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
1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285128612871288128912901291129212931294129512961297129812991300130113021303130413051306130713081309131013111312131313141315131613171318131913201321132213231324132513261327132813291330133113321333133413351336133713381339134013411342134313441345134613471348134913501351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384138513861387138813891390139113921393139413951396139713981399140014011402140314041405140614071408140914101411141214131414141514161417141814191420142114221423142414251426142714281429143014311432143314341435143614371438143914401441144214431444144514461447144814491450145114521453145414551456145714581459146014611462146314641465146614671468146914701471147214731474147514761477147814791480148114821483148414851486148714881489149014911492149314941495149614971498149915001501150215031504150515061507150815091510151115121513151415151516151715181519152015211522152315241525152615271528152915301531153215331534153515361537153815391540154115421543154415451546154715481549155015511552155315541555155615571558155915601561156215631564
  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), armv6)
  25. override ARCH=arm
  26. else ifeq ($(ARCH), armv7)
  27. override ARCH=arm
  28. else ifeq ($(ARCH), aarch64)
  29. override ARCH=arm64
  30. else ifeq ($(ARCH), zarch)
  31. override ARCH=zarch
  32. endif
  33. NETLIB_LAPACK_DIR = $(TOPDIR)/lapack-netlib
  34. # Default C compiler
  35. # - Only set if not specified on the command line or inherited from the environment.
  36. # - CC is an implicit variable so neither '?=' or 'ifndef' can be used.
  37. # http://stackoverflow.com/questions/4029274/mingw-and-make-variables
  38. # - Default value is 'cc' which is not always a valid command (e.g. MinGW).
  39. ifeq ($(origin CC),default)
  40. # Check if $(CC) refers to a valid command and set the value to gcc if not
  41. ifneq ($(findstring cmd.exe,$(SHELL)),)
  42. ifeq ($(shell where $(CC) 2>NUL),)
  43. CC = gcc
  44. endif
  45. else # POSIX-ish
  46. ifeq ($(shell command -v $(CC) 2>/dev/null),)
  47. ifeq ($(shell uname -s),Darwin)
  48. CC = clang
  49. # EXTRALIB += -Wl,-no_compact_unwind
  50. else
  51. CC = gcc
  52. endif # Darwin
  53. endif # CC exists
  54. endif # Shell is sane
  55. endif # CC is set to default
  56. # Default Fortran compiler (FC) is selected by f_check.
  57. ifndef MAKEFILE_RULE
  58. include $(TOPDIR)/Makefile.rule
  59. else
  60. include $(TOPDIR)/$(MAKEFILE_RULE)
  61. endif
  62. #
  63. # Beginning of system configuration
  64. #
  65. ifndef HOSTCC
  66. HOSTCC = $(CC)
  67. endif
  68. ifdef TARGET
  69. GETARCH_FLAGS := -DFORCE_$(TARGET)
  70. GETARCH_FLAGS += -DUSER_TARGET
  71. endif
  72. # Force fallbacks for 32bit
  73. ifeq ($(BINARY), 32)
  74. ifeq ($(TARGET), HASWELL)
  75. GETARCH_FLAGS := -DFORCE_NEHALEM
  76. endif
  77. ifeq ($(TARGET), SKYLAKEX)
  78. GETARCH_FLAGS := -DFORCE_NEHALEM
  79. endif
  80. ifeq ($(TARGET), COOPERLAKE)
  81. GETARCH_FLAGS := -DFORCE_NEHALEM
  82. endif
  83. ifeq ($(TARGET), SANDYBRIDGE)
  84. GETARCH_FLAGS := -DFORCE_NEHALEM
  85. endif
  86. ifeq ($(TARGET), BULLDOZER)
  87. GETARCH_FLAGS := -DFORCE_BARCELONA
  88. endif
  89. ifeq ($(TARGET), PILEDRIVER)
  90. GETARCH_FLAGS := -DFORCE_BARCELONA
  91. endif
  92. ifeq ($(TARGET), STEAMROLLER)
  93. GETARCH_FLAGS := -DFORCE_BARCELONA
  94. endif
  95. ifeq ($(TARGET), EXCAVATOR)
  96. GETARCH_FLAGS := -DFORCE_BARCELONA
  97. endif
  98. ifeq ($(TARGET), ZEN)
  99. GETARCH_FLAGS := -DFORCE_BARCELONA
  100. endif
  101. ifeq ($(TARGET), ARMV8)
  102. GETARCH_FLAGS := -DFORCE_ARMV7
  103. endif
  104. ifeq ($(TARGET), POWER8)
  105. GETARCH_FLAGS := -DFORCE_POWER6
  106. endif
  107. endif
  108. #TARGET_CORE will override TARGET which is used in DYNAMIC_ARCH=1.
  109. #
  110. ifdef TARGET_CORE
  111. GETARCH_FLAGS := -DFORCE_$(TARGET_CORE)
  112. endif
  113. # Force fallbacks for 32bit
  114. ifeq ($(BINARY), 32)
  115. ifeq ($(TARGET_CORE), HASWELL)
  116. GETARCH_FLAGS := -DFORCE_NEHALEM
  117. endif
  118. ifeq ($(TARGET_CORE), SKYLAKEX)
  119. GETARCH_FLAGS := -DFORCE_NEHALEM
  120. endif
  121. ifeq ($(TARGET_CORE), COOPERLAKE)
  122. GETARCH_FLAGS := -DFORCE_NEHALEM
  123. endif
  124. ifeq ($(TARGET_CORE), SANDYBRIDGE)
  125. GETARCH_FLAGS := -DFORCE_NEHALEM
  126. endif
  127. ifeq ($(TARGET_CORE), BULLDOZER)
  128. GETARCH_FLAGS := -DFORCE_BARCELONA
  129. endif
  130. ifeq ($(TARGET_CORE), PILEDRIVER)
  131. GETARCH_FLAGS := -DFORCE_BARCELONA
  132. endif
  133. ifeq ($(TARGET_CORE), STEAMROLLER)
  134. GETARCH_FLAGS := -DFORCE_BARCELONA
  135. endif
  136. ifeq ($(TARGET_CORE), EXCAVATOR)
  137. GETARCH_FLAGS := -DFORCE_BARCELONA
  138. endif
  139. ifeq ($(TARGET_CORE), ZEN)
  140. GETARCH_FLAGS := -DFORCE_BARCELONA
  141. endif
  142. endif
  143. # On x86_64 build getarch with march=native unless the compiler is PGI. This is required to detect AVX512 support in getarch.
  144. ifeq ($(HOSTARCH), x86_64)
  145. ifeq ($(findstring pgcc,$(HOSTCC)),)
  146. GETARCH_FLAGS += -march=native
  147. endif
  148. endif
  149. ifdef INTERFACE64
  150. ifneq ($(INTERFACE64), 0)
  151. GETARCH_FLAGS += -DUSE64BITINT
  152. endif
  153. endif
  154. ifndef GEMM_MULTITHREAD_THRESHOLD
  155. GEMM_MULTITHREAD_THRESHOLD=4
  156. endif
  157. GETARCH_FLAGS += -DGEMM_MULTITHREAD_THRESHOLD=$(GEMM_MULTITHREAD_THRESHOLD)
  158. ifeq ($(NO_AVX), 1)
  159. GETARCH_FLAGS += -DNO_AVX
  160. endif
  161. ifeq ($(BINARY), 32)
  162. GETARCH_FLAGS += -DNO_AVX -DNO_AVX2 -DNO_AVX512
  163. NO_AVX512 = 1
  164. endif
  165. ifeq ($(NO_AVX2), 1)
  166. GETARCH_FLAGS += -DNO_AVX2
  167. endif
  168. ifeq ($(NO_AVX512), 1)
  169. GETARCH_FLAGS += -DNO_AVX512
  170. endif
  171. ifeq ($(DEBUG), 1)
  172. GETARCH_FLAGS += -g
  173. endif
  174. ifeq ($(QUIET_MAKE), 1)
  175. MAKE += -s
  176. endif
  177. ifndef NO_PARALLEL_MAKE
  178. NO_PARALLEL_MAKE=0
  179. endif
  180. GETARCH_FLAGS += -DNO_PARALLEL_MAKE=$(NO_PARALLEL_MAKE)
  181. ifdef MAKE_NB_JOBS
  182. GETARCH_FLAGS += -DMAKE_NB_JOBS=$(MAKE_NB_JOBS)
  183. endif
  184. ifeq ($(HOSTCC), loongcc)
  185. GETARCH_FLAGS += -static
  186. endif
  187. #if don't use Fortran, it will only compile CBLAS.
  188. ifeq ($(ONLY_CBLAS), 1)
  189. NO_LAPACK = 1
  190. else
  191. ONLY_CBLAS = 0
  192. endif
  193. #For small matrix optimization
  194. ifeq ($(SMALL_MATRIX_OPT), 1)
  195. CCOMMON_OPT += -DSMALL_MATRIX_OPT
  196. endif
  197. # This operation is expensive, so execution should be once.
  198. ifndef GOTOBLAS_MAKEFILE
  199. export GOTOBLAS_MAKEFILE = 1
  200. # Generating Makefile.conf and config.h
  201. 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)
  202. ifndef TARGET_CORE
  203. include $(TOPDIR)/Makefile.conf
  204. else
  205. include $(TOPDIR)/Makefile_kernel.conf
  206. endif
  207. endif
  208. ifndef NUM_PARALLEL
  209. NUM_PARALLEL = 1
  210. endif
  211. ifndef NUM_THREADS
  212. NUM_THREADS = $(NUM_CORES)
  213. endif
  214. ifeq ($(NUM_THREADS), 1)
  215. override USE_THREAD = 0
  216. override USE_OPENMP = 0
  217. endif
  218. ifdef USE_THREAD
  219. ifeq ($(USE_THREAD), 0)
  220. SMP =
  221. else
  222. SMP = 1
  223. endif
  224. else
  225. ifeq ($(NUM_THREAD), 1)
  226. SMP =
  227. else
  228. SMP = 1
  229. endif
  230. endif
  231. ifeq ($(SMP), 1)
  232. USE_LOCKING =
  233. endif
  234. ifndef NEED_PIC
  235. NEED_PIC = 1
  236. endif
  237. ARFLAGS =
  238. CPP = $(COMPILER) -E
  239. AR ?= $(CROSS_SUFFIX)ar
  240. AS ?= $(CROSS_SUFFIX)as
  241. LD ?= $(CROSS_SUFFIX)ld
  242. RANLIB ?= $(CROSS_SUFFIX)ranlib
  243. NM = $(CROSS_SUFFIX)nm
  244. DLLWRAP = $(CROSS_SUFFIX)dllwrap
  245. OBJCOPY = $(CROSS_SUFFIX)objcopy
  246. OBJCONV = $(CROSS_SUFFIX)objconv
  247. # When fortran support was either not detected or actively deselected, only build BLAS.
  248. ifeq ($(NOFORTRAN), 1)
  249. NO_LAPACK = 1
  250. override FEXTRALIB =
  251. endif
  252. ifeq ($(C_COMPILER), GCC)
  253. GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
  254. GCCVERSIONGT4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 4)
  255. GCCVERSIONEQ5 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` = 5)
  256. GCCVERSIONGT5 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 5)
  257. GCCVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 7)
  258. GCCVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  259. GCCVERSIONGTEQ11 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 11)
  260. GCCVERSIONGTEQ10 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 10)
  261. # Note that the behavior of -dumpversion is compile-time-configurable for
  262. # gcc-7.x and newer. Use -dumpfullversion there
  263. ifeq ($(GCCVERSIONGTEQ7),1)
  264. GCCDUMPVERSION_PARAM := -dumpfullversion
  265. else
  266. GCCDUMPVERSION_PARAM := -dumpversion
  267. endif
  268. GCCMINORVERSIONGTEQ2 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 2)
  269. GCCMINORVERSIONGTEQ7 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 7)
  270. endif
  271. #
  272. # OS dependent settings
  273. #
  274. ifeq ($(OSNAME), Darwin)
  275. ifndef MACOSX_DEPLOYMENT_TARGET
  276. export MACOSX_DEPLOYMENT_TARGET=10.8
  277. endif
  278. MD5SUM = md5 -r
  279. endif
  280. ifneq (,$(findstring $(OSNAME), FreeBSD OpenBSD DragonFly))
  281. MD5SUM = md5 -r
  282. endif
  283. ifeq ($(OSNAME), NetBSD)
  284. MD5SUM = md5 -n
  285. endif
  286. ifeq ($(OSNAME), Linux)
  287. EXTRALIB += -lm
  288. NO_EXPRECISION = 1
  289. endif
  290. ifeq ($(OSNAME), Android)
  291. EXTRALIB += -lm
  292. endif
  293. ifeq ($(OSNAME), AIX)
  294. EXTRALIB += -lm
  295. endif
  296. ifeq ($(OSNAME), WINNT)
  297. NEED_PIC = 0
  298. NO_EXPRECISION = 1
  299. EXTRALIB += -defaultlib:advapi32
  300. SUFFIX = obj
  301. PSUFFIX = pobj
  302. LIBSUFFIX = a
  303. ifeq ($(C_COMPILER), CLANG)
  304. CCOMMON_OPT += -DMS_ABI
  305. endif
  306. #Version tests for supporting specific features (MS_ABI, POWER9 intrinsics)
  307. ifeq ($(GCCVERSIONGT4), 1)
  308. # GCC Major version > 4
  309. # It is compatible with MSVC ABI.
  310. CCOMMON_OPT += -DMS_ABI
  311. endif
  312. ifeq ($(GCCVERSIONGTEQ4), 1)
  313. ifeq ($(GCCMINORVERSIONGTEQ7), 1)
  314. # GCC Version >=4.7
  315. # It is compatible with MSVC ABI.
  316. CCOMMON_OPT += -DMS_ABI
  317. endif
  318. endif
  319. # Ensure the correct stack alignment on Win32
  320. # http://permalink.gmane.org/gmane.comp.lib.openblas.general/97
  321. ifeq ($(ARCH), x86)
  322. CCOMMON_OPT += -mincoming-stack-boundary=2
  323. FCOMMON_OPT += -mincoming-stack-boundary=2
  324. endif
  325. endif
  326. ifeq ($(OSNAME), Interix)
  327. NEED_PIC = 0
  328. NO_EXPRECISION = 1
  329. INTERIX_TOOL_DIR = /opt/gcc.3.3/i586-pc-interix3/bin
  330. endif
  331. ifeq ($(OSNAME), CYGWIN_NT)
  332. NEED_PIC = 0
  333. NO_EXPRECISION = 1
  334. OS_CYGWIN_NT = 1
  335. endif
  336. ifneq ($(OSNAME), WINNT)
  337. ifneq ($(OSNAME), CYGWIN_NT)
  338. ifneq ($(OSNAME), Interix)
  339. ifneq ($(OSNAME), Android)
  340. ifdef SMP
  341. EXTRALIB += -lpthread
  342. endif
  343. endif
  344. endif
  345. endif
  346. endif
  347. # ifeq logical or
  348. ifeq ($(OSNAME), $(filter $(OSNAME),WINNT CYGWIN_NT Interix))
  349. OS_WINDOWS=1
  350. endif
  351. ifdef QUAD_PRECISION
  352. CCOMMON_OPT += -DQUAD_PRECISION
  353. NO_EXPRECISION = 1
  354. endif
  355. ifneq ($(ARCH), x86)
  356. ifneq ($(ARCH), x86_64)
  357. NO_EXPRECISION = 1
  358. endif
  359. endif
  360. ifdef UTEST_CHECK
  361. CCOMMON_OPT += -DUTEST_CHECK
  362. SANITY_CHECK = 1
  363. endif
  364. ifdef SANITY_CHECK
  365. CCOMMON_OPT += -DSANITY_CHECK -DREFNAME=$(*F)f$(BU)
  366. endif
  367. MAX_STACK_ALLOC ?= 2048
  368. ifneq ($(MAX_STACK_ALLOC), 0)
  369. CCOMMON_OPT += -DMAX_STACK_ALLOC=$(MAX_STACK_ALLOC)
  370. endif
  371. ifdef USE_LOCKING
  372. ifneq ($(USE_LOCKING), 0)
  373. CCOMMON_OPT += -DUSE_LOCKING
  374. endif
  375. endif
  376. #
  377. # Architecture dependent settings
  378. #
  379. ifeq ($(ARCH), x86)
  380. ifndef BINARY
  381. NO_BINARY_MODE = 1
  382. endif
  383. ifeq ($(CORE), generic)
  384. NO_EXPRECISION = 1
  385. endif
  386. ifndef NO_EXPRECISION
  387. ifeq ($(F_COMPILER), GFORTRAN)
  388. # ifeq logical or. GCC or LSB
  389. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  390. EXPRECISION = 1
  391. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  392. FCOMMON_OPT += -m128bit-long-double
  393. endif
  394. ifeq ($(C_COMPILER), CLANG)
  395. EXPRECISION = 1
  396. CCOMMON_OPT += -DEXPRECISION
  397. FCOMMON_OPT += -m128bit-long-double
  398. endif
  399. endif
  400. endif
  401. endif
  402. ifeq ($(ARCH), x86_64)
  403. ifeq ($(CORE), generic)
  404. NO_EXPRECISION = 1
  405. endif
  406. ifndef NO_EXPRECISION
  407. ifeq ($(F_COMPILER), GFORTRAN)
  408. # ifeq logical or. GCC or LSB
  409. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  410. EXPRECISION = 1
  411. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  412. FCOMMON_OPT += -m128bit-long-double
  413. endif
  414. ifeq ($(C_COMPILER), CLANG)
  415. EXPRECISION = 1
  416. CCOMMON_OPT += -DEXPRECISION
  417. FCOMMON_OPT += -m128bit-long-double
  418. endif
  419. endif
  420. endif
  421. endif
  422. ifeq ($(C_COMPILER), INTEL)
  423. CCOMMON_OPT += -wd981
  424. endif
  425. ifeq ($(USE_OPENMP), 1)
  426. #check
  427. ifeq ($(USE_THREAD), 0)
  428. $(error OpenBLAS: Cannot set both USE_OPENMP=1 and USE_THREAD=0. The USE_THREAD=0 is only for building single thread version.)
  429. endif
  430. # ifeq logical or. GCC or LSB
  431. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  432. CCOMMON_OPT += -fopenmp
  433. endif
  434. ifeq ($(C_COMPILER), CLANG)
  435. CCOMMON_OPT += -fopenmp
  436. endif
  437. ifeq ($(C_COMPILER), INTEL)
  438. CCOMMON_OPT += -fopenmp
  439. endif
  440. ifeq ($(C_COMPILER), PGI)
  441. CCOMMON_OPT += -mp
  442. endif
  443. ifeq ($(C_COMPILER), OPEN64)
  444. CCOMMON_OPT += -mp
  445. CEXTRALIB += -lstdc++
  446. endif
  447. ifeq ($(C_COMPILER), PATHSCALE)
  448. CCOMMON_OPT += -mp
  449. endif
  450. endif
  451. ifeq ($(DYNAMIC_ARCH), 1)
  452. ifeq ($(ARCH), x86)
  453. DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
  454. CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
  455. endif
  456. ifeq ($(ARCH), x86_64)
  457. DYNAMIC_CORE = PRESCOTT CORE2
  458. ifeq ($(DYNAMIC_OLDER), 1)
  459. DYNAMIC_CORE += PENRYN DUNNINGTON
  460. endif
  461. DYNAMIC_CORE += NEHALEM
  462. ifeq ($(DYNAMIC_OLDER), 1)
  463. DYNAMIC_CORE += OPTERON OPTERON_SSE3
  464. endif
  465. DYNAMIC_CORE += BARCELONA
  466. ifeq ($(DYNAMIC_OLDER), 1)
  467. DYNAMIC_CORE += BOBCAT ATOM NANO
  468. endif
  469. ifneq ($(NO_AVX), 1)
  470. DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER STEAMROLLER EXCAVATOR
  471. endif
  472. ifneq ($(NO_AVX2), 1)
  473. DYNAMIC_CORE += HASWELL ZEN
  474. endif
  475. ifneq ($(NO_AVX512), 1)
  476. ifneq ($(NO_AVX2), 1)
  477. DYNAMIC_CORE += SKYLAKEX COOPERLAKE
  478. endif
  479. endif
  480. endif
  481. ifdef DYNAMIC_LIST
  482. override DYNAMIC_CORE = PRESCOTT $(DYNAMIC_LIST)
  483. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_PRESCOTT
  484. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  485. CCOMMON_OPT += $(XCCOMMON_OPT)
  486. #CCOMMON_OPT += -DDYNAMIC_LIST='$(DYNAMIC_LIST)'
  487. endif
  488. ifeq ($(ARCH), arm64)
  489. DYNAMIC_CORE = ARMV8
  490. DYNAMIC_CORE += CORTEXA53
  491. DYNAMIC_CORE += CORTEXA57
  492. DYNAMIC_CORE += CORTEXA72
  493. DYNAMIC_CORE += CORTEXA73
  494. DYNAMIC_CORE += NEOVERSEN1
  495. DYNAMIC_CORE += FALKOR
  496. DYNAMIC_CORE += THUNDERX
  497. DYNAMIC_CORE += THUNDERX2T99
  498. DYNAMIC_CORE += TSV110
  499. DYNAMIC_CORE += EMAG8180
  500. DYNAMIC_CORE += THUNDERX3T110
  501. endif
  502. ifeq ($(ARCH), zarch)
  503. DYNAMIC_CORE = ZARCH_GENERIC
  504. # Z13 is supported since gcc-5.2, gcc-6, and in RHEL 7.3 and newer
  505. ifeq ($(GCCVERSIONGT5), 1)
  506. ZARCH_SUPPORT_Z13 := 1
  507. else ifeq ($(GCCVERSIONEQ5), 1)
  508. ifeq ($(GCCMINORVERSIONGTEQ2), 1)
  509. ZARCH_SUPPORT_Z13 := 1
  510. endif
  511. endif
  512. ifeq ($(wildcard /etc/redhat-release), /etc/redhat-release)
  513. ifeq ($(shell source /etc/os-release ; expr $$VERSION_ID \>= "7.3"), 1)
  514. ZARCH_SUPPORT_Z13 := 1
  515. endif
  516. endif
  517. ifeq ($(ZARCH_SUPPORT_Z13), 1)
  518. DYNAMIC_CORE += Z13
  519. else
  520. $(info OpenBLAS: Not building Z13 kernels because gcc is older than 5.2 or 6.x)
  521. endif
  522. ifeq ($(GCCVERSIONGTEQ7), 1)
  523. DYNAMIC_CORE += Z14
  524. else
  525. $(info OpenBLAS: Not building Z14 kernels because gcc is older than 7.x)
  526. endif
  527. endif
  528. ifeq ($(ARCH), power)
  529. DYNAMIC_CORE = POWER6
  530. DYNAMIC_CORE += POWER8
  531. ifneq ($(C_COMPILER), GCC)
  532. DYNAMIC_CORE += POWER9
  533. DYNAMIC_CORE += POWER10
  534. endif
  535. ifeq ($(C_COMPILER), GCC)
  536. ifeq ($(GCCVERSIONGT5), 1)
  537. DYNAMIC_CORE += POWER9
  538. else
  539. $(info, OpenBLAS: Your gcc version is too old to build the POWER9 kernels.)
  540. endif
  541. ifeq ($(GCCVERSIONGTEQ11), 1)
  542. DYNAMIC_CORE += POWER10
  543. else ifeq ($(GCCVERSIONGTEQ10), 1)
  544. ifeq ($(GCCMINORVERSIONGTEQ2), 1)
  545. DYNAMIC_CORE += POWER10
  546. endif
  547. else
  548. $(info, OpenBLAS: Your gcc version is too old to build the POWER10 kernels.)
  549. endif
  550. endif
  551. endif
  552. # If DYNAMIC_CORE is not set, DYNAMIC_ARCH cannot do anything, so force it to empty
  553. ifndef DYNAMIC_CORE
  554. override DYNAMIC_ARCH=
  555. endif
  556. endif
  557. ifeq ($(ARCH), ia64)
  558. NO_BINARY_MODE = 1
  559. BINARY_DEFINED = 1
  560. ifeq ($(F_COMPILER), GFORTRAN)
  561. ifeq ($(C_COMPILER), GCC)
  562. # EXPRECISION = 1
  563. # CCOMMON_OPT += -DEXPRECISION
  564. endif
  565. endif
  566. endif
  567. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  568. NO_BINARY_MODE = 1
  569. endif
  570. ifeq ($(ARCH), alpha)
  571. NO_BINARY_MODE = 1
  572. BINARY_DEFINED = 1
  573. endif
  574. ifeq ($(ARCH), arm)
  575. NO_BINARY_MODE = 1
  576. BINARY_DEFINED = 1
  577. CCOMMON_OPT += -marm
  578. FCOMMON_OPT += -marm
  579. # If softfp abi is mentioned on the command line, force it.
  580. ifeq ($(ARM_SOFTFP_ABI), 1)
  581. CCOMMON_OPT += -mfloat-abi=softfp
  582. FCOMMON_OPT += -mfloat-abi=softfp
  583. endif
  584. ifeq ($(OSNAME), Android)
  585. ifeq ($(ARM_SOFTFP_ABI), 1)
  586. EXTRALIB += -lm
  587. else
  588. EXTRALIB += -Wl,-lm_hard
  589. endif
  590. endif
  591. endif
  592. ifeq ($(ARCH), arm64)
  593. NO_BINARY_MODE = 1
  594. BINARY_DEFINED = 1
  595. ifdef INTERFACE64
  596. ifneq ($(INTERFACE64), 0)
  597. ifeq ($(F_COMPILER), GFORTRAN)
  598. FCOMMON_OPT += -fdefault-integer-8
  599. endif
  600. ifeq ($(F_COMPILER), FLANG)
  601. FCOMMON_OPT += -i8
  602. endif
  603. endif
  604. endif
  605. endif
  606. #
  607. # C Compiler dependent settings
  608. #
  609. # ifeq logical or. GCC or CLANG or LSB
  610. # http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
  611. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG LSB))
  612. CCOMMON_OPT += -Wall
  613. COMMON_PROF += -fno-inline
  614. NO_UNINITIALIZED_WARN = -Wno-uninitialized
  615. ifeq ($(QUIET_MAKE), 1)
  616. CCOMMON_OPT += $(NO_UNINITIALIZED_WARN) -Wno-unused
  617. endif
  618. ifdef NO_BINARY_MODE
  619. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  620. ifdef BINARY64
  621. CCOMMON_OPT += -mabi=64
  622. else
  623. CCOMMON_OPT += -mabi=n32
  624. endif
  625. BINARY_DEFINED = 1
  626. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  627. CCOMMON_OPT += -mabi=32
  628. BINARY_DEFINED = 1
  629. endif
  630. ifeq ($(CORE), LOONGSON3A)
  631. CCOMMON_OPT += -march=mips64
  632. FCOMMON_OPT += -march=mips64
  633. endif
  634. ifeq ($(CORE), LOONGSON3B)
  635. CCOMMON_OPT += -march=mips64
  636. FCOMMON_OPT += -march=mips64
  637. endif
  638. ifeq ($(CORE), MIPS24K)
  639. CCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  640. FCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  641. endif
  642. ifeq ($(CORE), MIPS1004K)
  643. CCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  644. FCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  645. endif
  646. ifeq ($(CORE), P5600)
  647. CCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  648. FCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  649. endif
  650. ifeq ($(CORE), I6400)
  651. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  652. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  653. endif
  654. ifeq ($(CORE), P6600)
  655. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  656. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  657. endif
  658. ifeq ($(CORE), I6500)
  659. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  660. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  661. endif
  662. ifeq ($(OSNAME), AIX)
  663. BINARY_DEFINED = 1
  664. endif
  665. endif
  666. ifndef BINARY_DEFINED
  667. ifneq ($(OSNAME), AIX)
  668. ifdef BINARY64
  669. CCOMMON_OPT += -m64
  670. else
  671. CCOMMON_OPT += -m32
  672. endif
  673. endif
  674. endif
  675. endif
  676. ifeq ($(C_COMPILER), PGI)
  677. ifdef BINARY64
  678. ifeq ($(ARCH), x86_64)
  679. CCOMMON_OPT += -tp p7-64 -D__MMX__ -Mnollvm
  680. else
  681. ifeq ($(ARCH), power)
  682. ifeq ($(CORE), POWER8)
  683. CCOMMON_OPT += -tp pwr8
  684. endif
  685. ifeq ($(CORE), POWER9)
  686. CCOMMON_OPT += -tp pwr9
  687. endif
  688. endif
  689. endif
  690. else
  691. CCOMMON_OPT += -tp p7
  692. endif
  693. endif
  694. ifeq ($(C_COMPILER), PATHSCALE)
  695. ifdef BINARY64
  696. CCOMMON_OPT += -m64
  697. else
  698. CCOMMON_OPT += -m32
  699. endif
  700. endif
  701. #
  702. # Fortran Compiler dependent settings
  703. #
  704. ifeq ($(F_COMPILER), FLANG)
  705. CCOMMON_OPT += -DF_INTERFACE_FLANG
  706. FCOMMON_OPT += -Mrecursive -Kieee
  707. ifeq ($(OSNAME), Linux)
  708. ifeq ($(ARCH), x86_64)
  709. FLANG_VENDOR := $(shell expr `$(FC) --version|cut -f 1 -d "."|head -1`)
  710. ifeq ($(FLANG_VENDOR),AOCC)
  711. FCOMMON_OPT += -fno-unroll-loops
  712. endif
  713. endif
  714. endif
  715. ifdef BINARY64
  716. ifdef INTERFACE64
  717. ifneq ($(INTERFACE64), 0)
  718. FCOMMON_OPT += -i8
  719. endif
  720. endif
  721. FCOMMON_OPT += -Wall
  722. else
  723. FCOMMON_OPT += -Wall
  724. endif
  725. ifeq ($(USE_OPENMP), 1)
  726. FCOMMON_OPT += -fopenmp
  727. endif
  728. endif
  729. ifeq ($(F_COMPILER), G77)
  730. CCOMMON_OPT += -DF_INTERFACE_G77
  731. FCOMMON_OPT += -Wall
  732. ifndef NO_BINARY_MODE
  733. ifneq ($(OSNAME), AIX)
  734. ifdef BINARY64
  735. FCOMMON_OPT += -m64
  736. else
  737. FCOMMON_OPT += -m32
  738. endif
  739. endif
  740. endif
  741. endif
  742. ifeq ($(F_COMPILER), G95)
  743. CCOMMON_OPT += -DF_INTERFACE_G95
  744. FCOMMON_OPT += -Wall
  745. ifneq ($(OSNAME), AIX)
  746. ifndef NO_BINARY_MODE
  747. ifdef BINARY64
  748. FCOMMON_OPT += -m64
  749. else
  750. FCOMMON_OPT += -m32
  751. endif
  752. endif
  753. ifneq ($(NO_LAPACKE), 1)
  754. FCOMMON_OPT += -fno-second-underscore
  755. endif
  756. endif
  757. endif
  758. ifeq ($(F_COMPILER), GFORTRAN)
  759. CCOMMON_OPT += -DF_INTERFACE_GFORT
  760. FCOMMON_OPT += -Wall
  761. # make single-threaded LAPACK calls thread-safe #1847
  762. FCOMMON_OPT += -frecursive
  763. # work around ABI problem with passing single-character arguments
  764. FCOMMON_OPT += -fno-optimize-sibling-calls
  765. #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
  766. ifneq ($(NO_LAPACK), 1)
  767. EXTRALIB += -lgfortran
  768. endif
  769. ifdef NO_BINARY_MODE
  770. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  771. ifdef BINARY64
  772. FCOMMON_OPT += -mabi=64
  773. else
  774. FCOMMON_OPT += -mabi=n32
  775. endif
  776. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  777. FCOMMON_OPT += -mabi=32
  778. endif
  779. else
  780. ifdef BINARY64
  781. ifneq ($(OSNAME), AIX)
  782. FCOMMON_OPT += -m64
  783. endif
  784. ifdef INTERFACE64
  785. ifneq ($(INTERFACE64), 0)
  786. FCOMMON_OPT += -fdefault-integer-8
  787. endif
  788. endif
  789. else
  790. ifneq ($(OSNAME), AIX)
  791. FCOMMON_OPT += -m32
  792. endif
  793. endif
  794. endif
  795. ifeq ($(USE_OPENMP), 1)
  796. FCOMMON_OPT += -fopenmp
  797. endif
  798. endif
  799. ifeq ($(F_COMPILER), INTEL)
  800. CCOMMON_OPT += -DF_INTERFACE_INTEL
  801. ifdef INTERFACE64
  802. ifneq ($(INTERFACE64), 0)
  803. FCOMMON_OPT += -i8
  804. endif
  805. endif
  806. FCOMMON_OPT += -recursive -fp-model strict -assume protect-parens
  807. ifeq ($(USE_OPENMP), 1)
  808. FCOMMON_OPT += -fopenmp
  809. endif
  810. endif
  811. ifeq ($(F_COMPILER), FUJITSU)
  812. CCOMMON_OPT += -DF_INTERFACE_FUJITSU
  813. ifeq ($(USE_OPENMP), 1)
  814. FCOMMON_OPT += -openmp
  815. endif
  816. endif
  817. ifeq ($(F_COMPILER), IBM)
  818. CCOMMON_OPT += -DF_INTERFACE_IBM
  819. # FCOMMON_OPT += -qarch=440
  820. ifdef BINARY64
  821. FCOMMON_OPT += -q64
  822. ifdef INTERFACE64
  823. ifneq ($(INTERFACE64), 0)
  824. FCOMMON_OPT += -qintsize=8
  825. endif
  826. endif
  827. else
  828. FCOMMON_OPT += -q32
  829. endif
  830. ifeq ($(USE_OPENMP), 1)
  831. FCOMMON_OPT += -openmp
  832. endif
  833. endif
  834. ifeq ($(F_COMPILER), PGI)
  835. CCOMMON_OPT += -DF_INTERFACE_PGI
  836. COMMON_PROF += -DPGICOMPILER
  837. ifdef BINARY64
  838. ifdef INTERFACE64
  839. ifneq ($(INTERFACE64), 0)
  840. FCOMMON_OPT += -i8
  841. endif
  842. endif
  843. ifeq ($(ARCH), x86_64)
  844. FCOMMON_OPT += -tp p7-64
  845. else
  846. ifeq ($(ARCH), power)
  847. ifeq ($(CORE), POWER8)
  848. FCOMMON_OPT += -tp pwr8
  849. endif
  850. ifeq ($(CORE), POWER9)
  851. FCOMMON_OPT += -tp pwr9
  852. endif
  853. endif
  854. endif
  855. else
  856. FCOMMON_OPT += -tp p7
  857. endif
  858. FCOMMON_OPT += -Mrecursive
  859. ifeq ($(USE_OPENMP), 1)
  860. FCOMMON_OPT += -mp
  861. endif
  862. endif
  863. ifeq ($(F_COMPILER), PATHSCALE)
  864. CCOMMON_OPT += -DF_INTERFACE_PATHSCALE
  865. ifdef BINARY64
  866. ifdef INTERFACE64
  867. ifneq ($(INTERFACE64), 0)
  868. FCOMMON_OPT += -i8
  869. endif
  870. endif
  871. endif
  872. ifeq ($(USE_OPENMP), 1)
  873. FCOMMON_OPT += -mp
  874. endif
  875. endif
  876. ifeq ($(F_COMPILER), OPEN64)
  877. CCOMMON_OPT += -DF_INTERFACE_OPEN64
  878. ifdef BINARY64
  879. ifdef INTERFACE64
  880. ifneq ($(INTERFACE64), 0)
  881. FCOMMON_OPT += -i8
  882. endif
  883. endif
  884. endif
  885. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  886. ifndef BINARY64
  887. FCOMMON_OPT += -n32
  888. else
  889. FCOMMON_OPT += -n64
  890. endif
  891. ifeq ($(CORE), LOONGSON3A)
  892. FCOMMON_OPT += -loongson3 -static
  893. endif
  894. ifeq ($(CORE), LOONGSON3B)
  895. FCOMMON_OPT += -loongson3 -static
  896. endif
  897. else
  898. ifndef BINARY64
  899. FCOMMON_OPT += -m32
  900. else
  901. FCOMMON_OPT += -m64
  902. endif
  903. endif
  904. ifeq ($(USE_OPENMP), 1)
  905. FEXTRALIB += -lstdc++
  906. FCOMMON_OPT += -mp
  907. endif
  908. endif
  909. ifeq ($(C_COMPILER), OPEN64)
  910. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  911. ifndef BINARY64
  912. CCOMMON_OPT += -n32
  913. else
  914. CCOMMON_OPT += -n64
  915. endif
  916. ifeq ($(CORE), LOONGSON3A)
  917. CCOMMON_OPT += -loongson3 -static
  918. endif
  919. ifeq ($(CORE), LOONGSON3B)
  920. CCOMMON_OPT += -loongson3 -static
  921. endif
  922. else
  923. ifndef BINARY64
  924. CCOMMON_OPT += -m32
  925. else
  926. CCOMMON_OPT += -m64
  927. endif
  928. endif
  929. endif
  930. ifeq ($(C_COMPILER), SUN)
  931. CCOMMON_OPT += -w
  932. ifeq ($(ARCH), x86)
  933. CCOMMON_OPT += -m32
  934. else
  935. FCOMMON_OPT += -m64
  936. endif
  937. endif
  938. ifeq ($(F_COMPILER), SUN)
  939. CCOMMON_OPT += -DF_INTERFACE_SUN
  940. ifeq ($(ARCH), x86)
  941. FCOMMON_OPT += -m32
  942. else
  943. FCOMMON_OPT += -m64
  944. endif
  945. ifeq ($(USE_OPENMP), 1)
  946. FCOMMON_OPT += -xopenmp=parallel
  947. endif
  948. endif
  949. ifeq ($(F_COMPILER), COMPAQ)
  950. CCOMMON_OPT += -DF_INTERFACE_COMPAQ
  951. ifeq ($(USE_OPENMP), 1)
  952. FCOMMON_OPT += -openmp
  953. endif
  954. endif
  955. ifdef BINARY64
  956. ifdef INTERFACE64
  957. ifneq ($(INTERFACE64), 0)
  958. CCOMMON_OPT +=
  959. #-DUSE64BITINT
  960. endif
  961. endif
  962. endif
  963. ifeq ($(NEED_PIC), 1)
  964. ifeq ($(C_COMPILER), IBM)
  965. CCOMMON_OPT += -qpic=large
  966. else
  967. CCOMMON_OPT += -fPIC
  968. endif
  969. ifeq ($(F_COMPILER), SUN)
  970. FCOMMON_OPT += -pic
  971. else
  972. FCOMMON_OPT += -fPIC
  973. endif
  974. endif
  975. ifeq ($(DYNAMIC_ARCH), 1)
  976. CCOMMON_OPT += -DDYNAMIC_ARCH
  977. endif
  978. ifeq ($(DYNAMIC_OLDER), 1)
  979. CCOMMON_OPT += -DDYNAMIC_OLDER
  980. endif
  981. ifeq ($(NO_LAPACK), 1)
  982. CCOMMON_OPT += -DNO_LAPACK
  983. #Disable LAPACK C interface
  984. NO_LAPACKE = 1
  985. endif
  986. ifeq ($(NO_LAPACKE), 1)
  987. CCOMMON_OPT += -DNO_LAPACKE
  988. endif
  989. ifeq ($(NO_AVX), 1)
  990. CCOMMON_OPT += -DNO_AVX
  991. endif
  992. ifeq ($(ARCH), x86)
  993. CCOMMON_OPT += -DNO_AVX
  994. endif
  995. ifeq ($(NO_AVX2), 1)
  996. CCOMMON_OPT += -DNO_AVX2
  997. endif
  998. ifeq ($(NO_AVX512), 1)
  999. CCOMMON_OPT += -DNO_AVX512
  1000. endif
  1001. ifdef SMP
  1002. CCOMMON_OPT += -DSMP_SERVER
  1003. ifeq ($(ARCH), mips64)
  1004. ifneq ($(CORE), LOONGSON3B)
  1005. USE_SIMPLE_THREADED_LEVEL3 = 1
  1006. endif
  1007. endif
  1008. ifeq ($(USE_OPENMP), 1)
  1009. # USE_SIMPLE_THREADED_LEVEL3 = 1
  1010. # NO_AFFINITY = 1
  1011. CCOMMON_OPT += -DUSE_OPENMP
  1012. endif
  1013. ifeq ($(BIGNUMA), 1)
  1014. CCOMMON_OPT += -DBIGNUMA
  1015. endif
  1016. endif
  1017. ifeq ($(NO_WARMUP), 1)
  1018. CCOMMON_OPT += -DNO_WARMUP
  1019. endif
  1020. ifeq ($(CONSISTENT_FPCSR), 1)
  1021. CCOMMON_OPT += -DCONSISTENT_FPCSR
  1022. endif
  1023. # Only for development
  1024. # CCOMMON_OPT += -DPARAMTEST
  1025. # CCOMMON_OPT += -DPREFETCHTEST
  1026. # CCOMMON_OPT += -DNO_SWITCHING
  1027. # USE_PAPI = 1
  1028. ifdef USE_PAPI
  1029. CCOMMON_OPT += -DUSE_PAPI
  1030. EXTRALIB += -lpapi -lperfctr
  1031. endif
  1032. ifdef DYNAMIC_THREADS
  1033. CCOMMON_OPT += -DDYNAMIC_THREADS
  1034. endif
  1035. CCOMMON_OPT += -DMAX_CPU_NUMBER=$(NUM_THREADS)
  1036. CCOMMON_OPT += -DMAX_PARALLEL_NUMBER=$(NUM_PARALLEL)
  1037. ifdef USE_SIMPLE_THREADED_LEVEL3
  1038. CCOMMON_OPT += -DUSE_SIMPLE_THREADED_LEVEL3
  1039. endif
  1040. ifeq ($(USE_TLS), 1)
  1041. CCOMMON_OPT += -DUSE_TLS
  1042. endif
  1043. ifeq ($(BUILD_HALF), 1)
  1044. CCOMMON_OPT += -DBUILD_HALF
  1045. endif
  1046. CCOMMON_OPT += -DVERSION=\"$(VERSION)\"
  1047. ifndef SYMBOLPREFIX
  1048. SYMBOLPREFIX =
  1049. endif
  1050. ifndef SYMBOLSUFFIX
  1051. SYMBOLSUFFIX =
  1052. endif
  1053. ifndef LIBNAMESUFFIX
  1054. LIBNAMEBASE = $(SYMBOLPREFIX)openblas$(SYMBOLSUFFIX)
  1055. else
  1056. LIBNAMEBASE = $(SYMBOLPREFIX)openblas$(SYMBOLSUFFIX)_$(LIBNAMESUFFIX)
  1057. endif
  1058. ifeq ($(OSNAME), CYGWIN_NT)
  1059. LIBPREFIX = cyg$(LIBNAMEBASE)
  1060. else
  1061. LIBPREFIX = lib$(LIBNAMEBASE)
  1062. endif
  1063. KERNELDIR = $(TOPDIR)/kernel/$(ARCH)
  1064. include $(TOPDIR)/Makefile.$(ARCH)
  1065. ifneq ($(C_COMPILER), PGI)
  1066. CCOMMON_OPT += -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME
  1067. endif
  1068. CCOMMON_OPT += -DASMNAME=$(FU)$(*F) -DASMFNAME=$(FU)$(*F)$(BU) -DNAME=$(*F)$(BU) -DCNAME=$(*F) -DCHAR_NAME=\"$(*F)$(BU)\" -DCHAR_CNAME=\"$(*F)\"
  1069. ifeq ($(CORE), PPC440)
  1070. CCOMMON_OPT += -DALLOC_QALLOC
  1071. endif
  1072. ifeq ($(CORE), PPC440FP2)
  1073. STATIC_ALLOCATION = 1
  1074. endif
  1075. ifneq ($(OSNAME), Linux)
  1076. NO_AFFINITY = 1
  1077. endif
  1078. ifneq ($(ARCH), x86_64)
  1079. ifneq ($(ARCH), x86)
  1080. ifneq ($(CORE), LOONGSON3B)
  1081. NO_AFFINITY = 1
  1082. endif
  1083. endif
  1084. endif
  1085. ifdef NO_AFFINITY
  1086. ifeq ($(NO_AFFINITY), 0)
  1087. override undefine NO_AFFINITY
  1088. else
  1089. CCOMMON_OPT += -DNO_AFFINITY
  1090. endif
  1091. endif
  1092. ifdef FUNCTION_PROFILE
  1093. CCOMMON_OPT += -DFUNCTION_PROFILE
  1094. endif
  1095. ifdef HUGETLB_ALLOCATION
  1096. CCOMMON_OPT += -DALLOC_HUGETLB
  1097. endif
  1098. ifdef HUGETLBFILE_ALLOCATION
  1099. CCOMMON_OPT += -DALLOC_HUGETLBFILE -DHUGETLB_FILE_NAME=$(HUGETLBFILE_ALLOCATION)
  1100. endif
  1101. ifdef STATIC_ALLOCATION
  1102. CCOMMON_OPT += -DALLOC_STATIC
  1103. endif
  1104. ifdef DEVICEDRIVER_ALLOCATION
  1105. CCOMMON_OPT += -DALLOC_DEVICEDRIVER -DDEVICEDRIVER_NAME=\"/dev/mapper\"
  1106. endif
  1107. ifdef MIXED_MEMORY_ALLOCATION
  1108. CCOMMON_OPT += -DMIXED_MEMORY_ALLOCATION
  1109. endif
  1110. ifeq ($(OSNAME), SunOS)
  1111. TAR = gtar
  1112. PATCH = gpatch
  1113. GREP = ggrep
  1114. AWK = nawk
  1115. else
  1116. TAR = tar
  1117. PATCH = patch
  1118. GREP = grep
  1119. AWK = awk
  1120. endif
  1121. ifndef MD5SUM
  1122. MD5SUM = md5sum
  1123. endif
  1124. REVISION = -r$(VERSION)
  1125. MAJOR_VERSION = $(word 1,$(subst ., ,$(VERSION)))
  1126. ifeq ($(DEBUG), 1)
  1127. COMMON_OPT += -g
  1128. endif
  1129. ifeq ($(DEBUG), 1)
  1130. FCOMMON_OPT += -g
  1131. endif
  1132. ifndef COMMON_OPT
  1133. COMMON_OPT = -O2
  1134. endif
  1135. ifndef FCOMMON_OPT
  1136. FCOMMON_OPT = -O2 -frecursive
  1137. endif
  1138. override CFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
  1139. override PFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
  1140. override FFLAGS += $(COMMON_OPT) $(FCOMMON_OPT)
  1141. override FPFLAGS += $(FCOMMON_OPT) $(COMMON_PROF)
  1142. #MAKEOVERRIDES =
  1143. ifdef NEED_PIC
  1144. ifeq (,$(findstring PIC,$(FFLAGS)))
  1145. override FFLAGS += -fPIC
  1146. endif
  1147. endif
  1148. #For LAPACK Fortran codes.
  1149. #Disable -fopenmp for LAPACK Fortran codes on Windows.
  1150. ifdef OS_WINDOWS
  1151. LAPACK_FFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FFLAGS))
  1152. LAPACK_FPFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FPFLAGS))
  1153. else
  1154. LAPACK_FFLAGS := $(FFLAGS)
  1155. LAPACK_FPFLAGS := $(FPFLAGS)
  1156. endif
  1157. LAPACK_CFLAGS = $(CFLAGS)
  1158. LAPACK_CFLAGS += -DHAVE_LAPACK_CONFIG_H
  1159. ifdef INTERFACE64
  1160. ifneq ($(INTERFACE64), 0)
  1161. LAPACK_CFLAGS += -DLAPACK_ILP64
  1162. endif
  1163. endif
  1164. ifdef OS_WINDOWS
  1165. LAPACK_CFLAGS += -DOPENBLAS_OS_WINDOWS
  1166. endif
  1167. ifeq ($(C_COMPILER), LSB)
  1168. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1169. endif
  1170. ifndef SUFFIX
  1171. SUFFIX = o
  1172. endif
  1173. ifndef PSUFFIX
  1174. PSUFFIX = po
  1175. endif
  1176. ifndef LIBSUFFIX
  1177. LIBSUFFIX = a
  1178. endif
  1179. ifneq ($(DYNAMIC_ARCH), 1)
  1180. ifndef SMP
  1181. LIBNAME = $(LIBPREFIX)_$(LIBCORE)$(REVISION).$(LIBSUFFIX)
  1182. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)$(REVISION)_p.$(LIBSUFFIX)
  1183. else
  1184. LIBNAME = $(LIBPREFIX)_$(LIBCORE)p$(REVISION).$(LIBSUFFIX)
  1185. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)p$(REVISION)_p.$(LIBSUFFIX)
  1186. endif
  1187. else
  1188. ifndef SMP
  1189. LIBNAME = $(LIBPREFIX)$(REVISION).$(LIBSUFFIX)
  1190. LIBNAME_P = $(LIBPREFIX)$(REVISION)_p.$(LIBSUFFIX)
  1191. else
  1192. LIBNAME = $(LIBPREFIX)p$(REVISION).$(LIBSUFFIX)
  1193. LIBNAME_P = $(LIBPREFIX)p$(REVISION)_p.$(LIBSUFFIX)
  1194. endif
  1195. endif
  1196. LIBDLLNAME = $(LIBPREFIX).dll
  1197. IMPLIBNAME = lib$(LIBNAMEBASE).dll.a
  1198. ifneq ($(OSNAME), AIX)
  1199. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.so)
  1200. else
  1201. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.a)
  1202. endif
  1203. LIBDYNNAME = $(LIBNAME:.$(LIBSUFFIX)=.dylib)
  1204. LIBDEFNAME = $(LIBNAME:.$(LIBSUFFIX)=.def)
  1205. LIBEXPNAME = $(LIBNAME:.$(LIBSUFFIX)=.exp)
  1206. LIBZIPNAME = $(LIBNAME:.$(LIBSUFFIX)=.zip)
  1207. LIBS = $(TOPDIR)/$(LIBNAME)
  1208. LIBS_P = $(TOPDIR)/$(LIBNAME_P)
  1209. LIB_COMPONENTS = BLAS
  1210. ifneq ($(NO_CBLAS), 1)
  1211. LIB_COMPONENTS += CBLAS
  1212. endif
  1213. ifneq ($(NO_LAPACK), 1)
  1214. LIB_COMPONENTS += LAPACK
  1215. ifneq ($(NO_LAPACKE), 1)
  1216. LIB_COMPONENTS += LAPACKE
  1217. endif
  1218. ifeq ($(BUILD_RELAPACK), 1)
  1219. LIB_COMPONENTS += ReLAPACK
  1220. endif
  1221. endif
  1222. ifeq ($(ONLY_CBLAS), 1)
  1223. LIB_COMPONENTS = CBLAS
  1224. endif
  1225. export OSNAME
  1226. export ARCH
  1227. export CORE
  1228. export LIBCORE
  1229. export __BYTE_ORDER__
  1230. export ELF_VERSION
  1231. export PGCPATH
  1232. export CONFIG
  1233. export CC
  1234. export FC
  1235. export BU
  1236. export FU
  1237. export NEED2UNDERSCORES
  1238. export USE_THREAD
  1239. export NUM_THREADS
  1240. export NUM_CORES
  1241. export SMP
  1242. export MAKEFILE_RULE
  1243. export NEED_PIC
  1244. export BINARY
  1245. export BINARY32
  1246. export BINARY64
  1247. export F_COMPILER
  1248. export C_COMPILER
  1249. export USE_OPENMP
  1250. export CROSS
  1251. export CROSS_SUFFIX
  1252. export NOFORTRAN
  1253. export NO_FBLAS
  1254. export EXTRALIB
  1255. export CEXTRALIB
  1256. export FEXTRALIB
  1257. export HAVE_SSE
  1258. export HAVE_SSE2
  1259. export HAVE_SSE3
  1260. export HAVE_SSSE3
  1261. export HAVE_SSE4_1
  1262. export HAVE_SSE4_2
  1263. export HAVE_SSE4A
  1264. export HAVE_SSE5
  1265. export HAVE_AVX
  1266. export HAVE_VFP
  1267. export HAVE_VFPV3
  1268. export HAVE_VFPV4
  1269. export HAVE_NEON
  1270. export HAVE_MSA
  1271. export MSA_FLAGS
  1272. export KERNELDIR
  1273. export FUNCTION_PROFILE
  1274. export TARGET_CORE
  1275. export NO_AVX512
  1276. export BUILD_HALF
  1277. export SHGEMM_UNROLL_M
  1278. export SHGEMM_UNROLL_N
  1279. export SGEMM_UNROLL_M
  1280. export SGEMM_UNROLL_N
  1281. export DGEMM_UNROLL_M
  1282. export DGEMM_UNROLL_N
  1283. export QGEMM_UNROLL_M
  1284. export QGEMM_UNROLL_N
  1285. export CGEMM_UNROLL_M
  1286. export CGEMM_UNROLL_N
  1287. export ZGEMM_UNROLL_M
  1288. export ZGEMM_UNROLL_N
  1289. export XGEMM_UNROLL_M
  1290. export XGEMM_UNROLL_N
  1291. export CGEMM3M_UNROLL_M
  1292. export CGEMM3M_UNROLL_N
  1293. export ZGEMM3M_UNROLL_M
  1294. export ZGEMM3M_UNROLL_N
  1295. export XGEMM3M_UNROLL_M
  1296. export XGEMM3M_UNROLL_N
  1297. ifdef USE_CUDA
  1298. export CUDADIR
  1299. export CUCC
  1300. export CUFLAGS
  1301. export CULIB
  1302. endif
  1303. .SUFFIXES: .$(PSUFFIX) .$(SUFFIX) .f
  1304. .f.$(SUFFIX):
  1305. $(FC) $(FFLAGS) -c $< -o $(@F)
  1306. .f.$(PSUFFIX):
  1307. $(FC) $(FPFLAGS) -pg -c $< -o $(@F)
  1308. ifdef BINARY64
  1309. PATHSCALEPATH = /opt/pathscale/lib/3.1
  1310. PGIPATH = /opt/pgi/linux86-64/7.1-5/lib
  1311. else
  1312. PATHSCALEPATH = /opt/pathscale/lib/3.1/32
  1313. PGIPATH = /opt/pgi/linux86/7.1-5/lib
  1314. endif
  1315. ACMLPATH = /opt/acml/4.3.0
  1316. ifneq ($(OSNAME), Darwin)
  1317. MKLPATH = /opt/intel/mkl/10.2.2.025/lib
  1318. else
  1319. MKLPATH = /Library/Frameworks/Intel_MKL.framework/Versions/10.0.1.014/lib
  1320. endif
  1321. ATLASPATH = /opt/atlas/3.9.17/opteron
  1322. FLAMEPATH = $(HOME)/flame/lib
  1323. ifneq ($(OSNAME), SunOS)
  1324. SUNPATH = /opt/sunstudio12.1
  1325. else
  1326. SUNPATH = /opt/SUNWspro
  1327. endif