Oh.. by previous discussion I thought it was exactly the other way around,
that -profiling disabled optimizations that -g2 had. I've always had the
assumption that -g2 had full optimizations enabled, but apparently not.
Also, I see that using OUTLINING_LIMIT disables the simplifyIfs
optimization so it can affect performance as well.



2014-07-31 1:00 GMT+03:00 Alon Zakai <[email protected]>:

> -g2 (and -g1, anything but -g0) disables simplifyIfs, because it can
> generate lines thousands of characters long, due to commaification ( x = y;
> a = b; turns into  x = y, a = b; ). -profiling leaves simplifyIfs on.
> Basically -profiling is what you want when you are profiling (at least how
> I see things): full performance, just leaving function names intact (and
> basic indentation).
>
> - Alon
>
>
>
> On Wed, Jul 30, 2014 at 2:55 PM, Jukka Jylänki <[email protected]> wrote:
>
>> Can you remind the exact changes that -profiling does? There was some
>> optimization that -profiling didn't do, but -g2 should not affect any
>> optimizations? I recall you said that reading through the output code files
>> with -profiling was slightly easier than with -g2, but wouldn't -g2 be
>> better for users that only cared about Firefox/Chrome profiler results and
>> only need function names intact?
>>
>>
>> 2014-07-31 0:50 GMT+03:00 Alon Zakai <[email protected]>:
>>
>> -profiling is slightly better than -g2 here. -profiling is specifically
>>> intended for the use case of profiling the generated code.
>>>
>>> - Alon
>>>
>>>
>>>
>>>  On Wed, Jul 30, 2014 at 1:18 PM, Jukka Jylänki <[email protected]>
>>> wrote:
>>>
>>>>  Hi!
>>>>
>>>> Worth pinging here in a very recent and awesome advance in Firefox:
>>>> Finally as of last night, Firefox Nightly now supports profiling asm.js
>>>> execution. This means that when you use the CPU profiler after updating to
>>>> latest Nightly, you no longer need to disable asm.js before profiling in
>>>> order to get meaningful stack traces from asm.js functions!
>>>>
>>>> Remember to build all your profiling builds with the proper
>>>> optimization flags (-O2/-O3/-Os/-Oz) at both compile and link time, and add
>>>> in the linker flag -g2 in order to preserve the function names unminified.
>>>> These will ensure that you have the fastest output that retains profiling
>>>> info, at the expense of output size.
>>>>
>>>> For best profiling UI, make sure you are running the geckoprofiler.xpi
>>>> Firefox add-on, documented here:
>>>> https://developer.mozilla.org/en-US/docs/Mozilla/Performance/Profiling_with_the_Built-in_Profiler
>>>>
>>>> Thanks go to Luke Wagner for the hard work of implementing the asm.js
>>>> profiling support in to Firefox!
>>>>
>>>> Jukka
>>>>
>>>>  --
>>>> You received this message because you are subscribed to the Google
>>>> Groups "emscripten-discuss" 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.
>>>>
>>>
>>>  --
>>> You received this message because you are subscribed to the Google
>>> Groups "emscripten-discuss" 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.
>>>
>>
>>  --
>> You received this message because you are subscribed to the Google Groups
>> "emscripten-discuss" 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.
>>
>
>  --
> You received this message because you are subscribed to the Google Groups
> "emscripten-discuss" 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.
>

-- 
You received this message because you are subscribed to the Google Groups 
"emscripten-discuss" 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