ThreadFilter optim - fixes

- Fix removal of self in timerloop init
it was not using a slotID but a thread ID

- Add assertion to find other potential issues

Author: r1viollet

ddprof-lib/src/main/cpp/branch_hints.h

@@ -16,7 +16,7 @@
 
 #include <assert.h>
 
-#include "branch_hints.h"
+#include "arch_dd.h"
 #include "context.h"
 #include "counters.h"
 #include "engine.h"

ddprof-lib/src/main/cpp/javaApi.cpp

@@ -20,9 +20,11 @@
 // - These methods assume slot_id comes from registerThread() (undefined behavior otherwise)
 
 #include "threadFilter.h"
-#include "branch_hints.h"
+#include "arch_dd.h"
 
+#include <cassert>
 #include <cstdlib>
+#include <cstdio>
 #include <thread>
 #include <algorithm>
 #include <cstring>
@@ -181,10 +183,17 @@ void ThreadFilter::remove(SlotID slot_id) {
     int chunk_idx = slot_id >> kChunkShift;
     int slot_idx = slot_id & kChunkMask;
 
+    if (unlikely(chunk_idx >= kMaxChunks)) {
+        assert(false && "Invalid slot_id in ThreadFilter::remove - should not happen after wall clock fix");
+        return;
+    }
+
     ChunkStorage* chunk = _chunks[chunk_idx].load(std::memory_order_relaxed);
-    if (likely(chunk != nullptr)) {
-        chunk->slots[slot_idx].value.store(-1, std::memory_order_release);
+    if (unlikely(chunk == nullptr)) {
+        return;
     }
+    
+    chunk->slots[slot_idx].value.store(-1, std::memory_order_release);
 }
 
 void ThreadFilter::unregisterThread(SlotID slot_id) {

ddprof-lib/src/main/cpp/threadFilter.cpp

@@ -19,7 +19,7 @@
 
 #include <atomic>
 #include <unordered_set>
-#include "branch_hints.h"
+#include "arch_dd.h"
 
 // Simple fixed size thread ID table
 class ThreadIdTable {

ddprof-lib/src/main/cpp/threadIdTable.h

@@ -21,6 +21,7 @@
 #include "os.h"
 #include "profiler.h"
 #include "reservoirSampler.h"
+#include "thread.h"
 #include "threadFilter.h"
 #include "threadState.h"
 #include "tsc.h"
@@ -67,7 +68,15 @@ class BaseWallClock : public Engine {
       threads.reserve(reservoirSize);
       int self = OS::threadId();
       ThreadFilter* thread_filter = Profiler::instance()->threadFilter();
-      thread_filter->remove(self);
+      
+      // We don't want to profile ourselves in wall time
+      ProfiledThread* current = ProfiledThread::current();
+      if (current != nullptr) {
+        int slot_id = current->filterSlotId();
+        if (slot_id != -1) {
+          thread_filter->remove(slot_id);
+        }
+      }
 
       u64 startTime = TSC::ticks();
       WallClockEpochEvent epoch(startTime);