New stable CODES version

.gitignore

@@ -37,4 +37,10 @@
 # generated files from test runs
 ross.csv
 
-install-mastiff/include/codes/model-net-method.h
+install-mastiff/include/codes/model-net-method.h
+
+# commonly used building stuff
+/build*/
+/build*
+.cache
+compile_commands.json

CMakeLists.txt

@@ -1,4 +1,4 @@
-cmake_minimum_required(VERSION 3.10)
+cmake_minimum_required(VERSION 3.17)
 
 # set the project name and version
 project(codes LANGUAGES C CXX VERSION 2.0)
@@ -22,12 +22,13 @@ SET(CMAKE_INSTALL_RPATH_USE_LINK_PATH TRUE)
 
 set(ROSS_PKG_CONFIG_PATH "" CACHE PATH "Where is ROSS PKG_CONFIG is installed?")
 set(SWM_PKG_CONFIG_PATH "" CACHE PATH "Where is the SWM PKG_CONFIG installed?")
+set(UNION_PKG_CONFIG_PATH "" CACHE PATH "Where is the Union PKG_CONFIG installed?")
 set(ARGOBOTS_PKG_CONFIG_PATH "" CACHE PATH "Where is argobots PKG_COPNFIG installed? Necessary for SWM")
 set(DAMARIS_PKG_CONFIG_PATH "" CACHE PATH "Where is the damaris PKG_CONFIG installed?")
 
 
 find_package(PkgConfig REQUIRED)
-set(ENV{PKG_CONFIG_PATH} "${ROSS_PKG_CONFIG_PATH}:${SWM_PKG_CONFIG_PATH}:${ARGOBOTS_PKG_CONFIG_PATH}")
+set(ENV{PKG_CONFIG_PATH} "${ROSS_PKG_CONFIG_PATH}:${SWM_PKG_CONFIG_PATH}:${UNION_PKG_CONFIG_PATH}:${ARGOBOTS_PKG_CONFIG_PATH}")
 pkg_check_modules(ROSS REQUIRED IMPORTED_TARGET ross)
 
 # MPI
@@ -57,28 +58,50 @@ else(DUMPI_LIB)
         set(USE_DUMPI true)
 endif()
 
-## SWM
+# SWM and UNION (both require ARGOBOTS to function)
 pkg_check_modules(SWM IMPORTED_TARGET swm)
 if(NOT SWM_FOUND)
-        message(STATUS "SWM Library Not Found, Online workloads disabled")
+    message(STATUS "SWM Library Not Found, Online workloads disabled")
+
 else(SWM_FOUND)
-        message(STATUS "SWM Library Found: ${SWM_LIBRARIES}")
-        pkg_check_modules(ARGOBOTS REQUIRED IMPORTED_TARGET argobots)
-        if(NOT ARGOBOTS_FOUND)
-                message(STATUS "Argobots Library Not Found, Online workloads disabled")
-        else(ARGOBOTS_FOUND)
-                message(STATUS "Argobots Library Found: ${ARGOBOTS_LIBRARIES}")
-                set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${ARGOBOTS_CFLAGS} -I${ARGOBOTS_INCLUDE}")
-
-                pkg_get_variable(SWM_DATAROOTDIR swm datarootdir)
-                cmake_print_variables(SWM_DATAROOTDIR)
-
-                set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${SWM_CFLAGS} -I${SWM_INCLUDE}")
-                add_definitions(-DUSE_ONLINE=1)
-                set(USE_ONLINE true)
+    message(STATUS "SWM Library Found: ${SWM_LIBRARIES}")
+    pkg_check_modules(ARGOBOTS REQUIRED IMPORTED_TARGET argobots)
+    if(NOT ARGOBOTS_FOUND)
+        message(STATUS "Argobots Library Not Found, Online workloads disabled")
+
+    else(ARGOBOTS_FOUND)
+        message(STATUS "Argobots Library Found: ${ARGOBOTS_LIBRARIES}")
+        set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${ARGOBOTS_CFLAGS} -I${ARGOBOTS_INCLUDE}")
+
+        pkg_get_variable(SWM_DATAROOTDIR swm datarootdir)
+        cmake_print_variables(SWM_DATAROOTDIR)
+
+        set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${SWM_CFLAGS} -I${SWM_INCLUDE}")
+        add_definitions(-DUSE_ONLINE=1)
+        set(USE_ONLINE true)
+
+        pkg_check_modules(UNION IMPORTED_TARGET union)
+        if(NOT UNION_FOUND)
+            message(STATUS "UNION Library Not Found, SWM-only online workloads enabled")
+            add_definitions(-DUSE_SWM=1)
+            set(USE_SWM true)
+        else(UNION_FOUND)
+            message(STATUS "UNION Library Found: ${UNION_LIBRARIES}")
+            pkg_get_variable(UNION_DATAROOTDIR union datarootdir)
+            cmake_print_variables(UNION_DATAROOTDIR)
+
+            set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${UNION_INCLUDE}")
+            foreach(INCLUDE_OPT ${UNION_CFLAGS})
+                set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${INCLUDE_OPT}")
+            endforeach()
+
+            add_definitions(-DUSE_UNION=1)
+            set(USE_UNION true)
         endif()
+    endif()
 endif()
 
+
 ## RECORDER
 option(USE_RECORDER "use recorder io workload" ON)
 if(USE_RECORDER)
@@ -96,11 +119,34 @@ endif()
 #         set(USE_DAMARIS true)
 # endif()
 
+## TORCH loading ML models
+if((NOT DEFINED USE_TORCH) OR USE_TORCH)
+    find_package(Torch)
+    if(Torch_FOUND)
+        set(CMAKE_CXX_STANDARD 17)
+        add_definitions(-DUSE_TORCH)
+        set(USE_TORCH true)
+        message(STATUS "Loading TORCH models enabled.")
+    else()
+        set(USE_TORCH false)
+        message(STATUS "Torch library not found. Loading TORCH models disabled.")
+    endif()
+else()
+    message(STATUS "Loading TORCH models NOT enabled.")
+endif()
+
 cmake_print_variables(CMAKE_C_FLAGS)
 add_subdirectory(src)
 
 
-configure_file(codes_config.h.in codes_config.h)
-
+configure_file(codes_config.h.cmake.in codes_config.h)
 
+add_subdirectory(doc/example)
 
+string(COMPARE NOTEQUAL "RELEASE" "${CMAKE_BUILD_TYPE}" not_release)
+if(BUILD_TESTING AND not_release)
+    include(CTest)
+    set(CODES_SOURCE_DIR "${CMAKE_CURRENT_SOURCE_DIR}")
+    set(CODES_BINARY_DIR "${CMAKE_CURRENT_BINARY_DIR}")
+    add_subdirectory(tests)
+endif()

CODES-compile-instructions.sh

@@ -0,0 +1,136 @@
+#!/usr/bin bash -x
+
+# Switches
+swm_enable=1
+union_enable=1
+torch_enable=0
+
+# Uncomment below for MPICH
+#export PATH=/usr/local/mpich-4.1.2/bin/:"$PATH"
+# Note: remember to compile MPICH with nemesis not with UCX support
+
+################## Actual scripts starts from here ##################
+
+# SWM has to be enabled for UNION to work
+if [ $union_enable = 1 ]; then
+    swm_enable=1
+fi
+
+# What to compile
+CUR_DIR="$PWD"
+
+##### Downloading everything #####
+
+git clone https://github.com/codes-org/codes --depth=100 --branch=v1.5.0
+git clone https://github.com/ross-org/ross --depth=100 --branch=v8.1.0
+
+if [ $swm_enable = 1 ]; then
+    git clone https://github.com/pmodels/argobots --depth=1
+    git clone https://github.com/codes-org/swm-workloads --branch=v1.2
+fi
+
+if [ $union_enable = 1 ]; then
+    # Downloading conceptual
+    curl -L https://sourceforge.net/projects/conceptual/files/conceptual/1.5.1b/conceptual-1.5.1b.tar.gz -o conceptual-1.5.1b.tar.gz
+    tar xvf conceptual-1.5.1b.tar.gz
+    # Downloading union
+    git clone https://github.com/SPEAR-UIC/Union
+    pushd Union && git checkout 99b3df3 && popd
+fi
+
+##### COMPILING #####
+
+mkdir ross/build
+pushd ross/build
+cmake .. -DROSS_BUILD_MODELS=ON -DCMAKE_INSTALL_PREFIX="$(realpath ./bin)" \
+  -DCMAKE_C_COMPILER=mpicc -DCMAKE_BUILD_TYPE=Debug -DCMAKE_C_FLAGS="-g -Wall"
+#make VERBOSE=1
+make install -j4
+err=$?
+[[ $err -ne 0 ]] && exit $err
+popd
+
+if [ $swm_enable = 1 ]; then
+    pushd swm-workloads/swm
+    ./prepare.sh
+    mkdir build
+    pushd build
+    ../configure --disable-shared --prefix="$(realpath ./bin)" CC=mpicc CXX=mpicxx CFLAGS=-g CXXFLAGS=-g
+    #make V=1 && make install
+    make -j4 && make install
+    err=$?
+    [[ $err -ne 0 ]] && exit $err
+    popd && popd
+
+    pushd argobots
+    ./autogen.sh
+    mkdir build
+    pushd build
+    #../configure --enable-debug=all --disable-fast --disable-shared --prefix="$(realpath ./bin)" CC=mpicc CXX=mpicxx CFLAGS=-g CXXFLAGS=-g
+    ../configure --disable-shared --prefix="$(realpath ./bin)" CC=mpicc CXX=mpicxx CFLAGS=-g CXXFLAGS=-g
+    #make V=1 && make install
+    make -j4 && make install
+    err=$?
+    [[ $err -ne 0 ]] && exit $err
+    popd && popd
+fi
+
+if [ $union_enable = 1 ]; then
+    pushd conceptual-1.5.1b
+    PYTHON=python2 ./configure --prefix="$(realpath ./install)" LIBS=-lm
+    make -j4 && make install
+    err=$?
+    [[ $err -ne 0 ]] && exit $err
+    popd
+
+    pushd Union
+    # Python 2 override. Union expects Python 2 ONLY
+    mkdir -p python-override
+    ln -s /usr/bin/python2 python-override/python
+    # compiling
+    ./prepare.sh
+    PYTHON=python2 ./configure --disable-shared --with-conceptual="$(realpath ../conceptual-1.5.1b/install)" --with-conceptual-src="$(realpath ../conceptual-1.5.1b)" --prefix="$(realpath ./install)" CC=mpicc CXX=mpicxx
+    PATH="$PWD/python-override:$PATH" make -j4 && make install
+    err=$?
+    [[ $err -ne 0 ]] && exit $err
+    popd
+fi
+
+
+mkdir codes/build
+pushd codes/build
+
+make_args_codes=(
+    -DCMAKE_PREFIX_PATH="$(realpath "$CUR_DIR/ross/build/bin")"
+    -DCMAKE_CXX_COMPILER=mpicxx -DCMAKE_C_COMPILER=mpicc
+    -DCMAKE_C_FLAGS="-g -Wall"
+    -DCMAKE_CXX_FLAGS="-g -Wall"
+    -DCMAKE_BUILD_TYPE=Debug -DBUILD_TESTING=ON
+    -DCMAKE_INSTALL_PREFIX="$(realpath bin)"
+)
+if [ $swm_enable = 1 ]; then
+    make_args_codes=(
+        "${make_args_codes[@]}"
+        -DSWM_PKG_CONFIG_PATH="$(realpath "$CUR_DIR/swm-workloads/swm/build/maint")"
+        -DARGOBOTS_PKG_CONFIG_PATH="$(realpath "$CUR_DIR/argobots/build/maint")"
+    )
+fi
+if [ $union_enable = 1 ]; then
+    make_args_codes=(
+        "${make_args_codes[@]}"
+        -DUNION_PKG_CONFIG_PATH="$(realpath "$CUR_DIR/Union/install/lib/pkgconfig")"
+    )
+fi
+if [ $torch_enable = 1 ]; then
+    make_args_codes=("${make_args_codes[@]}" -DUSE_TORCH=true)
+else
+    make_args_codes=("${make_args_codes[@]}" -DUSE_TORCH=false)
+fi
+
+cmake .. "${make_args_codes[@]}"
+#make VERBOSE=1
+make -j4
+err=$?
+[[ $err -ne 0 ]] && exit $err
+
+popd

CONTRIBUTORS.md

@@ -20,6 +20,8 @@ Contributors to date (with affiliations at time of contribution)
 - Lee Savoie, Univ. of Arizona 
 - Ning Liu, Rensselaer Polytechnic Institute
 - Jason Cope, Argonne National Laboratory
+- Kevin A. Brown, Argonne National Laboratory
+- Elkin Cruz, Rensselaer Polytechnic Institute
 
 Contributions:
 
@@ -40,6 +42,8 @@ Neil McGlohon (RPI)
     - Merged 1-D dragonfly and 2-D dragonfly network models.
     - Updated adaptive routing in megafly and 1-D dragonfly network models. 
     - Extended slim fly network model's dual-rail mode to arbitrary number of rails (pending).
+    - Implemented Quality of Service (QoS) in 1-D dragonfly network.
+    - Implemented changes needed to allow ROSS's tiebreaker mechanism.
 
 Nikhil Jain, Abhinav Bhatele (LLNL)
     - Improvements in credit-based flow control of CODES dragonfly and torus network models.
@@ -78,3 +82,17 @@ Caitlin Ross (RPI):
     - Added instrumentation so that network models can report sampled
       statistics over virtual time (pending).
     - Bug reporter for CODES models.
+
+Elkin Cruz (RPI)
+    - Added network surrogate for 1-D Dragonfly model (dragonfly-dally).
+    - Added application surrogate for MPI replay (model-net-mpi-replay).
+    - Implemented API to allow network and application surrogates to switch as
+      simulation runs (aka, hybrid simulation).
+    - Added network and application level directors, which coordinate data
+      transference between model and predictor.
+    - Added simple averaged-based network and application predictors (they are
+      given simulation data and are in charge of predicting future states of the
+      simulation, skipping computation).
+    - Implemented necessary scaffolding to check for bugs in reversible
+      computation (to be used with SEQUENTIAL_ROLLBACK_CHECK option in ROSS).
+    - Fixed reversible computation bugs on 1-D Dragonfly network.