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

6 years ago
13 years ago
1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285128612871288128912901291129212931294129512961297129812991300130113021303130413051306130713081309131013111312131313141315131613171318131913201321132213231324132513261327132813291330133113321333133413351336133713381339134013411342134313441345134613471348134913501351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384138513861387138813891390139113921393139413951396139713981399140014011402140314041405140614071408140914101411141214131414141514161417141814191420142114221423142414251426142714281429143014311432143314341435143614371438143914401441144214431444144514461447144814491450145114521453145414551456145714581459146014611462146314641465146614671468146914701471147214731474147514761477147814791480148114821483148414851486148714881489149014911492149314941495149614971498149915001501150215031504150515061507150815091510151115121513151415151516151715181519152015211522152315241525152615271528152915301531153215331534153515361537153815391540154115421543154415451546154715481549155015511552155315541555
  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), COOPERLAKE)
  77. GETARCH_FLAGS := -DFORCE_NEHALEM
  78. endif
  79. ifeq ($(TARGET), SANDYBRIDGE)
  80. GETARCH_FLAGS := -DFORCE_NEHALEM
  81. endif
  82. ifeq ($(TARGET), BULLDOZER)
  83. GETARCH_FLAGS := -DFORCE_BARCELONA
  84. endif
  85. ifeq ($(TARGET), PILEDRIVER)
  86. GETARCH_FLAGS := -DFORCE_BARCELONA
  87. endif
  88. ifeq ($(TARGET), STEAMROLLER)
  89. GETARCH_FLAGS := -DFORCE_BARCELONA
  90. endif
  91. ifeq ($(TARGET), EXCAVATOR)
  92. GETARCH_FLAGS := -DFORCE_BARCELONA
  93. endif
  94. ifeq ($(TARGET), ZEN)
  95. GETARCH_FLAGS := -DFORCE_BARCELONA
  96. endif
  97. ifeq ($(TARGET), ARMV8)
  98. GETARCH_FLAGS := -DFORCE_ARMV7
  99. endif
  100. ifeq ($(TARGET), POWER8)
  101. GETARCH_FLAGS := -DFORCE_POWER6
  102. endif
  103. endif
  104. #TARGET_CORE will override TARGET which is used in DYNAMIC_ARCH=1.
  105. #
  106. ifdef TARGET_CORE
  107. GETARCH_FLAGS := -DFORCE_$(TARGET_CORE)
  108. endif
  109. # Force fallbacks for 32bit
  110. ifeq ($(BINARY), 32)
  111. ifeq ($(TARGET_CORE), HASWELL)
  112. GETARCH_FLAGS := -DFORCE_NEHALEM
  113. endif
  114. ifeq ($(TARGET_CORE), SKYLAKEX)
  115. GETARCH_FLAGS := -DFORCE_NEHALEM
  116. endif
  117. ifeq ($(TARGET_CORE), COOPERLAKE)
  118. GETARCH_FLAGS := -DFORCE_NEHALEM
  119. endif
  120. ifeq ($(TARGET_CORE), SANDYBRIDGE)
  121. GETARCH_FLAGS := -DFORCE_NEHALEM
  122. endif
  123. ifeq ($(TARGET_CORE), BULLDOZER)
  124. GETARCH_FLAGS := -DFORCE_BARCELONA
  125. endif
  126. ifeq ($(TARGET_CORE), PILEDRIVER)
  127. GETARCH_FLAGS := -DFORCE_BARCELONA
  128. endif
  129. ifeq ($(TARGET_CORE), STEAMROLLER)
  130. GETARCH_FLAGS := -DFORCE_BARCELONA
  131. endif
  132. ifeq ($(TARGET_CORE), EXCAVATOR)
  133. GETARCH_FLAGS := -DFORCE_BARCELONA
  134. endif
  135. ifeq ($(TARGET_CORE), ZEN)
  136. GETARCH_FLAGS := -DFORCE_BARCELONA
  137. endif
  138. endif
  139. # On x86_64 build getarch with march=native unless the compiler is PGI. This is required to detect AVX512 support in getarch.
  140. ifeq ($(HOSTARCH), x86_64)
  141. ifeq ($(findstring pgcc,$(HOSTCC)),)
  142. GETARCH_FLAGS += -march=native
  143. endif
  144. endif
  145. ifdef INTERFACE64
  146. ifneq ($(INTERFACE64), 0)
  147. GETARCH_FLAGS += -DUSE64BITINT
  148. endif
  149. endif
  150. ifndef GEMM_MULTITHREAD_THRESHOLD
  151. GEMM_MULTITHREAD_THRESHOLD=4
  152. endif
  153. GETARCH_FLAGS += -DGEMM_MULTITHREAD_THRESHOLD=$(GEMM_MULTITHREAD_THRESHOLD)
  154. ifeq ($(NO_AVX), 1)
  155. GETARCH_FLAGS += -DNO_AVX
  156. endif
  157. ifeq ($(BINARY), 32)
  158. GETARCH_FLAGS += -DNO_AVX -DNO_AVX2 -DNO_AVX512
  159. NO_AVX512 = 1
  160. endif
  161. ifeq ($(NO_AVX2), 1)
  162. GETARCH_FLAGS += -DNO_AVX2
  163. endif
  164. ifeq ($(NO_AVX512), 1)
  165. GETARCH_FLAGS += -DNO_AVX512
  166. endif
  167. ifeq ($(DEBUG), 1)
  168. GETARCH_FLAGS += -g
  169. endif
  170. ifeq ($(QUIET_MAKE), 1)
  171. MAKE += -s
  172. endif
  173. ifndef NO_PARALLEL_MAKE
  174. NO_PARALLEL_MAKE=0
  175. endif
  176. GETARCH_FLAGS += -DNO_PARALLEL_MAKE=$(NO_PARALLEL_MAKE)
  177. ifdef MAKE_NB_JOBS
  178. GETARCH_FLAGS += -DMAKE_NB_JOBS=$(MAKE_NB_JOBS)
  179. endif
  180. ifeq ($(HOSTCC), loongcc)
  181. GETARCH_FLAGS += -static
  182. endif
  183. #if don't use Fortran, it will only compile CBLAS.
  184. ifeq ($(ONLY_CBLAS), 1)
  185. NO_LAPACK = 1
  186. else
  187. ONLY_CBLAS = 0
  188. endif
  189. # This operation is expensive, so execution should be once.
  190. ifndef GOTOBLAS_MAKEFILE
  191. export GOTOBLAS_MAKEFILE = 1
  192. # Generating Makefile.conf and config.h
  193. 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)
  194. ifndef TARGET_CORE
  195. include $(TOPDIR)/Makefile.conf
  196. else
  197. include $(TOPDIR)/Makefile_kernel.conf
  198. endif
  199. endif
  200. ifndef NUM_PARALLEL
  201. NUM_PARALLEL = 1
  202. endif
  203. ifndef NUM_THREADS
  204. NUM_THREADS = $(NUM_CORES)
  205. endif
  206. ifeq ($(NUM_THREADS), 1)
  207. override USE_THREAD = 0
  208. override USE_OPENMP = 0
  209. endif
  210. ifdef USE_THREAD
  211. ifeq ($(USE_THREAD), 0)
  212. SMP =
  213. else
  214. SMP = 1
  215. endif
  216. else
  217. ifeq ($(NUM_THREAD), 1)
  218. SMP =
  219. else
  220. SMP = 1
  221. endif
  222. endif
  223. ifeq ($(SMP), 1)
  224. USE_LOCKING =
  225. endif
  226. ifndef NEED_PIC
  227. NEED_PIC = 1
  228. endif
  229. ARFLAGS =
  230. CPP = $(COMPILER) -E
  231. AR ?= $(CROSS_SUFFIX)ar
  232. AS ?= $(CROSS_SUFFIX)as
  233. LD ?= $(CROSS_SUFFIX)ld
  234. RANLIB ?= $(CROSS_SUFFIX)ranlib
  235. NM = $(CROSS_SUFFIX)nm
  236. DLLWRAP = $(CROSS_SUFFIX)dllwrap
  237. OBJCOPY = $(CROSS_SUFFIX)objcopy
  238. OBJCONV = $(CROSS_SUFFIX)objconv
  239. # When fortran support was either not detected or actively deselected, only build BLAS.
  240. ifeq ($(NOFORTRAN), 1)
  241. NO_LAPACK = 1
  242. override FEXTRALIB =
  243. endif
  244. ifeq ($(C_COMPILER), GCC)
  245. GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
  246. GCCVERSIONGT4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 4)
  247. GCCVERSIONEQ5 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` = 5)
  248. GCCVERSIONGT5 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 5)
  249. GCCVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 7)
  250. GCCVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  251. GCCVERSIONGTEQ11 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 11)
  252. GCCVERSIONGTEQ10 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 10)
  253. # Note that the behavior of -dumpversion is compile-time-configurable for
  254. # gcc-7.x and newer. Use -dumpfullversion there
  255. ifeq ($(GCCVERSIONGTEQ7),1)
  256. GCCDUMPVERSION_PARAM := -dumpfullversion
  257. else
  258. GCCDUMPVERSION_PARAM := -dumpversion
  259. endif
  260. GCCMINORVERSIONGTEQ2 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 2)
  261. GCCMINORVERSIONGTEQ7 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 7)
  262. endif
  263. #
  264. # OS dependent settings
  265. #
  266. ifeq ($(OSNAME), Darwin)
  267. ifndef MACOSX_DEPLOYMENT_TARGET
  268. export MACOSX_DEPLOYMENT_TARGET=10.8
  269. endif
  270. MD5SUM = md5 -r
  271. endif
  272. ifneq (,$(findstring $(OSNAME), FreeBSD OpenBSD DragonFly))
  273. MD5SUM = md5 -r
  274. endif
  275. ifeq ($(OSNAME), NetBSD)
  276. MD5SUM = md5 -n
  277. endif
  278. ifeq ($(OSNAME), Linux)
  279. EXTRALIB += -lm
  280. NO_EXPRECISION = 1
  281. endif
  282. ifeq ($(OSNAME), Android)
  283. EXTRALIB += -lm
  284. endif
  285. ifeq ($(OSNAME), AIX)
  286. EXTRALIB += -lm
  287. endif
  288. ifeq ($(OSNAME), WINNT)
  289. NEED_PIC = 0
  290. NO_EXPRECISION = 1
  291. EXTRALIB += -defaultlib:advapi32
  292. SUFFIX = obj
  293. PSUFFIX = pobj
  294. LIBSUFFIX = a
  295. ifeq ($(C_COMPILER), CLANG)
  296. CCOMMON_OPT += -DMS_ABI
  297. endif
  298. #Version tests for supporting specific features (MS_ABI, POWER9 intrinsics)
  299. ifeq ($(GCCVERSIONGT4), 1)
  300. # GCC Major version > 4
  301. # It is compatible with MSVC ABI.
  302. CCOMMON_OPT += -DMS_ABI
  303. endif
  304. ifeq ($(GCCVERSIONGTEQ4), 1)
  305. ifeq ($(GCCMINORVERSIONGTEQ7), 1)
  306. # GCC Version >=4.7
  307. # It is compatible with MSVC ABI.
  308. CCOMMON_OPT += -DMS_ABI
  309. endif
  310. endif
  311. # Ensure the correct stack alignment on Win32
  312. # http://permalink.gmane.org/gmane.comp.lib.openblas.general/97
  313. ifeq ($(ARCH), x86)
  314. CCOMMON_OPT += -mincoming-stack-boundary=2
  315. FCOMMON_OPT += -mincoming-stack-boundary=2
  316. endif
  317. endif
  318. ifeq ($(OSNAME), Interix)
  319. NEED_PIC = 0
  320. NO_EXPRECISION = 1
  321. INTERIX_TOOL_DIR = /opt/gcc.3.3/i586-pc-interix3/bin
  322. endif
  323. ifeq ($(OSNAME), CYGWIN_NT)
  324. NEED_PIC = 0
  325. NO_EXPRECISION = 1
  326. OS_CYGWIN_NT = 1
  327. endif
  328. ifneq ($(OSNAME), WINNT)
  329. ifneq ($(OSNAME), CYGWIN_NT)
  330. ifneq ($(OSNAME), Interix)
  331. ifneq ($(OSNAME), Android)
  332. ifdef SMP
  333. EXTRALIB += -lpthread
  334. endif
  335. endif
  336. endif
  337. endif
  338. endif
  339. # ifeq logical or
  340. ifeq ($(OSNAME), $(filter $(OSNAME),WINNT CYGWIN_NT Interix))
  341. OS_WINDOWS=1
  342. endif
  343. ifdef QUAD_PRECISION
  344. CCOMMON_OPT += -DQUAD_PRECISION
  345. NO_EXPRECISION = 1
  346. endif
  347. ifneq ($(ARCH), x86)
  348. ifneq ($(ARCH), x86_64)
  349. NO_EXPRECISION = 1
  350. endif
  351. endif
  352. ifdef UTEST_CHECK
  353. CCOMMON_OPT += -DUTEST_CHECK
  354. SANITY_CHECK = 1
  355. endif
  356. ifdef SANITY_CHECK
  357. CCOMMON_OPT += -DSANITY_CHECK -DREFNAME=$(*F)f$(BU)
  358. endif
  359. MAX_STACK_ALLOC ?= 2048
  360. ifneq ($(MAX_STACK_ALLOC), 0)
  361. CCOMMON_OPT += -DMAX_STACK_ALLOC=$(MAX_STACK_ALLOC)
  362. endif
  363. ifdef USE_LOCKING
  364. ifneq ($(USE_LOCKING), 0)
  365. CCOMMON_OPT += -DUSE_LOCKING
  366. endif
  367. endif
  368. #
  369. # Architecture dependent settings
  370. #
  371. ifeq ($(ARCH), x86)
  372. ifndef BINARY
  373. NO_BINARY_MODE = 1
  374. endif
  375. ifeq ($(CORE), generic)
  376. NO_EXPRECISION = 1
  377. endif
  378. ifndef NO_EXPRECISION
  379. ifeq ($(F_COMPILER), GFORTRAN)
  380. # ifeq logical or. GCC or LSB
  381. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  382. EXPRECISION = 1
  383. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  384. FCOMMON_OPT += -m128bit-long-double
  385. endif
  386. ifeq ($(C_COMPILER), CLANG)
  387. EXPRECISION = 1
  388. CCOMMON_OPT += -DEXPRECISION
  389. FCOMMON_OPT += -m128bit-long-double
  390. endif
  391. endif
  392. endif
  393. endif
  394. ifeq ($(ARCH), x86_64)
  395. ifeq ($(CORE), generic)
  396. NO_EXPRECISION = 1
  397. endif
  398. ifndef NO_EXPRECISION
  399. ifeq ($(F_COMPILER), GFORTRAN)
  400. # ifeq logical or. GCC or LSB
  401. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  402. EXPRECISION = 1
  403. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  404. FCOMMON_OPT += -m128bit-long-double
  405. endif
  406. ifeq ($(C_COMPILER), CLANG)
  407. EXPRECISION = 1
  408. CCOMMON_OPT += -DEXPRECISION
  409. FCOMMON_OPT += -m128bit-long-double
  410. endif
  411. endif
  412. endif
  413. endif
  414. ifeq ($(C_COMPILER), INTEL)
  415. CCOMMON_OPT += -wd981
  416. endif
  417. ifeq ($(USE_OPENMP), 1)
  418. #check
  419. ifeq ($(USE_THREAD), 0)
  420. $(error OpenBLAS: Cannot set both USE_OPENMP=1 and USE_THREAD=0. The USE_THREAD=0 is only for building single thread version.)
  421. endif
  422. # ifeq logical or. GCC or LSB
  423. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  424. CCOMMON_OPT += -fopenmp
  425. endif
  426. ifeq ($(C_COMPILER), CLANG)
  427. CCOMMON_OPT += -fopenmp
  428. endif
  429. ifeq ($(C_COMPILER), INTEL)
  430. CCOMMON_OPT += -fopenmp
  431. endif
  432. ifeq ($(C_COMPILER), PGI)
  433. CCOMMON_OPT += -mp
  434. endif
  435. ifeq ($(C_COMPILER), OPEN64)
  436. CCOMMON_OPT += -mp
  437. CEXTRALIB += -lstdc++
  438. endif
  439. ifeq ($(C_COMPILER), PATHSCALE)
  440. CCOMMON_OPT += -mp
  441. endif
  442. endif
  443. ifeq ($(DYNAMIC_ARCH), 1)
  444. ifeq ($(ARCH), x86)
  445. DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
  446. CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
  447. endif
  448. ifeq ($(ARCH), x86_64)
  449. DYNAMIC_CORE = PRESCOTT CORE2
  450. ifeq ($(DYNAMIC_OLDER), 1)
  451. DYNAMIC_CORE += PENRYN DUNNINGTON
  452. endif
  453. DYNAMIC_CORE += NEHALEM
  454. ifeq ($(DYNAMIC_OLDER), 1)
  455. DYNAMIC_CORE += OPTERON OPTERON_SSE3
  456. endif
  457. DYNAMIC_CORE += BARCELONA
  458. ifeq ($(DYNAMIC_OLDER), 1)
  459. DYNAMIC_CORE += BOBCAT ATOM NANO
  460. endif
  461. ifneq ($(NO_AVX), 1)
  462. DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER STEAMROLLER EXCAVATOR
  463. endif
  464. ifneq ($(NO_AVX2), 1)
  465. DYNAMIC_CORE += HASWELL ZEN
  466. endif
  467. ifneq ($(NO_AVX512), 1)
  468. ifneq ($(NO_AVX2), 1)
  469. DYNAMIC_CORE += SKYLAKEX COOPERLAKE
  470. endif
  471. endif
  472. endif
  473. ifdef DYNAMIC_LIST
  474. override DYNAMIC_CORE = PRESCOTT $(DYNAMIC_LIST)
  475. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_PRESCOTT
  476. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  477. CCOMMON_OPT += $(XCCOMMON_OPT)
  478. #CCOMMON_OPT += -DDYNAMIC_LIST='$(DYNAMIC_LIST)'
  479. endif
  480. ifeq ($(ARCH), arm64)
  481. DYNAMIC_CORE = ARMV8
  482. DYNAMIC_CORE += CORTEXA53
  483. DYNAMIC_CORE += CORTEXA57
  484. DYNAMIC_CORE += CORTEXA72
  485. DYNAMIC_CORE += CORTEXA73
  486. DYNAMIC_CORE += NEOVERSEN1
  487. DYNAMIC_CORE += FALKOR
  488. DYNAMIC_CORE += THUNDERX
  489. DYNAMIC_CORE += THUNDERX2T99
  490. DYNAMIC_CORE += TSV110
  491. DYNAMIC_CORE += EMAG8180
  492. DYNAMIC_CORE += THUNDERX3T110
  493. endif
  494. ifeq ($(ARCH), zarch)
  495. DYNAMIC_CORE = ZARCH_GENERIC
  496. # Z13 is supported since gcc-5.2, gcc-6, and in RHEL 7.3 and newer
  497. ifeq ($(GCCVERSIONGT5), 1)
  498. ZARCH_SUPPORT_Z13 := 1
  499. else ifeq ($(GCCVERSIONEQ5), 1)
  500. ifeq ($(GCCMINORVERSIONGTEQ2), 1)
  501. ZARCH_SUPPORT_Z13 := 1
  502. endif
  503. endif
  504. ifeq ($(wildcard /etc/redhat-release), /etc/redhat-release)
  505. ifeq ($(shell source /etc/os-release ; expr $$VERSION_ID \>= "7.3"), 1)
  506. ZARCH_SUPPORT_Z13 := 1
  507. endif
  508. endif
  509. ifeq ($(ZARCH_SUPPORT_Z13), 1)
  510. DYNAMIC_CORE += Z13
  511. else
  512. $(info OpenBLAS: Not building Z13 kernels because gcc is older than 5.2 or 6.x)
  513. endif
  514. ifeq ($(GCCVERSIONGTEQ7), 1)
  515. DYNAMIC_CORE += Z14
  516. else
  517. $(info OpenBLAS: Not building Z14 kernels because gcc is older than 7.x)
  518. endif
  519. endif
  520. ifeq ($(ARCH), power)
  521. DYNAMIC_CORE = POWER6
  522. DYNAMIC_CORE += POWER8
  523. ifneq ($(C_COMPILER), GCC)
  524. DYNAMIC_CORE += POWER9
  525. DYNAMIC_CORE += POWER10
  526. endif
  527. ifeq ($(C_COMPILER), GCC)
  528. ifeq ($(GCCVERSIONGT5), 1)
  529. DYNAMIC_CORE += POWER9
  530. else
  531. $(info, OpenBLAS: Your gcc version is too old to build the POWER9 kernels.)
  532. endif
  533. ifeq ($(GCCVERSIONGTEQ11), 1)
  534. DYNAMIC_CORE += POWER10
  535. else ifeq ($(GCCVERSIONGTEQ10), 1)
  536. ifeq ($(GCCMINORVERSIONGTEQ2), 1)
  537. DYNAMIC_CORE += POWER10
  538. endif
  539. else
  540. $(info, OpenBLAS: Your gcc version is too old to build the POWER10 kernels.)
  541. endif
  542. endif
  543. endif
  544. # If DYNAMIC_CORE is not set, DYNAMIC_ARCH cannot do anything, so force it to empty
  545. ifndef DYNAMIC_CORE
  546. override DYNAMIC_ARCH=
  547. endif
  548. endif
  549. ifeq ($(ARCH), ia64)
  550. NO_BINARY_MODE = 1
  551. BINARY_DEFINED = 1
  552. ifeq ($(F_COMPILER), GFORTRAN)
  553. ifeq ($(C_COMPILER), GCC)
  554. # EXPRECISION = 1
  555. # CCOMMON_OPT += -DEXPRECISION
  556. endif
  557. endif
  558. endif
  559. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  560. NO_BINARY_MODE = 1
  561. endif
  562. ifeq ($(ARCH), alpha)
  563. NO_BINARY_MODE = 1
  564. BINARY_DEFINED = 1
  565. endif
  566. ifeq ($(ARCH), arm)
  567. NO_BINARY_MODE = 1
  568. BINARY_DEFINED = 1
  569. CCOMMON_OPT += -marm
  570. FCOMMON_OPT += -marm
  571. # If softfp abi is mentioned on the command line, force it.
  572. ifeq ($(ARM_SOFTFP_ABI), 1)
  573. CCOMMON_OPT += -mfloat-abi=softfp
  574. FCOMMON_OPT += -mfloat-abi=softfp
  575. endif
  576. ifeq ($(OSNAME), Android)
  577. ifeq ($(ARM_SOFTFP_ABI), 1)
  578. EXTRALIB += -lm
  579. else
  580. EXTRALIB += -Wl,-lm_hard
  581. endif
  582. endif
  583. endif
  584. ifeq ($(ARCH), arm64)
  585. NO_BINARY_MODE = 1
  586. BINARY_DEFINED = 1
  587. ifdef INTERFACE64
  588. ifneq ($(INTERFACE64), 0)
  589. ifeq ($(F_COMPILER), GFORTRAN)
  590. FCOMMON_OPT += -fdefault-integer-8
  591. endif
  592. ifeq ($(F_COMPILER), FLANG)
  593. FCOMMON_OPT += -i8
  594. endif
  595. endif
  596. endif
  597. endif
  598. #
  599. # C Compiler dependent settings
  600. #
  601. # ifeq logical or. GCC or CLANG or LSB
  602. # http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
  603. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG LSB))
  604. CCOMMON_OPT += -Wall
  605. COMMON_PROF += -fno-inline
  606. NO_UNINITIALIZED_WARN = -Wno-uninitialized
  607. ifeq ($(QUIET_MAKE), 1)
  608. CCOMMON_OPT += $(NO_UNINITIALIZED_WARN) -Wno-unused
  609. endif
  610. ifdef NO_BINARY_MODE
  611. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  612. ifdef BINARY64
  613. CCOMMON_OPT += -mabi=64
  614. else
  615. CCOMMON_OPT += -mabi=n32
  616. endif
  617. BINARY_DEFINED = 1
  618. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  619. CCOMMON_OPT += -mabi=32
  620. BINARY_DEFINED = 1
  621. endif
  622. ifeq ($(CORE), LOONGSON3A)
  623. CCOMMON_OPT += -march=mips64
  624. FCOMMON_OPT += -march=mips64
  625. endif
  626. ifeq ($(CORE), LOONGSON3B)
  627. CCOMMON_OPT += -march=mips64
  628. FCOMMON_OPT += -march=mips64
  629. endif
  630. ifeq ($(CORE), MIPS24K)
  631. CCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  632. FCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  633. endif
  634. ifeq ($(CORE), MIPS1004K)
  635. CCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  636. FCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  637. endif
  638. ifeq ($(CORE), P5600)
  639. CCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  640. FCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  641. endif
  642. ifeq ($(CORE), I6400)
  643. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  644. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  645. endif
  646. ifeq ($(CORE), P6600)
  647. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  648. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  649. endif
  650. ifeq ($(CORE), I6500)
  651. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  652. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  653. endif
  654. ifeq ($(OSNAME), AIX)
  655. BINARY_DEFINED = 1
  656. endif
  657. endif
  658. ifndef BINARY_DEFINED
  659. ifneq ($(OSNAME), AIX)
  660. ifdef BINARY64
  661. CCOMMON_OPT += -m64
  662. else
  663. CCOMMON_OPT += -m32
  664. endif
  665. endif
  666. endif
  667. endif
  668. ifeq ($(C_COMPILER), PGI)
  669. ifdef BINARY64
  670. ifeq ($(ARCH), x86_64)
  671. CCOMMON_OPT += -tp p7-64 -D__MMX__ -Mnollvm
  672. else
  673. ifeq ($(ARCH), power)
  674. ifeq ($(CORE), POWER8)
  675. CCOMMON_OPT += -tp pwr8
  676. endif
  677. ifeq ($(CORE), POWER9)
  678. CCOMMON_OPT += -tp pwr9
  679. endif
  680. endif
  681. endif
  682. else
  683. CCOMMON_OPT += -tp p7
  684. endif
  685. endif
  686. ifeq ($(C_COMPILER), PATHSCALE)
  687. ifdef BINARY64
  688. CCOMMON_OPT += -m64
  689. else
  690. CCOMMON_OPT += -m32
  691. endif
  692. endif
  693. #
  694. # Fortran Compiler dependent settings
  695. #
  696. ifeq ($(F_COMPILER), FLANG)
  697. CCOMMON_OPT += -DF_INTERFACE_FLANG
  698. FCOMMON_OPT += -Mrecursive -Kieee
  699. ifeq ($(OSNAME), Linux)
  700. ifeq ($(ARCH), x86_64)
  701. FLANG_VENDOR := $(shell expr `$(FC) --version|cut -f 1 -d "."|head -1`)
  702. ifeq ($(FLANG_VENDOR),AOCC)
  703. FCOMMON_OPT += -fno-unroll-loops
  704. endif
  705. endif
  706. endif
  707. ifdef BINARY64
  708. ifdef INTERFACE64
  709. ifneq ($(INTERFACE64), 0)
  710. FCOMMON_OPT += -i8
  711. endif
  712. endif
  713. FCOMMON_OPT += -Wall
  714. else
  715. FCOMMON_OPT += -Wall
  716. endif
  717. ifeq ($(USE_OPENMP), 1)
  718. FCOMMON_OPT += -fopenmp
  719. endif
  720. endif
  721. ifeq ($(F_COMPILER), G77)
  722. CCOMMON_OPT += -DF_INTERFACE_G77
  723. FCOMMON_OPT += -Wall
  724. ifndef NO_BINARY_MODE
  725. ifneq ($(OSNAME), AIX)
  726. ifdef BINARY64
  727. FCOMMON_OPT += -m64
  728. else
  729. FCOMMON_OPT += -m32
  730. endif
  731. endif
  732. endif
  733. endif
  734. ifeq ($(F_COMPILER), G95)
  735. CCOMMON_OPT += -DF_INTERFACE_G95
  736. FCOMMON_OPT += -Wall
  737. ifneq ($(OSNAME), AIX)
  738. ifndef NO_BINARY_MODE
  739. ifdef BINARY64
  740. FCOMMON_OPT += -m64
  741. else
  742. FCOMMON_OPT += -m32
  743. endif
  744. endif
  745. ifneq ($(NO_LAPACKE), 1)
  746. FCOMMON_OPT += -fno-second-underscore
  747. endif
  748. endif
  749. endif
  750. ifeq ($(F_COMPILER), GFORTRAN)
  751. CCOMMON_OPT += -DF_INTERFACE_GFORT
  752. FCOMMON_OPT += -Wall
  753. # make single-threaded LAPACK calls thread-safe #1847
  754. FCOMMON_OPT += -frecursive
  755. # work around ABI problem with passing single-character arguments
  756. FCOMMON_OPT += -fno-optimize-sibling-calls
  757. #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
  758. ifneq ($(NO_LAPACK), 1)
  759. EXTRALIB += -lgfortran
  760. endif
  761. ifdef NO_BINARY_MODE
  762. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  763. ifdef BINARY64
  764. FCOMMON_OPT += -mabi=64
  765. else
  766. FCOMMON_OPT += -mabi=n32
  767. endif
  768. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  769. FCOMMON_OPT += -mabi=32
  770. endif
  771. else
  772. ifdef BINARY64
  773. ifneq ($(OSNAME), AIX)
  774. FCOMMON_OPT += -m64
  775. endif
  776. ifdef INTERFACE64
  777. ifneq ($(INTERFACE64), 0)
  778. FCOMMON_OPT += -fdefault-integer-8
  779. endif
  780. endif
  781. else
  782. ifneq ($(OSNAME), AIX)
  783. FCOMMON_OPT += -m32
  784. endif
  785. endif
  786. endif
  787. ifeq ($(USE_OPENMP), 1)
  788. FCOMMON_OPT += -fopenmp
  789. endif
  790. endif
  791. ifeq ($(F_COMPILER), INTEL)
  792. CCOMMON_OPT += -DF_INTERFACE_INTEL
  793. ifdef INTERFACE64
  794. ifneq ($(INTERFACE64), 0)
  795. FCOMMON_OPT += -i8
  796. endif
  797. endif
  798. FCOMMON_OPT += -recursive -fp-model strict -assume protect-parens
  799. ifeq ($(USE_OPENMP), 1)
  800. FCOMMON_OPT += -fopenmp
  801. endif
  802. endif
  803. ifeq ($(F_COMPILER), FUJITSU)
  804. CCOMMON_OPT += -DF_INTERFACE_FUJITSU
  805. ifeq ($(USE_OPENMP), 1)
  806. FCOMMON_OPT += -openmp
  807. endif
  808. endif
  809. ifeq ($(F_COMPILER), IBM)
  810. CCOMMON_OPT += -DF_INTERFACE_IBM
  811. # FCOMMON_OPT += -qarch=440
  812. ifdef BINARY64
  813. FCOMMON_OPT += -q64
  814. ifdef INTERFACE64
  815. ifneq ($(INTERFACE64), 0)
  816. FCOMMON_OPT += -qintsize=8
  817. endif
  818. endif
  819. else
  820. FCOMMON_OPT += -q32
  821. endif
  822. ifeq ($(USE_OPENMP), 1)
  823. FCOMMON_OPT += -openmp
  824. endif
  825. endif
  826. ifeq ($(F_COMPILER), PGI)
  827. CCOMMON_OPT += -DF_INTERFACE_PGI
  828. COMMON_PROF += -DPGICOMPILER
  829. ifdef BINARY64
  830. ifdef INTERFACE64
  831. ifneq ($(INTERFACE64), 0)
  832. FCOMMON_OPT += -i8
  833. endif
  834. endif
  835. ifeq ($(ARCH), x86_64)
  836. FCOMMON_OPT += -tp p7-64
  837. else
  838. ifeq ($(ARCH), power)
  839. ifeq ($(CORE), POWER8)
  840. FCOMMON_OPT += -tp pwr8
  841. endif
  842. ifeq ($(CORE), POWER9)
  843. FCOMMON_OPT += -tp pwr9
  844. endif
  845. endif
  846. endif
  847. else
  848. FCOMMON_OPT += -tp p7
  849. endif
  850. FCOMMON_OPT += -Mrecursive
  851. ifeq ($(USE_OPENMP), 1)
  852. FCOMMON_OPT += -mp
  853. endif
  854. endif
  855. ifeq ($(F_COMPILER), PATHSCALE)
  856. CCOMMON_OPT += -DF_INTERFACE_PATHSCALE
  857. ifdef BINARY64
  858. ifdef INTERFACE64
  859. ifneq ($(INTERFACE64), 0)
  860. FCOMMON_OPT += -i8
  861. endif
  862. endif
  863. endif
  864. ifeq ($(USE_OPENMP), 1)
  865. FCOMMON_OPT += -mp
  866. endif
  867. endif
  868. ifeq ($(F_COMPILER), OPEN64)
  869. CCOMMON_OPT += -DF_INTERFACE_OPEN64
  870. ifdef BINARY64
  871. ifdef INTERFACE64
  872. ifneq ($(INTERFACE64), 0)
  873. FCOMMON_OPT += -i8
  874. endif
  875. endif
  876. endif
  877. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  878. ifndef BINARY64
  879. FCOMMON_OPT += -n32
  880. else
  881. FCOMMON_OPT += -n64
  882. endif
  883. ifeq ($(CORE), LOONGSON3A)
  884. FCOMMON_OPT += -loongson3 -static
  885. endif
  886. ifeq ($(CORE), LOONGSON3B)
  887. FCOMMON_OPT += -loongson3 -static
  888. endif
  889. else
  890. ifndef BINARY64
  891. FCOMMON_OPT += -m32
  892. else
  893. FCOMMON_OPT += -m64
  894. endif
  895. endif
  896. ifeq ($(USE_OPENMP), 1)
  897. FEXTRALIB += -lstdc++
  898. FCOMMON_OPT += -mp
  899. endif
  900. endif
  901. ifeq ($(C_COMPILER), OPEN64)
  902. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  903. ifndef BINARY64
  904. CCOMMON_OPT += -n32
  905. else
  906. CCOMMON_OPT += -n64
  907. endif
  908. ifeq ($(CORE), LOONGSON3A)
  909. CCOMMON_OPT += -loongson3 -static
  910. endif
  911. ifeq ($(CORE), LOONGSON3B)
  912. CCOMMON_OPT += -loongson3 -static
  913. endif
  914. else
  915. ifndef BINARY64
  916. CCOMMON_OPT += -m32
  917. else
  918. CCOMMON_OPT += -m64
  919. endif
  920. endif
  921. endif
  922. ifeq ($(C_COMPILER), SUN)
  923. CCOMMON_OPT += -w
  924. ifeq ($(ARCH), x86)
  925. CCOMMON_OPT += -m32
  926. else
  927. FCOMMON_OPT += -m64
  928. endif
  929. endif
  930. ifeq ($(F_COMPILER), SUN)
  931. CCOMMON_OPT += -DF_INTERFACE_SUN
  932. ifeq ($(ARCH), x86)
  933. FCOMMON_OPT += -m32
  934. else
  935. FCOMMON_OPT += -m64
  936. endif
  937. ifeq ($(USE_OPENMP), 1)
  938. FCOMMON_OPT += -xopenmp=parallel
  939. endif
  940. endif
  941. ifeq ($(F_COMPILER), COMPAQ)
  942. CCOMMON_OPT += -DF_INTERFACE_COMPAQ
  943. ifeq ($(USE_OPENMP), 1)
  944. FCOMMON_OPT += -openmp
  945. endif
  946. endif
  947. ifdef BINARY64
  948. ifdef INTERFACE64
  949. ifneq ($(INTERFACE64), 0)
  950. CCOMMON_OPT +=
  951. #-DUSE64BITINT
  952. endif
  953. endif
  954. endif
  955. ifeq ($(NEED_PIC), 1)
  956. ifeq ($(C_COMPILER), IBM)
  957. CCOMMON_OPT += -qpic=large
  958. else
  959. CCOMMON_OPT += -fPIC
  960. endif
  961. ifeq ($(F_COMPILER), SUN)
  962. FCOMMON_OPT += -pic
  963. else
  964. FCOMMON_OPT += -fPIC
  965. endif
  966. endif
  967. ifeq ($(DYNAMIC_ARCH), 1)
  968. CCOMMON_OPT += -DDYNAMIC_ARCH
  969. endif
  970. ifeq ($(DYNAMIC_OLDER), 1)
  971. CCOMMON_OPT += -DDYNAMIC_OLDER
  972. endif
  973. ifeq ($(NO_LAPACK), 1)
  974. CCOMMON_OPT += -DNO_LAPACK
  975. #Disable LAPACK C interface
  976. NO_LAPACKE = 1
  977. endif
  978. ifeq ($(NO_LAPACKE), 1)
  979. CCOMMON_OPT += -DNO_LAPACKE
  980. endif
  981. ifeq ($(NO_AVX), 1)
  982. CCOMMON_OPT += -DNO_AVX
  983. endif
  984. ifeq ($(ARCH), x86)
  985. CCOMMON_OPT += -DNO_AVX
  986. endif
  987. ifeq ($(NO_AVX2), 1)
  988. CCOMMON_OPT += -DNO_AVX2
  989. endif
  990. ifeq ($(NO_AVX512), 1)
  991. CCOMMON_OPT += -DNO_AVX512
  992. endif
  993. ifdef SMP
  994. CCOMMON_OPT += -DSMP_SERVER
  995. ifeq ($(ARCH), mips64)
  996. ifneq ($(CORE), LOONGSON3B)
  997. USE_SIMPLE_THREADED_LEVEL3 = 1
  998. endif
  999. endif
  1000. ifeq ($(USE_OPENMP), 1)
  1001. # USE_SIMPLE_THREADED_LEVEL3 = 1
  1002. # NO_AFFINITY = 1
  1003. CCOMMON_OPT += -DUSE_OPENMP
  1004. endif
  1005. ifeq ($(BIGNUMA), 1)
  1006. CCOMMON_OPT += -DBIGNUMA
  1007. endif
  1008. endif
  1009. ifeq ($(NO_WARMUP), 1)
  1010. CCOMMON_OPT += -DNO_WARMUP
  1011. endif
  1012. ifeq ($(CONSISTENT_FPCSR), 1)
  1013. CCOMMON_OPT += -DCONSISTENT_FPCSR
  1014. endif
  1015. # Only for development
  1016. # CCOMMON_OPT += -DPARAMTEST
  1017. # CCOMMON_OPT += -DPREFETCHTEST
  1018. # CCOMMON_OPT += -DNO_SWITCHING
  1019. # USE_PAPI = 1
  1020. ifdef USE_PAPI
  1021. CCOMMON_OPT += -DUSE_PAPI
  1022. EXTRALIB += -lpapi -lperfctr
  1023. endif
  1024. ifdef DYNAMIC_THREADS
  1025. CCOMMON_OPT += -DDYNAMIC_THREADS
  1026. endif
  1027. CCOMMON_OPT += -DMAX_CPU_NUMBER=$(NUM_THREADS)
  1028. CCOMMON_OPT += -DMAX_PARALLEL_NUMBER=$(NUM_PARALLEL)
  1029. ifdef USE_SIMPLE_THREADED_LEVEL3
  1030. CCOMMON_OPT += -DUSE_SIMPLE_THREADED_LEVEL3
  1031. endif
  1032. ifeq ($(USE_TLS), 1)
  1033. CCOMMON_OPT += -DUSE_TLS
  1034. endif
  1035. ifeq ($(BUILD_HALF), 1)
  1036. CCOMMON_OPT += -DBUILD_HALF
  1037. endif
  1038. CCOMMON_OPT += -DVERSION=\"$(VERSION)\"
  1039. ifndef SYMBOLPREFIX
  1040. SYMBOLPREFIX =
  1041. endif
  1042. ifndef SYMBOLSUFFIX
  1043. SYMBOLSUFFIX =
  1044. endif
  1045. ifndef LIBNAMESUFFIX
  1046. LIBNAMEBASE = $(SYMBOLPREFIX)openblas$(SYMBOLSUFFIX)
  1047. else
  1048. LIBNAMEBASE = $(SYMBOLPREFIX)openblas$(SYMBOLSUFFIX)_$(LIBNAMESUFFIX)
  1049. endif
  1050. ifeq ($(OSNAME), CYGWIN_NT)
  1051. LIBPREFIX = cyg$(LIBNAMEBASE)
  1052. else
  1053. LIBPREFIX = lib$(LIBNAMEBASE)
  1054. endif
  1055. KERNELDIR = $(TOPDIR)/kernel/$(ARCH)
  1056. include $(TOPDIR)/Makefile.$(ARCH)
  1057. ifneq ($(C_COMPILER), PGI)
  1058. CCOMMON_OPT += -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME
  1059. endif
  1060. CCOMMON_OPT += -DASMNAME=$(FU)$(*F) -DASMFNAME=$(FU)$(*F)$(BU) -DNAME=$(*F)$(BU) -DCNAME=$(*F) -DCHAR_NAME=\"$(*F)$(BU)\" -DCHAR_CNAME=\"$(*F)\"
  1061. ifeq ($(CORE), PPC440)
  1062. CCOMMON_OPT += -DALLOC_QALLOC
  1063. endif
  1064. ifeq ($(CORE), PPC440FP2)
  1065. STATIC_ALLOCATION = 1
  1066. endif
  1067. ifneq ($(OSNAME), Linux)
  1068. NO_AFFINITY = 1
  1069. endif
  1070. ifneq ($(ARCH), x86_64)
  1071. ifneq ($(ARCH), x86)
  1072. ifneq ($(CORE), LOONGSON3B)
  1073. NO_AFFINITY = 1
  1074. endif
  1075. endif
  1076. endif
  1077. ifdef NO_AFFINITY
  1078. ifeq ($(NO_AFFINITY), 0)
  1079. override undefine NO_AFFINITY
  1080. else
  1081. CCOMMON_OPT += -DNO_AFFINITY
  1082. endif
  1083. endif
  1084. ifdef FUNCTION_PROFILE
  1085. CCOMMON_OPT += -DFUNCTION_PROFILE
  1086. endif
  1087. ifdef HUGETLB_ALLOCATION
  1088. CCOMMON_OPT += -DALLOC_HUGETLB
  1089. endif
  1090. ifdef HUGETLBFILE_ALLOCATION
  1091. CCOMMON_OPT += -DALLOC_HUGETLBFILE -DHUGETLB_FILE_NAME=$(HUGETLBFILE_ALLOCATION)
  1092. endif
  1093. ifdef STATIC_ALLOCATION
  1094. CCOMMON_OPT += -DALLOC_STATIC
  1095. endif
  1096. ifdef DEVICEDRIVER_ALLOCATION
  1097. CCOMMON_OPT += -DALLOC_DEVICEDRIVER -DDEVICEDRIVER_NAME=\"/dev/mapper\"
  1098. endif
  1099. ifdef MIXED_MEMORY_ALLOCATION
  1100. CCOMMON_OPT += -DMIXED_MEMORY_ALLOCATION
  1101. endif
  1102. ifeq ($(OSNAME), SunOS)
  1103. TAR = gtar
  1104. PATCH = gpatch
  1105. GREP = ggrep
  1106. AWK = nawk
  1107. else
  1108. TAR = tar
  1109. PATCH = patch
  1110. GREP = grep
  1111. AWK = awk
  1112. endif
  1113. ifndef MD5SUM
  1114. MD5SUM = md5sum
  1115. endif
  1116. REVISION = -r$(VERSION)
  1117. MAJOR_VERSION = $(word 1,$(subst ., ,$(VERSION)))
  1118. ifeq ($(DEBUG), 1)
  1119. COMMON_OPT += -g
  1120. endif
  1121. ifeq ($(DEBUG), 1)
  1122. FCOMMON_OPT += -g
  1123. endif
  1124. ifndef COMMON_OPT
  1125. COMMON_OPT = -O2
  1126. endif
  1127. ifndef FCOMMON_OPT
  1128. FCOMMON_OPT = -O2 -frecursive
  1129. endif
  1130. override CFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
  1131. override PFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
  1132. override FFLAGS += $(COMMON_OPT) $(FCOMMON_OPT)
  1133. override FPFLAGS += $(FCOMMON_OPT) $(COMMON_PROF)
  1134. #MAKEOVERRIDES =
  1135. ifdef NEED_PIC
  1136. ifeq (,$(findstring PIC,$(FFLAGS)))
  1137. override FFLAGS += -fPIC
  1138. endif
  1139. endif
  1140. #For LAPACK Fortran codes.
  1141. #Disable -fopenmp for LAPACK Fortran codes on Windows.
  1142. ifdef OS_WINDOWS
  1143. LAPACK_FFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FFLAGS))
  1144. LAPACK_FPFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FPFLAGS))
  1145. else
  1146. LAPACK_FFLAGS := $(FFLAGS)
  1147. LAPACK_FPFLAGS := $(FPFLAGS)
  1148. endif
  1149. LAPACK_CFLAGS = $(CFLAGS)
  1150. LAPACK_CFLAGS += -DHAVE_LAPACK_CONFIG_H
  1151. ifdef INTERFACE64
  1152. ifneq ($(INTERFACE64), 0)
  1153. LAPACK_CFLAGS += -DLAPACK_ILP64
  1154. endif
  1155. endif
  1156. ifdef OS_WINDOWS
  1157. LAPACK_CFLAGS += -DOPENBLAS_OS_WINDOWS
  1158. endif
  1159. ifeq ($(C_COMPILER), LSB)
  1160. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1161. endif
  1162. ifndef SUFFIX
  1163. SUFFIX = o
  1164. endif
  1165. ifndef PSUFFIX
  1166. PSUFFIX = po
  1167. endif
  1168. ifndef LIBSUFFIX
  1169. LIBSUFFIX = a
  1170. endif
  1171. ifneq ($(DYNAMIC_ARCH), 1)
  1172. ifndef SMP
  1173. LIBNAME = $(LIBPREFIX)_$(LIBCORE)$(REVISION).$(LIBSUFFIX)
  1174. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)$(REVISION)_p.$(LIBSUFFIX)
  1175. else
  1176. LIBNAME = $(LIBPREFIX)_$(LIBCORE)p$(REVISION).$(LIBSUFFIX)
  1177. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)p$(REVISION)_p.$(LIBSUFFIX)
  1178. endif
  1179. else
  1180. ifndef SMP
  1181. LIBNAME = $(LIBPREFIX)$(REVISION).$(LIBSUFFIX)
  1182. LIBNAME_P = $(LIBPREFIX)$(REVISION)_p.$(LIBSUFFIX)
  1183. else
  1184. LIBNAME = $(LIBPREFIX)p$(REVISION).$(LIBSUFFIX)
  1185. LIBNAME_P = $(LIBPREFIX)p$(REVISION)_p.$(LIBSUFFIX)
  1186. endif
  1187. endif
  1188. LIBDLLNAME = $(LIBPREFIX).dll
  1189. IMPLIBNAME = lib$(LIBNAMEBASE).dll.a
  1190. ifneq ($(OSNAME), AIX)
  1191. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.so)
  1192. else
  1193. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.a)
  1194. endif
  1195. LIBDYNNAME = $(LIBNAME:.$(LIBSUFFIX)=.dylib)
  1196. LIBDEFNAME = $(LIBNAME:.$(LIBSUFFIX)=.def)
  1197. LIBEXPNAME = $(LIBNAME:.$(LIBSUFFIX)=.exp)
  1198. LIBZIPNAME = $(LIBNAME:.$(LIBSUFFIX)=.zip)
  1199. LIBS = $(TOPDIR)/$(LIBNAME)
  1200. LIBS_P = $(TOPDIR)/$(LIBNAME_P)
  1201. LIB_COMPONENTS = BLAS
  1202. ifneq ($(NO_CBLAS), 1)
  1203. LIB_COMPONENTS += CBLAS
  1204. endif
  1205. ifneq ($(NO_LAPACK), 1)
  1206. LIB_COMPONENTS += LAPACK
  1207. ifneq ($(NO_LAPACKE), 1)
  1208. LIB_COMPONENTS += LAPACKE
  1209. endif
  1210. ifeq ($(BUILD_RELAPACK), 1)
  1211. LIB_COMPONENTS += ReLAPACK
  1212. endif
  1213. endif
  1214. ifeq ($(ONLY_CBLAS), 1)
  1215. LIB_COMPONENTS = CBLAS
  1216. endif
  1217. export OSNAME
  1218. export ARCH
  1219. export CORE
  1220. export LIBCORE
  1221. export __BYTE_ORDER__
  1222. export ELF_VERSION
  1223. export PGCPATH
  1224. export CONFIG
  1225. export CC
  1226. export FC
  1227. export BU
  1228. export FU
  1229. export NEED2UNDERSCORES
  1230. export USE_THREAD
  1231. export NUM_THREADS
  1232. export NUM_CORES
  1233. export SMP
  1234. export MAKEFILE_RULE
  1235. export NEED_PIC
  1236. export BINARY
  1237. export BINARY32
  1238. export BINARY64
  1239. export F_COMPILER
  1240. export C_COMPILER
  1241. export USE_OPENMP
  1242. export CROSS
  1243. export CROSS_SUFFIX
  1244. export NOFORTRAN
  1245. export NO_FBLAS
  1246. export EXTRALIB
  1247. export CEXTRALIB
  1248. export FEXTRALIB
  1249. export HAVE_SSE
  1250. export HAVE_SSE2
  1251. export HAVE_SSE3
  1252. export HAVE_SSSE3
  1253. export HAVE_SSE4_1
  1254. export HAVE_SSE4_2
  1255. export HAVE_SSE4A
  1256. export HAVE_SSE5
  1257. export HAVE_AVX
  1258. export HAVE_VFP
  1259. export HAVE_VFPV3
  1260. export HAVE_VFPV4
  1261. export HAVE_NEON
  1262. export HAVE_MSA
  1263. export MSA_FLAGS
  1264. export KERNELDIR
  1265. export FUNCTION_PROFILE
  1266. export TARGET_CORE
  1267. export NO_AVX512
  1268. export BUILD_HALF
  1269. export SHGEMM_UNROLL_M
  1270. export SHGEMM_UNROLL_N
  1271. export SGEMM_UNROLL_M
  1272. export SGEMM_UNROLL_N
  1273. export DGEMM_UNROLL_M
  1274. export DGEMM_UNROLL_N
  1275. export QGEMM_UNROLL_M
  1276. export QGEMM_UNROLL_N
  1277. export CGEMM_UNROLL_M
  1278. export CGEMM_UNROLL_N
  1279. export ZGEMM_UNROLL_M
  1280. export ZGEMM_UNROLL_N
  1281. export XGEMM_UNROLL_M
  1282. export XGEMM_UNROLL_N
  1283. export CGEMM3M_UNROLL_M
  1284. export CGEMM3M_UNROLL_N
  1285. export ZGEMM3M_UNROLL_M
  1286. export ZGEMM3M_UNROLL_N
  1287. export XGEMM3M_UNROLL_M
  1288. export XGEMM3M_UNROLL_N
  1289. ifdef USE_CUDA
  1290. export CUDADIR
  1291. export CUCC
  1292. export CUFLAGS
  1293. export CULIB
  1294. endif
  1295. .SUFFIXES: .$(PSUFFIX) .$(SUFFIX) .f
  1296. .f.$(SUFFIX):
  1297. $(FC) $(FFLAGS) -c $< -o $(@F)
  1298. .f.$(PSUFFIX):
  1299. $(FC) $(FPFLAGS) -pg -c $< -o $(@F)
  1300. ifdef BINARY64
  1301. PATHSCALEPATH = /opt/pathscale/lib/3.1
  1302. PGIPATH = /opt/pgi/linux86-64/7.1-5/lib
  1303. else
  1304. PATHSCALEPATH = /opt/pathscale/lib/3.1/32
  1305. PGIPATH = /opt/pgi/linux86/7.1-5/lib
  1306. endif
  1307. ACMLPATH = /opt/acml/4.3.0
  1308. ifneq ($(OSNAME), Darwin)
  1309. MKLPATH = /opt/intel/mkl/10.2.2.025/lib
  1310. else
  1311. MKLPATH = /Library/Frameworks/Intel_MKL.framework/Versions/10.0.1.014/lib
  1312. endif
  1313. ATLASPATH = /opt/atlas/3.9.17/opteron
  1314. FLAMEPATH = $(HOME)/flame/lib
  1315. ifneq ($(OSNAME), SunOS)
  1316. SUNPATH = /opt/sunstudio12.1
  1317. else
  1318. SUNPATH = /opt/SUNWspro
  1319. endif