Precise GC time measurements.

Precise GC time measurements.


BUG=

Committed: https://code.google.com/p/v8/source/detail?r=13689

[Michael Starzinger, 2013-02-11 15:56:31]

High level suggestion: I think it would make sense to print one or two decimal
places for all of the timings in --trace-gc-nvp which would increase precision
of our automated measurements, but I am not sure if all scripts that parse that
output can cope with such a change ATM.

On the other hand I am not sure if --trace-gc really needs that high of a
precision, since it is mainly intended to be read by humans and there the
millisecond granularity is good enough. WDYT?

https://codereview.chromium.org/12209090/diff/1/src/heap.h
File src/heap.h (right):

https://codereview.chromium.org/12209090/diff/1/src/heap.h#newcode2187
src/heap.h:2187: float max_gc_pause_;
IMHO, this should actually use double precision as all the other floating-point
millisecond-based timing counters use that as well.

[Michael Starzinger, 2013-02-13 12:16:10]

As discussed offline: Ulan verified that the scripts can handle decimal places
for --trace-gc-nvp, so we should print one or two decimal places for all of the
timings there. As for --trace-gc, I don't actually care that much whether we
have decimal places for the Scavenger or not, so I am fine with the change
there.

[Michael Starzinger, 2013-02-18 10:19:46]

https://codereview.chromium.org/12209090/diff/10001/src/heap.cc
File src/heap.cc (right):

https://codereview.chromium.org/12209090/diff/10001/src/heap.cc#newcode7099
src/heap.cc:7099: PrintF("mutator=%d ", static_cast<int>(spent_in_mutator_));
This should also get one decimal place printed.

https://codereview.chromium.org/12209090/diff/10001/src/heap.cc#newcode7113
src/heap.cc:7113: PrintF("external=%d ",
static_cast<int>(scopes_[Scope::EXTERNAL]));
Likewise for the scoped ones.

[Hannes Payer (out of office), 2013-02-18 12:22:24]

https://codereview.chromium.org/12209090/diff/10001/src/heap.cc
File src/heap.cc (right):

https://codereview.chromium.org/12209090/diff/10001/src/heap.cc#newcode7099
src/heap.cc:7099: PrintF("mutator=%d ", static_cast<int>(spent_in_mutator_));
On 2013/02/18 10:19:47, Michael Starzinger wrote:
> This should also get one decimal place printed.

Done.

https://codereview.chromium.org/12209090/diff/10001/src/heap.cc#newcode7113
src/heap.cc:7113: PrintF("external=%d ",
static_cast<int>(scopes_[Scope::EXTERNAL]));
On 2013/02/18 10:19:47, Michael Starzinger wrote:
> Likewise for the scoped ones.

Done.

[Michael Starzinger, 2013-02-19 11:42:11]

LGTM, just one more comment.

https://codereview.chromium.org/12209090/diff/13001/src/heap.cc
File src/heap.cc (right):

https://codereview.chromium.org/12209090/diff/13001/src/heap.cc#newcode7141
src/heap.cc:7141: PrintF("stepstook=%.f ", steps_took_since_last_gc_);
I think s/.f/.1f/ here, otherwise it will still not print any decimal places.
AFAIK, ".f" is the same as ".0f"

https://codereview.chromium.org/12209090/diff/13001/src/heap.cc#newcode7144
src/heap.cc:7144: PrintF("stepstook=%.f ", steps_took_);
Likewise.

[Hannes Payer (out of office), 2013-02-19 11:56:02]

https://codereview.chromium.org/12209090/diff/13001/src/heap.cc
File src/heap.cc (right):

https://codereview.chromium.org/12209090/diff/13001/src/heap.cc#newcode7141
src/heap.cc:7141: PrintF("stepstook=%.f ", steps_took_since_last_gc_);
On 2013/02/19 11:42:11, Michael Starzinger wrote:
> I think s/.f/.1f/ here, otherwise it will still not print any decimal places.
> AFAIK, ".f" is the same as ".0f"

Done.

https://codereview.chromium.org/12209090/diff/13001/src/heap.cc#newcode7144
src/heap.cc:7144: PrintF("stepstook=%.f ", steps_took_);
On 2013/02/19 11:42:11, Michael Starzinger wrote:
> Likewise.

Done.