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 41 kB

6 years ago
1 year ago
4 years ago
13 years ago
4 years ago
1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285128612871288128912901291129212931294129512961297129812991300130113021303130413051306130713081309131013111312131313141315131613171318131913201321132213231324132513261327132813291330133113321333133413351336133713381339134013411342134313441345134613471348134913501351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384138513861387138813891390139113921393139413951396139713981399140014011402140314041405140614071408140914101411141214131414141514161417141814191420142114221423142414251426142714281429143014311432143314341435143614371438143914401441144214431444144514461447144814491450145114521453145414551456145714581459146014611462146314641465146614671468146914701471147214731474147514761477147814791480148114821483148414851486148714881489149014911492149314941495149614971498149915001501150215031504150515061507150815091510151115121513151415151516151715181519152015211522152315241525152615271528152915301531153215331534153515361537153815391540154115421543154415451546154715481549155015511552155315541555155615571558155915601561156215631564156515661567156815691570157115721573157415751576157715781579158015811582158315841585158615871588158915901591159215931594159515961597159815991600160116021603160416051606160716081609161016111612161316141615161616171618161916201621162216231624162516261627162816291630163116321633163416351636163716381639164016411642164316441645164616471648164916501651165216531654165516561657165816591660166116621663166416651666166716681669167016711672167316741675167616771678167916801681168216831684168516861687168816891690169116921693169416951696169716981699170017011702170317041705170617071708170917101711171217131714171517161717171817191720172117221723172417251726172717281729173017311732173317341735173617371738173917401741174217431744174517461747174817491750175117521753175417551756175717581759176017611762176317641765176617671768176917701771177217731774177517761777177817791780178117821783178417851786178717881789179017911792179317941795179617971798179918001801180218031804180518061807180818091810181118121813181418151816181718181819182018211822182318241825182618271828182918301831183218331834183518361837183818391840184118421843184418451846184718481849185018511852185318541855185618571858185918601861186218631864186518661867186818691870187118721873187418751876187718781879188018811882188318841885188618871888188918901891189218931894189518961897189818991900190119021903190419051906190719081909191019111912191319141915191619171918191919201921192219231924192519261927192819291930193119321933193419351936193719381939
  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. else ifeq ($(ARCH), arm64)
  238. SMALL_MATRIX_OPT = 1
  239. endif
  240. ifeq ($(ARCH), loongarch64)
  241. SMALL_MATRIX_OPT = 1
  242. endif
  243. ifeq ($(ARCH), arm64)
  244. GEMM_GEMV_FORWARD = 1
  245. endif
  246. ifeq ($(ARCH), riscv)
  247. GEMM_GEMV_FORWARD = 1
  248. endif
  249. ifeq ($(ARCH), power)
  250. GEMM_GEMV_FORWARD = 1
  251. GEMM_GEMV_FORWARD_BF16 = 1
  252. endif
  253. ifeq ($(SMALL_MATRIX_OPT), 1)
  254. CCOMMON_OPT += -DSMALL_MATRIX_OPT
  255. endif
  256. ifneq ($(ONLY_CBLAS), 1)
  257. ifeq ($(GEMM_GEMV_FORWARD), 1)
  258. CCOMMON_OPT += -DGEMM_GEMV_FORWARD
  259. endif
  260. ifeq ($(GEMM_GEMV_FORWARD_BF16), 1)
  261. CCOMMON_OPT += -DGEMM_GEMV_FORWARD_BF16
  262. endif
  263. endif
  264. # This operation is expensive, so execution should be once.
  265. ifndef GOTOBLAS_MAKEFILE
  266. export GOTOBLAS_MAKEFILE = 1
  267. # Generating Makefile.conf and config.h
  268. 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)
  269. endif
  270. ifndef TARGET_CORE
  271. -include $(TOPDIR)/Makefile.conf
  272. else
  273. HAVE_NEON=
  274. HAVE_VFP=
  275. HAVE_VFPV3=
  276. HAVE_VFPV4=
  277. HAVE_MMX=
  278. HAVE_SSE=
  279. HAVE_SSE2=
  280. HAVE_SSE3=
  281. HAVE_SSSE3=
  282. HAVE_SSE4_1=
  283. HAVE_SSE4_2=
  284. HAVE_SSE4A=
  285. HAVE_SSE5=
  286. HAVE_AVX=
  287. HAVE_AVX2=
  288. HAVE_FMA3=
  289. include $(TOPDIR)/Makefile_kernel.conf
  290. endif
  291. ifndef NUM_PARALLEL
  292. NUM_PARALLEL = 1
  293. endif
  294. ifndef NUM_THREADS
  295. NUM_THREADS = $(NUM_CORES)
  296. endif
  297. ifeq ($(NUM_THREADS), 1)
  298. override USE_THREAD = 0
  299. override USE_OPENMP = 0
  300. endif
  301. ifdef USE_THREAD
  302. ifeq ($(USE_THREAD), 0)
  303. SMP =
  304. else
  305. SMP = 1
  306. endif
  307. else
  308. ifeq ($(NUM_THREADS), 1)
  309. SMP =
  310. else
  311. SMP = 1
  312. endif
  313. endif
  314. ifeq ($(SMP), 1)
  315. USE_LOCKING =
  316. endif
  317. ifndef NEED_PIC
  318. NEED_PIC = 1
  319. endif
  320. ARFLAGS =
  321. CPP = $(COMPILER) -E
  322. AR ?= $(CROSS_SUFFIX)ar
  323. AS ?= $(CROSS_SUFFIX)as
  324. LD ?= $(CROSS_SUFFIX)ld
  325. RANLIB ?= $(CROSS_SUFFIX)ranlib
  326. NM = $(CROSS_SUFFIX)nm
  327. DLLWRAP = $(CROSS_SUFFIX)dllwrap
  328. OBJCOPY = $(CROSS_SUFFIX)objcopy
  329. OBJCONV = $(CROSS_SUFFIX)objconv
  330. # When fortran support was either not detected or actively deselected, only build BLAS.
  331. ifeq ($(NOFORTRAN), 1)
  332. C_LAPACK = 1
  333. override FEXTRALIB =
  334. endif
  335. ifeq ($(C_COMPILER), GCC)
  336. GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
  337. GCCVERSIONGT4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 4)
  338. GCCVERSIONGT5 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 5)
  339. GCCVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 7)
  340. GCCVERSIONGTEQ8 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 8)
  341. GCCVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  342. GCCVERSIONGTEQ10 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 10)
  343. GCCVERSIONGTEQ11 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 11)
  344. GCCVERSIONGTEQ12 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 12)
  345. # Note that the behavior of -dumpversion is compile-time-configurable for
  346. # gcc-7.x and newer. Use -dumpfullversion there
  347. ifeq ($(GCCVERSIONGTEQ7),1)
  348. GCCDUMPVERSION_PARAM := -dumpfullversion
  349. else
  350. GCCDUMPVERSION_PARAM := -dumpversion
  351. endif
  352. GCCMINORVERSIONGTEQ1 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 1)
  353. GCCMINORVERSIONGTEQ2 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 2)
  354. GCCMINORVERSIONGTEQ4 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 4)
  355. GCCMINORVERSIONGTEQ7 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 7)
  356. endif
  357. ifeq ($(C_COMPILER), CLANG)
  358. CLANGVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  359. CLANGVERSIONGTEQ12 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 12)
  360. endif
  361. #
  362. # OS dependent settings
  363. #
  364. ifeq ($(OSNAME), Darwin)
  365. ifndef MACOSX_DEPLOYMENT_TARGET
  366. ifeq ($(ARCH), arm64)
  367. export MACOSX_DEPLOYMENT_TARGET=11.0
  368. ifeq ($(C_COMPILER), GCC)
  369. export NO_SVE = 1
  370. endif
  371. else
  372. export MACOSX_DEPLOYMENT_TARGET=10.8
  373. endif
  374. endif
  375. MD5SUM = md5 -r
  376. XCVER = $(shell pkgutil --pkg-info=com.apple.pkg.Xcode |awk '/version:/ {print $2}'|cut -d: -f2|cut -f1 -d.)
  377. ifeq (x$(XCVER)x,xx)
  378. XCVER = $(shell pkgutil --pkg-info=com.apple.pkg.CLTools_Executables |awk '/version:/ {print $2}'|cut -d: -f2|cut -f1 -d.)
  379. endif
  380. ifeq (x$(XCVER), x 15)
  381. CCOMMON_OPT += -Wl,-ld_classic
  382. FCOMMON_OPT += -Wl,-ld_classic
  383. endif
  384. endif
  385. ifneq (,$(findstring $(OSNAME), FreeBSD OpenBSD DragonFly))
  386. MD5SUM = md5 -r
  387. endif
  388. ifeq ($(OSNAME), NetBSD)
  389. MD5SUM = md5 -n
  390. endif
  391. ifeq ($(OSNAME), Linux)
  392. EXTRALIB += -lm
  393. #NO_EXPRECISION = 1
  394. endif
  395. ifeq ($(OSNAME), Android)
  396. EXTRALIB += -lm
  397. endif
  398. ifeq ($(OSNAME), AIX)
  399. EXTRALIB += -lm
  400. endif
  401. ifeq ($(OSNAME), $(filter $(OSNAME),FreeBSD OpenBSD NetBSD DragonFly))
  402. ifeq ($(ARCH), $(filter $(ARCH),arm arm64))
  403. EXTRALIB += -lm
  404. endif
  405. endif
  406. ifeq ($(OSNAME), WINNT)
  407. NEED_PIC = 0
  408. NO_EXPRECISION = 1
  409. EXTRALIB += -defaultlib:advapi32
  410. SUFFIX = obj
  411. PSUFFIX = pobj
  412. LIBSUFFIX = a
  413. ifeq ($(C_COMPILER), CLANG)
  414. CCOMMON_OPT += -DMS_ABI
  415. endif
  416. #Version tests for supporting specific features (MS_ABI, POWER9 intrinsics)
  417. ifeq ($(GCCVERSIONGT4), 1)
  418. # GCC Major version > 4
  419. # It is compatible with MSVC ABI.
  420. CCOMMON_OPT += -DMS_ABI
  421. endif
  422. ifeq ($(GCCVERSIONGTEQ4), 1)
  423. ifeq ($(GCCMINORVERSIONGTEQ7), 1)
  424. # GCC Version >=4.7
  425. # It is compatible with MSVC ABI.
  426. CCOMMON_OPT += -DMS_ABI
  427. endif
  428. endif
  429. # Ensure the correct stack alignment on Win32
  430. # http://permalink.gmane.org/gmane.comp.lib.openblas.general/97
  431. ifeq ($(ARCH), x86)
  432. CCOMMON_OPT += -mincoming-stack-boundary=2
  433. FCOMMON_OPT += -mincoming-stack-boundary=2
  434. endif
  435. endif
  436. ifeq ($(OSNAME), Interix)
  437. NEED_PIC = 0
  438. NO_EXPRECISION = 1
  439. INTERIX_TOOL_DIR = /opt/gcc.3.3/i586-pc-interix3/bin
  440. endif
  441. ifeq ($(OSNAME), CYGWIN_NT)
  442. NEED_PIC = 0
  443. NO_EXPRECISION = 1
  444. OS_CYGWIN_NT = 1
  445. endif
  446. ifneq ($(OSNAME), WINNT)
  447. ifneq ($(OSNAME), CYGWIN_NT)
  448. ifneq ($(OSNAME), Interix)
  449. ifneq ($(OSNAME), Android)
  450. ifdef SMP
  451. EXTRALIB += -lpthread
  452. endif
  453. endif
  454. endif
  455. endif
  456. endif
  457. # ifeq logical or
  458. ifeq ($(OSNAME), $(filter $(OSNAME),WINNT CYGWIN_NT Interix))
  459. OS_WINDOWS=1
  460. endif
  461. ifdef QUAD_PRECISION
  462. CCOMMON_OPT += -DQUAD_PRECISION
  463. NO_EXPRECISION = 1
  464. endif
  465. ifneq ($(ARCH), x86)
  466. ifneq ($(ARCH), x86_64)
  467. NO_EXPRECISION = 1
  468. endif
  469. endif
  470. ifdef UTEST_CHECK
  471. CCOMMON_OPT += -DUTEST_CHECK
  472. SANITY_CHECK = 1
  473. endif
  474. ifdef SANITY_CHECK
  475. CCOMMON_OPT += -DSANITY_CHECK -DREFNAME=$(*F)f$(BU)
  476. endif
  477. MAX_STACK_ALLOC ?= 2048
  478. ifneq ($(MAX_STACK_ALLOC), 0)
  479. CCOMMON_OPT += -DMAX_STACK_ALLOC=$(MAX_STACK_ALLOC)
  480. endif
  481. ifdef USE_LOCKING
  482. ifneq ($(USE_LOCKING), 0)
  483. CCOMMON_OPT += -DUSE_LOCKING
  484. endif
  485. endif
  486. #
  487. # Architecture dependent settings
  488. #
  489. ifeq ($(ARCH), x86)
  490. ifndef BINARY
  491. NO_BINARY_MODE = 1
  492. endif
  493. ifeq ($(CORE), generic)
  494. #NO_EXPRECISION = 1
  495. endif
  496. ifndef NO_EXPRECISION
  497. ifeq ($(F_COMPILER), GFORTRAN)
  498. # ifeq logical or. GCC or LSB
  499. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  500. EXPRECISION = 1
  501. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  502. FCOMMON_OPT += -m128bit-long-double
  503. endif
  504. ifeq ($(C_COMPILER), CLANG)
  505. EXPRECISION = 1
  506. CCOMMON_OPT += -DEXPRECISION
  507. FCOMMON_OPT += -m128bit-long-double
  508. endif
  509. endif
  510. endif
  511. endif
  512. ifeq ($(ARCH), x86_64)
  513. ifeq ($(CORE), generic)
  514. #NO_EXPRECISION = 1
  515. endif
  516. ifndef NO_EXPRECISION
  517. ifeq ($(F_COMPILER), GFORTRAN)
  518. # ifeq logical or. GCC or LSB
  519. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  520. EXPRECISION = 1
  521. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  522. FCOMMON_OPT += -m128bit-long-double
  523. endif
  524. ifeq ($(C_COMPILER), CLANG)
  525. EXPRECISION = 1
  526. CCOMMON_OPT += -DEXPRECISION
  527. FCOMMON_OPT += -m128bit-long-double
  528. endif
  529. endif
  530. endif
  531. endif
  532. ifeq ($(C_COMPILER), INTEL)
  533. CCOMMON_OPT += -wd981
  534. endif
  535. ifeq ($(USE_OPENMP), 1)
  536. #check
  537. ifeq ($(USE_THREAD), 0)
  538. $(error OpenBLAS: Cannot set both USE_OPENMP=1 and USE_THREAD=0. The USE_THREAD=0 is only for building single thread version.)
  539. endif
  540. # ifeq logical or. GCC or LSB
  541. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  542. CCOMMON_OPT += -fopenmp
  543. endif
  544. ifeq ($(C_COMPILER), CLANG)
  545. CCOMMON_OPT += -fopenmp
  546. ifeq ($(F_COMPILER), GFORTRAN)
  547. FEXTRALIB := $(subst -lgomp,-lomp,$(FEXTRALIB))
  548. endif
  549. endif
  550. ifeq ($(C_COMPILER), INTEL)
  551. CCOMMON_OPT += -fopenmp
  552. endif
  553. ifeq ($(C_COMPILER), PGI)
  554. CCOMMON_OPT += -mp
  555. endif
  556. ifeq ($(C_COMPILER), OPEN64)
  557. CCOMMON_OPT += -mp
  558. CEXTRALIB += -lstdc++
  559. endif
  560. ifeq ($(C_COMPILER), PATHSCALE)
  561. CCOMMON_OPT += -mp
  562. endif
  563. endif
  564. ifeq ($(DYNAMIC_ARCH), 1)
  565. ifeq ($(ARCH), x86)
  566. DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
  567. CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
  568. endif
  569. ifeq ($(ARCH), x86_64)
  570. DYNAMIC_CORE = PRESCOTT CORE2
  571. ifeq ($(DYNAMIC_OLDER), 1)
  572. DYNAMIC_CORE += PENRYN DUNNINGTON
  573. endif
  574. DYNAMIC_CORE += NEHALEM
  575. ifeq ($(DYNAMIC_OLDER), 1)
  576. DYNAMIC_CORE += OPTERON OPTERON_SSE3
  577. endif
  578. DYNAMIC_CORE += BARCELONA
  579. ifeq ($(DYNAMIC_OLDER), 1)
  580. DYNAMIC_CORE += BOBCAT ATOM NANO
  581. endif
  582. ifneq ($(NO_AVX), 1)
  583. DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER STEAMROLLER EXCAVATOR
  584. endif
  585. ifneq ($(NO_AVX2), 1)
  586. DYNAMIC_CORE += HASWELL ZEN
  587. endif
  588. ifneq ($(NO_AVX512), 1)
  589. ifneq ($(NO_AVX2), 1)
  590. DYNAMIC_CORE += SKYLAKEX COOPERLAKE SAPPHIRERAPIDS
  591. endif
  592. endif
  593. endif
  594. ifdef DYNAMIC_LIST
  595. override DYNAMIC_CORE = PRESCOTT $(DYNAMIC_LIST)
  596. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_PRESCOTT
  597. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  598. CCOMMON_OPT += $(XCCOMMON_OPT)
  599. #CCOMMON_OPT += -DDYNAMIC_LIST='$(DYNAMIC_LIST)'
  600. endif
  601. ifeq ($(ARCH), arm64)
  602. DYNAMIC_CORE = ARMV8
  603. DYNAMIC_CORE += CORTEXA53
  604. DYNAMIC_CORE += CORTEXA57
  605. DYNAMIC_CORE += NEOVERSEN1
  606. ifneq ($(NO_SVE), 1)
  607. DYNAMIC_CORE += NEOVERSEV1
  608. DYNAMIC_CORE += NEOVERSEN2
  609. DYNAMIC_CORE += ARMV8SVE
  610. DYNAMIC_CORE += A64FX
  611. endif
  612. DYNAMIC_CORE += THUNDERX
  613. DYNAMIC_CORE += THUNDERX2T99
  614. DYNAMIC_CORE += TSV110
  615. DYNAMIC_CORE += EMAG8180
  616. DYNAMIC_CORE += THUNDERX3T110
  617. ifdef DYNAMIC_LIST
  618. override DYNAMIC_CORE = ARMV8 $(DYNAMIC_LIST)
  619. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_ARMV8
  620. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  621. endif
  622. endif
  623. ifeq ($(ARCH), mips64)
  624. DYNAMIC_CORE = LOONGSON3R3 LOONGSON3R4 MIPS64_GENERIC
  625. ifdef DYNAMIC_LIST
  626. override DYNAMIC_CORE = MIPS64_GENERIC $(DYNAMIC_LIST)
  627. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_MIPS64_GENERIC
  628. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  629. endif
  630. endif
  631. ifeq ($(ARCH), loongarch64)
  632. DYNAMIC_CORE = LA64_GENERIC LA264 LA464
  633. endif
  634. ifeq ($(ARCH), riscv64)
  635. DYNAMIC_CORE = RISCV64_GENERIC
  636. DYNAMIC_CORE += RISCV64_ZVL128B
  637. DYNAMIC_CORE += RISCV64_ZVL256B
  638. ifdef DYNAMIC_LIST
  639. override DYNAMIC_CORE = RISCV64_GENERIC $(DYNAMIC_LIST)
  640. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_RISCV64_GENERIC
  641. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  642. endif
  643. endif
  644. ifeq ($(ARCH), zarch)
  645. DYNAMIC_CORE = ZARCH_GENERIC
  646. # if the compiler accepts -march=arch11 or -march=z13 and can compile a file
  647. # with z13-specific inline assembly, then we can include support for Z13.
  648. # note: -march=z13 is equivalent to -march=arch11 yet some compiler releases
  649. # only support one or the other.
  650. # note: LLVM version 6.x supported -march=z13 yet could not handle vector
  651. # registers in inline assembly, so the check for supporting the -march flag is
  652. # not enough.
  653. ZARCH_TEST_COMPILE=-c $(TOPDIR)/kernel/zarch/damin_z13.c -I$(TOPDIR) -o /dev/null > /dev/null 2> /dev/null
  654. ZARCH_CC_SUPPORTS_ARCH11=$(shell $(CC) -march=arch11 $(ZARCH_TEST_COMPILE) && echo 1)
  655. ZARCH_CC_SUPPORTS_Z13=$(shell $(CC) -march=z13 $(ZARCH_TEST_COMPILE) && echo 1)
  656. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH11), $(ZARCH_CC_SUPPORTS_Z13)), 1)
  657. DYNAMIC_CORE += Z13
  658. CCOMMON_OPT += -DDYN_Z13
  659. else
  660. $(info OpenBLAS: Not building Z13 kernels because the compiler $(CC) does not support it)
  661. endif
  662. # as above for z13, check for -march=arch12 and z14 support in the compiler.
  663. ZARCH_CC_SUPPORTS_ARCH12=$(shell $(CC) -march=arch12 $(ZARCH_TEST_COMPILE) && echo 1)
  664. ZARCH_CC_SUPPORTS_Z14=$(shell $(CC) -march=z14 $(ZARCH_TEST_COMPILE) && echo 1)
  665. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH12), $(ZARCH_CC_SUPPORTS_Z14)), 1)
  666. DYNAMIC_CORE += Z14
  667. CCOMMON_OPT += -DDYN_Z14
  668. else
  669. $(info OpenBLAS: Not building Z14 kernels because the compiler $(CC) does not support it)
  670. endif
  671. endif # ARCH zarch
  672. ifeq ($(ARCH), power)
  673. ifneq ($(C_COMPILER), PGI)
  674. DYNAMIC_CORE = POWER6
  675. DYNAMIC_CORE += POWER8
  676. ifneq ($(C_COMPILER), GCC)
  677. DYNAMIC_CORE += POWER9
  678. DYNAMIC_CORE += POWER10
  679. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  680. endif
  681. ifeq ($(C_COMPILER), GCC)
  682. ifeq ($(GCCVERSIONGT5), 1)
  683. DYNAMIC_CORE += POWER9
  684. else
  685. $(info, OpenBLAS: Your gcc version is too old to build the POWER9 kernels.)
  686. endif
  687. ifeq ($(OSNAME), AIX)
  688. LDVERSIONGTEQ35 := 1
  689. else
  690. LDVERSIONGTEQ35 := $(shell expr `$(CC) -Wl,--version 2> /dev/null | head -1 | cut -f2 -d "." | cut -f1 -d "-"` \>= 35)
  691. endif
  692. ifeq ($(GCCVERSIONGTEQ11)$(LDVERSIONGTEQ35), 11)
  693. DYNAMIC_CORE += POWER10
  694. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  695. else ifeq ($(GCCVERSIONGTEQ10), 1)
  696. ifeq ($(GCCMINORVERSIONGTEQ2)$(LDVERSIONGTEQ35), 11)
  697. DYNAMIC_CORE += POWER10
  698. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  699. endif
  700. else
  701. $(info, OpenBLAS: Your gcc version is too old to build the POWER10 kernels.)
  702. endif
  703. endif
  704. else
  705. DYNAMIC_CORE = POWER8
  706. DYNAMIC_CORE += POWER9
  707. endif
  708. endif
  709. # If DYNAMIC_CORE is not set, DYNAMIC_ARCH cannot do anything, so force it to empty
  710. ifndef DYNAMIC_CORE
  711. override DYNAMIC_ARCH=
  712. endif
  713. endif
  714. ifeq ($(ARCH), ia64)
  715. NO_BINARY_MODE = 1
  716. BINARY_DEFINED = 1
  717. ifeq ($(F_COMPILER), GFORTRAN)
  718. ifeq ($(C_COMPILER), GCC)
  719. EXPRECISION = 1
  720. CCOMMON_OPT += -DEXPRECISION
  721. endif
  722. endif
  723. endif
  724. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  725. NO_BINARY_MODE = 1
  726. endif
  727. ifeq ($(ARCH), alpha)
  728. NO_BINARY_MODE = 1
  729. BINARY_DEFINED = 1
  730. endif
  731. ifeq ($(ARCH), arm)
  732. NO_BINARY_MODE = 1
  733. BINARY_DEFINED = 1
  734. ifneq ($(EMBEDDED), 1)
  735. CCOMMON_OPT += -marm
  736. FCOMMON_OPT += -marm
  737. else
  738. CCOMMON_OPT += -DOS_EMBEDDED -mthumb -mcpu=cortex-m4 -mfloat-abi=hard -mfpu=fpv4-sp-d16
  739. endif
  740. # If softfp abi is mentioned on the command line, force it.
  741. ifeq ($(ARM_SOFTFP_ABI), 1)
  742. CCOMMON_OPT += -mfloat-abi=softfp
  743. FCOMMON_OPT += -mfloat-abi=softfp
  744. endif
  745. ifeq ($(OSNAME), Android)
  746. ifeq ($(ARM_SOFTFP_ABI), 1)
  747. EXTRALIB += -lm
  748. else
  749. EXTRALIB += -Wl,-lm_hard
  750. endif
  751. endif
  752. endif
  753. ifeq ($(ARCH), arm64)
  754. NO_BINARY_MODE = 1
  755. BINARY_DEFINED = 1
  756. ifdef INTERFACE64
  757. ifneq ($(INTERFACE64), 0)
  758. ifeq ($(F_COMPILER), GFORTRAN)
  759. FCOMMON_OPT += -fdefault-integer-8
  760. endif
  761. ifeq ($(F_COMPILER), FLANG)
  762. FCOMMON_OPT += -i8
  763. endif
  764. endif
  765. endif
  766. endif
  767. ifeq ($(ARCH), riscv64)
  768. NO_BINARY_MODE = 1
  769. BINARY_DEFINED = 1
  770. ifdef INTERFACE64
  771. ifneq ($(INTERFACE64), 0)
  772. ifeq ($(F_COMPILER), GFORTRAN)
  773. FCOMMON_OPT += -fdefault-integer-8
  774. endif
  775. ifeq ($(F_COMPILER), FLANG)
  776. FCOMMON_OPT += -i8
  777. endif
  778. endif
  779. endif
  780. endif
  781. ifeq ($(ARCH), loongarch64)
  782. NO_BINARY_MODE = 1
  783. BINARY_DEFINED = 1
  784. ifdef INTERFACE64
  785. ifneq ($(INTERFACE64), 0)
  786. ifeq ($(F_COMPILER), GFORTRAN)
  787. FCOMMON_OPT += -fdefault-integer-8
  788. endif
  789. ifeq ($(F_COMPILER), FLANG)
  790. FCOMMON_OPT += -i8
  791. endif
  792. endif
  793. endif
  794. endif
  795. ifeq ($(ARCH), csky)
  796. NO_BINARY_MODE = 1
  797. BINARY_DEFINED = 1
  798. endif
  799. #
  800. # C Compiler dependent settings
  801. #
  802. # ifeq logical or. GCC or CLANG or LSB
  803. # http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
  804. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG LSB))
  805. CCOMMON_OPT += -Wall
  806. COMMON_PROF += -fno-inline
  807. NO_UNINITIALIZED_WARN = -Wno-uninitialized
  808. ifeq ($(QUIET_MAKE), 1)
  809. CCOMMON_OPT += $(NO_UNINITIALIZED_WARN) -Wno-unused
  810. endif
  811. ifdef NO_BINARY_MODE
  812. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  813. ifdef BINARY64
  814. CCOMMON_OPT += -mabi=64
  815. else
  816. CCOMMON_OPT += -mabi=n32
  817. endif
  818. BINARY_DEFINED = 1
  819. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  820. CCOMMON_OPT += -mabi=32
  821. BINARY_DEFINED = 1
  822. endif
  823. ifneq (, $(filter $(CORE), MIPS64_GENERIC))
  824. CCOMMON_OPT += -DNO_MSA
  825. FCOMMON_OPT += -DNO_MSA
  826. endif
  827. ifneq (, $(filter $(CORE),LOONGSON3R3 LOONGSON3R4))
  828. CCOMMON_OPT += -march=loongson3a
  829. FCOMMON_OPT += -march=loongson3a
  830. endif
  831. ifeq ($(CORE), MIPS24K)
  832. CCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  833. FCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  834. endif
  835. ifeq ($(CORE), MIPS1004K)
  836. CCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  837. FCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  838. endif
  839. ifeq ($(CORE), P5600)
  840. CCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  841. FCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  842. endif
  843. ifeq ($(CORE), I6400)
  844. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  845. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  846. endif
  847. ifeq ($(CORE), P6600)
  848. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  849. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  850. endif
  851. ifeq ($(CORE), I6500)
  852. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  853. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  854. endif
  855. ifeq ($(OSNAME), AIX)
  856. BINARY_DEFINED = 1
  857. endif
  858. ifeq ($(ARCH), loongarch64)
  859. LA64_ABI=$(shell $(CC) -mabi=lp64d -c $(TOPDIR)/cpuid_loongarch64.c -o /dev/null > /dev/null 2> /dev/null && echo lp64d)
  860. LA64_ARCH=$(shell $(CC) -march=loongarch64 -c $(TOPDIR)/cpuid_loongarch64.c -o /dev/null > /dev/null 2> /dev/null && echo loongarch64)
  861. ifneq ($(LA64_ABI), lp64d)
  862. LA64_ABI=lp64
  863. endif
  864. ifneq ($(LA64_ARCH), loongarch64)
  865. CCOMMON_OPT += -mabi=$(LA64_ABI)
  866. FCOMMON_OPT += -mabi=$(LA64_ABI)
  867. else
  868. CCOMMON_OPT += -march=loongarch64 -mabi=$(LA64_ABI)
  869. FCOMMON_OPT += -march=loongarch64 -mabi=$(LA64_ABI)
  870. endif
  871. endif
  872. endif
  873. ifndef BINARY_DEFINED
  874. ifneq ($(OSNAME), AIX)
  875. ifdef BINARY64
  876. ifneq ($(ARCH), riscv64)
  877. CCOMMON_OPT += -m64
  878. endif
  879. else
  880. CCOMMON_OPT += -m32
  881. endif
  882. endif
  883. endif
  884. endif
  885. ifeq ($(C_COMPILER), PGI)
  886. PGCVERSIONGT20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \> 20)
  887. PGCVERSIONEQ20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` == 20)
  888. PGCMINORVERSIONGE11 := $(shell expr `$(CC) --version|sed -n "2p" |cut -d "-" -f 1 |sed -e "s/[^0-9.]//g" |cut -c 4-5` \>= 11)
  889. PGCVERSIONCHECK := $(PGCVERSIONGT20)$(PGCVERSIONEQ20)$(PGCMINORVERSIONGE11)
  890. ifeq ($(PGCVERSIONCHECK), $(filter $(PGCVERSIONCHECK), 100 101 011))
  891. NEWPGI := 1
  892. PGCVERSIONGT21 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \> 21)
  893. PGCVERSIONEQ21 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` == 21)
  894. PGCVERSIONCHECK2 := $(PGCVERSIONGT21)$(PGCVERSIONEQ21)$(PGCMINORVERSIONGE11)
  895. ifeq ($(PGCVERSIONCHECK2), $(filter $(PGCVERSIONCHECK2), 100 101 011))
  896. NEWPGI2 := 1
  897. endif
  898. endif
  899. ifdef BINARY64
  900. ifeq ($(ARCH), x86_64)
  901. ifeq (,$(findstring tp,$(CFLAGS)))
  902. ifneq ($(NEWPGI2),1)
  903. CCOMMON_OPT += -tp p7-64
  904. else
  905. CCOMMON_OPT += -tp px
  906. endif
  907. endif
  908. ifneq ($(NEWPGI),1)
  909. CCOMMON_OPT += -D__MMX__ -Mnollvm
  910. endif
  911. else
  912. ifeq ($(ARCH), power)
  913. ifeq (,$(findstring tp,$(CFLAGS)))
  914. ifeq ($(CORE), POWER8)
  915. CCOMMON_OPT += -tp pwr8
  916. endif
  917. ifeq ($(CORE), POWER9)
  918. CCOMMON_OPT += -tp pwr9
  919. endif
  920. endif
  921. endif
  922. endif
  923. else
  924. ifneq ($(NEWPGI2),1)
  925. ifeq (,$(findstring tp,$(CFLAGS)))
  926. CCOMMON_OPT += -tp p7
  927. else
  928. CCOMMON_OPT += -tp px
  929. endif
  930. endif
  931. endif
  932. endif
  933. ifeq ($(C_COMPILER), PATHSCALE)
  934. ifdef BINARY64
  935. CCOMMON_OPT += -m64
  936. else
  937. CCOMMON_OPT += -m32
  938. endif
  939. endif
  940. #
  941. # Fortran Compiler dependent settings
  942. #
  943. ifeq ($(F_COMPILER), NAG)
  944. FCOMMON_OPT += -dcfuns -recursive -ieee=full -w=obs -thread_safe
  945. ifdef INTERFACE64
  946. ifneq ($(INTERFACE64), 0)
  947. FCOMMON_OPT += -i8
  948. endif
  949. endif
  950. ifeq ($(USE_OPENMP), 1)
  951. FCOMMON_OPT += -openmp
  952. endif
  953. endif
  954. ifeq ($(F_COMPILER), FLANG)
  955. CCOMMON_OPT += -DF_INTERFACE_FLANG
  956. FCOMMON_OPT += -Mrecursive -Kieee
  957. ifeq ($(OSNAME), Linux)
  958. ifeq ($(ARCH), x86_64)
  959. FLANG_VENDOR := $(shell $(FC) --version|head -1 |cut -f 1 -d " ")
  960. ifeq ($(FLANG_VENDOR), AMD)
  961. FCOMMON_OPT += -fno-unroll-loops
  962. endif
  963. endif
  964. endif
  965. ifdef BINARY64
  966. ifdef INTERFACE64
  967. ifneq ($(INTERFACE64), 0)
  968. FCOMMON_OPT += -i8
  969. endif
  970. endif
  971. FCOMMON_OPT += -Wall
  972. else
  973. FCOMMON_OPT += -Wall
  974. endif
  975. ifeq ($(USE_OPENMP), 1)
  976. FCOMMON_OPT += -fopenmp
  977. endif
  978. endif
  979. ifeq ($(F_COMPILER), G77)
  980. CCOMMON_OPT += -DF_INTERFACE_G77
  981. FCOMMON_OPT += -Wall
  982. ifndef NO_BINARY_MODE
  983. ifneq ($(OSNAME), AIX)
  984. ifdef BINARY64
  985. FCOMMON_OPT += -m64
  986. else
  987. FCOMMON_OPT += -m32
  988. endif
  989. endif
  990. endif
  991. endif
  992. ifeq ($(F_COMPILER), G95)
  993. CCOMMON_OPT += -DF_INTERFACE_G95
  994. FCOMMON_OPT += -Wall
  995. ifneq ($(OSNAME), AIX)
  996. ifndef NO_BINARY_MODE
  997. ifdef BINARY64
  998. FCOMMON_OPT += -m64
  999. else
  1000. FCOMMON_OPT += -m32
  1001. endif
  1002. endif
  1003. ifneq ($(NO_LAPACKE), 1)
  1004. FCOMMON_OPT += -fno-second-underscore
  1005. endif
  1006. endif
  1007. endif
  1008. ifeq ($(F_COMPILER), $(filter $(F_COMPILER),GFORTRAN FLANGNEW))
  1009. CCOMMON_OPT += -DF_INTERFACE_GFORT
  1010. ifeq ($(F_COMPILER), GFORTRAN)
  1011. FCOMMON_OPT += -Wall
  1012. # make single-threaded LAPACK calls thread-safe #1847
  1013. FCOMMON_OPT += -frecursive
  1014. # work around ABI problem with passing single-character arguments
  1015. FCOMMON_OPT += -fno-optimize-sibling-calls
  1016. #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
  1017. ifneq ($(NOFORTRAN), 1)
  1018. ifneq ($(NOFORTRAN), 2)
  1019. ifneq ($(NO_LAPACK), 1)
  1020. EXTRALIB += -lgfortran
  1021. endif
  1022. endif
  1023. endif
  1024. endif
  1025. ifdef NO_BINARY_MODE
  1026. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  1027. ifdef BINARY64
  1028. FCOMMON_OPT += -mabi=64
  1029. else
  1030. FCOMMON_OPT += -mabi=n32
  1031. endif
  1032. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  1033. FCOMMON_OPT += -mabi=32
  1034. endif
  1035. else
  1036. ifdef BINARY64
  1037. ifneq ($(OSNAME), AIX)
  1038. ifneq ($(ARCH), riscv64)
  1039. FCOMMON_OPT += -m64
  1040. endif
  1041. endif
  1042. ifdef INTERFACE64
  1043. ifneq ($(INTERFACE64), 0)
  1044. FCOMMON_OPT += -fdefault-integer-8
  1045. endif
  1046. endif
  1047. else
  1048. ifneq ($(OSNAME), AIX)
  1049. FCOMMON_OPT += -m32
  1050. endif
  1051. endif
  1052. endif
  1053. ifeq ($(USE_OPENMP), 1)
  1054. FCOMMON_OPT += -fopenmp
  1055. endif
  1056. endif
  1057. ifeq ($(F_COMPILER), INTEL)
  1058. CCOMMON_OPT += -DF_INTERFACE_INTEL
  1059. ifdef INTERFACE64
  1060. ifneq ($(INTERFACE64), 0)
  1061. FCOMMON_OPT += -i8
  1062. endif
  1063. endif
  1064. FCOMMON_OPT += -recursive -fp-model strict -assume protect-parens
  1065. ifeq ($(USE_OPENMP), 1)
  1066. FCOMMON_OPT += -fopenmp
  1067. endif
  1068. endif
  1069. ifeq ($(F_COMPILER), FUJITSU)
  1070. CCOMMON_OPT += -DF_INTERFACE_FUJITSU
  1071. ifeq ($(USE_OPENMP), 1)
  1072. FCOMMON_OPT += -openmp
  1073. endif
  1074. endif
  1075. ifeq ($(F_COMPILER), IBM)
  1076. CCOMMON_OPT += -DF_INTERFACE_IBM
  1077. FEXTRALIB += -lxlf90
  1078. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG))
  1079. FCOMMON_OPT += -qextname -qzerosize
  1080. endif
  1081. # FCOMMON_OPT += -qarch=440
  1082. ifdef BINARY64
  1083. FCOMMON_OPT += -q64
  1084. ifdef INTERFACE64
  1085. ifneq ($(INTERFACE64), 0)
  1086. FCOMMON_OPT += -qintsize=8
  1087. endif
  1088. endif
  1089. else
  1090. FCOMMON_OPT += -q32
  1091. endif
  1092. endif
  1093. ifeq ($(F_COMPILER), PGI)
  1094. CCOMMON_OPT += -DF_INTERFACE_PGI
  1095. COMMON_PROF += -DPGICOMPILER
  1096. ifdef BINARY64
  1097. ifdef INTERFACE64
  1098. ifneq ($(INTERFACE64), 0)
  1099. FCOMMON_OPT += -i8
  1100. endif
  1101. endif
  1102. ifeq ($(ARCH), x86_64)
  1103. ifneq ($(NEWPGI2),1)
  1104. FCOMMON_OPT += -tp p7-64
  1105. else
  1106. FCOMMON_OPT += -tp px
  1107. endif
  1108. else
  1109. ifeq ($(ARCH), power)
  1110. ifeq ($(CORE), POWER6)
  1111. $(warning NVIDIA HPC compilers do not support POWER6.)
  1112. endif
  1113. ifeq ($(CORE), POWER8)
  1114. FCOMMON_OPT += -tp pwr8
  1115. endif
  1116. ifeq ($(CORE), POWER9)
  1117. FCOMMON_OPT += -tp pwr9
  1118. endif
  1119. ifeq ($(CORE), POWER10)
  1120. $(warning NVIDIA HPC compilers do not support POWER10.)
  1121. endif
  1122. endif
  1123. endif
  1124. else
  1125. FCOMMON_OPT += -tp p7
  1126. endif
  1127. FCOMMON_OPT += -Mrecursive -Kieee
  1128. ifeq ($(USE_OPENMP), 1)
  1129. FCOMMON_OPT += -mp
  1130. endif
  1131. endif
  1132. ifeq ($(F_COMPILER), PATHSCALE)
  1133. CCOMMON_OPT += -DF_INTERFACE_PATHSCALE
  1134. ifdef BINARY64
  1135. ifdef INTERFACE64
  1136. ifneq ($(INTERFACE64), 0)
  1137. FCOMMON_OPT += -i8
  1138. endif
  1139. endif
  1140. endif
  1141. ifeq ($(USE_OPENMP), 1)
  1142. FCOMMON_OPT += -mp
  1143. endif
  1144. endif
  1145. ifeq ($(F_COMPILER), OPEN64)
  1146. CCOMMON_OPT += -DF_INTERFACE_OPEN64
  1147. ifdef BINARY64
  1148. ifdef INTERFACE64
  1149. ifneq ($(INTERFACE64), 0)
  1150. FCOMMON_OPT += -i8
  1151. endif
  1152. endif
  1153. endif
  1154. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  1155. ifndef BINARY64
  1156. FCOMMON_OPT += -n32
  1157. else
  1158. FCOMMON_OPT += -n64
  1159. endif
  1160. ifeq ($(CORE), LOONGSON3R3)
  1161. FCOMMON_OPT += -loongson3 -static
  1162. endif
  1163. ifeq ($(CORE), LOONGSON3R4)
  1164. FCOMMON_OPT += -loongson3 -static
  1165. endif
  1166. else
  1167. ifndef BINARY64
  1168. FCOMMON_OPT += -m32
  1169. else
  1170. FCOMMON_OPT += -m64
  1171. endif
  1172. endif
  1173. ifeq ($(USE_OPENMP), 1)
  1174. FEXTRALIB += -lstdc++
  1175. FCOMMON_OPT += -mp
  1176. endif
  1177. endif
  1178. ifeq ($(C_COMPILER), OPEN64)
  1179. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  1180. ifndef BINARY64
  1181. CCOMMON_OPT += -n32
  1182. else
  1183. CCOMMON_OPT += -n64
  1184. endif
  1185. ifeq ($(CORE), LOONGSON3R3)
  1186. CCOMMON_OPT += -loongson3 -static
  1187. endif
  1188. ifeq ($(CORE), LOONGSON3R4)
  1189. CCOMMON_OPT += -loongson3 -static
  1190. endif
  1191. else
  1192. ifndef BINARY64
  1193. CCOMMON_OPT += -m32
  1194. else
  1195. CCOMMON_OPT += -m64
  1196. endif
  1197. endif
  1198. endif
  1199. ifeq ($(C_COMPILER), SUN)
  1200. CCOMMON_OPT += -w
  1201. ifeq ($(ARCH), x86)
  1202. CCOMMON_OPT += -m32
  1203. else
  1204. ifdef BINARY64
  1205. CCOMMON_OPT += -m64
  1206. else
  1207. CCOMMON_OPT += -m32
  1208. endif
  1209. endif
  1210. endif
  1211. ifeq ($(F_COMPILER), SUN)
  1212. CCOMMON_OPT += -DF_INTERFACE_SUN
  1213. FCOMMON_OPT += -ftrap=%none -xrecursive
  1214. ifeq ($(ARCH), x86)
  1215. FCOMMON_OPT += -m32
  1216. else
  1217. ifdef BINARY64
  1218. FCOMMON_OPT += -m64
  1219. else
  1220. FCOMMON_OPT += -m32
  1221. endif
  1222. endif
  1223. ifeq ($(USE_OPENMP), 1)
  1224. FCOMMON_OPT += -xopenmp=parallel
  1225. endif
  1226. endif
  1227. ifeq ($(F_COMPILER), COMPAQ)
  1228. CCOMMON_OPT += -DF_INTERFACE_COMPAQ
  1229. ifeq ($(USE_OPENMP), 1)
  1230. FCOMMON_OPT += -openmp
  1231. endif
  1232. endif
  1233. ifeq ($(F_COMPILER), CRAY)
  1234. CCOMMON_OPT += -DF_INTERFACE_INTEL
  1235. FCOMMON_OPT += -hnopattern
  1236. ifdef INTERFACE64
  1237. ifneq ($(INTERFACE64), 0)
  1238. FCOMMON_OPT += -s integer64
  1239. endif
  1240. endif
  1241. ifneq ($(USE_OPENMP), 1)
  1242. FCOMMON_OPT += -O noomp
  1243. endif
  1244. endif
  1245. ifdef BINARY64
  1246. ifdef INTERFACE64
  1247. ifneq ($(INTERFACE64), 0)
  1248. CCOMMON_OPT +=
  1249. #-DUSE64BITINT
  1250. endif
  1251. endif
  1252. endif
  1253. ifeq ($(NEED_PIC), 1)
  1254. ifeq ($(C_COMPILER), IBM)
  1255. CCOMMON_OPT += -qpic=large
  1256. else
  1257. CCOMMON_OPT += -fPIC
  1258. endif
  1259. ifeq ($(F_COMPILER), SUN)
  1260. FCOMMON_OPT += -pic
  1261. else ifeq ($(F_COMPILER), NAG)
  1262. FCOMMON_OPT += -PIC
  1263. else ifeq ($(F_COMPILER), IBM)
  1264. FCOMMON_OPT += -qpic=large
  1265. else
  1266. FCOMMON_OPT += -fPIC
  1267. endif
  1268. endif
  1269. ifeq ($(DYNAMIC_ARCH), 1)
  1270. CCOMMON_OPT += -DDYNAMIC_ARCH
  1271. endif
  1272. ifeq ($(DYNAMIC_OLDER), 1)
  1273. CCOMMON_OPT += -DDYNAMIC_OLDER
  1274. endif
  1275. ifeq ($(C_LAPACK), 1)
  1276. CCOMMON_OPT += -DC_LAPACK
  1277. endif
  1278. ifeq ($(NO_LAPACK), 1)
  1279. CCOMMON_OPT += -DNO_LAPACK
  1280. #Disable LAPACK C interface
  1281. NO_LAPACKE = 1
  1282. endif
  1283. ifeq ($(NO_LAPACKE), 1)
  1284. CCOMMON_OPT += -DNO_LAPACKE
  1285. endif
  1286. ifeq ($(NO_AVX), 1)
  1287. CCOMMON_OPT += -DNO_AVX
  1288. endif
  1289. ifeq ($(ARCH), x86)
  1290. CCOMMON_OPT += -DNO_AVX
  1291. endif
  1292. ifeq ($(NO_AVX2), 1)
  1293. CCOMMON_OPT += -DNO_AVX2
  1294. endif
  1295. ifeq ($(NO_AVX512), 1)
  1296. CCOMMON_OPT += -DNO_AVX512
  1297. endif
  1298. ifeq ($(NO_SVE), 1)
  1299. CCOMMON_OPT += -DNO_SVE
  1300. endif
  1301. ifdef SMP
  1302. CCOMMON_OPT += -DSMP_SERVER
  1303. ifeq ($(ARCH), mips64)
  1304. USE_SIMPLE_THREADED_LEVEL3 = 1
  1305. endif
  1306. ifeq ($(USE_OPENMP), 1)
  1307. # USE_SIMPLE_THREADED_LEVEL3 = 1
  1308. # NO_AFFINITY = 1
  1309. CCOMMON_OPT += -DUSE_OPENMP
  1310. endif
  1311. ifeq ($(BIGNUMA), 1)
  1312. CCOMMON_OPT += -DBIGNUMA
  1313. endif
  1314. endif
  1315. ifeq ($(NO_WARMUP), 1)
  1316. CCOMMON_OPT += -DNO_WARMUP
  1317. endif
  1318. ifeq ($(CONSISTENT_FPCSR), 1)
  1319. CCOMMON_OPT += -DCONSISTENT_FPCSR
  1320. endif
  1321. # Only for development
  1322. # CCOMMON_OPT += -DPARAMTEST
  1323. # CCOMMON_OPT += -DPREFETCHTEST
  1324. # CCOMMON_OPT += -DNO_SWITCHING
  1325. # USE_PAPI = 1
  1326. ifdef USE_PAPI
  1327. CCOMMON_OPT += -DUSE_PAPI
  1328. EXTRALIB += -lpapi -lperfctr
  1329. endif
  1330. ifdef BUFFERSIZE
  1331. CCOMMON_OPT += -DBUFFERSIZE=$(BUFFERSIZE)
  1332. endif
  1333. ifdef DYNAMIC_THREADS
  1334. CCOMMON_OPT += -DDYNAMIC_THREADS
  1335. endif
  1336. CCOMMON_OPT += -DMAX_CPU_NUMBER=$(NUM_THREADS)
  1337. CCOMMON_OPT += -DMAX_PARALLEL_NUMBER=$(NUM_PARALLEL)
  1338. ifdef USE_SIMPLE_THREADED_LEVEL3
  1339. CCOMMON_OPT += -DUSE_SIMPLE_THREADED_LEVEL3
  1340. endif
  1341. ifeq ($(USE_TLS), 1)
  1342. CCOMMON_OPT += -DUSE_TLS
  1343. endif
  1344. ifeq ($(BUILD_BFLOAT16), 1)
  1345. CCOMMON_OPT += -DBUILD_BFLOAT16
  1346. endif
  1347. ifeq ($(BUILD_SINGLE), 1)
  1348. CCOMMON_OPT += -DBUILD_SINGLE=1
  1349. endif
  1350. ifeq ($(BUILD_DOUBLE), 1)
  1351. CCOMMON_OPT += -DBUILD_DOUBLE=1
  1352. endif
  1353. ifeq ($(BUILD_COMPLEX), 1)
  1354. CCOMMON_OPT += -DBUILD_COMPLEX=1
  1355. endif
  1356. ifeq ($(BUILD_COMPLEX16), 1)
  1357. CCOMMON_OPT += -DBUILD_COMPLEX16=1
  1358. endif
  1359. CCOMMON_OPT += -DVERSION=\"$(VERSION)\"
  1360. ifndef SYMBOLPREFIX
  1361. SYMBOLPREFIX =
  1362. endif
  1363. ifndef SYMBOLSUFFIX
  1364. SYMBOLSUFFIX =
  1365. endif
  1366. ifndef LIBSONAMEBASE
  1367. LIBSONAMEBASE = openblas
  1368. endif
  1369. ifndef LIBNAMEPREFIX
  1370. LIBNAMEPREFIX =
  1371. endif
  1372. SYMPREFIX=$(SYMBOLPREFIX)
  1373. ifeq ($(SYMBOLPREFIX),$(LIBNAMEPREFIX))
  1374. SYMPREFIX=
  1375. endif
  1376. SYMSUFFIX=$(SYMBOLSUFFIX)
  1377. ifeq ($(SYMBOLSUFFIX),$(LIBNAMESUFFIX))
  1378. SYMSUFFIX=
  1379. endif
  1380. ifndef LIBNAMESUFFIX
  1381. LIBNAMEBASE = $(SYMPREFIX)$(LIBSONAMEBASE)$(SYMSUFFIX)
  1382. else
  1383. LIBNAMEBASE = $(SYMPREFIX)$(LIBSONAMEBASE)$(SYMSUFFIX)$(LIBNAMESUFFIX)
  1384. endif
  1385. ifeq ($(OSNAME), CYGWIN_NT)
  1386. LIBPREFIX = cyg$(LIBNAMEPREFIX)$(LIBNAMEBASE)
  1387. else
  1388. LIBPREFIX = lib$(LIBNAMEPREFIX)$(LIBNAMEBASE)
  1389. endif
  1390. KERNELDIR = $(TOPDIR)/kernel/$(ARCH)
  1391. include $(TOPDIR)/Makefile.$(ARCH)
  1392. ifneq ($(C_COMPILER), PGI)
  1393. ifneq ($(C_COMPILER), SUN)
  1394. CCOMMON_OPT += -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME
  1395. endif
  1396. endif
  1397. CCOMMON_OPT += -DASMNAME=$(FU)$(*F) -DASMFNAME=$(FU)$(*F)$(BU) -DNAME=$(*F)$(BU) -DCNAME=$(*F) -DCHAR_NAME=\"$(*F)$(BU)\" -DCHAR_CNAME=\"$(*F)\"
  1398. ifeq ($(CORE), PPC440)
  1399. CCOMMON_OPT += -DALLOC_QALLOC
  1400. endif
  1401. ifeq ($(CORE), PPC440FP2)
  1402. STATIC_ALLOCATION = 1
  1403. endif
  1404. ifneq ($(OSNAME), Linux)
  1405. NO_AFFINITY = 1
  1406. endif
  1407. ifneq ($(ARCH), x86_64)
  1408. ifneq ($(ARCH), x86)
  1409. NO_AFFINITY = 1
  1410. endif
  1411. endif
  1412. ifeq ($(ARCH), POWER)
  1413. ifeq ($(DEBUG), 1)
  1414. CCOMMON_OPT := $(filter-out -O%, $(CCOMMON_OPT)) -O0
  1415. FCOMMON_OPT := $(filter-out -O%, $(FCOMMON_OPT)) -O0
  1416. endif
  1417. endif
  1418. ifdef NO_AFFINITY
  1419. ifeq ($(NO_AFFINITY), 0)
  1420. override undefine NO_AFFINITY
  1421. else
  1422. CCOMMON_OPT += -DNO_AFFINITY
  1423. endif
  1424. endif
  1425. ifdef FUNCTION_PROFILE
  1426. CCOMMON_OPT += -DFUNCTION_PROFILE
  1427. endif
  1428. ifdef SHMEM_ALLOCATION
  1429. ifneq ($(SHMEM_ALLOCATION), 0)
  1430. CCOMMON_OPT += -DALLOC_SHM
  1431. endif
  1432. endif
  1433. ifdef HUGETLB_ALLOCATION
  1434. ifneq ($(HUGETLB_ALLOCATION), 0)
  1435. CCOMMON_OPT += -DALLOC_HUGETLB
  1436. endif
  1437. endif
  1438. ifdef HUGETLBFILE_ALLOCATION
  1439. ifneq ($(HUGETLBFILE_ALLOCATION), 0)
  1440. CCOMMON_OPT += -DALLOC_HUGETLBFILE -DHUGETLB_FILE_NAME=$(HUGETLBFILE_ALLOCATION)
  1441. endif
  1442. endif
  1443. ifdef STATIC_ALLOCATION
  1444. CCOMMON_OPT += -DALLOC_STATIC
  1445. endif
  1446. ifdef DEVICEDRIVER_ALLOCATION
  1447. CCOMMON_OPT += -DALLOC_DEVICEDRIVER -DDEVICEDRIVER_NAME=\"/dev/mapper\"
  1448. endif
  1449. ifdef MIXED_MEMORY_ALLOCATION
  1450. CCOMMON_OPT += -DMIXED_MEMORY_ALLOCATION
  1451. endif
  1452. ifeq ($(OSNAME), SunOS)
  1453. TAR = gtar
  1454. PATCH = gpatch
  1455. GREP = ggrep
  1456. AWK = nawk
  1457. else
  1458. TAR = tar
  1459. PATCH = patch
  1460. GREP = grep
  1461. AWK = awk
  1462. endif
  1463. ifndef MD5SUM
  1464. MD5SUM = md5sum
  1465. endif
  1466. REVISION = -r$(VERSION)
  1467. MAJOR_VERSION = $(word 1,$(subst ., ,$(VERSION)))
  1468. ifeq ($(DEBUG), 1)
  1469. COMMON_OPT += -g
  1470. endif
  1471. ifeq ($(DEBUG), 1)
  1472. FCOMMON_OPT += -g
  1473. endif
  1474. ifndef COMMON_OPT
  1475. COMMON_OPT = -O2
  1476. endif
  1477. ifndef FCOMMON_OPT
  1478. FCOMMON_OPT = -O2 -frecursive
  1479. endif
  1480. override CFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
  1481. override PFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
  1482. override FFLAGS += $(COMMON_OPT) $(FCOMMON_OPT)
  1483. override FPFLAGS += $(FCOMMON_OPT) $(COMMON_PROF)
  1484. #MAKEOVERRIDES =
  1485. ifeq ($(NEED_PIC), 1)
  1486. ifeq (,$(findstring PIC,$(FFLAGS)))
  1487. ifneq ($(F_COMPILER),IBM)
  1488. override FFLAGS += -fPIC
  1489. endif
  1490. endif
  1491. endif
  1492. #For LAPACK Fortran codes.
  1493. #Disable -fopenmp for LAPACK Fortran codes on Windows.
  1494. ifdef OS_WINDOWS
  1495. LAPACK_FFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FFLAGS))
  1496. LAPACK_FPFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FPFLAGS))
  1497. else
  1498. LAPACK_FFLAGS := $(FFLAGS)
  1499. LAPACK_FPFLAGS := $(FPFLAGS)
  1500. endif
  1501. ifeq ($(F_COMPILER),NAG)
  1502. LAPACK_FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1503. override FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1504. endif
  1505. ifeq ($(F_COMPILER),CRAY)
  1506. LAPACK_FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1507. override FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1508. endif
  1509. ifeq ($(F_COMPILER),FLANGNEW)
  1510. LAPACK_FFLAGS := $(filter-out -m32 -m64 -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 -mtune=% -mabi=% ,$(FFLAGS))
  1511. override FFLAGS := $(filter-out -m32 -m64 -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 -mtune=% -mabi=% ,$(FFLAGS))
  1512. endif
  1513. LAPACK_CFLAGS = $(CFLAGS)
  1514. LAPACK_CFLAGS += -DHAVE_LAPACK_CONFIG_H
  1515. ifdef INTERFACE64
  1516. ifneq ($(INTERFACE64), 0)
  1517. LAPACK_CFLAGS += -DLAPACK_ILP64
  1518. endif
  1519. endif
  1520. ifdef OS_WINDOWS
  1521. LAPACK_CFLAGS += -DOPENBLAS_OS_WINDOWS
  1522. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1523. endif
  1524. ifeq ($(C_COMPILER), LSB)
  1525. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1526. endif
  1527. ifndef SUFFIX
  1528. SUFFIX = o
  1529. endif
  1530. ifndef PSUFFIX
  1531. PSUFFIX = po
  1532. endif
  1533. ifndef LIBSUFFIX
  1534. LIBSUFFIX = a
  1535. endif
  1536. ifneq ($(DYNAMIC_ARCH), 1)
  1537. ifndef SMP
  1538. LIBNAME = $(LIBPREFIX)_$(LIBCORE)$(REVISION).$(LIBSUFFIX)
  1539. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)$(REVISION)_p.$(LIBSUFFIX)
  1540. else
  1541. LIBNAME = $(LIBPREFIX)_$(LIBCORE)p$(REVISION).$(LIBSUFFIX)
  1542. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)p$(REVISION)_p.$(LIBSUFFIX)
  1543. endif
  1544. else
  1545. ifndef SMP
  1546. LIBNAME = $(LIBPREFIX)$(REVISION).$(LIBSUFFIX)
  1547. LIBNAME_P = $(LIBPREFIX)$(REVISION)_p.$(LIBSUFFIX)
  1548. else
  1549. LIBNAME = $(LIBPREFIX)p$(REVISION).$(LIBSUFFIX)
  1550. LIBNAME_P = $(LIBPREFIX)p$(REVISION)_p.$(LIBSUFFIX)
  1551. endif
  1552. endif
  1553. ifeq ($(FIXED_LIBNAME),1)
  1554. LIBNAME = lib$(LIBNAMEPREFIX)$(LIBSONAMEBASE)$(LIBNAMESUFFIX).$(LIBSUFFIX)
  1555. LIBNAME_P = lib$(LIBNAMEPREFIX)$(LISOBNAMEBASE)$(LIBNAMESUFFIX)_p.$(LIBSUFFIX)
  1556. endif
  1557. LIBDLLNAME = $(LIBPREFIX).dll
  1558. IMPLIBNAME = lib$(LIBNAMEBASE).dll.a
  1559. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.so)
  1560. LIBDYNNAME = $(LIBNAME:.$(LIBSUFFIX)=.dylib)
  1561. LIBDEFNAME = $(LIBNAME:.$(LIBSUFFIX)=.def)
  1562. LIBEXPNAME = $(LIBNAME:.$(LIBSUFFIX)=.exp)
  1563. LIBZIPNAME = $(LIBNAME:.$(LIBSUFFIX)=.zip)
  1564. LIBS = $(TOPDIR)/$(LIBNAME)
  1565. LIBS_P = $(TOPDIR)/$(LIBNAME_P)
  1566. LIB_COMPONENTS = BLAS
  1567. ifneq ($(NO_CBLAS), 1)
  1568. LIB_COMPONENTS += CBLAS
  1569. endif
  1570. ifneq ($(NO_LAPACK), 1)
  1571. LIB_COMPONENTS += LAPACK
  1572. ifneq ($(NO_LAPACKE), 1)
  1573. LIB_COMPONENTS += LAPACKE
  1574. endif
  1575. ifeq ($(BUILD_RELAPACK), 1)
  1576. LIB_COMPONENTS += ReLAPACK
  1577. endif
  1578. endif
  1579. ifeq ($(ONLY_CBLAS), 1)
  1580. LIB_COMPONENTS = CBLAS
  1581. endif
  1582. export OSNAME
  1583. export ARCH
  1584. export CORE
  1585. export LIBCORE
  1586. export __BYTE_ORDER__
  1587. export ELF_VERSION
  1588. export PGCPATH
  1589. export CONFIG
  1590. export CC
  1591. export FC
  1592. export BU
  1593. export FU
  1594. export NEED2UNDERSCORES
  1595. export USE_THREAD
  1596. export NUM_THREADS
  1597. export NUM_CORES
  1598. export SMP
  1599. export MAKEFILE_RULE
  1600. export NEED_PIC
  1601. export BINARY
  1602. export BINARY32
  1603. export BINARY64
  1604. export F_COMPILER
  1605. export C_COMPILER
  1606. export USE_OPENMP
  1607. export CROSS
  1608. export CROSS_SUFFIX
  1609. export NOFORTRAN
  1610. export C_LAPACK
  1611. export NO_FBLAS
  1612. export EXTRALIB
  1613. export CEXTRALIB
  1614. export FEXTRALIB
  1615. export HAVE_SSE
  1616. export HAVE_SSE2
  1617. export HAVE_SSE3
  1618. export HAVE_SSSE3
  1619. export HAVE_SSE4_1
  1620. export HAVE_SSE4_2
  1621. export HAVE_SSE4A
  1622. export HAVE_SSE5
  1623. export HAVE_AVX
  1624. export HAVE_AVX2
  1625. export HAVE_FMA3
  1626. export HAVE_VFP
  1627. export HAVE_VFPV3
  1628. export HAVE_VFPV4
  1629. export HAVE_NEON
  1630. ifndef NO_MSA
  1631. export HAVE_MSA
  1632. export MSA_FLAGS
  1633. endif
  1634. export KERNELDIR
  1635. export FUNCTION_PROFILE
  1636. export TARGET_CORE
  1637. export NO_AVX512
  1638. export NO_AVX2
  1639. export BUILD_BFLOAT16
  1640. export NO_LSX
  1641. export NO_LASX
  1642. export SBGEMM_UNROLL_M
  1643. export SBGEMM_UNROLL_N
  1644. export SGEMM_UNROLL_M
  1645. export SGEMM_UNROLL_N
  1646. export DGEMM_UNROLL_M
  1647. export DGEMM_UNROLL_N
  1648. export QGEMM_UNROLL_M
  1649. export QGEMM_UNROLL_N
  1650. export CGEMM_UNROLL_M
  1651. export CGEMM_UNROLL_N
  1652. export ZGEMM_UNROLL_M
  1653. export ZGEMM_UNROLL_N
  1654. export XGEMM_UNROLL_M
  1655. export XGEMM_UNROLL_N
  1656. export CGEMM3M_UNROLL_M
  1657. export CGEMM3M_UNROLL_N
  1658. export ZGEMM3M_UNROLL_M
  1659. export ZGEMM3M_UNROLL_N
  1660. export XGEMM3M_UNROLL_M
  1661. export XGEMM3M_UNROLL_N
  1662. ifdef USE_CUDA
  1663. export CUDADIR
  1664. export CUCC
  1665. export CUFLAGS
  1666. export CULIB
  1667. endif
  1668. .SUFFIXES: .$(PSUFFIX) .$(SUFFIX) .f
  1669. .f.$(SUFFIX):
  1670. $(FC) $(FFLAGS) -c $< -o $(@F)
  1671. .f.$(PSUFFIX):
  1672. $(FC) $(FPFLAGS) -pg -c $< -o $(@F)
  1673. ifdef BINARY64
  1674. PATHSCALEPATH = /opt/pathscale/lib/3.1
  1675. PGIPATH = /opt/pgi/linux86-64/7.1-5/lib
  1676. else
  1677. PATHSCALEPATH = /opt/pathscale/lib/3.1/32
  1678. PGIPATH = /opt/pgi/linux86/7.1-5/lib
  1679. endif
  1680. ACMLPATH = /opt/acml/4.3.0
  1681. ifneq ($(OSNAME), Darwin)
  1682. MKLPATH = /opt/intel/mkl/10.2.2.025/lib
  1683. else
  1684. MKLPATH = /Library/Frameworks/Intel_MKL.framework/Versions/10.0.1.014/lib
  1685. endif
  1686. ATLASPATH = /opt/atlas/3.9.17/opteron
  1687. FLAMEPATH = $(HOME)/flame/lib
  1688. ifneq ($(OSNAME), SunOS)
  1689. SUNPATH = /opt/sunstudio12.1
  1690. else
  1691. SUNPATH = /opt/SUNWspro
  1692. endif