ELPA_2013.11.006 bugfix for kernels real blocking 6 and 4

Due to an error in a preprocessor statement, the results for
real matrices were wrong if the kernels "avx-real-block6" or
"avx-real-block4" were chosen. No other kernels are affected.

The test programms always correctly stated that the results for
these kernels are wrong.

ELPA_2013.11/Makefile.am

@@ -120,17 +120,23 @@ endif
 
 if WITH_AVX_REAL_BLOCK4
 if WITH_OPENMP
-  libelpa_mt_la_SOURCES += src/elpa2_kernels/elpa2_kernels_real_sse-avx_4hv.c
+  libelpa_mt_la_SOURCES += src/elpa2_kernels/elpa2_kernels_real_sse-avx_4hv.c \
+                           src/elpa2_kernels/elpa2_kernels_real_sse-avx_2hv.c
 else
-  libelpa_la_SOURCES += src/elpa2_kernels/elpa2_kernels_real_sse-avx_4hv.c
+  libelpa_la_SOURCES += src/elpa2_kernels/elpa2_kernels_real_sse-avx_4hv.c \
+                        src/elpa2_kernels/elpa2_kernels_real_sse-avx_2hv.c
 endif
 endif
 
 if WITH_AVX_REAL_BLOCK6
 if WITH_OPENMP
-  libelpa_mt_la_SOURCES += src/elpa2_kernels/elpa2_kernels_real_sse-avx_6hv.c
+  libelpa_mt_la_SOURCES += src/elpa2_kernels/elpa2_kernels_real_sse-avx_6hv.c \
+                           src/elpa2_kernels/elpa2_kernels_real_sse-avx_4hv.c \
+                           src/elpa2_kernels/elpa2_kernels_real_sse-avx_2hv.c
 else
-  libelpa_la_SOURCES += src/elpa2_kernels/elpa2_kernels_real_sse-avx_6hv.c
+  libelpa_la_SOURCES += src/elpa2_kernels/elpa2_kernels_real_sse-avx_6hv.c \
+                        src/elpa2_kernels/elpa2_kernels_real_sse-avx_4hv.c \
+                        src/elpa2_kernels/elpa2_kernels_real_sse-avx_2hv.c
 endif
 endif
 

ELPA_2013.11/Makefile.in

@@ -131,10 +131,20 @@ host_triplet = @host@
 
 @WITH_AVX_REAL_BLOCK2_TRUE@@WITH_OPENMP_TRUE@am__append_19 = src/elpa2_kernels/elpa2_kernels_real_sse-avx_2hv.c
 @WITH_AVX_REAL_BLOCK2_TRUE@@WITH_OPENMP_FALSE@am__append_20 = src/elpa2_kernels/elpa2_kernels_real_sse-avx_2hv.c
-@WITH_AVX_REAL_BLOCK4_TRUE@@WITH_OPENMP_TRUE@am__append_21 = src/elpa2_kernels/elpa2_kernels_real_sse-avx_4hv.c
-@WITH_AVX_REAL_BLOCK4_TRUE@@WITH_OPENMP_FALSE@am__append_22 = src/elpa2_kernels/elpa2_kernels_real_sse-avx_4hv.c
-@WITH_AVX_REAL_BLOCK6_TRUE@@WITH_OPENMP_TRUE@am__append_23 = src/elpa2_kernels/elpa2_kernels_real_sse-avx_6hv.c
-@WITH_AVX_REAL_BLOCK6_TRUE@@WITH_OPENMP_FALSE@am__append_24 = src/elpa2_kernels/elpa2_kernels_real_sse-avx_6hv.c
+@WITH_AVX_REAL_BLOCK4_TRUE@@WITH_OPENMP_TRUE@am__append_21 = src/elpa2_kernels/elpa2_kernels_real_sse-avx_4hv.c \
+@WITH_AVX_REAL_BLOCK4_TRUE@@WITH_OPENMP_TRUE@                           src/elpa2_kernels/elpa2_kernels_real_sse-avx_2hv.c
+
+@WITH_AVX_REAL_BLOCK4_TRUE@@WITH_OPENMP_FALSE@am__append_22 = src/elpa2_kernels/elpa2_kernels_real_sse-avx_4hv.c \
+@WITH_AVX_REAL_BLOCK4_TRUE@@WITH_OPENMP_FALSE@                        src/elpa2_kernels/elpa2_kernels_real_sse-avx_2hv.c
+
+@WITH_AVX_REAL_BLOCK6_TRUE@@WITH_OPENMP_TRUE@am__append_23 = src/elpa2_kernels/elpa2_kernels_real_sse-avx_6hv.c \
+@WITH_AVX_REAL_BLOCK6_TRUE@@WITH_OPENMP_TRUE@                           src/elpa2_kernels/elpa2_kernels_real_sse-avx_4hv.c \
+@WITH_AVX_REAL_BLOCK6_TRUE@@WITH_OPENMP_TRUE@                           src/elpa2_kernels/elpa2_kernels_real_sse-avx_2hv.c
+
+@WITH_AVX_REAL_BLOCK6_TRUE@@WITH_OPENMP_FALSE@am__append_24 = src/elpa2_kernels/elpa2_kernels_real_sse-avx_6hv.c \
+@WITH_AVX_REAL_BLOCK6_TRUE@@WITH_OPENMP_FALSE@                        src/elpa2_kernels/elpa2_kernels_real_sse-avx_4hv.c \
+@WITH_AVX_REAL_BLOCK6_TRUE@@WITH_OPENMP_FALSE@                        src/elpa2_kernels/elpa2_kernels_real_sse-avx_2hv.c
+
 bin_PROGRAMS = test_real$(EXEEXT) test_real2$(EXEEXT) \
 	test_complex$(EXEEXT) test_complex2$(EXEEXT)
 subdir = .
@@ -146,9 +156,9 @@ DIST_COMMON = $(srcdir)/Makefile.in $(srcdir)/Makefile.am \
 	ltmain.sh
 ACLOCAL_M4 = $(top_srcdir)/aclocal.m4
 am__aclocal_m4_deps = $(top_srcdir)/m4/ax_check_gnu_make.m4 \
-	$(top_srcdir)/m4/libtool.m4 $(top_srcdir)/m4/ltoptions.m4 \
-	$(top_srcdir)/m4/ltsugar.m4 $(top_srcdir)/m4/ltversion.m4 \
-	$(top_srcdir)/m4/lt~obsolete.m4 \
+	$(top_srcdir)/m4/ax_elpa_openmp.m4 $(top_srcdir)/m4/libtool.m4 \
+	$(top_srcdir)/m4/ltoptions.m4 $(top_srcdir)/m4/ltsugar.m4 \
+	$(top_srcdir)/m4/ltversion.m4 $(top_srcdir)/m4/lt~obsolete.m4 \
 	$(top_srcdir)/fdep/fortran_dependencies.m4 \
 	$(top_srcdir)/m4/ax_prog_fc_mpi.m4 $(top_srcdir)/configure.ac
 am__configure_deps = $(am__aclocal_m4_deps) $(CONFIGURE_DEPENDENCIES) \
@@ -223,8 +233,11 @@ am__dirstamp = $(am__leading_dot)dirstamp
 @WITH_AVX_COMPLEX_BLOCK2_TRUE@@WITH_OPENMP_FALSE@am__objects_9 = src/elpa2_kernels/elpa2_kernels_complex_sse-avx_2hv.lo \
 @WITH_AVX_COMPLEX_BLOCK2_TRUE@@WITH_OPENMP_FALSE@	src/elpa2_kernels/elpa2_kernels_complex_sse-avx_1hv.lo
 @WITH_AVX_REAL_BLOCK2_TRUE@@WITH_OPENMP_FALSE@am__objects_10 = src/elpa2_kernels/elpa2_kernels_real_sse-avx_2hv.lo
-@WITH_AVX_REAL_BLOCK4_TRUE@@WITH_OPENMP_FALSE@am__objects_11 = src/elpa2_kernels/elpa2_kernels_real_sse-avx_4hv.lo
-@WITH_AVX_REAL_BLOCK6_TRUE@@WITH_OPENMP_FALSE@am__objects_12 = src/elpa2_kernels/elpa2_kernels_real_sse-avx_6hv.lo
+@WITH_AVX_REAL_BLOCK4_TRUE@@WITH_OPENMP_FALSE@am__objects_11 = src/elpa2_kernels/elpa2_kernels_real_sse-avx_4hv.lo \
+@WITH_AVX_REAL_BLOCK4_TRUE@@WITH_OPENMP_FALSE@	src/elpa2_kernels/elpa2_kernels_real_sse-avx_2hv.lo
+@WITH_AVX_REAL_BLOCK6_TRUE@@WITH_OPENMP_FALSE@am__objects_12 = src/elpa2_kernels/elpa2_kernels_real_sse-avx_6hv.lo \
+@WITH_AVX_REAL_BLOCK6_TRUE@@WITH_OPENMP_FALSE@	src/elpa2_kernels/elpa2_kernels_real_sse-avx_4hv.lo \
+@WITH_AVX_REAL_BLOCK6_TRUE@@WITH_OPENMP_FALSE@	src/elpa2_kernels/elpa2_kernels_real_sse-avx_2hv.lo
 @WITH_OPENMP_FALSE@am_libelpa_la_OBJECTS = src/elpa1.lo src/elpa2.lo \
 @WITH_OPENMP_FALSE@	$(am__objects_1) $(am__objects_2) \
 @WITH_OPENMP_FALSE@	$(am__objects_3) $(am__objects_4) \
@@ -273,8 +286,11 @@ am__libelpa_mt_la_SOURCES_DIST = src/elpa1.F90 src/elpa2.F90 \
 @WITH_AVX_COMPLEX_BLOCK2_TRUE@@WITH_OPENMP_TRUE@am__objects_21 = src/elpa2_kernels/elpa2_kernels_complex_sse-avx_2hv.lo \
 @WITH_AVX_COMPLEX_BLOCK2_TRUE@@WITH_OPENMP_TRUE@	src/elpa2_kernels/elpa2_kernels_complex_sse-avx_1hv.lo
 @WITH_AVX_REAL_BLOCK2_TRUE@@WITH_OPENMP_TRUE@am__objects_22 = src/elpa2_kernels/elpa2_kernels_real_sse-avx_2hv.lo
-@WITH_AVX_REAL_BLOCK4_TRUE@@WITH_OPENMP_TRUE@am__objects_23 = src/elpa2_kernels/elpa2_kernels_real_sse-avx_4hv.lo
-@WITH_AVX_REAL_BLOCK6_TRUE@@WITH_OPENMP_TRUE@am__objects_24 = src/elpa2_kernels/elpa2_kernels_real_sse-avx_6hv.lo
+@WITH_AVX_REAL_BLOCK4_TRUE@@WITH_OPENMP_TRUE@am__objects_23 = src/elpa2_kernels/elpa2_kernels_real_sse-avx_4hv.lo \
+@WITH_AVX_REAL_BLOCK4_TRUE@@WITH_OPENMP_TRUE@	src/elpa2_kernels/elpa2_kernels_real_sse-avx_2hv.lo
+@WITH_AVX_REAL_BLOCK6_TRUE@@WITH_OPENMP_TRUE@am__objects_24 = src/elpa2_kernels/elpa2_kernels_real_sse-avx_6hv.lo \
+@WITH_AVX_REAL_BLOCK6_TRUE@@WITH_OPENMP_TRUE@	src/elpa2_kernels/elpa2_kernels_real_sse-avx_4hv.lo \
+@WITH_AVX_REAL_BLOCK6_TRUE@@WITH_OPENMP_TRUE@	src/elpa2_kernels/elpa2_kernels_real_sse-avx_2hv.lo
 @WITH_OPENMP_TRUE@am_libelpa_mt_la_OBJECTS = src/elpa1.lo src/elpa2.lo \
 @WITH_OPENMP_TRUE@	$(am__objects_13) $(am__objects_14) \
 @WITH_OPENMP_TRUE@	$(am__objects_15) $(am__objects_16) \
@@ -679,6 +695,7 @@ NM = @NM@
 NMEDIT = @NMEDIT@
 OBJDUMP = @OBJDUMP@
 OBJEXT = @OBJEXT@
+OPENMP_FCFLAGS = @OPENMP_FCFLAGS@
 OTOOL = @OTOOL@
 OTOOL64 = @OTOOL64@
 PACKAGE = @PACKAGE@

ELPA_2013.11/aclocal.m4

@@ -1149,6 +1149,7 @@ AC_SUBST([am__untar])
 ]) # _AM_PROG_TAR
 
 m4_include([m4/ax_check_gnu_make.m4])
+m4_include([m4/ax_elpa_openmp.m4])
 m4_include([m4/libtool.m4])
 m4_include([m4/ltoptions.m4])
 m4_include([m4/ltsugar.m4])

ELPA_2013.11/configure

 #! /bin/sh
 # Guess values for system-dependent variables and create Makefiles.
-# Generated by GNU Autoconf 2.69 for elpa 2013.11.005.
+# Generated by GNU Autoconf 2.69 for elpa 2013.11.006.
 #
 # Report bugs to <elpa-library@rzg.mpg.de>.
 #
@@ -590,8 +590,8 @@ MAKEFLAGS=
 # Identity of this package.
 PACKAGE_NAME='elpa'
 PACKAGE_TARNAME='elpa'
-PACKAGE_VERSION='2013.11.005'
-PACKAGE_STRING='elpa 2013.11.005'
+PACKAGE_VERSION='2013.11.006'
+PACKAGE_STRING='elpa 2013.11.006'
 PACKAGE_BUGREPORT='elpa-library@rzg.mpg.de'
 PACKAGE_URL=''
 
@@ -672,6 +672,7 @@ build_vendor
 build_cpu
 build
 LIBTOOL
+OPENMP_FCFLAGS
 WITH_OPENMP_FALSE
 WITH_OPENMP_TRUE
 FC_MODOUT
@@ -823,6 +824,7 @@ with_avx_real_block4
 with_avx_real_block6
 with_avx_optimization
 with_openmp
+enable_openmp
 enable_shared
 enable_static
 with_pic
@@ -1388,7 +1390,7 @@ if test "$ac_init_help" = "long"; then
   # Omit some internal or obsolete options to make the list less imposing.
   # This message is too long to be a string in the A/UX 3.1 sh.
   cat <<_ACEOF
-\`configure' configures elpa 2013.11.005 to adapt to many kinds of systems.
+\`configure' configures elpa 2013.11.006 to adapt to many kinds of systems.
 
 Usage: $0 [OPTION]... [VAR=VALUE]...
 
@@ -1458,7 +1460,7 @@ fi
 
 if test -n "$ac_init_help"; then
   case $ac_init_help in
-     short | recursive ) echo "Configuration of elpa 2013.11.005:";;
+     short | recursive ) echo "Configuration of elpa 2013.11.006:";;
    esac
   cat <<\_ACEOF
 
@@ -1472,6 +1474,7 @@ Optional Features:
                           do not reject slow dependency extractors
   --disable-dependency-tracking
                           speeds up one-time build
+  --disable-openmp        do not use OpenMP
   --enable-shared[=PKGS]  build shared libraries [default=yes]
   --enable-static[=PKGS]  build static libraries [default=yes]
   --enable-fast-install[=PKGS]
@@ -1596,7 +1599,7 @@ fi
 test -n "$ac_init_help" && exit $ac_status
 if $ac_init_version; then
   cat <<\_ACEOF
-elpa configure 2013.11.005
+elpa configure 2013.11.006
 generated by GNU Autoconf 2.69
 
 Copyright (C) 2012 Free Software Foundation, Inc.
@@ -2079,7 +2082,7 @@ cat >config.log <<_ACEOF
 This file contains any messages produced by compilers while
 running configure, to aid debugging if configure makes a mistake.
 
-It was created by elpa $as_me 2013.11.005, which was
+It was created by elpa $as_me 2013.11.006, which was
 generated by GNU Autoconf 2.69.  Invocation command line was
 
   $ $0 $@
@@ -2944,7 +2947,7 @@ fi
 
 # Define the identity of the package.
  PACKAGE='elpa'
- VERSION='2013.11.005'
+ VERSION='2013.11.006'
 
 
 cat >>confdefs.h <<_ACEOF
@@ -5911,7 +5914,7 @@ fi
 
 
 
-ELPA_LIB_VERSION=2013.11.005
+ELPA_LIB_VERSION=2013.11.006
 
 # this is the version of the API, should be changed in the major revision
 # if and only if the actual API changes
@@ -6144,7 +6147,61 @@ fi
 
 $as_echo "#define WITH_OPENMP 1" >>confdefs.h
 
-        AX_ELPA_OPENMP
+
+  OPENMP_FCFLAGS=
+  # Check whether --enable-openmp was given.
+if test "${enable_openmp+set}" = set; then :
+  enableval=$enable_openmp;
+fi
+
+  if test "$enable_openmp" != no; then
+    { $as_echo "$as_me:${as_lineno-$LINENO}: checking for $CC option to support OpenMP" >&5
+$as_echo_n "checking for $CC option to support OpenMP... " >&6; }
+if ${ac_cv_prog_fc_openmp+:} false; then :
+  $as_echo_n "(cached) " >&6
+else
+  cat > conftest.$ac_ext <<_ACEOF
+      program main
+      call omp_get_num_threads
+      end
+_ACEOF
+if ac_fn_fc_try_link "$LINENO"; then :
+  ac_cv_prog_fc_openmp='none needed'
+else
+  ac_cv_prog_fc_openmp='unsupported'
+	  	  	  	  	  	  	  	  	  	  	  	  	  for ac_option in -openmp -fopenmp -xopenmp -mp -omp -qsmp=omp; do
+	    ac_save_FCFLAGS=$FCFLAGS
+	    FCFLAGS="$FCFLAGS $ac_option"
+	    cat > conftest.$ac_ext <<_ACEOF
+      program main
+      call omp_get_num_threads
+      end
+_ACEOF
+if ac_fn_fc_try_link "$LINENO"; then :
+  ac_cv_prog_fc_openmp=$ac_option
+fi
+rm -f core conftest.err conftest.$ac_objext \
+    conftest$ac_exeext conftest.$ac_ext
+	    FCFLAGS=$ac_save_FCFLAGS
+	    if test "$ac_cv_prog_fc_openmp" != unsupported; then
+	      break
+	    fi
+	  done
+fi
+rm -f core conftest.err conftest.$ac_objext \
+    conftest$ac_exeext conftest.$ac_ext
+fi
+{ $as_echo "$as_me:${as_lineno-$LINENO}: result: $ac_cv_prog_fc_openmp" >&5
+$as_echo "$ac_cv_prog_fc_openmp" >&6; }
+    case $ac_cv_prog_fc_openmp in #(
+      "none needed" | unsupported)
+        ;; #(
+      *)
+        OPENMP_FCFLAGS=$ac_cv_prog_fc_openmp ;;
+    esac
+  fi
+
+
   fi
 
 FCFLAGS="$FCFLAGS $OPENMP_FCFLAGS $OPENMP_FFFLAGS"
@@ -20726,7 +20783,7 @@ cat >>$CONFIG_STATUS <<\_ACEOF || ac_write_fail=1
 # report actual input values of CONFIG_FILES etc. instead of their
 # values after options handling.
 ac_log="
-This file was extended by elpa $as_me 2013.11.005, which was
+This file was extended by elpa $as_me 2013.11.006, which was
 generated by GNU Autoconf 2.69.  Invocation command line was
 
   CONFIG_FILES    = $CONFIG_FILES
@@ -20792,7 +20849,7 @@ _ACEOF
 cat >>$CONFIG_STATUS <<_ACEOF || ac_write_fail=1
 ac_cs_config="`$as_echo "$ac_configure_args" | sed 's/^ //; s/[\\""\`\$]/\\\\&/g'`"
 ac_cs_version="\\
-elpa config.status 2013.11.005
+elpa config.status 2013.11.006
 configured by $0, generated by GNU Autoconf 2.69,
   with options \\"\$ac_cs_config\\"
 

ELPA_2013.11/configure.ac

 AC_PREREQ([2.69])
-AC_INIT([elpa],[2013.11.005], elpa-library@rzg.mpg.de)
+AC_INIT([elpa],[2013.11.006], elpa-library@rzg.mpg.de)
 AC_CONFIG_SRCDIR([src/elpa1.F90])
 
 AM_INIT_AUTOMAKE([foreign -Wall subdir-objects])
@@ -196,7 +196,7 @@ AX_PROG_FC_MPI([],[have_mpi=yes],[have_mpi=no
 		  fi])
 
 
-AC_SUBST([ELPA_LIB_VERSION], [2013.11.005])
+AC_SUBST([ELPA_LIB_VERSION], [2013.11.006])
 # this is the version of the API, should be changed in the major revision
 # if and only if the actual API changes
 AC_SUBST([ELPA_SO_VERSION], [0:0:0])

ELPA_2013.11/src/elpa2.F90

@@ -2431,7 +2431,7 @@ subroutine trans_ev_tridi_to_band_real(na, nev, nblk, nbw, q, ldq, mpi_comm_rows
 #endif
 #endif
 
-#if (defined(WITH_AVX_REAL_BLOCK4) && defined(WITH_AVX_REAL_BLOCK2)) || defined(WITH_AMD_BULLDOZER)
+#if defined(WITH_AVX_REAL_BLOCK4) || defined(WITH_AMD_BULLDOZER)
 
          ! X86 INTRINSIC CODE, USING 4 HOUSEHOLDER VECTORS
          do j = ncols, 4, -4
@@ -2470,7 +2470,7 @@ subroutine trans_ev_tridi_to_band_real(na, nev, nblk, nbw, q, ldq, mpi_comm_rows
 #endif
 
 
-#if (defined(WITH_AVX_REAL_BLOCK6) && defined(WITH_AVX_REAL_BLOCK4) && defined(WITH_AVX_REAL_BLOCK2))
+#if defined(WITH_AVX_REAL_BLOCK6)
          ! X86 INTRINSIC CODE, USING 6 HOUSEHOLDER VECTORS
          do j = ncols, 6, -6
              w(:,1) = bcast_buffer(1:nbw,j+off)

ELPA_2013.11/src/elpa2_kernels/README_elpa2_kernels.txt

@@ -112,12 +112,16 @@ Several
                                CFLAGS and CXXFLAGS automatically.
 
                                On Intel Sandybridge architectures the
-                               configure option "--with-intel-sandybride" 
-			       use the best combination.
+                               configure option "--with-avx-sandybride" 
+			       uses the best combination, which is a 
+                               combination of block2 for real matrices
+                               and block1 for complex matrices.
 
                                On AMD Bulldozer architectures the
                                configure option "--with-amd-bulldozer" 
-			       use the best combination.
+			       uses the best combination, which is a
+			       combination of block4 for real matrices 
+			       and block1 for complex matrices.
 
 			       Otherwise, you can try out your own
 			       combinations with the configure options