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 CCOMMON_OPT += -DQUAD_PRECISION
238 ifneq ($(ARCH), x86_64)
244 CCOMMON_OPT += -DUTEST_CHECK
249 CCOMMON_OPT += -DSANITY_CHECK -DREFNAME=$(*F)f$(BU)
253 # Architecture dependent settings
260 ifndef NO_EXPRECISION
261 ifeq ($(F_COMPILER), GFORTRAN)
262 # ifeq logical or. GCC or LSB
263 ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
265 CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
266 FCOMMON_OPT += -m128bit-long-double
268 ifeq ($(C_COMPILER), CLANG)
270 CCOMMON_OPT += -DEXPRECISION
271 FCOMMON_OPT += -m128bit-long-double
277 ifeq ($(ARCH), x86_64)
278 ifndef NO_EXPRECISION
279 ifeq ($(F_COMPILER), GFORTRAN)
280 # ifeq logical or. GCC or LSB
281 ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
283 CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
284 FCOMMON_OPT += -m128bit-long-double
286 ifeq ($(C_COMPILER), CLANG)
288 CCOMMON_OPT += -DEXPRECISION
289 FCOMMON_OPT += -m128bit-long-double
295 ifeq ($(C_COMPILER), INTEL)
296 CCOMMON_OPT += -wd981
299 ifeq ($(USE_OPENMP), 1)
300 # ifeq logical or. GCC or LSB
301 ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
302 CCOMMON_OPT += -fopenmp
305 ifeq ($(C_COMPILER), CLANG)
306 $(error OpenBLAS: Clang didn't support OpenMP yet.)
307 CCOMMON_OPT += -fopenmp
310 ifeq ($(C_COMPILER), INTEL)
311 CCOMMON_OPT += -openmp
314 ifeq ($(C_COMPILER), PGI)
318 ifeq ($(C_COMPILER), OPEN64)
320 CEXTRALIB += -lstdc++
323 ifeq ($(C_COMPILER), PATHSCALE)
329 ifeq ($(DYNAMIC_ARCH), 1)
331 DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
332 CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
334 DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER
338 ifeq ($(ARCH), x86_64)
339 DYNAMIC_CORE = PRESCOTT CORE2 PENRYN DUNNINGTON NEHALEM OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
341 DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER
354 ifeq ($(F_COMPILER), GFORTRAN)
355 ifeq ($(C_COMPILER), GCC)
357 # CCOMMON_OPT += -DEXPRECISION
362 ifeq ($(ARCH), mips64)
366 ifeq ($(ARCH), alpha)
372 # C Compiler dependent settings
376 # ifeq logical or. GCC or CLANG or LSB
377 # http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
378 ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG LSB))
380 COMMON_PROF += -fno-inline
381 NO_UNINITIALIZED_WARN = -Wno-uninitialized
383 ifeq ($(QUIET_MAKE), 1)
384 CCOMMON_OPT += $(NO_UNINITIALIZED_WARN) -Wno-unused
389 ifeq ($(ARCH), mips64)
391 CCOMMON_OPT += -mabi=64
393 CCOMMON_OPT += -mabi=n32
398 ifeq ($(CORE), LOONGSON3A)
399 CCOMMON_OPT += -march=mips64
400 FCOMMON_OPT += -march=mips64
403 ifeq ($(CORE), LOONGSON3B)
404 CCOMMON_OPT += -march=mips64
405 FCOMMON_OPT += -march=mips64
408 ifeq ($(OSNAME), AIX)
414 ifndef BINARY_DEFINED
424 ifeq ($(C_COMPILER), PGI)
426 CCOMMON_OPT += -tp p7-64
428 CCOMMON_OPT += -tp p7
432 ifeq ($(C_COMPILER), PATHSCALE)
441 # Fortran Compiler dependent settings
444 ifeq ($(F_COMPILER), G77)
445 CCOMMON_OPT += -DF_INTERFACE_G77
447 ifndef NO_BINARY_MODE
456 ifeq ($(F_COMPILER), G95)
457 CCOMMON_OPT += -DF_INTERFACE_G95
459 ifndef NO_BINARY_MODE
468 ifeq ($(F_COMPILER), GFORTRAN)
469 CCOMMON_OPT += -DF_INTERFACE_GFORT
471 #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
472 ifneq ($(NO_LAPACK), 1)
473 EXTRALIB += -lgfortran
476 ifeq ($(ARCH), mips64)
478 FCOMMON_OPT += -mabi=64
480 FCOMMON_OPT += -mabi=n32
487 FCOMMON_OPT += -fdefault-integer-8
494 FCOMMON_OPT += -fopenmp
498 ifeq ($(F_COMPILER), INTEL)
499 CCOMMON_OPT += -DF_INTERFACE_INTEL
504 FCOMMON_OPT += -openmp
508 ifeq ($(F_COMPILER), FUJITSU)
509 CCOMMON_OPT += -DF_INTERFACE_FUJITSU
511 FCOMMON_OPT += -openmp
515 ifeq ($(F_COMPILER), IBM)
516 CCOMMON_OPT += -DF_INTERFACE_IBM
517 # FCOMMON_OPT += -qarch=440
521 FCOMMON_OPT += -qintsize=8
527 FCOMMON_OPT += -openmp
531 ifeq ($(F_COMPILER), PGI)
532 CCOMMON_OPT += -DF_INTERFACE_PGI
533 COMMON_PROF += -DPGICOMPILER
538 FCOMMON_OPT += -tp p7-64
540 FCOMMON_OPT += -tp p7
547 ifeq ($(F_COMPILER), PATHSCALE)
548 CCOMMON_OPT += -DF_INTERFACE_PATHSCALE
555 ifneq ($(ARCH), mips64)
563 FCOMMON_OPT += -mabi=64
565 FCOMMON_OPT += -mabi=n32
574 ifeq ($(F_COMPILER), OPEN64)
575 CCOMMON_OPT += -DF_INTERFACE_OPEN64
582 ifeq ($(ARCH), mips64)
588 ifeq ($(CORE), LOONGSON3A)
589 FCOMMON_OPT += -loongson3 -static
592 ifeq ($(CORE), LOONGSON3B)
593 FCOMMON_OPT += -loongson3 -static
605 FEXTRALIB += -lstdc++
610 ifeq ($(C_COMPILER), OPEN64)
612 ifeq ($(ARCH), mips64)
618 ifeq ($(CORE), LOONGSON3A)
619 CCOMMON_OPT += -loongson3 -static
622 ifeq ($(CORE), LOONGSON3B)
623 CCOMMON_OPT += -loongson3 -static
636 ifeq ($(C_COMPILER), SUN)
645 ifeq ($(F_COMPILER), SUN)
646 CCOMMON_OPT += -DF_INTERFACE_SUN
653 FCOMMON_OPT += -xopenmp=parallel
657 ifeq ($(F_COMPILER), COMPAQ)
658 CCOMMON_OPT += -DF_INTERFACE_COMPAQ
660 FCOMMON_OPT += -openmp
671 ifeq ($(NEED_PIC), 1)
672 ifeq ($(C_COMPILER), IBM)
673 CCOMMON_OPT += -qpic=large
677 ifeq ($(F_COMPILER), SUN)
684 ifeq ($(DYNAMIC_ARCH), 1)
685 CCOMMON_OPT += -DDYNAMIC_ARCH
688 ifeq ($(NO_LAPACK), 1)
689 CCOMMON_OPT += -DNO_LAPACK
690 #Disable LAPACK C interface
694 ifeq ($(NO_LAPACKE), 1)
695 CCOMMON_OPT += -DNO_LAPACKE
699 CCOMMON_OPT += -DNO_AVX
703 CCOMMON_OPT += -DSMP_SERVER
705 ifeq ($(ARCH), mips64)
706 ifneq ($(CORE), LOONGSON3B)
707 USE_SIMPLE_THREADED_LEVEL3 = 1
711 ifeq ($(USE_OPENMP), 1)
712 # USE_SIMPLE_THREADED_LEVEL3 = 1
714 CCOMMON_OPT += -DUSE_OPENMP
719 ifeq ($(NO_WARMUP), 1)
720 CCOMMON_OPT += -DNO_WARMUP
723 ifeq ($(CONSISTENT_FPCSR), 1)
724 CCOMMON_OPT += -DCONSISTENT_FPCSR
727 # Only for development
728 # CCOMMON_OPT += -DPARAMTEST
729 # CCOMMON_OPT += -DPREFETCHTEST
730 # CCOMMON_OPT += -DNO_SWITCHING
734 CCOMMON_OPT += -DUSE_PAPI
735 EXTRALIB += -lpapi -lperfctr
738 ifdef DYNAMIC_THREADS
739 CCOMMON_OPT += -DDYNAMIC_THREADS
742 CCOMMON_OPT += -DMAX_CPU_NUMBER=$(NUM_THREADS)
744 ifdef USE_SIMPLE_THREADED_LEVEL3
745 CCOMMON_OPT += -DUSE_SIMPLE_THREADED_LEVEL3
749 LIBPREFIX = libopenblas
751 LIBPREFIX = libopenblas_$(LIBNAMESUFFIX)
754 KERNELDIR = $(TOPDIR)/kernel/$(ARCH)
756 include $(TOPDIR)/Makefile.$(ARCH)
758 CCOMMON_OPT += -DASMNAME=$(FU)$(*F) -DASMFNAME=$(FU)$(*F)$(BU) -DNAME=$(*F)$(BU) -DCNAME=$(*F) -DCHAR_NAME=\"$(*F)$(BU)\" -DCHAR_CNAME=\"$(*F)\"
760 ifeq ($(CORE), PPC440)
761 CCOMMON_OPT += -DALLOC_QALLOC
764 ifeq ($(CORE), PPC440FP2)
765 STATIC_ALLOCATION = 1
768 ifneq ($(OSNAME), Linux)
772 ifneq ($(ARCH), x86_64)
774 ifneq ($(CORE), LOONGSON3B)
781 CCOMMON_OPT += -DNO_AFFINITY
784 ifdef FUNCTION_PROFILE
785 CCOMMON_OPT += -DFUNCTION_PROFILE
788 ifdef HUGETLB_ALLOCATION
789 CCOMMON_OPT += -DALLOC_HUGETLB
792 ifdef HUGETLBFILE_ALLOCATION
793 CCOMMON_OPT += -DALLOC_HUGETLBFILE -DHUGETLB_FILE_NAME=$(HUGETLBFILE_ALLOCATION)
796 ifdef STATIC_ALLOCATION
797 CCOMMON_OPT += -DALLOC_STATIC
800 ifdef DEVICEDRIVER_ALLOCATION
801 CCOMMON_OPT += -DALLOC_DEVICEDRIVER -DDEVICEDRIVER_NAME=\"/dev/mapper\"
804 ifdef MIXED_MEMORY_ALLOCATION
805 CCOMMON_OPT += -DMIXED_MEMORY_ALLOCATION
808 ifeq ($(OSNAME), SunOS)
824 REVISION = -r$(VERSION)
825 MAJOR_VERSION = $(word 1,$(subst ., ,$(VERSION)))
836 override CFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
837 override PFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
839 override FFLAGS += $(COMMON_OPT) $(FCOMMON_OPT)
840 override FPFLAGS += $(COMMON_OPT) $(FCOMMON_OPT) $(COMMON_PROF)
843 LAPACK_CFLAGS = $(CFLAGS)
844 LAPACK_CFLAGS += -DHAVE_LAPACK_CONFIG_H
846 LAPACK_CFLAGS += -DLAPACK_ILP64
848 ifeq ($(C_COMPILER), LSB)
849 LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
864 ifneq ($(DYNAMIC_ARCH), 1)
866 LIBNAME = $(LIBPREFIX)_$(LIBCORE)$(REVISION).$(LIBSUFFIX)
867 LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)$(REVISION)_p.$(LIBSUFFIX)
869 LIBNAME = $(LIBPREFIX)_$(LIBCORE)p$(REVISION).$(LIBSUFFIX)
870 LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)p$(REVISION)_p.$(LIBSUFFIX)
874 LIBNAME = $(LIBPREFIX)$(REVISION).$(LIBSUFFIX)
875 LIBNAME_P = $(LIBPREFIX)$(REVISION)_p.$(LIBSUFFIX)
877 LIBNAME = $(LIBPREFIX)p$(REVISION).$(LIBSUFFIX)
878 LIBNAME_P = $(LIBPREFIX)p$(REVISION)_p.$(LIBSUFFIX)
883 LIBDLLNAME = $(LIBPREFIX).dll
884 LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.so)
885 LIBDYNNAME = $(LIBNAME:.$(LIBSUFFIX)=.dylib)
886 LIBDEFNAME = $(LIBNAME:.$(LIBSUFFIX)=.def)
887 LIBEXPNAME = $(LIBNAME:.$(LIBSUFFIX)=.exp)
888 LIBZIPNAME = $(LIBNAME:.$(LIBSUFFIX)=.zip)
890 LIBS = $(TOPDIR)/$(LIBNAME)
891 LIBS_P = $(TOPDIR)/$(LIBNAME_P)
894 LIB_COMPONENTS = BLAS
895 ifneq ($(NO_CBLAS), 1)
896 LIB_COMPONENTS += CBLAS
899 ifneq ($(NO_LAPACK), 1)
900 LIB_COMPONENTS += LAPACK
901 ifneq ($(NO_LAPACKE), 1)
902 LIB_COMPONENTS += LAPACKE
906 ifeq ($(ONLY_CBLAS), 1)
907 LIB_COMPONENTS = CBLAS
920 export NEED2UNDERSCORES
950 export FUNCTION_PROFILE
953 export SGEMM_UNROLL_M
954 export SGEMM_UNROLL_N
955 export DGEMM_UNROLL_M
956 export DGEMM_UNROLL_N
957 export QGEMM_UNROLL_M
958 export QGEMM_UNROLL_N
959 export CGEMM_UNROLL_M
960 export CGEMM_UNROLL_N
961 export ZGEMM_UNROLL_M
962 export ZGEMM_UNROLL_N
963 export XGEMM_UNROLL_M
964 export XGEMM_UNROLL_N
965 export CGEMM3M_UNROLL_M
966 export CGEMM3M_UNROLL_N
967 export ZGEMM3M_UNROLL_M
968 export ZGEMM3M_UNROLL_N
969 export XGEMM3M_UNROLL_M
970 export XGEMM3M_UNROLL_N
980 .SUFFIXES: .$(PSUFFIX) .$(SUFFIX) .f
983 $(FC) $(FFLAGS) -c $< -o $(@F)
986 $(FC) $(FPFLAGS) -pg -c $< -o $(@F)
990 PATHSCALEPATH = /opt/pathscale/lib/3.1
991 PGIPATH = /opt/pgi/linux86-64/7.1-5/lib
993 PATHSCALEPATH = /opt/pathscale/lib/3.1/32
994 PGIPATH = /opt/pgi/linux86/7.1-5/lib
997 ACMLPATH = /opt/acml/4.3.0
998 ifneq ($(OSNAME), Darwin)
999 MKLPATH = /opt/intel/mkl/10.2.2.025/lib
1001 MKLPATH = /Library/Frameworks/Intel_MKL.framework/Versions/10.0.1.014/lib
1003 ATLASPATH = /opt/atlas/3.9.17/opteron
1004 FLAMEPATH = $(HOME)/flame/lib
1005 ifneq ($(OSNAME), SunOS)
1006 SUNPATH = /opt/sunstudio12.1
1008 SUNPATH = /opt/SUNWspro