Merge branch 'dev' into dev-mlu-runtime

Author: kilinchange

.gitmodules

@@ -19,3 +19,6 @@
 [submodule "src/09python_ffi/pybind11"]
 	path = src/09python_ffi/pybind11
 	url = [email protected]:pybind/pybind11.git
+[submodule "3rd-party/cccl"]
+	path = 3rd-party/cccl
+	url = [email protected]:NVIDIA/cccl.git

3rd-party/backward-cpp

@@ -105,4 +105,5 @@ add_subdirectory(src/05computation)
 add_subdirectory(src/06frontend)
 add_subdirectory(src/07onnx)
 add_subdirectory(src/08communication)
+add_subdirectory(src/08-01llm)
 add_subdirectory(src/09python_ffi)

3rd-party/cccl

@@ -168,7 +168,7 @@ executor = compiler.compile("cuda", "default", [])  # -------- 编译模型
 - [fmt 10.1.1](https://github.com/fmtlib/fmt/releases/tag/10.1.0)
 - [fmtlog v2.2.1](https://github.com/MengRao/fmtlog/releases/tag/v2.2.1)
 - [googletest v1.14.0](https://github.com/google/googletest/releases/tag/v1.14.0)
-- [backward-cpp v1.6](https://github.com/bombela/backward-cpp/releases/tag/v1.6)
+- [backward-cpp master](https://github.com/bombela/backward-cpp)
 - [result master](https://github.com/willowell/result)
 - [abseil-cpp 20230802.1](https://github.com/abseil/abseil-cpp/releases/tag/20230802.1)
 

CMakeLists.txt

@@ -11,6 +11,5 @@ file(GLOB_RECURSE COMMON_TEST test/*.cpp)
 if(COMMON_TEST)
     add_executable(common_test ${COMMON_TEST})
     add_test(common_test common_test)
-    target_link_libraries(common_test common GTest::gtest_main ${BACKWARD_ENABLE})
-    add_backward(common_test)
+    target_link_libraries(common_test common GTest::gtest_main Backward::Object)
 endif()

README.md

@@ -2,6 +2,7 @@
 #define RC_HPP
 
 #include <functional>
+#include <utility>
 
 namespace refactor {
 
@@ -18,7 +19,7 @@ namespace refactor {
         T *_value;
         struct Counter {
             size_t strong, weak;
-        } * _counter;
+        } *_counter;
 
         Rc(T *ptr, Counter *counter) noexcept
             : _value(ptr), _counter(counter) { inc(); }

src/00common/CMakeLists.txt

@@ -11,6 +11,5 @@ file(GLOB_RECURSE GRAPH_TOPO_TEST test/*.cpp)
 if(GRAPH_TOPO_TEST)
     add_executable(graph_topo_test ${GRAPH_TOPO_TEST})
     add_test(graph_topo_test graph_topo_test)
-    target_link_libraries(graph_topo_test graph_topo GTest::gtest_main ${BACKWARD_ENABLE})
-    add_backward(graph_topo_test)
+    target_link_libraries(graph_topo_test graph_topo GTest::gtest_main Backward::Object)
 endif()

src/00common/include/common/rc.hpp

@@ -2,21 +2,18 @@ cmake_minimum_required(VERSION 3.12 FATAL_ERROR)
 project(hardware VERSION 0.0.0 LANGUAGES CXX)
 message(STATUS "Project " ${PROJECT_NAME} " version " ${PROJECT_VERSION})
 
-# Source files
 file(GLOB_RECURSE HARDWARE_SRC src/*.cc src/*.cpp)
+add_library(hardware STATIC ${HARDWARE_SRC} ${HARDWARE_CUDA_SRC})
+target_link_libraries(hardware PUBLIC common)
+target_include_directories(hardware PUBLIC include)
 
 if(USE_CUDA)
-    file(GLOB_RECURSE HARDWARE_CUDA_SRC src/devices/nvidia/*.cu)
+    target_include_directories(hardware PUBLIC ${CMAKE_CUDA_TOOLKIT_INCLUDE_DIRECTORIES})
 endif()
 
-add_library(hardware STATIC ${HARDWARE_SRC} ${HARDWARE_CUDA_SRC} ${HARDWARE_BANG_SRC})
-target_link_libraries(hardware PUBLIC common)
-target_include_directories(hardware PUBLIC include)
-
 file(GLOB_RECURSE HARDWARE_TEST test/*.cpp)
 if(HARDWARE_TEST)
     add_executable(hardware_test ${HARDWARE_TEST})
     add_test(hardware_test hardware_test)
-    target_link_libraries(hardware_test hardware GTest::gtest_main ${BACKWARD_ENABLE})
-    add_backward(hardware_test)
+    target_link_libraries(hardware_test hardware GTest::gtest_main Backward::Object)
 endif()

src/01graph_topo/CMakeLists.txt

@@ -52,7 +52,7 @@ namespace refactor::hardware {
 
         virtual ~Device() = default;
         virtual Type type() const noexcept = 0;
-        virtual void setContext() const noexcept;
+        virtual void setContext() const;
 
         Arc<Blob> malloc(size_t);
         Arc<Blob> absorb(Arc<Blob> &&);