Merge branch 'loongson3a' into loongson3b

.gitignore

@@ -1,16 +1,25 @@
 *.obj
 *.lib
 *.dll
+*.dylib
 *.def
 *.o
 lapack-3.1.1
 lapack-3.1.1.tgz
+lapack-3.4.1
+lapack-3.4.1.tgz
+lapack-3.4.2
+lapack-3.4.2.tgz
 *.so
 *.a
 .svn
 *~
+lib.grd
+nohup.out
 config.h
 Makefile.conf
+Makefile.conf_last
+config_last.h
 getarch
 getarch_2nd
 utest/openblas_utest

Changelog.txt

@@ -1,4 +1,115 @@
 OpenBLAS ChangeLog
+====================================================================
+Version 0.2.6
+2-Mar-2013
+common:
+	* Improved OpenMP performance slightly. (d744c9)
+	* Improved cblas.h compatibility with Intel MKL.(#185)
+	* Fixed the overflowing bug in single thread cholesky factorization.
+	* Fixed the overflowing buffer bug of multithreading hbmv and sbmv.(#174)
+
+x86/x86-64:
+	* Added AMD Bulldozer x86-64 S/DGEMM AVX kernels. (Thank Werner Saar)
+	  We will tune the performance in future.
+	* Auto-detect Intel Xeon E7540.
+	* Fixed the overflowing buffer bug of gemv. (#173)
+	* Fixed the bug of s/cdot about invalid reading NAN on x86_64. (#189)
+
+MIPS64:
+
+====================================================================
+Version 0.2.5
+26-Nov-2012
+common:
+	* Added NO_SHARED flag to disable generating the shared library.
+	* Compile LAPACKE with ILP64 modle when INTERFACE64=1 (#158)
+	* Export LAPACK 3.4.2 symbols in shared library. (#147)
+	* Only detect the number of physical CPU cores on Mac OSX. (#157)
+	* Fixed NetBSD build. (#155)
+	* Fixed compilation with TARGET=GENERIC. (#160)
+x86/x86-64:
+	* Restore the original CPU affinity when calling 
+	  openblas_set_num_threads(1) (#153)
+	* Fixed a SEGFAULT bug in dgemv_t when m is very large.(#154)
+MIPS64:
+
+====================================================================
+Version 0.2.4
+8-Oct-2012
+common:
+	* Upgraded LAPACK to 3.4.2 version. (#145)
+	* Provided support for passing CFLAGS, FFLAGS, PFLAGS, 
+	  FPFLAGS to make. (#137)
+	* f77blas.h:compatibility for compilers without C99 complex 
+	  number support. (#141)
+x86/x86-64:
+	* Added NO_AVX flag. Check OS supporting AVX on runtime. (#139)
+	* Fixed zdot incompatibility ABI issue with GCC 4.7 on 
+	  Windows 32-bit. (#140)
+MIPS64:
+	* Fixed the generation of shared library bug.
+	* Fixed the detection bug on the Loongson 3A server.
+====================================================================
+Version 0.2.3
+20-Aug-2012
+common:
+	* Fixed LAPACK unstable bug about ?laswp. (#130)
+	* Fixed the shared library bug about unloading the library on 
+	  Linux (#132).
+	* Fixed the compilation failure on BlueGene/P (TARGET=PPC440FP2)
+	  Please use gcc and IBM xlf. (#134)
+x86/x86-64:
+	* Supported goto_set_num_threads and openblas_set_num_threads 
+	  APIs in Windows. They can set the number of threads on runtime.
+	  
+====================================================================
+Version 0.2.2
+6-July-2012
+common:
+	* Fixed exporting DLL functions bug on Windows/MingW
+	* Support GNU Hurd (Thank Sylvestre Ledru)
+	* Support kfreebsd kernel (Thank Sylvestre Ledru)
+x86/x86-64:
+	* Support Intel Sandy Bridge 22nm desktop/mobile CPU
+SPARC:
+	* Improve the detection of SPARC (Thank Sylvestre Ledru)
+
+====================================================================
+Version 0.2.1
+30-Jun-2012
+common:
+x86/x86-64:
+	* Fixed the SEGFAULT bug about hyper-theading
+	* Support AMD Bulldozer by using GotoBLAS2 AMD Barcelona codes
+
+====================================================================
+Version 0.2.0
+26-Jun-2012
+common:
+	* Removed the limitation (64) of numbers of CPU cores.
+	  Now, it supports 256 cores at max.
+	* Supported clang compiler.
+	* Fixed some build bugs on FreeBSD
+x86/x86-64:
+	* Optimized Level-3 BLAS on Intel Sandy Bridge x86-64 by AVX instructions.
+	  Please use gcc >= 4.6 or clang >=3.1.
+	* Support AMD Bobcat by using GotoBLAS2 AMD Barcelona codes.
+
+====================================================================
+Version 0.1.1
+29-Apr-2012
+common:
+	* Upgraded LAPACK to 3.4.1 version. (Thank Zaheer Chothia)
+	* Supported LAPACKE, a C interface to LAPACKE. (Thank Zaheer Chothia)
+	* Fixed the build bug (MD5 and download) on Mac OSX.
+	* Auto download CUnit 2.1.2-2 from SF.net with UTEST_CHECK=1.
+	* Fxied the compatibility issue for compilers without C99 complex number
+	  (e.g. Visual Studio)
+x86/x86_64:
+	* Auto-detect Intel Sandy Bridge Core i7-3xxx & Xeon E7 Westmere-EX.
+	* Test alpha=Nan in dscale.
+	* Fixed a SEGFAULT bug in samax on x86 windows.
+	
 ====================================================================
 Version 0.1.0
 23-Mar-2012

GotoBLAS_03FAQ.txt

@@ -90,6 +90,15 @@
        number of threads will consume extra resource. I recommend you to
        specify minimum number of threads.
 
+1.9  Q I have segfaults when I compile with USE_OPENMP=1. What's wrong?
+
+     A This may be related to a bug in the Linux kernel 2.6.32. Try applying
+     the patch segaults.patch using
+
+     patch < segfaults.patch
+
+     and see if the crashes persist. Note that this patch will lead to many
+     compiler warnings.
 
 2. Architecture Specific issue or Implementation
 

LICENSE

@@ -1,4 +1,4 @@
-Copyright (c) 2011, Lab of Parallel Software and Computational Science,ICSAS
+Copyright (c) 2011,2012 Lab of Parallel Software and Computational Science,ISCAS
 All rights reserved.
 
 Redistribution and use in source and binary forms, with or without

Makefile

@@ -3,7 +3,7 @@ include ./Makefile.system
 
 BLASDIRS = interface driver/level2 driver/level3 driver/others
 
-ifndef DYNAMIC_ARCH
+ifneq ($(DYNAMIC_ARCH), 1)
 BLASDIRS += kernel 
 endif
 
@@ -26,7 +26,7 @@ endif
 
 SUBDIRS_ALL = $(SUBDIRS) test ctest utest exports benchmark ../laswp ../bench
 
-.PHONY : all libs netlib test ctest shared install 
+.PHONY : all libs netlib test ctest shared install
 .NOTPARALLEL : all libs prof lapack-test install
 
 all :: libs netlib tests shared
@@ -80,6 +80,7 @@ endif
 	@echo
 
 shared :
+ifndef NO_SHARED
 ifeq ($(OSNAME), Linux)
 	$(MAKE) -C exports so
 	-ln -fs $(LIBSONAME) $(LIBPREFIX).so
@@ -99,11 +100,10 @@ ifeq ($(OSNAME), Darwin)
 endif
 ifeq ($(OSNAME), WINNT)
 	$(MAKE) -C exports dll
-	-ln -fs $(LIBDLLNAME) $(LIBPREFIX).dll
 endif
 ifeq ($(OSNAME), CYGWIN_NT)
 	$(MAKE) -C exports dll
-	-ln -fs $(LIBDLLNAME) $(LIBPREFIX).dll
+endif
 endif
 
 tests :
@@ -147,7 +147,7 @@ ifeq ($(EXPRECISION), 1)
 	echo "#define EXPRECISION">> config_last.h
 endif
 ## 
-ifdef DYNAMIC_ARCH
+ifeq ($(DYNAMIC_ARCH), 1)
 	  $(MAKE) -C kernel commonlibs || exit 1
 	for d in $(DYNAMIC_CORE) ; \
 	do  $(MAKE) GOTOBLAS_MAKEFILE= -C kernel TARGET_CORE=$$d kernel || exit 1 ;\
@@ -165,7 +165,7 @@ prof_blas :
 	  $(MAKE) -C $$d prof || exit 1 ; \
 	fi; \
 	done
-ifdef DYNAMIC_ARCH
+ifeq ($(DYNAMIC_ARCH), 1)
 	  $(MAKE) -C kernel commonprof || exit 1
 endif
 
@@ -184,7 +184,7 @@ hpl :
 	  $(MAKE) -C $$d $(@F) || exit 1 ; \
 	fi; \
 	done
-ifdef DYNAMIC_ARCH
+ifeq ($(DYNAMIC_ARCH), 1)
 	  $(MAKE) -C kernel commonlibs || exit 1
 	for d in $(DYNAMIC_CORE) ; \
 	do  $(MAKE) GOTOBLAS_MAKEFILE= -C kernel TARGET_CORE=$$d kernel || exit 1 ;\
@@ -203,47 +203,73 @@ ifeq ($(NO_LAPACK), 1)
 netlib : 
 
 else
-netlib : lapack-3.4.0 patch.for_lapack-3.4.0 lapack-3.4.0/make.inc
+netlib : lapack-3.4.2 patch.for_lapack-3.4.2 $(NETLIB_LAPACK_DIR)/make.inc
 ifndef NOFORTRAN
-	-@$(MAKE) -C lapack-3.4.0 lapacklib
+	-@$(MAKE) -C $(NETLIB_LAPACK_DIR) lapacklib
+endif
+ifndef NO_LAPACKE
+	-@$(MAKE) -C $(NETLIB_LAPACK_DIR) lapackelib
 endif
 endif
 
-prof_lapack : lapack-3.4.0 lapack-3.4.0/make.inc
-	-@$(MAKE) -C lapack-3.4.0 lapack_prof
+prof_lapack : lapack-3.4.2 $(NETLIB_LAPACK_DIR)/make.inc
+	-@$(MAKE) -C $(NETLIB_LAPACK_DIR) lapack_prof
 
-lapack-3.4.0/make.inc :
+$(NETLIB_LAPACK_DIR)/make.inc :
 ifndef NOFORTRAN
-	-@echo "FORTRAN   = $(FC)" > lapack-3.4.0/make.inc
-	-@echo "OPTS      = $(FFLAGS)" >> lapack-3.4.0/make.inc
-	-@echo "POPTS     = $(FPFLAGS)" >> lapack-3.4.0/make.inc
-	-@echo "NOOPT     = $(FFLAGS) -O0" >> lapack-3.4.0/make.inc
-	-@echo "PNOOPT     = $(FPFLAGS) -O0" >> lapack-3.4.0/make.inc
-	-@echo "LOADOPTS  = $(FFLAGS) $(EXTRALIB)" >> lapack-3.4.0/make.inc
-	-@echo "ARCH      = $(AR)" >> lapack-3.4.0/make.inc
-	-@echo "RANLIB    = $(RANLIB)" >> lapack-3.4.0/make.inc
-	-@echo "LAPACKLIB = ../$(LIBNAME)" >> lapack-3.4.0/make.inc
-	-@echo "LAPACKLIB_P = ../$(LIBNAME_P)" >> lapack-3.4.0/make.inc
-	-@echo "SUFFIX     = $(SUFFIX)" >> lapack-3.4.0/make.inc
-	-@echo "PSUFFIX    = $(PSUFFIX)" >> lapack-3.4.0/make.inc
-#	-@echo "CEXTRALIB  = $(CEXTRALIB)" >> lapack-3.4.0/make.inc
-	-@cat  make.inc >> lapack-3.4.0/make.inc
-endif
-
-lapack-3.4.0 : lapack-3.4.0.tgz
+	-@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 "LOADOPTS    = $(FFLAGS) $(EXTRALIB)" >> $(NETLIB_LAPACK_DIR)/make.inc
+	-@echo "CC          = $(CC)" >> $(NETLIB_LAPACK_DIR)/make.inc
+ifdef INTERFACE64
+	-@echo "override CFLAGS      = $(CFLAGS) -DHAVE_LAPACK_CONFIG_H  -DLAPACK_ILP64" >> $(NETLIB_LAPACK_DIR)/make.inc
+else
+	-@echo "override CFLAGS      = $(CFLAGS)" >> $(NETLIB_LAPACK_DIR)/make.inc
+endif
+	-@echo "ARCH        = $(AR)" >> $(NETLIB_LAPACK_DIR)/make.inc
+	-@echo "ARCHFLAGS   = -ru" >> $(NETLIB_LAPACK_DIR)/make.inc
+	-@echo "RANLIB      = $(RANLIB)" >> $(NETLIB_LAPACK_DIR)/make.inc
+	-@echo "LAPACKLIB   = ../$(LIBNAME)" >> $(NETLIB_LAPACK_DIR)/make.inc
+	-@echo "LAPACKELIB  = ../$(LIBNAME)" >> $(NETLIB_LAPACK_DIR)/make.inc
+	-@echo "LAPACKLIB_P = ../$(LIBNAME_P)" >> $(NETLIB_LAPACK_DIR)/make.inc
+	-@echo "SUFFIX      = $(SUFFIX)" >> $(NETLIB_LAPACK_DIR)/make.inc
+	-@echo "PSUFFIX     = $(PSUFFIX)" >> $(NETLIB_LAPACK_DIR)/make.inc
+	-@echo "CEXTRALIB   = $(EXTRALIB)" >> $(NETLIB_LAPACK_DIR)/make.inc
+	-@cat  make.inc >> $(NETLIB_LAPACK_DIR)/make.inc
+endif
+
+lapack-3.4.2 : lapack-3.4.2.tgz
 ifndef NOFORTRAN
-	@if test `$(MD5SUM) lapack-3.4.0.tgz | $(AWK) '{print $$1}'` = 02d5706ec03ba885fc246e5fa10d8c70; then \
+ifndef NO_LAPACK
+	@if test `$(MD5SUM) lapack-3.4.2.tgz | $(AWK) '{print $$1}'` = 61bf1a8a4469d4bdb7604f5897179478; then \
 		echo $(TAR) zxf $< ;\
-		$(TAR) zxf $< && (cd lapack-3.4.0; $(PATCH) -p1 < ../patch.for_lapack-3.4.0) ;\
+		$(TAR) zxf $< && (cd $(NETLIB_LAPACK_DIR); $(PATCH) -p1 < ../patch.for_lapack-3.4.2) ;\
+		rm -f $(NETLIB_LAPACK_DIR)/lapacke/make.inc ;\
 	else \
-		echo "	lapack-3.4.0.tgz check sum is wrong (Please use orignal)." ;\
-		rm -rf lapack-3.4.0 ;\
+		rm -rf $(NETLIB_LAPACK_DIR) ;\
+		echo "	Cannot download lapack-3.4.2.tgz or the MD5 check sum is wrong (Please use orignal)."; \
+		exit 1; \
 	fi
 endif
+endif
+
+LAPACK_URL=http://www.netlib.org/lapack/lapack-3.4.2.tgz
 
-lapack-3.4.0.tgz :
+lapack-3.4.2.tgz :
 ifndef NOFORTRAN
-	-wget http://www.netlib.org/lapack/lapack-3.4.0.tgz
+#http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
+ifeq ($(OSNAME), $(filter $(OSNAME),Darwin NetBSD))
+	curl -O $(LAPACK_URL)
+else
+ifeq ($(OSNAME), FreeBSD)
+	fetch $(LAPACK_URL)
+else
+	wget -O $@ $(LAPACK_URL)
+endif
+endif
 endif
 
 large.tgz : 
@@ -256,21 +282,21 @@ ifndef NOFORTRAN
 	-wget http://www.netlib.org/lapack/timing/timing.tgz
 endif
 
-lapack-timing : lapack-3.4.0 large.tgz timing.tgz
+lapack-timing : lapack-3.4.2 large.tgz timing.tgz
 ifndef NOFORTRAN
-	(cd lapack-3.4.0; $(TAR) zxf ../timing.tgz TIMING)
-	(cd lapack-3.4.0/TIMING; $(TAR) zxf ../../large.tgz )
-	make -C lapack-3.4.0 tmglib
-	make -C lapack-3.4.0/TIMING
+	(cd $(NETLIB_LAPACK_DIR); $(TAR) zxf ../timing.tgz TIMING)
+	(cd $(NETLIB_LAPACK_DIR)/TIMING; $(TAR) zxf ../../large.tgz )
+	make -C $(NETLIB_LAPACK_DIR) tmglib
+	make -C $(NETLIB_LAPACK_DIR)/TIMING
 endif
 
 
 lapack-test :
-	$(MAKE) -C lapack-3.4.0 tmglib
-	$(MAKE) -C lapack-3.4.0/TESTING xeigtstc xeigtstd xeigtsts xeigtstz xlintstc xlintstd xlintstds xlintsts xlintstz xlintstzc
-	@rm	-f lapack-3.4.0/TESTING/*.out
-	$(MAKE) -j 1 -C lapack-3.4.0/TESTING
-	$(GREP) failed lapack-3.4.0/TESTING/*.out
+	$(MAKE) -C $(NETLIB_LAPACK_DIR) tmglib
+	$(MAKE) -C $(NETLIB_LAPACK_DIR)/TESTING xeigtstc xeigtstd xeigtsts xeigtstz xlintstc xlintstd xlintstds xlintsts xlintstz xlintstzc
+	@rm	-f $(NETLIB_LAPACK_DIR)/TESTING/*.out
+	$(MAKE) -j 1 -C $(NETLIB_LAPACK_DIR)/TESTING
+	$(GREP) failed $(NETLIB_LAPACK_DIR)/TESTING/*.out
 
 dummy :
 
@@ -288,10 +314,10 @@ clean ::
 #endif
 	@$(MAKE) -C reference clean
 	@rm -f *.$(LIBSUFFIX) *.so *~ *.exe getarch getarch_2nd *.dll *.lib *.$(SUFFIX) *.dwf $(LIBPREFIX).$(LIBSUFFIX) $(LIBPREFIX)_p.$(LIBSUFFIX) $(LIBPREFIX).so.$(MAJOR_VERSION) *.lnk myconfig.h
-	@rm -f Makefile.conf config.h Makefile_kernel.conf config_kernel.h st* *.dylib
-	@if test -d lapack-3.4.0; then \
-	echo deleting lapack-3.4.0; \
-	rm -rf lapack-3.4.0 ;\
+	@rm -f Makefile.conf config.h cblas_noconst.h Makefile_kernel.conf config_kernel.h st* *.dylib
+	@if test -d $(NETLIB_LAPACK_DIR); then \
+	echo deleting $(NETLIB_LAPACK_DIR); \
+	rm -rf $(NETLIB_LAPACK_DIR) ;\
 	fi
 	@rm -f *.grd Makefile.conf_last config_last.h
-	@echo Done.
+	@echo Done.

Makefile.install

@@ -23,7 +23,7 @@ install : 	lib.grd
 	@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_INCLUDE_DIR)/openblas_config.h
+	@echo \#endif  \/\* OPENBLAS_CONFIG_H \*\/ >> $(OPENBLAS_INCLUDE_DIR)/openblas_config.h
 
 	@echo Generating f77blas.h in $(OPENBLAS_INCLUDE_DIR)
 	@echo \#ifndef OPENBLAS_F77BLAS_H > $(OPENBLAS_INCLUDE_DIR)/f77blas.h
@@ -32,8 +32,18 @@ install : 	lib.grd
 	@cat common_interface.h >> $(OPENBLAS_INCLUDE_DIR)/f77blas.h
 	@echo \#endif >> $(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
+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
+endif
 
 #for install static library 
 	@echo Copy the static library to $(OPENBLAS_LIBRARY_DIR)
@@ -61,11 +71,9 @@ ifeq ($(OSNAME), Darwin)
 endif
 ifeq ($(OSNAME), WINNT)
 	-cp $(LIBDLLNAME) $(OPENBLAS_LIBRARY_DIR)
-	-ln -fs $(OPENBLAS_LIBRARY_DIR)/$(LIBDLLNAME) $(OPENBLAS_LIBRARY_DIR)/$(LIBPREFIX).dll
 endif
 ifeq ($(OSNAME), CYGWIN_NT)
 	-cp $(LIBDLLNAME) $(OPENBLAS_LIBRARY_DIR)
-	-ln -fs $(OPENBLAS_LIBRARY_DIR)/$(LIBDLLNAME) $(OPENBLAS_LIBRARY_DIR)/$(LIBPREFIX).dll
 endif
 
 	@echo Install OK!

Makefile.getarch → Makefile.prebuild

@@ -1,3 +1,5 @@
+# This is triggered by Makefile.system and runs before any of the code is built.
+
 export BINARY
 export USE_OPENMP
 
@@ -15,7 +17,7 @@ ifdef CPUIDEMU
 EXFLAGS = -DCPUIDEMU -DVENDOR=99
 endif
 
-all: getarch_2nd
+all: getarch_2nd cblas_noconst.h
 	./getarch_2nd  0 >> $(TARGET_MAKE)
 	./getarch_2nd  1 >> $(TARGET_CONF)
 
@@ -36,4 +38,7 @@ 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:

Makefile.rule

@@ -3,7 +3,7 @@
 #
 
 # This library's version
-VERSION = 0.1.0
+VERSION = 0.2.6
 
 # If you set the suffix, the library name will be libopenblas_$(LIBNAMESUFFIX).a
 # and libopenblas_$(LIBNAMESUFFIX).so. Meanwhile, the soname in shared library 
@@ -24,10 +24,13 @@ VERSION = 0.1.0
 # Fortran compiler. Default is g77.
 # FC = gfortran
 
-# Even you can specify cross compiler
+# Even you can specify cross compiler. Meanwhile, please set HOSTCC.
 # CC = x86_64-w64-mingw32-gcc
 # FC = x86_64-w64-mingw32-gfortran
 
+# If you use the cross compiler, please set this host compiler.
+# HOSTCC = gcc
+
 # If you need 32bit binary, define BINARY=32, otherwise define BINARY=64
 # BINARY=64
 
@@ -45,12 +48,19 @@ VERSION = 0.1.0
 # automatically detected by the the script.
 # NUM_THREADS = 24
 
+# if you don't need generate the shared library, please comment it in.
+# NO_SHARED = 1
+
 # If you don't need CBLAS interface, please comment it in.
 # NO_CBLAS = 1
 
-# If you don't need LAPACK, please comment it in.
+# If you don't need LAPACK, please comment it in. 
+# If you set NO_LAPACK=1, the library automatically sets NO_LAPACKE=1.
 # NO_LAPACK = 1
 
+# If you don't need LAPACKE (C Interface to LAPACK), please comment it in.
+# NO_LAPACKE = 1
+
 # If you want to use legacy threaded Level 3 implementation.
 # USE_SIMPLE_THREADED_LEVEL3 = 1
 
@@ -67,6 +77,10 @@ VERSION = 0.1.0
 # If you want to disable CPU/Memory affinity on Linux.
 # NO_AFFINITY = 1
 
+# Don't use AVX kernel on Sandy Bridge. It is compatible with old compilers 
+# and OS. However, the performance is low.
+# NO_AVX = 1
+
 # If you would like to know minute performance report of GotoBLAS.
 # FUNCTION_PROFILE = 1
 
@@ -90,8 +104,8 @@ VERSION = 0.1.0
 
 # If any gemm arguement m, n or k is less or equal this threshold, gemm will be execute 
 # with single thread. You can use this flag to avoid the overhead of multi-threading 
-# in small matrix sizes. The default value is 4.
-# GEMM_MULTITHREAD_THRESHOLD = 4
+# in small matrix sizes. The default value is 50.
+# GEMM_MULTITHREAD_THRESHOLD = 50
 
 # If you need santy check by comparing reference BLAS. It'll be very
 # slow (Not implemented yet).
@@ -104,19 +118,16 @@ VERSION = 0.1.0
 # The installation directory.
 # PREFIX = /opt/OpenBLAS
 
-# Common Optimization Flag; -O2 is enough.
-# DEBUG = 1
-
-ifeq ($(DEBUG), 1)
-COMMON_OPT += -g
-# -DDEBUG
-else
-COMMON_OPT += -O2
-endif
+# Common Optimization Flag; 
+# The default -O2 is enough.
+# COMMON_OPT = -O2
 
 # Profiling flags
 COMMON_PROF = -pg
 
+# Build Debug version
+# DEBUG = 1
+
 #
 #  End of user configuration 
 #

Makefile.system

@@ -9,8 +9,20 @@ ifndef TOPDIR
 TOPDIR = .
 endif
 
+ifndef NETLIB_LAPACK_DIR
+NETLIB_LAPACK_DIR = $(TOPDIR)/lapack-3.4.2
+endif
+
 # Default C compiler
+# - Only set if not specified on the command line or inherited from the environment.
+# - CC is an implicit variable so neither '?=' or 'ifndef' can be used.
+#   http://stackoverflow.com/questions/4029274/mingw-and-make-variables
+# - Default value is 'cc' which is not always a valid command (e.g. MinGW).
+ifeq ($(origin CC),default)
 CC = gcc
+endif
+
+# Default Fortran compiler (FC) is selected by f_check.
 
 ifndef MAKEFILE_RULE
 include $(TOPDIR)/Makefile.rule
@@ -41,16 +53,24 @@ GETARCH_FLAGS	+= -DUSE64BITINT
 endif
 
 ifndef GEMM_MULTITHREAD_THRESHOLD
-GEMM_MULTITHREAD_THRESHOLD=4
+GEMM_MULTITHREAD_THRESHOLD=50
 endif
 GETARCH_FLAGS	+= -DGEMM_MULTITHREAD_THRESHOLD=$(GEMM_MULTITHREAD_THRESHOLD) 
 
+ifeq ($(NO_AVX), 1)
+GETARCH_FLAGS	+= -DNO_AVX
+endif
+
+ifeq ($(DEBUG), 1)
+GETARCH_FLAGS	+= -g
+endif
+
 # This operation is expensive, so execution should be once.
 ifndef GOTOBLAS_MAKEFILE
 export GOTOBLAS_MAKEFILE = 1
 
 # Generating Makefile.conf and config.h
-DUMMY := $(shell $(MAKE) -C $(TOPDIR) -f Makefile.getarch CC="$(CC)" FC="$(FC)" HOSTCC="$(HOSTCC)" CFLAGS="$(GETARCH_FLAGS)" BINARY=$(BINARY) USE_OPENMP=$(USE_OPENMP) TARGET_CORE=$(TARGET_CORE) all)
+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)
 
 ifndef TARGET_CORE
 include $(TOPDIR)/Makefile.conf
@@ -101,6 +121,15 @@ DLLWRAP = $(CROSS_SUFFIX)dllwrap
 
 ifeq ($(OSNAME), Darwin)
 export MACOSX_DEPLOYMENT_TARGET=10.2
+MD5SUM = md5 -r
+endif
+
+ifeq ($(OSNAME), FreeBSD)
+MD5SUM = md5 -r
+endif
+
+ifeq ($(OSNAME), NetBSD)
+MD5SUM = md5 -n
 endif
 
 ifeq ($(OSNAME), Linux)
@@ -120,6 +149,26 @@ EXTRALIB        += -defaultlib:advapi32
 SUFFIX  = obj
 PSUFFIX = pobj
 LIBSUFFIX = lib
+ifeq ($(C_COMPILER), GCC)
+#Test for supporting MS_ABI
+GCCVERSIONGTEQ4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \>= 4)
+GCCVERSIONGT4 := $(shell expr `$(CC) -dumpversion | cut -f1 -d.` \> 4)
+GCCMINORVERSIONGTEQ7 := $(shell expr `$(CC) -dumpversion | cut -f2 -d.` \>= 7)
+ifeq ($(GCCVERSIONGT4), 1)
+# GCC Majar version > 4
+# It is compatible with MSVC ABI. 
+CCOMMON_OPT	+= -DMS_ABI
+endif
+
+ifeq ($(GCCVERSIONGTEQ4), 1)
+ifeq ($(GCCMINORVERSIONGTEQ7), 1)
+# GCC Version >=4.7
+# It is compatible with MSVC ABI. 
+CCOMMON_OPT	+= -DMS_ABI
+endif
+endif
+
+endif
 endif
 
 ifeq ($(OSNAME), Interix)
@@ -223,14 +272,20 @@ endif
 endif
 
 
-ifdef DYNAMIC_ARCH
+ifeq ($(DYNAMIC_ARCH), 1)
 ifeq ($(ARCH), x86)
 DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
-	       CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA ATOM NANO
+	       CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
+ifneq ($(NO_AVX), 1)
+DYNAMIC_CORE += SANDYBRIDGE BULLDOZER
+endif
 endif
 
 ifeq ($(ARCH), x86_64)
-DYNAMIC_CORE = PRESCOTT CORE2 PENRYN DUNNINGTON NEHALEM OPTERON OPTERON_SSE3 BARCELONA ATOM NANO
+DYNAMIC_CORE = PRESCOTT CORE2 PENRYN DUNNINGTON NEHALEM OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
+ifneq ($(NO_AVX), 1)
+DYNAMIC_CORE += SANDYBRIDGE BULLDOZER
+endif
 endif
 
 ifndef DYNAMIC_CORE
@@ -459,11 +514,28 @@ ifdef INTERFACE64
 FCOMMON_OPT += -i8
 endif
 endif
+
+ifeq ($(ARCH), mips64)
+ifndef BINARY64
+FCOMMON_OPT += -n32
+else
+FCOMMON_OPT += -n64
+endif
+ifeq ($(CORE), LOONGSON3A) 
+FCOMMON_OPT += -loongson3
+endif
+
+ifeq ($(CORE), LOONGSON3B) 
+FCOMMON_OPT += -loongson3
+endif
+
+else
 ifndef BINARY64
 FCOMMON_OPT += -m32
 else
 FCOMMON_OPT += -m64
 endif
+endif
 
 ifdef USE_OPENMP
 FEXTRALIB   += -lstdc++
@@ -472,12 +544,30 @@ endif
 endif
 
 ifeq ($(C_COMPILER), OPEN64)
+
+ifeq ($(ARCH), mips64)
+ifndef BINARY64
+CCOMMON_OPT += -n32
+else
+CCOMMON_OPT += -n64
+endif
+ifeq ($(CORE), LOONGSON3A) 
+CCOMMON_OPT += -loongson3
+endif
+
+ifeq ($(CORE), LOONGSON3B) 
+CCOMMON_OPT += -loongson3
+endif
+
+else
+
 ifndef BINARY64
 CCOMMON_OPT += -m32
 else
 CCOMMON_OPT += -m64
 endif
 endif
+endif
 
 ifeq ($(C_COMPILER), SUN)
 CCOMMON_OPT  += -w
@@ -533,6 +623,16 @@ endif
 
 ifeq ($(NO_LAPACK), 1)
 CCOMMON_OPT	+= -DNO_LAPACK
+#Disable LAPACK C interface
+NO_LAPACKE = 1
+endif
+
+ifeq ($(NO_LAPACKE), 1)
+CCOMMON_OPT	+= -DNO_LAPACKE
+endif
+
+ifeq ($(NO_AVX), 1)
+CCOMMON_OPT	+= -DNO_AVX
 endif
 
 ifdef SMP
@@ -651,17 +751,30 @@ PATCH	= patch
 GREP	= grep
 endif
 
+ifndef MD5SUM
 MD5SUM	= md5sum
+endif
+
 AWK	= awk
 
 REVISION = -r$(VERSION)
 MAJOR_VERSION = $(word 1,$(subst ., ,$(VERSION)))
 
-CFLAGS     = $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
-PFLAGS     = $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
+ifeq ($(DEBUG), 1)
+COMMON_OPT += -g
+endif
+
+ifndef COMMON_OPT
+COMMON_OPT = -O2
+endif
+
+
+override CFLAGS     += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR)
+override PFLAGS     += $(COMMON_OPT) $(CCOMMON_OPT) -I$(TOPDIR) -DPROFILE $(COMMON_PROF)
 
-FFLAGS     = $(COMMON_OPT) $(FCOMMON_OPT)
-FPFLAGS    = $(COMMON_OPT) $(FCOMMON_OPT) $(COMMON_PROF)
+override FFLAGS     += $(COMMON_OPT) $(FCOMMON_OPT)
+override FPFLAGS    += $(COMMON_OPT) $(FCOMMON_OPT) $(COMMON_PROF)
+#MAKEOVERRIDES =
 
 ifndef SUFFIX
 SUFFIX  = o
@@ -675,7 +788,7 @@ ifndef LIBSUFFIX
 LIBSUFFIX = a
 endif
 
-ifndef DYNAMIC_ARCH
+ifneq ($(DYNAMIC_ARCH), 1)
 ifndef SMP
 LIBNAME		= $(LIBPREFIX)_$(LIBCORE)$(REVISION).$(LIBSUFFIX)
 LIBNAME_P	= $(LIBPREFIX)_$(LIBCORE)$(REVISION)_p.$(LIBSUFFIX)
@@ -694,8 +807,8 @@ endif
 endif
 
 
+LIBDLLNAME   = $(LIBPREFIX).dll
 LIBSONAME    = $(LIBNAME:.$(LIBSUFFIX)=.so)
-LIBDLLNAME   = $(LIBNAME:.$(LIBSUFFIX)=.dll)
 LIBDYNNAME   = $(LIBNAME:.$(LIBSUFFIX)=.dylib)
 LIBDEFNAME   = $(LIBNAME:.$(LIBSUFFIX)=.def)
 LIBEXPNAME   = $(LIBNAME:.$(LIBSUFFIX)=.exp)
@@ -740,6 +853,7 @@ export HAVE_SSE4_1
 export HAVE_SSE4_2
 export HAVE_SSE4A
 export HAVE_SSE5
+export HAVE_AVX
 export KERNELDIR
 export FUNCTION_PROFILE
 export TARGET_CORE

Makefile.tail

@@ -22,19 +22,19 @@ BLASOBJS   += $(QBLASOBJS)   $(XBLASOBJS)
 BLASOBJS_P += $(QBLASOBJS_P) $(XBLASOBJS_P)
 endif
 
-$(SBLASOBJS) $(SBLASOBJS_P) : CFLAGS += -UDOUBLE  -UCOMPLEX
-$(DBLASOBJS) $(DBLASOBJS_P) : CFLAGS += -DDOUBLE  -UCOMPLEX
-$(QBLASOBJS) $(QBLASOBJS_P) : CFLAGS += -DXDOUBLE -UCOMPLEX
-$(CBLASOBJS) $(CBLASOBJS_P) : CFLAGS += -UDOUBLE  -DCOMPLEX
-$(ZBLASOBJS) $(ZBLASOBJS_P) : CFLAGS += -DDOUBLE  -DCOMPLEX
-$(XBLASOBJS) $(XBLASOBJS_P) : CFLAGS += -DXDOUBLE -DCOMPLEX
-
-$(SBLASOBJS_P) : CFLAGS += -DPROFILE $(COMMON_PROF)
-$(DBLASOBJS_P) : CFLAGS += -DPROFILE $(COMMON_PROF)
-$(QBLASOBJS_P) : CFLAGS += -DPROFILE $(COMMON_PROF)
-$(CBLASOBJS_P) : CFLAGS += -DPROFILE $(COMMON_PROF)
-$(ZBLASOBJS_P) : CFLAGS += -DPROFILE $(COMMON_PROF)
-$(XBLASOBJS_P) : CFLAGS += -DPROFILE $(COMMON_PROF)
+$(SBLASOBJS) $(SBLASOBJS_P) : override CFLAGS += -UDOUBLE  -UCOMPLEX
+$(DBLASOBJS) $(DBLASOBJS_P) : override CFLAGS += -DDOUBLE  -UCOMPLEX
+$(QBLASOBJS) $(QBLASOBJS_P) : override CFLAGS += -DXDOUBLE -UCOMPLEX
+$(CBLASOBJS) $(CBLASOBJS_P) : override CFLAGS += -UDOUBLE  -DCOMPLEX
+$(ZBLASOBJS) $(ZBLASOBJS_P) : override CFLAGS += -DDOUBLE  -DCOMPLEX
+$(XBLASOBJS) $(XBLASOBJS_P) : override CFLAGS += -DXDOUBLE -DCOMPLEX
+
+$(SBLASOBJS_P) : override CFLAGS += -DPROFILE $(COMMON_PROF)
+$(DBLASOBJS_P) : override CFLAGS += -DPROFILE $(COMMON_PROF)
+$(QBLASOBJS_P) : override CFLAGS += -DPROFILE $(COMMON_PROF)
+$(CBLASOBJS_P) : override CFLAGS += -DPROFILE $(COMMON_PROF)
+$(ZBLASOBJS_P) : override CFLAGS += -DPROFILE $(COMMON_PROF)
+$(XBLASOBJS_P) : override CFLAGS += -DPROFILE $(COMMON_PROF)
 
 libs    :: $(BLASOBJS) $(COMMONOBJS)
 	$(AR) $(ARFLAGS) -ru $(TOPDIR)/$(LIBNAME) $^

README

@@ -1,83 +0,0 @@
-OpenBLAS Readme
-
-1.Introduction
-OpenBLAS is an optimized BLAS library based on GotoBLAS2 1.13 BSD version. OpenBLAS is an open source project supported by Lab of Parallel Software and Computational Science, ISCAS.(http://www.rdcps.ac.cn)
-
-2.Intallation
-Download from project homepage. http://xianyi.github.com/OpenBLAS/
-Or,
-check out codes from git://github.com/xianyi/OpenBLAS.git
-1)Normal compile
-  (a) type "make" to detect the CPU automatically.
-  or
-  (b) type "make TARGET=xxx" to set target CPU, e.g. "make TARGET=NEHALEM". The full target list is in file TargetList.txt.
-
-2)Cross compile
-Please set CC and FC with the cross toolchains. Then, set HOSTCC with your host C compiler. At last, set TARGET explicitly.
-
-examples:
-On X86 box, compile this library for loongson3a CPU.
-make BINARY=64 CC=mips64el-unknown-linux-gnu-gcc FC=mips64el-unknown-linux-gnu-gfortran HOSTCC=gcc TARGET=LOONGSON3A
-
-3)Debug version
-make DEBUG=1
-
-4)Intall to the directory (Optional)
-e.g.
-make install PREFIX=your_installation_directory
-The default directory is /opt/OpenBLAS
-
-3.Support CPU & OS
-Please read GotoBLAS_01Readme.txt
-
-Additional support CPU:
-x86_64:
-	Intel Xeon 56xx (Westmere) //Used GotoBLAS2 Nehalem codes.
-MIPS64:
-	ICT Loongson 3A //The initial version used GotoBLAS2 MIPS64 kernels. Thus, the performance is not good.
-
-4.Usages
-Link with libopenblas.a or -lopenblas for shared library.
-
-4.1 Set the number of threads with environment variables. for example,
-export OPENBLAS_NUM_THREADS=4
- or
-export GOTO_NUM_THREADS=4
- or 
-export OMP_NUM_THREADS=4
-
-The priorities are OPENBLAS_NUM_THREADS > GOTO_NUM_THREADS > OMP_NUM_THREADS.
-
-If you compile this lib with USE_OPENMP=1, you should set OMP_NUM_THREADS environment variable. OpenBLAS ignores OPENBLAS_NUM_THREADS and GOTO_NUM_THREADS with USE_OPENMP=1.
-
-4.2 Set the number of threads with calling functions. for example,
-void goto_set_num_threads(int num_threads);
-or
-void openblas_set_num_threads(int num_threads);
-
-If you compile this lib with USE_OPENMP=1, you should use the above functions, too.
-
-5.Report Bugs
-Please add a issue in https://github.com/xianyi/OpenBLAS/issues
-
-6.To-Do List:
-Optimization on ICT Loongson 3A CPU
-
-7.Contact
-OpenBLAS users mailing list: http://list.rdcps.ac.cn/mailman/listinfo/openblas
-
-8.ChangeLog
-Please see Changelog.txt to obtain the differences between GotoBLAS2 1.13 BSD version.
-
-9.Known Issues
-* The number of CPUs/Cores should less than or equal to 8*sizeof(unsigned long). On 64 bits, the limit 
-  is 64. On 32 bits, it is 32.
-* On Loongson 3A. make test would be failed because of pthread_create error. The error code is EAGAIN. However, it will be OK when you run the same testcase on shell. I don't think this is a bug in OpenBLAS. 
-
-10. Specification of Git Branches
-We used the git branching model in this article (http://nvie.com/posts/a-successful-git-branching-model/). 
-Now, there are 4 branches in github.com.
-  * The master branch. This a main branch to reflect a production-ready state.
-  * The develop branch. This a main branch to reflect a state with the latest delivered development changes for the next release.
-  * The loongson3a branch. This is a feature branch. We develop Loongson3A codes on this branch. We will merge this feature to develop branch in future.
-  * The gh-pages branch. This is for web pages

README.md

@@ -0,0 +1,117 @@
+# OpenBLAS
+
+## Introduction
+OpenBLAS is an optimized BLAS library based on GotoBLAS2 1.13 BSD version. OpenBLAS is an open source project supported by Lab of Parallel Software and Computational Science, ISCAS <http://www.rdcps.ac.cn>.
+
+Please read the documents on OpenBLAS wiki pages <http://github.com/xianyi/OpenBLAS/wiki>.
+
+## Installation
+Download from project homepage. http://xianyi.github.com/OpenBLAS/
+
+Or, check out codes from git://github.com/xianyi/OpenBLAS.git
+### Normal compile
+  * type "make" to detect the CPU automatically.
+  or
+  * type "make TARGET=xxx" to set target CPU, e.g. "make TARGET=NEHALEM". The full target list is in file TargetList.txt.
+
+### Cross compile
+Please set CC and FC with the cross toolchains. Then, set HOSTCC with your host C compiler. At last, set TARGET explicitly.
+
+Examples:
+
+On X86 box, compile this library for loongson3a CPU.
+
+    make BINARY=64 CC=mips64el-unknown-linux-gnu-gcc FC=mips64el-unknown-linux-gnu-gfortran HOSTCC=gcc TARGET=LOONGSON3A
+
+On X86 box, compile this library for loongson3a CPU with loongcc (based on Open64) compiler.
+
+    make CC=loongcc FC=loongf95 HOSTCC=gcc TARGET=LOONGSON3A CROSS=1 CROSS_SUFFIX=mips64el-st-linux-gnu-   NO_LAPACKE=1 NO_SHARED=1 BINARY=32
+
+### Debug version
+
+    make DEBUG=1
+
+### Install to the directory (Optional)
+
+Example:
+
+    make install PREFIX=your_installation_directory
+
+The default directory is /opt/OpenBLAS
+
+## Support CPU & OS
+Please read GotoBLAS_01Readme.txt
+
+### Additional support CPU:
+
+#### x86/x86-64:
+- **Intel Xeon 56xx (Westmere)**: Used GotoBLAS2 Nehalem codes.
+- **Intel Sandy Bridge**: Optimized Level-3 BLAS with AVX on x86-64.
+- **AMD Bobcat**: Used GotoBLAS2 Barcelona codes.
+- **AMD Bulldozer**: x86-64 S/DGEMM AVX kernels. (Thank Werner Saar)
+
+#### MIPS64:
+- **ICT Loongson 3A**: Optimized Level-3 BLAS and the part of Level-1,2.
+- **ICT Loongson 3B**: Experimental
+
+### Support OS:
+- **GNU/Linux**
+- **MingWin/Windows**: Please read <https://github.com/xianyi/OpenBLAS/wiki/How-to-use-OpenBLAS-in-Microsoft-Visual-Studio>.
+- **Darwin/Mac OS X**: Experimental. Although GotoBLAS2 supports Darwin, we are the beginner on Mac OS X.
+- **FreeBSD**: Supportted by community. We didn't test the library on this OS.
+
+## Usages
+Link with libopenblas.a or -lopenblas for shared library.
+
+### Set the number of threads with environment variables. 
+
+Examples:
+
+    export OPENBLAS_NUM_THREADS=4
+
+ or
+
+    export GOTO_NUM_THREADS=4
+
+ or 
+
+    export OMP_NUM_THREADS=4
+
+The priorities are OPENBLAS_NUM_THREADS > GOTO_NUM_THREADS > OMP_NUM_THREADS.
+
+If you compile this lib with USE_OPENMP=1, you should set OMP_NUM_THREADS environment variable. OpenBLAS ignores OPENBLAS_NUM_THREADS and GOTO_NUM_THREADS with USE_OPENMP=1.
+
+### Set the number of threads on runtime. 
+
+We provided the below functions to controll the number of threads on runtime.
+
+    void goto_set_num_threads(int num_threads);
+
+    void openblas_set_num_threads(int num_threads);
+
+If you compile this lib with USE_OPENMP=1, you should use the above functions, too.
+
+## Report Bugs
+Please add a issue in https://github.com/xianyi/OpenBLAS/issues
+
+## Contact
+OpenBLAS users mailing list: http://list.rdcps.ac.cn/mailman/listinfo/openblas
+
+## ChangeLog
+Please see Changelog.txt to obtain the differences between GotoBLAS2 1.13 BSD version.
+
+## Troubleshooting
+* Please read [Faq](https://github.com/xianyi/OpenBLAS/wiki/Faq) at first.
+* Please use gcc version 4.6 and above to compile Sandy Bridge AVX kernels on Linux/MingW/BSD.
+* Please use Clang version 3.1 and above to compile the library on Sandy Bridge microarchitecture. The Clang 3.0 will generate the wrong AVX binary code.
+* The number of CPUs/Cores should less than or equal to 256. 
+* On Linux, OpenBLAS sets the processor affinity by default. This may cause [the conflict with R parallel](https://stat.ethz.ch/pipermail/r-sig-hpc/2012-April/001348.html). You can build the library with NO_AFFINITY=1.
+* On Loongson 3A. make test would be failed because of pthread_create error. The error code is EAGAIN. However, it will be OK when you run the same testcase on shell. 
+
+## Specification of Git Branches
+We used the git branching model in this article (http://nvie.com/posts/a-successful-git-branching-model/). 
+Now, there are 4 branches in github.com.
+  * The master branch. This a main branch to reflect a production-ready state.
+  * The develop branch. This a main branch to reflect a state with the latest delivered development changes for the next release.
+  * The loongson3a branch. This is a feature branch. We develop Loongson3A codes on this branch. We will merge this feature to develop branch in future.
+  * The gh-pages branch. This is for web pages

TargetList.txt

@@ -8,8 +8,8 @@ Supported List:
 1.X86/X86_64
 a)Intel CPU:
 P2
-COPPERMINE
 KATMAI
+COPPERMINE
 NORTHWOOD
 PRESCOTT
 BANIAS
@@ -18,6 +18,7 @@ CORE2
 PENRYN
 DUNNINGTON
 NEHALEM
+SANDYBRIDGE
 ATOM
 
 b)AMD CPU:
@@ -27,6 +28,8 @@ OPTERON_SSE3
 BARCELONA
 SHANGHAI
 ISTANBUL
+BOBCAT
+BULLDOZER
 
 c)VIA CPU:
 SSE_GENERIC
@@ -47,6 +50,7 @@ CELL
 3.MIPS64 CPU:
 SICORTEX
 LOONGSON3A
+LOONGSON3B
 
 4.IA64 CPU:
 ITANIUM2

c_check

@@ -43,14 +43,14 @@ $compiler = DEC       if ($data =~ /COMPILER_DEC/);
 $compiler = GCC       if ($compiler eq "");
 
 $os = Linux           if ($data =~ /OS_LINUX/);
-$os = FreeBSD         if ($data =~ /OS_FreeBSD/);
-$os = NetBSD          if ($data =~ /OS_NetBSD/);
-$os = Darwin          if ($data =~ /OS_Darwin/);
-$os = SunOS           if ($data =~ /OS_SunOS/);
+$os = FreeBSD         if ($data =~ /OS_FREEBSD/);
+$os = NetBSD          if ($data =~ /OS_NETBSD/);
+$os = Darwin          if ($data =~ /OS_DARWIN/);
+$os = SunOS           if ($data =~ /OS_SUNOS/);
 $os = AIX             if ($data =~ /OS_AIX/);
 $os = osf             if ($data =~ /OS_OSF/);
 $os = WINNT           if ($data =~ /OS_WINNT/);
-$os = CYGWIN_NT       if ($data =~ /OS_CYGWIN/);
+$os = CYGWIN_NT       if ($data =~ /OS_CYGWIN_NT/);
 $os = Interix         if ($data =~ /OS_INTERIX/);
 
 $architecture = x86    if ($data =~ /ARCH_X86/);
@@ -174,6 +174,8 @@ $linker_a = "";
     $link =~ s/\-Y\sP\,/\-Y/g;
     
     @flags = split(/[\s\,\n]/, $link);
+    # remove leading and trailing quotes from each flag.
+    @flags = map {s/^['"]|['"]$//g; $_} @flags;
 
     foreach $flags (@flags) {
 	if (

cblas.h

@@ -1,287 +1,293 @@
 #ifndef CBLAS_H
 #define CBLAS_H
 
+#include <stddef.h>
+#include "common.h"
+
 #ifdef __cplusplus
 extern "C" {
 	/* Assume C declarations for C++ */
 #endif  /* __cplusplus */
 
-#include <stddef.h>
-#include "common.h"
+/*Set the number of threads on runtime.*/
+void openblas_set_num_threads(int num_threads);
+void goto_set_num_threads(int num_threads);
+
+/*Get the build configure on runtime.*/
+char* openblas_get_config(void);
 
 #define CBLAS_INDEX size_t
 
-enum CBLAS_ORDER     {CblasRowMajor=101, CblasColMajor=102};
-enum CBLAS_TRANSPOSE {CblasNoTrans=111, CblasTrans=112, CblasConjTrans=113, CblasConjNoTrans=114};
-enum CBLAS_UPLO      {CblasUpper=121, CblasLower=122};
-enum CBLAS_DIAG      {CblasNonUnit=131, CblasUnit=132};
-enum CBLAS_SIDE      {CblasLeft=141, CblasRight=142};
-
-float  cblas_sdsdot(blasint n, float, float *x, blasint incx, float *y, blasint incy);
-double cblas_dsdot (blasint n, float *x, blasint incx, float *y, blasint incy);
-float  cblas_sdot(blasint n, float  *x, blasint incx, float  *y, blasint incy);
-double cblas_ddot(blasint n, double *x, blasint incx, double *y, blasint incy);
-
-float  _Complex cblas_cdotu(blasint n, float  *x, blasint incx, float  *y, blasint incy);
-float  _Complex cblas_cdotc(blasint n, float  *x, blasint incx, float  *y, blasint incy);
-double _Complex cblas_zdotu(blasint n, double *x, blasint incx, double *y, blasint incy);
-double _Complex cblas_zdotc(blasint n, double *x, blasint incx, double *y, blasint incy);
-
-void  cblas_cdotu_sub(blasint n, float  *x, blasint incx, float  *y, blasint incy, float  _Complex *ret);
-void  cblas_cdotc_sub(blasint n, float  *x, blasint incx, float  *y, blasint incy, float  _Complex *ret);
-void  cblas_zdotu_sub(blasint n, double *x, blasint incx, double *y, blasint incy, double _Complex *ret);
-void  cblas_zdotc_sub(blasint n, double *x, blasint incx, double *y, blasint incy, double _Complex *ret);
-
-float  cblas_sasum (blasint n, float  *x, blasint incx);
-double cblas_dasum (blasint n, double *x, blasint incx);
-float  cblas_scasum(blasint n, float  *x, blasint incx);
-double cblas_dzasum(blasint n, double *x, blasint incx);
-
-float  cblas_snrm2 (blasint N, float  *X, blasint incX);
-double cblas_dnrm2 (blasint N, double *X, blasint incX);
-float  cblas_scnrm2(blasint N, float  *X, blasint incX);
-double cblas_dznrm2(blasint N, double *X, blasint incX);
-
-CBLAS_INDEX cblas_isamax(blasint n, float  *x, blasint incx);
-CBLAS_INDEX cblas_idamax(blasint n, double *x, blasint incx);
-CBLAS_INDEX cblas_icamax(blasint n, float  *x, blasint incx);
-CBLAS_INDEX cblas_izamax(blasint n, double *x, blasint incx);
-
-void cblas_saxpy(blasint n, float, float *x, blasint incx, float *y, blasint incy);
-void cblas_daxpy(blasint n, double, double *x, blasint incx, double *y, blasint incy);
-void cblas_caxpy(blasint n, float *, float *x, blasint incx, float *y, blasint incy);
-void cblas_zaxpy(blasint n, double *, double *x, blasint incx, double *y, blasint incy);
-
-void cblas_scopy(blasint n, float *x, blasint incx, float *y, blasint incy);
-void cblas_dcopy(blasint n, double *x, blasint incx, double *y, blasint incy);
-void cblas_ccopy(blasint n, float *x, blasint incx, float *y, blasint incy);
-void cblas_zcopy(blasint n, double *x, blasint incx, double *y, blasint incy);
-
-void cblas_sswap(blasint n, float *x, blasint incx, float *y, blasint incy);
-void cblas_dswap(blasint n, double *x, blasint incx, double *y, blasint incy);
-void cblas_cswap(blasint n, float *x, blasint incx, float *y, blasint incy);
-void cblas_zswap(blasint n, double *x, blasint incx, double *y, blasint incy);
-
-void cblas_srot(blasint N, float *X, blasint incX, float *Y, blasint incY, float c, float s);
-void cblas_drot(blasint N, double *X, blasint incX, double *Y, blasint incY, double c, double  s);
+typedef enum CBLAS_ORDER     {CblasRowMajor=101, CblasColMajor=102} CBLAS_ORDER;
+typedef enum CBLAS_TRANSPOSE {CblasNoTrans=111, CblasTrans=112, CblasConjTrans=113, CblasConjNoTrans=114} CBLAS_TRANSPOSE;
+typedef enum CBLAS_UPLO      {CblasUpper=121, CblasLower=122} CBLAS_UPLO;
+typedef enum CBLAS_DIAG      {CblasNonUnit=131, CblasUnit=132} CBLAS_DIAG;
+typedef enum CBLAS_SIDE      {CblasLeft=141, CblasRight=142} CBLAS_SIDE;
+
+float  cblas_sdsdot(const blasint n, const float alpha, const float *x, const blasint incx, const float *y, const blasint incy);
+double cblas_dsdot (const blasint n, const float *x, const blasint incx, const float *y, const blasint incy);
+float  cblas_sdot(const blasint n, const float  *x, const blasint incx, const float  *y, const blasint incy);
+double cblas_ddot(const blasint n, const double *x, const blasint incx, const double *y, const blasint incy);
+
+openblas_complex_float  cblas_cdotu(const blasint n, const float  *x, const blasint incx, const float  *y, const blasint incy);
+openblas_complex_float  cblas_cdotc(const blasint n, const float  *x, const blasint incx, const float  *y, const blasint incy);
+openblas_complex_double cblas_zdotu(const blasint n, const double *x, const blasint incx, const double *y, const blasint incy);
+openblas_complex_double cblas_zdotc(const blasint n, const double *x, const blasint incx, const double *y, const blasint incy);
+
+void  cblas_cdotu_sub(const blasint n, const float  *x, const blasint incx, const float  *y, const blasint incy, openblas_complex_float  *ret);
+void  cblas_cdotc_sub(const blasint n, const float  *x, const blasint incx, const float  *y, const blasint incy, openblas_complex_float  *ret);
+void  cblas_zdotu_sub(const blasint n, const double *x, const blasint incx, const double *y, const blasint incy, openblas_complex_double *ret);
+void  cblas_zdotc_sub(const blasint n, const double *x, const blasint incx, const double *y, const blasint incy, openblas_complex_double *ret);
+
+float  cblas_sasum (const blasint n, const float  *x, const blasint incx);
+double cblas_dasum (const blasint n, const double *x, const blasint incx);
+float  cblas_scasum(const blasint n, const float  *x, const blasint incx);
+double cblas_dzasum(const blasint n, const double *x, const blasint incx);
+
+float  cblas_snrm2 (const blasint N, const float  *X, const blasint incX);
+double cblas_dnrm2 (const blasint N, const double *X, const blasint incX);
+float  cblas_scnrm2(const blasint N, const float  *X, const blasint incX);
+double cblas_dznrm2(const blasint N, const double *X, const blasint incX);
+
+CBLAS_INDEX cblas_isamax(const blasint n, const float  *x, const blasint incx);
+CBLAS_INDEX cblas_idamax(const blasint n, const double *x, const blasint incx);
+CBLAS_INDEX cblas_icamax(const blasint n, const float  *x, const blasint incx);
+CBLAS_INDEX cblas_izamax(const blasint n, const double *x, const blasint incx);
+
+void cblas_saxpy(const blasint n, const float alpha, const float *x, const blasint incx, float *y, const blasint incy);
+void cblas_daxpy(const blasint n, const double alpha, const double *x, const blasint incx, double *y, const blasint incy);
+void cblas_caxpy(const blasint n, const float *alpha, const float *x, const blasint incx, float *y, const blasint incy);
+void cblas_zaxpy(const blasint n, const double *alpha, const double *x, const blasint incx, double *y, const blasint incy);
+
+void cblas_scopy(const blasint n, const float *x, const blasint incx, float *y, const blasint incy);
+void cblas_dcopy(const blasint n, const double *x, const blasint incx, double *y, const blasint incy);
+void cblas_ccopy(const blasint n, const float *x, const blasint incx, float *y, const blasint incy);
+void cblas_zcopy(const blasint n, const double *x, const blasint incx, double *y, const blasint incy);
+
+void cblas_sswap(const blasint n, float *x, const blasint incx, float *y, const blasint incy);
+void cblas_dswap(const blasint n, double *x, const blasint incx, double *y, const blasint incy);
+void cblas_cswap(const blasint n, float *x, const blasint incx, float *y, const blasint incy);
+void cblas_zswap(const blasint n, double *x, const blasint incx, double *y, const blasint incy);
+
+void cblas_srot(const blasint N, float *X, const blasint incX, float *Y, const blasint incY, const float c, const float s);
+void cblas_drot(const blasint N, double *X, const blasint incX, double *Y, const blasint incY, const double c, const double  s);
 
 void cblas_srotg(float *a, float *b, float *c, float *s);
 void cblas_drotg(double *a, double *b, double *c, double *s);
 
-void cblas_srotm(blasint N, float *X, blasint incX, float *Y, blasint incY, float *P);
-void cblas_drotm(blasint N, double *X, blasint incX, double *Y, blasint incY, double *P);
-
-void cblas_srotmg(float *d1, float *d2, float *b1, float b2, float *P);
-void cblas_drotmg(double *d1, double *d2, double *b1, double b2, double *P);
-
-void cblas_sscal(blasint N, float alpha, float *X, blasint incX);
-void cblas_dscal(blasint N, double alpha, double *X, blasint incX);
-void cblas_cscal(blasint N, float *alpha, float *X, blasint incX);
-void cblas_zscal(blasint N, double *alpha, double *X, blasint incX);
-void cblas_csscal(blasint N, float alpha, float *X, blasint incX);
-void cblas_zdscal(blasint N, double alpha, double *X, blasint incX);
-
-void cblas_sgemv(enum CBLAS_ORDER order,  enum CBLAS_TRANSPOSE trans,  blasint m, blasint n,
-		 float alpha, float  *a, blasint lda,  float  *x, blasint incx,  float beta,  float  *y, blasint incy);
-void cblas_dgemv(enum CBLAS_ORDER order,  enum CBLAS_TRANSPOSE trans,  blasint m, blasint n,
-		 double alpha, double  *a, blasint lda,  double  *x, blasint incx,  double beta,  double  *y, blasint incy);
-void cblas_cgemv(enum CBLAS_ORDER order,  enum CBLAS_TRANSPOSE trans,  blasint m, blasint n,
-		 float *alpha, float  *a, blasint lda,  float  *x, blasint incx,  float *beta,  float  *y, blasint incy);
-void cblas_zgemv(enum CBLAS_ORDER order,  enum CBLAS_TRANSPOSE trans,  blasint m, blasint n,
-		 double *alpha, double  *a, blasint lda,  double  *x, blasint incx,  double *beta,  double  *y, blasint incy);
-
-void cblas_sger (enum CBLAS_ORDER order, blasint M, blasint N, float   alpha, float  *X, blasint incX, float  *Y, blasint incY, float  *A, blasint lda);
-void cblas_dger (enum CBLAS_ORDER order, blasint M, blasint N, double  alpha, double *X, blasint incX, double *Y, blasint incY, double *A, blasint lda);
-void cblas_cgeru(enum CBLAS_ORDER order, blasint M, blasint N, float  *alpha, float  *X, blasint incX, float  *Y, blasint incY, float  *A, blasint lda);
-void cblas_cgerc(enum CBLAS_ORDER order, blasint M, blasint N, float  *alpha, float  *X, blasint incX, float  *Y, blasint incY, float  *A, blasint lda);
-void cblas_zgeru(enum CBLAS_ORDER order, blasint M, blasint N, double *alpha, double *X, blasint incX, double *Y, blasint incY, double *A, blasint lda);
-void cblas_zgerc(enum CBLAS_ORDER order, blasint M, blasint N, double *alpha, double *X, blasint incX, double *Y, blasint incY, double *A, blasint lda);
-
-void cblas_strsv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag, blasint N, float *A, blasint lda, float *X, blasint incX);
-void cblas_dtrsv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag, blasint N, double *A, blasint lda, double *X, blasint incX);
-void cblas_ctrsv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag, blasint N, float *A, blasint lda, float *X, blasint incX);
-void cblas_ztrsv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag, blasint N, double *A, blasint lda, double *X, blasint incX);
-
-void cblas_strmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag, blasint N, float *A, blasint lda, float *X, blasint incX);
-void cblas_dtrmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag, blasint N, double *A, blasint lda, double *X, blasint incX);
-void cblas_ctrmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag, blasint N, float *A, blasint lda, float *X, blasint incX);
-void cblas_ztrmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag, blasint N, double *A, blasint lda, double *X, blasint incX);
-
-void cblas_ssyr(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, float alpha, float *X, blasint incX, float *A, blasint lda);
-void cblas_dsyr(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, double alpha, double *X, blasint incX, double *A, blasint lda);
-void cblas_cher(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, float alpha, float *X, blasint incX, float *A, blasint lda);
-void cblas_zher(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, double alpha, double *X, blasint incX, double *A, blasint lda);
-
-void cblas_ssyr2(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo,blasint N, float alpha, float *X,
-                blasint incX, float *Y, blasint incY, float *A, blasint lda);
-void cblas_dsyr2(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, double alpha, double *X,
-                blasint incX, double *Y, blasint incY, double *A, blasint lda);
-void cblas_cher2(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, float *alpha, float *X, blasint incX,
-                float *Y, blasint incY, float *A, blasint lda);
-void cblas_zher2(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, double *alpha, double *X, blasint incX,
-                double *Y, blasint incY, double *A, blasint lda);
-
-void cblas_sgbmv(enum CBLAS_ORDER order, enum CBLAS_TRANSPOSE TransA, blasint M, blasint N,
-                 blasint KL, blasint KU, float alpha, float *A, blasint lda, float *X, blasint incX, float beta, float *Y, blasint incY);
-void cblas_dgbmv(enum CBLAS_ORDER order, enum CBLAS_TRANSPOSE TransA, blasint M, blasint N,
-                 blasint KL, blasint KU, double alpha, double *A, blasint lda, double *X, blasint incX, double beta, double *Y, blasint incY);
-void cblas_cgbmv(enum CBLAS_ORDER order, enum CBLAS_TRANSPOSE TransA, blasint M, blasint N,
-                 blasint KL, blasint KU, float *alpha, float *A, blasint lda, float *X, blasint incX, float *beta, float *Y, blasint incY);
-void cblas_zgbmv(enum CBLAS_ORDER order, enum CBLAS_TRANSPOSE TransA, blasint M, blasint N,
-                 blasint KL, blasint KU, double *alpha, double *A, blasint lda, double *X, blasint incX, double *beta, double *Y, blasint incY);
-
-void cblas_ssbmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, blasint K, float alpha, float *A,
-                 blasint lda, float *X, blasint incX, float beta, float *Y, blasint incY);
-void cblas_dsbmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, blasint K, double alpha, double *A,
-                 blasint lda, double *X, blasint incX, double beta, double *Y, blasint incY);
-
-
-void cblas_stbmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag,
-                 blasint N, blasint K, float *A, blasint lda, float *X, blasint incX);
-void cblas_dtbmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag,
-                 blasint N, blasint K, double *A, blasint lda, double *X, blasint incX);
-void cblas_ctbmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag,
-                 blasint N, blasint K, float *A, blasint lda, float *X, blasint incX);
-void cblas_ztbmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag,
-                 blasint N, blasint K, double *A, blasint lda, double *X, blasint incX);
-
-void cblas_stbsv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag,
-                 blasint N, blasint K, float *A, blasint lda, float *X, blasint incX);
-void cblas_dtbsv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag,
-                 blasint N, blasint K, double *A, blasint lda, double *X, blasint incX);
-void cblas_ctbsv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag,
-                 blasint N, blasint K, float *A, blasint lda, float *X, blasint incX);
-void cblas_ztbsv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag,
-                 blasint N, blasint K, double *A, blasint lda, double *X, blasint incX);
-
-void cblas_stpmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag,
-                 blasint N, float *Ap, float *X, blasint incX);
-void cblas_dtpmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag,
-                 blasint N, double *Ap, double *X, blasint incX);
-void cblas_ctpmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag,
-                 blasint N, float *Ap, float *X, blasint incX);
-void cblas_ztpmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag,
-                 blasint N, double *Ap, double *X, blasint incX);
-
-void cblas_stpsv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag,
-                 blasint N, float *Ap, float *X, blasint incX);
-void cblas_dtpsv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag,
-                 blasint N, double *Ap, double *X, blasint incX);
-void cblas_ctpsv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag,
-                 blasint N, float *Ap, float *X, blasint incX);
-void cblas_ztpsv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag,
-                 blasint N, double *Ap, double *X, blasint incX);
-
-void cblas_ssymv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, float alpha, float *A,
-                 blasint lda, float *X, blasint incX, float beta, float *Y, blasint incY);
-void cblas_dsymv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, double alpha, double *A,
-                 blasint lda, double *X, blasint incX, double beta, double *Y, blasint incY);
-void cblas_chemv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, float *alpha, float *A,
-                 blasint lda, float *X, blasint incX, float *beta, float *Y, blasint incY);
-void cblas_zhemv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, double *alpha, double *A,
-                 blasint lda, double *X, blasint incX, double *beta, double *Y, blasint incY);
-
-
-void cblas_sspmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, float alpha, float *Ap,
-                 float *X, blasint incX, float beta, float *Y, blasint incY);
-void cblas_dspmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, double alpha, double *Ap,
-                 double *X, blasint incX, double beta, double *Y, blasint incY);
-
-void cblas_sspr(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, float alpha, float *X, blasint incX, float *Ap);
-void cblas_dspr(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, double alpha, double *X, blasint incX, double *Ap);
-
-void cblas_chpr(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, float alpha, float *X, blasint incX, float *A);
-void cblas_zhpr(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, double alpha, double *X,blasint incX, double *A);
-
-void cblas_sspr2(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, float alpha, float *X, blasint incX, float *Y, blasint incY, float *A);
-void cblas_dspr2(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, double alpha, double *X, blasint incX, double *Y, blasint incY, double *A);
-void cblas_chpr2(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, float *alpha, float *X, blasint incX, float *Y, blasint incY, float *Ap);
-void cblas_zhpr2(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, double *alpha, double *X, blasint incX, double *Y, blasint incY, double *Ap);
-
-void cblas_chbmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, blasint K,
-		 float *alpha, float *A, blasint lda, float *X, blasint incX, float *beta, float *Y, blasint incY);
-void cblas_zhbmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, blasint K,
-		 double *alpha, double *A, blasint lda, double *X, blasint incX, double *beta, double *Y, blasint incY);
-
-void cblas_chpmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N,
-		 float *alpha, float *Ap, float *X, blasint incX, float *beta, float *Y, blasint incY);
-void cblas_zhpmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N,
-		 double *alpha, double *Ap, double *X, blasint incX, double *beta, double *Y, blasint incY);
-
-void cblas_sgemm(enum CBLAS_ORDER Order, enum CBLAS_TRANSPOSE TransA, enum CBLAS_TRANSPOSE TransB, blasint M, blasint N, blasint K,
-		 float alpha, float *A, blasint lda, float *B, blasint ldb, float beta, float *C, blasint ldc);
-void cblas_dgemm(enum CBLAS_ORDER Order, enum CBLAS_TRANSPOSE TransA, enum CBLAS_TRANSPOSE TransB, blasint M, blasint N, blasint K,
-		 double alpha, double *A, blasint lda, double *B, blasint ldb, double beta, double *C, blasint ldc);
-void cblas_cgemm(enum CBLAS_ORDER Order, enum CBLAS_TRANSPOSE TransA, enum CBLAS_TRANSPOSE TransB, blasint M, blasint N, blasint K,
-		 float *alpha, float *A, blasint lda, float *B, blasint ldb, float *beta, float *C, blasint ldc);
-void cblas_zgemm(enum CBLAS_ORDER Order, enum CBLAS_TRANSPOSE TransA, enum CBLAS_TRANSPOSE TransB, blasint M, blasint N, blasint K,
-		 double *alpha, double *A, blasint lda, double *B, blasint ldb, double *beta, double *C, blasint ldc);
-
-void cblas_ssymm(enum CBLAS_ORDER Order, enum CBLAS_SIDE Side, enum CBLAS_UPLO Uplo, blasint M, blasint N,
-                 float alpha, float *A, blasint lda, float *B, blasint ldb, float beta, float *C, blasint ldc);
-void cblas_dsymm(enum CBLAS_ORDER Order, enum CBLAS_SIDE Side, enum CBLAS_UPLO Uplo, blasint M, blasint N,
-                 double alpha, double *A, blasint lda, double *B, blasint ldb, double beta, double *C, blasint ldc);
-void cblas_csymm(enum CBLAS_ORDER Order, enum CBLAS_SIDE Side, enum CBLAS_UPLO Uplo, blasint M, blasint N,
-                 float *alpha, float *A, blasint lda, float *B, blasint ldb, float *beta, float *C, blasint ldc);
-void cblas_zsymm(enum CBLAS_ORDER Order, enum CBLAS_SIDE Side, enum CBLAS_UPLO Uplo, blasint M, blasint N,
-                 double *alpha, double *A, blasint lda, double *B, blasint ldb, double *beta, double *C, blasint ldc);
-
-void cblas_ssyrk(enum CBLAS_ORDER Order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE Trans,
-		 blasint N, blasint K, float alpha, float *A, blasint lda, float beta, float *C, blasint ldc);
-void cblas_dsyrk(enum CBLAS_ORDER Order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE Trans,
-		 blasint N, blasint K, double alpha, double *A, blasint lda, double beta, double *C, blasint ldc);
-void cblas_csyrk(enum CBLAS_ORDER Order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE Trans,
-		 blasint N, blasint K, float *alpha, float *A, blasint lda, float *beta, float *C, blasint ldc);
-void cblas_zsyrk(enum CBLAS_ORDER Order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE Trans,
-		 blasint N, blasint K, double *alpha, double *A, blasint lda, double *beta, double *C, blasint ldc);
-
-void cblas_ssyr2k(enum CBLAS_ORDER Order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE Trans,
-		  blasint N, blasint K, float alpha, float *A, blasint lda, float *B, blasint ldb, float beta, float *C, blasint ldc);
-void cblas_dsyr2k(enum CBLAS_ORDER Order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE Trans,
-		  blasint N, blasint K, double alpha, double *A, blasint lda, double *B, blasint ldb, double beta, double *C, blasint ldc);
-void cblas_csyr2k(enum CBLAS_ORDER Order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE Trans,
-		  blasint N, blasint K, float *alpha, float *A, blasint lda, float *B, blasint ldb, float *beta, float *C, blasint ldc);
-void cblas_zsyr2k(enum CBLAS_ORDER Order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE Trans,
-		  blasint N, blasint K, double *alpha, double *A, blasint lda, double *B, blasint ldb, double *beta, double *C, blasint ldc);
-
-void cblas_strmm(enum CBLAS_ORDER Order, enum CBLAS_SIDE Side, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA,
-                 enum CBLAS_DIAG Diag, blasint M, blasint N, float alpha, float *A, blasint lda, float *B, blasint ldb);
-void cblas_dtrmm(enum CBLAS_ORDER Order, enum CBLAS_SIDE Side, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA,
-                 enum CBLAS_DIAG Diag, blasint M, blasint N, double alpha, double *A, blasint lda, double *B, blasint ldb);
-void cblas_ctrmm(enum CBLAS_ORDER Order, enum CBLAS_SIDE Side, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA,
-                 enum CBLAS_DIAG Diag, blasint M, blasint N, float *alpha, float *A, blasint lda, float *B, blasint ldb);
-void cblas_ztrmm(enum CBLAS_ORDER Order, enum CBLAS_SIDE Side, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA,
-                 enum CBLAS_DIAG Diag, blasint M, blasint N, double *alpha, double *A, blasint lda, double *B, blasint ldb);
-
-void cblas_strsm(enum CBLAS_ORDER Order, enum CBLAS_SIDE Side, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA,
-                 enum CBLAS_DIAG Diag, blasint M, blasint N, float alpha, float *A, blasint lda, float *B, blasint ldb);
-void cblas_dtrsm(enum CBLAS_ORDER Order, enum CBLAS_SIDE Side, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA,
-                 enum CBLAS_DIAG Diag, blasint M, blasint N, double alpha, double *A, blasint lda, double *B, blasint ldb);
-void cblas_ctrsm(enum CBLAS_ORDER Order, enum CBLAS_SIDE Side, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA,
-                 enum CBLAS_DIAG Diag, blasint M, blasint N, float *alpha, float *A, blasint lda, float *B, blasint ldb);
-void cblas_ztrsm(enum CBLAS_ORDER Order, enum CBLAS_SIDE Side, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA,
-                 enum CBLAS_DIAG Diag, blasint M, blasint N, double *alpha, double *A, blasint lda, double *B, blasint ldb);
-
-void cblas_chemm(enum CBLAS_ORDER Order, enum CBLAS_SIDE Side, enum CBLAS_UPLO Uplo, blasint M, blasint N,
-                 float *alpha, float *A, blasint lda, float *B, blasint ldb, float *beta, float *C, blasint ldc);
-void cblas_zhemm(enum CBLAS_ORDER Order, enum CBLAS_SIDE Side, enum CBLAS_UPLO Uplo, blasint M, blasint N,
-                 double *alpha, double *A, blasint lda, double *B, blasint ldb, double *beta, double *C, blasint ldc);
-
-void cblas_cherk(enum CBLAS_ORDER Order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE Trans, blasint N, blasint K,
-                 float alpha, float *A, blasint lda, float beta, float *C, blasint ldc);
-void cblas_zherk(enum CBLAS_ORDER Order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE Trans, blasint N, blasint K,
-                 double alpha, double *A, blasint lda, double beta, double *C, blasint ldc);
-
-void cblas_cher2k(enum CBLAS_ORDER Order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE Trans, blasint N, blasint K,
-                  float *alpha, float *A, blasint lda, float *B, blasint ldb, float beta, float *C, blasint ldc);
-void cblas_zher2k(enum CBLAS_ORDER Order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE Trans, blasint N, blasint K,
-                  double *alpha, double *A, blasint lda, double *B, blasint ldb, double beta, double *C, blasint ldc);
+void cblas_srotm(const blasint N, float *X, const blasint incX, float *Y, const blasint incY, const float *P);
+void cblas_drotm(const blasint N, double *X, const blasint incX, double *Y, const blasint incY, const double *P);
+
+void cblas_srotmg(float *d1, float *d2, float *b1, const float b2, float *P);
+void cblas_drotmg(double *d1, double *d2, double *b1, const double b2, double *P);
+
+void cblas_sscal(const blasint N, const float alpha, float *X, const blasint incX);
+void cblas_dscal(const blasint N, const double alpha, double *X, const blasint incX);
+void cblas_cscal(const blasint N, const float *alpha, float *X, const blasint incX);
+void cblas_zscal(const blasint N, const double *alpha, double *X, const blasint incX);
+void cblas_csscal(const blasint N, const float alpha, float *X, const blasint incX);
+void cblas_zdscal(const blasint N, const double alpha, double *X, const blasint incX);
+
+void cblas_sgemv(const enum CBLAS_ORDER order,  const enum CBLAS_TRANSPOSE trans,  const blasint m, const blasint n,
+		 const float alpha, const float  *a, const blasint lda,  const float  *x, const blasint incx,  const float beta,  float  *y, const blasint incy);
+void cblas_dgemv(const enum CBLAS_ORDER order,  const enum CBLAS_TRANSPOSE trans,  const blasint m, const blasint n,
+		 const double alpha, const double  *a, const blasint lda,  const double  *x, const blasint incx,  const double beta,  double  *y, const blasint incy);
+void cblas_cgemv(const enum CBLAS_ORDER order,  const enum CBLAS_TRANSPOSE trans,  const blasint m, const blasint n,
+		 const float *alpha, const float  *a, const blasint lda,  const float  *x, const blasint incx,  const float *beta,  float  *y, const blasint incy);
+void cblas_zgemv(const enum CBLAS_ORDER order,  const enum CBLAS_TRANSPOSE trans,  const blasint m, const blasint n,
+		 const double *alpha, const double  *a, const blasint lda,  const double  *x, const blasint incx,  const double *beta,  double  *y, const blasint incy);
+
+void cblas_sger (const enum CBLAS_ORDER order, const blasint M, const blasint N, const float   alpha, const float  *X, const blasint incX, const float  *Y, const blasint incY, float  *A, const blasint lda);
+void cblas_dger (const enum CBLAS_ORDER order, const blasint M, const blasint N, const double  alpha, const double *X, const blasint incX, const double *Y, const blasint incY, double *A, const blasint lda);
+void cblas_cgeru(const enum CBLAS_ORDER order, const blasint M, const blasint N, const float  *alpha, const float  *X, const blasint incX, const float  *Y, const blasint incY, float  *A, const blasint lda);
+void cblas_cgerc(const enum CBLAS_ORDER order, const blasint M, const blasint N, const float  *alpha, const float  *X, const blasint incX, const float  *Y, const blasint incY, float  *A, const blasint lda);
+void cblas_zgeru(const enum CBLAS_ORDER order, const blasint M, const blasint N, const double *alpha, const double *X, const blasint incX, const double *Y, const blasint incY, double *A, const blasint lda);
+void cblas_zgerc(const enum CBLAS_ORDER order, const blasint M, const blasint N, const double *alpha, const double *X, const blasint incX, const double *Y, const blasint incY, double *A, const blasint lda);
+
+void cblas_strsv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, const blasint N, const float *A, const blasint lda, float *X, const blasint incX);
+void cblas_dtrsv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, const blasint N, const double *A, const blasint lda, double *X, const blasint incX);
+void cblas_ctrsv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, const blasint N, const float *A, const blasint lda, float *X, const blasint incX);
+void cblas_ztrsv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, const blasint N, const double *A, const blasint lda, double *X, const blasint incX);
+
+void cblas_strmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, const blasint N, const float *A, const blasint lda, float *X, const blasint incX);
+void cblas_dtrmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, const blasint N, const double *A, const blasint lda, double *X, const blasint incX);
+void cblas_ctrmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, const blasint N, const float *A, const blasint lda, float *X, const blasint incX);
+void cblas_ztrmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag, const blasint N, const double *A, const blasint lda, double *X, const blasint incX);
+
+void cblas_ssyr(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const blasint N, const float alpha, const float *X, const blasint incX, float *A, const blasint lda);
+void cblas_dsyr(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const blasint N, const double alpha, const double *X, const blasint incX, double *A, const blasint lda);
+void cblas_cher(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const blasint N, const float alpha, const float *X, const blasint incX, float *A, const blasint lda);
+void cblas_zher(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const blasint N, const double alpha, const double *X, const blasint incX, double *A, const blasint lda);
+
+void cblas_ssyr2(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo,const blasint N, const float alpha, const float *X,
+                const blasint incX, const float *Y, const blasint incY, float *A, const blasint lda);
+void cblas_dsyr2(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const blasint N, const double alpha, const double *X,
+                const blasint incX, const double *Y, const blasint incY, double *A, const blasint lda);
+void cblas_cher2(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const blasint N, const float *alpha, const float *X, const blasint incX,
+                const float *Y, const blasint incY, float *A, const blasint lda);
+void cblas_zher2(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const blasint N, const double *alpha, const double *X, const blasint incX,
+                const double *Y, const blasint incY, double *A, const blasint lda);
+
+void cblas_sgbmv(const enum CBLAS_ORDER order, const enum CBLAS_TRANSPOSE TransA, const blasint M, const blasint N,
+                 const blasint KL, const blasint KU, const float alpha, const float *A, const blasint lda, const float *X, const blasint incX, const float beta, float *Y, const blasint incY);
+void cblas_dgbmv(const enum CBLAS_ORDER order, const enum CBLAS_TRANSPOSE TransA, const blasint M, const blasint N,
+                 const blasint KL, const blasint KU, const double alpha, const double *A, const blasint lda, const double *X, const blasint incX, const double beta, double *Y, const blasint incY);
+void cblas_cgbmv(const enum CBLAS_ORDER order, const enum CBLAS_TRANSPOSE TransA, const blasint M, const blasint N,
+                 const blasint KL, const blasint KU, const float *alpha, const float *A, const blasint lda, const float *X, const blasint incX, const float *beta, float *Y, const blasint incY);
+void cblas_zgbmv(const enum CBLAS_ORDER order, const enum CBLAS_TRANSPOSE TransA, const blasint M, const blasint N,
+                 const blasint KL, const blasint KU, const double *alpha, const double *A, const blasint lda, const double *X, const blasint incX, const double *beta, double *Y, const blasint incY);
+
+void cblas_ssbmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const blasint N, const blasint K, const float alpha, const float *A,
+                 const blasint lda, const float *X, const blasint incX, const float beta, float *Y, const blasint incY);
+void cblas_dsbmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const blasint N, const blasint K, const double alpha, const double *A,
+                 const blasint lda, const double *X, const blasint incX, const double beta, double *Y, const blasint incY);
+
+
+void cblas_stbmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+                 const blasint N, const blasint K, const float *A, const blasint lda, float *X, const blasint incX);
+void cblas_dtbmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+                 const blasint N, const blasint K, const double *A, const blasint lda, double *X, const blasint incX);
+void cblas_ctbmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+                 const blasint N, const blasint K, const float *A, const blasint lda, float *X, const blasint incX);
+void cblas_ztbmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+                 const blasint N, const blasint K, const double *A, const blasint lda, double *X, const blasint incX);
+
+void cblas_stbsv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+                 const blasint N, const blasint K, const float *A, const blasint lda, float *X, const blasint incX);
+void cblas_dtbsv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+                 const blasint N, const blasint K, const double *A, const blasint lda, double *X, const blasint incX);
+void cblas_ctbsv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+                 const blasint N, const blasint K, const float *A, const blasint lda, float *X, const blasint incX);
+void cblas_ztbsv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+                 const blasint N, const blasint K, const double *A, const blasint lda, double *X, const blasint incX);
+
+void cblas_stpmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+                 const blasint N, const float *Ap, float *X, const blasint incX);
+void cblas_dtpmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+                 const blasint N, const double *Ap, double *X, const blasint incX);
+void cblas_ctpmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+                 const blasint N, const float *Ap, float *X, const blasint incX);
+void cblas_ztpmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+                 const blasint N, const double *Ap, double *X, const blasint incX);
+
+void cblas_stpsv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+                 const blasint N, const float *Ap, float *X, const blasint incX);
+void cblas_dtpsv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+                 const blasint N, const double *Ap, double *X, const blasint incX);
+void cblas_ctpsv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+                 const blasint N, const float *Ap, float *X, const blasint incX);
+void cblas_ztpsv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+                 const blasint N, const double *Ap, double *X, const blasint incX);
+
+void cblas_ssymv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const blasint N, const float alpha, const float *A,
+                 const blasint lda, const float *X, const blasint incX, const float beta, float *Y, const blasint incY);
+void cblas_dsymv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const blasint N, const double alpha, const double *A,
+                 const blasint lda, const double *X, const blasint incX, const double beta, double *Y, const blasint incY);
+void cblas_chemv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const blasint N, const float *alpha, const float *A,
+                 const blasint lda, const float *X, const blasint incX, const float *beta, float *Y, const blasint incY);
+void cblas_zhemv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const blasint N, const double *alpha, const double *A,
+                 const blasint lda, const double *X, const blasint incX, const double *beta, double *Y, const blasint incY);
+
+
+void cblas_sspmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const blasint N, const float alpha, const float *Ap,
+                 const float *X, const blasint incX, const float beta, float *Y, const blasint incY);
+void cblas_dspmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const blasint N, const double alpha, const double *Ap,
+                 const double *X, const blasint incX, const double beta, double *Y, const blasint incY);
+
+void cblas_sspr(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const blasint N, const float alpha, const float *X, const blasint incX, float *Ap);
+void cblas_dspr(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const blasint N, const double alpha, const double *X, const blasint incX, double *Ap);
+
+void cblas_chpr(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const blasint N, const float alpha, const float *X, const blasint incX, float *A);
+void cblas_zhpr(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const blasint N, const double alpha, const double *X,const blasint incX, double *A);
+
+void cblas_sspr2(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const blasint N, const float alpha, const float *X, const blasint incX, const float *Y, const blasint incY, float *A);
+void cblas_dspr2(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const blasint N, const double alpha, const double *X, const blasint incX, const double *Y, const blasint incY, double *A);
+void cblas_chpr2(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const blasint N, const float *alpha, const float *X, const blasint incX, const float *Y, const blasint incY, float *Ap);
+void cblas_zhpr2(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const blasint N, const double *alpha, const double *X, const blasint incX, const double *Y, const blasint incY, double *Ap);
+
+void cblas_chbmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const blasint N, const blasint K,
+		 const float *alpha, const float *A, const blasint lda, const float *X, const blasint incX, const float *beta, float *Y, const blasint incY);
+void cblas_zhbmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const blasint N, const blasint K,
+		 const double *alpha, const double *A, const blasint lda, const double *X, const blasint incX, const double *beta, double *Y, const blasint incY);
+
+void cblas_chpmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const blasint N,
+		 const float *alpha, const float *Ap, const float *X, const blasint incX, const float *beta, float *Y, const blasint incY);
+void cblas_zhpmv(const enum CBLAS_ORDER order, const enum CBLAS_UPLO Uplo, const blasint N,
+		 const double *alpha, const double *Ap, const double *X, const blasint incX, const double *beta, double *Y, const blasint incY);
+
+void cblas_sgemm(const enum CBLAS_ORDER Order, const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_TRANSPOSE TransB, const blasint M, const blasint N, const blasint K,
+		 const float alpha, const float *A, const blasint lda, const float *B, const blasint ldb, const float beta, float *C, const blasint ldc);
+void cblas_dgemm(const enum CBLAS_ORDER Order, const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_TRANSPOSE TransB, const blasint M, const blasint N, const blasint K,
+		 const double alpha, const double *A, const blasint lda, const double *B, const blasint ldb, const double beta, double *C, const blasint ldc);
+void cblas_cgemm(const enum CBLAS_ORDER Order, const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_TRANSPOSE TransB, const blasint M, const blasint N, const blasint K,
+		 const float *alpha, const float *A, const blasint lda, const float *B, const blasint ldb, const float *beta, float *C, const blasint ldc);
+void cblas_zgemm(const enum CBLAS_ORDER Order, const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_TRANSPOSE TransB, const blasint M, const blasint N, const blasint K,
+		 const double *alpha, const double *A, const blasint lda, const double *B, const blasint ldb, const double *beta, double *C, const blasint ldc);
+
+void cblas_ssymm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, const enum CBLAS_UPLO Uplo, const blasint M, const blasint N,
+                 const float alpha, const float *A, const blasint lda, const float *B, const blasint ldb, const float beta, float *C, const blasint ldc);
+void cblas_dsymm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, const enum CBLAS_UPLO Uplo, const blasint M, const blasint N,
+                 const double alpha, const double *A, const blasint lda, const double *B, const blasint ldb, const double beta, double *C, const blasint ldc);
+void cblas_csymm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, const enum CBLAS_UPLO Uplo, const blasint M, const blasint N,
+                 const float *alpha, const float *A, const blasint lda, const float *B, const blasint ldb, const float *beta, float *C, const blasint ldc);
+void cblas_zsymm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, const enum CBLAS_UPLO Uplo, const blasint M, const blasint N,
+                 const double *alpha, const double *A, const blasint lda, const double *B, const blasint ldb, const double *beta, double *C, const blasint ldc);
+
+void cblas_ssyrk(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE Trans,
+		 const blasint N, const blasint K, const float alpha, const float *A, const blasint lda, const float beta, float *C, const blasint ldc);
+void cblas_dsyrk(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE Trans,
+		 const blasint N, const blasint K, const double alpha, const double *A, const blasint lda, const double beta, double *C, const blasint ldc);
+void cblas_csyrk(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE Trans,
+		 const blasint N, const blasint K, const float *alpha, const float *A, const blasint lda, const float *beta, float *C, const blasint ldc);
+void cblas_zsyrk(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE Trans,
+		 const blasint N, const blasint K, const double *alpha, const double *A, const blasint lda, const double *beta, double *C, const blasint ldc);
+
+void cblas_ssyr2k(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE Trans,
+		  const blasint N, const blasint K, const float alpha, const float *A, const blasint lda, const float *B, const blasint ldb, const float beta, float *C, const blasint ldc);
+void cblas_dsyr2k(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE Trans,
+		  const blasint N, const blasint K, const double alpha, const double *A, const blasint lda, const double *B, const blasint ldb, const double beta, double *C, const blasint ldc);
+void cblas_csyr2k(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE Trans,
+		  const blasint N, const blasint K, const float *alpha, const float *A, const blasint lda, const float *B, const blasint ldb, const float *beta, float *C, const blasint ldc);
+void cblas_zsyr2k(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE Trans,
+		  const blasint N, const blasint K, const double *alpha, const double *A, const blasint lda, const double *B, const blasint ldb, const double *beta, double *C, const blasint ldc);
+
+void cblas_strmm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA,
+                 const enum CBLAS_DIAG Diag, const blasint M, const blasint N, const float alpha, const float *A, const blasint lda, float *B, const blasint ldb);
+void cblas_dtrmm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA,
+                 const enum CBLAS_DIAG Diag, const blasint M, const blasint N, const double alpha, const double *A, const blasint lda, double *B, const blasint ldb);
+void cblas_ctrmm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA,
+                 const enum CBLAS_DIAG Diag, const blasint M, const blasint N, const float *alpha, const float *A, const blasint lda, float *B, const blasint ldb);
+void cblas_ztrmm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA,
+                 const enum CBLAS_DIAG Diag, const blasint M, const blasint N, const double *alpha, const double *A, const blasint lda, double *B, const blasint ldb);
+
+void cblas_strsm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA,
+                 const enum CBLAS_DIAG Diag, const blasint M, const blasint N, const float alpha, const float *A, const blasint lda, float *B, const blasint ldb);
+void cblas_dtrsm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA,
+                 const enum CBLAS_DIAG Diag, const blasint M, const blasint N, const double alpha, const double *A, const blasint lda, double *B, const blasint ldb);
+void cblas_ctrsm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA,
+                 const enum CBLAS_DIAG Diag, const blasint M, const blasint N, const float *alpha, const float *A, const blasint lda, float *B, const blasint ldb);
+void cblas_ztrsm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA,
+                 const enum CBLAS_DIAG Diag, const blasint M, const blasint N, const double *alpha, const double *A, const blasint lda, double *B, const blasint ldb);
+
+void cblas_chemm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, const enum CBLAS_UPLO Uplo, const blasint M, const blasint N,
+                 const float *alpha, const float *A, const blasint lda, const float *B, const blasint ldb, const float *beta, float *C, const blasint ldc);
+void cblas_zhemm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side, const enum CBLAS_UPLO Uplo, const blasint M, const blasint N,
+                 const double *alpha, const double *A, const blasint lda, const double *B, const blasint ldb, const double *beta, double *C, const blasint ldc);
+
+void cblas_cherk(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE Trans, const blasint N, const blasint K,
+                 const float alpha, const float *A, const blasint lda, const float beta, float *C, const blasint ldc);
+void cblas_zherk(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE Trans, const blasint N, const blasint K,
+                 const double alpha, const double *A, const blasint lda, const double beta, double *C, const blasint ldc);
+
+void cblas_cher2k(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE Trans, const blasint N, const blasint K,
+                  const float *alpha, const float *A, const blasint lda, const float *B, const blasint ldb, const float beta, float *C, const blasint ldc);
+void cblas_zher2k(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE Trans, const blasint N, const blasint K,
+                  const double *alpha, const double *A, const blasint lda, const double *B, const blasint ldb, const double beta, double *C, const blasint ldc);
 
 void cblas_xerbla(blasint p, char *rout, char *form, ...);
 
 #ifdef __cplusplus
 }
-     
 #endif  /* __cplusplus */
 
 #endif

common.h

@@ -68,7 +68,7 @@ extern "C" {
 #define SMP
 #endif
 
-#if defined(OS_WINNT) || defined(OS_CYGWIN_NT) || defined(OS_Interix)
+#if defined(OS_WINNT) || defined(OS_CYGWIN_NT) || defined(OS_INTERIX)
 #define WINDOWS_ABI
 #define OS_WINDOWS
 
@@ -89,7 +89,7 @@ extern "C" {
 #include <sched.h>
 #endif
 
-#ifdef OS_DARWIN
+#if defined(OS_DARWIN) || defined(OS_FREEBSD) || defined(OS_NETBSD)
 #include <sched.h>
 #endif
 
@@ -351,7 +351,12 @@ typedef int blasint;
 #endif
 
 #define MMAP_ACCESS (PROT_READ | PROT_WRITE)
+
+#ifdef __NetBSD__
+#define MMAP_POLICY (MAP_PRIVATE | MAP_ANON)
+#else
 #define MMAP_POLICY (MAP_PRIVATE | MAP_ANONYMOUS)
+#endif
 
 #include "param.h"
 #include "common_param.h"
@@ -374,6 +379,31 @@ typedef int blasint;
 #endif
 #endif
 
+#ifndef ASSEMBLER
+#ifndef NOINCLUDE
+/* Inclusion of a standard header file is needed for definition of __STDC_*
+   predefined macros with some compilers (e.g. GCC 4.7 on Linux).  This occurs
+   as a side effect of including either <features.h> or <stdc-predef.h>. */
+#include <stdio.h>
+#endif  // NOINCLUDE
+
+/* C99 supports complex floating numbers natively, which GCC also offers as an
+   extension since version 3.0.  If neither are available, use a compatible
+   structure as fallback (see Clause 6.2.5.13 of the C99 standard). */
+#if (defined(__STDC_IEC_559_COMPLEX__) || __STDC_VERSION__ >= 199901L || \
+     (__GNUC__ >= 3 && !defined(__cplusplus)))
+  #define OPENBLAS_COMPLEX_C99
+  typedef float _Complex openblas_complex_float;
+  typedef double _Complex openblas_complex_double;
+  typedef xdouble _Complex openblas_complex_xdouble;
+#else
+  #define OPENBLAS_COMPLEX_STRUCT
+  typedef struct { float real, imag; } openblas_complex_float;
+  typedef struct { double real, imag; } openblas_complex_double;
+  typedef struct { xdouble real, imag; } openblas_complex_xdouble;
+#endif
+#endif  // ASSEMBLER
+
 #ifndef IFLUSH
 #define IFLUSH
 #endif
@@ -528,7 +558,8 @@ typedef struct {
 #include "common_level3.h"
 #include "common_lapack.h"
 #ifdef CBLAS
-#include "cblas.h"
+/* This header file is generated from "cblas.h" (see Makefile.prebuild). */
+#include "cblas_noconst.h"
 #endif
 
 #ifndef ASSEMBLER

common_interface.h

@@ -45,6 +45,8 @@ extern "C" {
 
 int    BLASFUNC(xerbla)(char *, blasint *info, blasint);
 
+void    openblas_set_num_threads_(int *);
+
 FLOATRET  BLASFUNC(sdot)  (blasint *, float  *, blasint *, float  *, blasint *);
 FLOATRET  BLASFUNC(sdsdot)(blasint *, float  *,        float  *, blasint *, float  *, blasint *);
 
@@ -74,19 +76,19 @@ myxcomplex_t    BLASFUNC(xdotu)  (blasint *, xdouble  *, blasint *, xdouble  *,
 myxcomplex_t    BLASFUNC(xdotc)  (blasint *, xdouble  *, blasint *, xdouble  *, blasint *);
 
 #elif defined RETURN_BY_STACK
-void  BLASFUNC(cdotu)  (float   _Complex *,  blasint *, float  * , blasint *, float  *,  blasint *);
-void  BLASFUNC(cdotc)  (float   _Complex *,  blasint *, float  *,  blasint *, float  *,  blasint *);
-void  BLASFUNC(zdotu)  (double  _Complex *, blasint *, double  *, blasint *, double  *, blasint *);
-void  BLASFUNC(zdotc)  (double  _Complex *, blasint *, double  *, blasint *, double  *, blasint *);
-void  BLASFUNC(xdotu)  (xdouble _Complex *, blasint *, xdouble  *, blasint *, xdouble  *, blasint *);
-void  BLASFUNC(xdotc)  (xdouble _Complex *, blasint *, xdouble  *, blasint *, xdouble  *, blasint *);
+void  BLASFUNC(cdotu)  (openblas_complex_float   *,  blasint *, float  * , blasint *, float  *,  blasint *);
+void  BLASFUNC(cdotc)  (openblas_complex_float   *,  blasint *, float  *,  blasint *, float  *,  blasint *);
+void  BLASFUNC(zdotu)  (openblas_complex_double  *, blasint *, double  *, blasint *, double  *, blasint *);
+void  BLASFUNC(zdotc)  (openblas_complex_double  *, blasint *, double  *, blasint *, double  *, blasint *);
+void  BLASFUNC(xdotu)  (openblas_complex_xdouble *, blasint *, xdouble  *, blasint *, xdouble  *, blasint *);
+void  BLASFUNC(xdotc)  (openblas_complex_xdouble *, blasint *, xdouble  *, blasint *, xdouble  *, blasint *);
 #else
-float   _Complex BLASFUNC(cdotu)  (blasint *, float  *, blasint *, float  *, blasint *);
-float   _Complex BLASFUNC(cdotc)  (blasint *, float  *, blasint *, float  *, blasint *);
-double  _Complex BLASFUNC(zdotu)  (blasint *, double  *, blasint *, double  *, blasint *);
-double  _Complex BLASFUNC(zdotc)  (blasint *, double  *, blasint *, double  *, blasint *);
-xdouble _Complex BLASFUNC(xdotu)  (blasint *, xdouble  *, blasint *, xdouble  *, blasint *);
-xdouble _Complex BLASFUNC(xdotc)  (blasint *, xdouble  *, blasint *, xdouble  *, blasint *);
+openblas_complex_float   BLASFUNC(cdotu)  (blasint *, float  *, blasint *, float  *, blasint *);
+openblas_complex_float   BLASFUNC(cdotc)  (blasint *, float  *, blasint *, float  *, blasint *);
+openblas_complex_double  BLASFUNC(zdotu)  (blasint *, double  *, blasint *, double  *, blasint *);
+openblas_complex_double  BLASFUNC(zdotc)  (blasint *, double  *, blasint *, double  *, blasint *);
+openblas_complex_xdouble BLASFUNC(xdotu)  (blasint *, xdouble  *, blasint *, xdouble  *, blasint *);
+openblas_complex_xdouble BLASFUNC(xdotc)  (blasint *, xdouble  *, blasint *, xdouble  *, blasint *);
 #endif
 
 void    BLASFUNC(saxpy) (blasint *, float  *, float  *, blasint *, float  *, blasint *);
@@ -640,6 +642,8 @@ int BLASFUNC(zgemc)(char *, char *, blasint *, blasint *, blasint *, double *,
 int BLASFUNC(xgemc)(char *, char *, blasint *, blasint *, blasint *, xdouble *,
 	   xdouble *, blasint *, xdouble *, blasint *, xdouble *, blasint *, xdouble *, xdouble *, blasint *);
 
+/* Lapack routines */
+
 int BLASFUNC(sgetf2)(blasint *, blasint *, float  *, blasint *, blasint *, blasint *);
 int BLASFUNC(dgetf2)(blasint *, blasint *, double *, blasint *, blasint *, blasint *);
 int BLASFUNC(qgetf2)(blasint *, blasint *, xdouble *, blasint *, blasint *, blasint *);
@@ -675,6 +679,13 @@ int BLASFUNC(cgesv)(blasint *, blasint *, float  *, blasint *, blasint *, float
 int BLASFUNC(zgesv)(blasint *, blasint *, double *, blasint *, blasint *, double*, blasint *, blasint *);
 int BLASFUNC(xgesv)(blasint *, blasint *, xdouble *, blasint *, blasint *, xdouble*, blasint *, blasint *);
 
+int BLASFUNC(sgesvd)(char *, char *, blasint *, blasint *, float   *, blasint *, float   *, float   *, blasint *, float   *, blasint *, float   *, blasint *, blasint *);
+int BLASFUNC(dgesvd)(char *, char *, blasint *, blasint *, double  *, blasint *, double  *, double  *, blasint *, double  *, blasint *, double  *, blasint *, blasint *);
+int BLASFUNC(qgesvd)(char *, char *, blasint *, blasint *, xdouble *, blasint *, xdouble *, xdouble *, blasint *, xdouble *, blasint *, xdouble *, blasint *, blasint *);
+int BLASFUNC(cgesvd)(char *, char *, blasint *, blasint *, float   *, blasint *, float   *, float   *, blasint *, float   *, blasint *, float   *, blasint *, blasint *);
+int BLASFUNC(zgesvd)(char *, char *, blasint *, blasint *, double  *, blasint *, double  *, double  *, blasint *, double  *, blasint *, double  *, blasint *, blasint *);
+int BLASFUNC(xgesvd)(char *, char *, blasint *, blasint *, xdouble *, blasint *, xdouble *, xdouble *, blasint *, xdouble *, blasint *, xdouble *, blasint *, blasint *);
+
 int BLASFUNC(spotf2)(char *, blasint *, float  *, blasint *, blasint *);
 int BLASFUNC(dpotf2)(char *, blasint *, double *, blasint *, blasint *);
 int BLASFUNC(qpotf2)(char *, blasint *, xdouble *, blasint *, blasint *);
@@ -689,6 +700,13 @@ int BLASFUNC(cpotrf)(char *, blasint *, float  *, blasint *, blasint *);
 int BLASFUNC(zpotrf)(char *, blasint *, double *, blasint *, blasint *);
 int BLASFUNC(xpotrf)(char *, blasint *, xdouble *, blasint *, blasint *);
 
+int BLASFUNC(spotrs)(char *, blasint *, blasint *, float   *, blasint *, float   *, blasint *, blasint *);
+int BLASFUNC(dpotrs)(char *, blasint *, blasint *, double  *, blasint *, double  *, blasint *, blasint *);
+int BLASFUNC(qpotrs)(char *, blasint *, blasint *, xdouble *, blasint *, xdouble *, blasint *, blasint *);
+int BLASFUNC(cpotrs)(char *, blasint *, blasint *, float   *, blasint *, float   *, blasint *, blasint *);
+int BLASFUNC(zpotrs)(char *, blasint *, blasint *, double  *, blasint *, double  *, blasint *, blasint *);
+int BLASFUNC(xpotrs)(char *, blasint *, blasint *, xdouble *, blasint *, xdouble *, blasint *, blasint *);
+
 int BLASFUNC(slauu2)(char *, blasint *, float  *, blasint *, blasint *);
 int BLASFUNC(dlauu2)(char *, blasint *, double *, blasint *, blasint *);
 int BLASFUNC(qlauu2)(char *, blasint *, xdouble *, blasint *, blasint *);

common_linux.h

@@ -86,7 +86,13 @@ static inline int my_set_mempolicy(int mode, const unsigned long *addr, unsigned
   return syscall(SYS_set_mempolicy, mode, addr, flag);
 }
 
-static inline int my_gettid(void) { return syscall(SYS_gettid); }
+static inline int my_gettid(void) { 
+#ifdef SYS_gettid
+return syscall(SYS_gettid); 
+#else
+return getpid();
+#endif
+}
 
 #endif
 #endif

common_reference.h

@@ -63,5 +63,7 @@ double  _Complex BLASFUNC_REF(zdotc)  (blasint *, double  *, blasint *, double
 void BLASFUNC_REF(drotmg)(double *, double *, double *, double *, double *);
 
 double BLASFUNC_REF(dsdot)(blasint *, float *, blasint *, float *, blasint*);
+
+FLOATRET  BLASFUNC_REF(samax) (blasint *, float  *, blasint *);
 												   
 #endif

common_thread.h

@@ -135,7 +135,7 @@ static __inline int num_cpu_avail(int level) {
 	int openmp_nthreads=0;
 #endif
 
-  if ((blas_cpu_number == 1) 
+  if (blas_cpu_number == 1
 
 #ifdef USE_OPENMP
       || omp_in_parallel()

common_x86.h

@@ -254,7 +254,7 @@ static __inline int blas_quickdivide(unsigned int x, unsigned int y){
 #define PROFCODE
 #endif
 
-#if defined(OS_WINNT) || defined(OS_CYGWIN_NT) || defined(OS_INERIX)
+#if defined(OS_WINNT) || defined(OS_CYGWIN_NT) || defined(OS_INTERIX)
 #define SAVEREGISTERS \
 	subl	$32, %esp;\
 	movups	%xmm6,    0(%esp);\
@@ -269,7 +269,7 @@ static __inline int blas_quickdivide(unsigned int x, unsigned int y){
 #define RESTOREREGISTERS
 #endif
 
-#if defined(OS_WINNT) || defined(OS_CYGWIN_NT) || defined(OS_INERIX)
+#if defined(OS_WINNT) || defined(OS_CYGWIN_NT) || defined(OS_INTERIX)
 #define PROLOGUE \
 	.text; \
 	.align 16; \
@@ -282,7 +282,7 @@ REALNAME:
 #define EPILOGUE .end	 REALNAME
 #endif
 
-#if defined(OS_LINUX) || defined(OS_FreeBSD) || defined(OS_NetBSD) || defined(__ELF__)
+#if defined(OS_LINUX) || defined(OS_FREEBSD) || defined(OS_NETBSD) || defined(__ELF__)
 #define PROLOGUE \
 	.text; \
 	.align 16; \
@@ -356,4 +356,11 @@ REALNAME:
 
 #ifndef ALIGN_6
 #define ALIGN_6 .align 64
+
+// ffreep %st(0). 
+// Because Clang didn't support ffreep, we directly use the opcode.
+// Please check out http://www.sandpile.org/x86/opc_fpu.htm 
+#ifndef ffreep
+#define ffreep .byte 0xdf, 0xc0 #
+#endif
 #endif

common_x86_64.h

@@ -353,7 +353,7 @@ REALNAME:
 #define EPILOGUE .end	 REALNAME
 #endif
 
-#if defined(OS_LINUX) || defined(OS_FreeBSD) || defined(OS_NetBSD) || defined(__ELF__) || defined(C_PGI)
+#if defined(OS_LINUX) || defined(OS_FREEBSD) || defined(OS_NETBSD) || defined(__ELF__) || defined(C_PGI)
 #define PROLOGUE \
 	.text; \
 	.align 512; \
@@ -425,6 +425,7 @@ REALNAME:
 #define ALIGN_2 .align 2
 #define ALIGN_3 .align 3
 #define ALIGN_4 .align 4
+#define ALIGN_5 .align 5
 #define ffreep	fstp
 #endif
 
@@ -448,4 +449,10 @@ REALNAME:
 #define ALIGN_6 .align 64
 #endif
 
+// ffreep %st(0). 
+// Because Clang didn't support ffreep, we directly use the opcode.
+// Please check out http://www.sandpile.org/x86/opc_fpu.htm 
+#ifndef ffreep
+#define ffreep .byte 0xdf, 0xc0 #
+#endif
 #endif

cpuid.h

@@ -103,6 +103,9 @@
 #define CORE_NEHALEM	17
 #define CORE_ATOM	18
 #define CORE_NANO	19
+#define CORE_SANDYBRIDGE 20
+#define CORE_BOBCAT     21
+#define CORE_BULLDOZER  22
 
 #define HAVE_SSE      (1 <<  0)
 #define HAVE_SSE2     (1 <<  1)
@@ -122,6 +125,8 @@
 #define HAVE_MISALIGNSSE (1 << 15)
 #define HAVE_128BITFPU   (1 << 16)
 #define HAVE_FASTMOVU    (1 << 17)
+#define HAVE_AVX      (1 <<  18)
+#define HAVE_FMA4     (1 <<  19)
 
 #define CACHE_INFO_L1_I     1
 #define CACHE_INFO_L1_D     2
@@ -188,4 +193,7 @@ typedef struct {
 #define CPUTYPE_NSGEODE			41
 #define CPUTYPE_VIAC3			42
 #define CPUTYPE_NANO			43
+#define CPUTYPE_SANDYBRIDGE             44
+#define CPUTYPE_BOBCAT                  45
+#define CPUTYPE_BULLDOZER               46
 #endif

cpuid_mips.c

@@ -1,5 +1,5 @@
 /*****************************************************************************
-Copyright (c) 2011, Lab of Parallel Software and Computational Science,ICSAS
+Copyright (c) 2011,2012 Lab of Parallel Software and Computational Science,ISCAS
 All rights reserved.
 
 Redistribution and use in source and binary forms, with or without
@@ -101,12 +101,14 @@ int detect(void){
 
   fclose(infile);
 
+  if(p != NULL){
   if (strstr(p, "Loongson-3A")){
     return CPU_LOONGSON3A;
   }else if(strstr(p, "Loongson-3B")){
     return CPU_LOONGSON3B;
   }else if (strstr(p, "Loongson-3")){
     infile = fopen("/proc/cpuinfo", "r");
+    p = (char *)NULL;
     while (fgets(buffer, sizeof(buffer), infile)){
       if (!strncmp("system type", buffer, 11)){
 	p = strchr(buffer, ':') + 2;
@@ -119,6 +121,24 @@ int detect(void){
   }else{
     return CPU_SICORTEX;
   }
+  }
+  //Check model name for Loongson3
+  infile = fopen("/proc/cpuinfo", "r");
+  p = (char *)NULL;
+  while (fgets(buffer, sizeof(buffer), infile)){
+    if (!strncmp("model name", buffer, 10)){
+      p = strchr(buffer, ':') + 2;
+      break;
+    }
+  }
+  fclose(infile);
+  if(p != NULL){
+  if (strstr(p, "Loongson-3A")){
+    return CPU_LOONGSON3A;
+  }else if(strstr(p, "Loongson-3B")){
+    return CPU_LOONGSON3B;
+  }
+  }
 #endif
     return CPU_UNKNOWN;
 }

cpuid_x86.c

@@ -40,6 +40,13 @@
 #include <string.h>
 #include "cpuid.h"
 
+#ifdef NO_AVX
+#define CPUTYPE_SANDYBRIDGE CPUTYPE_NEHALEM
+#define CORE_SANDYBRIDGE CORE_NEHALEM
+#define CPUTYPE_BULLDOZER CPUTYPE_BARCELONA
+#define CORE_BULLDOZER CORE_BARCELONA
+#endif
+
 #ifndef CPUIDEMU
 
 #if defined(__APPLE__) && defined(__i386__)
@@ -109,6 +116,33 @@ static inline int have_excpuid(void){
   return eax & 0xffff;
 }
 
+#ifndef NO_AVX
+static inline void xgetbv(int op, int * eax, int * edx){
+  //Use binary code for xgetbv
+  __asm__ __volatile__
+    (".byte 0x0f, 0x01, 0xd0": "=a" (*eax), "=d" (*edx) : "c" (op) : "cc");
+}
+#endif
+
+int support_avx(){
+#ifndef NO_AVX
+  int eax, ebx, ecx, edx;
+  int ret=0;
+  
+  cpuid(1, &eax, &ebx, &ecx, &edx);
+  if ((ecx & (1 << 28)) != 0 && (ecx & (1 << 27)) != 0){
+    xgetbv(0, &eax, &edx);
+    if((eax & 6) == 6){
+      ret=1;  //OS support AVX
+    }
+  }
+  return ret;
+#else
+  return 0;
+#endif
+}
+
+
 int get_vendor(void){
   int eax, ebx, ecx, edx;
   char vendor[13];
@@ -189,11 +223,17 @@ int get_cputype(int gettype){
     if ((ecx & (1 <<  9)) != 0) feature |= HAVE_SSSE3;
     if ((ecx & (1 << 19)) != 0) feature |= HAVE_SSE4_1;
     if ((ecx & (1 << 20)) != 0) feature |= HAVE_SSE4_2;
+#ifndef NO_AVX
+    if (support_avx()) feature |= HAVE_AVX;
+#endif
 
     if (have_excpuid() >= 0x01) {
       cpuid(0x80000001, &eax, &ebx, &ecx, &edx);
       if ((ecx & (1 <<  6)) != 0) feature |= HAVE_SSE4A;
       if ((ecx & (1 <<  7)) != 0) feature |= HAVE_MISALIGNSSE;
+#ifndef NO_AVX
+      if ((ecx & (1 <<  16)) != 0) feature |= HAVE_FMA4;
+#endif
       if ((edx & (1 << 30)) != 0) feature |= HAVE_3DNOWEX;
       if ((edx & (1 << 31)) != 0) feature |= HAVE_3DNOW;
     }
@@ -974,21 +1014,44 @@ int get_cpuname(void){
 	  return CPUTYPE_DUNNINGTON;
 	}
 	break;
-	  case  2:
-		  switch (model) {
-		  case 5:
-			  //Intel Core (Clarkdale) / Core (Arrandale) 
-			  // Pentium (Clarkdale) / Pentium Mobile (Arrandale) 
-			  // Xeon (Clarkdale), 32nm
-			  return CPUTYPE_NEHALEM;
-		  case 10:
-                          //Intel Core i5-2000 /i7-2000 (Sandy Bridge)
-                          return CPUTYPE_NEHALEM;
-		  case 12:
-			  //Xeon Processor 5600 (Westmere-EP)
-			  return CPUTYPE_NEHALEM;
-		  }
-		  break;
+      case  2:
+	switch (model) {
+	case 5:
+	  //Intel Core (Clarkdale) / Core (Arrandale) 
+	  // Pentium (Clarkdale) / Pentium Mobile (Arrandale) 
+	  // Xeon (Clarkdale), 32nm
+	  return CPUTYPE_NEHALEM;
+	case 10:
+	  //Intel Core i5-2000 /i7-2000 (Sandy Bridge)
+	  if(support_avx())
+	    return CPUTYPE_SANDYBRIDGE;
+	  else
+	    return CPUTYPE_NEHALEM; //OS doesn't support AVX
+	case 12:
+	  //Xeon Processor 5600 (Westmere-EP)
+	  return CPUTYPE_NEHALEM;
+	case 13:
+	  //Intel Core i7-3000 / Xeon E5 (Sandy Bridge)
+	  if(support_avx())
+	    return CPUTYPE_SANDYBRIDGE;
+	  else
+	    return CPUTYPE_NEHALEM;
+	case 14:
+	  // Xeon E7540
+	case 15:
+	  //Xeon Processor E7 (Westmere-EX)
+	  return CPUTYPE_NEHALEM;
+	}
+	break;
+      case 3:
+	switch (model) {
+	case 10:
+	  if(support_avx())
+	    return CPUTYPE_SANDYBRIDGE;
+	  else
+	    return CPUTYPE_NEHALEM;
+	}
+	break;
       }
       break;
     case 0x7:
@@ -1021,6 +1084,13 @@ int get_cpuname(void){
       case  1:
       case 10:
 	return CPUTYPE_BARCELONA;
+      case  6:   //AMD Bulldozer Opteron 6200 / Opteron 4200 / AMD FX-Series
+	if(support_avx())
+	  return CPUTYPE_BULLDOZER;
+	else
+	  return CPUTYPE_BARCELONA; //OS don't support AVX.
+      case  5:
+	return CPUTYPE_BOBCAT;
       }
       break;
     }
@@ -1140,6 +1210,9 @@ static char *cpuname[] = {
   "NSGEODE",
   "VIAC3",
   "NANO",
+  "SANDYBRIDGE",
+  "BOBCAT",
+  "BULLDOZER",
 };
 
 static char *lowercpuname[] = {
@@ -1186,6 +1259,9 @@ static char *lowercpuname[] = {
   "tms3x00",
   "nsgeode",
   "nano",
+  "sandybridge",
+  "bobcat",
+  "bulldozer",
 };
 
 static char *corename[] = {
@@ -1209,6 +1285,9 @@ static char *corename[] = {
   "NEHALEM",
   "ATOM",
   "NANO",
+  "SANDYBRIDGE",
+  "BOBCAT",
+  "BULLDOZER",
 };
 
 static char *corename_lower[] = {
@@ -1232,6 +1311,9 @@ static char *corename_lower[] = {
   "nehalem",
   "atom",
   "nano",
+  "sandybridge",
+  "bobcat",
+  "bulldozer",
 };
 
 
@@ -1315,10 +1397,33 @@ int get_coretype(void){
 	  return CORE_NEHALEM;
 	case 10:
           //Intel Core i5-2000 /i7-2000 (Sandy Bridge)
-          return CORE_NEHALEM;
+	  if(support_avx())
+	    return CORE_SANDYBRIDGE;
+	  else
+	    return CORE_NEHALEM; //OS doesn't support AVX
 	case 12:
 	  //Xeon Processor 5600 (Westmere-EP)
 	  return CORE_NEHALEM;
+	case 13:
+          //Intel Core i7-3000 / Xeon E5 (Sandy Bridge)
+	  if(support_avx())
+	    return CORE_SANDYBRIDGE;
+	  else
+	    return CORE_NEHALEM; //OS doesn't support AVX
+	case 14:
+	  //Xeon E7540
+	case 15:
+	  //Xeon Processor E7 (Westmere-EX)
+	  return CORE_NEHALEM;
+	}
+	break;
+      case 3:
+	switch (model) {
+	case 10:
+	  if(support_avx())
+	    return CORE_SANDYBRIDGE;
+	  else
+	    return CORE_NEHALEM; //OS doesn't support AVX
 	}
 	break;
       }
@@ -1334,7 +1439,15 @@ int get_coretype(void){
     if (family <= 0x5) return CORE_80486;
     if (family <= 0xe) return CORE_ATHLON;
     if (family == 0xf){
-      if ((exfamily == 0) || (exfamily == 2)) return CORE_OPTERON; else return CORE_BARCELONA;
+      if ((exfamily == 0) || (exfamily == 2)) return CORE_OPTERON; 
+      else if (exfamily == 5) return CORE_BOBCAT; 
+      else if (exfamily == 6) {
+	//AMD Bulldozer Opteron 6200 / Opteron 4200 / AMD FX-Series
+	if(support_avx())
+	  return CORE_BULLDOZER;
+	else
+	  return CORE_BARCELONA; //OS don't support AVX. Use old kernels.
+      }else return CORE_BARCELONA;
     }
   }
 
@@ -1400,6 +1513,9 @@ void get_cpuconfig(void){
       printf("#define DTB_SIZE %d\n", info.size * 1024);
       printf("#define DTB_ASSOCIATIVE %d\n", info.associative);
       printf("#define DTB_DEFAULT_ENTRIES %d\n", info.linesize);
+    } else {
+      //fall back for some virtual machines.
+      printf("#define DTB_DEFAULT_ENTRIES 32\n");
     }
     
     features = get_cputype(GET_FEATURE);
@@ -1414,8 +1530,10 @@ void get_cpuconfig(void){
     if (features & HAVE_SSE4_2)   printf("#define HAVE_SSE4_2\n");
     if (features & HAVE_SSE4A)   printf("#define HAVE_SSE4A\n");
     if (features & HAVE_SSE5 )   printf("#define HAVE_SSSE5\n");
+    if (features & HAVE_AVX )    printf("#define HAVE_AVX\n");
     if (features & HAVE_3DNOWEX) printf("#define HAVE_3DNOWEX\n");
     if (features & HAVE_3DNOW)   printf("#define HAVE_3DNOW\n");
+    if (features & HAVE_FMA4 )    printf("#define HAVE_FMA4\n");
     if (features & HAVE_CFLUSH)  printf("#define HAVE_CFLUSH\n");
     if (features & HAVE_HIT)     printf("#define HAVE_HIT 1\n");
     if (features & HAVE_MISALIGNSSE) printf("#define HAVE_MISALIGNSSE\n");
@@ -1479,7 +1597,9 @@ void get_sse(void){
   if (features & HAVE_SSE4_2)   printf("HAVE_SSE4_2=1\n");
   if (features & HAVE_SSE4A)   printf("HAVE_SSE4A=1\n");
   if (features & HAVE_SSE5 )   printf("HAVE_SSSE5=1\n");
+  if (features & HAVE_AVX )    printf("HAVE_AVX=1\n");
   if (features & HAVE_3DNOWEX) printf("HAVE_3DNOWEX=1\n");
   if (features & HAVE_3DNOW)   printf("HAVE_3DNOW=1\n");
+  if (features & HAVE_FMA4 )    printf("HAVE_FMA4=1\n");
 
 }

ctest.c

@@ -34,20 +34,20 @@ COMPILER_GNU
 OS_LINUX
 #endif
 
-#if defined(__FreeBSD__)
-OS_FreeBSD
+#if defined(__FreeBSD__) || defined(__FreeBSD_kernel__)
+OS_FREEBSD
 #endif
 
 #if defined(__NetBSD__)
-OS_NetBSD
+OS_NETBSD
 #endif
 
 #if defined(__sun)
-OS_SunOS
+OS_SUNOS
 #endif
 
 #if defined(__APPLE__)
-OS_Darwin
+OS_DARWIN
 #endif
 
 #if defined(_AIX)
@@ -63,13 +63,18 @@ OS_WINNT
 #endif
 
 #if defined(__CYGWIN__)
-OS_CYGWIN
+OS_CYGWIN_NT
 #endif
 
 #if defined(__INTERIX)
 OS_INTERIX
 #endif
 
+#if defined(__gnu_hurd__)
+/* Hurd is very similar to GNU/Linux, it should work out of the box */
+OS_LINUX
+#endif
+
 #if defined(__i386) || defined(_X86)
 ARCH_X86
 #endif

ctest/Makefile

@@ -5,7 +5,7 @@
 TOPDIR = ..
 include $(TOPDIR)/Makefile.system
 
-CFLAGS += -DADD$(BU) -DCBLAS
+override CFLAGS += -DADD$(BU) -DCBLAS
 
 LIB = $(TOPDIR)/$(LIBNAME)
 

driver/level2/sbmv_thread.c

@@ -65,7 +65,6 @@ static int sbmv_kernel(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, F
 
   a = (FLOAT *)args -> a;
   x = (FLOAT *)args -> b;
-  y = (FLOAT *)args -> c;
 
   lda  = args -> lda;
   incx = args -> ldb;
@@ -76,6 +75,10 @@ static int sbmv_kernel(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, F
   n_from = 0;
   n_to   = n;
 
+  //Use y as each thread's n* COMPSIZE elements in sb buffer
+  y = buffer;   
+  buffer += ((COMPSIZE * n  + 1023) & ~1023);
+
   if (range_m) {
     n_from = *(range_m + 0);
     n_to   = *(range_m + 1);
@@ -83,7 +86,6 @@ static int sbmv_kernel(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, F
     a += n_from * lda  * COMPSIZE;
   }
 
-  if (range_n) y += *range_n * COMPSIZE;
 
   if (incx != 1) {
     COPY_K(n, x, incx, buffer, 1);
@@ -331,7 +333,7 @@ int CNAME(BLASLONG n, BLASLONG k, FLOAT *alpha, FLOAT *a, BLASLONG lda, FLOAT *x
 
   if (num_cpu) {
     queue[0].sa = NULL;
-    queue[0].sb = buffer + num_cpu * (((n + 255) & ~255) + 16) * COMPSIZE;
+    queue[0].sb = buffer;
     queue[num_cpu - 1].next = NULL;
   
     exec_blas(num_cpu, queue);
@@ -344,7 +346,7 @@ int CNAME(BLASLONG n, BLASLONG k, FLOAT *alpha, FLOAT *a, BLASLONG lda, FLOAT *x
 #else
 	    ONE, ZERO,
 #endif
-	    buffer + range_n[i] * COMPSIZE, 1, buffer, 1, NULL, 0);
+	    (FLOAT*)(queue[i].sb), 1, buffer, 1, NULL, 0);
   }
 
   AXPYU_K(n, 0, 0,

driver/level3/gemm_thread_n.c

@@ -71,7 +71,7 @@ int CNAME(int mode, blas_arg_t *arg, BLASLONG *range_m, BLASLONG *range_n, int (
     queue[num_cpu].args    = arg;
     queue[num_cpu].range_m = range_m;
     queue[num_cpu].range_n = &range[num_cpu];
-#if defined(LOONGSON3A)
+#if 0   //defined(LOONGSON3A)
     queue[num_cpu].sa      = sa	+ GEMM_OFFSET_A1 * num_cpu;
     queue[num_cpu].sb      = queue[num_cpu].sa + GEMM_OFFSET_A1 * 5;
 #else
@@ -83,7 +83,7 @@ int CNAME(int mode, blas_arg_t *arg, BLASLONG *range_m, BLASLONG *range_n, int (
   }
   
   if (num_cpu) {
-#if defined(LOONGSON3A)
+#if 0 //defined(LOONGSON3A)
     queue[0].sa = sa;
     queue[0].sb = sa + GEMM_OFFSET_A1 * 5;
 #else

driver/others/Makefile

@@ -1,12 +1,12 @@
 TOPDIR	= ../..
 include ../../Makefile.system
 
-COMMONOBJS	 = memory.$(SUFFIX) xerbla.$(SUFFIX) c_abs.$(SUFFIX) z_abs.$(SUFFIX)
+COMMONOBJS	 = memory.$(SUFFIX) xerbla.$(SUFFIX) c_abs.$(SUFFIX) z_abs.$(SUFFIX) openblas_set_num_threads.$(SUFFIX) openblas_get_config.$(SUFFIX)
 
 COMMONOBJS	+= slamch.$(SUFFIX) slamc3.$(SUFFIX) dlamch.$(SUFFIX)  dlamc3.$(SUFFIX)
 
 ifdef SMP
-COMMONOBJS	+= blas_server.$(SUFFIX) divtable.$(SUFFIX) blasL1thread.$(SUFFIX) openblas_set_num_threads.$(SUFFIX)
+COMMONOBJS	+= blas_server.$(SUFFIX) divtable.$(SUFFIX) blasL1thread.$(SUFFIX) 
 ifndef NO_AFFINITY
 COMMONOBJS	+= init.$(SUFFIX) 
 endif
@@ -14,7 +14,7 @@ endif
 
 # COMMONOBJS	+=  info.$(SUFFIX)
 
-ifdef DYNAMIC_ARCH
+ifeq ($(DYNAMIC_ARCH), 1)
 COMMONOBJS	+=  dynamic.$(SUFFIX)
 else
 COMMONOBJS	+=  parameter.$(SUFFIX)
@@ -70,7 +70,7 @@ ifndef BLAS_SERVER
 BLAS_SERVER = blas_server.c
 endif
 
-ifdef DYNAMIC_ARCH
+ifeq ($(DYNAMIC_ARCH), 1)
 HPLOBJS = memory.$(SUFFIX) xerbla.$(SUFFIX) dynamic.$(SUFFIX)
 else
 HPLOBJS = memory.$(SUFFIX) xerbla.$(SUFFIX) parameter.$(SUFFIX)
@@ -103,6 +103,9 @@ blas_server.$(SUFFIX) : $(BLAS_SERVER) ../../common.h ../../common_thread.h ../.
 openblas_set_num_threads.$(SUFFIX) : openblas_set_num_threads.c
 	$(CC) $(CFLAGS) -c $< -o $(@F)
 
+openblas_get_config.$(SUFFIX) : openblas_get_config.c
+	$(CC) $(CFLAGS) -c $< -o $(@F)
+
 blasL1thread.$(SUFFIX) : blas_l1_thread.c ../../common.h ../../common_thread.h
 	$(CC) $(CFLAGS) -c $< -o $(@F)
 
@@ -215,7 +218,7 @@ info.$(SUFFIX) : info.c info.h ../../common.h ../../param.h
 	$(CC) $(CFLAGS) -c $< -o $(@F)
 
 
-hpl   : CFLAGS += -DHPL
-hpl_p : CFLAGS += -DHPL
+hpl   : override CFLAGS += -DHPL
+hpl_p : override CFLAGS += -DHPL
 
 include $(TOPDIR)/Makefile.tail

driver/others/blas_server.c

@@ -385,6 +385,7 @@ static int blas_thread_server(void *arg){
 					+ GEMM_ALIGN) & ~GEMM_ALIGN)) + GEMM_OFFSET_B);
 	  }
 	}
+	queue->sb=sb;
       }
 	
 #ifdef MONITOR
@@ -435,7 +436,7 @@ static int blas_thread_server(void *arg){
 
   blas_memory_free(buffer);
 
-  pthread_exit(NULL);
+  //pthread_exit(NULL);
 
   return 0;
 }
@@ -770,6 +771,19 @@ void goto_set_num_threads(int num_threads) {
 
   if (num_threads < 1) num_threads = blas_num_threads;
 
+#ifndef NO_AFFINITY
+  if (num_threads == 1) {
+    if (blas_cpu_number == 1){
+      //OpenBLAS is already single thread.
+      return; 
+    }else{
+      //From multi-threads to single thread
+      //Restore the original affinity mask
+      gotoblas_set_affinity(-1);
+    }
+  }
+#endif
+
   if (num_threads > MAX_CPU_NUMBER) num_threads = MAX_CPU_NUMBER;
 
   if (num_threads > blas_num_threads) {
@@ -800,6 +814,13 @@ void goto_set_num_threads(int num_threads) {
     UNLOCK_COMMAND(&server_lock);
   }
 
+#ifndef NO_AFFINITY
+  if(blas_cpu_number == 1 && num_threads > 1){
+    //Restore the thread 0 affinity.
+    gotoblas_set_affinity(0);
+  }
+#endif
+
   blas_cpu_number  = num_threads;
 
 #if defined(ARCH_MIPS64) 

driver/others/blas_server_omp.c

@@ -49,8 +49,12 @@
 
 int blas_server_avail = 0;
 
+static void * blas_thread_buffer[MAX_CPU_NUMBER];
+
 void goto_set_num_threads(int num_threads) {
 
+  int i=0;
+
   if (num_threads < 1) num_threads = blas_num_threads;
 
   if (num_threads > MAX_CPU_NUMBER) num_threads = MAX_CPU_NUMBER;
@@ -62,7 +66,19 @@ void goto_set_num_threads(int num_threads) {
   blas_cpu_number  = num_threads;
 
   omp_set_num_threads(blas_cpu_number);
-  
+
+  //adjust buffer for each thread
+  for(i=0; i<blas_cpu_number; i++){
+    if(blas_thread_buffer[i]==NULL){
+      blas_thread_buffer[i]=blas_memory_alloc(2);
+    }
+  }
+  for(; i<MAX_CPU_NUMBER; i++){
+    if(blas_thread_buffer[i]!=NULL){
+      blas_memory_free(blas_thread_buffer[i]);
+      blas_thread_buffer[i]=NULL;
+    }
+  }
 #if defined(ARCH_MIPS64) 
   //set parameters for different number of threads.
   blas_set_parameter();
@@ -76,17 +92,33 @@ void openblas_set_num_threads(int num_threads) {
 
 int blas_thread_init(void){
 
+  int i=0;
+
   blas_get_cpu_number();
 
   blas_server_avail = 1;
 
+  for(i=0; i<blas_num_threads; i++){
+    blas_thread_buffer[i]=blas_memory_alloc(2);
+  }
+  for(; i<MAX_CPU_NUMBER; i++){
+      blas_thread_buffer[i]=NULL;
+  }
+
   return 0;
 }
 
 int BLASFUNC(blas_thread_shutdown)(void){
-
+  int i=0;
   blas_server_avail = 0;
 
+  for(i=0; i<MAX_CPU_NUMBER; i++){
+    if(blas_thread_buffer[i]!=NULL){
+      blas_memory_free(blas_thread_buffer[i]);
+      blas_thread_buffer[i]=NULL;
+    }
+  }
+
   return 0;
 }
 
@@ -177,7 +209,8 @@ static void legacy_exec(void *func, int mode, blas_arg_t *args, void *sb){
 static void exec_threads(blas_queue_t *queue){
 
   void *buffer, *sa, *sb;
-
+  int pos=0, release_flag=0;
+  
   buffer = NULL;
   sa = queue -> sa;
   sb = queue -> sb;
@@ -189,7 +222,14 @@ static void exec_threads(blas_queue_t *queue){
 
   if ((sa == NULL) && (sb == NULL) && ((queue -> mode & BLAS_PTHREAD) == 0)) {
 
-    buffer = blas_memory_alloc(2);
+    pos = omp_get_thread_num();
+    buffer = blas_thread_buffer[pos];
+
+    //fallback
+    if(buffer==NULL) {
+      buffer = blas_memory_alloc(2);
+      release_flag=1;
+    }
 
     if (sa == NULL) sa = (void *)((BLASLONG)buffer + GEMM_OFFSET_A);
     
@@ -224,6 +264,7 @@ static void exec_threads(blas_queue_t *queue){
 					    + GEMM_ALIGN) & ~GEMM_ALIGN)) + GEMM_OFFSET_B);
 	  }
       }
+      queue->sb=sb;
     }
   }
 
@@ -241,7 +282,7 @@ static void exec_threads(blas_queue_t *queue){
 
     }
 
-  if (buffer != NULL) blas_memory_free(buffer);
+  if (release_flag) blas_memory_free(buffer);
 
 }
 

driver/others/blas_server_win32.c

@@ -63,6 +63,8 @@ static blas_pool_t   pool;
 static HANDLE	    blas_threads   [MAX_CPU_NUMBER];
 static DWORD	    blas_threads_id[MAX_CPU_NUMBER];
 
+
+
 static void legacy_exec(void *func, int mode, blas_arg_t *args, void *sb){
 
       if (!(mode & BLAS_COMPLEX)){
@@ -179,7 +181,7 @@ static DWORD WINAPI blas_thread_server(void *arg){
     
     do {
       action = WaitForMultipleObjects(2, handles, FALSE, INFINITE);
-    } while ((action != WAIT_OBJECT_0) && (action == WAIT_OBJECT_0 + 1));
+    } while ((action != WAIT_OBJECT_0) && (action != WAIT_OBJECT_0 + 1));
     
     if (action == WAIT_OBJECT_0 + 1) break;
 
@@ -251,6 +253,7 @@ static DWORD WINAPI blas_thread_server(void *arg){
 					  + GEMM_ALIGN) & ~GEMM_ALIGN)) + GEMM_OFFSET_B);
 	    }
 	}
+	queue->sb=sb;
       }
 	
 #ifdef MONITOR
@@ -263,7 +266,9 @@ static DWORD WINAPI blas_thread_server(void *arg){
       } else {
 	legacy_exec(routine, queue -> mode, queue -> args, sb);
       }
-    }
+    }else{
+		continue; //if queue == NULL
+	}
     
 #ifdef SMP_DEBUG
     fprintf(STDERR, "Server[%2ld] Finished!\n", cpu);
@@ -425,7 +430,7 @@ int exec_blas(BLASLONG num, blas_queue_t *queue){
 /* Shutdown procedure, but user don't have to call this routine. The */
 /* kernel automatically kill threads.                                */
 
-int blas_thread_shutdown_(void){
+int BLASFUNC(blas_thread_shutdown)(void){
 
   int i;
 
@@ -437,7 +442,7 @@ int blas_thread_shutdown_(void){
 
     SetEvent(pool.killed);
     
-    for(i = 0; i < blas_cpu_number - 1; i++){
+    for(i = 0; i < blas_num_threads - 1; i++){
       WaitForSingleObject(blas_threads[i], INFINITE);
     }
     
@@ -448,3 +453,47 @@ int blas_thread_shutdown_(void){
   
   return 0;
 }
+
+void goto_set_num_threads(int num_threads)
+{
+	 long i;
+
+	if (num_threads < 1) num_threads = blas_cpu_number;
+
+	if (num_threads > MAX_CPU_NUMBER) num_threads = MAX_CPU_NUMBER;
+
+	if (num_threads > blas_num_threads) {
+
+		LOCK_COMMAND(&server_lock);
+		
+		//increased_threads = 1;
+	    if (!blas_server_avail){
+
+			InitializeCriticalSection(&pool.lock);
+			pool.filled = CreateEvent(NULL, FALSE, FALSE, NULL);
+			pool.killed = CreateEvent(NULL, TRUE,  FALSE, NULL);
+
+			pool.shutdown = 0;
+			pool.queue    = NULL;
+			blas_server_avail = 1;
+		}
+		
+		for(i = blas_num_threads - 1; i < num_threads - 1; i++){	  
+		  
+			blas_threads[i] = CreateThread(NULL, 0, 
+				     blas_thread_server, (void *)i,
+				     0, &blas_threads_id[i]);
+		}
+	
+		blas_num_threads = num_threads;
+
+		UNLOCK_COMMAND(&server_lock);
+	}
+
+	blas_cpu_number  = num_threads;
+}
+
+void openblas_set_num_threads(int num)
+{
+	goto_set_num_threads(num);
+}

driver/others/dynamic.c

@@ -60,6 +60,16 @@ extern gotoblas_t  gotoblas_NEHALEM;
 extern gotoblas_t  gotoblas_OPTERON;
 extern gotoblas_t  gotoblas_OPTERON_SSE3;
 extern gotoblas_t  gotoblas_BARCELONA;
+extern gotoblas_t  gotoblas_BOBCAT;
+#ifndef NO_AVX
+extern gotoblas_t  gotoblas_SANDYBRIDGE;
+extern gotoblas_t  gotoblas_BULLDOZER;
+#else
+//Use NEHALEM kernels for sandy bridge
+#define gotoblas_SANDYBRIDGE gotoblas_NEHALEM
+#define gotoblas_BULLDOZER gotoblas_BARCELONA
+#endif
+
 
 #define VENDOR_INTEL      1
 #define VENDOR_AMD        2
@@ -68,6 +78,32 @@ extern gotoblas_t  gotoblas_BARCELONA;
 
 #define BITMASK(a, b, c) ((((a) >> (b)) & (c)))
 
+#ifndef NO_AVX
+static inline void xgetbv(int op, int * eax, int * edx){
+  //Use binary code for xgetbv
+  __asm__ __volatile__
+    (".byte 0x0f, 0x01, 0xd0": "=a" (*eax), "=d" (*edx) : "c" (op) : "cc");
+}
+#endif
+
+int support_avx(){
+#ifndef NO_AVX
+  int eax, ebx, ecx, edx;
+  int ret=0;
+  
+  cpuid(1, &eax, &ebx, &ecx, &edx);
+  if ((ecx & (1 << 28)) != 0 && (ecx & (1 << 27)) != 0){
+    xgetbv(0, &eax, &edx);
+    if((eax & 6) == 6){
+      ret=1;  //OS support AVX
+    }
+  }
+  return ret;
+#else
+  return 0;
+#endif
+}
+
 static int get_vendor(void){
   int eax, ebx, ecx, edx;
   char vendor[13];
@@ -122,15 +158,39 @@ static gotoblas_t *get_coretype(void){
 	if (model == 12) return &gotoblas_ATOM;
 	return NULL;
 
-	  case 2:
-		  //Intel Core (Clarkdale) / Core (Arrandale)
-		  // Pentium (Clarkdale) / Pentium Mobile (Arrandale)
-		  // Xeon (Clarkdale), 32nm
-		  if (model ==  5) return &gotoblas_NEHALEM;
+      case 2:
+	//Intel Core (Clarkdale) / Core (Arrandale)
+	// Pentium (Clarkdale) / Pentium Mobile (Arrandale)
+	// Xeon (Clarkdale), 32nm
+	if (model ==  5) return &gotoblas_NEHALEM;
 		  
-		  //Intel Xeon Processor 5600 (Westmere-EP)
-		  if (model == 12) return &gotoblas_NEHALEM;
-		  return NULL;
+	//Intel Xeon Processor 5600 (Westmere-EP)
+	//Xeon Processor E7 (Westmere-EX)
+	//Xeon E7540
+	if (model == 12 || model == 14 || model == 15) return &gotoblas_NEHALEM;
+
+	//Intel Core i5-2000 /i7-2000 (Sandy Bridge)
+	//Intel Core i7-3000 / Xeon E5
+	if (model == 10 || model == 13) {
+	  if(support_avx())
+	    return &gotoblas_SANDYBRIDGE;
+	  else{
+	    fprintf(stderr, "OpenBLAS : Your OS does not support AVX instructions. OpenBLAS is using Nehalem kernels as a fallback, which may give poorer performance.\n");
+	    return &gotoblas_NEHALEM; //OS doesn't support AVX. Use old kernels.
+	  }
+	}
+	return NULL;
+      case 3:
+	//Intel Sandy Bridge 22nm (Ivy Bridge?)
+	if (model == 10) {
+	  if(support_avx())
+	    return &gotoblas_SANDYBRIDGE;
+	  else{
+	    fprintf(stderr, "OpenBLAS : Your OS does not support AVX instructions. OpenBLAS is using Nehalem kernels as a fallback, which may give poorer performance.\n");
+	    return &gotoblas_NEHALEM; //OS doesn't support AVX. Use old kernels.
+	  }
+	}
+	return NULL;
       }
       case 0xf:
       if (model <= 0x2) return &gotoblas_NORTHWOOD;
@@ -144,7 +204,17 @@ static gotoblas_t *get_coretype(void){
       if ((exfamily == 0) || (exfamily == 2)) {
 	if (ecx & (1 <<  0)) return &gotoblas_OPTERON_SSE3; 
 	else return &gotoblas_OPTERON;
-      }  else {
+      }  else if (exfamily == 5) {
+	return &gotoblas_BOBCAT;
+      } else if (exfamily == 6) {
+	//AMD Bulldozer Opteron 6200 / Opteron 4200 / AMD FX-Series
+	  if(support_avx())
+	    return &gotoblas_BULLDOZER;
+	  else{
+	    fprintf(stderr, "OpenBLAS : Your OS does not support AVX instructions. OpenBLAS is using Barcelona kernels as a fallback, which may give poorer performance.\n");
+	    return &gotoblas_BARCELONA; //OS doesn't support AVX. Use old kernels.
+	  }	
+      } else {
 	return &gotoblas_BARCELONA;
       }
     }
@@ -178,6 +248,9 @@ static char *corename[] = {
     "Opteron(SSE3)",
     "Barcelona",
     "Nano",
+    "Sandybridge",
+    "Bobcat",
+    "Bulldozer",
 };
 
 char *gotoblas_corename(void) {
@@ -197,7 +270,10 @@ char *gotoblas_corename(void) {
   if (gotoblas == &gotoblas_OPTERON)      return corename[13];
   if (gotoblas == &gotoblas_BARCELONA)    return corename[14];
   if (gotoblas == &gotoblas_NANO)         return corename[15];
-  
+  if (gotoblas == &gotoblas_SANDYBRIDGE)  return corename[16];
+  if (gotoblas == &gotoblas_BOBCAT)       return corename[17];
+  if (gotoblas == &gotoblas_BULLDOZER)    return corename[18];
+
   return corename[0];
 }
 
@@ -211,12 +287,21 @@ void gotoblas_dynamic_init(void) {
   if (gotoblas == NULL) gotoblas = &gotoblas_KATMAI;
 #else
   if (gotoblas == NULL) gotoblas = &gotoblas_PRESCOTT;
+  /* sanity check, if 64bit pointer we can't have a 32 bit cpu */
+  if (sizeof(void*) == 8) {
+      if (gotoblas == &gotoblas_KATMAI ||
+          gotoblas == &gotoblas_COPPERMINE ||
+          gotoblas == &gotoblas_NORTHWOOD ||
+          gotoblas == &gotoblas_BANIAS ||
+          gotoblas == &gotoblas_ATHLON)
+          gotoblas = &gotoblas_PRESCOTT;
+  }
 #endif
   
   if (gotoblas && gotoblas -> init) {
     gotoblas -> init();
   } else {
-    fprintf(stderr, "GotoBLAS : Architecture Initialization failed. No initialization function found.\n");
+    fprintf(stderr, "OpenBLAS : Architecture Initialization failed. No initialization function found.\n");
     exit(1);
   }
   

driver/others/init.c

@@ -1,5 +1,5 @@
 /*****************************************************************************
-Copyright (c) 2011, Lab of Parallel Software and Computational Science,ICSAS
+Copyright (c) 2011,2012 Lab of Parallel Software and Computational Science,ISCAS
 All rights reserved.
 
 Redistribution and use in source and binary forms, with or without
@@ -85,6 +85,11 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 #define MAX_NODES	16
 #define MAX_CPUS	256
+#define NCPUBITS        (8*sizeof(unsigned long))
+#define MAX_BITMASK_LEN (MAX_CPUS/NCPUBITS)
+#define CPUELT(cpu)	((cpu) / NCPUBITS)
+#define CPUMASK(cpu)	((unsigned long) 1UL << ((cpu) % NCPUBITS))
+
 
 #define SH_MAGIC	0x510510
 
@@ -103,10 +108,10 @@ typedef struct {
   int num_nodes;
   int num_procs;
   int final_num_procs;
-  unsigned long avail;
-  
+  unsigned long avail [MAX_BITMASK_LEN];
+  int avail_count;
   unsigned long cpu_info   [MAX_CPUS];
-  unsigned long node_info  [MAX_NODES];
+  unsigned long node_info  [MAX_NODES][MAX_BITMASK_LEN];
   int cpu_use[MAX_CPUS];
 
 } shm_t;
@@ -126,7 +131,8 @@ static shm_t *common = (void *)-1;
 static int shmid, pshmid;
 static void *paddr;
 
-static unsigned long lprocmask, lnodemask;
+static unsigned long lprocmask[MAX_BITMASK_LEN], lnodemask;
+static int lprocmask_count = 0;
 static int numprocs = 1;
 static int numnodes = 1;
 
@@ -177,70 +183,114 @@ static inline int rcount(unsigned long number) {
   than sizeof(unsigned long). On 64 bits, the limit 
   is 64. On 32 bits, it is 32.
 ***/
-static inline unsigned long get_cpumap(int node) {
+static inline void get_cpumap(int node, unsigned long * node_info) {
 
   int infile;
-  unsigned long affinity;
+  unsigned long affinity[32];
   char name[160];
   char cpumap[160];
-  char *p, *dummy;
+  char *dummy;
   int i=0;
+  int count=0;
+  int k=0;
 
   sprintf(name, CPUMAP_NAME, node);
   
   infile = open(name, O_RDONLY);
+  for(i=0; i<32; i++){
+    affinity[i] = 0;
+  }
 
-  affinity = 0;
-    
   if (infile != -1) {
     
     read(infile, cpumap, sizeof(cpumap));
-    p = cpumap;
-    while (*p != '\n' && i<160){
-      if(*p != ',') {
-	name[i++]=*p;
-      }
-      p++;
-    }
-    p = name;
 
-    //    while ((*p == '0') || (*p == ',')) p++;
+    for(i=0; i<160; i++){
+      if(cpumap[i] == '\n')
+	break;
+      if(cpumap[i] != ','){
+	name[k++]=cpumap[i];
+	
+	//Enough data for Hex
+	if(k >= NCPUBITS/4){
+	  affinity[count++] = strtoul(name, &dummy, 16);
+	  k=0;
+	}
+      }
 
-    affinity = strtoul(p, &dummy, 16);
-   
+    }
+    if(k!=0){
+      name[k]='\0';
+      affinity[count++] = strtoul(name, &dummy, 16);
+      k=0;
+    }
+    // 0-63bit -> node_info[0], 64-128bit -> node_info[1] ....
+    // revert the sequence
+    for(i=0; i<count && i<MAX_BITMASK_LEN; i++){
+      node_info[i]=affinity[count-i-1];
+    }
     close(infile);
   }
 
-  return affinity;
+  return ;
 }
 
-static inline unsigned long get_share(int cpu, int level) {
+static inline void get_share(int cpu, int level, unsigned long * share) {
 
   int infile;
-  unsigned long affinity;
+  unsigned long affinity[32];
+  char cpumap[160];
   char name[160];
-  char *p;
-  
+  char *dummy;
+  int count=0;
+  int i=0,k=0;
+  int bitmask_idx = 0;
+
   sprintf(name, SHARE_NAME, cpu, level);
   
   infile = open(name, O_RDONLY);
 
-  affinity = (1UL << cpu);
-    
+  //  Init share
+  for(i=0; i<MAX_BITMASK_LEN; i++){
+    share[i]=0;
+  }
+  bitmask_idx = CPUELT(cpu);
+  share[bitmask_idx] = CPUMASK(cpu);
+
   if (infile != -1) {
     
-    read(infile, name, sizeof(name));
-   
-    p = name;
+    read(infile, cpumap, sizeof(cpumap));
 
-    while ((*p == '0') || (*p == ',')) p++;
+    for(i=0; i<160; i++){
+      if(cpumap[i] == '\n')
+	break;
+      if(cpumap[i] != ','){
+	name[k++]=cpumap[i];
+	
+	//Enough data 
+	if(k >= NCPUBITS/4){
+	  affinity[count++] = strtoul(name, &dummy, 16);
+	  k=0;
+	}
+      }
 
-    affinity = strtol(p, &p, 16);
+    }
+    if(k!=0){
+      name[k]='\0';
+      affinity[count++] = strtoul(name, &dummy, 16);
+      k=0;
+    }
+    // 0-63bit -> node_info[0], 64-128bit -> node_info[1] ....
+    // revert the sequence
+    for(i=0; i<count && i<MAX_BITMASK_LEN; i++){
+      share[i]=affinity[count-i-1];
+    }
+   
    
     close(infile);
   }
 
-  return affinity;
+  return ;
 }
 
 static int numa_check(void) {
@@ -248,6 +298,7 @@ static int numa_check(void) {
   DIR *dp;
   struct dirent *dir;
   int node;
+  int j;
 
   common -> num_nodes = 0;
 
@@ -258,7 +309,9 @@ static int numa_check(void) {
     return 0;
   }
 
-  for (node = 0; node < MAX_NODES; node ++) common -> node_info[node] = 0;
+  for (node = 0; node < MAX_NODES; node ++) {
+    for (j = 0; j<MAX_BITMASK_LEN; j++) common -> node_info[node][j] = 0;
+  }
 
   while ((dir = readdir(dp)) != NULL) {
     if (*(unsigned int *) dir -> d_name == 0x065646f6eU) {
@@ -266,12 +319,12 @@ static int numa_check(void) {
       node = atoi(&dir -> d_name[4]);
 
       if (node > MAX_NODES) {
-	fprintf(stderr, "\nGotoBLAS Warining : MAX_NODES (NUMA) is too small. Terminated.\n");
+	fprintf(stderr, "\nOpenBLAS Warning : MAX_NODES (NUMA) is too small. Terminated.\n");
 	exit(1);
       }
 
       common -> num_nodes ++;
-      common -> node_info[node] = get_cpumap(node);
+      get_cpumap(node, common->node_info[node]);
 
     }
   }
@@ -284,7 +337,7 @@ static int numa_check(void) {
   fprintf(stderr, "Numa found : number of Nodes = %2d\n", common -> num_nodes);
 
   for (node = 0; node < common -> num_nodes; node ++)
-    fprintf(stderr, "MASK (%2d) : %08lx\n", node, common -> node_info[node]);
+    fprintf(stderr, "MASK (%2d) : %08lx\n", node, common -> node_info[node][0]);
 #endif
 
   return common -> num_nodes;
@@ -296,11 +349,13 @@ static void numa_mapping(void) {
   int i, j, h;
   unsigned long work, bit;
   int count = 0;
+  int bitmask_idx = 0;
 
   for (node = 0; node < common -> num_nodes; node ++) {
     core = 0;
     for (cpu = 0; cpu < common -> num_procs; cpu ++) {
-      if (common -> node_info[node] & common -> avail & (1UL << cpu)) {
+      bitmask_idx = CPUELT(cpu);
+      if (common -> node_info[node][bitmask_idx] & common -> avail[bitmask_idx] & CPUMASK(cpu)) {
 	common -> cpu_info[count] = WRITE_CORE(core) | WRITE_NODE(node) | WRITE_CPU(cpu);
 	count ++;
 	core ++;
@@ -357,58 +412,92 @@ static void numa_mapping(void) {
 
 static void disable_hyperthread(void) {
 
-  unsigned long share;
+  unsigned long share[MAX_BITMASK_LEN];
   int cpu;
+  int bitmask_idx = 0;
+  int i=0, count=0;
+  bitmask_idx = CPUELT(common -> num_procs);
 
-  if(common->num_procs > 64){
-    fprintf(stderr, "\nOpenBLAS Warining : The number of CPU/Cores(%d) is beyond the limit(64). Terminated.\n", common->num_procs);
-    exit(1);
-  }else if(common->num_procs == 64){
-    common -> avail = 0xFFFFFFFFFFFFFFFFUL;
-  }else
-    common -> avail = (1UL << common -> num_procs) - 1;
+  for(i=0; i< bitmask_idx; i++){
+    common -> avail[count++] = 0xFFFFFFFFFFFFFFFFUL;
+  }
+  if(CPUMASK(common -> num_procs) != 1){
+    common -> avail[count++] = CPUMASK(common -> num_procs) - 1;
+  }
+  common -> avail_count = count;
+
+  /* if(common->num_procs > 64){ */
+  /*   fprintf(stderr, "\nOpenBLAS Warning : The number of CPU/Cores(%d) is beyond the limit(64). Terminated.\n", common->num_procs); */
+  /*   exit(1); */
+  /* }else if(common->num_procs == 64){ */
+  /*   common -> avail = 0xFFFFFFFFFFFFFFFFUL; */
+  /* }else */
+  /*   common -> avail = (1UL << common -> num_procs) - 1; */
 
 #ifdef DEBUG
-  fprintf(stderr, "\nAvail CPUs    : %04lx.\n", common -> avail);
+  fprintf(stderr, "\nAvail CPUs    : ");
+  for(i=0; i<count; i++)
+    fprintf(stderr, "%04lx ", common -> avail[i]);
+  fprintf(stderr, ".\n");
 #endif
 
   for (cpu = 0; cpu < common -> num_procs; cpu ++) {
-    
-    share = (get_share(cpu, 1) & common -> avail);
-    
-    if (popcount(share) > 1) {
+
+    get_share(cpu, 1, share);
+
+    //When the shared cpu are in different element of share & avail array, this may be a bug.
+    for (i = 0; i < count ; i++){
+
+      share[i] &= common->avail[i];
+
+      if (popcount(share[i]) > 1) {
       
 #ifdef DEBUG
-      fprintf(stderr, "Detected Hyper Threading on CPU %4x; disabled CPU %04lx.\n",
-	      cpu, share & ~(1UL << cpu));
+	fprintf(stderr, "Detected Hyper Threading on CPU %4x; disabled CPU %04lx.\n",
+		cpu, share[i] & ~(CPUMASK(cpu)));
 #endif
       
-      common -> avail &= ~((share & ~(1UL << cpu)));
+	common -> avail[i] &= ~((share[i] & ~ CPUMASK(cpu)));
+      }
     }
   }
 }
 
 static void disable_affinity(void) {
-
+  int i=0;
+  int bitmask_idx=0;
+  int count=0;
 #ifdef DEBUG
-    fprintf(stderr, "Final all available CPUs  : %04lx.\n\n", common -> avail);
+    fprintf(stderr, "Final all available CPUs  : %04lx.\n\n", common -> avail[0]);
     fprintf(stderr, "CPU mask                  : %04lx.\n\n", *(unsigned long *)&cpu_orig_mask[0]);
 #endif
 
-  if(common->final_num_procs > 64){
-    fprintf(stderr, "\nOpenBLAS Warining : The number of CPU/Cores(%d) is beyond the limit(64). Terminated.\n", common->final_num_procs);
-    exit(1);
-  }else if(common->final_num_procs == 64){
-    lprocmask = 0xFFFFFFFFFFFFFFFFUL;
-  }else
-    lprocmask = (1UL << common -> final_num_procs) - 1;
+  /* if(common->final_num_procs > 64){ */
+  /*   fprintf(stderr, "\nOpenBLAS Warining : The number of CPU/Cores(%d) is beyond the limit(64). Terminated.\n", common->final_num_procs); */
+  /*   exit(1); */
+  /* }else if(common->final_num_procs == 64){ */
+  /*   lprocmask = 0xFFFFFFFFFFFFFFFFUL; */
+  /* }else */
+  /*   lprocmask = (1UL << common -> final_num_procs) - 1; */
+
+  bitmask_idx = CPUELT(common -> final_num_procs);
+
+  for(i=0; i< bitmask_idx; i++){
+    lprocmask[count++] = 0xFFFFFFFFFFFFFFFFUL;
+  }
+  if(CPUMASK(common -> final_num_procs) != 1){
+    lprocmask[count++] = CPUMASK(common -> final_num_procs) - 1;
+  }
+  lprocmask_count = count;
 
 #ifndef USE_OPENMP
-  lprocmask &= *(unsigned long *)&cpu_orig_mask[0];
+  for(i=0; i< count; i++){
+    lprocmask[i] &= ((unsigned long *)&cpu_orig_mask[0])[i];
+  }
 #endif
 
 #ifdef DEBUG
-    fprintf(stderr, "I choose these CPUs  : %04lx.\n\n", lprocmask);
+    fprintf(stderr, "I choose these CPUs  : %04lx.\n\n", lprocmask[0]);
 #endif
 
 }
@@ -498,7 +587,7 @@ static void create_pshmem(void) {
 static void local_cpu_map(void) {
 
   int cpu, id, mapping;
-
+  int bitmask_idx = 0;
   cpu = 0;
   mapping = 0;
 
@@ -508,8 +597,9 @@ static void local_cpu_map(void) {
     if (id > 0) {
       if (is_dead(id)) common -> cpu_use[cpu] = 0;
     }
-
-    if ((common -> cpu_use[cpu] == 0) && (lprocmask & (1UL << cpu))) {
+    
+    bitmask_idx = CPUELT(cpu);
+    if ((common -> cpu_use[cpu] == 0) && (lprocmask[bitmask_idx] & CPUMASK(cpu))) {
 
       common -> cpu_use[cpu] = pshmid;
       cpu_mapping[mapping] = READ_CPU(common -> cpu_info[cpu]);
@@ -595,6 +685,7 @@ void gotoblas_affinity_init(void) {
 #ifndef USE_OPENMP
   cpu_set_t cpu_mask;
 #endif
+  int i;
 
   if (initialized) return;
 
@@ -646,6 +737,11 @@ void gotoblas_affinity_init(void) {
 
     common -> num_procs = get_nprocs();
 
+    if(common -> num_procs > MAX_CPUS) {
+      fprintf(stderr, "\nOpenBLAS Warining : The number of CPU/Cores(%d) is beyond the limit(%d). Terminated.\n", common->num_procs, MAX_CPUS);
+      exit(1);
+    }
+
     for (cpu = 0; cpu < common -> num_procs; cpu++) common -> cpu_info[cpu] = cpu;
     
     numa_check();
@@ -654,7 +750,8 @@ void gotoblas_affinity_init(void) {
 
     if (common -> num_nodes > 1) numa_mapping();
 
-    common -> final_num_procs = popcount(common -> avail);
+    common -> final_num_procs = 0;
+    for(i = 0; i < common -> avail_count; i++) common -> final_num_procs += popcount(common -> avail[i]);
 
     for (cpu = 0; cpu < common -> final_num_procs; cpu ++) common -> cpu_use[cpu] =  0;
 
@@ -664,7 +761,8 @@ void gotoblas_affinity_init(void) {
 
   disable_affinity();
 
-  num_avail = popcount(lprocmask);
+  num_avail = 0;
+  for(i=0; i<lprocmask_count; i++) num_avail += popcount(lprocmask[i]);
 
   if ((numprocs <= 0) || (numprocs > num_avail)) numprocs = num_avail;
 

driver/others/memory.c

@@ -1,5 +1,5 @@
 /*****************************************************************************
-Copyright (c) 2011, Lab of Parallel Software and Computational Science,ICSAS
+Copyright (c) 2011,2012 Lab of Parallel Software and Computational Science,ISCAS
 All rights reserved.
 
 Redistribution and use in source and binary forms, with or without
@@ -103,7 +103,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #include <sys/syscall.h>
 #endif
 
-#if defined(OS_FreeBSD) || defined(OS_Darwin)
+#if defined(OS_FREEBSD) || defined(OS_DARWIN)
 #include <sys/sysctl.h>
 #endif
 
@@ -185,7 +185,7 @@ int get_num_procs(void) {
 
 #endif
 
-#if defined(OS_FreeBSD) || defined(OS_Darwin)
+#if defined(OS_FREEBSD) 
 
 int get_num_procs(void) {
   
@@ -206,7 +206,27 @@ int get_num_procs(void) {
 
 #endif
 
+#if defined(OS_DARWIN)
+int get_num_procs(void) {
+  static int nums = 0;
+  size_t len;
+  if (nums == 0){
+    len = sizeof(int);
+    sysctlbyname("hw.physicalcpu", &nums, &len, NULL, 0);
+  }
+  return nums;
+}
+#endif
+
+/*
+OpenBLAS uses the numbers of CPU cores in multithreading. 
+It can be set by openblas_set_num_threads(int num_threads);
+*/
 int blas_cpu_number  = 0;
+/*
+The numbers of threads in the thread pool. 
+This value is equal or large than blas_cpu_number. This means some threads are sleep.
+*/
 int blas_num_threads = 0;
 
 int  goto_get_num_procs  (void) {
@@ -215,7 +235,7 @@ int  goto_get_num_procs  (void) {
 
 int blas_get_cpu_number(void){
   char *p;
-#if defined(OS_LINUX) || defined(OS_WINDOWS) || defined(OS_FreeBSD) || defined(OS_Darwin)
+#if defined(OS_LINUX) || defined(OS_WINDOWS) || defined(OS_FREEBSD) || defined(OS_DARWIN)
   int max_num;
 #endif
   int blas_goto_num   = 0;
@@ -223,7 +243,7 @@ int blas_get_cpu_number(void){
 
   if (blas_num_threads) return blas_num_threads;
 
-#if defined(OS_LINUX) || defined(OS_WINDOWS) || defined(OS_FreeBSD) || defined(OS_Darwin)
+#if defined(OS_LINUX) || defined(OS_WINDOWS) || defined(OS_FREEBSD) || defined(OS_DARWIN)
   max_num = get_num_procs();
 #endif
 
@@ -250,7 +270,7 @@ int blas_get_cpu_number(void){
   else if (blas_omp_num > 0) blas_num_threads = blas_omp_num;
   else blas_num_threads = MAX_CPU_NUMBER;
 
-#if defined(OS_LINUX) || defined(OS_WINDOWS) || defined(OS_FreeBSD) || defined(OS_Darwin)
+#if defined(OS_LINUX) || defined(OS_WINDOWS) || defined(OS_FREEBSD) || defined(OS_DARWIN)
   if (blas_num_threads > max_num) blas_num_threads = max_num;
 #endif
 
@@ -1128,7 +1148,7 @@ static BLASULONG   init_lock = 0UL;
 static void _touch_memory(blas_arg_t *arg, BLASLONG *range_m, BLASLONG *range_n, 
 			  void *sa, void *sb, BLASLONG pos) {
 
-#ifndef ARCH_POWER
+#if !defined(ARCH_POWER) && !defined(ARCH_SPARC)
 
   long size;
   BLASULONG buffer;
@@ -1289,6 +1309,7 @@ void DESTRUCTOR gotoblas_quit(void) {
    moncontrol (1);
 #endif
 
+   blas_shutdown();
 }
 
 #if (defined(C_PGI) || (!defined(C_SUN) && defined(F_INTERFACE_SUN))) && (defined(ARCH_X86) || defined(ARCH_X86_64))

driver/others/openblas_get_config.c

@@ -0,0 +1,59 @@
+/*****************************************************************************
+Copyright (c) 2011,2012 Lab of Parallel Software and Computational Science,ISCAS
+All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+
+   1. Redistributions of source code must retain the above copyright
+      notice, this list of conditions and the following disclaimer.
+
+   2. Redistributions in binary form must reproduce the above copyright
+      notice, this list of conditions and the following disclaimer in
+      the documentation and/or other materials provided with the
+      distribution.
+   3. Neither the name of the ISCAS nor the names of its contributors may 
+      be used to endorse or promote products derived from this software 
+      without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 
+ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE 
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR 
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER 
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, 
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE 
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+**********************************************************************************/
+
+#include "common.h"
+
+static char* openblas_config_str=""
+#ifdef USE64BITINT
+  "USE64BITINT "
+#endif
+#ifdef NO_CBLAS
+  "NO_CBLAS "
+#endif
+#ifdef NO_LAPACK
+  "NO_LAPACK "
+#endif
+#ifdef NO_LAPACKE
+  "NO_LAPACKE "
+#endif
+#ifdef DYNAMIC_ARCH
+  "DYNAMIC_ARCH "
+#endif
+#ifdef NO_AFFINITY
+  "NO_AFFINITY "
+#endif
+  ;
+
+char* CNAME() {
+  return openblas_config_str;
+}
+

driver/others/openblas_set_num_threads.c

@@ -1,5 +1,5 @@
 /*****************************************************************************
-Copyright (c) 2011, Lab of Parallel Software and Computational Science,ICSAS
+Copyright (c) 2011,2012 Lab of Parallel Software and Computational Science,ISCAS
 All rights reserved.
 
 Redistribution and use in source and binary forms, with or without
@@ -33,13 +33,20 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #include "common.h"
 
 #ifdef SMP_SERVER
-#ifdef OS_LINUX
 
 extern  void openblas_set_num_threads(int num_threads) ;
 
-void NAME(int* num_threads){
+void openblas_set_num_threads_(int* num_threads){
 	openblas_set_num_threads(*num_threads);
 }
 
-#endif
+#else
+//Single thread
+
+void openblas_set_num_threads(int num_threads) {
+}
+
+void openblas_set_num_threads_(int* num_threads){
+
+}
 #endif

driver/others/parameter.c

@@ -163,9 +163,9 @@ int get_L2_size(void){
 
   int eax, ebx, ecx, edx;
 
-#if defined(ATHLON) || defined(OPTERON) || defined(BARCELONA) || \
+#if defined(ATHLON) || defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER) || \
     defined(CORE_PRESCOTT) || defined(CORE_CORE2) || defined(PENRYN) || defined(DUNNINGTON) || \
-    defined(CORE_NEHALEM) || defined(ATOM) || defined(GENERIC)
+  defined(CORE_NEHALEM) || defined(CORE_SANDYBRIDGE) || defined(ATOM) || defined(GENERIC)
 
   cpuid(0x80000006, &eax, &ebx, &ecx, &edx);
 
@@ -384,6 +384,17 @@ void blas_set_parameter(void){
 #endif
 #endif
 
+#if defined(SANDYBRIDGE)
+  sgemm_p = 1024;
+  dgemm_p =  512;
+  cgemm_p =  512;
+  zgemm_p =  256;
+#ifdef EXPRECISION
+  qgemm_p =  256;
+  xgemm_p =  128;
+#endif
+#endif
+
 #if defined(CORE_PRESCOTT)  || defined(GENERIC)
   size >>= 6;
 
@@ -435,7 +446,7 @@ void blas_set_parameter(void){
 #endif
 #endif
 
-#if defined(CORE_BARCELONA)
+#if defined(CORE_BARCELONA) || defined(CORE_BOBCAT)
   size >>= 8;
 
   sgemm_p = 232 * size;

exports/Makefile

@@ -10,10 +10,23 @@ ifndef NO_CBLAS
 NO_CBLAS	= 0
 endif
 
+ifndef NO_LAPACK
+NO_LAPACK	= 0
+endif
+
+ifndef NO_LAPACKE
+NO_LAPACKE	= 0
+endif
+
 ifeq ($(OSNAME), WINNT)
 ifeq ($(F_COMPILER), GFORTRAN)
 EXTRALIB += -lgfortran
 endif
+ifeq ($(USE_OPENMP), 1)
+ifeq ($(C_COMPILER), GCC)
+EXTRALIB += -lgomp
+endif
+endif
 endif
 
 ifeq ($(OSNAME), CYGWIN_NT)
@@ -58,15 +71,20 @@ dll  : ../$(LIBDLLNAME)
 
 dll2 : libgoto2_shared.dll
 
+# On Windows, we only generate a DLL without a version suffix.  This is because
+# applications which link against the dynamic library reference a fixed DLL name
+# in their import table.  By instead using a stable name it is possible to
+# upgrade between library versions, without needing to re-link an application.
+# For more details see: https://github.com/xianyi/OpenBLAS/issues/127.
 ../$(LIBDLLNAME) : ../$(LIBNAME) libopenblas.def dllinit.$(SUFFIX)
 	$(RANLIB) ../$(LIBNAME)
 ifeq ($(BINARY32), 1)
 	$(DLLWRAP) -o ../$(LIBDLLNAME) --def libopenblas.def \
-	--entry _dllinit@12 -s dllinit.$(SUFFIX) --dllname $(@F) ../$(LIBNAME) $(FEXTRALIB)
+	--entry _dllinit@12 -s dllinit.$(SUFFIX) --dllname $(@F) ../$(LIBNAME) $(EXTRALIB)
 	-lib /machine:i386 /def:libopenblas.def
 else
 	$(DLLWRAP) -o ../$(LIBDLLNAME) --def libopenblas.def \
-	--entry $(FU)dllinit    -s dllinit.$(SUFFIX) --dllname $(@F) ../$(LIBNAME) $(FEXTRALIB)
+	--entry $(FU)dllinit    -s dllinit.$(SUFFIX) --dllname $(@F) ../$(LIBNAME) $(EXTRALIB)
 	-lib /machine:X64 /def:libopenblas.def 
 endif
 
@@ -76,13 +94,13 @@ libgoto2_shared.dll : ../$(LIBNAME) libgoto2_shared.def
 	-Wl,--out-implib,libgoto2_shared.lib $(FEXTRALIB)
 
 libopenblas.def : gensymbol
-	perl ./gensymbol win2k    $(ARCH) dummy $(EXPRECISION) $(NO_CBLAS) $(NO_LAPACK) > $(@F)
+	perl ./gensymbol win2k    $(ARCH) dummy $(EXPRECISION) $(NO_CBLAS) $(NO_LAPACK) $(NO_LAPACKE) > $(@F)
 
 libgoto2_shared.def : gensymbol
-	perl ./gensymbol win2k    $(ARCH) dummy $(EXPRECISION) $(NO_CBLAS) $(NO_LAPACK)  > $(@F)
+	perl ./gensymbol win2k    $(ARCH) dummy $(EXPRECISION) $(NO_CBLAS) $(NO_LAPACK) $(NO_LAPACKE) > $(@F)
 
 libgoto_hpl.def : gensymbol
-	perl ./gensymbol win2khpl $(ARCH) dummy $(EXPRECISION) $(NO_CBLAS) $(NO_LAPACK)  > $(@F)
+	perl ./gensymbol win2khpl $(ARCH) dummy $(EXPRECISION) $(NO_CBLAS) $(NO_LAPACK) $(NO_LAPACKE) > $(@F)
 
 $(LIBDYNNAME) : ../$(LIBNAME) osx.def
 	$(CC) $(CFLAGS) -all_load -headerpad_max_install_names -install_name $(CURDIR)/../$(LIBDYNNAME) -dynamiclib -o ../$(LIBDYNNAME) $< -Wl,-exported_symbols_list,osx.def  $(FEXTRALIB)
@@ -106,14 +124,15 @@ so : ../$(LIBSONAME)
 
 endif
 
-ifeq ($(OSNAME), FreeBSD)
+#http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
+ifeq ($(OSNAME), $(filter $(OSNAME),FreeBSD NetBSD))
 
 so : ../$(LIBSONAME)
 
 ../$(LIBSONAME) : ../$(LIBNAME) linux.def linktest.c
 	$(CC) $(CFLAGS)  -shared -o ../$(LIBSONAME) \
 	-Wl,--whole-archive ../$(LIBNAME) -Wl,--no-whole-archive \
-	-Wl,--retain-symbols-file=linux.def $(EXTRALIB)
+	-Wl,--retain-symbols-file=linux.def $(FEXTRALIB) $(EXTRALIB)
 	$(CC) $(CFLAGS) -w -o linktest linktest.c ../$(LIBSONAME) $(FEXTRALIB) && echo OK.
 	rm -f linktest
 
@@ -163,23 +182,23 @@ static : ../$(LIBNAME)
 	rm -f goto.$(SUFFIX)
 
 linux.def : gensymbol ../Makefile.system ../getarch.c
-	perl ./gensymbol linux $(ARCH) $(BU) $(EXPRECISION) $(NO_CBLAS)  $(NO_LAPACK) > $(@F)
+	perl ./gensymbol linux $(ARCH) $(BU) $(EXPRECISION) $(NO_CBLAS)  $(NO_LAPACK) $(NO_LAPACKE) > $(@F)
 
 osx.def : gensymbol ../Makefile.system ../getarch.c
-	perl ./gensymbol osx $(ARCH) $(BU) $(EXPRECISION) $(NO_CBLAS)  $(NO_LAPACK) > $(@F)
+	perl ./gensymbol osx $(ARCH) $(BU) $(EXPRECISION) $(NO_CBLAS)  $(NO_LAPACK) $(NO_LAPACKE) > $(@F)
 
 aix.def : gensymbol ../Makefile.system ../getarch.c
-	perl ./gensymbol aix $(ARCH) $(BU) $(EXPRECISION) $(NO_CBLAS)  $(NO_LAPACK) > $(@F)
+	perl ./gensymbol aix $(ARCH) $(BU) $(EXPRECISION) $(NO_CBLAS)  $(NO_LAPACK) $(NO_LAPACKE) > $(@F)
 
 symbol.S : gensymbol
-	perl ./gensymbol win2kasm noarch dummy $(EXPRECISION) $(NO_CBLAS)  $(NO_LAPACK) > symbol.S
+	perl ./gensymbol win2kasm noarch dummy $(EXPRECISION) $(NO_CBLAS)  $(NO_LAPACK) $(NO_LAPACKE) > symbol.S
 
 test : linktest.c
 	$(CC) $(CFLAGS) -w -o linktest linktest.c ../$(LIBSONAME) -lm && echo OK.
 	rm -f linktest
 
 linktest.c : gensymbol ../Makefile.system ../getarch.c
-	perl ./gensymbol linktest  $(ARCH) $(BU) $(EXPRECISION) $(NO_CBLAS) $(NO_LAPACK) > linktest.c
+	perl ./gensymbol linktest  $(ARCH) $(BU) $(EXPRECISION) $(NO_CBLAS) $(NO_LAPACK) $(NO_LAPACKE) > linktest.c
 
 clean ::
 	@rm -f *.def *.dylib __.SYMDEF*

f_check

@@ -32,11 +32,12 @@ if ($compiler eq "") {
 	      "pgf95", "pgf90", "pgf77",
               "ifort");
     
+OUTER:
     foreach $lists (@lists) {
         foreach $path (@path) {
-            if (-f $path . "/" . $lists) {
+            if (-x $path . "/" . $lists) {
                 $compiler = $lists;
-                break;
+                last OUTER;
             }
         }
     }
@@ -210,6 +211,10 @@ if (!$?) {
 	if ($?) {
 	    $link = `$compiler $openmp -q32 -v ftest2.f 2>&1 && rm -f a.out a.exe`;
 	}
+       #For gfortran MIPS
+	if ($?) {
+	    $link = `$compiler $openmp -mabi=n32 -v ftest2.f 2>&1 && rm -f a.out a.exe`;
+	}
 	$binary = "" if ($?);
     }
     
@@ -218,6 +223,10 @@ if (!$?) {
 	if ($?) {
 	    $link = `$compiler $openmp -q64 -v ftest2.f 2>&1 && rm -f a.out a.exe`;
 	}
+       #For gfortran MIPS
+	if ($?) {
+	    $link = `$compiler $openmp -mabi=64 -v ftest2.f 2>&1 && rm -f a.out a.exe`;
+	}
 	$binary = "" if ($?);
     }
     
@@ -237,6 +246,8 @@ if ($link ne "") {
     $link =~ s/\-rpath\s+/\-rpath\@/g;
 
     @flags = split(/[\s\,\n]/, $link);
+    # remove leading and trailing quotes from each flag.
+    @flags = map {s/^['"]|['"]$//g; $_} @flags;
 
     foreach $flags (@flags) {
 	if (

getarch.c

@@ -1,5 +1,5 @@
 /*****************************************************************************
-Copyright (c) 2011, Lab of Parallel Software and Computational Science,ICSAS
+Copyright (c) 2011,2012 Lab of Parallel Software and Computational Science,ISCAS
 All rights reserved.
 
 Redistribution and use in source and binary forms, with or without
@@ -96,12 +96,16 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 /* #define FORCE_PENRYN		*/
 /* #define FORCE_DUNNINGTON	*/
 /* #define FORCE_NEHALEM	*/
+/* #define FORCE_SANDYBRIDGE	*/
+/* #define FORCE_ATOM		*/
 /* #define FORCE_ATHLON		*/
 /* #define FORCE_OPTERON	*/
 /* #define FORCE_OPTERON_SSE3	*/
 /* #define FORCE_BARCELONA	*/
 /* #define FORCE_SHANGHAI	*/
 /* #define FORCE_ISTANBUL	*/
+/* #define FORCE_BOBCAT		*/
+/* #define FORCE_BULLDOZER	*/
 /* #define FORCE_SSE_GENERIC	*/
 /* #define FORCE_VIAC3		*/
 /* #define FORCE_NANO		*/
@@ -116,12 +120,12 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 /* #define FORCE_PPC440FP2	*/
 /* #define FORCE_CELL		*/
 /* #define FORCE_SICORTEX	*/
-/* #define FORCE_LOONGSON3A      */
-/* #define FORCE_LOONGSON3B      */
+/* #define FORCE_LOONGSON3A	*/
+/* #define FORCE_LOONGSON3B	*/
 /* #define FORCE_ITANIUM2	*/
-/* #define FORCE_GENERIC	*/
 /* #define FORCE_SPARC		*/
 /* #define FORCE_SPARCV7	*/
+/* #define FORCE_GENERIC	*/
 
 #ifdef FORCE_P2
 #define FORCE
@@ -137,32 +141,32 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #define CORENAME  "P5"
 #endif
 
-#ifdef FORCE_COPPERMINE
+#ifdef FORCE_KATMAI
 #define FORCE
 #define FORCE_INTEL
 #define ARCHITECTURE    "X86"
 #define SUBARCHITECTURE "PENTIUM3"
 #define ARCHCONFIG   "-DPENTIUM3 " \
 		     "-DL1_DATA_SIZE=16384 -DL1_DATA_LINESIZE=32 " \
-		     "-DL2_SIZE=262144 -DL2_LINESIZE=32 " \
+		     "-DL2_SIZE=524288 -DL2_LINESIZE=32 " \
 		     "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 " \
 		     "-DHAVE_CMOV -DHAVE_MMX -DHAVE_SSE "
-#define LIBNAME   "coppermine"
-#define CORENAME  "COPPERMINE"
+#define LIBNAME   "katmai"
+#define CORENAME  "KATMAI"
 #endif
 
-#ifdef FORCE_KATMAI
+#ifdef FORCE_COPPERMINE
 #define FORCE
 #define FORCE_INTEL
 #define ARCHITECTURE    "X86"
 #define SUBARCHITECTURE "PENTIUM3"
 #define ARCHCONFIG   "-DPENTIUM3 " \
 		     "-DL1_DATA_SIZE=16384 -DL1_DATA_LINESIZE=32 " \
-		     "-DL2_SIZE=524288 -DL2_LINESIZE=32 " \
+		     "-DL2_SIZE=262144 -DL2_LINESIZE=32 " \
 		     "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 " \
 		     "-DHAVE_CMOV -DHAVE_MMX -DHAVE_SSE "
-#define LIBNAME   "katmai"
-#define CORENAME  "KATMAI"
+#define LIBNAME   "coppermine"
+#define CORENAME  "COPPERMINE"
 #endif
 
 #ifdef FORCE_NORTHWOOD
@@ -278,6 +282,20 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #define CORENAME  "NEHALEM"
 #endif
 
+#ifdef FORCE_SANDYBRIDGE
+#define FORCE
+#define FORCE_INTEL
+#define ARCHITECTURE    "X86"
+#define SUBARCHITECTURE "SANDYBRIDGE"
+#define ARCHCONFIG   "-DSANDYBRIDGE " \
+		     "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=64 " \
+		     "-DL2_SIZE=262144 -DL2_LINESIZE=64 " \
+		     "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 " \
+		     "-DHAVE_CMOV -DHAVE_MMX -DHAVE_SSE -DHAVE_SSE2 -DHAVE_SSE3 -DHAVE_SSSE3 -DHAVE_SSE4_1 -DHAVE_SSE4_2 -DHAVE_AVX"
+#define LIBNAME   "sandybridge"
+#define CORENAME  "SANDYBRIDGE"
+#endif
+
 #ifdef FORCE_ATOM
 #define FORCE
 #define FORCE_INTEL
@@ -342,13 +360,44 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #define ARCHCONFIG   "-DBARCELONA " \
 		     "-DL1_DATA_SIZE=65536 -DL1_DATA_LINESIZE=64 " \
 		     "-DL2_SIZE=524288 -DL2_LINESIZE=64  -DL3_SIZE=2097152 " \
-		     "-DDTB_DEFAULT_ENTRIES=48 -DDTB_SIZE=4096 -DHAVE_3DNOW " \
-		     "-DHAVE_3DNOWEX -DHAVE_MMX -DHAVE_SSE -DHAVE_SSE2 -DHAVE_SSE3 " \
+		     "-DDTB_DEFAULT_ENTRIES=48 -DDTB_SIZE=4096 " \
+		     "-DHAVE_MMX -DHAVE_SSE -DHAVE_SSE2 -DHAVE_SSE3 " \
 		     "-DHAVE_SSE4A -DHAVE_MISALIGNSSE -DHAVE_128BITFPU -DHAVE_FASTMOVU"
 #define LIBNAME   "barcelona"
 #define CORENAME  "BARCELONA"
 #endif
 
+#if defined(FORCE_BOBCAT) 
+#define FORCE
+#define FORCE_INTEL
+#define ARCHITECTURE    "X86"
+#define SUBARCHITECTURE "BOBCAT"
+#define ARCHCONFIG   "-DBOBCAT " \
+		     "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=64 " \
+		     "-DL2_SIZE=524288 -DL2_LINESIZE=64 " \
+		     "-DDTB_DEFAULT_ENTRIES=40 -DDTB_SIZE=4096 " \
+		     "-DHAVE_MMX -DHAVE_SSE -DHAVE_SSE2 -DHAVE_SSE3 -DHAVE_SSSE3 " \
+		     "-DHAVE_SSE4A -DHAVE_MISALIGNSSE -DHAVE_CFLUSH -DHAVE_CMOV"
+#define LIBNAME   "bobcat"
+#define CORENAME  "BOBCAT"
+#endif
+
+#if defined (FORCE_BULLDOZER)
+#define FORCE
+#define FORCE_INTEL
+#define ARCHITECTURE    "X86"
+#define SUBARCHITECTURE "BULLDOZER"
+#define ARCHCONFIG   "-DBULLDOZER " \
+		     "-DL1_DATA_SIZE=49152 -DL1_DATA_LINESIZE=64 " \
+		     "-DL2_SIZE=1024000 -DL2_LINESIZE=64  -DL3_SIZE=16777216 " \
+		     "-DDTB_DEFAULT_ENTRIES=32 -DDTB_SIZE=4096 " \
+		     "-DHAVE_MMX -DHAVE_SSE -DHAVE_SSE2 -DHAVE_SSE3 " \
+		     "-DHAVE_SSE4A -DHAVE_MISALIGNSSE -DHAVE_128BITFPU -DHAVE_FASTMOVU " \
+                     "-DHAVE_AVX -DHAVE_FMA4"
+#define LIBNAME   "bulldozer"
+#define CORENAME  "BULLDOZER"
+#endif
+
 #ifdef FORCE_SSE_GENERIC
 #define FORCE
 #define FORCE_INTEL

getarch_2nd.c

@@ -34,7 +34,7 @@ int main(int argc, char **argv) {
 #ifdef USE64BITINT
 	printf("#define USE64BITINT\n");
 #endif
-	printf("#define GEMM_MULTITHREAD_THRESHOLD\t%ld\n", GEMM_MULTITHREAD_THRESHOLD);
+	printf("#define GEMM_MULTITHREAD_THRESHOLD\t%ld\n", (long int)GEMM_MULTITHREAD_THRESHOLD);
   }
 
   return 0;

interface/Makefile

@@ -318,7 +318,7 @@ CZBLAS3OBJS   = \
 
 ifndef NO_CBLAS
 
-CFLAGS += -I.
+override CFLAGS += -I.
 
 SBLAS1OBJS   += $(CSBLAS1OBJS)
 SBLAS2OBJS   += $(CSBLAS2OBJS)
@@ -400,7 +400,7 @@ all :: libs
 
 ifdef FUNCTION_PROFILE
 $(BLASOBJS) $(BLASOBJS_P) : functable.h
-$(BLASOBJS) $(BLASOBJS_P) : CFLAGS += -DPROFILE_FUNC_NAME=interface_$(*F)
+$(BLASOBJS) $(BLASOBJS_P) : override CFLAGS += -DPROFILE_FUNC_NAME=interface_$(*F)
 
 functable.h : Makefile
 	./create $(FUNCALLFILES) > functable.h
@@ -420,7 +420,7 @@ level3 : $(SBLAS3OBJS) $(DBLAS3OBJS) $(QBLAS3OBJS) $(CBLAS3OBJS) $(ZBLAS3OBJS) $
 	$(AR) $(ARFLAGS) -ru $(TOPDIR)/$(LIBNAME) $^
 
 $(CSBLASOBJS) $(CSBLASOBJS_P) $(CDBLASOBJS) $(CDBLASOBJS_P) $(CQBLASOBJS) $(CQBLASOBJS_P) \
-$(CCBLASOBJS) $(CCBLASOBJS_P) $(CZBLASOBJS) $(CZBLASOBJS_P) $(CXBLASOBJS) $(CXBLASOBJS_P) : CFLAGS += -DCBLAS
+$(CCBLASOBJS) $(CCBLASOBJS_P) $(CZBLASOBJS) $(CZBLASOBJS_P) $(CXBLASOBJS) $(CXBLASOBJS_P) : override CFLAGS += -DCBLAS
 
 srot.$(SUFFIX) srot.$(PSUFFIX) : rot.c
 	$(CC) $(CFLAGS) -c $< -o $(@F)

kernel/Makefile

@@ -6,7 +6,7 @@ TOPDIR	= ..
 include $(TOPDIR)/Makefile.system
 
 ifdef TARGET_CORE
-CFLAGS += -DBUILD_KERNEL -DTABLE_NAME=gotoblas_$(TARGET_CORE)
+override CFLAGS += -DBUILD_KERNEL -DTABLE_NAME=gotoblas_$(TARGET_CORE)
 BUILD_KERNEL = 1
 KDIR =
 TSUFFIX = _$(TARGET_CORE)
@@ -48,7 +48,7 @@ HPLOBJS = \
 
 COMMONOBJS	+= lsame.$(SUFFIX) scabs1.$(SUFFIX) dcabs1.$(SUFFIX)
 
-ifdef DYNAMIC_ARCH
+ifeq ($(DYNAMIC_ARCH), 1)
 SBLASOBJS	+= setparam$(TSUFFIX).$(SUFFIX)
 CCOMMON_OPT	+= -DTS=$(TSUFFIX)
 endif

kernel/generic/zgemm_ncopy_4_sandy.c

@@ -0,0 +1,235 @@
+/*****************************************************************************
+  Copyright (c) 2011,2012 Lab of Parallel Software and Computational Science,ISCAS
+  All rights reserved.
+
+  Redistribution and use in source and binary forms, with or without
+  modification, are permitted provided that the following conditions are
+met:
+
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the ISCAS nor the names of its contributors may 
+be used to endorse or promote products derived from this software 
+without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 
+ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE 
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR 
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER 
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, 
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE 
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+ **********************************************************************************/
+
+#include <stdio.h>
+#include "common.h"
+
+int CNAME(BLASLONG row,BLASLONG col,FLOAT* src,BLASLONG srcdim,FLOAT* dest) 
+{
+   BLASLONG i,j;
+   BLASLONG idx=0;
+   BLASLONG ii;
+   FLOAT *src0,*src1,*src2,*src3,*dest0;
+   for (j=0; j<col/4; j+=1) 
+     {
+        src0 = src;
+        src1 = src0+2*srcdim;
+        src2 = src1+2*srcdim;
+        src3 = src2+2*srcdim;
+        src = src3+2*srcdim;
+        dest0 = dest;
+        ii = (row<<3);
+        dest = dest+ii;
+        for (i=0; i<row/4; i+=1) 
+          {
+             dest0[0] = src0[0];
+             dest0[1] = src0[1];
+             dest0[2] = src1[0];
+             dest0[3] = src1[1];
+             dest0[4] = src2[0];
+             dest0[5] = src2[1];
+             dest0[6] = src3[0];
+             dest0[7] = src3[1];
+             dest0[8] = src0[2];
+             dest0[9] = src0[3];
+             dest0[10] = src1[2];
+             dest0[11] = src1[3];
+             dest0[12] = src2[2];
+             dest0[13] = src2[3];
+             dest0[14] = src3[2];
+             dest0[15] = src3[3];
+             dest0[16] = src0[4];
+             dest0[17] = src0[5];
+             dest0[18] = src1[4];
+             dest0[19] = src1[5];
+             dest0[20] = src2[4];
+             dest0[21] = src2[5];
+             dest0[22] = src3[4];
+             dest0[23] = src3[5];
+             dest0[24] = src0[6];
+             dest0[25] = src0[7];
+             dest0[26] = src1[6];
+             dest0[27] = src1[7];
+             dest0[28] = src2[6];
+             dest0[29] = src2[7];
+             dest0[30] = src3[6];
+             dest0[31] = src3[7];
+             src0 = src0+8;
+             src1 = src1+8;
+             src2 = src2+8;
+             src3 = src3+8;
+             ii = (4<<3);
+             dest0 = dest0+ii;
+          }
+        if (row&2)  
+          {
+             dest0[0] = src0[0];
+             dest0[1] = src0[1];
+             dest0[2] = src1[0];
+             dest0[3] = src1[1];
+             dest0[4] = src2[0];
+             dest0[5] = src2[1];
+             dest0[6] = src3[0];
+             dest0[7] = src3[1];
+             dest0[8] = src0[2];
+             dest0[9] = src0[3];
+             dest0[10] = src1[2];
+             dest0[11] = src1[3];
+             dest0[12] = src2[2];
+             dest0[13] = src2[3];
+             dest0[14] = src3[2];
+             dest0[15] = src3[3];
+             src0 = src0+4;
+             src1 = src1+4;
+             src2 = src2+4;
+             src3 = src3+4;
+             ii = (2<<3);
+             dest0 = dest0+ii;
+          }
+        if (row&1)  
+          {
+             dest0[0] = src0[0];
+             dest0[1] = src0[1];
+             dest0[2] = src1[0];
+             dest0[3] = src1[1];
+             dest0[4] = src2[0];
+             dest0[5] = src2[1];
+             dest0[6] = src3[0];
+             dest0[7] = src3[1];
+             src0 = src0+2;
+             src1 = src1+2;
+             src2 = src2+2;
+             src3 = src3+2;
+             ii = (1<<3);
+             dest0 = dest0+ii;
+          }
+     }
+   if (col&2)  
+     {
+        src0 = src;
+        src1 = src0+2*srcdim;
+        src = src1+2*srcdim;
+        dest0 = dest;
+        ii = (row<<2);
+        dest = dest+ii;
+        for (i=0; i<row/4; i+=1) 
+          {
+             dest0[0] = src0[0];
+             dest0[1] = src0[1];
+             dest0[2] = src1[0];
+             dest0[3] = src1[1];
+             dest0[4] = src0[2];
+             dest0[5] = src0[3];
+             dest0[6] = src1[2];
+             dest0[7] = src1[3];
+             dest0[8] = src0[4];
+             dest0[9] = src0[5];
+             dest0[10] = src1[4];
+             dest0[11] = src1[5];
+             dest0[12] = src0[6];
+             dest0[13] = src0[7];
+             dest0[14] = src1[6];
+             dest0[15] = src1[7];
+             src0 = src0+8;
+             src1 = src1+8;
+             ii = (4<<2);
+             dest0 = dest0+ii;
+          }
+        if (row&2)  
+          {
+             dest0[0] = src0[0];
+             dest0[1] = src0[1];
+             dest0[2] = src1[0];
+             dest0[3] = src1[1];
+             dest0[4] = src0[2];
+             dest0[5] = src0[3];
+             dest0[6] = src1[2];
+             dest0[7] = src1[3];
+             src0 = src0+4;
+             src1 = src1+4;
+             ii = (2<<2);
+             dest0 = dest0+ii;
+          }
+        if (row&1)  
+          {
+             dest0[0] = src0[0];
+             dest0[1] = src0[1];
+             dest0[2] = src1[0];
+             dest0[3] = src1[1];
+             src0 = src0+2;
+             src1 = src1+2;
+             ii = (1<<2);
+             dest0 = dest0+ii;
+          }
+     }
+   if (col&1)  
+     {
+        src0 = src;
+        src = src0+2*srcdim;
+        dest0 = dest;
+        ii = (row<<1);
+        dest = dest+ii;
+        for (i=0; i<row/4; i+=1) 
+          {
+             dest0[0] = src0[0];
+             dest0[1] = src0[1];
+             dest0[2] = src0[2];
+             dest0[3] = src0[3];
+             dest0[4] = src0[4];
+             dest0[5] = src0[5];
+             dest0[6] = src0[6];
+             dest0[7] = src0[7];
+             src0 = src0+8;
+             ii = (4<<1);
+             dest0 = dest0+ii;
+          }
+        if (row&2)  
+          {
+             dest0[0] = src0[0];
+             dest0[1] = src0[1];
+             dest0[2] = src0[2];
+             dest0[3] = src0[3];
+             src0 = src0+4;
+             ii = (2<<1);
+             dest0 = dest0+ii;
+          }
+        if (row&1)  
+          {
+             dest0[0] = src0[0];
+             dest0[1] = src0[1];
+             src0 = src0+2;
+             ii = (1<<1);
+             dest0 = dest0+ii;
+          }
+     }
+   return 0;
+}

kernel/generic/zgemm_ncopy_8_sandy.c

@@ -0,0 +1,401 @@
+/*****************************************************************************
+  Copyright (c) 2011,2012 Lab of Parallel Software and Computational Science,ISCAS
+  All rights reserved.
+
+  Redistribution and use in source and binary forms, with or without
+  modification, are permitted provided that the following conditions are
+met:
+
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the ISCAS nor the names of its contributors may 
+be used to endorse or promote products derived from this software 
+without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 
+ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE 
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR 
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER 
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, 
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE 
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+ **********************************************************************************/
+
+#include <stdio.h>
+#include "common.h"
+
+int CNAME(BLASLONG row,BLASLONG col,FLOAT* src,BLASLONG srcdim,FLOAT* dest) 
+{
+   BLASLONG i,j;
+   BLASLONG idx=0;
+   BLASLONG ii;
+   FLOAT *src0,*src1,*src2,*src3,*src4,*src5,*src6,*src7,*dest0;
+   for (j=0; j<col/8; j+=1) 
+     {
+        src0 = src;
+        src1 = src0+2*srcdim;
+        src2 = src1+2*srcdim;
+        src3 = src2+2*srcdim;
+        src4 = src3+2*srcdim;
+        src5 = src4+2*srcdim;
+        src6 = src5+2*srcdim;
+        src7 = src6+2*srcdim;
+        src = src7+2*srcdim;
+        dest0 = dest;
+        ii = (row<<4);
+        dest = dest+ii;
+        for (i=0; i<row/4; i+=1) 
+          {
+             dest0[0] = src0[0];
+             dest0[1] = src0[1];
+             dest0[2] = src1[0];
+             dest0[3] = src1[1];
+             dest0[4] = src2[0];
+             dest0[5] = src2[1];
+             dest0[6] = src3[0];
+             dest0[7] = src3[1];
+             dest0[8] = src4[0];
+             dest0[9] = src4[1];
+             dest0[10] = src5[0];
+             dest0[11] = src5[1];
+             dest0[12] = src6[0];
+             dest0[13] = src6[1];
+             dest0[14] = src7[0];
+             dest0[15] = src7[1];
+             dest0[16] = src0[2];
+             dest0[17] = src0[3];
+             dest0[18] = src1[2];
+             dest0[19] = src1[3];
+             dest0[20] = src2[2];
+             dest0[21] = src2[3];
+             dest0[22] = src3[2];
+             dest0[23] = src3[3];
+             dest0[24] = src4[2];
+             dest0[25] = src4[3];
+             dest0[26] = src5[2];
+             dest0[27] = src5[3];
+             dest0[28] = src6[2];
+             dest0[29] = src6[3];
+             dest0[30] = src7[2];
+             dest0[31] = src7[3];
+             dest0[32] = src0[4];
+             dest0[33] = src0[5];
+             dest0[34] = src1[4];
+             dest0[35] = src1[5];
+             dest0[36] = src2[4];
+             dest0[37] = src2[5];
+             dest0[38] = src3[4];
+             dest0[39] = src3[5];
+             dest0[40] = src4[4];
+             dest0[41] = src4[5];
+             dest0[42] = src5[4];
+             dest0[43] = src5[5];
+             dest0[44] = src6[4];
+             dest0[45] = src6[5];
+             dest0[46] = src7[4];
+             dest0[47] = src7[5];
+             dest0[48] = src0[6];
+             dest0[49] = src0[7];
+             dest0[50] = src1[6];
+             dest0[51] = src1[7];
+             dest0[52] = src2[6];
+             dest0[53] = src2[7];
+             dest0[54] = src3[6];
+             dest0[55] = src3[7];
+             dest0[56] = src4[6];
+             dest0[57] = src4[7];
+             dest0[58] = src5[6];
+             dest0[59] = src5[7];
+             dest0[60] = src6[6];
+             dest0[61] = src6[7];
+             dest0[62] = src7[6];
+             dest0[63] = src7[7];
+             src0 = src0+8;
+             src1 = src1+8;
+             src2 = src2+8;
+             src3 = src3+8;
+             src4 = src4+8;
+             src5 = src5+8;
+             src6 = src6+8;
+             src7 = src7+8;
+             ii = (4<<4);
+             dest0 = dest0+ii;
+          }
+        if (row&2)  
+          {
+             dest0[0] = src0[0];
+             dest0[1] = src0[1];
+             dest0[2] = src1[0];
+             dest0[3] = src1[1];
+             dest0[4] = src2[0];
+             dest0[5] = src2[1];
+             dest0[6] = src3[0];
+             dest0[7] = src3[1];
+             dest0[8] = src4[0];
+             dest0[9] = src4[1];
+             dest0[10] = src5[0];
+             dest0[11] = src5[1];
+             dest0[12] = src6[0];
+             dest0[13] = src6[1];
+             dest0[14] = src7[0];
+             dest0[15] = src7[1];
+             dest0[16] = src0[2];
+             dest0[17] = src0[3];
+             dest0[18] = src1[2];
+             dest0[19] = src1[3];
+             dest0[20] = src2[2];
+             dest0[21] = src2[3];
+             dest0[22] = src3[2];
+             dest0[23] = src3[3];
+             dest0[24] = src4[2];
+             dest0[25] = src4[3];
+             dest0[26] = src5[2];
+             dest0[27] = src5[3];
+             dest0[28] = src6[2];
+             dest0[29] = src6[3];
+             dest0[30] = src7[2];
+             dest0[31] = src7[3];
+             src0 = src0+4;
+             src1 = src1+4;
+             src2 = src2+4;
+             src3 = src3+4;
+             src4 = src4+4;
+             src5 = src5+4;
+             src6 = src6+4;
+             src7 = src7+4;
+             ii = (2<<4);
+             dest0 = dest0+ii;
+          }
+        if (row&1)  
+          {
+             dest0[0] = src0[0];
+             dest0[1] = src0[1];
+             dest0[2] = src1[0];
+             dest0[3] = src1[1];
+             dest0[4] = src2[0];
+             dest0[5] = src2[1];
+             dest0[6] = src3[0];
+             dest0[7] = src3[1];
+             dest0[8] = src4[0];
+             dest0[9] = src4[1];
+             dest0[10] = src5[0];
+             dest0[11] = src5[1];
+             dest0[12] = src6[0];
+             dest0[13] = src6[1];
+             dest0[14] = src7[0];
+             dest0[15] = src7[1];
+             src0 = src0+2;
+             src1 = src1+2;
+             src2 = src2+2;
+             src3 = src3+2;
+             src4 = src4+2;
+             src5 = src5+2;
+             src6 = src6+2;
+             src7 = src7+2;
+             ii = (1<<4);
+             dest0 = dest0+ii;
+          }
+     }
+   if (col&4)  
+     {
+        src0 = src;
+        src1 = src0+2*srcdim;
+        src2 = src1+2*srcdim;
+        src3 = src2+2*srcdim;
+        src = src3+2*srcdim;
+        dest0 = dest;
+        ii = (row<<3);
+        dest = dest+ii;
+        for (i=0; i<row/4; i+=1) 
+          {
+             dest0[0] = src0[0];
+             dest0[1] = src0[1];
+             dest0[2] = src1[0];
+             dest0[3] = src1[1];
+             dest0[4] = src2[0];
+             dest0[5] = src2[1];
+             dest0[6] = src3[0];
+             dest0[7] = src3[1];
+             dest0[8] = src0[2];
+             dest0[9] = src0[3];
+             dest0[10] = src1[2];
+             dest0[11] = src1[3];
+             dest0[12] = src2[2];
+             dest0[13] = src2[3];
+             dest0[14] = src3[2];
+             dest0[15] = src3[3];
+             dest0[16] = src0[4];
+             dest0[17] = src0[5];
+             dest0[18] = src1[4];
+             dest0[19] = src1[5];
+             dest0[20] = src2[4];
+             dest0[21] = src2[5];
+             dest0[22] = src3[4];
+             dest0[23] = src3[5];
+             dest0[24] = src0[6];
+             dest0[25] = src0[7];
+             dest0[26] = src1[6];
+             dest0[27] = src1[7];
+             dest0[28] = src2[6];
+             dest0[29] = src2[7];
+             dest0[30] = src3[6];
+             dest0[31] = src3[7];
+             src0 = src0+8;
+             src1 = src1+8;
+             src2 = src2+8;
+             src3 = src3+8;
+             ii = (4<<3);
+             dest0 = dest0+ii;
+          }
+        if (row&2)  
+          {
+             dest0[0] = src0[0];
+             dest0[1] = src0[1];
+             dest0[2] = src1[0];
+             dest0[3] = src1[1];
+             dest0[4] = src2[0];
+             dest0[5] = src2[1];
+             dest0[6] = src3[0];
+             dest0[7] = src3[1];
+             dest0[8] = src0[2];
+             dest0[9] = src0[3];
+             dest0[10] = src1[2];
+             dest0[11] = src1[3];
+             dest0[12] = src2[2];
+             dest0[13] = src2[3];
+             dest0[14] = src3[2];
+             dest0[15] = src3[3];
+             src0 = src0+4;
+             src1 = src1+4;
+             src2 = src2+4;
+             src3 = src3+4;
+             ii = (2<<3);
+             dest0 = dest0+ii;
+          }
+        if (row&1)  
+          {
+             dest0[0] = src0[0];
+             dest0[1] = src0[1];
+             dest0[2] = src1[0];
+             dest0[3] = src1[1];
+             dest0[4] = src2[0];
+             dest0[5] = src2[1];
+             dest0[6] = src3[0];
+             dest0[7] = src3[1];
+             src0 = src0+2;
+             src1 = src1+2;
+             src2 = src2+2;
+             src3 = src3+2;
+             ii = (1<<3);
+             dest0 = dest0+ii;
+          }
+     }
+   if (col&2)  
+     {
+        src0 = src;
+        src1 = src0+2*srcdim;
+        src = src1+2*srcdim;
+        dest0 = dest;
+        ii = (row<<2);
+        dest = dest+ii;
+        for (i=0; i<row/4; i+=1) 
+          {
+             dest0[0] = src0[0];
+             dest0[1] = src0[1];
+             dest0[2] = src1[0];
+             dest0[3] = src1[1];
+             dest0[4] = src0[2];
+             dest0[5] = src0[3];
+             dest0[6] = src1[2];
+             dest0[7] = src1[3];
+             dest0[8] = src0[4];
+             dest0[9] = src0[5];
+             dest0[10] = src1[4];
+             dest0[11] = src1[5];
+             dest0[12] = src0[6];
+             dest0[13] = src0[7];
+             dest0[14] = src1[6];
+             dest0[15] = src1[7];
+             src0 = src0+8;
+             src1 = src1+8;
+             ii = (4<<2);
+             dest0 = dest0+ii;
+          }
+        if (row&2)  
+          {
+             dest0[0] = src0[0];
+             dest0[1] = src0[1];
+             dest0[2] = src1[0];
+             dest0[3] = src1[1];
+             dest0[4] = src0[2];
+             dest0[5] = src0[3];
+             dest0[6] = src1[2];
+             dest0[7] = src1[3];
+             src0 = src0+4;
+             src1 = src1+4;
+             ii = (2<<2);
+             dest0 = dest0+ii;
+          }
+        if (row&1)  
+          {
+             dest0[0] = src0[0];
+             dest0[1] = src0[1];
+             dest0[2] = src1[0];
+             dest0[3] = src1[1];
+             src0 = src0+2;
+             src1 = src1+2;
+             ii = (1<<2);
+             dest0 = dest0+ii;
+          }
+     }
+   if (col&1)  
+     {
+        src0 = src;
+        src = src0+2*srcdim;
+        dest0 = dest;
+        ii = (row<<1);
+        dest = dest+ii;
+        for (i=0; i<row/4; i+=1) 
+          {
+             dest0[0] = src0[0];
+             dest0[1] = src0[1];
+             dest0[2] = src0[2];
+             dest0[3] = src0[3];
+             dest0[4] = src0[4];
+             dest0[5] = src0[5];
+             dest0[6] = src0[6];
+             dest0[7] = src0[7];
+             src0 = src0+8;
+             ii = (4<<1);
+             dest0 = dest0+ii;
+          }
+        if (row&2)  
+          {
+             dest0[0] = src0[0];
+             dest0[1] = src0[1];
+             dest0[2] = src0[2];
+             dest0[3] = src0[3];
+             src0 = src0+4;
+             ii = (2<<1);
+             dest0 = dest0+ii;
+          }
+        if (row&1)  
+          {
+             dest0[0] = src0[0];
+             dest0[1] = src0[1];
+             src0 = src0+2;
+             ii = (1<<1);
+             dest0 = dest0+ii;
+          }
+     }
+   return 0;
+}

kernel/generic/zgemm_tcopy_4_sandy.c

@@ -0,0 +1,237 @@
+/*****************************************************************************
+  Copyright (c) 2011,2012 Lab of Parallel Software and Computational Science,ISCAS
+  All rights reserved.
+
+  Redistribution and use in source and binary forms, with or without
+  modification, are permitted provided that the following conditions are
+met:
+
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the ISCAS nor the names of its contributors may 
+be used to endorse or promote products derived from this software 
+without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 
+ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE 
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR 
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER 
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, 
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE 
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+ **********************************************************************************/
+
+#include <stdio.h>
+#include "common.h"
+
+int CNAME(BLASLONG row,BLASLONG col,FLOAT* src,BLASLONG srcdim,FLOAT* dest) 
+{
+   BLASLONG i,j;
+   BLASLONG idx=0;
+   BLASLONG ii;
+   FLOAT *src0,*src1,*src2,*src3,*dest0;
+   FLOAT *dest1,*dest2;
+   ii = col&-4;
+   ii = ii*(2*row);
+   dest2 = dest+ii;
+   ii = col&-2;
+   ii = ii*(2*row);
+   dest1 = dest+ii;
+   for (j=0; j<row/4; j+=1) 
+     {
+        src0 = src;
+        src1 = src0+2*srcdim;
+        src2 = src1+2*srcdim;
+        src3 = src2+2*srcdim;
+        src = src3+2*srcdim;
+        dest0 = dest;
+        ii = (4<<3);
+        dest = dest+ii;
+        for (i=0; i<col/4; i+=1) 
+          {
+             dest0[0] = src0[0];
+             dest0[1] = src0[1];
+             dest0[2] = src0[2];
+             dest0[3] = src0[3];
+             dest0[4] = src0[4];
+             dest0[5] = src0[5];
+             dest0[6] = src0[6];
+             dest0[7] = src0[7];
+             dest0[8] = src1[0];
+             dest0[9] = src1[1];
+             dest0[10] = src1[2];
+             dest0[11] = src1[3];
+             dest0[12] = src1[4];
+             dest0[13] = src1[5];
+             dest0[14] = src1[6];
+             dest0[15] = src1[7];
+             dest0[16] = src2[0];
+             dest0[17] = src2[1];
+             dest0[18] = src2[2];
+             dest0[19] = src2[3];
+             dest0[20] = src2[4];
+             dest0[21] = src2[5];
+             dest0[22] = src2[6];
+             dest0[23] = src2[7];
+             dest0[24] = src3[0];
+             dest0[25] = src3[1];
+             dest0[26] = src3[2];
+             dest0[27] = src3[3];
+             dest0[28] = src3[4];
+             dest0[29] = src3[5];
+             dest0[30] = src3[6];
+             dest0[31] = src3[7];
+             src0 = src0+8;
+             src1 = src1+8;
+             src2 = src2+8;
+             src3 = src3+8;
+             ii = (row<<3);
+             dest0 = dest0+ii;
+          }
+        if (col&2)  
+          {
+             dest2[0] = src0[0];
+             dest2[1] = src0[1];
+             dest2[2] = src0[2];
+             dest2[3] = src0[3];
+             dest2[4] = src1[0];
+             dest2[5] = src1[1];
+             dest2[6] = src1[2];
+             dest2[7] = src1[3];
+             dest2[8] = src2[0];
+             dest2[9] = src2[1];
+             dest2[10] = src2[2];
+             dest2[11] = src2[3];
+             dest2[12] = src3[0];
+             dest2[13] = src3[1];
+             dest2[14] = src3[2];
+             dest2[15] = src3[3];
+             src0 = src0+4;
+             src1 = src1+4;
+             src2 = src2+4;
+             src3 = src3+4;
+             dest2 = dest2+16;
+          }
+        if (col&1)  
+          {
+             dest1[0] = src0[0];
+             dest1[1] = src0[1];
+             dest1[2] = src1[0];
+             dest1[3] = src1[1];
+             dest1[4] = src2[0];
+             dest1[5] = src2[1];
+             dest1[6] = src3[0];
+             dest1[7] = src3[1];
+             src0 = src0+2;
+             src1 = src1+2;
+             src2 = src2+2;
+             src3 = src3+2;
+             dest1 = dest1+8;
+          }
+     }
+   if (row&2)  
+     {
+        src0 = src;
+        src1 = src0+2*srcdim;
+        src = src1+2*srcdim;
+        dest0 = dest;
+        ii = (2<<3);
+        dest = dest+ii;
+        for (i=0; i<col/4; i+=1) 
+          {
+             dest0[0] = src0[0];
+             dest0[1] = src0[1];
+             dest0[2] = src0[2];
+             dest0[3] = src0[3];
+             dest0[4] = src0[4];
+             dest0[5] = src0[5];
+             dest0[6] = src0[6];
+             dest0[7] = src0[7];
+             dest0[8] = src1[0];
+             dest0[9] = src1[1];
+             dest0[10] = src1[2];
+             dest0[11] = src1[3];
+             dest0[12] = src1[4];
+             dest0[13] = src1[5];
+             dest0[14] = src1[6];
+             dest0[15] = src1[7];
+             src0 = src0+8;
+             src1 = src1+8;
+             ii = (row<<3);
+             dest0 = dest0+ii;
+          }
+        if (col&2)  
+          {
+             dest2[0] = src0[0];
+             dest2[1] = src0[1];
+             dest2[2] = src0[2];
+             dest2[3] = src0[3];
+             dest2[4] = src1[0];
+             dest2[5] = src1[1];
+             dest2[6] = src1[2];
+             dest2[7] = src1[3];
+             src0 = src0+4;
+             src1 = src1+4;
+             dest2 = dest2+8;
+          }
+        if (col&1)  
+          {
+             dest1[0] = src0[0];
+             dest1[1] = src0[1];
+             dest1[2] = src1[0];
+             dest1[3] = src1[1];
+             src0 = src0+2;
+             src1 = src1+2;
+             dest1 = dest1+4;
+          }
+     }
+   if (row&1)  
+     {
+        src0 = src;
+        src = src0+2*srcdim;
+        dest0 = dest;
+        ii = (1<<3);
+        dest = dest+ii;
+        for (i=0; i<col/4; i+=1) 
+          {
+             dest0[0] = src0[0];
+             dest0[1] = src0[1];
+             dest0[2] = src0[2];
+             dest0[3] = src0[3];
+             dest0[4] = src0[4];
+             dest0[5] = src0[5];
+             dest0[6] = src0[6];
+             dest0[7] = src0[7];
+             src0 = src0+8;
+             ii = (row<<3);
+             dest0 = dest0+ii;
+          }
+        if (col&2)  
+          {
+             dest2[0] = src0[0];
+             dest2[1] = src0[1];
+             dest2[2] = src0[2];
+             dest2[3] = src0[3];
+             src0 = src0+4;
+             dest2 = dest2+4;
+          }
+        if (col&1)  
+          {
+             dest1[0] = src0[0];
+             dest1[1] = src0[1];
+             src0 = src0+2;
+             dest1 = dest1+2;
+          }
+     }
+
+   return 0;
+}

kernel/generic/zgemm_tcopy_8_sandy.c

@@ -0,0 +1,370 @@
+/*****************************************************************************
+  Copyright (c) 2011,2012 Lab of Parallel Software and Computational Science,ISCAS
+  All rights reserved.
+
+  Redistribution and use in source and binary forms, with or without
+  modification, are permitted provided that the following conditions are
+met:
+
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the ISCAS nor the names of its contributors may 
+be used to endorse or promote products derived from this software 
+without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 
+ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE 
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR 
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER 
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, 
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE 
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+ **********************************************************************************/
+
+#include <stdio.h>
+#include "common.h"
+
+int CNAME(BLASLONG row,BLASLONG col,FLOAT* src,BLASLONG srcdim,FLOAT* dest) 
+{
+   BLASLONG i,j;
+   BLASLONG idx=0;
+   BLASLONG ii;
+   FLOAT *src0,*src1,*src2,*src3,*dest0;
+   FLOAT *dest1,*dest2,*dest4;
+   ii = col&-8;
+   ii = ii*(2*row);
+   dest4 = dest+ii;
+   ii = col&-4;
+   ii = ii*(2*row);
+   dest2 = dest+ii;
+   ii = col&-2;
+   ii = ii*(2*row);
+   dest1 = dest+ii;
+   for (j=0; j<row/4; j+=1) 
+     {
+        src0 = src;
+        src1 = src0+2*srcdim;
+        src2 = src1+2*srcdim;
+        src3 = src2+2*srcdim;
+        src = src3+2*srcdim;
+        dest0 = dest;
+        ii = (4<<4);
+        dest = dest+ii;
+        for (i=0; i<col/8; i+=1) 
+          {
+             dest0[0] = src0[0];
+             dest0[1] = src0[1];
+             dest0[2] = src0[2];
+             dest0[3] = src0[3];
+             dest0[4] = src0[4];
+             dest0[5] = src0[5];
+             dest0[6] = src0[6];
+             dest0[7] = src0[7];
+             dest0[8] = src0[8];
+             dest0[9] = src0[9];
+             dest0[10] = src0[10];
+             dest0[11] = src0[11];
+             dest0[12] = src0[12];
+             dest0[13] = src0[13];
+             dest0[14] = src0[14];
+             dest0[15] = src0[15];
+             dest0[16] = src1[0];
+             dest0[17] = src1[1];
+             dest0[18] = src1[2];
+             dest0[19] = src1[3];
+             dest0[20] = src1[4];
+             dest0[21] = src1[5];
+             dest0[22] = src1[6];
+             dest0[23] = src1[7];
+             dest0[24] = src1[8];
+             dest0[25] = src1[9];
+             dest0[26] = src1[10];
+             dest0[27] = src1[11];
+             dest0[28] = src1[12];
+             dest0[29] = src1[13];
+             dest0[30] = src1[14];
+             dest0[31] = src1[15];
+             dest0[32] = src2[0];
+             dest0[33] = src2[1];
+             dest0[34] = src2[2];
+             dest0[35] = src2[3];
+             dest0[36] = src2[4];
+             dest0[37] = src2[5];
+             dest0[38] = src2[6];
+             dest0[39] = src2[7];
+             dest0[40] = src2[8];
+             dest0[41] = src2[9];
+             dest0[42] = src2[10];
+             dest0[43] = src2[11];
+             dest0[44] = src2[12];
+             dest0[45] = src2[13];
+             dest0[46] = src2[14];
+             dest0[47] = src2[15];
+             dest0[48] = src3[0];
+             dest0[49] = src3[1];
+             dest0[50] = src3[2];
+             dest0[51] = src3[3];
+             dest0[52] = src3[4];
+             dest0[53] = src3[5];
+             dest0[54] = src3[6];
+             dest0[55] = src3[7];
+             dest0[56] = src3[8];
+             dest0[57] = src3[9];
+             dest0[58] = src3[10];
+             dest0[59] = src3[11];
+             dest0[60] = src3[12];
+             dest0[61] = src3[13];
+             dest0[62] = src3[14];
+             dest0[63] = src3[15];
+             src0 = src0+16;
+             src1 = src1+16;
+             src2 = src2+16;
+             src3 = src3+16;
+             ii = (row<<4);
+             dest0 = dest0+ii;
+          }
+        if (col&4)  
+          {
+             dest4[0] = src0[0];
+             dest4[1] = src0[1];
+             dest4[2] = src0[2];
+             dest4[3] = src0[3];
+             dest4[4] = src0[4];
+             dest4[5] = src0[5];
+             dest4[6] = src0[6];
+             dest4[7] = src0[7];
+             dest4[8] = src1[0];
+             dest4[9] = src1[1];
+             dest4[10] = src1[2];
+             dest4[11] = src1[3];
+             dest4[12] = src1[4];
+             dest4[13] = src1[5];
+             dest4[14] = src1[6];
+             dest4[15] = src1[7];
+             dest4[16] = src2[0];
+             dest4[17] = src2[1];
+             dest4[18] = src2[2];
+             dest4[19] = src2[3];
+             dest4[20] = src2[4];
+             dest4[21] = src2[5];
+             dest4[22] = src2[6];
+             dest4[23] = src2[7];
+             dest4[24] = src3[0];
+             dest4[25] = src3[1];
+             dest4[26] = src3[2];
+             dest4[27] = src3[3];
+             dest4[28] = src3[4];
+             dest4[29] = src3[5];
+             dest4[30] = src3[6];
+             dest4[31] = src3[7];
+             src0 = src0+8;
+             src1 = src1+8;
+             src2 = src2+8;
+             src3 = src3+8;
+             dest4 = dest4+32;
+          }
+        if (col&2)  
+          {
+             dest2[0] = src0[0];
+             dest2[1] = src0[1];
+             dest2[2] = src0[2];
+             dest2[3] = src0[3];
+             dest2[4] = src1[0];
+             dest2[5] = src1[1];
+             dest2[6] = src1[2];
+             dest2[7] = src1[3];
+             dest2[8] = src2[0];
+             dest2[9] = src2[1];
+             dest2[10] = src2[2];
+             dest2[11] = src2[3];
+             dest2[12] = src3[0];
+             dest2[13] = src3[1];
+             dest2[14] = src3[2];
+             dest2[15] = src3[3];
+             src0 = src0+4;
+             src1 = src1+4;
+             src2 = src2+4;
+             src3 = src3+4;
+             dest2 = dest2+16;
+          }
+        if (col&1)  
+          {
+             dest1[0] = src0[0];
+             dest1[1] = src0[1];
+             dest1[2] = src1[0];
+             dest1[3] = src1[1];
+             dest1[4] = src2[0];
+             dest1[5] = src2[1];
+             dest1[6] = src3[0];
+             dest1[7] = src3[1];
+             src0 = src0+2;
+             src1 = src1+2;
+             src2 = src2+2;
+             src3 = src3+2;
+             dest1 = dest1+8;
+          }
+     }
+   if (row&2)  
+     {
+        src0 = src;
+        src1 = src0+2*srcdim;
+        src = src1+2*srcdim;
+        dest0 = dest;
+        ii = (2<<4);
+        dest = dest+ii;
+        for (i=0; i<col/8; i+=1) 
+          {
+             dest0[0] = src0[0];
+             dest0[1] = src0[1];
+             dest0[2] = src0[2];
+             dest0[3] = src0[3];
+             dest0[4] = src0[4];
+             dest0[5] = src0[5];
+             dest0[6] = src0[6];
+             dest0[7] = src0[7];
+             dest0[8] = src0[8];
+             dest0[9] = src0[9];
+             dest0[10] = src0[10];
+             dest0[11] = src0[11];
+             dest0[12] = src0[12];
+             dest0[13] = src0[13];
+             dest0[14] = src0[14];
+             dest0[15] = src0[15];
+             dest0[16] = src1[0];
+             dest0[17] = src1[1];
+             dest0[18] = src1[2];
+             dest0[19] = src1[3];
+             dest0[20] = src1[4];
+             dest0[21] = src1[5];
+             dest0[22] = src1[6];
+             dest0[23] = src1[7];
+             dest0[24] = src1[8];
+             dest0[25] = src1[9];
+             dest0[26] = src1[10];
+             dest0[27] = src1[11];
+             dest0[28] = src1[12];
+             dest0[29] = src1[13];
+             dest0[30] = src1[14];
+             dest0[31] = src1[15];
+             src0 = src0+16;
+             src1 = src1+16;
+             ii = (row<<4);
+             dest0 = dest0+ii;
+          }
+        if (col&4)  
+          {
+             dest4[0] = src0[0];
+             dest4[1] = src0[1];
+             dest4[2] = src0[2];
+             dest4[3] = src0[3];
+             dest4[4] = src0[4];
+             dest4[5] = src0[5];
+             dest4[6] = src0[6];
+             dest4[7] = src0[7];
+             dest4[8] = src1[0];
+             dest4[9] = src1[1];
+             dest4[10] = src1[2];
+             dest4[11] = src1[3];
+             dest4[12] = src1[4];
+             dest4[13] = src1[5];
+             dest4[14] = src1[6];
+             dest4[15] = src1[7];
+             src0 = src0+8;
+             src1 = src1+8;
+             dest4 = dest4+16;
+          }
+        if (col&2)  
+          {
+             dest2[0] = src0[0];
+             dest2[1] = src0[1];
+             dest2[2] = src0[2];
+             dest2[3] = src0[3];
+             dest2[4] = src1[0];
+             dest2[5] = src1[1];
+             dest2[6] = src1[2];
+             dest2[7] = src1[3];
+             src0 = src0+4;
+             src1 = src1+4;
+             dest2 = dest2+8;
+          }
+        if (col&1)  
+          {
+             dest1[0] = src0[0];
+             dest1[1] = src0[1];
+             dest1[2] = src1[0];
+             dest1[3] = src1[1];
+             src0 = src0+2;
+             src1 = src1+2;
+             dest1 = dest1+4;
+          }
+     }
+   if (row&1)  
+     {
+        src0 = src;
+        src = src0+2*srcdim;
+        dest0 = dest;
+        ii = (1<<4);
+        dest = dest+ii;
+        for (i=0; i<col/8; i+=1) 
+          {
+             dest0[0] = src0[0];
+             dest0[1] = src0[1];
+             dest0[2] = src0[2];
+             dest0[3] = src0[3];
+             dest0[4] = src0[4];
+             dest0[5] = src0[5];
+             dest0[6] = src0[6];
+             dest0[7] = src0[7];
+             dest0[8] = src0[8];
+             dest0[9] = src0[9];
+             dest0[10] = src0[10];
+             dest0[11] = src0[11];
+             dest0[12] = src0[12];
+             dest0[13] = src0[13];
+             dest0[14] = src0[14];
+             dest0[15] = src0[15];
+             src0 = src0+16;
+             ii = (row<<4);
+             dest0 = dest0+ii;
+          }
+        if (col&4)  
+          {
+             dest4[0] = src0[0];
+             dest4[1] = src0[1];
+             dest4[2] = src0[2];
+             dest4[3] = src0[3];
+             dest4[4] = src0[4];
+             dest4[5] = src0[5];
+             dest4[6] = src0[6];
+             dest4[7] = src0[7];
+             src0 = src0+8;
+             dest4 = dest4+8;
+          }
+        if (col&2)  
+          {
+             dest2[0] = src0[0];
+             dest2[1] = src0[1];
+             dest2[2] = src0[2];
+             dest2[3] = src0[3];
+             src0 = src0+4;
+             dest2 = dest2+4;
+          }
+        if (col&1)  
+          {
+             dest1[0] = src0[0];
+             dest1[1] = src0[1];
+             src0 = src0+2;
+             dest1 = dest1+2;
+          }
+     }
+   return 0;
+}

kernel/setparam-ref.c

@@ -634,10 +634,10 @@ static void init_parameter(void) {
   TABLE_NAME.xgemm_q = XGEMM_DEFAULT_Q;
 #endif
 
-#if defined(CORE_KATMAI)  || defined(CORE_COPPERMINE) || defined(CORE_BANIAS) || defined(CORE_YONAH)
+#if defined(CORE_KATMAI)  || defined(CORE_COPPERMINE) || defined(CORE_BANIAS) || defined(CORE_YONAH) || defined(CORE_ATHLON)
 
 #ifdef DEBUG
-  fprintf(stderr, "Katmai, Coppermine, Banias\n");
+  fprintf(stderr, "Katmai, Coppermine, Banias, Athlon\n");
 #endif
 
   TABLE_NAME.sgemm_p =  64 * (l2 >> 7);
@@ -746,6 +746,22 @@ static void init_parameter(void) {
 #endif
 #endif
 
+#ifdef SANDYBRIDGE
+
+#ifdef DEBUG
+  fprintf(stderr, "Sandybridge\n");
+#endif
+
+  TABLE_NAME.sgemm_p = SGEMM_DEFAULT_P;
+  TABLE_NAME.dgemm_p = DGEMM_DEFAULT_P;
+  TABLE_NAME.cgemm_p = CGEMM_DEFAULT_P;
+  TABLE_NAME.zgemm_p = ZGEMM_DEFAULT_P;
+#ifdef EXPRECISION
+  TABLE_NAME.qgemm_p = QGEMM_DEFAULT_P;
+  TABLE_NAME.xgemm_p = XGEMM_DEFAULT_P;
+#endif
+#endif
+
 #ifdef OPTERON
 
 #ifdef DEBUG
@@ -778,6 +794,38 @@ static void init_parameter(void) {
 #endif
 #endif
 
+#ifdef BOBCAT
+
+#ifdef DEBUG
+  fprintf(stderr, "Bobcate\n");
+#endif
+
+  TABLE_NAME.sgemm_p = SGEMM_DEFAULT_P;
+  TABLE_NAME.dgemm_p = DGEMM_DEFAULT_P;
+  TABLE_NAME.cgemm_p = CGEMM_DEFAULT_P;
+  TABLE_NAME.zgemm_p = ZGEMM_DEFAULT_P;
+#ifdef EXPRECISION
+  TABLE_NAME.qgemm_p = QGEMM_DEFAULT_P;
+  TABLE_NAME.xgemm_p = XGEMM_DEFAULT_P;
+#endif
+#endif
+
+#ifdef BULLDOZER
+
+#ifdef DEBUG
+  fprintf(stderr, "Bulldozer\n");
+#endif
+
+  TABLE_NAME.sgemm_p = SGEMM_DEFAULT_P;
+  TABLE_NAME.dgemm_p = DGEMM_DEFAULT_P;
+  TABLE_NAME.cgemm_p = CGEMM_DEFAULT_P;
+  TABLE_NAME.zgemm_p = ZGEMM_DEFAULT_P;
+#ifdef EXPRECISION
+  TABLE_NAME.qgemm_p = QGEMM_DEFAULT_P;
+  TABLE_NAME.xgemm_p = XGEMM_DEFAULT_P;
+#endif
+#endif
+
 #ifdef NANO
 
 #ifdef DEBUG

kernel/x86/KERNEL.BOBCAT

@@ -0,0 +1,59 @@
+SGEMMKERNEL    =  gemm_kernel_4x4_barcelona.S
+SGEMMINCOPY    =  
+SGEMMITCOPY    =  
+SGEMMONCOPY    =  ../generic/gemm_ncopy_4.c
+SGEMMOTCOPY    =  ../generic/gemm_tcopy_4.c
+SGEMMINCOPYOBJ =  
+SGEMMITCOPYOBJ =  
+SGEMMONCOPYOBJ =  sgemm_oncopy$(TSUFFIX).$(SUFFIX)
+SGEMMOTCOPYOBJ =  sgemm_otcopy$(TSUFFIX).$(SUFFIX)
+DGEMMKERNEL    =  gemm_kernel_2x4_barcelona.S
+DGEMMINCOPY    =  ../generic/gemm_ncopy_2.c
+DGEMMITCOPY    =  ../generic/gemm_tcopy_2.c
+DGEMMONCOPY    =  ../generic/gemm_ncopy_4.c
+DGEMMOTCOPY    =  ../generic/gemm_tcopy_4.c
+DGEMMINCOPYOBJ =  dgemm_incopy$(TSUFFIX).$(SUFFIX)
+DGEMMITCOPYOBJ =  dgemm_itcopy$(TSUFFIX).$(SUFFIX)
+DGEMMONCOPYOBJ =  dgemm_oncopy$(TSUFFIX).$(SUFFIX)
+DGEMMOTCOPYOBJ =  dgemm_otcopy$(TSUFFIX).$(SUFFIX)
+CGEMMKERNEL    =  zgemm_kernel_2x2_barcelona.S
+CGEMMINCOPY    =  
+CGEMMITCOPY    =  
+CGEMMONCOPY    =  ../generic/zgemm_ncopy_2.c
+CGEMMOTCOPY    =  ../generic/zgemm_tcopy_2.c
+CGEMMINCOPYOBJ =
+CGEMMITCOPYOBJ =  
+CGEMMONCOPYOBJ =  cgemm_oncopy$(TSUFFIX).$(SUFFIX)
+CGEMMOTCOPYOBJ =  cgemm_otcopy$(TSUFFIX).$(SUFFIX)
+ZGEMMKERNEL    =  zgemm_kernel_1x2_barcelona.S
+ZGEMMINCOPY    =  ../generic/zgemm_ncopy_1.c
+ZGEMMITCOPY    =  ../generic/zgemm_tcopy_1.c
+ZGEMMONCOPY    =  ../generic/zgemm_ncopy_2.c
+ZGEMMOTCOPY    =  ../generic/zgemm_tcopy_2.c
+ZGEMMINCOPYOBJ =  zgemm_incopy$(TSUFFIX).$(SUFFIX)
+ZGEMMITCOPYOBJ =  zgemm_itcopy$(TSUFFIX).$(SUFFIX)
+ZGEMMONCOPYOBJ =  zgemm_oncopy$(TSUFFIX).$(SUFFIX)
+ZGEMMOTCOPYOBJ =  zgemm_otcopy$(TSUFFIX).$(SUFFIX)
+
+STRSMKERNEL_LN	=  trsm_kernel_LN_4x4_sse.S
+STRSMKERNEL_LT	=  trsm_kernel_LT_4x4_sse.S
+STRSMKERNEL_RN	=  trsm_kernel_LT_4x4_sse.S
+STRSMKERNEL_RT	=  trsm_kernel_RT_4x4_sse.S
+
+DTRSMKERNEL_LN	=  trsm_kernel_LN_2x4_sse2.S
+DTRSMKERNEL_LT	=  trsm_kernel_LT_2x4_sse2.S
+DTRSMKERNEL_RN	=  trsm_kernel_LT_2x4_sse2.S
+DTRSMKERNEL_RT	=  trsm_kernel_RT_2x4_sse2.S
+
+CTRSMKERNEL_LN	=  ztrsm_kernel_LN_2x2_sse.S
+CTRSMKERNEL_LT	=  ztrsm_kernel_LT_2x2_sse.S
+CTRSMKERNEL_RN	=  ztrsm_kernel_LT_2x2_sse.S
+CTRSMKERNEL_RT	=  ztrsm_kernel_RT_2x2_sse.S
+
+ZTRSMKERNEL_LN	=  ztrsm_kernel_LT_1x2_sse2.S
+ZTRSMKERNEL_LT	=  ztrsm_kernel_LT_1x2_sse2.S
+ZTRSMKERNEL_RN	=  ztrsm_kernel_LT_1x2_sse2.S
+ZTRSMKERNEL_RT	=  ztrsm_kernel_RT_1x2_sse2.S
+
+CGEMM3MKERNEL    =  zgemm3m_kernel_4x4_barcelona.S
+ZGEMM3MKERNEL    =  zgemm3m_kernel_2x4_barcelona.S

kernel/x86/KERNEL.BULLDOZER

@@ -0,0 +1,59 @@
+SGEMMKERNEL    =  gemm_kernel_4x4_barcelona.S
+SGEMMINCOPY    =  
+SGEMMITCOPY    =  
+SGEMMONCOPY    =  ../generic/gemm_ncopy_4.c
+SGEMMOTCOPY    =  ../generic/gemm_tcopy_4.c
+SGEMMINCOPYOBJ =  
+SGEMMITCOPYOBJ =  
+SGEMMONCOPYOBJ =  sgemm_oncopy$(TSUFFIX).$(SUFFIX)
+SGEMMOTCOPYOBJ =  sgemm_otcopy$(TSUFFIX).$(SUFFIX)
+DGEMMKERNEL    =  gemm_kernel_2x4_barcelona.S
+DGEMMINCOPY    =  ../generic/gemm_ncopy_2.c
+DGEMMITCOPY    =  ../generic/gemm_tcopy_2.c
+DGEMMONCOPY    =  ../generic/gemm_ncopy_4.c
+DGEMMOTCOPY    =  ../generic/gemm_tcopy_4.c
+DGEMMINCOPYOBJ =  dgemm_incopy$(TSUFFIX).$(SUFFIX)
+DGEMMITCOPYOBJ =  dgemm_itcopy$(TSUFFIX).$(SUFFIX)
+DGEMMONCOPYOBJ =  dgemm_oncopy$(TSUFFIX).$(SUFFIX)
+DGEMMOTCOPYOBJ =  dgemm_otcopy$(TSUFFIX).$(SUFFIX)
+CGEMMKERNEL    =  zgemm_kernel_2x2_barcelona.S
+CGEMMINCOPY    =  
+CGEMMITCOPY    =  
+CGEMMONCOPY    =  ../generic/zgemm_ncopy_2.c
+CGEMMOTCOPY    =  ../generic/zgemm_tcopy_2.c
+CGEMMINCOPYOBJ =
+CGEMMITCOPYOBJ =  
+CGEMMONCOPYOBJ =  cgemm_oncopy$(TSUFFIX).$(SUFFIX)
+CGEMMOTCOPYOBJ =  cgemm_otcopy$(TSUFFIX).$(SUFFIX)
+ZGEMMKERNEL    =  zgemm_kernel_1x2_barcelona.S
+ZGEMMINCOPY    =  ../generic/zgemm_ncopy_1.c
+ZGEMMITCOPY    =  ../generic/zgemm_tcopy_1.c
+ZGEMMONCOPY    =  ../generic/zgemm_ncopy_2.c
+ZGEMMOTCOPY    =  ../generic/zgemm_tcopy_2.c
+ZGEMMINCOPYOBJ =  zgemm_incopy$(TSUFFIX).$(SUFFIX)
+ZGEMMITCOPYOBJ =  zgemm_itcopy$(TSUFFIX).$(SUFFIX)
+ZGEMMONCOPYOBJ =  zgemm_oncopy$(TSUFFIX).$(SUFFIX)
+ZGEMMOTCOPYOBJ =  zgemm_otcopy$(TSUFFIX).$(SUFFIX)
+
+STRSMKERNEL_LN	=  trsm_kernel_LN_4x4_sse.S
+STRSMKERNEL_LT	=  trsm_kernel_LT_4x4_sse.S
+STRSMKERNEL_RN	=  trsm_kernel_LT_4x4_sse.S
+STRSMKERNEL_RT	=  trsm_kernel_RT_4x4_sse.S
+
+DTRSMKERNEL_LN	=  trsm_kernel_LN_2x4_sse2.S
+DTRSMKERNEL_LT	=  trsm_kernel_LT_2x4_sse2.S
+DTRSMKERNEL_RN	=  trsm_kernel_LT_2x4_sse2.S
+DTRSMKERNEL_RT	=  trsm_kernel_RT_2x4_sse2.S
+
+CTRSMKERNEL_LN	=  ztrsm_kernel_LN_2x2_sse.S
+CTRSMKERNEL_LT	=  ztrsm_kernel_LT_2x2_sse.S
+CTRSMKERNEL_RN	=  ztrsm_kernel_LT_2x2_sse.S
+CTRSMKERNEL_RT	=  ztrsm_kernel_RT_2x2_sse.S
+
+ZTRSMKERNEL_LN	=  ztrsm_kernel_LT_1x2_sse2.S
+ZTRSMKERNEL_LT	=  ztrsm_kernel_LT_1x2_sse2.S
+ZTRSMKERNEL_RN	=  ztrsm_kernel_LT_1x2_sse2.S
+ZTRSMKERNEL_RT	=  ztrsm_kernel_RT_1x2_sse2.S
+
+CGEMM3MKERNEL    =  zgemm3m_kernel_4x4_barcelona.S
+ZGEMM3MKERNEL    =  zgemm3m_kernel_2x4_barcelona.S

kernel/x86/KERNEL.SANDYBRIDGE

@@ -0,0 +1 @@
+include $(KERNELDIR)/KERNEL.PENRYN

kernel/x86/amax_sse.S

@@ -495,7 +495,6 @@
 	ALIGN_4
 	
 .L999:
-	RESTOREREGISTERS
 
 	subl	$8, %esp
 	movss	%xmm0, (%esp)

kernel/x86/gemm_kernel_2x4_penryn.S

@@ -76,6 +76,12 @@
 #define PREFETCHB     prefetcht0
 #endif
 
+#ifdef SANDYBRIDGE
+#define PREFETCHSIZE  (8 * 1 - 4)
+#define PREFETCHW     prefetcht0
+#define PREFETCHB     prefetcht0
+#endif
+
 #ifndef PREFETCH
 #define PREFETCH      prefetcht0
 #endif

kernel/x86/gemm_kernel_4x4_barcelona.S

@@ -596,7 +596,7 @@
 .L22:
 	mulps	%xmm0, %xmm2
 	addps	%xmm2, %xmm4
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movsd	 4 * SIZE(BB), %xmm2
@@ -842,7 +842,7 @@
 .L32:
 	mulss	%xmm0, %xmm2
 	addss	%xmm2, %xmm4
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movss	 4 * SIZE(BB), %xmm2
@@ -1168,7 +1168,7 @@
 
 .L52:
 	mulps	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	mulps	 4 * SIZE(BB), %xmm0
@@ -1198,7 +1198,7 @@
 	addps	%xmm0, %xmm5
 	movaps	32 * SIZE(AA), %xmm0
 
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 16) * SIZE(AA)
 #endif
 	mulps	%xmm1, %xmm2
@@ -1347,7 +1347,7 @@
 	ALIGN_4
 
 .L62:
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 
@@ -1531,7 +1531,7 @@
 
 .L72:
 	mulss	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	mulss	 4 * SIZE(BB), %xmm0
@@ -1778,7 +1778,7 @@
 
 .L92:
 	mulps	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movaps	 4 * SIZE(AA), %xmm0
@@ -1793,7 +1793,7 @@
 	mulps	12 * SIZE(BB), %xmm0
 	addps	%xmm0, %xmm7
 	movaps	32 * SIZE(AA), %xmm0
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 16) * SIZE(AA)
 #endif
 	mulps	%xmm1, %xmm3
@@ -1924,7 +1924,7 @@
 
 .L102:
 	mulps	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movsd	 2 * SIZE(AA), %xmm0
@@ -2069,7 +2069,7 @@
 
 .L112:
 	mulss	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movss	 1 * SIZE(AA), %xmm0

kernel/x86/gemm_kernel_4x4_penryn.S

@@ -69,6 +69,12 @@
 #define PREFETCHB     prefetcht0
 #endif
 
+#ifdef SANDYBRIDGE
+#define PREFETCHSIZE  (16 * 1 - 8)
+#define PREFETCHW     prefetcht0
+#define PREFETCHB     prefetcht0
+#endif
+
 #ifndef PREFETCH
 #define PREFETCH      prefetcht0
 #endif
@@ -262,7 +268,7 @@
 	movaps	-16 * SIZE(AA), %xmm0
 
 	addps	%xmm2, %xmm7
-#ifndef NEHALEM
+#if !(defined(NEHALEM) || defined(SANDYBRIDGE))
 	PREFETCH (PREFETCHSIZE + 16) * SIZE(AA)
 #endif
 	pshufd	$0x93, %xmm1, %xmm2

kernel/x86/gemv_n_sse.S

@@ -58,7 +58,7 @@
 #define PREFETCHSIZE	(16 * 4)
 #endif
 
-#if defined(CORE2) || defined(PENRYN) || defined(DUNNINGTON) || defined(NEHALEM)
+#if defined(CORE2) || defined(PENRYN) || defined(DUNNINGTON) || defined(NEHALEM) || defined(SANDYBRIDGE)
 #define PREFETCH	prefetcht0
 #define PREFETCHW	prefetcht0
 #define PREFETCHSIZE	(16 * 7)
@@ -89,17 +89,22 @@
 #endif
 
 #define STACKSIZE	16
-
-#define M		 4 + STACKSIZE(%esp)
-#define N		 8 + STACKSIZE(%esp)
-#define ALPHA		16 + STACKSIZE(%esp)
-#define A		20 + STACKSIZE(%esp)
-#define STACK_LDA	24 + STACKSIZE(%esp)
-#define STACK_X		28 + STACKSIZE(%esp)
-#define STACK_INCX	32 + STACKSIZE(%esp)
-#define Y		36 + STACKSIZE(%esp)
-#define STACK_INCY	40 + STACKSIZE(%esp)
-#define BUFFER		44 + STACKSIZE(%esp)
+#define ARGS	16
+
+#define M		 4 + STACKSIZE+ARGS(%esp)
+#define N		 8 + STACKSIZE+ARGS(%esp)
+#define ALPHA		16 + STACKSIZE+ARGS(%esp)
+#define A		20 + STACKSIZE+ARGS(%esp)
+#define STACK_LDA	24 + STACKSIZE+ARGS(%esp)
+#define STACK_X		28 + STACKSIZE+ARGS(%esp)
+#define STACK_INCX	32 + STACKSIZE+ARGS(%esp)
+#define Y		36 + STACKSIZE+ARGS(%esp)
+#define STACK_INCY	40 + STACKSIZE+ARGS(%esp)
+#define BUFFER		44 + STACKSIZE+ARGS(%esp)
+#define MMM	0+ARGS(%esp)
+#define YY	4+ARGS(%esp)
+#define AA	8+ARGS(%esp)
+#define LDAX	12+ARGS(%esp)
 	
 #define I	%eax
 #define J	%ebx
@@ -114,6 +119,7 @@
 
 	PROLOGUE
 
+	subl	$ARGS,%esp
 	pushl	%ebp
 	pushl	%edi
 	pushl	%esi
@@ -121,7 +127,34 @@
 
 	PROFCODE
 
+	movl	Y,J
+	movl	J,YY				# backup Y
+	movl	A,J
+	movl	J,AA				# backup A
+	movl	M,J
+	movl	J,MMM				# backup MM
+.L0t:
+	xorl	J,J
+	addl	$1,J
+	sall	$21,J
+	subl	J,MMM
+	movl	J,M
+	jge		.L00t
+	ALIGN_4
+
+	movl	MMM,%eax
+	addl	J,%eax
+	jle		.L999x
+	movl	%eax,M
+
+.L00t:
+	movl	AA,%eax
+	movl	%eax,A
+
+	movl	YY,J
+	movl	J,Y
 	movl	STACK_LDA,  LDA
+
 	movl	STACK_X,    X
 	movl	STACK_INCX, INCX
 
@@ -651,12 +684,22 @@
 	addss	0 * SIZE(X), %xmm0
 	movss	%xmm0, (Y1)
 	ALIGN_3
-
 .L999:
+	movl	M,J
+	leal	(,J,SIZE),%eax
+	addl	%eax,AA
+	movl	YY,J
+	addl	%eax,J
+	movl	J,YY
+	jmp		.L0t
+	ALIGN_4
+
+.L999x:
 	popl	%ebx
 	popl	%esi
 	popl	%edi	
 	popl	%ebp
+	addl	$ARGS,%esp
 	ret
 
 	EPILOGUE

kernel/x86/gemv_n_sse2.S

@@ -45,7 +45,7 @@
 #define PREFETCHSIZE	(8 * 2)
 #endif
 
-#if defined(CORE2) || defined(PENRYN) || defined(DUNNINGTON) || defined(NEHALEM)
+#if defined(CORE2) || defined(PENRYN) || defined(DUNNINGTON) || defined(NEHALEM) || defined(SANDYBRIDGE)
 #define PREFETCH	prefetcht0
 #define PREFETCHW	prefetcht0
 #define PREFETCHSIZE	(8 * 7)
@@ -76,17 +76,22 @@
 #endif
 
 #define STACKSIZE	16
-
-#define M		 4 + STACKSIZE(%esp)
-#define N		 8 + STACKSIZE(%esp)
-#define ALPHA		16 + STACKSIZE(%esp)
-#define A		24 + STACKSIZE(%esp)
-#define STACK_LDA	28 + STACKSIZE(%esp)
-#define STACK_X		32 + STACKSIZE(%esp)
-#define STACK_INCX	36 + STACKSIZE(%esp)
-#define Y		40 + STACKSIZE(%esp)
-#define STACK_INCY	44 + STACKSIZE(%esp)
-#define BUFFER		48 + STACKSIZE(%esp)
+#define ARGS	16
+
+#define M		 4 + STACKSIZE+ARGS(%esp)
+#define N		 8 + STACKSIZE+ARGS(%esp)
+#define ALPHA		16 + STACKSIZE+ARGS(%esp)
+#define A		24 + STACKSIZE+ARGS(%esp)
+#define STACK_LDA	28 + STACKSIZE+ARGS(%esp)
+#define STACK_X		32 + STACKSIZE+ARGS(%esp)
+#define STACK_INCX	36 + STACKSIZE+ARGS(%esp)
+#define Y		40 + STACKSIZE+ARGS(%esp)
+#define STACK_INCY	44 + STACKSIZE+ARGS(%esp)
+#define BUFFER		48 + STACKSIZE+ARGS(%esp)
+
+#define MMM	0+ARGS(%esp)
+#define YY	4+ARGS(%esp)
+#define AA	8+ARGS(%esp)
 	
 #define I	%eax
 #define J	%ebx
@@ -101,6 +106,8 @@
 
 	PROLOGUE
 
+
+	subl	$ARGS,%esp
 	pushl	%ebp
 	pushl	%edi
 	pushl	%esi
@@ -108,6 +115,33 @@
 
 	PROFCODE
 
+	movl	Y,J
+	movl	J,YY				# backup Y
+	movl	A,J
+	movl	J,AA				# backup A
+	movl	M,J
+	movl	J,MMM				# backup MM
+.L0t:
+	xorl	J,J
+	addl	$1,J
+	sall	$20,J
+	subl	J,MMM
+	movl	J,M
+	jge		.L00t
+	ALIGN_4
+
+	movl	MMM,%eax
+	addl	J,%eax
+	jle		.L999x
+	movl	%eax,M
+
+.L00t:
+	movl	AA,%eax
+	movl	%eax,A
+
+	movl	YY,J
+	movl	J,Y
+
 	movl	STACK_LDA,  LDA
 	movl	STACK_X,    X
 	movl	STACK_INCX, INCX
@@ -677,10 +711,22 @@
 	ALIGN_3
 
 .L999:
+	movl	M,J
+	leal	(,J,SIZE),%eax
+	addl	%eax,AA
+	movl	YY,J
+	addl	%eax,J
+	movl	J,YY
+	jmp		.L0t
+	ALIGN_4
+
+.L999x:
+
 	popl	%ebx
 	popl	%esi
 	popl	%edi	
 	popl	%ebp
+	addl	$ARGS,%esp
 	ret
 
 	EPILOGUE

kernel/x86/gemv_t_sse.S

@@ -58,7 +58,7 @@
 #define PREFETCHSIZE	(16 * 4)
 #endif
 
-#if defined(CORE2) || defined(PENRYN) || defined(DUNNINGTON) || defined(NEHALEM)
+#if defined(CORE2) || defined(PENRYN) || defined(DUNNINGTON) || defined(NEHALEM) || defined(SANDYBRIDGE)
 #define PREFETCH	prefetcht0
 #define PREFETCHW	prefetcht0
 #define PREFETCHSIZE	(16 * 7)
@@ -89,17 +89,24 @@
 #endif
 
 #define STACKSIZE	16
-
-#define M		 4 + STACKSIZE(%esp)
-#define N		 8 + STACKSIZE(%esp)
-#define ALPHA		16 + STACKSIZE(%esp)
-#define A		20 + STACKSIZE(%esp)
-#define STACK_LDA	24 + STACKSIZE(%esp)
-#define STACK_X		28 + STACKSIZE(%esp)
-#define STACK_INCX	32 + STACKSIZE(%esp)
-#define Y		36 + STACKSIZE(%esp)
-#define STACK_INCY	40 + STACKSIZE(%esp)
-#define BUFFER		44 + STACKSIZE(%esp)
+#define ARGS	20
+
+#define M		 4 + STACKSIZE+ARGS(%esp)
+#define N		 8 + STACKSIZE+ARGS(%esp)
+#define ALPHA		16 + STACKSIZE+ARGS(%esp)
+#define A		20 + STACKSIZE+ARGS(%esp)
+#define STACK_LDA	24 + STACKSIZE+ARGS(%esp)
+#define STACK_X		28 + STACKSIZE+ARGS(%esp)
+#define STACK_INCX	32 + STACKSIZE+ARGS(%esp)
+#define Y		36 + STACKSIZE+ARGS(%esp)
+#define STACK_INCY	40 + STACKSIZE+ARGS(%esp)
+#define BUFFER		44 + STACKSIZE+ARGS(%esp)
+
+#define MMM	0+STACKSIZE(%esp)
+#define NN	4+STACKSIZE(%esp)
+#define AA	8+STACKSIZE(%esp)
+#define LDAX	12+STACKSIZE(%esp)
+#define XX	16+STACKSIZE(%esp)
 	
 #define I	%eax
 #define J	%ebx
@@ -114,6 +121,7 @@
 
 	PROLOGUE
 
+	subl	$ARGS,%esp
 	pushl	%ebp
 	pushl	%edi
 	pushl	%esi
@@ -122,7 +130,42 @@
 	PROFCODE
 
 	movl	STACK_LDA,  LDA
+	movl	LDA,LDAX			# backup LDA
 	movl	STACK_X,    X
+	movl	X,XX
+	movl	N,J
+	movl	J,NN				# backup N
+	movl	A,J
+	movl	J,AA				# backup A
+    movl	M,J
+	movl	J,MMM				# mov M to MMM
+.L0t:
+	xorl	J,J
+	addl	$1,J
+	sall    $22,J                           # J=2^24*sizeof(float)=buffer size(16MB)
+	subl    $8, J                           # Don't use last 8 float in the buffer.
+	                                        # Now, split M by block J
+	subl	J,MMM				# MMM=MMM-J
+	movl	J,M		
+	jge		.L00t
+	ALIGN_4
+	
+	movl	MMM,%eax
+	addl	J,%eax
+	jle		.L999x
+	movl	%eax,M
+
+.L00t:
+	movl	AA,%eax
+	movl	%eax,A			 	# mov AA to A
+
+	movl	NN,%eax
+	movl	%eax,N				# reset N
+
+
+	movl	LDAX,  LDA			# reset LDA
+	movl	XX,X
+
 	movl	STACK_INCX, INCX
 	movl	STACK_INCY, INCY
 
@@ -198,6 +241,20 @@
 	jg	.L06
 	ALIGN_4
 
+//Padding zero to prevent loading the dirty number from buffer.
+	movl	M,  I
+	movl	$8, J
+	andl	$7, I
+	xorps	%xmm0, %xmm0
+	subl	I, J
+	ALIGN_2
+.L07:
+	movss	%xmm0, 0 * SIZE(Y1)
+	addl	$SIZE, Y1
+	decl	J
+	jg	.L07
+	ALIGN_4
+
 .L10:
 	movl	Y, Y1
 
@@ -628,10 +685,22 @@
 	ALIGN_4
  	
 .L999:
+	movl	M,J
+	leal	(,J,SIZE),%eax
+	addl	%eax,AA
+	movl	XX,J
+	addl	%eax,J
+	movl	J,XX
+	jmp		.L0t
+	ALIGN_4
+
+.L999x:
 	popl	%ebx
 	popl	%esi
 	popl	%edi	
 	popl	%ebp
+
+	addl	$ARGS,%esp
 	ret
 
 	EPILOGUE

kernel/x86/gemv_t_sse2.S

@@ -45,7 +45,7 @@
 #define PREFETCHSIZE	(8 * 2)
 #endif
 
-#if defined(CORE2) || defined(PENRYN) || defined(DUNNINGTON) || defined(NEHALEM)
+#if defined(CORE2) || defined(PENRYN) || defined(DUNNINGTON) || defined(NEHALEM) || defined(SANDYBRIDGE)
 #define PREFETCH	prefetcht0
 #define PREFETCHW	prefetcht0
 #define PREFETCHSIZE	(8 * 7)
@@ -76,18 +76,24 @@
 #endif
 
 #define STACKSIZE	16
+#define ARGS	16
+
+#define M		 4 + STACKSIZE+ARGS(%esp)
+#define N		 8 + STACKSIZE+ARGS(%esp)
+#define ALPHA		16 + STACKSIZE+ARGS(%esp)
+#define A		24 + STACKSIZE+ARGS(%esp)
+#define STACK_LDA	28 + STACKSIZE+ARGS(%esp)
+#define STACK_X		32 + STACKSIZE+ARGS(%esp)
+#define STACK_INCX	36 + STACKSIZE+ARGS(%esp)
+#define Y		40 + STACKSIZE+ARGS(%esp)
+#define STACK_INCY	44 + STACKSIZE+ARGS(%esp)
+#define BUFFER		48 + STACKSIZE+ARGS(%esp)
+
+#define MMM	0+STACKSIZE(%esp)
+#define AA	4+STACKSIZE(%esp)
+#define LDAX 8+STACKSIZE(%esp)
+#define NN	12+STACKSIZE(%esp)
 
-#define M		 4 + STACKSIZE(%esp)
-#define N		 8 + STACKSIZE(%esp)
-#define ALPHA		16 + STACKSIZE(%esp)
-#define A		24 + STACKSIZE(%esp)
-#define STACK_LDA	28 + STACKSIZE(%esp)
-#define STACK_X		32 + STACKSIZE(%esp)
-#define STACK_INCX	36 + STACKSIZE(%esp)
-#define Y		40 + STACKSIZE(%esp)
-#define STACK_INCY	44 + STACKSIZE(%esp)
-#define BUFFER		48 + STACKSIZE(%esp)
-	
 #define I	%eax
 #define J	%ebx
 
@@ -101,6 +107,8 @@
 
 	PROLOGUE
 
+	subl	$ARGS,%esp
+
 	pushl	%ebp
 	pushl	%edi
 	pushl	%esi
@@ -108,7 +116,40 @@
 
 	PROFCODE
 
+
 	movl	STACK_LDA,  LDA
+	movl	LDA,LDAX			# backup LDA
+	movl	N,J
+	movl	J,NN				# backup N
+	movl	A,J
+	movl	J,AA				# backup A
+    movl	M,J
+	movl	J,MMM				# mov M to MMM
+.L0t:
+	xorl	J,J
+	addl	$1,J
+	sall    $21,J                           # J=2^21*sizeof(double)=buffer size(16MB)
+	subl    $4, J                           # Don't use last 4 double in the buffer.
+	                                        # Now, split M by block J
+	subl	J,MMM				# MMM=MMM-J
+	movl	J,M		
+	jge		.L00t
+	ALIGN_4
+	
+	movl	MMM,%eax
+	addl	J,%eax
+	jle		.L999x
+	movl	%eax,M
+
+.L00t:
+	movl	AA,%eax
+	movl	%eax,A			 	# mov AA to A
+
+	movl	NN,%eax
+	movl	%eax,N				# reset N
+
+
+	movl	LDAX,  LDA			# reset LDA
 	movl	STACK_X,    X
 	movl	STACK_INCX, INCX
 	movl	STACK_INCY, INCY
@@ -117,6 +158,7 @@
 	leal	(,INCY, SIZE), INCY
 	leal	(,LDA,  SIZE), LDA
 
+
 	subl	$-16 * SIZE, A
 
 	cmpl	$0, N
@@ -560,10 +602,19 @@
 	ALIGN_4
 	
 .L999:
+	movl 	M,J
+	leal 	(,J,SIZE),%eax
+	addl	%eax,AA
+	jmp		.L0t
+	ALIGN_4
+
+.L999x:
 	popl	%ebx
 	popl	%esi
 	popl	%edi	
 	popl	%ebp
+
+	addl	$ARGS,%esp
 	ret
 
 	EPILOGUE

kernel/x86/scal_sse.S

@@ -269,7 +269,7 @@
 	sarl	$5, I
 	jle	.L113
 
-#if defined(BARCELONA)
+#if defined(BARCELONA) || defined(BULLDOZER)
 
 	movaps	%xmm0, %xmm1
 	mulps	-32 * SIZE(X), %xmm1

kernel/x86/scal_sse2.S

@@ -76,7 +76,8 @@
 	xorps	%xmm1, %xmm1
 	comisd	%xmm0, %xmm1
 	jne	.L100		# Alpha != ZERO
-
+        jp      .L100           # For Alpha = NaN
+	
 /* Alpha == ZERO */
 	cmpl	$SIZE, INCX
 	jne	.L50
@@ -252,7 +253,7 @@
 	sarl	$4, I
 	jle	.L113
 
-#if defined(BARCELONA)
+#if defined(BARCELONA) || defined(BULLDOZER)
 
 	movaps  %xmm0, %xmm1
 	mulpd	-16 * SIZE(X), %xmm1

kernel/x86/trsm_kernel_LN_2x4_penryn.S

@@ -62,7 +62,7 @@
 #define PREFETCHSIZE  (8 * 21 + 4)
 #endif
 
-#ifdef NEHALEM
+#if defined(NEHALEM) || defined(SANDYBRIDGE)
 #define PREFETCH     prefetcht0
 #define PREFETCHSIZE  (8 * 21 + 4)
 #endif

kernel/x86/trsm_kernel_LN_2x4_sse2.S

@@ -69,7 +69,7 @@
 #define STACK_ALIGN	4096
 #define STACK_OFFSET	1024
 
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 #define PREFETCH     prefetch
 #define PREFETCHSIZE (8 * 10 + 4)
 #endif
@@ -439,7 +439,7 @@
 .L22:
 	mulsd	%xmm0, %xmm2
 	addsd	%xmm2, %xmm4
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	PREFETCH (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movlpd	 2 * SIZE(BB), %xmm2
@@ -488,7 +488,7 @@
 	movlpd	40 * SIZE(BB), %xmm3
 	addsd	%xmm0, %xmm7
 	movlpd	 8 * SIZE(AA), %xmm0
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	PREFETCH (PREFETCHSIZE  + 8) * SIZE(AA)
 #endif
 	mulsd	%xmm1, %xmm2
@@ -1697,7 +1697,7 @@
 
 .L42:
 	mulpd	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	mulpd	 2 * SIZE(BB), %xmm0
@@ -1727,7 +1727,7 @@
 	addpd	%xmm0, %xmm7
 	movapd	16 * SIZE(AA), %xmm0
 
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 8) * SIZE(AA)
 #endif
 	mulpd	%xmm1, %xmm2

kernel/x86/trsm_kernel_LN_4x4_penryn.S

@@ -62,7 +62,7 @@
 #define PREFETCHSIZE  (8 * 21 + 4)
 #endif
 
-#ifdef NEHALEM
+#if defined(NEHALEM) || defined(SANDYBRIDGE)
 #define PREFETCH     prefetcht0
 #define PREFETCHSIZE  (8 * 21 + 4)
 #endif

kernel/x86/trsm_kernel_LN_4x4_sse.S

@@ -64,7 +64,7 @@
 #define BORIG	60(%esp)
 #define BUFFER 128(%esp)
 
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 #define PREFETCH     prefetch
 #define PREFETCHW    prefetchw
 #define PREFETCHSIZE (16 * 10 + 8)
@@ -437,7 +437,7 @@
 .L32:
 	mulss	%xmm0, %xmm2
 	addss	%xmm2, %xmm4
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movss	 4 * SIZE(BB), %xmm2
@@ -833,7 +833,7 @@
 .L22:
 	mulps	%xmm0, %xmm2
 	addps	%xmm2, %xmm4
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movaps	 4 * SIZE(BB), %xmm2
@@ -1848,7 +1848,7 @@
 
 .L72:
 	mulss	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	mulss	 4 * SIZE(BB), %xmm0
@@ -2109,7 +2109,7 @@
 	ALIGN_4
 
 .L62:
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 
@@ -2429,7 +2429,7 @@
 
 .L52:
 	mulps	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	mulps	 4 * SIZE(BB), %xmm0
@@ -2459,7 +2459,7 @@
 	addps	%xmm0, %xmm5
 	movaps	32 * SIZE(AA), %xmm0
 
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 16) * SIZE(AA)
 #endif
 	mulps	%xmm1, %xmm2
@@ -2952,7 +2952,7 @@
 
 .L112:
 	mulss	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movss	 1 * SIZE(AA), %xmm0
@@ -3148,7 +3148,7 @@
 
 .L102:
 	mulps	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movsd	 2 * SIZE(AA), %xmm0
@@ -3389,7 +3389,7 @@
 
 .L92:
 	mulps	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movaps	 4 * SIZE(AA), %xmm0
@@ -3404,7 +3404,7 @@
 	mulps	12 * SIZE(BB), %xmm0
 	addps	%xmm0, %xmm7
 	movaps	32 * SIZE(AA), %xmm0
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 16) * SIZE(AA)
 #endif
 	mulps	%xmm1, %xmm3

kernel/x86/trsm_kernel_LT_2x4_penryn.S

@@ -62,7 +62,7 @@
 #define PREFETCHSIZE  (8 * 21 + 4)
 #endif
 
-#ifdef NEHALEM
+#if defined(NEHALEM) || defined(SANDYBRIDGE)
 #define PREFETCH     prefetcht0
 #define PREFETCHSIZE  (8 * 21 + 4)
 #endif

kernel/x86/trsm_kernel_LT_2x4_sse2.S

@@ -69,7 +69,7 @@
 #define STACK_ALIGN	4096
 #define STACK_OFFSET	1024
 
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 #define PREFETCH     prefetch
 #define PREFETCHSIZE (8 * 10 + 4)
 #endif
@@ -910,7 +910,7 @@
 .L22:
 	mulsd	%xmm0, %xmm2
 	addsd	%xmm2, %xmm4
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	PREFETCH (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movlpd	 2 * SIZE(BB), %xmm2
@@ -959,7 +959,7 @@
 	movlpd	40 * SIZE(BB), %xmm3
 	addsd	%xmm0, %xmm7
 	movlpd	 8 * SIZE(AA), %xmm0
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	PREFETCH (PREFETCHSIZE  + 8) * SIZE(AA)
 #endif
 	mulsd	%xmm1, %xmm2
@@ -1439,7 +1439,7 @@
 
 .L42:
 	mulpd	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	mulpd	 2 * SIZE(BB), %xmm0
@@ -1469,7 +1469,7 @@
 	addpd	%xmm0, %xmm7
 	movapd	16 * SIZE(AA), %xmm0
 
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 8) * SIZE(AA)
 #endif
 	mulpd	%xmm1, %xmm2

kernel/x86/trsm_kernel_LT_4x4_penryn.S

@@ -62,7 +62,7 @@
 #define PREFETCHSIZE  (8 * 21 + 4)
 #endif
 
-#ifdef NEHALEM
+#if defined(NEHALEM) || defined(SANDYBRIDGE)
 #define PREFETCH     prefetcht0
 #define PREFETCHSIZE  (8 * 21 + 4)
 #endif

kernel/x86/trsm_kernel_LT_4x4_sse.S

@@ -64,7 +64,7 @@
 #define BORIG	60(%esp)
 #define BUFFER 128(%esp)
 
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 #define PREFETCH     prefetch
 #define PREFETCHW    prefetchw
 #define PREFETCHSIZE (16 * 10 + 8)
@@ -872,7 +872,7 @@
 .L22:
 	mulps	%xmm0, %xmm2
 	addps	%xmm2, %xmm4
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movaps	 4 * SIZE(BB), %xmm2
@@ -1316,7 +1316,7 @@
 .L32:
 	mulss	%xmm0, %xmm2
 	addss	%xmm2, %xmm4
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movss	 4 * SIZE(BB), %xmm2
@@ -1855,7 +1855,7 @@
 
 .L52:
 	mulps	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	mulps	 4 * SIZE(BB), %xmm0
@@ -1885,7 +1885,7 @@
 	addps	%xmm0, %xmm5
 	movaps	32 * SIZE(AA), %xmm0
 
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 16) * SIZE(AA)
 #endif
 	mulps	%xmm1, %xmm2
@@ -2249,7 +2249,7 @@
 	ALIGN_4
 
 .L62:
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 
@@ -2562,7 +2562,7 @@
 
 .L72:
 	mulss	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	mulss	 4 * SIZE(BB), %xmm0
@@ -2957,7 +2957,7 @@
 
 .L92:
 	mulps	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movaps	 4 * SIZE(AA), %xmm0
@@ -2972,7 +2972,7 @@
 	mulps	12 * SIZE(BB), %xmm0
 	addps	%xmm0, %xmm7
 	movaps	32 * SIZE(AA), %xmm0
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 16) * SIZE(AA)
 #endif
 	mulps	%xmm1, %xmm3
@@ -3280,7 +3280,7 @@
 
 .L102:
 	mulps	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movsd	 2 * SIZE(AA), %xmm0
@@ -3515,7 +3515,7 @@
 
 .L112:
 	mulss	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movss	 1 * SIZE(AA), %xmm0

kernel/x86/trsm_kernel_RT_2x4_penryn.S

@@ -62,7 +62,7 @@
 #define PREFETCHSIZE  (8 * 21 + 4)
 #endif
 
-#ifdef NEHALEM
+#if defined(NEHALEM) || defined(SANDYBRIDGE)
 #define PREFETCH     prefetcht0
 #define PREFETCHSIZE  (8 * 21 + 4)
 #endif

kernel/x86/trsm_kernel_RT_2x4_sse2.S

@@ -69,7 +69,7 @@
 #define STACK_ALIGN	4096
 #define STACK_OFFSET	1024
 
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 #define PREFETCH     prefetch
 #define PREFETCHSIZE (8 * 10 + 4)
 #endif
@@ -1036,7 +1036,7 @@
 
 .L42:
 	mulpd	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	mulpd	 2 * SIZE(BB), %xmm0
@@ -1066,7 +1066,7 @@
 	addpd	%xmm0, %xmm7
 	movapd	16 * SIZE(AA), %xmm0
 
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 8) * SIZE(AA)
 #endif
 	mulpd	%xmm1, %xmm2
@@ -2224,7 +2224,7 @@
 .L22:
 	mulsd	%xmm0, %xmm2
 	addsd	%xmm2, %xmm4
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	PREFETCH (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movlpd	 2 * SIZE(BB), %xmm2
@@ -2273,7 +2273,7 @@
 	movlpd	40 * SIZE(BB), %xmm3
 	addsd	%xmm0, %xmm7
 	movlpd	 8 * SIZE(AA), %xmm0
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	PREFETCH (PREFETCHSIZE  + 8) * SIZE(AA)
 #endif
 	mulsd	%xmm1, %xmm2

kernel/x86/trsm_kernel_RT_4x4_penryn.S

@@ -62,7 +62,7 @@
 #define PREFETCHSIZE  (8 * 21 + 4)
 #endif
 
-#ifdef NEHALEM
+#if defined(NEHALEM) || defined(SANDYBRIDGE)
 #define PREFETCH     prefetcht0
 #define PREFETCHSIZE  (8 * 21 + 4)
 #endif

kernel/x86/trsm_kernel_RT_4x4_sse.S

@@ -64,7 +64,7 @@
 #define BORIG	60(%esp)
 #define BUFFER 128(%esp)
 
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 #define PREFETCH     prefetch
 #define PREFETCHW    prefetchw
 #define PREFETCHSIZE (16 * 10 + 8)
@@ -439,7 +439,7 @@
 
 .L92:
 	mulps	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movaps	 4 * SIZE(AA), %xmm0
@@ -454,7 +454,7 @@
 	mulps	12 * SIZE(BB), %xmm0
 	addps	%xmm0, %xmm7
 	movaps	32 * SIZE(AA), %xmm0
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 16) * SIZE(AA)
 #endif
 	mulps	%xmm1, %xmm3
@@ -758,7 +758,7 @@
 
 .L102:
 	mulps	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movsd	 2 * SIZE(AA), %xmm0
@@ -993,7 +993,7 @@
 
 .L112:
 	mulss	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movss	 1 * SIZE(AA), %xmm0
@@ -1324,7 +1324,7 @@
 
 .L52:
 	mulps	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	mulps	 4 * SIZE(BB), %xmm0
@@ -1354,7 +1354,7 @@
 	addps	%xmm0, %xmm5
 	movaps	32 * SIZE(AA), %xmm0
 
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 16) * SIZE(AA)
 #endif
 	mulps	%xmm1, %xmm2
@@ -1718,7 +1718,7 @@
 	ALIGN_4
 
 .L62:
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 
@@ -2031,7 +2031,7 @@
 
 .L72:
 	mulss	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	mulss	 4 * SIZE(BB), %xmm0
@@ -2859,7 +2859,7 @@
 .L22:
 	mulps	%xmm0, %xmm2
 	addps	%xmm2, %xmm4
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movaps	 4 * SIZE(BB), %xmm2
@@ -3303,7 +3303,7 @@
 .L32:
 	mulss	%xmm0, %xmm2
 	addss	%xmm2, %xmm4
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movss	 4 * SIZE(BB), %xmm2

kernel/x86/zdot_sse2.S

@@ -1541,6 +1541,16 @@
 	popl	%ebx
 	popl	%esi
 	popl	%edi
-/*remove the hidden return value address from the stack.*/
+#if defined(OS_WINNT) || defined(OS_CYGWIN_NT) || defined(OS_INTERIX)
+#ifdef MS_ABI
+/* For MingW GCC >= 4.7. It is compatible with MSVC ABI. http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36834 */
+	ret
+#else
+/* remove the hidden return value address from the stack. For MingW GCC < 4.7 */
 	ret	$0x4
+#endif
+#else
+/*remove the hidden return value address from the stack on Linux.*/
+	ret $0x4
+#endif
 	EPILOGUE

kernel/x86/zgemm3m_kernel_4x4_barcelona.S

@@ -74,7 +74,7 @@
 #define	BB	%ecx
 #define LDC	%ebp
 
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 #define movsd	movlps
 #endif
 
@@ -625,7 +625,7 @@
 .L22:
 	mulps	%xmm0, %xmm2
 	addps	%xmm2, %xmm4
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movsd	 4 * SIZE(BB), %xmm2
@@ -870,7 +870,7 @@
 .L32:
 	mulss	%xmm0, %xmm2
 	addss	%xmm2, %xmm4
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movss	 4 * SIZE(BB), %xmm2
@@ -1173,7 +1173,7 @@
 
 .L52:
 	mulps	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	mulps	 4 * SIZE(BB), %xmm0
@@ -1203,7 +1203,7 @@
 	addps	%xmm0, %xmm5
 	movaps	32 * SIZE(AA), %xmm0
 
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 16) * SIZE(AA)
 #endif
 	mulps	%xmm1, %xmm2
@@ -1359,7 +1359,7 @@
 	ALIGN_4
 
 .L62:
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 
@@ -1536,7 +1536,7 @@
 
 .L72:
 	mulss	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	mulss	 4 * SIZE(BB), %xmm0
@@ -1794,7 +1794,7 @@
 
 .L92:
 	mulps	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movaps	 4 * SIZE(AA), %xmm0
@@ -1809,7 +1809,7 @@
 	mulps	12 * SIZE(BB), %xmm0
 	addps	%xmm0, %xmm7
 	movaps	32 * SIZE(AA), %xmm0
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 16) * SIZE(AA)
 #endif
 	mulps	%xmm1, %xmm3
@@ -1936,7 +1936,7 @@
 
 .L102:
 	mulps	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movsd	 2 * SIZE(AA), %xmm0
@@ -2069,7 +2069,7 @@
 
 .L112:
 	mulss	%xmm0, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BULLDOZER)
 	prefetcht0 (PREFETCHSIZE  + 0) * SIZE(AA)
 #endif
 	movss	 1 * SIZE(AA), %xmm0

kernel/x86/zgemm_kernel_1x2_penryn.S

@@ -64,7 +64,7 @@
 #define PREFETCHB     prefetcht0
 #endif
 
-#ifdef NEHALEM
+#if defined(NEHALEM) || defined(SANDYBRIDGE)
 #define PREFETCHSIZE  (8 * 1 - 4)
 #define PREFETCHW     prefetcht0
 #define PREFETCHB     prefetcht0

kernel/x86/zgemm_kernel_2x2_penryn.S

@@ -64,7 +64,7 @@
 #define PREFETCHB     prefetcht0
 #endif
 
-#ifdef NEHALEM
+#if defined(NEHALEM) || defined(SANDYBRIDGE)
 #define PREFETCHSIZE  (16 * 1 + 8)
 #define PREFETCHW     prefetcht0
 #define PREFETCHB     prefetcht0

kernel/x86/zgemv_n_sse.S

@@ -58,7 +58,7 @@
 #define PREFETCHSIZE	(16 * 2)
 #endif
 
-#if defined(CORE2) || defined(PENRYN) || defined(DUNNINGTON) || defined(NEHALEM)
+#if defined(CORE2) || defined(PENRYN) || defined(DUNNINGTON) || defined(NEHALEM) || defined(SANDYBRIDGE)
 #define PREFETCH	prefetcht0
 #define PREFETCHW	prefetcht0
 #define PREFETCHSIZE	(16 * 7)
@@ -71,7 +71,7 @@
 #define movsd		movlps
 #endif
 
-#ifdef BARCELONA
+#if defined(BARCELONA)  || defined(BULLDOZER)
 #define PREFETCH	prefetchnta
 #define PREFETCHW	prefetchw
 #define PREFETCHSIZE	(16 * 5)

kernel/x86/zgemv_n_sse2.S

@@ -45,7 +45,7 @@
 #define PREFETCHSIZE	(8 * 2)
 #endif
 
-#if defined(CORE2) || defined(PENRYN) || defined(DUNNINGTON) || defined(NEHALEM)
+#if defined(CORE2) || defined(PENRYN) || defined(DUNNINGTON) || defined(NEHALEM) || defined(SANDYBRIDGE)
 #define PREFETCH	prefetcht0
 #define PREFETCHW	prefetcht0
 #define PREFETCHSIZE	(8 * 7)
@@ -58,7 +58,7 @@
 #define movsd		movlps
 #endif
 
-#ifdef BARCELONA
+#if defined(BARCELONA) || defined(BULLDOZER)
 #define PREFETCH	prefetchnta
 #define PREFETCHW	prefetchw
 #define PREFETCHSIZE	(8 * 5)

kernel/x86/zgemv_t_sse.S

@@ -58,7 +58,7 @@
 #define PREFETCHSIZE	(16 * 2)
 #endif
 
-#if defined(CORE2) || defined(PENRYN) || defined(DUNNINGTON) || defined(NEHALEM)
+#if defined(CORE2) || defined(PENRYN) || defined(DUNNINGTON) || defined(NEHALEM) || defined(SANDYBRIDGE)
 #define PREFETCH	prefetcht0
 #define PREFETCHW	prefetcht0
 #define PREFETCHSIZE	(16 * 7)
@@ -71,7 +71,7 @@
 #define movsd		movlps
 #endif
 
-#ifdef BARCELONA
+#if defined(BARCELONA) || defined(BULLDOZER)
 #define PREFETCH	prefetchnta
 #define PREFETCHW	prefetchw
 #define PREFETCHSIZE	(16 * 5)

kernel/x86/zgemv_t_sse2.S

@@ -45,7 +45,7 @@
 #define PREFETCHSIZE	(8 * 2)
 #endif
 
-#if defined(CORE2) || defined(PENRYN) || defined(DUNNINGTON) || defined(NEHALEM)
+#if defined(CORE2) || defined(PENRYN) || defined(DUNNINGTON) || defined(NEHALEM) || defined(SANDYBRIDGE)
 #define PREFETCH	prefetcht0
 #define PREFETCHW	prefetcht0
 #define PREFETCHSIZE	(8 * 7)
@@ -58,7 +58,7 @@
 #define movsd		movlps
 #endif
 
-#ifdef BARCELONA
+#if defined(BARCELONA) || defined(BULLDOZER)
 #define PREFETCH	prefetchnta
 #define PREFETCHW	prefetchw
 #define PREFETCHSIZE	(8 * 5)

kernel/x86/zscal_sse.S

@@ -55,7 +55,7 @@
 #define XX	%edi
 #define FLAG	%ebp
 
-#if defined(NEHALEM) || defined(PENRYN) || defined(DUNNINGTON)
+#if defined(NEHALEM) || defined(PENRYN) || defined(DUNNINGTON) || defined(SANDYBRIDGE)
 #define  USE_PSHUFD
 #else
 #define USE_PSHUFD_HALF
@@ -697,7 +697,7 @@
 	cmpl	$2 * SIZE, INCX
 	jne	.L120
 
-#if defined(ALIGNED_ACCESS) && !defined(NEHALEM)
+#if defined(ALIGNED_ACCESS) && !defined(NEHALEM) &&  !defined(SANDYBRIDGE)
 
 	PSHUFD2($0, %xmm0, %xmm6)
 	PSHUFD2($0, %xmm1, %xmm1)

kernel/x86/zscal_sse2.S

@@ -57,7 +57,7 @@
 
 #include "l1param.h"
 
-#if defined(NEHALEM) || defined(PENRYN) || defined(DUNNINGTON)
+#if defined(NEHALEM) || defined(PENRYN) || defined(DUNNINGTON) || defined(SANDYBRIDGE)
 #define  USE_PSHUFD
 #else
 #define USE_PSHUFD_HALF
@@ -860,7 +860,7 @@
 	cmpl	$2 * SIZE, INCX
 	jne	.L220
 
-#if defined(ALIGNED_ACCESS) && !defined(NEHALEM)
+#if defined(ALIGNED_ACCESS) && !defined(NEHALEM) && !defined(SANDYBRIDGE)
 
 #ifdef HAVE_SSE3
 	movddup	%xmm0, %xmm6

kernel/x86/ztrsm_kernel_LN_2x2_penryn.S

@@ -61,7 +61,7 @@
 #define PREFETCHSIZE 84
 #endif
 
-#ifdef NEHALEM
+#if defined(NEHALEM) || defined(SANDYBRIDGE)
 #define PREFETCH	prefetcht1
 #define PREFETCHSIZE 84
 #endif

kernel/x86/ztrsm_kernel_LN_2x2_sse.S

@@ -75,7 +75,7 @@
 #define STACK_ALIGN	4096
 #define STACK_OFFSET	1024
 
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 #define PREFETCHSIZE (16 * 10 + 8)
 #define WPREFETCHSIZE 112
 #define PREFETCH      prefetch
@@ -533,7 +533,7 @@
 	addps	%xmm0, %xmm7
 	movsd	16 * SIZE(AA), %xmm0
 	mulps	%xmm1, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht1     (PREFETCHSIZE + 16) * SIZE(AA)
 #endif
 	addps	%xmm2, %xmm4

kernel/x86/ztrsm_kernel_LT_1x2_penryn.S

@@ -63,7 +63,7 @@
 #define PREFETCHSIZE 84
 #endif
 
-#ifdef NEHALEM
+#if defined(NEHALEM) || defined(SANDYBRIDGE)
 #define PREFETCH	prefetcht1
 #define PREFETCHSIZE 84
 #endif

kernel/x86/ztrsm_kernel_LT_2x2_penryn.S

@@ -61,7 +61,7 @@
 #define PREFETCHSIZE 84
 #endif
 
-#ifdef NEHALEM
+#if defined(NEHALEM) || defined(SANDYBRIDGE)
 #define PREFETCH	prefetcht1
 #define PREFETCHSIZE 84
 #endif

kernel/x86/ztrsm_kernel_LT_2x2_sse.S

@@ -75,7 +75,7 @@
 #define STACK_ALIGN	4096
 #define STACK_OFFSET	1024
 
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 #define PREFETCHSIZE (16 * 10 + 8)
 #define WPREFETCHSIZE 112
 #define PREFETCH      prefetch
@@ -994,7 +994,7 @@
 	addps	%xmm0, %xmm7
 	movsd	16 * SIZE(AA), %xmm0
 	mulps	%xmm1, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht1     (PREFETCHSIZE + 16) * SIZE(AA)
 #endif
 	addps	%xmm2, %xmm4

kernel/x86/ztrsm_kernel_RT_1x2_penryn.S

@@ -63,7 +63,7 @@
 #define PREFETCHSIZE 84
 #endif
 
-#ifdef NEHALEM
+#if defined(NEHALEM) || defined(SANDYBRIDGE)
 #define PREFETCH	prefetcht1
 #define PREFETCHSIZE 84
 #endif

kernel/x86/ztrsm_kernel_RT_2x2_penryn.S

@@ -61,7 +61,7 @@
 #define PREFETCHSIZE 84
 #endif
 
-#ifdef NEHALEM
+#if defined(NEHALEM) || defined(SANDYBRIDGE)
 #define PREFETCH	prefetcht1
 #define PREFETCHSIZE 84
 #endif

kernel/x86/ztrsm_kernel_RT_2x2_sse.S

@@ -75,7 +75,7 @@
 #define STACK_ALIGN	4096
 #define STACK_OFFSET	1024
 
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 #define PREFETCHSIZE (16 * 10 + 8)
 #define WPREFETCHSIZE 112
 #define PREFETCH      prefetch
@@ -1820,7 +1820,7 @@
 	addps	%xmm0, %xmm7
 	movsd	16 * SIZE(AA), %xmm0
 	mulps	%xmm1, %xmm2
-#if defined(OPTERON) || defined(BARCELONA)
+#if defined(OPTERON) || defined(BARCELONA) || defined(BOBCAT) || defined(BULLDOZER)
 	prefetcht1     (PREFETCHSIZE + 16) * SIZE(AA)
 #endif
 	addps	%xmm2, %xmm4

kernel/x86_64/KERNEL.BOBCAT

@@ -0,0 +1,62 @@
+ZGEMVNKERNEL = zgemv_n_dup.S
+ZGEMVTKERNEL = zgemv_t_dup.S
+
+SGEMMKERNEL    =  gemm_kernel_8x4_barcelona.S
+SGEMMINCOPY    =  ../generic/gemm_ncopy_8.c
+SGEMMITCOPY    =  ../generic/gemm_tcopy_8.c
+SGEMMONCOPY    =  gemm_ncopy_4_opteron.S
+SGEMMOTCOPY    =  gemm_tcopy_4_opteron.S
+SGEMMINCOPYOBJ =  sgemm_incopy$(TSUFFIX).$(SUFFIX) 
+SGEMMITCOPYOBJ =  sgemm_itcopy$(TSUFFIX).$(SUFFIX) 
+SGEMMONCOPYOBJ =  sgemm_oncopy$(TSUFFIX).$(SUFFIX)
+SGEMMOTCOPYOBJ =  sgemm_otcopy$(TSUFFIX).$(SUFFIX)
+DGEMMKERNEL    =  gemm_kernel_4x4_barcelona.S
+DGEMMINCOPY    =
+DGEMMITCOPY    =
+DGEMMONCOPY    =  gemm_ncopy_4_opteron.S
+DGEMMOTCOPY    =  gemm_tcopy_4_opteron.S
+DGEMMINCOPYOBJ =
+DGEMMITCOPYOBJ =
+DGEMMONCOPYOBJ =  dgemm_oncopy$(TSUFFIX).$(SUFFIX)
+DGEMMOTCOPYOBJ =  dgemm_otcopy$(TSUFFIX).$(SUFFIX)
+CGEMMKERNEL    =  zgemm_kernel_4x2_barcelona.S
+CGEMMINCOPY    =  ../generic/zgemm_ncopy_4.c
+CGEMMITCOPY    =  ../generic/zgemm_tcopy_4.c
+CGEMMONCOPY    =  zgemm_ncopy_2.S
+CGEMMOTCOPY    =  zgemm_tcopy_2.S
+CGEMMINCOPYOBJ =  cgemm_incopy$(TSUFFIX).$(SUFFIX)
+CGEMMITCOPYOBJ =  cgemm_itcopy$(TSUFFIX).$(SUFFIX)
+CGEMMONCOPYOBJ =  cgemm_oncopy$(TSUFFIX).$(SUFFIX)
+CGEMMOTCOPYOBJ =  cgemm_otcopy$(TSUFFIX).$(SUFFIX)
+ZGEMMKERNEL    =  zgemm_kernel_2x2_barcelona.S
+ZGEMMINCOPY    =
+ZGEMMITCOPY    =
+ZGEMMONCOPY    =  zgemm_ncopy_2.S
+ZGEMMOTCOPY    =  zgemm_tcopy_2.S
+ZGEMMINCOPYOBJ =
+ZGEMMITCOPYOBJ =
+ZGEMMONCOPYOBJ =  zgemm_oncopy$(TSUFFIX).$(SUFFIX)
+ZGEMMOTCOPYOBJ =  zgemm_otcopy$(TSUFFIX).$(SUFFIX)
+
+STRSMKERNEL_LN	=  trsm_kernel_LN_8x4_sse.S
+STRSMKERNEL_LT	=  trsm_kernel_LT_8x4_sse.S
+STRSMKERNEL_RN	=  trsm_kernel_LT_8x4_sse.S
+STRSMKERNEL_RT	=  trsm_kernel_RT_8x4_sse.S
+
+DTRSMKERNEL_LN	=  trsm_kernel_LN_4x4_barcelona.S
+DTRSMKERNEL_LT	=  trsm_kernel_LT_4x4_barcelona.S
+DTRSMKERNEL_RN	=  trsm_kernel_LT_4x4_barcelona.S
+DTRSMKERNEL_RT	=  trsm_kernel_RT_4x4_barcelona.S
+
+CTRSMKERNEL_LN	=  ztrsm_kernel_LN_4x2_sse.S
+CTRSMKERNEL_LT	=  ztrsm_kernel_LT_4x2_sse.S
+CTRSMKERNEL_RN	=  ztrsm_kernel_LT_4x2_sse.S
+CTRSMKERNEL_RT	=  ztrsm_kernel_RT_4x2_sse.S
+
+ZTRSMKERNEL_LN	=  ztrsm_kernel_LN_2x2_sse2.S
+ZTRSMKERNEL_LT	=  ztrsm_kernel_LT_2x2_sse2.S
+ZTRSMKERNEL_RN	=  ztrsm_kernel_LT_2x2_sse2.S
+ZTRSMKERNEL_RT	=  ztrsm_kernel_RT_2x2_sse2.S
+
+CGEMM3MKERNEL    =  zgemm3m_kernel_8x4_barcelona.S
+ZGEMM3MKERNEL    =  zgemm3m_kernel_4x4_barcelona.S

kernel/x86_64/KERNEL.BULLDOZER

@@ -0,0 +1,62 @@
+ZGEMVNKERNEL = zgemv_n_dup.S
+ZGEMVTKERNEL = zgemv_t_dup.S
+
+SGEMMKERNEL    =  gemm_kernel_8x4_barcelona.S
+SGEMMINCOPY    =  ../generic/gemm_ncopy_8.c
+SGEMMITCOPY    =  ../generic/gemm_tcopy_8.c
+SGEMMONCOPY    =  gemm_ncopy_4_opteron.S
+SGEMMOTCOPY    =  gemm_tcopy_4_opteron.S
+SGEMMINCOPYOBJ =  sgemm_incopy$(TSUFFIX).$(SUFFIX) 
+SGEMMITCOPYOBJ =  sgemm_itcopy$(TSUFFIX).$(SUFFIX) 
+SGEMMONCOPYOBJ =  sgemm_oncopy$(TSUFFIX).$(SUFFIX)
+SGEMMOTCOPYOBJ =  sgemm_otcopy$(TSUFFIX).$(SUFFIX)
+DGEMMKERNEL    =  dgemm_kernel_4x4_bulldozer.S
+DGEMMINCOPY    =
+DGEMMITCOPY    =
+DGEMMONCOPY    =  gemm_ncopy_4_opteron.S
+DGEMMOTCOPY    =  gemm_tcopy_4_opteron.S
+DGEMMINCOPYOBJ =
+DGEMMITCOPYOBJ =
+DGEMMONCOPYOBJ =  dgemm_oncopy$(TSUFFIX).$(SUFFIX)
+DGEMMOTCOPYOBJ =  dgemm_otcopy$(TSUFFIX).$(SUFFIX)
+CGEMMKERNEL    =  zgemm_kernel_4x2_barcelona.S
+CGEMMINCOPY    =  ../generic/zgemm_ncopy_4.c
+CGEMMITCOPY    =  ../generic/zgemm_tcopy_4.c
+CGEMMONCOPY    =  zgemm_ncopy_2.S
+CGEMMOTCOPY    =  zgemm_tcopy_2.S
+CGEMMINCOPYOBJ =  cgemm_incopy$(TSUFFIX).$(SUFFIX)
+CGEMMITCOPYOBJ =  cgemm_itcopy$(TSUFFIX).$(SUFFIX)
+CGEMMONCOPYOBJ =  cgemm_oncopy$(TSUFFIX).$(SUFFIX)
+CGEMMOTCOPYOBJ =  cgemm_otcopy$(TSUFFIX).$(SUFFIX)
+ZGEMMKERNEL    =  zgemm_kernel_2x2_barcelona.S
+ZGEMMINCOPY    =
+ZGEMMITCOPY    =
+ZGEMMONCOPY    =  zgemm_ncopy_2.S
+ZGEMMOTCOPY    =  zgemm_tcopy_2.S
+ZGEMMINCOPYOBJ =
+ZGEMMITCOPYOBJ =
+ZGEMMONCOPYOBJ =  zgemm_oncopy$(TSUFFIX).$(SUFFIX)
+ZGEMMOTCOPYOBJ =  zgemm_otcopy$(TSUFFIX).$(SUFFIX)
+
+STRSMKERNEL_LN	=  trsm_kernel_LN_8x4_sse.S
+STRSMKERNEL_LT	=  trsm_kernel_LT_8x4_sse.S
+STRSMKERNEL_RN	=  trsm_kernel_LT_8x4_sse.S
+STRSMKERNEL_RT	=  trsm_kernel_RT_8x4_sse.S
+
+DTRSMKERNEL_LN	=  trsm_kernel_LN_4x4_barcelona.S
+DTRSMKERNEL_LT	=  trsm_kernel_LT_4x4_barcelona.S
+DTRSMKERNEL_RN	=  trsm_kernel_LT_4x4_barcelona.S
+DTRSMKERNEL_RT	=  trsm_kernel_RT_4x4_barcelona.S
+
+CTRSMKERNEL_LN	=  ztrsm_kernel_LN_4x2_sse.S
+CTRSMKERNEL_LT	=  ztrsm_kernel_LT_4x2_sse.S
+CTRSMKERNEL_RN	=  ztrsm_kernel_LT_4x2_sse.S
+CTRSMKERNEL_RT	=  ztrsm_kernel_RT_4x2_sse.S
+
+ZTRSMKERNEL_LN	=  ztrsm_kernel_LN_2x2_sse2.S
+ZTRSMKERNEL_LT	=  ztrsm_kernel_LT_2x2_sse2.S
+ZTRSMKERNEL_RN	=  ztrsm_kernel_LT_2x2_sse2.S
+ZTRSMKERNEL_RT	=  ztrsm_kernel_RT_2x2_sse2.S
+
+CGEMM3MKERNEL    =  zgemm3m_kernel_8x4_barcelona.S
+ZGEMM3MKERNEL    =  zgemm3m_kernel_4x4_barcelona.S

kernel/x86_64/KERNEL.SANDYBRIDGE

@@ -0,0 +1,84 @@
+SGEMMKERNEL    =  sgemm_kernel_8x8_sandy.S
+SGEMMINCOPY    =  
+SGEMMITCOPY    =  
+SGEMMONCOPY    =  ../generic/gemm_ncopy_8.c
+SGEMMOTCOPY    =  ../generic/gemm_tcopy_8.c
+SGEMMINCOPYOBJ =   
+SGEMMITCOPYOBJ =   
+SGEMMONCOPYOBJ =  sgemm_oncopy$(TSUFFIX).$(SUFFIX)
+SGEMMOTCOPYOBJ =  sgemm_otcopy$(TSUFFIX).$(SUFFIX)
+DGEMMKERNEL    =  dgemm_kernel_4x8_sandy.S
+DGEMMINCOPY    =  ../generic/gemm_ncopy_8.c
+DGEMMITCOPY    =  ../generic/gemm_tcopy_8.c
+#DGEMMONCOPY    =  gemm_ncopy_4.S
+DGEMMONCOPY    =  ../generic/gemm_ncopy_4.c
+DGEMMOTCOPY    =  ../generic/gemm_tcopy_4.c
+#DGEMMOTCOPY    =  gemm_tcopy_4.S
+DGEMMINCOPYOBJ =  dgemm_incopy$(TSUFFIX).$(SUFFIX)
+DGEMMITCOPYOBJ =  dgemm_itcopy$(TSUFFIX).$(SUFFIX)
+DGEMMONCOPYOBJ =  dgemm_oncopy$(TSUFFIX).$(SUFFIX)
+DGEMMOTCOPYOBJ =  dgemm_otcopy$(TSUFFIX).$(SUFFIX)
+#CGEMMKERNEL    =  zgemm_kernel_2x4_nehalem.S
+CGEMMKERNEL    =  cgemm_kernel_4x8_sandy.S
+CGEMMINCOPY    =  ../generic/zgemm_ncopy_8_sandy.c
+CGEMMITCOPY    =  ../generic/zgemm_tcopy_8_sandy.c
+CGEMMONCOPY    =  ../generic/zgemm_ncopy_4_sandy.c
+CGEMMOTCOPY    =  ../generic/zgemm_tcopy_4_sandy.c
+CGEMMINCOPYOBJ =  cgemm_incopy$(TSUFFIX).$(SUFFIX)
+CGEMMITCOPYOBJ =  cgemm_itcopy$(TSUFFIX).$(SUFFIX)
+CGEMMONCOPYOBJ =  cgemm_oncopy$(TSUFFIX).$(SUFFIX)
+CGEMMOTCOPYOBJ =  cgemm_otcopy$(TSUFFIX).$(SUFFIX)
+#ZGEMMKERNEL    =  zgemm_kernel_1x4_nehalem.S
+ZGEMMKERNEL    =  zgemm_kernel_4x4_sandy.S
+ZGEMMINCOPY    =  
+ZGEMMITCOPY    = 
+ZGEMMONCOPY    =  ../generic/zgemm_ncopy_4.c
+ZGEMMOTCOPY    =  ../generic/zgemm_tcopy_4.c
+ZGEMMINCOPYOBJ =  
+ZGEMMITCOPYOBJ =  
+ZGEMMONCOPYOBJ =  zgemm_oncopy$(TSUFFIX).$(SUFFIX)
+ZGEMMOTCOPYOBJ =  zgemm_otcopy$(TSUFFIX).$(SUFFIX)
+
+#STRSMKERNEL_LN	=  trsm_kernel_LN_4x8_nehalem.S
+#STRSMKERNEL_LT	=  trsm_kernel_LT_4x8_nehalem.S
+#STRSMKERNEL_RN	=  trsm_kernel_LT_4x8_nehalem.S
+#STRSMKERNEL_RT	=  trsm_kernel_RT_4x8_nehalem.S
+
+#DTRSMKERNEL_LN	=  trsm_kernel_LN_2x8_nehalem.S
+#DTRSMKERNEL_LT	=  trsm_kernel_LT_2x8_nehalem.S
+#DTRSMKERNEL_RN	=  trsm_kernel_LT_2x8_nehalem.S
+#DTRSMKERNEL_RT	=  trsm_kernel_RT_2x8_nehalem.S
+
+#CTRSMKERNEL_LN	=  ztrsm_kernel_LN_2x4_nehalem.S
+#CTRSMKERNEL_LT	=  ztrsm_kernel_LT_2x4_nehalem.S
+#CTRSMKERNEL_RN	=  ztrsm_kernel_LT_2x4_nehalem.S
+#CTRSMKERNEL_RT	=  ztrsm_kernel_RT_2x4_nehalem.S
+
+#ZTRSMKERNEL_LN	=  ztrsm_kernel_LT_1x4_nehalem.S
+#ZTRSMKERNEL_LT	=  ztrsm_kernel_LT_1x4_nehalem.S
+#ZTRSMKERNEL_RN	=  ztrsm_kernel_LT_1x4_nehalem.S
+#ZTRSMKERNEL_RT	=  ztrsm_kernel_RT_1x4_nehalem.S
+STRSMKERNEL_LN	=  ../generic/trsm_kernel_LN.c
+STRSMKERNEL_LT	=  ../generic/trsm_kernel_LT.c
+STRSMKERNEL_RN	=  ../generic/trsm_kernel_RN.c
+STRSMKERNEL_RT	=  ../generic/trsm_kernel_RT.c
+
+DTRSMKERNEL_LN	= ../generic/trsm_kernel_LN.c
+DTRSMKERNEL_LT	= ../generic/trsm_kernel_LT.c
+DTRSMKERNEL_RN	= ../generic/trsm_kernel_RN.c
+DTRSMKERNEL_RT	= ../generic/trsm_kernel_RT.c
+
+CTRSMKERNEL_LN	= ../generic/trsm_kernel_LN.c
+CTRSMKERNEL_LT	= ../generic/trsm_kernel_LT.c
+CTRSMKERNEL_RN	= ../generic/trsm_kernel_RN.c
+CTRSMKERNEL_RT	= ../generic/trsm_kernel_RT.c
+
+ZTRSMKERNEL_LN	= ../generic/trsm_kernel_LN.c
+ZTRSMKERNEL_LT	= ../generic/trsm_kernel_LT.c
+ZTRSMKERNEL_RN	= ../generic/trsm_kernel_RN.c
+ZTRSMKERNEL_RT	= ../generic/trsm_kernel_RT.c
+
+
+
+CGEMM3MKERNEL    =  zgemm3m_kernel_4x8_nehalem.S
+ZGEMM3MKERNEL    =  zgemm3m_kernel_2x8_nehalem.S

kernel/x86_64/axpy_sse.S

@@ -69,7 +69,7 @@
 #endif
 	movaps	%xmm0,  ALPHA
 #else
-	movaps	%xmm3,  ALPHA
+	
 
 	movq	40(%rsp), X
 	movq	48(%rsp), INCX
@@ -79,6 +79,10 @@
 
 	SAVEREGISTERS
 
+#ifdef WINDOWS_ABI
+	movaps	%xmm3,  ALPHA
+#endif
+	
 	shufps	$0, ALPHA, ALPHA
 
 	leaq	(, INCX, SIZE), INCX

kernel/x86_64/axpy_sse2.S

@@ -69,7 +69,6 @@
 #endif
 	movaps	%xmm0,  ALPHA
 #else
-	movaps	%xmm3,  ALPHA
 
 	movq	40(%rsp), X
 	movq	48(%rsp), INCX
@@ -79,6 +78,10 @@
 
 	SAVEREGISTERS
 
+#ifdef WINDOWS_ABI
+	movaps	%xmm3,  ALPHA
+#endif
+
 	unpcklpd ALPHA, ALPHA
 
 	leaq	(, INCX, SIZE), INCX