diff --git a/Makefile.in b/Makefile.in
index ea71c6bcd..81ee25c15 100644
--- a/Makefile.in
+++ b/Makefile.in
@@ -11,6 +11,7 @@ instbasedir     := $(DESTDIR)$(prefix)
 bmarkdir        := $(abs_top_src_dir)/benchmarks
 isa_src_dir     := $(abs_top_src_dir)/isa
 debug_src_dir   := $(abs_top_src_dir)/debug
+enable_multi_threads := @MULTI_THREADS@
 
 all: benchmarks isa
 
diff --git a/benchmarks/Makefile b/benchmarks/Makefile
index fb0d2977d..698d5c2ff 100644
--- a/benchmarks/Makefile
+++ b/benchmarks/Makefile
@@ -43,7 +43,7 @@ RISCV_LINK_OPTS ?= -static -nostdlib -nostartfiles -lm -lgcc -T $(src_dir)/commo
 RISCV_OBJDUMP ?= $(RISCV_PREFIX)objdump --disassemble-all --disassemble-zeroes --section=.text --section=.text.startup --section=.data
 RISCV_SIM ?= spike --isa=rv$(XLEN)gc
 
-incs  += -I$(src_dir)/../env -I$(src_dir)/common $(addprefix -I$(src_dir)/, $(bmarks))
+incs  += -I$(src_dir)/../env -I$(src_dir)/common -I$(src_dir)/../ $(addprefix -I$(src_dir)/, $(bmarks))
 objs  :=
 
 define compile_template
diff --git a/benchmarks/common/crt.S b/benchmarks/common/crt.S
index d75e81e06..10d6a9572 100644
--- a/benchmarks/common/crt.S
+++ b/benchmarks/common/crt.S
@@ -1,6 +1,7 @@
 # See LICENSE for license details.
 
 #include "encoding.h"
+#include "config.h"
 
 #if __riscv_xlen == 64
 # define LREG ld
diff --git a/benchmarks/common/syscalls.c b/benchmarks/common/syscalls.c
index 0a7d6b728..2c8c5a93a 100644
--- a/benchmarks/common/syscalls.c
+++ b/benchmarks/common/syscalls.c
@@ -95,13 +95,15 @@ int __attribute__((weak)) main(int argc, char** argv)
 
 static void init_tls()
 {
+#if ENABLE_THREADS
   register void* thread_pointer asm("tp");
   extern char _tls_data;
-  extern __thread char _tdata_begin, _tdata_end, _tbss_end;
+  extern THREAD_LOCAL char _tdata_begin, _tdata_end, _tbss_end;
   size_t tdata_size = &_tdata_end - &_tdata_begin;
   memcpy(thread_pointer, &_tls_data, tdata_size);
   size_t tbss_size = &_tbss_end - &_tdata_end;
   memset(thread_pointer + tdata_size, 0, tbss_size);
+#endif
 }
 
 void _init(int cid, int nc)
@@ -126,8 +128,8 @@ void _init(int cid, int nc)
 #undef putchar
 int putchar(int ch)
 {
-  static __thread char buf[64] __attribute__((aligned(64)));
-  static __thread int buflen = 0;
+  static THREAD_LOCAL char buf[64] __attribute__((aligned(64)));
+  static THREAD_LOCAL int buflen = 0;
 
   buf[buflen++] = ch;
 
diff --git a/benchmarks/common/util.h b/benchmarks/common/util.h
index 081cfd634..da894d436 100644
--- a/benchmarks/common/util.h
+++ b/benchmarks/common/util.h
@@ -6,9 +6,20 @@
 extern void setStats(int enable);
 
 #include <stdint.h>
+#include "config.h"
 
 #define static_assert(cond) switch(0) { case 0: case !!(long)(cond): ; }
 
+#ifndef ENABLE_THREADS
+#define ENABLE_THREADS 1
+#endif
+
+#if ENABLE_THREADS
+#define THREAD_LOCAL __thread
+#else
+#define THREAD_LOCAL
+#endif
+
 static int verify(int n, const volatile int* test, const int* verify)
 {
   int i;
@@ -45,7 +56,7 @@ static void __attribute__((noinline)) barrier(int ncores)
 {
   static volatile int sense;
   static volatile int count;
-  static __thread int threadsense;
+  static THREAD_LOCAL int threadsense;
 
   __sync_synchronize();
 
diff --git a/config.in b/config.in
new file mode 100644
index 000000000..74e9bb2e8
--- /dev/null
+++ b/config.in
@@ -0,0 +1,22 @@
+/* config.in.  Generated from configure.ac by autoheader.  */
+
+/* Define to 1 if you want to test with multi-thread. */
+#undef ENABLE_THREADS
+
+/* Define to the address where bug reports for this package should be sent. */
+#undef PACKAGE_BUGREPORT
+
+/* Define to the full name of this package. */
+#undef PACKAGE_NAME
+
+/* Define to the full name and version of this package. */
+#undef PACKAGE_STRING
+
+/* Define to the one symbol short name of this package. */
+#undef PACKAGE_TARNAME
+
+/* Define to the home page for this package. */
+#undef PACKAGE_URL
+
+/* Define to the version of this package. */
+#undef PACKAGE_VERSION
diff --git a/configure b/configure
index db0969e93..6464c3651 100755
--- a/configure
+++ b/configure
@@ -633,6 +633,7 @@ ac_subst_files=''
 ac_user_opts='
 enable_option_checking
 with_xlen
+enable_threads
 '
       ac_precious_vars='build_alias
 host_alias
@@ -1247,6 +1248,12 @@ if test -n "$ac_init_help"; then
    esac
   cat <<\_ACEOF
 
+Optional Features:
+  --disable-option-checking  ignore unrecognized --enable/--with options
+  --disable-FEATURE       do not include FEATURE (same as --enable-FEATURE=no)
+  --enable-FEATURE[=ARG]  include FEATURE [ARG=yes]
+  --enable-threads        enable test for multi-thread.
+
 Optional Packages:
   --with-PACKAGE[=ARG]    use PACKAGE [ARG=yes]
   --without-PACKAGE       do not use PACKAGE (same as --with-PACKAGE=no)
@@ -2534,6 +2541,27 @@ else
 fi
 
 
+ac_enable_threads=unset
+# Check whether --enable-threads was given.
+if test "${enable_threads+set}" = set; then :
+  enableval=$enable_threads; case "${enableval}" in
+  yes)  ac_enable_threads=1 ;;
+  no)   ac_enable_threads=0 ;;
+esac
+fi
+
+if test "${ac_enable_threads}" = unset; then
+  ac_enable_threads=1
+fi
+
+
+cat >>confdefs.h <<_ACEOF
+#define ENABLE_THREADS $ac_enable_threads
+_ACEOF
+
+
+ac_config_headers="$ac_config_headers config.h:config.in"
+
 
 ac_config_files="$ac_config_files Makefile"
 
@@ -2627,43 +2655,7 @@ test "x$prefix" = xNONE && prefix=$ac_default_prefix
 # Let make expand exec_prefix.
 test "x$exec_prefix" = xNONE && exec_prefix='${prefix}'
 
-# Transform confdefs.h into DEFS.
-# Protect against shell expansion while executing Makefile rules.
-# Protect against Makefile macro expansion.
-#
-# If the first sed substitution is executed (which looks for macros that
-# take arguments), then branch to the quote section.  Otherwise,
-# look for a macro that doesn't take arguments.
-ac_script='
-:mline
-/\\$/{
- N
- s,\\\n,,
- b mline
-}
-t clear
-:clear
-s/^[	 ]*#[	 ]*define[	 ][	 ]*\([^	 (][^	 (]*([^)]*)\)[	 ]*\(.*\)/-D\1=\2/g
-t quote
-s/^[	 ]*#[	 ]*define[	 ][	 ]*\([^	 ][^	 ]*\)[	 ]*\(.*\)/-D\1=\2/g
-t quote
-b any
-:quote
-s/[	 `~#$^&*(){}\\|;'\''"<>?]/\\&/g
-s/\[/\\&/g
-s/\]/\\&/g
-s/\$/$$/g
-H
-:any
-${
-	g
-	s/^\n//
-	s/\n/ /g
-	p
-}
-'
-DEFS=`sed -n "$ac_script" confdefs.h`
-
+DEFS=-DHAVE_CONFIG_H
 
 ac_libobjs=
 ac_ltlibobjs=
@@ -3097,11 +3089,15 @@ case $ac_config_files in *"
 "*) set x $ac_config_files; shift; ac_config_files=$*;;
 esac
 
+case $ac_config_headers in *"
+"*) set x $ac_config_headers; shift; ac_config_headers=$*;;
+esac
 
 
 cat >>$CONFIG_STATUS <<_ACEOF || ac_write_fail=1
 # Files that config.status was made for.
 config_files="$ac_config_files"
+config_headers="$ac_config_headers"
 
 _ACEOF
 
@@ -3122,10 +3118,15 @@ Usage: $0 [OPTION]... [TAG]...
       --recheck    update $as_me by reconfiguring in the same conditions
       --file=FILE[:TEMPLATE]
                    instantiate the configuration file FILE
+      --header=FILE[:TEMPLATE]
+                   instantiate the configuration header FILE
 
 Configuration files:
 $config_files
 
+Configuration headers:
+$config_headers
+
 Report bugs to the package provider."
 
 _ACEOF
@@ -3186,7 +3187,18 @@ do
     esac
     as_fn_append CONFIG_FILES " '$ac_optarg'"
     ac_need_defaults=false;;
-  --he | --h |  --help | --hel | -h )
+  --header | --heade | --head | --hea )
+    $ac_shift
+    case $ac_optarg in
+    *\'*) ac_optarg=`$as_echo "$ac_optarg" | sed "s/'/'\\\\\\\\''/g"` ;;
+    esac
+    as_fn_append CONFIG_HEADERS " '$ac_optarg'"
+    ac_need_defaults=false;;
+  --he | --h)
+    # Conflict between --help and --header
+    as_fn_error $? "ambiguous option: \`$1'
+Try \`$0 --help' for more information.";;
+  --help | --hel | -h )
     $as_echo "$ac_cs_usage"; exit ;;
   -q | -quiet | --quiet | --quie | --qui | --qu | --q \
   | -silent | --silent | --silen | --sile | --sil | --si | --s)
@@ -3242,6 +3254,7 @@ cat >>$CONFIG_STATUS <<\_ACEOF || ac_write_fail=1
 for ac_config_target in $ac_config_targets
 do
   case $ac_config_target in
+    "config.h") CONFIG_HEADERS="$CONFIG_HEADERS config.h:config.in" ;;
     "Makefile") CONFIG_FILES="$CONFIG_FILES Makefile" ;;
 
   *) as_fn_error $? "invalid argument: \`$ac_config_target'" "$LINENO" 5;;
@@ -3255,6 +3268,7 @@ done
 # bizarre bug on SunOS 4.1.3.
 if $ac_need_defaults; then
   test "${CONFIG_FILES+set}" = set || CONFIG_FILES=$config_files
+  test "${CONFIG_HEADERS+set}" = set || CONFIG_HEADERS=$config_headers
 fi
 
 # Have a temporary directory for convenience.  Make it in the build tree
@@ -3442,8 +3456,116 @@ fi
 cat >>$CONFIG_STATUS <<\_ACEOF || ac_write_fail=1
 fi # test -n "$CONFIG_FILES"
 
+# Set up the scripts for CONFIG_HEADERS section.
+# No need to generate them if there are no CONFIG_HEADERS.
+# This happens for instance with `./config.status Makefile'.
+if test -n "$CONFIG_HEADERS"; then
+cat >"$ac_tmp/defines.awk" <<\_ACAWK ||
+BEGIN {
+_ACEOF
+
+# Transform confdefs.h into an awk script `defines.awk', embedded as
+# here-document in config.status, that substitutes the proper values into
+# config.h.in to produce config.h.
 
-eval set X "  :F $CONFIG_FILES      "
+# Create a delimiter string that does not exist in confdefs.h, to ease
+# handling of long lines.
+ac_delim='%!_!# '
+for ac_last_try in false false :; do
+  ac_tt=`sed -n "/$ac_delim/p" confdefs.h`
+  if test -z "$ac_tt"; then
+    break
+  elif $ac_last_try; then
+    as_fn_error $? "could not make $CONFIG_HEADERS" "$LINENO" 5
+  else
+    ac_delim="$ac_delim!$ac_delim _$ac_delim!! "
+  fi
+done
+
+# For the awk script, D is an array of macro values keyed by name,
+# likewise P contains macro parameters if any.  Preserve backslash
+# newline sequences.
+
+ac_word_re=[_$as_cr_Letters][_$as_cr_alnum]*
+sed -n '
+s/.\{148\}/&'"$ac_delim"'/g
+t rset
+:rset
+s/^[	 ]*#[	 ]*define[	 ][	 ]*/ /
+t def
+d
+:def
+s/\\$//
+t bsnl
+s/["\\]/\\&/g
+s/^ \('"$ac_word_re"'\)\(([^()]*)\)[	 ]*\(.*\)/P["\1"]="\2"\
+D["\1"]=" \3"/p
+s/^ \('"$ac_word_re"'\)[	 ]*\(.*\)/D["\1"]=" \2"/p
+d
+:bsnl
+s/["\\]/\\&/g
+s/^ \('"$ac_word_re"'\)\(([^()]*)\)[	 ]*\(.*\)/P["\1"]="\2"\
+D["\1"]=" \3\\\\\\n"\\/p
+t cont
+s/^ \('"$ac_word_re"'\)[	 ]*\(.*\)/D["\1"]=" \2\\\\\\n"\\/p
+t cont
+d
+:cont
+n
+s/.\{148\}/&'"$ac_delim"'/g
+t clear
+:clear
+s/\\$//
+t bsnlc
+s/["\\]/\\&/g; s/^/"/; s/$/"/p
+d
+:bsnlc
+s/["\\]/\\&/g; s/^/"/; s/$/\\\\\\n"\\/p
+b cont
+' <confdefs.h | sed '
+s/'"$ac_delim"'/"\\\
+"/g' >>$CONFIG_STATUS || ac_write_fail=1
+
+cat >>$CONFIG_STATUS <<_ACEOF || ac_write_fail=1
+  for (key in D) D_is_set[key] = 1
+  FS = ""
+}
+/^[\t ]*#[\t ]*(define|undef)[\t ]+$ac_word_re([\t (]|\$)/ {
+  line = \$ 0
+  split(line, arg, " ")
+  if (arg[1] == "#") {
+    defundef = arg[2]
+    mac1 = arg[3]
+  } else {
+    defundef = substr(arg[1], 2)
+    mac1 = arg[2]
+  }
+  split(mac1, mac2, "(") #)
+  macro = mac2[1]
+  prefix = substr(line, 1, index(line, defundef) - 1)
+  if (D_is_set[macro]) {
+    # Preserve the white space surrounding the "#".
+    print prefix "define", macro P[macro] D[macro]
+    next
+  } else {
+    # Replace #undef with comments.  This is necessary, for example,
+    # in the case of _POSIX_SOURCE, which is predefined and required
+    # on some systems where configure will not decide to define it.
+    if (defundef == "undef") {
+      print "/*", prefix defundef, macro, "*/"
+      next
+    }
+  }
+}
+{ print }
+_ACAWK
+_ACEOF
+cat >>$CONFIG_STATUS <<\_ACEOF || ac_write_fail=1
+  as_fn_error $? "could not setup config headers machinery" "$LINENO" 5
+fi # test -n "$CONFIG_HEADERS"
+
+
+eval set X "  :F $CONFIG_FILES  :H $CONFIG_HEADERS    "
 shift
 for ac_tag
 do
@@ -3651,7 +3773,30 @@ which seems to be undefined.  Please make sure it is defined" >&2;}
   esac \
   || as_fn_error $? "could not create $ac_file" "$LINENO" 5
  ;;
-
+  :H)
+  #
+  # CONFIG_HEADER
+  #
+  if test x"$ac_file" != x-; then
+    {
+      $as_echo "/* $configure_input  */" \
+      && eval '$AWK -f "$ac_tmp/defines.awk"' "$ac_file_inputs"
+    } >"$ac_tmp/config.h" \
+      || as_fn_error $? "could not create $ac_file" "$LINENO" 5
+    if diff "$ac_file" "$ac_tmp/config.h" >/dev/null 2>&1; then
+      { $as_echo "$as_me:${as_lineno-$LINENO}: $ac_file is unchanged" >&5
+$as_echo "$as_me: $ac_file is unchanged" >&6;}
+    else
+      rm -f "$ac_file"
+      mv "$ac_tmp/config.h" "$ac_file" \
+	|| as_fn_error $? "could not create $ac_file" "$LINENO" 5
+    fi
+  else
+    $as_echo "/* $configure_input  */" \
+      && eval '$AWK -f "$ac_tmp/defines.awk"' "$ac_file_inputs" \
+      || as_fn_error $? "could not create -" "$LINENO" 5
+  fi
+ ;;
 
 
   esac
diff --git a/configure.ac b/configure.ac
index ebfd64979..9114632ae 100644
--- a/configure.ac
+++ b/configure.ac
@@ -10,6 +10,24 @@ AC_ARG_WITH(xlen,
         AC_SUBST(XLEN, 64)
         )
 
+ac_enable_threads=unset
+AC_ARG_ENABLE(threads,
+        AS_HELP_STRING([--enable-threads],
+        [enable test for multi-thread.]),
+[case "${enableval}" in
+  yes)  ac_enable_threads=1 ;;
+  no)   ac_enable_threads=0 ;;
+esac])dnl
+
+if test "${ac_enable_threads}" = unset; then
+  ac_enable_threads=1
+fi
+
+AC_DEFINE_UNQUOTED(ENABLE_THREADS,
+  $ac_enable_threads,
+  [Define to 1 if you want to test with multi-thread.])
+
+AC_CONFIG_HEADERS(config.h:config.in)
 
 AC_OUTPUT(
     Makefile
diff --git a/mt/av_matmul.c b/mt/av_matmul.c
index 0f0dbaed2..381e4f827 100644
--- a/mt/av_matmul.c
+++ b/mt/av_matmul.c
@@ -9,10 +9,10 @@ void __attribute__((noinline)) matmul(const int coreid, const int ncores, const
 	//-----------------------------------------------------------------version 2.16, optimize v2.15 get rid of tempb. MSI 83K.w/ test one 81K.
 
 	
-	static __thread data_t TempA[8];
-	static __thread data_t TempB[8];
-	static __thread data_t TempC[8];
-	static __thread int j,m,n;
+	static THREAD_LOCAL data_t TempA[8];
+	static THREAD_LOCAL data_t TempB[8];
+	static THREAD_LOCAL data_t TempC[8];
+	static THREAD_LOCAL int j,m,n;
 	
 			if(coreid == 1 || ncores == 1 )
 	{
diff --git a/mt/ay_matmul.c b/mt/ay_matmul.c
index 45fb19425..949285754 100644
--- a/mt/ay_matmul.c
+++ b/mt/ay_matmul.c
@@ -6,11 +6,11 @@
 void __attribute__((noinline)) matmul(const int coreid, const int ncores, const int lda,  const data_t A[], const data_t B[], data_t C[] )
 {
   if(coreid > 1) return; 
-  static __thread int i, j, k;
-  static __thread data_t tempA0, tempA1, tempA2, tempA3, tempA4, tempA5, tempA6, tempA7;
-  static __thread data_t tempC0, tempC1, tempC2, tempC3, tempC4, tempC5, tempC6, tempC7, tempC8, tempC9, tempC10, tempC11, tempC12, tempC13, tempC14, tempC15;
+  static THREAD_LOCAL int i, j, k;
+  static THREAD_LOCAL data_t tempA0, tempA1, tempA2, tempA3, tempA4, tempA5, tempA6, tempA7;
+  static THREAD_LOCAL data_t tempC0, tempC1, tempC2, tempC3, tempC4, tempC5, tempC6, tempC7, tempC8, tempC9, tempC10, tempC11, tempC12, tempC13, tempC14, tempC15;
 
-  static __thread int start, end, jStride, jToRow, jToCol;
+  static THREAD_LOCAL int start, end, jStride, jToRow, jToCol;
   
   start = coreid << 9;
   end = ((ncores == 1) ? 2 : (coreid+1) ) << 9;
diff --git a/mt/az_matmul.c b/mt/az_matmul.c
index 1668fb019..2679558bb 100644
--- a/mt/az_matmul.c
+++ b/mt/az_matmul.c
@@ -7,11 +7,11 @@ void __attribute__((noinline)) matmul(const int coreid, const int ncores, const
 {
    
     if(coreid > 1) return; 
-    static __thread int i, j, k;
-    static __thread data_t tempA0, tempA1, tempA2, tempA3, tempA4, tempA5, tempA6, tempA7;
-    static __thread data_t tempC0, tempC1, tempC2, tempC3, tempC4, tempC5, tempC6, tempC7; //tempC8, tempC9, tempC10, tempC11, tempC12, tempC13, tempC14, tempC15;
+    static THREAD_LOCAL int i, j, k;
+    static THREAD_LOCAL data_t tempA0, tempA1, tempA2, tempA3, tempA4, tempA5, tempA6, tempA7;
+    static THREAD_LOCAL data_t tempC0, tempC1, tempC2, tempC3, tempC4, tempC5, tempC6, tempC7; //tempC8, tempC9, tempC10, tempC11, tempC12, tempC13, tempC14, tempC15;
 
-    static __thread int start, end, jStride, jToRow, jToCol, iToRow;
+    static THREAD_LOCAL int start, end, jStride, jToRow, jToCol, iToRow;
 
     start = coreid << 9;
     end = ((ncores == 1) ? 2 : (coreid+1)) << 9;
diff --git a/mt/br_matmul.c b/mt/br_matmul.c
index 13263e4f9..b3f7361bc 100644
--- a/mt/br_matmul.c
+++ b/mt/br_matmul.c
@@ -17,7 +17,7 @@ void __attribute__((noinline)) matmul(const int coreid, const int ncores, const
   int jBLOCK = 32;
   int iBLOCK = 16;
   int kBLOCK = 32;
-  static __thread int tB[4096]; //__thread
+  static THREAD_LOCAL int tB[4096]; //THREAD_LOCAL
   int startInd = coreid*(lda/ncores);
   int endInd = (coreid+1)*(lda/ncores);
 
diff --git a/mt/cv_matmul.c b/mt/cv_matmul.c
index 2a0691940..5ec0984ad 100644
--- a/mt/cv_matmul.c
+++ b/mt/cv_matmul.c
@@ -7,9 +7,9 @@ void __attribute__((noinline)) matmul(const int coreid, const int ncores, const
 {
 	//----------------------------------------------------------------version 2.11 optmize j,use core 1 j from 0 to 15 MSI 98k i = j*lda
 	//----------------------------------------------------------------version 2.12 not use i = j *lda MSI  95k
-	static __thread data_t TempA[8];
-	static __thread data_t TempB[8];
-	static __thread int j,m,n,i,k;
+	static THREAD_LOCAL data_t TempA[8];
+	static THREAD_LOCAL data_t TempB[8];
+	static THREAD_LOCAL int j,m,n,i,k;
 	
 	if(coreid == 1 || ncores == 1)
 	{
diff --git a/mt/cy_matmul.c b/mt/cy_matmul.c
index b0713cd43..a758952ad 100644
--- a/mt/cy_matmul.c
+++ b/mt/cy_matmul.c
@@ -6,11 +6,11 @@
 void __attribute__((noinline)) matmul(const int coreid, const int ncores, const int lda,  const data_t A[], const data_t B[], data_t C[] )
 {
    if(coreid > 1) return; 
-  static __thread int i, j, k;
-  static __thread data_t tempA0, tempA1, tempA2, tempA3, tempA4, tempA5, tempA6, tempA7;
-  static __thread data_t tempC0, tempC1, tempC2, tempC3, tempC4, tempC5, tempC6, tempC7, tempC8, tempC9, tempC10, tempC11, tempC12, tempC13, tempC14, tempC15;
+  static THREAD_LOCAL int i, j, k;
+  static THREAD_LOCAL data_t tempA0, tempA1, tempA2, tempA3, tempA4, tempA5, tempA6, tempA7;
+  static THREAD_LOCAL data_t tempC0, tempC1, tempC2, tempC3, tempC4, tempC5, tempC6, tempC7, tempC8, tempC9, tempC10, tempC11, tempC12, tempC13, tempC14, tempC15;
 
-  static __thread int start, end, jStride, jToRow, jToCol;
+  static THREAD_LOCAL int start, end, jStride, jToRow, jToCol;
   static data_t A1[1024], B1[1024];;
   
   start = coreid << 9;
diff --git a/mt/dr_matmul.c b/mt/dr_matmul.c
index 8a3d57ce1..c59722d72 100644
--- a/mt/dr_matmul.c
+++ b/mt/dr_matmul.c
@@ -17,7 +17,7 @@ void __attribute__((noinline)) matmul(const int coreid, const int ncores, const
   int jBLOCK = 32;
   int iBLOCK = 16;
   int kBLOCK = 32;
-  static __thread int tB[4096]; //__thread
+  static THREAD_LOCAL int tB[4096]; //THREAD_LOCAL
   int startInd = coreid*(lda/ncores);
   int endInd = (coreid+1)*(lda/ncores);