On Mon, Jan 4, 2016 at 11:13 AM, Matt Esch <[email protected]> wrote:
> When I say CPU profiler I mean the programmatic API.
>
> I'm familiar with most of the tools currently available for generating flame
> graphs and have even written one myself based on ptrace for linux, but
> nothing is really good enough. I am looking for something that can be
> switched on and off on a production system without needing to start the
> process with special flags.
>
> The cpu profiler supplied by v8-profiler.h generates a cpu profile artifact.
> To generate the cpu statistics v8 backtraces on an interval and then
> generate statistics from the backtraces. I would like access to the raw data
> that's driving the cpu profiler. The raw set of backtraces with resolved js
> symbols is enough to generate a flame graph and do other interesting things.
> The other symbols for the native code can be resolved in the standard way
> after the profile is generated.

The raw data isn't currently exposed.  You'd have to patch V8; maybe
you can get a patch accepted upstream.

FWIW, in StrongLoop's strong-agent product, we basically ended up
writing a C++-and-JS profiler from scratch based on V8's JITCodeEvent
API.  That was in part because we needed to be compatible with old
versions of node.js, it was easier to write a new profiler than
retrofit the old software with the requisite changes.

-- 
-- 
v8-users mailing list
[email protected]
http://groups.google.com/group/v8-users
--- 
You received this message because you are subscribed to the Google Groups 
"v8-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/d/optout.

Reply via email to