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 29 kB

7 years ago
13 years ago
1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285128612871288128912901291129212931294129512961297129812991300130113021303130413051306130713081309131013111312131313141315131613171318131913201321132213231324132513261327132813291330133113321333133413351336133713381339134013411342134313441345134613471348134913501351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384138513861387138813891390139113921393139413951396139713981399140014011402140314041405140614071408140914101411141214131414141514161417141814191420142114221423142414251426142714281429143014311432143314341435143614371438143914401441144214431444144514461447144814491450145114521453145414551456145714581459146014611462146314641465146614671468146914701471147214731474147514761477147814791480148114821483148414851486148714881489149014911492149314941495149614971498149915001501150215031504150515061507150815091510151115121513151415151516
  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. endif
  98. #TARGET_CORE will override TARGET which is used in DYNAMIC_ARCH=1.
  99. #
  100. ifdef TARGET_CORE
  101. GETARCH_FLAGS := -DFORCE_$(TARGET_CORE)
  102. endif
  103. # Force fallbacks for 32bit
  104. ifeq ($(BINARY), 32)
  105. ifeq ($(TARGET_CORE), HASWELL)
  106. GETARCH_FLAGS := -DFORCE_NEHALEM
  107. endif
  108. ifeq ($(TARGET_CORE), SKYLAKEX)
  109. GETARCH_FLAGS := -DFORCE_NEHALEM
  110. endif
  111. ifeq ($(TARGET_CORE), SANDYBRIDGE)
  112. GETARCH_FLAGS := -DFORCE_NEHALEM
  113. endif
  114. ifeq ($(TARGET_CORE), BULLDOZER)
  115. GETARCH_FLAGS := -DFORCE_BARCELONA
  116. endif
  117. ifeq ($(TARGET_CORE), PILEDRIVER)
  118. GETARCH_FLAGS := -DFORCE_BARCELONA
  119. endif
  120. ifeq ($(TARGET_CORE), STEAMROLLER)
  121. GETARCH_FLAGS := -DFORCE_BARCELONA
  122. endif
  123. ifeq ($(TARGET_CORE), EXCAVATOR)
  124. GETARCH_FLAGS := -DFORCE_BARCELONA
  125. endif
  126. ifeq ($(TARGET_CORE), ZEN)
  127. GETARCH_FLAGS := -DFORCE_BARCELONA
  128. endif
  129. endif
  130. # On x86_64 build getarch with march=native unless the compiler is PGI. This is required to detect AVX512 support in getarch.
  131. ifeq ($(HOSTARCH), x86_64)
  132. ifeq ($(findstring pgcc,$(HOSTCC)),)
  133. GETARCH_FLAGS += -march=native
  134. endif
  135. endif
  136. ifdef INTERFACE64
  137. ifneq ($(INTERFACE64), 0)
  138. GETARCH_FLAGS += -DUSE64BITINT
  139. endif
  140. endif
  141. ifndef GEMM_MULTITHREAD_THRESHOLD
  142. GEMM_MULTITHREAD_THRESHOLD=4
  143. endif
  144. GETARCH_FLAGS += -DGEMM_MULTITHREAD_THRESHOLD=$(GEMM_MULTITHREAD_THRESHOLD)
  145. ifeq ($(NO_AVX), 1)
  146. GETARCH_FLAGS += -DNO_AVX
  147. endif
  148. ifeq ($(BINARY), 32)
  149. GETARCH_FLAGS += -DNO_AVX -DNO_AVX2 -DNO_AVX512
  150. NO_AVX512 = 1
  151. endif
  152. ifeq ($(NO_AVX2), 1)
  153. GETARCH_FLAGS += -DNO_AVX2
  154. endif
  155. ifeq ($(NO_AVX512), 1)
  156. GETARCH_FLAGS += -DNO_AVX512
  157. endif
  158. ifeq ($(DEBUG), 1)
  159. GETARCH_FLAGS += -g
  160. endif
  161. ifeq ($(QUIET_MAKE), 1)
  162. MAKE += -s
  163. endif
  164. ifndef NO_PARALLEL_MAKE
  165. NO_PARALLEL_MAKE=0
  166. endif
  167. GETARCH_FLAGS += -DNO_PARALLEL_MAKE=$(NO_PARALLEL_MAKE)
  168. ifdef MAKE_NB_JOBS
  169. GETARCH_FLAGS += -DMAKE_NB_JOBS=$(MAKE_NB_JOBS)
  170. endif
  171. ifeq ($(HOSTCC), loongcc)
  172. GETARCH_FLAGS += -static
  173. endif
  174. #if don't use Fortran, it will only compile CBLAS.
  175. ifeq ($(ONLY_CBLAS), 1)
  176. NO_LAPACK = 1
  177. else
  178. ONLY_CBLAS = 0
  179. endif
  180. # This operation is expensive, so execution should be once.
  181. ifndef GOTOBLAS_MAKEFILE
  182. export GOTOBLAS_MAKEFILE = 1
  183. # Generating Makefile.conf and config.h
  184. 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)
  185. ifndef TARGET_CORE
  186. include $(TOPDIR)/Makefile.conf
  187. else
  188. include $(TOPDIR)/Makefile_kernel.conf
  189. endif
  190. endif
  191. ifndef NUM_PARALLEL
  192. NUM_PARALLEL = 1
  193. endif
  194. ifndef NUM_THREADS
  195. NUM_THREADS = $(NUM_CORES)
  196. endif
  197. ifeq ($(NUM_THREADS), 1)
  198. override USE_THREAD = 0
  199. override USE_OPENMP = 0
  200. endif
  201. ifdef USE_THREAD
  202. ifeq ($(USE_THREAD), 0)
  203. SMP =
  204. else
  205. SMP = 1
  206. endif
  207. else
  208. ifeq ($(NUM_THREAD), 1)
  209. SMP =
  210. else
  211. SMP = 1
  212. endif
  213. endif
  214. ifeq ($(SMP), 1)
  215. USE_LOCKING =
  216. endif
  217. ifndef NEED_PIC
  218. NEED_PIC = 1
  219. endif
  220. ARFLAGS =
  221. CPP = $(COMPILER) -E
  222. AR ?= $(CROSS_SUFFIX)ar
  223. AS ?= $(CROSS_SUFFIX)as
  224. LD ?= $(CROSS_SUFFIX)ld
  225. RANLIB ?= $(CROSS_SUFFIX)ranlib
  226. NM = $(CROSS_SUFFIX)nm
  227. DLLWRAP = $(CROSS_SUFFIX)dllwrap
  228. OBJCOPY = $(CROSS_SUFFIX)objcopy
  229. OBJCONV = $(CROSS_SUFFIX)objconv
  230. # When fortran support was either not detected or actively deselected, only build BLAS.
  231. ifeq ($(NOFORTRAN), 1)
  232. NO_LAPACK = 1
  233. override FEXTRALIB =
  234. endif
  235. ifeq ($(C_COMPILER), GCC)
  236. GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
  237. GCCVERSIONGT4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 4)
  238. GCCVERSIONEQ5 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` = 5)
  239. GCCVERSIONGT5 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 5)
  240. GCCVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 7)
  241. GCCVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  242. # Note that the behavior of -dumpversion is compile-time-configurable for
  243. # gcc-7.x and newer. Use -dumpfullversion there
  244. ifeq ($(GCCVERSIONGTEQ7),1)
  245. GCCDUMPVERSION_PARAM := -dumpfullversion
  246. else
  247. GCCDUMPVERSION_PARAM := -dumpversion
  248. endif
  249. GCCMINORVERSIONGTEQ2 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 2)
  250. GCCMINORVERSIONGTEQ7 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 7)
  251. endif
  252. #
  253. # OS dependent settings
  254. #
  255. ifeq ($(OSNAME), Darwin)
  256. ifndef MACOSX_DEPLOYMENT_TARGET
  257. export MACOSX_DEPLOYMENT_TARGET=10.8
  258. endif
  259. MD5SUM = md5 -r
  260. endif
  261. ifneq (,$(findstring $(OSNAME), FreeBSD OpenBSD DragonFly))
  262. MD5SUM = md5 -r
  263. endif
  264. ifeq ($(OSNAME), NetBSD)
  265. MD5SUM = md5 -n
  266. endif
  267. ifeq ($(OSNAME), Linux)
  268. EXTRALIB += -lm
  269. NO_EXPRECISION = 1
  270. endif
  271. ifeq ($(OSNAME), Android)
  272. EXTRALIB += -lm
  273. endif
  274. ifeq ($(OSNAME), AIX)
  275. EXTRALIB += -lm
  276. endif
  277. ifeq ($(OSNAME), WINNT)
  278. NEED_PIC = 0
  279. NO_EXPRECISION = 1
  280. EXTRALIB += -defaultlib:advapi32
  281. SUFFIX = obj
  282. PSUFFIX = pobj
  283. LIBSUFFIX = a
  284. ifeq ($(C_COMPILER), CLANG)
  285. CCOMMON_OPT += -DMS_ABI
  286. endif
  287. #Version tests for supporting specific features (MS_ABI, POWER9 intrinsics)
  288. ifeq ($(GCCVERSIONGT4), 1)
  289. # GCC Major version > 4
  290. # It is compatible with MSVC ABI.
  291. CCOMMON_OPT += -DMS_ABI
  292. endif
  293. ifeq ($(GCCVERSIONGTEQ4), 1)
  294. ifeq ($(GCCMINORVERSIONGTEQ7), 1)
  295. # GCC Version >=4.7
  296. # It is compatible with MSVC ABI.
  297. CCOMMON_OPT += -DMS_ABI
  298. endif
  299. endif
  300. # Ensure the correct stack alignment on Win32
  301. # http://permalink.gmane.org/gmane.comp.lib.openblas.general/97
  302. ifeq ($(ARCH), x86)
  303. CCOMMON_OPT += -mincoming-stack-boundary=2
  304. FCOMMON_OPT += -mincoming-stack-boundary=2
  305. endif
  306. endif
  307. ifeq ($(OSNAME), Interix)
  308. NEED_PIC = 0
  309. NO_EXPRECISION = 1
  310. INTERIX_TOOL_DIR = /opt/gcc.3.3/i586-pc-interix3/bin
  311. endif
  312. ifeq ($(OSNAME), CYGWIN_NT)
  313. NEED_PIC = 0
  314. NO_EXPRECISION = 1
  315. OS_CYGWIN_NT = 1
  316. endif
  317. ifneq ($(OSNAME), WINNT)
  318. ifneq ($(OSNAME), CYGWIN_NT)
  319. ifneq ($(OSNAME), Interix)
  320. ifneq ($(OSNAME), Android)
  321. ifdef SMP
  322. EXTRALIB += -lpthread
  323. endif
  324. endif
  325. endif
  326. endif
  327. endif
  328. # ifeq logical or
  329. ifeq ($(OSNAME), $(filter $(OSNAME),WINNT CYGWIN_NT Interix))
  330. OS_WINDOWS=1
  331. endif
  332. ifdef QUAD_PRECISION
  333. CCOMMON_OPT += -DQUAD_PRECISION
  334. NO_EXPRECISION = 1
  335. endif
  336. ifneq ($(ARCH), x86)
  337. ifneq ($(ARCH), x86_64)
  338. NO_EXPRECISION = 1
  339. endif
  340. endif
  341. ifdef UTEST_CHECK
  342. CCOMMON_OPT += -DUTEST_CHECK
  343. SANITY_CHECK = 1
  344. endif
  345. ifdef SANITY_CHECK
  346. CCOMMON_OPT += -DSANITY_CHECK -DREFNAME=$(*F)f$(BU)
  347. endif
  348. MAX_STACK_ALLOC ?= 2048
  349. ifneq ($(MAX_STACK_ALLOC), 0)
  350. CCOMMON_OPT += -DMAX_STACK_ALLOC=$(MAX_STACK_ALLOC)
  351. endif
  352. ifdef USE_LOCKING
  353. ifneq ($(USE_LOCKING), 0)
  354. CCOMMON_OPT += -DUSE_LOCKING
  355. endif
  356. endif
  357. #
  358. # Architecture dependent settings
  359. #
  360. ifeq ($(ARCH), x86)
  361. ifndef BINARY
  362. NO_BINARY_MODE = 1
  363. endif
  364. ifeq ($(CORE), generic)
  365. NO_EXPRECISION = 1
  366. endif
  367. ifndef NO_EXPRECISION
  368. ifeq ($(F_COMPILER), GFORTRAN)
  369. # ifeq logical or. GCC or LSB
  370. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  371. EXPRECISION = 1
  372. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  373. FCOMMON_OPT += -m128bit-long-double
  374. endif
  375. ifeq ($(C_COMPILER), CLANG)
  376. EXPRECISION = 1
  377. CCOMMON_OPT += -DEXPRECISION
  378. FCOMMON_OPT += -m128bit-long-double
  379. endif
  380. endif
  381. endif
  382. endif
  383. ifeq ($(ARCH), x86_64)
  384. ifeq ($(CORE), generic)
  385. NO_EXPRECISION = 1
  386. endif
  387. ifndef NO_EXPRECISION
  388. ifeq ($(F_COMPILER), GFORTRAN)
  389. # ifeq logical or. GCC or LSB
  390. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  391. EXPRECISION = 1
  392. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  393. FCOMMON_OPT += -m128bit-long-double
  394. endif
  395. ifeq ($(C_COMPILER), CLANG)
  396. EXPRECISION = 1
  397. CCOMMON_OPT += -DEXPRECISION
  398. FCOMMON_OPT += -m128bit-long-double
  399. endif
  400. endif
  401. endif
  402. endif
  403. ifeq ($(C_COMPILER), INTEL)
  404. CCOMMON_OPT += -wd981
  405. endif
  406. ifeq ($(USE_OPENMP), 1)
  407. #check
  408. ifeq ($(USE_THREAD), 0)
  409. $(error OpenBLAS: Cannot set both USE_OPENMP=1 and USE_THREAD=0. The USE_THREAD=0 is only for building single thread version.)
  410. endif
  411. # ifeq logical or. GCC or LSB
  412. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  413. CCOMMON_OPT += -fopenmp
  414. endif
  415. ifeq ($(C_COMPILER), CLANG)
  416. CCOMMON_OPT += -fopenmp
  417. endif
  418. ifeq ($(C_COMPILER), INTEL)
  419. CCOMMON_OPT += -fopenmp
  420. endif
  421. ifeq ($(C_COMPILER), PGI)
  422. CCOMMON_OPT += -mp
  423. endif
  424. ifeq ($(C_COMPILER), OPEN64)
  425. CCOMMON_OPT += -mp
  426. CEXTRALIB += -lstdc++
  427. endif
  428. ifeq ($(C_COMPILER), PATHSCALE)
  429. CCOMMON_OPT += -mp
  430. endif
  431. endif
  432. ifeq ($(DYNAMIC_ARCH), 1)
  433. ifeq ($(ARCH), x86)
  434. DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
  435. CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
  436. endif
  437. ifeq ($(ARCH), x86_64)
  438. DYNAMIC_CORE = PRESCOTT CORE2
  439. ifeq ($(DYNAMIC_OLDER), 1)
  440. DYNAMIC_CORE += PENRYN DUNNINGTON
  441. endif
  442. DYNAMIC_CORE += NEHALEM
  443. ifeq ($(DYNAMIC_OLDER), 1)
  444. DYNAMIC_CORE += OPTERON OPTERON_SSE3
  445. endif
  446. DYNAMIC_CORE += BARCELONA
  447. ifeq ($(DYNAMIC_OLDER), 1)
  448. DYNAMIC_CORE += BOBCAT ATOM NANO
  449. endif
  450. ifneq ($(NO_AVX), 1)
  451. DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER STEAMROLLER EXCAVATOR
  452. endif
  453. ifneq ($(NO_AVX2), 1)
  454. DYNAMIC_CORE += HASWELL ZEN
  455. endif
  456. ifneq ($(NO_AVX512), 1)
  457. ifneq ($(NO_AVX2), 1)
  458. DYNAMIC_CORE += SKYLAKEX
  459. endif
  460. endif
  461. endif
  462. ifdef DYNAMIC_LIST
  463. override DYNAMIC_CORE = PRESCOTT $(DYNAMIC_LIST)
  464. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_PRESCOTT
  465. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  466. CCOMMON_OPT += $(XCCOMMON_OPT)
  467. #CCOMMON_OPT += -DDYNAMIC_LIST='$(DYNAMIC_LIST)'
  468. endif
  469. ifeq ($(ARCH), arm64)
  470. DYNAMIC_CORE = ARMV8
  471. DYNAMIC_CORE += CORTEXA53
  472. DYNAMIC_CORE += CORTEXA57
  473. DYNAMIC_CORE += CORTEXA72
  474. DYNAMIC_CORE += CORTEXA73
  475. DYNAMIC_CORE += NEOVERSEN1
  476. DYNAMIC_CORE += FALKOR
  477. DYNAMIC_CORE += THUNDERX
  478. DYNAMIC_CORE += THUNDERX2T99
  479. DYNAMIC_CORE += TSV110
  480. DYNAMIC_CORE += EMAG8180
  481. endif
  482. ifeq ($(ARCH), zarch)
  483. DYNAMIC_CORE = ZARCH_GENERIC
  484. # Z13 is supported since gcc-5.2, gcc-6, and in RHEL 7.3 and newer
  485. ifeq ($(GCCVERSIONGT5), 1)
  486. ZARCH_SUPPORT_Z13 := 1
  487. else ifeq ($(GCCVERSIONEQ5), 1)
  488. ifeq ($(GCCMINORVERSIONGTEQ2), 1)
  489. ZARCH_SUPPORT_Z13 := 1
  490. endif
  491. endif
  492. ifeq ($(wildcard /etc/redhat-release), /etc/redhat-release)
  493. ifeq ($(shell source /etc/os-release ; expr $$VERSION_ID \>= "7.3"), 1)
  494. ZARCH_SUPPORT_Z13 := 1
  495. endif
  496. endif
  497. ifeq ($(ZARCH_SUPPORT_Z13), 1)
  498. DYNAMIC_CORE += Z13
  499. else
  500. $(info OpenBLAS: Not building Z13 kernels because gcc is older than 5.2 or 6.x)
  501. endif
  502. ifeq ($(GCCVERSIONGTEQ7), 1)
  503. DYNAMIC_CORE += Z14
  504. else
  505. $(info OpenBLAS: Not building Z14 kernels because gcc is older than 7.x)
  506. endif
  507. endif
  508. ifeq ($(ARCH), power)
  509. DYNAMIC_CORE = POWER6
  510. DYNAMIC_CORE += POWER8
  511. ifneq ($(C_COMPILER), GCC)
  512. DYNAMIC_CORE += POWER9
  513. DYNAMIC_CORE += POWER10
  514. endif
  515. ifeq ($(C_COMPILER), GCC)
  516. ifeq ($(GCCVERSIONGT5), 1)
  517. DYNAMIC_CORE += POWER9
  518. else
  519. $(info, OpenBLAS: Your gcc version is too old to build the POWER9 kernels.)
  520. endif
  521. GCCVERSIONGTEQ11 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 11)
  522. ifeq ($(GCCVERSIONGTEQ11), 1)
  523. DYNAMIC_CORE += POWER10
  524. else
  525. $(info, OpenBLAS: Your gcc version is too old to build the POWER10 kernels.)
  526. endif
  527. endif
  528. endif
  529. # If DYNAMIC_CORE is not set, DYNAMIC_ARCH cannot do anything, so force it to empty
  530. ifndef DYNAMIC_CORE
  531. override DYNAMIC_ARCH=
  532. endif
  533. endif
  534. ifeq ($(ARCH), ia64)
  535. NO_BINARY_MODE = 1
  536. BINARY_DEFINED = 1
  537. ifeq ($(F_COMPILER), GFORTRAN)
  538. ifeq ($(C_COMPILER), GCC)
  539. # EXPRECISION = 1
  540. # CCOMMON_OPT += -DEXPRECISION
  541. endif
  542. endif
  543. endif
  544. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  545. NO_BINARY_MODE = 1
  546. endif
  547. ifeq ($(ARCH), alpha)
  548. NO_BINARY_MODE = 1
  549. BINARY_DEFINED = 1
  550. endif
  551. ifeq ($(ARCH), arm)
  552. NO_BINARY_MODE = 1
  553. BINARY_DEFINED = 1
  554. CCOMMON_OPT += -marm
  555. FCOMMON_OPT += -marm
  556. # If softfp abi is mentioned on the command line, force it.
  557. ifeq ($(ARM_SOFTFP_ABI), 1)
  558. CCOMMON_OPT += -mfloat-abi=softfp
  559. FCOMMON_OPT += -mfloat-abi=softfp
  560. endif
  561. ifeq ($(OSNAME), Android)
  562. ifeq ($(ARM_SOFTFP_ABI), 1)
  563. EXTRALIB += -lm
  564. else
  565. EXTRALIB += -Wl,-lm_hard
  566. endif
  567. endif
  568. endif
  569. ifeq ($(ARCH), arm64)
  570. NO_BINARY_MODE = 1
  571. BINARY_DEFINED = 1
  572. ifdef INTERFACE64
  573. ifneq ($(INTERFACE64), 0)
  574. ifeq ($(F_COMPILER), GFORTRAN)
  575. FCOMMON_OPT += -fdefault-integer-8
  576. endif
  577. ifeq ($(F_COMPILER), FLANG)
  578. FCOMMON_OPT += -i8
  579. endif
  580. endif
  581. endif
  582. endif
  583. #
  584. # C Compiler dependent settings
  585. #
  586. # ifeq logical or. GCC or CLANG or LSB
  587. # http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
  588. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG LSB))
  589. CCOMMON_OPT += -Wall
  590. COMMON_PROF += -fno-inline
  591. NO_UNINITIALIZED_WARN = -Wno-uninitialized
  592. ifeq ($(QUIET_MAKE), 1)
  593. CCOMMON_OPT += $(NO_UNINITIALIZED_WARN) -Wno-unused
  594. endif
  595. ifdef NO_BINARY_MODE
  596. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  597. ifdef BINARY64
  598. CCOMMON_OPT += -mabi=64
  599. else
  600. CCOMMON_OPT += -mabi=n32
  601. endif
  602. BINARY_DEFINED = 1
  603. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  604. CCOMMON_OPT += -mabi=32
  605. BINARY_DEFINED = 1
  606. endif
  607. ifeq ($(CORE), LOONGSON3A)
  608. CCOMMON_OPT += -march=mips64
  609. FCOMMON_OPT += -march=mips64
  610. endif
  611. ifeq ($(CORE), LOONGSON3B)
  612. CCOMMON_OPT += -march=mips64
  613. FCOMMON_OPT += -march=mips64
  614. endif
  615. ifeq ($(CORE), MIPS24K)
  616. CCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  617. FCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  618. endif
  619. ifeq ($(CORE), MIPS1004K)
  620. CCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  621. FCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  622. endif
  623. ifeq ($(CORE), P5600)
  624. CCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  625. FCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  626. endif
  627. ifeq ($(CORE), I6400)
  628. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  629. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  630. endif
  631. ifeq ($(CORE), P6600)
  632. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  633. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  634. endif
  635. ifeq ($(CORE), I6500)
  636. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  637. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  638. endif
  639. ifeq ($(OSNAME), AIX)
  640. BINARY_DEFINED = 1
  641. endif
  642. endif
  643. ifndef BINARY_DEFINED
  644. ifneq ($(OSNAME), AIX)
  645. ifdef BINARY64
  646. CCOMMON_OPT += -m64
  647. else
  648. CCOMMON_OPT += -m32
  649. endif
  650. endif
  651. endif
  652. endif
  653. ifeq ($(C_COMPILER), PGI)
  654. ifdef BINARY64
  655. CCOMMON_OPT += -tp p7-64 -D__MMX__ -Mnollvm
  656. else
  657. CCOMMON_OPT += -tp p7
  658. endif
  659. endif
  660. ifeq ($(C_COMPILER), PATHSCALE)
  661. ifdef BINARY64
  662. CCOMMON_OPT += -m64
  663. else
  664. CCOMMON_OPT += -m32
  665. endif
  666. endif
  667. #
  668. # Fortran Compiler dependent settings
  669. #
  670. ifeq ($(F_COMPILER), FLANG)
  671. CCOMMON_OPT += -DF_INTERFACE_FLANG
  672. FCOMMON_OPT += -Mrecursive -Kieee
  673. ifeq ($(OSNAME), Linux)
  674. ifeq ($(ARCH), x86_64)
  675. FLANG_VENDOR := $(shell expr `$(FC) --version|cut -f 1 -d "."|head -1`)
  676. ifeq ($(FLANG_VENDOR),AOCC)
  677. FCOMMON_OPT += -fno-unroll-loops
  678. endif
  679. endif
  680. endif
  681. ifdef BINARY64
  682. ifdef INTERFACE64
  683. ifneq ($(INTERFACE64), 0)
  684. FCOMMON_OPT += -i8
  685. endif
  686. endif
  687. FCOMMON_OPT += -Wall
  688. else
  689. FCOMMON_OPT += -Wall
  690. endif
  691. ifeq ($(USE_OPENMP), 1)
  692. FCOMMON_OPT += -fopenmp
  693. endif
  694. endif
  695. ifeq ($(F_COMPILER), G77)
  696. CCOMMON_OPT += -DF_INTERFACE_G77
  697. FCOMMON_OPT += -Wall
  698. ifndef NO_BINARY_MODE
  699. ifneq ($(OSNAME), AIX)
  700. ifdef BINARY64
  701. FCOMMON_OPT += -m64
  702. else
  703. FCOMMON_OPT += -m32
  704. endif
  705. endif
  706. endif
  707. endif
  708. ifeq ($(F_COMPILER), G95)
  709. CCOMMON_OPT += -DF_INTERFACE_G95
  710. FCOMMON_OPT += -Wall
  711. ifneq ($(OSNAME), AIX)
  712. ifndef NO_BINARY_MODE
  713. ifdef BINARY64
  714. FCOMMON_OPT += -m64
  715. else
  716. FCOMMON_OPT += -m32
  717. endif
  718. endif
  719. ifneq ($(NO_LAPACKE), 1)
  720. FCOMMON_OPT += -fno-second-underscore
  721. endif
  722. endif
  723. endif
  724. ifeq ($(F_COMPILER), GFORTRAN)
  725. CCOMMON_OPT += -DF_INTERFACE_GFORT
  726. FCOMMON_OPT += -Wall
  727. # make single-threaded LAPACK calls thread-safe #1847
  728. FCOMMON_OPT += -frecursive
  729. # work around ABI problem with passing single-character arguments
  730. FCOMMON_OPT += -fno-optimize-sibling-calls
  731. #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
  732. ifneq ($(NO_LAPACK), 1)
  733. EXTRALIB += -lgfortran
  734. endif
  735. ifdef NO_BINARY_MODE
  736. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  737. ifdef BINARY64
  738. FCOMMON_OPT += -mabi=64
  739. else
  740. FCOMMON_OPT += -mabi=n32
  741. endif
  742. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  743. FCOMMON_OPT += -mabi=32
  744. endif
  745. else
  746. ifdef BINARY64
  747. ifneq ($(OSNAME), AIX)
  748. FCOMMON_OPT += -m64
  749. endif
  750. ifdef INTERFACE64
  751. ifneq ($(INTERFACE64), 0)
  752. FCOMMON_OPT += -fdefault-integer-8
  753. endif
  754. endif
  755. else
  756. ifneq ($(OSNAME), AIX)
  757. FCOMMON_OPT += -m32
  758. endif
  759. endif
  760. endif
  761. ifeq ($(USE_OPENMP), 1)
  762. FCOMMON_OPT += -fopenmp
  763. endif
  764. endif
  765. ifeq ($(F_COMPILER), INTEL)
  766. CCOMMON_OPT += -DF_INTERFACE_INTEL
  767. ifdef INTERFACE64
  768. ifneq ($(INTERFACE64), 0)
  769. FCOMMON_OPT += -i8
  770. endif
  771. endif
  772. FCOMMON_OPT += -recursive -fp-model strict -assume protect-parens
  773. ifeq ($(USE_OPENMP), 1)
  774. FCOMMON_OPT += -fopenmp
  775. endif
  776. endif
  777. ifeq ($(F_COMPILER), FUJITSU)
  778. CCOMMON_OPT += -DF_INTERFACE_FUJITSU
  779. ifeq ($(USE_OPENMP), 1)
  780. FCOMMON_OPT += -openmp
  781. endif
  782. endif
  783. ifeq ($(F_COMPILER), IBM)
  784. CCOMMON_OPT += -DF_INTERFACE_IBM
  785. # FCOMMON_OPT += -qarch=440
  786. ifdef BINARY64
  787. FCOMMON_OPT += -q64
  788. ifdef INTERFACE64
  789. ifneq ($(INTERFACE64), 0)
  790. FCOMMON_OPT += -qintsize=8
  791. endif
  792. endif
  793. else
  794. FCOMMON_OPT += -q32
  795. endif
  796. ifeq ($(USE_OPENMP), 1)
  797. FCOMMON_OPT += -openmp
  798. endif
  799. endif
  800. ifeq ($(F_COMPILER), PGI)
  801. CCOMMON_OPT += -DF_INTERFACE_PGI
  802. COMMON_PROF += -DPGICOMPILER
  803. ifdef BINARY64
  804. ifdef INTERFACE64
  805. ifneq ($(INTERFACE64), 0)
  806. FCOMMON_OPT += -i8
  807. endif
  808. endif
  809. FCOMMON_OPT += -tp p7-64
  810. else
  811. FCOMMON_OPT += -tp p7
  812. endif
  813. FCOMMON_OPT += -Mrecursive
  814. ifeq ($(USE_OPENMP), 1)
  815. FCOMMON_OPT += -mp
  816. endif
  817. endif
  818. ifeq ($(F_COMPILER), PATHSCALE)
  819. CCOMMON_OPT += -DF_INTERFACE_PATHSCALE
  820. ifdef BINARY64
  821. ifdef INTERFACE64
  822. ifneq ($(INTERFACE64), 0)
  823. FCOMMON_OPT += -i8
  824. endif
  825. endif
  826. endif
  827. ifeq ($(USE_OPENMP), 1)
  828. FCOMMON_OPT += -mp
  829. endif
  830. endif
  831. ifeq ($(F_COMPILER), OPEN64)
  832. CCOMMON_OPT += -DF_INTERFACE_OPEN64
  833. ifdef BINARY64
  834. ifdef INTERFACE64
  835. ifneq ($(INTERFACE64), 0)
  836. FCOMMON_OPT += -i8
  837. endif
  838. endif
  839. endif
  840. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  841. ifndef BINARY64
  842. FCOMMON_OPT += -n32
  843. else
  844. FCOMMON_OPT += -n64
  845. endif
  846. ifeq ($(CORE), LOONGSON3A)
  847. FCOMMON_OPT += -loongson3 -static
  848. endif
  849. ifeq ($(CORE), LOONGSON3B)
  850. FCOMMON_OPT += -loongson3 -static
  851. endif
  852. else
  853. ifndef BINARY64
  854. FCOMMON_OPT += -m32
  855. else
  856. FCOMMON_OPT += -m64
  857. endif
  858. endif
  859. ifeq ($(USE_OPENMP), 1)
  860. FEXTRALIB += -lstdc++
  861. FCOMMON_OPT += -mp
  862. endif
  863. endif
  864. ifeq ($(C_COMPILER), OPEN64)
  865. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  866. ifndef BINARY64
  867. CCOMMON_OPT += -n32
  868. else
  869. CCOMMON_OPT += -n64
  870. endif
  871. ifeq ($(CORE), LOONGSON3A)
  872. CCOMMON_OPT += -loongson3 -static
  873. endif
  874. ifeq ($(CORE), LOONGSON3B)
  875. CCOMMON_OPT += -loongson3 -static
  876. endif
  877. else
  878. ifndef BINARY64
  879. CCOMMON_OPT += -m32
  880. else
  881. CCOMMON_OPT += -m64
  882. endif
  883. endif
  884. endif
  885. ifeq ($(C_COMPILER), SUN)
  886. CCOMMON_OPT += -w
  887. ifeq ($(ARCH), x86)
  888. CCOMMON_OPT += -m32
  889. else
  890. FCOMMON_OPT += -m64
  891. endif
  892. endif
  893. ifeq ($(F_COMPILER), SUN)
  894. CCOMMON_OPT += -DF_INTERFACE_SUN
  895. ifeq ($(ARCH), x86)
  896. FCOMMON_OPT += -m32
  897. else
  898. FCOMMON_OPT += -m64
  899. endif
  900. ifeq ($(USE_OPENMP), 1)
  901. FCOMMON_OPT += -xopenmp=parallel
  902. endif
  903. endif
  904. ifeq ($(F_COMPILER), COMPAQ)
  905. CCOMMON_OPT += -DF_INTERFACE_COMPAQ
  906. ifeq ($(USE_OPENMP), 1)
  907. FCOMMON_OPT += -openmp
  908. endif
  909. endif
  910. ifdef BINARY64
  911. ifdef INTERFACE64
  912. ifneq ($(INTERFACE64), 0)
  913. CCOMMON_OPT +=
  914. #-DUSE64BITINT
  915. endif
  916. endif
  917. endif
  918. ifeq ($(NEED_PIC), 1)
  919. ifeq ($(C_COMPILER), IBM)
  920. CCOMMON_OPT += -qpic=large
  921. else
  922. CCOMMON_OPT += -fPIC
  923. endif
  924. ifeq ($(F_COMPILER), SUN)
  925. FCOMMON_OPT += -pic
  926. else
  927. FCOMMON_OPT += -fPIC
  928. endif
  929. endif
  930. ifeq ($(DYNAMIC_ARCH), 1)
  931. CCOMMON_OPT += -DDYNAMIC_ARCH
  932. endif
  933. ifeq ($(DYNAMIC_OLDER), 1)
  934. CCOMMON_OPT += -DDYNAMIC_OLDER
  935. endif
  936. ifeq ($(NO_LAPACK), 1)
  937. CCOMMON_OPT += -DNO_LAPACK
  938. #Disable LAPACK C interface
  939. NO_LAPACKE = 1
  940. endif
  941. ifeq ($(NO_LAPACKE), 1)
  942. CCOMMON_OPT += -DNO_LAPACKE
  943. endif
  944. ifeq ($(NO_AVX), 1)
  945. CCOMMON_OPT += -DNO_AVX
  946. endif
  947. ifeq ($(ARCH), x86)
  948. CCOMMON_OPT += -DNO_AVX
  949. endif
  950. ifeq ($(NO_AVX2), 1)
  951. CCOMMON_OPT += -DNO_AVX2
  952. endif
  953. ifeq ($(NO_AVX512), 1)
  954. CCOMMON_OPT += -DNO_AVX512
  955. endif
  956. ifdef SMP
  957. CCOMMON_OPT += -DSMP_SERVER
  958. ifeq ($(ARCH), mips64)
  959. ifneq ($(CORE), LOONGSON3B)
  960. USE_SIMPLE_THREADED_LEVEL3 = 1
  961. endif
  962. endif
  963. ifeq ($(USE_OPENMP), 1)
  964. # USE_SIMPLE_THREADED_LEVEL3 = 1
  965. # NO_AFFINITY = 1
  966. CCOMMON_OPT += -DUSE_OPENMP
  967. endif
  968. ifeq ($(BIGNUMA), 1)
  969. CCOMMON_OPT += -DBIGNUMA
  970. endif
  971. endif
  972. ifeq ($(NO_WARMUP), 1)
  973. CCOMMON_OPT += -DNO_WARMUP
  974. endif
  975. ifeq ($(CONSISTENT_FPCSR), 1)
  976. CCOMMON_OPT += -DCONSISTENT_FPCSR
  977. endif
  978. # Only for development
  979. # CCOMMON_OPT += -DPARAMTEST
  980. # CCOMMON_OPT += -DPREFETCHTEST
  981. # CCOMMON_OPT += -DNO_SWITCHING
  982. # USE_PAPI = 1
  983. ifdef USE_PAPI
  984. CCOMMON_OPT += -DUSE_PAPI
  985. EXTRALIB += -lpapi -lperfctr
  986. endif
  987. ifdef DYNAMIC_THREADS
  988. CCOMMON_OPT += -DDYNAMIC_THREADS
  989. endif
  990. CCOMMON_OPT += -DMAX_CPU_NUMBER=$(NUM_THREADS)
  991. CCOMMON_OPT += -DMAX_PARALLEL_NUMBER=$(NUM_PARALLEL)
  992. ifdef USE_SIMPLE_THREADED_LEVEL3
  993. CCOMMON_OPT += -DUSE_SIMPLE_THREADED_LEVEL3
  994. endif
  995. ifeq ($(USE_TLS), 1)
  996. CCOMMON_OPT += -DUSE_TLS
  997. endif
  998. ifeq ($(BUILD_HALF), 1)
  999. CCOMMON_OPT += -DBUILD_HALF
  1000. endif
  1001. CCOMMON_OPT += -DVERSION=\"$(VERSION)\"
  1002. ifndef SYMBOLPREFIX
  1003. SYMBOLPREFIX =
  1004. endif
  1005. ifndef SYMBOLSUFFIX
  1006. SYMBOLSUFFIX =
  1007. endif
  1008. ifndef LIBNAMESUFFIX
  1009. LIBNAMEBASE = $(SYMBOLPREFIX)openblas$(SYMBOLSUFFIX)
  1010. else
  1011. LIBNAMEBASE = $(SYMBOLPREFIX)openblas$(SYMBOLSUFFIX)_$(LIBNAMESUFFIX)
  1012. endif
  1013. ifeq ($(OSNAME), CYGWIN_NT)
  1014. LIBPREFIX = cyg$(LIBNAMEBASE)
  1015. else
  1016. LIBPREFIX = lib$(LIBNAMEBASE)
  1017. endif
  1018. KERNELDIR = $(TOPDIR)/kernel/$(ARCH)
  1019. include $(TOPDIR)/Makefile.$(ARCH)
  1020. CCOMMON_OPT += -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME
  1021. CCOMMON_OPT += -DASMNAME=$(FU)$(*F) -DASMFNAME=$(FU)$(*F)$(BU) -DNAME=$(*F)$(BU) -DCNAME=$(*F) -DCHAR_NAME=\"$(*F)$(BU)\" -DCHAR_CNAME=\"$(*F)\"
  1022. ifeq ($(CORE), PPC440)
  1023. CCOMMON_OPT += -DALLOC_QALLOC
  1024. endif
  1025. ifeq ($(CORE), PPC440FP2)
  1026. STATIC_ALLOCATION = 1
  1027. endif
  1028. ifneq ($(OSNAME), Linux)
  1029. NO_AFFINITY = 1
  1030. endif
  1031. ifneq ($(ARCH), x86_64)
  1032. ifneq ($(ARCH), x86)
  1033. ifneq ($(CORE), LOONGSON3B)
  1034. NO_AFFINITY = 1
  1035. endif
  1036. endif
  1037. endif
  1038. ifdef NO_AFFINITY
  1039. ifeq ($(NO_AFFINITY), 0)
  1040. override undefine NO_AFFINITY
  1041. else
  1042. CCOMMON_OPT += -DNO_AFFINITY
  1043. endif
  1044. endif
  1045. ifdef FUNCTION_PROFILE
  1046. CCOMMON_OPT += -DFUNCTION_PROFILE
  1047. endif
  1048. ifdef HUGETLB_ALLOCATION
  1049. CCOMMON_OPT += -DALLOC_HUGETLB
  1050. endif
  1051. ifdef HUGETLBFILE_ALLOCATION
  1052. CCOMMON_OPT += -DALLOC_HUGETLBFILE -DHUGETLB_FILE_NAME=$(HUGETLBFILE_ALLOCATION)
  1053. endif
  1054. ifdef STATIC_ALLOCATION
  1055. CCOMMON_OPT += -DALLOC_STATIC
  1056. endif
  1057. ifdef DEVICEDRIVER_ALLOCATION
  1058. CCOMMON_OPT += -DALLOC_DEVICEDRIVER -DDEVICEDRIVER_NAME=\"/dev/mapper\"
  1059. endif
  1060. ifdef MIXED_MEMORY_ALLOCATION
  1061. CCOMMON_OPT += -DMIXED_MEMORY_ALLOCATION
  1062. endif
  1063. ifeq ($(OSNAME), SunOS)
  1064. TAR = gtar
  1065. PATCH = gpatch
  1066. GREP = ggrep
  1067. AWK = nawk
  1068. else
  1069. TAR = tar
  1070. PATCH = patch
  1071. GREP = grep
  1072. AWK = awk
  1073. endif
  1074. ifndef MD5SUM
  1075. MD5SUM = md5sum
  1076. endif
  1077. REVISION = -r$(VERSION)
  1078. MAJOR_VERSION = $(word 1,$(subst ., ,$(VERSION)))
  1079. ifeq ($(DEBUG), 1)
  1080. COMMON_OPT += -g
  1081. endif
  1082. ifeq ($(DEBUG), 1)
  1083. FCOMMON_OPT += -g
  1084. endif
  1085. ifndef COMMON_OPT
  1086. COMMON_OPT = -O2
  1087. endif
  1088. ifndef FCOMMON_OPT
  1089. FCOMMON_OPT = -O2 -frecursive
  1090. endif
  1091. override CFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
  1092. override PFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
  1093. override FFLAGS += $(COMMON_OPT) $(FCOMMON_OPT)
  1094. override FPFLAGS += $(FCOMMON_OPT) $(COMMON_PROF)
  1095. #MAKEOVERRIDES =
  1096. ifdef NEED_PIC
  1097. ifeq (,$(findstring PIC,$(FFLAGS)))
  1098. override FFLAGS += -fPIC
  1099. endif
  1100. endif
  1101. #For LAPACK Fortran codes.
  1102. #Disable -fopenmp for LAPACK Fortran codes on Windows.
  1103. ifdef OS_WINDOWS
  1104. LAPACK_FFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FFLAGS))
  1105. LAPACK_FPFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FPFLAGS))
  1106. else
  1107. LAPACK_FFLAGS := $(FFLAGS)
  1108. LAPACK_FPFLAGS := $(FPFLAGS)
  1109. endif
  1110. LAPACK_CFLAGS = $(CFLAGS)
  1111. LAPACK_CFLAGS += -DHAVE_LAPACK_CONFIG_H
  1112. ifdef INTERFACE64
  1113. ifneq ($(INTERFACE64), 0)
  1114. LAPACK_CFLAGS += -DLAPACK_ILP64
  1115. endif
  1116. endif
  1117. ifdef OS_WINDOWS
  1118. LAPACK_CFLAGS += -DOPENBLAS_OS_WINDOWS
  1119. endif
  1120. ifeq ($(C_COMPILER), LSB)
  1121. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1122. endif
  1123. ifndef SUFFIX
  1124. SUFFIX = o
  1125. endif
  1126. ifndef PSUFFIX
  1127. PSUFFIX = po
  1128. endif
  1129. ifndef LIBSUFFIX
  1130. LIBSUFFIX = a
  1131. endif
  1132. ifneq ($(DYNAMIC_ARCH), 1)
  1133. ifndef SMP
  1134. LIBNAME = $(LIBPREFIX)_$(LIBCORE)$(REVISION).$(LIBSUFFIX)
  1135. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)$(REVISION)_p.$(LIBSUFFIX)
  1136. else
  1137. LIBNAME = $(LIBPREFIX)_$(LIBCORE)p$(REVISION).$(LIBSUFFIX)
  1138. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)p$(REVISION)_p.$(LIBSUFFIX)
  1139. endif
  1140. else
  1141. ifndef SMP
  1142. LIBNAME = $(LIBPREFIX)$(REVISION).$(LIBSUFFIX)
  1143. LIBNAME_P = $(LIBPREFIX)$(REVISION)_p.$(LIBSUFFIX)
  1144. else
  1145. LIBNAME = $(LIBPREFIX)p$(REVISION).$(LIBSUFFIX)
  1146. LIBNAME_P = $(LIBPREFIX)p$(REVISION)_p.$(LIBSUFFIX)
  1147. endif
  1148. endif
  1149. LIBDLLNAME = $(LIBPREFIX).dll
  1150. IMPLIBNAME = lib$(LIBNAMEBASE).dll.a
  1151. ifneq ($(OSNAME), AIX)
  1152. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.so)
  1153. else
  1154. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.a)
  1155. endif
  1156. LIBDYNNAME = $(LIBNAME:.$(LIBSUFFIX)=.dylib)
  1157. LIBDEFNAME = $(LIBNAME:.$(LIBSUFFIX)=.def)
  1158. LIBEXPNAME = $(LIBNAME:.$(LIBSUFFIX)=.exp)
  1159. LIBZIPNAME = $(LIBNAME:.$(LIBSUFFIX)=.zip)
  1160. LIBS = $(TOPDIR)/$(LIBNAME)
  1161. LIBS_P = $(TOPDIR)/$(LIBNAME_P)
  1162. LIB_COMPONENTS = BLAS
  1163. ifneq ($(NO_CBLAS), 1)
  1164. LIB_COMPONENTS += CBLAS
  1165. endif
  1166. ifneq ($(NO_LAPACK), 1)
  1167. LIB_COMPONENTS += LAPACK
  1168. ifneq ($(NO_LAPACKE), 1)
  1169. LIB_COMPONENTS += LAPACKE
  1170. endif
  1171. ifeq ($(BUILD_RELAPACK), 1)
  1172. LIB_COMPONENTS += ReLAPACK
  1173. endif
  1174. endif
  1175. ifeq ($(ONLY_CBLAS), 1)
  1176. LIB_COMPONENTS = CBLAS
  1177. endif
  1178. export OSNAME
  1179. export ARCH
  1180. export CORE
  1181. export LIBCORE
  1182. export __BYTE_ORDER__
  1183. export ELF_VERSION
  1184. export PGCPATH
  1185. export CONFIG
  1186. export CC
  1187. export FC
  1188. export BU
  1189. export FU
  1190. export NEED2UNDERSCORES
  1191. export USE_THREAD
  1192. export NUM_THREADS
  1193. export NUM_CORES
  1194. export SMP
  1195. export MAKEFILE_RULE
  1196. export NEED_PIC
  1197. export BINARY
  1198. export BINARY32
  1199. export BINARY64
  1200. export F_COMPILER
  1201. export C_COMPILER
  1202. export USE_OPENMP
  1203. export CROSS
  1204. export CROSS_SUFFIX
  1205. export NOFORTRAN
  1206. export NO_FBLAS
  1207. export EXTRALIB
  1208. export CEXTRALIB
  1209. export FEXTRALIB
  1210. export HAVE_SSE
  1211. export HAVE_SSE2
  1212. export HAVE_SSE3
  1213. export HAVE_SSSE3
  1214. export HAVE_SSE4_1
  1215. export HAVE_SSE4_2
  1216. export HAVE_SSE4A
  1217. export HAVE_SSE5
  1218. export HAVE_AVX
  1219. export HAVE_VFP
  1220. export HAVE_VFPV3
  1221. export HAVE_VFPV4
  1222. export HAVE_NEON
  1223. export HAVE_MSA
  1224. export MSA_FLAGS
  1225. export KERNELDIR
  1226. export FUNCTION_PROFILE
  1227. export TARGET_CORE
  1228. export NO_AVX512
  1229. export BUILD_HALF
  1230. export SHGEMM_UNROLL_M
  1231. export SHGEMM_UNROLL_N
  1232. export SGEMM_UNROLL_M
  1233. export SGEMM_UNROLL_N
  1234. export DGEMM_UNROLL_M
  1235. export DGEMM_UNROLL_N
  1236. export QGEMM_UNROLL_M
  1237. export QGEMM_UNROLL_N
  1238. export CGEMM_UNROLL_M
  1239. export CGEMM_UNROLL_N
  1240. export ZGEMM_UNROLL_M
  1241. export ZGEMM_UNROLL_N
  1242. export XGEMM_UNROLL_M
  1243. export XGEMM_UNROLL_N
  1244. export CGEMM3M_UNROLL_M
  1245. export CGEMM3M_UNROLL_N
  1246. export ZGEMM3M_UNROLL_M
  1247. export ZGEMM3M_UNROLL_N
  1248. export XGEMM3M_UNROLL_M
  1249. export XGEMM3M_UNROLL_N
  1250. ifdef USE_CUDA
  1251. export CUDADIR
  1252. export CUCC
  1253. export CUFLAGS
  1254. export CULIB
  1255. endif
  1256. .SUFFIXES: .$(PSUFFIX) .$(SUFFIX) .f
  1257. .f.$(SUFFIX):
  1258. $(FC) $(FFLAGS) -c $< -o $(@F)
  1259. .f.$(PSUFFIX):
  1260. $(FC) $(FPFLAGS) -pg -c $< -o $(@F)
  1261. ifdef BINARY64
  1262. PATHSCALEPATH = /opt/pathscale/lib/3.1
  1263. PGIPATH = /opt/pgi/linux86-64/7.1-5/lib
  1264. else
  1265. PATHSCALEPATH = /opt/pathscale/lib/3.1/32
  1266. PGIPATH = /opt/pgi/linux86/7.1-5/lib
  1267. endif
  1268. ACMLPATH = /opt/acml/4.3.0
  1269. ifneq ($(OSNAME), Darwin)
  1270. MKLPATH = /opt/intel/mkl/10.2.2.025/lib
  1271. else
  1272. MKLPATH = /Library/Frameworks/Intel_MKL.framework/Versions/10.0.1.014/lib
  1273. endif
  1274. ATLASPATH = /opt/atlas/3.9.17/opteron
  1275. FLAMEPATH = $(HOME)/flame/lib
  1276. ifneq ($(OSNAME), SunOS)
  1277. SUNPATH = /opt/sunstudio12.1
  1278. else
  1279. SUNPATH = /opt/SUNWspro
  1280. endif