Add GCTracer metrics for a scavenger GC for DOM wrappers

Add GCTracer metrics for a scavenger GC for DOM wrappers

This patch adds the following three metrics for the --trace_gc_nvp option.

nodes_died_in_new_space_; // Number of died nodes in the new space.
nodes_copied_in_new_space_; // Number of copied nodes to the new space.
nodes_promoted; // Number of promoted nodes to the old space.

BUG=
TEST=Manually confirmed that the "--trace_gc --trace_gc_nvp" option prints the metrics

Committed: http://code.google.com/p/v8/source/detail?r=13159

[haraken, 2012-11-08 11:28:12]

These metrics are important to measure scavenger GC behaviors for DOM wrappers.

Could you take a look?

[Michael Starzinger, 2012-11-16 13:45:51]

https://codereview.chromium.org/11365146/diff/1/src/global-handles.cc
File src/global-handles.cc (right):

https://codereview.chromium.org/11365146/diff/1/src/global-handles.cc#newcode555
src/global-handles.cc:555: GarbageCollector collector, GCTracer* tracer) {
I am kind of hesitant to let the GCTracer "escape" into the global handles
subsystem. Is it important for the measurements to have per GC-cycle counts?

If not, then I would suggest to use normal per-isolate counters instead (i.e.
v8-counters.h) to gather those numbers.

I wouldn't use histograms though, because apparently they are automatically
exposed via UMA and these plain measurements are not useful when aggregated over
all pages and all users.

[haraken, 2012-11-19 07:45:43]

Thanks for review!

https://codereview.chromium.org/11365146/diff/1/src/global-handles.cc
File src/global-handles.cc (right):

https://codereview.chromium.org/11365146/diff/1/src/global-handles.cc#newcode555
src/global-handles.cc:555: GarbageCollector collector, GCTracer* tracer) {
> Is it important for the measurements to have per GC-cycle counts?

I think yes. What I am interested in is how many global handles are
disposed/copied/promoted in each scavenger/full GC.

> I am kind of hesitant to let the GCTracer "escape" into the global handles
> subsystem.

Then how about adding a couple of methods to global-handles.cc and having
GCTracer count up numbers in Heap::PerformGarbageCollection()?

> I wouldn't use histograms though, because apparently they are automatically
> exposed via UMA and these plain measurements are not useful when aggregated
over
> all pages and all users.

Agreed. I am not interested in UMAs for these metrics.

[Michael Starzinger, 2012-12-05 14:26:14]

I see. So letting the GCTracer escape is probably the simplest solution after
all. I can live with it. One round of reviews, then we should be good to go.

https://codereview.chromium.org/11365146/diff/1/src/heap.cc
File src/heap.cc (right):

https://codereview.chromium.org/11365146/diff/1/src/heap.cc#newcode7200
src/heap.cc:7200: PrintF("copied_nodes_in_new=%" V8_PTR_PREFIX "d ",
copied_nodes_in_new_space_);
More than 80 characters.

https://codereview.chromium.org/11365146/diff/1/src/heap.h
File src/heap.h (right):

https://codereview.chromium.org/11365146/diff/1/src/heap.h#newcode2617
src/heap.h:2617: int died_nodes_in_new_space_;
Can we rename this to "nodes_died_in_new_space_" so that it is using a common
prefix for all the global-handle related counters.

https://codereview.chromium.org/11365146/diff/1/src/heap.h#newcode2620
src/heap.h:2620: int copied_nodes_in_new_space_;
Likewise with "nodes_copied_in_new_space_".

https://codereview.chromium.org/11365146/diff/1/src/heap.h#newcode2623
src/heap.h:2623: int promoted_nodes_;
Likewise with "nodes_promoted_"

[haraken, 2012-12-06 11:08:15]

Thanks for reviewing!

https://codereview.chromium.org/11365146/diff/1/src/heap.cc
File src/heap.cc (right):

https://codereview.chromium.org/11365146/diff/1/src/heap.cc#newcode7200
src/heap.cc:7200: PrintF("copied_nodes_in_new=%" V8_PTR_PREFIX "d ",
copied_nodes_in_new_space_);
On 2012/12/05 14:26:14, Michael Starzinger wrote:
> More than 80 characters.

Done.

https://codereview.chromium.org/11365146/diff/1/src/heap.h
File src/heap.h (right):

https://codereview.chromium.org/11365146/diff/1/src/heap.h#newcode2617
src/heap.h:2617: int died_nodes_in_new_space_;
On 2012/12/05 14:26:14, Michael Starzinger wrote:
> Can we rename this to "nodes_died_in_new_space_" so that it is using a common
> prefix for all the global-handle related counters.

Done.

https://codereview.chromium.org/11365146/diff/1/src/heap.h#newcode2620
src/heap.h:2620: int copied_nodes_in_new_space_;
On 2012/12/05 14:26:14, Michael Starzinger wrote:
> Likewise with "nodes_copied_in_new_space_".

Done.

https://codereview.chromium.org/11365146/diff/1/src/heap.h#newcode2623
src/heap.h:2623: int promoted_nodes_;
On 2012/12/05 14:26:14, Michael Starzinger wrote:
> Likewise with "nodes_promoted_"

Done.

[Michael Starzinger, 2012-12-07 09:07:58]

LGTM. I'll land this for you.