Merge branch 'release-0.2.9'

Author: xianyi

BACKERS.md

@@ -0,0 +1,39 @@
+Thank you for the support.
+
+### [2013.8] [Testbed for OpenBLAS project](https://www.bountysource.com/fundraisers/443-testbed-for-openblas-project)
+
+https://www.bountysource.com/fundraisers/443-testbed-for-openblas-project/pledges
+
+In chronological order:
+
+* aeberspaecher
+* fmolina
+* saullocastro
+* xianyi
+* cuda
+* carter
+* StefanKarpinski
+* staticfloat
+* sebastien-villemot
+* JeffBezanson
+* ihnorton
+* simonp0420
+* andrioni
+* Tim Holy
+* ivarne
+* johnmyleswhite
+* traz
+* Jean-Francis Roy
+* bkalpert
+* Anirban
+* pgermain
+* alexandre.lacoste.18
+* foges
+* ssam
+* WestleyArgentum
+* daniebmariani
+* pjpuglia
+* albarrentine
+* Alexander Vogt
+
+

CONTRIBUTORS.md

@@ -10,13 +10,26 @@
   * Optimize BLAS3 on ICT Loongson 3A.
   * Optimize BLAS3 on Intel Sandy Bridge.
 
+* Werner Saar <[email protected]>
+  * [2013-03-04] Optimize AVX and FMA4 DGEMM on AMD Bulldozer
+  * [2013-04-27] Optimize AVX and FMA4 TRSM on AMD Bulldozer
+  * [2013-06-09] Optimize AVX and FMA4 SGEMM on AMD Bulldozer
+  * [2013-06-11] Optimize AVX and FMA4 ZGEMM on AMD Bulldozer
+  * [2013-06-12] Optimize AVX and FMA4 CGEMM on AMD Bulldozer
+  * [2013-06-16] Optimize dgemv_n kernel on AMD Bulldozer
+  * [2013-06-20] Optimize ddot, daxpy kernel on AMD Bulldozer
+  * [2013-06-21] Optimize dcopy kernel on AMD Bulldozer
+  * Porting and Optimization on ARM Cortex-A9
+  * Optimization on AMD Piledriver
+  * Optimization on Intel Haswell
+
+## Previous Developers
+
 * Zaheer Chothia <[email protected]>
   * Improve the compatibility about complex number
   * Build LAPACKE: C interface to LAPACK
   * Improve the windows build.
 
-## Previous Developers
-
 * Chen Shaohu <[email protected]>
   * Optimize GEMV on the Loongson 3A processor. 
 
@@ -52,16 +65,7 @@ In chronological order:
 
 * Sébastien Villemot <[email protected]>
   * [2012-11-14] Fix compilation with TARGET=GENERIC. Patch applied to Debian package.
-
-* Werner Saar <[email protected]>
-  * [2013-03-04] Optimize AVX and FMA4 DGEMM on AMD Bulldozer
-  * [2013-04-27] Optimize AVX and FMA4 TRSM on AMD Bulldozer
-  * [2013-06-09] Optimize AVX and FMA4 SGEMM on AMD Bulldozer
-  * [2013-06-11] Optimize AVX and FMA4 ZGEMM on AMD Bulldozer
-  * [2013-06-12] Optimize AVX and FMA4 CGEMM on AMD Bulldozer
-  * [2013-06-16] Optimize dgemv_n kernel on AMD Bulldozer
-  * [2013-06-20] Optimize ddot, daxpy kernel on AMD Bulldozer
-  * [2013-06-21] Optimize dcopy kernel on AMD Bulldozer
+  * [2013-08-28] Avoid failure on qemu guests declaring an Athlon CPU without 3dnow!
 
 * Kang-Che Sung <[email protected]>
   * [2013-05-17] Fix typo in the document. Re-order the architecture list in getarch.c.
@@ -79,9 +83,36 @@ In chronological order:
   * [2013-07-11] create openblas_get_parallel to retrieve information which parallelization 
     model is used by OpenBLAS.
 
+* Elliot Saba <[email protected]>
+  * [2013-07-22] Add in return value for `interface/trtri.c`
+
 * Sébastien Fabbro <[email protected]>
   * [2013-07-24] Modify makefile to respect user's LDFLAGS
   * [2013-07-24] Add stack markings for GNU as arch-independent for assembler files
 
+* Viral B. Shah <[email protected]>
+  * [2013-08-21] Patch LAPACK XLASD4.f as discussed in JuliaLang/julia#2340
+
+* Lars Buitinck <https://github.com/larsmans>
+  * [2013-08-28] get rid of the generated cblas_noconst.h file
+  * [2013-08-28] Missing threshold in gemm.c
+  * [2013-08-28] fix default prefix handling in makefiles
+
+* yieldthought <https://github.com/yieldthought>
+  * [2013-10-08] Remove -Wl,--retain-symbols-file from dynamic link line to fix tool support
+
+* Keno Fischer <https://github.com/loladiro>
+  * [2013-10-23] Use FC instead of CC to link the dynamic library on OS X
+
+* Christopher Meng <[email protected]>
+  * [2013-12-09] Add DESTDIR support for easier building on RPM based distros.
+                 Use install command instead of cp to install files with permissions control.
+
+* Lucas Beyer <[email protected]>
+  * [2013-12-10] Added support for NO_SHARED in make install.
+
+* carlkl <https://github.com/carlkl>
+  * [2013-12-13] Fixed LAPACKE building bug on Windows
+
 * [Your name or handle] <[email or website]>
   * [Date] [Brief summary of your changes]

Changelog.txt

@@ -1,4 +1,31 @@
 OpenBLAS ChangeLog
+====================================================================
+Version 0.2.9.rc2
+06-Mar-2014
+common:
+	* Added OPENBLAS_VERBOSE environment variable.(#338)
+	* Make OpenBLAS thread-pool resilient to fork via pthread_atfork.
+	  (#294, Thank Olivier Grisel)
+	* Rewrote rotmg
+	* Fixed sdsdot bug.
+x86/x86-64:
+	* Detect Intel Haswell for new Macbook.
+
+====================================================================
+Version 0.2.9.rc1
+13-Jan-2013
+common:
+	* Update LAPACK to 3.5.0 version
+	* Fixed compatiable issues with Clang and Pathscale compilers.
+
+x86/x86-64:
+	* Optimization on Intel Haswell.
+	* Enable optimization kernels on AMD Bulldozer and Piledriver.
+
+ARM:
+	* Support ARMv6 and ARMv7 ISA.
+	* Optimization on ARM Cortex-A9.
+
 ====================================================================
 Version 0.2.8
 01-Aug-2013

Makefile

@@ -15,10 +15,6 @@ ifdef SANITY_CHECK
 BLASDIRS += reference
 endif
 
-ifndef PREFIX
-PREFIX = /opt/OpenBLAS
-endif
-
 SUBDIRS	= $(BLASDIRS)
 ifneq ($(NO_LAPACK), 1)
 SUBDIRS	+= lapack
@@ -31,7 +27,7 @@ SUBDIRS_ALL = $(SUBDIRS) test ctest utest exports benchmark ../laswp ../bench
 
 all :: libs netlib tests shared
 	@echo
-	@echo " OpenBLAS build complete."
+	@echo " OpenBLAS build complete. ($(LIB_COMPONENTS))"
 	@echo
 	@echo "  OS               ... $(OSNAME)             "
 	@echo "  Architecture     ... $(ARCH)               "
@@ -44,7 +40,9 @@ ifdef INTERFACE64
 	@echo "  Use 64 bits int    (equivalent to \"-i8\" in Fortran)      "
 endif
 	@echo "  C compiler       ... $(C_COMPILER)  (command line : $(CC))"
+ifndef NOFORTRAN
 	@echo "  Fortran compiler ... $(F_COMPILER)  (command line : $(FC))"
+endif
 ifneq ($(OSNAME), AIX)
 	@echo -n "  Library Name     ... $(LIBNAME)"
 else
@@ -221,10 +219,10 @@ prof_lapack : lapack_prebuild
 lapack_prebuild :
 ifndef NOFORTRAN
 	-@echo "FORTRAN     = $(FC)" > $(NETLIB_LAPACK_DIR)/make.inc
-	-@echo "OPTS        = $(FFLAGS)" >> $(NETLIB_LAPACK_DIR)/make.inc
-	-@echo "POPTS       = $(FPFLAGS)" >> $(NETLIB_LAPACK_DIR)/make.inc
-	-@echo "NOOPT       = $(FFLAGS) -O0" >> $(NETLIB_LAPACK_DIR)/make.inc
-	-@echo "PNOOPT      = $(FPFLAGS) -O0" >> $(NETLIB_LAPACK_DIR)/make.inc
+	-@echo "OPTS        = $(LAPACK_FFLAGS)" >> $(NETLIB_LAPACK_DIR)/make.inc
+	-@echo "POPTS       = $(LAPACK_FPFLAGS)" >> $(NETLIB_LAPACK_DIR)/make.inc
+	-@echo "NOOPT       = $(LAPACK_FFLAGS) -O0" >> $(NETLIB_LAPACK_DIR)/make.inc
+	-@echo "PNOOPT      = $(LAPACK_FPFLAGS) -O0" >> $(NETLIB_LAPACK_DIR)/make.inc
 	-@echo "LOADOPTS    = $(FFLAGS) $(EXTRALIB)" >> $(NETLIB_LAPACK_DIR)/make.inc
 	-@echo "CC          = $(CC)" >> $(NETLIB_LAPACK_DIR)/make.inc
 	-@echo "override CFLAGS      = $(LAPACK_CFLAGS)" >> $(NETLIB_LAPACK_DIR)/make.inc
@@ -320,7 +318,7 @@ clean ::
 ifeq ($(OSNAME), Darwin)
 	@rm -rf getarch.dSYM getarch_2nd.dSYM
 endif
-	@rm -f Makefile.conf config.h cblas_noconst.h Makefile_kernel.conf config_kernel.h st* *.dylib
+	@rm -f Makefile.conf config.h Makefile_kernel.conf config_kernel.h st* *.dylib
 	@touch $(NETLIB_LAPACK_DIR)/make.inc
 	@$(MAKE) -C $(NETLIB_LAPACK_DIR) clean
 	@rm -f $(NETLIB_LAPACK_DIR)/make.inc $(NETLIB_LAPACK_DIR)/lapacke/include/lapacke_mangling.h

Makefile.arm

@@ -0,0 +1,12 @@
+
+ifeq ($(CORE), ARMV7)
+CCOMMON_OPT += -marm -mfpu=vfpv3  -mfloat-abi=hard -march=armv7-a
+FCOMMON_OPT += -marm -mfpu=vfpv3  -mfloat-abi=hard -march=armv7-a
+endif
+
+ifeq ($(CORE), ARMV6)
+CCOMMON_OPT += -marm -mfpu=vfp -mfloat-abi=hard  -march=armv6
+FCOMMON_OPT += -marm -mfpu=vfp -mfloat-abi=hard  -march=armv6
+endif
+
+

Makefile.arm64

@@ -0,0 +1,7 @@
+
+ifeq ($(CORE), ARMV8)
+CCOMMON_OPT += -march=armv8-a
+FCOMMON_OPT += -march=armv8-a
+endif
+
+

Makefile.install

@@ -3,9 +3,11 @@ export GOTOBLAS_MAKEFILE = 1
 -include $(TOPDIR)/Makefile.conf_last
 include ./Makefile.system
 
-OPENBLAS_INCLUDE_DIR:=$(PREFIX)/include
-OPENBLAS_LIBRARY_DIR:=$(PREFIX)/lib
-OPENBLAS_BUILD_DIR:=$(CURDIR)
+PREFIX ?= /opt/OpenBLAS
+
+OPENBLAS_INCLUDE_DIR := $(PREFIX)/include
+OPENBLAS_LIBRARY_DIR := $(PREFIX)/lib
+OPENBLAS_BUILD_DIR := $(CURDIR)
 
 .PHONY : install
 .NOTPARALLEL : install
@@ -14,71 +16,73 @@ lib.grd :
 	$(error OpenBLAS: Please run "make" firstly)
 
 install : 	lib.grd
-	@-mkdir -p $(PREFIX)
-	@-mkdir -p $(OPENBLAS_INCLUDE_DIR)
-	@-mkdir -p $(OPENBLAS_LIBRARY_DIR)
-	@echo Generating openblas_config.h in $(OPENBLAS_INCLUDE_DIR)
+	@-mkdir -p $(DESTDIR)$(PREFIX)
+	@-mkdir -p $(DESTDIR)$(OPENBLAS_INCLUDE_DIR)
+	@-mkdir -p $(DESTDIR)$(OPENBLAS_LIBRARY_DIR)
+	@echo Generating openblas_config.h in $(DESTDIR)$(OPENBLAS_INCLUDE_DIR)
 #for inc 
-	@echo \#ifndef OPENBLAS_CONFIG_H > $(OPENBLAS_INCLUDE_DIR)/openblas_config.h
-	@echo \#define OPENBLAS_CONFIG_H >> $(OPENBLAS_INCLUDE_DIR)/openblas_config.h
-	@cat config_last.h >> $(OPENBLAS_INCLUDE_DIR)/openblas_config.h
-	@echo \#define VERSION \" OpenBLAS $(VERSION) \" >> $(OPENBLAS_INCLUDE_DIR)/openblas_config.h
-	@cat openblas_config_template.h >> $(OPENBLAS_INCLUDE_DIR)/openblas_config.h
-	@echo \#endif  \/\* OPENBLAS_CONFIG_H \*\/ >> $(OPENBLAS_INCLUDE_DIR)/openblas_config.h
+	@echo \#ifndef OPENBLAS_CONFIG_H > $(DESTDIR)$(OPENBLAS_INCLUDE_DIR)/openblas_config.h
+	@echo \#define OPENBLAS_CONFIG_H >> $(DESTDIR)$(OPENBLAS_INCLUDE_DIR)/openblas_config.h
+	@awk 'NF {print $$1, "OPENBLAS_"$$2, $$3}' config_last.h >> $(DESTDIR)$(OPENBLAS_INCLUDE_DIR)/openblas_config.h
+	@echo \#define OPENBLAS_VERSION \" OpenBLAS $(VERSION) \" >> $(DESTDIR)$(OPENBLAS_INCLUDE_DIR)/openblas_config.h
+	@cat openblas_config_template.h >> $(DESTDIR)$(OPENBLAS_INCLUDE_DIR)/openblas_config.h
+	@echo \#endif  \/\* OPENBLAS_CONFIG_H \*\/ >> $(DESTDIR)$(OPENBLAS_INCLUDE_DIR)/openblas_config.h
 
-	@echo Generating f77blas.h in $(OPENBLAS_INCLUDE_DIR)
-	@echo \#ifndef OPENBLAS_F77BLAS_H > $(OPENBLAS_INCLUDE_DIR)/f77blas.h
-	@echo \#define OPENBLAS_F77BLAS_H >> $(OPENBLAS_INCLUDE_DIR)/f77blas.h
-	@echo \#include \"openblas_config.h\" >> $(OPENBLAS_INCLUDE_DIR)/f77blas.h
-	@cat common_interface.h >> $(OPENBLAS_INCLUDE_DIR)/f77blas.h
-	@echo \#endif >> $(OPENBLAS_INCLUDE_DIR)/f77blas.h
+	@echo Generating f77blas.h in $(DESTDIR)$(OPENBLAS_INCLUDE_DIR)
+	@echo \#ifndef OPENBLAS_F77BLAS_H > $(DESTDIR)$(OPENBLAS_INCLUDE_DIR)/f77blas.h
+	@echo \#define OPENBLAS_F77BLAS_H >> $(DESTDIR)$(OPENBLAS_INCLUDE_DIR)/f77blas.h
+	@echo \#include \"openblas_config.h\" >> $(DESTDIR)$(OPENBLAS_INCLUDE_DIR)/f77blas.h
+	@cat common_interface.h >> $(DESTDIR)$(OPENBLAS_INCLUDE_DIR)/f77blas.h
+	@echo \#endif >> $(DESTDIR)$(OPENBLAS_INCLUDE_DIR)/f77blas.h
 
 ifndef NO_CBLAS
-	@echo Generating cblas.h in $(OPENBLAS_INCLUDE_DIR)
-	@sed 's/common/openblas_config/g' cblas.h > $(OPENBLAS_INCLUDE_DIR)/cblas.h
+	@echo Generating cblas.h in $(DESTDIR)$(OPENBLAS_INCLUDE_DIR)
+	@sed 's/common/openblas_config/g' cblas.h > $(DESTDIR)$(OPENBLAS_INCLUDE_DIR)/cblas.h
 endif
 
 ifndef NO_LAPACKE
-	@echo Copying LAPACKE header files to $(OPENBLAS_LIBRARY_DIR)
-	@-cp $(NETLIB_LAPACK_DIR)/lapacke/include/lapacke.h $(OPENBLAS_INCLUDE_DIR)/lapacke.h
-	@-cp $(NETLIB_LAPACK_DIR)/lapacke/include/lapacke_config.h $(OPENBLAS_INCLUDE_DIR)/lapacke_config.h
-	@-cp $(NETLIB_LAPACK_DIR)/lapacke/include/lapacke_mangling_with_flags.h $(OPENBLAS_INCLUDE_DIR)/lapacke_mangling.h
-	@-cp $(NETLIB_LAPACK_DIR)/lapacke/include/lapacke_utils.h $(OPENBLAS_INCLUDE_DIR)/lapacke_utils.h
+	@echo Copying LAPACKE header files to $(DESTDIR)$(OPENBLAS_LIBRARY_DIR)
+	@-install -pDm644 $(NETLIB_LAPACK_DIR)/lapacke/include/lapacke.h $(DESTDIR)$(OPENBLAS_INCLUDE_DIR)/lapacke.h
+	@-install -pDm644 $(NETLIB_LAPACK_DIR)/lapacke/include/lapacke_config.h $(DESTDIR)$(OPENBLAS_INCLUDE_DIR)/lapacke_config.h
+	@-install -pDm644 $(NETLIB_LAPACK_DIR)/lapacke/include/lapacke_mangling_with_flags.h $(DESTDIR)$(OPENBLAS_INCLUDE_DIR)/lapacke_mangling.h
+	@-install -pDm644 $(NETLIB_LAPACK_DIR)/lapacke/include/lapacke_utils.h $(DESTDIR)$(OPENBLAS_INCLUDE_DIR)/lapacke_utils.h
 endif
 
 #for install static library 
-	@echo Copy the static library to $(OPENBLAS_LIBRARY_DIR)
-	@cp $(LIBNAME) $(OPENBLAS_LIBRARY_DIR)
-	@cd $(OPENBLAS_LIBRARY_DIR) ; \
+	@echo Copying the static library to $(DESTDIR)$(OPENBLAS_LIBRARY_DIR)
+	@install -pm644 $(LIBNAME) $(DESTDIR)$(OPENBLAS_LIBRARY_DIR)
+	@cd $(DESTDIR)$(OPENBLAS_LIBRARY_DIR) ; \
 	ln -fs $(LIBNAME) $(LIBPREFIX).$(LIBSUFFIX)
 #for install shared library 
-	@echo Copy the shared library to $(OPENBLAS_LIBRARY_DIR)
+ifndef NO_SHARED
+	@echo Copying the shared library to $(DESTDIR)$(OPENBLAS_LIBRARY_DIR)
 ifeq ($(OSNAME), Linux)
-	@cp $(LIBSONAME) $(OPENBLAS_LIBRARY_DIR)
-	@cd $(OPENBLAS_LIBRARY_DIR) ; \
+	@install -pm755 $(LIBSONAME) $(DESTDIR)$(OPENBLAS_LIBRARY_DIR)
+	@cd $(DESTDIR)$(OPENBLAS_LIBRARY_DIR) ; \
 	ln -fs $(LIBSONAME) $(LIBPREFIX).so ; \
 	ln -fs $(LIBSONAME) $(LIBPREFIX).so.$(MAJOR_VERSION)
 endif
 ifeq ($(OSNAME), FreeBSD)
-	@cp $(LIBSONAME) $(OPENBLAS_LIBRARY_DIR)
-	@cd $(OPENBLAS_LIBRARY_DIR) ; \
+	@cp $(LIBSONAME) $(DESTDIR)$(OPENBLAS_LIBRARY_DIR)
+	@cd $(DESTDIR)$(OPENBLAS_LIBRARY_DIR) ; \
 	ln -fs $(LIBSONAME) $(LIBPREFIX).so
 endif
 ifeq ($(OSNAME), NetBSD)
-	@cp $(LIBSONAME) $(OPENBLAS_LIBRARY_DIR)
-	@cd $(OPENBLAS_LIBRARY_DIR) ; \
+	@cp $(LIBSONAME) $(DESTDIR)$(OPENBLAS_LIBRARY_DIR)
+	@cd $(DESTDIR)$(OPENBLAS_LIBRARY_DIR) ; \
 	ln -fs $(LIBSONAME) $(LIBPREFIX).so
 endif
 ifeq ($(OSNAME), Darwin)     
-	@-cp $(LIBDYNNAME) $(OPENBLAS_LIBRARY_DIR)
-	@-install_name_tool -id $(OPENBLAS_LIBRARY_DIR)/$(LIBDYNNAME) $(OPENBLAS_LIBRARY_DIR)/$(LIBDYNNAME)
-	@-ln -fs $(OPENBLAS_LIBRARY_DIR)/$(LIBDYNNAME) $(OPENBLAS_LIBRARY_DIR)/$(LIBPREFIX).dylib
+	@-cp $(LIBDYNNAME) $(DESTDIR)$(OPENBLAS_LIBRARY_DIR)
+	@-install_name_tool -id $(DESTDIR)$(OPENBLAS_LIBRARY_DIR)/$(LIBDYNNAME) $(DESTDIR)$(OPENBLAS_LIBRARY_DIR)/$(LIBDYNNAME)
+	@-ln -fs $(DESTDIR)$(OPENBLAS_LIBRARY_DIR)/$(LIBDYNNAME) $(DESTDIR)$(OPENBLAS_LIBRARY_DIR)/$(LIBPREFIX).dylib
 endif
 ifeq ($(OSNAME), WINNT)
 	@-cp $(LIBDLLNAME) $(OPENBLAS_LIBRARY_DIR)
 endif
 ifeq ($(OSNAME), CYGWIN_NT)
 	@-cp $(LIBDLLNAME) $(OPENBLAS_LIBRARY_DIR)
+endif
 endif
 
 	@echo Install OK!

Makefile.prebuild

@@ -17,13 +17,23 @@ ifdef CPUIDEMU
 EXFLAGS = -DCPUIDEMU -DVENDOR=99
 endif
 
-all: getarch_2nd cblas_noconst.h
+all: getarch_2nd
 	./getarch_2nd  0 >> $(TARGET_MAKE)
 	./getarch_2nd  1 >> $(TARGET_CONF)
 
 config.h : c_check f_check getarch
 	perl ./c_check $(TARGET_MAKE) $(TARGET_CONF) $(CC)
+ifneq ($(ONLY_CBLAS), 1)
 	perl ./f_check $(TARGET_MAKE) $(TARGET_CONF) $(FC)
+else
+#When we only build CBLAS, we set NOFORTRAN=2
+	echo "NOFORTRAN=2" >> $(TARGET_MAKE)
+	echo "NO_FBLAS=1" >> $(TARGET_MAKE)
+	echo "F_COMPILER=GFORTRAN"  >> $(TARGET_MAKE)
+	echo "BU=_"  >> $(TARGET_MAKE)
+	echo "#define BUNDERSCORE _" >> $(TARGET_CONF)
+	echo "#define NEEDBUNDERSCORE 1" >> $(TARGET_CONF)
+endif
 	./getarch 0 >> $(TARGET_MAKE)
 	./getarch 1 >> $(TARGET_CONF)
 
@@ -38,7 +48,4 @@ else
 	$(HOSTCC) -I. $(CFLAGS) -DBUILD_KERNEL -o $(@F) getarch_2nd.c
 endif
 
-cblas_noconst.h : cblas.h
-	perl -ane ' s/\bconst\b\s*//g; print; ' < cblas.h > cblas_noconst.h	
-
 dummy: