  959. endif
  960. endif
  961. ifeq ($(DYNAMIC_ARCH), 1)
  962. CCOMMON_OPT += -DDYNAMIC_ARCH
  963. endif
  964. ifeq ($(DYNAMIC_OLDER), 1)
  965. CCOMMON_OPT += -DDYNAMIC_OLDER
  966. endif
  967. ifeq ($(NO_LAPACK), 1)
  968. CCOMMON_OPT += -DNO_LAPACK
  969. #Disable LAPACK C interface
  970. NO_LAPACKE = 1
  971. endif
  972. ifeq ($(NO_LAPACKE), 1)
  973. CCOMMON_OPT += -DNO_LAPACKE
  974. endif
  975. ifeq ($(NO_AVX), 1)
  976. CCOMMON_OPT += -DNO_AVX
  977. endif
  978. ifeq ($(ARCH), x86)
  979. CCOMMON_OPT += -DNO_AVX
  980. endif
  981. ifeq ($(NO_AVX2), 1)
  982. CCOMMON_OPT += -DNO_AVX2
  983. endif
  984. ifeq ($(NO_AVX512), 1)
  985. CCOMMON_OPT += -DNO_AVX512
  986. endif
  987. ifdef SMP
  988. CCOMMON_OPT += -DSMP_SERVER
  989. ifeq ($(ARCH), mips64)
  990. ifneq ($(CORE), LOONGSON3B)
  991. USE_SIMPLE_THREADED_LEVEL3 = 1
  992. endif
  993. endif
  994. ifeq ($(USE_OPENMP), 1)
  995. # USE_SIMPLE_THREADED_LEVEL3 = 1
  996. # NO_AFFINITY = 1
  997. CCOMMON_OPT += -DUSE_OPENMP
  998. endif
  999. ifeq ($(BIGNUMA), 1)
  1000. CCOMMON_OPT += -DBIGNUMA
  1001. endif
  1002. endif
  1003. ifeq ($(NO_WARMUP), 1)
  1004. CCOMMON_OPT += -DNO_WARMUP
  1005. endif
  1006. ifeq ($(CONSISTENT_FPCSR), 1)
  1007. CCOMMON_OPT += -DCONSISTENT_FPCSR
  1008. endif
  1009. # Only for development
  1010. # CCOMMON_OPT += -DPARAMTEST
  1011. # CCOMMON_OPT += -DPREFETCHTEST
  1012. # CCOMMON_OPT += -DNO_SWITCHING
  1013. # USE_PAPI = 1
  1014. ifdef USE_PAPI
  1015. CCOMMON_OPT += -DUSE_PAPI
  1016. EXTRALIB += -lpapi -lperfctr
  1017. endif
  1018. ifdef DYNAMIC_THREADS
  1019. CCOMMON_OPT += -DDYNAMIC_THREADS
  1020. endif
  1021. CCOMMON_OPT += -DMAX_CPU_NUMBER=$(NUM_THREADS)
  1022. CCOMMON_OPT += -DMAX_PARALLEL_NUMBER=$(NUM_PARALLEL)
  1023. ifdef USE_SIMPLE_THREADED_LEVEL3
  1024. CCOMMON_OPT += -DUSE_SIMPLE_THREADED_LEVEL3
  1025. endif
  1026. ifeq ($(USE_TLS), 1)
  1027. CCOMMON_OPT += -DUSE_TLS
  1028. endif
  1029. ifeq ($(BUILD_HALF), 1)
  1030. CCOMMON_OPT += -DBUILD_HALF
  1031. endif
  1032. CCOMMON_OPT += -DVERSION=\"$(VERSION)\"
  1033. ifndef SYMBOLPREFIX
  1034. SYMBOLPREFIX =
  1035. endif
  1036. ifndef SYMBOLSUFFIX
  1037. SYMBOLSUFFIX =
  1038. endif
  1039. ifndef LIBNAMESUFFIX
  1040. LIBNAMEBASE = $(SYMBOLPREFIX)openblas$(SYMBOLSUFFIX)
  1041. else
  1042. LIBNAMEBASE = $(SYMBOLPREFIX)openblas$(SYMBOLSUFFIX)_$(LIBNAMESUFFIX)
  1043. endif
  1044. ifeq ($(OSNAME), CYGWIN_NT)
  1045. LIBPREFIX = cyg$(LIBNAMEBASE)
  1046. else
  1047. LIBPREFIX = lib$(LIBNAMEBASE)
  1048. endif
  1049. KERNELDIR = $(TOPDIR)/kernel/$(ARCH)
  1050. include $(TOPDIR)/Makefile.$(ARCH)
  1051. ifneq ($(C_COMPILER), PGI)
  1052. CCOMMON_OPT += -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME
  1053. endif
  1054. CCOMMON_OPT += -DASMNAME=$(FU)$(*F) -DASMFNAME=$(FU)$(*F)$(BU) -DNAME=$(*F)$(BU) -DCNAME=$(*F) -DCHAR_NAME=\"$(*F)$(BU)\" -DCHAR_CNAME=\"$(*F)\"
  1055. ifeq ($(CORE), PPC440)
  1056. CCOMMON_OPT += -DALLOC_QALLOC
  1057. endif
  1058. ifeq ($(CORE), PPC440FP2)
  1059. STATIC_ALLOCATION = 1
  1060. endif
  1061. ifneq ($(OSNAME), Linux)
  1062. NO_AFFINITY = 1
  1063. endif
  1064. ifneq ($(ARCH), x86_64)
  1065. ifneq ($(ARCH), x86)
  1066. ifneq ($(CORE), LOONGSON3B)
  1067. NO_AFFINITY = 1
  1068. endif
  1069. endif
  1070. endif
  1071. ifdef NO_AFFINITY
  1072. ifeq ($(NO_AFFINITY), 0)
  1073. override undefine NO_AFFINITY
  1074. else
  1075. CCOMMON_OPT += -DNO_AFFINITY
  1076. endif
  1077. endif
  1078. ifdef FUNCTION_PROFILE
  1079. CCOMMON_OPT += -DFUNCTION_PROFILE
  1080. endif
  1081. ifdef HUGETLB_ALLOCATION
  1082. CCOMMON_OPT += -DALLOC_HUGETLB
  1083. endif
  1084. ifdef HUGETLBFILE_ALLOCATION
  1085. CCOMMON_OPT += -DALLOC_HUGETLBFILE -DHUGETLB_FILE_NAME=$(HUGETLBFILE_ALLOCATION)
  1086. endif
  1087. ifdef STATIC_ALLOCATION
  1088. CCOMMON_OPT += -DALLOC_STATIC
  1089. endif
  1090. ifdef DEVICEDRIVER_ALLOCATION
  1091. CCOMMON_OPT += -DALLOC_DEVICEDRIVER -DDEVICEDRIVER_NAME=\"/dev/mapper\"
  1092. endif
  1093. ifdef MIXED_MEMORY_ALLOCATION
  1094. CCOMMON_OPT += -DMIXED_MEMORY_ALLOCATION
  1095. endif
  1096. ifeq ($(OSNAME), SunOS)
  1097. TAR = gtar
  1098. PATCH = gpatch
  1099. GREP = ggrep
  1100. AWK = nawk
  1101. else
  1102. TAR = tar
  1103. PATCH = patch
  1104. GREP = grep
  1105. AWK = awk
  1106. endif
  1107. ifndef MD5SUM
  1108. MD5SUM = md5sum
  1109. endif
  1110. REVISION = -r$(VERSION)
  1111. MAJOR_VERSION = $(word 1,$(subst ., ,$(VERSION)))
  1112. ifeq ($(DEBUG), 1)
  1113. COMMON_OPT += -g
  1114. endif
  1115. ifeq ($(DEBUG), 1)
  1116. FCOMMON_OPT += -g
  1117. endif
  1118. ifndef COMMON_OPT
  1119. COMMON_OPT = -O2
  1120. endif
  1121. ifndef FCOMMON_OPT
  1122. FCOMMON_OPT = -O2 -frecursive
  1123. endif
  1124. override CFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
  1125. override PFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
  1126. override FFLAGS += $(COMMON_OPT) $(FCOMMON_OPT)
  1127. override FPFLAGS += $(FCOMMON_OPT) $(COMMON_PROF)
  1128. #MAKEOVERRIDES =
  1129. ifdef NEED_PIC
  1130. ifeq (,$(findstring PIC,$(FFLAGS)))
  1131. override FFLAGS += -fPIC
  1132. endif
  1133. endif
  1134. #For LAPACK Fortran codes.
  1135. #Disable -fopenmp for LAPACK Fortran codes on Windows.
  1136. ifdef OS_WINDOWS
  1137. LAPACK_FFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FFLAGS))
  1138. LAPACK_FPFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FPFLAGS))
  1139. else
  1140. LAPACK_FFLAGS := $(FFLAGS)
  1141. LAPACK_FPFLAGS := $(FPFLAGS)
  1142. endif
  1143. LAPACK_CFLAGS = $(CFLAGS)
  1144. LAPACK_CFLAGS += -DHAVE_LAPACK_CONFIG_H
  1145. ifdef INTERFACE64
  1146. ifneq ($(INTERFACE64), 0)
  1147. LAPACK_CFLAGS += -DLAPACK_ILP64
  1148. endif
  1149. endif
  1150. ifdef OS_WINDOWS
  1151. LAPACK_CFLAGS += -DOPENBLAS_OS_WINDOWS
  1152. endif
  1153. ifeq ($(C_COMPILER), LSB)
  1154. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1155. endif
  1156. ifndef SUFFIX
  1157. SUFFIX = o
  1158. endif
  1159. ifndef PSUFFIX
  1160. PSUFFIX = po
  1161. endif
  1162. ifndef LIBSUFFIX
  1163. LIBSUFFIX = a
  1164. endif
  1165. ifneq ($(DYNAMIC_ARCH), 1)
  1166. ifndef SMP
  1167. LIBNAME = $(LIBPREFIX)_$(LIBCORE)$(REVISION).$(LIBSUFFIX)
  1168. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)$(REVISION)_p.$(LIBSUFFIX)
  1169. else
  1170. LIBNAME = $(LIBPREFIX)_$(LIBCORE)p$(REVISION).$(LIBSUFFIX)
  1171. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)p$(REVISION)_p.$(LIBSUFFIX)
  1172. endif
  1173. else
  1174. ifndef SMP
  1175. LIBNAME = $(LIBPREFIX)$(REVISION).$(LIBSUFFIX)
  1176. LIBNAME_P = $(LIBPREFIX)$(REVISION)_p.$(LIBSUFFIX)
  1177. else
  1178. LIBNAME = $(LIBPREFIX)p$(REVISION).$(LIBSUFFIX)
  1179. LIBNAME_P = $(LIBPREFIX)p$(REVISION)_p.$(LIBSUFFIX)
  1180. endif
  1181. endif
  1182. LIBDLLNAME = $(LIBPREFIX).dll
  1183. IMPLIBNAME = lib$(LIBNAMEBASE).dll.a
  1184. ifneq ($(OSNAME), AIX)
  1185. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.so)
  1186. else
  1187. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.a)
  1188. endif
  1189. LIBDYNNAME = $(LIBNAME:.$(LIBSUFFIX)=.dylib)
  1190. LIBDEFNAME = $(LIBNAME:.$(LIBSUFFIX)=.def)
  1191. LIBEXPNAME = $(LIBNAME:.$(LIBSUFFIX)=.exp)
  1192. LIBZIPNAME = $(LIBNAME:.$(LIBSUFFIX)=.zip)
  1193. LIBS = $(TOPDIR)/$(LIBNAME)
  1194. LIBS_P = $(TOPDIR)/$(LIBNAME_P)
  1195. LIB_COMPONENTS = BLAS
  1196. ifneq ($(NO_CBLAS), 1)
  1197. LIB_COMPONENTS += CBLAS
  1198. endif
  1199. ifneq ($(NO_LAPACK), 1)
  1200. LIB_COMPONENTS += LAPACK
  1201. ifneq ($(NO_LAPACKE), 1)
  1202. LIB_COMPONENTS += LAPACKE
  1203. endif
  1204. ifeq ($(BUILD_RELAPACK), 1)
  1205. LIB_COMPONENTS += ReLAPACK
  1206. endif
  1207. endif
  1208. ifeq ($(ONLY_CBLAS), 1)
  1209. LIB_COMPONENTS = CBLAS
  1210. endif
  1211. export OSNAME
  1212. export ARCH
  1213. export CORE
  1214. export LIBCORE
  1215. export __BYTE_ORDER__
  1216. export ELF_VERSION
  1217. export PGCPATH
  1218. export CONFIG
  1219. export CC
  1220. export FC
  1221. export BU
  1222. export FU
  1223. export NEED2UNDERSCORES
  1224. export USE_THREAD
  1225. export NUM_THREADS
  1226. export NUM_CORES
  1227. export SMP
  1228. export MAKEFILE_RULE
  1229. export NEED_PIC
  1230. export BINARY
  1231. export BINARY32
  1232. export BINARY64
  1233. export F_COMPILER
  1234. export C_COMPILER
  1235. export USE_OPENMP
  1236. export CROSS
  1237. export CROSS_SUFFIX
  1238. export NOFORTRAN
  1239. export NO_FBLAS
  1240. export EXTRALIB
  1241. export CEXTRALIB
  1242. export FEXTRALIB
  1243. export HAVE_SSE
  1244. export HAVE_SSE2
  1245. export HAVE_SSE3
  1246. export HAVE_SSSE3
  1247. export HAVE_SSE4_1
  1248. export HAVE_SSE4_2
  1249. export HAVE_SSE4A
  1250. export HAVE_SSE5
  1251. export HAVE_AVX
  1252. export HAVE_VFP
  1253. export HAVE_VFPV3
  1254. export HAVE_VFPV4
  1255. export HAVE_NEON
  1256. export HAVE_MSA
  1257. export MSA_FLAGS
  1258. export KERNELDIR
  1259. export FUNCTION_PROFILE
  1260. export TARGET_CORE
  1261. export NO_AVX512
  1262. export BUILD_HALF
  1263. export SHGEMM_UNROLL_M
  1264. export SHGEMM_UNROLL_N
  1265. export SGEMM_UNROLL_M
  1266. export SGEMM_UNROLL_N
  1267. export DGEMM_UNROLL_M
  1268. export DGEMM_UNROLL_N
  1269. export QGEMM_UNROLL_M
  1270. export QGEMM_UNROLL_N
  1271. export CGEMM_UNROLL_M
  1272. export CGEMM_UNROLL_N
  1273. export ZGEMM_UNROLL_M
  1274. export ZGEMM_UNROLL_N
  1275. export XGEMM_UNROLL_M
  1276. export XGEMM_UNROLL_N
  1277. export CGEMM3M_UNROLL_M
  1278. export CGEMM3M_UNROLL_N
  1279. export ZGEMM3M_UNROLL_M
  1280. export ZGEMM3M_UNROLL_N
  1281. export XGEMM3M_UNROLL_M
  1282. export XGEMM3M_UNROLL_N
  1283. ifdef USE_CUDA
  1284. export CUDADIR
  1285. export CUCC
  1286. export CUFLAGS
  1287. export CULIB
  1288. endif
  1289. .SUFFIXES: .$(PSUFFIX) .$(SUFFIX) .f
  1290. .f.$(SUFFIX):
  1291. $(FC) $(FFLAGS) -c $< -o $(@F)
  1292. .f.$(PSUFFIX):
  1293. $(FC) $(FPFLAGS) -pg -c $< -o $(@F)
  1294. ifdef BINARY64
  1295. PATHSCALEPATH = /opt/pathscale/lib/3.1
  1296. PGIPATH = /opt/pgi/linux86-64/7.1-5/lib
  1297. else
  1298. PATHSCALEPATH = /opt/pathscale/lib/3.1/32
  1299. PGIPATH = /opt/pgi/linux86/7.1-5/lib
  1300. endif
  1301. ACMLPATH = /opt/acml/4.3.0
  1302. ifneq ($(OSNAME), Darwin)
  1303. MKLPATH = /opt/intel/mkl/10.2.2.025/lib
  1304. else
  1305. MKLPATH = /Library/Frameworks/Intel_MKL.framework/Versions/10.0.1.014/lib
  1306. endif
  1307. ATLASPATH = /opt/atlas/3.9.17/opteron
  1308. FLAMEPATH = $(HOME)/flame/lib
  1309. ifneq ($(OSNAME), SunOS)
  1310. SUNPATH = /opt/sunstudio12.1
  1311. else
  1312. SUNPATH = /opt/SUNWspro
  1313. endif