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 32 kB

7 years ago
13 years ago
1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285128612871288128912901291129212931294129512961297129812991300130113021303130413051306130713081309131013111312131313141315131613171318131913201321132213231324132513261327132813291330133113321333133413351336133713381339134013411342134313441345134613471348134913501351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384138513861387138813891390139113921393139413951396139713981399140014011402140314041405140614071408140914101411141214131414141514161417141814191420142114221423142414251426142714281429143014311432143314341435143614371438143914401441144214431444144514461447144814491450145114521453145414551456145714581459146014611462146314641465146614671468146914701471147214731474147514761477147814791480148114821483148414851486148714881489149014911492149314941495149614971498149915001501150215031504150515061507150815091510151115121513151415151516151715181519152015211522152315241525152615271528152915301531153215331534153515361537153815391540154115421543154415451546154715481549155015511552155315541555155615571558155915601561156215631564156515661567156815691570157115721573157415751576157715781579158015811582158315841585158615871588158915901591159215931594159515961597159815991600160116021603160416051606
  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. ifneq ($(BUILD_SINGLE),1)
  66. ifneq ($(BUILD_DOUBLE),1)
  67. ifneq ($(BUILD_COMPLEX),1)
  68. ifneq ($(BUILD_COMPLEX16),1)
  69. override BUILD_SINGLE=1
  70. override BUILD_DOUBLE=1
  71. override BUILD_COMPLEX=1
  72. override BUILD_COMPLEX16=1
  73. endif
  74. endif
  75. endif
  76. endif
  77. ifndef HOSTCC
  78. HOSTCC = $(CC)
  79. endif
  80. ifdef TARGET
  81. GETARCH_FLAGS := -DFORCE_$(TARGET)
  82. GETARCH_FLAGS += -DUSER_TARGET
  83. ifeq ($(TARGET), GENERIC)
  84. ifeq ($(DYNAMIC_ARCH), 1)
  85. override NO_EXPRECISION=1
  86. export NO_EXPRECiSION
  87. endif
  88. endif
  89. endif
  90. # Force fallbacks for 32bit
  91. ifeq ($(BINARY), 32)
  92. ifeq ($(TARGET), HASWELL)
  93. GETARCH_FLAGS := -DFORCE_NEHALEM
  94. endif
  95. ifeq ($(TARGET), SKYLAKEX)
  96. GETARCH_FLAGS := -DFORCE_NEHALEM
  97. endif
  98. ifeq ($(TARGET), COOPERLAKE)
  99. GETARCH_FLAGS := -DFORCE_NEHALEM
  100. endif
  101. ifeq ($(TARGET), SANDYBRIDGE)
  102. GETARCH_FLAGS := -DFORCE_NEHALEM
  103. endif
  104. ifeq ($(TARGET), BULLDOZER)
  105. GETARCH_FLAGS := -DFORCE_BARCELONA
  106. endif
  107. ifeq ($(TARGET), PILEDRIVER)
  108. GETARCH_FLAGS := -DFORCE_BARCELONA
  109. endif
  110. ifeq ($(TARGET), STEAMROLLER)
  111. GETARCH_FLAGS := -DFORCE_BARCELONA
  112. endif
  113. ifeq ($(TARGET), EXCAVATOR)
  114. GETARCH_FLAGS := -DFORCE_BARCELONA
  115. endif
  116. ifeq ($(TARGET), ZEN)
  117. GETARCH_FLAGS := -DFORCE_BARCELONA
  118. endif
  119. ifeq ($(TARGET), ARMV8)
  120. GETARCH_FLAGS := -DFORCE_ARMV7
  121. endif
  122. ifeq ($(TARGET), POWER8)
  123. GETARCH_FLAGS := -DFORCE_POWER6
  124. endif
  125. endif
  126. #TARGET_CORE will override TARGET which is used in DYNAMIC_ARCH=1.
  127. #
  128. ifdef TARGET_CORE
  129. GETARCH_FLAGS := -DFORCE_$(TARGET_CORE)
  130. endif
  131. # Force fallbacks for 32bit
  132. ifeq ($(BINARY), 32)
  133. ifeq ($(TARGET_CORE), HASWELL)
  134. GETARCH_FLAGS := -DFORCE_NEHALEM
  135. endif
  136. ifeq ($(TARGET_CORE), SKYLAKEX)
  137. GETARCH_FLAGS := -DFORCE_NEHALEM
  138. endif
  139. ifeq ($(TARGET_CORE), COOPERLAKE)
  140. GETARCH_FLAGS := -DFORCE_NEHALEM
  141. endif
  142. ifeq ($(TARGET_CORE), SANDYBRIDGE)
  143. GETARCH_FLAGS := -DFORCE_NEHALEM
  144. endif
  145. ifeq ($(TARGET_CORE), BULLDOZER)
  146. GETARCH_FLAGS := -DFORCE_BARCELONA
  147. endif
  148. ifeq ($(TARGET_CORE), PILEDRIVER)
  149. GETARCH_FLAGS := -DFORCE_BARCELONA
  150. endif
  151. ifeq ($(TARGET_CORE), STEAMROLLER)
  152. GETARCH_FLAGS := -DFORCE_BARCELONA
  153. endif
  154. ifeq ($(TARGET_CORE), EXCAVATOR)
  155. GETARCH_FLAGS := -DFORCE_BARCELONA
  156. endif
  157. ifeq ($(TARGET_CORE), ZEN)
  158. GETARCH_FLAGS := -DFORCE_BARCELONA
  159. endif
  160. endif
  161. # On x86_64 build getarch with march=native unless the compiler is PGI. This is required to detect AVX512 support in getarch.
  162. ifeq ($(HOSTARCH), x86_64)
  163. ifeq ($(findstring pgcc,$(HOSTCC)),)
  164. GETARCH_FLAGS += -march=native
  165. endif
  166. endif
  167. ifdef INTERFACE64
  168. ifneq ($(INTERFACE64), 0)
  169. GETARCH_FLAGS += -DUSE64BITINT
  170. endif
  171. endif
  172. ifndef GEMM_MULTITHREAD_THRESHOLD
  173. GEMM_MULTITHREAD_THRESHOLD=4
  174. endif
  175. GETARCH_FLAGS += -DGEMM_MULTITHREAD_THRESHOLD=$(GEMM_MULTITHREAD_THRESHOLD)
  176. ifeq ($(NO_AVX), 1)
  177. GETARCH_FLAGS += -DNO_AVX
  178. endif
  179. ifeq ($(BINARY), 32)
  180. GETARCH_FLAGS += -DNO_AVX -DNO_AVX2 -DNO_AVX512
  181. NO_AVX512 = 1
  182. endif
  183. ifeq ($(NO_AVX2), 1)
  184. GETARCH_FLAGS += -DNO_AVX2
  185. endif
  186. ifeq ($(NO_AVX512), 1)
  187. GETARCH_FLAGS += -DNO_AVX512
  188. endif
  189. ifeq ($(DEBUG), 1)
  190. GETARCH_FLAGS += -g
  191. endif
  192. ifeq ($(QUIET_MAKE), 1)
  193. MAKE += -s
  194. endif
  195. ifndef NO_PARALLEL_MAKE
  196. NO_PARALLEL_MAKE=0
  197. endif
  198. GETARCH_FLAGS += -DNO_PARALLEL_MAKE=$(NO_PARALLEL_MAKE)
  199. ifdef MAKE_NB_JOBS
  200. GETARCH_FLAGS += -DMAKE_NB_JOBS=$(MAKE_NB_JOBS)
  201. endif
  202. ifeq ($(HOSTCC), loongcc)
  203. GETARCH_FLAGS += -static
  204. endif
  205. #if don't use Fortran, it will only compile CBLAS.
  206. ifeq ($(ONLY_CBLAS), 1)
  207. NO_LAPACK = 1
  208. else
  209. ONLY_CBLAS = 0
  210. endif
  211. # This operation is expensive, so execution should be once.
  212. ifndef GOTOBLAS_MAKEFILE
  213. export GOTOBLAS_MAKEFILE = 1
  214. # Generating Makefile.conf and config.h
  215. 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)
  216. ifndef TARGET_CORE
  217. include $(TOPDIR)/Makefile.conf
  218. else
  219. include $(TOPDIR)/Makefile_kernel.conf
  220. endif
  221. endif
  222. ifndef NUM_PARALLEL
  223. NUM_PARALLEL = 1
  224. endif
  225. ifndef NUM_THREADS
  226. NUM_THREADS = $(NUM_CORES)
  227. endif
  228. ifeq ($(NUM_THREADS), 1)
  229. override USE_THREAD = 0
  230. override USE_OPENMP = 0
  231. endif
  232. ifdef USE_THREAD
  233. ifeq ($(USE_THREAD), 0)
  234. SMP =
  235. else
  236. SMP = 1
  237. endif
  238. else
  239. ifeq ($(NUM_THREAD), 1)
  240. SMP =
  241. else
  242. SMP = 1
  243. endif
  244. endif
  245. ifeq ($(SMP), 1)
  246. USE_LOCKING =
  247. endif
  248. ifndef NEED_PIC
  249. NEED_PIC = 1
  250. endif
  251. ARFLAGS =
  252. CPP = $(COMPILER) -E
  253. AR ?= $(CROSS_SUFFIX)ar
  254. AS ?= $(CROSS_SUFFIX)as
  255. LD ?= $(CROSS_SUFFIX)ld
  256. RANLIB ?= $(CROSS_SUFFIX)ranlib
  257. NM = $(CROSS_SUFFIX)nm
  258. DLLWRAP = $(CROSS_SUFFIX)dllwrap
  259. OBJCOPY = $(CROSS_SUFFIX)objcopy
  260. OBJCONV = $(CROSS_SUFFIX)objconv
  261. # When fortran support was either not detected or actively deselected, only build BLAS.
  262. ifeq ($(NOFORTRAN), 1)
  263. NO_LAPACK = 1
  264. override FEXTRALIB =
  265. endif
  266. ifeq ($(C_COMPILER), GCC)
  267. GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
  268. GCCVERSIONGT4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 4)
  269. GCCVERSIONGT5 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 5)
  270. GCCVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 7)
  271. GCCVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  272. GCCVERSIONGTEQ11 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 11)
  273. GCCVERSIONGTEQ10 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 10)
  274. # Note that the behavior of -dumpversion is compile-time-configurable for
  275. # gcc-7.x and newer. Use -dumpfullversion there
  276. ifeq ($(GCCVERSIONGTEQ7),1)
  277. GCCDUMPVERSION_PARAM := -dumpfullversion
  278. else
  279. GCCDUMPVERSION_PARAM := -dumpversion
  280. endif
  281. GCCMINORVERSIONGTEQ1 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 1)
  282. GCCMINORVERSIONGTEQ2 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 2)
  283. GCCMINORVERSIONGTEQ7 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 7)
  284. endif
  285. #
  286. # OS dependent settings
  287. #
  288. ifeq ($(OSNAME), Darwin)
  289. ifndef MACOSX_DEPLOYMENT_TARGET
  290. export MACOSX_DEPLOYMENT_TARGET=10.8
  291. endif
  292. MD5SUM = md5 -r
  293. endif
  294. ifneq (,$(findstring $(OSNAME), FreeBSD OpenBSD DragonFly))
  295. MD5SUM = md5 -r
  296. endif
  297. ifeq ($(OSNAME), NetBSD)
  298. MD5SUM = md5 -n
  299. endif
  300. ifeq ($(OSNAME), Linux)
  301. EXTRALIB += -lm
  302. NO_EXPRECISION = 1
  303. endif
  304. ifeq ($(OSNAME), Android)
  305. EXTRALIB += -lm
  306. endif
  307. ifeq ($(OSNAME), AIX)
  308. EXTRALIB += -lm
  309. endif
  310. ifeq ($(OSNAME), WINNT)
  311. NEED_PIC = 0
  312. NO_EXPRECISION = 1
  313. EXTRALIB += -defaultlib:advapi32
  314. SUFFIX = obj
  315. PSUFFIX = pobj
  316. LIBSUFFIX = a
  317. ifeq ($(C_COMPILER), CLANG)
  318. CCOMMON_OPT += -DMS_ABI
  319. endif
  320. #Version tests for supporting specific features (MS_ABI, POWER9 intrinsics)
  321. ifeq ($(GCCVERSIONGT4), 1)
  322. # GCC Major version > 4
  323. # It is compatible with MSVC ABI.
  324. CCOMMON_OPT += -DMS_ABI
  325. endif
  326. ifeq ($(GCCVERSIONGTEQ4), 1)
  327. ifeq ($(GCCMINORVERSIONGTEQ7), 1)
  328. # GCC Version >=4.7
  329. # It is compatible with MSVC ABI.
  330. CCOMMON_OPT += -DMS_ABI
  331. endif
  332. endif
  333. # Ensure the correct stack alignment on Win32
  334. # http://permalink.gmane.org/gmane.comp.lib.openblas.general/97
  335. ifeq ($(ARCH), x86)
  336. CCOMMON_OPT += -mincoming-stack-boundary=2
  337. FCOMMON_OPT += -mincoming-stack-boundary=2
  338. endif
  339. endif
  340. ifeq ($(OSNAME), Interix)
  341. NEED_PIC = 0
  342. NO_EXPRECISION = 1
  343. INTERIX_TOOL_DIR = /opt/gcc.3.3/i586-pc-interix3/bin
  344. endif
  345. ifeq ($(OSNAME), CYGWIN_NT)
  346. NEED_PIC = 0
  347. NO_EXPRECISION = 1
  348. OS_CYGWIN_NT = 1
  349. endif
  350. ifneq ($(OSNAME), WINNT)
  351. ifneq ($(OSNAME), CYGWIN_NT)
  352. ifneq ($(OSNAME), Interix)
  353. ifneq ($(OSNAME), Android)
  354. ifdef SMP
  355. EXTRALIB += -lpthread
  356. endif
  357. endif
  358. endif
  359. endif
  360. endif
  361. # ifeq logical or
  362. ifeq ($(OSNAME), $(filter $(OSNAME),WINNT CYGWIN_NT Interix))
  363. OS_WINDOWS=1
  364. endif
  365. ifdef QUAD_PRECISION
  366. CCOMMON_OPT += -DQUAD_PRECISION
  367. NO_EXPRECISION = 1
  368. endif
  369. ifneq ($(ARCH), x86)
  370. ifneq ($(ARCH), x86_64)
  371. NO_EXPRECISION = 1
  372. endif
  373. endif
  374. ifdef UTEST_CHECK
  375. CCOMMON_OPT += -DUTEST_CHECK
  376. SANITY_CHECK = 1
  377. endif
  378. ifdef SANITY_CHECK
  379. CCOMMON_OPT += -DSANITY_CHECK -DREFNAME=$(*F)f$(BU)
  380. endif
  381. MAX_STACK_ALLOC ?= 2048
  382. ifneq ($(MAX_STACK_ALLOC), 0)
  383. CCOMMON_OPT += -DMAX_STACK_ALLOC=$(MAX_STACK_ALLOC)
  384. endif
  385. ifdef USE_LOCKING
  386. ifneq ($(USE_LOCKING), 0)
  387. CCOMMON_OPT += -DUSE_LOCKING
  388. endif
  389. endif
  390. #
  391. # Architecture dependent settings
  392. #
  393. ifeq ($(ARCH), x86)
  394. ifndef BINARY
  395. NO_BINARY_MODE = 1
  396. endif
  397. ifeq ($(CORE), generic)
  398. NO_EXPRECISION = 1
  399. endif
  400. ifndef NO_EXPRECISION
  401. ifeq ($(F_COMPILER), GFORTRAN)
  402. # ifeq logical or. GCC or LSB
  403. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  404. EXPRECISION = 1
  405. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  406. FCOMMON_OPT += -m128bit-long-double
  407. endif
  408. ifeq ($(C_COMPILER), CLANG)
  409. EXPRECISION = 1
  410. CCOMMON_OPT += -DEXPRECISION
  411. FCOMMON_OPT += -m128bit-long-double
  412. endif
  413. endif
  414. endif
  415. endif
  416. ifeq ($(ARCH), x86_64)
  417. ifeq ($(CORE), generic)
  418. NO_EXPRECISION = 1
  419. endif
  420. ifndef NO_EXPRECISION
  421. ifeq ($(F_COMPILER), GFORTRAN)
  422. # ifeq logical or. GCC or LSB
  423. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  424. EXPRECISION = 1
  425. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  426. FCOMMON_OPT += -m128bit-long-double
  427. endif
  428. ifeq ($(C_COMPILER), CLANG)
  429. EXPRECISION = 1
  430. CCOMMON_OPT += -DEXPRECISION
  431. FCOMMON_OPT += -m128bit-long-double
  432. endif
  433. endif
  434. endif
  435. endif
  436. ifeq ($(C_COMPILER), INTEL)
  437. CCOMMON_OPT += -wd981
  438. endif
  439. ifeq ($(USE_OPENMP), 1)
  440. #check
  441. ifeq ($(USE_THREAD), 0)
  442. $(error OpenBLAS: Cannot set both USE_OPENMP=1 and USE_THREAD=0. The USE_THREAD=0 is only for building single thread version.)
  443. endif
  444. # ifeq logical or. GCC or LSB
  445. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  446. CCOMMON_OPT += -fopenmp
  447. endif
  448. ifeq ($(C_COMPILER), CLANG)
  449. CCOMMON_OPT += -fopenmp
  450. endif
  451. ifeq ($(C_COMPILER), INTEL)
  452. CCOMMON_OPT += -fopenmp
  453. endif
  454. ifeq ($(C_COMPILER), PGI)
  455. CCOMMON_OPT += -mp
  456. endif
  457. ifeq ($(C_COMPILER), OPEN64)
  458. CCOMMON_OPT += -mp
  459. CEXTRALIB += -lstdc++
  460. endif
  461. ifeq ($(C_COMPILER), PATHSCALE)
  462. CCOMMON_OPT += -mp
  463. endif
  464. endif
  465. ifeq ($(DYNAMIC_ARCH), 1)
  466. ifeq ($(ARCH), x86)
  467. DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
  468. CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
  469. endif
  470. ifeq ($(ARCH), x86_64)
  471. DYNAMIC_CORE = PRESCOTT CORE2
  472. ifeq ($(DYNAMIC_OLDER), 1)
  473. DYNAMIC_CORE += PENRYN DUNNINGTON
  474. endif
  475. DYNAMIC_CORE += NEHALEM
  476. ifeq ($(DYNAMIC_OLDER), 1)
  477. DYNAMIC_CORE += OPTERON OPTERON_SSE3
  478. endif
  479. DYNAMIC_CORE += BARCELONA
  480. ifeq ($(DYNAMIC_OLDER), 1)
  481. DYNAMIC_CORE += BOBCAT ATOM NANO
  482. endif
  483. ifneq ($(NO_AVX), 1)
  484. DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER STEAMROLLER EXCAVATOR
  485. endif
  486. ifneq ($(NO_AVX2), 1)
  487. DYNAMIC_CORE += HASWELL ZEN
  488. endif
  489. ifneq ($(NO_AVX512), 1)
  490. ifneq ($(NO_AVX2), 1)
  491. DYNAMIC_CORE += SKYLAKEX COOPERLAKE
  492. endif
  493. endif
  494. endif
  495. ifdef DYNAMIC_LIST
  496. override DYNAMIC_CORE = PRESCOTT $(DYNAMIC_LIST)
  497. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_PRESCOTT
  498. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  499. CCOMMON_OPT += $(XCCOMMON_OPT)
  500. #CCOMMON_OPT += -DDYNAMIC_LIST='$(DYNAMIC_LIST)'
  501. endif
  502. ifeq ($(ARCH), arm64)
  503. DYNAMIC_CORE = ARMV8
  504. DYNAMIC_CORE += CORTEXA53
  505. DYNAMIC_CORE += CORTEXA57
  506. DYNAMIC_CORE += CORTEXA72
  507. DYNAMIC_CORE += CORTEXA73
  508. DYNAMIC_CORE += NEOVERSEN1
  509. DYNAMIC_CORE += FALKOR
  510. DYNAMIC_CORE += THUNDERX
  511. DYNAMIC_CORE += THUNDERX2T99
  512. DYNAMIC_CORE += TSV110
  513. DYNAMIC_CORE += EMAG8180
  514. DYNAMIC_CORE += THUNDERX3T110
  515. endif
  516. ifeq ($(ARCH), zarch)
  517. DYNAMIC_CORE = ZARCH_GENERIC
  518. # if the compiler accepts -march=arch11 or -march=z13 and can compile a file
  519. # with z13-specific inline assembly, then we can include support for Z13.
  520. # note: -march=z13 is equivalent to -march=arch11 yet some compiler releases
  521. # only support one or the other.
  522. # note: LLVM version 6.x supported -march=z13 yet could not handle vector
  523. # registers in inline assembly, so the check for supporting the -march flag is
  524. # not enough.
  525. ZARCH_TEST_COMPILE=-c $(TOPDIR)/kernel/zarch/damin_z13.c -I$(TOPDIR) -o /dev/null > /dev/null 2> /dev/null
  526. ZARCH_CC_SUPPORTS_ARCH11=$(shell $(CC) -march=arch11 $(ZARCH_TEST_COMPILE) && echo 1)
  527. ZARCH_CC_SUPPORTS_Z13=$(shell $(CC) -march=z13 $(ZARCH_TEST_COMPILE) && echo 1)
  528. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH11), $(ZARCH_CC_SUPPORTS_Z13)), 1)
  529. DYNAMIC_CORE += Z13
  530. CCOMMON_OPT += -DDYN_Z13
  531. else
  532. $(info OpenBLAS: Not building Z13 kernels because the compiler $(CC) does not support it)
  533. endif
  534. # as above for z13, check for -march=arch12 and z14 support in the compiler.
  535. ZARCH_CC_SUPPORTS_ARCH12=$(shell $(CC) -march=arch12 $(ZARCH_TEST_COMPILE) && echo 1)
  536. ZARCH_CC_SUPPORTS_Z14=$(shell $(CC) -march=z14 $(ZARCH_TEST_COMPILE) && echo 1)
  537. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH12), $(ZARCH_CC_SUPPORTS_Z14)), 1)
  538. DYNAMIC_CORE += Z14
  539. CCOMMON_OPT += -DDYN_Z14
  540. else
  541. $(info OpenBLAS: Not building Z14 kernels because the compiler $(CC) does not support it)
  542. endif
  543. endif # ARCH zarch
  544. ifeq ($(ARCH), power)
  545. DYNAMIC_CORE = POWER6
  546. DYNAMIC_CORE += POWER8
  547. ifneq ($(C_COMPILER), GCC)
  548. DYNAMIC_CORE += POWER9
  549. DYNAMIC_CORE += POWER10
  550. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  551. endif
  552. ifeq ($(C_COMPILER), GCC)
  553. ifeq ($(GCCVERSIONGT5), 1)
  554. DYNAMIC_CORE += POWER9
  555. else
  556. $(info, OpenBLAS: Your gcc version is too old to build the POWER9 kernels.)
  557. endif
  558. LDVERSIONGTEQ35 := $(shell expr `ld --version | head -1 | cut -f2 -d "." | cut -f1 -d "-"` >= 35)
  559. ifeq ($(GCCVERSIONGTEQ11)$(LDVERSIONGTEQ35), 11)
  560. DYNAMIC_CORE += POWER10
  561. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  562. else ifeq ($(GCCVERSIONGTEQ10), 1)
  563. ifeq ($(GCCMINORVERSIONGTEQ2)$(LDVERSIONGTEQ35), 11)
  564. DYNAMIC_CORE += POWER10
  565. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  566. endif
  567. else
  568. $(info, OpenBLAS: Your gcc version is too old to build the POWER10 kernels.)
  569. endif
  570. endif
  571. endif
  572. # If DYNAMIC_CORE is not set, DYNAMIC_ARCH cannot do anything, so force it to empty
  573. ifndef DYNAMIC_CORE
  574. override DYNAMIC_ARCH=
  575. endif
  576. endif
  577. ifeq ($(ARCH), ia64)
  578. NO_BINARY_MODE = 1
  579. BINARY_DEFINED = 1
  580. ifeq ($(F_COMPILER), GFORTRAN)
  581. ifeq ($(C_COMPILER), GCC)
  582. # EXPRECISION = 1
  583. # CCOMMON_OPT += -DEXPRECISION
  584. endif
  585. endif
  586. endif
  587. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  588. NO_BINARY_MODE = 1
  589. endif
  590. ifeq ($(ARCH), alpha)
  591. NO_BINARY_MODE = 1
  592. BINARY_DEFINED = 1
  593. endif
  594. ifeq ($(ARCH), arm)
  595. NO_BINARY_MODE = 1
  596. BINARY_DEFINED = 1
  597. CCOMMON_OPT += -marm
  598. FCOMMON_OPT += -marm
  599. # If softfp abi is mentioned on the command line, force it.
  600. ifeq ($(ARM_SOFTFP_ABI), 1)
  601. CCOMMON_OPT += -mfloat-abi=softfp
  602. FCOMMON_OPT += -mfloat-abi=softfp
  603. endif
  604. ifeq ($(OSNAME), Android)
  605. ifeq ($(ARM_SOFTFP_ABI), 1)
  606. EXTRALIB += -lm
  607. else
  608. EXTRALIB += -Wl,-lm_hard
  609. endif
  610. endif
  611. endif
  612. ifeq ($(ARCH), arm64)
  613. NO_BINARY_MODE = 1
  614. BINARY_DEFINED = 1
  615. ifdef INTERFACE64
  616. ifneq ($(INTERFACE64), 0)
  617. ifeq ($(F_COMPILER), GFORTRAN)
  618. FCOMMON_OPT += -fdefault-integer-8
  619. endif
  620. ifeq ($(F_COMPILER), FLANG)
  621. FCOMMON_OPT += -i8
  622. endif
  623. endif
  624. endif
  625. endif
  626. #
  627. # C Compiler dependent settings
  628. #
  629. # ifeq logical or. GCC or CLANG or LSB
  630. # http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
  631. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG LSB))
  632. CCOMMON_OPT += -Wall
  633. COMMON_PROF += -fno-inline
  634. NO_UNINITIALIZED_WARN = -Wno-uninitialized
  635. ifeq ($(QUIET_MAKE), 1)
  636. CCOMMON_OPT += $(NO_UNINITIALIZED_WARN) -Wno-unused
  637. endif
  638. ifdef NO_BINARY_MODE
  639. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  640. ifdef BINARY64
  641. CCOMMON_OPT += -mabi=64
  642. else
  643. CCOMMON_OPT += -mabi=n32
  644. endif
  645. BINARY_DEFINED = 1
  646. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  647. CCOMMON_OPT += -mabi=32
  648. BINARY_DEFINED = 1
  649. endif
  650. ifeq ($(CORE), LOONGSON3A)
  651. CCOMMON_OPT += -march=mips64
  652. FCOMMON_OPT += -march=mips64
  653. endif
  654. ifeq ($(CORE), LOONGSON3B)
  655. CCOMMON_OPT += -march=mips64
  656. FCOMMON_OPT += -march=mips64
  657. endif
  658. ifeq ($(CORE), MIPS24K)
  659. CCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  660. FCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  661. endif
  662. ifeq ($(CORE), MIPS1004K)
  663. CCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  664. FCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  665. endif
  666. ifeq ($(CORE), P5600)
  667. CCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  668. FCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  669. endif
  670. ifeq ($(CORE), I6400)
  671. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  672. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  673. endif
  674. ifeq ($(CORE), P6600)
  675. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  676. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  677. endif
  678. ifeq ($(CORE), I6500)
  679. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  680. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  681. endif
  682. ifeq ($(OSNAME), AIX)
  683. BINARY_DEFINED = 1
  684. endif
  685. endif
  686. ifndef BINARY_DEFINED
  687. ifneq ($(OSNAME), AIX)
  688. ifdef BINARY64
  689. CCOMMON_OPT += -m64
  690. else
  691. CCOMMON_OPT += -m32
  692. endif
  693. endif
  694. endif
  695. endif
  696. ifeq ($(C_COMPILER), PGI)
  697. ifdef BINARY64
  698. ifeq ($(ARCH), x86_64)
  699. CCOMMON_OPT += -tp p7-64 -D__MMX__ -Mnollvm
  700. else
  701. ifeq ($(ARCH), power)
  702. ifeq ($(CORE), POWER8)
  703. CCOMMON_OPT += -tp pwr8
  704. endif
  705. ifeq ($(CORE), POWER9)
  706. CCOMMON_OPT += -tp pwr9
  707. endif
  708. endif
  709. endif
  710. else
  711. CCOMMON_OPT += -tp p7
  712. endif
  713. endif
  714. ifeq ($(C_COMPILER), PATHSCALE)
  715. ifdef BINARY64
  716. CCOMMON_OPT += -m64
  717. else
  718. CCOMMON_OPT += -m32
  719. endif
  720. endif
  721. #
  722. # Fortran Compiler dependent settings
  723. #
  724. ifeq ($(F_COMPILER), FLANG)
  725. CCOMMON_OPT += -DF_INTERFACE_FLANG
  726. FCOMMON_OPT += -Mrecursive -Kieee
  727. ifeq ($(OSNAME), Linux)
  728. ifeq ($(ARCH), x86_64)
  729. FLANG_VENDOR := $(shell `$(FC) --version|cut -f 1 -d "."|head -1`)
  730. ifeq ($(FLANG_VENDOR),AOCC)
  731. FCOMMON_OPT += -fno-unroll-loops
  732. endif
  733. endif
  734. endif
  735. ifdef BINARY64
  736. ifdef INTERFACE64
  737. ifneq ($(INTERFACE64), 0)
  738. FCOMMON_OPT += -i8
  739. endif
  740. endif
  741. FCOMMON_OPT += -Wall
  742. else
  743. FCOMMON_OPT += -Wall
  744. endif
  745. ifeq ($(USE_OPENMP), 1)
  746. FCOMMON_OPT += -fopenmp
  747. endif
  748. endif
  749. ifeq ($(F_COMPILER), G77)
  750. CCOMMON_OPT += -DF_INTERFACE_G77
  751. FCOMMON_OPT += -Wall
  752. ifndef NO_BINARY_MODE
  753. ifneq ($(OSNAME), AIX)
  754. ifdef BINARY64
  755. FCOMMON_OPT += -m64
  756. else
  757. FCOMMON_OPT += -m32
  758. endif
  759. endif
  760. endif
  761. endif
  762. ifeq ($(F_COMPILER), G95)
  763. CCOMMON_OPT += -DF_INTERFACE_G95
  764. FCOMMON_OPT += -Wall
  765. ifneq ($(OSNAME), AIX)
  766. ifndef NO_BINARY_MODE
  767. ifdef BINARY64
  768. FCOMMON_OPT += -m64
  769. else
  770. FCOMMON_OPT += -m32
  771. endif
  772. endif
  773. ifneq ($(NO_LAPACKE), 1)
  774. FCOMMON_OPT += -fno-second-underscore
  775. endif
  776. endif
  777. endif
  778. ifeq ($(F_COMPILER), GFORTRAN)
  779. CCOMMON_OPT += -DF_INTERFACE_GFORT
  780. FCOMMON_OPT += -Wall
  781. # make single-threaded LAPACK calls thread-safe #1847
  782. FCOMMON_OPT += -frecursive
  783. # work around ABI problem with passing single-character arguments
  784. FCOMMON_OPT += -fno-optimize-sibling-calls
  785. #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
  786. ifneq ($(NO_LAPACK), 1)
  787. EXTRALIB += -lgfortran
  788. endif
  789. ifdef NO_BINARY_MODE
  790. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  791. ifdef BINARY64
  792. FCOMMON_OPT += -mabi=64
  793. else
  794. FCOMMON_OPT += -mabi=n32
  795. endif
  796. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  797. FCOMMON_OPT += -mabi=32
  798. endif
  799. else
  800. ifdef BINARY64
  801. ifneq ($(OSNAME), AIX)
  802. FCOMMON_OPT += -m64
  803. endif
  804. ifdef INTERFACE64
  805. ifneq ($(INTERFACE64), 0)
  806. FCOMMON_OPT += -fdefault-integer-8
  807. endif
  808. endif
  809. else
  810. ifneq ($(OSNAME), AIX)
  811. FCOMMON_OPT += -m32
  812. endif
  813. endif
  814. endif
  815. ifeq ($(USE_OPENMP), 1)
  816. FCOMMON_OPT += -fopenmp
  817. endif
  818. endif
  819. ifeq ($(F_COMPILER), INTEL)
  820. CCOMMON_OPT += -DF_INTERFACE_INTEL
  821. ifdef INTERFACE64
  822. ifneq ($(INTERFACE64), 0)
  823. FCOMMON_OPT += -i8
  824. endif
  825. endif
  826. FCOMMON_OPT += -recursive -fp-model strict -assume protect-parens
  827. ifeq ($(USE_OPENMP), 1)
  828. FCOMMON_OPT += -fopenmp
  829. endif
  830. endif
  831. ifeq ($(F_COMPILER), FUJITSU)
  832. CCOMMON_OPT += -DF_INTERFACE_FUJITSU
  833. ifeq ($(USE_OPENMP), 1)
  834. FCOMMON_OPT += -openmp
  835. endif
  836. endif
  837. ifeq ($(F_COMPILER), IBM)
  838. CCOMMON_OPT += -DF_INTERFACE_IBM
  839. # FCOMMON_OPT += -qarch=440
  840. ifdef BINARY64
  841. FCOMMON_OPT += -q64
  842. ifdef INTERFACE64
  843. ifneq ($(INTERFACE64), 0)
  844. FCOMMON_OPT += -qintsize=8
  845. endif
  846. endif
  847. else
  848. FCOMMON_OPT += -q32
  849. endif
  850. ifeq ($(USE_OPENMP), 1)
  851. FCOMMON_OPT += -openmp
  852. endif
  853. endif
  854. ifeq ($(F_COMPILER), PGI)
  855. CCOMMON_OPT += -DF_INTERFACE_PGI
  856. COMMON_PROF += -DPGICOMPILER
  857. ifdef BINARY64
  858. ifdef INTERFACE64
  859. ifneq ($(INTERFACE64), 0)
  860. FCOMMON_OPT += -i8
  861. endif
  862. endif
  863. ifeq ($(ARCH), x86_64)
  864. FCOMMON_OPT += -tp p7-64
  865. else
  866. ifeq ($(ARCH), power)
  867. ifeq ($(CORE), POWER8)
  868. FCOMMON_OPT += -tp pwr8
  869. endif
  870. ifeq ($(CORE), POWER9)
  871. FCOMMON_OPT += -tp pwr9
  872. endif
  873. endif
  874. endif
  875. else
  876. FCOMMON_OPT += -tp p7
  877. endif
  878. FCOMMON_OPT += -Mrecursive
  879. ifeq ($(USE_OPENMP), 1)
  880. FCOMMON_OPT += -mp
  881. endif
  882. endif
  883. ifeq ($(F_COMPILER), PATHSCALE)
  884. CCOMMON_OPT += -DF_INTERFACE_PATHSCALE
  885. ifdef BINARY64
  886. ifdef INTERFACE64
  887. ifneq ($(INTERFACE64), 0)
  888. FCOMMON_OPT += -i8
  889. endif
  890. endif
  891. endif
  892. ifeq ($(USE_OPENMP), 1)
  893. FCOMMON_OPT += -mp
  894. endif
  895. endif
  896. ifeq ($(F_COMPILER), OPEN64)
  897. CCOMMON_OPT += -DF_INTERFACE_OPEN64
  898. ifdef BINARY64
  899. ifdef INTERFACE64
  900. ifneq ($(INTERFACE64), 0)
  901. FCOMMON_OPT += -i8
  902. endif
  903. endif
  904. endif
  905. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  906. ifndef BINARY64
  907. FCOMMON_OPT += -n32
  908. else
  909. FCOMMON_OPT += -n64
  910. endif
  911. ifeq ($(CORE), LOONGSON3A)
  912. FCOMMON_OPT += -loongson3 -static
  913. endif
  914. ifeq ($(CORE), LOONGSON3B)
  915. FCOMMON_OPT += -loongson3 -static
  916. endif
  917. else
  918. ifndef BINARY64
  919. FCOMMON_OPT += -m32
  920. else
  921. FCOMMON_OPT += -m64
  922. endif
  923. endif
  924. ifeq ($(USE_OPENMP), 1)
  925. FEXTRALIB += -lstdc++
  926. FCOMMON_OPT += -mp
  927. endif
  928. endif
  929. ifeq ($(C_COMPILER), OPEN64)
  930. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  931. ifndef BINARY64
  932. CCOMMON_OPT += -n32
  933. else
  934. CCOMMON_OPT += -n64
  935. endif
  936. ifeq ($(CORE), LOONGSON3A)
  937. CCOMMON_OPT += -loongson3 -static
  938. endif
  939. ifeq ($(CORE), LOONGSON3B)
  940. CCOMMON_OPT += -loongson3 -static
  941. endif
  942. else
  943. ifndef BINARY64
  944. CCOMMON_OPT += -m32
  945. else
  946. CCOMMON_OPT += -m64
  947. endif
  948. endif
  949. endif
  950. ifeq ($(C_COMPILER), SUN)
  951. CCOMMON_OPT += -w
  952. ifeq ($(ARCH), x86)
  953. CCOMMON_OPT += -m32
  954. else
  955. ifdef BINARY64
  956. CCOMMON_OPT += -m64
  957. else
  958. CCOMMON_OPT += -m32
  959. endif
  960. endif
  961. endif
  962. ifeq ($(F_COMPILER), SUN)
  963. CCOMMON_OPT += -DF_INTERFACE_SUN
  964. FCOMMON_OPT += -ftrap=%none -xrecursive
  965. ifeq ($(ARCH), x86)
  966. FCOMMON_OPT += -m32
  967. else
  968. ifdef BINARY64
  969. FCOMMON_OPT += -m64
  970. else
  971. FCOMMON_OPT += -m32
  972. endif
  973. endif
  974. ifeq ($(USE_OPENMP), 1)
  975. FCOMMON_OPT += -xopenmp=parallel
  976. endif
  977. endif
  978. ifeq ($(F_COMPILER), COMPAQ)
  979. CCOMMON_OPT += -DF_INTERFACE_COMPAQ
  980. ifeq ($(USE_OPENMP), 1)
  981. FCOMMON_OPT += -openmp
  982. endif
  983. endif
  984. ifdef BINARY64
  985. ifdef INTERFACE64
  986. ifneq ($(INTERFACE64), 0)
  987. CCOMMON_OPT +=
  988. #-DUSE64BITINT
  989. endif
  990. endif
  991. endif
  992. ifeq ($(NEED_PIC), 1)
  993. ifeq ($(C_COMPILER), IBM)
  994. CCOMMON_OPT += -qpic=large
  995. else
  996. CCOMMON_OPT += -fPIC
  997. endif
  998. ifeq ($(F_COMPILER), SUN)
  999. FCOMMON_OPT += -pic
  1000. else
  1001. FCOMMON_OPT += -fPIC
  1002. endif
  1003. endif
  1004. ifeq ($(DYNAMIC_ARCH), 1)
  1005. CCOMMON_OPT += -DDYNAMIC_ARCH
  1006. endif
  1007. ifeq ($(DYNAMIC_OLDER), 1)
  1008. CCOMMON_OPT += -DDYNAMIC_OLDER
  1009. endif
  1010. ifeq ($(NO_LAPACK), 1)
  1011. CCOMMON_OPT += -DNO_LAPACK
  1012. #Disable LAPACK C interface
  1013. NO_LAPACKE = 1
  1014. endif
  1015. ifeq ($(NO_LAPACKE), 1)
  1016. CCOMMON_OPT += -DNO_LAPACKE
  1017. endif
  1018. ifeq ($(NO_AVX), 1)
  1019. CCOMMON_OPT += -DNO_AVX
  1020. endif
  1021. ifeq ($(ARCH), x86)
  1022. CCOMMON_OPT += -DNO_AVX
  1023. endif
  1024. ifeq ($(NO_AVX2), 1)
  1025. CCOMMON_OPT += -DNO_AVX2
  1026. endif
  1027. ifeq ($(NO_AVX512), 1)
  1028. CCOMMON_OPT += -DNO_AVX512
  1029. endif
  1030. ifdef SMP
  1031. CCOMMON_OPT += -DSMP_SERVER
  1032. ifeq ($(ARCH), mips64)
  1033. ifneq ($(CORE), LOONGSON3B)
  1034. USE_SIMPLE_THREADED_LEVEL3 = 1
  1035. endif
  1036. endif
  1037. ifeq ($(USE_OPENMP), 1)
  1038. # USE_SIMPLE_THREADED_LEVEL3 = 1
  1039. # NO_AFFINITY = 1
  1040. CCOMMON_OPT += -DUSE_OPENMP
  1041. endif
  1042. ifeq ($(BIGNUMA), 1)
  1043. CCOMMON_OPT += -DBIGNUMA
  1044. endif
  1045. endif
  1046. ifeq ($(NO_WARMUP), 1)
  1047. CCOMMON_OPT += -DNO_WARMUP
  1048. endif
  1049. ifeq ($(CONSISTENT_FPCSR), 1)
  1050. CCOMMON_OPT += -DCONSISTENT_FPCSR
  1051. endif
  1052. # Only for development
  1053. # CCOMMON_OPT += -DPARAMTEST
  1054. # CCOMMON_OPT += -DPREFETCHTEST
  1055. # CCOMMON_OPT += -DNO_SWITCHING
  1056. # USE_PAPI = 1
  1057. ifdef USE_PAPI
  1058. CCOMMON_OPT += -DUSE_PAPI
  1059. EXTRALIB += -lpapi -lperfctr
  1060. endif
  1061. ifdef DYNAMIC_THREADS
  1062. CCOMMON_OPT += -DDYNAMIC_THREADS
  1063. endif
  1064. CCOMMON_OPT += -DMAX_CPU_NUMBER=$(NUM_THREADS)
  1065. CCOMMON_OPT += -DMAX_PARALLEL_NUMBER=$(NUM_PARALLEL)
  1066. ifdef USE_SIMPLE_THREADED_LEVEL3
  1067. CCOMMON_OPT += -DUSE_SIMPLE_THREADED_LEVEL3
  1068. endif
  1069. ifeq ($(USE_TLS), 1)
  1070. CCOMMON_OPT += -DUSE_TLS
  1071. endif
  1072. ifeq ($(BUILD_BFLOAT16), 1)
  1073. CCOMMON_OPT += -DBUILD_BFLOAT16
  1074. endif
  1075. ifeq ($(BUILD_SINGLE), 1)
  1076. CCOMMON_OPT += -DBUILD_SINGLE=1
  1077. endif
  1078. ifeq ($(BUILD_DOUBLE), 1)
  1079. CCOMMON_OPT += -DBUILD_DOUBLE=1
  1080. endif
  1081. ifeq ($(BUILD_COMPLEX), 1)
  1082. CCOMMON_OPT += -DBUILD_COMPLEX=1
  1083. endif
  1084. ifeq ($(BUILD_COMPLEX16), 1)
  1085. CCOMMON_OPT += -DBUILD_COMPLEX16=1
  1086. endif
  1087. CCOMMON_OPT += -DVERSION=\"$(VERSION)\"
  1088. ifndef SYMBOLPREFIX
  1089. SYMBOLPREFIX =
  1090. endif
  1091. ifndef SYMBOLSUFFIX
  1092. SYMBOLSUFFIX =
  1093. endif
  1094. ifndef LIBNAMESUFFIX
  1095. LIBNAMEBASE = $(SYMBOLPREFIX)openblas$(SYMBOLSUFFIX)
  1096. else
  1097. LIBNAMEBASE = $(SYMBOLPREFIX)openblas$(SYMBOLSUFFIX)_$(LIBNAMESUFFIX)
  1098. endif
  1099. ifeq ($(OSNAME), CYGWIN_NT)
  1100. LIBPREFIX = cyg$(LIBNAMEBASE)
  1101. else
  1102. LIBPREFIX = lib$(LIBNAMEBASE)
  1103. endif
  1104. KERNELDIR = $(TOPDIR)/kernel/$(ARCH)
  1105. include $(TOPDIR)/Makefile.$(ARCH)
  1106. ifneq ($(C_COMPILER), PGI)
  1107. ifneq ($(C_COMPILER), SUN)
  1108. CCOMMON_OPT += -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME
  1109. endif
  1110. endif
  1111. CCOMMON_OPT += -DASMNAME=$(FU)$(*F) -DASMFNAME=$(FU)$(*F)$(BU) -DNAME=$(*F)$(BU) -DCNAME=$(*F) -DCHAR_NAME=\"$(*F)$(BU)\" -DCHAR_CNAME=\"$(*F)\"
  1112. ifeq ($(CORE), PPC440)
  1113. CCOMMON_OPT += -DALLOC_QALLOC
  1114. endif
  1115. ifeq ($(CORE), PPC440FP2)
  1116. STATIC_ALLOCATION = 1
  1117. endif
  1118. ifneq ($(OSNAME), Linux)
  1119. NO_AFFINITY = 1
  1120. endif
  1121. ifneq ($(ARCH), x86_64)
  1122. ifneq ($(ARCH), x86)
  1123. ifneq ($(CORE), LOONGSON3B)
  1124. NO_AFFINITY = 1
  1125. endif
  1126. endif
  1127. endif
  1128. ifdef NO_AFFINITY
  1129. ifeq ($(NO_AFFINITY), 0)
  1130. override undefine NO_AFFINITY
  1131. else
  1132. CCOMMON_OPT += -DNO_AFFINITY
  1133. endif
  1134. endif
  1135. ifdef FUNCTION_PROFILE
  1136. CCOMMON_OPT += -DFUNCTION_PROFILE
  1137. endif
  1138. ifdef HUGETLB_ALLOCATION
  1139. CCOMMON_OPT += -DALLOC_HUGETLB
  1140. endif
  1141. ifdef HUGETLBFILE_ALLOCATION
  1142. CCOMMON_OPT += -DALLOC_HUGETLBFILE -DHUGETLB_FILE_NAME=$(HUGETLBFILE_ALLOCATION)
  1143. endif
  1144. ifdef STATIC_ALLOCATION
  1145. CCOMMON_OPT += -DALLOC_STATIC
  1146. endif
  1147. ifdef DEVICEDRIVER_ALLOCATION
  1148. CCOMMON_OPT += -DALLOC_DEVICEDRIVER -DDEVICEDRIVER_NAME=\"/dev/mapper\"
  1149. endif
  1150. ifdef MIXED_MEMORY_ALLOCATION
  1151. CCOMMON_OPT += -DMIXED_MEMORY_ALLOCATION
  1152. endif
  1153. ifeq ($(OSNAME), SunOS)
  1154. TAR = gtar
  1155. PATCH = gpatch
  1156. GREP = ggrep
  1157. AWK = nawk
  1158. else
  1159. TAR = tar
  1160. PATCH = patch
  1161. GREP = grep
  1162. AWK = awk
  1163. endif
  1164. ifndef MD5SUM
  1165. MD5SUM = md5sum
  1166. endif
  1167. REVISION = -r$(VERSION)
  1168. MAJOR_VERSION = $(word 1,$(subst ., ,$(VERSION)))
  1169. ifeq ($(DEBUG), 1)
  1170. COMMON_OPT += -g
  1171. endif
  1172. ifeq ($(DEBUG), 1)
  1173. FCOMMON_OPT += -g
  1174. endif
  1175. ifndef COMMON_OPT
  1176. COMMON_OPT = -O2
  1177. endif
  1178. ifndef FCOMMON_OPT
  1179. FCOMMON_OPT = -O2 -frecursive
  1180. endif
  1181. override CFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
  1182. override PFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
  1183. override FFLAGS += $(COMMON_OPT) $(FCOMMON_OPT)
  1184. override FPFLAGS += $(FCOMMON_OPT) $(COMMON_PROF)
  1185. #MAKEOVERRIDES =
  1186. ifdef NEED_PIC
  1187. ifeq (,$(findstring PIC,$(FFLAGS)))
  1188. override FFLAGS += -fPIC
  1189. endif
  1190. endif
  1191. #For LAPACK Fortran codes.
  1192. #Disable -fopenmp for LAPACK Fortran codes on Windows.
  1193. ifdef OS_WINDOWS
  1194. LAPACK_FFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FFLAGS))
  1195. LAPACK_FPFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FPFLAGS))
  1196. else
  1197. LAPACK_FFLAGS := $(FFLAGS)
  1198. LAPACK_FPFLAGS := $(FPFLAGS)
  1199. endif
  1200. LAPACK_CFLAGS = $(CFLAGS)
  1201. LAPACK_CFLAGS += -DHAVE_LAPACK_CONFIG_H
  1202. ifdef INTERFACE64
  1203. ifneq ($(INTERFACE64), 0)
  1204. LAPACK_CFLAGS += -DLAPACK_ILP64
  1205. endif
  1206. endif
  1207. ifdef OS_WINDOWS
  1208. LAPACK_CFLAGS += -DOPENBLAS_OS_WINDOWS
  1209. endif
  1210. ifeq ($(C_COMPILER), LSB)
  1211. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1212. endif
  1213. ifndef SUFFIX
  1214. SUFFIX = o
  1215. endif
  1216. ifndef PSUFFIX
  1217. PSUFFIX = po
  1218. endif
  1219. ifndef LIBSUFFIX
  1220. LIBSUFFIX = a
  1221. endif
  1222. ifneq ($(DYNAMIC_ARCH), 1)
  1223. ifndef SMP
  1224. LIBNAME = $(LIBPREFIX)_$(LIBCORE)$(REVISION).$(LIBSUFFIX)
  1225. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)$(REVISION)_p.$(LIBSUFFIX)
  1226. else
  1227. LIBNAME = $(LIBPREFIX)_$(LIBCORE)p$(REVISION).$(LIBSUFFIX)
  1228. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)p$(REVISION)_p.$(LIBSUFFIX)
  1229. endif
  1230. else
  1231. ifndef SMP
  1232. LIBNAME = $(LIBPREFIX)$(REVISION).$(LIBSUFFIX)
  1233. LIBNAME_P = $(LIBPREFIX)$(REVISION)_p.$(LIBSUFFIX)
  1234. else
  1235. LIBNAME = $(LIBPREFIX)p$(REVISION).$(LIBSUFFIX)
  1236. LIBNAME_P = $(LIBPREFIX)p$(REVISION)_p.$(LIBSUFFIX)
  1237. endif
  1238. endif
  1239. LIBDLLNAME = $(LIBPREFIX).dll
  1240. IMPLIBNAME = lib$(LIBNAMEBASE).dll.a
  1241. ifneq ($(OSNAME), AIX)
  1242. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.so)
  1243. else
  1244. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.a)
  1245. endif
  1246. LIBDYNNAME = $(LIBNAME:.$(LIBSUFFIX)=.dylib)
  1247. LIBDEFNAME = $(LIBNAME:.$(LIBSUFFIX)=.def)
  1248. LIBEXPNAME = $(LIBNAME:.$(LIBSUFFIX)=.exp)
  1249. LIBZIPNAME = $(LIBNAME:.$(LIBSUFFIX)=.zip)
  1250. LIBS = $(TOPDIR)/$(LIBNAME)
  1251. LIBS_P = $(TOPDIR)/$(LIBNAME_P)
  1252. LIB_COMPONENTS = BLAS
  1253. ifneq ($(NO_CBLAS), 1)
  1254. LIB_COMPONENTS += CBLAS
  1255. endif
  1256. ifneq ($(NO_LAPACK), 1)
  1257. LIB_COMPONENTS += LAPACK
  1258. ifneq ($(NO_LAPACKE), 1)
  1259. LIB_COMPONENTS += LAPACKE
  1260. endif
  1261. ifeq ($(BUILD_RELAPACK), 1)
  1262. LIB_COMPONENTS += ReLAPACK
  1263. endif
  1264. endif
  1265. ifeq ($(ONLY_CBLAS), 1)
  1266. LIB_COMPONENTS = CBLAS
  1267. endif
  1268. export OSNAME
  1269. export ARCH
  1270. export CORE
  1271. export LIBCORE
  1272. export __BYTE_ORDER__
  1273. export ELF_VERSION
  1274. export PGCPATH
  1275. export CONFIG
  1276. export CC
  1277. export FC
  1278. export BU
  1279. export FU
  1280. export NEED2UNDERSCORES
  1281. export USE_THREAD
  1282. export NUM_THREADS
  1283. export NUM_CORES
  1284. export SMP
  1285. export MAKEFILE_RULE
  1286. export NEED_PIC
  1287. export BINARY
  1288. export BINARY32
  1289. export BINARY64
  1290. export F_COMPILER
  1291. export C_COMPILER
  1292. export USE_OPENMP
  1293. export CROSS
  1294. export CROSS_SUFFIX
  1295. export NOFORTRAN
  1296. export NO_FBLAS
  1297. export EXTRALIB
  1298. export CEXTRALIB
  1299. export FEXTRALIB
  1300. export HAVE_SSE
  1301. export HAVE_SSE2
  1302. export HAVE_SSE3
  1303. export HAVE_SSSE3
  1304. export HAVE_SSE4_1
  1305. export HAVE_SSE4_2
  1306. export HAVE_SSE4A
  1307. export HAVE_SSE5
  1308. export HAVE_AVX
  1309. export HAVE_VFP
  1310. export HAVE_VFPV3
  1311. export HAVE_VFPV4
  1312. export HAVE_NEON
  1313. export HAVE_MSA
  1314. export MSA_FLAGS
  1315. export KERNELDIR
  1316. export FUNCTION_PROFILE
  1317. export TARGET_CORE
  1318. export NO_AVX512
  1319. export BUILD_BFLOAT16
  1320. export SBGEMM_UNROLL_M
  1321. export SBGEMM_UNROLL_N
  1322. export SGEMM_UNROLL_M
  1323. export SGEMM_UNROLL_N
  1324. export DGEMM_UNROLL_M
  1325. export DGEMM_UNROLL_N
  1326. export QGEMM_UNROLL_M
  1327. export QGEMM_UNROLL_N
  1328. export CGEMM_UNROLL_M
  1329. export CGEMM_UNROLL_N
  1330. export ZGEMM_UNROLL_M
  1331. export ZGEMM_UNROLL_N
  1332. export XGEMM_UNROLL_M
  1333. export XGEMM_UNROLL_N
  1334. export CGEMM3M_UNROLL_M
  1335. export CGEMM3M_UNROLL_N
  1336. export ZGEMM3M_UNROLL_M
  1337. export ZGEMM3M_UNROLL_N
  1338. export XGEMM3M_UNROLL_M
  1339. export XGEMM3M_UNROLL_N
  1340. ifdef USE_CUDA
  1341. export CUDADIR
  1342. export CUCC
  1343. export CUFLAGS
  1344. export CULIB
  1345. endif
  1346. .SUFFIXES: .$(PSUFFIX) .$(SUFFIX) .f
  1347. .f.$(SUFFIX):
  1348. $(FC) $(FFLAGS) -c $< -o $(@F)
  1349. .f.$(PSUFFIX):
  1350. $(FC) $(FPFLAGS) -pg -c $< -o $(@F)
  1351. ifdef BINARY64
  1352. PATHSCALEPATH = /opt/pathscale/lib/3.1
  1353. PGIPATH = /opt/pgi/linux86-64/7.1-5/lib
  1354. else
  1355. PATHSCALEPATH = /opt/pathscale/lib/3.1/32
  1356. PGIPATH = /opt/pgi/linux86/7.1-5/lib
  1357. endif
  1358. ACMLPATH = /opt/acml/4.3.0
  1359. ifneq ($(OSNAME), Darwin)
  1360. MKLPATH = /opt/intel/mkl/10.2.2.025/lib
  1361. else
  1362. MKLPATH = /Library/Frameworks/Intel_MKL.framework/Versions/10.0.1.014/lib
  1363. endif
  1364. ATLASPATH = /opt/atlas/3.9.17/opteron
  1365. FLAMEPATH = $(HOME)/flame/lib
  1366. ifneq ($(OSNAME), SunOS)
  1367. SUNPATH = /opt/sunstudio12.1
  1368. else
  1369. SUNPATH = /opt/SUNWspro
  1370. endif