Thread and task specific profiling

- Adds thread and task ids to profile samples
- Implements thread and task selection for Profile.print()
- Implements thread and task groupby options for Profile.print()
- Add include_meta to Profile.fetch() which defaults to false to ensure backwards compat with external profiling tooling

contrib/generate_precompile.jl

@@ -222,7 +222,11 @@ Profile = get(Base.loaded_modules,
           nothing)
 if Profile !== nothing
     hardcoded_precompile_statements *= """
-    precompile(Tuple{typeof(Profile.tree!), Profile.StackFrameTree{UInt64}, Vector{UInt64}, Dict{UInt64, Vector{Base.StackTraces.StackFrame}}, Bool, Symbol})
+    precompile(Tuple{typeof(Profile.tree!), Profile.StackFrameTree{UInt64}, Vector{UInt64}, Dict{UInt64, Vector{Base.StackTraces.StackFrame}}, Bool, Symbol, Int, UInt})
+    precompile(Tuple{typeof(Profile.tree!), Profile.StackFrameTree{UInt64}, Vector{UInt64}, Dict{UInt64, Vector{Base.StackTraces.StackFrame}}, Bool, Symbol, Int, UnitRange{UInt}})
+    precompile(Tuple{typeof(Profile.tree!), Profile.StackFrameTree{UInt64}, Vector{UInt64}, Dict{UInt64, Vector{Base.StackTraces.StackFrame}}, Bool, Symbol, UnitRange{Int}, UInt})
+    precompile(Tuple{typeof(Profile.tree!), Profile.StackFrameTree{UInt64}, Vector{UInt64}, Dict{UInt64, Vector{Base.StackTraces.StackFrame}}, Bool, Symbol, UnitRange{Int}, UnitRange{UInt}})
+    precompile(Tuple{typeof(Profile.tree!), Profile.StackFrameTree{UInt64}, Vector{UInt64}, Dict{UInt64, Vector{Base.StackTraces.StackFrame}}, Bool, Symbol, Vector{Int}, Vector{UInt}})
     """
 end
 

src/signals-mach.c

@@ -588,6 +588,13 @@ void *mach_profile_listener(void *arg)
 #else
                 bt_size_cur += rec_backtrace_ctx((jl_bt_element_t*)bt_data_prof + bt_size_cur, bt_size_max - bt_size_cur - 1, uc, NULL);
 #endif
+                jl_ptls_t ptls = jl_all_tls_states[i];
+
+                // store threadid but add 1 as 0 is preserved to indicate end of block
+                bt_data_prof[bt_size_cur++].uintptr = ptls->tid + 1;
+
+                // store task id
+                bt_data_prof[bt_size_cur++].uintptr = ptls->current_task;
 
                 // Mark the end of this block with 0
                 bt_data_prof[bt_size_cur++].uintptr = 0;

src/signals-unix.c

@@ -783,6 +783,14 @@ static void *signal_listener(void *arg)
                     }
                     jl_set_safe_restore(old_buf);
 
+                    jl_ptls_t ptls = jl_all_tls_states[i];
+
+                    // store threadid but add 1 as 0 is preserved to indicate end of block
+                    bt_data_prof[bt_size_cur++].uintptr = ptls->tid + 1;
+
+                    // store task id
+                    bt_data_prof[bt_size_cur++].uintptr = ptls->current_task;
+
                     // Mark the end of this block with 0
                     bt_data_prof[bt_size_cur++].uintptr = 0;
                 }

src/signals-win.c

@@ -360,6 +360,15 @@ static DWORD WINAPI profile_bt( LPVOID lparam )
                     // Get backtrace data
                     bt_size_cur += rec_backtrace_ctx((jl_bt_element_t*)bt_data_prof + bt_size_cur,
                             bt_size_max - bt_size_cur - 1, &ctxThread, NULL);
+
+                    jl_ptls_t ptls = jl_all_tls_states[0]; // given only profiling hMainThread
+
+                    // store threadid but add 1 as 0 is preserved to indicate end of block
+                    bt_data_prof[bt_size_cur++].uintptr = ptls->tid + 1;
+
+                    // store task id
+                    bt_data_prof[bt_size_cur++].uintptr = ptls->current_task;
+
                     // Mark the end of this block with 0
                     bt_data_prof[bt_size_cur++].uintptr = 0;
                 }