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 42 kB

7 years ago
1 year ago
4 years ago
13 years ago
4 years ago
1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285128612871288128912901291129212931294129512961297129812991300130113021303130413051306130713081309131013111312131313141315131613171318131913201321132213231324132513261327132813291330133113321333133413351336133713381339134013411342134313441345134613471348134913501351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384138513861387138813891390139113921393139413951396139713981399140014011402140314041405140614071408140914101411141214131414141514161417141814191420142114221423142414251426142714281429143014311432143314341435143614371438143914401441144214431444144514461447144814491450145114521453145414551456145714581459146014611462146314641465146614671468146914701471147214731474147514761477147814791480148114821483148414851486148714881489149014911492149314941495149614971498149915001501150215031504150515061507150815091510151115121513151415151516151715181519152015211522152315241525152615271528152915301531153215331534153515361537153815391540154115421543154415451546154715481549155015511552155315541555155615571558155915601561156215631564156515661567156815691570157115721573157415751576157715781579158015811582158315841585158615871588158915901591159215931594159515961597159815991600160116021603160416051606160716081609161016111612161316141615161616171618161916201621162216231624162516261627162816291630163116321633163416351636163716381639164016411642164316441645164616471648164916501651165216531654165516561657165816591660166116621663166416651666166716681669167016711672167316741675167616771678167916801681168216831684168516861687168816891690169116921693169416951696169716981699170017011702170317041705170617071708170917101711171217131714171517161717171817191720172117221723172417251726172717281729173017311732173317341735173617371738173917401741174217431744174517461747174817491750175117521753175417551756175717581759176017611762176317641765176617671768176917701771177217731774177517761777177817791780178117821783178417851786178717881789179017911792179317941795179617971798179918001801180218031804180518061807180818091810181118121813181418151816181718181819182018211822182318241825182618271828182918301831183218331834183518361837183818391840184118421843184418451846184718481849185018511852185318541855185618571858185918601861186218631864186518661867186818691870187118721873187418751876187718781879188018811882188318841885188618871888188918901891189218931894189518961897189818991900190119021903190419051906190719081909191019111912191319141915191619171918191919201921192219231924192519261927192819291930193119321933193419351936193719381939194019411942194319441945194619471948194919501951195219531954195519561957195819591960196119621963196419651966196719681969197019711972197319741975
  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 ($(ARCH), arm64)
  238. SMALL_MATRIX_OPT = 1
  239. BUILD_BFLOAT16 = 1
  240. endif
  241. ifeq ($(ARCH), loongarch64)
  242. SMALL_MATRIX_OPT = 1
  243. endif
  244. ifeq ($(ARCH), arm64)
  245. GEMM_GEMV_FORWARD = 1
  246. SBGEMM_GEMV_FORWARD = 1
  247. BGEMM_GEMV_FORWARD = 1
  248. endif
  249. ifeq ($(ARCH), riscv)
  250. GEMM_GEMV_FORWARD = 1
  251. endif
  252. ifeq ($(ARCH), power)
  253. GEMM_GEMV_FORWARD = 1
  254. SBGEMM_GEMV_FORWARD = 1
  255. endif
  256. ifeq ($(SMALL_MATRIX_OPT), 1)
  257. CCOMMON_OPT += -DSMALL_MATRIX_OPT
  258. endif
  259. ifneq ($(ONLY_CBLAS), 1)
  260. ifeq ($(GEMM_GEMV_FORWARD), 1)
  261. CCOMMON_OPT += -DGEMM_GEMV_FORWARD
  262. endif
  263. ifeq ($(SBGEMM_GEMV_FORWARD), 1)
  264. CCOMMON_OPT += -DSBGEMM_GEMV_FORWARD
  265. endif
  266. ifeq ($(BGEMM_GEMV_FORWARD), 1)
  267. CCOMMON_OPT += -DBGEMM_GEMV_FORWARD
  268. endif
  269. endif
  270. # This operation is expensive, so execution should be once.
  271. ifndef GOTOBLAS_MAKEFILE
  272. export GOTOBLAS_MAKEFILE = 1
  273. # Generating Makefile.conf and config.h
  274. 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)
  275. endif
  276. ifndef TARGET_CORE
  277. -include $(TOPDIR)/Makefile.conf
  278. else
  279. HAVE_NEON=
  280. HAVE_VFP=
  281. HAVE_VFPV3=
  282. HAVE_VFPV4=
  283. HAVE_MMX=
  284. HAVE_SSE=
  285. HAVE_SSE2=
  286. HAVE_SSE3=
  287. HAVE_SSSE3=
  288. HAVE_SSE4_1=
  289. HAVE_SSE4_2=
  290. HAVE_SSE4A=
  291. HAVE_SSE5=
  292. HAVE_AVX=
  293. HAVE_AVX2=
  294. HAVE_FMA3=
  295. include $(TOPDIR)/Makefile_kernel.conf
  296. endif
  297. ifndef NUM_PARALLEL
  298. NUM_PARALLEL = 1
  299. endif
  300. ifndef NUM_THREADS
  301. NUM_THREADS = $(NUM_CORES)
  302. endif
  303. ifeq ($(NUM_THREADS), 1)
  304. override USE_THREAD = 0
  305. override USE_OPENMP = 0
  306. endif
  307. ifdef USE_THREAD
  308. ifeq ($(USE_THREAD), 0)
  309. SMP =
  310. else
  311. SMP = 1
  312. endif
  313. else
  314. ifeq ($(NUM_THREADS), 1)
  315. SMP =
  316. else
  317. SMP = 1
  318. endif
  319. endif
  320. ifeq ($(SMP), 1)
  321. USE_LOCKING =
  322. endif
  323. ifndef NEED_PIC
  324. NEED_PIC = 1
  325. endif
  326. ARFLAGS =
  327. CPP = $(COMPILER) -E
  328. AR ?= $(CROSS_SUFFIX)ar
  329. AS ?= $(CROSS_SUFFIX)as
  330. LD ?= $(CROSS_SUFFIX)ld
  331. RANLIB ?= $(CROSS_SUFFIX)ranlib
  332. NM = $(CROSS_SUFFIX)nm
  333. DLLWRAP = $(CROSS_SUFFIX)dllwrap
  334. OBJCOPY = $(CROSS_SUFFIX)objcopy
  335. OBJCONV = $(CROSS_SUFFIX)objconv
  336. # When fortran support was either not detected or actively deselected, only build BLAS.
  337. ifeq ($(NOFORTRAN), 1)
  338. C_LAPACK = 1
  339. override FEXTRALIB =
  340. endif
  341. ifeq ($(C_COMPILER), GCC)
  342. GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
  343. GCCVERSIONGT4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 4)
  344. GCCVERSIONGT5 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 5)
  345. GCCVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 7)
  346. GCCVERSIONGTEQ8 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 8)
  347. GCCVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  348. GCCVERSIONGTEQ10 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 10)
  349. GCCVERSIONGTEQ11 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 11)
  350. GCCVERSIONGTEQ12 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 12)
  351. GCCVERSIONGTEQ13 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 13)
  352. GCCVERSIONGTEQ14 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 14)
  353. # Note that the behavior of -dumpversion is compile-time-configurable for
  354. # gcc-7.x and newer. Use -dumpfullversion there
  355. ifeq ($(GCCVERSIONGTEQ7),1)
  356. GCCDUMPVERSION_PARAM := -dumpfullversion
  357. else
  358. GCCDUMPVERSION_PARAM := -dumpversion
  359. endif
  360. GCCMINORVERSIONGTEQ1 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 1)
  361. GCCMINORVERSIONGTEQ2 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 2)
  362. GCCMINORVERSIONGTEQ4 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 4)
  363. GCCMINORVERSIONGTEQ7 := $(shell expr `$(CC) $(GCCDUMPVERSION_PARAM) | cut -f2 -d.` \>= 7)
  364. endif
  365. ifeq ($(C_COMPILER), CLANG)
  366. CLANGVERSIONGTEQ9 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 9)
  367. CLANGVERSIONGTEQ12 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 12)
  368. endif
  369. #
  370. # OS dependent settings
  371. #
  372. ifeq ($(OSNAME), Darwin)
  373. ifndef MACOSX_DEPLOYMENT_TARGET
  374. ifeq ($(ARCH), arm64)
  375. export MACOSX_DEPLOYMENT_TARGET=11.0
  376. ifeq ($(C_COMPILER), GCC)
  377. export NO_SVE = 1
  378. export NO_SME = 1
  379. endif
  380. else
  381. export MACOSX_DEPLOYMENT_TARGET=10.8
  382. endif
  383. endif
  384. MD5SUM = md5 -r
  385. XCVER = $(shell pkgutil --pkg-info=com.apple.pkg.Xcode |awk '/version:/ {print $2}'|cut -d: -f2|cut -f1 -d.)
  386. ifeq (x$(XCVER)x,xx)
  387. XCVER = $(shell pkgutil --pkg-info=com.apple.pkg.CLTools_Executables |awk '/version:/ {print $2}'|cut -d: -f2|cut -f1 -d.)
  388. endif
  389. ifeq (x$(XCVER), x 15)
  390. CCOMMON_OPT += -Wl,-ld_classic
  391. FCOMMON_OPT += -Wl,-ld_classic
  392. endif
  393. ifeq (x$(XCVER), x 16)
  394. ifeq ($(F_COMPILER), GFORTRAN)
  395. override CEXTRALIB := $(filter-out(-lto_library, $(CEXTRALIB)))
  396. endif
  397. endif
  398. endif
  399. ifneq (,$(findstring $(OSNAME), FreeBSD OpenBSD DragonFly))
  400. MD5SUM = md5 -r
  401. endif
  402. ifeq ($(OSNAME), NetBSD)
  403. MD5SUM = md5 -n
  404. endif
  405. ifeq ($(OSNAME), Linux)
  406. EXTRALIB += -lm
  407. #NO_EXPRECISION = 1
  408. endif
  409. ifeq ($(OSNAME), Android)
  410. EXTRALIB += -lm
  411. endif
  412. ifeq ($(OSNAME), AIX)
  413. EXTRALIB += -lm
  414. endif
  415. ifeq ($(OSNAME), $(filter $(OSNAME),FreeBSD OpenBSD NetBSD DragonFly))
  416. ifeq ($(ARCH), $(filter $(ARCH),arm arm64))
  417. EXTRALIB += -lm
  418. endif
  419. endif
  420. ifeq ($(OSNAME), WINNT)
  421. NEED_PIC = 0
  422. NO_EXPRECISION = 1
  423. EXTRALIB += -defaultlib:advapi32
  424. SUFFIX = obj
  425. PSUFFIX = pobj
  426. LIBSUFFIX = a
  427. ifeq ($(C_COMPILER), CLANG)
  428. CCOMMON_OPT += -DMS_ABI
  429. endif
  430. #Version tests for supporting specific features (MS_ABI, POWER9 intrinsics)
  431. ifeq ($(GCCVERSIONGT4), 1)
  432. # GCC Major version > 4
  433. # It is compatible with MSVC ABI.
  434. CCOMMON_OPT += -DMS_ABI
  435. endif
  436. ifeq ($(GCCVERSIONGTEQ4), 1)
  437. ifeq ($(GCCMINORVERSIONGTEQ7), 1)
  438. # GCC Version >=4.7
  439. # It is compatible with MSVC ABI.
  440. CCOMMON_OPT += -DMS_ABI
  441. endif
  442. endif
  443. # Ensure the correct stack alignment on Win32
  444. # http://permalink.gmane.org/gmane.comp.lib.openblas.general/97
  445. ifeq ($(ARCH), x86)
  446. CCOMMON_OPT += -mincoming-stack-boundary=2
  447. FCOMMON_OPT += -mincoming-stack-boundary=2
  448. endif
  449. endif
  450. ifeq ($(OSNAME), Interix)
  451. NEED_PIC = 0
  452. NO_EXPRECISION = 1
  453. INTERIX_TOOL_DIR = /opt/gcc.3.3/i586-pc-interix3/bin
  454. endif
  455. ifeq ($(OSNAME), CYGWIN_NT)
  456. NEED_PIC = 0
  457. NO_EXPRECISION = 1
  458. OS_CYGWIN_NT = 1
  459. endif
  460. ifneq ($(OSNAME), WINNT)
  461. ifneq ($(OSNAME), CYGWIN_NT)
  462. ifneq ($(OSNAME), Interix)
  463. ifneq ($(OSNAME), Android)
  464. ifdef SMP
  465. EXTRALIB += -lpthread
  466. endif
  467. endif
  468. endif
  469. endif
  470. endif
  471. # ifeq logical or
  472. ifeq ($(OSNAME), $(filter $(OSNAME),WINNT CYGWIN_NT Interix))
  473. OS_WINDOWS=1
  474. endif
  475. ifdef QUAD_PRECISION
  476. CCOMMON_OPT += -DQUAD_PRECISION
  477. NO_EXPRECISION = 1
  478. endif
  479. ifneq ($(ARCH), x86)
  480. ifneq ($(ARCH), x86_64)
  481. NO_EXPRECISION = 1
  482. endif
  483. endif
  484. ifdef UTEST_CHECK
  485. CCOMMON_OPT += -DUTEST_CHECK
  486. SANITY_CHECK = 1
  487. endif
  488. ifdef SANITY_CHECK
  489. CCOMMON_OPT += -DSANITY_CHECK -DREFNAME=$(*F)f$(BU)
  490. endif
  491. MAX_STACK_ALLOC ?= 2048
  492. ifneq ($(MAX_STACK_ALLOC), 0)
  493. CCOMMON_OPT += -DMAX_STACK_ALLOC=$(MAX_STACK_ALLOC)
  494. endif
  495. ifdef USE_LOCKING
  496. ifneq ($(USE_LOCKING), 0)
  497. CCOMMON_OPT += -DUSE_LOCKING
  498. endif
  499. endif
  500. #
  501. # Architecture dependent settings
  502. #
  503. ifeq ($(ARCH), x86)
  504. ifndef BINARY
  505. NO_BINARY_MODE = 1
  506. endif
  507. ifeq ($(CORE), generic)
  508. #NO_EXPRECISION = 1
  509. endif
  510. ifndef NO_EXPRECISION
  511. ifeq ($(F_COMPILER), GFORTRAN)
  512. # ifeq logical or. GCC or LSB
  513. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  514. EXPRECISION = 1
  515. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  516. FCOMMON_OPT += -m128bit-long-double
  517. endif
  518. ifeq ($(C_COMPILER), CLANG)
  519. EXPRECISION = 1
  520. CCOMMON_OPT += -DEXPRECISION
  521. FCOMMON_OPT += -m128bit-long-double
  522. endif
  523. endif
  524. endif
  525. endif
  526. ifeq ($(ARCH), x86_64)
  527. ifeq ($(CORE), generic)
  528. #NO_EXPRECISION = 1
  529. endif
  530. ifndef NO_EXPRECISION
  531. ifeq ($(F_COMPILER), GFORTRAN)
  532. # ifeq logical or. GCC or LSB
  533. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  534. EXPRECISION = 1
  535. CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
  536. FCOMMON_OPT += -m128bit-long-double
  537. endif
  538. ifeq ($(C_COMPILER), CLANG)
  539. EXPRECISION = 1
  540. CCOMMON_OPT += -DEXPRECISION
  541. FCOMMON_OPT += -m128bit-long-double
  542. endif
  543. endif
  544. endif
  545. endif
  546. ifeq ($(C_COMPILER), INTEL)
  547. CCOMMON_OPT += -wd981
  548. endif
  549. ifeq ($(USE_OPENMP), 1)
  550. #check
  551. ifeq ($(USE_THREAD), 0)
  552. $(error OpenBLAS: Cannot set both USE_OPENMP=1 and USE_THREAD=0. The USE_THREAD=0 is only for building single thread version.)
  553. endif
  554. # ifeq logical or. GCC or LSB
  555. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
  556. CCOMMON_OPT += -fopenmp
  557. endif
  558. ifeq ($(C_COMPILER), CLANG)
  559. CCOMMON_OPT += -fopenmp
  560. ifeq ($(F_COMPILER), GFORTRAN)
  561. FEXTRALIB := $(subst -lgomp,-lomp,$(FEXTRALIB))
  562. endif
  563. endif
  564. ifeq ($(C_COMPILER), INTEL)
  565. CCOMMON_OPT += -fopenmp
  566. endif
  567. ifeq ($(C_COMPILER), PGI)
  568. CCOMMON_OPT += -mp
  569. endif
  570. ifeq ($(C_COMPILER), OPEN64)
  571. CCOMMON_OPT += -mp
  572. CEXTRALIB += -lstdc++
  573. endif
  574. ifeq ($(C_COMPILER), PATHSCALE)
  575. CCOMMON_OPT += -mp
  576. endif
  577. endif
  578. ifeq ($(DYNAMIC_ARCH), 1)
  579. ifeq ($(ARCH), x86)
  580. DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
  581. CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
  582. endif
  583. ifeq ($(ARCH), x86_64)
  584. DYNAMIC_CORE = PRESCOTT CORE2
  585. ifeq ($(DYNAMIC_OLDER), 1)
  586. DYNAMIC_CORE += PENRYN DUNNINGTON
  587. endif
  588. DYNAMIC_CORE += NEHALEM
  589. ifeq ($(DYNAMIC_OLDER), 1)
  590. DYNAMIC_CORE += OPTERON OPTERON_SSE3
  591. endif
  592. DYNAMIC_CORE += BARCELONA
  593. ifeq ($(DYNAMIC_OLDER), 1)
  594. DYNAMIC_CORE += BOBCAT ATOM NANO
  595. endif
  596. ifneq ($(NO_AVX), 1)
  597. DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER STEAMROLLER EXCAVATOR
  598. endif
  599. ifneq ($(NO_AVX2), 1)
  600. DYNAMIC_CORE += HASWELL ZEN
  601. endif
  602. ifneq ($(NO_AVX512), 1)
  603. ifneq ($(NO_AVX2), 1)
  604. DYNAMIC_CORE += SKYLAKEX COOPERLAKE SAPPHIRERAPIDS
  605. endif
  606. endif
  607. endif
  608. ifdef DYNAMIC_LIST
  609. override DYNAMIC_CORE = PRESCOTT $(DYNAMIC_LIST)
  610. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_PRESCOTT
  611. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  612. CCOMMON_OPT += $(XCCOMMON_OPT)
  613. #CCOMMON_OPT += -DDYNAMIC_LIST='$(DYNAMIC_LIST)'
  614. endif
  615. ifeq ($(ARCH), arm64)
  616. DYNAMIC_CORE = ARMV8
  617. DYNAMIC_CORE += CORTEXA53
  618. DYNAMIC_CORE += CORTEXA57
  619. DYNAMIC_CORE += NEOVERSEN1
  620. ifneq ($(NO_SVE), 1)
  621. DYNAMIC_CORE += NEOVERSEV1
  622. DYNAMIC_CORE += NEOVERSEN2
  623. DYNAMIC_CORE += ARMV8SVE
  624. DYNAMIC_CORE += A64FX
  625. endif
  626. ifneq ($(NO_SME), 1)
  627. DYNAMIC_CORE += ARMV9SME
  628. endif
  629. DYNAMIC_CORE += THUNDERX
  630. DYNAMIC_CORE += THUNDERX2T99
  631. DYNAMIC_CORE += TSV110
  632. DYNAMIC_CORE += EMAG8180
  633. DYNAMIC_CORE += THUNDERX3T110
  634. ifdef DYNAMIC_LIST
  635. override DYNAMIC_CORE = ARMV8 $(DYNAMIC_LIST)
  636. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_ARMV8
  637. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  638. endif
  639. endif
  640. ifeq ($(ARCH), mips64)
  641. DYNAMIC_CORE = LOONGSON3R3 LOONGSON3R4 MIPS64_GENERIC
  642. ifdef DYNAMIC_LIST
  643. override DYNAMIC_CORE = MIPS64_GENERIC $(DYNAMIC_LIST)
  644. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_MIPS64_GENERIC
  645. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  646. endif
  647. endif
  648. ifeq ($(ARCH), loongarch64)
  649. DYNAMIC_CORE = LA64_GENERIC LA264 LA464
  650. endif
  651. ifeq ($(ARCH), riscv64)
  652. DYNAMIC_CORE = RISCV64_GENERIC
  653. DYNAMIC_CORE += RISCV64_ZVL128B
  654. DYNAMIC_CORE += RISCV64_ZVL256B
  655. ifdef DYNAMIC_LIST
  656. override DYNAMIC_CORE = RISCV64_GENERIC $(DYNAMIC_LIST)
  657. XCCOMMON_OPT = -DDYNAMIC_LIST -DDYN_RISCV64_GENERIC
  658. XCCOMMON_OPT += $(foreach dcore,$(DYNAMIC_LIST),-DDYN_$(dcore))
  659. endif
  660. endif
  661. ifeq ($(ARCH), zarch)
  662. DYNAMIC_CORE = ZARCH_GENERIC
  663. # if the compiler accepts -march=arch11 or -march=z13 and can compile a file
  664. # with z13-specific inline assembly, then we can include support for Z13.
  665. # note: -march=z13 is equivalent to -march=arch11 yet some compiler releases
  666. # only support one or the other.
  667. # note: LLVM version 6.x supported -march=z13 yet could not handle vector
  668. # registers in inline assembly, so the check for supporting the -march flag is
  669. # not enough.
  670. ZARCH_TEST_COMPILE=-c $(TOPDIR)/kernel/zarch/damin_z13.c -I$(TOPDIR) -o /dev/null > /dev/null 2> /dev/null
  671. ZARCH_CC_SUPPORTS_ARCH11=$(shell $(CC) -march=arch11 $(ZARCH_TEST_COMPILE) && echo 1)
  672. ZARCH_CC_SUPPORTS_Z13=$(shell $(CC) -march=z13 $(ZARCH_TEST_COMPILE) && echo 1)
  673. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH11), $(ZARCH_CC_SUPPORTS_Z13)), 1)
  674. DYNAMIC_CORE += Z13
  675. CCOMMON_OPT += -DDYN_Z13
  676. else
  677. $(info OpenBLAS: Not building Z13 kernels because the compiler $(CC) does not support it)
  678. endif
  679. # as above for z13, check for -march=arch12 and z14 support in the compiler.
  680. ZARCH_CC_SUPPORTS_ARCH12=$(shell $(CC) -march=arch12 $(ZARCH_TEST_COMPILE) && echo 1)
  681. ZARCH_CC_SUPPORTS_Z14=$(shell $(CC) -march=z14 $(ZARCH_TEST_COMPILE) && echo 1)
  682. ifeq ($(or $(ZARCH_CC_SUPPORTS_ARCH12), $(ZARCH_CC_SUPPORTS_Z14)), 1)
  683. DYNAMIC_CORE += Z14
  684. CCOMMON_OPT += -DDYN_Z14
  685. else
  686. $(info OpenBLAS: Not building Z14 kernels because the compiler $(CC) does not support it)
  687. endif
  688. endif # ARCH zarch
  689. ifeq ($(ARCH), power)
  690. ifneq ($(C_COMPILER), PGI)
  691. DYNAMIC_CORE = POWER6
  692. DYNAMIC_CORE += POWER8
  693. ifneq ($(C_COMPILER), GCC)
  694. DYNAMIC_CORE += POWER9
  695. DYNAMIC_CORE += POWER10
  696. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  697. endif
  698. ifeq ($(C_COMPILER), GCC)
  699. ifeq ($(GCCVERSIONGT5), 1)
  700. DYNAMIC_CORE += POWER9
  701. else
  702. $(info, OpenBLAS: Your gcc version is too old to build the POWER9 kernels.)
  703. endif
  704. ifeq ($(OSNAME), AIX)
  705. LDVERSIONGTEQ35 := 1
  706. else
  707. LDVERSIONGTEQ35 := $(shell expr `$(CC) -Wl,--version 2> /dev/null | head -1 | cut -f2 -d "." | cut -f1 -d "-"` \>= 35)
  708. endif
  709. ifeq ($(GCCVERSIONGTEQ11)$(LDVERSIONGTEQ35), 11)
  710. DYNAMIC_CORE += POWER10
  711. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  712. else ifeq ($(GCCVERSIONGTEQ10), 1)
  713. ifeq ($(GCCMINORVERSIONGTEQ2)$(LDVERSIONGTEQ35), 11)
  714. DYNAMIC_CORE += POWER10
  715. CCOMMON_OPT += -DHAVE_P10_SUPPORT
  716. endif
  717. else
  718. $(info, OpenBLAS: Your gcc version is too old to build the POWER10 kernels.)
  719. endif
  720. endif
  721. else
  722. DYNAMIC_CORE = POWER8
  723. DYNAMIC_CORE += POWER9
  724. endif
  725. endif
  726. # If DYNAMIC_CORE is not set, DYNAMIC_ARCH cannot do anything, so force it to empty
  727. ifndef DYNAMIC_CORE
  728. override DYNAMIC_ARCH=
  729. endif
  730. endif
  731. ifeq ($(ARCH), ia64)
  732. NO_BINARY_MODE = 1
  733. BINARY_DEFINED = 1
  734. ifeq ($(F_COMPILER), GFORTRAN)
  735. ifeq ($(C_COMPILER), GCC)
  736. EXPRECISION = 1
  737. CCOMMON_OPT += -DEXPRECISION
  738. endif
  739. endif
  740. endif
  741. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  742. NO_BINARY_MODE = 1
  743. endif
  744. ifeq ($(ARCH), alpha)
  745. NO_BINARY_MODE = 1
  746. BINARY_DEFINED = 1
  747. endif
  748. ifeq ($(ARCH), arm)
  749. NO_BINARY_MODE = 1
  750. BINARY_DEFINED = 1
  751. ifneq ($(EMBEDDED), 1)
  752. CCOMMON_OPT += -marm
  753. FCOMMON_OPT += -marm
  754. else
  755. CCOMMON_OPT += -DOS_EMBEDDED -mthumb -mcpu=cortex-m4 -mfloat-abi=hard -mfpu=fpv4-sp-d16
  756. endif
  757. # If softfp abi is mentioned on the command line, force it.
  758. ifeq ($(ARM_SOFTFP_ABI), 1)
  759. CCOMMON_OPT += -mfloat-abi=softfp
  760. FCOMMON_OPT += -mfloat-abi=softfp
  761. endif
  762. ifeq ($(OSNAME), Android)
  763. ifeq ($(ARM_SOFTFP_ABI), 1)
  764. EXTRALIB += -lm
  765. else
  766. EXTRALIB += -Wl,-lm_hard
  767. endif
  768. endif
  769. endif
  770. ifeq ($(ARCH), arm64)
  771. NO_BINARY_MODE = 1
  772. BINARY_DEFINED = 1
  773. ifdef INTERFACE64
  774. ifneq ($(INTERFACE64), 0)
  775. ifeq ($(F_COMPILER), GFORTRAN)
  776. FCOMMON_OPT += -fdefault-integer-8
  777. endif
  778. ifeq ($(F_COMPILER), FLANG)
  779. FCOMMON_OPT += -i8
  780. endif
  781. endif
  782. endif
  783. endif
  784. ifeq ($(ARCH), riscv64)
  785. NO_BINARY_MODE = 1
  786. BINARY_DEFINED = 1
  787. ifdef INTERFACE64
  788. ifneq ($(INTERFACE64), 0)
  789. ifeq ($(F_COMPILER), GFORTRAN)
  790. FCOMMON_OPT += -fdefault-integer-8
  791. endif
  792. ifeq ($(F_COMPILER), FLANG)
  793. FCOMMON_OPT += -i8
  794. endif
  795. endif
  796. endif
  797. endif
  798. ifeq ($(ARCH), loongarch64)
  799. NO_BINARY_MODE = 1
  800. BINARY_DEFINED = 1
  801. ifdef INTERFACE64
  802. ifneq ($(INTERFACE64), 0)
  803. ifeq ($(F_COMPILER), GFORTRAN)
  804. FCOMMON_OPT += -fdefault-integer-8
  805. endif
  806. ifeq ($(F_COMPILER), FLANG)
  807. FCOMMON_OPT += -i8
  808. endif
  809. endif
  810. endif
  811. endif
  812. ifeq ($(ARCH), csky)
  813. NO_BINARY_MODE = 1
  814. BINARY_DEFINED = 1
  815. endif
  816. #
  817. # C Compiler dependent settings
  818. #
  819. # ifeq logical or. GCC or CLANG or LSB
  820. # http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
  821. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG LSB))
  822. CCOMMON_OPT += -Wall
  823. COMMON_PROF += -fno-inline
  824. NO_UNINITIALIZED_WARN = -Wno-uninitialized
  825. ifeq ($(QUIET_MAKE), 1)
  826. CCOMMON_OPT += $(NO_UNINITIALIZED_WARN) -Wno-unused
  827. endif
  828. ifdef NO_BINARY_MODE
  829. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  830. ifdef BINARY64
  831. CCOMMON_OPT += -mabi=64
  832. else
  833. CCOMMON_OPT += -mabi=n32
  834. endif
  835. BINARY_DEFINED = 1
  836. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  837. CCOMMON_OPT += -mabi=32
  838. BINARY_DEFINED = 1
  839. endif
  840. ifneq (, $(filter $(CORE), MIPS64_GENERIC))
  841. CCOMMON_OPT += -DNO_MSA
  842. FCOMMON_OPT += -DNO_MSA
  843. endif
  844. ifneq (, $(filter $(CORE),LOONGSON3R3 LOONGSON3R4))
  845. CCOMMON_OPT += -march=loongson3a
  846. FCOMMON_OPT += -march=loongson3a
  847. endif
  848. ifeq ($(CORE), MIPS24K)
  849. CCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  850. FCOMMON_OPT += -mips32r2 -mtune=24kc $(MSA_FLAGS)
  851. endif
  852. ifeq ($(CORE), MIPS1004K)
  853. CCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  854. FCOMMON_OPT += -mips32r2 $(MSA_FLAGS)
  855. endif
  856. ifeq ($(CORE), P5600)
  857. CCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  858. FCOMMON_OPT += -mips32r5 -mnan=2008 -mtune=p5600 $(MSA_FLAGS)
  859. endif
  860. ifeq ($(CORE), I6400)
  861. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  862. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  863. endif
  864. ifeq ($(CORE), P6600)
  865. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  866. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=p6600 $(MSA_FLAGS)
  867. endif
  868. ifeq ($(CORE), I6500)
  869. CCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  870. FCOMMON_OPT += -mips64r6 -mnan=2008 -mtune=i6400 $(MSA_FLAGS)
  871. endif
  872. ifeq ($(OSNAME), AIX)
  873. BINARY_DEFINED = 1
  874. endif
  875. ifeq ($(ARCH), loongarch64)
  876. LA64_ABI=$(shell $(CC) -mabi=lp64d -c $(TOPDIR)/cpuid_loongarch64.c -o /dev/null > /dev/null 2> /dev/null && echo lp64d)
  877. LA64_ARCH=$(shell $(CC) -march=loongarch64 -c $(TOPDIR)/cpuid_loongarch64.c -o /dev/null > /dev/null 2> /dev/null && echo loongarch64)
  878. ifneq ($(LA64_ABI), lp64d)
  879. LA64_ABI=lp64
  880. endif
  881. ifneq ($(LA64_ARCH), loongarch64)
  882. CCOMMON_OPT += -mabi=$(LA64_ABI)
  883. FCOMMON_OPT += -mabi=$(LA64_ABI)
  884. else
  885. CCOMMON_OPT += -march=loongarch64 -mabi=$(LA64_ABI)
  886. FCOMMON_OPT += -march=loongarch64 -mabi=$(LA64_ABI)
  887. endif
  888. endif
  889. endif
  890. ifndef BINARY_DEFINED
  891. ifneq ($(OSNAME), AIX)
  892. ifdef BINARY64
  893. ifneq ($(ARCH), riscv64)
  894. CCOMMON_OPT += -m64
  895. endif
  896. else
  897. CCOMMON_OPT += -m32
  898. endif
  899. endif
  900. endif
  901. endif
  902. ifeq ($(C_COMPILER), PGI)
  903. PGCVERSIONGT20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \> 20)
  904. PGCVERSIONEQ20 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` == 20)
  905. PGCMINORVERSIONGE11 := $(shell expr `$(CC) --version|sed -n "2p" |cut -d "-" -f 1 |sed -e "s/[^0-9.]//g" |cut -c 4-5` \>= 11)
  906. PGCVERSIONCHECK := $(PGCVERSIONGT20)$(PGCVERSIONEQ20)$(PGCMINORVERSIONGE11)
  907. ifeq ($(PGCVERSIONCHECK), $(filter $(PGCVERSIONCHECK), 100 101 011))
  908. NEWPGI := 1
  909. PGCVERSIONGT21 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` \> 21)
  910. PGCVERSIONEQ21 := $(shell expr `$(CC) --version|sed -n "2p" |sed -e "s/[^0-9.]//g" |cut -d "." -f 1` == 21)
  911. PGCVERSIONCHECK2 := $(PGCVERSIONGT21)$(PGCVERSIONEQ21)$(PGCMINORVERSIONGE11)
  912. ifeq ($(PGCVERSIONCHECK2), $(filter $(PGCVERSIONCHECK2), 100 101 011))
  913. NEWPGI2 := 1
  914. endif
  915. endif
  916. ifdef BINARY64
  917. ifeq ($(ARCH), x86_64)
  918. ifeq (,$(findstring tp,$(CFLAGS)))
  919. ifneq ($(NEWPGI2),1)
  920. CCOMMON_OPT += -tp p7-64
  921. else
  922. CCOMMON_OPT += -tp px
  923. endif
  924. endif
  925. ifneq ($(NEWPGI),1)
  926. CCOMMON_OPT += -D__MMX__ -Mnollvm
  927. endif
  928. else
  929. ifeq ($(ARCH), power)
  930. ifeq (,$(findstring tp,$(CFLAGS)))
  931. ifeq ($(CORE), POWER8)
  932. CCOMMON_OPT += -tp pwr8
  933. endif
  934. ifeq ($(CORE), POWER9)
  935. CCOMMON_OPT += -tp pwr9
  936. endif
  937. endif
  938. endif
  939. endif
  940. else
  941. ifneq ($(NEWPGI2),1)
  942. ifeq (,$(findstring tp,$(CFLAGS)))
  943. CCOMMON_OPT += -tp p7
  944. else
  945. CCOMMON_OPT += -tp px
  946. endif
  947. endif
  948. endif
  949. endif
  950. ifeq ($(C_COMPILER), PATHSCALE)
  951. ifdef BINARY64
  952. CCOMMON_OPT += -m64
  953. else
  954. CCOMMON_OPT += -m32
  955. endif
  956. endif
  957. #
  958. # Fortran Compiler dependent settings
  959. #
  960. ifeq ($(F_COMPILER), NAG)
  961. FCOMMON_OPT += -dcfuns -recursive -ieee=full -w=obs -thread_safe
  962. ifdef INTERFACE64
  963. ifneq ($(INTERFACE64), 0)
  964. FCOMMON_OPT += -i8
  965. endif
  966. endif
  967. ifeq ($(USE_OPENMP), 1)
  968. FCOMMON_OPT += -openmp
  969. endif
  970. endif
  971. ifeq ($(F_COMPILER), FLANG)
  972. CCOMMON_OPT += -DF_INTERFACE_FLANG
  973. FCOMMON_OPT += -Mrecursive -Kieee
  974. ifeq ($(OSNAME), Linux)
  975. ifeq ($(ARCH), x86_64)
  976. FLANG_VENDOR := $(shell $(FC) --version|head -1 |cut -f 1 -d " ")
  977. ifeq ($(FLANG_VENDOR), AMD)
  978. FCOMMON_OPT += -fno-unroll-loops
  979. endif
  980. endif
  981. endif
  982. ifdef BINARY64
  983. ifdef INTERFACE64
  984. ifneq ($(INTERFACE64), 0)
  985. FCOMMON_OPT += -i8
  986. endif
  987. endif
  988. FCOMMON_OPT += -Wall
  989. else
  990. FCOMMON_OPT += -Wall
  991. endif
  992. ifeq ($(USE_OPENMP), 1)
  993. FCOMMON_OPT += -fopenmp
  994. endif
  995. endif
  996. ifeq ($(F_COMPILER), G77)
  997. CCOMMON_OPT += -DF_INTERFACE_G77
  998. FCOMMON_OPT += -Wall
  999. ifndef NO_BINARY_MODE
  1000. ifneq ($(OSNAME), AIX)
  1001. ifdef BINARY64
  1002. FCOMMON_OPT += -m64
  1003. else
  1004. FCOMMON_OPT += -m32
  1005. endif
  1006. endif
  1007. endif
  1008. endif
  1009. ifeq ($(F_COMPILER), G95)
  1010. CCOMMON_OPT += -DF_INTERFACE_G95
  1011. FCOMMON_OPT += -Wall
  1012. ifneq ($(OSNAME), AIX)
  1013. ifndef NO_BINARY_MODE
  1014. ifdef BINARY64
  1015. FCOMMON_OPT += -m64
  1016. else
  1017. FCOMMON_OPT += -m32
  1018. endif
  1019. endif
  1020. ifneq ($(NO_LAPACKE), 1)
  1021. FCOMMON_OPT += -fno-second-underscore
  1022. endif
  1023. endif
  1024. endif
  1025. ifeq ($(F_COMPILER), $(filter $(F_COMPILER),GFORTRAN FLANGNEW))
  1026. CCOMMON_OPT += -DF_INTERFACE_GFORT
  1027. ifeq ($(F_COMPILER), GFORTRAN)
  1028. FCOMMON_OPT += -Wall
  1029. # make single-threaded LAPACK calls thread-safe #1847
  1030. FCOMMON_OPT += -frecursive
  1031. # work around ABI problem with passing single-character arguments
  1032. FCOMMON_OPT += -fno-optimize-sibling-calls
  1033. #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
  1034. ifneq ($(NOFORTRAN), 1)
  1035. ifneq ($(NOFORTRAN), 2)
  1036. ifneq ($(NO_LAPACK), 1)
  1037. EXTRALIB += -lgfortran
  1038. endif
  1039. endif
  1040. endif
  1041. endif
  1042. ifdef NO_BINARY_MODE
  1043. ifeq ($(ARCH), $(filter $(ARCH),mips64))
  1044. ifdef BINARY64
  1045. FCOMMON_OPT += -mabi=64
  1046. else
  1047. FCOMMON_OPT += -mabi=n32
  1048. endif
  1049. else ifeq ($(ARCH), $(filter $(ARCH),mips))
  1050. FCOMMON_OPT += -mabi=32
  1051. endif
  1052. ifeq ($(ARCH), $(filter $(ARCH),loongarch64))
  1053. ifdef INTERFACE64
  1054. ifneq ($(INTERFACE64), 0)
  1055. FCOMMON_OPT += -fdefault-integer-8
  1056. endif
  1057. endif
  1058. endif
  1059. else
  1060. ifdef BINARY64
  1061. ifneq ($(OSNAME), AIX)
  1062. ifneq ($(ARCH), riscv64)
  1063. FCOMMON_OPT += -m64
  1064. endif
  1065. endif
  1066. ifdef INTERFACE64
  1067. ifneq ($(INTERFACE64), 0)
  1068. FCOMMON_OPT += -fdefault-integer-8
  1069. endif
  1070. endif
  1071. else
  1072. ifneq ($(OSNAME), AIX)
  1073. FCOMMON_OPT += -m32
  1074. endif
  1075. endif
  1076. endif
  1077. ifeq ($(USE_OPENMP), 1)
  1078. FCOMMON_OPT += -fopenmp
  1079. endif
  1080. endif
  1081. ifeq ($(F_COMPILER), INTEL)
  1082. CCOMMON_OPT += -DF_INTERFACE_INTEL
  1083. ifdef INTERFACE64
  1084. ifneq ($(INTERFACE64), 0)
  1085. FCOMMON_OPT += -i8
  1086. endif
  1087. endif
  1088. FCOMMON_OPT += -recursive -fp-model strict -assume protect-parens
  1089. ifeq ($(USE_OPENMP), 1)
  1090. FCOMMON_OPT += -fopenmp
  1091. endif
  1092. endif
  1093. ifeq ($(F_COMPILER), FUJITSU)
  1094. CCOMMON_OPT += -DF_INTERFACE_FUJITSU
  1095. ifeq ($(USE_OPENMP), 1)
  1096. FCOMMON_OPT += -openmp
  1097. endif
  1098. endif
  1099. ifeq ($(F_COMPILER), IBM)
  1100. CCOMMON_OPT += -DF_INTERFACE_IBM
  1101. FEXTRALIB += -lxlf90
  1102. ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG))
  1103. FCOMMON_OPT += -qextname -qzerosize
  1104. endif
  1105. # FCOMMON_OPT += -qarch=440
  1106. ifdef BINARY64
  1107. FCOMMON_OPT += -q64
  1108. ifdef INTERFACE64
  1109. ifneq ($(INTERFACE64), 0)
  1110. FCOMMON_OPT += -qintsize=8
  1111. endif
  1112. endif
  1113. else
  1114. FCOMMON_OPT += -q32
  1115. endif
  1116. endif
  1117. ifeq ($(F_COMPILER), PGI)
  1118. CCOMMON_OPT += -DF_INTERFACE_PGI
  1119. COMMON_PROF += -DPGICOMPILER
  1120. ifdef BINARY64
  1121. ifdef INTERFACE64
  1122. ifneq ($(INTERFACE64), 0)
  1123. FCOMMON_OPT += -i8
  1124. endif
  1125. endif
  1126. ifeq ($(ARCH), x86_64)
  1127. ifneq ($(NEWPGI2),1)
  1128. FCOMMON_OPT += -tp p7-64
  1129. else
  1130. FCOMMON_OPT += -tp px
  1131. endif
  1132. else
  1133. ifeq ($(ARCH), power)
  1134. ifeq ($(CORE), POWER6)
  1135. $(warning NVIDIA HPC compilers do not support POWER6.)
  1136. endif
  1137. ifeq ($(CORE), POWER8)
  1138. FCOMMON_OPT += -tp pwr8
  1139. endif
  1140. ifeq ($(CORE), POWER9)
  1141. FCOMMON_OPT += -tp pwr9
  1142. endif
  1143. ifeq ($(CORE), POWER10)
  1144. $(warning NVIDIA HPC compilers do not support POWER10.)
  1145. endif
  1146. endif
  1147. endif
  1148. else
  1149. FCOMMON_OPT += -tp p7
  1150. endif
  1151. FCOMMON_OPT += -Mrecursive -Kieee
  1152. ifeq ($(USE_OPENMP), 1)
  1153. FCOMMON_OPT += -mp
  1154. endif
  1155. endif
  1156. ifeq ($(F_COMPILER), PATHSCALE)
  1157. CCOMMON_OPT += -DF_INTERFACE_PATHSCALE
  1158. ifdef BINARY64
  1159. ifdef INTERFACE64
  1160. ifneq ($(INTERFACE64), 0)
  1161. FCOMMON_OPT += -i8
  1162. endif
  1163. endif
  1164. endif
  1165. ifeq ($(USE_OPENMP), 1)
  1166. FCOMMON_OPT += -mp
  1167. endif
  1168. endif
  1169. ifeq ($(F_COMPILER), OPEN64)
  1170. CCOMMON_OPT += -DF_INTERFACE_OPEN64
  1171. ifdef BINARY64
  1172. ifdef INTERFACE64
  1173. ifneq ($(INTERFACE64), 0)
  1174. FCOMMON_OPT += -i8
  1175. endif
  1176. endif
  1177. endif
  1178. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  1179. ifndef BINARY64
  1180. FCOMMON_OPT += -n32
  1181. else
  1182. FCOMMON_OPT += -n64
  1183. endif
  1184. ifeq ($(CORE), LOONGSON3R3)
  1185. FCOMMON_OPT += -loongson3 -static
  1186. endif
  1187. ifeq ($(CORE), LOONGSON3R4)
  1188. FCOMMON_OPT += -loongson3 -static
  1189. endif
  1190. else
  1191. ifndef BINARY64
  1192. FCOMMON_OPT += -m32
  1193. else
  1194. FCOMMON_OPT += -m64
  1195. endif
  1196. endif
  1197. ifeq ($(USE_OPENMP), 1)
  1198. FEXTRALIB += -lstdc++
  1199. FCOMMON_OPT += -mp
  1200. endif
  1201. endif
  1202. ifeq ($(C_COMPILER), OPEN64)
  1203. ifeq ($(ARCH), $(filter $(ARCH),mips64 mips))
  1204. ifndef BINARY64
  1205. CCOMMON_OPT += -n32
  1206. else
  1207. CCOMMON_OPT += -n64
  1208. endif
  1209. ifeq ($(CORE), LOONGSON3R3)
  1210. CCOMMON_OPT += -loongson3 -static
  1211. endif
  1212. ifeq ($(CORE), LOONGSON3R4)
  1213. CCOMMON_OPT += -loongson3 -static
  1214. endif
  1215. else
  1216. ifndef BINARY64
  1217. CCOMMON_OPT += -m32
  1218. else
  1219. CCOMMON_OPT += -m64
  1220. endif
  1221. endif
  1222. endif
  1223. ifeq ($(C_COMPILER), SUN)
  1224. CCOMMON_OPT += -w
  1225. ifeq ($(ARCH), x86)
  1226. CCOMMON_OPT += -m32
  1227. else
  1228. ifdef BINARY64
  1229. CCOMMON_OPT += -m64
  1230. else
  1231. CCOMMON_OPT += -m32
  1232. endif
  1233. endif
  1234. endif
  1235. ifeq ($(F_COMPILER), SUN)
  1236. CCOMMON_OPT += -DF_INTERFACE_SUN
  1237. FCOMMON_OPT += -ftrap=%none -xrecursive
  1238. ifeq ($(ARCH), x86)
  1239. FCOMMON_OPT += -m32
  1240. else
  1241. ifdef BINARY64
  1242. FCOMMON_OPT += -m64
  1243. else
  1244. FCOMMON_OPT += -m32
  1245. endif
  1246. endif
  1247. ifeq ($(USE_OPENMP), 1)
  1248. FCOMMON_OPT += -xopenmp=parallel
  1249. endif
  1250. endif
  1251. ifeq ($(F_COMPILER), COMPAQ)
  1252. CCOMMON_OPT += -DF_INTERFACE_COMPAQ
  1253. ifeq ($(USE_OPENMP), 1)
  1254. FCOMMON_OPT += -openmp
  1255. endif
  1256. endif
  1257. ifeq ($(F_COMPILER), CRAY)
  1258. CCOMMON_OPT += -DF_INTERFACE_INTEL
  1259. FCOMMON_OPT += -hnopattern
  1260. ifdef INTERFACE64
  1261. ifneq ($(INTERFACE64), 0)
  1262. FCOMMON_OPT += -s integer64
  1263. endif
  1264. endif
  1265. ifneq ($(USE_OPENMP), 1)
  1266. FCOMMON_OPT += -O noomp
  1267. endif
  1268. endif
  1269. ifdef BINARY64
  1270. ifdef INTERFACE64
  1271. ifneq ($(INTERFACE64), 0)
  1272. CCOMMON_OPT +=
  1273. #-DUSE64BITINT
  1274. endif
  1275. endif
  1276. endif
  1277. ifeq ($(NEED_PIC), 1)
  1278. ifeq ($(C_COMPILER), IBM)
  1279. CCOMMON_OPT += -qpic=large
  1280. else
  1281. CCOMMON_OPT += -fPIC
  1282. endif
  1283. ifeq ($(F_COMPILER), SUN)
  1284. FCOMMON_OPT += -pic
  1285. else ifeq ($(F_COMPILER), NAG)
  1286. FCOMMON_OPT += -PIC
  1287. else ifeq ($(F_COMPILER), IBM)
  1288. FCOMMON_OPT += -qpic=large
  1289. else
  1290. FCOMMON_OPT += -fPIC
  1291. endif
  1292. endif
  1293. ifeq ($(DYNAMIC_ARCH), 1)
  1294. CCOMMON_OPT += -DDYNAMIC_ARCH
  1295. endif
  1296. ifeq ($(DYNAMIC_OLDER), 1)
  1297. CCOMMON_OPT += -DDYNAMIC_OLDER
  1298. endif
  1299. ifeq ($(C_LAPACK), 1)
  1300. CCOMMON_OPT += -DC_LAPACK
  1301. endif
  1302. ifeq ($(NO_LAPACK), 1)
  1303. CCOMMON_OPT += -DNO_LAPACK
  1304. #Disable LAPACK C interface
  1305. NO_LAPACKE = 1
  1306. endif
  1307. ifeq ($(NO_LAPACKE), 1)
  1308. CCOMMON_OPT += -DNO_LAPACKE
  1309. endif
  1310. ifeq ($(NO_AVX), 1)
  1311. CCOMMON_OPT += -DNO_AVX
  1312. endif
  1313. ifeq ($(ARCH), x86)
  1314. CCOMMON_OPT += -DNO_AVX
  1315. endif
  1316. ifeq ($(NO_AVX2), 1)
  1317. CCOMMON_OPT += -DNO_AVX2
  1318. endif
  1319. ifeq ($(NO_AVX512), 1)
  1320. CCOMMON_OPT += -DNO_AVX512
  1321. endif
  1322. ifeq ($(NO_SVE), 1)
  1323. CCOMMON_OPT += -DNO_SVE
  1324. endif
  1325. ifeq ($(NO_SME), 1)
  1326. CCOMMON_OPT += -DNO_SME
  1327. endif
  1328. ifdef SMP
  1329. CCOMMON_OPT += -DSMP_SERVER
  1330. ifeq ($(ARCH), mips64)
  1331. USE_SIMPLE_THREADED_LEVEL3 = 1
  1332. endif
  1333. ifeq ($(USE_OPENMP), 1)
  1334. # USE_SIMPLE_THREADED_LEVEL3 = 1
  1335. # NO_AFFINITY = 1
  1336. CCOMMON_OPT += -DUSE_OPENMP
  1337. endif
  1338. ifeq ($(BIGNUMA), 1)
  1339. CCOMMON_OPT += -DBIGNUMA
  1340. endif
  1341. endif
  1342. ifeq ($(NO_WARMUP), 1)
  1343. CCOMMON_OPT += -DNO_WARMUP
  1344. endif
  1345. ifeq ($(CONSISTENT_FPCSR), 1)
  1346. CCOMMON_OPT += -DCONSISTENT_FPCSR
  1347. endif
  1348. # Only for development
  1349. # CCOMMON_OPT += -DPARAMTEST
  1350. # CCOMMON_OPT += -DPREFETCHTEST
  1351. # CCOMMON_OPT += -DNO_SWITCHING
  1352. # USE_PAPI = 1
  1353. ifdef USE_PAPI
  1354. CCOMMON_OPT += -DUSE_PAPI
  1355. EXTRALIB += -lpapi -lperfctr
  1356. endif
  1357. ifdef BUFFERSIZE
  1358. CCOMMON_OPT += -DBUFFERSIZE=$(BUFFERSIZE)
  1359. endif
  1360. ifdef DYNAMIC_THREADS
  1361. CCOMMON_OPT += -DDYNAMIC_THREADS
  1362. endif
  1363. CCOMMON_OPT += -DMAX_CPU_NUMBER=$(NUM_THREADS)
  1364. CCOMMON_OPT += -DMAX_PARALLEL_NUMBER=$(NUM_PARALLEL)
  1365. ifdef USE_SIMPLE_THREADED_LEVEL3
  1366. CCOMMON_OPT += -DUSE_SIMPLE_THREADED_LEVEL3
  1367. endif
  1368. ifeq ($(USE_TLS), 1)
  1369. CCOMMON_OPT += -DUSE_TLS
  1370. endif
  1371. ifeq ($(BUILD_BFLOAT16), 1)
  1372. CCOMMON_OPT += -DBUILD_BFLOAT16
  1373. endif
  1374. ifeq ($(BUILD_HFLOAT16), 1)
  1375. CCOMMON_OPT += -DBUILD_HFLOAT16
  1376. endif
  1377. ifeq ($(BUILD_SINGLE), 1)
  1378. CCOMMON_OPT += -DBUILD_SINGLE=1
  1379. endif
  1380. ifeq ($(BUILD_DOUBLE), 1)
  1381. CCOMMON_OPT += -DBUILD_DOUBLE=1
  1382. endif
  1383. ifeq ($(BUILD_COMPLEX), 1)
  1384. CCOMMON_OPT += -DBUILD_COMPLEX=1
  1385. endif
  1386. ifeq ($(BUILD_COMPLEX16), 1)
  1387. CCOMMON_OPT += -DBUILD_COMPLEX16=1
  1388. endif
  1389. CCOMMON_OPT += -DVERSION=\"$(VERSION)\"
  1390. ifndef SYMBOLPREFIX
  1391. SYMBOLPREFIX =
  1392. endif
  1393. ifndef SYMBOLSUFFIX
  1394. SYMBOLSUFFIX =
  1395. endif
  1396. ifndef LIBSONAMEBASE
  1397. LIBSONAMEBASE = openblas
  1398. endif
  1399. ifndef LIBNAMEPREFIX
  1400. LIBNAMEPREFIX =
  1401. endif
  1402. SYMPREFIX=$(SYMBOLPREFIX)
  1403. ifeq ($(SYMBOLPREFIX),$(LIBNAMEPREFIX))
  1404. SYMPREFIX=
  1405. endif
  1406. SYMSUFFIX=$(SYMBOLSUFFIX)
  1407. ifeq ($(SYMBOLSUFFIX),$(LIBNAMESUFFIX))
  1408. SYMSUFFIX=
  1409. endif
  1410. ifndef LIBNAMESUFFIX
  1411. LIBNAMEBASE = $(SYMPREFIX)$(LIBSONAMEBASE)$(SYMSUFFIX)
  1412. else
  1413. LIBNAMEBASE = $(SYMPREFIX)$(LIBSONAMEBASE)$(SYMSUFFIX)$(LIBNAMESUFFIX)
  1414. endif
  1415. ifeq ($(OSNAME), CYGWIN_NT)
  1416. LIBPREFIX = cyg$(LIBNAMEPREFIX)$(LIBNAMEBASE)
  1417. else
  1418. LIBPREFIX = lib$(LIBNAMEPREFIX)$(LIBNAMEBASE)
  1419. endif
  1420. KERNELDIR = $(TOPDIR)/kernel/$(ARCH)
  1421. include $(TOPDIR)/Makefile.$(ARCH)
  1422. ifneq ($(C_COMPILER), PGI)
  1423. ifneq ($(C_COMPILER), SUN)
  1424. CCOMMON_OPT += -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME
  1425. endif
  1426. endif
  1427. CCOMMON_OPT += -DASMNAME=$(FU)$(*F) -DASMFNAME=$(FU)$(*F)$(BU) -DNAME=$(*F)$(BU) -DCNAME=$(*F) -DCHAR_NAME=\"$(*F)$(BU)\" -DCHAR_CNAME=\"$(*F)\"
  1428. ifeq ($(CORE), PPC440)
  1429. CCOMMON_OPT += -DALLOC_QALLOC
  1430. endif
  1431. ifeq ($(CORE), PPC440FP2)
  1432. STATIC_ALLOCATION = 1
  1433. endif
  1434. ifneq ($(OSNAME), Linux)
  1435. NO_AFFINITY = 1
  1436. endif
  1437. ifneq ($(ARCH), x86_64)
  1438. ifneq ($(ARCH), x86)
  1439. NO_AFFINITY = 1
  1440. endif
  1441. endif
  1442. ifeq ($(ARCH), POWER)
  1443. ifeq ($(DEBUG), 1)
  1444. CCOMMON_OPT := $(filter-out -O%, $(CCOMMON_OPT)) -O0
  1445. FCOMMON_OPT := $(filter-out -O%, $(FCOMMON_OPT)) -O0
  1446. endif
  1447. endif
  1448. ifdef NO_AFFINITY
  1449. ifeq ($(NO_AFFINITY), 0)
  1450. override undefine NO_AFFINITY
  1451. else
  1452. CCOMMON_OPT += -DNO_AFFINITY
  1453. endif
  1454. endif
  1455. ifdef FUNCTION_PROFILE
  1456. CCOMMON_OPT += -DFUNCTION_PROFILE
  1457. endif
  1458. ifdef SHMEM_ALLOCATION
  1459. ifneq ($(SHMEM_ALLOCATION), 0)
  1460. CCOMMON_OPT += -DALLOC_SHM
  1461. endif
  1462. endif
  1463. ifdef HUGETLB_ALLOCATION
  1464. ifneq ($(HUGETLB_ALLOCATION), 0)
  1465. CCOMMON_OPT += -DALLOC_HUGETLB
  1466. endif
  1467. endif
  1468. ifdef HUGETLBFILE_ALLOCATION
  1469. ifneq ($(HUGETLBFILE_ALLOCATION), 0)
  1470. CCOMMON_OPT += -DALLOC_HUGETLBFILE -DHUGETLB_FILE_NAME=$(HUGETLBFILE_ALLOCATION)
  1471. endif
  1472. endif
  1473. ifdef STATIC_ALLOCATION
  1474. CCOMMON_OPT += -DALLOC_STATIC
  1475. endif
  1476. ifdef DEVICEDRIVER_ALLOCATION
  1477. CCOMMON_OPT += -DALLOC_DEVICEDRIVER -DDEVICEDRIVER_NAME=\"/dev/mapper\"
  1478. endif
  1479. ifdef MIXED_MEMORY_ALLOCATION
  1480. CCOMMON_OPT += -DMIXED_MEMORY_ALLOCATION
  1481. endif
  1482. ifeq ($(OSNAME), SunOS)
  1483. TAR = gtar
  1484. PATCH = gpatch
  1485. GREP = ggrep
  1486. AWK = nawk
  1487. else
  1488. TAR = tar
  1489. PATCH = patch
  1490. GREP = grep
  1491. AWK = awk
  1492. endif
  1493. ifndef MD5SUM
  1494. MD5SUM = md5sum
  1495. endif
  1496. REVISION = -r$(VERSION)
  1497. MAJOR_VERSION = $(word 1,$(subst ., ,$(VERSION)))
  1498. ifeq ($(DEBUG), 1)
  1499. COMMON_OPT += -g
  1500. endif
  1501. ifeq ($(DEBUG), 1)
  1502. FCOMMON_OPT += -g
  1503. endif
  1504. ifndef COMMON_OPT
  1505. COMMON_OPT = -O2
  1506. endif
  1507. ifndef FCOMMON_OPT
  1508. FCOMMON_OPT = -O2 -frecursive
  1509. endif
  1510. override CFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
  1511. override PFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
  1512. override FFLAGS += $(COMMON_OPT) $(FCOMMON_OPT)
  1513. override FPFLAGS += $(FCOMMON_OPT) $(COMMON_PROF)
  1514. #MAKEOVERRIDES =
  1515. ifeq ($(NEED_PIC), 1)
  1516. ifeq (,$(findstring PIC,$(FFLAGS)))
  1517. ifneq ($(F_COMPILER),IBM)
  1518. override FFLAGS += -fPIC
  1519. endif
  1520. endif
  1521. endif
  1522. #For LAPACK Fortran codes.
  1523. #Disable -fopenmp for LAPACK Fortran codes on Windows.
  1524. ifdef OS_WINDOWS
  1525. LAPACK_FFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FFLAGS))
  1526. LAPACK_FPFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FPFLAGS))
  1527. else
  1528. LAPACK_FFLAGS := $(FFLAGS)
  1529. LAPACK_FPFLAGS := $(FPFLAGS)
  1530. endif
  1531. ifeq ($(F_COMPILER),NAG)
  1532. LAPACK_FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1533. override FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1534. endif
  1535. ifeq ($(F_COMPILER),CRAY)
  1536. LAPACK_FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1537. override FFLAGS := $(filter-out -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 ,$(FFLAGS))
  1538. endif
  1539. ifeq ($(F_COMPILER),FLANGNEW)
  1540. LAPACK_FFLAGS := $(filter-out -m32 -m64 -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 -mtune=% -mabi=% ,$(FFLAGS))
  1541. override FFLAGS := $(filter-out -m32 -m64 -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mskylake-avx512 -mtune=% -mabi=% ,$(FFLAGS))
  1542. endif
  1543. LAPACK_CFLAGS = $(CFLAGS)
  1544. LAPACK_CFLAGS += -DHAVE_LAPACK_CONFIG_H
  1545. ifdef INTERFACE64
  1546. ifneq ($(INTERFACE64), 0)
  1547. LAPACK_CFLAGS += -DLAPACK_ILP64
  1548. endif
  1549. endif
  1550. ifdef OS_WINDOWS
  1551. LAPACK_CFLAGS += -DOPENBLAS_OS_WINDOWS
  1552. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1553. endif
  1554. ifeq ($(C_COMPILER), LSB)
  1555. LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
  1556. endif
  1557. ifndef SUFFIX
  1558. SUFFIX = o
  1559. endif
  1560. ifndef PSUFFIX
  1561. PSUFFIX = po
  1562. endif
  1563. ifndef LIBSUFFIX
  1564. LIBSUFFIX = a
  1565. endif
  1566. ifneq ($(DYNAMIC_ARCH), 1)
  1567. ifndef SMP
  1568. LIBNAME = $(LIBPREFIX)_$(LIBCORE)$(REVISION).$(LIBSUFFIX)
  1569. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)$(REVISION)_p.$(LIBSUFFIX)
  1570. else
  1571. LIBNAME = $(LIBPREFIX)_$(LIBCORE)p$(REVISION).$(LIBSUFFIX)
  1572. LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)p$(REVISION)_p.$(LIBSUFFIX)
  1573. endif
  1574. else
  1575. ifndef SMP
  1576. LIBNAME = $(LIBPREFIX)$(REVISION).$(LIBSUFFIX)
  1577. LIBNAME_P = $(LIBPREFIX)$(REVISION)_p.$(LIBSUFFIX)
  1578. else
  1579. LIBNAME = $(LIBPREFIX)p$(REVISION).$(LIBSUFFIX)
  1580. LIBNAME_P = $(LIBPREFIX)p$(REVISION)_p.$(LIBSUFFIX)
  1581. endif
  1582. endif
  1583. ifeq ($(FIXED_LIBNAME),1)
  1584. LIBNAME = lib$(LIBNAMEPREFIX)$(LIBSONAMEBASE)$(LIBNAMESUFFIX).$(LIBSUFFIX)
  1585. LIBNAME_P = lib$(LIBNAMEPREFIX)$(LISOBNAMEBASE)$(LIBNAMESUFFIX)_p.$(LIBSUFFIX)
  1586. endif
  1587. LIBDLLNAME = $(LIBPREFIX).dll
  1588. IMPLIBNAME = lib$(LIBNAMEBASE).dll.a
  1589. LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.so)
  1590. LIBDYNNAME = $(LIBNAME:.$(LIBSUFFIX)=.dylib)
  1591. LIBDEFNAME = $(LIBNAME:.$(LIBSUFFIX)=.def)
  1592. LIBEXPNAME = $(LIBNAME:.$(LIBSUFFIX)=.exp)
  1593. LIBZIPNAME = $(LIBNAME:.$(LIBSUFFIX)=.zip)
  1594. LIBS = $(TOPDIR)/$(LIBNAME)
  1595. LIBS_P = $(TOPDIR)/$(LIBNAME_P)
  1596. LIB_COMPONENTS = BLAS
  1597. ifneq ($(NO_CBLAS), 1)
  1598. LIB_COMPONENTS += CBLAS
  1599. endif
  1600. ifneq ($(NO_LAPACK), 1)
  1601. LIB_COMPONENTS += LAPACK
  1602. ifneq ($(NO_LAPACKE), 1)
  1603. LIB_COMPONENTS += LAPACKE
  1604. endif
  1605. ifeq ($(BUILD_RELAPACK), 1)
  1606. LIB_COMPONENTS += ReLAPACK
  1607. endif
  1608. endif
  1609. ifeq ($(ONLY_CBLAS), 1)
  1610. LIB_COMPONENTS = CBLAS
  1611. endif
  1612. export OSNAME
  1613. export ARCH
  1614. export CORE
  1615. export LIBCORE
  1616. export __BYTE_ORDER__
  1617. export ELF_VERSION
  1618. export PGCPATH
  1619. export CONFIG
  1620. export CC
  1621. export FC
  1622. export BU
  1623. export FU
  1624. export NEED2UNDERSCORES
  1625. export USE_THREAD
  1626. export NUM_THREADS
  1627. export NUM_CORES
  1628. export SMP
  1629. export MAKEFILE_RULE
  1630. export NEED_PIC
  1631. export BINARY
  1632. export BINARY32
  1633. export BINARY64
  1634. export F_COMPILER
  1635. export C_COMPILER
  1636. export USE_OPENMP
  1637. export CROSS
  1638. export CROSS_SUFFIX
  1639. export NOFORTRAN
  1640. export C_LAPACK
  1641. export NO_FBLAS
  1642. export EXTRALIB
  1643. export CEXTRALIB
  1644. export FEXTRALIB
  1645. export HAVE_SSE
  1646. export HAVE_SSE2
  1647. export HAVE_SSE3
  1648. export HAVE_SSSE3
  1649. export HAVE_SSE4_1
  1650. export HAVE_SSE4_2
  1651. export HAVE_SSE4A
  1652. export HAVE_SSE5
  1653. export HAVE_AVX
  1654. export HAVE_AVX2
  1655. export HAVE_FMA3
  1656. export HAVE_VFP
  1657. export HAVE_VFPV3
  1658. export HAVE_VFPV4
  1659. export HAVE_NEON
  1660. ifndef NO_MSA
  1661. export HAVE_MSA
  1662. export MSA_FLAGS
  1663. endif
  1664. export KERNELDIR
  1665. export FUNCTION_PROFILE
  1666. export TARGET_CORE
  1667. export NO_AVX512
  1668. export NO_AVX2
  1669. export BUILD_BFLOAT16
  1670. export BUILD_HFLOAT16
  1671. export NO_LSX
  1672. export NO_LASX
  1673. export BGEMM_UNROLL_M
  1674. export BGEMM_UNROLL_N
  1675. export SBGEMM_UNROLL_M
  1676. export SBGEMM_UNROLL_N
  1677. export SHGEMM_UNROLL_M
  1678. export SHGEMM_UNROLL_N
  1679. export SGEMM_UNROLL_M
  1680. export SGEMM_UNROLL_N
  1681. export DGEMM_UNROLL_M
  1682. export DGEMM_UNROLL_N
  1683. export QGEMM_UNROLL_M
  1684. export QGEMM_UNROLL_N
  1685. export CGEMM_UNROLL_M
  1686. export CGEMM_UNROLL_N
  1687. export ZGEMM_UNROLL_M
  1688. export ZGEMM_UNROLL_N
  1689. export XGEMM_UNROLL_M
  1690. export XGEMM_UNROLL_N
  1691. export CGEMM3M_UNROLL_M
  1692. export CGEMM3M_UNROLL_N
  1693. export ZGEMM3M_UNROLL_M
  1694. export ZGEMM3M_UNROLL_N
  1695. export XGEMM3M_UNROLL_M
  1696. export XGEMM3M_UNROLL_N
  1697. ifdef USE_CUDA
  1698. export CUDADIR
  1699. export CUCC
  1700. export CUFLAGS
  1701. export CULIB
  1702. endif
  1703. .SUFFIXES: .$(PSUFFIX) .$(SUFFIX) .f
  1704. .f.$(SUFFIX):
  1705. $(FC) $(FFLAGS) -c $< -o $(@F)
  1706. .f.$(PSUFFIX):
  1707. $(FC) $(FPFLAGS) -pg -c $< -o $(@F)
  1708. ifdef BINARY64
  1709. PATHSCALEPATH = /opt/pathscale/lib/3.1
  1710. PGIPATH = /opt/pgi/linux86-64/7.1-5/lib
  1711. else
  1712. PATHSCALEPATH = /opt/pathscale/lib/3.1/32
  1713. PGIPATH = /opt/pgi/linux86/7.1-5/lib
  1714. endif
  1715. ACMLPATH = /opt/acml/4.3.0
  1716. ifneq ($(OSNAME), Darwin)
  1717. MKLPATH = /opt/intel/mkl/10.2.2.025/lib
  1718. else
  1719. MKLPATH = /Library/Frameworks/Intel_MKL.framework/Versions/10.0.1.014/lib
  1720. endif
  1721. ATLASPATH = /opt/atlas/3.9.17/opteron
  1722. FLAMEPATH = $(HOME)/flame/lib
  1723. ifneq ($(OSNAME), SunOS)
  1724. SUNPATH = /opt/sunstudio12.1
  1725. else
  1726. SUNPATH = /opt/SUNWspro
  1727. endif