C++ profiles processor: align browser mode with the old implementation, sample VM state.

C++ profiles processor: align browser mode with the old implementation, sample VM state.

In browser (DevTools) mode, only non-native JS code and callbacks are reported.
Also, added "(garbage collector)" entry which accumulates samples count in GC state.
Trying to display "(compiler)" and "(external)" only brings confusion,
because it ends up in displaying scripts code under "(compiler)" node, and DOM
event handlers under "(external)" node, which looks weird.

Committed: http://code.google.com/p/v8/source/detail?r=4357

[mnaganov (inactive), 2010-04-07 10:09:50]

Sending message manually because upload had finished with 500 error.

[Søren Thygesen Gjesse, 2010-04-07 11:09:20]

LGTM

Maybe we should consider renaming the flag --prof-browser-mode to something like
--prof-js-only.

http://codereview.chromium.org/1523015/diff/2001/3001
File src/cpu-profiler.cc (right):

http://codereview.chromium.org/1523015/diff/2001/3001#newcode231
src/cpu-profiler.cc:231: bool ProfilerEventsProcessor::IsNameOfNative(String*
name) {
This seems like a bit of a hack. The script object have a type (Script::Type).
Will it be be possible to push that through?

[mnaganov (inactive), 2010-04-07 11:29:15]

On 2010/04/07 11:09:20, Søren Gjesse wrote:
> LGTM
> 
> Maybe we should consider renaming the flag --prof-browser-mode to something
like
> --prof-js-only.
> 

We actually can have 3 orthogonal choices:
 - whether to leave only user JS code;
 - whether we need to add "(program)";
 - should we add shared libraries code entries to code map;

The "browser mode" should set these to "true", "true", "false". It looks like
for a VM developer the values should be opposite. Do you see any need for an
intermediate state? If not, then a single "browser mode" flag is enough.

> http://codereview.chromium.org/1523015/diff/2001/3001
> File src/cpu-profiler.cc (right):
> 
> http://codereview.chromium.org/1523015/diff/2001/3001#newcode231
> src/cpu-profiler.cc:231: bool ProfilerEventsProcessor::IsNameOfNative(String*
> name) {
> This seems like a bit of a hack. The script object have a type (Script::Type).
> Will it be be possible to push that through?

Great hint, I didn't know that, thanks! Yes, I'll change code to use that field
instead of matching against script name.

[Søren Thygesen Gjesse, 2010-04-07 11:55:22]

On 2010/04/07 11:29:15, Michail Naganov wrote:
> On 2010/04/07 11:09:20, Søren Gjesse wrote:
> > LGTM
> > 
> > Maybe we should consider renaming the flag --prof-browser-mode to something
> like
> > --prof-js-only.
> > 
> 
> We actually can have 3 orthogonal choices:
>  - whether to leave only user JS code;
>  - whether we need to add "(program)";
>  - should we add shared libraries code entries to code map;
> 
> The "browser mode" should set these to "true", "true", "false". It looks like
> for a VM developer the values should be opposite. Do you see any need for an
> intermediate state? If not, then a single "browser mode" flag is enough.

I do think that the one flag should be sufficient. It was mainly the name that I
was wondering about, as that mode might not be a browser only thing, but also
useful when V8 is used in other places e.g. node.js.

> 
> > http://codereview.chromium.org/1523015/diff/2001/3001
> > File src/cpu-profiler.cc (right):
> > 
> > http://codereview.chromium.org/1523015/diff/2001/3001#newcode231
> > src/cpu-profiler.cc:231: bool
ProfilerEventsProcessor::IsNameOfNative(String*
> > name) {
> > This seems like a bit of a hack. The script object have a type
(Script::Type).
> > Will it be be possible to push that through?
> 
> Great hint, I didn't know that, thanks! Yes, I'll change code to use that
field
> instead of matching against script name.