2 # Include user definition
5 # TO suppress recursive includes
12 NETLIB_LAPACK_DIR = $(TOPDIR)/lapack-netlib
15 # - Only set if not specified on the command line or inherited from the environment.
16 # - CC is an implicit variable so neither '?=' or 'ifndef' can be used.
17 # http://stackoverflow.com/questions/4029274/mingw-and-make-variables
18 # - Default value is 'cc' which is not always a valid command (e.g. MinGW).
19 ifeq ($(origin CC),default)
21 # Change the default compile to clang on Mac OSX.
22 # http://stackoverflow.com/questions/714100/os-detecting-makefile
23 UNAME_S := $(shell uname -s)
24 ifeq ($(UNAME_S),Darwin)
29 # Default Fortran compiler (FC) is selected by f_check.
32 include $(TOPDIR)/Makefile.rule
34 include $(TOPDIR)/$(MAKEFILE_RULE)
38 # Beginning of system configuration
46 GETARCH_FLAGS := -DFORCE_$(TARGET)
49 #TARGET_CORE will override TARGET which is used in DYNAMIC_ARCH=1.
52 GETARCH_FLAGS := -DFORCE_$(TARGET_CORE)
56 GETARCH_FLAGS += -DUSE64BITINT
59 ifndef GEMM_MULTITHREAD_THRESHOLD
60 GEMM_MULTITHREAD_THRESHOLD=4
62 GETARCH_FLAGS += -DGEMM_MULTITHREAD_THRESHOLD=$(GEMM_MULTITHREAD_THRESHOLD)
65 GETARCH_FLAGS += -DNO_AVX
72 ifeq ($(QUIET_MAKE), 1)
76 ifndef NO_PARALLEL_MAKE
79 GETARCH_FLAGS += -DNO_PARALLEL_MAKE=$(NO_PARALLEL_MAKE)
81 ifeq ($(HOSTCC), loongcc)
82 GETARCH_FLAGS += -static
85 #if don't use Fortran, it will only compile CBLAS.
86 ifeq ($(ONLY_CBLAS), 1)
92 # This operation is expensive, so execution should be once.
93 ifndef GOTOBLAS_MAKEFILE
94 export GOTOBLAS_MAKEFILE = 1
96 # Generating Makefile.conf and config.h
97 DUMMY := $(shell $(MAKE) -C $(TOPDIR) -f Makefile.prebuild CC="$(CC)" FC="$(FC)" HOSTCC="$(HOSTCC)" CFLAGS="$(GETARCH_FLAGS)" BINARY=$(BINARY) USE_OPENMP=$(USE_OPENMP) TARGET_CORE=$(TARGET_CORE) ONLY_CBLAS=$(ONLY_CBLAS) all)
100 include $(TOPDIR)/Makefile.conf
102 include $(TOPDIR)/Makefile_kernel.conf
108 NUM_THREADS = $(NUM_CORES)
111 ifeq ($(NUM_THREADS), 1)
112 override USE_THREAD = 0
116 ifeq ($(USE_THREAD), 0)
122 ifeq ($(NUM_THREAD), 1)
135 AR = $(CROSS_SUFFIX)ar
136 AS = $(CROSS_SUFFIX)as
137 LD = $(CROSS_SUFFIX)ld
138 RANLIB = $(CROSS_SUFFIX)ranlib
139 NM = $(CROSS_SUFFIX)nm
140 DLLWRAP = $(CROSS_SUFFIX)dllwrap
143 # OS dependent settings
146 ifeq ($(OSNAME), Darwin)
147 export MACOSX_DEPLOYMENT_TARGET=10.2
151 ifeq ($(OSNAME), FreeBSD)
155 ifeq ($(OSNAME), NetBSD)
159 ifeq ($(OSNAME), Linux)
163 ifeq ($(OSNAME), AIX)
167 ifeq ($(OSNAME), WINNT)
171 EXTRALIB += -defaultlib:advapi32
177 ifeq ($(C_COMPILER), CLANG)
178 CCOMMON_OPT += -DMS_ABI
181 ifeq ($(C_COMPILER), GCC)
182 #Test for supporting MS_ABI
183 GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
184 GCCVERSIONGT4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 4)
185 GCCMINORVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f2 -d.` \>= 7)
186 ifeq ($(GCCVERSIONGT4), 1)
187 # GCC Majar version > 4
188 # It is compatible with MSVC ABI.
189 CCOMMON_OPT += -DMS_ABI
192 ifeq ($(GCCVERSIONGTEQ4), 1)
193 ifeq ($(GCCMINORVERSIONGTEQ7), 1)
195 # It is compatible with MSVC ABI.
196 CCOMMON_OPT += -DMS_ABI
201 # Ensure the correct stack alignment on Win32
202 # http://permalink.gmane.org/gmane.comp.lib.openblas.general/97
204 CCOMMON_OPT += -mincoming-stack-boundary=2
205 FCOMMON_OPT += -mincoming-stack-boundary=2
210 ifeq ($(OSNAME), Interix)
214 INTERIX_TOOL_DIR = /opt/gcc.3.3/i586-pc-interix3/bin
217 ifeq ($(OSNAME), CYGWIN_NT)
222 ifneq ($(OSNAME), WINNT)
223 ifneq ($(OSNAME), CYGWIN_NT)
224 ifneq ($(OSNAME), Interix)
226 EXTRALIB += -lpthread
233 ifeq ($(OSNAME), $(filter $(OSNAME),WINNT CYGWIN_NT Interix))
238 CCOMMON_OPT += -DQUAD_PRECISION
243 ifneq ($(ARCH), x86_64)
249 CCOMMON_OPT += -DUTEST_CHECK
254 CCOMMON_OPT += -DSANITY_CHECK -DREFNAME=$(*F)f$(BU)
258 # Architecture dependent settings
265 ifndef NO_EXPRECISION
266 ifeq ($(F_COMPILER), GFORTRAN)
267 # ifeq logical or. GCC or LSB
268 ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
270 CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
271 FCOMMON_OPT += -m128bit-long-double
273 ifeq ($(C_COMPILER), CLANG)
275 CCOMMON_OPT += -DEXPRECISION
276 FCOMMON_OPT += -m128bit-long-double
282 ifeq ($(ARCH), x86_64)
283 ifndef NO_EXPRECISION
284 ifeq ($(F_COMPILER), GFORTRAN)
285 # ifeq logical or. GCC or LSB
286 ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
288 CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
289 FCOMMON_OPT += -m128bit-long-double
291 ifeq ($(C_COMPILER), CLANG)
293 CCOMMON_OPT += -DEXPRECISION
294 FCOMMON_OPT += -m128bit-long-double
300 ifeq ($(C_COMPILER), INTEL)
301 CCOMMON_OPT += -wd981
304 ifeq ($(USE_OPENMP), 1)
305 # ifeq logical or. GCC or LSB
306 ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
307 CCOMMON_OPT += -fopenmp
310 ifeq ($(C_COMPILER), CLANG)
311 $(error OpenBLAS: Clang didn't support OpenMP yet.)
312 CCOMMON_OPT += -fopenmp
315 ifeq ($(C_COMPILER), INTEL)
316 CCOMMON_OPT += -openmp
319 ifeq ($(C_COMPILER), PGI)
323 ifeq ($(C_COMPILER), OPEN64)
325 CEXTRALIB += -lstdc++
328 ifeq ($(C_COMPILER), PATHSCALE)
334 ifeq ($(DYNAMIC_ARCH), 1)
336 DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
337 CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
339 DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER
343 ifeq ($(ARCH), x86_64)
344 DYNAMIC_CORE = PRESCOTT CORE2 PENRYN DUNNINGTON NEHALEM OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
346 DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER
359 ifeq ($(F_COMPILER), GFORTRAN)
360 ifeq ($(C_COMPILER), GCC)
362 # CCOMMON_OPT += -DEXPRECISION
367 ifeq ($(ARCH), mips64)
371 ifeq ($(ARCH), alpha)
377 # C Compiler dependent settings
381 # ifeq logical or. GCC or CLANG or LSB
382 # http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
383 ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG LSB))
385 COMMON_PROF += -fno-inline
386 NO_UNINITIALIZED_WARN = -Wno-uninitialized
388 ifeq ($(QUIET_MAKE), 1)
389 CCOMMON_OPT += $(NO_UNINITIALIZED_WARN) -Wno-unused
394 ifeq ($(ARCH), mips64)
396 CCOMMON_OPT += -mabi=64
398 CCOMMON_OPT += -mabi=n32
403 ifeq ($(CORE), LOONGSON3A)
404 CCOMMON_OPT += -march=mips64
405 FCOMMON_OPT += -march=mips64
408 ifeq ($(CORE), LOONGSON3B)
409 CCOMMON_OPT += -march=mips64
410 FCOMMON_OPT += -march=mips64
413 ifeq ($(OSNAME), AIX)
419 ifndef BINARY_DEFINED
429 ifeq ($(C_COMPILER), PGI)
431 CCOMMON_OPT += -tp p7-64
433 CCOMMON_OPT += -tp p7
437 ifeq ($(C_COMPILER), PATHSCALE)
446 # Fortran Compiler dependent settings
449 ifeq ($(F_COMPILER), G77)
450 CCOMMON_OPT += -DF_INTERFACE_G77
452 ifndef NO_BINARY_MODE
461 ifeq ($(F_COMPILER), G95)
462 CCOMMON_OPT += -DF_INTERFACE_G95
464 ifndef NO_BINARY_MODE
473 ifeq ($(F_COMPILER), GFORTRAN)
474 CCOMMON_OPT += -DF_INTERFACE_GFORT
476 #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
477 ifneq ($(NO_LAPACK), 1)
478 EXTRALIB += -lgfortran
481 ifeq ($(ARCH), mips64)
483 FCOMMON_OPT += -mabi=64
485 FCOMMON_OPT += -mabi=n32
492 FCOMMON_OPT += -fdefault-integer-8
499 FCOMMON_OPT += -fopenmp
503 ifeq ($(F_COMPILER), INTEL)
504 CCOMMON_OPT += -DF_INTERFACE_INTEL
509 FCOMMON_OPT += -openmp
513 ifeq ($(F_COMPILER), FUJITSU)
514 CCOMMON_OPT += -DF_INTERFACE_FUJITSU
516 FCOMMON_OPT += -openmp
520 ifeq ($(F_COMPILER), IBM)
521 CCOMMON_OPT += -DF_INTERFACE_IBM
522 # FCOMMON_OPT += -qarch=440
526 FCOMMON_OPT += -qintsize=8
532 FCOMMON_OPT += -openmp
536 ifeq ($(F_COMPILER), PGI)
537 CCOMMON_OPT += -DF_INTERFACE_PGI
538 COMMON_PROF += -DPGICOMPILER
543 FCOMMON_OPT += -tp p7-64
545 FCOMMON_OPT += -tp p7
552 ifeq ($(F_COMPILER), PATHSCALE)
553 CCOMMON_OPT += -DF_INTERFACE_PATHSCALE
560 ifneq ($(ARCH), mips64)
568 FCOMMON_OPT += -mabi=64
570 FCOMMON_OPT += -mabi=n32
579 ifeq ($(F_COMPILER), OPEN64)
580 CCOMMON_OPT += -DF_INTERFACE_OPEN64
587 ifeq ($(ARCH), mips64)
593 ifeq ($(CORE), LOONGSON3A)
594 FCOMMON_OPT += -loongson3 -static
597 ifeq ($(CORE), LOONGSON3B)
598 FCOMMON_OPT += -loongson3 -static
610 FEXTRALIB += -lstdc++
615 ifeq ($(C_COMPILER), OPEN64)
617 ifeq ($(ARCH), mips64)
623 ifeq ($(CORE), LOONGSON3A)
624 CCOMMON_OPT += -loongson3 -static
627 ifeq ($(CORE), LOONGSON3B)
628 CCOMMON_OPT += -loongson3 -static
641 ifeq ($(C_COMPILER), SUN)
650 ifeq ($(F_COMPILER), SUN)
651 CCOMMON_OPT += -DF_INTERFACE_SUN
658 FCOMMON_OPT += -xopenmp=parallel
662 ifeq ($(F_COMPILER), COMPAQ)
663 CCOMMON_OPT += -DF_INTERFACE_COMPAQ
665 FCOMMON_OPT += -openmp
676 ifeq ($(NEED_PIC), 1)
677 ifeq ($(C_COMPILER), IBM)
678 CCOMMON_OPT += -qpic=large
682 ifeq ($(F_COMPILER), SUN)
689 ifeq ($(DYNAMIC_ARCH), 1)
690 CCOMMON_OPT += -DDYNAMIC_ARCH
693 ifeq ($(NO_LAPACK), 1)
694 CCOMMON_OPT += -DNO_LAPACK
695 #Disable LAPACK C interface
699 ifeq ($(NO_LAPACKE), 1)
700 CCOMMON_OPT += -DNO_LAPACKE
704 CCOMMON_OPT += -DNO_AVX
708 CCOMMON_OPT += -DSMP_SERVER
710 ifeq ($(ARCH), mips64)
711 ifneq ($(CORE), LOONGSON3B)
712 USE_SIMPLE_THREADED_LEVEL3 = 1
716 ifeq ($(USE_OPENMP), 1)
717 # USE_SIMPLE_THREADED_LEVEL3 = 1
719 CCOMMON_OPT += -DUSE_OPENMP
724 ifeq ($(NO_WARMUP), 1)
725 CCOMMON_OPT += -DNO_WARMUP
728 ifeq ($(CONSISTENT_FPCSR), 1)
729 CCOMMON_OPT += -DCONSISTENT_FPCSR
732 # Only for development
733 # CCOMMON_OPT += -DPARAMTEST
734 # CCOMMON_OPT += -DPREFETCHTEST
735 # CCOMMON_OPT += -DNO_SWITCHING
739 CCOMMON_OPT += -DUSE_PAPI
740 EXTRALIB += -lpapi -lperfctr
743 ifdef DYNAMIC_THREADS
744 CCOMMON_OPT += -DDYNAMIC_THREADS
747 CCOMMON_OPT += -DMAX_CPU_NUMBER=$(NUM_THREADS)
749 ifdef USE_SIMPLE_THREADED_LEVEL3
750 CCOMMON_OPT += -DUSE_SIMPLE_THREADED_LEVEL3
754 LIBPREFIX = libopenblas
756 LIBPREFIX = libopenblas_$(LIBNAMESUFFIX)
759 KERNELDIR = $(TOPDIR)/kernel/$(ARCH)
761 include $(TOPDIR)/Makefile.$(ARCH)
763 CCOMMON_OPT += -DASMNAME=$(FU)$(*F) -DASMFNAME=$(FU)$(*F)$(BU) -DNAME=$(*F)$(BU) -DCNAME=$(*F) -DCHAR_NAME=\"$(*F)$(BU)\" -DCHAR_CNAME=\"$(*F)\"
765 ifeq ($(CORE), PPC440)
766 CCOMMON_OPT += -DALLOC_QALLOC
769 ifeq ($(CORE), PPC440FP2)
770 STATIC_ALLOCATION = 1
773 ifneq ($(OSNAME), Linux)
777 ifneq ($(ARCH), x86_64)
779 ifneq ($(CORE), LOONGSON3B)
786 CCOMMON_OPT += -DNO_AFFINITY
789 ifdef FUNCTION_PROFILE
790 CCOMMON_OPT += -DFUNCTION_PROFILE
793 ifdef HUGETLB_ALLOCATION
794 CCOMMON_OPT += -DALLOC_HUGETLB
797 ifdef HUGETLBFILE_ALLOCATION
798 CCOMMON_OPT += -DALLOC_HUGETLBFILE -DHUGETLB_FILE_NAME=$(HUGETLBFILE_ALLOCATION)
801 ifdef STATIC_ALLOCATION
802 CCOMMON_OPT += -DALLOC_STATIC
805 ifdef DEVICEDRIVER_ALLOCATION
806 CCOMMON_OPT += -DALLOC_DEVICEDRIVER -DDEVICEDRIVER_NAME=\"/dev/mapper\"
809 ifdef MIXED_MEMORY_ALLOCATION
810 CCOMMON_OPT += -DMIXED_MEMORY_ALLOCATION
813 ifeq ($(OSNAME), SunOS)
829 REVISION = -r$(VERSION)
830 MAJOR_VERSION = $(word 1,$(subst ., ,$(VERSION)))
841 override CFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
842 override PFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
844 override FFLAGS += $(COMMON_OPT) $(FCOMMON_OPT)
845 override FPFLAGS += $(COMMON_OPT) $(FCOMMON_OPT) $(COMMON_PROF)
848 #For LAPACK Fortran codes.
849 LAPACK_FFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FFLAGS))
850 LAPACK_FPFLAGS := $(filter-out -fopenmp -mp -openmp -xopenmp=parallel,$(FPFLAGS))
852 LAPACK_CFLAGS = $(CFLAGS)
853 LAPACK_CFLAGS += -DHAVE_LAPACK_CONFIG_H
855 LAPACK_CFLAGS += -DLAPACK_ILP64
858 LAPACK_CFLAGS += -DOPENBLAS_OS_WINDOWS
860 ifeq ($(C_COMPILER), LSB)
861 LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
876 ifneq ($(DYNAMIC_ARCH), 1)
878 LIBNAME = $(LIBPREFIX)_$(LIBCORE)$(REVISION).$(LIBSUFFIX)
879 LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)$(REVISION)_p.$(LIBSUFFIX)
881 LIBNAME = $(LIBPREFIX)_$(LIBCORE)p$(REVISION).$(LIBSUFFIX)
882 LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)p$(REVISION)_p.$(LIBSUFFIX)
886 LIBNAME = $(LIBPREFIX)$(REVISION).$(LIBSUFFIX)
887 LIBNAME_P = $(LIBPREFIX)$(REVISION)_p.$(LIBSUFFIX)
889 LIBNAME = $(LIBPREFIX)p$(REVISION).$(LIBSUFFIX)
890 LIBNAME_P = $(LIBPREFIX)p$(REVISION)_p.$(LIBSUFFIX)
895 LIBDLLNAME = $(LIBPREFIX).dll
896 LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.so)
897 LIBDYNNAME = $(LIBNAME:.$(LIBSUFFIX)=.dylib)
898 LIBDEFNAME = $(LIBNAME:.$(LIBSUFFIX)=.def)
899 LIBEXPNAME = $(LIBNAME:.$(LIBSUFFIX)=.exp)
900 LIBZIPNAME = $(LIBNAME:.$(LIBSUFFIX)=.zip)
902 LIBS = $(TOPDIR)/$(LIBNAME)
903 LIBS_P = $(TOPDIR)/$(LIBNAME_P)
906 LIB_COMPONENTS = BLAS
907 ifneq ($(NO_CBLAS), 1)
908 LIB_COMPONENTS += CBLAS
911 ifneq ($(NO_LAPACK), 1)
912 LIB_COMPONENTS += LAPACK
913 ifneq ($(NO_LAPACKE), 1)
914 LIB_COMPONENTS += LAPACKE
918 ifeq ($(ONLY_CBLAS), 1)
919 LIB_COMPONENTS = CBLAS
932 export NEED2UNDERSCORES
962 export FUNCTION_PROFILE
965 export SGEMM_UNROLL_M
966 export SGEMM_UNROLL_N
967 export DGEMM_UNROLL_M
968 export DGEMM_UNROLL_N
969 export QGEMM_UNROLL_M
970 export QGEMM_UNROLL_N
971 export CGEMM_UNROLL_M
972 export CGEMM_UNROLL_N
973 export ZGEMM_UNROLL_M
974 export ZGEMM_UNROLL_N
975 export XGEMM_UNROLL_M
976 export XGEMM_UNROLL_N
977 export CGEMM3M_UNROLL_M
978 export CGEMM3M_UNROLL_N
979 export ZGEMM3M_UNROLL_M
980 export ZGEMM3M_UNROLL_N
981 export XGEMM3M_UNROLL_M
982 export XGEMM3M_UNROLL_N
992 .SUFFIXES: .$(PSUFFIX) .$(SUFFIX) .f
995 $(FC) $(FFLAGS) -c $< -o $(@F)
998 $(FC) $(FPFLAGS) -pg -c $< -o $(@F)
1002 PATHSCALEPATH = /opt/pathscale/lib/3.1
1003 PGIPATH = /opt/pgi/linux86-64/7.1-5/lib
1005 PATHSCALEPATH = /opt/pathscale/lib/3.1/32
1006 PGIPATH = /opt/pgi/linux86/7.1-5/lib
1009 ACMLPATH = /opt/acml/4.3.0
1010 ifneq ($(OSNAME), Darwin)
1011 MKLPATH = /opt/intel/mkl/10.2.2.025/lib
1013 MKLPATH = /Library/Frameworks/Intel_MKL.framework/Versions/10.0.1.014/lib
1015 ATLASPATH = /opt/atlas/3.9.17/opteron
1016 FLAMEPATH = $(HOME)/flame/lib
1017 ifneq ($(OSNAME), SunOS)
1018 SUNPATH = /opt/sunstudio12.1
1020 SUNPATH = /opt/SUNWspro