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 39 kB

6 years ago
4 years ago
13 years ago
4 years ago
1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285128612871288128912901291129212931294129512961297129812991300130113021303130413051306130713081309131013111312131313141315131613171318131913201321132213231324132513261327132813291330133113321333133413351336133713381339134013411342134313441345134613471348134913501351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384138513861387138813891390139113921393139413951396139713981399140014011402140314041405140614071408140914101411141214131414141514161417141814191420142114221423142414251426142714281429143014311432143314341435143614371438143914401441144214431444144514461447144814491450145114521453145414551456145714581459146014611462146314641465146614671468146914701471147214731474147514761477147814791480148114821483148414851486148714881489149014911492149314941495149614971498149915001501150215031504150515061507150815091510151115121513151415151516151715181519152015211522152315241525152615271528152915301531153215331534153515361537153815391540154115421543154415451546154715481549155015511552155315541555155615571558155915601561156215631564156515661567156815691570157115721573157415751576157715781579158015811582158315841585158615871588158915901591159215931594159515961597159815991600160116021603160416051606160716081609161016111612161316141615161616171618161916201621162216231624162516261627162816291630163116321633163416351636163716381639164016411642164316441645164616471648164916501651165216531654165516561657165816591660166116621663166416651666166716681669167016711672167316741675167616771678167916801681168216831684168516861687168816891690169116921693169416951696169716981699170017011702170317041705170617071708170917101711171217131714171517161717171817191720172117221723172417251726172717281729173017311732173317341735173617371738173917401741174217431744174517461747174817491750175117521753175417551756175717581759176017611762176317641765176617671768176917701771177217731774177517761777177817791780178117821783178417851786178717881789179017911792179317941795179617971798179918001801180218031804180518061807180818091810181118121813181418151816181718181819182018211822182318241825182618271828182918301831183218331834183518361837183818391840184118421843184418451846184718481849185018511852185318541855185618571858185918601861186218631864186518661867186818691870187118721873187418751876187718781879
  1. #
  2. # Include user definition
  3. #
  4. # TO suppress recursive includes
  5. INCLUDED = 1
  6. ifndef TOPDIR
  7. TOPDIR = .
  8. endif
  9. ifndef RELAPACK_REPLACE
  10. RELAPACK_REPLACE=0
  11. endif
  12. # we need to use the host system's architecture for getarch compile options even especially when cross-compiling
  13. HOSTARCH := $(shell uname -m)
  14. ifeq ($(HOSTARCH), amd64)
  15. HOSTARCH=x86_64
  16. endif
  17. # Catch conflicting usage of ARCH in some BSD environments
  18. ifeq ($(ARCH), amd64)
  19. override ARCH=x86_64
  20. else ifeq ($(ARCH), powerpc64)
  21. override ARCH=power
  22. else ifeq ($(ARCH), powerpc64le)
  23. override ARCH=power
  24. else ifeq ($(ARCH), powerpc)
  25. override ARCH=power
  26. else ifeq ($(ARCH), i386)
  27. override ARCH=x86
  28. else ifeq ($(ARCH), armv6)
  29. override ARCH=arm
  30. else ifeq ($(ARCH), armv7)
  31. override ARCH=arm
  32. else ifeq ($(ARCH), aarch64)
  33. override ARCH=arm64
  34. else ifeq ($(ARCH), mipsel)
  35. override ARCH=mips
  36. else ifeq ($(ARCH), mips64el)
  37. override ARCH=mips64
  38. else ifeq ($(ARCH), zarch)
  39. override ARCH=zarch
  40. endif
  41. NETLIB_LAPACK_DIR = $(TOPDIR)/lapack-netlib
  42. # Default C compiler
  43. # - Only set if not specified on the command line or inherited from the environment.
  44. # - CC is an implicit variable so neither '?=' or 'ifndef' can be used.
  45. # http://stackoverflow.com/questions/4029274/mingw-and-make-variables
  46. # - Default value is 'cc' which is not always a valid command (e.g. MinGW).
  47. ifeq ($(origin CC),default)
  48. # Check if $(CC) refers to a valid command and set the value to gcc if not
  49. ifneq ($(findstring cmd.exe,$(SHELL)),)
  50. ifeq ($(shell where $(CC) 2>NUL),)
  51. CC = gcc
  52. endif
  53. else # POSIX-ish
  54. ifeq ($(shell command -v $(CC) 2>/dev/null),)
  55. ifeq ($(shell uname -s),Darwin)
  56. CC = clang
  57. # EXTRALIB += -Wl,-no_compact_unwind
  58. else
  59. CC = gcc
  60. endif # Darwin
  61. endif # CC exists
  62. endif # Shell is sane
  63. endif # CC is set to default
  64. # Default Fortran compiler (FC) is selected by f_check.
  65. ifndef MAKEFILE_RULE
  66. include $(TOPDIR)/Makefile.rule
  67. else
  68. include $(TOPDIR)/$(MAKEFILE_RULE)
  69. endif
  70. #
  71. # Beginning of system configuration
  72. #
  73. ifneq ($(BUILD_SINGLE),1)
  74. ifneq ($(BUILD_DOUBLE),1)
  75. ifneq ($(BUILD_COMPLEX),1)
  76. ifneq ($(BUILD_COMPLEX16),1)
  77. override BUILD_SINGLE=1
  78. override BUILD_DOUBLE=1
  79. override BUILD_COMPLEX=1
  80. override BUILD_COMPLEX16=1
  81. endif
  82. endif
  83. endif
  84. endif
  85. ifndef HOSTCC
  86. HOSTCC = $(CC)
  87. endif
  88. ifdef TARGET
  89. GETARCH_FLAGS := -DFORCE_$(TARGET)
  90. GETARCH_FLAGS += -DUSER_TARGET
  91. ifeq ($(TARGET), GENERIC)
  92. ifeq ($(DYNAMIC_ARCH), 1)
  93. override NO_EXPRECISION=1
  94. export NO_EXPRECISION
  95. endif
  96. endif
  97. endif
  98. # Force fallbacks for 32bit
  99. ifeq ($(BINARY), 32)
  100. ifeq ($(TARGET), HASWELL)
  101. GETARCH_FLAGS := -DFORCE_NEHALEM
  102. endif
  103. ifeq ($(TARGET), SKYLAKEX)
  104. GETARCH_FLAGS := -DFORCE_NEHALEM
  105. endif
  106. ifeq ($(TARGET), COOPERLAKE)
  107. GETARCH_FLAGS := -DFORCE_NEHALEM
  108. endif
  109. ifeq ($(TARGET), SAPPHIRERAPIDS)
  110. GETARCH_FLAGS := -DFORCE_NEHALEM
  111. endif
  112. ifeq ($(TARGET), SANDYBRIDGE)
  113. GETARCH_FLAGS := -DFORCE_NEHALEM
  114. endif
  115. ifeq ($(TARGET), BULLDOZER)
  116. GETARCH_FLAGS := -DFORCE_BARCELONA
  117. endif
  118. ifeq ($(TARGET), PILEDRIVER)
  119. GETARCH_FLAGS := -DFORCE_BARCELONA
  120. endif
  121. ifeq ($(TARGET), STEAMROLLER)
  122. GETARCH_FLAGS := -DFORCE_BARCELONA
  123. endif
  124. ifeq ($(TARGET), EXCAVATOR)
  125. GETARCH_FLAGS := -DFORCE_BARCELONA
  126. endif
  127. ifeq ($(TARGET), ZEN)
  128. GETARCH_FLAGS := -DFORCE_BARCELONA
  129. endif
  130. ifeq ($(TARGET), ARMV8)
  131. GETARCH_FLAGS := -DFORCE_ARMV7
  132. endif
  133. ifeq ($(TARGET), POWER8)
  134. GETARCH_FLAGS := -DFORCE_POWER6
  135. endif
  136. ifeq ($(TARGET), POWER9)
  137. GETARCH_FLAGS := -DFORCE_POWER6
  138. endif
  139. ifeq ($(TARGET), POWER10)
  140. GETARCH_FLAGS := -DFORCE_POWER6
  141. endif
  142. endif
  143. #TARGET_CORE will override TARGET which is used in DYNAMIC_ARCH=1.
  144. #
  145. ifdef TARGET_CORE
  146. GETARCH_FLAGS := -DFORCE_$(TARGET_CORE)
  147. endif
  148. # Force fallbacks for 32bit
  149. ifeq ($(BINARY), 32)
  150. ifeq ($(TARGET_CORE), HASWELL)
  151. GETARCH_FLAGS := -DFORCE_NEHALEM
  152. endif
  153. ifeq ($(TARGET_CORE), SKYLAKEX)
  154. GETARCH_FLAGS := -DFORCE_NEHALEM
  155. endif
  156. ifeq ($(TARGET_CORE), COOPERLAKE)
  157. GETARCH_FLAGS := -DFORCE_NEHALEM
  158. endif
  159. ifeq ($(TARGET_CORE), SAPPHIRERAPIDS)
  160. GETARCH_FLAGS := -DFORCE_NEHALEM
  161. endif
  162. ifeq ($(TARGET_CORE), SANDYBRIDGE)
  163. GETARCH_FLAGS := -DFORCE_NEHALEM
  164. endif
  165. ifeq ($(TARGET_CORE), BULLDOZER)
  166. GETARCH_FLAGS := -DFORCE_BARCELONA
  167. endif
  168. ifeq ($(TARGET_CORE), PILEDRIVER)
  169. GETARCH_FLAGS := -DFORCE_BARCELONA
  170. endif
  171. ifeq ($(TARGET_CORE), STEAMROLLER)
  172. GETARCH_FLAGS := -DFORCE_BARCELONA
  173. endif
  174. ifeq ($(TARGET_CORE), EXCAVATOR)
  175. GETARCH_FLAGS := -DFORCE_BARCELONA
  176. endif
  177. ifeq ($(TARGET_CORE), ZEN)
  178. GETARCH_FLAGS := -DFORCE_BARCELONA
  179. endif
  180. endif
  181. # On x86_64 build getarch with march=native unless the compiler is PGI. This is required to detect AVX512 support in getarch.
  182. ifeq ($(HOSTARCH), x86_64)
  183. ifeq ($(findstring pgcc,$(HOSTCC))$(findstring nvc,$(HOSTCC)),)
  184. GETARCH_FLAGS += -march=native
  185. endif
  186. endif
  187. ifdef INTERFACE64
  188. ifneq ($(INTERFACE64), 0)
  189. GETARCH_FLAGS += -DUSE64BITINT
  190. endif
  191. endif
  192. ifndef GEMM_MULTITHREAD_THRESHOLD
  193. GEMM_MULTITHREAD_THRESHOLD=4
  194. endif
  195. GETARCH_FLAGS += -DGEMM_MULTITHREAD_THRESHOLD=$(GEMM_MULTITHREAD_THRESHOLD)
  196. ifeq ($(NO_AVX), 1)
  197. GETARCH_FLAGS += -DNO_AVX
  198. endif
  199. ifeq ($(BINARY), 32)
  200. GETARCH_FLAGS += -DNO_AVX -DNO_AVX2 -DNO_AVX512
  201. NO_AVX512 = 1
  202. endif
  203. ifeq ($(NO_AVX2), 1)
  204. GETARCH_FLAGS += -DNO_AVX2
  205. endif
  206. ifeq ($(NO_AVX512), 1)
  207. GETARCH_FLAGS += -DNO_AVX512
  208. endif
  209. ifeq ($(DEBUG), 1)
  210. GETARCH_FLAGS += -g
  211. endif
  212. ifeq ($(QUIET_MAKE), 1)
  213. MAKE += -s
  214. endif
  215. ifndef NO_PARALLEL_MAKE
  216. NO_PARALLEL_MAKE=0
  217. endif
  218. GETARCH_FLAGS += -DNO_PARALLEL_MAKE=$(NO_PARALLEL_MAKE)
  219. ifdef MAKE_NB_JOBS
  220. GETARCH_FLAGS += -DMAKE_NB_JOBS=$(MAKE_NB_JOBS)
  221. endif
  222. ifeq ($(HOSTCC), loongcc)
  223. GETARCH_FLAGS += -static
  224. endif
  225. #if don't use Fortran, it will only compile CBLAS.
  226. ifeq ($(ONLY_CBLAS), 1)
  227. NO_LAPACK = 1
  228. else
  229. ONLY_CBLAS = 0
  230. endif
  231. #For small matrix optimization
  232. ifeq ($(ARCH), x86_64)
  233. SMALL_MATRIX_OPT = 1
  234. else ifeq ($(ARCH), power)
  235. SMALL_MATRIX_OPT = 1
  236. BUILD_BFLOAT16 = 1
  237. endif
  238. ifeq ($(SMALL_MATRIX_OPT), 1)
  239. CCOMMON_OPT += -DSMALL_MATRIX_OPT
  240. endif
  241. # This operation is expensive, so execution should be once.
  242. ifndef GOTOBLAS_MAKEFILE
  243. export GOTOBLAS_MAKEFILE = 1
  244. # Generating Makefile.conf and config.h
  245. 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) DYNAMIC_ARCH=$(DYNAMIC_ARCH) TARGET_CORE=$(TARGET_CORE) ONLY_CBLAS=$(ONLY_CBLAS) TARGET=$(TARGET) all)
  246. endif
  247. ifndef TARGET_CORE
  248. -include $(TOPDIR)/Makefile.conf
  249. else
  250. HAVE_NEON=
  251. HAVE_VFP=
  252. HAVE_VFPV3=
  253. HAVE_VFPV4=
  254. HAVE_MMX=
  255. HAVE_SSE=
  256. HAVE_SSE2=
  257. HAVE_SSE3=
  258. HAVE_SSSE3=
  259. HAVE_SSE4_1=
  260. HAVE_SSE4_2=
  261. HAVE_SSE4A=
  262. HAVE_SSE5=
  263. HAVE_AVX=
  264. HAVE_AVX2=
  265. HAVE_FMA3=
  266. include $(TOPDIR)/Makefile_kernel.conf
  267. endif
  268. ifndef NUM_PARALLEL
  269. NUM_PARALLEL = 1
  270. endif
  271. ifndef NUM_THREADS
  272. NUM_THREADS = $(NUM_CORES)
  273. endif
  274. ifeq ($(NUM_THREADS), 1)
  275. override USE_THREAD = 0
  276. override USE_OPENMP = 0
  277. endif
  278. ifdef USE_THREAD
  279. ifeq ($(USE_THREAD), 0)
  280. SMP =
  281. else
  282. SMP = 1
  283. endif
  284. else
  285. ifeq ($(NUM_THREADS), 1)
  286. SMP =
  287. else
  288. SMP = 1
  289. endif
  290. endif
  291. ifeq ($(SMP), 1)
  292. USE_LOCKING =
  293. endif
  294. ifndef NEED_PIC
  295. NEED_PIC = 1
  296. endif
  297. ARFLAGS =
  298. CPP = $(COMPILER) -E
  299. AR ?= $(CROSS_SUFFIX)ar
  300. AS ?= $(CROSS_SUFFIX)as
  301. LD ?= $(CROSS_SUFFIX)ld
  302. RANLIB ?= $(CROSS_SUFFIX)ranlib
  303. NM = $(CROSS_SUFFIX)nm
  304. DLLWRAP = $(CROSS_SUFFIX)dllwrap
  305. OBJCOPY = $(CROSS_SUFFIX)objcopy
  306. OBJCONV = $(CROSS_SUFFIX)objconv
  307. # When fortran support was either not detected or actively deselected, only build BLAS.
  308. ifeq ($(NOFORTRAN), 1)
  309. C_LAPACK = 1
  310. override FEXTRALIB =
  311. endif
  312. ifeq ($(C_COMPILER), GCC)
  313. GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
  314. GCCVERSIONGT4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 4)
  315. GCCVERSIONGT5 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 5)
  316. GCCVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 7)
  317. GCCVERSIONGTEQ8 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 8)
  318. GCCVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  319. GCCVERSIONGTEQ10 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 10)
  320. GCCVERSIONGTEQ11 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 11)
  321. GCCVERSIONGTEQ12 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 12)
  322. # Note that the behavior of -dumpversion is compile-time-configurable for
  323. # gcc-7.x and newer. Use -dumpfullversion there
  324. ifeq ($(GCCVERSIONGTEQ7),1)
  325. GCCDUMPVERSION_PARAM := -dumpfullversion
  326. else
  327. GCCDUMPVERSION_PARAM := -dumpversion
  328. endif
  329. GCCMINORVERSIONGTEQ1 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 1)
  330. GCCMINORVERSIONGTEQ2 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 2)
  331. GCCMINORVERSIONGTEQ4 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 4)
  332. GCCMINORVERSIONGTEQ7 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 7)
  333. endif
  334. ifeq ($(C_COMPILER), CLANG)
  335. CLANGVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  336. CLANGVERSIONGTEQ12 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 12)
  337. endif
  338. #
  339. # OS dependent settings
  340. #
  341. ifeq ($(OSNAME), Darwin)
  342. ifndef MACOSX_DEPLOYMENT_TARGET
  343. ifeq ($(ARCH), arm64)
  344. export MACOSX_DEPLOYMENT_TARGET=11.0
  345. ifeq ($(C_COMPILER), GCC)
  346. export NO_SVE = 1
  347. endif
  348. else
  349. export MACOSX_DEPLOYMENT_TARGET=10.8
  350. endif
  351. endif
  352. MD5SUM = md5 -r
  353. XCVER = $(shell pkgutil --pkg-info=com.apple.pkg.Xcode |awk '/version:/ {print $2}'|cut -d: -f2|cut -f1 -d.)
  354. ifeq (x$(XCVER)x,xx)
  355. XCVER = $(shell pkgutil --pkg-info=com.apple.pkg.CLTools_Executables |awk '/version:/ {print $2}'|cut -d: -f2|cut -f1 -d.)
  356. endif
  357. ifeq (x$(XCVER), x 15)
  358. CCOMMON_OPT += -Wl,-ld_classic
  359. FCOMMON_OPT += -Wl,-ld_classic
  360. endif
  361. endif
  362. ifneq (,$(findstring $(OSNAME), FreeBSD OpenBSD DragonFly))
  363. MD5SUM = md5 -r
  364. endif
  365. ifeq ($(OSNAME), NetBSD)
  366. MD5SUM = md5 -n
  367. endif
  368. ifeq ($(OSNAME), Linux)
  369. EXTRALIB += -lm
  370. NO_EXPRECISION = 1
  371. endif
  372. ifeq ($(OSNAME), Android)
  373. EXTRALIB += -lm
  374. endif
  375. ifeq ($(OSNAME), AIX)
  376. EXTRALIB += -lm
  377. endif
  378. ifeq ($(OSNAME), $(filter $(OSNAME),FreeBSD OpenBSD NetBSD DragonFly))
  379. ifeq ($(ARCH), $(filter $(ARCH),arm arm64))
  380. EXTRALIB += -lm
  381. endif
  382. endif
  383. ifeq ($(OSNAME), WINNT)
  384. NEED_PIC = 0
  385. NO_EXPRECISION = 1
  386. EXTRALIB += -defaultlib:advapi32
  387. SUFFIX = obj
  388. PSUFFIX = pobj
  389. LIBSUFFIX = a
  390. ifeq ($(C_COMPILER), CLANG)
  391. CCOMMON_OPT += -DMS_ABI
  392. endif
  393. #Version tests for supporting specific features (MS_ABI, POWER9 intrinsics)
  394. ifeq ($(GCCVERSIONGT4), 1)
  395. # GCC Major version > 4
  396. # It is compatible with MSVC ABI.
  397. CCOMMON_OPT += -DMS_ABI
  398. endif
  399. ifeq ($(GCCVERSIONGTEQ4), 1)
  400. ifeq ($(GCCMINORVERSIONGTEQ7), 1)
  401. # GCC Version >=4.7
  402. # It is compatible with MSVC ABI.
  403. CCOMMON_OPT += -DMS_ABI
  404. endif
  405. endif
  406. # Ensure the correct stack alignment on Win32
  407. # http://permalink.gmane.org/gmane.comp.lib.openblas.general/97
  408. ifeq ($(ARCH), x86)
  409. CCOMMON_OPT += -mincoming-stack-boundary=2
  410. FCOMMON_OPT += -mincoming-stack-boundary=2
  411. endif
  412. endif
  413. ifeq ($(OSNAME), Interix)
  414. NEED_PIC = 0
  415. NO_EXPRECISION = 1
  416. INTERIX_TOOL_DIR = /opt/gcc.3.3/i586-pc-interix3/bin
  417. endif
  418. ifeq ($(OSNAME), CYGWIN_NT)
  419. NEED_PIC = 0
  420. NO_EXPRECISION = 1
  421. OS_CYGWIN_NT = 1
  422. endif
  423. ifneq ($(OSNAME), WINNT)
  424. ifneq ($(OSNAME), CYGWIN_NT)
  425. ifneq ($(OSNAME), Interix)
  426. ifneq ($(OSNAME), Android)
  427. ifdef SMP
  428. EXTRALIB += -lpthread
  429. endif
  430. endif
  431. endif
  432. endif
  433. endif
  434. # ifeq logical or
  435. ifeq ($(OSNAME), $(filter $(OSNAME),WINNT CYGWIN_NT Interix))
  436. OS_WINDOWS=1
  437. endif
  438. ifdef QUAD_PRECISION
  439. CCOMMON_OPT += -DQUAD_PRECISION
  440. NO_EXPRECISION = 1
  441. endif
  442. ifneq ($(ARCH), x86)
  443. ifneq ($(ARCH), x86_64)
  444. NO_EXPRECISION = 1
  445. endif
  446. endif
  447. ifdef UTEST_CHECK
  448. CCOMMON_OPT += -DUTEST_CHECK
  449. SANITY_CHECK = 1
  450. endif
  451. ifdef SANITY_CHECK
  452. CCOMMON_OPT += -DSANITY_CHECK -DREFNAME=$(*F)f$(BU)
  453. endif
  454. MAX_STACK_ALLOC ?= 2048
  455. ifneq ($(MAX_STACK_ALLOC), 0)
  456. CCOMMON_OPT += -DMAX_STACK_ALLOC=$(MAX_STACK_ALLOC)
  457. endif
  458. ifdef USE_LOCKING
  459. ifneq ($(USE_LOCKING), 0)
  460. CCOMMON_OPT += -DUSE_LOCKING
  461. endif
  462. endif
  463. #
  464. # Architecture dependent settings
  465. #
  466. ifeq ($(ARCH), x86)
  467. ifndef BINARY
  468. NO_BINARY_MODE = 1
  469. endif
  470. ifeq ($(CORE), generic)
  471. NO_EXPRECISION = 1
  472. endif
  473. ifndef NO_EXPRECISION
  474. ifeq ($(F_COMPILER), GFORTRAN)
  475. # ifeq logical or. GCC or LSB
  476. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  477. EXPRECISION = 1
  478. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  479. FCOMMON_OPT += -m128bit-long-double
  480. endif
  481. ifeq ($(C_COMPILER), CLANG)
  482. EXPRECISION = 1
  483. CCOMMON_OPT += -DEXPRECISION
  484. FCOMMON_OPT += -m128bit-long-double
  485. endif
  486. endif
  487. endif
  488. endif
  489. ifeq ($(ARCH), x86_64)
  490. ifeq ($(CORE), generic)
  491. NO_EXPRECISION = 1
  492. endif
  493. ifndef NO_EXPRECISION
  494. ifeq ($(F_COMPILER), GFORTRAN)
  495. # ifeq logical or. GCC or LSB
  496. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  497. EXPRECISION = 1
  498. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  499. FCOMMON_OPT += -m128bit-long-double
  500. endif
  501. ifeq ($(C_COMPILER), CLANG)
  502. EXPRECISION = 1
  503. CCOMMON_OPT += -DEXPRECISION
  504. FCOMMON_OPT += -m128bit-long-double
  505. endif
  506. endif
  507. endif
  508. endif
  509. ifeq ($(C_COMPILER), INTEL)
  510. CCOMMON_OPT += -wd981
  511. endif
  512. ifeq ($(USE_OPENMP), 1)
  513. #check
  514. ifeq ($(USE_THREAD), 0)
  515. $(error OpenBLAS: Cannot set both USE_OPENMP=1 and USE_THREAD=0. The USE_THREAD=0 is only for building single thread version.)
  516. endif
  517. # ifeq logical or. GCC or LSB
  518. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  519. CCOMMON_OPT += -fopenmp
  520. endif
  521. ifeq ($(C_COMPILER), CLANG)
  522. CCOMMON_OPT += -fopenmp
  523. ifeq ($(F_COMPILER), GFORTRAN)
  524. FEXTRALIB := $(subst -lgomp,-lomp,$(FEXTRALIB))
  525. endif
  526. endif
  527. ifeq ($(C_COMPILER), INTEL)
  528. CCOMMON_OPT += -fopenmp
  529. endif
  530. ifeq ($(C_COMPILER), PGI)
  531. CCOMMON_OPT += -mp
  532. endif
  533. ifeq ($(C_COMPILER), OPEN64)
  534. CCOMMON_OPT += -mp
  535. CEXTRALIB += -lstdc++
  536. endif
  537. ifeq ($(C_COMPILER), PATHSCALE)
  538. CCOMMON_OPT += -mp
  539. endif
  540. endif
  541. ifeq ($(DYNAMIC_ARCH), 1)
  542. ifeq ($(ARCH), x86)
  543. DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
  544. CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
  545. endif
  546. ifeq ($(ARCH), x86_64)
  547. DYNAMIC_CORE = PRESCOTT CORE2
  548. ifeq ($(DYNAMIC_OLDER), 1)
  549. DYNAMIC_CORE += PENRYN DUNNINGTON
  550. endif
  551. DYNAMIC_CORE += NEHALEM
  552. ifeq ($(DYNAMIC_OLDER), 1)
  553. DYNAMIC_CORE += OPTERON OPTERON_SSE3
  554. endif
  555. DYNAMIC_CORE += BARCELONA
  556. ifeq ($(DYNAMIC_OLDER), 1)
  557. DYNAMIC_CORE += BOBCAT ATOM NANO
  558. endif
  559. ifneq ($(NO_AVX), 1)
  560. DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER STEAMROLLER EXCAVATOR
  561. endif
  562. ifneq ($(NO_AVX2), 1)
  563. DYNAMIC_CORE += HASWELL ZEN
  564. endif
  565. ifneq ($(NO_AVX512), 1)
  566. ifneq ($(NO_AVX2), 1)
  567. DYNAMIC_CORE += SKYLAKEX COOPERLAKE SAPPHIRERAPIDS
  568. endif
  569. endif
  570. endif
  571. ifdef DYNAMIC_LIST
  572. override DYNAMIC_CORE = PRESCOTT $(DYNAMIC_LIST)
  573. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_PRESCOTT
  574. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  575. CCOMMON_OPT += $(XCCOMMON_OPT)
  576. #CCOMMON_OPT += -DDYNAMIC_LIST='$(DYNAMIC_LIST)'
  577. endif
  578. ifeq ($(ARCH), arm64)
  579. DYNAMIC_CORE = ARMV8
  580. DYNAMIC_CORE += CORTEXA53
  581. DYNAMIC_CORE += CORTEXA57
  582. DYNAMIC_CORE += NEOVERSEN1
  583. ifneq ($(NO_SVE), 1)
  584. DYNAMIC_CORE += NEOVERSEV1
  585. DYNAMIC_CORE += NEOVERSEN2
  586. DYNAMIC_CORE += ARMV8SVE
  587. endif
  588. DYNAMIC_CORE += THUNDERX
  589. DYNAMIC_CORE += THUNDERX2T99
  590. DYNAMIC_CORE += TSV110
  591. DYNAMIC_CORE += EMAG8180
  592. DYNAMIC_CORE += THUNDERX3T110
  593. ifdef DYNAMIC_LIST
  594. override DYNAMIC_CORE = ARMV8 $(DYNAMIC_LIST)
  595. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_ARMV8
  596. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  597. endif
  598. endif
  599. ifeq ($(ARCH), mips64)
  600. DYNAMIC_CORE = LOONGSON3R3 LOONGSON3R4 MIPS64_GENERIC
  601. ifdef DYNAMIC_LIST
  602. override DYNAMIC_CORE = MIPS64_GENERIC $(DYNAMIC_LIST)
  603. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_MIPS64_GENERIC
  604. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  605. endif
  606. endif
  607. ifeq ($(ARCH), loongarch64)
  608. DYNAMIC_CORE = LOONGSON3R5 LOONGSON2K1000 LOONGSONGENERIC
  609. endif
  610. ifeq ($(ARCH), zarch)
  611. DYNAMIC_CORE = ZARCH_GENERIC
  612. # if the compiler accepts -march=arch11 or -march=z13 and can compile a file
  613. # with z13-specific inline assembly, then we can include support for Z13.
  614. # note: -march=z13 is equivalent to -march=arch11 yet some compiler releases
  615. # only support one or the other.
  616. # note: LLVM version 6.x supported -march=z13 yet could not handle vector
  617. # registers in inline assembly, so the check for supporting the -march flag is
  618. # not enough.
  619. ZARCH_TEST_COMPILE=-c $(TOPDIR)/kernel/zarch/damin_z13.c -I$(TOPDIR) -o /dev/null > /dev/null 2> /dev/null
  620. ZARCH_CC_SUPPORTS_ARCH11=$(shell $(CC) -march=arch11 $(ZARCH_TEST_COMPILE) && echo 1)
  621. ZARCH_CC_SUPPORTS_Z13=$(shell $(CC) -march=z13 $(ZARCH_TEST_COMPILE) && echo 1)
  622. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH11), $(ZARCH_CC_SUPPORTS_Z13)), 1)
  623. DYNAMIC_CORE += Z13
  624. CCOMMON_OPT += -DDYN_Z13
  625. else
  626. $(info OpenBLAS: Not building Z13 kernels because the compiler $(CC) does not support it)
  627. endif
  628. # as above for z13, check for -march=arch12 and z14 support in the compiler.
  629. ZARCH_CC_SUPPORTS_ARCH12=$(shell $(CC) -march=arch12 $(ZARCH_TEST_COMPILE) && echo 1)
  630. ZARCH_CC_SUPPORTS_Z14=$(shell $(CC) -march=z14 $(ZARCH_TEST_COMPILE) && echo 1)
  631. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH12), $(ZARCH_CC_SUPPORTS_Z14)), 1)
  632. DYNAMIC_CORE += Z14
  633. CCOMMON_OPT += -DDYN_Z14
  634. else
  635. $(info OpenBLAS: Not building Z14 kernels because the compiler $(CC) does not support it)
  636. endif
  637. endif # ARCH zarch
  638. ifeq ($(ARCH), power)
  639. ifneq ($(C_COMPILER), PGI)
  640. DYNAMIC_CORE = POWER6
  641. DYNAMIC_CORE += POWER8
  642. ifneq ($(C_COMPILER), GCC)
  643. DYNAMIC_CORE += POWER9
  644. DYNAMIC_CORE += POWER10
  645. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  646. endif
  647. ifeq ($(C_COMPILER), GCC)
  648. ifeq ($(GCCVERSIONGT5), 1)
  649. DYNAMIC_CORE += POWER9
  650. else
  651. $(info, OpenBLAS: Your gcc version is too old to build the POWER9 kernels.)
  652. endif
  653. ifeq ($(OSNAME), AIX)
  654. LDVERSIONGTEQ35 := 1
  655. else
  656. LDVERSIONGTEQ35 := $(shell expr `$(CC) -Wl,--version 2> /dev/null | head -1 | cut -f2 -d "." | cut -f1 -d "-"` \>= 35)
  657. endif
  658. ifeq ($(GCCVERSIONGTEQ11)$(LDVERSIONGTEQ35), 11)
  659. DYNAMIC_CORE += POWER10
  660. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  661. else ifeq ($(GCCVERSIONGTEQ10), 1)
  662. ifeq ($(GCCMINORVERSIONGTEQ2)$(LDVERSIONGTEQ35), 11)
  663. DYNAMIC_CORE += POWER10
  664. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  665. endif
  666. else
  667. $(info, OpenBLAS: Your gcc version is too old to build the POWER10 kernels.)
  668. endif
  669. endif
  670. else
  671. DYNAMIC_CORE = POWER8
  672. DYNAMIC_CORE += POWER9
  673. endif
  674. endif
  675. # If DYNAMIC_CORE is not set, DYNAMIC_ARCH cannot do anything, so force it to empty
  676. ifndef DYNAMIC_CORE
  677. override DYNAMIC_ARCH=
  678. endif
  679. endif
  680. ifeq ($(ARCH), ia64)
  681. NO_BINARY_MODE = 1
  682. BINARY_DEFINED = 1
  683. ifeq ($(F_COMPILER), GFORTRAN)
  684. ifeq ($(C_COMPILER), GCC)
  685. # EXPRECISION = 1
  686. # CCOMMON_OPT += -DEXPRECISION
  687. endif
  688. endif
  689. endif
  690. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  691. NO_BINARY_MODE = 1
  692. endif
  693. ifeq ($(ARCH), alpha)
  694. NO_BINARY_MODE = 1
  695. BINARY_DEFINED = 1
  696. endif
  697. ifeq ($(ARCH), arm)
  698. NO_BINARY_MODE = 1
  699. BINARY_DEFINED = 1
  700. CCOMMON_OPT += -marm
  701. FCOMMON_OPT += -marm
  702. # If softfp abi is mentioned on the command line, force it.
  703. ifeq ($(ARM_SOFTFP_ABI), 1)
  704. CCOMMON_OPT += -mfloat-abi=softfp
  705. FCOMMON_OPT += -mfloat-abi=softfp
  706. endif
  707. ifeq ($(OSNAME), Android)
  708. ifeq ($(ARM_SOFTFP_ABI), 1)
  709. EXTRALIB += -lm
  710. else
  711. EXTRALIB += -Wl,-lm_hard
  712. endif
  713. endif
  714. endif
  715. ifeq ($(ARCH), arm64)
  716. NO_BINARY_MODE = 1
  717. BINARY_DEFINED = 1
  718. ifdef INTERFACE64
  719. ifneq ($(INTERFACE64), 0)
  720. ifeq ($(F_COMPILER), GFORTRAN)
  721. FCOMMON_OPT += -fdefault-integer-8
  722. endif
  723. ifeq ($(F_COMPILER), FLANG)
  724. FCOMMON_OPT += -i8
  725. endif
  726. endif
  727. endif
  728. endif
  729. ifeq ($(ARCH), riscv64)
  730. NO_BINARY_MODE = 1
  731. BINARY_DEFINED = 1
  732. ifdef INTERFACE64
  733. ifneq ($(INTERFACE64), 0)
  734. ifeq ($(F_COMPILER), GFORTRAN)
  735. FCOMMON_OPT += -fdefault-integer-8
  736. endif
  737. ifeq ($(F_COMPILER), FLANG)
  738. FCOMMON_OPT += -i8
  739. endif
  740. endif
  741. endif
  742. endif
  743. ifeq ($(ARCH), loongarch64)
  744. NO_BINARY_MODE = 1
  745. BINARY_DEFINED = 1
  746. ifdef INTERFACE64
  747. ifneq ($(INTERFACE64), 0)
  748. ifeq ($(F_COMPILER), GFORTRAN)
  749. FCOMMON_OPT += -fdefault-integer-8
  750. endif
  751. ifeq ($(F_COMPILER), FLANG)
  752. FCOMMON_OPT += -i8
  753. endif
  754. endif
  755. endif
  756. endif
  757. ifeq ($(ARCH), csky)
  758. NO_BINARY_MODE = 1
  759. BINARY_DEFINED = 1
  760. endif
  761. #
  762. # C Compiler dependent settings
  763. #
  764. # ifeq logical or. GCC or CLANG or LSB
  765. # http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
  766. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG LSB))
  767. CCOMMON_OPT += -Wall
  768. COMMON_PROF += -fno-inline
  769. NO_UNINITIALIZED_WARN = -Wno-uninitialized
  770. ifeq ($(QUIET_MAKE), 1)
  771. CCOMMON_OPT += $(NO_UNINITIALIZED_WARN) -Wno-unused
  772. endif
  773. ifdef NO_BINARY_MODE
  774. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  775. ifdef BINARY64
  776. CCOMMON_OPT += -mabi=64
  777. else
  778. CCOMMON_OPT += -mabi=n32
  779. endif
  780. BINARY_DEFINED = 1
  781. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  782. CCOMMON_OPT += -mabi=32
  783. BINARY_DEFINED = 1
  784. endif
  785. ifneq (, $(filter $(CORE), MIPS64_GENERIC))
  786. CCOMMON_OPT += -DNO_MSA
  787. FCOMMON_OPT += -DNO_MSA
  788. endif
  789. ifneq (, $(filter $(CORE),LOONGSON3R3 LOONGSON3R4))
  790. CCOMMON_OPT += -march=loongson3a
  791. FCOMMON_OPT += -march=loongson3a
  792. endif
  793. ifeq ($(CORE), MIPS24K)
  794. CCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  795. FCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  796. endif
  797. ifeq ($(CORE), MIPS1004K)
  798. CCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  799. FCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  800. endif
  801. ifeq ($(CORE), P5600)
  802. CCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  803. FCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  804. endif
  805. ifeq ($(CORE), I6400)
  806. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  807. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  808. endif
  809. ifeq ($(CORE), P6600)
  810. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  811. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  812. endif
  813. ifeq ($(CORE), I6500)
  814. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  815. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  816. endif
  817. ifeq ($(OSNAME), AIX)
  818. BINARY_DEFINED = 1
  819. endif
  820. ifeq ($(ARCH), loongarch64)
  821. LA64_ABI=$(shell $(CC) -mabi=lp64d -c $(TOPDIR)/cpuid_loongarch64.c -o /dev/null > /dev/null 2> /dev/null && echo lp64d)
  822. ifneq ($(LA64_ABI), lp64d)
  823. LA64_ABI=lp64
  824. endif
  825. CCOMMON_OPT += -march=loongarch64 -mabi=$(LA64_ABI)
  826. FCOMMON_OPT += -march=loongarch64 -mabi=$(LA64_ABI)
  827. endif
  828. endif
  829. ifndef BINARY_DEFINED
  830. ifneq ($(OSNAME), AIX)
  831. ifdef BINARY64
  832. ifneq ($(ARCH), riscv64)
  833. CCOMMON_OPT += -m64
  834. endif
  835. else
  836. CCOMMON_OPT += -m32
  837. endif
  838. endif
  839. endif
  840. endif
  841. ifeq ($(C_COMPILER), PGI)
  842. PGCVERSIONGT20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \> 20)
  843. PGCVERSIONEQ20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` == 20)
  844. PGCMINORVERSIONGE11 := $(shell expr `$(CC) --version|sed -n "2p" |cut -d "-" -f 1 |sed -e "s/[^0-9.]//g" |cut -c 4-5` \>= 11)
  845. PGCVERSIONCHECK := $(PGCVERSIONGT20)$(PGCVERSIONEQ20)$(PGCMINORVERSIONGE11)
  846. ifeq ($(PGCVERSIONCHECK), $(filter $(PGCVERSIONCHECK), 100 101 011))
  847. NEWPGI := 1
  848. PGCVERSIONGT21 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \> 21)
  849. PGCVERSIONEQ21 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` == 21)
  850. PGCVERSIONCHECK2 := $(PGCVERSIONGT21)$(PGCVERSIONEQ21)$(PGCMINORVERSIONGE11)
  851. ifeq ($(PGCVERSIONCHECK2), $(filter $(PGCVERSIONCHECK2), 100 101 011))
  852. NEWPGI2 := 1
  853. endif
  854. endif
  855. ifdef BINARY64
  856. ifeq ($(ARCH), x86_64)
  857. ifeq (,$(findstring tp,$(CFLAGS)))
  858. ifneq ($(NEWPGI2),1)
  859. CCOMMON_OPT += -tp p7-64
  860. else
  861. CCOMMON_OPT += -tp px
  862. endif
  863. endif
  864. ifneq ($(NEWPGI),1)
  865. CCOMMON_OPT += -D__MMX__ -Mnollvm
  866. endif
  867. else
  868. ifeq ($(ARCH), power)
  869. ifeq (,$(findstring tp,$(CFLAGS)))
  870. ifeq ($(CORE), POWER8)
  871. CCOMMON_OPT += -tp pwr8
  872. endif
  873. ifeq ($(CORE), POWER9)
  874. CCOMMON_OPT += -tp pwr9
  875. endif
  876. endif
  877. endif
  878. endif
  879. else
  880. ifneq ($(NEWPGI2),1)
  881. ifeq (,$(findstring tp,$(CFLAGS)))
  882. CCOMMON_OPT += -tp p7
  883. else
  884. CCOMMON_OPT += -tp px
  885. endif
  886. endif
  887. endif
  888. endif
  889. ifeq ($(C_COMPILER), PATHSCALE)
  890. ifdef BINARY64
  891. CCOMMON_OPT += -m64
  892. else
  893. CCOMMON_OPT += -m32
  894. endif
  895. endif
  896. #
  897. # Fortran Compiler dependent settings
  898. #
  899. ifeq ($(F_COMPILER), NAG)
  900. FCOMMON_OPT += -dcfuns -recursive -ieee=full -w=obs -thread_safe
  901. ifdef INTERFACE64
  902. ifneq ($(INTERFACE64), 0)
  903. FCOMMON_OPT += -i8
  904. endif
  905. endif
  906. ifeq ($(USE_OPENMP), 1)
  907. FCOMMON_OPT += -openmp
  908. endif
  909. endif
  910. ifeq ($(F_COMPILER), FLANG)
  911. CCOMMON_OPT += -DF_INTERFACE_FLANG
  912. FCOMMON_OPT += -Mrecursive -Kieee
  913. ifeq ($(OSNAME), Linux)
  914. ifeq ($(ARCH), x86_64)
  915. FLANG_VENDOR := $(shell $(FC) --version|head -1 |cut -f 1 -d " ")
  916. ifeq ($(FLANG_VENDOR), AMD)
  917. FCOMMON_OPT += -fno-unroll-loops
  918. endif
  919. endif
  920. endif
  921. ifdef BINARY64
  922. ifdef INTERFACE64
  923. ifneq ($(INTERFACE64), 0)
  924. FCOMMON_OPT += -i8
  925. endif
  926. endif
  927. FCOMMON_OPT += -Wall
  928. else
  929. FCOMMON_OPT += -Wall
  930. endif
  931. ifeq ($(USE_OPENMP), 1)
  932. FCOMMON_OPT += -fopenmp
  933. endif
  934. endif
  935. ifeq ($(F_COMPILER), G77)
  936. CCOMMON_OPT += -DF_INTERFACE_G77
  937. FCOMMON_OPT += -Wall
  938. ifndef NO_BINARY_MODE
  939. ifneq ($(OSNAME), AIX)
  940. ifdef BINARY64
  941. FCOMMON_OPT += -m64
  942. else
  943. FCOMMON_OPT += -m32
  944. endif
  945. endif
  946. endif
  947. endif
  948. ifeq ($(F_COMPILER), G95)
  949. CCOMMON_OPT += -DF_INTERFACE_G95
  950. FCOMMON_OPT += -Wall
  951. ifneq ($(OSNAME), AIX)
  952. ifndef NO_BINARY_MODE
  953. ifdef BINARY64
  954. FCOMMON_OPT += -m64
  955. else
  956. FCOMMON_OPT += -m32
  957. endif
  958. endif
  959. ifneq ($(NO_LAPACKE), 1)
  960. FCOMMON_OPT += -fno-second-underscore
  961. endif
  962. endif
  963. endif
  964. ifeq ($(F_COMPILER), $(filter $(F_COMPILER),GFORTRAN FLANGNEW))
  965. CCOMMON_OPT += -DF_INTERFACE_GFORT
  966. ifeq ($(F_COMPILER), GFORTRAN)
  967. FCOMMON_OPT += -Wall
  968. # make single-threaded LAPACK calls thread-safe #1847
  969. FCOMMON_OPT += -frecursive
  970. # work around ABI problem with passing single-character arguments
  971. FCOMMON_OPT += -fno-optimize-sibling-calls
  972. #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
  973. ifneq ($(NOFORTRAN), 1)
  974. ifneq ($(NOFORTRAN), 2)
  975. ifneq ($(NO_LAPACK), 1)
  976. EXTRALIB += -lgfortran
  977. endif
  978. endif
  979. endif
  980. endif
  981. ifdef NO_BINARY_MODE
  982. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  983. ifdef BINARY64
  984. FCOMMON_OPT += -mabi=64
  985. else
  986. FCOMMON_OPT += -mabi=n32
  987. endif
  988. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  989. FCOMMON_OPT += -mabi=32
  990. endif
  991. else
  992. ifdef BINARY64
  993. ifneq ($(OSNAME), AIX)
  994. ifneq ($(ARCH), riscv64)
  995. FCOMMON_OPT += -m64
  996. endif
  997. endif
  998. ifdef INTERFACE64
  999. ifneq ($(INTERFACE64), 0)
  1000. FCOMMON_OPT += -fdefault-integer-8
  1001. endif
  1002. endif
  1003. else
  1004. ifneq ($(OSNAME), AIX)
  1005. FCOMMON_OPT += -m32
  1006. endif
  1007. endif
  1008. endif
  1009. ifeq ($(USE_OPENMP), 1)
  1010. FCOMMON_OPT += -fopenmp
  1011. endif
  1012. endif
  1013. ifeq ($(F_COMPILER), INTEL)
  1014. CCOMMON_OPT += -DF_INTERFACE_INTEL
  1015. ifdef INTERFACE64
  1016. ifneq ($(INTERFACE64), 0)
  1017. FCOMMON_OPT += -i8
  1018. endif
  1019. endif
  1020. FCOMMON_OPT += -recursive -fp-model strict -assume protect-parens
  1021. ifeq ($(USE_OPENMP), 1)
  1022. FCOMMON_OPT += -fopenmp
  1023. endif
  1024. endif
  1025. ifeq ($(F_COMPILER), FUJITSU)
  1026. CCOMMON_OPT += -DF_INTERFACE_FUJITSU
  1027. ifeq ($(USE_OPENMP), 1)
  1028. FCOMMON_OPT += -openmp
  1029. endif
  1030. endif
  1031. ifeq ($(F_COMPILER), IBM)
  1032. CCOMMON_OPT += -DF_INTERFACE_IBM
  1033. FEXTRALIB += -lxlf90
  1034. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG))
  1035. FCOMMON_OPT += -qextname -qzerosize
  1036. endif
  1037. # FCOMMON_OPT += -qarch=440
  1038. ifdef BINARY64
  1039. FCOMMON_OPT += -q64
  1040. ifdef INTERFACE64
  1041. ifneq ($(INTERFACE64), 0)
  1042. FCOMMON_OPT += -qintsize=8
  1043. endif
  1044. endif
  1045. else
  1046. FCOMMON_OPT += -q32
  1047. endif
  1048. ifeq ($(USE_OPENMP), 1)
  1049. FCOMMON_OPT += -openmp
  1050. endif
  1051. endif
  1052. ifeq ($(F_COMPILER), PGI)
  1053. CCOMMON_OPT += -DF_INTERFACE_PGI
  1054. COMMON_PROF += -DPGICOMPILER
  1055. ifdef BINARY64
  1056. ifdef INTERFACE64
  1057. ifneq ($(INTERFACE64), 0)
  1058. FCOMMON_OPT += -i8
  1059. endif
  1060. endif
  1061. ifeq ($(ARCH), x86_64)
  1062. ifneq ($(NEWPGI2),1)
  1063. FCOMMON_OPT += -tp p7-64
  1064. else
  1065. FCOMMON_OPT += -tp px
  1066. endif
  1067. else
  1068. ifeq ($(ARCH), power)
  1069. ifeq ($(CORE), POWER6)
  1070. $(warning NVIDIA HPC compilers do not support POWER6.)
  1071. endif
  1072. ifeq ($(CORE), POWER8)
  1073. FCOMMON_OPT += -tp pwr8
  1074. endif
  1075. ifeq ($(CORE), POWER9)
  1076. FCOMMON_OPT += -tp pwr9
  1077. endif
  1078. ifeq ($(CORE), POWER10)
  1079. $(warning NVIDIA HPC compilers do not support POWER10.)
  1080. endif
  1081. endif
  1082. endif
  1083. else
  1084. FCOMMON_OPT += -tp p7
  1085. endif
  1086. FCOMMON_OPT += -Mrecursive -Kieee
  1087. ifeq ($(USE_OPENMP), 1)
  1088. FCOMMON_OPT += -mp
  1089. endif
  1090. endif
  1091. ifeq ($(F_COMPILER), PATHSCALE)
  1092. CCOMMON_OPT += -DF_INTERFACE_PATHSCALE
  1093. ifdef BINARY64
  1094. ifdef INTERFACE64
  1095. ifneq ($(INTERFACE64), 0)
  1096. FCOMMON_OPT += -i8
  1097. endif
  1098. endif
  1099. endif
  1100. ifeq ($(USE_OPENMP), 1)
  1101. FCOMMON_OPT += -mp
  1102. endif
  1103. endif
  1104. ifeq ($(F_COMPILER), OPEN64)
  1105. CCOMMON_OPT += -DF_INTERFACE_OPEN64
  1106. ifdef BINARY64
  1107. ifdef INTERFACE64
  1108. ifneq ($(INTERFACE64), 0)
  1109. FCOMMON_OPT += -i8
  1110. endif
  1111. endif
  1112. endif
  1113. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  1114. ifndef BINARY64
  1115. FCOMMON_OPT += -n32
  1116. else
  1117. FCOMMON_OPT += -n64
  1118. endif
  1119. ifeq ($(CORE), LOONGSON3R3)
  1120. FCOMMON_OPT += -loongson3 -static
  1121. endif
  1122. ifeq ($(CORE), LOONGSON3R4)
  1123. FCOMMON_OPT += -loongson3 -static
  1124. endif
  1125. else
  1126. ifndef BINARY64
  1127. FCOMMON_OPT += -m32
  1128. else
  1129. FCOMMON_OPT += -m64
  1130. endif
  1131. endif
  1132. ifeq ($(USE_OPENMP), 1)
  1133. FEXTRALIB += -lstdc++
  1134. FCOMMON_OPT += -mp
  1135. endif
  1136. endif
  1137. ifeq ($(C_COMPILER), OPEN64)
  1138. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  1139. ifndef BINARY64
  1140. CCOMMON_OPT += -n32
  1141. else
  1142. CCOMMON_OPT += -n64
  1143. endif
  1144. ifeq ($(CORE), LOONGSON3R3)
  1145. CCOMMON_OPT += -loongson3 -static
  1146. endif
  1147. ifeq ($(CORE), LOONGSON3R4)
  1148. CCOMMON_OPT += -loongson3 -static
  1149. endif
  1150. else
  1151. ifndef BINARY64
  1152. CCOMMON_OPT += -m32
  1153. else
  1154. CCOMMON_OPT += -m64
  1155. endif
  1156. endif
  1157. endif
  1158. ifeq ($(C_COMPILER), SUN)
  1159. CCOMMON_OPT += -w
  1160. ifeq ($(ARCH), x86)
  1161. CCOMMON_OPT += -m32
  1162. else
  1163. ifdef BINARY64
  1164. CCOMMON_OPT += -m64
  1165. else
  1166. CCOMMON_OPT += -m32
  1167. endif
  1168. endif
  1169. endif
  1170. ifeq ($(F_COMPILER), SUN)
  1171. CCOMMON_OPT += -DF_INTERFACE_SUN
  1172. FCOMMON_OPT += -ftrap=%none -xrecursive
  1173. ifeq ($(ARCH), x86)
  1174. FCOMMON_OPT += -m32
  1175. else
  1176. ifdef BINARY64
  1177. FCOMMON_OPT += -m64
  1178. else
  1179. FCOMMON_OPT += -m32
  1180. endif
  1181. endif
  1182. ifeq ($(USE_OPENMP), 1)
  1183. FCOMMON_OPT += -xopenmp=parallel
  1184. endif
  1185. endif
  1186. ifeq ($(F_COMPILER), COMPAQ)
  1187. CCOMMON_OPT += -DF_INTERFACE_COMPAQ
  1188. ifeq ($(USE_OPENMP), 1)
  1189. FCOMMON_OPT += -openmp
  1190. endif
  1191. endif
  1192. ifeq ($(F_COMPILER), CRAY)
  1193. CCOMMON_OPT += -DF_INTERFACE_INTEL
  1194. FCOMMON_OPT += -hnopattern
  1195. ifdef INTERFACE64
  1196. ifneq ($(INTERFACE64), 0)
  1197. FCOMMON_OPT += -s integer64
  1198. endif
  1199. endif
  1200. ifneq ($(USE_OPENMP), 1)
  1201. FCOMMON_OPT += -O noomp
  1202. endif
  1203. endif
  1204. ifdef BINARY64
  1205. ifdef INTERFACE64
  1206. ifneq ($(INTERFACE64), 0)
  1207. CCOMMON_OPT +=
  1208. #-DUSE64BITINT
  1209. endif
  1210. endif
  1211. endif
  1212. ifeq ($(NEED_PIC), 1)
  1213. ifeq ($(C_COMPILER), IBM)
  1214. CCOMMON_OPT += -qpic=large
  1215. else
  1216. CCOMMON_OPT += -fPIC
  1217. endif
  1218. ifeq ($(F_COMPILER), SUN)
  1219. FCOMMON_OPT += -pic
  1220. else ifeq ($(F_COMPILER), NAG)
  1221. FCOMMON_OPT += -PIC
  1222. else ifeq ($(F_COMPILER), IBM)
  1223. FCOMMON_OPT += -qpic=large
  1224. else
  1225. FCOMMON_OPT += -fPIC
  1226. endif
  1227. endif
  1228. ifeq ($(DYNAMIC_ARCH), 1)
  1229. CCOMMON_OPT += -DDYNAMIC_ARCH
  1230. endif
  1231. ifeq ($(DYNAMIC_OLDER), 1)
  1232. CCOMMON_OPT += -DDYNAMIC_OLDER
  1233. endif
  1234. ifeq ($(C_LAPACK), 1)
  1235. CCOMMON_OPT += -DC_LAPACK
  1236. endif
  1237. ifeq ($(NO_LAPACK), 1)
  1238. CCOMMON_OPT += -DNO_LAPACK
  1239. #Disable LAPACK C interface
  1240. NO_LAPACKE = 1
  1241. endif
  1242. ifeq ($(NO_LAPACKE), 1)
  1243. CCOMMON_OPT += -DNO_LAPACKE
  1244. endif
  1245. ifeq ($(NO_AVX), 1)
  1246. CCOMMON_OPT += -DNO_AVX
  1247. endif
  1248. ifeq ($(ARCH), x86)
  1249. CCOMMON_OPT += -DNO_AVX
  1250. endif
  1251. ifeq ($(NO_AVX2), 1)
  1252. CCOMMON_OPT += -DNO_AVX2
  1253. endif
  1254. ifeq ($(NO_AVX512), 1)
  1255. CCOMMON_OPT += -DNO_AVX512
  1256. endif
  1257. ifeq ($(NO_SVE), 1)
  1258. CCOMMON_OPT += -DNO_SVE
  1259. endif
  1260. ifdef SMP
  1261. CCOMMON_OPT += -DSMP_SERVER
  1262. ifeq ($(ARCH), mips64)
  1263. USE_SIMPLE_THREADED_LEVEL3 = 1
  1264. endif
  1265. ifeq ($(USE_OPENMP), 1)
  1266. # USE_SIMPLE_THREADED_LEVEL3 = 1
  1267. # NO_AFFINITY = 1
  1268. CCOMMON_OPT += -DUSE_OPENMP
  1269. endif
  1270. ifeq ($(BIGNUMA), 1)
  1271. CCOMMON_OPT += -DBIGNUMA
  1272. endif
  1273. endif
  1274. ifeq ($(NO_WARMUP), 1)
  1275. CCOMMON_OPT += -DNO_WARMUP
  1276. endif
  1277. ifeq ($(CONSISTENT_FPCSR), 1)
  1278. CCOMMON_OPT += -DCONSISTENT_FPCSR
  1279. endif
  1280. # Only for development
  1281. # CCOMMON_OPT += -DPARAMTEST
  1282. # CCOMMON_OPT += -DPREFETCHTEST
  1283. # CCOMMON_OPT += -DNO_SWITCHING
  1284. # USE_PAPI = 1
  1285. ifdef USE_PAPI
  1286. CCOMMON_OPT += -DUSE_PAPI
  1287. EXTRALIB += -lpapi -lperfctr
  1288. endif
  1289. ifdef BUFFERSIZE
  1290. CCOMMON_OPT += -DBUFFERSIZE=$(BUFFERSIZE)
  1291. endif
  1292. ifdef DYNAMIC_THREADS
  1293. CCOMMON_OPT += -DDYNAMIC_THREADS
  1294. endif
  1295. CCOMMON_OPT += -DMAX_CPU_NUMBER=$(NUM_THREADS)
  1296. CCOMMON_OPT += -DMAX_PARALLEL_NUMBER=$(NUM_PARALLEL)
  1297. ifdef USE_SIMPLE_THREADED_LEVEL3
  1298. CCOMMON_OPT += -DUSE_SIMPLE_THREADED_LEVEL3
  1299. endif
  1300. ifeq ($(USE_TLS), 1)
  1301. CCOMMON_OPT += -DUSE_TLS
  1302. endif
  1303. ifeq ($(BUILD_BFLOAT16), 1)
  1304. CCOMMON_OPT += -DBUILD_BFLOAT16
  1305. endif
  1306. ifeq ($(BUILD_SINGLE), 1)
  1307. CCOMMON_OPT += -DBUILD_SINGLE=1
  1308. endif
  1309. ifeq ($(BUILD_DOUBLE), 1)
  1310. CCOMMON_OPT += -DBUILD_DOUBLE=1
  1311. endif
  1312. ifeq ($(BUILD_COMPLEX), 1)
  1313. CCOMMON_OPT += -DBUILD_COMPLEX=1
  1314. endif
  1315. ifeq ($(BUILD_COMPLEX16), 1)
  1316. CCOMMON_OPT += -DBUILD_COMPLEX16=1
  1317. endif
  1318. CCOMMON_OPT += -DVERSION=\"$(VERSION)\"
  1319. ifndef SYMBOLPREFIX
  1320. SYMBOLPREFIX =
  1321. endif
  1322. ifndef SYMBOLSUFFIX
  1323. SYMBOLSUFFIX =
  1324. endif
  1325. ifndef LIBSONAMEBASE
  1326. LIBSONAMEBASE = openblas
  1327. endif
  1328. ifndef LIBNAMEPREFIX
  1329. LIBNAMEPREFIX =
  1330. endif
  1331. SYMPREFIX=$(SYMBOLPREFIX)
  1332. ifeq ($(SYMBOLPREFIX),$(LIBNAMEPREFIX))
  1333. SYMPREFIX=
  1334. endif
  1335. SYMSUFFIX=$(SYMBOLSUFFIX)
  1336. ifeq ($(SYMBOLSUFFIX),$(LIBNAMESUFFIX))
  1337. SYMSUFFIX=
  1338. endif
  1339. ifndef LIBNAMESUFFIX
  1340. LIBNAMEBASE = $(SYMPREFIX)$(LIBSONAMEBASE)$(SYMSUFFIX)
  1341. else
  1342. LIBNAMEBASE = $(SYMPREFIX)$(LIBSONAMEBASE)$(SYMSUFFIX)$(LIBNAMESUFFIX)
  1343. endif
  1344. ifeq ($(OSNAME), CYGWIN_NT)
  1345. LIBPREFIX = cyg$(LIBNAMEPREFIX)$(LIBNAMEBASE)
  1346. else
  1347. LIBPREFIX = lib$(LIBNAMEPREFIX)$(LIBNAMEBASE)
  1348. endif
  1349. KERNELDIR = $(TOPDIR)/kernel/$(ARCH)
  1350. include $(TOPDIR)/Makefile.$(ARCH)
  1351. ifneq ($(C_COMPILER), PGI)
  1352. ifneq ($(C_COMPILER), SUN)
  1353. CCOMMON_OPT += -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME
  1354. endif
  1355. endif
  1356. CCOMMON_OPT += -DASMNAME=$(FU)$(*F) -DASMFNAME=$(FU)$(*F)$(BU) -DNAME=$(*F)$(BU) -DCNAME=$(*F) -DCHAR_NAME=\"$(*F)$(BU)\" -DCHAR_CNAME=\"$(*F)\"
  1357. ifeq ($(CORE), PPC440)
  1358. CCOMMON_OPT += -DALLOC_QALLOC
  1359. endif
  1360. ifeq ($(CORE), PPC440FP2)
  1361. STATIC_ALLOCATION = 1
  1362. endif
  1363. ifneq ($(OSNAME), Linux)
  1364. NO_AFFINITY = 1
  1365. endif
  1366. ifneq ($(ARCH), x86_64)
  1367. ifneq ($(ARCH), x86)
  1368. NO_AFFINITY = 1
  1369. endif
  1370. endif
  1371. ifdef NO_AFFINITY
  1372. ifeq ($(NO_AFFINITY), 0)
  1373. override undefine NO_AFFINITY
  1374. else
  1375. CCOMMON_OPT += -DNO_AFFINITY
  1376. endif
  1377. endif
  1378. ifdef FUNCTION_PROFILE
  1379. CCOMMON_OPT += -DFUNCTION_PROFILE
  1380. endif
  1381. ifdef HUGETLB_ALLOCATION
  1382. CCOMMON_OPT += -DALLOC_HUGETLB
  1383. endif
  1384. ifdef HUGETLBFILE_ALLOCATION
  1385. CCOMMON_OPT += -DALLOC_HUGETLBFILE -DHUGETLB_FILE_NAME=$(HUGETLBFILE_ALLOCATION)
  1386. endif
  1387. ifdef STATIC_ALLOCATION
  1388. CCOMMON_OPT += -DALLOC_STATIC
  1389. endif
  1390. ifdef DEVICEDRIVER_ALLOCATION
  1391. CCOMMON_OPT += -DALLOC_DEVICEDRIVER -DDEVICEDRIVER_NAME=\"/dev/mapper\"
  1392. endif
  1393. ifdef MIXED_MEMORY_ALLOCATION
  1394. CCOMMON_OPT += -DMIXED_MEMORY_ALLOCATION
  1395. endif
  1396. ifeq ($(OSNAME), SunOS)
  1397. TAR = gtar
  1398. PATCH = gpatch
  1399. GREP = ggrep
  1400. AWK = nawk
  1401. else
  1402. TAR = tar
  1403. PATCH = patch
  1404. GREP = grep
  1405. AWK = awk
  1406. endif
  1407. ifndef MD5SUM
  1408. MD5SUM = md5sum
  1409. endif
  1410. REVISION = -r$(VERSION)
  1411. MAJOR_VERSION = $(word 1,$(subst ., ,$(VERSION)))
  1412. ifeq ($(DEBUG), 1)
  1413. COMMON_OPT += -g
  1414. endif
  1415. ifeq ($(DEBUG), 1)
  1416. FCOMMON_OPT += -g
  1417. endif
  1418. ifndef COMMON_OPT
  1419. COMMON_OPT = -O2
  1420. endif
  1421. ifndef FCOMMON_OPT
  1422. FCOMMON_OPT = -O2 -frecursive
  1423. endif
  1424. override CFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
  1425. override PFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
  1426. override FFLAGS += $(COMMON_OPT) $(FCOMMON_OPT)
  1427. override FPFLAGS += $(FCOMMON_OPT) $(COMMON_PROF)
  1428. #MAKEOVERRIDES =
  1429. ifeq ($(NEED_PIC), 1)
  1430. ifeq (,$(findstring PIC,$(FFLAGS)))
  1431. ifneq ($(F_COMPILER),IBM)
  1432. override FFLAGS += -fPIC
  1433. endif
  1434. endif
  1435. endif
  1436. #For LAPACK Fortran codes.
  1437. #Disable -fopenmp for LAPACK Fortran codes on Windows.
  1438. ifdef OS_WINDOWS
  1439. LAPACK_FFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FFLAGS))
  1440. LAPACK_FPFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FPFLAGS))
  1441. else
  1442. LAPACK_FFLAGS := $(FFLAGS)
  1443. LAPACK_FPFLAGS := $(FPFLAGS)
  1444. endif
  1445. ifeq ($(F_COMPILER),NAG)
  1446. LAPACK_FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1447. override FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1448. endif
  1449. ifeq ($(F_COMPILER),CRAY)
  1450. LAPACK_FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1451. override FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1452. endif
  1453. ifeq ($(F_COMPILER),FLANGNEW)
  1454. LAPACK_FFLAGS := $(filter-out -m32 -m64 -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1455. override FFLAGS := $(filter-out -m32 -m64 -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1456. endif
  1457. LAPACK_CFLAGS = $(CFLAGS)
  1458. LAPACK_CFLAGS += -DHAVE_LAPACK_CONFIG_H
  1459. ifdef INTERFACE64
  1460. ifneq ($(INTERFACE64), 0)
  1461. LAPACK_CFLAGS += -DLAPACK_ILP64
  1462. endif
  1463. endif
  1464. ifdef OS_WINDOWS
  1465. LAPACK_CFLAGS += -DOPENBLAS_OS_WINDOWS
  1466. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1467. endif
  1468. ifeq ($(C_COMPILER), LSB)
  1469. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1470. endif
  1471. ifndef SUFFIX
  1472. SUFFIX = o
  1473. endif
  1474. ifndef PSUFFIX
  1475. PSUFFIX = po
  1476. endif
  1477. ifndef LIBSUFFIX
  1478. LIBSUFFIX = a
  1479. endif
  1480. ifneq ($(DYNAMIC_ARCH), 1)
  1481. ifndef SMP
  1482. LIBNAME = $(LIBPREFIX)_$(LIBCORE)$(REVISION).$(LIBSUFFIX)
  1483. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)$(REVISION)_p.$(LIBSUFFIX)
  1484. else
  1485. LIBNAME = $(LIBPREFIX)_$(LIBCORE)p$(REVISION).$(LIBSUFFIX)
  1486. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)p$(REVISION)_p.$(LIBSUFFIX)
  1487. endif
  1488. else
  1489. ifndef SMP
  1490. LIBNAME = $(LIBPREFIX)$(REVISION).$(LIBSUFFIX)
  1491. LIBNAME_P = $(LIBPREFIX)$(REVISION)_p.$(LIBSUFFIX)
  1492. else
  1493. LIBNAME = $(LIBPREFIX)p$(REVISION).$(LIBSUFFIX)
  1494. LIBNAME_P = $(LIBPREFIX)p$(REVISION)_p.$(LIBSUFFIX)
  1495. endif
  1496. endif
  1497. ifeq ($(FIXED_LIBNAME),1)
  1498. LIBNAME = lib$(LIBNAMEPREFIX)$(LIBSONAMEBASE)$(LIBNAMESUFFIX).$(LIBSUFFIX)
  1499. LIBNAME_P = lib$(LIBNAMEPREFIX)$(LISOBNAMEBASE)$(LIBNAMESUFFIX)_p.$(LIBSUFFIX)
  1500. endif
  1501. LIBDLLNAME = $(LIBPREFIX).dll
  1502. IMPLIBNAME = lib$(LIBNAMEBASE).dll.a
  1503. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.so)
  1504. LIBDYNNAME = $(LIBNAME:.$(LIBSUFFIX)=.dylib)
  1505. LIBDEFNAME = $(LIBNAME:.$(LIBSUFFIX)=.def)
  1506. LIBEXPNAME = $(LIBNAME:.$(LIBSUFFIX)=.exp)
  1507. LIBZIPNAME = $(LIBNAME:.$(LIBSUFFIX)=.zip)
  1508. LIBS = $(TOPDIR)/$(LIBNAME)
  1509. LIBS_P = $(TOPDIR)/$(LIBNAME_P)
  1510. LIB_COMPONENTS = BLAS
  1511. ifneq ($(NO_CBLAS), 1)
  1512. LIB_COMPONENTS += CBLAS
  1513. endif
  1514. ifneq ($(NO_LAPACK), 1)
  1515. LIB_COMPONENTS += LAPACK
  1516. ifneq ($(NO_LAPACKE), 1)
  1517. LIB_COMPONENTS += LAPACKE
  1518. endif
  1519. ifeq ($(BUILD_RELAPACK), 1)
  1520. LIB_COMPONENTS += ReLAPACK
  1521. endif
  1522. endif
  1523. ifeq ($(ONLY_CBLAS), 1)
  1524. LIB_COMPONENTS = CBLAS
  1525. endif
  1526. export OSNAME
  1527. export ARCH
  1528. export CORE
  1529. export LIBCORE
  1530. export __BYTE_ORDER__
  1531. export ELF_VERSION
  1532. export PGCPATH
  1533. export CONFIG
  1534. export CC
  1535. export FC
  1536. export BU
  1537. export FU
  1538. export NEED2UNDERSCORES
  1539. export USE_THREAD
  1540. export NUM_THREADS
  1541. export NUM_CORES
  1542. export SMP
  1543. export MAKEFILE_RULE
  1544. export NEED_PIC
  1545. export BINARY
  1546. export BINARY32
  1547. export BINARY64
  1548. export F_COMPILER
  1549. export C_COMPILER
  1550. export USE_OPENMP
  1551. export CROSS
  1552. export CROSS_SUFFIX
  1553. export NOFORTRAN
  1554. export C_LAPACK
  1555. export NO_FBLAS
  1556. export EXTRALIB
  1557. export CEXTRALIB
  1558. export FEXTRALIB
  1559. export HAVE_SSE
  1560. export HAVE_SSE2
  1561. export HAVE_SSE3
  1562. export HAVE_SSSE3
  1563. export HAVE_SSE4_1
  1564. export HAVE_SSE4_2
  1565. export HAVE_SSE4A
  1566. export HAVE_SSE5
  1567. export HAVE_AVX
  1568. export HAVE_AVX2
  1569. export HAVE_FMA3
  1570. export HAVE_VFP
  1571. export HAVE_VFPV3
  1572. export HAVE_VFPV4
  1573. export HAVE_NEON
  1574. ifndef NO_MSA
  1575. export HAVE_MSA
  1576. export MSA_FLAGS
  1577. endif
  1578. export KERNELDIR
  1579. export FUNCTION_PROFILE
  1580. export TARGET_CORE
  1581. export NO_AVX512
  1582. export NO_AVX2
  1583. export BUILD_BFLOAT16
  1584. export NO_LSX
  1585. export NO_LASX
  1586. export SBGEMM_UNROLL_M
  1587. export SBGEMM_UNROLL_N
  1588. export SGEMM_UNROLL_M
  1589. export SGEMM_UNROLL_N
  1590. export DGEMM_UNROLL_M
  1591. export DGEMM_UNROLL_N
  1592. export QGEMM_UNROLL_M
  1593. export QGEMM_UNROLL_N
  1594. export CGEMM_UNROLL_M
  1595. export CGEMM_UNROLL_N
  1596. export ZGEMM_UNROLL_M
  1597. export ZGEMM_UNROLL_N
  1598. export XGEMM_UNROLL_M
  1599. export XGEMM_UNROLL_N
  1600. export CGEMM3M_UNROLL_M
  1601. export CGEMM3M_UNROLL_N
  1602. export ZGEMM3M_UNROLL_M
  1603. export ZGEMM3M_UNROLL_N
  1604. export XGEMM3M_UNROLL_M
  1605. export XGEMM3M_UNROLL_N
  1606. ifdef USE_CUDA
  1607. export CUDADIR
  1608. export CUCC
  1609. export CUFLAGS
  1610. export CULIB
  1611. endif
  1612. .SUFFIXES: .$(PSUFFIX) .$(SUFFIX) .f
  1613. .f.$(SUFFIX):
  1614. $(FC) $(FFLAGS) -c $< -o $(@F)
  1615. .f.$(PSUFFIX):
  1616. $(FC) $(FPFLAGS) -pg -c $< -o $(@F)
  1617. ifdef BINARY64
  1618. PATHSCALEPATH = /opt/pathscale/lib/3.1
  1619. PGIPATH = /opt/pgi/linux86-64/7.1-5/lib
  1620. else
  1621. PATHSCALEPATH = /opt/pathscale/lib/3.1/32
  1622. PGIPATH = /opt/pgi/linux86/7.1-5/lib
  1623. endif
  1624. ACMLPATH = /opt/acml/4.3.0
  1625. ifneq ($(OSNAME), Darwin)
  1626. MKLPATH = /opt/intel/mkl/10.2.2.025/lib
  1627. else
  1628. MKLPATH = /Library/Frameworks/Intel_MKL.framework/Versions/10.0.1.014/lib
  1629. endif
  1630. ATLASPATH = /opt/atlas/3.9.17/opteron
  1631. FLAMEPATH = $(HOME)/flame/lib
  1632. ifneq ($(OSNAME), SunOS)
  1633. SUNPATH = /opt/sunstudio12.1
  1634. else
  1635. SUNPATH = /opt/SUNWspro
  1636. endif