Fix thread index values (#287)

* Update PTL

- PTL submodule waits for threads to start before proceeding

* Initialize perfetto after init_bundle

- perfetto thread creation after pthread_create wrapped

* backtrace component update

- exclude gotcha call-tree

* callchain component update

- callchain::get sorts based on timestamp
- callchain::sample supports duplicate IPs (recursion)

* Bump version to 1.10.1

[ROCm/rocprofiler-systems commit: de9f0e4c10]

projects/rocprofiler-systems/VERSION

@@ -1 +1 @@
-1.10.0
+1.10.1

projects/rocprofiler-systems/source/lib/omnitrace/library.cpp

@@ -492,13 +492,6 @@ omnitrace_init_tooling_hidden()
 
     OMNITRACE_SCOPED_SAMPLING_ON_CHILD_THREADS(false);
 
-    // perfetto initialization
-    if(get_use_perfetto())
-    {
-        OMNITRACE_VERBOSE_F(1, "Setting up Perfetto...\n");
-        omnitrace::perfetto::setup();
-    }
-
     // ideally these have already been started
     omnitrace_preinit_hidden();
 
@@ -507,6 +500,13 @@ omnitrace_init_tooling_hidden()
 
     if(get_use_sampling()) sampling::block_signals();
 
+    // perfetto initialization
+    if(get_use_perfetto())
+    {
+        OMNITRACE_VERBOSE_F(1, "Setting up Perfetto...\n");
+        omnitrace::perfetto::setup();
+    }
+
     tasking::setup();
 
     if(get_use_causal()) causal::start_experimenting();

projects/rocprofiler-systems/source/lib/omnitrace/library/components/backtrace.cpp

@@ -120,7 +120,6 @@ backtrace::filter_and_patch(const std::vector<entry_type>& _data)
         const auto _npos          = std::string::npos;
         if(_keep_internal) return 1;
         if(_lbl.find("omnitrace_main") != _npos) return 0;
-        if(_lbl.find("omnitrace::common::") != _npos) return 0;
         if(_lbl.find("omnitrace::") != _npos) return 0;
         if(_lbl.find("tim::") != _npos) return 0;
         if(_lbl.find("DYNINST_") != _npos) return 0;
@@ -129,6 +128,7 @@ backtrace::filter_and_patch(const std::vector<entry_type>& _data)
         if(_lbl.find("roctracer_") != _npos) return -1;
         if(_lbl.find("perfetto::") != _npos) return -1;
         if(_lbl.find("protozero::") == 0) return -1;
+        if(_lbl.find("gotcha_") != _npos) return -1;
         return 1;
     };
 

projects/rocprofiler-systems/source/lib/omnitrace/library/components/callchain.cpp

@@ -120,6 +120,9 @@ callchain::get() const
             itr.second.pop_back();
     }
 
+    std::sort(_v.begin(), _v.end(),
+              [](const auto& _lhs, const auto& _rhs) { return _lhs.first < _rhs.first; });
+
     return _v;
 }
 
@@ -193,9 +196,15 @@ callchain::sample(int signo)
             auto _data      = record{};
             _data.timestamp = itr.get_time();
             _data.data.emplace_back(_ip);
+            bool _skip_ip = true;
             for(auto ditr : itr.get_callchain())
             {
-                if(ditr != _ip) _data.data.emplace_back(ditr);
+                // skip the first instance of current IP but allow after that since this
+                // might be a recursive call
+                if(ditr == _ip && _skip_ip)
+                    _skip_ip = false;
+                else
+                    _data.data.emplace_back(ditr);
                 if(_data.data.size() == _data.data.capacity()) break;
             }
             if(!_data.data.empty()) m_data.emplace_back(_data);

projects/rocprofiler-systems/source/lib/omnitrace/library/ptl.cpp

@@ -96,8 +96,9 @@ get_thread_pool_state()
 PTL::ThreadPool&
 get_thread_pool()
 {
-    static auto* _v = (get_thread_pool_state() = State::Active,
-                       new PTL::ThreadPool{ _thread_pool_cfg() });
+    static auto  _cfg = _thread_pool_cfg();
+    static auto* _v =
+        (get_thread_pool_state() = State::Active, new PTL::ThreadPool{ _cfg });
     return *_v;
 }
 }  // namespace
@@ -145,6 +146,7 @@ void
 setup()
 {
     OMNITRACE_SCOPED_THREAD_STATE(ThreadState::Internal);
+    OMNITRACE_SCOPED_SAMPLING_ON_CHILD_THREADS(false);
     (void) get_thread_pool();
 }