Implement firestarter kernel

.clang-format

@@ -47,9 +47,7 @@ UseTab:          Never
 
 SpacesInParentheses: false
 SpacesInAngles:  false
-SpaceInEmptyParentheses: false
 SpacesInCStyleCastParentheses: false
 SpaceAfterControlStatementKeyword: true
-SpaceBeforeAssignmentOperators: true
 ContinuationIndentWidth: 4
 ...

.gitignore

@@ -1,2 +1,3 @@
 build*
+.cache/
 python

CMakeLists.txt

@@ -70,12 +70,9 @@ else()
     message(STATUS "X86Adapt wasn't found, DDCM not available.")
 endif()
 
-if (USE_FIRESTARTER)
-    include(cmake/BuildFirestarter.cmake)
-    message(STATUS "Building WITH firestarter.")
-else()
-    message(STATUS "Building WITHOUT firestarter.")
-endif()
+# Allow the use of external metrics for firestarter
+set(FIRESTARTER_LINK_STATIC OFF)
+add_subdirectory(lib/FIRESTARTER)
 
 target_include_directories(roco2 INTERFACE include)
 
@@ -96,7 +93,6 @@ macro(Roco2Configuration CONF )
 
     target_link_libraries(
         roco2_${CONF}
-        roco2_main
         roco2_core
     )
 

include/roco2/kernels/firestarter.hpp

@@ -1,9 +1,11 @@
 #ifndef INCLUDE_ROCO2_KERNELS_FIRESTARTER_HPP
 #define INCLUDE_ROCO2_KERNELS_FIRESTARTER_HPP
 
+#include <roco2/chrono/util.hpp>
 #include <roco2/kernels/base_kernel.hpp>
 
-#include <roco2/chrono/util.hpp>
+#include <firestarter/Payload/CompiledPayload.hpp>
+#include <firestarter/Payload/Payload.hpp>
 
 namespace roco2
 {
@@ -24,15 +26,27 @@ namespace kernels
         }
 
     private:
+        /// Runs the firestarter kernel until a timepoint
+        /// \arg until The timepoint until the kernel should execute the high load function of
+        /// firestarter
         void run_kernel(roco2::chrono::time_point until) override;
 
-        int (*firestarter_function_)(void*);
-        int (*firestarter_init_)(void*);
-        int base_function_;
+        /// This function terminates the execution of firestarter by writing the load variable to
+        /// ::firestarter::LoadThreadWorkType::LoadStop after the time of the experiment elapsed.
+        /// \arg until The timepoint until the kernel should execute the high load function of
+        /// firestarter
+        /// \arg load_var the reference to the variable that termination the load of firestarter
+        static void stop_kernel(roco2::chrono::time_point until,
+                                ::firestarter::LoadThreadWorkType& load_var);
+
+        /// The variable shared across all threads that termination the load of firestarter
+        ::firestarter::LoadThreadWorkType load_var = ::firestarter::LoadThreadWorkType::LoadHigh;
 
-        const static param_type loop_count = 10000;
+        /// The unique ptr to the load and init function for firestarter
+        ::firestarter::payload::CompiledPayload::UniquePtr compiled_payload_ptr = { nullptr,
+                                                                                    nullptr };
     };
-}
-}
+} // namespace kernels
+} // namespace roco2
 
 #endif // INCLUDE_ROCO2_KERNELS_FIRESTARTER_HPP

include/roco2/memory/thread_local.hpp

@@ -1,15 +1,13 @@
 #ifndef INCLUDE_ROCO2_MEMORY_HPP
 #define INCLUDE_ROCO2_MEMORY_HPP
 
+#include <firestarter/LoadWorkerMemory.hpp>
 #include <roco2/memory/alignment_allocator.hpp>
-#include <roco2/log.hpp>
 
 #include <omp.h>
 
-#include <vector>
-#include <chrono>
-#include <thread>
 #include <memory>
+#include <vector>
 
 namespace roco2
 {
@@ -21,9 +19,8 @@ namespace detail
     {
         thread_local_memory()
         : vec_A(vec_size), vec_B(vec_size), vec_C(vec_size), vec_F(vec_size),
-          mat_A(mat_size * mat_size), mat_B(mat_size * mat_size),
-          mat_C(mat_size * mat_size), mem_buffer(mem_size),
-          firestarter_buffer(firestarter_size)
+          mat_A(mat_size * mat_size), mat_B(mat_size * mat_size), mat_C(mat_size * mat_size),
+          mem_buffer(mem_size)
         {
             for (std::size_t i = 0; i < vec_A.size(); ++i)
             {
@@ -45,12 +42,6 @@ namespace detail
                 mem_buffer[i] = i * 23 + 42;
             }
 
-            for (std::size_t i = 0; i < firestarter_buffer.size(); ++i)
-            {
-                firestarter_buffer[i] =
-                    0.25 + static_cast<double>(i % 9267) * 0.24738995982e-4;
-            }
-
             log::debug() << "Memory allocated and touched.";
         }
 
@@ -64,25 +55,20 @@ namespace detail
         std::vector<double> mat_C;
 
         std::vector<std::uint64_t> mem_buffer;
-        std::vector<double, AlignmentAllocator<double, 32>> firestarter_buffer;
+
+        ::firestarter::LoadWorkerMemory::UniquePtr firestarter_memory = { nullptr, nullptr };
 
         const static std::size_t vec_size = 1024;
         const static std::size_t mat_size = 512;
 
         // size of mem_buffer equals 64MB
-        const static std::size_t mem_size =
-            64 * 1024 * 1024 / sizeof(mem_buffer[0]);
-
-        // size of mem_buffer equals 160MB
-        const static std::size_t firestarter_size =
-            512 * 1024 * 1024 / sizeof(firestarter_buffer[0]);
+        const static std::size_t mem_size = 64 * 1024 * 1024 / sizeof(mem_buffer[0]);
     };
-}
+} // namespace detail
 
 static detail::thread_local_memory& thread_local_memory()
 {
-    static std::vector<std::unique_ptr<
-        detail::thread_local_memory>> memory(omp_get_max_threads());
+    static std::vector<std::unique_ptr<detail::thread_local_memory>> memory(omp_get_max_threads());
 
     auto& tld = memory[omp_get_thread_num()];
 
@@ -93,5 +79,5 @@ static detail::thread_local_memory& thread_local_memory()
 
     return *tld;
 }
-}
+} // namespace roco2
 #endif // INCLUDE_ROCO2_MEMORY_HPP

src/CMakeLists.txt

@@ -1,5 +1,6 @@
 set(roco2_core_SOURCES)
 list(APPEND roco2_core_SOURCES
+    main.cpp
     metrics.cpp
     cpu.cpp
     fastcpufreq.c
@@ -13,27 +14,25 @@ else()
     message(STATUS "Building WITHOUT asm kernels.")
 endif()
 
-if (USE_FIRESTARTER)
-    list(APPEND roco2_core_SOURCES firestarter_kernel.cpp)
-endif()
+list(APPEND roco2_core_SOURCES firestarter_kernel.cpp)
 
 add_library(roco2_core STATIC ${roco2_core_SOURCES})
-target_link_libraries(roco2_core PUBLIC roco2)
-if (USE_FIRESTARTER)
-    target_compile_definitions(roco2_core PUBLIC "HAS_FIRESTARTER")
-    target_link_libraries(roco2_core PUBLIC firestarter)
-    add_dependencies(roco2_core libfirestarter)
-endif()
+
+target_link_libraries(roco2_core PUBLIC roco2 Nitro::env)
+
+target_link_libraries(roco2_core PUBLIC
+    "-Wl,--whole-archive"
+    firestartercombined
+    "-Wl,--no-whole-archive"
+    )
+
 if (USE_SCOREP)
     option(ROCO2_HIGHLOW_INSTRUMENT_PHASES "instrument the individual high & low phases of the high-low-kernel" OFF)
     if (ROCO2_HIGHLOW_INSTRUMENT_PHASES)
         target_compile_definitions(roco2_core PUBLIC "ROCO2_HIGHLOW_INSTRUMENT_PHASES")
     endif()
 endif()
 
-add_library(roco2_main STATIC main.cpp)
-target_link_libraries(roco2_main PUBLIC roco2)
-
 add_subdirectory(configurations)
 
 add_subdirectory(test)

src/configurations/ariel/CMakeLists.txt

@@ -1,3 +1,3 @@
-if (X86Adapt_FOUND AND CpuFreq_FOUND AND USE_FIRESTARTER AND USE_ASM_KERNELS)
+if (X86Adapt_FOUND AND CpuFreq_FOUND AND USE_ASM_KERNELS)
     Roco2Configuration(ariel)
 endif()

src/configurations/charon/CMakeLists.txt

@@ -1,3 +1,3 @@
-if(CpuFreq_FOUND AND USE_FIRESTARTER AND USE_ASM_KERNELS)
+if(CpuFreq_FOUND AND USE_ASM_KERNELS)
     Roco2Configuration(charon)
 endif()

src/configurations/charon/experiment.cpp

@@ -85,9 +85,8 @@ void run_experiments(roco2::chrono::time_point starting_point, bool eta_only)
 
     roco2::experiments::const_lenght exp(experiment_startpoint, experiment_duration);
 
-    auto experiment = [&](auto& kernel, const auto& on) {
-        plan.push_back(roco2::task::experiment_task(exp, kernel, on));
-    };
+    auto experiment = [&](auto& kernel, const auto& on)
+    { plan.push_back(roco2::task::experiment_task(exp, kernel, on)); };
 
     auto setting = [&](auto lambda) { plan.push_back(roco2::task::make_lambda_task(lambda)); };
 

src/configurations/charon_turbo/CMakeLists.txt

@@ -1,4 +1,4 @@
-if(CpuFreq_FOUND AND USE_FIRESTARTER AND USE_ASM_KERNELS)
+if(CpuFreq_FOUND AND USE_ASM_KERNELS)
     Roco2Configuration(charon_turbo)
 endif()
 

src/configurations/conway/CMakeLists.txt

@@ -1,4 +1,4 @@
-if(CpuFreq_FOUND AND USE_FIRESTARTER AND USE_ASM_KERNELS)
+if(X86Adapt_FOUND AND CpuFreq_FOUND AND USE_ASM_KERNELS)
     Roco2Configuration(conway)
 endif()
 

src/configurations/example/CMakeLists.txt

@@ -1,3 +1,3 @@
-if(CpuFreq_FOUND AND USE_FIRESTARTER AND USE_ASM_KERNELS)
+if(X86Adapt_FOUND AND CpuFreq_FOUND AND USE_ASM_KERNELS)
     Roco2Configuration(example)
 endif()