From b360f7ae3c09b227f71407757efe012dda733745 Mon Sep 17 00:00:00 2001
From: Andreas Marek <amarek@rzg.mpg.de>
Date: Mon, 3 Mar 2014 18:21:04 +0100
Subject: [PATCH] ELPA_2013.11.007: OpenMP for complex block2 kernel

By mistake the complex kernel with blocking 2 was not called
in parallel if OpenMP was used.
---
 ELPA_2013.11/configure     | 20 ++++++++++----------
 ELPA_2013.11/configure.ac  |  2 +-
 ELPA_2013.11/src/elpa2.F90 | 19 ++++++++++++++++++-
 3 files changed, 29 insertions(+), 12 deletions(-)

diff --git a/ELPA_2013.11/configure b/ELPA_2013.11/configure
index ec508849..f921bbb8 100755
--- a/ELPA_2013.11/configure
+++ b/ELPA_2013.11/configure
@@ -1,6 +1,6 @@
 #! /bin/sh
 # Guess values for system-dependent variables and create Makefiles.
-# Generated by GNU Autoconf 2.69 for elpa 2013.11.006.
+# Generated by GNU Autoconf 2.69 for elpa 2013.11.007.
 #
 # 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='2013.11.006'
-PACKAGE_STRING='elpa 2013.11.006'
+PACKAGE_VERSION='2013.11.007'
+PACKAGE_STRING='elpa 2013.11.007'
 PACKAGE_BUGREPORT='elpa-library@rzg.mpg.de'
 PACKAGE_URL=''
 
@@ -1390,7 +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 2013.11.006 to adapt to many kinds of systems.
+\`configure' configures elpa 2013.11.007 to adapt to many kinds of systems.
 
 Usage: $0 [OPTION]... [VAR=VALUE]...
 
@@ -1460,7 +1460,7 @@ fi
 
 if test -n "$ac_init_help"; then
   case $ac_init_help in
-     short | recursive ) echo "Configuration of elpa 2013.11.006:";;
+     short | recursive ) echo "Configuration of elpa 2013.11.007:";;
    esac
   cat <<\_ACEOF
 
@@ -1599,7 +1599,7 @@ fi
 test -n "$ac_init_help" && exit $ac_status
 if $ac_init_version; then
   cat <<\_ACEOF
-elpa configure 2013.11.006
+elpa configure 2013.11.007
 generated by GNU Autoconf 2.69
 
 Copyright (C) 2012 Free Software Foundation, Inc.
@@ -2082,7 +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 2013.11.006, which was
+It was created by elpa $as_me 2013.11.007, which was
 generated by GNU Autoconf 2.69.  Invocation command line was
 
   $ $0 $@
@@ -2947,7 +2947,7 @@ fi
 
 # Define the identity of the package.
  PACKAGE='elpa'
- VERSION='2013.11.006'
+ VERSION='2013.11.007'
 
 
 cat >>confdefs.h <<_ACEOF
@@ -20783,7 +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 2013.11.006, which was
+This file was extended by elpa $as_me 2013.11.007, which was
 generated by GNU Autoconf 2.69.  Invocation command line was
 
   CONFIG_FILES    = $CONFIG_FILES
@@ -20849,7 +20849,7 @@ _ACEOF
 cat >>$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 2013.11.006
+elpa config.status 2013.11.007
 configured by $0, generated by GNU Autoconf 2.69,
   with options \\"\$ac_cs_config\\"
 
diff --git a/ELPA_2013.11/configure.ac b/ELPA_2013.11/configure.ac
index e812f6ec..45b13861 100644
--- a/ELPA_2013.11/configure.ac
+++ b/ELPA_2013.11/configure.ac
@@ -1,5 +1,5 @@
 AC_PREREQ([2.69])
-AC_INIT([elpa],[2013.11.006], elpa-library@rzg.mpg.de)
+AC_INIT([elpa],[2013.11.007], elpa-library@rzg.mpg.de)
 AC_CONFIG_SRCDIR([src/elpa1.F90])
 
 AM_INIT_AUTOMAKE([foreign -Wall subdir-objects])
diff --git a/ELPA_2013.11/src/elpa2.F90 b/ELPA_2013.11/src/elpa2.F90
index 9aa3dd2f..32f703b8 100644
--- a/ELPA_2013.11/src/elpa2.F90
+++ b/ELPA_2013.11/src/elpa2.F90
@@ -4580,14 +4580,31 @@ contains
         complex*16 w(nbw,2)
 #if defined(WITH_AVX_COMPLEX_BLOCK2)
         ttt = mpi_wtime()
+#ifdef WITH_OPENMP
+        if(istripe<stripe_count) then
+          nl = stripe_width
+        else
+          noff = (my_thread-1)*thread_width + (istripe-1)*stripe_width
+          nl = min(my_thread*thread_width-noff, l_nev-noff)
+          if(nl<=0) return
+        endif
+#else
         nl = merge(stripe_width, last_stripe_width, istripe<stripe_count)
+#endif
         do j = ncols, 2, -2
             w(:,1) = bcast_buffer(1:nbw,j+off)
             w(:,2) = bcast_buffer(1:nbw,j+off-1)
+#ifdef WITH_OPENMP
+            call double_hh_trafo_complex(a(1,j+off+a_off-1,istripe,my_thread), w, nbw, nl, stripe_width, nbw)
+#else
             call double_hh_trafo_complex(a(1,j+off+a_off-1,istripe), w, nbw, nl, stripe_width, nbw)
+#endif
         enddo
+#ifdef WITH_OPENMP
+        if(j==1) call single_hh_trafo_complex(a(1,1+off+a_off,istripe,my_thread),bcast_buffer(1,off+1), nbw, nl, stripe_width,my_thread)
+#else
         if(j==1) call single_hh_trafo_complex(a(1,1+off+a_off,istripe),bcast_buffer(1,off+1), nbw, nl, stripe_width)
-
+#endif
 
 #endif
 
-- 
GitLab