feat(dfr): add default configuration file for HPX runtime.

compiler/CMakeLists.txt

@@ -83,7 +83,10 @@ if(CONCRETELANG_PARALLEL_EXECUTION_ENABLED)
   find_package(HPX REQUIRED CONFIG)
   include_directories(SYSTEM ${HPX_INCLUDE_DIRS})
   list(APPEND CMAKE_MODULE_PATH "${HPX_CMAKE_DIR}")
-  add_compile_options(-DCONCRETELANG_PARALLEL_EXECUTION_ENABLED)
+  add_compile_options(
+    -DCONCRETELANG_PARALLEL_EXECUTION_ENABLED
+    -DHPX_DEFAULT_CONFIG_FILE="${PROJECT_SOURCE_DIR}/hpx.ini"
+  )
 
 else()
   message(STATUS "ConcreteLang parallel execution disabled.")

compiler/hpx.ini

@@ -0,0 +1,27 @@
+[hpx]
+location = ${HPX_LOCATION:$[system.prefix]}
+component_path = $[hpx.location]/lib/hpx:$[system.executable_prefix]/lib/hpx:$[system.executable_prefix]/../lib/hpx
+master_ini_path = $[hpx.location]/share/hpx-<version>:$[system.executable_prefix]/share/hpx-<version>:$[system.executable_prefix]/../share/hpx-<version>
+ini_path = $[hpx.master_ini_path]/ini
+os_threads = 1
+localities = 1
+program_name =
+cmd_line =
+lock_detection = ${HPX_LOCK_DETECTION:0}
+throw_on_held_lock = ${HPX_THROW_ON_HELD_LOCK:1}
+minimal_deadlock_detection = <debug>
+spinlock_deadlock_detection = <debug>
+spinlock_deadlock_detection_limit = ${HPX_SPINLOCK_DEADLOCK_DETECTION_LIMIT:1000000}
+max_background_threads = ${HPX_MAX_BACKGROUND_THREADS:$[hpx.os_threads]}
+max_idle_loop_count = ${HPX_MAX_IDLE_LOOP_COUNT:<hpx_idle_loop_count_max>}
+max_busy_loop_count = ${HPX_MAX_BUSY_LOOP_COUNT:<hpx_busy_loop_count_max>}
+max_idle_backoff_time = ${HPX_MAX_IDLE_BACKOFF_TIME:<hpx_idle_backoff_time_max>}
+exception_verbosity = ${HPX_EXCEPTION_VERBOSITY:1}
+default_stack_size = 0x20000000
+
+[hpx.stacks]
+small_size = 0x8000000
+medium_size = 0x10000000
+large_size = 0x20000000
+huge_size = 0x40000000
+use_guard_pages = ${HPX_THREAD_GUARD_PAGE:3}

compiler/lib/Runtime/DFRuntime.cpp

@@ -791,12 +791,21 @@ static inline void _dfr_start_impl(int argc, char *argv[]) {
     if (nCores < 1)
       nCores = 1;
 
+    // We do not directly handle this, but we should take into account
+    // the choices made by the OpenMP runtime if we would be mixing
+    // loop & dataflow parallelism.
     char *env = getenv("OMP_NUM_THREADS");
     if (env != nullptr)
       nOMPThreads = strtoul(env, NULL, 10);
     else
       nOMPThreads = nCores;
 
+    // Unless specified, we will consider that within each node loop
+    // parallelism is the priority, so we would allocate either
+    // ncores/OMP_NUM_THREADS or ncores-OMP_NUM_THREADS+1.  Both make
+    // sense depending on whether we have very regular computation or
+    // not - the latter being more conservative in that we will
+    // exploit all cores, but we may oversubscribe.
     env = getenv("DFR_NUM_THREADS");
     if (env != nullptr)
       nHPXThreads = strtoul(env, NULL, 10);
@@ -805,7 +814,16 @@ static inline void _dfr_start_impl(int argc, char *argv[]) {
     if (nHPXThreads < 1)
       nHPXThreads = 1;
 
+    // If the user does not provide their own config file, one is by
+    // default located at the root of the concrete-compiler directory.
     env = getenv("HPX_CONFIG_FILE");
+    // If no file is provided, try and check that the default is
+    // available - otherwise use a basic default configuration.
+#ifdef HPX_DEFAULT_CONFIG_FILE
+    if (env == nullptr)
+      if (access(HPX_DEFAULT_CONFIG_FILE, F_OK) == 0)
+        env = const_cast<char *>(HPX_DEFAULT_CONFIG_FILE);
+#endif
     if (env != nullptr) {
       int _argc = 3;
       char *_argv[3] = {const_cast<char *>("__dummy_dfr_HPX_program_name__"),
@@ -813,6 +831,11 @@ static inline void _dfr_start_impl(int argc, char *argv[]) {
                         const_cast<char *>(env)};
       hpx::start(nullptr, _argc, _argv);
     } else {
+      // Last resort configuration in case no config file could be
+      // identified, provide some default values that make (some)
+      // sense for homomorphic computations (stacks need to reflect
+      // the size of ciphertexts rather than simple cleartext
+      // scalars).
       std::string numHPXThreads = std::to_string(nHPXThreads);
       int _argc = 7;
       char *_argv[7] = {