Merge branch 'hotfix-v0.2.8'

xianyi · xianyi · commit 835293cc1afd · 2013-08-01T23:53:12.000+08:00
diff --git a/CONTRIBUTORS.md b/CONTRIBUTORS.md
@@ -79,5 +79,9 @@ In chronological order:
   * [2013-07-11] create openblas_get_parallel to retrieve information which parallelization 
     model is used by OpenBLAS.
 
+* Sébastien Fabbro <bicatali@gentoo.org>
+  * [2013-07-24] Modify makefile to respect user's LDFLAGS
+  * [2013-07-24] Add stack markings for GNU as arch-independent for assembler files
+
 * [Your name or handle] <[email or website]>
   * [Date] [Brief summary of your changes]
diff --git a/Changelog.txt b/Changelog.txt
@@ -1,4 +1,16 @@
 OpenBLAS ChangeLog
+====================================================================
+Version 0.2.8
+01-Aug-2013
+common:
+	* Support Open64 5.0. (#266)
+	* Add executable stack markings. (#262, Thank Sébastien Fabbro)
+	* Respect user's LDFLAGS (Thank Sébastien Fabbro)
+
+x86/x86-64:
+	* Rollback bulldozer and piledriver kernels to barcelona kernels (#263)
+	  We will fix the compuational error bug in bulldozer and piledriver kernels.
+
 ====================================================================
 Version 0.2.7
 20-Jul-2013
diff --git a/Makefile.generic b/Makefile.generic
@@ -1,6 +1 @@
 COPT	= -Wall -O2 # -DGEMMTEST
-ifdef BINARY64
-else
-# LDFLAGS		 = -m elf32ppc
-LDFLAGS		 = -m elf_i386
-endif
diff --git a/Makefile.power b/Makefile.power
@@ -17,13 +17,7 @@ endif
 endif
 
 ifdef BINARY64
-ifeq ($(OSNAME), Linux)
-LDFLAGS		= -m elf64ppc
-endif
 
-ifeq ($(OSNAME), Darwin)
-LDFLAGS		= -arch ppc64
-endif
 
 ifeq ($(OSNAME), AIX)
 CCOMMON_OPT	+= -mpowerpc64 -maix64
@@ -34,17 +28,12 @@ ifeq ($(COMPILER_F77), xlf)
 FCOMMON_OPT	+= -q64
 endif
 ARFLAGS		= -X 64
-LDFLAGS		= -b64
 ASFLAGS		= -a64
 endif
 else
-ifeq ($(OSNAME), Linux)
-LDFLAGS		 = -m elf32ppc
-endif
 ifeq ($(OSNAME), AIX)
 CCOMMON_OPT	+= -Wa,-a32
 ARFLAGS		= -X 32
-LDFLAGS		= -b32
 ASFLAGS		= -a32
 endif
 endif
diff --git a/Makefile.rule b/Makefile.rule
@@ -3,7 +3,7 @@
 #
 
 # This library's version
-VERSION = 0.2.7
+VERSION = 0.2.8
 
 # If you set the suffix, the library name will be libopenblas_$(LIBNAMESUFFIX).a
 # and libopenblas_$(LIBNAMESUFFIX).so. Meanwhile, the soname in shared library 
diff --git a/Makefile.sparc b/Makefile.sparc
@@ -10,7 +10,6 @@ endif
 ifeq ($(COMPILER_F77), f90)
 FCOMMON_OPT += -xarch=v9
 endif
-LDFLAGS	= -64
 else
 
 CCOMMON_OPT += -mcpu=v9
diff --git a/Makefile.system b/Makefile.system
@@ -324,14 +324,16 @@ ifeq ($(ARCH), x86)
 DYNAMIC_CORE = KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS \
 	       CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
 ifneq ($(NO_AVX), 1)
-DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER
+DYNAMIC_CORE += SANDYBRIDGE 
+#BULLDOZER PILEDRIVER
 endif
 endif
 
 ifeq ($(ARCH), x86_64)
 DYNAMIC_CORE = PRESCOTT CORE2 PENRYN DUNNINGTON NEHALEM OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
 ifneq ($(NO_AVX), 1)
-DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER
+DYNAMIC_CORE += SANDYBRIDGE 
+#BULLDOZER PILEDRIVER
 endif
 endif
 
@@ -895,6 +897,7 @@ export CC
 export FC
 export BU
 export FU
+export NEED2UNDERSCORES
 export USE_THREAD
 export NUM_THREADS
 export NUM_CORES
diff --git a/Makefile.x86 b/Makefile.x86
@@ -1,8 +1,5 @@
 # COMPILER_PREFIX = mingw32-
 
-ifeq ($(OSNAME), Linux)
-LDFLAGS     = -melf_i386
-endif
 
 ifeq ($(OSNAME), Interix)
 ARFLAGS		= -m x86
diff --git a/Makefile.x86_64 b/Makefile.x86_64
@@ -2,25 +2,12 @@
 
 ifeq ($(OSNAME), SunOS)
 ifdef BINARY64
-LDFLAGS = -64
 ifeq ($(F_COMPILER), SUN)
 FCOMMON_OPT += -m64
 endif
 endif
 endif
 
-ifeq ($(OSNAME), FreeBSD)
-LDFLAGS		 = -m elf_x86_64_fbsd
-endif
-
-ifeq ($(OSNAME), Linux)
-LDFLAGS		 = -m elf_x86_64
-endif
-
-ifeq ($(OSNAME), Darwin)
-LDFLAGS		 = 
-endif
-
 ifeq ($(OSNAME), Interix)
 ARFLAGS		= -m x64
 endif
diff --git a/common_alpha.h b/common_alpha.h
@@ -150,9 +150,17 @@ static __inline int blas_quickdivide(unsigned int x, unsigned int y){
 #define PROFCODE .prologue 0
 #endif
 
+#if defined(__linux__) && defined(__ELF__)
+#define GNUSTACK .section .note.GNU-stack,"",%progbits
+#else
+#define GNUSTACK
+#endif
+
 #define EPILOGUE \
 	.end REALNAME; \
-	.ident VERSION
+	.ident VERSION; \
+	GNUSTACK
+
 #endif
 
 #ifdef DOUBLE
diff --git a/common_ia64.h b/common_ia64.h
@@ -379,8 +379,15 @@ static __inline int blas_quickdivide(unsigned int x, unsigned int y){
 #define PROFCODE
 #endif
 
+#if defined(__linux__) && defined(__ELF__)
+#define GNUSTACK .section .note.GNU-stack,"",%progbits
+#else
+#define GNUSTACK
+#endif
+
 #define EPILOGUE \
-	.endp REALNAME
+        .endp REALNAME ; \
+        GNUSTACK
 
 #define START_ADDRESS 0x20000fc800000000UL
 
diff --git a/common_mips64.h b/common_mips64.h
@@ -235,10 +235,17 @@ REALNAME: ;\
 	.set	noreorder ;\
 	.set	nomacro
 
+#if defined(__linux__) && defined(__ELF__)
+#define GNUSTACK .section .note.GNU-stack,"",%progbits
+#else
+#define GNUSTACK
+#endif
+
 #define EPILOGUE \
 	.set	macro ;\
 	.set	reorder ;\
-	.end	REALNAME
+	.end	REALNAME ;\
+	GNUSTACK
 
 #define PROFCODE
 #endif
diff --git a/common_sparc.h b/common_sparc.h
@@ -199,8 +199,17 @@ static __inline int blas_quickdivide(blasint x, blasint y){
 	.type	REALNAME, #function; \
 	.proc	07; \
 REALNAME:;
+
+#if defined(__linux__) && defined(__ELF__)
+#define GNUSTACK .section .note.GNU-stack,"",%progbits
+#else
+#define GNUSTACK
+#endif
+
 #define EPILOGUE \
-	.size	 REALNAME, .-REALNAME
+        .size	 REALNAME, .-REALNAME; \
+        GNUSTACK
+
 #endif
 
 #endif
diff --git a/common_x86.h b/common_x86.h
@@ -301,7 +301,9 @@ static __inline int blas_quickdivide(unsigned int x, unsigned int y){
 #define PROFCODE
 #endif
 
-#define EPILOGUE .size	 REALNAME, .-REALNAME
+#define EPILOGUE \
+        .size	 REALNAME, .-REALNAME; \
+        .section .note.GNU-stack,"",%progbits
 
 #endif
 
diff --git a/common_x86_64.h b/common_x86_64.h
@@ -372,7 +372,10 @@ static __inline int blas_quickdivide(unsigned int x, unsigned int y){
 #define PROFCODE
 #endif
 
-#define EPILOGUE .size	 REALNAME, .-REALNAME
+#define EPILOGUE \
+        .size	 REALNAME, .-REALNAME; \
+        .section .note.GNU-stack,"",%progbits
+
 
 #endif
 
diff --git a/cpuid.h b/cpuid.h
@@ -105,8 +105,8 @@
 #define CORE_NANO	19
 #define CORE_SANDYBRIDGE 20
 #define CORE_BOBCAT     21
-#define CORE_BULLDOZER  22
-#define CORE_PILEDRIVER  23
+#define CORE_BULLDOZER CORE_BARCELONA
+#define CORE_PILEDRIVER CORE_BARCELONA
 #define CORE_HASWELL CORE_SANDYBRIDGE
 
 #define HAVE_SSE      (1 <<  0)
@@ -198,8 +198,8 @@ typedef struct {
 #define CPUTYPE_NANO			43
 #define CPUTYPE_SANDYBRIDGE             44
 #define CPUTYPE_BOBCAT                  45
-#define CPUTYPE_BULLDOZER               46
-#define CPUTYPE_PILEDRIVER              47
+#define CPUTYPE_BULLDOZER CPUTYPE_BARCELONA
+#define CPUTYPE_PILEDRIVER CPUTYPE_BARCELONA
 // this define is because BLAS doesn't have haswell specific optimizations yet
 #define CPUTYPE_HASWELL CPUTYPE_SANDYBRIDGE 
 
diff --git a/cpuid_x86.c b/cpuid_x86.c
@@ -229,8 +229,8 @@ int get_cputype(int gettype){
     if ((ecx & (1 << 20)) != 0) feature |= HAVE_SSE4_2;
 #ifndef NO_AVX
     if (support_avx()) feature |= HAVE_AVX;
+    if ((ecx & (1 << 12)) != 0) feature |= HAVE_FMA3;
 #endif
-    if ((ecx & (1 << 20)) != 0) feature |= HAVE_FMA3;
 
     if (have_excpuid() >= 0x01) {
       cpuid(0x80000001, &eax, &ebx, &ecx, &edx);
diff --git a/ctest/Makefile b/ctest/Makefile
@@ -77,7 +77,7 @@ endif
 clean ::
 	rm -f x* 
 
-FLDFLAGS = $(FFLAGS:-fPIC=)
+FLDFLAGS = $(FFLAGS:-fPIC=) $(LDFLAGS)
 CEXTRALIB = 
 
 # Single real
diff --git a/driver/others/dynamic.c b/driver/others/dynamic.c
@@ -63,14 +63,16 @@ extern gotoblas_t  gotoblas_BARCELONA;
 extern gotoblas_t  gotoblas_BOBCAT;
 #ifndef NO_AVX
 extern gotoblas_t  gotoblas_SANDYBRIDGE;
-extern gotoblas_t  gotoblas_BULLDOZER;
-extern gotoblas_t  gotoblas_PILEDRIVER;
+//extern gotoblas_t  gotoblas_BULLDOZER;
+//extern gotoblas_t  gotoblas_PILEDRIVER;
 #else
 //Use NEHALEM kernels for sandy bridge
 #define gotoblas_SANDYBRIDGE gotoblas_NEHALEM
+#endif
+
 #define gotoblas_BULLDOZER gotoblas_BARCELONA
 #define gotoblas_PILEDRIVER gotoblas_BARCELONA
-#endif
+
 //Use sandy bridge kernels for haswell.
 #define gotoblas_HASWELL gotoblas_SANDYBRIDGE
 
diff --git a/driver/others/init.c b/driver/others/init.c
@@ -83,6 +83,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #include <dirent.h>
 #include <dlfcn.h>
 #include <unistd.h>
+#include <string.h>
 
 #define MAX_NODES	16
 #define MAX_CPUS	256
@@ -315,7 +316,7 @@ static int numa_check(void) {
   }
 
   while ((dir = readdir(dp)) != NULL) {
-    if (*(unsigned int *) dir -> d_name == 0x065646f6eU) {
+    if (strncmp(dir->d_name, "node", 4)==0) {
 
       node = atoi(&dir -> d_name[4]);
 
diff --git a/exports/Makefile b/exports/Makefile
diff --git a/exports/gensymbol b/exports/gensymbol
diff --git a/f_check b/f_check
diff --git a/ftest3.f b/ftest3.f
diff --git a/getarch.c b/getarch.c
diff --git a/getarch_2nd.c b/getarch_2nd.c
diff --git a/interface/trtri.c b/interface/trtri.c
diff --git a/lapack/trtri/dtrtri_lapack.f b/lapack/trtri/dtrtri_lapack.f
diff --git a/test/Makefile b/test/Makefile

Original file line number	Diff line number	Diff line change
`@@ -3,7 +3,7 @@`
`3`	`3`	`#`
`4`	`4`
`5`	`5`	`# This library's version`
`6`		`-VERSION = 0.2.7`
	`6`	`+VERSION = 0.2.8`
`7`	`7`
`8`	`8`	`# If you set the suffix, the library name will be libopenblas_$(LIBNAMESUFFIX).a`
`9`	`9`	`# and libopenblas_$(LIBNAMESUFFIX).so. Meanwhile, the soname in shared library`