Corrected some compilation errors (zolotarev.h) and SSE4 vsplat and conj to make cshift test pass.

Makefile.in

@@ -1,7 +1,7 @@
-# Makefile.in generated by automake 1.15 from Makefile.am.
+# Makefile.in generated by automake 1.14.1 from Makefile.am.
 # @configure_input@
 
-# Copyright (C) 1994-2014 Free Software Foundation, Inc.
+# Copyright (C) 1994-2013 Free Software Foundation, Inc.
 
 # This Makefile.in is free software; the Free Software Foundation
 # gives unlimited permission to copy and/or distribute it,
@@ -14,17 +14,7 @@
 
 @SET_MAKE@
 VPATH = @srcdir@
-am__is_gnu_make = { \
+am__is_gnu_make = test -n '$(MAKEFILE_LIST)' && test -n '$(MAKELEVEL)'
-  if test -z '$(MAKELEVEL)'; then \
-    false; \
-  elif test -n '$(MAKE_HOST)'; then \
-    true; \
-  elif test -n '$(MAKE_VERSION)' && test -n '$(CURDIR)'; then \
-    true; \
-  else \
-    false; \
-  fi; \
-}
 am__make_running_with_option = \
   case $${target_option-} in \
       ?) ;; \
@@ -86,12 +76,14 @@ NORMAL_UNINSTALL = :
 PRE_UNINSTALL = :
 POST_UNINSTALL = :
 subdir = .
+DIST_COMMON = INSTALL NEWS README AUTHORS ChangeLog \
+	$(srcdir)/Makefile.in $(srcdir)/Makefile.am \
+	$(top_srcdir)/configure $(am__configure_deps) COPYING TODO \
+	compile depcomp install-sh missing
 ACLOCAL_M4 = $(top_srcdir)/aclocal.m4
 am__aclocal_m4_deps = $(top_srcdir)/configure.ac
 am__configure_deps = $(am__aclocal_m4_deps) $(CONFIGURE_DEPENDENCIES) \
 	$(ACLOCAL_M4)
-DIST_COMMON = $(srcdir)/Makefile.am $(top_srcdir)/configure \
-	$(am__configure_deps) $(am__DIST_COMMON)
 am__CONFIG_DISTCLEAN_FILES = config.status config.cache config.log \
  configure.lineno config.status.lineno
 mkinstalldirs = $(install_sh) -d
@@ -154,8 +146,6 @@ ETAGS = etags
 CTAGS = ctags
 CSCOPE = cscope
 DIST_SUBDIRS = $(SUBDIRS)
-am__DIST_COMMON = $(srcdir)/Makefile.in AUTHORS COPYING ChangeLog \
-	INSTALL NEWS README TODO compile depcomp install-sh missing
 DISTFILES = $(DIST_COMMON) $(DIST_SOURCES) $(TEXINFOS) $(EXTRA_DIST)
 distdir = $(PACKAGE)-$(VERSION)
 top_distdir = $(distdir)
@@ -309,6 +299,7 @@ $(srcdir)/Makefile.in:  $(srcdir)/Makefile.am  $(am__configure_deps)
 	echo ' cd $(top_srcdir) && $(AUTOMAKE) --gnu Makefile'; \
 	$(am__cd) $(top_srcdir) && \
 	  $(AUTOMAKE) --gnu Makefile
+.PRECIOUS: Makefile
 Makefile: $(srcdir)/Makefile.in $(top_builddir)/config.status
 	@case '$?' in \
 	  *config.status*) \
@@ -515,15 +506,15 @@ dist-xz: distdir
 	$(am__post_remove_distdir)
 
 dist-tarZ: distdir
-	@echo WARNING: "Support for distribution archives compressed with" \
+	@echo WARNING: "Support for shar distribution archives is" \
-		       "legacy program 'compress' is deprecated." >&2
+	               "deprecated." >&2
 	@echo WARNING: "It will be removed altogether in Automake 2.0" >&2
 	tardir=$(distdir) && $(am__tar) | compress -c >$(distdir).tar.Z
 	$(am__post_remove_distdir)
 
 dist-shar: distdir
-	@echo WARNING: "Support for shar distribution archives is" \
+	@echo WARNING: "Support for distribution archives compressed with" \
-	               "deprecated." >&2
+		       "legacy program 'compress' is deprecated." >&2
 	@echo WARNING: "It will be removed altogether in Automake 2.0" >&2
 	shar $(distdir) | GZIP=$(GZIP_ENV) gzip -c >$(distdir).shar.gz
 	$(am__post_remove_distdir)
@@ -559,17 +550,17 @@ distcheck: dist
 	esac
 	chmod -R a-w $(distdir)
 	chmod u+w $(distdir)
-	mkdir $(distdir)/_build $(distdir)/_build/sub $(distdir)/_inst
+	mkdir $(distdir)/_build $(distdir)/_inst
 	chmod a-w $(distdir)
 	test -d $(distdir)/_build || exit 0; \
 	dc_install_base=`$(am__cd) $(distdir)/_inst && pwd | sed -e 's,^[^:\\/]:[\\/],/,'` \
 	  && dc_destdir="$${TMPDIR-/tmp}/am-dc-$$$$/" \
 	  && am__cwd=`pwd` \
-	  && $(am__cd) $(distdir)/_build/sub \
+	  && $(am__cd) $(distdir)/_build \
-	  && ../../configure \
+	  && ../configure \
 	    $(AM_DISTCHECK_CONFIGURE_FLAGS) \
 	    $(DISTCHECK_CONFIGURE_FLAGS) \
-	    --srcdir=../.. --prefix="$$dc_install_base" \
+	    --srcdir=.. --prefix="$$dc_install_base" \
 	  && $(MAKE) $(AM_MAKEFLAGS) \
 	  && $(MAKE) $(AM_MAKEFLAGS) dvi \
 	  && $(MAKE) $(AM_MAKEFLAGS) check \
@@ -743,8 +734,6 @@ uninstall-am:
 	maintainer-clean-generic mostlyclean mostlyclean-generic pdf \
 	pdf-am ps ps-am tags tags-am uninstall uninstall-am
 
-.PRECIOUS: Makefile
-
 
 # Tell versions [3.59,3.63) of GNU make to not export all variables.
 # Otherwise a system limit (for SysV at least) may be exceeded.

aclocal.m4

@@ -1,6 +1,6 @@
-# generated automatically by aclocal 1.15 -*- Autoconf -*-
+# generated automatically by aclocal 1.14.1 -*- Autoconf -*-
 
-# Copyright (C) 1996-2014 Free Software Foundation, Inc.
+# Copyright (C) 1996-2013 Free Software Foundation, Inc.
 
 # This file is free software; the Free Software Foundation
 # gives unlimited permission to copy and/or distribute it,
@@ -20,7 +20,7 @@ You have another version of autoconf.  It may work, but is not guaranteed to.
 If you have problems, you may need to regenerate the build system entirely.
 To do so, use the procedure documented by the package, typically 'autoreconf'.])])
 
-# Copyright (C) 2002-2014 Free Software Foundation, Inc.
+# Copyright (C) 2002-2013 Free Software Foundation, Inc.
 #
 # This file is free software; the Free Software Foundation
 # gives unlimited permission to copy and/or distribute it,
@@ -32,10 +32,10 @@ To do so, use the procedure documented by the package, typically 'autoreconf'.])
 # generated from the m4 files accompanying Automake X.Y.
 # (This private macro should not be called outside this file.)
 AC_DEFUN([AM_AUTOMAKE_VERSION],
-[am__api_version='1.15'
+[am__api_version='1.14'
 dnl Some users find AM_AUTOMAKE_VERSION and mistake it for a way to
 dnl require some minimum version.  Point them to the right macro.
-m4_if([$1], [1.15], [],
+m4_if([$1], [1.14.1], [],
       [AC_FATAL([Do not call $0, use AM_INIT_AUTOMAKE([$1]).])])dnl
 ])
 
@@ -51,14 +51,14 @@ m4_define([_AM_AUTOCONF_VERSION], [])
 # Call AM_AUTOMAKE_VERSION and AM_AUTOMAKE_VERSION so they can be traced.
 # This function is AC_REQUIREd by AM_INIT_AUTOMAKE.
 AC_DEFUN([AM_SET_CURRENT_AUTOMAKE_VERSION],
-[AM_AUTOMAKE_VERSION([1.15])dnl
+[AM_AUTOMAKE_VERSION([1.14.1])dnl
 m4_ifndef([AC_AUTOCONF_VERSION],
   [m4_copy([m4_PACKAGE_VERSION], [AC_AUTOCONF_VERSION])])dnl
 _AM_AUTOCONF_VERSION(m4_defn([AC_AUTOCONF_VERSION]))])
 
 # AM_AUX_DIR_EXPAND                                         -*- Autoconf -*-
 
-# Copyright (C) 2001-2014 Free Software Foundation, Inc.
+# Copyright (C) 2001-2013 Free Software Foundation, Inc.
 #
 # This file is free software; the Free Software Foundation
 # gives unlimited permission to copy and/or distribute it,
@@ -103,14 +103,15 @@ _AM_AUTOCONF_VERSION(m4_defn([AC_AUTOCONF_VERSION]))])
 # configured tree to be moved without reconfiguration.
 
 AC_DEFUN([AM_AUX_DIR_EXPAND],
-[AC_REQUIRE([AC_CONFIG_AUX_DIR_DEFAULT])dnl
+[dnl Rely on autoconf to set up CDPATH properly.
-# Expand $ac_aux_dir to an absolute path.
+AC_PREREQ([2.50])dnl
-am_aux_dir=`cd "$ac_aux_dir" && pwd`
+# expand $ac_aux_dir to an absolute path
+am_aux_dir=`cd $ac_aux_dir && pwd`
 ])
 
 # AM_CONDITIONAL                                            -*- Autoconf -*-
 
-# Copyright (C) 1997-2014 Free Software Foundation, Inc.
+# Copyright (C) 1997-2013 Free Software Foundation, Inc.
 #
 # This file is free software; the Free Software Foundation
 # gives unlimited permission to copy and/or distribute it,
@@ -141,7 +142,7 @@ AC_CONFIG_COMMANDS_PRE(
 Usually this means the macro was only invoked conditionally.]])
 fi])])
 
-# Copyright (C) 1999-2014 Free Software Foundation, Inc.
+# Copyright (C) 1999-2013 Free Software Foundation, Inc.
 #
 # This file is free software; the Free Software Foundation
 # gives unlimited permission to copy and/or distribute it,
@@ -332,7 +333,7 @@ _AM_SUBST_NOTMAKE([am__nodep])dnl
 
 # Generate code to set up dependency tracking.              -*- Autoconf -*-
 
-# Copyright (C) 1999-2014 Free Software Foundation, Inc.
+# Copyright (C) 1999-2013 Free Software Foundation, Inc.
 #
 # This file is free software; the Free Software Foundation
 # gives unlimited permission to copy and/or distribute it,
@@ -408,7 +409,7 @@ AC_DEFUN([AM_OUTPUT_DEPENDENCY_COMMANDS],
 
 # Do all the work for Automake.                             -*- Autoconf -*-
 
-# Copyright (C) 1996-2014 Free Software Foundation, Inc.
+# Copyright (C) 1996-2013 Free Software Foundation, Inc.
 #
 # This file is free software; the Free Software Foundation
 # gives unlimited permission to copy and/or distribute it,
@@ -498,8 +499,8 @@ AC_REQUIRE([AC_PROG_MKDIR_P])dnl
 # <http://lists.gnu.org/archive/html/automake/2012-07/msg00001.html>
 # <http://lists.gnu.org/archive/html/automake/2012-07/msg00014.html>
 AC_SUBST([mkdir_p], ['$(MKDIR_P)'])
-# We need awk for the "check" target (and possibly the TAP driver).  The
+# We need awk for the "check" target.  The system "awk" is bad on
-# system "awk" is bad on some platforms.
+# some platforms.
 AC_REQUIRE([AC_PROG_AWK])dnl
 AC_REQUIRE([AC_PROG_MAKE_SET])dnl
 AC_REQUIRE([AM_SET_LEADING_DOT])dnl
@@ -572,11 +573,7 @@ to "yes", and re-run configure.
 END
     AC_MSG_ERROR([Your 'rm' program is bad, sorry.])
   fi
-fi
+fi])
-dnl The trailing newline in this macro's definition is deliberate, for
-dnl backward compatibility and to allow trailing 'dnl'-style comments
-dnl after the AM_INIT_AUTOMAKE invocation. See automake bug#16841.
-])
 
 dnl Hook into '_AC_COMPILER_EXEEXT' early to learn its expansion.  Do not
 dnl add the conditional right here, as _AC_COMPILER_EXEEXT may be further
@@ -605,7 +602,7 @@ for _am_header in $config_headers :; do
 done
 echo "timestamp for $_am_arg" >`AS_DIRNAME(["$_am_arg"])`/stamp-h[]$_am_stamp_count])
 
-# Copyright (C) 2001-2014 Free Software Foundation, Inc.
+# Copyright (C) 2001-2013 Free Software Foundation, Inc.
 #
 # This file is free software; the Free Software Foundation
 # gives unlimited permission to copy and/or distribute it,
@@ -616,7 +613,7 @@ echo "timestamp for $_am_arg" >`AS_DIRNAME(["$_am_arg"])`/stamp-h[]$_am_stamp_co
 # Define $install_sh.
 AC_DEFUN([AM_PROG_INSTALL_SH],
 [AC_REQUIRE([AM_AUX_DIR_EXPAND])dnl
-if test x"${install_sh+set}" != xset; then
+if test x"${install_sh}" != xset; then
   case $am_aux_dir in
   *\ * | *\	*)
     install_sh="\${SHELL} '$am_aux_dir/install-sh'" ;;
@@ -626,7 +623,7 @@ if test x"${install_sh+set}" != xset; then
 fi
 AC_SUBST([install_sh])])
 
-# Copyright (C) 2003-2014 Free Software Foundation, Inc.
+# Copyright (C) 2003-2013 Free Software Foundation, Inc.
 #
 # This file is free software; the Free Software Foundation
 # gives unlimited permission to copy and/or distribute it,
@@ -647,7 +644,7 @@ AC_SUBST([am__leading_dot])])
 
 # Check to see how 'make' treats includes.	            -*- Autoconf -*-
 
-# Copyright (C) 2001-2014 Free Software Foundation, Inc.
+# Copyright (C) 2001-2013 Free Software Foundation, Inc.
 #
 # This file is free software; the Free Software Foundation
 # gives unlimited permission to copy and/or distribute it,
@@ -697,7 +694,7 @@ rm -f confinc confmf
 
 # Fake the existence of programs that GNU maintainers use.  -*- Autoconf -*-
 
-# Copyright (C) 1997-2014 Free Software Foundation, Inc.
+# Copyright (C) 1997-2013 Free Software Foundation, Inc.
 #
 # This file is free software; the Free Software Foundation
 # gives unlimited permission to copy and/or distribute it,
@@ -736,7 +733,7 @@ fi
 
 # Helper functions for option handling.                     -*- Autoconf -*-
 
-# Copyright (C) 2001-2014 Free Software Foundation, Inc.
+# Copyright (C) 2001-2013 Free Software Foundation, Inc.
 #
 # This file is free software; the Free Software Foundation
 # gives unlimited permission to copy and/or distribute it,
@@ -767,7 +764,7 @@ AC_DEFUN([_AM_IF_OPTION],
 
 # Check to make sure that the build environment is sane.    -*- Autoconf -*-
 
-# Copyright (C) 1996-2014 Free Software Foundation, Inc.
+# Copyright (C) 1996-2013 Free Software Foundation, Inc.
 #
 # This file is free software; the Free Software Foundation
 # gives unlimited permission to copy and/or distribute it,
@@ -848,7 +845,7 @@ AC_CONFIG_COMMANDS_PRE(
 rm -f conftest.file
 ])
 
-# Copyright (C) 2009-2014 Free Software Foundation, Inc.
+# Copyright (C) 2009-2013 Free Software Foundation, Inc.
 #
 # This file is free software; the Free Software Foundation
 # gives unlimited permission to copy and/or distribute it,
@@ -908,7 +905,7 @@ AC_SUBST([AM_BACKSLASH])dnl
 _AM_SUBST_NOTMAKE([AM_BACKSLASH])dnl
 ])
 
-# Copyright (C) 2001-2014 Free Software Foundation, Inc.
+# Copyright (C) 2001-2013 Free Software Foundation, Inc.
 #
 # This file is free software; the Free Software Foundation
 # gives unlimited permission to copy and/or distribute it,
@@ -936,7 +933,7 @@ fi
 INSTALL_STRIP_PROGRAM="\$(install_sh) -c -s"
 AC_SUBST([INSTALL_STRIP_PROGRAM])])
 
-# Copyright (C) 2006-2014 Free Software Foundation, Inc.
+# Copyright (C) 2006-2013 Free Software Foundation, Inc.
 #
 # This file is free software; the Free Software Foundation
 # gives unlimited permission to copy and/or distribute it,
@@ -955,7 +952,7 @@ AC_DEFUN([AM_SUBST_NOTMAKE], [_AM_SUBST_NOTMAKE($@)])
 
 # Check how to create a tarball.                            -*- Autoconf -*-
 
-# Copyright (C) 2004-2014 Free Software Foundation, Inc.
+# Copyright (C) 2004-2013 Free Software Foundation, Inc.
 #
 # This file is free software; the Free Software Foundation
 # gives unlimited permission to copy and/or distribute it,

configure

@@ -2309,7 +2309,7 @@ ac_link='$CC -o conftest$ac_exeext $CFLAGS $CPPFLAGS $LDFLAGS conftest.$ac_ext $
 ac_compiler_gnu=$ac_cv_c_compiler_gnu
 
 
-am__api_version='1.15'
+am__api_version='1.14'
 
 ac_aux_dir=
 for ac_dir in "$srcdir" "$srcdir/.." "$srcdir/../.."; do
@@ -2510,8 +2510,8 @@ test "$program_suffix" != NONE &&
 ac_script='s/[\\$]/&&/g;s/;s,x,x,$//'
 program_transform_name=`$as_echo "$program_transform_name" | sed "$ac_script"`
 
-# Expand $ac_aux_dir to an absolute path.
+# expand $ac_aux_dir to an absolute path
-am_aux_dir=`cd "$ac_aux_dir" && pwd`
+am_aux_dir=`cd $ac_aux_dir && pwd`
 
 if test x"${MISSING+set}" != xset; then
   case $am_aux_dir in
@@ -2530,7 +2530,7 @@ else
 $as_echo "$as_me: WARNING: 'missing' script is too old or missing" >&2;}
 fi
 
-if test x"${install_sh+set}" != xset; then
+if test x"${install_sh}" != xset; then
   case $am_aux_dir in
   *\ * | *\	*)
     install_sh="\${SHELL} '$am_aux_dir/install-sh'" ;;
@@ -2858,8 +2858,8 @@ MAKEINFO=${MAKEINFO-"${am_missing_run}makeinfo"}
 # <http://lists.gnu.org/archive/html/automake/2012-07/msg00014.html>
 mkdir_p='$(MKDIR_P)'
 
-# We need awk for the "check" target (and possibly the TAP driver).  The
+# We need awk for the "check" target.  The system "awk" is bad on
-# system "awk" is bad on some platforms.
+# some platforms.
 # Always define AMTAR for backward compatibility.  Yes, it's still used
 # in the wild :-(  We should find a proper way to deprecate it ...
 AMTAR='$${TAR-tar}'
@@ -2918,7 +2918,6 @@ END
 fi
 
 
-
 ac_config_headers="$ac_config_headers lib/Grid_config.h"
 
 

lib/Grid_config.h

@@ -0,0 +1,121 @@
+/* lib/Grid_config.h.  Generated from Grid_config.h.in by configure.  */
+/* lib/Grid_config.h.in.  Generated from configure.ac by autoheader.  */
+
+/* AVX */
+/* #undef AVX1 */
+
+/* AVX2 */
+/* #undef AVX2 */
+
+/* AVX512 */
+/* #undef AVX512 */
+
+/* GRID_COMMS_MPI */
+/* #undef GRID_COMMS_MPI */
+
+/* GRID_COMMS_NONE */
+#define GRID_COMMS_NONE 1
+
+/* Define to 1 if you have the declaration of `be64toh', and to 0 if you
+   don't. */
+#define HAVE_DECL_BE64TOH 1
+
+/* Define to 1 if you have the declaration of `ntohll', and to 0 if you don't.
+   */
+#define HAVE_DECL_NTOHLL 0
+
+/* Define to 1 if you have the <endian.h> header file. */
+#define HAVE_ENDIAN_H 1
+
+/* Define to 1 if you have the `gettimeofday' function. */
+#define HAVE_GETTIMEOFDAY 1
+
+/* Define to 1 if you have the <gmp.h> header file. */
+#define HAVE_GMP_H 1
+
+/* Define to 1 if you have the <inttypes.h> header file. */
+#define HAVE_INTTYPES_H 1
+
+/* Define to 1 if you have the <malloc.h> header file. */
+#define HAVE_MALLOC_H 1
+
+/* Define to 1 if you have the <malloc/malloc.h> header file. */
+/* #undef HAVE_MALLOC_MALLOC_H */
+
+/* Define to 1 if you have the <memory.h> header file. */
+#define HAVE_MEMORY_H 1
+
+/* Define to 1 if you have the <mm_malloc.h> header file. */
+#define HAVE_MM_MALLOC_H 1
+
+/* Define to 1 if you have the <stdint.h> header file. */
+#define HAVE_STDINT_H 1
+
+/* Define to 1 if you have the <stdlib.h> header file. */
+#define HAVE_STDLIB_H 1
+
+/* Define to 1 if you have the <strings.h> header file. */
+#define HAVE_STRINGS_H 1
+
+/* Define to 1 if you have the <string.h> header file. */
+#define HAVE_STRING_H 1
+
+/* Define to 1 if you have the <sys/stat.h> header file. */
+#define HAVE_SYS_STAT_H 1
+
+/* Define to 1 if you have the <sys/types.h> header file. */
+#define HAVE_SYS_TYPES_H 1
+
+/* Define to 1 if you have the <unistd.h> header file. */
+#define HAVE_UNISTD_H 1
+
+/* Name of package */
+#define PACKAGE "grid"
+
+/* Define to the address where bug reports for this package should be sent. */
+#define PACKAGE_BUGREPORT "paboyle@ph.ed.ac.uk"
+
+/* Define to the full name of this package. */
+#define PACKAGE_NAME "Grid"
+
+/* Define to the full name and version of this package. */
+#define PACKAGE_STRING "Grid 1.0"
+
+/* Define to the one symbol short name of this package. */
+#define PACKAGE_TARNAME "grid"
+
+/* Define to the home page for this package. */
+#define PACKAGE_URL ""
+
+/* Define to the version of this package. */
+#define PACKAGE_VERSION "1.0"
+
+/* SSE4 */
+#define SSE4 1
+
+/* Define to 1 if you have the ANSI C header files. */
+#define STDC_HEADERS 1
+
+/* Version number of package */
+#define VERSION "1.0"
+
+/* Define for Solaris 2.5.1 so the uint32_t typedef from <sys/synch.h>,
+   <pthread.h>, or <semaphore.h> is not used. If the typedef were allowed, the
+   #define below would cause a syntax error. */
+/* #undef _UINT32_T */
+
+/* Define for Solaris 2.5.1 so the uint64_t typedef from <sys/synch.h>,
+   <pthread.h>, or <semaphore.h> is not used. If the typedef were allowed, the
+   #define below would cause a syntax error. */
+/* #undef _UINT64_T */
+
+/* Define to `unsigned int' if <sys/types.h> does not define. */
+/* #undef size_t */
+
+/* Define to the type of an unsigned integer type of width exactly 32 bits if
+   such a type exists and the standard includes do not define it. */
+/* #undef uint32_t */
+
+/* Define to the type of an unsigned integer type of width exactly 64 bits if
+   such a type exists and the standard includes do not define it. */
+/* #undef uint64_t */

lib/algorithms/approx/Zolotarev.cc

@@ -1,5 +1,5 @@
 /* -*- Mode: C; comment-column: 22; fill-column: 79; compile-command: "gcc -o zolotarev zolotarev.c -ansi -pedantic -lm -DTEST"; -*- */
-#define VERSION Source Time-stamp: <19-OCT-2004 09:33:22.00 adk@MISSCONTRARY>
+#define VERSION Source Time-stamp: <2015-05-18 16:32:08 neo>
 
 /* These C routines evalute the optimal rational approximation to the signum
  * function for epsilon < |x| < 1 using Zolotarev's theorem.
@@ -26,13 +26,13 @@
 #define INTERNAL_PRECISION double
 #endif
 
-#include "zolotarev.h"
+#include "Zolotarev.h"
 #define ZOLOTAREV_INTERNAL
 #undef ZOLOTAREV_DATA
 #define ZOLOTAREV_DATA izd
 #undef ZPRECISION
 #define ZPRECISION INTERNAL_PRECISION
-#include "zolotarev.h"
+#include "Zolotarev.h"
 #undef ZOLOTAREV_INTERNAL
 
 /* The ANSI standard appears not to know what pi is */

lib/simd/Grid_vComplexD.h

@@ -345,6 +345,9 @@ friend inline void vstore(const vComplexD &ret, ComplexD *a){
 // REDUCE FIXME must be a cleaner implementation
        friend inline ComplexD Reduce(const vComplexD & in)
        { 
+#if defined SSE4
+         return ComplexD(in.v[0], in.v[1]); // inefficient
+#endif
 #if defined (AVX1) || defined(AVX2)
 	 //            return std::complex<double>(_mm256_mask_reduce_add_pd(0x55, in.v),_mm256_mask_reduce_add_pd(0xAA, in.v));
 	 __attribute__ ((aligned(32))) double c_[4];

lib/simd/Grid_vComplexF.h

@@ -219,7 +219,7 @@ namespace Grid {
             ret.v = _mm256_set_ps(b,a,b,a,b,a,b,a);
 #endif
 #ifdef SSE4
-            ret.v = _mm_set_ps(a,b,a,b);
+            ret.v = _mm_set_ps(b,a,b,a);
 #endif
 #ifdef AVX512
             ret.v = _mm512_set_ps(b,a,b,a,b,a,b,a,b,a,b,a,b,a,b,a);
@@ -354,9 +354,7 @@ namespace Grid {
 
 #endif
 #ifdef SSE4
-	    cvec tmp;
+	    ret.v = _mm_xor_ps(_mm_addsub_ps(ret.v,in.v), _mm_set1_ps(-0.f));
-	    tmp = _mm_addsub_ps(ret.v,_mm_shuffle_ps(in.v,in.v,_MM_SHUFFLE(2,3,0,1))); // ymm1 <- br,bi
-	    ret.v=_mm_shuffle_ps(tmp,tmp,_MM_SHUFFLE(2,3,0,1));
 #endif
 #ifdef AVX512
             ret.v = _mm512_mask_sub_ps(in.v,0xaaaa,ret.v,in.v); // Zero out 0+real 0-imag 

lib/simd/Grid_vector_types.h

@@ -0,0 +1,299 @@
+#ifndef GRID_VECTOR_TYPES
+#define GRID_VECTOR_TYPES
+
+
+namespace Grid {
+
+  // To take the floating point type of real/complex type
+  template <typename T> 
+    struct RealPart {
+      typedef T type;
+    };
+  template <typename T> 
+    struct RealPart< std::complex<T> >{
+    typedef T type;
+  };
+  ////////////////////////////////////////////////////////
+
+  // Check for complexity with type traits
+  template <typename T> 
+    struct is_complex : std::false_type {};
+  template < typename T > 
+    struct is_complex< std::complex<T> >: std::true_type {};
+  ////////////////////////////////////////////////////////
+
+
+  // Define the operation templates functors
+  template < class SIMD, class Operation > 
+    SIMD binary(SIMD src_1, SIMD src_2, Operation op){
+    return op(src_1, src_2);
+  }
+
+  template < class SIMD, class Operation > 
+    SIMD unary(SIMD src, Operation op){
+    return op(src);
+  }
+  ///////////////////////////////////////////////
+
+  /*
+    @brief Grid_simd class for the SIMD vector type operations
+
+   */
+  template < class Scalar_type, class Vector_type > 
+    class Grid_simd {
+    
+  public:
+    typedef typename RealPart < Scalar_type >::type Real; 
+    Vector_type v;
+    
+
+    static inline int Nsimd(void) { return sizeof(Vector_type)/sizeof(Scalar_type);}
+
+    // Constructors
+    Grid_simd & operator = ( Zero & z){
+      vzero(*this);
+      return (*this);
+    }
+    Grid_simd(){};
+    
+    
+    //Enable if complex type
+    template < class S = Scalar_type > 
+    Grid_simd(typename std::enable_if< is_complex < S >::value, S>::type a){
+      vsplat(*this,a);
+    };
+    
+
+    Grid_simd(Real a){
+      vsplat(*this,Scalar_type(a));
+    };
+
+
+
+       
+    ///////////////////////////////////////////////
+    // mac, mult, sub, add, adj
+    // Should do an AVX2 version with mac.
+    ///////////////////////////////////////////////
+    friend inline void mac (Grid_simd * __restrict__ y,const Grid_simd * __restrict__ a,const Grid_simd *__restrict__ x){ *y = (*a)*(*x)+(*y); };
+    friend inline void mult(Grid_simd * __restrict__ y,const Grid_simd * __restrict__ l,const Grid_simd *__restrict__ r){ *y = (*l) * (*r); }
+    friend inline void sub (Grid_simd * __restrict__ y,const Grid_simd * __restrict__ l,const Grid_simd *__restrict__ r){ *y = (*l) - (*r); }
+    friend inline void add (Grid_simd * __restrict__ y,const Grid_simd * __restrict__ l,const Grid_simd *__restrict__ r){ *y = (*l) + (*r); }
+    friend inline Grid_simd adj(const Grid_simd &in){ return conj(in); }
+        
+    //////////////////////////////////
+    // Initialise to 1,0,i
+    //////////////////////////////////
+    // if not complex overload here 
+    friend inline void vone(Grid_simd &ret)      { vsplat(ret,1.0); }
+    friend inline void vzero(Grid_simd &ret)     { vsplat(ret,0.0); }
+
+    // overload for complex type
+    template <  class S = Scalar_type,typename std::enable_if < is_complex < S >::value, int >::type = 0 > 
+    friend inline void vone(Grid_simd &ret)      { vsplat(ret,1.0,0.0); }
+    template < class S = Scalar_type,typename std::enable_if < is_complex < S >::value, int >::type = 0 > 
+    friend inline void vzero(Grid_simd &ret)     { vsplat(ret,0.0,0.0); }
+
+    // For integral type
+    template <  class S = Scalar_type,typename std::enable_if < std::is_integral < S >::value, int >::type = 0 > 
+    friend inline void vone(Grid_simd &ret)      { vsplat(ret,1); }
+    template <  class S = Scalar_type,typename std::enable_if < std::is_integral < S >::value, int >::type = 0 > 
+    friend inline void vzero(Grid_simd &ret)      { vsplat(ret,0); }
+
+
+    
+    // do not compile if real or integer, send an error message from the compiler
+    template < class S = Scalar_type,typename std::enable_if < is_complex < S >::value, int >::type = 0 > 
+    friend inline void vcomplex_i(Grid_simd &ret){ vsplat(ret,0.0,1.0);}
+       
+
+    
+   
+    ////////////////////////////////////
+    // Arithmetic operator overloads +,-,*
+    ////////////////////////////////////
+    friend inline Grid_simd operator + (Grid_simd a, Grid_simd b)
+    {
+      vComplexF ret;
+      // FIXME call the binary op
+      return ret;
+    };
+        
+    friend inline Grid_simd operator - (Grid_simd a, Grid_simd b)
+    {
+      vComplexF ret;
+      // FIXME call the binary op
+      return ret;
+    };
+        
+    friend inline Grid_simd operator * (Grid_simd a, Grid_simd b)
+      {
+	vComplexF ret;
+	// FIXME call the binary op
+	return ret;
+      };
+      
+
+    ////////////////////////////////////////////////////////////////////////
+    // FIXME:  gonna remove these load/store, get, set, prefetch
+    ////////////////////////////////////////////////////////////////////////
+    friend inline void vset(Grid_simd &ret, Scalar_type *a){
+      // FIXME set 
+    }
+        
+    ///////////////////////
+    // Splat
+    ///////////////////////
+    // overload if complex
+    template < class S = Scalar_type > 
+    friend inline void vsplat(Grid_simd &ret, typename std::enable_if< is_complex < S >::value, S>::type c){
+      Real a= real(c);
+      Real b= imag(c);
+      vsplat(ret,a,b);
+    }
+
+    // this only for the complex version
+    template < class S = Scalar_type, typename std::enable_if < is_complex < S >::value, int >::type = 0 > 
+    friend inline void vsplat(Grid_simd &ret,Real a, Real b){
+      // FIXME add operator
+    }    
+
+    //if real fill with a, if complex fill with a in the real part
+    friend inline void vsplat(Grid_simd &ret,Real a){
+      // FIXME add operator
+    }    
+
+
+
+    friend inline void vstore(const Grid_simd &ret, Scalar_type *a){
+      //FIXME
+    }
+    friend inline void vprefetch(const Grid_simd &v)
+    {
+      _mm_prefetch((const char*)&v.v,_MM_HINT_T0);
+    }
+
+
+
+    friend inline Scalar_type Reduce(const Grid_simd & in)
+    {
+      // FIXME add operator
+    }
+
+    friend inline Grid_simd operator * (const Scalar_type &a, Grid_simd b){
+      Grid_simd va;
+      vsplat(va,a);
+      return va*b;
+    }
+    friend inline Grid_simd operator * (Grid_simd b,const Scalar_type &a){
+      return a*b;
+    }
+
+    ///////////////////////
+    // Conjugate
+    ///////////////////////
+								     
+    friend inline Grid_simd  conj(const Grid_simd  &in){
+      Grid_simd  ret ; vzero(ret);
+      // FIXME add operator
+      return ret;
+    }
+    friend inline Grid_simd timesMinusI(const Grid_simd &in){
+      Grid_simd ret; 
+      vzero(ret);
+      // FIXME add operator
+      return ret;
+    }
+    friend inline Grid_simd timesI(const Grid_simd &in){
+      Grid_simd ret; vzero(ret);
+      // FIXME add operator
+      return ret;
+    }
+        
+    // Unary negation
+    friend inline Grid_simd operator -(const Grid_simd &r) {
+      vComplexF ret;
+      vzero(ret);
+      ret = ret - r;
+      return ret;
+    }
+    // *=,+=,-= operators
+    inline Grid_simd &operator *=(const Grid_simd &r) {
+      *this = (*this)*r;
+      return *this;
+    }
+    inline Grid_simd &operator +=(const Grid_simd &r) {
+      *this = *this+r;
+      return *this;
+    }
+    inline Grid_simd &operator -=(const Grid_simd &r) {
+      *this = *this-r;
+      return *this;
+    }
+
+
+
+    friend inline void permute(Grid_simd &y,Grid_simd b,int perm)
+    {
+      Gpermute<Grid_simd>(y,b,perm);
+    }
+    friend inline void merge(Grid_simd &y,std::vector<Scalar_type *> &extracted)
+    {
+      Gmerge<Grid_simd,Scalar_type >(y,extracted);
+    }
+    friend inline void extract(const Grid_simd &y,std::vector<Scalar_type *> &extracted)
+    {
+      Gextract<Grid_simd,Scalar_type>(y,extracted);
+    }
+    friend inline void merge(Grid_simd &y,std::vector<Scalar_type > &extracted)
+    {
+      Gmerge<Grid_simd,Scalar_type >(y,extracted);
+    }
+    friend inline void extract(const Grid_simd &y,std::vector<Scalar_type > &extracted)
+    {
+      Gextract<Grid_simd,Scalar_type>(y,extracted);
+    }
+
+
+  };// end of Grid_simd class definition 
+
+
+
+
+
+
+  template<class scalar_type, class vector_type > 
+    inline Grid_simd< scalar_type, vector_type>  innerProduct(const Grid_simd< scalar_type, vector_type> & l, const Grid_simd< scalar_type, vector_type> & r) 
+  {
+    return conj(l)*r; 
+  }
+
+  template<class scalar_type, class vector_type >
+  inline void zeroit(Grid_simd< scalar_type, vector_type> &z){ vzero(z);}
+
+
+  template<class scalar_type, class vector_type >
+  inline Grid_simd< scalar_type, vector_type> outerProduct(const Grid_simd< scalar_type, vector_type> &l, const Grid_simd< scalar_type, vector_type>& r)
+  {
+    return l*r;
+  }
+
+
+  template<class scalar_type, class vector_type >
+  inline Grid_simd< scalar_type, vector_type> trace(const Grid_simd< scalar_type, vector_type> &arg){
+    return arg;
+  }
+
+
+  // Define available types (now change names to avoid clashing)
+  typedef __m128 SIMD_type;// decided at compilation time
+  typedef Grid_simd< float                 , SIMD_type > MyRealF;
+  typedef Grid_simd< double                , SIMD_type > MyRealD;
+  typedef Grid_simd< std::complex< float > , SIMD_type > MyComplexF;
+  typedef Grid_simd< std::complex< double >, SIMD_type > MyComplexD;
+
+
+
+}
+
+#endif

lib/stamp-h1

@@ -0,0 +1 @@
+timestamp for lib/Grid_config.h

tests/Grid_cshift.cc

@@ -34,6 +34,8 @@ int main (int argc, char ** argv)
     U=lex;
   }
 
+
+
   TComplex cm;
   
   for(int dir=0;dir<4;dir++){