upgraded baryon and dibaryon benchmarks for generic flavors

benchmarks/compile_and_run_benchmarks_mpi.sh

@@ -1,9 +1,8 @@
 #!/bin/bash
 
 # Machine specific directories
-MPIINCLUDE="-I/gscratch/sw/intel-2020u1/compilers_and_libraries_2020.1.217/linux/mpi/intel64/include"
-MPILIB="-L/gscratch/sw/intel-2020u1/compilers_and_libraries_2020.1.217/linux/mpi/intel64/lib -L/gscratch/sw/intel-2020u1/compilers_and_libraries_2020.1.217/linux/mpi/intel64/lib/release"
-
+MPIINCLUDE="-I/opt/ohpc/pub/compiler/intel/compilers_and_libraries_2019.4.243/linux/mpi/intel64/include"
+MPILIB="-L/opt/ohpc/pub/compiler/intel/compilers_and_libraries_2019.4.243/linux/mpi/intel64/lib"
 
 CXX="mpicxx -DWITH_MPI"
 DXX="mpicxx -DWITH_MPI"
@@ -23,9 +22,11 @@ KERNEL_FOLDER=$1
 KERNEL=$2
 source configure_paths.sh
 
-CXXFLAGS="-std=c++11 -O3 -fno-rtti -mavx2 -m64 $MPIINCLUDE $MPILIB -lmpi"
-CORES=128
-#CORES=2
+CXXFLAGS="-std=c++11 -O3 -fno-rtti -mavx512f -m64 $MPIINCLUDE $MPILIB -lmpi"
+#CORES=128
+#CORES=216
+#CORES=1
+CORES=8
 
 # Compile options
 # - Make ${CXX} dump generated assembly
@@ -74,7 +75,7 @@ echo "Running ${KERNEL} wrapper"
 #VTUNE_METRIC=memory-access
 #PROFILING_COMMAND="amplxe-cl -collect ${VTUNE_METRIC} -result-dir vtune_results -quiet"
 #rm -rf vtune_results
-RUN_REF=1 RUN_TIRAMISU=1 HL_NUM_THREADS=$CORES LD_LIBRARY_PATH=${LD_LIBRARY_PATH}:${HALIDE_LIB_DIRECTORY}:${ISL_LIB_DIRECTORY}:${TIRAMISU_ROOT}/build/:${MKL_PREFIX}/lib/${MKL_LIB_PATH_SUFFIX} DYLD_LIBRARY_PATH=${DYLD_LIBRARY_PATH}:${HALIDE_LIB_DIRECTORY}:${TIRAMISU_ROOT}/build/:${MKL_PREFIX}/lib/${MKL_LIB_PATH_SUFFIX} ${PROFILING_COMMAND} mpirun -np ${CORES} ./${KERNEL}_wrapper
+#RUN_REF=1 RUN_TIRAMISU=1 HL_NUM_THREADS=$CORES LD_LIBRARY_PATH=${LD_LIBRARY_PATH}:${HALIDE_LIB_DIRECTORY}:${ISL_LIB_DIRECTORY}:${TIRAMISU_ROOT}/build/:${MKL_PREFIX}/lib/${MKL_LIB_PATH_SUFFIX} DYLD_LIBRARY_PATH=${DYLD_LIBRARY_PATH}:${HALIDE_LIB_DIRECTORY}:${TIRAMISU_ROOT}/build/:${MKL_PREFIX}/lib/${MKL_LIB_PATH_SUFFIX} ${PROFILING_COMMAND} mpirun -np ${CORES} ./${KERNEL}_wrapper
 
 #rm -rf ${KERNEL}_generator ${KERNEL}_wrapper generated_${KERNEL}.o generated_${KERNEL}.o.h
 

benchmarks/run_benchmarks_mpi.sh

@@ -0,0 +1,70 @@
+#!/bin/bash
+
+# Machine specific directories
+MPIINCLUDE="-I/opt/ohpc/pub/compiler/intel/compilers_and_libraries_2019.4.243/linux/mpi/intel64/include"
+MPILIB="-L/opt/ohpc/pub/compiler/intel/compilers_and_libraries_2019.4.243/linux/mpi/intel64/lib"
+
+CXX="mpicxx -DWITH_MPI"
+DXX="mpicxx -DWITH_MPI"
+
+set -x
+
+if [ $# -eq 0 ]; then
+ echo "Usage: TIRAMISU_SMALL=1 script.sh <KERNEL_FOLDER> <KERNEL_NAME_WITHOUT_EXTENSION>"
+ echo "Example: script.sh level1/axpy axpy"
+ exit
+fi
+
+# Define data sizes, possible value: -DTIRAMISU_XLARGE, -DTIRAMISU_LARGE, -DTIRAMISU_MEDIUM, -DTIRAMISU_SMALL
+DEFINED_SIZE="-DTIRAMISU_LARGE"
+
+KERNEL_FOLDER=$1
+KERNEL=$2
+source configure_paths.sh
+
+CXXFLAGS="-m64 -std=c++11 -O3 -fno-rtti -mavx2 $MPIINCLUDE $MPILIB -lmpi"
+CORES=8
+
+# Compile options
+# - Make ${CXX} dump generated assembly
+# CXXFLAGS: -g -Wa,-alh
+# - Get info about ${CXX} vectorization
+# CXXFLAGS -fopt-info-vec
+# - Pass options to the llvm compiler
+# HL_LLVM_ARGS="-help" 
+# - Set thread number for Halide
+# HL_NUM_THREADS=32
+# - Execution env variables
+# OMP_NUM_THREADS=48
+# to set the number of threads to use by OpenMP.
+# - Command to run Vtune
+# source /data/scratch/yunming/intel_parallel_studio_cluster/parallel_studio_xe_2017/install/vtune_amplifier_xe/amplxe-vars.sh
+# amplxe-cl -collect hpc-performance -result-dir vtune_results -quiet ./binary
+# Guide: https://software.intel.com/en-us/vtune-amplifier-help-amplxe-cl-command-syntax
+
+INCLUDES="-I${MKL_PREFIX}/include/ -I${TIRAMISU_ROOT}/include/ -I${HALIDE_SOURCE_DIRECTORY}/include/ -I${ISL_INCLUDE_DIRECTORY} -I${TIRAMISU_ROOT}/benchmarks/"
+LIBRARIES="-ltiramisu ${MKL_FLAGS} -lHalide -lisl -lz -lpthread -ldl ${EXTRA_LIBRARIES}"
+LIBRARIES_DIR="-L${MKL_PREFIX}/lib/${MKL_LIB_PATH_SUFFIX} -L${HALIDE_LIB_DIRECTORY}/ -L${ISL_LIB_DIRECTORY}/ -L${TIRAMISU_ROOT}/build/"
+
+echo "Compiling ${KERNEL}"
+
+cd ${KERNEL_FOLDER}
+
+echo "Compiling ${KERNEL} wrapper"
+${DXX} ${LANKA_OPTIONS} $CXXFLAGS ${INCLUDES} ${DEFINED_SIZE} ${KERNEL}_wrapper.cpp ${LIBRARIES_DIR} ${LIBRARIES} generated_${KERNEL}.o ${LIBRARIES} -o ${KERNEL}_wrapper
+
+echo "Running ${KERNEL} wrapper"
+# To enable profiling:
+## Perf:
+#PROFILING_COMMAND="perf stat -e cycles,instructions,cache-misses,L1-icache-load-misses,LLC-load-misses,dTLB-load-misses,cpu-migrations,context-switches,bus-cycles,cache-references,minor-faults"
+## Vtune:
+#VTUNE_METRIC=hpc-performance
+#VTUNE_METRIC=memory-access
+#PROFILING_COMMAND="amplxe-cl -collect ${VTUNE_METRIC} -result-dir vtune_results -quiet"
+#rm -rf vtune_results
+
+RUN_REF=1 RUN_TIRAMISU=1 HL_NUM_THREADS=$CORES LD_LIBRARY_PATH=${LD_LIBRARY_PATH}:${HALIDE_LIB_DIRECTORY}:${ISL_LIB_DIRECTORY}:${TIRAMISU_ROOT}/build/:${MKL_PREFIX}/lib/${MKL_LIB_PATH_SUFFIX} DYLD_LIBRARY_PATH=${DYLD_LIBRARY_PATH}:${HALIDE_LIB_DIRECTORY}:${TIRAMISU_ROOT}/build/:${MKL_PREFIX}/lib/${MKL_LIB_PATH_SUFFIX} ${PROFILING_COMMAND} srun -n ${CORES} --mpi=pmi2 ./${KERNEL}_wrapper
+
+#rm -rf ${KERNEL}_generator ${KERNEL}_wrapper generated_${KERNEL}.o generated_${KERNEL}.o.h
+
+cd -