Merge pull request #13 from xianyi/develop

resync with upstream

Makefile.system

@@ -769,6 +769,9 @@ else
 FCOMMON_OPT += -m32
 endif
 endif
+ifneq ($(NO_LAPACKE), 1)
+FCOMMON_OPT += -fno-second-underscore
+endif
 endif
 endif
 

cmake/arch.cmake

@@ -73,7 +73,7 @@ if (DYNAMIC_ARCH)
     endif ()
     if (NOT NO_AVX512)
       set(DYNAMIC_CORE ${DYNAMIC_CORE} SKYLAKEX)
-      string(REGEX REPLACE "-march=native" "" CMAKE_C_FLAGS ${CMAKE_C_FLAGS})
+      string(REGEX REPLACE "-march=native" "" CMAKE_C_FLAGS "${CMAKE_C_FLAGS}")
    endif ()
     if (DYNAMIC_LIST)
 	set(DYNAMIC_CORE PRESCOTT ${DYNAMIC_LIST})

common_arm64.h

@@ -78,7 +78,18 @@ static void __inline blas_lock(volatile BLASULONG *address){
 
 #define BLAS_LOCK_DEFINED
 
+#if !defined(OS_DARWIN) && !defined (OS_ANDROID)
+static __inline BLASULONG rpcc(void){
+  BLASULONG ret = 0;
+ 
+  __asm__ __volatile__ ("isb; mrs %0,cntvct_el0":"=r"(ret));
 
+  return ret;
+}
+
+#define RPCC_DEFINED
+#define RPCC64BIT
+#endif 
 
 static inline int blas_quickdivide(blasint x, blasint y){
   return x / y;

common_thread.h

@@ -194,10 +194,6 @@ int trsm_thread(int mode, BLASLONG m, BLASLONG n,
 
 int syrk_thread(int mode, blas_arg_t *, BLASLONG *, BLASLONG *, int (*function)(), void *, void *, BLASLONG);
 
-int beta_thread(int mode, BLASLONG m, BLASLONG n,
-		double alpha_r, double alpha_i,
-		void *c, BLASLONG ldc, int (*fuction)());
-
 int getrf_thread(int mode, BLASLONG m, BLASLONG n, BLASLONG k,
 		 void *offsetA, BLASLONG lda,
 		 void *offsetB, BLASLONG jb,

cpuid_x86.c

@@ -1197,7 +1197,11 @@ int get_cpuname(void){
 	case  3:
 	case  5:
 	case  6:
+#if defined(__x86_64__) || defined(__amd64__)
+	  return CPUTYPE_CORE2;
+#else
 	  return CPUTYPE_PENTIUM2;
+#endif
 	case  7:
 	case  8:
 	case 10:
@@ -1379,6 +1383,8 @@ int get_cpuname(void){
       break;
       case 7: // family 6 exmodel 7
         switch (model) {
+	case 10: // Goldmont Plus
+	    return CPUTYPE_NEHALEM;
         case 14: // Ice Lake
           if(support_avx512())
             return CPUTYPE_SKYLAKEX;

driver/others/memory.c

@@ -129,7 +129,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #include <unistd.h>
 #endif
 
-#if defined(OS_FREEBSD) || defined(OS_OPENBSD) || defined(OS_DRAGONFLY) || defined(OS_DARWIN)
+#if defined(OS_FREEBSD) || defined(OS_OPENBSD) || defined(OS_NETBSD) || defined(OS_DRAGONFLY) || defined(OS_DARWIN)
 #include <sys/sysctl.h>
 #include <sys/resource.h>
 #endif
@@ -192,7 +192,7 @@ void goto_set_num_threads(int num_threads) {};
 
 #else
 
-#if defined(OS_LINUX) || defined(OS_SUNOS) || defined(OS_NETBSD)
+#if defined(OS_LINUX) || defined(OS_SUNOS)
 #ifndef NO_AFFINITY
 int get_num_procs(void);
 #else
@@ -312,7 +312,7 @@ int get_num_procs(void) {
 
 #endif
 
-#if defined(OS_FREEBSD) || defined(OS_OPENBSD) || defined(OS_DRAGONFLY)
+#if defined(OS_FREEBSD) || defined(OS_OPENBSD) || defined(OS_NETBSD) || defined(OS_DRAGONFLY)
 
 int get_num_procs(void) {
 
@@ -404,7 +404,7 @@ extern int openblas_goto_num_threads_env();
 extern int openblas_omp_num_threads_env();
 
 int blas_get_cpu_number(void){
-#if defined(OS_LINUX) || defined(OS_WINDOWS) || defined(OS_FREEBSD) || defined(OS_OPENBSD) || defined(OS_DRAGONFLY) || defined(OS_DARWIN) || defined(OS_ANDROID)
+#if defined(OS_LINUX) || defined(OS_WINDOWS) || defined(OS_FREEBSD) || defined(OS_OPENBSD) || defined(OS_NETBSD) || defined(OS_DRAGONFLY) || defined(OS_DARWIN) || defined(OS_ANDROID)
   int max_num;
 #endif
   int blas_goto_num   = 0;
@@ -412,7 +412,7 @@ 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_OPENBSD) || defined(OS_DRAGONFLY) || defined(OS_DARWIN) || defined(OS_ANDROID)
+#if defined(OS_LINUX) || defined(OS_WINDOWS) || defined(OS_FREEBSD) || defined(OS_OPENBSD) || defined(OS_NETBSD) || defined(OS_DRAGONFLY) || defined(OS_DARWIN) || defined(OS_ANDROID)
   max_num = get_num_procs();
 #endif
 
@@ -436,7 +436,7 @@ 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_OPENBSD) || defined(OS_DRAGONFLY) || defined(OS_DARWIN) || defined(OS_ANDROID)
+#if defined(OS_LINUX) || defined(OS_WINDOWS) || defined(OS_FREEBSD) || defined(OS_OPENBSD) || defined(OS_NETBSD) || defined(OS_DRAGONFLY) || defined(OS_DARWIN) || defined(OS_ANDROID)
   if (blas_num_threads > max_num) blas_num_threads = max_num;
 #endif
 
@@ -1673,7 +1673,7 @@ void gotoblas_dummy_for_PGI(void) {
 #include <sys/resource.h>
 #endif
 
-#if defined(OS_FREEBSD) || defined(OS_OPENBSD) || defined(OS_DRAGONFLY) || defined(OS_DARWIN)
+#if defined(OS_FREEBSD) || defined(OS_OPENBSD) || defined(OS_NETBSD) || defined(OS_DRAGONFLY) || defined(OS_DARWIN)
 #include <sys/sysctl.h>
 #include <sys/resource.h>
 #endif
@@ -1736,7 +1736,7 @@ void goto_set_num_threads(int num_threads) {};
 
 #else
 
-#if defined(OS_LINUX) || defined(OS_SUNOS) || defined(OS_NETBSD)
+#if defined(OS_LINUX) || defined(OS_SUNOS)
 #ifndef NO_AFFINITY
 int get_num_procs(void);
 #else
@@ -1855,7 +1855,7 @@ int get_num_procs(void) {
 
 #endif
 
-#if defined(OS_FREEBSD) || defined(OS_OPENBSD) || defined(OS_DRAGONFLY)
+#if defined(OS_FREEBSD) || defined(OS_OPENBSD) || defined(OS_NETBSD) || defined(OS_DRAGONFLY)
 
 int get_num_procs(void) {
 
@@ -1945,7 +1945,7 @@ extern int openblas_goto_num_threads_env();
 extern int openblas_omp_num_threads_env();
 
 int blas_get_cpu_number(void){
-#if defined(OS_LINUX) || defined(OS_WINDOWS) || defined(OS_FREEBSD) || defined(OS_OPENBSD) || defined(OS_DRAGONFLY) || defined(OS_DARWIN) || defined(OS_ANDROID)
+#if defined(OS_LINUX) || defined(OS_WINDOWS) || defined(OS_FREEBSD) || defined(OS_OPENBSD) || defined(OS_NETBSD) || defined(OS_DRAGONFLY) || defined(OS_DARWIN) || defined(OS_ANDROID)
   int max_num;
 #endif
   int blas_goto_num   = 0;
@@ -1953,7 +1953,7 @@ 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_OPENBSD) || defined(OS_DRAGONFLY) || defined(OS_DARWIN) || defined(OS_ANDROID)
+#if defined(OS_LINUX) || defined(OS_WINDOWS) || defined(OS_FREEBSD) || defined(OS_OPENBSD) || defined(OS_NETBSD) || defined(OS_DRAGONFLY) || defined(OS_DARWIN) || defined(OS_ANDROID)
   max_num = get_num_procs();
 #endif
 
@@ -1977,7 +1977,7 @@ 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_OPENBSD) || defined(OS_DRAGONFLY) || defined(OS_DARWIN) || defined(OS_ANDROID)
+#if defined(OS_LINUX) || defined(OS_WINDOWS) || defined(OS_FREEBSD) || defined(OS_OPENBSD) || defined(OS_NETBSD) || defined(OS_DRAGONFLY) || defined(OS_DARWIN) || defined(OS_ANDROID)
   if (blas_num_threads > max_num) blas_num_threads = max_num;
 #endif
 

exports/gensymbol

@@ -618,19 +618,6 @@
     # functions added for lapack-3.7.0
 
     slarfy,
-    slasyf_rk,
-    ssyconvf_rook,
-    ssytf2_rk,
-    ssytrf_rk,
-    ssytrs_3,
-    ssytri_3,
-    ssytri_3x,
-    ssycon_3,
-    ssysv_rk,
-    slasyf_aa,
-    ssysv_aa,
-    ssytrf_aa,
-    ssytrs_aa,
     strevc3,
     sgelqt,
     sgelqt3,
@@ -647,33 +634,8 @@
     stplqt,
     stplqt2,
     stpmlqt,
-    ssytrd_2stage,
-    ssytrd_sy2sb,
-    ssytrd_sb2st,
-    ssb2st_kernels,
-    ssyevd_2stage,
-    ssyev_2stage,
-    ssyevx_2stage,
-    ssyevr_2stage,
-    ssbev_2stage,
-    ssbevx_2stage,
-    ssbevd_2stage,
-    ssygv_2stage,
     dlarfy,
-    dlasyf_rk,
     dsyconvf,
-    dsyconvf_rook,
-    dsytf2_rk,
-    dsytrf_rk,
-    dsytrs_3,
-    dsytri_3,
-    dsytri_3x,
-    dsycon_3,
-    dsysv_rk,
-    dlasyf_aa,
-    dsysv_aa,
-    dsytrf_aa,
-    dsytrs_aa,
     dtrevc3,
     dgelqt,
     dgelqt3,
@@ -690,45 +652,8 @@
     dtplqt,
     dtplqt2,
     dtpmlqt,
-    dsytrd_2stage,
-    dsytrd_sy2sb,
-    dsytrd_sb2st,
-    dsb2st_kernels,
-    dsyevd_2stage,
-    dsyev_2stage,
-    dsyevx_2stage,
-    dsyevr_2stage,
-    dsbev_2stage,
-    dsbevx_2stage,
-    dsbevd_2stage,
-    dsygv_2stage,
-    chetf2_rk,
-    chetrf_rk,
-    chetri_3,
-    chetri_3x,
-    chetrs_3,
-    checon_3,
-    chesv_rk,
-    chesv_aa,
-    chetrf_aa,
-    chetrs_aa,
-    clahef_aa,
-    clahef_rk,
     clarfy,
-    clasyf_rk,
-    clasyf_aa,
     csyconvf,
-    csyconvf_rook,
-    csytf2_rk,
-    csytrf_rk,
-    csytrf_aa,
-    csytrs_3,
-    csytrs_aa,
-    csytri_3,
-    csytri_3x,
-    csycon_3,
-    csysv_rk,
-    csysv_aa,
     ctrevc3,
     cgelqt,
     cgelqt3,
@@ -745,45 +670,8 @@
     ctplqt,
     ctplqt2,
     ctpmlqt,
-    chetrd_2stage,
-    chetrd_he2hb,
-    chetrd_hb2st,
-    chb2st_kernels,
-    cheevd_2stage,
-    cheev_2stage,
-    cheevx_2stage,
-    cheevr_2stage,
-    chbev_2stage,
-    chbevx_2stage,
-    chbevd_2stage,
-    chegv_2stage,
-    zhetf2_rk,
-    zhetrf_rk,
-    zhetri_3,
-    zhetri_3x,
-    zhetrs_3,
-    zhecon_3,
-    zhesv_rk,
-    zhesv_aa,
-    zhetrf_aa,
-    zhetrs_aa,
-    zlahef_aa,
-    zlahef_rk,
     zlarfy,
-    zlasyf_rk,
-    zlasyf_aa,
     zsyconvf,
-    zsyconvf_rook,
-    zsytrs_aa,
-    zsytf2_rk,
-    zsytrf_rk,
-    zsytrf_aa,
-    zsytrs_3,
-    zsytri_3,
-    zsytri_3x,
-    zsycon_3,
-    zsysv_rk,
-    zsysv_aa,
     ztrevc3,
     ztplqt,
     ztplqt2,
@@ -800,43 +688,13 @@
     zlaswlq,
     zlamswlq,
     zgemlq,
-    zhetrd_2stage,
-    zhetrd_he2hb,
-    zhetrd_hb2st,
-    zhb2st_kernels,
-    zheevd_2stage,
-    zheev_2stage,
-    zheevx_2stage,
-    zheevr_2stage,
-    zhbev_2stage,
-    zhbevx_2stage,
-    zhbevd_2stage,
-    zhegv_2stage,
     sladiv1,
     dladiv1,
     iparam2stage,
     
     # functions added for lapack-3.8.0
 
-    ilaenv2stage,
-    ssysv_aa_2stage,
-    ssytrf_aa_2stage,
-    ssytrs_aa_2stage,
-    chesv_aa_2stage,
-    chetrf_aa_2stage,
-    chetrs_aa_2stage,
-    csysv_aa_2stage,
-    csytrf_aa_2stage,
-    csytrs_aa_2stage,
-    dsysv_aa_2stage,
-    dsytrf_aa_2stage,
-    dsytrs_aa_2stage,
-    zhesv_aa_2stage,
-    zhetrf_aa_2stage,
-    zhetrs_aa_2stage,
-    zsysv_aa_2stage,
-    zsytrf_aa_2stage,
-    zsytrs_aa_2stage
+    ilaenv2stage
 );
 
 @lapack_extendedprecision_objs = (
@@ -3509,6 +3367,59 @@
     zlahef_rook, zlasyf_rook,
     zsytf2_rook, zsytrf_rook, zsytrs_rook,
     zsytri_rook, zsycon_rook, zsysv_rook,
+# 3.7.0
+    slasyf_rk, ssyconvf_rook, ssytf2_rk,
+    ssytrf_rk, ssytrs_3, ssytri_3,
+    ssytri_3x, ssycon_3, ssysv_rk,
+    slasyf_aa, ssysv_aa, ssytrf_aa,
+    ssytrs_aa, ssytrd_2stage, ssytrd_sy2sb,
+    ssytrd_sb2st, ssb2st_kernels, ssyevd_2stage,
+    ssyev_2stage, ssyevx_2stage, ssyevr_2stage,
+    ssbev_2stage, ssbevx_2stage, ssbevd_2stage,
+    ssygv_2stage, dlasyf_rk, dsyconvf_rook,
+    dsytf2_rk, dsytrf_rk, dsytrs_3,
+    dsytri_3, dsytri_3x, dsycon_3,
+    dsysv_rk, dlasyf_aa, dsysv_aa,
+    dsytrf_aa, dsytrs_aa, dsytrd_2stage,
+    dsytrd_sy2sb, dsytrd_sb2st, dsb2st_kernels,
+    dsyevd_2stage, dsyev_2stage, dsyevx_2stage,
+    dsyevr_2stage, dsbev_2stage, dsbevx_2stage,
+    dsbevd_2stage, dsygv_2stage, chetf2_rk,
+    chetrf_rk, chetri_3, chetri_3x,
+    chetrs_3, checon_3, chesv_rk,
+    chesv_aa, chetrf_aa, chetrs_aa,
+    clahef_aa, clahef_rk, clasyf_rk,
+    clasyf_aa, csytf2_rk, csytrf_rk,
+    csytrf_aa, csytrs_3, csytrs_aa,
+    csytri_3, csytri_3x, csycon_3,
+    csysv_rk, csysv_aa, csyconvf_rook,
+    chetrd_2stage, chetrd_he2hb, chetrd_hb2st,
+    chb2st_kernels, cheevd_2stage, cheev_2stage,
+    cheevx_2stage, cheevr_2stage, chbev_2stage,
+    chbevx_2stage, chbevd_2stage, chegv_2stage,
+    zhetf2_rk, zhetrf_rk, zhetri_3,
+    zhetri_3x, zhetrs_3, zhecon_3,
+    zhesv_rk, zhesv_aa, zhetrf_aa,
+    zhetrs_aa, zlahef_aa, zlahef_rk,
+    zlasyf_rk, zlasyf_aa, zsyconvf_rook,
+    zsytrs_aa, zsytf2_rk, zsytrf_rk,
+    zsytrf_aa, zsytrs_3, zsytri_3,
+    zsytri_3x, zsycon_3, zsysv_rk,
+    zsysv_aa, zhetrd_2stage, zhetrd_he2hb,
+    zhetrd_hb2st, zhb2st_kernels, zheevd_2stage,
+    zheev_2stage, zheevx_2stage, zheevr_2stage,
+    zhbev_2stage, zhbevx_2stage, zhbevd_2stage,
+    zhegv_2stage,
+# 3.8.0
+    ssysv_aa_2stage, ssytrf_aa_2stage,
+    ssytrs_aa_2stage, chesv_aa_2stage,
+    chetrf_aa_2stage, chetrs_aa_2stage,
+    csysv_aa_2stage, csytrf_aa_2stage,
+    csytrs_aa_2stage, dsysv_aa_2stage,
+    dsytrf_aa_2stage, dsytrs_aa_2stage,
+    zhesv_aa_2stage, zhetrf_aa_2stage,
+    zhetrs_aa_2stage, zsysv_aa_2stage,
+    zsytrf_aa_2stage, zsytrs_aa_2stage
 );
 
 

f_check

@@ -19,7 +19,7 @@ $nofortran = 0;
 
 $compiler = join(" ", @ARGV);
 $compiler_bin = shift(@ARGV);
-
+ 
 # f77 is too ambiguous
 $compiler = "" if $compiler eq "f77";
 
@@ -130,6 +130,11 @@ if ($compiler eq "") {
 	if ($data =~ / zho_ge__/) {
 	    $need2bu       = 1;
 	}
+	if ($vendor =~ /G95/) {
+    	  if ($ENV{NO_LAPACKE} != 1) {
+		$need2bu = "";
+	  }
+	}
     }
 
     if ($vendor eq "") {
@@ -277,6 +282,8 @@ $linker_a = "";
 if ($link ne "") {
 
     $link =~ s/\-Y\sP\,/\-Y/g;
+    
+    $link =~ s/\-R\s*/\-rpath\@/g;
 
     $link =~ s/\-rpath\s+/\-rpath\@/g;
 

getarch.c

@@ -82,7 +82,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #ifdef OS_WINDOWS
 #include <windows.h>
 #endif
-#if defined(__FreeBSD__) || defined(__OpenBSD__) || defined(__DragonFly__) || defined(__APPLE__)
+#if defined(__FreeBSD__) || defined(__OpenBSD__) || defined(__NetBSD__) || defined(__DragonFly__) || defined(__APPLE__)
 #include <sys/types.h>
 #include <sys/sysctl.h>
 #endif
@@ -1201,7 +1201,7 @@ static int get_num_cores(void) {
 
 #ifdef OS_WINDOWS
   SYSTEM_INFO sysinfo;
-#elif defined(__FreeBSD__) || defined(__OpenBSD__) || defined(__DragonFly__) || defined(__APPLE__)
+#elif defined(__FreeBSD__) || defined(__OpenBSD__) || defined(__NetBSD__) || defined(__DragonFly__) || defined(__APPLE__)
   int m[2], count;
   size_t len;
 #endif
@@ -1215,7 +1215,7 @@ static int get_num_cores(void) {
   GetSystemInfo(&sysinfo);
   return sysinfo.dwNumberOfProcessors;
 
-#elif defined(__FreeBSD__) || defined(__OpenBSD__) || defined(__DragonFly__) || defined(__APPLE__)
+#elif defined(__FreeBSD__) || defined(__OpenBSD__) || defined(__NetBSD__) || defined(__DragonFly__) || defined(__APPLE__)
   m[0] = CTL_HW;
   m[1] = HW_NCPU;
   len = sizeof(int);