8080112: Replace and remove the last usages of CollectedHeap::n_par_threads()

Reviewed-by: jmasa, kbarrett

hotspot/src/share/vm/gc/cms/compactibleFreeListSpace.cpp

@@ -641,6 +641,7 @@ void CompactibleFreeListSpace::set_end(HeapWord* value) {
 class FreeListSpace_DCTOC : public Filtering_DCTOC {
   CompactibleFreeListSpace* _cfls;
   CMSCollector* _collector;
+  bool _parallel;
 protected:
   // Override.
 #define walk_mem_region_with_cl_DECL(ClosureType)                       \
@@ -661,9 +662,10 @@ public:
                       CMSCollector* collector,
                       ExtendedOopClosure* cl,
                       CardTableModRefBS::PrecisionStyle precision,
-                      HeapWord* boundary) :
+                      HeapWord* boundary,
+                      bool parallel) :
     Filtering_DCTOC(sp, cl, precision, boundary),
-    _cfls(sp), _collector(collector) {}
+    _cfls(sp), _collector(collector), _parallel(parallel) {}
 };
 
 // We de-virtualize the block-related calls below, since we know that our
@@ -674,10 +676,7 @@ void FreeListSpace_DCTOC::walk_mem_region_with_cl(MemRegion mr,
                                                  HeapWord* bottom,              \
                                                  HeapWord* top,                 \
                                                  ClosureType* cl) {             \
-   bool is_par = GenCollectedHeap::heap()->n_par_threads() > 0;                 \
+   if (_parallel) {                                                             \
-   if (is_par) {                                                                \
-     assert(GenCollectedHeap::heap()->n_par_threads() ==                        \
-            GenCollectedHeap::heap()->workers()->active_workers(), "Mismatch"); \
      walk_mem_region_with_cl_par(mr, bottom, top, cl);                          \
    } else {                                                                     \
      walk_mem_region_with_cl_nopar(mr, bottom, top, cl);                        \
@@ -747,8 +746,9 @@ FreeListSpace_DCTOC__walk_mem_region_with_cl_DEFN(FilteringClosure)
 DirtyCardToOopClosure*
 CompactibleFreeListSpace::new_dcto_cl(ExtendedOopClosure* cl,
                                       CardTableModRefBS::PrecisionStyle precision,
-                                      HeapWord* boundary) {
+                                      HeapWord* boundary,
-  return new FreeListSpace_DCTOC(this, _collector, cl, precision, boundary);
+                                      bool parallel) {
+  return new FreeListSpace_DCTOC(this, _collector, cl, precision, boundary, parallel);
 }
 
 
@@ -1897,11 +1897,9 @@ CompactibleFreeListSpace::splitChunkAndReturnRemainder(FreeChunk* chunk,
   assert(chunk->is_free() && ffc->is_free(), "Error");
   _bt.split_block((HeapWord*)chunk, chunk->size(), new_size);
   if (rem_sz < SmallForDictionary) {
-    bool is_par = (GenCollectedHeap::heap()->n_par_threads() > 0);
+    // The freeList lock is held, but multiple GC task threads might be executing in parallel.
+    bool is_par = Thread::current()->is_GC_task_thread();
     if (is_par) _indexedFreeListParLocks[rem_sz]->lock();
-    assert(!is_par ||
-           (GenCollectedHeap::heap()->n_par_threads() ==
-            GenCollectedHeap::heap()->workers()->active_workers()), "Mismatch");
     returnChunkToFreeList(ffc);
     split(size, rem_sz);
     if (is_par) _indexedFreeListParLocks[rem_sz]->unlock();
@@ -1972,8 +1970,6 @@ void CompactibleFreeListSpace::save_marks() {
 
 bool CompactibleFreeListSpace::no_allocs_since_save_marks() {
   assert(_promoInfo.tracking(), "No preceding save_marks?");
-  assert(GenCollectedHeap::heap()->n_par_threads() == 0,
-         "Shouldn't be called if using parallel gc.");
   return _promoInfo.noPromotions();
 }
 
@@ -1981,8 +1977,6 @@ bool CompactibleFreeListSpace::no_allocs_since_save_marks() {
                                                                             \
 void CompactibleFreeListSpace::                                             \
 oop_since_save_marks_iterate##nv_suffix(OopClosureType* blk) {              \
-  assert(GenCollectedHeap::heap()->n_par_threads() == 0,                    \
-         "Shouldn't be called (yet) during parallel part of gc.");          \
   _promoInfo.promoted_oops_iterate##nv_suffix(blk);                         \
   /*                                                                        \
    * This also restores any displaced headers and removes the elements from \

hotspot/src/share/vm/gc/cms/compactibleFreeListSpace.hpp

@@ -438,7 +438,8 @@ class CompactibleFreeListSpace: public CompactibleSpace {
   // Override: provides a DCTO_CL specific to this kind of space.
   DirtyCardToOopClosure* new_dcto_cl(ExtendedOopClosure* cl,
                                      CardTableModRefBS::PrecisionStyle precision,
-                                     HeapWord* boundary);
+                                     HeapWord* boundary,
+                                     bool parallel);
 
   void blk_iterate(BlkClosure* cl);
   void blk_iterate_careful(BlkClosureCareful* cl);

hotspot/src/share/vm/gc/cms/parCardTableModRefBS.cpp

@@ -39,16 +39,11 @@
 void CardTableModRefBS::non_clean_card_iterate_parallel_work(Space* sp, MemRegion mr,
                                                              OopsInGenClosure* cl,
                                                              CardTableRS* ct,
-                                                             int n_threads) {
+                                                             uint n_threads) {
-  assert(n_threads > 0, "Error: expected n_threads > 0");
+  assert(n_threads > 0, "expected n_threads > 0");
-  assert((n_threads == 1 && ParallelGCThreads == 0) ||
+  assert(n_threads <= ParallelGCThreads,
-         n_threads <= (int)ParallelGCThreads,
+         err_msg("n_threads: %u > ParallelGCThreads: " UINTX_FORMAT, n_threads, ParallelGCThreads));
-         "# worker threads != # requested!");
+
-  assert(!Thread::current()->is_VM_thread() || (n_threads == 1), "There is only 1 VM thread");
-  assert(UseDynamicNumberOfGCThreads ||
-         !FLAG_IS_DEFAULT(ParallelGCThreads) ||
-         n_threads == (int)ParallelGCThreads,
-         "# worker threads != # requested!");
   // Make sure the LNC array is valid for the space.
   jbyte**   lowest_non_clean;
   uintptr_t lowest_non_clean_base_chunk_index;
@@ -66,7 +61,8 @@ void CardTableModRefBS::non_clean_card_iterate_parallel_work(Space* sp, MemRegio
 
   uint stride = 0;
   while (!pst->is_task_claimed(/* reference */ stride)) {
-    process_stride(sp, mr, stride, n_strides, cl, ct,
+    process_stride(sp, mr, stride, n_strides,
+                   cl, ct,
                    lowest_non_clean,
                    lowest_non_clean_base_chunk_index,
                    lowest_non_clean_chunk_size);
@@ -132,9 +128,13 @@ process_stride(Space* sp,
     assert(chunk_mr.word_size() > 0, "[chunk_card_start > used_end)");
     assert(used.contains(chunk_mr), "chunk_mr should be subset of used");
 
+    // This function is used by the parallel card table iteration.
+    const bool parallel = true;
+
     DirtyCardToOopClosure* dcto_cl = sp->new_dcto_cl(cl, precision(),
-                                                     cl->gen_boundary());
+                                                     cl->gen_boundary(),
-    ClearNoncleanCardWrapper clear_cl(dcto_cl, ct);
+                                                     parallel);
+    ClearNoncleanCardWrapper clear_cl(dcto_cl, ct, parallel);
 
 
     // Process the chunk.

hotspot/src/share/vm/gc/g1/concurrentMark.cpp

@@ -1938,15 +1938,11 @@ void ConcurrentMark::cleanup() {
 
   HeapRegionRemSet::reset_for_cleanup_tasks();
 
-  uint n_workers;
-
   // Do counting once more with the world stopped for good measure.
   G1ParFinalCountTask g1_par_count_task(g1h, &_region_bm, &_card_bm);
 
   g1h->set_par_threads();
-  n_workers = g1h->n_par_threads();
+  uint n_workers = _g1h->workers()->active_workers();
-  assert(g1h->n_par_threads() == n_workers,
-         "Should not have been reset");
   g1h->workers()->run_task(&g1_par_count_task);
   // Done with the parallel phase so reset to 0.
   g1h->set_par_threads(0);

hotspot/src/share/vm/gc/serial/defNewGeneration.cpp

@@ -455,7 +455,7 @@ void DefNewGeneration::compute_new_size() {
   }
 }
 
-void DefNewGeneration::younger_refs_iterate(OopsInGenClosure* cl) {
+void DefNewGeneration::younger_refs_iterate(OopsInGenClosure* cl, uint n_threads) {
   assert(false, "NYI -- are you sure you want to call this?");
 }
 

hotspot/src/share/vm/gc/serial/defNewGeneration.hpp

@@ -255,7 +255,7 @@ protected:
   // Iteration
   void object_iterate(ObjectClosure* blk);
 
-  void younger_refs_iterate(OopsInGenClosure* cl);
+  void younger_refs_iterate(OopsInGenClosure* cl, uint n_threads);
 
   void space_iterate(SpaceClosure* blk, bool usedOnly = false);
 

hotspot/src/share/vm/gc/shared/cardGeneration.cpp

@@ -353,8 +353,8 @@ void CardGeneration::space_iterate(SpaceClosure* blk,
   blk->do_space(space());
 }
 
-void CardGeneration::younger_refs_iterate(OopsInGenClosure* blk) {
+void CardGeneration::younger_refs_iterate(OopsInGenClosure* blk, uint n_threads) {
   blk->set_generation(this);
-  younger_refs_in_space_iterate(space(), blk);
+  younger_refs_in_space_iterate(space(), blk, n_threads);
   blk->reset_generation();
 }

hotspot/src/share/vm/gc/shared/cardGeneration.hpp

@@ -89,7 +89,7 @@ class CardGeneration: public Generation {
 
   void space_iterate(SpaceClosure* blk, bool usedOnly = false);
 
-  void younger_refs_iterate(OopsInGenClosure* blk);
+  void younger_refs_iterate(OopsInGenClosure* blk, uint n_threads);
 
   bool is_in(const void* p) const;
 

hotspot/src/share/vm/gc/shared/cardTableModRefBS.cpp

@@ -440,31 +440,11 @@ void CardTableModRefBS::write_ref_field_work(void* field, oop newVal, bool relea
 void CardTableModRefBS::non_clean_card_iterate_possibly_parallel(Space* sp,
                                                                  MemRegion mr,
                                                                  OopsInGenClosure* cl,
-                                                                 CardTableRS* ct) {
+                                                                 CardTableRS* ct,
+                                                                 uint n_threads) {
   if (!mr.is_empty()) {
-    // Caller (process_roots()) claims that all GC threads
+    if (n_threads > 0) {
-    // execute this call.  With UseDynamicNumberOfGCThreads now all
-    // active GC threads execute this call.  The number of active GC
-    // threads needs to be passed to par_non_clean_card_iterate_work()
-    // to get proper partitioning and termination.
-    //
-    // This is an example of where n_par_threads() is used instead
-    // of workers()->active_workers().  n_par_threads can be set to 0 to
-    // turn off parallelism.  For example when this code is called as
-    // part of verification during root processing then n_par_threads()
-    // may have been set to 0. active_workers is not overloaded with
-    // the meaning that it is a switch to disable parallelism and so keeps
-    // the meaning of the number of active gc workers. If parallelism has
-    // not been shut off by setting n_par_threads to 0, then n_par_threads
-    // should be equal to active_workers.  When a different mechanism for
-    // shutting off parallelism is used, then active_workers can be used in
-    // place of n_par_threads.
-    int n_threads =  GenCollectedHeap::heap()->n_par_threads();
-    bool is_par = n_threads > 0;
-    if (is_par) {
 #if INCLUDE_ALL_GCS
-      assert(GenCollectedHeap::heap()->n_par_threads() ==
-             GenCollectedHeap::heap()->workers()->active_workers(), "Mismatch");
       non_clean_card_iterate_parallel_work(sp, mr, cl, ct, n_threads);
 #else  // INCLUDE_ALL_GCS
       fatal("Parallel gc not supported here.");
@@ -472,8 +452,11 @@ void CardTableModRefBS::non_clean_card_iterate_possibly_parallel(Space* sp,
     } else {
       // clear_cl finds contiguous dirty ranges of cards to process and clear.
 
-      DirtyCardToOopClosure* dcto_cl = sp->new_dcto_cl(cl, precision(), cl->gen_boundary());
+      // This is the single-threaded version.
-      ClearNoncleanCardWrapper clear_cl(dcto_cl, ct);
+      const bool parallel = false;
+
+      DirtyCardToOopClosure* dcto_cl = sp->new_dcto_cl(cl, precision(), cl->gen_boundary(), parallel);
+      ClearNoncleanCardWrapper clear_cl(dcto_cl, ct, parallel);
 
       clear_cl.do_MemRegion(mr);
     }

hotspot/src/share/vm/gc/shared/cardTableModRefBS.hpp

@@ -178,14 +178,15 @@ class CardTableModRefBS: public ModRefBarrierSet {
   // region mr in the given space and apply cl to any dirty sub-regions
   // of mr. Clears the dirty cards as they are processed.
   void non_clean_card_iterate_possibly_parallel(Space* sp, MemRegion mr,
-                                                OopsInGenClosure* cl, CardTableRS* ct);
+                                                OopsInGenClosure* cl, CardTableRS* ct,
+                                                uint n_threads);
 
  private:
   // Work method used to implement non_clean_card_iterate_possibly_parallel()
   // above in the parallel case.
   void non_clean_card_iterate_parallel_work(Space* sp, MemRegion mr,
                                             OopsInGenClosure* cl, CardTableRS* ct,
-                                            int n_threads);
+                                            uint n_threads);
 
  protected:
   // Dirty the bytes corresponding to "mr" (not all of which must be

hotspot/src/share/vm/gc/shared/cardTableRS.cpp

@@ -102,9 +102,10 @@ void CardTableRS::prepare_for_younger_refs_iterate(bool parallel) {
 }
 
 void CardTableRS::younger_refs_iterate(Generation* g,
-                                       OopsInGenClosure* blk) {
+                                       OopsInGenClosure* blk,
+                                       uint n_threads) {
   _last_cur_val_in_gen[g->level()+1] = cur_youngergen_card_val();
-  g->younger_refs_iterate(blk);
+  g->younger_refs_iterate(blk, n_threads);
 }
 
 inline bool ClearNoncleanCardWrapper::clear_card(jbyte* entry) {
@@ -164,15 +165,8 @@ inline bool ClearNoncleanCardWrapper::clear_card_serial(jbyte* entry) {
 }
 
 ClearNoncleanCardWrapper::ClearNoncleanCardWrapper(
-  DirtyCardToOopClosure* dirty_card_closure, CardTableRS* ct) :
+  DirtyCardToOopClosure* dirty_card_closure, CardTableRS* ct, bool is_par) :
-    _dirty_card_closure(dirty_card_closure), _ct(ct) {
+    _dirty_card_closure(dirty_card_closure), _ct(ct), _is_par(is_par) {
-    // Cannot yet substitute active_workers for n_par_threads
-    // in the case where parallelism is being turned off by
-    // setting n_par_threads to 0.
-    _is_par = (GenCollectedHeap::heap()->n_par_threads() > 0);
-    assert(!_is_par ||
-           (GenCollectedHeap::heap()->n_par_threads() ==
-            GenCollectedHeap::heap()->workers()->active_workers()), "Mismatch");
 }
 
 bool ClearNoncleanCardWrapper::is_word_aligned(jbyte* entry) {
@@ -272,7 +266,8 @@ void CardTableRS::write_ref_field_gc_par(void* field, oop new_val) {
 }
 
 void CardTableRS::younger_refs_in_space_iterate(Space* sp,
-                                                OopsInGenClosure* cl) {
+                                                OopsInGenClosure* cl,
+                                                uint n_threads) {
   const MemRegion urasm = sp->used_region_at_save_marks();
 #ifdef ASSERT
   // Convert the assertion check to a warning if we are running
@@ -301,7 +296,7 @@ void CardTableRS::younger_refs_in_space_iterate(Space* sp,
     ShouldNotReachHere();
   }
 #endif
-  _ct_bs->non_clean_card_iterate_possibly_parallel(sp, urasm, cl, this);
+  _ct_bs->non_clean_card_iterate_possibly_parallel(sp, urasm, cl, this, n_threads);
 }
 
 void CardTableRS::clear_into_younger(Generation* old_gen) {

hotspot/src/share/vm/gc/shared/cardTableRS.hpp

@@ -56,7 +56,7 @@ class CardTableRS: public GenRemSet {
 
   CardTableModRefBSForCTRS* _ct_bs;
 
-  virtual void younger_refs_in_space_iterate(Space* sp, OopsInGenClosure* cl);
+  virtual void younger_refs_in_space_iterate(Space* sp, OopsInGenClosure* cl, uint n_threads);
 
   void verify_space(Space* s, HeapWord* gen_start);
 
@@ -116,7 +116,7 @@ public:
   // Card table entries are cleared before application; "blk" is
   // responsible for dirtying if the oop is still older-to-younger after
   // closure application.
-  void younger_refs_iterate(Generation* g, OopsInGenClosure* blk);
+  void younger_refs_iterate(Generation* g, OopsInGenClosure* blk, uint n_threads);
 
   void inline_write_ref_field_gc(void* field, oop new_val) {
     jbyte* byte = _ct_bs->byte_for(field);
@@ -183,7 +183,7 @@ private:
   bool is_word_aligned(jbyte* entry);
 
 public:
-  ClearNoncleanCardWrapper(DirtyCardToOopClosure* dirty_card_closure, CardTableRS* ct);
+  ClearNoncleanCardWrapper(DirtyCardToOopClosure* dirty_card_closure, CardTableRS* ct, bool is_par);
   void do_MemRegion(MemRegion mr);
 };
 

hotspot/src/share/vm/gc/shared/collectedHeap.cpp

@@ -160,8 +160,7 @@ void CollectedHeap::trace_heap_after_gc(const GCTracer* gc_tracer) {
 // Memory state functions.
 
 
-CollectedHeap::CollectedHeap() : _n_par_threads(0)
+CollectedHeap::CollectedHeap() {
-{
   const size_t max_len = size_t(arrayOopDesc::max_array_length(T_INT));
   const size_t elements_per_word = HeapWordSize / sizeof(jint);
   _filler_array_max_size = align_object_size(filler_array_hdr_size() +

hotspot/src/share/vm/gc/shared/collectedHeap.hpp

@@ -101,7 +101,6 @@ class CollectedHeap : public CHeapObj<mtInternal> {
  protected:
   BarrierSet* _barrier_set;
   bool _is_gc_active;
-  uint _n_par_threads;
 
   unsigned int _total_collections;          // ... started
   unsigned int _total_full_collections;     // ... started
@@ -291,11 +290,8 @@ class CollectedHeap : public CHeapObj<mtInternal> {
   }
   GCCause::Cause gc_cause() { return _gc_cause; }
 
-  // Number of threads currently working on GC tasks.
-  uint n_par_threads() { return _n_par_threads; }
-
   // May be overridden to set additional parallelism.
-  virtual void set_par_threads(uint t) { _n_par_threads = t; };
+  virtual void set_par_threads(uint t) { (void)t; };
 
   // General obj/array allocation facilities.
   inline static oop obj_allocate(KlassHandle klass, int size, TRAPS);

hotspot/src/share/vm/gc/shared/genCollectedHeap.cpp

@@ -700,7 +700,7 @@ void GenCollectedHeap::gen_process_roots(StrongRootsScope* scope,
   // older-gen scanning.
   if (level == 0) {
     older_gens->set_generation(_old_gen);
-    rem_set()->younger_refs_iterate(_old_gen, older_gens);
+    rem_set()->younger_refs_iterate(_old_gen, older_gens, scope->n_threads());
     older_gens->reset_generation();
   }
 

hotspot/src/share/vm/gc/shared/genRemSet.hpp

@@ -77,10 +77,11 @@ public:
   //  1) that are in objects allocated in "g" at the time of the last call
   //     to "save_Marks", and
   //  2) that point to objects in younger generations.
-  virtual void younger_refs_iterate(Generation* g, OopsInGenClosure* blk) = 0;
+  virtual void younger_refs_iterate(Generation* g, OopsInGenClosure* blk, uint n_threads) = 0;
 
   virtual void younger_refs_in_space_iterate(Space* sp,
-                                             OopsInGenClosure* cl) = 0;
+                                             OopsInGenClosure* cl,
+                                             uint n_threads) = 0;
 
   // This method is used to notify the remembered set that "new_val" has
   // been written into "field" by the garbage collector.

hotspot/src/share/vm/gc/shared/generation.cpp

@@ -293,9 +293,10 @@ void Generation::oop_iterate(ExtendedOopClosure* cl) {
 }
 
 void Generation::younger_refs_in_space_iterate(Space* sp,
-                                               OopsInGenClosure* cl) {
+                                               OopsInGenClosure* cl,
+                                               uint n_threads) {
   GenRemSet* rs = GenCollectedHeap::heap()->rem_set();
-  rs->younger_refs_in_space_iterate(sp, cl);
+  rs->younger_refs_in_space_iterate(sp, cl, n_threads);
 }
 
 class GenerationObjIterateClosure : public SpaceClosure {

hotspot/src/share/vm/gc/shared/generation.hpp

@@ -122,7 +122,7 @@ class Generation: public CHeapObj<mtGC> {
   // The iteration is only over objects allocated at the start of the
   // iterations; objects allocated as a result of applying the closure are
   // not included.
-  void younger_refs_in_space_iterate(Space* sp, OopsInGenClosure* cl);
+  void younger_refs_in_space_iterate(Space* sp, OopsInGenClosure* cl, uint n_threads);
 
  public:
   // The set of possible generation kinds.
@@ -526,7 +526,7 @@ class Generation: public CHeapObj<mtGC> {
   // in the current generation that contain pointers to objects in younger
   // generations. Objects allocated since the last "save_marks" call are
   // excluded.
-  virtual void younger_refs_iterate(OopsInGenClosure* cl) = 0;
+  virtual void younger_refs_iterate(OopsInGenClosure* cl, uint n_threads) = 0;
 
   // Inform a generation that it longer contains references to objects
   // in any younger generation.    [e.g. Because younger gens are empty,

hotspot/src/share/vm/gc/shared/space.cpp

@@ -181,7 +181,8 @@ void DirtyCardToOopClosure::do_MemRegion(MemRegion mr) {
 
 DirtyCardToOopClosure* Space::new_dcto_cl(ExtendedOopClosure* cl,
                                           CardTableModRefBS::PrecisionStyle precision,
-                                          HeapWord* boundary) {
+                                          HeapWord* boundary,
+                                          bool parallel) {
   return new DirtyCardToOopClosure(this, cl, precision, boundary);
 }
 
@@ -260,7 +261,8 @@ ContiguousSpaceDCTOC__walk_mem_region_with_cl_DEFN(FilteringClosure)
 DirtyCardToOopClosure*
 ContiguousSpace::new_dcto_cl(ExtendedOopClosure* cl,
                              CardTableModRefBS::PrecisionStyle precision,
-                             HeapWord* boundary) {
+                             HeapWord* boundary,
+                             bool parallel) {
   return new ContiguousSpaceDCTOC(this, cl, precision, boundary);
 }
 

hotspot/src/share/vm/gc/shared/space.hpp

@@ -183,7 +183,8 @@ class Space: public CHeapObj<mtGC> {
   // operate. ResourceArea allocated.
   virtual DirtyCardToOopClosure* new_dcto_cl(ExtendedOopClosure* cl,
                                              CardTableModRefBS::PrecisionStyle precision,
-                                             HeapWord* boundary = NULL);
+                                             HeapWord* boundary,
+                                             bool parallel);
 
   // If "p" is in the space, returns the address of the start of the
   // "block" that contains "p".  We say "block" instead of "object" since
@@ -629,7 +630,8 @@ class ContiguousSpace: public CompactibleSpace {
   // Override.
   DirtyCardToOopClosure* new_dcto_cl(ExtendedOopClosure* cl,
                                      CardTableModRefBS::PrecisionStyle precision,
-                                     HeapWord* boundary = NULL);
+                                     HeapWord* boundary,
+                                     bool parallel);
 
   // Apply "blk->do_oop" to the addresses of all reference fields in objects
   // starting with the _saved_mark_word, which was noted during a generation's