Commit d871fe90 authored by Giovanni La Mura's avatar Giovanni La Mura
Browse files

Merge branch 'parallel_angles_gmu' into 'master' 

Implement parallel directional routines

See merge request giacomo.mulas/np_tmcode!52
parents 244cbc04 ca5e3b80

build/Makefile.am

+1 −1
Original line number Diff line number Diff line 
LDADD=libnptm/libnptm.la -L/usr/lib64 ${USER_LDFLAGS} ${HDF5_LDFLAGS} ${LAPACKLDFLAGS} ${MAGMALDFLAGS}

lib_LTLIBRARIES=libnptm/libnptm.la

libnptm_libnptm_la_SOURCES=../src/libnptm/algebraic.cpp ../src/libnptm/clu_subs.cpp ../src/libnptm/Commons.cpp ../src/libnptm/Configuration.cpp ../src/libnptm/file_io.cpp ../src/libnptm/lapack_calls.cpp ../src/libnptm/logging.cpp ../src/libnptm/magma_calls.cpp ../src/libnptm/Parsers.cpp ../src/libnptm/sph_subs.cpp ../src/libnptm/tfrfme.cpp ../src/libnptm/TransitionMatrix.cpp ../src/libnptm/tra_subs.cpp ../src/libnptm/types.cpp

libnptm_libnptm_la_SOURCES=../src/libnptm/algebraic.cpp ../src/libnptm/clu_subs.cpp ../src/libnptm/Commons.cpp ../src/libnptm/Configuration.cpp ../src/libnptm/file_io.cpp ../src/libnptm/lapack_calls.cpp ../src/libnptm/logging.cpp ../src/libnptm/magma_calls.cpp ../src/libnptm/Parsers.cpp ../src/libnptm/sph_subs.cpp ../src/libnptm/tfrfme.cpp ../src/libnptm/TransitionMatrix.cpp ../src/libnptm/tra_subs.cpp

bin_PROGRAMS=cluster/edfb_clu cluster/clu cluster/np_cluster sphere/edfb_sph sphere/sph sphere/np_sphere trapping/frfme trapping/lffft trapping/np_trapping testing/test_TEDF testing/test_TTMS

cluster_edfb_clu_SOURCES=../src/cluster/edfb_clu.f

cluster_clu_SOURCES=../src/cluster/clu.f

build/Makefile.in

+3 −9
Original line number Diff line number Diff line
@@ -145,8 +145,7 @@ am_libnptm_libnptm_la_OBJECTS = ../src/libnptm/algebraic.lo \ 
	../src/libnptm/lapack_calls.lo ../src/libnptm/logging.lo \

	../src/libnptm/magma_calls.lo ../src/libnptm/Parsers.lo \

	../src/libnptm/sph_subs.lo ../src/libnptm/tfrfme.lo \

	../src/libnptm/TransitionMatrix.lo ../src/libnptm/tra_subs.lo \

	../src/libnptm/types.lo

	../src/libnptm/TransitionMatrix.lo ../src/libnptm/tra_subs.lo

libnptm_libnptm_la_OBJECTS = $(am_libnptm_libnptm_la_OBJECTS)

AM_V_lt = $(am__v_lt_@AM_V@)

am__v_lt_ = $(am__v_lt_@AM_DEFAULT_V@)
@@ -254,7 +253,6 @@ am__depfiles_remade = ../src/cluster/$(DEPDIR)/cluster.Po \ 
	../src/libnptm/$(DEPDIR)/sph_subs.Plo \

	../src/libnptm/$(DEPDIR)/tfrfme.Plo \

	../src/libnptm/$(DEPDIR)/tra_subs.Plo \

	../src/libnptm/$(DEPDIR)/types.Plo \

	../src/sphere/$(DEPDIR)/np_sphere.Po \

	../src/sphere/$(DEPDIR)/sphere.Po \

	../src/testing/$(DEPDIR)/test_TEDF.Po \
@@ -445,6 +443,7 @@ NMEDIT = @NMEDIT@ 
NVTXFLAGS = @NVTXFLAGS@

OBJDUMP = @OBJDUMP@

OBJEXT = @OBJEXT@

OFFLOADFLAGS = @OFFLOADFLAGS@

OMPFLAGS = @OMPFLAGS@

OTOOL = @OTOOL@

OTOOL64 = @OTOOL64@
@@ -521,7 +520,7 @@ top_builddir = @top_builddir@ 
top_srcdir = @top_srcdir@

LDADD = libnptm/libnptm.la -L/usr/lib64 ${USER_LDFLAGS} ${HDF5_LDFLAGS} ${LAPACKLDFLAGS} ${MAGMALDFLAGS}

lib_LTLIBRARIES = libnptm/libnptm.la

libnptm_libnptm_la_SOURCES = ../src/libnptm/algebraic.cpp ../src/libnptm/clu_subs.cpp ../src/libnptm/Commons.cpp ../src/libnptm/Configuration.cpp ../src/libnptm/file_io.cpp ../src/libnptm/lapack_calls.cpp ../src/libnptm/logging.cpp ../src/libnptm/magma_calls.cpp ../src/libnptm/Parsers.cpp ../src/libnptm/sph_subs.cpp ../src/libnptm/tfrfme.cpp ../src/libnptm/TransitionMatrix.cpp ../src/libnptm/tra_subs.cpp ../src/libnptm/types.cpp

libnptm_libnptm_la_SOURCES = ../src/libnptm/algebraic.cpp ../src/libnptm/clu_subs.cpp ../src/libnptm/Commons.cpp ../src/libnptm/Configuration.cpp ../src/libnptm/file_io.cpp ../src/libnptm/lapack_calls.cpp ../src/libnptm/logging.cpp ../src/libnptm/magma_calls.cpp ../src/libnptm/Parsers.cpp ../src/libnptm/sph_subs.cpp ../src/libnptm/tfrfme.cpp ../src/libnptm/TransitionMatrix.cpp ../src/libnptm/tra_subs.cpp

cluster_edfb_clu_SOURCES = ../src/cluster/edfb_clu.f

cluster_clu_SOURCES = ../src/cluster/clu.f

cluster_np_cluster_SOURCES = ../src/cluster/np_cluster.cpp ../src/cluster/cluster.cpp
@@ -686,8 +685,6 @@ clean-libLTLIBRARIES: 
	../src/libnptm/$(DEPDIR)/$(am__dirstamp)

../src/libnptm/tra_subs.lo: ../src/libnptm/$(am__dirstamp) \

	../src/libnptm/$(DEPDIR)/$(am__dirstamp)

../src/libnptm/types.lo: ../src/libnptm/$(am__dirstamp) \

	../src/libnptm/$(DEPDIR)/$(am__dirstamp)

libnptm/$(am__dirstamp):

	@$(MKDIR_P) libnptm

	@: > libnptm/$(am__dirstamp)
@@ -833,7 +830,6 @@ distclean-compile: 
@AMDEP_TRUE@@am__include@ @am__quote@../src/libnptm/$(DEPDIR)/sph_subs.Plo@am__quote@ # am--include-marker

@AMDEP_TRUE@@am__include@ @am__quote@../src/libnptm/$(DEPDIR)/tfrfme.Plo@am__quote@ # am--include-marker

@AMDEP_TRUE@@am__include@ @am__quote@../src/libnptm/$(DEPDIR)/tra_subs.Plo@am__quote@ # am--include-marker

@AMDEP_TRUE@@am__include@ @am__quote@../src/libnptm/$(DEPDIR)/types.Plo@am__quote@ # am--include-marker

@AMDEP_TRUE@@am__include@ @am__quote@../src/sphere/$(DEPDIR)/np_sphere.Po@am__quote@ # am--include-marker

@AMDEP_TRUE@@am__include@ @am__quote@../src/sphere/$(DEPDIR)/sphere.Po@am__quote@ # am--include-marker

@AMDEP_TRUE@@am__include@ @am__quote@../src/testing/$(DEPDIR)/test_TEDF.Po@am__quote@ # am--include-marker
@@ -1202,7 +1198,6 @@ distclean: distclean-am 
	-rm -f ../src/libnptm/$(DEPDIR)/sph_subs.Plo

	-rm -f ../src/libnptm/$(DEPDIR)/tfrfme.Plo

	-rm -f ../src/libnptm/$(DEPDIR)/tra_subs.Plo

	-rm -f ../src/libnptm/$(DEPDIR)/types.Plo

	-rm -f ../src/sphere/$(DEPDIR)/np_sphere.Po

	-rm -f ../src/sphere/$(DEPDIR)/sphere.Po

	-rm -f ../src/testing/$(DEPDIR)/test_TEDF.Po
@@ -1272,7 +1267,6 @@ maintainer-clean: maintainer-clean-am 
	-rm -f ../src/libnptm/$(DEPDIR)/sph_subs.Plo

	-rm -f ../src/libnptm/$(DEPDIR)/tfrfme.Plo

	-rm -f ../src/libnptm/$(DEPDIR)/tra_subs.Plo

	-rm -f ../src/libnptm/$(DEPDIR)/types.Plo

	-rm -f ../src/sphere/$(DEPDIR)/np_sphere.Po

	-rm -f ../src/sphere/$(DEPDIR)/sphere.Po

	-rm -f ../src/testing/$(DEPDIR)/test_TEDF.Po

build/configure

+114 −1
Original line number Diff line number Diff line
@@ -664,6 +664,7 @@ MAGMAFLAGS 
LAPACKLDFLAGS

LAPACKFLAGS

OMPFLAGS

OFFLOADFLAGS

HDF5_LDFLAGS

HDF5_LIB

HDF5_INCLUDE
@@ -810,6 +811,7 @@ with_aix_soname 
with_gnu_ld

with_sysroot

enable_libtool_lock

enable_offload

enable_openmp

with_lapack

with_magma
@@ -1472,6 +1474,8 @@ Optional Features: 
  --enable-fast-install[=PKGS]

                          optimize for fast installation [default=yes]

  --disable-libtool-lock  avoid locking (might break parallel builds)

  --enable-offload        enable target offloading (requires g++ version >=

                          13) [default=auto]

  --enable-openmp         enable OpneMP multi-threading [default=yes]

  --enable-nvtx           use NVTX profiling [default=no]










@@ -24883,6 +24887,105 @@ esac













fi





























# Configure the optional features













# Check whether --enable-offload was given.













if test ${enable_offload+y}













then :













  enableval=$enable_offload;













    if test "x$enableval" != "xno"; then

























    cat > np_test_offload.cpp <<EOF













#include <omp.h>













#pragma omp requires unified_shared_memory

























#pragma omp begin declare target device_type(any)













void fill_with_ones(int *array) {













#pragma omp target teams distribute parallel for













  for (int i = 0; i < 1000; i++) {













    for (int j = 0; j < 1000; j++) {













      array[(1000 * i) + j] = 1;













    }













  }













}













#pragma omp end declare target













int main(int argc, char** argv) {













  int *numbers = new int[1000000]();













  fill_with_ones(numbers);













  delete[] numbers;













  return 0;













}













EOF













    $CXX -fcf-protection=check -foffload=default -foffload=nvptx-none="-O3 -ggdb -fopt-info -lm -latomic -mgomp" -fopenmp -c np_test_offload.cpp > /dev/null 2>&1













    export CXX_SUPPORTS_OFFLOAD=$?













    rm np_test_offload.cpp













    if test "x$CXX_SUPPORTS_OFFLOAD" = "x0"; then













      rm np_test_offload.o













    fi





































      if test "x$CXX_SUPPORTS_OFFLOAD" = "x0"; then













        { printf "%s\n" "$as_me:${as_lineno-$LINENO}: Enabling offload." >&5













printf "%s\n" "$as_me: Enabling offload." >&6;}













        OFFLOADFLAGS="-fcf-protection=check -foffload=default -foffload=nvptx-none=\"-O3 -ggdb -fopt-info -lm -latomic -mgomp\" -fopenmp -DUSE_TARGET_OFFLOAD "

























      else













        as_fn_error $? "Target offload was requested, but it is not supported!" "$LINENO" 5













      fi













    else













      { printf "%s\n" "$as_me:${as_lineno-$LINENO}: Disabling offload." >&5













printf "%s\n" "$as_me: Disabling offload." >&6;}













      OFFLOADFLAGS=""

























    fi

























else case e in #(













  e)

























    cat > np_test_offload.cpp <<EOF













#include <omp.h>













#pragma omp requires unified_shared_memory

























#pragma omp begin declare target device_type(any)













void fill_with_ones(int *array) {













#pragma omp target teams distribute parallel for













  for (int i = 0; i < 1000; i++) {













    for (int j = 0; j < 1000; j++) {













      array[(1000 * i) + j] = 1;













    }













  }













}













#pragma omp end declare target













int main(int argc, char** argv) {













  int *numbers = new int[1000000]();













  fill_with_ones(numbers);













  delete[] numbers;













  return 0;













}













EOF













    $CXX -fcf-protection=check -foffload=default -foffload=nvptx-none="-O3 -ggdb -fopt-info -lm -latomic -mgomp" -fopenmp -c np_test_offload.cpp > /dev/null 2>&1













    export CXX_SUPPORTS_OFFLOAD=$?













    rm np_test_offload.cpp













    if test "x$CXX_SUPPORTS_OFFLOAD" = "x0"; then













      rm np_test_offload.o













    fi





































    if test "x$CXX_SUPPORTS_OFFLOAD" = "x0"; then













      { printf "%s\n" "$as_me:${as_lineno-$LINENO}: Enabling offload." >&5













printf "%s\n" "$as_me: Enabling offload." >&6;}













      OFFLOADFLAGS="-fcf-protection=check -foffload=default -foffload=nvptx-none=\"-O3 -ggdb -fopt-info -lm -latomic -mgomp\" -fopenmp -DUSE_TARGET_OFFLOAD "

























    else













      { printf "%s\n" "$as_me:${as_lineno-$LINENO}: Disabling offload." >&5













printf "%s\n" "$as_me: Disabling offload." >&6;}













      OFFLOADFLAGS=""

























    fi

























 ;;













esac













fi







































# Check whether --enable-openmp was given.















if test ${enable_openmp+y}















then :










@@ -25175,7 +25278,17 @@ esac













fi









































CXXFLAGS="$CLANGFLAGS -O3 -ggdb $USER_INCLUDE -I$HDF5_INCLUDE $OMPFLAGS $MPIFLAGS $LAPACKFLAGS $MAGMAFLAGS $NVTXFLAGS"













if test "x$OFFLOADFLAGS" != "x"













then :













  OMPFLAGS=""

























else case e in #(













  e) OMPFLAGS=$OMPFLAGS

























 ;;













esac













fi













CXXFLAGS="$CLANGFLAGS -O3 -ggdb $OFFLOADFLAGS $USER_INCLUDE -I$HDF5_INCLUDE $OMPFLAGS $MPIFLAGS $LAPACKFLAGS $MAGMAFLAGS $NVTXFLAGS"















SUBDIRS="cluster libnptm sphere testing trapping"





























# Generate the output

























build/configure.ac



+68
−1




























Original line number
Diff line number
Diff line








@@ -93,6 +93,39 @@ EOF













    fi















  ]















)



























m4_define(













  [M4_TEST_OFFLOAD],













  [













    cat > np_test_offload.cpp <<EOF













#include <omp.h>













#pragma omp requires unified_shared_memory



























#pragma omp begin declare target device_type(any)













void fill_with_ones(int *array) {













#pragma omp target teams distribute parallel for













  for (int i = 0; i < 1000; i++) {













    for (int j = 0; j < 1000; j++) {













      array[[(1000 * i) + j]] = 1;













    }













  }













}













#pragma omp end declare target













int main(int argc, char** argv) {













  int *numbers = new int[[1000000]]();













  fill_with_ones(numbers);













  delete[[]] numbers;













  return 0;













}













EOF













    $CXX -fcf-protection=check -foffload=default -foffload=nvptx-none="-O3 -ggdb -fopt-info -lm -latomic -mgomp" -fopenmp -c np_test_offload.cpp > /dev/null 2>&1













    export CXX_SUPPORTS_OFFLOAD=$?













    rm np_test_offload.cpp













    if test "x$CXX_SUPPORTS_OFFLOAD" = "x0"; then













      rm np_test_offload.o













    fi













  ]













)















# END CAPABILITY TESTING MACROS































# autoconf setup initialization










@@ -219,6 +252,35 @@ AS_IF(













)































# Configure the optional features













AC_ARG_ENABLE(













  [offload],













  [AS_HELP_STRING([--enable-offload], [enable target offloading (requires g++ version >= 13) [default=auto]])],













  [













    if test "x$enableval" != "xno"; then













      M4_TEST_OFFLOAD













      if test "x$CXX_SUPPORTS_OFFLOAD" = "x0"; then













        AC_MSG_NOTICE([Enabling offload.])













        AC_SUBST([OFFLOADFLAGS], ["-fcf-protection=check -foffload=default -foffload=nvptx-none=\"-O3 -ggdb -fopt-info -lm -latomic -mgomp\" -fopenmp -DUSE_TARGET_OFFLOAD "])













      else













        AC_MSG_ERROR([Target offload was requested, but it is not supported!])













      fi













    else













      AC_MSG_NOTICE([Disabling offload.])













      AC_SUBST([OFFLOADFLAGS], [""])













    fi













  ],













  [













    M4_TEST_OFFLOAD













    if test "x$CXX_SUPPORTS_OFFLOAD" = "x0"; then













      AC_MSG_NOTICE([Enabling offload.])













      AC_SUBST([OFFLOADFLAGS], ["-fcf-protection=check -foffload=default -foffload=nvptx-none=\"-O3 -ggdb -fopt-info -lm -latomic -mgomp\" -fopenmp -DUSE_TARGET_OFFLOAD "])













    else













      AC_MSG_NOTICE([Disabling offload.])













      AC_SUBST([OFFLOADFLAGS], [""])













    fi













  ]













)





























AC_ARG_ENABLE(















  [openmp],















  [AS_HELP_STRING([--enable-openmp], [enable OpneMP multi-threading [default=yes]])],










@@ -346,7 +408,12 @@ AC_ARG_WITH(













  ]















)





























CXXFLAGS="$CLANGFLAGS -O3 -ggdb $USER_INCLUDE -I$HDF5_INCLUDE $OMPFLAGS $MPIFLAGS $LAPACKFLAGS $MAGMAFLAGS $NVTXFLAGS"













AS_IF(













  [test "x$OFFLOADFLAGS" != "x"],













  [AC_SUBST([OMPFLAGS], [""])],













  [AC_SUBST([OMPFLAGS], [$OMPFLAGS])]













)













CXXFLAGS="$CLANGFLAGS -O3 -ggdb $OFFLOADFLAGS $USER_INCLUDE -I$HDF5_INCLUDE $OMPFLAGS $MPIFLAGS $LAPACKFLAGS $MAGMAFLAGS $NVTXFLAGS"















SUBDIRS="cluster libnptm sphere testing trapping"































# Generate the output

































src/cluster/cluster.cpp



+9
−1




























Original line number
Diff line number
Diff line








@@ -306,7 +306,15 @@ void cluster(const string& config_file, const string& data_file, const string& o













#ifdef USE_NVTX















      nvtxRangePush("First iteration");















#endif













      int jer = cluster_jxi488_cycle(jxi488, sconf, gconf, p_scattering_angles, cid, p_output, output_path, vtppoanp);













      // use these pragmas, which should have no effect on parallelism, just to push OMP nested levels at the same level also in the first wavelength iteration













      int jer = 0;













#pragma omp parallel













      {













#pragma omp single













	{













	  jer = cluster_jxi488_cycle(jxi488, sconf, gconf, p_scattering_angles, cid, p_output, output_path, vtppoanp);













	}













      }















#ifdef USE_NVTX















      nvtxRangePop();















#endif