Perform scavenge in idle tasks.

src/heap/spaces.cc

 // Copyright 2011 the V8 project authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #include "src/heap/spaces.h"
 
 #include "src/base/bits.h"
 #include "src/base/platform/platform.h"
 #include "src/full-codegen/full-codegen.h"
 #include "src/heap/slots-buffer.h"
@@ skipping 1450 matching lines @@
   }
 }
 
 
 void NewSpace::UpdateInlineAllocationLimit(int size_in_bytes) {
   if (heap()->inline_allocation_disabled()) {
     // Lowest limit when linear allocation was disabled.
     Address high = to_space_.page_high();
     Address new_top = allocation_info_.top() + size_in_bytes;
     allocation_info_.set_limit(Min(new_top, high));
   } else if (inline_allocation_limit_step() == 0) {

[Hannes Payer (out of office), 2015/09/21 09:00:57]

This case is now dead code.

[ulan, 2015/09/23 09:58:38]

Done.

     // Normal limit is the end of the current page.
     allocation_info_.set_limit(to_space_.page_high());
   } else {
     // Lower limit during incremental marking.
     Address high = to_space_.page_high();
     Address new_top = allocation_info_.top() + size_in_bytes;
     Address new_limit = new_top + inline_allocation_limit_step_;
     allocation_info_.set_limit(Min(new_limit, high));
   }
   DCHECK_SEMISPACE_ALLOCATION_INFO(allocation_info_, to_space_);
@@ skipping 63 matching lines @@
     top_on_previous_step_ = old_top;
 
     high = to_space_.page_high();
     filler_size = Heap::GetFillToAlign(old_top, alignment);
     aligned_size_in_bytes = size_in_bytes + filler_size;
   }
 
   DCHECK(old_top + aligned_size_in_bytes < high);
 
   if (allocation_info_.limit() < high) {
     // Either the limit has been lowered because linear allocation was disabled

[Hannes Payer (out of office), 2015/09/21 09:00:57]

Update this comment (maybe some other comments also needs updates after your
change).

[ulan, 2015/09/23 09:58:38]

Done.

     // or because incremental marking wants to get a chance to do a step. Set
     // the new limit accordingly.
     Address new_top = old_top + aligned_size_in_bytes;
     int bytes_allocated = static_cast<int>(new_top - top_on_previous_step_);
+    heap()->ScheduleIdleScavenge(bytes_allocated);

[Hannes Payer (out of office), 2015/09/21 09:00:57]

Hmm, when incremental marking is on, steps will be really small and we are going
to trigger many idle tasks for scavenge. I don't think we should do that. Can we
just invoke this method on ScavengeJob::kBytesAllocatedBeforeNextIdleTask
allocations? WDYT?

[ulan, 2015/09/23 09:58:38]

The function already checks the bytes allocated. Changed the name accordingly.

     heap()->incremental_marking()->Step(bytes_allocated,
                                         IncrementalMarking::GC_VIA_STACK_GUARD);
     UpdateInlineAllocationLimit(aligned_size_in_bytes);
     top_on_previous_step_ = new_top;
   }
   return true;
 }
 
 
 #ifdef VERIFY_HEAP
@@ skipping 1614 matching lines @@
     object->ShortPrint();
     PrintF("\n");
   }
   printf(" --------------------------------------\n");
   printf(" Marked: %x, LiveCount: %x\n", mark_size, LiveBytes());
 }
 
 #endif  // DEBUG
 }  // namespace internal
 }  // namespace v8