Make new space allocation throughput estimation more accurate.

Make new space allocation throughput estimation more accurate.

Sample new space allocation throughput at scavenge and at idle notificatioon.

This will allow better estimation of mutator idleness for switching between
latency and memory modes in idle notification handler.

BUG=chromium:486005
LOG=NO
TEST=cctest/test-heap/NewSpaceAllocationThroughput

Committed: https://crrev.com/57ce9724be99df36011b3191b1372c9d1007dfba
Cr-Commit-Position: refs/heads/master@{#28537}

[ulan, 2015-05-12 14:28:51]

ulan@chromium.org changed reviewers:
+ hpayer@chromium.org

[ulan, 2015-05-12 14:28:52]

PTAL. I will hook this up with mutator idleness detection in the next CL to fix
486005

[Hannes Payer (out of office), 2015-05-12 17:04:17]

lgtm

https://codereview.chromium.org/1125193005/diff/60001/src/heap/gc-tracer.h
File src/heap/gc-tracer.h (right):

https://codereview.chromium.org/1125193005/diff/60001/src/heap/gc-tracer.h#ne...
src/heap/gc-tracer.h:282: typedef RingBuffer<AllocationEvent,
kRingBufferMaxSize> AllocationEventBuffer;
I am wondering about the size of the ring buffer. You get now finer grained
samples. I think the buffer should be bigger for that to average out spikes.

https://codereview.chromium.org/1125193005/diff/60001/src/heap/gc-tracer.h#ne...
src/heap/gc-tracer.h:488: double new_space_allocation_time_ms_;
Please describe the added fields.

https://codereview.chromium.org/1125193005/diff/60001/src/heap/heap.h
File src/heap/heap.h (right):

https://codereview.chromium.org/1125193005/diff/60001/src/heap/heap.h#newcode...
src/heap/heap.h:2169: size_t crankshaft_codegen_bytes_generated_;
newline

https://codereview.chromium.org/1125193005/diff/60001/src/heap/spaces.h
File src/heap/spaces.h (right):

https://codereview.chromium.org/1125193005/diff/60001/src/heap/spaces.h#newco...
src/heap/spaces.h:2439: return static_cast<int>(addr - bottom()) /
Page::kPageSize;
cast to intptr_t

[ulan, 2015-05-18 12:45:51]

https://codereview.chromium.org/1125193005/diff/60001/src/heap/gc-tracer.h
File src/heap/gc-tracer.h (right):

https://codereview.chromium.org/1125193005/diff/60001/src/heap/gc-tracer.h#ne...
src/heap/gc-tracer.h:282: typedef RingBuffer<AllocationEvent,
kRingBufferMaxSize> AllocationEventBuffer;
On 2015/05/12 17:04:17, Hannes Payer wrote:
> I am wondering about the size of the ring buffer. You get now finer grained
> samples. I think the buffer should be bigger for that to average out spikes. 

Adding a new allocation event for each idle notification made the result more
noisy. I reworked this part:
- Add allocation events to the buffer only after GC (as before).
- Separately accumulate (duration_ms, allocated_bytes) since the last GC (see
new_space_allocation_duration_since_gc_, new_space_allocation_in_bytes_since_gc_
in gc-tracer.h)

https://codereview.chromium.org/1125193005/diff/60001/src/heap/gc-tracer.h#ne...
src/heap/gc-tracer.h:488: double new_space_allocation_time_ms_;
On 2015/05/12 17:04:17, Hannes Payer wrote:
> Please describe the added fields.

Done.

https://codereview.chromium.org/1125193005/diff/60001/src/heap/heap.h
File src/heap/heap.h (right):

https://codereview.chromium.org/1125193005/diff/60001/src/heap/heap.h#newcode...
src/heap/heap.h:2169: size_t crankshaft_codegen_bytes_generated_;
On 2015/05/12 17:04:17, Hannes Payer wrote:
> newline

Done.

https://codereview.chromium.org/1125193005/diff/60001/src/heap/spaces.h
File src/heap/spaces.h (right):

https://codereview.chromium.org/1125193005/diff/60001/src/heap/spaces.h#newco...
src/heap/spaces.h:2439: return static_cast<int>(addr - bottom()) /
Page::kPageSize;
On 2015/05/12 17:04:17, Hannes Payer wrote:
> cast to intptr_t

Done.

[ulan, 2015-05-21 09:14:33]

The CQ bit was checked by ulan@chromium.org

[ulan, 2015-05-21 09:14:33]

The patchset sent to the CQ was uploaded after l-g-t-m from hpayer@chromium.org
Link to the patchset: https://codereview.chromium.org/1125193005/#ps140001
(title: "Rebase")

[commit-bot: I haz the power, 2015-05-21 09:14:39]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1125193005/140001

[commit-bot: I haz the power, 2015-05-21 09:29:25]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2015-05-21 09:29:26]

Try jobs failed on following builders:
  v8_linux_gcc_compile_rel on tryserver.v8 (JOB_FAILED,
http://build.chromium.org/p/tryserver.v8/builders/v8_linux_gcc_compile_rel/bu...)

[ulan, 2015-05-21 09:31:33]

The CQ bit was checked by ulan@chromium.org

[ulan, 2015-05-21 09:31:34]

The patchset sent to the CQ was uploaded after l-g-t-m from hpayer@chromium.org
Link to the patchset: https://codereview.chromium.org/1125193005/#ps160001
(title: "Fix test")

[commit-bot: I haz the power, 2015-05-21 09:31:43]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1125193005/160001

[commit-bot: I haz the power, 2015-05-21 10:19:17]

Committed patchset #9 (id:160001)

[commit-bot: I haz the power, 2015-05-21 10:19:30]

Patchset 9 (id:??) landed as
https://crrev.com/57ce9724be99df36011b3191b1372c9d1007dfba
Cr-Commit-Position: refs/heads/master@{#28537}