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 33 kB

7 years ago
13 years ago
1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285128612871288128912901291129212931294129512961297129812991300130113021303130413051306130713081309131013111312131313141315131613171318131913201321132213231324132513261327132813291330133113321333133413351336133713381339134013411342134313441345134613471348134913501351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384138513861387138813891390139113921393139413951396139713981399140014011402140314041405140614071408140914101411141214131414141514161417141814191420142114221423142414251426142714281429143014311432143314341435143614371438143914401441144214431444144514461447144814491450145114521453145414551456145714581459146014611462146314641465146614671468146914701471147214731474147514761477147814791480148114821483148414851486148714881489149014911492149314941495149614971498149915001501150215031504150515061507150815091510151115121513151415151516151715181519152015211522152315241525152615271528152915301531153215331534153515361537153815391540154115421543154415451546154715481549155015511552155315541555155615571558155915601561156215631564156515661567156815691570157115721573157415751576157715781579158015811582158315841585158615871588158915901591159215931594159515961597159815991600160116021603160416051606160716081609161016111612161316141615161616171618161916201621162216231624162516261627162816291630163116321633163416351636163716381639164016411642164316441645164616471648164916501651165216531654
  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), powerpc64le)
  21. override ARCH=power
  22. else ifeq ($(ARCH), powerpc)
  23. override ARCH=power
  24. else ifeq ($(ARCH), i386)
  25. override ARCH=x86
  26. else ifeq ($(ARCH), armv6)
  27. override ARCH=arm
  28. else ifeq ($(ARCH), armv7)
  29. override ARCH=arm
  30. else ifeq ($(ARCH), aarch64)
  31. override ARCH=arm64
  32. else ifeq ($(ARCH), zarch)
  33. override ARCH=zarch
  34. endif
  35. NETLIB_LAPACK_DIR = $(TOPDIR)/lapack-netlib
  36. # Default C compiler
  37. # - Only set if not specified on the command line or inherited from the environment.
  38. # - CC is an implicit variable so neither '?=' or 'ifndef' can be used.
  39. # http://stackoverflow.com/questions/4029274/mingw-and-make-variables
  40. # - Default value is 'cc' which is not always a valid command (e.g. MinGW).
  41. ifeq ($(origin CC),default)
  42. # Check if $(CC) refers to a valid command and set the value to gcc if not
  43. ifneq ($(findstring cmd.exe,$(SHELL)),)
  44. ifeq ($(shell where $(CC) 2>NUL),)
  45. CC = gcc
  46. endif
  47. else # POSIX-ish
  48. ifeq ($(shell command -v $(CC) 2>/dev/null),)
  49. ifeq ($(shell uname -s),Darwin)
  50. CC = clang
  51. # EXTRALIB += -Wl,-no_compact_unwind
  52. else
  53. CC = gcc
  54. endif # Darwin
  55. endif # CC exists
  56. endif # Shell is sane
  57. endif # CC is set to default
  58. # Default Fortran compiler (FC) is selected by f_check.
  59. ifndef MAKEFILE_RULE
  60. include $(TOPDIR)/Makefile.rule
  61. else
  62. include $(TOPDIR)/$(MAKEFILE_RULE)
  63. endif
  64. #
  65. # Beginning of system configuration
  66. #
  67. ifneq ($(BUILD_SINGLE),1)
  68. ifneq ($(BUILD_DOUBLE),1)
  69. ifneq ($(BUILD_COMPLEX),1)
  70. ifneq ($(BUILD_COMPLEX16),1)
  71. override BUILD_SINGLE=1
  72. override BUILD_DOUBLE=1
  73. override BUILD_COMPLEX=1
  74. override BUILD_COMPLEX16=1
  75. endif
  76. endif
  77. endif
  78. endif
  79. ifndef HOSTCC
  80. HOSTCC = $(CC)
  81. endif
  82. ifdef TARGET
  83. GETARCH_FLAGS := -DFORCE_$(TARGET)
  84. GETARCH_FLAGS += -DUSER_TARGET
  85. ifeq ($(TARGET), GENERIC)
  86. ifeq ($(DYNAMIC_ARCH), 1)
  87. override NO_EXPRECISION=1
  88. export NO_EXPRECiSION
  89. endif
  90. endif
  91. endif
  92. # Force fallbacks for 32bit
  93. ifeq ($(BINARY), 32)
  94. ifeq ($(TARGET), HASWELL)
  95. GETARCH_FLAGS := -DFORCE_NEHALEM
  96. endif
  97. ifeq ($(TARGET), SKYLAKEX)
  98. GETARCH_FLAGS := -DFORCE_NEHALEM
  99. endif
  100. ifeq ($(TARGET), COOPERLAKE)
  101. GETARCH_FLAGS := -DFORCE_NEHALEM
  102. endif
  103. ifeq ($(TARGET), SANDYBRIDGE)
  104. GETARCH_FLAGS := -DFORCE_NEHALEM
  105. endif
  106. ifeq ($(TARGET), BULLDOZER)
  107. GETARCH_FLAGS := -DFORCE_BARCELONA
  108. endif
  109. ifeq ($(TARGET), PILEDRIVER)
  110. GETARCH_FLAGS := -DFORCE_BARCELONA
  111. endif
  112. ifeq ($(TARGET), STEAMROLLER)
  113. GETARCH_FLAGS := -DFORCE_BARCELONA
  114. endif
  115. ifeq ($(TARGET), EXCAVATOR)
  116. GETARCH_FLAGS := -DFORCE_BARCELONA
  117. endif
  118. ifeq ($(TARGET), ZEN)
  119. GETARCH_FLAGS := -DFORCE_BARCELONA
  120. endif
  121. ifeq ($(TARGET), ARMV8)
  122. GETARCH_FLAGS := -DFORCE_ARMV7
  123. endif
  124. ifeq ($(TARGET), POWER8)
  125. GETARCH_FLAGS := -DFORCE_POWER6
  126. endif
  127. endif
  128. #TARGET_CORE will override TARGET which is used in DYNAMIC_ARCH=1.
  129. #
  130. ifdef TARGET_CORE
  131. GETARCH_FLAGS := -DFORCE_$(TARGET_CORE)
  132. endif
  133. # Force fallbacks for 32bit
  134. ifeq ($(BINARY), 32)
  135. ifeq ($(TARGET_CORE), HASWELL)
  136. GETARCH_FLAGS := -DFORCE_NEHALEM
  137. endif
  138. ifeq ($(TARGET_CORE), SKYLAKEX)
  139. GETARCH_FLAGS := -DFORCE_NEHALEM
  140. endif
  141. ifeq ($(TARGET_CORE), COOPERLAKE)
  142. GETARCH_FLAGS := -DFORCE_NEHALEM
  143. endif
  144. ifeq ($(TARGET_CORE), SANDYBRIDGE)
  145. GETARCH_FLAGS := -DFORCE_NEHALEM
  146. endif
  147. ifeq ($(TARGET_CORE), BULLDOZER)
  148. GETARCH_FLAGS := -DFORCE_BARCELONA
  149. endif
  150. ifeq ($(TARGET_CORE), PILEDRIVER)
  151. GETARCH_FLAGS := -DFORCE_BARCELONA
  152. endif
  153. ifeq ($(TARGET_CORE), STEAMROLLER)
  154. GETARCH_FLAGS := -DFORCE_BARCELONA
  155. endif
  156. ifeq ($(TARGET_CORE), EXCAVATOR)
  157. GETARCH_FLAGS := -DFORCE_BARCELONA
  158. endif
  159. ifeq ($(TARGET_CORE), ZEN)
  160. GETARCH_FLAGS := -DFORCE_BARCELONA
  161. endif
  162. endif
  163. # On x86_64 build getarch with march=native unless the compiler is PGI. This is required to detect AVX512 support in getarch.
  164. ifeq ($(HOSTARCH), x86_64)
  165. ifeq ($(findstring pgcc,$(HOSTCC))$(findstring nvc,$(HOSTCC)),)
  166. GETARCH_FLAGS += -march=native
  167. endif
  168. endif
  169. ifdef INTERFACE64
  170. ifneq ($(INTERFACE64), 0)
  171. GETARCH_FLAGS += -DUSE64BITINT
  172. endif
  173. endif
  174. ifndef GEMM_MULTITHREAD_THRESHOLD
  175. GEMM_MULTITHREAD_THRESHOLD=4
  176. endif
  177. GETARCH_FLAGS += -DGEMM_MULTITHREAD_THRESHOLD=$(GEMM_MULTITHREAD_THRESHOLD)
  178. ifeq ($(NO_AVX), 1)
  179. GETARCH_FLAGS += -DNO_AVX
  180. endif
  181. ifeq ($(BINARY), 32)
  182. GETARCH_FLAGS += -DNO_AVX -DNO_AVX2 -DNO_AVX512
  183. NO_AVX512 = 1
  184. endif
  185. ifeq ($(NO_AVX2), 1)
  186. GETARCH_FLAGS += -DNO_AVX2
  187. endif
  188. ifeq ($(NO_AVX512), 1)
  189. GETARCH_FLAGS += -DNO_AVX512
  190. endif
  191. ifeq ($(DEBUG), 1)
  192. GETARCH_FLAGS += -g
  193. endif
  194. ifeq ($(QUIET_MAKE), 1)
  195. MAKE += -s
  196. endif
  197. ifndef NO_PARALLEL_MAKE
  198. NO_PARALLEL_MAKE=0
  199. endif
  200. GETARCH_FLAGS += -DNO_PARALLEL_MAKE=$(NO_PARALLEL_MAKE)
  201. ifdef MAKE_NB_JOBS
  202. GETARCH_FLAGS += -DMAKE_NB_JOBS=$(MAKE_NB_JOBS)
  203. endif
  204. ifeq ($(HOSTCC), loongcc)
  205. GETARCH_FLAGS += -static
  206. endif
  207. #if don't use Fortran, it will only compile CBLAS.
  208. ifeq ($(ONLY_CBLAS), 1)
  209. NO_LAPACK = 1
  210. else
  211. ONLY_CBLAS = 0
  212. endif
  213. # This operation is expensive, so execution should be once.
  214. ifndef GOTOBLAS_MAKEFILE
  215. export GOTOBLAS_MAKEFILE = 1
  216. # Generating Makefile.conf and config.h
  217. 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)
  218. ifndef TARGET_CORE
  219. include $(TOPDIR)/Makefile.conf
  220. else
  221. HAVE_NEON=
  222. HAVE_VFP=
  223. HAVE_VFPV3=
  224. HAVE_VFPV4=
  225. HAVE_MMX=
  226. HAVE_SSE=
  227. HAVE_SSE2=
  228. HAVE_SSE3=
  229. HAVE_SSSE3=
  230. HAVE_SSE4_1=
  231. HAVE_SSE4_2=
  232. HAVE_SSE4A=
  233. HAVE_SSE5=
  234. HAVE_AVX=
  235. HAVE_AVX2=
  236. HAVE_FMA3=
  237. include $(TOPDIR)/Makefile_kernel.conf
  238. endif
  239. endif
  240. ifndef NUM_PARALLEL
  241. NUM_PARALLEL = 1
  242. endif
  243. ifndef NUM_THREADS
  244. NUM_THREADS = $(NUM_CORES)
  245. endif
  246. ifeq ($(NUM_THREADS), 1)
  247. override USE_THREAD = 0
  248. override USE_OPENMP = 0
  249. endif
  250. ifdef USE_THREAD
  251. ifeq ($(USE_THREAD), 0)
  252. SMP =
  253. else
  254. SMP = 1
  255. endif
  256. else
  257. ifeq ($(NUM_THREAD), 1)
  258. SMP =
  259. else
  260. SMP = 1
  261. endif
  262. endif
  263. ifeq ($(SMP), 1)
  264. USE_LOCKING =
  265. endif
  266. ifndef NEED_PIC
  267. NEED_PIC = 1
  268. endif
  269. ARFLAGS =
  270. CPP = $(COMPILER) -E
  271. AR ?= $(CROSS_SUFFIX)ar
  272. AS ?= $(CROSS_SUFFIX)as
  273. LD ?= $(CROSS_SUFFIX)ld
  274. RANLIB ?= $(CROSS_SUFFIX)ranlib
  275. NM = $(CROSS_SUFFIX)nm
  276. DLLWRAP = $(CROSS_SUFFIX)dllwrap
  277. OBJCOPY = $(CROSS_SUFFIX)objcopy
  278. OBJCONV = $(CROSS_SUFFIX)objconv
  279. # When fortran support was either not detected or actively deselected, only build BLAS.
  280. ifeq ($(NOFORTRAN), 1)
  281. NO_LAPACK = 1
  282. override FEXTRALIB =
  283. endif
  284. ifeq ($(C_COMPILER), GCC)
  285. GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
  286. GCCVERSIONGT4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 4)
  287. GCCVERSIONGT5 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 5)
  288. GCCVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 7)
  289. GCCVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  290. GCCVERSIONGTEQ11 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 11)
  291. GCCVERSIONGTEQ10 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 10)
  292. # Note that the behavior of -dumpversion is compile-time-configurable for
  293. # gcc-7.x and newer. Use -dumpfullversion there
  294. ifeq ($(GCCVERSIONGTEQ7),1)
  295. GCCDUMPVERSION_PARAM := -dumpfullversion
  296. else
  297. GCCDUMPVERSION_PARAM := -dumpversion
  298. endif
  299. GCCMINORVERSIONGTEQ1 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 1)
  300. GCCMINORVERSIONGTEQ2 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 2)
  301. GCCMINORVERSIONGTEQ7 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 7)
  302. endif
  303. #
  304. # OS dependent settings
  305. #
  306. ifeq ($(OSNAME), Darwin)
  307. ifndef MACOSX_DEPLOYMENT_TARGET
  308. export MACOSX_DEPLOYMENT_TARGET=10.8
  309. endif
  310. MD5SUM = md5 -r
  311. endif
  312. ifneq (,$(findstring $(OSNAME), FreeBSD OpenBSD DragonFly))
  313. MD5SUM = md5 -r
  314. endif
  315. ifeq ($(OSNAME), NetBSD)
  316. MD5SUM = md5 -n
  317. endif
  318. ifeq ($(OSNAME), Linux)
  319. EXTRALIB += -lm
  320. NO_EXPRECISION = 1
  321. endif
  322. ifeq ($(OSNAME), Android)
  323. EXTRALIB += -lm
  324. endif
  325. ifeq ($(OSNAME), AIX)
  326. EXTRALIB += -lm
  327. endif
  328. ifeq ($(OSNAME), WINNT)
  329. NEED_PIC = 0
  330. NO_EXPRECISION = 1
  331. EXTRALIB += -defaultlib:advapi32
  332. SUFFIX = obj
  333. PSUFFIX = pobj
  334. LIBSUFFIX = a
  335. ifeq ($(C_COMPILER), CLANG)
  336. CCOMMON_OPT += -DMS_ABI
  337. endif
  338. #Version tests for supporting specific features (MS_ABI, POWER9 intrinsics)
  339. ifeq ($(GCCVERSIONGT4), 1)
  340. # GCC Major version > 4
  341. # It is compatible with MSVC ABI.
  342. CCOMMON_OPT += -DMS_ABI
  343. endif
  344. ifeq ($(GCCVERSIONGTEQ4), 1)
  345. ifeq ($(GCCMINORVERSIONGTEQ7), 1)
  346. # GCC Version >=4.7
  347. # It is compatible with MSVC ABI.
  348. CCOMMON_OPT += -DMS_ABI
  349. endif
  350. endif
  351. # Ensure the correct stack alignment on Win32
  352. # http://permalink.gmane.org/gmane.comp.lib.openblas.general/97
  353. ifeq ($(ARCH), x86)
  354. CCOMMON_OPT += -mincoming-stack-boundary=2
  355. FCOMMON_OPT += -mincoming-stack-boundary=2
  356. endif
  357. endif
  358. ifeq ($(OSNAME), Interix)
  359. NEED_PIC = 0
  360. NO_EXPRECISION = 1
  361. INTERIX_TOOL_DIR = /opt/gcc.3.3/i586-pc-interix3/bin
  362. endif
  363. ifeq ($(OSNAME), CYGWIN_NT)
  364. NEED_PIC = 0
  365. NO_EXPRECISION = 1
  366. OS_CYGWIN_NT = 1
  367. endif
  368. ifneq ($(OSNAME), WINNT)
  369. ifneq ($(OSNAME), CYGWIN_NT)
  370. ifneq ($(OSNAME), Interix)
  371. ifneq ($(OSNAME), Android)
  372. ifdef SMP
  373. EXTRALIB += -lpthread
  374. endif
  375. endif
  376. endif
  377. endif
  378. endif
  379. # ifeq logical or
  380. ifeq ($(OSNAME), $(filter $(OSNAME),WINNT CYGWIN_NT Interix))
  381. OS_WINDOWS=1
  382. endif
  383. ifdef QUAD_PRECISION
  384. CCOMMON_OPT += -DQUAD_PRECISION
  385. NO_EXPRECISION = 1
  386. endif
  387. ifneq ($(ARCH), x86)
  388. ifneq ($(ARCH), x86_64)
  389. NO_EXPRECISION = 1
  390. endif
  391. endif
  392. ifdef UTEST_CHECK
  393. CCOMMON_OPT += -DUTEST_CHECK
  394. SANITY_CHECK = 1
  395. endif
  396. ifdef SANITY_CHECK
  397. CCOMMON_OPT += -DSANITY_CHECK -DREFNAME=$(*F)f$(BU)
  398. endif
  399. MAX_STACK_ALLOC ?= 2048
  400. ifneq ($(MAX_STACK_ALLOC), 0)
  401. CCOMMON_OPT += -DMAX_STACK_ALLOC=$(MAX_STACK_ALLOC)
  402. endif
  403. ifdef USE_LOCKING
  404. ifneq ($(USE_LOCKING), 0)
  405. CCOMMON_OPT += -DUSE_LOCKING
  406. endif
  407. endif
  408. #
  409. # Architecture dependent settings
  410. #
  411. ifeq ($(ARCH), x86)
  412. ifndef BINARY
  413. NO_BINARY_MODE = 1
  414. endif
  415. ifeq ($(CORE), generic)
  416. NO_EXPRECISION = 1
  417. endif
  418. ifndef NO_EXPRECISION
  419. ifeq ($(F_COMPILER), GFORTRAN)
  420. # ifeq logical or. GCC or LSB
  421. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  422. EXPRECISION = 1
  423. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  424. FCOMMON_OPT += -m128bit-long-double
  425. endif
  426. ifeq ($(C_COMPILER), CLANG)
  427. EXPRECISION = 1
  428. CCOMMON_OPT += -DEXPRECISION
  429. FCOMMON_OPT += -m128bit-long-double
  430. endif
  431. endif
  432. endif
  433. endif
  434. ifeq ($(ARCH), x86_64)
  435. ifeq ($(CORE), generic)
  436. NO_EXPRECISION = 1
  437. endif
  438. ifndef NO_EXPRECISION
  439. ifeq ($(F_COMPILER), GFORTRAN)
  440. # ifeq logical or. GCC or LSB
  441. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  442. EXPRECISION = 1
  443. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  444. FCOMMON_OPT += -m128bit-long-double
  445. endif
  446. ifeq ($(C_COMPILER), CLANG)
  447. EXPRECISION = 1
  448. CCOMMON_OPT += -DEXPRECISION
  449. FCOMMON_OPT += -m128bit-long-double
  450. endif
  451. endif
  452. endif
  453. endif
  454. ifeq ($(C_COMPILER), INTEL)
  455. CCOMMON_OPT += -wd981
  456. endif
  457. ifeq ($(USE_OPENMP), 1)
  458. #check
  459. ifeq ($(USE_THREAD), 0)
  460. $(error OpenBLAS: Cannot set both USE_OPENMP=1 and USE_THREAD=0. The USE_THREAD=0 is only for building single thread version.)
  461. endif
  462. # ifeq logical or. GCC or LSB
  463. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  464. CCOMMON_OPT += -fopenmp
  465. endif
  466. ifeq ($(C_COMPILER), CLANG)
  467. CCOMMON_OPT += -fopenmp
  468. endif
  469. ifeq ($(C_COMPILER), INTEL)
  470. CCOMMON_OPT += -fopenmp
  471. endif
  472. ifeq ($(C_COMPILER), PGI)
  473. CCOMMON_OPT += -mp
  474. endif
  475. ifeq ($(C_COMPILER), OPEN64)
  476. CCOMMON_OPT += -mp
  477. CEXTRALIB += -lstdc++
  478. endif
  479. ifeq ($(C_COMPILER), PATHSCALE)
  480. CCOMMON_OPT += -mp
  481. endif
  482. endif
  483. ifeq ($(DYNAMIC_ARCH), 1)
  484. ifeq ($(ARCH), x86)
  485. DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
  486. CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
  487. endif
  488. ifeq ($(ARCH), x86_64)
  489. DYNAMIC_CORE = PRESCOTT CORE2
  490. ifeq ($(DYNAMIC_OLDER), 1)
  491. DYNAMIC_CORE += PENRYN DUNNINGTON
  492. endif
  493. DYNAMIC_CORE += NEHALEM
  494. ifeq ($(DYNAMIC_OLDER), 1)
  495. DYNAMIC_CORE += OPTERON OPTERON_SSE3
  496. endif
  497. DYNAMIC_CORE += BARCELONA
  498. ifeq ($(DYNAMIC_OLDER), 1)
  499. DYNAMIC_CORE += BOBCAT ATOM NANO
  500. endif
  501. ifneq ($(NO_AVX), 1)
  502. DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER STEAMROLLER EXCAVATOR
  503. endif
  504. ifneq ($(NO_AVX2), 1)
  505. DYNAMIC_CORE += HASWELL ZEN
  506. endif
  507. ifneq ($(NO_AVX512), 1)
  508. ifneq ($(NO_AVX2), 1)
  509. DYNAMIC_CORE += SKYLAKEX COOPERLAKE
  510. endif
  511. endif
  512. endif
  513. ifdef DYNAMIC_LIST
  514. override DYNAMIC_CORE = PRESCOTT $(DYNAMIC_LIST)
  515. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_PRESCOTT
  516. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  517. CCOMMON_OPT += $(XCCOMMON_OPT)
  518. #CCOMMON_OPT += -DDYNAMIC_LIST='$(DYNAMIC_LIST)'
  519. endif
  520. ifeq ($(ARCH), arm64)
  521. DYNAMIC_CORE = ARMV8
  522. DYNAMIC_CORE += CORTEXA53
  523. DYNAMIC_CORE += CORTEXA57
  524. DYNAMIC_CORE += CORTEXA72
  525. DYNAMIC_CORE += CORTEXA73
  526. DYNAMIC_CORE += NEOVERSEN1
  527. DYNAMIC_CORE += FALKOR
  528. DYNAMIC_CORE += THUNDERX
  529. DYNAMIC_CORE += THUNDERX2T99
  530. DYNAMIC_CORE += TSV110
  531. DYNAMIC_CORE += EMAG8180
  532. DYNAMIC_CORE += THUNDERX3T110
  533. endif
  534. ifeq ($(ARCH), mips64)
  535. DYNAMIC_CORE = LOONGSON3R3 LOONGSON3R4
  536. endif
  537. ifeq ($(ARCH), zarch)
  538. DYNAMIC_CORE = ZARCH_GENERIC
  539. # if the compiler accepts -march=arch11 or -march=z13 and can compile a file
  540. # with z13-specific inline assembly, then we can include support for Z13.
  541. # note: -march=z13 is equivalent to -march=arch11 yet some compiler releases
  542. # only support one or the other.
  543. # note: LLVM version 6.x supported -march=z13 yet could not handle vector
  544. # registers in inline assembly, so the check for supporting the -march flag is
  545. # not enough.
  546. ZARCH_TEST_COMPILE=-c $(TOPDIR)/kernel/zarch/damin_z13.c -I$(TOPDIR) -o /dev/null > /dev/null 2> /dev/null
  547. ZARCH_CC_SUPPORTS_ARCH11=$(shell $(CC) -march=arch11 $(ZARCH_TEST_COMPILE) && echo 1)
  548. ZARCH_CC_SUPPORTS_Z13=$(shell $(CC) -march=z13 $(ZARCH_TEST_COMPILE) && echo 1)
  549. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH11), $(ZARCH_CC_SUPPORTS_Z13)), 1)
  550. DYNAMIC_CORE += Z13
  551. CCOMMON_OPT += -DDYN_Z13
  552. else
  553. $(info OpenBLAS: Not building Z13 kernels because the compiler $(CC) does not support it)
  554. endif
  555. # as above for z13, check for -march=arch12 and z14 support in the compiler.
  556. ZARCH_CC_SUPPORTS_ARCH12=$(shell $(CC) -march=arch12 $(ZARCH_TEST_COMPILE) && echo 1)
  557. ZARCH_CC_SUPPORTS_Z14=$(shell $(CC) -march=z14 $(ZARCH_TEST_COMPILE) && echo 1)
  558. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH12), $(ZARCH_CC_SUPPORTS_Z14)), 1)
  559. DYNAMIC_CORE += Z14
  560. CCOMMON_OPT += -DDYN_Z14
  561. else
  562. $(info OpenBLAS: Not building Z14 kernels because the compiler $(CC) does not support it)
  563. endif
  564. endif # ARCH zarch
  565. ifeq ($(ARCH), power)
  566. ifneq ($(C_COMPILER), PGI)
  567. DYNAMIC_CORE = POWER6
  568. DYNAMIC_CORE += POWER8
  569. ifneq ($(C_COMPILER), GCC)
  570. DYNAMIC_CORE += POWER9
  571. DYNAMIC_CORE += POWER10
  572. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  573. endif
  574. ifeq ($(C_COMPILER), GCC)
  575. ifeq ($(GCCVERSIONGT5), 1)
  576. DYNAMIC_CORE += POWER9
  577. else
  578. $(info, OpenBLAS: Your gcc version is too old to build the POWER9 kernels.)
  579. endif
  580. LDVERSIONGTEQ35 := $(shell expr `$(CC) -Wl,--version 2> /dev/null | head -1 | cut -f2 -d "." | cut -f1 -d "-"` \>= 35)
  581. ifeq ($(GCCVERSIONGTEQ11)$(LDVERSIONGTEQ35), 11)
  582. DYNAMIC_CORE += POWER10
  583. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  584. else ifeq ($(GCCVERSIONGTEQ10), 1)
  585. ifeq ($(GCCMINORVERSIONGTEQ2)$(LDVERSIONGTEQ35), 11)
  586. DYNAMIC_CORE += POWER10
  587. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  588. endif
  589. else
  590. $(info, OpenBLAS: Your gcc version is too old to build the POWER10 kernels.)
  591. endif
  592. endif
  593. else
  594. DYNAMIC_CORE = POWER8
  595. DYNAMIC_CORE += POWER9
  596. endif
  597. endif
  598. # If DYNAMIC_CORE is not set, DYNAMIC_ARCH cannot do anything, so force it to empty
  599. ifndef DYNAMIC_CORE
  600. override DYNAMIC_ARCH=
  601. endif
  602. endif
  603. ifeq ($(ARCH), ia64)
  604. NO_BINARY_MODE = 1
  605. BINARY_DEFINED = 1
  606. ifeq ($(F_COMPILER), GFORTRAN)
  607. ifeq ($(C_COMPILER), GCC)
  608. # EXPRECISION = 1
  609. # CCOMMON_OPT += -DEXPRECISION
  610. endif
  611. endif
  612. endif
  613. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  614. NO_BINARY_MODE = 1
  615. endif
  616. ifeq ($(ARCH), alpha)
  617. NO_BINARY_MODE = 1
  618. BINARY_DEFINED = 1
  619. endif
  620. ifeq ($(ARCH), arm)
  621. NO_BINARY_MODE = 1
  622. BINARY_DEFINED = 1
  623. CCOMMON_OPT += -marm
  624. FCOMMON_OPT += -marm
  625. # If softfp abi is mentioned on the command line, force it.
  626. ifeq ($(ARM_SOFTFP_ABI), 1)
  627. CCOMMON_OPT += -mfloat-abi=softfp
  628. FCOMMON_OPT += -mfloat-abi=softfp
  629. endif
  630. ifeq ($(OSNAME), Android)
  631. ifeq ($(ARM_SOFTFP_ABI), 1)
  632. EXTRALIB += -lm
  633. else
  634. EXTRALIB += -Wl,-lm_hard
  635. endif
  636. endif
  637. endif
  638. ifeq ($(ARCH), arm64)
  639. NO_BINARY_MODE = 1
  640. BINARY_DEFINED = 1
  641. ifdef INTERFACE64
  642. ifneq ($(INTERFACE64), 0)
  643. ifeq ($(F_COMPILER), GFORTRAN)
  644. FCOMMON_OPT += -fdefault-integer-8
  645. endif
  646. ifeq ($(F_COMPILER), FLANG)
  647. FCOMMON_OPT += -i8
  648. endif
  649. endif
  650. endif
  651. endif
  652. ifeq ($(ARCH), riscv64)
  653. NO_BINARY_MODE = 1
  654. BINARY_DEFINED = 1
  655. endif
  656. #
  657. # C Compiler dependent settings
  658. #
  659. # ifeq logical or. GCC or CLANG or LSB
  660. # http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
  661. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG LSB))
  662. CCOMMON_OPT += -Wall
  663. COMMON_PROF += -fno-inline
  664. NO_UNINITIALIZED_WARN = -Wno-uninitialized
  665. ifeq ($(QUIET_MAKE), 1)
  666. CCOMMON_OPT += $(NO_UNINITIALIZED_WARN) -Wno-unused
  667. endif
  668. ifdef NO_BINARY_MODE
  669. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  670. ifdef BINARY64
  671. CCOMMON_OPT += -mabi=64
  672. else
  673. CCOMMON_OPT += -mabi=n32
  674. endif
  675. BINARY_DEFINED = 1
  676. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  677. CCOMMON_OPT += -mabi=32
  678. BINARY_DEFINED = 1
  679. endif
  680. ifeq ($(CORE), $(filter $(CORE),LOONGSON3R3 LOONGSON3R4))
  681. CCOMMON_OPT += -march=loongson3a
  682. FCOMMON_OPT += -march=loongson3a
  683. endif
  684. ifeq ($(CORE), MIPS24K)
  685. CCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  686. FCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  687. endif
  688. ifeq ($(CORE), MIPS1004K)
  689. CCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  690. FCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  691. endif
  692. ifeq ($(CORE), P5600)
  693. CCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  694. FCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  695. endif
  696. ifeq ($(CORE), I6400)
  697. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  698. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  699. endif
  700. ifeq ($(CORE), P6600)
  701. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  702. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  703. endif
  704. ifeq ($(CORE), I6500)
  705. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  706. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  707. endif
  708. ifeq ($(OSNAME), AIX)
  709. BINARY_DEFINED = 1
  710. endif
  711. endif
  712. ifndef BINARY_DEFINED
  713. ifneq ($(OSNAME), AIX)
  714. ifdef BINARY64
  715. ifneq ($(ARCH), riscv64)
  716. CCOMMON_OPT += -m64
  717. endif
  718. else
  719. CCOMMON_OPT += -m32
  720. endif
  721. endif
  722. endif
  723. endif
  724. ifeq ($(C_COMPILER), PGI)
  725. PGCVERSIONGT20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \> 20)
  726. PGCVERSIONGTEQ20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \>= 20)
  727. PGCMINORVERSIONGE11 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -c 4-5` == 11)
  728. PGCVERSIONCHECK := $(PGCVERSIONGT20)$(PGCVERSIONEQ20)$(PGCMINORVERSIONGE11)
  729. ifeq ($(PGCVERSIONCHECK), $(filter $(PGCVERSIONCHECK), 110 111 011))
  730. NEWPGI := 1
  731. endif
  732. ifdef BINARY64
  733. ifeq ($(ARCH), x86_64)
  734. CCOMMON_OPT += -tp p7-64
  735. ifneq ($(NEWPGI),1)
  736. CCOMMON_OPT += -D__MMX__ -Mnollvm
  737. endif
  738. else
  739. ifeq ($(ARCH), power)
  740. ifeq ($(CORE), POWER8)
  741. CCOMMON_OPT += -tp pwr8
  742. endif
  743. ifeq ($(CORE), POWER9)
  744. CCOMMON_OPT += -tp pwr9
  745. endif
  746. endif
  747. endif
  748. else
  749. CCOMMON_OPT += -tp p7
  750. endif
  751. endif
  752. ifeq ($(C_COMPILER), PATHSCALE)
  753. ifdef BINARY64
  754. CCOMMON_OPT += -m64
  755. else
  756. CCOMMON_OPT += -m32
  757. endif
  758. endif
  759. #
  760. # Fortran Compiler dependent settings
  761. #
  762. ifeq ($(F_COMPILER), FLANG)
  763. CCOMMON_OPT += -DF_INTERFACE_FLANG
  764. FCOMMON_OPT += -Mrecursive -Kieee
  765. ifeq ($(OSNAME), Linux)
  766. ifeq ($(ARCH), x86_64)
  767. FLANG_VENDOR := $(shell `$(FC) --version|cut -f 1 -d "."|head -1`)
  768. ifeq ($(FLANG_VENDOR),AOCC)
  769. FCOMMON_OPT += -fno-unroll-loops
  770. endif
  771. endif
  772. endif
  773. ifdef BINARY64
  774. ifdef INTERFACE64
  775. ifneq ($(INTERFACE64), 0)
  776. FCOMMON_OPT += -i8
  777. endif
  778. endif
  779. FCOMMON_OPT += -Wall
  780. else
  781. FCOMMON_OPT += -Wall
  782. endif
  783. ifeq ($(USE_OPENMP), 1)
  784. FCOMMON_OPT += -fopenmp
  785. endif
  786. endif
  787. ifeq ($(F_COMPILER), G77)
  788. CCOMMON_OPT += -DF_INTERFACE_G77
  789. FCOMMON_OPT += -Wall
  790. ifndef NO_BINARY_MODE
  791. ifneq ($(OSNAME), AIX)
  792. ifdef BINARY64
  793. FCOMMON_OPT += -m64
  794. else
  795. FCOMMON_OPT += -m32
  796. endif
  797. endif
  798. endif
  799. endif
  800. ifeq ($(F_COMPILER), G95)
  801. CCOMMON_OPT += -DF_INTERFACE_G95
  802. FCOMMON_OPT += -Wall
  803. ifneq ($(OSNAME), AIX)
  804. ifndef NO_BINARY_MODE
  805. ifdef BINARY64
  806. FCOMMON_OPT += -m64
  807. else
  808. FCOMMON_OPT += -m32
  809. endif
  810. endif
  811. ifneq ($(NO_LAPACKE), 1)
  812. FCOMMON_OPT += -fno-second-underscore
  813. endif
  814. endif
  815. endif
  816. ifeq ($(F_COMPILER), GFORTRAN)
  817. CCOMMON_OPT += -DF_INTERFACE_GFORT
  818. FCOMMON_OPT += -Wall
  819. # make single-threaded LAPACK calls thread-safe #1847
  820. FCOMMON_OPT += -frecursive
  821. # work around ABI problem with passing single-character arguments
  822. FCOMMON_OPT += -fno-optimize-sibling-calls
  823. #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
  824. ifneq ($(NO_LAPACK), 1)
  825. EXTRALIB += -lgfortran
  826. endif
  827. ifdef NO_BINARY_MODE
  828. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  829. ifdef BINARY64
  830. FCOMMON_OPT += -mabi=64
  831. else
  832. FCOMMON_OPT += -mabi=n32
  833. endif
  834. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  835. FCOMMON_OPT += -mabi=32
  836. endif
  837. else
  838. ifdef BINARY64
  839. ifneq ($(OSNAME), AIX)
  840. ifneq ($(ARCH), riscv64)
  841. FCOMMON_OPT += -m64
  842. endif
  843. endif
  844. ifdef INTERFACE64
  845. ifneq ($(INTERFACE64), 0)
  846. FCOMMON_OPT += -fdefault-integer-8
  847. endif
  848. endif
  849. else
  850. ifneq ($(OSNAME), AIX)
  851. FCOMMON_OPT += -m32
  852. endif
  853. endif
  854. endif
  855. ifeq ($(USE_OPENMP), 1)
  856. FCOMMON_OPT += -fopenmp
  857. endif
  858. endif
  859. ifeq ($(F_COMPILER), INTEL)
  860. CCOMMON_OPT += -DF_INTERFACE_INTEL
  861. ifdef INTERFACE64
  862. ifneq ($(INTERFACE64), 0)
  863. FCOMMON_OPT += -i8
  864. endif
  865. endif
  866. FCOMMON_OPT += -recursive -fp-model strict -assume protect-parens
  867. ifeq ($(USE_OPENMP), 1)
  868. FCOMMON_OPT += -fopenmp
  869. endif
  870. endif
  871. ifeq ($(F_COMPILER), FUJITSU)
  872. CCOMMON_OPT += -DF_INTERFACE_FUJITSU
  873. ifeq ($(USE_OPENMP), 1)
  874. FCOMMON_OPT += -openmp
  875. endif
  876. endif
  877. ifeq ($(F_COMPILER), IBM)
  878. CCOMMON_OPT += -DF_INTERFACE_IBM
  879. # FCOMMON_OPT += -qarch=440
  880. ifdef BINARY64
  881. FCOMMON_OPT += -q64
  882. ifdef INTERFACE64
  883. ifneq ($(INTERFACE64), 0)
  884. FCOMMON_OPT += -qintsize=8
  885. endif
  886. endif
  887. else
  888. FCOMMON_OPT += -q32
  889. endif
  890. ifeq ($(USE_OPENMP), 1)
  891. FCOMMON_OPT += -openmp
  892. endif
  893. endif
  894. ifeq ($(F_COMPILER), PGI)
  895. CCOMMON_OPT += -DF_INTERFACE_PGI
  896. COMMON_PROF += -DPGICOMPILER
  897. ifdef BINARY64
  898. ifdef INTERFACE64
  899. ifneq ($(INTERFACE64), 0)
  900. FCOMMON_OPT += -i8
  901. endif
  902. endif
  903. ifeq ($(ARCH), x86_64)
  904. FCOMMON_OPT += -tp p7-64
  905. else
  906. ifeq ($(ARCH), power)
  907. ifeq ($(CORE), POWER6)
  908. $(warning NVIDIA HPC compilers do not support POWER6.)
  909. endif
  910. ifeq ($(CORE), POWER8)
  911. FCOMMON_OPT += -tp pwr8
  912. endif
  913. ifeq ($(CORE), POWER9)
  914. FCOMMON_OPT += -tp pwr9
  915. endif
  916. ifeq ($(CORE), POWER10)
  917. $(warning NVIDIA HPC compilers do not support POWER10.)
  918. endif
  919. endif
  920. endif
  921. else
  922. FCOMMON_OPT += -tp p7
  923. endif
  924. FCOMMON_OPT += -Mrecursive -Kieee
  925. ifeq ($(USE_OPENMP), 1)
  926. FCOMMON_OPT += -mp
  927. endif
  928. endif
  929. ifeq ($(F_COMPILER), PATHSCALE)
  930. CCOMMON_OPT += -DF_INTERFACE_PATHSCALE
  931. ifdef BINARY64
  932. ifdef INTERFACE64
  933. ifneq ($(INTERFACE64), 0)
  934. FCOMMON_OPT += -i8
  935. endif
  936. endif
  937. endif
  938. ifeq ($(USE_OPENMP), 1)
  939. FCOMMON_OPT += -mp
  940. endif
  941. endif
  942. ifeq ($(F_COMPILER), OPEN64)
  943. CCOMMON_OPT += -DF_INTERFACE_OPEN64
  944. ifdef BINARY64
  945. ifdef INTERFACE64
  946. ifneq ($(INTERFACE64), 0)
  947. FCOMMON_OPT += -i8
  948. endif
  949. endif
  950. endif
  951. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  952. ifndef BINARY64
  953. FCOMMON_OPT += -n32
  954. else
  955. FCOMMON_OPT += -n64
  956. endif
  957. ifeq ($(CORE), LOONGSON3R3)
  958. FCOMMON_OPT += -loongson3 -static
  959. endif
  960. ifeq ($(CORE), LOONGSON3R4)
  961. FCOMMON_OPT += -loongson3 -static
  962. endif
  963. else
  964. ifndef BINARY64
  965. FCOMMON_OPT += -m32
  966. else
  967. FCOMMON_OPT += -m64
  968. endif
  969. endif
  970. ifeq ($(USE_OPENMP), 1)
  971. FEXTRALIB += -lstdc++
  972. FCOMMON_OPT += -mp
  973. endif
  974. endif
  975. ifeq ($(C_COMPILER), OPEN64)
  976. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  977. ifndef BINARY64
  978. CCOMMON_OPT += -n32
  979. else
  980. CCOMMON_OPT += -n64
  981. endif
  982. ifeq ($(CORE), LOONGSON3R3)
  983. CCOMMON_OPT += -loongson3 -static
  984. endif
  985. ifeq ($(CORE), LOONGSON3R4)
  986. CCOMMON_OPT += -loongson3 -static
  987. endif
  988. else
  989. ifndef BINARY64
  990. CCOMMON_OPT += -m32
  991. else
  992. CCOMMON_OPT += -m64
  993. endif
  994. endif
  995. endif
  996. ifeq ($(C_COMPILER), SUN)
  997. CCOMMON_OPT += -w
  998. ifeq ($(ARCH), x86)
  999. CCOMMON_OPT += -m32
  1000. else
  1001. ifdef BINARY64
  1002. CCOMMON_OPT += -m64
  1003. else
  1004. CCOMMON_OPT += -m32
  1005. endif
  1006. endif
  1007. endif
  1008. ifeq ($(F_COMPILER), SUN)
  1009. CCOMMON_OPT += -DF_INTERFACE_SUN
  1010. FCOMMON_OPT += -ftrap=%none -xrecursive
  1011. ifeq ($(ARCH), x86)
  1012. FCOMMON_OPT += -m32
  1013. else
  1014. ifdef BINARY64
  1015. FCOMMON_OPT += -m64
  1016. else
  1017. FCOMMON_OPT += -m32
  1018. endif
  1019. endif
  1020. ifeq ($(USE_OPENMP), 1)
  1021. FCOMMON_OPT += -xopenmp=parallel
  1022. endif
  1023. endif
  1024. ifeq ($(F_COMPILER), COMPAQ)
  1025. CCOMMON_OPT += -DF_INTERFACE_COMPAQ
  1026. ifeq ($(USE_OPENMP), 1)
  1027. FCOMMON_OPT += -openmp
  1028. endif
  1029. endif
  1030. ifdef BINARY64
  1031. ifdef INTERFACE64
  1032. ifneq ($(INTERFACE64), 0)
  1033. CCOMMON_OPT +=
  1034. #-DUSE64BITINT
  1035. endif
  1036. endif
  1037. endif
  1038. ifeq ($(NEED_PIC), 1)
  1039. ifeq ($(C_COMPILER), IBM)
  1040. CCOMMON_OPT += -qpic=large
  1041. else
  1042. CCOMMON_OPT += -fPIC
  1043. endif
  1044. ifeq ($(F_COMPILER), SUN)
  1045. FCOMMON_OPT += -pic
  1046. else
  1047. FCOMMON_OPT += -fPIC
  1048. endif
  1049. endif
  1050. ifeq ($(DYNAMIC_ARCH), 1)
  1051. CCOMMON_OPT += -DDYNAMIC_ARCH
  1052. endif
  1053. ifeq ($(DYNAMIC_OLDER), 1)
  1054. CCOMMON_OPT += -DDYNAMIC_OLDER
  1055. endif
  1056. ifeq ($(NO_LAPACK), 1)
  1057. CCOMMON_OPT += -DNO_LAPACK
  1058. #Disable LAPACK C interface
  1059. NO_LAPACKE = 1
  1060. endif
  1061. ifeq ($(NO_LAPACKE), 1)
  1062. CCOMMON_OPT += -DNO_LAPACKE
  1063. endif
  1064. ifeq ($(NO_AVX), 1)
  1065. CCOMMON_OPT += -DNO_AVX
  1066. endif
  1067. ifeq ($(ARCH), x86)
  1068. CCOMMON_OPT += -DNO_AVX
  1069. endif
  1070. ifeq ($(NO_AVX2), 1)
  1071. CCOMMON_OPT += -DNO_AVX2
  1072. endif
  1073. ifeq ($(NO_AVX512), 1)
  1074. CCOMMON_OPT += -DNO_AVX512
  1075. endif
  1076. ifdef SMP
  1077. CCOMMON_OPT += -DSMP_SERVER
  1078. ifeq ($(ARCH), mips64)
  1079. USE_SIMPLE_THREADED_LEVEL3 = 1
  1080. endif
  1081. ifeq ($(USE_OPENMP), 1)
  1082. # USE_SIMPLE_THREADED_LEVEL3 = 1
  1083. # NO_AFFINITY = 1
  1084. CCOMMON_OPT += -DUSE_OPENMP
  1085. endif
  1086. ifeq ($(BIGNUMA), 1)
  1087. CCOMMON_OPT += -DBIGNUMA
  1088. endif
  1089. endif
  1090. ifeq ($(NO_WARMUP), 1)
  1091. CCOMMON_OPT += -DNO_WARMUP
  1092. endif
  1093. ifeq ($(CONSISTENT_FPCSR), 1)
  1094. CCOMMON_OPT += -DCONSISTENT_FPCSR
  1095. endif
  1096. # Only for development
  1097. # CCOMMON_OPT += -DPARAMTEST
  1098. # CCOMMON_OPT += -DPREFETCHTEST
  1099. # CCOMMON_OPT += -DNO_SWITCHING
  1100. # USE_PAPI = 1
  1101. ifdef USE_PAPI
  1102. CCOMMON_OPT += -DUSE_PAPI
  1103. EXTRALIB += -lpapi -lperfctr
  1104. endif
  1105. ifdef DYNAMIC_THREADS
  1106. CCOMMON_OPT += -DDYNAMIC_THREADS
  1107. endif
  1108. CCOMMON_OPT += -DMAX_CPU_NUMBER=$(NUM_THREADS)
  1109. CCOMMON_OPT += -DMAX_PARALLEL_NUMBER=$(NUM_PARALLEL)
  1110. ifdef USE_SIMPLE_THREADED_LEVEL3
  1111. CCOMMON_OPT += -DUSE_SIMPLE_THREADED_LEVEL3
  1112. endif
  1113. ifeq ($(USE_TLS), 1)
  1114. CCOMMON_OPT += -DUSE_TLS
  1115. endif
  1116. ifeq ($(BUILD_BFLOAT16), 1)
  1117. CCOMMON_OPT += -DBUILD_BFLOAT16
  1118. endif
  1119. ifeq ($(BUILD_SINGLE), 1)
  1120. CCOMMON_OPT += -DBUILD_SINGLE=1
  1121. endif
  1122. ifeq ($(BUILD_DOUBLE), 1)
  1123. CCOMMON_OPT += -DBUILD_DOUBLE=1
  1124. endif
  1125. ifeq ($(BUILD_COMPLEX), 1)
  1126. CCOMMON_OPT += -DBUILD_COMPLEX=1
  1127. endif
  1128. ifeq ($(BUILD_COMPLEX16), 1)
  1129. CCOMMON_OPT += -DBUILD_COMPLEX16=1
  1130. endif
  1131. CCOMMON_OPT += -DVERSION=\"$(VERSION)\"
  1132. ifndef SYMBOLPREFIX
  1133. SYMBOLPREFIX =
  1134. endif
  1135. ifndef SYMBOLSUFFIX
  1136. SYMBOLSUFFIX =
  1137. endif
  1138. ifndef LIBSONAMEBASE
  1139. LIBSONAMEBASE = openblas
  1140. endif
  1141. ifndef LIBNAMESUFFIX
  1142. LIBNAMEBASE = $(SYMBOLPREFIX)$(LIBSONAMEBASE)$(SYMBOLSUFFIX)
  1143. else
  1144. LIBNAMEBASE = $(SYMBOLPREFIX)$(LIBSONAMEBASE)$(SYMBOLSUFFIX)_$(LIBNAMESUFFIX)
  1145. endif
  1146. ifeq ($(OSNAME), CYGWIN_NT)
  1147. LIBPREFIX = cyg$(LIBNAMEBASE)
  1148. else
  1149. LIBPREFIX = lib$(LIBNAMEBASE)
  1150. endif
  1151. KERNELDIR = $(TOPDIR)/kernel/$(ARCH)
  1152. include $(TOPDIR)/Makefile.$(ARCH)
  1153. ifneq ($(C_COMPILER), PGI)
  1154. ifneq ($(C_COMPILER), SUN)
  1155. CCOMMON_OPT += -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME
  1156. endif
  1157. endif
  1158. CCOMMON_OPT += -DASMNAME=$(FU)$(*F) -DASMFNAME=$(FU)$(*F)$(BU) -DNAME=$(*F)$(BU) -DCNAME=$(*F) -DCHAR_NAME=\"$(*F)$(BU)\" -DCHAR_CNAME=\"$(*F)\"
  1159. ifeq ($(CORE), PPC440)
  1160. CCOMMON_OPT += -DALLOC_QALLOC
  1161. endif
  1162. ifeq ($(CORE), PPC440FP2)
  1163. STATIC_ALLOCATION = 1
  1164. endif
  1165. ifneq ($(OSNAME), Linux)
  1166. NO_AFFINITY = 1
  1167. endif
  1168. ifneq ($(ARCH), x86_64)
  1169. ifneq ($(ARCH), x86)
  1170. NO_AFFINITY = 1
  1171. endif
  1172. endif
  1173. ifdef NO_AFFINITY
  1174. ifeq ($(NO_AFFINITY), 0)
  1175. override undefine NO_AFFINITY
  1176. else
  1177. CCOMMON_OPT += -DNO_AFFINITY
  1178. endif
  1179. endif
  1180. ifdef FUNCTION_PROFILE
  1181. CCOMMON_OPT += -DFUNCTION_PROFILE
  1182. endif
  1183. ifdef HUGETLB_ALLOCATION
  1184. CCOMMON_OPT += -DALLOC_HUGETLB
  1185. endif
  1186. ifdef HUGETLBFILE_ALLOCATION
  1187. CCOMMON_OPT += -DALLOC_HUGETLBFILE -DHUGETLB_FILE_NAME=$(HUGETLBFILE_ALLOCATION)
  1188. endif
  1189. ifdef STATIC_ALLOCATION
  1190. CCOMMON_OPT += -DALLOC_STATIC
  1191. endif
  1192. ifdef DEVICEDRIVER_ALLOCATION
  1193. CCOMMON_OPT += -DALLOC_DEVICEDRIVER -DDEVICEDRIVER_NAME=\"/dev/mapper\"
  1194. endif
  1195. ifdef MIXED_MEMORY_ALLOCATION
  1196. CCOMMON_OPT += -DMIXED_MEMORY_ALLOCATION
  1197. endif
  1198. ifeq ($(OSNAME), SunOS)
  1199. TAR = gtar
  1200. PATCH = gpatch
  1201. GREP = ggrep
  1202. AWK = nawk
  1203. else
  1204. TAR = tar
  1205. PATCH = patch
  1206. GREP = grep
  1207. AWK = awk
  1208. endif
  1209. ifndef MD5SUM
  1210. MD5SUM = md5sum
  1211. endif
  1212. REVISION = -r$(VERSION)
  1213. MAJOR_VERSION = $(word 1,$(subst ., ,$(VERSION)))
  1214. ifeq ($(DEBUG), 1)
  1215. COMMON_OPT += -g
  1216. endif
  1217. ifeq ($(DEBUG), 1)
  1218. FCOMMON_OPT += -g
  1219. endif
  1220. ifndef COMMON_OPT
  1221. COMMON_OPT = -O2
  1222. endif
  1223. ifndef FCOMMON_OPT
  1224. FCOMMON_OPT = -O2 -frecursive
  1225. endif
  1226. override CFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
  1227. override PFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
  1228. override FFLAGS += $(COMMON_OPT) $(FCOMMON_OPT)
  1229. override FPFLAGS += $(FCOMMON_OPT) $(COMMON_PROF)
  1230. #MAKEOVERRIDES =
  1231. ifdef NEED_PIC
  1232. ifeq (,$(findstring PIC,$(FFLAGS)))
  1233. override FFLAGS += -fPIC
  1234. endif
  1235. endif
  1236. #For LAPACK Fortran codes.
  1237. #Disable -fopenmp for LAPACK Fortran codes on Windows.
  1238. ifdef OS_WINDOWS
  1239. LAPACK_FFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FFLAGS))
  1240. LAPACK_FPFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FPFLAGS))
  1241. else
  1242. LAPACK_FFLAGS := $(FFLAGS)
  1243. LAPACK_FPFLAGS := $(FPFLAGS)
  1244. endif
  1245. LAPACK_CFLAGS = $(CFLAGS)
  1246. LAPACK_CFLAGS += -DHAVE_LAPACK_CONFIG_H
  1247. ifdef INTERFACE64
  1248. ifneq ($(INTERFACE64), 0)
  1249. LAPACK_CFLAGS += -DLAPACK_ILP64
  1250. endif
  1251. endif
  1252. ifdef OS_WINDOWS
  1253. LAPACK_CFLAGS += -DOPENBLAS_OS_WINDOWS
  1254. endif
  1255. ifeq ($(C_COMPILER), LSB)
  1256. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1257. endif
  1258. ifndef SUFFIX
  1259. SUFFIX = o
  1260. endif
  1261. ifndef PSUFFIX
  1262. PSUFFIX = po
  1263. endif
  1264. ifndef LIBSUFFIX
  1265. LIBSUFFIX = a
  1266. endif
  1267. ifneq ($(DYNAMIC_ARCH), 1)
  1268. ifndef SMP
  1269. LIBNAME = $(LIBPREFIX)_$(LIBCORE)$(REVISION).$(LIBSUFFIX)
  1270. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)$(REVISION)_p.$(LIBSUFFIX)
  1271. else
  1272. LIBNAME = $(LIBPREFIX)_$(LIBCORE)p$(REVISION).$(LIBSUFFIX)
  1273. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)p$(REVISION)_p.$(LIBSUFFIX)
  1274. endif
  1275. else
  1276. ifndef SMP
  1277. LIBNAME = $(LIBPREFIX)$(REVISION).$(LIBSUFFIX)
  1278. LIBNAME_P = $(LIBPREFIX)$(REVISION)_p.$(LIBSUFFIX)
  1279. else
  1280. LIBNAME = $(LIBPREFIX)p$(REVISION).$(LIBSUFFIX)
  1281. LIBNAME_P = $(LIBPREFIX)p$(REVISION)_p.$(LIBSUFFIX)
  1282. endif
  1283. endif
  1284. LIBDLLNAME = $(LIBPREFIX).dll
  1285. IMPLIBNAME = lib$(LIBNAMEBASE).dll.a
  1286. ifneq ($(OSNAME), AIX)
  1287. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.so)
  1288. else
  1289. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.a)
  1290. endif
  1291. LIBDYNNAME = $(LIBNAME:.$(LIBSUFFIX)=.dylib)
  1292. LIBDEFNAME = $(LIBNAME:.$(LIBSUFFIX)=.def)
  1293. LIBEXPNAME = $(LIBNAME:.$(LIBSUFFIX)=.exp)
  1294. LIBZIPNAME = $(LIBNAME:.$(LIBSUFFIX)=.zip)
  1295. LIBS = $(TOPDIR)/$(LIBNAME)
  1296. LIBS_P = $(TOPDIR)/$(LIBNAME_P)
  1297. LIB_COMPONENTS = BLAS
  1298. ifneq ($(NO_CBLAS), 1)
  1299. LIB_COMPONENTS += CBLAS
  1300. endif
  1301. ifneq ($(NO_LAPACK), 1)
  1302. LIB_COMPONENTS += LAPACK
  1303. ifneq ($(NO_LAPACKE), 1)
  1304. LIB_COMPONENTS += LAPACKE
  1305. endif
  1306. ifeq ($(BUILD_RELAPACK), 1)
  1307. LIB_COMPONENTS += ReLAPACK
  1308. endif
  1309. endif
  1310. ifeq ($(ONLY_CBLAS), 1)
  1311. LIB_COMPONENTS = CBLAS
  1312. endif
  1313. export OSNAME
  1314. export ARCH
  1315. export CORE
  1316. export LIBCORE
  1317. export __BYTE_ORDER__
  1318. export ELF_VERSION
  1319. export PGCPATH
  1320. export CONFIG
  1321. export CC
  1322. export FC
  1323. export BU
  1324. export FU
  1325. export NEED2UNDERSCORES
  1326. export USE_THREAD
  1327. export NUM_THREADS
  1328. export NUM_CORES
  1329. export SMP
  1330. export MAKEFILE_RULE
  1331. export NEED_PIC
  1332. export BINARY
  1333. export BINARY32
  1334. export BINARY64
  1335. export F_COMPILER
  1336. export C_COMPILER
  1337. export USE_OPENMP
  1338. export CROSS
  1339. export CROSS_SUFFIX
  1340. export NOFORTRAN
  1341. export NO_FBLAS
  1342. export EXTRALIB
  1343. export CEXTRALIB
  1344. export FEXTRALIB
  1345. export HAVE_SSE
  1346. export HAVE_SSE2
  1347. export HAVE_SSE3
  1348. export HAVE_SSSE3
  1349. export HAVE_SSE4_1
  1350. export HAVE_SSE4_2
  1351. export HAVE_SSE4A
  1352. export HAVE_SSE5
  1353. export HAVE_AVX
  1354. export HAVE_AVX2
  1355. export HAVE_FMA3
  1356. export HAVE_VFP
  1357. export HAVE_VFPV3
  1358. export HAVE_VFPV4
  1359. export HAVE_NEON
  1360. export HAVE_MSA
  1361. export MSA_FLAGS
  1362. export KERNELDIR
  1363. export FUNCTION_PROFILE
  1364. export TARGET_CORE
  1365. export NO_AVX512
  1366. export NO_AVX2
  1367. export BUILD_BFLOAT16
  1368. export SBGEMM_UNROLL_M
  1369. export SBGEMM_UNROLL_N
  1370. export SGEMM_UNROLL_M
  1371. export SGEMM_UNROLL_N
  1372. export DGEMM_UNROLL_M
  1373. export DGEMM_UNROLL_N
  1374. export QGEMM_UNROLL_M
  1375. export QGEMM_UNROLL_N
  1376. export CGEMM_UNROLL_M
  1377. export CGEMM_UNROLL_N
  1378. export ZGEMM_UNROLL_M
  1379. export ZGEMM_UNROLL_N
  1380. export XGEMM_UNROLL_M
  1381. export XGEMM_UNROLL_N
  1382. export CGEMM3M_UNROLL_M
  1383. export CGEMM3M_UNROLL_N
  1384. export ZGEMM3M_UNROLL_M
  1385. export ZGEMM3M_UNROLL_N
  1386. export XGEMM3M_UNROLL_M
  1387. export XGEMM3M_UNROLL_N
  1388. ifdef USE_CUDA
  1389. export CUDADIR
  1390. export CUCC
  1391. export CUFLAGS
  1392. export CULIB
  1393. endif
  1394. .SUFFIXES: .$(PSUFFIX) .$(SUFFIX) .f
  1395. .f.$(SUFFIX):
  1396. $(FC) $(FFLAGS) -c $< -o $(@F)
  1397. .f.$(PSUFFIX):
  1398. $(FC) $(FPFLAGS) -pg -c $< -o $(@F)
  1399. ifdef BINARY64
  1400. PATHSCALEPATH = /opt/pathscale/lib/3.1
  1401. PGIPATH = /opt/pgi/linux86-64/7.1-5/lib
  1402. else
  1403. PATHSCALEPATH = /opt/pathscale/lib/3.1/32
  1404. PGIPATH = /opt/pgi/linux86/7.1-5/lib
  1405. endif
  1406. ACMLPATH = /opt/acml/4.3.0
  1407. ifneq ($(OSNAME), Darwin)
  1408. MKLPATH = /opt/intel/mkl/10.2.2.025/lib
  1409. else
  1410. MKLPATH = /Library/Frameworks/Intel_MKL.framework/Versions/10.0.1.014/lib
  1411. endif
  1412. ATLASPATH = /opt/atlas/3.9.17/opteron
  1413. FLAMEPATH = $(HOME)/flame/lib
  1414. ifneq ($(OSNAME), SunOS)
  1415. SUNPATH = /opt/sunstudio12.1
  1416. else
  1417. SUNPATH = /opt/SUNWspro
  1418. endif