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
328 #BULLDOZER PILEDRIVER
332 ifeq ($(ARCH), x86_64)
333 DYNAMIC_CORE = PRESCOTT CORE2 PENRYN DUNNINGTON NEHALEM OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
335 DYNAMIC_CORE += SANDYBRIDGE
336 #BULLDOZER PILEDRIVER
349 ifeq ($(F_COMPILER), GFORTRAN)
350 ifeq ($(C_COMPILER), GCC)
352 # CCOMMON_OPT += -DEXPRECISION
357 ifeq ($(ARCH), mips64)
361 ifeq ($(ARCH), alpha)
367 # C Compiler dependent settings
371 # ifeq logical or. GCC or CLANG or LSB
372 # http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
373 ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC CLANG LSB))
375 COMMON_PROF += -fno-inline
376 NO_UNINITIALIZED_WARN = -Wno-uninitialized
378 ifeq ($(QUIET_MAKE), 1)
379 CCOMMON_OPT += $(NO_UNINITIALIZED_WARN) -Wno-unused
384 ifeq ($(ARCH), mips64)
386 CCOMMON_OPT += -mabi=64
388 CCOMMON_OPT += -mabi=n32
393 ifeq ($(CORE), LOONGSON3A)
394 CCOMMON_OPT += -march=mips64
395 FCOMMON_OPT += -march=mips64
398 ifeq ($(CORE), LOONGSON3B)
399 CCOMMON_OPT += -march=mips64
400 FCOMMON_OPT += -march=mips64
403 ifeq ($(OSNAME), AIX)
409 ifndef BINARY_DEFINED
419 ifeq ($(C_COMPILER), PGI)
421 CCOMMON_OPT += -tp p7-64
423 CCOMMON_OPT += -tp p7
427 ifeq ($(C_COMPILER), PATHSCALE)
436 # Fortran Compiler dependent settings
439 ifeq ($(F_COMPILER), G77)
440 CCOMMON_OPT += -DF_INTERFACE_G77
442 ifndef NO_BINARY_MODE
451 ifeq ($(F_COMPILER), G95)
452 CCOMMON_OPT += -DF_INTERFACE_G95
454 ifndef NO_BINARY_MODE
463 ifeq ($(F_COMPILER), GFORTRAN)
464 CCOMMON_OPT += -DF_INTERFACE_GFORT
466 #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
467 ifneq ($(NO_LAPACK), 1)
468 ifneq ($(C_COMPILER), LSB)
469 EXTRALIB += -lgfortran
473 ifeq ($(ARCH), mips64)
475 FCOMMON_OPT += -mabi=64
477 FCOMMON_OPT += -mabi=n32
484 FCOMMON_OPT += -fdefault-integer-8
491 FCOMMON_OPT += -fopenmp
495 ifeq ($(F_COMPILER), INTEL)
496 CCOMMON_OPT += -DF_INTERFACE_INTEL
501 FCOMMON_OPT += -openmp
505 ifeq ($(F_COMPILER), FUJITSU)
506 CCOMMON_OPT += -DF_INTERFACE_FUJITSU
508 FCOMMON_OPT += -openmp
512 ifeq ($(F_COMPILER), IBM)
513 CCOMMON_OPT += -DF_INTERFACE_IBM
514 # FCOMMON_OPT += -qarch=440
518 FCOMMON_OPT += -qintsize=8
524 FCOMMON_OPT += -openmp
528 ifeq ($(F_COMPILER), PGI)
529 CCOMMON_OPT += -DF_INTERFACE_PGI
530 COMMON_PROF += -DPGICOMPILER
535 FCOMMON_OPT += -tp p7-64
537 FCOMMON_OPT += -tp p7
544 ifeq ($(F_COMPILER), PATHSCALE)
545 CCOMMON_OPT += -DF_INTERFACE_PATHSCALE
552 ifneq ($(ARCH), mips64)
560 FCOMMON_OPT += -mabi=64
562 FCOMMON_OPT += -mabi=n32
571 ifeq ($(F_COMPILER), OPEN64)
572 CCOMMON_OPT += -DF_INTERFACE_OPEN64
579 ifeq ($(ARCH), mips64)
585 ifeq ($(CORE), LOONGSON3A)
586 FCOMMON_OPT += -loongson3 -static
589 ifeq ($(CORE), LOONGSON3B)
590 FCOMMON_OPT += -loongson3 -static
602 FEXTRALIB += -lstdc++
607 ifeq ($(C_COMPILER), OPEN64)
609 ifeq ($(ARCH), mips64)
615 ifeq ($(CORE), LOONGSON3A)
616 CCOMMON_OPT += -loongson3 -static
619 ifeq ($(CORE), LOONGSON3B)
620 CCOMMON_OPT += -loongson3 -static
633 ifeq ($(C_COMPILER), SUN)
642 ifeq ($(F_COMPILER), SUN)
643 CCOMMON_OPT += -DF_INTERFACE_SUN
650 FCOMMON_OPT += -xopenmp=parallel
654 ifeq ($(F_COMPILER), COMPAQ)
655 CCOMMON_OPT += -DF_INTERFACE_COMPAQ
657 FCOMMON_OPT += -openmp
668 ifeq ($(NEED_PIC), 1)
669 ifeq ($(C_COMPILER), IBM)
670 CCOMMON_OPT += -qpic=large
674 ifeq ($(F_COMPILER), SUN)
681 ifeq ($(DYNAMIC_ARCH), 1)
682 CCOMMON_OPT += -DDYNAMIC_ARCH
685 ifeq ($(NO_LAPACK), 1)
686 CCOMMON_OPT += -DNO_LAPACK
687 #Disable LAPACK C interface
691 ifeq ($(NO_LAPACKE), 1)
692 CCOMMON_OPT += -DNO_LAPACKE
696 CCOMMON_OPT += -DNO_AVX
700 CCOMMON_OPT += -DSMP_SERVER
702 ifeq ($(ARCH), mips64)
703 ifneq ($(CORE), LOONGSON3B)
704 USE_SIMPLE_THREADED_LEVEL3 = 1
708 ifeq ($(USE_OPENMP), 1)
709 # USE_SIMPLE_THREADED_LEVEL3 = 1
711 CCOMMON_OPT += -DUSE_OPENMP
716 ifeq ($(NO_WARMUP), 1)
717 CCOMMON_OPT += -DNO_WARMUP
720 ifeq ($(CONSISTENT_FPCSR), 1)
721 CCOMMON_OPT += -DCONSISTENT_FPCSR
724 # Only for development
725 # CCOMMON_OPT += -DPARAMTEST
726 # CCOMMON_OPT += -DPREFETCHTEST
727 # CCOMMON_OPT += -DNO_SWITCHING
731 CCOMMON_OPT += -DUSE_PAPI
732 EXTRALIB += -lpapi -lperfctr
735 ifdef DYNAMIC_THREADS
736 CCOMMON_OPT += -DDYNAMIC_THREADS
739 CCOMMON_OPT += -DMAX_CPU_NUMBER=$(NUM_THREADS)
741 ifdef USE_SIMPLE_THREADED_LEVEL3
742 CCOMMON_OPT += -DUSE_SIMPLE_THREADED_LEVEL3
746 LIBPREFIX = libopenblas
748 LIBPREFIX = libopenblas_$(LIBNAMESUFFIX)
751 KERNELDIR = $(TOPDIR)/kernel/$(ARCH)
753 include $(TOPDIR)/Makefile.$(ARCH)
755 CCOMMON_OPT += -DASMNAME=$(FU)$(*F) -DASMFNAME=$(FU)$(*F)$(BU) -DNAME=$(*F)$(BU) -DCNAME=$(*F) -DCHAR_NAME=\"$(*F)$(BU)\" -DCHAR_CNAME=\"$(*F)\"
757 ifeq ($(CORE), PPC440)
758 CCOMMON_OPT += -DALLOC_QALLOC
761 ifeq ($(CORE), PPC440FP2)
762 STATIC_ALLOCATION = 1
765 ifneq ($(OSNAME), Linux)
769 ifneq ($(ARCH), x86_64)
771 ifneq ($(CORE), LOONGSON3B)
778 CCOMMON_OPT += -DNO_AFFINITY
781 ifdef FUNCTION_PROFILE
782 CCOMMON_OPT += -DFUNCTION_PROFILE
785 ifdef HUGETLB_ALLOCATION
786 CCOMMON_OPT += -DALLOC_HUGETLB
789 ifdef HUGETLBFILE_ALLOCATION
790 CCOMMON_OPT += -DALLOC_HUGETLBFILE -DHUGETLB_FILE_NAME=$(HUGETLBFILE_ALLOCATION)
793 ifdef STATIC_ALLOCATION
794 CCOMMON_OPT += -DALLOC_STATIC
797 ifdef DEVICEDRIVER_ALLOCATION
798 CCOMMON_OPT += -DALLOC_DEVICEDRIVER -DDEVICEDRIVER_NAME=\"/dev/mapper\"
801 ifdef MIXED_MEMORY_ALLOCATION
802 CCOMMON_OPT += -DMIXED_MEMORY_ALLOCATION
805 ifeq ($(OSNAME), SunOS)
821 REVISION = -r$(VERSION)
822 MAJOR_VERSION = $(word 1,$(subst ., ,$(VERSION)))
833 override CFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
834 override PFLAGS += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
836 override FFLAGS += $(COMMON_OPT) $(FCOMMON_OPT)
837 override FPFLAGS += $(COMMON_OPT) $(FCOMMON_OPT) $(COMMON_PROF)
840 LAPACK_CFLAGS = $(CFLAGS)
841 LAPACK_CFLAGS += -DHAVE_LAPACK_CONFIG_H
843 LAPACK_CFLAGS += -DLAPACK_ILP64
845 ifeq ($(C_COMPILER), LSB)
846 LAPACK_CFLAGS += -DLAPACK_COMPLEX_STRUCTURE
861 ifneq ($(DYNAMIC_ARCH), 1)
863 LIBNAME = $(LIBPREFIX)_$(LIBCORE)$(REVISION).$(LIBSUFFIX)
864 LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)$(REVISION)_p.$(LIBSUFFIX)
866 LIBNAME = $(LIBPREFIX)_$(LIBCORE)p$(REVISION).$(LIBSUFFIX)
867 LIBNAME_P = $(LIBPREFIX)_$(LIBCORE)p$(REVISION)_p.$(LIBSUFFIX)
871 LIBNAME = $(LIBPREFIX)$(REVISION).$(LIBSUFFIX)
872 LIBNAME_P = $(LIBPREFIX)$(REVISION)_p.$(LIBSUFFIX)
874 LIBNAME = $(LIBPREFIX)p$(REVISION).$(LIBSUFFIX)
875 LIBNAME_P = $(LIBPREFIX)p$(REVISION)_p.$(LIBSUFFIX)
880 LIBDLLNAME = $(LIBPREFIX).dll
881 LIBSONAME = $(LIBNAME:.$(LIBSUFFIX)=.so)
882 LIBDYNNAME = $(LIBNAME:.$(LIBSUFFIX)=.dylib)
883 LIBDEFNAME = $(LIBNAME:.$(LIBSUFFIX)=.def)
884 LIBEXPNAME = $(LIBNAME:.$(LIBSUFFIX)=.exp)
885 LIBZIPNAME = $(LIBNAME:.$(LIBSUFFIX)=.zip)
887 LIBS = $(TOPDIR)/$(LIBNAME)
888 LIBS_P = $(TOPDIR)/$(LIBNAME_P)
900 export NEED2UNDERSCORES
929 export FUNCTION_PROFILE
932 export SGEMM_UNROLL_M
933 export SGEMM_UNROLL_N
934 export DGEMM_UNROLL_M
935 export DGEMM_UNROLL_N
936 export QGEMM_UNROLL_M
937 export QGEMM_UNROLL_N
938 export CGEMM_UNROLL_M
939 export CGEMM_UNROLL_N
940 export ZGEMM_UNROLL_M
941 export ZGEMM_UNROLL_N
942 export XGEMM_UNROLL_M
943 export XGEMM_UNROLL_N
944 export CGEMM3M_UNROLL_M
945 export CGEMM3M_UNROLL_N
946 export ZGEMM3M_UNROLL_M
947 export ZGEMM3M_UNROLL_N
948 export XGEMM3M_UNROLL_M
949 export XGEMM3M_UNROLL_N
959 .SUFFIXES: .$(PSUFFIX) .$(SUFFIX) .f
962 $(FC) $(FFLAGS) -c $< -o $(@F)
965 $(FC) $(FPFLAGS) -pg -c $< -o $(@F)
969 PATHSCALEPATH = /opt/pathscale/lib/3.1
970 PGIPATH = /opt/pgi/linux86-64/7.1-5/lib
972 PATHSCALEPATH = /opt/pathscale/lib/3.1/32
973 PGIPATH = /opt/pgi/linux86/7.1-5/lib
976 ACMLPATH = /opt/acml/4.3.0
977 ifneq ($(OSNAME), Darwin)
978 MKLPATH = /opt/intel/mkl/10.2.2.025/lib
980 MKLPATH = /Library/Frameworks/Intel_MKL.framework/Versions/10.0.1.014/lib
982 ATLASPATH = /opt/atlas/3.9.17/opteron
983 FLAMEPATH = $(HOME)/flame/lib
984 ifneq ($(OSNAME), SunOS)
985 SUNPATH = /opt/sunstudio12.1
987 SUNPATH = /opt/SUNWspro