Track size of incremental marking steps in GCTracer.

Track size of incremental marking steps in GCTracer.

- Track size of bytes marked incrementally
- Add method to compute marking speed in bytes/millisecond
- Distinguish between cumulative and differential marking stats.
- Fix methods that compute incremental marking stats.

R=hpayer@chromium.org
BUG=

Committed: https://code.google.com/p/v8/source/detail?r=22649

[ernstm, 2014-07-28 12:43:58]

PTAL

[Hannes Payer (out of office), 2014-07-28 13:04:13]

lgtm, just nits

https://codereview.chromium.org/420213002/diff/1/src/gc-tracer.cc
File src/gc-tracer.cc (right):

https://codereview.chromium.org/420213002/diff/1/src/gc-tracer.cc#newcode277
src/gc-tracer.cc:277: PrintF("stepscount=%d ",
current_.incremental_marking_steps);
Since you are here, can we make the names here uniform:
steps_count
steps_took

same in the else case

https://codereview.chromium.org/420213002/diff/1/src/gc-tracer.cc#newcode282
src/gc-tracer.cc:282: PrintF("longeststep=%.1f ",
current_.longest_incremental_marking_step);
Let's print the MarkingSpeedInBytesPerMillisecond here in the else case.
PrintF("marking_throughput=%d\n", MarkingSpeedInBytesPerMillisecond());

https://codereview.chromium.org/420213002/diff/1/src/gc-tracer.cc#newcode320
src/gc-tracer.cc:320: // We haven't complete an entire round of incremental
marking, yet.
completed

https://codereview.chromium.org/420213002/diff/1/src/gc-tracer.cc#newcode343
src/gc-tracer.cc:343: // We haven't complete an entire round of incremental
marking, yet.
completed

https://codereview.chromium.org/420213002/diff/1/src/gc-tracer.h
File src/gc-tracer.h (right):

https://codereview.chromium.org/420213002/diff/1/src/gc-tracer.h#newcode255
src/gc-tracer.h:255: double MarkingSpeedInBytesPerMillisecond() const;
No need for double here, lets cast it to e.g. intptr_t. I assume that the result
will always be a large number.

https://codereview.chromium.org/420213002/diff/1/src/incremental-marking.cc
File src/incremental-marking.cc (right):

https://codereview.chromium.org/420213002/diff/1/src/incremental-marking.cc#n...
src/incremental-marking.cc:960: heap_->AddMarkingTime(duration);
Can we remove the AddMarkingTime call? In a separate CL?

[ernstm, 2014-07-28 14:43:26]

https://codereview.chromium.org/420213002/diff/1/src/gc-tracer.cc
File src/gc-tracer.cc (right):

https://codereview.chromium.org/420213002/diff/1/src/gc-tracer.cc#newcode277
src/gc-tracer.cc:277: PrintF("stepscount=%d ",
current_.incremental_marking_steps);
On 2014/07/28 13:04:12, Hannes Payer wrote:
> Since you are here, can we make the names here uniform:
> steps_count
> steps_took
> 
> same in the else case

Done.

https://codereview.chromium.org/420213002/diff/1/src/gc-tracer.cc#newcode282
src/gc-tracer.cc:282: PrintF("longeststep=%.1f ",
current_.longest_incremental_marking_step);
On 2014/07/28 13:04:12, Hannes Payer wrote:
> Let's print the MarkingSpeedInBytesPerMillisecond here in the else case.
> PrintF("marking_throughput=%d\n", MarkingSpeedInBytesPerMillisecond());

Done.

https://codereview.chromium.org/420213002/diff/1/src/gc-tracer.cc#newcode320
src/gc-tracer.cc:320: // We haven't complete an entire round of incremental
marking, yet.
On 2014/07/28 13:04:12, Hannes Payer wrote:
> completed

Done.

https://codereview.chromium.org/420213002/diff/1/src/gc-tracer.cc#newcode343
src/gc-tracer.cc:343: // We haven't complete an entire round of incremental
marking, yet.
On 2014/07/28 13:04:13, Hannes Payer wrote:
> completed

Done.

https://codereview.chromium.org/420213002/diff/1/src/gc-tracer.h
File src/gc-tracer.h (right):

https://codereview.chromium.org/420213002/diff/1/src/gc-tracer.h#newcode255
src/gc-tracer.h:255: double MarkingSpeedInBytesPerMillisecond() const;
On 2014/07/28 13:04:13, Hannes Payer wrote:
> No need for double here, lets cast it to e.g. intptr_t. I assume that the
result
> will always be a large number.

Done.

https://codereview.chromium.org/420213002/diff/1/src/incremental-marking.cc
File src/incremental-marking.cc (right):

https://codereview.chromium.org/420213002/diff/1/src/incremental-marking.cc#n...
src/incremental-marking.cc:960: heap_->AddMarkingTime(duration);
On 2014/07/28 13:04:13, Hannes Payer wrote:
> Can we remove the AddMarkingTime call? In a separate CL?

I'd prefer to remove this in a separate CL.

[ernstm, 2014-07-28 14:49:01]

Committed patchset #2 manually as r22649 (presubmit successful).