c/c++中类似于java jprofiler/eclispe memoryanalysis的代码性能瓶颈以及内存泄露分析工具
visual studio有自带的,可以看MSDN,不过一般来说,我们比较关注linux下的,搜了下,比较好用的应该有gprof和valgrind(可以分析内存潜在的泄露,如针对json-c/log4c进行分析),先记录。可参考如下:
http://blog.csdn.net/clarstyle/article/details/41747817?utm_source=itdadao&utm_medium=referral
对于性能分析,推荐gcc perf命令行工具(https://zhuanlan.zhihu.com/p/186208907,使用上这个更合适),没有之一。
perf record -a --call-graph dwarf -p PID
perf report
上述命令生成的就是调用链。
# To display the perf.data header info, please use --header/--header-only options. # # # Total Lost Samples: 0 # # Samples: 411K of event 'cycles:uppp' # Event count (approx.): 415626554486 # # Children Self Command Shared Object Symbol # ........ ........ ....... .................. ....................................... # 86.20% 7.98% lightdb lightdb [.] ReadBuffer_common | |--80.75%--ReadBuffer_common | | | --80.75%--BufferAlloc (inlined) | | | |--38.27%--StrategyGetBuffer | | | | | --37.50%--GetBufferFromRing (inlined) | | | | | |--14.59%--GetCurrCandidateNums | | | | | | | --11.59%--GetPgwrCandidateNums (inlined) | | | | | | | --0.89%--pg_atomic_read_u64 (inlined) | | | pg_atomic_read_u64_impl (inlined) | | | | | |--9.72%--LockBufHdr | | | | | | | |--8.18%--pg_atomic_fetch_or_u32 (inlined) | | | | pg_atomic_fetch_or_u32_impl (inlined) | | | | | | | --0.54%--init_spin_delay (inlined) | | | | | --3.85%--pg_atomic_read_u32 (inlined) | | pg_atomic_read_u32_impl (inlined)
除了动态链路链路外,还可用gnu cflow对单个静态c源文件(虽然单个,但是相比直接看源代码,还是方便很多的)分析调用链路。如下:
[lightdb@lightdb-dev plan]$ /usr/local/cflow/bin/cflow planner.c | more expression_planner() <Expr *expression_planner (Expr *expr) at planner.c:6319>: eval_const_expressions() fix_opfuncids() expression_planner_with_deps() <Expr *expression_planner_with_deps (Expr *expr, List **relationOids, List **invalItems) at planner.c:6346>: MemSet() eval_const_expressions() fix_opfuncids() extract_query_dependencies_walker() plan_cluster_use_sort() <bool plan_cluster_use_sort (Oid tableOid, Oid indexOid) at planner.c:6399>: makeNode() list_make1() setup_simple_rel_arrays() build_simple_rel() foreach() lfirst_node() get_relation_data_width() cost_qual_eval() create_seqscan_path() cost_sort() create_index_path() plan_create_index_workers() <int plan_create_index_workers (Oid tableOid, Oid indexOid) at planner.c:6517>: makeNode() list_make1() setup_simple_rel_arrays() build_simple_rel() table_open() index_open() is_parallel_safe() RelationGetIndexExpressions() RelationGetIndexPredicate() Min() estimate_rel_size() compute_parallel_worker() index_close() table_close() planner() <PlannedStmt *planner (Query *parse, const char *query_string, int cursorOptions, ParamListInfo boundParams) at planner.c:277>: standard_planner() <PlannedStmt *standard_planner (Query *parse, const char *query_string, int cursorOptions, ParamListInfo boundParams) at planner.c:290>: makeNode() IsParallelWorker() max_parallel_hazard() subquery_planner() <PlannerInfo *subquery_planner (PlannerGlobal *glob, Query *parse, PlannerInfo *parent_root, bool hasRecursion, double tuple_fraction) at planner.c:608>: makeNode() memset() assign_special_exec_param() rt_fetch() SS_process_ctes() replace_empty_jointree() pull_up_sublinks() preprocess_function_rtes() preprocess_rownum() pull_up_subqueries() flatten_simple_union_all() foreach() lfirst_node() has_subclass()
perf使用可参见:https://blog.csdn.net/jasonactions/article/details/109332167、https://perf.wiki.kernel.org/index.php/Main_Page