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 # This operation is expensive, so execution should be once.
86 ifndef GOTOBLAS_MAKEFILE
87 export GOTOBLAS_MAKEFILE = 1
89 # Generating Makefile.conf and config.h
90 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) all)
93 include $(TOPDIR)/Makefile.conf
95 include $(TOPDIR)/Makefile_kernel.conf
101 NUM_THREADS = $(NUM_CORES)
104 ifeq ($(NUM_THREADS), 1)
105 override USE_THREAD = 0
109 ifeq ($(USE_THREAD), 0)
115 ifeq ($(NUM_THREAD), 1)
128 AR = $(CROSS_SUFFIX)ar
129 AS = $(CROSS_SUFFIX)as
130 LD = $(CROSS_SUFFIX)ld
131 RANLIB = $(CROSS_SUFFIX)ranlib
132 NM = $(CROSS_SUFFIX)nm
133 DLLWRAP = $(CROSS_SUFFIX)dllwrap
136 # OS dependent settings
139 ifeq ($(OSNAME), Darwin)
140 export MACOSX_DEPLOYMENT_TARGET=10.2
144 ifeq ($(OSNAME), FreeBSD)
148 ifeq ($(OSNAME), NetBSD)
152 ifeq ($(OSNAME), Linux)
156 ifeq ($(OSNAME), AIX)
160 ifeq ($(OSNAME), WINNT)
164 EXTRALIB += -defaultlib:advapi32
170 ifeq ($(C_COMPILER), CLANG)
171 CCOMMON_OPT += -DMS_ABI
174 ifeq ($(C_COMPILER), GCC)
175 #Test for supporting MS_ABI
176 GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
177 GCCVERSIONGT4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 4)
178 GCCMINORVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f2 -d.` \>= 7)
179 ifeq ($(GCCVERSIONGT4), 1)
180 # GCC Majar version > 4
181 # It is compatible with MSVC ABI.
182 CCOMMON_OPT += -DMS_ABI
185 ifeq ($(GCCVERSIONGTEQ4), 1)
186 ifeq ($(GCCMINORVERSIONGTEQ7), 1)
188 # It is compatible with MSVC ABI.
189 CCOMMON_OPT += -DMS_ABI
194 # Ensure the correct stack alignment on Win32
195 # http://permalink.gmane.org/gmane.comp.lib.openblas.general/97
197 CCOMMON_OPT += -mincoming-stack-boundary=2
198 FCOMMON_OPT += -mincoming-stack-boundary=2
203 ifeq ($(OSNAME), Interix)
207 INTERIX_TOOL_DIR = /opt/gcc.3.3/i586-pc-interix3/bin
210 ifeq ($(OSNAME), CYGWIN_NT)
215 ifneq ($(OSNAME), WINNT)
216 ifneq ($(OSNAME), CYGWIN_NT)
217 ifneq ($(OSNAME), Interix)
219 EXTRALIB += -lpthread
226 CCOMMON_OPT += -DQUAD_PRECISION
231 ifneq ($(ARCH), x86_64)
237 CCOMMON_OPT += -DUTEST_CHECK
242 CCOMMON_OPT += -DSANITY_CHECK -DREFNAME=$(*F)f$(BU)
246 # Architecture dependent settings
253 ifndef NO_EXPRECISION
254 ifeq ($(F_COMPILER), GFORTRAN)
255 # ifeq logical or. GCC or LSB
256 ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
258 CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
259 FCOMMON_OPT += -m128bit-long-double
261 ifeq ($(C_COMPILER), CLANG)
263 CCOMMON_OPT += -DEXPRECISION
264 FCOMMON_OPT += -m128bit-long-double
270 ifeq ($(ARCH), x86_64)
271 ifndef NO_EXPRECISION
272 ifeq ($(F_COMPILER), GFORTRAN)
273 # ifeq logical or. GCC or LSB
274 ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
276 CCOMMON_OPT += -DEXPRECISION -m128bit-long-double
277 FCOMMON_OPT += -m128bit-long-double
279 ifeq ($(C_COMPILER), CLANG)
281 CCOMMON_OPT += -DEXPRECISION
282 FCOMMON_OPT += -m128bit-long-double
288 ifeq ($(C_COMPILER), INTEL)
289 CCOMMON_OPT += -wd981
292 ifeq ($(USE_OPENMP), 1)
293 # ifeq logical or. GCC or LSB
294 ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB))
295 CCOMMON_OPT += -fopenmp
298 ifeq ($(C_COMPILER), CLANG)
299 $(error OpenBLAS: Clang didn't support OpenMP yet.)
300 CCOMMON_OPT += -fopenmp
303 ifeq ($(C_COMPILER), INTEL)
304 CCOMMON_OPT += -openmp
307 ifeq ($(C_COMPILER), PGI)
311 ifeq ($(C_COMPILER), OPEN64)
313 CEXTRALIB += -lstdc++
316 ifeq ($(C_COMPILER), PATHSCALE)
322 ifeq ($(DYNAMIC_ARCH), 1)
324 DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
325 CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
327 DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER
331 ifeq ($(ARCH), x86_64)
332 DYNAMIC_CORE = PRESCOTT CORE2 PENRYN DUNNINGTON NEHALEM OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
334 DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER
347 ifeq ($(F_COMPILER), GFORTRAN)
348 ifeq ($(C_COMPILER), GCC)
350 # CCOMMON_OPT += -DEXPRECISION
355 ifeq ($(ARCH), mips64)
359 ifeq ($(ARCH), alpha)
365 # C Compiler dependent settings
369 # ifeq logical or. GCC or CLANG or LSB
370 # http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
371 ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG LSB))
373 COMMON_PROF += -fno-inline
374 NO_UNINITIALIZED_WARN = -Wno-uninitialized
376 ifeq ($(QUIET_MAKE), 1)
377 CCOMMON_OPT += $(NO_UNINITIALIZED_WARN) -Wno-unused
382 ifeq ($(ARCH), mips64)
384 CCOMMON_OPT += -mabi=64
386 CCOMMON_OPT += -mabi=n32
391 ifeq ($(CORE), LOONGSON3A)
392 CCOMMON_OPT += -march=mips64
393 FCOMMON_OPT += -march=mips64
396 ifeq ($(CORE), LOONGSON3B)
397 CCOMMON_OPT += -march=mips64
398 FCOMMON_OPT += -march=mips64
401 ifeq ($(OSNAME), AIX)
407 ifndef BINARY_DEFINED
417 ifeq ($(C_COMPILER), PGI)
419 CCOMMON_OPT += -tp p7-64
421 CCOMMON_OPT += -tp p7
425 ifeq ($(C_COMPILER), PATHSCALE)
434 # Fortran Compiler dependent settings
437 ifeq ($(F_COMPILER), G77)
438 CCOMMON_OPT += -DF_INTERFACE_G77
440 ifndef NO_BINARY_MODE
449 ifeq ($(F_COMPILER), G95)
450 CCOMMON_OPT += -DF_INTERFACE_G95
452 ifndef NO_BINARY_MODE
461 ifeq ($(F_COMPILER), GFORTRAN)
462 CCOMMON_OPT += -DF_INTERFACE_GFORT
464 #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
465 ifneq ($(NO_LAPACK), 1)
466 ifneq ($(C_COMPILER), LSB)
467 EXTRALIB += -lgfortran
471 ifeq ($(ARCH), mips64)
473 FCOMMON_OPT += -mabi=64
475 FCOMMON_OPT += -mabi=n32
482 FCOMMON_OPT += -fdefault-integer-8
489 FCOMMON_OPT += -fopenmp
493 ifeq ($(F_COMPILER), INTEL)
494 CCOMMON_OPT += -DF_INTERFACE_INTEL
499 FCOMMON_OPT += -openmp
503 ifeq ($(F_COMPILER), FUJITSU)
504 CCOMMON_OPT += -DF_INTERFACE_FUJITSU
506 FCOMMON_OPT += -openmp
510 ifeq ($(F_COMPILER), IBM)
511 CCOMMON_OPT += -DF_INTERFACE_IBM
512 # FCOMMON_OPT += -qarch=440
516 FCOMMON_OPT += -qintsize=8
522 FCOMMON_OPT += -openmp
526 ifeq ($(F_COMPILER), PGI)
527 CCOMMON_OPT += -DF_INTERFACE_PGI
528 COMMON_PROF += -DPGICOMPILER
533 FCOMMON_OPT += -tp p7-64
535 FCOMMON_OPT += -tp p7
542 ifeq ($(F_COMPILER), PATHSCALE)
543 CCOMMON_OPT += -DF_INTERFACE_PATHSCALE
550 ifneq ($(ARCH), mips64)
558 FCOMMON_OPT += -mabi=64
560 FCOMMON_OPT += -mabi=n32
569 ifeq ($(F_COMPILER), OPEN64)
570 CCOMMON_OPT += -DF_INTERFACE_OPEN64
577 ifeq ($(ARCH), mips64)
583 ifeq ($(CORE), LOONGSON3A)
584 FCOMMON_OPT += -loongson3 -static
587 ifeq ($(CORE), LOONGSON3B)
588 FCOMMON_OPT += -loongson3 -static
600 FEXTRALIB += -lstdc++
605 ifeq ($(C_COMPILER), OPEN64)
607 ifeq ($(ARCH), mips64)
613 ifeq ($(CORE), LOONGSON3A)
614 CCOMMON_OPT += -loongson3 -static
617 ifeq ($(CORE), LOONGSON3B)
618 CCOMMON_OPT += -loongson3 -static
631 ifeq ($(C_COMPILER), SUN)
640 ifeq ($(F_COMPILER), SUN)
641 CCOMMON_OPT += -DF_INTERFACE_SUN
648 FCOMMON_OPT += -xopenmp=parallel
652 ifeq ($(F_COMPILER), COMPAQ)
653 CCOMMON_OPT += -DF_INTERFACE_COMPAQ
655 FCOMMON_OPT += -openmp
666 ifeq ($(NEED_PIC), 1)
667 ifeq ($(C_COMPILER), IBM)
668 CCOMMON_OPT += -qpic=large
672 ifeq ($(F_COMPILER), SUN)
679 ifeq ($(DYNAMIC_ARCH), 1)
680 CCOMMON_OPT += -DDYNAMIC_ARCH
683 ifeq ($(NO_LAPACK), 1)
684 CCOMMON_OPT += -DNO_LAPACK
685 #Disable LAPACK C interface
689 ifeq ($(NO_LAPACKE), 1)
690 CCOMMON_OPT += -DNO_LAPACKE
694 CCOMMON_OPT += -DNO_AVX
698 CCOMMON_OPT += -DSMP_SERVER
700 ifeq ($(ARCH), mips64)
701 ifneq ($(CORE), LOONGSON3B)
702 USE_SIMPLE_THREADED_LEVEL3 = 1
706 ifeq ($(USE_OPENMP), 1)
707 # USE_SIMPLE_THREADED_LEVEL3 = 1
709 CCOMMON_OPT += -DUSE_OPENMP
714 ifeq ($(NO_WARMUP), 1)
715 CCOMMON_OPT += -DNO_WARMUP
718 ifeq ($(CONSISTENT_FPCSR), 1)
719 CCOMMON_OPT += -DCONSISTENT_FPCSR
722 # Only for development
723 # CCOMMON_OPT += -DPARAMTEST
724 # CCOMMON_OPT += -DPREFETCHTEST
725 # CCOMMON_OPT += -DNO_SWITCHING
729 CCOMMON_OPT += -DUSE_PAPI
730 EXTRALIB += -lpapi -lperfctr
733 ifdef DYNAMIC_THREADS
734 CCOMMON_OPT += -DDYNAMIC_THREADS
737 CCOMMON_OPT += -DMAX_CPU_NUMBER=$(NUM_THREADS)
739 ifdef USE_SIMPLE_THREADED_LEVEL3
740 CCOMMON_OPT += -DUSE_SIMPLE_THREADED_LEVEL3
744 LIBPREFIX = libopenblas
746 LIBPREFIX = libopenblas_$(LIBNAMESUFFIX)
749 KERNELDIR = $(TOPDIR)/kernel/$(ARCH)
751 include $(TOPDIR)/Makefile.$(ARCH)
753 CCOMMON_OPT += -DASMNAME=$(FU)$(*F) -DASMFNAME=$(FU)$(*F)$(BU) -DNAME=$(*F)$(BU) -DCNAME=$(*F) -DCHAR_NAME=\"$(*F)$(BU)\" -DCHAR_CNAME=\"$(*F)\"
755 ifeq ($(CORE), PPC440)
756 CCOMMON_OPT += -DALLOC_QALLOC
759 ifeq ($(CORE), PPC440FP2)
760 STATIC_ALLOCATION = 1
763 ifneq ($(OSNAME), Linux)
767 ifneq ($(ARCH), x86_64)
769 ifneq ($(CORE), LOONGSON3B)
776 CCOMMON_OPT += -DNO_AFFINITY
779 ifdef FUNCTION_PROFILE
780 CCOMMON_OPT += -DFUNCTION_PROFILE
783 ifdef HUGETLB_ALLOCATION
784 CCOMMON_OPT += -DALLOC_HUGETLB
787 ifdef HUGETLBFILE_ALLOCATION
788 CCOMMON_OPT += -DALLOC_HUGETLBFILE -DHUGETLB_FILE_NAME=$(HUGETLBFILE_ALLOCATION)
791 ifdef STATIC_ALLOCATION
792 CCOMMON_OPT += -DALLOC_STATIC
795 ifdef DEVICEDRIVER_ALLOCATION
796 CCOMMON_OPT += -DALLOC_DEVICEDRIVER -DDEVICEDRIVER_NAME=\"/dev/mapper\"
799 ifdef MIXED_MEMORY_ALLOCATION
800 CCOMMON_OPT += -DMIXED_MEMORY_ALLOCATION
803 ifeq ($(OSNAME), SunOS)
819 REVISION = -r$(VERSION)
820 MAJOR_VERSION = $(word 1,$(subst ., ,$(VERSION)))
831 override CFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
832 override PFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
834 override FFLAGS += $(COMMON_OPT) $(FCOMMON_OPT)
835 override FPFLAGS += $(COMMON_OPT) $(FCOMMON_OPT) $(COMMON_PROF)
838 LAPACK_CFLAGS = $(CFLAGS)
839 LAPACK_CFLAGS += -DHAVE_LAPACK_CONFIG_H
841 LAPACK_CFLAGS += -DLAPACK_ILP64
843 ifeq ($(C_COMPILER), LSB)
844 LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
859 ifneq ($(DYNAMIC_ARCH), 1)
861 LIBNAME = $(LIBPREFIX)_$(LIBCORE)$(REVISION).$(LIBSUFFIX)
862 LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)$(REVISION)_p.$(LIBSUFFIX)
864 LIBNAME = $(LIBPREFIX)_$(LIBCORE)p$(REVISION).$(LIBSUFFIX)
865 LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)p$(REVISION)_p.$(LIBSUFFIX)
869 LIBNAME = $(LIBPREFIX)$(REVISION).$(LIBSUFFIX)
870 LIBNAME_P = $(LIBPREFIX)$(REVISION)_p.$(LIBSUFFIX)
872 LIBNAME = $(LIBPREFIX)p$(REVISION).$(LIBSUFFIX)
873 LIBNAME_P = $(LIBPREFIX)p$(REVISION)_p.$(LIBSUFFIX)
878 LIBDLLNAME = $(LIBPREFIX).dll
879 LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.so)
880 LIBDYNNAME = $(LIBNAME:.$(LIBSUFFIX)=.dylib)
881 LIBDEFNAME = $(LIBNAME:.$(LIBSUFFIX)=.def)
882 LIBEXPNAME = $(LIBNAME:.$(LIBSUFFIX)=.exp)
883 LIBZIPNAME = $(LIBNAME:.$(LIBSUFFIX)=.zip)
885 LIBS = $(TOPDIR)/$(LIBNAME)
886 LIBS_P = $(TOPDIR)/$(LIBNAME_P)
926 export FUNCTION_PROFILE
929 export SGEMM_UNROLL_M
930 export SGEMM_UNROLL_N
931 export DGEMM_UNROLL_M
932 export DGEMM_UNROLL_N
933 export QGEMM_UNROLL_M
934 export QGEMM_UNROLL_N
935 export CGEMM_UNROLL_M
936 export CGEMM_UNROLL_N
937 export ZGEMM_UNROLL_M
938 export ZGEMM_UNROLL_N
939 export XGEMM_UNROLL_M
940 export XGEMM_UNROLL_N
941 export CGEMM3M_UNROLL_M
942 export CGEMM3M_UNROLL_N
943 export ZGEMM3M_UNROLL_M
944 export ZGEMM3M_UNROLL_N
945 export XGEMM3M_UNROLL_M
946 export XGEMM3M_UNROLL_N
956 .SUFFIXES: .$(PSUFFIX) .$(SUFFIX) .f
959 $(FC) $(FFLAGS) -c $< -o $(@F)
962 $(FC) $(FPFLAGS) -pg -c $< -o $(@F)
966 PATHSCALEPATH = /opt/pathscale/lib/3.1
967 PGIPATH = /opt/pgi/linux86-64/7.1-5/lib
969 PATHSCALEPATH = /opt/pathscale/lib/3.1/32
970 PGIPATH = /opt/pgi/linux86/7.1-5/lib
973 ACMLPATH = /opt/acml/4.3.0
974 ifneq ($(OSNAME), Darwin)
975 MKLPATH = /opt/intel/mkl/10.2.2.025/lib
977 MKLPATH = /Library/Frameworks/Intel_MKL.framework/Versions/10.0.1.014/lib
979 ATLASPATH = /opt/atlas/3.9.17/opteron
980 FLAMEPATH = $(HOME)/flame/lib
981 ifneq ($(OSNAME), SunOS)
982 SUNPATH = /opt/sunstudio12.1
984 SUNPATH = /opt/SUNWspro